University of Southern California Dissertations and Theses

Biological and chemical detection using optical resonant cavities

(USC Thesis Other)

Biological and chemical detection using optical resonant cavities

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

BIOLOGICAL AND CHEMICAL DETECTION USING OPTICAL RESONANT
CAVITIES

by
Maria Vladimirovna Chistiakova

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)

December 2014

Copyright 2014 Maria Vladimirovna Chistiakova
ii

Acknowledgements
There are many people I would like to thank that have helped me over the last five years
of my PhD. But first and foremost, I would like to say the biggest thanks to my advisor, Andrea
Armani. Thank you for finding my application worth interest. Thank you for giving me a call to
ask if I would join your lab group. Without that, I would not have even known that I was
accepted to USC for the PhD program. Thank you for believing me to be a good candidate for
your (almost) brand new lab. I feel honored to have been a part of the beginning and to watch the
group grow. Thank you for teaching me how to be a researcher and how to troubleshoot
experiments to determine just which part is not working correctly. Thanks also for all the
valuable skills you have taught me that I will continue to use throughout my career. Skills such
as giving presentations to in-field and out of field audience, reviewing manuscripts for scientific
errors and validity, and how to work on a project independently and more importantly, with a
group. I am glad I got to be in a group with one of the leaders of the field! And also thank you
for all the extra things you do for the lab, such as the lab parties and the lab Olympics, the group
lunches and all the other outings, the envelopes and the rewards inside them, the group shirts and
the group coffee mugs. I will always have the souvenirs to remind me of my time here in the
group.
I would also like to thank all the past and present members of the Armani group. It has
been a great environment to work in with such pleasant colleagues. Thank you to Professor
Heather Hunt for getting us organized, helping to build the labs, and teaching us how to maintain
our notebooks. Thank you to Dr. Jason Gamba for always having helpful suggestions on any
issue at hand. Thank you to Dr. Rasheeda Hawk for teaching me surface functionalization and
for trusting me as a collaborator. Thank you to Dr. Cecilia Lopez for always being friendly and
for giving great advice on how to tackle many situations. Thank you to Dr. Ashley Maker for her
iii

encouragements and help with this thesis and the projects within it. Thank you to Dr. Ce Shi for
being a great office mate and an awesome partner to complete double-blind experiments with.
Thank you to Dr. Hong Seok Choi for welcoming me to the group and for being a mentor and for
introducing me to Korean BBQ. Thank you to Dr. Xiaomin Zhang for always being so nice and
letting me watch you test when I first joined the group. A very big thank you to Matthew
Reddick for all the LabView programs you have written for our lab and all the ones you fixed.
They were invaluable in my data taking and analyzing. Thank you to the undergrads of the lab
that didn't let us forget what it was like being an undergrad ourselves a few short years before:
Leah, Jason, Alexei, Lili, Sam K-L, Grace, Tara, Garrison, Emma, Max, Andre, Yuting, Audrey,
Brian, Sara, Nic, Christine, Daniel, Yingchao, Bradley, Shehzad, Mansi, Chai, Dania, Carol, and
Lindsay. Thank you to our visiting scholars Tobi, Abian, and Imran. You guys came to our lab
and fit right in and so you are missed.
To all my fellow grad students, I would like to thank you all for your collaborations,
advice, opinions, and help. Giant thanks goes to Soheil Soltani for always helping with any
problems, for always knowing the theory of resonators and beyond, for giving advice on all
kinds of issues, and for helping fix setup two multiple times. Thank you to Victoria Sun for
being very supportive and encouraging no matter what. Thank you to Simin Mehrabani for being
an amazing office mate and for being a responsible lab leader. Thank you to Mark Harrison for
answering any random questions I had and for being willing to train me in the clean room even
though I never made it there. Thank you to Kelvin Kuo for collaborating with me, for showing
me how to make sol-gels, and for being a good office mate even though it was short-term. Thank
you to Eda Gungor for being a great post-doc and leading the group well. Thank you to Vinh
Diep for taking over the carbon nanotube project and for letting me try the home-made beer and
iv

cider. Thank you to Sahar Elyahoodayan for tea time (!) and the chipses. Thank you to Michele
Lee for being friendly and always recommending new places to eat. Thank you to Sam
McBirney for taking over setup 2 and I wish you luck with it in the future. Thank you also to the
new members of the group that will continue to grow our research and take care of our lab:
Akshay, Erick, Tushar, Xiaoqin, Amanda, and Alexa.
I would next like to thank my family for supporting me through my life. Thanks
especially to my mom, who has always gently pushed me to always challenge myself and to not
settle for what everyone else was doing. I am grateful for your own ambitions and
accomplishments at becoming an engineer. Because of your success, I never questioned whether
it would be possible for me to achieve the same. Thank you also to my dad for supporting me
and sending me gifts. And thank you to the rest of my family in Russia for being a part of my life
and for always welcoming me home. I would also like to thank Hamed Barghi for being with me
through the good times and the bad and always being my biggest supporter.

v

Abstract
The ability to detect specific molecules in solution is critical for a variety of applications,
including analytical and medical diagnostic measurements. Optical microcavities have become a
popular platform for the detection of minute quantities of material adsorbing to their surfaces.
These devices trap light by guiding it in circular resonant paths near the periphery of the cavity.
The wavelengths required to excite these modes are affected by the surrounding environment,
and can be used to measure changes at the surface of the resonator including the adsorption of
material. Since optical microcavity biosensors were first demonstrated, a great number of
questions have come to light regarding how these devices are used, how their data is interpreted,
and how they may be improved. Here I describe my preliminary work addressing some of these
questions.
Specifically, I have explored mass transport considerations suggesting that the sensor
geometry can play a significant role in the transient response of the sensor. Through this work I
have helped identify flow-induced artifacts unique to this type of sensor and presented methods
to correct for them. Additionally, I have demonstrated how undoped optical microcavities can be
used as narrow-linewidth lasers. This work will enable an important paradigm shift in optical
microcavity biosensing, improving accuracy and reliability of the measurement by monitoring
changes to the lasing wavelength that occur upon adsorption rather than changes in the resonant
wavelength.
I describe my progress and future plans developing novel applications for optical
microcavities in sensing and imaging. I have shown gas differentiation capabilities of silica
microspheres coated with carbon nanotube clusters. The application could greatly increase safety
of gas exposure and enable surface science studies. I have also contributed to a project focused
on measuring the hybridization kinetics of DNA molecules, which could help with disease
vi

prevention or forensics. The microcavity can detect the vibrations resulting from ultrasound
pulses, which change depending on what material the US travels through. I discuss how the
image can be reconstructed from the oscillatory transmission changes detected by the
microsphere. Finally, I discuss future work to be performed on these projects and possible
improvements in experimental design.

vii

Table of Contents

Acknowledgements ......................................................................................................................... ii
List of Figures ............................................................................................................................... xii
List of Tables ............................................................................................................................... xxi
Abstract....... ....................................................................... .............................................................v
Chapter 1 Introduction and Thesis Chapter Overview .................................................................. 1
1.1 Motivation ...................................................................................................................... 1
1.2 Thesis Chapter Overview ............................................................................................... 2
References ................................................................................................................................. 4
Chapter 2 Background ................................................................................................................... 6
2.1 Whispering-Gallery Mode Optical Resonators .............................................................. 6
2.1.1 Total Internal Reflection ....................................................................................... 8
2.1.2 Resonant Wavelength ............................................................................................ 9
2.1.3 Evanescent Field .................................................................................................. 10
2.1.4 Quality factor ....................................................................................................... 13
2.1.5 Calculating Quality Factor .................................................................................. 14
2.1.6 Free Spectral Range ............................................................................................. 16
2.1.7 Intrinsic Quality Factor and its Loss Mechanisms .............................................. 16
2.1.8 Finesse ................................................................................................................. 19
2.1.9 Coupling .............................................................................................................. 19
2.1.10 Polarization and Mode Numbers ......................................................................... 22
2.2 Testing Procedures ....................................................................................................... 23
2.2.1 Pulling a Taper .................................................................................................... 23
2.2.2 Fabricating Silica Microspheres .......................................................................... 24
2.2.3 Finding Resonance Wavelengths ........................................................................ 25
2.2.4 Measuring the Quality Factor: Broad Scan and Fine Scan ................................. 26
2.2.5 Finding Intrinsic Quality Factor .......................................................................... 29
2.2.6 Resonance Shift Measurements ........................................................................... 30
2.2.7 Temperature Stage Measurements ...................................................................... 31
2.3 Lasing Phenomena in Microresonators ........................................................................ 32
viii

2.3.1 Overview of Lasing in Microresonators .............................................................. 32
2.3.2 Purcell Factor ...................................................................................................... 33
2.3.3 Mode Volume ...................................................................................................... 34
2.3.4 Circulating Power ................................................................................................ 35
2.3.5 Threshold Power and Efficiency of Laser ........................................................... 35
2.3.6 Lasing Measurements .......................................................................................... 36
2.3.7 Raman Lasing ...................................................................................................... 37
2.4 Microresonators as Biodetectors .................................................................................. 39
2.5 Ultrasound Imaging ...................................................................................................... 42
References ............................................................................................................................... 45
Chapter 3 Effects of Flow on Whispering Gallery Mode Microresonator Sensors .................... 49
3.1 Overview and Motivation ............................................................................................. 49
3.2 Sample Holder Design for Flow Experiments ............................................................. 49
3.3 Solution Preparation ..................................................................................................... 52
3.4 Initial Experiments ....................................................................................................... 54
3.5 Two-Syringe Method ................................................................................................... 59
3.6 Perturbation Flow Experiments .................................................................................... 61
3.7 Checking the Repeatability of Wavelength Shift ......................................................... 64
3.8 Comparing Initial Slopes of Buffer and Protein Injections .......................................... 66
References ............................................................................................................................... 73
Chapter 4 Raman Microlaser ....................................................................................................... 75
4.1 Overview/Motivation ................................................................................................... 75
4.2 Testing Set-up ............................................................................................................... 76
4.3 Threshold Power for Raman Lasers ............................................................................. 78
4.4 Finite Element Method Mode Volume Simulations ..................................................... 81
4.5 Raman Lasing in Air .................................................................................................... 86
4.6 Threshold Measurement in Air ..................................................................................... 87
4.7 Results in Buffer ........................................................................................................... 94
4.8 Detection Experiments Using Raman Peaks .............................................................. 100
4.9 Raman Peaks Close to Pump ...................................................................................... 100
4.10 Temperature-Dependent Shift of Raman Lasing Peaks in Air ................................... 101
ix

4.11 Temperature-Dependent Shift of Raman Lasing Peaks in Buffer .............................. 106
4.12 Ultraviolet Light Induced Shift of Raman Lasing Peaks ........................................... 110
4.13 Conclusion .................................................................................................................. 113
References ............................................................................................................................. 113
Chapter 5 Determining the Rate of DNA Hybridization ........................................................... 115
5.1 Overview .................................................................................................................... 115
5.2 Background and Motivation ....................................................................................... 115
5.3 Testing Approach ....................................................................................................... 117
5.4 Device Fabrication and Functionalization .................................................................. 119
5.5 Verifying Functionalization on Silica Microtoroids .................................................. 123
5.6 Device Characterization ............................................................................................. 125
5.7 DNA Hybridization Detection .................................................................................... 126
5.8 Sensor Working Range ............................................................................................... 129
5.9 Conclusion .................................................................................................................. 129
References ............................................................................................................................. 130
Chapter 6 Carbon Nanotube Gas Detection .............................................................................. 132
6.1 Overview .................................................................................................................... 132
6.2 Background and Motivation ....................................................................................... 132
6.3 Device Fabrication ..................................................................................................... 135
6.4 Initial Experiments ..................................................................................................... 137
6.5 Verifying the Surface Coverage ................................................................................. 138
6.6 Controlled Gas Adsorption ......................................................................................... 141
6.7 Sensor Characterization .............................................................................................. 142
6.8 Characterization of CNT-Covered Device ................................................................. 145
6.9 Response in Ambient Environment ............................................................................ 148
6.10 Carbon Monoxide Desorption Measurements ............................................................ 149
6.11 Carbon Dioxide Desorption Measurements ............................................................... 150
6.12 Direct Comparison of All Measurements ................................................................... 153
6.13 Step-wise Temperature Response .............................................................................. 154
6.14 Conclusion .................................................................................................................. 156
References ............................................................................................................................. 157
x

Chapter 7 Photoelastic Detection of Ultrasound and Imaging Techniques .............................. 160
7.1 Overview and Motivation ........................................................................................... 160
7.2 Background ................................................................................................................ 160
7.3 Photoelastic Theory .................................................................................................... 162
7.4 Proof of Concept Experiment ..................................................................................... 164
7.5 Optimizing the Experimental Setup ........................................................................... 166
7.6 Control Experiments ................................................................................................... 170
7.7 Optical Simulation Design and Parameters ................................................................ 172
7.8 Acoustic Simulation Design and Parameters ............................................................. 173
7.9 Simulation Results ...................................................................................................... 175
7.10 Converting Simulation Data to Match Experiment .................................................... 177
7.11 Experimental Results and Analysis ............................................................................ 179
7.12 Comparing Simulation and Experimental Data .......................................................... 181
7.13 Ultrasound Imaging Techniques ................................................................................ 182
7.14 Axial Resolution Experiment ..................................................................................... 187
7.15 Conclusion .................................................................................................................. 188
References ............................................................................................................................. 189
Chapter 8 Future Work .............................................................................................................. 191
8.1 Overview .................................................................................................................... 191
8.2 Effects of Flow on Whispering Gallery Mode Microresonator Sensors .................... 191
8.3 Raman Microlaser ...................................................................................................... 192
8.4 DNA Hybridization .................................................................................................... 192
8.5 Carbon Nanotube Gas Detection ................................................................................ 192
8.6 Ultrasound Detection and Imaging ............................................................................. 193
References ............................................................................................................................. 194
Appendix Other Functionalization and Detection Techniques ................................................. 195
A. 1 Lipid Bilayer Project .................................................................................................. 195
A.1.1 Objective ........................................................................................................... 195
A.1.1 Approach ........................................................................................................... 195
A. 2 Biotin Functionalization and Streptavidin Detection ................................................. 198
A.2.1 Objective ........................................................................................................... 198
xi

A.2.2 Functionalization Method .................................................................................. 198
A.2.3 Imaging the Functionalized Surface .................................................................. 200
A.2.4 Detecting Streptavidin ....................................................................................... 201
A. 3 Biodetection: Parameters Affecting the Resonance Shift .......................................... 202
A.3.1 Objective ........................................................................................................... 202
A.3.2 Sensing Experimental Procedure ....................................................................... 203
A.3.3 Results ............................................................................................................... 204
References ............................................................................................................................. 206

xii

List of Figures
Figure 2-1 Renderings of four types of whispering gallery mode microcavities with a waveguide
and the mode highlighted. a) Microring or Flat Microdisk [1-8]. b) Wedged
Microdisk [9, 10]. c) Microtoroid [11-13]. d) Microsphere [14-19]. ........................... 7
Figure 2-2 Snell's law and total internal reflection diagram. .......................................................... 8
Figure 2-3 Schematic of light propagation within a circular microcavity showing total internal
reflection in an a) on-resonance and b) off-resonance condition. ................................. 9
Figure 2-4 Finite element method simulation result showing the distribution of the electric field
in a microsphere cavity. a) A rendering of a microsphere, where the blue square
represents the simulation area. b) Result for device in air. c) Result for device in a
buffer environment. In b) and c), the redder areas indicate high intensity field and
bluer areas show smaller intensity. d) Normalized mode intensity along the red cut
lines in b and c, showing how far the field exists outside the cavity. ......................... 11
Figure 2-5 Methods of calculating Q. a) The linewidth method, where Q is the center wavelength
divided by the FWHM. b) The ring-down method, where the frequency is multiplied
by the photon lifetime. ................................................................................................ 15
Figure 2-6 Representative schematic of different coupling conditions. ....................................... 21
Figure 2-7 Optical fiber holder used for taper pulling. The clamps on top hold the fiber in place.
..................................................................................................................................... 23
Figure 2-8 Tapering of an optical fiber. a) Single mode optical fiber showing the core and
cladding regions. b) Tapered optical fiber with a thinned merged region of core and
cladding. ...................................................................................................................... 24
Figure 2-9 Fabrication of microspheres. Starting with a cladding free optical fiber, the tip is
melted to produce a sphere. ........................................................................................ 25
Figure 2-10 Testing setup. The laser couples light to the microsphere through a tapered fiber,
which is then connected to a photodetector (PD). The output is observed on an
oscilloscope and the laser can be controlled with a function generator. ..................... 26
Figure 2-11 Broad scan. a) A representative schematic of a broad scan showing the free spectral
range of a device with critically coupled peaks. b) A set of data taken on a
microsphere showing an FSR of 0.78 nm. .................................................................. 27
Figure 2-12 Fine scan. a) A theoretical schematic of a quality factor calculation showing the data
as black dots and the Lorentzian fit as a red line. The graph also shows an example of
where the FWHM is measured. b) An example of experimental data showing an
under-coupled resonance peak on a microsphere with a fit Lorentzian function. ...... 29
Figure 2-13 Intrinsic quality factor calculation showing four points of coupling and a linear fit..
..................................................................................................................................... 30
Figure 2-14 View of minimum tracking program. Only the global minimum peak will be tracked.
..................................................................................................................................... 31
Figure 2-15 Temperature stage and controller used for temperature-dependent experiments. .... 32
xiii

Figure 2-16 a) Energy level diagram showing different photon absorption and emission
phenomena [48]. b) Detailed energy level diagram of Raman and Rayleigh scattering.
..................................................................................................................................... 39
Figure 2-17 Resolution in ultrasound imaging. a) Axial resolution describing the distance at
which particles laying parallel to the incoming ultrasound beam can be distinguished
as individual objects. b) Lateral resolution is the distance between two particles that
lie perpendicular to the US beam for them to be distinguished as individual objects.
The lateral resolution changes for the transducer in the focal range. ......................... 43
Figure 3-1 Picture of the steel sample holder with all the parts assembled for testing in water.
The glass slides are used to suspend the microsphere above the steel surface and the
cover slide above the sphere and taper. The injection needle is placed underneath the
sphere and bent to avoid movement. Tubing is connected to the injection tube. The
tape in the diagram wraps around the sample holder but during testing runs, only
enough tape to cover the carbon tape was used. ......................................................... 51
Figure 3-2 Rendering of the liquid chamber setup for microsphere detection testing. The
placement of the taper is shown. The components are not to scale relevant to each
other in the image........................................................................................................ 51
Figure 3-3 Data recorded with minimum tracking program shown in the data cropping program.
The black line is the resonance shift (Lambda axis) and the red line is the
transmission data (Intensity axis). Both are plotted as a function of time. ................. 56
Figure 3-4 Resonance shift data obtained for different flow rates, showing where the flow was
started and where it was shut off. All flow rates are in L/min. ................................ 57
Figure 3-5 Shift in wavelength vs. time for injection of HEPES buffer into HEPES buffer
without any protein. The overall shift in picometers is greater than the shifts seen
when protein was injected in the same time range. Flow rates are in L/min. ........... 58
Figure 3-6 Shift response from water injected into water............................................................. 58
Figure 3-7 Two pump experiment flow layout. One solution was injected at a time, but the
overall flow rate was maintained constant. ................................................................. 60
Figure 3-8 Two pump experiment with buffer and protein showing effects on the resonant shift
when the flow was switched from one to the other. Red arrows indicate when the
protein solution was turned on and blue arrows indicate the buffer solution. ............ 61
Figure 3-9 Two pump perturbation flow layout. Te buffer solution was injected continuously
while the protein solution flow was added as a 1% perturbation. .............................. 63
Figure 3-10 Perturbation flow experiment results. The 1% perturbation of protein flow did not
produce any major changes in the shift diagram. Here, the red arrows refer to addition
of protein, while blue arrows indicate when only buffer was flowing. a) Results for a
buffer flow rate of 1000 L/min where protein was added at 10 L/min. b) Results
for a flow rate of buffer at 250 L/min with an addition of protein at 2.5 L/min. ... 63
Figure 3-11 Injection of buffer or protein, as noted, at 100 L/min. The red lines are fits to the
initial linear part of the resonance shift curves. a) Four injections of buffer followed
xiv

by two protein injections. b) Four protein injections followed by two buffer
injections. .................................................................................................................... 64
Figure 3-12 Initial slopes of the binding curves for buffer and protein injections. ...................... 65
Figure 3-13 a) At lower flow rates (250 L/min), the protein solution sometimes produces a
larger shift and bigger initial slope. b) At higher flow rates (1000 L/min), the buffer
flow always has the bigger effect. The blue and purple lines are the linear fits to the
initial part of the shift. ................................................................................................. 68
Figure 3-14 The difference between buffer and protein initial slopes of the wavelength shift
curve versus flow rate for two sphere sizes. ............................................................... 69
Figure 3-15 Difference between buffer and protein initial slopes for two sizes of toroids. All
sizes are in microns. .................................................................................................... 70
Figure 3-16 Additional data for a higher (1 nM) concentration of protein (blue and red half-
colored star data points). ............................................................................................. 71
Figure 3-17 The data for wavelength shifts resulting from buffer injections at two flow rates. a)
Wavelength shift resulting from 100 L/min flow rate. The three lines fit to the initial
part of the curve show very different slopes. b) Wavelength shift resulting from 500
L/min flow rate. Fewer data points appear at the higher flow rate because the
resonance shifts much faster. The slope still changes if fewer or more points are
added in the fit. ........................................................................................................... 72
Figure 3-18 The data for protein shifts at two flow rates showing the linear fits to the initial
slope. a) Shift for a 100 L/min flow rate of protein solution. b) Shift for a from 500
L/min flow rate. ........................................................................................................ 73
Figure 4-1 Coupling light into a silica microsphere. a) No power coupled into sphere. b) Power
starting to couple into microsphere. c) and d) Maximum power coupled into
microsphere with (c) and without (d) the microscope light illuminating the device....
..................................................................................................................................... 77
Figure 4-2 Schematic of testing setup, showing the tapered optical fiber, laser, photodetector
(PD), and the spectrograph and its fiber, which is positioned near the microsphere
[15]. ............................................................................................................................. 78
Figure 4-3 Absorption coefficient of water at different wavelengths [16]. .................................. 81
Figure 4-4 a) SEM image of a silica microsphere showing a blue square that represents the
simulation area. b) Simulation area of a 200 m diameter silica microsphere and its
surroundings. ............................................................................................................... 82
Figure 4-5 Free triangular mesh used in the simulation geometry. .............................................. 83
Figure 4-6 Simulation results for the first four eigenvalues found by the model. The units of the
mode are V/m
2
. ........................................................................................................... 84
Figure 4-7 a-b). COMSOL simulation showing the cross section of the mode around the equator
of a 200 m diameter microsphere in air and in buffer, respectively. The mode shifts
outward as the device is placed in buffer rather than air. c). Normalized mode field
intensity along the cut line shown in (a) and (b). The vertical blue line represents the
xv

microsphere boundary. The mode shifts closer to the boundary of the device and more
of the field is propagating in the environment as well. The mode size thus should also
increase [15]. ............................................................................................................... 85
Figure 4-8 COMSOL simulation results for effective mode volume as a function of the refractive
index of the environment for spheres of diameters a) 180, b) 190, and c) 200 m. The
mode volume increases as the size of the device is increased and has a significant
dependence on the refractive index around the device as well [15]. ......................... 86
Figure 4-9 Cascaded Raman lasing at three pump wavelengths in air. a) A pump of 633 nm
results in two Raman peaks. b) Pump of 771 nm gives five lasing peaks that
potentially display multi-mode lasing. c) Lasing emission with an 850 nm pump,
clearly showing multi-mode behavior [15]. ............................................................... 87
Figure 4-10 Correlation between power and voltage seen on the o-scope. .................................. 88
Figure 4-11 A broadening resonance peak. The inset shows a zoomed in view with the same axes
labels. .......................................................................................................................... 89
Figure 4-12 a) Spectrograph output with the pump peak and two lasing peaks. b) Intensity vs.
power graph for the first lasing peak. c) Intensity vs. power graph for the second
lasing peak. ................................................................................................................. 90
Figure 4-13 Lasing in air. a) Three emission peaks were detected by the spectrograph at 801,
831, and 864 nm. b-d) Intensity as a function of power going into the sphere for all
three lasing lines. The threshold is found at the x-intercept when fitting a line to the
stimulated emission portion of the graph. b) The first lasing peak shows a threshold of
432 W. The apparent saturation at input powers greater than 0.9 mW is due to the
saturation of the spectrograph. c) The second lasing peak at 831 nm has a threshold of
491 W. d) Third lasing peak shows a threshold of 685 W [15]. ............................ 92
Figure 4-14 Lasing in air. a-d) The lasing peak intensity is plotted as a function of input power
going into the sphere for all four detected emission peaks. The threshold for a)-d) is
157 W, 220 W, 259 W, and 301 W, respectively. Inset: Lasing intensity graph
showing four lasing peaks at 801, 829, 862, and 895 nm [22]. .................................. 93
Figure 4-15 Rendering of testing set-up for buffer experiments, indicating locations of device,
taper, and spectrograph fiber tip. The chamber is filled with HEPES buffer [15]. .... 94
Figure 4-16 Normalized transmission spectra. a) Intrinsic quality factor of the device immersed
in HEPES buffer, Q
0
= 1.57x10
7
. b) Broadened resonance peak of microsphere in
buffer taken during a measurement [15]. .................................................................... 95
Figure 4-17 Lasing in buffer. (a) Emission peak at 806 nm. (b) Emission intensity versus input
power in HEPES buffer. The threshold is 1.94 mW [15]. ......................................... 96
Figure 4-18 Cascaded Raman lasing in HEPES buffer. a) Threshold graph for the first lasing
peak shown in the inset with a threshold value of 0.76 mW. Inset: Two lasing peaks
achieved in HEPES buffer. b) Threshold graph for the second lasing peak showing a
threshold of 0.824 mW [22]. ....................................................................................... 97
xvi

Figure 4-19 Plot of threshold vs. peak number for two different experiments. Squares (circles)
show data for a 189 m (38 m) sphere. Experimental results agree well with the
theoretical fit. Inset: Slope efficiency for each lasing peak in Figure 4-14, showing a
decreasing trend in agreement with theory [22]. ........................................................ 99
Figure 4-20 Raman lasing peaks around the pump wavelength of 777.8 nm seen on the OSA...
................................................................................................................................... 101
Figure 4-21 Raman peaks detected on OSA at three time points. .............................................. 102
Figure 4-22 Changes in Raman lasing peak wavelength as the temperature is changed. a)
Tracking two lasing peaks vs. time as the temperature is changed. b) Zoomed in view
of the lower wavelength peak from a). ..................................................................... 103
Figure 4-23 Data analysis for averaged scans. The scans were averaged and a Gaussian was fit to
the averaged data....................................................................................................... 104
Figure 4-24 Temperature experiments on averaged data sets. a) Resonance shift for increasing
and decreasing temperature. b) Raman lasing peak shift with temperature changes in
both directions. .......................................................................................................... 105
Figure 4-25 Comparing resonance and Raman lasing wavelength shifts vs. temperature. ........ 106
Figure 4-26 Raman lasing peaks on the OSA when testing in HEPEs buffer with a pump
wavelength of 774.7 nm and 780.395 nm. ................................................................ 107
Figure 4-27 Temperature induced resonance and Raman lasing peak shifts. ............................ 108
Figure 4-28 Representative resonance peaks in HEPEs buffer for which Raman lasing was
observed on the OSA. ............................................................................................... 109
Figure 4-29 Shifts in resonance in HEPEs buffer with UV. ....................................................... 111
Figure 4-30 Raman lasing peak shifts for two different lasing peaks. The black squares show
shifts for increasing UV intensity and the red circles for decreasing intensity. ....... 112
Figure 5-1 Silica toroidal optical microcavity. a) Scanning electron micrograph of a toroidal
optical microcavity. b) 2D COMSOL finite element method simulation of the
equatorial cross section of a toroidal optical cavity. The optical field extends into the
environment, enabling excitation of fluorophores located near the surface of the
device [15]. ............................................................................................................... 118
Figure 5-2 Epoxide functionalization process. The surface of the cavity is hydroxylated and then
GPTMS is used to covalently attach epoxide groups. Aminated ssDNA binds to the
epoxides, forming the ssDNA functionalized devices. In the last step, the complement
ssDNA-Cy5 is hybridized to the surface. Cy5 emits in the red. To perform the
control fluorescent imaging experiments, the aminated ssDNA was labeled with 6-
FAM, a green fluorophore [15]. The devices can also be recycled if placed under
oxygen plasma once again, which will remove the functionalized components. ..... 119
Figure 5-3 Bright field and fluorescence images of the microscope showing functionalization has
occurred. a) A bright field image of a functionalized microsphere showing a smooth
surface that indicated that the sphere was not damaged during the process. b) A bright
field image of a control sphere that was not exposed to oxygen plasma also shows a
xvii

smooth surface. c) A fluorescent image of the functionalized sphere shows the Cy5
dye present on the complementary strand of DNA and that it has bound to the surface
of the sphere. d) A fluorescent image of the control sphere showing no fluorescence,
indicating that DNA does not bind to bare silica and that functionalization does not
occur without the oxygen plasma step. ..................................................................... 122
Figure 5-4 Multicolor fluorescent imaging. a) Bright field image of a microtoroid after GPTMS
vapor deposition and incubation with amine modified ss-DNA-6-FAM. b)
Fluorescent image of the same microtoroid showing the attachment of ss-DNA-6-
FAM. c) Fluorescent image of a microtoroid after incubation with complement ss-
DNA-Cy5. The filters are adjusted to isolate the Cy5 emission from the FAM
emission. d) Overlap of parts b and c, verifying that hybridization occurred. The
specific sequences used are: 3’ –NH
2
-GCC GGA TAG CGT AAA GGT TA-FAM and
5’-CGG CCT ACT GCA TTT CCA AT/Cy5-3’ [15]. ............................................. 124
Figure 5-5 Optical device characterization. a) A transmission spectra used to determine the
quality factor of the cavity. Based on the linewidth, this device has a Q of 2.2 x 10
7

in water. b) A rendering of the testing setup, which can simultaneously record the
emission from the fluorescent dye using a fiber coupled spectrograph and inject the
ssDNA-Cy5 using a syringe pump. The emission from the 633nm tunable laser is
partially blocked using a red filter. The quality factor can also be measured in this
configuration [15]. .................................................................................................... 126
Figure 5-6 Detection of 2 M ssDNA-Cy5. a) Emission spectrum with and without the ssDNA-
Cy5 present. While the 633 nm laser line is present in both spectra, the fluorescent
emission is only present when the ssDNA-Cy5 is injected, as expected. There are no
secondary lines or other noise sources present in this wavelength range. As such, the
signal fidelity is extremely high. The arrow indicates the 670 nm wavelength that is
tracked in the detection experiments. b) The maximum of the emission at 670 nm,
indicated in part a, is monitored and recorded while the ssDNA-Cy5 is injected. A
strong but transient signal is generated when the molecule nonspecifically binds to the
surface and/or moves within the evanescent field. The second stable peak is the result
of the hybridization [15]. .......................................................................................... 128
Figure 5-7 The working range of the device as it is sequentially exposed to several different
ssDNA-Cy5 solutions [15]. ....................................................................................... 129
Figure 6-1 Overview of the spherical resonant cavity fabrication process. a) Optical microscope
image of the optical fiber tip after the polymer cladding has been removed and the
end-face has been cleaved. b) Optical microscope image of the fabricated cavity after
the CO
2
reflow process. c) Rendering showing the sphere with its fiber optic stem, the
tapered fiber waveguide, and an excited mode within the resonator [36]. ............... 136
Figure 6-2 Resonance shifts resulting from CNTs adsorbing gas from the air. ......................... 138
Figure 6-3 CNT clusters on silica spheres under the JSM SEM. The CNT structure cannot be
resolved with this instrument. ................................................................................... 139
xviii

Figure 6-4 Scanning electron microscope images of the silica sphere with CNT clusters. a) Image
of the sphere surface. The clusters are barely identifiable. b) Magnification of the
region indicated in part a. c) Magnification of the region indicated in part b [36].
................................................................................................................................. ..141
Figure 6-5 a) Schematic of the testing setup with the major components indicated. PD is the
photodetector, F-Gen is the function generator, and O-scope is the oscilloscope. The
PCI cards are integrated directly into a computer and are controlled using a LabView
interface. b) Schematic of the temperature stage with the silica sphere and fiber
waveguide shown. TC is the thermocouple [36]. ..................................................... 144
Figure 6-6 Representative transmission spectrum of a resonance peak of a) a bare silica resonant
cavity with a Q of 1.85x10
8
and b) a CNT cluster-covered resonant cavity with a Q
of 4.0x10
6
[36]. ......................................................................................................... 146
Figure 6-7 a) As the CNT concentration increases, the quality factor decreases, due to increasing
material loss. A line of the form y=ax
b
is fit to the data with a b value of -1.5 [36]. b)
UV-Vis measurements of different concentrations of CNTs. The response is linear in
the low concentration, but saturates at about 1 mg/mL. ........................................... 147
Figure 6-8 Comparison between bare silica and CNT-covered spheres. a) Resonance shifts for a
step of 0.5
o
C. Inset: Histograms of the noise in the sensing signal generated by the
bare SiO
2
and CNT cluster-covered SiO
2
spheres. b) dn/dT data points for bare and
CNT-covered spheres with the linear fits shown. Temperature is the final
temperature minus the starting temperature (20
o
C) [36]. ......................................... 149
Figure 6-9n vs. T of heating cycles for CO-saturated a) silica and b) CNT cluster cavity
sensors. Linear fits to experimental data are shown [36]. ........................................ 150
Figure 6-10 n vs. T of heating cycles for CO
2
-saturated a) silica and b) CNT cluster cavity
sensors. n vs. T of heating cycles with a high temperature hold for CO
2
-saturated
c) silica and d) CNT cluster cavity sensors. Linear fits to experimental data are
shown [36]. ............................................................................................................... 152
Figure 6-11 Summary of dn/dT vs. cycle number for a) bare silica and b) CNT cluster cavity
sensors. dn/dT values from the control experiments are included as the dashed lines
[36]. ........................................................................................................................... 153
Figure 6-12 Step-wise testing of CNT-covered spheres. a) Response of CNTs saturated with
CO
2
. b) Response of CNT spheres saturated with CO.............................................. 155
Figure 6-13 Zoom-in graphs of the temperature steps showing an overshoot and then a cooling
back down to the set point at the top of the peak. a) Step from 23
o
C to 29
o
C for CNT
sphere saturated with CO. b) Step from 23
o
C to 31
o
C for the same sphere. ............ 156
Figure 7-1 a) The resonant wavelength increases and decreases in response to the ultrasound
pulse. b) If the transmission values at a single wavelength ( 
o
) are plotted, the
characteristic damped oscillator curve is clear. The positive values indicate high
pressure and the negative values indicate low pressure [8]. ..................................... 161
xix

Figure 7-2 Two ultrasound pulses detected by the microsphere. a) The entire optical transmission
spectrum. b) and c) show zoomed in view of the dampened oscillations. ................ 165
Figure 7-3 Top and side view of initial holder setup for imaging in pulse/echo mode. ............. 167
Figure 7-4 Optimized sample holder. a) Image of PDMS holder on a PEEK rod. b) Rendering of
the sensor setup [8]. .................................................................................................. 169
Figure 7-5 a) Schematic of the testing setup. b) Resonance peak used in the experiment. The red
dashed line shows the fit line used to convert simulation data [8]. .......................... 170
Figure 7-6 Response from ultrasound pulse for under-coupled (red), over-coupled (black), and
critically-coupled (blue) regimes. ............................................................................. 171
Figure 7-7 Control testing of ultrasound response. a) Response from microsphere to a pulse from
the transducer when it is placed in front of the sphere. b) Transducer response to an
echo coming from the silica microsphere [8]. .......................................................... 172
Figure 7-8 a) Microscope image of a silica microsphere. b) RF simulation results showing a
small part of the sphere in a water environment with the mode distribution plotted at
V/m
2
[8]. ................................................................................................................... 173
Figure 7-9 a) Geometry of the 2D FEM simulation in COMSOL. b) Ultrasound pulse shape used
in the model [8]. ........................................................................................................ 174
Figure 7-10 Simulation results. a) FEM result showing the initial pulse as it passes by the silica
microsphere. b) FEM result for the echo coming from the steel sphere as it reaches
the silica surface. c) Pressure variations recorded with a 0.225 s pulse at the point
just outside the silica surface boundary in the water. d) Pressure variation within the
silica material showing the effects of the initial pulse and the secondary echo [8]....
................................................................................................................................... 176
Figure 7-11 Experimental data. a) The entire signal received by the silica microsphere. b)
Zoomed-in graph of the steel echo. c) Zoomed-in echo from the water-air interface
[8]. ............................................................................................................................. 180
Figure 7-12 Comparing experimental and simulation results. a) Overlapping the experimental
data (black) with simulation data (red). b) Accuracy as a function of pulse width
specified in the simulation [8]................................................................................... 182
Figure 7-13 Time domain data shown a) before and b) after normalization. ............................. 183
Figure 7-14 Imagesc plots showing the data a) without normalization and b) with normalization.
The distance in the y-axis is round-trip distance from the silica sphere. The x-axis
distance is how far the imaging object was moved from left to right. Note: these
figures do not represent an image, but a series of A-line scans. ............................... 185
Figure 7-15 Imagesc reconstruction of A-line scans from a wire of 270 m diameter. The x-axis
distance is the distance the wire was moved in the up/down direction. ................... 186
Figure 7-16 Schematic showing how the wire (red circle) is 0.4 mm farther from the microsphere
when it is at the top or bottom of the liquid chamber compared to when it is exactly in
the center. .................................................................................................................. 187
xx

Figure 7-17 a) Imagesc representation of A-line scans of a wire. The x-axis is the distance the
wire was moved away from the sphere. b) Agreement of experimentally found
location versus the set location of the wire. The red line is the linear fit to the data
with an R
2
> 0.999. ................................................................................................... 188
Figure 8-1 Side view diagram of the new testing liquid chamber for imaging objects with
ultrasound. ................................................................................................................. 194
Figure A-1-1 Fluorescent microscope images of a lipid bilayer coated silica microsphere with
different focal planes shown. a) The covering seems smooth and even. b) Focusing on
a different part of the sphere, the coating still looks smooth. c) At a third focal plane,
some spots can be seen in the coating. ...................................................................... 197
Figure A-2-1 Fluorescent imaging of microspheres functionalized with biotin. a) A bright field
image of a sphere that was biotinylated but not exposed to the Texas-Red-Avidin
solution. b) Fluorescence image of the control sphere showing no fluorescence. c)
Bright field image of a biotinylated sphere that had been incubated with Texas-Red-
Avidin. The surface is still smooth after all the functionalization steps. d)
Fluorescence imaging shows a uniform coverage of the sphere and step indicating
successful functionalization. ..................................................................................... 201
Figure A-2-2 Resonance shift of a microsphere as streptavidin is injected. After injection is
stopped, the resonance does not return to the original position due to the strong
binding of streptavidin to biotin. ............................................................................... 202
Figure A-3-1 Results of wavelength shift with varying scan range. All legends show the
operating wavelengths in nm. a) Results for a 10 aM concentration of protein. b)
Results for a 1 nM concentration. c) Results with a protein concentration of 1 M..
................................................................................................................................... 205
Figure A-3-2 Results of wavelength shift as a function of power in the resonator (P
in
) multiplied
by the loaded Q (Q
L
). The diameter of the microspheres is also indicated in the
legend. No apparent trend appears. ........................................................................... 206

xxi

List of Tables
Table 4-1 Cascaded Raman lasing results in air and buffer for various size spheres at a pump
wavelength of around 770 nm [15]. ............................................................................ 98
Table 6-1 Evolution of dn/dT values through thermo-cycling of devices [36]. .......................... 154

1

Chapter 1 Introduction and Thesis Chapter Overview
1.1 Motivation
Chemical and biological sensing techniques have been developing and improving for
centuries. Higher sensitivity and selectivity in detectors is desirable. Miniaturization of the
sensor has great implications for a variety of fields, especially medicine and electronics. While
minimizing the footprint of the sensor, its efficiency must be maintained to preserve its
performance.
To achieve the aforementioned goals, optical resonators have been developed over the
years to achieve a better and faster sensing response. Of these resonators, the whispering gallery
mode (WGM) silica resonators show the most promise for detection capabilities. Because these
devices are able to store light through total internal reflection, the power within the resonators
build up enormously and has the potential to act as an excitation source for applications such as
lasing. Additionally, the existence of the evanescent field just outside the periphery of the device
makes it extremely sensitive to the surroundings, detecting infinitesimal changes. Most
importantly, these devices are tiny with a diameter range from 30 to 200 microns.
Specifically, silica microspheres show one of the highest recirculation times out of all the
WGM resonators, giving them ultra high quality factors on the order of 100 million. As such,
the linewidths of its resonance peaks are very narrow and the changes in resonance wavelength
can easily be detected. This is due to the low material loss and low doping of the fused silica
material, as well as the smooth surface of the device from to the reflow fabrication technique.
Because of this high quality, silica microspheres have many applications in biosensing [1-5],
chemical sensing [6-8], laser design [9-12], and communications [13-15].
The selectivity and sometimes the sensitivity of the device can be enhanced with
functionalization. For example, silica microspheres alone cannot detect the presence of CO
2
,
2

however, by coating the device with carbon nanotube clusters, the sphere can now sense the
changes within the nanotubes and can detect the gas [6]. When functionalizing the device, it is
important to choose biocompatible materials when detecting biological compounds or molecules.
Silica is inherently biocompatible and thus makes a great starting platform to work from.
Further background on WGM resonators and specific properties of silica microspheres
are presented first. Then, each project is outlined and described in its own chapter. The effects of
flow experiments on silica devices are explored initially. Then, the lasing applications of bare
silica spheres are shown, including lasing peak tracking for sensing applications. The biological
detection experiments are presented next, with detection of DNA hybridization rates. The
chemical sensing capabilities of silica microspheres are showcased through carbon nanotube
attachment. Another medical application is shown next with ultrasound detection on silica
microspheres. Finally, the future of these projects is discussed and an appendix of other
functionalization and detection methods is presented.
1.2 Thesis Chapter Overview
The outline of this thesis is detailed below.
Chapter 2 details the background information on whispering gallery mode optical
resonators. The basic testing procedures used for most experiments are outlined. Aspects of
lasing phenomena in resonators is discussed and the general theory is summarized. Some basic
information about sensing and ultrasound imaging is also presented.
In Chapter 3, a project that seeks to explain the effects of flow on the detection
mechanism is presented. The design of the sample holder for fluid experiments is described. Two
injection setups, two-syringe method and the perturbation method, are presented and the results
3

from each are discussed. Finally, a slope analysis method is outlined and tested on experimental
data.
Chapter 4 presents data on cascaded Raman lasing in a silica microsphere. For the first
time, cascaded Raman lasing on undoped silica spheres was shown in an aqueous environment
[10]. Due to the high recirculation of light in the cavity, the thresholds for lasing in air and in
water were sub-mW [11]. Finite element method simulations were performed to explain the
increase in threshold for lasing in water. Results for tracking the lasing peak with changes in the
refractive index are also presented.
In Chapter 5, an experiment on determining the hybridization rate of DNA using optical
resonators is discussed [1]. The resonators are functionalized with single-stranded DNA by the
epoxide method. The verification of coverage is presented and explained. The hybridization is
monitored via fluorescence detection by a spectrograph. Two regimes of the hybridization
process are determined. The sensor working range is also analyzed.
Chapter 6 describes the gas adsorption mechanism of carbon nanotube clusters on silica
microspheres. By adding carbon nanotube clusters to the surface of the device and controllably
adsorbing a particular gas to the nanotubes, the desorption parameters of the gas could be
studied. In the experiment, desorption of carbon monoxide and carbon dioxide could be
distinguished at ambient conditions, an improvement over standard thermal desorption
techniques [6].
Chapter 7 outlines an experiment and simulation of ultrasound detection by silica
microspheres [16]. The ultimate goal is to use an optical system for ultrasound imaging. Because
resonators are sensitive to small intensity ultrasound pulses, echo based images can be
4

constructed. The photo-elastic response to ultrasound is shown in a finite element simulation
model and verified experimentally.
Finally, Chapter 8 discusses the potential future work on the preceding projects and how
the experiments can be improved. An appendix summarizing other functionalization and
detection methods is also included. A project on attaching lipid bilayers to silica microspheres
discusses the techniques used to achieve a uniform coverage. An amine group functionalization
method is also discussed, which allows for the attachment of specific biomarkers to the surface
of silica. Additionally, a study of experimental parameter effects on biodetection is provided.
References
1. Hawk, R.M., M.V. Chistiakova, and A.M. Armani, Monitoring DNA hybridization using
optical microcavities. Optics Letters, 2013. 38(22): p. 4690-4693.
2. Freeman, L.M. and A.M. Armani, Photobleaching of Cy5 Conjugated Lipid Bilayers
Determined With Optical Microresonators. Ieee Journal of Selected Topics in Quantum
Electronics, 2012. 18(3): p. 1160-1165.
3. Soteropulos, C.E., H.K. Hunt, and A.M. Armani, Determination of binding kinetics using
whispering gallery mode microcavities. Applied Physics Letters, 2011. 99(10).
4. Freeman, L.M., et al., Excitation of Cy5 in self-assembled lipid bilayers using optical
microresonators. Applied Physics Letters, 2011. 98(14).
5. Carrier, J.R., M. Boissinot, and C.N. Allen, Dielectric resonating microspheres for
biosensing: An optical approach to a biological problem. American Journal of Physics,
2014. 82(5).
6. Chistiakova, M.V. and A.M. Armani, Optical detection of CO and CO
2
temperature
dependent desorption from carbon nanotube clusters. Nanotechnology, 2014.
7. Pal, A., et al., A high-Q low threshold thulium-doped silica microsphere laser in the 2 mu
m wavelength region designed for gas sensing applications. Laser Physics Letters, 2013.
10(8).
8. Matejec, V., et al., Preparation of spherical optical microresonators and their resonance
spectra in air and gaseous acetone. Photonics, Devices, and Systems V, 2011. 8306.
9. Sulaiman, A., S.W. Harun, and H. Ahmad, Erbium-Doped Fiber Laser With a Microfiber
Coupled to Silica Microsphere. Ieee Photonics Journal, 2012. 4(4): p. 1065-1070.
10. Chistiakova, M.V. and A.M. Armani, Cascaded Raman microlaser in air and buffer.
Optics Letters, 2012. 37(19): p. 4068-4070.
11. Chistiakova, M.V. and A.M. Armani, Microcavity-based cascaded Raman microlaser in
air and in buffer. Optical Components and Materials X, 2013. 8621.
12. Ward, J. and O. Benson, WGM microresonators: sensing, lasing and fundamental optics
with microspheres. Laser & Photonics Reviews, 2011. 5(4): p. 553-570.
5

13. Farnesi, D., et al., Optical Frequency Conversion in Silica-Whispering-Gallery-Mode
Microspherical Resonators. Physical Review Letters, 2014. 112(9).
14. Kandas, I., et al., High quality factor silica microspheres functionalized with self-
assembled nanomaterials. Optics Express, 2013. 21(18): p. 20601-20610.
15. Zhao, P., et al., Iron-oxide nanoparticles embedded silica microsphere resonator
exhibiting broadband all-optical wavelength tunability. Optics Letters, 2014. 39(13): p.
3845-3848.
16. Chistiakova, M.V. and A.M. Armani, Photoelastic ultrasound detection using ultra-high-Q
silica optical resonators. Optics Express, 2014.

6

Chapter 2 Background
2.1 Whispering-Gallery Mode Optical Resonators
Optical microcavity devices are able to store light through the use of mirror-like surfaces.
One type of optical cavity is the Fabry-Perot (FP) interferometer. Conventional FP resonators
consist of two or more mirrors that allow the recirculation, or continuous reflection, of light. The
result is a longer effective path length to use in spectroscopic instruments. Longer recirculation is
achieved with more perfect resonators, ones that have reduced losses, and thus better reflections.
To produce low loss resonators, it is possible to use higher reflectivity surfaces to result in
smaller scattering loss. Such optical resonators are major components of lasers, surrounding the
gain medium and providing feedback for the laser light. They can also be used in optical
parametric oscillators and some interferometers.
The disadvantage to using conventional FP cavities is their bulky size, cost, and
instability due to vibrations. To overcome such obstacles, monolithic microresonators have been
produced. Microcavities are small in size (micron-scale), relatively cheap to fabricate, and are
much more resistant to vibrations and other outside disturbances.
A subset of optical microresonators includes the whispering gallery mode (WGM)
resonators, which support a traveling wave optical mode rather than a standing wave. WGM
resonators get their name from an architectural feature, like that found in St. Paul’s C athedral in
London, where sound carries around the periphery of the gallery and can be easily heard at the
perimeter, even at an audible level of a whisper. WGM microcavities have various geometries:
flat microrings and microdisks [1-8], wedged microdisks [9, 10], microtoroids [11-13], and
microspheres [14-19]. These cavities can be fabricated from a variety of materials, such as
silicon, silica, silicon nitride, as well as many others, including doped materials for specific
7

functions. They also have the advantage of being produced from one material, unlike the FP
resonators, which must be built from different parts.
The most common geometries of cavities are shown in Figure 2-1 along with the
references to those who studied them. The figure shows renderings of the microcavities along
with a propagating mode around the equator. In the Armani lab, microtoroids and microspheres
are the most utilized resonator geometries. Most of the research I was involved in has focused on
utilizing the microsphere for its immense sensing capabilities.

Figure 2-1 Renderings of four types of whispering gallery mode microcavities with a waveguide and
the mode highlighted. a) Microring or Flat Microdisk [1-8]. b) Wedged Microdisk [9, 10]. c)
Microtoroid [11-13]. d) Microsphere [14-19].
Optical WGM resonators guide light in a similar fashion to the galleries they are named
after. When light of a particular wavelength enters the device, it stays confined due to total
internal reflection, recirculating around the periphery. When an integral number of wavelengths
fits around the perimeter of the device, the cavity is said to be on resonance. Light can enter the
cavity at other wavelengths as well, but will not build up inside since a non-integral number of
wavelengths fit and thus will be out of phase with other light coming in. Thus the light will
instead be lost to the outside.
8

2.1.1 Total Internal Reflection
The confinement of light inside a cavity is due to the refractive index contrast between
the material of the cavity and its surroundings, which leads to total internal reflection. According
to Snell’ s law, the angle of refraction is related to the refractive indices by:

(2.1)
where 
1
is the angle of incidence, 
2
is the angle of refraction, v
1
and v
2
are phase velocities in
the different media, and n
1
and n
2
are refractive indices in each medium. When light travels from
a medium of higher refractive index, such as water (n = 1.33), into a medium with lower
refractive index, for instance, air (n = 1), there is an angle of incidence above which the sine of
the angle of refraction is forced to be greater than one. This is not possible, and instead, the light
is completely reflected from the interface back into the same medium. Figure 2-2 shows a
diagram with incident and refracted or reflected rays of light. At the critical angle, which can be
found by setting the angle of refraction to 90
o
, the light will propagate along the interface
between the media.

Figure 2-2 Snell's law and total internal reflection diagram.
9

Since microresonators are made of a material with a higher refractive index than air
usually, light inside the cavity is trapped through total internal reflection. A demonstrative
schematic of light propagation inside a WGM resonator is shown in Figure 2-3. The left
schematic shows on-resonance circulation with the round trip in the cavity equal to an integral
number of wavelengths, thus it is able to constructively interfere with the light entering the
device on the next turn, and continually build up in intensity. The right side of Figure 2-3 shows
how the light is not in phase with itself upon completing a revolution, resulting in non-
continuous circulation, or off-resonance condition and losses.

Figure 2-3 Schematic of light propagation within a circular microcavity showing total internal
reflection in an a) on-resonance and b) off-resonance condition.
2.1.2 Resonant Wavelength
A resonance occurs at particular wavelengths, called the resonant wavelengths ( 
R
). The
resonance criterion is described by the following equation:

(2.2)
where M is the mode number, R
eff
is the effective radius of the mode, and n
eff
is the effective
refractive index. The effective refractive index can be calculated from the optical field
10

distribution in each material region, such as silica for the device material, and air or water as the
surrounding medium. The refractive index is a ratio of maximum speed of light in a material to
the speed of light in vacuum. As examples, the refractive index of air is 1.06, of water is 1.33,
and of silica is 1.45. If 98% of the field is in the device and 2% is outside in air, the effective
refractive index is found as 0.98*1.45+0.02*1.06, which equates to 1.442. When there is a
change in n
eff
, the resonance wavelength will change, as can be seen when the equation is
rewritten as:

(2.3)
This relationship shows how these devices can be used to probe the surrounding environment
and output a detectable change in signal, seen as a resonance wavelength shift.
2.1.3 Evanescent Field
As the light propagates within a microcavity, a small portion of that light propagates just
outside of the device and is known as the evanescent field. The evanescent field is an
electromagnetic field that decays exponentially to zero away from the surface of the microcavity.
The existence of this field does not lead to losses from the cavity, assuming it is operating in a
lossless environment. However, the presence of an evanescent field outside the microcavity
makes it sensitive to its environment, providing for a sensor operation. It is this field that is
responsible for interactions between the device and its surroundings. When new molecules
appear within the field, they change the effective refractive index through which the light travels
and since the resonance wavelength is dependent on the effective refractive index, there is a
detectable change in the resonance wavelength. The sensitivity increases with quality factor, a
measure of light storage capability of the cavity, which is discussed in detail in the next section.
Effectively, photons spend more time inside the cavity, sampling the surrounding outside
11

multiple times, leading to higher sensitivity. The sensitivity makes even single molecule
detection possible with these devices [20].

Figure 2-4 Finite element method simulation result showing the distribution of the electric field in a
microsphere cavity. a) A rendering of a microsphere, where the blue square represents the
simulation area. b) Result for device in air. c) Result for device in a buffer environment. In b) and
c), the redder areas indicate high intensity field and bluer areas show smaller intensity. d)
Normalized mode intensity along the red cut lines in b and c, showing how far the field exists
outside the cavity.
Though finite element method (FEM) simulations are described in greater detail in
another section, a result is presented here to further clarify the existence of the evanescent field.
Through the finite element method, the distribution of the electric field can be computed. This
method is a numerical technique to solve partial differential equations (PDE) or integral
12

equations. By assuming steady state or approximating PDEs as a system of ordinary differential
equations, the problem is simplified into a numerical integration. For running simulations on
microcavities, COMSOL Multiphysics software is used as the PDE solver. The solutions of
Maxwell’s equations, along with known associated expressions for WG modes are provided by
the user [21].
Figure 2-4 shows how the electric field is distributed in a microsphere. Figure 2-4a shows
a rendering of a microsphere with light circulating inside the cavity. The blue square is the
simulation region and is shown as a reference for the other figures. Figure 2-4b and c show
results of FEM simulations performed in Comsol on the cross section of the sphere shown by the
blue square in Figure 2-4a.
Figure 2-4b shows the result for a microsphere surrounded by air, while Figure 2-4c
shows the result when the sphere is surrounded by buffer. These simulation results show that the
electric field is concentrated around the equator of the microsphere in both cases with a small
fraction leaking into the outside. Under careful examination of Figure 2-4b and c, it can be seen
that the field shifts toward the outside of the cavity when the cavity is in the buffer environment.
Figure 2-4d makes it easier to see that the field does extend farther into the environment
if the index is increased. Figure 2-4d shows the normalized intensity of the field along the red
lines drawn through the middle of the mode in Figure 2-4b and c, along the radial axis of the
microcavity. The blue line in Figure 2-4d shows where the physical boundary of the device is
located. In both cases, the field extends outside the boundary of the device. When the device is in
buffer, the field extends even further, indicating enhanced sensing capabilities farther from the
surface. The shape of the intensity curves closely resembles an exponential decay function away
from the boundary of the device. Any changes that occur within this evanescent tail will interact
13

with the electric field and produce detectable signals as a resonance wavelength shift. The
changes could result from new objects in the vicinity, such as molecules, particles, or bigger
objects, and also from changes in the surroundings, including temperature, humidity, and
pressure.
2.1.4 Quality factor
The sensitivity of detection is highly determined by the quality factor of the device. The
quality factor (Q) describes the amount of time light spends inside the cavity. The quality factor
is also a measure of the energy stored in the resonator divided by the amount of energy lost per
circulation inside the cavity. The total Q is a summation of all the loss mechanisms present in the
microcavity. In dielectric cavities, these losses include material, surface scattering, coupling,
radiation, and contamination losses [22]:

1 1 1 1 1 1      
    
cont rad coup ss mat tot
Q Q Q Q Q Q

(2.4)
The first term, 1/Q
mat
, represents the loss due to material adsorption and is intrinsic to the
material from which the cavity is made. By manipulating the material, this loss can be changed,
but will affect other aspects of the resonator as well. The loss due to surface roughness is
represented by 1/Q
ss
and can be minimized by careful fabrication that evens the surface of the
device to atomic smoothness. The 1/Q
coup
term relates to the coupling loss or the loss caused by
inefficient coupling from a waveguide into the microcavity. Coupling loss is minimized by using
a high efficiency waveguide, such as a tapered optical fiber. The bending, or radiation, loss,
1/Q
rad
, is determined by the size of the resonator. A small cavity will force the light to approach
the wall at such an angle that total internal reflection is no longer achieved and some light is
refracted. A larger resonator avoids this problem, and as a result, has a smaller 1/Q
rad
term. The
minimum resonator size to avoid radiation loss limitations is determined by the cavity material
14

and the surrounding environment, as well as the testing wavelength. For example, when the
cavity is tested in water rather than in air, the refractive index contrast is lowered and thus the
device must be made bigger to effectively confine the light. Finally, 1/Q
cont
is the contamination
loss, which is minimized by keeping the cavity in a clean and dry environment such as a
desiccator.
The loss mechanisms are described as either intrinsic to the device itself, or extrinsic and
depend on outside factors. Material, surface scattering, radiation, and contamination losses are
intrinsic to the cavity. Coupling loss is the only extrinsic loss.
The common resonator geometries shown in Figure 2-1 have different dominant loss
sources and, therefore, different Qs. Of the devices fabricated from silica, microspheres have the
highest Q since they are made from pure fused silica and then thermally reflowed to an
atomically smooth surface. The choice of material and fabrication limit the scattering and
material losses. Microtoroids, on the other hand, have slightly lower Q factors because the
thermally grown silica they are made from has impurities (e.g., boron inclusions), which result in
higher scattering and absorption losses. Microdisks have high surface roughness since the
photolithography and etching processes are rough on the material and since there is no reflow
step to even out the surface. Microrings are usually made with a polymer or silicon, which have
higher material losses than silica, and also have rough edges.
2.1.5 Calculating Quality Factor
There are two methods of calculating the quality factor: the linewidth method and the
ring-down method. The linewidth method is implemented by slowly scanning through a small
range of wavelengths with a tunable laser until a drop in transmission is observed, signifying that
light is entering the cavity rather than continuing through the optical fiber waveguide. An
15

example is shown in Figure 2-5a. The Q is found by dividing the center wavelength, 
0
, of the
dip by the full width at half maximum (FWHM), , of the peak:

(2.5)
The other method of calculating Q is the ring-down method, in which the device is
brought to resonance with a laser and the transmission from the device is measured (rather than
from the waveguide). The laser is then turned off and the decay of transmission from the cavity
is recorded, shown in Figure 2-5b. The quality factor is the angular frequency, , times the
photon lifetime, 
0
, in the cavity:

(2.6)
where the photon lifetime is found as the time it takes the signal to decay to 1/e of the maximum.
The angular frequency is found from the resonance wavelength:

c
0

 

(2.7)
where c is the speed of light.

Figure 2-5 Methods of calculating Q. a) The linewidth method, where Q is the center wavelength
divided by the FWHM. b) The ring-down method, where the frequency is multiplied by the photon
lifetime.
16

In the Armani lab, it is more convenient to use the linewidth measurement for the quality
factor range that is usually present (~10
5
-10
8
). The ring-down method can also be used and is
essential for measuring higher Qs.
2.1.6 Free Spectral Range
As an initial test to characterize the device, the free spectral range (FSR) is found by
sweeping the laser over a wide range of wavelengths while maintaining constant coupling
between the sphere and the taper. The FSR of a cavity represents the spacing between successive
longitudinal modes. As mentioned before, the on-resonance condition is such that upon one
revolution inside the device, the round trip in the cavity is equal to an integral number of
wavelengths. The FSR is the spacing between resonance wavelengths when the integral number
per revolution is increased by one (one more period).
The FSR can be calculated theoretically by:

(2.8)
where  is the wavelength of the input light, R is the radius of the device, and n
eff
is the effective
refractive index. When testing a 200 micron diameter microsphere in air with the 765 nm laser,
the FSR should be about 0.64 nm theoretically. Smaller devices have larger FSRs.
2.1.7 Intrinsic Quality Factor and its Loss Mechanisms
As mentioned above, the intrinsic quality factor takes into account all losses except
coupling loss. Scattering, material, and radiation losses can be computed individually through the
use of equations found in reference [23]. Using the Rayleigh scattering model, the surface-
scattering-limited quality factor can be estimated by the inverse of:
17

(2.9)
where  is the root-mean-square size of the surface roughness and B is the correlation length of
the surface inhomogeneities. For silica microspheres,   0.3 nm and B 3 nm, giving a Q
-1
SS
<<
1 x 10
-10
as long as the diameter is greater than 100 m [23]. At larger wavelengths, the surface
scattering losses decrease because the inhomogeneities get less significant compared to the size
of the wavelength. If silica microspheres were surface-scattering loss-limited, the quality factor
would easily exceed 10
10
.
The material-limited Q can be calculated from the refractive index, n, the coefficient of
material loss, , and wavelength of operation, , using the following expression:

(2.10)
This quantity evaluates to 0.9 x 10
10
for a high-purity fused silica microsphere and an input
laser wavelength of 633 nm. Material losses, especially due to absorption, are often the limiting
loss source in optical microcavity biosensors [23]. Compared to surface scattering, material loss
is the bigger loss mechanism for silica microspheres and often the limiting loss mechanism.
The equation used to calculate radiation loss is complicated and requires a lot of
knowledge about the device surface and material properties. However, in most cases, this loss
can be ignored since the device size is large enough to eliminate radiation loss. The equation for
the quality factor associated with radiation loss can be calculated from [24]:

(2.11)
18

with

and

where c is the speed of light in vacuum, 
0
is the permittivity of free space, Z
0
is the impedance
of free space (~377), N
s
is the normalization constant found in [24], n
s
is the refractive index of
the material of the cavity (n
s
= 1.45 for silica microspheres), n
0
is the surrounding medium index
(1.06 for air, or 1.3 for water), l is one of the three integers that describe the mode inside the
sphere, k is the wave vector, R
0
is the radius of the sphere, and the j values are Spherical Bessel
functions, formulas for the computation of which have been found [25].
One might assume that minimizing the losses produces a more perfect device, when in
fact a few disadvantages exist for very high quality factors. With narrower bandwidths, it is
much harder to correctly phase match the resonator with the waveguide, producing coupling loss
and making it very difficult to fully excite the fundamental modes. High Qs could also result in
heating of the device from the high circulating power within the cavity. This heat distorts the
resonance linewidth, which is detrimental for sensing applications. This heating could also result
in a resonant shift, which could be mistaken for one due to sensing. As the laser frequency is
swept across the cavity resonance, optical power coupled into the resonator is partially absorbed
and converted to heat, altering the optical properties of the bulk medium and shifting the
resonant frequency either along or opposite to the direction of laser scanning. The direction of
the shift is determined by optical properties of the device. The resulting linewidth looks like
either a broadened or a narrowed peak. The change in refractive index can be calculated from the
resonant frequency shift caused by the temperature change [26]:

(2.12)
19

where  is the resonant frequency resulting from a shift away from the original frequency, 
0
.
This relationship is used to characterize optical properties of new materials. Temperature-
dependent shifts will be further discussed in the section about gas sensing with carbon nanotube
coated microspheres.
2.1.8 Finesse
Another measure of cavity perfection is finesse, which is similar to Q, but without
propagation effects within the cavity. For a Fabry-Perot resonator, the Q can be related to finesse
(F) by the following equation [27]:

(2.13)
where t=2d/c is the time a photon takes to go around the resonator once, F= /L, where L is the
total mirror loss, c is the speed of light in vacuum, and d is the diameter of the FP resonator.
For an optical resonator, the finesse can be defined as the ratio of the free spectral range
of the cavity ( 
FSR
) to the full width at half max of the resonance peak ( 
0
) multiplied by Q:

(2.14)
where  is the width of the peak in frequency units. Since the free spectral range is much larger
than a typical resonance linewidth, Q is always much greater than F. It is preferred to work with
Q, since it is more widely referenced in literature than finesse.
2.1.9 Coupling
The only extrinsic loss mechanism in microcavities is coupling loss. To minimize the
loss, a high efficiency waveguide must be used. A tapered optical fiber is ideal to couple light
into microcavities evanescently. Optical fiber is made to confine and propagate light of a
particular wavelength range in single-mode configuration. Fiber spools were purchased from the
20

Newport Corporation to match the wavelength ranges of the lasers used in lab. Optical fibers
have a cutoff frequency, which corresponds to the lowest wavelength that can propagate in the
fiber at single-mode operation.
Multi-mode operation is detrimental to testing conditions due to multiple loss
mechanisms. When operating in single-mode, the mode travels near the center of the fiber and
loss to the cladding is minimal. With multi-mode traveling through the fiber, more light is lost to
the cladding since higher order modes travel closer to the edge of the fiber core. The other loss
occurs when coupling into the microcavity. Even though light is not actually lost, less of the light
enters the cavity. This is because the mode in the taper couples with the mode in the microcavity.
While higher order modes are able to couple into the device, the fundamental mode couples best.
Therefore, a large fraction of the light will continue to travel down the optical fiber instead of
coupling due to mode-mismatch.
There are three regimes of coupling: under-coupled, critically coupled, and over-coupled.
In the critically-coupled regime, the coupling loss is the same as the cavity loss and the resulting
transmission at the output of the waveguide is zero at resonance since at this point, all input
power is coupled into the cavity. Critical coupling sometimes cannot be achieved due to parasitic
losses, or coupling into multiple modes at once, preventing the power from fully going into one
mode alone. As the distance between the taper and resonator is increased, the system enters the
under-coupled regime. The overlap of the fiber taper mode and the resonator mode is lowered,
resulting in less power entering the device. If the taper is taken out far enough, there is no
overlap of the modes, and thus no coupling.
The last possible coupling condition is the over-coupled one. In this regime the taper is
very close to the device, already past the critical coupling distance. The modes overlap even
21

more, but the transfer of energy is not as efficient, causing the transmission to drop from
maximum, but not reaching zero. The coupling loss is increased in this case and is bigger than
the cavity losses. Figure 2-6 shows a representative schematic with different coupling conditions
and how coupling affects the same resonance peak.

Figure 2-6 Representative schematic of different coupling conditions.
There are a few other ways to couple light into microcavities without the use of a tapered
optical fiber waveguide. A prism coupler is flexible and versatile, capable of efficiently coupling
light evanescently into and out of a cavity [28]. The drawbacks of using a prism are its bulky size
and the need for collimation of light when used with optical fiber. A side polished bent fiber
coupler is another possibility, but it is usually not used due to residual phase mismatch [29]. An
angle polished fiber tip is a preferred coupling device, since it allows for more efficient coupling,
but it is hard to position and requires delicate cutting and polishing [30]. A polished half block
coupler provides the most robust way to couple light, but the coupling efficiency is low due to
light leaking into the cladding radiation mode [31].
Out of all of the mentioned coupling apparatuses, the tapered fiber provides the best
method of energy transfer in and out of the device due to easier phase matching and efficient, air-
guided, evanescent coupling. Its major disadvantage is parasitic losses, or excitation of higher
22

order taper modes or radiation modes in the cavity by either the taper mode or the target
resonator mode.
The efficiency of coupling depends on many factors, such as the wavelength, the shape
and size of the microcavity, the thickness of the waveguide in the coupling region, and the gap
between the waveguide and resonator. All of these factors can be controlled to maximize
efficiency of coupling. The use of a nano-scale motorized stage to move the microdevice, allows
for good control of coupling distance. The piezo stage has three axis controls with 1nm step
sizes, so the device can be positioned with high precision.
2.1.10 Polarization and Mode Numbers
It is convenient to consider WGMs of two different polarizations: transverse electric (TE)
and transverse magnetic (TM). In TE modes, the electric field is perpendicular to the direction of
propagation while the magnetic field is parallel to it, whereas in TM modes, the magnetic field is
perpendicular. The modes are characterized by four quantum numbers: n, the radial mode
number, l, the polar or angular mode number, m, the azimuthal mode number, and p, the
polarization [32]. The angular mode number is equal to about the number of wavelengths that fit
into the optical length of the equator. Mode number n is the number of field maxima along the
radius of the sphere. The number of maxima in the azimuthal variation of the resonant field
around the equator is 2l. The value l-m+1 is the number of field maxima in the polar direction
(perpendicular to the equatorial plane).
The resonant wavelength can be found from n and l. The fundamental mode is defined by
an n value of 1 and when m is also equivalent to l. This mode is closest to the surface of the
cavity, especially near the equator. The TE modes are generally more confined within the cavity,
making them harder to excite, since there is less overlap with the waveguide mode. Since one
23

polarization can have a higher Q, it could be in the interest of sensitivity or increasing lasing
power to try to selectively excite one over the other by controlling input polarization.
2.2 Testing Procedures
2.2.1 Pulling a Taper
To ensure minimal loss through multi-mode tapers, it is essential to use fiber meant for
the wavelength of interest. Optical fiber is made from three parts: the inner core, the cladding,
and the protective polymer coating. For different wavelengths, the core diameter and index is
changed. The protective coating is there to prevent damages to the cladding, but it is removed
with fiber strippers before pulling a tapered optical fiber. The fiber is also cleaned with isopropyl
alcohol to remove any residual cladding or dust particles.

Figure 2-7 Optical fiber holder used for taper pulling. The clamps on top hold the fiber in place.
A specialized holder is used to hold the taper in place and for taper pulling. The holder is
shown in Figure 2-7. There are two movable pillars that can be locked in place with screws. On
top of each pillar is a clamp that holds the piece of fiber in place. The clamp has a magnet that
sticks to the magnet on the pillar. The fiber is also covered with O-Ring cord to avoid damage to
the fiber. The micrometer on the side is in case the pulled fiber becomes loose and manual
tightening is necessary.
24

Using a hydrogen torch and movable stages, the optical fiber is heated and pulled apart
until the vestigial core merges with the cladding and is thin enough to be able to couple light out
evanescently. The light is guided due to the index contrast between the core of the fiber and the
air rather than the cladding (Figure 2-8). Figure 2-8a shows the optical fiber without the
protective polymer layer, so only the core and cladding are seen. Figure 2-8b shows the taper
after pulling with rainbow colors in the taper, indicating that it can be used to couple 633 nm
light into a microcavity. Both images are taken at the same magnification and on the same
camera, therefore, the size difference can be directly compared. The width of the tapered region
is about 500 nm and is ideal for testing with 633 nm light. Coupling occurs when there is phase
matching between the light going through the taper and the mode of the microcavity.

Figure 2-8 Tapering of an optical fiber. a) Single mode optical fiber showing the core and cladding
regions. b) Tapered optical fiber with a thinned merged region of core and cladding.
2.2.2 Fabricating Silica Microspheres
To fabricate microspheres, an optical fiber piece is used that has been stripped of the
protective jacket layer. The fiber is placed in front of a carbon dioxide (CO
2
) laser. The heat
from the laser melts the silica and surface tension forces the molten material into a spherical
shape. Figure 2-9 shows the different steps in sphere fabrication. The size of the sphere is
a)
b)
25

controlled by choosing the initial size of the optical fiber. By tapering the fiber with a hydrogen
torch, thinner fibers and thus smaller spheres can be made. Since the sphere is still attached to
the fiber tip, it can be easily positioned it in the testing setup.

Figure 2-9 Fabrication of microspheres. Starting with a cladding free optical fiber, the tip is melted
to produce a sphere.
2.2.3 Finding Resonance Wavelengths
The overall testing setup is shown in Figure 2-10. In the lab, there are several different
tunable lasers that can be used. However, the majority of the work in this thesis used lasers with
center wavelengths of 633 nm, 765 nm, and 980 nm as the light source. The light is transmitted
through optical fiber to the tapered region and out to a photodetector. The output is read by a
high speed digitizer/O-scope PCI card on the computer. When the power in the photodetector
drops to zero, the light is coupling into the cavity, signifying that it is on resonance. The output is
observed on an oscilloscope front panel program where the display options can be changed. A
function generator PCI card can also be controlled, which regulates the output of the laser
through frequency modulation. The function sent to the laser selects a range of wavelengths over
which to scan. A triangular waveform is used to scan the laser in the forward and reverse
directions. Both directions need to be seen to know if some non-linear effect is taking place. This
is detected when the resonance peaks in the forward and reverse range are not symmetrical.
26

Figure 2-10 Testing setup. The laser couples light to the microsphere through a tapered fiber, which
is then connected to a photodetector (PD). The output is observed on an oscilloscope and the laser
can be controlled with a function generator.
The solid blue line in the figure between the laser and the PD represents an optical signal,
while the dashed red lines represent electrical signals. Three computers are shown for clarity
while in reality all signals are delivered to a single computer.
2.2.4 Measuring the Quality Factor: Broad Scan and Fine Scan
The quality factor can be measured in two ways. The first is through the broad scan
method, which is also used to find the FSR of the device. To measure the FSR of the devices
experimentally, a relatively fast scan rate (0.2-0.5 nm/sec) and a large scan range (a few
nanometers) are used to detect the repeated fundamental modes. The mode is fundamental when
the taper and device are phase matched, allowing for all of the power in the taper to enter the
resonator, causing a drop in transmission to zero power. The device must also be critically
coupled to the taper by adjusting how far away it is from the taper. The maximum power will be
coupled into the resonator only under critical coupling conditions. When the dips in this broad
scan do not reach zero transmission, this signifies that either the taper was not of the correct
27

width to give phase matching or the coupling was not critical. By measuring the distance
between the fundamental peaks in the scan, the FSR is found. Figure 2-11a shows the
representative broad scan schematic that would result from this experiment under ideal
conditions.

Figure 2-11 Broad scan. a) A representative schematic of a broad scan showing the free spectral
range of a device with critically coupled peaks. b) A set of data taken on a microsphere showing an
FSR of 0.78 nm.
Microspheres have many modes that light could couple into because of their spherical
shape. Because of this, when acquiring a broad scan, the resulting data shows many dips in
transmission. It is important to also record the taper scan, which is the scan of the laser without
coupling into the device, recording the oscillations that may occur in the laser with changing
wavelength. The data is then normalized by dividing the coupling scan by the taper scan. The
result taken on a sphere of 192 micron diameter is shown in Figure 2-11b. The upper region of
the graph is noisy because normalization is made difficult due to the many resonances that occur.
28

Nevertheless, an FSR of 0.78 nm can be seen, which is close to the expected 0.7 nm predicted
theoretically for this size device at this wavelength.
The quality factor of the peaks recorded in broad scan can be measured, but they will
contain coupling losses, since broad scan is done at critical or close to critical-coupling.
Additionally, for ultra-high-Q peaks, there will not be a sufficient number of data points for an
accurate fit. Therefore, when measuring Q, the coupling loss must be reduced, placing the system
as far into the under-coupled regime as possible while still seeing a dip. If the Q of the device is
around 10
6
, broad scan can still used to record and analyze the data. If, however, the quality
factors are much higher, it is necessary to rely on fine scan.
In fine scan, the laser is manually stepped through a range of wavelengths with a step size
of 0.001nm. Because the approximate location of the fundamental peaks was determined in
broad scan, the laser is placed close to that spot and stepped until the peak appears on the O-
scope. To more closely estimate the intrinsic Q of the device, coupling is reduced so that the
resonance is barely visible over the noise-level. Figure 2-12a shows a representative schematic
with data points shown as black dots and the Lorentzian fit as the red line. The quality factor is
found by dividing the width of the peak at half maximum by the center wavelength of the peak.
Figure 2-12b shows experimental data of a resonance with a Lorentzian fit. The Q was calculated
to be 1.85x10
8
. The coupling is only about 8%, so the coupling losses are minimized.
29

Figure 2-12 Fine scan. a) A theoretical schematic of a quality factor calculation showing the data as
black dots and the Lorentzian fit as a red line. The graph also shows an example of where the
FWHM is measured. b) An example of experimental data showing an under-coupled resonance
peak on a microsphere with a fit Lorentzian function.
2.2.5 Finding Intrinsic Quality Factor
To experimentally find the intrinsic quality factor, the Q at various degrees of coupling is
measured in a fine scan. The dip in transmission can be changed to a percentage of coupling by
considering how close the dip is to zero transmission. Considering normalized transmission,
when the peak reaches 0.2 on the scale, this signifies coupling of 80%. A Lorentzian curve can
be fit to each recorded Q. By plotting the Q vs. the percentage coupling, a straight line results as
long as non-linear effects are not present in the resonance. By extending the line to zero percent
coupling, the intrinsic Q can be estimated. At this point, there would be no coupling loss, which
increases linearly with coupling.
a) b)
30

Figure 2-13 Intrinsic quality factor calculation showing four points of coupling and a linear fit.
An example data set for calculating intrinsic Q is shown in Figure 2-13. Four points
appear in the graph at the different coupling percentages. As the coupling increases, the Q
decreases. Fitting a line to these points gives the intrinsic Q as the y-intercept and in this case it
is 2.26x10
8
. Therefore, even with low coupling, such as 8% shown in Figure 2-12b, coupling
losses affect the calculated Q.
2.2.6 Resonance Shift Measurements
After a resonance peak with a good Q that does not display non-linear effects is found,
detection experiments can be performed with the device by tracking the resonance peak. A
modified O-scope program designed by Matthew Reddick is used that records the minimum
position of the resonance over time. The program records both the x and y-positions, but only on
the global minimum point within the O-scope window. If the time axis is converted into
wavelength, the wavelength change can be plotted over time to analyze the detection
mechanisms.
31

Figure 2-14 View of minimum tracking program. Only the global minimum peak will be tracked.
Figure 2-14 shows the front panel view of the O-scope program. Other resonances can
still see on the screen, but only the minimum of the peak indicated by the blue arrow will be
recorded. This is because that peak contains the global minimum point. If coupling conditions
change, a different resonance could become the minimum, so it is important to maintain coupling
as best as possible.
2.2.7 Temperature Stage Measurements
Another piece of equipment that could be used in experiments is the temperature stage.
The stage is designed with a heating element and a thermocouple connected to the same
controller [33]. The feedback loop makes it possible for the controller to automatically adjust the
heating element to maintain a constant temperature. Figure 2-15 shows the stage and the
controller with labels indicating the important components. The heating element is not visible
because it is embedded into the aluminum interior of the temperature stage. The thermocouple is
attached to the same part of the stage as where the samples are placed during testing to more
32

accurately measure the temperatures the devices are exposed to. The temperature controller is
sensitive to 0.1 degrees Celsius.

Figure 2-15 Temperature stage and controller used for temperature-dependent experiments.
2.3 Lasing Phenomena in Microresonators
2.3.1 Overview of Lasing in Microresonators
WGM optical microcavities have attracted momentous attention as a platform for
creating ultra-low threshold microlasers [34] with applications in biodetection [35], quantum
optics [36], and telecommunications [37]. Because WGM cavities accumulate and build-up
optical power of a particular wavelength, the threshold for lasing emission is lowered. Ultra-high
Q results in extremely large optical field intensities, lowering the threshold for nonlinear
phenomena, such as the Kerr effect, stimulated Brillouin scattering, and multi-order Stokes
emission, also known as Raman scattering. A higher Q gives a narrower bandwidth, making the
device suitable for other uses: mechanical oscillators [38], cavity quantum electrodynamics
measurements [39], and add-drop filters [40].
One example of a nonlinear optical phenomenon is described by the Kerr effect. For
especially intense optical fields, the refractive index of a material becomes a function of the
33

electric field to which it is exposed. The Kerr coefficient, K, relates these two quantities
according to the expression n = KE
2
.
Stimulated Brillouin scattering (SBS) occurs for intense beams of light in which the
variations in the electric field produce acoustic vibrations in the material, resulting in a change of
the energy and path of the light. The light scatters in the opposite direction from the incoming
beam, with frequency shifts on the order of 1-10 GHz, equivalent to 1-10 pm wavelength shifts
in visible light.
Stokes emission occurs when a molecule or atom that absorbs a photon and enters an
excited state, then releases a photon of lower energy, examples include fluorescence and Raman
scattering. If the emitted photon has more energy than the incident one, the emission is called
anti-Stokes. Multi-order Stokes emission occurs when the initial Stokes emission is high enough
in intensity to serve as the pump source for further Stokes emissions [41]. Since a cavity can
build up the light intensity in the material, many orders of Stokes emission can occur in the
device.
Specifically, silica microcavities, including microspheres and microtoroids, have
extremely high quality factors, or photon lifetimes, and make ideal candidates in this field.
Lasing occurs when enough power is applied to a material to produce stimulated emission. If not
enough energy is applied, then the system could be undergoing spontaneous emission. Lasing
begins with spontaneous emission, but builds up to stimulated emission.
2.3.2 Purcell Factor
The Purcell factor describes how the spontaneous emission rate is enhanced in a resonant
cavity. It can also suggest how good a specific resonator would be for lasing applications with
34

higher factors indicating better lasing capabilities [42]. The Purcell factor is given by the
following relation:

3
2
4
3









eff m
P
n V
Q
F



(2.15)
where Q is the quality factor, V
m
is the mode volume,  is the resonance wavelength, and n
eff
is
the effective refractive index.
2.3.3 Mode Volume
The mode volume mentioned in the Purcell factor equation indicates how much space a
certain mode of a resonator occupies. It can be calculated by using the following formula:

(2.16)
where V
Q
is the quantization volume of the field, chosen as the region near the optical mode, | Ē|
is the strength of the electric field, n is the refractive index, r is the position along the radius, and
|Ē
max
| is the maximum electric field strength [43]. The smaller the mode volume, the bigger the
Purcell factor, implying better lasing potential. However, it is also important to remember that
making the cavity too small will increase the radiation loss source in the Q. A smaller Q
decreases the Purcell factor. These two parameters must be optimized for lasing applications.
For silica microspheres, the mode volume can be approximated by two simple equations
for the two mode polarizations [44]:

6 / 7
6 / 11






 
n
D A V
m


(2.17)
35

where A is 1.02 for TE modes and 1.08 for TM modes. D is the diameter of the sphere,  is the
wavelength, and n is the refractive index. The TM mode volume is a bit larger due to its farther
extent out of the sphere.
2.3.4 Circulating Power
The power circulating inside the cavity (P
cav
) can be orders of magnitude larger than the
input power and can be calculated according to [45]:

(2.18)
where Q, the quality factor, is on the order of 100 million in microsphere cavities, the input
power, P
in
, is about 1 mW to 10 mW for Newport tunable lasers, and the diameter, D, is about
200 m for microspheres, giving a circulating power of 10-100 W. Circulating intensity can be
found by dividing the cavity power by the effective volume: I
cav
= P
cav
/(V
eff
/D ) and gives
intensities on the order of 1 GW/cm
2
.
2.3.5 Threshold Power and Efficiency of Laser
One of the motivations for pursuing microcavity-based lasers is the resonant
amplification of the input power, which is proportional to the quality factor (Q), or photon
lifetime within the cavity. This amplification enables the development of ultra-low threshold
lasers, which operate at otherwise unachievable wavelengths and in complex environments. The
threshold for stimulated emission rather than spontaneous emission has the form:

2
th
P
gQ
q V
S D
eff
 
 (2.19)
where V
eff
is the effective mode volume, q is the quantum efficiency, 
D
is the pump wavelength,

S
is the lasing wavelength, g is the gain coefficient, and Q is the quality factor. According to
36

this equation, the smallest mode volume and highest Q device is the best. However, the two
parameters must be optimized to maintain the highest Q and produce the lowest threshold.
The lasing threshold is also dependent on the coupling condition. The threshold will be
minimized at critical coupling and increase when the device is over or under-coupled. If the
device is doped with a rare-earth metal, for example, the threshold will change with the dopant
concentration. Too much doping will increase the threshold since dopants introduce impurities
that lower the Q. The threshold power can be used to find the efficiency of a laser:

(2.20)
where P
laser
is the output power of the lasing emission, P
pump
is the power being pumped into the
device, 
laser
is the wavelength of lasing emission, and 
pump
is the wavelength of the pump laser
[46].
2.3.6 Lasing Measurements
Lasing emission from microresonators can be detected in one of two ways. A fiber-
coupled spectrograph is used that detects light coming off the cavity, or an optical spectrum
analyzer (OSA) can be used that detects lasing emission that is coupled back into the taper
waveguide.
The spectrograph is made up of the fiber tip, which leads to the chamber with the grating,
and a CCD camera attached at the output, which needs to be cooled when acquiring data to
reduce thermal noise. The spectrograph fiber tip is placed as close as possible to the resonator
because the fiber tip is able to detect only a small fraction of the available light. The advantage
of using the spectrograph is that not all lasing emissions will be able to couple back into the
taper, since the taper is fabricated to carry the wavelength of the pump wavelength, but the
emission wavelength could be very different. This is a particular issue if the emission
37

wavelength is lower than the pump wavelength. Additionally, the operating range of the
spectrograph is from the UV to the near-IR (~1700 nm), whereas the OSA operated from the
visible (600 nm) to the near-IR (~2200 nm). Also, the OSA can give quantitative values (power)
instead of counts. Therefore, they are complementary in many ways.
The resolution of the spectrograph depends on the grating used inside. Some gratings
allow the user to look at a wide range of wavelengths at the same time, but the resolution is
lower. The highest resolution grating used was 0.16 nm, but the other two had 0.17 nm and 0.76
nm resolution. The wavelength range also changes with the grating and the three available
gratings allow observation of the range from 250 nm to 1700 nm, though not all at once or on the
same grating.
The OSA has better resolution than the spectrograph (6.8 pm) [47]. However, this value
is several times larger than the lasing linewidth, and therefore, it limits the ultimate sensitivity of
the device. To get the emission to the OSA, the output from the tapered optical fiber that is
usually sent to the detector is split so that it reaches both the detector and the OSA. The splitter is
designed for a certain wavelength and if the splitter does not match well with the lasing
wavelength, the signal may not make it to the OSA.
2.3.7 Raman Lasing
Raman, as opposed to Rayleigh (elastic) scattering, is the inelastic scattering of a photon
from an atom or molecule, resulting in a change in energy and wavelength of the emitted photon.
Raman scattering can occur as either Stokes (system absorbs energy) or anti-Stokes (system
loses energy) scattering. The difference in energy is equivalent to the difference between
vibrational or rotational energy levels of the system. The absolute value of the difference in
energy does not depend on whether the process is Stokes or not, since the energy difference
38

comes from different vibrational or rotational levels. With enough excitation power, the
spectrum will be symmetric around the pump wavelength, with Stokes emission on the red and
anti-Stokes on the blue side.
The intensities of emitted photons are proportional to the number of molecules that
occupy the vibrational states at the start of the process. At conditions of thermal equilibrium, the
distribution of molecules in different energy states is given by the Boltzmann relation:

(2.21)
where N
0
, and N
1
, is the number of atoms or molecules in the lower, or higher, vibrational state,
respectively, g
0
, and g
1
, is the degeneracy of the lower, or higher, vibrational state, E

is the
difference between the energy states, k is the Boltzmann constant, and T is the temperature. From
this distribution at room temperature, it is apparent that more atoms or molecules will occupy the
lower energy state, producing a more intense anti-Stokes spectrum.
Raman scattering is different from other types of energy exchanging scattering, such as
fluorescence, due to the presence of virtual energy states. Fluorescence occurs by the process of
transferring the system into an excited state and dropping it back to a number of lower states.
The other difference is that Raman scattering can occur at any incident wavelength, with the
emission peak at the same distance from the pump for any excitation frequency. With high
intensity build up, the resulting Raman peaks serve as subsequent excitation frequencies to
produce higher-order Raman peaks. Figure 2-16a shows the energy level diagram for several
spectroscopy phenomena in which photons are absorbed and emitted by the system. Ultraviolet-
visible (UV-Vis) and fluorescent spectroscopy both work by exciting the system to an excited
state. Infrared (IR) spectroscopy operates at the high wavelength range and results in an
excitation in vibrational states.
39

In comparison, Raman scattering is excited to a virtual state and decays to different
vibrational states. Figure 2-16b shows the detailed instances of Rayleigh and Raman scattering.
For elastic scattering, the system returns to the same state from which it started and no new
emissions are observed. In Stokes scattering, the system begins at the ground state but ends up at
a higher energy vibrational or rotational state, producing an emission different from the pump
source. For Stokes, the emission will be of lower energy and higher wavelength. Anti-Stokes
scattering is observed only when a sufficient fraction of the material is already in an excited
rotational or vibrational state. This type of scattering results in an overall decrease of energy in
the system since the emitted photons have a higher energy than the pump source and appear at
lower wavelengths.

Figure 2-16 a) Energy level diagram showing different photon absorption and emission phenomena
[48]. b) Detailed energy level diagram of Raman and Rayleigh scattering.
2.4 Microresonators as Biodetectors
Another important application of microcavities is biodetection. Microtoroids have been
shown to detect at the single molecule level without requiring any labeling of the analyte
40

molecules [20]. The extreme sensitivity comes from the presence of the evanescent field around
the microresonator surface, which can detect minute changes in the environment.
Getting a solution to the surface of the device is thus an important part of sensing
experiments. Counting on diffusion alone is inefficient since the microresonators have such a
tiny surface area with which the molecules interact. Without convection, the response time for a
certain degree of binding is inconveniently high. Usually a flow is required to provide fast
response times and allow for small quantities of low concentration solutions to be used. The
transient response of an optical microcavity biosensor will depend strongly on the flow
conditions to which it is exposed. High liquid flow rates lead to faster adsorption and a more
dense coverage of molecules on the surface. Since the evanescent field tail decays exponentially
with distance, the molecules that adsorbed to the device will be the ones detected. As molecules
approach the surface, the effective refractive index changes, resulting in a detectable resonant
frequency shift. By tracking the shift, the initial binding rate of the molecules to the device
surface can be found [49].
A common method of testing these devices is the injection of the analyte solution over
the surface of the device. As with any flow system, the boundary layer around the surface must
be considered to determine the effects of the flow. The boundary layer is the layer of fluid near
the surface of an object, where viscous effects are significant. When there is no flow, the
concentration of analyte near the surface of the cavity is essentially zero and only by random
motion can the molecules get to the surface. When a small flow rate is applied, the boundary
layer gets a lot smaller than the no-flow case and the molecules don’t have as far to travel to g et
to the surface. At very high flow rates, the boundary layer become miniscule and the
concentration near the surface approaches that of the injected solution very quickly.
41

The changes in boundary layer cause a change in the flux. Fick’s Law of diffusion relates
the diffusive flux to the concentration by:

(2.22)
where D
AB
is the diffusion coefficient, c is the concentration, r is the distance from the surface, c
0

is the bulk concentration, and  is the thickness of the boundary layer. At r = 0, or at the surface
of the cavity, the equation reduces to the ratio of bulk concentration and boundary layer
thickness, all multiplied by the diffusion coefficient. This approximation holds because the
concentration near the surface approaches zero and thus the difference between the bulk
concentration and that of the surface is just equivalent to c
0
. What this simplification implies is
that increase in flow rate, and thus a shrinking boundary layer thickness, gives a higher flux.
Therefore, at higher flow rates, the solution should reach the surface faster, and more molecules
should be able to adsorb to the surface.
An important factor to consider when performing biodetection experiments is that the
surrounding environment greatly affects the microresonator. Since most biodetection
experiments are performed in an aqueous environment, the absorption of light in the medium has
to be minimized. Water will absorb light of higher wavelengths (≥ 980 nm) a lot more than light
with wavelengths in the visible range. Larger absorption of light in the water leads to a reduced
quality factor since the overall material loss is higher. For this reason, 633 nm and 765 nm lasers
are used for most biodetection experiments. Different aqueous environments absorb light in
varying degrees. For example, heavy water (D
2
O) absorbs less than regular water and leads to
increased quality factors [50].
42

The great advantage of using WGM resonators is the ability to detect molecules without
labeling them [51, 52]. Any molecules that reach the surface will be detected. To improve
selectivity, the device itself can be functionalized so that only specific molecules will bind to the
surface during detection. Additionally, resonators coated with polymer films allow for numerous
applications, such as measuring the optical properties of the polymer [33] or harnessing the
properties of the polymer to create a new type of sensor [53].
2.5 Ultrasound Imaging
Ultrasound imaging has seen a lot of progress in recent years in terms of higher
resolution images and smaller signal transmitters and receivers. Ultrasound (US) is sound with
an oscillating frequency higher than what can be heard by the human ear, in the range of 20 kHz
to 200 MHz. For tissue and organ imaging purposes, frequencies in the range of 1 MHz to 20
MHz are typically used. As frequency is increased, the resolution is increased, but the
penetration depth is decreased. US has the advantage of providing information about soft tissue,
whereas rival techniques, such as x-ray imaging cannot. US is used as a diagnostics tool in
medicine because of its ability to image tissue, as well as provide information on flow magnitude
and direction of blood flow, and unlike x-rays, ultrasound waves are not harmful to biological
tissues.
The most important aspects of ultrasound imaging are spatial and temporal resolution.
Spatial resolution is the ability to distinguish points in close proximity to one another. Temporal
resolution depends on how quickly the image frame can be captured and processed by the
equipment. Spatial resolution can be further split into axial and lateral resolution. Axial
resolution is the smallest distance between adjacent particles that can be distinguished in the
direction of the ultrasound wave (Figure 2-17a). This resolution is dependent on the spatial pulse
43

length of the ultrasound. Since the spatial pulse length is equivalent to the number of cycles in a
pulse multiplied by the wavelength of the wave, the resolution can be improved by introducing
damping on the transducer as this decreases the number of cycles. Also, it is apparent that
working at higher frequency also increases resolution since the wavelength will be smaller.

Figure 2-17 Resolution in ultrasound imaging. a) Axial resolution describing the distance at which
particles laying parallel to the incoming ultrasound beam can be distinguished as individual
objects. b) Lateral resolution is the distance between two particles that lie perpendicular to the US
beam for them to be distinguished as individual objects. The lateral resolution changes for the
transducer in the focal range.
Lateral resolution describes how well objects perpendicular to the ultrasound beam can
be resolved. This resolution greatly depends on the width of the ultrasound beam (Figure 2-17b).
The beam with is governed mainly by the size of the transducer, so smaller transducers produce
better lateral resolution. The lateral resolution may also vary with the distance from the
b) a)
44

transducer due to the beam converging and diverging as it travels away from the transducer. In
Figure 2-17b, the two objects at point 1 and 3 will not be resolved since they both lie within the
ultrasound beam; however, the objects at point 2 will be seen as two individual particles since
they lie in the focal point of the transducer where the beam is narrow.
The most common imaging technique is pulse-echo, in which a transducer sends an
ultrasound pulse, then receives and interprets the echoes resulting from reflections at interfaces
within the substance being imaged. For higher resolution imaging, high-frequency transducers
are required. However, it becomes more difficult to produce small piezoelectric elements as well
as high-density element arrays. By incorporating resonant optical cavities, the size issue is
solved. Optical cavities maintain high sensitivity even with reduced sizes, have high frequency
and wideband response, as well as provide for easy fabrication.
Polymer microring resonators have been studied for their ability to detect ultrasound.
Through the elasto-optic effect, in the presence of ultrasound, the refractive index of the polymer
is changed, thus shifting the resonance wavelength. This shift can be best observed as an
amplitude modulation at a fixed wavelength of the laser positioned at the maximum slope within
a resonance dip. Besides detecting ultrasound echoes, these devices can also serve as detectors in
photoacoustic imaging experiments. When a material absorbs laser energy, it undergoes thermal
expansion and generates a broadband acoustic wave. This wave can serve as the ultrasound
pulse, and has been previously used for imaging with the polymer ring as the detector [54, 55].
The highest resolution achieved in photoacoustic imaging was with a polystyrene microring. The
lateral and axial resolutions were 150 and 90 m, respectively [55]. Lateral and axial resolution
describe how close two objects can be physically perpendicular or along the beam path and still
be displayed separately on the image. Current medical equipment has resolution of 1mm or less.
45

One way to establish the performance of the sensor is with the noise equivalent pressure
value, which is a measure of the minimum detectable pressure. The best performing polymer
microrings have surpassed the state-of-the-art piezoelectric hydrophones with NEPs as low as
29Pa [56] (compared to 5kPa for the hydrophone [55]).
There is still room for improvement in this area. Higher input power into the device could
allow for greater sensitivity. Powers of 20-30 W were achieved in the output of a fiber coupled
bus waveguide responsible for bringing light to the microring [55]. Tapered optical fibers have
low loss and could thus deliver a lot of power to the device. With the high power 765 nm laser,
up to 10 mW power in the cavity can be achieved. Higher quality factors would also increase
performance. The highest achieved intrinsic Q in water, using a 780 nm laser, is 3x10
5
for
polymer microrings [56]. Microspheres are able to maintain ultra high Qs of 10
8
in water when
using visible wavelength lasers, such as 633 or 765 nm. With these improvements, we hope to
further lower the detectable pressure limit and increase sensitivity of ultrasound imaging.
References
1. Guo, J. and E. Lindner, Cyclic voltammograms at coplanar and shallow recessed
microdisk electrode arrays: guidelines for design and experiment. Anal Chem, 2009.
81(1): p. 130-8.
2. Kwon, M.S. and W.H. Steier, Microring-resonator-based sensor measuring both the
concentration and temperature of a solution. Opt Express, 2008. 16(13): p. 9372-7.
3. Luo, L.W., et al., Power insensitive silicon microring resonators. Opt Lett. 37(4): p. 590-
2.
4. Rafizadeh, D., et al., Waveguide-coupled AlGaAs / GaAs microcavity ring and disk
resonators with high f inesse and 21.6-nm f ree spectral range. Opt Lett, 1997. 22(16): p.
1244-6.
5. Kippelen, B., Optics in energy: the power of optical solutions. Opt Express. 18(9): p. A1-
2.
6. Tien, M.C., et al., Ultra-high quality factor planar Si3N4 ring resonators on Si
substrates. Opt Express. 19(14): p. 13551-6.
7. Jiang, X., et al., High-Q double-disk microcavities for cavity optomechanics. Opt
Express, 2009. 17(23): p. 20911-9.
8. Fan, X., et al., Inhomogeneously broadened ring-cavity optical-bistability system with a
Gaussian transverse intensity profile. Phys Rev A, 1991. 44(3): p. 2048-2056.
46

9. Painter, O., et al., Two-dimensional photonic band-Gap defect mode laser. Science, 1999.
284(5421): p. 1819-21.
10. Srinivasan, K., et al., Optomechanical transduction of an integrated silicon cantilever
probe using a microdisk resonator. Nano Lett. 11(2): p. 791-7.
11. Vernooy, D.W., et al., High-Q measurements of fused-silica microspheres in the near
infrared. Opt Lett, 1998. 23(4): p. 247-9.
12. Sandoghdar, V.V., et al., Very low threshold whispering-gallery-mode microsphere laser.
Phys Rev A, 1996. 54(3): p. R1777-R1780.
13. Liang, W., et al., Whispering-gallery-mode-resonator-based ultranarrow linewidth
external-cavity semiconductor laser. Opt Lett. 35(16): p. 2822-4.
14. Armani, D.K., et al., Ultra-high-Q toroid microcavity on a chip. Nature, 2003.
421(6926): p. 925-928.
15. Lin, Z.R., et al., Generation of quantum-dot cluster States with a superconducting
transmission line resonator. Phys Rev Lett, 2008. 101(23): p. 230501.
16. Chern, G.D., et al., Direct evidence of open ray orbits in a square two-dimensional
resonator of dye-doped polymers. Opt Lett, 2004. 29(14): p. 1674-6.
17. Kippenberg, T.J., S.M. Spillane, and K.J. Vahala, Kerr-nonlinearity optical parametric
oscillation in an ultrahigh-Q toroid microcavity. Phys Rev Lett, 2004. 93(8): p. 083904.
18. Kippenberg, T.J., et al., Purcell-factor-enhanced scattering from Si nanocrystals in an
optical microcavity. Phys Rev Lett, 2009. 103(2): p. 027406.
19. Bayindir, M. and E. Ozbay, Band-dropping via coupled photonic crystal waveguides. Opt
Express, 2002. 10(22): p. 1279-84.
20. Armani, A.M., et al., Label-free, single-molecule detection with optical microcavities.
Science, 2007. 317(5839): p. 783-787.
21. Oxborrow, M., Traceable 2-D Finite-Element Simulation of the Whispering-Gallery
Modes of Axisymmetric Electromagnetic Resonators. Microwave Theory and Techniques,
IEEE Transactions on, 2007. 55(6): p. 1209-1218.
22. Matsko, A.B. and V.S. Ilchenko, Optical resonators with whispering-gallery modes-part
I: basics. Selected Topics in Quantum Electronics, IEEE Journal of, 2006. 12(1): p. 3-14.
23. Gorodetsky, M.L., A.A. Savchenkov, and V.S. Ilchenko, Ultimate Q of optical
microsphere resonators. Opt Lett, 1996. 21(7): p. 453-5.
24. Little, B.E., J.P. Laine, and H.A. Haus, Analytic theory of coupling from tapered fibers
and half-blocks into microsphere resonators. Lightwave Technology, Journal of, 1999.
17(4): p. 704-715.
25. Abramowitz, M. and I.A. Stegun, Handbook of mathematical functions, with formulas,
graphs, and mathematical tables. 1965, New York,: Dover Publications. xiv, 1046 p.
26. Choi, H.S. and A.M. Armani, Thermally stable hybrid organic/inorganic resonant
cavities. Linear and Nonlinear Optics of Organic Materials Xi, 2011. 8113.
27. Vahala, K.J., Optical microcavities. Nature, 2003. 424(6950): p. 839-46.
28. Gorodetsky, M.L. and V.S. Ilchenko, High-Q Optical Whispering-Gallery
Microresonators - Precession Approach for Spherical Mode Analysis and Emission
Patterns with Prism Couplers. Optics Communications, 1994. 113(1-3): p. 133-143.
29. Tseng, S.M. and C.L. Chen, Side-Polished Fibers. Applied Optics, 1992. 31(18): p. 3438-
3447.
47

30. Ilchenko, V.S., X.S. Yao, and L. Maleki, Pigtailing the high-Q microsphere cavity: a
simple fiber coupler for optical whispering-gallery modes. Optics Letters, 1999. 24(11):
p. 723-725.
31. Sharma, A., J. Kompella, and P.K. Mishra, Analysis of Fiber Directional-Couplers and
Coupler Half-Blocks Using a New Simple-Model for Single-Mode Fibers. Journal of
Lightwave Technology, 1990. 8(2): p. 143-151.
32. Barber, P.W. and R.K. Chang, Optical effects associated with small particles. Advanced
series in applied physics. 1988, Singapore ; Teaneck, NJ, USA: World Scientific. xx, 336
p.
33. Choi, H.S., S. Ismail, and A.M. Armani, Studying polymer thin films with hybrid optical
microcavities. Optics Letters, 2011. 36(11): p. 2152-2154.
34. Min, B., et al., Ultralow threshold on-chip microcavity nanocrystal quantum dot lasers.
Applied Physics Letters, 2006. 89(19): p. 191124.
35. He, L.N., et al., Detecting single viruses and nanoparticles using whispering gallery
microlasers. Nature Nanotechnology, 2011. 6(7): p. 428-432.
36. Kippenberg, T.J., R. Holzwarth, and S.A. Diddams, Microresonator-Based Optical
Frequency Combs. Science, 2011. 332(6029): p. 555-559.
37. Ilchenko, V.S. and A.B. Matsko, Optical Resonators With Whispering-Gallery Modes —
Part II: Applications. IEEE Journal of Selected Topics in Quantum Electronics, 2006.
12(1): p. 15-32.
38. Hossein-Zadeh, M. and K.J. Vahala, Fiber-taper coupling to Whispering-Gallery modes
of fluidic resonators embedded in a liquid medium. Opt Express, 2006. 14(22): p. 10800-
10.
39. Aoki, T., et al., Observation of strong coupling between one atom and a monolithic
microresonator. Nature, 2006. 443(7112): p. 671-4.
40. Rokhsari, H. and K.J. Vahala, Ultralow loss, high Q, four port resonant couplers for
quantum optics and photonics. Phys Rev Lett, 2004. 92(25 Pt 1): p. 253905.
41. Qian, S.X. and R.K. Chang, Multiorder stokes emission from micrometer-size droplets.
Phys Rev Lett, 1986. 56(9): p. 926-929.
42. Checoury, X., et al., Deterministic measurement of the Purcell factor in microcavities
through Raman emission. Physical Review A, 2010. 81(3).
43. Kippenberg, T.J., S.M. Spillane, and K.J. Vahala, Demonstration of ultra-high-Q small
mode volume toroid microcavities on a chip. Applied Physics Letters, 2004. 85(25): p.
6113-6115.
44. Spillane, S.M., Fiber-coupled ultra-high-Q microresonators for nonlinear and quantum
optics. 2004, California Institute of Technology: Ann Arbor. p. 143-143 p.
45. Kippenberg, T.J., et al., Ultralow-threshold microcavity Raman laser on a
microelectronic chip. Optics Letters, 2004. 29(11): p. 1224-1226.
46. Yang, L., et al., Ultra-low threshold Microlasers on a Si chip by Solgel Method. 2004.
47. Maker, A.J. and A.M. Armani, Heterodyned toroidal microlaser sensor. Applied Physics
Letters, 2013. 103(12).
48. Raman Spectroscopy. 01/25/2007; Available from:
http://web.nmsu.edu/~snsm/classes/chem435/Lab15/.
49. Soteropulos, C.E., H.K. Hunt, and A.M. Armani, Determination of binding kinetics using
whispering gallery mode microcavities. Applied Physics Letters, 2011. 99(10).
48

50. Armani, A.M. and K.J. Vahala, Heavy water detection using ultra-high-Q microcavities.
Optics Letters, 2006. 31(12): p. 1896-1898.
51. Mehrabani, S., A. Maker, and A. Armani, Hybrid Integrated Label-Free Chemical and
Biological Sensors. Sensors. 14(4): p. 5890-5928.
52. Hunt, H.K. and A.M. Armani, Label-free biological and chemical sensors. Nanoscale,
2010. 2(9): p. 1544-1559.
53. Mehrabani, S., et al., Hybrid microcavity humidity sensor. Applied Physics Letters, 2013.
102(24).
54. Hsieh, B.Y., et al., All-Optical Generation and Detection of Acoustic Waves for
Intravascular Ultrasound and Photoacoustic Imaging. 2011 Ieee International
Ultrasonics Symposium (Ius), 2012: p. 1175-1178.
55. Maxwell, A., et al., Polymer microring resonators for high-frequency ultrasound
detection and imaging. Ieee Journal of Selected Topics in Quantum Electronics, 2008.
14(1): p. 191-197.
56. Xie, Z.X., et al., Pure optical photoacoustic microscopy. Optics Express, 2011. 19(10): p.
9027-9034.

49

Chapter 3 Effects of Flow on Whispering Gallery Mode Microresonator
Sensors
3.1 Overview and Motivation
Optical cavity microresonators have been shown to detect temperature changes [1-3],
humidity changes [4], as well as UV changes [5]. An important application for such sensitive
devices is biosensing, which involves a fluid environment [6]. The transient response of an
optical microcavity biosensor will depend strongly on the flow conditions to which it is exposed.
High liquid flow rates lead to faster adsorption and a more dense coverage of molecules. This is
because the boundary layer shrinks as flow rates increase [7]. Therefore, it was of interest to
study the effects that flow has on microresonators during detection.
Since the evanescent field tail decays exponentially with distance, the molecules that
adsorbed to the device will be the only ones detected by the device [8, 9]. As molecules approach
the surface, the effective refractive index changes and thus the environment surrounding the
sphere is different. The change in the refractive index causes a detectable resonant frequency
shifts. By tracking the shift the initial binding rate of the molecules to the device surface can be
measured, effectively creating a sensor. Since the boundary layer changes around different
shapes, the experiment was to be performed on two resonator geometries: the toroid and
microsphere. It will be shown experimentally how the initial binding rate changes with different
injection speeds [10].
3.2 Sample Holder Design for Flow Experiments
The flow cell is designed to integrate with the existing testing setup and study how fluid
flow changes the detection signal. Using an existing steel sample holder, two carefully cut glass
slides are glued down with super glue to start the chamber. The first one is used to suspend the
50

sphere above the steel surface and the second serves to separate the top cover slide from the
device. The fiber tail of the microsphere is attached to the lower piece of glass with double sided
carbon tape. The carbon tape is then covered by a strip of clear tape. This prevents any of the
glue from the tape from ending up in the water or buffer while at the same time ensuring that the
sphere won’t move. Making sure the tape is not exposed is important, since contaminants could
be introduced into the solution and produce false results in detection. The chamber holds about
350 L of liquid. The sides and front of the chamber are left open for excess fluid to escape from
and for the tapered optical fiber to pass through.
The holder has a hole made to fit a 23-gauge injection tube, which has an outer diameter
of 0.0253 inches and an inner one of 0.0173 inches. One end of the tube is covered with tubing
that will deliver the solution of interest. The other end of the tube is inserted through the hole in
the steel and bent ~90 degrees to secure it in place. The injection point is positioned underneath
the sphere to prevent significant fluctuations in the optical fiber waveguide, which is placed
above the sphere. The tube and tubing are filled with solution prior to attachment to the sample
holder. The other end of the tubing is attached to a syringe filled with the solution. After
assembly, the sample holder can be attached to the nanostage that controls the position of the
sphere in three axes. I then bring the sphere close to the taper, making use of the microscope
view from above. The cover slip is placed on last and is secured with super glue. One source of
contaminants in the water could be from the glue, therefore, it is vital that the glue is not exposed
to the water. To accomplish this, only a small droplet is used to hold the glass together such that
none of the glue leaks out on the sides. The detailed sample holder with its constituents is shown
in Figure 3-1. The microsphere is difficult to see, but appears as a thin line seen through the
cover slide. The rendering of the same stage is shown in Figure 3-2.
51

Figure 3-1 Picture of the steel sample holder with all the parts assembled for testing in water. The
glass slides are used to suspend the microsphere above the steel surface and the cover slide above
the sphere and taper. The injection needle is placed underneath the sphere and bent to avoid
movement. Tubing is connected to the injection tube. The tape in the diagram wraps around the
sample holder but during testing runs, only enough tape to cover the carbon tape was used.

Figure 3-2 Rendering of the liquid chamber setup for microsphere detection testing. The placement
of the taper is shown. The components are not to scale relevant to each other in the image.
When testing toroids, a sample holder without a hole is used. The wafer with the toroids
is glued to the surface of the holder and the injection needle is placed at the front of the chamber
52

close to the edge of the wafer. The taper is positioned on the opposite side of the toroid to
minimize fluctuations.
3.3 Solution Preparation
After assembling the sample holder and aligning it in place with the optical fiber taper,
the chamber is ready for testing. The chamber is then filled with hydroxyethyl
piperazineethanesulfonic acid (HEPES) buffer through the tubing using a syringe pump. The use
of the syringe pump helps to avoid unwanted bubbles in the chamber. It is also important to
ensure that the tubing and the syringe have no bubbles either. HEPES is a zwitterionic buffer,
meaning it is composed of neutral molecules that have a positive and a negative electrical charge
at different locations within that molecule [11]. Because buffers can interconvert between acidic
and basic forms, donating or accepting protons as needed, HEPES maintains physiological pH
despite changes of dissolved CO
2
[12]. There are other buffers that are ideal for biochemical
testing, such as those known as Good’s buffers [13]. The list includes buffers with desirable
values of parameters such as their pKa number, their solubility, their influence on dissociation,
among others. HEPEs is in the middle range of pKa, or the acid dissociation constant, with a
value of 7.55. The pKa value reveals the extent of dissociation of molecules, or the release of
hydrogen atoms from the molecules. The larger the pKa value, the more dissociation occurs. The
pKa value can be thought of as the pH at which the number of molecules of conjugate base and
weak acid are equal.
To make the buffer, standard techniques are used [14]. HEPES powder (minimum 95%
titration – Sigma Aldrich) and sodium chloride (crystal – JT Baker) are dissolved in millipore-
filtered water to make 50 mM HEPES and 140 mM sodium chloride. The pH is monitored with a
pH meter and 1M sodium hydroxide is added drop-wise to bring the solution to a pH of about
53

7.5. These conditions mimic the physiological environment that proteins need to avoid
denaturation. The prepared buffer is then used to make the desired dilutions of the target protein.
The choice of protein came from its convenience. A protein that had high-affinity specific
interactions, one that was cheap, and readily available in high purity was needed. Streptavidin
fits all of these criteria. Streptavidin is found in the Streptomyces avidinii bacterium and can be
extracted and purified [15]. This protein has a high affinity for vitamin B7, also called biotin.
Their non-covalent interaction is one of the strongest in biology, with a dissociation constant on
the order of 10
-14
mol/L [16]. This pair is frequently used for specific detection [17]. The surface
of the detector can be functionalized with biotin groups, which will selectively bind streptavidin
molecules from a complex solution [10].
The storage of the protein is very important, as it is easily denatured. After receiving
streptavidin in powder form, enough Milli-Q water is added to reconstitute it to 0.2 mM
concentration. The buffer is then cleaned through vacuum filtration to get rid of larger debris.
Most of the protein solution is added to the correct amount of buffer to make a 20 M solution.
The solution is separated into 5 L aliquots that are frozen and stored until they are used. It is
best to keep the protein frozen to extend its shelf-life. While the freezing and defrosting process
will result in some denaturation, the majority of the protein in solution will maintain its viability.
Repeated freezing and thawing will damage the protein more. Another issue with protein storage
is its affinity for container walls. The protein will adsorb to the surface of the container and in
the process may denture. For this reason as well, it is best to store small frozen aliquots, rather
than large liquid solutions.
The protein solution for testing must be made on the same day, and any excess is not
stored for more than 24 hours. When needed, an aliquot is removed from the freezer and allowed
54

to melt. It is then centrifuged for ten seconds to get all the liquid to the bottom of the tube. One
microliter is diluted with filtered buffer to make a 10 nM concentration. Then this dilution is
used to make the smaller concentrations of 1-10 pM for testing.
The HEPES buffer to be used for the day is filtered with the use of vacuum filtration.
Only then can it be used to dilute the streptavidin aliquot to the desired concentration. HEPES is
stored at room temperature, but the protein solutions are refrigerated if they are to be used the
following day. It is important to use the same buffer for the experiment and for the dilution, since
slight changes in pH will alter the salt compositions in the buffer and change the refractive index
around the device. If different buffers are used, the data cannot be directly compared.
3.4 Initial Experiments
In initial experiments, all variables were kept as constant as possible. These variables
included input power, percent power coupled into the device, the quality factor, device size, and
protein concentration. Only the flow rate was varied. The laser centered at 765 nm was used to
excite the cavity. Even though the loss through absorption is higher at this wavelength in water,
it is easier to control and maintain the input power.
To record data, it is necessary to track a single resonance peak and how it moves in both
the time and transmission axes of the oscilloscope. Recording both axes gives a better sense of
how the resonator was coupling to the taper and outliers can be easily spotted and cropped out.
With microspheres, it is more difficult to look at one resonance at a time, since the spherical
shape is able to support a lot of modes and the taper is capable of coupling into many of them.
The fundamental peak had to be chosen;. This peak is identified as the one that dips lowest in
transmission and is the most isolated in the resonance spectrum. As mentioned in the background
chapter, a LabView program written by Matthew Reddick is used to track the minimum. While a
55

typical transmission spectrum will have multiple resonances on the screen, only the global
minimum point will be recorded over time.
Both x and y data corresponding to the minimum are stored continually. The y data, or
transmission, is recorded as a reference to make sure that the coupling is constant. The x-axis
data gives the shift of resonance in the time axis of the oscilloscope. Since the slope of the
function generator signal sent to the laser is known, the time axis can be converted to
wavelength. A data cropping program also written by Matthew Reddick is used to exclude
outlier points due to changes in coupling. Depending on the coupling regime, the shape of the
resonance peak can change and produce a perceived shift that is not relevant to the detection.
The program interface is shown in Figure 3-3 with data from both axes. The left y-axis displays
the values for Lambda (the black line), which is the resonance shift, or the time axis changes
already converted to nanometers. The y-axis on the right shows the Intensity, or the value on the
transmission axis (the red line).
Cropping is allowed only on the y-axis (transmission) data. By tracking the coupling
efficiency in the y-axis, the errors from perceived shift due to coupling changes can be
eliminated. In Figure 3-3, the intensity data is fairly constant throughout the time range. A large
change can be seen in the intensity and the wavelength shift data at around 1500 seconds. The
green bars shown in the figure can be moved to selectively crop data that lies outside the
horizontal lines but within the vertical ones. When cropping the transmission data, the
corresponding lambda data is immediately updated. It is easy to see how the shift data is
affected, since the same time points are removed from both data sets.
56

Figure 3-3 Data recorded with minimum tracking program shown in the data cropping program.
The black line is the resonance shift (Lambda axis) and the red line is the transmission data
(Intensity axis). Both are plotted as a function of time.
The shift in resonant wavelength over time gives information on the initial binding rate,
the equilibrium point at which adsorption is equal to desorption, and the rate of dissociation. The
first experiment was performed with a 1 pM solution of streptavidin in HEPES buffer. The
wavelength shift response for different flow rates was recorded. The results for five different
flow rates are shown in Figure 3-4. The initial slope and the overall shift increase with flow rate,
as one would expect by considering the behavior of the boundary layer. At higher flow rates, the
boundary layer is smaller and solution is able to get to the surface faster. The initial binding rate
can be estimated from the slope of the first few seconds of injection. The equilibrium point,
where the rates of adsorption and desorption are the same can be found by extrapolating the data
to see at which point the shift levels off (not shown). The dissociation constants can be
calculated from fitting a theoretical curve to the part of the graph right after flow is turned off
(not shown). The diameter of the sphere was 117 m and the quality factor was 3.46x10
6
.
57

Figure 3-4 Resonance shift data obtained for different flow rates, showing where the flow was
started and where it was shut off. All flow rates are in L/min.
To verify that the shift is due to protein adsorption, the experiment was repeated with
HEPES buffer alone. Previously, buffer did not seem to produce a noticeable shift, but was only
tested at lower flow rates. When no shift was observed from buffer, it was assumed the protein
binding was responsible for the shift in Figure 3-4. However, upon repeating the experiment at
high flow rates, a significant effect due to flow alone was seen. As can be seen in Figure 3-5, the
overall shifts were even greater than those seen in Figure 3-4, when protein was injected. The
sphere diameter in this case was different at 177 m, as was the Q at 8.2x10
5
. In other words, the
diameter was approximately 50% larger and the Q was about five times smaller. Therefore, the
absolute shift values should not be directly compared. Nevertheless, the shift from buffer is not
negligible.
58

Figure 3-5 Shift in wavelength vs. time for injection of HEPES buffer into HEPES buffer without
any protein. The overall shift in picometers is greater than the shifts seen when protein was injected
in the same time range. Flow rates are in L/min.
For example, even for experiments of water injected into water, the shift was large at
higher flow rates (Figure 3-6).

Figure 3-6 Shift response from water injected into water.
59

The contribution to the signal due to the fluid flow must be isolated and subtracted from
the overall signal in order to draw any conclusions from this data, which required an
experimental setup change.
3.5 Two-Syringe Method
To introduce two solutions into the flow cell, two separate syringe pumps featuring
digitally-controlled and programmable stepper motors were used. The first solution was just
HEPES buffer, while the second was a streptavidin solution with 1 nM concentration in the same
HEPES buffer. The tubing from each syringe was connected to a T-shaped 23-gauge splitter to
combine the flows in the arrangement shown in Figure 3-7.
The pumps were first tested to see if they behaved the same. To do this, the same buffer
was used in both syringes, and the flow was switched from one to the other. To begin, the flow
of one pump is started and continued until the resonance shift is stabilized. Then, simultaneously,
the first pump is turned off and the second pump is turned on at the same flow rate, in the initial
case 100 L/min. The microsphere should thus be exposed to a constant flow rate. This
experiment showed that there is a small delay when switching between pumps, seen as small
blue-shifts (shifts to lower wavelength) in the data. However, the shift returned to the same place
within seconds. Therefore, the pumps operate in the same way since there was no extreme shift
of the resonance wavelength when switching from one to the other.
The next set of experiments was with different flow rates, but still using only buffer. The
overall initial steady state shift increased with flow rate as expected. However, the effect of
switching between syringe pumps was still just a blue shift and didn't vary much with flow rate.
Finally, one of the buffer syringes was replaced with one that contained a 1 nM solution
of streptavidin in the same buffer. The experiment consisted of switching from one solution to
60

the other while pumping at the same rate. First, the buffer was injected until the sensor reached a
new steady state resonance wavelength. At this point, flow of pure buffer was turned off while
flow of the protein solution was simultaneously turned on. The switch resulted in a dip in shift
wavelength, since the short delay of flow allowed the resonance to slightly blue-shift.
Nevertheless, it was possible to distinguish between buffer and protein injections.

Figure 3-7 Two pump experiment flow layout. One solution was injected at a time, but the overall
flow rate was maintained constant.
Figure 3-8 shows how the shift changed when alternating between solutions. Figure 3-8a
shows the results when buffer was injected first (red arrow), indicated by the ON in the graph.
Once the flow stabilized, the flow was switched to the protein solution (blue arrow). The flow
was switched three more time until it was turned off, as indicated by the OFF on the graph.
Figure 3-8b is different in that the protein solution was injected first, then the flow was switched
to the buffer solution. The switching was done three additional times as well. The blue and red
arrows indicate the protein and buffer solutions start time in each graph. The inserts show
zoomed in views of the four injections after the initial steady state was reached. The flow rate
was set as 250 L/min for both pumps.
61

Figure 3-8 Two pump experiment with buffer and protein showing effects on the resonant shift
when the flow was switched from one to the other. Red arrows indicate when the protein solution
was turned on and blue arrows indicate the buffer solution.
From the graphs, it is apparent that there is a difference between buffer and protein in the
plots. Whenever protein is injected, the shift increases and when the flow is switched to buffer,
the shift decreases again. The largest source of error in these experiments is the blue shift that
occurs when the flow is switched. Because of the blue shifts, it is difficult to compare the effects
directly on the graph.
3.6 Perturbation Flow Experiments
To further distinguish flow effects and to eliminate the slight delay resulting from
switching between syringe pumps (which causes the blue shift), a perturbation flow
configuration was used instead. Using the same configuration as in Figure 3-7, the protein
solution was to be added to an already flowing buffer solution as a perturbation. The experiment
required two syringe pumps again for the buffer and the protein solution. The buffer was injected
until there was no more shift in resonance, at which point the protein solution was started at 1%
flow rate of the buffer. Since the protein solution was to be flowed at 1/100
th
of the flow rate of
62

buffer, the concentration was made to be 100 times bigger (e.g. 100 pM instead of 1 pM). The
experiment involved flowing buffer until the resonance peak stabilized, then introducing protein
flow to possibly detect molecule adsorption.
The buffer flow was never stopped during the experiment and only the protein solution
was stopped and restarted. The experiment was performed at a variety of flow rates, such as 100,
250, 500, and 1000 L/min for the buffer solution. These experiments did not produce any
noticeable changes from pure buffer injection and addition of protein flow. It is possible that the
high velocity buffer was pushing into the protein section of the tubing too much for the protein to
be able to add to the flow.
The arrangement of the flow channels was altered to better allow the protein solution
access to the flow cell. A T-splitter was again used, but the flow was designed such that the
buffer had to make a 90 degree turn when flowing while the protein solution flowed straight
when it was turned on (shown in Figure 3-10). The results from this experiment at two flow rates
are shown in Figure 3-10. There was no apparent additional shift due to the protein solution. It
may be possible that the protein was being leached into the buffer injection due to the pressure
gradient and large flow sweeping past the entrance point. One possible way to guard against this
is to install a valve on the protein side to effectively shut off access to it until it needs to be
added.
63

Figure 3-9 Two pump perturbation flow layout. Te buffer solution was injected continuously while
the protein solution flow was added as a 1% perturbation.

Figure 3-10 Perturbation flow experiment results. The 1% perturbation of protein flow did not
produce any major changes in the shift diagram. Here, the red arrows refer to addition of protein,
while blue arrows indicate when only buffer was flowing. a) Results for a buffer flow rate of 1000
L/min where protein was added at 10 L/min. b) Results for a flow rate of buffer at 250 L/min
with an addition of protein at 2.5 L/min.
64

3.7 Checking the Repeatability of Wavelength Shift
Without changing the layout of the tubing and syringes, an experiment was performed to
see whether there was repeatability in the initial wavelength shift data throughout multiple
injections. Buffer was injected for one minute at 100 L/min. Then the flow was stopped and the
resonance was allowed to shift back to the original position. The injection was repeated three
additional times while the data was saved continuously. After the resonance returned to the
starting position after the fourth buffer injection, a 100 pM protein solution was injected at 100
L/min for the same amount of time (one minute). Again, after the flow was turned off, the
resonance was allowed to shift back and a second protein injection was performed.
The data for these injections is shown in Figure 3-11a, where the buffer and protein
injections are marked. The interesting thing to note on this graph is the similarity between the
four injections of buffer, and how different these are from the ones resulting from protein. The
red lines are fits to the linear part of the binding curves.

Figure 3-11 Injection of buffer or protein, as noted, at 100 L/min. The red lines are fits to the
initial linear part of the resonance shift curves. a) Four injections of buffer followed by two protein
injections. b) Four protein injections followed by two buffer injections.
65

The second part of the experiment, shown in Figure 3-11b, consisted of four protein
injections in a row, with time for the system to recover between each one, and two buffer
injections. It is important to note that after the protein injection, the resonance did not return to
the original position, possibly signifying that molecules were still adsorbed to the surface of the
sphere. Another strange effect is the high shift in resonance that resulted from injecting buffer
after the protein injections. The environment around the sphere was drastically changed with this
injection, signifying that either proteins were on the surface or close to the surface of the device.
The values for the slopes in Figure 3-11 are shown in Figure 3-12. It is easy to see that
the four initial injections of buffer had very similar slopes, shown by the black squares. The
protein slopes varied a bit more than the buffer slopes, but did not show dramatic changes (red
circles). The two points for the 5th and 6th injection of buffer differ greatly from the rest of the
buffer slopes, signifying that the sphere’s environment was changed drastically .

Figure 3-12 Initial slopes of the binding curves for buffer and protein injections.
It is important to separate the buffer and protein solutions and perform the experiments
with one solution at a time. It also seems that between protein solutions, it would be useful to
66

flush the chamber with buffer to remove any remaining protein in solution and hopefully help
clear off the surface of the sphere. A series of experiments were performed with multiple
injections of each solution at various flow rates. The experiments were repeated for three
different sizes of microspheres and two sizes of toroids to determine how the device size affects
the results. The details are shown in the following section.
3.8 Comparing Initial Slopes of Buffer and Protein Injections
The experiment was changed to further separate buffer flow from protein flow without
any mixing. Two syringe pumps were used again, but only one liquid at a time was injected into
the testing chamber. The goal was to measure the initial slope of the resonance shift curve and
compare it for the different liquids and flow rates. The surface of the device was also modified to
make it more sensitive to molecule adsorption. The reflowed devices were treated with oxygen
plasma to form hydroxyl groups then exposed to APTMS vapor under vacuum for 15 minutes to
form silane groups on the surface. The shift from treated devices was much larger for protein
injections than from untreated devices. However, this was tested at higher streptavidin
concentrations (~1 nM). The concentration used for the following experiments was lower at 10
pM.
The experiments were first performed by flowing buffer three to four times at each flow
rate, followed by protein solution flows. The buffer injections were performed by flowing the
fluid for 30-60 seconds, depending on the flow rate, followed by a three minute waiting time for
the resonance to shift back before injecting again. The buffer was injected for different amounts
of time, since at higher flow rates, the initial linear part of the shift changes to steady state faster
than at slower flow rates. The waiting time was later changed to four minutes since three was not
always enough to return exactly to the initial condition.
67

To switch to protein flow, the same tubing is used as for buffer, to keep the flow length
the same. The luer tip is switched from one syringe to another. Then enough protein solution is
pushed though to clear out the buffer from the tubing. The aqueous chamber where the sphere is
located is then flushed with buffer to get rid of any protein molecules that might have ended up
there from filling the tubing. After every injection of protein, the chamber was flushed with
buffer again to clear it out.
One problem with the experiment was that over time, the coupling condition or the power
through the taper changes. To address this issue, the experiment was performed in a slightly
different order. Instead of completing all the buffer injections first, the injections were alternated
between buffer and protein. For example, first buffer was injected three to four times at one flow
rate, such as 100 L/min, then the tubing was attached to the protein syringe and the protein
solution was injected three to four times at the same flow rate. The chamber was still flushed
with buffer between each injection. The biggest issue with this method of testing is the high
possibility of leaving behind some protein in the chamber or tubing after its injection. This could
skew the results.
The few representative results of the experiment with alternating buffer/protein flows are
shown below. Two parameters are considered in these data sets: the total shift in resonance, or
the steady state point after injection is started, and the initial slope of the resonance shift curve.
Sometimes at lower flow rates (100-250 L/min), the protein solution produces larger numbers
in both parameters (Figure 3-13a). The overall resonance shift is larger for protein here, and the
initial slope, shown as the purple line, is steeper. For comparison, the slope of the protein line is
1.12 and the one for buffer is 0.516 pm/sec. At these lower flow rates the data is not consistent.
Even within the same data sets there are injections that contradict this behavior in that the buffer
68

slope and shift are larger than protein. Therefore, an average of the slopes is used to compare
data. All four injection slopes are calculated and averaged at every flow rate for protein and for
buffer.
At higher flow rates the buffer solution always dominates over the protein. The data is
usually much more consistent. At a flow rate of 1000 L/min, the buffer shift is much higher and
the initial slope of the buffer curve, the blue line in Figure 3-13b, is steeper. The slopes are 8.92
and 4.26 pm/sec for buffer and protein, respectively. The saturation is reached much faster at
higher flow rates as can be noted in Figure 3-13. At the low flow rate, it takes about 5 seconds to
cover the linear part of the curve and over 20 seconds to begin to see saturation. On the other
hand, at the high flow rate, the linear part makes up a fraction of a second and saturation can be
noticed within three seconds. The overall shift is larger for both solutions at higher flow rates, as
expected from previous experiments.

Figure 3-13 a) At lower flow rates (250 L/min), the protein solution sometimes produces a larger
shift and bigger initial slope. b) At higher flow rates (1000 L/min), the buffer flow always has the
bigger effect. The blue and purple lines are the linear fits to the initial part of the shift.
69

Different size spheres were tested to see if the size has an effect on the slopes. As
mentioned, the slopes are averaged at every flow rate. To display the results, the difference
between the slopes is plotted as a function of flow rate, chosen as buffer minus protein slopes.
The results for two sizes of spheres (~120 and 190 m diameter) are shown in Figure 3-14. The
two sets of data lie in two distinct regions of the plot. The larger spheres show a larger difference
in the slopes than the smaller spheres. However, not every sphere size is consistent with the trend
(the 121 m sphere line is below 118 and 117 m), though this could be accounted for by
experimental error. Larger spheres have a larger mode volume and a larger evanescent field,
which would imply greater interaction with the environment. Possibly, the bigger spheres are
more sensitive to buffer even when protein is present.

Figure 3-14 The difference between buffer and protein initial slopes of the wavelength shift curve
versus flow rate for two sphere sizes.
The experiment was also performed on toroids of two sizes. Kelvin Kuo took the data and
I analyzed it. The major radius of the toroids was kept constant at 40 m, but the minor radius
was changed from about 2-4 m to 8 m. For the small minor-radius toroids, the difference in
shift slopes was smaller than for the big minor-radius toroids (Figure 3-15). The difference in
70

slopes was much smaller for the 4 m radius toroids than it was for even the 120 m diameter
spheres. The 8 m radius toroid results were comparable to the 120 m diameter spheres results.
The resonance shifts for the toroids were smaller than for the spheres. The toroid mode volume is
much smaller than that of spheres, which could account for the differences.

Figure 3-15 Difference between buffer and protein initial slopes for two sizes of toroids. All
sizes are in microns.
An additional higher concentration solution was tested with two spheres and analyzed to
add to the data previously taken with the 10 pM concentrations. A 1 nM solution was used as the
new concentration to see if a different response could be detected. The spheres were tested in the
same way as before, alternating between buffer and protein solutions at different flow rates. A
line was fit to the initial resonance shift. The averaged slope of the protein injections was
subtracted from that of the buffer. Adding the data to the plot of previous results, it can be seen
that for two sizes of spheres, the new lines appear above the lower concentration ones, as shown
in Figure 3-16. The two new lines are the blue and red half-colored stars and lie above the other
data points.
71

Figure 3-16 Additional data for a higher (1 nM) concentration of protein (blue and red half-colored
star data points).
The results suggest that with the higher concentration of protein, an even bigger effect is
seen on the resonance from the buffer alone. This would mean that the protein is not coming in
contact with the surface as much as the lower concentration protein does. This suggests that
detection is harder with higher concentrations of protein, which is very counter-intuitive.
Let's take a closer look at the way the data is analyzed. Figure 3-17 shows a set of data
for buffer for flow rates of 100 L/min (a) and 500 L/min (b). The plots show multiple linear
fits of the initial slope. The blue line in Figure 3-17a is fit to the most amount of data in the curve
and has a slope of 0.323 pm/sec. The red line is fit to a smaller region of the curve and the slope
for this line increases to 0.458 pm/sec. Finally, the cyan line is fit to the initial, steepest part of
the curve with even fewer points, giving a slope of 0.564 pm/sec. As can be seen, fitting the line
to different parts of the curve drastically changes the results. Similarly, in Figure 3-17b, the blue
line is fit to more data and gives a slope of 8.71 pm/sec, while the red line is fit to the steepest
part and gives a slope of 10.64 pm/sec.
72

Figure 3-17 The data for wavelength shifts resulting from buffer injections at two flow rates. a)
Wavelength shift resulting from 100 L/min flow rate. The three lines fit to the initial part of the
curve show very different slopes. b) Wavelength shift resulting from 500 L/min flow rate. Fewer
data points appear at the higher flow rate because the resonance shifts much faster. The slope still
changes if fewer or more points are added in the fit.
The same analysis is shown for protein in Figure 3-18. Again, it can be seen that the
slope changes a lot depending on how many points are used in the fit. In Figure 3-18a, the slopes
of the 100 L/min lines are 0.097 (blue), 0.116 (red), and 0.147 pm/sec (cyan). In Figure 3-18b,
the slopes of the 500 L/min lines are 1.21 (blue) and 1.99 pm/sec (red).
73

Figure 3-18 The data for protein shifts at two flow rates showing the linear fits to the initial slope.
a) Shift for a 100 L/min flow rate of protein solution. b) Shift for a from 500 L/min flow rate.
Considering the large difference in results if the initial slope is fit to more or fewer data
points, it was decided that convincing data could not be acquired to make any conclusive claims
about the effects of flow on these devices.
References

1. Choi, H.S., et al., Thermo-optic Coefficient of Polyisobutylene Ultrathin Films Measured
with Integrated Photonic Devices. Langmuir, 2012. 28(1): p. 849-854.
2. Choi, H.S., S. Ismail, and A.M. Armani, Studying polymer thin films with hybrid optical
microcavities. Optics Letters, 2011. 36(11): p. 2152-2154.
3. Choi, H.S. and A.M. Armani, Thermal nonlinear effects in hybrid optical
microresonators. Applied Physics Letters, 2010. 97(22).
4. Mehrabani, S., et al., Hybrid microcavity humidity sensor. Applied Physics Letters, 2013.
102(24).
5. Harker, A., S. Mehrabani, and A.M. Armani, Ultraviolet light detection using an optical
microcavity. Optics Letters, 2013. 38(17): p. 3422-3425.
6. Mehrabani, S., A. Maker, and A. Armani, Hybrid Integrated Label-Free Chemical and
Biological Sensors. Sensors. 14(4): p. 5890-5928.
7. Murthy, C.R. and A.M. Armani, Mass Transport Effects in Suspended Waveguide
Biosensors Integrated in Microfluidic Channels. Sensors, 2012. 12(11): p. 14327-14343.
8. Lopez-Yglesias, X., J.M. Gamba, and R.C. Flagan, The physics of extreme sensitivity in
whispering gallery mode optical biosensors. Journal of Applied Physics, 2012. 111(8).
9. Vollmer, F. and S. Arnold, Whispering-gallery-mode biosensing: label-free detection
down to single molecules. Nature Methods, 2008. 5(7): p. 591-596.
74

10. Soteropulos, C.E., H.K. Hunt, and A.M. Armani, Determination of binding kinetics using
whispering gallery mode microcavities. Applied Physics Letters, 2011. 99(10).
11. Bushey, M.M. and J.W. Jorgenson, Capillary Electrophoresis of Proteins in Buffers
Containing High-Concentrations of Zwitterionic Salts. Journal of Chromatography, 1989.
480: p. 301-310.
12. Shipman, C., Evaluation of 4-(2-Hydroxyethyl)-1-Piperazineethanesulfonic Acid (Hepes)
as a Tissue Culture Buffer. Proceedings of the Society for Experimental Biology and
Medicine, 1969. 130(1): p. 305-&.
13. Good, N.E., et al., Hydrogen Ion Buffers for Biological Research. Biochemistry, 1966.
5(2): p. 467-&.
14. Elving, P.J., J.M. Markowitz, and I. Rosenthal, Preparation of Buffer Systems of
Constant Ionic Strength. Analytical Chemistry, 1956. 28(7): p. 1179-1180.
15. Argarana, C.E., et al., Molecular-Cloning and Nucleotide-Sequence of the Streptavidin
Gene. Nucleic Acids Research, 1986. 14(4): p. 1871-1882.
16. Chilkoti, A. and P.S. Stayton, Molecular Origins of the Slow Streptavidin-Biotin
Dissociation Kinetics. Journal of the American Chemical Society, 1995. 117(43): p.
10622-10628.
17. Star, A., et al., Electronic detection of specific protein binding using nanotube FET
devices. Nano Letters, 2003. 3(4): p. 459-463.
75

Chapter 4 Raman Microlaser
4.1 Overview/Motivation
Lasing effects typically occur due to the doping of materials with rare earth metals, such
as erbium and ytterbium. Microlasers based on rare-earth doping [1, 2] and Raman [3-5] have
been demonstrated, but the emphasis of microcavity-based laser research has focused on lasing
in the near-IR and in air or vacuum. However, many biological or healthcare applications require
operation in biocompatible environments, such as water or buffer, in order to ensure the stability
of the biomolecule or process of interest. One example of a microlaser operating in a buffer has
been achieved with the use of rare-earth metals to increase the gain of the material [6]. Given the
high absorption of water in the near-IR, the lasing threshold will significantly increase because
the quality factor will decrease in this wavelength range [7]. Therefore, there is significant
interest in developing devices that operate closer to the visible range.
While there are numerous methods of fabricating a doped microlaser that emit in the
visible, most, if not all, of these dopants are not biocompatible and could interfere with the
biodetection experiments [8]. These dopants are also expensive and at times difficult to handle,
thus avoiding them is a very attractive alternative. Therefore, it is preferable to form a microlaser
without dopants. One approach is to use the inherent Raman gain of silica as the basis of the
laser [9-11]. However, because the Raman gain coefficient of silica fibers is low (approximately
10
−13
m/W), the threshold power necessary to achieve lasing is very high [12]. Therefore, by
combining the resonant amplification of a microcavity with the inherent biocompatibility of
silica, it is possible to create a low threshold, dopant-free microlaser-based biosensor. Since
silica is inherently biocompatible, the biggest challenge to produce the dopant-free laser-based
biosensor is achieving lasing in a biologically significant medium.
76

This chapter presents a cascaded Raman microlaser that operates in the visible range for
an undoped ultra-high Q silica microsphere in both air and buffer environments. To the best of
our knowledge, this is the first successful undoped silica Raman microlaser in this regime. Using
a tunable laser (Newport, San Jose, CA), the resonance is set at around 771 nm for the pump
wavelength. At the near visible wavelengths the absorption of water is significantly reduced,
decreasing the loss from the microcavity, and allowing the Q to remain high. With the ultra-high
Q, lasing in both air and buffer is presented with sub-milliwatt thresholds, an important advance
over previous methods that required dopants or were only achieved in air or vacuum
environments.
4.2 Testing Set-up
The general setup is the same as described in section 2.2.3. Not mentioned, but also
present, are two optical columns with cameras at the end used for side view and top view
alignment of the device and taper. Figure 4-1 shows the top view of the silica microsphere and a
tapered fiber as seen by the camera at different power inputs of 633 nm light. Figure 4-1a shows
the sphere and taper aligned but far enough away that no power couples into the sphere from the
taper. Figure 4-1b shows an under-coupled regime where a little light is coupled into the
microsphere. Figure 4-1c and d show the maximum coupling of light from the taper into the
microsphere with (c) and without (d) the light of the microscope camera.
77

Figure 4-1 Coupling light into a silica microsphere. a) No power coupled into sphere. b) Power
starting to couple into microsphere. c) and d) Maximum power coupled into microsphere with (c)
and without (d) the microscope light illuminating the device.
When testing the microsphere in air, its fiber tail is held in place by a fiber chuck, which
minimizes vibrations and allows for fine positioning with the piezoelectric stage. As mentioned,
the microspheres are fabricated from an optical fiber tip with a carbon dioxide (CO
2
) laser. The
surface tension in the molten silica produces a spherical shape that is atomically smooth,
resulting in negligible scattering loss due to surface roughness, again leading to an ultra-high Q.
The size of the fabricated device ranges from ~180-210 m from protective-layer-free optical
fibers. To reduce the size of the microspheres, the fiber is first tapered to a desired thickness by
heating it with a hydrogen torch and pulling apart with a motorized stage. If cut in the middle,
the two ends of the fiber will be much smaller than the original fiber and when reflowed with the
CO
2
laser, will produce smaller spheres.
Light is coupled into the microsphere by using a low loss, high efficiency tapered optical
fiber as the waveguide [13]. The fiber is pulled until its thickness is comparable to the
wavelength of light intended to pass through it. An external cavity tunable laser is used to couple
a) b)
c) d)
78

light into the tapered fiber and then the cavity. The output from the fiber is converted to an
electrical signal by a photodetector (PD) and the transmission is recorded and displayed by a
high speed digitizer/oscilloscope PCI card inside the computer.
The new component to the set-up is the spectrograph and its fiber tip to detect lasing
emission. A fiber coupled spectrograph (Andor Shamrock spectrograph SR-163 with Newton
CCD detector) was used for the experiments [14]. The entire setup was covered in blackout
curtains during measurements to minimize external light sources. The spectrograph fiber tip was
placed as close as possible to the device in each experiment. The background noise of the
spectrograph is at 600 counts, which could be subtracted from the spectrum measurements to set
the baseline at zero. The detector saturates at 65,000 counts. The grating of the spectrograph
(SR1-GRT-0300-760) has a resolution of 0.76 nm and can detect emissions from 420 to 1700
nm. The Andor Solis software was used to record emissions on the computer. Figure 4-2 shows
the layout of the testing set with the major components shown, including the spectrograph and its
tip.

Figure 4-2 Schematic of testing setup, showing the tapered optical fiber, laser, photodetector (PD),
and the spectrograph and its fiber, which is positioned near the microsphere [15].
4.3 Threshold Power for Raman Lasers
For microcavity based Raman lasers, the threshold power required for lasing is related to
the number of lasing lines, or the Stokes order number (N), by [11]:
79

(4.1)

(4.2)
where n
eff
is the effective refractive index, V
eff
is the effective optical mode volume, 
p
and 
R
are
the pump laser and the Raman lasing wavelengths,  is the overlap factor between the optical
mode and the Raman mode, B is a correction factor to account for backscattering within the
device, g is the Raman gain coefficient, Q
e
is extrinsic loss of the cavity, Q
T,R
is the total quality
factor of the cavity at the Raman wavelength and Q
T
is the loaded quality factor of the cavity [4].
There are several key features of this equation worthy of comment. First, as the
environment around the laser changes, the effective refractive index will change, resulting in a
shift in 
R
. This shift can provide a mechanism for biodetection, the primary motivation for the
present work. Additionally, for cavities larger than 30 m, V
eff
increases as the size of the cavity
increases. Therefore, the majority of the previous work using silica cavities has focused on small
diameter devices. However, when working in a low refractive index contrast environment, such
as water or buffer, the radiation loss increases, resulting in lower Q
P
values. Therefore, it is
necessary to balance these two competing effects [7].
Now let’s look at each term in turn to discuss how the threshold can be minimized. The
three Qs are related to the quality factor, which is an inverse of a sum of losses. Appearing in the
numerator, Q
e
is one such loss and represents the extrinsic cavity loss, or the loss due to
inefficient coupling between the waveguide and the resonator. Q
T,R
and Q
T
are quality factors
that account for intrinsic scattering, radiation, absorption, and contamination losses, and the
extrinsic coupling loss. Since all the Q terms in the equations depend on how the light is
80

introduced into the resonator, an efficient and low loss coupling technique is required to
minimize the threshold. Ideally, the coupling condition should be changed accordingly for each
lasing peak to find the absolute minimum threshold for that peak.
The effective refractive index, n
eff
, is found by summing the fraction of light traveling in
each region multiplied by the refractive index of that region. In this case the regions are the silica
resonator and the surrounding environment (either air or water). The refractive index of fused
silica is constant for a given wavelength; therefore, the environment will have the biggest impact
on n
eff
. As the resonator is transferred to liquid surroundings from air, the environmental
refractive index is increased, in turn increasing n
eff
, which will also increase the threshold
requirement.
The effective mode volume, V
eff
, has a similar effect on the threshold. V
eff
is highly
dependent on the difference in the refractive index of the environment and that of the device, as
will be shown with simulation results in the following subsection. V
eff
is also proportional to
D
11/6
, as shown by Eqn. (2.17), where D is the diameter of the sphere. Therefore, theoretically it
is better to have a smaller diameter device to minimize threshold. However, for smaller devices,
the loaded quality factor, Q
T
, is decreased, especially in an aqueous environment, due to an
increased radiation loss from a smaller microsphere. Figure 4-3 shows the wavelength
dependence of the absorption coefficient in water. The absorption is minimum at a wavelength of
500 nm. However, at 633 nm and even at 765 nm, the absorption is still relatively low. It is
important to note that the y-axis is on a log scale, making the absorption at 765 nm an order of
magnitude smaller than at 980 nm and almost 3 orders of magnitude smaller than at 1560 nm.
81

Figure 4-3 Absorption coefficient of water at different wavelengths [16].
4.4 Finite Element Method Mode Volume Simulations
To better understand the effect of refractive index on the mode volume and predict the
behavior of the microsphere, simulations were performed in COMSOL Multiphysics. The
software operates by the finite element method. The method operates by dividing a large domain
into small subdomains, or finite elements. Simple equations are then solved for boundary value
problems of differential equations at these elements. The combination of these solutions gives a
good approximation over the large complex domain.
To simulate silica microspheres, the methods developed by Oxborrow were utilized [17].
Because of the symmetry of the microspheres, only a small equatorial cross section just near the
boundary of the device needs to be considered as the simulation geometry . This is also because
the optical field does not penetrate very deeply into the microsphere, and even less into the
surrounding medium. Figure 4-4a shows a scanning electron microscope (SEM) image of a silica
microsphere. The blue square in the figure represents the region in which the simulation is
performed. The actual geometry parameters of a 200 m diameter microsphere are shown in
Figure 4-4b.
82

Figure 4-4 a) SEM image of a silica microsphere showing a blue square that represents the
simulation area. b) Simulation area of a 200 m diameter silica microsphere and its surroundings.
To create the finite elements within the geometry, a mesh is implemented. The mesh
divides the geometry into small parts over which the equations of interest are solved. In this case,
a free triangular mesh was used. The size distribution was specified for the silica and the
surroundings. The mesh was set to extra fine in the silica material, a setting which produces
maximum elements of 0.3 m size in the specified geometry. The sphere surroundings were
given a less stringent limitation on size, since the field does not penetrate far into the
surroundings and does not need to be solved for with such high precision far from the surface. As
can be seen in Figure 4-5, the mesh is much finer inside the silica, and less dense outside. The
mesh is also finer near the boundaries of the geometry and the interface between materials.

83

Figure 4-5 Free triangular mesh used in the simulation geometry.
After the mesh is defined and built, the simulation can be evaluated. The physics model
used was the axisymmetric weak form partial differential equation (PDE). The study was set to
find four eigenvalues near a defined value of resonance frequency. The resonance frequency is
related to the effective refractive index, n
eff
, to the radius of the device, R, and to the azimuthal
mode order, M, by the following equation:

(4.3)
The frequency can of course then be found from the wavelength using the speed of light. Setting
the defined resonance frequency close to its predicted value greatly lowers the computation time
used by the simulation.
The results for the four eigenvalues are usually the fundamental and first higher order
mode with TE and TM polarization. The results are shown in Figure 4-6. The TE and TM modes
can be differentiated by considering the direction of the radial component of the electric field.
For TE, the component points in the up/down direction in the simulation geometry, while for
TM, the component points left/right. Also, the TE mode extends further out of the cavity because
the field intensity is usually higher for this kind of polarization.
84

Figure 4-6 Simulation results for the first four eigenvalues found by the model. The units of the
mode are V/m
2
.
The fundamental TE mode is the one used for data analysis. The wavelength of interest is
chosen to match experimental conditions. The azimuthal mode order is changed accordingly.
Since the refractive index of silica changes with wavelength, the value was calculated from [18],
which uses established values of fused silica at different wavelengths [19, 20]. The refractive
index of the surroundings can then be varied to gather information about the mode under the
different conditions.
The results show that the mode propagates closer to the environment when the device is
surrounded by a higher refractive index medium, such as buffer, than it does in air (Figure 4-7a
and b). The normalized mode profile along the equator of the sphere as a function of radius in
Figure 4-7c shows that the mode shifts closer to the boundary between the device and
environment. The graph also shows that more of the mode intensity, called the evanescent field,
is present outside the device, which while helpful for biodetection, also produces more loss from
the cavity. These changes account for a larger V
eff
at a higher refractive index of the environment.
85

Figure 4-7 a-b). COMSOL simulation showing the cross section of the mode around the equator of
a 200 m diameter microsphere in air and in buffer, respectively. The mode shifts outward as the
device is placed in buffer rather than air. c). Normalized mode field intensity along the cut line
shown in (a) and (b). The vertical blue line represents the microsphere boundary. The mode shifts
closer to the boundary of the device and more of the field is propagating in the environment as well.
The mode size thus should also increase [15].
To further classify V
eff
changes, three sizes of spheres were considered with diameters of
180, 190, and 200 m, since fabrication of microspheres results in a range of sizes. The
refractive index of the environment was changed from 1.00, which is the index for air or
vacuum, to 1.3417, a range that includes the refractive index of the buffer used (HEPES, 1.341)
[21]. The effective mode volume was found by using Eqn. (2.16). Different definitions of V
Q

result in negligible difference when calculating the effective mode volume. As pictured in Figure
4-8, for each sphere size, the effective mode volume increases by the same trend. As the
86

refractive index of the environment is increased, V
eff
increases in a non-linear fashion, with the
biggest changes at higher indices. Also note that the mode volume increases significantly when
the size of the device is increased. Since the mode volume is significantly larger in buffer than
air, it is expected to have a large impact on the threshold of lasing, P
thres
.

Figure 4-8 COMSOL simulation results for effective mode volume as a function of the refractive
index of the environment for spheres of diameters a) 180, b) 190, and c) 200 m. The mode volume
increases as the size of the device is increased and has a significant dependence on the refractive
index around the device as well [15].
4.5 Raman Lasing in Air
Lasing can occur with any pump wavelength as long as the threshold requirement is met.
When using a single mode, tunable external cavity continuous wavelength 633nm laser that is
able to scan between 634 and 637 nm, lasing at 655 nm and 676 nm was achieved. Similar
experiments were performed using a 776 nm and 850 nm tunable laser as well, and lasing was
observed (Figure 4-9). For the 776 nm pump, five lasing peaks appeared at 806, 836, 865, 902,
and 931 nm. The emission in Figure 4-9b has structure that is below the resolution of the
spectrograph grating. This most likely indicated multi-mode lasing behavior. The structure
becomes more apparent in Figure 4-9c, where multiple peaks appear at each emission
wavelength. The pump for this data set was 850 nm and the lasing peaks appear at 880, 886, 912,
919, 926, 948, 955, and 962 nm.
87

Figure 4-9 Cascaded Raman lasing at three pump wavelengths in air. a) A pump of 633 nm results
in two Raman peaks. b) Pump of 771 nm gives five lasing peaks that potentially display multi-mode
lasing. c) Lasing emission with an 850 nm pump, clearly showing multi-mode behavior [15].
It is apparent that the first emission peak is not always a constant distance from the pump.
This could be due to the excitation of different vibrational or rotational modes of the silica. The
distance between the pump and subsequent peaks is the same, indicating that the peaks serve as
the new pump source to produce newer red-shifted peaks.
4.6 Threshold Measurement in Air
Because the threshold of lasing is determined by the power in the resonator, the input
power must be changed to find the point at which lasing begins. Since the O-scope displays the
signal as a voltage, a calibration curve had to be constructed to convert the signal to units of
power. This was done by recording the voltage reading on the O-scope, then directing the output
signal to a power meter instead of the photodetector. Recording the voltage at each power
produced the linear plot shown in Figure 4-10. With the help of this linear relation, it is much
easier to determine the power of the signal without measuring with the power meter at every
point.
88

Figure 4-10 Correlation between power and voltage seen on the o-scope.
Since the pump light is very bright, a background data scan is recorded and subtracted
from subsequent signal scans. This scan is recorded with the sphere far from the taper so that no
coupling occurs. Scanning over the range of the pump laser shows which resonances give lasing.
Usually lasing occurs with broadened resonance peaks. These peaks are ultra-high Q resonances
and often cause the sphere to vibrate (mechanical vibration), or heat through absorption
(thermal), causing the peak to shift (or broaden). This behavior is clearly visible on the O-scope
(Figure 4-11).
To regulate the coupling, the position of the microsphere with respect to the taper is
observed on the microscope and controlled with a three-axis piezo stage with step sizes of 100
nm. During quality factor measurements, the laser scan rate and scan range are optimized to
minimize linewidth ( ) distortion due to non-linear effects, such as heating or mechanical
vibrations. At low input powers and in the under-coupled regime, the intrinsic Q of the
resonators can be found. The Qs used in the present series of experiments are always above
1x10
7
. An example measurement is shown in Figure 4-16a.
89

Figure 4-11 A broadening resonance peak. The inset shows a zoomed in view with the same axes
labels.
For every emission spectrum, the resonance spectrum or the linewidth is also recorded.
During these measurements, the coupling was increased, and approached or met critical coupling
in many experiments. Additionally, the scan range and rate were changed to maximize the time
that the cavity spent on-resonance. As a result, significant broadening of the linewidth occurred
(Figure 4-16b). All experiments were performed multiple times using different resonant cavities,
and representative results are shown.
The lasing intensity and linewidth were recorded at different values of input power. The
dip in transmission of the resonance peak gave the fraction of the input power coupled into the
cavity. Therefore, the input power was calculated as the output power multiplied by this fraction.
Plotting the lasing intensity counts as a function of input power shows two distinct regions in the
graph: the spontaneous emission, where the slope is small, and the stimulated emission, the large
slope portion of the graph. Fitting a line to the stimulated emission part of the data, gives the
lasing threshold at the x-intercept. Data points above saturation of the detector were not included
in the fit.
90

A sample output of the spectrograph is shown in Figure 4-12a. The left most peak is the
pump and the ones to the right are lasing peaks. The peaks shown are the maximum achieved
lasing in this experiment. Figure 4-12b and c show the lasing intensity as a function of power
input for both lasing peaks. The red lines are fit to the stimulated emission section of the graph.
The x-intercepts are 0.27 mW and 0.28 mW, which give the threshold power required to achieve
lasing. As expected, the threshold is higher for the second peak, which takes more power to
excite.

Figure 4-12 a) Spectrograph output with the pump peak and two lasing peaks. b) Intensity vs.
power graph for the first lasing peak. c) Intensity vs. power graph for the second lasing peak.
The threshold graphs show that some intensity points stop increasing with input power.
The intensity was found as the maximum of the lasing peaks, but at a certain point, the maximum
would not increase while the width of the peak would. As is apparent in Figure 4-9, some peaks
show multi-mode behavior that the spectrograph cannot resolve. Because of the presence of
multi-mode peaks, the area under the peak, rather than the maximum was considered. For
experiments where the mode structure wasn’t as apparent, the analysis of the data did not
produce different results in the threshold.
Two data sets are presented here to show results of calculating lasing threshold from the
area under the peaks. In the first experiment, the intrinsic Q of the 197 m diameter microsphere
91

in air was 4.36x10
7
and coupling was maintained at about 65%. The lasing spectrum and
threshold graphs for operation in air are shown in Figure 4-13 (a)-(d).
The pump wavelength was set to 772 nm. The emission peaks appeared at 801, 831, and
864 nm, indicating Stokes Raman scattering. It is apparent that the detector has reached
saturation, as the top of the first lasing peak is flat. The threshold values for each peak were
calculated as 433, 491, and 685 W, respectively. As expected, the threshold increases for higher
oreder emission peaks. In Figure 4-13b and c, points above 0.9 mW were not included in the fit,
since at this point, the emission had saturated the detector and the intensity was no longer
accurate. It is important to note that the apparent saturation is the saturation of the spectrograph,
which occurs at 65,000 counts, and not the device.
Another thing to note is the decrease in slope of the threshold line in higher order lasing
peaks. This slope is the differential efficiency of the device and is predicted to decrease with
order, consistent with observation.
The slight curvature at the low power end of the curve might be explained by thermal
effects which distort the resonance lineshape. As more power is coupled into the sphere, it is able
to more easily and uniformly populate the higher energy vibrational/rotational states, thus
changing the threshold at each point in a non-linear fashion.

92

Figure 4-13 Lasing in air. a) Three emission peaks were detected by the spectrograph at 801, 831,
and 864 nm. b-d) Intensity as a function of power going into the sphere for all three lasing lines.
The threshold is found at the x-intercept when fitting a line to the stimulated emission portion of
the graph. b) The first lasing peak shows a threshold of 432 W. The apparent saturation at input
powers greater than 0.9 mW is due to the saturation of the spectrograph. c) The second lasing peak
at 831 nm has a threshold of 491 W. d) Third lasing peak shows a threshold of 685 W [15].
The second set of data comes from a device with a loaded Q of 2x10
8
in air. The coupling
was maintained at 88% coupling. The pump wavelength was 771 nm. Four lasing peaks were
generated in this case. The threshold for the initial lasing peak is 157 W. As expected, the
93

threshold slightly increases for the higher order peaks and is 220 W, 259 W, and 301 W,
respectively.

Figure 4-14 Lasing in air. a-d) The lasing peak intensity is plotted as a function of input power
going into the sphere for all four detected emission peaks. The threshold for a)-d) is 157 W, 220
W, 259 W, and 301 W, respectively. Inset: Lasing intensity graph showing four lasing peaks at
801, 829, 862, and 895 nm [22].
94

4.7 Results in Buffer
To suspend the sphere in a liquid, a chamber of about 350 L volume was designed. The
tip of the spectrograph fiber is placed close to the device, however, the fiber was slightly further
from the sphere during water experiments to avoid liquid damage. To minimize any distortion
due to differing volumes of liquid, a glass cover slide was used as the top of our liquid chamber.
A rendering of the liquid chamber is shown in Figure 4-15.
Although there are many biologically relevant fluids, for the present series of
experiments, hydroxyethyl piperazine-ethanesulfonic acid (HEPES, 0.2 M, Electron Microscopy
Sciences) buffer was used since it is ideal to hold biologically important proteins and molecules
[23]. This buffer consists of neutral molecules that help to maintain physiological pH even with
fluctuations of dissolved CO
2
.

Figure 4-15 Rendering of testing set-up for buffer experiments, indicating locations of device, taper,
and spectrograph fiber tip. The chamber is filled with HEPES buffer [15].
The microsphere used in the following experiment had a diameter of 194 m. Similar to
air experiments, during Q measurements, the linewidth ( ) distortion from non-linear effects
was minimized to produce a Lorentzian dip in transmission (Figure 4-16a). The loaded quality
95

factor in buffer, found by dividing the center wavelength by the full width at half maximum
linewidth (Q = /), was 1.57x10
7
. When recording the lasing spectrum, the laser scan range
and scan rate were set to maximize the lasing intensity. Even immersed in buffer, at high input
powers, the linewidth exhibits thermal broadening as seen in Figure 4-16b.

Figure 4-16 Normalized transmission spectra. a) Intrinsic quality factor of the device immersed in
HEPES buffer, Q
0
= 1.57x10
7
. b) Broadened resonance peak of microsphere in buffer taken during
a measurement [15].
Figure 4-17 shows the lasing results when this device was immersed in HEPES. As the
spectra in Figure 4-16 show, the measurements were taken in the critical-coupling condition. The
laser emission appeared at 806 nm, slightly higher than the emission in air, but only because the
pump wavelength was at a slightly higher 775 nm. As a result of the slight decrease in Q due to
the HEPES, the lasing threshold increases to 1.94 mW (Figure 4-17). The significant change in
threshold is due to the increased mode volume and the increased material absorption of the
buffer. Therefore, due to limitations on the testing laser, additional, higher wavelength lasing
lines in HEPES were not achieved with this device.
96

Figure 4-17 Lasing in buffer. (a) Emission peak at 806 nm. (b) Emission intensity versus input
power in HEPES buffer. The threshold is 1.94 mW [15].
An additional experiment was performed in HEPES to improve on the results. The device
used showed a higher intrinsic Q when immersed in buffer with a value of 9x10
7
. The sphere had
a diameter of 207 m. With this device, cascaded Raman lasing could be achieved (Figure 4-18).
Two lasing peaks at 800 and 827 nm were observed as shown in the inset of Figure 4-18a. The
threshold was lower than in the previous experiment with only 760 W required for the first
lasing peak and 824 W for the second.
97

Figure 4-18 Cascaded Raman lasing in HEPES buffer. a) Threshold graph for the first lasing peak
shown in the inset with a threshold value of 0.76 mW. Inset: Two lasing peaks achieved in HEPES
buffer. b) Threshold graph for the second lasing peak showing a threshold of 0.824 mW [22].
Table 4-1 shows results of some experiments done with microspheres of varying
diameters using a tunable laser around 770 nm as the pump. Other parameters not listed that
affect the results are: quality factor, thickness and loss of the waveguide, the fraction of light
coupling into the device for a given resonance, and the presence of thermal broadening. The
results show that cascaded behavior was observed to varying degrees with as little as three lasing
peaks and as many as ten lasing peaks (not all are in the table) in air. In water, up to two lasing
peaks have been observed with this pump laser.
98

Table 4-1 Cascaded Raman lasing results in air and buffer for various size spheres at a pump
wavelength of around 770 nm [15].

To verify the general behavior predicted by Equations (4.1) and (4.2), two data sets using
two different spheres were plotted (Figure 4-19). The experimental results are fit to a third order
power law, with all other variables floating. There is excellent agreement, even in devices with
nine emission lines. The slope of the threshold graph, or differential efficiency of the device,
shown in Figure 4-14 was also calculated (Figure 4-19, inset). As expected, the efficiency
decreases at the higher order lasing lines.
The results presented here indicate that dopant-free microlasers are able to achieve low
threshold lasing. Since laser emissions were achieved in a biologically significant media (HEPES
buffer), these silica microcavity devices can serve as biodetectors capable of detecting tiny
concentrations. By tracking the shift of the lasing peak, which has a narrower linewidth than the
resonance, molecules that come close to the surface of the device and interact with the
evanescent field can be detected. To perform such laser-based biodetection, a spectrograph
grating of higher resolution is required, since a wavelength shift of 1-10 pm is usually observed
for concentrations of 1 pM or lower.
99

Figure 4-19 Plot of threshold vs. peak number for two different experiments. Squares (circles) show
data for a 189 m (38 m) sphere. Experimental results agree well with the theoretical fit. Inset:
Slope efficiency for each lasing peak in Figure 4-14, showing a decreasing trend in agreement with
theory [22].
Label-free biodetection could also be possible using this method. By bioconjugating the
surface of the device with a specific marker, we can maintain high quality factors, but also make
the surface sensitive to one specific kind of molecule.
In conclusion, a dopant-free silica microcavity-based Raman laser was shown to operate
in air and buffer environments. Cascaded behavior was achievable and sub-mW lasing thresholds
are possible in both environments. The biocompatibility of this microlaser allows for its use in
many applications in the field of biodetection. By using a pump wavelength in the visible, lasing
in buffer as well as in air was achieved, which represents a significant advance over previous
work that was restricted to air or vacuum operation [4]. This type of biocompatible microlaser
will readily find applications within the biodetection community [24, 25].
100

4.8 Detection Experiments Using Raman Peaks
As stated previously, the resolution of the spectrograph grating was not high enough for
biodetection experiments. As an example, silica microresonators can be used to detect
temperature changes [26]. Looking over the theory of the silica material, according to the
thermo-optic coefficient if silica, a shift of about 8 pm per degree C with a 765 nm pump laser is
expected. The best grating for the spectrograph has a resolution of 0.16 nm (SR1-GRT-1200-
750, Andor), so a smaller shift than this cannot be detected. The temperature would need to
increase by over 20 degrees to see a shift in the lasing peak. However, simultaneously keeping
the device on resonance would be a challenge. Therefore, since the OSA has higher resolution
than this, it would need to be used for future detection measurements.
4.9 Raman Peaks Close to Pump
While studying the resolution of the OSA, some Raman peaks were observed in air near
the pump wavelength that could not be distinguished using the spectrograph in previous
experiments. When the Q and power are high enough, about 1x10
8
for Q and over 5 mW for
input power, modes within nanometers of the pump can be excited. The transmission out of the
taper is split using a 90/10 splitter where 90% of the power is sent to the OSA and the other 10%
is sent to the photodetector. Figure 4-20 shows examples of what these peaks look like on the
OSA. The sensitivity of the OSA needs to be increased to more than -65 dBm to see these peaks.
This discovery is important because these Raman peaks can be used for heterodyning, since
another 765 nm laser is available in the lab as the reference [27].
101

.
Figure 4-20 Raman lasing peaks around the pump wavelength of 777.8 nm seen on the OSA.
These peaks signify that other vibrational or rotational states of the silica material can be
excited with this wavelength. In addition, both Stokes and anti-Stokes peaks appear in this case,
indicating that some atoms in the material are already in the excited state.
4.10 Temperature-Dependent Shift of Raman Lasing Peaks in Air
As mentioned, the OSA receives the signal after it is split at the output of the tapered
optical fiber with a 633 nm splitter. The split ratio is 90/10, with 90% of the light going into the
OSA and the other 10% going to the photodetector where transmission dips are observed on the
Oscope/digitizer. For silica, cascaded Raman peaks are observed at intervals of 30 nm away from
the pump. Within the clumps of peaks, multiple modes at each interval are seen that could not be
distinguished using the spectrograph. These are the peaks that can be tracked for detection
experiments.
Because silica microspheres have very high quality factors, the material is easily excited
to produce Raman lasing peaks from the build-up of power within. Therefore, multiple resonance
peaks could produce lasing independently. Because the function generator sweeps the laser
102

through a range of wavelengths at 100 Hz and the OSA receives data at a different frequency, the
lasing peaks will not always correspond to the same resonance on the OSA. As an example,
Figure 4-21 shows the cascaded Raman lasing output of the OSA at three time points within
seconds of each other. As can be seen, the amplitude and position of the peaks vary. Even though
the pump laser is kept at a constant wavelength, due to the presence of the function generator, the
OSA output changes. To be able to track only one peak, a small wavelength range under
increased sensitivity of the OSA must be analyzed.

Figure 4-21 Raman peaks detected on OSA at three time points.
A peak tracking program was written in LabView that communicates with the OSA and
records the maximum position on the screen over time. While tracking the maximum peak, the
temperature can be changed and the shifts can be detected as they occur. During temperature
experiments, the microsphere is suspended between the aluminum temperature stage and a glass
slide covered in aluminum foil. The cover slide creates a sort of heating chamber for the sphere
and helps to keep the temperature constant. The temperature stage is connected to a controller
that changes the temperature based on the readings received from the thermocouple on the stage
through a feedback loop.
103

Figure 4-22 Changes in Raman lasing peak wavelength as the temperature is changed. a) Tracking
two lasing peaks vs. time as the temperature is changed. b) Zoomed in view of the lower wavelength
peak from a).
The temperature was decreased from 29 to 28.1 degrees C, then increased to 29.7
O
C (blue
squares in Figure 4-22). The black circles show the wavelength over time of the maximum
intensity laser peak. As can be inferred from the graph, two peaks were competing for the
maximum intensity, since two sets of points appear at 801 and 799 nm (Figure 4-22). Looking at
just one of the peaks in Figure 4-22b, it seems like the noise in the peak wavelength is too large
to differentiate a shift in the data.
With increased sensitivity on the OSA, the lasing peaks are more stable and appear in the
same position most of the time. However, recording the maximum position over time at
increased sensitivity still did not produce a noticeable shift in the data. Even for a temperature
change of 4
o
C, which should produce a 0.032 nm shift, the peak maximum fluctuations are still
larger than this value. Without any temperature changes, the spread of the maximum position is
about 0.2 nm and these fluctuations are too high to detect shifts.
To address this issue of peak fluctuation, other functions of the OSA that could help to
reduce the noise were utilized. For example, the OSA is able to average scans as it takes them.
104

To minimize the fluctuations in the data, the maximum of the average of 50 scans was recorded
over time. The fluctuations decrease to about 0.08 nm, but are still not good enough for small
shift measurements. The inherent resolution of the OSA can be found from the digitization of
data in a small wavelength range. In a scan of 0.2 nm, the data is digitized as 13 points per 0.1
nm. That means the data is recorded every 7.69 pm, which is just smaller than the 8 pm shift
expected for a one degree temperature change.

Figure 4-23 Data analysis for averaged scans. The scans were averaged and a Gaussian was fit to
the averaged data.
To further improve on this resolution, a Gaussian function was fit to the averaged peaks,
giving a better approximation for the maximum position. First, the OSA was set to average ten
scans before displaying the data on the screen. A total of ten averaged scans would then be
recorded one after another. During data analysis, the ten recorded scans were averaged using
OriginPro software and then a Gaussian peak was fit to the averaged graph. Figure 4-23 shows
the details of the data analysis. Figure 4-23a shows the ten averaged scans saved from the OSA
with one lasing peak present. Figure 4-23b shows the average of those ten scans and the fitted
Gaussian peak. From the fit, the maximum position of the peak is recorded.
105

Using the method mentioned, a temperature experiment was performed. The temperature
was allowed to stabilize at every step point before the data was recorded. After taking data in the
manner explained, the temperature set point would be changed to the next value. This method
produced excellent results. At the same time as the OSA scans were saved, the resonance data
was also recorded. In this way, both the Raman lasing peak and the resonance peak could be
tracked simultaneously. The temperature was increased by one degree with steps of 0.2
o
C. Then
it was decreased back to the starting point with the same step size. Figure 4-24 shows the data for
the shift in the resonance a) and the Raman lasing b) peaks as the temperature was changed. Only
the maximum of the Gaussian peak is plotted at each temperature for the Raman data.

Figure 4-24 Temperature experiments on averaged data sets. a) Resonance shift for increasing and
decreasing temperature. b) Raman lasing peak shift with temperature changes in both directions.
To produce two points at the high temperature, the stage was heated 0.2 degrees higher
than the end point and then cooled back down to the 26 degree mark. Since the heating causes
the resonance wavelengths to shift, the resonance shifted too far to still be observable on the O-
Scope. The lasing peak also disappeared. As the temperature was lowered back to 26 degrees, the
resonance peak reappeared, but the lasing peak appeared in a different location, indicated by the
106

red line going off the screen in Figure 4-24b. However, after this point, the lasing peak appeared
in the same or close to the same position as during increasing temperature experiments.
Figure 4-25 shows the agreement between the resonance and Raman lasing wavelength
shift. The resonance wavelength axis is shown on the right side of the graph, while the Raman
wavelength axis is on the left side. It is apparent that both sets of data follow the same trend.
Only the increasing temperature Raman wavelength shifts do not line up well with the others, but
still appear close. It is also important to notice that the shift in both regimes are about 8 pm,
exactly what is expected from a change of 1 degree Celsius.

Figure 4-25 Comparing resonance and Raman lasing wavelength shifts vs. temperature.
4.11 Temperature-Dependent Shift of Raman Lasing Peaks in Buffer
In the experiments done in air, a splitter was used to direct the signal to both the OSA and
the photodetector; however, the splitter was meant for 633 nm light rather than the 765 nm pump
or the 800 nm Raman. This is not an ideal way to guide light of these wavelengths. This is also
the most likely problem of why lasing could not be detected when the microsphere was tested in
buffer. While the absorption of 765 nm light is not very high in an aqueous environment, it is not
107

negligible, increasing the threshold of lasing. Lasing could be detected in buffer by using the
spectrograph, but lasing peaks on the OSA could not be seen even with high Qs and high input
powers.
A 50/50 splitter designed for 780 nm light was purchased from Newport to address this
issue. Using this splitter, lasing peaks were detected in the HEPES buffer very easily. Figure
4-26 shows two examples of Raman lasing peaks on the OSA when testing in water.
Nevertheless. even by using the new splitter, high Qs and high power was needed to produce
lasing output. Though the splitter supports wavelengths closer to the lasing emission, it is still
not designed for the exact wavelength in the output. Also introducing the splitter into the setup
introduces loss as the output of the tapered optical fiber, which ends in a bare fiber adapter, must
couple light into the splitter.

Figure 4-26 Raman lasing peaks on the OSA when testing in HEPEs buffer with a pump
wavelength of 774.7 nm and 780.395 nm.
Lasing did not occur for other buffers used, such as phosphate buffered saline (PBS) or
saline-sodium citrate (SSC). Any buffer used was vacuum filtered before every use to ensure that
108

it was clean. Since the HEPES buffer is kept in a 4
o
C refrigerator, the buffer was allowed to
come to room temperature before experiments were performed.
The data was recorded in the same way as for the experiments done in air. The
microsphere used had a Q of 1.78x10
8
. Ten averaged scans were recorded at each temperature,
averaged, and fit with a Gaussian peak to find the maximum. The results are shown in Figure
4-27. The figure shows both the resonance and Raman lasing peaks shifts with temperature. Only
three points were obtained because after this the lasing peak disappeared. Working with such
high power, there is heating of the buffer around the device. This causes a red shift to occur
continuously when testing in buffer. The power couldn't be lowered though because the lasing
could not be detected at lower power on the OSA.

Figure 4-27 Temperature induced resonance and Raman lasing peak shifts.
The heating also causes the buffer to evaporate. Adding more buffer during the
experiment is not possible. When adding buffer at the beginning of the experiment, the
temperature stage and sample are kept in place while the optical fiber is moved sideways until
the thicker part of the taper is in the sample chamber. The buffer is then added to the chamber.
109

This method further insures that if any impurities remain in the buffer or on the sample holder,
they will stick to the thick part of the taper where the light does not interact with the
environment. If buffer is added to the thin part, usually the power decreases because even with
filtration and a clean sample holder, some impurities stick to the taper and ruin its performance.
When testing in buffer, the resonance peaks that produce lasing are broadened peaks that
also show signs of mechanical vibrations. If the time on resonance is adjusted by changing the
scan range on the function generator, more lasing peaks occur with increased amplitudes.
However, decreasing the scan range doesn't allow for good tracking of the resonance and it
quickly shifts out of the range. Figure 4-28 shows two representative resonance peaks that
resulted in Raman lasing output on the OSA.

Figure 4-28 Representative resonance peaks in HEPEs buffer for which Raman lasing was observed
on the OSA.
Resonances that show broadening and mechanical vibrations have very high Qs and
result in heating of the buffer. The background shift of the heating can be measured by tracking
the resonance peak shift over time. The shift occurs as 1.2 pm/min and is continuous.
Simultaneously, the position of the Raman lasing peak was recorded over time but with averaged
110

scans as before. The Raman lasing peak does not follow the same trend as the resonance shift,
but instead appears to be random in response. This could be because the continuous resonance
shift makes it hard to get good averaged scans, since the average will contain more error.
Therefore, it is not viable to record Raman lasing peaks as they shift in buffer under the current
setup. Also because of the thermal shift, using temperature to heat the sample was possible, but
cooling back down would not occur, since the buffer is heated not only by the temperature stage,
but also by the sample itself.
4.12 Ultraviolet Light Induced Shift of Raman Lasing Peaks
In addition to temperature studies, ultraviolet (UV) light was also used to see Raman
lasing shifts, since it had been shown that silica devices have a response to UV [28]. In air, the
stem of the microsphere was attached to a sample holder, and the sphere was suspended in air.
The sphere was then exposed to various intensities of UV light while the resonance peak was
tracked on the O-Scope.
When the UV was turned on, the resonance shifted as expected but the fluctuations in the
resonance became large. Without UV, the noise in the resonance is 2.7%, whereas with UV, the
noise increases to 18.2%. Such large fluctuation would not produce stable lasing peaks. Besides
increasing the fluctuations in the data, UV light also changes the resonance depth or coupling
condition.
The spheres were also tested in buffer to see if these fluctuations occur in an aqueous
environment as well. The sphere was suspended above the sample holder and a glass slide was
placed on top to make the chamber. The buffer was added and the UV source was placed above
the chamber. The resonance was recorded over time. Figure 4-29 shows two examples of the
resonance shift over time when tracked in buffer.
111

Figure 4-29 Shifts in resonance in HEPEs buffer with UV.
As is evident in both graphs, there is a background thermal shift that occurs in the buffer
with a constant slope. More importantly though, we see no more huge fluctuations in the signal
even when the UV is on for a long time. This means that something in the testing setup is
stabilizing the signal. It turns out that placing the sphere above the steel holder, rather than
suspending it in air, produces the smooth curves. Repeating the experiment in air produces the
same smooth curve without the thermal shift. Therefore, the presence of buffer enhances thermal
effects due to absorption.
Since the shift is smooth as long as the sphere was suspended above the steel holder, a
detection experiment was performed to measure the shift in Raman lasing peaks in air. The UV
source intensity could be changed by 1%, so the same procedure could be used as for the
temperature experiments. The intensity was adjusted, given time to stabilize, and averaged scans
were recorded on the OSA before moving on to a new intensity set point. Since the intensity of
the light reaching the sample depends on the distance of the UV source, the source was placed
further away to produces smaller shifts, which are easier to track continuously. The intensity
percentage was measured on a UV detector at the conclusion of the experiment at the same
112

location as the microsphere. The values were as follows: 1% - 8 mW/cm
2
, 2 - 10, 3 - 12, 4 - 14, 5
- 16, 6% - 18 mW/cm
2
. The intensity was increased from 0 to 6% and then decreased, allowing
for the resonance to stabilize at each point. When the UV intensity was decreased, the resonance
did not shift back to the previous position. This could be due to the resonance changing when it
is exposed to UV for a long time.
At the same time as the resonance was tracked, the Raman lasing signal was also
recorded. The Gaussian fit data for two different Raman peaks vs. changing UV intensity is
shown in Figure 4-30. The black squares indicate the shift as the intensity was increased, while
the red circles show the shift for decreasing intensity. The shifts of the Raman peaks did not
match up with the UV changes very well. This again could be due to the UV light changing the
resonance peak and thus changing the lasing peak from prolonged exposure.

Figure 4-30 Raman lasing peak shifts for two different lasing peaks. The black squares show shifts
for increasing UV intensity and the red circles for decreasing intensity.
Raman lasing peaks could not be accurately tracked using UV even in air. The lasing
peaks could also not be tracked well in buffer with temperature changes due to the large
background thermal shifts. Trying to track the Raman peak without averaging is not possible
113

since the peak width is too big and produces fluctuations in the maximum position that are much
larger than the expected shift.
4.13 Conclusion

In conclusion, I showed that cascaded Raman lasing was possible in air and in an aqueous
environment. I tried to demonstrate a shift of the Raman wavelength in air and buffer. I started
with OSA noise measurements that showed Raman peak variations. Since the data is unstable, I
needed to average over numerous scans. I was able to show that the shift of resonance
wavelength is consistent to Raman wavelength shift as a function of temperature in air. I wanted
to repeat experiment on lasing peak in HEPES buffer, but the background thermal shift was too
large in the current set up. Also, I tried to use UV intensity to track Raman lasing peaks, but even
air, the shift was too unpredictable.
References
1. von Klitzing, W., et al., Very low threshold green lasing in microspheres by up-
conversion of IR photons. Journal of Optics B-Quantum and Semiclassical Optics, 2000.
2(2): p. 204-206.
2. Hsu, H.S., C. Cai, and A.M. Armani, Ultra-low-threshold Er:Yb sol-gel microlaser on
silicon. Opt Express, 2009. 17(25): p. 23265-71.
3. Mazzei, A., et al., Influence of a controllable scatterer on the lasing properties of an
ultralow threshold Raman microlaser. Applied Physics Letters, 2006. 89(10).
4. Spillane, S.M., T.J. Kippenberg, and K.J. Vahala, Ultralow-threshold Raman laser using
a spherical dielectric microcavity. Nature, 2002. 415(6872): p. 621-3.
5. Lin, H.B. and A.J. Campillo, Microcavity enhanced Raman gain. Optics
Communications, 1997. 133(1-6): p. 287-292.
6. He, L.N., et al., Detecting single viruses and nanoparticles using whispering gallery
microlasers. Nature Nanotechnology, 2011. 6(7): p. 428-432.
7. Armani, A.M., et al., Ultra-high-Q microcavity operation in H
2
O and D
2
O. Applied
Physics Letters, 2005. 87(15): p. 151118.
8. Feyerabend, F., et al., Evaluation of short-term effects of rare earth and other elements
used in magnesium alloys on primary cells and cell lines. Acta Biomaterialia, 2010. 6(5):
p. 1834-1842.
9. Grudinin, I.S. and L. Maleki, Ultralow-threshold Raman lasing with CaF2 resonators.
Optics Letters, 2007. 32(2): p. 166-168.
114

10. Carmon, T., et al., Feedback control of ultra-high-Q microcavities: application to micro-
Raman lasers and microparametric oscillators. Optics Express, 2005. 13(9): p. 3558-
3566.
11. Min, B., T.J. Kippenberg, and K.J. Vahala, Compact, fiber-compatible, cascaded Raman
laser. Optics Letters, 2003. 28(17): p. 1507-1509.
12. Stolen, R.H. and E.P. Ippen, Raman gain in glass optical waveguides. Applied Physics
Letters, 1973. 22(6): p. 276-278.
13. Knight, J.C., et al., Phase-matched excitation of whispering-gallery-mode resonances by
a fiber taper. Opt Lett, 1997. 22(15): p. 1129-31.
14. Freeman, L.M., et al., Excitation of Cy5 in Self-Assembled Lipid Bilayers using Optical
Microresonators. Applied Physics Letters, 2011. 98(15): p. 143703.
15. Chistiakova, M.V. and A.M. Armani, Microcavity-based cascaded Raman microlaser in
air and in buffer. Optical Components and Materials X, 2013. 8621.
16. Hale, G.M. and M.R. Querry, Optical-Constants of Water in 200-Nm to 200-Mum
Wavelength Region. Applied Optics, 1973. 12(3): p. 555-563.
17. Oxborrow, M. How to simulate the whispering-gallery modes of dielectric
microresonators in FEMLAB/COMSOL. 2007.
18. Polyanskiy, M. Refractive Index of SiO2. 2013; Available from:
http://refractiveindex.info/legacy/?group=CRYSTALS&material=SiO2&option=F_SILI
CA.
19. Bass, M., V.N. Mahajan, and Optical Society of America., Handbook of optics. 3rd ed,
New York: McGraw-Hill. v. <1, 4-5>.
20. Malitson, I.H., Interspecimen Comparison of Refractive Index of Fused Silica. Journal of
the Optical Society of America, 1965. 55(10P1): p. 1205-&.
21. Koerner, M.M., et al., Electrodynamics of Lipid Membrane Interactions in the Presence
of Zwitterionic Buffers. Biophysical Journal, 2011. 101(2): p. 362-369.
22. Chistiakova, M.V. and A.M. Armani, Cascaded Raman microlaser in air and buffer.
Optics Letters, 2012. 37(19): p. 4068-4070.
23. Radi, A., et al., Reagentless, reusable, ultrasensitive electrochemical molecular beacon
aptasensor. Journal of the American Chemical Society, 2006. 128(1): p. 117-124.
24. Hunt, H.K. and A.M. Armani, Label-free biological and chemical sensors. Nanoscale,
2010. 2(9): p. 1544-1559.
25. Washburn, A.L. and R.C. Bailey, Photonics-on-a-Chip: Integrated Waveguides as
Enabling Detection Elements for Lab-on-a-Chip Biosensing Applications. Analyst, 2011.
136: p. 227-236.
26. Choi, H.S., S. Ismail, and A.M. Armani, Studying polymer thin films with hybrid optical
microcavities. Optics Letters, 2011. 36(11): p. 2152-2154.
27. Maker, A.J. and A.M. Armani, Heterodyned toroidal microlaser sensor. Applied Physics
Letters, 2013. 103(12).
28. Harker, A., S. Mehrabani, and A.M. Armani, Ultraviolet light detection using an optical
microcavity. Optics Letters, 2013. 38(17): p. 3422-3425.

115

Chapter 5 Determining the Rate of DNA Hybridization
5.1 Overview
The development of DNA analysis methods is rapidly expanding as interest in
characterizing subtle variations increases in biomedicine. A promising approach is based on
evanescent field sensors that monitor the hybridization process in real time. However, one
challenge is discriminating between nonspecific and specific attachment. Here, a hybridization
sensor based on an integrated toroidal optical microcavity is demonstrated. The surface is
functionalized with ssDNA using an epoxide method, and the evanescent wave of the
microresonator excites a fluorescent label on the complementary ssDNA during hybridization.
Based on a temporal analysis, the different binding regimes can be identified.
5.2 Background and Motivation
As the genome is deciphered, the relationship between subtle variations in DNA and
various disease states is becoming increasingly apparent [1-6]. To this end, researchers are
developing novel diagnostic platforms based on DNA hybridization assays [5-10]. DNA
hybridization or renaturation is the process in which a single strand of nucleotides forms a
noncovalent association with its complement strand [11]. The hybridization rate is governed by
several parameters, including the degree of similarity between the strands and environmental
factors, such as pH and temperature.
Because there are many factors that can simultaneously contribute to hybridization, the
measurement of DNA hybridization is nontrivial. For example, this process is dependent on the
concentration of DNA, the temperature, and the cation concentration. At the very minimum,
there are two conditions that must be met in order for hybridization to occur. For one, the salt
concentration must be high enough that the phosphate groups of the DNA strands are not
116

repelled. Second, the temperature of the sample must be at least 20°C
―
25°C below the melting
temperature (T
m
). In this temperature range, random intrastrand hydrogen bonding is disrupted,
favoring interstrand hydrogen bonding between complementary base pairs.
The conventional method for studying hybridization is to measure the reverse process,
specifically the melting temperature. In this approach, single-stranded DNA (ssDNA) is
immobilized on a surface and hybridized to the complement strand. The sample is then
gradually heated, denaturing or melting the hybridized DNA, while recording the fluorescence or
optical absorbance spectra. From this signal, the binding energy can be determined. While this
measurement does provide an accurate reading of the dissociation of the DNA, it is unable to
yield information about the behavior of DNA at ambient temperatures and physiological pH, and
requires a significant quantity of the DNA.
Recently, researchers have recognized this challenge and begun to develop new
technologies to address it. One approach is the combination of evanescent wave optical sensors
with fluorescently labeled ssDNA (ssDNA-dye) [9, 10]. In this approach, the ssDNA is
immobilized directly on the surface of the optical device and exposed to the ssDNA-dye
complement. The optical field excites the dye molecules and the fluorescence is detected using a
variety of methods, including spectrographs and power meters. Using this approach, researchers
have shown the ability to detect the hybridization process using these sensor devices
However, there are actually two phases in the absorption and hybridization process.
Specifically, when molecules enter the evanescent field, some will be oriented correctly and
begin to hybridize immediately; however, many will bind non-specifically to the device surface,
rapidly dissociating and diffusing away. These molecules will produce a transient signal.
Therefore, the detection signal has two components: 1) the hybridized DNA and 2) non-
117

hybridized DNA. This two-phase process follows directly from mass transport/kinetics theory
and sensor response theory [12].
Clearly, during the initial phase of the experiment, the second component (nonspecific
binding) is significantly larger than the first (hybridized DNA). However, as the experiment
progresses, it is anticipated that the signal from the second component will go to zero, leaving
only the signal from the hybridized DNA. In previous work, this dual-phase detection signal was
not resolvable, and possibly incorrectly interpreted as a single signal, for several different
reasons. For example, without sufficiently high sensitivity and/or fast temporal resolution, the
transient signal would not be recorded by the detector (e.g. spectrograph). However, it is critical
to fully understand the temporal aspect of the sensor response, given the time dependence of the
DNA hybridization process.
5.3 Testing Approach
In the present work, an evanescent wave sensor is presented in combination with a high
speed spectrograph to detect both phases of binding. The sensor is based on an integrated optical
resonant cavity, specifically a toroidal microcavity (Figure 5-1a) [13, 14]. An oligonucleotide
with 20 nucleotides (20-mer) was used. The 20-mer ssDNA is attached to the surface of the
device using an epoxide approach, and its complement is labeled with the cyanine dye 5 (Cy5),
which fluorescently emits between 650 nm and 670 nm when excited with a wavelength around
635 nm. As the DNA hybridizes, the emission is detected on the spectrograph and analyzed.
118

Figure 5-1 Silica toroidal optical microcavity. a) Scanning electron micrograph of a toroidal optical
microcavity. b) 2D COMSOL finite element method simulation of the equatorial cross section of a
toroidal optical cavity. The optical field extends into the environment, enabling excitation of
fluorophores located near the surface of the device [15].
COMSOL Multiphysics finite element method modeling of the optical field distribution
inside the toroid and the environment was performed to verify that the functionalized DNA fell
within the evanescent field of the toroid. By leveraging the symmetry of the device, the
modeling challenge simplifies to a 2D problem, as shown in Figure 5-1b. The black line in the
figure indicates the boundary of the toroid ring. The environment around the ring was modeled
as buffer. The evanescent field was found to extend approximately 100 nm into the environment
completely overlapping the ssDNA and fluorophore. Because of the long photon lifetime within
the cavity (high quality factor), very low input power is needed to generate a strong fluorescent
signal. In previous work, optical cavities have demonstrated the ability to excite fluorescent dyes
embedded within lipid bilayers and detect DNA [16-20].
119

5.4 Device Fabrication and Functionalization
The silica toroidal cavities are fabricated on silicon wafers using a simple three-step
method involving photolithography, two etching processes (buffered oxide etch and xenon
difluoride etch), and a carbon dioxide laser reflow step [13, 14].

Figure 5-2 Epoxide functionalization process. The surface of the cavity is hydroxylated and then
GPTMS is used to covalently attach epoxide groups. Aminated ssDNA binds to the epoxides,
forming the ssDNA functionalized devices. In the last step, the complement ssDNA-Cy5 is
hybridized to the surface. Cy5 emits in the red. To perform the control fluorescent imaging
experiments, the aminated ssDNA was labeled with 6-FAM, a green fluorophore [15]. The devices
can also be recycled if placed under oxygen plasma once again, which will remove the
functionalized components.
120

To immobilize and orient the ssDNA on the surface of the cavity, an epoxide approach is
used, which relies on an amine-initiated nucleophilic ring-opening reaction of an epoxide.
Specifically, the surface of the cavity is hydroxylated using an oxygen plasma (120 Watts, 200
mTorr, 30 sccm for 5 minutes). Then the epoxy linker (3-glycidoxypropylmethyldiethoxysilane,
GPTMS) is added by vapor deposition under vacuum for 45 minutes. Finally, the amine
modified oligonucleotides (IDT, Coralville Iowa) are covalently linked to the epoxide at 37 C in
a humid environment for six hours or overnight (Figure 5-2) [21].
The functionalization method was optimized for both microtoroids and microspheres for
possible detection experiments in the future. The epoxide method was used to attach the amine-
modified DNA to the device. This procedure could also be used for antibodies or other proteins,
since they have an amine end, which breaks up the epoxide ring to bind to the surface.
The steps to functionalize microspheres are similar to those used for the toroids. First, the
oxygen plasma treatment is performed to put hydroxyl groups on the surface. Second, the vapor
deposition method is used to deposit GPTMS on the surface of the devices under vacuum for 45
minutes. Third, the devices are incubated in a 35
o
C controlled-humidity water bath within a
solution of DNA strands for at least three hours or overnight. The difficulty arises in that the
spheres are not fabricated on a chip and can thus be damaged by knocking the samples into the
walls of the solution holder. To avoid damage, the spheres were suspended along the wall of the
tube. Since part of the fiber tail of the spheres still has cladding, it helps to keep the spheres away
from the walls. The solution was then added and the tube was kept tilted so that the spheres were
exposed to DNA as much as possible.
The samples were then imaged with a fluorescent microscope to detect the fluorophore
tags on the DNA to verify attachment. The strand being used for sphere functionalization was
121

marked with fluorescein isothiocyanate (FITC). No fluorescence was detecting using the
described method of functionalization. The incubation time was increased, but did not help. A
baking method was also tried in which the devices were incubated at 60
o
C for two hours instead
of incubating at 35
o
C. The baking method did not produce fluorescence even when the time was
increased to 15 hours. A heating block designed for biological procedures was then used for
incubation at 60
o
C, but gave similar results.
At this point the fluorophore on the DNA was tested to see if photobleaching had
occurred. A drop of the DNA solution was placed on a glass slide and imaged with the
fluorescent microscope. The fluorescence was not evenly distributed along the droplet. Some
locations showed fluorescence while others did not, indicating at least partial photobleaching.
Since the FITC fluorophore was not reliable, the attachment of DNA had to be shown by
adding another step to the process. At the end of the previous procedure, the devices would be
incubated in a solution with the complementary DNA strand that had a cyanine (Cy5)
fluorophore attached. The fluorophore was confirmed as active since a droplet of this solution
showed strong uniform fluorescence under the microscope. The procedure was only altered
slightly in that the first incubation step was done at 37
o
C instead of 35
o
C for six hours. The
devices were also washed between incubation steps to remove any excess salts. Incubation with
the complementary strand was done at 60
o
C overnight.
After the added functionalization step, the Cy5 fluorescence could be seen on the
microspheres. Figure 5-3 shows the bright field and fluorescence images of two spheres, one
functionalized and one used as a control. Figure 5-3a and c show images of a sphere that was
functionalized following the procedure mentioned. Figure 5-3b and d show images of a control
sphere that was not exposed to oxygen plasma but was exposed to the other functionalization
122

steps. The functionalized sphere showed fluorescence from the Cy5 flurophore, while the control
sphere does not. This shows that the hydroxyl groups are an important step in functionalization
and that DNA does not bind if the hydroxylation step is skipped. The stem of the microsphere
does not show fluorescence most likely because the small quantities of the DNA solutions that
were used did not cover the stem but only the sphere, indicating that not the entire device was
immersed in solution.

Figure 5-3 Bright field and fluorescence images of the microscope showing functionalization has
occurred. a) A bright field image of a functionalized microsphere showing a smooth surface that
indicated that the sphere was not damaged during the process. b) A bright field image of a control
sphere that was not exposed to oxygen plasma also shows a smooth surface. c) A fluorescent image
of the functionalized sphere shows the Cy5 dye present on the complementary strand of DNA and
123

that it has bound to the surface of the sphere. d) A fluorescent image of the control sphere showing
no fluorescence, indicating that DNA does not bind to bare silica and that functionalization does not
occur without the oxygen plasma step.
One concern with using the epoxide method for functionalizing the device with proteins
is that not all epoxide groups may be covered in the functionalization step, leaving a few rings
behind. A false reading could then result when trying to perform detection since incorrect or non-
specific binding could occur. This problem could be avoided by incubating with BSA after
incubation with the primary protein before the detection experiments. This would saturate the
epoxide rings and would prevent false detection signals of non-specific binding during
experiments. When using DNA, this is not an issue since ssDNA does not have naturally
occurring amine groups. Therefore, it will not bind to any remaining epoxide rings.
5.5 Verifying Functionalization on Silica Microtoroids
To verify the activity of the surface chemistry on the toroids, a multi-color fluorescent
microscopy study is performed in which the surface immobilized amine modified ssDNA is
labeled with 6-carboxyfluorescein (6-FAM), a derivative of fluorescein and highly compatible
with oligonucleotides, and the complement ssDNA is labeled with Cy5 (ssDNA-Cy5). As such,
6-FAM emits in the green and Cy5 emits in the red. Imaging is performed at each step in the
process: (1) after hydroxylation and attachment of epoxide silane, (2) after attachment of the
amine modified, 6-FAM labeled ssDNA, and (3) after hybridization of complement ssDNA
(ssDNA-Cy5) at room temperature. These reaction conditions are identical to those used in the
optical device experiments and therefore serve as a control measurement verifying hybridization.
The images are combined using Nikon NIS element basic research imaging software.
124

Figure 5-4 Multicolor fluorescent imaging. a) Bright field image of a microtoroid after GPTMS
vapor deposition and incubation with amine modified ss-DNA-6-FAM. b) Fluorescent image of the
same microtoroid showing the attachment of ss-DNA-6-FAM. c) Fluorescent image of a
microtoroid after incubation with complement ss-DNA-Cy5. The filters are adjusted to isolate the
Cy5 emission from the FAM emission. d) Overlap of parts b and c, verifying that hybridization
occurred. The specific sequences used are: 3’–NH
2
-GCC GGA TAG CGT AAA GGT TA-FAM
and 5’-CGG CCT ACT GCA TTT CCA AT/Cy5-3’ [15].
The bright field image in Figure 5-4a shows that the toroid was not damaged in the
functionalization process and retains the smooth walls. The fluorescence imaging is done with a
125

fluorescent microscope that can automatically adjust the excitation and detection wavelengths for
any type of dye specified. The images are false colored by the microscope software to distinguish
parts that show fluorescence from those that do not. The presence of green fluorescence in Figure
5-4b indicates that the functionalization approach uniformly conjugates ssDNA to the surface of
the device. Figure 5-4c shows the emission from the ssDNA-Cy5 where it has hybridized to the
6-FAM labeled ssDNA verifying that hybridization occurs at room temperature. Figure 5-4d is
the overlap of these two images, showing precisely where the two ssDNAs are co-located.
Superposing fluorescent images is a classic method in dual-color fluorescent microscopy. It
allows for easy comparison of the location of two different probes, in this case, green and red.
Where the two probes overlap, one expects to see yellow which indicates hybridization in Figure
5-4d.
5.6 Device Characterization
Two different optical measurements are performed: (1) device characterization and (2)
DNA hybridization detection. In both experiments, the toroids are placed on a steel sample
holder and a raised cover slide is secured on top, creating a chamber that can be filled with
liquid. For device characterization, the bioconjugated resonant cavity is initially immersed in
nuclease free water and the DNA hybridization detection experiments are measured in 100 L of
2X sodium citrate (SSC) with 0.2% sodium dodecyl sulfate (SDS) buffer (pH 7.0) at ambient
temperature. Light is coupled into the device from a tunable narrow linewidth laser centered at
633 nm (Velocity series, Newport) using a tapered optical fiber waveguide.
To measure the photon lifetime (quality factor) of the bioconjugated cavity, the cavity is
aligned to the waveguide using a high precision three-axis nano-positioning stage (Optosigma)
and monitored on top and side view machine vision systems. The transmission spectrum is
126

recorded using a high speed digitizer/oscilloscope (National Instruments) and fit to a Lorentzian.
The quality factor is calculated using the expression: Q =   , where  is the full width half-
maximum of the Lorentzian fit and  is the resonant wavelength, resulting in a Q of 2.2 x 10
7
for
the device tested (Figure 5-5a) [13, 14]. The quality factor stays high even after
functionalization, indicating that the attached molecules do not cause excessive loss from the
cavity.

Figure 5-5 Optical device characterization. a) A transmission spectra used to determine the quality
factor of the cavity. Based on the linewidth, this device has a Q of 2.2 x 10
7
in water. b) A rendering
of the testing setup, which can simultaneously record the emission from the fluorescent dye using a
fiber coupled spectrograph and inject the ssDNA-Cy5 using a syringe pump. The emission from
the 633nm tunable laser is partially blocked using a red filter. The quality factor can also be
measured in this configuration [15].
5.7 DNA Hybridization Detection
To perform the hybridization detection experiments, the testing set-up is slightly
modified, and a fiber-coupled imaging spectrograph (Andor Shamrock SR-163 with a Newton
CCD detector) is integrated to the top (Figure 5-5b) [16]. A filter that blocks light below 650 nm
is placed between the spectrograph tip and the toroid to avoid saturation of the spectrograph
127

detector with the laser pump light. The complement ssDNA-Cy5 is injected into the sample
chamber using a syringe pump at a flow rate of 50 L/min for two minutes. Cy5 was specifically
selected as the fluorophore because 633 nm falls within the absorption range, allowing the
evanescent field of the cavity to efficiently excite the dye. The emission spectra are continuously
recorded for 30 minutes, and the peak emission intensity of the dye (670 nm) is monitored as it
evolves throughout the experiment.
To verify and characterize the sensing method, two different detection measurements are
performed using two different functionalized devices. The first is detection of a single solution
of 2 M ssDNA-Cy5. The second is the characterization of the working range. For this
measurement, a series of ssDNA-Cy5 solutions are made with concentrations ranging from 1 nM
to 2 M. The solutions are then sequentially injected into the volume around the toroidal cavity.
Control experiments are also performed without the ssDNA-Cy5 present.
Figure 5-6a shows a pair of representative excitation/emission spectra. The excitation
source at 633 nm, which originates from the optical cavity, is clearly identifiable in both spectra.
Even with the presence of the filter, some of the excitation source light is still detected.
However, since the spectrograph saturates at 65,000 counts, most of the light is indeed filtered
out. The second peak in Figure 5-6a, which corresponds to the emission from the fluorescent
dye, only occurs when the ssDNA-Cy5 is present. The peak emission wavelength of the dye
agrees with known values. It is also important to note that the signal to noise ratio of this
measurement is extremely high.
128

Figure 5-6 Detection of 2 M ssDNA-Cy5. a) Emission spectrum with and without the ssDNA-Cy5
present. While the 633 nm laser line is present in both spectra, the fluorescent emission is only
present when the ssDNA-Cy5 is injected, as expected. There are no secondary lines or other noise
sources present in this wavelength range. As such, the signal fidelity is extremely high. The arrow
indicates the 670 nm wavelength that is tracked in the detection experiments. b) The maximum of
the emission at 670 nm, indicated in part a, is monitored and recorded while the ssDNA-Cy5 is
injected. A strong but transient signal is generated when the molecule nonspecifically binds to the
surface and/or moves within the evanescent field. The second stable peak is the result of the
hybridization [15].
Representative hybridization detection results are contained in Figure 5-6b. There are
two clearly identifiable peaks. The first transient peak is very large in magnitude but fades
quickly. This peak is due to the ssDNA-Cy5 nonspecifically binding to the surface of the device
or diffusing through the evanescent field in solution. Additionally, since the solution is injected,
the flow pushes the DNA molecules onto the surface of the toroid, causing the rapid spike. In
contrast, the second peak is smaller but stable, indicating that the DNA has hybridized to the
surface. It is known that the hybridization reaction takes approximately 15 to 20 minutes under
the conditions specified. This is why there is a decrease between the two peaks, as the reaction
129

has not finished. Also, the signal does not decrease down to zero, indicating that at least some
hybridization has already occurred by this time.
5.8 Sensor Working Range
Figure 5-7 shows the characterization of the working range of the device. As expected, it
has a sigmoidal response. While the limit of detection is moderate, it is important to note that a
significant amount of the potential detection signal (Cy5 emission) is lost due to scattering at the
different interfaces and the optical absorption of water.

Figure 5-7 The working range of the device as it is sequentially exposed to several different ssDNA-
Cy5 solutions [15].
5.9 Conclusion
In conclusion, an epoxide based method for attaching ssDNA to the surface of integrated
optical cavities has been developed and verified without degrading the optical performance of
the device. Using the evanescent field of the cavity as an excitation source, both nonspecific
binding and the final hybridization process were detected. A more complete understanding of
the various transport processes that give rise to the detection signals in DNA hybridization
130

sensors as well as new diagnostic methods will aid in developing improved biological models for
disease.
References
1. Conrad, D.F., et al., Origins and functional impact of copy number variation in the
human genome. Nature, 2010. 464(7289): p. 704-712.
2. Hanahan, D. and R.A. Weinberg, Hallmarks of Cancer: The Next Generation. Cell, 2011.
144(5): p. 646-674.
3. Jackson, S.P. and J. Bartek, The DNA-damage response in human biology and disease.
Nature, 2009. 461(7267): p. 1071-1078.
4. Lister, R., et al., Human DNA methylomes at base resolution show widespread
epigenomic differences. Nature, 2009. 462(7271): p. 315-322.
5. Lujambio, A., et al., A microRNA DNA methylation signature for human cancer
metastasis. Proceedings of the National Academy of Sciences of the United States of
America, 2008. 105(36): p. 13556-13561.
6. Zhang, L., et al., microRNAs exhibit high frequency genomic alterations in human
cancer. Proceedings of the National Academy of Sciences of the United States of
America, 2006. 103(24): p. 9136-9141.
7. Stoughton, R.B., Applications of DNA microarrays in biology. Annual Review of
Biochemistry, 2005. 74: p. 53-82.
8. Lockhart, D.J., et al., Expression monitoring by hybridization to high-density
oligonucleotide arrays. Nature Biotechnology, 1996. 14(13): p. 1675-1680.
9. Long, F., et al., Ultrasensitive quantum dots-based DNA detection and hybridization
kinetics analysis with evanescent wave biosensing platform. Biosensors and
Bioelectronics, 2011. 26: p. 2390-2395.
10. Malic, L., T. Veres, and M. Tabrizian, Nanostructured digital microfluidics for enhanced
surface plasmon resonance imaging. Biosensors and Bioelectronics, 2011. 26: p. 2053-
2059.
11. Thomas, P.S., Hybridization of denatured RNA and small DNA fragments transferred to
nitrocellulose. Proceedings of the National Academy of Sciences of the United States of
America-Biological Sciences, 1980. 77(9): p. 5201-5205.
12. Gervais, T. and K.F. Jensen, Mass transport and surface reactions in microfluidic
systems. Chemical Engineering Science, 2006. 61: p. 1102-1121.
13. Armani, D.K., et al., Ultra-high-Q toroid microcavity on a chip. Nature, 2003.
421(6926): p. 925-928.
14. Zhang, X., H.-S. Choi, and A.M. Armani, Ultimate quality factor of silica microtoroid
resonant cavities. Applied Physics Letters, 2010. 96(15): p. 153304.
15. Hawk, R.M., M.V. Chistiakova, and A.M. Armani, Monitoring DNA hybridization using
optical microcavities. Optics Letters, 2013. 38(22): p. 4690-4693.
16. Freeman, L.M. and A.M. Armani, Photobleaching of Cy5 conjugated lipid bilayers
determined with optical microresonators. IEEE Journal of Selected Topics in Quantum
Electronics, 2012. 18(3): p. 1160-1165.
17. Freeman, L.M., et al., Excitation of Cy5 in Self-Assembled Lipid Bilayers using Optical
Microresonators. Applied Physics Letters, 2011. 98(15): p. 143703.
131

18. Hunt, H.K. and A.M. Armani, Label-free biological and chemical sensors. Nanoscale,
2010. 2(9): p. 1544-1559.
19. Luchansky, M.S. and R.C. Bailey, High-Q Optical Sensors for Chemical and Biological
Analysis. Analytical Chemistry, 2012. 84: p. 793-821.
20. Vollmer, F., et al., Multiplexed DNA quantification by spectroscopic shift of two
microsphere cavities. Biophysical Journal, 2003. 85(3): p. 1974-1979.
21. Carré, A., W. Birch, and V. LaCarriére, Silanes and Coupling Agents, K.L. Mittal, Editor.
2007. p. 14.

132

Chapter 6 Carbon Nanotube Gas Detection
6.1 Overview
The development of new materials relies on high precision methods to quantify gas
adsorption to and desorption from surfaces. One commonly used approach is temperature
programmed desorption spectroscopy, which measures gas release as a function of temperature.
While this approach is very accurate, it requires complex and bulky instrumentation, and it is
limited to performing experiments under high vacuum, thus restricting experimental scope. An
alternative approach is to integrate the surface of interest directly onto a detector face, creating
an active substrate. One surface that has applications in numerous areas is the carbon nanotube
(CNT). As such, an active substrate that integrates a CNT surface on a sensor and is able to
perform measurements in ambient environments will have significant impact.
In the present work, an active substrate has been developed that combines an optical
sensor with a CNT substrate. The optical sensor is able to accurately probe the temperature
dependent desorption of carbon monoxide and carbon dioxide gases from the CNT cluster
surface. This active substrate will enable a wide range of temperature dependent desorption
measurements to be performed from a scientifically interesting material system.
6.2 Background and Motivation
Understanding and quantifying the thermodynamic behavior of gas absorption to or
desorption from a substrate is critical in numerous fields, such as developing efficient catalysts
for bioengineering and CO
2
conversion and designing effective filters for water and air
purification [1-11]. One commonly used method for characterizing these interactions is based
on temperature programmed desorption (TPD) spectroscopy [12, 13]. In these measurements, a
substrate is saturated with the chemical of interest and is placed in an isolation chamber. The
133

vacuum chamber temperature is increased at a fixed rate, and at a critical temperature, which is
related to the activation energy for desorption (E
ads
), the adsorbed chemicals will desorb,
allowing several thermodynamic parameters to be determined. Typically, the detection of the
desorbing chemicals is performed by monitoring the mass of the desorbing molecules as a
function of temperature using a mass spectrometer. Therefore, due to the complexity of the
different components in the instrument, miniaturization of the TPD system has not been
aggressively pursued despite its relevance in a wide range of industries.
One approach to solve this challenge is to develop an active substrate, in which the
substrate is directly interfaced with the sensor in a single platform. The development of this type
of hybrid device requires the optimization of several parameters, particularly the interface
between the sensor element and the substrate material as well as the sensor readout process.
Recently, researchers have increased efforts in the design and development of these types of
hybrid sensor platforms [11, 14]. However, because most of the sensors were designed for
bio/chemical detection, they have faced problems maintaining their performance at the high
temperatures needed for TPD measurements. Therefore, to successfully create an active substrate
suitable for TPD, it is necessary to develop a sensor that is compatible both with the high
temperatures required and with the substrate of interest.
Of the numerous nano-material-based substrates currently being studied, carbon
nanotubes (CNT) have several interesting properties. They have an extremely large volume to
surface area ratio, a nano-porous structure, and a very chemically inert surface [15]. They are
also inexpensive and can be bought with different specifications of diameter, length, and other
physical parameters. As a result, researchers are interested in their ability to serve as filters or
134

nano-porous membranes for air purification as well as the potential to decorate or functionalize
CNTs with nano-catalysts to make dense, active matrix-like structures [16-19].
Within the sensor community, there is similar interest in CNT-based devices. Because
the physical properties, such as the length, diameter, and wall structure, can be easily tuned, there
are many methods to optimize the performance of a CNT-based sensor [20]. For the application
of TPD spectroscopy, CNTs offer a unique opportunity; specifically, as a result of their high
thermal conductivity, they can be exposed to very high temperatures with minimal to no impact
on their structure or performance. However, simply transforming the existing CNT nanowire
sensors into a TPD active substrate is not feasible because the background noise in an electrical
sensor increases as the temperature increases. Therefore, it is necessary to combine the favorable
thermal properties of the CNT substrate with an alternative sensing approach.
One such technique is based on evanescent field optical sensors; for example,
waveguides and interferometers [20, 21]. In these sensors, the optical field is not completely
confined within the device, but evanesces into the environment. As a result, it is able to sample
or monitor the environment at the surface of the device. This type of detection relies on the
interaction strength between the photons and the analytes of interest [22]. Therefore, by
combining an evanescent field sensor with a CNT coating, an active substrate capable of
studying gas-CNT interactions can be formed.
Of the different types of evanescent field sensors, whispering gallery mode cavities are
ideally suited for this application. In these devices, light of a well-defined wavelength, also
known as the resonant wavelength, is confined in circular orbits at the device periphery [23].The
photon lifetime inside the cavity is proportional to the quality factor (Q) of the device. In high Q
cavities, photon lifetimes in excess of 500 ns can be obtained, enabling very strong photon-
135

analyte interactions. The resonant wavelength is determined by the refractive index of the device
and the device geometry. Therefore, a change to either parameter results in a detectable change
in the resonant wavelength. These types of cavities can be fabricated from a range of materials
(e.g. silica, silicon, silicon nitride) [24-30]. Previous work has demonstrated their utility in
biological and chemical detection applications as well as material characterization [31-35].
In the present work, CNTs were combined with an optical resonant sensor platform to
form an active substrate capable of detecting the temperature-dependent dissociation of CO and
CO
2
. In this configuration, the CNTs are the substrate while the optical sensor provides a readout
mechanism.
6.3 Device Fabrication
First, the silica microspheres are fabricated, as described previously [27]. The outer
polymer protective layer is removed from the entire stem to avoid possible melting at higher
temperatures (Figure 6-1a). The silica surface is cleaned with isopropanol and melted with the
CO
2
laser (Figure 6-1b). The residual stem allows for easy maneuverability of the device.
Fabricated spheres range in size from 190 to 210 m in diameter. A rendering of a spherical
resonant cavity with an excited mode is shown in Figure 6-1c. The tapered optical fiber
waveguide, which couples light from the laser into the cavity, is also shown.
It should be noted that the final device size is determined by the initial fiber diameter.
While the diameter is related to the resonant wavelength, because this technique is a differential
detection method, a change in the resonant wavelength is monitored, not the absolute
wavelength. Therefore, as long as a device supports resonant modes, it can be used as a sensor.
136

Figure 6-1 Overview of the spherical resonant cavity fabrication process. a) Optical microscope
image of the optical fiber tip after the polymer cladding has been removed and the end-face has
been cleaved. b) Optical microscope image of the fabricated cavity after the CO
2
reflow process. c)
Rendering showing the sphere with its fiber optic stem, the tapered fiber waveguide, and an excited
mode within the resonator [36].
The multi-walled carbon nanotubes used are >95 wt. % purity (Nanostructure and
Amorphous Materials, Inc, Texas, USA). The inner and outer diameters are reported by the
manufacturer as 2-5 nm and <8 nm, respectively. The lengths range between 10 m and 30 m.
Multi-walled carbon nanotubes have multiple layers of graphite in tubes rather than their single-
walled counterparts. This makes them more stable and less fragile.
To deposit the CNTs on the surface of the silica sphere, the CNTs are suspended in a
reagent alcohol consisting of 89-91% ethanol, 4-6% methanol and 4-6% of isopropyl alcohol
(BDH, distributed by VWR). Immediately before the deposition process, the CNT solution is
sonicated for five minutes using an Elmasonic S15H sonicator (Germany) to thoroughly disperse
the CNTs in the solution. In initial experiments, the concentration of the CNT solution was
varied from 0.3 to 3 mg/mL to study the effect of the CNT concentration on the quality factor.
The concentration directly affects the surface density on the spherical cavities. In parallel, the
optical absorption spectra of all of the CNT solutions are measured using a UV-Vis
spectrophotometer (Beckman).
137

The spheres are suspended in the solution for 15 minutes to allow the CNTs to deposit on
the surface. The nanotubes naturally stick to the glass surface. The spheres are subsequently
taken out of the solution and any remaining ethanol is allowed to evaporate off the surface of the
silica spheres.
6.4 Initial Experiments
Using a tube furnace, the spheres with CNTs on the surface were heated under nitrogen
for a few hours. This method is used to desorb any gas that may be on the CNTs to begin with.
The heat helps to release the adsorbed gas and the nitrogen carries it away. The nitrogen also
keeps the CNTs from burning if high temperatures are applied.
After heating, the spheres are removed from the furnace and are immediately tested on
the setup to see if any resonance shift occurs. Figure 6-2 shows results of testing a microsphere
with CNTs (black line), as well as a sphere without CNTs on the surface (red line). Both spheres
were heated to 230
o
C in the furnace before testing. The CNT-covered device shows a large shift
in the resonance, whereas the bare sphere shows negligible shift. The initial decrease in
wavelength shift in both curves could be due to the cooling of the spheres as they had just been
taken out of the hot furnace. The black line is much noisier due to a decrease in the quality
factor, making the resonance broader, and introducing noise into the minimum of the peak. The
gases being adsorbed are most likely water molecules from the air, and minute quantities of
carbon dioxide and carbon monoxide.
138

Figure 6-2 Resonance shifts resulting from CNTs adsorbing gas from the air.
The gas adsorption is determined by the boiling point of the gas. The higher the boiling
point, the more likely the gas molecules will adsorb to the CNTs. This is because after the gas
adsorbs to the carbon nanotube furnace, it resembles its liquid state, therefore, the easier the
transition is, the more likely they will adsorb. The process is exothermic and very little energy is
required to start. Gases with very low boiling points, such as nitrogen, will not adsorb to CNTs
under ambient conditions; whereas, gases such as CO, CO
2
and water will adsorb readily.
Additionally, gases that more readily adsorb to the CNTs are more difficult to remove because
the activation energy to transition from liquid to gas must be overcome.
6.5 Verifying the Surface Coverage
Below are some scanning electron microscope (SEM) images taken at the Center for
Electron Microscopy and Microanalysis (CEMMA) at USC using the JSM-7001F-LV SEM
(Figure 6-3). The SEM is not equipped to take images of such small structures, and individual
nanotubes cannot be seen. Only the clumped structures of the CNTs can be resolved.
139

Figure 6-3 CNT clusters on silica spheres under the JSM SEM. The CNT structure cannot be
resolved with this instrument.
Additional analysis was performed on the JSM SEM in the form of energy-dispersive X-
ray spectroscopy (EDS). This method is used for elemental analysis of a sample. The sample is
excited by a beam of charged particles with high energy, such as electrons or protons. An X-ray
beam can also be used as the source. In response, the electrons of the material will be excited
into higher energy level shells. Upon relaxation back to the ground state, the atoms may emit X-
rays, which are detected by the instrument. Because the X-rays contain information about the
difference in energy between shells, the elements of the material can be deduced.
The JSM-7001F SEM uses an electron beam for EDS measurements. When the
instrument was focused on a spot that did not contain CNTs, the EDS showed the following
140

elements present: oxygen, aluminum, silicon, and gold. The oxygen and silicon make up the
silica device. The holder on which the sample was placed is made of aluminum. The gold peak
was small, but also comes from the holder, since gold-covered devices are frequently imaged on
the SEM. The holder elements can be detected due to scattering of the excitation beam.
In contrast, when the SEM was focused on two different CNT clumps, a carbon peak
appeared in both spectrums. When looking at a large clump, the carbon peak was bigger than the
gold peak. However, when looking at a smaller clump, the carbon peak was a bit smaller than the
gold peak, but still identified by the instrument. While this method could very well detect a
carbon presence, the SEM could still not resolve the CNT structure.
To better verify and prove the deposition of the CNTs, the silica resonant cavity devices
were imaged using a higher resolution SEM, the Hitachi S-4800, which is co-owned by Professor
O'Brien and Professor Dapkus. Figure 6-4(a-c) shows different magnifications of a single CNT
cluster on the surface of the silica sphere for a 1mg/mL solution. The SEM is able to distinguish
individual tubes within the cluster. Based on an analysis of the SEM images, the CNT clusters
cover approximately 1-2% of the sphere surface. However, this low density does not negatively
impact the efficacy of the CNT cluster cavity sensor in the present studies because the exposed
silica surface is not reactive to either CO or CO
2
. This stability is verified in control experiments.
141

Figure 6-4 Scanning electron microscope images of the silica sphere with CNT clusters. a) Image of
the sphere surface. The clusters are barely identifiable. b) Magnification of the region indicated in
part a. c) Magnification of the region indicated in part b [36].
6.6 Controlled Gas Adsorption
Similar to temperature, UV has been shown to regenerate certain materials. As a quick
check, UV light effects were studied in their ability to desorbs light from the CNTs. Using
spheres covered with CNTs, the resonance peak was tracked as it was exposed to UV. After each
UV pulse, the resonance returned to the original position, indicating that gas does not desorb
from UV exposure, independent of UV intensity. Therefore, a furnace had to be used to
regenerate the CNT surface.
In addition, instead of adsorbing unknown gases from the atmosphere, the CNTs were
saturated with a known gas of interest. The gases used during the CNT saturation process are
from Gilmore (El Monte, CA, USA). The argon is 99.996%, the carbon monoxide (CO) is
99.9%, and the carbon dioxide (CO
2
) is 99.99% pure. To controllably adsorb CO or CO
2
to the
CNT surface, the CNT surface must first be activated or regenerated by removing the water that
binds immediately upon the exposure of CNTs to the ambient environment. To accomplish this,
the CNT-loaded spheres are thermally annealed at 100
o
C for ten hours in a tube furnace
(Lindberg) under a continuous argon flow (30 sccm).
142

After the sphere is returned to room temperature within the furnace, still under
continuous argon flow, the gas flow is changed from argon to the gas of interest (CO or CO
2
) at a
flow rate of 20 sccm. To verify that the CNTs are saturated, the exposure time is varied from 2.5
to five hours. The devices are then removed from the tube furnace, and the silica spheres are
immediately tested.
6.7 Sensor Characterization
The present work has two distinctly different, yet inter-related, components: 1)
development and characterization of a CNT nano-cluster optical device and 2) demonstration of
its application in distinguishing between CO and CO
2
. Given that the CNT cluster functionalized
cavity is a new device architecture, it is necessary to characterize its basic operational behavior
before moving forward with the detection experiments.
The resonant cavities are characterized on a custom testing setup (Figure 6-5a) described
previously. Light from a 635 nm tunable laser (Newport, Santa Clara, USA) is coupled into the
sphere through a tapered optical fiber waveguide [37]. The resonant wavelengths of the cavity
are identified by modulating the laser wavelength using a triangle wave generated by a function
generator PCI card and monitoring the transmission on the oscilloscope for characteristic dips,
indicating that power is coupled into the cavity. The spectra are fit to a Lorentzian, and the full-
width-half-max, or linewidth ( ), is directly related to the loaded quality factor (Q) of the
device, according the expression: Q=   , where  is the resonance wavelength.
The loaded Q is comprised of intrinsic loss and extrinsic loss. The intrinsic loss is related
to the inherent loss of the cavity (e.g. material loss, surface roughness), and the extrinsic loss is
related to the system loss (e.g. coupling loss) [27]. By coupling in low amounts of power and
using high efficiency waveguides, the loaded Q can be approximated as the material-limited Q
143

(Q
mat
) [37, 38]. The analytical expression for Q
mat
is: Q
mat
=2 n
eff
/ , where n
eff
is the effective
refractive index,  is the resonant wavelength, and  is the effective material loss. Given that the
CNT concentration scales linearly with material absorption in the lower concentration range, it is
expected that Q
mat
will scale as 1/[CNT]. Therefore, by measuring the Q over a range of CNT
concentrations, the device can be verified as behaving according to conventional optical cavity
physics. The quality factor is measured at both 633 nm and 765 nm, and the sensing experiments
are performed at 635 nm.
The testing set-up is slightly modified for the second series of experiments. Specifically,
a temperature stage is integrated into the previously described testing set-up (Figure 6-5b). The
temperature stage consists of a heating element with a thermocouple (TC) controlled feedback
loop [39]. The temperature stage has a resolution of 0.1
o
C and cooling occurs through convection
alone.
To perform the detection experiments, the cavity surface is saturated with the appropriate
gas and then is mounted on the temperature stage between two glass slides using double-sided
carbon tape. It is important to suspend the sphere in the air so that it is not in contact with any
surfaces, which would negatively impact its optical performance and its sensing ability. To
reduce ambient thermal fluctuations due to random air currents, a cover slip covered in foil is
mounted on top of the glass slides, effectively creating a heating chamber. The sides are left open
for the waveguide to pass through.
144

Figure 6-5 a) Schematic of the testing setup with the major components indicated. PD is the
photodetector, F-Gen is the function generator, and O-scope is the oscilloscope. The PCI cards are
integrated directly into a computer and are controlled using a LabView interface. b) Schematic of
the temperature stage with the silica sphere and fiber waveguide shown. TC is the thermocouple
[36].
Once the resonant wavelength is found, the temperature is increased from room
temperature (20
o
C) to 100
o
C in steps of 0.5
o
C. The heating is done incrementally to allow the
resonance to stabilize between temperature changes. To track the resonance wavelength, an
automated LabView program is used that records the resonant wavelength position at pre-defined
intervals for a fixed period of time.
In the present series of experiments, the temperature and the resonance shift are recorded
for approximately two minutes as they reach stability. From this information, the change in
temperature ( T) and the resonance shift ( ) can be calculated for each step. As mentioned
previously, the resonant wavelength is determined by the device geometry (radius) and the
refractive index (n). Both of these parameters exhibit temperature dependent behaviors according
to the following expression [25]:
145









 


eff
eff
eff
n dT
dn
T
1
0
 

(6.1)
where 
o
, 
eff
, dn
eff
/dT, and n
eff
are the initial resonant wavelength, effective expansion
coefficient, effective thermo-optic coefficient, and effective refractive index of the cavity. In
silica, the expansion coefficient ( 
eff
) is nearly an order of magnitude smaller than the thermo-
optic coefficient; as such, the above expression reduces to:

eff
eff
n dT
dn
T
1
0





(6.2)
Additionally, given that the change in geometry plays a minimal role in the overall shift, the 
can directly be converted to a n using the expression:

eff
n n
0

 
 
(6.3)
where n
eff
is the refractive index of the device, 

is the resonance wavelength, and  is the
resonance shift. The change in the refractive index ( n) is plotted as a function of the change in
temperature ( T). The slopes can then be directly compared, which are the dn/dT values, or
thermo-optic coefficients, in all data sets.
6.8 Characterization of CNT-Covered Device
To thoroughly determine the CNT cluster-induced response, all of the measurements and
subsequent data analysis were performed using both a CNT cluster functionalized resonant
cavity and a bare silica resonant cavity. It is important to note that these results are consistent for
all of the CO and CO
2
annealing times. As such, the CNTs can be considered to be saturated.
Given this consistency, only the three hour saturation time results as given as representative data.
146

The bare silica resonant cavities used in the experiments have Q’s in the range of 10
7
-10
8
.
Figure 6-6a shows a representative Q of a bare silica device, which was calculated as 1.85x10
8
.
Previous work has shown that the Q of silica spheres with large diameters is ultimately limited
by the material loss from a monolayer of water present on its surface [27]. However, in imperfect
devices, surface roughness can also play a role.
The quality factors of the cavities are characterized at both 633 nm and 765 nm.
However, over this narrow range, minimal wavelength-dependent behavior is observed and
similar Q’s are obtained. To enable direct comparison with the bare sphere, a representative
resonance peak of a CNT cluster covered cavity at 765 nm is shown in Figure 6-6b. The CNT
solution concentration used for this device is 1 mg/mL. It is important to note that the general
shape of both spectra in Figure 6-6 are symmetric, indicating that the devices are not undergoing
any non-linear behavior, such as thermal bistability or opto-mechanical vibrations, which could
interfere with the subsequent detection measurement [25, 32].

Figure 6-6 Representative transmission spectrum of a resonance peak of a) a bare silica resonant
cavity with a Q of 1.85x10
8
and b) a CNT cluster-covered resonant cavity with a Q of 4.0x10
6
[36].
147

Figure 6-7a shows the results from the systematic study of the effect of the initial CNT
concentration on the final cavity quality factor. The data is fit to an equation of the form y=ax
b
,
and, from the fit, b = -1.5 with an R
2
of 0.957. As such, the devices are clearly exhibiting
material limited loss. Therefore, while surface scattering might contribute slightly, the dominant
loss mechanism is the increase in material loss from the CNT-clusters. Material loss dominated
Q factors have been observed in previous work using metal nanoparticle-coated silica devices
[40].
The results from the UV-Vis measurements are shown in Figure 6-7b. At low
concentrations, the optical absorption scales linearly; however, the absorption saturates at
approximately 1 mg/mL. Based on previous work [41, 42], the flattening out in the absorption
spectrum indicates that the solution is composed of nano-clusters of CNTs, rather than individual
tubes.

Figure 6-7 a) As the CNT concentration increases, the quality factor decreases, due to increasing
material loss. A line of the form y=ax
b
is fit to the data with a b value of -1.5 [36]. b) UV-Vis
measurements of different concentrations of CNTs. The response is linear in the low concentration,
but saturates at about 1 mg/mL.
148

These results show that it is possible to pre-determine the Q by controlling the
concentration of the CNT clusters. However, the collection efficiency or limit of detection is
directly related to the CNT cluster concentration. Therefore, it is necessary to optimize these two
competing parameters. A related consideration is that as the Q decreases, it becomes more
challenging to identify and to track the resonant wavelength, resulting in noisier data.
6.9 Response in Ambient Environment
To determine the effect of the CNT clusters on the silica cavity thermo-optic coefficient
(dn/dT), the dn/dT of the cavity with and without CNT clusters was measured. These devices
were only exposed to, and thus saturated with, atmospheric gases, most likely water, given its
high boiling point and abundance in the atmosphere. Therefore, these devices represent a
baseline reading. To compensate for the increase in noise due to the reduced Q, data was taken
more frequently (with smaller temperature steps) for the CNT cluster-covered sphere. The
temperature stage was calibrated by using ice water to set the zero point. The PID values of the
controller were also reset using the built-in Auto-PID function.
Figure 6-8a shows representative detection measurements during which the sensors are
exposed to a temperature change of 0.5
o
C. The CNT covered sphere has a larger response than
the bare sphere. Figure 6-8a inset shows the histogram of the noise distribution for the CNT
cluster and the bare silica sensors. The 3  for the CNT cluster (0.2302 pm) is significantly
larger than the bare silica (0.1186 pm). This degradation in performance is related to the
decrease in Q discussed previously.
Figure 6-8b summarizes all of the resonant wavelength shift experimental results for the
bare and CNT-cluster spheres and the values are summarized in Table 6-1 as ambient
measurements. To determine the dn/dT, the  has been converted to n, and dn/dT is simply the
149

slope of this graph. Based on these results, the dn/dT of the bare silica device is 1.19x10
-5
(
o
C
-1
),
which is consistent with the dn/dT of fused silica [43]. In contrast, when the CNT clusters are
present on the silica device, the dn/dT value increases to 1.31x10
-5
(
o
C
-1
).

Figure 6-8 Comparison between bare silica and CNT-covered spheres. a) Resonance shifts for a
step of 0.5
o
C. Inset: Histograms of the noise in the sensing signal generated by the bare SiO
2
and
CNT cluster-covered SiO
2
spheres. b) dn/dT data points for bare and CNT-covered spheres with the
linear fits shown. Temperature is the final temperature minus the starting temperature (20
o
C)
[36].
6.10 Carbon Monoxide Desorption Measurements
The first gas examined was carbon monoxide (CO). CO has a low boiling point of -192
o
C
[44]. Therefore, it is expected that it will readily adsorb and desorb from the CNT clusters.
During testing, the bare silica and the CNT-cluster cavities are heated to 100
o
C in 3
o
C
increments, and the resonant wavelength is monitored. Then, the stage is allowed to cool to
room temperature, and the heating process is performed one additional time for a bare silica
device, or two additional times for the CNT-cluster covered device.
150

The temperature dependent shift results for the SiO
2
and the CNT cluster cavities are
shown in Figure 6-9a and Figure 6-9b, and the values are summarized in Table 6-1. The slope of
the first heating cycle of the CNT cluster cavity is lower than the other two, signifying
desorption of the CO gas. In contrast, the slopes of the SiO
2
device are nearly identical to those
in Figure 6-8b, indicating that CO is not absorbed to the silica during the annealing process.
Therefore, the signal detected in Figure 6-9b is solely from CO desorption from the CNT-
clusters, and not the underlying silica. Additionally, the final dn/dT values for the CNT cluster
device agree with the values measured for CNT devices exposed to atmospheric gases, indicating
that the surface is once again primarily saturated with water. As the CO desorbs from the surface,
water vapor is free to adsorb to the CNTs.

Figure 6-9 n vs. T of heating cycles for CO-saturated a) silica and b) CNT cluster cavity sensors.
Linear fits to experimental data are shown [36].
6.11 Carbon Dioxide Desorption Measurements
In contrast to CO, CO
2
has a significantly higher boiling point (-78.5
o
C) [44]. Based on
the thermodynamics of the system (E
des
~ H
abs
), the higher boiling point means that it takes more
energy to desorb the gas from the CNT clusters. Additionally, the gas more readily adsorbs than
151

CO to the CNT clusters. To thoroughly explore this system, the heating sequence was performed
in two distinctly different methods.
In the first series, the control (bare) and the CNT cluster cavities are heated to 100
o
C in
3
o
C increments, and the resonant wavelength is monitored. Then, the stage is allowed to cool to
room temperature as before, and the heating process is performed either one (for bare silica) or
two (for CNT-cluster) additional times. In the second series of experiments, the bare silica and
the CNT cluster cavities are heated to 100
o
C in 3
o
C increments, and then the temperature is
increased and held at 120
o
C for 90 minutes. After, the stage is allowed to cool to room
temperature, and the heating process is performed either one (for bare silica) or two (for CNT-
cluster) additional times without the additional hold at 120
o
C.
The results of the heating cycles for the CO
2
experiment without the 90 minute hold for
both the bare silica and the CNT cluster cavity are shown in Figure 6-10a and Figure 6-10b and
are summarized in Table 6-1. The slope for the silica cavity is unchanged from the control
experiments in Figure 6-8b, indicating that minimal CO
2
absorbed during the annealing process
or desorbed during the sensing experiment. In contrast, the slope of the CNT cluster device
decreases slightly as compared to Figure 6-8b; however, it stays nearly constant throughout all of
the cycles, suggesting that minimal CO
2
is coming off the surface. The slope of the first heating
line is slightly lower than the subsequent two lines, indicating that some CO
2
may be desorbing
at the lower temperature range.
152

Figure 6-10 n vs. T of heating cycles for CO
2
-saturated a) silica and b) CNT cluster cavity
sensors. n vs. T of heating cycles with a high temperature hold for CO
2
-saturated c) silica and d)
CNT cluster cavity sensors. Linear fits to experimental data are shown [36].
Figure 6-10c and Figure 6-10d show the results for the bare silica and the CNT cluster
devices with the additional 90 minute hold. Similar to the previous experiments, the silica cavity
exhibits comparable behavior to the control devices. However, the response of the CNT cluster
devices is significantly different. The first slope is similar to the previous CO
2
experiment, but
the second and third slopes increase to the same value seen in the control experiment. Therefore,
the CO
2
is able to desorb from the CNTs after the high temperature hold.
153

6.12 Direct Comparison of All Measurements
To enable direct comparison, the results from all of the experiments are summarized in
Table 6-1 and in Figure 6-11a and Figure 6-11b. In both graphs, the two dashed lines represent
the slopes from the initial control measurements using bare SiO
2
(green dashed) and CNT cluster
cavities (purple dotted).

Figure 6-11 Summary of dn/dT vs. cycle number for a) bare silica and b) CNT cluster cavity
sensors. dn/dT values from the control experiments are included as the dashed lines [36].
In all measurements with the CNT cluster sensors, the slopes of the first heating cycle are
lower, closer to the bare sphere slope than the CNT-covered slope. However, after one thermal
cycle, the dn/dT values increase. These results are probably due to the fluid nature of the gases
once they adsorb to the CNTs. Since layers of adsorbed gases more closely resemble a fluid than
their gaseous state [45], the density can be considered to explain the observed trend. At their
boiling points, the liquids have the following densities: water: 53.172 kmol/m
3
, CO: 28.299
kmol/m
3
, CO
2
: 26.83 kmol/m
3
[44]. When a more dense material is present on the surface of the
CNTs, as is the case with water, the change in temperature will have a larger effect on the
refractive index of this material, and thus a more pronounced resonance shift is observed. On the
154

other hand, a less dense material, like CO
2
, will not be impacted as much by temperature
changes. The densities of CO and CO
2
are very similar, and the initial dn/dT values observed for
the two gases are similar likewise.
Table 6-1 Evolution of dn/dT values through thermo-cycling of devices [36].
Device Ambient
(dn/dT x10
-5
)
CO
(dn/dT x10
-5
)
CO
2
, no hold
(dn/dT x10
-5
)
CO
2
, hold
(dn/dT x10
-5
)
SiO
2
Cavity
Cycle 1 1.19 1.1933 1.1944 1.1943
Cycle 2 1.1909 1.1943 1.1948
CNT Cluster
Cycle 1 1.31 1.2525 1.2151 1.2348
Cycle 2 1.3143 1.2432 1.3144
Cycle 3 1.3132 1.2488 1.3065

6.13 Step-wise Temperature Response
The step response of the CNTs saturated with CO or CO
2
to temperature changes was
studied. Starting at 23
o
C, the temperature was changed by two degrees to 25
o
C. After the
resonance stabilized, the temperature was changed back to 23
o
C. Then the temperature was
increased to 27
o
C, decrease back to room temperature, increase to 29
o
C, and so on with the
highest set point at 33
o
C. At this point the same steps are performed, but in reverse. The results
are shown in Figure 6-12. For both gases, it is apparent that the resonance shift does not come
back to zero as the spheres are cooled to room temperature. This could imply that minimal gas is
desorbing from the CNTs, changing the overall effective refractive index, and thus the resonance
wavelength. It is important to notice that as the temperature steps are increased, the baseline
resonance gets farther from zero with each step. However, once the steps start to decrease, the
zero points stays level. Therefore, more gas does not come off at these low temperatures.
155

Figure 6-12 Step-wise testing of CNT-covered spheres. a) Response of CNTs saturated with CO
2
. b)
Response of CNT spheres saturated with CO.
The temperature stage has an overshoot if the temperature is set much higher than the
current temperature. At these instances, the temperature overshoots the set point, but then slowly
cools back to the set point. That's why some of the peaks in Figure 6-12 look split at the top. The
graphs showing the details for the third and fourth peaks in Figure 6-12b are shown in Figure
6-13. The resonance was allowed to stabilize before the temperature was changed again.
156

Figure 6-13 Zoom-in graphs of the temperature steps showing an overshoot and then a cooling back
down to the set point at the top of the peak. a) Step from 23
o
C to 29
o
C for CNT sphere saturated
with CO. b) Step from 23
o
C to 31
o
C for the same sphere.
The slopes of dn/dT lines were calculated for these experiments as well. For increasing
temperature steps, or the forward line, the slope for CO-saturated CNTs was 1.215x10
-5
. For the
decreasing steps, or reverse line, the slope increased to 1.226x10
-5
. For CO
2
, the forward and
reverse slopes were 1.18x10
-5
and 1.211x10
-5
, respectively. In both cases, the slope increased for
the reverse line. This could indicate that some gas has desorbed from the surface but is not a
conclusive study. The experiment must be performed at higher temperatures, where the
difference between forward and reverse lines would be more pronounced.
6.14 Conclusion
In conclusion, an active substrate that integrates an optical sensor with a CNT cluster was
developed, enabling direct readout of the temperature-dependent desorption of gases from the
CNT surface. This substrate platform was demonstrated by studying the temperature-dependent
desorption of CO and CO
2
from CNTs. By studying the changes in the dn/dT values over several
heating cycles, the time of desorption of the gas from the sensor surface could be determined.
157

Given its immunity to system leaks and the ability to selectively differentiate between
different types of gases, this approach has advantages over the commonly used volumetric
method of detection [46]. Additionally, unlike the TPD approach which requires vacuum testing
chambers, this method is able to operate in ambient environments [47]. Theoretically, a similar
analysis method to that used in TPD measurements can also be applied in this approach.
Namely, by evaluating the slope of the dn/dT line at each temperature point, the desorption
temperature can be identified as the point where the slope changes significantly. The
development of the full theoretical paradigm is currently ongoing. Therefore, this type of active
substrate platform is the first step in the development of a small TPD-like system and could
accelerate surface science investigations, enabling the development of improved materials in a
wide range of applications.
References
1. Chen, C.Y., H.X. Li, and M.E. Davis, Studies on mesoporous materials. 1. Synthesis and
characterization of MCM-41. Microporous Materials, 1993. 2(1): p. 17-26.
2. Dillon, A.C., et al., Storage of hydrogen in single-walled carbon nanotubes. Nature,
1997. 386(6623): p. 377-379.
3. Gong, J.L. and C.B. Mullins, Surface Science Investigations of Oxidative Chemistry on
Gold. Accounts of Chemical Research, 2009. 42(8): p. 1063-1073.
4. Henderson, M.A., et al., Interaction of molecular oxygen with the vacuum-annealed
TiO
2
(110) surface: Molecular and dissociative channels. Journal of Physical Chemistry
B, 1999. 103(25): p. 5328-5337.
5. Henry, C.R., Surface studies of supported model catalysts. Surface Science Reports,
1998. 31(7-8): p. 235-325.
6. Kahlich, M.J., H.A. Gasteiger, and R.J. Behm, Kinetics of the selective CO oxidation in
H-2-rich gas on Pt/Al
2
O
3
. Journal of Catalysis, 1997. 171(1): p. 93-105.
7. Kundu, S., et al., Thermal Stability and Reducibility of Oxygen-Containing Functional
Groups on Multiwalled Carbon Nanotube Surfaces: A Quantitative High-Resolution XPS
and TPD/TPR Study. Journal of Physical Chemistry C, 2008. 112(43): p. 16869-16878.
8. Lee, I., et al., Tuning selectivity in catalysis by controlling particle shape. Nature
Materials, 2009. 8(2): p. 132-138.
9. Pevida, C., et al., Surface modification of activated carbons for CO
2
capture. Applied
Surface Science, 2008. 254(22): p. 7165-7172.
10. Zheng, Y.G., et al., Interaction of ammonia with intrazeolitic silver ions: Development of
an ammonia sensor. Sensors and Actuators B-Chemical, 2014. 193: p. 542-551.
158

11. Li, D.B., et al., Direct in situ determination of the polarization dependence of
physisorption on ferroelectric surfaces. Nature Materials, 2008. 7(6): p. 473-477.
12. Bray, J.M., et al., First-principles reaction site model for coverage-sensitive surface
reactions: Pt(111)-O temperature programmed desorption. Surface Science, 2014. 622:
p. L1-L6.
13. Topsoe, N.Y., K. Pedersen, and E.G. Derouane, Infrared and temperature-programmed
desorption study of the acidic properties of ZSM-5-type zeolites. Journal of Catalysis,
1981. 70(1): p. 41-52.
14. Mehrabani, S., A.J. Maker, and A.M. Armani, Hybrid integrated label-free chemical and
biological sensors. Sensors, 2014. 14(4): p. 5890-5928.
15. Barghi, S.H., T.T. Tsotsis, and M. Sahimi, Chemisorption, physisorption and hysteresis
during hydrogen storage in carbon nanotubes. International Journal of Hydrogen Energy.
39(3): p. 1390-1397.
16. Yildiz, O. and P.D. Bradford, Aligned carbon nanotube sheet high efficiency particulate
air filters. Carbon, 2013. 64: p. 295-304.
17. Wang, H., et al., Development of stretchable membrane based nanofilters using patterned
arrays of vertically grown carbon nanotubes. Nanoscale, 2013. 5(18): p. 8488-8493.
18. Zhang, K. and S.R. Wang, Thermal and electronic transport of semiconducting
nanoparticle-functionalized carbon nanotubes. Carbon, 2014. 69: p. 46-54.
19. Soylemez, S., et al., Development of an efficient immobilization matrix based on a
conducting polymer and functionalized multiwall carbon nanotubes: synthesis and its
application to ethanol biosensors. Journal of Materials Chemistry B, 2014. 2(5): p. 511-
521.
20. Hunt, H.K. and A.M. Armani, Label-free biological and chemical sensors. Nanoscale,
2010. 2(9): p. 1544-1559.
21. Luchansky, M.S. and R.C. Bailey, High-Q Optical Sensors for Chemical and Biological
Analysis. Analytical Chemistry, 2012. 84: p. 793-821.
22. Blair, S. and Y. Chen, Biosensing with optical microcavities. Biomedical Instrumentation
Based on Micro- and Nanotechnology, 2001. 2(22): p. 125-133.
23. Matsko, A.B. and V.S. Ilchenko, Optical Resonators with Whispering-Gallery Modes-
Part I: Basics. IEEE Journal of Selected Topics in Quantum Electronics, 2006. 12(1): p.
3-14.
24. Armani, D.K., et al., Ultra-high-Q toroid microcavity on a chip. Nature, 2003.
421(6926): p. 925-928.
25. Choi, H.-S. and A.M. Armani, Thermal non-linear effects in hybrid optical
microresonators. Appl. Phys. Lett., 2010. 97(22): p. 223306.
26. Gondarenko, A., J.S. Levy, and M. Lipson, High confinement micron-scale silicon nitride
high Q ring resonator. Opt. Express, 2009. 17: p. 11366-11370.
27. Gorodetsky, M.L., A.A. Savchenkov, and V.S. Ilchenko, Ultimate Q of optical
microsphere resonators. Optics Letters, 1996. 21(7): p. 453-455.
28. Shah Hosseini, E., et al., High quality planar silicon nitride microdisk resonators for
integrated photonics in the visible wavelength range. Opt. Express, 2009. 17: p. 14543-
14551.
29. Shankar, R., I. Bulu, and M. Loncar, Integrated high-quality factor silicon-on-sapphire
ring resonators for the mid-infrared. Applied Physics Letters, 2013. 102(5).
159

30. Xu, Q.F. and M. Lipson, All-optical logic based on silicon micro-ring resonators. Optics
Express, 2007. 15(3): p. 924-929.
31. Choi, H.S., S. Ismail, and A.M. Armani, Studying polymer thin films with hybrid optical
microcavities. Optics Letters, 2011. 36(12): p. 2152-2154.
32. Ilchenko, V.S. and A.B. Matsko, Optical Resonators With Whispering-Gallery Modes —
Part II: Applications. IEEE Journal of Selected Topics in Quantum Electronics, 2006.
12(1): p. 15-32.
33. Robinson, J.T., L. Chen, and M. Lipson, On-chip gas detection in silicon optical
microcavities. Optics Express, 2008. 16(6): p. 4296-4301.
34. Hawk, R.M., M.V. Chistiakova, and A.M. Armani, Monitoring DNA hybridization using
optical microcavities. Optics Letters, 2013. 38(22): p. 4690-4693.
35. Washburn, A., L.C. Gunn, and R.C. Bailey, Label-free Quantification of a Cancer
Biomarker in Complex Media using Silicon Photonic Microring Resonators. Analytical
Chemistry, 2009. 81.
36. Chistiakova, M.V. and A.M. Armani, Optical detection of CO and CO
2
temperature
dependent desorption from carbon nanotube clusters. Nanotechnology, 2014.
37. Cai, M., O. Painter, and K.J. Vahala, Observation of critical coupling in a fiber taper to a
silica-microsphere whispering-gallery mode system. Physical Review Letters, 2000.
85(1): p. 74-77.
38. Yariv, A., Universal relations for coupling of optical power between microresonators
and dielectric waveguides. Electronics Letters, 2000. 36(11): p. 321-333.
39. Choi, H.S., S. Ismail, and A.M. Armani, Studying polymer thin films with hybrid optical
microcavities. Optics Letters, 2011. 36(11): p. 2152-2154.
40. Shi, C., H.S. Choi, and A.M. Armani, Optical microcavities with a thiol-functionalized
gold nanoparticle polymer thin film coating. Applied Physics Letters, 2012. 100(1).
41. Wittung, P., et al., Absorption Flattening in the Optical-Spectra of Liposome-Entrapped
Substances. Febs Letters, 1994. 352(1): p. 37-40.
42. Berberan-Santos, M.N., Beer's law revisited. Journal of Chemical Education, 1990.
67(9): p. 757.
43. Weber, M.J., Handbook of Optical Materials. 2003, Boca Raton, FL: CRC.
44. Perry, R.H., D.W. Green, and J.O. Maloney, Perry's chemical engineers' handbook. 7th
ed. 1997, New York: McGraw-Hill. 1 v. (various pagings).
45. Thommes, M., Physical Adsorption Characterization of Nanoporous Materials. Chemie
Ingenieur Technik, 2010. 82(7): p. 1059-1073.
46. Panella, B., M. Hirscher, and S. Roth, Hydrogen adsorption in different carbon
nanostructures. Carbon, 2005. 43(10): p. 2209-2214.
47. Tal-Gutelmacher, E., D. Eliezer, and E. Abramov, Thermal desorption spectroscopy
(TDS) - Application in quantitative study of hydrogen evolution and trapping in
crystalline and non-crystalline materials. Materials Science and Engineering a-Structural
Materials Properties Microstructure and Processing, 2007. 445: p. 625-631.

160

Chapter 7 Photoelastic Detection of Ultrasound and Imaging Techniques
7.1 Overview and Motivation
As a result of its non-invasive and non-destructive nature, ultrasound imaging has found a
variety of applications in a wide range of fields, including healthcare and electronics. One
accurate and sensitive approach for detecting ultrasound waves is based on optical microcavities.
Previous research using polymer microring resonators demonstrated detection based on the
deformation of the cavity induced by the ultrasound wave. An alternative detection approach is
based on the photoelastic effect in which an ultrasound wave induces a strain in the material that
is converted to a refractive index change. In the present chapter, photoelastic-based ultrasound
detection is experimentally demonstrated using ultra high quality factor silica optical
microcavities. A finite element method model that includes both the acoustics and optics
components of this system is developed, and the predictive accuracy of the model is determined.
A few images reconstructed from echoes are also shown.
7.2 Background
The development of non-invasive and non-destructive imaging methods is in high-
demand both in the medical field and in the electronics industry. Ultrasound imaging is one
emerging solution. While there are many variations of acoustic microscopes and imaging
systems, one approach leverages the relative strengths of ultrasound excitation with optical
detection of the ultrasound wave. Using this approach, it is possible to image blood vessels in
tissue in 3D [1], image single capillaries in vivo [2], and identify tumors in breast tissue [3]. All
of these experiments rely on having a sensitive (low noise and high resolution) method of
detecting the ultrasound wave.
161

One ultrasound sensing technique is based on whispering gallery mode optical resonant
cavities. These devices are able to confine light at specific optical frequencies that are defined, in
part, by the device geometry and the refractive index [4]. Therefore, when an ultrasound wave
hits the surface of the cavity, this resonant frequency changes. A schematic highlighting the
general principles of this device is contained in Figure 7-1. In previous work using microring
cavities, the focus has been on optimizing the quality factor (Q) of the cavity, as a higher Q
device has improved sensitivity in response to an impulse [5-7]. This previous work focused on
using the effect of material deformation for detection. In this approach, an incoming ultrasound
wave changes the device radius. As a result, several other parameters also contribute to the
signal. For example, the material constants of the device and its surroundings as well as the
intensity of the initial ultrasound pulse. While polymer microring devices have a strong response
due to their low bulk modulus, they have moderate quality factors that place a fundamental limit
on their ultimate performance.

Figure 7-1 a) The resonant wavelength increases and decreases in response to the ultrasound pulse.
b) If the transmission values at a single wavelength ( 
o
) are plotted, the characteristic damped
oscillator curve is clear. The positive values indicate high pressure and the negative values indicate
low pressure [8].
162

Therefore, to develop the ideal ultrasound sensor, it is necessary to explore alternative
detection mechanisms that include both optical and mechanical changes. However, within the
broader field of resonant cavity ultrasound imaging, there is limited analysis enabling direct
comparison of different material systems and device geometries. Therefore, to move the field
forward, a more rigorous understanding of these inter-dependencies is needed.
In the present work, COMSOL Multiphysics simulations are developed and performed,
using the transient pressure acoustics and RF modules. Subsequently, the model is verified
experimentally using a 40 MHz ultrasound transducer and silica ultra-high-Q spherical cavities.
Given the geometry and material change as well as the three order of magnitude increase in Q,
these devices have significantly improved sensitivity as compared to the previous work with
microrings. Moreover, as a result of the change in device size and material, a distinct detection
mechanism from that observed previously is responsible for the signal generation.
7.3 Photoelastic Theory
The two key parameters that need to be considered when designing an ultrasound sensor
are the sensor sensitivity and the sensor response. The sensitivity is determined by cavity Q
whereas the response is determined by the cavity material.
Sensitivity describes the smallest detectable signal change. In these measurements, it is
more common to measure transmission than wavelength changes. However, as shown in Figure
7-1, as long as the change in wavelength is less than half of the linewidth of the resonance, there
is a linear relationship between transmission and wavelength, and, on the linear portion of the
spectra, the Q and the slope are linearly proportional. As such, an increase in the Q results in an
increase in resolution.
163

As mentioned, the cavity’s resonance wavelength is defined by the refractive index and
the device geometry. Therefore, a change to either parameter will shift the resonant wavelength.
The magnitude of both effects are determined by fundamental material constants; therefore, it is
possible to predict which mechanism will be the dominant one in a given device system.
The refractive index change is dominated by the photoelastic effect. In the photoelastic
effect, the acoustic wave causes a strain in the cavity, which in turn alters the refractive index.
This behavior can be defined by:

S
p n
n
2
3

 
(7.1)
where n is the refractive index in the absence of a sound wave, p is the photoelastic constant, also
known as the strain-optic coefficient, and S is the strain, or relative displacement [9]. The
relation has a negative sign to indicate that a positive strain produces a decrease in the refractive
index. The photoelastic constants (p) are tabulated as a strain-optic tensor [10]. For fused silica,
when the acoustic field and the resulting compressions are in the same direction as the electric
field distribution, signifying TE waves in our configuration, p is 0.121 at 633 nm. For the TM
waves, p increases to 0.270 because the electric field is perpendicular to the ultrasound pulse.
Since TM modes are more commonly used with silica microspheres, the p value of 0.27 was
used. For either electric field orientation, the p value in water is 0.31.
The cavity can also mechanically deform in response to an ultrasound pulse. In this case,
the density changes due to molecular vibrations caused by the sound wave. In regions under
compression, the density increases. This change leads to a localized decrease in the device size
and an increase in the refractive index, simultaneously. The following equation relates the
changing density in a material with an input pressure:
164






 


K
P P
f
) (
1
0
0


(7.2)
where 
0
is the density in the absence of ultrasound, P is the new pressure the material is
exposed to, P
0
is the original or atmospheric pressure, and K is the Bulk modulus.
Since the resonance wavelength is also dependent on the geometry of the device, a
change in radius will likewise produce a shift. Radius change effects are more apparent in
devices with a size similar to that of the ultrasound pulse wavelength and with low K values. As
the pulse hits the device, contraction and expansion will occur and will dominate the resonance
wavelength change in devices with a size similar to the pulse wavelength. In previous work with
polymer microrings, this effect was the dominant detection mechanism [5].
While there are analytical expressions for all of the key physical effects, because the
ultrasound wave is time-dependent, the expressions become increasingly complex. For example,
the pressure (P) is time-dependent. Therefore, in previous work, researchers have relied on
proportionalities or used best fits. However, this approach does not allow for intelligent design
of the ideal sensor device.
7.4 Proof of Concept Experiment
The first test of the validity of this experiment was to detect ultrasound pulses sent
directly at the microsphere. The sphere was suspended in water and the transducer tip was placed
close to it from the side. The transducer used in the experiments is fabricated in the Ultrasound
Transducer Resource Center at the University of Southern California and generates 40 MHz
pulses. To detect ultrasound, a tunable laser was placed precisely at the wavelength of the
resonance where the slope of the Lorentzian is maximum. If the wavelength is set to a lower
slope, the response will not be as pronounced and could be lost in the noise. As the ultrasound
165

pulse hits the device, the resonance peak of the device shifts due to a change in refractive index
of both the material from which the device is made and the material surrounding the device.
A signal was easily detected on the O-scope as shown in Figure 7-2. Since the pulse
generator was set to 20 kHz, a pulse was seen every 50 s (Figure 7-2a). The O-scope was set to
trigger off the changes in the transmission channel and therefore, the first response appears at
zero seconds. Zooming in on the responses at zero and 50 microseconds in Figure 7-2a, a
decrease in the magnitude of the oscillation signal is observed (Figure 7-2b and c). This
corresponds to the pulse shape sent from the transducer. When the transducer was pointed away
from the sphere, no signal was detected, showing that the ultrasound was in fact responsible for
the response shown.

Figure 7-2 Two ultrasound pulses detected by the microsphere. a) The entire optical transmission
spectrum. b) and c) show zoomed in view of the dampened oscillations.
The response lasts for less than a microsecond; therefore, a high speed photodetector
must be used with a high speed digitizer. The digitizer/O-scope PCI card is capable of sampling
at 250 MSamples/second (250 x 10
6
). However, a faster photodetector must be used than is
typically required for other resonator experiments. Such a detector has limitations in that it can
only accept up to 80 W of power as its input. The power coming out of the tapered fiber can be
lowered by decreasing the amount of power sent into the taper. This is done by changing the
166

coupling condition between the angle polished cable (APC) patch cable and the APC connector
on the spool. However, the power coupling into the cavity should be high to get the maximum
response from the sphere. Therefore, to reduce the power between the sphere and the detector, an
attenuator is placed between the output of the taper and the photodetector.
7.5 Optimizing the Experimental Setup
After confirming that the microsphere shows a response for ultrasound pulses, the
experimental setup was slightly modified. The goal was to try imaging samples with the
pulse/echo technique. The transducer had to be pointed at the target to be imaged and the sphere
was to receive the echo that bounces off the target. The sphere and transducer were suspended in
water to allow for ultrasound propagation, since air reflects ultrasound very strongly. The
imaging target was not fully immersed, but just in contact with the water on the opposite side of
where the sphere and transducer were located. The layout of the initial experiments is shown in
Figure 7-3. The top view shows the transducer to the right of the microsphere, which is not
included in the side view diagram. The side view shows two spacers from which the sphere is
suspended above the bottom of the sample holder. These were glass slides cut to fit the size of
the holder. The top of the holder consisted of a cover slide glued to the top glass spacer. The
space between the steel holder and the glass cover slide was filled with distilled water to
maximize the propagation of the ultrasound.
167

Figure 7-3 Top and side view of initial holder setup for imaging in pulse/echo mode.
A few issues arose when trying to take data in this way. The main one was that the sphere
showed no response to the echoes that were expected to hit it. The biggest problem was that the
position of the transducer had to be optimized so that the ultrasound could bounce off the
imaging target at the right angle to reach the spheres. To avoid working with angles, the setup
was redesigned such that the transducer was placed parallel to the sphere, at the same height and
slightly behind. The parallel alignment is important for echo detection since the waves return to
the same place and a slight angle might result in non-uniform data collection.
With the modified setup, a response was easily identifiable on the sphere. The next issue
to arise was the abundance of stray echoes detected by the device. If the chamber material is not
impedance-matched, strong reflections from the boundaries can be generated, significantly
interfering with the primary signal and complicating signal analysis. Ultrasound reflects from
materials depending on the difference in their acoustic impedances. When the difference is small,
more ultrasound is transmitted and less is reflected. On the other hand, a big difference results in
a large reflection. Since the holder is made of steel, the impedance (46.7) is very different than
that of the water (1.48), resulting in most of the ultrasound reflecting from the surface of the
steel. The acoustic impedance can be calculated as the density of the material multiplied by the
168

acoustic velocity through the material and is written in units of g/cm
2
*
s (x10
5
). The reflection
from air is also large since the impedance of air is 0.0004, very different from water.
To meet the impedance-matching requirements, a sample chamber constructed entirely
from polydimethylsiloxane (PDMS) was designed (Figure 7-4a). It was connected to the nano-
positioner using a PEEK rod. These materials are impedance-matched to water. Specifically, the
testing chamber is formed from a pair of thin PDMS sheets separated by a pair of PDMS slabs.
In addition to meeting the impedance requirements, PDMS is transparent and allows for imaging
of the system during the experiment using the top view camera. The stem of the silica
microsphere is attached to the first PDMS slab and the top sheet is attached to the second slab. In
this configuration, the sphere is suspended in the middle of the chamber. With this holder, the
echoes seen in the output file were much less abundant than before.
It was also decided that the imaging object should be fully submerged rather than simply
in contact with the water. The advantage of this layout is that the echo from the air can be
detected separately from the echo of the imaging target. Numerous objects were used as the
imaging object, such as a steel wire of 270 m diameter, silica fibers of varying diameter, and 1
mm diameter steel spheres. To keep the object in place and to easily change its position, the
imaging object was embedded into a PDMS strip and attached to a separate three-axis stage with
its own micro-positioning controllers. A rendering showing the detailed components of the stage
is shown in Figure 7-4b. In this rendering, a steel sphere is shown as the imaging object. It is
placed directly in front of the transducer and microsphere cavity. The focal length of the
transducer is between 1.5-3 mm. Therefore, the object is placed within this distance.
169

Figure 7-4 Optimized sample holder. a) Image of PDMS holder on a PEEK rod. b) Rendering of the
sensor setup [8].
The overall testing setup is shown in Figure 7-5. A narrow linewidth tunable laser
centered at 775 nm is coupled into a tapered optical fiber, which is used to evanescently couple
the laser light into the silica microcavity, as can be seen in Figure 7-4b [11, 12]. The taper is
aligned with the sphere using high precision nanopositioning stages, and the alignment is
monitored using a top view machine vision system (Navitar). The transmission through the
optical fiber is received by a photodetector (PD) and sent to a high speed O-scope/digitizer PCI
card.
During these experiments, the laser is operated in two different modes: 1) frequency
modulated and 2) fixed wavelength. To modulate the frequency, a function generator PCI card
sends a 100 Hz/1 VPP signal to the laser controller. This modulation allows for precise scanning
across a narrow wavelength range (~0.03 nm), which is necessary to measure the quality factor
of the cavity (Q=   , where  is the linewidth). A representative Q spectrum is shown in
Figure 7-5b.
170

Figure 7-5 a) Schematic of the testing setup. b) Resonance peak used in the experiment. The red
dashed line shows the fit line used to convert simulation data [8].
7.6 Control Experiments

To perform an ultrasound measurement, a resonance is first identified by modulating the
laser using the function generator. Subsequently, the laser wavelength is set to the value at the
steepest part of the left side of the resonance. Using a function generation, the transducer and the
oscilloscope are simultaneously triggered. This approach allows the entire response to be
recorded on the oscilloscope. Only 20 s are displayed and recorded, since this range spans more
than the entire testing chamber. For the ultrasound experiments, a high speed detector (150 MHz)
is used. As mentioned, to adjust the power to fall within the operational range of the detector, an
attenuator designed for 850 nm is placed in line, after the cavity and before the detector.
Given the total system performance, the limiting factor in the data acquisition rate is the
oscilloscope PCI card, which can record at 250 MSamples/second for several hours (limited by
the total RAM of the computer). Compared to other data acquisition methods, which rely on
either high speed cables or on-board memory on external equipment, by directly acquiring and
recording onto the computer, significantly faster and/or longer acquisition times are possible.
171

The coupling condition is important to this experiment. The ideal coupling is critical.
Figure 7-6 shows the response from the sphere as an ultrasound pulse is sent at it. The red line at
the top is from an under-coupled regime. The black line at the bottom shows the response when
the sphere is in the over-coupled regime with the taper. The middle blue line the show critical-
coupling response, in which the ultrasound response is easily identifiable.

Figure 7-6 Response from ultrasound pulse for under-coupled (red), over-coupled (black), and
critically-coupled (blue) regimes.
To make sure the system responds as expected, a control experiment is performed. The
transducer is placed directly in front of the silica microsphere, instead of behind it. The
transducer is pulsed and the response in the microsphere is recorded. Figure 7-7a shows three
sample pulses recorded without changing parameters. As can be seen, the signal is stable over all
the recordings. The transducer is able to operate in a transmit and receive mode in which it
records any ultrasound pulses it receives after sending the initial pulse. Therefore, it can behave
as a reference or control sensor. Figure 7-7b shows the signal recorded multiple times by the
transducer as an echo returning from the silica microsphere. The signal is once again consistent
over the separate recordings and is similar to that detected by the microcavity. Both signals in
172

Figure 7-7a and Figure 7-7b show 40 MHz response, consistent with the transducer output. It is
also of interest to note that the microsphere response shows a longer ringing, or more sensitive
output, than the transducer.

Figure 7-7 Control testing of ultrasound response. a) Response from microsphere to a pulse from
the transducer when it is placed in front of the sphere. b) Transducer response to an echo coming
from the silica microsphere [8].
7.7 Optical Simulation Design and Parameters
To develop a generalizable ultrasound response model, the Multiphysics capabilities of
COMSOL Multiphysics 4.3a were leveraged, specifically the acoustics and resonant frequency
(RF) modules. First, the RF module was used to model the mode volume of the resonance under
specific conditions. This model establishes the distribution of the optical mode within the silica
and the surrounding medium (water). Once again, the 2D simulation geometry is set as the
equatorial cross section of a silica microsphere shown in Figure 7-8a. The refractive index of the
water was set to 1.332. The mesh size of the simulation was 0.021 m
2
. Due to the symmetry of
the device, the computation is greatly simplified. Figure 7-8b shows the results of the simulation
with the mode intensity plotted. From this model, mode volume and mode distributions can be
calculated [13]. For the present geometry, approximately, 1.5% of the mode is within the water
173

surrounding the resonator, while the rest is inside the silica. Therefore, all values in Equation
(7.1) and Equation (7.2) must be calculated for both water and silica.

Figure 7-8 a) Microscope image of a silica microsphere. b) RF simulation results showing a small
part of the sphere in a water environment with the mode distribution plotted at V/m
2
[8].
7.8 Acoustic Simulation Design and Parameters
The simulation geometry is shown in Figure 7-9a. All parameters are specified to match
subsequent experimental values. A few of the material properties are modified to account for
ultrasound effects. The density of all materials is changed to the relation shown in Equation
(7.2). The speed of sound ( 
s
) in a material was defined in terms of density ( ) and Bulk modulus
(K) as follows:


K
v
S
 (7.3)
The characteristic acoustic impedance (Z
o
) was also changed to the following:

.
0

S
v Z  (7.4)
All other material constants are defined using the built-in COMSOL library.
174

The 1 mm wide ultrasound source enters the system, defined as liquid water, from the
upper boundary. The source matches the physical 1 mm diameter transducer. Similar to the
experimental conditions, the silica microsphere resonator is placed adjacent to the transducer and
the steel sphere imaging object is directly in front of the transducer. However, to reduce memory
requirements, the separation distance between the steel sphere and the resonant cavity is reduced.
The boundary matching layer (BML) absorbs the ultrasound waves, eliminating reflections and
pressure build-up within the simulation area, which would result in a non-physical result. The
BML is defined as a polyurethane polymer with optimized density, Bulk modulus, and porosity
parameters. Without the defined BML, only planar parts of waves were transmitted through the
boundaries, resulting in too many echoes.

Figure 7-9 a) Geometry of the 2D FEM simulation in COMSOL. b) Ultrasound pulse shape used in
the model [8].
The two red points on the left of the silica microsphere in Figure 7-9a indicate the
location of the pressure monitors which are located on either side of the water-silica interface.
Because the optical cavity is an evanescent field sensor, only changes to the refractive index
which occur within the optical mode volume are detected.
175

The transducer used in the experimental component of the present work exhibited a time-
dependent intensity profile which was a composite of an exponential growth and an exponential
decay. Therefore, to more accurately mimic experimental conditions, a shaped ultrasonic pulse is
defined with this envelope on top of a sine wave of 40 MHz, as indicated in Figure 7-9b. The
amplitude and width was adjusted to match the experimental data. The boundary was set to a
plane wave radiation condition with a defined incident pressure field.
The mesh size is determined by the ultrasound wavelength. Therefore, to maintain 12
degrees of freedom per wavelength, the maximum mesh size is set to three microns. The
minimum mesh size was set as the maximum divided by 20. A free triangular geometry was used
to construct the mesh. The size of the time steps used in the simulation was determined by the
Courant-Friedrichs-Lewy (CFL) condition to be 0.2 ns [14].
7.9 Simulation Results
As seen in Figure 7-10a and Figure 7-10b, the ultrasound waves first interact with the
silica microsphere cavity, and then the steel imaging object. The ultrasound echo from the steel
object is then reflected onto the silica cavity. It is important to note that the BML is behaving as
expected, and there are negligible reflections.
To quantify this behavior, the pressure is evaluated using the previously mentioned
power monitors. Figure 7-10c and Figure 7-10d show the initial and reflected pulses in water and
silica resulting from an initial pulse width of 0.225 s. Over large separation distances, the
amplitude of the echo should decrease due to scattering and absorption losses in the water.
However, to reduce memory requirements, the simulation area is decreased, and these effects
were not observed. The time between the two signals is determined by the distance between the
cavity and the steel sphere and the ultrasound speed in water (1480 m/s), thus allowing a precise
176

measurement of this distance. Additionally, the initial pulse and the echo mirror each other in
relative amplitude and frequency.

Figure 7-10 Simulation results. a) FEM result showing the initial pulse as it passes by the silica
microsphere. b) FEM result for the echo coming from the steel sphere as it reaches the silica
surface. c) Pressure variations recorded with a 0.225 s pulse at the point just outside the silica
surface boundary in the water. d) Pressure variation within the silica material showing the effects
of the initial pulse and the secondary echo [8].
To convert changes in pressure to changes in refractive index, first the pressure-
dependent density was calculated from Equation (7.2). The characteristic impedance (Equation
(7.4)) was computed next, followed by the particle velocity:
177

(7.5)
From there, the sound intensity could be calculated:
(7.6)
Finally, the strain in the medium was computed using the relation:

(7.7)
Each value was calculated for both silica and water, taking into account the different ultrasound
propagation velocity. To preserve the sinusoid shape resulting from the ultrasound pulse, the
strain was multiplied by +/- 1, depending on whether the pressure at that time point had a
negative or a positive sign. The changes in refractive index were then computed using Equation
(7.1).
Based on these simulations and calculations, it becomes evident that there are multiple
parameters, in addition to Q, which should be considered when designing a resonant cavity-
based ultrasound sensor. For example, when using the photoelastic effect for detection, the pulse
length and the bulk modulus plays a dominant role in the fidelity and magnitude of the signal.
Specifically, if the Bulk modulus is increased, the refractive index change will be smaller. The
density also contributes and has a similar effect to the Bulk modulus, in that its increase will
cause a decrease in the overall refractive index changes.
7.10 Converting Simulation Data to Match Experiment
In order to directly compare the simulation results with the experimental results it is
necessary to convert the simulation data into the transmission change that would be recorded
during experiments. The first step is to convert the pressure changes to refractive index changes.
178

The conversion of the simulation data relies on the relationship between wavelength
change and refractive index discussed previously ( = n
eff
/n
eff
+ R/R). In direct contrast to
previous work, because the silica sphere is larger than the ultrasound wavelength and the bulk
modulus is very high, the change in the radius is negligible in comparison to the refractive index
change. Therefore, the expression is simplified to = n
eff
/n
eff
.
By combining the information from the mode volume simulations with the results from
the acoustic modeling, the effective refractive index is calculated by Equation (7.1). Specifically,
the time-dependent refractive index of the material (silica or water) is multiplied by the fraction
of the optical mode in that material. These values are then combined to form the n
eff
, and the
resonance shift is determined.
The last step is to convert the resonance shift into a transmission change. A spectrum is
typically fit to a Lorentzian:

(7.8)
where y
0
is the baseline of the curve, w is the full-width at half-maximum of the peak, x
c
is the
center of the peak, and A can be found by solving the equation at the center:

(7.9)
where y
c
is the y-axis value of the tip of the peak. The equation is modified to the following:

(7.10)
In this equation, 
start
is the starting wavelength within the resonance, the wavelength of the
steepest part that the laser is set to. The center of the peak shifts as the pressure changes.
Therefore, x
c
becomes the argument in the parentheses. The 
0
is the original center wavelength
179

when no ultrasound is present. The effective refractive index, n
0
, is calculated when no
ultrasound is present as well. Finally, n, is calculated by Equation (7.1) as described. All other
values are found from the resonance peak used in the experiments.
Trying to convert the simulation data using the Lorentz equation produced some
unwanted results. For example, the denominator produces such small values that division leads
to very large values of the fraction overall. As a result, the calculated transmission graph looks
nothing like what is obtained from experiments.
To avoid the complexities of the Lorentz equation, only the linear portion of the peak was
considered instead for the relationship between transmission and wavelength. As indicated in
Figure 7-5b, the laser was fixed to the left side of the peak; therefore, this side was fit with a
linear equation with >98% agreement. The linear fit is shown in Figure 7-5b as the red dashed
line. Additionally, the center wavelength of the resonance is known from the laser controller. To
correctly attribute a transmission value for a corresponding wavelength shift, the shift is
subtracted from the starting wavelength, and the transmission is calculated at that point. This
relationship results in a conversion that is directly comparable to experimental data.
7.11 Experimental Results and Analysis
To verify the simulations, a series of experiments was performed using an ultra-high-Q
silica microsphere resonant cavity. Results from a 192 m diameter cavity with a loaded Q of
9.5x10
7
are shown in Figure 7-11. In Figure 7-11a, three different signals are clearly identifiable.
The initial signal is from the pulse leaving the transducer and passing by the microsphere. The
second set of oscillations is due to the echo from the steel sphere. The last signal is the echo
returning from the air-water interface behind the steel sphere. Based on the time between the
initial pulse and the two subsequent pulses and the speed of ultrasound in water, the distances
180

can be calculated. Specifically, the front face of the steel sphere is approximately 2.2 mm away
and the air-water interface is approximately 3.9 mm. These values agree with the experimental
design.
One interesting point is that the second steel-water interface is not detected. However, if
one looks at the simulations, it is clear that the amount of energy that enters the steel is small
because of the large difference in acoustic impedance between steel and water. Additionally, any
ultrasound that entered the steel sphere does not leave because of the large loss or attenuation of
the wave as it propagates through the steel. Therefore, any generated signal is too small to detect.

Figure 7-11 Experimental data. a) The entire signal received by the silica microsphere. b) Zoomed-
in graph of the steel echo. c) Zoomed-in echo from the water-air interface [8].
Figure 7-11b and Figure 7-11c show the echoes from the steel sphere and the air
interface. The echo from the steel sphere is not as clean as the theoretically expected signal;
however, the echo from the air interface is very clear. It can be noted that the baseline signal
between the initial pulse and the steel echo in Figure 7-11a is much noisier than the rest of the
baseline. This is due to the long low-amplitude ringing of the ultrasound pulse that lasts longer
than a microsecond. Because of this ringing, the microsphere response during the time the echo
returns from the steel sphere is a combination of the continued ringing of the initial pulse and the
returning echo. The combination of the two signals produces the non-ideal shape seen in Figure
181

7-11b. By the time the echo from the air interface returns, on the other hand, the ringing has
stopped and the microsphere has only one response resulting from the echo alone.
Additionally, and to a smaller degree, some interference occurs from the echoes coming
from the boundaries that exist within our chamber. Even with careful attention to impedance
matching, extra echoes could still be produced that interfere and change the total wave as it
returns to the resonator.
7.12 Comparing Simulation and Experimental Data
Figure 7-12a shows an overlay of experimental and simulation results. In this specific
simulation, the pulse length was 0.25 s. While there is general agreement, a more quantitative
analysis of the predictive accuracy is important. There are several ways to quantitatively
characterize the predictive nature of a model. One approach, based on computer science analysis
methods, is to calculate the Accuracy. In this method, the true positives (TP), true negatives
(TN), false positives (FP) and false negatives (FN) are determined, and then the Accuracy is
calculated using the standard formula:

.
FN FP TN TP
TN TP
Accuracy
  

 (7.11)

All quantities are defined with respect to the experimental results. Specifically, a true
positive (TP) is if the simulation correctly predicts the presence of an experimental peak while a
false positive (FP) is if the simulation predicts a peak that did not occur in the experiment. Both
the peaks and the troughs are considered “peaks”. A true negative (TN) occurs if both the
simulation and experimental values are on the zero axis. Similarly, a false negative (FN) is if
only the simulation data is zero. For example, the accuracy of the simulation shown in Figure
7-12a is 58%.
182

While there are several parameters that control the accuracy, one key value is the pulse
width in the simulation. Keeping the pulse shape the same, the pulse duration used in the
simulations is changed from 0.1 s to 1 s, and the accuracy analysis detailed above is
performed (Figure 7-12b). Based on this analysis, there are clearly additional ways to improve
the predictive. For example, additional parameters that contribute to the microcavity response
are: the shape of the ultrasound wave, the wave propagation distance, and boundary reflections.
In the current simulation, BMLs are used in conjunction with plane wave radiation with scaled
distances. Therefore, some aspects of the experiment are not fully captured.

Figure 7-12 Comparing experimental and simulation results. a) Overlapping the experimental data
(black) with simulation data (red). b) Accuracy as a function of pulse width specified in the
simulation [8].
7.13 Ultrasound Imaging Techniques
After establishing the theoretical groundwork on ultrasound detection with silica
microspheres, imaging analysis experiments were performed. To collect data for image
formation, the imaging target was moved rather than moving the sphere, the taper, and the
transducer, which must all be kept together. At first, the step sizes were controlled with a manual
183

actuator that has marking at every 10 m. However, the difficulty of controlled step sizes and the
jarring movement produced excessive noise. Therefore, motorized actuators were used that had
controllable speed allowing for smooth movements. The imaging target was moved in front of
the microsphere and an echo transmission graph was recorded at each position. Each
transmission graph is called an A-line plot.
The first step in analysis is to normalize all the data by dividing each A-line scan by its
mean value. When recording data, the transmission can vary slightly because the position of the
sphere relative to the taper could change. The data before normalization is shown in Figure
7-13a. The initial pulse and an echo can be barely identified in the figure because the baseline for
each A-line scan is different. After normalization, the initial pulse and echo are more pronounced
in the data (Figure 7-13b).

Figure 7-13 Time domain data shown a) before and b) after normalization.
The data is then imported as a matrix into MATLAB and the "imagesc" function is
implemented. This function plots the magnitude of each data point on a grey scale. The points
with the largest amplitude will be closest to black, while the minimum amplitude will be closer
to white. The average values in the data will be the average grey color. The output of the
184

function is a 2D plot of these colored points. The y-axis of the imagesc plot corresponds to the
time axis in the A-line scans. However, the units are counts, rather than seconds. The recording
frequency is 250 MHz, therefore one count corresponds to four nanoseconds. Knowing the speed
of sound in water, the axis can then be changed to distance. The x-axis in the imagesc plot is also
in counts but corresponds to each A-line scan. Usually around 100 A-line scans are made for this
image analysis. If the imaging object is moved 10 m between the scans, the x-axis would have
step sizes of 10 m for each count.
An example of the output of imagesc is shown in Figure 7-14. The resulting data is
presented as un-normalized and normalized data to show how normalization smoothes out the
color plot. The imaging object is the steel sphere embedded in a PDMS sheet. The object was
moved from left to right in front of the silica sphere and transducer with steps of 20 m. Two
features can be made out in Figure 7-14, the PDMS surface, which is at relatively the same
distance from the silica throughout and the steel sphere with large varying distance. This result
makes sense since the steel sphere is embedded in the PDMS and as it is moved left or right, the
front face of the PDMS is still at the same distance away. However, as the sphere is moved
around, the echo is closer or further since the travel time varies. It is important to note that the
distance on the y-axis is the round-trip distance from the detector (silica sphere) to the imaging
object. The zero point distance is set as the time of the center of the initial pulse.
The results indicate that the ultrasound pulse diverges as it leaves the transducer since an
echo is detected when the steel sphere is not directly in front of the transducer and silica sphere.
An echo can be detected even when the steel sphere is 2 mm out of the direct plane of the silica
sphere. Additionally, it is apparent that the PDMS used in experiments was not as well
impedance-matched as it could be.
185

Figure 7-14 Imagesc plots showing the data a) without normalization and b) with normalization.
The distance in the y-axis is round-trip distance from the silica sphere. The x-axis distance is how
far the imaging object was moved from left to right. Note: these figures do not represent an image,
but a series of A-line scans.
Similar results can be obtained from a wire as the imaging object. Since the wire is long,
it can be attached to a holder without requiring PDMS. The wire was moved up and down in
front of the silica sphere and transducer, rather than left and right. The imagesc representation of
the A-line scans is shown in Figure 7-15. The x-axis here is the distance the wire was moved in
the up/down direction relative to the silica sphere. The y-axis is still the round-trip distance away
from the sphere.
186

Figure 7-15 Imagesc reconstruction of A-line scans from a wire of 270 m diameter. The x-axis
distance is the distance the wire was moved in the up/down direction.
There are a few things to notice in this figure. The first is that there are many extraneous
echoes appearing. Since the echoes vary in position as the wire is moved, it seems that most of
these extra echoes are a result of the wire. As the ultrasound pulse hits the wire, some ultrasound
is reflected while a fraction is transmitted. Since the wire diameter is larger than the wavelength
of ultrasound, the wave can travel within the wire easily. At every boundary between the wire,
some sound is again transmitted and some is reflected. As the sound waves travel around the
water chamber, the silica sphere picks up all the resulting echoes.
The second thing to note is the curvature seen on the figure. The curvature is due to the
wire being closer and farther from the microsphere as it was moved. To better get a sense of how
the curvature comes about from the position of the wire, the layout is diagrammed in Figure
7-16. The diagram is shown as a side-view with the wire represented as a dot since the length of
it would lie in and out of the page. From the x-axis of Figure 7-15, it is seen that the wire was
moved about 3.25 mm in total. When the wire was closest to the silica sphere, it was about 3.4
187

mm away (half the distance on the y-axis of 6.8 mm in Figure 7-15). By basic geometry, the
largest distance the wire will be from the sphere is about 3.8 mm, or 7.6 mm round-trip distance,
which is the exact distance the furthest echo appears in Figure 7-15 at point 0.

Figure 7-16 Schematic showing how the wire (red circle) is 0.4 mm farther from the microsphere
when it is at the top or bottom of the liquid chamber compared to when it is exactly in the center.
7.14 Axial Resolution Experiment
The axial resolution was analyzed with the same wire. The wire was placed in front of the
sphere at a distance of about 2.8 mm and moved forward away from the sphere with steps of 10
m. The detected distance of the wire was found by looking at the maximum intensity echo in
each A-line scan. The agreement between the set location and the detected location was very
good (>99.9%). Figure 7-17a shows the imagesc output as the wire is moved forward. The x-axis
is the distance the wire was moved forward (away from the sphere). Since the y-axis is also
distance from the sphere, it is clear to see that the wire moved from 5.6 to 7.6 mm round-trip,
indicating agreement of a total of 1 mm movement. In this figure it is also easy to see a
188

secondary, less intense echo. The distance of this echo exactly matched the width of the wire
(270 m).
Figure 7-17b shows the graph of the calculated distance versus the actual distance the
wire was moved. The red line shows a linear fit to the data points with an R
2
of more than 0.999.
Since the steps size of 10 m can be detected by the sphere, the axial resolution should be good
to at least 10 m. A more rigorous test would be to detect separate objects that are 10 m apart
to verify this resolution.

Figure 7-17 a) Imagesc representation of A-line scans of a wire. The x-axis is the distance the wire
was moved away from the sphere. b) Agreement of experimentally found location versus the set
location of the wire. The red line is the linear fit to the data with an R
2
> 0.999.
7.15 Conclusion
In conclusion, ultrasound detection based on the photoelastic effect was demonstrated
using ultra-high-Q silica microcavities. A COMSOL Multiphysics model of ultrasound
propagation and interaction with the optical field of the cavity is developed and experimentally
verified. When compared to previous work, the Q factors are nearly three orders of magnitude
higher [15]. However, the bulk modulus of silica is also significantly higher, resulting in
189

previously unobserved and unanticipated interference effects, which could be reduced or
eliminated if a transducer with a shorter pulse output is used. These results demonstrate the
delicate balance between the improved sensitivity enabled by ultra-high Q factors and the
deleterious effects of silica. These findings significantly expand the field of optical microcavity-
based ultrasound detection, both in terms of material selection for sensor design and approaches
for performing detection, and they will enable advances in the field of ultrasound imaging for
medical applications [16-18].
References
1. Hoelen, C.G.A., et al., Three-dimensional photoacoustic imaging of blood vessels in
tissue. Optics Letters, 1998. 23(8): p. 648-650.
2. Maslov, K., et al., Optical-resolution photoacoustic microscopy for in vivo imaging of
single capillaries. Optics Letters, 2008. 33(9): p. 929-931.
3. Manohar, S., et al., Initial results of in vivo non-invasive cancer imaging in the human
breast using near-infrared photoacoustics. Optics Express, 2007. 15(19): p. 12277-
12285.
4. Matsko, A.B. and V.S. Ilchenko, Optical Resonators with Whispering-Gallery Modes-
Part I: Basics. IEEE Journal of Selected Topics in Quantum Electronics, 2006. 12(1): p.
3-14.
5. Chao, C.Y., et al., High-frequency ultrasound sensors using polymer microring
resonators. Ieee Transactions on Ultrasonics Ferroelectrics and Frequency Control, 2007.
54(5): p. 957-965.
6. Huang, S.W., et al., Low-noise wideband ultrasound detection using polymer microring
resonators. Applied Physics Letters, 2008. 92(19).
7. Li, H., et al., A transparent broadband ultrasonic detector based on an optical micro-ring
resonator for photoacoustic microscopy. Scientific Reports, 2014. 4.
8. Chistiakova, M.V. and A.M. Armani, Photoelastic ultrasound detection using ultra-high-Q
silica optical resonators. Optics Express, 2014.
9. Saleh, B.E.A. and M.C. Teich, Fundamentals of photonics. Wiley series in pure and
applied optics. 1991, New York: Wiley. xviii, 966 p.
10. Yariv, A. and P. Yeh, Optical waves in crystals : propagation and control of laser
radiation. Wiley series in pure and applied optics,. 1984, New York: Wiley. xi, 589 p.
11. Fielding, A.J. and C.C. Davis, Tapered single-mode optical fiber evanescent coupling.
Ieee Photonics Technology Letters, 2002. 14(1): p. 53-55.
12. Cai, M., O. Painter, and K.J. Vahala, Observation of critical coupling in a fiber taper to a
silica-microsphere whispering-gallery mode system. Physical Review Letters, 2000.
85(1): p. 74-77.
13. Chistiakova, M.V. and A.M. Armani, Cascaded Raman microlaser in air and buffer.
Optics Letters, 2012. 37(19): p. 4068-4070.
190

14. Koues, O.I., et al., Regulation of acetylation at the major histocompatibility complex
class II proximal promoter by the 19S proteasomal ATPase Sug1. Molecular and Cellular
Biology, 2008. 28(19): p. 5837-5850.
15. Ling, T., S.L. Chen, and L.J. Guo, High-sensitivity and wide-directivity ultrasound
detection using high Q polymer microring resonators. Applied Physics Letters, 2011.
98(20).
16. Heinritz, H., et al., Imaging of Superficial Skin Tumors in the Head and Neck - a
Comparison of High-Frequency Ultrasound with Computed-Tomography and Magnetic-
Resonance-Imaging. Hno, 1995. 43(1): p. 6-11.
17. Liu, W., et al., In vivo corneal neovascularization imaging by optical-resolution
photoacoustic microscopy. p. 81-86.
18. Plag, C., et al., High frequency ultrasound imaging of whole blood gelation and
retraction during in vitro coagulation. Journal of the Acoustical Society of America,
2012. 131(5): p. 4196-4202.

191

Chapter 8 Future Work
8.1 Overview
Though the projects in Chapters 3 to 7 have produced some interesting results and
conclusions, some improvement could be implemented for most of them. Further developments
on each project are presented here. Also, possible side projects and future directions are
discussed in this chapter.
8.2 Effects of Flow on Whispering Gallery Mode Microresonator Sensors
It was shown that it is difficult to isolate flow effects from detection effects on a
microresonator. To conduct this experiment in a way such that the effects are isolated, it is
necessary to maintain constant parameters. For one, the size of the device must be precisely
controlled. This could be accomplished by using the same fiber for spheres and by controlling
the etching on the toroids. The power input can be regulated with an attenuator to maintain
constant power. While the quality factor is not under complete control, it is important to test
devices with a high Q of about the same value. For linear fits to the initial slope of the data, the
analysis must be normalized throughout the experiments. Either the steepest part of each curve
should be fit, or the entire region before the shift begins to level out. Another possibility is to fit
a sigmoidal curve to the entire shift curve and use this fit to approximate the slope in the linear
region.
As was discussed, another large problem in the experiment is the isolation of protein flow
from buffer flow. One possible design for an injection experiment is to build a concentric needle
in which the outer chamber would contain buffer and the inner chamber would contain the
protein solution. In such a configuration, leaching effects should be minimized because the
buffer flow would not flow along the other solution and thus would not pull it along. The
192

experiment would be similar to the perturbation flow design in which the protein flow would be
started only once the system had reached steady state under buffer flow.
Another solution is to use a valve in the system so that leaching effects are once again
avoided. If the protein flow is completely isolated by the use of a valve, it would be easier to
conclude that no protein is flowing when it isn't supposed to.
8.3 Raman Microlaser
It was shown that bare silica microspheres are capable of producing cascading Raman
peaks in air and in buffer. Also, that the Raman peak could be tracked in air to detect temperature
changes. In the future, lasing peak tracking can be improved in a number of ways. First, lasing
peaks could be heterodyned using existing equipment in lab, as has been shown in a different
application already [1]. Second, a higher resolution OSA or spectrograph may aid in data
acquisition. Lastly, the splitter can be optimized to hold the wavelength of lasing to minimize
loss.
8.4 DNA Hybridization
In this chapter, the rate of hybridization was recorded in real time, showing two regimes
in the process. In future experiments, this method could be used to determine how the rate of
hybridization changes with differing numbers of matched base-pairs. In this way, the method can
be used as a tool for DNA comparison. Additionally, other parameters that affect the rate of
hybridization can be explored, such as temperature, DNA concentration, and cation
concentration [2].
8.5 Carbon Nanotube Gas Detection
It was shown that microspheres covered with carbon nanotube (CNT) clusters could
differentiate between carbon monoxide and carbon dioxide desorption. It is important to develop
193

a full theoretical paradigm to understand the detection limitations better. The system discussed in
this chapter could replace or aid with current TPD systems, greatly simplifying the process and
necessary equipment. The same device could also be used for adsorption experiments too. The
device needs to be isolated from the environment in a chamber that can be heated so that the
carbon nanotubes can be regenerated. Once the heating is complete, any gas can be injected into
the chamber and the response of the CNTs can be easily monitored as a resulting resonance shift
from the microcavity. In this way, gas sensors can be developed and further experiments can be
performed to figure out the sensitivity and selectivity of the device. Additionally, carbon
nanotubes can be grown from the silica surface via catalysts or other functionalization methods
[3]. This would allow for more control of the concentration of CNTs on the surface as well as
their properties.
8.6 Ultrasound Detection and Imaging
In the chapter about ultrasound, the theory of ultrasound detection on microspheres was
presented. A simple way to increase resolution in imaging experiments is to use a smaller
transducer [4]. Another approach that needs to be analyzed is the production of ultrasound pulses
using the photoacoustic effect. When electromagnetic energy hits a material, it could generate an
acoustic wave. This effect has been observed by shining laser light onto a material. The material
heats up by a small amount and causes a thermal expansion by the thermo-elastic mechanism and
acoustic waves are produced. By producing acoustic waves, the need for a transducer is
eliminated and makes the system more versatile and compact [5].
For future experiments, a new testing setup was made to reduce echoes and give more
freedom of movement for the imaged object. The diagram of the setup is shown in Figure 8-1. A
big chamber filled with water will be used to image objects. The chamber is too small to hold the
194

full-sized taper holder, so a smaller one was machined. The taper will still be pulled using the
regular taper holder, but then will be glued to the smaller aluminum one using UV curing epoxy.
After placing the taper in the chamber, the object to be imaged can be positioned underneath it.
The sphere will be attached to the transducer and will be introduced from the top. Using the side-
view camera, the sphere and the taper can be lined up before filling the chamber with water.

Figure 8-1 Side view diagram of the new testing liquid chamber for imaging objects with
ultrasound.
References
1. Maker, A.J. and A.M. Armani, Heterodyned toroidal microlaser sensor. Applied Physics
Letters, 2013. 103(12).
2. Liu, J., et al., Nanomaterial-Assisted Signal Enhancement of Hybridization for DNA
Biosensors: A Review. Sensors, 2009. 9(9): p. 7343-7364.
3. Tan, L.-L., et al., Growth of carbon nanotubes over non-metallic based catalysts: A
review on the recent developments. Catalysis Today, 2013. 217(0): p. 1-12.
4. Xiang, L., et al., Micromachined PIN-PMN-PT crystal composite transducer for high-
frequency intravascular ultrasound (IVUS) imaging. Ultrasonics, Ferroelectrics, and
Frequency Control, IEEE Transactions on, 2014. 61(7): p. 1171-1178.
5. Dong, B., et al., Photoacoustic probe using a microring resonator ultrasonic sensor for
endoscopic applications. Optics Letters, 2014. 39(15): p. 4372-4375.

195

Appendix Other Functionalization and Detection Techniques
A. 1 Lipid Bilayer Project
A.1.1 Objective
The goal of attaching lipid bilayers to microspheres, is to study how a molecule, such as a
protein, travels through the bilayer of a cell. Understanding of cell membrane transport will aid
in drug delivery and design. Current theoretical models try to predict transport through the
bilayer, but they are not complete and are unable to accurately predict the path of a molecule
through the bilayer or the motion within the bilayer. Microspheres can detect the molecule as it is
traveling and could help to exactly trace out the path it takes. This sensitivity is due to the
evanescent field that decays exponentially outside the sphere.
The bilayer is made when small unilamellar vesicles deposit on the surface of a silica
sphere and fuse together to make a single, even layer of lipids. Molecular movement is then
tracked by measuring the resonance shift when the sphere is coupled to a tapered fiber. Due to
the evanescent tail that exists outside of the microresonator, the presence of a molecule can be
detected as it gets close to the surface.
A.1.1 Approach
The first step in achieving the stated goal was to successfully attach a lipid bilayer to the
surface of the silica sphere. A 2:2:1 mixture of two lipids, DOPC and DPPC, and cholesterol is
made in chloroform. The lipids form unilamellar vesicles in the solution. The chloroform is
evaporated out as much as possible using a needle-thin stream of nitrogen or argon. The resulting
dry layer is then put in a degassing chamber under vacuum for an additional hour to remove any
remaining liquid.
196

The lipids are then dissolved in a salt buffer solution containing 100 mM sodium
phosphate (Na
3
PO
4
) and 150 mM NaCl with a pH of 7.09 by placing the tube on a vortexer. The
tube is then suspended in a sonicator for 45 minutes to break up the large vesicles into smaller
ones. Large vesicles produce a blotchy and uneven lipid layer on the surface of the sphere,
indicating the formation of a multi-layer structure. Therefore, the large vesicles should be broken
down in size.
After sonication, the spheres are placed into the solution for 36 hours. As the lipids come
in contact with the surface of the sphere, the vesicles rupture and cover the surface, forming a
bilayer.
All lipids are stored in glass bottles to prevent interactions the lipids (and chloroform)
may have with plastic containers. To image the bilayers, a low concentration of a Texas-Red-
labeled lipid is introduced into the initial solution to color the bilayer under a fluorescence
microscope. If the dye is used, the vials are covered in foil to prevent photobleaching. The dye
makes it possible to study the attachment of lipid bilayers and analyze its uniformity.
The microspheres need to be kept clean and have very little surface roughness. The
microspheres are made immediately before they are placed in the solution coming out of the
sonicator. The spheres are suspended from the sides of the glass vessel. Since the polymer-
covered section of the fiber is thicker than the sphere, the sphere does not touch the side of the
vial.
After the allotted aging time, a fluorescence microscope is used to view the coverage on
the spheres. Texas Red absorbs at 589 nm and fluoresces at 615 nm, so it appears red or orange
in the microscope software. Below are several images of the results (Figure A-1-1). Figure
A-1-1a shows fluorescence on a silica microsphere successfully covered in a lipid bilayer. The
197

coating looks uniform and even. However, if the focus of the microscope is scanned through the
entire sphere, some abnormalities can be discovered. Figure A-1-1b shows a different focal plane
of the same sphere, and the coverage still looks good. Looking at yet another plane in Figure
A-1-1c, a few spots are detected on the surface, indicating contamination or other problems with
the coverage.

Figure A-1-1 Fluorescent microscope images of a lipid bilayer coated silica microsphere with
different focal planes shown. a) The covering seems smooth and even. b) Focusing on a different
part of the sphere, the coating still looks smooth. c) At a third focal plane, some spots can be seen in
the coating.
The next step was to ensure that the quality factor of the microresonator remained high
after the bilayer attachment. The coated spheres had Qs on the order of 10
6
, indicating that the
sphere is still sensitive to the environment around it and will be able to detect a molecule’s path
through the bilayer.
In previous work, Lindsay Freeman was able to verify that the lipids attached as a single
layer by using a photobleaching technique known as FRAP [1]. She and her co-authors were also
able to show excitation of Cy5 fluorophores in the bilayer using a resonant cavity using a
198

modification of FRET [2]. Currently, a molecule is being chosen to traverse the bilayer to gather
data on its path.
A. 2 Biotin Functionalization and Streptavidin Detection

A.2.1 Objective
Specific detection is possible when the surface of the sensing device is functionalized
with a marker. For example, biotin can be attached to the surface to specifically detect
streptavidin in solution, which tightly binds to this marker [3, 4].
A.2.2 Functionalization Method
Functionalization begins with a clean fabricated sphere (made by melting the tip of an
optical fiber with a CO
2
laser). The surface is hydroxylated by the process of oxygen plasma
etching. The samples are placed in a chamber under vacuum. Oxygen gas is then introduced
under low pressure and a high frequency electric field excites the gas into plasma through
dielectric breakdown. It is presumed that the silicon surface becomes covered with a monolayer
of bridge-bonded oxygen atoms. These bonds are highly strained, so the introduction of water
(even from the air), breaks the bonds to form hydroxyl groups.
The next step is vapor deposition of 3-aminopropyl-trimethoxysilane (APTMS) onto the
surface of the device. Since the APTMS is stored in the fridge, it is first brought to room
temperature. The samples are then placed in a dessicator with the open bottle of APTMS. The
chamber is placed under vacuum and the reaction is allowed to take place for 15 minutes, an
ideal time to form a single monolayer of amine groups on the surface via the following chemical
reaction:
199

The spheres are then ready to be functionalized with a specific marker. Since many
markers can be attached to N-hydroxysuccinimide (NHS), there are endless possibilities for
functionalization and detection. NHS-biotin is used in the present series of experiments since
streptavidin is the molecule to be detected. The biotin-avidin interaction is extremely strong,
even though it is non-covalent. Streptavidin has four biotin binding pockets. The biotin solution
is made in an organic solvent such as dimethyl sulfoxide (DMSO) in a 10 mM concentration.
The devices are submerged in the NHS-biotin/DMSO solution and placed on a tilt tray for 30
minutes. The following reaction takes place:

200

The samples are then rinsed by placing them in deionized water and back on the tilt tray for five
additional minutes. They are then allowed to air dry for one hour before testing. The devices can
also be recycled by repeating the oxygen plasma step to clear the surface [5].
A.2.3 Imaging the Functionalized Surface
To test the extent of biotinylation, a Texas Red-Avidin protein conjugate is used to
visualize the surface. The biotin functionalized spheres are incubated in a 10 g/mL Texas Red-
Avidin in Phosphate Buffered Saline (PBS) solution for 30 minutes on a tilt tray. The spheres are
then rinsed for five minutes before imaging on the fluorescent microscope.
201

Figure A-2-1 Fluorescent imaging of microspheres functionalized with biotin. a) A bright field
image of a sphere that was biotinylated but not exposed to the Texas-Red-Avidin solution. b)
Fluorescence image of the control sphere showing no fluorescence. c) Bright field image of a
biotinylated sphere that had been incubated with Texas-Red-Avidin. The surface is still smooth
after all the functionalization steps. d) Fluorescence imaging shows a uniform coverage of the
sphere and step indicating successful functionalization.
A.2.4 Detecting Streptavidin
The functionalized devices are very good at detecting streptavidin. Since biotin and
streptavidin have one of the strongest biological bonds, it is very difficult for the two to come
apart. Therefore, once the molecules bind, it can be assumed most of them will stay bound on the
202

surface of the sphere. Figure A-2-2 shows the result of one detection experiment. Streptavidin
was injected on a functionalized sphere at 100 mL/min. Once the peak shift reached equilibrium,
the flow was stopped. The resonance peak began to shift back to the original position; however,
it is easy to see that it did not shift all the way back. This is because molecules that were not
covalently bound to the biotin on the surface come off the surface and float away. Also, the
stoppage of flow also produces the blue-shift effect. The change of the resonance wavelength
from the initial value to the final value indicates that streptavidin bound to the surface.

Figure A-2-2 Resonance shift of a microsphere as streptavidin is injected. After injection is stopped,
the resonance does not return to the original position due to the strong binding of streptavidin to
biotin.
A. 3 Biodetection: Parameters Affecting the Resonance Shift
A.3.1 Objective
Microresonators are sensitive to the environment around them. Any time a molecule
binds to the surface, the phenomenon can be detected through a resonance shift. Classifying how
this shift is affected by the experimental setup is an important step to be able to factor out these
effects.
203

The cause of a changing refractive index when molecules bind to the surface of the
microsphere needs to be determined. Two factors could contribute to the change, individually or
in combination. Temperature effects may occur when the whispering gallery mode heats the
particles on the surface of the device, which in turn heats the silica material, causing a red shift in
the wavelength. The other possibility is that the optical field interacts with the molecules, which
have a different polarizability than the device, resulting in a similar shift. Theories and
simulations have been presented that support both causes.
The changes in resonance shift were studied by varying a series of parameters:
concentration, input power, quality factor, wavelength, geometry, and laser scan rate and speed.
This is a complete set of variables to understand how the shift is affected.
A.3.2 Sensing Experimental Procedure
To begin, the power is recorded at the input to the fiber spool and at the output of the
taper. The solution concentration, the testing wavelength, the flow rate, and the device size are
also pre-recorded for each experiment. The resonance is positioned on the left side of the screen,
so that its movement forward can be recorded without losing the peak. The scan parameters from
the function generator (hertz and volts-peak-to-peak) are also recorded.
The Q is measured in the under-coupled and critically-coupled regimes. During critical
coupling, the percent of power coupled into the cavity is between 30 and 70%. The resonance
shift is recorded at critical coupling. The software starts recording data before the protein
solution is injected to detect the base position of the peak. The solution of interest is then injected
for 60 seconds and stopped. The peak is continually tracked as it shifts back to the original
position. The quality factor is again recorded in both coupling regimes. The chamber is then
flushed with buffer, and the experiment can be repeated with changed parameters.
204

After experiments are over, the percent of power coupled into the device (or the
transmission) is determined as C
%
= 1 - y
R
/y
o
where y
R
is the transmission value at the minimum
of the resonance and y
o
is the baseline transmission. The power coupled into the cavity is found
by multiplying the input power by the percent of coupling. The total power in the cavity is
determined by the following equation:

(A.1)
where P
in
is the power coupled into the cavity. The power interacting with the environment is
found by multiplying the percent of the power in the environment by the power of the cavity.
The percent of time spent on resonance is calculated by:

(A.2)
where w is the FWHM of the resonance, VPP is the voltage setting of the function generator (in
volts-peak-to-peak), F is the frequency set on the function generator (in hertz), and x
0
and x
F
are
the start and end values on the x-axis of the digitizer. Finally, the approximate time spent on
resonance is estimated as the FWHM of the resonance.
A.3.3 Results
The scan range is changed by setting the voltage peak-to-peak on the function generator.
As this voltage is decreased, the device spends more time on resonance because the rate of the
scan is the same. The results of changing the scan range for three wavelengths and three
concentrations are shown in Figure A-3-1. The concentrations shown are 10 aM (Figure A-3-1a),
1 nM (Figure A-3-1b), and 1 M (Figure A-3-1c), although more concentrations were studied in
the experiments. The three wavelengths chosen were 633 nm, 765 nm, and 1300 nm. While at
the lower wavelengths, the absorption of water is low, at 1300 nm it increases considerably,
205

lowering the Q and increasing the noise of detection. As can be seen in Figure A-3-1, the
wavelength shift varies much more for the high wavelength than the other two. This could be due
to the increased noise from tracking a wider resonance peak.

Figure A-3-1 Results of wavelength shift with varying scan range. All legends show the operating
wavelengths in nm. a) Results for a 10 aM concentration of protein. b) Results for a 1 nM
concentration. c) Results with a protein concentration of 1 M.
There appears to be no trend in the data to suggest that the scan range has an effect on the
wavelength shift. For all three concentrations, the shift appears fairly constant for 633 and 765
nm and even though the shift varies for 1300 nm, there doesn't seem to be a dependence on the
scan range.
There is a change in shift with increasing concentration of protein. For all scan range
values, the shift in resonance at 633 and 765 nm is greater for higher concentrations. For 1300
nm, it is more difficult to determine whether there is a difference between the 1 nM and 1 M
concentrations, but the shifts definitely increase compared to the 10 aM concentration.
To better understand the effects of variables on the wavelength shift, experimental
parameters were kept as consistent as possible. For example, in the data shown in Figure A-3-2,
the following parameters were the same for each point: the protein concentration was kept
constant at 1 pM, the flow rate was 200 L/min, and the volume injected was 1 mL. Since the
quality factor, the power, and the coupling cannot be made exactly the same for each experiment,
206

the data was plotted as a function of the combination of these values. The x-axis in Figure A-3-2
is shown as P
in
*Q
L
, where P
in
is the power input to the cavity multiplied by the percent coupling
and Q
L
is the loaded quality factor. Even plotting the data in this way, it is possible to see that a
trend is too difficult to extract from the data to determine just how each variable affects the
overall shift. To better perform experiments of this sort, the variables must be separated further,
with more care taken to monitor potential confounds, such as device size. Additionally, it is
necessary to maintain more consistency and vary only one parameter at a time. This work has
inspired a more systematic study investigating the dependence of the shift on circulating
intensity, which includes not just Q and P
in
, but also device-specific variables, like radius.

Figure A-3-2 Results of wavelength shift as a function of power in the resonator (P
in
) multiplied by
the loaded Q (Q
L
). The diameter of the microspheres is also indicated in the legend. No apparent
trend appears.
References
1. Freeman, L.M. and A.M. Armani, Photobleaching of Cy5 Conjugated Lipid Bilayers
Determined With Optical Microresonators. Ieee Journal of Selected Topics in Quantum
Electronics, 2012. 18(3): p. 1160-1165.
207

2. Freeman, L.M., et al., Evanescent field excitation of Cy5-conjugated lipid bilayers using
optical microcavities. Nanobiosystems: Processing, Characterization, and Applications
Iv, 2011. 8103.
3. Hunt, H.K. and A.M. Armani, Bioconjugation of ultra-high-Q optical microcavities for
label-free sensing. Frontiers in Biological Detection: From Nanosensors to Systems Iii,
2011. 7888.
4. Soteropulos, C.E., H.K. Hunt, and A.M. Armani, Determination of binding kinetics using
whispering gallery mode microcavities. Applied Physics Letters, 2011. 99(10).
5. Hunt, H.K. and A.M. Armani, Recycling microcavity optical biosensors. Optics Letters,
2011. 36(7): p. 1092-1094.

Abstract (if available)

Abstract The ability to detect specific molecules in solution is critical for a variety of applications, including analytical and medical diagnostic measurements. Optical microcavities have become a popular platform for the detection of minute quantities of material adsorbing to their surfaces. These devices trap light by guiding it in circular resonant paths near the periphery of the cavity. The wavelengths required to excite these modes are affected by the surrounding environment, and can be used to measure changes at the surface of the resonator including the adsorption of material. Since optical microcavity biosensors were first demonstrated, a great number of questions have come to light regarding how these devices are used, how their data is interpreted, and how they may be improved. Here I describe my preliminary work addressing some of these questions. ❧ Specifically, I have explored mass transport considerations suggesting that the sensor geometry can play a significant role in the transient response of the sensor. Through this work I have helped identify flow-induced artifacts unique to this type of sensor and presented methods to correct for them. Additionally, I have demonstrated how undoped optical microcavities can be used as narrow-linewidth lasers. This work will enable an important paradigm shift in optical microcavity biosensing, improving accuracy and reliability of the measurement by monitoring changes to the lasing wavelength that occur upon adsorption rather than changes in the resonant wavelength. ❧ I describe my progress and future plans developing novel applications for optical microcavities in sensing and imaging. I have shown gas differentiation capabilities of silica microspheres coated with carbon nanotube clusters. The application could greatly increase safety of gas exposure and enable surface science studies. I have also contributed to a project focused on measuring the hybridization kinetics of DNA molecules, which could help with disease prevention or forensics. The microcavity can detect the vibrations resulting from ultrasound pulses, which change depending on what material the US travels through. I discuss how the image can be reconstructed from the oscillatory transmission changes detected by the microsphere. Finally, I discuss future work to be performed on these projects and possible improvements in experimental design.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Optical simulation and development of novel whispering gallery mode microresonators

PDF

Fabrication and characterization of yoroidal resonators for optical process improvement

PDF

Developing improved silica materials and devices for integrated optics applications

PDF

Surface-enhanced on-chip whispering gallery mode resonators for improved nonlinear optics

PDF

Integration of practical high sensitivity whispering gallery mode resonator sensors

PDF

Applications of optical microring and microdisk resonators as physical, chemical, and biological sensors

PDF

Development of integrated waveguide biosensors and portable optical biomaterial analysis systems

PDF

Development of high-frequency and high-field optically detected magnetic resonance

PDF

Materials development and characterization of optically-active core-shell nanomaterials

PDF

Resonant light-matter interactions in nanophotonic structures: for manipulating optical forces and thermal emission

PDF

Silica sol-gel thin film coatings for integrated photonic devices

PDF

Nonlinear optical nanomaterials in integrated photonic devices

PDF

Development of hybrid microsensors for environmental monitoring and biodetection

PDF

Organic materials for linear and nonlinear optical microdevices

PDF

One-dimensional nanomaterials for electronic and sensing applications

PDF

Hydrogen storage in carbon and silicon carbide nanotubes

PDF

Formation of polymer gels, films, and particles via initiated chemical vapor deposition onto liquid substrates

PDF

Microresonator-based Kerr frequency comb for high-speed optical communications and signal processing

PDF

Hybrid lipid-based nanostructures

PDF

Printed and flexible carbon nanotube macroelectronics

Asset Metadata

Creator Chistiakova, Maria Vladimirovna (author)

Core Title Biological and chemical detection using optical resonant cavities

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Chemical Engineering

Publication Date 11/03/2014

Defense Date 10/20/2014

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag biosensing,carbon dioxide,carbon monoxide,carbon nanotubes,COMSOL simulations,detection,imaging,microresonators,microspheres,microtoroids,OAI-PMH Harvest,sensing,ultrasound,whispering gallery mode resonators

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Armani, Andrea M. (committee chair), Malmstadt, Noah (committee member), Shing, Katherine (committee member), Thompson, Barry C. (committee member)

Creator Email c_masha@yahoo.com,chistiak@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-512920

Unique identifier UC11297882

Identifier etd-Chistiakov-3057.pdf (filename),usctheses-c3-512920 (legacy record id)

Legacy Identifier etd-Chistiakov-3057.pdf

Dmrecord 512920

Document Type Dissertation

Format application/pdf (imt)

Rights Chistiakova, Maria Vladimirovna

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