University of Southern California Dissertations and Theses

Development of integrated waveguide biosensors and portable optical biomaterial analysis systems

(USC Thesis Other)

Development of integrated waveguide biosensors and portable optical biomaterial analysis systems

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content DEVELOPMENT OF INTEGRATED WAVEGUIDE BIOSENSORS AND PORTABLE OPTICAL BIOMATERIAL ANALYSIS SYSTEMS by Mark Christopher Harrison 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 (Electrical Engineering - Electrophysics) December 2015 Copyright 2015 Mark Christopher Harrison ii Acknowledgements Completing a PhD is a large undertaking, but it is not something that is done alone. During my time at USC, I have been fortunate to receive the help and encouragement of a great number of people who have played a role, large or small, in the completion of my PhD, and I owe them a great deal of gratitude. First and foremost, I would like to thank my advisor, Prof. Andrea Armani, for all she has done for me. She has been a mentor in the best sense of the word, teaching me most of what I know about conducting research as well as offering guidance and support in other areas of my life. Andrea has always made ample time for all of her graduate students, and I feel fortunate to be counted among them. From asking her countless questions when I began as a PhD student to practicing presentations for conferences to ecstatically sharing results with her as I completed projects, she has always cheerfully participated in the research process with me. I have also appreciated her focus on both the theoretical and experimental sides of research in creating a complete research story. I know that because of the training and mentorship I received from Andrea, I have been well-prepared for success in my future endeavors. I have also encountered many great teachers and professors during my time at USC. I would like to thank my qualifying and dissertation committee members: Prof. William Steier, Prof. Alan Willner, Prof. Megan McCain, Dr. Joe Touch, and Prof. Wei Wu. Additionally, I would like to thank Prof. Aluizio Prata for all he taught me of electromagnetic theory as well as Prof. Steve Cronin for giving me my first taste of research as an undergraduate. I know that a great deal of the reason I have enjoyed my time in graduate school is due to the other fantastic members of the Armani Research Group. In no particular order, thank you to Dr. Tushar Rane, Hyungwoo Choi, Vinh Diep (especially for the homebrew sessions), Soheil Soltani (especially for all the helpful E&M conversations), Alexa Hudnut, Amanda Cordes, Kelvin Kuo (especially for all the encouragement and times spent working out together), Dr. Xiaoqin Shen, Andre, Sam McBirney, Victoria Sun (especially for mischief with dry ice), Michele Lee, Erick Moen, Akshay Panchavati, and Dr. Eda Gungor. Thank you to past members of the group, Dr. Ashley Maker, Dr. Maria Chistiakova, Dr. Simin iii Mehrabani, Dr. Xiaomin Zhang, Dr. Ce Shi, Dr. Hong Seok Choi, Matthew Reddick, Dr. Jason Gamba, Dr. Rasheeda Hawk (especially for teaching me so much about biology), Prof. Cecilia Lopez, and Prof. Heather Hunt. Thank you to the undergraduates, past and present, in the group: Danny, Grace, Lili, Gumi, Martin, Linda, Garrison, Leah, Emma, Tara, Sam K-L, Max, Nishita, Brian, Nic, Bradley and Chai. A special thanks to the undergraduates (and high school student!) I have had the privilege of working with directly: Lea Fang, Alexei Naumann, Audrey Harker, and Samantha Wathugala. There have been several visiting scholars in the group, and I have appreciated the chance to get to know and work with them, so thank you Tobi Weinhold, Abian Socorro, Imran Cheema, and Dr. Yong-Won Song. I am blessed with a loving and caring family who has been very supportive of me throughout my time as a PhD student. Thank you to my grandparents and extended family for their love, support and friendship throughout my life. Thank you to my parents for their love, support, encouragement, and everything else they have done to help me become the man I am today. Thank you to my brothers Joel, Paul and Andrew. There are too many things I could thank you for individually, so I will simply thank the three of you for your brotherhood and all that entails. Finally, I must thank my lovely and compassionate wife, Nicole. Without you, I do not know if I would have made it to the end of the PhD, and I certainly would not have enjoyed the process nearly as much as I have. I love you, Niki. iv Table of Contents Acknowledgements ....................................................................................................................................... ii Table of Contents ......................................................................................................................................... iv List of Figures ............................................................................................................................................ viii List of Tables ............................................................................................................................................. xiv Abstract ....................................................................................................................................................... xv Chapter 1: Overview ..................................................................................................................................... 1 1.1 Motivation ..................................................................................................................................... 1 1.2 Chapter overview .......................................................................................................................... 3 Chapter 1 References ................................................................................................................................ 5 Chapter 2: Background ................................................................................................................................. 7 2.1 Integrated Optical Devices ............................................................................................................ 7 2.2 Optical Waveguides ...................................................................................................................... 9 2.2.1 Waveguide Modes and Polarization.................................................................................... 10 2.2.2 Evanescent Field ................................................................................................................. 11 2.2.3 Optical Birefringence .......................................................................................................... 12 2.3 Waveguide Loss Mechanisms ..................................................................................................... 14 2.3.1 Absorption Loss .................................................................................................................. 14 2.3.2 Scattering Loss .................................................................................................................... 15 2.3.3 Propagation Loss ................................................................................................................. 15 2.3.4 Bending Loss ...................................................................................................................... 16 2.3.5 Coupling Loss ..................................................................................................................... 17 2.3.6 Insertion Loss ...................................................................................................................... 18 2.4 Suspended Silica-on-Silicon Waveguides and Devices .............................................................. 18 2.4.1 Waveguide Geometry and Fabrication................................................................................ 19 2.4.2 Waveguide Splitter Devices ................................................................................................ 21 2.4.3 Waveguide Splitter Fabrication .......................................................................................... 23 2.4.4 Advantages and Disadvantages of This Design .................................................................. 23 2.5 Waveguide and Waveguide Device Characterization ................................................................. 24 2.5.1 Loss measurements ............................................................................................................. 26 2.5.2 Waveguide Splitter Output .................................................................................................. 27 Chapter 2 References .............................................................................................................................. 28 Chapter 3: Using embedded waveguide sensors for spatiotemporal fluorescent measurements ................ 30 3.1 Introduction ................................................................................................................................. 30 3.1.1 Embedded waveguide sensors ............................................................................................ 30 3.1.2 Fluorescent back-coupling .................................................................................................. 31 v 3.2 Experimental Methods ................................................................................................................ 32 3.2.1 Computer modeling and simulations .................................................................................. 32 3.2.2 Device design and fabrication ............................................................................................. 34 3.2.3 Experimental setup and measurements ............................................................................... 35 3.3 Data and Results ......................................................................................................................... 37 3.3.1 Transmission loss measurements ........................................................................................ 37 3.3.2 Simulation and modeling .................................................................................................... 37 3.3.3 Spectrograph sensing measurements................................................................................... 43 3.3.4 Spatiotemporal fluorescence measurements ....................................................................... 46 3.4 Conclusion and Future Outlook .................................................................................................. 50 Chapter 3 References .............................................................................................................................. 50 Chapter 4: Characterization of silica suspended waveguides and performance improvements with inorganic coatings ....................................................................................................................................... 52 4.1 Introduction ................................................................................................................................. 52 4.2 Trapezoidal waveguides .............................................................................................................. 52 4.2.1 Fabricating bent devices ...................................................................................................... 53 4.2.2 Bending Losses in a Trapezoidal Device ............................................................................ 54 4.2.3 Simulation Models .............................................................................................................. 55 4.2.4 Experimental methods ......................................................................................................... 59 4.2.5 Data and Results.................................................................................................................. 59 4.2.6 Conclusion and Future Outlook .......................................................................................... 63 4.3 Sol-gel coated waveguides .......................................................................................................... 64 4.3.1 Strategies for enabling single-mode operation .................................................................... 65 4.3.2 Silica sol-gel thin films ....................................................................................................... 67 4.3.3 Simulations and refractive index constraints ...................................................................... 69 4.3.4 Data and results ................................................................................................................... 70 4.3.5 Conclusions and future work .............................................................................................. 72 Chapter 4 References .............................................................................................................................. 73 Chapter 5: Waveguide Splitter Biosensors ................................................................................................. 74 5.1 Introduction ................................................................................................................................. 74 5.1.1 Sensing Mechanism ............................................................................................................ 74 5.1.2 Microfluidics ....................................................................................................................... 76 5.1.3 Lab-on-a-chip ...................................................................................................................... 76 5.2 Computer modeling and simulations .......................................................................................... 77 5.2.1 Models and methods ........................................................................................................... 77 5.2.2 Data and results ................................................................................................................... 78 5.3 Surface functionalization ............................................................................................................ 81 vi 5.3.1 Experimental methods ......................................................................................................... 81 5.3.2 Results ................................................................................................................................. 83 5.4 Sensing experiments ................................................................................................................... 84 5.4.1 Experimental methods ......................................................................................................... 84 5.4.2 Data and results ................................................................................................................... 87 5.5 Microfluidics fabrication and integration ................................................................................... 91 5.5.1 Experimental methods ......................................................................................................... 91 5.5.2 Results ................................................................................................................................. 93 5.6 Conclusions and future work ...................................................................................................... 94 Chapter 5 References .............................................................................................................................. 94 Chapter 6: Portable, fiber-based polarimetric stress sensor ........................................................................ 96 6.1 Introduction ................................................................................................................................. 96 6.2 Theoretical Analysis ................................................................................................................... 97 6.2.1 Sensor Operation ................................................................................................................. 97 6.2.2 Transfer Matrices ................................................................................................................ 99 6.3 Experimental Methods .............................................................................................................. 100 6.3.1 Experimental Setup ........................................................................................................... 100 6.3.2 Calibration ......................................................................................................................... 101 6.3.3 PDMS preparation............................................................................................................. 111 6.3.4 Sensor operation and experiments performed ................................................................... 111 6.4 Data and Results ....................................................................................................................... 113 6.4.1 PDMS compression experiments ...................................................................................... 113 6.4.2 Noise measurements and sensitivity ................................................................................. 114 6.5 Conclusions ............................................................................................................................... 118 Chapter 6 References ............................................................................................................................ 119 Chapter 7: Future Work ............................................................................................................................ 121 7.1 Introduction ............................................................................................................................... 121 7.2 Polymer waveguides ................................................................................................................. 121 7.3 Waveguide splitter biosensor .................................................................................................... 122 7.4 Polarimetric fiber stress sensor ................................................................................................. 122 Chapter 7 References ............................................................................................................................ 124 Appendix A: Graphene coated waveguide splitters .................................................................................. 125 A.1 Introduction ............................................................................................................................... 125 A.2 Theory ....................................................................................................................................... 126 A.3 Fabrication methods .................................................................................................................. 127 A.3.1 Electrodes .......................................................................................................................... 127 vii A.3.2 Graphene growth and deposition ...................................................................................... 127 A.4 Experiments .............................................................................................................................. 128 A.5 Future Work .............................................................................................................................. 129 Appendix A References ........................................................................................................................ 129 Appendix B: Mentoring Projects .............................................................................................................. 131 B.1 Polymer waveguide coatings .................................................................................................... 131 B.1.1 Liquid crystal polymers .................................................................................................... 131 B.1.2 Liquid Crystal Coatings on Suspended Silica Waveguides .............................................. 132 B.1.3 Device Coating Experiments ............................................................................................ 133 B.1.4 Conclusions and Future Work ........................................................................................... 135 B.2 Sol-gel recipes ........................................................................................................................... 135 B.2.1 Experimental methods ....................................................................................................... 135 B.2.2 Data and results ................................................................................................................. 136 B.2.3 Conclusions ....................................................................................................................... 138 Appendix B References ........................................................................................................................ 139 Bibliography ............................................................................................................................................. 140 viii List of Figures Figure 2-1: Examples of typical integrated photonic devices. Pictured are computer renderings of a) a waveguide, b) a coupler or splitter, and c) a resonator or laser. ............................................................. 8 Figure 2-2: Total internal reflection in a waveguide. Light at a shallow enough angle will be reflected and continue to propagate in the waveguide core. ........................................................................................ 9 Figure 2-3: Hybrid modes of a step-index optical fiber. At the left is the fundamental mode, and the mode-orders increase from left to right. Step-index optical fibers can also have higher-order modes with nodes in the radial direction. ........................................................................................................ 10 Figure 2-4: Cartoon indicating the locations of the confined and evanescent fields in a fundamental waveguide mode. The hashed areas indicate the evanescent field of the optical field distribution. ... 12 Figure 2-5: Cartoon diagram depicting un-polarized light passing through a birefringent block of material. Since the parallel and perpendicular polarizations see two different refractive indices (n 1 and n 2 ), they are refracted at different angles (θ 1 and θ 2 ) through the block of material. .......................................... 13 Figure 2-6: Bending loss in a waveguide. Going around a bend, the guided light will not satisfy the critical angle condition, and some light will be transmitted out of the waveguide core and lost. ........ 17 Figure 2-7: Cross section of waveguide geometry. a) A schematic cross section. The blue section represents SiO 2 and the green section represents Si. b) SEM image of the end-facet of a waveguide. .............................................................................................................................................................. 20 Figure 2-8: Schematic diagrams showing several steps in the waveguide fabrication process. These diagrams show the waveguides after a) the first photolithography step, b) the second photolithography step, c) the XeF 2 etching step, and d) the CO 2 laser reflow step. ......................................................... 21 Figure 2-9: Scanning electron micrograph (SEM) of a waveguide splitter device. The waveguide, tapered and splitting region can all be seen, and the bump which causes splitting to occur is indicated with an arrow on the bottom waveguide channel. ............................................................................................. 22 Figure 2-10: Schematic diagram of the waveguide testing setup................................................................ 25 Figure 2-11: Diagram showing how propagation loss is calculated from measurements. The transmission loss measured from several waveguide lengths is plotted and fit to a straight line. The slope of the line is equal to propagation loss and the y-intercept is equal to the coupling loss in the system. ........ 26 Figure 2-12: A three-dimensional plot of arbitrary intensity vs. position for a waveguide splitter. The intensity is measured by the beam profiler and two distinct peaks of roughly equal power can be clearly seen. .......................................................................................................................................... 27 Figure 3-1: Schematic diagram indicating emission paths for fluorescent radiation which can be experimentally monitored. This diagram excludes laser light used to excite the fluorophores. .......... 32 Figure 3-2: Schematic diagram of device geometry used for simulations. D is distance between dipoles, ϕ 1 and ϕ 2 are angles of alternating dipoles, t is the thickness of the cladding in the sensing well, L is the length of the simulation, and n 1 , n 2 , and n 3 are the refractive indices of the cladding, waveguide core, and sensing well, respectively. The locations of power monitors in the simulation are indicated by red lines, and the boundaries of the simulation region are indicated by a dashed line. ................... 33 Figure 3-3: Schematic of testing set-up. Light is coupled into the device from the left via a lensed fiber, and output is collected with a filter and spectrograph a) from the output of the waveguide and b) from above the sensing well of the device. ................................................................................................... 36 ix Figure 3-4: Results from 2-dipole simulation (Model 1). Graphs show the percentage of optical power, for different dipole spacing and angles, a) coupled into the waveguide, b) radiating above the waveguide, and c) lost into the substrate below the waveguide. .......................................................... 39 Figure 3-5: Results from the first ten-dipoles simulation (Model 2). Graphs show the percentage of optical power a) back-coupled into the waveguide, b) radiated above the waveguide and c) lost into the substrate below the waveguide for three different dipole orientations, indicated on the graphs using arrows. ........................................................................................................................................ 41 Figure 3-6: Results from the second 10-dipole simulation (Model 3). Distribution of radiated dipole power for increasing refractive indices of the sensing well environment. ........................................... 42 Figure 3-7: Spectrograph spectra taken from the output of the waveguide (through), and above the waveguide (above), with fluorescent dye on the device. ...................................................................... 44 Figure 3-8: Emission peak of LD-700 laser dye dissolved in toluene. ....................................................... 45 Figure 3-9: Spatiotemporal fluorescence measurements of dry laser film on the waveguide device. Graph a) shows normalized intensity vs. time and the graphs are in order of position further down the waveguide with the rear being the far left of the waveguide. Graph b) shows normalized intensity vs. position and the graphs are in order of increasing time with time increasing from the rear to the front. These two graphs show the same data, but rotated 90 degrees with respect to each other. ................. 48 Figure 4-1: Fabrication procedure and scanning electron micrograph (SEM) image of the trapezoidal waveguide. (a) 80- m wide silica pads are created through a photolithography and buffered HF etching step. (b) After another photolithography and buffered HF etching step, two 8- m trapezoids on each side of the 80- m trapezoids are created. (c) XeF 2 etching is used to isotropically undercut the silica. (d) SEM side view image of a single waveguide. ................................................................ 53 Figure 4-2: Scanning electron micrograph (SEM) images of the curved trapezoidal waveguide. (a) An overview of the S-curve waveguide geometry depicting the inner bending radius (R). (b) SEM of the bending section and (c) cleaved end of a curved waveguide device. ................................................... 54 Figure 4-3: Finite element method simulation of the electric field distribution (TE mode) of (a) trapezoidal and (b) rectangular cross-sections at 1550nm. (c) Comparison of the intensity of the mode profile along the x (horizontal) direction, normalized to the highest intensity in the trapezoidal device. .............................................................................................................................................................. 55 Figure 4-4: Finite-difference time-domain simulation showing the optical field intensity inside a bent trapezoidal waveguide with an inner radius of 50 m at 1550nm. It is clear that the optical field leaks into the silica membrane in-between the two waveguide arms. ........................................................... 58 Figure 4-5: (a) Propagation loss of the trapezoidal waveguides at 658, 980 and 1550 nm. (b) Transmission at four different polarization states. ...................................................................................................... 60 Figure 4-6: Experimentally measured bending losses at 658, 980 and 1550 nm, along with loss calculated from the FDTD simulations.................................................................................................................. 62 Figure 4-7: Output response of the trapezoidal silica waveguides when input with power up to 200 mW. .............................................................................................................................................................. 63 Figure 4-8: Two routes for single-mode behavior. Left: Shrinking device dimensions will result in structural defects. Right: A cladding layer may be added by spin-coating another material on the waveguide device. ................................................................................................................................ 66 Figure 4-9: Hydrolysis and condensation reactions showing formation of a silica sol-gel matrix. ............ 68 x Figure 4-10: COMSOL simulation results of a) the fundamental mode in an air-clad waveguide and b) the fundamental mode in a single-mode, sol-gel cladding waveguide. ...................................................... 70 Figure 4-11: SEM images of a sol-gel coated waveguide taken from a) the input end and b) above. The sol-gel thickness is clearly not uniform and has several cracks, including around the waveguide channel, where it might affect the mode. Additionally, because the thermal silica and silica sol-gel have such similar conductivity values, they appear as the same color in these images. ...................... 71 Figure 4-12: Measured losses from sol-gel coated waveguides measured at three different wavelengths. The data was fit to linear curves from which the loss can be measured. Given the large spread in measured loss values, the linear fits are likely not very accurate. ........................................................ 72 Figure 5-1: Examples of several photonic sensor geometries: a) Mach-Zender interferometric waveguide sensor, b) resonant optical microcavity sensor, and c) photonic crystal waveguide sensor. ................ 74 Figure 5-2: 3D images of the waveguide splitter output as captured by the beam profiler. Each of the two peaks is the output from one of the splitter waveguide arms. The two images show the output of the waveguide splitter a) before and b) after an analyte is introduced to the surface of the device, causing the splitting ratio to change. ................................................................................................................. 75 Figure 5-3: Schematic of model used for simulation with important parameters marked on the drawing. The light and dark colors represent thin and thick parts of the membrane, respectively (T 1 = 900 nm and T 2 = 2μm). The black lines are the waveguide channels and the red line indicates the direction of insertion of the optical mode. ............................................................................................................... 77 Figure 5-4: Results from the simulations in which the background refractive index was modified. For increasing refractive index, ΔPower decreases, and the splitting ratio stays fairly constant across a wide wavelength range. ........................................................................................................................ 79 Figure 5-5: Results from the simulations in which an adlayer with increasing thickness was added to the sensor. For increasing thickness, the change in ΔPower varies at different wavelengths. Additionally, ΔPower does not stay constant across the entire wavelength range as seen when operating without an adlayer. ............................................................................................................... 80 Figure 5-6: Results from the simulations in which an adlayer with increasing refractive index was added to the sensor. For increasing index, the change in ΔPower varies at different wavelengths. Additionally, ΔPower does not stay constant across the entire wavelength range as seen when operating without an adlayer. ............................................................................................................... 81 Figure 5-7: Surface functionalization procedure. The steps are as follows: a) the device is oxygen plasma treated to hydroxylate the silica surface, b) a pressure vapor deposition step is performed to attach GPTMS to the OH groups, and c) the sample is incubated with an anti-CREB solution in a warm, humid environment to attach the antibodies to the surface of the device. ............................................ 82 Figure 5-8: Optical micrographs of the splitting region of a functionalized waveguide splitter a) under a bright field light and b) under fluorescent excitation and a filter. ........................................................ 84 Figure 5-9: Cross-sectional profile of the output of the waveguide splitter as measured by the beam profiler. Integrating the area under each peak is used to calculate the splitting ratio. ......................... 85 Figure 5-10: Sensing data showing the results of an increasing concentrations experiment. Approximately 30 seconds after adding PBS and starting data acquisition, a 1 fM droplet of CREB is added. Approximately 3 minutes later, a 1 pM droplet of CREB is added followed by a 1 nM droplet 3 minutes after that. After each subsequent droplet is added, the sensor responds with a shift in ΔPower quickly, and slowly settles as the sensor reaches steady-state.............................................................. 88 Figure 5-11: Sensing data showing the results of a competitive binding experiment. Approximately 30 seconds after adding PBS and starting data acquisition, a 1 nM droplet of BSA is added. xi Approximately 3.5 minutes later, a 1 nM droplet of CREB is added. After each droplet is added, the sensor responds with a shift in ΔPower quickly, and slowly settles as the sensor reaches steady-state. After the CREB is added the sensor has a stronger response than when the BSA was added. ............ 89 Figure 5-12: Results from glycerol sweep experiments are shown. For different concentrations of glycerol, ΔPower has large variations. In addition, ΔPower has large variations across all wavelengths. ......................................................................................................................................... 90 Figure 5-13: Diagrams showing a top and side view of the microfluidic channels. The channels were 100 μm by 40 μm. There are two channels for the two devices per sample and one channel for the solution being tested to flow across the sensing or coupling region of the device. .............................. 91 Figure 5-14: Schematic diagram of soft lithography process. A bare wafer is coated with SU-8 photoresist dvia a spin-coating procedure. UV lithography is used to transfer a pattern from a mask to the photoresist. After developing, a master mold is created. The mold is covered in PDMS and cured. Once the PDMS is removed from the master, microfluidic channels are left in the PDMS and it is ready to be attached to a device. .................................................................................................... 92 Figure 6-1: Schematic diagram displaying the important angles used in the theoretical analysis as light travels through the sensor system. The variable α represents the angle between the applied force and the fast and slow axes of the PM fiber, the variable β represents the angle between the linearly polarized light and the fast and slow axes of the PM fiber, the variable γ represents the angle between the polarimeter axes and the fast and slow axes of the PM fiber, and the variable ϕ represents the angle of rotation that the fast and slow axes go when stress is applied to the PM fiber. ...................... 98 Figure 6-2: Diagram of the experimental setup. A CW laser is attached to an in-line polarizer, which is in turn attached to a PM fiber. The PM fiber is run under the sample to be tested and attached to a polarimeter. The polarimeter is attached to a laptop computer, which records the data it measures. The sample is placed on a load-frame, which provides reference measurements as well as compression for the sample. ............................................................................................................... 100 Figure 6-3: Plot of raw data on the surface of the Poincaré sphere. As stress is applied to the fiber, the polarization state changes will trace out a circle on the surface of the sphere. The arrow on the graph indicates how the polarization changes with applied stress. .............................................................. 103 Figure 6-4: Diagram indicating the effect of angles a) β and b) γ on the circular trace from stress on the fiber. a) With changing β, the circular trace will have different horizontal positions on the Poincaré sphere, and therefore different radii. b) With changing γ, the circular trace will rotate around on the equator, about the vertical axis of the Poincaré sphere. ..................................................................... 104 Figure 6-5: Diagrams indicating the effects of the angles δ and α on the circular trace. a) A circular trace for some given values of δ and α. b) If δ is changed, the starting point of the circular trace is rotated to a new position around the circle, but the arc-length of the circular trace remains constant. In this graph, only δ has been changed and α retains its original value. c) If α is changed, the circular trace traces out more of the circle. The starting point remains the same, but the arc-length of the circular trace increases. This graph has the same value of δ, and therefore the same starting point, as in b). Changing α can also cause the circular trace to trace out less of the circle and for the arc-length to decrease. ............................................................................................................................................. 105 Figure 6-6: a) Graph showing the circular trace of polarization state changes on the Poincaré sphere. The phase angle of the traced out circle can be used to generate ΔPol. b) A graph showing analyzed raw data (ΔPol) vs. time. The stokes parameters measured by the polarimeter and shown on the graph in a) are used to calculate ΔPol............................................................................................................... 106 xii Figure 6-7: The circular trace created by stressing the optical fiber is always the intersection of the Poincaré sphere with a vertical plane. The circular trace lies on both the sphere and the vertical plane. ............................................................................................................................................................ 107 Figure 6-8: A view of the Poincaré sphere, vertical plane, and circular trace from above. From this angle, it is easy to see that the x-coordinate of the circular trace will be a rotation of s 1 and s 2 based on the angle θ center . ......................................................................................................................................... 108 Figure 6-9: Graph depicting a calibration curve generated from the Matlab script and fitting algorithm. The black square points show the relationship between ΔPol and applied force, f [N/m], as calculated by the fitting algorithm. The red dashed line is a 3 rd -order polynomial fit to the relationship derived from the theoretical fit of the circular trace. The polynomial fit equation is shown on the graph and is used to calibrate the raw data measured by the polarimeter and turn it into applied force, which can in turn be converted to applied stress. .................................................................................................... 110 Figure 6-10: Stress-strain curves for 5:1 base:curing agent PDMS samples measured with a) a 980 nm laser and PM fiber b) a 1550 nm laser and PM fiber. The trace of red circles is data measured by the Instron load frame and the trace of black squares is data measured by the fiber sensor. The blue and green dashed lines are 3 rd -order polynomial fits for the load frame and fiber sensor, respectively. Despite some noise, there is very good agreement between the reference and fiber sensor data. ...... 113 Figure 6-11: These plots show the calculated values for the Young’s modulus from both the load-frame measurements as well as the fiber sensor data for both a) 980 nm and b) 1550 nm testing wavelengths. The Young’s modulus values calculated from runs that were self-consistent were averaged and the average values are presented here with their standard deviations. ..................................................... 114 Figure 6-12: Noise histograms characterizing the level of noise measured by the fiber sensor in four different environments and two different wavelengths, shown with their normal distribution curves. Plots a-d) show data taken at 980 nm and plots e-h) show data taken at 1550 nm. Plots a,e) were taken in the materials analysis lab, plots b,f) were taken in a chemistry lab on a countertop, plots c,g) were taken inside of a laminar flowhood, and plots d,h) were taken on top of a vibration-isolating optical table. ....................................................................................................................................... 116 Figure 6-13: Surface plots of the sensitivity curves given in a) Equation 6-11 and b) Equation 6-12. a) Shows the sensitivity of the sensor operating at 980 nm for a given applied stress and interaction length. b) Shows the sensitivity of the sensor operating at 1550 nm for a given applied stress and interaction length. ............................................................................................................................... 118 Figure A-1: Diagram showing the unique band structure of graphene at a Dirac point. a) Normally, the Fermi energy is right where the valence and conduction bands meet, and any energy transition is allowed, meaning graphene has no band gap and will absorb incident photons of any wavelength. b) If the graphene is biased electrically, the Fermi energy will move, and some transitions will not be allowed, inducing a band gap of a desired energy level. .................................................................... 126 Figure A-2: Cross-section of waveguide splitter device with electrodes on either side of the splitting region. ................................................................................................................................................. 127 Figure A-3: Images of graphene coated waveguide splitter devices at a) 20x zoom and b) 50x zoom. ... 128 Figure B-1: A schematic diagram of a trapezoidal waveguide with a homeotropic liquid crystal coating. The liquid crystal molecules, shown in yellow, align themselves perpendicularly to the silica substrate, which in this case is the waveguide itself. .......................................................................... 133 Figure B-2: a) An SEM image close up of a waveguide channel for a device that has been spin-coated with a LC polymer. The liquid crystal builds up underneath the waveguide arm and does not coat the top of the waveguide channel. b) An SEM image of the end of a waveguide that has been dip-coated with a LC polymer. The liquid crystal build-up is even more severe in this case. ............................ 134 xiii Figure B-3: Graphs showing results of thickness models. For each graph, thickness (in angstroms) is shown as a function of a) time annealed and spin speed, b) temperature annealed and spin speed, c) spin speed and water amount and d) temperature annealed and water amount. ................................. 137 xiv List of Tables Table 3-1: Values of key simulation parameters......................................................................................... 33 Table 4-1: Effective refractive indices for TE and TM modes ................................................................... 56 Table 4-2: Chemicals and ratios used for silica sol-gel synthesis. .............................................................. 69 Table 5-1: Refractive Indices of Glycerol Solutions................................................................................... 86 Table 6-1: Noise levels given in ΔPol and stress for different wavelengths and measuring environments ............................................................................................................................................................ 117 Table B-1: Parameters varied independently for phase one ...................................................................... 136 Table B-2: Parameters co-varied in different permutations for phase two ............................................... 136 Table B-3: Refractive index values at 633 nm. Red numbers indicate large standard deviations and ranges.................................................................................................................................................. 138 xv Abstract Optical and photonic devices have numerous applications that range from telecommunications to various types of sensing. They are widely studied in a number of geometries and are often designed to accomplish a specific application. However, sometimes specific devices can be improved upon or intelligently designed to enable new applications. To accelerate this process, a combination of modeling and experimental efforts can be applied. The focus of this dissertation is the improvement and modification of existing waveguides and waveguide devices for new applications by the use of complementary modeling and experimental research to better understand their behavior. In the first part of the dissertation, fluorescent waveguide sensors are modeled and developed for use in spatiotemporal fluorescent measurements. The experimental measurements are enabled by the unique geometry of the devices which was better understood after modeling them. Subsequently, suspended, silica-on-silicon waveguides are explored in a new geometry and their behavior is modeled as well. Improvements to these devices are attempted via the use of conformal coatings of different materials. Next, a suspended silica-on-silicon waveguide splitter is developed for use as a biosensor, with models used to predict its sensing behavior. Finally, an optical fiber-based polarimetric stress sensor for use with visco-elastic materials is developed by improving upon previous work and generalizing the theoretical analysis and modeling of these types of polarimetric sensors. 1 Chapter 1: Overview 1.1 Motivation Integrated optical devices have become an area of intensive and interesting research over the last few decades. There are a number of advantages gained by integrating optical devices onto a silicon wafer, including manufacturability, miniaturization and uniformity [1, 2]. They have proven to be a useful platform for many applications, from communications to sensing. Specifically, integrated optical waveguides and waveguide devices have proven to be very useful for a number of applications within these fields, including as biosensors, modulators, and power splitters [3-8]. Additionally, many models have been developed to better understand the behavior of these devices. Integrated optical waveguides are used to guide and confine light along their length and have been used for numerous applications, including biological and chemical sensing. One method of using waveguides for sensing is as a fluorescent sensor. In this approach, the optical field of the waveguide interacts and excites a bound fluorescent probe molecule. The emission from the probe can be detected using several different methods. These devices are well understood and typically operate with light radiating above the device. However, with appropriate design, these devices can be made to operate with light coupling back into the waveguide [9]. In order to truly unlock the capabilities of such a device, its operation needs to be understood in a more detailed way [10]. This is investigated in this dissertation by using simulations and experiments to develop a rigorous theoretical model of the device operation. This model also allows the device to be used in manner of operation it was not originally intended for. Taking advantage of the long length of these special waveguides, spatiotemporal fluorescent measurements are performed which allow for the characteristics of fluorescent dyes to be characterized in space and time. Additionally, waveguides are important components of integrated photonic circuits, as they serve as conduits for guiding signals in-between the areas of the circuit that will process the signals, making them the photonic circuit analogue to a wire in an electrical circuit [11, 12]. However, waveguides differ 2 from wires in many ways as a signal-carrying medium. Many of these differences are due to the signal carrier (photons vs. electrons) having a much smaller wavelength than in wires. This can lead to important tradeoffs. For example, waveguides can have significant loss if they have bends that are too sharp, but straight segments typically have much lower propagation loss than wires [11, 13-15]. Because of this, optical waveguides have primarily found commercial use as a means to send signals over very long distances (as in the case of optical fiber) but have been commercially used less frequently in small- scale integrated optical circuits. One goal of researchers is to reduce the minimum tolerable bend radius of integrated waveguides without increasing the propagation loss of the waveguide too much, allowing for more compact integrated circuits [16-18]. These characteristics of waveguides typically depend on the size and geometry of the device as well as the refractive indices of the materials used to fabricate the device. The propagation and bending loss of suspended silica-on-silicon waveguides is investigated in this dissertation using experiments and modeling, and improvements are sought by applying coatings of various materials. One common and important photonic circuit component is a waveguide device known as a power splitter. This device does what its name implies: it splits the power from an incoming signal into two or more outgoing signals. The ratios of the splitting can depend on the geometry of the device and the refractive indices of the materials as well as the wavelength of light used to carry the signal [4, 19]. These waveguide splitter devices can be used for purposes other than simply splitting power by modifying them slightly. For example, with the addition of an electro-optic material, one could fabricate a tunable splitter or modulator [20], or with a surface chemistry functionalization, one could fabricate a biosensing device [21]. The feasibility of using suspended silica-on-silicon waveguide splitters as biosensors is investigated in this dissertation, leveraging the unique geometry of the splitters as an advantage to gain high sensitivity. Optical fiber has been used not only for communications applications but also for sensing. Some of these sensors have required free-space components, but recent advances in making in-line components as well as more powerful measurement and characterization tools have opened up new options for 3 designing fiber optic sensors. Polarimetric stress sensors are one class of device that strongly relied on free space components for operation [22, 23]. In this dissertation, improvements are sought to these devices to remove free space components, enhance their ease of use, and enable them for use with visco- elastic materials [24]. This is accomplished by expanding upon the theoretical models used to understand these sensors and using new equipment. 1.2 Chapter overview Chapter 2 gives background on integrated optical waveguides and other waveguide devices. Basic features and operation of these devices are explored, especially as they relate to applications developed later in this dissertation. These features include modes and polarization in waveguides, as well as different loss mechanisms. Optical birefringence is examined with a general overview useful for understanding its use in manipulating polarization states of light. The fabrication and geometry of suspended silica-on-silicon waveguides and waveguide splitters are described. Different approaches for characterizing waveguides are discussed as well as the specific waveguide testing setup used in these experiments. Chapter 3 explores a rigorous theoretical model for a specific fluorescent embedded waveguide biosensor design which relies on back-coupling of fluorescent radiation for detection. A finite-difference time-domain (FDTD) simulation model is developed to understand operation of this device as intended compared to how fluorescent waveguide biosensors are normally operated. The simulations are verified by performing a series of experiments using fluorescent laser dye and a spectrograph. Finally, using a special LabVIEW program for data analysis, spatiotemporal fluorescent measurements are performed using these waveguide sensors. Chapter 4 discusses improvements to suspended silica-on-silicon waveguides and investigates their bending characteristics. Trapezoidal waveguides are fabricated and characterized with finite element method (FEM) simulations. Their propagation loss is found experimentally via a series of cut-back measurements. The bending loss of the devices and minimum tolerable bend radius are investigated with 4 experimental measurements as well as a series of FDTD simulations. Power-dependent measurements are made, indicating very low non-linearities even at high input power. Further improvements are sought to the suspended waveguides by using sol-gel thin film coatings to enable single-mode operation which will lower the propagation loss. Chapter 5 details investigations into using the suspended silica-on-silicon waveguide splitters as biosensors. Several FDTD simulation models are developed of the splitters as biosensors. Devices are fabricated and functionalized to be specific to an analyte, and the functionalization is verified. Several detection experiments are performed to assess the ability of the sensor to detect the analyte, the specificity of the functionalized sensors, and the sensitivity of the sensor to background index changes. In an attempt to improve the sensor, microfluidic channels are fabricated from polydimethylsiloxane (PDMS), and an attempt is made to integrate them with the waveguide sensors. Chapter 6 presents the development of a portable, fiber-optic system for measuring the Young’s modulus of visco-elastic materials. A generalized theoretical analysis of the sensor operation consisting of a series of transfer matrices is developed. The design of the sensor is outlined, highlighting the removal of free-space components and addition of a polarimeter. This combination increases the sensitivity and ease-of-use of the sensor. Experiments are performed in a materials characterization lab with reference data taken alongside the sensor data. Details about the data analysis process and fitting algorithm developed to calibrate the sensor are outlined. The sensor is calibrated and the Young’s modulus as measured by the sensor and the reference data are compared. Finally, noise measurements are performed in order to assess the sensitivity of the polarimetric sensor in different environments, and the ultimate sensitivity of the sensor is calculated. Chapter 7 outlines future directions of the work contained in this dissertation. It includes strategies for further improving suspended silica-on-silicon waveguides and enabling the silica suspended splitters as biosensors with repeatable results. Additionally, future steps for the development of the portable fiber- based sensor are outlined, including a path for use as a tool to measure the Young’s modulus of biomaterials, such as tissue, for diagnostic purposes. In Appendix A, research involving coating graphene 5 splitters is explored. This work was never fully investigated, and strategies for completing the research are discussed. In Appendix B, projects that were undertaken by students I mentored which did not result in any publications are detailed. These include investigations into various waveguide coatings and investigating the tuning of sol-gel thin film properties via modifying synthesis procedures. Chapter 1 References [1] A. Scandurra, "Silicon Photonics: The System on Chip Perspective," Silicon Photonics Ii: Components and Integration, vol. 119, pp. 143-168, 2011. [2] C. Kopp, S. Bernabe, B. Ben Bakir, J. M. Fedeli, R. Orobtchouk, F. Schrank, et al., "Silicon Photonic Circuits: On-CMOS Integration, Fiber Optical Coupling, and Packaging," Ieee Journal of Selected Topics in Quantum Electronics, vol. 17, pp. 498-509, May-Jun 2011. [3] A. Crespi, Y. Gu, B. Ngamsom, H. J. W. M. Hoekstra, C. Dongre, M. Pollnau, et al., "Three- dimensional Mach-Zehnder interferometer in a microfluidic chip for spatially-resolved label-free detection," Lab on a Chip, vol. 10, pp. 1167-1173, 2010 2010. [4] X. M. Zhang and A. M. Armani, "Suspended bridge-like silica 2 x 2 beam splitter on silicon," Optics Letters, vol. 36, pp. 3012-3014, Aug 1 2011. [5] R. Song, H. C. Song, W. H. Steier, and C. H. Cox, "Analysis and demonstration of Mach-Zehnder polymer modulators using in-plane coplanar waveguide structure," Ieee Journal of Quantum Electronics, vol. 43, pp. 633-640, Jul-Aug 2007. [6] F. S. Ligler, "Perspective on Optical Biosensors and Integrated Sensor Systems," Analytical Chemistry, vol. 81, pp. 519-526, Jan 2009. [7] N. Bamiedakis, T. Hutter, R. V. Penty, I. H. White, and S. R. Elliott, "PCB-Integrated Optical Waveguide Sensors: An Ammonia Gas Sensor," Journal of Lightwave Technology, vol. 31, pp. 1628-1635, May 2013. [8] J. W. Kim, K. J. Kim, M. C. Oh, J. K. Seo, Y. O. Noh, and H. J. Lee, "Polarization-Splitting Waveguide Devices Incorporating Perfluorinated Birefringent Polymers," Journal of Lightwave Technology, vol. 29, pp. 1842-1846, Jun 2011. [9] R. Duer, R. Lund, R. Tanaka, D. A. Christensen, and J. N. Herron, "In-Plane Parallel Scanning: A Microarray Technology for Point-of-Care Testing," Analytical Chemistry, vol. 82, Nov 1 2010. [10] M. C. Harrison and A. M. Armani, "Spatiotemporal Fluorescent Detection Measurements Using Embedded Waveguide Sensors," IEEE Journal of Selected Topics in Quantum Electronics, vol. PP, 2014. [11] K. Okamoto, "Recent progress of integrated optics planar lightwave circuits," Optical and Quantum Electronics, vol. 31, pp. 107-129, Feb 1999. [12] C. J. Wang and L. Y. Lin, "Nanoscale waveguiding methods," Nanoscale Research Letters, vol. 2, pp. 219-229, May 2007. [13] Y. A. Vlasov and S. J. McNab, "Losses in single-mode silicon-on-insulator strip waveguides and bends," Optics Express, vol. 12, pp. 1622-1631, Apr 2004. [14] J. F. Bauters, M. J. R. Heck, D. John, D. X. Dai, M. C. Tien, J. S. Barton, et al., "Ultra-low-loss high-aspect-ratio Si3N4 waveguides," Optics Express, vol. 19, pp. 3163-3174, Feb 2011. [15] S. Romero-Garcia, F. Merget, F. Zhong, H. Finkelstein, and J. Witzens, "Silicon nitride CMOS- compatible platform for integrated photonics applications at visible wavelengths," Optics Express, vol. 21, pp. 14036-14046, Jun 2013. [16] C. Xiong, W. H. P. Pernice, and H. X. Tang, "Low-Loss, Silicon Integrated, Aluminum Nitride Photonic Circuits and Their Use for Electro-Optic Signal Processing," Nano Letters, vol. 12, pp. 3562-3568, Jul 2012. 6 [17] X. L. Zuo and Z. J. Sun, "Low-loss plasmonic hybrid optical ridge waveguide on silicon-on- insulator substrate," Optics Letters, vol. 36, pp. 2946-2948, Aug 2011. [18] X. Zhang, M. Harrison, A. Harker, and A. M. Armani, "Serpentine low loss trapezoidal silica waveguides on silicon," Optics Express, vol. 20, pp. 22298-22307, Sep 24 2012. [19] S. Y. Lin, E. Chow, J. Bur, S. G. Johnson, and J. D. Joannopoulos, "Low-loss, wide-angle Y splitter at similar to 1.6-mu m wavelengths built with a two-dimensional photonic crystal," Optics Letters, vol. 27, pp. 1400-1402, Aug 2002. [20] M. Liu, X. B. Yin, E. Ulin-Avila, B. S. Geng, T. Zentgraf, L. Ju, et al., "A graphene-based broadband optical modulator," Nature, vol. 474, pp. 64-67, Jun 2 2011. [21] D. Duval, A. Belen Gonzalez-Guerrero, S. Dante, J. Osmond, R. Monge, L. J. Fernandez, et al., "Nanophotonic lab-on-a-chip platforms including novel bimodal interferometers, microfluidics and grating couplers," Lab on a Chip, vol. 12, 2012 2012. [22] T. H. Chua and C. L. Chen, "FIBER POLARIMETRIC STRESS SENSORS," Applied Optics, vol. 28, pp. 3158-3165, Aug 1989. [23] R. C. Gauthier and J. Dhliwayo, "BIREFRINGENT FIBEROPTIC PRESSURE SENSOR," Optics and Laser Technology, vol. 24, pp. 139-143, Jun 1992. [24] M. C. Harrison and A. M. Armani, "Portable polarimetric fiber stress sensor system for visco- elastic and biomimetic material analysis," Applied Physics Letters, vol. 106, p. 191105, 2015. 7 Chapter 2: Background This dissertation deals with a broad variety of topics, but they are all connected by the fundamental theme of optical devices. This chapter provides a brief review of the background necessary to understand the common areas connecting the various chapters. It assumes that the reader has a basic background in electrical engineering, specifically in the area of electricity and magnetism. Background that is specific to a certain chapter will not be given here, but will instead be included in the chapter of relevance. 2.1 Integrated Optical Devices In our increasingly inter-connected world, it is important for communication technology to be efficient, reliable, and cost-effective. Computing and communication devices which route and process information have been the subject of intensive research by both industry and academia. Initial research in optical communications began using free space optical components, such as lenses, mirrors and gratings. However, these systems are extremely large and require constant alignment. In order for this technology to move from a benchtop system, it needs to be reduced in size and made more robust. One approach is to replace the free space optical components with the on-chip or integrated equivalents [1-6]. Integrated devices have proven themselves exceptionally useful in the electronics industry. Therefore, it is straightforward to apply the same approach for translating optics. Integrated devices, whether they are electronic or optical, are devices or components that are fabricated directly onto a silicon (or other semiconductor) wafer. Processors, light emitting diodes (LEDs) and charge-coupled device (CCDs) are all examples of electronic integrated devices which are commonly used throughout society. Examples of integrated optical devices include waveguides, power splitters, lasers, and resonators (Figure 2-1). Similar to electronic devices, one of the largest benefits of integrated optical devices is the ability to fabricate multiple components on to a single chip, creating higher complexity devices [7]. Fabricating these devices together onto a single chip also makes the final optical system more compact, allowing more components in a smaller footprint. 8 The methods used to fabricate integrated devices have several advantages. Optical devices require precise alignment and have strict fabrication tolerances. This precision is relatively easy to achieve and automate with modern microfabrication methods. When fabricating integrated devices, typically many copies of the same devices or device systems are fabricated on a wafer simultaneously. The combination of automation and mass-production allows integrated optical devices to be fabricated inexpensively and reliably. Therefore, by developing integrated optical devices, these systems can translate from a research setting to practical, commercial use. Figure 2-1: Examples of typical integrated photonic devices. Pictured are computer renderings of a) a waveguide, b) a coupler or splitter, and c) a resonator or laser. Methods for fabricating integrated electronic devices have been investigated and used in industry for several decades. Therefore, by leveraging these existing techniques, optics and photonics researchers have been able to quickly develop methods for integrating optical devices, greatly accelerating this research field. Furthermore, because integrated optical and electronic devices largely rely upon the same fabrication technologies and platforms, they can be integrated together to create hybrid opto-electronic devices. These opto-electronic devices are already studied in relation to improving the speed and efficiency of communications networks [8]; however, they have numerous applications beyond communications. Therefore, studying hybrid opto-electronic devices will be beneficial as integrated optical devices begin to outperform and replace traditional electronic devices that perform similar functions for high-performance computing and communications. Despite the numerous advantages and benefits of integrated optical devices, there are still many challenges to overcome in bringing them to widespread use outside of laboratories. Many optical devices have a larger footprint, consume more power, and/or are less efficient than their electronic counterparts. (c) (b) (a) 9 These are all obstacles to address before optical devices replace integrated electronic devices in high- powered computing and communications. There are also other uses for integrated optical devices which are being pursued through research. For example, many of these photonic devices make efficient sensors for a variety of applications. These applications include biosensing and gas sensing [9-12]. 2.2 Optical Waveguides Optical waveguides are typically fabricated from dielectric materials and rely on total internal reflection to guide a light signal along their length. They consist of two parts: the higher refractive index core, and the lower refractive index cladding. Based on Snell’s law, by injecting light into the waveguide core at shallow angles, the light will be confined within the device. If the coupling conditions meet the criteria for total internal reflection (TIR) at the core:cladding interface (incident angle>critical angle), then the light will travel along the length of the waveguide without leaking out of the core (Figure 2-2). In this way, waveguides are able to transmit light-encoded signals over long distances. Optical waveguides can exist in a variety of geometries, from non-integrated waveguides such as optical fiber to numerous integrated geometries. Additionally, they can typically be fabricated with very low loss, as is the case with optical fiber, which is a common and ubiquitous example of an optical waveguide. Figure 2-2: Total internal reflection in a waveguide. Light at a shallow enough angle will be reflected and continue to propagate in the waveguide core. 10 2.2.1 Waveguide Modes and Polarization Based on how light couples into a waveguide, different optical field distributions can exist that propagate within the waveguide core. These optical field distributions are referred to as the modes of the waveguide (Figure 2-3). Depending on the size of the core and the difference in refractive index between core and cladding (refractive index contrast), there can be different numbers of supported modes. For certain geometries and refractive index contrasts, only one mode will be supported, and these waveguides are referred to as single-mode. This mode will always be the lowest-order mode of the waveguide, and it is also referred to as the fundamental mode. Researchers often distinguish between transverse-electric (TE) and transverse-magnetic (TM) modes. TE modes have all electric field components perpendicular to the direction of propagation (but not magnetic field components), and TM modes have all magnetic field components perpendicular to the direction of propagation (but not electric field components). TE and TM modes are generally considered to be orthogonal polarizations in waveguides, because their primary electric and magnetic fields will be pointing in different, orthogonal directions. Hybrid modes, or those with both magnetic and electric field components in the direction of propagation, can also exist. Furthermore, several waveguide modes can also coexist in the waveguide at the same time. Understanding waveguide modes is vital to understanding other important properties of waveguides which will be discussed next. Figure 2-3: Hybrid modes of a step-index optical fiber. At the left is the fundamental mode, and the mode-orders increase from left to right. Step-index optical fibers can also have higher-order modes with nodes in the radial direction. 11 Depending on the application, the polarization can be important. However, TE and TM modes are not always both supported, depending on the geometry and material of a particular waveguide. Furthermore, just as superpositions of modes can exist in a waveguide, superpositions of polarizations (or polarization modes) may also exist. This can lead to many different polarization states aside from simple horizontal and vertical linear polarization. In optical fiber, since the modes supported by the fiber are similar to plane-wave modes, any arbitrary polarization state may exist, including circular and elliptical polarizations. 2.2.2 Evanescent Field Evanescent fields arise as a necessity because electric and magnetic fields cannot be discontinuous, even at a boundary. Because of this fact, part of the optical field in a dielectric optical waveguide will extend beyond the boundary of the waveguide core (Figure 2-4). This portion of the optical field is referred to as the evanescent field. The length of the field depends on the relative dielectric constants of the waveguide material and the cladding as well as the optical mode area and profile of the specific mode. In all cases, the evanescent field decays exponentially from the core- cladding boundary towards the cladding and travels along with the rest of the lightwave signal. The evanescent field does not represent loss in the waveguide, but it can be involved in loss mechanisms of the waveguide. It is difficult to measure the evanescent field length directly. Therefore, in order to determine the evanescent field length, one will typically either solve Maxwell’s equations for the waveguide geometry to determine a solution for the particular mode, or in the case of complicated or un- solvable geometries, use a simulation to determine the extent of the evanescent field. Typically, the waveguide cladding will be thick enough so as to completely contain the evanescent field, preventing the optical field from being disturbed by the surrounding environment. 12 Figure 2-4: Cartoon indicating the locations of the confined and evanescent fields in a fundamental waveguide mode. The hashed areas indicate the evanescent field of the optical field distribution. Furthermore, the evanescent field is often exploited in waveguide devices. If the waveguide cladding is intentionally removed, then the evanescent field is allowed to interact with the environment. This interaction can result in a variety of effects on the light signal depending on the waveguide or device geometry [10], and it can be a high fidelity approach for detection. These effects can be a phase shift or signal attenuation in the case of a standard waveguide, or some other output shift in the case of waveguide devices. Some of the waveguide devices which are pursued in this dissertation operate by exploiting the reach of the evanescent field outside of the waveguide core. 2.2.3 Optical Birefringence Optical birefringence, also referred to as anisotropy, is when a material or structure has a different refractive index along different axes. Birefringence is not a property unique to optical waveguide structures, but it may be present in them. There are many naturally occurring birefringent crystals (such as calcite) that have been used for different applications. Birefringent materials may be uniaxial or biaxial. Uniaxial materials have two axes where the refractive index is the same and one where it is different. Biaxial crystals have three different refractive indices for all three axes. An optical wave 13 passing through a birefringent material will experience the refractive index along the axis which is aligned with its polarization. If the polarization is in-between axes, then each component of the polarization will see the refractive index along the axis with which it is aligned. In other words, if the polarization is broken into components along each refractive index axis, then each component will experience the refractive index of its corresponding axis (Figure 2-5). This property is often usefully exploited in devices in which control over the distinct polarization states is desired. Figure 2-5: Cartoon diagram depicting un-polarized light passing through a birefringent block of material. Since the parallel and perpendicular polarizations see two different refractive indices (n 1 and n 2 ), they are refracted at different angles (θ 1 and θ 2 ) through the block of material. Birefringence may be caused by a number of mechanisms. In crystals, it typically arises from the particular crystal structure of the material. Birefringence can also be caused by internal stresses within a material, which may have internal or external causes. For waveguides, the birefringence may be due to either of these factors. A device may be fabricated with specific materials in order to give it birefringent properties or it may have some structural component which induces the stress necessary to create birefringence. In a waveguide, typically the birefringence will be uniaxial, but since the light is primarily 14 polarized in a horizontal or vertical direction (TE or TM), the birefringent axes are often referred to as the fast and slow axes. The fast axis will be the axis with a lower refractive index and the slow axis will be the axis with a higher refractive index. 2.3 Waveguide Loss Mechanisms There are several mechanisms which can result in loss in a waveguide. These can be intrinsic to the waveguide, such as material loss, scattering loss or bending loss. Alternatively, they can be extrinsic to the device, such as coupling loss. Each loss mechanism arises from unique material and geometrical considerations and will be discussed individually in subsequent sections. The signal level of waveguides can be measured in absolute power, typically W or mW, or in dBm, which is a unit of measurement defined as decibels relative to 1 mW of power. Decibels (dB) are a relative unit of measurement and are defined as follows: 𝑑𝐵 = −10 log ( 𝑃 2 𝑃 1 ) (2.1) In the case of dBm, P 1 = 1 mW and P 2 is equal to the power being measured. Loss is typically measured in dB [13]. This is useful because loss defined this way is a relative measurement that is independent of any absolute power levels and can therefore be applied no matter what the operating power is. Several of the loss mechanisms are distance dependent. Because of this, loss can often be reported in dB/(unit length), where unit length can be anywhere from nm to km. In almost all circumstances, it is desirable to minimize loss, as this can increase the signal-to-noise ratio, reduce the amount of amplification needed in transmitting signals, and improve device operation. 2.3.1 Absorption Loss All materials will absorb some part of the light that travels through it, even if only slightly. This material absorption results in absorption loss in waveguides. Materials with higher refractive indices usually have higher absorption as well [14]. Like the refractive index of a material, absorption is 15 wavelength-dependent. Absorption loss is the primary component of loss in most waveguides, and it describes how much the signal is attenuated as it travels down the length of the waveguide. It is typically given in dB/(unit length). For non-integrated optical waveguides, such as optical fiber, the absorption loss is typically very low. For example, at 1550nm, Corning SMF-28 fiber has losses of ~0.1 dB/km [13]. For integrated optical waveguides, the propagation loss can have more variation depending on the specific geometry and materials used, but it can be anywhere in the dB/µm to dB/m range. Although this is much higher than non-integrated waveguides, it is acceptable because integrated waveguides are typically used to route signals short distances on single chips. In contrast, non-integrated waveguides are typically used to send signals long distances, taking advantage of their low absorption losses. 2.3.2 Scattering Loss Scattering loss occurs due to surface roughness or other artifacts on the boundary between the core and cladding of optical waveguides. Photons can scatter off of the rough surface of waveguides and out of the guided modes. Higher-order modes are especially susceptible to scattering losses, as they tend to have a higher percentage of their optical field in proximity to the core-cladding boundary [13]. Surface roughness is typically a result of imperfections from the fabrication processes, and it can often be minimized by modifying or optimizing the fabrication procedures. Frequently, integrated waveguides have no cladding layer. This effectively makes the surrounding air the cladding layer, which can lead to an apparent amplification of the scattering losses because of the sharp contrast in index. If any foreign particles (dirt, water, etc.) end up on the waveguide boundary, they can cause additional scattering losses. Because these losses occur as the light travels down the waveguide, they are typically given in dB/(unit length). 2.3.3 Propagation Loss In practice, when measuring losses in a waveguide, it is difficult to separate absorption loss from scattering loss. The absorption loss could theoretically be calculated if you know what modes are 16 propagating in your waveguide (and how much power is in each mode) and the material loss of the materials used to fabricate the waveguide. However, this is not always feasible. Therefore, when measuring losses experimentally, researchers often refer to the propagation loss or transmission loss. These terms mean the same thing and refer to a combination of absorption loss and scattering loss [13]. The propagation loss combines these losses into one value which says how much attenuation there is in the waveguide as light travels down it, and it is given in units of dB/(unit length). 2.3.4 Bending Loss Unlike bends in an electrical circuit, bends in waveguides can cause loss of varying degrees, depending on the bend radius. There are a few mechanisms that give rise to this loss, and some of these mechanisms can be visualized in different ways. One way to conceptualize bending losses is to return to a ray-optics, particle-view of light travelling down the waveguide. In this approach, light will bounce off the sides of the waveguide core at large enough angles to enable total internal reflection. However, when a bend is introduced, the angle of incidence of light on the waveguide core boundary is decreased. Therefore, the total internal reflection condition may no longer be satisfied, and some of the light may be able to propagate through the core-cladding interface and radiate away from the device (Figure 2-6). Therefore, as the bend radius becomes smaller, the incident angle further decreases, resulting in an increase in bending loss[13]. Conceptualizing losses from a wave-view of light reveals another way to look at this radiation loss. Solving Maxwell’s equations for a curved section of waveguide reveals that the modes supported by this curved section are similar to, but different from, the modes supported by a straight section of waveguide. The evanescent field on the outside edge of these modes does not simply decay exponentially as it did in the straight waveguide, but it has a small sinusoidal component. This sinusoidal component corresponds to the radiation losses from the waveguide bend, and it grows larger as the bend decreases in size. Additionally, the modes supported by a curved section of waveguide typically have a higher optical field distribution towards the outside edge. This results in increased interaction with the surface of the 17 waveguide and higher scattering losses in a bent section. Finally, because the mode shapes of the straight waveguide section and the curved waveguide section are not the same, there will be some loss as the light transitions from the straight waveguide mode to the bent waveguide mode, similar to the coupling loss. This loss will be high in waveguides with tighter bends, as the mode mismatch will be larger. Figure 2-6: Bending loss in a waveguide. Going around a bend, the guided light will not satisfy the critical angle condition, and some light will be transmitted out of the waveguide core and lost. All of these factors combine to produce the bending loss of a waveguide. The bending loss is often given in dB/(unit length) for a specific bending radius of the waveguide. Sometimes it can be given as dB/(unit length · unit radius), if that relation can be determined from experimental measurements or theoretical analysis. Most often, bending losses are not reported directly. Instead, researchers will report a critical radius or minimum tolerable bending radius [13]. The measured bending losses for radii above the bend radius are small enough to be negligible. For practical applications, a minimum tolerable bending radius is usually adequate. 2.3.5 Coupling Loss In order to get light into a propagating mode in an optical waveguide, it must be coupled in from an outside source. This can be done via a variety of different mechanisms, such as butt-coupling and gratings, but none of them are completely ideal. In other words, there will always be some amount of power lost. This loss during the transfer is called coupling loss. 18 Most coupling loss occurs due to mode mismatches. The light being coupled into the waveguide will never quite match the mode shape of the supported modes of the waveguide. This mode mismatch prevents all of the power incident on the waveguide input from transferring into the guided mode of the waveguide, resulting in loss [13]. Additionally, the input surface may have some roughness on it or may simply reflect some of the incident light, which will result in some scattering losses, especially for butt- coupling methods. There are methods to reduce the loss due to both mode mismatch and surface roughness/reflection. For example, by coating the endface of the waveguide with an anti-reflection coating, the coupling efficiency can be improved. However, this increases the overall fabrication complexity, and the additional fabrication step increases the probability of damage to the device. Since coupling losses are constant (independent of length), they are typically given in dB. 2.3.6 Insertion Loss The insertion loss is the total loss of an optical component if it were placed into a photonic circuit. It is a combination of all the loss mechanisms present in that particular optical device and represents the total loss of the device. Insertion loss is a useful concept because it describes how much loss your device would add to a system if it were to be inserted into that system. Insertion loss usually includes coupling loss, as the device may be inserted into a system which is not fully integrated. However, if the device were inserted into a fully integrated photonic circuit or system, the coupling loss may be reduced or eliminated. 2.4 Suspended Silica-on-Silicon Waveguides and Devices Researchers use numerous different material systems to fabricate waveguides and waveguide devices, each with their own inherent advantages and disadvantages [3]. Often, CMOS-compatible materials and fabrication methods are used in order to keep costs down and ensure easy integration with integrated electrical components. Typically, the waveguide geometry, device design and application are the deciding factors for what materials system is used. 19 Silicon-based materials systems are highly attractive due to their CMOS compatibility and the wealth of fabrication techniques that can be borrowed from decades of silicon-based electronics manufacturing [7, 15]. One of the major hurdles in creating silicon-based materials systems is that silicon has a higher refractive index than many of its derivative materials, such as silicon dioxide (also referred to as silica) and silicon nitride. This can create difficulties for certain waveguide designs, as the waveguide core must have a higher refractive index than its surrounding cladding in order to effectively guide light. Therefore, if silicon is used as a substrate, the refractive index contrast is inverted. One common solution is to use silicon-on-insulator (SOI) material systems, in which a layer of silicon dioxide (which has a much lower refractive index) is grown on the silicon substrate and further fabrication takes place on top of this new low-index layer. In this dissertation, however, a different approach is taken to solving this challenge. 2.4.1 Waveguide Geometry and Fabrication Many of the waveguides and waveguide devices discussed in this dissertation have a unique, suspended silica-on-silicon geometry [16]. This geometry mimics that of a silica microtoroid resonator, another type of optical device that is studied in my research group. The waveguide consists of two parallel, circular, silica waveguide channels connected by a silica membrane that is suspended on top of a silicon pillar (Figure 2-7). In this configuration, the silica waveguide channels act as the waveguide core while the surrounding air acts as the cladding. The membrane is fabricated to be thin enough that, for the wavelengths of light at which we normally operate the waveguide (visible and near-IR), no light leaks into the membrane for the propagating modes. The diameter of the waveguide channels is approximately 5 µm, and the channels are approximately 80-60 µm apart from each other. The silica membrane, which is supported by the silicon pillar, is fabricated to be approximately 1 µm thick. 20 Figure 2-7: Cross section of waveguide geometry. a) A schematic cross section. The blue section represents SiO 2 and the green section represents Si. b) SEM image of the end-facet of a waveguide. The waveguide fabrication begins with standard photolithographic techniques. Photoresist is deposited via spin-coating onto a blank silicon wafer that has a 2 µm thick layer of thermally grown silicon dioxide on it. Using a UV mask aligner, an initial pattern of rectangles is transferred onto the photoresist layer. The exposed photoresist is washed away using developer, and the pattern of rectangles is transferred to the silica layer using a buffered HF etch that isotropically etches the silica. These steps are then repeated to transfer a pattern that protects the edges of the rectangular pads (where the waveguides will be) and allows us to etch the membrane down to approximately 1 µm thick. Next, a dry etch step is performed using xenon difluoride (XeF 2 ) gas. The XeF 2 isotropically etches silicon, but does not etch silicon dioxide, leaving the rectangular pads elevated on silicon pillars. After that, we use a high- powered CO 2 laser to reflow the edges of the silica pads into round waveguide channels. The silica portion of the device is opaque to the wavelength of the CO 2 laser, while the silicon portion is transparent to it. Because of this, the silica absorbs the energy from the laser, heats up, and melts, forming round channels due to surface tension. The laser simply passes through the silicon, which acts as a heat sink, causing the silica membrane layer in contact with and immediately adjacent to the silicon pillar to remain un-melted. The final step in the fabrication process is to cleave the ends of the waveguides in order to form smooth input and output facets for high-efficiency coupling of light into and out of the waveguides [16]. (b) (a) 50 μm 21 Figure 2-8: Schematic diagrams showing several steps in the waveguide fabrication process. These diagrams show the waveguides after a) the first photolithography step, b) the second photolithography step, c) the XeF 2 etching step, and d) the CO 2 laser reflow step. 2.4.2 Waveguide Splitter Devices The waveguide splitter devices discussed in this dissertation are based off of, and very similar to, the straight waveguides [17]. They are also fabricated out of silica that is suspended on a silicon pillar. The outside edges of the device are very similar to the waveguides, with channels approximately 5 µm in diameter spaced approximately 80 µm apart, supported by a 1 µm thick membrane held up on a silicon pillar. Towards the center of the device, the waveguide channel spacing begins to taper down, and the membrane thickens to 2 µm adjacent to the waveguide channels (Figure 2-9). After this tapered region is the coupling or splitting region, where the waveguide channels are approximately 20 µm apart and the membrane is 2 µm thick across its entire width. The splitting region is also completely suspended above the silicon substrate, with no silicon pillar for support. An important feature to note is a bump that occurs at the beginning of the tapered region (Figure 2-9), which arises as a natural consequence of the fabrication procedures and is important to the operation of these devices [18]. 22 The splitter devices split light in one waveguide channel from the visible to near-IR wavelength range into two output channels with roughly equal power. After light is coupled into the input channel, it travels down the waveguide normally until it encounters the bump at the beginning of the tapered region. The bump causes new modes to be excited in which part of the optical field is contained in the narrow 2 µm thick membrane in the tapered region (Figure 2-9). Once the light reaches the suspended splitting region, it excites new modes which span across the 20 µm wide membrane and both waveguide channels. The light contained in those modes is split by the abrupt thinning of the 2 µm thick membrane to 1 µm at the end of the splitting region. The light continues to travel up the output tapered region and the output waveguide channels until it reaches the end of the device. Figure 2-9: Scanning electron micrograph (SEM) of a waveguide splitter device. The waveguide, tapered and splitting region can all be seen, and the bump which causes splitting to occur is indicated with an arrow on the bottom waveguide channel. The splitting ratio, or amount of light split into each output waveguide channel, depends on the wavelength of the light being used, the length of the splitting region, and the effective refractive indices of the modes initially excited. It is possible to change the splitting ratio by changing any of these three variables. However, for the default design of the device with a fixed splitting region length and air as the waveguide cladding, the splitting ratio remains around 50/50 for wavelengths in the visible to near-IR 300 μm 23 range. This type of broadband behavior is challenging to obtain in an on-chip splitter, and it is one of the unique features of this specific device. 2.4.3 Waveguide Splitter Fabrication The waveguide splitter devices are fabricated using a similar procedure as the straight waveguides [17]. In fact, the first part of the fabrication process follows the steps for the straight waveguides exactly, using a different mask design for the splitters. When the XeF 2 step is performed, a silicon pillar that is 2-4 µm wide is left in the splitting region. This is important so that when performing the CO 2 laser reflow step the splitting region does not completely melt away. If the silicon pillar in the splitting region is too thick, the silica will not reflow properly and waveguide channels will not completely form. In contrast, if the pillar is too thin, the channels may end up touching or part of the splitting region may melt away. After the devices are reflowed and the ends are cleaved, they undergo an additional XeF 2 etching step. This final XeF 2 etching removes the silicon pillar completely and leaves the splitting region of the devices completely suspended above the silicon substrate. 2.4.4 Advantages and Disadvantages of This Design There are several advantages to the suspended silica-on-silicon material system and geometry employed for these waveguides and devices, some of which have already been touched on briefly. One big advantage is that the devices are based on a true silicon-substrate materials system, which allows easy integration with silicon-based electronic devices. Furthermore, silica is a very attractive material to use as the waveguide core. Because it has a lower refractive index than silicon and silicon nitride, silicon dioxide is usually relegated to the role of cladding. However, silicon dioxide has less absorption loss in the visible and near-IR wavelength ranges. Additionally, its refractive index has less variation across these wavelength ranges and lower X 2 and X 3 coefficients (nonlinear behavior), which enables more reliable broadband operation from devices fabricated in silica. The difference in refractive index between air and silica, the cladding and core of our waveguide system, is much higher than in many other 24 integrated waveguides. This results in higher confinement of the optical field in the waveguide core and prevents the light from interacting too much with its surroundings. However, the lack of a deposited cladding layer allows the light to interact with the area immediately around the waveguide, which can be exploited for many useful applications. There are some disadvantages to this material system and geometry, as well. One that results from the high-index contrast between the core and cladding and the physical dimensions of the waveguide channels is that these devices operate in multi-mode fashion. Higher order modes typically have higher loss and can decrease device performance. Additionally, it can be difficult to controllably excite a known set of modes when coupling light into a multi-mode device, which can make it difficult to theoretically predict the exact behavior of a given device. Unfortunately, based on the developed fabrication method, it is not possible to reduce the dimensions of the device in order to achieve single-mode behavior. Additionally, due to the material properties of silica and size of the CO 2 laser used for reflow, the waveguide channels can often have a “wavy” characteristic that is difficult to eliminate. This can cause more loss in the waveguides plus additional problems in the splitter devices. Because the splitting relies upon the bump at the beginning of the tapered region to inject modes into the membrane, any difference in mode shape before the light reaches the bump can change how much light couples into the membrane. Waviness in the waveguide devices can change the mode shape, and therefore can slightly alter the behavior of the splitters, making it more difficult to reliably predict how they will behave. However, given that this device was invented in 2011 and is therefore incredibly new, it is not surprising that there are several aspects to their operation which still pose challenges. Therefore, these waveguides and waveguide splitters offer unique advantages that make their study and use for specific applications worthwhile. 2.5 Waveguide and Waveguide Device Characterization Both the waveguides and waveguide splitter devices were characterized on the same, custom waveguide testing setup built onto an optical table. The test setup uses a tapered, lensed fiber to couple 25 light into the waveguide at the input facet (OZ Optics). The lensed fiber is designed for operation at a free-space wavelength of 1550 nm and, at that wavelength, it has a spot size of 2 µm and a focal length of 12 µm ± 2 µm. The lensed fiber is mounted on a three-axis stage controlled by a motorized stage controller (Newport) with 30 nm resolution along each axis. The waveguide sample is affixed to a chuck that is mounted to the optical table, preventing it from moving while the lensed fiber is aligned. At the output end of the waveguide, an aspheric lens (Thorlabs) mounted on a three-axis stage is used to focus the output light from the waveguide on to a beam profiler (Thorlabs) which is attached to a computer workstation (Figure 2-10). Figure 2-10: Schematic diagram of the waveguide testing setup. The beam profiler can be used to measure the profile of the focused output beam and can also be used to measure the power output from the waveguide. Additionally, the output light may be focused on to a free-space power meter in order to measure the output power from the waveguide. The aspheric lens may be replaced by another lensed fiber mounted on a three-axis stage with a motorized controller if the output from the waveguides or splitter devices needs to be coupled back into a guided fiber mode. Two cameras attached to the computer workstation are used to monitor the top- and side-views of the waveguide, assisting with aligning the lensed fibers. A 4-wavelength continuous wave (CW) diode laser (Thorlabs) with fiber-coupled lasers at wavelengths of 405 nm, 633 nm, 980 nm and 1550 nm is used as a 26 laser source for the waveguide characterization. Additionally, a CW laser (Agilent) tunable between 1520 nm and 1630 nm with a resolution of 0.1 pm at 1550 nm can be used as a source as well. 2.5.1 Loss measurements Loss measurements are performed in a very straightforward manner. First, the power coming out of the lensed fiber, P in , is focused using the aspheric lens and measured. Next, the power coming out of waveguides of different lengths, P out , is measured and recorded. The measured powers, in mW, are used to determine the loss for each device according to the following formula: 𝐿𝑜𝑠𝑠 (𝑑𝐵 ) = −10log ( 𝑃 𝑜𝑢𝑡 𝑃 𝑖𝑛 ) (2.2) These losses are then plotted on a graph, where they should have a linear trend. By fitting a line to the measured values, both the coupling loss and the propagation loss for the wavelength of light used in the measurements are determined. The y-intercept of the fitted line is approximately equal to the coupling loss as it is the loss when the waveguide has zero length (Figure 2-11). Additionally, the slope of the fitted line is the propagation loss in units of [dB/cm]. Figure 2-11: Diagram showing how propagation loss is calculated from measurements. The transmission loss measured from several waveguide lengths is plotted and fit to a straight line. The slope of the line is equal to propagation loss and the y-intercept is equal to the coupling loss in the system. 27 2.5.2 Waveguide Splitter Output The output from the waveguide splitters is a little more unique and requires a little bit more work to analyze. Since the splitter splits the input signal into two outputs, the beam profiler is used to look at the outputs so that they can be resolved as two separate light beams (Figure 2-12). The beam profiler measures an x- and y-cross section of the incident light beam and produces two intensity vs. position graphs: one in the x-direction and one in the y-direction. The power of each beam is obtained by numerically integrating each intensity peak. Since the splitting ratio, the ratio of power split into each output channel, is a typical figure of merit for splitter devices, the two powers determined from the numerical integration are added and each is divided by the total to obtain a percentage of power in that peak. These percentages may then be compared, giving the splitting ratio. Figure 2-12: A three-dimensional plot of arbitrary intensity vs. position for a waveguide splitter. The intensity is measured by the beam profiler and two distinct peaks of roughly equal power can be clearly seen. Often, data obtained from the waveguide splitters using the beam profiler comes in large quantities. This is due to a number of reasons, including taking wavelength-sweeps of the devices and measuring the output change of the device over a period of time. In order to analyze data more efficiently, a custom Matlab script that automates the process is used. For each splitting ratio datapoint (typically one point in time or one wavelength in a sweep), a file is saved from the beam profiler which has an intensity vs. position profile. Using that profile, the script numerically integrates the outputs, 28 correcting for baseline noise from the beam profiler, and then uses that power to find the splitting ratio as indicated above. Finally, the script graphs each point vs. time or wavelength, with the x-axis scaled appropriately. Using this Matlab script, data analysis for any given data set can be accomplished in under a minute. Chapter 2 References [1] M. Liu, X. B. Yin, E. Ulin-Avila, B. S. Geng, T. Zentgraf, L. Ju, et al., "A graphene-based broadband optical modulator," Nature, vol. 474, pp. 64-67, Jun 2 2011. [2] K. Okamoto, "Recent progress of integrated optics planar lightwave circuits," Optical and Quantum Electronics, vol. 31, pp. 107-129, Feb 1999. [3] S. Romero-Garcia, F. Merget, F. Zhong, H. Finkelstein, and J. Witzens, "Silicon nitride CMOS- compatible platform for integrated photonics applications at visible wavelengths," Optics Express, vol. 21, pp. 14036-14046, Jun 2013. [4] A. Scandurra, "Silicon Photonics: The System on Chip Perspective," Silicon Photonics Ii: Components and Integration, vol. 119, pp. 143-168, 2011. [5] D. Pan, H. Wei, and H. X. Xu, "Optical interferometric logic gates based on metal slot waveguide network realizing whole fundamental logic operations," Optics Express, vol. 21, pp. 9556-9562, Apr 2013. [6] A. Abbasi, M. Noshad, R. Ranjbar, and R. Kheradmand, "Ultra compact and fast All Optical Flip Flop design in photonic crystal platform," Optics Communications, vol. 285, pp. 5073-5078, Nov 2012. [7] C. Kopp, S. Bernabe, B. Ben Bakir, J. M. Fedeli, R. Orobtchouk, F. Schrank, et al., "Silicon Photonic Circuits: On-CMOS Integration, Fiber Optical Coupling, and Packaging," Ieee Journal of Selected Topics in Quantum Electronics, vol. 17, pp. 498-509, May-Jun 2011. [8] M. K. Iyer, P. A. V. Ramana, K. Sudharsanam, C. J. Leo, M. Sivakumar, B. L. S. Pong, et al., "Design and development of optoelectronic mixed signal system-on-package (SOP)," Ieee Transactions on Advanced Packaging, vol. 27, pp. 278-285, May 2004. [9] A. M. Armani, "Label-free, single-molecule detection with optical microcavities (August, pg 783, 2007)," Science, vol. 334, pp. 1496-1496, Dec 2011. [10] A. L. Washburn and R. C. Bailey, "Photonics-on-a-chip: recent advances in integrated waveguides as enabling detection elements for real-world, lab-on-a-chip biosensing applications," Analyst, vol. 136, pp. 227-236, 2011. [11] S. Mehrabani, P. Kwong, M. Gupta, and A. M. Armani, "Hybrid microcavity humidity sensor," Applied Physics Letters, vol. 102, Jun 17 2013. [12] N. Bamiedakis, T. Hutter, R. V. Penty, I. H. White, and S. R. Elliott, "PCB-Integrated Optical Waveguide Sensors: An Ammonia Gas Sensor," Journal of Lightwave Technology, vol. 31, pp. 1628-1635, May 2013. [13] D. Derickson, Fiber Optic Test and Measurement: Prentice Hall, 1998. [14] B. E. A. S. a. M. C. Teich, Fundamentals of Photonics, 2nd ed. Hoboken, New Jersey: John Wiley & Sons, 2007. [15] T. Miya, "Silica-based planar lightwave circuits: Passive and thermally active devices," Ieee Journal of Selected Topics in Quantum Electronics, vol. 6, pp. 38-45, Jan-Feb 2000. [16] A. J. Maker and A. M. Armani, "Low-loss silica-on-silicon waveguides," Optics Letters, vol. 36, pp. 3729-3731, Oct 1 2011. [17] X. M. Zhang and A. M. Armani, "Suspended bridge-like silica 2 x 2 beam splitter on silicon," Optics Letters, vol. 36, pp. 3012-3014, Aug 1 2011. 29 [18] S. Soltani and A. M. Armani, "Optimal design of suspended silica on-chip splitter," Optics Express, vol. 21, pp. 7748-7757, Mar 25 2013. 30 Chapter 3: Using embedded waveguide sensors for spatiotemporal fluorescent measurements 3.1 Introduction Waveguides and waveguide devices are not only useful in photonic circuits, but are also useful for a wide variety of other applications. Many different waveguide geometries have proven themselves useful as various sensors [1-5]. Optical waveguide sensors have become so widely researched, in part due to their advantages over electrical sensing platforms. For example, optical waveguide sensors can perform over a wide pH range, whereas electrical sensors may be strongly affected by the pH of the environment [6, 7]. Because optical waveguides can employ various sensing mechanisms, they usually take advantage of the unique strengths and attributes of the sensing platform. Many researchers are working on making more compact and robust devices and integrating them with additional components, such as detectors or laser sources, to make the device more capable and useful. Advances in making devices more compact, as well as giving them additional functionality, will help to propel these miniaturized sensing platforms closer to commercialization. 3.1.1 Embedded waveguide sensors Embedded waveguides are a sub-type of integrated waveguides that are partially or completely embedded in the substrate material, which acts as a cladding. They are a commonly used photonic component and can be fabricated in a number of ways. Typically, they have a rectangular cross section and are made with silicon-on-insulator (SOI) material systems. Embedded waveguides, along with other waveguide types, are often used in fluorescent sensing systems. Fluorescent sensing systems rely on the evanescent field of the waveguide to excite a fluorescent moiety. The emission from the fluorophore acts as the sensing signal. These systems often rely on sandwich assays to perform their sensing. In these assays, the waveguide surface is functionalized with an antibody that will bind specifically to the molecule of interest. Once the molecule of interest (antigen) binds to its respective antibody, the device is exposed to a secondary solution that contains secondary antibodies that have been fluorescently labeled. When light is coupled into the waveguide, the 31 evanescent field will excite any fluorophores attached to the bound antibodies, which will then emit a fluorescent signal. Fluorescent sensing schemes often have a specific sensing area where the waveguide cladding has been thinned to allow greater evanescent field interaction with the surrounding environment, including bound fluorophores [8, 9]. This type of fluorescent sensing with an embedded waveguide sensor lends itself well to integration with microfluidic schemes, and the sensing is most often performed in solution. Additionally, this type of integration with microfluidic channels, creating an optofluidic device, is desirable for producing lab-on-a-chip (LOC) type systems. Conventionally, the fluorescent signal is collected and monitored from above the waveguide. However, this sort of “overhead” detection is problematic when pursuing more integrated and densely packed LOC devices. In order to condense these embedded waveguide sensor systems further, other detection schemes must be investigated. 3.1.2 Fluorescent back-coupling Another fluorescent detection mechanism for embedded waveguide sensors is based on back- coupling of fluorescent emission into the waveguide. Ordinarily, when a dipole such as a fluorophore radiates, each photon is emitted in a single, non-controlled direction. However, dipoles can be made to radiate preferentially in one direction by placing them on top of a dielectric stack [10-12]. This preferential radiation depends on the refractive indices and thicknesses of the dielectric layers. Therefore, by appropriately designing an embedded waveguide system, dipole radiation from a fluorophore can be made to preferentially couple back into the waveguide (Figure 3-1). The preferential coupling is strong enough to result in a detectable signal. Currently, there is a waveguide detection system that exploits preferential fluorescent back- coupling based on stacked dielectric waveguide devices [12]. Because the refractive indices of the stacked dielectric layer influence preferential dipole radiation (and therefore preferential coupling) of fluorescence into the waveguide, these systems are sensitive to changes in the refractive index of the surrounding environment. Thus, subtle changes in the environment can significantly affect the 32 performance of these embedded waveguide sensors. Previously, no rigorous theoretical model for the dependence of fluorescent back-coupling on environmental factors and how it affects embedded waveguide sensors based on this effect existed. However, such a model is developed in this chapter and experimentally verified. Figure 3-1: Schematic diagram indicating emission paths for fluorescent radiation which can be experimentally monitored. This diagram excludes laser light used to excite the fluorophores. 3.2 Experimental Methods 3.2.1 Computer modeling and simulations Two- and three-dimensional finite-difference time-domain (FDTD) simulations are used to model the embedded waveguide sensors based on fluorescent back-coupling. The simulations were used to investigate the mechanism of fluorophore radiation coupling into the waveguide, as well as to characterize the amount of light coupled into the waveguide versus how much light is radiated in other directions. For most simulations, a 2D model was used in order to minimize computation time. However, for a subset of the modeling design parameters, a 3D simulation was performed to verify the accuracy of the 2D simulations, and similar results were found. The model used in the simulations is given in Figure 3-2 and is based off of experimentally obtained design parameters. In addition, the number of dipoles, dipole spacing, dipole orientation, cladding thickness, and cladding and sensing well indices of refraction were varied to account for experimental and fabrication variations. 33 Figure 3-2: Schematic diagram of device geometry used for simulations. D is distance between dipoles, ϕ 1 and ϕ 2 are angles of alternating dipoles, t is the thickness of the cladding in the sensing well, L is the length of the simulation, and n 1 , n 2 , and n 3 are the refractive indices of the cladding, waveguide core, and sensing well, respectively. The locations of power monitors in the simulation are indicated by red lines, and the boundaries of the simulation region are indicated by a dashed line. Three different simulation models were run, and the parameters varied in these models are given in Table 3-1. The embedded waveguide devices were designed to operate with fluorescent emission sources operating at 642 nm wavelength, so that is the wavelength that was used for the dipole radiation sources in the model. The simulation was centered on the sensing well region and run at a length (L) of 500 µm in order to allow the light that is coupled into the waveguide to settle into a steady-state propagating mode. Power was monitored in five places: the total amount of power radiated from the fluorophores, the power radiated above the waveguide, the power radiated below the waveguide, and the power coupled into the waveguide radiating to the right and left, as can be seen in Figure 3-2. The power monitors around the waveguide extended 0.2 µm into the cladding in order to capture power from the evanescent field, which also extends about 0.2 µm into the cladding. The waveguide cladding (n 1 ) was modeled as silica with a refractive index of 1.4355 and the waveguide core (n 2 ) was modeled as silicon nitride with a refractive index of 1.98. These materials match the materials used in the physical devices. Table 3-1: Values of key simulation parameters Model t (µm) D (µm) # dipoles ϕ 1 (º) ϕ 2 (º) n 1 n 3 1 0.010 0.1-0.8 2 0-90 90 1.4355 1.33 2 0.008-0.012 0.1 10 90, 90, 0 90, 0, 0 1.4-1.6 1.33 3 0.010 0.1 10 90 0 1.4355 1-1.4 34 In the first set of simulations, the distance between adjacent dipoles (D) and the relative angle of the dipoles (φ 1 and φ 2 ) were varied. These variations were used to study how interactions between individual fluorophores affected the efficiency of dipole coupling into the waveguide. The second set of simulations used 10 dipoles oriented in three different relative angle (φ 1 and φ 2 ) patterns. The thickness of the cladding layer (t) was varied from 8 to 12 nm and the refractive index of the cladding above the waveguide (n 1 ) was varied from 1.4 to 1.6 in these simulations. The ranges of thickness and refractive index were selected to represent the fabrication tolerances of the material deposition. The last set of simulations was run with a varying refractive index of the sensing well (n 3 ) and 10 dipoles in a fixed orientation. The refractive index range of the sensing well, from 1 to 1.4, was chosen to cover the typical range found in common biological solutions. All 2D simulations were run using the sweep function available in Lumerical. These sweeps were set to give outputs of the power transmitted through each power monitor mentioned above. The total power measured from the fluorophore dipoles was used to normalize the results. Additionally, the 2D simulations were run with a simulation length of 3 ps and a minimum mesh size of 0.25 nm. The simulation accuracy was set to 3 for these simulations to ensure accurate results with a reasonable amount of computation time. 3.2.2 Device design and fabrication Devices for this project were provided by a third party company. They were fabricated using a procedure described in a previous work [12] with the addition of one extra step. In order to fully characterize these devices using our waveguide testing setup, the end of the device was cleaved and polished, exposing both input and output ends of the waveguide. The main body of the waveguide is 0.2 µm thick and 100 µm wide, and approximately 8.6 mm long after being cleaved. Additionally, there is a sensing well, as shown in Figure 3-2, where the upper cladding has been mostly removed, which allows the evanescent field of the guided light to excite fluorophores. The length of the sensing well is approximately 4.6 mm. 35 3.2.3 Experimental setup and measurements The first set of experimental measurements performed were simple transmission loss measurements using the testing setup and methods described above. These measurements were performed in order to characterize the uniformity of the devices and to check for fabrication errors or imperfections. For the biosensing experiments, LD-700 perchlorate dye dissolved in toluene was used to simulate emission from a fluorescently labeled biolayer. This dye, when dissolved in toluene, absorbs at around 658 nm and emits at around 680 nm. It was selected due to its high quantum yield and robustness in air and liquid environments [13]. Two different sets of experiments were performed in order to verify the results from the simulations. In one set, the dye was suspended in liquid toluene solution in the sensing well, and in the second, the dye was deposited on the surface of the sensing well as a dry film. By using two experiments, it was possible to study the effect of the refractive index of the environment on the behavior of the device. Furthermore, in the case of the dry film experiments, the laser dye is located precisely at the surface of the sensing well, which is where fluorescently-labeled molecules would bind in a typical biosensing experiment with a surface-functionalized device. Simulations determined that the evanescent field extends 0.2 µm beyond the core of the waveguide, which is sufficient penetration to excite the fluorescent dye in both the dry and wet experiments. 36 Figure 3-3: Schematic of testing set-up. Light is coupled into the device from the left via a lensed fiber, and output is collected with a filter and spectrograph a) from the output of the waveguide and b) from above the sensing well of the device. The modified waveguide testing setup shown in Figure 3-3 was used to measure the fluorescence from the dye. The main difference between the testing setup used here and that described in the background of this dissertation is that the aspheric lens and beam profiler were replaced with a bandpass filter (Thorlabs, 680 nm center wavelength, 10 nm FWHM) and a fiber-coupled spectrograph (Andor). A 658 nm CW laser from the testing setup was used as the input to the device. In these complimentary measurements, both the power and wavelengths of light radiating out of the end of the device and the power and wavelengths of light radiating from the top of the device are monitored. The bandpass filter was used to isolate the dye-emitted fluorescence from the input laser light. The spectrograph and filter are first placed at the output of the waveguide and used to monitor the transmitted power through end emission. Next, they are placed above the sensing well to monitor the power radiated perpendicular to the waveguide (vertical emission). In both cases, the filter was placed between the device and the spectrograph tip. Additionally, background scans were used to normalize the data and remove pump laser light not filtered out, and data was acquired at a rate of 1 sample/second. 37 Additional spatiotemporal fluorescence measurements were performed using the dry films. These experiments studied the spatiotemporal excitation of the dye using the top-view camera in conjunction with the bandpass filter to monitor the fluorescence visually. In order to capture the fluorescence signal on video, the camera settings were optimized at 1 frame/second, 997.56 ms exposure time, and 5 MHz pixel clock. The video data produced by these measurements was analyzed with a custom LabView program that measured the intensity of whole columns of pixels at every horizontal pixel position for every frame of a given video. This allowed the fluorescent decay of horizontal positions on the waveguide to be measured with an accuracy of one pixel. Therefore, the analysis resolution is only limited by the resolution of the camera used to make the videos for analysis. Additionally, it is possible to measure the bright fluorescing area as it propagates down the length of the waveguide and measure its propagation rate. 3.3 Data and Results 3.3.1 Transmission loss measurements The transmission loss measurements verified that the fabrication process of the devices was uniform across various wafers. The coupling loss of these devices was estimated with a simulation because it is not possible to easily vary their length without destroying their biosensing capabilities. Using this coupling loss, the transmission loss was calculated to be approximately 0.3 dB/cm [14]. 3.3.2 Simulation and modeling Findings from the simulations are summarized in Figure 3-4, Figure 3-5, and Figure 3-6. The data is presented in three percentages: the percentage of optical power coupled into the waveguides and propagating to the left and right (end emission), the percentage of optical power radiated above the dipoles and the waveguide (vertical emission), and the percentage of optical power radiated below the waveguide and lost into the substrate. The percentage was calculated by dividing the power picked up by the monitors in each of the three locations just mentioned by the total power being radiated from the dipoles. Calculating the optical power as a percentage has two important benefits. First, it allows a 38 comparison of results from different simulations, where dipole orientations may have affected the total power being output by the dipoles. Second, it provides a unitless number which can be used to compare the simulation results to the experimental results, making the simulations more useful. The results from the two-dipole simulation (Table 3-1, Model 1) are shown in Figure 3-4. These results demonstrate that the percentage of power radiated to the various regions varies dramatically with a limited number of dipoles. There are a few trends; for example, the dipoles tend to couple more light into the waveguide when they are both oriented perpendicularly (φ = 90°) to the waveguide. However, this trend is not true for every dipole spacing that was simulated. The general trend is reasonable since dipoles radiate less energy in the direction of their axis of oscillation. Thus, when a dipole is oriented parallel to the waveguide, one would expect less light to be coupled in to the waveguide since less light is being radiated in that direction. Unfortunately, these simulations by themselves do not provide much insight into the practical operation of the devices. They do indicate that the distance and angle between two dipoles can have a very large impact on how much radiated light gets coupled into the waveguides. This highlights a unique aspect of the modeling, which was studying how the coupling is affected by the placement and angle of dipoles. The placement and angle of the dipoles not only affects the amount of power coupled into the waveguides (end emission), but also the amount of power radiated above the waveguide (vertical emission) and lost into the substrate below. In particular, the two dipole simulations demonstrate a need to study how the embedded waveguide sensor devices react to being covered in a large number of fluorophores in order to accurately model their behavior [14]. 39 Figure 3-4: Results from 2-dipole simulation (Model 1). Graphs show the percentage of optical power, for different dipole spacing and angles, a) coupled into the waveguide, b) radiating above the waveguide, and c) lost into the substrate below the waveguide. 40 Figure 3-5 shows the results of the first ten-dipole simulation (Table 3-1, Model 2). For clarity, the graphs include the orientation patterns of the alternating dipoles indicated with arrows. These graphs clearly demonstrate that, with a larger number of dipoles, end emission, vertical emission and power lost into the substrate are all more consistent across variations that were simulated. Certain parameters, such as the thickness of the cladding layer above the waveguide, seem to have no effect on the distribution of power radiated from the dipoles at all. Varying the refractive index of the cladding layer had some impact on the power distribution, but even its effects were minimal, typically resulting in variations of less than a percentage point. 41 Figure 3-5: Results from the first ten-dipoles simulation (Model 2). Graphs show the percentage of optical power a) back-coupled into the waveguide, b) radiated above the waveguide and c) lost into the substrate below the waveguide for three different dipole orientations, indicated on the graphs using arrows. The results from this simulation begin to provide insight into the operation of these waveguide sensor devices as well as their tolerance to fabrication variations. For instance, as previously mentioned, the thickness of the cladding layer in the sensing well has little impact on the device performance, as long 42 as it is sufficiently thin. One important exception to tolerance to device variations is the case when all dipoles are oriented parallel to the waveguide. In this case, very little power is coupled into the waveguide, mostly due to the fact that very little power is being radiated in that direction. Fortunately, this is a somewhat unrealistic scenario. In normal operation, it is expected that the fluorophores will be randomly oriented, and the average distribution of radiated light coupled into the waveguides and radiated above and below the waveguides will tend to look more like the percentages found in Figure 3-5 where the dipoles are in an alternating pattern. The second ten-dipole simulation (Table 3-1, Model 3) studied the impact of the refractive index of the environment on device behavior. Figure 3-6 shows that with increasing refractive index, the light radiated below the waveguide decreases, while the light coupled into the waveguide (end emission) and the light radiated above the waveguide (vertical emission) increase. Because less light is radiated into the substrate, the overall efficiency of the device is increased, even though the ratio of end emission to vertical emission remains fairly constant. Figure 3-6: Results from the second 10-dipole simulation (Model 3). Distribution of radiated dipole power for increasing refractive indices of the sensing well environment. 43 3.3.3 Spectrograph sensing measurements The results of the spectrographic sensing measurements were directly compared to the simulation results. In order to make this comparison and because the spectrograph measures data in units of photon counts, the following formula was used to normalize the data: 𝑂 = 𝑊 (𝑊 + 𝐴 ) (3.1) In this formula, the output of the device, O, is equal to the amount of light coming from the waveguide (end emission), W, divided by the sum of that light and the amount of light radiated above the waveguide (vertical emission), A. W and A are measured in photon counts in the case of the spectrograph or percentage in the case of the simulations, and in both cases the resulting O is a unitless number. Because the simulation results are in percentage of total light radiated from the dipoles and the formula for O gives a value relative to the total light measured, it allows for a direct comparison between the simulation and experimental results. 44 Figure 3-7: Spectrograph spectra taken from the output of the waveguide (through), and above the waveguide (above), with fluorescent dye on the device. Figure 3-7 presents spectrograph data taken from the end of the waveguide (end emission) and from above the waveguide (vertical emission) in both air and liquid environments, as described in the methods section. The figure also shows spectra taken from before the fluorescent dye was applied to the device, from which one can see that the 658 nm pump laser is completely filtered out from the area of the spectrum where the fluorescent light appears. This indicates that the signal picked up with the spectrograph is completely from the fluorescent dye. The inset of Figure 3-7 shows artifacts that show up around the center laser wavelength which arise from how the spectrograph software applies background scans, which was used to normalize the data. In this same area on the graphs with dye, there is a dip in the spectra. This dip shows up because the dye is absorbing the pump laser light, reducing the light around the pump wavelength in comparison to the background spectra. Finally, both of the spectra taken with dye from through the waveguide (end emission) and above the waveguide (vertical emission) show fluorescent peaks around 680 nm. This is consistent with spectrofluorometer data (Figure 3-8) taken of 45 the dye dissolved in toluene and indicates that the light being measured with the spectrograph is indeed from the fluorescent dye [14]. Figure 3-8: Emission peak of LD-700 laser dye dissolved in toluene. From the data in Figure 3-7, the output is calculated using formula 3.1 given above for both end emission and vertical emission. The bandpass filter used in our experiments cuts off a portion of the fluorescence peak, so integrating the peaks to obtain the total amount of power radiated in either direction is not the correct approach. Instead, the peak fluorescence counts, approximately 8100 for light radiated above the waveguide and approximately 1400 for light radiated through the waveguide, are used as an indicator of the amount of light radiated in these two directions. Because only light coupled into the waveguide was collected from one direction in the experiments but both directions in the simulations, the peak counts are doubled from light through the waveguide before calculating the output, O. From the peak counts mentioned above, the output, O, was calculated to be 0.257 from the experiment and in the range from 0.278 to 0.327 from the simulations [14]. The results from the spectrograph measurements show agreement with the simulations, confirming the simulation models. The experimental results also shed light on the performance of the 46 devices and how to improve the performance and efficiency. For example, dipole spacing and orientation with respect to the waveguide is a very important factor in overall efficiency of the device. One could improve the performance by designing a surface functionalization that caused the fluorophores to attach to the device with an optimum spacing between each other, and between the fluorophores and the cladding surface. Additionally, there is a slight discrepancy between the experimental and simulation results for a couple of reasons. One factor is that there was a small difference between the refractive index in the sensing well between the experiment and simulations. The second reason has to do with the limited numerical aperture of the spectrograph tip. Alignment of the spectrograph was very important due to its limited numerical aperture, and slight variations in alignment had an effect on the measured results. Therefore, experimental error due to imperfect alignment and the refractive index difference in the sensing well account for the small difference seen in the output results between the simulations and experiments. Finally, it is important to note that the data taken from spectrograph measurements and presented in Figure 3-7 is only from experiments with the fluorescent dye suspended in toluene. This is because in experiments with a dry fluorescent dye film, the amount of light back-coupled in to the waveguide was below the detection threshold of the spectrograph. Thus, only the spatiotemporal fluorescence measurements were made with dry laser dye films. 3.3.4 Spatiotemporal fluorescence measurements Spatiotemporal fluorescence measurements were performed with a camera on the testing setup The data was analyzed with a custom LabView program to produce the results in Figure 3-9. All of these measurements were performed with a dry laser dye film since our camera would not have been able to focus on the waveguide with liquid toluene on its surface. Figure 3-9a presents a series of normalized intensity vs. time graphs generated at every point down the length of the waveguide. Each graph shows a fluorescence excitation and decay curve, and for graphs at positions further down the waveguide this 47 curve is shifted back in time. These graphs demonstrate that the pump laser light is almost completely absorbed by the fluorescent dye. Once the dye begins to bleach and the fluorescent signal begins to decay, the pump laser light can travel further down the waveguide and begin to excite dye at those positions. The graph also shows two distinct decay regions that, when fit to an exponential decay curve, have slightly different time constants. The reason for these distinct decay regions is unknown, but the fact that they have different time constants could indicate that the dye has multiple paths to bleaching [15, 16]. 48 Figure 3-9: Spatiotemporal fluorescence measurements of dry laser film on the waveguide device. Graph a) shows normalized intensity vs. time and the graphs are in order of position further down the waveguide with the rear being the far left of the waveguide. Graph b) shows normalized intensity vs. position and the graphs are in order of increasing time with time increasing from the rear to the front. These two graphs show the same data, but rotated 90 degrees with respect to each other. 49 Figure 3-9b shows normalized intensity vs. position graphs for every point of time throughout the experiment. Each graph shows where the bulk of the emitted fluorescence is on the waveguide. The bulk of fluorescence clearly moves down the length of the waveguide as time progresses. These graphs further demonstrate that the pump laser light is almost completely absorbed by the dye towards the input end of the waveguide sensor, and it is not until this dye begins to bleach that the pump light can travel further down the waveguide to excite more dye. Interestingly, the fluorescence peak shape remains relatively unchanged as it travels down the waveguide; however, there are a few bright fluorescent spots left over after it passes. These bright spots are likely due to clumps of dye that formed as the toluene and laser dye solution was drying on the surface, and the spots and other noise indicate that the dye probably did not completely bleach before the bright fluorescent peak moved further down the device. The data in Figure 3-9 seems to depend on the properties of the dye used, which leads to the conclusion that the density of dye on the surface also played a role in this spatiotemporal fluorescence behavior. These spatiotemporal fluorescence measurements demonstrate how this embedded waveguide sensor system could be used for a variety of interesting applications [1, 2, 17]. The devices could be utilized to measure the fluorescent decay curves of fluorophores in a number of different environments. This kind of information could help researchers understand how fluorescent signals in different waveguide sensing environments change over time and allow them to optimize their biosensing methods and devices. Additionally, these devices could be used to perform cascading assays in which different sections further along the waveguide are functionalized with different receptors for different analytes. One would then get a different signal for each analyte as the fluorescent signal propagated down the waveguide. This opens up the door for complex multiplexing using a single device and would allow for the detection of a complex analyte or for a unique signature of analytes in a target solution. Both complex multiplexing and rapid fluorescence characterization are useful avenues for improving the effectiveness and cost of LOC devices and technology [2, 18]. 50 3.4 Conclusion and Future Outlook In this chapter, a more rigorous theoretical model was developed for back-coupling and vertical detection schemes for embedded waveguide sensors and the model was verified experimentally. The viability of operating the waveguide in either of these modes was demonstrated, although in certain experiments, one mode may be more advantageous than the other. Additionally, it has been demonstrated that the back-coupling operation of the waveguide sensors is tolerant to variations in fabrication, making this method desirable for many types of sensing applications, many of which could be beneficial for LOC devices and research. Finally, using the unique geometry of the waveguide and the properties of fluorescent laser dye, spatiotemporal fluorescence measurements were performed. These measurements were used to produce a series of fluorescent decay curves at arbitrary points along the waveguide and could be useful for fluorescent dye characterization. Furthermore, this spatiotemporal measurement could be used for complex multiplexing that could allow cascading assays that would be attractive for increasing the scope and effectiveness of photonic biosensing devices and LOC devices in general. Chapter 3 References [1] M. J. Lochhead, K. Todorof, M. Delaney, J. T. Ives, C. Greef, K. Moll, et al., "Rapid Multiplexed Immunoassay for Simultaneous Serodiagnosis of HIV-1 and Coinfections," Journal of Clinical Microbiology, vol. 49, pp. 3584-3590, Oct 2011. [2] S. Mandal, J. M. Goddard, and D. Erickson, "A multiplexed optofluidic biomolecular sensor for low mass detection," Lab on a Chip, vol. 9, 2009 2009. [3] M. Medina-Sanchez, S. Miserere, and A. Merkoci, "Nanomaterials and lab-on-a-chip technologies," Lab on a Chip, vol. 12, pp. 1932-1943, 2012. [4] V. M. N. Passaro, C. de Tullio, B. Troia, M. La Notte, G. Giannoccaro, and F. De Leonardis, "Recent Advances in Integrated Photonic Sensors," Sensors, vol. 12, pp. 15558-15598, Nov 2012. [5] C. A. Rowe, S. B. Scruggs, M. J. Feldstein, J. P. Golden, and F. S. Ligler, "An array immunosensor for simultaneous detection of clinical analytes," Analytical Chemistry, vol. 71, pp. 433-439, Jan 1999. [6] A. H. Liu, T. Watanabe, I. Honma, J. Wang, and H. S. Zhou, "Effect of solution pH and ionic strength on the stability of poly(acrylic acid)-encapsulated multiwalled carbon nanotubes aqueous dispersion and its application for NADH sensor," Biosensors & Bioelectronics, vol. 22, pp. 694- 699, Dec 2006. [7] N. Lloret, R. S. Frederiksen, T. C. Moller, N. I. Rieben, S. Upadhyay, L. De Vico, et al., "Effects of buffer composition and dilution on nanowire field-effect biosensors," Nanotechnology, vol. 24, p. 9, Jan 2013. [8] G. L. Duveneck, A. P. Abel, M. A. Bopp, G. M. Kresbach, and M. Ehrat, "Planar waveguides for ultra-high sensitivity of the analysis of nucleic acids," Analytica Chimica Acta, vol. 469, pp. 49- 61, Sep 2002. [9] H. Mukundan, A. S. Anderson, W. K. Grace, K. M. Grace, N. Hartman, J. S. Martinez, et al., "Waveguide-Based Biosensors for Pathogen Detection," Sensors, vol. 9, Jul 2009. 51 [10] L. Polerecki, J. Hamrle, and B. D. MacCraith, "Theory of the radiation of dipoles placed within a multilayer system," Applied Optics, vol. 39, Aug 1 2000. [11] H. Hoekstra and H. B. H. Elrofai, "Theory of optical spontaneous emission rates in layered structures," Physical Review E, vol. 71, Apr 2005. [12] R. Duer, R. Lund, R. Tanaka, D. A. Christensen, and J. N. Herron, "In-Plane Parallel Scanning: A Microarray Technology for Point-of-Care Testing," Analytical Chemistry, vol. 82, Nov 1 2010. [13] T. F. Johnston, R. H. Brady, and W. Proffitt, "POWERFUL SINGLE-FREQUENCY RING DYE-LASER SPANNING THE VISIBLE SPECTRUM," Applied Optics, vol. 21, pp. 2307- 2316, 1982. [14] M. C. A. Harrison, A. M., "Spatiotemporal Fluorescent Detection Measurements Using Embedded Waveguide Sensors," IEEE Journal of Selected Topics in Quantum Electronics, vol. PP, 2014. [15] C. Eggeling, J. Widengren, L. Brand, J. Schaffer, S. Felekyan, and C. A. M. Seidel, "Analysis of photobleaching in single-molecule multicolor excitation and forster resonance energy transfer measurement," Journal of Physical Chemistry A, vol. 110, pp. 2979-2995, Mar 2006. [16] T. Gensch, M. Bohmer, and P. F. Aramendia, "Single molecule blinking and photobleaching separated by wide-field fluorescence microscopy," Journal of Physical Chemistry A, vol. 109, pp. 6652-6658, Aug 2005. [17] M. Lehnhardt, T. Riedl, T. Rabe, and W. Kowalsky, "Room temperature lifetime of triplet excitons in fluorescent host/guest systems," Organic Electronics, vol. 12, pp. 486-491, Mar 2011. [18] H. K. Hunt and A. M. Armani, "Label-free biological and chemical sensors," Nanoscale, vol. 2, pp. 1544-1559, 2010. 52 Chapter 4: Characterization of silica suspended waveguides and performance improvements with inorganic coatings 4.1 Introduction As fundamental components in photonic circuits, waveguides are fabricated from numerous material systems and in a wide variety of geometries [1-4]. The goal of much of this research is to minimize propagation loss and bending loss. The suspended silica-on-silicon waveguides briefly described in Chapter 2 help to minimize propagation loss by using low-loss silica as the waveguide core. Additionally, the high index contrast between the waveguide core (silica) and cladding (air) helps to reduce bending losses. Tight bends are desirable, because they allow more compact devices which in turn allow fabrication of denser photonic circuits. In the first part of this chapter, improvements to the previously described devices were sought for the reflowed waveguides by eschewing the reflow step. Avoiding this step minimizes defects that were thought to arise from the reflow process, allowing for an understanding of propagation and bending losses in waveguides with a trapezoidal cross-section. In the second part of the chapter, one approach for further reducing the propagation loss by using silica sol-gel coatings is explored. 4.2 Trapezoidal waveguides One source of loss for the silica-on-silicon suspended waveguides arises due to the CO 2 laser reflow step. Although this step removes the residual lithographical defects in the waveguide surface, which helps to reduce overall propagation loss, it also results in a nanometer to micrometer waviness in the cylindrical waveguides [5]. Ultimately, this waviness causes more propagation loss than is reduced by removing the surface roughness and the performance of the waveguide suffers. In order to address this issue and explore other waveguide geometries, suspended silica-on-silicon waveguides were fabricated without the reflow step, resulting in devices with a trapezoidal cross-section (Figure 4-1). The trapezoidal shape arises from the isotropic nature of the BOE etchant used. In addition to removing the CO 2 laser reflow step, the devices are cleaved before the XeF 2 etching step. This allows for more reliable 53 fabrication of smooth input and output facets, where cleaving pre-silicon etch is significant. In the future, this could enable further improvements such as the deposition of anti-reflection coatings or polishing of the endface. Both would reduce coupling losses and are not possible with the cylindrical waveguides. Figure 4-1: Fabrication procedure and scanning electron micrograph (SEM) image of the trapezoidal waveguide. (a) 80- m wide silica pads are created through a photolithography and buffered HF etching step. (b) After another photolithography and buffered HF etching step, two 8- m trapezoids on each side of the 80- m trapezoids are created. (c) XeF 2 etching is used to isotropically undercut the silica. (d) SEM side view image of a single waveguide. 4.2.1 Fabricating bent devices In addition to straight devices, curved trapezoidal devices were fabricated and studied. These devices were fabricated using the same process as the straight devices (outlined above), with a different mask to produce the bent shape. The waveguides were fabricated with an inside bend radius (R) that varied from 125 μm to 400 μm with 25 μm intervals (Figure 4-2). The separation between the waveguides was approximately 75 m. Therefore, the outside bend radius ranged from 200 m to 475 m. In this dissertation, when referring to the bend radius, the inner bend radius is being specified. As much as was possible, the overall length of the waveguide arms was kept constant to assist in the calculation of bending losses. In order to ensure that loss measurements would be consistent, the waveguides were fabricated with two 90-degree bends in opposite directions, ensuring that light traveling down the device encounters both an inside bend and an outside bend before exiting the device. 54 Figure 4-2: Scanning electron micrograph (SEM) images of the curved trapezoidal waveguide. (a) An overview of the S- curve waveguide geometry depicting the inner bending radius (R). (b) SEM of the bending section and (c) cleaved end of a curved waveguide device. 4.2.2 Bending Losses in a Trapezoidal Device Due to the unique suspended structure of these devices, there are issues that arise in bent devices that are not present in other waveguide geometries. Because of the dual-channel nature of the devices, the direction of the bend of the waveguide is important. Depending on the bend direction, the light signal in the waveguide can travel through either an outside bend or an inside bend. Unlike most waveguides, the membrane in this structure causes outside bends and inside bends to have different losses. In particular, the inside bends tend to have more loss because they introduce a new loss mechanism [6]. When travelling through an outside bend, the waveguide mode will be distributed towards the outside edge of the waveguide, near the core-cladding interface, just as in other waveguide devices. However, for an inside bend, the mode will be distributed towards the waveguide membrane. This distribution causes light to leak into the membrane, where it will hit the silicon pillar and subsequently leak into the silicon substrate. The membrane loss mechanism of the inner bend tends to be the limiting loss mechanism for the waveguide device. Due to this new loss mechanism, it was important to fully characterize the bending losses of these devices, both through experiment and theory, to study whether this mechanism would prevent their use in practical applications. 55 4.2.3 Simulation Models The optical field profile of the devices was modeled using COMSOL multiphysics at wavelengths of 658, 980, and 1550 nm. COMSOL simulates structures using finite element method (FEM) analysis and was only used to model the optical field of the straight devices. These simulations demonstrate that the optical field of the fundamental mode does not leak into the membrane of the straight waveguides. Additionally, the optical field profiles of rectangular cross-section waveguides were plotted for comparison at the same wavelengths (Figure 4-3). Although the effective mode area for both geometries varies, the electric field distribution in both devices is symmetric, which further indicates that there is no leakage of the optical field into the membrane. Figure 4-3: Finite element method simulation of the electric field distribution (TE mode) of (a) trapezoidal and (b) rectangular cross-sections at 1550nm. (c) Comparison of the intensity of the mode profile along the x (horizontal) direction, normalized to the highest intensity in the trapezoidal device. The FEM simulations were used to calculate the effective refractive indices of the trapezoidal waveguides for both TE and TM polarizations at all three wavelengths simulated. The results are in Table 4-1. The refractive indices are very similar for all three wavelengths and for both TE and TM modes, indicating that the devices should have polarization-independent behavior. Finally, using these effective 56 refractive indices, the refractive index contrast between the core and the cladding was calculated as follows: 𝑐𝑜𝑛𝑡𝑟𝑎𝑠𝑡 = 𝑛 𝑐𝑜𝑟𝑒 2 − 𝑛 𝑐𝑙𝑎𝑑 2 2𝑛 𝑐𝑜𝑟𝑒 2 (4.1) The values for the index contrast for the TE mode are 25.6%, 25.3% and 24.5% for 658, 980, and 1550 nm, respectively. The TM mode behavior was similar. Table 4-1: Effective refractive indices for TE and TM modes Wavelength (nm) TE TM 658 1.4313 1.4305 980 1.4222 1.4199 1550 1.4011 1.3933 Due to the complex three-dimensional shape of the bent device and the desire to include both inside and outside bends in the simulation, they were simulated with finite difference time domain (FDTD) simulations to more accurately capture the device behavior. Specifically, a commercial software package called Lumerical FDTD was used to simulate the bending behavior of the devices. Lumerical allows the user to draw arbitrary 3D geometries, specify the materials that make up the geometry and a light source, and then see how that light propagates through the structure. It uses the geometry and light source as the boundary and initial conditions, respectively, while solving Maxwell’s equations in space and time. Thus, using Lumerical one can see the time-evolution of light in a photonic structure. Additionally, Lumerical is not limited to time-harmonic solutions, making it ideal for complex geometries and large waveguides which can support multiple modes, such as the ones studied in this work. Using Lumerical FDTD, devices were simulated with similar geometrical values to those that were fabricated. Devices with two 90-degree bends were simulated with inner bend radii ranging from 50 μm to 175 μm in 25 μm intervals, along with one device with a bend radius of 250 μm. The simulations used light with a wavelength of 1550 nm, and in the simulations, the light travels first through an inside bend and then through an outside bend. Due to limits of the computer system used to run the simulations, all dimensions of the device as well as the wavelength of light in the simulation were scaled down by a 57 factor of 3, thus the simulations were run with 516.667 nm wavelength light. Using the mesh object in Lumerical, the maximum mesh step was forced to 0.1 μm, and the simulation accuracy was set to the lowest setting so that the simulations would complete within a reasonable amount of time on the computer system being used. Although Lumerical has a material library containing refractive index and absorption values at various wavelengths of each material, due to the wavelength scaling, the SiO 2 material was not used and all materials were set to the user-defined dielectric material with a refractive index of 1.444 (the refractive index of SiO 2 at 1550 nm) [7]. Because the absorption of silica at 1550 nm is so low and the devices were simulated over short distances, no absorption loss is included. Using a mode source object, the fundamental mode of the waveguide was injected into the arm that goes through an inner bend first. By using the fundamental mode of the straight waveguide, the simulation captured the mode mismatch loss that occurs when the waveguide transitions from a straight to a curved section. The experimental devices also started with a straight section. The loss from traveling through the bends was determined using two 2D Y-normal frequency- domain power monitors placed close to the source and at the very end of the waveguide. Both of these monitors were placed and sized such that they contained only the light and evanescent field from the waveguide arm that was guiding the light and as little of the silica membrane as possible. The output monitor was used to measure the power leaving the waveguide arm at the end of the structure (P out ) while the monitor near the source was used to normalize the output for the power that actually coupled into the waveguide (P in ). In Lumerical, the power from a source is usually normalized to 1, but the input monitor ensured accuracy in the loss calculations. Using the difference between P out and P in , the overall loss from the device bends was calculated in dB and then normalized by dividing the loss by the arc-length of the waveguide arms, yielding a result in dB/cm. Calculating the loss in this way allows the simulation results to be directly compared to the experimental results. In addition, one 2D Z-normal monitor captured the field profile overview for the entire structure (Figure 4-4). 58 Figure 4-4: Finite-difference time-domain simulation showing the optical field intensity inside a bent trapezoidal waveguide with an inner radius of 50 m at 1550nm. It is clear that the optical field leaks into the silica membrane in- between the two waveguide arms. The silicon pillar of the device was left out of the simulation due to computational constraints. A significant portion of the light in the simulations leaks through the silica membrane and in some cases reaches the other waveguide arm. In reality, this light would likely be lost into the silicon pillar and would never reach the opposite waveguide arm. However, because just the light in the waveguide arm of interest was monitored, the light that reached the opposite arm in the simulation is still lost, and the simulations still retain a reasonable degree of accuracy. 59 4.2.4 Experimental methods The straight and bent waveguide devices were characterized using the waveguide testing setup described in Chapter 2. Propagation loss measurements were made with the methods described in that chapter as well. Additionally, an erbium-doped fiber amplifier (EDFA) was used to perform high-power measurements with the straight waveguides at a wavelength of 1550 nm. The EDFA output was connected to a 99:1 fiber coupler, with 99% of the power directed to the waveguide through the test setup and 1% of the power directed to a power meter to monitor the power input to the waveguide. Additionally, a free-space neutral-density filter was placed in the output path of the waveguide between the aspheric lens and beam profiler to avoid saturating the beam profiler with high power. Finally, polarization-dependence measurements were made at 1550 nm using an in-line polarization controller between the laser and lensed fiber. The polarization of the input light was varied using this controller, and transmission measurements were made at several different polarization states. 4.2.5 Data and Results 4.2.5.1 Propagation Loss The results from our propagation loss measurements are given in Figure 4-5a. The propagation loss at 658, 980 and 1550 nm is 0.69, 0.59 and 0.41 dB/cm, respectively. Additionally, the coupling loss for these three wavelengths is 2.3, 1.6 and 1.3 dB, respectively [6]. Due to the size of the waveguide, higher order modes are supported, and these loss measurements reflect that. These results were obtained from multiple devices, verifying the reproducibility of the measurements. Figure 4-5b shows the results from the polarization state measurements. The transmission remains nearly constant at various polarization states, verifying the simulation results that the waveguide is polarization independent. 60 Figure 4-5: (a) Propagation loss of the trapezoidal waveguides at 658, 980 and 1550 nm. (b) Transmission at four different polarization states. The propagation loss is proportional to the scattered electric field amplitude and the interface roughness [8], which creates two dominant loss mechanisms: 1) reduction in optical mode confinement and 2) surface scattering due to roughness. The first mechanism usually dominates at longer wavelengths, whereas the second mechanism tends to dominate at shorter wavelengths. The simulations show that the optical field is well confined within the waveguide device and the loss increases with decreasing wavelength, indicating that the second mechanism is responsible for the loss at all wavelengths. Therefore, the primary source of loss for these trapezoidal waveguides is surface scattering due to interface roughness between the core and the cladding [6]. 4.2.5.2 Bending Loss Bending loss was measured experimentally in the same way the propagation loss was measured. However, the losses measured from the bent devices consist of three parts: 𝛼 𝑡𝑜𝑡𝑎𝑙 = 𝛼 𝑝𝑟𝑜𝑝 + 𝛼 𝑐𝑜𝑢𝑝𝑙 + 𝛼 𝑏𝑒𝑛𝑑 (4.2) 61 In this equation, α total is the total measured loss from the device, α prop is propagation loss over the length of the device, α coupl is the coupling loss, and α bend is the bending loss. Because α coupl is a constant, the coupling losses measured earlier may be subtracted from α total to remove that loss from the total loss. The propagation loss over the length of the device (α prop ) is determined by a simple calculation. First, the length of the waveguide arms is measured under an optical microscope (L) and added that to the arc- length of each bent section (R x π/2) to obtain the total length of the device. Second, this total length is multiplied by the propagation loss measured from the straight waveguide devices, 0.69, 0.59 and 0.41 dB/cm for 658, 980 and 1550 nm, respectively. Once α prop and α coupl have been found, they are subtracted from α total to obtain α bend . Finally, the bending loss is normalized by dividing it by the arc-length of the waveguide bend, because each radius tested has a different length. The results from the experimental measurements of bending loss are given in Figure 4-6. In the graph, each point represents a unique device, and at least 3 devices were tested for each radius. Additionally, the graph has exponential fits for all three wavelengths. The exponential fit for all three wavelengths is similar, but the shorter wavelengths have slightly higher loss. This increased loss arises because shorter wavelengths can more easily leak into the membrane. In most waveguides, shorter wavelengths correspond to lower losses due to better confinement. If light only traveled through outside bends in these devices, then lower loss for shorter wavelengths might also be observed for the bending loss. The critical radius was calculated and although it has a slight wavelength dependence, it was less than 375 µm for all three wavelengths [6]. 62 Figure 4-6: Experimentally measured bending losses at 658, 980 and 1550 nm, along with loss calculated from the FDTD simulations. The simulations provided valuable insight into the operation of curved trapezoidal waveguides, and the results are shown in Figure 4-6 alongside the experimental data. The simulation data is also fit to an exponential curve. As predicted, light traveling through inside bends had more loss than that traveling through outside bends, and the loss from the inside bend is the limiting factor for determining a critical radius. The loss from the simulation results is systematically lower than the loss from the experiments, but the overall trend of the data is very similar. This discrepancy is likely due to the fact that the simulations did not include the silicon pillar, and corroborates the hypothesis that leakage of light into the membrane and subsequently into the silicon pillar accounts for much of the bending loss. These results further suggest that limiting the light signal to outside bends would reduce the bending loss and the critical radius. 63 4.2.5.3 Power Dependence Silica has a low non-linear coefficient from visible to near-IR wavelengths, which makes it a good candidate for high-power applications. Figure 4-7 shows the response of a straight trapezoidal waveguide device of input powers up to 200 mW, which was the maximum power that could be achieved with the system used for the experiment. The behavior is very linear, which is significantly different than most silicon and polymer waveguide devices, as those materials tend to have higher non-linear coefficients. This very linear response shows potential for the trapezoidal silica waveguides to be used for high-power applications. Figure 4-7: Output response of the trapezoidal silica waveguides when input with power up to 200 mW. 4.2.6 Conclusion and Future Outlook In summary, suspended, trapezoidal, silica-on-silicon waveguide devices were fabricated with both straight and curved geometries, and these devices were experimentally characterized. In addition, FEM and FDTD simulation models were developed to further study this waveguide device geometry. The simulations confirmed polarization independent behavior and good optical field confinement, as well as provided insight into the dominant loss mechanism for bending losses of these devices. The measured 64 propagation losses were 0.69, 0.59 and 0.41 dB/cm for 658, 980 and 1550 nm wavelengths, respectively, and this low loss is attributed to the low material loss of silica and the high optical field confinement due to a large core-cladding refractive index contrast. The critical bending radius was found to be less than 375 µm for all three wavelengths, and linear behavior was observed at input powers up to 200 mW. Future research can focus on improving the performance characteristics of these devices. For example, by lowering the surface roughness of the device, one can reduce further reduce the propagation loss. Also, by fabricating the device with a thinner membrane, the critical radius can be improved. Due to the simple, low cost fabrication and desirable characteristics such as linear behavior and low loss, these waveguides are suitable for many photonic applications, including as interconnects in photonic circuits. 4.3 Sol-gel coated waveguides Although the suspended silica-on-silicon waveguides have fairly low loss [5, 6], due to the material properties of silica, it is reasonable to expect that the loss of the devices could be even lower. In the case of trapezoidal devices, the transmission loss is limited by the surface roughness of the waveguides. Techniques have already been used to minimize the surface roughness, such as using a buffered oxide etchant for a slower and more uniform etch. Additional techniques might be able to minimize the surface roughness further, but at the cost of a more expensive and time-consuming fabrication process. The CO 2 laser-reflowed waveguides take advantage of the reflow step to greatly reduce the surface roughness of the waveguides. However, these devices have another issue that becomes the limiting factor for propagation loss—the reflow often makes the waveguide channels wavy, and this waviness is difficult to control [5]. Unfortunately, the waviness creates surface structures that act as very small microbends, and causes the optical power to be lost from the waveguide at a rate higher than it would be lost with a perfectly straight waveguide [9]. Because it is difficult to control how wavy the channels are simply through optimization of the reflow process, it is difficult to address this issue directly. Another potential disadvantage with the suspended waveguides is their size and core-cladding index contrast. While a high core-cladding index contrast can be beneficial to increase mode 65 confinement, in these devices it also causes the waveguides to operate in a multi-mode fashion and is an underlying contributor to the first issue of high loss. The multi-mode operation of the suspended waveguides is disadvantageous for a few reasons. First, higher order modes tend to have higher loss than lower order modes [10]. This occurs because higher order modes tend to have a higher power distribution towards the core-cladding interface, which leads to more interaction with the interface or surface of the waveguide. As a result, these modes are more susceptible to surface scattering for the trapezoidal devices and the waviness effects of the reflowed devices. Additionally, higher order modes are more susceptible to bending losses. In the ray-optics view of waveguides, higher order modes are those with a steeper reflection angle within the waveguide. Steeper angles are more likely to violate the critical angle condition once they encounter a bend, leading to higher bending loss. Secondly, it can be difficult to predict the behavior of multi-mode waveguides. Often it is possible to numerically or analytically solve Maxwell’s equations to determine several orders of modes for a given waveguide device. However, when coupling light into the device, it is very difficult to determine and control how much power will couple into each mode of the waveguide. This power varies based on the shape of the beam of light that is being coupled into the waveguide as well as its alignment to the waveguide. With a single-mode device, regardless of the alignment of the input optics or the shape of the input beam, only one mode can be excited, making the behavior of the waveguide much more predictable. Finally, multi-mode waveguides are more susceptible to dispersion [10]. Since different modes have different effective refractive indices, they will propagate through the waveguide at slightly different speeds, and optical pulses will spread out over time in a multi-mode device. Although this phenomenon occurs in single-mode waveguides as well, it is much more pronounced in multi-mode devices. 4.3.1 Strategies for enabling single-mode operation Making the suspended silica-on-silicon devices operate in a single-mode fashion would improve the devices by alleviating the aforementioned disadvantages, resulting in more predictable behavior and lower device loss. There are two ways to make the waveguides single-mode (Figure 4-8). The first 66 option is to shrink the dimensions of the device. However, in the case of the cylindrical or reflowed waveguides, this approach is problematic given the fabrication procedures currently used to make these waveguides. The silicon pillar acts as a heat-sink during the reflow step, preventing the silica membrane from melting and allowing it to support the waveguide channels. The size of the waveguide channels is related to the amount of silica to be reflowed on the outer edges of the device, as the total volume of silica will be conserved. Therefore to make smaller waveguide channels, the silica needs to have a smaller undercut portion, resulting in less exposed silica during the reflow process. However, in order to obtain a small enough waveguide channel diameter, the requisite amount of undercut would cause the overhanging silica to be too close to the silicon membrane, preventing it from reflowing properly. The second difficulty with shrinking device dimensions is maintaining the structural integrity of the silica membrane. Shrinking the waveguide channel dimensions would necessitate shrinking the membrane dimensions as well, and if the membrane is too thin it will not support the weight of the waveguide channels, causing the device to break. Figure 4-8: Two routes for single-mode behavior. Left: Shrinking device dimensions will result in structural defects. Right: A cladding layer may be added by spin-coating another material on the waveguide device. The second option for creating single-mode behavior is to reduce the refractive index contrast between the core and the cladding. This would allow us to keep our current fabrication procedures with one additional step of adding a coating to the device. This coating would replace air as the cladding layer, and for single-mode operation, it would need to have a refractive index similar to, but lower than the refractive index of the core. This option is more attractive than shrinking the device dimensions for a number of reasons. The first is that modifications to the current fabrication procedures are minimized. The second advantage is that the testing setup and procedures currently used for testing these devices will 67 not need to be modified. Shrinking the device dimensions would make coupling light into the waveguide more difficult because the input facet would be a smaller target. Additionally, the spot size of the lensed fibers used might be too large for a device with much smaller dimensions, which would increase the coupling loss significantly. The third advantage of using a coating is that reducing the core-cladding index contrast will make variations in the interface between these layers less prominent, reducing the effect of waviness and surface roughness, and consequently reducing propagation loss. There is one minor disadvantage to lowering the core-cladding index contrast—the critical radius for bending losses will be increased. Nevertheless, the advantages outweigh this one disadvantage, and for these reasons, pursuing a cladding layer is the best option to enable single-mode behavior in these waveguide devices. In order to generate a single-mode waveguide device, silica sol-gel thin films will be pursued as the low- index cladding layer. 4.3.2 Silica sol-gel thin films Fortunately, significant research has been done in the area of silica sol-gels[11-13]. Sol-gels are liquid-phase solutions with tunable properties that can be spin-coated onto surfaces and annealed to form a solid silica thin film. Silica sol-gels are fabricated using a liquid silica precursor, typically tetraethylorthosilicate (TEOS) or methyltriethoxysilane (MTES). Using an acid or base as a catalyst, the silica precursor undergoes hydrolysis and condensation reactions and begins to form a silica matrix (Figure 4-9). The optical properties of the sol-gel and resulting silica thin film may be tuned by adding dopants to the solution, using a different precursor, altering the synthesis procedure or altering the deposition and annealing parameters. 68 Figure 4-9: Hydrolysis and condensation reactions showing formation of a silica sol-gel matrix. In this work, sol-gels were prepared according to the following procedures: first, TEOS, ethanol solvent (EtOH), water and hydrochloric acid (HCl) are mixed in the ratios shown in Table 4-2. The ingredients were added in a specific order: TEOS to ethanol, then water, followed by HCl, with 5 minutes of stirring after each addition. After the HCl has been added, the hydrolysis reaction begins and the solution was stirred for an additional 2 hours to allow the hydrolysis reaction to complete. After this final stirring step, the sol-gel was aged 24 hours at room temperature, filtered through a 0.45 µm syringe filter, and stored in a refrigerator until the next step. After aging, the sol-gels are ready to be spin-coated onto either a bare silicon wafer (for characterization) or a waveguide device. First, the target sample was cleaned thoroughly and the room-temperature sol-gel was spin-coated onto the target sample at 7000 RPM for 30 seconds. The sol-gel was then dried by placing the sample on a 75° C hot plate for 5 minutes, which allows the solvent to evaporate. The final step was to anneal the samples in a tube furnace at 1000° C for 1 hour, with a ramp rate of 5° C/min, which densifies the silica thin film. 69 Table 4-2: Chemicals and ratios used for silica sol-gel synthesis. Chemicals TEOS EtOH HCl H 2 O Vendor Alfa Aesar, 99.999% EMD, 200 proof EMD Clean room Molecular Weight 208.33 46.07 36.46 18.02 Purity (mass) 100% 100% 36% 100% Molar ratio 1.00 4.000 0.10 2.00 Desired Chemical Mass 5.000 4.423 0.088 0.709 Desired Solution Mass 5.000 4.423 0.243 0.709 The properties of the resulting silica thin film may be altered by adding dopants when synthesizing the sol-gel solution or by altering any one of the synthesis parameters above. Because the desired result is a refractive index close to that of silica, investigations into changing the standard synthesis parameters were pursued. The parameters that were varied in search of a suitable formula are as follows: H 2 O amount, annealing temperature, annealing time, annealing ramp rate, aging time, and spin- coat spin speed. 4.3.3 Simulations and refractive index constraints As previously mentioned, the refractive index of the new cladding layer must be similar to, but slightly less than, the refractive index of the core. The exact value of this difference to ensure single- mode operation depends not only on the geometry of the waveguide device, but also on the wavelength of the light being used. Because of the difficulty in fabricating a waveguide device with a sol-gel coating that is effective across a wide range of wavelengths, this work optimized the device to operate at 1550 nm, a wavelength commonly used in optical communications. Even though the device may not have true single-mode operation at shorter, visible wavelengths, it will still reap some of the benefits of the cladding layer and will hopefully have lower loss at those wavelengths as well. In order to determine the exact refractive index needed for the cladding layer to enforce single- mode operation at a wavelength of 1550 nm, COMSOL simulations were performed. The device was simulated with a cladding layer of varying refractive indices. Using these simulations, a refractive index contrast threshold was determined, below which the device will have single-mode operation. An example 70 simulation result is given in Figure 4-10 compared with a simulation result with no cladding. From this simulation, you can see that the optical field has a slightly wider distribution in the device with a cladding layer, but does not significantly change shape. It was determined from the simulations that the refractive index difference between the core and the cladding must be less than 0.033 in order to obtain single-mode operation at 1550 nm. Figure 4-10: COMSOL simulation results of a) the fundamental mode in an air-clad waveguide and b) the fundamental mode in a single-mode, sol-gel cladding waveguide. Several experiments were performed to determine what sol-gel recipes might produce the required core-cladding contrast. Ellipsometry and Fourier Transform Infrared Spectroscopy (FTIR) were used to characterize the properties of the resulting films from each recipe. With ellipsometry, the thin film thickness and refractive index of the sol-gel were measured, and with FTIR, characteristic absorption spectra were measured, which are helpful in detecting unwanted solvent residue in the film. Using these techniques, it was determined that sol-gels synthesized with 2 times the amount of water given in Table 4-2 yield a refractive index close to that desired for single-mode operation. 4.3.4 Data and results Unfortunately, while an appropriate sol-gel recipe was identified, there were issues which ultimately prevented a sol-gel thin film approach from working. The main problem occurred when coating the sol-gel on waveguide devices. The coatings did not evenly coat the waveguides, resulting in cracking of the thin film during annealing (Figure 4-11). Although an uneven coating alone would not be (a) (b) 71 problematic, providing that the coating is thick enough to use in its thinnest part, the cracking gave rise to additional scattering losses, resulting in devices with higher loss, instead of lowering the loss. Several experiments were performed in an attempt to alleviate the cracking. The waveguides were treated with an O 2 plasma cleaning step before spin coating. Different spin speeds and annealing temperatures were tested, as well as the use of trapezoidal waveguides. However, while some of these procedures seemed to mildly alleviate the cracking, none of them diminished the cracking enough to eliminate the extra surface scattering. Figure 4-11: SEM images of a sol-gel coated waveguide taken from a) the input end and b) above. The sol-gel thickness is clearly not uniform and has several cracks, including around the waveguide channel, where it might affect the mode. Additionally, because the thermal silica and silica sol-gel have such similar conductivity values, they appear as the same color in these images. Despite the cracking, loss measurements were performed using trapezoidal waveguides to evaluate the performance of coated devices (Figure 4-12). As can be seen in the figure, the loss values do decrease slightly with decreasing length when fit to a linear curve. However, the losses are higher overall than they were for uncoated devices. Additionally, the spread in measured losses for each length is much larger than for the uncoated devices. This is likely due to the fact that cracking varied between devices depending on how many cracks were formed during cleaving. Additionally, when the cut-back method was used to measure the loss of these waveguides, non-realistic (unphysical) results were obtained (loss stayed approximately constant with very wide distribution in values). This behavior likely resulted from the stress on the device every time the waveguide was cleaved to a shorter length, and the higher number of cracks introduced into the thin film coating due to that process. These cracks increased a) b) 72 the scattering loss, resulting in a wide range of measured loss values for each length. The result is the trend seen in Figure 4-12—the loss is overall very high, and even though the linear fits indicate low loss values, they are suspect due to the large spread in measured loss values and the high error. Figure 4-12: Measured losses from sol-gel coated waveguides measured at three different wavelengths. The data was fit to linear curves from which the loss can be measured. Given the large spread in measured loss values, the linear fits are likely not very accurate. 4.3.5 Conclusions and future work Although lowering the index contrast of the silica suspended waveguides remains an attractive option for creating single-mode operation and reducing the propagation loss, the use of silica sol-gels for this purpose is not viable. The thin film coats the devices unevenly due to their unique geometry, which results in cracks during the annealing process. Despite several attempts to reduce the cracking, it remained and caused additional propagation losses. Another material with a refractive index lower than silica might serve as an appropriate cladding layer. For example, there are several low-index polymers [14, 15] which could be spin-coated onto the 73 waveguide devices and are not in danger of cracking. As long as the index is suitably low and the coating is suitably thick, there is a good potential for lowering the number of supported modes in the device and therefore lowering the propagation losses in the silica suspended waveguides. Chapter 4 References [1] C. Kopp, S. Bernabe, B. Ben Bakir, J. M. Fedeli, R. Orobtchouk, F. Schrank, et al., "Silicon Photonic Circuits: On-CMOS Integration, Fiber Optical Coupling, and Packaging," IEEE Journal of Selected Topics in Quantum Electronics, vol. 17, pp. 498-509, May-Jun 2011. [2] A. Himeno, K. Kato, and T. Miya, "Silica-based planar lightwave circuits," IEEE Journal of Selected Topics in Quantum Electronics, vol. 4, pp. 913-924, Nov-Dec 1998. [3] J. F. Bauters, M. J. R. Heck, D. John, D. X. Dai, M. C. Tien, J. S. Barton, et al., "Ultra-low-loss high-aspect-ratio Si3N4 waveguides," Optics Express, vol. 19, pp. 3163-3174, Feb 2011. [4] J. F. Bauters, M. J. R. Heck, D. D. John, J. S. Barton, C. M. Bruinink, A. Leinse, et al., "Planar waveguides with less than 0.1 dB/m propagation loss fabricated with wafer bonding," Optics Express, vol. 19, pp. 24090-24101, Nov 21 2011. [5] A. J. Maker and A. M. Armani, "Low-loss silica-on-silicon waveguides," Optics Letters, vol. 36, pp. 3729-3731, Oct 1 2011. [6] X. Zhang, M. Harrison, A. Harker, and A. M. Armani, "Serpentine low loss trapezoidal silica waveguides on silicon," Optics Express, vol. 20, pp. 22298-22307, Sep 24 2012. [7] I. H. Malitson, "INTERSPECIMEN COMPARISON OF REFRACTIVE INDEX OF FUSED SILICA," Journal of the Optical Society of America, vol. 55, pp. 1205-&, 1965. [8] P. K. Tien, "LIGHT WAVES IN THIN FILMS AND INTEGRATED OPTICS," Applied Optics, vol. 10, pp. 2395-&, 1971. [9] C. R. Pollock, Fundamentals of Optoelectronics. Chicago: Irwin, 1995. [10] D. Derickson, Fiber Optic Test and Measurement: Prentice Hall, 1998. [11] H. S. Hsu, C. Cai, and A. M. Armani, "Ultra-low-threshold Er:Yb sol-gel microlaser on silicon," Optics Express, vol. 17, pp. 23265-23271, Dec 7 2009. [12] B. A. Rose, A. J. Maker, and A. M. Armani, "Characterization of thermo-optic coefficient and material loss of high refractive index silica sol-gel films in the visible and near-IR," Optical Materials Express, vol. 2, pp. 671-681, May 1 2012. [13] L. C. Klein, "SOL-GEL OPTICAL-MATERIALS," Annual Review of Materials Science, vol. 23, pp. 437-452, 1993. [14] J. Liang, E. Toussaere, R. Hierle, R. Levenson, J. Zyss, A. V. Ochs, et al., "Low loss, low refractive index fluorinated self-crosslinking polymer waveguides for optical applications," Optical Materials, vol. 9, pp. 230-235, 1998. [15] A. Sulaiman, S. W. Harun, K. S. Lim, F. Ahmad, and H. Ahmad, "Microfiber Mach-Zehnder interferometer embedded in low index polymer," Optics and Laser Technology, vol. 44, pp. 1186- 1189, 2012. 74 Chapter 5: Waveguide Splitter Biosensors 5.1 Introduction Optical devices have proven themselves to be robust and sensitive sensors in a wide variety of applications [1-4]. They exist in many form factors (Figure 5-1) and many of them have distinct advantages over their electrical counterparts [5, 6]. For example, optical devices tend to be more robust against environmental factors, such as pH, than electrical sensors [7, 8]. These advantages make optical devices particularly useful in the realm of biosensing, where specificity for a target molecule of interest is often imparted to devices using surface functionalization techniques. Additionally, many researchers have begun to investigate how to take these tiny devices and integrate all the necessary components for their operation on to the same chip, making them lab-on-a-chip (LOC) devices [9, 10]. In this chapter, the viability of suspended waveguide splitters as biosensors is explored via simulations, experiments, and integration with microfluidic channels. Figure 5-1: Examples of several photonic sensor geometries: a) Mach-Zender interferometric waveguide sensor, b) resonant optical microcavity sensor, and c) photonic crystal waveguide sensor. 5.1.1 Sensing Mechanism The waveguide splitters described in Chapter 2 operate as biosensors by detecting changes in the effective refractive index. When target molecules, or analytes, bind to the surface of the splitter, the a) c) b) 75 evanescent field of the light in the waveguide will interact with these molecules. This will cause the guided light to experience a change in the effective refractive index, which will result in a shift of the splitting ratio (Figure 5-2). Additionally, a shift in the background refractive index, for example a change in the refractive index of the water solution containing the analyte of interest, will also produce a shift in the splitting ratio of the output. Figure 5-2: 3D images of the waveguide splitter output as captured by the beam profiler. Each of the two peaks is the output from one of the splitter waveguide arms. The two images show the output of the waveguide splitter a) before and b) after an analyte is introduced to the surface of the device, causing the splitting ratio to change. The waveguide splitter demonstrates unique sensing properties that make it stand out amongst other optical waveguide-based biosensors, such as fluorescent waveguide sensors or Mach-Zender interferometer sensors. First is the fact that the device has two outputs that change in a complimentary fashion with respect to each other. Specifically, if the signal from one output increases, the signal from the other output will simultaneously decrease. This differential signal is beneficial because it inherently causes a two-fold increase to the sensitivity of the device. Secondly, the waveguide splitters are completely suspended in the coupling region, which is the region where sensing takes place. Typically, waveguide sensor devices are embedded into the substrate or lie directly on top of it, limiting the area available for sensing to one side of the device. The advantage of having the sensing region of the waveguide sensor device fully suspended is that there is more surface area available for sensing to occur. 76 This results in higher collection efficiency, improving the limit of detection and decreasing the response time. Finally, compared to fluorescent waveguide sensors, the waveguide splitter would operate as a label-free biosensor, reducing the complexity of the detection assay and equipment required to perform bio-detection experiments. 5.1.2 Microfluidics Microfluidics is an area of research involving channels for fluid flow that have dimensions on the micrometer scale. Microfluidic channels are used in many areas of research in order to manipulate small quantities of liquids or samples [11, 12]. Additionally, researchers have begun integrating microfluidics with optical sensors, creating useful optofluidic devices, which perform better than their non-integrated counterparts [4, 5, 13]. They are frequently made from polydimethylsiloxane (PDMS) which is a flexible elastomer that adheres easily to several substrates. Microfluidics is attractive for optical sensors because it can further reduce the amount of sample solution necessary to perform detection measurements, and deliver the solution directly to the sensing area of the sensor being used. 5.1.3 Lab-on-a-chip The waveguide splitter devices are also attractive as biosensors because of their potential as LOC devices. The testing setup used to characterize the devices consists of parts that have the potential to be miniaturized and packaged on the same chip. Researchers have successfully integrated laser sources and photodiodes onto devices with waveguides, the two components needed to replace the lensed fiber and beam profiler of the testing set up to have a fully integrated LOC device. Additionally, several advantages are gained by making an optofluidic device by integrating the waveguide splitter with microfluidic channels. The device would be more robust against damage and have improved efficiency by delivering the sample solution directly to the active sensing area of the device. Therefore, if the viability of the waveguide splitter devices as biosensors can be demonstrated, there is a clear path forward to making these sensors into compact and complete LOC devices. 77 5.2 Computer modeling and simulations 5.2.1 Models and methods Using a model developed in Lumerical FDTD of the waveguide splitter device as a starting point [14], three different models for how the splitter operates as a biosensor were developed. These models were used to investigate the behavior of the biosensor devices and predict the output when used in sensing experiments. In all three simulation models, light with a wavelength range from 1520 nm to 1630 nm is used, as this range matches that of the tunable laser used for sensing experiments. A schematic of the basic simulation is shown in Figure 5-3. In the simulations performed here, all the parameters shown were kept constant, with R 1 = R 2 ≈ 200 μm, R 3 = R 4 = 250 μm, L = 160 μm, L’ = 155 μm, ϕ = 10°, W = 22 μm, and L p ≈ L’/cos(ϕ). Building a simulation model is useful when predicting how the device will react to different analytes or how it will react to an analyte in a complex biological solution. Figure 5-3: Schematic of model used for simulation with important parameters marked on the drawing. The light and dark colors represent thin and thick parts of the membrane, respectively (T 1 = 900 nm and T 2 = 2μm). The black lines are the waveguide channels and the red line indicates the direction of insertion of the optical mode. The first computational model developed is a bulk refractive index sensing model. The model is the same as in Figure 5-3, with one important change. In this set of simulations, the background refractive index is increased from 1 to 1.3 in increments of 0.05. These simulations investigate how the waveguide splitters react to a change in the background refractive index and provide valuable insight to how the splitting ratio of the device could change in response to random environmental fluctuations or in different media. The information will be valuable for biosensing applications since the detection will be 78 performed in an aqueous environment, but the exact refractive index may change depending on whether the environment is water, a biological buffer, or a complex biological fluid. The second simulation model developed was one of an adlayer forming on the waveguide splitter surface. When analytes bind to the surface of the waveguide biosensor, they will slowly form an adlayer until, theoretically, the entire functionalized area is covered in a uniform monolayer. This adlayer will form one protein molecule at a time in a random distribution across the functionalized area. Unfortunately, it is not possible to simulate this exact mechanism in Lumerical. Instead, adlayer formation is first simulated by having a solid adlayer slowly increase in thickness from 0 to 50 nm, with a refractive index similar to that of the CREB protein. This is achieved by starting with the base model in Figure 5-3 and adding a layer with a refractive index equal to 1.49, and increasing the thickness of that solid layer from 0 to 50 nm in 10 nm increments. Although this adlayer formation is different, it should affect the effective refractive index of the light signal traveling down the splitter device in a similar way to the actual adlayer formation would and provide useful results that will allow us to predict device behavior at saturation. In the third model an adlayer forming on the surface of the waveguide splitter was simulated using a different mechanism. In this set of simulations, an adlayer of constant thickness 50 μm was modeled with increasing refractive indices. The refractive index of the adlayer was increased from 1 to 1.45 in increments of 0.05. This model should also affect the effective refractive index of the light signal travelling through the splitter similarly to actual adlayer formation. Simulating a different adlayer formation mechanism should give additional insight into how the adlayer actually develops on the surface of the biosensor, as well as the actual operation of the sensor. 5.2.2 Data and results In order to take advantage of the differential output of the sensor, the percent of total power coming from one arm is subtracted from the other. This subtraction results in a unit of change referred to in this chapter as ΔPower. This is a unit-less number which is simply a measure of the change in power 79 between both arms. This unit of measurement can range from -1 to 1, with a value of 0 corresponding to a 50:50 splitting ratio. Results from the background index simulations are given in Figure 5-4. Aside from the shift from a background index of 1 to 1.05, there is a trend of decreasing ΔPower with increasing refractive index. Additionally, ΔPower stays fairly constant over the entire wavelength range for every background index value. This means that the splitting ratio stays relatively constant, just as in normal operation. Both of these factors are very positive, as they indicate that the behavior of the sensor is stable and predictable across a fairly wide range of background indices. Figure 5-4: Results from the simulations in which the background refractive index was modified. For increasing refractive index, ΔPower decreases, and the splitting ratio stays fairly constant across a wide wavelength range. The results from the second set of simulations in which an adlayer thickness was increased are given in Figure 5-5. In this set of simulations, ΔPower does not change with any constant trend. At shorter wavelengths, it generally increases, but at the higher wavelengths it decreases before increasing. 80 In addition, there is a lot more variation of ΔPower across the wavelength range simulated. While this is undesirable, since it eliminates broadband operation, it does not eliminate the splitter as a candidate for biosensing experiments. This result means that in practice, during sensing experiments, the wavelength should be set to a specific value and does not change during the course of the experiment. The results from these simulations indicate what sort of behavior can be expected from the splitter when operating at different wavelengths. Figure 5-5: Results from the simulations in which an adlayer with increasing thickness was added to the sensor. For increasing thickness, the change in ΔPower varies at different wavelengths. Additionally, ΔPower does not stay constant across the entire wavelength range as seen when operating without an adlayer. The results from the third set of simulations are given in Figure 5-6. In this set of simulations, ΔPower has a much different reaction than in the previous set. However, ΔPower still changes in a different way across different wavelengths. For each value of adlayer index, ΔPower does tend to decrease with increasing wavelength. Between this model and the second model the splitter operation can 81 begin to be understood. Although the ΔPower change is different for increasing adlayer at different wavelengths in both simulation sets, patterns can emerge when the sensor is locked to a single wavelength. Both the second and third sets of simulations indicate the response of the sensor will not be very predictable, but will depend greatly on the exact mechanism of adlayer formation. Figure 5-6: Results from the simulations in which an adlayer with increasing refractive index was added to the sensor. For increasing index, the change in ΔPower varies at different wavelengths. Additionally, ΔPower does not stay constant across the entire wavelength range as seen when operating without an adlayer. 5.3 Surface functionalization 5.3.1 Experimental methods A sensor is not very useful if it detects molecules indiscriminately. Therefore, specificity is of great concern when designing optical biosensors. As a proof-of-concept, the waveguide splitter biosensors were functionalized to be specific to the cAMP response element-binding (CREB) protein. CREB has been shown to regulate genes involved in metabolic regulation, long-term memory, 82 depression, and tumor suppression [15]. Additionally, it has been a target protein to treat memory loss, post-traumatic stress disorder and head trauma [15]. These factors make CREB an interesting and worthwhile protein to detect with a novel optical biosensor device. In order to make optical biosensors specific to their target analyte, a surface-chemistry functionalization was used. Using the steps outlined in Figure 5-7, anti-CREB antibodies were attached to the surface of the waveguide device. The process for surface functionalization begins with treating the samples with an oxygen plasma in an O 2 plasma asher. The samples are exposed for 5 minutes at 120 W of power and 30 sccm flow rate. This treatment hydroxylates the surface, attaching OH groups to the device. Next, (3-Glycidoxypropyl)methyldiethoxysilane (GPTMS) is attached to the OH groups using chemical vapor deposition. The device is put in a desiccator with a small amount of GPTMS and the desiccator is brought under vacuum for 30 or 40 minutes using a pump. After that, 5 μL drops of diluted antibody solutions are placed on to the device. The device is incubated with the antibody in a humid environment of roughly 37º C for at least 3 hours. Finally, the samples are rinsed of excess, unattached antibodies and buffer residue by immersing them in DI water and setting them on a tilt tray for 5 minutes, then allowing them to dry. Figure 5-7: Surface functionalization procedure. The steps are as follows: a) the device is oxygen plasma treated to hydroxylate the silica surface, b) a pressure vapor deposition step is performed to attach GPTMS to the OH groups, and c) the sample is incubated with an anti-CREB solution in a warm, humid environment to attach the antibodies to the surface of the device. 83 The anti-CREB antibodies used for functionalization are complimentary to the CREB protein, which means that the antibodies will bind specifically to CREB, but will not bind with a strong affinity to other proteins. This kind of surface functionalization is fairly common for optical biosensors and imparts specificity upon the device, allowing it to be used to detect the protein of interest in a complex biological solution. A unique feature of the fabrication process for the waveguide splitters is that, due to the XeF 2 etching step, the surface functionalization is selective to the silica on the devices, preventing the silicon substrate from being functionalized and increasing efficiency by limiting the protein in solution to binding to active areas of the device [16]. The functionalization process does cause some loss in the device, but not enough to degrade the signal beyond practical use. Additionally, as the first step in the process is oxygen plasma treatment, by cleaning the samples after use and repeating this step, the waveguide splitter biosensors may be re-functionalized and re-used [17]. These unique features of the functionalization process enhance the versatility and usefulness of the biosensor devices. 5.3.2 Results In order to verify the functionalization procedure, imaging with fluorescent labels was used. Waveguide splitters were functionalized following the procedures outlined above and subsequently incubated with CREB followed by a secondary antibody that was fluorescently labeled. These incubation steps were performed in the same way that the anti-CREB incubation step was performed. After these steps, the device was viewed under a fluorescent microscope. Anywhere that the anti-CREB antibodies have attached to the surface of the device should have a fluorescent signal, while non-functionalized surfaces should be dark. The result of this verification step is shown in Figure 5-8. In this pair of bright field and fluorescent images, it is clear that the device is intact (part a) and that the device surface is fluorescent (part b), indicating that the surface functionalization procedures work and produce an evenly functionalized surface, with minimal clumping. 84 Figure 5-8: Optical micrographs of the splitting region of a functionalized waveguide splitter a) under a bright field light and b) under fluorescent excitation and a filter. 5.4 Sensing experiments 5.4.1 Experimental methods Sensing experiments were performed using the waveguide testing setup described in Chapter 2. All sensing experiments were performed with the tunable 1520-1630 nm laser. Before an attempt was made to integrate the waveguide splitters with microfluidics for sample delivery, the following procedures were used to perform sensing experiments. A micro-pipette was used to place droplets of solution of a specific volume on the surface of the device during each experiment. The sensing experiments began with placing a 1 μL droplet of phosphate buffered saline (PBS, or buffer) on to the device and adjusting the wavelength on the tunable laser so that the splitting ratio was close to 50/50. This preparatory step brings the background index of the device close to that of the sample solution and reduces any change that might result from a background refractive index change while allowing the splitting ratio to be set at the point where the device will be most sensitive. At that point, data collection began and 0.5 μL droplets of increasing concentrations of CREB in PBS (1 fM, 1 pM, 1 nM) were deposited on the device, with time in-between each droplet to allow the device to reach steady-state. About 3 minutes were left in-between each droplet of CREB solution and the total length of these experiments was 10 minutes. 85 Figure 5-9: Cross-sectional profile of the output of the waveguide splitter as measured by the beam profiler. Integrating the area under each peak is used to calculate the splitting ratio. The data being saved was cross-sectional profiles of the splitter output as captured by the beam profiler (Figure 5-9). These intensity vs. position profiles were saved at a rate of 4 Hz, which is the fastest the computer system could acquire data without dropping any data points. The data was analyzed using a custom Matlab script, which automatically calculated the percent of power coming out of each arm of the device by integrating each intensity peak to obtain power and dividing the result by the sum of power from both peaks. The ΔPower quantity mentioned above was calculated in the same way, the percent of power from one arm is subtracted from the other. Once again, this is a unit-less number which measures the change in power between both arms and takes advantage of the differential output. In addition to the sensing experiments described above, competitive binding experiments were performed in order to assess the specificity of the sensor with the functionalization process used and the 86 ability of the device to perform detection in a complex solution. The procedures for these experiments are nearly identical to those described above. However, after the 1 μL droplet of buffer was placed on the device, a 0.5 μL droplet of 1 nM bovine serum albumin (BSA, a non-specific binder) was placed on the device followed by a 0.5 μL droplet of 1 nM CREB solution. As before, the timing of the droplets was spaced out by 3 minutes and the total length of the experiments was 10 minutes. In these experiments, the BSA should bind weakly to the surface of the device, resulting in a signal change from the sensor. However, the CREB binds more strongly. Therefore, once it is introduced into the system, it can displace the BSA and cause an increased response from the sensor. The data collected from these experiments was analyzed in the same way as before with the same Matlab script. Finally, experiments were performed in which solutions of different refractive indices were placed on the sample to assess the sensitivity of the devices to fluctuations in the background index. For these experiments glycerol solutions in DI water of varying concentration (0.1% wt to 0.5% wt) were used. Glycerol solutions were chosen because their refractive index changes with different concentration and is well-characterized. The refractive index of each glycerol solution was calculated and the results are in Table 5-1. A 1.5 μL drop of glycerol solution was placed on the surface of the device, and a wavelength sweep was performed from 1520 nm to 1630 nm, measuring the splitting ratio as before. After each wavelength sweep for a particular concentration of glycerol was performed, the device was removed from the testing setup and cleaned with solvents and an oxygen plasma treatment. The splitting ratio and ΔPower were calculated using the Matlab script as before. Table 5-1: Refractive Indices of Glycerol Solutions % wt 0.1 0.2 0.3 0.4 0.5 Refractive Index 1.33312 1.33324 1.33336 1.33348 1.3336 One major advantage to these methods is that data can be collected in real-time and analyzed quickly. The splitting ratio change can be viewed as it happens on the beam profiler output. However, while the experiment is ongoing, this change can only be viewed qualitatively and cannot be quantified. As soon as the experiment has ended, the data can be analyzed by the Matlab script and quantitative 87 results can be obtained. The script runs very quickly, completing in less than a minute. Therefore, very shortly after the experiments have been performed, quantitative results may be obtained, with no additional effort needed for data analysis. 5.4.2 Data and results Several sets of sensing experiments were performed. An example of results from an experiment with increasing concentrations of CREB is shown in Figure 5-10. The figure shows ΔPower vs. time, and the places where new droplets of CREB solution were added to the surface of the device are indicated on the graph. After each droplet of solution is added, ΔPower shifts rapidly then slowly settles as the sensor reaches steady-state. This settling occurs because the CREB solution reaches the device by diffusion. Additionally, the 1 nM concentration droplet shows the strongest response. These results indicate that the splitter biosensor is sensitive and can distinguish between concentrations of CREB solution. The 1 fM droplet likely shows a stronger response than the 1 pM droplet because no CREB was present in the solution before the 1 fM solution was added, causing an overall stronger response even though the concentration was lower. 88 Figure 5-10: Sensing data showing the results of an increasing concentrations experiment. Approximately 30 seconds after adding PBS and starting data acquisition, a 1 fM droplet of CREB is added. Approximately 3 minutes later, a 1 pM droplet of CREB is added followed by a 1 nM droplet 3 minutes after that. After each subsequent droplet is added, the sensor responds with a shift in ΔPower quickly, and slowly settles as the sensor reaches steady-state. Example results from a competitive binding experiment are given in Figure 5-11. This figure also shows ΔPower vs. time and indicates where droplets of 1 nM BSA and 1 nM CREB were added to the droplet already on the device. After the BSA is added, there is a very strong response, which is to be expected as it was added in a high concentration. However, after the CREB is added, there is a response beyond the response of the BSA. This indicates that the CREB was more favorable than the BSA in steady-state. This proves that the device can be used to detect CREB specifically due to the functionalization procedures, even in the presence of a non-specific binder like BSA. There is some noise after the CREB was added due to a disturbance in the lensed fiber coupling. 89 Figure 5-11: Sensing data showing the results of a competitive binding experiment. Approximately 30 seconds after adding PBS and starting data acquisition, a 1 nM droplet of BSA is added. Approximately 3.5 minutes later, a 1 nM droplet of CREB is added. After each droplet is added, the sensor responds with a shift in ΔPower quickly, and slowly settles as the sensor reaches steady-state. After the CREB is added the sensor has a stronger response than when the BSA was added. Results from the glycerol experiment are shown in Figure 5-12. The refractive indices for each of the glycerol concentrations are given in Table 5-1. The index range of the glycerol concentrations is much smaller than that of the first simulation set in which the background index was varied. Despite this, the ΔPower variation is much greater in the experiments. Additionally, there is a lot of variation over different wavelengths. This indicates that the sensors are likely more sensitive in practice than the model accounts for. This sensitivity is probably due to variations in fabrication; particularly in the CO 2 laser reflow process which can cause waviness in the waveguide channels. These random variations are very difficult to account for in a simulation model, so experiments like those shown in Figure 5-12 are important to understand the actual behavior and sensitivity of the waveguide splitter sensors. 90 Figure 5-12: Results from glycerol sweep experiments are shown. For different concentrations of glycerol, ΔPower has large variations. In addition, ΔPower has large variations across all wavelengths. Unfortunately, the positive results shown in Figure 5-10 and Figure 5-11 were never repeated. Several experiments were performed with the same procedures over the course of many months, but the shifts in ΔPower were always different. One reason this happened could be due to the previously mentioned variations in fabrication between the different waveguide samples. The waviness in waveguide channels created during the CO 2 laser reflow can have an effect on the splitting ratio of the splitters. This effect seems to be amplified when the splitters are placed in an aqueous environment, which could cause difficulties with using the splitters as biosensors. In addition, during the sensing experiments, the droplets placed on the surface of the device often did not wet the device evenly. In some cases, the water droplet would expand across the surface of the device and result in a slow drift in ΔPower throughout the experiment. This uneven wetting could also be a source of error and difference in the 91 experiments. Uneven wetting of the sample surface was one of the key motivations for integrating the devices with microfluidics. 5.5 Microfluidics fabrication and integration 5.5.1 Experimental methods To improve the reproducibility of the results and accelerate the experiments, microfluidic channels were designed to deliver sample solutions directly to the sensing area of the sensor devices. The channels were designed to have the dimensions as shown in Figure 5-13. The design consists of two channels in a cross pattern. The first channel is designed as a space to go over the splitter and is 40 um by 100 um wide. The second channel is designed to carry the sample solution to the sensing area of the device and is also 40 um by 100 um wide. The channels are designed such that solution traveling down the main channel will not be able to flow into the channel containing the waveguide splitter due to a high pressure differential. This would help keep the input and output channels of the device clean and direct the sample solution only to the area where sensing occurs. Figure 5-13: Diagrams showing a top and side view of the microfluidic channels. The channels were 100 μm by 40 μm. There are two channels for the two devices per sample and one channel for the solution being tested to flow across the sensing or coupling region of the device. 92 The microfluidic channels were created with PDMS using a soft-lithography process (Figure 5-14). First, a transparency mask with the channel pattern was designed and printed. Second, the pattern from the mask was transferred to a wafer that was coated with SU-8 photoresist using UV photolithography. The photoresist was deposited onto the wafer using a spin-coating technique, with the rotation speed set to achieve a thickness of 40 um. After the wafer was developed, a master mold of SU-8 remained in the pattern of the microfluidic channels. The master mold was placed in a petri dish and covered with PDMS. After the PDMS was cured, it was removed from the master mold and cut into appropriate sizes using a razor blade. Figure 5-14: Schematic diagram of soft lithography process. A bare wafer is coated with SU-8 photoresist dvia a spin- coating procedure. UV lithography is used to transfer a pattern from a mask to the photoresist. After developing, a master mold is created. The mold is covered in PDMS and cured. Once the PDMS is removed from the master, microfluidic channels are left in the PDMS and it is ready to be attached to a device. A number of methods were tried to attach the microfluidic channels to the devices, including oxygen plasma bonding, glue, and uncured PDMS. The glue method was very straightforward: glue was applied to the silicon surface and the PDMS channels were placed on top of the glue. The wafer and PDMS were clamped together until the glue was allowed to dry. The uncured PDMS method was similar. A bit of uncured PDMS was used in the place of glue, and after the microfluidic channels were attached, the device was placed in an oven to cure. For oxygen plasma bonding, the PDMS channels are placed in an O 2 plasma asher face-up along with the sample. After a short exposure of 30 seconds at 50 W and 30 sccm, the sample and wafer are attached, clamped together, and allowed to sit overnight. The O 2 plasma treatment should activate the surface of the wafer and the PDMS by bombarding it with free radicals, 93 which will allow a chemical bond to form between the two substances. This is a common procedure performed with silica and PDMS, but it is not commonly performed with silicon [18, 19]. The microfluidic channels were operated using syringe pumps. Before attaching the channels to the splitter sensors, inlet and outlet channels were punched in the PDMS using luer stubs. Once the PDMS is attached to the splitter, a metal tip can be inserted into the inlet and outlet channels. Plastic tubing was attached to the metal tips, which was in turn attached to a syringe placed in the syringe pump. Using the syringe pumps allowed for controlled flow rates to be used when performing experiments and the tubing allowed the sample to be placed on the testing setup without any microfluidic components getting in the way of the optical components. 5.5.2 Results There were many problems when it came to integrating the microfluidic channels with the waveguide splitters. The first main problem was that none of the microfluidic integration methods attempted worked with a high degree of repeatability. For example, the most promising method required bonding the PDMS to the silicon. However, a tight enough seal between the PDMS and silicon wafer was never achieved. Even when flowing at low flow rates, microfluidic channels are subject to very high pressure due to their small size. In the integrated devices that were made, this pressure was too great for any of the bonding methods to sustain. The microfluidic channels consistently leaked, and virtually no devices could be tested. A second large problem was that it was very difficult to align the channels with the devices. Since the PDMS channels were held with tweezers and attached by hand, precision was very important. A stereoscope was used to assist with the alignment, but the size of channels and waveguides was very small, with a low margin for error. Frequently the spaces for the waveguide devices would be aligned, but the channel for the solution would not be properly aligned, causing the solution to run into the silicon pillar of the device instead of the suspended sensing region. This also contributed to the first problem, by causing higher pressure due to the outflow channel being blocked by the splitter. Because of the 94 difficulty in properly attaching the PDMS to the silicon substrate, microfluidic integration for the waveguide splitter sensors was abandoned. 5.6 Conclusions and future work In this chapter, the viability of using waveguide splitter biosensors was explored. A theoretical model for operation was developed using FDTD simulations. These simulations indicated complicated operational parameters that would not easily be reduced to an equation describing the sensing response of the device. A surface functionalization procedure was developed for use with the CREB protein, verified, and sensing experiments were performed with functionalized devices. Unfortunately, although individual experiments yielded positive results, due to variations in fabrication and other factors, no results were able to be repeated. Finally, microfluidic channels were developed for controlled sample delivery and to alleviate some of the problems from the sensing experiments. Despite a number of different attempted procedures, it was not possible to adequately attach the PDMS microfluidic channels to the waveguide splitter devices. The high pressure in the channels consistently caused the PDMS to detach from the wafer surface. Due to the issues faced during the course of the investigations, it was determined that the suspended waveguide splitters are not a good candidate for biosensing applications. Chapter 5 References [1] C. Dongre, R. Dekker, H. Hoekstra, M. Pollnau, R. Martinez-Vazcluez, R. Osellame, et al., "Fluorescence monitoring of microchip capillary electrophoresis separation with monolithically integrated waveguides," Optics Letters, vol. 33, pp. 2503-2505, Nov 2008. [2] A. M. Armani, "Label-free, single-molecule detection with optical microcavities (August, pg 783, 2007)," Science, vol. 334, pp. 1496-1496, Dec 2011. [3] A. Crespi, Y. Gu, B. Ngamsom, H. J. W. M. Hoekstra, C. Dongre, M. Pollnau, et al., "Three- dimensional Mach-Zehnder interferometer in a microfluidic chip for spatially-resolved label-free detection," Lab on a Chip, vol. 10, pp. 1167-1173, 2010 2010. [4] D. Duval, A. Belen Gonzalez-Guerrero, S. Dante, J. Osmond, R. Monge, L. J. Fernandez, et al., "Nanophotonic lab-on-a-chip platforms including novel bimodal interferometers, microfluidics and grating couplers," Lab on a Chip, vol. 12, 2012 2012. [5] D. Erickson, S. Mandal, A. H. J. Yang, and B. Cordovez, "Nanobiosensors: optofluidic, electrical and mechanical approaches to biomolecular detection at the nanoscale," Microfluidics and Nanofluidics, vol. 4, Jan 2008. [6] J. Wang, "Electrochemical biosensors: Towards point-of-care cancer diagnostics," Biosensors & Bioelectronics, vol. 21, pp. 1887-1892, Apr 2006. [7] A. H. Liu, T. Watanabe, I. Honma, J. Wang, and H. S. Zhou, "Effect of solution pH and ionic strength on the stability of poly(acrylic acid)-encapsulated multiwalled carbon nanotubes aqueous 95 dispersion and its application for NADH sensor," Biosensors & Bioelectronics, vol. 22, pp. 694- 699, Dec 2006. [8] N. Lloret, R. S. Frederiksen, T. C. Moller, N. I. Rieben, S. Upadhyay, L. De Vico, et al., "Effects of buffer composition and dilution on nanowire field-effect biosensors," Nanotechnology, vol. 24, p. 9, Jan 2013. [9] N. Bamiedakis, T. Hutter, R. V. Penty, I. H. White, and S. R. Elliott, "PCB-Integrated Optical Waveguide Sensors: An Ammonia Gas Sensor," Journal of Lightwave Technology, vol. 31, pp. 1628-1635, May 2013. [10] A. L. Washburn and R. C. Bailey, "Photonics-on-a-chip: recent advances in integrated waveguides as enabling detection elements for real-world, lab-on-a-chip biosensing applications," Analyst, vol. 136, pp. 227-236, 2011. [11] J. Y. Chen, Y. F. Liu, T. Ye, X. Xiang, X. H. Ji, and Z. K. He, "A novel droplet dosing strategy- based versatile microscale biosensor for detection of DNA, protein and ion," Sensors and Actuators B-Chemical, vol. 215, pp. 206-214, Aug 2015. [12] J. C. Chou, J. L. Chen, Y. H. Liao, J. T. Chen, C. Y. Lin, J. W. Lin, et al., "Fabrication and Characteristic Analysis of a Remote Real-Time Monitoring Applied to Glucose Sensor System Based on Microfluidic Framework," IEEE Sensors Journal, vol. 15, pp. 3234-3240, Jun 2015. [13] F. Ceyssens, D. Witters, T. Van Grimbergen, K. Knez, J. Lammertyn, and R. Puers, "Integrating optical waveguides in electrowetting-on-dielectric digital microfluidic chips," Sensors and Actuators B-Chemical, vol. 181, pp. 166-171, May 2013. [14] S. Soltani and A. M. Armani, "Optimal design of suspended silica on-chip splitter," Optics Express, vol. 21, pp. 7748-7757, Mar 25 2013. [15] A. J. Shaywitz and M. E. Greenberg, "CREB: A stimulus-induced transcription factor activated by a diverse array of extracellular signals," in Annual Review of Biochemistry. vol. 68, C. C. Richardson, Ed., ed, 1999, pp. 821-861. [16] B. W. Biggs, H. K. Hunt, and A. M. Armani, "Selective patterning of Si-based biosensor surfaces using isotropic silicon etchants," Journal of Colloid and Interface Science, vol. 369, pp. 477-481, Mar 1 2012. [17] H. K. Hunt and A. M. Armani, "Recycling microcavity optical biosensors," Optics Letters, vol. 36, p. 1092, 2011. [18] R. Gajasinghe, S. U. Senveli, S. Rawal, A. Williams, A. Zheng, R. H. Datar, et al., "Experimental study of PDMS bonding to various substrates for monolithic microfluidic applications," Journal of Micromechanics and Microengineering, vol. 24, p. 11, Jul 2014. [19] L. C. Xiong, P. Chen, and Q. S. Zhou, "Adhesion promotion between PDMS and glass by oxygen plasma pre-treatment," Journal of Adhesion Science and Technology, vol. 28, pp. 1046-1054, Jun 2014. 96 Chapter 6: Portable, fiber-based polarimetric stress sensor 6.1 Introduction Rapid and non-destructive materials characterization methods have greatly expanded knowledge of the fundamental behavior of materials, and enabled the development of numerous applications throughout society [1-3]. Additionally, the mechanical properties of a material can change over time providing information about the condition of a material, making monitoring of material behavior an important research goal. The Young’s modulus of a material is one such property that changes within a material over time. This information has been used to predict mechanical failure of materials. For example, by using mechanical deformation diagnostic measurements as a part of failure analysis, key aircraft components, such as helicopter blades, can have their remaining lifetimes predicted and appropriate preventative maintenance can be performed [4]. This type of analysis has recently been transferred to visco-elastic materials and therefore has gained interest withinin the bio-domain for studying how biological tissues change over time [5-8]. Visco-elastic and biological materials have complex sample handling requirements and exhibit significantly different mechanical behaviors. For example, biosafety cabinets are required for experiments with human tissue samples. These constraints make conventional measurement instrumentation, such as a load-frame or load cell, difficult to use in biological studies. As a result, several alternative methods have been sought out to solve these unique challenges, such as atomic force microscopy (AFM), sonoelastography, and nano-indentation [9-14]. In fact, these techniques have successfully characterized the Young’s modulus of biomimetic samples as well as tissue in previous work [15-20]. However, all of these methods have tradeoffs that limit their utility: AFM is very sensitive to environmental vibrations, sonoelastography requires manual compression of samples that is uncontrolled, and nano-indentation has a large footprint and generates results that require complex analysis methods. Therefore, a system which can meet the following requirements is necessary: 1) is small enough to fit inside a biosafety cabinet, 2) has high sensitivity comparable to other methods, 3) uses sterile or disposable sensors or parts, and 4) analyzes samples in a quick and non-destructive manner. 97 A method which meets the requirements for small footprint, disposability and non-destructive testing is a compliance sensor based on the inherent optical properties of polarization-maintaining (PM) optical fiber and how they change when external forces are applied [21-23]. Additionally, these types of sensors demonstrate a high noise-tolerance, with a comparable theoretical sensitivity. Previous work with polarimetric stress and pressure sensors used an analyzer and photodetector to probe the polarization state, which limits the amount of information the sensor provides. Furthermore, these sensors used free- space optical components, which are not portable and require highly skilled users due to their need for alignment [24-26]. In order for a compliance sensor to become widely used within the field of biology it must be easy to use, rapid, and highly portable. Addressing these issues by replacing the free space components of past sensors will allow a portable and compact polarimetric sensor to be made. In this chapter, an all-optical fiber sensor system is presented which solves the challenges listed above. A polarimeter is used to probe and measure the arbitrary polarization state of the light coming out of the fiber. Additionally, all the free-space components are replaced with fiber-optic counterparts. Due to these changes, no alignment is required and more information is obtained from the sensor. However, these changes require an expansion and generalization of the previous theoretical analysis of polarimetric sensors [24, 25, 27]. The fiber sensor is characterized using a series of polydimethylsiloxane (PDMS) samples prepared at various base:curing agent ratios as biomimetic samples. The Young’s modulus is measured and verified using a reference measurement from a commercial load-frame system. The result is an improved, portable, flexible, and easy-to-use optical fiber stress sensor validated with a visco-elastic material similar to tissue. 6.2 Theoretical Analysis 6.2.1 Sensor Operation The PM fiber sensor detects stress via the photo-elastic effect. As stress is applied to the fiber, the birefringence of the fiber changes in a known way; the beat length changes and the fast and slow axes are rotated by an angle ϕ (Figure 6-1). The more in depth analysis of the modified sensor expands upon that 98 in Chua,et.al.[27], with a two-dimensional cross-section of fiber that is acted upon by a force, f, in [N/m]. The force, f, is applied at an angle α with respect to the fast and slow axes of the fiber, and the equations for a normalized force, F, rotation angle, ϕ, and stressed beat length of the fiber (L b ) are given by: 𝐹 = 2𝑁 3 (1 + 𝜎 )(𝑝 12 − 𝑝 11 )𝐿 𝑏 0 𝑓 /(𝜆𝜋𝑏𝑌 ) (6.1) tan(2𝜙 ) = 𝐹 sin(2𝛼 ) /(1 + 𝐹 cos(2𝛼 )) (6.2) 𝐿 𝑏 = 𝐿 𝑏 0 (1 + 𝐹 2 + 2𝐹 cos(2𝛼 )) −1 2 ⁄ (6.3) In these equations, σ represents Poisson’s ratio for the fiber, p ij are the photoelastic constants of the fiber, L b0 is the un-stressed beat length of the fiber, Y is the Young’s modulus of the fiber, and b is the radius of the fiber cladding in meters. The following values for these variables are used based on the fiber utilized in the sensor system: σ = 0.17, p 11 = 0.121, p 12 = 0.27, L b0 = 2 mm, Y = 7.3x10 10 , and b = 62.5 µm [27]. Finally, the fast and slow axis of the fiber have refractive index N and N + ΔN 0 , respectively. The variable ΔN 0 is related to L b0 by ΔN 0 = λ/L b0 , where λ is the free space wavelength of light. Figure 6-1: Schematic diagram displaying the important angles used in the theoretical analysis as light travels through the sensor system. The variable α represents the angle between the applied force and the fast and slow axes of the PM fiber, the variable β represents the angle between the linearly polarized light and the fast and slow axes of the PM fiber, the variable γ represents the angle between the polarimeter axes and the fast and slow axes of the PM fiber, and the variable ϕ represents the angle of rotation that the fast and slow axes go when stress is applied to the PM fiber. 99 6.2.2 Transfer Matrices As can be seen in Figure 6-1, the light experiences multiple transformations as it travels through the sensor system. The specific changes that the light undergoes can be represented by the following set of transfer matrices, which describe the output based on linearly polarized light being sent into the system: [ 𝐸 𝑥 𝐸 𝑦 ] = [ cos 𝛾 sin 𝛾 −sin 𝛾 cos 𝛾 ] [ 1 0 0 𝑒 𝑗𝛿 ] [ cos 𝜙 − sin 𝜙 sin 𝜙 cos 𝜙 ] × [ 𝑒 −𝑗𝑘 𝑁 𝑠 𝑙 0 0 𝑒 −𝑗𝑘 𝑁 𝑓 𝑙 ] [ cos 𝜙 sin 𝜙 −sin 𝜙 cos 𝜙 ] [ cos 𝛽 sin 𝛽 − sin 𝛽 cos 𝛽 ] [ 𝐸 𝑥 0 0 ] (6.4) On the left side of equation 6.4, E x and E y are the x and y components of the electric field at the output of the PM fiber. On the far right of the equation, E x0 is the linearly polarized input light at the input of the PM fiber. The matrices read from right to left as light travels further down the system. First, linearly polarized light exits the in-line polarizer and enters the PM fiber, but may be at an angle β with respect to the fast and slow axes of the PM fiber. Eventually, the light will encounter the stressed section of the fiber, where the fast and slow axes are rotated at an angle ϕ with respect to the axes in the unstressed fiber. The light will then accumulate phase based on the length of the stressed region, l, and on the refractive indices of the fast and slow axes in the stressed region, N s and N f . The values N s and N f can be related to the stressed beat length of the fiber by 𝑁 𝑠 − 𝑁 𝑓 = 2𝜋 𝑘 𝐿 𝑏 ⁄ . Exiting the stressed section, there is a rotation of –ϕ as the light transitions back to the original rotation of the fast and slow axes of the fiber. When the light reaches the polarimeter, the fast and slow axes may be rotated at an angle γ with respect to the x and y axes of the polarimeter, where the light is measured. The variable δ in the transfer matrices still needs to be accounted for. This variable accounts for extra phase accumulated by the light as it travels down the fiber. Because the linearly polarized light is not aligned with the fast and slow axes of the PM fiber, the light will accumulate a phase difference before and after the stressed section of the fiber. The phase accumulated is related to the refractive indices of the fast and slow axes (and therefore the un-stressed beat length) as well as the length of the un- stressed fiber. This phase difference could have been accounted for with a matrix similar to the one for 100 phase accumulation in the stressed section of the fiber, but that would have required an accurate measurement of the fiber length, and ultimately, is unnecessary as it is a constant for a given measurement. Instead, this variable can easily be ascertained as part of the fitting algorithm process used to calibrate the sensor (and described in a section below). Therefore, the variable δ minimizes the complexity of taking measurements and keeps the stress sensor an easy-to-use system. 6.3 Experimental Methods 6.3.1 Experimental Setup Figure 6-2 shows a diagram of the testing setup. A laser light source at 980 nm or 1550 nm is used to couple light into an in-line polarizer. The polarizer is connected to a length of PM fiber, which is then connected to a polarimeter. A laptop computer reads and records the output of the fiber sensor as measured by the polarimeter. The PM fiber is secured on top of a load-frame (Instron) compression stage using tape, with the PDMS sample under test placed directly on top of the fiber. The load-frame compresses the sample, causing stress in the fiber, and the polarimeter monitors the polarization state change as stress is applied. At the same time, the load-frame takes a reference measurement of the stress and strain in the sample as it is compressed at a compression rate or crosshead velocity of 0.1 mm/s. Finally, a small amount of mineral oil is placed on the top and bottom of the sample to reduce friction and therefore reduce barreling of the sample, creating a more accurate measurement. Figure 6-2: Diagram of the experimental setup. A CW laser is attached to an in-line polarizer, which is in turn attached to a PM fiber. The PM fiber is run under the sample to be tested and attached to a polarimeter. The polarimeter is attached to a laptop computer, which records the data it measures. The sample is placed on a load-frame, which provides reference measurements as well as compression for the sample. 101 6.3.2 Calibration Every time the sensor is set up, the angles α, β, γ, and δ will be different, but α must be known to relate polarization state change to applied force or stress for that unique setup. Therefore, a calibration step must be performed to ensure accurate measurement of the Young’s Modulus. Calibration for the stress sensor is a multi-step process. Due to the simplicity of the setup, the theoretical analysis is generalized and made more complex. This carries over into the calibration process. However, in practice, the complex parts of the calibration are carried out by Matlab scripts and a fitting algorithm, keeping the calibration process executed by the user simple. In order to perform a calibration, a user of the sensor needs only to apply a known force to the fiber while acquiring data with the polarimeter. Once this calibration data is obtained, the fitting algorithm which produces a calibration curve may be run. Since none of the important angles should change as long as the setup does not change, this calibration only needs to be done once every time the system is set up. 6.3.2.1 Fitting algorithm The polarimeter measures raw data in the form of normalized stokes parameters. Multiplying out the matrices in equation 6.4Error! Reference source not found. will yield equations for the polarization state measured by the polarimeter in the form of E x and E y . Plugging in the values from equations 6.1, 6.2, and 6.3Error! Reference source not found. will yield the x and y components of the polarization state as a function of the angles α, β, γ, and δ. The equations for the Stokes parameters as a function of E x and E y are given below: 𝑠 0 = |𝐸 𝑥 | 2 + |𝐸 𝑦 | 2 (6.5) 𝑠 1 = |𝐸 𝑥 | 2 − |𝐸 𝑦 | 2 (6.6) 𝑠 2 = 2𝑅𝑒 {𝐸 𝑥 𝐸 𝑦 ∗ } (6.7) 𝑠 3 = −2𝐼𝑚 {𝐸 𝑥 𝐸 𝑦 ∗ } (6.8) Plugging in the values for E x and E y obtained by multiplying out the transfer matrices (equation 6.4Error! Reference source not found.), values for the Stokes parameters as a function of the angles α, 102 β, γ, and δ can be calculated. The polarimeter measures normalized stokes parameters, however, meaning that s 0 = 1and the vector <s 1 , s 2 , s 3 > always lies on the surface of the Poincaré sphere. Therefore, what is left is a system of three equations (s 1 , s 2 , s 3 ) and four unknowns (α, β, γ, δ), which cannot be solved directly. Instead of solving the system of equations, the system can be fit to a set of experimental data. This is done by using knowledge of how the four unknown angles will affect the polarization trace generated by polarimeter along with a known stress on the PM fiber. The fit is performed using an algorithm described below that was developed in Matlab, as well as with the reference data measured by the load-frame. As shown in Figure 6-3, as stress is applied to the fiber, the changing polarization state will cause the Stokes parameters to trace out a line that follows a circular path on the surface of the Poincaré sphere. Additionally, the circular trace will always have its center on the equator (s 3 = 0) of the Poincaré sphere. These two features are always present given the experimental setup and are consistent with the theoretical analysis. Furthermore, they can be leveraged to solve for β and γ, which are the first two angles that are fit in the algorithm. 103 Figure 6-3: Plot of raw data on the surface of the Poincaré sphere. As stress is applied to the fiber, the polarization state changes will trace out a circle on the surface of the sphere. The arrow on the graph indicates how the polarization changes with applied stress. When β changes, the radius of the circular trace on the Poincaré sphere will change (Figure 6-4a). This changing radius changes the horizontal position of the circle on the sphere, since the circular trace is always on the surface of the sphere. To solve for β, first the experimental values of the Stokes parameter vector <s 1 , s 2 , s 3 >, which corresponds to a vector <x, y, z> in Cartesian coordinates, are converted to spherical coordinates. The radial component (r) of the spherical coordinates will always be equal to one because <s 1 , s 2 , s 3 > always lies on the surface of the sphere; therefore, only θ (polar angle) and ϕ (azimuthal angle) need to be determined. These angles θ and ϕ are used as x and y coordinates in a 2D plane and fit to a circle, which yields a radius (R) of the circle in the 2D plane. This resulting radius (R) is the maximum value of ϕ [− 𝜋 2 , 𝜋 2 ], and β is equal to R/2. 104 Figure 6-4: Diagram indicating the effect of angles a) β and b) γ on the circular trace from stress on the fiber. a) With changing β, the circular trace will have different horizontal positions on the Poincaré sphere, and therefore different radii. b) With changing γ, the circular trace will rotate around on the equator, about the vertical axis of the Poincaré sphere. As γ changes, the center of the circular trace rotates around the equator of the Poincaré sphere (Figure 6-4b). From the 2D circle that was previously fit using θ and ϕ, the center coordinates are examined. The center ϕ-coordinate (the center y-coordinate) will always be equal to zero and the center θ-coordinate (the center x-coordinate) will give the location of the center of the circular trace. The angle γ is simply equal to -θ center /2. When δ changes, the starting point of the circular trace rotates around the circle (Figure 6-5a,b). The location and size of the circle will be fully determined by β and γ, but the starting position of the trace itself on that circle will be determined by δ. Due to this dependence, δ must be fit after β and γ are determined. To fit δ, the theoretical value of <s 1 , s 2 , s 3 > with applied force f = 0 is swept through varying values of δ to find the value of δ that makes it closest to the initial experimental value of <s 1 , s 2 , s 3 >. The range of δ before it starts repeating values of <s 1 , s 2 , s 3 > is 0 to 2π. The value of δ that yielded the best <s 1 , s 2 , s 3 > match is taken as the best-fit value for δ. 105 Figure 6-5: Diagrams indicating the effects of the angles δ and α on the circular trace. a) A circular trace for some given values of δ and α. b) If δ is changed, the starting point of the circular trace is rotated to a new position around the circle, but the arc-length of the circular trace remains constant. In this graph, only δ has been changed and α retains its original value. c) If α is changed, the circular trace traces out more of the circle. The starting point remains the same, but the arc-length of the circular trace increases. This graph has the same value of δ, and therefore the same starting point, as in b). Changing α can also cause the circular trace to trace out less of the circle and for the arc-length to decrease. The angle α is the most important angle to solve for as it determines the relationship between polarization state change and applied force. However, α cannot be determined until β, γ, and δ have been found. When α changes, the circular trace will complete more or less of the circle defined by β and γ (Figure 6-5b,c). The stress applied to the fiber at the final point of the circular trace in the experimental data set must be known in order to find α. The reference data from the load frame provides this final stress value for the present work, but the value could be determined in another manner, such as carefully placing a known weight on top of the fiber. This final stress must be multiplied by the diameter of the fiber in order to obtain the applied force, f [N/m], at the final point of the circular trace. Fitting α is similar to fitting δ. The theoretical value of <s 1 , s 2 , s 3 > with f equal to the final applied force in [N/m] is swept through varying values of α to find the value of α that makes it closest to the final experimental value of <s 1 , s 2 , s 3 >. The range of α before it starts repeating values is 0 to π/2. The value of α that yielded the best <s 1 , s 2 , s 3 > match is taken as the best-fit value for α. 6.3.2.2 Data processing As previously mentioned the polarimeter collects raw data in the form of three stokes parameters (s 1 , s 2 , s 3 ). In order to be useful, the collected data must be converted to a single variable, which will be referred to as ΔPol, which represents the change in polarization state. Furthermore, this conversion 106 should ideally preserve information contained in the Stokes parameters (s 1 , s 2 , s 3 ) and therefore cannot be done by throwing out any of the variables. Figure 6-6: a) Graph showing the circular trace of polarization state changes on the Poincaré sphere. The phase angle of the traced out circle can be used to generate ΔPol. b) A graph showing analyzed raw data (ΔPol) vs. time. The stokes parameters measured by the polarimeter and shown on the graph in a) are used to calculate ΔPol. The fact that the changing polarization state of light traveling through the fiber as it is stressed traces out a circle on the Poincaré sphere is leveraged to convert (s 1 , s 2 , s 3 ) to ΔPol (Figure 6-6). The phase angle of the circular trace is used for ΔPol and can be calculated such that it maintains all the information contained in the Stokes parameters. A step-by-step explanation of the process for calculating the phase angle of the circular trace follows, demonstrating how no information is lost from the stokes parameters (s 1 , s 2 , s 3 ) in this conversion process. 107 Figure 6-7: The circular trace created by stressing the optical fiber is always the intersection of the Poincaré sphere with a vertical plane. The circular trace lies on both the sphere and the vertical plane. The center of the circular trace always lies on the equator of the Poincaré sphere (s 3 = 0). Therefore the circular trace is always the intersection of a vertical plane, parallel to the s 3 -axis, and the sphere (Figure 6-7). The x-y coordinates of the circular trace on the vertical plane can be used to calculate the phase angle that will serve as ΔPol. The y-coordinate of the circular trace is given simply by s 3 since the plane is vertical and parallel to the s 3 -axis. 108 Figure 6-8: A view of the Poincaré sphere, vertical plane, and circular trace from above. From this angle, it is easy to see that the x-coordinate of the circular trace will be a rotation of s 1 and s 2 based on the angle θ center . Determining the x-coordinate of the circular trace requires rotating s 1 and s 2 to the reference frame of the intersecting vertical plane (Figure 6-8). In the previous section, the Stokes parameters were converted to spherical coordinates and the center of the circular trace was found in terms of the θ- coordinate. The same θ center value is the angle of rotation needed to convert s 1 and s 2 to the reference frame of the vertical plane. Therefore, the x-coordinate of the circular trace on the vertical plane is given by: 𝑥 − 𝑐𝑜𝑜𝑟𝑑𝑖𝑛𝑎𝑡𝑒 = −𝑠 1 × sin(𝜃 𝑐𝑒𝑛𝑡𝑒𝑟 ) + 𝑠 2 × cos(𝜃 𝑐𝑒𝑛𝑡𝑒𝑟 ) (6.9) The x-y coordinates of the circular trace on the vertical plane have now been found, and were calculated using all three stokes parameters, s 1 , s 2 , and s 3 . Thus, the x-y coordinates contain all the polarization state information that was contained in the Stokes parameters. The phase angle of the circular trace is given by: 𝑝 ℎ𝑎𝑠𝑒 = tan −1 ( 𝑦 − 𝑐𝑜𝑜𝑟𝑑𝑖𝑛𝑎𝑡𝑒 𝑥 − 𝑐𝑜𝑜𝑟𝑑𝑖𝑛𝑎𝑡𝑒 ) (6.10) Finally, ΔPol is calculated by normalizing the phase. The initial phase angle is subtracted from every subsequent phase, and all of the phase angles are divided by π. 109 The variable ΔPol is used to determine the stress on the fiber, and therefore the stress on the sample, measured by the fiber sensor by using a calibration curve. However, the strain on the sample must also be calculated. When the polarimeter measures the Stokes parameters, it measures them with respect to time, and this time axis is used to determine the strain. Multiplying the time axis by the compression rate (0.1 mm/s) yields the extension, or change in thickness, of the sample. Dividing this extension by the initial, measured thickness (height) of the sample yields strain. In addition to calibrating the data from the sensor, the raw data from the load-frame must also be slightly processed in order to obtain a stress-strain curve. The raw data from the load-frame comes in the form of force on the sample from the load-cell and extension from zero compression. To create a stress- strain curve the stress and the strain must be determined from these values. Strain is calculated by dividing the extension by the initial, uncompressed thickness (height) of the sample being tested. Stress is calculated by diving the measured force by the cross-sectional area (width x depth) of the sample being tested. Calculating stress and strain in the way described technically yields engineering stress and engineering strain, but these values are commonly used to report stress and strain in the literature. 6.3.2.3 Calibration curve After the data processing has taken place, the data still requires calibration, as it is left in the form of ΔPol vs. strain. The calibration combines the output from the fitting algorithm with that from the data processing to generate a stress-strain curve from the data measured by the fiber sensor. As mentioned above, the fitting algorithm requires the maximum stress value applied by the load frame in order to calculate a relationship between ΔPol and f. After the value for f in [N/m] of the maximum stress value has been calculated, and plugged into the fitting algorithm script, a theoretical circular trace that very closely matches the experimental data is generated. The theoretical circular trace relates the applied force, f [N/m], to the polarization state in the form of stokes parameters (s 1 , s 2 , s 3 ). 110 Figure 6-9: Graph depicting a calibration curve generated from the Matlab script and fitting algorithm. The black square points show the relationship between ΔPol and applied force, f [N/m], as calculated by the fitting algorithm. The red dashed line is a 3 rd -order polynomial fit to the relationship derived from the theoretical fit of the circular trace. The polynomial fit equation is shown on the graph and is used to calibrate the raw data measured by the polarimeter and turn it into applied force, which can in turn be converted to applied stress. Using the same data processing method described in the previous section, the theoretical value of <s 1 , s 2 , s 3 > is converted to ΔPol, generating a relationship between ΔPol and applied force, f. This relationship is then plotted, with f on the y-axis and ΔPol on the x-axis (Figure 6-9). The relationship is subsequently fit to a polynomial, usually 2 nd or 3 rd order, creating a calibration curve. The calibration curve gives an equation that determines f as a function of ΔPol. Therefore by plugging in experimental values of ΔPol, the resulting applied force, f [N/m], is calculated. The applied force must then be converted to stress by dividing it by the width of the fiber. Because the calibration curve is valid for a given set-up configuration, only one calibration curve needs to be generated per sample. In theory, a new calibration curve does not need to be generated as long as the system configuration does not change, even when changing samples. In practice, removing and replacing samples disturbed the fiber slightly (even 111 though it was taped down), which shifts α, so a new calibration curve was generated for each sample. Initial calibration measurements like this are frequently performed in many fields; for example, background or normalization spectra in UV Vis spectroscopy. 6.3.3 PDMS preparation In order to characterize and verify the operation of the fiber stress sensor system, six different polydimethylsiloxane (PDMS) samples were used. The samples were prepared with different base:curing agent ratios, from 5:1 up to 30:1, which yield different stiffness. The range of base:curing agent ratios was chosen as it spans about an order of magnitude in Young’s modulus and also has overlap with common biomaterials such as tissue [10, 17]. The samples were prepared as recommended by the manufacturer (aside from change the base:curing agent ratio) and then cut with a razor blade into a sample size of roughly 18 mm x 18 mm x 5 mm. The Instron load frame sample platen used to compress the sample fit completely over this size sample. 6.3.4 Sensor operation and experiments performed The sensor was tested with all six PDMS samples and for each sample ten successive measurements were performed. Two compression measurements were performed simultaneously: one by the Instron load-frame and one by the fiber stress sensor. Because the measurements were performed simultaneously, the Instron data serves as a very good reference measurement. When the sample is loaded onto the sensor and before any measurements were performed, the sample was pre-loaded slightly to make sure there was uniform contact between the sample and the PM fiber, even if there were slight irregularities in the sample. The setup was not disturbed in any way between the ten successive measurements so that the variables α, β, γ, and δ remained constant. Both 980 nm and 1550 nm lasers were used with multiple samples to evaluate the wavelength- dependent response of the fiber stress sensor. There are advantages and disadvantages to both wavelengths. Because more complete wavelengths will fit into the same length of fiber, shorter wavelengths should be more sensitive to perturbations like stress on the fiber. However, the overall 112 sensitivity of the sensor will vary slightly every time the sensor is set up because it depends on α as well as other angles that will vary when the sensor is set up. Considering a case where all variables relevant to sensitivity (α, β, γ, δ, l) are held constant can help shed light on how different wavelengths affect sensitivity. Performing some calculations using the theoretical analysis confirms that testing with 980 nm wavelength is more sensitive to applied force, f [N/m]. However, the theoretical sensitivity to stress does not give the complete picture of the sensitivity of the fiber sensor. For example, the applied force, f [N/m], does not take into account the width of the fiber, which is included in stress. In the experimental data presented, the diameter of the 1550 nm PM fiber used for testing was 400 μm, whereas the 980 nm PM fiber had a 245 μm diameter. Therefore, the 1550 nm fiber experiences more stress for the same applied force and the net effect is that both the 1550 nm and 980 nm fibers have a very similar theoretical sensitivity for the same applied force. In order to establish the background noise level and determine the robustness of the fiber sensor to different environments, noise measurements were taken in four locations. The four locations were the materials analysis lab where the primary experiments were performed, an optical device characterization lab on top of a vibration-isolating optical table, a countertop in a synthetic chemistry lab, and inside a laminar flowhood, which mimics the inside of a biosafety cabinet. In the materials analysis lab, the sensor was set up in the exact same way as it was for measuring samples, with the fiber taped down to the load frame. However, no sample was placed on the fiber and no compression was performed while data was collected. In the other three locations, the fiber stress sensor was set up in a similar way to the materials analysis lab, using a similar length of fiber that was taped down to the surface where measurements were performed. In all four locations, the noise was measured with 980 nm and 1550 nm wavelengths of light. Several sets of data were taken, and the polarization state changes were analyzed the same way as the compression data described in Section 6.3.2.2. The noise was evaluated by examining the distribution on histogram plots and fitting the distribution to a normal curve. The noise threshold was determined by using the standard deviation of the normal distribution curve fit. 113 6.4 Data and Results 6.4.1 PDMS compression experiments Because PDMS is a visco-elastic material, the stress-strain curves measured for PDMS were nonlinear [28, 29]. The Young’s modulus was therefore determined in the following manner: the data is fit to a 3 rd -order polynomial, and the derivative is taken at a defined strain. In the measurements presented in this chapter, both the reference data and the fiber sensor data were fit to a polynomial, and the Young’s modulus is given as the derivative of these polynomials at 30% strain. Representative data taken with 980 nm and 1550 nm lasers for a pair of 5:1 PDMS samples of the same approximate size is given in Figure 6-10. The stress-strain curve from the polarimetric sensor is laid on top of the curve from the load-frame, and polynomial fits for each data set are given as dashed lines. There is excellent qualitative agreement over most of the measurement range, aside from two small deviations: one for high strain values of the 980 nm graph (Figure 6-10a), and for low strain values of the 1550 nm graph (Figure 6-10b). These deviations could be due to a variety of different factors including noise in the testing environment or the sample shifting slightly in the oil used to reduce barreling. Because the Young’s modulus is determined from the polynomial fit of the whole dataset, small deviations like this do not significantly impact the measurement results overall. Figure 6-10: Stress-strain curves for 5:1 base:curing agent PDMS samples measured with a) a 980 nm laser and PM fiber b) a 1550 nm laser and PM fiber. The trace of red circles is data measured by the Instron load frame and the trace of black squares is data measured by the fiber sensor. The blue and green dashed lines are 3 rd -order polynomial fits for the load frame and fiber sensor, respectively. Despite some noise, there is very good agreement between the reference and fiber sensor data. 114 The data is compared quantitatively by examining the Young’s modulus of the reference and sensor data. The values obtained for the Young’s modulus for all the base:curing agent ratios, determined by taking the derivative of the polynomial fits of the polarimetric sensor and load-frame reference at 30% strain as previously mentioned, are given in Figure 6-11. The values presented in the figure are determined from the average of several measurements and shown with their standard deviations in error bars, which indicate that the measurements have good repeatability. There are a couple of important observations to be made from this graph. First, for most data points, there is very good agreement between the load-frame reference data and the polarimetric sensor data. Secondly, the deviation within individual datasets, indicated by the error bars, is very low for the sensor as well as the reference data. The good agreement and high accuracy are especially interesting because of the reduction in complexity of the overall sensor system compared to the load-frame and previous polarimetric sensor systems, as well as the reduction in footprint compared to the load-frame and previous polarimetric sensor systems. Figure 6-11: These plots show the calculated values for the Young’s modulus from both the load-frame measurements as well as the fiber sensor data for both a) 980 nm and b) 1550 nm testing wavelengths. The Young’s modulus values calculated from runs that were self-consistent were averaged and the average values are presented here with their standard deviations. 6.4.2 Noise measurements and sensitivity The ultimate theoretical sensitivity of the fiber-based sensor is determined by the noise level in the system. The two main noise sources are optical noise inherent in the setup and components used and environmental noise due to vibrations and temperature changes around the PM fiber. Because the first 115 source, optical noise inherent in the setup, is very difficult to determine, only the environmental noise was characterized in the four environments listed in Section 6.3.4 and the results of these measurements are shown in Figure 6-12. From the noise histograms, it is clear to see that the materials analysis lab was the noisiest environment, with mechanical testing equipment in near-constant use, and represents a worst-case scenario for operation of the sensor. The noise levels decreasing for the countertop and optical table are intuitive, and it is not surprising that the noise is lowest on the vibration-isolating optical table. What is somewhat surprising is that the laminar flowhood showed a lower noise level than the countertop. The constant airflow of the flowhood could be expected to cause extra vibrations and noise, but it seems to have stabilized the fiber instead, reducing noise due to environmental vibrations. 116 Figure 6-12: Noise histograms characterizing the level of noise measured by the fiber sensor in four different environments and two different wavelengths, shown with their normal distribution curves. Plots a-d) show data taken at 980 nm and plots e-h) show data taken at 1550 nm. Plots a,e) were taken in the materials analysis lab, plots b,f) were taken in a chemistry lab on a countertop, plots c,g) were taken inside of a laminar flowhood, and plots d,h) were taken on top of a vibration-isolating optical table. 117 In order to determine how the noise levels in the fiber stress sensor relate to ultimate sensitivity, some sensitivity calculations were performed. Using the transfer matrices (equation 6.4), a relationship between ΔPol, applied stress (σ), and interaction length (l) was calculated with Matlab assuming the optimum value for α and the other angles. A plot of this relationship for 980 nm and 1550 nm is shown in Figure 6-13. After this relationship was determined, an ideal operating curve was generated by fitting it to a function (with R 2 =1). The ideal operating curves predict the polarization change from the fiber (ΔPol) for a given interaction length (l) and applied stress (σ). Because α was fixed, these curves are, in fact, idealized, and in practice the sensitivity will vary as this angle will not always be optimized. The calculated ideal operating curves are given below for 980 nm (equation 6.11) and 1550 nm (equation 6.12Error! Reference source not found.). Δ𝑃𝑜𝑙 980 = (3.785 × 10 −5 )𝜎𝑙 (6.11) ∆𝑃𝑜𝑙 1550 = (3.907 × 10 −5 )𝜎𝑙 (6.12) The ideal operating curves given above were used, in conjunction with the standard deviation of the noise measurements, to determine the ultimate theoretical sensitivity of the fiber stress sensor. Plugging in the measured standard deviations for ΔPol and 18 mm for l (the approximate length of samples tested), the applied stress is calculated. Thus, the stress calculated corresponds directly to the standard noise level in the system; therefore, any stress above this calculated value should be detectable, and this value represents the minimum sensitivity. The standard deviations from the noise measurements in ΔPol are given along with the corresponding calculated minimum detectable stress in (Table 6-1) for both wavelengths in each environment where the noise was characterized. Table 6-1: Noise levels given in ΔPol and stress for different wavelengths and measuring environments ΔPol Stress (kPa) Location 980 nm 1550 nm 980 nm 1550 nm Mat. Analysis Lab 0.02073 0.01606 30 23 Countertop 0.01814 0.01189 26 17 Flowhood 0.007 0.00423 10 6 Optical Table 0.00333 0.00303 5 4 118 From the table, a few trends become evident. First, the decreasing noise from materials analysis lab down to optical table is more apparent. Secondly, it is plain to see that the noise levels and minimum detectable stress are lower in every environment for 1550 nm than for 980 nm. As mentioned previously, the 1550 nm PM fiber used for the experiments presented here was actually thicker than the 980 nm fiber. This thickness is likely the major contributing factor to lower noise from environmental sources for the 1550 nm experiments. Additionally, the longer wavelength of light itself may also have made 1550 nm less sensitive to small perturbations. Importantly, in all locations and for both wavelengths, the sensor is still sensitive enough to measure the Young’s modulus of biomaterials. Figure 6-13: Surface plots of the sensitivity curves given in a) Error! Reference source not found. and b) Error! Reference source not found.. a) Shows the sensitivity of the sensor operating at 980 nm for a given applied stress and interaction length. b) Shows the sensitivity of the sensor operating at 1550 nm for a given applied stress and interaction length. 6.5 Conclusions In this chapter, a portable, easy-to-use fiber-based polarimetric stress sensor was presented. A generalized theoretical analysis was used to reduce complexity in the experimental setup, and a fitting algorithm is described which handles the resulting additional complexity generated in the theoretical analysis. Using this analysis and fitting algorithm, the sensor is calibrated and used to accurately characterize the Young’s modulus of visco-elastic PDMS samples. The sensitivity of the sensor is explored via noise measurements in various locations. In addition, some alternative calculations based on the new theoretical analysis were performed to understand the relative noise and sensitivity. These sensitivity measurements, in conjunction with the PDMS testing, indicate that the sensor is sensitive 119 enough to accurately characterize the Young’s modulus of biomimetic and visco-elastic materials. The result is a flexible, portable, and simple-to-use tool that requires only a simple calibration that will be valuable to researchers interested in mechanically characterizing soft and visco-elastic materials, such as tissue [7, 9]. Chapter 6 References [1] F. Cellini, S. Khapli, S. D. Peterson, and M. Porfiri, "Mechanochromic polyurethane strain sensor," Applied Physics Letters, vol. 105, p. 061907, 2014. [2] H. Zhou, H. Zhang, Y. Pei, H.-S. Chen, H. Zhao, and D. Fang, "Scaling relationship among indentation properties of electromagnetic materials at micro- and nanoscale," Applied Physics Letters, vol. 106, p. 081904, 2015. [3] H. Fu, S. Xu, R. Xu, J. Jiang, Y. Zhang, J. A. Rogers, et al., "Lateral buckling and mechanical stretchability of fractal interconnects partially bonded onto an elastomeric substrate," Applied Physics Letters, vol. 106, p. 091902, 2015. [4] G. J. Kacprzynski, A. Sarlashkar, M. J. Roemer, A. Hess, and W. Hardman, "Predicting remaining life by fusing the physics of failure modeling with diagnostics," JOM, vol. 56, pp. 29- 35, 2004. [5] K. Hoyt, B. Castaneda, M. Zhang, P. Nigwekar, P. A. di Sant'Agnese, J. V. Joseph, et al., "Tissue elasticity properties as biomarkers for prostate cancer," Cancer Biomarkers, vol. 4, pp. 213-225, 2008. [6] I. D. Johnston, D. K. McCluskey, C. K. L. Tan, and M. C. Tracey, "Mechanical characterization of bulk Sylgard 184 for microfluidics and microengineering," Journal of Micromechanics and Microengineering, vol. 24, p. 7, Mar 2014. [7] T. A. Krouskop, T. M. Wheeler, F. Kallel, B. S. Garra, and T. Hall, "Elastic moduli of breast and prostate tissues under compression," Ultrasonic Imaging, vol. 20, pp. 260-274, Oct 1998. [8] N. S. Lu, C. Lu, S. X. Yang, and J. Rogers, "Highly Sensitive Skin-Mountable Strain Gauges Based Entirely on Elastomers," Advanced Functional Materials, vol. 22, pp. 4044-4050, Oct 2012. [9] A. Samani, J. Bishop, C. Luginbuhl, and D. B. Plewes, "Measuring the elastic modulus of ex vivo small tissue samples," Physics in Medicine and Biology, vol. 48, pp. 2183-2198, Jul 2003. [10] A. Samani, J. Zubovits, and D. Plewes, "Elastic moduli of normal and pathological human breast tissues: an inversion-technique-based investigation of 169 samples," Physics in Medicine and Biology, vol. 52, pp. 1565-1576, Mar 2007. [11] P. N. T. Wells and H. D. Liang, "Medical ultrasound: imaging of soft tissue strain and elasticity," Journal of the Royal Society Interface, vol. 8, pp. 1521-1549, Nov 2011. [12] U. Zaleska-Dorobisz, K. Kaczorowski, A. Pawlus, A. Puchalska, and M. Inglot, "Ultrasound Elastography - Review of Techniques and its Clinical Applications," Advances in Clinical and Experimental Medicine, vol. 23, pp. 645-655, Jul-Aug 2014. [13] L. Penuela, F. Wolf, R. Raiteri, D. Wendt, I. Martin, and A. Barbero, "Atomic force microscopy to investigate spatial patterns of response to interleukin-1beta in engineered cartilage tissue elasticity," Journal of Biomechanics, vol. 47, pp. 2157-2164, Jun 2014. [14] M. Horimizu, T. Kawase, T. Tanaka, K. Okuda, M. Nagata, D. M. Burns, et al., "Biomechanical evaluation by AFM of cultured human cell-multilayered periosteal sheets," Micron, vol. 48, pp. 1-10, May 2013. [15] B. S. Garra, "Imaging and estimation of tissue elasticity by ultrasound," Ultrasound Q, vol. 23, pp. 255-68, Dec 2007. 120 [16] D. W. Good, A. Khan, S. Hammer, P. Scanlan, W. M. Shu, S. Phipps, et al., "Tissue Quality Assessment Using a Novel Direct Elasticity Assessment Device (The E-Finger): A Cadaveric Study of Prostatectomy Dissection," Plos One, vol. 9, p. 8, Nov 2014. [17] D. W. Good, G. D. Stewart, S. Hammer, P. Scanlan, W. Shu, S. Phipps, et al., "Elasticity as a biomarker for prostate cancer: a systematic review," Bju International, vol. 113, pp. 523-534, Apr 2014. [18] S. R. Mousavi, A. Sadeghi-Naini, G. J. Czarnota, and A. Samani, "Towards clinical prostate ultrasound elastography using full inversion approach," Medical Physics, vol. 41, p. 12, Mar 2014. [19] J. Ophir, I. Cespedes, H. Ponnekanti, Y. Yazdi, and X. Li, "ELASTOGRAPHY - A QUANTITATIVE METHOD FOR IMAGING THE ELASTICITY OF BIOLOGICAL TISSUES," Ultrasonic Imaging, vol. 13, pp. 111-134, Apr 1991. [20] B. M. Ahn, J. Kim, L. Ian, K. H. Rha, and H. J. Kim, "Mechanical Property Characterization of Prostate Cancer Using a Minimally Motorized Indenter in an Ex Vivo Indentation Experiment," Urology, vol. 76, pp. 1007-1011, Oct 2010. [21] J. Calero, S. P. Wu, C. Pope, S. L. Chuang, and J. P. Murtha, "THEORY AND EXPERIMENTS ON BIREFRINGENT OPTICAL FIBERS EMBEDDED IN CONCRETE STRUCTURES," Journal of Lightwave Technology, vol. 12, pp. 1081-1091, Jun 1994. [22] T. K. Noh, U. C. Ryu, and Y. W. Lee, "Compact and wide range polarimetric strain sensor based on polarization-maintaining photonic crystal fiber," Sensors and Actuators a-Physical, vol. 213, pp. 89-93, Jul 2014. [23] Y. Verbandt, B. Verwilghen, P. Cloetens, L. VanKempen, H. Thienpont, I. Veretennicoff, et al., "Polarimetric optical fiber sensors: Aspects of sensitivity and practical implementation," Optical Review, vol. 4, pp. 75-79, Jan-Feb 1997. [24] R. C. Gauthier and J. Dhliwayo, "BIREFRINGENT FIBEROPTIC PRESSURE SENSOR," Optics and Laser Technology, vol. 24, pp. 139-143, Jun 1992. [25] G. Liu and S. L. Chuang, "Polarimetric optical fiber weight sensor," Sensors and Actuators a- Physical, vol. 69, pp. 143-147, Aug 1998. [26] V. M. Murukeshan, P. Y. Chan, L. S. Ong, and A. Asundi, "Effects of different parameters on the performance of a fiber polarimetric sensor for smart structure applications," Sensors and Actuators a-Physical, vol. 80, pp. 249-255, Mar 2000. [27] T. H. Chua and C. L. Chen, "FIBER POLARIMETRIC STRESS SENSORS," Applied Optics, vol. 28, pp. 3158-3165, Aug 1989. [28] I. K. Lin, K.-S. Ou, Y.-M. Liao, Y. Liu, K.-S. Chen, and X. Zhang, "Viscoelastic Characterization and Modeling of Polymer Transducers for Biological Applications," Journal of Microelectromechanical Systems, vol. 18; 16, pp. 1087-1099, 2009. [29] A. Mata, A. J. Fleischman, and S. Roy, "Characterization of Polydimethylsiloxane (PDMS) Properties for Biomedical Micro/Nanosystems," Biomedical Microdevices, vol. 7, pp. 281-293, 2005. 121 Chapter 7: Future Work 7.1 Introduction In the previous chapters, several research projects were described which helped advance knowledge in their respective areas, even if only slightly. However, the nature of research is such that it is always moving forward, and it is important to consider future directions and applications of projects and research that have been completed. In this chapter, such future applications and projects of the work previously described in this dissertation will be discussed. 7.2 Polymer waveguides Although the silica suspended waveguides presented in Chapter 4 had low loss, it is reasonable to expect that the propagation loss could be further reduced. One attempt at this was made by using a silica sol-gel thin film to reduce the index contrast of the waveguide and therefore reduce the number of supported modes. Reducing the number of supported modes should reduce the propagation loss, since higher order modes tend to have higher loss. This should also reduce dispersion in the waveguide as well. Unfortunately, the silica sol-gel thin film coating did not work. Other materials could be used to create a low-index cladding coating besides silica sol-gels. For example, low-index polymers [1, 2] could be spun-coat onto the waveguides to be used as a cladding. As long as the index is suitably low and the coating is suitably thick, then there is good potential for lowering the number of supported modes in the device and therefore lowering the propagation losses in the silica suspended waveguides. Additionally, the possibility of making devices in a similar geometry, but wholly out of optical polymers, could be explored. The unique, suspended geometry has some advantages, and polymers are versatile materials. Instead of a CO 2 laser reflow, the devices could be reflowed thermally. This could also be a route towards integrating silica microtoroid resonators with a waveguide on a single chip, as polymers may be manipulated more easily than silica to make sure the waveguide is close enough for efficient coupling. Finally, the possibility of coating the silica suspended waveguides with a birefringent polymer coating was explored in Appendix B. Due to significant accumulation under the waveguide arms, this 122 coating was unsuccessful. However, it may be possible to fabricate a waveguide from a birefringent polymer in the same geometry as the silica suspended waveguides. This would require an appropriate etchant for the polymer used, as well as a polymer or liquid crystal polymer that would align itself appropriately to be birefringent. One option for alignment would be to use a silicon wafer with a very thin thermal silica layer as the base, and spin-coat a homeotropically-aligning liquid crystal polymer onto the silica substrate. The polymer would then be appropriately aligned, and fabrication could proceed with the polymer etchant. An additional, short buffered HF step would be needed to etch away the thermal silica base to expose the underlying silicon for subsequent XeF 2 etching. 7.3 Waveguide splitter biosensor In Chapter 5, it was determined that the silica suspended waveguide splitters were not a very good candidate for biosensing applications. This conclusion was based on the fact that there was variation in the output during sensing experiments due to uneven wetting of the functionalized device surface. Additionally, attempts at integrating the device with microfluidic channels in order to solve this problem failed. One possibility for obtaining consistent data is to use a detergent, such as sodium dodecyl sulfate (SDS), in order to promote wetting of the device surface. The detergent works by masking negative charges and therefore inhibiting hydrophobicity caused by polar water molecules, promoting even wetting of the sample surface [3]. This would eliminate any variations due to the size of the liquid drop on the surface of the splitter device or any spreading of the drop. There still might be device-to-device variation if the waviness of the splitter devices due to the CO 2 laser reflow process during fabrication turns out to cause differences in output during sensing experiments. In that case, the CO 2 laser reflow process would need to be adjusted in order to reduce the waviness in the waveguide arms. 7.4 Polarimetric fiber stress sensor In Chapter 6, a small, portable optical-fiber based stress sensor was demonstrated and characterized. The sensor was characterized at 980 nm and 1550 nm wavelengths of light, but shorter wavelengths were shown to provide higher sensitivity. Therefore, it would be worthwhile to test the capabilities of the sensor at a shorter wavelength, such as 635 nm. Additionally, in that chapter, the 123 sensor was operated on a load frame stage which not only provided compression to the sample and reference data, but was also used to calibrate the sensor. In order to use the fiber stress sensor in other locations, a method for calibrating the sensor and compressing the sample without the load frame must be developed. In fact, work on solving these problems has already begun. In order to calibrate the stress sensor, a known force must be applied to the fiber. Therefore, a calibration weight can be used very easily to perform this calibration. However, there are two important restrictions to be considered when using a calibration weight. First, because the calibration must be performed after the sensor is set up, the weight must fit within the area where sensing will be performed; in other words, the calibration weight cannot be too large. Second, the weight cannot be placed directly on the fiber. If it is, it will lean to one side and its full weight will not stress the fiber as desired. Therefore, for future development, it will be beneficial to explore other options for calibrating the sensor. If the calibration could be automated, by applying a known force via a compression stage, that could further extend the usefulness of the stress sensor and make it easier for the end user to operate. A small compression stage has been built to allow compression of the sample without using the load frame. This stage consists of a one-axis translation stage mounted to a base using an L-bracket, a motorized, computer-controlled micrometer, another L-bracket used to mount a custom-machined compression platen. The one-axis translation stage is controlled using the micrometer and the custom platen is mounted so that when the stage moves, it will provide compression to the sample. In this setup, the micrometer can be controlled to move at a known rate, which will produce a known compression rate. The strain on the sample can therefore be calculated in the same way it was calculated with the load frame measurements. Even though this strain stage is semi-automated, it requires input from the user. It could be fully automated based on the dimensions of the sample or even by choosing a maximum stress to be applied if the sensor is calibrated properly. This automation of the setup would involve additional engineering work, but would likely not change the required components of the sensor if designed properly. 124 The fiber stress sensor presented in Chapter 6 was designed to be used with biological samples, such as tissue, and testing it with these samples is an important next step for research with the device. The goal of this work is to ultimately use the fiber sensor as a diagnostic tool, measuring the Young’s modulus of cancerous tissue samples in order to gain more information about them. In order to do that, not only do cancerous samples need to be measured by the sensor, but the same samples should undergo standard pathology analysis in order to determine how their stiffness correlates with the status of the sample [4, 5]. In this way, a map could be built up which relates the stiffness of a biopsied tissue sample with the aggressiveness of the tumor or some other relevant pathological information. A study such as this must be conducted carefully, as it requires expertise in cancerous tissues as well as approval from institutional research advisory boards, since it involves handling human tissue samples. Chapter 7 References [1] J. Liang, E. Toussaere, R. Hierle, R. Levenson, J. Zyss, A. V. Ochs, et al., "Low loss, low refractive index fluorinated self-crosslinking polymer waveguides for optical applications," Optical Materials, vol. 9, pp. 230-235, 1998. [2] A. Sulaiman, S. W. Harun, K. S. Lim, F. Ahmad, and H. Ahmad, "Microfiber Mach-Zehnder interferometer embedded in low index polymer," Optics and Laser Technology, vol. 44, pp. 1186- 1189, 2012. [3] E. E. Dreger, G. Klein, G. Miles, L. Shedlovsky, and J. Ross, "Sodium Alcohol Sulfates.Properties Involving Surface Activity," Industrial & Engineering Chemistry, vol. 36, pp. 610-617, 1944/07/01 1944. [4] M. J. Paszek, N. Zahir, K. R. Johnson, J. N. Lakins, G. I. Rozenberg, A. Gefen, et al., "Tensional homeostasis and the malignant phenotype," Cancer Cell, vol. 8, pp. 241-254, Sep 2005. [5] R. W. Tilghman, C. R. Cowan, J. D. Mih, Y. Koryakina, D. Gioeli, J. K. Slack-Davis, et al., "Matrix Rigidity Regulates Cancer Cell Growth and Cellular Phenotype," Plos One, vol. 5, p. 13, Sep 2010. 125 Appendix A: Graphene coated waveguide splitters A.1 Introduction Active optical devices are devices whose output signal can be controlled by varying a second signal—typically electrical—going to the device. An electrical analogue would be an electrical circuit involving transistor elements. The signal passing through a transistor changes based on a second signal going to the transistor gate. These types of devices are attractive for several reasons. First and foremost, they can be used to make active photonic circuits, which are useful in the communications field. Presently in optical communications, optical waveguides (typically optical fibers) are used to transmit signals over long distances, but most processing and routing of these signals still takes place in the electrical domain. Therefore, signals must be converted from the optical to the electrical domain to be processed and then back to an optical signal again before they can be re-transmitted. This conversion process causes a loss of speed and bandwidth in optical communication links. Therefore, if it could be replaced by an active photonic circuit which processes optical signals in the appropriate ways, great improvements could be made in the area of optical communications. Active optical devices often rely upon nonlinear effects in order to work. There are a wide variety of effects and optical material properties that are classified as nonlinear, such as thermo-optic effects, saturable absorption and electro-optic effects. One nonlinear optical material that has recently been the focus of numerous research activities in the areas of optics and electronics is graphene. Graphene is a single sheet of carbon atoms arranged in a hexagonal crystal lattice and has very unique optical and electronic properties due to its very unique bandgap structure [1]. Graphene is a saturable absorber, has been shown to exhibit third-order nonlinear optical effects, and also can have a very strong interaction with light despite being very thin [2-5]. It has also been used to make optical modulators by taking advantage of changing its absorption with an applied bias [6, 7]. The unique properties of graphene and its novelty as a material have caused many researchers to look for new and unique ways to use graphene in specific optical and electronic applications. This appendix investigates 126 the possibility of coating silica suspended waveguide splitter devices with a layer of graphene in order to turn them into active devices by leveraging the unique electro-optical properties of graphene. A.2 Theory Graphene is a very unique material due to its optical and electronic properties. It has an electronic band structure unlike any other material, with its conduction and valence bands meeting at Dirac points (Figure A-1). This gives graphene the unique ability to have its band gap tuned via electrical bias. By applying a bias voltage across a sheet of graphene, the Fermi level can be raised or lowered, and the band gap of the material can be effectively changed. Because the band gap determines the absorption behavior of a material, it is possible to tune the absorption wavelength of graphene by applying a bias voltage. This feature clearly has important implications for optics. Additionally, even a single sheet of graphene has strong absorption. Figure A-1: Diagram showing the unique band structure of graphene at a Dirac point. a) Normally, the Fermi energy is right where the valence and conduction bands meet, and any energy transition is allowed, meaning graphene has no band gap and will absorb incident photons of any wavelength. b) If the graphene is biased electrically, the Fermi energy will move, and some transitions will not be allowed, inducing a band gap of a desired energy level. These electrical and optical properties can be leveraged to make active devices because the splitting ratio of the splitters depends on the refractive index of the device itself and its environment. Generally speaking, modifying the absorption of a material will also modify its refractive index, even if only 127 slightly. Therefore, tuning the absorption of graphene on top of a suspended splitter device will change the splitting ratio. This is the mechanism used for controlling the output of the waveguide splitters. A.3 Fabrication methods A.3.1 Electrodes In order to use the graphene-coated waveguide splitters as an active device, electrodes are added to the devices. This requires a slight modification to the fabrication procedures. Before the XeF 2 etching step, a third photolithography step is performed to create a photoresist shadow mask. The shadow mask is used for a thermal evaporator deposition of chromium onto the device to serve as electrode contacts for the graphene. After the electrodes are deposited, the photoresist shadow mask is washed off by placing the devices in an acetone bath and sonicating. Finally, the devices are cleaned and fabrication proceeds as usual. Because they are added before the XeF 2 step, when the silicon substrate is etched away, the electrodes will remain at the same elevation as the silica sol-gel as shown in Figure A-2. This makes it easier to deposit graphene, since the graphene does not have to sag to reach the electrodes and does not hamper any other part of the fabrication process. Figure A-2: Cross-section of waveguide splitter device with electrodes on either side of the splitting region. A.3.2 Graphene growth and deposition This work was pursued in collaboration with a group at the Korean Institute of Science and Technology (KIST). The group at KIST fabricated and deposited the graphene onto the waveguide splitter devices provided. The graphene was grown using a CVD method. After growth, the graphene is transferred to a layer of polymer. Some graphene was transferred to a polyimide (PI) layer and some was transferred to a polymethyl methacrylate (PMMA) layer. The graphene and polymer were deposited together onto the splitter devices, with the polymer layer providing mechanical support to the graphene. 128 In all cases, the graphene itself was in contact with the device, and the polymer layer was above it (Figure A-3). After graphene has been deposited, the waveguide splitters were returned to be tested using the waveguide testing setup. Figure A-3: Images of graphene coated waveguide splitter devices at a) 20x zoom and b) 50x zoom. A.4 Experiments In order to use the electrodes, a probe station was added to the waveguide setup that consists of two probes and a high-voltage source. This allowed bias voltages to be added to the device while testing. The probe station was integrated into the waveguide testing setup in such a way as to allow the setup to operate without any additional interference. The cameras on the setup were used to align the probes and make contact with the electrodes. When the devices arrived back to USC, the graphene/polymer layer extended over the surface of several samples, so each device had to be cut out individually. This was attempted very carefully with razorblades and tweezers, but once the graphene/polymer layer was cut, the graphene became loose. The PMMA samples seemed to be slightly more attached to the substrate than the PI samples, but samples with both polymers had the problem of the graphene/polymer layer coming loose. Additionally, the devices were also damaged, and the output signal from them was not measurable. The plan had been to test them both with DC and AC bias voltages as well as pumping them with high power in order to obtain modulator or nonlinear behavior from the active devices. 129 A.5 Future Work Due to unsuccessful coating attempts and the distance of the collaboration, this project was abandoned. If a better method for graphene deposition was developed, it might be possible to perform further investigations. The graphene layer would not need to extend beyond the sample containing the device, however. Additionally, it would need to be more securely attached, and it would be preferable to not use a polymer backing. The polymer adds extra loss to the system that could potentially be avoided if a better deposition method was developed. Also, there would not be a need to scratch through the polymer with a probe tip to make contact with the graphene if it were removed. Another route for making modulator devices is to use a different non-linear material. If the suspended waveguide splitters could be fabricated using an electro-optic material, such as lithium niobate, or some hybrid material containing a material with electro-optic properties, a modulator could be built. The device would require the same electrodes as the graphene device, which would be used to generate an electric field to modulate the material as opposed to biasing graphene. However, an electro-optic device would also require electric poling, which could be achieved using the electrodes already fabricated on the device. Fabricating a device with the suspended splitter geometry having electro-optic properties is a difficult challenge that would need to be addressed for this route to be feasible. If a better graphene deposition method is developed or a route for imparting electro-optic nonlinear properties is developed, then the devices could be characterized as modulators. By measuring their output characteristics at different input voltages, a picture of their operation could be built. Additionally, their insertion loss and efficiency could be characterized, giving insight into their usefulness as communications components. At that point, any information gained could be used to make modifications to the design in order to decrease loss and improve efficiency. Appendix A References [1] P. Avouris, "Graphene: Electronic and Photonic Properties and Devices," Nano Letters, vol. 10, pp. 4285-4294, Nov 2010. [2] G. K. Lim, Z. L. Chen, J. Clark, R. G. S. Goh, W. H. Ng, H. W. Tan, et al., "Giant broadband nonlinear optical absorption response in dispersed graphene single sheets," Nature Photonics, vol. 5, pp. 554-560, Sep 2011. 130 [3] T. Gu, N. Petrone, J. F. McMillan, A. van der Zande, M. Yu, G. Q. Lo, et al., "Regenerative oscillation and four-wave mixing in graphene optoelectronics," Nature Photonics, vol. 6, pp. 554- 559, Aug 2012. [4] K. Kim, S. H. Cho, and C. W. Lee, "NONLINEAR OPTICS Graphene-silicon fusion," Nature Photonics, vol. 6, pp. 502-503, Aug 2012. [5] Q. Ye, J. Wang, Z. B. Liu, Z. C. Deng, X. T. Kong, F. Xing, et al., "Polarization-dependent optical absorption of graphene under total internal reflection," Applied Physics Letters, vol. 102, p. 4, Jan 2013. [6] A. Majumdar, J. Kim, J. Vuckovic, and F. Wang, "Electrical Control of Silicon Photonic Crystal Cavity by Graphene," Nano Letters, vol. 13, pp. 515-518, Feb 2013. [7] M. Liu, X. B. Yin, E. Ulin-Avila, B. S. Geng, T. Zentgraf, L. Ju, et al., "A graphene-based broadband optical modulator," Nature, vol. 474, pp. 64-67, Jun 2 2011. 131 Appendix B: Mentoring Projects This appendix presents work that was undertaken by undergraduate and high school students that were being mentored by the author. The work presented here did not lead to publications, but the findings may be of interest and could be expanded upon for future work. B.1 Polymer waveguide coatings Additional functionality could be imparted to the silica suspended trapezoidal waveguides described in Chapter 4 by using a birefringent polymer coating. Birefringent liquid crystal polymers have been used to make polarization-dependent waveguides and passive waveguide devices [1-3]. The trapezoidal waveguides already have a geometry that makes them susceptible to polarization-dependent behavior. However, they are large enough to support modes of both TE and TM polarized light. Alexei Naumann, an undergraduate student at USC, investigated adding birefringent polymer coatings to suspended silica waveguides in order to impart polarization-dependent behavior in 2014. B.1.1 Liquid crystal polymers Liquid crystals (LCs) are a class of liquid materials whose molecules exhibit crystal-like long- range directional order and orientation. They exist in many phases, each of which has molecules which align themselves in a unique way. Additionally, these materials only enter their liquid crystal phase under certain temperature and pressure conditions. Due to their long-range directional order, LCs often display unique optical properties. For example, many liquid crystal phases exhibit birefringence. Liquid crystal polymers, also known as polymerizable liquid crystals, are a class of materials that exhibit liquid crystal-like properties, but which can also be polymerized. The polymerization freezes the LC molecules in place, locking the molecular orientation and therefore the optical properties in place as well. Because of these properties, LC polymers are often employed in conformal coatings. They can be applied in liquid form, either by dip-coating or spin coating, and then polymerized, imparting their unique optical properties to the device being coated. 132 Because liquid crystals exist in several varieties, it is important to pick an appropriate liquid crystal for the application. In the realm of optical devices and waveguides, a birefringent LC polymer is typically desired [2]. These polymers are useful for imparting polarization-dependent properties onto the waveguide or device. Often, waveguides are fabricated from the LC polymers themselves. In these and other cases, care must be taken to make sure that the LC molecules align themselves appropriately to give the desired polarization-dependent behavior [1, 3]. This typically requires the use of an alignment layer, which encourages the liquid crystal molecules to line up in a particular direction. These alignment layers are frequently made from rubbed polymer layers. In other words, researchers use a machine to physically rub a deposited polymer layer to create microscopic grooves in the surface for LC molecules to gather, encouraging a specific alignment direction. B.1.2 Liquid Crystal Coatings on Suspended Silica Waveguides For the suspended trapezoidal silica waveguides discussed in Chapter 4, certain considerations must be taken into account when attempting to coat them with liquid crystal polymers. Perhaps the most obvious is that the suspended silica devices cannot utilize a rubbed alignment layer or be rubbed themselves. The devices which use a rubbed alignment layer often are fabricated directly out of the LC polymer. In this case, the suspended silica waveguides will be coated in the LC polymer. If another polymer were deposited as an alignment layer, it would need to be very thin to maximize the optical interaction with the LC polymer, which is not practical. If the device was rubbed directly, it would degrade the surface and likely increase loss. In both cases, if the device was rubbed, it would almost certainly break, due to the suspended nature of the waveguide structure. Therefore, a coating made of a LC polymer that does not need a rubbed alignment layer will be pursued. 133 Figure B-1: A schematic diagram of a trapezoidal waveguide with a homeotropic liquid crystal coating. The liquid crystal molecules, shown in yellow, align themselves perpendicularly to the silica substrate, which in this case is the waveguide itself. In order to create LC polymer coated devices, a homeotropically aligned liquid crystal was used. These liquid crystals align themselves perpendicularly to the substrate (Figure B-1), and do not require an alignment layer. In order to avoid design and synthesis of a polymer for the application, a LC polymer was purchased from EMD Millipore (RMS 03-015) that exhibits birefringence, is homeotropically aligned, and requires UV curing. This polymer has many of the necessary requirements for the birefringent coatings on the suspended waveguides. Unfortunately, many potential LC polymers, including the one being used in the present work, have a higher refractive index than silica. This means that the polymer coating must be thin enough that it does not act as a waveguide core itself. This is in contrast to the previous section, in which thick coatings were desired. Additionally, because a thin coating is desired, it is preferable that the coating is very uniform so that device behavior is consistent. B.1.3 Device Coating Experiments Alexei attempted to coat the waveguides using both a spinner and a dip-coat method. The spin speed for spinning was chosen based on information provided by the manufacturer. Devices were spin- coated at 5000 RPM for 30 seconds and cured under a UV lamp for approximately three minutes, during which polymer coating solidified. This process resulted in layers approximately 500 nm thick when spin- coated on bare silicon wafers with a thermal oxide layer. However, difficulties arose when attempting to 134 spin-coat the LC polymer onto devices. The coatings on devices were very non-uniform, and in many cases, did not cover the waveguide arms (Figure B-2a). Figure B-2: a) An SEM image close up of a waveguide channel for a device that has been spin-coated with a LC polymer. The liquid crystal builds up underneath the waveguide arm and does not coat the top of the waveguide channel. b) An SEM image of the end of a waveguide that has been dip-coated with a LC polymer. The liquid crystal build-up is even more severe in this case. In an attempt to generate a more even coating, a few other experiments were conducted. The devices were treated with O 2 plasma before spin-coating, which did not improve the results. Additionally, the wafer was dip-coated by hand (Figure B-2b). In this case, the coating build-up underneath the waveguide arms was even thicker, and the waveguide arms still remained relatively bare. Devices that were coated via spin-coating and dip-coating methods were tested with the waveguide testing setup, but in both cases the loss was very high and no polarization-dependent behavior was observed. Finally, as an additional experiment, a silica microsphere resonator was dip-coated with the LC polymer. The quality factor of the device was measured before coating, and the coated device was exposed to the same UV curing parameters as the waveguides. However, when attempting to measure the quality factor after coating, the tapered fiber used for coupling light into the device touched the sphere (a common occurrence), and removed some of the LC polymer coating when directed back away from the sphere. This indicates that the polymer surface coating was not fully cured. Although this problem could have been easily fixed, no additional experiments were pursued with microsphere resonators due to time constraints. (a) (b) 135 B.1.4 Conclusions and Future Work In conclusion, attempting to create polarization-dependent devices was not successful. Due to complications with the liquid crystal polymer, it was not possible to create the coating necessary to enable proper operation of the suspended silica waveguides. However, with a liquid crystal engineered specifically for this purpose, it still might be possible to create a polarization-dependent waveguide. This would require a polymer that can be coated very thinly (most likely, by spin-coating) on the waveguide devices without accumulating under the waveguide arms. Additionally, a polymer that attaches to silica selectively could be used. Finally, it could be possible to fabricate suspended devices from a LC polymer with the same geometry by using similar fabrication methods to the silica suspended devices. This would require an appropriate etchant for the polymer and the ability of the LC molecules to be aligned properly on top of a silicon substrate. B.2 Sol-gel recipes Silica sol-gels are an area of active research and have been used for many projects in the Armani Research Lab [4-7]. Frequently, researchers seek to tune their properties by adding dopants [8, 9]. However, in certain cases, such as the silica sol-gel coated waveguides presented in Chapter 4, it is desirable to slightly modify the properties without large changes. Samantha Wathugala, a high school student who worked in the lab in the summer of 2013, investigated how the properties of silica sol-gels could be tuned by just varying the normal synthesis parameters. B.2.1 Experimental methods Using the procedures outlined in section 4.3.2, Samantha systematically varied the synthesis parameters and characterized their effects on the resulting silica thin-film. She varied the water amount, the aging time, the spin speed of coating, the annealing temperature and the annealing time. The amount she varied each parameter is give in Table B-1. After synthesizing sol-gels with different recipes, Samantha characterized them using an optical microscope, SEM, ellipsometer, and FTIR spectrometer. 136 She compared the results from different recipes to the standard recipe, to each other, and to thermally grown silicon dioxide. Table B-1: Parameters varied independently for phase one Variable Description Control (standard) variations units A Water amount 1 0.5, 1, 1.5, 2, 2.5 x standard amount B Time aged 24 18, 24, 48, 72 Hours C Spin speed 7 3, 5, 7, 8 kRPM D Anneal temp. 1000 400, 600, 800, 1000, 1100 °C E Anneal time 1 0.5, 1, 1.5, 2, 2.5 hours During the first phase of this project, each parameter was varied independently. However, in the second phase, Samantha performed a five-dimensional co-variance analysis by co-varying the parameters in different permutations so she could perform a statistical analysis. For this phase, the interaction samples can be described by every permutation of two non-control values given in Table B-2. In total, between both phases, 175 different samples were synthesized and prepared. Table B-2: Parameters co-varied in different permutations for phase two Variable Description Low value High value units A Water amount 0.5 1.5 x standard amount B Time aged 18 30 Hours C Spin speed 5 8 kRPM D Anneal temp. 800 1100 °C E Anneal time 0.5 1.5 hours After synthesizing the samples, they were examined via microscope and FTIR for consistency and quality. Good samples had their refractive index at 830 nm, 635 nm and 405 nm and their thickness measured by the ellipsometer. The values for refractive index and thickness from good samples were used to produce regression models with MATLAB. Algorithms were used to fit the data to various models for each parameter varied using the co-variance analysis. Three continuous models were used for the thickness. For the refractive index, a discrete model was used. B.2.2 Data and results Some representative models for thickness co-varying two variables are shown in Figure B-3. The values of the statistical data for the refractive index at 633 nm are given in Table B-3. The refractive index for the other wavelengths the sol-gel films were tested at should follow the same trends as 635 nm. 137 Figure B-3: Graphs showing results of thickness models. For each graph, thickness (in angstroms) is shown as a function of a) time annealed and spin speed, b) temperature annealed and spin speed, c) spin speed and water amount and d) temperature annealed and water amount. By examining these figures, an idea of how thickness and refractive index vary with synthesis parameters can be developed. Generally, the following trends hold: increasing the amount of water decreases the thickness and slightly raises the refractive index, increasing the aging time decreases the thickness slightly and does not change the refractive index much, increasing the spin speed has a negligible effect on the thickness and refractive index, increasing the anneal temperature decreases the thickness and refractive index, and increasing the anneal time has a small effect on the thickness but increases the refractive index. However, as shown in the graphs in Figure B-3, co-varying relationships has a complex effect on the thickness. The relationship for the refractive index is complex as well. By using these ranges, though, recipes that yield appropriate values of refractive index and thickness can be chosen and utilized for future experiments. 138 Table B-3: Refractive index values at 633 nm. Red numbers indicate large standard deviations and ranges. B.2.3 Conclusions In conclusion, Samantha investigated in detail how various synthesis parameters affect silica sol- gel thin films. She found several trends using independent variables and ran a co-variance experiment and statistical analysis to develop more complex relationships between the different parameters. Her findings will be useful for future projects involving silica sol-gels, especially those where refractive index and thickness are important, but use of dopants is not desirable. 139 Appendix B References [1] O. Castany, S. Abbas, I. Hardy, M. Gadonna, and L. Dupont, "Polarization Splitter Made of a Polymer Waveguide Coupler With Localized Anisotropic Segments," IEEE Photonics Technology Letters, vol. 25, pp. 468-471, Mar 2013. [2] J. W. Kim, S. H. Park, W. S. Chu, and M. C. Oh, "Integrated-optic polarization controllers incorporating polymer waveguide birefringence modulators," Optics Express, vol. 20, pp. 12443- 12448, May 2012. [3] G. Nabil, W. F. Ho, and H. P. Chan, "Experimental study on the performance of a variable optical attenuator using polymer dispersed liquid crystal," Applied Optics, vol. 52, pp. E15-E21, Aug 2013. [4] H. S. Hsu, C. Cai, and A. M. Armani, "Ultra-low-threshold Er:Yb sol-gel microlaser on silicon," Optics Express, vol. 17, pp. 23265-23271, Dec 7 2009. [5] B. A. Rose, A. J. Maker, and A. M. Armani, "Characterization of thermo-optic coefficient and material loss of high refractive index silica sol-gel films in the visible and near-IR," Optical Materials Express, vol. 2, pp. 671-681, May 1 2012. [6] A. J. Maker, B. A. Rose, and A. M. Armani, "Tailoring the behavior of optical microcavities with high refractive index sol-gel coatings," Optics Letters, vol. 37, pp. 2844-2846, Jul 2012. [7] S. Mehrabani and A. M. Armani, "Blue upconversion laser based on thulium-doped silica microcavity," Optics Letters, vol. 38, pp. 4346-4349, Nov 2013. [8] R. L. Dumas, I. Tejedor-Tejedor, and M. A. Anderson, "Dependence of SiO2 gel structure on gelation conditions and sol reaction temperature as followed by FTIR and nitrogen adsorption measurements," Journal of Porous Materials, vol. 5, pp. 95-101, May 1998. [9] J. Livage and C. Sanchez, "SOL-GEL CHEMISTRY," Journal of Non-Crystalline Solids, vol. 145, pp. 11-19, Aug 1992. 140 Bibliography A. Abbasi, M. Noshad, R. Ranjbar, and R. Kheradmand, "Ultra compact and fast All Optical Flip Flop design in photonic crystal platform," Optics Communications, vol. 285, pp. 5073-5078, Nov 2012. B. M. Ahn, J. Kim, L. Ian, K. H. Rha, and H. J. Kim, "Mechanical Property Characterization of Prostate Cancer Using a Minimally Motorized Indenter in an Ex Vivo Indentation Experiment," Urology, vol. 76, pp. 1007-1011, Oct 2010. A. M. Armani, "Label-free, single-molecule detection with optical microcavities (August, pg 783, 2007)," Science, vol. 334, pp. 1496-1496, Dec 2011. P. Avouris, "Graphene: Electronic and Photonic Properties and Devices," Nano Letters, vol. 10, pp. 4285- 4294, Nov 2010. N. Bamiedakis, T. Hutter, R. V. Penty, I. H. White, and S. R. Elliott, "PCB-Integrated Optical Waveguide Sensors: An Ammonia Gas Sensor," Journal of Lightwave Technology, vol. 31, pp. 1628-1635, May 2013. J. F. Bauters, M. J. R. Heck, D. John, D. X. Dai, M. C. Tien, J. S. Barton, et al., "Ultra-low-loss high-aspect- ratio Si3N4 waveguides," Optics Express, vol. 19, pp. 3163-3174, Feb 2011. J. F. Bauters, M. J. R. Heck, D. D. John, J. S. Barton, C. M. Bruinink, A. Leinse, et al., "Planar waveguides with less than 0.1 dB/m propagation loss fabricated with wafer bonding," Optics Express, vol. 19, pp. 24090-24101, Nov 21 2011. B. W. Biggs, H. K. Hunt, and A. M. Armani, "Selective patterning of Si-based biosensor surfaces using isotropic silicon etchants," Journal of Colloid and Interface Science, vol. 369, pp. 477-481, Mar 1 2012. J. Calero, S. P. Wu, C. Pope, S. L. Chuang, and J. P. Murtha, "THEORY AND EXPERIMENTS ON BIREFRINGENT OPTICAL FIBERS EMBEDDED IN CONCRETE STRUCTURES," Journal of Lightwave Technology, vol. 12, pp. 1081-1091, Jun 1994. O. Castany, S. Abbas, I. Hardy, M. Gadonna, and L. Dupont, "Polarization Splitter Made of a Polymer Waveguide Coupler With Localized Anisotropic Segments," IEEE Photonics Technology Letters, vol. 25, pp. 468-471, Mar 2013. F. Cellini, S. Khapli, S. D. Peterson, and M. Porfiri, "Mechanochromic polyurethane strain sensor," Applied Physics Letters, vol. 105, p. 061907, 2014. F. Ceyssens, D. Witters, T. Van Grimbergen, K. Knez, J. Lammertyn, and R. Puers, "Integrating optical waveguides in electrowetting-on-dielectric digital microfluidic chips," Sensors and Actuators B- Chemical, vol. 181, pp. 166-171, May 2013. J. Y. Chen, Y. F. Liu, T. Ye, X. Xiang, X. H. Ji, and Z. K. He, "A novel droplet dosing strategy-based versatile microscale biosensor for detection of DNA, protein and ion," Sensors and Actuators B-Chemical, vol. 215, pp. 206-214, Aug 2015. J. C. Chou, J. L. Chen, Y. H. Liao, J. T. Chen, C. Y. Lin, J. W. Lin, et al., "Fabrication and Characteristic Analysis of a Remote Real-Time Monitoring Applied to Glucose Sensor System Based on Microfluidic Framework," IEEE Sensors Journal, vol. 15, pp. 3234-3240, Jun 2015. T. H. Chua and C. L. Chen, "FIBER POLARIMETRIC STRESS SENSORS," Applied Optics, vol. 28, pp. 3158- 3165, Aug 1989. A. Crespi, Y. Gu, B. Ngamsom, H. J. W. M. Hoekstra, C. Dongre, M. Pollnau, et al., "Three-dimensional Mach-Zehnder interferometer in a microfluidic chip for spatially-resolved label-free detection," Lab on a Chip, vol. 10, pp. 1167-1173, 2010 2010. D. Derickson, Fiber Optic Test and Measurement: Prentice Hall, 1998. C. Dongre, R. Dekker, H. Hoekstra, M. Pollnau, R. Martinez-Vazcluez, R. Osellame, et al., "Fluorescence monitoring of microchip capillary electrophoresis separation with monolithically integrated waveguides," Optics Letters, vol. 33, pp. 2503-2505, Nov 2008. 141 E. E. Dreger, G. Klein, G. Miles, L. Shedlovsky, and J. Ross, "Sodium Alcohol Sulfates.Properties Involving Surface Activity," Industrial & Engineering Chemistry, vol. 36, pp. 610-617, 1944/07/01 1944. R. Duer, R. Lund, R. Tanaka, D. A. Christensen, and J. N. Herron, "In-Plane Parallel Scanning: A Microarray Technology for Point-of-Care Testing," Analytical Chemistry, vol. 82, Nov 1 2010. R. L. Dumas, I. Tejedor-Tejedor, and M. A. Anderson, "Dependence of SiO2 gel structure on gelation conditions and sol reaction temperature as followed by FTIR and nitrogen adsorption measurements," Journal of Porous Materials, vol. 5, pp. 95-101, May 1998. D. Duval, A. Belen Gonzalez-Guerrero, S. Dante, J. Osmond, R. Monge, L. J. Fernandez, et al., "Nanophotonic lab-on-a-chip platforms including novel bimodal interferometers, microfluidics and grating couplers," Lab on a Chip, vol. 12, 2012 2012. G. L. Duveneck, A. P. Abel, M. A. Bopp, G. M. Kresbach, and M. Ehrat, "Planar waveguides for ultra-high sensitivity of the analysis of nucleic acids," Analytica Chimica Acta, vol. 469, pp. 49-61, Sep 2002. C. Eggeling, J. Widengren, L. Brand, J. Schaffer, S. Felekyan, and C. A. M. Seidel, "Analysis of photobleaching in single-molecule multicolor excitation and forster resonance energy transfer measurement," Journal of Physical Chemistry A, vol. 110, pp. 2979-2995, Mar 2006. D. Erickson, S. Mandal, A. H. J. Yang, and B. Cordovez, "Nanobiosensors: optofluidic, electrical and mechanical approaches to biomolecular detection at the nanoscale," Microfluidics and Nanofluidics, vol. 4, Jan 2008. H. Fu, S. Xu, R. Xu, J. Jiang, Y. Zhang, J. A. Rogers, et al., "Lateral buckling and mechanical stretchability of fractal interconnects partially bonded onto an elastomeric substrate," Applied Physics Letters, vol. 106, p. 091902, 2015. R. Gajasinghe, S. U. Senveli, S. Rawal, A. Williams, A. Zheng, R. H. Datar, et al., "Experimental study of PDMS bonding to various substrates for monolithic microfluidic applications," Journal of Micromechanics and Microengineering, vol. 24, p. 11, Jul 2014. B. S. Garra, "Imaging and estimation of tissue elasticity by ultrasound," Ultrasound Q, vol. 23, pp. 255- 68, Dec 2007. R. C. Gauthier and J. Dhliwayo, "BIREFRINGENT FIBEROPTIC PRESSURE SENSOR," Optics and Laser Technology, vol. 24, pp. 139-143, Jun 1992. T. Gensch, M. Bohmer, and P. F. Aramendia, "Single molecule blinking and photobleaching separated by wide-field fluorescence microscopy," Journal of Physical Chemistry A, vol. 109, pp. 6652-6658, Aug 2005. D. W. Good, A. Khan, S. Hammer, P. Scanlan, W. M. Shu, S. Phipps, et al., "Tissue Quality Assessment Using a Novel Direct Elasticity Assessment Device (The E-Finger): A Cadaveric Study of Prostatectomy Dissection," Plos One, vol. 9, p. 8, Nov 2014. D. W. Good, G. D. Stewart, S. Hammer, P. Scanlan, W. Shu, S. Phipps, et al., "Elasticity as a biomarker for prostate cancer: a systematic review," Bju International, vol. 113, pp. 523-534, Apr 2014. T. Gu, N. Petrone, J. F. McMillan, A. van der Zande, M. Yu, G. Q. Lo, et al., "Regenerative oscillation and four-wave mixing in graphene optoelectronics," Nature Photonics, vol. 6, pp. 554-559, Aug 2012. M. C. Harrison and A. M. Armani, "Spatiotemporal Fluorescent Detection Measurements Using Embedded Waveguide Sensors," IEEE Journal of Selected Topics in Quantum Electronics, vol. PP, 2014. M. C. Harrison and A. M. Armani, "Portable polarimetric fiber stress sensor system for visco-elastic and biomimetic material analysis," Applied Physics Letters, vol. 106, p. 191105, 2015. A. Himeno, K. Kato, and T. Miya, "Silica-based planar lightwave circuits," IEEE Journal of Selected Topics in Quantum Electronics, vol. 4, pp. 913-924, Nov-Dec 1998. H. Hoekstra and H. B. H. Elrofai, "Theory of optical spontaneous emission rates in layered structures," Physical Review E, vol. 71, Apr 2005. 142 M. Horimizu, T. Kawase, T. Tanaka, K. Okuda, M. Nagata, D. M. Burns, et al., "Biomechanical evaluation by AFM of cultured human cell-multilayered periosteal sheets," Micron, vol. 48, pp. 1-10, May 2013. K. Hoyt, B. Castaneda, M. Zhang, P. Nigwekar, P. A. di Sant'Agnese, J. V. Joseph, et al., "Tissue elasticity properties as biomarkers for prostate cancer," Cancer Biomarkers, vol. 4, pp. 213-225, 2008. H. S. Hsu, C. Cai, and A. M. Armani, "Ultra-low-threshold Er:Yb sol-gel microlaser on silicon," Optics Express, vol. 17, pp. 23265-23271, Dec 7 2009. H. K. Hunt and A. M. Armani, "Label-free biological and chemical sensors," Nanoscale, vol. 2, pp. 1544- 1559, 2010. H. K. Hunt and A. M. Armani, "Recycling microcavity optical biosensors," Optics Letters, vol. 36, p. 1092, 2011. M. K. Iyer, P. A. V. Ramana, K. Sudharsanam, C. J. Leo, M. Sivakumar, B. L. S. Pong, et al., "Design and development of optoelectronic mixed signal system-on-package (SOP)," IEEE Transactions on Advanced Packaging, vol. 27, pp. 278-285, May 2004. I. D. Johnston, D. K. McCluskey, C. K. L. Tan, and M. C. Tracey, "Mechanical characterization of bulk Sylgard 184 for microfluidics and microengineering," Journal of Micromechanics and Microengineering, vol. 24, p. 7, Mar 2014. T. F. Johnston, R. H. Brady, and W. Proffitt, "POWERFUL SINGLE-FREQUENCY RING DYE-LASER SPANNING THE VISIBLE SPECTRUM," Applied Optics, vol. 21, pp. 2307-2316, 1982. G. J. Kacprzynski, A. Sarlashkar, M. J. Roemer, A. Hess, and W. Hardman, "Predicting remaining life by fusing the physics of failure modeling with diagnostics," JOM, vol. 56, pp. 29-35, 2004. J. W. Kim, S. H. Park, W. S. Chu, and M. C. Oh, "Integrated-optic polarization controllers incorporating polymer waveguide birefringence modulators," Optics Express, vol. 20, pp. 12443-12448, May 2012. K. Kim, S. H. Cho, and C. W. Lee, "NONLINEAR OPTICS Graphene-silicon fusion," Nature Photonics, vol. 6, pp. 502-503, Aug 2012. L. C. Klein, "SOL-GEL OPTICAL-MATERIALS," Annual Review of Materials Science, vol. 23, pp. 437-452, 1993. C. Kopp, S. Bernabe, B. Ben Bakir, J. M. Fedeli, R. Orobtchouk, F. Schrank, et al., "Silicon Photonic Circuits: On-CMOS Integration, Fiber Optical Coupling, and Packaging," IEEE Journal of Selected Topics in Quantum Electronics, vol. 17, pp. 498-509, May-Jun 2011. T. A. Krouskop, T. M. Wheeler, F. Kallel, B. S. Garra, and T. Hall, "Elastic moduli of breast and prostate tissues under compression," Ultrasonic Imaging, vol. 20, pp. 260-274, Oct 1998. M. Lehnhardt, T. Riedl, T. Rabe, and W. Kowalsky, "Room temperature lifetime of triplet excitons in fluorescent host/guest systems," Organic Electronics, vol. 12, pp. 486-491, Mar 2011. J. Liang, E. Toussaere, R. Hierle, R. Levenson, J. Zyss, A. V. Ochs, et al., "Low loss, low refractive index fluorinated self-crosslinking polymer waveguides for optical applications," Optical Materials, vol. 9, pp. 230-235, 1998. G. K. Lim, Z. L. Chen, J. Clark, R. G. S. Goh, W. H. Ng, H. W. Tan, et al., "Giant broadband nonlinear optical absorption response in dispersed graphene single sheets," Nature Photonics, vol. 5, pp. 554-560, Sep 2011. I. K. Lin, K.-S. Ou, Y.-M. Liao, Y. Liu, K.-S. Chen, and X. Zhang, "Viscoelastic Characterization and Modeling of Polymer Transducers for Biological Applications," Journal of Microelectromechanical Systems, vol. 18; 16, pp. 1087-1099, 2009. A. H. Liu, T. Watanabe, I. Honma, J. Wang, and H. S. Zhou, "Effect of solution pH and ionic strength on the stability of poly(acrylic acid)-encapsulated multiwalled carbon nanotubes aqueous dispersion and its application for NADH sensor," Biosensors & Bioelectronics, vol. 22, pp. 694- 699, Dec 2006. 143 G. Liu and S. L. Chuang, "Polarimetric optical fiber weight sensor," Sensors and Actuators a-Physical, vol. 69, pp. 143-147, Aug 1998. M. Liu, X. B. Yin, E. Ulin-Avila, B. S. Geng, T. Zentgraf, L. Ju, et al., "A graphene-based broadband optical modulator," Nature, vol. 474, pp. 64-67, Jun 2 2011. J. Livage and C. Sanchez, "SOL-GEL CHEMISTRY," Journal of Non-Crystalline Solids, vol. 145, pp. 11-19, Aug 1992. N. Lloret, R. S. Frederiksen, T. C. Moller, N. I. Rieben, S. Upadhyay, L. De Vico, et al., "Effects of buffer composition and dilution on nanowire field-effect biosensors," Nanotechnology, vol. 24, p. 9, Jan 2013. M. J. Lochhead, K. Todorof, M. Delaney, J. T. Ives, C. Greef, K. Moll, et al., "Rapid Multiplexed Immunoassay for Simultaneous Serodiagnosis of HIV-1 and Coinfections," Journal of Clinical Microbiology, vol. 49, pp. 3584-3590, Oct 2011. N. S. Lu, C. Lu, S. X. Yang, and J. Rogers, "Highly Sensitive Skin-Mountable Strain Gauges Based Entirely on Elastomers," Advanced Functional Materials, vol. 22, pp. 4044-4050, Oct 2012. A. Majumdar, J. Kim, J. Vuckovic, and F. Wang, "Electrical Control of Silicon Photonic Crystal Cavity by Graphene," Nano Letters, vol. 13, pp. 515-518, Feb 2013. A. J. Maker and A. M. Armani, "Low-loss silica-on-silicon waveguides," Optics Letters, vol. 36, pp. 3729- 3731, Oct 1 2011. A. J. Maker, B. A. Rose, and A. M. Armani, "Tailoring the behavior of optical microcavities with high refractive index sol-gel coatings," Optics Letters, vol. 37, pp. 2844-2846, Jul 2012. I. H. Malitson, "INTERSPECIMEN COMPARISON OF REFRACTIVE INDEX OF FUSED SILICA," Journal of the Optical Society of America, vol. 55, pp. 1205-&, 1965. S. Mandal, J. M. Goddard, and D. Erickson, "A multiplexed optofluidic biomolecular sensor for low mass detection," Lab on a Chip, vol. 9, 2009 2009. A. Mata, A. J. Fleischman, and S. Roy, "Characterization of Polydimethylsiloxane (PDMS) Properties for Biomedical Micro/Nanosystems," Biomedical Microdevices, vol. 7, pp. 281-293, 2005. M. Medina-Sanchez, S. Miserere, and A. Merkoci, "Nanomaterials and lab-on-a-chip technologies," Lab on a Chip, vol. 12, pp. 1932-1943, 2012. S. Mehrabani and A. M. Armani, "Blue upconversion laser based on thulium-doped silica microcavity," Optics Letters, vol. 38, pp. 4346-4349, Nov 2013. S. Mehrabani, P. Kwong, M. Gupta, and A. M. Armani, "Hybrid microcavity humidity sensor," Applied Physics Letters, vol. 102, Jun 17 2013. T. Miya, "Silica-based planar lightwave circuits: Passive and thermally active devices," IEEE Journal of Selected Topics in Quantum Electronics, vol. 6, pp. 38-45, Jan-Feb 2000. S. R. Mousavi, A. Sadeghi-Naini, G. J. Czarnota, and A. Samani, "Towards clinical prostate ultrasound elastography using full inversion approach," Medical Physics, vol. 41, p. 12, Mar 2014. H. Mukundan, A. S. Anderson, W. K. Grace, K. M. Grace, N. Hartman, J. S. Martinez, et al., "Waveguide- Based Biosensors for Pathogen Detection," Sensors, vol. 9, Jul 2009. V. M. Murukeshan, P. Y. Chan, L. S. Ong, and A. Asundi, "Effects of different parameters on the performance of a fiber polarimetric sensor for smart structure applications," Sensors and Actuators a-Physical, vol. 80, pp. 249-255, Mar 2000. G. Nabil, W. F. Ho, and H. P. Chan, "Experimental study on the performance of a variable optical attenuator using polymer dispersed liquid crystal," Applied Optics, vol. 52, pp. E15-E21, Aug 2013. T. K. Noh, U. C. Ryu, and Y. W. Lee, "Compact and wide range polarimetric strain sensor based on polarization-maintaining photonic crystal fiber," Sensors and Actuators a-Physical, vol. 213, pp. 89-93, Jul 2014. 144 K. Okamoto, "Recent progress of integrated optics planar lightwave circuits," Optical and Quantum Electronics, vol. 31, pp. 107-129, Feb 1999. J. Ophir, I. Cespedes, H. Ponnekanti, Y. Yazdi, and X. Li, "ELASTOGRAPHY - A QUANTITATIVE METHOD FOR IMAGING THE ELASTICITY OF BIOLOGICAL TISSUES," Ultrasonic Imaging, vol. 13, pp. 111- 134, Apr 1991. D. Pan, H. Wei, and H. X. Xu, "Optical interferometric logic gates based on metal slot waveguide network realizing whole fundamental logic operations," Optics Express, vol. 21, pp. 9556-9562, Apr 2013. V. M. N. Passaro, C. de Tullio, B. Troia, M. La Notte, G. Giannoccaro, and F. De Leonardis, "Recent Advances in Integrated Photonic Sensors," Sensors, vol. 12, pp. 15558-15598, Nov 2012. M. J. Paszek, N. Zahir, K. R. Johnson, J. N. Lakins, G. I. Rozenberg, A. Gefen, et al., "Tensional homeostasis and the malignant phenotype," Cancer Cell, vol. 8, pp. 241-254, Sep 2005. L. Penuela, F. Wolf, R. Raiteri, D. Wendt, I. Martin, and A. Barbero, "Atomic force microscopy to investigate spatial patterns of response to interleukin-1beta in engineered cartilage tissue elasticity," Journal of Biomechanics, vol. 47, pp. 2157-2164, Jun 2014. L. Polerecki, J. Hamrle, and B. D. MacCraith, "Theory of the radiation of dipoles placed within a multilayer system," Applied Optics, vol. 39, Aug 1 2000. C. R. Pollock, Fundamentals of Optoelectronics. Chicago: Irwin, 1995. S. Romero-Garcia, F. Merget, F. Zhong, H. Finkelstein, and J. Witzens, "Silicon nitride CMOS-compatible platform for integrated photonics applications at visible wavelengths," Optics Express, vol. 21, pp. 14036-14046, Jun 2013. B. A. Rose, A. J. Maker, and A. M. Armani, "Characterization of thermo-optic coefficient and material loss of high refractive index silica sol-gel films in the visible and near-IR," Optical Materials Express, vol. 2, pp. 671-681, May 1 2012. C. A. Rowe, S. B. Scruggs, M. J. Feldstein, J. P. Golden, and F. S. Ligler, "An array immunosensor for simultaneous detection of clinical analytes," Analytical Chemistry, vol. 71, pp. 433-439, Jan 1999. A. Samani, J. Bishop, C. Luginbuhl, and D. B. Plewes, "Measuring the elastic modulus of ex vivo small tissue samples," Physics in Medicine and Biology, vol. 48, pp. 2183-2198, Jul 2003. A. Samani, J. Zubovits, and D. Plewes, "Elastic moduli of normal and pathological human breast tissues: an inversion-technique-based investigation of 169 samples," Physics in Medicine and Biology, vol. 52, pp. 1565-1576, Mar 2007. A. Scandurra, "Silicon Photonics: The System on Chip Perspective," Silicon Photonics Ii: Components and Integration, vol. 119, pp. 143-168, 2011. A. J. Shaywitz and M. E. Greenberg, "CREB: A stimulus-induced transcription factor activated by a diverse array of extracellular signals," in Annual Review of Biochemistry. vol. 68, C. C. Richardson, Ed., ed, 1999, pp. 821-861. S. Soltani and A. M. Armani, "Optimal design of suspended silica on-chip splitter," Optics Express, vol. 21, pp. 7748-7757, Mar 25 2013. A. Sulaiman, S. W. Harun, K. S. Lim, F. Ahmad, and H. Ahmad, "Microfiber Mach-Zehnder interferometer embedded in low index polymer," Optics and Laser Technology, vol. 44, pp. 1186-1189, 2012. B. E. A. S. a. M. C. Teich, Fundamentals of Photonics, 2nd ed. Hoboken, New Jersey: John Wiley & Sons, 2007. P. K. Tien, "LIGHT WAVES IN THIN FILMS AND INTEGRATED OPTICS," Applied Optics, vol. 10, pp. 2395-&, 1971. R. W. Tilghman, C. R. Cowan, J. D. Mih, Y. Koryakina, D. Gioeli, J. K. Slack-Davis, et al., "Matrix Rigidity Regulates Cancer Cell Growth and Cellular Phenotype," Plos One, vol. 5, p. 13, Sep 2010. Y. Verbandt, B. Verwilghen, P. Cloetens, L. VanKempen, H. Thienpont, I. Veretennicoff, et al., "Polarimetric optical fiber sensors: Aspects of sensitivity and practical implementation," Optical Review, vol. 4, pp. 75-79, Jan-Feb 1997. 145 J. Wang, "Electrochemical biosensors: Towards point-of-care cancer diagnostics," Biosensors & Bioelectronics, vol. 21, pp. 1887-1892, Apr 2006. A. L. Washburn and R. C. Bailey, "Photonics-on-a-chip: recent advances in integrated waveguides as enabling detection elements for real-world, lab-on-a-chip biosensing applications," Analyst, vol. 136, pp. 227-236, 2011. S. I. Wathugala, "Dependence of Silica Sol-gel Thin Film Material Properties on Fabrication Methods," in Intel International Science and Engineering Fair, Report Submitted to Palos Verdes Peninsula High School, Rolling Hills Estates, CA ed, 2014. P. N. T. Wells and H. D. Liang, "Medical ultrasound: imaging of soft tissue strain and elasticity," Journal of the Royal Society Interface, vol. 8, pp. 1521-1549, Nov 2011. L. C. Xiong, P. Chen, and Q. S. Zhou, "Adhesion promotion between PDMS and glass by oxygen plasma pre-treatment," Journal of Adhesion Science and Technology, vol. 28, pp. 1046-1054, Jun 2014. Q. Ye, J. Wang, Z. B. Liu, Z. C. Deng, X. T. Kong, F. Xing, et al., "Polarization-dependent optical absorption of graphene under total internal reflection," Applied Physics Letters, vol. 102, p. 4, Jan 2013. A. Y. a. P. Yeh, Optical Waves in Crystals. Hoboken, New Jersey: John Wiley & Sons, 2003. U. Zaleska-Dorobisz, K. Kaczorowski, A. Pawlus, A. Puchalska, and M. Inglot, "Ultrasound Elastography - Review of Techniques and its Clinical Applications," Advances in Clinical and Experimental Medicine, vol. 23, pp. 645-655, Jul-Aug 2014. X. Zhang, M. Harrison, A. Harker, and A. M. Armani, "Serpentine low loss trapezoidal silica waveguides on silicon," Optics Express, vol. 20, pp. 22298-22307, Sep 24 2012. X. M. Zhang and A. M. Armani, "Suspended bridge-like silica 2 x 2 beam splitter on silicon," Optics Letters, vol. 36, pp. 3012-3014, Aug 1 2011. H. Zhou, H. Zhang, Y. Pei, H.-S. Chen, H. Zhao, and D. Fang, "Scaling relationship among indentation properties of electromagnetic materials at micro- and nanoscale," Applied Physics Letters, vol. 106, p. 081904, 2015.

Abstract (if available)

Abstract Optical and photonic devices have numerous applications that range from telecommunications to various types of sensing. They are widely studied in a number of geometries and are often designed to accomplish a specific application. However, sometimes specific devices can be improved upon or intelligently designed to enable new applications. To accelerate this process, a combination of modeling and experimental efforts can be applied. The focus of this dissertation is the improvement and modification of existing waveguides and waveguide devices for new applications by the use of complementary modeling and experimental research to better understand their behavior. ❧ In the first part of the dissertation, fluorescent waveguide sensors are modeled and developed for use in spatiotemporal fluorescent measurements. The experimental measurements are enabled by the unique geometry of the devices which was better understood after modeling them. Subsequently, suspended, silica-on-silicon waveguides are explored in a new geometry and their behavior is modeled as well. Improvements to these devices are attempted via the use of conformal coatings of different materials. Next, a suspended silica-on-silicon waveguide splitter is developed for use as a biosensor, with models used to predict its sensing behavior. Finally, an optical fiber-based polarimetric stress sensor for use with visco-elastic materials is developed by improving upon previous work and generalizing the theoretical analysis and modeling of these types of polarimetric sensors.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Developing improved silica materials and devices for integrated optics applications

PDF

Integrated silicon waveguides and specialty optical fibers for optical communications system applications

PDF

Optical simulation and development of novel whispering gallery mode microresonators

PDF

Silicon integrated devices for optical system applications

PDF

Application of optical forces in microphotonic systems

PDF

Analysis and design of cleaved edge lithium niobate waveguide resonators

PDF

Development of optical devices for applications in photonic integrated circuit and sensing

PDF

Development of novel optical materials for whispering gallery mode resonator for nonlinear optics

PDF

Integration of practical high sensitivity whispering gallery mode resonator sensors

PDF

Titania and hybrid titania - silica sol-gel thin films and their applications in integrated optical devices

PDF

Active integrated photonic devices in single crystal LiNbO₃ micro-platelets and a hybrid Si-LiNbO₃ platform

PDF

Silicon-based photonic crystal waveguides and couplers

PDF

High-speed and reconfigurable all-optical signal processing for phase and amplitude modulated signals

PDF

Microdisk-waveguide system for the control of the polarization state of light

PDF

Nonlinear optical nanomaterials in integrated photonic devices

PDF

Optical modulators for integrated photonic systems

PDF

Fabrication and characterization of yoroidal resonators for optical process improvement

PDF

Biological and chemical detection using optical resonant cavities

PDF

High-resolution optical instrumentation for biosensing and biomechanical characterization

PDF

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

Asset Metadata

Creator Harrison, Mark Christopher (author)

Core Title Development of integrated waveguide biosensors and portable optical biomaterial analysis systems

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Electrical Engineering

Publication Date 09/10/2015

Defense Date 08/27/2015

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

Tag bio-optics,biosensing,integrated optics,integrated photonics,OAI-PMH Harvest,optics,photonics,waveguides

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Armani, Andrea M. (committee chair), McCain, Megan (committee member), Steier, William H. (committee member), Willner, Alan (committee member)

Creator Email markchar@usc.edu,markchristopherharrison@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-176415

Unique identifier UC11272484

Identifier etd-HarrisonMa-3890.pdf (filename),usctheses-c40-176415 (legacy record id)

Legacy Identifier etd-HarrisonMa-3890.pdf

Dmrecord 176415

Document Type Dissertation

Format application/pdf (imt)

Rights Harrison, Mark Christopher

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