UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Optical resonator sensors and systems Talebi Fard, Sahba 2015

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2015_may_talebifard_sahba.pdf [ 16.93MB ]
JSON: 24-1.0166188.json
JSON-LD: 24-1.0166188-ld.json
RDF/XML (Pretty): 24-1.0166188-rdf.xml
RDF/JSON: 24-1.0166188-rdf.json
Turtle: 24-1.0166188-turtle.txt
N-Triples: 24-1.0166188-rdf-ntriples.txt
Original Record: 24-1.0166188-source.json
Full Text

Full Text

Optical Resonator Sensors andSystemsbySahba Talebi FardB.Sc., Carleton University, 2006M.A.Sc., The University of British Columbia, 2008A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinThe Faculty of Graduate and Postdoctoral Studies(Biomedical Engineering)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)April , 2015c© Sahba Talebi Fard 2015AbstractSilicon-on-insulator (SOI)-based sensors are attractive for sensing applica-tions in environmental safety, oil and gas, medical research, and clinical ap-plications. Since these devices are typically developed using Complementarymetal-oxide-semiconductor (CMOS)-compatible multi-project-wafer (MPW)shuttles, they bring the potential for having sensing systems on chips (SSOCs),and for mass fabrication and low cost production. The objective of this thesisis to improve the sensitivity, accuracy, and repeatability of sensors fabricatedon the SOI platform. Such sensors have the potential to be the key compo-nents of an SSOC.One can increase the sensitivity of a resonator sensor by increasing theinteraction between the evanescent field of the guided mode and the ana-lyte. In this thesis, two methods for increasing this interaction in micro-ringresonator-based sensors are investigated: 1) using the transverse electric (TE)guided mode in ultra-thin strip waveguides and 2) using the quasi-transversemagnetic (TM) guided mode in thin strip waveguides. Using analyses andsimulations, micro-ring sensors were designed to be fabricated within theconstraints of a MPW CMOS-compatible process. Using the TE sensors, thetemperature-induced errors were reduced by a factor of three; and the TMsensors exhibited twice the sensitivity of the best SOI micro-ring resonator-based sensors reported to date.Moving towards the actual implementation of an SSOC, a system of sen-sors was design to correct for unwanted variations in the measurements. Thissystem drew on multivariate techniques to achieve improvements that re-sulted in measurements that were more repeatable and more accurate in theiiAbstractpresence of environmental variations. The capability of this system is in-vestigated by designing a cascade of previously developed micro-ring sensorswith various waveguide thicknesses. With this system of sensors, we achievedan R2 value of predictions over 0.996 in the presence of a 2 K temperaturedrift. This approach significantly improved the repeatability and reliabil-ity of the measurements in the presence of undesirable variations and drifts.In another move towards achieving an SSOC, integrating photodetectors inresonator sensors was investigated. To accomplish this, ion-implantation onmicro-ring sensors was used. Such integrated photodetector-sensors were de-signed, fabricated, and tested. Their measured sensitivities were within 90%of the expected values.iiiPrefaceThe following publications contribute to parts of the chapters of this thesis:• Sahba Talebi Fard, Samantha M. Grist, Valentina Donzella, ShonA. Schmidt, Jonas Flueckiger, Xu Wang, Wei Shi, Andrew Millspaugh,Mitchell Webb, Daniel M. Ratner, Karen C. Cheung, Lukas Chros-towski, “Label-free silicon photonic biosensors for use in clinical di-agnostics”, Proc. SPIE, Silicon Photonics VIII, 8629:86290914 (In-vited), SPIE OPTO, International Society for Optics and Photonics,2/02/2013.This was an invited conference paper. I was the lead author, pre-pared the manuscript and presented it at the SPIE conference. I haveperformed data analysis of the measurement results for the devices pre-sented in this invited conference paper.• Sahba Talebi Fard, Valentina Donzella, Shon A. Schmidt, JonasFlueckiger, Samantha M. Grist, Pouria TalebiFard, Yichen Wu, Rick J.Bojko, Ezra Kwok, Nicolas A. F. Jaeger, Daniel M. Ratner, and LukasChrostowski, ”Performance of ultra-thin SOI-based resonators for sens-ing applications.” Optics Express 22, no. 12 (2014): 14166-14179.The author identified the research problem and proposed methods toimprove, performed literature review and contributed the idea. The au-thor then calculated the parameters using simulations and analyticalrelations, designed the devices, and prepared the layout to be submit-ted for fabrication. Upon arrival of the fabricated devices, the authorivPrefaceperformed experiments to test and characterize the devices, analyzedthe measurement results and prepared the manuscript.• Sahba Talebi Fard, Kyle Murray, Michael Caverley, Valentina Donzella,Jonas Flueckiger, Samantha M. Grist, Edgar Huante-Ceron, Shon A.Schmidt, Ezra Kwok, Nicolas A. F. Jaeger, Andrew P. Knights, andLukas Chrostowski, ”Silicon-on-insulator sensors using integrated resonance-enhanced defect-mediated photodetectors.” Optics Express 22, no. 23(2014): 28517-28529.This paper demonstrates a ring resonator biosensor with an integratedion-implanted detector, which exploits the resonance enhancement ofthe ring for increased responsivity. For this project, we brought to-gether the expertise of two groups, one at McMaster university, andone at the University of British Columbia (UBC), where the McMastergroup’s expertise is in ion-implanted defect-mediated based photode-tectors, and UBC’s expertise is in resonator-based sensors. For thispaper/project, the author performed literature search and, using sim-ulations and analytical equations, calculated the appropriate param-eters to design the devices. The author contributed to the layout ofthe design to be submitted for fabrication. Upon arrival of the fab-ricated chips, the author designed and performed experiments to testand characterize the devices, analyzed the measurement results andprepared the manuscript.• Sahba Talebi Fard, et al., “Optimized Sensitivity of Silicon-on-InsulatorStrip Waveguide Resonator Sensor”, In review (2014).The author identified the research problem and proposed methods toimprove, performed literature review and contributed the idea. The authorthen calculated the parameters using simulations and analytical relations,designed the devices, and prepared the layout to be submitted for fabrication.Upon arrival of the fabricated devices, the author performed experimentsvPrefaceto test and characterize the devices, analyzed the measurement results andprepared the manuscript.For the following publication, I have contributed the sections on thin andultra-thin TE resonator sensors and TM racetrack resonator sensors.• S. Schmidt, J. Flueckiger, W. Wu, S. M. Grist, S. Talebi Fard, V.Donzella, P. Khumwan, E. R. Thompson, Q. Wang, P. Kulik, X. Wang,A. Sherwali, J. Kirk, K. C. Cheung, L. Chrostowski, and D. Ratner,“Improving the performance of silicon photonic rings, disks, and bragggratings for use in label-free biosensing”, Proc. SPIE, Biosensing andNanomedicine VII 9166, 91660M (2014).For the following publications, I have contributed to the data analysis andrelated methods.• Xu Wang, Jonas Flueckiger, Shon Schmidt, Samantha Grist, SahbaTalebi Fard, James Kirk, Matt Doerfler, Karen C. Cheung, DanielM. Ratner, and Lukas Chrostowski, “A silicon photonic biosensor usingphase-shifted bragg gratings in slot waveguide”, Journal of Biophoton-ics, 04/2013 2013.• Samantha M. Grist, Shon A. Schmidt, Jonas Flueckiger, ValentinaDonzella, Wei Shi, Sahba Talebi Fard, James T. Kirk, Daniel M.Ratner, Karen C. Cheung, and Lukas Chrostowski, “Silicon photonicmicro-disk resonators for label-free biosensing”, Optics Express, 21:79948006,03/2013 2013.For the following papers, I have done the pioneering work and initial inves-tigations, performing simulation, design, and layout:• Valentina Donzella, Ahmed Sherwali, Jonas Flueckiger, Sahba TalebiFard, Samantha M Grist, and Lukas Chrostowski, “Sub-wavelengthgrating components for integrated optics applications on SOI chips”,Optics Express, 22, no. 17 (2014/8/25): 21037-21050.viPreface• Valentina Donzella, Ahmed Sherwali, Jonas Flueckiger, Samantha MGrist, Sahba Talebi Fard, and Lukas Chrostowski, “Design and fab-rication of SOI micro-ring resonators based on sub-wavelength gratingwaveguides”, Optics Express (2014).viiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiiiList of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . xviAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . xviiDedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviii1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Silicon Photonics as Sensors . . . . . . . . . . . . . . . . . . . 21.2 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.3 Contributions and Some Applications . . . . . . . . . . . . . 61.4 Methodology and Collaborations . . . . . . . . . . . . . . . . 91.5 Organization of the Thesis . . . . . . . . . . . . . . . . . . . 112 Methodology and Review of the Sensors . . . . . . . . . . . 132.1 Sensor’s Figures of Merit . . . . . . . . . . . . . . . . . . . . 142.2 Methods and Materials . . . . . . . . . . . . . . . . . . . . . 19viiiTable of Contents2.2.1 Sensor Design . . . . . . . . . . . . . . . . . . . . . . 192.2.2 Layout Design . . . . . . . . . . . . . . . . . . . . . . 222.2.3 Fabrication . . . . . . . . . . . . . . . . . . . . . . . . 232.2.4 Experimental Setup . . . . . . . . . . . . . . . . . . . 242.2.5 Microfluidic Integration . . . . . . . . . . . . . . . . . 262.2.6 Sensor Characterization . . . . . . . . . . . . . . . . . 282.3 Sensor Designs and Calibration Results . . . . . . . . . . . . 322.3.1 Optical Sensor Designs . . . . . . . . . . . . . . . . . 332.3.2 Disk Resonators . . . . . . . . . . . . . . . . . . . . . 342.3.3 Strip Waveguide Bragg Grating . . . . . . . . . . . . . 362.3.4 Slot Waveguide Ring Resonator . . . . . . . . . . . . . 382.3.5 Slot Waveguide Bragg Grating . . . . . . . . . . . . . 382.4 Biological Results . . . . . . . . . . . . . . . . . . . . . . . . 402.5 Discussion and Analysis . . . . . . . . . . . . . . . . . . . . . 422.6 Conclusion and Future Work . . . . . . . . . . . . . . . . . . 453 Ultra-Thin Resonator Sensors . . . . . . . . . . . . . . . . . . 473.1 Introduction and Background . . . . . . . . . . . . . . . . . . 483.2 Design Methods and Analysis . . . . . . . . . . . . . . . . . . 503.2.1 Bulk Sensitivities of TE Resonator Sensors . . . . . . 543.2.2 Temperature Sensitivities of TE Resonator Sensors . . 563.3 Experimental Methods and Materials . . . . . . . . . . . . . . 563.4 Performance of the Ultra-Thin TE Resonator Sensors . . . . . 583.4.1 Sensitivity Analysis Results . . . . . . . . . . . . . . . 583.4.2 Q Factor and Intrinsic Limit of Detection (ILOD) . . 623.5 Conclusion and Discussion . . . . . . . . . . . . . . . . . . . . 654 Optimized Sensitivity . . . . . . . . . . . . . . . . . . . . . . . 684.1 Introduction and Background . . . . . . . . . . . . . . . . . . 694.2 Methods and Materials . . . . . . . . . . . . . . . . . . . . . 714.3 Theory: Waveguide’s Sensitivity . . . . . . . . . . . . . . . . 72ixTable of Contents4.4 Resonator’s Sensitivity - Effect of Dispersion . . . . . . . . . 744.5 Resonator’s Bulk Sensitivity . . . . . . . . . . . . . . . . . . . 764.6 Resonator’s Temperature Sensitivity - Simulation Results . . 804.7 Resonator’s Surface Sensitivity - Simulations and Experiments 814.8 Bio-Sensing Demonstration . . . . . . . . . . . . . . . . . . . 824.9 Discussion, Summary and Conclusion . . . . . . . . . . . . . 855 Integrated Photodetector Sensors . . . . . . . . . . . . . . . . 885.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 885.2 Resonator Sensors and Defect-Mediated Photodetectors . . . 895.2.1 Defect-Mediated Photodetectors . . . . . . . . . . . . 905.2.2 Evanescent Field-Based Resonator Sensors . . . . . . . 915.2.3 Defect-Mediated Photodetector Resonator Sensor . . . 925.3 Design Considerations and Layout . . . . . . . . . . . . . . . 945.4 Experimental Methods and Materials . . . . . . . . . . . . . . 985.4.1 Experimental Setup . . . . . . . . . . . . . . . . . . . 985.4.2 Reagents and Microfluidic Setup . . . . . . . . . . . . 995.5 Performance Analysis and Results . . . . . . . . . . . . . . . 1005.6 Analysis and Discussion . . . . . . . . . . . . . . . . . . . . . 1055.6.1 Performance Tradeoffs . . . . . . . . . . . . . . . . . . 1055.6.2 Discussion and Analysis of Device Performance . . . . 1065.7 Summary and Conclusion . . . . . . . . . . . . . . . . . . . . 1086 Differential Sensing System . . . . . . . . . . . . . . . . . . . . 1116.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 1126.2 System Design and Methods . . . . . . . . . . . . . . . . . . 1156.3 Characterization and Performance of the Sensors and System 1206.4 Discussions and Conclusions . . . . . . . . . . . . . . . . . . . 1257 Conclusions and Future Work . . . . . . . . . . . . . . . . . . 1297.1 Summary and Conclusion . . . . . . . . . . . . . . . . . . . . 129xTable of Contents7.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . 131Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133AppendicesA List of Publications . . . . . . . . . . . . . . . . . . . . . . . . . 147B Derivation of Sensitivity Formula . . . . . . . . . . . . . . . . 150xiList of Tables2.1 Measured refractive index (RI) of various concentrations ofsalt solutions, in refractive index unit (RIU) . . . . . . . . . . 302.2 Summary of the performance and characteristics of our siliconphotonics sensors . . . . . . . . . . . . . . . . . . . . . . . . . 433.1 Thermo optics coefficient of the materials in the system . . . . 575.1 Summary of the performances of designs A and B devices . . . 1036.1 Sensitivity vectors of a few sensors . . . . . . . . . . . . . . . 1276.2 Determinant of the sensitivity matrix for various combinationsof a system of two resonator sensors . . . . . . . . . . . . . . . 127xiiList of Figures1.1 Research and development cycle . . . . . . . . . . . . . . . . . 102.1 Q and S demonstrations . . . . . . . . . . . . . . . . . . . . . 172.2 Schematic representation of the cross-section of the waveguide. 202.3 TE convergence test (out-of-plane) . . . . . . . . . . . . . . . 212.4 TE convergence test (in-plane) . . . . . . . . . . . . . . . . . . 212.5 TM convergence test (out-of-plane) . . . . . . . . . . . . . . . 222.6 TM convergence test (in-plane) . . . . . . . . . . . . . . . . . 232.7 Schematic representation of our testing platform . . . . . . . . 242.8 Picture representation of coupling light into the chip . . . . . 252.9 Schematic representation of our testing platform . . . . . . . . 262.10 The experimental setup . . . . . . . . . . . . . . . . . . . . . . 272.11 SEM images of the fabricated devices . . . . . . . . . . . . . . 332.12 Mode profiles (TE and TM) for 10 µm disk resonator . . . . . 342.13 Spectral shift and sensitivity for a 3 µm radius disk resonator 352.14 Strip Bragg waveguide and the propagating TE mode . . . . . 372.15 Spectral shift and sensitivity for a strip Bragg resonator . . . . 38xiiiList of Figures2.16 Slot Bragg waveguide and the propagating TE mode . . . . . 392.17 Spectral shift and sensitivity for a slot Bragg resonator . . . . 402.18 Biosensing results for a disk and a Bragg resonator. . . . . . . 412.19 Summary of the performance of various sensors. . . . . . . . . 453.1 Simulation results for the cross-section of the SOI waveguides 513.2 Simulated evanescent field penetration depth (TE mode) . . . 523.3 Sensitivities of TE resonator sensors . . . . . . . . . . . . . . . 543.4 SEM image of ultra-thin TE resonator sensors . . . . . . . . . 573.5 Experimental bulk sensitivity results . . . . . . . . . . . . . . 593.6 Experimental temperature sensitivities . . . . . . . . . . . . . 613.7 Prediction results of glucose concentrations while temperaturevaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623.8 Prediction results of glucose concentrations at constant tem-perature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633.9 Performance summary of the sensors (simulation and experi-ment) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 644.1 Simulated evanescent field penetration depth (TM mode) . . . 704.2 Sensitivities of the waveguide as functions of slab thicknesses . 734.3 Surface Sensitivities of the waveguide as functions of slab thick-nesses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 744.4 Group indices of slab waveguides as functions of slab thick-nesses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75xivList of Figures4.5 Calculated/simulated resonator’s sensitivities. . . . . . . . . . 774.6 Contour plot of sensitivity in nm/RIU as functions of waveg-uide withs and thicknesses. . . . . . . . . . . . . . . . . . . . . 784.7 Sensitivities of TM sensors (simulation and experiment). . . . 794.8 Measured bulk sensitivities of TM sensors. . . . . . . . . . . . 804.9 Temperature sensitivities of TM waveguides. . . . . . . . . . . 804.10 Calculated resonator’s surface sensitivities . . . . . . . . . . . 824.11 Surface Sensing Experiment . . . . . . . . . . . . . . . . . . . 845.1 Schematic representations of proposed designs for sensing. . . 935.2 Schematic of detector cross-section . . . . . . . . . . . . . . . 955.3 Schematic presentation and microscope pictures of designs Aand B devices. . . . . . . . . . . . . . . . . . . . . . . . . . . . 965.4 IV curves for designs A and B. . . . . . . . . . . . . . . . . . . 1005.5 Measured electrical/optical responses of photodetector sensors 1015.6 Experimental sensitivity results for the photodetector sensor . 1046.1 Response of each sensor to concentration variations. . . . . . . 1226.2 Response of the system of sensors to concentration variations. 1236.3 Response of the sensors and the system to concentration vari-ations at random temperature. . . . . . . . . . . . . . . . . . . 124xvList of AbbreviationsBSA Bovine Serum AlbumenCMOS Complementary Metal-Oxide-SemiconductorEF Evanescent FieldELISA Enzyme-Linked Immunosorbent AssayEMI Electromagnetic InterferenceGC Grating CouplerIME Institute of MicroelectronicsLOC Lab On a ChipLOD Limit Of DetectionMPW Multi Project WaferPBS Phosphate Buffered SalinePDMS polydimethylsiloxaneSEM Scanning Electron MicroscopeSOI Silicon On InsulatorSPR Surface Plasmon ResonanceSSOC Sensing System On ChipTE Transverse ElectricTM Transverse MagneticxviAcknowledgementsI would like to acknowledge Profs. Lukas Chrostowski, Ezra Kwok, and Nico-las Jaeger for their support, guidance and encouragements throughout mygraduate studies. I give thanks to the rest of my thesis committee membersfor their invaluable comments and support during my PhD program. Fur-ther, I would like to acknowledge the University of British Columbia, NSERCas well as the BC Innovation Council for their financial support.A very special gratitude and love goes to my parents, Sirous and Nahid,who are the embodiments of determination and selflessness, for their all-embracing support, prayers, and abiding love that words cannot express. Itis impossible to imagine any of my achievements without them. I furtheracknowledge them, and all my teachers, for planting the seeds of love ofscience, humbleness, and striving for excellence in my heart. They taughtme to apply my knowledge to solve real world problems that effect humanity.I acknowledge my spouse for his insightful and invaluable scientific commentsas well as his loving care and emotional support. I am thankful for having asupportive partner with whom I can discuss interesting scientific matters, realworld problems, and service to humanity. Particular appreciation goes to mybrothers who were my best friends and the point of trust and reliance, and allmy close friends and families, who never stopped supporting and encouragingme. I thank my parents-in-law for their understanding and prayers, and mygrand parents for their continuous thoughts of love and well wishes. At theend, I like to appreciate my sister for her beautiful smile that is always in mymind and enlightens my path. I am blessed and thankful for having peoplearound me who shower me with their love and care.xviiDedicationTo my parents, Nahid and Sirous, who have sacrificed their lives with endlesslove, support, and kindnessTo my spouse, my eternal partner and friend, who brought me the real-life experience of the joy of love and careTo my siblings, who have always been there for me“Arts, crafts and sciences uplift the world of being, and are conducive toits exaltation. Knowledge is as wings to man’s life, and a ladder for his as-cent. Its acquisition is incumbent upon everyone. The knowledge of suchsciences, however, should be acquired as can profit the peoples of the earth,. . . .” –Baha’u’llah“Regard man as a mine rich in gems of inestimable value. Education can,alone, cause it to reveal its treasures, and enable mankind to benefit there-from.” –Baha’u’llahxviiiChapter 1IntroductionThe healthcare efficiency and costs is an ever-increasing public issue. Moreeasily accessible diagnosis and monitoring devices can significantly improvethe outcomes. The current diagnostic gold-standard for biological analysisand protein quantification is Enzyme-Linked Immunosorbent Assay (ELISA)[1]. This method is highly sensitive [2], but requires trained operators andconsumption of significant volumes of costly reagents and enzyme-labeledantibodies for optical detection, in well-equipped facilities. Therefore, themethod is expensive, time consuming, and bound to sophisticated labs, whichlimits the accessibility and usage of this method for frequent monitoring andearly detection, in home healthcare settings or rural areas. Researchers havefocused on development of nano- and micro-scale sensors [2–4], with ELISA-like sensitivities and capabilities for Lab on Chip (LOC) applications, toprovide affordable diagnostics to public and leverage the cost benefits ofearly detection [5].Researchers have investigated and compared the performances of varioussensing methods to ELISA [2]. Silicon photonics-based resonator sensors notonly match and exceed the performances of ELISA [6], they have advantages11.1. Silicon Photonics as Sensorssuch as immunity from electromagnetic interference (EMI) and mechanicalforces. Genalyte’s system is a larger scale example of such technology thatis currently being used for basic biomedical research [7–11]. However, thissystem would have been more beneficial and widely used if the sensors hadhigher sensitivity and larger evanescent field penetration depths to detectlarger molecules or molecules further away from the surface. The other ad-vantage of silicon photonics-based sensors (critical to their commercializa-tion) is that their fabrication processes are compatible with CMOS circuitsproviding the opportunity for integration of electronics on chip as well asleveraging the fabrication economies of scale.1.1 Silicon Photonics as SensorsJust as electronic integration has dramatically changed everyday life (e.g.with tablets, smart phones, laptops and all kinds of electronic devices), pho-tonics integration is quickly evolving and revolutionizing several fields, fromoptical intra- and inter- chip connections [12], modulators and filters [13, 14],and high speed telecommunications [15], to environmental monitoring [16]and healthcare [17]. The fabrication of silicon photonics chips could exploitcurrent CMOS foundries, resulting in a great number of potential advantages,including integration with control electronics, availability of electro-optic de-vices, and lowered cost facilitated by large scale production and leveragingof existing electronic facilities.21.1. Silicon Photonics as SensorsOne of the most promising applications of silicon photonics is in the fieldof sensors and biosensor such as environmental and healthcare applications.The environmental monitoring application of these sensors include water andair quality monitoring, as well as oil and gas sensing. The biosensing appli-cations of these sensors range from basic medical research [8] to bioterrordetection [18, 19] to smart home healthcare [20] diagnostics. Sensitive, reli-able, and inexpensive silicon photonic sensors integrated with microfluidicsand electronics could lead to the development of integrated electro-optic mi-crofluidic chips (Sensing System on a Chip, SSOC, or Lab on a Chip, LOC).These chips will eventually be able to perform completely automated bio-logical and sensing analysis and present numerous advantages, such as smallsample volume requirement, portability, and ease of use. Silicon photonicsis poised to revolutionize biosensing applications, specifically in medical di-agnostics. In principle, we can have either disposable or partially disposablesensors depending on the application and the the associated safety and costs.The core component of a SSOC or LOC is the sensing device, which allowsthe detection and quantification of target molecules. Several optical sensingmechanisms have been proposed and developed so far [21, 22]; in particularSurface Plasmon Resonance (SPR) has been exploited in well-developed com-mercial devices [23]. Those commercial devices present some drawbacks andonly a very small part of the device is actually integrated on a chip. One wayto improve current sensing mechanisms is using silicon nanophotonic opti-cal resonators that offer smaller footprints [24] while keeping high sensitivity31.1. Silicon Photonics as Sensors(comparable with commercial devices [25]). The resonant condition providesenhanced light - functionalized area interaction because the light travels mul-tiple times and resonates within the sensing volume, and therefore interactsmultiple times with the analyte, causing the sensor response to be amplified[26]. Evanescent field (EF) optical sensors exploit the interaction betweenthe evanescent field of the waveguide mode and the material surrounding thewaveguide core (e.g. analyte in the cladding medium) for detection. AmongEF optical sensors, waveguide-based ring [7, 9, 27], disk [28], photonic crys-tal [29], and Bragg grating [30] resonators have shown excellent potential forsensing applications as well as for SSOC and LOC integration.Our group is working on characterizing and comparing various types ofresonator-based EF optical sensors. Most of the silicon photonics chips thatare fabricated on a Silicon-on-Insulator (SOI) wafer are waveguide-basedsensors. The high refractive index of silicon confines the light in planarnanoscaled waveguides. Due to the high index of Si (nSi is about 3.46 at1550 nm, while nSiO2 is about 1.43), the light is well confined in the siliconwaveguide, resulting in low propagation and bending losses for the TE polar-ization [31]. Nevertheless, some modal energy propagates in the surroundingmedia and interacts with the analyte. The overlap of electric field and ana-lyte is dictated by the shape of the propagating optical mode and determinesthe sensitivity of the device. Their basic working mechanism is the detec-tion of a change in the effective refractive index for light propagation in thestructure [32], due to a variation of analyte concentration (bulk sensing) or41.1. Silicon Photonics as Sensorsbinding of molecules on the waveguide functionalized surfaces (surface sens-ing). The addition of a resonant structure allows for sensing this refractiveindex change by a shift in the resonance wavelength, providing a measurableresponse.The geometry (ring, disk, Bragg grating, as well as waveguide type andsize) and polarization of light (TM vs TE) can be optimized to design thesensing element for a given analyte and application. For example, the TMand TE have different distributions of the field in the waveguide and outsideof it, so they can potentially allow for sensing particles with different sizesor distances from the sensor surface [33]. The field distribution for the TMmode has more evanescent field extending into the surrounding media (andthus in the analyte) so it can potentially provide higher sensitivity and it canallow sensing of particles farther from the waveguide surface, for exampleattached to long functionalization chemistries.In our group, the performance and sensing capabilities of these deviceswere characterized. Comparing sensing performance across these devices willhelp validate architectures suitable for the application at hand. The mostpromising sensors for each application will then be identified for further studyand development. We discuss the sensors’ comparative advantages for sensingapplications and provide an outlook for future work in this field.51.2. Objective1.2 ObjectiveThe objectives of this thesis are to investigate and design novel optical sen-sors and systems of sensors (as the core of SSOC) on the silicon platform andto explore their capabilities for on-chip sensing systems. It is vital to havesensitive, accurate, repeatable, and reliable sensors at the core of a reliablesystem. The long-term objective of this work is to use these sensors and sys-tems of sensors for development of lab-on-chip and further optimize them forvarious applications, such as biomedical and environmental sensing systems.1.3 Contributions and Some ApplicationsIn this thesis,• Methods to improve the sensitivity of the sensors as single devices isproposed and demonstrated.• Optimum waveguide dimension to achieve the highest sensitivity isdemonstrated analytically and experimentally.• One new method to integrate a detector with a sensor is investigated.• Novel design of a system of sensors, mimicking multivariate techniques,to correct for unwanted variations and to accomplish a more repeatableand accurate measurement in the presence of environmental variations,is demonstrated and achieved.61.3. Contributions and Some ApplicationsThere are numerous applications for this work in the sensing field. Somepossible applications of this work include, but not limited to: medical andclinical diagnostic and/or monitoring, research and development, and envi-ronmental safety.These sensors can be used in medical and clinical application for patientsthat require real-time monitoring, such as glucose monitoring for diabetespatients, as well as patients that require less frequent monitoring, such asTSH for patients with thyroid deceases or CTNI for people with Cardiovas-cular Diseases (CVD) who are prone to Myocardial Infarction (MI) or heartattack. Their small size, immunity from electromagnetic interference, sensi-tivity to adsorbed bimolecular layers at their surface, and compatibility withestablished, high volume CMOS foundry processes make them an attractivetechnology for lab-on-chip applications [17, 21, 34–36]. This makes themuseful for smart home healthcare [20] diagnostics.These sensors are also promising for basic medical research and diagnosis[8] since they are sensitive to adsorbed bimolecular layers at their surface.Genalyte’s diagnostic platform leverages a TE mode silicon photonic ringresonator and has demonstrated significant progress towards realizing thefirst commercially available, highly multiplexed, silicon photonic biosensorfor clinical and research use [7–11]. There still exist many medical diagnosticapplications where TE mode ring resonators cannot achieve the sensitivityrequired for a definitive clinical diagnosis without secondary amplification[7, 37]. Therefore, developing sensors with higher sensitivities and larger pen-71.3. Contributions and Some Applicationsetration depths is highly desirable. A TM-mode based sensor, for example,provides larger penetration depths, allowing for sensing molecules attach-ing to long functionalized chemistries further from the surface of the sensor.An example is glycoprotein receptor molecules that play an important rolein initiating cellular binding and communication. These membrane-boundproteins, which can be immobilized on a silicon photonic biosensor, oftenbind small ligands (10-15 nm in size) or in the case of bacteria, the adhesinsat the end of their long fimbria. Since fimbria are often 100’s nm long, asensor capable of detecting bound mass several 100’s nm from the sensor’ssurface is needed, especially for bacterial adhesion and other cellular diag-nostic applications. Therefore, silicon photonic biosensors with Quasi-TMguided mode, which naturally extend their sensing field 100’s of nm from thesensor’s surface, are ideally suited for these kinds of biosensing applications.Furthermore, these sensors can be used for water and air quality monitor-ing applications, as well as oil and gas sensing. Examples include detection ofthe presence of Organophosphorus (OP) in water, that is extremely harmfulto human, or detection of various toxic molecules in the air. Organophos-phorus (OP) pesticides have been used as insecticide on broad range of crops(vegetables, fruits, etc). Although designed to kill pests, but OP pesticidescan be extremely harmful to humans due to their significant contribution tocancer mortality [38], and their potential in effecting humans’ nervous systemat low concentrations (they inhibit the nervous system enzyme, AChE [39]).81.4. Methodology and Collaborations1.4 Methodology and CollaborationsTo achieve the objectives, it was required to go through the research anddevelopment cycle multiple times. Each research and development cycle con-sists of:1. Define research problem through observations, experience and litera-ture reviews.2. Propose methods and designs to address the problem.3. Simulation and analysis of the design to predict its feasibility and per-formance, and to calculate and optimize the parameters.4. Design of the mask layout based on the process design kit (PDK) pro-vided by the foundry services.5. Design is sent for fabrication6. Measurement and tests7. Analysis of the results and characterization of the devices8. Optimizing and improving the design based on the results and repeatingthe cycle9. Publication of the results.Fabrication of the devices are achieved through Multi Project Wafer(MPW) runs offered by commercial CMOS-photonics fabrication facilities,91.4. Methodology and CollaborationsResearch ProblemPropose MethodsIdeasSimulationCalculationDesignMask DesignMeasurementsIdeas to Optimize the DesignAnalysisCharacterizationDesign VerificationFabricationPublicationsFigure 1.1: Schematic representation of our research and development cycle.such as The Institute of Microelectronics (IME) foundry in Singapore, andIMEC [40]. For initial investigations, we usually used E-Beam Lithography(EBL) System at the University of Washington - Washington Nanofabrica-tion Facility (UW WNF) that has a quick turnaround time.Design and performance of biomolecular binding experiments are donein collaborations with Dr. Ratner’s group at the University of Washington(UW) in Seattle. Specifically, the sandwich assays explained in section 2.2.6are designed at UW, and the experiments incorporating these assays areperformed on their site.101.5. Organization of the ThesisThe design and fabrication of the microfluidic channels in poly dimethyl-siloxane (PDMS) was performed by our collaborator’s group at UBC (Dr.Karen Cheung).The work on the integrated photodetector sensor (chapter 5) was in col-laboration with McMaster university. The layouts of the designs were dis-cussed with Prof. Andrew Knights, and the ion-implantations of the deviceswere performed at McMaster university.1.5 Organization of the ThesisTo achieve the objectives, this PhD thesis is organized into following chapters.Chapter 2 includes an overview of some of the sensors developed in ourgroup. This chapter also includes the figures of merit for evaluating theperformance of the sensors, as well as detailed materials and methods usedthroughout this thesis. Sensors with higher sensitivities to the refractive in-dex changes of the analyte, or molecular adsorption, are more desirable formost sensing applications. Therefore, in the rest of this thesis, we proposeand investigate methods to improve and optimize the sensitivity of the sensor(chapters 3 and 4). One novel method of integrating a detector with a sensordevice is investigated (chapter 5). This work was done in collaboration withMcMaster University. In addition, a system design of multiple sensors capa-ble of correcting for some ‘unwanted’ common changes and drifts is proposedand investigated (chapter 6). At the end, chapter 7 concludes the thesis with111.5. Organization of the Thesisa brief summary as well as discussions and suggestions for future researchdirections.12Chapter 2Methodology and Review ofthe Sensors 1Our group has worked on developing various SOI-based resonator sensorssuch as TE and TM disk [28], slot ring, Bragg grating [30], and sub-wavelengthgratings (SWG). These devices have been investigated for wavelengths aroundλ=1550 nm (conventional wavelength window in fiber-optic communication)and λ=1220 nm, where the water absorption is greatly decreased, offeringimproved limits of detection. Each one of these have advantages as well asdrawbacks, for example, devices such as slot waveguide Bragg grating sen-sors have shown high sensitivities and high quality factors and may presentadvantages for specific biosensing applications, however they are usually longresulting in larger footprints; also cascading these devices is challenging.To compare the performance of these sensors, we use a set of criteria thatwill be explained in section 2.1.To characterize the device and measure the performance, we used an1Parts of of this chapter have been published in [34]:S. Talebi Fard, et al., “Label-free silicon photonic biosensors for use in clinical diagnostics”,Proc. SPIE, Silicon Photonics VIII, 8629:86290914 (Invited), SPIE OPTO, InternationalSociety for Optics and Photonics, 2/02/2013.132.1. Sensor’s Figures of Meritautomated test setup, which is explained in section 2.2.In section 2.3, we compare the performance of different optical resonatordevices, i.e. rings, disks and slot-waveguide Bragg gratings, that have beendesigned and fabricated, using electron beam (e-beam) lithography rapidprototyping as well as standard CMOS foundry fabrication processes on SOIchips.2.1 Sensor’s Figures of MeritTo compare the performance of the devices, the set of criteria is included inthis section [17, 34, 35, 41, 42].In evanescent field (EF) sensors, changes in refractive index of the cladding(resulting from molecular binding events or concentration variation) changethe effective index of the propagating mode. The ratio of the change in effec-tive index of the propagating mode to the changes in refractive index of thecladding medium, is defined as the mode sensitivity, Smode, or waveguide’sbulk sensitivity:Smodebulk =δneff(ncl, nco, nbox, ω)δncl=neff(ncl + δncl, nco, nbox, ω)− neff(ncl, nco, nbox, ω)δncl(2.1)where neff is the effective index of the waveguide, ncl is the refractive indexof the cladding medium (including the biomolecules of interest), nco is therefractive index of the waveguide core material, nbox is the refractive index142.1. Sensor’s Figures of Meritof the buried oxide (BOX) material, and ω = 2pif is the optical frequency.This change in the effective index, in turn, results in a change in the resonantwavelength(s) of the resonator, which can be measured.Waveguide’s (or modal) surface sensitivity is defined as the change ineffective index of the waveguide as biomolecules adsorb to the surface of thewaveguide’s core:SmodeSurf =δneffδt(2.2)where t is the thickness of the bimolecular layer adsorbed to the surface.The sensor’s refractive index sensitivity is defined as the shift in resonantwavelength as a function of the change in refractive index of the cladding[17, 34, 43]:S =∆λres∆ncl=λresngδneffδncl(2.3)where λres is a resonant wavelength, and ng is the group index. The wave-length shift is the result of contributions from various factors:1. The change in the refractive index of the cladding medium (∆ncl).2. The effect of material and waveguide dispersions (ng): due to the wave-length dependence of the effective index, a change in wavelength resultsin a further change in the effective index. ng takes this effect into ac-count.3. As the index changes, the mode profiles slightly change, and thereforethe mode’s effective index and sensitivity ( δneffδncl ) changes.152.1. Sensor’s Figures of MeritWhen comparing different wavelength resonators, normalized sensitivity(S ′) is used:S ′ =Sλres=1λres∆λres∆ncl(2.4)Surface sensitivity of a resonator is defined as:Ssurf =∆λres∆t=λresngδneffδt(2.5)Another important property of optical waveguide resonator-based sensorsis their quality factor, Q, defined as the number of optical oscillations untilthe resonating energy decays to 1/e of its maximum value. The higher thenumber of these oscillations, the more light interacts with analyte. Highquality factor means improved minimum detectable wavelength shift (leadingto improved limit of detection, equation 2.9). The intrinsic quality factor (Qi)of a resonator is defined and approximated as [17, 34]:Qi = ωrEdE/dt= ωrτp =ωrα[m−1] ×cng=2pi × ngλres × α[m−1]=2pi × ng × 4.34λres × α[dB/m](2.6)where α is the loss in the resonator and ωr = 2pifr is the resonant frequency.For a critically coupled resonator:Q =Qi2(2.7)162.1. Sensor’s Figures of Merit1549 1549.5 1550 1550.5 1551 1551.5−30−25−20−15−10−5−30Wavelength (nm)Power (dBm)3$dB$bandwidth$ ΔλresΔλ3dB λres2λres1 ncl +Δnclfor$nclfor$Figure 2.1: Schematic demonstration of Q value and sensitivity (S) of a resonatorsensor. The red spectrum is the shifted version of the blue spectrum correspondingto a change in the refractive index of the cladding (δncl).Experimentally, we can measure Q:Q ≈λres∆λ3dB(2.8)where ∆λ3dB is the 3dB bandwidth of the resonance. A higher Q indicates asharper resonance. Figure 2.1 schematically demonstrates how Q and S arecalculated from the given spectra.Since shifts in the resonant wavelength of sharper peaks are easier to de-tect, both the resonator’s Q and sensitivity impact its sensing ability. Theabsolute limit of detection (minimum detectable refractive index change incladding) of a resonator-based optical biosensor is affected by both the in-172.1. Sensor’s Figures of Merittrinsic characteristics of the resonator and the characteristics of the mea-surement system (thermal stability, optical input laser noise, etc), both ofwhich should be optimized. To compare different resonators, independentof their operating wavelength and system optimizations, we have used thepreviously defined intrinsic limit of detection (ILOD) [17, 34, 42]. This takesinto account bulk sensing characteristics of the resonators.The minimum refractive index unit change that can be detected by theresonator, not taking into account the other system components (eg. laser,readout hardware), is called Intrinsic Limit of Detection (ILOD) as definedby the equation 2.9.∆nmin = ILOD =λresQS=1QS ′(2.9)However, researchers have demonstrated that the system’s LOD can betypically 100 times better than iLOD [42]. We have utilized commonly usedmethods such as R2 and standard deviation to compare systems’ perfor-mances. From the system’s point of view, there are challenges, as describedin chapter 6, and there are various sources of error that can impact thesystem’s performance and contribute to the noise. These sources of errorinclude, but not limited to, laser noise (amplitude relative intensity noise,phase noise, wavelength accuracy and repeatability, laser wavelength drift),detector thermal noise, shot noise, thermal fluctuations and thermal gradi-ents on the chip, etc. Chapter 6 identifies thermal variations as being the182.2. Methods and Materialsdominant contributor to error and can be reduced significantly using oursystem’s design.2.2 Methods and MaterialsThis section will include details of the design and fabrication of the sensors,the experimental setup, and the reagents used for sensor characterization andbio-experiments.If there are slight variations to these for any specific chapter, it will beincluded in a the methods section of that chapter.2.2.1 Sensor DesignTo design the parameters of the various sensor types, simulations were con-ducted using Lumerical’s MODE (fully-vectorial 2D eigenmode calculations),as well as analytical modelling in MATLAB.Lumerical MODE solutions was used to calculate the effective index of thewaveguide. Figure 2.2 shows the cross-section of the waveguide schematically.Simulation AreaTo determine the sensitivity, we calculated the change in neff of the waveg-uide as the refractive index of the cladding changes. To observe these changesand also to have reasonably accurate estimation of neff , we aim for errorsin the orders of or less than 10−5. This value also ensures that the mode192.2. Methods and Materials!!!!!!!Cladding!(H2O)! !!Core!(Si)!!!!!!!!BOX!(SiO2)!Width! Thickness!2!–!3!μm!Figure 2.2: Schematic representation of the cross-section of the waveguide.perturbation due to boundary values is minimal.To determine the approximate dimensions for simulation area, the error inneff is estimated as a function of the dimensions of the simulation area, basedon the method explained in [44]. We first determine the simulation span forthe out-of-plane direction by calculating the neff of the slab waveguide asthe simulation span increases (figure 2.3). We then use the neff of the slabto calculate the neff in the in-plane direction, and to determine the error inneff as the simulation span in the in-plane direction increases (figure 2.4).We have performed these convergance tests for various waveguide thick-nesses under study. As the waveguides get thinner, the mode is less confined(this is the cause of higher sensitivity). Therefore, larger simulation spansare required to get the desired error.Our investigation of TM mode is for waveguide core thicknesses of 220,and 150 nm. To estimate the required simulation area for TM calculations,202.2. Methods and Materials024622. (µm)neff  220 nm 150 nm 90 nm(a)024610−1510−1010−5100Span (µm)error  220 nm 150 nm 90 nm(b)Figure 2.3: Estimated error in simulating effective index for the first TE mode assimulation span increases, out-of-plane, for various waveguide thicknesses.02461.71.81.922. (µm)neff  220 nm 150 nm 90 nm(a)01234567810−1510−1010−5100Span (µm)error  220 nm 150 nm 90 nm(b)Figure 2.4: Estimated error in simulating effective index for the first TE mode assimulation span increases, in-plane, for various waveguide thicknesses.212.2. Methods and Materialsthe area on the top and bottom (out-of-plane simulation span) is calculated,then the neff of the slab is used to calculate the required area on two sides(in-plane simulation span). Figure 2.5 and 2.6 shows plots of the calculatederror as the simulation area in these two directions increase.024681012141. (µm)neff  220 nm 150 nm(a)0246810121410−1510−1010−5100Span (µm)error  220 nm 150 nm(b)Figure 2.5: Estimated error in simulating effective index for the first TM mode assimulation span increases, out-of-plane, for various waveguide thicknesses.2.2.2 Layout DesignLight is coupled into and out of the chip using grating couplers (GCs) [45].The sensors designed for biological applications were all aligned towards thecenter of a large chip (e.g. 16 × 25 mm) containing multiple designs, to allowfor poly dimethylsiloxane (PDMS) bonding on chip. This design facilitatesthe exposure of the sensors to various reagents, when the microfluidic chan-nels in PDMS are aligned to the sensors. On the chip, PDMS channels arereversibly bonded, in order to allow the flow of liquids and biological fluids222.2. Methods and Materials0246810121416181. (µm)neff  220 nm 150 nm(a)02468101214161810−1510−1010−5100Span (µm)error  220 nm 150 nm(b)Figure 2.6: Estimated error in simulating effective index for the first TM mode assimulation span increases, in-plane, for various waveguide thicknesses.on the top of the devices. Sometimes, sensors are fabricated with referencesensors, which are not subjected to change in analyte concentration or so-lution (they can be under PDMS, or in a different channel), thus they canallow correction for temperature drift or other effects.2.2.3 FabricationOptical devices were fabricated on SOI chips using various foundry services.For initial investigations, we usually used an E-Beam Lithography (EBL)System at the University of Washington - Washington Nanofabrication Fa-cility (UW WNF) that has a quick turnaround time. In addition, variousfoundries with multi-project wafer (MPW) services such as IMEC [40] andIME in Singapore were used.232.2. Methods and Materials2.2.4 Experimental SetupOur automated measurement setup is currently capable of automaticallymeasuring hundreds of devices within only a few hours. Figure 2.7 showsthe schematics of our setup.Figure 2.7: Schematic representation of our testing platform [courtesy of UW team]The setup consists of a tunable laser (Agilent 81682A, Agilent Technolo-gies, Inc., USA) as the optical source that operates with an output rangeof 1460 nm to 1580 nm. An array of polarization maintaining (PM) opticalfibers (PLC Connections, LLC., USA) is used to couple light from the tun-able laser to the silicon waveguides on the chip through a GC (figure 2.8).The guided mode in the waveguide will then travel through the biosensor andafter interacting with the reagents will go to an output GC, which couplesthe light from the waveguide into another PM optical fiber.A vacuum is used to immobilize the photonic chip onto the motorizedstage (Thorlabs, USA). The efficiency of light coupling from the fiber array242.2. Methods and Materials!!Figure 2.8: Schematic representation and a picture of how light gets coupled intothe chip. [courtesy of UW team]to the Si waveguides and back out from the chip is very sensitive to thealignment (rotational and linear alignment) of the fiber array itself to thefiber grating couplers. By monitoring the output power and moving thestage in plane, the alignment is automatically optimized with a MATLABscript by maximizing collected output power.The stage temperature is controlled with a Peltier element and a feedbackcontroller (LDC501, Stanford Research System, USA) with a typical stabilityof 1-10mK.The intensity of the output light is then measured with an optical power252.2. Methods and Materialssensor (Agilent 81635A, Agilent Technologies, Inc., USA). All of the abovementioned processes are controlled with MATLAB programs that control thetunable laser and optical detectors as well as a motorized stage holding thechip. A MATLAB script is also used to sweep the laser wavelength andto acquire the transmission spectrum. The motorized stage is controlledthrough a script to align the fiber array to the GCs on the chip; and movefrom one GC to another to interrogate various biosensors.Figure 2.10 and 2.9 are two picture representation of the test setup. Oneusing PDMS as microfluidic channel, and the other using the silicone gasketwith Polytetrafluoroethylene (PTFE) flow cell.(a) (b) Figure 2.9: Schematic representation of our testing platform [courtesy of UW team]2.2.5 Microfluidic IntegrationCustom design and fabrication of the microfluidic channels in poly dimethyl-siloxane (PDMS, Sylgard 184, Dow Corning, USA), using soft-lithographyprocedures, provided the means to expose specific set of devices on SOI chip262.2. Methods and MaterialsFigure 2.10: The experimental setup: fiber array is used to both bring the lightfrom a tunable laser source (operating in 1460 nm to 1580 nm wavelength range)to, and collect the output light from the chip to be transferred to the optical powersensor. The chip is immobilized on the motorized stage using vacuum. [34]to the fluidic reagents. The mold masters were fabricated with standard pho-tolithography in SU-8 2075 (MicroChem, USA). Uncured PDMS was thenpoured onto the molds to a thickness of about 1 cm. Before curing for 2hours at 80oC, the PDMS was degassed in a desiccator for 10min to removeunwanted air bubbles.These fabricated polydimethylsiloxane (PDMS) microfluidic channels wereused to deliver the refractive index titrations to the sensors (figure 2.10).Inlet and outlet holes are punched into the PDMS layer to access the mi-crochannels and to connect to the syringe pump. The inlet and outlet holeswere punched with a 0.5 mm coring tool (Schmidt Press, CORP, USA) priorto bonding and connected to tygon micro-bore tubing using 22 gauge blunt272.2. Methods and Materialsneedles (Nordson EFD, CORP., USA) to serve as the fluidic inlet and outletof the device. The PDMS fluidic block was then aligned to the devices understudy on the SOI substrate using a stereo microscope. The reversible bondformed between the SOI substrate and the PDMS block is strong enoughto form a seal to withstand the pressure used to drive the flow during ourexperiment; however, to minimize the risk of leakage, the fluids were sup-plied to the channels under a negative pressure (the syringe pump was setto withdraw rather than inject). A syringe pump (KDS-230, KD Scientific,Inc., USA) operating in withdraw mode (negative pressure) was used to con-trol the flow rate in the device, usually at a constant flow rate of about10 µL/min. Maintaining a slow flow is required in these experiments both toavoid creating air bubbles as well as to allow the system to maintain thermalequilibrium.In some recent experiments at the university of Washington, a differentmethod is used to deliver various solutions to the sensor. To deliver reagentsto the sensor, a 500 nm thick laser-cut silicone gasket (Grace BioLabs, Bend,Oregon) defined 300 µm wide channels over the optical sensors and matedwith a custom, PTFE (Teflon R©) flow cell to connect the Tygon tubing (figure2.9).2.2.6 Sensor CharacterizationTo characterize the sensor devices and demonstrate their bulk sensitivitycapabilities, aqueous solutions of salt and aqueous solutions of glucose were282.2. Methods and Materialsused, interchangeably. A standard sandwich assay was used to demonstratethe surface sensitivity capabilities of the sensors.Currently, to characterize and demonstrate the performances of our de-vices, we have to perform multiple steps and multiple scans at each step.This can take a few hours for characterization of each device and even longer(sometimes days) when surface modifications with bio-reagents are required.However, when these devices are developed and ready to be used for sensingapplications, ideally only one scan is needed to determine/predict the valueof interest, which takes less than a minute.Refractive Index Calibration Using Salt SolutionsTo characterize the performance and bulk sensitivity of the devices, a set ofaqueous solutions of NaCl were used with various concentrations (7 samplesin the range of 0 to 2M). To ensure the accuracy of the characterization,refractive index (RI) of these solutions was measured and characterized usinga Reichert AR200 digital refractometer (Depew, NY), in refractive index unit(RIU).The measured solutions are included in table 2.1.The sensor was subjected to the various solutions and its response mea-sured to determine the resonant shift while the optical stage was thermallytuned to 25 degrees Celsius to limit the impact of thermal drift on the mea-surements.292.2. Methods and MaterialsTable 2.1: Measured refractive index (RI) of various concentrations of salt solu-tions, in refractive index unit (RIU)Reagent Measured RI (RIU)Ultrapure, deionized water 1.333462.5 mM NaCl 1.3335125 mM NaCl 1.3344250 mM NaCl 1.3354500 mM NaCl 1.33751 M NaCl 1.3430Refractive Index Calibration Using Glucose SolutionsAqueous solutions of glucose can also be used to determine the bulk sensitiv-ity characteristics of the sensors. Therefore, aqueous solutions of D-Glucose(D16-500, Fisher Chemicals, Fisher Scientific, Inc.) in distilled water withvarious concentrations were prepared using the multiple dilution method (0to 2000 mg/dL). The refractive index of glucose solutions can be estimatedaccording to the following Equation [46]:nglucose(λ) = nH2O(λ) + 1.515× 10−6Cglucose (2.10)where nglucose is the refractive index of the aqueous solution of glucose, andCglucose is the concentration of glucose in mg/dL. These aqueous solutionswere used to characterize the performance and sensitivity of our resonators.302.2. Methods and MaterialsBio-Assay Experimental ReagentsTo test the performance and response of the devices to bio-molecules, we useda modified sandwich assay [47, 48] involving well-characterized molecules withhigh binding affinities including: anti-Streptavidin (antiSA, Vector Labs;Burlingame, CA), streptavidin (SA, Vector Labs; Burlingame, CA), and Bi-otinylated Bovine Serum Albumen (biotin-BSA, bBSA), which was conju-gated per the manufacturer’s instructions using a commercial biotinylationkit (SoluLink; San Diego, CA). During the experiments, the optical stagewas thermally controlled at 30oC to minimize thermal drift. Reagents wereintroduced to the sensor arrays using a reversibly bonded PDMS flow celland Chemyx Nexus 3000 Syringe Pump (Houston, TX) at 10 µL/min. Se-lected peaks were tracked every 45 seconds using an Agilent 8164A 1550 nmmainframe with a tunable laser (Agilent 81682A) with an integrated detec-tor (Agilent 81635A). Each sensor was exposed to phosphate-buffered saline(PBS) for at least 20 minutes prior to other reagents to establish an initialsignal baseline.312.3. Sensor Designs and Calibration Results2.3 Sensor Designs and Calibration Results 2Using these metrics, we analyze the performance of ring, disk, and Bragggrating resonators fabricated in SOI, and integrated with microfluidics. Wealso demonstrate a biosensing experiment.In addition to investigating these proposed devices at around 1550 nmwavelength, which is commonly used for telecommunication applications,some of these devices are under investigation for wavelengths around 1220nm. At 1220 nm, light absorption from water is strongly decreased with re-spect to the 1550 nm range, thus increasing the quality factor and improvingthe limit of detection [17].When dealing with resonator sensors, the sensitivity is determined by theshift of the resonant peak wavelength as the refractive index of the solutionchanges. A set of aqueous solutions of NaCl, as explained in section 2.2.6 isused to characterize the sensor devices under investigation.Resonator sensors with a high quality factor are desirable. A high qual-ity factor of the sensor means improved accuracy of the detection due tothe improved minimum detectable wavelength shift. This will consequentlyimprove the Limit of Detection (LOD) based on equation 2.9.2Parts of of this section have been published in [34]:S. Talebi Fard, et al., “Label-free silicon photonic biosensors for use in clinical diagnostics”,Proc. SPIE, Silicon Photonics VIII, 8629:86290914 (Invited), SPIE OPTO, InternationalSociety for Optics and Photonics, 2/02/2013.322.3. Sensor Designs and Calibration Results2.3.1 Optical Sensor DesignsOptical resonators have shown promises as biosensors due to their longerinteraction with the analyte surrounding the resonators. The following res-onators have been investigated in our group: disk resonators; slot waveguidering resonators; strip waveguide Bragg gratings; and slot waveguide Bragggratings. Figure 2.11 provides an SEM image of each of these resonators.Also, since they are integrated in waveguides, they often have compact foot-print and thus are easily integrated with microfluidic channels.5 µm 500 nm 5 µm 500 nm (a) (c) (b) (d) Figure 2.11: SEM images of the fabricated devices on SOI wafers, using E-Beamlithography process at the University of Washington. a) disk resonators; b) slotwaveguide ring resonator; c) strip waveguide Bragg grating; d) slot waveguideBragg grating.332.3. Sensor Designs and Calibration Results2.3.2 Disk ResonatorsDisk resonators offer potential advantages of improved limits of detection dueto lower scattering losses and thus higher quality factors [17]. The natureof disk resonators means that there is only one sidewall surface from whichscattering can occur. These reduced losses increase the resonator qualityfactor and in doing so have the potential to improve the limit of detectionas defined in Equation 2.9. Additionally, disk resonators offer very smalldevice footprints; this is advantageous for multiplexing (many disks can fitin a small area, and the disks offer a wide free spectral range, FSR), and thesensing surface area is comparatively small.Figure 2.12 shows the schematics of the 10 µm disk resonators and thesimulated mode profiles for the first three TE and TM modes.(a) (b) (c) (d) (e) (f ) 10 µm Si SiO2 Figure 2.12: Mode profiles (TE and TM) for 10 µm disk resonator. a-c are thefirst to third TE modes; and d-f are the first to third TM modes.342.3. Sensor Designs and Calibration ResultsWe have investigated 10 µm disk resonators supporting both TE and TMmodes. As Figure 2.12 illustrates, the TM modes in our disk geometrieshave electric fields that penetrate further into the analyte as well as morefield traveling in the analyte itself; this leads to more interaction of light withanalyte and therefore higher refractive index sensitivity. The results of thesensitivity analysis of the 3 µm disk are shown in Figure 2.13, demonstratingthe peak shift and refractive index sensitivity calibration for the two modespresent in the disk.(a) (b) Wavelength (nm) Transmitted Power (dBm) Figure 2.13: a) Shows how the spectra shifts as the refractive index of the mediumchanges (using various concentrations of NaCl). b) Sensitivity for the fundamental(black) second (blue) TE modes in 3 µm radius disk resonator. The error barsindicate a 99% confidence interval (within 3 standard deviation). We observesensitivities of 26 nm/RIU for the fundamental mode and 29 nm/RIU for thesecond mode.352.3. Sensor Designs and Calibration Results2.3.3 Strip Waveguide Bragg GratingIntegrated waveguide Bragg gratings are also promising candidates for biosen-sors. Compared with other resonant structures (e.g. ring or disk), waveguideBragg gratings usually operate at only one particular wavelength (Braggwavelength) and thus are not limited to FSR for the maximum range of thepeak wavelength shift.Figure 2.14 represents a schematic of a strip Bragg grating sensor withphase shift; with inset being the mode profile of the main propagating TEmode. The Bragg gratings are realized with corrugations on the lateral side-walls of the strip. As light travels through the waveguide, the optical modeexperiences periodic modulation of the effective refractive index, and theBragg condition depends on the grating period and the effective refractiveindex of the medium.The phase shift region in the strip Bragg grating sensors constructs acavity with two Bragg reflector mirrors and induces a resonant peak in thestop band of the Bragg gratings. The Q factor of the resonance can be veryhigh (100,000 in air [49] ).362.3. Sensor Designs and Calibration ResultsFigure 2.14: Schematic of a strip Bragg waveguide and the propagating TE modein the waveguide cross-section.Figure 2.15 shows the experimental results of the sensitivity analysis ofstrip waveguide Bragg grating sensor. A sensitivity of 59 nm/RIU is mea-sured, which is close to the simulated value of about 55 nm/RIU. The qualityfactor (Q) of this device is measured to be 27600, which leads to an intrinsiclimit of detection of 9.3× 10−4 RIU.372.3. Sensor Designs and Calibration Resultsa)1516 1516.5 1517 1517.5 1518−50−45−40−35−30−25Wavelength (nm)Power (dBm)  0 mM62.5 mM125 mM250 mM500 mM1 M2 Mb)0 5 10 15 20x 10−3− Index Change (RIU)Peak Wavelength Shift (nm)  y = 59*x + 0.044Figure 2.15: a) Shows how the spectra shifts as the refractive index of the mediumchanges (using various concentrations of NaCl). b) Shows the sensitivity of thestrip Bragg sensors. Error bars show experimental results for peak wavelengthshift versus refractive index change of the medium. The error bars indicate a 99%confidence interval (within 3 standard deviation). Red line is a linear fit to thesepoints indicating the sensitivity of about 59 nm/RIU.2.3.4 Slot Waveguide Ring ResonatorA slot waveguide racetrack resonator with a 30 µm radius, 300 nm waveguidewidths and 130 nm slot was fabricated and yielded quality factors of about1450 and sensitivities of 263 nm/RIU. The low Q is partly due to high bend-ing, mode mismatch, and scattering losses. This limited Q suggests that it isnecessary to eliminate the bent regions in the slot device in order to obtainhigh ILODs.2.3.5 Slot Waveguide Bragg GratingTo enhance the light-analyte interaction, we also applied the phase-shiftedBragg grating structure in a slot waveguide [50], as shown in Figure 2.16.382.3. Sensor Designs and Calibration ResultsThe electric field of the slot waveguide is concentrated inside the small low-index slot region. This unique property makes the slot waveguide much moresensitive to the surrounding fluidics than conventional strip waveguides thatuse only weak evanescent field tails. The Bragg gratings are constructed bycorrugating the outer sidewalls of the slot waveguide, where the evanescentfield decays.SiO2Si Phase shiftMicrofluidic channelFigure 2.16: Schematic of a slot Bragg waveguide and the propagating TE modein the waveguide cross-section.Figure 2.17 shows the experimental results of the sensitivity analysis ofa slot waveguide Bragg grating sensor. A sensitivity of 340 nm/RIU is mea-sured, which is in excellent agreement with simulation results [30]. Thequality factor (Q) of this device is measured to be about 15000, which leadsto a limit of detection of 3.0×10−4 RIU. Compared to the slot waveguide ringresonators, the slot Bragg grating resonators exhibit significantly enhanced Qfactors since they do not suffer from the bending and mode mismatch lossesas in ring structures.392.4. Biological Resultsa)1528 1528.5 1529 1529.5 1530 1530.5 1531 1531.5 1532−55−50−45−40−35−30−25Wavelength (nm)Power (dBm)  0 mM62.5 mM125 mM250 mM500 mM0 mMb)−1 0 1 2 3 4 5x 10−3−0.500.511.52Refractive Index Change (RIU)Peak Wavelength Shift (nm)S=340nm/RIU  ExperimentSimulationFigure 2.17: a) Shows how the spectra shifts as the refractive index of the mediumchanges (using various concentrations of NaCl). b) Shows the sensitivity of theSlot Bragg Sensors. Error bars show experimental results for peak wavelengthshift versus refractive index change of the medium. The error bars indicate a 99%confidence interval (within 3 standard deviation). Red line is a linear fit to thesepoints indicating the sensitivity of about 340 nm/RIU.2.4 Biological ResultsAs a first step towards demonstrating the sensor’s biosensing capabilities, weperformed a modified “sandwich” assay using streptavidin (SA) as a modelprotein and its binding partners, biotin and a monoclonal anti-SA antibody.For the purposes of this study, these molecules were selected based on theiravailability and ease of use. Additionally, it is worth noting that the specificmolecular recognition event between SA and biotin is one of the strongestnon-covalent bonds known [51] and anti-SA has been used to demonstratespecific binding to immobilized SA on silicon photonic sensors previously[52].Figure 2.18 shows results from the biosensing experiments for a 3 µmradius TE-mode resonant disk and a single-waveguide Bragg interferometer402.4. Biological Results!"#$!%#$!&#$Figure 2.18: Wavelength shift during biosensing demonstration for the 3 µm radiusTE-mode disk sensor (a) and single waveguide Bragg interferometer sensor (b).(c) Illustrates reagent sequencing corresponding to regions [i, ii, and iii] in (a)and (b). Region i = Biotinylated Bovine Serum Albumen (b-BSA) (2 mg/mL), ii= streptavidin (SA) (1.8 µM), iii = anti-streptavidin (anti-SA) (125 µg/mL). APBS-wash preceded and followed the introduction of each reagent in steps i-iii.sensor. Wavelength shifts resulting from the molecular binding events forthe disk resonator and Bragg sensor are shown in Figure 2.18 (a and b)respectively. Figure 2.18 (c) illustrates the idealized sequence of molecularinteractions resulting in the sandwich assay. These sequential binding intera-tions correspond to the wavelength shifts shown in regions i, ii, and iii shownin Figure 2.18 (a, b).After establishing a signal baseline in PBS buffer, biotinylated-BSA (b-BSA) was adsorbed to the oxide of the sensor surface (shown in region i in412.5. Discussion and AnalysisFigure 2.18 (a, b)). Next, SA was introduced to the functionalized sensor,binding irreversibly to the immobilized b-BSA, resulting in the wavelengthshift observed in region ii (Figure 2.18 (a, b)). Finally, Anti-SA bound thecapture SA, serving as a final signal amplification step, resulting in the ad-ditional resonant peak shift shown in region iii (Figure 2.18 (a, b, and c)).These binding results are in good agreement with the formation of a multiaddlayer biomolecular system consisting of b-BSA, SA and anti-SA. Ideally,the refractive index change (and subsequent wavelength shift) resulting fromeach additional layer would correlate precisely to the molecule’s mass of eachprotein. However, steric hindrance due to dense molecular packing limits1:1 stoichiometries of binding. This, coupled with the exponential decay ofthe evanescent sensing field, causes the sensor’s response to each addlayer todeviate from the ideal. With those limitations in mind, the sandwich assayclearly demonstrates expected responses to the biological interactions anddemonstrates the platform’s suitability for interrogating molecules in othermulti-layer biological assays.2.5 Discussion and AnalysisTable 2.2 presents a summary of geometrical specifics, resonance wavelengthsand light polarizations, as well as sensing performance in terms of Q, S, andILOD (equations 2.5, 2.6, 2.8, and 2.9) of our fabricated SOI sensors. Inaddition, Figure 2.19 presents a comparison of the limits of detection and422.5. Discussion and Analysissensitivities of devices summarized in Table 2.2 to our previously-presenteddevices [17] and to limits of detection of various sensor configurations re-ported in literature. The blue line on Figure 2.19 represents the theoreticallimit to the obtainable sensor performance (in water).Table 2.2: Summary of the performance and characteristics of our silicon photonicssensorsSensor Specifications λ TE/TM Qwater S ILOD(nm) ( nmRIU ) (RIU)Disk 3 µm radius 1527 TE 32300 26 1.8×10−3Disk 10 µm radius 1512 TE 131000 21 5.5×10−4Disk 10 µm radius 1543 TM 16000 142 6.8×10−4Slot Ring 30 µm radius 1500 TE 1450 263 1.3×10−2Strip Bragg 0.5 µm width 1517 TE 27600 59 9.3×10−4∼112 µm lengthSlot Bragg 0.5 µm width 1530 TE 15000 340 3.0×10−4∼112 µm lengthIt is worth noting that the lowest limits of detection (about one order ofmagnitude less than other sensor configurations) are achieved with the TEand TM modes [28] in the 10 µm radius disks as well as the slot Bragg grat-ing resonator sensors. On the other hand, their refractive index sensitivitiesvary by an order of magnitude (with TM disk and the slot Bragg with betterperformance), but their different quality factors equalize the limits of detec-tion to the same order of magnitude. Concerning Bragg sensors, they bothhave a very low limit of detection and high Q, but the sensitivity is about 6times improved for the configuration with the slot waveguide. This is likely432.5. Discussion and Analysisdue to an enhanced interaction between the electromagnetic field and light,causing a wider shift of its resonance wavelength. We can observe in Figure2.19 that Q factor of slot Bragg is very close to the theoretical limit, so onecan conclude that it is mainly limited by the water absorption.In addition to ILODs, often resonator characteristics can influence thechoice of best resonator type for a given application. Depending on theshape and size of the microchannels, number of multiplexed sensors, as wellas shape of target molecule, one can determine the best sensor for the par-ticular application. For example, the 3 µm radius disks have the smallestfootprint, while Bragg sensors have long but very narrow shape; this per-haps makes 3 µm disks better suited for multiplexing applications. Indeed,the ultimate best choice of sensor type will be dictated by the requirementsof the application; whether sharp peaks or large peak shifts are preferable,whether it is advantageous to track multiple peaks at the same time. Obvi-ously, the sizes and expected concentrations of the analyte or of the specifictarget molecule will all play a role in determining the best sensor type. Forexample, slot waveguides may be better suited for sensing small moleculeswith low concentrations, because they offer high sensitivity, but they requirethat molecules can flow also in the slot (that has a cross section of 150 nm× 220 nm), where much of the field is concentrated.442.6. Conclusion and Future Work103 104 10510−210−1100Q factorNormalized Sensitivity [/RIU]  1.0e−026.3e−034.0e−032.5e−031.6e−031.0e−031.0e−036.3e−044.0e−042.5e−041.6e−041.0e−04Ring TMRing TERing TERing TE1D PCSlot Ring TERing TEDisk TE, 2.5 µmDisk TM, 10 µmDisk TE, 10 µmDisk TE, 3 µmSlot Ring TEStrip BraggSlot BraggLimit at 1550 nmFigure 2.19: Experimental results for performance and figure of merit (ILOD) ofvarious sensors. All the red filled points are devices developed by our group atUBC/UW; and black hallowed ones represent devices developed in other groups[9, 27, 29, 53–55] for comparison purposes. The y-axis is the sensitivity normalizedto the peak resonant wavelength (equation 2.4). The blue line is the theoraticallimit for 1550 nm in water.2.6 Conclusion and Future WorkIn this section, we have discussed the emerging and promising role of in-tegrated optical biosensors, and we described various individual resonatorsensors that have been modelled and fabricated by our group. Those sen-sors have been characterized and validated using a standard ‘sandwich’ assay.Having variety of sensors characterized have prepared us for performing moremeaningful bio-assay experiments using more custom designed sensors to de-tect essential aspects under study.452.6. Conclusion and Future WorkHaving sensors with small footprints, it would be very easy to have high-throughput multiplexed and simultaneous analysis, with hundreds of sensorseach-one ‘tuned’ for a different specific target. All those sensors can bedesigned and aligned such that they can be easily integrated with microfluidicchannels, as well as they can have built-in normalization with integratedreference sensors. Considering that all those sensors would be fabricatedwith standard SOI chip processes that are CMOS compatible, facilitatingintegration with on-chip electronics, they would be very promising in termsof whole systems integration. Furthermore, our integration of these deviceswith microfluidics and bioassays represents a step forward towards realizingLab on Chip systems.A final remark on the wavelength we have used, that is 1550 nm, com-monly used for telecommunications, thus it has been very well characterizedand offers several low-cost components. Since our aim is biological sens-ing, it is worth exploring different wavelength regions, which may offer someadvantages or the possibility of gaining multi-wavelength information. Forexample we are start to investigate 1220 nm as a working wavelength, sinceit presents a significantly reduced water absorption, which in turn may im-prove the figure of merit, ILOD. As seen in the work by Chrostowski et al.(Figure 8b)[17], the detection limit is lower at around 1220 nm compared to1550 nm, making the wavelength window of 1.2 to 1.3 µm a more desirablerange for sensing applications in the presence of water.46Chapter 3Ultra-Thin Resonator Sensors 3This chapter presents simulation and experimental results of ultra-thin opti-cal ring resonators, having larger Evanescent Field (EF) penetration depths,and therefore larger sensitivities, as compared to conventional Silicon-on-Insulator (SOI)-based resonator sensors. Having higher sensitivities to thechanges in the refractive indices of the cladding media is desirable for sensingapplications, as the interactions of interest take place in this region. Usingultra-thin waveguides (<100 nm thick) shows promise to enhance sensitivityfor both bulk and surface sensing, due to increased penetration of the EFinto the cladding. In this work, the designs and characterization of ultra-thin resonator sensors, within the constraints of a multi-project wafer servicethat offers three waveguide thicknesses (90nm, 150nm, and 220nm), are pre-sented. These services typically allow efficient integration of biosensors withon-chip detectors, moving towards the implementation of lab-on-chip (LoC)systems. Also, higher temperature stability of ultra-thin resonator sensorswere characterized and, in the presence of intentional environmental (temper-3Parts of of this chapter have been published in [35]:S. Talebi Fard, et al., ”Performance of ultra-thin SOI-based resonators for sensing appli-cations.” Optics Express 22, no. 12 (2014): 14166-14179.473.1. Introduction and Backgroundature) fluctuations, were compared to standard transverse electric SOI-basedresonator sensors.3.1 Introduction and BackgroundAs described in section 2.3, our group has previously compared the ILODsof waveguide-based resonator sensors fabricated in silicon-on-insulator (SOI)operating near wavelength of λ = 1550 nm. Ring [17], disk [28], and Bragggrating [30] resonators have been demonstrated with ILODs approaching thetheoretical limit (due to the optical absorption of water) of 2.5 × 10−4 RIU[17].One technique for improving the performance of a waveguide-based res-onator sensor is to increase its sensitivity, S. This can be done by increas-ing the interaction between the propagating optical mode and the claddingmedium, since the sensing mechanism relies upon the interaction of theevanescent tail of the guided mode in the waveguide with molecules in thecladding medium. When a larger portion of the optical field travels outside ofthe silicon waveguide core, both the mode sensitivity (Smode) and S increase,resulting in a larger shift in the resonant wavelength. This also results inhigher loss due to the optical absorption of water, resulting in a lowered Q,and, thus, there is a trade-off that results in limited improvements in theILOD.The use of thinner waveguide cores was proposed to increase the interac-483.1. Introduction and Backgroundtion of the EF of the optical mode with the cladding. Theoretical analysis[56, 57] demonstrates that thinner SOI waveguide cores are more respon-sive (have higher sensitivities) both to bulk cladding concentrations and thinadsorbed biomolecule layers. Analysis to determine the optimum waveg-uide thicknesses for specific biological applications has also been performed[56, 57]. In addition, thin and ultra-thin waveguides have been fabricated inCMOS-compatible processes and exhibited low losses; for example a loss ofabout 2 dB/cm was reported for a 50 nm thick strip waveguide [58]. Ultra-thin waveguide resonator sensors (based on using ultra-thin silicon cores) alsooffer the potential for improved thermal stability for biosensing applications.The refractive indices of the silicon core and silicon dioxide (buried oxide,BOX) materials increase with increasing temperature [59, 60], while the in-dex of the water cladding decreases with increasing temperature [61, 62], thusmoving more of the propagating field to the cladding decreases the overalleffect of temperature on the modal effective index.This chapter includes the first experimental demonstration of ultra-thinwaveguide resonator sensors fabricated in a CMOS-compatible SOI process.We have successfully simulated, fabricated, and tested ultra-thin waveguideresonators using waveguide core thicknesses (90 nm) available in commer-cial Multi Project Wafer (MPW) runs offered by Optoelectronic Systems inSilicon (OpSIS) and/or CMC Microsystems fabricated by the Institute ofMicroelectronics (IME) in Singapore. We focused on these thicknesses asthey offered the best potential for integration with future CMOS-compatible493.2. Design Methods and Analysisprocesses. Using both simulated and experimental results, we have demon-strated that ultra-thin ring resonators have higher sensitivities to the changesin cladding medium and lower sensitivities to temperature. We also comparedthese ultra-thin resonators with standard 220 nm thick ring resonators andobserved significantly more stable responses.These sensors were tested with glucose solutions, from which a predictionmodel was created to predict glucose concentrations. These sensors couldalso be used to study molecular bindings for chemical and biological researchpurposes [9].3.2 Design Methods and AnalysisWe have used Lumerical MODE Solutions and analytical models in MATLABto design optical resonators with various thicknesses. MODE is used tocalculate the effective index for each waveguide thickness and, ultimately,the change in effective index as the cladding is altered. In these MODEsimulations, a 4 × 4 µm simulation area was used. This results in errors lessthan 10−7 in calculations of effective index (neff) based on the convergencetest (figures 2.3 and 2.4). The resulted errors in sensitivity calculations areinsignificant (< 0.1%). Figures 3.1(a-c) illustrate the mode profile of thepropagating TE fundamental mode for waveguide thicknesses of 220, 150,and 90 nm.The thinner the waveguides, the less confined are the propagating modes503.2. Design Methods and Analysis(a) TE mode profile for 90 nm thicksilicon core(b) TE mode profile for 150 nm thicksilicon core(c) TE mode profile for 220 nm thicksilicon core	  	  	  	  	  	  	  Cladding	  (H2O)	   	  	  Core	  (Si)	  	  	  	  	  	  	  	  BOX	  (SiO2)	  Width	   Thickness	   4	  μm	  4	  μm	  (d) Waveguide cross section pa-rametersFigure 3.1: Simulation results for the cross-section of the silicon waveguides (SOI)with water as the cladding medium. (a-c) the mode profile of the fundamental TEmode for silicon thicknesses of 220, 150, and 90 nm, in log scale, calculated usingMODE Solutions. (d) Schematic of the waveguide’s cross section and parameters.to their silicon cores and the higher are their penetration depths into thesurrounding media. The evanescent field in the top cladding at the centre ofthe silicon slab can be approximated by:E(d) ≈ E0e−d( 2piλ0)√N2eff−n2cl (3.1)where E0 is the electric field at the surface of the silicon core, d is the perpen-dicular distance from the surface, Neff is the effective index of the waveguide,513.2. Design Methods and Analysis500 600 700 800 900 1000 1100100120140160180200220240260Width (nm)EF Penetration Depth (nm)   90 nm thick150 nm thick220 nm thickFigure 3.2: Simulated 1/e penetration depths of the evanescent field of the firstguided TE mode into the solution for various waveguide core thicknesses andwidths.ncl is the refractive index of the cladding, and λ0 is the operating wavelength.The penetration depth (de) is defined as the distance from the silicon-cladding interface (at the centre of the silicon) at which the evanescent fielddecays to E0/e. This penetration depth (de) of the mode is calculated usingMODE and Equation 3.1 for the waveguide thicknesses under investigation,and is plotted in Figure 3.2.Racetrack resonators are usually designed and optimized at critical cou-pling, resulting in maximum Extinction Ratio (ER). Analytical modellingin MATLAB is used to calculate the parameters and predict the responseof the ring resonator, to achieve an optimal design. This model considersestimated and simulated losses as well as simulated mode profiles for various523.2. Design Methods and Analysissections of racetrack in the calculations. The simulation area of 4 × 4 µm,assuming a standard SOI wafer (with 2 or 3 µm of SiO2 BOX between thewaveguide and silicon substrate), includes the waveguide core, its SiO2 BOXand cladding. Propagation losses (scattering loss + substrate leakage) wereestimated to calculate the critical coupling lengths for racetrack resonators.Researchers have investigated these losses, analytically and experimentally.They reported a substrate leakage of 1 dB/cm for 100 nm slab waveguideover a 1 µm SiO2 BOX [63], and a substrate loss of 3-4 dB/cm for a 500 ×120 nm waveguide over 1.4 µm SiO2 BOX [64, 65] (with reported total loss of7 dB/cm for racetrack structure). In addition, a propagation loss of 2 dB/cmfor a 500 × 50 nm waveguide on a standard SOI wafer is reported [58]. Ithas been demonstrated that the losses due to substrate leakage decreasedexponentially as the thickness of the oxide layer increased [31, 63]. Prop-agation losses, both scattering loss and substrate leakage, decreased as thewaveguide width increased (higher mode confinement) [31, 66]. Therefore,for our case with a standard SOI wafer (with 2 or 3 µm of SiO2 BOX) andwider waveguides, we assumed a total propagation losses of 2-5 dB/cm.Based on the higher simulated sensitivities for ultra-thin 90 nm resonatorsensors, we decided to fabricate a few variations of these ultra-thin sensors aswell as conventional 220 nm thick resonators for comparison purposes. In thisresearch, resonators with radii of 10, 20, and 30 µm and waveguide widthsof 800 nm to 950 nm were designed for critical coupling, in order to achievethe best ER and optimized response. The goal was to use 90 nm waveguide533.2. Design Methods and Analysiscores, the thinnest offered by the standard MPW foundries. When usingultra-thin waveguides, a wider silicon core is necessary to guide the modeand has the advantage of reduced scattering losses due to sidewall roughness[58]. These reduced losses result in slightly higher Q values, but slightly lowersensitivities (as a small portion of the evanescent field is travelling outsideof the core). This trade-off between Q and S leads to an optimum point forthe ILOD. In addition, the single mode conditions for waveguides with 90nm silicon cores are not disturbed until the waveguide widths reach 900 nm,after which the waveguides are multi-mode.3.2.1 Sensitivities of TE Resonator Sensors toCladding Refractive Indices500 600 700 800 900 1000 1100050100150200250Waveguide Width (nm)Sensitivity (nm/RIU)   90nm thick150nm thick220nm thick(a)500 600 700 800 900 1000 11000. Width (nm)Sensitivity (nm/K)   90nm thick150nm thick220nm thick(b)Figure 3.3: Sensitivities of TE waveguide resonators with various thicknesses to(a) the analyte in the cladding medium and (b) to the temperatureAn EF sensor operates by detecting changes in the effective index of543.2. Design Methods and Analysisthe waveguide. The effective index of the waveguide can be affected by achange in the refractive index of the core and the refractive index of theBOX and/or the refractive index of cladding. We are most interested in thechange in effective index of the waveguide as a function of the change inrefractive index of the analyte in the cladding medium (equations 2.1 and2.5).If thinner cores are used in resonators, the mode is less confined to thecore and a larger portion of the evanescent field travels outside the core,resulting in more interaction with the cladding. Therefore, the dependenceof the effective index of the waveguide on the refractive index of the claddingmedium is increased in thinner waveguides, resulting in higher sensitivity.MODE is used to calculate the effective index of strip waveguides with variouswidths, thicknesses, and claddings. For each case, the change in effectiveindex as a function of the refractive index of the cladding is calculated. Wethen use the sensitivity relation for resonators (equation 2.5) to find theestimated sensitivity for each case at λ0 = 1550 nm. Figure 3.3(a) showsthe sensitivity that can be achieved in a strip waveguide resonator for threewaveguide thicknesses: 90, 150, and 220 nm.An ideal optical sensor will have high sensitivity (S) to cladding refrac-tive index changes and low sensitivity to other factors such as temperaturevariation or system noise.553.3. Experimental Methods and Materials3.2.2 Sensitivities of TE Resonator Sensors toTemperature VariationsThe sensitivity of a TE resonator sensor to temperature is defined as thewavelength shift in resonator’s response caused by a temperature change ofthe waveguide (core, buried oxide, and cladding). The change in refractiveindex as a function of a change in temperature ( dndT ) for silicon can be approx-imated for T around 295 K, and λ0 around 1.5 µm, to be dndT ≈ 1.8× 10−4/K[59, 60]. Since the waveguide thickness is changing in our study, the contri-bution of the temperature variation of the silicon dioxide (SiO2) BOX wouldbe significant and different for each case. Therefore, the sensitivity is esti-mated assuming that the temperature of the substrate and BOX is changingby the same amount as the temperature of the core. The sensitivity of therefractive index of silicon dioxide to temperature is about dndT ≈ 2.8×10−5/K[44]. Most solutions used for biological applications are aqueous. The depen-dence of the refractive index of water to temperature is dndT ≈ −9.9× 10−5/K[61, 62]. Table 3.1 shows the summary of these thermo optics (TO) coefficientvalues. Figure 3.3(b) shows the sensitivity of strip waveguide resonators totemperature variations of waveguide (when the cladding is water).3.3 Experimental Methods and MaterialsThe above mentioned designed ultra-thin racetrack resonator sensors werefabricated on a SOI chip using the E-Beam Lithography (EBL) System at563.3. Experimental Methods and MaterialsTable 3.1: Thermo optics coefficient of the materials in the systemMaterial TO coefficientdndT (1oC )Si waveguide 1.8× 10−4 [44, 59, 60, 67]Water (cladding) −9.9× 10−5 [61, 62]SiO2 (BOX) 2.8× 10−5 [44]the University of Washington - Washington Nanofabrication Facility (UWWNF). Figure 3.4 shows SEM images of two of these sensors. On-chip Grat-ing Couplers (GCs) are used to couple light into and out of the SOI chip [68].These sensors were tested using techniques and reagents described in section2.2.Figure 3.4: SEM image of ultra-thin TE resonator sensors for two radii of 10 µmand 30 µm, fabricated on SOI wafers, using E-Beam lithography process at theUniversity of Washington573.4. Performance of the Ultra-Thin TE Resonator Sensors3.4 Performance of the Ultra-Thin TEResonator SensorsIn this section, we report on the performance of our fabricated devices.3.4.1 Sensitivity Analysis ResultsThe optical responses of two ring resonators, one with the standard 220nm thick silicon core and one with 90 nm ultra-thin silicon cores, in thepresence of the various glucose concentrations (section 2.2.6) were measured.These two devices were measured consecutively in the presence of the samesolutions under the same experimental conditions. These experiments wererepeated on three days, at three different temperatures between 298 and299 K inclusive, therefore subjected to an intentional temperature variationsof 1 K. From the data, the wavelength shifts as functions of the change inrefractive index of the glucose concentrations were calculated and plotted forboth sensors.Figure 3.5 summarizes the wavelength shift responses of our two sensors,in the presence of the various concentrations of analyte, and plots theseshifts as functions of refractive index change. The slopes of the best linearfits to these points represent the sensitivities in nm/RIU, which is the peakwavelength shift as a function of refractive index change. Figure 3.5(a) isthe response of the traditional 220 nm thick resonator sensor, and Figure3.5(b) is the response of a 90 nm ultra-thin resonator sensor. The slope of583.4. Performance of the Ultra-Thin TE Resonator Sensors0 1 2 3 x 10−300. Index Change (ΔRIU)Peak Wavelength Shift (nm)(a) Measured sensitivity of a 220 nm thickTE sensor0 1 2 3 x 10−300. Index Change (ΔRIU)Peak Wavelength Shift (nm)(b) Measured sensitivity of a 90 nm ultra-thin TE sensorFigure 3.5: Peak wavelength shift as a function of the change in refractive indexof the glucose solutions. There are about 25 measurements for each concentration(i.e. each refractive index point), measured over three days subjected to intentionaltemperature variations of 1 K. The dashed line is the best linear fit to thesemeasurements, representing the sensitivity in nm/RIU. The estimated sensitivitiesbased on these linear fits for (a) is 38.2 nm/RIU and for (b) is 133 nm/RIU.the dashed lines in Figures 3.5(a,b) is the sensitivity of these sensors that are38.2 and 133 nm/RIU for the standard 220 nm thick and 90 nm ultra-thinresonator sensors, respectively. The wavelength shift in the 220 nm thickresonator shows strong perturbations (large errors) due to environmentalvariations such as temperature (average variations of 1 K were induced) ornoise, whereas the wavelength shift in the 90 nm ultra-thin resonator sensorshows a strong linear relation with the changes in the refractive index of theanalyte; i.e. lower sensitivity to temperature and noise. It is evident that theultra-thin resonator sensor shows higher sensitivity as well as better stability(significantly smaller errors). Having higher sensitivity and better stability593.4. Performance of the Ultra-Thin TE Resonator Sensorsallows improved predictions. Note that the temperature was varied in theseexperiments to demonstrate and compare the performance of the sensors inthe presence of intentional temperature variation; the errors are due to thesevariations. In typical experiments, where the temperature is not varied, theerrors are significantly smaller (Figure 3.7).To further demonstrate the lower sensitivity of ultra-thin resonators totemperature variations, and to characterize the temperature sensitivity ofthese sensors, the responses of the sensors at three different temperatureswere measured and the corresponding resonant wavelength shifts were plot-ted as functions of temperature change. The shift in resonant wavelengthas a function of temperature ( dλdT ) denotes the sensitivity of a sensor to tem-perature. Figure 3.6 shows the experimental results for the temperaturesensitivity of a 220 nm thick and a 90 nm ultra-thin resonator sensor, to be69 and 49 pm/K respectively.One potential application of these sensors is the use of their bulk sensitiv-ity in predicting concentrations of analyte in their cladding medium. Figure3.7 shows the results of using linear models, based on their sensitivities andmeasured wavelength shifts, to predict the concentrations of the analyte inthe cladding medium using our two sensors. The red (light) error bars showthe prediction results using the conventional 220 nm thick resonator withthe goodness of prediction (R2 value) of 0.85, and the thick black error barsare the result of predictions using the ultra-thin resonator sensor, with anR2 value of 0.993. The error bars are the result of intentional temperature603.4. Performance of the Ultra-Thin TE Resonator Sensors0 1 2 3 400. Change, ΔT (K)Peak Wavelength Shift (nm)Ultra-ThinConventionalFigure 3.6: Peak wavelength shift as a function of temperature change, demon-strates the sensitivity of our sensors to temperature change. Sensitivities of the220 nm thick (conventional) and 90 nm ultra-thin resonators are measured to be69 and 49pm/K, respectively.variations of 1 K. Temperature controlling these sensors would significantlyreduce the error bars, and the R2 value for ultra-thin sensors increasing to0.998 (figure 3.8). Figure 3.7 demonstrates significant improvement in ultra-thin resonator’s ability to measure glucose concentrations in the presence oftemperature variations, as compared to conventional 220 nm thick sensor.613.4. Performance of the Ultra-Thin TE Resonator Sensors−20 0 20 40 60 80 100 120−20020406080100120Known Concentration (mM)Estimated Concentration (mM)(a)−20 0 20 40 60 80 100 120−30−20−10010203040Known Concentration (mM)Error in Estimated Concentration (mM)  ConventionalUltra−Thin(b)Figure 3.7: The results of predictions of glucose concentrations using the two char-acterized sensors based on their responses on two different days under intentionaltemperature fluctuations of about 1 K. a) The red (light) error bars show the pre-dicted results using the 220 nm thick resonator(R2 = 0.85), and the thick blackerror bars show the predicted results using the 90 nm ultra-thin resonator sensor(R2 = 0.993). b) The bars represent the error in predicted concentrations usingour two sensors. The bars are based on the maximum +/- errors.3.4.2 Q Factor and Intrinsic Limit of Detection(ILOD)The Q factor of an optical waveguide resonator, being a measure of number ofoptical oscillations until the resonating energy decays to 1/e of its max value,is defined in Equation 2.6. This approximation is used to measure the Qvalues of our devices studied in this chapter. The Q of a resonator is inverselyproportional to the losses that are affecting the propagating mode (equation2.6). Various loss components contribute to the Q factor in ring resonators:scattering loss, bend loss, mode-mismatch loss, radiation loss in bends, andmaterial absorption, including water absorption. The Q factor is lower in623.4. Performance of the Ultra-Thin TE Resonator Sensors−20 0 20 40 60 80 100 120−20020406080100120Known Concentration (mM)Estimated Concentration (mM)Figure 3.8: The results of predictions of glucose concentrations using ultra-thinresonator sensors at constant temperature (R2 = 0.998). The light dashed line isfor: Estimated Concentrations = Known Concentrations.the thinner waveguide resonators, because of their higher losses due to theirincreased interaction with the biomolecules in the cladding medium. Theseincreased interactions are desirable for sensing applications. The highest Qfactor that was achieved by the ultra-thin resonators in water was around24,000, and the maximum Extinction Ratio (ER) measured was 28 dB.Figure 3.9 summarizes the modelling and experimental results of the EFsensors, presenting the intrinsic limit of detection of the EF sensors as func-tions of their Q factors and normalized sensitivities. The thin black dashedlines represent contours of constant ILOD. The thick light blue line is thetheoretical limit of detection for (unloaded) resonant sensors due to waterabsorption at 1550 nm [17]; its locus represents different proportions of lighttravelling in the water versus in the waveguide core, namely a high Q / low633.4. Performance of the Ultra-Thin TE Resonator Sensors104 10510−1Q factorNormalized Sensitivity [/RIU]  7.5e−045.6e−044.2e−043.2e−042.4e−041.8e−0490 X 800nm90 X 850nm90 X 900nm90 X 950nm220 X 500nmLimit: unloaded resonatorsLimit: CC resonatorsFigure 3.9: The intrinsic limit of detection (ILOD) of the silicon photonic evanes-cent field sensors, as functions of the normalized sensitivity (S’) and quality factor(Q). The thin black dashed lines represent contours of constant ILOD. The thicklight blue line is the theoretical limit of detection for (unloaded) resonant sensorsdue to water absorption at 1550 nm [17]. The corresponding light blue markers arethe modelling results for the specific unloaded resonators considered in this study,where only optical absorption due to water around the waveguide is considered.The thick black line is the theoretical limit for a critically coupled (CC) resonator.The corresponding black markers are the modelling results for critically coupledresonators, where the Q factor is determined from the simulated optical spectra;these also include additional losses (e.g. bend loss and mode-mismatch loss) exceptfor the waveguide scattering loss. The experimental normalized sensitivities and Qvalues for each fabricated device are plotted (white markers), with the waveguidedimensions noted in the legend.sensitivity for highly confined modes (e.g., thick silicon waveguides [17] anddisk resonators [28]), and a low Q / high sensitivity for weakly guided modes(e.g., thin silicon waveguides, TM polarized waveguides [54], and slot waveg-uides [55, 69]). The differences between the model (black markers) and theexperimental results (white markers) are attributed to scattering losses and643.5. Conclusion and Discussionthe excess losses of the directional couplers (the couplers were assumed to beideal in the model, i.e., κ2 + t2 = 1). Note that the difference between theexperimental and modelling results is much larger for the narrow waveguides(500 nm) as compared to the wide waveguides (800-950 nm). It is knownthat wide waveguides have much lower optical scattering loss [66, 70, 71].Scattering losses are usually induced by sidewall roughness. The interactionof the propagating mode with the sidewall roughness causes a strong scatter-ing effect [66, 70, 72]. Thus, we expect that the scattering losses of the thinand wide waveguides (800-950 nm) should be relatively small and, hence,we expect good agreement between the model and experiments. It is seenthat the experimental results for the 90 nm ultra-thin sensors agree very wellwith the model, both in terms of sensitivity and quality factor. These sensorsoffer performance that matches the theoretical limit for a critically coupledresonator.3.5 Conclusion and DiscussionWe have investigated, both by simulations and experiments, ultra-thin TEresonator sensors within the constraint of available thicknesses in standardMPW foundries and services. We obtained sensitivities over 100 nm/RIUwith the ultra-thin TE resonator sensors. We have demonstrated, by exper-iment and simulation, the increased stability of these ultra-thin resonators,as compared to the traditional 220 nm thick resonators, in the presence653.5. Conclusion and Discussionof temperature variations. We report Q factors on the order of 15,000 to25,000, with the ILODs on the order of 5 × 10−4 RIU. In addition, goodagreement between experimental results and simulations was demonstrated.The bulk sensitivity and capability of these sensors in predicting glucoseconcentrations, in the presence of intentional 1 K temperature fluctuations,was demonstrated and showed a manifold improvement in these predictionsas compared to traditional 220 nm thick resonator sensors. One method toquantify this improved performance, in the presence of temperature drift,is to compare their relative sensitivities (temperature sensitivity/bulk sensi-tivity). The calculated temperature-induced errors in the estimation of thechange in refractive index of the cladding (due to 1K temperature change)are approximately 0.0015 and 0.0004 RIU/K, for the case of our conventional220 nm thick resonator sensor and ultra-thin resonator sensor respectively.This means more than three times improvement in the performance (73%improvement).The ultra-thin resonator sensors developed here, using the smallest avail-able thickness offered by MPW foundries, can be integrated with the on-chipdetectors also offered by these standard foundries. Furthermore, given thatthey are fabricated using SOI technology, they are well positioned to be in-tegrated with CMOS electronics to produce a lab-on-chip. Additionally, ascompared to conventional 220 nm resonator sensors, our ultra-thin resonatorsensors, with larger evanescent fields, have unique advantages for bio sensing,e.g., by sampling more of the measurand, these sensors provide the capability663.5. Conclusion and Discussionto sense larger particles.67Chapter 4Optimized Sensitivity 4Evanescent field sensors have shown promise for biological sensing appli-cations. In particular, Silicon-on-Insulator (SOI)-nano-photonic based res-onator sensors have many advantages, including exquisite sensitivity andcompatibility with today’s high volume CMOS foundries. We have investi-gated the optimum design parameters within the fabrication constraints ofMulti-Project Wafer (MPW) foundries that result in highest sensitivity for aresonator sensor. We have demonstrated the optimum waveguide thicknessto achieve the maximum bulk sensitivity with a SOI-based resonator sensorto be 165 nm using quasi-TM guided mode. The closest thickness offered byMPW foundry services is 150 nm. Therefore, resonators with 150 nm thicksilicon core were fabricated and showed a sensitivity of 269 nm/RIU, whereasa similar resonator sensor with 220nm thickness showed sensitivity of around200 nm/RIU.4A version of this chapter will be published in:S. Talebi Fard, et al.,“Optimized Sensitivity of Silicon-on-Insulator Strip Waveguide Res-onator Sensor”, In review (2015).684.1. Introduction and Background4.1 Introduction and BackgroundResearchers have investigated the optimum waveguide thicknesses for maxi-mum sensitivities both analytically and experimentally for a waveguide slab[56] but not a waveguide resonator used for sensing which has different char-acteristics. It is known that the propagating TM-modes generally exhibithigher sensitivities due to an increased overlap of their evanescent field withthe cladding [36, 54, 56]. Figure 4.1 shows the evanescent field penetrationsdepth of a propagating TM mode for the two waveguide thicknesses offeredby MPW foundries (150 and 220 nm), based on equations 3.1. Comparingfigures 4.1 and 3.1 show that the evanescent field penetration depth for quasi-TM waveguides with regular thickness of 220 nm is close and comparable tothe penetration depth for TE guided mode in ultra-thin waveguides with 90nm thickness.The sensitivity can be defined in two ways: (1) bulk sensitivity and (2)surface sensitivity. For a waveguide, the homogenous or bulk sensitivity is de-fined as the sensitivity to the changes in the refractive index of the cladding,or aqueous solution surrounding the waveguide, assuming a homogenous so-lution ( δneffδnc , for the case of waveguide’s sensitivity, where neff is the effectiveindex of the waveguide, and nc is the refractive index of the cladding). Theother type of sensitivity, commonly referred to as surface sensitivity, is de-fined as the sensitivity to the adsorbed bimolecular layer to the surface of thesilicon core ( δneffδt , for the case of waveguide’s sensitivity, where t denotes the694.1. Introduction and Backgroundthickness of the adsorbed biomolecule). Both sensitivities were investigatedfor a simplified slab waveguide with silicon core on silicon dioxide substrateand aqueous cladding [56]. The results suggested a maximum bulk sensitivity( δneffδnc ) and surface sensitivity (δneffδt ) at waveguide thicknesses of 190 nm and210 nm respectively. The molecular add-layer used to determine the surfacesensitivity was assumed to be a protein adlayer with refractive index of 1.48[56, 73]. We have verified these optimum thicknesses as well using analyticalmethods (using analytical expressions by Yariv [74]) and fully-vectorial 2Deigenmode calculations. Note that the optimum value for surface sensitivitywould be different depending on the distance of the newly adsorbed layerfrom surface of the waveguide core.400 500 600 700 800 900 1000 1100 1200 1300150200250300350400450Width (nm)EF Penetration Depth (nm)  150 nm thick220 nm thickFigure 4.1: Simulated 1/e penetration depths of the evanescent field of the firstguided TM mode into the solution for various waveguide core thicknesses andwidths.In the case of resonators and interferometers, the sensitivity also depends704.2. Methods and Materialson the group index of the waveguide (ng). In this chapter, we have calculatedand experimentally validated optimum thicknesses to achieve the highest sen-sitivities for a biosensing resonator, including the dispersion effect accountedby ng. We fabricated TM resonators with the standard etch layer thicknessesoffered by MPW foundries and verified experimental observations with theo-retical simulations. Until now, the highest experimental sensitivity reportedfor quasi-TM ring resonators was 135 nm/RIU [54] but using the optimumwaveguide thickness to maximize sensitivity, we demonstrate quasi-TM res-onators with a bulk sensitivity of 270 nm/RIU.4.2 Methods and MaterialsEmploying fully-vectorial 2D eigenmode (using Lumerical MODE Solutions.)calculations and analytical equations in MATLAB, we have determined ourdesign parameters for a racetrack at, or close to, critical coupling [35, 44].TM resonators with 150 nm thick silicon core, close to thickness with max-imum sensitivity, and 220 nm thick resonators as a conventional referencefor comparison purposes were fabricated at The Institute of Microelectronics(IME) foundry in Singapore [75].Fabricated devices were characterized using a custom test platform de-veloped by our lab to sequence solutions over the sensor [28, 34, 76]. Forbulk sensitivity measurements, a set of NaCl refractive index (RI) titrationsranging from 62.5 mM to 1 M were subjected to the sensors to measure714.3. Theory: Waveguide’s Sensitivitytheir response. The chip was thermally tuned to 25 ◦C to limit the impactof thermal noise and drift on the measurements. For surface sensitivity mea-surements, standard sandwich assay involving well characterized moleculeswere used to demonstrate the sensors ability to detect biological interactionsas described in [76]. Figure 4.11(a) is schematic representation of our Bioassay material.4.3 Theory: Waveguide’s SensitivityThe homogenous sensitivity of a waveguide is defined as the change in effec-tive index as a function of a change in refractive index of cladding (equation4.1). We have calculated homogenous sensitivity (or bulk sensitivity) of a slabwaveguide, with silicon as the core, silicon dioxide as the substrate, and aque-ous cladding; using two methods: analytically (using analytical expressionsby Yariv [74]); and with 1D MODE simulations. A matching results betweenthese two methods were achieved (figure 4.2), verifying that our simulationsresults (markers in fig. 4.2) are accurately aligned with the analytical models(solid lines in fig. 4.2). These analysis results of waveguide’s homogenoussensitivity for a slab waveguide suggest that a 190 nm (36 nm) thickness isoptimal for TM (TE) bulk sensing, these values were also identified by otherresearchers [56].Waveguide bulk Sensitivity =δneffδnc(4.1)724.3. Theory: Waveguide’s Sensitivity10020030040050000. Waveguide Thickness (nm)δ neff/δ nc  TE AnalyTM AnalyTE MODETM MODEFigure 4.2: Sensitivities of the waveguide as functions of slab thicknesses. Solidlines show the results of analytical calculations and the markers show the resultsof 1D MODE calculations.where neff is the effective index of the waveguide and nc denotes the refractiveindex of the cladding.The surface sensitivity of a waveguide is defined as a change in effectiveindex of the waveguide, as the thickness of the molecular layer adsorbed tothe surface of the silicon core changes (eq. 4.2). To investigate the surfacesensitivity, we start with a slab waveguide using 1D MODE. We assumed themolecular add-layer to be a protein with refractive index of 1.48 [56, 73].Waveguide Surface Sensitivity =δneffδt(4.2)where t denotes the thickness of the adsorbed biomolecule. This is assumingthat a homogenous layer is aded evenly across the surface. Figure 4.3 showsthe simulation results for slab waveguide surface sensitivity, as a function734.4. Resonator’s Sensitivity - Effect of Dispersionof slab thickness. These results suggest that a 210 nm (65 nm) thicknessis optimal for TM (TE) surface sensing, these values were also identified byother researchers [56].5010015020025000. Waveguide Thickness (nm)δ neff/δ t (µm−1)  TE MODETM MODEFigure 4.3: Surface Sensitivities of the waveguide as functions of slab thicknesses.A 10 nm thick molecular layer with refractive index of 1.48 is used.4.4 Resonator’s Sensitivity - Effect ofDispersionFor resonators, sensitivity is defined as the shift in resonant wavelength dueto a refractive index change in the cladding, which results from a change inconcentration of the analyte in cladding or from biomolecular adsorptbion tothe surface of the silicon waveguide. To include the effect of this dispersion,the group index (ng) is used since it relates to the effective index of the744.4. Resonator’s Sensitivity - Effect of Dispersionwaveguide (neff) to changes in the resonant wavelength (equation 4.3) [44]:ng(λ) = neff(λ)− λδneffδλ(4.3)Simulations of various waveguide core thicknesses were used to calculateng and ultimately determine the resonator sensitivity (δneffδnc). Using the simu-lations results and equation 4.3, values for ng as functions of slab thicknessesis shown in fig. 4.4. It can be observed that group index (ng) of a TM prop-agating mode varies significantly as slab thicknesses vary between 100 nm to300 nm (fig. 4.4).1002003004005001.522.533.544.5Slab Waveguide Thickness (nm)ng  TE MODETM MODEFigure 4.4: Group indices of slab waveguides as functions of slab thicknesses.The next two sections discuss the fully-vectorial 2D eigenmode calcula-tions (simulations) and compare them with the observed, experimental re-sults for both kinds of sensitivities (bulk and surface) when considering theeffect of dispersion (including ng).754.5. Resonator’s Bulk Sensitivity4.5 Resonator’s Bulk Sensitivity - Theory,Simulation and Experimental ResultsThe bulk sensitivity of a resonator is defined as the shift in resonant peakthat is caused by a change in refractive index of cladding, ∆λres∆nc (equation4.4) [35].Resonator’s Sensitivity = S =∆λres∆nc=λresngδneffδnc(4.4)where λres is the resonant wavelength of the resonator and ng is the groupindex.To calculate the sensitivity of the resonators for the case of a slab waveg-uide, equation 4.4 along with the above simulated ng results and simulatedwaveguide sensitivities ( δneffδnc ) are used (figure 4.5, dashed lines with hallowedmarkers). For these calculations, we have considered a perturbation of 0.01RIU for the refractive index of the cladding, assuming that δneffδnc is nearlyconstant over small ranges. These results indicate that the maximum sensi-tivity for a quasi-TM mode happens when the silicon slab thickness is around155 nm. Conditions for maximum sensitivity for a waveguide resonator andwaveguide alone differ because of the effect of ng in the sensitivity of theresonators. The optimal thickness, which happens to be close to one of theetch-depths offered through a standard MPW processes, provides an advan-tage for improved sensitivity of a biosensor at the economies of scale offered764.5. Resonator’s Bulk Sensitivitythrough mass fabrication.100200300400500050100150200250300Slab Waveguide Thickness (nm)Sensitivity (nm/RIU)  TE slabTM slabTM strip 900Exp. 900TM strip 750Exp. 750Figure 4.5: Calculated resonator’s sensitivities, based on simulations, as functionsof silicon core thicknesses. The hallowed markers are the simulated sensitivities forthe case of a slab waveguide, the black filled markers are the simulated sensitivitiesfor the case of rectangular waveguides with waveguide widths of 750 nm and 900nm, and the red markers are averages of our experimental results for TM ringresonators with 150 and 220 nm thick silicon cores.The bulk sensitivity of the fabricated resonator sensors were experimen-tally validated using refractive index standards described above. The slopeof the best fit line through the different resonant wavelengths at each con-centration results in its bulk sensitivity. Each sensor was measured severaltimes on different days and the experimental observations were comparedwith simulation results in figures 4.5 and 4.7.To better understand and compare the simulation results with the ex-perimental observations, we included simulated sensitivities for rectangularwaveguides and interpolated the experimental observations at those designedpoints (figure 4.5). The sensitivity for the waveguide is plotted for a waveg-774.5. Resonator’s Bulk Sensitivityuide core thickness of 220 nm and width of 750 nm, as well as for waveguidethickness of 150 nm and width of 900 nm.140 160 180 200 220 240 26040060080010001200... ... Inf80100100120120140140160160160180180200 2002002002202202202402402402602602602602602602702702702802808290290300 340Waveguide Width (nm)Waveguide Thickness (nm)Resonator‘s Sensitivity (nm/RIU)Figure 4.6: Contour plot of sensitivity in nm/RIU as functions of waveguide withsand thicknesses. The cross-section corresponding to the dashed line representingthe slab is plotted in figure 4.5, and the other two cross-sections representing thethicknesses of 150 and 220 nm are plotted in figure 4.7. The red markers showthe fabricated TM resonator devices (Star: Width = 900 nm and Thickness = 150nm, Triangle: Width = 750 nm and Thickness = 220 nm )To further improve our model, we simulated the sensitivity of a TM modepropagating in a rectangular waveguide as a function of waveguide width andthickness. These results are illustrated in figure 4.6 as contour plots. Thesimulation area was fixed to 4×4 µm and only structures with less than 2%error in their sensitivity (based on the error calculated in the convergencetest) were considered. These results indicate that the maximum bulk sensi-tivity of 363 nm/RIU is achieved at a waveguide thickness of 165 nm. Theclosest thickness offered by MPW foundries to this optimum thickness is 150nm.784.5. Resonator’s Bulk Sensitivity4005006007008009001000110012001300150200250300350mode 2mode 3TM1 isTM1 is Waveguide Width (nm)Resonator‘s Sensitivity (nm/RIU)  150nm(Sim)220nm(Sim)150nm(Exp)220nm(Exp)Figure 4.7: Sensitivities of TM waveguide resonators to the aqueous cladding forthe optimum silicon core thickness of 150 nm and conventional thickness of 220nm as functions of widths. The filled markers show corresponding experimentalresults.The two vertical dashed lines in figure 4.6 represent the cross-sections forthe thicknesses offered by MPW foundries (150 and 220 nm). The sensitivi-ties for these two thicknesses, are plotted as functions of waveguide widths infigure 4.7, where hallowed markers indicate the simulations and filled markersindicate the experimental results. The slight variations, between the exper-imental and simulation results, partly come from the imprecision of salt-solutions (measurement error) and any thermal noise in tuning the stage,and can partly be explained with the non-uniformity that exist on SOI chips[77].794.6. Resonator’s Temperature Sensitivity - Simulation Results00.0020.0040.0060.0080.0100.511.522.5Refractive Index Change (RIU)Wavelength Shift (nm)  220 nm 150 nmFigure 4.8: Measured resonant peak wavelength shift as indices of refraction of thecladding media changes. The slope of the line indicates the sensitivities (nm/RIU)for the optimum silicon core thickness of 150 nm and conventional thickness of 220nm, to be 269 and 193 respectively.4.6 Resonator’s Temperature Sensitivity -Simulation Results400500600700800900100000. Width (nm)Temperature Sensitivity (nm/K)  150nm thick220nm thickFigure 4.9: Temperature sensitivities of TM guided mode in a strip waveguideresonator sensor as functions of waveguide core widths, for the two waveguide corethicknesses of 150 nm and 220 nm.804.7. Resonator’s Surface Sensitivity - Simulations and ExperimentsSensitivities of strip waveguide resonator sensors for TM guided modeswere calculated as described in section 3.2.2. Figure 4.9 shows the resultsin these simulations/calculations for two silicon thickness of 150 nm and 220nm.4.7 Resonator’s Surface Sensitivity -Simulations and ExperimentsSurface sensitivity of a resonator is defined as the sensitivity to the adsorbedbimolecular layer to the surface of the silicon core. Therefore, waveguide’ssurface sensitivity is defined as δneffδt , where t denotes the thickness of theadsorbed biomolecule. Subsequently, surface sensitivity of resonator is de-scribed as δλδt . The resonant wavelength shift as a function of the thicknessof the add-layer is approximately linear for small add-layer thicknesses (<30 nm). We approximate the simulated surface sensitivity of our fabricateddevices based on a 10 nm protein-like add-layer with refractive index of 1.48[78]. Using our simulation/analytical model for slab waveguide, the maxi-mum surface sensitivity of a TM resonator sensor is calculated to be at thethicknesses around 185 nm (figure 4.10). The surface sensitivity for TM res-onator sensors with waveguide dimensions of 750x220 nm and 900x150 nm(our fabricated devices) are simulated to be 300, and 245 pm/nm, respec-tively.814.8. Bio-Sensing Demonstration5010015020025000. Waveguide Thickness (nm)Surface Sensitivity (nm/nm)  TE MODETM MODEFigure 4.10: Calculated resonator’s surface sensitivities, based on simulations, asfunctions of silicon core thicknesses. The hallowed markers are the simulatedsensitivities for the case of a slab waveguide.4.8 Bio-Sensing DemonstrationA sandwich assay representing a model biological system was performedto demonstrate the biosensing capability of our quasi-TM mode ring res-onator’s and verify the surface sensitivity characterized using the electrostaticpolymers. Figure 4.11(b) shows the result of the biological sandwich useddemonstrate the sensors capability to detect molecular binding at its surface.The details of each reagent and step is described in detail in our previousmanuscript [76] with the relevant highlights restated here. The introductionof each reagent was followed by a 20 minute rinse using PBS buffer to removeany unbound species in the channel (shown by the short, black dashed linein Figure 4.11(b)). After achieving a signal baseline in a PBS buffer at 37degrees Celsius, Protein-A (1 mg/mL) was passively adsorbed to the sensors824.8. Bio-Sensing Demonstrationnative oxide surface prior to Region B to facilitate the immobilization andorientation of the capture antibodies [79, 80]. Region B shows the robustimmobilization of the IgG capture antibody, anti-streptavidin (10 µg/mL),to the Protein-A add-layer.The unintended introduction of an air bubble mid-way through the an-tiSA rinse cycle (Region B), indicated by the blue-dashed vertical line, doesnot impact the viability of the subsequent biological species or their function-ality. To show that subsequent molecular binding interactions are specific andto prevent unwanted, non-specific adsorption to the sensor, BSA (20 µg/mL)was introduced next (Region C) to block any exposed surfaces remaining onthe sensor (Region C). The slight negative shift in resonant wavelength afterthe BSA block and challenge (Region C) suggests that a small portion ofthe antibody add-layer lifts off but the original sensor coverage (Region B)was robust. Next, the functionalized and blocked sensor was subjected to10 µg/mL of SA, the model analyte captured by the antibody (Region D).The resulting shift in the sensors resonant wavelength after the buffer rinsesuggests specific and irreversible binding interaction as expected. To fur-ther demonstrate the sensors capability to detect molecular binding, biotinconjugated with BSA was introduced as a final amplification step. Biotinand SA have one of the highest non-covalent binding interactions known [81]resulting in another, permanent resonant shift (Region E).Protein A (with a diameter of approximately 3 nm [82, 83] and refractiveindex of 1.48 [78]) can approximately form a thin layer of 1-3 nm thick834.8. Bio-Sensing Demonstration(a)(b)Figure 4.11: Surface Sensing Experiment [courtesy of UW team]. a) Schematicrepresentation of the bio-assay material. b) Biosensing demonstration using amodel biological system.[28, 84]. Based on our simulated sensitivity values, a layer of 1 nm thickprotein layer would have covered around 50% of our sensor’s surface to resultin the observed shift. Subsequently, if a 3 nm layer was formed, about 16%surface coverage results in the observed shift. These results are within therange of surface coverage that was determined by other researchers [84].844.9. Discussion, Summary and Conclusion4.9 Discussion, Summary and ConclusionWe have investigated the optimum thickness of the waveguide core that re-sults in the highest sensitivity for a strip waveguide resonator sensor, withinthe constraints of MPW foundries. The results indicate that guided quasi-TM modes have higher sensitivities to the changes in refractive indices of thecladding media, since larger evanescent field component is traveling abovethe waveguide, where the target molecules exist. This affects the term δneffδncin the sensitivity relation and researchers have investigated the maximumsensitivity based on this term. However, for the case of the resonators, ngalso affects the sensitivity of the device (equation 4.4). Our investigationsdetermined that the optimum thickness for TM resonator sensors is around165 nm, which is close to one of the thicknesses offered by MPW foundries(150 nm). The compatibility of these resonators with the standard CMOSprocesses and MPW foundries, in terms of their minimum feature size require-ments as well as offered thicknesses, make them a cost-effective candidate fora sensor.The measured Q value, from the spectra of these resonators in water,are 10100 and 4500 for 220 nm thick and 150 nm thick resonator sensorsrespectively. These give intrinsic limit of detection (iLoD) of approximately7.5 × 10−4 RIU and 1.2 × 10−3 RIU for 220 nm thick and 150 nm thickresonator sensors respectively. The extinction ratio for both these deviceswas measured to be around 30 dB. The lower Q value associated with 150854.9. Discussion, Summary and Conclusionnm thickness, compared to the conventional 220 nm thick silicon core, ispartly due to lower group index associated with this thickness, and partlydue to higher intrinsic losses (bend loss, mode-mismatch loss, substrate leak-age loss). Assuming that these resonators are close to critical coupling, theirQ values translate to distributed losses of 34 dB/cm and 48 dB/cm for theresonators with 220 nm and 150 nm thick silicon core respectively. Basedon simulations, material absorption losses, including water absorption, areresponsible for approximately 23 dB/cm and 19 dB/cm of these losses for theresonators with 220 nm and 150 nm thick silicon core, respectively. The esti-mated losses based on the dry measurements of these devices are 14 dB/cmand 28 dB/cm for the resonators with 220 nm and 150 nm thick siliconcore, respectively. These values approximately exclude the losses due towater absorption. As mentioned in chapter 3, the scattering losses are usu-ally induces by the interaction of the propagating mode with the sidewallroughness [66, 70, 72]. Therefore, TM modes experience less scattering losscaused by sidewall roughness because the higher proportion of the guidedmode is above and below the waveguide and, hence, the guided mode inter-acts less with the roughness in the sidewalls (resulted from fabrication). Apropagation loss of ∼ 3 dB/cm for TE guided mode in a strip waveguide,dominated by sidewall scattering, is demonstrated experimentally and ana-lytically [44, 66, 70, 72]. It has also been shown that thinner waveguides,fabricated in CMOS-compatible processes, exhibited lower scattering losses,in the order of 2 dB/cm [58]. Therefore, we estimate the scattering losses864.9. Discussion, Summary and Conclusionfor TM modes to be less than 3 dB/cm. The remaining losses (∼ 9 dB/cmand 26 dB/cm for the resonators with 220 nm and 150 nm thick siliconcore respectively) are contributed by bend losses (radiation loss, and modemismatch loss), substrate leakage, and coupling loss. Propagation losses, in-cluding the scattering loss and substrate leakage, as well as bend losses canbe improved by using wider waveguides. The loss due to substrate leakagecan be improved using SOI wafers with thicker silicon oxide substrates (e.g.SOI wafers with 3 µm thick silicon oxide substrate).Simulations supported by experimental results in this chapter demon-strate a novel approach to achieve higher sensitivity within the constraintsof MPW foundries, which has numerous advantages such as possibility ofmass fabrication and integration with CMOS circuitry for potential system-on-chip implementations.87Chapter 5Integrated PhotodetectorSensors 5A resonance-enhanced, defect-mediated, ring resonator photodetector hasbeen implemented as a single unit biosensor on a silicon-on-insulator plat-form, providing a cost effective means of integrating ring resonator sensorswith photodetectors for lab-on-chip applications. This method overcomesthe challenge of integrating hybrid photodetectors on the chip. The demon-strated responsivity of the photodetector-sensor was 90 mA/W. Devices werecharacterized using refractive index modified solutions and showed sensitivi-ties of 30 nm/RIU.5.1 IntroductionVarious types of silicon photonic resonators have been previously demon-strated with sensitivities sufficient for bio-molecule detection [17, 28, 30, 34,5A version of this chapter have been published in [85]:S. Talebi Fard, et al., ”Silicon-on-insulator sensors using integrated resonance-enhanceddefect-mediated photodetectors.” Optics Express 22, no. 23 (2014): 28517-28529.885.2. Resonator Sensors and Defect-Mediated Photodetectors35], however, these demonstrations did not use an integrated photodetector.Hybrid photodetectors have been demonstrated using III-V materials [86–88] or germanium [89–93] on the SOI platform. However, these detectorsrequired the use of relatively complicated and expensive processes, which isnot in keeping with the goal of providing an inexpensive lab-on-chip. Toavoid the need for implementing such hybrid photodetectors, we suggest theuse of defect-mediated based photodetectors on SOI [94–98]. In this chap-ter, we propose integrating such photodetectors [94–98] with well-developedand characterized resonator-based sensors [17, 28, 30, 34, 35]. To achievethis, we brought together the expertise of two groups, one at McMasteruniversity, and one at the University of British Columbia (UBC), wherethe McMaster group’s expertise is in ion-implanted defect-mediated basedphotodetectors [94–98], and UBC’s expertise is in resonator-based sensors[17, 28, 30, 34, 35, 99].Here, we investigate and demonstrate a ring resonator sensor with anintegrated detector, which exploits the resonance enhancement of the ringfor increased responsivity.5.2 Resonator Sensors and Defect-MediatedPhotodetectorsIn this section, we review the performance and characterization of defect-mediated photodetectors, which have been characterized by Knights et al.895.2. Resonator Sensors and Defect-Mediated Photodetectors[94–98], as well as the resonator sensors developed in our group [17, 28, 30,34, 35]. Further, we discuss the novel integration of these photodetectorswith a resonator-based sensor.5.2.1 Defect-Mediated PhotodetectorsSilicon is a relatively poor material for photodetection in the near-infraredbecause the photon energy is less than the silicon bandgap resulting in negli-gible optical absorption. To overcome this, defect-mediated silicon waveguidedetectors have been proposed and characterized [94–98, 100]. These photode-tectors are formed by creating a p-i-n diode across a waveguide structure.An inert ion implantation into the intrinsic region causes damage to the sili-con lattice which introduces deep levels in the bandgap through which lightwith energy less than that of the bandgap can be absorbed. The latticedamage-induced absorption, which increases the propagation loss to 20-100dB/cm [96], causes the absorbed light to produce electron hole pairs thatcan be extracted by applying a reverse bias to the p-i-n diode. A subse-quent low temperature anneal can be used to partially repair the damage toachieve an optimum defect concentration. The increased absorption due tothe presence of the defects is modest, so, if a linear detector were to be used,it would have to be long in order to absorb a large fraction of the incidentlight and, thereby, achieve a high responsivity. Alternatively, the effectiveresponsivity can be increased by using a resonant structure which makes useof the intensity buildup within the resonator to achieve increased absorption905.2. Resonator Sensors and Defect-Mediated Photodetectorswithin a small footprint [101]. In these resonant detectors, the increased op-tical loss associated with the defects reduces the quality factor and thereforeintroduces a tradeoff between the responsivity and the precision of sensing.Defect-mediated silicon photodetectors fabricated in this manner have shownresponsivities of 0.5 to 10 A/W [100].5.2.2 Evanescent Field-Based Resonator SensorsEvanescent Field (EF) sensors operate by detecting a change in the effectiverefractive index of a waveguide [9, 32, 102, 103] caused by changes in the con-centration of an analyte in the cladding (bulk sensing) or as molecules bindto a functionalized surface of the waveguide core (surface sensing). Opticalresonators, having small footprints and longer interactions with the ana-lytes in their cladding media (resulting in higher sensitivities), are promisingcandidates for sensing applications. SOI-based resonator sensors with smallfootprints, each integrated with a detector, would make high-throughputand/or simultaneous analysis of multiple analytes possible, where many sen-sors, each designed to measure a different target could be integrated on asingle chip. These chips could be designed, with various sensors aligned, touse microfluidic channels to deliver various targets to the sensors, all in asingle system. Various resonator sensors have been developed and charac-terized , some of which include disk resonators[28]; strip waveguide and slotwaveguide ring resonators [17]; strip waveguide Bragg gratings [34]; and slotwaveguide Bragg gratings [30]. These resonator sensors were shown to have915.2. Resonator Sensors and Defect-Mediated Photodetectorssensitivities and limits of detection approaching the theoretical maximumvalue (based on the case of a resonator in water). We have also characterizedand validated these sensors using a standard ‘sandwich’ assay [34]. Hav-ing fabricated these sensors using a CMOS compatible SOI wafers, througha Multi-Project Wafer (MPW) foundry, provides the potential for integra-tion of these sensors with on-chip electronics, and makes the integration of acomplete system on a chip possible.5.2.3 Defect-Mediated Photodetector ResonatorSensorWe propose integrating defect-mediated photodetectors with previously demon-strated resonator-based biosensors, such as ring [35], disk [28], and phase-shifted Bragg grating resonators [30], to create a biosensor with integratedphotodetectors (photodetector-sensor). Figure 5.1 shows schematic drawingsof how defect mediated photodetectors could be integrated with each type ofbiosensor. A top view of each sensor shows the doped and detection regions.In each case the defect-mediated detector is introduced across the waveguidethat forms the resonator cavity.The 3D rendering below each schematic view provides perspective onhow the defect region and detector can be oriented in a fluidic channel forbiosensing applications. The biomolecules would be introduced across theentire device area but only those in proximity to the resonator waveguide925.2. Resonator Sensors and Defect-Mediated Photodetectors(a) Ring resonator sensor	 (b) Disk resonator sensor	 (c) Bragg grating sensor	p++	n++	Implanted region	In/out waveguide	 Implanted region	In/out waveguide	p++	n++	Implanted region	Bragg grating resonator	In/out waveguide	n++	p++	p++	 n++	Implanted region	 In/out waveguide	p++	 n++	Implanted region	p++	 n++	Implanted region	Bragg grating resonator	Flow channel	 Flow channel	 Flow channel	Figure 5.1: Schematic representations of proposed designs for biosensing: a)a defect-mediated ring resonator photodetector-sensor; b) a defect-mediateddisk resonator photodetector-sensor; and c) a defect-mediated Bragg gratingphotodetector-sensor. The region of the detector with defects is highlighted inbold cyan. The remaining waveguide is shown in beige.are interrogated. The ring resonator, Fig. 5.1(a), and disk resonator, Fig.5.1(b), share similar configurations except that the disks are not etched inthe center and only have one exterior side-wall. The ring, however, has twosidewalls formed at the inner and outer radius of the ring. The n++ implantsare shown in red while the p++ implants are shown in blue. This forms a PINstructure, with the detection occurring in the “I” region which has defects.The operation of each type of defect-mediated photodetector resonatorsensor is very similar. The defect-mediated photodetector in the resonancecavity senses the optical power. Since the buildup of power inside the res-935.3. Design Considerations and Layoutonator is most significant at the resonance wavelength, by measuring thephotocurrent as the wavelength is varied, one can find the wavelength forwhich the intensity buildup is maximum and determine the resonance wave-length. Furthermore, since each resonator forms an evanescent field-basedsensor as described in Section 5.2.2 the resonance wavelength shifts as therefractive index of the cladding changes. Thus, in this configuration theresonator allows the device to sense refractive index changes while the in-tegrated detector allows the simultaneous readout of the sensor response inone device.5.3 Design Considerations and LayoutAs a proof of concept, two variations of ring resonator photodetector sensorslike that depicted in Fig. 5.1(a) were designed and fabricated. A cross-sectionof the detector is shown in Fig. 5.2. In order to integrate the photodetectorwith the ring resonator, rib waveguides, as opposed to strip waveguides, areused to allow for electrical contact with the junction to extract photocarriersthat are generated in the waveguide. Metal contacts are made to the highlydoped n++ and p++ slab regions far away from the optical mode that isconfined to the waveguide such that the metal does not induce significantloss. The cladding oxide is removed above the resonator so that the analytetest solution can interact with the optical mode.We fabricated two device variations (designs A and B) that have differ-945.3. Design Considerations and LayoutFigure 5.2: Schematic of detector cross-sectionent tradeoffs between sensitivity, responsivity, and quality factor. The tworesonator designs are shown schematically along with microscopic images offabricated devices in Fig. 5.3.In design A, the oxide cladding is removed over 75% of the ring area toexpose the ring to the analyte. Electrical contact from the metal routingto the ring interior is made through a contact via in the remaining oxidecladding as shown in Fig. 5.3(a). This design allows contact to be made tothe center of the ring without introducing significant optical loss because themetal trace only crosses over the waveguides when it is above the claddingoxide, at the expense of a reduced sensitivity due to the smaller region overwhich the analyte interacts with the optical mode. In design B, the oxidecladding is removed over the entire ring, so the analyte can interact with theentire ring. In this case electrical contact to the silicon is made through a viaoutside the ring and contact is made to the ring interior by a conductive traceformed of highly doped silicon shown as the region of p++ doping crossingthrough the ring waveguide in Fig. 5.3(c). In this design a greater sensitivitycan be achieved by allowing the optical mode to interact with the analyte955.3. Design Considerations and Layout(a) (b)(c) (d)Figure 5.3: a) Schematic representation of design A (where the metal contactpasses above the coupling region of the ring) ; b) Microscopic picture of designA; c) Schematic representation of design B (where the highly-doped silicon wirepasses through the waveguide); d) Microscopic picture of design B. In Figs (a) and(c), the gray overlay indicates where the cladding oxide was removed.over the entire length of the ring. This benefit comes at the cost of an increaseof absorption loss and lower quality factor because the highly doped tracemust cross the waveguide.In both designs A and B the resonator is in a racetrack configuration withradius 40 µm and coupling lengths are 4.2 µm and 5.1 µm, respectively, each965.3. Design Considerations and Layoutwith a 200 nm coupling gap. The racetracks were designed for critical cou-pling using Lumerical MODE Solutions and analytical models in MATLAB.The bend and mode-mismatch losses were simulated using MODE, and prop-agation loss was assumed to be around 3 dB/cm [44]. The induced loss dueto implanted defects and proximity of highly doped regions were estimatedto be around 10 dB/cm to 30 dB/cm [104, 105]. A few coupling lengths werecalculated based on these assumed values of losses (where the assumed lossdue to defects were varied from 10 to 30 dB/cm). For design A, a couplinglength of 4.2 µm, corresponding to the assumed average loss of 20 dB/cm dueto defects, resulted in a spectrum close to critical coupling. In this design,75% of the resonator is implanted. For design B, a coupling length of 5.1 µm,corresponding to the assumed loss of 30 dB/cm, resulted in a spectrum closeto critical coupling. In this design, the entire ring is implanted, and in addi-tion a trace of p++ doping crosses the waveguide and therefore higher lossesare expected compared to design A. The bus and resonator waveguides are500 nm wide rib waveguides, formed by a 130 nm thick waveguide on a 90nm thick slab. As shown in Figs. 5.2 and 5.3, the p++ and n++ dopingregions are 500 nm from the edges of the waveguides. We choose the dis-tance from the waveguide to the p++ and n++ contact regions based on twoconsiderations: the doped regions must be sufficiently close to the waveguideto extract the generated carriers efficiently before they undergo carrier re-combination, and the distance must be large enough so that the optical lossfrom the interaction with the highly doped regions is not too high.975.4. Experimental Methods and Materials5.4 Experimental Methods and MaterialsThe devices were fabricated at The Institute of Microelectronics in Singapore,up to the point at which the ion implantation was performed to introducedefects in the waveguide. For this step, implantation was performed at Mc-Master University. Boron ions were implanted at an energy of 30 keV anddose of 5 × 1012 cm−2. No post-implantation high temperature anneal wasperformed such that the boron remained inactive and thus only structural de-fects were introduced to the waveguide (i.e. no chemical doping took place).On-chip grating couplers (GCs) facilitate the efficient coupling of light intoand out of the SOI chip [68], and electrical contact to the detectors is madeby routing metal traces to the device from large on-chip contact pads thatcan be externally probed. This sensor was characterized using the techniquesand reagents described in this section.5.4.1 Experimental SetupOur measurement setup used: a tunable laser (Agilent 81682A, Agilent Tech-nologies, Inc., USA) with a wavelength range of 1460 nm to 1580 nm as theoptical source; an optical power sensor (Agilent 81635A, Agilent Technolo-gies, Inc., USA) to measure the output light intensity; and a source mea-surement unit (Keithley 2602, Keithley Instruments, USA) to perform theelectrical characterization. Light was coupled into and out of the chip usingGCs and an array of polarization maintaining (PM) optical fibers (PLC Con-985.4. Experimental Methods and Materialsnections, LLC., USA). The PM fiber array was aligned to the GCs on thechip using a motorized stage that was controlled using an automatic align-ment routine, to obtain maximum optical coupling to the devices. Electricalcontact to the detectors was made by routing metal traces to the device fromlarge on-chip contact pads that were externally probed, using a needle.5.4.2 Reagents and Microfluidic SetupTo determine the bulk sensitivity, and to characterize the performance of ourphotodetector-sensors, standard aqueous solutions of sodium chloride (withconcentrations of 0, 500 mM, 1 M, and 2 M) were characterized using aReichert AR200 Digital Refractometer (Depew, NY). Polydimethylsiloxane(PDMS) microfluidic channels, with widths and heights of 200 µm and 80 µm,respectively, were aligned to the sensors and reversibly bonded on the chipto facilitate the delivery of standard solutions to the cladding media of ourphotodetector-sensors. Prior to bonding the PDMS to the chip, inlet andoutlet holes for microfluidic channels were punched in the PDMS using a0.5 mm coring tool (Schmidt Press, CORP, USA). After bonding the PDMSto the chip, the channels were linked to tygon micro-bore tubing using 21gauge blunt needles (Nordson EFD, CORP., USA) through the punched inletand outlet holes. These tubes were used as the fluidic inlet and outlets andwere connected to a syringe. To maintain a constant flow rate of the reagentsover our photodetector-sensor, a syringe pump (KDS-230, KD Scientific, Inc.,USA) was set to operate in withdraw mode (negative pressure) at a constant995.5. Performance Analysis and Resultsrate of 19 µL/min.5.5 Performance Analysis and ResultsPrior to device characterization, the detectors were forward biased to producea forward current of 1 mA for 5 minutes [106]. The current-voltage measure-ment for both designs is plotted in Fig. 5.4. The implanted photodetector-sensors were first tested under various reverse bias voltage (-2 to -35 V)conditions to determine a suitable operation point. In order to achieve a rea-sonable signal-to-noise (SNR), without unduly stressing the device, we chosea bias voltage of -10 V for operation.−30−20−100−0.2−0.15−0.1−0.0500.05Bias Voltage (V)Current (mA)  A BFigure 5.4: IV curves for designs A and B.To further characterize the devices, we measured both the optical andthe electrical responses of our photodetector-sensors; to do this we swept thelaser’s wavelength while simultaneously measuring the optical power trans-1005.5. Performance Analysis and Resultsmitted through the ring and the current generated by the integrated detector(with a bias of -10 V). The optical power at the output of the laser was 0dBm. There are various losses that contribute to the overall loss in the sys-tem. The primary dominant loss is associated with coupling from the fibersto the waveguides through the grating couplers. There are long ( > 5 mm)routing waveguides from the input GC to the device and from device to theoutput GC which also contribute to the loss. However, due to the symmetryof the design, the loss from the output of the optical power to the device isequal to the loss from device to the detector connected to the output fiber.Therefore, the total loss is divided by two to approximate the power levelof the optical signal that reaches the device. Figure 5.5 shows the measuredelectrical and optical response of design A and design B devices. It can beseen that when the ring is on resonance there is an increase in the photocur-rent due to the resonant buildup of energy in the ring.15461546.51547024Wavelength λ (nm)Current magnitude (µA)  15461546.51547012 Optical Power (µW)current magnitudeoptical power(a) Design A1548.515491549.5024Wavelength λ (nm)Current magnitude (µA)  1548.515491549.5012 Optical Power (µW)current magnitudeoptical power(b) Design BFigure 5.5: Measured electrical and optical responses of our implanted detectorresonator sensors: a) Design A, and b) Design B.1015.5. Performance Analysis and ResultsTo estimate the responsivity of a device, the losses in the path from thelaser output to the input of the device and from the device output to thedetector were assumed to be equal, due to the symmetry of the design, giv-ing an estimate of the power seen at the ring which was 0.045 mW and0.04 mW and a calculated responsivity of approximately 0.09 A/W and 0.11A/W for design A and design B, respectively. To calculate the dark currentof a single device, the neighbouring devices had to be taken into account,since, to save space on the chip, several resonator sensors were designed tobe wired together and to share common electrodes. This had the drawbackthat the dark current from the detectors wired in parallel was summed andthe contribution from a single detector could not be directly measured. Toestimate the dark current of only the detector under test, the dark currentper unit length of each detector was assumed to be equal and the measuredcurrent was scaled by the ratio of the length of the detector under test to thecombined lengths of all of the detectors that were in parallel. For the devicesreported on here, the measured dark current calculated in this manner was17 nA for the design A device and 30 nA for the design B device. The corre-sponding Noise Equivalent Power (NEP) for these two designs are 0.19 µWand 0.27 µW. Considering that the peak currents for designs A and B deviceswere 4 µA and 4.5 µA, respectively, the highest SNRs that were achieved were235 and 160, respectively. Table 5.1 summarizes the characterization results.The relatively low quality factors (Q’s) for these photodetector-sensors, ascompared to ring resonators without integrated detectors, is due to the loss1025.5. Performance Analysis and ResultsTable 5.1: Summary of the performances of designs A and B devicesDark Current NEP SNR Responsivity Q(nA) (1 µW) (A/W)Design A 17 0.19 235 0.09 5800Design B 30 0.27 160 0.11 4500introduced by the ion-implantation in the silicon waveguides. As expected,design B has a lower Q than design A due to the loss associated with thehighly doped silicon trace crossing the waveguide. Subsequently, for the pur-poses of this project, we proceed with the sensitivity analysis and experimentson the design A device only.In order to characterize the sensitivity of the photodetector-sensor, theoptical and electrical responses were measured simultaneously while the res-onator was exposed to various salt solution concentrations. From the opticalresponses of the ring resonator in water, the Q for the design A device wasmeasured to be approximately 4400; and the extinction ratio was measuredto be approximately 30 dB. The measurements were repeated several timesfor each concentration. The location of the resonant peak is the impor-tant information for sensitivity measurements, which can also be found bycurve fitting the response spectra. Figure 5.6(a) shows how the normalizedelectrical response shifts as a function of wavelength as we change the con-centration of the solution. As we can see, the changing concentration of the1035.5. Performance Analysis and Resultsreagent results in a resonance wavelength shift of the ring resonator. Figure5.6(b) shows the resonance wavelength shift plotted against the salt solu-tion’s change in refractive index for various concentrations; the slope of thelinear best fit gives a sensitivity (S) of approximately 30 nm/RIU. Basedon measured values of Q and S, the intrinsic limit-of-detection (iLoD) forthis sensor is 1.1 × 10−2 RIU. Considering that 75% of the ring is exposedto the solution, and that the simulated sensitivity for this waveguide is 40nm/RIU, this yields a predicted sensitivity of 30 nm/RIU, in agreement withour experiments.154815491550155100. (nm)Normalized Current  0 M 0.5 M 1 M 2 M 0 M(a)00.0050.010.0150.0200. index change (RIU)Wavelength shift (nm)(b)Figure 5.6: a) Normalized electrical response of photodetector-sensor design Awhen exposed to various salt solution concentrations. b) Resonant wavelengthshift as a function of the change in refractive index of the solution, for designA. The slope of the best fit line, defined as the sensitivity, is approximately 30nm/RIU.1045.6. Analysis and Discussion5.6 Analysis and DiscussionIn this chapter, we have investigated and demonstrated the integration ofa resonator sensor and detector into one device. If, instead, the resonance-enhanced defect-mediated photodetector and the resonator sensor were de-signed as two separated, cascaded devices, then, the resonant wavelengths ofthe two would need to be matched and therefore, a wavelength locking mech-anism would be required. However, having them as a single photodetector-sensor overcomes this wavelength mismatch issue and thus reduces the com-plexity of making measurements.5.6.1 Performance TradeoffsThe various advantages of using this novel approach for easier and cost-effective realization of a system on chip come with tradeoff that some of thesensing performance is sacrificed. Researchers developing defect-mediatedphotodetectors have investigated the optimum defect concentration (Nt =2 × 1017 to 2 × 1018 cm−3 ), and consequently the optimum ion implan-tation dosage, for maximum efficiency and detector responsivity in variousstructures [98, 104, 105]. Although these defect-mediated detectors havebeen shown to have performances well comparable with and sometimes bet-ter than Ge detectors [106], the integration of the detector into the sensorreduces the Q factor of the resonator (by contributing extra loss on the or-der of 20 to 30 dB/cm depending on the defect concentration [104–106])1055.6. Analysis and Discussionwhich, in turn, degrades the performance of the sensor. We can quantify thisreduction in performance by comparing the performance of our integratedphotodetector-sensor to a hypothetical device fabricated using the same fab-rication process without having an integrated detector. If we were to designthis sensor without an integrated detector, we could have designed the res-onator to have 100% of its surface exposed to the solution, which would in-crease the sensor’s sensitivity to 40 nm/RIU. In addition, the introduction ofdefects in the waveguide degraded the Q by about 1000 on average (observedexperimentally and confirmed through calculations). Therefore without thedefects, the Q would be about about 5400 which would increase the iLoDfrom 1.1× 10−2 (for the integrated detector) to 0.7× 10−2 RIU. In exchangefor this slightly lowered performance, the integrated device offers a host ofadvantages including reduced system footprint and less expensive fabricationprocess; furthermore, the iLoD of the integrated device could potentially beimproved by optimizing the device geometries.5.6.2 Discussion and Analysis of Device Performancefor Sensing ApplicationsIntegrated photodetector-sensors, like the one presented here, can be usefulfor many sensing applications, including the detection of aqueous solutions ofchanging concentration, gas mixtures, or molecules adsorbed to the surface ofthe sensor. We have previously studied the SOI-based resonator sensors and1065.6. Analysis and Discussionhave demonstrated both their bulk refractive index sensitivity and surfacebiosensing capabilities through experiments and simulations [17, 28, 34, 35,76].For applications which require detecting the concentration of specificmolecules in the solution, a change in concentration results in a changein refractive index of the solution. In this application, the sensor’s sen-sitivity to the concentration of specific molecules can be calculated fromthe sensor’s bulk refractive index. For example, the relationship betweenglucose concentration and RI unit change is described as ∆nglucose(λ) =2.7× 10−5∆Cglucose[mM ] [35, 46]. This translates to a sensitivity of approx-imately 810 pm/M for a device with sensitivity of 30 nm/RIU. Using thesame method, the sensitivity for other aqueous solutions and gas mixturescan also be determined.We have previously demonstrated the capability of our SOI-based res-onator sensors in responding to the adsorbed molecular layers using standardsandwich assays [28, 34, 76]. The experimental results from these previousdemonstrations agree with the simulations. In addition, we have confirmedthat sensors offering impressive bulk refractive index sensitivities and iLoDshave also performed well in the surface sensing experiments. In the integrateddevice, we have observed a sensitivity of 30 nm/RIU which is 75% of the sim-ulated value of 40 nm/RIU. Considering that only 75% of the resonator isexposed to the solutions, this shows that our device’s performance agreeswith simulations. Based on the agreement between previous experiments1075.7. Summary and Conclusionand simulations, we expect that we can reliably predict the performance ofdevices in other applications such as biosensing. For example, if there is abimolecular add-layer, with refractive index of 1.48 (to mimic proteins), thesensitivity of the non-integrated device (with full surface exposed) to theadd-layer thickness is 0.12 nm/nm based on simulations. This sensitivity re-lation is approximately linear for small add-layer thicknesses of 0 to 30 nm.However, since 75% of the resonator is exposed in the integrated case, weexpect a sensitivity of about 90 pm/nm to the thickness of add-layer withrefractive index of 1.48.The integrated device offers many advantages for potential system on chipapplications as well as CMOS integration with a small footprint and less ex-pensive mass fabrication process; these advantages might outweigh the sac-rificed performance for many applications (particularly low-cost applicationsor those requiring disposable chips). In addition, the sensing performance ofthe device could be improved by optimizing the geometries and/or using TMpolarization that can potentially provide higher sensitivity [56, 103].5.7 Summary and ConclusionTo avoid using expensive hybrid fabrication of photodetectors for resonatorbiosensors and system integration on SOI, e.g., using III-V materials or ger-manium, we proposed the use of a defect-mediated photodetector-sensor.The absorption of the photodetectors were enhanced using resonant struc-1085.7. Summary and Conclusiontures, giving improved responsivity and SNR, compared to a linear detector.Resonant structures were also shown to be promising biosensors. In this pa-per, we proposed and demonstrated a novel approach to integrate a sensorwith a photodetector by integrating, as a single device, our well-developedresonator-based sensors with defect-mediated photodetectors.The single-device integration of a photodetector-sensor was investigatedby implementing two designs. In each design, the photodetector-sensor couldinteract with the target analyte through a window in its oxide cladding,where the cladding otherwise provides isolation for the electrical contactsmade with the photodetector. In one of these designs, metal routing had topass over the coupling region, resulting in only a portion of the total length ofthe photodetector-sensor being exposed to the analyte; whereas in the otherdesign, highly doped silicon was used to route the electrical signal, allowingthe full length of photodetector-sensor to be exposed. In the second design,in order to make contact to the inside of the ring, a portion of the highlydoped silicon had to pass through the resonator waveguide. This resultedin additional loss in the resonator, and, therefore, a relatively lower Q forthe second design. Hence, we performed sensitivity analysis on the designwith the higher Q. Our photodetector-sensor had a measured responsivity of0.09 A/W, sensitivity of 30 nm/RIU, quality factor of 4400 (in water), andextinction ratio of 30 dB.The integration of a sensor with a photodetector removes the need for anexternal photodetector to interrogate the sensor and provides a significant1095.7. Summary and Conclusionstep toward the development of CMOS compatible, SOI-based systems ona chip. Implementing a defect-mediated photodetector in a resonator-basedsensor (photodetector-sensor) allows for reductions in the device footprint,optical loss, complexity, and cost.There are two factors to consider when choosing the defect concentra-tion for a detector-sensor: 1) a defect concentration that is too high reducesthe Q and iLoD (assuming constant S), thus, degrading the sensor’s perfor-mance [17, 34]. 2) a defect concentration that is too low degrades the de-tector’s performance by reducing the responsivity and SNR of the detector[105], while there is an optimal point for implant concentration and detec-tor’s performance [105]. These opposing trends impact the overall system’sperformance.In this chapter, we demonstrated a basic proof of concept for an inte-grated ion-implanted photodetector-sensor. Clearly, we can improve thesephotodetector sensors based on concepts discussed in the previous chapters.For example, the lower sensitivity resulting from the TE guided mode in therib structure can be improved by operating using the TM mode. Simulationsshow six times improvements in sensitivity when using propagating TM modein rib waveguides.110Chapter 6Differential Sensing SystemEvanescent Field (EF) optical sensors have been widely investigated and usedas biological and chemical sensors by researchers. However, the stability, re-peatability, and accuracy of these sensors have always been a concern. Thesesensors, like all other optical devices, are affected by noises from variouscomponents/instruments in the experimental setup and the device’s imper-fections. In addition, the response of these sensors is significantly affectedby common and inevitable environmental variations. In the experimentalsetup, the noise or error due to instrumentation (laser source and detector)can be classified to 1) amplitude intensity noise and 2) wavelength noise.The noise in the amplitude of the intensity are usually due to photodetec-tor noises (shot noise, etc), and output power variations of the light source(tuneable laser) [107–109]. The effect of amplitude noise for the resonatorswith high extinction ratio (>15dB) is insignificant and can be further re-duced to a more negligible value using fitting methods (e.g. lorentzian fit)[42, 107]. The wavelength noise in the instrumentation of the setup is mostlydue to light source’s wavelength repeatability, stability, and accuracy. Forthe case of our light source, the tuneable laser, the worst case of these noises1116.1. Introductionis ∼ ±1pm for repeatability and stability [110].Various groups have investigated the sources of noise in resonator sen-sors and temperature has been determined to have the most dominant andsignificant affect [107, 109, 111–113]. The resonant wavelength of the EFresonator sensors is strongly affected by temperature variations (∼ 80 pm/Kfor standard 220nm SOI sensor).EF optical sensors can have a more stable performance in the presence ofa reliable reference sensor that can capture and correct for dominant envi-ronmental variations such as temperature changes. To address the challengesof current EF sensors, a system of partially orthogonal sensors with varioussensitivities is proposed and investigated. This system mimics a multivariatesystem. The aim of this design is to correct for correlated errors and driftsdue to environmental variations.6.1 IntroductionEvanescent field optical sensors based on Silicon-on-Insulator (SOI) haveshown promises for biosensing applications [17, 34], ranging from basic med-ical research to home healthcare. These sensors provide label-free detectionand estimation of concentration, presence of a molecule, or a molecular bind-ing event. Fabricated devices seem to be remarkably affected by processvariations such as those caused by environmental and external conditions.Among fabricated sensors, we have observed a lot of variability, from excel-1126.1. Introductionlent performance to unpredictable or poor response. From a system’s pointof view, the variability could strongly reduce the usage of these sensors be-cause of poor repeatability. It also decreases the high throughput that isalways associated with microelectronics and SOI-based integrated systems,fabricated through commercially available foundry services.Data processing and chemometrics can improve the system’s performance[114, 115], but a well-inclusive reference is required to be able to remove asmuch correlated and orthogonal noises as possible. (Orthogonal noise meansthat the undesirable variations are independent of the variation of interest.Correlated noise means that the variations affect the properties that aremeasured to predict/calculate the change of interest.) One of the majorchallenges of most sensing systems (e.g. bio-sensors) is to find a properreference signal that accounts for most of the variations that are not relevantto the variations of the analyte of interest.Researchers have proposed and tried various methods of having a ref-erence sensor. Methods include having identical sensors but covering thereference sensor with a cladding material such as SiO2 or PMMA. There-fore, it does not come into contact with the analyte under study and onlypicks up the temperature variations [54]. However, a reference sensor undera material with different thermal properties might not reflect the temper-ature variations accurately and timely. In fact, it has been observed thatdrifts are hypothesized to be due to leakage of fluid into the cladding of thecovered reference sensor [54]. This can also mean a change in thermal prop-1136.1. Introductionerties of the material and the system relations. The other common methodof implementing a reference signal is having a reference sensor in anotherfluidic channel which does not experience the analyte changes [116]. Thismethod is sometimes used as sensing reference and, other times, as biologicalreference or control experiment. The difference between the sensing signal(from analyte channel) and reference signal (from buffer solution) determinesthe corrected sensing signal. This method assumes that the buffer solutionin the reference channel experiences the same variations as the analyte inthe sensing channel. Using this method as sensing or temperature referencemight not be the most ideal as the two sensors can be far away. Thus, it isdifficult to make sure that the temperature of solutions in both channels arethe same. This method might require advanced temperature control. Havinga separate channel for biological control experiments is a common practice.This control channel can also benefit from a reference temperature sensor tocorrect for the possible variations between the temperatures of the solutionsin two channels.In this chapter, we propose and investigate a system model capable ofcorrecting the common external and correlated drifts and noises (i.e. tem-perature variations). This method can be beneficial and applicable to anysensing application, including biological applications.1146.2. System Design and Methods6.2 System Design and MethodsA system with a single resonator would respond to both cladding changes(including the change of interest) and changes in other components and prop-erties of the system (noise in the system). Changes in optical propertiesof the waveguide due to environmental variations such as temperature andcommon mode shifts due to the source or detector, are examples of systemnoises. The error bars due to noises (unwanted variations and drifts) in oursystem are correlated and dominated by common mode contributions fromeach sensor. Sensors with various and significantly different sensitivities tocladding medium and to other contributing variations (e.g. temperature) inone channel would allow the rejection of these common mode variations.Therefore, to address the challenges of bio-sensor systems, we proposea combination of multiple sensors with significantly different sensitivities inthe same channels; i.e. both data channel (i.e. on the same waveguide bus)and fluidic channel. In such a design, all sensors in a single channel captureboth the unwanted variations that causes drifts, such as temperature, andvariations of interest, such as concentrations. Each sensor’s response to thesevariations differ based on their specific sensitivity to that variation (S). Asa result, solving for the variation of interest in the system of equations (i.e.a weighted difference between the resonant peaks of these sensors) wouldcorrect a dominant part of the common/correlated noises. Thus, a moreaccurate and direct indication of the variations of interest in the analyte1156.2. System Design and Methodswill be obtained. This can be achieved with any set of sensors that can becascaded and have significantly different sensitivities to their surroundingchanges.We present the general case of our system proposal with the n dimensionalmatrix representation below:∆λ = S ∆Var (6.1)∆λS1∆λS2...∆λSn=S1S2...Sn∆V arx∆V ar2...∆V arn(6.2)∆λS1∆λS2...∆λSn=( δλδV ar1 )S1 · · · (δλδV arn)S1( δλδV ar1 )S2 · · · (δλδV arn)S2... · · ·...( δλδV ar1 )Sn · · · (δλδV arn)Sn∆V arx∆V ar2...∆V arn(6.3)( δλδV ari )Si  (δλδV ari)Sj for i 6= jwhere V ar1 to V arn are the n varying properties of the system. V ar1 in-cludes, and dominated by, the variable of interest (V arx), but it is alsoaffected by all other variable properties. This represents a multivariate sys-tem for sensing applications, where a higher degree of orthogonality betweensignals is desirable. The determinant of the matrix determines the degree ofthe orthogonality between its row vectors. Therefore, a system that has an1166.2. System Design and MethodsS matrix with larger determinant (which means that the sensitivity charac-teristics of the sensors have higher degrees of partial orthogonality) is moredesirable.For the case of a sensing system, the purpose is to measure the change ofinterest in the cladding medium as a result of changes in concentrations orbiomolecular/chemical reactions. For each of the evanescent field sensors inthe system, the peak resonance wavelength shifts according to equation 6.4:∆λSi = (δλδT)Si ∆T + (δλδncl)Si ∆nbio + ∆λe (6.4)where ∆λSi is the measured peak resonant wavelength shift in the i-th sensor;( δλδT )Si and (δλδncl)Si are the sensitivity of the i-th sensor to the temperaturevariations and the refractive index variations of the cladding medium re-spectively; ∆nbio is the refractive index change of interest, which is due tobiomolecule reactions or concentration variations; and ∆λe is the error inresonant shift. Assuming that the system of sensors is mostly experienc-ing common/correlated noises, and the random noise is negligible (∆λe ≈ 0),having two sensors will solve for the two unknowns ∆T and ∆nbio in equation6.4.The change in refractive index of the cladding (∆ncl) is a function oftemperature, concentrations, and/or biological reactions:∆ncl = (δnclδTcl)∆T + ∆nbio (6.5)1176.2. System Design and Methodswhere ∆ncl is the change in refractive index of the cladding due to tempera-ture (∆T ) and biochemical changes (∆nbio) such as concentration variations,and δnclδTcl is the sensitivity of the refractive index of the cladding to the temper-ature variation of the cladding. Note that the wavelength shift contributedby the change in refractive index of analyte due to temperature is includedin the first term of equation 6.4.To demonstrate the capabilities of the proposed system of resonator sen-sors, we have designed and investigated a system of two resonator sensors.For this case, we can rewrite our matrix representation in equations 6.2 and6.3. The first varying property is the change in the refractive index of thecladding (V ar1 = ∆ncl), which is affected by temperature (V ar2 = ∆T ), andthe change of interest (V arx = ∆nbio). Therefore,∆λS1∆λS2=S1S2∆nbio∆T(6.6)whereS =S1S2=( δλδncl )S1 (δλδT )S1( δλδncl )S2 (δλδT )S2(6.7)1186.2. System Design and MethodsTherefore, ∆λS1∆λS2=( δλδncl )S1 (δλδT )S1( δλδncl )S2 (δλδT )S2∆nbio∆T(6.8)To achieve the purposes of noise corrections with this system design, itis required to obtain two sensors with significantly different sensitivity prop-erties (oppositely behaving); i.e. 1) a sensor with elevated sensitivity to thebiomolecules in the cladding medium, and lower sensitivities to other systemvariations such as temperature variations, 2) a sensor with lower sensitivitiesto the cladding and higher sensitivity to temperature variations for example.A differential measurement is achieved when the responses of these sensorsare significantly different. In other words, we are designing a sensitivity ma-trix (S) with a higher determinant. One way to build such a system is tohave different waveguide thicknesses, which will allow different relative shifts,and thus, eliminating spurious effects and improving measurements repeata-bility. To investigate the proposed system design, we used ring resonatorswith various waveguide thicknesses [35, 99].To demonstrate the capability of a two-resonator system, we have de-signed a cascade of two resonators with thicknesses of 220nm and 90nm (byusing methods explained in Sections 2.2 and 3.2). These were fabricated inThe Institute of Microelectronics in Singapore (IME). The resonator sensorwith thinner waveguide core (90 nm thick silicon) serves as the sensor withhigher sensitivity to cladding and lower sensitivity to temperature. In con-1196.3. Characterization and Performance of the Sensors and Systemtrast, the resonator with the regular thickness of 220nm serves as the sensorwith lower sensitivity to cladding and higher sensitivity to temperature (orreference sensor):(δλδncl)S90 > (δλδncl)S220 (6.9)(δλδT)S90 < (δλδT)S220 (6.10)where, ( δλδncl )S90 and (δλδncl)S220 are the sensitivities of the sensor to the vari-ations of the refractive index of the cladding, and ( δλδT )S90 and (δλδT )S220 arethe sensitivities of the sensor to the temperature variations, for the sensorwith silicon core thickness of 90 nm and 220 nm, respectively. According tosimulated sensitivities, presented in figure 3.3, these sensitivity values for ourdesigned system with two cascaded resonators are: ( δλδncl )S90 ≈ 120 nm/RIU,( δλδncl )S220 ≈ 56 nm/RIU, (δλδT )S90 ≈ 0.06 nm/K, and (δλδT )S220 ≈ 0.075 nm/K.Based on these values for our designed sensors and equation 6.4, a system oftwo equations and two unknowns can be established to predict the variationsof interest in the presence of temperature drifts.6.3 Characterization and Performance ofthe Sensors and SystemTo demonstrate and experimentally validate the capabilities of our system,various concentrations of analytes over a range of temperatures were flownover our system of cascaded resonator sensors (using methods explained in1206.3. Characterization and Performance of the Sensors and SystemSection 2.2). Specifically, four sets of data were collected. For each data set,various RI solutions were introduced to the cladding medium of our cascadedresonator sensors. Three sets were collected at controlled temperatures of24, 25, and 26 ◦C. The last set was collected at a random temperature (i.e.without temperature controlling the experimental setup). Maintenance ofslow flow is required in these experiments to avoid bubble generation and toallow the system to reach a steady-state temperature. If we assume negligible∆λe with controlled experiments, we can achieve temperature correctionsusing the response of two sensors.Figure 6.1 (a,b) shows the wavelength shift responses of each of the tworesonator sensors individually for the four sets of data explained above (threecontrolled and one random temperature). The results in figure 6.1 assumesa zero concentration at 25 ◦C as the reference point to demonstrate the rela-tive shifts. When the temperature is controlled, a well-defined linear relationbetween the concentration and wavelength shift is observed. The linear rela-tion describes the sensitivity of the sensor to the refractive index of cladding.Therefore, the shift of interest is the slope, and the shift due to temperaturevariations is the vertical distance (i.e. the offset between the lines for varioustemperatures).∆λtotal = ∆λwanted + ∆λunwanted (6.11)Assuming that our sensing system had only one of these resonator sen-1216.3. Characterization and Performance of the Sensors and System00.0020.0040.0060.0080.0100. Index Change (RIU)Wavelength Shift (nm)  24o C 25o C 26o C Rand(a) 220 nm thick silicon core00.0020.0040.0060.0080.0100.511.52Refractive Index Change (RIU)Wavelength Shift (nm)  24o C 25o C 26o C Rand(b) 90 nm thick silicon coreFigure 6.1: Response of each single resonator sensor, a) sensor with 220 nm thicksilicon core, b) sensor with 90 nm thick silicon core, to the four data sets. Eachsensor’s response to the three data sets with controlled temperature has a clearbest fit line. However, it is clear that in the absence of a repeatable and accu-rate temperature controlling mechanism, the predictions are not reliable and theaccuracy of predictions is significantly reduced.sors, it is clear that without a well repeatable temperature controlled experi-ment, predicting the refractive index of the cladding accurately is challenging.Thus, the error of prediction/estimation is high.Based on this set of calibration data, the ‘unwanted’ shift for both sensorsare approximately similar. Therefore, subtracting the resonance wavelengthsof the two sensors lproduces only the ‘wanted’ shift. Although not optimal,this differential measurement system improves the prediction results signifi-cantly. These promising results are due to significantly different sensitivitiesof the two resonator sensors to RI changes, where one is almost twice of theother.Figure 6.2 shows the results of our differential measurement system, which1226.3. Characterization and Performance of the Sensors and System00.0020.0040.0060.0080.0100. Index Change (RIU)Differential Shift (nm)  24o C 25o C 26o C Rand(a)00.0020.0040.0060.0080.0100. Index Change (RIU)Differential Shift (nm)(b)Figure 6.2: Response of the differential system of cascaded resonators to the fourdata sets. a) observing how each set fits. Dashed lines are the best fit lines foreach data set (to compare with figure 6.1). b) observing how all the four sets fit.Dashed line is the best fit line to the three data sets with controlled temperature.in this case is the difference between the resonance peak of the two resonatorsensors. It can be well observed that the response of the system to thechange in refractive index of cladding is the same at various temperatures.For initial demonstration, the differential calculation is performed on eachtemperature and the best-fit line was plotted for comparison to figure 6.1(a,b). In figure 6.2 (b), the linear line is the best fit to the data from thethree temperatures 24 ◦C, 25 ◦C, and 26 ◦C. The result of the random set isplotted as purple stars. Assuming that a linear model was built with theresults from the three temperatures and was used to estimate the randomset, we calculate the R2 value of prediction to be 0.996. If we only usedthe data at one temperature and used the sensors individually to build themodel, we would have achieved R2 values of 0.9-0.95 and 0.4-0.74 for 90 nm1236.3. Characterization and Performance of the Sensors and Systemand 220 nm thick resonator sensors respectively.To observe the effect of differential system on a set of data where thetemperature of the stage is not controlled, the response of the sensors tochanges of concentration over time is plotted in figure 6.3 (a). Figure 6.3 (b)translates the same data in figure 6.3 (a) into our usual sensitivity plot to ob-serve how the degree of the accuracy of concentration prediction/estimationimproves with the differential system.102030405000. StepWavelength Shift (nm)  220 nm 90 nm diff(a) step00.0020.0040.0060.0080.0100.511.52Refractive Index Change (RIU)Wavelength Shift (nm)  220 nm 90 nm diff(b) linear fitFigure 6.3: Response of the differential system (black) as well as each of theresonator sensors individually to a set of RI measurements where the temperatureis not controlled. a) is the response over time as the reagent (RI solution) changes.b) plots the shift as a function of change in refractive index of the cladding, andthe linear best fit line through the points to estimate the sensitivity and observethe accuracy of measurements.1246.4. Discussions and Conclusions6.4 Discussions and ConclusionsThe differential system would perform best if there are multiple sensors withsignificantly different sensitivities to the cladding material (change in refrac-tive index of the cladding) and the temperature of the waveguide. We assumethat the system reaches a thermally steady state, meaning that the tempera-ture of the silicon waveguides and the cladding are the same. Also, assumingthat the dominant factor in the resonant shifts are temperature and con-centration, we can solve equation 6.4 for our two sensors to obtain the twounknowns, ∆T and ∆nbio. Furthermore, the resonant wavelength shift dueto temperature variations can be extracted in order to obtain ∆T .One might think that having multiple identical sensors and averagingthem would have similar effect. However, this would only reduce the randomnoise but not the undesired correlated shifts and drifts due to environmentalvariations that all sensors experience similarly. Therefore, it is importantto have significantly different sensors that would respond to these commonvariations differently.Regarding the optimization of our proposed system of differential mea-surement, let us write the two equations for the two sensors: Sensor1 andSensor2. Sensor1 is assumed to have a higher sensitivity to the change inrefractive index of the cladding.∆λSensor1 = ST−1∆T + SC−1∆C (6.12)1256.4. Discussions and Conclusions∆λSensor2 = ST−2∆T + SC−2∆C (6.13)where ST−1 and ST−2 are the sensitivities of Sensor1 and Sensor2, respec-tively, to temperature variations; SC−1 and SC−2 are the sensitivities of Sen-sor1 and Sensor2, respectively, to concentrations; and ∆C is the concen-tration changes (change of interest). To remove the effect of temperaturevariations (the first term in the equations 6.12 and 6.13), we have to calcu-late a weighted difference of the responses of the two sensors:∆λSystem = (SC−1 −ST−1ST−2SC−2)∆C (6.14)Therefore, for improved higher sensitivity of our differential system, largerSC−1 and ST−2 as well as smaller ST−1 and SC−2 are preferred. Comparingthis with the matrix representations in equations 6.6-6.8, it can be realizedthat maximizing the determinant of the S matrix results in the optimumsystem performance.The comparison of various two-sensor systems can be illustrated with thesensors that are developed in this thesis. The sensitivity matrices and theirdeterminants can be derived from table 6.2.Table 6.1 summarizes the specification of some of the previously discussedsensors. Table 6.2 is a summary of the calculated determinant of the sen-sitivity matrix (S) for various possible pairs with the sensors in table 6.1.Generally, for this specific case of extracting concentration changes in thepresence of temperature variations, pairing the sensors with highest and low-1266.4. Discussions and Conclusionsest ratio of temperature sensitivity/bulk sensitivity would result in the bestperformance (compare tables 6.1 and 6.2).Table 6.1: Sensitivity vectors of a few sensorsName Pol. (Thick, Width) Sbulk ST Ratio(nm, nm) ∆λres∆ncl [nmRIU ]∆λres∆T [nmK ]STSbulk[RIUK ]TE-220 TE (220, 500) 56 0.08 1.4× 10−3TE-90 TE (90, 900) 120 0.07 5.8× 10−4TM-220 TM (220, 750) 179 0.04 2.2× 10−4TM-150 TM (150, 900) 274 0.008 2.9× 10−5Table 6.2: Determinant of the sensitivity matrix for various combinations of asystem of two resonator sensorsTE-220 TE-90 TM-220 TM-150TE-220 0 5.7 12.1 21.5TE-90 5.7 0 7.7 18.2TM-220 12.1 7.7 0 9.5TM-150 21.5 18.2 9.5 0In general, a better performance is achieved with a system of partiallyorthogonal sensors. This means that each sensor has significantly highersensitivity to one parameter and significantly lower sensitivities to the othervarying parameters in the system. For example, extending our system design1276.4. Discussions and Conclusionsto n dimensions, if n sources of undesirable drifts and variations in the systemare identified, designing n sensors, where each one is significantly more sen-sitive to one of these variations, is ideally desirable. With this n dimensionalproblem presented in equations 6.1-6.3, it can be observed that designinga diagonally-dominant matrix S, with maximized determinant value, is re-quired for the best performance of this proposed system of sensors.128Chapter 7Conclusions and Future WorkIn this thesis, we have investigated, developed, and demonstrated novel SOI-based optical sensors and sensing systems for the ultimate long-term goalof full sensing system on-chip advancement. The SOI-based sensor, beingthe core component of the sensing system, allows detection and quantifi-cation of target molecules. Therefore, improving the sensitivity, accuracy,repeatability, and reliability of the sensing system, within the constraints ofMPW foundries, are critical steps towards our long-term goal. Efficient SOI-based sensors and system of sensors that are fabricated with standard CMOSprocesses, create a potential for integration of sensors with electronics anddevelop a sensing system-on-chip.7.1 Summary and ConclusionIn summary, we have studied, theoretically and experimentally, SOI-basedresonator sensors and system. We have proposed and demonstrated methodsto improve the sensitivity of the sensors as single devices, investigated a novelmethod to integrate a detector with a sensor, and achieved a novel design of a1297.1. Summary and Conclusionsystem of sensors, mimicking multivariate techniques, to correct for unwantedvariations and to accomplish a more repeatable and accurate measurementin the presence of environmental variations.Contributions of this thesis are:• Proposed and investigated methods to improve the sensitivity of SOI-based sensors as a single device. The effect of waveguide core thicknesson the sensitivity of our resonators was investigated both theoreticallyand experimentally. It was found that, for a TE guided mode, thesensitivity of a resonator sensor increases as the waveguide thicknessdecreases (from our conventional 220 nm thick silicon core) to approx-imately 35 nm thick silicon core. Therefore, the smallest thicknessavailable through MPW foundry services, 90 nm thick silicon core, wasused to design ultra-thin resonator sensors.• Demonstrated analytically, by using fully-vectorial 2D eigenmode solverand analytical equations coded in MATLAB, the relationship betweensensitivity of the sensor and the dimensions of the waveguide core.• Designed and experimentally characterized resonator sensors with thethicknesses that were available through MPW foundries and were closeto our analytical findings. Sensors with sensitivities of around 270nm/RIU were accomplished.• Investigation and demonstration of a novel method to integrate a detec-tor with resonator sensors. Defect-mediated, ion-implanted, resonant-1307.2. Future Workenhanced photodetector sensors were designed and characterized.• Proposed and demonstrated a novel design of a system of sensors. Withthis design, each sensor in the system, having various sensitivities todifferent variables, creates an equation with multiple unknowns. Therelations and variability between these equations constitutes a systemof equations and unknowns. The sensitivity components of all thesensors in the system generates a sensitivity matrix (S). This system issuitable for training a multivariate model to predict the variations ofinterest more accurately in the presence of ‘unwanted’ variations. Theperformance of this multivariate system depends on the determinant ofS. The higher the determinant, the higher is the degree of orthogonality,and therefore the system is more efficient. The R2 value of predictionwas improved from 0.5 to Future WorkIn this thesis, we have focused on optimizing sensor designs within the con-straints of MPW foundries to prepare for future development of sensing sys-tem on-chip. To achieve this long-term goal, suggested improvements andfuture work include:• Having achieved promising results with the design of a system of twosensors, the design of the system of cascaded sensors can be expandedto more sensors.1317.2. Future Work• The proposed system of sensors, where each sensor is more sensitive toa specific variation, provides a set of suitable equations for developmentof a prediction model. Therefore, a prediction model can be developed,by using multivariate techniques, based on the system of equations thatare formed as a results of our system of cascaded sensors.• Investigation of different options for integrating photodetector with asystem of resonator sensors on the same chip.• A more thorough study of noises involved in the system, both frominstrumentation and environmental variations, will be required to de-termine the specification of other components of the sensing system onchip (such as on-chip laser and detector). The overall performance ofthe sensing system is limited by the combination of the performancesof all components.• Integration of the sensing systems with electronics on the same chip, asa step towards lab-on-chip (LoC) or sensing systems on chips (SSOCs)132Bibliography[1] R. M. Lequin, “Enzyme immunoassay (eia)/enzyme-linked immunosor-bent assay (elisa),” Clinical chemistry, vol. 51, no. 12, pp. 2415–2418,2005.[2] J. L. Arlett, E. B. Myers, and M. L. Roukes, “Comparative advantagesof mechanical biosensors,” Nat Nano, vol. 6, no. 4, pp. 203–215, 042011. [Online]. Available: http://dx.doi.org/10.1038/nnano.2011.44[3] V. Gubala, L. F. Harris, A. J. Ricco, M. X. Tan, and D. E. Williams,“Point of care diagnostics: status and future,” Analytical chemistry,vol. 84, no. 2, pp. 487–515, 2011.[4] C. D. Chin, V. Linder, and S. K. Sia, “Commercialization of microflu-idic point-of-care diagnostic devices,” Lab on a Chip, vol. 12, no. 12,pp. 2118–2134, 2012.[5] R. Etzioni, N. Urban, S. Ramsey, M. McIntosh, S. Schwartz, B. Reid,J. Radich, G. Anderson, and L. Hartwell, “The case for early detec-tion,” Nature Reviews Cancer, vol. 3, no. 4, pp. 243–252, 2003.[6] J. T. Kindt, M. S. Luchansky, A. J. Qavi, S.-H. Lee, and R. C. Bailey,“Subpicogram per milliliter detection of interleukins using silicon pho-tonic microring resonators and an enzymatic signal enhancement strat-egy,” Analytical Chemistry, vol. 85, no. 22, pp. 10 653–10 657, 2013,pMID: 24171505.[7] M. S. Luchansky and R. C. Bailey, “Silicon photonic microring res-onators for quantitative cytokine detection and T-cell secretion analy-sis,” Anal. Chem., vol. 82, no. 5, pp. 1975–1981, 2010.[8] A. L. Washburn, L. C. Gunn, and R. C. Bailey, “Label-free quanti-tation of a cancer biomarker in complex media using silicon photonic133Chapter 7. Bibliographymicroring resonators,” Anal. Chem., vol. 81, no. 22, pp. 9499–9506,2009.[9] M. Iqbal, M. A. Gleeson, B. Spaugh, F. Tybor, W. G. Gunn,M. Hochberg, T. Baehr-Jones, R. C. Bailey, and L. C. Gunn, “Label-free biosensor arrays based on silicon ring resonators and high-speed op-tical scanning instrumentation,” IEEE J. Quantum Electron., vol. 16,no. 3, pp. 654–661, 2010.[10] A. J. Qavi and R. C. Bailey, “Multiplexed detection and label-freequantitation of MicroRNAs using arrays of silicon photonic microringresonators,” Angewandte Chemie, vol. 122, no. 27, pp. 4712–4715, 2010.[11] A. L. Washburn, M. S. Luchansky, A. L. Bowman, and R. C. Bai-ley, “Quantitative, label-free detection of five protein biomarkers usingmultiplexed arrays of silicon photonic microring resonators,” Analyticalchemistry, vol. 82, no. 1, pp. 69–72, 2009.[12] J. K. Doylend and A. P. Knights, “The evolution of silicon photonics asan enabling technology for optical interconnection,” Laser & Photon.Rev., vol. 6, no. 4, pp. 504–525, 2012.[13] W. Shi, H. Yun, C. Lin, J. Flueckiger, N. A. F. Jaeger, and L. Chros-towski, “Coupler-apodized bragg-grating add-drop filter,” Opt. Lett.,vol. 38, pp. 3068–3070, 08/2013 2013.[14] W. Shi, X. Wang, W. Zhang, H. Yun, N. A. F. Jaeger, and L. Chros-towski, “Integrated microring add-drop filters with contradirectionalcouplers,” CLEO (Conference on Lasers and Electro-Optics) 2012, p.JW4A.91, 05/2012 2012.[15] R. Soref, “The Past, Present, and Future of Silicon Photonics,” SelectedTopics in Quantum Electronics, IEEE Journal of, vol. 12, no. 6, pp.1678–1687, Nov. 2006.[16] N. A. Yebo, D. Taillaert, J. Roels, D. Lahem, M. Debliquy, P. Lom-mens, Z. Hens, and R. Baets, “Integrated optical gas sensors on silicon-on-insulator platform,” in Proceedings of the 2009 Annual Symposiumof the IEEE Photonics Benelux Chapter, S. Beri, P. Tassin, G. Craggs,X. Leijtens, and J. Danckaert, Eds. VUB Press, 2009, pp. 153–156.134Chapter 7. Bibliography[17] L. Chrostowski, S. Grist, J. Flueckiger, W. Shi, X. Wang, E. Ouellet,H. Yun, M. Webb, B. Nie, Z. Liang, K. C. Cheung, S. A. Schmidt,D. M. Ratner, and N. A. F. Jaeger, “Silicon photonic resonator sensorsand devices,” Proc. SPIE, vol. 8236, no. 1, pp. 823 620–823 620–16, Feb.2012.[18] O. Scheler, J. T. Kindt, A. J. Qavi, L. Kaplinski, B. Glynn, T. Barry,A. Kurg, and R. C. Bailey, “Label-free, multiplexed detection of bac-terial tmRNA using silicon photonic microring resonators,” Biosens.Bioelectron., vol. 36, no. 1, pp. 56–61, 2012.[19] W. W. Shia and R. C. Bailey, “Single domain antibodies for the detec-tion of ricin using silicon photonic microring resonator arrays,” Anal.Chem., vol. 85, no. 2, pp. 805–810, 2012.[20] M. R. Tomita, L. S. Russ, R. Sridhar, and B. J. Naughton, “Smarthome with healthcare technologies for community-dwelling olderadults,” Available as on 13th of March, 2012.[21] X. Fan, I. White, S. Shopova, H. Zhu, J. Suter, and Y. Sun, “Sensitiveoptical biosensors for unlabeled targets: A review,” Analytica ChimicaActa, vol. 620, no. 1-2, pp. 8–26, Jul. 2008.[22] F. Vollmer and L. Yang, “Review Label-free detection with high-Q mi-crocavities: a review of biosensing mechanisms for integrated devices,”Nanophotonics, vol. 1, no. 3-4, Jan. 2012.[23] Https://www.biacore.com/lifesciences/company/index.html, “Gehealthcare,” 2013-01-23. [Online]. Available: Https://www.biacore.com/lifesciences/company/index.html[24] V. M. Passaro, F. Dell’olio, B. Casamassima, and F. D. Leonardis,“Guided-Wave Optical Biosensors,” Sensors, vol. 7, pp. 508–539, 2007.[25] C. F. Donzella V, “Optical biosensors to analyze novel biomarkers inoncology,” J Biophotonics, vol. 4, no. 6, pp. 442–452, 2011.[26] F. Vollmer, S. Arnold, D. Braun, I. Teraoka, and A. Libchaber, “Mul-tiplexed DNA quantification by spectroscopic shift of two microspherecavities.” Biophysical journal, vol. 85, no. 3, pp. 1974–1979, Sep. 2003.135Chapter 7. Bibliography[27] C. L. Arce, K. De Vos, T. Claes, K. Komorowska, D. Van Thourhout,and P. Bienstman, “Silicon-on-insulator microring resonator sensor in-tegrated on an optical fiber facet,” IEEE Photon. Tech. Lett., vol. 23,no. 13, pp. 890–892, 2011.[28] S. M. Grist, S. A. Schmidt, J. Flueckiger, V. Donzella, W. Shi,S. Talebi Fard, J. T. Kirk, D. M. Ratner, K. C. Cheung, and L. Chros-towski, “Silicon photonic micro-disk resonators for label-free biosens-ing,” Opt. Express, vol. 21, pp. 7994–8006, 03/2013 2013.[29] S. Mandal and D. Erickson, “Nanoscale optofluidic sensor arrays,” Opt.Express, vol. 16, no. 3, p. 1623, 2008.[30] X. Wang, J. Flueckiger, S. Schmidt, S. Grist, S. Talebi Fard, J. Kirk,M. Doerfler, K. C. Cheung, D. M. Ratner, and L. Chrostowski, “Asilicon photonic biosensor using phase-shifted bragg gratings in slotwaveguide,” J. Biophotonics, 04/2013 2013.[31] P. Dumon, W. Bogaerts, V. Wiaux, J. Wouters, S. Beckx, J. Van Camp-enhout, D. Taillaert, B. Luyssaert, P. Bienstman, D. Van Thourhout,and R. Baets, “Low-loss SOI photonic wires and ring resonators fabri-cated with deep UV lithography,” IEEE Photon. Tech. Lett., vol. 16,no. 5, pp. 1328–1330, May 2004.[32] H. Mukundan, A. S. Anderson, W. K. Grace, K. M. Grace, N. Hartman,J. S. Martinez, and B. I. Swanson, “Waveguide-based biosensors forpathogen detection,” Sensors, vol. 9, no. 7, pp. 5783–5809, 2009.[33] F. Dell’Olio and V. M. Passaro, “Optical sensing by optimized siliconslot waveguides,” Opt. Express, vol. 15, no. 8, pp. 4977–4993, Apr.2007.[34] S. Talebi Fard, S. M. Grist, V. Donzella, S. A. Schmidt, J. Flueckiger,X. Wang, W. Shi, A. Millspaugh, M. Webb, D. M. Ratner, K. C.Cheung, and L. Chrostowski, “Label-free silicon photonic biosensorsfor use in clinical diagnostics,” Proc. SPIE, Silicon Photonics VIII,vol. 8629, pp. 862 909–14, 02/2013 2013.[35] S. Talebi Fard, V. Donzella, S. A. Schmidt, J. Flueckiger, S. M. Grist,P. TalebiFard, Y. Wu, R. J. Bojko, E. Kwok, N. A. Jaeger, D. M. Rat-136Chapter 7. Bibliographyner, and L. Chrostowski, “Performance of ultra-thin SOI-based res-onators for sensing applications,” Opt. Express, vol. 22, no. 12, pp.14 166–14 179, 2014.[36] K. Zinoviev, L. G. Carrascosa, J. Sa´nchez del Rı´o, B. Sepu´lveda,C. Domı´nguez, and L. M. Lechuga, “Silicon photonic biosensors for lab-on-a-chip applications,” Advances in Optical Technologies, vol. 2008,2008.[37] J. Arlett, E. Myers, and M. Roukes, “Comparative advantages of me-chanical biosensors,” Nature nanotechnology, vol. 6, no. 4, pp. 203–215,2011.[38] C. W. Heath, “Pesticides and cancer risk,” Cancer, vol. 80, no. 10, pp.1887–1888, 1997.[39] S. Laschi, D. Ogon´czyk, I. Palchetti, and M. Mascini, “Evaluation ofpesticide-induced acetylcholinesterase inhibition by means of dispos-able carbon-modified electrochemical biosensors,” Enzyme and Micro-bial Technology, vol. 40, no. 3, pp. 485–489, 2007.[40] http://www.epixfab.eu., “epixfab: The silicon photonics platform,”2013-01-22. [Online]. Available: http://www.epixfab.eu.[41] I. M. White and X. Fan, “On the performance quantification of reso-nant refractive index sensors,” Opt. Express, vol. 16, no. 2, pp. 1020–1028, 2008.[42] T. Yoshie, L. Tang, and S.-Y. Su, “Optical microcavity: Sensing downto single molecules and atoms,” Sensors, vol. 11, no. 2, pp. 1972–1991,2011.[43] J. Ackert, J. Doylend, D. Logan, P. Jessop, R. Vafaei, L. Chros-towski, and A. Knights, “Defect-mediated resonance shift of silicon-on-insulator racetrack resonators,” Opt. Express, vol. 19, no. 13, pp.11 969–11 976, 2011.[44] L. Chrostowski and M. Hochberg, Silicon Photonics Design, ser. Cam-bridge University Press, 2015.137Chapter 7. Bibliography[45] Y. Wang, X. Wang, J. Flueckiger, H. Yun, W. Shi, R. Bojko, N. A. F.Jaeger, and L. Chrostowski, “Focusing sub-wavelength grating cou-plers with low back reflections for rapid prototyping of silicon photoniccircuits,” Opt. Express, vol. 22, no. 17, pp. 20 652–20 662, Aug 2014.[46] V. V. Tuchin, I. L. Maksimova, D. A. Zimnyakov, I. L. Kon, A. H.Mavlyutov, and A. A. Mishin, “Light propagation in tissues with con-trolled optical properties,” J. Biomed. Opt., vol. 2, no. 4, pp. 401–417,1997.[47] R. Sokoloff, K. Norton, C. Gasior, K. Marker, and L. Grauer, “Adual-monoclonal sandwich assay for prostate-specific membrane anti-gen: levels in tissues, seminal fluid and urine.” Prostate, vol. 43, no. 2,pp. 150–7, 2000.[48] U. B. Nielsen and B. H. Geierstanger, “Multiplexed sandwich assays inmicroarray format,” Journal of immunological methods, vol. 290, pp.107–120, 2004.[49] X. Wang, W. Shi, S. Grist, H. Yun, and N. A. F. J. Chrostowski,“Narrow-Band Transmission Filter using Phase-Shifted Bragg Gratingsin SOI Waveguide,” 10/2011 2011, p. ThZ1.[50] V. R. Almeida, Q. Xu, C. A. Barrios, and M. Lipson, “Guiding andconfining light in void nanostructure,” Opt. Lett., vol. 29, no. 11, pp.1209–1211, 2004.[51] E. L. Florin, V. T. Moy, and H. E. Gaub, “Adhesion forces betweenindividual ligand-receptor pairs,” Science (New York, NY), vol. 264, p.415, 1994.[52] S. Mandal, J. M. Goddard, and D. Erickson, “A multiplexed optofluidicbiomolecular sensor for low mass detection,” Lab on a Chip, vol. 9, pp.2924–2932, 2009.[53] K. D. Vos, I. Bartolozzi, E. Schacht, P. Bienstman, and R. Baets,“Silicon-on-insulator microring resonator for silicon-on-insulator mi-croring resonator for sensitive and label-free biosensing,” Opt. Express,vol. 15, no. 12, pp. 7610–7615, 2007.138Chapter 7. Bibliography[54] D.-X. Xu, M. Vachon, A. Densmore, R. Ma, S. Janz, A. Delaˆge, J. La-pointe, P. Cheben, J. Schmid, E. Post et al., “Real-time cancellation oftemperature induced resonance shifts in SOI wire waveguide ring res-onator label-free biosensor arrays,” Opt. Express, vol. 18, no. 22, pp.22 867–22 879, 2010.[55] T. Claes, J. G. Molera, K. De Vos, E. Schacht, R. Baets, and P. Bienst-man, “Label-free biosensing with a slot-waveguide-based ring resonatorin silicon on insulator,” IEEE J. Photonics, vol. 1, no. 3, pp. 197–204,2009.[56] A. Densmore, D.-X. Xu, P. Waldron, S. Janz, P. Cheben, J. Lapointe,A. Delaˆge, B. Lamontagne, J. Schmid, and E. Post, “A silicon-on-insulator photonic wire based evanescent field sensor,” IEEE Photon.Tech. Lett., vol. 18, no. 23, pp. 2520–2522, 2006.[57] G. Veldhuis, O. Parriaux, H. Hoekstra, and P. Lambeck, “Sensitivityenhancement in evanescent optical waveguide sensors,” J. LightwaveTech., vol. 18, no. 5, p. 677, 2000.[58] M. Gould, A. Pomerene, C. Hill, S. Ocheltree, Y. Zhang, T. Baehr-Jones, and M. Hochberg, “Ultra-thin silicon-on-insulator strip waveg-uides and mode couplers,” Appl. Phys. Lett., vol. 101, no. 22, pp.221 106–221 106, 2012.[59] H. Li, “Refractive index of silicon and germanium and its wavelengthand temperature derivatives,” J. Phys. and Chem. Ref. data, vol. 9,no. 3, pp. 561–658, 1980.[60] B. J. Frey, D. B. Leviton, and T. J. Madison, “Temperature-dependentrefractive index of silicon and germanium,” in Astronomical Telescopesand Instrumentation. International Society for Optics and Photonics,2006, pp. 62 732J–62 732J.[61] A. N. Bashkatov and E. A. Genina, “Water refractive index in de-pendence on temperature and wavelength: a simple approximation,”in Saratov Fall Meeting 2002: Optical Technologies in Biophysics andMedicine IV. International Society for Optics and Photonics, 2003,pp. 393–395.139Chapter 7. Bibliography[62] G. Abbate, U. Bernini, E. Ragozzino, and F. Somma, “The tempera-ture dependence of the refractive index of water,” J. Phys. D: Appl.Phys., vol. 11, no. 8, p. 1167, 1978.[63] W. Bogaerts, V. Wiaux, D. Taillaert, S. Beckx, B. Luyssaert, P. Bien-stman, and R. Baets, “Fabrication of photonic crystals in silicon-on-insulator using 248-nm deep uv lithography,” IEEE journal of selectedtopics in quantum electronics, vol. 8, no. 4, pp. 928–934, 2002.[64] M. Hochberg, “Integrated ultrafast nonlinear optical devices in silicon,”Ph.D. dissertation, California Institute of Technology, 2006.[65] T. Baehr-Jones, M. Hochberg, C. Walker, and A. Scherer, “High Qring resonators in thin silicon-on-insulator,” Appl. Phys. lett., vol. 85,no. 16, pp. 3346–3347, 2004.[66] F. Grillot, L. Vivien, S. Laval, and E. Cassan, “Propagation loss insingle-mode ultrasmall square silicon-on-insulator optical waveguides,”J. Lightwave Tech., vol. 24, no. 2, p. 891, 2006.[67] J. McCaulley, V. Donnelly, M. Vernon, and I. Taha, “Temperaturedependence of the near-infrared refractive index of silicon, gallium ar-senide, and indium phosphide,” Phys. Rev. B, vol. 49, no. 11, p. 7408,1994.[68] Y. Wang, J. Flueckiger, C. Lin, and L. Chrostowski, “Universal gratingcoupler design,” Proc. SPIE, vol. 8915, pp. 89 150Y–89 150Y–7, 2013.[69] X. Wang, S. Grist, J. Flueckiger, N. A. Jaeger, and L. Chrostowski,“Silicon photonic slot waveguide bragg gratings and resonators,” Opt.Express, vol. 21, no. 16, pp. 19 029–19 039, 2013.[70] F. Payne and J. Lacey, “A theoretical analysis of scattering loss fromplanar optical waveguides,” Opt. and Quantum Electron., vol. 26,no. 10, pp. 977–986, 1994.[71] P. Dong, W. Qian, S. Liao, H. Liang, C.-C. Kung, N.-N. Feng, R. Shafi-iha, J. Fong, D. Feng, A. V. Krishnamoorthy et al., “Low loss shallow-ridge silicon waveguides,” Opt. Express, vol. 18, no. 14, pp. 14 474–14 479, 2010.140Chapter 7. Bibliography[72] F. Grillot, L. Vivien, S. Laval, D. Pascal, and E. Cassan, “Size influenceon the propagation loss induced by sidewall roughness in ultrasmall soiwaveguides,” Photonics Technology Letters, IEEE, vol. 16, no. 7, pp.1661–1663, 2004.[73] J. Vo¨ro¨s, “The density and refractive index of adsorbing protein lay-ers,” Biophysical journal, vol. 87, no. 1, pp. 553–561, 2004.[74] A. Yariv and P. Yeh, Photonics: Optical Electronics in Modern Com-munications (The Oxford Series in Electrical and Computer Engineer-ing). Oxford University Press, Inc., 2006.[75] “IME,” 2015-01. [Online]. Available: http://www.a-star.edu.sg/ime/[76] S. Schmidt, J. Flueckiger, W. Wu, S. M. Grist, S. Talebi Fard,V. Donzella, P. Khumwan, E. R. Thompson, Q. Wang, P. Kulik,X. Wang, A. Sherwali, J. Kirk, K. C. Cheung, L. Chrostowski, andD. Ratner, “Improving the performance of silicon photonic rings,disks, and bragg gratings for use in label-free biosensing,” Proc. SPIE,Biosensing and Nanomedicine VII, vol. 9166, p. 91660M, 08/2014 2014.[77] L. Chrostowski, X. Wang, J. Flueckiger, Y. Wu, Y. Wang, andS. Talebi Fard, “Impact of fabrication non-uniformity on chip-scalesilicon photonic integrated circuits,” in Optical Fiber CommunicationConference. Optical Society of America, 2014, pp. Th2A–37.[78] I. Teraoka and S. Arnold, “Theory of resonance shifts in TE and TMwhispering gallery modes by nonradial perturbations for sensing appli-cations,” JOSA B, vol. 23, no. 7, pp. 1381–1389, 2006.[79] G. Mot¸a, I. Moraru, J. Sjo¨quist, and V. Ghet¸ie, “Protein a as a molec-ular probe for the detection of antigen induced conformational changein Fc region of rabbit antibody,” Molecular immunology, vol. 18, no. 5,pp. 373–378, 1981.[80] W. L. DeLano, M. H. Ultsch, J. A. Wells et al., “Convergent solutionsto binding at a protein-protein interface,” Science, vol. 287, no. 5456,pp. 1279–1283, 2000.141Chapter 7. Bibliography[81] P. C. Weber, D. Ohlendorf, J. Wendoloski, and F. Salemme, “Struc-tural origins of high-affinity biotin binding to streptavidin,” Science,vol. 243, no. 4887, pp. 85–88, 1989.[82] M. Graille, E. A. Stura, A. L. Corper, B. J. Sutton, M. J. Taussig, J.-B.Charbonnier, and G. J. Silverman, “Crystal structure of a staphylo-coccus aureus protein A domain complexed with the fab fragment of ahuman IgM antibody: structural basis for recognition of b-cell recep-tors and superantigen activity,” Proceedings of the National Academyof Sciences, vol. 97, no. 10, pp. 5399–5404, 2000.[83] H. P. Erickson, “Size and shape of protein molecules at the nanome-ter level determined by sedimentation, gel filtration, and electron mi-croscopy,” Biological procedures online, vol. 11, no. 1, pp. 32–51, 2009.[84] M. C. Coen, R. Lehmann, P. Gro¨ning, M. Bielmann, C. Galli, andL. Schlapbach, “Adsorption and bioactivity of protein A on siliconsurfaces studied by AFM and XPS,” Journal of colloid and interfacescience, vol. 233, no. 2, pp. 180–189, 2001.[85] S. Talebi Fard, K. Murray, M. Caverley, V. Donzella, J. Flueck-iger, S. M. Grist, E. Huante-Ceron, S. A. Schmidt, E. Kwok, N. A.Jaeger et al., “Silicon-on-insulator sensors using integrated resonance-enhanced defect-mediated photodetectors,” Opt. express, vol. 22,no. 23, pp. 28 517–28 529, 2014.[86] A. Gassenq, N. Hattasan, L. Cerutti, J. B. Rodriguez, E. Tournie´,and G. Roelkens, “Study of evanescently-coupled and grating-assistedGaInAsSb photodiodes integrated on a silicon photonic chip,” Opt.Express, vol. 20, no. 11, pp. 11 665–11 672, 2012.[87] H. Park, A. W. Fang, R. Jones, O. Cohen, O. Raday, M. N. Sysak, M. J.Paniccia, and J. E. Bowers, “A hybrid AlGaInAs-silicon evanescentwaveguide photodetector,” Opt. Express, vol. 15, no. 10, p. 60446052,2007.[88] J. Brouckaert, G. Roelkens, D. Van Thourhout, and R. Baets, “Com-pact InAlAs-InGaAs metal-semiconductor-metal photodetectors inte-grated on silicon-on-insulator waveguides,” IEEE Photon. Tech. Lett.,vol. 19, no. 19, pp. 1484–1486, Oct. 2007.142Chapter 7. Bibliography[89] J. Michel, J. Liu, and L. C. Kimerling, “High-performance Ge-on-Siphotodetectors,” Nature Photonics, vol. 4, no. 8, pp. 527–534, 2010.[90] L. Vivien, J. Osmond, J.-M. Fdli, D. Marris-Morini, P. Crozat, J.-F.Damlencourt, E. Cassan, Y. Lecunff, and S. Laval, “42 GHz PIN ger-manium photodetector integrated in a silicon-on-insulator waveguide,”Opt. Express, vol. 17, no. 8, pp. 6252–6257, 2009.[91] L. Virot, L. Vivien, A. Polzer, D. Marris-Morini, J. Osmond, J. M.Hartmann, P. Crozat, E. Cassan, C. Baudot, C. Kopp, F. Boeuf,H. Zimmermann, and J. M. Fdli, “40 Gbit/s germanium waveguidephotodetector on silicon,” L. Vivien, S. K. Honkanen, L. Pavesi, andS. Pelli, Eds., Jun. 2012, pp. 84 310A–84 310A–7.[92] C. T. DeRose, D. C. Trotter, W. A. Zortman, A. L. Starbuck,M. Fisher, M. R. Watts, and P. S. Davids, “Ultra compact 45 GHzCMOS compatible germanium waveguide photodiode with low darkcurrent,” Opt. Express, vol. 19, no. 25, pp. 24 897–24 904, 2011.[93] L. Vivien, A. Polzer, D. Marris-Morini, J. Osmond, J. M. Hartmann,P. Crozat, E. Cassan, C. Kopp, H. Zimmermann, and J. M. Fdli, “Zero-bias 40Gbit/s germanium waveguide photodetector on silicon,” Opt.Express, vol. 20, no. 2, pp. 1096–1101, 2012.[94] J. Ackert, A. Knights, M. Fiorentino, R. Beausoleil, and P. Jessop, “De-fect enhanced silicon-on-insulator microdisk photodetector,” in OpticalInterconnects Conference, 2012 IEEE. IEEE, 2012, pp. 76–77.[95] J. J. Ackert, M. Fiorentino, D. F. Logan, R. G. Beausoleil, P. E.Jessop, and A. P. Knights, “Silicon-on-insulator microring resonatordefect-based photodetector with 3.5-GHz bandwidth,” J. Nanophoton-ics, vol. 5, no. 1, pp. 059 507–059 507, 2011.[96] D. F. Logan, P. Velha, M. Sorel, R. M. De La Rue, P. E. Jessop,and A. P. Knights, “Monitoring and tuning micro-ring properties usingdefect-enhanced silicon photodiodes at 1550 nm,” IEEE Photon. Tech.Lett., vol. 24, no. 4, pp. 261–263, 2012.[97] D. F. Logan, P. Velha, M. Sorel, R. De La Rue, A. P. Knights, andP. E. Jessop, “Defect-enhanced silicon-on-insulator waveguide resonant143Chapter 7. Bibliographyphotodetector with high sensitivity at 1.55 µm,” IEEE Photon. Tech.Lett., vol. 22, no. 20, pp. 1530–1532, 2010.[98] D. F. Logan, P. E. Jessop, and A. P. Knights, “Modeling defect en-hanced detection at 1550 nm in integrated silicon waveguide photode-tectors,” J. Lightwave Tech., vol. 27, no. 7, pp. 930–937, 2009.[99] S. Talebi Fard, V. Donzella, S. Schmidt, D. Ratner, R. Bojko, andL. Chrostowski, “Sensitivity analysis of thin waveguide SOI ring res-onators for sensing applications,” in Photonics Conference (IPC), 2013IEEE. IEEE, 2013, pp. 616–617.[100] R. R. Grote, K. Padmaraju, B. Souhan, J. B. Driscoll, K. Bergman, andR. Osgood, “10 Gb/s error-free operation of all-silicon ion-implanted-waveguide photodiodes at 1.55 µm,” IEEE Photon. Tech. Lett., vol. 25,no. 1, pp. 67–70, 2013.[101] J. Doylend, P. Jessop, and A. Knights, “Silicon photonic resonator-enhanced defect-mediated photodiode for sub-bandgap detection,”Opt. Express, vol. 18, no. 14, pp. 14 671–14 678, 2010.[102] M. S. Luchansky, A. L. Washburn, T. A. Martin, M. Iqbal, L. C. Gunn,and R. C. Bailey, “Characterization of the evanescent field profile andbound mass sensitivity of a label-free silicon photonic microring res-onator biosensing platform,” Biosens. Bioelectron., vol. 26, no. 4, pp.1283–1291, 2010.[103] A. Densmore, M. Vachon, D.-X. Xu, S. Janz, R. Ma, Y.-H. Li, G. Lop-inski, A. Delaˆge, J. Lapointe, C. Luebbert, Q. Y. Liu, P. Cheben, andJ. H. Schmid, “Silicon photonic wire biosensor array for multiplexed,real-time and label-free molecular detection,” Opt. Lett., vol. 34, no. 23,pp. 3598–3600, 2009.[104] P. J. Foster, J. K. Doylend, P. Mascher, A. P. Knights, and P. G. Cole-man, “Optical attenuation in defect-engineered silicon rib waveguides,”J. Appl. Phys., vol. 99, no. 7, pp. –, 2006.[105] D. Logan, K. Murray, J. Ackert, P. Velha, M. Sorel, R. M. De La Rue,P. Jessop, and A. Knights, “Analysis of resonance enhancement indefect-mediated silicon micro-ring photodiodes operating at 1550 nm,”J. Optics, vol. 13, no. 12, p. 125503, 2011.144Chapter 7. Bibliography[106] M. Geis, S. Spector, M. Grein, J. Yoon, D. Lennon, and T. Lyszczarz,“Silicon waveguide infrared photodiodes with >35 GHz bandwidth andphototransistors with 50 AW−1 response,” Opt. Express, vol. 17, no. 7,pp. 5193–5204, 2009.[107] K. De Vos, “Label-free silicon photonics biosensor platform withmi-croring resonators.” Ph.D. dissertation, Ghent University, 2010.[108] M. A´. Guille´n-Torres, E. Cretu, N. A. Jaeger, and L. Chrostowski,“Ring resonator optical gyroscopesparameter optimization and robust-ness analysis,” Journal of Lightwave Technology, vol. 30, no. 12, pp.1802–1817, 2012.[109] F. Ghasemi, M. Chamanzar, A. A. Eftekhar, and A. Adibi, “Anefficient technique for the reduction of wavelength noise in resonance-based integrated photonic sensors,” Analyst, vol. 139, pp. 5901–5910,2014. [Online]. Available: http://dx.doi.org/10.1039/C4AN01292E[110] Agilent 81682A Tunable Laser Modules User’s Guide, Sixth Edition.[111] J. Schmid, M. Ibrahim, P. Cheben, J. Lapointe, S. Janz, P. Bock,A. Densmore, B. Lamontagne, R. Ma, W. Ye et al., “Temperature-independent silicon subwavelength grating waveguides,” Optics letters,vol. 36, no. 11, pp. 2110–2112, 2011.[112] J. Teng, P. Dumon, W. Bogaerts, H. Zhang, X. Jian, X. Han, M. Zhao,G. Morthier, and R. Baets, “Athermal silicon-on-insulator ring res-onators by overlaying a polymer cladding on narrowed waveguides,”Optics express, vol. 17, no. 17, pp. 14 627–14 633, 2009.[113] V. Raghunathan, W. N. Ye, J. Hu, T. Izuhara, J. Michel, and L. Kimer-ling, “Athermal operation of silicon waveguides: spectral, second orderand footprint dependencies,” Optics express, vol. 18, no. 17, pp. 17 631–17 639, 2010.[114] S. Talebi Fard, L. Chrostowski, E. Kwok, and M.-C. Amann, “Chemo-metric approach for improving VCSEL-based glucose predictions,”IEEE Trans. Biomed. Eng., vol. 57, no. 3, pp. 578–585, 2010.[115] S. Talebi Fard, W. Hofmann, P. Talebi Fard, E. Kwok, M.-C. Amann,and L. Chrostowski, “Optical glucose monitoring using vertical cavity145Chapter 7. Bibliographysurface emitting lasers (vcsels),” in Proc. SPIE, vol. 7397, 2009, p.739704.[116] K. B. Gylfason, C. F. Carlborg, A. Kazmierczak, F. Dortu, L. Vivien,C. A. Barrios, W. van der Wijngaart, G. Stemme et al., “On-chip tem-perature compensation in an integrated slot-waveguide ring resonatorrefractive index sensor array,” Opt. Express, vol. 18, no. 4, pp. 3226–3237, 2010.146Appendix AList of Publications1. Sahba Talebi Fard, et al., “Optimized Sensitivity of Silicon-on-Insulator Strip Waveguide Resonator Sensor”, In review (2014).2. Sahba Talebi Fard, V. Donzella, S. Schmidt, J. Flueckiger, S. Grist,P. TalebiFard, Y. Wu, R. Bojko, E. Kwok, N. Jaeger, D. Ratner, andL. Chrostowski, ”Performance of ultra-thin SOI-based resonators forsensing applications.” Optics Express 22, no. 12 (2014): 14166-14179.3. Sahba Talebi Fard, K. Murray, M. Caverley, V. Donzella, J. Flueck-iger, S. Grist, E. Huante-Ceron, S. Schmidt, E. Kwok, N. Jaeger, A.Knights, and L. Chrostowski, ”Silicon-on-insulator sensors using in-tegrated resonance-enhanced defect-mediated photodetectors.” OpticsExpress 22, no. 23 (2014): 28517-28529.4. Sahba Talebi Fard, S. Grist, V. Donzella, S. Schmidt, J. Flueckiger,X. Wang, W. Shi, A. Millspaugh, M. Webb, D. Ratner, K. Cheung, L.Chrostowski, “Label-free silicon photonic biosensors for use in clinicaldiagnostics”, Proc. SPIE, Silicon Photonics VIII, 8629:86290914 (In-vited), SPIE OPTO, International Society for Optics and Photonics,2/02/2013.5. S. Schmidt, J. Flueckiger, W. Wu, S. M. Grist, Sahba Talebi Fard, V.Donzella, P. Khumwan, E. R. Thompson, Q. Wang, P. Kulik, X. Wang,A. Sherwali, J. Kirk, K. C. Cheung, L. Chrostowski, and D. Ratner,“Improving the performance of silicon photonic rings, disks, and bragggratings for use in label-free biosensing”, Proc. SPIE, Biosensing andNanomedicine VII 9166, 91660M (2014).6. Sahba Talebi Fard, E. Kwok, and L. Chrostowski, “Optical Glu-cose Monitoring Sensors”, Canadian Medical and Biological Engineer-ing Conference (2014).147Appendix A. List of Publications7. V. Donzella, A. Sherwali, J. Flueckiger, S. Grist, Sahba Talebi Fard,and L. Chrostowski, “Design and fabrication of SOI micro-ring res-onators based on sub-wavelength grating waveguides”, Optics Express(2014).8. C. Fu, Sahba Talebi Fard, K. Lee, S. Hong, L. Lee “NanoplasmonicOptoporation for Large-scale Precision Gene Regulation in Stem Cells”,in review (2014).9. L. Chrostowski, X. Wang, J. Flueckiger, Y. Wu, Y. Wang, SahbaTalebi Fard, “Impact of fabrication non-uniformity on chip-scale sili-con photonic integrated circuits”, Optical Fiber Communication Con-ference, Th2A. 37. 2014/3/9.10. V. Donzella, A. Sherwali, J. Flueckiger, Sahba Talebi Fard, S. Grist,and L. Chrostowski, “Sub-wavelength grating components for inte-grated optics applications on SOI chips”, Optics Express, 22, no. 17(2014/8/25): 21037-21050.11. S. Moghaddamjoo, A. Tashakor, Sahba Talebi Fard, Y. Tashakkor,E. Kwok, Y. Li, N. Tavassoli, G. Tibbits, E. Grant, and A. Rawicz,“Characterization of Cardiac Troponin I Raman Signature in BovineSerum Albumin and Human Blood Serum for the Potential Diagnosisof Myocardial Infarction”, in review12. X. Wang, J. Flueckiger, S. Schmidt, S. Grist, Sahba Talebi Fard,J. Kirk, M. Doerfler, K. Cheung, D. Ratner, and L. Chrostowski, “Asilicon photonic biosensor using phase-shifted bragg gratings in slotwaveguide”, Journal of Biophotonics, 04/2013 2013.13. S. Grist, S. Schmidt, J. Flueckiger, V. Donzella, W. Shi, Sahba TalebiFard, J. Kirk, D. Ratner, K. Cheung, and L. Chrostowski, “Silicon pho-tonic micro-disk resonators for label-free biosensing”, Optics Express,21:79948006, 03/2013 2013.14. V. Donzella, Sahba Talebifard, L. Chrostowski, “Study of waveg-uide crosstalk in silicon photonics integrated circuits”, Photonics North(2013)15. Sahba Talebifard, V. Donzella, S. A. Schmidt, D. M. Ratner, R. J.Bojko, L. Chrostowski, “Sensitivity analysis of thin waveguide SOI ring148Appendix A. List of Publicationsresonators for sensing applications.” International Photonics Confer-ence (IPC), 2013 IEEE, pages 616-617.16. V. Donzella, Sahba Talebifard, L. Chrostowski, “Modeling of asym-metric slot racetracks for improved bio-sensors performance”,NumericalSimulation of Optoelectronic Devices (NUSOD), 201317. W. Shi, H. Yun, C. Lin, M. Greenberg, X. Wang, Y. Wang, SahbaTalebi Fard, J. Flueckiger, N. Jaeger, L. Chrostowski, “Ultra-compact,flat-top demultiplexer using anti-reflection contra-directional couplersfor CWDM networks on silicon”, Optics express 21 (6), 6733-6738,2013/3/25.18. V. Donzella, Sahba Talebifard, L. Chrostowski, “Fabrication andexperimental characterization of cascaded SOI micro-rings for high-throughput label-free molecular sensing”, Industrial & Engineering Chem-istry Research, 201319. Sahba Talebi Fard, L. Chrostowski, E. Kwok, M.C. Amann, “Chemo-metric Approach for Improving VCSEL-based Glucose Predictions”,IEEE Transactions on Biomedical Engineering, vol. 57, issue 3, pp.578-585, 02/2010.20. Sahba Talebi Fard, W. Hofmann, P. Talebi Fard, E. Kwok, M.C.Amann, L. Chrostowski, “Optical Glucose Monitoring using VerticalCavity Surface Emitting Lasers (VCSELs)”, SPIE Optics and Photon-ics Symposium - Bio-Sensing-II - Novel and Bio-inspired Sensors, vol.7397, pp. 739704-1 to 739704-11 (Invited), 2/08/2009.21. Sahba Talebi Fard, “Glucose Monitoring”, Master’s Thesis, The Uni-versity of British Columbia, Aug. 2008.22. Sahba Talebi Fard, W. Hofmann, P. Talebi Fard, G. Bohm, M. Ort-siefer, E. Kwok, M.C. Amann, L. Chrostowski, “Optical AbsorptionGlucose Measurements Using 2.3 µm Vertical Cavity SemiconductorLasers”, IEEE Photonics Technology Letters, vol. 20, issue 11, pp.930–932, 06/2008.23. Sahba Talebi Fard, L. Chrostowski, E. Kwok, “Measuring Blood Glu-cose Using Vertical Cavity Semiconductor Lasers (VCSELs)”, Cana-dian Medical and Biological Engineering Society Conference, 06/2007.149Appendix BDerivation of SensitivityFormulaHere we review the derivation of the sensitivity formula for resonator sensorsbased on [43].Considering dispersion in a resonator, the mode condition is given byequation B.1:mpi =2neffLλres(B.1)where, L is the length of the resonator, λres is the resonant wavelength, andneff is the effective index of the waveguide.The group index (ng) is defined as:ng = neff − λresδneffδλ(B.2)Now, equating the mode at initial wavelength and the shifted wavelength(using equation B.1), we have:ng + λresδneffδλλres=ng + (λres + ∆λres)δneffδλ + ∆nclδneffδnclλres + ∆λres(B.3)Solving this equation, we can find wavelength shift as a function of thechange in refractive index of the cladding, which defines sensitivity of theresonator.∆λres∆ncl=λresngδneffδncl(B.4)150


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items