<%BANNER%>

Low Voltage, MEMS-Based Reflective and Refractive Optical Scanners for Endoscopic Biomedical Imaging


PAGE 1

LOW VOLTAGE, MEMS-BASED REFLEC TIVE AND REFRACTIVE OPTICAL SCANNERS FOR ENDOSCOPIC BIOMEDICAL IMAGING By ANKUR JAIN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006

PAGE 2

Copyright 2006 by Ankur Jain

PAGE 3

To my parents, Ranjan and Poonam, to my br other Prateek, and to my fiance Kavitha for their constant love, unwavering sup port, confidence and encouragement.

PAGE 4

ACKNOWLEDGMENTS I would like to thank my advisor, Dr. Huikai Xie, for the constant support and guidance he has given me over the past few years. I first met Huikai in August 2002, and subsequently joined his research group as a PhD student in the fall semester. I am grateful for all the insight he has provided, and am thankful to him for introducing me to the areas of MEMS and endoscopic biomedical imaging. I have personally gained technical expertise, as well as professional know-how through my interactions with him, and I will forever be indebted to him for mentoring me towards becoming a better microsystems technology engineer. The research presented in this dissertation was also painstakingly reviewed by other members of my PhD committee, Dr. Toshikazu Nishida, Dr. Ramakant Srivastava, and Dr. William Ditto, and for that I am grateful. I have enjoyed many conversations with Dr. Nishida, both personally and professionally. Working as a graduate teaching assistant for Dr. Srivastava was a pleasure, and I am grateful for our personal friendship. Dr. Ditto has provided me with unique insights related to the biomedical application aspect of this project. I want to acknowledge technical and personal discussions with Dr. Mark Sheplak and Dr. David Arnold, as their advice helped improve my research work and their pleasant company at MEMS conferences is always welcome. Further appreciation goes out to Dr. Peter Zory for his valuable friendship, for always being a mentor, and for his indispensable lessons on how to maintain a good research notebook. iv

PAGE 5

The Biophotonics and Microsystems Lab (BML) located in 136 Larsen Hall was home to this project, and I am indebted to my BML group members. Special thanks go out to Hongwei Qu for teaching me the ropes in the cleanroom, and to Shane Todd for helping me with electrothermal modeling. Hongwei has always been a pillar of support within BML, and I enjoyed working with him on various projects. Shane worked with me on various micromirror projects, and I have benefited greatly from our professional interactions and personal friendship. I would like to acknowledge support from Anthony Kopa, both personally and also for using my 2-D micromirror for imaging purposes. Other BML members who aided me during the course of my research include Deyou Fang, Maojiao He, Mi Huang, Mingliang Wang, Ben Caswell and the newcomers Kemiao Alex Jia and Lei Wu. Alex and Lei have proven to be worthy successors for my project, and I will value their camaraderie. All BML members work great as a team, and I have so many good memories about the multilingual jokes told in the lab, and the parties and sports that we all participated in. BML is just part of the much bigger microsystems group at the University of Florida, known as the Interdisciplinary Microsystems Group (IMG). I am grateful to all IMG members for their support-group-like environment and technical expertise. In particular, I would like to thank my colleagues Venkat Chandrasekaran, Stephen Horowitz, Anurag Kasyap, Chris Bahr, David Martin, Ryan Holman, Erin Patrick, Israel Boniche, Janhavi Agashe, Sheetal Shetye, Jian Jackie Chan Liu, Yawei Li, Vijay Chandrasekharan, Lee Hunt, Tai-An Chen, Karthik Kadirvel, Robert Taylor, Brandon Bertolucci, Champak Das and Zheng Xia, to name a few. Venkat introduced me to the world of wire-bonding, while Dave, Ryan and Chris kept our IMG server running 24/7. v

PAGE 6

Thanks go to Anurag for helping me with the vibrometer, to Erin for help with the PCB milling machine, and to Israel and Janhavi for AutoCAD assistance. I thank Brandon and the Ultimate Frisbee gang for relentlessly organizing sporting events that helped to upkeep the morale of IMG. Finally, credit is due to all other IMG members for general technical assistance, and for maintaining a lively work environment in the office, lab and even inside the cleanrooms. This work would not be complete without help from our external collaborators. I thank Dr. Yingtian Pan and Zhenguo Wang at the State University of New York at Stony Brook for validating the use of my micromirrors for endoscopic optical coherence tomographic (OCT) imaging. I am grateful that they invited me to visit their lab so that I could witness OCT imaging using my endoscopically-packaged micromirrors. Two-photon excitation fluorescence imaging and second harmonic generation nonlinear optical imaging experiments using my micromirrors were demonstrated in collaboration with Dr. Min Gu and Ling Fu at the Swinburne University of Technology, Australia. I thank Ling for the many hours she has put into this project, and for her personal friendship. I would also like to thank Dr. Michael Bass and Te-Yuan Chung from CREOL, University of Central Florida, Orlando, for letting me use their thermal imager for my research. The MEMS device fabrication was done using the facilities provided by the University of Florida Nanofabrication Facilities (UFNF) and by the UF Microfabritech center. Therefore, I appreciate the support provided by the UFNF staff Al Ogden, Ivan Kravchenko, Bill Lewis and the UF Microfabritech staff. Scanning electron microscopy (SEM) and white light profilometry were performed using the equipment at the Major vi

PAGE 7

Analytical Instrumentation Center (MAIC) at the University of Florida. I wish to thank Dr. Luisa Dempere, Wayne Acree, Andrew Gerger and Brad Willenberg of the MAIC for their assistance. Special thanks go to Tanya Riedhammer who helped me with the Variable-Pressure SEM for imaging my photoresist microlenses. I also want to acknowledge our administrative assistant, Joyce White, for her help and support. Finally, I am eternally grateful to my family and friends for their constant support and encouragement. I would like to thank my parents, Ranjan and Poonam, and my brother, Prateek, for their confidence in me, for their endless love and support, and for keeping me debt-free all through graduate school. I want to thank my fiance Kavitha, for all her love, support, advice, and also for all the car rides to school she provided that ultimately helped my research. Thanks are due to my friends Anuradha Ventakesan and Boman Irani who kept me going throughout graduate school. I also want to acknowledge my friends Kanak Behari Agarwal and Himanshu Kaul whose learned advice helped me through the mid-PhD crisis. The MEMS-based endoscopic biomedical imaging project at the University of Florida has been supported by the National Science Foundation Biophotonics Program through award number BES-0423557, and by the Florida Photonics Center for Excellence. vii

PAGE 8

TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................iv LIST OF TABLES .............................................................................................................xi LIST OF FIGURES ..........................................................................................................xii ABSTRACT ...................................................................................................................xviii CHAPTER 1 INTRODUCTION........................................................................................................1 1.1 Limitations of Conventional Cancer Diagnosis Methodologies............................1 1.2 Emerging Optical Coherence Tomography...........................................................3 1.3 MEMS-based OCT................................................................................................4 1.4 MEMS-based OCM...............................................................................................6 1.5 Research Objectives...............................................................................................9 1.6 Dissertation Overview.........................................................................................10 2 OPTICAL BIOIMAGING METHODOLOGIES.......................................................12 2.1 Optical Coherence Tomography..........................................................................12 2.1.1 OCT System Design...................................................................................13 2.1.2 Key Imaging Parameters............................................................................17 2.1.3 Internal Organ OCT Imaging.....................................................................20 2.1.4 MEMS-based OCT.....................................................................................22 2.2 Optical Coherence Microscopy...........................................................................24 2.2.1 Bench-Top OCM........................................................................................26 2.2.2 MEMS-based OCM....................................................................................27 2.3 Non Linear Optical Imaging................................................................................29 2.3.1 Two-Photon Excitation Fluorescence Imaging..........................................30 2.3.2 Second Harmonic Generation Imaging......................................................33 2.3.3 Nonlinear Optical Imaging System Design................................................35 2.3.4 Endoscopic Nonlinear Optical Imaging.....................................................37 2.3.5 MEMS-based Endoscopic Nonlinear Optical Imaging..............................39 3 ELECTROTHERMAL MICROMIRRORS AND ENDOSCOPIC OCT IMAGING...................................................................................................................40 viii

PAGE 9

3.1 Scanning Micromirrors........................................................................................40 3.2 Electrothermal Actuation and Design..................................................................44 3.3 Microfabrication Process.....................................................................................46 3.4 Bimorph Actuation and Theoretical Analysis.....................................................50 3.5 Electrothermal Micromirrors...............................................................................54 3.5.1 One-Dimensional Electrothermal Micromirror..........................................54 3.5.2 Two-dimensional Electrothermal Micromirror..........................................57 3.5.2.1 Device design.................................................................................57 3.5.2.2 Device characterization..................................................................59 3.5.2.3 Laser scanning experiment.............................................................62 3.6 Micromirror Packaging........................................................................................63 3.7 MEMS-based Endoscopic OCT Imaging............................................................65 3.7.1 MEMS-based OCT System Design............................................................65 3.7.2 OCT Imaging Results.................................................................................71 3.8 Summary..............................................................................................................73 4 LARGE-VERTICAL-DISPLACEMENT MICROMIRRORS AND NON-LINEAR OPTICAL IMAGING.................................................................................75 4.1 LVD Microactuator Design.................................................................................77 4.2 1-D LVD Micromirror.........................................................................................80 4.2.1 Fabricated Device.......................................................................................80 4.2.2 Equivalent Circuit Model...........................................................................81 4.2.3 Experimental Results..................................................................................86 4.2.3.1 Static response................................................................................86 4.2.3.2 Frequency response/resonant scanning..........................................89 4.3 2-D LVD Micromirror.........................................................................................92 4.3.1 Mirror Design.............................................................................................92 4.3.2 Experimental Results..................................................................................94 4.3.2.1 Bi-directional scanning..................................................................94 4.3.2.2 Two-dimensional dynamic scanning..............................................98 4.3.2.3 Vertical displacement motion.......................................................100 4.4 MEMS Mirror-based Nonlinear Endoscopy......................................................102 4.4.1 Nonlinear Optical Imaging System..........................................................103 4.4.2 Experimental Results................................................................................104 4.5 Summary............................................................................................................107 5 MICROLENS SCANNERS AND OPTICAL CONFOCAL MICROSCOPY........109 5.1 LVD Microlens Scanner....................................................................................109 5.1.1 Microlens Scanner Design........................................................................111 5.1.2 Fabricated Microlens Scanner..................................................................111 5.1.3 Experimental Results................................................................................115 5.2 Millimeter-Range LVD Microlens Scanner......................................................119 5.2.1 Millimeter-Range Scanner Design...........................................................120 5.2.2 Fabrication Process...................................................................................121 5.2.3 Experimental Results................................................................................126 ix

PAGE 10

5.3 LVD Microlens Packaging................................................................................132 5.4 Summary............................................................................................................134 6 CONCLUSIONS AND FUTURE WORK...............................................................135 6.1 Research Effort Accomplishments....................................................................137 6.2 Future Work.......................................................................................................138 APPENDIX A NON-CMOS, WAFER LEVEL FABRICATION PROCESS.................................140 B ARTICLES GENERATED BY THIS RESEARCH EFFORT................................145 LIST OF REFERENCES.................................................................................................148 BIOGRAPHICAL SKETCH...........................................................................................162 x

PAGE 11

LIST OF TABLES Table page 3-1 Thermomechanical properties of some possible bimorph materials at room temperature..............................................................................................................45 4-1 Parameters used by the equivalent circuit model of the 1-D LVD micromirror......84 4-2 Actuator characteristics for the 2-D LVD micromirror...........................................96 5-1 Microlens characteristics........................................................................................113 5-2 Estimated microlens parameters for various desired focal lengths........................126 xi

PAGE 12

LIST OF FIGURES Figure page 1-1 Schematic of a MEMS-based OCT/OCM system. (a) System block diagram. (b) Optical delay line that uses the LVD micromirror as a reference mirrors for transverse and axial scanning. (c) OCT endoscope that uses a 1-D or 2-D LVD micromirror for transverse scanning of tissue. (d) OCM endoscope that uses a LVD microlens for axial scanning of tissue...............................................................7 2-1 OCT schematic.........................................................................................................13 2-2 OCT tissue scanning modes. (a) Conventional longitudinal scanning (A-scan). (b) En-face scanning (B-scan)..................................................................................15 2-3 Comparison between histology, ultrasound and OCT images of biological tissue. (a) HE-stained histology, (b) 50-MHz ultrasound, and (c) OCT image of a nevus.........................................................................................................................16 2-4 Comparison between ultrasound and OCT images of human coronary artery plaques. (a) In vivo OCT image with axial imaging resolution of 13 m. (b) 30 MHz intravascular ultrasound image of the same artery with a lower resolution of 100 m................................................................................................17 2-5 Gaussian beam optics...............................................................................................19 2-6 Existing endoscopic OCT probe designs.(a) Radial imaging probe design. (b) Forward imaging probe design.................................................................................20 2-7 MEMS-based endoscopic OCT system....................................................................24 2-8 High NA scanning approaches for OCT. (a) Reference mirror and focusing optics placed on a movable stage to achieve dynamic focusing. (b) Lateral and axial scanning achieved by displacement of the fiber tip.........................................25 2-9 MEMS-based endoscopic OCM schematic..............................................................26 2-10 Difference between OCM and OCT.........................................................................27 2-11 1-m axial resolution by 3-m lateral resolution tomogram of a tadpole...............28 xii

PAGE 13

2-12 Energy band diagrams illustrating (a) one photon and, (b) two photon excitation fluorescence phenomena..........................................................................................31 2-13 Optical sectioning ability of TPEF imaging. (a) Single-photon excitation of fluorescein by 488 nm light. (b) Two-photon excitation using 960 nm light..........32 2-14 Schematic of a nonlinear microscope.......................................................................36 2-15 (a) Pulse train from a mode-locked Ti:Sapphire laser at 80 MHz. (b) The laser pulses typically have a FWHM duration of 100 fs, and (c) a spectral FWHM bandwidth of ~ 10 nm..............................................................................................36 2-16 Two-photon fluorescence and SHG signals emitted by a sample excited by 800 nm light.............................................................................................................37 2-17 Two-photon imaging of the hamster cheek pouch tissue at an excitation wavelength of 780 nm. (a) Normal tissue. Precancerous tissue: (b) Moderate Dysplasia. (c) Carcinoma in situ. Cancerous tissue: (d) nonpapillary, and (e) papillary squamous cell carcinoma..........................................................................38 3-1 Side view of a bimorph beam...................................................................................45 3-2 Electrothermal micromirror basic structure. (a) Top view. (b) Cross-sectional side view...................................................................................................................47 3-3 DRIE CMOS-MEMS fabrication process flow.......................................................48 3-4 SEM of a fabricated 1-D micromirror with initial tilt angle....................................50 3-5 Bimorph actuation mechanism. Side views of: (a) Initial position of mirror at zero bias. (b) Downward rotation of mirror plate on application of bias voltage to polysilicon resistor...............................................................................................51 3-6 SEM of 1-D micromirror..........................................................................................55 3-7 1-D mirror characterization. (a) Rotational static response. (b) Plot of the heater resistance versus applied current..............................................................................56 3-8 Frequency response of the 1-D mirror....................................................................56 3-9 Schematic of the 2-D mirror design. (a) Top view showing the axes of rotation. (b) Cross-sectional view of A-A.............................................................................58 3-10 SEM of a fabricated 2-D micromirror......................................................................59 3-11 2-D Mirror Characterization. (a) Rotation angle vs. current, and (b) Polysilicon resistance vs. current for the two actuators..............................................................60 xiii

PAGE 14

3-12 Thermal images of a device biased at 10 V. (a) Temperature distribution across the mirror actuator only. (b) Thermograph of the entire device...............................61 3-13 Laser scanning using the 2-D mirror. (a) Schematic of experimental setup. (b) 4 x 4 pixel images scanned by the micromirror..........................................................63 3-14 Micromirror package. (a) Packaged micromirror on a custom PCB. (b) Picture of the PCB package. (c) Picture of a packaged mirror alongside a US dime coin...64 3-15 Endoscopic OCT probe designs. (a) Side-imaging configuration. (b) Forward-imaging configuration..............................................................................................65 3-16 Schematic of the MEMS-based endoscopic OCT system........................................66 3-17 Photographs of the 5-mm diameter MEMS-based OCT endoscope at the State University of New York at Stony Brook..................................................................69 3-18 Photograph of the portable, MEMS-based endoscopic OCT system at the State University of New York at Stony Brook..................................................................70 3-19 Comparison of OCT with histological image. (a) OCT image, and (b) histological image of rat bladder..............................................................................71 3-20 Bench-top versus MEMS-based endoscopic OCT imaging of rat bladder. (a) Bench-top OCT image. Size: 6 mm by 2.7 mm. (b) Endoscopic MEMS-based OCT image. Size: 4 mm by 2.7 mm.........................................................................72 4-1 Design schematic of the LVD mirror. (a) Top view. (b) Cross-sectional view across A-A..............................................................................................................77 4-2 Coventor simulations. (a) Device side-view. (b) 3-D model of the LVD micromirror illustrating the initial curling of the bimorph actuators.......................78 4-3 Wiring schematic for the LVD actuators.................................................................79 4-4 SEM images of the LVD micromirror.....................................................................81 4-5 Line scan of the surface profile of the LVD micromirror........................................81 4-6 Equivalent circuit model of the LVD micromirror device.......................................82 4-7 SEM of the burn pattern of the mirror actuator........................................................85 4-8 LVD mirror characterization. Plots of the (a) rotation angle versus applied voltage, and (b) polysilicon heater resistance versus applied voltage for the two actuators. (c) Plot showing the linear correlation between rotation angle and polysilicon resistance of the actuators......................................................................88 xiv

PAGE 15

4-9 Piston motion mode. (a) Vertical displacement of the mirror plate as a function of the frame actuator voltage. (b) Plot of the mirror actuator voltage versus frame actuator voltage that was used to drive the LVD device to obtain less than 0.03 tilting of the mirror plate.................................................................................90 4-10 Frequency response of the LVD micromirror device...............................................91 4-11 2-D LVD micromirror design. (a) Top view of the 2-D micromirror, highlighting the 4 bimorph actuators. (b) Top view of the actuator area boxed in part (a). (c) Cross-sectional view of the bimorph actuator as seen across A-A......93 4-12 SEM of a fabricated 2-D LVD mirror......................................................................93 4-13 Static 2-D line scans. (a) Plot showing the optical angles scanned in 2-D space when each actuator is individually actuated. (b) Plot of the effective optical angle scanned versus actuation voltage for each actuator........................................95 4-14 Static characterization. (a) Plot showing the linear scan pattern during static 2-D scanning of Act1 and Act4 only. Act4 was actuated at different Act1 bias voltages. (b) Linear plot of actuator resistance versus optical scan angle for each actuator.....................................................................................................................96 4-15 Tilt angle stability of the mirror plate versus time...................................................97 4-16 Initial tilt angle of the mirror plate in x and y directions at different environmental temperatures.....................................................................................98 4-17 Photographs of 2-D scan patterns obtained by exciting actuators 1 and 4 only. (a)-(e) Lissajous figures scanned by the micromirror by varying only the phase of the two excitation signals. (f) Lissajous figure scanned at an excitation frequency ratio of 1:10.............................................................................................99 4-18 2-D raster scanning pattern obtained using actuators 1 and 4..................................99 4-19 Piston motion mode. (a) Vertical displacement of the mirror plate as a function of Act4 voltage. (b) Corresponding plot of the relationship between Act3 and Act4 voltages that is required to generate the vertical displacement. (c) Linear increase of Act3 and Act4 resistance with vertical displacement..........................101 4-20 Tilting of the mirror plate in the negative y-direction (due to thermal coupling) as a function of vertical position of the mirror.......................................................102 4-21 Schematic of the nonlinear optical imaging system...............................................103 4-22 Cross-sectional view of the double-clad photonic crystal fiber.............................104 xv

PAGE 16

4-23 Second harmonic generation imaging. (a) Series of SHG line profiles from rat tail tendon taken at 10 m axial steps. (b) SHG line profile of an unstained rat esophagus tissue.....................................................................................................105 4-24 2-D raster scan pattern scanned by actuators 1 and 4 only....................................106 4-25 TPEF imaging of 10-m diameter fluorescent microbeads. (a) Bench-top TPEF imaging system. (b) MEMS-based TPEF imaging................................................106 4-26 In vitro imaging of rat stomach epithelial surface stained with 1% acridine orange in Ringers solution....................................................................................107 5-1 Design schematic of the LVD microlens scanner. (a) Top view. (b) Cross-sectional side view. (c) 3-D illustration of the scanner..........................................112 5-2 SEMs of: (a) Fabricated LVD lens holder, and (b) LVD microlens scanner.........113 5-3 Microlens fabrication process. (a) Backside Si etch. (b) Oxide etch. (c) Deep Si trench etch. (d) Si undercut. (e) Microlens formation by reflow of photoresist.....114 5-4 SEM of a fabricated PR microlens.........................................................................115 5-5 Vertical displacement experiment. (a) Vertical displacement of the microlens as a function of frame actuator voltage. (b) Plot of the ratio of the applied voltages to the lens and frame actuators that was used to obtain the vertical displacement shown in (a)............................................................................................................116 5-6 Plot showing the increase in polysilicon heater resistances versus vertical displacement for the two actuators.........................................................................117 5-7 Microlens imaging quality. (a) Schematic of the imaging experiment apparatus. (b) Photo of the test pattern. (c) Snap-shot images of the test pattern as obtained through the PR microlens.......................................................................................118 5-8 CCD image of a 4 m focused beam spot (top), and its corresponding intensity profile (bottom)......................................................................................................119 5-9 Top view of the millimeter-range LVD microlens scanner...................................120 5-10 Modified fabrication process for mm-LVD microlens scanner. (a) Backside Si etch. (b) Oxide etch. (c) Spin on photoresist. (d) Anisotropic photoresist etch to expose metal-2 layer. (e) Metal wet etch followed by photoresist removal. (f) Deep Si trench etch. (g) Silicon undercut. (h) Microlens formation by reflow of photoresist..............................................................................................................121 5-11 Modified process for die-level fabrication of the microlens scanner.....................122 xvi

PAGE 17

5-12 Photograph of the CMOS die after the wet etch of lithographically exposed aluminum layers.....................................................................................................123 5-13 (a) SEM of a fabricated scanner before microlens formation. Close-up views of (b) lens actuator; and (c) frame actuator bimorph regions.....................................124 5-14 SEM of a fabricated microlens scanner with integrated polymer microlens.........124 5-15 SEMs of (a) convex microlens, and (b) ball-type microlens..................................125 5-16 Imaging using the photoresist microlens. (a) Photograph of the test pattern on a chrome mask. (b) Corresponding image of the test pattern as seen at the focal plane of the photoresist microlens..........................................................................126 5-17 Vertical displacement of the microlens scanner. (a) Microlens displacement as a function of frame actuator voltage. (b) Displacement versus applied electrical power. (c) Corresponding linear relationship between the two voltages...............128 5-18 Change in resistance versus vertical displacement of the microlens......................129 5-19 Change of initial microlens elevation with ambient temperature...........................129 5-20 Lateral shift of the microlens during vertical displacement actuation. (a) Illustration of the lateral shift. (b) Characterized plot of the lateral shift...............130 5-21 Dynamic response of the microlens scanner. (a) Mechanical response when a square excitation was applied to both actuators at t = 0. This can be fitted to an exponential envelope using the damping ratio, (b) Frequency response showing the resonant peaks....................................................................................131 5-22 Microlens package design. (a) Microlens package schematic. (b) Forward-imaging OCM endoscope.......................................................................................132 5-23 Packaged OCM endoscope. (a) With the Lucite end cap. (b) Without the end cap to show the packaged microlens scanner. .............................................................133 A-1 Non-CMOS, wafer-level fabrication process illustrating the steps required to fabricate the 1-D LVD micromirror device............................................................140 xvii

PAGE 18

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy LOW VOLTAGE, MEMS-BASED REFLECTIVE AND REFRACTIVE OPTICAL SCANNERS FOR ENDOSCOPIC BIOMEDICAL IMAGING By Ankur Jain August 2006 Chair: Huikai Xie Major Department: Electrical and Computer Engineering Imaging technologies such as optical coherence tomography (OCT), two-photon excitation fluorescence microscopy (TPEF), and second harmonic generation (SHG) microscopy require optical scanners to transversely scan a focused laser beam onto the tissue specimen being imaged. However, for in vivo early-cancer detection of internal organs the optical scanners must be integrated into slender endoscopes. The goal of this work is to develop millimeter-sized MEMS optical scanners packaged inside endoscopes to enable endoscopic biomedical imaging. This work reports MEMS micromirrors and microlens scanners fabricated using post-CMOS micromachining processes, which can provide large scan ranges at low driving voltages. Several 1-D and 2-D micromirror scanners have been designed, fabricated and characterized. Scanning micromirrors, as large as 1.3 by 1.1 mm 2 have demonstrated optical scan angles greater than 40 at actuation voltages below 20Vdc. The xviii

PAGE 19

maximum scanning speed of these devices is in the range of 200 to 500Hz, which is adequate for real-time bio-imaging. A new electrothermal microactuator design is reported which enables large vertical displacements (LVD). This LVD microactuator uses two sets of electrothermal bimorph thin-film beams to provide out-of-plane elevation to the micromirror, while keeping the mirror parallel to the substrate. LVD micromirrors have demonstrated large bi-directional scanning ability (>) as well as large vertical piston motion (~0.5mm) at low driving voltages (<15V). A 1-D LVD micromirror has the ability to scan optical angles greater than 170 at its resonance frequency of 2.4kHz. Polymer microlenses integrated with the LVD microactuators have been developed for endoscopic optical coherence microscopy which requires microlenses to axially scan their focal planes by 0.5 to 2 mm. A modified fabrication process allows larger polymer lenses with better thermal isolation to be integrated. A maximum vertical displacement of 0.71mm was obtained. These scanners have been packaged inside 5-mm diameter endoscopes to enable in vivo imaging. Endoscopic OCT with transverse and axial resolutions of 15m and 12m, respectively has been demonstrated at imaging speeds of 2 to 6 frames/second. TPEF and SHG imaging with imaging resolution greater than 1.5m has been obtained. These results show the potential for the use of MEMS-based endoscopy for early-cancer detection. xix

PAGE 20

CHAPTER 1 INTRODUCTION Cancer is a leading killer disease worldwide, and it accounts for as many as one-quarter of deaths in the United States of America. For the year 2005, the American Cancer Society estimates 570,280 cancer deaths in the US, and expects more than 2 million new cancer cases to be diagnosed [1]. Although cancers in the breast, colon, rectum, cervix, prostrate, skin and the oral cavity are readily treatable provided they are diagnosed at a pre-invasive stage, early lesions in these tissues are often almost impossible to detect without regular screening. A study has estimated that the 5-year survival rate of patients with these types of cancers can increase to 95% if the cancers are diagnosed during their localized precancerous stage [1]. Therefore, early detection of many of these precancerous lesions is essential in order to greatly reduce patient morbidity and mortality. The goal of this research effort is to develop endoscopic imaging modalities that can detect and diagnose in vivo precancerous lesions. It is proposed to achieve endoscopic imaging through the use of miniature optical scanners packaged inside endoscopes. This chapter discusses the limitations of existing cancer diagnostic techniques and introduces a new cancer detection technique along with the objectives of the project. 1.1 Limitations of Conventional Cancer Diagnosis Methodologies Cancer researchers have estimated that more than 85% of all cancers originate inside the epithelium layer that lines the internal surfaces of organs throughout the human 1

PAGE 21

2 body [2]; therefore it is of paramount importance to develop methods that can diagnose cancers just below the tissue surface. The existing diagnosis of cancers is carried out through visual inspection of the tissue surface followed by random tissue biopsy. Internal organ cancer screening is conducted by using special biopsy endoscopes that are equipped with cameras for visually inspecting the internal organ tissue surfaces. Since precancers originate below the tissue surface, conventional endoscopes that only image the tissue surface are unable to make an accurate diagnosis. Therefore, this current practice of white-light endoscopy often requires biopsies for ex vivo histological analysis and clinical diagnosis of suspect tissue. This biopsy procedure creates significant delay in clinical diagnosis, with the added risk and cost of the medical procedure. Another limitation is the biopsy tissue sampling density. A study performed by Reid et al. on the early detection of high-grade dysplasia in Barretts esophagus proved that by reducing the tissue biopsy sampling interval from 2 cm to 1 cm, the success in detecting cancer was doubled [3]. However, even this practice of biopsy over-sampling suffers from substantial limitations since there is a practical limit in the number of biopsies that can be performed, thereby diminishing its diagnostic potential. Imaging techniques such as radiography, computed tomography (CT), magnetic resonance imaging (MRI), and ultrasound allow noninvasive investigation of large-scale structures in the human body and also permit three-dimensional (3-D) visualization. However, the imaging resolution of these existing diagnostic techniques makes the detection and diagnosis of many precancers difficult if not impossible. For example, bronchial cancers are not commonly detected at curable stages since the precancerous Ali Fazel, M.D., Personal Communication, Gainesville, FL, 2004.

PAGE 22

3 lesions are generally smaller than the detection limit of current imaging modalities. The spatial resolution of approaches such as conventional radiography, CT, and MRI is generally restricted to a few millimeters in standard clinical practice [4], thereby preventing the detection of lesions less than 1 cm in diameter [5]. However, for detecting cancer in its early stages, an imaging technology with a higher resolution (< 20 m) is necessary for accurate diagnosis. In addition, clinical screening procedures such as the random biopsy procedure for the diagnosis of cancer can be improved by using a high-resolution, non-invasive imaging technique to identify biopsy sites that correspond to the most severe disease. 1.2 Emerging Optical Coherence Tomography Optical coherence tomography (OCT) is an emerging diagnostic medical imaging technology that produces high-resolution, cross-sectional images of biological samples [6-8]. Optical coherence tomography combines the operating principles of ultrasound with the imaging performance of a microscope. It uses advanced photonics and fiber optics to focus an infrared light beam into a sample, and then uses low-coherence interferometry to measure the echo time delay of the reflected light to determine tissue microstructure. The OCT imaging depth is limited by optical attenuation from tissue absorption and scattering to about 2 to 3 mm. This is the same scale as that generally imaged using biopsy and histology. A very attractive feature of OCT imaging is the high resolution. Although ultrasound imaging has greater imaging depth, OCT has a much higher imaging resolution of 10 m or less [9]. An OCT system with 1 m axial resolution has also been demonstrated [10], which is about two orders of magnitude higher than that of standard ultrasound imaging. Even though high-frequency ultrasonic

PAGE 23

4 imaging has been shown to obtain a resolution of about 10 m [11], the simplicity of OCT systems permits a much lower hardware cost. Since OCT uses infrared light it is much safer to use than CT systems which use harmful x-rays. OCT imaging is minimally-invasive and has the potential to eliminate risky and time-consuming biopsy procedures; therefore it is also known as optical biopsy. Optical coherence tomography has been proved to be clinically useful in the field of ophthalmology, and has great potential for use in cardiovascular, gastrointestinal and pulmonary imaging through the use of endoscopes and catheters [5, 12, 13]. Endoscopic OCT systems have been demonstrated to be able to detect in vivo cancers at a very early stage [14, 15]. For these internal organ applications, the imaging probe must be small, and fast image scanning is required. Various methodologies have been proposed to transversely scan the optical beam across the internal tissue surface. Some endoscopic OCT devices use a rotating hollow cable that carries a single-mode optical fiber, while others use a galvanometric plate or piezoelectric transducer, that swings the distal fiber tip to perform in vivo transverse scanning of tissue [12, 14, 16]. 1.3 MEMS-based OCT Micro-Electro-Mechanical Systems (MEMS) or Microsystem Technology (MST) is another emerging technology that makes miniature sensors and actuators through batch-fabrication micromachining processes. Micromirrors manufactured using this technology have been widely used for optical displays and optical switching [17, 18]. MEMS-based transducers have been also widely used by the automobile industry which uses accelerometers and other inertial sensors for deploying safety air-bags and other vehicle stability applications. The small size, fast speed and low power consumption of MEMS

PAGE 24

5 mirrors make them ideal for use in an endoscopic OCT imaging probe. In fact, researchers have started to use MEMS mirrors for the transverse scanning of endoscopic OCT systems [19-23]. Xie et al. demonstrated a 5 mm diameter MEMS-based OCT endoscope that used a 1-D electrothermal mirror to scan the light beam onto the biological tissue [19]. By performing 1-D transverse scans of the tissue, high resolution cross-sectional 2-D images were obtained. Zara et al. also reported an endoscopic OCT system based on MEMS mirrors in 2002 [24], in which the MEMS mirror has large deflection angle but requires elaborate assembly. Tran et al. [25] and Herz et al. [26] demonstrated radial endoscopic-OCT imaging using MEMS micromotors packaged inside endoscopes that rotated a prism or mirror. More recently Fan et al. [22] and McCormick et al. [23] separately demonstrated 3-D endoscopic OCT imaging through the use of 2-D electrostatic micromirrors packaged inside fiber-optic endoscopes. However, the high voltages required for electrostatic actuation may be a concern due to electrical safety issues during internal organ imaging. Even though the electrothermal micromirrors used by Xie et al. [19] operate at low voltages, the large initial tilt angle of the mirror plate complicated the endoscope package design. Another limitation of all these existing micromirror-based OCT endoscopes is that their lateral resolution is restricted to a few tens of microns in order to provide the necessary millimeter-range depth of focus. These lateral resolutions are not sufficient since a much high lateral resolution (< 10 m) is required for the detection of in vivo precancers. These issues regarding the use of existing MEMS scanners for OCT imaging will be discussed in greater detail in Chapters 2 and 3.

PAGE 25

6 Proposed solution: In prior work, 1-D and 2-D electrothermal micromirrors with large rotation angles at low actuation voltages were designed for transverse scanning in OCT imaging [27-30]. However the unidirectional operation, non-stationary center of rotation and large initial tilt angle of these mirrors complicated device packaging and optical design. These issues can be resolved by using a novel microactuator design that uses two complementarily-oriented electrothermal actuators to keep the mirror parallel to the substrate, and these actuators also provide bi-directional scanning capability to the mirror. This actuator pair can also generate large, out-of-plane, piston motion at low actuation voltages (< 15V). MEMS devices using this novel microactuator design are referred to as large-vertical-displacement (LVD) microdevices. It is proposed to use micromirrors integrated with either one or two sets of LVD microactuators to perform 1-D or 2-D transverse scanning, respectively. The fabricated mirrors will be packaged inside endoscopes to perform OCT imaging. Also, further miniaturization of the overall OCT system is possible by replacing the bulky axial scanning mirror with a phase-only LVD micromirror. Figure 1-1 shows the schematic of a MEMS-based OCT system in which the LVD micromirror can be used for axial reference scanning as well as endoscopic transverse bi-directional scanning applications. Further details about the LVD micromirrors are provided in Chapter 4. 1.4 MEMS-based OCM Optical coherence microscopy (OCM) is an extension of OCT imaging technique, and it allows for ultrahigh-resolution cross-sectional imaging of highly-scattering tissue by combining the imaging capabilities of OCT technology and high numerical-aperture (NA) confocal microscopy. In OCM, the high imaging resolution in the axial direction is provided by low-coherence interferometry, while the micron imaging resolution in the

PAGE 26

7 Figure 1-1: Schematic of a MEMS-based OCT/OCM system. (a) System block diagram. (b) Optical delay line that uses the LVD micromirror as a reference mirrors for transverse and axial scanning. (c) OCT endoscope that uses a 1-D or 2-D LVD micromirror for transverse scanning of tissue. (d) OCM endoscope that uses a LVD microlens for axial scanning of tissue. CL: collimating lens. FM: Fixed mirror. Ax ial Scann i ng Op tica l Delay Line Pi st on -m ot i on m i crom irror (a) (b) (c) Fiber B r oa dba n d Li ght So urce Photodetector 50: 50 Si gnal Processing Refere nce arm Fiber coupler Sam p le arm Tissue Sam p le Optical Delay Line Tissue Scanning Opti cs OCT En dosc o pe Tissue Sam p le CL Bi-d irection a l m i crom irror f o r trans v erse sca n s Fiber OCM En dosc o pe Tissue Scan ning Op tics FM (d) Tissue CL Fiber Tissue Scan ning Op tics Microlens scanner fo r a x ial an d trans v erse sca n s

PAGE 27

8 lateral direction is provided by confocal microscopy. Bench-top OCM systems have demonstrated lateral imaging resolutions better than 3 m [10], and thereby have the ability to detect precancerous lesions at the cellular level (similar to histology). Therefore, in order to detect and diagnose precancers in internal organs, endoscopic OCM apparatus is highly desirable. Unfortunately, the high-NA optical components of OCM systems are bulky, therefore existing OCM systems are restricted to bench-top set-ups just like standard microscopes. Researchers have been investigating various methodologies to develop an endoscopic OCM system, and till date none have been reported in literature. Endoscopic OCM essentially requires a scanning mechanism which can vertically displace a highly-focused light spot by up to 2 millimeters inside tissue. Fitting a high-NA optical scanner that meets this requirement into a millimeter-scale endoscope has been a challenge. Proposed solution: This project proposes to integrate a high-NA microlens with an LVD microactuator to form a LVD microlens scanner which can then be fitted into an endoscope for OCM imaging. This endoscopic OCM probe will then be used to obtain high-resolution images in both lateral and longitudinal directions. Since high lateral resolution results in a reduced depth-of-focus, the LVD microactuator will be used for vertically displacing the microlens in order to focus a light spot at different depths inside tissue. A schematic of an OCM endoscope is shown in Figure 1-1(d). The LVD microlens scanner design allows it to axially displace the focal plane of the scanning microlens by up to a few millimeters. These LVD scanners are appropriate for endoscopic OCM systems since the scanners are small enough to fit inside millimeter-sized endoscopes,

PAGE 28

9 and also need low voltages for actuation. Details about MEMS-based OCM and LVD microlens scanners are presented in Chapter 5. 1.5 Research Objectives The main goal of this research project is to develop miniature optical scanners for an endoscopic imaging modality that can detect and diagnose in vivo precancerous lesions. This main goal has been further subdivided into two approaches for this research effort. First, this work aims to extend and improve on the MEMS-based endoscopic OCT imaging technology developed by Xie [31], by developing novel reflective optical micro-scanners. The aim of this approach is to fabricate reflective scanners that are capable of providing large bi-directional optical scans (>20) at low actuation voltages (<20 V), and they also should be small enough to fit inside a 5-mm diameter endoscope. A two-dimensional optical scanner will also be developed in order to enable three-dimensional OCT imaging. These micromirrors can also be used with other endoscopic imaging techniques such as two-photon excitation fluorescence and second harmonic generation microscopy for in vivo visualization of precancers. Secondly, this project aims to develop a MEMS-based OCM system which uses a refractive micro-scanner to provide ultrahigh resolution endoscopic imaging for the detection of early cancers. The objective of this approach is to develop microlens scanners which can focus a light beam at different depths inside biological tissue. This scanner should be capable of providing millimeter-range displacements at actuation voltages below 20 V, and should also be small enough to fit inside 5-mm diameter endoscopes. A prototype microlens scanner was initially developed to demonstrate proof

PAGE 29

10 of-concept, and this scanner design was then scaled to develop millimeter-scale scanners that meet the requirements of this project. 1.6 Dissertation Overview This dissertation is divided into six chapters. The first chapter provides the motivation for this work and a background of current imaging methods used for the detection of cancer. Chapter 2 provides the background information on three emerging biomedical imaging technologies that can perform in vivo detection of precancerous lesions. These include optical coherence tomography and microscopy, and nonlinear optical imaging techniques of two-photon excitation fluorescence and second harmonic generation. The required OCT scanning-probe characteristics for endoscopic OCT imaging are also explained. Chapter 3 provides a comprehensive literature review of various MEMS micromirror design structures and their limitations for use in endoscopic OCT imaging systems. Electrothermally-actuated micromirror designs are also introduced and their principles of operation and fabrication process are explained in great detail. Endoscopic OCT imaging using these micromirrors is also demonstrated. Chapter 4 presents a novel large-vertical-displacement (LVD) microactuator design that has the ability to perform bi-directional rotational motion as well as generate large vertical displacements. 1-D and 2-D LVD micromirrors using this actuator design are demonstrated. Nonlinear optical endoscopy using these devices is also presented. In Chapter 5, a novel LVD microlens scanner design along with experimental results is presented. This microlens scanner will be used for endoscopic optical coherence microscopy (OCM) imaging, and has been packaged inside a 5-mm diameter endoscope.

PAGE 30

11 Finally, Chapter 6 summarizes the entire research effort and lists suggestions for future work, along with a list of accomplishments for this project.

PAGE 31

CHAPTER 2 OPTICAL BIOIMAGING METHODOLOGIES 2.1 Optical Coherence Tomography Optical coherence tomography (OCT) is an emerging in vivo diagnostic medical imaging technology that produces high-resolution, cross-sectional images of biological samples [6-8]. Optical coherence tomographic imaging technology is an optical analogy of the more conventional ultrasonic pulse-echo imaging technology which measures the intensity and echo delay of acoustic waves to determine tissue microstructure. Since the speed of light is many orders of magnitude faster than that of acoustic waves, a direct measurement of optical echoes cannot be obtained electronically as in ultrasound imaging. Therefore OCT uses an optical measurement technique known as low-coherence interferometry to measure the optical delay information in the back-reflected signal from tissue. Low-coherence interferometry was initially developed and demonstrated for optical-coherence domain reflectometry (OCDR), a 1-D optical ranging technique used for locating faults in fiber optic cables [32-34]. Optical coherence tomography is based on the principles of low-coherence interferometery, and it uses advanced photonics and fiber-optics to image high-resolution cellular structure of tissues at depths greater than conventional microscopes. This section presents the operating principle of OCT imaging technology and also discusses the scanning probe requirements for endoscopic OCT imaging. 12

PAGE 32

13 2.1.1 OCT System Design The schematic of an OCT system is shown in Figure 2-1. The core of this system is a Michelson interferometer, which uses a broadband light source (BBS) to provide a low-coherence infrared light beam. This low-coherence infrared light beam is split at a fiber coupler into the reference and sample arms of the interferometer. The light in the sample arm is focused onto the sample; and the reflected light containing time-of-flight information is collected by the same optical fiber. The reflections from the sample and reference arms are then combined at the coupler and their optical interference is detected by a photodetector. Figure 2-1: OCT schematic. Optical interference is detected by the photodetector only when the optical path difference of the reference and sample arms is within the coherence length of the light source. That is, only the light reflected back from a particular depth of the sample is detected. This is called coherence gating. The amplitude of the interferometric signal (detected by the photodetector) provides a direct measure of the intensity of backz 1 z 3 z 2 n z Reference mirror Axial scan Broadband V z 1 z 3 z 2 z Light Source Beam splitter Photo-detector Transverse scan z Tissue

PAGE 33

14 scattered light from a tissue segment as thin as the coherence length of the BBS. Since OCT imaging provides tissue microstructure information pixel by pixel, scanning mechanisms are required to scan the tissue in axial and lateral directions. The depth information of the tissue sample is acquired through the axial (z-axis) scanning of an optical delay line (i.e., the reference mirror in Figure 2-1), while the lateral information is obtained by transversely scanning the light beam in the sample arm of the interferometer. Multiple longitudinal scans are performed at different lateral locations to provide a two-dimensional data set which contains the back-scattering information of a tissue cross-section. This 2-D data is then displayed as a grayscale or false color OCT image. Since multiple longitudinal scans are performed at different lateral positions, this scanning mechanism is similar to the A-scan image scanning method used in ultrasound. En-face scanning is another method in which the tissue is transversely scanned at different longitudinal locations to generate a B-scan OCT image. A schematic of these scanning operations is shown in Figure 2-2. Depending on the coherence length of the employed broadband light source, OCT can provide cellular or even sub-cellular resolutions (1~20 m), which are one or two orders of magnitude higher than that of commonly used ultrasound imaging (~100 m) [9]. Figures 2-3 and 2-4 show the difference in tissue image resolutions obtained using ultrasound and OCT technologies [9, 35]. Since more than 85% of all cancers originate in the tissue epithelial layer, which is within the penetration depth (a few mm) of infrared light [2, 36, 37], malignant or premalignant changes of epithelia can be detected at a very early stage without performing biopsies. Also by using OCT imaging along with conventional biopsy, highly

PAGE 34

15 Figure 2-2: OCT tissue scanning modes. (a) Conventional longitudinal scanning (A-scan). (b) En-face scanning (B-scan). Adapted from Podoleanu et al. [38]. x z SLOW FA S T SLOW FA S T (a) (b) x Tissue Surface z Tissue Surface

PAGE 35

16 (a) (b) (c) Figure 2-3: Comparison between histology, ultrasound and OCT images of biological tissue. (a) HE-stained histology, (b) 50-MHz ultrasound, and (c) OCT image of a nevus. 1996 IEEE. Reprinted, with permission, from Pan et al. [35].

PAGE 36

17 Figure 2-4: Comparison between ultrasound and OCT images of human coronary artery plaques. (a) In vivo OCT image with axial imaging resolution of 13 m. (b) 30 MHz intravascular ultrasound image of the same artery with a lower resolution of 100 m. Reprinted from Jang et al. [39], Copyright 2002, with permission from The American College of Cardiology Foundation. suspicious tissue areas can be easily identified which can reduce the randomness of biopsies. Optical coherence tomography has been applied to a wide variety of biological tissue and organ systems including eyes, skin, teeth and gastrointestinal and respiratory tracts [5, 12, 14, 40-42]. Researchers have also demonstrated 3-D OCT imaging by 2-D lateral scanning of tissue [23, 42, 43]. 2.1.2 Key Imaging Parameters The performance of an OCT system is mainly determined by its axial and transverse imaging resolutions, dynamic range and also by its imaging speed. OCT achieves a very high axial resolution because the axial and lateral resolutions are independent of each other, unlike the case with conventional or confocal microscopy. In OCT the axial resolution is determined by the coherence length of the broadband light source. For a light source with a Gaussian spectrum, the coherence length (l c ) in air

PAGE 37

18 is given by [8]: 2202ln(2)0.44cl 0 (2-1) where, and are the center wavelength and the full-width at half-maximum (FWHM) spectral bandwidth of the light source, respectively. In order to obtain high axial resolution, a short temporal coherence length is desired, therefore, a light source with a broad emission bandwidth, i.e., a broadband light source (BBS) should be used. The BBS should operate in a spectral range that allows adequate penetration of light into tissue. Researchers have determined that BBSs that emit infrared light with a center wavelength between 1200 nm to 1800 nm achieve the deepest penetration in most tissues [44]. Another requirement for the BBS is that the irradiance of the emitted light should be high enough to provide a wide dynamic range. A wide dynamic range provides high detection sensitivity by enabling OCT imaging of weakly backscattering microstructures present deep inside the tissue. As shown in Figure 2-5, the lateral resolution of an OCT system is determined by the spot-size of the focused optical beam on the tissue. The diameter of the focused spot-size of a Gaussian beam is given by: 00421fx D NA (2-2) where, D is the diameter of the beam, NA is its numerical aperture, and f is the focal length of the lens. The lateral resolution of an OCT system is also affected by the depth of focus of the optical beam. The depth of focus, also known as the confocal parameter or the Rayleigh range, of an optical beam is the longitudinal distance within which the optical beam is considered

PAGE 38

19 Figure 2-5: Gaussian beam optics. to be in focus. The depth of focus (DOF) of a Gaussian beam is given by: 222822fDOFxD 1NA (2-3) This implies that a smaller spot-size will increase the lateral resolution but at the cost of reduced depth of focus. Since OCT imaging can penetrate tissue depths up to a few millimeters [36, 41], a depth of focus of a few millimeters is required. To overcome this depth of focus limitation, many researchers limit the numerical aperture of the scanning optics to obtain a DOF of approximately 1 mm [8]. At a center wavelength of 1300 nm, a lateral spot size of about 29 m is achieved with a DOF of 1 mm. In order to improve lateral resolution without sacrificing depth of focus, some researchers have developed novel methods to scan the reference path-length and position of the focused sample optical-beam simultaneously [45, 46]. These methods that increase the lateral resolution without reducing the DOF will be discussed in Section 2.2. The focal length of the scanning optics is of the same order of magnitude as the working distance between the OCT scanner and the tissue sample, which is typically a few millimeters. Therefore, by plugging in a focal length of a few millimeters in Equation (2-2), one can see that in order to obtain a lateral resolution better than 20 m, the diameter of the optical beam on the scanning optics should be large (> 1 mm). This is

PAGE 39

20 an important requirement to consider while designing optical scanners for endoscopic OCT imaging. 2.1.3 Internal Organ OCT Imaging It is challenging to realize in vivo imaging of internal organs due to the size limitations of conventional OCT systems. Fiber optic endoscopes specifically designed for endoscopic OCT imaging are required for imaging internal organs. Some key factors that have to be considered for in vivo OCT imaging are endoscope sizes and imaging speed. Therefore for internal organ applications, miniature OCT imaging endoscopic probes with diameters of a few millimeters must be developed. Also the transverse scanning mechanism should be fast enough to provide real-time images. Figure 2-6: Existing endoscopic OCT probe designs.(a) Radial imaging probe design. Adapted from Li et al. [47]. (b) Forward imaging probe design. Adapted from Boppart et al. [16]. Various methodologies have been presented to transversely scan the optical beam across the tissue surface to obtain 2-D OCT images. Many researchers have demonstrated radial OCT scans by rotating a hollow cable that carries a single-mode fiber and a microprism [12, 13, 47, 48]. Bouma and Tearney used a galvanometer to linearly translate a optical fiber above the surface of in vivo tissue [49]. Sergeev et al. used a galvanometric plate to swing the distal fiber tip to perform transverse scanning [14]. Piezoelectric Cantilev e r Cleaved Fib e r Tip Gear Prism Station a ry F i ber DC Motor ( a ) ( b )

PAGE 40

21 Bopart et al. [16, 50] and Li et al. [16, 50] also demonstrated transverse scanning by swinging a cantilevered fiber tip, but by using piezoelectric methods. A schematic drawing of these types of endoscopes is shown in Figure 2-6. These existing methods are complex and use bulky components, thereby miniaturization of the endoscopic OCT system becomes difficult. Furthermore, rotating a fiber is slow and it also introduces complexities due to non-linear optical coupling. Therefore, it may be advantageous to replace these existing scanning mechanisms with a MEMS-based scanning solution to perform OCT imaging. This MEMS-based solution may also result in smaller probe sizes with faster scanning speeds, and at potentially lower cost. The MEMS-based device should meet the following requirements for use in OCT probes: Large scanning angle: A large scan angle should be provided by the device in order to image large tissue areas. Large scanning angle combined with fast scanning speed will also reduce the OCT imaging time. Fast scanning speed: A fast scanning speed is desired to enable real-time OCT imaging, as well as to reduce the time required to conduct the endoscopic procedure. Low operating voltage: Low voltage operation of the MEMS scanner is essential for electrical safety of the patient during internal organ imaging. High resolution: The scanner should be able to provide lateral resolutions (spot sizes) better than a few tens of microns at working distances of a few millimeters. Small size: Endoscope diameters should be smaller than a few millimeters, therefore MEMS devices should be small enough to fit inside. However, there exists a

PAGE 41

22 trade-off between the small device size and high resolution requirements since a larger scanner size is needed for high spatial resolution. The small size, fast speed and low power consumption of MEMS mirrors makes them an ideal choice for use in endoscopic OCT imaging probes. The micromirror must be large (~1 mm) and optically-flat to maintain high light coupling efficiency and spatial resolution, and should also have large angle of rotation to meet the scanning range requirements. In fact, OCT endoscopes using micromirrors actuated electrostatically or electrothermally, have already been reported in literature [24, 51, 52]. Using these systems two dimensional, high resolution, cross-sectional images were obtained. As shown in Figure 2-1, 3-D OCT imaging requires 2-D transverse scanning. However, almost all existing OCT systems have only 1-D transverse scanning. In those cases, 3-D imaging is typically obtained by physically pulling the entire imaging probe. Since endoscopic catheters are flexible, the pull-out length may not be exactly the same as the physical displacement of the imaging probe. Therefore, there also exists a requirement for miniature scanners that can transversely scan the tissue surface in 2-D, thereby enabling endoscopic 3-D OCT imaging. 2.1.4 MEMS-based OCT As discussed earlier in this section, the key to making compact OCT probes is to miniaturize the scanning mirrors. In fact, there are a few groups who are working on MEMS micromirror based OCT. Pan et al. assembled the first MEMS micromirror based endoscopic OCT system in 2001 [53]. Zara et al. also reported an endoscopic OCT system based on MEMS mirrors in 2002 [24], in which the MEMS mirror has large deflection angle but requires elaborate assembly.

PAGE 42

23 Xie et al. reported a MEMS-based endoscopic OCT system that used a single-axis, single-crystal-silicon, electrothermal micromirror to scan the light beam onto the tissue [51]. Figure 2-7 shows a schematic of the MEMS-based OCT setup reported by Xie et al. [51], where the scanning micromirror along with the focusing optics was packaged inside a 5-mm diameter endoscope. The collimated light in the sample arm of the Michelsons interferometer is reflected off a beam steering 1-D micromirror and focused into the tissue. The same mirror collects the back-scattered light from the tissue, and the tissue microstructure is determined by low coherence interferometry. Two-dimensional (2-D) (i.e., x-z) cross-sectional images are obtained by combining the transverse scanning of the 1-D micromirror in the x-direction with the axial scanning of the reference mirror in the z-direction, as shown in Figure 2-7. This MEMS-based system acquired OCT images at a rate of 5 frames per second, thereby demonstrating the potential for real-time clinical diagnosis of cancers. Since the micromirror is packaged inside a flexible endoscope, no mechanical movement of the endoscope is necessary for OCT imaging. Other researchers have also used MEMS-based solutions to address the transverse scanning requirements of endoscopic OCT systems. Qi et al. used a MEMS deformable mirror to tune the focus of the OCT objective lens [54]. Tran et al. demonstrated an endoscopic OCT catheter using a MEMS micromotor to rotate a prism [25], while Herz et al. used a MEMS micromotor to rotate a mirror [26]. However, all these efforts are focused on only transverse scanning in one direction. 3-D OCT images can be obtained by using a MEMS scanner that can transversely scan the tissue in the x-y plane, i.e., in two dimensions. Yeow et al. demonstrated 3-D OCT imaging by using a 2-D electrostatic

PAGE 43

24 Axial Scanning Broadband Light Source Figure 2-7: MEMS-based endoscopic OCT system. MM: Micromirror. CM: Collimator. micromirror that scanned angles smaller than .5 in both transverse directions [21]. Fan et al. packaged a 2-D electrostatic micromirror inside a 5-mm diameter endoscope for in vivo OCT imaging [22]. McCormick et al. also demonstrated 3-D endoscopic OCT imaging using a 2-D scanning MEMS micromirror that used electrostatic actuation [23]. However, the high voltages required by electrostatic actuation for the abovementioned scanners may be a concern due to electrical safety issues during internal organ imaging. MEMS mirrors will be discussed in detail in the next chapter. 2.2 Optical Coherence Microscopy As discussed in the previous section, the lateral resolution of an OCT system was restricted to a few tens of microns due to the millimeter-range depth-of-focus requirement for the scanning optics. This low lateral imaging resolution was due to the use of focusing optics with relatively low numerical apertures (NA). Researchers have demonstrated Photodetector 50:50 Signal Processing Refer ence Mirror Tissue Sam p le Fiber Endoscope CM MM x z y

PAGE 44

25 various OCT system architectures that use high NA optics to obtain high-lateral resolution images without sacrificing the required depth of focus [10, 45, 46, 54, 55]. Figure 2-8 illustrates some of these different scanning approaches. Drexler et al. demonstrated a bench-top OCT system that used an x-z scanning stage to traverse the tissue sample in the transverse and axial directions to perform OCT imaging with resolutions better than 3 m [10]. The depth of focus of their imaging optics was less than 100 m, therefore nine separate OCT images were fused together to form a millimeter-deep ultrahigh-resolution tomogram. Other researchers have developed novel methods to scan the reference path-length and the position of the focused sample optical-beam simultaneously, thereby improving lateral resolution without sacrificing the depth of focus [45, 46]. Schmitt et al. placed the reference scanning mirror and the focusing optics on the same movable stage, which allowed the use of lenses with high NA [45]. Since these modified OCT systems use high NA optics to obtain OCT images with high lateral resolutions (< 10 m), they are also referred to as optical coherence microscopes. Moving Stage Tissue Tissue Optical Fiber Fixed Lenses Combined sample & reference beams (a) (b) Figure 2-8: High NA scanning approaches for OCT. (a) Reference mirror and focusing optics placed on a movable stage to achieve dynamic focusing. Adapted from Schmitt et al. [45]. (b) Lateral and axial scanning achieved by displacement of the fiber tip. Adapted from Schmitt [8].

PAGE 45

26 2.2.1 Bench-Top OCM Optical coherence microscopy (OCM) is an extension of OCT technology and it allows for ultrahigh-resolution cross-sectional imaging of highly-scattering tissue by combining the imaging capabilities of OCT and high NA confocal microscopy. In OCM, the high imaging resolution in the axial direction is provided by low-coherence interferometry, while the micron imaging resolution in the lateral direction is provided by confocal microscopy. Confocal microscopy is an optical technique that is used for imaging thin optical sections of relatively transparent tissue with very high resolution. Researchers have demonstrated bench-top OCM systems that obtained imaging resolutions better than 10 m in both, the axial and transverse dimensions [56-58]. A schematic of an OCM system is shown in Figure 2-9. As seen in the figure, the OCM system architecture is similar to that of OCT systems, the only difference being the imaging methodology of the tissue scanning optics. Figure 2-10 shows the difference Axial Scanning Figure 2-9: MEMS-based endoscopic OCM schematic. Broadband Light Source Photodetector 50:50 Signal Processing Endoscope Tissue sam p le Micro l ens Scanner Reference Mirro r Collim ating Lens

PAGE 46

27 OCM OCT Low NA Figure 2-10: Difference between OCM and OCT. between OCM and OCT imaging methodologies. OCT relies on large depth of focus to obtain imaging depth at the price of low transverse resolution, while OCM uses higher NA lenses (NA>0.2) to achieve high transverse resolution but with smaller imaging depth. The OCM imaging depth can be extended by using a moving lens or stage, as mentioned above. Figure 2-11 shows the tissue image acquired using an OCM system, demonstrating that cellular imaging is possible using this technology. Therefore, OCM technology is very promising for the early detection of cancer. 2.2.2 MEMS-based OCM Although bench-top OCM systems allow for ultrahigh lateral and axial resolutions, endoscopic OCM probes with ultrahigh imaging resolutions are needed for in vivo detection of precancers in internal organs. Since the methods shown in Figure 2-8 require the use of mechanical stages with stepper motors, they are bulky and slow, and therefore cannot be used for ultrahigh-resolution endoscopic OCM imaging. A MEMS-based dynamic focusing micromirror has been proposed by Qi et al. that could potentially be used for endoscopic OCM [54]. They demonstrated a MEMS deformable mirror to focus a high NA objective lens at different depths inside biological tissue. However their micromirror requires a high ac voltage of 400 V (peak to peak) to produce a 1.25-mm focus scan range. Ding et al. used an axicon lens to obtain OCT images with a lateral High NA Imaging Depth Lateral Resolu t ion

PAGE 47

28 Figure 2-11: 1-m axial resolution by 3-m lateral resolution tomogram of a tadpole. Reprinted, with permission, from Drexler et al. [10]. resolution of 10 m with a depth of focus of 6 mm [59]. However, the axicon lens significantly reduced the optical signal intensity, which will result in a lower sensitivity for the OCT system. Kwon et al. demonstrated a microlens scanner for micro-confocal imaging that used electrostatic vertical-comb-drives; however its vertical scan range is restricted to less than 55 m and therefore unsuitable for OCM imaging [60-62]. In order to obtain high lateral resolution without compromising the axial scanning range and small size requirement of endoscopic OCM systems, a MEMS microlens can be used to scan along the optical axis. This MEMS scanner should be able to axially displace the focal plane of the scanning microlens by up to a few millimeters. Other requirements are that the scanner should be small enough to fit inside millimeter-sized

PAGE 48

29 endoscopes, and it should also use low voltage for actuation. A schematic of a MEMS-based OCM system is shown in Figure 2-9. The overall system design is similar to the MEMS-based OCT system architecture presented in Section 2.1.4. Infrared light in the sample arm of the Michelson interferometer is first collimated by a GRIN lens placed inside the hollow endoscope. Then the collimated light is focused by a high NA polymer microlens into the tissue, as shown in Figure 2-9. In this OCM scanner design, the polymer microlens is attached to a MEMS microactuator, which enables vertical displacement of the microlens. A vertical displacement of the microlens results in vertical displacement of the focused beam-spot inside the tissue. By changing the vertical position of the focused beam-spot, it is possible to scan axially into the tissue. Since a high NA microlens is used, this system will provide OCM images with high lateral resolution. Unlike the axicon lens used by Ding et al. [59], the smaller depth of focus of the polymer microlens will maintain a strong optical signal intensity, which will help to improve the overall OCT system sensitivity. 2.3 Non Linear Optical Imaging The imaging methods described in the preceding sections were linear optical imaging methods, since the magnitude of the observed signals from tissue changes linearly with incident light intensity. The well-known optical phenomena of reflection, refraction, and diffraction occur in the linear domain since the intensity of reflected, refracted or diffracted light changes linearly with the magnitude of the incident light. Other naturally occurring linear events are the absorption of light and photochemical reactions such as in the photosynthesis process in plants and bacteria [63]. This section introduces another class of imaging modalities that use the nonlinear optical properties of tissue for high-resolution imaging. The nonlinear optical imaging

PAGE 49

30 modalities presented in the following sections have the potential to detect and diagnose in vivo cancers at a very early stage. 2.3.1 Two-Photon Excitation Fluorescence Imaging Two-photon excitation fluorescence (TPEF) microscopy is a nonlinear optical imaging technique which can provide high resolution imaging at the cellular level. TPEF microscopy is a new form of scanning far-field fluorescence optical microscopy. Far-field fluorescence optical microscopy is typically a one-photon excitation fluorescence based microscopy technique, in which illumination is focused into a diffraction-limited spot scanned on the tissue specimen, thereby confining the excitation focal region. The diagram in Figure 2-12(a) depicts the phenomena of fluorescence when a single photon is absorbed by a fluorescent molecule, and so the molecule is excited to a higher energy state. The excited molecule now returns to its ground energy state by emission of a fluorescent photon at a characteristic wavelength. As seen in Figure 2-12(a), the energy of the fluorescing photon is less than the energy of the excitation photon, therefore the fluorescence emission is shifted towards a longer wavelength than that used for excitation. This means that in order to obtain fluorescence from samples that exhibit fluorescence in the blue-green wavelengths (~ 450 nm), the sample would have to be excited at a lower ultraviolet (UV) wavelength of about 350 nm. However, exciting tissue samples at UV or blue wavelengths is undesirable due to problems due to photobleaching and phototoxicity [64]. Another problem with one-photon excitation is that the entire thickness of the sample within the hourglass-shaped region of the focused light spot is excited, which results in poor optical sectioning ability. This is shown in Figure 2-13(a).

PAGE 50

31 Figure 2-12: Energy band diagrams illustrating (a) one photon and, (b) two photon excitation fluorescence phenomena. Two-photon excitation fluorescence (TPEF) microscopy provides an inherent optical sectioning ability to improve axial imaging resolution, and it is also less affected by the effects of photobleaching and phototoxicity. For two-photon excitation to occur, the fluorescent molecule should simultaneously absorb two photons of a longer wavelength to reach its excited state. As shown in Figure 2-12(b), two photons with lower energy are simultaneously absorbed to provide the energy needed to prime the fluorescence process. The fluorescent molecule now emits a single photon of fluorescence as if it were excited by a single higher energy photon. This phenomenon of TPEF depends on two photons interacting simultaneously with the molecule, and it results in a quadratic dependence on the intensity of incident excitation light. In contrast, conventional fluorescence is linearly dependent with the excitation light intensity. The reason that TPEF is referred to as a nonlinear imaging method is due to the fact that the rate of occurrence depends nonlinearly on the incident light intensity. Since light intensity is the highest at the focal spot, the largest probability of observing TPEF is at Thermal relaxation Thermal relaxation h f h Fluorescence h f h h Fluorescence (a) (b)

PAGE 51

32 this location. Axially away from the focal plane, the TPEF probability drops off rapidly with decreasing light intensity. As seen in Figure 2-13(b), no significant amount of fluorescence is emitted from regions away from the focal plane, and this demonstrates TPEFs inherent optical sectioning ability. Therefore, TPEF microscopy can image tissue with very high resolution in all three dimensions. TPEF theory: The probability, p that a molecule absorbs two photons simultaneously to reach its excited state has been computed as [64]: 2Ip (2.4) where, is a proportionality factor, and I is the intensity of the incident laser beam. The timescale for the keyword simultaneous for TPEF is the same timescale of molecular Figure 2-13: Optical sectioning ability of TPEF imaging. (a) Single-photon excitation of fluorescein by 488 nm light. (b) Two-photon excitation using 960 nm light. Reprinted by permission from Macmillan Publishers Ltd: Nature Biotechnology, Zipfel et al. [65], copyright 2003.

PAGE 52

33 energy fluctuations at photon energy scale, and using Heisenbergs uncertainty principle this has been computed to mean a temporal window of 10 -16 s or 0.1 fs [64]. The emitted fluorescence intensity, I f (t) from the molecule is proportional to the molecular cross-section and also to the square of the incident intensity I(t) 2 [64]: 22222)()(.)(cNAtPtItIf (2.5) where, P(t) is the optical power of the incident light, c is the speed of light, is the Planck quantum of action, is the two-photon absorption cross-section, and NA is the numerical aperture of the focusing objective. The time averaged fluorescence intensity emitted from a fluorophore when excited with a pulsed laser beam with pulse width p repetition rate f p and average power P 0 can be computed from Equation (2.5) as: 2220,2)(cNAfPtIpppf (2.6) The number of photons absorbed by a fluorophore per pulse is given by [66]: 222202cNAfPnppa (2.7) Equation (2.7) does not account for saturation effects, and was computed with the paraxial approximation assumption. 2.3.2 Second Harmonic Generation Imaging Second harmonic generation (SHG) is also a nonlinear optical process, similar to TPEF, and it can be used for high resolution imaging of tissue microstructure. SHG converts an input optical wave into an output optical wave of twice the input frequency,

PAGE 53

34 therefore this phenomenon is also commonly known as frequency doubling. This is the same process used to produce green light at a wavelength of 532 nm from a Nd-YAG laser operating at 1.06 m [63]. Similar to TPEF, the probability of SHG is proportional to the square of the intensity of the incident excitation light. Thus, SHG imaging has the same intrinsic optical sectional ability as TPEF imaging. However, unlike TPEF, SHG is confined to imaging only highly polarizable materials that lack a center of symmetry. SHG imaging can be used for bioimaging purposes since biological materials can be highly polarizable and the cellular membranes contain SHG-active constituents which are asymmetrically distributed [67]. The second-harmonic light emitted from the noncentrosymmetric, highly polarizable material is exactly half the wavelength of the incident excitation light. Therefore, the SHG process within the nonlinear optical material converts two incident photons into one exiting photon at exactly half the wavelength (or twice the energy). As described in Section 2.3.1, in TPEF some of the incident energy of the photon is lost during thermal relaxation of the excited state (Figure 2-12(b)), but in the case of SHG, there is no excited state and so SHG is energy conserving and it also preserves the coherence of the incident laser light. Since SHG does not involve excitation of molecules, it should not suffer from photobleaching or phototoxicity effects (which limit the usefulness of fluorescence microscopy). Another advantage of SHG is that it uses excitation wavelengths in the near-infrared range which allow for excellent depth penetration, thereby permitting imaging of thick tissue samples [68]. SHG theory: The nonlinear polarization for a material can be expressed as [68]: ...)3()2()1(EEEEEEP (2-8)

PAGE 54

35 where, P is the induced polarization vector, E represents the electric field vector, (i) is the ith order nonlinear susceptibility tensor, and represents a combined tensor product and integral over frequencies. The first term in the series, (1) describes normal absorption and reflection of light. The second term describes the sum and difference frequency generation; and thereby also describes SHG. The third term describes two-photon absorption (the probability of which is linearly proportional to the imaginary part of the third-order nonlinear susceptibility tensor), as well as third harmonic generation and coherent anti-Stokes Raman scattering. The portion of the polarization that contributes to SHG is: EEP)2()2( (2-9) The intensity of the SHG signals, I SHG emitted from such materials will scale as follows [68]: 2)2(20PISHG (2-10) where P 0 and are the laser pulse energy and pulse width, respectively. This term shows the nonlinear dependence of the SHG emission intensity to the incident light intensity. 2.3.3 Nonlinear Optical Imaging System Design The schematic of a nonlinear optical imaging system is shown in Figure 2-14. The light source generally consists of a pump laser and a Ti:Sapphire laser which generates ~100 femtosecond long laser pulses at around 1W power at a repetition rate of 80 MHz. A laser pulse train output of such a laser system is depicted in Figure 2-15. This laser light is focused by a microscope objective lens and scanned laterally on the tissue sample using an XY beam scanner. In a fluorescence microscopy system, the dichroic mirror is used

PAGE 55

36 Figure 2-14: Schematic of a nonlinear microscope. Figure 2-15: (a) Pulse train from a mode-locked Ti:Sapphire laser at 80 MHz. (b) The laser pulses typically have a FWHM duration of 100 fs, and (c) a spectral FWHM bandwidth of ~ 10 nm. Adapted from Zipfel et al. [65]. for separating the excitation and emission light beams. This dichroic mirror reflects light with wavelengths longer than 800 nm, while it transmits light with shorter wavelengths. The emission signal from the tissue specimen is collected by the same focusing optics, passes through the dichroic mirror, and is detected by a photomultiplier tube (PMT) as 800 820 780 0 100 200 -100 -200 Tim e (ns) Tim e (fs) Wavelength (nm) Laser pulse train ~12 ns (a) (b) (c) Beam Scanner Ti:Sapphire Laser PMT 40x Dichroic Mirro r Filte r Pum p Lase r Objectiv e Tissue

PAGE 56

37 shown in Figure 2-14. A bandpass filter is inserted in the light path before the PMT to help differentiate between the TPEF and SHG signals. Figure 2-16 illustrates the SHG and TPEF emissions when excited with near-infrared light. SHG TPEF Illumination 400 520 800 Wavelength (nm) Figure 2-16: Two-photon fluorescence and SHG signals emitted by a sample excited by 800 nm light. Researchers have used nonlinear optical microscopes to image and identify cancerous tissue with very high resolution as shown in Figure 2-17. The hamster cheek pouch biopsies were imaged using a bench-top system with lateral and axial imaging resolutions of 0.35 and 1.25 m, respectively [69]. 2.3.4 Endoscopic Nonlinear Optical Imaging As stated in the previous section, researchers have successfully demonstrated high resolution imaging of tissue using bench-top nonlinear microscopes [69, Zipfel, 2003 #241]. However in order to demonstrate in vivo imaging, lateral beam scanning endoscopes are required. Jung and Schnitzer developed a free-space multiphoton endoscope using GRIN lenses [70]; however, the lack of a flexible optical fiber prevents its use for endoscopic in vivo imaging. Helmchen et al. used a piezoelectric bending element to transversely scan a cantilevered fiber tip [71], but the 1.3-cm diameter, 7.5-cm long endoscope is too bulky

PAGE 57

38 Figure 2-17: Two-photon imaging of the hamster cheek pouch tissue at an excitation wavelength of 780 nm. (a) Normal tissue. Precancerous tissue: (b) Moderate Dysplasia. (c) Carcinoma in situ. Cancerous tissue: (d) nonpapillary, and (e) papillary squamous cell carcinoma. The top image is at the surface of the tissue, and each subsequent image in the montage represents a 10-m axial step. Scale bar represents 30 m. Reprinted, with permission, from Skala et al. [69]. to be used for internal organ imaging. Flusberg et al. [72] also used a piezoelectric actuator, along with a MEMS micromotor, to create a multiphoton microscope designed for imaging peripheral organs of small animals. Gobel et al. demonstrated in vivo TPEF imaging using a fiber bundle and GRIN lens, but averaging and the use of a Gaussian blur filter were needed to improve image quality [73]. Bird and Gu developed a radially scanning endoscope, similar to the OCT endoscope design illustrated in Figure 2-6(a),

PAGE 58

39 that required an external motor to physically rotate the endoscope [74]. Myaing et al. also adapted a piezoelectric OCT endoscope for in vivo, endoscopic TPEF imaging [75]. 2.3.5 MEMS-based Endoscopic Nonlinear Optical Imaging It is clear from the above discussion that the design requirements for TPEF endoscopes are almost the same as that for OCT endoscopes, which are listed in Section 2.1.3. The main difference being that the nonlinear imaging probes should be able to provide spot-sizes in the micron range, and should also be able to laterally scan higher power laser beams. MEMS-based scanners, packaged inside endoscopes with high numerical aperture optics, are very suitable for endoscopic nonlinear optical imaging as they can provide large scan ranges with high imaging resolution. L. Fu et al. used the micromirrors developed by this research effort to demonstrate the first-ever MEMS-based nonlinear optical endoscopy system [76, 77]. Recently, Piyawattanametha et al. used an electrostatic micromirror to transversely scan the proximal end of a free-space, GRIN-lens endoscope [78]. The high voltage requirement (up to 160 V) of this MEMS scanner is a safety concern for in vivo internal organ imaging. Ideally, the MEMS-based endoscopes should be capable of providing large scan range and high imaging resolution at a fast scan speed, but at low operating voltage. MEMS micromirrors are discussed in the next chapter, while the endoscopic TPEF and SHG imaging results obtained by this research effort are reported in Section 4.4.

PAGE 59

CHAPTER 3 ELECTROTHERMAL MICROMIRRORS AND ENDOSCOPIC OCT IMAGING As mentioned in the previous chapter, system miniaturization is the key for OCT to become practical for clinical use in imaging visceral organs. We can see from Figure 2-7 that the miniaturization of OCT imaging systems is largely determined by the axial scanning and transverse scanning mirrors. MEMS technology leverages integrated circuits (IC) technology to manufacture micro-scale devices and systems [79, 80], and thus is the natural choice to make scanning microdevices, i.e., MEMS micromirrors. This chapter introduces different types of MEMS micromirrors, and justifies the selection of electrothermal actuation as the preferred choice of micromirror actuation for internal organ OCT imaging probes. The basic operating principles, fabrication process and characterization results of 1-D and 2-D electrothermal micromirrors are presented. Finally, OCT imaging using these micromirrors packaged inside endoscopic probes is also demonstrated. 3.1 Scanning Micromirrors Rotational scanning micromirrors are widely used for a variety of applications, such as optical displays [81, 82], biomedical imaging [20, 51, 83], barcode scanning [84, 85], optical switching [18, 86-88], and laser beam steering [85, 89]. There are numerous commercially available MEMS scanning micromirrors ranging from Texas Instruments DMDs (Digital Micromirror Devices) [17] to Lucent Technologies optical switch [18]. Most of these commercially-available micromirrors are surface micromachined and their size is limited to about 0.1 mm due to curling that is caused by residual stresses in thin40

PAGE 60

41 film structures. For biomedical imaging applications, relatively large mirrors (>0.5 mm) are required to maintain high spatial resolution. Therefore, bulk-micromachining processes are often used to make relatively large, flat single-crystal silicon (SCS) micromirrors. These micromirrors can be actuated using electromagnetic, piezoelectric, electrostatic or electrothermal techniques. Fast scanning speeds and low power consumption make electrostatically-actuated micromirrors the most popular amongst all scanning mirrors. Electrostatic micromirrors can be further subdivided into two categories based on electrode placement. The first type of mirror design uses the electrostatic force created by parallel-plate electrodes placed underneath the mirror to generate rotation. Micromirrors using this approach have demonstrated rotation angles of at 142 V [18], .5 [90], and at 70 V [91]. Since most of these devices are fabricated using surface micromachining techniques, there is a trade-off between mirror-plate size and the maximum allowed rotation angle due to the small gap size between the electrodes. Other researchers have used bulk micromachining methods which achieve larger electrode gaps thereby permitting larger mirror sizes; but this significantly increases the actuation voltage. Parallel-plate actuation using bulk micromachining have yielded 2-D mirrors that rotate at 160-170 V [92], at 200 V [93], and at 40V [94]. Since the tradeoff between the mirror size and rotation angle limits the applications of parallel-plate electrostatic actuation to small micromirrors, a second category of electrostatic mirrors have been developed that use electrostatic comb fingers to rotate the mirror plate. A number of vertical comb drive (VCD) designs based on single-crystal silicon (SCS) have been reported for achieving larger rotation angles with large mirror

PAGE 61

42 sizes [95-101]. For instance, Conant et al. reported a fast-scanning VCD micromirror by using silicon-on-insulator (SOI) wafers [96]. Xie et al. demonstrated a curled-hinge VCD micromirror that rotated .7 at 18V [97]. Patterson et al. reported a VCD design in which photoresist re-flow was used to tilt comb fingers, but the device fabrication uniformity and yield may be concerns [98]. Krishnamoorthy et al. used SOI wafers to fabricate self-aligned VCD micromirrors [99]. Milanovi et al. used lateral comb drives to generate torsional rotation [100, 101]. Kim and Lin reported an electrostatic micromirror with a pre-tilted mirror using localized plastic deformation of silicon by Joule heating [95]. 2-D electrostatic mirrors using comb drives have also been reported to produce mechanical rotation angles of 5.5 at a resonance of 720 Hz and 16 V voltage [102], up to 11 at 100 V [103], 6.2 at 55 V [104], and 10 at 140 V [101]. Although the high resonant frequencies of electrostatic mirrors allow for high speed scanning, the scan area is limited by the small rotation angles. Also, the high voltages required for larger angular actuation is a deterring factor for their use in certain applications, such as in endoscopes for internal biomedical imaging. On the other hand, electrothermal actuation can generate large rotational displacements at low drive voltages. Electrothermally-actuated micromirrors use thin-film bimorph cantilevers (composed of materials with different coefficients of thermal expansion) that are attached to a mirror plate. Joule heating of these bimorph structures result in rotation of the mirror plate. Micromirrors based on the bending of bimorph or multimorph structures have been reported [51, 105-108]. Metals are often used as the top layer of bimorph structures due to their large thermal expansion coefficients and high reflectivity. The commonly used bottom layers include silicon dioxide [51, 105, 106,

PAGE 62

43 108] and silicon [107, 109-111]. Heating sources can be provided by polysilicon [51, 105], diffusion [107], or metal resistors [106, 108]. 2-D electrothermal mirrors have reported mirror rotation of ~15 at a resonant frequency of 1.3 kHz [84], and also rotation angles as large as 40 at 15 V [30]. There is also an interesting report in which a clamped-clamped polysilicon beam was used as the thermal actuator [112]. In this case, the buckling of the clamped-clamped beam due to thermally-induced stress is used for actuation, and the polysilicon beam itself functions as a thermal resistor. The disadvantages of thermal actuation include high power consumption, relatively slow speed and poor temperature stability. Even though electrothermal micromirrors generally consume more electrical power than others, they are the best suited choice for some applications that require large optical angles at low driving voltages. Electromagnetic micromirrors rotate due to the Lorentz force generated by the interaction of an external magnetic field with electric current flowing through a coil on the mirror plate. Electromagnetic micromirrors have been demonstrated using metallic coils [113-115] or magnetic materials such as Permalloy [116]. Although electromagnetic micromirrors can achieve large rotation angles of 10 [86], 15.7 [114], and 23 [87] at low actuation voltages, they are bulkier than other micromirrors since they require large external magnets. Therefore it is challenging to compactly package these electromagnetic micromirrors for applications with stringent size restrictions, such as endoscopic imaging. Piezoelectric actuation is another mechanism that can generate large forces and have low power consumption and high bandwidth. In piezoelectric mirrors, mirror rotation is brought about by the piezoelectric bending of thin-film PZT actuators/cantilevers on application of an electric voltage. Piezoelectrically actuated

PAGE 63

44 micromirrors with rotation angles of 2.3 at 4.5 V [117], 2.2 at 60 V [118], 3.5 at 40 V [119], and upto 5.5 at 16 V [88], have been reported. Even though some piezoelectric mirrors operate at low voltages, they are limited to the area they can scan. Other drawbacks of piezoelectric actuation include small displacements and charge leakage and hysteresis effects which often require a feedback control loop. As mentioned in Chapter 2, micromirrors specifically designed for use inside endoscopic probes for internal organ biomedical imaging must meet requirements of small size, fast scanning speed, large scan angles, and low operating voltage. Electrothermal actuation was chosen as the preferred actuation method since it meets all the above mentioned requirements. The following sections present the fundamentals of electrothermal actuation, and also the fabrication process used to fabricate 1-D and 2-D micromirrors that use electrothermal actuation to achieve large angular displacements at low driving voltages for endoscopic optical coherence tomographic imaging. 3.2 Electrothermal Actuation and Design All electrothermal micromirror designs described in this dissertation use the same basic design structure which is based on electrothermal bimorph actuation. A bimorph structure, illustrated in Figure 3-1, consists of two thin-film layers that have different coefficients of thermal expansion. A temperature change induces stress in the two layers due to the difference in their thermal expansion coefficients, thereby resulting in bending of the bimorph beam. This temperature change can be brought about by resistive Joule heating. Although any two materials with different coefficients of thermal expansion can be used to form bimorph structures, their mechanical properties and material compatibility

PAGE 64

45 Layer 2 Layer 1 Anchor Figure 3-1: Side view of a bimorph beam. must also be considered. Table 3-1 lists some materials that may be used in conjunction with others to form bimorph structures. Researchers have also reported bimorphs that were formed by a layer of metal and polyimide polymer [120, 121]. Table 3-1: Thermomechanical properties of some possible bimorph materials at room temperature. Material Coefficient of Thermal Expansion [10 -6 /K] Youngs Modulus [10 11 N/m 2 ] Specific Heat [10 3 J/kgK] Thermal Conductivity [W/mK] Density [10 3 kg/m 3 ] Layer 1 Si 2.6 1.62 0.691 170 2.42 SiO 2 0.4 0.74 0.84 1.1 2.66 Si 3 N 4 2.8 1.55 0.711 18.5 3.19 SiC 3.5 4.57 86.5 3.2 Poly-Si 2.3 1.60 0.754 2.33 Layer 2 Al 23.0 0.69 0.9 235 2.692 Au 14.3 0.8 0.129 318 19.4 Pt 8.9 1.47 0.133 73 21.5 Cu 16.7 1.2 0.387 401 8.95 Ni 12.8 2.1 0.444 91 9.04 Pb 28.7 0.16 0.128 35 11.48 Material properties obtained from [110], and those marked obtained from Memsnet.org

PAGE 65

46 All electrothermal MEMS devices presented in this dissertation are fabricated using a CMOS-MEMS process [122], therefore bimorph actuators are composed of silicon dioxide (SiO2) and aluminum (Al) thin-film layers which are provided by the CMOS process. As given in Table 3-1, the coefficients of thermal expansion for SiO2 and Al are 0.410 -6 /K and 2310 -6 /K, respectively. It is this large difference in the thermal expansion coefficients of the two materials that attributes to the large actuation range of the fabricated MEMS optical scanners. The basic structure of an electrothermal micromirror is shown in Figure 3-2. It consists of a mirror plate attached to the substrate by a bimorph beam actuator. The mirror plate is composed of an Al top layer which forms broadband, highly reflective surface of the mirror, a single-crystal silicon (SCS) membrane bottom layer which adds stiffness to the mirror to ensure surface flatness, and a SiO 2 layer in between the Al and SCS layers. The bimorph actuator is composed of a top aluminum layer, a bottom silicon dioxide layer, and within the SiO 2 layer is embedded an electrically-resistive polysilicon layer to provide Joule heating. A unique micromachining process which can be used to fabricate the bimorph structure shown in Figure 3-3 is detailed in the next section. 3.3 Microfabrication Process It is a widely known fact that thin-film deposition processes generate residual stress and stress gradients, which cause curling of the resultant thin-film microstructures. Micromirrors made up of thin-film layers are typically small in size, in order to reduce their optical quality degradation due to curling. Therefore, this thin-film curling limits the useful size of micromirrors to about 10-100 m. In order to increase mirror sizes without sacrificing the mirror flatness, single-crystal silicon (SCS) based mirrors are desirable. As

PAGE 66

47 (a) (b) Figure 3-2: Electrothermal micromirror basic structure. (a) Top view. (b) Cross-sectional side view. introduced in the previous section, the bimorph micromirrors presented here require thin-film bimorph structures for actuation and SCS structures for large size and flatness. The micromirrors were fabricated by a deep-reactive-ion-etch (DRIE) CMOS-MEMS process [122]. The basic idea of this process is to introduce an SCS layer underneath CMOS multi-layer structures in such a way that the mechanical properties are dominated by the SCS layer, electrical connections provided by the CMOS interconnect metal layers,

PAGE 67

48 Step 3: Deep Si Etch Step 1: Backside Etch metal-3 Figure 3-3: DRIE CMOS-MEMS fabrication process flow. heating source provided by polysilicon and high reflectivity by the top metal layer. This maskless post-CMOS micromachining process has also been successfully used to fabricate integrated MEMS accelerometers and gyroscopes [123, 124]. The process flow is shown in Figure 3-3, which is completely CMOS-compatible and involves only four dry-etch steps. The process starts with CMOS wafers or chips that are fabricated at virtually any CMOS foundry. As a demonstration of foundry-CMOS compatibility, the devices presented in this dissertation were fabricated using the Agilent 0.5-m or the AMI 0.5-m 3-metal CMOS processes available through the MOSIS foundry service [125]. The first step of the post-CMOS fabrication process is to perform a backside DRIE step to form a 30 m to 50 m-thick SCS membrane. This etch step is carried out by the Bosch process [126 1996] on a Surface Technology Systems (STS) inductively-coupled-plasma (ICP) etcher. The etching chemistry used is SF 6 /O 2 with the following parameters: 600 W coil power, 12 W platen power, 130 sccm SF 6 flow, 13 sccm O 2 flow, and 37 mT chamber pressure. This step controls the thickness of the microstructure and forms a cavity (~200 m deep) that allows the microstructure to move metal-1 metal-2 CMOS-region oxide SCS membrane poly-Si ~40 m Step 2: Oxide Etch Step 4: Si Undercut bimorph beam mirror ~5 m ~40 m

PAGE 68

49 freely in a wide range. The depth of the cavity is determined by the thickness of the CMOS chips (which is typically around 250 m). The second step is a frontside anisotropic oxide etch that uses the CMOS interconnect metal (i.e., aluminum) as an etching mask. This oxide RIE etch is performed in a Unaxis Shuttlelock ICP etcher with the following process conditions: 600 W coil power, 100 W platen power, 15 sccm SF 6 flow, 5 sccm Ar flow, and a chamber pressure of 5 mT. The oxide etch is followed by a deep silicon trench etch using the STS ICP etcher to release the microstructure. At the end of this step, a 30 m to 50 m thick SCS layer remains underneath the CMOS layer, resulting in a flat released microstructure. Finally, a brief isotropic silicon etch is performed to undercut the silicon from under the thin-film bimorph beams. Any beam with a half-width less than the silicon undercut will have no SCS layer underneath. This isotropic silicon etch is attained using the same STS ICP etcher but by reducing the platen power to 2 W. These undercut thin-film beams can be used to form electrically isolated SCS islands, purposely curled-up structures or z-compliant springs. In the micromirrors, these 2-m-thick thin-film beams form bimorph actuators with an embedded polysilicon heater. As the top aluminum layer is used as an etching mask, CMOS circuits under it will remain unaffected by the fabrication process. Thus, this maskless post-CMOS process is completely compatible with foundry CMOS processes, and CMOS circuits can be integrated with MEMS devices. When the mirror is released from the substrate during fabrication, the bimorph actuator is no longer constrained and will curl up. This bimorph curling occurs due to the residual tensile and compressive stresses present in the aluminum and silicon dioxide layers, respectively. As a result of the bimorph curvature,

PAGE 69

50 Aluminum Package Side-wall Aluminum/Oxide Bimorph Actuator Mirror Single-Crystal Silicon Figure 3-4: SEM of a fabricated 1-D micromirror with initial tilt angle. the attached mirror tilts upward and away from the substrate with an angle equal to the tangential angle at the end of the bimorph. The SEM of a fabricated 1-D micromirror, demonstrating the initial tilting of the mirror plate, is presented in Figure 3-4. 3.4 Bimorph Actuation and Theoretical Analysis Before looking into the actual electrothermal micromirrors, bimorph actuation theory and electrothermomechanical analysis are presented in this section. The electrothermal micromirror is actuated by applying an electrical current to the polysilicon resistor. The electrical power dissipated by the resistor as heat raises the temperature of the bimorph actuator. Since the top Al layer has a greater coefficient of thermal expansion than the bottom SiO 2 layer, the increase in temperature causes the top metal layer to expand more than the bottom SiO 2 layer. This in turn increases the radius of curvature of the bimorph actuator by bending the bimorph in the downward direction. Therefore, the tilt angle of the mirror decreases from its initial value. Side-view schematics illustrating the released micromirror structure and the electrothermal actuation mechanism are shown in Figure 3-5.

PAGE 70

51 (b) (a) Figure 3-5: Bimorph actuation mechanism. Side views of: (a) Initial position of mirror at zero bias. (b) Downward rotation of mirror plate on application of bias voltage to polysilicon resistor. The bimorph beam curls up after being released due to the tensile stress in the aluminum layer and compressive residual stress in the bottom silicon dioxide layer. Therefore, the radius of curvature of the bimorph beam is determined by both the initial curling and also due to the temperature change from the polysilicon heating, and is given by [105]: 0111Trrr (3-1) where r is the actual radius of curvature, r 0 is the initial radius of curvature and r T is the radius of curvature due to the temperature change. The minus sign is due to the fact that the initial curling of the bimorph is caused by residual stresses due to cooling from high processing temperature to room temperature, while the thermally induced curvature is caused by thermal heating. By ignoring the thin polysilicon layer, r T is readily derived as [127]:

PAGE 71

52 3226()()1446AloxAloxT 3 A lAloxoxAloxAloxoxoxAlAlttTrEtEtttttEtEt (3-2) where T is the temperature change of the bimorph beams, t a a and E a are the thickness, thermal coefficient of expansion, and Young's modulus of the metal (a=Al) and oxide (a=ox) layers, respectively. Equation (3-2) can be rewritten as 1rTTTr (3-3) where TAl ox is the difference in the coefficients of thermal expansion of Al and SiO 2 and r is the curvature coefficient of the bimorph beam as is given by: 3226()446Aloxr 3 A lAloxoxAloxAloxoxoxAlAlttEtEtttttEtEt (3-4) As shown in Figure 3-5, the tangential angle at the tip of a curled beam is equal to the arc angle. Using simple geometry, we get bTTLr (3-5) or TrbTL T (3-6) where L b is the length of the bimorph beam. Equation (3-6) relates the angular change of the bimorph actuator to its temperature increase. Using Equations (3-1) and (3-3), we can express both the initial curvature and thermal-induced curvature in terms of the curvature coefficient as: 01rTTr (3-7)

PAGE 72

53 where 0 is the linear strain difference caused by residual stress. Therefore, the actual tilt angle of the beam tip relative to the substrate plane is given by 00TrbTTL T0 (3-8) where 0rbL is the initial tilt angle. So, increasing the curvature coefficient will simultaneously increase the actuation angle and the initial tilt angle. Sometimes large initial tilt angle may be undesired, in which case a compromise has to be made. Equation (3-6) gives the relationship between the actuation angle and the temperature change which is uniform along the entire bimorph beam. However, in most cases, the beam temperature distribution is not uniform, so the radius of curvature varies along the beam. Therefore, the tilt angle at the tip of the bimorph is an accumulation of the gradual curvature changes, i.e., 0000011()bbbLLLrTTdxdxTxdxrxrr (3-9) or 001bLTrbTrbbLTxdxLL TT (3-10) where 01bLbTTL xdx is the average beam temperature difference above the substrate (or ambient) temperature. Thus, the actuation angle is linearly proportional to the average temperature of the bimorph beam. This is valid as long as the increased temperature does not change the material and mechanical properties of the bimorph layers. This analysis assumes that the width of the bimorph layers is equal for both layers, the materials are isotropic and continuously distributed, and that the radius of curvature is constant along the bimorph beam.

PAGE 73

54 3.5 Electrothermal Micromirrors 3.5.1 One-Dimensional Electrothermal Micromirror An electrothermal micromirror consists of a thin-film bimorph structure and a bulk-silicon mirror plate. Instead of using a continuous bimorph mesh that was used in the previous designs [19, 28], the bimorph actuator was split into an array of bimorph beams, as shown in Figure 3.1. Since each beam has a relatively small width, the silicon undercut of the structure will remove all of the silicon underneath the beams, leaving a majority of the silicon underneath the mirror. This design was created to further improve the micromirror scanning performance by reducing the overall stress of the bimorph upon actuation. The buckling phenomenon observed by Xie et al. [19] is not present in this device. Another difference in the new design is that thermal isolation regions were added to isolate the bimorph beam array from the substrate and mirror plate regions. The thermal isolation regions are useful for two things. First, the thermal isolation region between the bimorph array and the substrate increases the average temperature of the bimorph array for a given bias, yielding a greater angular response of the mirror. Second, the thermal isolation layer between the bimorph and the mirror plate lowers the heat flux between the two regions upon actuation, resulting in a faster thermal response time of the bimorph. A lumped element model of this micromirror has been developed by Todd and Xie, and interested readers may refer to [128]. A single-axis micromirror with a bimorph actuator using this beam design has been fabricated [27]. A scanning electron micrograph (SEM) of a released device is shown in Figure 3-6. The micromirror is 1mm by 1mm in size, coated with aluminum, and thermally actuated by an integrated polysilicon heater.

PAGE 74

55 Bimorph Actuator Mirror Figure 3-6: SEM of 1-D micromirror. Inset: Close-up of the bimorph actuator beams. The bimorph beam array consists of 64 beams. The embedded poly-Si resistors of adjacent beams are connected in parallel giving a total of 32 resistors embedded in the bimorph beam array. The 32 resistors are connected in series yielding a total resistance of 1.15 k for the bimorph beam array. Two voltage contacts are present on the furthest right and left resistors of the beam array. Figure 3-7 shows the measured rotation angles at different currents and also the current dependence of the polysilicon resistor. A rotation angle of 31 is achieved at 9 mA or 18 V. The response curve is smooth over the whole scanning range. Thus, this mirror design also eliminates the discontinuity problem observed in the micromirror design reported by Xie et al. [19]. The resistance of the polysilicon resistor changes significantly with current. There are two effects attributed to the resistance change: piezoresistive effect and temperature dependence of polysilicon resistance. The temperature coefficient of the piezoresistivity is given by the product of the coefficient of thermal expansion and the gauge factor which are respectively ~2.510 -6 /K and ~30 for polysilicon, while the temperature coefficient of resistivity of the polysilicon used in these micromirrors was measured to be

PAGE 75

56 Rotation Angle (degree) (a) A pp lied Current ( mA ) Heater Resistance (k) (b) Applied Current (mA) Figure 3-7: 1-D mirror characterization. (a) Rotational static response. (b) Plot of the heater resistance versus applied current. Voltage Applied: (0.5 + 0.5cost) V Optical Scan Angle (degree) Frequency (Hz) Figure 3-8: Frequency response of the 1-D mirror.

PAGE 76

57 about 5.910 -3 /K. Therefore, the piezoresistive effect can be ignored. The resonant frequency of the mirror is 380 Hz, as obtained in the frequency response of Figure 3-8. The radius of curvature of the mirror surface is about 50 cm. 3.5.2 Two-dimensional Electrothermal Micromirror 3.5.2.1 Device design A two-dimensional (2-D) optical scanner was designed and fabricated by extending the 1-D mirror design concept presented in the previous section. This 2-D mirror uses a combination of two 1-D electrothermal actuators, to provide it two-dimensional scanning capability. The schematic drawing of this 2-D micromirror device is illustrated in Figure 3-9. The mirror is attached to a movable, rigid silicon frame by a set of bimorph aluminum/silicon dioxide thin-film beams. As before, a polysilicon resistor is embedded within the silicon dioxide layer to form the heater for the bimorph actuator. This movable silicon frame is connected to the silicon substrate by another set of identical bimorph thin-film beams that are oriented perpendicular to the first. In order to differentiate between the two actuators, the first set of actuators that rotate the mirror is defined as the mirror actuator, while the second set of beams that actuate the rigid silicon frame is defined as the frame actuator. The orthogonal orientation of these two actuators results in two perpendicular axes of rotation for the mirror plate. As shown in the cross-sectional view of the device (Figure 3-9 (b)), the top layer of the mirror is aluminum. Thus, the mirror has high reflectivity. A 40 m thick SCS layer backing the mirror plate keeps it optically flat. The mirror plate is 1 mm by 1 mm in size. This size is chosen for the micromirror to fit the available space in the OCT imaging probe. Each side of the rectangular frame is 75 m wide, and it also has a 40 m thick

PAGE 77

58 (b) (a) Figure 3-9: Schematic of the 2-D mirror design. (a) Top view showing the axes of rotation. (b) Cross-sectional view of A-A. SCS layer under it to provide rigidity to the structure. The heating element in the bimorph beams is a set of 200 m long, 7 m wide, polysilicon strips oriented along the beams. This is the same actuator design used by the 1-D mirror design of Section 3.5.1. The polysilicon layer from the CMOS process permits a maximum current of 1 mA per micron width. Therefore, only a maximum current of 7 mA can flow through the 68 polysilicon heater of each individual bimorph beam. In order to increase this current limit to a higher value, the polysilicon resistors in two adjacent beams are connected in parallel. This reduces the beam pair resistance to 34 and increases the maximum current to 14 mA. The fabricated mirror has 32 and 38 pairs of bimorph beams in the mirror and frame actuators, respectively. This results in mirror and frame actuator resistances of 1.1 k and 1.3 k, respectively. The SEM of a fabricated micromirror [29, 30] is shown in Figure 3-10. After fabrication, the initial tilt angles of the mirror and frame, with respect to the substrate, are 42 and 16, respectively. These initial tilt angles are due to the residual stresses present in the bimorph beams. The maximum actuation angles, allowed by this device design, are Frame Mirror Second Axis Mirror Actuato r First Axis A A A A Bimorph Beams ~5 m Mirror/Frame Frame Actuator Oxide Silicon Metal Poly-Si ~40 m Silicon Bimorph Actuator

PAGE 78

59 Figure 3-10: SEM of a fabricated 2-D micromirror. limited by the substrate contact points and also by the maximum electrical current that can be passed through the beams. Calculations based on a 500 m thick silicon wafer show that the mirror can tilt up to -22, while the frame can tilt up to -17 below the chip surface. Therefore, the maximum allowed rotation angles for the mirror and frame are 64 and 33, respectively. 3.5.2.2 Device characterization Various experiments were performed to determine the characteristics of this device. These experiments include static response, frequency response, long-term stability, and thermal imaging of the device. A simple experimental setup with a helium-neon (HeNe) laser and a dc current source was used to measure the static deflection angles. The mechanical rotation angle of the mirror was obtained by measuring the displacement of the reflected laser beam on a screen. Figure 3-11(a) shows the measured angles of rotation at different currents for the two independent axes. The mirror rotates 40 at an applied current of 6.3mA (or 15V, corresponding to an applied power of 95 mW), while the frame rotates by 25 at a current Mirro r Bimorph Actuators 1mm Frame

PAGE 79

60 Rotation Angle (degree) (a) Applied Current (mA) Resistance (k) (b) Applied Current (mA) Figure 3-11: 2-D Mirror Characterization. (a) Rotation angle vs. current, and (b) Polysilicon resistance vs. current for the two actuators. I 1 : current in mirror actuator. I 2 : current in frame actuator. A 7 mA frame actuator current is required for aligning the rotation axis of the mirror actuator with the substrate. of 8mA (or 17V, corresponding to a power of 135 mW). Mirror rotation angles up to 50 have been observed at higher currents, but the high stress induced in the bimorph actuator results in mirror instability. It has been observed that thermal damage in the polysilicon heater occurs at this point. The mirror instability limits the usable scan range of the mirror actuator to 40. The dc current dependence of the resistors is plotted in Figure 3-11(b). The resistances of the polysilicon heaters change significantly with current because

PAGE 80

61 Actuator Temperature (C) Figure 3-12: Thermal images of a device biased at 10 V. (a) Temperature distribution across the mirror actuator only [highlighted in (b)]. (b) Thermograph of the entire device. the heating effect of the current causes temperature change, which in turn induces stress change in the bimorph beams. The measured open circuit polysilicon resistances of the mirror and frame actuators at room temperature are 1.09 k and 1.26 k, respectively. The temperature distribution on the surface of the device was observed using an infrared thermal camera (FLIR ThermaCAM PM290). The temperature distribution profile of the entire mirror actuator is shown in Figure 3-12(a). Figure 3-12(b) shows this distribution over the entire device, and as expected, the mirror actuator has a higher temperature than the frame actuator due to the thermal isolation provided by the frame. Even though the actuator temperatures can be as high as 120C, the mirror plate and silicon substrate dissipate heat and remain at relatively lower temperatures (~40C). So there will be no thermal damage to tissue during endoscopic OCT imaging. Actuator Length (m) Actuator Width (m) (b) (a)

PAGE 81

62 The resonant frequencies of the mirror and frame actuator structures were measured to be 445 Hz and 259 Hz, respectively. The radius of curvature of the mirror was measured by a Wyko NT1000 white light profilometer to be 0.33 m. The mirror can be made optically flatter by using a thicker SCS layer backing the mirror plate. The long-term stability of the mirror was evaluated by scanning the mirror to steer a laser beam onto a fixed screen. The mirror was continuously scanned at 5 Hz, and the scan length and angular position of the reflected laser beam were monitored for over 2 million cycles. For the entire duration of the experiment, the observed angular drift was about 0.8; which is mostly due to fluctuations in ambient temperature. 3.5.2.3 Laser scanning experiment To further study the scanning behavior of the 2-D micromirror, a laser scanning experiment was performed, which simulates the 2-D transverse scanning for 3-D OCT imaging [129]. In this experiment, a simple visual display was successfully demonstrated by using this 2-D micromirror. The objective of this beam scanning experiment was to scan a pixel field with the micromirror and then to illuminate the selected pixels with a laser diode, thereby creating a projection display. The experimental setup is shown in Figure 3-13(a). By using a microprocessor to control the mirror and laser, 44 pixel-images were obtained at 10 frames per second. A sample image projected on a screen is shown in Figure 3-13(b). An active notch filter was incorporated into the amplifier to remove frequency content from the driving signals which could excite the mirrors resonant vibration modes. The 44 pixel resolution is largely limited by attempts to stabilize the mirror for each pixel. These techniques that were developed for a high-resolution projection display can be directly employed to control the laser beam scanning

PAGE 82

63 Scree n Scann i ng 2D M i crom irro r Laser Diode Figure 3-13: Laser scanning usi ng the 2-D m i rror. (a) Schem a tic of experim e ntal setup. (b) 4 x 4 pixel im ages scanned by the m i crom irror. in an endoscopic OCT system because the sam e basic operatio n of the device is required for both system s. Due to the large angu lar displacem ents by the two actuato rs the cen ter of the m i rror plate does not rem a in stationary in the ve rtical d i rection. For example, at a ro tation angle of 20 (optical ang l e of 40 ), th e cente r of the m i rror plate displaces downwards by 170 m This vertical dis p lacem ent of the m i rror plate do es n o t affect the working of the laser scanning display, but needs to be accounted for during OCT i m aging since it changes the optical path length of the scanni ng arm of the low-coherence interferom eter. 3.6 Micromirror Packaging For endoscopic OCT imaging, the m i crom irrors must be packaged inside endoscopes with diam eters rang ing from 3-5mm. It is proposed to use a sim ilar package design as used by Xie et al. for their MEMS-based endoscopi c-OCT system [51]. In this packaging s c hem e the m i crom irror is glued on to a sem i circular p i ece of printed circuit board (PCB) using a therm a lly-conductive epoxy. A picture of this custom-built PCB (a) Am p lifiers Micr o p rocess o r (b )

PAGE 83

64 package is shown in Figure 3-14. This PCB package has flexible electrical wires soldered to the surface gold-coated contacts. A wire bonder is used to wire-bond gold wire from the bond pads on the micromirror chip to the gold contact area on the PCB. A customized holder has been manufactured to hold the PCB package during this wire-bonding step. The packaged micromirror and PCB are then placed on a machined ferrule. This ferrule provides through holes for passage of the electrical wires from the PCB. The ferrule with the packaged micromirror is then fitted inside a hollow endoscope tube. There are two primary ferrule placement configurations. For forward-imaging probes, a stationary reflective mirror is required that directs the collimated light beam from the fiber on to the micromirror plate. The reflected light from the micromirror is focused in tissue through an optical window located on the distal end of the probe. The second configuration provides side-imaging OCT probes, in which the light beam exits the cylindrical probe through its side. For this arrangement, the ferrule is fitted into the Electrical Wires Gold Contacts Custom PCB Package Micromirro r Figure 3-14: Micromirror package. (a) Packaged micromirror on a custom PCB. (b) Picture of the PCB package. (c) Picture of a packaged mirror alongside a US dime coin.

PAGE 84

65 endoscope tube such that the micromirror is placed at an angle of 45 with respect to the optical fiber. The reflected light from the micromirror is scanned on to the tissue through an optical window located on the side of the endoscope. A schematic of these configurations is illustrated in Figure 3-15. Tissue Figure 3-15: Endoscopic OCT probe designs. (a) Side-im a ging configuration. (b) Forward-im aging configuration. 3.7 MEMS-based Endoscopic OCT Imaging 3.7.1 MEMS-based OCT System Design The electrotherm al m i crom irror packaged on the PCB, shown in Figure 3-14, has been installed into a custom m ade 5-mm dia m eter endoscope tube for in v i vo OCT im aging. The OCT system work was perfor m e d in collaboration with Dr. Yingtian Pan W i nd ow GRI N Lens Optical Fiber Micr o M i rro r (a) C ont r o l Lines W in do w GRI N Lens Optical Fiber C ont r o l Lines Mi c r om irr or Tissue F e rr u l e (b )

PAGE 85

66 and Zhenguo Wang of the State University of New York at Stony Brook who performed the endoscope system construction and testing. The schematic of this endoscopic OCT system is shown in Figure 3-16. A high-power, broadband light source (AFC Technology) with an output power of 13 mW, a central wavelength ( 0 ) of 1310 nm, and a full width half maximum (FWHM) spectral bandwidth () of 80 nm has been used. The coherence length that determines the axial resolution of the OCT system is 9.7 m. The pigtailed output from the broadband light Rapid Scanning Optical Delay Line Figure 3-16: Schematic of the MEMS-based endoscopic OCT system. CM: collimating GRIN lens, MM: micromirror, AO: acousto-optic Broadband Light Source Photodetector 50:50 Signal Processing Tissue Sample MM CM Endoscope Reference Arm f f CM Fiber AO Modulator Sam p le Ar m Fiber Coupler z x y Co mputer

PAGE 86

67 source is coupled into a fiber optic Michelson interferometer, where the input light beam is equally divided into the two arms of the interferometer using a fiber coupler. A fiber-optic polarization controller (FPC) in the reference arm of the fiber-optic interferometer ensures that the polarization of the exiting light beam from the non-polarization maintaining, single-mode fiber (SMF-28) is almost linearly polarized. The exiting light from this fiber is collimated into a 2-mm diameter optical beam by an angle-polished GRIN lens used as a collimator (CM). A rapid-scanning optical delay (RSOD) line utilizing a grating, lens and scanning mirror is used for axial scanning. In the RSOD, the temporal profile of a broadband light is linearly distributed at the Fourier focal plane of a grating-lens pair, and by placing a mirror at the focal plane and titling it rapidly results in fast group delay. To provide a stable and appropriately elevated Doppler frequency shift, a fiber-optic acousto-optic (AO) modulator is inserted into the reference arm before the RSOD. In this A-O modulator, two crystals are configured with one upshifted to 56 MHz and the other downshifted to 55 MHz to frequency modulate the light to 2 MHz for heterodyne detection. By carefully choosing the parameters of each component (e.g., f = 80 mm/35 mm for the scan lens, g = 450 lines/mm for the diffraction grating, 4 mm VM500 galvanometric mirror tilted at 4.2 and with 4 kHz repetition rate, and 2 MHz A-O frequency modulation), the high-speed depth scanner allows the acquisition of 4 K axial scans per second with an optical delay window of 2.8 mm (higher path length delay is possible by increasing the tilting angle). The high and stable Doppler frequency shift results in increased signal to noise performance of the signal processing electronics. Moreover, the dispersion induced by unbalanced fiber lengths and optical components between two arms of the Michelson interferometer can be

PAGE 87

68 minimized by slightly moving the grating along the optical axis, which can greatly enhance the axial resolution as has been observed during the alignment. The sample arm of the Michelson interferometer is connected to the fiber-optic MEMS-based OCT endoscope through FC/APC fiber connector. The design schematic of the forward-imaging MEMS-based OCT endoscope is shown in Figure 3-15(b). The light from the fiber is collimated by a 0.25-pitch selfoc GRIN lens to a 1.1-mm diameter optical beam, which is then reflected by a fixed mirror onto the surface of the tilted micromirror. The MEMS micromirror is used for transversely scanning this light beam onto a fixed laser doublet exit lens. The 5-mm diameter laser doublet has a focal length of 10 mm, and it focuses the light beam into a 12-m diameter spot at its image plane. Figure 3-17 shows photographs of the packaged OCT endoscopes. Since the MEMS mirror has an initial tilt angle of ~ 20, the custom-machined ferrule on which the micromirror sits is tilted by about 10 to keep the reflected beam on the center of the optical axis of the end lens. The backscattered light from tissue is collected by the same sample-arm optical path, and the combined interferometric signal from the sample and reference arms of the interferometer is detected by a photodetector. The detected signal is pre-amplified using a low-noise, transimpedance amplifier (Femto HCA-10M-100K), bandpass filtered and demodulated prior to being digitized by a 5 MHz, 12 bit A/D converter. Both depth scan and lateral MEMS scan are synchronized with the image data acquisition via two 16-bit D/A channels. All these components have been assembled into a readily transportable trolley console of Figure 3-18 to permit portable OCT imaging.

PAGE 88

69 25 cm Figure 3-17: Photographs of the 5-mm diameter MEMS-based OCT endoscope at the State University of New York at Stony Brook. Photographs by Z. Wang. Used with permission.

PAGE 89

70 Figure 3-18: Photograph of the portable, MEMS-based endoscopic OCT system at the State University of New York at Stony Brook. Photograph by Z. Wang. Used with permission.

PAGE 90

71 3.7.2 OCT Imaging Results (a) (b) Figure 3-19: Comparison of OCT with histological image. (a) OCT image, and (b) histological image of rat bladder. Imaged by Z. Wang. Used with permission. To demonstrate the ability of these MEMS mirrors for endoscopic OCT imaging the packaged endoscope shown in Figure 3-17, was connected to the portable OCT system shown in Figure 3-18. Figure 3-19 is a comparison between an OCT image and a histology photograph of a rat bladder. The OCT image was acquired at an imaging speed of 4 frames per second, and covers an area of 2.9.7 mm 2 [52, 130]. The lateral and axial resolutions are 15 m and 12 m, respectively. Figure 3-20 compares the OCT image quality for rat bladder tissue imaged using bench-top and endoscopic MEMSbased OCT systems.

PAGE 91

72 (a) (b) Figure 3-20: Bench-top versus MEMS-based endoscopic OCT imaging of rat bladder. (a) Bench-top OCT image. Size: 6 mm by 2.7 mm. (b) Endoscopic MEMS-based OCT image. Size: 4 mm by 2.7 mm. Imaged by Z. Wang. Used with permission.

PAGE 92

73 As can be see in these images, the morphological details of the rat bladder wall, e.g., the epithelium, submucosa and the upper muscularis layer are readily delineated. Because most transitional cell carcinomas originate in the epithelium, these results demonstrate the potential of MEMS-based endoscopic OCT for early detection and staging of bladder cancers. Also, as a wide variety of inner organs (e.g., cervix, colon, joints) can be accessed and imaged by front-view endoscopic OCT, the results suggest the potential applications of this technique for noninvasive or minimally invasive imaging diagnosis in these tissues. 3.8 Summary This chapter reviewed the different types of micromirrors that have been reported in literature. The selection of electrothermal actuation as the technique for micromirror actuation for this research project was justified. Theoretical analysis about electrothermal bimorph actuation, along with the fabrication process used to fabricate these micromirrors was also presented. This chapter also presented 1-D and 2-D electrothermal micromirror designs, along with experimental results. The low driving voltage and large rotation angles of these devices make them very suitable for use in endoscopic OCT imaging systems. A forward-imaging OCT endoscope using an electrothermal MEMS mirror for endoscopic light steering to achieve biomedical imaging at transverse and axial resolutions of roughly 15 m and 12 m, respectively, has been demonstrated. Cross-sectional OCT images covering an area of 4.0 .7 mm 2 can be acquired at 2-16 frames/s and with close to 100 dB dynamic range. The 5-mm diameter large OCT endoscope was chosen to fully use the internal clearance of a 22 Fr endoscope. Smaller OCT scopes can also be developed to accommodate various types of endoscopes.

PAGE 93

74 However, some optical design and packaging issues for these micromirrors include their unidirectional operation mode and non-stationary center of rotation of the mirror plate. These issues can be resolved by using the novel microactuator design presented in the next chapter.

PAGE 94

CHAPTER 4 LARGE-VERTICAL-DISPLACEMENT MICROMIRRORS AND NON-LINEAR OPTICAL IMAGING The 1-D and 2-D scanning micromirrors presented in Chapter 3 provide large rotation angles for transverse scanning in OCT imaging, but the unidirectional operation, non-stationary center of rotation, and large initial tilt angle of those micromirrors complicated the device packaging and optical design. These issues can be resolved by a novel mirror design that uses two complementarily-oriented electrothermal actuators, to keep the mirror surface parallel to the substrate and also to provide it bi-directional scanning capability. This chapter presents a new large-vertical-displacement (LVD) micromirror design that can perform rotational scans, as well as generate large piston motion at low driving voltages. Out-of-plane displacement of the micromirror is provided by a pair of electrothermal actuators. It is well known that there is large z-displacement at the tip of a long rotational beam. The innovation of this LVD device is converting the large tip displacement into a pure z-axis displacement of a flat micromirror. The LVD microactuator design can potentially achieve maximum vertical displacements of a few millimeters with millimeter-sized devices. Since this device can also perform bi-directional scans, it can also be used in the sample arm of an endoscopic OCT system to transversely scan the tissue surface. Further miniaturization of the OCT system is also possible by using the large piston-motion of LVD micromirrors to perform the millimeter-range axial scans that are currently scanned in the reference arm of Figure 2-7. Piston-motion micromirrors are also 75

PAGE 95

76 required by various other applications such as wave-front shaping in adaptive optics [131], interferometry systems [132], and spatial light modulators [133]. Numerous piston-motion actuation designs have been reported in literature. An electrostatic deformable micromirror reported by Helmbrecht et al. displaced up to 6 m [131]. Lee et al. and Kwon et al. presented devices that used electrostatic vertical comb drives (VCDs) to demonstrate maximum static vertical displacements of 7.5 m [134], and 20 m [60], respectively. Milanovi et al. reported an electrostatic VCD micromirror that generated piston motion of 60 m, but at a high actuation voltage of 130 V [101]. Cugat et al. reported a deformable micromirror with electromagnetic actuation that displaced up to 20 m for use in adaptive optics [135]. Yee et al. developed a piezoelectric micromirror for high-precision tracking of a laser beam for high-density optical data storage [136]. Tuantranont et al. demonstrated a 2 m displacement using an electrothermally-actuated trampoline-type micromirror for light phase modulation [137]. Wan et al. explored a new type of micromirror which utilized electrocapillary actuation to push a mirror vertically up to 8 m using a mercury droplet in a metal-plated microhole [138]. However, these actuators can only generate up to a few tens of microns of vertical displacement, and therefore cannot be used to meet the millimeter-range axial scanning requirement of OCT systems. In this chapter, two novel large-vertical-displacement (LVD) micromirror designs are presented that can generate bi-directional scans, and also perform large out-of-plane vertical displacement. The first mirror design uses one set of LVD microactuators to perform 1-D bi-directional scans. The second mirror design uses two sets of LVD microactuators to enable 2-D bi-directional rotational scanning.

PAGE 96

77 4.1 LVD Microactuator Design The LVD microactuator consists of two complementary electrothermal actuators in a folded structure which allows a mirror plate to remain parallel to the substrate surface, while still taking advantage of the large stroke lengths provided by the actuators. The schematic drawing of the LVD micromirror is illustrated in Figure 4-1 which is very similar to the 2-D electrothermal micromirror design which was presented in Section 3.5.2. The only difference being that the inner and outer bimorph actuators are aligned along the same axis in the LVD micromirror, instead of orthogonal to each other as in case of the 2-D micromirror. The inner and outer bimorph actuators are still referred to as mirror actuator and frame actuator, respectively. yx(a)Mirror Actuator Frame ActuatorMirror FrameEmbedded Polysilicon Heater Thermal Isolation Substrate AA yx(a)Mirror Actuator Frame ActuatorMirror FrameEmbedded Polysilicon Heater Thermal Isolation Substrate AA yx yx(a)Mirror Actuator Frame ActuatorMirror FrameEmbedded Polysilicon Heater Thermal Isolation Substrate AAMirror Actuator Frame ActuatorMirror FrameEmbedded Polysilicon Heater Thermal Isolation Substrate AA zx(b) Substrate l Zmir Lf -metal, -oxide -silicon, -polysilicon, Bimorph Mirror ActuatorFrame Actuator Mirror Frame Substrate l Zmir Lf -metal, -metal, -oxide -oxide -silicon, -silicon, -polysilicon, -polysilicon, Bimorph Mirror ActuatorFrame Actuator Mirror Frame zx(b) Substrate l Zmir Lf -metal, -oxide -silicon, -polysilicon, Bimorph Mirror ActuatorFrame Actuator Mirror Frame Substrate l Zmir Lf -metal, -metal, -oxide -oxide -silicon, -silicon, -polysilicon, -polysilicon, Bimorph Mirror ActuatorFrame Actuator Mirror Frame zx zx(b) Substrate l Zmir Lf -metal, -oxide -silicon, -polysilicon, Bimorph Mirror ActuatorFrame Actuator Mirror Frame Substrate l Zmir Lf -metal, -metal, -oxide -oxide -silicon, -silicon, -polysilicon, -polysilicon, Bimorph Mirror ActuatorFrame Actuator Mirror Frame Figure 4-1: Design schematic of the LVD mirror. (a) Top view. (b) Cross-sectional view across A-A.

PAGE 97

78 (a) Mirror Frame Actuator Figure 4-2: Coventor simulations. (a) Device side-view. (b) 3-D model of the LVD micromirror illustrating the initial curling of the bimorph actuators. The mirror surface is parallel to the substrate plane since the curling of the two actuators compensate each other. FEM thermomechanical simulation was conducted using CoventorWare [139]. The simulation results are shown in Figure 4-2, where the curlings of the two sets of bimorph beams compensate each other resulting in a zero initial tilt. The initial elevation of the mirror plate above the substrate plane, z mirror due to the curling of the thermal actuators can be calculated from: sinmirrorffzLW (4-1) where L f and W f are the length and beam width of the frame, respectively. the initial tilt angle of the frame, can be computed from = l/r ; where l and r are the length and (b) Frame Mirror Si Bimorph Beams Substrate Plane Frame Mirror Actuator

PAGE 98

79 radius-of-curvature of the thermal actuator, respectively. For a frame with L f = 0.5 mm, W f = 40 m, and = 17, the initial rest position of the mirror Z mirror is 135 m. The simulation results in Figure 4-2 show that the mirror plate is located 0.132 mm above and parallel to the substrate plane. There is no substrate underneath the mirror plate. The mirror and frame actuators rotate the mirror in opposite angular directions. Therefore, there exist two basic modes of operation: (1) Bi-directional scanning by alternatively applying voltages to the mirror and frame actuators; and (2) Large piston motion by simultaneously applying voltages to both actuators. Equal angular rotations by the two actuators will result in pure vertical displacement of the mirror. Large z-axis displacement is achieved via the angular amplification due to the long arm length of the frame. In order to enable independent electrical excitation for each actuator, a wiring schematic as shown in Figure 4-3 is used. The metal-1 aluminum layer on top of the bimorph beams is electrically divided into several paths to carry the actuation current for the inner actuator. For example, the metal-1 layer on the frame actuator has been divided Figure 4-3: Wiring schematic for the LVD actuators. Inset: Section of a frame actuator bimorph beam showing that the mirror actuator current (i 1 ) is carried by metal-1 layer. V1 V2 Frame MA Mirror FA i 1 i 1 metal-1 i 2 i 2 oxide poly-Si

PAGE 99

80 into two electrical paths that carry current in and out for the mirror actuator. As seen in the inset of Figure 4-3, the actuation current which flows though the polysilicon heater is electrically isolated from the current flowing through the metal-1 layer by a thin oxide layer. 4.2 1-D LVD Micromirror 4.2.1 Fabricated Device A fabricated 0.7 mm by 0.32 mm LVD micromirror device [140-142] is shown in Figure 4-4. The fabrication process is exactly the same as the one described in Section 3.3. This LVD device has its mirror plate elevated about 100 m above the silicon substrate plane. The initial tilt angle of the frame with respect to the substrate surface is 13. The heating element in the 10-m wide bimorph beams is a set of 200 m long, 7 m wide, polysilicon strips oriented along the beams. The gaps between the beams are 9 m and used to undercut silicon to form thin-film bimorph beams. The frame actuator and mirror actuator are constituted of 20 and 12 bimorph beams, respectively. The measured open circuit polysilicon resistances of the mirror and frame actuators are 240 and 365, respectively. The mirror plate is 190 m by 190 m. This small mirror size is just used to demonstrate the proof of concept. Since the mirror plate is supported by bulk silicon, much larger mirrors can be made. The quality of the mirror surface was determined using a Wyko NT1000 white-light optical profilometer. A line scan of surface heights across the mirror plate is shown in Figure 4-5. The peak-to-valley surface deformations are within 40 nm over the 190 m mirror plate. The optical quality of the mirror is better than /20 for near-infrared light.

PAGE 100

81 xz y xz y Frame Actuator Mirror Actuator Figure 4-4: SEM images of the LVD micromirror. Figure 4-5: Line scan of the surface profile of the LVD micromirror. 4.2.2 Equivalent Circuit Model Modeling an electrothermal actuator involves multiple domains (i.e., electrical, thermal and mechanical). Since most of the actuators electrical properties are temperature dependent, electrothermal coupling makes the modeling very challenging especially in the case where self-heating elements are used. For instance, in the case of a Mirror Frame Mirror SCS Frame 0 50 100 150 200 -50 -40 -30 -20 -10 0 10 20 30 40 50 Mirror Plate (m)Mirror Deformation (nm)

PAGE 101

82 bimorph cantilever beam with an embedded resistor, electrical current causes Joule heating of the resistor which generates a temperature distribution along the beam. Conversely, this temperature gradient along the beam produces non-uniform Joule heating of the resistor due to a finite temperature coefficient of resistance of the resistor material. The operation of the micromirror has been modeled by using an equivalent electrical circuit, as presented in Figure 4-6, by partitioning the individual components of the device into electrical and thermal domains. This multi-domain model [128] simulates the electro-thermal behavior of the LVD actuators by considering the change in the polysilicon heater resistances, thermal coupling between the actuators, and heat loss by thermal convection. The unit of the product of the across and through variables in the thermal domain is Watts-Kelvin, while the thermal current has dimensions of power. Joule heating generated by the electrical resistor is modeled as a dependent current source in this thermal domain. Additional details about this electrothermal model developed by Todd and Xie are available in [128]. This model is valid when only one of the two actuators is activated. The subscript i in the parameters denotes the actuator that is activated and j denotes the other actuator Electrical DomainThermal Domain iTmax_ EiiRV2iAR_iViR0EiR iRiTR..0 2).1(iiRxiisoR_CFR0T jisoR_ jAR_ 2jR 2jR jT iT iixR3112 iiixxR3112 2.iiRx Figure 4-6: Equivalent circuit model of the LVD micromirror device. Voltage is only applied to actuator i, and actuator j is always grounded.

PAGE 102

83 with zero applied voltage. Therefore we just need to consider two cases: 1) the mirror actuator is activated, i.e., i = MA and j = FA; and 2) the frame actuator is activated, i.e., i = FA and j = MA. In the electrical domain, the electrical resistance of actuator i, R Ei is temperature-dependent and is given by 01EiiRi R R T (4-2) where R 0i is the electrical resistance of the polysilicon heater at the substrate temperature T 0 R is the thermal coefficient of resistance of polysilicon, and 0iiTTT where iT is the average temperature of actuator i. In the thermal domain, the thermal resistance, R T is directly analogous to electrical resistance, and can be computed for different parts of the mirror by using the following expression: wtRT (4-3) where , w and t are the thermal conductivity, length, width and thickness of the thermal resistor, respectively. The various thermal resistors used in the equivalent circuit model of Figure 4-6 are defined in Table 4-1. In the circuit model shown in Figure 4-6, using Kirchoffs current law, it can be easily shown that the average temperature of the bimorph beams of the actuator i is given by: 2041112RTiiiiRiRTxVR (4-4)

PAGE 103

84 Table 4-1: Parameters used by the equivalent circuit model of the 1-D LVD micromirror. Parameter Description Electrical Domain V i Voltage applied to actuator i R Ei Electrical resistance of actuator i at temperature iT R 0i Resistance of actuator i at substrate temperature, T 0 R Thermal coefficient of resistance of polysilicon Thermal Domain R i Thermal resistance of actuator i R j Thermal resistance of actuator j R CF Resistance of frame due to convection R CM Resistance of mirror plate due to convection R iso_i Resistance due to thermal isolation of actuator i R iso_j Resistance due to thermal isolation of actuator j T max_i Maximum temperature of actuator i iT Average temperature of actuator i jT Average temperature of actuator j Where: i ={ m f } j = { f m } and i j x m = 0.5 x f = 0.66 R Am = R iso_m + R CM R Af = R iso_f where x i denotes the position of the maximum temperature of the actuator i. and can be determined by thermal imaging or by the burn pattern of actuator i. Todd and Xie determined the maximum temperature position on the 1-D micromirror actuator by using high-resolution thermal imaging [128]. The value of x i can also be determined by the position of the burn pattern of actuator i. For this analysis, electrical current flowing through the actuator was increased till burn marks appeared on the surface of the bimorph metal layer.

PAGE 104

85 The value of x i was obtained from the SEM image of Figure 4-7 as: iiblxL (4-5) where, l i is the distance of the actuator i burn pattern from its anchor location, and L b is the length of the bimorph actuator. Figure 4-7: SEM of the burn pattern of the mirror actuator. The value of x i is calculated as: x i = l i /L b [Equation (4-5)] Using the analysis presented by Todd and Xie [128], TiR the equivalent average thermal resistance of the actuator i, can be expressed as : _1123iTi A iiRRx R (4-6) The average temperature, j T, of the inactive actuator j can be computed in a similar manner. As discussed in Section 3.4, the rotation angle of each actuator is given by: 03226()()()446kAloxAloxk 3 A lAloxoxAloxAloxoxoxAlAllttTTEtEtttttEtEt (4-7) where, k = MA or FA, l is the length of the bimorph actuator, and ta a and Ea are the thickness, thermal coefficient of expansion, and Young's modulus of the metal (a=Al) and oxide (a=ox) layers, respectively. l i L b

PAGE 105

86 Therefore the net rotation of the mirror is obtained by plugging the values of iTan d j Tinto Equation (4-7) and solving as: rotij ( 4-8) As seen in Equation (4-8), this model considers the rotation of the inactive actuator j due totuators. From Equationetween the average temperature change and a thermal coupling between the two ac (4-4) it is observed that the average temperature of the bimorph actuator becomes approximately linear with voltage when the applied voltage is above a certain threshold value. This linear relationship b pplied voltage is represented as: 0iTiRixRTVVR when i 04iRTiRV R x (4-9) Combining Equation (4-9) with E quation (3-6) yields the linear relation between the actuation angle and the applied voltage as given below. 0RiRwhere L iTiTrbTxRLV (4-10) ion is very important for practical use, and experimental verification will be presented in the next section. 4.2.3 Experimental Results from Section 3.5.2.2 was used to characterize this LVD micromirror. The mirror plate of the LVD micromirror rotates 26.5 when 3 V dc is applied to the mirror actuator. The mirror plate and the frame both rotate when a voltage is applied only to the frame actuator due to b is the length of the bimorph beam. This linear relat 4.2.3.1 Static response The same experimental setup used for the 2-D micromirror characterization

PAGE 106

87 thermal coupling effects. The mirror plate rotates -16.5 when 5.5 V dc is applied to the V (since P = V2/R). Note that the electrical resistance, R is not constant and in ee that r itional perature coefficient of resistance (TCR) of polysilicon. As evident from Figure 4-8(c), there exists a linear correlation between the frame actuator. As shown in Figure 4-8(a), the rotation angles of the actuators vary linearly with applied voltages after the applied voltages are above the small threshold voltages. The rotation-angle response plot can be explained using Equation (4-4) which was derived from the equivalent circuit model of the device. Rotation by a bimorph actuator is proportional to its applied electrical power, i.e., it is proportional to the square of the applied voltage, creases linearly with increasing temperature due to a non-zero thermal coefficientof resistance of the polysilicon heater. Therefore, the angular response has an initial nonlinear V 2 -like dependence, as seen in Figure 4-8(a) and in Equation (4-4) at low voltages. Consistent with the analysis presented in the preceding section, one can safter a threshold voltage is crossed (Equation (4-9)), the rotation angle can be approximated to be linearly proportional to the applied voltage. The simulation results using the equivalent circuit model match the experimental data to within 8% in the 15 rotation range, as shown in Figure 4-8(a). The same actuation voltage causes a larger rotation angle by the mirror actuatothan the frame actuator due to the polysilicon resistance difference between the two actuators, and also because the mechanical structure of the frame provides addthermal isolation to the mirror actuator. The polysilicon resistances of the bimorph actuators change with applied voltages, which are plotted in Figure 4-8(b). The large resistance change is caused by the large tem

PAGE 107

88 Figure 4-8: LVD mirror characterization. Plots of the (a) rotation angle versus applied voltage, and (b) polysilicon heater resistance versus applied voltage for the two actuators. (c) Plot showing the linear correlation between rotation angle and polysilicon resistance of the actuators. 0 1 2 3 4 5 6 -30 -20 -10 0 10 20 30 Experimental Model Applied Actuator Voltage (V) Rotation Angle (degree) Mirror Actuator (a) Frame Actuator 0.2 0.3 0.4 0.5 0.6 0.7 -20 -10 0 10 20 30 Mirror ActuatorFrame Actuator Actuator Resistance (k) Rotation Angle (degree) (c) (b)

PAGE 108

89 rotation angle and the polysilicon resistance for each of the two actuators. The nonlinearity is less than 2% and is within the rotation-angle measurement error. This linear relatiip allows for independent control of rotation angle of each actuator by monitoring its polysilicon heater resistance. Thermal coupling between the two actuators can also be accounted for by monitoring their individual polysilicon heater resistances. Large piston motion of the mirror can be achieved by equal but opposite angular rotations of the two actuators. By usinersus actuation-voltage data, a mirror-actuator drive-voltage versus frame-actuator drive-voltage plot for same angular rotation values can be obtained. The slope of this experimentally determined plot provides the driving voltage ratio for the two actuators that would maintain no tilting of the mirror plate. A voltage divider was used to drive the two actuators with a voltage ratio of 3:7 (determined from experiment). A maximum vertical displacement of 200 m was obtained. The vertical displacement of the mirror as a function of the drive voltage is shown in Figure 4-9(a). By using a linearly-fitted voltage ratio, about 1 tilting of the mirror plate was observed during the full-range vertical actuation. A 2nd-order polynomiaing, as shown in Figure 4-9(b), can be used to further reduce the tilting to less than 0g of the mirror plate during this experiment was monitored by using a quadrant photodetector. 4.2.3.2 Frequency response/resonant scanning The frequency response of the LVD micromirror was measured using a Polytec OFV-511 laser Doppler vibrometer, as shown in Figure re observed at 1.12 nd2.76 kHz fe onsh g the rotation-angle v l-fitt.03. The tiltin 4-10. Resonant peaks we kHz and 2.62 kHz. These results are a close match to the modes observed at 1.14 kHz arom simulations using CoventorWare. When current is passed only through th

PAGE 109

90 0 1 2 3 4 5 6 7 0 150 200 Figure 4-9: Piston motion mode. (a) Vertical displacement of the mirror plate as a versus frame actuator voltage that was used to drive the LVD device to obtain he mechanical resonance of the mirror structure (Qresonance frequency of 2.6 kHz, the function of the frame actuator voltage. (b) Plot of the mirror actuator voltage less than 0.03 tilting of the mirror plate. mirror actuator at its resonance, t factor of 25) generates bi-directional scans. At the optical angle scanned by the mirror is 170 at a dc plus ac drive voltage of (0.6 + 0.6sint) V. Scan angles greater than 170 were observed visually at marginally higher voltages, but could not be monitored since the reflected light beam is blocked by the package sidewall. This large-angle scanning is stable and repeatable. 50 100 Frame Actuator Voltage (V) Vertic al Displacement (m) (a) 0 1 2 3 4 5 6 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 Experimental DataFitted to "y=0.04x2 + 0.016x" R2 = 0.9992 Frame Actuator Voltage (V) Mirror Actuator Voltage (V) (b)

PAGE 110

91 Figure 4-10: Frequency response of the LVD micromirror device. Frequency doubling: In response to the applied ac voltage signal, the power, Pac dissipated by the actuator is given by: 1 2 3 4 5 0.1 1 10 100 1000 FrequDis p lacet ( a. u. ) 1.12 kHz2.62 kHz men ency (kHz) tPTRVPTRtVP TRV Pacacacacacacac2cos1)(.2)(sin)( ~ 2222 (4-11) where R is the polysilicon resistance of the actuator. Therefore, frequency doubling is observed when the device is actuated by only an ac voltage signal. Note that the polysilicon resistance R doeally increases with increasing power due In oring, a dc bias is added to the applied s not remain constant, but actu to a non-zero thermal coefficient of resistance. der to reduce the effect of frequency doubl ac voltage signal, and the power dissipated is given by: t RVRVVRVVRtVVacacdcacdcacdc2sin2222By increasing the dc bias V tPdcac2cossin (4-12) dc above the threshold voltage given in Equation (4-9), the device operates in the linear region. In this operation region, the angular response is

PAGE 111

92 directly proportional to the applied voltage, so the effect of frequency doubling can be significantly reduced4.3 2-D LVD Micromirror 4.3.1 Mirror Design schematic view of the 2-D micromirror with its nomenclature is illustrated in Figure 4-11. The 0.5mm-by-0.5mm mirror plate is attached to a rigid silicon frame by a set of bimorph aluminum/silicon dioxide thin-film beams. This first set of bimorph beams embenother set of identical bimorph thin-film beams, known as actuator 2 (Act2). Act1 and Act2 together form a large-vertical-displacement (LVD) microactuator set, in which the curls of the two sets of bimorpis In order to enable 2-D scanning, a second set oD actuators results in two perpendicular axes of rotation for the micromirror. As seen in the SEM of the 2-D LVD mirror [143] in Figure 4-12, the inie a A is referred to as actuator 1 (Act1). As shown in Figure 4-11(c), polysilicon resistors dded in the bimorph beams are used for electrothermal actuation. This frame is connected to a second outer frame by a h beams compensate each other resulting in zero initial tilt of the mirror plate. A detailed analyson LVD microactuators has already been reported in the preceding sections. Bi-directional 1-D line scanning along the y-axis is possible using this LVD microstructure by alternately applying voltage to actuators Act1 and Act2. f LVD actuators (Act3 and Act4) is attached to the first, as shown in Figure 4-11(a). The orthogonal orientation of the two sets of LV tial tilt angle of a fabricated mirror plate is less than 0.5, and its rest position is 1.24 mm abovthe substrate plane. Each side of the three rectangular frames is 40 m wide, and has

PAGE 112

93 highlighting the 4 bimorph actuators. (b) Top view of the actuator area boxed A. 40-m-thick SCS layer under it to provide rigidity and thermal conduction to the substrate. The heating element in the bimorph beams is a set of 200-m-long, 7-m-wide polysilicon strips oriented along the beams. The primary mirror-rotation directions for actuators 1, 2, 3, and 4 are along the +y, y, +x, and x axes, respectively. However, the Figure 4-12: SEM of a fabricated 2-D LVD mirror. The cross-sectional view of A-A is shown in Figure 4-11(c). Figure 4-11: 2-D LVD micromirror design. (a) Top view of the 2-D micromirror, in part (a). (c) Cross-sectional view of the bimorph actuator as seen across AxyzAct3 Act2 Act1 Mirror Act4 Single-Crystal Silicon (SCS) 1m m Frame Substrate A A SCS Oxide Silicon Substrate MetalPolysilicon Frame Bimorph (c) Actuator Substrate Frame A (b) A Mirror xz y xy x(a) Act2 Act1 Act3 Frame Substra te z Act4

PAGE 113

94 heating of the active actuators will cause unwanted rotations by the other actuators due to thermal cou. This issue discussed in the next section. 4.3.2 Experimental Results 4.3.2.1 Bi-directional scanning n experimental setup with a laser beam incident on the mirror and dc voltages applied to the four actuators was used to deterhe static 2-D scanning response of each actuator. The optical angle scanned by the mirror was determi the displacemeam on a calibrated x-y screen. Figure 4-13(a) shows only Act4 scansaxes. The corresponding scan-angle versus actuation-voltage characteristics for each of the four actuators are shown in Figure 4-13(b). Act4 scans along the x axis, while Act1, Act2 and Act3 scan 1-D lines angled at +60, -66, and -28 with respect to the x-axis, respectively. This 2-D micromirror device scans optical angles greater than in the x-direction, and over in the y-direction at dc actuation voltages less than 12 V. The deviation of a line scan from its primary axis in Figure 4-13(a) is caused by thermal coupling between the actuators. Sinc is directly connected to the silicon substrate it is affected by thermal coupling, and this can be observed in Figure 4-14(a), where Acned consistently along the x direction for different Act1 bias voltages. The thermal coupling between the actuators can b extending the LVD electrothermal model reported in Section 4.2.2. The resistances of the polysilicon heaters embedded in all four actuators increase pling will be A mine t ned by measuring ent of the reflected laser b the static 2-D line scans obtained by actuating each actuator individually, in which along its primary axis while the other scan lines deviate from their primary e Act4 least t4 scan e modeled by significantly with applied voltage because of Joule heating. The measured polysilicon

PAGE 114

95 Optical angle scanned in Figure 4-13: Static 2-D line scans. (a) Plot showing the optical angles scanned in 2-Dspace when each actuator is individually actuated. (b) Plot of the effective optical angle scanned versus actuation voltage for each actuator. Act4 scans along x, while Act1, Act2 and Act3 scan at +60, -66, -28 with respect to the x-axis, respectively. resistance of each actuator is listed in Table 4-2. A linear correlation between the opticscan angle and the polysilicon resistance for each of the four actuators was obseshown in Figure 4-14(b). This correlation allows al rved, as for independent control of the rotation angle. of each actuator by monitoring the resistance of each individual polysilicon heater y direction ( de g rees ) +x -x -y Optical angle scanned in x direction (degrees) +y (a) Actuation Voltage (V) e (degrees (b) Optical Angl)

PAGE 115

96 tic s Heater Maximum Static Mirror Rotation Resonant Figure 4-14: Static characterization. (a) Plot showing the linear scan pattern during sta2-D scanning of Act1 and Act4 only. Act4 was actuated at different Act1 biavoltages. (b) Linear plot of actuator resistance versus optical scan angle for each actuator. Table 4-2: Actuator characteristics for the 2-D LVD micromirror. Actuator Resistance Optical Scan AngleDirection Frequency Act1 1.3 k 36 at 8.4 V -30 to +y-axis 870 Hz Act2 1.3 k 41 at 9.1 V 24 to -y-axis 452 Hz Act3 0.62 k 66 at 8.5 V -28 to +x-axis 312 Hz Act4 0.62 k 38 at 11.4 V along x-axis 170 Hz Optical angle of each actuator (degrees) (b) Actuator Poly-Si Resistance (k) Act1 scan range Act4 scan range 0 V2.7V 4.4V 6V Act1 bias voltages Optical angle scanned inOptical angle scanned in y direction (degrees) (a) x direction (degrees)

PAGE 116

97 Thermaupling between the actuators can also be accounted for by monitoring the individolysilicon heaces. The angular stability of the micro maximum xy tilt angle) was experim determined with respect to time, as shown in Figure 4-15. From this plot it dmicromirror stability for Act1 t4 were within 0.6 and 0.16, respectively, for their entire optical scan ranges (which are 36 and 38). The initial tilt angle of the mirror plate at different environmental temperatures was also documented, and the results are presented in Figure 4-16. For this experiment, the micromirror was heated by a thin-film micro-heater placed under a packaged device, and the tilting of the mirror plate at different temperatures was optically monitored. Theoretically, a uniform change in device temperature would cause equal rotational displaces by all four actuators, thereby negating any net mirror tilting. However, as observed in Figure 4-16, there is significant mirror tilting in the x direction and this is igure 4-15: Tilt angle stability of the mirror plate versus time. Act1 and Act4 were excited to rotate the mirror to its maximum scan angle. l coual pentallyetermined that the ter resistan mirror (at its and A c was ment F

PAGE 117

98 Figure 4-16: Initial tilt angle of the mirror plate in x and y directions at different environmental temperatures. attributed to the fact that Act4 is heated more than the other actuators since it is direcconnected to the substrate, which in turn is directly connected to the heat source. 4.3.2.2 Two-dimensional dynamic scanning tly is 2-D scanning using this device was demonstrated by simultaneously exciting both Act1and Act4 actuators with small ac voltage signals. The frequency and phase of the ac signals were varied in order to generate the Lissajous figures shown in Figure 4-17. Thdevice exhibits resonant peaks at 870 Hz, 452 Hz, 312 Hz, and 170 Hz due to the different actuators as summarized in Table 4-2. A 2-D raster-scanning pattern was generated by the micromirror when Act1 was supplied with 1 Vdc plus 1 Vac at its resonance of 870 Hz, and Act4 was supplied with 2 Vdc plus 2 Vac at 15 Hz. As shown in Figure 4-18, 58 parallel lines were scanned in a raster-scan pattern by the laser beam covering a 14 by 50 parallelogram angular area.

PAGE 118

99 (a) (b) (c) 14 14 (d) (e) (f) Figure 4-18: 2-D raster scanning pattern obtained using actuators 1 and 4. Figure 4-17: Photographs of 2-D scan patterns obtained by exciting actuators 1 and 4 only. (a)-(e) Lissajous figures scanned by the micromirror by varying only the phase of the two excitation signals. (f) Lissajous figure scanned at an excitation frequency ratio of 1:10. 14 50 Reflection from Frame

PAGE 119

100 4.3.2.3 Vertical displacement motion Vertical displacement of the mirror plate can be achieved by equal but opposite angular rotations of only one set of LVD microactuators. For this experiment, only actuators 3 and 4 were used. The Act3 actuation voltage was increased in small increments to rotate the mirror plate in the +x direction. This +x-direction mirror rotation was compensated by supplying voltage to Act4, which rotates the mirror along the x direction. The opposite rotations brought about by actuators 3 and 4, result in pure vertical displacement of the micromirror along the +z direction. It should be noted that due to thermal coupling, the inactive actuators 1 and 2 will tilt the mirror plate along the y direction. The vertical displacement of the center of the mirror plate as a function of Act4 voltage is shown in ed at an Act4 tween the Act3 voltage and the corresponding Act4 voltage that is required to vertically displace the micromirror as obtained in Figure 4-19(a). Using this voltage relationship, the x-axis tilting of the mirror plate was less tng the entire 0.53 mm actuation range. An almost linear correlation between the vertical displacement of the mirror plate and the actuator resistances was observed, as shown in Figure 4-19(c). This linear relationship provides a closed-loop feedback path for determining mirror displacement by monitoring the actuator resistances. The z-axis vertical scan provided by actuators 3 and 4, in combination with the y-axis resonance scan provided by Act1 will generate 2-D scans in the y-z plane. As mentioned above, thermapling between the active Act3 and Act4 actuators causes rotation by Act1 and Act2 in the orthogonal y-dection. This y-axis tilting of the Figure 4-19(a). It can be seen that a maximum z-displacement of 0.53 mm was observvoltage of 15 V. Figure 4-19(b) shows the almost linear relationship be han 0.8 duri l cou ir

PAGE 120

101 (b) (a) Vertical dislacement o the mirror plate (m) Act3 Actuation Voltage (V) Act4 Actuation Voltage (V) pf (c) Actuator Poly-Si Resistance (k) Figur (c) Linear increase of Act3 and Act4 resistance with vertical displacement. Vertical Dis p lacement of Mirror Plate ( m ) Act4 Actuation Voltage (V) e 4-19: Piston motion mode. (a) Vertical displacement of the mirror plate as a function of Act4 voltage. (b) Corresponding plot of the relationship between Act3 and Act4 voltages that is required to generate the vertical displacement.

PAGE 121

102 Figure 4-2ilting of the mirror plate in the negative y-direction (due to thermal pling) as a function of vertical position of the mirror. mirror plate was monitored and is plotted as a function of vertical displacement in Figure 4-20. A maum y-direction mirror-plate tilting of -7 was observed. This -y mirror tilting is my due to the heating of Act2 by the Act3 and Act4 active actuators. Even with thermal coupling it is possible to scan a 2-D depth scan using this mirror; however the scanned area would be displace towards the reasing z-displacemerved in Figure 4-20). 4.4 MEMS Mirror-based Nonlinear Endoscopy Using MEMS mirrors for nonlinear optical imaging was demonstrated in collaboratiith Ms. Ling Fu and Prof. Min Gu of the Centre for Micro-Photonics, Swinburne University of Technology, Hawthorn, Australia. MEMS mirrors have been used to facilitate endoscopic beam steering because of their small size, potentially low cost and excellent micro beam Vertical Displacement of Mirror ( m)Tilting of mirror plate in y direction (degrees) 0: Tcouximainl y direction with inc ent (as obson w manipulating capacity.

PAGE 122

103 4 .4.1 Nonlinear Optical Imaging System A schematic of the nonlinear imaging system is depicted in Figure 4-21. A Ti:Sapphire laser source at 800 nm (Spectra Physics Mai Tai) is used to generate ~80 fs pulses at a repetition rate of 80 MHz, and an output power of 850 mW. The laser pulses pass through an iris diaphragm, and are focused into a double-clad photonic-crystal-fiber using acope objective lens (40x /0.65 NA). The double-clad photonic crystal fiber (Crystal Fiber A/S) has a core diameter of 20 m, an inner cladding diameter of 165 m, an outer diameter of 5 of 0.6 at a wavelength of is used for from the laser s-RIN lens Figure 4-21: Schematic of the nonlinear optical imaging system. mi cros 50 m, and a numerical aperture (NA) 800 nm. This double-clad photonic crystal fiber is used for two purposes: the central core single-mode propagation of the 800 nm near-infrared light source to the tissue sample, while the high NA inner cladding is used for efficient multimode propagation of visible light from the sample to the photodetector. The crossection of this fiber is shown in Figure 4-22. The excitation laser beam exiting from the double-clad photonic-crystal-fiber is scanned by a MEMS mirror onto the surface of the tissue sample. A 0.2-pitch G Dichroic Prechirp Unit Ti:Sapphire Laser PMT 40x Mirror Bandpass Filter Iris Filter Micromirror Double-Clad Objective Tissue Photonic Crystal Fiber GRIN Lens

PAGE 123

104 Fu. Used with permission. with a diameter of 1 mm (GRINTECH) is used to focus the laser beam to a diffractlimited spot-size of ~1 m onto the tissue sample at a working distance of about 0.15 mm. The two-photon excitation fluoresce Figure 4-22: Cross-sectional view of the double-clad photonic crystal fiber. Imaged by L. ion-nce and second harmonic generated optical signais e dnaplified by an amplifier (Oriel, Model #70710), and then digitized using a data acquisition scans are synchronized with the image acquisition data through the use of a Labview program. 4.4.2 Experimental Results The 1-D micromirror was used for SHG imaging as shown in the ure 4-se, a 0.2IN lens with a diameter of 1 mm was used, and the resulting field-of-view on the sample was 35 m. This corresponds to an optical scanning ls from the tissue are collected using the same GRIN lens and micromirror, and thUV/visible spectrum signal propagates through the inner cladding of the double-clad photonic crystal fiber before being detected by a photo multiplier tube (PMT) (Oriel, Model # 77348). Th etected sig l is pre-am card. The lateral MEMS results of Fig 23. In this ca -pitch GR

PAGE 124

105 range requirement of only ~ 6 by the 1-D micromirror. A Labview program provides the 5V actuation signal to the micromirror to perform this angular scan. Since the laser beamis scanned at the back-surface of the GRIN lens, the GRIN lens is under-filled and this results in an axial resolution of about 10 m. The optical power of the excitation laser beam on the tissue is approximately 30 mW. Figure 4-23(a) shows a series of SHG line profiles from the rat tail tendon, which are spaced at 10 m in the axial direction. A ondimensional stage (Melles Griot) was used for the axial scanning. Figure 4-23(b) shows a SHG line profile from a rat esophageal tissue which was removed from a euthanized rat, e-any stainincan Figure 4-23: Second harmonic generation imaging. (a) Series of SHG line profiles from rat esophagus tissue. Obtained by L. Fu. Used with permission. imt mersed in Hanks balanced salt solution (no phenol red) and imaged directly withoug. In order to obtain more meaningful 2-D images, the 2-D LVD micromirror from Section 4.3 was used with the same nonlinear imaging system. The lateral raster spattern of this mirror was first characterized. Subsequently, a Labview program was written to supply control voltages to Actuators 1 and 4 only to obtain the 2-D scan rat tail tendon taken at 10 m axial steps. (b) SHG line profile of an unstained 0 10 20 30 40 0 0.5 1 1.5 2 2.5 Lateralposition( m) SHG fromop eshagus (a.u.) Lateral position (m)

PAGE 125

106 Figure 4-24: 2-D raster scan pattern scanned by actuators 1 and 4 only. pattern shown in Figure 4-24. The nonlinear optical image was reconstructed according to the scanning calibration data. The 1-D mirror in the setup was replaced by the 2-D LVD mirror, and to demonstrate 2-D TPEF imaging 10 m diameter fluorescent imaged using a bench-top TPEF system, while Figure 4-25(b) is the TPEF image acquired using the 2-D LVD micromirror. This figure has a field of view of 50 m, and a microbeads were imaged as shown in Figure 4-25. Figure 4-25(a) shows the beads lateral resolution of 1.5 m. (a) (b) Figure 4-25: TPEF imaging of 10-m diameter fluorescent microbeads. (a) Bench-top TPEF imaging system. (b) MEMS-based TPEF imaging. Imaged by L. Fu. Used with permission.

PAGE 126

107 Figure 4-26: In vitro imaging of rat stomach epithelial surface stained with 1% acridine orange in Ringers solution. Scale bar represents 10 m, and image spacing is 10 m. Imaged by L. Fu. Used with permission. There are a number of promising clinical applications for this nonlinea r optical 2These results demonstrate the potential of MEMS-based endoscopic TPEF and SHG imaging systems for the early detection and staging of gastrointestinal cancers. verification endoscope, but its applications in the areas of the gastrointestinal tract and the oral cavityhave been highlighted as shown in Figure 4-26. This figure shows the in vitro TPEF imaging of rat stomach epithelial surface (stained with 1% acridine orange in Ringers solution) at different penetration depths. The field of view of these images is 50 by 50 m, and the lateral imaging resolution is 1.5 m. The columnar mucosal tissue microstructure of the rat stomach is clearly visible. Also, as a wide variety of inner organs (e.g., cervix, colon) can be accessed and imaged by similar endoscopic nonlinear optical imaging systems, the results suggest the potential applications of this technique for noninvasive or minimally invasive imaging diagnosis in these tissues. 4.5 Summary This chapter presented a novel LVD microactuator design along with experimental This actuator design used a complementary configuration of two actuators

PAGE 127

108 for mdisplath greater than 170 at resonance. A 2-D LVD micromirror was also successfully demonstratthan 12 V. 2-D dynam This aking micromirrors that are capable of high-speed vertical scanning as well as 1-D bi-directional rotational scanning. A maximum vertical displacement of 200 m has been achieved with a microdevice of only 0.32mm by 0.7mm in size. Much larger vertical cements in the millimeter-range can be achieved by simply increasing the lengand/or the initial tilt angle of the frame. This 1-D mirror also scanned optical angles ed to be able to scan optical angles larger than at driving voltages less ic scanning using this device was demonstrated by obtaining a 14by 50 angular raster scan pattern. Nonlinear optical endoscopy using these micromirrors has been demonstrated, and imaging resolution as high as 1.5 m has been achieved. result validates the potential of using these scanners for in vivo detection of precancers.

PAGE 128

CHAPTER 5 MICROLENS SCANNERS AND OPTICAL CONFOCAL MICROSCOPY This chapter presents a novel microlens scanner design that can displace its focal The electrothermal scanning micromirrors presented in Chapters 3 and 4 cannot be used for ultrahigh-resolution OCT imaging since their transverse resolution was limited point up to a few millimeters. This electrothermally-actuated microlens scanner uses the large-vertical-displacement (LVD) microactuator design, which was introduced in Chapter 4, to generate large out-of-plane displacements at low actuation voltages. 5.1 LVD Microlens Scanner y the required depth-of-focus of the optical beam. This is due to the fact that the rse resolution (spot size of the focused beam) is inversely proportional to numerical aperture (NA), while the depth-of-focus is inversely proportional to the square of NA. However, in order to obtain high transverse resolution without compromising the axial scanning range and small size requirement of endoscopic OCT/OCM systems, a MEMS microlens can be used to scan along the optical axis. Various microlens scanner designs have been reported in literature. Hoshino and Shimoyama developed a pneumatically-actuated microlens arrays that demonstrated a maximum vertical displacement of 11 m [144]. Kim et al. demonstrated an electrostatically-actuated microlens scanner with a maximum displacement of 7 m for optical data storage applications [145]. Kwon et al. presented a microlens scanner for confocal imaging that b transve 109

PAGE 129

110 used electrostatic vertical-comb-drives to produce a maximum displacement of 55 m at resonance [62]. These abovementioned scanners generate vertical displacements only up to a few tens of microns, but OCM im requires an axial scanning depth of at leasnother method by which the focal point of the microlens can be scanned axially is by vater Peseux demonstrated a 6-mmat the interface between two immiscible liquids [148]. By using the princis to 0 V. Although these tunable lenses meet the millimeter-range scanning requiremay isplace its m aging typically t a millimeter. A rying the focal length of a tunable lens. Graham developed a variable focus lens using a deformable chamber filled with liquids with different refractive indices [146]. Sugiura and Morita also demonstrated a variable-focus, liquid-filled, 27-mm diamelens that varied its focus by changing the liquid volume in the lens [147]. Berge and diameter, variable lens formed ple of electrowetting [149] the contact angle of one of the liquids was changed, thereby changing the focal length of the lens. However, this lens requires a high dc voltage of 210 V in order to double its focal length. Krupenkin et al. also used electrowetting to change the curvature of a liquid microlenvary its focal length [150]. They reported a 0.5 mm increase in focal length at an applied voltage of 8 ent of OCM systems, their bulky-size and/or high operating voltages are a deterring factor in their use for endoscopic imaging applications. Also, liquid lenses mnot be unsuitable for OCM imaging since infrared light may be absorbed by the lenses. In this section a novel microlens scanner design is presented which can dfocal point up to a few millimeters. A miniature MEMS scanner that can displace its focal point by up to 280 has been fabricated and tested to demonstrate the proof-of-concept

PAGE 130

111 of this scanner design. This microlens scanner design can be easily scaled to meet the 2-mm axial scanning requirement of OCM systems. 5.1.1 Microlens Scanner Design The schematic o f the LVD microlens scanner is illustrated in Figure 5-1. The lens holdes of h the e, r ampliscanner was fabricated [151, 152]. A fabricated lens holder shown in Figure 5-2(a) is r consists of a square plate with a circular opening in the middle. A polymer microlens is formed on top of this plate. The lens holder is attached to a rigid silicon frame by a set of aluminum/silicon-dioxide bimorph actuation beams. Since these beamactuate the lens holder plate, they are collectively referred as the lens actuator. A 30-50 m thick single-crystal-silicon (SCS) layer under the lens holder prevents curling the thin-film layers of the plate. The rigid silicon frame is attached to the substrate by another set of identical bimorph actuation beams, referred to as the frame actuator. The lens and frame actuators together form one set of the LVD microactuator, in whicout-of-plane curls of the two sets of bimorph beams compensate each other. Therefore, the lens holder and the lens are parallel to the substrate surface. Polysilicon resistors embedded in the bimorph beams provide heating for electrothermal actuation. As shown in Figure 5-1(b), the initial elevation of the lens holder above the substrate surface isgiven by: H = L f sin where L f and are the length and initial tilt angle of the framrespectively. Equal angular rotations by the two actuators will result in pure vertical displacement of the microlens. Large z-axis displacement can be achieved via the angula fication due to the long arm length of the frame. 5.1.2 Fabricated Microlens Scanner As a proof of concept, a miniature version of the millimeter-range microlens

PAGE 131

112 D microlens scanner. (a) Top view. (b) Cross-tration of the scanner. Figure 5-1: Design schematic of the LVsectional side view. (c) 3-D illus Lf Lens actuato r Frame actuato r (a) S ubstrate F rame l y x PR lens (b) z x Lens actuator Frame actuator PR microlens y z x oxide aluminum polysilicon Lens holder Frame Al SiO 2 Si (c)

PAGE 132

113 Figure 5-2: SEMs of: (a) Fabricated LVD lens holder, and (b) LVD microlens scanner. only 0.7 mm by 0.32 mm in size. The lens holder plate is 200 m by 200 m in size, and it has a circular opening with a diameter of 140 m, on which a photoresist (PR) microlens is formed. The initial rest position of the lens holder is 120 m above the substrate surface. This LVD lens holder was fabricated using the same DRIE CMOS-MEMS process presented in Section 3.3, and its processing steps are illustrated in Figure 5-3. After the four DRIE CMOS-MEMS dry etch steps, a fifth step has been added that integrates a polymer microlens with the device. In this final processing step, a droplet of Shipley Microposit 1805 photoresist (PR) is dispensed using ction system and then baked in an oven (120C for 30 minutes) to form a microlens due to surface tension. Since there are no substrate or thin-film layers directly above or below the microlens structure, large verticaltuation range is allowed. Table 5-1: Microlens characteristics asu Par a nano-inje ac Mered ameters Lens Diamete r 210 m Lens Hei g ht58 m Focal Len g th f188 m Estimated Parameters Numerical A p erture ~ 0.35 Microlens Volume~ 1.1nLRadius of Curvature~ 124 m Lense -holder Plat SCS Frame Actuato r Lens Actuato r Frame ( a ) 200m ~120 m ( b ) PR Microlens

PAGE 133

114 oxidesilicon substrate Si trench etch. (d) Si undercut. (e) Microlens formation by reflow of morph beams. A fabricated e measured and estimated parameters of the microlens are listed in Table 5ved on the bottom surface of the microlens within an acche esmeters were calculated using analysis presented by Daly a flat bottom lens surface. Estimations using the smal20he bottom radius-of-curvature increase the focal length by less than 12%. The PR microlens has a diameter of Figure 5-3: Microlens fabrication process. (a) Backside Si etch. (b) Oxide etch. (c) Deepphotoresist. The initial rest position of this fabricated microlens is 120 m above the substrate surface. The weight of the microlens causes less than 1 m vertical displacement of the lens-holder plate. This is attributed to the small volume of the microlens and also due to the large stiffness of the bi microlens is shown in Figure 5-4. Th -1. Since no cur vatur e was obser uracy of 2 m, t timated para [153] with the assumption of lest possible value (~1 0 m) for t (b) (d) frame actuator(e) bimorph beam frame lens holder lens actuator PR microlens ~ 5 m ~ 40 m (c) metal-3metal-2metal-1 poly-SiSCS membrane (a) -3-2-1 poly-SiSCS membrane (a) oxidesilicon substrate oxidesilicon substrate (b) (b) (d) frame actuator(e) bimorph beam frame lens holder lens actuator PR microlens (d) frame actuator(e) bimorph beam frame lens holder lens actuator PR microlens ~ 5 m ~ 40 m (c) ~ 5 m ~ 40 m (c)

PAGE 134

115 about 210 m. Since the circular opening on taller diameter of 140 mthis helps tocal aberrahe measured focal length of the microlens is about 188 m. A 0.2 nanoliter variation in the dispensed volume of the PR dropleocal length of the formicrolens to within 10%. Microlenses with longer focal lengths can be fabricated by reducing the volume of the photoresist drople by increasing the size of the lensabrication process also allows for the formation of mrolenses tha optically transparent for the visible spectrum of light by using UV-curable polymers instead of photoresist. 5.1.3 Experimental Results Vertical displacement of the microlens is accomplished by equal angular rotations of the two actuators. A plot of the vertical displacement of the microlens as a function of the drive voltage of the frame actuator is shown in Figure 5-5(a). A maximum vertical displacement of 280 m was obtained when a dc voltage of 10V (corresponding to a dc current of 12 mA) was applied. Figure 5-5(b) shows the driving voltage ratios that are needed for the two actuators to generate pure vertical displacement. This plot clearly he l e ns holder has a sm reduce spheri tio n T t will change the f ed m t, or ho ld er. This f ic t are Figure 5-4: SEM of a fabricated PR microlens. shows two regions of device operation, which are related to the position of the microlens PR crolens Mi Frame

PAGE 135

116 with respect to the substrate surface. The voltage ratio increases linearly with displacement until the microlens displaces below the substrate surface, at which poinratio becomes constant. A voltage divider can be used to drive the lens and frame actuators with a voltage ratio of 1:2.3 after the lens displaces below the substrate surface. The tilting of the microlens is less than 1 during the entire 280 m vertical scan range t this s 0 2 4 6 8 10 0 50 100 150 200 250 300 Frame Actuator Voltage (V)Vertical Displacemem) Figure 5-5: Vertical displacement experiment. (a) Vertical displacement of the microlenas a function of frame actuator voltage. (b) Plot of the ratio of the applied voltages to the lens and frame actuators that was used to obtain the verticaldisplacement shown in (a). nt ( above substrate surface (a) below substrate surface 0 50 100 150 200 250 300 0 0.1 0.2 0.3 0.4 0.5 0.6 Vertical Displacement (m)uagerage Lens ActFrame Act atoruato Volt Volt (b) below substrate surface above substrate surface

PAGE 136

117 0 50 100 150 200 250 300 0 0.2 0.4 0.6 0.8 1 Vertical DispActuat Resi orstance (k) Frame Actuator Lens Actuator below substrate surfaceabove substrate surface lacement (m) Figure 5-6: Plot showing the increase in polysilicon heater resistances versus vertical displacement for the two actuators. Since the focal length is smaller than the thickness of the substrate, the incident light beam must enter from the substrate side. Longer focal lengths are possible by reducing the volume of the PR droplet, but at the expense of reduced NA. For example, a 0.3 nanoliter droplet will form a 0.5 mm focal length microlens, but with a reduced NA of 0.14. A longer focal length gives the microlens a larger working distance. The open circuit polysilicand frame actuators are 240 and 360 respectively. The electrical resistance of the polysilicon heaters embedded inside the bimorph beams increases with an increase in applied voltage due to Joule heating. A linear correlation between the vertical displacement of the microlens and the polysilicon resistance of the two actuators was observed, as shown in Figure 5-6. This correlation allows for closed-loop control of the vertical position of the microlens by monitoring the resistance change of each individual polysilicon heater. Although researchersincreasing able change in focal length was observed (within on resistances of the lens have reported significant variations in focal length of polymer lenses with temperature [154], no notice

PAGE 137

118 1 mm rn the ttern. r over 7 billio p m by hown in Figure 5-7: Microlens imaging quality. (a) Schematic of the imaging experiment apparatus. (b) Photo of the test pattern. (c) Snap-shot images of the test patteas obtained through the PR microlens. The triangular markers indicatelocation where two images were merged to obtain the complete test pa2 m) during the actuation range of this device. The LVD actuators exhibited no failure or performance degradation after being scanned at their full actuation range fo n cycles. The resonant frequencies of the lens and frame actuator structures are 2.15kHz and 1.1 kHz, respectively. The imaging quality of the PR microlens was tested using an experimental setucomprising of a microscope, CCD (charge coupled device) camera, chrome photomask and a light source, as shown in Figure 5-7(a). A test pattern on the mask (placed 100 mbelow the microlens) was imaged by the microlens. The test-pattern was demagnified about 100 times, and its image was formed at the focal plane of the microlens as sFigure 5-7(a). This image was then captured by the CCD camera installed on the microscope using a 40 objective. Figure 5-7(b) shows the pictures of the test pattern, 10 m 10 m 10 m 40x CCD 40x CCD ffobj(b) (c) =633nm (a)

PAGE 138

119 -20 and Figuree (~1.1m) of this microlens. A CCD image and its intensity profile of a 4 m sized spot are shown in Figure 5-8. 5.2 Millimeter-Range LVD Microlens Scanner The LVD microlens scanner presented in Section 5.1 was fabricated to validate the ment. -10 0 10 20 2 -2 0.5 1.0 Transverse Distance ( m) arbitrary units 10 m Figure 5-8: CCD image of a 4 m focused beam spot (top), and its corresponding intensity profile (bottom). 5-7(c) shows the photographs of the images captured using the microlens. Th field of view of the captured photos is 40 m by 30 m, and features smaller than 2 m can be easily distinguished. The measured spot size when imaging a pinhole is less than 2 m using a 633 nm collimated light source, which is close to the theoretical resolution concept of using a LVD microactuator to vertically displace an integrated microlens. As mentioned in Section 2.2, OCM requires vertical displacement of the focused beam-spot by 0.5 mm to 2 mm; therefore the LVD scanner design was scaled to meet this require

PAGE 139

120 The larger version of this scanner (referred to as the mm-LVD microlens device) was fabricated using the AMI 0.5-m 3-metal CMOS process. 5.2.1 Millimeter-Range Scanner Design A schematic of the mm-LVD microlens device is shown in Figure 5-9. In order to amplify the vertical displacements generated by the LVD microactuator, the frame length (Lf) was increased to 1.6 mm. The initial tilt angle provided by a 200-m long bimorph actuator that is fabricated using the AMI 0.5-m process is ~ 40. This initial tilt angle along with the increased frame length will increase the initial elevation of the microlens holder to an estimated value of about 1 mm. The diameter (D) of the circular opening on the lens holder has been increased to 600 m, and a transparent silicondioxide mesh photoresistrge opening during the microlens formation structure has been added in this opening. This oxide mesh was added to prevent the droplet from falling through the la step of the fabrication process. The fabrication process has been slightly modified to allow for the formation of this oxide mesh. Figure 5-9: Top view of the millimeter-range LVD microlens scanner. L f Lens Actuator Substrate Frame Lens Holder Frame Actuato r l Oxide Mesh D

PAGE 140

121 5.2.2 Fabrication Process The modified wafer-level fabrication process is shown in Figure 5-10. The first step hickness of the SCS membrane. The next step is the is the backside silicon etch that defines the t frontside oxide etch which uses the metal layers as an etching mask. After the oxideetch, photoresist is spun on the device as shown in Figure 5-10(c). The photoresist is then etched uniformly across the entire device until the metal-2 and metal-3 layers are completely exposed. This is shown in Figure 5-10(d), where the metal-1 layer is still protected by photoresist. Next, a wet metal etch is done to remove the exposed metal layers. After this, the remaining photoresist is removed using an oxygen plasma clean (e) (f) (g) bimorph beam frame lens(e) (e) holder metal-1 CMOS regionpoly-SiSCS membrane(a) oxidesisu licon bstrate (b) (c) (d) metal-2metal-3 photo-resist ~ 5m ~ 40 m (h) PR microlens lens actuatorframe actuator (f) (f) (g) bimorph beam frame lens(g) -holder bimorph beam frame lens -holder metal-1 CMOS regionpoly-SiSCS membrane(a) oxidesisu licon bstrate (b) (c) (d) metal-2metal-3 photo-resist ~ 5m ~ 40 m metal-1 CMOS regionpoly-SiSCS membrane(a) oxidesisu licon bstrate (b) (c) (d) metal-2metal-3 photo-resist ~ 5m ~ 40 m (h) PR microlens lens actuatorframe actuator(h) PR microlens lens actuatorframe actuator Figure 5-10: Modified fabrication process for mm-LVD microlens scanner. (a) Backside Si etch. (b) Oxide etch. (c) Spin on photoresist. (d) Anisotropic photoresist etch to expose metal-2 layer. (Metal wet etch followed by photoresist removal. (f) Deep Si trench etch. (g) Silicon undercut. (h) Micrlens e) o formation by reflow of photoresist.

PAGE 141

122 process to obtain the device as shown in Figure 5-10(e). The next 3 steps are exactly the same mple irst ot required to be planar), a photomask with an 800-m circular opening is used (Figure 5-11(c)). This circular opening is aligned to the lens holder within a lateral tolerance of about 20 m. After the metal layers on the mesh and the lens-holder area are exposed, aluminum wet-etchant is used (Figure 5-11(d)). The other steps are exactly the same as before. Figure 5-12 shows the photograph of a device after the lithographic and Figure 5-11: Modified process for die-level fabrication of the microlens scanner. as before, in which a silicon trench etch is performed followed by a silicon undercut to release the device. The last step involves formation of the microlens by precisely dispensing photoresist onto the oxide mesh of the lens holder. It is challenging to fabricate millimeter sized CMOS dies using this wafer-level fabrication process because spinning on a planar layer of photoresist is difficult. Therefore, a modified process was used to fabricate on the die level that used one silithography step. The die-level fabrication process is illustrated in Figure 5-11. The ffew steps are the same, but after spinning on a layer of positive photoresist (which is n oxidesilicon substrate photo-resist 1. Backside DRIE & Oxide Etch2. Spin on Photoresist oxidesilicon substrate photo-resist 1. Backside DRIE & Oxide Etch2. Spin on Photoresist m ~ 40 m ~ 5 ~ 40 3. Open Mesh using Lithography m ~ 40 m ~ 5 ~ 40 3. Open Mesh using Lithography 4. Metal Etch + PR Removal 4. Metal Etch + PR Removal lens actuatorframe actuator bimorph beam frame lens-holder lens actuatorframe actuator 5. DRIE & Si Undercut6. Microlens Formation lens actuatorframe actuator bimorph beam frame lens-holder lens actuatorframe actuator 5. DRIE & Si Undercut6. Microlens Formation

PAGE 142

123 Figure 5-12: Photograph of the CMOS die after the wet etch of lithographically exposed aluminum layers. aluminum wet-etch steps. The process modifications allow the use of the metal-2 layer to define the structure for the 3-m thick oxide mesh within the lens holder. Also, the PR microlens is now formed on a layer of insulating oxide instead of a more thermally-conductive Al metal layer, resulting in better thermal isolation from the electrically-heated lens-actuato r. Thermally isolating PR microlenses from heating sources prevents the situations observed by Glebov et al. [154], where the focal length of a polymer microlens changed significantly with increasing heat. Thermal isolation also prevents carbonization of the PR microlens, which tends to occur at temperatures greater than 200C. The initial rest position of a fabricated microlens [155] is 1.2 mm above the substrate surface, as shown in Figure 5-13. In order to form a polymer microlens on the elevated lens-holder, droplets of PR are first dispensed using a nanoliter-injection system and then baked in an oven to form a microlens due to surface tension. Figure 5-14 shows a fabricated microlens scanner integrated with a photoresist microlens. oxide silicon aluminum

PAGE 143

124 r; and (c) frame actuator bimorph regions. Figure 5-14: SEM of a fabricated microlens scanner with integrated polymer microlens. Carbon Tape Lens Actuator Frame Frame Actuator Lens Holder 1 mm Substrate (a) (b) (c) Figure 5-13: (a) SEM of a fabricated scanner before microlens formation. Close-up views of (b) lens actuato Lens Actuator Polymer Actuator Frame Frame Microlens Substrate Focal length can be controlled by varying the quantity and/or volume of the dispensed dropets. PR microlenses with focal lengths between 0.5 to 3 mm with numerical apertures (NA) ranging from 0.1 to 0.35 have been successfully fabricated. Figure 5-15 shows two fabricated PR microlenses with different focal lengths and lens

PAGE 144

125 (a) (b) Figure 5-15: SEMs of (a) convex microlens, and (b) ball-type microlens. sThe conhas a focal length o1 Nnd a lens diameter of and a diameter of ~0.4 mm. To test the imaging quality of these microlenses, a chrome test mask was placed ~100 mm away from the ball-type microlens, and the image at its focal plane (which is de-magnified by about 100) was captured using a CCD microscope camera. The test pattern image obtained by a PR microlens with a numerical aperture of 0.35 is shown in Figure 5-16, where features as small as 1 m can be resolved. Table 5-2 provides a theoretical estimate of the optical properties of the microlens for different desired focal length values. The focal length of the microlens in conjunction with the package design determines t distance between the scanner and in vivo tissue. As seen in Table 5-2, a longer focal length (i.e., a longer working distance) reduces the lateral imaging resolution which in turn is determined by the focused spot size. A slight variation in the volume of the dispensed photoresist droplet will change the focal length significantly; therefore a more precise photoresist dispensing technique is required for good repeatability. A commercially available inkjet dispensing system which dispenses droplets izes. vex microlens f 3 mm, 0. A a 0.8 mm. The ball-type microlens of Figure 5-15(b), has a focal length ~0.5 mm, 0.35 NA, he working

PAGE 145

126 on a chrome mask. (b) Corresponding image of the test pattern as seen at the hotoresist microlens. Table 5-2: Estimated microlens parameters for various desired focal lengths. Focal Length Spot Size Focus Aperture Volume 100 m 1 m (a)(b) Figure 5-16: Imaging using the photoresist microlens. (a) Photograph of the test pattern focal plane of the p Targeted (mm) Focused (m) Depth of (m) Numerical (NA) Required Photoresist (nL) 1.0 2.8 9.3 0.29 10.6 1.5 4.2 20.8 0.20 6.7 2.0 5.6 37.1 0.15 5.0 2.5 6.9 57.9 0.12 4.0 3.0 8.3 83.4 0.10 3.3 Theseof 1310 nm. with an accuracy of few tens of picoliters may be used. 5.2.3 Experimental Results Vertical displacement of the PR microlens was achieved by simultaneously exciting both actuators and tuning the actuation voltages such that the opposite angular rotations of the two actuators offset any net tilting of the microlens. The static displacements of the microlens and its corresponding driving voltage plots are shown in values were computed for a microlens diameter of 600 m, and light wavelength

PAGE 146

127 Figure 5-17, where the vertical displacem of the microlens were observed to within an accuracy of m by using a 40 objective lens (0.65 NA). Figuraximum vertical displacement of 0.71 mm was obtained at a dc voltage of 23 V applied to the frame actuator. Figure 5-17(b) displays the same vertical displacement data, but with respect to the electrical power supplied to the two actuators. The plot of the two driving voltages that are required to obtain this vertical displacement is shown in Figure 5-17(c). The slope of this linear plot yields the driving voltage ratio for the two actuators for LVD actuation. The driving voltage applied to the divider. Us.71mm travel range is less than 0.4. stanlysers emothsigni with apoltage doule heatinle heating raisetempee of the bihs, whichn increases the heater resistancesthermafficient ofstivity of licon. The open-circuit, room-temture n between the vertical displacement of the microlens and ar relationship between the resistors allows for closed-loop feedback control of the vertical position of the microlens by monitoring the polysilicon resistance change of each actuator. The initial elevations of the microlens at different operating temperatures were also documented, and the results are presented in Figure 5-19. The initial lens-holder elevation ents e 5-17(a) shows that a m lens actuator is 43% of the frame actuator voltage, which is provided by a voltage ing this constant voltage ratio, the maximum tilt of the microlens in the entire 0 The resi ces of the po ilicon heat bedded in bg. Jou actuators increase ficantly plied v ue to J s the ratur morp in tur due to the l coe resi polysi pera electrical resistances of the lens and frame actuators are 1.17 k and 1.35 k, spectively. A linear correlatio re the heater resistance of the two actuators was observed, as shown in Figure 5-18. This experimentally determined line

PAGE 147

128 Frame Actuator Voltage (V) Vertical Displacement ((a) pment ((b) o o m) Figure 5-17electrical power. (c) Corresponding linear relationship between the two Frame Actuaator Voltage (V) Electrical Power (mW) Vertic tor Voltage (V) Mirror Actu(c) al Dislacem) : Vertical displacement of the microlens scanner. (a) Microlens displacement as a function of frame actuator voltage. (b) Displacement versus applied voltages.

PAGE 148

129 Figure 5-18: Change in resistance versus vertical displacement of the microlens. of a fabricated microlens is 1.2 mm at a room temperature (23C), and it decreases with increasing temperature. Since this device will be packaged inside an endoscope for in vivo OCM imaging, the maximum ambient temperature shall not exceed 40C. Therefore, the vertical scan range of the device will, in the worst case, be reduced by up to 20 m, which is less than 3% of the entire scan range. This can be seen in the plots of Figure 5Figure 5-19 Vert ical Dis p lacement ( m ) Actuator Resistance (k) : Change of initial microlens elevation with ambient temperature. Ambient Temperature (C) Microlens Elevation (m)

PAGE 149

130 17(a) where an increase in ambient temperature to 40C reduced the vertical scan range by ~17 m. There isificant lateral shifting of the microlens during vertical actuation, as shown in Figure 5-20. A maximum lateral shift of 425 m is observed for the entire vertical scan range. This lateral shift is mainly due to the rotational displacement of the Figure 5-20: Lateral shift of the microlens during vertical displacement actuation. (a) hift. sign Microlens Lateral Shift ( m) Microlens Elevation (m) (b) (a) Illustration of the lateral shift. (b) Characterized plot of the lateral s

PAGE 150

131 frame actuator, and is amplified by the long length of the silicon frame. This lateral shift will be accounted f or and corrected during the OCM image formation stage. a Figure 5-21: Dynamic response of the microlens scanner. (a) Mechanical response when to e The dynamic response of the device was characterized using the laser Doppler vibrometer. Figure 5-21(a) shows the measured velocity response of the device whenpulse excitation was applied to both actuators at time, t = 0. The settling time of this Time (seconds) Velocity (m/s) (Velocity Response (a.u.) ( te Envelope 226 Hz 609 Hz a) b) Frequency (Hz) a square excitation was applied to both actuators at t = 0. This can be fittedan exponential envelope using the damping ratio, (b) Frequency responsshowing the resonant peaks.

PAGE 151

132 device is around 300 ms, and the quality factor was determined from the envelope by fitting an exponential plot. The Q-factor for this device has been measured to be around 65 to ved at 226 Hz and 609 Hz, for the frame and lens actuators, respectively. 5.3 LVD Microlens Packaging To package this microlens scanner into a 5-mm diameter endoscope, a custom holder was fabricated by TMR Engineering, Micanopy, FL. A schematic of this package design is shown in Figure 5-22(a). The LVD microlens scanner was glued on to this hollow cylindrical plastic package. Three copper contacts are available for wire-bonding the microlens to the electrical feed-through wires. The center of the package is machined Figure 5-22 75. This plot provides the settling time of the device after being subjected to environmental shock and vibrations. Figure 5-21(b) is the frequency response of the device measured using the same laser Doppler vibrometer. Resonant peaks were obser Copper Contacts Copper Electrical GRIN Lens Stop Optical WindowGRIN LensOptical FiberCopper WireMicrolens Scanner Tissue 5 mm Control Lines Optical Fiber Optical WindowGRIN LensOptical FiberCopper WireMicrolens Scanner Tissue 5 mm Control Lines Optical Fiber Feed-through (a) (b) : Microlens package design. (a) Microlens package schematic. (b) Forward-imaging OCM endoscope.

PAGE 152

133 (a) WiGRIN Lens Optical ndow 9 mm 5mm Electrical Wires Optical Fiber Aluminum Probe Lucite Cap Figure 5-23: Packaged OCM endoscope. (a) With the Lucite end cap. (b) Without the end cap to show the packaged microlens scanner. to provide automatic alignmtween the collimating lens and the microlens, and also to provide a stopbarrier for the collimating lens. A packaged microlens scanner is then inserted into a forward-imaging endoscope, as shown in Figure 5-22(b). The packaged OCM endoscope of Figure 5-23 is about 15 mm long, and has a diameter of 5 mm. Figure 5-23(b) shows the inside of the endoscope by removing the Aluminum Microlens Scanner Probe US 1 cent coin (b) ent be

PAGE 153

134 distal Lucite end cap. This Lucite cap has a rectangular optical window that allows the focused light beam to exit this forward imaging endoscope. 5.4 Summary This chapter presented a tunable microlens scanner which used microactuator to verticallhotoresist microlens by up to 280 m at a low actuation voltage of 10 V. Thmpatible fabrication process also allows for the formation of optically transparent microlenses ensing ultraviolet-curable polymers instead of photoresist. Microlenses with longer focal lengths can be fabricated by reducing the volume of the dispensed droplet, or by increasing the size of the lens-holder. In order to increase the vertical scanning range of this device, a millimeter-range LVD microlens scanner has been designed and fabricated. A modified fabrication process, eme for this molens scanner device was also presented. This device demonstrated vertical displacement of the focal plane of an integrated polymer microlens by up to 0.71 mm at an actuation voltage of 23 V. The lateral imaging resolution can be increased to as high as 1 m using these devices. an LVD y displace a p e CMOS-co by microdisp along with a packaging sch m-LVD micr

PAGE 154

CHAPTER 6 CONCLUSIONS AND FUTURE WORK This research effort has successfully developed and demonstrated various MEMS optical scanners specifically designed for use in endoscopic biomedical imaging applications. The integration of MEMS-based endoscopes with these bioimaging modalities can potentially detect and diagnose internal cancers at an early stage using minimally invasive techniques. Reflective micromirror scanners have been presented that used electrothermaactuation to provide large rotational scans at voltages below 20 V. A single-axis micromirror was successfully packaged inside a 5-mm diameter endoscope with other optical components to provide endoscopic laser-beam steering. OCT images obtainedusing this MEMS-based OCT endoscope achieved imaging resolutions of 15 m and l 12 mrse and axial directions, respectively. In vivo cross-sectional OCT images covering an area of 4.0 .7 mm2 were acquired at an imaging speed of 2 to 16 frames per second. The optical design, alignment, and packaging issues were addressed through the design of another class of electrothermal micromirrors which used the large-vertical-displacement (LVD) microactuator. The LVD microactuator used two complementarily-oriented electrothermal actuators, to keep the mirror surface parallel to the substrate and also to provide it bi-directional scanning capability. LVD micromirrors demonstrated large rotational scans, as well as the ability to generate large vertical piston motion at low in the transve 135

PAGE 155

136 driving voltages. The LVD microactuator design can potentially achieve maximum vertical displacements of a few millimeters with millimeter-sized devices. Scanning LVD micromirrors have demonstrated endoscopiility through their use in nonlinear optical endosster-scan pattern to obtain two-dimensional two-photon excitation fluorescence and second harmonic genera71 mm at a voltage of 23 V. This microlens scanner was successfully packags. al ising imaging resolution which is critical to enable early d c imaging capab copy. The micromirrors were scanned in a ra tion images with very high resolution (~1.5 m). Finally, a microlens scanner using the LVD microactuator was reported. This device can physically scan the focal plane of an integrated polymer microlens and is therefore suitable for axial focus scanning applications. A fabricated device demonstrated an axial displacement of as high as 0. ed inside a 5-mm endoscope for high-resolution, endoscopic OCM imaging. Till date, OCM imaging has been possible only on bench-top systemHowever, the scan-range and imaging resolution provided by the microlens scanner demonstrates the potential of developing endoscopic OCM systems using this technology. The large actuation range, low driving voltages, and small footprint of these opticmicroscanners make them very suitable for endoscopic biomedical imaging applications. These experiments show very prom etection of cancerous markers. The high resolution along with the scanning speed, open up the possibility to make compact, high-performance endoscopes for future clinical applications through the use of microsystem technology.

PAGE 156

137 6.1 Research Effort Accomplishments The following tasks have been accomplished for this project: Four variations of 1-D micromirror devices were designed, fabricated, characterized and packaged for project collaborators at the State University of New York, Stony Brook, NY. The 2-D micromirror devices were characterized, and a 4 by 4 pixel laser scan ned. z. rototype e demonstrated a maximum vertical displacement of 280 m at an actuation the 1-D micromirror scanners to image in vivo tissue with resolutions around 10 m. pattern at 10 frames per second was demonstrated. The large-vertical-displacement (LVD) microactuator was invented and desig The 1-D LVD micromirror was developed, and its bi-directional scanning and large piston motion capabilities at low actuation voltages was demonstrated. This device also generated optical scans greater than 170 at its resonance of 2.6 kH A 2-D LVD micromirror was developed, and static as well as dynamic 2-D bi-directional scanning was demonstrated. A LVD microlens scanner was fabricated and this sub-millimeter size pdevic voltage of 10 V. A millimeter-range LVD microlens scanner has been designed and fabricated, and it demonstrated a vertical displacement of 0.71 mm at 23 V. The millimeter-range LVD microlens scanner has been successfully packaged inside a 5-mm diameter endoscope for OCM imaging. Endoscopic OCT imaging was demonstrated using

PAGE 157

138 Nonlinear opticathe 1-D and 2-D t 1.5 m. Tdemonsw great potential for their intended application of in vivo developFcapilCT imagesnonlinewitnItD scanners that do not suffer from latethe micAive lumped element model for the LVD microactuation scheme sho actuatol the LVD piston mooptimizAll MEMS devices pusing variations of the DRIE CMOS-ercial volume production, the feasibility of wafer-level processing must be investigated. Since l endoscopy has been accomplished using micromirrors to obtain images with lateral resolution of abou 6.2 Future Work he optical scanners and the endoscopic biomedical imaging techniques trated in this research work sho detection of precancerous lesions. However to realize this end goal, additional ment and verification is required. rom an imaging standpoint, experimental demonstration is needed to verify the abity of the 2-D micromirrors for lateral scanning in OCT systems to obtain 3D O in real-time. Till date, the 2-D micromirrors have only been demonstrated for ar optical imaging. The mm-range LVD microlens scanner should be integrated h a existing OCT setup to demonstrate OCM imaging capability. is possible to design millimeter range LV ral shifting effects. The new design should be able to maintain the lateral position of rolens for the entire vertical displacement range. more comprehens uldbe developed. The existing static model is valid only when one of the two rs is electrically activated, and is therefore unsuitable to mode tion operation mode. A dynamic model is also needed which will help to design ed devices in the future. resented in this dissertation were fabricated at the die-level MEMS process. For comm

PAGE 158

139 elecns, present troic circuits are not required by these devices, a non-CMOS fabrication procesed in Appendix A, that uses SOI wafers can be easily developed.

PAGE 159

APPENDIX A NON-CMOS, WAFER LEVEL FABRICATION PROCESS The MEMS devices presented in this dissertation have been fabricated at the die level using variations of a post-CMOS micromachining process. Since on-chip electrical circuitry is not required for these devices, there is no added advantage in using the CMOS process. From a volume manufacturability point of view, these optical microscanners can be batch fabricated at the wafer level using only five lithography steps. The non-CMOS, wafer level fabrication process is presented in this appendix. Figure A-1 illustrates the proposed wafer-level fabrication process that uses the device layer of a silicon-on-insulator (SOI) wafer to define the single-crystal-silicon (SCS) thickness under the microstructure. This process uses only 5 lithographic steps, and can simultaneously produce both thin-film and bulk-Si microstructures. All devices presented in this dissertation can be fabricated using this non-CMOS process. Figure A-1: Non-CMOS, wafer-level fabrication process illustrating the steps required to fabricate the 1-D LVD micromirror device. Step 1: Deposit silicondioxide on SOI wafer SOI Device Layer SiO2 Silicon 140

PAGE 160

141 Step 2: Deposit platinum layer and pattern (Mask 1) Step 4: Pattern and etch SiO2 to open contact to buried Pt layer (Mask 2) Pt Step 3: Deposit silicondioxide layer : SiO2: Si : Pt Legend Figure A-1 Continued

PAGE 161

142 Step 5: Deposit aluminum layer and pattern (Mask 3) Al Figure A-1. Continued Step 6: Deposit another silicondioxide layer Step 7: Pattern and anisotropically etch SiO 2 till Si is exposed (Mask 4) : SiO2: Si : Pt Legend : Al

PAGE 162

143 Figure A-1. Continued Step 8: Pattern and etch backside Si to stop at the buried S iO2 layer (Mask 5) Step 10: Frontside anisotropic etch of Si using the top SiO2 layer as the etch mask : SiO2: Si : Pt Legend : Al Step 9: Backside etch of the buried SiO2 layer only

PAGE 163

144 Step 12: Isotropic etch of Si to form the thin-film bimorph actuator beams : SiO2: Si : Pt Legend : Al Step 11: Frontside timed dry etch of SiO2 to expose the Al layer SCS Mirror PlateBond Pad Mirror Actuator Frame Actuato r Frame Figure A-1. Continued

PAGE 164

APPENDIX B ARTICLES GENERATED BY THIS RESEARCH EFFORT The referred archival journal (J) and conference (C) publications generated by this research effort (for each individual micro-scanner) are listed below: 1-D Micromirror C1: A. Jain, S. T. Todd, G. K. Fedder, and H. Xie, A Large-Scanning-Angle, Electrothermal SCS Micromirror for Biomedical Imaging, Frontiers in Optics, C2: H. Xie, A. Jain, T. Xie, Y. Pan, and G. K. Fedder, A Single-Crystal Silicon-based Micromirrcanning AnBiomedical Applications, Conference nd Electro-Optics (CLEOaltimore, MD, June 2003. 2-D Micromirror The 87th OSA Annual Meeting, Tucson, AZ, October 2003. or with Large S gle for ) 2003, B on Lasers a J1: A. Jain, A. Kopa, Y. Pan, G. K. Fedder, and H. Xie, A Two-Axis Electrothermal Micromirror for Endoscopic Optical Coherence Tomography, IEEE Journal of Selected Topics in Quantum Electronics 10, pp. 636-642 (2004). C3: A. Jain, T. Xie, Y. Pan, G. K. Fednd H. Xie, A Two-Axis Electrothermal SCS Micromirror for Biomedical Imaging, 2003 IEEE/LEOS International Conference on Optical MEMS, Waikoloa, HI, August 2003. C4: A. Kopa, A. Jain, and H. Xie, Laser Scanning Display using a 2-D Micromirror, Optics in the South East (OISE) 2003, Orlando, FL, November 2003. 1-D Lmirror der, a VD Micro J2: A. Jain, H. Qu, S. Todd, and H. Xie, "A Thermal Bimorph Micromirror with Large Bi-Directional and Vertical Actuation," Sensors and Actuators A 122, pp. 9-15 C5: Jain, H. Qu, S. Todd, G. K. Fedder, and H. Xie, Electrothermal SCS Micromirror with Large-Vertical-Displacement Actuation, 2004 Solid-State ensor, Actuator and Microsystems Workshop, Hilton Head Isl., SC, June 2004. (2005). A S 145

PAGE 165

146 C6: A. Jain, S. Todd, and H. Xie, "An Electrothermally-Actuated, Dual-Mode Micromirror for Large Bi-Directional Scanning," IEEE International Electron Devices Meeting (IEDM) 2004, pp. 47-50, San Francisco, CA, December 2004. 2-D LVD Micromirror J3: A. Jain i-Directional 2-D Scanning Applications, Sensors and Actuators A, 2006 (In press). C7: A. Jain and H. Xie, An Electrothermal SCS Micromirror for Large Bi-Directional ors and Microsystems (Transducers), pp. 988-991, Seoul, Korea, June 2005. LVD Microlens Scanner and H. Xie, A Single-Crystal-Silicon Micromirror for Large B 2-D Scanning, 13th International Conference on Solid-State Sensors, Actuat J4: A. Jain and H. Xie, An Electrothermal Microlens Scanner with Low-Voltage, ers 17, C8: A. Jain and H. Xie, "A Tunable Microlens Scanner with Large-Vertical-C9: A. Jain and H. Xie, Half-Millimeter-Range Vertically Scanning Microlenses for Focusing Applications, 2006 Solid-State Sensor, Actuator and Microsystems Workshop, Hilton Head Isl., SC, June 2006. C10: n Optical MEMS, Big Sky, MT, August 2006. Biom Large-Vertical-Displacement Actuation, IEEE Photonics Technology Lettpp. 1971-1973 (2005). Displacement Actuation," 18th IEEE International Conference on Micro Electro Mechanical Systems (MEMS 2005), pp. 92-95, Miami, FL, January 2005. Microscopic A. Jain and H. Xie, Endoscopic Microprobe with a LVD Microlens Scanner for Confocal Imaging, 2006 IEEE/LEOS International Conference o edical Imaging J5: L. Fu, A. Jain, H. Xie, C. Cranfield, and M. Gu, Nonlinear Optical Endoscopy dler, D. Chan, A. Jain, H. Xie, Z. L. Wu, and Y. T. Pan, Cystoscopic Optical Coherence Tomography for Urinary Bladder Imaging in C12: Jain, H. Xie, C. Cranfield, and M. Gu, Integration of a Double-clad Photonic Crystal Fiber, a GRIN lens and a MEMS Mirror for Nonlinear Optical based on a Double-clad Photonic Crystal Fiber and a MEMS Mirror, Optics Express 14, pp. 1027-1032 (2006). C11: Z. G. Wang, H. A vivo, Proceedings of SPIE 6079, pp. 91-99, 2006. L. Fu, A Endoscopy, Biomedical Optics 2006, Ft. Lauderdale, FL, March 2006.

PAGE 166

147 Book Chapter H. Xie, S. Todd, and A. Jain, Single-Crystal Silicon Based Electrothermal MEM S Mirrors for Biomedical Imaging Applications, in MEMS/NEMS Handbook: pplications, edited by C. Leondes, Springer NY, 2005 (In press). Techniques and A

PAGE 167

LIST OF REFERENCES [1] American Cancer Society,"Cancer Facts and Figures 2005," Atlanta, GA 2005. [2] V. Backman, L. T. Perelman, J. T. Arendt, R. Gurjar, M. G. Muller, Q. Zhang, G. Zonios, E. Kline, T. McGillican, T. Valdez, J. V. Dam, M. Wallace, K. Badizadegan, J. M. Crawford, M. Fitzmaurice, S. Kabani, H. S. Levin, M. Seiler, R.R.Dasari, I. Itzkan, and M. S. Feld, "Detection of Preinvasive Cancer Cells," Nature, vol. 406, pp. 35-36, 2000. [3] B. J. Reid, P. L. Blount, Z. Feng, and D. S. Levine, "Optimizing Endoscopic Biopsy Detection of Early Cancers in Barrett's High-Grade Dysplasia," The American Journal of Gastroenterology, vol. 95, pp. 3089-3096, 2000. [4] W. Drexler, "Ultrahigh-Resolution Optical Coherence Tomography," Journal of Biomedical Optics, vol. 9, pp. 47-74, 2004. [5] C. Pitris, M. E. Brezinski, B. E. Bouma, G. J. Tearney, J. F. Southern, and J. G. Fujimoto, "High Resolution Imaging of the Upper Respiratory Tract with Optical Coherence Tomography: A Feasibility Study," American Journal Of Respiratory And Critical Care Medicine, vol. 157, pp. 1640-1644, 1998. [6] D. Huang, E. A. Swanson, C. P. Lin, J. S. Schuman, W. G. Stinson, W. Chang, M. R. Hee, T. Flotte, K. Gregory, C. A. Puliafito, and J. G. Fujimoto, "Optical Coherence Tomography," Science, vol. 254, pp. 1178-1181, 1991. [7] B. E. Bouma and G. J. Tearney, The Handbook of Optical Coherence Tomography. New York: Marcel Dekker, Inc., 2002. [8] J. M. Schmitt, "Optical Coherence Tomography (OCT): A Review," IEEE Journal of Selected Topics in Quantum Electronics, vol. 5, pp. 1205-1215, 1999. [9] I.-K. Jang, G. Tearney, and B. Bouma, "Visualization of Tissue Prolapse Between Coronary Stent Struts by Optical Coherence Tomography: Comparison with Intravascular Ultrasound," Circulation, vol. 104, pp. 2754, 2001. [10] W. Drexler, U. Morgner, F. X. Kartner, C. Pitris, S. A. Boppart, X. D. Li, E. P. Ippen, and J. G. Fujimoto, "In vivo Ultrahigh-Resolution Optical Coherence Tomography," Optics Letters, vol. 24, pp. 1221-1223, 1999. 148

PAGE 168

149 [11] C. Passmann and H. Ermert, "A 100-MHz Ultrasound Imaging System for Dermatologic and Ophthalmologic Diagnostics," IEEE Transactions on Ultrasonics, Ferroelectrics and Frequency Control, vol. 43, pp. 545-552, 1996. [12] G. J. Tearney, M. E. Brezinski, B. E. Bouma, S. A. Boppart, C. Pitris, J. F. Southern, and J. G. Fujimoto, "In Vivoal Biopsy with Optical Coherence Tomography," -2039, 1997. [14] Kuranov, T Imaging of Precancer and Cancer States of [15] ary of [16] earney, J. G. Fujimoto, and M. E. [17] eystone, CO, Aug. 5-[18] ches," Journal [19] S Micromirror for Optical Coherence iomedical [21] n, B. ner for 3-D Optical Coherence Tomography," Sensors and Actuators A, vol. 117, pp. 331-340, 2005. Endoscopic Optic Science, vol. 276, pp. 2037 [13]B. E. Bouma and G. J. Tearney, "Clinical Imaging with Optical Coherence Tomography," Academic Radiology, vol. 9, pp. 942-953, 2002. A. M. Sergeev, V. M. Gelikonov, G. V. Gelikonov, F. I. Feldchtein, R. VN. D. Gladkova, N. M. Shakhova, L. B. Snopova, A. V. Shakhov, I. A. Kuznetzova, A. N. Denisenko, V. V. Pochinko, Y. P. Chumakov, and O. S. Streltzova, "In vivo Endoscopic OC Human Mucosa," Optics Express, vol. 1, pp. 432-440, 1997. T.-Q. Xie, M. L. Zeidel, and Y.-T. Pan, "Detection of Tumorigenesis in UrinBladder with Optical Coherence Tomography: Optical Characterization Morphological Changes," Optics Express, vol. 10, pp. 1431-1443, 2002. S. A. Boppart, B. E. Bouma, C. Pitris, G. J. T Brezinski, "Forward-Imaging Instruments for Optical Coherence Tomography," Optics Letters, vol. 22, pp. 1618-1620, 1997. L. J. Hornbeck, "Digital Light Processing and MEMS: An Overview," Technical Digest of the IEEE/LEOS 1996 Summer Topical Meetings, K 9, pp. 7-8, 1996. D. S. Greywall, P. A. Busch, F. Pardo, D. W. Carr, G. Bogart, and H. T. Soh"Crystalline Silicon Tilting Mirrors for Optical Cross-Connect Swit of Microelectromechanical Systems, vol. 12, pp. 708-712, 2003. H. Xie, Y. Pan, and G. K. Fedder, "A SC Tomographic Imaging," Technical Digest of the 15th IEEE International Conference on Micro Electro Mechanical Systems (MEMS '02), Las Vegas, NV, Jan 20-24, pp. 495-499, 2002. [20]J. M. Zara, J. A. Izatt, K. D. Rao, S. Yazdanfar, and S. W. Smith, "Scanning Mirror for Optical Coherence Tomography Using an Electrostatic MEMS Actuator,"Technical Digest of the 2002 IEEE International Symposium on B Imaging, Washington, DC, Jul. 7-10, pp. 297300, 2002. J. T. W. Yeow, V. X. D. Yang, A. Chahwan, M. L. Gordon, B. Qi, I. A. VitkiC. Wilson, and A. A. Goldenberg, "Micromachined 2-D Scan

PAGE 169

150 [22] L. S. Fan, W. Piyawattanametha, M. C. Wu, A. D. Aguirre, P. R. Herz, Y. Cand J. G. Fujimoto, "High-resolution 3D OCT Imaging with a MEMS ScEndoscope," Proceedings of the SPIE, vol. 5719, pp. 140-143, 2005. hen, anning th rosystems (Transducers 2005), Seoul, Korea, Jun. 5-9, pp. 1644-1648, 2005. [24] "Electrostatic Micromachine Scanning Mirror for Optical Coherence Tomography," Optics [25] ical ," Optics Letters, vol. 29, pp. 2261-2263, 2004. [27] d iomedical Applications," Technical Digest of the Conference on Lasers and Electro-Optics (CLEO 2003), Baltimore, [28] d, G. K. Fedder, and H. Xie, "A Large-Scanning-Angle, Electrothermal SCS Micromirror for Biomedical Imaging," Technical Digest of the [29] CS Electrothermal Micromirror for Biomedical Imaging," Technical Digest of the 2003 ug. [30] Fedder, and H. Xie, "A Two-Axis Electrothermal Micromirror for Endoscopic Optical Coherence Tomography," IEEE Journal of [31] S n University, Pittsburgh, PA, 2002 [32] 158-160, 1987. [23] D. T. McCormick, W. Jung, Z. Chen, and N. C. Tien, "3-D MEMS Based Minimally Invasive Optical Coherence Tomography," Technical Digest of the 13International Conference on Solid-State Sensors, Actuators and Mic J. M. Zara, S. Yazdanfar, K. D. Rao, J. A. Izatt, and S. W. Smith Letters, vol. 22, pp. 628-630, 2003. P. H. Tran, D. S. Mukai, M. Brenner, and Z. Chen, "In vivo Endoscopic OptCoherence Tomography by use of a Rotational Microelectromechanical System Probe," Optics Letters, vol. 29, pp. 1236-1238, 2004. [26] P. R. Herz, Y. Chen, A. D. Aguirre, K. Schneider, P. Hsiung, J. G. Fujimoto, K. Madden, J. Schmitt, J. Goodnow, and C. Petersen, "Micromotor EndoscopeCatheter for invivo, Ultrahigh-Resolution Optical Coherence Tomography H. Xie, A. Jain, T. Xie, Y. Pan, and G. K. Fedder, "A Single-Crystal Silicon-BaseMicromirror with Large Scanning Angle for B MD, Jun. 1-6, 2003. A. Jain, S. T. Tod 2003 OSA Annual Meeting, Tuscon, AZ, Oct. 6, 2003. A. Jain, T. Xie, Y. Pan, G. K. Fedder, and H. Xie, "A Two-Axis S IEEE/LEOS International Conference on Optical MEMS, Waikoloa, Hawaii, A18-21, pp. 14-15, 2003. A. Jain, A. Kopa, Y. Pan, G. K Selected Topics in Quantum Electronics, vol. 10, pp. 636-642, 2004. H. Xie, Gyroscope and Micromirror Design Using Vertical Axis CMOS-MEMActuation and Sensing, Ph.D. Dissertation, Carnegie Mello R. C. Youngquist, S. Carr, and D. E. N. Davies, "Optical Coherence-Domain Reflectometry: A New Optical Evaluation Technique," Optics Letters, vol. 12, pp.

PAGE 170

151 [33] K. Takada, I. Yokohama, K. Chida, and J. Noda, "New Measurement System for Fault Location in Optical Waveguide Devices Based on an Interferometric Technique," Applied Optics, vol. 26, pp. 1603-1606, 1987. [34] P. Beaud, J. Schutz, W. Hodel, H. P. Weber, H. H. Gilgen, and R. P. Salath"Optical Reflectometry with Micrometer Resolution for the Investigation of Integrated Optical Devices," IEEE Journal of Quantum Electronics, vol. 25, pp755-759, 1989. e, opics 029-1034, 1996. 1-[38] ckson, and S. Dunne, "Three Dimensional OCT Images from Retina and Skin," Optics Express, vol. 7, pp. 292-298, 2000. [39] lization of Coronary Atherosclerotic Plaques in Patients Using Optical Coherence Tomography: Comparison With Intravascular Ultrasound," [40] J. G. Fujimoto, "Optical Coherence Microscopy in Scattering Media," Optics Letters, vol. 19, pp. 590-592, [41] icroscopy," Dermatology, vol. 191, pp. 93-98, 1995. s Express, vol. 3, pp. 239-250, 1998. [43] E. Ungersma, R. Wang, M. E. Williams, and S. E. Fraser, "An Optical Coherence Microscope for 3-Dimensional Imaging in Developmental Biology," [35] Y. Pan, E. Lankenau, J. Welzel, R. Birngruber, and R. Engelhardt, "Optical Coherence-Gated Imaging of Biological Tissues," IEEE Journal of Selected Tin Quantum Electronics, vol. 2, pp. 1 [36] J. M. Schmitt and A. Knuttel, "Model of Optical Coherence Tomography of Heterogeneous Tissue," Journal of Optical Society of America A, vol. 14, pp. 1231242, 1997. [37] J. Welzel, "Optical Coherence Tomography in Dermatology: A Review," Skin Research and Technology, vol. 7, pp. 1-9, 2001. A. G. Podoleanu, J. A. Rogers, D. A. Ja I.-K. Jang, B. E. Bouma, D.-H. Kang, S.-J. Park, S.-W. Park, K.-B. Seung, K.-B. Choi, M. Shishkov, K. Schlendorf, E. Pomerantsev, S. L. Houser, H. T. Artez, andG. J. Tearney, "Visua Journal of the American College of Cardiology, vol. 39, pp. 604-609, 2002J. A. Izatt, M. R. Hee, G. M. Owen, E. A. Swanson, and 1994. J. Schmitt, M. Yadlowsky, and R. Bonner, "Subsurface Imaging of Living Skin with Optical Coherence M [42] F. I. Feldchtein, V. M. Gelikonov, R. R. Iksanov, G. V. Gelikonov, R. V. Kuranov,A. M. Sergeev, N. Gladkova, M. N. Ourutina, D. H. Reitze, and J. A. Warren, "Invivo OCT Imaging of Hard and Soft Tissue of the Oral Cavity," Optic B. M. Hoeling, A. D. Fernandez, R. C. Haskell, E. Huang, W. R. Myers, D. C. Peterson, S Optics Express, vol. 6, pp. 136-146, 2000.

PAGE 171

152 [44] J. G. Fujimoto, M. E. Brezinski, G. J. Tearney, S. A. Boppart, B. Bouma, M. R. Hee, J. F. Southern, and E. A. Swanson, "Optical Biopsy and Imaging Using Optical Coherence Tomography," Nature Medicine, vol. 1, p p. 970-972, 1995. with 42, pp. [46] R. Windecker, M. Fleischer, and B. Franze, "Two Methods for Fast Coherence 8, [47] X. Li, C. Chudoba, T. Ko, C. Pitris, and J. G. Fujimoto, "Imaging Needle for 0. [48] o, and J. G. Fujimoto, "Intraluminal Fiber-Optic Doppler Imaging Catheter for Structural and Functional Optical Coherence Tomography," Optics [49] B. E. Bouma and G. J. Tearney, "Power-Efficient Nonreciprocal Interferometer and f -5, -241, 2003. Detection of Bladder Cancers," Applied Optics, vol. 42, pp. 6422-6426, 2003. [53] p. [45] J. M. Schmitt, S. L. Lee, and K. M. Yung, "An Optical Coherence MicroscopeEnhanced Resolving Power in Thick Tissue," Optics Communications, vol. 1203-207, 1997. Tomography and Topometry," Journal of Modern Optics, vol. 44, pp. 967-971997. Optical Coherence Tomography," Optics Letters, vol. 25, pp. 1520-1522, 200X. Li, T. H. K Letters, vol. 26, pp. 1906-1908, 2001. Linear Scanning Fiber-Optic Catheter for Optical Coherence Tomography," OpticsLetters, vol. 24, pp. 531-533, 1999. [50] X. D. Li, X. M. Liu, Y. C. Chen, M. J. Cobb, and M. B. Kimmey, "Development oa Fast Scanning Miniature Probe and Methods of Dispersion Management for High-Resolution Optical Coherence Tomography," Technical Digest of the 26th Annual International Conference of the IEEE EMBS, San Francisco, CA, Sep. 1pp. 5289-5291, 2004. [51] H. Xie, Y. Pan, and G. K. Fedder, "Endoscopic Optical Coherence Tomography Imaging with a CMOS-MEMS Micromirror," Sensors and Actuators A, vol. 103, pp. 237 [52] T. Xie, H. Xie, G. K. Fedder, and Y. Pan, "Endoscopic Optical Coherence Tomography with a Modified Microelectromechanical Systems Mirror for Y. Pan, H. Xie, and G. K. Fedder, "Endoscopic Optical Coherence Tomography Based on a Microelectromechanical Mirror," Optics Letters, vol. 26, pp. 1966-1968, 2001. [54] B. Qi, A. P. Himmer, L. M. Gordon, X. D. Yang, L. D. Dickensheets, and I. A. Vitkin, "Dynamic Focus Control in High-Speed Optical Coherence Tomography Based on a Microelectromechanical Mirror," Optics Communications, vol. 232, p123-128, 2004.

PAGE 172

153 [55] F. Lexer, C. K. Hitzenberger, W. Drexler, S. Molebny, H. Sattmann, M. Sticker, and A. F. Fercher, "Dynamic Coherent Focus OCT with Depth-Independent Transversal Resolution," Journal of Modern Optics, vol. 46, pp. 541-553, 1999. Selected Topics in Quantum Electronics, vol. 2, pp. 1017-1028, 1996. [57] ons, R. C. Haskell, D. C. Petersen, R. Wang, and J. I. Medford, "Optical Coherence Microscopy. A Technology for Rapid, in vivo, Non-Destructive 2000. [58] A. D. Aguirre, P. Hsiung, T. H. Ko, I. Hartl, and J. G. Fujimoto, "High-Resolution [59] Z. Ding, H. Ren, Y. Zhao, J. S. Nelson, and Z. Chen, "High-resolution Optical [60] S. Kwon, V. Milanovic, and L. P. Lee, "Vertical Microlens Scanner for 3D [61] y," Technical Digest of the 15thIEEE International Conference on Micro Electro Mechanical Systems (MEMS 2002), Las Vegas, NV, [62] vic, and L. P. Lee, "Large-Displacement Vertical Microlens Scanner with Low Driving Voltage," IEEE Photonics Technology Letters, vol. 14, [63] 1. [65] W. R. Zipfel, R. M. Williams, and W. W. Webb, "Nonlinear Magic: Multiphoton [66] trickler, and W. W. Webb, "Two-Photon Laser Scanning Fluorescence Microscopy," Science, vol. 248, pp. 73-76, 1990. [56] J. A. Izatt, M. D. Kulkarni, H.-W. Wang, K. Kobayashi, and J. M. V. Sivak, "Optical Coherence Tomography and Microscopy in Gastrointestinal Tissues," IEEE Journal of J. W. Hettinger, M. d. l. P. Mattozzi, W. R. Myers, M. E. Williams, A. Reeves, RL. Pars Visualization of Plants and Plant Cells," Plant Physiology, vol. 123, pp. 3-16, Optical Coherence Microscopy for High-Speed, in vivo Cellular Imaging," OpticsLetters, vol. 28, pp. 2064-2066, 2003. Coherence Tomography Over a Large Depth Range with an Axicon Lens," Optics Letters, vol. 27, pp. 243-245, 2002. Imaging," Technical Digest of the 2002 Solid-State Sensor and Actuator Workshop, Hilton Head Isl., SC, Jun. 2-6, pp. 227-230, 2002. S. Kwon and L. P. Lee, "Stacked Two Dimensional Micro-Lens Scanner for Micro Confocal Imaging Arra Jan. 20-24, pp. 483-486, 2002. S. Kwon, V. Milano pp. 1572-1574, 2002. D. F. Eaton, "Nonlinear Optical Materials," Science, vol. 253, pp. 281-287, 199 [64] A. Diaspro, Confocal and Two-Photon Microscopy: Foundations, Applications, and Advances. New York: Wiley-Liss, 2002. Microscopy in the Biosciences," Nature Biotechnology, vol. 21, pp. 1369-1377, 2003. W. Denk, J. H. S

PAGE 173

154 [67] G. Marriott and I. Parker, "Biophotonics, Part B," in Methods in Enzymology, vol361, J. N. Abelson and M. I. Simon, Eds. San Diego: Academic Press, 2003. [68] P. J. Campagnola and L. M. Loew, "Second-Harmonic Imaging Microscopy for [69] kala, J. M. Squirrell, K. M. Vrotos, J. C. Eickhoff, A. Gendron-Fitzpatrick, K. W. Eliceiri, and N. Ramanujam, "Multiphoton Microscopy of Endogenous [70] pp. 902-904, 2003. [71] tion Brain Imaging in Freely Moving Animals," Neuron, vol. 31, pp. 903-912, 2001. [72] 3.9 gram Two-Photon Fluorescence Microendoscope," Optics Letters, vol. 30, pp. 2272-2274, 2005. [73] Twoand a Gradient-Index Lens Objective," Optics Letters, vol. 29, pp. 2521-2523, 2004. [74] tic [75] X. Li, "Fiber-Optic Scanning Two-Photon [76] C. Cranfield, and M. Gu, "Nonlinear Optical Endoscopy ystal Fiber and a MEMS Mirror," Optics EMS Mirror for Nonlinear Optical Endoscopy," Technical Digest of the Biomedical Optics 2006, Ft. Lauderdale, FL, [78] awattanametha, R. P. J. Barretto, T. H. Ko, B. A. Flusberg, E. D. Cocker, H. Ra, D. Lee, O. Solgaard, and M. J. Schnitzer, "Fast-Scanning Two-Photon -2020, 2006. Visualizing Biomolecular Arrays in Cells, Tissues and Organisms," Nature Biotechnology, vol. 21, pp. 1356-1360, 2003. M. C. S Fluorescence Differentiates Normal, Precancerous, and Cancerous Squamous Epithelial Tissues," Cancer Research, vol. 65, pp. 1180-1186, 2005. J. C. Jung and M. J. Schnitzer, "Multiphoton Endoscopy," Optics Letters, vol. 28, F. Helmchen, M. S. Fee, D. W. Tank, and W. Denk, "A Miniature Head-Mounted Two-Photon Microscope: High-Resolu B. A. Flusberg, J. C. Jung, E. D. Cocker, E. P. Anderson, and M. J. Schnitzer, "In vivo Brain Imaging Using a Portable W. Gobel, J. N. D. Kerr, A. Nimmerjahn, and F. Helmchen, "MiniaturizedPhoton Microscope Based on a Flexible Coherent Fiber Bundle D. Bird and M. Gu, "Two-Photon Fluorescence Endoscopy with a Micro-OpScanning Head," Optics Letters, vol. 28, pp. 1552-1554, 2003. M. T. Myaing, D. J. MacDonald, and Fluorescence Endoscope," Optics Letters, vol. 31, pp. 1076-1078, 2006. L. Fu, A. Jain, H. Xie, Based on a Double-Clad Photonic Cr Express, vol. 14, pp. 1027-1032, 2006. [77] L. Fu, A. Jain, H. Xie, C. Cranfield, and M. Gu, "Integration of a Double-clad Photonic Crystal Fiber, a GRIN Lens and a M Mar. 19-22, 2006. W. Piy Fluorescence Imaging Based on a Microelectromechanical Systems Two-Dimensional Scanning Mirror," Optics Letters, vol. 31, pp. 2018

PAGE 174

155 [79] J. M. Bustillo, R. T. Howe, and R. S. Muller, "Surface Micromachining for Microelectromechanical Systems," Proceedings of the IEEE, vol. 86, pp. 1552 -1574, 1998. [80] 1536-1551, 1998. 998. [82] and R. S. Muller, "A Raster-Scanning Full-Motion Video Display Using Polysilicon [83] T. W. Yeow, K. Y. Lim, B. Wilson, and A. A. Goldenberg, "A Low-Voltage Biology, Madison, WI, May 2-4, pp. 205-207, 2002. [84] rs and Actuators A, vol. 85, pp. 424-429, 2000. [85] e-urnal of Microelectromechanical Systems, vol. 7, pp. 27-37, 1998. [86] irkos, J. E. Odhner, A. Pareek, M. Waelti, and M. Zai, "Electromagnetically Actuated Mirror Arrays for nce EMS 2003), Kyoto, Japan, Jan. 19-23, [88] al national Conference on Micro Electro Mechanical Systems [89] "Large n pp. 715-722, 2000. G. T. A. Kovacs, N. I. Maluf, and K. E. Petersen, "Bulk Micromachining of Silicon," Proceedings of the IEEE, vol. 86, pp. [81] P. F. V. Kessel, L. J. Hornbeck, R. E. Meier, and M. R. Douglass, "A MEMS-Based Projection Display," Proceedings of the IEEE, vol. 98, pp. 1687-1704, 1R. A. Conant, P. Hagelin, U. Krishnamoorthy, O. Solgaard, K. Y. Lau Micromachined Mirrors," Sensors and Actuators A, vol. 83, pp. 291-296, 2000. Electrostatically Actuated MEMS Scanner for in vivo Biomedical Imaging," Technical Digest of the 2nd Annual International IEEE-EMB Special Topic Conference on Microtechnologies in Medicine & S. Schweizer, P. Cousseau, G. Lammel, S. Calmes, and P. Renaud, "Two-Dimensional Thermally Actuated Optical Microprojector," Senso M.-H. Kiang, O. Solgaard, K. Y. Lau, and R. S. Muller, "Electrostatic CombdrivActuated Micromirrors for Laser-Beam Scanning and Positioning," Jo J. J. Bernstein, W. P. Taylor, J. D. Brazzle, C. J. Corcoran, G. K use in 3-D Optical Switching Applications," Journal of Microelectromechanical Systems, vol. 13, pp. 526535, 2004. [87] A. Baba, H. Okano, H. Uetsuka, and M. Esashi, "2 Axes Optical Switch with Holding Mechanism," Technical Digest of the 16th IEEE International Confereon Micro Electro Mechanical Systems (M pp. 251-254, 2003. S. J. Kim, Y. H. Cho, H. J. Nam, and J. U. Bu, "Piezoelectrically Pushed RotationMicromirrors for Wide-Angle Optical Switch Applications," Technical Digest of the 16th IEEE Inter (MEMS 2003), Kyoto, Japan, Jan. 19-23, pp. 263-266, 2003. H. Schenk, P. Durr, T. Haase, D. Kunze, U. Sobe, H. Lakner, and H. Kuck,Deflection Micromechanical Scanning Mirrors for Linear Scans and PatterGeneration," IEEE Journal of Selected Topics in Quantum Electronics, vol. 6,

PAGE 175

156 [90] G.-D. Su, H. Toshiyoshi, and M. C. Wu, "Surface-Micromachined 2-D OpticScanners with High-Performance Single-Crystalline Silicon Micromirrors," IEEPhotonics Te al E chnology Letters, vol. 13, pp. 606-608, 2001. ar Digest of the 1998 IEEE/LEOS Summer Topical Meetings, Monterey, CA, Jul. 20-24, pp. [92] T. Kudrle, C. Wang, M. Bancu, J. Hsiao, A. Pareek, M. Waelti, G. Kirkos, T. 2005. [93] M. R. Dokmeci, A. Pareek, S. Bakshi, M. Waelti, C. D. Fung, K. H. Heng, and C. nal of [94] ulk-Micromachined Two-Dimensional Electrostatic Torsion Micromirror," Journal of Microelectromechanical Systems, vol. 12, pp. 692-701, [95] icated Scanning Micromirrors Using Localized Plastic Deformation of Silicon," Technical Digest of the 17th IEEE International [96] R. A. Conant, J. T. Nee, K. Y. Lau, and R. S. Muller, "A Flat High-Frequency [97] MOS-MEMS Mirror with Curled-Hinge Comb Drive," Journal of Microelectromechanical Systems, vol. 12, pp. 450-457, [98] or with Angular Comb Drive Actuation," Technical Digest of the 15th IEEE International Conference on Micro Electro Mechanical Systems [99] ic tromechanical Systems, vol. 12, pp. 458-464, 2003. [100] 247, 2003. [91] L. Fan and M. C. Wu, "Two-Dimensional Optical Scanner with Large AngulRotation Realized by Self-Assembled Micro-Elevator," Technical II/107-II/108, 1998. Shone, C. Fung, and C. Mastrangelo, "Single-Crystal Silicon Micromirror Array with Polysilicon Flexures," Sensors and Actuators A, vol. 119, pp. 559-566, H. Mastrangelo, "Two-Axis Single-Crystal Silicon Micromirror Arrays," JourMicroelectromechanical Systems, vol. 13, pp. 1006-1017, 2004. Z. Hao, B. Wingfield, M. Whitley, J. Brooks, and J. A. Hammer, "A Design Methodology for a B 2003. J. Kim and L. Lin, "Batch-Fabr Conference on Micro Electro Mechanical Systems (MEMS 2004), Maastricht, TheNetherlands, 25-29 Jan., pp. 494 497, 2004. Scanning Micromirror," Technical Digest of the 2000 Solid-State Sensor & Actuator Workshop, Hilton Head, SC, Jun. 2-6, pp. 6-9, 2000. H. Xie, Y. Pan, and G. K. Fedder, "A C 2003. P. R. Patterson, D. Hah, N. Hung, H. Toshiyoshi, R. Chao, and M. C. Wu, "A Scanning Micromirr (MEMS 2002), Jan. 20-24, pp. 544-547, 2002. U. Krishnamoorthy, D. Lee, and O. Solgaard, "Self-Aligned Vertical ElectrostatCombdrives for Micromirror Actuation," Journal of Microelec V. Milanovic, M. Last, and K. S. J. Pister, "Laterally Actuated Torsional Micromirrors for Large Static Deflection," IEEE Photonics Technology Letters, vol. 15, pp. 245

PAGE 176

157 [101] V. Milanovic, G. A. Mathus, and D. T. McCormick, "Gimbal-Less Monolithic Silicon Actuators for Tip-Tilt-Piston Micromirror Applications," IEEE Journal of Selected Topics in Quantum Electronics, vol. 10, pp. 462-47 1, 2004. 2D-9, [103] S. Kwon, V. Milanovic, and L. P. Lee, "Vertical Combdrive Based 2-D Gimbaled [104] W. Piyawattanametha, P. R. Patterson, D. Hah, H. Toshiyoshi, and M. C. Wu, "A [106] mel, S. Schweizer, and P. Renaud, Optical Microscanners and Microspectrometers Using Thermal Bimorph Actuators. Boston: Kluwer [107] res for Medical Applications," Sensors and Actuators A, vol. 31, pp. 29-34, 1992. [108] ptical rs and Actuators A, vol. 706, pp. 470-477, 1999. [109] on a Porous Silicon Tunable Filter," Technical Digest of the The 14th IEEE International Conference on Micro Electro Mechanical Systems (MEMS 2001), [110] pp. 758-763, 1988. 4, pp. 1036-1041 vol.2, 2002. ion," 2002), Las Vegas, NV, Jan. 20-24, pp. 698-701, 2002. [102] H. Schenk, P. Durr, D. Kunze, H. Lakner, and H. Kuck, "A Resonantly Excited Micro-Scanning-Mirror with Large Deflection," Sensors and Actuators A, vol. 8pp. 104-111, 2001. Micromirrors with Large Static Rotation by Backside Island Isolation," IEEE Journal of Selected Topics in Quantum Electronics, vol. 10, pp. 498-504, 2004. 2-D Scanner by Surface and Bulk Micromachined Angular Vertical Comb Actuators," Technical Digest of the 2003 IEEE/LEOS International Conference on Optical MEMS, Waikoloa, Hawaii, Aug. 18-21, pp. 93-94, 2003. [105] J. Bhler, J. Funk, O. Paul, F.-P. Steiner, and H. Baltes, "Thermally ActuatedCMOS Micromirrors," Sensors and Actuators A, vol. 46-47, pp. 572-575, 1995. G. Lam Academic, 2002. R. Buser, N. F. d. Rooij, H. Tischauser, A. Dommann, and G. Staufert, "Biaxial Scanning Mirror Activated by Bimorph Structu S. Schweizer, S. Calmes, M. Laudon, and P. Renaud, "Thermally Actuated OMicroscanner with Large Angle and Low Consumption," Senso G. Lammel, S. Schweizer, and P. Renaud, "MEMS Infrared Gas Spectrometer Based Interlaken, Switzerland, Jan. 21-25, pp. 578-581, 2001. W. Riethmuller and W. Benecke, "Thermally Excited Silicon Microactuators," IEEE Transactions on Electron Devices, vol. 35 [111] Q. Liu and Q.-A. Huang, "Micro-Electro-Mechanical Digital-to-Analog Converter Based on a Novel Bimorph Thermal Actuator," Technical Digest of the IEEE Sensors 2002, Orlando, FL, Jun. 12-1 [112] M. Sinclair, "A High Frequency Resonant Scanner Using Thermal ActuatTechnical Digest of the 15th IEEE International Conference on Micro Electro Mechanical Systems (MEMS

PAGE 177

158 [113] J. W. Judy and R. S. Muller, "Magnetically Actuated, Addressable Microstructures," Journal of Microelectromechanical Systems, vol. 6, pp. 249-256,1997. echnical Digest of the 15th IEEE International Conference on Micro Electro Mechanical Systems (MEMS 2002), Las Vegas, NV, Jan. 20-24, pp. [115] ai, "Two Axis-of-Rotation Mirror Array Using Electromagnetic MEMS," Technical [116] ning Mirror," Optical Engineering, vol. 36, pp. 1399-1407, 1997. [117] al Micro Scanner Integrated with PZT Thin Film Actuator," Technical Digest of the ransducers [118] N. Kikuchi, Y. Haga, M. Maeda, W. Makishi, and M. Esashi, "Piezoelectric 2-D d-ucers 03), Boston, MA, Jun. 8-[119] with Hinge Structure," Technical Digest of the 2002 IEEE/LEOS International Conference on Optical MEMS, Lugano, Switzerland, [120] f ternational Conference on Solid-State Sensors and Actuators (Transducers 1995 and Eurosensors IX), [122] H. Xie, L. Erdmann, X. Zhu, K. Gabriel, and G. K. Fedder, "Post-CMOS f [114] I. J. Cho, K. S. Yun, H. K. Lee, J. B. Yoon, and E. Yoon, "A Low-Voltage Two-Axis Electromagnetically Actuated Micromirror with Bulk Silicon Mirror Plates and Torsion Bars," T 540-543, 2002. J. Bernstein, W. P. Taylor, J. Brazzle, G. Kirkos, J. Odhner, A. Pareek, and M. Z Digest of the 16th IEEE International Conference on Micro Electro Mechanical Systems (MEMS 2003), Kyoto, Japan, Jan. 19-23, pp. 275278, 2003. R. A. Miller and Y.-C. Tai, "Micromachined Electromagnetic Scan T. Kawabata, M. Ikeda, H. Goto, M. Matsumoto, and T. Yada, "The 2-Dimension 1997 International Conference on Solid State Sensors and Actuators (T'97), Chicago, IL, Jun. 16-19, pp. 339-342, 1997. Micro Scanner for Minimally Invasive Therapy Fabricated Using Femtosecond Laser Ablation," Technical Digest of the 12th International Conference on SoliState Sensors, Actuators and Microsystems (Transd 12, pp. 603606, 2003. H.-J. Nam, Y.-S. Kim, S.-M. Cho, Y. Yee, and J.-U. Bu, "Low Voltage PZT Actuated Tilting Micromirror Aug. 20-23, pp. 89-90, 2002. G. Lin, C.-J. Kim, S. Konishi, and H. Fujita, "Design, Fabrication, And Testing OA C-shape Actuator," Technical Digest of the The 8th In Stockholm, Sweden, Jun. 25-29, pp. 416-419, 1995. [121] M. Ataka, A. Omodaka, N. Takeshima, and H. Fujita, "Fabrication and Operation of Polyimide Bimorph Actuators for a Ciliary Motion System," Journal of Microelectromechanical Systems, vol. 2, pp. 146-150, 1993. Processing for High-Aspect-Ratio Integrated Silicon Microstructures," Journal oMicroelectromechanical Systems, vol. 11, pp. 93-101, 2002.

PAGE 178

159 [123] H. Qu, D. Fang, and H. Xie, "A Single-Crystal Silicon 3-Axis CMOAccelerometer," Technical Digest of the IEEE Sensors 2004, Vienna, Austria, Oct. 24-27, S-MEMS pp. 661-664, 2004. [125] [126] Anisotropically Etching Silicon." US-Patent [127] [129] [131] ive-Optics Arrays," Technical Digest of the 11th 01 [132] b Drive with In Situ Capacitive Position [133] tial Sensors and [134] hSelf-Aligned Dual[124] H. Xie and G. K. Fedder, "Fabrication, Characterization, and Analysis of a DRIECMOS-MEMS Gyroscope," IEEE Sensors Journal, vol. 3, pp. 622-631, 2003. MOSIS Integrated Circuit Fabrication Service, Marina del Rey, CA. F. Laermer and A. Schilp, "Method of No. 5501893: Robert Bosch GmbH (Stuttgart, DE), March 26, 1996. J. Bhler, Deformable Micromirror Arrays by CMOS Technology, Ph.D. Thesis,The Swiss Federal Institute of Technology, 1997 [128] S. T. Todd and H. Xie, "An Analytical Electrothermal Model of a 1-D Electrothermal MEMS Micromirror," Proceedings of the SPIE, vol. 5649, pp. 344-353, 2004. A. Kopa, A. Jain, and H. Xie, "Laser Scanning Display Using a 2-D Micromirror,"Technical Digest of the Optics in the Southeast (OISE) 2003, Orlando, FL, Nov. 12-13, 2003. [130] Z. G. Wang, H. Adler, D. Chan, A. Jain, H. Xie, Z. L. Wu, and Y. T. Pan, "Cystoscopic Optical Coherence Tomography for Urinary Bladder Imaging in vivo," Proceedings of the SPIE, vol. 6079, pp. 91-99, 2006. M. A. Helmbrecht, U. Srinivasan, C. Rembe, R. T. Howe, and R. S. Muller, "Micromirrors for Adapt International Conference on Solid State Sensors and Actuators (Transducers and Eurosensors XV), Munich, Germany, Jun. 10-14, 2001. A. P. Lee, C. F. McConaghy, G. Sommargren, P. Krulevitch, and E. W. Campbell, "Vertical-Actuated Electrostatic Com Correction for Application in Phase Shifting Diffraction Interferometry," Journal ofMicroelectromechanical Systems, vol. 12, pp. 960-971, 2003. S. W. Chung and Y. K. Kim, "Design and Fabrication of 10 x 10 Micro-SpaLight Modulator Array for Phase and Amplitude Modulation," Actuators A, vol. 78, pp. 63-70, 1999. D. Lee, U. Krishnamoorthy, K. Yu, and O. Solgaard, "High-Resolution, HigSpeed Microscanner in Single-Crystalline Silicon Actuated by Mode Vertical Electrostatic Combdrive with Capability for Phased Array Operation," Technical Digest of the 12th International Conference on Solid State Sensors, Actuators and Microsystems (Transducers '03), Boston, MA, Jun. 8-12, pp. 576-579, 2003.

PAGE 179

160 [135] O. Cugat, P. Mounaix, S. Basrour, C. Divoux, and G. Reyne, "Deformable Magnetic Mirror for Adaptive Optics: First Results," Technical Digest of the 13th IEEE Annual International Conference on Micro Electro Mechanical Systems (MEMS 2000), Miyazaki, Japan, Jan. 23-27, pp. 485-490, 2000. [136] ted Micromirror for Nano-Tracking of Laser Beam for High-Density Optical Data ce on Micro [137] Phase-Only Sensors and Actuators CoventorWare 2003 Reference Manual[140] e l and Vertical Actuation," Sensors and Actuators A, 2005. f the op, Hilton Head Isl., SC, Jun. 6-10, pp. 228-231, 2004. [142] San Francisco, CA, [143] nal ), Seoul, Korea, Jun. [144] of the The 14th IEEE International [145] egrated tors and Microsystems (Transducers ), Boston, MA, Jun. 8-12, pp. 607-610, 2003. Y. Yee, H.-J. Nam, S.-H. Lee, J. U. Bu, Y.-S. Jeon, and S.-M. Cho, "PZT Actua Storage," Technical Digest of the 13th Annual International Conferen Electro Mechanical Systems (MEMS 2000), Jan. 23-27, pp. 435-440, 2000. A. Tuantranont, L. A. Liew, V. M. Bright, W. Zhang, and Y. C. Lee, Micromirror Array Fabricated by Standard CMOS Process," A, vol. 89, pp. 124-134, 2001. [138] Z.-L. Wan, A. Feinerman, H.-J. Zeng, and G. Friedman, "Electrocapillary Piston Motion and a Prototype of Phase-Manipulating Micromirror," Journal of Microelectromechanical Systems, vol. 13, pp. 620-627, 2004. [139] Cary, NC: Coventor, Inc., 2003. A. Jain, H. Qu, S. Todd, and H. Xie, "A Thermal Bimorph Micromirror with LargBi-Directiona [141] A. Jain, H. Qu, S. Todd, G. K. Fedder, and H. Xie, "Electrothermal SCS Micromirror with Large-Vertical-Displacement Actuation," Technical Digest o2004 Solid-State Sensor, Actuator and Microsystems Worksh A. Jain, S. Todd, and H. Xie, An Electrothermally-Actuated, Dual-ModeMicromirror for Large Bi-Directional Scanning," Technical Digest of the 2004 IEEE International Electron Devices Meeting (IEDM 2004), Dec. 13-15, pp. 47-50, 2004. A. Jain and H. Xie, "An Electrothermal SCS Micromirror for Large Bi-Directio2-D Scanning," Technical Digest of the 13th International Conference on Solid-State Sensors, Actuators and Microsystems (Transducers 2005 5-9, pp. 988-991, 2005. K. Hoshino and I. Shimoyama, "An Elastic Thin-Film Microlens Array withPneumatic Actuator," Technical Digest Conference on Micro Electro Mechanical Systems (MEMS 2001), Interlaken, Switzerland, Jan. 21-25, pp. 321-324, 2001. S.-H. Kim, Y. Yee, J. Choi, H. Kwon, M.-H. Ha, C. Oh, and J. U. Bu, "IntMicro Optical Flying Head with Lens Positioning Actuator for Small Form FactorData Storage," Technical Digest of the 12th International Conference on Solid-State Sensors, Actua

PAGE 180

161 [146] R. Graham, "A Variable Focus Lens and Its Uses," J. O. S. A., vol. 30, pp. 51940. N. Sugiura and S. Morita, "Variable-Focus Liquid-Filled Optical 60-563, [147] Lens," Applied Optics, vol. 32, pp. 4181-4186, 1993. [148] age: echanical Systems (MEMS 2005), Miami Beach, FL, Jan. 30 Feb. 3, pp. 227-230, 2005. [150] d Physics age Large-Vertical-Displacement Actuation," IEEE Photonics Technology Letters, vol. [152] A. Jain and H. Xie, "A Tunable Microlens Scanner with Large-Vertical-l [153] D. Daly, Microlens Arrays. London: Taylor & Francis, 2001. [154] vol. 16, pp. 1107-1109, 2004. B. Berge and J. Peseux, "Variable Focal Length Controlled by an External VoltAn Application of Electrowetting," The European Physical Journal E, vol. 3, pp. 159-163, 2000. [149] B. Berge, "Liquid Lens Technology: Principle of Electrowetting Based Lenses and Applications to Imaging," Technical Digest of the 18th IEEE International Conference on Micro Electro M T. Krupenkin, S. Yang, and P. Mach, "Tunable Liquid Microlens," ApplieLetters, vol. 82, pp. 316-318, 2003. [151] A. Jain and H. Xie, "An Electrothermal Microlens Scanner with Low-Volt 17, pp. 1971-1973, 2005. Displacement Actuation," Technical Digest of the 18th IEEE InternationaConference on Micro Electro Mechanical Systems (MEMS 2005), Miami Beach, FL, Jan. 30 Feb. 3, pp. 92-95, 2005. A. L. Glebov, L. Huang, S. Aoki, M. Lee, and K. Yokouchi, "Planar Hybrid Polymer-Silica Microlenses with Tunable Beamwidth and Focal Length," IEEE Photonics Technology Letters [155] A. Jain and H. Xie, "Half-Millimeter-Range Vertically Scanning Microlenses for Microscopic Focusing Applications," Technical Digest of the 2006 Solid State Sensors, Actuators and Microsystems Workshop, Hilton Head Isl., SC, pp. 74-77, 2006.

PAGE 181

lled at the Birla Institute of Technology and Science, Pilani, India, As anan optoelectronic arrayed waveguide multiplexer, and also the design of an infrared transcmunication. He completed his bachelors gradu. degree in electrical and computer engineering from the semic2 to pursue a Ph.D. biome carbo and he has also co-authored a book chapter in the MEMS/NEMS Handbook. He is a member of the Institute of Electrical and Electronics Engineers and the Optical Society of America, and is also a member of the Eta Kappa Nu and Tau Beta Pi honor societies. After receiving his PhD degree in 2006, he plans to pursue a career in the area of optical microsystems. BIOGRAPHICAL SKETCH Ankur Jain enro in the fall of 1996 for the B.E. (Honors) degree in electrical and electronics engineering. undergraduate at BITS, he participated in various projects including the design of eiver system for indoor wireless com degree in 2000, and was awarded the Motorola Student of the Year Gold Medal uponation. He received the M.S University of Florida in 2002 where he specialized in the areas of photonics, onductor device theory, and computer systems and networking. Ankur joined the Interdisciplinary Microsystems Group in 200 degree that involved the development of optical MEMS scanners for endoscopic dical imaging systems. His research interests include optical microsystems, endoscopic biomedical imaging, photonic devices, CMOS-MEMS microfabrication, andn nanotubes. His doctoral research has contributed to over 20 research publications, 162


Permanent Link: http://ufdc.ufl.edu/UFE0015728/00001

Material Information

Title: Low Voltage, MEMS-Based Reflective and Refractive Optical Scanners for Endoscopic Biomedical Imaging
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0015728:00001

Permanent Link: http://ufdc.ufl.edu/UFE0015728/00001

Material Information

Title: Low Voltage, MEMS-Based Reflective and Refractive Optical Scanners for Endoscopic Biomedical Imaging
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0015728:00001


This item has the following downloads:


Full Text












LOW VOLTAGE, MEMS-BASED REFLECTIVE AND REFRACTIVE OPTICAL
SCANNERS FOR ENDOSCOPIC BIOMEDICAL IMAGING















By

ANKUR JAIN


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2006

































Copyright 2006

by

Ankur Jain

































To my parents, Ranjan and Poonam, to my brother Prateek, and to my fiancee Kavitha for
their constant love, unwavering support, confidence and encouragement.















ACKNOWLEDGMENTS

I would like to thank my advisor, Dr. Huikai Xie, for the constant support and

guidance he has given me over the past few years. I first met Huikai in August 2002, and

subsequently joined his research group as a PhD student in the fall semester. I am grateful

for all the insight he has provided, and am thankful to him for introducing me to the areas

of MEMS and endoscopic biomedical imaging. I have personally gained technical

expertise, as well as professional know-how through my interactions with him, and I will

forever be indebted to him for mentoring me towards becoming a better microsystems

technology engineer. The research presented in this dissertation was also painstakingly

reviewed by other members of my PhD committee, Dr. Toshikazu Nishida, Dr. Ramakant

Srivastava, and Dr. William Ditto, and for that I am grateful. I have enjoyed many

conversations with Dr. Nishida, both personally and professionally. Working as a

graduate teaching assistant for Dr. Srivastava was a pleasure, and I am grateful for our

personal friendship. Dr. Ditto has provided me with unique insights related to the

biomedical application aspect of this project.

I want to acknowledge technical and personal discussions with Dr. Mark Sheplak

and Dr. David Arnold, as their advice helped improve my research work and their

pleasant company at MEMS conferences is always welcome. Further appreciation goes

out to Dr. Peter Zory for his valuable friendship, for always being a mentor, and for his

indispensable lessons on how to maintain a good research notebook.









The Biophotonics and Microsystems Lab (BML) located in 136 Larsen Hall was

home to this project, and I am indebted to my BML group members. Special thanks go

out to Hongwei Qu for teaching me the ropes in the cleanroom, and to Shane Todd for

helping me with electrothermal modeling. Hongwei has always been a pillar of support

within BML, and I enjoyed working with him on various projects. Shane worked with me

on various micromirror projects, and I have benefited greatly from our professional

interactions and personal friendship. I would like to acknowledge support from Anthony

Kopa, both personally and also for using my 2-D micromirror for imaging purposes.

Other BML members who aided me during the course of my research include Deyou

Fang, Maojiao He, Mi Huang, Mingliang Wang, Ben Caswell and the newcomers

Kemiao "Alex" Jia and Lei Wu. Alex and Lei have proven to be worthy successors for

my project, and I will value their camaraderie. All BML members work great as a team,

and I have so many good memories about the multilingual jokes told in the lab, and the

parties and sports that we all participated in.

BML is just part of the much bigger microsystems group at the University of

Florida, known as the Interdisciplinary Microsystems Group (IMG). I am grateful to all

IMG members for their support-group-like environment and technical expertise. In

particular, I would like to thank my colleagues Venkat Chandrasekaran, Stephen

Horowitz, Anurag Kasyap, Chris Bahr, David Martin, Ryan Holman, Erin Patrick, Israel

Boniche, Janhavi Agashe, Sheetal Shetye, Jian "Jackie Chan" Liu, Yawei Li, Vijay

Chandrasekharan, Lee Hunt, Tai-An Chen, Karthik Kadirvel, Robert Taylor, Brandon

Bertolucci, Champak Das and Zheng Xia, to name a few. Venkat introduced me to the

world of wire-bonding, while Dave, Ryan and Chris kept our IMG server running 24/7.









Thanks go to Anurag for helping me with the vibrometer, to Erin for help with the PCB

milling machine, and to Israel and Janhavi for AutoCAD assistance. I thank Brandon and

the Ultimate Frisbee gang for relentlessly organizing sporting events that helped to

upkeep the morale of IMG. Finally, credit is due to all other IMG members for general

technical assistance, and for maintaining a lively work environment in the office, lab and

even inside the cleanrooms.

This work would not be complete without help from our external collaborators. I

thank Dr. Yingtian Pan and Zhenguo Wang at the State University of New York at Stony

Brook for validating the use of my micromirrors for endoscopic optical coherence

tomographic (OCT) imaging. I am grateful that they invited me to visit their lab so that I

could witness OCT imaging using my endoscopically-packaged micromirrors. Two-

photon excitation fluorescence imaging and second harmonic generation nonlinear

optical imaging experiments using my micromirrors were demonstrated in collaboration

with Dr. Min Gu and Ling Fu at the Swinburne University of Technology, Australia. I

thank Ling for the many hours she has put into this project, and for her personal

friendship. I would also like to thank Dr. Michael Bass and Te-Yuan Chung from

CREOL, University of Central Florida, Orlando, for letting me use their thermal imager

for my research.

The MEMS device fabrication was done using the facilities provided by the

University of Florida Nanofabrication Facilities (UFNF) and by the UF Microfabritech

center. Therefore, I appreciate the support provided by the UFNF staff Al Ogden, Ivan

Kravchenko, Bill Lewis and the UF Microfabritech staff. Scanning electron microscopy

(SEM) and white light profilometry were performed using the equipment at the Major









Analytical Instrumentation Center (MAIC) at the University of Florida. I wish to thank

Dr. Luisa Dempere, Wayne Acree, Andrew Gerger and Brad Willenberg of the MAIC for

their assistance. Special thanks go to Tanya Riedhammer who helped me with the

Variable-Pressure SEM for imaging my photoresist microlenses. I also want to

acknowledge our administrative assistant, Joyce White, for her help and support.

Finally, I am eternally grateful to my family and friends for their constant support

and encouragement. I would like to thank my parents, Ranjan and Poonam, and my

brother, Prateek, for their confidence in me, for their endless love and support, and for

keeping me debt-free all through graduate school. I want to thank my fiancee Kavitha, for

all her love, support, advice, and also for all the car rides to school she provided that

ultimately helped my research. Thanks are due to my friends Anuradha Ventakesan and

Boman Irani who kept me going throughout graduate school. I also want to acknowledge

my friends Kanak Behari Agarwal and Himanshu Kaul whose learned advice helped me

through the "mid-PhD crisis".

The MEMS-based endoscopic biomedical imaging project at the University of

Florida has been supported by the National Science Foundation Biophotonics Program

through award number BES-0423557, and by the Florida Photonics Center for

Excellence.
















TABLE OF CONTENTS



A C K N O W L E D G M E N T S ................................................................................................. iv

LIST OF TABLES ....................................................... ............ ....... ....... xi

LIST OF FIGURE S ......... ..................................... ........... xii

A B ST R A C T ................................................................................................................... xviii

CHAPTER

1 IN TR OD U CTION ............................................... .. ......................... ..

1.1 Limitations of Conventional Cancer Diagnosis Methodologies..........................1
1.2 Emerging Optical Coherence Tomography ........................................................3
1.3 M EM S-based OCT ...................................... ........................ ....
1.4 M E M S-based O C M .................................................................... .....................6
1.5 R research O bjectives........... .................................................................... .. .. ....
1.6 D issertation Overview ........................................................... ............... 10

2 OPTICAL BIOIMAGING METHODOLOGIES.....................................................12

2.1 O ptical Coherence Tom ography................................... .................................... 12
2.1.1 O CT System D esign ..................................................... .................. 13
2.1.2 K ey Im aging Param eters ........................................ ........................ 17
2.1.3 Internal Organ OCT Im aging .......................................... ............... 20
2 .1.4 M E M S-based O C T ......................................................................... ..... 22
2.2 Optical Coherence M icroscopy ........................................ ....................... 24
2 .2 .1 B ench-T op O C M .......................................................................... .... ... 26
2.2.2 M EM S-based O CM ..................................................... ...................27
2.3 N on L near O optical Im aging ...............................................................................29
2.3.1 Two-Photon Excitation Fluorescence Imaging .......................................30
2.3.2 Second Harmonic Generation Imaging ............... ................... ...........33
2.3.3 Nonlinear Optical Imaging System Design..............................................35
2.3.4 Endoscopic Nonlinear Optical Imaging ............................................. 37
2.3.5 MEMS-based Endoscopic Nonlinear Optical Imaging ...........................39

3 ELECTROTHERMAL MICROMIRRORS AND ENDOSCOPIC OCT
IM A G IN G ...................................... ................................................. 4 0









3.1 Scanning M icrom irrors ......................................................... ............... 40
3.2 Electrothermal Actuation and Design....................................... ............... 44
3.3 M icrofabrication Process .............................................. ........................... 46
3.4 Bimorph Actuation and Theoretical Analysis ............................................. 50
3.5 Electrotherm al M icrom irrors ........................................ .......................... 54
3.5.1 One-Dimensional Electrothermal Micromirror................. ............. ...54
3.5.2 Two-dimensional Electrothermal Micromirror ........................................57
3.5.2.1 D evice design .......................... ............. .... .. .............. 57
3.5.2.2 D evice characterization ............................. .............................. 59
3.5.2.3 Laser scanning experim ent.................................. ............... 62
3.6 M icromirror Packaging.......................................................................... 63
3.7 M EM S-based Endoscopic OCT Imaging .................................... .................65
3.7.1 M EM S-based OCT System Design................................. ............... 65
3.7.2 O C T Im aging R results ..............................................................................71
3.8 Sum m ary ................................................................. ..... .........73

4 LARGE-VERTICAL-DISPLACEMENT MICROMIRRORS AND NON-
LINEAR OPTICAL IMAGING................................................... .................. 75

4.1 LVD M icroactuator D esign ........................................ ........................... 77
4.2 1-D L V D M icrom irror........................................................................... .... ... 80
4.2.1 Fabricated D evice ......................................................... .............. 80
4.2.2 Equivalent Circuit M odel ........................................ ........ ............... 81
4.2.3 Experim ental Results................................... ......... ... ............... 86
4.2.3.1 Static response ............ .. ... ..... ......................... .... ... ......... 86
4.2.3.2 Frequency response/resonant scanning .......................................89
4 .3 2-D L V D M icrom irror.............................................................. .....................92
4.3.1 M irror D esign ............................................. .. ...... .. ............ 92
4.3.2 Experim ental Results ......................................................... ............... 94
4.3.2.1 Bi-directional scanning ...................................... ............... 94
4.3.2.2 Two-dimensional dynamic scanning................ .............. 98
4.3.2.3 Vertical displacement m otion..................................................100
4.4 MEMS Mirror-based Nonlinear Endoscopy......................................................102
4.4.1 N onlinear Optical Im aging System .................................. ............... 103
4.4.2 Experim ental R results ..................................................... ...... ......... 104
4.5 Sum m ary ..................................... ................................ ......... 107

5 MICROLENS SCANNERS AND OPTICAL CONFOCAL MICROSCOPY........ 109

5.1 L V D M icrolens Scanner ......................................................... ..................... 109
5.1.1 M icrolens Scanner D esign................................................ ..................111
5.1.2 Fabricated M icrolens Scanner ................................................................. 111
5.1.3 Experimental Results............... ........................ .... ............... 115
5.2 M illimeter-Range LVD M icrolens Scanner ................................... ................119
5.2.1 M illim eter-Range Scanner D esign ..........................................................120
5.2.2 Fabrication Process........................................................ ............... 12 1
5.2.3 Experim ental Results...................................... ......... ............... 126










5.3 LV D M icrolens Packaging .................................................................... ...... 132
5 .4 S u m m a ry ...................................................................................................1 3 4

6 CONCLUSIONS AND FUTURE WORK ........................................................135

6.1 Research Effort Accomplishm ents .......................................... ............... 137
6.2 Future W ork ................................................................................... .. .......... 138

APPENDIX

A NON-CMOS, WAFER LEVEL FABRICATION PROCESS...............................140

B ARTICLES GENERATED BY THIS RESEARCH EFFORT.............................145

L IST O F R E F E R E N C E S ......... .. ............... ................. .............................................. 148

B IO G R A PH ICA L SK ETCH ......... ................. ...................................... .....................162






































x
















LIST OF TABLES

Table p

3-1 Thermomechanical properties of some possible bimorph materials at room
tem p eratu re ..............................................................................................................4 5

4-1 Parameters used by the equivalent circuit model of the 1-D LVD micromirror......84

4-2 Actuator characteristics for the 2-D LVD micromirror. ........................................96

5-1 M icrolens characteristics ......................................................... ............... 113

5-2 Estimated microlens parameters for various desired focal lengths......................126
















LIST OF FIGURES


Figure page

1-1 Schematic of a MEMS-based OCT/OCM system. (a) System block diagram. (b)
Optical delay line that uses the LVD micromirror as a reference mirrors for
transverse and axial scanning. (c) OCT endoscope that uses a 1-D or 2-D LVD
micromirror for transverse scanning of tissue. (d) OCM endoscope that uses a
LVD microlens for axial scanning of tissue.............. ............................................7

2-1 OCT schematic................................... .........................................13

2-2 OCT tissue scanning modes. (a) Conventional longitudinal scanning (A-scan).
(b) E n-face scanning (B -scan) ..................................................................... ... .. 15

2-3 Comparison between histology, ultrasound and OCT images of biological tissue.
(a) HE-stained histology, (b) 50-MHz ultrasound, and (c) OCT image of a
n e v u s .......................................................................... 1 6

2-4 Comparison between ultrasound and OCT images of human coronary artery
plaques. (a) In vivo OCT image with axial imaging resolution of 13 atm.
(b) 30 MHz intravascular ultrasound image of the same artery with a lower
resolution of 100 atm ......................... ....... ...... ...... ............ 17

2-5 Gaussian beam optics. ...... ........................... ..........................................19

2-6 Existing endoscopic OCT probe designs.(a) Radial imaging probe design. (b)
Forw ard im aging probe design .................................................................... .. ..... 20

2-7 MEMS-based endoscopic OCT system....................................... ............... 24

2-8 High NA scanning approaches for OCT. (a) Reference mirror and focusing
optics placed on a movable stage to achieve dynamic focusing. (b) Lateral and
axial scanning achieved by displacement of the fiber tip.................. ...............25

2-9 M EM S-based endoscopic OCM schematic................................... ............... 26

2-10 Difference between OCM and OCT............................ ...... ..................... 27

2-11 1-[tm axial resolution by 3-[tm lateral resolution tomogram of a tadpole. ..............28









2-12 Energy band diagrams illustrating (a) one photon and, (b) two photon excitation
fluorescence phenom ena. ........................................ ....................................... 1

2-13 Optical sectioning ability of TPEF imaging. (a) Single-photon excitation of
fluorescein by 488 nm light. (b) Two-photon excitation using 960 nm light. .........32

2-14 Schem atic of a nonlinear microscope ..................................................... ........... 36

2-15 (a) Pulse train from a mode-locked Ti:Sapphire laser at 80 MHz. (b) The laser
pulses typically have a FWHM duration of 100 fs, and (c) a spectral FWHM
band idth of 10 nm .................................................................. ....................36

2-16 Two-photon fluorescence and SHG signals emitted by a sample excited by
800 nm light. .........................................................................37

2-17 Two-photon imaging of the hamster cheek pouch tissue at an excitation
wavelength of 780 nm. (a) Normal tissue. Precancerous tissue: (b) Moderate
Dysplasia. (c) Carcinoma in situ. Cancerous tissue: (d) nonpapillary, and (e)
papillary squam ous cell carcinom a. .............................................. ............... 38

3-1 Side view of a bimorph beam. ................................ .............. ....... .......... 45

3-2 Electrothermal micromirror basic structure. (a) Top view. (b) Cross-sectional
sid e v iew ...................................... ................................. ......... ...... 4 7

3-3 DRIE CMOS-MEMS fabrication process flow. ................... ............................. 48

3-4 SEM of a fabricated 1-D micromirror with initial tilt angle. ............................50

3-5 Bimorph actuation mechanism. Side views of: (a) Initial position of mirror at
zero bias. (b) Downward rotation of mirror plate on application of bias voltage
to poly silicon resistor. .......................... ...................... ... .. ...... .... ....... ..5 1

3-6 SEM of 1-D m icrom irror....................................................................... 55

3-7 1-D mirror characterization. (a) Rotational static response. (b) Plot of the heater
resistance versus applied current ................................................................. .. ..... 56

3-8 Frequency response of the 1-D mirror. ...................................... ............... 56

3-9 Schematic of the 2-D mirror design. (a) Top view showing the axes of rotation.
(b) Cross-sectional view of A -A '...................................................... ....................58

3-10 SEM of a fabricated 2-D micromirror................................ 59

3-11 2-D Mirror Characterization. (a) Rotation angle vs. current, and (b) Polysilicon
resistance vs. current for the two actuators. ................................... .............. 60









3-12 Thermal images of a device biased at 10 V. (a) Temperature distribution across
the mirror actuator only. (b) Thermograph of the entire device ............................61

3-13 Laser scanning using the 2-D mirror. (a) Schematic of experimental setup. (b) 4
x 4 pixel images scanned by the micromirror. ................................. ............... 63

3-14 Micromirror package. (a) Packaged micromirror on a custom PCB. (b) Picture
of the PCB package. (c) Picture of a packaged mirror alongside a US dime coin...64

3-15 Endoscopic OCT probe designs. (a) Side-imaging configuration. (b) Forward-
im aging configuration. ......................................... ................. .. .. ... 65

3-16 Schematic of the MEMS-based endoscopic OCT system................... ............66

3-17 Photographs of the 5-mm diameter MEMS-based OCT endoscope at the State
University of New York at Stony Brook.............. ........... ................... 69

3-18 Photograph of the portable, MEMS-based endoscopic OCT system at the State
University of New York at Stony Brook.............. ........... ................... 70

3-19 Comparison of OCT with histological image. (a) OCT image, and (b)
histological im age of rat bladder ....................................... ............. ... 71

3-20 Bench-top versus MEMS-based endoscopic OCT imaging of rat bladder. (a)
Bench-top OCT image. Size: 6 mm by 2.7 mm. (b) Endoscopic MEMS-based
O C T im age. Size: 4 m m by 2.7 m m .............................................. .....................72

4-1 Design schematic of the LVD mirror. (a) Top view. (b) Cross-sectional view
a cro ss A -A '. ....................................................... ................ 7 7

4-2 Coventor simulations. (a) Device side-view. (b) 3-D model of the LVD
micromirror illustrating the initial curling of the bimorph actuators ....................78

4-3 W iring schematic for the LVD actuators. ..................................... ............... 79

4-4 SEM images of the LVD micromirror. ..................................... ............... 81

4-5 Line scan of the surface profile of the LVD micromirror .......................................81

4-6 Equivalent circuit model of the LVD micromirror device .................................... 82

4-7 SEM of the burn pattern of the mirror actuator...............................85

4-8 LVD mirror characterization. Plots of the (a) rotation angle versus applied
voltage, and (b) polysilicon heater resistance versus applied voltage for the two
actuators. (c) Plot showing the linear correlation between rotation angle and
polysilicon resistance of the actuators....................................... ......................... 88









4-9 Piston motion mode. (a) Vertical displacement of the mirror plate as a function
of the frame actuator voltage. (b) Plot of the mirror actuator voltage versus
frame actuator voltage that was used to drive the LVD device to obtain less than
0.030 tilting of the m irror plate...................................................................... .. .. 90

4-10 Frequency response of the LVD micromirror device....................................91

4-11 2-D LVD micromirror design. (a) Top view of the 2-D micromirror,
highlighting the 4 bimorph actuators. (b) Top view of the actuator area boxed in
part (a). (c) Cross-sectional view of the bimorph actuator as seen across A-A'......93

4-12 SEM of a fabricated 2-D LVD mirror ........................................... ............... 93

4-13 Static 2-D line scans. (a) Plot showing the optical angles scanned in 2-D space
when each actuator is individually actuated. (b) Plot of the effective optical
angle scanned versus actuation voltage for each actuator .....................................95

4-14 Static characterization. (a) Plot showing the linear scan pattern during static 2-D
scanning of Act] and Act4 only. Act4 was actuated at different Act] bias
voltages. (b) Linear plot of actuator resistance versus optical scan angle for each
actu ator. .............................................................................. 96

4-15 Tilt angle stability of the mirror plate versus time. ................................................97

4-16 Initial tilt angle of the mirror plate in x and y directions at different
environm ental tem peratures. ............................................ ............................ 98

4-17 Photographs of 2-D scan patterns obtained by exciting actuators 1 and 4 only.
(a)-(e) Lissajous figures scanned by the micromirror by varying only the phase
of the two excitation signals. (f) Lissajous figure scanned at an excitation
frequency ratio of 1:10. .......................... ...................... .... ...... .... ...... ...... 99

4-18 2-D raster scanning pattern obtained using actuators 1 and 4 ..............................99

4-19 Piston motion mode. (a) Vertical displacement of the mirror plate as a function
of Act4 voltage. (b) Corresponding plot of the relationship between Act3 and
Act4 voltages that is required to generate the vertical displacement. (c) Linear
increase of Act3 and Act4 resistance with vertical displacement........................101

4-20 Tilting of the mirror plate in the negative y-direction (due to thermal coupling)
as a function of vertical position of the mirror............... .... .................102

4-21 Schematic of the nonlinear optical imaging system ............................................103

4-22 Cross-sectional view of the double-clad photonic crystal fiber.............................104









4-23 Second harmonic generation imaging. (a) Series of SHG line profiles from rat
tail tendon taken at 10 |tm axial steps. (b) SHG line profile of an unstained rat
esophagu s tissue. .................................................................... 105

4-24 2-D raster scan pattern scanned by actuators 1 and 4 only. ..................................106

4-25 TPEF imaging of 10-[tm diameter fluorescent microbeads. (a) Bench-top TPEF
imaging system. (b) MEMS-based TPEF imaging. .............................................106

4-26 In vitro imaging of rat stomach epithelial surface stained with 1% acridine
orange in R singer's solution. ............. ............ .... .............. .................. 107

5-1 Design schematic of the LVD microlens scanner. (a) Top view. (b) Cross-
sectional side view. (c) 3-D illustration of the scanner. .............. .....................112

5-2 SEMs of: (a) Fabricated LVD lens holder, and (b) LVD microlens scanner.........113

5-3 Microlens fabrication process. (a) Backside Si etch. (b) Oxide etch. (c) Deep Si
trench etch. (d) Si undercut. (e) Microlens formation by reflow of photoresist..... 114

5-4 SEM of a fabricated PR microlens. ................................ 115

5-5 Vertical displacement experiment. (a) Vertical displacement of the microlens as
a function of frame actuator voltage. (b) Plot of the ratio of the applied voltages
to the lens and frame actuators that was used to obtain the vertical displacement
sh o w n in (a ) ...................................... ............................................. 1 1 6

5-6 Plot showing the increase in polysilicon heater resistances versus vertical
displacem ent for the tw o actuators ............... ...................................... ............... 117

5-7 Microlens imaging quality. (a) Schematic of the imaging experiment apparatus.
(b) Photo of the test pattern. (c) Snap-shot images of the test pattern as obtained
through the PR m icrolens. ......................................................... ...............118

5-8 CCD image of a 4 |tm focused beam spot (top), and its corresponding intensity
profile (bottom ). ......................................... ......... ....... .. ........ .. .. 119

5-9 Top view of the millimeter-range LVD microlens scanner. .............................120

5-10 Modified fabrication process for mm-LVD microlens scanner. (a) Backside Si
etch. (b) Oxide etch. (c) Spin on photoresist. (d) Anisotropic photoresist etch to
expose metal-2 layer. (e) Metal wet etch followed by photoresist removal. (f)
Deep Si trench etch. (g) Silicon undercut. (h) Microlens formation by reflow of
photoresist. .......................................... ............................ 12 1

5-11 Modified process for die-level fabrication of the microlens scanner ....................122









5-12 Photograph of the CMOS die after the wet etch of lithographically exposed
alum inum layers. ...................... .. ........................ .. .. .. .... .. ........... 123

5-13 (a) SEM of a fabricated scanner before microlens formation. Close-up views of
(b) lens actuator; and (c) frame actuator bimorph regions ...................................124

5-14 SEM of a fabricated microlens scanner with integrated polymer microlens. ........124

5-15 SEMs of (a) convex microlens, and (b) ball-type microlens..............................125

5-16 Imaging using the photoresist microlens. (a) Photograph of the test pattern on a
chrome mask. (b) Corresponding image of the test pattern as seen at the focal
plane of the photoresist m icrolens ...................................................................... 126

5-17 Vertical displacement of the microlens scanner. (a) Microlens displacement as a
function of frame actuator voltage. (b) Displacement versus applied electrical
power. (c) Corresponding linear relationship between the two voltages .............128

5-18 Change in resistance versus vertical displacement of the microlens......................129

5-19 Change of initial microlens elevation with ambient temperature...........................129

5-20 Lateral shift of the microlens during vertical displacement actuation. (a)
Illustration of the lateral shift. (b) Characterized plot of the lateral shift.............130

5-21 Dynamic response of the microlens scanner. (a) Mechanical response when a
square excitation was applied to both actuators at t = 0. This can be fitted to an
exponential envelope using the damping ratio, (b) Frequency response
show ing the resonant peaks ....................................................................... ....... 13 1

5-22 Microlens package design. (a) Microlens package schematic. (b) Forward-
im aging O C M endoscope ........................................................................ ......... 132

5-23 Packaged OCM endoscope. (a) With the Lucite end cap. (b) Without the end cap
to show the packaged microlens scanner. ..................................................133

A-1 Non-CMOS, wafer-level fabrication process illustrating the steps required to
fabricate the 1-D LVD m icrom irror device .........................................................140















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

LOW VOLTAGE, MEMS-BASED REFLECTIVE AND REFRACTIVE OPTICAL
SCANNERS FOR ENDOSCOPIC BIOMEDICAL IMAGING

By

Ankur Jain

August 2006

Chair: Huikai Xie
Major Department: Electrical and Computer Engineering

Imaging technologies such as optical coherence tomography (OCT), two-photon

excitation fluorescence microscopy (TPEF), and second harmonic generation (SHG)

microscopy require optical scanners to transversely scan a focused laser beam onto the

tissue specimen being imaged. However, for in vivo early-cancer detection of internal

organs the optical scanners must be integrated into slender endoscopes. The goal of this

work is to develop millimeter-sized MEMS optical scanners packaged inside endoscopes

to enable endoscopic biomedical imaging.

This work reports MEMS micromirrors and microlens scanners fabricated using

post-CMOS micromachining processes, which can provide large scan ranges at low

driving voltages. Several 1-D and 2-D micromirror scanners have been designed,

fabricated and characterized. Scanning micromirrors, as large as 1.3 by 1.1 mm2, have

demonstrated optical scan angles greater than 400 at actuation voltages below 20Vdc. The


xviii









maximum scanning speed of these devices is in the range of 200 to 500Hz, which is

adequate for real-time bio-imaging.

A new electrothermal microactuator design is reported which enables large vertical

displacements (LVD). This LVD microactuator uses two sets of electrothermal bimorph

thin-film beams to provide out-of-plane elevation to the micromirror, while keeping the

mirror parallel to the substrate. LVD micromirrors have demonstrated large bi-directional

scanning ability (>400) as well as large vertical piston motion (-0.5mm) at low driving

voltages (<15V). A 1-D LVD micromirror has the ability to scan optical angles greater

than 1700 at its resonance frequency of 2.4kHz.

Polymer microlenses integrated with the LVD microactuators have been developed

for endoscopic optical coherence microscopy which requires microlenses to axially scan

their focal planes by 0.5 to 2 mm. A modified fabrication process allows larger polymer

lenses with better thermal isolation to be integrated. A maximum vertical displacement of

0.71mm was obtained.

These scanners have been packaged inside 5-mm diameter endoscopes to enable in

vivo imaging. Endoscopic OCT with transverse and axial resolutions of 15[tm and 12[tm,

respectively has been demonstrated at imaging speeds of 2 to 6 frames/second. TPEF and

SHG imaging with imaging resolution greater than 1.5[ tm has been obtained. These

results show the potential for the use of MEMS-based endoscopy for early-cancer

detection














CHAPTER 1
INTRODUCTION

Cancer is a leading killer disease worldwide, and it accounts for as many as one-

quarter of deaths in the United States of America. For the year 2005, the American

Cancer Society estimates 570,280 cancer deaths in the US, and expects more than 2

million new cancer cases to be diagnosed [1]. Although cancers in the breast, colon,

rectum, cervix, prostrate, skin and the oral cavity are readily treatable provided they are

diagnosed at a pre-invasive stage, early lesions in these tissues are often almost

impossible to detect without regular screening. A study has estimated that the 5-year

survival rate of patients with these types of cancers can increase to 95% if the cancers are

diagnosed during their localized precancerous stage [1]. Therefore, early detection of

many of these precancerous lesions is essential in order to greatly reduce patient

morbidity and mortality.

The goal of this research effort is to develop endoscopic imaging modalities that

can detect and diagnose in vivo precancerous lesions. It is proposed to achieve

endoscopic imaging through the use of miniature optical scanners packaged inside

endoscopes. This chapter discusses the limitations of existing cancer diagnostic

techniques and introduces a new cancer detection technique along with the objectives of

the project.

1.1 Limitations of Conventional Cancer Diagnosis Methodologies

Cancer researchers have estimated that more than 85% of all cancers originate

inside the epithelium layer that lines the internal surfaces of organs throughout the human









body [2]; therefore it is of paramount importance to develop methods that can diagnose

cancers just below the tissue surface. The existing diagnosis of cancers is carried out

through visual inspection of the tissue surface followed by random tissue biopsy. Internal

organ cancer screening is conducted by using special biopsy endoscopes that are

equipped with cameras for visually inspecting the internal organ tissue surfaces.* Since

precancers originate below the tissue surface, conventional endoscopes that only image

the tissue surface are unable to make an accurate diagnosis. Therefore, this current

practice of white-light endoscopy often requires biopsies for ex vivo histological analysis

and clinical diagnosis of suspect tissue. This biopsy procedure creates significant delay in

clinical diagnosis, with the added risk and cost of the medical procedure. Another

limitation is the biopsy tissue sampling density. A study performed by Reid et al. on the

early detection of high-grade dysplasia in Barrett's esophagus proved that by reducing the

tissue biopsy sampling interval from 2 cm to 1 cm, the success in detecting cancer was

doubled [3]. However, even this practice of biopsy over-sampling suffers from

substantial limitations since there is a practical limit in the number of biopsies that can be

performed, thereby diminishing its diagnostic potential.

Imaging techniques such as radiography, computed tomography (CT), magnetic

resonance imaging (MRI), and ultrasound allow noninvasive investigation of large-scale

structures in the human body and also permit three-dimensional (3-D) visualization.

However, the imaging resolution of these existing diagnostic techniques makes the

detection and diagnosis of many precancers difficult if not impossible. For example,

bronchial cancers are not commonly detected at curable stages since the precancerous


Ali Fazel, M.D., Personal Communication, Gainesville, FL, 2004.









lesions are generally smaller than the detection limit of current imaging modalities. The

spatial resolution of approaches such as conventional radiography, CT, and MRI is

generally restricted to a few millimeters in standard clinical practice [4], thereby

preventing the detection of lesions less than 1 cm in diameter [5].

However, for detecting cancer in its early stages, an imaging technology with a

higher resolution (< 20 [tm) is necessary for accurate diagnosis. In addition, clinical

screening procedures such as the random biopsy procedure for the diagnosis of cancer

can be improved by using a high-resolution, non-invasive imaging technique to identify

biopsy sites that correspond to the most severe disease.

1.2 Emerging Optical Coherence Tomography

Optical coherence tomography (OCT) is an emerging diagnostic medical imaging

technology that produces high-resolution, cross-sectional images of biological samples

[6-8]. Optical coherence tomography combines the operating principles of ultrasound

with the imaging performance of a microscope. It uses advanced photonics and fiber

optics to focus an infrared light beam into a sample, and then uses low-coherence

interferometry to measure the echo time delay of the reflected light to determine tissue

microstructure. The OCT imaging depth is limited by optical attenuation from tissue

absorption and scattering to about 2 to 3 mm. This is the same scale as that generally

imaged using biopsy and histology. A very attractive feature of OCT imaging is the high

resolution. Although ultrasound imaging has greater imaging depth, OCT has a much

higher imaging resolution of 10 |tm or less [9]. An OCT system with 1 |tm axial

resolution has also been demonstrated [10], which is about two orders of magnitude

higher than that of standard ultrasound imaging. Even though high-frequency ultrasonic









imaging has been shown to obtain a resolution of about 10 |tm [11], the simplicity of

OCT systems permits a much lower hardware cost. Since OCT uses infrared light it is

much safer to use than CT systems which use harmful x-rays. OCT imaging is

minimally-invasive and has the potential to eliminate risky and time-consuming biopsy

procedures; therefore it is also known as optical biopsy.

Optical coherence tomography has been proved to be clinically useful in the field

of ophthalmology, and has great potential for use in cardiovascular, gastrointestinal and

pulmonary imaging through the use of endoscopes and catheters [5, 12, 13]. Endoscopic

OCT systems have been demonstrated to be able to detect in vivo cancers at a very early

stage [14, 15]. For these internal organ applications, the imaging probe must be small,

and fast image scanning is required. Various methodologies have been proposed to

transversely scan the optical beam across the internal tissue surface. Some endoscopic

OCT devices use a rotating hollow cable that carries a single-mode optical fiber, while

others use a galvanometric plate or piezoelectric transducer, that swings the distal fiber

tip to perform in vivo transverse scanning of tissue [12, 14, 16].

1.3 MEMS-based OCT

Micro-Electro-Mechanical Systems (MEMS) or Microsystem Technology (MST) is

another emerging technology that makes miniature sensors and actuators through batch-

fabrication micromachining processes. Micromirrors manufactured using this technology

have been widely used for optical displays and optical switching [17, 18]. MEMS-based

transducers have been also widely used by the automobile industry which uses

accelerometers and other inertial sensors for deploying safety air-bags and other vehicle

stability applications. The small size, fast speed and low power consumption of MEMS









mirrors make them ideal for use in an endoscopic OCT imaging probe. In fact,

researchers have started to use MEMS mirrors for the transverse scanning of endoscopic

OCT systems [19-23]. Xie et al. demonstrated a 5 mm diameter MEMS-based OCT

endoscope that used a 1-D electrothermal mirror to scan the light beam onto the

biological tissue [19]. By performing 1-D transverse scans of the tissue, high resolution

cross-sectional 2-D images were obtained. Zara et al. also reported an endoscopic OCT

system based on MEMS mirrors in 2002 [24], in which the MEMS mirror has large

deflection angle but requires elaborate assembly. Tran et al. [25] and Herz et al. [26]

demonstrated radial endoscopic-OCT imaging using MEMS micromotors packaged

inside endoscopes that rotated a prism or mirror. More recently Fan et al. [22] and

McCormick et al. [23] separately demonstrated 3-D endoscopic OCT imaging through

the use of 2-D electrostatic micromirrors packaged inside fiber-optic endoscopes.

However, the high voltages required for electrostatic actuation may be a concern due to

electrical safety issues during internal organ imaging. Even though the electrothermal

micromirrors used by Xie et al. [19] operate at low voltages, the large initial tilt angle of

the mirror plate complicated the endoscope package design. Another limitation of all

these existing micromirror-based OCT endoscopes is that their lateral resolution is

restricted to a few tens of microns in order to provide the necessary millimeter-range

depth of focus. These lateral resolutions are not sufficient since a much high lateral

resolution (< 10 [tm) is required for the detection of in vivo precancers. These issues

regarding the use of existing MEMS scanners for OCT imaging will be discussed in

greater detail in Chapters 2 and 3.









Proposed solution: In prior work, 1-D and 2-D electrothermal micromirrors with

large rotation angles at low actuation voltages were designed for transverse scanning in

OCT imaging [27-30]. However the unidirectional operation, non-stationary center of

rotation and large initial tilt angle of these mirrors complicated device packaging and

optical design. These issues can be resolved by using a novel microactuator design that

uses two complementarily-oriented electrothermal actuators to keep the mirror parallel to

the substrate, and these actuators also provide bi-directional scanning capability to the

mirror. This actuator pair can also generate large, out-of-plane, piston motion at low

actuation voltages (< 15V). MEMS devices using this novel microactuator design are

referred to as large-vertical-displacement (LVD) microdevices. It is proposed to use

micromirrors integrated with either one or two sets of LVD microactuators to perform 1-

D or 2-D transverse scanning, respectively. The fabricated mirrors will be packaged

inside endoscopes to perform OCT imaging. Also, further miniaturization of the overall

OCT system is possible by replacing the bulky axial scanning mirror with a phase-only

LVD micromirror. Figure 1-1 shows the schematic of a MEMS-based OCT system in

which the LVD micromirror can be used for axial reference scanning as well as

endoscopic transverse bi-directional scanning applications. Further details about the LVD

micromirrors are provided in Chapter 4.

1.4 MEMS-based OCM

Optical coherence microscopy (OCM) is an extension of OCT imaging technique,

and it allows for ultrahigh-resolution cross-sectional imaging of highly-scattering tissue

by combining the imaging capabilities of OCT technology and high numerical-aperture

(NA) confocal microscopy. In OCM, the high imaging resolution in the axial direction is

provided by low-coherence interferometry, while the micron imaging resolution in the











Optical Delay Line
Axial Scanning

Fiber I I
I I


Piston-motion
micromirror


(b)


OCT
Endoscope


Tissue
Scanning
Optics


Microlens scanner
for axial and
transverse scans


Figure 1-1: Schematic of a MEMS-based OCT/OCM system. (a) System block diagram.
(b) Optical delay line that uses the LVD micromirror as a reference mirrors
for transverse and axial scanning. (c) OCT endoscope that uses a 1-D or 2-D
LVD micromirror for transverse scanning of tissue. (d) OCM endoscope that
uses a LVD microlens for axial scanning of tissue. CL: collimating lens. FM:
Fixed mirror.









lateral direction is provided by confocal microscopy. Bench-top OCM systems have

demonstrated lateral imaging resolutions better than 3 |tm [10], and thereby have the

ability to detect precancerous lesions at the cellular level (similar to histology).

Therefore, in order to detect and diagnose precancers in internal organs, endoscopic

OCM apparatus is highly desirable. Unfortunately, the high-NA optical components of

OCM systems are bulky, therefore existing OCM systems are restricted to bench-top set-

ups just like standard microscopes. Researchers have been investigating various

methodologies to develop an endoscopic OCM system, and till date none have been

reported in literature. Endoscopic OCM essentially requires a scanning mechanism which

can vertically displace a highly-focused light spot by up to 2 millimeters inside tissue.

Fitting a high-NA optical scanner that meets this requirement into a millimeter-scale

endoscope has been a challenge.

Proposed solution: This project proposes to integrate a high-NA microlens with an

LVD microactuator to form a LVD microlens scanner which can then be fitted into an

endoscope for OCM imaging. This endoscopic OCM probe will then be used to obtain

high-resolution images in both lateral and longitudinal directions. Since high lateral

resolution results in a reduced depth-of-focus, the LVD microactuator will be used for

vertically displacing the microlens in order to focus a light spot at different depths inside

tissue. A schematic of an OCM endoscope is shown in Figure 1-1(d). The LVD microlens

scanner design allows it to axially displace the focal plane of the scanning microlens by

up to a few millimeters. These LVD scanners are appropriate for endoscopic OCM

systems since the scanners are small enough to fit inside millimeter-sized endoscopes,









and also need low voltages for actuation. Details about MEMS-based OCM and LVD

microlens scanners are presented in Chapter 5.

1.5 Research Objectives

The main goal of this research project is to develop miniature optical scanners for

an endoscopic imaging modality that can detect and diagnose in vivo precancerous

lesions. This main goal has been further subdivided into two approaches for this research

effort. First, this work aims to extend and improve on the MEMS-based endoscopic OCT

imaging technology developed by Xie [31], by developing novel reflective optical micro-

scanners. The aim of this approach is to fabricate reflective scanners that are capable of

providing large bi-directional optical scans (>200) at low actuation voltages (<20 V), and

they also should be small enough to fit inside a 5-mm diameter endoscope. A two-

dimensional optical scanner will also be developed in order to enable three-dimensional

OCT imaging. These micromirrors can also be used with other endoscopic imaging

techniques such as two-photon excitation fluorescence and second harmonic generation

microscopy for in vivo visualization of precancers.

Secondly, this project aims to develop a MEMS-based OCM system which uses a

refractive micro-scanner to provide ultrahigh resolution endoscopic imaging for the

detection of early cancers. The objective of this approach is to develop microlens

scanners which can focus a light beam at different depths inside biological tissue. This

scanner should be capable of providing millimeter-range displacements at actuation

voltages below 20 V, and should also be small enough to fit inside 5-mm diameter

endoscopes. A prototype microlens scanner was initially developed to demonstrate proof-









of-concept, and this scanner design was then scaled to develop millimeter-scale scanners

that meet the requirements of this project.

1.6 Dissertation Overview

This dissertation is divided into six chapters. The first chapter provides the

motivation for this work and a background of current imaging methods used for the

detection of cancer.

Chapter 2 provides the background information on three emerging biomedical

imaging technologies that can perform in vivo detection of precancerous lesions. These

include optical coherence tomography and microscopy, and nonlinear optical imaging

techniques of two-photon excitation fluorescence and second harmonic generation. The

required OCT scanning-probe characteristics for endoscopic OCT imaging are also

explained.

Chapter 3 provides a comprehensive literature review of various MEMS

micromirror design structures and their limitations for use in endoscopic OCT imaging

systems. Electrothermally-actuated micromirror designs are also introduced and their

principles of operation and fabrication process are explained in great detail. Endoscopic

OCT imaging using these micromirrors is also demonstrated.

Chapter 4 presents a novel large-vertical-displacement (LVD) microactuator design

that has the ability to perform bi-directional rotational motion as well as generate large

vertical displacements. 1-D and 2-D LVD micromirrors using this actuator design are

demonstrated. Nonlinear optical endoscopy using these devices is also presented.

In Chapter 5, a novel LVD microlens scanner design along with experimental

results is presented. This microlens scanner will be used for endoscopic optical coherence

microscopy (OCM) imaging, and has been packaged inside a 5-mm diameter endoscope.






11


Finally, Chapter 6 summarizes the entire research effort and lists suggestions for

future work, along with a list of accomplishments for this project.














CHAPTER 2
OPTICAL BIOIMAGING METHODOLOGIES

2.1 Optical Coherence Tomography

Optical coherence tomography (OCT) is an emerging in vivo diagnostic medical

imaging technology that produces high-resolution, cross-sectional images of biological

samples [6-8]. Optical coherence tomographic imaging technology is an optical analogy

of the more conventional ultrasonic pulse-echo imaging technology which measures the

intensity and echo delay of acoustic waves to determine tissue microstructure. Since the

speed of light is many orders of magnitude faster than that of acoustic waves, a direct

measurement of optical echoes cannot be obtained electronically as in ultrasound

imaging. Therefore OCT uses an optical measurement technique known as low-

coherence interferometry to measure the optical delay information in the back-reflected

signal from tissue. Low-coherence interferometry was initially developed and

demonstrated for optical-coherence domain reflectometry (OCDR), a 1-D optical ranging

technique used for locating faults in fiber optic cables [32-34]. Optical coherence

tomography is based on the principles of low-coherence interferometery, and it uses

advanced photonics and fiber-optics to image high-resolution cellular structure of tissues

at depths greater than conventional microscopes. This section presents the operating

principle of OCT imaging technology and also discusses the scanning probe requirements

for endoscopic OCT imaging.









2.1.1 OCT System Design

The schematic of an OCT system is shown in Figure 2-1. The core of this system is

a Michelson interferometer, which uses a broadband light source (BBS) to provide a low-

coherence infrared light beam.

This low-coherence infrared light beam is split at a fiber coupler into the reference

and sample arms of the interferometer. The light in the sample arm is focused onto the

sample; and the reflected light containing time-of-flight information is collected by the

same optical fiber. The reflections from the sample and reference arms are then combined

at the coupler and their optical interference is detected by a photodetector.




n Reference
mirror
Broadband Axial scan
Light Source Tissue



Bem Transverse
7 Beam sa
scan
splitter

Photo- VA
detector Z1 2 Z3
-00 z
Z1 Z2 Z3

Figure 2-1: OCT schematic.

Optical interference is detected by the photodetector only when the optical path

difference of the reference and sample arms is within the coherence length of the light

source. That is, only the light reflected back from a particular depth of the sample is

detected. This is called coherence gating. The amplitude of the interferometric signal

(detected by the photodetector) provides a direct measure of the intensity of back-









scattered light from a tissue segment as thin as the coherence length of the BBS. Since

OCT imaging provides tissue microstructure information pixel by pixel, scanning

mechanisms are required to scan the tissue in axial and lateral directions. The depth

information of the tissue sample is acquired through the axial (z-axis) scanning of an

optical delay line (i.e., the reference mirror in Figure 2-1), while the lateral information is

obtained by transversely scanning the light beam in the sample arm of the interferometer.

Multiple longitudinal scans are performed at different lateral locations to provide a two-

dimensional data set which contains the back-scattering information of a tissue cross-

section. This 2-D data is then displayed as a grayscale or false color OCT image. Since

multiple longitudinal scans are performed at different lateral positions, this scanning

mechanism is similar to the A-scan image scanning method used in ultrasound. En-face

scanning is another method in which the tissue is transversely scanned at different

longitudinal locations to generate a B-scan OCT image. A schematic of these scanning

operations is shown in Figure 2-2.

Depending on the coherence length of the employed broadband light source, OCT

can provide cellular or even sub-cellular resolutions (1-20 [tm), which are one or two

orders of magnitude higher than that of commonly used ultrasound imaging (-100 [tm)

[9]. Figures 2-3 and 2-4 show the difference in tissue image resolutions obtained using

ultrasound and OCT technologies [9, 35].

Since more than 85% of all cancers originate in the tissue epithelial layer, which is

within the penetration depth (a few mm) of infrared light [2, 36, 37], malignant or

premalignant changes of epithelia can be detected at a very early stage without

performing biopsies. Also by using OCT imaging along with conventional biopsy, highly





























(a)
x































X
"SLOW.....






Tissue






x



Figure 2-2: OCT tissue scanning modes. (a) Conventional longitudinal scanning (A-
scan). (b) En-face scanning (B-scan). Adapted from Podoleanu et al. [38].













= Y.i..,' p' ut ,.op. w
WE..1
.... ,- .....," ...J ., -
IL^. -'-T ,. ,s- .;.-,
S-'
N, ':

-r,
,i ; -,- .'..
,.i .
i "


4 ,


~pf;- ~I;;


--- -- -










(b)

















Figure 2-3: Comparison between histology, ultrasound and OCT images of biological
tissue. (a) HE-stained histology, (b) 50-MHz ultrasound, and (c) OCT image
of a nevus. 1996 IEEE. Reprinted, with permission, from Pan et al. [35].


IR-."


* ':


.t
ilc:
.I;.
r ; j~'























Figure 2-4: Comparison between ultrasound and OCT images of human coronary artery
plaques. (a) In vivo OCT image with axial imaging resolution of 13 |tm.
(b) 30 MHz intravascular ultrasound image of the same artery with a lower
resolution of 100 |tm. Reprinted from Jang et al. [39], Copyright 2002, with
permission from The American College of Cardiology Foundation.

suspicious tissue areas can be easily identified which can reduce the randomness of

biopsies. Optical coherence tomography has been applied to a wide variety of biological

tissue and organ systems including eyes, skin, teeth and gastrointestinal and respiratory

tracts [5, 12, 14, 40-42]. Researchers have also demonstrated 3-D OCT imaging by 2-D

lateral scanning of tissue [23, 42, 43].

2.1.2 Key Imaging Parameters

The performance of an OCT system is mainly determined by its axial and

transverse imaging resolutions, dynamic range and also by its imaging speed. OCT

achieves a very high axial resolution because the axial and lateral resolutions are

independent of each other, unlike the case with conventional or confocal microscopy.

In OCT the axial resolution is determined by the coherence length of the broadband

light source. For a light source with a Gaussian spectrum, the coherence length (lo) in air


- -ifW
.yE. I









is given by [8]:

21n(2) A/2 A 22
SA2 0.44 (2-1)
;Tz AA AA

where, X and AX are the center wavelength and the full-width at half-maximum (FWHM)

spectral bandwidth of the light source, respectively. In order to obtain high axial

resolution, a short temporal coherence length is desired, therefore, a light source with a

broad emission bandwidth, i.e., a broadband light source (BBS) should be used. The

BBS should operate in a spectral range that allows adequate penetration of light into

tissue. Researchers have determined that BBS's that emit infrared light with a center

wavelength between 1200 nm to 1800 nm achieve the deepest penetration in most tissues

[44]. Another requirement for the BBS is that the irradiance of the emitted light should be

high enough to provide a wide dynamic range. A wide dynamic range provides high

detection sensitivity by enabling OCT imaging of weakly backscattering microstructures

present deep inside the tissue.

As shown in Figure 2-5, the lateral resolution of an OCT system is determined by

the spot-size of the focused optical beam on the tissue. The diameter of the focused spot-

size of a Gaussian beam is given by:

4 2 f 20( 1
Ax = D--- -t- (2-2)
nx D n NA (2-2)

where, D is the diameter of the beam, NA is its numerical aperture, andfis the

focal length of the lens. The lateral resolution of an OCT system is also affected by the

depth of focus of the optical beam.

The depth of focus, also known as the confocal parameter or the Rayleigh range, of

an optical beam is the longitudinal distance within which the optical beam is considered










14 DOF -


-- f --1
Figure 2-5: Gaussian beam optics.

to be in focus. The depth of focus (DOF) of a Gaussian beam is given by:

28A = fY 2Z 2 2A 1 2
DOF = \D =-(Ax) -- (2-3)
D) 2A NA

This implies that a smaller spot-size will increase the lateral resolution but at the

cost of reduced depth of focus. Since OCT imaging can penetrate tissue depths up to a

few millimeters [36, 41], a depth of focus of a few millimeters is required. To overcome

this depth of focus limitation, many researchers limit the numerical aperture of the

scanning optics to obtain a DOF of approximately 1 mm [8]. At a center wavelength of

1300 nm, a lateral spot size of about 29 |tm is achieved with a DOF of 1 mm. In order to

improve lateral resolution without sacrificing depth of focus, some researchers have

developed novel methods to scan the reference path-length and position of the focused

sample optical-beam simultaneously [45, 46]. These methods that increase the lateral

resolution without reducing the DOF will be discussed in Section 2.2.

The focal length of the scanning optics is of the same order of magnitude as the

working distance between the OCT scanner and the tissue sample, which is typically a

few millimeters. Therefore, by plugging in a focal length of a few millimeters in

Equation (2-2), one can see that in order to obtain a lateral resolution better than 20 |tm,

the diameter of the optical beam on the scanning optics should be large (> 1 mm). This is


Ax









an important requirement to consider while designing optical scanners for endoscopic

OCT imaging.

2.1.3 Internal Organ OCT Imaging

It is challenging to realize in vivo imaging of internal organs due to the size

limitations of conventional OCT systems. Fiber optic endoscopes specifically designed

for endoscopic OCT imaging are required for imaging internal organs. Some key factors

that have to be considered for in vivo OCT imaging are endoscope sizes and imaging

speed. Therefore for internal organ applications, miniature OCT imaging endoscopic

probes with diameters of a few millimeters must be developed. Also the transverse

scanning mechanism should be fast enough to provide real-time images.


StatGear PF r rism Cleaved Fiber Tip
Stationary Fiber i --


Piezoelectric
DC Motor ] Cantilever

(a) (b)
Figure 2-6: Existing endoscopic OCT probe designs.(a) Radial imaging probe design.
Adapted from Li et al. [47]. (b) Forward imaging probe design. Adapted from
Boppart et al. [16].

Various methodologies have been presented to transversely scan the optical beam

across the tissue surface to obtain 2-D OCT images. Many researchers have demonstrated

radial OCT scans by rotating a hollow cable that carries a single-mode fiber and a

microprism [12, 13, 47, 48]. Bouma and Tearney used a galvanometer to linearly

translate a optical fiber above the surface of in vivo tissue [49]. Sergeev et al. used a

galvanometric plate to swing the distal fiber tip to perform transverse scanning [14].









Bopart et al. [16, 50] and Li et al. [16, 50] also demonstrated transverse scanning by

swinging a cantilevered fiber tip, but by using piezoelectric methods. A schematic

drawing of these types of endoscopes is shown in Figure 2-6. These existing methods are

complex and use bulky components, thereby miniaturization of the endoscopic OCT

system becomes difficult. Furthermore, rotating a fiber is slow and it also introduces

complexities due to non-linear optical coupling. Therefore, it may be advantageous to

replace these existing scanning mechanisms with a MEMS-based scanning solution to

perform OCT imaging. This MEMS-based solution may also result in smaller probe sizes

with faster scanning speeds, and at potentially lower cost.

The MEMS-based device should meet the following requirements for use in OCT

probes:

Large scanning angle: A large scan angle should be provided by the device in

order to image large tissue areas. Large scanning angle combined with fast scanning

speed will also reduce the OCT imaging time.

Fast scanning speed: A fast scanning speed is desired to enable real-time OCT

imaging, as well as to reduce the time required to conduct the endoscopic procedure.

Low operating voltage: Low voltage operation of the MEMS scanner is essential

for electrical safety of the patient during internal organ imaging.

High resolution: The scanner should be able to provide lateral resolutions (spot

sizes) better than a few tens of microns at working distances of a few millimeters.

Small size: Endoscope diameters should be smaller than a few millimeters,

therefore MEMS devices should be small enough to fit inside. However, there exists a









trade-off between the small device size and high resolution requirements since a larger

scanner size is needed for high spatial resolution.

The small size, fast speed and low power consumption of MEMS mirrors makes

them an ideal choice for use in endoscopic OCT imaging probes. The micromirror must

be large (-1 mm) and optically-flat to maintain high light coupling efficiency and spatial

resolution, and should also have large angle of rotation to meet the scanning range

requirements. In fact, OCT endoscopes using micromirrors actuated electrostatically or

electrothermally, have already been reported in literature [24, 51, 52]. Using these

systems two dimensional, high resolution, cross-sectional images were obtained.

As shown in Figure 2-1, 3-D OCT imaging requires 2-D transverse scanning.

However, almost all existing OCT systems have only 1-D transverse scanning. In those

cases, 3-D imaging is typically obtained by physically pulling the entire imaging probe.

Since endoscopic catheters are flexible, the pull-out length may not be exactly the same

as the physical displacement of the imaging probe. Therefore, there also exists a

requirement for miniature scanners that can transversely scan the tissue surface in 2-D,

thereby enabling endoscopic 3-D OCT imaging.

2.1.4 MEMS-based OCT

As discussed earlier in this section, the key to making compact OCT probes is to

miniaturize the scanning mirrors. In fact, there are a few groups who are working on

MEMS micromirror based OCT. Pan et al. assembled the first MEMS micromirror based

endoscopic OCT system in 2001 [53]. Zara et al. also reported an endoscopic OCT

system based on MEMS mirrors in 2002 [24], in which the MEMS mirror has large

deflection angle but requires elaborate assembly.









Xie et al. reported a MEMS-based endoscopic OCT system that used a single-axis,

single-crystal-silicon, electrothermal micromirror to scan the light beam onto the tissue

[51]. Figure 2-7 shows a schematic of the MEMS-based OCT setup reported by Xie et al.

[51], where the scanning micromirror along with the focusing optics was packaged inside

a 5-mm diameter endoscope. The collimated light in the sample arm of the Michelson's

interferometer is reflected off a beam steering 1-D micromirror and focused into the

tissue. The same mirror collects the back-scattered light from the tissue, and the tissue

microstructure is determined by low coherence interferometry. Two-dimensional (2-D)

(i.e., x-z) cross-sectional images are obtained by combining the transverse scanning of the

1-D micromirror in the x-direction with the axial scanning of the reference mirror in the

z-direction, as shown in Figure 2-7. This MEMS-based system acquired OCT images at a

rate of 5 frames per second, thereby demonstrating the potential for real-time clinical

diagnosis of cancers. Since the micromirror is packaged inside a flexible endoscope, no

mechanical movement of the endoscope is necessary for OCT imaging.

Other researchers have also used MEMS-based solutions to address the transverse

scanning requirements of endoscopic OCT systems. Qi et al. used a MEMS deformable

mirror to tune the focus of the OCT objective lens [54]. Tran et al. demonstrated an

endoscopic OCT catheter using a MEMS micromotor to rotate a prism [25], while Herz

et al. used a MEMS micromotor to rotate a mirror [26]. However, all these efforts are

focused on only transverse scanning in one direction. 3-D OCT images can be obtained

by using a MEMS scanner that can transversely scan the tissue in the x-y plane, i.e., in

two dimensions. Yeow et al. demonstrated 3-D OCT imaging by using a 2-D electrostatic









Axial Scanning

Broadband Light f-I IReference
Source Mirror

50:50

Photodetector iber Endoscope



Signal CM MM
Processing


Y x Tissue Sample

Figure 2-7: MEMS-based endoscopic OCT system. MM: Micromirror. CM: Collimator.

micromirror that scanned angles smaller than +0.50 in both transverse directions [21]. Fan

et al. packaged a 2-D electrostatic micromirror inside a 5-mm diameter endoscope for in

vivo OCT imaging [22]. McCormick et al. also demonstrated 3-D endoscopic OCT

imaging using a 2-D scanning MEMS micromirror that used electrostatic actuation [23].

However, the high voltages required by electrostatic actuation for the abovementioned

scanners may be a concern due to electrical safety issues during internal organ imaging.

MEMS mirrors will be discussed in detail in the next chapter.

2.2 Optical Coherence Microscopy

As discussed in the previous section, the lateral resolution of an OCT system was

restricted to a few tens of microns due to the millimeter-range depth-of-focus requirement

for the scanning optics. This low lateral imaging resolution was due to the use of focusing

optics with relatively low numerical apertures (NA). Researchers have demonstrated









various OCT system architectures that use high NA optics to obtain high-lateral

resolution images without sacrificing the required depth of focus [10, 45, 46, 54, 55].

Figure 2-8 illustrates some of these different scanning approaches. Drexler et al.

demonstrated a bench-top OCT system that used an x-z scanning stage to traverse the

tissue sample in the transverse and axial directions to perform OCT imaging with

resolutions better than 3 ptm [10]. The depth of focus of their imaging optics was less

than 100 Ltm, therefore nine separate OCT images were fused together to form a

millimeter-deep ultrahigh-resolution tomogram. Other researchers have developed novel

methods to scan the reference path-length and the position of the focused sample optical-

beam simultaneously, thereby improving lateral resolution without sacrificing the depth

of focus [45, 46]. Schmitt et al. placed the reference scanning mirror and the focusing

optics on the same movable stage, which allowed the use of lenses with high NA [45].

Since these modified OCT systems use high NA optics to obtain OCT images with high

lateral resolutions (< 10 tpm), they are also referred to as optical coherence microscopes.



Moving Stage



,,T issue
issue Tissue
F-7 Optical
Fiber
Fixed Lenses

Combined sample
& reference beams
(a) (b)
Figure 2-8: High NA scanning approaches for OCT. (a) Reference mirror and focusing
optics placed on a movable stage to achieve dynamic focusing. Adapted from
Schmitt et al. [45]. (b) Lateral and axial scanning achieved by displacement of
the fiber tip. Adapted from Schmitt [8].









2.2.1 Bench-Top OCM

Optical coherence microscopy (OCM) is an extension of OCT technology and it

allows for ultrahigh-resolution cross-sectional imaging of highly-scattering tissue by

combining the imaging capabilities of OCT and high NA confocal microscopy. In OCM,

the high imaging resolution in the axial direction is provided by low-coherence

interferometry, while the micron imaging resolution in the lateral direction is provided by

confocal microscopy. Confocal microscopy is an optical technique that is used for

imaging thin optical sections of relatively transparent tissue with very high resolution.

Researchers have demonstrated bench-top OCM systems that obtained imaging

resolutions better than 10 |tm in both, the axial and transverse dimensions [56-58]. A

schematic of an OCM system is shown in Figure 2-9. As seen in the figure, the OCM

system architecture is similar to that of OCT systems, the only difference being the

imaging methodology of the tissue scanning optics. Figure 2-10 shows the difference


Axial Scanning

' H |Reference
1 I Mirror


Endoscope


Collimating
Lens


Microlens
Scanner


Figure 2-9: MEMS-based endoscopic OCM schematic.









OCT OCM

Low NA High NA

Imaging Lateral -
Depth Resolution/\


Figure 2-10: Difference between OCM and OCT.

between OCM and OCT imaging methodologies. OCT relies on large depth of focus to

obtain imaging depth at the price of low transverse resolution, while OCM uses higher

NA lenses (NA>0.2) to achieve high transverse resolution but with smaller imaging

depth. The OCM imaging depth can be extended by using a moving lens or stage, as

mentioned above. Figure 2-11 shows the tissue image acquired using an OCM system,

demonstrating that cellular imaging is possible using this technology. Therefore, OCM

technology is very promising for the early detection of cancer.

2.2.2 MEMS-based OCM

Although bench-top OCM systems allow for ultrahigh lateral and axial resolutions,

endoscopic OCM probes with ultrahigh imaging resolutions are needed for in vivo

detection of precancers in internal organs. Since the methods shown in Figure 2-8 require

the use of mechanical stages with stepper motors, they are bulky and slow, and therefore

cannot be used for ultrahigh-resolution endoscopic OCM imaging. A MEMS-based

dynamic focusing micromirror has been proposed by Qi et al. that could potentially be

used for endoscopic OCM [54]. They demonstrated a MEMS deformable mirror to focus

a high NA objective lens at different depths inside biological tissue. However their

micromirror requires a high ac voltage of 400 V (peak to peak) to produce a 1.25-mm

focus scan range. Ding et al. used an axicon lens to obtain OCT images with a lateral




































Figure 2-11: 1-[tm axial resolution by 3 -tm lateral resolution tomogram of a tadpole.
Reprinted, with permission, from Drexler et al. [10].

resolution of 10 |tm with a depth of focus of 6 mm [59]. However, the axicon lens

significantly reduced the optical signal intensity, which will result in a lower sensitivity

for the OCT system. Kwon et al. demonstrated a microlens scanner for micro-confocal

imaging that used electrostatic vertical-comb-drives; however its vertical scan range is

restricted to less than 55 |tm and therefore unsuitable for OCM imaging [60-62].

In order to obtain high lateral resolution without compromising the axial scanning

range and small size requirement of endoscopic OCM systems, a MEMS microlens can

be used to scan along the optical axis. This MEMS scanner should be able to axially

displace the focal plane of the scanning microlens by up to a few millimeters. Other

requirements are that the scanner should be small enough to fit inside millimeter-sized









endoscopes, and it should also use low voltage for actuation. A schematic of a MEMS-

based OCM system is shown in Figure 2-9. The overall system design is similar to the

MEMS-based OCT system architecture presented in Section 2.1.4. Infrared light in the

sample arm of the Michelson interferometer is first collimated by a GRIN lens placed

inside the hollow endoscope. Then the collimated light is focused by a high NA polymer

microlens into the tissue, as shown in Figure 2-9. In this OCM scanner design, the

polymer microlens is attached to a MEMS microactuator, which enables vertical

displacement of the microlens. A vertical displacement of the microlens results in vertical

displacement of the focused beam-spot inside the tissue. By changing the vertical

position of the focused beam-spot, it is possible to scan axially into the tissue. Since a

high NA microlens is used, this system will provide OCM images with high lateral

resolution. Unlike the axicon lens used by Ding et al. [59], the smaller depth of focus of

the polymer microlens will maintain a strong optical signal intensity, which will help to

improve the overall OCT system sensitivity.

2.3 Non Linear Optical Imaging

The imaging methods described in the preceding sections were linear optical

imaging methods, since the magnitude of the observed signals from tissue changes

linearly with incident light intensity. The well-known optical phenomena of reflection,

refraction, and diffraction occur in the linear domain since the intensity of reflected,

refracted or diffracted light changes linearly with the magnitude of the incident light.

Other naturally occurring linear events are the absorption of light and photochemical

reactions such as in the photosynthesis process in plants and bacteria [63].

This section introduces another class of imaging modalities that use the nonlinear

optical properties of tissue for high-resolution imaging. The nonlinear optical imaging









modalities presented in the following sections have the potential to detect and diagnose

in vivo cancers at a very early stage.

2.3.1 Two-Photon Excitation Fluorescence Imaging

Two-photon excitation fluorescence (TPEF) microscopy is a nonlinear optical

imaging technique which can provide high resolution imaging at the cellular level. TPEF

microscopy is a new form of scanning far-field fluorescence optical microscopy.

Far-field fluorescence optical microscopy is typically a one-photon excitation

fluorescence based microscopy technique, in which illumination is focused into a

diffraction-limited spot scanned on the tissue specimen, thereby confining the excitation

focal region. The diagram in Figure 2-12(a) depicts the phenomena of fluorescence when

a single photon is absorbed by a fluorescent molecule, and so the molecule is excited to a

higher energy state. The excited molecule now returns to its ground energy state by

emission of a fluorescent photon at a characteristic wavelength. As seen in Figure 2-

12(a), the energy of the fluorescing photon is less than the energy of the excitation

photon, therefore the fluorescence emission is shifted towards a longer wavelength than

that used for excitation. This means that in order to obtain fluorescence from samples that

exhibit fluorescence in the blue-green wavelengths (- 450 nm), the sample would have to

be excited at a lower ultraviolet (UV) wavelength of about 350 nm. However, exciting

tissue samples at UV or blue wavelengths is undesirable due to problems due to

photobleaching and phototoxicity [64]. Another problem with one-photon excitation is

that the entire thickness of the sample within the hourglass-shaped region of the focused

light spot is excited, which results in poor optical sectioning ability. This is shown in

Figure 2-13(a).










Thermal Thermal
relaxation relaxation



hv

Shvf h hvf

o o
h- -




(a) (b)

Figure 2-12: Energy band diagrams illustrating (a) one photon and, (b) two photon
excitation fluorescence phenomena.

Two-photon excitation fluorescence (TPEF) microscopy provides an inherent

optical sectioning ability to improve axial imaging resolution, and it is also less affected

by the effects of photobleaching and phototoxicity. For two-photon excitation to occur,

the fluorescent molecule should simultaneously absorb two photons of a longer

wavelength to reach its excited state. As shown in Figure 2-12(b), two photons with

lower energy are simultaneously absorbed to provide the energy needed to prime the

fluorescence process. The fluorescent molecule now emits a single photon of

fluorescence as if it were excited by a single higher energy photon. This phenomenon of

TPEF depends on two photons interacting simultaneously with the molecule, and it

results in a quadratic dependence on the intensity of incident excitation light. In contrast,

conventional fluorescence is linearly dependent with the excitation light intensity. The

reason that TPEF is referred to as a nonlinear imaging method is due to the fact that the

rate of occurrence depends nonlinearly on the incident light intensity. Since light

intensity is the highest at the focal spot, the largest probability of observing TPEF is at









this location. Axially away from the focal plane, the TPEF probability drops off rapidly

with decreasing light intensity. As seen in Figure 2-13(b), no significant amount of

fluorescence is emitted from regions away from the focal plane, and this demonstrates

TPEFs' inherent optical sectioning ability. Therefore, TPEF microscopy can image tissue

with very high resolution in all three dimensions.

TPEF theory: The probability, p that a molecule absorbs two photons

simultaneously to reach its excited state has been computed as [64]:

P OC KJ2 (2.4)

where, K is a proportionality factor, and I is the intensity of the incident laser beam. The

timescale for the keyword 'simultaneous' for TPEF is the same timescale of molecular


Figure 2-13: Optical sectioning ability of TPEF imaging. (a) Single-photon excitation of
fluorescein by 488 nm light. (b) Two-photon excitation using 960 nm light.
Reprinted by permission from Macmillan Publishers Ltd: Nature
Biotechnology, Zipfel et al. [65], copyright 2003.








energy fluctuations at photon energy scale, and using Heisenberg's uncertainty principle

this has been computed to mean a temporal window of 10-16 s or 0.1 fs [64]. The emitted

fluorescence intensity, If(t) from the molecule is proportional to the molecular cross-

section 6, and also to the square of the incident intensity I(t)2 [64]:

I NA2 2
if(t) S.I(t)2 P(t) 2 NA (2.5)


where, P(t) is the optical power of the incident light, c is the speed of light, h is the

Planck quantum of action, 6 is the two-photon absorption cross-section, and NA is the

numerical aperture of the focusing objective. The time averaged fluorescence intensity

emitted from a fluorophore when excited with a pulsed laser beam with pulse width Cp,

repetition ratefp, and average power Po can be computed from Equation (2.5) as:


2PS NA2 2
() f 2hcZ (2.6)


The number of photons absorbed by a fluorophore per pulse is given by [66]:

P0 ( NA2 2
na pf 2 2hc (2.7)


Equation (2.7) does not account for saturation effects, and was computed with the

paraxial approximation assumption.

2.3.2 Second Harmonic Generation Imaging

Second harmonic generation (SHG) is also a nonlinear optical process, similar to

TPEF, and it can be used for high resolution imaging of tissue microstructure. SHG

converts an input optical wave into an output optical wave of twice the input frequency,









therefore this phenomenon is also commonly known as frequency doubling. This is the

same process used to produce green light at a wavelength of 532 nm from a Nd-YAG

laser operating at 1.06 |tm [63].

Similar to TPEF, the probability of SHG is proportional to the square of the

intensity of the incident excitation light. Thus, SHG imaging has the same intrinsic

optical sectional ability as TPEF imaging. However, unlike TPEF, SHG is confined to

imaging only highly polarizable materials that lack a center of symmetry. SHG imaging

can be used for bioimaging purposes since biological materials can be highly polarizable

and the cellular membranes contain SHG-active constituents which are asymmetrically

distributed [67]. The second-harmonic light emitted from the noncentrosymmetric, highly

polarizable material is exactly half the wavelength of the incident excitation light.

Therefore, the SHG process within the nonlinear optical material converts two incident

photons into one exiting photon at exactly half the wavelength (or twice the energy). As

described in Section 2.3.1, in TPEF some of the incident energy of the photon is lost

during thermal relaxation of the excited state (Figure 2-12(b)), but in the case of SHG,

there is no excited state and so SHG is energy conserving and it also preserves the

coherence of the incident laser light. Since SHG does not involve excitation of molecules,

it should not suffer from photobleaching or phototoxicity effects (which limit the

usefulness of fluorescence microscopy). Another advantage of SHG is that it uses

excitation wavelengths in the near-infrared range which allow for excellent depth

penetration, thereby permitting imaging of thick tissue samples [68].

SHG theory: The nonlinear polarization for a material can be expressed as [68]:

P = Z() E + (2) E E + (3) EEE+... (2-8)









where, P is the induced polarization vector, E represents the electric field vector, X(i) is

the ith order nonlinear susceptibility tensor, and 0 represents a combined tensor product

and integral over frequencies. The first term in the series, X(1) describes normal

absorption and reflection of light. The second term describes the sum and difference

frequency generation; and thereby also describes SHG. The third term describes two-

photon absorption (the probability of which is linearly proportional to the imaginary part

of the third-order nonlinear susceptibility tensor), as well as third harmonic generation

and coherent anti-Stokes Raman scattering. The portion of the polarization that

contributes to SHG is:

p(2) = (2) O E E (2-9)

The intensity of the SHG signals, ISHG emitted from such materials will scale as

follows [68]:


ISHG C o2 (2) 2 (2-10)

where Po and c are the laser pulse energy and pulse width, respectively. This term shows

the nonlinear dependence of the SHG emission intensity to the incident light intensity.

2.3.3 Nonlinear Optical Imaging System Design

The schematic of a nonlinear optical imaging system is shown in Figure 2-14. The light

source generally consists of a pump laser and a Ti:Sapphire laser which generates -100

femtosecond long laser pulses at around 1W power at a repetition rate of 80 MHz. A laser

pulse train output of such a laser system is depicted in Figure 2-15. This laser light is

focused by a microscope objective lens and scanned laterally on the tissue sample using

an XY beam scanner. In a fluorescence microscopy system, the dichroic mirror is used



















Dichroic
Mirror


40x
H Objective




Tissue

Figure 2-14: Schematic of a nonlinear microscope.


(a) (b) (c)


Laser pulse train

12 ns

-200 -100 0 100 200 780 800 820
Time (ns) Time (fs) Wavelength (nm)

Figure 2-15: (a) Pulse train from a mode-locked Ti:Sapphire laser at 80 MHz. (b) The
laser pulses typically have a FWHM duration of 100 fs, and (c) a spectral
FWHM bandwidth of 10 nm. Adapted from Zipfel et al. [65].

for separating the excitation and emission light beams. This dichroic mirror reflects light

with wavelengths longer than 800 nm, while it transmits light with shorter wavelengths.

The emission signal from the tissue specimen is collected by the same focusing optics,

passes through the dichroic mirror, and is detected by a photomultiplier tube (PMT) as








shown in Figure 2-14. A bandpass filter is inserted in the light path before the PMT to

help differentiate between the TPEF and SHG signals. Figure 2-16 illustrates the SHG
and TPEF emissions when excited with near-infrared light.


SHG TPEF Illumination



A A
I I I
400 520 800
Wavelength (nm)

Figure 2-16: Two-photon fluorescence and SHG signals emitted by a sample excited by
800 nm light.
Researchers have used nonlinear optical microscopes to image and identify

cancerous tissue with very high resolution as shown in Figure 2-17. The hamster cheek

pouch biopsies were imaged using a bench-top system with lateral and axial imaging

resolutions of 0.35 and 1.25 |tm, respectively [69].

2.3.4 Endoscopic Nonlinear Optical Imaging
As stated in the previous section, researchers have successfully demonstrated high
resolution imaging of tissue using bench-top nonlinear microscopes [69, Zipfel, 2003
#241]. However in order to demonstrate in vivo imaging, lateral beam scanning

endoscopes are required.

Jung and Schnitzer developed a free-space multiphoton endoscope using GRIN

lenses [70]; however, the lack of a flexible optical fiber prevents its use for endoscopic in
vivo imaging. Helmchen et al. used a piezoelectric bending element to transversely scan a

cantilevered fiber tip [71], but the 1.3-cm diameter, 7.5-cm long endoscope is too bulky










A. B. C. D. E.





























wavelength of 780 nm. (a) Normal tissue. Precancerous tissue: (b) Moderate
Dysplasia. (c) Carcinoma in situ. Cancerous tissue: (d) nonpapillary, and (e)
papillary squamous cell carcinoma. The top image is at the surface of the
tissue, and each subsequent image in the montage represents a 10-[tm axial
step. Scale bar represents 30 |jm. Reprinted, with permission, from Skala eta!.
[69].
















to be used for internal organ imaging. Flusberg et a!. [72] also used a piezoelectric

actuator, along with a MEMS micromotor, to create a multiphoton microscope designed

for imaging peripheral organs of small animals. Gobel et a!. demonstrated in vivo TPEF
Figure 2-17 Two-photon imaging using a fiber bundle and GRN lens, but av hamsteraging cheek pouch tissue atof an excitation
wavelength of 780 nm. (a) Normal tissue. Precancerous tissue: (b) Moderate
Dysplasia. (c) Carcinoma in situ. Cancerous tissue: (d) nonpapillary, and (e)
papillary squamous cell carcinoma. The top image is at the surface of the
tissue, and each subsequent image in the montage represents a 10-tm axial
step. Scale bar represents 30 tm. Reprinted, with permission, from Skala et al.
[69].

to be used for internal organ imaging. Flusberg et al. [72] also used a piezoelectric

actuator, along with a MEMS micromotor, to create a multiphoton microscope designed

for imaging peripheral organs of small animals. Gobel et al. demonstrated in vivo TPEF

imaging using a fiber bundle and GRIN lens, but averaging and the use of a Gaussian

blur filter were needed to improve image quality [73]. Bird and Gu developed a radially

scanning endoscope, similar to the OCT endoscope design illustrated in Figure 2-6(a),









that required an external motor to physically rotate the endoscope [74]. Myaing et al. also

adapted a piezoelectric OCT endoscope for in vivo, endoscopic TPEF imaging [75].

2.3.5 MEMS-based Endoscopic Nonlinear Optical Imaging

It is clear from the above discussion that the design requirements for TPEF

endoscopes are almost the same as that for OCT endoscopes, which are listed in Section

2.1.3. The main difference being that the nonlinear imaging probes should be able to

provide spot-sizes in the micron range, and should also be able to laterally scan higher

power laser beams. MEMS-based scanners, packaged inside endoscopes with high

numerical aperture optics, are very suitable for endoscopic nonlinear optical imaging as

they can provide large scan ranges with high imaging resolution.

L. Fu et al. used the micromirrors developed by this research effort to demonstrate

the first-ever MEMS-based nonlinear optical endoscopy system [76, 77]. Recently,

Piyawattanametha et al. used an electrostatic micromirror to transversely scan the

proximal end of a free-space, GRIN-lens endoscope [78]. The high voltage requirement

(up to 160 V) of this MEMS scanner is a safety concern for in vivo internal organ

imaging.

Ideally, the MEMS-based endoscopes should be capable of providing large scan

range and high imaging resolution at a fast scan speed, but at low operating voltage.

MEMS micromirrors are discussed in the next chapter, while the endoscopic TPEF and

SHG imaging results obtained by this research effort are reported in Section 4.4.














CHAPTER 3
ELECTROTHERMAL MICROMIRRORS AND ENDOSCOPIC OCT IMAGING

As mentioned in the previous chapter, system miniaturization is the key for OCT to

become practical for clinical use in imaging visceral organs. We can see from Figure 2-7

that the miniaturization of OCT imaging systems is largely determined by the axial

scanning and transverse scanning mirrors. MEMS technology leverages integrated

circuits (IC) technology to manufacture micro-scale devices and systems [79, 80], and

thus is the natural choice to make scanning microdevices, i.e., MEMS micromirrors. This

chapter introduces different types of MEMS micromirrors, and justifies the selection of

electrothermal actuation as the preferred choice of micromirror actuation for internal

organ OCT imaging probes. The basic operating principles, fabrication process and

characterization results of 1-D and 2-D electrothermal micromirrors are presented.

Finally, OCT imaging using these micromirrors packaged inside endoscopic probes is

also demonstrated.

3.1 Scanning Micromirrors

Rotational scanning micromirrors are widely used for a variety of applications,

such as optical displays [81, 82], biomedical imaging [20, 51, 83], barcode scanning [84,

85], optical switching [18, 86-88], and laser beam steering [85, 89]. There are numerous

commercially available MEMS scanning micromirrors ranging from Texas Instruments'

DMDs (Digital Micromirror Devices) [17] to Lucent Technologies' optical switch [18].

Most of these commercially-available micromirrors are surface micromachined and their

size is limited to about 0.1 mm due to curling that is caused by residual stresses in thin-









film structures. For biomedical imaging applications, relatively large mirrors (>0.5 mm)

are required to maintain high spatial resolution. Therefore, bulk-micromachining

processes are often used to make relatively large, flat single-crystal silicon (SCS)

micromirrors. These micromirrors can be actuated using electromagnetic, piezoelectric,

electrostatic or electrothermal techniques.

Fast scanning speeds and low power consumption make electrostatically-actuated

micromirrors the most popular amongst all scanning mirrors. Electrostatic micromirrors

can be further subdivided into two categories based on electrode placement. The first type

of mirror design uses the electrostatic force created by parallel-plate electrodes placed

underneath the mirror to generate rotation. Micromirrors using this approach have

demonstrated rotation angles of 8 at 142 V [18], 7.5 [90], and 7 at 70 V [91].

Since most of these devices are fabricated using surface micromachining techniques,

there is a trade-off between mirror-plate size and the maximum allowed rotation angle

due to the small gap size between the electrodes. Other researchers have used bulk

micromachining methods which achieve larger electrode gaps thereby permitting larger

mirror sizes; but this significantly increases the actuation voltage. Parallel-plate actuation

using bulk micromachining have yielded 2-D mirrors that rotate 5 at 160-170 V [92],

+5 at 200 V [93], and 3 at 40V [94].

Since the tradeoff between the mirror size and rotation angle limits the applications

of parallel-plate electrostatic actuation to small micromirrors, a second category of

electrostatic mirrors have been developed that use electrostatic comb fingers to rotate the

mirror plate. A number of vertical comb drive (VCD) designs based on single-crystal

silicon (SCS) have been reported for achieving larger rotation angles with large mirror









sizes [95-101]. For instance, Conant et al. reported a fast-scanning VCD micromirror by

using silicon-on-insulator (SOI) wafers [96]. Xie et al. demonstrated a curled-hinge VCD

micromirror that rotated +4.70 at 18V [97]. Patterson et al. reported a VCD design in

which photoresist re-flow was used to tilt comb fingers, but the device fabrication

uniformity and yield may be concerns [98]. Krishnamoorthy et al. used SOI wafers to

fabricate self-aligned VCD micromirrors [99]. Milanovic et al. used lateral comb drives

to generate torsional rotation [100, 101]. Kim and Lin reported an electrostatic

micromirror with a pre-tilted mirror using localized plastic deformation of silicon by

Joule heating [95]. 2-D electrostatic mirrors using comb drives have also been reported to

produce mechanical rotation angles of 5.50 at a resonance of 720 Hz and 16 V voltage

[102], up to 11 at 100 V [103], 6.20 at 55 V [104], and 10 at 140 V [101]. Although

the high resonant frequencies of electrostatic mirrors allow for high speed scanning, the

scan area is limited by the small rotation angles. Also, the high voltages required for

larger angular actuation is a deterring factor for their use in certain applications, such as

in endoscopes for internal biomedical imaging.

On the other hand, electrothermal actuation can generate large rotational

displacements at low drive voltages. Electrothermally-actuated micromirrors use thin-

film bimorph cantilevers (composed of materials with different coefficients of thermal

expansion) that are attached to a mirror plate. Joule heating of these bimorph structures

result in rotation of the mirror plate. Micromirrors based on the bending of bimorph or

multimorph structures have been reported [51, 105-108]. Metals are often used as the top

layer of bimorph structures due to their large thermal expansion coefficients and high

reflectivity. The commonly used bottom layers include silicon dioxide [51, 105, 106,









108] and silicon [107, 109-111]. Heating sources can be provided by polysilicon [51,

105], diffusion [107], or metal resistors [106, 108]. 2-D electrothermal mirrors have

reported mirror rotation of-150 at a resonant frequency of 1.3 kHz [84], and also rotation

angles as large as 400 at 15 V [30]. There is also an interesting report in which a

clamped-clamped polysilicon beam was used as the thermal actuator [112]. In this case,

the buckling of the clamped-clamped beam due to thermally-induced stress is used for

actuation, and the polysilicon beam itself functions as a thermal resistor. The

disadvantages of thermal actuation include high power consumption, relatively slow

speed and poor temperature stability. Even though electrothermal micromirrors generally

consume more electrical power than others, they are the best suited choice for some

applications that require large optical angles at low driving voltages.

Electromagnetic micromirrors rotate due to the Lorentz force generated by the

interaction of an external magnetic field with electric current flowing through a coil on

the mirror plate. Electromagnetic micromirrors have been demonstrated using metallic

coils [113-115] or magnetic materials such as Permalloy [116]. Although electromagnetic

micromirrors can achieve large rotation angles of 100 [86], 15.70 [114], and 230 [87] at

low actuation voltages, they are bulkier than other micromirrors since they require large

external magnets. Therefore it is challenging to compactly package these electromagnetic

micromirrors for applications with stringent size restrictions, such as endoscopic imaging.

Piezoelectric actuation is another mechanism that can generate large forces and

have low power consumption and high bandwidth. In piezoelectric mirrors, mirror

rotation is brought about by the piezoelectric bending of thin-film PZT

actuators/cantilevers on application of an electric voltage. Piezoelectrically actuated









micromirrors with rotation angles of 2.30 at 4.5 V [117], 2.20 at 60 V [118], 3.50 at 40 V

[119], and upto 5.5 at 16 V [88], have been reported. Even though some piezoelectric

mirrors operate at low voltages, they are limited to the area they can scan. Other

drawbacks of piezoelectric actuation include small displacements and charge leakage and

hysteresis effects which often require a feedback control loop.

As mentioned in Chapter 2, micromirrors specifically designed for use inside

endoscopic probes for internal organ biomedical imaging must meet requirements of

small size, fast scanning speed, large scan angles, and low operating voltage.

Electrothermal actuation was chosen as the preferred actuation method since it meets all

the above mentioned requirements. The following sections present the fundamentals of

electrothermal actuation, and also the fabrication process used to fabricate 1-D and 2-D

micromirrors that use electrothermal actuation to achieve large angular displacements at

low driving voltages for endoscopic optical coherence tomographic imaging.

3.2 Electrothermal Actuation and Design

All electrothermal micromirror designs described in this dissertation use the same

basic design structure which is based on electrothermal bimorph actuation. A bimorph

structure, illustrated in Figure 3-1, consists of two thin-film layers that have different

coefficients of thermal expansion. A temperature change induces stress in the two layers

due to the difference in their thermal expansion coefficients, thereby resulting in bending

of the bimorph beam. This temperature change can be brought about by resistive Joule

heating.

Although any two materials with different coefficients of thermal expansion can be

used to form bimorph structures, their mechanical properties and material compatibility










Layer 2


00 Layer 1
Anchor



Figure 3-1: Side view of a bimorph beam.

must also be considered. Table 3-1 lists some materials that may be used in conjunction

with others to form bimorph structures. Researchers have also reported bimorphs that

were formed by a layer of metal and polyimide polymer [120, 121].

Table 3-1: Thermomechanical properties of some possible bimorph materials at room
temperature.
Coefficient of Young's Specific Thermal Density
.Thermal Modulus Heat Conductivity
Material
Expansion
[10-6 /K] [1011N/m2] [103 J/kgK] [W/mK] [103 kg/m3]

Layer 1
Si 2.6 1.62 0.691 170 2.42
Si02 0.4 0.74 0.84 1.1i 2.66
Si3N4 2.8 1.55 0.711 18.5 3.19W
SiC 3.5 4.57 86.5 3.2
Poly-Si 2.3 1.60W 0.754 2.33

Layer 2
Al 23.0 0.69 0.9 235 2.692
Au 14.3 0.8 0.129 318 19.4
Pt 8.9 1.47 0.133 73 21.5
Cu 16.7 1.2 0.387 401 8.95
Ni 12.8 2.1 0.444 91 9.04
Pb 28.7 0.16 0.128 35 11.48


Material properties obtained from [110], and those marked t obtained from Memsnet.org









All electrothermal MEMS devices presented in this dissertation are fabricated using

a CMOS-MEMS process [122], therefore bimorph actuators are composed of silicon

dioxide (SiO2) and aluminum (Al) thin-film layers which are provided by the CMOS

process. As given in Table 3-1, the coefficients of thermal expansion for SiO2 and Al are

0.4x10-6/K and 23 x 106/K, respectively. It is this large difference in the thermal

expansion coefficients of the two materials that attributes to the large actuation range of

the fabricated MEMS optical scanners.

The basic structure of an electrothermal micromirror is shown in Figure 3-2. It

consists of a mirror plate attached to the substrate by a bimorph beam actuator. The

mirror plate is composed of an Al top layer which forms broadband, highly reflective

surface of the mirror, a single-crystal silicon (SCS) membrane bottom layer which adds

stiffness to the mirror to ensure surface flatness, and a SiO2 layer in between the Al and

SCS layers. The bimorph actuator is composed of a top aluminum layer, a bottom silicon

dioxide layer, and within the SiO2 layer is embedded an electrically-resistive polysilicon

layer to provide Joule heating. A unique micromachining process which can be used to

fabricate the bimorph structure shown in Figure 3-3 is detailed in the next section.

3.3 Microfabrication Process

It is a widely known fact that thin-film deposition processes generate residual stress

and stress gradients, which cause curling of the resultant thin-film microstructures.

Micromirrors made up of thin-film layers are typically small in size, in order to reduce

their optical quality degradation due to curling. Therefore, this thin-film curling limits the

useful size of micromirrors to about 10-100 |tm. In order to increase mirror sizes without

sacrificing the mirror flatness, single-crystal silicon (SCS) based mirrors are desirable. As










(a)

V Bimorph Actuator











Embedded
Poly-Si Resistor




(b) riIiTOr




Bimorph
Actuator Al
Al". SC.'


Polysilicon
Silicon
Substrate


Figure 3-2: Electrothermal micromirror basic structure. (a) Top view. (b) Cross-sectional
side view.

introduced in the previous section, the bimorph micromirrors presented here require thin-

film bimorph structures for actuation and SCS structures for large size and flatness. The

micromirrors were fabricated by a deep-reactive-ion-etch (DRIE) CMOS-MEMS process

[122]. The basic idea of this process is to introduce an SCS layer underneath CMOS

multi-layer structures in such a way that the mechanical properties are dominated by the

SCS layer, electrical connections provided by the CMOS interconnect metal layers,






48


Step 1: Backside Etch metal-3 Step 3: Deep Si Etch
r Tlv.- gon .- metal-2
--metal-1 i _- --- oxide
*- poly-Si
SCS membrane



mirror
Step 2: Oxide Etch Step 4: Si Undercut mrror bimorph
-5=T -j -beam

-40gtm -40tm
T T


Figure 3-3: DRIE CMOS-MEMS fabrication process flow.

heating source provided by polysilicon and high reflectivity by the top metal layer. This

maskless post-CMOS micromachining process has also been successfully used to

fabricate integrated MEMS accelerometers and gyroscopes [123, 124].

The process flow is shown in Figure 3-3, which is completely CMOS-compatible

and involves only four dry-etch steps. The process starts with CMOS wafers or chips that

are fabricated at virtually any CMOS foundry. As a demonstration of foundry-CMOS

compatibility, the devices presented in this dissertation were fabricated using the Agilent

0.5-pm or the AMI 0.5-[tm 3-metal CMOS processes available through the MOSIS

foundry service [125]. The first step of the post-CMOS fabrication process is to perform

a backside DRIE step to form a 30 |tm to 50 [tm-thick SCS membrane. This etch step is

carried out by the Bosch process [126 1996] on a Surface Technology Systems (STS)

inductively-coupled-plasma (ICP) etcher. The etching chemistry used is SF6/02 with the

following parameters: 600 W coil power, 12 W platen power, 130 sccm SF6 flow, 13

sccm 02 flow, and 37 mT chamber pressure. This step controls the thickness of the

microstructure and forms a cavity (-200 am deep) that allows the microstructure to move









freely in a wide range. The depth of the cavity is determined by the thickness of the

CMOS chips (which is typically around 250 [tm). The second step is a frontside

anisotropic oxide etch that uses the CMOS interconnect metal (i.e., aluminum) as an

etching mask. This oxide RIE etch is performed in a Unaxis Shuttlelock ICP etcher with

the following process conditions: 600 W coil power, 100 W platen power, 15 sccm SF6

flow, 5 sccm Ar flow, and a chamber pressure of 5 mT. The oxide etch is followed by a

deep silicon trench etch using the STS ICP etcher to release the microstructure. At the

end of this step, a 30 |tm to 50 |tm thick SCS layer remains underneath the CMOS layer,

resulting in a flat released microstructure. Finally, a brief isotropic silicon etch is

performed to undercut the silicon from under the thin-film bimorph beams. Any beam

with a half-width less than the silicon undercut will have no SCS layer underneath. This

isotropic silicon etch is attained using the same STS ICP etcher but by reducing the

platen power to 2 W. These undercut thin-film beams can be used to form electrically

isolated SCS islands, purposely curled-up structures or z-compliant springs. In the

micromirrors, these 2-[tm-thick thin-film beams form bimorph actuators with an

embedded polysilicon heater.

As the top aluminum layer is used as an etching mask, CMOS circuits under it will

remain unaffected by the fabrication process. Thus, this maskless post-CMOS process is

completely compatible with foundry CMOS processes, and CMOS circuits can be

integrated with MEMS devices. When the mirror is released from the substrate during

fabrication, the bimorph actuator is no longer constrained and will curl up. This bimorph

curling occurs due to the residual tensile and compressive stresses present in the

aluminum and silicon dioxide layers, respectively. As a result of the bimorph curvature,
















K


Figure 3-4: SEM of a fabricated 1-D micromirror with initial tilt angle.

the attached mirror tilts upward and away from the substrate with an angle equal to the

tangential angle at the end of the bimorph. The SEM of a fabricated 1-D micromirror,

demonstrating the initial tilting of the mirror plate, is presented in Figure 3-4.

3.4 Bimorph Actuation and Theoretical Analysis

Before looking into the actual electrothermal micromirrors, bimorph actuation

theory and electrothermomechanical analysis are presented in this section.

The electrothermal micromirror is actuated by applying an electrical current to the

polysilicon resistor. The electrical power dissipated by the resistor as heat raises the

temperature of the bimorph actuator. Since the top Al layer has a greater coefficient of

thermal expansion than the bottom Si02 layer, the increase in temperature causes the top

metal layer to expand more than the bottom Si02 layer. This in turn increases the radius

of curvature of the bimorph actuator by bending the bimorph in the downward direction.

Therefore, the tilt angle of the mirror decreases from its initial value. Side-view

schematics illustrating the released micromirror structure and the electrothermal actuation

mechanism are shown in Figure 3-5.











(a) (b)
Mirror

/
r Mirror Plate Tilts



SBias Applied 10
SL Polysilicon Rei edmu '

Bimorph
Silicon Actuator Silicon 'Reduced
Substrate Substrate Bimorph Curl



Figure 3-5: Bimorph actuation mechanism. Side views of: (a) Initial position of mirror at
zero bias. (b) Downward rotation of mirror plate on application of bias voltage
to polysilicon resistor.

The bimorph beam curls up after being released due to the tensile stress in the

aluminum layer and compressive residual stress in the bottom silicon dioxide layer.

Therefore, the radius of curvature of the bimorph beam is determined by both the initial

curling and also due to the temperature change from the polysilicon heating, and is given

by [105]:

1- 1 (3-1)
r ro r

where r is the actual radius of curvature, ro is the initial radius of curvature and rT is the

radius of curvature due to the temperature change. The minus sign is due to the fact that

the initial curling of the bimorph is caused by residual stresses due to cooling from high

processing temperature to room temperature, while the thermally induced curvature is

caused by thermal heating.

By ignoring the thin polysilicon layer, rT is readily derived as [127]:









1 6(tA, + tox)(aAl ao)AT

4tA + 4tx +6tAox + 4E Et (3-2)
Eoxtox E,4tt4

where AT is the temperature change of the bimorph beams, ta, aa and Ea are the

thickness, thermal coefficient of expansion, and Young's modulus of the metal (a=Al)

and oxide (a=ox) layers, respectively.

Equation (3-2) can be rewritten as

1
= fAaAT (3-3)
r,

where Aa, = a'A -ao is the difference in the coefficients of thermal expansion of Al

and SiO2, and /f is the curvature coefficient of the bimorph beam as is given by:

6(tA( + tox)
SEA 3 3x (3-4)
4t2 + 4to2 + 6tox + +
Eoxtox EAltAl

As shown in Figure 3-5, the tangential angle at the tip of a curled beam is equal to

the arc angle. Using simple geometry, we get

b6 =Lb (3-5)
rT

or 0, = 8rLbATAT (3-6)

where Lb is the length of the bimorph beam. Equation (3-6) relates the angular change of

the bimorph actuator to its temperature increase.

Using Equations (3-1) and (3-3), we can express both the initial curvature and

thermal-induced curvature in terms of the curvature coefficient as:

1
-= f8 (AE0 AaTAT) (3-7)
r









where AE0 is the linear strain difference caused by residual stress. Therefore, the actual

tilt angle of the beam tip relative to the substrate plane is given by

0(AT) = 00 -, = aLb (A0 AaAT) (3-8)

where 00 = fLbAo is the initial tilt angle. So, increasing the curvature coefficient will

simultaneously increase the actuation angle and the initial tilt angle. Sometimes large

initial tilt angle may be undesired, in which case a compromise has to be made.

Equation (3-6) gives the relationship between the actuation angle and the

temperature change which is uniform along the entire bimorph beam. However, in most

cases, the beam temperature distribution is not uniform, so the radius of curvature varies

along the beam. Therefore, the tilt angle at the tip of the bimorph is an accumulation of

the gradual curvature changes, i.e.,


0= b 1 0 dx o 0 Af a,rAT(x)dx (3-9)



or O 00-0= fLbTAcI L AT(x)dx = ,LbAcrTT (3-10)
[Lb

where AT = f L AT(x)dx is the average beam temperature difference above the
Lb

substrate (or ambient) temperature. Thus, the actuation angle is linearly proportional to

the average temperature of the bimorph beam. This is valid as long as the increased

temperature does not change the material and mechanical properties of the bimorph

layers. This analysis assumes that the width of the bimorph layers is equal for both layers,

the materials are isotropic and continuously distributed, and that the radius of curvature is

constant along the bimorph beam.









3.5 Electrothermal Micromirrors

3.5.1 One-Dimensional Electrothermal Micromirror

An electrothermal micromirror consists of a thin-film bimorph structure and a bulk-

silicon mirror plate. Instead of using a continuous bimorph mesh that was used in the

previous designs [19, 28], the bimorph actuator was split into an array of bimorph beams,

as shown in Figure 3.1. Since each beam has a relatively small width, the silicon

undercut of the structure will remove all of the silicon underneath the beams, leaving a

majority of the silicon underneath the mirror. This design was created to further improve

the micromirror scanning performance by reducing the overall stress of the bimorph upon

actuation. The buckling phenomenon observed by Xie et al. [19] is not present in this

device. Another difference in the new design is that thermal isolation regions were added

to isolate the bimorph beam array from the substrate and mirror plate regions. The

thermal isolation regions are useful for two things. First, the thermal isolation region

between the bimorph array and the substrate increases the average temperature of the

bimorph array for a given bias, yielding a greater angular response of the mirror. Second,

the thermal isolation layer between the bimorph and the mirror plate lowers the heat flux

between the two regions upon actuation, resulting in a faster thermal response time of the

bimorph. A lumped element model of this micromirror has been developed by Todd and

Xie, and interested readers may refer to [128].

A single-axis micromirror with a bimorph actuator using this beam design has been

fabricated [27]. A scanning electron micrograph (SEM) of a released device is shown in

Figure 3-6. The micromirror is 1mm by 1mm in size, coated with aluminum, and

thermally actuated by an integrated polysilicon heater.



























Figure 3-6: SEM of 1-D micromirror. Inset: Close-up of the bimorph actuator beams.

The bimorph beam array consists of 64 beams. The embedded poly-Si resistors of

adjacent beams are connected in parallel giving a total of 32 resistors embedded in the

bimorph beam array. The 32 resistors are connected in series yielding a total resistance

of 1.15 kM for the bimorph beam array. Two voltage contacts are present on the furthest

right and left resistors of the beam array. Figure 3-7 shows the measured rotation angles

at different currents and also the current dependence of the polysilicon resistor. A rotation

angle of 310 is achieved at 9 mA or 18 V. The response curve is smooth over the whole

scanning range. Thus, this mirror design also eliminates the discontinuity problem

observed in the micromirror design reported by Xie et al. [19].

The resistance of the polysilicon resistor changes significantly with current. There

are two effects attributed to the resistance change: piezoresistive effect and temperature

dependence of polysilicon resistance. The temperature coefficient of the piezoresistivity

is given by the product of the coefficient of thermal expansion and the gauge factor

which are respectively -2.5x10-6/K and -30 for polysilicon, while the temperature

coefficient of resistivity of the polysilicon used in these micromirrors was measured to be


= c=-~-'
r ~ -~LLc--
I
I


























Applied Current (mA)


4 6
Applied Current (mA)


Figure 3-7: 1-D mirror characterization. (a) Rotational static response. (b) Plot of the
heater resistance versus applied current.


40 Voltage Applied:
(0.5 + 0.5coso3t) V
0) 35
-a
^ 30
S 25
S 20 ..
C.)
S 15
o 10

O i i,
Oo5-

10 102
Frequency (Hz)

Figure 3-8: Frequency response of the 1-D mirror.









about 5.9x 10-/K. Therefore, the piezoresistive effect can be ignored. The resonant

frequency of the mirror is 380 Hz, as obtained in the frequency response of Figure 3-8.

The radius of curvature of the mirror surface is about 50 cm.

3.5.2 Two-dimensional Electrothermal Micromirror

3.5.2.1 Device design

A two-dimensional (2-D) optical scanner was designed and fabricated by extending

the 1-D mirror design concept presented in the previous section. This 2-D mirror uses a

combination of two 1-D electrothermal actuators, to provide it two-dimensional scanning

capability. The schematic drawing of this 2-D micromirror device is illustrated in Figure

3-9. The mirror is attached to a movable, rigid silicon frame by a set of bimorph

aluminum/silicon dioxide thin-film beams. As before, a polysilicon resistor is embedded

within the silicon dioxide layer to form the heater for the bimorph actuator. This movable

silicon frame is connected to the silicon substrate by another set of identical bimorph

thin-film beams that are oriented perpendicular to the first. In order to differentiate

between the two actuators, the first set of actuators that rotate the mirror is defined as the

mirror actuator, while the second set of beams that actuate the rigid silicon frame is

defined as the frame actuator. The orthogonal orientation of these two actuators results in

two perpendicular axes of rotation for the mirror plate.

As shown in the cross-sectional view of the device (Figure 3-9 (b)), the top layer of

the mirror is aluminum. Thus, the mirror has high reflectivity. A 40 |tm thick SCS layer

backing the mirror plate keeps it optically flat. The mirror plate is 1 mm by 1 mm in size.

This size is chosen for the micromirror to fit the available space in the OCT imaging

probe. Each side of the rectangular frame is 75 |tm wide, and it also has a 40 |tm thick











(a) Second Axis (b)
Mirror Actuator
..... Mirror/Frame 5 gm
-40gLm
First Axis Metal
Oxide ........0 Silicon
Silicon Poly-Si

f l Bimorph
Actuator
Frame Bimorph Beams
Actuator
Figure 3-9: Schematic of the 2-D mirror design. (a) Top view showing the axes of
rotation. (b) Cross-sectional view of A-A'.

SCS layer under it to provide rigidity to the structure.

The heating element in the bimorph beams is a set of 200 |tm long, 7 |tm wide,

polysilicon strips oriented along the beams. This is the same actuator design used by the

1-D mirror design of Section 3.5.1. The polysilicon layer from the CMOS process

permits a maximum current of 1 mA per micron width. Therefore, only a maximum

current of 7 mA can flow through the 68 Q polysilicon heater of each individual bimorph

beam. In order to increase this current limit to a higher value, the polysilicon resistors in

two adjacent beams are connected in parallel. This reduces the beam pair resistance to

34 Q and increases the maximum current to 14 mA. The fabricated mirror has 32 and 38

pairs of bimorph beams in the mirror and frame actuators, respectively. This results in

mirror and frame actuator resistances of 1.1 kM and 1.3 kM, respectively.

The SEM of a fabricated micromirror [29, 30] is shown in Figure 3-10. After

fabrication, the initial tilt angles of the mirror and frame, with respect to the substrate, are

420 and 160, respectively. These initial tilt angles are due to the residual stresses present

in the bimorph beams. The maximum actuation angles, allowed by this device design, are


























Figure 3-10: SEM of a fabricated 2-D micromirror.

limited by the substrate contact points and also by the maximum electrical current that

can be passed through the beams. Calculations based on a 500 |tm thick silicon wafer

show that the mirror can tilt up to -220, while the frame can tilt up to -170 below the chip

surface. Therefore, the maximum allowed rotation angles for the mirror and frame are

640 and 330, respectively.

3.5.2.2 Device characterization

Various experiments were performed to determine the characteristics of this device.

These experiments include static response, frequency response, long-term stability, and

thermal imaging of the device.

A simple experimental setup with a helium-neon (HeNe) laser and a dc current

source was used to measure the static deflection angles. The mechanical rotation angle of

the mirror was obtained by measuring the displacement of the reflected laser beam on a

screen. Figure 3-11 (a) shows the measured angles of rotation at different currents for the

two independent axes. The mirror rotates 400 at an applied current of 6.3mA (or 15V,

corresponding to an applied power of 95 mW), while the frame rotates by 250 at a current













--- Frame rotation at I = OmA
S 40
35-
30

(a) 25 p
20-
o-
15




0 1 2 3 4 5 6 7 8 9
Applied Current (mA)


--- Mirror resistance at 12 = 7mA
3.5 .- Frame resistance at 11 = OmA

S 3-

(b ) .

2 2-


1.5


0 1 2 3 4 5 6 7 8 9
Applied Current (mA)

Figure 3-11: 2-D Mirror Characterization. (a) Rotation angle vs. current, and
(b) Polysilicon resistance vs. current for the two actuators. Ii: current in mirror
actuator. 12: current in frame actuator. A 7 mA frame actuator current is
required for aligning the rotation axis of the mirror actuator with the substrate.

of 8mA (or 17V, corresponding to a power of 135 mW). Mirror rotation angles up to 500


have been observed at higher currents, but the high stress induced in the bimorph actuator

results in mirror instability. It has been observed that thermal damage in the polysilicon

heater occurs at this point. The mirror instability limits the usable scan range of the


mirror actuator to 400. The dc current dependence of the resistors is plotted in Figure 3-


1 (b). The resistances of the polysilicon heaters change significantly with current because







61





120


100100

880

60








Figure 3-12: Thermal images of a device biased at 10 V. (a) Temperature distribution
200 12

Actuator 200o Actuator
Length (jim) Width (jim) (0.8"C
(a) (b)

Figure 3-12: Thermal images of a device biased at 10 V. (a) Temperature distribution
across the mirror actuator only [highlighted in (b)]. (b) Thermograph of the
entire device.

the heating effect of the current causes temperature change, which in turn induces

stress change in the bimorph beams. The measured open circuit polysilicon resistances of

the mirror and frame actuators at room temperature are 1.09 kM and 1.26 kM,

respectively.

The temperature distribution on the surface of the device was observed using an

infrared thermal camera (FLIR ThermaCAM PM290). The temperature distribution

profile of the entire mirror actuator is shown in Figure 3-12(a). Figure 3-12(b) shows this

distribution over the entire device, and as expected, the mirror actuator has a higher

temperature than the frame actuator due to the thermal isolation provided by the frame.

Even though the actuator temperatures can be as high as 1200C, the mirror plate and

silicon substrate dissipate heat and remain at relatively lower temperatures (-400C). So

there will be no thermal damage to tissue during endoscopic OCT imaging.









The resonant frequencies of the mirror and frame actuator structures were measured

to be 445 Hz and 259 Hz, respectively. The radius of curvature of the mirror was

measured by a Wyko NT1000 white light profilometer to be 0.33 m. The mirror can be

made optically flatter by using a thicker SCS layer backing the mirror plate. The long-

term stability of the mirror was evaluated by scanning the mirror to steer a laser beam

onto a fixed screen. The mirror was continuously scanned at 5 Hz, and the scan length

and angular position of the reflected laser beam were monitored for over 2 million cycles.

For the entire duration of the experiment, the observed angular drift was about 0.8;

which is mostly due to fluctuations in ambient temperature.

3.5.2.3 Laser scanning experiment

To further study the scanning behavior of the 2-D micromirror, a laser scanning

experiment was performed, which simulates the 2-D transverse scanning for 3-D OCT

imaging [129]. In this experiment, a simple visual display was successfully demonstrated

by using this 2-D micromirror. The objective of this beam scanning experiment was to

scan a pixel field with the micromirror and then to illuminate the selected pixels with a

laser diode, thereby creating a projection display. The experimental setup is shown in

Figure 3-13(a). By using a microprocessor to control the mirror and laser, 4x4 pixel-

images were obtained at 10 frames per second. A sample image projected on a screen is

shown in Figure 3-13(b). An active notch filter was incorporated into the amplifier to

remove frequency content from the driving signals which could excite the mirror's

resonant vibration modes. The 4x4 pixel resolution is largely limited by attempts to

stabilize the mirror for each pixel. These techniques that were developed for a high-

resolution projection display can be directly employed to control the laser beam scanning














Scanning 2-D
Micromirror




Laser Diode


Microprocessor
Amplifiers
(a) (b)

Figure 3-13: Laser scanning using the 2-D mirror. (a) Schematic of experimental setup.
(b) 4 x 4 pixel images scanned by the micromirror.

in an endoscopic OCT system because the same basic operation of the device is required

for both systems.

Due to the large angular displacements by the two actuators, the center of the

mirror plate does not remain stationary in the vertical direction. For example, at a rotation

angle of 200 (optical angle of 400), the center of the mirror plate displaces downwards by

170 [tm. This vertical displacement of the mirror plate does not affect the working of the

laser scanning display, but needs to be accounted for during OCT imaging since it

changes the optical path length of the scanning arm of the low-coherence interferometer.

3.6 Micromirror Packaging

For endoscopic OCT imaging, the micromirrors must be packaged inside

endoscopes with diameters ranging from 3-5mm. It is proposed to use a similar package

design as used by Xie et al. for their MEMS-based endoscopic-OCT system [51]. In this

packaging scheme, the micromirror is glued onto a semicircular piece of printed circuit

board (PCB) using a thermally-conductive epoxy. A picture of this custom-built PCB









package is shown in Figure 3-14. This PCB package has flexible electrical wires soldered

to the surface gold-coated contacts. A wire bonder is used to wire-bond gold wire from

the bond pads on the micromirror chip to the gold contact area on the PCB. A

customized holder has been manufactured to hold the PCB package during this wire-

bonding step. The packaged micromirror and PCB are then placed on a machined ferrule.

This ferrule provides through holes for passage of the electrical wires from the PCB. The

ferrule with the packaged micromirror is then fitted inside a hollow endoscope tube.

There are two primary ferrule placement configurations. For forward-imaging

probes, a stationary reflective mirror is required that directs the collimated light beam

from the fiber on to the micromirror plate. The reflected light from the micromirror is

focused in tissue through an optical window located on the distal end of the probe. The

second configuration provides side-imaging OCT probes, in which the light beam exits

the cylindrical probe through its side. For this arrangement, the ferrule is fitted into the


Figure 3-14: Micromirror package. (a) Packaged micromirror on a custom PCB. (b)
Picture of the PCB package. (c) Picture of a packaged mirror alongside a US
dime coin.










endoscope tube such that the micromirror is placed at an angle of 450 with respect to the

optical fiber. The reflected light from the micromirror is scanned on to the tissue through

an optical window located on the side of the endoscope. A schematic of these

configurations is illustrated in Figure 3-15.




Tissue





(a)
Window 1 .


Control
Lines GRIN Micro-
Lens Mirror



Tissue
Ferrule Micromirror
Control
(b) Lines

Optical Fiber GRIN Window
Lens

Figure 3-15: Endoscopic OCT probe designs. (a) Side-imaging configuration. (b)
Forward-imaging configuration.

3.7 MEMS-based Endoscopic OCT Imaging

3.7.1 MEMS-based OCT System Design

The electrothermal micromirror packaged on the PCB, shown in Figure 3-14, has

been installed into a custom-made 5-mm diameter endoscope tube for in vivo OCT

imaging. The OCT system work was performed in collaboration with Dr. Yingtian Pan









and Zhenguo Wang of the State University of New York at Stony Brook who performed

the endoscope system construction and testing.

The schematic of this endoscopic OCT system is shown in Figure 3-16. A high-

power, broadband light source (AFC Technology) with an output power of 13 mW, a

central wavelength (Xo) of 1310 nm, and a full width half maximum (FWHM) spectral

bandwidth (AX) of 80 nm has been used. The coherence length that determines the axial

resolution of the OCT system is 9.7 itm. The pigtailed output from the broadband light


Rapid Scanning Optical Delay Line
r---------------------------------------




--------- ----- -------- -->



Broadband Light Fiber Reference CM
Source A ArmM odl
S c Am AO Modulator


Figure 3-16: Schematic of the MEMS-based endoscopic OCT system. CM: collimating
GRIN lens, MM: micromirror, AO: acousto-optic









source is coupled into a fiber optic Michelson interferometer, where the input light beam

is equally divided into the two arms of the interferometer using a fiber coupler.

A fiber-optic polarization controller (FPC) in the reference arm of the fiber-optic

interferometer ensures that the polarization of the exiting light beam from the non-

polarization maintaining, single-mode fiber (SMF-28) is almost linearly polarized. The

exiting light from this fiber is collimated into a 2-mm diameter optical beam by an angle-

polished GRIN lens used as a collimator (CM). A rapid-scanning optical delay (RSOD)

line utilizing a grating, lens and scanning mirror is used for axial scanning.

In the RSOD, the temporal profile of a broadband light is linearly distributed at the

Fourier focal plane of a grating-lens pair, and by placing a mirror at the focal plane and

titling it rapidly results in fast group delay. To provide a stable and appropriately elevated

Doppler frequency shift, a fiber-optic acousto-optic (AO) modulator is inserted into the

reference arm before the RSOD. In this A-O modulator, two crystals are configured with

one upshifted to 56 MHz and the other downshifted to 55 MHz to frequency modulate the

light to 2 MHz for heterodyne detection. By carefully choosing the parameters of each

component (e.g.,f= 80 mm/(p35 mm for the scan lens, g = 450 lines/mm for the

diffraction grating, 4 mm VM500 galvanometric mirror tilted at 4.20 and with 4 kHz

repetition rate, and 2 MHz A-O frequency modulation), the high-speed depth scanner

allows the acquisition of 4 K axial scans per second with an optical delay window of 2.8

mm (higher path length delay is possible by increasing the tilting angle). The high and

stable Doppler frequency shift results in increased signal to noise performance of the

signal processing electronics. Moreover, the dispersion induced by unbalanced fiber

lengths and optical components between two arms of the Michelson interferometer can be









minimized by slightly moving the grating along the optical axis, which can greatly

enhance the axial resolution as has been observed during the alignment.

The sample arm of the Michelson interferometer is connected to the fiber-optic

MEMS-based OCT endoscope through FC/APC fiber connector. The design schematic of

the forward-imaging MEMS-based OCT endoscope is shown in Figure 3-15(b). The light

from the fiber is collimated by a 0.25-pitch selfoc GRIN lens to a 1.1-mm diameter

optical beam, which is then reflected by a fixed mirror onto the surface of the tilted

micromirror. The MEMS micromirror is used for transversely scanning this light beam

onto a fixed laser doublet exit lens. The 5-mm diameter laser doublet has a focal length of

10 mm, and it focuses the light beam into a 12-jtm diameter spot at its image plane.

Figure 3-17 shows photographs of the packaged OCT endoscopes.

Since the MEMS mirror has an initial tilt angle of- 200, the custom-machined

ferrule on which the micromirror sits is tilted by about 100 to keep the reflected beam on

the center of the optical axis of the end lens. The backscattered light from tissue is

collected by the same sample-arm optical path, and the combined interferometric signal

from the sample and reference arms of the interferometer is detected by a photodetector.

The detected signal is pre-amplified using a low-noise, transimpedance amplifier (Femto

HCA-10M-100K), bandpass filtered and demodulated prior to being digitized by a 5

MHz, 12 bit A/D converter. Both depth scan and lateral MEMS scan are synchronized

with the image data acquisition via two 16-bit D/A channels. All these components have

been assembled into a readily transportable trolley console of Figure 3-18 to permit

portable OCT imaging.























































Figure 3-17: Photographs of the 5-mm diameter MEMS-based OCT endoscope at the
State University of New York at Stony Brook. Photographs by Z. Wang. Used
with permission.











pp.


Figure 3-18: Photograph of the portable, MEMS-based endoscopic OCT system at the
State University of New York at Stony Brook. Photograph by Z. Wang. Used
with permission.









3.7.2 OCT Imaging Results


Figure 3-19: Comparison of OCT with histological image. (a) OCT image, and (b)
histological image of rat bladder. Imaged by Z. Wang. Used with permission.

To demonstrate the ability of these MEMS mirrors for endoscopic OCT imaging

the packaged endoscope shown in Figure 3-17, was connected to the portable OCT

system shown in Figure 3-18. Figure 3-19 is a comparison between an OCT image and a

histology photograph of a rat bladder. The OCT image was acquired at an imaging speed

of 4 frames per second, and covers an area of 2.9x2.7 mm2 [52, 130]. The lateral and

axial resolutions are 15 |tm and 12 |tm, respectively. Figure 3-20 compares the OCT

image quality for rat bladder tissue imaged using bench-top and endoscopic MEMS-

based OCT systems.






















(a)




















(b)











Figure 3-20: Bench-top versus MEMS-based endoscopic OCT imaging of rat bladder. (a)
Bench-top OCT image. Size: 6 mm by 2.7 mm. (b) Endoscopic MEMS-based
OCT image. Size: 4 mm by 2.7 mm. Imaged by Z. Wang. Used with
permission.









As can be see in these images, the morphological details of the rat bladder wall,

e.g., the epithelium, submucosa and the upper muscularis layer are readily delineated.

Because most transitional cell carcinomas originate in the epithelium, these results

demonstrate the potential of MEMS-based endoscopic OCT for early detection and

staging of bladder cancers. Also, as a wide variety of inner organs (e.g., cervix, colon,

joints) can be accessed and imaged by front-view endoscopic OCT, the results suggest

the potential applications of this technique for noninvasive or minimally invasive

imaging diagnosis in these tissues.

3.8 Summary

This chapter reviewed the different types of micromirrors that have been reported

in literature. The selection of electrothermal actuation as the technique for micromirror

actuation for this research project was justified. Theoretical analysis about electrothermal

bimorph actuation, along with the fabrication process used to fabricate these

micromirrors was also presented. This chapter also presented 1-D and 2-D electrothermal

micromirror designs, along with experimental results. The low driving voltage and large

rotation angles of these devices make them very suitable for use in endoscopic OCT

imaging systems. A forward-imaging OCT endoscope using an electrothermal MEMS

mirror for endoscopic light steering to achieve biomedical imaging at transverse and axial

resolutions of roughly 15 [tm and 12 atm, respectively, has been demonstrated. Cross-

sectional OCT images covering an area of 4.0 x2.7 mm2 can be acquired at 2-16 frames/s

and with close to 100 dB dynamic range. The 5-mm diameter large OCT endoscope was

chosen to fully use the internal clearance of a 22 Fr endoscope. Smaller OCT scopes can

also be developed to accommodate various types of endoscopes.






74


However, some optical design and packaging issues for these micromirrors include

their unidirectional operation mode and non-stationary center of rotation of the mirror

plate. These issues can be resolved by using the novel microactuator design presented in

the next chapter.














CHAPTER 4
LARGE-VERTICAL-DISPLACEMENT MICROMIRRORS AND NON-LINEAR
OPTICAL IMAGING

The 1-D and 2-D scanning micromirrors presented in Chapter 3 provide large

rotation angles for transverse scanning in OCT imaging, but the unidirectional operation,

non-stationary center of rotation, and large initial tilt angle of those micromirrors

complicated the device packaging and optical design. These issues can be resolved by a

novel mirror design that uses two complementarily-oriented electrothermal actuators, to

keep the mirror surface parallel to the substrate and also to provide it bi-directional

scanning capability. This chapter presents a new large-vertical-displacement (LVD)

micromirror design that can perform rotational scans, as well as generate large piston

motion at low driving voltages. Out-of-plane displacement of the micromirror is provided

by a pair of electrothermal actuators. It is well known that there is large z-displacement at

the tip of a long rotational beam. The innovation of this LVD device is converting the

large tip displacement into a pure z-axis displacement of a flat micromirror. The LVD

microactuator design can potentially achieve maximum vertical displacements of a few

millimeters with millimeter-sized devices. Since this device can also perform bi-

directional scans, it can also be used in the sample arm of an endoscopic OCT system to

transversely scan the tissue surface.

Further miniaturization of the OCT system is also possible by using the large

piston-motion of LVD micromirrors to perform the millimeter-range axial scans that are

currently scanned in the reference arm of Figure 2-7. Piston-motion micromirrors are also









required by various other applications such as wave-front shaping in adaptive optics

[131], interferometry systems [132], and spatial light modulators [133]. Numerous

piston-motion actuation designs have been reported in literature. An electrostatic

deformable micromirror reported by Helmbrecht et al. displaced up to 6 |tm [131]. Lee et

al. and Kwon et al. presented devices that used electrostatic vertical comb drives (VCDs)

to demonstrate maximum static vertical displacements of 7.5 |tm [134], and 20 |tm [60],

respectively. Milanovic et al. reported an electrostatic VCD micromirror that generated

piston motion of 60 |tm, but at a high actuation voltage of 130 V [101]. Cugat et al.

reported a deformable micromirror with electromagnetic actuation that displaced up to

20 |tm for use in adaptive optics [135]. Yee et al. developed a piezoelectric micromirror

for high-precision tracking of a laser beam for high-density optical data storage [136].

Tuantranont et al. demonstrated a 2 |tm displacement using an electrothermally-actuated

trampoline-type micromirror for light phase modulation [137]. Wan et al. explored a new

type of micromirror which utilized electrocapillary actuation to push a mirror vertically

up to 8 |tm using a mercury droplet in a metal-plated microhole [138]. However, these

actuators can only generate up to a few tens of microns of vertical displacement, and

therefore cannot be used to meet the millimeter-range axial scanning requirement of OCT

systems.

In this chapter, two novel large-vertical-displacement (LVD) micromirror designs

are presented that can generate bi-directional scans, and also perform large out-of-plane

vertical displacement. The first mirror design uses one set of LVD microactuators to

perform 1-D bi-directional scans. The second mirror design uses two sets of LVD

microactuators to enable 2-D bi-directional rotational scanning.










4.1 LVD Microactuator Design

The LVD microactuator consists of two complementary electrothermal actuators in

a folded structure which allows a mirror plate to remain parallel to the substrate surface,

while still taking advantage of the large stroke lengths provided by the actuators. The

schematic drawing of the LVD micromirror is illustrated in Figure 4-1 which is very

similar to the 2-D electrothermal micromirror design which was presented in Section

3.5.2. The only difference being that the inner and outer bimorph actuators are aligned

along the same axis in the LVD micromirror, instead of orthogonal to each other as in

case of the 2-D micromirror. The inner and outer bimorph actuators are still referred to as

mirror actuator and frame actuator, respectively.




(a)
Frame Actuator
Mirror Actuator



A A'
t I L



Embedded Thermal
Poly silicon Heater Isolation
(b)
Bimorph
Mirror Actuator
Mirror


Frame L f z
p Actuator L



silicon, metal, polysilicon, -oxide


Figure 4-1: Design schematic of the LVD mirror. (a) Top view. (b) Cross-sectional view
across A-A'.




































Figure 4-2: Coventor simulations. (a) Device side-view. (b) 3-D model of the LVD
micromirror illustrating the initial curling of the bimorph actuators. The
mirror surface is parallel to the substrate plane since the curling of the two
actuators compensate each other.

FEM thermomechanical simulation was conducted using CoventorWare [139]. The

simulation results are shown in Figure 4-2, where the curlings of the two sets of bimorph

beams compensate each other resulting in a zero initial tilt. The initial elevation of the

mirror plate above the substrate plane, Zmrror, due to the curling of the thermal actuators

can be calculated from:

z,, = (L W,) sin 0 (4-1)

where Lf and Wfare the length and beam width of the frame, respectively. 0, the initial

tilt angle of the frame, can be computed from 0 = 1/r ; where I and r are the length and









radius-of-curvature of the thermal actuator, respectively. For a frame with Lf= 0.5 mm,

Wf= 40 |tm, and 0= 170, the initial rest position of the mirror Zmirror is 135 |tm. The

simulation results in Figure 4-2 show that the mirror plate is located 0.132 mm above and

parallel to the substrate plane. There is no substrate underneath the mirror plate.

The mirror and frame actuators rotate the mirror in opposite angular directions.

Therefore, there exist two basic modes of operation: (1) Bi-directional scanning by

alternatively applying voltages to the mirror and frame actuators; and (2) Large piston

motion by simultaneously applying voltages to both actuators. Equal angular rotations by

the two actuators will result in pure vertical displacement of the mirror. Large z-axis

displacement is achieved via the angular amplification due to the long arm length of the

frame.

In order to enable independent electrical excitation for each actuator, a wiring

schematic as shown in Figure 4-3 is used. The metal-1 aluminum layer on top of the

bimorph beams is electrically divided into several paths to carry the actuation current for

the inner actuator. For example, the metal-1 layer on the frame actuator has been divided


F r~niin




hinor metal-1





oxide _
V2 V1 poly-Si

Figure 4-3: Wiring schematic for the LVD actuators. Inset: Section of a frame actuator
bimorph beam showing that the mirror actuator current (il) is carried by
metal-1 layer.









into two electrical paths that carry current in and out for the mirror actuator. As seen in

the inset of Figure 4-3, the actuation current which flows though the polysilicon heater is

electrically isolated from the current flowing through the metal-1 layer by a thin oxide

layer.

4.2 1-D LVD Micromirror

4.2.1 Fabricated Device

A fabricated 0.7 mm by 0.32 mm LVD micromirror device [140-142] is shown in

Figure 4-4. The fabrication process is exactly the same as the one described in Section

3.3. This LVD device has its mirror plate elevated about 100 |tm above the silicon

substrate plane. The initial tilt angle of the frame with respect to the substrate surface is

130. The heating element in the 10-[tm wide bimorph beams is a set of 200 |tm long,

7 |tm wide, polysilicon strips oriented along the beams. The gaps between the beams are

9 |tm and used to undercut silicon to form thin-film bimorph beams. The frame actuator

and mirror actuator are constituted of 20 and 12 bimorph beams, respectively. The

measured open circuit polysilicon resistances of the mirror and frame actuators are 2400

and 3650, respectively.

The mirror plate is 190 |tm by 190 |tm. This small mirror size is just used to

demonstrate the proof of concept. Since the mirror plate is supported by bulk silicon,

much larger mirrors can be made. The quality of the mirror surface was determined using

a Wyko NT 1000 white-light optical profilometer. A line scan of surface heights across

the mirror plate is shown in Figure 4-5. The peak-to-valley surface deformations are

within 40 nm over the 190 |tm mirror plate. The optical quality of the mirror is better

than V/20 for near-infrared light.


























7,2M 0


z








Figure 4-4: SEM images of the LVD micromirror.


50
40------- ------- -----------------





o 20 - ---- ------ -- ----- -
30








1 -30- ---- ------ ----- --- ----
0 --





-30 ---------T-------- --------- ---------
-40 -
-50 -
0 50 100 150 200
Mirror Plate (1tm)

Figure 4-5: Line scan of the surface profile of the LVD micromirror.


4.2.2 Equivalent Circuit Model


Modeling an electrothermal actuator involves multiple domains (i.e., electrical,


thermal and mechanical). Since most of the actuator's electrical properties are


temperature dependent, electrothermal coupling makes the modeling very challenging


especially in the case where self-heating elements are used. For instance, in the case of a