Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Design of a MEMS-based optical accelerometer with large measurable range and high sensitivity Zeng, Yiyi 2008

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

Item Metadata

Download

Media
24-ubc_2008_fall_zeng_yiyi.pdf [ 2.66MB ]
Metadata
JSON: 24-1.0066397.json
JSON-LD: 24-1.0066397-ld.json
RDF/XML (Pretty): 24-1.0066397-rdf.xml
RDF/JSON: 24-1.0066397-rdf.json
Turtle: 24-1.0066397-turtle.txt
N-Triples: 24-1.0066397-rdf-ntriples.txt
Original Record: 24-1.0066397-source.json
Full Text
24-1.0066397-fulltext.txt
Citation
24-1.0066397.ris

Full Text

DESIGN OF A MEMS-BASED OPTICAL ACCELEROMETER WITH LARGE MEASURABLE RANGE AND HIGH SENSITIVITY  by  YIYI ZENG B. ENG., Tsinghua University, 2005  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE STUDIES (Electrical and Computer Engineering)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) April 2008  ©Yiyi Zeng, 2008  Abstract  MEMS Accelerometers are broadly used in the area of vibration sensor. Their applications range from seismic disturbances, to automotive industry such as airbag systems, active suspension, and smart braking. Traditionally, the acceleration is detected electrically by measuring either capacitive variations or piezoelectric signals. Those approaches suffer from a number of drawbacks, such as low sensitivity due to low signalto-noise ratio (SNR), small dynamic range, high temperature sensitivity, etc. In this thesis, a MEMS-based optical accelerometer is designed and analyzed. The device can be fabricated on a silicon-on-insulator (SOI) wafer, on which a double-leg single-mode optical rib waveguide is used to propagate 1.55μm laser beam. The device integrates the waveguide with a mechanical oscillator, and is able to detect in-plane vibrations of the oscillator by taking advantages of optical interference. According to the analysis, the maximum working range of the oscillator can be as large as 50μm and the acceleration sensitivity can be below 1μg/Hz1/2. Device fabrication and characterization are also carried out and described in the thesis. All necessary fabrication steps and details as well as characterization setups are given. Due to several fabrication challenges in UBC (e.g. malfunctioned equipment), a complete device has not been fabricated. More fabrication and characterizations are to be continued as future work.  ii  Table of Contents Abstract ............................................................................................................................... ii Table of Contents............................................................................................................... iii List of Tables ...................................................................................................................... v List of Figures .................................................................................................................... vi Acknowledgements........................................................................................................... xii Chapter I. Introduction.................................................................................................... 1 1.1 Overview and Motivation ..................................................................................... 1 1.2 Capacitive Accelerometers ................................................................................... 3 1.3 Piezoelectric Accelerometers................................................................................ 8 1.4 Optical Accelerometers....................................................................................... 10 1.5 Laser Doppler Vibrometer (LDV) ...................................................................... 12 Chapter II. Device Design and Simulations ................................................................... 16 2.1 Device Overview ................................................................................................ 16 2.2 Wafer Profile....................................................................................................... 18 2.3 Capacitive Accelerometer Design....................................................................... 20 2.4 Optical Waveguide Design ................................................................................. 22 2.4.1 Geometric Design ........................................................................................ 22 2.4.2 Layout Design.............................................................................................. 24 2.4.3 BeamPROP Simulations .............................................................................. 25 2.4.4 Coupling Loss Calculation........................................................................... 30 2.5 Device Performance............................................................................................ 34 2.6 Noise and Sensitivity .......................................................................................... 38 Chapter III. Device Fabrication ................................................................................... 43 3.1 General Process Flow ......................................................................................... 43 3.2 Ohmic Contact .................................................................................................... 45 3.3 Dry Etch .............................................................................................................. 49 3.3.1 Electron Cyclotron Resonance (ECR) Plasma Etch .................................... 49 3.3.2 Deep Reactive Ion Etch (DRIE) .................................................................. 55 3.4 Sacrificial Etch.................................................................................................... 61 Chapter IV. Characterization and Future Work ........................................................... 64 4.1 Characterization .................................................................................................. 64 4.1.1 Waveguide Characterization........................................................................ 64 4.1.2 Oscillator Characterization with LDV ......................................................... 65 4.1.3 Device Characterization............................................................................... 67 4.2 Conclusion and Future Work .............................................................................. 69 iii  References......................................................................................................................... 71 Appendix A: Code of coupling loss calculation ............................................................... 76  iv  List of Tables  Table 3.1 Commonly used contact metals and their barrier heights when contacting with a p-type Si……………………………...…………………………………………………..47  Table 3.2 Comparisons between ECR etch in UBC cleanroom and DRIE in University of Sherbrooke……………………………………………………………………………….60  v  List of Figures  Figure 1.1 A basic spring-mass system which can be used as an accelerometer………….3  Figure 1.2 A differential capacitive sensing setup…………...……………………………4  Figure 1.3 Effective electric circuit of the differential capacitive sensing setup……...…..5  Figure 1.4 A compression disk made of piezoelectric material sandwiched between two electrodes………………………………………………………………………………….8  Figure 1.5 Basic principle of a piezoelectric accelerometer………………...…………….9  Figure 1.6 Working principle of a light intensity modulation based optical accelerometer…………………………………………………………………………….10  Figure 1.7 Working principle of a folded-pinwheel interferometric accelerometer......…11  Figure 1.8 Principle of laser Doppler vibrometry. The reference beam is modulated by a Bragg cell in order to determine the moving direction……...…………………………...13  Figure 1.9 A Laser Doppler Vibrometer stage………………………………...…………15  vi  Figure2.1 Top view of a basic optical accelerometer……...…………………………….16  Figure 2.2 Cross-sectional view of the optical accelerometer. The reflection wall is not included in this figure……………………………………………………………………17  Figure 2.3 Resistivity versus doping concentration of silicon. The top blue curve represents p-type Si, and the bottom red curve represents n-type Si…………………….19  Figure 2.4 Sheet resistance versus doping concentration of silicon. The top blue curve represents p-type Si, and the bottom red curve represents n-type Si…………………….19  Figure 2.5 Top view of the capacitive accelerometer……………………………...…….20  Figure 2.6 Cross-sectional view of the capacitive accelerometer………………...……...20  Figure 2.7 Simulation results showing the resonance frequencies of the oscillator. Both amplitude and phase are simulated. In the amplitude plot, the top green curve corresponds to y-axis (i.e. in-plane) oscillation, and the bottom blue curve corresponds to z-axis (i.e. vertical) oscillation……………………………………………………………………….21  Figure 2.8 Cross-sectional view of rib waveguide in SOI………………...……………..22  Figure 2.9 Critical a/b value versus b for an SOI rib waveguide…………..……………23  vii  Figure 2.10 Layout of the Y-shape rib waveguide. The thin lines represent ribs, which have a uniform width of 9μm. The total length for each path is 9mm, and the separation angle is 2.5º for both legs………………………………………………………...………24  Figure 2.11 Output profile of the rib waveguide. The tilting angle is 8.1º which is same as that of typical APC fibers……………………………………………...………………...25  Figure 2.12 Mode profiles after 4000μm propagation for the rib waveguide designed above. (a) Fundamental fiber mode. (b) and (c) High-order modes……………………..26  Figure 2.13 Propagation loss vs. distance for the designed rib waveguide. (a) Fundamental fiber mode. (b) and (c) High-order modes…..…………………………….27  Figure 2.14 Mode profiles after 4000μm propagation for a similar rib waveguide with 2μm slab height. (a) Fundamental fiber mode. (b) and (c) High-order modes…………..28  Figure 2.15 Propagation loss vs. distance for the similar rib waveguide with 2μm slab height. (a) Fundamental fiber mode. (b) and (c) High-order modes……................…..…29  Figure 2.16 A typical Gaussian beam with a waist w0…..………………...…………….30  viii  Figure 2.17 Illustration shows the Gaussian beam reflected by the movable reflector couples into the waveguide. The coupling loss is a function of propagation distance of the beam……………………………………………………………………………………..31  Figure 2.18 Coupling loss versus distance between proof mass and waveguide output...33  Figure 2.19 Detector power versus oscillator position………….……………………….34  Figure 2.20 Plot of the first three Bessel functions of first kind for integer orders α…....36  Figure 2.21 Detector current versus input power for a typical photodiode. The solid line represents signal, and the dash-dot line represents noise………………………………...40  Figure 2.22 Detector current versus input power at m=1.5×10-4 for a 10mW laser. The solid line represents signal, and the dash-dot line represents noise……………...……....41  Figure 3.1 The energy band diagram of metal, p-type semiconductor, and the contact between the two in equilibrium…………………………………………………………..45  Figure 3.2 I-V curve of typical Ohmic and Schottky contacts…………...…...…………46  Figure 3.3 Voltage drop versus bias current for a 30nm/200nm Cr/Cu contact….…...…48  ix  Figure 3.4 A 70nm Al layer is deposited on Si using lift-off technique. The minimum feature size is 3μm……………………………………………………………………….50  Figure 3.5 Top view of proof mass and comb-drive fingers after ECR etch……...……..52  Figure 3.6 ECR Etch profile scanned by Atomic Force Microscope (AFM)…...……….52  Figure 3.7 ECR Etch profile observed by Scanning Electron Microscope (SEM) in UBC cleanroom. The etch depth is 4.5um in 50 minutes, resulting in 90nm/min etch rate…...53  Figure 3.8 Bosch process illustration. (a) Passivation mode: Teflon-like fluorocarbon polymer film is deposited on Si sample. (b) Etching mode: the horizontal part of the polymer film is removed by ionic bombardment, and Si is isotropically etched by reactions with fluorine species. (c) Next passivation mode: the resulting sidewalls consist of a series of very small scallops shape………………………………………………….56  Figure 3.9 Microloading effect and notching effect. The notching effect has happened for large features, while etching has not reached the SiO2 layer for small features…………57  Figure 3.10 A 13um deep vertical Si etch done by Bosch process in University of Sherbrooke. The profile is observed by SEM in UBC cleanroom……………………….59  x  Figure 3.11 Around 2μm SiO2 is etched away by BOE in 55 minutes with almost no damage to Si. The etch profile is observed by SEM in UBC cleanroom………………...61  Figure 3.12 SEM image of a device after the first dry etch and BOE release. The picture is taken in Simon Fraser University (SFU). The etching is done by DRIE in Sherbrooke.........................................................................................................................62  Figure 3.13 Another SEM image of the device after the first dry etch and BOE release. The dry etch is done by Bosch process in University of Sherbrooke. The picture is taken in SFU……………………………………………………………………………………63  Figure 4.1 Schematic view of optical waveguide characterization stage…………..……64  Figure 4.2 Optical waveguide characterization stage…………...……………………….65  Figure 4.3 Schematic view of oscillator characterization setup. The z-axis vibration is measured by LDV………………………………………………………………………..66  Figure 4.4 Oscillator characterization stage…...…………………………...…………….67  Figure 4.5 Schematic view of device characterization setup……………………….........68  xi  Acknowledgements My deepest appreciation goes to my supervisor Dr. Lukas Chrostowski, and Dr. Edmond Cretu, for proposing this project and providing continuing support during my research. Their friendly and constructive guidance has been the key to every success of this project. I would also like to thank Mario Beaudoin and individuals who offered me trainings and helps in the cleanroom, thank Julie Verstraeten and Dr. Vincent Aimez for their help in fabrication at University of Sherbrooke, and thank Miguel Angel GuillenTorres, Mrigank Sharma, Akila Kannan for their helpful discussions and cooperation regarding my research. I would like to thank Dr. Nicolas Jaeger and Dr. Mu Chiao for kindly providing simulation and characterization tools, and thank all the individuals in Dr. Lukas Chrostowski’s research group, department office, and purchasing office. Finally, I would like to thank my father (Ange Zeng), my mother (Yanmou Mo), and my girlfriend (Li Liu) for their love, trust, and mental support.  xii  Chapter I. 1.1  Introduction  Overview and Motivation Accelerometers are sensors that measure acceleration, which are broadly used in  automobile suspension systems, vehicle air bag systems, anti-lock brake systems (ABS), vibrometers, and hard disk drives, etc [1]-[3]. Microelectromechanical Systems (MEMS) is a technique for fabricating mechanical and electrical devices through the semiconductor process. The advantages of MEMS accelerometers include large scale production, lower costs, and small size and weight. There are three main kinds of accelerometers: capacitive accelerometers, piezoelectric accelerometers and optical accelerometers. A capacitive accelerometer measures capacitance of electrodes which changes as a function of acceleration. A piezoelectric accelerometer measures electrical charge signal produced when a piezoelectric material is stressed, which is proportional to vibration acceleration. Both of them detect the acceleration electrically, and therefore suffer from low signal-to-noise ratio (SNR), which limits the sensitivity. They also have respective inherent drawbacks such as low dynamic range for capacitive sensors, high temperature sensitivity of piezoelectric disks, etc. On the other hand, a variety of optical accelerometers have been developed, most of which are based on light intensity modulation, and none of them is designed for detecting in-plane vibrations. Optical interferometry and Doppler Effect have also been used to detect vibration with high sensitivity. However, the device is not compact and potable, and is not able to detect in-plane vibrations either. Therefore, the goal of this  1  project is to develop a highly sensitive MEMS-based optical accelerometer which is able to detect in-plane acceleration. Instead of light intensity modulation, optical interferometry and phase (frequency) modulation are used for the detection.  2  1.2  Capacitive Accelerometers Capacitive accelerometers detect displacements and accelerations by measuring  capacitance variations. Firstly, consider a basic spring-mass system shown in Figure 1.1.  Figure 1.1 A basic spring-mass system which can be used as an accelerometer [32].  According to Hooke’s law and Newton’s second law of motion, we have JG G G G F = − kx = m  x + a0  (  )  Therefore we obtain a differential equation G k G G  x + x + a0 = 0 m  The solution of this equation is given by  G G JJG a0 x = x0 cos ω0t − 2  ω0  JJG G G G  x = −ω02 x0 cos ω0t = −ω02 x − a0 where  3  ω0 =  k m  is the resonance frequency of the system. It can be seen that, if the resonance frequency is known, the acceleration at which the system is excited can be determined by detecting the maximum displacement of the system. This is the basic detection principle for capacitive accelerometers. It also can be seen that, compliant spring or large proof mass results in low resonance frequency, which improves the sensitivity. Figure 1.2 shows a simple differential capacitive sensing scheme. Comparing with single-sided capacitive sensing, differential measurement scheme is more reliable because it is able to cancel out common voltage noise fluctuation and nonlinear terms. The effective electric circuit is shown if Figure 1.3.  Figure 1.2 A differential capacitive sensing setup [32].  4  Figure 1.3 Effective electric circuit of the differential capacitive sensing setup [32].  Voltage VS and –VS are applied on two ends of the circuit. Assume VS = VDC + Vac sin ωt For a capacitance C, the impedance is given by Z =−j  1 ωC  Therefore, the output voltage VO is given by −j VO = VS − 2VS  = VS − 2VS =  −j  1 ωC1  1 1 −j ωC1 ω C2  C2 C1 + C2  C1 − C2 VS C1 + C2  5  On the other hand, the relationship between capacitance and the gap between two plates is given by C=  εA d  Assuming the original distance is d0 for both capacitors and the original capacitance is C0, for a mass displacement of Δd, the capacitance C1 and C2 are given by C1 =  εA d 0 − Δd  C 2 = C0  = C0  d0 d 0 − Δd  d0 d 0 + Δd  If the displacement is small enough, we obtain ⎛ d0 d0 ⎞ − C1 − C2 = C0 ⎜ ⎟ ⎝ d 0 − Δd d 0 + Δd ⎠ 2d Δd = C0 2 0 2 d 0 − Δd ≈ C0 C1 + C2 = C0  2 Δd d0 2d 02 ≈ 2C0 d 02 − Δd 2  Therefore, the output voltage can be denoted by VO =  C1 − C2 Δd VS ≈ VS C1 + C2 d0  which means the output voltage is linear proportional to the displacement. Therefore, the displacement as well as the acceleration can be determined by detecting the voltage amplitude. The device sensitivity is limited due to small signal and relatively large noise coming from electric circuits. The signal-to-noise ratio (SNR) can not be improved by  6  amplifiers. Besides, the parasitic capacitance on electrodes makes it difficult to detect device capacitance variations. “Silicon Designs” manufactured a MEMS capacitive accelerometer with 2000ng/Hz1/2 noise floor at 600Hz oscillating frequency, which is by far the most sensitive commercial capacitive sensor [33]. For the sensing scheme shown in Figure 1.2, the gap between two capacitor plates can not be very large in order to form detectable capacitance, so the dynamic range is limited, typically a few microns. On the other hand, for a narrow gap, although the sensitivity can be high because the capacitance is inversely proportional to the distance, the small displacement approximation might not be satisfied, resulting in poor justification of accelerations. In another typical sensing scheme, the proof mass moves laterally, so the capacitance and the output are always linearly proportional to the displacement, and the dynamic range can be much larger. However, the sensitivity is compromised in this case.  7  1.3  Piezoelectric Accelerometers Another commonly used accelerometer is called piezoelectric accelerometer, in  which the active element is a piezoelectric material. Figure 1.4 shows a typical compression disk made of piezoelectric material sandwiched between two electrodes.  Figure 1.4 A compression disk made of piezoelectric material sandwiched between two electrodes [34].  A force F applied perpendicular to the disk causes a charge production q and a voltage u at the electrodes, which are given by q = d 33 F u=  d33 d F e33 A  where d33 and e33 are piezoelectric constants, d and A are thickness and surface area of the disk.  8  A basic piezoelectric accelerometer is formed if one side of such a disk is attached to a seismic mass and the other side is connected to a sensor, as is shown in Figure 1.5. When the device is subjected to vibration, a force is generated which acts on the piezoelectric disk. Due to the piezoelectric effect, a charge output is generated, which can be denoted as q = d33 F = d33 ma It shows the output charge signal is linear proportional to the acceleration of the mass.  Figure 1.5 Basic principle of a piezoelectric accelerometer [34].  Comparing with capacitive accelerometers, the piezoelectric accelerometer has a better sensitivity, which can be around 1000ng/Hz1/2 [34]. However, the piezoelectric device doesn’t response to very low oscillating frequency or constant accelerations. Besides, it is sensitive to temperature variations [34], which is another drawback.  9  1.4  Optical Accelerometers Optical accelerometers based on light intensity modulation have been largely  developed [35][36]. Figure 1.6 shows a device structure developed by J. A. Plaza, et al. A waveguide is defined on the movable mass, and is aligned with the two waveguides on the frames. Therefore, the transmitted light intensity at the output will become lower when there is an offset of the mass, and the intensity loss is a function of the offset, which can be used to determine the acceleration. The device is reported to give 2.5dB loss after 1μm vertical displacement. The minimum detectable acceleration is not reported. This device is designed for detecting vertical vibrations.  Figure 1.6 Working principle of a light intensity modulation based optical accelerometer [36].  Figure 1.7 shows another approach made by N. C. Loh. When the laser source shines on the interdigitated fingers, the dominant one reflected from the grating is the 0th mode. As the tip gets out-of-plane displaced, the interference between the light reflecting  10  off the reference fingers and the moving fingers causes the 0th mode intensity to decrease, while the 1st mode is enhanced. When the cantilever is deflected by the amount of λ/4, the 0th mode is minimized and the 1st mode is maximized .The cantilever deflection can be measured from the intensity of the 0th mode, 1st mode or the difference between the two modes. The device is reported to have a nano-g resolution with an 80Hz resonant proof mass. The intensity of the diffracted modes depends on the out-of-plane offset between the two sets of fingers and is given by, ⎛ 2π d ⎞ I ( d ) = I 0 sin 2 ⎜ ⎟ ⎝ λ ⎠  Figure 1.7 Working principle of a folded-pinwheel interferometric accelerometer [35].  11  1.5  Laser Doppler Vibrometer (LDV) Optical interferometry has been commercially used to detect vertical vibrations.  The optical detection takes advantages of optical interference and the Doppler Effect, and the device is called “Laser Doppler Vibrometer”. The principle of LDV is shown in Figure 1.8. It requires two coherent laser beams from a same laser source to overlap, generating interference patterns. Assuming the light intensities are I1 and I2 respectively, the detected intensity is not only a sum of two intensities, but also contains an “interference” term whose intensity is determined by the phase difference between the two beams. The detected intensity can be written as I total =  1 1 I1 + I 2 + I1 I 2 cos ⎡⎣ 2π ( l1 − l2 ) λ ⎤⎦ 2 2  where l1-l2 is the path length difference between the two beams. If the two beams have the same intensity and the path length difference is an integer multiple of the wavelength, the total intensity becomes 2 times of each one, referred as bright; if the difference contains half of the wavelength, the total intensity becomes zero, referred as dark. As the object is moving, bright and dark patterns are generated on the detector. One complete bright-dark changing on the detector corresponds to a displacement of exactly half of the wavelength used. If He-Ne laser is used, this corresponds to a displacement of 316nm.  12  Figure 1.8 Principle of laser Doppler vibrometry. The reference beam is modulated by a Bragg cell in order to determine the moving direction [5].  The Doppler Effect can be used to determine the velocity of the moving object. The effect is the shift in frequency and wavelength of waves, which results from a source or an observer moving respect to the medium, or even a moving medium. For waves traveling at the speed of light, the relationship between observed frequency f’ and emitted frequency f is given by f′= f +  fv v = f+ c λ  where v is the moving velocity. If the object is moving toward the source, the frequency becomes higher, called “blue shift”; otherwise, the frequency becomes lower, called “red shift”. When the measurement object is moving at a velocity v, the optical intensity on detector can be written as  13  1 1 I1 + I 2 + I1 I 2 cos [ 2π vt λ ] 2 2 1 1 = I1 + I 2 + I1 I 2 cos [ 2πΔft ] 2 2  I total =  where the modulation frequency Δf =  v  λ which is linearly proportional to the velocity. This means there will be a frequency shift on the spectrum, and the shift amount is same as Doppler frequency shift. The absolute value of the velocity can thus be determined. However, the moving direction is not able to be determined in the case described above. To determine the moving direction, the reference beam needs to be modulated by a Bragg cell shown in Figure 1.8 so that the optical frequency is shifted by an amount equal to the modulation frequency F. In this case, the peak will appear at the frequency f peak = F −  v  λ  on the detector. Therefore, different moving directions result in different directions of frequency shift regarding F, and the velocity value is given by the shift amount. An actual LDV stage is shown in Figure 1.9.  14  Figure 1.9 A Laser Doppler Vibrometer stage.  The LDV is able to optically measure the vibration out of the plane with subangstrom sensitivity. However, it has not been a compact device. Therefore, to achieve a compact device which is able to detect vibrations sensitively is highly desirable, and is the ultimate goal of this project.  15  Chapter II. 2.1  Device Design and Simulations  Device Overview A basic optical accelerometer consists of a traditional capacitive accelerometer, a  fixed reflection wall, and a Y-shape optical rib waveguide, which is shown in Figure 2.1.  Figure2.1 Top view of a basic optical accelerometer.  16  The presence of a capacitive accelerometer enables both electrical actuation and measurement which can be used to calibrate the optical readouts. The two branches of the Y-shape optical coupler separated at a same angle, which divides the optical signal equally in theory. The beam reflected by the movable mass meets the one reflected by the fixed wall, generating interference patterns which are detected by the external measurement system. The silicon-on-insulator (SOI) wafer is used in this project for several reasons: 1) Silicon is transparent to 1.55μm wavelength, which is one of the typical wavelengths for semiconductor lasers. 2) The contrast of refractive indices between silicon and the silicon dioxide (SiO2) insulator is very high, which better confines optical signals in the waveguide. 3) SiO2 layer can be easily removed to release the movable structures, keeping the device undamaged. The cross-sectional view of the device is shown in Figure 2.2.  Figure 2.2 Cross-sectional view of the optical accelerometer. The reflection wall is not included in this figure.  17  2.2  Wafer Profile An SOI wafer is used in this project. A typical SOI wafer consists of a Si device  layer on top, a buried oxide (BOX) layer underneath, and a Si substrate. Device is fabricated in the top Si layer, and the movable structures can be released in the end by etching away the BOX layer. The specifications of the wafer used in this project are listed below: Type: p-type Crystal orientation: <100> Wafer thickness: 625μm Top Si thickness: 13μm BOX thickness: 1000nm Minimum resistivity: 1Ω·cm Maximum resistivity: 30Ω·cm The resistivity indicates the doping concentration of the wafer, which is shown in Figure 2.3. With the resistivity range given above, the doping concentration of the wafer is calculated to be 4.47×1014~1.50×1016cm-3, which means this is a lightly doped p-type SOI wafer. The relationship between doping concentration and sheet resistance for a 13μm thick silicon layer is shown in Figure 2.4, and the sheet resistance for this wafer is calculated to be 0.77~23.07kΩ/sq [6].  18  Figure 2.3 Resistivity versus doping concentration of silicon. The top blue curve represents p-type Si, and the bottom red curve represents n-type Si.  Figure 2.4 Sheet resistance versus doping concentration of silicon. The top blue curve represents p-type Si, and the bottom red curve represents n-type Si.  19  2.3  Capacitive Accelerometer Design The capacitive accelerometer layout is shown in Figure 2.5. The proof mass,  comb drive fingers and four supporting beams need to be released eventually to achieve movable mass and capacitive measurement of displacement. Metal contacts will be made on contact pads. The cross-sectional view of the device is shown in Figure 2.6.  Figure 2.5 Top view of the capacitive accelerometer.  Figure 2.6 Cross-sectional view of the capacitive accelerometer. 20  The oscillator performance is simulated using a 3D MEMS and semiconductor software “Coventor”, done by Mrigank Sharma. The in-plane resonance frequency of the movable structure is 18.166 kHz according to the simulation, as is shown in Figure 2.7.  Figure 2.7 Simulation results showing the resonance frequencies of the oscillator. Both amplitude and phase are simulated. In the amplitude plot, the top green curve corresponds to y-axis (i.e. in-plane) oscillation, and the bottom blue curve corresponds to z-axis (i.e. vertical) oscillation.  21  2.4  Optical Waveguide Design The ideal waveguide in this project should have several features: 1) Able to keep  single optical mode inside which is crucial to optical interferometry; 2) Small propagation and coupling losses; 3) Reasonably large dimensions in order to result in small output beam divergence. It is known the core size of a single-mode fiber is around 9μm, so the waveguide dimensions should be comparable in order to reduce the coupling loss. Moreover, for the Gaussian beam, larger beam waist results in smaller divergence, so the preferable waveguide dimensions are as large as possible. However, for a typical rectangular waveguide in SOI, the dimensions cannot be 9μm or even half of it in order to ensure single propagation mode. Therefore, the waveguide geometry has to be carefully designed to achieve all the three features.  2.4.1 Geometric Design It is indicated that rib waveguides are able to propagate a single optical mode with relatively large geometric dimensions, typically several microns [7]-[12]. The crosssectional view of a rib waveguide in SOI wafer is shown in Figure 2.8.  Figure 2.8 Cross-sectional view of rib waveguide in SOI.  22  The waveguide consists of a rib with 2aλ width and 2bλ height, and two wings on both sides with a height of 2brλ, where λ is the wavelength. With an optimized geometric design, only the fundamental optical mode can be kept in the rib area, since the highorder modes will gradually leak through the wings when propagating in the waveguide, and eventually do no exist in the rib area. For a specific height ratio r, the single-mode condition is given by a r ≤ 0.3 + b 1− r2  in the limit of large b [7]. The critical values of a/b as a function of b for r=0.5 and r=0.8 are shown in Figure 2.9. The shade area indicates the single-mode region for r=0.5.  Figure 2.9 Critical a/b value versus b for an SOI rib waveguide.  Based on this theory, the designed waveguide dimensions are as follows, Rib width: 9μm Rib height: 13μm Wing height: 6.5μm  23  Given the wavelength 1.55μm, the values of a and b are calculated to be 2.90 and 4.19, respectively. The ratio a/b=0.69, r=0.5. This design meets the single-mode condition, therefore should be able to propagate only the fundamental mode.  2.4.2 Layout Design To fabricate a rib waveguide, two etch steps are required. In the first step, the whole waveguide area is etched down to the BOX layer, which is 13μm; in the second step, the wing areas are etched by 6.5μm, leaving the rib protected. The Y-shape waveguide layout is shown in Figure 2.10.  Figure 2.10 Layout of the Y-shape rib waveguide. The thin lines represent ribs, which have a uniform width of 9μm. The total length for each path is 9mm, and the separation angle is 2.5º for both legs.  The total length for each path is 9mm, which ensures sufficient leaks of highorder modes. On the other hand, the length does not exceed the coherent length of the laser source, which can reach some 100m for semiconductor lasers [37], and therefore the interference would be significant. The separation angle is 2.5º for both legs, and equal power separation and small separation loss are expected from this design. The width of each wing decreases to 20μm at the output due to the space limitation as it comes close to the movable structure. 24  The output profile is shown in Figure 2.11. It is designed to be angular in order to minimize the unwanted reflection signal from the interface.  Figure 2.11 Output profile of the rib waveguide. The tilting angle is 8.1º which is same as that of typical APC fibers.  2.4.3 BeamPROP Simulations To verify single-mode condition described in 2.4.1, beam propagation is simulated by “RSoft BeamPROP” for the rib waveguide designed above. Both fundamental mode and high-order modes are excited at the input, and the mode profiles and powers in the rib region are monitored. The input modes are set to be a little bit offaxis which is close to the real case. The refractive indices for Si and SiO2 are set to be  25  3.518 and 1.5277 [13]-[14]. The simulation results are shown in Figure 2.12 and Figure 2.13, respectively.  (a)  (b)  (c)  Figure 2.12 Mode profiles after 4000μm propagation for the rib waveguide designed above. (a) Fundamental fiber mode. (b) and (c) High-order modes.  26  (a)  (b)  (c)  Figure 2.13 Propagation loss vs. distance for the designed rib waveguide. (a) Fundamental fiber mode. (b) and (c) High-order modes.  The results show that after 18000μm propagation, which is the total path length in the waveguide, the fundamental mode is still confined in the rib region, and the propagation loss is kept to be less than 3dB. On the other hand, high-order modes have  27  leaked through the wings within 4000μm propagation, so the rib region is very lossy for them. As a comparison, a rib waveguide with same rib region dimensions but only 2μm slab width is also simulated, and the results are shown in Figure 2.14 and Figure 2.15, respectively.  (a)  (b)  (c)  Figure 2.14 Mode profiles after 4000μm propagation for a similar rib waveguide with 2μm slab height. (a) Fundamental fiber mode. (b) and (c) High-order modes. 28  (a)  (b)  (c)  Figure 2.15 Propagation loss vs. distance for the similar rib waveguide with 2μm slab height. (a) Fundamental fiber mode. (b) and (c) High-order modes.  It can be seen that both fundamental mode and high-order modes are confined in the rib region, and the clear interference signal may not be obtained in this case. The simulation results verify the single mode operation for the designed rib waveguide.  29  2.4.4 Coupling Loss Calculation One of the main losses during the laser beam propagation comes from couplings. The couplings include fiber to waveguide, waveguide to fiber, and waveguide to waveguide when beam is reflected by the movable structure or the fixed reflection wall. Because the proof mass is moving during the measurement, the last coupling loss varies as a function of distance between waveguide output and the oscillator. This loss also determines the measurement range of the device. The relationship between this coupling loss and the distance is calculated in this section. It is approximated that the actual laser beam in free space is a Gaussian beam whose transverse electric field and intensity distributions are described by Gaussian functions. A typical Gaussian beam is shown in Figure 2.16.  Figure 2.16 A typical Gaussian beam with a waist w0.  Figure 2.17 illustrates the coupling where the laser beam coming out of the waveguide is reflected by the oscillator and hits back into the waveguide. The ratio  30  between output optical power and the reflected power that will be confined in the waveguide region defines the coupling loss.  Figure 2.17 Illustration shows the Gaussian beam reflected by the movable reflector couples into the waveguide. The coupling loss is a function of propagation distance of the beam.  It is approximated that the beam waist is located right at the waveguide output. According to the properties of Gaussian beam, the spot radius for both dimensions are wx(z) and wy(z), and are given by  31  wx ( z ) = w0 x 1 +  z2 z02x  z2 1+ 2 z0 y  wy ( z ) = w0 y  where the Rayleigh ranges for both dimensions z0x and z0y are given by  π w02x z0 x = λ π w02y z0 y = λ The beam intensity at a specific position is then given by ⎧⎪ ⎡ x 2 ⎡ w0 x ⎤ ⎡ w0 y ⎤ y 2 ⎤ ⎫⎪ + 2 I ( x, y , z ) = I 0 ⎢ ⎥ exp ⎨−2 ⎢ 2 ⎥⎬ ⎥⎢ ⎪⎩ ⎣⎢ wx ( z ) wy ( z ) ⎦⎥ ⎪⎭ ⎣ wx ( z ) ⎦ ⎣⎢ wy ( z ) ⎦⎥ Therefore the beam power is given by  Ptotal = ∫  ∞  1 ∫ I ( x, y, z ) dxdy = 2 π I w ∞  0  −∞ −∞  0x  w0 y  Assuming the beam waists w0x and w0y are equal to half of the width and height of the rib region, and are located at the output interface. Thus, the coupling loss in dB can be calculated by Loss ( z ) = 10 log  Ptotal w0 y  w0 x  − w0 y  − w0 x  ∫ ∫  I ( x, y, 2 z ) dxdy  The parameter 2z written on the right side corresponds to the double pass of the laser beam through the distance. The loss function is shown in Figure 2.18. The curve shows that at around 50μm the loss reaches 3dB. This means potentially the optical accelerometer has a large dynamic range, thus is able to measure very large acceleration values. The code of this calculation is given in appendix A.  32  Figure 2.18 Coupling loss versus distance between proof mass and waveguide output.  33  2.5  Device Performance The optical interference signals are detected by a PIN photodetector followed by  an RF spectrum analyzer. Suppose the optical powers from the two waveguide branches are P1 and P2 respectively, the detected electrical signal is proportional to P=  ⎡ (l − l ) ⎤ 1 1 P1 + P2 + P1 P2 cos ⎢ 2π 2 1 ⎥ 2 2 λ ⎦ ⎣  where l1 and l2 are optical path from two branches. Suppose the original distance between the waveguide output and the oscillator is 20μm, and so is the distance between the other output and the fixed reflector. If the coupling loss is included, the normalized power as a function of oscillator position is given by Figure 2.19.  Figure 2.19 Detector power versus oscillator position.  34  Consider the case in which the proof mass oscillates during the measurement. Assuming the oscillating frequency is ωm, the above equation must be written as ⎛ d ⎞ P = P1 + P2 + 2 P1 P2 cos ⎜ 4π sin ωmt ⎟ ⎝ λ ⎠ ⎛ 4π dωm t ⎞ cos ωmτ dτ ⎟ = P1 + P2 + 2 P1 P2 cos ⎜ ∫ τ = 0 ⎝ λ ⎠ where d is the largest displacement of the mass. This can be treated as phase modulation (PM) or frequency modulation (FM) where the carrier’s base frequency is zero. The modulation index is defined to be  m=  4π dωm  λ  ⋅  1  ωm  =  4π d  λ  = 8.2 ×106 d  It needs to be noted that, due to the coupling loss dependence of proof mass position, there should also be an amplitude modulation (AM) component as well as a combination of AM and FM. A typical amplitude modulation can be denoted by PAM = ( P0 + ΔP sin ωmt ) cos ωt  and the modulation index is defined by m=ΔP/P0, which is the ratio between the largest power variation and the carrier’s power. The AM spectrum shows a central peak at ω and two sidebands at ω+ωm and ω-ωm respectively, with a ratio of m/2 to the central peak in amplitude According to Figure 2.18, the largest amplitude modulation index could be ~104×d, which is much smaller than the frequency modulation index. It is to be found out later that, a FM spectrum can be seen as an AM spectrum with the same modulation index, in the limit of small index value. In this case, the amplitude modulation factor is ignored.  35  In frequency domain, the harmonic distribution of a sinusoid wave modulated by another sinusoid wave can be represented by Bessel functions of the first kind, which is defined to be 2 n +α −1) ( ⎛ x⎞ Jα ( x ) = ∑ ⎜ ⎟ n = 0 n !Γ ( n + α + 1) ⎝ 2 ⎠ ∞  n  The first three functions for integer orders α are shown in Figure 2.20.  Figure 2.20 Plot of the first three Bessel functions of first kind for integer orders α. For a modulation index m, the amplitude of the DC component is proportional to J0(m), the amplitude of the first harmonic component is proportional to J1(m), and the amplitude of the second harmonic component is proportional to J2(m), etc. Therefore, if the displacement of the proof mass is much smaller than the wavelength, on the RF  36  spectrum analyzer there should be a dominating DC signal as well as a small peak at the oscillating frequency 2fm. On the other hand, if the reference beam is modulated (e.g. by a Bragg cell) at a frequency fB, the spectrum will show a high peak at fB and two small B  B  peaks at fB+fm and fB-fm, respectively. Moreover, for very small modulation index, we B  B  have 2 n +1 −1) ( m ⎛m⎞ ≈ J1 ( m ) = ∑ ⎜ ⎟ 2 n = 0 n !Γ ( n + 1 + 1) ⎝ 2 ⎠ J i ( m ) ≈ 0,i>1 ∞  n  so the amplitude of the first harmonic signal is proportional to the index, and therefore proportional to the mass displacement. This feature sets up a simple relationship between the displacement (also acceleration) and the peak value of the first harmonic signal, which makes this optical accelerometer an ideal sensor of small accelerations. The result also shows that an FM can be seen as an AM in the case of sufficiently small modulation index value. As the modulation index grows large, i.e., the displacement is comparable to the wavelength, higher harmonic signals start to show up on the spectrum analyzer, which makes the readout to be complicated. On the other hand, the large displacement can be easily detected by capacitance variations.  37  2.6  Noise and Sensitivity The noise mainly comes from the laser and the photodetector, which limits the  sensitivity of the device. The three most significant noises are thermal noise, shot noise and laser noise. Thermal noise describes the fluctuations in the voltage across a dissipative circuit element, in this case the effective resistor of the photodetector, which are mostly caused by the thermal motion of the charge carriers. The mean-square noise current amplitude can be written as [15] 2 = iNA  4kT Δν RL  where Δν is the bandwidth of the detector circuit, T is the temperature, and RL is the output impedance of the detector. Shot noise comes from random generation and flow of mobile charge carriers, and the mean-square noise current amplitude is given by [15] 2 = iNS  2η e 2 P0 Δν hν  where P0 is the optical power, e is the charge of a electron, and η is the quantum efficiency. Laser noise describes the fluctuations in the output power of the laser, which may be due to temperature variations, acoustic variations, spontaneous emission of radiation into the laser mode, etc. The mean-square noise current amplitude is denoted by [15]  η 2e2 i = RIN ) P02 Δν 2 ( ( hν ) 2 NL  38  where relative intensity noise (RIN) is defined as the relative fluctuation power in a 1 Hz bandwidth. The typical RIN value is 10-16Hz-1. Thus, the total noise power can be written as 2 2 iN2 = iNL + iNS + iN2 A  Consider a photodiode to be the photodetector, thus the mean-square signal current amplitude is given by [15]  ⎛ η eP0 ⎞ is2 = m ⎜ ⎟ ⎝ hν ⎠  2  where m is the modulation index. Note that the equation is valid for amplitude modulation. However, since a very light frequency modulation can be seen as an amplitude modulation with the same modulation index, the same equation is used for our case. The signal and noise currents are calculated as functions of input optical power P0 for a typical photodiode detector, as is shown in Figure 2.21. The parameters are set to be Temperature: T=290K Detector output impedance: RL=1000Ω Quantum efficiency: η=0.5 Modulation index: m=2×10-4 Bandwidth of detector circuit: Δν=20kHz  39  Figure 2.21 Detector current versus input power for a typical photodiode. The solid line represents signal, and the dash-dot line represents noise.  It can be seen that when the input power is higher than -41.9dB, the signal current surpasses the noise current, thus can be detected by the diode. Therefore, the minimum detectable power is around 75nW. Now let us determine the possible sensitivity of the optical accelerometer. Assuming that the input laser power is 10mW, the propagation and coupling losses are 10dB in total, and the detector bandwidth is 20 kHz, at a minimum modulation index of 1.5×10-4, the output optical power at the device oscillating frequency is -41.2dBm, which can be detected by the photodiode, as is shown in Figure2.22. The displacement of proof mass for this modulation index is calculated to be 0.18Å. If the oscillating frequency of 40  the mass is 1kHz, the acceleration is calculated to be 58.4μg, and the sensitivity is 1848ng/Hz1/2. With a higher laser power, a narrower detection bandwidth, or a lower operating frequency, the sensitivity can be further reduced. For example, if the laser power is doubled, i.e. 20mW, the detection bandwidth is 2kHz, and the oscillating frequency is 1kHz, then the sensitivity is calculated to be 685ng/Hz1/2, which is below the minimum values for capacitive and piezoelectric accelerometers. The sensitivity can also be improved by applying shorter wavelength.  Figure 2.22 Detector current versus input power at m=1.5×10-4 for a 10mW laser. The solid line represents signal, and the dash-dot line represents noise.  41  On the other hand, given a requirement to measure a minimum detectable displacement, the minimum optical power required can also be calculated. For example, to achieve a minimum detectable displacement of 1Å at 2kHz detection bandwidth, the minimum optical power required is -6.0dBm.  42  Chapter III. 3.1  Device Fabrication  General Process Flow To fabricate a device described in Chapter II, a metallization and two dry etch  steps are required. The metallization process forms contact pads for the capacitive accelerometer. The first dry etch process, which is 13um in depth, creates the capacitive accelerometer and waveguide area; the second one, which is 6.5um in depth, defines the rib of the waveguide by etching away a part of the wings. The metallization is preferred to be made before the patterns are etched out for high lift-off quality. Therefore, the general fabrication process is as follows, 1. Photoresist coating: leave the wafer under vapor HMDS environment for 30s for good adhesion, then spin AZP4110 photoresist on wafer for 40s at 2000rpm, which gives around 1.6um thickness; 2. Soft bake: put the wafer on a hot plate for 10min at 90ºC; 3. Photolithography #1 for metallization: 13s photolithography using the Canon mask aligner in UBC cleanroom; 4. Developing: develop the wafer with 1:3 AZ400K for 40s, then immerse it into DI water for 30s, then blow dry; 5. Metal evaporation: evaporate metals on the wafer using the Dee Wong thermal evaporator in UBC cleanroom; 6. Lift-off: immerse the wafer in a beaker of Acetone, and then put the beaker in the ultrasonic bath for 10min to get contact pads, then blow dry; 7. Photolithography #2 for dry etch #1: same recipe as process 1, 2, 3, and 4;  43  8. Dry etch #1: etch all the way down to the BOX layer (13um) to create the capacitive accelerometer and waveguide area on the wafer using either PowerQuest Electron Cyclotron Resonance (ECR) system in UBC cleanroom or Bosch etch in the University of Sherbrooke; 9. Photolithography #3 for dry etch #2: same recipe as process 1, 2, 3, and 4; 10. Dry etch #2: 6.5um etch to create a 9um rib on waveguide area, same recipe as process 8; 11. Sacrificial etch: immerse the wafer into 1:10 Buffered Oxide Etch (BOE) solution for 4 hours to obtain suspended mass and beam, then immerse it into a tray of DI water for 5min and repeat 3 times, then leave it in a beaker of low surface tension solvent (e.g. Isopropanol); 12. Critical Point Drying (CPD): dry the wafer using a Tousimis Autosamdri-815 1’’ Wafer and MEMS Dryer in order to prevent the released structure from collapse.  44  3.2  Ohmic Contact For measurement and characterization purpose, the ideal capacitive accelerometer  should have a bias independent resistance. Since the device layer is semiconductor, the contact metal has to be carefully selected to avoid large resistance in between. Figure 3.1 illustrates the energy band diagram as metal contacts with a p-type semiconductor. A Schottky barrier, which has a height of φB , is formed between the two materials, which blocks the electron flow from semiconductor to metal at low bias voltage. The barrier height φB is determined by the work function of the metal φm , the electron’s affinity and bandgap of the semiconductor χ s and Eg , which is given by  φ B = E g − ( φm − χ s )  Figure 3.1 The energy band diagram of metal, p-type semiconductor, and the contact between the two in equilibrium. 45  As we want the contact to be Ohmic, the Schottky barrier should be overcome. A typical I-V curve for Ohmic and Shottky contacts is shown in Figure 3.2. There are two ways to make a contact Ohmic enough to get signals into and out of the semiconductor: 1. lower the barrier height; 2. narrow down the barrier width. For a p-type semiconductor, the first way can be achieved by choosing metal with a large work function. The second way works because a very narrow barrier makes it easy enough for electrons to tunnel through, and this can be achieved by doping the semiconductor heavily. Since we use a lightly doped p-type Si wafer in this project, the barrier height has to be minimized.  Figure 3.2 I-V curve of typical Ohmic and Schottky contacts.  The barrier heights between p-Si and several commonly used metals are listed in Table 3.1. It can be seen that Platinum may result in the lowest barrier height. However,  46  Both Platinum and Palladium are not available in UBC cleanroom. Chromium has a low barrier height but is easy to be oxidized, thus can be used as an adhesion layer. Therefore, a 30nm/200nm Cr/Cu layer is deposited on the device to form contact pads.  Table 3.1 Commonly used contact metals and their barrier heights when contacting with a p-type Si [16]. Metal  Barrier Height (eV)  Aluminum (Al)  0.92  Titanium (Ti)  1.22  Chromium (Cr)  0.59  Copper (Cu)  0.77  Palladium (Pd)  0.37  Tungsten (W)  0.67  Platinum (Pt)  -0.15  Gold (Au)  0.87  The contact characterization stage consists of a Keithley 2602 SYSTEM SourceMeter and two probes. Electric current generated by the SourceMeter flows through two separate contact pads and the semiconductor in between. The contact separation is 380μm. The voltage drop is then measured by the same equipment. A GPIB controller is used to automatically scan and collect the data, which is saved in the computer. For the above Cr/Cu contact, the I-V curve is shown in Figure 3.3.  47  Figure 3.3 Voltage drop versus bias current for a 30nm/200nm Cr/Cu contact.  The contact looks not perfectly ohmic, but definitely not a Schottky one because there is not a threshold voltage. The average resistivity is around 20kΩ, which is relatively high but reasonable for such a lightly doped Si wafer.  48  3.3  Dry Etch Highly anisotropic etch is required for this project, as the ideal device, especially  the optical waveguide, should have vertical sidewalls. This process is done in both UBC cleanroom and University of Sherbrooke. An electron cyclotron resonance (ECR) etching system with chlorine (Cl2) source is used in UBC, while a standard deep reactive ion etch (DRIE) process with sulfur hexafluoride (SF6) source is done in University of Sherbrooke. The two commonly used dry etch methods are introduced and compared in this section.  3.3.1 Electron Cyclotron Resonance (ECR) Plasma Etch In an ECR plasma etch system, high-density ionized gas plasmas are generated by superimposing a static magnetic field and a RF electromagnetic field. The free electrons in the chamber are heated by microwaves and in turn collide with the molecules of the gas to cause ionization, like an avalanche. The ions are vertically delivered to the sample being etched, and react both chemically and physically with the surface material, which creates highly anisotropic etching profile. In UBC cleanroom, chlorine gas is used to etch Si samples. The chlorine plasma takes away Si material mainly by chemical reaction which generates SiCl4 gas, but also by kinetic energy transfer which knocks off the Si atoms. There are quite a few parameters that may take effect on the etching conditions, such as gas flow rate, chamber pressure and temperature, microwave and RF power, etc. Generally, the etch rate is increased with higher gas flow rate, microwave power or RF power, as the plasma density or ions energy is increased. On the other hand, very high  49  plasma density or energy may lead to the degradation of etch profile, which is a trade-off factor. Although chlorine does not chemically react with photoresist, it may still physically etch away the material fast. Moreover, the etch rate for Si is relatively low, which results in a low Si:photoresist selectivity. Therefore, the photoresist layer is not thick enough for a 13um deep Si etching. In this case, metal has to be used for etch mask, and nickel is an ideal candidate. The Ni etch mask is also prepared using lift-off technique, and titanium is used as adhesion layer. A 20nm/200nm Ti/Ni layer is deposited using thermal evaporator to form etch mask. Figure 3.4 shows a 70nm aluminum deposition results using lift-off technique. The patterns look clear on the figure.  Figure 3.4 A 70nm Al layer is deposited on Si using lift-off technique. The minimum feature size is 3μm.  50  There is another important phenomenon called “microloading effect”, which means the etch rate can depend on feature size. This is because the etch species and etch products cannot move into and out of very narrow trenches as quickly as the wide area case. This effect can be minimized by reducing either the chamber pressure or the etch rate. Also, cares should be taken in the layout design. The etch recipe is composed and modified based on the research work done by Prof. S. W. Pang’s group in University of Michigan [17]-[20]. They managed to achieve ~100μm deep high-aspect-ratio etch of Si by ECR. The recipe used in the etch test is: Cl2 flow rate: 20sccm He flow rate: 5sccm Chamber pressure: 3mtorr Microwave power: 200W RF power: 50W Temperature: 0.9ºC The Helium gas is used for cooling down the sample from backside. A 50 minutes etch has been done using the above recipe, and a 4.5um deep trench is obtained for a 9um opening area, resulting in a 90nm/min etch rate. The top view and the etch profiles are shown in Figure 3.5, Figure 3.6, and Figure 3.7.  51  Figure 3.5 Top view of proof mass and comb-drive fingers after ECR etch.  Figure 3.6 ECR Etch profile scanned by Atomic Force Microscope (AFM). 52  Figure 3.7 ECR Etch profile observed by Scanning Electron Microscope (SEM) in UBC cleanroom. The etch depth is 4.5um in 50 minutes, resulting in 90nm/min etch rate.  It can be seen from SEM image that the sidewall is close to be vertical. However, the etching could not be continued because the Ni mask was gone. The etch rate for Ni is 4nm/min, so the selectivity is determined to be 22.5. To complete a 13um etch, a Ni mask thicker than 578nm is required. However, the high intrinsic stress limits the evaporated Ni to be no more than 200nm. In this case, electroplating should be used to prepare thick Ni mask. The Ni is electroplated following the work in [18]. A 25nm/5nm Ti/Ni layer is first evaporated on Si as a seed layer. Photoresist is deposited on the seed layer and patterned by the mask aligner. After electroplating, the photoresist is washed away by acetone, and the seed layer can be eliminated by 1:1 HCl:H2O followed by a 5% HF dip. 53  The ideal Ni electroplating solution is so called “Watts bath”, which consists of 310g/l Nickel Sulphate crystal (NiSO4·6H2O), 50g/l Nickel Chloride crystal (NiCl2·6H2O), and 40g/l Boric Acid (H3BO3), and the suggested current density value range is 30-100A/ft2 [21]-[22]. Boric acid is used as a buffer, to ensure no hydroxides or basic salt precipitates are formed near the electrodes due to pH rise near the cathode. Only NiCl2 is available and therefore used to prepare electroplating solution and make deposition tests. The concentration used is as high as 400g/l for reasonable deposition rate. The result shows that the deposited layer is very rough with a large amount of NiCl2 crystal covered on top. It is also known that deposits made with high concentration of NiCl2 show high tensile stress and are brittle, thus prone to fracture or peel off. Therefore, it seems “Watts bath” should be prepared for future Ni electroplating tests.  54  3.3.2 Deep Reactive Ion Etch (DRIE) DRIE is a modified version of Reactive Ion Etch (RIE) for creating deep anisotropic features. In a typical RIE system, plasma is initiated by applying a strong RF electromagnetic field to the wafer platter. The oscillating electric field ionizes the gas molecules to create the plasma. Then the free electrons are driven by the field up and down, and are absorbed into the wafer platter which causes the platter to build up charge due to its DC isolation. Thus, the positive ions tend to drift toward the wafer platter, where they collide with the sample material to be etched. Same as ECR system, both chemical and physical reaction take place on the uncovered surface of the sample and etch away the material. Because the ions are mostly driven vertically to the sample, the etch profile can be very anisotropic. However, when a trench is created, the unprotected sidewalls may also be attacked by reactive ions due to the presence of scattering, which results in undercut profiles. As the etch goes deep, the undercut effect becomes more and more serious. Therefore, DRIE is developed to solve this problem. There are two main DRIE technologies: cryogenic process and Bosch process. In cryogenic process, the sample is chilled to a very low temperature. The low temperature slows down the chemical reaction, which is the main reaction taken place on the sidewalls. On the other hand, the ions continue to bombard vertically facing the sample surface and etch the material away. This process can produce trenches with vertical sidewalls. However, the etch rate might be greatly limited without sufficient chemical reaction. The Bosch process consists of a repeated switch between two modes of short periods, as is shown in Figure 3.8. In the passivation mode, octafluorocyclobutane (C4F8)  55  is dissociated in the plasma to deposit a Teflon-like fluorocarbon polymer film on the surface and sidewalls of the Si sample, which is a chemically inert layer [23]. In the etching mode, sulphur hexafluoride (SF6) is dissociated in the plasma to form fluorine species, and the Si is isotropically etched by fluorine after the polymer film is removed by ionic bombardment. The ionic bombardment is highly directional and thus preferentially removes the horizontal part of the polymer film, keeping the sidewalls protected against chemical reaction. These etching and passivation steps are repeated many times, which results in a large number of small isotropic etch taking place only at the bottom of the etched trench. Comparing with cryogenic process, the etching done by Bosch process has not only highly anisotropic profile, but also very high etch rate [24][26]. Therefore, Bosch process is the most common DRIE technique currently.  Figure 3.8 Bosch process illustration. (a) Passivation mode: Teflon-like fluorocarbon polymer film is deposited on Si sample. (b) Etching mode: the horizontal part of the polymer film is removed by ionic bombardment, and Si is isotropically etched by reactions with fluorine species. (c) Next passivation mode: the resulting sidewalls consist of a series of very small scallops shape.  56  The Bosch process also has very high selectivity. The DRIE equipment in University of Sherbrooke has a selectivity of 75 for photoresist masks. Therefore, a 1.6um photoresist layer is thick enough for 13um deep Si etching. The selectivity for SiO2 is even close to 150, which makes the etching stop at the BOX layer when dealing with SOI wafers. This is a helpful feature for fabricating movable structures. However, the BOX layer charges when exposed to the ionic bombardment. This makes the ions deflect and start to etch the sidewalls close to the bottom, which eventually creates notches there. This phenomenon is called notching effect. When a small notch is created, the passivation layer can not be deposited inside the notch, which makes it grow bigger and bigger. When there are both small and big features to be released on the same sample, both microloading and notching phenomena take effect, which may seriously degrade the etch profile, as is shown in Figure 3.9. Besides, the sidewall profile is highly dependent on the plasma parameters [27]-[29]. Therefore, both layout design and etch recipe have to be carefully made to minimize these two effects [30]-[31].  Figure 3.9 Microloading effect and notching effect. The notching effect has happened for large features, while etching has not reached the SiO2 layer for small features.  57  A better optimized Si etch recipe used in University of Sherbrooke is as follows,  where O2 flow is added during the etching mode to remove organic species resulting of the polymer sputtering. The etch rate is around 2.08μm/min. A 13um Si etch is done by Bosch process in University of Sherbrooke, and the sidewall profile is shown in Figure 3.10. The surface looks clean, and the etching has reached the BOX layer, which looks fainter in the figure. Also, the sidewall is truly vertical, and the scallops are too small to be seen.  58  Figure 3.10 A 13um deep vertical Si etch done by Bosch process in University of Sherbrooke. The profile is observed by SEM in UBC cleanroom.  The comparisons between ECR etch in UBC cleanroom and DRIE in University of Sherbrooke are shown in Table 3.2. The Bosch process has higher etch rate, better anisotropy, and higher selectivity. On the other hand, higher etch rate leads to more obvious microloading effect, as there is no sufficient time for etch species and products to exchange in narrow trenches; higher selectivity results in more obvious notching effect. With a careful design of layout and etch recipe, Bosch process is currently the best way for deep Si etch.  59  Table 3.2 Comparisons between ECR etch in UBC cleanroom and DRIE in University of Sherbrooke. ECR in UBC cleanroom  DRIE in Sherbrooke  Gas  Cl2  SF6  Etch rate  ~90nm/min  ~2μm/min  Anisotropy  high  truly vertical  Selectivity  22.5 for Ni mask  75 for photoresist mask  Microloading effect  low  high  Notching effect  low  high  60  3.4  Sacrificial Etch The isotropic sacrificial etch is done by 1:10 BOE solution, in order to remove the  BOX layer underneath the structures to be released. A 55 minutes etch test is done to determine the etch rate of SiO2 and Si:SiO2 selectivity, and the etch result is shown in Figure 3.11.  Figure 3.11 Around 2μm SiO2 is etched away by BOE in 55 minutes with almost no damage to Si. The etch profile is observed by SEM in UBC cleanroom. After a 55 minutes BOE etch, around 2μm SiO2 underneath Si is etched away, resulting in a ~364Å/min etch rate. On the other hand, the etch rate of Si is almost zero, which means the selectivity is almost infinite. Therefore, to make sure the structures to be released are completely free of SiO2, a slightly longer release time is preferred. A 5 hours  61  BOE etch was performed in order to fully release the oscillator. Figure 3.12 and Figure 3.13 show the SEM images of a device after the first dry etch step and BOE release.  Figure 3.12 SEM image of a device after the first dry etch and BOE release. The picture is taken in Simon Fraser University (SFU). The etching is done by DRIE in Sherbrooke.  62  Figure 3.13 Another SEM image of the device after the first dry etch and BOE release. The dry etch is done by Bosch process in University of Sherbrooke. The picture is taken in SFU.  63  Chapter IV. 4.1  Characterization and Future Work  Characterization The device consists of two main components: a capacitive accelerometer and a Y-  shape rib waveguide. These two components should be characterized separately before characterizing the whole device. For optical waveguide, the most important feature to be characterized is single-mode propagation. For the oscillator, the resonance frequency needs to be characterized to compare with the simulation results. At last, calibration should be made with electrical actuation before optical measurements are carried out.  4.1.1 Waveguide Characterization The schematic view of waveguide characterization stage is shown in Figure 4.1. A 1.55μm laser beam is coupled into the waveguide through a lensed single-mode optical fiber. The tip position is finely controlled by a three-axis micropositioner. A lens is placed close to the waveguide output, which is controlled by another three-axis micropositioner and projects the light spot to the infrared card. If there is only fundamental mode in the waveguide, a single red spot should appear on the IR card. The actual stage is shown in Figure 4.2. IR card  Micropositioner  Lens  Y-shape rib waveguide  Micropositioner  Lensed fiber  1550 nm DFB laser  Figure 4.1 Schematic view of optical waveguide characterization stage.  64  Figure 4.2 Optical waveguide characterization stage.  The laser beam needs to be carefully coupled into the rib region. If it is coupled into the wing region which can be treated as a slab waveguide, there should be a series of red spots appearing on the IR card, because slab waveguide with the designed dimensions is not able to propagate single mode inside.  4.1.2 Oscillator Characterization with LDV Although the movable structure oscillates mainly in the plane, there is also vibration in the vertical direction, and the z-axis resonance frequency is simulated to be 29.017 kHz, as is shown previously in Figure 2.7. This vibration can be detected using an LDV, which is introduced in Section 1.4. 65  The oscillator characterization setup is shown in Figure 4.3. DC bias and AC signal are applied on the oscillator through probes and contact pads, and both continuous and pulsed signals can be used to actuate the device. The z-axis vibrations are measured by the LDV on top, which takes advantages of optical inteferometry and Doppler Effect. Both the actuation signal and the vibration signal are monitored by an oscilloscope. The actual characterization setup is shown in Figure 4.4. DC voltage source  Combdrive actuation probe VDC + vac  + +  Bias T  Proofmass probe  Probe Station  Proofmass (grounded)  – +  ac voltage function generator  LDV measurement axis* (Z)  Combdrive actuation probe -VDC + vac  Oscilloscope CH 1  +  CH 2  +  Figure 4.3 Schematic view of oscillator characterization setup. The z-axis vibration is measured by LDV.  66  Figure 4.4 Oscillator characterization stage.  4.1.3 Device Characterization The characterization setup for the whole device is shown in Figure 4.5. A 1.55μm laser beam goes through an optical circulator and couples into the waveguide. The reflected interference patterns then pass through the circulator and are detected by a PIN photodetector. The spectrum is shown on a RF analyzer. When the device is working, there should be a peak at the oscillating frequency, which is described in Section 2.5. The acceleration can also be electrically measured by the bridge circuit, and calibration can be made based on results from both measurements.  67  Figure 4.5 Schematic view of device characterization setup.  68  4.2  Conclusion and Future Work In this thesis, A MEMS-based optical accelerometer has been designed. Moreover,  analysis and simulations have been finalized which indicate that the acceleration sensitivity can be below 1μg/Hz1/2, which is lower than the minimum values for capacitive and piezoelectric accelerometers. Moreover, the measurable range is determined to be 50μm. However, a complete device has not been made and the characterization stages have not been tested. Therefore, the future work includes: 1. Fabrication of complete optical accelerometers. 2. Set up a device characterization stage. 3. Waveguide, oscillator and device characterizations. 4. Design optimization. A number of improvements in design can be made to reduce fabrication difficulty and obtain better device performance. These may include: 1. Use a highly doped SOI wafer for better electrical conductivity, and the device layer is preferred to be 9μm thick. 2. Reduce the resonance frequency of the oscillator so that the minimum detectable acceleration of the device is further reduced given the same minimum detectable displacement. 3. Try to prevent large feature size difference on the layout in order to minimize microloading and notching effects. 4. Deposit SiO2 instead of Ni as etch mask. The SiO2 mask can be microns thick and the removal is easy and thorough. However, DRIE is still the best choice.  69  5. Create a V-groove on the sample, in front of the waveguide input, for efficient and stable fiber/waveguide coupling.  70  References [1] H. Xie and G. Fedder, “Integrated micromechanical gyroscopes”, Journal of  Aerospace Engineering, 16 (2), pp. 65-75 (April 2003). [2] D. A. Horsley, R. Horowitz, A. P. Pisano. “Microfabricated electrostatic actuators for hard disk drives”, IEEE/ASME Transactions on Mechatronics, 3, (3), pp. 175-183 (September 1998). [3] A. Bertolini, R. DeSalvo, F. Fidecaro and A. Takamori. “Monolithic folded pendulum accelerometers for seismic monitoring and active isolation systems”, IEEE Nuclear  Science Symposium Conference 2004, pp. 4644-4648 (October 2004). [4] S. D. Senturia, “Microsystem design”, Kluwer Academic Publishers, Boston, USA (2001). [5] Polytec LM INFO Special, issue 1/2003, Polytec GmbH. D-Waldbronn. [6] B. V. Zeghbroeck, “Principle of semiconductor devices”, http://ecewww.colorado.edu/~bart/book/ [7] R. A. Soref, J. Shmidtchen and K. Petermann, “”Large single-mode rib waveguide in GeSi-Si and Si-on-SiO2”, IEEE Journal of Quantum Electronics, Vol 27, No. 8, pp. 1971-1974 (August 1991). [8] A. G. Rickman, G. T. Reed and F. Namavar, “Silicon-on-insulator optical rib waveguide loss and mode characteristics”, Journal of Lightwave Technology, Vol. 12, No. 10, pp. 1771-1776 (October 1994). [9] S. P. Pogossian, L. Vescan and A. Vonsovici, “The single-mode condition for semiconductor rib waveguides with large cross section”, Journal of Lightwave  Technology, Vol. 16, No. 10, pp. 1851-1853 (October 1998). 71  [10] L. Lousteau, A. B. Seddon, T. M. Benson, A. Vukovic and P. Sewell, “The singlemode condition for silicon-on-insulator rib waveguides with large cross section”, Journal  of Lightwave Technology, Vol. 22, No. 8, pp. 1923-1929 (August 2004). [11] Y. Wang, X. Cheng, Z. Lin, F. Gao, F. Zhang, “Monolithic beam splitter in siliconon insulator”, Optical Society of America, Vol. 12, No. 21, pp. 5154-5159 (18 Oct. 2004). [12] K. Solehmainen, T. Aalto, J. Dekker, M. Kapulainen, M. Harjanne, K. Kukli, P. Heimala, K. Kolari and M. Leskela, “Dry-etched silicon-on-insulator waveguides with low propagation and fiber-coupling losses”, Journal of Lightwave Technology, Vol. 23, No. 111, pp. 3875-3880 (November 2005). [13] S. Adachi, “Model dielectric constants of Si and Ge,” Physics Review B, Vol. 38, No. 18, pp. 12966-12976 (15 Dec. 1988). [14] I. H. Malitson, “Interspecimen comparison of the refractive index of fused silica,” J.  Optical Society of America Vol. 55, No. 10, pp. 1205-1209 (October 1965). [15] A. Yariv, “Optical electronics in modern communications”, Oxford University Press, New York, USA (March 1997). [16] http://www.ee.byu.edu/cleanroom/ohmic-schottky.phtml [17] W. H. Juan and S. W. Pang, “High-aspect-ratio Si etch for microsensor fabrication”,  Journal of Vacuum Society A, Vo. 13, No. 3, pp. 834-838 (19 Dec. 1994). [18] W. H. Juan and S. W. Pang, “Control of etch profile for fabrication of Si microsensors”, Journal of Vacuum Society A, Vol. 14, No. 3, pp. 1189-1193 (11 Mar. 1996).  72  [19] J. W. Weigold, W. H. Juan and S. W. Pang, “Dry etch of deep Si trenches for released resonators in a Cl2 plasma”, Journal of Electrochemical Society, Vol. 145, No. 5, pp. 1767-1771 (May 1998). [20] W. H. Juan and S. W. Pang, “High-aspect-ratio Si vertical micromirror arrays for optical switching”, Journal of Microelectromechanical Systems, Vol. 7, No. 2, pp. 207213 (June 1998). [21] J. K. Dennis, T. E. Such, “Nickel and Chromium Plating”, Newnes Butterworths, London, England (1972). [22] R. Brugger, “Nickel plating”, Robert Drapper, Teddington, England (1970). [23] G. Gormley, K. Yallup, W. A. Nevin, J. Bhardwaj, H. Ashraft, P. Huggett, S. Blackstone, “State of the art deep silicon anisotropic etching on SOI bonded substrates for dielectric isolation and MEMS applications”, Fifth International Symposium on  Semiconductor Waver Bonding: Science, Technology and Applications, Hawaii, USA, pp.17-22 (1999). [24] K. Yu, M. Feldbaum, T. Pandhumsoporn and P. Gadgil, “Deep anisotropic ICP plasma etching designed for high volume MEMS manufacturing”, Alcatel Comptech Inc., Fremont, USA. [25] P. T. Docker, P. Kinnell and M. C. L. Ward, “A dry sigle-step process ofr the manufacture of released MEMS structures”, Journal of Micromechanics and  Microengineering, Vol. 13, pp. 790-794 (9 Jul. 2003). [26] M. Wasilik, N. Chen, “Deep reactive ion etch conditioning recipe”, Berkeley Sensor & Actuator Center, Berkeley, USA.  73  [27] K. Chen, A. A. Ayon, X. Zhang and S. M. Spearing, “Effect of process parameters on the surface morphology and mechanical performance of silicon structures after deep reactive ion etching (DRIE)”, Journal of Micromechanics and Microengineering, Vol. 11, pp. 264-275 (2002). [28] A. A. Ayon, R. Braff, C. C. Lin, H. H. Sawin and M. A. Schmidt, “Characterization of a time multiplexed inductively coupled plasma etcher”, Journal of Electrochemical  Society, Vol. 146, No. 1, pp. 339-349 (1999). [29] J. Min, G. Lee, J. Lee and S. H. Moona, “Dependences of bottom and sidewall etch rates on bias voltage and source power during the etching of poly-Si and fluorocarbon polymer using SF6, C4F8, and O2 plasmas”, Journal of Vacuum Science and Technology  B, Vol. 22, No. 3, pp. 893-901 (2004). [30] J. Kiihamaki, “Fabrication of SOI micromechanical devices”, VTT Publications 559, ESPOO 2005. [31] R. J. Shul, S. J. Pearton, “Handbook of advanced plasma processing techniques”,  Springer, New York, USA (2000). [32] http://www.ee.ucla.edu/~wu/ee250b/Case%20studyCapacitive%20Accelerometer.pdf. [33] http://www.silicondesigns.com/. [34] http://www.mmf.de/piezoelectric_principle.htm. [35] Nin C. Loh, Martin A. Schmidt, and Scott R. Manalis, “Sub-10cm3 interferometric accelerometer with nano-g resolution”, Journal of Microelectromechanical Systems, Vol. 11, No. 3, pp. 182-187 (2002).  74  [36] Jose A. Plaza, Andreu Llobera, Carlos Domingues, Jaume Esteve, Inigo Salinas, Jorge Garcia and J. Berganzo, “BESOI-based integrated optical silicon accelerometer”,  Journal of Microelectromechanical Systems, Vol. 13, No. 2, pp. 355-364 (2004). [37] Jean-Francois Cliche and Michel Tetu, “Effect of laser linewidth reduction systems on coherence length and interferometer noise”, 2005 Digest of the LEOS Summer Topical Meetings, pp. 121-122.  75  Appendix A: Code of coupling loss calculation clear all w=9e-6;h=13e-6; w0x=w/2;w0y=h/2;lamda=1.55e-6; z0x=pi*w0x^2/lamda; z0y=pi*w0y^2/lamda; distance=1e-4;pts=1001; Ptotal=0.5*pi*w0x*w0y; z=linspace(0,distance,pts);P=zeros(1,pts); z2=2*z; dwx=w0x/100;dwy=w0y/100; for y=-w0y:dwy:w0y for x=-w0x:dwx:w0x wx=w0x*sqrt(1+z2.^2/z0x^2); wy=w0y*sqrt(1+z2.^2/z0y^2); I=(w0x./wx).*(w0y./wy).*exp(-2*(x^2./wx.^2+y^2./wy.^2)); P=P+I*dwx*dwy; end end loss=10*log10(P(1)./P); plot(z*1e6,loss) grid on  76  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0066397/manifest

Comment

Related Items