University of Southern California Dissertations and Theses

Piezoelectric ultrasonic and acoustic microelectromechanical systems (MEMS) for biomedical, manipulation, and actuation applications

(USC Thesis Other)

Piezoelectric ultrasonic and acoustic microelectromechanical systems (MEMS) for biomedical, manipulation, and actuation applications

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content PIEZOELECTRIC ULTRASONIC AND ACOUSTIC MICROELECTROMECHANICAL SYSTEMS (MEMS) FOR BIOMEDICAL, MANIPULATION, AND ACTUATION APPLICATIONS by Yongkui Tang A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (ELECTRICAL ENGINEERING) August 2021 Copyright 2021 Yongkui Tang ii Acknowledgements First and foremost, I would like to express my sincere gratitude towards my advisor, Prof. Eun Sok Kim. In seven years, he has offered tremendous help, thoughtful guidance, insightful inspirations, and warm encouragement, which helped me grow from a layman to an experienced researcher in the MEMS field. He is also an excellent role model with never-ending enthusiasm and meticulous attitude towards research, which inspired me to always strive for excellence and to explore a wide variety of research projects. As a mentor, he is caring, patient, and understanding both in research and in daily life, which helped me learn from my past mistakes to become a better academician and a better person. He always puts his students at the highest preference, even during times with very limited research funds. He has provided me with opportunities to attend multiple international conferences, which enabled me to learn from and communicate with talented researchers from all over the world and improve my presentation skills. All the works mentioned in the thesis would not have been possible without Dr. Kim’s guidance. I would also like to thank Prof. Wei Wu and Prof. Ellis Meng for serving as my thesis committee and for sharing useful advice on improving this thesis. At the same time, I truly appreciate all the help and support I’ve received from all the current and former members of the USC MEMS Group whom I have worked with. My special thanks go to Dr. Yufeng Wang, who trained, helped, and inspired me in so many ways during my first, and the toughest two years of Ph.D. research. I would also like to express my special thanks to Dr. Anton Shkel, Hai Liu, Jaehoon Lee, and Dr. Yunqi Cao for their friendship, discussion, and help in both research and life. I am also grateful for other students and friends I met at USC who offered enormous help and have fought together with me throughout the years, especially Runze Li and Dr. Xuejun Qian from Prof. Qifa Zhou’s lab; Dr. Boxiang Song, Dr. Yifei Wang, Dr. Yuhan Yao, iii Hao Yang, Deming Meng, and Pan Hu from Prof. Wei Wu’s lab; Dr. Xiaodong Yan from Prof. Han Wang’s lab; Huandong Chen from Prof. Jayakanth Ravichandran’s lab; Dr. Debarghya Sarkar and Jun Tao from Prof. Rehan Kapadia’s lab. In addition, I am extremely thankful for all the collaborators I have worked with, whose expertise and hard work have not only led to fruitful results, but also broadened my knowledge and skills in bioscience and biomedical experiments. These people include Prof. Mitchell E. Gross, Dr. Chun-Peng Liao, Leng-Ying (Joanna) Chen, Ailin Zhang, and Kiran Sriram from Keck School of Medicine at University of Southern California (USC); Prof. Su-Youne Chang from Mayo Clinic, as well as Prof. Jiang (John) F. Zhong and Xuelian (Lili) Chen from Herman Ostrow School of Dentistry at USC. Most importantly, I want to express my deepest gratitude to my girlfriend, Shi Qiu, whom I met in the most difficult times of my Ph.D. studies, and who have encouraged and helped me through numerous hardships in life and research, and have created and shared so many happy moments with me. I am truly in debt to her never-ending support and understanding. Also, I would like to deeply thank my parents for their endless love and support for me, even since I was born. iv Table of Contents Acknowledgements ......................................................................................................................... ii List of Figures .............................................................................................................................. viii List of Tables ................................................................................................................................ xx Abstract ........................................................................................................................................ xxi Chapter 1 Introduction .................................................................................................................... 1 1.1 Applications of Sound .......................................................................................................... 1 1.2 Overview of Acoustic Transducers ....................................................................................... 2 1.2.1 Electromagnetic Acoustic Transducers .......................................................................... 3 1.2.2 Electrostatic Acoustic Transducers ................................................................................ 4 1.2.3 Piezoelectric Acoustic Transducers ............................................................................... 5 1.3 Overview of Acoustic Lenses ............................................................................................... 6 1.4 Scope of the Thesis ............................................................................................................... 8 1.5 Overview of the Thesis Chapters .......................................................................................... 9 Chapter 2 Design, Fabrication, and Characterization of Self-Focusing Aoucstic Transducers (SFAT) .......................................................................................................................................... 11 2.1 Introduction ......................................................................................................................... 11 2.2 Working Principle of SFAT ................................................................................................ 12 2.2.1 SFAT Based on an Air-Cavity Lens ............................................................................ 12 2.2.2 SFAT Based on Patterned Electrode Rings ................................................................. 14 2.3 Simulation of Acoustic Pressure ......................................................................................... 15 2.4 Design Parameters of SFATs .............................................................................................. 17 2.4.1 Operating Frequency .................................................................................................... 17 2.4.2 Peak Acoustic Pressure ................................................................................................ 18 2.4.3 Focal Diameter (Focal Size) ........................................................................................ 18 2.4.4 Depth of Focus (Focal Depth) ...................................................................................... 20 2.4.5 Device Size .................................................................................................................. 21 2.5 Microfabrication Methods for SFATs ................................................................................ 21 2.5.1 Fabrication of SFAT with Patterned Electrode Rings ................................................. 21 2.5.2 Fabrication of SFAT with Parylene Air-Cavity Lens .................................................. 22 2.5.3 Fabrication of SFAT with PDMS Air-Cavity Lens ..................................................... 23 2.5.4 Fabrication of SFAT with SU-8 Air-Cavity Lens ........................................................ 25 2.5.5 Fabrication of SFAT with SU-8/PDMS Air-Cavity Lens ............................................ 27 2.5.6 Comparison of Different Fabrication Methods ............................................................ 28 2.6 Characterization of SFAT ................................................................................................... 29 2.6.1 Measurement of Electrical Impedance ......................................................................... 29 2.6.2 Measurement of Acoustic Pressure .............................................................................. 30 v Chapter 3 SFATs for Nozzleless Droplet Ejection and Droplet-Ejection-Assisted Object Delivery ....................................................................................................................................................... 32 3.1 Introduction ......................................................................................................................... 32 3.2 On-Demand, Nozzleless Ejection of Sub-Millimeter-Sized Liquid Droplets .................... 33 3.2.1 Device Design and Fabrication .................................................................................... 34 3.2.2 Ejection Characterization ............................................................................................. 35 3.3 Varying Droplet Size through Electrical Tuning of Focal Diameter .................................. 41 3.3.1 Device Design and Tuning Principle ........................................................................... 42 3.3.2 Measurement Results ................................................................................................... 45 3.3.3 Summary ...................................................................................................................... 49 3.4 Acoustic Droplet Ejector with Pulse-Width-Modulated Droplet Size for Picking and Placing Semiconductor Chips ................................................................................................................ 49 3.4.1 Background and Motivation ........................................................................................ 49 3.4.2 Device Design .............................................................................................................. 51 3.4.3 Simulation of Acoustic Pressure and Body Force ....................................................... 52 3.4.4 Simulation of Droplet Ejection Process ....................................................................... 54 3.4.5 Characterization of Droplet Diameter .......................................................................... 59 3.4.6 Ejection of Silicon Chips ............................................................................................. 61 3.4.7 Automatic Loading of Silicon Chips ........................................................................... 63 3.4.8 Assembly of Silicon Chips ........................................................................................... 65 3.4.9 Summary ...................................................................................................................... 65 3.5 Extraction and Delivery of Microparticles and Cells Based on Droplet Ejection .............. 67 3.5.1 Introduction .................................................................................................................. 67 3.5.2 Device Design .............................................................................................................. 68 3.5.3 Focusing through Agarose-Gel-Filled Petri Dish ........................................................ 70 3.5.4 Characterization of Acoustic Pressure and Droplet Ejection ....................................... 72 3.5.5 Droplet-Assisted Particle Ejection ............................................................................... 75 3.5.6 Droplet-Assisted Cell Extraction and Ejection ............................................................ 78 3.5.7 Summary ...................................................................................................................... 79 Chapter 4 SFAT for In Vivo Non-Thermal Selective Cancer Treatment with High-Frequency Medium-Intensity Focused Ultrasound......................................................................................... 81 4.1 Introduction ......................................................................................................................... 81 4.2 Device Design ..................................................................................................................... 84 4.3 Tumor Treatment System ................................................................................................... 87 4.4 In Vivo Mice Treatment Protocol ........................................................................................ 89 4.5 Experimental Results .......................................................................................................... 92 4.5.1 Characterization of Treatment Pressure ....................................................................... 92 4.5.2 Temperature Measurement and Thermal Toxicity Tests ............................................. 98 4.5.3 Treatment Results ...................................................................................................... 100 4.6 Discussions ....................................................................................................................... 102 4.7 Summary ........................................................................................................................... 105 Chapter 5 SFATs for Localized Neurostimulation ..................................................................... 107 5.1 Introduction ....................................................................................................................... 107 vi 5.2 The Patch-clamp Technique for Studying Neuronal Activities ........................................ 109 5.3 Device Design ................................................................................................................... 110 5.4 Experimental Results ........................................................................................................ 117 5.5 Summary ........................................................................................................................... 121 Chapter 6 Acoustic Tweezers Based on Modified Fresnel Acoustic Lenses.............................. 122 6.1 Introduction ....................................................................................................................... 122 6.2 Ring-Focusing Fresnel Lens for Long Depth-of-Focus Focusing and Multi-Beam Acoustic Trapping .................................................................................................................................. 123 6.2.1 Background and Motivation ...................................................................................... 124 6.2.2 Device Design ............................................................................................................ 125 6.2.3 Simulation of Acoustic Pressure ................................................................................ 129 6.2.4 Simulation of Acoustic Radiation Force (ARF) ........................................................ 133 6.2.5 Measurement of Acoustic Pressure ............................................................................ 135 6.2.6 Trapping of Polyethylene (PE) Microspheres ............................................................ 136 6.2.7 Summary .................................................................................................................... 139 6.3 Acoustic Tweezers Based on Multi-Foci Linear Fresnel Lens for Large Volume Particle Trapping .................................................................................................................................. 140 6.3.1 Device Design ............................................................................................................ 140 6.3.2 Experimental Results ................................................................................................. 143 6.3.3 Summary .................................................................................................................... 147 Chapter 7 Piezoelectric Micromachined Ultrasonic Transducer (PMUT) Based on Dome-Shaped Diaphragm................................................................................................................................... 148 7.1 Introduction ....................................................................................................................... 148 7.2 Device Overview .............................................................................................................. 149 7.3 Fabrication and Packaging ................................................................................................ 151 7.4 Experimental Results ........................................................................................................ 154 7.4.1 Measurement of Sound Output .................................................................................. 154 7.4.2 Increasing Sound Output through Structure Modification ......................................... 156 7.4.3 Increasing Sound Output through Phase Control ....................................................... 157 7.5 Summary ........................................................................................................................... 161 Chapter 8 Acoustic Propeller Based on Air Jets from Acoustic Streaming ............................... 162 8.1 Introduction ....................................................................................................................... 162 8.2 Device Overview .............................................................................................................. 163 8.3 Fabrication ........................................................................................................................ 164 8.4 Experimental Results ........................................................................................................ 166 8.4.1 Optimization of Design Parameters ........................................................................... 166 8.4.2 Acoustic Thruster Demonstrations............................................................................. 169 8.4.3 Wireless Thruster Operation ...................................................................................... 172 8.4.4 Acoustic Lifter Demonstrations ................................................................................. 172 8.5 Summary ........................................................................................................................... 174 Chapter 9 Summary and Future Directions ................................................................................ 175 vii 9.1 Conclusion ........................................................................................................................ 175 9.2 Future Work on Self-Focusing Acoustic Transducers (SFATs) ....................................... 176 9.2.1 Design and Microfabrication ..................................................................................... 176 9.2.2 Droplet Ejection ......................................................................................................... 177 9.2.3 Tumor Treatment ....................................................................................................... 177 9.2.4 Neurostimulation ........................................................................................................ 178 9.3 Future Work on Acoustic Tweezers ................................................................................. 179 9.4 Future Work on Piezoelectric Micromachined Ultrasonic Transducers (PMUT) ............ 179 9.5 Future Work on Acoustic Propellers ................................................................................ 180 References ................................................................................................................................... 181 viii List of Figures Figure 2.1 (a) Cross-sectional (across the dashed line in (b)) diagram of a typical SFAT based on PZT substrate with a Fresnel air-cavity acoustic lens made of Parylene on top, showing how the Fresnel acoustic lens focuses ultrasound in water by blocking destructively interfering acoustic waves. (b) Top-view diagram of the same SFAT, showing the relative positions of the top electrode, air-cavity rings, soldering pad, and Parylene-coated regions. ...................................... 14 Figure 2.2 (a) Cross-sectional (across the dashed line in (b)) diagram and (b) top-view diagram of an SFAT based on a PZT sheet with its top and bottom electrodes patterned into overlapping Fresnel half-wavelength rings connected by narrow rectangular electrodes. The rectangular electrodes and soldering pads on the top and bottom electrodes don’t overlap. ........................... 15 Figure 2.3 FEM-simulated normalized acoustic pressure of an SFAT working at 20.70 MHz, with six constructive Fresnel rings designed for 5 mm focal length in water (a) on the central vertical plane and (b) on the lateral focal plane (at Z = 5 mm), with same color bar scale but different dimension scales............................................................................................................. 17 Figure 2.4 (a) Simulated peak acoustic pressure versus the number of constructive Fresnel rings and operating frequency, when the vibration from lens surface is 1 nm, and the designed focal length is 8 mm. (b) The calculated active area of non-air-cavity regions and simulated peak pressure versus the designed focal length, when the operating frequency is 2.32 MHz, vibration amplitude from the lens is 1 nm, and the number of constructive Fresnel rings is 6. .................. 18 Figure 2.5 FEM-simulated relative acoustic pressure distribution in the focal planes of different SFATs working at (a) 6.90 MHz, (b) 11.65 MHz, and (c) 20.99 MHz. The SFATs have the same number of rings designed for the same focal length. .................................................................... 20 Figure 2.6 Simulated focal depth versus the product of the outermost ring width (ΔR) and the F#, along with linear fitting lines, for SFATs designed at 2.32 MHz, 6.60 MHz, and 11.00 MHz, respectively. .................................................................................................................................. 21 Figure 2.7 Fabrication process for an SFAT with patterned electrode rings: (a) pattern photoresists on both side of a nickel-coated PZT sheet through photolithography; (b) pattern electrodes through wet etching; (c) deposit Parylene for encapsulation. (d) Photo of a fabricated device before final Parylene deposition. ....................................................................................... 22 Figure 2.8 Fabrication process for an SFAT with a Parylene air-cavity lens: (a) Pattern electrodes on both sides of PZT; (b) spin-coat and pattern photoresist as sacrificial layer for air cavities; (c) deposit Parylene; (d) pattern release holes; (e) remove photoresist with acetone, clean and air dry; (f) deposit Parylene to seal the air cavities. (g) Photo of a fabricated device before wires are soldered. ............................................................................................................. 23 Figure 2.9 Fabrication process of an SFAT with a PDMS air-cavity lens: on glass plate, (a) create silicon mold through photolithography of SU-8; (b) replicate PDMS membrane from the ix silicon mold; on PZT sheet, (c) pattern top/bottom electrodes, (d) deposit Parylene, (e) spin-coat SU-8 photoresist or UV-curable adhesive, (f) bond the PDMS (peeled off from the silicon mold) onto PZT after alignment, then cure the photoresist or UV-curable adhesive, (g) solder electrical wires (not shown), and deposit Parylene for electrical encapsulation. (h) Photo of a fabricated device before wire soldering and the final Parylene deposition. .................................................. 25 Figure 2.10 Fabrication process of an SFAT with a SU-8 air-cavity lens: on the PZT sheet, (a) pattern top and bottom electrodes, (b) deposit Parylene, (c) pattern the bottom SU-8 layer; (d) spin-coat SU-8 on a PET film supported by a carrier wafer, and soft bake for a prolonged period of time, (e) peel off the SU-8 coated PET film, (f) bond the top and bottom SU-8 layers with a laminator, (g) pattern the top SU-8 through photolithography, (h) peel off the PET film. (i) Photo of a fabricated SFAT before wire soldering. ................................................................................ 27 Figure 2.11 F Fabrication process of an SFAT with a SU-8 air-cavity lens: (a) pattern top and bottom electrodes of a PZT sheet, (b) deposit Parylene, (c) pattern the bottom SU-8 layer, (d) attach a piece of PDMS membrane as the top layer of the lens, (e) the PDMS and SU-8 are permanently encapsulated within a Parylene layer. (f) A photo of a fabricated device, showing the gaps on the SU-8 layer to allow pressure equalization during the final Parylene deposition. 28 Figure 2.12 Measurement setup for measuring acoustic pressure with hydrophone. .................. 31 Figure 3.1 Simulation of the normalized vertical particle displacement at (a) the focal plane (xy plane at Z = 5 mm), and (b) the central vertical plane perpendicular to the transducer (xz plane at Y = 0). ........................................................................................................................................... 34 Figure 3.2 Photos of (a) the front side of a fabricated device with air-cavity-based acoustic reflector lens (light grey areas) on nickel electrode (dark grey areas) sealed with Parylene, and (b) packaged device with acrylic liquid reservoir. ........................................................................ 35 Figure 3.3 Measurement set-up for capturing photos of droplet ejection. ................................... 36 Figure 3.4 Photos of the ejected droplets taken with optical strobing at different time points after actuation with 18.18 µs pulse width. Smaller droplets in the background are the ones falling downward after being ejected and reaching the highest point. ..................................................... 37 Figure 3.5 Photos showing ejection pattern 700 μs after actuation at different pulse widths. Smaller droplets in the background are the ones falling downward after being ejected and reaching the highest point. ............................................................................................................ 38 Figure 3.6 Photos of the ejected droplets taken with optical strobing at different time points after actuation with (a) 22.72 µs and (b) 31.82 µs pulse width showing the generation of satellite droplets (marked with numbers greater than 1) at longer pulse widths. Smaller droplets in the background are the ones falling downward after being ejected and reaching the highest point. . 39 Figure 3.7 Photos taken at different times while droplets are continuously being ejected, showing the stability of the ejection. ............................................................................................ 40 x Figure 3.8 Photos of the ejected droplets taken with optical strobing at different time points after actuation with (a) IPA as medium, 17.24 μs pulse width; (b) hydrocarbon oil as medium, 107.76 μs pulse width. The right photo in (b) was taken at higher vertical level than that of the left photo, because the droplet separates from oil column at a very late stage. Smaller droplets in the background are the ones falling downward after being ejected and reaching the highest point. . 41 Figure 3.9 (a) Cross-sectional-view schematic of the transducer with electrically tunable focal size. (b) Top view of the transudcer, showing white annular-ring areas that represent air-cavities that block acoustic waves; (c) bottom view of the transducer, showing the bottom electrodes that are patterned into six annular rings, so that we can individually select corresponding Fresnel rings on the top side to actuate. Photos of (d) top side of the transducer, showing eight Fresnel rings (dark grey ring areas) separated by eight air-cavity rings (light grey ring areas) with filled release holes (at 0°, 90°, 180°, 270° positions of each ring) on a circular Nickel electrode; and (e) bottom side of the transducer, showing six electrode rings with wires soldered. ......................... 43 Figure 3.10 Cross-sectional-view schematic of the transducer, showing how the focal size changes with (a) 4 Fresnel rings and (b) 2 Fresnel rings actuated from center. With more Fresnel rings actuated from center, the focal size will be smaller. ............................................................ 44 Figure 3.11 Simulated relative acoustic pressure distribution in central vertical plane with (a) 2 Fresnel rings, (b) 4 Fresnel rings, and (c) 8 Fresnel rings actuated from center; and that in focal plane 6 mm from device with (d) 2 Fresnel rings, (e) 4 Fresnel rings, and (f) 8 Fresnel rings actuated from center, showing the focal size becoming smaller with larger number of actuated rings............................................................................................................................................... 45 Figure 3.12 Measured normalized acoustic pressure: (a) along the central vertical axis with 8 rings being actuated, showing focal length of 6 mm and (b) along a central lateral axis at the focal plane with different numbers of rings actuated from center, showing varied focal sizes (diameters). ................................................................................................................................... 47 Figure 3.13 Photos showing sub-mm-sized water droplets of different diameters ejected by our focal-size-tunable transducer, when (a) 2 rings, (b) 3 rings, (c) 4 rings, (d) 5 rings, (e) 6 rings, (f) 8 rings were actuated from the center. The arcs at the bottom of each photo are part of air-cavity rings on the top of the transducer and the red area is from LED illumination. ............................ 48 Figure 3.14 Measured and simulated focal diameter, outermost ring width estimation, and diameter of ejected droplets when different number of rings from the center were actuated....... 49 Figure 3.15 Top-view photo of the ejector on 2-mm-thick PZT substrate (brown), showing five air-cavity rings (light grey with holes, for blocking out-of-phase waves) alternating with five Parylene-covered electrode regions (dark grey circle and four rings, where in-phase waves could pass). The holes on air-cavity ring are release-holes sealed after releasing sacrificial layers to create the air cavities ..................................................................................................................... 51 Figure 3.16 (a) Defined 2D simulation area with axisymmetry, with related simulation settings shown in Table I. Simulated relative acoustic pressure distribution ignoring reflection from the xi SPT-air interface: (b) over the central vertical plane, (c) over focal plane at Z = 22 mm. Hydrophone measurement of acoustic pressure in SPT solution (d) along the central vertical axis and (e) along the central lateral axis on the focal plane, with 85 V pp applied on the ejector. (f) Simulated acoustic pressure (color-bar unit: MPa) over central vertical plane with acoustic reflection from the SPT-air interface (20.04 mm above the device’s top surface), with 400 V pp applied to the transducer (pressure values normalized from measurement data in (d) and (e)). (g) Magnitude (color-bar unit: ×10 6 N/m 3 ) and direction (white arrows) of the acoustic-field-induced body force near the SPT-air interface, calculated from the pressure distribution in (f). .............. 54 Figure 3.17 Simulation snapshots of ejected droplets with 400 V pp applied on the transducer with driving pulse width of (a) 517 μs, (b) 862 μs, (c) 1,724 μs and (d) 2,586 μs. Simulation snapshots of ejected droplets with 1,724 μs driving pulse width and driving voltage of (e) 310 Vpp, (f) 340 Vpp, (g) 370 Vpp, and (h) 430 Vpp. Simulation snapshots of the water columns formed at the moment when acoustic signal is turned off with driving voltages and pulse widths of (i) 400 Vpp, 517 μs; (j) 400 Vpp, 2,586 μs; (k) 310 Vpp, 1,724 μs; and (l) 400 Vpp, 1,724 μs, respectively. .................................................................................................................................. 57 Figure 3.18 (a) Simulated main droplet diameter and the equivalent diameter calculated from VPW (bulging water column volume when acoustic signal is turned off) versus driving pulse width (with 400 Vpp applied) and driving voltage (with 1,724 μs pulse width). (b) VPW (in μL) versus uPW (maximum fluid speed along the central vertical axis when acoustic signal is turned off, in m/s) multiplied by pulse width (in ms) at different driving conditions (driving voltage in Vpp and pulse width in ms), fitted by a linear trendline. ............................................................... 59 Figure 3.19 Photos of the ejected droplets with 394 V pp applied on the transducer with driving pulse width of (a) 517 μs, (b) 862 μs, (c) 1,724 μs, (d) 2,586 μs and (e) 6,034 μs. Photos of the ejected droplets with 1,724 μs driving pulse width and driving voltage of (f) 283 V pp, (g) 312 Vpp, (h) 339 Vpp, and (i) 368 Vpp. Scale bar length is 2 mm in all photos. All photos are taken with optical strobing, except that (e) is taken with a high-speed camera. .................................... 61 Figure 3.20 Graph showing measured main droplet diameter with different driving voltages and driving pulse widths. ..................................................................................................................... 61 Figure 3.21 (a) Cross-sectional diagram (across the center line along the channel on cover) of the ejection setup. (b) Top-view photo of a laser-machined channel-embedded plastic cover (designed for chips with 1,600 μm side length) held by an acrylic holder, aligned to the ejector at the container bottom, with silicon chips floating in the channel. Photos of ejected droplets of differernt sizes carrying 0.4-mm-thick silicon chips having side length of (c) 700 μm, (d) 1,600 μm (with a satellite droplet), and (e) 3,100 μm (with satellite droplets). ..................................... 63 Figure 3.22 Photos showing the semi-automatic chip loading mechanism: (a) chips are dumped into the inlet of the flow channel, also showing the ejection trajectory of liquid droplets under a weak ejection; (b) chips moving in along the channel as an ejected liquid droplet flies in air; (c) chips moving in further; (d) front chip loaded at the ejection site after seven ejections. Photos of 400-μm-thick square silicon chips with side length of (e) 700 μm and (f) 1,600 μm ejected into xii 4×3 arrays with an interval of 5 mm collected on filter paper, with red crosses showing the centers of ejected chips in five other repeated trials with the center of the left top chip aligned together. ........................................................................................................................................ 64 Figure 3.23 FEM-simulated relative acoustic pressure distribution: on the central vertical planes for SFATs (designed for three different harmonics at 6.60 MHz, 11.00 MHz, and 20.96 MHz) working at (a) 6.90 MHz, (b) 11.65 MHz, and (c) 20.99 MHz, respectively; and in the focal planes (dashed lines) for the same SFATs working at (d) 6.90 MHz, (e) 11.65 MHz, and (f) 20.99 MHz. ................................................................................................................................... 69 Figure 3.24 Photos of fabricated devices on PZT substrates working at (a) 6.90 MHz, (b) 11.65 MHz, (c) 20.99 MHz, showing the air-cavities (shiny areas), and the same devices working at (d) 6.90 MHz, (e) 11.65 MHz, (f) 20.99 MHz under a digital microscope, showing air cavities (light grey areas), Parylene covered electrode (dark grey areas), and sealed release holes. ................. 69 Figure 3.25 FEM-simulated relative acoustic pressure distribution when the bottom of a Petri dish (with 0.75-mm-thick bottom plate, red line) filled with 0.98-mm-thick 1% agarose gel (yellow lines) is 1.5 mm above SFAT surface in water: on the central vertical plane for SFATs working at (a) 6.90 MHz, (b) 11.65 MHz, and (c) 20.99 MHz, respectively; and in the focal planes (dashed lines) for the same devices working at (d) 6.90 MHz, (e) 11.65 MHz, and (f) 20.99 MHz. The color ranges are adjusted to be the same as those in Figure 3.23. ..................... 71 Figure 3.26 FEM-simulated normalized peak pressure versus 1% (w/v) agarose gel thickness. The navy dashed vertical line shows the selected thickness of 0.98 mm. In this simulation, the top of the gel is fixed at 4.5 mm above SFAT surface. (b) Experimental results on the relationship between gel volume before solidifying and the resulted gel thickness after solidifying at the center of a Petri dish with 90 mm diameter. ..................................................... 72 Figure 3.27 Measured peak acoustic pressure at the focal point and focal length from different SFATs with and without the agarose-gel-filled Petri dish. ........................................................... 73 Figure 3.28 Measurement set-up for capturing photos of droplet ejection with optical strobing; (b) cross-sectional diagram showing the ejection set-up with a Petri dish filled with agarose gel. ....................................................................................................................................................... 74 Figure 3.29 Photos showing water droplets ejected without the Petri dish by our SFATs working at (a) 6.90 MHz, (b) 11.65 MHz, and (b) 20.99 MHz and that with the Petri dish for SFATs working at (d) 6.90 MHz, (e) 11.65 MHz, and (f) 20.99 MHz. The background circles are ripples on the water surface. ..................................................................................................................... 75 Figure 3.30 (a) Cross-sectional diagram showing the droplet-assisted particle ejection set-up. (b) Photo of an ejected droplet carrying fluorescent microspheres under black light, ejected from the Petri dish by the 6.90 MHz SFAT and flies above the beaker edge. ............................................ 77 Figure 3.31 Microscope photos of (a) microsphere monolayer on the gel surface; collected microsphere agglomerates on plastic cover slips, ejected by SFATs working at (b) 6.90 MHz, (c) xiii 11.65 MHz, (d) 20.99 MHz, respectively. (e) Diameters of collected microsphere agglomerate on cover slips, ejected droplets without Petri dish and ejected droplets with Petri dish, versus SFAT operating frequency. ..................................................................................................................... 77 Figure 3.32 (a) Top-view photo of the 20.12 MHz SFAT used for cell ejection. (b) Simulated relative acoustic pressure distribution on the central vertical plane above an SFAT working at 20.12 MHz with a Fresnel lens designed for 20.96 MHz and 5-mm focal length. (b) Cross- sectional diagram showing the droplet-assisted cell ejection set-up. (c) Microscope photos of 100% confluency human retinal pigment epithelium (RPE) mono-layer cells (d) before and (e) after an ejection of cells by SFAT. (f) Photo of the same mono-layer cells when the cells are re- cultured (for 4 days) after the cell ejection. .................................................................................. 79 Figure 4.1 (a) Top-view photo of a fabricated SFAT before wires are soldered. (b) Microscope photo of part of the transducer (the dashed rectangular area in (a)), showing parts of five air- cavity rings with sealed release holes on the top electrode. The outermost air-cavity ring (top one in photo) covers part of the electrode and PZT. ............................................................................ 84 Figure 4.2 FEM-simulated normalized acoustic pressure (e) on the central vertical plane and (f) on the lateral focal plane (at Z = 5 mm), with same color bar scale but different dimension scales. ....................................................................................................................................................... 86 Figure 4.3 Photos of (a) a packaged SFAT on an acrylic holder, with soldered electrical wires connected to an SMA adapter for electrical connection and a hollow water-cooling block in a close-loop water-cooling system to dissipate heat from the transducer surface (Inset: front-view of the transducer clamped on the holder); (b) the 3D-printer-modified three-axis positioning system with a holding platform with the transducer attached and a heating platform where the mouse will be placed, showing how the movement in each axis is realized; (c) a close-up view of the dashed rectangular area in (b) showing the transducer holder on the right, and the laser diode in the black housing with its driving circuit on the left, all held by the movable holding platform with screws; (d) a mouse lying on the heating platform of the positioning system, anesthetized by isoflurane gas coming from an anesthesia nose cone where its nose is placed (as excessive gas is drawn away to the large exhaust pipe nearby), with a red laser dot aligned to the center of its tumor (highlighted with a dark skin marker). (e) Side-view cross-sectional diagram of the in vivo tumor treatment setup on mice. (f) Top-view diagram of the circular raster scan pattern of the transducer covering a circle of 4.8 mm diameter (dotted circle), showing the scan route (yellow- highlighted solid line) with 204 treatment spots (red dots). The width of the yellow line and the diameter of small solid circles are drawn in scale to indicate the focal size of the transducer, showing the actual treated area. .................................................................................................... 89 Figure 4.4 (a) Hydrophone measurement (black) and normalized simulation (red) of the output acoustic pressure in water along the central vertical axis, with 211 V pp applied on the transducer. (b) Cross-sectional diagram of the hydrophone measurement setup with a piece of tissue slice attached to the top of the SFAT with ultrasound transmission gel. (c) Natural logarithm of the measured acoustic pressure at the focal point when tumor (grey dots) or skin (red dots) slices of xiv different thicknesses are placed on the SFAT, with linear fitting lines whose slopes are equal to the attenuation coefficients in the tissues (unit: np/mm). ............................................................. 95 Figure 4.5 Simulated acoustic pressure distributions during treatment for 0.4 mm skin thickness, if the distance between the SFAT and the bottom of skin is (a) 4.6 mm, (b) 3.7 mm, and (c) 3.1 mm; and similar simulations for 0.9 mm skin thickness, if the distance between the SFAT and the bottom of skin is (d) 4.6 mm, (e) 3.7 mm, and (f) 3.1 mm; all sharing the same color bar in (a) with unit being MPa. (g) Simulated maximum acoustic pressure (and corresponding mechanical index) in tumor tissue during treatment versus different skin thicknesses as a function of SFAT-skin distances. ................................................................................................. 97 Figure 4.6 (a) Typical temperature change on the skin above the treated tumor center (grey) and on an untreated skin area nearby with no tumor underneath (red) during the treatment with 1.45 ms pulse width. (b) Representative histology images of untreated normal skin (upper panel) compared with skin harvested 24 hours after treatment (lower panel) in thermal toxicity experiment with a maximum temperature of 35 °C. No histologic changes are noted. ............. 100 Figure 4.7 Tumor weights of control (n = 5) and treated (n = 8) tumors shown in boxplots ( * : student’s t-test p < 0.05). ............................................................................................................ 101 Figure 4.8 Representative cross-sectional histologic images of control and treated B16F10 tumors, showing matched sections representing control (untreated) and ultrasound-treated tumors that are stained for H&E ((a) and (b)), Ki67 (an indicator of cell proliferation, (c) and (d)), and cleaved caspase-3 (an indicator of cell apoptosis, (e) and (f)). *: viable tumor; ▲: normal skin overlying tumor; solid arrow and solid arrow head: area of necrosis and apoptosis; open arrow and open arrow head: large area of proliferating cancer cells without apoptosis. ...................... 102 Figure 5.1 [178] (a) Schematic diagram showing the setup of patch-clamp experiment: current flux within the patch pipette, sensing electrode and amplifier; (b) A phase contrast image of a patch pipette attached to the membrane of a cultured murine hippocampal neuron. ................. 110 Figure 5.2 (a) Cross-sectional schematic view and (b) top-view photo of a 127-μm-thick electrode-ring SFAT designed for neurostimulation experiments, aCSF is short for artificial cerebrospinal fluid; (c) Microscope photo of the SFAT with infrared illumination from the back, showing good visibility through the translucent PZT substrate; (d) Bottom view of an SFAT with a 0.5-mm-thick polyester sheet attached on the back for improved mechanical robustness and easiness of handling. ................................................................................................................... 112 Figure 5.3 FEM-simulated relative acoustic pressure distribution of the SFAT (a) in the central vertical plane, (b) at the focal plane where Z = 0.4 mm). ........................................................... 113 Figure 5.4 (a) Design and (b) microscope photo of 0.4-mm-focal-length SFAT on 127-mm-thick PZTs with 7 rings (with 8 th -15 th rings cleared) for 18.4 MHz. (c) Design and (d) microscope photo of an SFAT designed for the same focal length and frequency with 12 rings (with 8 th -10 th rings cleared). (e) Microscope photo of an SFAT with a drilled hole and circular slits through xv micro-powder blasting, with a piece of blue tape on the other side to demonstrate the improved visibility. ..................................................................................................................................... 114 Figure 5.5 Top-view photos of SFATs based on 15-ring Parylene air-cavity lenses designed for the same 0.4 mm focal length, at operating frequencies of (a) 18.4 MHz (the fundamental thickness-mode resonant frequency of a 127-μm-thick PZT); and (b) 54.0 MHz (the 3 rd harmonic thickness-mode resonant frequency of a 127-μm-thick PZT). (c) The electrode pattern (black) below the air-cavity lens, with exposed areas to let light pass through. ..................................... 115 Figure 5.6 Photos of (a) a control SFAT with a large Parylene-sealed air cavity covering most of the electrode regions; (b) an air-cavity cover made of three pieces of acetate sheets, before being attached to an SFAT; and (c) a control SFAT with an acetate-sheet-sealed air cavity. ............. 116 Figure 5.7 Cross-sectional-view diagram of the rat-brain-tissue patch-clamp experiment setup with an SFAT. Photos of the experiment setup (b) without a microscope lens, showing a piece of brain slice on top of an SFAT patched by a micropipette; and (c) with a microscope lens. (d) Microscope photo showing a neuron cell within the brain tissue slice patched by a micropipette during the experiment. ................................................................................................................ 118 Figure 5.8 Measured membrane potential from a patched neuronal cell: (a) by focused ultrasound delivered by SFAT driven at 2, 3 and 4 V pp and (b) enlarged traces. Arrows show stimulation artifact. Action potentials (APs) were not generated by 2 V pp, but 3 Vpp could generate APs, but not all stimuli could generate AP. Each ultrasonic stimulation with 4 V pp could generate AP, the peak of which was 80-100 mV. ....................................................................... 119 Figure 5.9 Comparison of action potentials (APs) between electrically-evoked and ultrasound stimulation evoked: (a) Representative cellular responses to electrical and ultrasound stimulation. For electrical stimulation, 100 pA current was injected through the patch pipette. (b) Enlarged traces. Filled circle is a single AP. Arrow is an artifact. ............................................................. 120 Figure 5.10 APs without and with Tetrodotoxin (TTX), a Na+ channel blocker: (a) AP generation evoked by ultrasound stimulation. (b) Ultrasound evoked APs were abolished by TTX (0.3 μM), the vertical axis is not at the same scale. .................................................................... 121 Figure 6.1 (a) Cross-sectional schematic (across A-A’ in (c)) of the transducer, illustrating how the ring-focusing air-cavity Fresnel lens is designed to generate long depth-of-focus focal zone and many trapping zones. (b) Axisymmetric cross-sectional schematic (across A-O in (c)) of the transducer, showing how the ring radii are calculated. Top-view diagram (c) and photo (d) of the 2.32-MHz transducer on PZT (brown in (d)) designed for a focal ring with F R = 8 mm (red dashed circle in (c)) and FZ = 12 mm, showing sound-blocking air cavities (black in (b) and (c), shiny grey in (d)) and sound-passing non-air-cavity rings (white in (b) and (c), dark grey in (d)). ..................................................................................................................................................... 127 Figure 6.2 The amplitude (a) and phase (b) of a modified Airy function versus lateral distance. (c) The transmission function of the ring-focusing Fresnel lens, which closely resembles the Airy phase pattern shown in (b). ......................................................................................................... 129 xvi Figure 6.3 Simulated acoustic pressure amplitude in XZ plane for the ring-focusing transducer (a) with outer four Fresnel rings actuated to create inwards-bending Airy-like beams; (b) with inner four Fresnel rings actuated to create outwards-bending Airy-like beams; (c) with all rings actuated. (d) Bessel-beam-like radial pressure amplitude distributions (red) within the focal zone at Z = 29.3 mm (upper) and Z = 34.2 mm (lower) compared with scaled Bessel functions of the first kind (black). (e) Simulated acoustic pressure amplitude in XZ plane for a normal Fresnel lens with similar focal length and aperture size. (f) Relative pressure isosurfaces of 0.15 showing multiple bottle beams on the central axis and many “acoustical belts” in off-axis areas; (g)-(j) Acoustic pressure amplitude (in XZ plane (g) and in XY plane (i)) and phase (in XZ plane (h) and in XY plane (j)) distribution for a bottle beam located at Z = 14.85 mm on the central axis (highlighted in (f)). (k) Simulated acoustic pressure amplitude in XZ plane for the ring-focusing transducer during trapping experiments with acoustic reflection from water surface (58.0 mm above the transducer surface). The pressure values in all figures are normalized to the values in (c). ............................................................................................................................................... 131 Figure 6.4 Simulated acoustic pressure amplitude (without reflection, normalized to the values in Figure 6.3c) in XZ plane for the ring-focusing transducer (a) with two copper rings blocking some acoustic waves; (b) with two PE microspheres and one PE ring in potential trapping regions (axes rescaled to maintain circular (and square) cross sections of the objects). ............ 132 Figure 6.5 Simulated acoustic radiation potential (colorbar unit: 10 -14 Joules) and acoustic radiation force (ARF, white arrows) for 70-μm-diameter PE microspheres, showing fully three- dimensional trapping forces towards the center of bottle beams at Z = 21.2 mm (a) and 25.3 mm (b). (c) Same plot for off-axis Airy-shaped acoustical belt regions, showing ARF for particle trapping and rotation (with arrow length logarithmically normalized). (d) Simulated vertical ARF exerted on a 1-mm-diameter PE microsphere centered at different positions of the central axis (red) and the pressure amplitude (black), along with the required vertical lifting force (blue dashed line) and potential stable trapping positions (blue circles). (e) Simulated acoustic pressure amplitude showing regions with pressure amplitude higher than 0.045 MPa for potential trapping capability (colorbar unit: MPa). In all figures, the acoustic pressure distribution is normalized from Fig. 3k with a peak pressure of 0.226 MPa, which corresponds to the case where 35 Vpp is applied on the transducer (according to measurement data in Figure 6.6). ...... 135 Figure 6.6 Measurement (red, with 150 Vpp applied on the transducer) and normalized simulation (black) of acoustic pressure along the central axis. .................................................. 136 Figure 6.7 (a) Photo of the experimental set-up for all the trapping experiments. Side-view photos showing (b) trapping of three 1-mm-diameter PE microspheres (1.025 g/cm 3 density) through picking and releasing them one by one into water with a pipette; (c) trapping process of five out of six 1-mm-diameter PE microspheres released at the same time from water surface, including two sticking with each other; (d) two 1-mm-diameter PE microspheres (glued together) rotating clockwise while being trapped. The transducer is driven with continuous sinusoidal signals of 35 Vpp in (b) and 30 Vpp in (c) and (d). ...................................................................... 137 xvii Figure 6.8 Side-view photos showing (a) positions of a trapped 1-mm-diameter microsphere (1.025 g/cm 3 density) before and after the transducer is moved to the right for 2.5 mm at an average speed of 60 μm/s. The microsphere (green dash line) remains trapped while following transducers' movement (marked by the left edge of the acrylic holder below it, red dash line). Side-view photos showing (b) trapping process of multiple 350-μm-diameter PE microspheres (1.13 g/cm 3 density) after blowing with a pipette to make them float above transducer surface; (c) trapping of six 1-mm-diameter PE microspheres using the same technique as in (b). The transducer is driven with continuous sinusoidal signals of 30 V pp in all figures. ....................... 139 Figure 6.9 (a) Side-view schematic of the acoustic tweezer with multi-foci linear Fresnel lens based on air cavities, showing how air cavities prevent destructive waves from reaching the designed focal point; also indicated are multiple focal lengths from multiple pairs of Fresnel air- cavity lens and the resulting Bessel beam region with negative axial radiation force for particle trapping; (b) Top view of the Fresnel lens design, showing seven pairs of rectangular air cavities (in white) for blocking destructive acoustic waves and non-air-cavity regions (in black) to let pass constructive waves. (c) Photo of the front side of the fabricated acoustic tweezers, showing seven pairs of Fresnel bands (in dark grey areas where Parylene covers top Nickel electrode) separated by seven air-cavity bands (light grey areas) with filled release holes (three circles on each rectangular air cavity). ........................................................................................................ 142 Figure 6.10 FEM-simulated absolute sound pressure in (a) central XZ and (b) central YZ trapping planes perpendicular to the device, and (c) in a XY trapping plane at Z = 6.75 mm. .. 143 Figure 6.11 Hydrophone measurement (red) and COMSOL simulation (black) of the relative acoustic pressure along the central vertical axis vs the distance from the tweezers. .................. 144 Figure 6.12 Experiment set-up for characterizing particle trapping by the acoustic tweezers (not in scale). ...................................................................................................................................... 145 Figure 6.13 Photos showing a trapping process: (a) before trapping a large Delrin plastic piece, (b) with the piece trapped, (c) with the piece still trapped, as a nearby microsphere (that is not trapped) goes away. .................................................................................................................... 146 Figure 6.14 (a)(b)Photos showing trapping of a 1.9 mg rectangular silicon chip while un- trapped microspheres moving around. Photos showing trapping process of a 3.9 mg rectangular silicon chip along with two microspheres: (c) before trapping, (d) trapping of a silicon chip and two microspheres, (e) an un-trapped microsphere goes away as three objects still trapped. ..... 147 Figure 7.1 (a) Schematic cross-section view of the piezoelectric dome-shaped diaphragm transducer, before breaking the wafer into individual chips. The pre-diced grooves were used for front-back alignment as well as post-process wafer breaking. Photos of the front side of the transducer taken with a camera (b) and with a scanning electron microscope (SEM). (d) Photo of the back side of the transducer. ................................................................................................... 150 xviii Figure 7.2 (a) Diagram showing the dicing lines (black) on one side of the wafer and the two wafer flats (red) for dicing alignment; (b) Cross-section photo of a diced 2-mm thick wafer, showing front-to-back alignment error of 48 μm........................................................................ 151 Figure 7.3 Brief microfabrication process: (a) Dice both sides, deposit SiN, apply tape, and deposit Al; (b) Pattern Al, tape and SiN; (c) Isotropically etch Si in HNA etchant; (d) Remove tape and SiN; (e) Deposit LPCVD SiN; (f) Release the diaphragm through KOH etching from backside and deposit Al bottom electrode; (g) Deposit PECVD SiN through a silicon shadow mask; (h) Deposit ZnO and top Ti/Au electrodes with another shadow mask. .......................... 153 Figure 7.4 Photos of (a) diced chips with the fabricated dome-diaphragm transducers from a 3- inch wafer; (b) a chip mounted and wire-bonded onto SMT adaptor. ........................................ 154 Figure 7.5 Setup for measuring the sound pressure level produced by the dome-diaphragm transducer. ................................................................................................................................... 155 Figure 7.6 (a) Measured sound pressure level (SPL) versus frequency over 10 kHz – 40 kHz range at 5mm away from the transducer with 30 V pp drive voltage. (b) Measured sound pressure versus drive voltage at 18.5 kHz, 5 mm away; (c) measured sound pressure level at different distances from 5 mm to 45 mm between the microphone and the transducer at 18.5 kHz, with 30 Vpp drive voltage. ........................................................................................................................ 156 Figure 7.7 Diagram (a) and photo (b) of the laser-cutted cantilever diaphragm. (c) Measured sound pressure level (SPL) of the cantilever-like-diaphragm transducer (red line) and unmodified-diaphragm transducer (blue line) versus frequency over 10 kHz – 40 kHz range at 5mm away from the transducer with 30 Vpp drive voltage. (d) Measured sound pressure of the cantilever-diaphragm transducer versus drive voltage at 20.4 kHz, 5 mm away. ...................... 157 Figure 7.8 COMSOL model of the dome-diaphragm structure with top electrode multilayer stacks (incorporated as one layer). .............................................................................................. 159 Figure 7.9 COMSOL simulation of (a) side view and (b) top view of the first resonant vibration mode (loudspeaker mode); (c) side view and (d) top view of the second resonant vibration mode (rocking mode). ........................................................................................................................... 160 Figure 8.1 Photos of the piezoelectric speakers: (a) front sides with (right) and without cover attached (left), (b) backsides with (right) and without cover attached (left), (c) cross-sections with (right) and without cover (left), compared with a U.S. quarter coin (middle). ................... 163 Figure 8.2 Cross-section schematics showing how zero-net-synthetic jets are generated (not in scale): (a) suction of air; (b) vortex ring generation during air ejection; c) vortex ring leaving the orifice when air is sucked in at the beginning of the next cycle. ................................................ 164 Figure 8.3 (a) photo of laser-drilled holes on a 0.5-mm-thick Polyester sheet taken with a digital microscope; (b) an off-center surface profile of an orifice scanned with a surface profilometer, showing the sloped nozzle-like profile. ...................................................................................... 165 xix Figure 8.4 Side-view photos of the device suspended with its wires (a) before (b) and immediately after applying continuous driving signal of 35 V pp. The initial momentum will push the device to a large angle before the device stabilizes at a smaller angle after swinging back and forth. ............................................................................................................................................ 166 Figure 8.5 Measured average stabilization angle versus (a) orifice diameter, (b) orifice diameter and cover orientation; (c) orifice interval; (d) orifice array size (row × column) and (e) tape thickness and type. Unless specified, the default experiment conditions are: 420 μm orifice diameter, “Type A” cover orientation, 1.2 mm orifice interval, 10×6 array size and 0.86-mm- thick “Type X” tape. ................................................................................................................... 169 Figure 8.6 Photos of device positions (a) before and (b) after one pulse of driving signal (4.01 kHz, 49 Vpp, 750 ms pulse width). The purple lines show the initial position of the device, each grid on the paper is 5.08×5.08 mm 2 . ........................................................................................... 170 Figure 8.7 Photos of device positions (a) before; (b) during one pulse of driving signal (1.74 kHz, 58 Vpp, 575 ms pulse width) and (c) during an unbalanced jump with one side lifted much high than the other. ..................................................................................................................... 171 Figure 8.8 Photos of device positions (a) before; (b) 0.84 sec after and (c) 1.79 sec after continuous drive signal (2.08 kHz, 35 V pp). ............................................................................... 171 Figure 8.9 Photos of device positions (a) before and (b) after 5 pulses of acoustic drive from a tweeter 1 cm above the device. ................................................................................................... 172 Figure 8.10 Photos of (a) a laser-cut Acrylic base with a large rectangular opening; (b) backside of the device attached onto the base with foam tape. ................................................................. 173 Figure 8.11 Photos of (a) the acoustic lifter with plastic pieces of 299 mg before actuating the device; (b) the plastic pieces being lifted up by the air jets from the device when driving signal (2.08 kHz, 49 Vpp) was applied. ................................................................................................ 173 Figure 8.12 Photos of (a) the metal piece sitting on the device before device was activated; (b) the metal piece rotated about 180° after 1.85 sec of the device operation.................................. 174 xx List of Tables Table 2.1 Comparison of the Microfabrication Processes Involved in Difference Fabrication Methods of SFATs ........................................................................................................................ 29 Table 3.1 Corresponding Relationship between Top Fresnel Rings and bottom Electrode Rings, with Inner and Outer Radii and Widths of Each Fresnel rings ..................................................... 45 Table 3.2 Key Simulation Settings ............................................................................................... 53 Table 3.3 Simulation Results ....................................................................................................... 71 Table 3.4 Droplet Diameters and Driving Conditions ................................................................. 75 Table 4.1 Design parameters of the transducer used for in vivo tumor treatment........................ 85 Table 4.2 Material Properties Used in The Simulation of the Treatment Acoustic Pressure ...... 96 Table 6.1 Dimensions of the Acoustic Tweezers. ...................................................................... 142 Table 7.1 Physical Dimensions of the Transducer ..................................................................... 150 Table 7.2 Material Properties Used in Simulation ..................................................................... 159 xxi Abstract With thoughtful engineering and arrangement, acoustic waves can be used as powerful and versatile tools for many applications. For generating acoustic waves and ultrasound, piezoelectric acoustic/ultrasonic transducers have been widely used due to their low loss, low cost, low power consumption, high acoustic output, and easy operation. Moreover, with acoustic lenses, the generated acoustic waves can be effectively modulated, realizing focusing effect or creating complex beam patterns without relying on an array of transducers. With microelectromechanical system (MEMS) technologies, piezoelectric acoustic transducers and acoustic lenses with smaller footprints, higher performance, and lower cost could be mass-produced with high precision through microfabrication processes. In this thesis, our research on the design, fabrication, and applications of four types of piezoelectric acoustic/ultrasonic transducers (two of them are equipped with acoustic lenses) based on MEMS technologies. The first type is self-focusing acoustic transducers (SFAT), which are single-element, planar ultrasonic transducers based on piezoelectric lead zirconate titanate (PZT) substrates vibrating in thickness mode with microfabricated Fresnel acoustic lenses on top to generate focused ultrasound. The design parameters, as well as microfabrication and characterization methods of SFATs are introduced. Exploiting their small size, electrical tunability, and microfabrication compatibility, different types of SFATs tailored for applications such as in vivo selective cancer treatment, neurostimulation, and acoustic droplet ejection (for cell extraction/delivery and semiconductor chip pick-and-placement) have been developed. The second type is single-element microfabricated acoustic tweezers, which are based on modified Fresnel acoustic lenses to achieve trapping and manipulation of large (sub-millimeter- or millimeter-sized) particles in liquid environments, with two designs demonstrated. The first design xxii is based on a ring-focusing Fresnel lens, which is capable of generating multiple acoustic trapping beams such as bottle beams (with three-dimensional trapping force) and quasi-Airy beams (with the capability to rotate trapped particles). The second design relies on a multi-foci linear Fresnel lens to generate large, cuboid-shaped trapping zones that can trap plastic and silicon chips with up to 5.3×0.67×0.51 mm 3 in size and 3.98 mg in weight. The other two types are both airborne acoustic transducers based on diaphragms vibrating in flexural mode. The third type is a piezoelectric micromachined ultrasonic transducer (PMUT) based on a hemispherical dome-shaped diaphragm driven by eight piezoelectric zinc oxide (ZnO) actuation elements designed to increase the sound output through releasing residual stress and transforming in-plane vibration to large radial deflection. The fourth type is an acoustic propeller capable of generating propulsion force from synthetic air jets generated by acoustic waves passing through small orifices on a thin plastic covering a mini membrane-type speaker. With different driving conditions and operating frequencies, the transducer that weighs 603 mg can rotate, jump, move, or lift objects, propelled by the air jets. The device could also be driven wirelessly with acoustic signals instead of electrical power. Finally, a summary of current research accomplishments, as well as future research directions will be presented. 1 1Chapter 1 Introduction 1.1 Applications of Sound Acoustic waves and ultrasound are powerful and versatile tools for a large variety types of applications. In our daily lives, as a signal perceived by one of our major senses, audible sound (with frequencies between 20 Hz and 20 kHz) is used for communication, perception, entertainment, and art, and thus considered an indispensable part of human culture. With frequencies higher than the human hearing limit, ultrasound has also been widely used in almost every aspect of our lives. Ultrasound has been extensively used in biomedical applications. For diagnosis purposes, ultrasound imaging is a safe, portable, and cost-efficient medical imaging modality [1]. Ultrasound can also be used to measure the stiffness of tissues (ultrasound elastography [2]) and the rate of blood flow [3]. Apart from diagnosis, ultrasound is also an effective tool for non-invasive therapy for cancer treatment [4], [5], neuromodulation [6]–[8], physiotherapy [9], [10], drug delivery [11], [12], etc. The acoustic-field-induced forces (such as acoustic radiation force) and acoustic streaming effects have also been utilized to achieve control and manipulation of objects. One important example is acoustic “tweezers”, which is a type of acoustic transducers capable of trapping [13], [14], levitating [15]–[18], or manipulating (such as rotating [19], [20] and moving [14], [18], [21]) objects in mid-air or liquid environments without relying on physical contact. Such devices have found a wide range of applications in life sciences [22]. 2 The high energy contained in acoustic waves could also enable actuation applications. For example, thrust force induced by acoustic waves could be used for propulsion [23]–[25]. In addition, liquid droplets could be ejected by focused ultrasound, and be used as carriers of liquids or solids [26]–[29]. High power ultrasound has also been used for cleaning [30], [31] and sonication [32], [33]. Additionally, ultrasound has been used for sensing and measurement, in applications such as nondestructive testing (NDT) (to evaluate and examine the properties of materials and detect potential defects) [34], [35], range finding (to measure the distance between the sensor and target objects) [36]–[38], and flow measurement [39], [40]. Acoustic transducers also allow simultaneous actuation and sensing. For example, in ultrasound haptics [39], [40], acoustic waves could both generate tactile sensations and detect human gestures, achieving human-computer interaction. In wireless communication, piezoelectric acoustic filters based on high-frequency (100 MHz to over 10 GHz) surface acoustic waves (SAW) and bulk acoustic waves (BAW) have been intensively used to avoid cross-talk and optimize the quality and throughput of different wireless transmission bands [41], [42]. Acoustic waves could also be used for wireless power transfer and delivery [43] to devices in hard-to-access regions such as biomedical implants in human bodies [44] and sensors in pipelines [45]. For fabrication and manufacturing, ultrasound has also been demonstrated its usefulness in applications such as soldering [46], welding [47], ultrasound impact treatment (UIT) (to enhance mechanical and physical properties of metals) [48]. 1.2 Overview of Acoustic Transducers 3 Acoustic transducers are devices capable of converting electrical energy into acoustic energy in the form of mechanical vibration (transmitters) or vice versa (receivers), or both (transceivers), depending on the application. The mechanical vibration is generally induced by surface traction (such as membrane deflection) or body distortion (such as the thickness or radial vibration of a piezoelectric disk/sheet) [49]. The energy conversion usually relies on electromagnetic, electrostatic, and piezoelectric effects. 1.2.1 Electromagnetic Acoustic Transducers Electromagnetic acoustic transducers typically consist of at least one permanent magnet and a conductive material (such as a coil) attached to a diaphragm, or a conductive diaphragm such as a corrugated ribbon [50]. Sound generation and sensing are usually based on electromagnetic induction. When an alternating current is applied through a conductive coil, a varying magnetic field will be generated around the conductive material. The induced magnetic field will interact with the static magnetic field from a permanent magnet nearby, inducing vibrational motions on the conductive material and the diaphragm, thus generating acoustic waves. Similarly, when incoming acoustic waves cause relative motion between a conductor and a permanent magnet, electrical current will be induced in the conductive material, allowing us to sense the audio signal [51]. Apart from electromagnetic induction, some electromagnetic acoustic transducers are based on magnetostriction, in which a ferromagnetic material undergoes dimensional change when a nearby external magnetic field is created, which can be induced through applying alternating current on a conductive material. These types of devices have been widely used as loudspeakers and microphones working at audio frequencies. Another type of electromagnetic acoustic transducers [52] don’t rely on a vibrating diaphragm, but could remotely generate or sense acoustic waves with different modes within a nearby 4 conductive (utilizing Lorentz force generated by induced surface eddy current) or ferromagnetic material (utilizing magnetostriction effect), and have been used as a non-contact method for NDT. One major limitation of electromagnetic acoustic transducers is their low energy transduction with non-trivial heat generation [53]. Moreover, the requirement of a permanent magnet poses several limitations in size reduction, array integration (due to its incompatibility with microfabrication processes), and range of applications. 1.2.2 Electrostatic Acoustic Transducers For electrostatic acoustic transducers, the mechanism of the energy transduction is the vibration of a thin plate under electrostatic forces [54]. This type of device is based on a capacitor structure, in which a thin metal membrane is suspended above a fixed back electrode (backplate) with a small gap in between. The capacitor is biased with a direct current (DC) voltage, and when the membrane vibrates in response to impinging acoustic waves, the gap height (and thus the device capacitance) varies, generating a varying output voltage (proportional to the amplitude of the acoustic waves) superimposed on the bias voltage. On the other hand, if an alternating current (AC) voltage is applied with the DC bias, driven by electrostatic force, the membrane will vibrate and generate sonic waves. The need of an external DC bias voltage could be avoided if the backplate is made of an electret, which is a stable dielectric material with a quasi-permanently embedded electric charge [55]. Common examples of electrostatic acoustic transducers include condenser microphones and loudspeakers. Recent advances in MEMS and micromachining technologies have led to the development of capacitive micromachined ultrasonic transducers (CMUTs) that can form two-dimensional (2D) or three-dimensional (3D) arrays used for advanced applications such as medical imaging and therapy [56], [57]. 5 The major advantages of electrostatic acoustic transducers include good impedance match to air and wide bandwidth [58]. On top of these, CMUTs also offer benefits such as low cost, high electromechanical coupling coefficient (can be made to be close to unity), and high levels of integration that can be compatible with complementary metal-oxide-semiconductor (CMOS) integrated circuits (IC) [59]. The limitations of electrostatic acoustic transducers are the need for a bias voltage, limited output pressure and parasitic capacitance, and as well as long-term reliability [60]. 1.2.3 Piezoelectric Acoustic Transducers Acoustic transduction could also be realized by piezoelectric acoustic transducers. Piezoelectric effect refers to the phenomenon in which a material, upon the application of an electrical field, changes its physical dimensions (reverse piezoelectric effect) and vice versa (direct piezoelectric effect) [61]. Typically, a piezoelectric material is sandwiched by two electrodes, where electrical potential could be applied or detected to generate or sense acoustic waves, respectively. Some piezoelectric acoustic transducers are based on the body distortion (e.g., thickness- mode or radial-mode vibration) of bulk piezoelectric materials such as quartz, lithium niobate (LiNbO3), lead zirconate titanate (PZT), lead magnesium niobate-lead titanate (PMN-PT), and polyvinylidene difluoride (PVDF) [1]. Thanks to the desirable properties of these piezoelectric materials such as low electrical and mechanical loss, high electromechanical coupling coefficient, and low cost, single-element or arrays of piezoelectric transducers with very high sensitivity have become the most common tools used in medical imaging. Transducers based on composite piezoelectric materials (such as PZT rods embedded in a low-density polymer) [62] have also been developed to achieve further improve electromechanical coupling coefficient, sensitivity and 6 bandwidth. However, this type of transducer requires an acoustic matching layer to ensure good energy transmission into the target medium, and are usually fabricated through macro-machining and mechanical dicing with relatively large size, high die-to-die variation, and incompatibility with CMOS IC [63]. In recent years, based on the MEMS technology, piezoelectric micromachined ultrasonic transducers (PMUT) with smaller size, higher element density, and potential CMOS IC compatibility have been developed and used in applications such as intravascular ultrasound (IVUS) diagnosis, fingerprint sensing, and range-finding [64]. Similar to a CMUT, a PMUT is based on a piezoelectric membrane that vibrates in flexural modes, and thus doesn’t require an acoustic matching layer. Thin-film piezoelectric materials such as zinc oxide (ZnO), aluminum nitride (AlN), and PZT, which are compatible with microfabrication processes are commonly used. Compared to CMUTs, PMUTs do not require a DC bise voltage to operate, thus having much smaller power consumption and less requirement on electronics. Also, the deflection of the membrane is less limited since there is no vacuum gap between the top and bottom electrodes [65]. However, the performance of PMUTs is limited by the hard-to-control intrinsic stress within the diaphragms, and poor electromechanical coupling coefficient that comes from the difficulties in manufacturing high-performance piezoelectric thin films [65]. In this thesis, we will present piezoelectric transducers that overcome these limitations, including microfabricated piezoelectric transducers based on bulk PZT sheets, and a PMUT based on a dome-shaped diaphragm with improved sound output. 1.3 Overview of Acoustic Lenses Based on similar working principles as the optical lenses, acoustic lenses can be used to modulate the acoustic waves from and to acoustic transducers, realizing focusing effect or creating 7 certain beam patterns (such as Bessel beams [66] and Airy beams [67]) without relying on an array of transducers, which requires advanced manufacturing techniques and complex control circuits. Acoustic lenses made of metal could be fabricated through macro-machining (such as milling) [67]–[70]. However, due to the limited manufacturing precision, they are vulnerable to fabrication defects such as curvature error and surface roughness. Also, the high acoustic impedance of metals results in poor acoustic matching between the transducer and commonly used sound-bearing media (such as water and soft tissues). Polymer acoustic lenses with lower acoustic impedances could be manufactured through 3D printing [66] and replica molding [71]. However, these fabrication techniques have low throughput and are not suitable for massively parallel manufacture. Acoustic lenses based on metamaterials can achieve complex modulation (such as converting spherical waves into plane waves [72]) of acoustic waves with extraordinary performances (such as very high transmission [73]). Some metamaterial-based acoustic lenses are based on complex 3D structures that are hard to fabricate [72]–[74], others rely on planar structures such as multilayer stacks of different materials [75], [76], and 2D arrays of structures [77], requiring critical control over the thicknesses and dimensions of the lens structures. Fresnel acoustic lens is a type of planar acoustic lens that can be mass-produced with microfabricated. For example, the phase of acoustic waves could be modified when they pass through Fresnel lenses having annular-ring structures of different thicknesses fabricated through etching and bonding [78], [79]. However, it’s challenging to precisely control the layer thicknesses, especially when there are multiple layers, resulting in undesirable errors. On the other hand, our fully-microfabricated air-cavity Fresnel lenses are easy to fabricate [80], [81], since the modulation of phase is determined by the lateral patterns of sound-blocking air cavities for filtering out out- of-phase waves, thus thickness control is no longer critical. Apart from generating focused 8 ultrasound, through modifying the air-cavity patterns, the lens could also produce other acoustic beam patterns such as Bessel beams [82], [83], bottle beams [83], [84] and quasi-Airy beams [83] for applications such as acoustic trapping. 1.4 Scope of the Thesis My Ph.D. research has been focused on developing high-performance piezoelectric microfabricated acoustic/ultrasonic transducers as well as acoustic lenses with MEMS technology, and exploring novel applications of these ultrasonic/acoustic microelectromechanical systems. More specifically, in this thesis, I will present four types of piezoelectric acoustic or ultrasonic microelectromechanical systems that effectively generate acoustic waves for biomedical (such as cancer treatment and neuromodulation), manipulation (such as acoustic trapping), and actuation (such as droplet ejection and in-air propulsion) applications. The first two types are both ultrasonic transducers based on piezoelectric lead zirconate titanate (PZT) substrates vibrating in thickness mode with microfabricated Fresnel acoustic lenses on top, namely self-focusing acoustic transducers (SFAT) and single-element acoustic “tweezers”. The SFAT has been applied to in vivo selective cancer treatment, neurostimulation, and nozzleless droplet ejection (for transferring semiconductor chips, microspheres and cells). While two different types of acoustic tweezers have been developed to trap large (sub-millimeter- or millimeter-sized) particles. The other two types are both airborne acoustic/ultrasonic transducers based on membranes vibrating in flexural mode, with one based on a dome-shaped diaphragm with piezoelectric zinc oxide (ZnO) actuation elements to achieve higher sound output, and the other being a synthetic jet acoustic propeller based on a membrane-type mini piezoelectric speaker covered by a laser-micromachined plastic sheet having sub-millimeter-sized orifices. 9 1.5 Overview of the Thesis Chapters In chapter 1, an overview of the application of acoustic waves and ultrasound, as well as reviews on different types of acoustic transducers and acoustic lenses are described as an introduction to the background and motivation of this thesis. In chapter 2, the design, fabrication, and characterization methods for self-focusing acoustic transducers (SFATs) based on piezoelectric PZT substrates and microfabricated Fresnel acoustic lenses will be presented. From chapter 3 to chapter 5, applications of SFATs will be described. Chapter 3 demonstrates different types of SFATs used for nozzleless acoustic droplet ejection (ADE) with droplet diameters varying from 100 μm to 2.49 mm. Methods to realize electrical tuning of droplet size have been discussed. Acoustic droplet-assisted transfer and delivery of semiconductor chips, microspheres, and cells have been demonstrated. Chapter 4 describes an in vivo cancer treatment system based on an SFAT, a mechanical scanning platform, and a water-cooling system is developed. The SFAT generates pulsed high- frequency medium-intensity focused ultrasound which shows good therapeutic effects on the non- thermal treatment of subcutaneous tumors on mice. Chapter 5 describes the design of SFATs based on thin PZT substrates which ensure good visibility for patch-clamp experiments used to study the ultrasound-induced neuron activities on rat brain tissues. In addition, the preliminary results from the patch-clamp experiments are discussed. Chapter 6 presents the design, simulation, and demonstration of two types of single-element, microfabricated acoustic tweezers based on modified Fresnel air-cavity acoustic lenses, one is based on a ring-focusing Fresnel acoustic lens, and the other is based on a multi-foci linear Fresnel lens. 10 In chapter 7, the design and fabrication of a piezoelectric micromachined ultrasonic transducer based on a dome-shaped diaphragm as well as methods to improve its sound output will be covered. In chapter 8, an acoustic propeller based on a mini piezoelectric speaker with a plastic cover having tiny orifices is presented, along with discussions on the optimization of design parameters and demonstrations of using the device as an acoustic thruster that can move, jump, or rotate, as well as an acoustic lifter that can lift and rotate objects. Finally, chapter 9 presents conclusions and future research directions. 11 2Chapter 2 Design, Fabrication, and Characterization of Self-Focusing Acoustic Transducers (SFATs) 2.1 Introduction Through effectively focusing acoustic waves into a small volume, focused ultrasound (FUS) with higher intensity and better spatial resolution can be achieved, leading to the development of a wide range of applications including tumor ablation [85]–[87], neuromodulation [6], [88], [89], droplet ejection [28], [90], acoustic trapping [91], microfluidic mixing [92], and jet propulsion [93]. To effectively generate FUS, the acoustic waves generated by a transducer have to be focused, and curved surfaces on the transducer itself [94] or acoustic lens attached to it [69], [95] have been used for focusing. However, curvature error and surface roughness of these usually macro- machined or 3D-printed curved surfaces are hard to control. Also, these fabrication processes are time-consuming without massively parallel processing capability. Alternatively, focusing can be realized by controlling the time delay of the driving signal applied on each element of a phased array [96], which involves complicated driving electronics with many electrical connections, making the system bulky and complex. Another method to focus ultrasound is to use a thin, planar, microfabricated Fresnel acoustic lens. Single- [78] and multi-layer [79] acoustic Fresnel lenses have been microfabricated through etching and bonding to thin-film transducers, with challenges in precisely controlling the layer thickness. On the other hand, our fully-microfabricated self- focusing acoustic transducers (SFATs) [80], [81] based on annular-ring air-cavity Fresnel lenses are easy to be manufactured since the layer thickness control is not critical, and no bonding of the 12 lens is needed. The SFATs have a small form factor, and can be mass-produced into arrays. Moreover, with small modifications in design, their focusing parameters such as focal diameter [97] and focal length [98] could be electrically tuned. 2.2 Working Principle of SFAT 2.2.1 SFAT Based on an Air-Cavity Lens A typical SFAT consists of two parts, an ultrasonic sound source to generate ultrasound waves and an acoustic lens for focusing them (Figure 2.1a). The sound source is a bulk sheet of piezoelectric material sandwiched by its top and bottom circular electrodes overlapping with each other in the chip center. PZT such as PZT-5A and PZT-4 are usually chosen as the substrate material due to their high d33 coefficients (high vibration amplitude with same applied voltage), electromechanical coupling coefficients (good conversion efficiency between electrical and acoustic energy), and Curie temperatures (compatible to the baking procedures used in microfabrication processes) [99]. The electrodes are usually made of thin layers of nickel, copper or silver. For electrical connections, the top and bottom circular electrodes are extended into two non-overlapping rectangular soldering pads, where electrical wires are soldered (Figure 2.1b). When sinusoidal voltage signals with frequency equal to the PZT’s thickness-mode resonant frequency are applied onto the electrodes through the soldered wires, the PZT sandwiched between the circular regions of the electrodes vibrates in thickness direction, generating ultrasound waves of the same frequency. The generated acoustic waves are then focused through a microfabricated Fresnel acoustic lens on the top electrode. The lens is typically made of Parylene-sealed annular-ring air cavities alternating with non-air-cavity regions with Parylene uniformly coated on the electrode (Figure 13 2.1a and 2.1b). Alternatively, the Parylene could be replaced with other materials such as SU-8 photoresist and Polydimethylsiloxane (PDMS), as described in later sections in this chapter. The acoustic waves are almost completely blocked by the air cavities due to the large acoustic impedance difference between air (0.4 MRayl) and solid (over 1 MRayl) [1], but can propagate through the non-air-cavity areas. To focus ultrasound waves at a focal point at a distance F (focal length) above the center of the transducer’s top surface, the annular rings are designed into Fresnel half-wavelength bands (FHWB) [78] so that all the acoustic waves arrive at the focal point with a net phase difference less than 180  after passing through the lens. This is achieved by choosing boundary radii Rn so that the path-length from the focal point to any ring boundary is longer than F by integer multiples of the half-wavelength ( /2) in the surrounding medium (which is typically a liquid such as water, Figure 2.1a), as shown in the equation below: 𝑅 + 𝐹 − 𝐹 = 𝑛𝜆 2 ⁄ , 𝑛 = 0,1,2, ⋯, (2.1) from which we have: 𝑅 = 𝑛𝜆 × (𝐹 + 𝑛𝜆 4 ⁄ ), 𝑛 = 0,1,2, ⋯. (2.2) For focusing, the non-air-cavity regions are assigned for areas where 𝑅 < 𝑅 < 𝑅 , 𝑛 = 0,2,4, ⋯, which include the circle in the center (which is essentially a “ring” with zero inner diameter) and every other ring outwards, while air cavity rings are assigned for areas where 𝑅 < 𝑅 < 𝑅 , 𝑛 = 1,3,5, ⋯. Here, the assignment of air-cavity and non-air-cavity regions could be reversed without affecting the focusing effect. 14 Figure 2.1 (a) Cross-sectional (across the dashed line in (b)) diagram of a typical SFAT based on PZT substrate with a Fresnel air-cavity acoustic lens made of Parylene on top, showing how the Fresnel acoustic lens focuses ultrasound in water by blocking destructively interfering acoustic waves. (b) Top-view diagram of the same SFAT, showing the relative positions of the top electrode, air-cavity rings, soldering pad, and Parylene-coated regions. 2.2.2 SFAT Based on Patterned Electrode Rings Another way to realize Fresnel focusing of acoustic waves is to pattern the top and bottom electrodes covering a piezoelectric sheet (such as PZT) into overlapping FHWB ring patterns, which are connected by a narrow rectangular electrode (Figure 2.2). Ideally, thickness-mode vibration will only happen in the PZT regions covered by the Fresnel electrode rings, thus only in- phase acoustic waves could be generated and will result in FUS. To avoid sound generation from non-Fresnel-ring regions, the rectangular electrode and soldering pads from the top electrode are designed not to overlap with their counterparts on the bottom electrode. The most attractive feature of this type of SFAT is the easy and fast fabrication. However, it suffers from poor focusing performance resulted from fringing electrical fields in non-electrode regions and non-thickness vibration modes (especially when the ring width is comparable or smaller than the thickness of the piezoelectric sheet). Also, electrode-ring SFAT is much more sensitive to fabrication errors such as reduced electrode ring width from lateral undercut during wet etching and front-to-back alignment errors between the top and bottom electrode patterns. As (a) (b) R n = nλ × (F + nλ 4 ⁄ ) 15 a comparison, in SFATs based on Fresnel air-cavity lenses, the width of air cavities could be precisely defined with photolithography, and small front-to-back alignment errors can be tolerated by designing the bottom electrode to be larger than the top one, while having a wide outermost air- cavity ring at the top. As a result, with the same FHWB pattern and applied voltage, the peak acoustic pressure generated by an electrode-ring SFAT is about 2.6 times lower compared to the peak pressure generated by an SFAT with an air-cavity lens made of PDMS [100]. Moreover, the smaller electrode area on the electrode-ring SFATs means larger series resistance of the electrode, and thus more heat generation during device operation, limiting the electric power that can be applied. Due to the aforementioned reasons, electrode-ring SFATs are seldomly used unless they are more suited to a specific application. Figure 2.2 (a) Cross-sectional (across the dashed line in (b)) diagram and (b) top-view diagram of an SFAT based on a PZT sheet with its top and bottom electrodes patterned into overlapping Fresnel half-wavelength rings connected by narrow rectangular electrodes. The rectangular electrodes and soldering pads on the top and bottom electrodes don’t overlap. 2.3 Simulation of Acoustic Pressure (a) (b) 16 To verify the design, the distribution of acoustic pressure generated by an SFAT could be simulated with the finite-element method (FEM) [101] using the Pressure Acoustics module of COMSOL Multiphysics (COMSOL Inc.) in the frequency domain. For simplicity, only the volume above the piezoelectric sheet is considered (including the medium and the Fresnel lens). The vibration of the piezoelectric sheet is modelled with normal displacement boundary conditions defined on the top electrode. To save computation time and memory, two-dimensional (2D) axial symmetry is defined, where only a half of the volume cross- section is modelled (Figure 2.3). The full simulation results can be created through mirroring the simulated data along the central vertical axis (R = 0). For cases where the SFAT is not axial symmetric (for example, when only sections of rings are modelled), 3D simulation models could be used. For the FEM analysis, a free triangular mesh is used, and the maximum element size is typically set to be less than 1/10 of the wavelength in the medium, to ensure the accuracy of the simulation. For simplicity, all materials modelled in the simulations are assumed to have isotropic and homogeneous material properties. At outer boundaries of the simulation volume where acoustic waves could pass through without reflection, spherical wave radiation boundary conditions or perfectly matched layers could be used. To define the air-water boundary where the acoustic reflectance is close to unity, sound hard boundary conditions have been used. Figure 2.3 shows the typical acoustic pressure distributions of an SFAT working at 20.70 MHz, with six constructive Fresnel rings designed for 5 mm focal length in water. From the simulation results, we can see a strong focusing effect at the designed focal length of 5 mm with an ellipsoid focal zone. 17 Figure 2.3 FEM-simulated normalized acoustic pressure of an SFAT working at 20.70 MHz, with six constructive Fresnel rings designed for 5 mm focal length in water (a) on the central vertical plane and (b) on the lateral focal plane (at Z = 5 mm), with same color bar scale but different dimension scales. 2.4 Design Parameters of SFATs 2.4.1 Operating Frequency For efficient ultrasound generation, an SFAT could only operate near the fundamental or odd harmonic resonant frequencies of the piezoelectric sheet, which is related to its thickness. The relationship between the resonant frequencies and the thickness of the piezoelectric sheet is given by [1]: 𝑓 = ( 1 2 + 𝑛 )𝑐 𝑡 , 𝑛 = 0, 1, 2, 3 ⋯. (2.3) in which t is the thickness of the piezoelectric sheet and c is the sound velocity within the piezoelectric material. The frequency is usually chosen according to the target application. As mentioned in later sections, the operating frequency is also related to the peak pressure, focal size, and depth of focus of the SFAT. Acoustic attenuation within the transmission medium is also related to the operating frequency, and the acoustic attenuation coefficient is typically proportional to the frequency square [1]. (a) (b) Z (mm) X (mm) Y (mm) X (mm) 18 2.4.2 Peak Acoustic Pressure According to acoustic theory, the peak pressure at the focal point is proportional to the vibration displacement at the lens surface and the operating frequency [102]. When vibration amplitude and designed focal length are the same, the simulated peak pressure is proportional to the number of constructive Fresnel rings (Figure 2.4a). However, if the number of Fresnel rings are fixed, the simulated peak acoustic pressure almost doesn’t change as the designed focal length becomes longer, even when the active area is much larger (Figure 2.4b). Figure 2.4 (a) Simulated peak acoustic pressure versus the number of constructive Fresnel rings and operating frequency, when the vibration from lens surface is 1 nm, and the designed focal length is 8 mm. (b) The calculated active area of non-air-cavity regions and simulated peak pressure versus the designed focal length, when the operating frequency is 2.32 MHz, vibration amplitude from the lens is 1 nm, and the number of constructive Fresnel rings is 6. 2.4.3 Focal Diameter (Focal Size) On the focal plane where Z = F, the focal diameter (or focal size) is defined by the lateral off- center distance where the acoustic pressures drop to -3 dB of the peak value. The focal diameter of an SFAT can be approximated by the width of its outermost ring band (if its boundary radii are much larger than its width) [103] and becomes smaller if the designed operating frequency is (a) (b) Active Device Area (mm 2 ) Sim. Peak. Pressure (Pa) Number of Constructive Rings Sim. Peak. Pressure (Pa) Focal Length (mm) 19 higher, as explained below. For focal length F much longer than the wavelength  (which is true in our cases), equation (1) can be approximated as 𝑅 ≅ √𝑛𝜆𝐹 (2.4) For total N Fresnel ring boundaries, we get from (2.4): 𝑅 − 𝑅 ≅ 𝑁𝜆𝐹 − (𝑁 − 1)𝜆𝐹 = 𝜆𝐹 . (2.5) From algebra relationship, we also have: 𝑅 − 𝑅 = 𝑅 − (𝑅 − 𝛥 𝑅 ) = 2𝑅 𝛥 𝑅 − 𝛥 𝑅 , (2.6) where 𝛥𝑅 is the outermost ring width. If 𝑅 ≫ Δ𝑅 , which is usually true for N > 3, 𝛥𝑅 term in (2.6) could be ignored, and by comparing (2.5) and (2.6) we get 𝛥 𝑅 ≅ 𝜆𝐹 2𝑅 ⁄ ≅ 𝜆𝐹 2√𝑛𝜆𝐹 ⁄ = 𝜆 𝐹 4𝑁 ⁄ = 𝑐𝐹 4𝑁𝑓 ⁄ (2.7) where f and c are frequency and sound velocity in medium, respectively. This suggests that the focal diameter is positively related to the sound velocity in medium and the designed focal length, while being negatively related to the number of Fresnel rings and the frequency. For example, for SFATs with the six number of constructive Fresnel rings designed for the same focal length (5 mm) in water, the higher the frequency, the smaller the focal diameter (Figure 2.5). The focal diameters of these SFATs working at 6.90 MHz, 11.65 MHz, and 20.99 MHz are simulated to be 190.9, 144.2 and 102.4 μm, respectively. 20 Figure 2.5 FEM-simulated relative acoustic pressure distribution in the focal planes of different SFATs working at (a) 6.90 MHz, (b) 11.65 MHz, and (c) 20.99 MHz. The SFATs have the same number of rings designed for the same focal length. 2.4.4 Depth of Focus (Focal Depth) The depth of focus (DoF) or focal depth is defined by the axial distance between two positions where the acoustic pressures drop to -3 dB of the peak value. The DoF is usually longer than the focal diameter. From the simulation, we find a linear relationship between the DoF and the multiplication product of the outermost ring width ΔR (approximately equal to the focal diameter) and the F#, which is defined by the focal length divided by the diameter of the lens (Figure 2.6). (a) (b) Y (mm) X (mm) (c) f = 6.90 MHz f = 11.65 MHz f = 20.99 MHz 21 Figure 2.6 Simulated focal depth versus the product of the outermost ring width (ΔR) and the F#, along with linear fitting lines, for SFATs designed at 2.32 MHz, 6.60 MHz, and 11.00 MHz, respectively. 2.4.5 Device Size According to (2.2), the radius of the active area is related to the wavelength, the number of rings, and the designed focal length. 2.5 Microfabrication Methods for SFATs 2.5.1 Fabrication of SFAT with Patterned Electrode Rings An SFAT with patterned electrode could be fabricated according to steps shown in Figure 2.7. First, photoresist is spin-coated and patterned on the top and bottom electrodes of a PZT sheet through photolithography. The front-to-back alignment is achieved through aligning one corner of the PZT sheet to reference corners on the masks for top and bottom electrodes, respectively. If the thickness of the PZT is thin (e.g. only 127 μm thick), double-sided alignment based on IR imaging could be used to achieve better front-to-back alignment precisions. Then the electrodes are Simulated Focal Depth (μm) ΔR × F# (μm) 22 patterned through wet etching. After wires are soldered on the soldering pads, Parylene D is deposited to offer electrical encapsulation and acoustic matching. Figure 2.7 Fabrication process for an SFAT with patterned electrode rings: (a) pattern photoresists on both side of a nickel-coated PZT sheet through photolithography; (b) pattern electrodes through wet etching; (c) deposit Parylene for encapsulation. (d) Photo of a fabricated device before final Parylene deposition. 2.5.2 Fabrication of SFAT with Parylene Air-Cavity Lens An SFAT with Parylene Air-Cavity Lens could be fabricated according to the steps briefly described in Figure 2.8 [80]. Firstly, predeposited front and back nickel electrodes on a PZT sheet are patterned through photolithography and wet etching (Figure 2.8a). The front-to-back alignment is achieved by aligning one corner of the PZT sheet to a reference corner on the masks. Then AZ 5214 photoresist is spin-coated at 1,200 rpm to form a sacrificial layer for air cavities with a thickness of around 3.5 μm, and is patterned into Fresnel half-wavelength annular rings (Figure 2.8b). After that, 3.5 μm Parylene D is deposited (Figure 2.8c), followed by the patterning of “release holes” through O2 reactive ion etching (RIE) to expose the photoresist sacrificial layer (Figure 2.8d), which is then removed by soaking the substrate in acetone for two days (Figure 2.8e), as the acetone dissolves the photoresist sacrificial layer through the release holes. After cleaning with methanol, isopropyl alcohol (IPA), and DI water, followed by air drying, we deposit (d) 23 another thick (much thicker than the height of the air cavities) Parylene D to fill and seal the open holes (Figure 2.8f). Figure 2.8 Fabrication process for an SFAT with a Parylene air-cavity lens: (a) Pattern electrodes on both sides of PZT; (b) spin-coat and pattern photoresist as sacrificial layer for air cavities; (c) deposit Parylene; (d) pattern release holes; (e) remove photoresist with acetone, clean and air dry; (f) deposit Parylene to seal the air cavities. (g) Photo of a fabricated device before wires are soldered. 2.5.3 Fabrication of SFAT with PDMS Air-Cavity Lens The fabrication process of Parylene air-cavity lens is time-consuming with many steps including sacrificial layer release which can take several days. Also, the acoustic energy transmission through the lens is limited at low operating frequencies because it’s hard to deposit very thick Parylene. To address this issue, we have developed fast microfabrication methods for Fresnel air-cavity lenses made of different materials including PDMS and SU-8 photoresist, whose acoustic impedances are close to that of water, and can be made very thick to ensure good acoustic energy transmission. For the SFAT with PDMS air-cavity lens, first, the PDMS lens is made through soft lithography. To create the mold for casting the PDMS membrane, through photolithography, SU- 8 photoresist is patterned on a glass plate (Figure 2.9a). The SU-8 and glass plate is then treated (g) 24 with a hydrophobic silane (SIH5841, Gelest, Inc.) to prevent the casted PDMS from permanently sticking to the mold surface. To create the PDMS membrane, we first mix the base polymer and curing agent of PDMS (Sylgard 184) at a weight ratio of 10:1. After degassing in a vacuum chamber to remove any air bubbles, the PDMS mixture is poured onto the SU-8/glass mold. The thickness of the PDMS could be controlled through controlling the spin-speed during spin-coating. It can also be controlled with a custom-made precision clamping system through adjusting the vertical distance between the mold and another hydrophobically treated glass on top. The lens thickness is chosen through simulation to achieve the maximal focal pressure. Finally, the PDMS mixture is baked at 60 °C for 4 hours for curing. After curing, the PDMS membrane is carefully peeled off from the mold, the space defined with SU-8 will later become the air-cavities after bonding the membrane to the PZT sheet (Figure 2.9b). On the PZT sheet, first, the top and bottom electrodes are patterned through photolithography and wet etching (Nickel Etchant TFG, Transene) at 30 °C (Figure 2.9c). The front-to-back alignment is achieved through aligning one corner of the PZT sheet to reference corners on the masks for top and bottom electrodes, respectively. The PZT is then uniformly coated with 10-μm- thick Parylene-D (Special Coating Systems) through chemical vapor deposition (Figure 2.9d), during which the soldering areas on the electrodes are masked with low-tack low-residue dicing tape (Semiconductor Equipment Corporation) so they can be exposed after tape removal. This ensures good electrical encapsulation on both electrodes except the soldering areas, which will later be sealed through another Parylene deposition. To bond the PDMS membrane onto the Parylene-covered PZT, a thin layer of SU-8 or UV- curable adhesive (NOA 60, Norland Products) is spin-coated onto the PZT surface (Figure 2.9e). To improve adhesion, before spin-coating, both the PDMS membrane and the Parylene-coated 25 PZT are treated under 35W 265 mTorr O2 plasma for 1 min. Then the PDMS membrane is aligned under a stereo microscope to bond to the PZT substrate, using alignment marks created at the four corners of the PZT and PDMS, and the SU-8 or UV-adhesive is fully cured with UV exposure (Figure 2.9f). Compared to conventional O 2 plasma bonding and partially-curing bonding [104], this bonding technique offers easy, time-efficient bonding between multiple surfaces with high repeatability and without any heating or mechanical force on the PZT. Lastly, the wires are soldered onto the electrodes, and encapsulated with another deposition of Parylene (Figure 2.9g). Figure 2.9 Fabrication process of an SFAT with a PDMS air-cavity lens: on glass plate, (a) create silicon mold through photolithography of SU-8; (b) replicate PDMS membrane from the silicon mold; on PZT sheet, (c) pattern top/bottom electrodes, (d) deposit Parylene, (e) spin-coat SU-8 photoresist or UV-curable adhesive, (f) bond the PDMS (peeled off from the silicon mold) onto PZT after alignment, then cure the photoresist or UV-curable adhesive, (g) solder electrical wires (not shown), and deposit Parylene for electrical encapsulation. (h) Photo of a fabricated device before wire soldering and the final Parylene deposition. 2.5.4 Fabrication of SFAT with SU-8 Air-Cavity Lens (h) 26 The fabrication method for an SFAT with SU-8 air-cavity lens is demonstrated in Figure 2.10, which involves the bonding of two layers of SU-8 photoresist. On a PZT sheet, the top and bottom electrodes are patterned (Figure 2.10a), followed by a Parylene deposition (Figure 2.10b). Then the bottom layer of SU-8 is patterned into constructive Fresnel rings and the outermost support structures, which will later be bonded to a flat top layer of SU-8 (Figure 2.10c). To prepare the top layer, SU-8 is spin-coated on a hydrophobically-treated polyester (PET) film attached to a carrier glass wafer. After long soft baking (to prevent underfilling during the bonding process), the SU- 8-coated PET film is carefully peeled off from the carrier wafer (Figure 2.10d and 4.10e). To enhance adhesion, the bottom SU-8 layer is then treated with O 2 plasma for 1 min. After that, with two layers of SU-8 facing each other, both the PET film and the PZT sheet are put into a commercial laminator heated to 65 °C which will melt the un-crosslinked top SU-8 layer, bonding it to the bottom crosslinked SU-8 with minimal underfilling (Figure 2.10f). Next, the top SU-8 layer is patterned and crosslinked through photolithography (Figure 2.10g). Finally, the PET film is peeled off. Due to the PET’s hydrophobicity and thus poor adhesion to the SU-8, no SU-8 will adhere to the PET film (Figure 2.10h). 27 Figure 2.10 Fabrication process of an SFAT with a SU-8 air-cavity lens: on the PZT sheet, (a) pattern top and bottom electrodes, (b) deposit Parylene, (c) pattern the bottom SU-8 layer; (d) spin-coat SU-8 on a PET film supported by a carrier wafer, and soft bake for a prolonged period of time, (e) peel off the SU-8 coated PET film, (f) bond the top and bottom SU-8 layers with a laminator, (g) pattern the top SU-8 through photolithography, (h) peel off the PET film. (i) Photo of a fabricated SFAT before wire soldering. 2.5.5 Fabrication of SFAT with SU-8/PDMS Air-Cavity Lens A Fresnel air-cavity lens could also be made with SU-8 and PDMS. The first three steps (Figure 2.11a to 4.11c) are very similar to the ones involved in fabricating SU-8 air-cavity lens, in which the bottom layer of SU-8 is patterned. Then a piece of flat PDMS membrane without any patterns is aligned and attached to the SU-8 surface as the top layer of the air-cavity lens. The (i) 28 “bonding” between PDMS and SU-8 is not achieved with any type of epoxy, but with a Parylene deposition that seals them together permanently. To balance the pressure difference between inside the air-cavities and the deposition chamber during the low-pressure CVD of Parylene, tiny gaps are created on the SU-8 patterns, which will be sealed by the thick Parylene (Figure 2.11f). Figure 2.11 F Fabrication process of an SFAT with a SU-8 air-cavity lens: (a) pattern top and bottom electrodes of a PZT sheet, (b) deposit Parylene, (c) pattern the bottom SU-8 layer, (d) attach a piece of PDMS membrane as the top layer of the lens, (e) the PDMS and SU-8 are permanently encapsulated within a Parylene layer. (f) A photo of a fabricated device, showing the gaps on the SU-8 layer to allow pressure equalization during the final Parylene deposition. 2.5.6 Comparison of Different Fabrication Methods Table 2.1 shows the comparison of the different fabrication methods mentioned in previous sections, from which we can see that the fabrications for PDMS, SU-8, and SU-8/PDMS Fresnel air-cavity lenses are not only faster, but also ensures the same or better acoustic transmission through the lens. (f) 29 Table 2.1 Comparison of the Microfabrication Processes Involved in Difference Fabrication Methods of SFATs Patterned Electrodes Parylene Air- Cavity Lens PDMS Air- Cavity Lens SU-8 Air- Cavity Lens SU-8/PDMS Air-Cavity Lens Photolithography 2 4 (1 needs tight alignment) 2+0.5 (mold fabrication) 2 2 Parylene Depo. 1 2-3 1-2 1-2 1-2 RIE & Sacrificial Layer Release 0 1 (time- consuming) 0 0 0 PDMS Casting 0 0 0.25 (4 lenses per run) 0 0.25 (4 lenses per run) Bonding 0 0 1 1 1 Fabrication Time ~0.5-1.0 day ~2.0-4.0 days ~0.5-1.0 day ~0.5-1.5 day ~0.5-1.5 day Theoretical Max. Acoustic Transmittance 15.4%* (limited by Parylene thickness) 15.4% (limited by Parylene thickness) 15.7% 17-52% 17-67% 2.6 Characterization of SFAT 2.6.1 Measurement of Electrical Impedance The electrical impedance of the SFAT could be measured with an impedance analyzer (such as 4294A, Keysight Technologies, Inc.) or a network analyzer (such as 8753d, Hewlett Packard, Inc.). If measured with a network analyzer, the one-port s 11 parameter is first measured, then converted to the impedance using the equation below [105]: 𝑍 = 50 Ω × (1 + 𝑠 ) (1 − 𝑠 ) ⁄ (8) 30 2.6.2 Measurement of Acoustic Pressure The acoustic pressure produced by an SFAT could be measured by a hydrophone (Figure 2.12). During the measurement, an SFAT is placed in a beaker or a water tank and immersed in deionized (DI) water. The water surface should be much higher than the transducer to avoid reflection from the air-water interface. To drive the SFAT, a function generator (AFG-3252, Tektronix, Inc.) is used to produce pulsed sinusoidal voltage signals of the operating frequency, which are amplified by a power amplifier (75A250, Amplifier Research Corp.) and applied to the SFAT. The shortest pulse width possible at which the acoustic pressure is independent of the pulse width is chosen to avoid damage to the hydrophone from high acoustic energy. During the measurement, a commercial hydrophone (HGL-0085, Onda Corp.) fixed onto a five-axis movable stage (Prior Scientific Instruments Ltd.) is aligned to the SFAT, and is slowly scanned in the water to measure the acoustic pressure at different positions. An oscilloscope (MDO3014, Tektronix, Inc.) is used to simultaneously monitor the applied voltage after a 40-dB voltage attenuator (100- SA-MFN-40, Bird Technologies) and the signal from hydrophone after a 20-dB pre-amplifier (AH-2010, Onda Corp.). 31 Figure 2.12 Measurement setup for measuring acoustic pressure with hydrophone. 32 3Chapter 3 SFATs for Nozzleless Droplet Ejection and Droplet-Ejection- Assisted Object Delivery 3.1 Introduction Droplet ejectors have broad application in inkjet printing, drug/gene delivery, in situ DNA/protein synthesis, etc. Most commercial droplet ejectors are currently based on droplet- forming nozzles, which tend to get clogged with viscous liquid or particle-containing liquid, causing reliability problems and increasing maintenance costs. Some of the commercial ejectors use air bubbles formed through rapid heating (which is not acceptable to many biochemical solutions), while some others use membrane deflection to push liquid out of a nozzle to eject liquid droplets. No matter which driving mechanism is used, the ejection direction of a nozzle-based droplet ejector will always be limited at the perpendicular direction to the nozzle plane, unless some special arrangement(s) is incorporated near the nozzle. Thus, nozzle-less and heatless droplet ejections are highly desirable. Acoustic droplet ejection (ADE) was first reported in scientific literature in 1927 [106], where oil droplets were observed to be ejected from the liquid surface by continuous acoustic waves of high energy generated by a piezoelectric quartz resonator. In 1989, this phenomenon was systemically studied with both experiments and numerical simulations using tone bursts of focused acoustic energy [27]. The high energy from focused acoustic waves can overcome restraining forces from surface tension and gravity to eject liquid droplets at high speed without relying on physical contact or nozzles, which reduces maintenance cost and the risk of contact contamination. The technology also works with a wide range of liquids such as water [27], isopropyl alcohol (IPA) 33 and ethanol [79], acetone and photoresist [107], ink [108], and various reagents used in life sciences [109]–[112], and the ejection direction, as well as droplet volume, can be controlled with high accuracy and precision [26]. These attractive features lead to the development of simple and cost-efficient ADE tools that use ejected droplets to transfer liquids or solids for applications such as inkjet printing [7][108], bio-reagent transferring [109], [110], cell transferring [111], [112], DNA [90] and protein [113] synthesis, and crystallography [114]. For this droplet ejection, our fully-microfabricated self-focusing acoustic transducers (SFATs) with annular-ring air-cavity Fresnel lenses are easy to be manufactured since the layer thickness control is not critical, and no bonding of the lens is needed. With our SFAT-based droplet ejectors, the demonstrated ejection rate is up to 8 kHz [115], and the transducers can be formed into arrays for parallel ejection of multiple droplets with tunable ejection directions [90], [113]. Moreover, the size of droplets ejected by the SFATs can be tuned through electrically controlling the number of activated Fresnel rings [97] or designing transducers working at different frequencies [29], [115]. These advantages make SFAT-based droplet ejectors a useful tool for many applications. In this chapter, we will demonstrate the ejection of large (sub-mm-sized and mm-sized) droplets using SFATs, as well as the tuning of the droplet size. We also demonstrate the applications of semiconductor chip pick-and-place, and microsphere/cell delivery. 3.2 On-Demand, Nozzleless Ejection of Sub-Millimeter-Sized Liquid Droplets The previously demonstrated acoustic droplet ejectors were all for small droplet size (5 – 120 μm in diameter, which corresponds to 0.065 pL – 0.91 nL in volume) and short focal length (0.4 – 0.8 mm) [116]–[120], and are not suitable for applications that require large droplets for high 34 throughput applications. Thus, we would like to invent a new SFAT that is capable of ejecting sub- mm-sized or even mm-sized droplets. 3.2.1 Device Design and Fabrication The SFAT ejector is built on a 1-mm-thick PZT substrate of which the thickness mode fundamental frequency is 2.32 MHz, with air-cavity-based acoustic lens reflectors designed for 5 mm focal length. To verify the design, simulation has been done to show the distribution of vertical particle displacement on both the focal plane (i.e., the xy plane at Z = 5 mm) (Figure 3.1a) and the central vertical plane perpendicular to the transducer at (i.e., in the xz plane at Y = 0) (Figure 3.1b). From both plots in Figure 1.1, we can clearly see the significant focusing effect at the designed focal point with over 5 times larger particle displacement. The focal spot at the focal plane is circular with 396 μm in diameter. But the focal spot at the central vertical plane is elliptical with a focal depth of 1282 μm along the wave propagation direction (Z). In other words, the focal depth is inherently more than three times of the diameter of the focal size at the focal plane, which provides tolerance on the needed liquid height for ejection to happen, as ejection of large droplets cause some fluctuation of the liquid level. Figure 3.1 Simulation of the normalized vertical particle displacement at (a) the focal plane (xy plane at Z = 5 mm), and (b) the central vertical plane perpendicular to the transducer (xz plane at Y = 0). (a) (b) 35 The device is fabricated according to steps described in chapter 2, with the diameter of the largest annular ring being 18.8 mm (Figure 3.2a). After fabrication, an 8 × 8 cm 2 liquid reservoir (Figure 3.2b) made of laser-cut acrylic sheets is attached to the fabricated device after two wires are soldered on its top and bottom electrodes. The reservoir is designed large enough to maintain a stable water level during ejection because the relative water level change is small and ejected droplets will fall back and get collected by it. Two notches are made on two face-to-face reservoir walls and sealed by transparent tape for a brighter view during microscope observation of the ejection process (described in later sections). Figure 3.2 Photos of (a) the front side of a fabricated device with air-cavity-based acoustic reflector lens (light grey areas) on nickel electrode (dark grey areas) sealed with Parylene, and (b) packaged device with acrylic liquid reservoir. 3.2.2 Ejection Characterization The ejector immersed in DI water is tested with pulsed 2.32 MHz sinusoidal signals of 150 Vpp at a pulse repetition frequency (PRF) of 60 Hz. Triggered by a pulse generator, a function generator generates a train of sinusoidal pulses, which is then amplified by a power amplifier to drive the device. The liquid level is adjusted until the ejected droplets can fly to the highest level before falling down due to gravity, and turn out to be around 5 mm, as predicted by the simulation. (a) (b) 36 To capture the ejected droplets, four red light-emitting diode (LED) driven by pulsed signals with the same PRF serves as a light source for stroboscopic observation of the ejection process with a certain delay after device actuation (Figure 3.3), while a digital camera (STC- MCCM401U3V, Sentech Co., Ltd.) whose frame rate is set to be the same as the PRF is attached at the end of a long-range microscope (Zoom 6000, Navitar , Inc. or K2 Distamax, Infinity Photo- Optical Company) focused on the water surface where ejection happens for image recording. Figure 3.3 Measurement set-up for capturing photos of droplet ejection. When the water level is at the designed focal plane, efficient ejection of droplets is observed. At the voltage level of 150 Vpp, a minimum of 18.18 μs (40 cycles of 2.32 MHz sinusoidal wave) pulse width is needed for ejection to happen. Figure 3.4 shows the time procession of the droplet ejection obtained with this minimum pulse width. From the photos we can see that a single droplet breaks off from a cone-shaped water column of 600 μm height 1,000 μs after actuation, and then flies upwards. The flat water surface is recovered at around 4,500 μs which means the maximum ejection rate can be 222 Hz. The droplet diameter is measured to be 285 μm, and the volume is estimated to be 12.1 nL, which is 13.3 times larger than our previously reported value. At this 37 pulse width, the ejected droplet travels 7 cm upward, which suggests an initial velocity of 1.17 m/s if evaporation can be ignored. Figure 3.4 Photos of the ejected droplets taken with optical strobing at different time points after actuation with 18.18 µs pulse width. Smaller droplets in the background are the ones falling downward after being ejected and reaching the highest point. To see the influence of pulse width, we observed the ejection pattern 700 μs after actuation at different pulse widths as shown in Figure 3.5. From the first five photos, we can see that wider pulse width results in higher water column and faster droplet speed. However, if the pulse width is too wide, the water column turns into a strange shape with a lot of water mist being generated because of too much energy at the focal point, making the process less stable. If we further increase the pulse width, stable droplet ejections disappear, since the vibration at the water surface is too violent. The droplets fly as high as 30 cm before falling down, when the ejector is driven with 81.81 μs pulse width. The upward traveling distance indicates an initial velocity of 2.42 m/s. Also, within this small range, the droplet size is observed to be almost independent of the pulse width. 300 μs after actuation 700 μs after actuation 1,000 μs after actuation 1,200 μs after actuation 2,360 μs after actuation 4,500 μs after actuation 38 Figure 3.5 Photos showing ejection pattern 700 μs after actuation at different pulse widths. Smaller droplets in the background are the ones falling downward after being ejected and reaching the highest point. We took a further look at the time procession of droplet ejection at wider pulse widths. For 22.72 μs pulse width (50 cycles of 2.32 MHz sinusoidal wave) (Figure 3.6a), at 1000 μs we can see a second satellite droplet being generated due to higher energy at longer pulse width. However, as the major droplet goes up, the satellite droplet actually falls down because of its low initial speed. If we increase pulse width to 31.82 μs (70 cycles of 2.32 MHz sinusoidal wave), six satellite droplets will be generated with one droplet falling down eventually (Figure 3.6b). 18.18 μs pulse width 22.27 μs pulse width 45.45 μs pulse width 63.63 μs pulse width 81.81 μs pulse width 27.27 μs pulse width 39 Figure 3.6 Photos of the ejected droplets taken with optical strobing at different time points after actuation with (a) 22.72 µs and (b) 31.82 µs pulse width showing the generation of satellite droplets (marked with numbers greater than 1) at longer pulse widths. Smaller droplets in the background are the ones falling downward after being ejected and reaching the highest point. To confirm the reliability of the device, we operated it continuously for three minutes and monitored the ejection pattern every half minute. From Figure 3.7, it is clearly seen that the ejection is very stable throughout the whole process and droplet size remains the same. Water temperature (a) (b) 40 is also measured before and after the operation with an infrared thermometer and we only saw a temperature rise of 0.2 °C. Figure 3.7 Photos taken at different times while droplets are continuously being ejected, showing the stability of the ejection. The droplet ejector also worked well with different liquid media such as IPA (Isopropyl Alcohol) and hydrocarbon oil. Due to difference in sound velocity in media, the focal length was shifted, but stable ejection of single liquid droplets was still observed. For IPA (Figure 3.8a), with 2.32 MHz, 150 Vpp, and 60 Hz PRF (same as before), the minimum pulse width needed is slightly smaller than that for water (17.24 vs 18.18 μs), possibly due to its lower surface tension (23.0 vs 72.8 mN/m) and density (0.77 vs 1.00 g/cc). For oil (Figure 3.8b), the PRF had to be reduced to 30 Hz to ensure stable ejection, and the minimum pulse width needed is 107.76 μs. The diameter of IPA and oil droplets are 339 μm and 592 μm, respectively, both are larger than water droplets, due to different focusing behavior in media of different physical properties. 41 Figure 3.8 Photos of the ejected droplets taken with optical strobing at different time points after actuation with (a) IPA as medium, 17.24 μs pulse width; (b) hydrocarbon oil as medium, 107.76 μs pulse width. The right photo in (b) was taken at a higher vertical level than that of the left photo, because the droplet separates from oil column at a very late stage. Smaller droplets in the background are the ones falling downward after being ejected and reaching the highest point. 3.3 Varying Droplet Size through Electrical Tuning of Focal Diameter The ability to tune the diameter of the ejected droplets is highly desirable, because it broadens the range of applications and making the ejector more versatile. Since the diameter of the ejected droplets is proportional to the focal diameter of the generated focused ultrasound, tuning of the droplet size could be realized through changing the focal diameter. Electrical turning of focal diameter can be obtained with ultrasonic phased array transducer (or simply termed as phased array) by applying different phase delays on the array elements to change the focusing characteristics. However, phased arrays require complicated control circuits and multiple power amplifiers. For an inexpensive single-element-focusing ultrasonic transducer with electrical (a) (b) 42 tunability of the focal size, we developed a new design of SFAT which utilizes the combination of (1) Fresnel annual-ring air-cavity acoustic lens on the top and (2) annual-ring patterned electrodes on the bottom of a piezoelectric substrate, respectively. 3.3.1 Device Design and Tuning Principle The modified SFAT (Figure 3.9) is built on a 1-mm-thick PZT sheet, whose fundamental thickness-mode resonant frequency is 2.32 MHz. On the top side of the PZT, there is a Parylene Fresnel designed for 6 mm focal length (Figure 3.9b and 3.9d). On the PZT’s bottom side (Figure 3.9c and 3.9e), the nickel electrode is patterned into 6 annular rings overlapping with the first two, the third, the fourth, the fifth, the sixth, and the last two (of the 8 Fresnel rings) on the top, so that we may electrically select any combination of the 6 electrode rings to produce acoustic waves (on corresponding annular regions) that will pass through the air-cavity lens for focusing (Figure 3.10). The bottom electrode ring width are designed to be slightly wider than its corresponding top Fresnel ring width, so that small errors in top-bottom alignment during fabrication would not affect device operation. The corresponding relationship between top Fresnel rings and bottom electrode rings, as well as the radii of all Fresnel rings are shown in Table 3.1. The number of annular electrode rings is chosen to be less than that of annular air-cavity-lens rings, in order to ensure (1) enough focusing when the smallest number of the electrodes is chosen and (2) enough width on the outermost electrode (which is the narrowest among all the patterned electrodes) for wire connection. According to (2.7), as the number of Fresnel rings being actuated from the center increases, the width of the outermost ring decreases, and consequently, the focal size becomes smaller (due to better focusing effect), as shown in Figure 3.10. Although the output acoustic intensity will vary 43 when the number of actuated rings changes, this could be compensated by adjusting the voltage applied on the device. Figure 3.9 (a) Cross-sectional-view schematic of the transducer with electrically tunable focal size. (b) Top view of the transudcer, showing white annular-ring areas that represent air-cavities that block acoustic waves; (c) bottom view of the transducer, showing the bottom electrodes that are patterned into six annular rings, so that we can individually select corresponding Fresnel rings on the top side to actuate. Photos of (d) top side of the transducer, showing eight Fresnel rings (dark grey ring areas) separated by eight air-cavity rings (light grey ring areas) with filled release holes (a) (b) (c) Bottom electrode rings Top air-cavity rings (d) (e) 6 th 4 th 7 th & 8 th 1 st & 2 nd 5 th 3 th 44 (at 0°, 90°, 180°, 270° positions of each ring) on a circular Nickel electrode; and (e) bottom side of the transducer, showing six electrode rings with wires soldered. To demonstrate the effectiveness of SFAT-based focal size tuning, we simulated the relative acoustic pressure distribution in the central vertical plane perpendicular to the transducer (Figure 4.9a to 4.9c) and in the focal plane 6 mm from transducer (Figure 4.9d to 4.9f), when 2, 4, and 8 of Fresnel rings from the center are actuated. The focal diameters in these three cases are simulated to be 660, 468 and 368 μm, respectively, which are close to outermost ring width estimations (Table 3.1) while the focal length for all cases remain the same (6 mm). Figure 3.10 Cross-sectional-view schematic of the transducer, showing how the focal size changes with (a) 4 Fresnel rings and (b) 2 Fresnel rings actuated from center. With more Fresnel rings actuated from center, the focal size will be smaller. (a) (b) 45 Table 3.1 Corresponding Relationship between Top Fresnel Rings and bottom Electrode Rings, with Inner and Outer Radii and Widths of Each Fresnel rings Ring Type Ring Sequence Top Fresnel Rings 1 st 2 nd 3 rd 4 th 5 th 6 th 7 th 8 th Bottom Electrode Rings 1 st 2 nd 3 rd 4 th 5 th 6 th Boundary Radii of Top Fresnel Rings (Inner, Outer, Ring Width in mm) 1 st 2 nd 3 rd 4 th 0.000, 1.982, 1.982 2.839, 3.521, 0.682 4.116, 4.656, 0.540 5.160, 5.637, 0.477 5 th 6 th 7 th 8 th 6.094, 6.534, 0.449 6.961, 7.377, 0.416 7.783, 8.183, 0.400 8.575, 8.961, 0.386 Figure 3.11 Simulated relative acoustic pressure distribution in central vertical plane with (a) 2 Fresnel rings, (b) 4 Fresnel rings, and (c) 8 Fresnel rings actuated from center; and that in focal plane 6 mm from device with (d) 2 Fresnel rings, (e) 4 Fresnel rings, and (f) 8 Fresnel rings actuated from center, showing the focal size becoming smaller with larger number of actuated rings. 3.3.2 Measurement Results (a) (b) (c) (d) (e) (f) 46 To experimentally determine the focal size, first, after alignment, a vertical scan of acoustic pressure along the center line was done to find the focal length with a hydrophone (Onda HGL- 0085) fixed onto a motorized 3-axis stage, and then a lateral scan of acoustic pressure along a central lateral axis was done at the focal plane with the same setup. During measurement, the top electrode and the inactivated bottom electrodes were connected to the ground, while the actuated bottom electrodes were connected to the driving signal. The transducer was driven with 2.32 MHz pulsed sinusoidal signal with a pulse width of 6.03 μs and the voltage level was adjusted in each case to keep the maximal intensity level the same, as we varied the number of the actuated electrode rings from the center. From the result of the vertical scan (Fig. 7a), we have confirmed that the focal length is 6 mm. And by controlling the number of Fresnel rings being driven from the center, the beam profiles in the focal plane were varied (Fig. 7b), from which the -3dB focal diameter was calculated. The measurement and simulation (as well as outmost ring width estimation of the focal sizes) of focal diameter are in good agreement over a wide range (371 - 866 μm) of the focal size (Fig. 10). The deviation from theory might be due to fringing fields between adjacent electrode rings and non- thickness vibration modes (since electrode width is comparable to or less than the PZT thickness), which were not considered in simulation or calculation. 47 Figure 3.12 Measured normalized acoustic pressure: (a) along the central vertical axis with 8 rings being actuated, showing focal length of 6 mm and (b) along a central lateral axis at the focal plane with different numbers of rings actuated from center, showing varied focal sizes (diameters). The transducer has been tested as a droplet ejector capable of ejecting sub-mm-sized droplets, whose dimension could be electrically controlled. During the tests, the transducer placed in a beaker filled with water was driven with 2.32 MHz pulsed sinusoidal signals of 200 V pp (for driving 5, 6, and 8 Fresnel rings from center) or 250 V pp (for driving 2, 3, and 4 Fresnel rings from center), at a pulse repetition frequency (PRF) of 10 Hz. The photos are taken with red light-emitting diodes (LED) as a light source to stroboscopically observe the ejection process with a certain delay after device actuation. A camera whose frame rate was set to be the same as PRF (10 Hz) was attached at the end of a long-range microscope focused on the water surface where ejection happens. The camera was connected to a computer to capture the ejection process. We were able to observe ejection of single water droplet per pulse in all cases when we actuated two, three, four, five, six, and eight Fresnel rings from center (Fig. 3.13a to 3.13f), with droplet diameter ranging from 294 to 560 m, which corresponds to volumes from 13.3 nL to 92.0 nL. The droplet diameter follows the trend of the focal size when different number of rings are actuated (Figure 3.13g), and the diameter and volume of the largest droplets are about 2.0 and 7.6 (a) (b) 48 times larger than our previously reported values in [81]. During an operation of 5 min, no temperature rise was detected. Figure 3.13 Photos showing sub-mm-sized water droplets of different diameters ejected by our focal-size-tunable transducer, when (a) 2 rings, (b) 3 rings, (c) 4 rings, (d) 5 rings, (e) 6 rings, (f) 8 rings were actuated from the center. The arcs at the bottom of each photo are part of air-cavity rings on the top of the transducer and the red area is from LED illumination. (a) (b) (c) (d) (e) (f) 2 rings actuated 5 rings actuated 6 rings actuated 8 rings actuated 3 rings actuated 4 rings actuated (g) 49 Figure 3.14 Measured and simulated focal diameter, outermost ring width estimation, and diameter of ejected droplets when different number of rings from the center were actuated. 3.3.3 Summary We designed, simulated, fabricated (on a 1-mm-thick PZT), and characterized a single- element planar focusing ultrasonic transducer with electrically tunable focal size. As a modification of our previously demonstrated SFAT, this new design has Fresnel air-cavity rings (for focusing ultrasound) on the top and patterned annular-ring electrodes (for selecting the number of Fresnel rings being actuated from center) on the bottom, which allow tuning of the focal size. The transducer was able to electrically tune the focal size from 371 to 866 μm, while keeping the focal length at 6 mm. When tested as a droplet ejector, the transducer ejected droplets with diameter from 294 to 560 μm (13.3 to 92.0 nL in volume), with one droplet per pulse. 3.4 Acoustic Droplet Ejector with Pulse-Width-Modulated Droplet Size for Picking and Placing Semiconductor Chips 3.4.1 Background and Motivation 50 In semiconductor packaging such as the assembly of surface mount devices (SMDs) or micro- light-emitting diodes (micro-LEDs), individual chips need to be precisely picked and placed onto various substrates, and a sophisticated system equipped with robotic arms each carrying multiple nozzles with vacuum suction is commonly used for such tasks [121]. The system is usually expensive and bulky due to its complexity. With many moving parts, mechanical failures become a concern, and building a system with massively parallel processing capability is highly challenging. The nozzles in the system, handling tens of thousands parts per hour, suffer from wear and tear that can cause deformation and damage, which may lead to loss of vacuum (and thus holding power), shifts in picking or placing positions, or even part damage, increasing the assembly failure rate. In addition, the nozzle size limits the system’s ability to handle very small chips, and different nozzle size is needed in order to handle chips of different sizes. As a result, a cheaper and smaller system having no moving parts or nozzles, and with the capability for massively parallel processing and handling very small chips is highly desirable. To realize such a system, we come up with the idea of utilizing liquid droplets ejected by high-intensity acoustic waves to carry semiconductor chips to the target position. However, our previously developed SFAT-based droplet ejectors have relatively small droplet size (up to 560 μm [97]), which is not large enough to generate droplets for carrying most semiconductor chips. Such ejectors, especially if they are capable of ejecting droplets at an electrically-tunable angle, offer unique opportunities in mid-air merging of droplets for non- contact mixing and biochemical reactions. In this work, as a proof-of-concept demonstration, we present a simple and small system based on our newly designed SFAT-based droplet ejector, which achieves a focal size as large as 1 mm, and can eject droplets to carry silicon chips floating on liquid surface. We also demonstrate 51 a new and easy way to control the droplet size through changing the driving pulse width and voltage, which can vary the droplet diameter for transferring chips of various sizes. Moreover, we design a channel-embedded plastic cover to load silicon chips automatically onto the ejection site by the local fluid flow generated during droplet dejection. 3.4.2 Device Design In order to make focal diameter (which can be estimated by ΔR) large, according to (2.7), we design an SFAT with long focal length of 22 mm, low operating frequency of 1.16 MHz (the fundamental thickness-mode resonant frequency of 2-mm-thick PZT), and only five non-air-cavity Fresnel rings (N = 9), as shown in Figure 3.15. Figure 3.15 Top-view photo of the ejector on 2-mm-thick PZT substrate (brown), showing five air-cavity rings (light grey with holes, for blocking out-of-phase waves) alternating with five Parylene-covered electrode regions (dark grey circle and four rings, where in-phase waves could pass). The holes on air-cavity rings are release-holes sealed after releasing sacrificial layers to create the air cavities To keep silicon chips (having a density of 2.32 g/cm 3 ) afloat on the liquid medium, we choose sodium polytungstate (SPT, Geoliquids Inc.) solution as the liquid medium, with its density adjusted to 2.50 g/cm 3 by mixing SPT powder with DI water at a weight ratio of 3.386:1, resulting in a sound velocity of 1,372 m/s [122]. Compared to other types of heavy liquids such as 5 mm 52 halogenated hydrocarbons, SPT solution is non-toxic, easy to make (water-soluble), and chemically inert to common materials used on semiconductor chips. 3.4.3 Simulation of Acoustic Pressure and Body Force Acoustic fields produced by the SFAT have been simulated through finite-element-method (FEM) with COMSOL Multiphysics, using the settings summarized in Table 3.2. Without reflection from SPT solution’s top surface, acoustic simulations in the frequency domain at 1.16 MHz show that the focal length is at the targeted 22 mm, while focal depth and diameter are 5 mm (Figure 3.16b) and 1 mm (Figure 3.16c), respectively. The large focal depth is desirable, because it tolerates the liquid level change during ejection. These results are confirmed with hydrophone measurement of acoustic pressure. With an SFAT immersed in SPT solution driven with 1.16 MHz sinusoidal pulsed signals, a hydrophone (HGL-0085, Onda Corp.) is scanned along the central vertical axis (Figure 3.16d) and along the central lateral axis at the focal plane (Figure 3.16e) to measure pressure distribution, with liquid’s top surface well above the scan path to minimize reflection. The measured focal length, focal depth and focal size are 21.7, 5.03, and 1.0 mm, respectively, close to the simulated values. With 85 V pp applied, the peak pressure at the focal point is measured to be 1.29 MPa. In the actual case, acoustic reflection from the SPT’s top surface needs to be considered. By comparing models with different liquid heights in simulation, we find that when the liquid level is 20.04 mm above the transducer, focal zones with very high pressure appear near the SPT-air interface. Through normalizing the simulated pressure value with hydrophone measurement data (assuming the peak pressure is proportional to the applied voltage), the peak pressure near the SPT’s top surface is 8.7 MPa, when 400 V pp is applied on the transducer (Figure 3.16f). 53 Table 3.2 Key Simulation Settings * All boundary and area notations are defined in Figure 3.16a. Acoustic Simulation Fluid Dynamics Simulation Simulation Area Area ABPO Area CDNM Material Ignoring SPT-air interface reflection: SPT: Area ABPO Normal cases: Air: Area ABJG SPT: Area GJPO At t = 0, Air: Area CDIG SPT: Area GINM Physics Modules Pressure Acoustics, Frequency domain Laminar Flow/Level Set, Time domain Boundary Conditions Normal displacement (non-air-cavity rings): OR 1, R 2R 3, R 4R 5, R 6R 7, R 8R 9; Sound hard boundaries (air-cavity rings): R 1R 2, R 3R 4, R 5R 6, R 7R 8, R 9P; Perfectly matched boundaries (no reflection): ABJP Volume force: Area EFLK, value derived from acoustic simulation, duration equal to pulse width; Gravity: Area CDNM; Open boundary: CDNM; Level set initial interface: GHI Mesh Type/Size Free triangular, 25 μm maximum element size, with three rounds of adaptive mesh refinement Free triangular, 50 μm maximum element size, with three rounds of adaptive mesh refinement From Nyborg’s analysis of Navier-Stokes equation of fluid mechanics and the continuity equation using the method of successive approximations [123], the acoustic-field-induced steady body force 𝐹 ⃗ exerted on the SPT solution (which is considered as an incompressible Newtonian fluid) can be evaluated using the equation below: 𝐹 ⃗ = 𝜌 ⟨𝑣 ⃗ (𝛻 ⋅ 𝑣 ⃗) + (𝑣 ⃗ ⋅ 𝛻 )𝑣 ⃗⟩ (3.9) where the angle brackets denote time-averaging; 𝜌 is SPT density (2.50 kg/m 3 ); and 𝑣 ⃗ is sinusoidal particle velocity induced by acoustic field, calculated from simulated acoustic pressure [124]. From (4), the magnitude and direction of the body force near the SPT-air interface are 54 evaluated (Figure 3.16g), in which we clearly see strong force pointing from the center of the focal zone closest to SPT’s top surface into the air, and the peak force magnitude is as high as 3.38×10 7 N/m 3 . Figure 3.16 (a) Defined 2D simulation area with axisymmetry, with related simulation settings shown in Table I. Simulated relative acoustic pressure distribution ignoring reflection from the SPT-air interface: (b) over the central vertical plane, (c) over focal plane at Z = 22 mm. Hydrophone measurement of acoustic pressure in SPT solution (d) along the central vertical axis and (e) along the central lateral axis on the focal plane, with 85 V pp applied on the ejector. (f) Simulated acoustic pressure (color-bar unit: MPa) over central vertical plane with acoustic reflection from the SPT- air interface (20.04 mm above the device’s top surface), with 400 V pp applied to the transducer (pressure values normalized from measurement data in (d) and (e)). (g) Magnitude (color-bar unit: ×10 6 N/m 3 ) and direction (white arrows) of the acoustic-field-induced body force near the SPT-air interface, calculated from the pressure distribution in (f). 3.4.4 Simulation of Droplet Ejection Process We then use the calculated body force to simulate the droplet ejection process in time domain, using settings shown in Tabel. For simplicity, we only model the body force near the focal zone (Area EFLK in Figure 3.16), due to the dominantly higher force magnitude in this region. To model the pulsed driving signal, the force value is multiplied by a window function whose duration (a) (b) (c) (d) (e) (f) 55 is equal to the driving pulse width. In the simulation, fluid velocity 𝑢 ⃗ is calculated through solving the Navier-Stokes equation (3.2) and the continuity equation (3.3): 𝜌𝜕 𝑢 ⃗ 𝜕𝑡 ⁄ + 𝜌 𝑢 ⃗(𝛻 ⋅ 𝑢 ⃗) − 𝜂 ∇ 𝑢 ⃗ + ∇𝑝 = 𝐹 ⃗ + 𝐹 ⃗ + 𝜌𝑔 , (3.10) 𝛻 ⋅ 𝑢 ⃗ = 0, (3.11) where 𝜌 is mass density (1.204×10 -3 g/cm 3 for air [27] and 2.5 g/cm 3 for SPT); η is dynamic viscosity (1.825×10 -2 mPa∙s for air [125] and 10.2 mPa∙s for SPT [126]); p is pressure; g is standard gravity (9.8 m/s 2 ); and 𝐹 ⃗ is the body force induced by surface tension calculated from (3.7). The boundary profile change of the SPT-air two-phase system is numerically solved with a conservative level set method [127]. A field function 𝜙 (𝑥 ⃗, 𝑡 ) having a value between 0 and 1 at any spatial coordinate 𝑥 ⃗ is used to model the two-phase fluid system consisting of air and SPT. The regions where 𝜙 = 0 are considered as air, and the regions where 𝜙 = 1 are considered as SPT, while the transition regions with 𝜙 values in between are modelled with a smoothed Heaviside step function [127], with 𝜙 = 0.5 defined as the boundary between the two media. The governing equation calculating the time-evolution of the field function is shown below: 𝜕𝜙 (𝑥 ⃗, 𝑡 ) 𝜕𝑡 ⁄ + 𝑢 ⃗ ∙ 𝛻𝜙 = 𝛾𝛻 ∙ [𝜖 𝛻𝜙 − 𝜙 (1 − 𝜙 )(𝛻𝜙 |𝛻𝜙 | ⁄ )], (3.12) where reinitialization parameter 𝛾 and interface thickness-controlling parameter 𝜖 are two non- physical parameters related to the stability and accuracy of the numerical calculation, with the former set as the maximum fluid velocity during the ejection process, while the latter being the half of the maximum finite-element mesh size. The body force from surface tension 𝐹 ⃗ is calculated from the interface unit normal vector 𝑛 and the mean boundary curvature 𝜅 at the SPT- air boundary, which are functions of 𝜙 , as shown below [128]: 56 𝑛 = (𝛻𝜙 |𝛻𝜙 | ⁄ ) | . , (3.13) 𝜅 = (−𝛻 ⋅ 𝑛 ) | . , (3.14) 𝐹 ⃗ = 𝜎𝜅 𝛿 𝑛 , (3.15) where 𝜎 is the surface tension at the SPT-air interface (78 mN/m [126]), and 𝛿 is a smoothed Dirac delta function centered at the boundary where 𝜙 = 0.5 [128]. In each time step of the simulation, the flow velocity 𝑢 ⃗ is solved using (3.2) and (3.3), and then plugged into (3.4) to solve the field function 𝜙 . Once the field function is updated, the surface-tension-induced body force 𝐹 ⃗ is calculated using (10), and is fed back to (3.3) for further updating 𝑢 ⃗. This process repeats as time proceeds. Using the model mentioned above, we simulate different cases with the same 400 V pp applied on the transducer but with driving pulse width from 517 to 2,586 μs (Figure 3.17a to 3.17d). In all four cases, droplets are ejected by the focused ultrasound (with satellite droplets generated in the latter three cases), and the simulated diameter of the main droplets increases with the driving pulse width, ranging from 932.6 to 1690.5 μm (Figure 3.18a). Similarly, we keep the driving pulse width at 1,724 μs, while varying the driving voltage from 310 to 430 V pp. In this case, the droplet size also increases with applied voltage (ranging from 1327.3 to 1451.5 μm, as shown in Figure 3.17c, 3.17e to 3.17h), but the increase is small compared to the case with increased pulse widths (Figure 3.18a). 57 Figure 3.17 Simulation snapshots of ejected droplets with 400 V pp applied on the transducer with a driving pulse width of (a) 517 μs, (b) 862 μs, (c) 1,724 μs and (d) 2,586 μs. Simulation snapshots of ejected droplets with 1,724 μs driving pulse width and driving voltage of (e) 310 V pp, (f) 340 V pp, (g) 370 V pp, and (h) 430 V pp. Simulation snapshots of the water columns formed at the moment when the acoustic signal is turned off with driving voltages and pulse widths of (i) 400 V pp, 517 μs; (j) 400 V pp, 2,586 μs; (k) 310 V pp, 1,724 μs; and (l) 400 V pp, 1,724 μs, respectively. These results suggest a new and simple way of controlling the droplet size through tuning the driving conditions, especially the pulse width. To understand this phenomenon, we have simulated the volume (defined as VPW) of the bulging water column above the initial water surface (Z = 20.04 mm) at the moment when the acoustic signal (as well as the induced body force) is turned off after an activation period equal to the driving pulse width, with four examples shown in Figure 58 3.17i−3.17l. To compare VPW and the main droplet volume in each case, we calculate the equivalent diameter from VPW assuming that the same amount of volume forms a spherical droplet, and plot the results in Figure 3.18a. Interestingly, from the graph, we see that the equivalent diameter derived from VPW and the simulated main droplet diameter are very close. The water column will continue to rise after the acoustic signal is off, and the main droplet will later break up from its tip. It seems that only the initial volume pushed up during the acoustic drive contributes to the volume of the main droplet. To study how VPW (which is roughly equal to the main droplet volume) is affected by the driving conditions, we also simulate the maximum upward fluid speed (defined as uPW) along the central vertical axis (R = 0) at the same moment when VPW is calculated. By plotting the product of uPW and the driving pulse width versus VPW, we find a good linear relationship between these two plotted values (Figure 3.18b). These findings explain the dominant effect of the pulse width on the droplet size: while both increasing applied voltage and increasing pulse width will lead to higher uPW due to higher input energy, with longer pulse width, there will be more liquid injected into the initial water column during the acoustic drive, thus increasing VPW (and droplet size). These findings may lead to new insights on the droplet size control, and in-depth theoretical analysis regarding the physics behind them will be our future work. 59 Figure 3.18 (a) Simulated main droplet diameter and the equivalent diameter calculated from V PW (bulging water column volume when the acoustic signal is turned off) versus driving pulse width (with 400 V pp applied) and driving voltage (with 1,724 μs pulse width). (b) V PW (in μL) versus u PW (maximum fluid speed along the central vertical axis when the acoustic signal is turned off, in m/s) multiplied by pulse width (in ms) at different driving conditions (driving voltage in V pp and pulse width in ms), fitted by a linear trendline. With long pulse widths (and high driving voltages), the ejection process is less stable [129]. As a result, satellite droplets are generated, and the ejection direction is less repeatable. However, in our application, the former concern is not an issue as long as the main droplet can carry a semiconductor chip, and the latter problem can be minimized as discussed in the next section. 3.4.5 Characterization of Droplet Diameter During droplet ejection experiments, a 1.16 MHz pulsed sinusoidal voltage signal from a function generator (AFG3252, Tektronix, Inc.) is amplified by a power amplifier (75A250, Amplifier Research Corp.), and is delivered to the SFAT, which is placed at the bottom of a plastic container filled with SPT solution. The liquid level is adjusted until the ejected droplets can fly to the highest level before falling down due to gravity, and turn out to be around 20 mm, as predicted by the simulation in Figure 3.16f. With the driving condition of 394 V pp and 2 Hz pulse repetition frequency (PRF) while varying the driving pulse width from 517 to 2,586 μs, we take stroboscopic snapshots of the droplet ejection process (Figure 3.19a to 3.19d) with a long-range microscope (a) (b) 60 lens (Zoom 6000, Navitar Inc.) attached to a digital camera (STC-MCCM401U3V, Sentech Co., Ltd.) which is connected to a computer to save the photos. The background during droplet ejection is illuminated by a strobing LED light source flickering at the same 2 Hz frequency with some delay after the onset of the driving voltage pulse. The droplet ejections are stable and repeatable as photos (with frame height of 15.5 mm) taken at different times during the ejection look almost identical, and the main droplet size increases with longer pulse width, ranging from 950 to 1,604 μm, which are close to the simulation results. With much longer pulse width of 6,034 μs, the ejection process is recorded with a high-speed camera (DSC-RX100M6, Sony Corp.) at 960 frames per second, since the ejection is less stable so that no clear image may be captured with the strobing method, and the main droplet diameter is observed to be 2,490 μm (Figure 3.19e). Similarly, we vary the driving voltage from 283 to 368 Vpp, and change the pulse width in each case to measure the diameter of the ejected main droplets. In the case where the pulse width is 1,724 μs, the main droplet diameter varies from 1241.4 to 1370.2 μm, as the driving voltage changes from 268 to 394 Vpp (Figure 3.19f to 3.19i and 3.19c). All the measurement results (Figure 3.20) show that higher driving voltage and longer pulse width indeed increase the droplet size, with the pulse width having more effect on the droplet size than the voltage, especially at lower driving voltages, agreeing with simulation results in Figure 3.18a. 61 Figure 3.19 Photos of the ejected droplets with 394 V pp applied on the transducer with a driving pulse width of (a) 517 μs, (b) 862 μs, (c) 1,724 μs, (d) 2,586 μs and (e) 6,034 μs. Photos of the ejected droplets with 1,724 μs driving pulse width and driving voltage of (f) 283 V pp, (g) 312 V pp, (h) 339 V pp, and (i) 368 V pp. Scale bar length is 2 mm in all photos. All photos are taken with optical strobing, except that (e) is taken with a high-speed camera. Figure 3.20 Graph showing measured main droplet diameter with different driving voltages and driving pulse widths. 3.4.6 Ejection of Silicon Chips 62 We then build a setup for chip ejection based on the ejector (Figure 3.21a). To simulate semiconductor chips, 400-μm-thick silicon-nitride-covered silicon wafers are diced into square pieces with a dicing saw. To guide silicon chips floating on the liquid surface to the ejection site, a 500-μm-thick polyester cover is laser-machined to create a flow channel and an engraved circle to align the cover to the center Fresnel circle of the ejector (Figure 3.21b). The cover is held at the liquid surface (adjusted to around 20 mm), by a laser-machined acrylic holder fixed on a 5-axis precision stage, which allows fine-tuning of the liquid level held by surface tension between liquid and the cover as well as the alignment between the cover and the transducer. The surface tension helps also to maintain liquid height during ejection, in spite of some loss of liquid in the container from the ejection. When silicon chips are dumped onto the SPT solution (whose high density keeps them afloat) near the inlet of the cover, the chips are automatically loaded onto the ejection site, as we eject the chip one by one through operating the droplet ejector in two modes described in the next subsection. With 394 Vpp applied to the ejector, we successfully eject droplets of different sizes to carry square silicon chips having side lengths of 700, 1600 and 3100 μm (Figure 3.21c to 3.21e), using pulse widths of 1293, 3017, and 6034 μs, respectively. 63 Figure 3.21 (a) Cross-sectional diagram (across the center line along the channel on cover) of the ejection setup. (b) Top-view photo of a laser-machined channel-embedded plastic cover (designed for chips with 1,600 μm side length) held by an acrylic holder, aligned to the ejector at the container bottom, with silicon chips floating in the channel. Photos of ejected droplets of differernt sizes carrying 0.4-mm-thick silicon chips having side length of (c) 700 μm, (d) 1,600 μm (with a satellite droplet), and (e) 3,100 μm (with satellite droplets). 3.4.7 Automatic Loading of Silicon Chips The chips loaded into the inlet of the embedded flow channel on the cover can be drawn to the ejection site automatically through an operating mode of weak droplet ejection. When the device is driven with 1,724 μs pulse width, 10 Hz PRF and around 200 V pp voltage (about half of the typical ejection voltage), due to the long pulse width and the cover not being perfectly parallel to the ejector, droplets are ejected to the side of the cover from the ejection site (Figure 3.22a to 3.22c). The voltage is adjusted so that the droplet does not fly too high, but flies just out of the cover, so that the total liquid level remains constant. As a result, the temporarily reduced liquid (a) (b) (c) (d) (e) 2 mm 2 mm 2 mm 64 level in the local area in the flow channel draws the silicon chips to the end of the channel, which is the ejection site (Figure 3.22d), with the silicon chips lined up along the relatively narrow channel. When a silicon chip is in place, the loading stops automatically since the chip blocks the weak ejection. Then the driving condition is changed to the one for the regular ejection (the only manual task at this point, which can be automated), and we can keep ejecting the chips lined up in the channel, until loading is needed. This process of loading and ejection can be achieved with a computer interfaced to a function generator with pre-stored conditions. Figure 3.22 Photos showing the semi-automatic chip loading mechanism: (a) chips are dumped into the inlet of the flow channel, also showing the ejection trajectory of liquid droplets under a weak ejection; (b) chips moving in along the channel as an ejected liquid droplet flies in air; (c) chips moving in further; (d) front chip loaded at the ejection site after seven ejections. Photos of 400-μm-thick square silicon chips with side length of (e) 700 μm and (f) 1,600 μm ejected into 4×3 arrays with an interval of 5 mm collected on filter paper, with red crosses showing the centers of ejected chips in five other repeated trials with the center of the left top chip aligned together. (a) (b) (c) (d) (e) (f) 65 3.4.8 Assembly of Silicon Chips As a demonstration, a piece of filter paper is held above the liquid surface with a second movable stage which is manually moved laterally at an interval of 5 mm after each ejection to collect the ejected chips. Once hitting the filter paper, the chip is held onto the paper by the surface tension of the SPT and is “glued” in place after water evaporates from the liquid. With a relatively large chip, the ejection is less repeatable. To ensure good positioning precision and alignment, the distance between the liquid surface and filter paper is kept short so that the chip positioning error caused by variation in ejection direction is minimized. However, if the distance is too short, sometimes the ejected chip (flying up at high speed) is reflected back after hitting the paper. Thus, we find 4 mm distance to be a good compromise. In addition, since shorter driving pulse width improves ejection stability and thus repeatability, pulse widths are kept as low as possible, to 1,293 and 2,069 μs for the 700 and 1,600 μm chips, respectively. Moreover, to reduce chip rotation after ejection and to ensure repeatable ejection direction, good alignment between the chip center and the ejection site is crucial. For this, the opening width of the guiding channel near the ejection site is designed to be close to the side length of the chip and the position of the channel-embedded cover is carefully adjusted. Before each ejection, we also make sure that there is always a second chip right behind the loaded chip in the channel to prevent the loaded chip from going back. With these measures, we successfully eject silicon chips with side length of 700 μm (Figure 3.22e) and 1,600 μm (Figure 3.22f) into 4×3 arrays with the same interval. Five other trial sets indicate that the repeatability is good. 3.4.9 Summary 66 This section describes a micromachined acoustic droplet ejector based on a focusing ultrasonic transducer with Fresnel air-cavity lens. With small footprint and two-wire electrical interface, the transducer is capable of generating high-intensity focused ultrasound with 1-mm focal size, which can eject large liquid droplets with diameter from 850 to 2,490 μm, controlled by the driving pulse width and voltage, with the former having more tuning effect than the latter. An FEM simulation model calculating the time evolution of the acoustic-field-induced liquid motion during droplet ejection is developed, and is confirmed with experiments. With the transducer, a proof-of-concept semiconductor chip pick-and-place system has successfully been demonstrated to eject 400-μm-thick square silicon chips with side lengths ranging from 700 to 3,100 μm, carried by SPT droplets of different sizes. As ejected droplets generate lateral liquid flow towards the ejection site, the chips are automatically loaded through a microchannel-embedded plastic cover, and can be ejected one after another into arrays with good repeatability. Our experiments demonstrate that a droplet-ejector-based system that is much smaller and cheaper than the conventional approach with robotic arms can be a new possible tool for on demand semiconductor chip pick and placement. With easily adjustable focal size [14] and tunable droplet diameter, ejector-based pick-and-place systems are able to handle very small chips that robotic arms cannot handle reliably. In addition, since the transducer can be massively microfabricated, a similar system can potentially have massively parallel processing capability with an array of ejectors. 67 3.5 Extraction and Delivery of Microparticles and Cells Based on Droplet Ejection 3.5.1 Introduction There is an unmet need to extract cell(s) from mono-layer cells cultured on a solid surface for regenerative medicine. When using a pipette, scoop, or knife, it is difficult to control the number of cells extracted, due to large tool size and poor precision and repeatability of the manual operation. Micromanipulation offers better precision and is suitable for rare cells, but with low throughput [130]. Laser capture microdissection (LCM) has higher throughput but is still time- consuming [131], and requires a complicated and expensive system. Moreover, all the methods mentioned above may cause unwanted damage on the extracted cells and on the extraction edges of remaining cells, resulting in loss of rare cells, scars on the tissue grown out of the cells, or contamination from accidentally damaged neighboring cells. A focused ultrasound (FUS) offers a solution to this need, as it can produce a large, yet undamaging, focused extraction force which can eject cells contained in liquid droplets with minimal impact on the cells. Ultrasound propagates through different types of liquids and solids (without much reflection at the interfaces of materials with similar acoustic impedances), and the FUS transducer does not have to be in physical contact with the substrate where cells are grown. The number of cells that are ejected by a FUS transducer depends on the focal size of the FUS, which can be very small (as small as the size of a single cell) and is very precise and repeatable. In this document, we first present our proof-of-concept demonstration of FUS-based ejection of particles (to simulate cells) from a solid surface with the FUS transducer not in direct contact with the particle-containing solid substrate. Specifically, we use our self-focusing acoustic transducers (SFATs) based on Fresnel air-cavity lens, and show that different amounts of 68 microspheres can be ejected out of the surface of a Petri dish filled with agarose gel through varying the focal size of SFAT. We then demonstrate that cells grown on a Petri dish can be ejected from a monolayer of cells, without damaging surrounding cells. For these experiments, we have designed SFATs to operate at different frequencies, and used multiple SFATs with different focal sizes. However, with a special design, the focal size of a single SFAT can be electrically tuned, as shown in an early section. 3.5.2 Device Design According (2.7), the focal size of SFAT can be approximated by the width of its outermost ring band (if its boundary radii are much larger than its width), and becomes smaller if the designed operating frequency is higher, which is verified in finite element method (FEM) simulations (Figure 3.23). Thus, we designed SFATs with 6 constructive rings working at the 3 rd (6.60 MHz), 5 th (11.00 MHz), and 9 th (20.96 MHz) harmonic thickness-mode resonant frequencies on 1-mm- thick PZT substrates with 5 mm focal length (Figure 3.24). The actual measured resonant frequencies are 6.90, 11.65, and 20.99 MHz, respectively, which result in focal lengths that are slightly different from the designed values, but have negligible impact on the focusing efficiency. 69 Figure 3.23 FEM-simulated relative acoustic pressure distribution: on the central vertical planes for SFATs (designed for three different harmonics at 6.60 MHz, 11.00 MHz, and 20.96 MHz) working at (a) 6.90 MHz, (b) 11.65 MHz, and (c) 20.99 MHz, respectively; and in the focal planes (dashed lines) for the same SFATs working at (d) 6.90 MHz, (e) 11.65 MHz, and (f) 20.99 MHz. Figure 3.24 Photos of fabricated devices on PZT substrates working at (a) 6.90 MHz, (b) 11.65 MHz, (c) 20.99 MHz, showing the air-cavities (shiny areas), and the same devices working at (d) 6.90 MHz, (e) 11.65 MHz, (f) 20.99 MHz 70 under a digital microscope, showing air cavities (light grey areas), Parylene covered electrode (dark grey areas), and sealed release holes. 3.5.3 Focusing through Agarose-Gel-Filled Petri Dish When a Petri dish (made of Polystyrene with its bottom plate being 0.75 mm thick) containing agarose gel is immersed in water between SFAT and the water’s top surface (Fig. 7b), the acoustic waves produced by the SFAT propagate through the water, Petri dish’s bottom substrate, and agarose gel, interfering with each other. The waves constructively interfere at the focal point with a slightly larger focal size and slightly attenuated peak pressure at a slightly closer focal point (Figure 3.25 and Table 3.3), compared to the case of having no Petri dish. This is due to the fact that acoustic impedances of Petri dish (2.49 MRayl, attenuation coefficient 0.285 dB/cm-MHz [132]) and agarose gel (1.58 MRayl, attenuation coefficient 0.07 dB/cm-MHz [133]) are close to the water’s acoustic impedance (1.48 MRayl), so that there is little reflection at the interfaces of different media. 71 Figure 3.25 FEM-simulated relative acoustic pressure distribution when the bottom of a Petri dish (with 0.75-mm- thick bottom plate, red line) filled with 0.98-mm-thick 1% agarose gel (yellow lines) is 1.5 mm above SFAT surface in water: on the central vertical plane for SFATs working at (a) 6.90 MHz, (b) 11.65 MHz, and (c) 20.99 MHz, respectively; and in the focal planes (dashed lines) for the same devices working at (d) 6.90 MHz, (e) 11.65 MHz, and (f) 20.99 MHz. The color ranges are adjusted to be the same as those in Figure 3.23. Table 3.3 Simulation Results Focusing Parameters Working Frequency (MHz) 6.90 11.65 20.99 Focal Length (mm) In Water 5.34 5.39 5.08 Through Dish & Gel 4.59 4.69 4.38 Focal Size (μm) In Water 190.9 144.2 102.4 Through Dish & Gel 198.6 149.0 103.4 Normalized Peak Pressure In water 100% 100% 100% Through Dish & Gel 73.8% 88.9% 88.8% The effects of 1% (w/v) agarose gel thickness on the peak acoustic pressure have been simulated (Figure 3.26a), from which we see that the pressure changes periodically as the gel thickness increases and reaches local maxima at each incremental increase of the gel thickness 72 every half wavelength. A thickness of 0.98 mm was thus chosen so that almost optimal peak pressure could be achieved for all three frequencies. The gel thickness is realized by pouring 7.2 mL melted agarose gel solution into a 90-mm-diameter Petri dish, based on our experiments shown in Figure 3.26b. Figure 3.26 FEM-simulated normalized peak pressure versus 1% (w/v) agarose gel thickness. The navy dashed vertical line shows the selected thickness of 0.98 mm. In this simulation, the top of the gel is fixed at 4.5 mm above SFAT surface. (b) Experimental results on the relationship between gel volume before solidifying and the resulted gel thickness after solidifying at the center of a Petri dish with 90 mm diameter. 3.5.4 Characterization of Acoustic Pressure and Droplet Ejection To measure the peak acoustic pressure at the focal point, a commercial hydrophone (Onda HGL-0085) fixed onto a manual 3-axis stage is scanned around along the central vertical axis to find the focal point, while the transducer is driven with pulsed sinusoidal signal with 50 V pp at the resonant frequencies of each device. Then the same experiments are repeated with a Petri dish filled with 0.98-mm-thick agarose gel with the bottom of the Petri dish about 1.5 mm above the surfaces of SFATs. From the measurements (Figure 3.27), we see that the agarose-gel-filled Petri dish attenuates the acoustic pressure by 31.9%, 18.4%, and 3.6% for the 3 rd , 5 th , and 9 th harmonic (a) (b) 73 SFATs, respectively, and the measured focal lengths are close to the simulated values (Figure 3.25 and Table 3.3). Figure 3.27 Measured peak acoustic pressure at the focal point and focal length from different SFATs with and without the agarose-gel-filled Petri dish. The SFATs are then tested to eject water droplets in water with and without a Petri dish filled with agarose gel. During the tests, each transducer placed in a beaker filled with water is driven by pulsed sinusoidal signals of 200 Vpp (for the 6.90 MHz and 11.65 MHz transducers) or 400 Vpp (for the 20.99 MHz transducer) at the resonant frequency, at a pulse repetition frequency (PRF) of 20 Hz. The pulse width is selected so that only one droplet (without satellite droplets) may be ejected per ejection (Table 3.4). To capture the ejected droplets, a red light-emitting diode (LED) driven by pulsed signals with the same PRF serves as a light source for stroboscopic observation of the ejection process with a certain delay after device actuation (Figure 3.28a), while a camera whose frame rate is set to be the same as the PRF is attached at the end of a long-range microscope focused on the water surface where ejection happens for image recording. In experiments without a Petri dish, the water surface is adjusted to the focal plane. In experiments with Petri dishes, the 74 dish is held by a 5-axis precision manual stage, with a thin layer of water above the gel. Then the position of the Petri dish is adjusted until ejection can happen (Figure 3.28b). Figure 3.28 Measurement set-up for capturing photos of droplet ejection with optical strobing; (b) cross-sectional diagram showing the ejection set-up with a Petri dish filled with agarose gel. We have successfully observed the ejection of a single water droplet per pulse without (Figure 3.29a to 3.29c) and with the Petri dish (Figure 3.29d to 3.29f). The diameter of ejected droplets and minimum pulse width needed for ejection to happen are summarized in Table 3.4, from which we see that the droplet diameter (related to focal size) and the needed minimum pulse width are larger and longer with the Petri dish (than without it) due to some acoustic loss in the Petri dish and agarose gel. (a) (b) 75 Figure 3.29 Photos showing water droplets ejected without the Petri dish by our SFATs working at (a) 6.90 MHz, (b) 11.65 MHz, and (b) 20.99 MHz and that with the Petri dish for SFATs working at (d) 6.90 MHz, (e) 11.65 MHz, and (f) 20.99 MHz. The background circles are ripples on the water surface. Table 3.4 Droplet Diameters and Driving Conditions Ejection Parameters Working Frequency (MHz) 6.90 11.65 20.99 Droplet Diameter (μm) Without Dish 300 120 95 Through Dish & Gel 340 190 105 Minimal Pulse Width Needed (μs) Without Dish 37.7 16.3 15.3 Through Dish & Gel 94.2 60.1 33.4 Driving Voltage (Vpp) 200 200 400 3.5.5 Droplet-Assisted Particle Ejection 76 We choose 10-μm-diameter Polystyrene microspheres to simulate grown cells. To embed the microspheres onto agarose gel through self-assembly, we pour a thin layer of water on top of the gel, and fully suspend microspheres in methanol with sonication. Then the methanol with the microspheres is poured into the water layer, and the microspheres form a uniform layer at water/methanol boundary, most of where a monolayer of microspheres is formed (Figure 3.31a). When the solution is almost dried, we gently press the microspheres against the gel with a spatula to increase the microspheres’ adhesion on the gel. Then we use the set-up shown in Figure 3.30a, and collect the ejected droplets (Figure 3.30b) with a coverslip held above the water surface (Figure 3.31b to 3.31d). All the SFAT-ejected droplets contain microspheres, and the diameter of collected microsphere agglomerates on coverslips (determines the number of microspheres per droplet) is related to droplet size (Figure 3.31e). Assuming the collected microspheres are in monolayer with a filling factor of 0.9069, with 6.90 MHz, 11.65 MHz and 20.99 MHz SFATs, the estimated numbers of microspheres per ejected droplet are: 746, 498, and 167, respectively. More and fewer number of microspheres per droplet could be easily achieved by designing transducers at lower and higher frequencies, respectively. During 10 minutes of operation (2 droplets per second), no temperature rise or visible gel damage is observed. 77 Figure 3.30 (a) Cross-sectional diagram showing the droplet-assisted particle ejection set-up. (b) Photo of an ejected droplet carrying fluorescent microspheres under black light, ejected from the Petri dish by the 6.90 MHz SFAT and flies above the beaker edge. Figure 3.31 Microscope photos of (a) microsphere monolayer on the gel surface; collected microsphere agglomerates on plastic cover slips, ejected by SFATs working at (b) 6.90 MHz, (c) 11.65 MHz, (d) 20.99 MHz, respectively. (e) Diameters of collected microsphere agglomerate on cover slips, ejected droplets without Petri dish and ejected droplets with Petri dish, versus SFAT operating frequency. (a) (b) (a) (b) (c) (d) (e) 78 3.5.6 Droplet-Assisted Cell Extraction and Ejection We have tested the ejection of human retinal pigment epithelium (RPE) cells using an SFAT built on a PZT-4 substrate (Figure 3.32a). The resonant frequency is measured to be 20.12 MHz, and the focal length is simulated to be 4.86 mm (Figure 3.32b). The experiment set-up (Figure 3.32c) is similar to that for the particle ejection (Figure 3.30a), except that (1) the monolayer of cells is cultured directly on the inner bottom of a Petri dish without any agarose gel (Figure 3.32d) and (2) the cells are immersed in a shallow layer (about a few hundreds of micrometers above the cells) of phosphate-buffered saline (PBS) solution to keep the cells alive, while also creating a liquid-air interface close enough to the cells for droplet ejection. We circle each intended ejection spot with a permanent marker at the outer bottom of the Petri dish, which allows us to visually align the transducer center to the ejection spot. The vertical distance between the SFAT and the Petri dish is first adjusted to about 4.8 mm, and then slowly increased and decreased through scanning the Petri dish up and down around the initial position, while the SFAT is driven with 20.12 MHz pulsed sinusoidal drive of 300 V pp, with 248 μs pulse width at 50 Hz PRF, to produce visible droplet ejection. We have successfully observed ejection of cells (Figure 3.32e) from the cell monolayer with the ejection spot diameter being about 100 μm, close to the simulated focal diameter of the SFAT, without any visible damage to the cells surrounding the ejection spot. After four days of re-culturing, the new cells grown out of the remaining cells fill in the empty spot left by the previous ejection (Figure 3.32f), without any scar or damage. 79 Figure 3.32 (a) Top-view photo of the 20.12 MHz SFAT used for cell ejection. (b) Simulated relative acoustic pressure distribution on the central vertical plane above an SFAT working at 20.12 MHz with a Fresnel lens designed for 20.96 MHz and 5-mm focal length. (b) Cross-sectional diagram showing the droplet-assisted cell ejection set-up. (c) Microscope photos of 100% confluency human retinal pigment epithelium (RPE) mono-layer cells (d) before and (e) after an ejection of cells by SFAT. (f) Photo of the same mono-layer cells when the cells are re-cultured (for 4 days) after the cell ejection. 3.5.7 Summary We have designed and fabricated SFATs (working at three different harmonic resonant frequencies having different focal sizes) that can produce high-intensity focused ultrasound through and from Petri dish filled with agarose gel, and can eject droplets of different sizes to carry different numbers of particles embedded on the gel surface. The transducers could be useful for non-contact, precise extraction of cells without causing damage to the ejected cells, surrounding cells, or the culture medium. Single-cell extraction capability can be achieved through operating the transducer at a high frequency for a small focal (a) (b) (c) (d) (e) (f) 1 mm Before Ejection After Ejection After 4-day recovery 80 size, and high throughput cell extraction through an array of the transducers, which can be parallelly-microfabricated in the same batch. 81 4Chapter 4 SFAT for In Vivo Non-Thermal, Selective Cancer Treatment with High-Frequency Medium-Intensity Focused Ultrasound 4.1 Introduction Cancer, as one of the leading causes of death, claims almost 10 million lives each year [134]. Surgery to remove visible tumors is a straightforward and commonly used cancer treatment technique [135]. However, in some cases, tumors are not amenable to surgery because of medical comorbidities or critical structures nearby preventing an adequate resection. Also, malignant cells can spread outside of the tissue of origin, simple surgery may not be sufficient treatment for many cancers, thus has to be combined with radiation therapy and chemotherapy. However, in some cases, radiation therapy and chemotherapy are not very effective and could lead to serious side effects by affecting normal cells in the body [136], [137]. Focused ultrasound (FUS), as an emerging tool for cancer treatment, has received great attention in both medical research and clinical trials. With the energy of ultrasound focused onto a small volume of tissue that can be deep inside the body, the treatment efficacy and precision is greatly enhanced with less side effects. For example, high-intensity focused ultrasound (HIFU) has demonstrated good therapeutic effects in the treatment of tumors in the prostate [136], pancreas [85], breast [138], and brain [87]. In most of these applications, low-frequency (< 4 MHz) focused ultrasound of high intensity (with spatial-peak-pulse-average intensity I SPPA usually ranging from 1,000 to 10,000 W/cm 2 ) induces rapid heating in tissue [139], raising its temperature above 60 °C in seconds, to cause irreversible cell damage (coagulative necrosis) [140]. Apart from direct heating, HIFU, especially with high pressure (> 10 MPa), short pulse width (< 20 μs) at a low duty 82 cycle (< 1%) [141], has been shown to induce inertial acoustic cavitation, in which submicron/micron-sized gas bubbles form from cavitation nuclei and collapse rapidly after growth, causing destructive mechanical damage from shock waves or high-speed microjets [142]. Although both heat and cavitation effectively destroy tumor cells, nearby normal tissue in the target area may also be affected by the non-specific damage during the treatment, causing unwanted side effects [143]. As a result, most clinical treatment with HIFU must operate under the guidance of external imaging methods [85] such as magnetic resonance imaging (MRI) [3], [6]–[7] or ultrasound imaging [141], greatly increasing the cost and complexity of the procedure. Even with imaging guidance, dealing with invasive cancer cells near critical structures such as neural circuits and blood vessels in brain [87], [144], is still highly challenging. Thus, an ultrasound treatment that could selectively destroy cancer cells without damaging benign cells would be highly desirable. For this purpose, external agents have been used to increase cancer cells’ sensitivity to ultrasound treatment. For example, when monolayer co-cultures of both normal and cancer cells are exposed to 4 MHz FUS having a power of 8 W and a duration of 3 s, the addition of gold or magnetic nanoparticles greatly enhances ultrasound-induced killing of cancer cells while largely sparing their normal counterparts [145]. It has also been shown in in vivo experiments with a murine model that, when combined with pulsed ultrasound (3 MPa, 0.94 MHz, with 10 Hz pulse repetition frequency and 0.19% duty cycle), intravenously injected microbubbles selectively and persistently reduced tumor perfusion by depleting the neovasculature, resulting in a substantial and permanent cessation of tumor blood flow without a significant effect on normal tissue [146]. However, as the mechanisms and potential risks remain unclear, the addition of these agents may bring other undesirable complications to the treatment. 83 Without additional agents, selective cancer treatment with ultrasound alone has also been demonstrated in vitro with low-intensity ultrasound without relying on temperature rise. For example, experiments with monolayer cell cultures have shown that, compared to benign cells, some types of malignant cells are much more sensitive to the damaging effect from two to four minutes of exposure to low intensity (0.33 W/cm 2 ) continuous-wave (CW) ultrasound at 2 MHz [147] and 0.02 MHz [148], respectively. In another cell suspension model, pulsed ultrasound of low frequency (0.50 to 0.67 MHz) and low intensity (spatial-peak-temporal-average intensity I SPTA < 5 W/cm 2 , peak negative pressure < 1.2 MPa) has demonstrated specific killing effect on cancer cells suspended in phosphate buffered saline (PBS) when the pulse duration is longer than 20 ms, while most healthy cells remain undamaged [149]. However, the treatment efficacy drops significantly when the cells are suspended in a more rigid medium such as agarose and acrylamide gels. Compared to low-frequency ultrasound, high-frequency focused ultrasound has better targeting precision due to its smaller focal volume at a shorter wavelength. Previously, using pulsed high-frequency (18 MHz) low-intensity (ISPPA < 15.14 W/cm 2 ) focused ultrasound generated by microfabricated self-focusing acoustic transducers (SFAT), we have demonstrated selective cytolysis on both monolayers [150] and three-dimensional (3D) spheroids [151] of cancer cells with high spatial resolution of 100 and 150 μm, respectively. In both cases, we found that the acoustic intensity thresholds (AIT) for cytolysis of cancerous cells are substantially lower than those for benign cells, due to less organized actin cytoskeletal pattern (known to be associated with decreased cell stiffness) compared to benign cells [150]. Utilizing such a difference through keeping treatment acoustic intensity higher than the AIT of cancer cells and lower than that of normal cells, we successfully destroyed cancer cells without harming benign cells. 84 To further confirm the effectiveness of this non-thermal selective cancer treatment with high- frequency focused ultrasound, we have developed an SFAT along with a treatment system for in vivo treatment of B16F10 subcutaneous melanoma tumor in mice. 4.2 Device Design The SFAT (Figure 4.1) is based on a 1-mm-thick PZT-4 sheet (DL-47, DeL Piezo Specialties, LLC) working at its 9th harmonic thickness-mode resonant frequency of 20.70 MHz. On the top, there is a Parylene Fresnel lens with six constructive Fresnel rings for a focal length of 5 mm in water. A total Parylene thickness of 26 μm (which equals to quarter wavelength) is chosen to ensure the highest acoustic energy transmission [102] from the transducer to the medium. Figure 4.1 (a) Top-view photo of a fabricated SFAT before wires are soldered. (b) Microscope photo of part of the transducer (the dashed rectangular area in (a)), showing parts of five air-cavity rings with sealed release holes on the top electrode. The outermost air-cavity ring (top one in photo) covers part of the electrode and PZT. The design parameters of the transducers are summarized in Table 4.1. The operating frequency is chosen for keeping the frequency and spatial resolution similar to those used in our previous successful in vitro selective tumor treatment [150], [151]. To reduce unwanted dielectric (a) Sealed release holes 250 μm 1 mm (b) 85 heating (which is proportional to the PZT’s loss tangent [152]) during the transducer operation, PZT-4 instead of commonly used PZT-5A is selected due to its lower loss tangent, which is measured to be 0.15, 2.1 times lower at 20.70 MHz compared to PZT-5A. A thick PZT substrate operating at the 9th harmonic frequency rather than a nine-times-thinner PZT operating at its fundamental frequency is chosen due to the mechanical sturdiness of the thicker PZT for easy handling and packaging. Compared to commonly used focusing acoustic transducers based on a curved surface or multi-element phased array, the planar SFAT is microfabricated with high precision, has small footprint, and could be easily operated without complex driving electronics. Table 4.1 Design parameters of the transducer used for in vivo tumor treatment. Substrate Material Transducer Dimensions Air Cavity Height Parylene Thickness PZT-4 16 × 16 × 1 mm 3 3.5 μm 26.0 μm Electrode Material Working Frequency Simulated Focal Length Simulated Focal Diameter/Depth 10-μm-thick Silver 20.70 MHz (9 th harmonic) 5 mm 96 μm/790 μm Boundary radii of non-air-cavity Fresnel circle and rings (inner, outer in μm) 1 st circle 0, 595 2 nd ring 843, 1035 3 rd ring 1197, 1340 4 th ring 1471, 1591 5 th ring 1704, 1811 6 th ring 1912, 2009 The design of the SFAT has been verified through simulating the relative output acoustic pressure distribution in water with the finite-element method (FEM) using the Pressure Acoustics 86 module of COMSOL Multiphysics (COMSOL Inc.) at 20.70 MHz. Over the central vertical plane (Figure 4.2a), strong focusing effect happens 5 mm above the transducer center with 790 μm focal depth. On the focal plane (Z = 5 mm), the focal size is simulated to be 96 μm (Figure 4.2b). Figure 4.2 FEM-simulated normalized acoustic pressure (e) on the central vertical plane and (f) on the lateral focal plane (at Z = 5 mm), with same color bar scale but different dimension scales. After fabrication, the transducer is packaged onto a laser-machined acrylic holder (Figure 4.3a). For an easy electrical connection, the wires soldered on the transducer are connected to a subminiature version A (SMA) adapter attached on the side of the holder, passing through a ferrite core tube (2673000701, Fair-Rite Products Corp.) for shielding electromagnetic interference that might affect the operation of other equipment used for the tumor treatment. During transducer operation, a small portion of the input electrical power is dissipated in undesirable Joule and dielectric heating on and in the PZT, respectively, and raises the temperature significantly, which is confirmed to be not from heating effect due to acoustic energy at the focal point. To reduce Joule heating from the series resistance of the electrodes, thick (10 μm) silver electrode with low electrical resistivity is chosen along with large soldering pad area (94.5 mm 2 ) (a) (b) Z (mm) X (mm) Y (mm) X (mm) 87 (inset of Figure 4.3a), which reduces the resistance to mere 5 mΩ, making Joule heating almost negligible. 4.3 Tumor Treatment System To dissipate the heat generated on the transducer from dielectric heating, a compact water- cooling system is custom-built. As shown in Figure 4.3a, the transducer is attached to a small hollow water-cooling block made of nickel-coated copper (MCX Ram Block, Alphacool International GmbH) with thermally conductive paste (Kryonaut, Thermal Grizzly GmbH) having a high thermal conductivity of 12.5 W/(m·K). The transducer/water block assembly is held together by a clamping mechanism consisting of two laser-machined acrylic sheets with screws and nuts. On the acrylic sheet closer to the transducer surface, a 10 10 mm 2 opening is cut out to let the ultrasound waves pass through and to store ultrasound transmission gel (Scan Ultrasound Gel, Parker Laboratories) which serves as the coupling medium between the SFAT and the treated tumor. The inlet and the outlet of the water-cooling block is connected to a refrigerated liquid circulator (Neslab NTE 740, Thermo Fisher Scientific, Inc.) through two pieces of plastic tubing, forming a water-filled close loop. Driven by the pump of the circulator, the warmed water in the water-cooling block (due to the PZT’s dielectric heating) goes through the tubing and the refrigeration unit of the circulator, and then gets cooled down rapidly there. The cooled water is then pumped back into the water-cooling block, resulting in a closed loop in which the water temperature near the refrigeration unit is digitally controlled to be 8 °C, which is determined from thermal toxicity experiments described in the next section to avoid any potential thermal damage. In order to treat a tumor volume (varying from 30 to 1,500 mm 3 ) that is much larger than the volume of the focused ultrasound (about 4.2  10 -3 mm 3 ) generated by the transducer, a three-axis positioning and scanning system (Figure 4.3b) has been developed through modifying a 88 commercial 3D printer (Alunar M508) whose print head is removed and replaced with a customized holding platform consisting of an acrylic sheet and several metal blocks (Figure 4.3c). The transducer holder can be fixed onto the holding platform with the SFAT facing down towards the heating platform, and with inlet and outlet of the attached water-cooling block facing the back of the positioning system. As an aid for the alignment between the transducer and targeted tumor, also attached onto the holding platform are a low power (< 1 mW) laser diode (VLM-650-03 LPT, Quarton, Inc.) fixed vertically facing down in a metal housing (Fixed Laser Mounting Stand, Adafruit Industries, LLC) and its driving circuit (12 mm Coin Cell Breakout, Adafruit Industries, LLC) with a switch and a coin cell battery (Figure 4.3c). Since the lateral position difference between the focused laser spot and the transducer center is fixed and can be calibrated, once the laser focal point is positioned on the target (tumor center) (Figure 4.3d), the transducer can be brought to the same position through moving it by the calibrated distances. In the positioning system, the movement in X and Z directions is realized by moving the holding platform, while Y direction movement is achieved through moving the heating platform where the mouse is placed and kept warm during treatment (Figure 4.3b and 4.3d). 89 Figure 4.3 Photos of (a) a packaged SFAT on an acrylic holder, with soldered electrical wires connected to an SMA adapter for electrical connection and a hollow water-cooling block in a close-loop water-cooling system to dissipate heat from the transducer surface (Inset: front-view of the transducer clamped on the holder); (b) the 3D-printer- modified three-axis positioning system with a holding platform with the transducer attached and a heating platform where the mouse will be placed, showing how the movement in each axis is realized; (c) a close-up view of the dashed rectangular area in (b) showing the transducer holder on the right, and the laser diode in the black housing with its driving circuit on the left, all held by the movable holding platform with screws; (d) a mouse lying on the heating platform of the positioning system, anesthetized by isoflurane gas coming from an anesthesia nose cone where its nose is placed (as excessive gas is drawn away to the large exhaust pipe nearby), with a red laser dot aligned to the center of its tumor (highlighted with a dark skin marker). (e) Side-view cross-sectional diagram of the in vivo tumor treatment setup on mice. (f) Top-view diagram of the circular raster scan pattern of the transducer covering a circle of 4.8 mm diameter (dotted circle), showing the scan route (yellow-highlighted solid line) with 204 treatment spots (red dots). The width of the yellow line and the diameter of small solid circles are drawn in scale to indicate the focal size of the transducer, showing the actual treated area. 4.4 In Vivo Mice Treatment Protocol Animal experimentation is conducted in accordance with the ethical federal guidelines mandated by the University of Southern California Institutional Animal Care and Use Committee (Protocol 20542, approved on April 29 th , 2016). To test the effectiveness of SFAT treatment, the B16F10 melanoma tumor model [153] is grown in immune-competent C57/B6 living mice. 90 Before injection, B16F10 cells are maintained in Dulbecco's modified eagle medium (DMEM) supplemented with 10% v/v fetal bovine serum (FBS) and 1% penicillin-streptomycin in a humidified chamber at 37 °C under 5% CO2 and passed in fresh media every 2 to 3 days. Cells are harvested and diluted to 1  10 6 cells/mL in media with 50% Matrigel by volume. On Day 0, each mouse is anesthetized and injected subcutaneously with 1  10 5 cells in 100 μL into a shaved area on one flank. The mice are then randomized either for treatment or as controls to remain as untreated. Throughout the treatment period, all the mice are housed and maintained under identical conditions. Treatment is administered immediately (< 30 minutes) after injection on Day 0. For every seven days, treatment is repeated once per day on each animal for five consecutive days followed by two days without treatment. In each treatment, the treated animal with its tumor area freshly shaved is positioned on the heating platform (heated to 37 °C) of the positioning system while continuously receiving 1–4% isoflurane gas through a nose cone system to keep it sleeping (Figure 4.3d). Before the treatment, ultrasound transmission gel is applied into the shallow open window on the front acrylic sheet on the transducer surface to ensure good acoustic coupling between the transducer and the mouse skin. Then the SFAT with its top surface facing downward is attached to the movable holding platform. The surface tension between the ultrasound gel and the PZT/acrylic sheet keeps the gel from dripping down or flowing away. Next, with the aid of a laser diode, the transducer center is aligned to the center of the tumor (manually indicated by a skin marker) (Figure 4.3d). Then the laser is immediately turned off to avoid any potential heating effect, and the SFAT is lowered until the bottom of the front acrylic sheet (3 mm thick) reaches the top of the tumor, then raised 1.6 mm so that its focal point is at the top of the tumor, 0.4 mm below the skin surface (since the designed focal length is 5 mm and skin thickness is about 0.4 91 mm). During the treatment, the SFAT is mechanically scanned according to a pre-stored automatic scanning program written in G code, while simultaneously driven with pulsed sinusoidal electrical signals to produce focused ultrasound inside the tumor (Figure 4.3e). At each XY plane, the mechanical scanning pattern covers a circular area with 4.8 mm diameter, where the SFAT is raster scanned at a speed of 2 mm/s between spots with 0.3 mm spacing, and stops at each spot for 0.4 second for treatment, for a total of 204 spots per plane (Figure 4.3f). The same scan pattern is repeated at six XY planes with 0.3 mm spacing in the Z direction, covering 1.5 mm height and taking 11.5 min in total. When the mechanical scanning program is finished, the electrical signal is turned off, and the SFAT is then moved up and to the side through pre-programed movement. After that, the treated mouse is removed from the heating platform and allowed to recovery from anesthesia in a cage. During treatment period, tumor growth is monitored at least twice per week until the volume reaches about 1,500 mm 3 three weeks after injection, when animals are euthanatized and fresh tumors harvested. Two pieces of excised tumors from the control group are stored in phosphate- buffered saline (PBS) at 4 °C and are used within 12 hours in the ex vivo experiments for characterizing their acoustic properties (described in Supplementary Method I). The remaining excised tumors are fixed in 4% paraformaldehyde overnight at 4 °C, embedded in paraffin, and 5- μm-thick sections are prepared for histologic analyses. Sections are routinely stained with hematoxylin and eosin (H&E). In addition, immunohistochemistry (IHC) is performed with primary antibodies for Ki67 (MA5-14520, 1:50, Thermo Fisher Scientific, Inc.) and cleaved caspase-3 (9964S, 1:200, Cell Signaling Technology, Inc.), and then developed with DAB Poly Define Detection System (DS9800, Leica Biosystem GmbH) in BOND-III Automated IHC Stainer 92 (Leica Biosystems). Whole slide images are scanned by VS120 Virtual Slide Microscope (Olympus Corp.) and analyzed in OlyVIA software (Olympus Corp.). 4.5 Experimental Results 4.5.1 Characterization of Treatment Pressure Since it’s hard to directly measure the acoustic pressure within the tissue, the treatment acoustic pressure is estimated through FEM simulation along with measured parameters including the measured peak acoustic pressure in water, plus the acoustic attenuation coefficients and sound velocities within B16F10 melanoma tumor tissue and mouse skin. To characterize the acoustic intensity produced by the transducer, a commercial hydrophone (HGL-0085, Onda Corp.) fixed onto a five-axis movable stage (Prior Scientific Instruments Ltd.) is aligned to and scanned along the central vertical axis of an un-packaged SFAT facing up in water. During the hydrophone measurement (as well as the tumor treatment), a function generator (AFG-3252, Tektronix, Inc.) is used to produce 20.70 MHz pulsed sinusoidal voltage signals, which are amplified by a power amplifier (75A250, Amplifier Research Corp.) and applied to the SFAT. For hydrophone tests, the driving pulse width is set to be 2.9 μs, at which the acoustic pressure is independent of the pulse width. An oscilloscope (MDO3014, Tektronix, Inc.) is used to simultaneously monitor the applied voltage after a 40-dB voltage attenuator (100-SA-MFN-40, Bird Technologies) and the signal from hydrophone after a 20-dB pre-amplifier (AH-2010, Onda Corp.). At the focal point, as the applied voltage increases, the measured acoustic pressure increases almost linearly, and reaches 4.53 MPa when 211 V pp is applied on the SFAT (Figure 4.4a). The measured focal length and focal depth are 5.0 and 0.8 mm, respectively, which closely match with the simulated values (Figure 4.4a). Similar tests are repeated with 3-mm-thick 93 ultrasound transmission gel applied on the SFAT surface, and the measured acoustic pressure changes only by 2% compared to the previous case without the gel, suggesting that the acoustic properties of water and ultrasound gel are very similar. The attenuation coefficients and sound velocities within B16F10 tumor tissues and mouse skin has been measured through similar hydrophone tests. Freshly excised tumor tissues are dissected with a scalpel to produce tumor and normal skin samples of varying thicknesses. With the hydrophone measurement setup shown in Figure 4.4b, a tumor or skin slice is attached on the SFAT’s top surface with a thin (< 1-mm-thick) layer of ultrasound transmission gel. In each measurement involving a tumor/skin slice, the acoustic pressure at the focal point is measured with the hydrophone. Taking the peak acoustic pressure measured in water with no tissue/skin above the transducer to be P0 (measured to be 4.53 MPa as shown in Figure 4.4a), the attenuated focal- point pressure Patten after passing through a thin layer of tissue with thickness d and attenuation α can be expressed with 𝑃 = 𝑃 × 𝑒 . (4.16) For simplicity, in (1), we ignore the acoustic loss due to reflection from the tissue slices (which is small, since tissues have similar acoustic impedances to that of water/ultrasound gel), and we assume the path-length from anywhere in the small transducer active area to the focal point is equal to the focal length (paraxial approximation). By taking the natural logarithm on both sides of (1), we have 𝑙𝑛 (𝑃 ) = 𝑙𝑛 (𝑃 ) − 𝛼𝑑 , (4.17) where ln(P0) is a constant. Thus, by plotting ln(Patten) versus d, we can extract the attenuation coefficient (α) through a linear fitting. From Figure 4.4c, the attenuation coefficients in tumor and skin in the mouse are estimated to be 0.321 and 1.091 np/mm at 20.70 MHz, respectively. 94 Additionally, to estimate the sound velocities in tumor or skin tissues (c1) , we measure the focal length (F) and the time-of-flight for the first ultrasound beam from the transducer center to arrive the hydrophone (t1) after passing through a tissue with thickness d. Assuming that the sound velocities in the ultrasonic gel and water are the same (c0 = 1480 m/s [1]), we have 𝑡 = 𝑑 𝑐 ⁄ + (𝐹 − 𝑑 ) 𝑐 ⁄ , (4.18) from which we get 𝑐 = 𝑐 𝑑 (𝑐 𝑡 − 𝐹 + 𝑑 ) ⁄ . (4.19) According to the measurements, the average measured sound velocities in the tumor and skin are 1,521 and 1,558 m/s, respectively. 95 Figure 4.4 (a) Hydrophone measurement (black) and normalized simulation (red) of the output acoustic pressure in water along the central vertical axis, with 211 V pp applied on the transducer. (b) Cross-sectional diagram of the hydrophone measurement setup with a piece of tissue slice attached to the top of the SFAT with ultrasound transmission gel. (c) Natural logarithm of the measured acoustic pressure at the focal point when tumor (grey dots) or skin (red dots) slices of different thicknesses are placed on the SFAT, with linear fitting lines whose slopes are equal to the attenuation coefficients in the tissues (unit: np/mm). (a) (b) (c) 96 Table 4.2 Material Properties Used in The Simulation of the Treatment Acoustic Pressure Material Thickness (mm) Mass Density (kg/m 3 ) Sound Velocity (m/s) Attenuation Coefficient (np/mm) B16F10 Tumor 5 (11 in diameter) 1060 * 1521 ** 0.321 ** Mouse Skin 0.4–1.4 1060 * 1558 ** 1.091 ** Ultrasound Gel *** NA 1000 * 1480 ** 0.110 ** * The average density of soft tissues [154] ** From measurement *** Used the values for water [1] due to similar acoustic properties Using the material properties from Table 4.2, we simulate the treatment acoustic pressure in tumor which is modelled as an ellipsoid whose depth is 5 mm and diameter 11 mm, and the skin thickness is assumed to be uniform across the tumor, ranging from 0.4 to 1.5 mm in different simulations. For simplicity, all materials modelled in the simulations are assumed to have isotropic and homogeneous material properties. The SFAT is aligned to the center of the tumor, and the distance between the SFAT and the bottom surface of the tumor skin is varied over 3.1, 3.4, 3.7, 4.0, 4.3, and 4.6 mm, which correspond to the cases where the SFAT is positioned in the center of the six XY scan planes during treatment. From the simulation, we see that the maximum acoustic pressure within the tumor tissue varies with skin thickness and the SFAT-skin distance, and the focal zone extends up to about 1.5 mm into the tumor (Figure 4.5a and 4.5f). 97 Figure 4.5 Simulated acoustic pressure distributions during treatment for 0.4 mm skin thickness, if the distance between the SFAT and the bottom of skin is (a) 4.6 mm, (b) 3.7 mm, and (c) 3.1 mm; and similar simulations for 0.9 mm skin thickness, if the distance between the SFAT and the bottom of skin is (d) 4.6 mm, (e) 3.7 mm, and (f) 3.1 mm; all sharing the same color bar in (a) with unit being MPa. (g) Simulated maximum acoustic pressure (and corresponding mechanical index) in tumor tissue during treatment versus different skin thicknesses as a function of SFAT-skin distances. When skin thickness is 0.4 mm, the maximum treatment pressure varies from 2.1 to 3.0 MPa (Figure 4.5g), which corresponds to a mechanical index (MI, calculated by dividing the peak (a) b) (c) (d) (e) (f) (g) (g) 98 negative pressure in MPa by the square root of frequency in MHz [155]) of 0.46 to 0.66, and is the highest among all cases. The highest MI value of 0.66 is lower than the threshold value of 0.71 where cavitation may happen for short-pulse (a few cycles), low-duty-cycle (< 1%) ultrasound [155], and much lower than the United States Food and Drug Administration (FDA) safety limit of 1.9 for diagnostic ultrasound [156]. 4.5.2 Temperature Measurement and Thermal Toxicity Tests The device driving condition is determined through experiments in which skin temperature is monitored in real time to ensure there is no thermal damage. During the tests, we keep the same frequency, driving voltage and pulse repetition frequency (20.70 MHz, 211 V pp and 60 Hz, respectively), while varying the pulse width. In the treatment setup, we attach one k-type thermocouple from a digital datalogging thermometer (HH506RA, Omega Engineering Inc.) onto the skin area above the tumor target to monitor the treatment temperature, while a separate thermocouple is positioned over untreated skin to measure the body temperature of the mouse. During treatment, the body temperature slowly drops from about 34 to 32 °C as the mouse is being anesthetized and kept warm by the heating platform (Figure 4.6a). Before the treatment, the skin temperature in the center of the tumor is relatively low (24 to 25 °C) due to the cooled transducer surface and ultrasound gel. After the SFAT is turned on and being mechanically scanned, the temperature increases, then saturates and changes slightly during the treatment, as the distance between the SFAT and the thermocouple varies during mechanical scanning (the periodical temperature drops in Fig. 5a happen when the ultrasound gel cooled by the relatively cold non- active surface area of SFAT reaches the thermocouple). From the experiments, a pulse width of 1.45 ms is chosen so that the maximum skin temperature in the treatment area remains below 35 °C throughout the 11.5 min treatment, which is less than 1 °C higher than measured normal body 99 temperature (Figure 4.6a). With this temperature, the estimated cumulative equivalent minutes at 43 °C (CEM43, the accepted metric for thermal dose assessment, which estimates the equivalent exposure time at 43 °C for a thermal exposure of a certain temperature for a given time) for a 11.5 min treatment is only 0.01 s, which is far from causing any thermal damage [157]. Thermal toxicity tests on normal mouse skin with the same treatment condition and duration are also conducted to examine the potential thermal damage on normal tissues. The 11.5 min treatment with a maximum temperature of 35 °C measured directly in the treated area does not seem to cause any lasting visible or microscopic lesions. No histologic effect is observed on normal, shaved mouse skin in non-tumor bearing animals 24 hours after the treatment (Fig. 5b). With a pulse width of 1.45 ms, for an estimated treatment pressure of 2.1 to 3.0 MPa (assuming a skin thickness of 0.4 mm), the calculated ISPPA and ISPTA are 136.8–279.1 and 11.9– 24.3 W/cm 2 , respectively, which are about an order of magnitude lower than those used in thermal- based HIFU treatment. Compared to the FDA safety limits for diagnostic ultrasound (I SPPA < 190 W/cm 2 , ISPTA < 720 mW/cm 2 , MI < 1.9) [156], our values are close in ISPPA, much higher in ISPTA, and much lower in MI. Thus, we define our treatment intensity to be “medium intensity” when compared to HIFU and diagnostic ultrasound. 100 Figure 4.6 (a) Typical temperature change on the skin above the treated tumor center (grey) and on an untreated skin area nearby with no tumor underneath (red) during the treatment with 1.45 ms pulse width. (b) Representative histology images of untreated normal skin (upper panel) compared with skin harvested 24 hours after treatment (lower panel) in thermal toxicity experiment with a maximum temperature of 35 °C. No histologic changes are noted. 4.5.3 Treatment Results The tumor weights measured at the end of the treatment are summarized in Figure 4.7. We observe a significant difference in mean tumor weights according to treatment assignment. Specifically, the mean ± standard deviation weights of the treated versus the untreated tumors are 390 ± 200 mg (n = 8) versus 885 ± 506 mg (n = 5), respectively (student’s t-test, p = 0.03). Histologic analysis is used to identify effects of focal ultrasound treatment on tumors. While areas of viable cancer cells are readily found in both treatment and control groups (indicated by * in Figure 4.8a and 4.8b), the tumor area is smaller in the treated tumors. Moreover, in the expected treatment areas of the treated tumors, large areas of necrosis are identified, which can be seen in a large pale area of necrotic cells lacking nuclear detail (solid arrow in Figure 4.8b). Cells in the same area also exhibit strong expression of cleaved caspase-3, suggesting ongoing cell apoptosis (solid arrowhead, Figure 4.8f). In contrast, the untreated tumor exhibits a large continuous area of highly proliferative cancer cells indicated by strong Ki67 expression (open arrow, Figure 4.8c) without cleaved caspase-3 expression (open arrowhead, Figure 4.8e). In all cases, no damage is (a) (b) 101 found in the skin area positioned directly under the transducer during the treatment (solid triangles in Figure 4.8a and 4.8b). Figure 4.7 Tumor weights of control (n = 5) and treated (n = 8) tumors shown in boxplots ( * : student’s t-test p < 0.05). 102 Figure 4.8 Representative cross-sectional histologic images of control and treated B16F10 tumors, showing matched sections representing control (untreated) and ultrasound-treated tumors that are stained for H&E ((a) and (b)), Ki67 (an indicator of cell proliferation, (c) and (d)), and cleaved caspase-3 (an indicator of cell apoptosis, (e) and (f)). *: viable tumor; ▲: normal skin overlying tumor; solid arrow and solid arrow head: area of necrosis and apoptosis; open arrow and open arrow head: large area of proliferating cancer cells without apoptosis. 4.6 Discussions In an in vivo treatment using the B16F10 cell model on mice, selective cancer treatment has been achieved with pHFMIFU generated by an SFAT, when the heat generated during treatment is too small to cause any damage. Previously, we have suggested that this selectivity might be a result of the disorganized cytoskeletal structure observed in cancer cells [150], [158], resulting a 103 reduced cell stiffness [158], [159], making them much easier to deform under mechanical stress, and thus easier to be damaged. It has been demonstrated in in vitro experiments that the cytoskeleton could be disrupted by very low intensity (290 kPa) of ultrasound at 1 MHz [160]. Due to the less organized cytoskeletal structure of cancer cells, the acoustic intensity threshold for permanent damage on cancer cells is likely lower compared to that for benign cells, which has been confirmed by our previous in vitro experiments [16]. Another possible cause for this selectivity is the different natural mechanical resonant frequencies of cancerous cells and healthy cells, resulted from their differences in material properties (such as stiffness) and cell geometries (such as nuclei size) [161], [162]. As a result, cancer and normal cells will effectively react to ultrasound of different frequencies. The exact mechanism of damage from our non-thermal pHFMIFU treatment is so far unclear, but there are several theories available to explain the effect. According to the low MI (< 0.66), the chance of inertial cavitation is small. However, with the relatively long pulse width of 1.45 ms, stable cavitation might be induced, in which microbubbles are formed and periodically oscillates without collapsing. These oscillating microbubbles could move at high speed in response to ultrasound-induced acoustic radiation force [163], generate microstreaming motion around them [11], and scatter acoustic waves [164], causing bioeffects such as increased cell membrane permeability and perturbed cytoskeleton structure [165], which could cause destructive damage to cells. Apart from cavitation, ultrasound itself could induce damage through acoustic radiation force which may produce displacements of cells to generate shear strain [166], or through generating acoustic microstreaming flow inside or around cells [167]. In another theory, the acoustic waves could react on the lipid bilayer cell membrane, causing it to expand and contract, thus transforming the oscillating acoustic pressure waves into smaller-scale intracellular deformations [168]. In our 104 future work, we plan to carry out experiments to study the underlying mechanisms of this selective cancer treatment. While we observed a decrease in the average tumor weight in the treated group, the results are not uniform or consistent across all tumors, and we do not see a complete destruction of cancer cells in the treated animals, which is possibly caused by a few factors. The first reason is the discrepancy between the treated volume and the actual tumor volume. To avoid an impractically long treatment time, we limit the number of treatment spots, covering only a small portion of the total treatment volume. With the current scanning pattern, the total effectively treated volume consisting all the treatment spots (where the transducer stops for 400 ms) covers only 8.9% of the total treatment volume. Even with the inclusion of the scanned volume (where the transducer is moved at a relatively fast speed of 2 mm/s between treatment spots), the value increases only to 37.2%, suggesting a large volume of tumor being left untreated (Figure 4.3f). Moreover, while the treatment target is established immediately prior to each activation, the shape and extent of the total treatment volume is fixed once treatment is started and did not vary over the entire course of the experiment. During the later phase of the tumor growth experiment, tumor diameters of 5-12 mm were observed in some of the treated tumors. However, the treatment pattern constrained by the 4.8 mm circular raster pattern was held constant. We also note the difficulty of employing our rigid positioning system for anesthetized, living animals. The treatment volume could also be affected by the respiratory and involuntary motion from the mice observed during treatment, especially considering the short treatment time on each treatment spot (corresponding to only 24 pulses per spot). Finally, the shaved mouse skin is not perfectly uniform over the tumor bearing areas which may result in inhomogeneous acoustic doses delivered to each tumor. As future work, to ensure better treatment consistency, we will realize a full treatment coverage of tumor without 105 increasing the treatment time through modifying the Fresnel lens design to double the focal depth of the ultrasound while maintaining the small focal diameter [83] and using an array of transducers. In addition, a customized mechanical scanning pattern to better fit the tumor shape could be implemented through 3D shape scan of tumors. Moreover, treatment conditions (such as acoustic intensity, pulse width, PRF, and treatment duration) can be optimized to ensure the best selective therapeutic effects on tumor with shorter treatment time and lower acoustic power. 4.7 Summary To further confirm the non-thermal, selective killing effects of high-frequency focused ultrasound previously demonstrated in in vitro experiments involving monolayers and spheroids of cancer cells, this study examines the therapeutic effects of pulsed (60 Hz PRF, 1.45 ms pulse width) high-frequency (20.7 MHz) focused ultrasound with medium-intensity (peak pressure < 3.0 MPa, ISPPA < 279.1 W/cm 2 , ISPTA < 24.3 W/cm 2 , MI < 0.66) in an in vivo subcutaneous B16F10 melanoma tumor growth model in mice. The ultrasound is generated by an SFAT with Fresnel air- cavity lens designed for 5 mm focal length, with a focal diameter and focal depth of 96 and 790 μm, respectively. A three-axis positioning and scanning system has been developed to realize automatic mechanical scanning of the transducer to cover a larger cylindrical treatment volume with 1,224 treatment points in 11.5 min with 400 ms duration per spot. Throughout the treatment, the skin temperature in the treated area is kept below 35 °C, which is too low to generate any thermal damage. After three weeks of ultrasound treatment, the treated tumors have significantly less weight compared to the untreated ones, and histologic analyses have revealed cell necrosis and apoptosis in the treated area underneath the skin. In addition, no damage has been detected in normal skin or tissues in the treatment and thermal toxicity tests, suggesting a successful non- thermal, selective tumor treatment. 106 In contrast to low-frequency high-intensity ultrasound, our in vivo experiment demonstrates the potential use of pHFMIFU as a new tool for selective cancer treatment with much better spatial resolution, in regions where it is critical to keep the surrounding normal cells and tissues undamaged from the cancer treatment. The potential mechanisms of this selective killing effect as well as the possible causes of and solutions to the variation in treatment results are analyzed and proposed. 107 5Chapter 5 SFATs for Localized Neurostimulation 5.1 Introduction Neurostimulation technology can greatly benefit those who suffer from neurologic diseases like epilepsy, or profound losses to various sense organs, as well as for reduction of severe, chronic pain which would otherwise require constant, high-dose opioid therapy (such as neuropathic pain and spinal cord injury). Neurostimulation is usually achieved by implanting microelectrodes or applying surface electrodes to realize electrical stimulation through charge-balanced biphasic constant current or capacitively coupled charge injection. However, a consistent finding from the histologic evaluation of electrode implantation sites is the gradual loss over time of neurons within approximately 50-70 μm of the microelectrodes [169]. If excessive, this neuronal loss can reduce the effectiveness of the applied electrical stimulus and consequently, reduce the ability of microstimulation to achieve a graded activation of neuronal populations. Surface electrodes, while less invasive, lack the ability to generate a highly localized electrical field. Other non-invasive techniques such as transcranial magnetic stimulation (TMS) [170] and transcranial electrical stimulation (TES) were also explored [171] by transcranially applying magnetic field or electric currents, but more experiments are still needed to confirm their efficacy and potential side effects. Acoustic energy is another candidate for transcranial neurostimulation. Ample evidence supports that acoustic energy modulates neuronal activities in both the peripheral and central nervous systems. Although the mechanism for reversible suppression of neural activity by ultrasound is not well understood, in 1958 William Fry, et al. [172] reported interesting neural effects when a cat’s eye was stimulated with light pulses. Light-evoked cortical potentials were 108 recorded at fixed time intervals before, during, and after irradiation from a focused ultrasonic beam onto the lateral geniculate nuclei. The primary and secondary responses were reduced by > 3 times and to almost zero responses, respectively, when the acoustic intensity was properly chosen. The responses returned to normal 30 minutes after the ultrasonic irradiation. The results were repeatable, showing a specific range of acoustic intensities that could cause suppression of neural activity without damaging neural tissue. Recent reports show that acoustic energy is: 1) capable of stimulating voltage-gated Na + and Ca 2+ channels sufficient to evoke action potentials and trigger synaptic transmission in hippocampal neurons [173]; 2) can change membrane permeability, altering the ionic flux through the cell membrane [174]; 3) can induce reversible increases in the internal Ca2+ concentration of fibroblasts [175]; and 4) can modulate potassium ion influx and efflux in rat thymocytes [176]. A recent report in a normal intact mouse showed that acoustic energy stimulates neuronal activity via modulation of sodium channels, without a significant rise in the brain temperature (<0.01°C) [177]. With focused ultrasound, SFAT can exert a body force on micron-sized areas from 5 to 200 μm in diameter, which makes it an attractive alternative for ultrasound neurostimulation. It also has the benefits of (1) the small (10-micron) focal stimulation area for high spatial resolution; (2) ability to focus the stimulus onto multiple points using a single transducer or an array of transducers; (3) small transducer size; and (4) low physical profile suitable for chronic implantation. With a typical focal length of 0.4 – 10 mm (although the focal length can be made to be longer than a few cm), SFAT can be placed on the surface of the cerebral cortex (just beneath the skull). The proximal placement of the transducer coupled with the transducer’s ability to target multiple points over a large area offers unprecedented ability to generate various responses in the cerebral cortex. However, the precise effects of ultrasound on neuromodulation are not fully 109 understood. For example, given that cortical neurons lack mechanoreceptors, it is not clear whether they can be stimulated at acoustic intensities low enough to avoid cavitation or heat. Thus, we would like to investigate focused ultrasound stimulation (FUS) based on our SFAT technology as a novel neuronal stimulation therapy for neurologic diseases by investigating its effects on neurons using patch-clamp technology for potential application in the treatment of neurologic diseases such as epilepsy. Our SFAT approach, if successful, will be the first MEMS- technology-based system for highly localized neurostimulation that involves neither heat, cavitation nor mechanoreceptors. 5.2 The Patch-clamp Technique for Studying Neuronal Activities The patch-clamp technique allows the investigation of a small set or even single ion channels, which affects our neural activities by letting ions going into or out of neuron cells, and thus changing their membrane potential [178]. In the experiment, a glass pipette containing electrolyte solution is tightly sealed onto the cell membrane through suction and thus isolates a membrane patch electrically (Figure 5.1). Currents fluxing through the ion channels in this patch hence flow into the pipette and can be recorded by an electrode that is connected to a highly sensitive differential amplifier. In the voltage-clamp configuration, a current is injected into the cell via a negative feedback loop to compensate for changes in membrane potential. Recording this current allows conclusions about the input resistance of the membrane. The membrane potential can be manipulated independently of ionic currents and this allows investigation of the current-voltage relationships of membrane channels. Alternatively, in the current-clamp configuration, current passing across the membrane is controlled and the resulting changes in voltage are recorded, generally in the form of action potentials. 110 Figure 5.1 [178] (a) Schematic diagram showing the setup of patch-clamp experiment: current flux within the patch pipette, sensing electrode and amplifier; (b) A phase contrast image of a patch pipette attached to the membrane of a cultured murine hippocampal neuron. The major challenge of using the patch-clamp technique to study the effects of ultrasound lies in that it requires good visibility under the microscope to precisely place the micropipette. However, most ultrasonic transducers are opaque and would block the light source beneath the tissue placed over it, resulting in a type of “blind patch-clamp experiment.” Such a “blind patch- clamp experiment” adds complexity and uncertainty to experiments that are already challenging and that, to our knowledge, has not been done before. As a result, additional setups such as an acoustic reflector cone have to be used to redirect the ultrasound waves from an ultrasonic transducer placed on the side of the illumination path to brain tissue [179], adding complexity to the system and the acoustic field. Other researchers emulate the effects of ultrasound through mechanical stimulations such poking [180], which cannot accurately represent the actual ultrasound stimulation. To solve this problem, we have developed SFATs based on translucent 127-μm-thick PZT sheets, which ensures good visibility during patch-clamp experiments, without additional modification of the setup. 5.3 Device Design (a) (b) 111 For easy handling and fabrication, and to ensure the best light transmission through the transducer, we first developed lens-less SFATs with patterned Fresnel electrode rings based on 127-μm-thick PZT-5A sheets. On the SFATs, there are 15 constructive Fresnel rings designed for 18.4 MHz operating frequency (fundamental thickness-mode resonant frequency of the PZT) and 0.4 mm focal length in rat brain tissue (assuming a sound velocity of 1540 m/s within the tissue) (Figure 5.2a and 5.2b). The high frequency is selected to ensure a small focal diameter (and thus better spatial resolution), so that the simulation effect will be highly localized. The focal length is chosen to be the same as the thickness of a rat brain tissue used in the patch-clamp experiment so that the focal point will be near the top surface of the tissue, where the micro-pipette will be patched (Figure 5.2a and 5.2b). Due to the relatively small thickness of the PZT, back-illuminated infrared light can pass through the transducer where there is no electrode without losing much intensity, ensuring good visibility under a microscope (Figure 5.2c). Instead of selecting the center circle as the first Fresnel ring, the second ring from the center is chosen to be the initial Fresnel ring, followed by every other ring outward. This design allows light to pass through the electrode-less center of the chip, illuminating the region that will be patched during the experiment (Figure 5.2b and 5.2c). In some devices, a piece of laser-machined 0.5-mm-thick polyester sheet is attached to the bottom with superglue, in an effort to minimize the device tilting despite the solder bumps at the bottom, while also increasing the mechanical robustness. The sheet adds more attenuation to the light passing through the device, but this effect can be compensated by increasing the illumination intensity. 112 Figure 5.2 (a) Cross-sectional schematic view and (b) top-view photo of a 127-μm-thick electrode-ring SFAT designed for neurostimulation experiments, aCSF is short for artificial cerebrospinal fluid; (c) Microscope photo of the SFAT with infrared illumination from the back, showing good visibility through the translucent PZT substrate; (d) Bottom view of an SFAT with a 0.5-mm-thick polyester sheet attached on the back for improved mechanical robustness and easiness of handling. According to the simulation (Figure 5.3), the depth of focus of the designed SFAT is 1.25 mm, and the focal diameter is only 34 μm. The long depth of focus is desirable, since the actual tissue thickness may slightly vary from the designed 0.4 mm. (a) (b) (c) (d) (f) 113 Figure 5.3 FEM-simulated relative acoustic pressure distribution of the SFAT (a) in the central vertical plane, (b) at the focal plane where Z = 0.4 mm). To ensure that the stimulation effect is truly at the focal point aligned to the center of the device, we also would like to patch the non-center areas of the brain tissue during ultrasound generation, where no ultrasound-induced neuronal activities are expected. To achieve this, we have developed SFATs with some electrode rings cleared (Figure 5.4a to 5.4d), to realize good visibility of cells at these regions. To further enhance the visibility through the transducer, we have also etched through-chip via-holes and circular slits on the PZT substrate, creating areas where light could directly pass through without attenuation (Figure 5.4e). Since the etch rate of PZT using typical microfabrication dry etching processes such as reactive ion etching (RIE) or inductively coupled plasma (ICP) RIE is too slow to etch through 127-μm-thick PZT, we used micro-powder blasting [181] to achieve such task, where a focused jet of micron-scale abrasive particles is used to ablate material from a substrate. During the blasting process, a 100-μm-thick dry film photoresist is used to define the opening patterns (through photolithography) and protect the unexposed areas. (a) (b) 114 Figure 5.4 (a) Design and (b) microscope photo of 0.4-mm-focal-length SFAT on 127-mm-thick PZTs with 7 rings (with 8 th -15 th rings cleared) for 18.4 MHz. (c) Design and (d) microscope photo of an SFAT designed for the same focal length and frequency with 12 rings (with 8 th -10 th rings cleared). (e) Microscope photo of an SFAT with a drilled hole and circular slits through micro-powder blasting, with a piece of blue tape on the other side to demonstrate the improved visibility. (a) (b) (c) (d) (e) 115 In addition, two types of SFATs based on Parylene air-cavity lenses have also been developed. The first type shares the same Fresnel lens design as the electrode-ring SFAT mentioned earlier, working at 18.4 MHz (Figure 5.4a). Another type is designed for the same focal length and has the same number of rings, but works at 54.0 MHz, which is the 3 rd harmonic resonant frequency of the 127-μm-thick PZT (Figure 5.4b). This device can be used to examine the effect of operating frequency on the ultrasound-induced neuron stimulation effects. At this high frequency, the width of the Fresnel rings can be as narrow as 16 μm, which is hard to fabricate if patterned electrode rings are used, due to the lateral undercut during wet etching. To ensure good light transmission, the circular electrode below the air-cavity lenses has been patterned to expose the center and several peripheral areas (Figure 5.4c). Figure 5.5 Top-view photos of SFATs based on 15-ring Parylene air-cavity lenses designed for the same 0.4 mm focal length, at operating frequencies of (a) 18.4 MHz (the fundamental thickness-mode resonant frequency of a 127- μm-thick PZT); and (b) 54.0 MHz (the 3 rd harmonic thickness-mode resonant frequency of a 127-μm-thick PZT). (c) The electrode pattern (black) below the air-cavity lens, with exposed areas to let light pass through. To verify that the induced neurostimulation effects are truly from focused ultrasound rather than other factors such as electromagnetic interference or heat generation, we developed control devices which have the same design (18.4 MHz, 0.4 mm focal length, 15 rings) as the electro-ring SFATs, but with a large Parylene-sealed air cavity (fabricated through sacrificial layer releasing (a) (b) (c) 116 technique) covering most electrode regions, to block acoustic waves coming from the transducer (Figure 5.6a). However, during experiments when a water-immersed control SFAT is actuated, we still notice acoustic waves coming from the device, generating vibrations on the water surface 0.4 mm above the transducer surface. This might be due to acoustic waves leaking from the sealed release holes of the air cavity, or because of the small thickness. To completely block acoustic waves, we laser-machined an air-cavity cover made of three acetate sheets, creating a large air cavity when it is bonded to the cover SFAT. The openings from the acetate-film air cavity are then sealed with waterproof silicone sealant and another deposition of Parylene. Figure 5.6 Photos of (a) a control SFAT with a large Parylene-sealed air cavity covering most of the electrode regions; (b) an air-cavity cover made of three pieces of acetate sheets, before being attached to an SFAT; and (c) a control SFAT with an acetate-sheet-sealed air cavity. (a) (b) (c) 117 5.4 Experimental Results To verify the effect of focused ultrasound on neural activities, we carried out ex vivo patch- clamp experiment (carried out by Prof. Su-Youne Chang from Mayo Clinic, MN, USA) involving rat (Sprague-Dawley rat with 250–300 g body weight) hippocampal brain tissue slices. During the experiment (Figure 5.7a to 5.7c), the SFAT is placed at the bottom of a transparent fluid chamber, with a piece of brain slice having a thickness of around 0.4 mm placed on top. The chamber has a fluid inlet and an outlet, which enable the continuous perfuse of artificial cerebrospinal fluid (aCSF) through the chamber at a rate of 2 ml/min to keep the brain tissue functioning. Thanks to the unique design of SFATs, we could visualize cells in the slice placed on top of an SFAT between the microscope lens (connected to an infrared-imaging camera) and the infrared (IR) illuminating light source from underneath the transducer. The SFAT reduced the visibility greatly, but we could visualize cells near the focal point (Figure 5.7d)., and we are able to precisely patch cells near that region (Figure 5.7b and 5.7c). Inside the micropipette patched to a neuron cell, there is an electrolyte and a recording electrode, which is connected to an amplifier and a signal recording system, allowing us to monitor the cell membrane potential (an indicator of the neural activity of the cell) in real-time and record the data on a computer. In the meanwhile, the SFAT is driven with pulsed sinusoidal voltage signals of 18.4 MHz, generating pulsed focused ultrasound. In different experiments, driving conditions such as driving voltage, pulse repetitional frequency (PRF), and pulse width are varied to examine their effects on neurostimulation. 118 Figure 5.7 Cross-sectional-view diagram of the rat-brain-tissue patch-clamp experiment setup with an SFAT. Photos of the experiment setup (b) without a microscope lens, showing a piece of brain slice on top of an SFAT patched by a micropipette; and (c) with a microscope lens. (d) Microscope photo showing a neuron cell within the brain tissue slice patched by a micropipette during the experiment. In our initial experiments, we are able to generate action potential (a sudden surge in membrane potential, indicating neural activity) with 17.6 MHz pulsed ultrasound with 10 Hz PRF and a pulse width of 284 μs (corresponding to 5000 cycles per pulse), as the applied voltage on the SFAT is varied (Figure 5.8). We also find out the success rate of inducing action potential (AP) generation is dependent on the stimulation intensity. If we increased the acoustic intensity through applying higher voltage on the SFAT, we could generate action potentials (Figure 5.8). (a) (c) (d) (b) Neuron cell Micropipette tip 119 Figure 5.8 Measured membrane potential from a patched neuronal cell: (a) by focused ultrasound delivered by SFAT driven at 2, 3 and 4 V pp and (b) enlarged traces. Arrows show stimulation artifact. Action potentials (APs) were not generated by 2 V pp, but 3 V pp could generate APs, but not all stimuli could generate AP. Each ultrasonic stimulation with 4 V pp could generate AP, the peak of which was 80-100 mV. Compared to the action potentials (APs) generated by electrical stimulation using 100 pA current, the ultrasound-evoked APs (with 17.6 MHz, 5 Hz PRF focused ultrasound) seem to be cleaner with a hyperpolarization (cell membrane potential becomes more negative) period after each stimulation (Figure 5.9). (a) (b) 120 Figure 5.9 Comparison of action potentials (APs) between electrically-evoked and ultrasound stimulation evoked: (a) Representative cellular responses to electrical and ultrasound stimulation. For electrical stimulation, 100 pA current was injected through the patch pipette. (b) Enlarged traces. Filled circle is a single AP. Arrow is an artifact. To further confirm the generation of action potentials and to examine the mechanism involved in the ultrasound-induced neurostimulation, we utilized a selective ion channel blocker to disable a certain type of ion channel. For example, tetrodotoxin (TTX, 0.3 µM), a voltage-gated Na+ channel blocker, was applied onto the slice. We noticed that TTX successfully abolished APs generated by the focused ultrasound (Figure 5.10), suggesting the potential involvement of Na+ ion channels during the stimulation process. (a) (b) 121 Figure 5.10 APs without and with Tetrodotoxin (TTX), a Na+ channel blocker: (a) AP generation evoked by ultrasound stimulation. (b) Ultrasound evoked APs were abolished by TTX (0.3 μM), the vertical axis is not at the same scale. 5.5 Summary We have designed and fabricated SFATs working at high frequencies (18 MHz and 54 MHz) with very high spatial resolution for rat-brain patch-clamp experiments. The special design of the SFATs ensures good visibility under the microscope with IR illumination, allowing the micropipette to be precisely patched on neuron cells near the focal zone. Some initial experiments have been carried out, and the encouraging preliminary data show that it is possible to stimulate neurons with pulsed focused ultrasound, and the success rate of AP generation depends on the acoustic intensity. A selective ion channel blocker is used to examine the ion channels involved in the ultrasound stimulation process. More experiments involving control devices, different driving conditions, and different types of ion blockers will be carried out to further verify the stimulation effects and to study the underlying mechanism. (a) (b) 122 6Chapter 6 Acoustic Tweezers Based on Modified Fresnel Acoustic Lenses 6.1 Introduction Contactless micromanipulation or tweezing of particles in liquid is highly desirable in many applications where the particles are sensitive to physical contact force or potential contamination, such as cell manipulation, microinjection into cells, protein crystallization, etc. Contactless trapping of cells, for example, allows defect-free enrichment of cells through cell-cell or cell- chemical interactions [182] and boosts the sensitivity of bioassays [183]. Contactless trapping can be obtained through optical tweezing [184], dielectrophoresis trapping [185], and magnetic trapping [186]. Optical tweezers trap particles at the focal point of a laser beam with high precision, but require optical transparency in liquid and is incapable of trapping particle whose size is much greater than the light’s wavelength or whose mass is substantially large. Also, high energy contained in focused laser beams may cause potential damage to living cells or bioparticles. Dielectrophoresis tweezers trap particles by non-uniform electrical field created by microeletrodes, but they introduce current-induced heating and electrical field which may interact with (or damage) cells. Furthermore, the permittivity of the particle has to be different from that of the liquid medium, requiring a low-salt, non-physiological environment for cells. Magnetic trapping is based on non-uniform magnetic force for keeping the particles in place, but requires that particles be either paramagnetic or ferromagnetic, which is not the case for most bioparticles or cells, otherwise magnetic beads have to be attached to the particles to be trapped. Also, micromagnets are usually not sufficient to generate a high magnetic field, so large permanent magnets are needed, making integration with small lab-on-a-chip systems very hard. 123 Moreover, the efficiency of magnetic trapping will be significantly impaired by a slight misalignment of the magnets. Acoustic tweezers, on the other hand, work on many kinds of materials in a wide range of liquid media, regardless of their optical or electromagnetic properties, with very little heat. Researchers have also studied both short- and long-term influences of the ultrasound used in acoustic trapping on cells, and reported almost no effect of the ultrasound on the cells’ viability [182], [187]–[189]. Thanks to these attractive features, acoustic tweezers have been used for trapping microspheres [21], cells [182], organisms [21], or embryos [190]. Conventional acoustic trapping utilizes standing waves [182], [191], bulk acoustic waves (BAW) [192], surface acoustic waves (SAW) [21], [193], Gaussian beam [13][91], [194], and Bessel beam based on annular-ring Fresnel lens [82], etc. However, most acoustic tweezers have limited trapping volume, making them unable to trap large particles such as embryos. Also, many acoustic tweezers depend on additional reflectors [182], [191], multiple transducers [195], [196], or multi-element phased arrays [18], [19], which make the system unnecessarily bulky and complex. To address these issues, we aimed to develop single-element microfabricated acoustic tweezers that are able to trap large (mm-sized or sub-mm-sized) particles in liquid environments. In this chapter, two types of such acoustic tweezers will be demonstrated. One type is based on a ring-focusing Fresnel acoustic lens with dual functionality of generating long depth-of-focus focused Bessel-like beam and creating multiple acoustic trapping beams such as bottle beams and Airy-like beams. The other type is based on a multi-foci linear Fresnel acoustic lens for creating long cuboid-shaped trapping beams for large particle trapping. 6.2 Ring-Focusing Fresnel Lens for Long Depth-of-Focus Focusing and Multi- Beam Acoustic Trapping 124 6.2.1 Background and Motivation Narrowly focused acoustic beams with long depth-of-focus can be useful for ultrasound imaging [197] and nondestructive testing (NDT) [198], as it enhances the imaging quality with higher signal-to-noise ratio (SNR). It can also increase treatment volume in ultrasound therapeutics [199], while maintaining fine spatial resolution. To achieve this kind of focused acoustic beams, different methods have been explored. A straightforward method is to change the shape of the output wave front through modifying the surface profile of a transducer [199], [200] or an acoustic lens attached to a flat transducer [66], [201]. However, these devices are usually macro-machined (with limited fabrication accuracy and consistency), or 3D-printed (which is time-consuming and not mass-producible). Another method is to construct thin, planar acoustic meta-surfaces consisting of multilayer periodic stacks of different materials [75] which effectively modulates the transmitted wave front. But the fabrication of these meta-surfaces is nontrivial, since layer thickness control is critical. A third approach is to encode the amplitude/phase distributions of the output waves, such as pulsing each individual elements in a phased array with different time delay and amplitude [197], [198], [202], which requires complicated control with limited acoustic intensity unless extra external power amplifiers are used. Alternatively, without multiple transducer elements, a single-focusing planar Fresnel acoustic lens which can be mass-produced with high-precision microfabrication offers a simple and effective way of focusing ultrasound through selectively allowing in-phase acoustic waves to interfere constructively at the designed focal point [81], but with limited depth-of-focus. To extend the depth-of-focus of acoustic Fresnel lens, in this paper, we have modified the lens design to focus ultrasounds over a ring (instead of a single point), generating bending quasi-Airy beams to create a limited-diffracting, self-healing Bessel-beam-like focused ultrasound beam with long depth-of- 125 focus. Moreover, the same lens generates multiple bottle beams and Airy-beam-shaped “acoustical belts” for effective acoustic trapping of multiple relatively large objects (up to 2 mm in length) immersed in water. Other types of acoustic trapping transducers (or “acoustic tweezers”) have been reported. The ones based on standing waves trap multiple objects in pressure nodal or anti-nodal points with good precision and controllability, but rely on acoustic reflectors [182], [191] or additional transducers [195], or have limited trapping volume [21] (surface-acoustic-wave-based). Another type of acoustic tweezers rely on hydrodynamic forces from the acoustic-field-induced streaming, and offers high throughput [203] but has limited repeatability due to the nonlinear nature of the streaming fluid. Other acoustic tweezers are based on travelling-wave trapping beams such as zeroth-order Bessel beams ([84], [190], [204]), vortex beams (with limited trapping size [18], [205]) and twin-trap beams (with weak vertical trapping force [18], [196]). In comparison, our transducer generates bottle beams with stable, fully three-dimensional trapping forces with minimal field disturbance from the trapped object, along with Airy beams with large trapping zones and the capability of rotating the trapped objects [206]. To generate acoustic bottle beams and Airy beams, transducers based on macro-machined corrugated piston transducer (for Airy beam [67]), phased arrays (for both [18], [19] or for bottle beam [207]) and 3D-printed acoustic holographic lens (for bottle beam [208] or Airy beam [209]) have been reported, with limitations mentioned in the first paragraph of this section. 6.2.2 Device Design The transducer (Figure 6.1a and 6.1d) is built on a 1-mm-thick PZT-5A substrate sandwiched by two overlapping circular nickel electrodes. When 2.32 MHz sinusoidal signal is applied across the electrodes, the PZT will vibrate in its fundamental thickness-mode resonance and effectively 126 generate ultrasound waves, which then pass through a modified Fresnel lens consisting of Parylene-sealed annular-ring air cavities (shiny grey circle and rings in Figure 6.1d) alternating with non-air-cavity ring areas uniformly coated with Parylene (dark grey rings in Figure 6.1d) on the top electrode. The air-cavity rings in the lens almost completely block acoustic waves due to the large acoustic impedance mismatch between air (0.4 kRayl) and solid (over 1 MRayl), whereas the non-air-cavity ring areas allow the waves to pass through. The radii of the ring boundaries of the Fresnel lens are chosen to make the waves arrive in-phase (with a net phase difference less than 180°) at an annular region centered on the central vertical axis with radius FR and height FZ. In other words, in any axisymmetric cross-section of the lens (Fig. 1b), the path-length (Ln) from “focal point” (point of intersection between the focal ring and the axisymmetric cross-sectional plane) to any ring boundary is longer than FZ by integer multiples of the half-wavelength (λ) so that: 𝐿 − 𝐹 = 𝑛 𝜆 2 , 𝑛 = 1,2, ⋯, (6.20) where 𝐿 = 𝑅 + 𝐹 , 𝑛 = 1,2, ⋯. Through solving (1), we can calculate the ring boundary radii as follows. For 𝐹 ≥ 𝑅 , ≥ 0: 𝑅 , = 𝐹 − 𝑛𝜆 × (𝐹 + ), 𝑛 = 1,2, ⋯; (6.21) For 𝑅 ≥ 𝑅 𝑜𝑢𝑡𝑒𝑟 , ≥ 𝐹 : 𝑅 , = 𝐹 + 𝑛𝜆 × (𝐹 + 𝑛𝜆 4 ), 𝑛 = 1,2, ⋯ (6.22) 127 The Rmax is set by the transducer's aperture radius. It is worth noting that (2) and (3) are close to the radii equation for a single-focusing Fresnel lens [81], except that for the ring-focusing Fresnel lens, the ring patterns are shifted radially by FR. Figure 6.1 (a) Cross-sectional schematic (across A-A’ in (c)) of the transducer, illustrating how the ring-focusing air- cavity Fresnel lens is designed to generate long depth-of-focus focal zone and many trapping zones. (b) Axisymmetric cross-sectional schematic (across A-O in (c)) of the transducer, showing how the ring radii are calculated. Top-view diagram (c) and photo (d) of the 2.32-MHz transducer on PZT (brown in (d)) designed for a focal ring with F R = 8 mm (red dashed circle in (c)) and F Z = 12 mm, showing sound-blocking air cavities (black in (b) and (c), shiny grey in (d)) and sound-passing non-air-cavity rings (white in (b) and (c), dark grey in (d)). (a) (b) (c) (d) 128 After being focused at the focal ring, the waves then propagate farther, and arrive at a narrow region along the vertical axis with constructive wave interference to create a narrow Bessel-beam- like focal zone with long depth-of-focus. This process also produces bottle beams and quasi-Airy beams where radiation force toward the inner region exists and thus, particles can be trapped (Figure 6.1a). For the ring-focusing Fresnel lens on this transducer working in water (λ = 638 μm at 2.32 MHz), we choose FR = 8 mm, FZ = 12 mm and Rmax = 2FR = 16 mm. According to calculation, there are eight sound-passing non-air-cavity Fresnel rings with symmetric ring widths with respect to the focal ring, with the middle two rings merged together to form a wider ring (Figure 6.1c and 6.1d). Interestingly, in any axisymmetric cross section, the passing zones on the ring-focusing Fresnel lens (Figure 6.2c) aligns well with the zero-phase zones (Figure 6.2b) of a modified Airy function (Figure 6.2a). This may be the reason why our modified Fresnel lens can generate Airy- like beams (which are a type of bending, nondiffracting, self-healing acoustic beams [67], [206], [207]), even without amplitude modification, which has been shown to be less important than phase modulation for maintaining the unique nature of Airy beams [67]. 129 Figure 6.2 The amplitude (a) and phase (b) of a modified Airy function versus lateral distance. (c) The transmission function of the ring-focusing Fresnel lens, which closely resembles the Airy phase pattern shown in (b). 6.2.3 Simulation of Acoustic Pressure To verify the design, simulations based on the finite-element method (FEM) are carried out using COMSOL Multiphysics. The simulations are done in frequency domain at 2.32 MHz with two-dimensional (2D) axisymmetry defined to save computation time and memory. We first simulate the relative acoustic pressure generated by the ring-focusing transducer in water. To demonstrate how the focal zone and trapping zones are generated, we simulate the contributions of the outer (with radii larger than FR which is 8 mm) and inner (with radii less than 8 mm) Fresnel rings. When only the outer four Fresnel rings are activated, we observe many quasi- Airy beams bending inwards radially towards the central axis, some of which arrive in phase to create a focal zone with long depth-of-focus (Figure 6.3a). Similarly, when only the inner four ((a) (b) (c) 130 Fresnel rings are activated, we see Airy-like beams generated, but with most beams bending outwards (Figure 6.3b). When all the Fresnel rings on the transducer are actuated, these two aforementioned effects combine (Figure 6.3c). First, focused ultrasound is generated in the designed focal ring region and also in the focal zone centered at Z = 31.33 mm with depth-of-focus of 9.838 mm and focal diameter of 573 μm (Figure 6.3c). From the lateral beam profiles at different axial positions within the focal zone (Figure 6.3d), we see Bessel-beam-like pressure distributions. For comparison, we also simulate the acoustic pressure from a single-focusing Fresnel half-wave-band transducer with similar aperture size (31.9 mm) and focal length (31.33 mm), when the vibration amplitude from the PZT surface is the same (Figure 6.3e). This transducer generates a single focal zone with depth- of-depth of 5.018 mm and focal diameter of 727 μm. From these we see that the ring-focusing transducer produces depth-of-focus almost twice that of a typical Fresnel lens, with thinner beam width, albeit a lower peak pressure (by a factor of 1.73). For low-intensity applications (such as particle trapping and neural stimulation), the lower peak pressure is not a concern, and can be compensated by applying a higher driving voltage. Second, the acoustic beams interfere destructively on some parts of the central axis to create low-pressure zones, while also interfere constructively in their surrounding regions, thus generating multiple bottle beams on the central axis (Figure 6.3c and 6.3f). By simulating the pressure amplitude and phase near a bottle beam at Z = 14.85 mm, we clearly see high-pressure regions surrounding low-pressure regions (Figure 6.3g and 6.3i) with a phase singularity at the center of the beam (Figure 6.3h and 6.3j). In a similar fashion, in the off-axis regions, Airy-beam-shaped “acoustical belts” characterized by high- pressure beams embracing low-pressure beams are also generated (Figure 6.3c and 6.3f). 131 According to analysis in the next subsection, these two acoustic beams can generate acoustic radiation forces towards the beam center, which can effectively trap particles. In the actual trapping experiments where there is acoustic reflection from the water surface (58 mm above the transducer), the acoustic pressure distribution remains similar, but the pressure amplitude (especially in acoustical belt areas) becomes higher due to half-wavelength resonances (Figure 6.3k), which increase the trapping force in these areas. Figure 6.3 Simulated acoustic pressure amplitude in XZ plane for the ring-focusing transducer (a) with outer four Fresnel rings actuated to create inwards-bending Airy-like beams; (b) with inner four Fresnel rings actuated to create outwards-bending Airy-like beams; (c) with all rings actuated. (d) Bessel-beam-like radial pressure amplitude distributions (red) within the focal zone at Z = 29.3 mm (upper) and Z = 34.2 mm (lower) compared with scaled Bessel functions of the first kind (black). (e) Simulated acoustic pressure amplitude in XZ plane for a normal Fresnel lens with similar focal length and aperture size. (f) Relative pressure isosurfaces of 0.15 showing multiple bottle beams on the central axis and many “acoustical belts” in off-axis areas; (g)-(j) Acoustic pressure amplitude (in XZ plane (g) and in XY plane (i)) and phase (in XZ plane (h) and in XY plane (j)) distribution for a bottle beam located at Z = 14.85 mm on the central axis (highlighted in (f)). (k) Simulated acoustic pressure amplitude in XZ plane for the ring-focusing transducer during trapping experiments with acoustic reflection from the water surface (58.0 mm above the transducer surface). The pressure values in all figures are normalized to the values in (c). 132 The interaction between the generated acoustic field and solid objects is also studied. To demonstrate the self-healing property of the quasi-Bessel and quasi-Airy beams, we simulate the acoustic pressure distribution with the presence of two sound-blocking copper rings placed at Z = 2 mm and 6 mm (with ring width of 1.5 mm and thickness of 1 mm, whose inner radii are 2 mm and 12 mm, respectively, shown in Figure 6.4a). Compared to the case with no obstruction (Figure 6.3c), the beam pattern remains similar with less than 20% loss in peak acoustic pressure. To demonstrate the trapping beams’ ability to circumvent obstacles placed at trapping zones, we simulate another scenario where two 1-mm-diameter polyethylene (PE) microspheres are placed in the center of two bottle beams at Z = 14.85 mm and 24.80 mm on the central axis, and one PE ring with a square 0.7-mm-side-length cross section and inner radius of 5.35 mm is placed in an Airy acoustical belt region at Z = 21.95 mm. The simulated pressure pattern with the obstacles (Figure 6.4b) is similar to that for a case with no obstacles (Figure 6.3c). Thus, with multiple self- healing and obstacle-circumventing bottle beams and acoustical belts, the transducer is capable of trapping multiple objects simultaneously. Figure 6.4 Simulated acoustic pressure amplitude (without reflection, normalized to the values in Figure 6.3c) in XZ plane for the ring-focusing transducer (a) with two copper rings blocking some acoustic waves; (b) with two PE microspheres and one PE ring in potential trapping regions (axes rescaled to maintain circular (and square) cross sections of the objects). (a) (b) 133 6.2.4 Simulation of Acoustic Radiation Force (ARF) To demonstrate how these acoustic beams can trap particles, we first simulate the acoustic radiation force (ARF) acting on a 70-μm-diameter PE microsphere (with density of 1.130 g/cm 3 and acoustic velocity of 2460 m/s) immersed in water (with reflection from the water surface 58 mm above the transducer), with acoustic pressure values in Fig. 3k normalized (based on measurement data in Figure 6.6, with 35 Vpp applied on transducer) to have a peak pressure of 0.226 MPa. The acoustic radiation potential is calculated from the simulated acoustic pressure and velocity fields along with the medium and particle properties, and then the ARF is calculated by taking the negative spatial derivatives of the radiation potential (as indicated by Eqs. 27 in [210]). From the simulation results, we clearly see ARF pointing from high-potential (also high-pressure) shells of bottle beams at Z = 21.2 mm (Figure 6.5a) and 25.3 mm (Figure 6.5b), towards their center low-potential (low-pressure) regions. Similarly, in Airy-shaped acoustical belt regions (Figure 6.5c), ARF points towards large low-potential zones, and in some regions, forms a vortex pattern which is capable of rotating the trapped object. For a large microsphere whose diameter is comparable to or larger than the wavelength, ARF needs to be calculated through integrating the second-order momentum fluxes generated by the first-order pressure and velocity fields over the microsphere surface [211]. Using Equations 3–5 in [211], we simulate the vertical ARF exerted on a 1-mm-diameter PE microsphere with density of 1.025 g/cm 3 centered on different positions of the vertical axis, with 35 V pp applied to the transducer. The simulated vertical ARF ranges from 8.7 nN to 2.2 μN (Figure 6.5d). To lift the microspheres, the gravitational force (-5.26 μN) and buoyant force (5.13 μN) need to be balanced, requiring a vertical ARF of 0.13 μN. In Figure 6.5d, we identify six positions with such ARF and 134 with a negative spatial force gradient (so that the force is restoring) as potential stable trapping positions. Comparing the vertical ARF with the pressure amplitude along the central axis, we notice that the vertical ARF is almost proportional to the pressure amplitude (Figure 6.5d), and the ARF needed for lifting the microspheres corresponds to the pressure amplitude of 0.03–0.06 MPa. By visualizing only the regions where the pressure amplitude is larger than 0.045 MPa (Figure 6.5e), we find many potential trapping positions, most of which in off-axis regions are due to the increased pressure amplitude resulted from the acoustic reflection from the water surface 58 mm above the transducer (Figure 6.3k). With reduced reflection from the water surface, the number of the trapping zones will be lower. 135 Figure 6.5 Simulated acoustic radiation potential (colorbar unit: 10 -14 Joules) and acoustic radiation force (ARF, white arrows) for 70-μm-diameter PE microspheres, showing fully three-dimensional trapping forces towards the center of bottle beams at Z = 21.2 mm (a) and 25.3 mm (b). (c) Same plot for off-axis Airy-shaped acoustical belt regions, showing ARF for particle trapping and rotation (with arrow length logarithmically normalized). (d) Simulated vertical ARF exerted on a 1-mm-diameter PE microsphere centered at different positions of the central axis (red) and the pressure amplitude (black), along with the required vertical lifting force (blue dashed line) and potential stable trapping positions (blue circles). (e) Simulated acoustic pressure amplitude showing regions with pressure amplitude higher than 0.045 MPa for potential trapping capability (colorbar unit: MPa). In all figures, the acoustic pressure distribution is normalized from Fig. 3k with a peak pressure of 0.226 MPa, which corresponds to the case where 35 V pp is applied on the transducer (according to measurement data in Figure 6.6). 6.2.5 Measurement of Acoustic Pressure The transducer is microfabricated according to steps described in [81]. After fabrication, we measure the acoustic pressure by mechanically scanning a hydrophone (HGL-0085, Onda Corp.) aligned to the center of the transducer along its central vertical axis. Immersed in water, the transducer is driven at 150 Vpp with 2.32 MHz sinusoidal pulsed signals, and the measured peak 136 pressure is 0.8 MPa with depth-of-focus of 9.9 mm (15.5λ), in agreement with the simulation (Figure 6.6). The minor difference between simulation and measurement is likely due to the slight misalignment between the scanning axis of the hydrophone and the transducer's central axis. Figure 6.6 Measurement (red, with 150 V pp applied on the transducer) and normalized simulation (black) of acoustic pressure along the central axis. 6.2.6 Trapping of Polyethylene (PE) Microspheres As shown in Figure 6.7a, the transducer (facing up) is immersed in water in a glass beaker (150 mm in diameter to eliminate reflection from the sidewalls) that is placed on a laser-machined acrylic holder attached to a precision 5-axis manually-movable stage. The PE microspheres are pre-wet in 0.1% Tween-80 solution (a surfactant to overcome their hydrophobicity, purchased from Sigma-Aldrich, Inc.) and then slowly released into water from a glass pipette (Figure 6.7b) attached to a precision manual syringe pump (VWR International, LLC.). Unless specified otherwise, the microspheres have diameter of 1 mm with density of 1.025 g/cm 3 . During trapping experiments, a function generator (AFG3252, Tektronix, Inc.) generates 2.32 MHz continuous 137 sinusoidal signal, which gets amplified by a power amplifier (75A250, Amplifier Research Corp.) and delivered to the transducer. Figure 6.7 (a) Photo of the experimental set-up for all the trapping experiments. Side-view photos showing (b) trapping of three 1-mm-diameter PE microspheres (1.025 g/cm 3 density) through picking and releasing them one by one into water with a pipette; (c) trapping process of five out of six 1-mm-diameter PE microspheres released at the same time from water surface, including two sticking with each other; (d) two 1-mm-diameter PE microspheres (glued together) rotating clockwise while being trapped. The transducer is driven with continuous sinusoidal signals of 35 V pp in (b) and 30 V pp in (c) and (d). We have successfully trapped multiple microspheres in two ways. First, with 35 V pp applied on the transducer, we put microspheres in different trapping zones by picking and releasing them with a pipette in sequence (one microsphere at a time), as shown in Figure 6.7b. With the self- healing property of the acoustic trapping beams and strong trapping force, the disturbance during and after the pick-and-release procedure does not affect the trapping efficiency of the previously trapped microspheres. Comparing the third and fourth photo in Figure 6.7b, we see that while falling down, microsphere #2 is pushed laterally towards the center and then get trapped there. Second, as shown in Figure 6.7c, we release six microspheres into water simultaneously, and five 138 of them get firmly trapped as they fall down due to gravity, when the transducer is driven with 30 Vpp. Two of the five trapped microspheres (1.07 mg in weight) stick together throughout the whole process, lifted by buoyance and a vertical lifting force of 0.257 μN from the transducer. Sometimes, when trapped in the acoustical belts, particles (especially of odd shapes) are found rotating. For example, with 30 Vpp applied on the transducer, two 1-mm-diameter microspheres glued together are trapped while rotating around the same position at an angular speed of 8.57 °/ms (Figure 6.7d). While one microsphere is firmly trapped, we move the transducer (driven with 30 V pp) at an average speed of 60 μm/s for 2.5 mm (Figure 6.8a). The trapped microsphere closely follows the transducer’s lateral movement throughout the process. However, when we move the transducer vertically by merely tens of micrometers at about the same speed, the trapped microsphere follows the transducer’s movement a little, and then falls off the trapping zones. This suggests that in some trapping regions (especially the off-axis regions), the variation in vertical ARF is larger than the lateral one. In another set of experiments, we first let the microspheres fall on the transducer surface; while the transducer is being actuated (30 Vpp), we blow with water flow from a pipette to get the microspheres to float above the transducer and move around fast; and then we stop the blow to observe the trapping of microspheres. In this way, we successfully trap many 350-μm-diameter PE microspheres with density of 1.13 g/cm 3 (Figure 6.8b) and six 1-mm-diameter PE microspheres (Figure 6.8c), despite the initially large fluid disturbance and fast particle movement. 139 Figure 6.8 Side-view photos showing (a) positions of a trapped 1-mm-diameter microsphere (1.025 g/cm 3 density) before and after the transducer is moved to the right for 2.5 mm at an average speed of 60 μm/s. The microsphere (green dash line) remains trapped while following transducers' movement (marked by the left edge of the acrylic holder below it, red dash line). Side-view photos showing (b) trapping process of multiple 350-μm-diameter PE microspheres (1.13 g/cm 3 density) after blowing with a pipette to make them float above transducer surface; (c) trapping of six 1-mm-diameter PE microspheres using the same technique as in (b). The transducer is driven with continuous sinusoidal signals of 30 V pp in all figures. 6.2.7 Summary We have designed and microfabricated a novel transducer based on a ring-focusing air-cavity Fresnel acoustic lens, and successfully demonstrated its ability to generate a Bessel-beam-like focal zone with a long focal depth of 9.9 mm which is 15.5 times the wavelength at 2.32 MHz. Furthermore, with the same transducer, we present a brand-new way of generating bottle beams and Airy-like “acoustical belts” which firmly trap multiple large (0.35–1.00 mm in diameter) polyethylene microspheres with density larger than that of water, with the capability of rotating 140 the trapped particle. To our knowledge, it is the first demonstration of generating such beams with a single-element, microfabricated acoustic lens. 6.3 Acoustic Tweezers Based on Multi-Foci Linear Fresnel Lens for Large Volume Particle Trapping Traditionally, the trapping zones of acoustic tweezers are designed to be relatively small volume (largest diameter  1 mm) to achieve high-resolution, making them not suitable for trapping large objects such as embryos, larvae, etc. This paper describes a new type of acoustic tweezers based on a multi-foci linear Fresnel lens with rectangular-shaped air cavities to create long cuboid shaped trapping zones for large particle trapping. 6.3.1 Device Design Over the top electrode of a 1-mm-thick PZT (whose thickness-mode fundamental resonant frequency is 2.32 MHz), we form seven pairs of rectangular air cavities sealed in Parylene, which also serves as electrical insulation and acoustic matching layer (Figure 6.9). The widths of each symmetric air-cavity pair (Figure 6.9b) are designed into Fresnel half-wavelength band (FHWB) sources, with the distance from the n th band boundary to the center line given by (2.2) [78]. On top of the transducer, the air cavities (white areas in Figure 6.9b) cover all regions from which acoustic waves will destructively interfere at the designed focal lengths (R n<R<Rn+1, n=1,3,5,…), so that due to acoustic impedance mismatch between air (only 0.4 kRayl) and liquid/solid (over 1 MRayl), almost all destructive waves may be reflected back. Other waves in non-air-cavity regions (Rn<R<Rn+1, n=0,2,4,…) pass through the Parylene layer of the lens, to interfere constructively at designed focal lengths. 141 The widths of each symmetric air-cavity pair from center to outside are designed for 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, and 7.3 mm focal lengths, which emulates the design of an axicon lens where acoustic waves farther away from device center will be focused at a longer distance, so that Bessel beam region (orange region in Figure 6.9a) where the radiation pressure is lower than the surrounding regions (so that particle could be trapped) may be created. The length of the air cavities was chosen to be 8.0 mm, but can be longer for longer focal zone. The device dimensions are listed in Table 6.1. We have simulated the sound pressure map at XZ (Figure 6.10a), YZ (Figure 6.10b) central planes perpendicular to the device, as well as the XY plane at Z = 6.75 mm (Figure 6.10c) where a trapping zone is located, with COMSOL Multiphysics software using Pressure Acoustic model in frequency domain. Simulation results suggested three major trapping zones at 1.70, 6.75 and 11.65 mm away from the device center, along with other trapping zones on the sides. The major trapping zones (about 8 × 0.6 × 1.2 mm 3 in size) have a very low acoustic intensity zone surrounded by high intensity regions, thus having a potential well to trap particle. One issue with using rectangular bands (rather than annular Fresnel rings) for focusing is that there are quite a few satellite focal points, which are not necessarily bad for acoustic tweezing, since they allow trapping of multiple particles at the same time. 142 Figure 6.9 (a) Side-view schematic of the acoustic tweezer with multi-foci linear Fresnel lens based on air cavities, showing how air cavities prevent destructive waves from reaching the designed focal point; also indicated are multiple focal lengths from multiple pairs of Fresnel air-cavity lens and the resulting Bessel beam region with negative axial radiation force for particle trapping; (b) Top view of the Fresnel lens design, showing seven pairs of rectangular air cavities (in white) for blocking destructive acoustic waves and non-air-cavity regions (in black) to let pass constructive waves. (c) Photo of the front side of the fabricated acoustic tweezers, showing seven pairs of Fresnel bands (in dark grey areas where Parylene covers top Nickel electrode) separated by seven air-cavity bands (light grey areas) with filled release holes (three circles on each rectangular air cavity). Table 6.1 Dimensions of the Acoustic Tweezers. PZT thickness Transducer size Active-area length (a) (b) (c) 143 1.0 mm 32 x 16 mm 2 18 mm Active-area width Parylene thickness Air-cavity height 8 mm 21 μm 3.5 μm Figure 6.10 FEM-simulated absolute sound pressure in (a) central XZ and (b) central YZ trapping planes perpendicular to the device, and (c) in a XY trapping plane at Z = 6.75 mm. 6.3.2 Experimental Results The transducer is fabricated using the same technique mentioned in [81]. The acoustic pressure field produced by the fabricated tweezers was measured vertically along the center line with a commercial hydrophone (Onda HGL-0085) fixed onto a manual movable stage with a) b) (c) 144 precision of 0.1 mm. The measured pressure, when the tweezers was driven with pulsed sinusoidal signal at 2.32 MHz with 3.88 μs pulse width, is shown in Figure 6.11, which is in good agreement with simulated data, suggesting two trapping zones near 6.8 mm and 11.6 mm where there are sharp pressure (and acoustic potential energy) dips. When the hydrophone was placed less than 5.5 mm from the transducer, electromagnetic interference affected the acoustic signal so much that we could not measure the pressure accurately. Figure 6.11 Hydrophone measurement (red) and COMSOL simulation (black) of the relative acoustic pressure along the central vertical axis vs the distance from the tweezers. Trapping experiments were done with various objects using the set-up shown in Figure 6.12. During the experiment, the acoustic tweezers was immersed into DI water in a beaker, with water level adjusted to the height of trapping zones, and then the object to be trapped was put on the water surface above the device, which was driven with continuous 2.32 MHz sinusoidal signals. The trapping process was recorded with a digital microscope connected to a computer. 145 Figure 6.12 Experiment set-up for characterizing particle trapping by the acoustic tweezers (not in scale). In the first experiment, Delrin plastic piece (3.16×0.9×0.53 mm 3 and 1.99 mg) was put on a water surface above the device (Fig. 7a). Before the device was actuated, the piece moved freely around on the water surface. When the device was actuated, the Delrin piece was immediately attracted to the trapping zone and held in place at 1.70, 6.75, and 11.65 mm tapping planes, with a minimum required voltage of 20, 25, and 40 V pp, respectively. A typical trapping process is shown in Figure 6.13, where the Delrin piece is trapped, and a nearby microsphere that is not in the trapping zone moves away. The trapping is stable, and lasts as long as the device is actuated and there is no major disturbance on the water surface. 146 Figure 6.13 Photos showing a trapping process: (a) before trapping a large Delrin plastic piece, (b) with the piece trapped, (c) with the piece still trapped, as a nearby microsphere (that is not trapped) goes away. In another experiment, a silicon chip of similar size and weight (3.5×0.5×0.5 mm 3 and 1.90 mg) was also firmly trapped when the water level was at all three trapping zones (Figure 6.14a to 6.14c). We also demonstrated stable trapping of a larger silicon chip (5.3×0.67×0.51 mm 3 and 3.98 mg), which could only be trapped at 1.7 mm trapping zone with 25 V pp applied, along with two 500-μm-diameter Polyethylene microspheres (Figure 6.14d and 6.14e). (a) (b) (c) 147 Figure 6.14 (a)(b)Photos showing trapping of a 1.9 mg rectangular silicon chip while un-trapped microspheres moving around. Photos showing trapping process of a 3.9 mg rectangular silicon chip along with two microspheres: (c) before trapping, (d) trapping of a silicon chip and two microspheres, (e) an un-trapped microsphere goes away as three objects still trapped. 6.3.3 Summary In summary, we designed and fabricated a novel acoustic tweezers based on multi-foci linear Fresnel air-cavity lens to create cuboid Bessel beam regions of large volume with negative axial radiation force, which was confirmed with simulation and hydrophone measurement. The device was able to trap plastic and silicon chips with up to 5.3×0.67×0.51 mm 3 in size and 3.98 mg in weight. (a) (b) (c) (d) (e) 148 7Chapter 7 Piezoelectric Micromachined Ultrasonic Transducer (PMUT) Based on Dome-Shaped Diaphragm 7.1 Introduction Due to their simple structure, low power consumption, low cost, and good sensitivity, piezoelectric ultrasonic transducers have been used in many applications such as automobile collision avoidance [212], nondestructive evaluation (NDE) [35], medical imaging [213], flow rate measurement [214], soldering [46], therapy [215], tissue ablation [133], etc. Compared with traditional ultrasonic transducers based on bulk piezoelectric substrates, ultrasonic transducers based on micromachined diaphragms have the following attractive features: low cost due to batch fabrication, smaller size and weight that makes dense array operation possible, compatibility with CMOS IC (complementary metal-oxide-semiconductor integrated circuits) fabrication, and long- term reliability (if the diaphragms are actuated piezoelectrically). Among various micromachined ultrasonic transducers, capacitive (CMUT) [56]and piezoelectric (PMUT) micromachined ultrasonic transducers [64] are most commonly used. A typical CMUT structure involves an air gap between two electrodes, and the air gap spacing will change when acoustic pressure (receiving mode) or electrostatic force (transmitting mode) is applied. CMUT has the merit of high sensitivity, but needs external DC bias to operate, and suffers from resonant frequency shift over time caused by dielectric charging effects, as well as dynamic range limitation caused by pull-in effect (top thin electrode collapsing when air gap is too small). PMUTs, on the other hand, are based on the deflection of a thin-film diaphragm due to piezoelectric effect when acoustic pressure (receiving mode, acoustic pressure-induced 149 mechanical deflection causes polarization change in the film) or voltage (transmitting mode, electric voltage piezoelectrically causes mechanical strain) is applied. Compared to CMUT, PMUT does not require external bias voltage, is easier to fabricate since no air gap is involved, and has long-term reliability and a wider dynamic range. However, most PMUTs are based on flat, four- edge-clamped diaphragms, which suffer from poor sound production and sensitivity, as the diaphragm displacement is small due to residual stress and geometrically constraining boundary condition. To solve these problems, we designed a microfabricated ultrasonic transducer which has (1) a hemispherical dome-shaped diaphragm that readily releases residual stress through its shape change and transforms in-plane diaphragm vibration to large radial deflection, (2) a thin, flexible, flat diaphragm supporting the dome diaphragm at its center and (3) embedded piezoelectric elements actuating the dome diaphragm. 7.2 Device Overview The schematic cross-section of the dome-diaphragm transducer is shown in Figure 7.1a, while Table 7.1 shows all the device dimensions. The 2.2-μm-thick silicon nitride dome-shaped diaphragm with a diameter of 3.8 mm and depth of 0.85 mm is supported at the center of a 4.4×4.4 mm 2 flat square diaphragm of the same material with four edges clamped on a 2-mm-thick silicon substrate. On top of the diaphragm, patterned zinc oxide (ZnO) piezoelectric layer and silicon nitride encapsulation layer are sandwiched by the blanket-deposited aluminum bottom electrode and the patterned titanium/gold top electrode, forming eight piezoelectric multi-layer stacks at 0°, 45°, 90°, 135°, 180°, 225°, 270°, 315° directions, respectively. Having piezoelectric stacks at eight directions allows individual control of each driving position to maximize the sound output at different vibration modes where the phase of each position may vary. A novel dicing-based front- 150 back alignment method is utilized which will be described in detail in the fabrication section. The photos of the fabricated device are shown in Figure 7.1b to 7.1d. Figure 7.1 (a) Schematic cross-section view of the piezoelectric dome-shaped diaphragm transducer, before breaking the wafer into individual chips. The pre-diced grooves were used for front-back alignment as well as post-process wafer breaking. Photos of the front side of the transducer taken with a camera (b) and with a scanning electron microscope (SEM). (d) Photo of the back side of the transducer. Table 7.1 Physical Dimensions of the Transducer Wafer thickness Chip edge length Bottom electrode edge length 2.0 mm 12.5 mm 10.0 mm Square edge length Dome diameter Dome depth 4.4 mm 3.8 mm 0.85 mm LPCVD SiN diaphragm thickness PECVD SiN thickness Top Ti electrode thickness 2.2 μm 400 nm 20 nm (a) (b) (c) (d) 151 Bottom Al electrode thickness ZnO thickness Top Au electrode thickness 400 nm 500 nm 380 nm 7.3 Fabrication and Packaging To achieve a large and deep dome-shaped diaphragm, a thick silicon substrate has to be used, which makes the front-back alignment challenging. Also, due to the fragility of the dome diaphragm, dicing the wafer after diaphragm formation could cause damage and reduce yield. To solve this problem, a novel dicing-based technique was invented. As the first step of fabrication, we partially diced both sides of a silicon wafer at a fixed interval of the chip edge length, aligning with the two orthogonal wafer flats (red lines in Figure 7.2a) of the (100) p-type wafer, to form perpendicular dicing lines (black lines in Figure 7.2a) on both sides which create alignment crosses as well as pre-defined grooves for chip separation after the fabrication process. This simple, photolithography-free technique provides front-to-back alignment error less than 50 μm (Figure 7.2b) and also allows easy, damage-free wafer breaking after fabrication (Figure 7.4a). Figure 7.2 (a) Diagram showing the dicing lines (black) on one side of the wafer and the two wafer flats (red) for dicing alignment; (b) Cross-section photo of a diced 2-mm thick wafer, showing front-to-back alignment error of 48 μm. za) (b) 152 The next major step after dicing is to form a large hemispherical dome on a silicon wafer by isotropic silicon etching with a 65 μm thick tape as an etch mask (Figure 7.3a to 7.3d). First, 1 μm thick silicon nitride is deposited by low-pressure chemical vapor deposition (LPCVD) over a 2- mm thick silicon wafer to prevent contamination from the adhesive residues of the tape (used as an etch mask during silicon isotropic etching). Then a 65 m thick tape is applied and patterned (along with underlying silicon nitride) by reactive ion etching (RIE) with aluminum as an etch mask during the RIE. With the patterned tape acting as an etch mask, the silicon substrate is etched in HNA (49% HF: 70% HNO3: 99.7% Acetic acid = 4:7:11) at 30 °C in a Teflon beaker with no stirring. This specific etching condition is inspired and modified from [216] to achieve the best isotropicity and large hemispherical etch cavity with a diameter of 3.8 mm is achieved. We can further enlarge the dome size by simply having a large circular opening. After forming the spherical etch cavity, the tape is detached in acetone and isopropyl alcohol (IPA) solution with ultrasonic treatment, followed by removal of the silicon nitride with hot phosphoric acid and wafer cleaning with “piranha” (98% H 2SO4:30% H2O2 = 3:1) and buffered oxide etchant (40% NH4F: 49% HF = 6:1) (Figure 7.3d). After depositing a 2.2-μm-thick LPCVD silicon nitride and patterning it on the wafer backside, we release the whole diaphragm by KOH etching from the backside, obtaining a square flat diaphragm around the dome diaphragm (Figure 7.3e and 7.3f). The last major step is to deposit all the layers needed for piezoelectric actuation (Figure 7.3f to 7.3h). After the bottom aluminum electrode is blanket-deposited by sputtering, patterned silicon nitride for encapsulation is deposited by plasma-enhanced chemical vapor deposition (PECVD) through a silicon shadow-mask fabricated according to the steps described in [217]. Then zinc oxide (ZnO) is deposited by sputtering through another shadow mask, followed by electron beam 153 evaporation deposition of a top electrode made of titanium and gold through the same shadow mask. Figure 7.3 Brief microfabrication process: (a) Dice both sides, deposit SiN, apply tape, and deposit Al; (b) Pattern Al, tape and SiN; (c) Isotropically etch Si in HNA etchant; (d) Remove tape and SiN; (e) Deposit LPCVD SiN; (f) Release the diaphragm through KOH etching from backside and deposit Al bottom electrode; (g) Deposit PECVD SiN through a silicon shadow mask; (h) Deposit ZnO and top Ti/Au electrodes with another shadow mask. After fabrication, the 3-inch wafer is broken into 16 chips with 12.5 mm edge length (Figure 7.4a). The pre-defined dicing lines are instrumental in the die separation as no diaphragm, although it is relatively large and fragile (4.4 x 4.4 mm 2 square diaphragm supporting a 3.8-mm-diameter piezoelectric dome diaphragm), is damaged during the process. Then the chips are mounted to 154 surface-mount technology (SMT) adaptors with double-sided tape and electrical connections were made through gold wire-bonding (Figure 7.4b). Figure 7.4 Photos of (a) diced chips with the fabricated dome-diaphragm transducers from a 3-inch wafer; (b) a chip mounted and wire-bonded onto SMT adaptor. 7.4 Experimental Results 7.4.1 Measurement of Sound Output Using the setup shown in Figure 7.5, we tested the output sound pressure level of our fabricated transducers with various drive voltages (controlled by a function generator and operational amplifier circuit) and at different distances (controlled by a micromanipulator) with a precision microphone (G.R.A.S 40AO) located in the open field. The measured sound data were collected by an oscilloscope and processed by a computer. (a) (b) 155 Figure 7.5 Setup for measuring the sound pressure level produced by the dome-diaphragm transducer. The output sound pressure level (SPL) versus frequency measured at 5 mm away with a drive voltage of 30 Vpp is shown in Figure 7.6a, from which we see two resonant peaks at 18.5 kHz and 30.7 kHz with the sound pressure levels of 88.21 dB and 87.54 dB, respectively. Although the measurement is stopped at 40 kHz due to the upper frequency limit of our microphone, it is reasonable to believe that our transducer can produce ultrasonic output at an even higher frequency. Figure 7.6b shows the highly linear relationship between the measured sound pressure (in Pa) and the drive voltage. The sound output at 18.5 kHz decreases as the distance between the transducer and the pick-up point is increased from 5 mm to 45 mm (Figure 7.6c). The sound output pressure drops by a factor of about 6, as the distance increases by a factor of 9, which indicates that the sound is directional. 156 Figure 7.6 (a) Measured sound pressure level (SPL) versus frequency over 10 kHz – 40 kHz range at 5mm away from the transducer with 30 V pp drive voltage. (b) Measured sound pressure versus drive voltage at 18.5 kHz, 5 mm away; (c) measured sound pressure level at different distances from 5 mm to 45 mm between the microphone and the transducer at 18.5 kHz, with 30 V pp drive voltage. 7.4.2 Increasing Sound Output through Structure Modification To further enhance the sound pressure level, we used laser-cutting, as a post-fabrication step, to turn the square diaphragm into a cantilever-shaped one (Figure 7.7a and 7.7b). By turning a square diaphragm (with its four edges clamped) into a cantilever-like diaphragm (with its three edges free), we can increase the vibration amplitude of the diaphragm substantially, since the diaphragm is no longer bound by the deflection-limiting boundary conditions. As long as the gap between the cantilever-like diaphragm and the surrounding substrate is small enough to avoid the pressure equalization effect (which depends on the frequency of the sound wave), the sound pressure also is expected to improve substantially by the post-fabrication step. (a) (b) (c) 157 The sound output pressure levels versus frequency for the cantilever-like-diaphragm and unmodified-diaphragm transducers measured at the same condition (5mm away, with 30 V pp drive voltage) are shown in Figure 7.7c for comparison. From this graph, we can see that the maximum SPL from the cantilever device is 95.12 dB at 20.4 kHz, which is 6.91 dB larger than that of the initial transducer. Also, there is a plateau of SPL larger than 80 dB at 19.3 kHz ~ 24.5 kHz. According to Figure 7.7d, the linearity of this modified transducer is also very good. Figure 7.7 Diagram (a) and photo (b) of the laser-cutted cantilever diaphragm. (c) Measured sound pressure level (SPL) of the cantilever-like-diaphragm transducer (red line) and unmodified-diaphragm transducer (blue line) versus frequency over 10 kHz – 40 kHz range at 5mm away from the transducer with 30 V pp drive voltage. (d) Measured sound pressure of the cantilever-diaphragm transducer versus drive voltage at 20.4 kHz, 5 mm away. 7.4.3 Increasing Sound Output through Phase Control (a) (b) (c) (d) 158 Another way of increasing sound output level without modifying the dome structure is to apply signals with different phases on top electrodes according to the vibration phase in the specific resonant mode at each electrode location. First, the vibration modes were studied with LDDM (Laser Doppler Displacement Meter), an electro-optical device that directs a laser beam onto the vibration surface, and then converts the vibration amplitude and phase on the area where the beam is focused into electrical signals from the Doppler shift of the reflected laser beam frequency. Although the diameter of the focused laser dot is small (~500 μm), by probing multiple areas of the dome diaphragm, we were able to get enough information of vibration amplitude and phase which, when combined with simulation, will tell us what the resonant modes are. From LDDM measurement we found that at 18.5 kHz, the whole dome diaphragm is vibrating with the same phase and similar amplitude, suggesting that the whole dome vibrates up and down (“loudspeaker mode”); while at 30.7 kHz, there is 180° difference in vibration phase between one half of the dome and the other, suggesting that one half of the dome moving up and the other half moving down alternately (“rocking mode”). To verify, simulation of resonant modes was carried out with COMSOL Multiphysics software, and the dome diaphragm was built based on the Solid Mechanics Module (Figure 7.8). Since each layer in the piezoelectric multilayer stacks (top electrode/ZnO/PECVD SiN) is too thin to render a reasonable computation time, they were incorporated into one effective layer whose thickness equals the total thickness of the stack and whose mechanical properties were calculated as effective properties shown in Table 7.2 as an estimation ( Effective value = ∑( × ) . Similarly, LPCVD Silicon Nitride was incorporated with the aluminum bottom electrode as an effective layer. In the simulation, the properties of PECVD 159 Silicon Nitride and LPCVD Silicon Nitride were found in [218] and [219], respectively, while other properties were from COMSOL material database. Figure 7.8 COMSOL model of the dome-diaphragm structure with top electrode multilayer stacks (incorporated as one layer). Table 7.2 Material Properties Used in Simulation Layer Material Young’s modulus (GPa) Poisson’s ratio Density(kg/m 3 ) Layer Thickness (nm) Multilayer stacks Au 70 0.44 19300 380 Ti 116 0.321 4506 20 ZnO 210 0.33 5676 500 PECVD SiN 160 0.253 2500 400 Effective Layer 132.9 0.341 7260 1700 Dome diaphragm Al 70 0.35 2700 400 LPCVD SiN 290 0.27 3000 2200 Effective 256 0.282 2950 2600 160 Layer From the simulation, we confirmed that the first two resonant modes, although happening at slightly different frequencies than measured ones (15.1 and 19.2 kHz simulated vs. 18.5 kHz and 30.7 kHz measured) due to approximation of layer properties, are truly loudspeaker mode and rocking mode, respectively. Figure 7.9 COMSOL simulation of (a) side view and (b) top view of the first resonant vibration mode (loudspeaker mode); (c) side view and (d) top view of the second resonant vibration mode (rocking mode). The above discoveries suggest that in loudspeaker mode, to maximize sound output, the phase of applied signals on all top electrodes should be the same; while in rocking mode, we should have a 180° phase difference between signals applied on one half of the electrodes and those applied on the other half. This was confirmed by experiments: in loudspeaker mode, the sound output is proportional to the number of actuated electrodes and reaches its maximum when all electrodes were driven with signals of the same phase. In rocking mode, having a 180° phase difference (a) b) (c) (d) 161 between signals applied on electrodes at two different halves resulted in 50% higher output sound pressure (3.5 dB higher in sound pressure level) than the case when signals applied all have the same phase. 7.5 Summary The design and microfabrication process of an ultrasonic transducer based on a dome-shaped diaphragm supported at the center of a square diaphragm with ZnO piezoelectric layers on top has been demonstrated. In fabrication, we invented a novel dicing-based technique for front-back alignment and damage-free post-process chip separation on thick silicon wafers which can ensure less than 50 μm alignment error and 100% yield. Shadow masks are used for patterned layer deposition on the 3D diaphragm with good step coverage. The maximum sound pressure level (SPL) tested at 5 mm away with 30 Vpp drive voltage from 10 kHz to 40 kHz is 88.21 dB and the linearity is tested to be good. To further increase the output sound pressure, we modified the diaphragm to be like a cantilever by laser cutting, and increased the maximum SPL by about 7 dB without impairing the linearity. Another method to improve sound output is to through tuning the phases of the applied driving signals on different actuation elements. For example, at the second resonant frequency at which the diaphragm vibrates in rocking mode, without change the diaphragm structure, through having 180° phase difference between signals applied on electrodes at two different halves, the sound pressure level is increased by 3.5 dB. 162 8Chapter 8 Acoustic Propeller Based on Air Jets from Acoustic Streaming 8.1 Introduction Compared to traditional microrobots for navigating in or through air, acoustic propulsion technology offers attractive features such as no error-prone moving parts, low power consumption, low manufacturing cost, etc. Compared to an underwater acoustic propeller, though, air-borne acoustic propeller typically produces substantially smaller propulsion force, because, for a given particle displacement, the acoustic intensity is much lower in air than in water (due to 3,500 times lower acoustic impedance). Thus, for propulsion in air, a chamber whose cover has one or more small orifices is usually attached to an acoustic transducer, so that when air is entrained into and ejected out of the chamber through the orifice(s), turbulent jets may be synthesized, forming vortex rings propagating away from the cover. The synthetic jets involve zero net mass flux, and thus require no net mass injection into the system, though non-zero momentum is transferred [220], producing propulsion or thrust force. This kind of synthetic jets was demonstrated in both macro- [221] and micro- [222] scales, and has been studied experimentally and numerically [223]–[225]. Silicon bulk micromachining was used to produce an orifice, through which small air-jet- induced force was produced when a nearby membrane was actuated [222], [226]. Here we report an acoustic propulsion technique based on a piezoelectric speaker (with a cover having holes made through laser machining) that produces propulsion force large enough to make the whole device to jump high and long. 163 8.2 Device Overview The device is based on a card type piezoelectric speaker (APS2513S-T-R, PUI Audio), which is 25.2×16.6 mm 2 in area, 0.37 mm (or 0.93 mm including the surface-mounted resistor on the back) in thickness, and 385 mg in weight with a piezoelectric membrane (12 × 18 mm 2 ) encapsulated in the center of a thin PCB housing (left ones on Figure 8.1a, 8.1b, and 8.1c). On the speaker, we attach a 0.5-mm-thick Polyester cover (22.5 ×16.6 mm 2 ) with an array of small circular orifices in the center that is laser-machined, and seal the cover with a 0.9-mm-thick laser- cut double-sided foam tape, forming the acoustic chamber (right ones in Figure 8.1a, 8.1b, and 8.1c). Figure 8.1 Photos of the piezoelectric speakers: (a) front sides with (right) and without cover attached (left), (b) backsides with (right) and without cover attached (left), (c) cross-sections with (right) and without cover (left), compared with a U.S. quarter coin (middle). (a) (b) (c) 164 As illustrated in Figure 8.2, when the speaker is driven with a sinusoidal voltage signal, the piezoelectric membrane vibrates sinusoidally, generating a monotone sound. As the membrane vibrates, air is drawn into the chamber (Figure 8.2a) and then pushed out through the orifices in each vibration cycle of the membrane. If the orifice size is small enough and if the flow rate of air passing through the orifices is large enough, the acoustic streaming effect produces vortex rings, which propagate away from the orifice (Figure 8.2b). When the vibration amplitude and frequency are high, the vortex ring travels fast enough, so that it is not affected by the suction stage of the next vibration cycle (Figure 8.2c), maintaining zero mass flux into the chamber while offering momentum delivery to produce thrust force. Figure 8.2 Cross-section schematics showing how zero-net-synthetic jets are generated (not in scale): (a) suction of air; (b) vortex ring generation during air ejection; c) vortex ring leaving the orifice when air is sucked in at the beginning of the next cycle. 8.3 Fabrication On a 0.5-mm-thick Polyester cover, circular orifices were laser-drilled with a commercial laser machine (LG-500, Jamieson Laser). The laser machine shines a focused laser light with controlled power for a certain amount of time on a spot, producing enough heat to melt the material and create a via hole. The portion closer to the top surface of the material receives more laser energy, and thus the opening is larger near the top, creating a nozzle-like profile (Figure 8.3b). (a) (b) (c) 165 The diameter of the orifices could be controlled by laser drilling parameters (larger power and longer radiation time result in larger orifice) and was selected according to optimization experiments (described in Experimental Results section) to determine the best design for the strongest propulsion force. We found that enabling air blowing (available in the machine) during the laser drilling made the orifices 5% smaller with better uniformity in diameter over a cover area. Although the shape and diameter of the melted area near an orifice are relatively non-uniform, the opening diameter is relatively uniform within 5% among the orifices (Figure 8.3a). Figure 8.3 (a) photo of laser-drilled holes on a 0.5-mm-thick Polyester sheet taken with a digital microscope; (b) an off-center surface profile of an orifice scanned with a surface profilometer, showing the sloped nozzle-like profile. After drilling holes, the piece is cut with cutting function of the laser machine, in which it shines laser light while travelling along a designed route with a set speed. A 0.9-mm-thick double- sided Polyethylene foam tape with rubber adhesive is also cut with the same method. The cover with orifices is attached onto the speaker with the foam tape which serves as an adhesion layer as well as the acoustic chamber wall. Once the design is determined, the whole process takes less than a minute to finish. After that, two ultra-flexible 36 American Wire Gauge (AWG) wires are soldered to provide electrical connection with little added tension and weight. (a) (b) 166 8.4 Experimental Results 8.4.1 Optimization of Design Parameters To evaluate the thrust force from air jets, the device was suspended by its wires fixed on a post with tape, forming a pendulum-like structure (Figure 8.4a). When the device was driven with 35 Vpp continuous sinusoidal signal, thrust force generated when air jets being ejected out through the orifices on the cover will push the device towards its backside direction (the side without cover), dragging the wires to the same direction (Figure 8.4b). After stabilization, the stabilization angle between balance positions before and after device actuation was measured with a digital protractor while viewing from the side, from which we could compare the thrust force: the larger the stabilization angle, the stronger the force. Figure 8.4 Side-view photos of the device suspended with its wires (a) before (b) and immediately after applying continuous driving signal of 35 V pp. The initial momentum will push the device to a large angle before the device stabilizes at a smaller angle after swinging back and forth. The following design parameters have been studied to find out the best design: (a) (b) 167 1. Orifice diameter: 230, 325, 370, 420, 465, and 560 μm, controlled by varied laser power and radiation time during laser drilling process. 2. Cover orientation: since the laser drilled hole looks like a nozzle, there are two ways of attaching the cover onto the speaker, with either larger (“Type A”) or smaller opening (“Type V”) facing the speaker. 3. Orifice interval: the distance between each orifice (equal in rows and columns), 1.2, 1.5, and 1.8 mm. 4. Array size (row×column): 12×8, 9×9, 9×7, 10×6, 7×7, 8×6, 6×6, 5×5, and 4×4. 5. Tape thickness and type: 0.86-mm-thick “Type X” (“Repositionable Foam Mounting Tape”, McMaster-Carr,), 1.65-mm-thick “Type X” and 0.90-mm-thick “Type Y” (“Light Duty Foam Mounting Tape”, McMaster-Carr), which is denser and provides stronger adhesion. To reduce the error from manual alignment between the cover and the speaker, as well as potential inconsistencies from the laser drilling process, for each design, two experiments were repeated with two covers with the same design parameters made and assembled onto the same speaker, during which stabilization angle was measured and averaged. In different designs of the speakers, the frequency of the driving signal was adjusted to their corresponding resonant frequencies so that the strongest thrust force might be generated. The measurement results are shown in Figure 8.5. From Figure 8.5a, we could see that there is an optimal orifice diameter (420 μm), as expected since too small orifice diameter limits the outcoming air jets while too large holes reduce jet speed. In Figure 8.5b, larger openings of the cover facing the speaker give strong thrust force, for different orifice diameters (230, 325, and 420 μm). According to Bernoulli Principle, this sloped nozzle structure helps ejected air to accelerate when it travels from a larger opening to a smaller one. 168 As for the orifice interval from 1.2 to 1.8 mm, smaller interval results in stronger thrust force, according to Figure 8.5c, since the diaphragm displacement is largest near its center, and denser orifices near the center generate stronger force. According to Figure 8.5d, a 7×7 array gives the best performance among the various array sizes, as the result of a tradeoff between having more orifices (thus larger propulsion area) and having less air jets from each orifice. From Figure 8.5e, we can conclude that a thinner tape works better for the type X, and the type Y tape with almost same thickness works better than the type X. From the above experiments, we came up with the optimized design parameters: 420 μm orifice diameter, “Type A” cover orientation, 1.2 mm orifice interval, 7×7 array size and 0.90- mm-thick “Type Y” tape. With this design, when driven with 35 V pp 2.08 kHz sinusoidal signal, the largest stable angle is 4.05°, while the dynamic angle immediately after device actuation could be over 12.8° (Figure 8.4b). At 4.01 kHz, there is a second resonance which gives stabilization angle of 1.85°. 169 Figure 8.5 Measured average stabilization angle versus (a) orifice diameter, (b) orifice diameter and cover orientation; (c) orifice interval; (d) orifice array size (row × column) and (e) tape thickness and type. Unless specified, the default experiment conditions are: 420 μm orifice diameter, “Type A” cover orientation, 1.2 mm orifice interval, 10×6 array size and 0.86-mm-thick “Type X” tape. 8.4.2 Acoustic Thruster Demonstrations To demonstrate the device’s ability to jump long, with the cover side facing down, the device was driven with pulsed sinusoidal signal of 4.01 kHz, 1 sec pulse repetition frequency, and 49 V pp. (a) (b) (c) (d) (e) 170 After each driving pulse, the device was slightly lifted up and moved away, with the travel distance being proportional to the pulse width. With 750 ms pulse width, the device moves about 100 mm (Figure 8.6) after one pulse. Figure 8.6 Photos of device positions (a) before and (b) after one pulse of driving signal (4.01 kHz, 49 V pp, 750 ms pulse width). The purple lines show the initial position of the device, each grid on the paper is 5.08×5.08 mm 2 . When driven with 1.74 kHz, 58 Vpp, and 575 ms pulse width, the device could be lifted up 2.8 mm above its initial position (Figure 8.7a and 8.7b), suggesting a lift force of more than 6.03 mN since the device weighs 603 mg. Sometimes the jet is unbalanced, causing one side being lifted up much higher (6.0 mm above) than the other (Figure 8.7c). (a) (b) 171 Figure 8.7 Photos of device positions (a) before; (b) during one pulse of driving signal (1.74 kHz, 58 V pp, 575 ms pulse width) and (c) during an unbalanced jump with one side lifted much high than the other. Interestingly, at the previously determined resonant frequency of 2.08 kHz, with 35 V pp continuous drive, the device rotates at a speed of 3.58 sec per round (1.76 rad/s), dragging the wires while being slightly lifted up (Figure 8.8), exhibiting different behavior than just vertical lifting. This might have been due to the interaction between the cover and the ground surface. Figure 8.8 Photos of device positions (a) before; (b) 0.84 sec after and (c) 1.79 sec after continuous drive signal (2.08 kHz, 35 V pp). (a) (b) (c) (a) (b) (c) 172 8.4.3 Wireless Thruster Operation We also tried to drive the device wirelessly with acoustic waves by applying a pulsed square wave of 4.01 kHz, 15 V0-p, and 750 ms pulse width on a tweeter (FT17H, Fostex) about 1 cm above the device. After 5 pulses of driving, the device moved about 15 mm (averaging 3.0 mm/pulse, as shown in Figure 8.9), as the acoustic waves drove the piezoelectric membrane to vibrate, producing air jets out of the orifices while generating vibration on the cover. Figure 8.9 Photos of device positions (a) before and (b) after 5 pulses of acoustic drive from a tweeter 1 cm above the device. 8.4.4 Acoustic Lifter Demonstrations Finally, to test the device’s capability to lift up a solid object, a 6-mm-thick Acrylic base was laser cut with a rectangular opening (Figure 8.10a), and attached to the backside of the device with the foam tape, leaving the center area clear for the membrane to vibrate (Figure 8.10b). (a) (b) 173 Figure 8.10 Photos of (a) a laser-cut Acrylic base with a large rectangular opening; (b) backside of the device attached onto the base with foam tape. In the first experiment, two pieces of laser-cut plastic sheets (weighing 299 mg with the bottom area being 10×10 mm 2 ) were bonded together by a double-sided tape (not visible) while sandwiching another flexible Polyester tape (green one in Figure 8.11). The plastic pieces were aligned to the center of the orifice array. The device, when driven with 49 V pp 2.08 kHz sinusoidal signal, could push up the 299 mg combined pieces by the air jets (Figure 8.11b). Figure 8.11 Photos of (a) the acoustic lifter with plastic pieces of 299 mg before actuating the device; (b) the plastic pieces being lifted up by the air jets from the device when driving signal (2.08 kHz, 49 Vpp) was applied. In another experiment, a circular 20.4 g metal piece was placed over the center of the device. While it was too heavy to be lifted, it could be rotated at a speed of 3.70 sec per round (1.70 rad/s), (a) (b) (a) (b) 174 as shown in Figure 8.12, when the device was driven under the same condition as above, likely due to both the air jets and the cover vibration induced by them. Figure 8.12 Photos of (a) the metal piece sitting on the device before device was activated; (b) the metal piece rotated about 180° after 1.85 sec of the device operation. 8.5 Summary This chapter describes acoustic thruster and lifter capable of generating propulsion force on- demand from synthetic air jets generated by acoustic waves passing through orifices on a thin plastic (shaped with laser-machining) covering a sound source (audio speaker). The orifice array pattern was optimized through experiments to produce a large thrust force. By varying driving conditions and operating frequencies, the transducer that weighs 603 mg can rotate, or jump 2.8 mm high and 100 mm far, propelled by the air jets. When tested as an acoustic lifter, the transducer is capable of lifting a plastic piece of 299 mg and rotating a metal piece of 20.4 g. The device could also be driven with acoustic signals instead of electrical power. (a) (b) 175 9Chapter 9 Summary and Future Directions 9.1 Conclusion In conclusion, this thesis presents the design, fabrication, and applications of four types of piezoelectric acoustic/ultrasonic microelectromechanical systems (MEMS). The first type is self-focusing acoustic transducer (SFAT), which is a type of single-element, planar ultrasonic transducer based on piezoelectric lead zirconate titanate (PZT) substrates vibrating in thickness mode with microfabricated Fresnel acoustic lenses on top to generate focused ultrasound. The design parameters as well as microfabrication and characterization methods of SFATs are introduced. Exploiting their small size, electrical tunability, and microfabrication-compatibility, different types of SFATs tailored for applications such as in vivo selective cancer treatment, neurostimulation, and acoustic droplet ejection (for cell extraction/delivery and semiconductor chip pick-and-placement) have been developed. The second type is microfabricated single-element PZT-based acoustic tweezers, which is similar to SFAT but with modified Fresnel acoustic lenses to achieve trapping and manipulation of large (sub-millimeter- or millimeter-sized) particles in liquid environments, with two designs demonstrated. The first design is based on a ring-focusing Fresnel lens, which is capable of generating a Bessel-like focused acoustic beam with long depth-of-focus and multiple acoustic trapping beams such as bottle beams (with three-dimensional trapping force) and quasi-Airy beams (with the capability to rotate trapped particles). The second design relies on a multi-foci linear Fresnel lens to generate large, cuboid-shaped trapping zones that can trap plastic and silicon chips with up to 5.3×0.67×0.51 mm 3 in size and 3.98 mg in weight. 176 The other two types are both airborne acoustic/ultrasonic transducers based on diaphragms vibrating in flexural mode. The third type is a piezoelectric micromachined ultrasonic transducer (PMUT) based on a hemispherical dome-shaped diaphragm that readily releases residual stress through its shape change and transforms in-plane diaphragm vibration to large radial deflection. The diaphragm is actuated by multiple piezoelectric zinc oxide (ZnO) actuation elements, each can be individually controlled to maximize the sound output through matching the phases of applied electrical signals and those of the mechanical vibrations at different locations at a specific resonant mode. The fourth type is an acoustic propeller capable of generating propulsion force on-demand from synthetic air jets generated by acoustic waves passing through orifices on a thin plastic (shaped with laser-machining) covering a sound source (a mini membrane-type speaker). The orifice array pattern has been optimized through experiments to produce large thrust force. By varying driving conditions and operating frequencies, the transducer that weighs 603 mg can rotate, or jump 2.8 mm high and 100 mm far, propelled by the air jets. When tested as an acoustic lifter, the transducer is capable of lifting a plastic piece of 299 mg and rotating a metal piece of 20.4 g. The device could also be driven with acoustic signals instead of electrical power. 9.2 Future Work on Self-Focusing Acoustic Transducers (SFATs) 9.2.1 Design and Microfabrication In the future, SFATs with better tunability (such as tunable focal length, focal position) will be developed. This could be achieved through creating annular-ring electrode arrays on PZT with individual phase/amplitude control, or be achieved through controlled lens shape/size deformation with stretchable lens materials. 177 Even faster and easier microfabrication techniques of SFATs will be explored. For example, the air-cavity reflectors could be created through developing 3D superhydrophobic nanostructures. Currently, although the air-cavity Fresnel lens could effectively focus ultrasound, half of the energy is reflected by the air cavities, losing a significant amount of output acoustic energy. Different designs and microfabrication methods will be explored to fully utilize all the acoustic energy. 9.2.2 Droplet Ejection As future work, a 3D acoustic printing system based on an array of SFAT-based acoustic droplet ejectors will be developed and each ejector will have controllable ejection direction and droplet size. With this system, on-demand, massively parallel fabrication of multi-layer, multi- material, and multi-phase printing could be realized to fabricate heterogeneous 3D structures with well-controlled properties. Currently, the relative position between the transducer and the air-liquid interface has to be manually adjusted, which is time- and labor-consuming. In the future, we hope to develop a feedback system that allows the sensing and automatic adjustment of liquid levels during droplet ejection. For cell ejection experiments, we will further verify the viability of the ejected cells. In addition, currently, the droplet size is relatively large (about 100 μm in diameter), limiting the extraction resolution. In the future, we will develop high-frequency SFAT-based droplet ejectors with droplet sizes smaller than 10 μm to achieve single-cell extraction. 9.2.3 Tumor Treatment 178 As future work, to ensure better treatment consistency, we will realize a full treatment coverage of tumor without increasing the treatment time using ring-focusing Fresnel acoustic lens to increase the focal volume while maintaining the small focal diameter and using an array of transducers. In addition, a customized mechanical scanning pattern to better fit the tumor shape could be implemented through 3D shape scan of tumors. Moreover, treatment conditions (such as acoustic intensity, pulse width, PRF, and treatment duration) can be optimized through design of experiments (DoE) procedures to ensure the best selective therapeutic effects on the tumor with shorter treatment time and lower acoustic power. 9.2.4 Neurostimulation We will continue the patch-clamp experiments to further verify the effects of focused ultrasound with different driving parameters. To ensure even better visibility under the microscope, transparent transducers based on materials such as Polyvinylidene fluoride (PVDF) will be developed. For studying the underlying mechanisms, different types of ion channel blockers and techniques such as calcium imaging will be used. Once we understand the working mechanism of single-spot FUS and confirmed its therapeutic effects, we will design and fabricate SFAT arrays to examine the effect of multi-spot neurostimulation. A lens-less SFAT working at 17.7 MHz with 0.4 mm designed focal length and 6 FWHB rings has a device area of only 1.6 mm, in a 10×10 mm 2 area, we could have 25 SFAT devices that could be individually accessed. First, we would like to see if more stimulation spots will result in more influence on ion channel activity and better therapeutic effects on epileptic activities. Then we will study how different patterns of multi-spot FUS will affect neural responses, to find out the locations where FUS-stimulation is the most effective. 179 Finally, we will develop transducers for in vivo transcranial neuromodulation experiments to evaluate the efficacy of focused ultrasound on epileptic neural activities. 9.3 Future Work on Acoustic Tweezers In the future, we will explore other effective ways to trap large particles. For example, we can modify the Fresnel lens to create vortex beams, whose diameter could be easily controlled by design and is not related to the wavelength, thus achieving large trapping volumes. Also, the rotation of the trapped particles could be realized. We would also like to explore a system with multiple acoustic tweezers, which could realize contactless, precise control and manipulation of objects, just like a system with multiple high-precision robotic arms. Moreover, we could combine SFATs and acoustic tweezers on a single chip, achieving a single-chip pumpless lab-on-a-chip system that can realize fluid pumping, mixing, acoustic trapping, and acoustic droplet ejection. 9.4 Future Work on Piezoelectric Micromachined Ultrasonic Transducers (PMUT) To further improve the sound output, we would like to optimize the design of the dome diameter and depth, as well as the patterns of piezoelectric elements. We will also explore different diaphragm (such as Parylene and polyimide which are more flexible) and piezoelectric materials (such as PZT with higher piezoelectric and electromechanical coupling coefficients) to increase the vibration amplitude on the diaphragm. In addition, to fully utilize the advantage of PMUTs, we will organize them into arrays that are capable of both transmitting and receiving acoustic signals. 180 9.5 Future Work on Acoustic Propellers The demonstrated acoustic propellers work in audio frequencies, which makes their operation very noisy. Also, the transducer is not fully microfabricated, limiting the size and weight, and thus the ability to move. Moreover, there is no control over the movement directions with only a single device. In the future, I hope to develop an array of fully microfabricated acoustic propellers working at ultrasonic frequencies with a feedback system to precisely control the movement mode and directions. 181 References [1] K. K. Shung, Diagnostic ultrasound: Imaging and blood flow measurements, 1st ed. Boca Raton, FL, USA: CRC Press, 2005. [2] R. M. S. Sigrist, J. Liau, A. El Kaffas, M. C. Chammas, and J. K. Willmann, “Ultrasound elastography: Review of techniques and clinical applications,” Theranostics, vol. 7, no. 5, pp. 1303–1329, 2017, doi: 10.7150/thno.18650. [3] R. W. Gill, “Measurement of blood flow by ultrasound: Accuracy and sources of error,” Ultrasound Med. Biol., vol. 11, no. 4, pp. 625–641, 1985, doi: 10.1016/0301-5629(85)90035-3. [4] H. Lukka et al., “High-intensity Focused Ultrasound for Prostate Cancer: A Systematic Review,” Clin. Oncol., vol. 23, no. 2, pp. 117–127, 2011, doi: 10.1016/j.clon.2010.09.002. [5] M. C. L. Peek et al., “Systematic review of high-intensity focused ultrasound ablation in the treatment of breast cancer,” Br. J. Surg., vol. 102, no. 8, pp. 873–882, 2015, doi: 10.1002/bjs.9793. [6] J. Blackmore, S. Shrivastava, J. Sallet, C. R. Butler, and R. O. Cleveland, “Ultrasound neuromodulation: A review of results, mechanisms and safety,” Ultrasound Med. Biol., vol. 45, no. 7, pp. 1509–1536, 2019, doi: 10.1016/j.ultrasmedbio.2018.12.015. [7] D. P. Darrow, “Focused ultrasound for neuromodulation,” Neurotherapeutics, vol. 16, no. 1, pp. 88–99, 2019, doi: 10.1007/s13311-018-00691-3. [8] A. Fomenko, C. Neudorfer, R. F. Dallapiazza, S. K. Kalia, and A. M. Lozano, “Low- intensity ultrasound neuromodulation: An overview of mechanisms and emerging human applications,” Brain Stimul., vol. 11, no. 6, pp. 1209–1217, 2018, doi: 10.1016/j.brs.2018.08.013. [9] T. Watson, “Ultrasound in contemporary physiotherapy practice,” Ultrasonics, vol. 48, no. 4, pp. 321–329, 2008, doi: 10.1016/j.ultras.2008.02.004. [10] S. J. Warden and J. M. McMeeken, “Ultrasound usage and dosage in sports physiotherapy,” Ultrasound Med. Biol., vol. 28, no. 8, pp. 1075–1080, 2002, doi: 10.1016/S0301-5629(02)00552- 5. [11] F. P. Curra and L. A. Crum, “Therapeutic ultrasound: Surgery and drug delivery,” Acoust. Sci. Technol., vol. 24, no. 6, pp. 343–348, 2003, doi: 10.1250/ast.24.343. [12] S. Mitragotri, “Healing sound: The use of ultrasound in drug delivery and other therapeutic applications,” Nat. Rev. Drug Discov., vol. 4, no. 3, pp. 255–260, Mar. 2005, doi: 10.1038/nrd1662. [13] S. C. Takatori, R. De Dier, J. Vermant, and J. F. Brady, “Acoustic trapping of active matter,” Nat. Commun., vol. 7, no. 10694, Apr. 2016, doi: 10.1038/ncomms10694. [14] C. R. P. Courtney et al., “Independent trapping and manipulation of microparticles using 182 dexterous acoustic tweezers,” Appl. Phys. Lett., vol. 104, no. 15, pp. 1–5, 2014, doi: 10.1063/1.4870489. [15] A. L. Yarin, M. Pfaffenlehner, and C. Tropea, “On the acoustic levitation of droplets,” J. Fluid Mech., vol. 356, pp. 65–91, 1998, doi: 10.1017/S0022112097007829. [16] W. J. Xie, C. D. Cao, Y. J. Lü, Z. Y. Hong, and B. Wei, “Acoustic method for levitation of small living animals,” Appl. Phys. Lett., vol. 89, no. 21, pp. 1–4, 2006, doi: 10.1063/1.2396893. [17] M. A. B. Andrade, N. Pérez, and J. C. Adamowski, “Review of progress in acoustic levitation,” Brazilian J. Phys., vol. 48, no. 2, pp. 190–213, 2018, doi: 10.1007/s13538-017-0552- 6. [18] A. Marzo, S. A. Seah, B. W. Drinkwater, D. R. Sahoo, B. Long, and S. Subramanian, “Holographic acoustic elements for manipulation of levitated objects.,” Nat. Commun., vol. 6, no. 8861, Oct. 2015, doi: 10.1038/ncomms9661. [19] A. Marzo and B. W. Drinkwater, “Holographic acoustic tweezers,” Proc. Natl. Acad. Sci. U. S. A., vol. 116, no. 1, pp. 84–89, 2019, doi: 10.1073/pnas.1813047115. [20] L. Zhao and E. S. Kim, “Acoustic tweezers with electrical controllability on rotation of trapped particle,” in Proc. IEEE Int. Ultrason. Symp., 2019, pp. 663–666, doi: 10.1109/ULTSYM.2019.8925781. [21] X. Ding et al., “On-chip manipulation of single microparticles, cells, and organisms using surface acoustic waves,” Proc. Natl. Acad. Sci. U. S. A., vol. 109, no. 28, pp. 11105–11109, 2012, doi: 10.1073/pnas.1209288109. [22] A. Ozcelik et al., “Acoustic tweezers for the life sciences,” Nat. Methods, vol. 15, no. 12, pp. 1021–1028, 2018, doi: 10.1038/s41592-018-0222-9. [23] N. Zhang, Y. Wen, and J. Friend, “MHz-order surface acoustic wave thruster for underwater silent propulsion,” Micromachines, vol. 11, no. 4, pp. 1–13, 2020, doi: 10.3390/MI11040419. [24] E. M. Allison, G. S. Springer, and J. Van Dam, “Ultrasonic propulsion,” J. Propuls. Power, vol. 24, no. 3, pp. 547–553, 2008, doi: 10.2514/1.30044. [25] H. Yu and E. S. Kim, “Micropropulsion of air and liquid jet by acoustic streaming,” in Proc. IEEE 16th Int. Conf. on Micro Electro Mech. Syst., 2003, pp. 76–79, doi: 10.1109/MEMSYS.2003.1189691. [26] B. Hadimioglu, R. Stearns, and R. Ellson, “Moving Liquids with Sound: The Physics of Acoustic Droplet Ejection for Robust Laboratory Automation in Life Sciences,” J. Lab. Autom., vol. 21, no. 1, pp. 4–18, 2016, doi: 10.1177/2211068215615096. [27] S. A. Elrod et al., “Nozzleless droplet formation with focused acoustic beams,” J. Appl. Phys., vol. 65, no. 9, pp. 3441–3447, 1989, doi: 10.1063/1.342663. 183 [28] C. Y. Lee, H. Yu, W. Pang, and E. S. Kim, “Droplet-based microreactions with oil encapsulation,” J. Microelectromechanical Syst., vol. 17, no. 1, pp. 147–156, 2008, doi: 10.1109/JMEMS.2007.911873. [29] Y. Tang and E. S. Kim, “Acoustic Droplet-Assisted Particle Ejection through and from Agarose-Gel-Filled Petri Dish,” in Proc. IEEE Int. Ultrason. Symp. (IUS), 2019, pp. 64–67, doi: 10.1109/ULTSYM.2019.8925932. [30] T. J. Mason, “Ultrasonic cleaning: An historical perspective,” Ultrason. Sonochem., vol. 29, pp. 519–523, 2016, doi: 10.1016/j.ultsonch.2015.05.004. [31] F. J. Fuchs, “Ultrasonic cleaning and washing of surfaces,” in Power Ultrasonics: Applications of High-Intensity Ultrasound, Elsevier Inc., 2015, pp. 577–609. [32] I. Majid, G. A. Nayik, and V. Nanda, “Ultrasonication and food technology: A review,” Cogent Food Agric., vol. 1, no. 1, p. 1071022, 2015, doi: 10.1080/23311932.2015.1071022. [33] S. Bel Haaj, A. Magnin, C. Pétrier, and S. Boufi, “Starch nanoparticles formation via high power ultrasonication,” Carbohydr. Polym., vol. 92, no. 2, pp. 1625–1632, 2013, doi: 10.1016/j.carbpol.2012.11.022. [34] M. Brigante and M. A. Sumbatyan, “Acoustic methods for the nondestructive testing of concrete: A review of foreign publications in the experimental field,” Russ. J. Nondestruct. Test., vol. 49, no. 2, pp. 100–111, 2013, doi: 10.1134/S1061830913020034. [35] M. Kobayashi and C. K. Jen, “Piezoelectric thick bismuth titanate/lead zirconate titanate composite film transducers for smart NDE of metals,” Smart Mater. Struct., vol. 13, no. 4, pp. 951–956, 2004, doi: 10.1088/0964-1726/13/4/033. [36] M. M. Saad, C. J. Bleakley, and S. Dobson, “Robust high-accuracy ultrasonic range measurement system,” IEEE Trans. Instrum. Meas., vol. 60, no. 10, pp. 3334–3341, 2011, doi: 10.1109/TIM.2011.2128950. [37] M. Drumheller, “Mobile Robot Localization Using Sonar,” IEEE Trans. Pattern Anal. Mach. Intell., vol. PAMI-9, no. 2, pp. 325–332, Mar. 1987, doi: 10.1109/TPAMI.1987.4767907. [38] R. J. Przybyla et al., “In-air rangefinding with an AlN piezoelectric micromachined ultrasound transducer,” IEEE Sens. J., vol. 11, no. 11, pp. 2690–2697, 2011, doi: 10.1109/JSEN.2011.2157490. [39] Y. Inoue, H. Kikura, H. Murakawa, M. Aritomi, and M. Mori, “A study of ultrasonic propagation for ultrasonic flow rate measurement,” Flow Meas. Instrum., vol. 19, no. 3–4, pp. 223–232, Jun. 2008, doi: 10.1016/j.flowmeasinst.2007.06.013. [40] K. Tezuka, M. Mori, T. Suzuki, and T. Kanamine, “Ultrasonic pulse-Doppler flow meter application for hydraulic power plants,” Flow Meas. Instrum., vol. 19, no. 3–4, pp. 155–162, Jun. 2008, doi: 10.1016/j.flowmeasinst.2007.06.004. 184 [41] R. Aigner, “SAW and BAW technologies for RF filter applications: A review of the relative strengths and weaknesses,” in Proc. IEEE Int. Ultrason. Symp. (IUS), Nov. 2008, pp. 582–589, doi: 10.1109/ULTSYM.2008.0140. [42] D. C. Malocha, “SAW/BAW acoustoelectronic technology for filters and communication systems,” in Proc. IEEE 11th Annual Wirel. Microw. Technol. Conf., Apr. 2010, pp. 1–7, doi: 10.1109/WAMICON.2010.5461848. [43] V. F. G. Tseng, S. S. Bedair, and N. Lazarus, “Acoustic Power Transfer and Communication with a Wireless Sensor Embedded Within Metal,” IEEE Sens. J., vol. 18, no. 13, pp. 5550–5558, 2018, doi: 10.1109/JSEN.2018.2839558. [44] H. Basaeri, D. B. Christensen, and S. Roundy, “A review of acoustic power transfer for bio- medical implants,” Smart Mater. Struct., vol. 25, no. 12, 2016, doi: 10.1088/0964- 1726/25/12/123001. [45] M. E. Kiziroglou, D. E. Boyle, S. W. Wright, and E. M. Yeatman, “Acoustic power delivery to pipeline monitoring wireless sensors,” Ultrasonics, vol. 77, pp. 54–60, 2017, doi: 10.1016/j.ultras.2017.01.017. [46] J. H. Kim, J. Lee, and C. D. Yoo, “Soldering method using longitudinal ultrasonic,” IEEE Trans. Components Packag. Technol., vol. 28, no. 3, pp. 493–498, 2005, doi: 10.1109/TCAPT.2005.848576. [47] S. ichi Matsuoka and H. Imai, “Direct welding of different metals used ultrasonic vibration,” J. Mater. Process. Technol., vol. 209, no. 2, pp. 954–960, 2009, doi: 10.1016/j.jmatprotec.2008.03.006. [48] S. Roy, J. W. Fisher, and B. T. Yen, “Fatigue resistance of welded details enhanced by ultrasonic impact treatment (UIT),” Int. J. Fatigue, vol. 25, no. 9–11, pp. 1239–1247, 2003, doi: 10.1016/S0142-1123(03)00151-8. [49] G. Bradfield, “Ultrasonic transducers: 1. Introduction to ultrasonic transducers Part A,” Ultrasonics, vol. 8, no. 2, pp. 112–123, Apr. 1970, doi: 10.1016/0041-624X(70)90052-1. [50] A. Fox, “Why And How Do Speakers Use Magnets & Electromagnetism?” https://mynewmicrophone.com/why-and-how-do-speakers-use-magnets-electromagnetism/ (accessed May 24, 2021). [51] W. C. Jones and L. W. Giles, “A moving coil microphone for high quality sound reproduction,” J. Soc. Motion Pict. Eng., vol. 17, no. 6, pp. 977–993, Dec. 1931, doi: 10.5594/J07490. [52] H. Hirao, Masahiko, Ogi, Electromagnetic Acoustic Transducers: Noncontacting Ultrasonic Measurements using EMATs, 2nd Editio. Tokyo, Japan: Springer Japan, 2017. [53] P. J. Latimer and H. L. Whaley, “Electromagnetic Transducers for Generation and Detection of Ultrasonic Waves,” in Acousto-Ultrasonics: Theory and Application, J. C. Duke, Ed. 185 Boston, MA: Springer US, 1988, pp. 209–220. [54] F. V. Hunt, Electroacoustics: The analysis of transduction, and its historical background, 2nd Editio. Cambridge, MA: Harvard University Press, 1982. [55] G. M. Sessler and J. E. West, “Electret transducers: a review,” J. Acoust. Soc. Am., vol. 53, no. 6, pp. 1589–1600, Jun. 1973, doi: 10.1121/1.1913507. [56] Ö. Oralkan et al., “Capacitive micromachined ultrasonic transducers: Next-generation arrays for acoustic imaging?,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 49, no. 11, pp. 1596–1610, 2002, doi: 10.1109/TUFFC.2002.1049742. [57] B. T. Khuri-Yakub and Ö. Oralkan, “Capacitive micromachined ultrasonic transducers for medical imaging and therapy,” J. Micromechanics Microengineering, vol. 21, no. 5, pp. 54004– 54014, 2011, doi: 10.1088/0960-1317/21/5/054004. [58] A. S. Ergun, G. G. Yaralioglu, and B. T. Khuri-Yakub, “Capacitive Micromachined Ultrasonic Transducers: Theory and Technology,” J. Aerosp. Eng., vol. 16, no. 2, pp. 76–84, Apr. 2003, doi: 10.1061/(ASCE)0893-1321(2003)16:2(76). [59] M. S. Salim, M. F. Abd Malek, R. B. W. Heng, K. M. Juni, and N. Sabri, “Capacitive Micromachined Ultrasonic Transducers: Technology and Application,” J. Med. Ultrasound, vol. 20, no. 1, pp. 8–31, 2012, doi: 10.1016/j.jmu.2012.02.001. [60] A. S. Ergun, G. G. Yaralioglu, O. Oralkan, and B. T. Khuri-Yakub, “MEMS/NEMS techniques and applications of capacitive micromachined ultrasonic transducers,” in MEMS/NEMS: Handbook Techniques and Applications, C. T. Leondes, Ed. Boston, MA: Springer US, 2006, pp. 553–615. [61] G. S. Kino, Acoustic Waves: Devices, Imaging, and Analog Signal Processing. Englewood Cliffs, NJ: Prentice Hall, 1987. [62] R. E. Newnham, L. J. Bowen, K. A. Klicker, and L. E. Cross, “Composite piezoelectric transducers,” Mater. Des., vol. 2, no. 2, pp. 93–106, 1980, doi: 10.1016/0261-3069(80)90019-9. [63] V. Marra, “Which transducer type is best for ultrasonic fingerprint sensing: CMUT, PMUT or PZT? | OnScale.” https://onscale.com/blog/which-transducer-type-is-best-for-ultrasonic- fingerprint-sensing-cmut-pmut-or-pzt/ (accessed May 25, 2021). [64] Y. Qiu et al., “Piezoelectric micromachined ultrasound transducer (PMUT) arrays for integrated sensing, actuation and imaging,” Sensors (Switzerland), vol. 15, no. 4, pp. 8020–8041, 2015, doi: 10.3390/s150408020. [65] J. Jung, W. Lee, W. Kang, E. Shin, J. Ryu, and H. Choi, “Review of piezoelectric micromachined ultrasonic transducers and their applications,” J. Micromechanics Microengineering, vol. 27, no. 11, 2017, doi: 10.1088/1361-6439/aa851b. [66] S. Jiménez-Gambín, N. Jiménez, J. M. Benlloch, and F. Camarena, “Generating Bessel 186 beams with broad depth-of-field by using phase-only acoustic holograms,” Sci. Rep., vol. 9, no. 1, pp. 1–13, 2019, doi: 10.1038/s41598-019-56369-z. [67] Z. Lin, X. Guo, J. Tu, Q. Ma, J. Wu, and D. Zhang, “Acoustic non-diffracting Airy beam,” J. Appl. Phys., vol. 117, no. 10, 2015, doi: 10.1063/1.4914295. [68] T. P. Nguyen et al., “Design, fabrication, and evaluation of multifocal point transducer for high-frequency ultrasound applications,” Sensors (Switzerland), vol. 19, no. 3, pp. 5–15, 2019, doi: 10.3390/s19030609. [69] J. Woodacre and J. Brown, “An evaluation of PZT5A, PIN-PMN-PT Single Crystal, and High-Dielectric PZT for a 5mm x 5mm Histotripsy Transducer,” Proc. IEEE Int. Ultrason. Symp., pp. 2596–2599, 2019, doi: 10.1109/ULTSYM.2019.8925688. [70] P. Katchadjian, C. Desimone, and A. Garcia, “Design and some practical applications of ultrasonic transducers with axicon lenses,” AIP Conf. Proc., vol. 1511, no. January, pp. 916–923, 2013, doi: 10.1063/1.4789142. [71] F. Acquaticci, J. F. Guarracino, S. N. Gwirc, and S. E. Lew, “A polydimethylsiloxane-based axicon lens for focused ultrasonic brain stimulation techniques,” Acoust. Sci. Technol., vol. 40, no. 2, pp. 116–126, 2019, doi: 10.1250/ast.40.116. [72] S. Liang, T. Liu, H. Gao, Z. Gu, S. An, and J. Zhu, “Acoustic metasurface by layered concentric structures,” Phys. Rev. Res., vol. 2, no. 4, pp. 1–9, 2020, doi: 10.1103/physrevresearch.2.043362. [73] M. Molerón, M. Serra-Garcia, and C. Daraio, “Acoustic Fresnel lenses with extraordinary transmission,” Appl. Phys. Lett., vol. 105, no. 11, 2014, doi: 10.1063/1.4896276. [74] C. M. Donahue, P. W. J. Anzel, L. Bonanomi, T. A. Keller, and C. Daraio, “Experimental realization of a nonlinear acoustic lens with a tunable focus,” Appl. Phys. Lett., vol. 104, no. 1, 2014, doi: 10.1063/1.4857635. [75] S. Zhai, H. Chen, C. Ding, F. Shen, C. Luo, and X. Zhao, “Manipulation of transmitted wave front using ultrathin planar acoustic metasurfaces,” Appl. Phys. A Mater. Sci. Process., vol. 120, no. 4, pp. 1283–1289, 2015, doi: 10.1007/s00339-015-9379-6. [76] Z. Xu, W. Xu, M. Qian, Q. Cheng, and X. Liu, “A flat acoustic lens to generate a Bessel- like beam,” Ultrasonics, vol. 80, pp. 66–71, 2017, doi: 10.1016/j.ultras.2017.05.005. [77] H. Y. Dong, J. Wang, G. Y. Song, W. X. Tang, and Q. Cheng, “A broadband planar acoustic metamaterial lens,” Phys. Lett. Sect. A Gen. At. Solid State Phys., vol. 383, no. 16, pp. 1955–1959, 2019, doi: 10.1016/j.physleta.2019.03.014. [78] K. Yamada and H. Shimizu, “Planar-Structure Focusing Lens for Acoustic Microscope,” in Proc. IEEE Int. Ultrason. Symp. (IUS), 1985, pp. 755–758, doi: 10.1109/ULTSYM.1985.198612. [79] T. Hamazaki and N. Morita, “Ejection Characteristics and Drop Modulation of Acoustic 187 Inkjet Printing Using Fresnel Lens,” J. Fluid Sci. Technol., vol. 4, no. 1, pp. 25–36, 2009, doi: 10.1299/jfst.4.25. [80] C.-Y. Lee, H. Y. Yu, and E. S. Kim, “Acoustic Ejector with Novel Lens Employing Air- Reflectors,” in Proc. IEEE 19th Conf. on Micro Electro Mech. Syst., 2002, no. January, pp. 170– 173, doi: 10.1109/MEMSYS.2006.1627763. [81] Y. Tang, L. Wang, Y. Wang, and E. S. Kim, “On-demand, heatless ejection of sub-MM- sized liquid droplets,” in Proc. IEEE 30th Conf. on Micro Electro Mech. Syst., Jan. 2017, pp. 1196–1199, doi: 10.1109/MEMSYS.2017.7863630. [82] Y. Choe, J. W. Kim, K. K. Shung, and E. S. Kim, “Microparticle trapping in an ultrasonic Bessel beam,” Appl. Phys. Lett., vol. 99, no. 23, pp. 2009–2012, 2011, doi: 10.1063/1.3665615. [83] Y. Tang and E. S. Kim, “Ring-Focusing Fresnel Acoustic Lens for Long Depth-of-Focus Focused Ultrasound With Multiple Trapping Zones,” J. Microelectromechanical Syst., vol. 29, no. 5, pp. 692–698, Oct. 2020, doi: 10.1109/JMEMS.2020.3000715. [84] Y. Tang and E. S. Kim, “Acoustic Tweezers Based on Linear Fresnel Lens with Air Cavities for Large Volume Particle Trapping,” in IEEE 32nd Int. Conf. on Micro Electro Mech. Syst. (MEMS), 2019, vol. 2019-Janua, doi: 10.1109/MEMSYS.2019.8870692. [85] Y. Zhou, “High-Intensity Focused Ultrasound Treatment for Advanced Pancreatic Cancer,” Gastroenterol. Res. Pract., vol. 2014, pp. 1–11, 2014, doi: 10.1155/2014/205325. [86] S. Madersbacher, M. Pedevilla, L. Vingers, M. Susani, and M. Marberger, “Effect of high- intensity focused ultrasound on human prostate cancer in vivo.,” Cancer Res., vol. 55, no. 15, pp. 3346–51, Aug. 1995, [Online]. Available: http://www.ncbi.nlm.nih.gov/pubmed/7542168. [87] Z. Ram et al., “Magnetic resonance imaging-guided, high-intensity focused ultrasound for brain tumor therapy.,” Neurosurgery, vol. 59, no. 5, pp. 949–55; discussion 955-6, Nov. 2006, doi: 10.1227/01.NEU.0000254439.02736.D8. [88] D. P. Darrow, P. O’Brien, T. J. Richner, T. I. Netoff, and E. S. Ebbini, “Reversible neuroinhibition by focused ultrasound is mediated by a thermal mechanism,” Brain Stimul., vol. 12, no. 6, pp. 1439–1447, 2019, doi: 10.1016/j.brs.2019.07.015. [89] W. Legon et al., “Transcranial focused ultrasound modulates the activity of primary somatosensory cortex in humans,” Nat. Neurosci., vol. 17, no. 2, pp. 322–329, 2014, doi: 10.1038/nn.3620. [90] J. W. Kwon, S. Kamal-Bahl, and E. S. Kim, “In situ DNA synthesis on glass substrate for microarray fabrication using self-focusing acoustic transducer,” IEEE Trans. Autom. Sci. Eng., vol. 3, no. 2, pp. 152–158, 2006, doi: 10.1109/TASE.2006.871483. [91] J. Lee, S. Y. Teh, A. Lee, H. H. Kim, C. Lee, and K. K. Shung, “Single beam acoustic trapping,” Appl. Phys. Lett., vol. 95, no. 7, pp. 21–24, 2009, doi: 10.1063/1.3206910. 188 [92] H. Yu, J. W. Kwon, and E. S. Kim, “Microfluidic mixer and transporter based on PZT self- focusing acoustic transducers,” J. Microelectromechanical Syst., vol. 15, no. 4, pp. 1015–1024, 2006, doi: 10.1109/JMEMS.2006.878890. [93] H. Yu and E. S. Kim, “Micropropulsion of air and liquid jet by acoustic streaming,” in Proc. IEEE 16th Int. Conf. on Micro Electro Mech. Syst., 2003, pp. 76–79, doi: 10.1109/MEMSYS.2003.1189691. [94] G. R. Lockwood, D. H. Turnbull, and F. S. Foster, “Fabrication of high frequency spherically shaped ceramic transducers,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 41, no. 2, pp. 231–235, 1994, doi: 10.1109/58.279136. [95] M. Akiyama and T. Kamakura, “Elliptically curved acoustic lens for emitting strongly focused finite-amplitude beams: Application of the spheroidal beam equation model to the theoretical prediction,” Acoust. Sci. Technol., vol. 26, no. 3, pp. 279–284, 2005, doi: 10.1250/ast.26.279. [96] M. Lu, M. Wan, F. Xu, X. Wang, and X. Chang, “Design and experiment of 256-element ultrasound phased array for noninvasive focused ultrasound surgery,” Ultrasonics, vol. 44, pp. e325–e330, 2006, doi: https://doi.org/10.1016/j.ultras.2006.07.015. [97] Y. Tang and E. S. Kim, “Electrical Tuning of Focal Size with Single Focused Ultrasonic Transducer,” in Proc. IEEE Int. Ultrason. Symp., 2018, doi: 10.1109/ULTSYM.2018.8579883. [98] L. Zhao and E. S. Kim, “Focused Ultrasonic Transducer with Electrically Controllable Focal-Point Location,” in Proc. IEEE Int. Ultrason. Symp., Oct. 2018, vol. 2018-Octob, no. January, pp. 1–3, doi: 10.1109/ULTSYM.2018.8580054. [99] Boston Piezo-Optics Inc., “Properties of Ceramic Materials (PZT).” https://www.bostonpiezooptics.com/ceramic-materials-pzt (accessed May 25, 2021). [100] Y. Tang, S. Liu, and E. S. Kim, “MEMS Focused Ultrasonic Transducer with Air-Cavity Lens Based on Polydimethylsiloxane (PDMS) Membrane,” in IEEE 33rd Int. Conf. on Micro Electro Mech. Syst. (MEMS), Jan. 2020, pp. 58–61, doi: 10.1109/MEMS46641.2020.9056313. [101] O. C. Zienkiewicz, R. L. Taylor, and J. Z. Zhu, The Finite Element Method: Its Basis and Fundamentals, 7th editio. Oxford, UK: Butterworth-Heinemann, 2013. [102] L. E. Kinsler, A. R. Frey, A. B. Coppens, and J. V. Sanders, Fundamentals of Acoustics. New York, NY, USA: John Wiley & Sons, 1982. [103] H. H. Barrett and F. A. Horrigan, “Fresnel zone plate imaging of gamma rays; theory.,” Appl. Opt., vol. 12, no. 11, pp. 2686–702, Nov. 1973, doi: 10.1364/AO.12.002686. [104] M. A. Eddings, M. A. Johnson, and B. K. Gale, “Determining the optimal PDMS–PDMS bonding technique for microfluidic devices,” J. Micromechanics Microengineering, vol. 18, no. 6, p. 067001, Jun. 2008, doi: 10.1088/0960-1317/18/6/067001. 189 [105] J. A. Dobrowolski, Scattering Parameters in RF and Microwave Circuit Analysis and Design. Norwood, MA, USA: Artech House, 2016. [106] R. W. Wood and A. L. Loomis, “The physical and biological effects of high-frequency sound-waves of great intensity,” Phil. Mag., vol. 4, no. 22, pp. 417–436, Jan. 1927, doi: 10.1016/S0016-0032(28)92379-0. [107] U. Demirci, “Acoustic picoliter droplets for emerging applications in semiconductor industry and biotechnology,” J. Microelectromechanical Syst., vol. 15, no. 4, pp. 957–966, 2006, doi: 10.1109/JMEMS.2006.878879. [108] B. Hadimioglu, S. Elrod, and R. Sprague, “Acoustic ink printing: an application of ultrasonics for photographic quality printing at high speed,” in Proc. IEEE Int. Ultrason. Symp. (IUS), 2001, vol. 1, pp. 627–635, doi: 10.1109/ULTSYM.2001.991808. [109] E. K. Sackmann et al., “Technologies That Enable Accurate and Precise Nano- to Milliliter- Scale Liquid Dispensing of Aqueous Reagents Using Acoustic Droplet Ejection,” J. Lab. Autom., vol. 21, no. 1, pp. 166–177, 2016, doi: 10.1177/2211068215602191. [110] B. Colin, B. Deprez, and C. Couturier, “High-Throughput DNA Plasmid Transfection Using Acoustic Droplet Ejection Technology,” SLAS Discov. Adv. Sci. Drug Discov., vol. 24, no. 4, pp. 492–500, Oct. 2018, doi: 10.1177/2472555218803064. [111] U. Demirci and G. Montesano, “Cell encapsulating droplet vitrification,” Lab Chip, vol. 7, no. 11, pp. 1428–1433, 2007, doi: 10.1039/B705809H. [112] Y. Fang et al., “Rapid generation of multiplexed cell cocultures using acoustic droplet ejection followed by aqueous two-phase exclusion patterning,” Tissue Eng. - Part C Methods, vol. 18, no. 9, pp. 647–657, 2012, doi: 10.1089/ten.tec.2011.0709. [113] Y. Choe, S. Chen, and E. S. Kim, “Peptide Synthesis on Glass Substrate Using Acoustic Droplet Ejector,” IEEE Trans. Biomed. Eng., vol. 61, no. 3, pp. 705–710, 2014, doi: 10.1109/TBME.2013.2287218. [114] C. G. Roessler et al., “Acoustic Injectors for Drop-On-Demand Serial Femtosecond Crystallography,” Structure, vol. 24, no. 4, pp. 631–640, 2016, doi: https://doi.org/10.1016/j.str.2016.02.007. [115] C. Y. Lee, H. Yu, and E. S. Kim, “Harmonic operation of acoustic transducer for droplet ejection application,” in Proc. 14th Int. Conf. Solid-State Sens. Actuators and Microsyst., 2007, pp. 1283–1286, doi: 10.1109/SENSOR.2007.4300372. [116] J. W. Kwon, H. Yu, Q. Zou, and E. S. Kim, “Directional droplet ejection by nozzleless acoustic ejectors built on ZnO and PZT,” J. Micromechanics Microengineering, vol. 16, no. 12, pp. 2697–2704, 2006, doi: 10.1088/0960-1317/16/12/024. [117] D. Huang and E. S. Kim, “Micromachined acoustic-wave liquid ejector,” J. Microelectromechanical Syst., vol. 10, no. 3, pp. 442–449, 2001, doi: 10.1109/84.946804. 190 [118] H. Yu, J. W. Kwon, and E. S. Kim, “Chembio extraction on a chip by nanoliter droplet ejection,” Lab Chip, vol. 5, no. 3, pp. 344–349, 2005, doi: 10.1039/B413697G. [119] C. Y. Lee, W. Pang, S. C. Hill, H. Yu, and E. S. Kim, “Airborne particle generation through acoustic ejection of particles-in-droplets,” Aerosol Sci. Technol., vol. 42, no. 10, pp. 832–841, 2008, doi: 10.1080/02786820802345709. [120] C. Y. Lee, W. Pang, H. Yu, and E. S. Kim, “Subpicoliter droplet generation based on a nozzle-free acoustic transducer,” Appl. Phys. Lett., vol. 93, no. 3, pp. 2006–2009, 2008, doi: 10.1063/1.2958342. [121] “SMT Placement Machines,” in Development and Evaluation of Setup Strategies in Printed Circuit Board Assembly, Wiesbaden: Gabler, 2008, pp. 15–30. [122] M. Sahin, K. Ayranci, E. Kosun, and E. Ayranci, “Density, sound velocity and viscosity properties of aqueous sodium metatungstate solutions and an application of these solutions in heavy mineral separations,” Chem. Geol., vol. 264, no. 1–4, pp. 96–100, 2009, doi: 10.1016/j.chemgeo.2009.02.016. [123] W. L. Nyborg, “Acoustic Streaming due to Attenuated Plane Waves,” J. Acoust. Soc. Am., vol. 25, no. 1, pp. 68–75, Jan. 1953, doi: 10.1121/1.1907010. [124] COMSOL Inc, “COMSOL Multiphysics Version 5.5 User’s Guide on Acoustic Module.” Burlington, MA, USA, 2019. [125] Y. Cengel and J. Cimbala, Fluid Mechanics: Fundamentals and Applications, 3rd Editio. New York, NY, USA: McGraw-Hill, 2014. [126] S. Natsui, R. Nashimoto, D. Nakajima, T. Kumagai, T. Kikuchi, and R. O. Suzuki, “Observation of Interface Deformation in Sodium Polytungstate Solution–Silicone Oil System due to Single Rising Bubble,” ISIJ Int., vol. 57, no. 2, pp. 394–396, 2017, doi: 10.2355/isijinternational.ISIJINT-2016-521. [127] E. Olsson and G. Kreiss, “A conservative level set method for two phase flow,” J. Comput. Phys., vol. 210, no. 1, pp. 225–246, 2005, doi: 10.1016/j.jcp.2005.04.007. [128] COMSOL Inc, “COMSOL Multiphysics Version 5.5 User’s Guide on CFD Module.” Burlington, MA, USA, 2019. [129] Y. Zhong, H. Fang, Q. Ma, and X. Dong, “Analysis of droplet stability after ejection from an inkjet nozzle,” J. Fluid Mech., vol. 845, pp. 378–391, 2018, doi: DOI: 10.1017/jfm.2018.251. [130] W. E. Lowry et al., “Generation of human induced pluripotent stem cells from dermal fibroblasts.,” Proc. Natl. Acad. Sci. U. S. A., vol. 105, no. 8, pp. 2883–8, Feb. 2008, doi: 10.1073/pnas.0711983105. [131] D. Frumkin, A. Wasserstrom, S. Itzkovitz, A. Harmelin, G. Rechavi, and E. Shapiro, “Amplification of multiple genomic loci from single cells isolated by laser micro-dissection of 191 tissues,” BMC Biotechnol., vol. 8, no. 1, p. 17, 2008, doi: 10.1186/1472-6750-8-17. [132] L. P. Adjadj, G. Storti, and M. Morbidelli, “Ultrasound Attenuation in Polystyrene Latexes,” Langmuir, vol. 19, no. 9, pp. 3953–3957, Apr. 2003, doi: 10.1021/la026893l. [133] S. Yang et al., “High-intensity focused ultrasound ablation: An in vitro agarose gel model,” Int. J. Clin. Exp. Med., vol. 10, no. 11, pp. 15302–15308, 2017, [Online]. Available: http://www.ijcem.com/files/ijcem0047133.pdf. [134] H. Sung et al., “Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries.,” CA. Cancer J. Clin., vol. 71, no. 3, pp. 209–249, May 2021, doi: 10.3322/caac.21660. [135] R. Sullivan et al., “Global cancer surgery: Delivering safe, affordable, and timely cancer surgery,” Lancet Oncol., vol. 16, no. 11, pp. 1193–1224, 2015, doi: 10.1016/S1470- 2045(15)00223-5. [136] S. Guillaumier et al., “A Multicentre Study of 5-year Outcomes Following Focal Therapy in Treating Clinically Significant Nonmetastatic Prostate Cancer,” Eur. Urol., vol. 74, no. 4, pp. 422–429, Oct. 2018, doi: 10.1016/j.eururo.2018.06.006. [137] R. R. Love, H. Leventhal, D. V. Easterling, and D. R. Nerenz, “Side effects and emotional distress during cancer chemotherapy.,” Cancer, vol. 63, no. 3, pp. 604–12, Feb. 1989, doi: 10.1002/1097-0142(19890201)63:3<604::aid-cncr2820630334>3.0.co;2-2. [138] L. G. Merckel et al., “First clinical experience with a dedicated MRI-guided high-intensity focused ultrasound system for breast cancer ablation,” Eur. Radiol., vol. 26, no. 11, pp. 4037–4046, 2016, doi: 10.1007/s00330-016-4222-9. [139] G. T. Clement, “Perspectives in clinical uses of high-intensity focused ultrasound,” Ultrasonics, vol. 42, no. 10, pp. 1087–1093, 2004, doi: 10.1016/j.ultras.2004.04.003. [140] V. A. Salgaonkar, S. Scott, J. Kurhanewicz, and C. J. Diederich, “High-intensity interstitial ultrasound for thermal ablation of focal cancer targets in prostate,” in AIP Conf. Proc., 2017, vol. 1821, no. March, p. 160002, doi: 10.1063/1.4977655. [141] A. R. Smolock et al., “Robotically assisted sonic therapy as a noninvasive nonthermal ablation modality: Proof of concept in a porcine liver model,” Radiology, vol. 287, no. 2, pp. 485– 493, 2018, doi: 10.1148/radiol.2018171544. [142] C. C. Church, E. L. Carstensen, W. L. Nyborg, P. L. Carson, L. A. Frizzell, and M. R. Bailey, “The Risk of Exposure to Diagnostic Ultrasound in Postnatal Subjects,” J. Ultrasound Med., vol. 27, no. 4, pp. 565–592, Apr. 2008, doi: 10.7863/jum.2008.27.4.565. [143] A. I. So, “HIFU ablation is not a proven standard treatment for localized prostate cancer,” Can. Urol. Assoc. J., vol. 5, no. 6, pp. 424–426, Dec. 2011, doi: 10.5489/cuaj.11232. [144] G. F. Woodworth, G. P. Dunn, E. A. Nance, J. Hanes, and H. Brem, “Emerging Insights 192 into Barriers to Effective Brain Tumor Therapeutics,” Front. Oncol., vol. 4, no. July, pp. 1–14, Jul. 2014, doi: 10.3389/fonc.2014.00126. [145] O. K. Kosheleva, T. C. Lai, N. G. Chen, M. Hsiao, and C. H. Chen, “Selective killing of cancer cells by nanoparticle-assisted ultrasound,” J. Nanobiotechnology, vol. 14, no. 1, pp. 1–11, 2016, doi: 10.1186/s12951-016-0194-9. [146] J. Wang et al., “Selective depletion of tumor neovasculature by microbubble destruction with appropriate ultrasound pressure,” Int. J. Cancer, vol. 137, no. 10, pp. 2478–2491, 2015, doi: 10.1002/ijc.29597. [147] F. Lejbkowicz, M. Zwiran, and S. Salzberg, “The response of normal and malignant cells to ultrasound in vitro.,” Ultrasound Med. Biol., vol. 19, no. 1, pp. 75–82, 1993, doi: 10.1016/0301- 5629(93)90020-o. [148] F. Lejbkowicz and S. Salzberg, “Distinct sensitivity of normal and malignant cells to ultrasound in vitro.,” Environ. Health Perspect., vol. 105, no. suppl 6, pp. 1575–1578, Dec. 1997, doi: 10.1289/ehp.97105s61575. [149] D. R. Mittelstein et al., “Selective ablation of cancer cells with low intensity pulsed ultrasound,” Appl. Phys. Lett., vol. 116, no. 1, 2020, doi: 10.1063/1.5128627. [150] L. Wang, Y. J. Li, A. Lin, Y. Choe, M. E. Gross, and E. S. Kim, “A self-focusing acoustic transducer that exploits cytoskeletal differences for selective cytolysis of cancer cells,” J. Microelectromechanical Syst., vol. 22, no. 3, pp. 542–552, 2013, doi: 10.1109/JMEMS.2012.2229696. [151] L. Wang, C. P. Liao, M. Gross, and E. S. Kim, “Self focusing acoustic transducer (SFAT) with 10-mm focal length for cancer-specific localized cytolysis of 3D cell spheroids in 3D Matrigel,” in Proc. 17th Int. Conf. Solid-State Sens. Actuators and Microsyst., 2013, no. June, pp. 653–656, doi: 10.1109/Transducers.2013.6626851. [152] G. L. Johnson, “Solid state Tesla coil,” 2001. https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.628.6214&rep=rep1&type=pdf (accessed May 21, 2021). [153] G. Poste, J. Doll, I. R. Hart, and I. J. Fidler, “In vitro selection of murine B16 melanoma variants with enhanced tissue-invasive properties.,” Cancer Res., vol. 40, no. 5, pp. 1636–1644, May 1980. [154] V. A. Salgaonkar, S. Scott, J. Kurhanewicz, and C. J. Diederich, “High-intensity interstitial ultrasound for thermal ablation of focal cancer targets in prostate,” AIP Conf. Proc., vol. 1821, no. March, 2017, doi: 10.1063/1.4977655. [155] R. E. Apfel and C. K. Holland, “Gauging the likelihood of cavitation from short-pulse, low- duty cycle diagnostic ultrasound,” Ultrasound Med. Biol., vol. 17, no. 2, pp. 179–185, Jan. 1991, doi: 10.1016/0301-5629(91)90125-G. 193 [156] T. R. Nelson, J. B. Fowlkes, J. S. Abramowicz, and C. C. Church, “Ultrasound biosafety considerations for the practicing sonographer and sonologist,” J. Ultrasound Med., vol. 28, no. 2, pp. 139–150, 2009, doi: 10.7863/jum.2009.28.2.139. [157] P. S. Yarmolenko et al., “Thresholds for thermal damage to normal tissues: An update,” Int. J. Hyperth., vol. 27, no. 4, pp. 320–343, 2011, doi: 10.3109/02656736.2010.534527. [158] M. Lekka, “Discrimination Between Normal and Cancerous Cells Using AFM.,” Bionanoscience, vol. 6, no. 1, pp. 65–80, 2016, doi: 10.1007/s12668-016-0191-3. [159] S. E. Cross, Y. S. Jin, J. Rao, and J. K. Gimzewski, “Nanomechanical analysis of cells from cancer patients,” Nat. Nanotechnol., vol. 2, no. 12, pp. 780–783, 2007, doi: 10.1038/nnano.2007.388. [160] N. Mizrahi et al., “Low intensity ultrasound perturbs cytoskeleton dynamics,” Soft Matter, vol. 8, no. 8, pp. 2438–2443, 2012, doi: 10.1039/c2sm07246g. [161] S. Heyden and M. Ortiz, “Oncotripsy: Targeting cancer cells selectively via resonant harmonic excitation,” J. Mech. Phys. Solids, vol. 92, pp. 164–175, 2016, doi: 10.1016/j.jmps.2016.04.016. [162] M. Fraldi, A. Cutolo, A. R. Carotenuto, S. Palumbo, and N. Pugno, “A lesson from earthquake engineering for selectively damaging cancer cell structures,” J. Mech. Behav. Biomed. Mater., vol. 119, no. April, 2021, doi: 10.1016/j.jmbbm.2021.104533. [163] D. L. Miller, R. M. Thomas, and A. R. Williams, “Mechanisms for hemolysis by ultrasonic cavitation in the rotating exposure system,” Ultrasound Med. Biol., vol. 17, no. 2, pp. 171–178, 1991, doi: 10.1016/0301-5629(91)90124-F. [164] E. P. Stride and C. C. Coussios, “Cavitation and contrast: The use of bubbles in ultrasound imaging and therapy,” Proc. Inst. Mech. Eng. Part H J. Eng. Med., vol. 224, no. 2, pp. 171–191, 2010, doi: 10.1243/09544119JEIM622. [165] M. Wang, Y. Zhang, C. Cai, J. Tu, X. Guo, and D. Zhang, “Sonoporation-induced cell membrane permeabilization and cytoskeleton disassembly at varied acoustic and microbubble-cell parameters,” Sci. Rep., vol. 8, no. 1, pp. 1–12, 2018, doi: 10.1038/s41598-018-22056-8. [166] H. A. Hancock et al., “Investigations into Pulsed High-Intensity Focused Ultrasound- Enhanced Delivery: Preliminary Evidence for a Novel Mechanism,” Ultrasound Med. Biol., vol. 35, no. 10, pp. 1722–1736, 2009, doi: 10.1016/j.ultrasmedbio.2009.04.020. [167] V. Pereno, J. Lei, D. Carugo, and E. Stride, “Microstreaming inside Model Cells Induced by Ultrasound and Microbubbles,” Langmuir, vol. 36, no. 23, pp. 6388–6398, 2020, doi: 10.1021/acs.langmuir.0c00536. [168] B. Krasovitski, V. Frenkel, S. Shoham, and E. Kimmel, “Intramembrane cavitation as a unifying mechanism for ultrasound-induced bioeffects,” Proc. Natl. Acad. Sci. U. S. A., vol. 108, no. 8, pp. 3258–3263, 2011, doi: 10.1073/pnas.1015771108. 194 [169] D. McCreery, V. Pikov, and P. R. Troyk, “Neuronal loss due to prolonged controlled- current stimulation with chronically implanted microelectrodes in the cat cerebral cortex,” J. Neural Eng., vol. 7, no. 3, p. 036005, Jun. 2010, doi: 10.1088/1741-2560/7/3/036005. [170] M. Hallett, “Transcranial magnetic stimulation and the human brain,” Nature, vol. 406, no. 6792, pp. 147–150, Jul. 2000, doi: 10.1038/35018000. [171] J. Mindes, M. J. Dubin, and M. Altemus, “Cranial electrical stimulation (CES): Comparison with cranial nerve stimulation, transcranial direct current stimulation and transcranial low voltage pulsed electromagnetic fields,” Brain Stimul., vol. 7, no. 2, p. e13, Mar. 2014, doi: 10.1016/j.brs.2014.01.047. [172] F. J. Fry, H. W. Ades, and W. J. Fry, “Production of Reversible Changes in the Central Nervous System by Ultrasound,” Science (80-. )., vol. 127, no. 3289, pp. 83–84, Jan. 1958, doi: 10.1126/science.127.3289.83. [173] W. J. Tyler, Y. Tufail, M. Finsterwald, M. L. Tauchmann, E. J. Olson, and C. Majestic, “Remote Excitation of Neuronal Circuits Using Low-Intensity, Low-Frequency Ultrasound,” PLoS One, vol. 3, no. 10, p. e3511, Oct. 2008, doi: 10.1371/journal.pone.0003511. [174] M. A. Dinno, M. Dyson, S. R. Young, A. J. Mortimer, J. Hart, and L. A. Crum, “The significance of membrane changes in the safe and effective use of therapeutic and diagnostic ultrasound,” Phys. Med. Biol., vol. 34, no. 11, pp. 1543–1552, Nov. 1989, doi: 10.1088/0031- 9155/34/11/003. [175] A. J. Mortimer and M. Dyson, “The effect of therapeutic ultrasound on calcium uptake in fibroblasts,” Ultrasound Med. Biol., vol. 14, no. 6, pp. 499–506, Jan. 1988, doi: 10.1016/0301- 5629(88)90111-1. [176] I. V. Chapman, N. A. MacNally, and S. Tucker, “Ultrasound-induced changes in rates of influx and efflux of potassium ions in rat thymocytes in vitro,” Ultrasound Med. Biol., vol. 6, no. 1, pp. 47–49, Jan. 1980, doi: 10.1016/0301-5629(80)90063-0. [177] Y. Tufail et al., “Transcranial Pulsed Ultrasound Stimulates Intact Brain Circuits,” Neuron, vol. 66, no. 5, pp. 681–694, Jun. 2010, doi: 10.1016/j.neuron.2010.05.008. [178] S. Veitinger, “The Patch-Clamp Technique,” 2011. https://www.leica- microsystems.com/science-lab/the-patch-clamp-technique/ (accessed May 29, 2021). [179] T. J. Manuel et al., “Ultrasound neuromodulation depends on pulse repetition frequency and can modulate inhibitory effects of TTX,” Sci. Rep., vol. 10, no. 1, p. 15347, Dec. 2020, doi: 10.1038/s41598-020-72189-y. [180] S.-J. Oh et al., “Ultrasonic Neuromodulation via Astrocytic TRPA1,” Curr. Biol., vol. 29, no. 20, pp. 3386-3401.e8, Oct. 2019, doi: 10.1016/j.cub.2019.08.021. [181] I. Misri, P. Hareesh, S. Yang, and D. L. DeVoe, “Microfabrication of bulk PZT transducers by dry film photolithography and micro powder blasting,” J. Micromechanics Microengineering, 195 vol. 22, no. 8, p. 085017, Aug. 2012, doi: 10.1088/0960-1317/22/8/085017. [182] B. Vanherberghen et al., “Ultrasound-controlled cell aggregation in a multi-well chip,” Lab Chip, vol. 10, no. 20, p. 2727, 2010, doi: 10.1039/c004707d. [183] J. A. Molina‐Bolívar and F. Galisteo‐González, “Latex Immunoagglutination Assays,” J. Macromol. Sci. Part C Polym. Rev., vol. 45, no. 1, pp. 59–98, Jan. 2005, doi: 10.1081/MC- 200045819. [184] D. G. Grier, “A revolution in optical manipulation,” Nature, vol. 424, no. 6950, pp. 810– 816, Aug. 2003, doi: 10.1038/nature01935. [185] A. Rosenthal and J. Voldman, “Dielectrophoretic Traps for Single-Particle Patterning,” Biophys. J., vol. 88, no. 3, pp. 2193–2205, Mar. 2005, doi: 10.1529/biophysj.104.049684. [186] Z. M. Wang, R. G. Wu, Z. P. Wang, and R. V. Ramanujan, “Magnetic Trapping of Bacteria at Low Magnetic Fields,” Sci. Rep., vol. 6, no. 1, p. 26945, Jun. 2016, doi: 10.1038/srep26945. [187] D. Bazou, R. Kearney, F. Mansergh, C. Bourdon, J. Farrar, and M. Wride, “Gene Expression Analysis of Mouse Embryonic Stem Cells Following Levitation in an Ultrasound Standing Wave Trap,” Ultrasound Med. Biol., vol. 37, no. 2, pp. 321–330, Feb. 2011, doi: 10.1016/j.ultrasmedbio.2010.10.019. [188] J. Hultström, O. Manneberg, K. Dopf, H. M. Hertz, H. Brismar, and M. Wiklund, “Proliferation and viability of adherent cells manipulated by standing-wave ultrasound in a microfluidic chip,” Ultrasound Med. Biol., vol. 33, no. 1, pp. 145–151, Jan. 2007, doi: 10.1016/j.ultrasmedbio.2006.07.024. [189] A. E. Christakou et al., “Aggregation and long-term positioning of cells by ultrasound in a multi-well microchip for high-resolution imaging of the natural killer cell immune synapse,” in Proc. of 15th Int. Conf. on Miniaturized Syst. for Chemistry and Life Sciences (μTAS 2011), 2011, pp. 329–331, [Online]. Available: https://www.aphys.kth.se/polopoly_fs/1.489917.1600689448!/Christakou, MicroTAS_2011.pdf. [190] L. Zhao and E. S. Kim, “Acoustic Tweezers for Trapping Late-Stage Zebrafish Embryos,” in Proc. IEEE 32nd Int. Conf. on Micro Electro Mech. Syst. (MEMS), Jan. 2019, pp. 57–60, doi: 10.1109/MEMSYS.2019.8870615. [191] S. Radel, J. Schnöller, A. Dominguez, B. Lendl, M. Gröschl, and E. Benes, “Raman spectroscopy of particles in suspension concentrated by an ultrasonic standing wave,” e i Elektrotechnik und Informationstechnik, vol. 125, no. 3, pp. 82–85, Mar. 2008, doi: 10.1007/s00502-008-0515-2. [192] W. Liu et al., “Directly trapping of nanoscale biomolecules using bulk acoustic wave resonators,” in 2016 IEEE Sensors Conf., Oct. 2016, pp. 1–3, doi: 10.1109/ICSENS.2016.7808615. [193] J. Shi, D. Ahmed, X. Mao, S.-C. S. Lin, A. Lawit, and T. J. Huang, “Acoustic tweezers: patterning cells and microparticles using standing surface acoustic waves (SSAW),” Lab Chip, vol. 196 9, no. 20, p. 2890, 2009, doi: 10.1039/b910595f. [194] J. Lee, S.-Y. Teh, A. Lee, H. H. Kim, C. Lee, and K. K. Shung, “Transverse Acoustic Trapping Using a Gaussian Focused Ultrasound,” Ultrasound Med. Biol., vol. 36, no. 2, pp. 350– 355, Feb. 2010, doi: 10.1016/j.ultrasmedbio.2009.10.005. [195] J. Wu, “Acoustical tweezers,” J. Acoust. Soc. Am., vol. 89, no. 5, pp. 2140–2143, May 1991, doi: 10.1121/1.400907. [196] A. Franklin, A. Marzo, R. Malkin, and B. W. Drinkwater, “Three-dimensional ultrasonic trapping of micro-particles in water with a simple and compact two-element transducer,” Appl. Phys. Lett., vol. 111, no. 9, p. 094101, Aug. 2017, doi: 10.1063/1.4992092. [197] Y. Zheng and S. D. Silverstain, “Novel transmit aperture for very large depth of focus in medical ultrasound B-scan,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 53, no. 6, pp. 1079–1087, Jun. 2006, doi: 10.1109/TUFFC.2006.1642507. [198] A. Lamarre and F. Mainguy, “Dynamic focusing of phased arrays for nondestructivetesting: Characterization and application,” in Proc. Int. Symp. on Adv. Sensors for Metal Processing, 1999, pp. 217–227, [Online]. Available: https://www.osti.gov/etdeweb/biblio/20151258. [199] J. S. Jeong, J. M. Cannata, and K. K. Shung, “Dual-Focus Therapeutic Ultrasound Transducer for Production of Broad Tissue Lesions,” Ultrasound Med. Biol., vol. 36, no. 11, pp. 1836–1848, Nov. 2010, doi: 10.1016/j.ultrasmedbio.2010.08.008. [200] S. Nagai and K. Iizuka, “A practical ultrasound axicon for non-destructive testing,” Ultrasonics, vol. 20, no. 6, pp. 265–270, Nov. 1982, doi: 10.1016/0041-624X(82)90047-6. [201] A. Karpelson, “Piezotransducers with acoustic lenses forming narrow weakly diverging ultrasonic beams,” in Proc. 9th European Conf. on NDT, 2006, pp. 1–12, [Online]. Available: https://www.ndt.net/article/ecndt2006/doc/Mo.2.7.2.pdf. [202] J.-Y. Lu and J. F. Greenleaf, “A study of two-dimensional array transducers for limited diffraction beams,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 41, no. 5, pp. 724–739, Sep. 1994, doi: 10.1109/58.308509. [203] D. J. Collins et al., “Selective particle and cell capture in a continuous flow using micro- vortex acoustic streaming,” Lab Chip, vol. 17, no. 10, pp. 1769–1777, 2017, doi: 10.1039/C7LC00215G. [204] Y. Choe, J. W. Kim, K. K. Shung, and E. S. Kim, “Microparticle trapping in an ultrasonic Bessel beam,” Appl. Phys. Lett., vol. 99, no. 23, p. 233704, Dec. 2011, doi: 10.1063/1.3665615. [205] D. Baresch, J. L. Thomas, and R. Marchiano, “Spherical vortex beams of high radial degree for enhanced single-beam tweezers,” J. Appl. Phys., vol. 113, no. 18, 2013, doi: 10.1063/1.4803078. [206] F. G. Mitri, “Airy acoustical–sheet spinner tweezers,” J. Appl. Phys., vol. 120, no. 10, p. 197 104901, Sep. 2016, doi: 10.1063/1.4962397. [207] P. Zhang et al., “Generation of acoustic self-bending and bottle beams by phase engineering,” Nat. Commun., vol. 5, no. 1, p. 4316, Sep. 2014, doi: 10.1038/ncomms5316. [208] Q. Zhou, J. Zhang, X. Ren, Z. Xu, and X. Liu, “Multi-bottle beam generation using acoustic holographic lens,” Appl. Phys. Lett., vol. 116, no. 13, p. 133502, Mar. 2020, doi: 10.1063/5.0003379. [209] U. Bar-Ziv, A. Postan, and M. Segev, “Observation of shape-preserving accelerating underwater acoustic beams,” Phys. Rev. B, vol. 92, no. 10, p. 100301, Sep. 2015, doi: 10.1103/PhysRevB.92.100301. [210] H. Bruus, “Acoustofluidics 7: The acoustic radiation force on small particles,” Lab Chip, vol. 12, no. 6, p. 1014, 2012, doi: 10.1039/c2lc21068a. [211] P. Glynne-Jones, P. P. Mishra, R. J. Boltryk, and M. Hill, “Efficient finite element modeling of radiation forces on elastic particles of arbitrary size and geometry,” J. Acoust. Soc. Am., vol. 133, no. 4, pp. 1885–1893, Apr. 2013, doi: 10.1121/1.4794393. [212] K.-T. Song, C.-H. Chen, and C.-H. C. Huang, “Design and experimental study of an ultrasonic sensor system for lateral collision avoidance at low speeds,” in Proc. IEEE Intelll. Vehic. Symp., 2004, pp. 647–652, doi: 10.1109/IVS.2004.1336460. [213] Y.-F. Wang et al., “High-density pMUT array for 3-D ultrasonic imaging based on reverse- bonding structure,” in Proc. IEEE 24th Int. Conf. on Micro Electro Mech. Syst., Jan. 2011, pp. 1035–1038, doi: 10.1109/MEMSYS.2011.5734605. [214] A. Schibler and R. Henning, “Measurement of Functional Residual Capacity in Rabbits and Children Using an Ultrasonic Flow Meter,” Pediatr. Res., vol. 49, no. 4, pp. 581–588, Apr. 2001, doi: 10.1203/00006450-200104000-00022. [215] P. M. Duggan, G. C. Liggins, and S. B. Barnett, “Ultrasonic heating of the brain of the fetal sheep in utero,” Ultrasound Med. Biol., vol. 21, no. 4, pp. 553–560, Jan. 1995, doi: 10.1016/0301- 5629(94)00143-2. [216] C.-H. Han and E. S. Kim, “Fabrication of dome-shaped diaphragm with circular clamped boundary on silicon substrate,” in Proc. IEEE 12th Int. Conf. on Micro Electro Mech. Syst., 1999, pp. 505–510, doi: 10.1109/MEMSYS.1999.746880. [217] M. Apanius, P. B. Kaul, and A. R. Abramson, “Silicon shadow mask fabrication for patterned metal deposition with microscale dimensions using a novel corner compensation scheme,” Sensors Actuators A Phys., vol. 140, no. 2, pp. 168–175, Nov. 2007, doi: 10.1016/j.sna.2007.06.028. [218] “Material Property Database-PECVD Silicon Nitride.” http://www.mit.edu/~6.777/matprops/pecvd_sin.htm. 198 [219] “Material Property Database-LPCVD Silicon Nitride,” [Online]. Available: http://www.mit.edu/~6.777/matprops/lpcvd_sin.htm. [220] B. L. Smith and A. Glezer, “The formation and evolution of synthetic jets,” Phys. Fluids, vol. 10, no. 9, pp. 2281–2297, Sep. 1998, doi: 10.1063/1.869828. [221] U. Ingård and S. Labate, “Acoustic Circulation Effects and the Nonlinear Impedance of Orifices,” J. Acoust. Soc. Am., vol. 22, no. 2, pp. 211–218, Mar. 1950, doi: 10.1121/1.1906591. [222] S. G. Sawant, B. George, L. S. Ukeiley, and D. P. Arnold, “Microfabricated Electrodynamic Synthetic Jet Actuators,” J. Microelectromechanical Syst., vol. 27, no. 1, pp. 95–105, Feb. 2018, doi: 10.1109/JMEMS.2017.2769445. [223] S. G. Mallinson, J. A. Reizes, and G. Hong, “An experimental and numerical study of synthetic jet flow,” Aeronaut. J., vol. 105, no. 1043, pp. 41–49, Jan. 2001, doi: 10.1017/S0001924000095968. [224] J. E. Cater and J. Soria, “The evolution of round zero-net-mass-flux jets,” J. Fluid Mech., vol. 472, pp. 167–200, Dec. 2002, doi: 10.1017/S0022112002002264. [225] M. M. Kanase, L. D. Mangate, and M. B. Chaudhari, “Acoustic aspects of synthetic jet generated by acoustic actuator,” J. Low Freq. Noise, Vib. Act. Control, vol. 37, no. 1, pp. 31–47, Mar. 2018, doi: 10.1177/1461348418757879. [226] D. J. Coe, M. G. Allen, B. L. Smith, and A. Glezer, “Addressable Micromachined Jet Arrays,” in Proc. 4th Int. Conf. Solid-State Sens. Actuators and Microsyst., 1995, vol. 2, pp. 329– 332, doi: 10.1109/SENSOR.1995.721813.

Abstract (if available)

Abstract With thoughtful engineering and arrangement, acoustic waves can be used as powerful and versatile tools for many applications. For generating acoustic waves and ultrasound, piezoelectric acoustic/ultrasonic transducers have been widely used due to their low loss, low cost, low power consumption, high acoustic output, and easy operation. Moreover, with acoustic lenses, the generated acoustic waves can be effectively modulated, realizing focusing effect or creating complex beam patterns without relying on an array of transducers. With microelectromechanical system (MEMS) technologies, piezoelectric acoustic transducers and acoustic lenses with smaller footprints, higher performance, and lower cost could be mass-produced with high precision through microfabrication processes. In this thesis, our research on the design, fabrication, and applications of four types of piezoelectric acoustic/ultrasonic transducers (two of them are equipped with acoustic lenses) based on MEMS technologies. ? The first type is self-focusing acoustic transducers (SFAT), which are single-element, planar ultrasonic transducers based on piezoelectric lead zirconate titanate (PZT) substrates vibrating in thickness mode with microfabricated Fresnel acoustic lenses on top to generate focused ultrasound. The design parameters, as well as microfabrication and characterization methods of SFATs, are introduced. Exploiting their small size, electrical tunability, and microfabrication compatibility, different types of SFATs tailored for applications such as in vivo selective cancer treatment, neurostimulation, and acoustic droplet ejection (for cell extraction/delivery and semiconductor chip pick-and-placement) have been developed. ? The second type is single-element microfabricated acoustic tweezers, which are based on modified Fresnel acoustic lenses to achieve trapping and manipulation of large (sub-millimeter- or millimeter-sized) particles in liquid environments, with two designs demonstrated. The first design is based on a ring-focusing Fresnel lens, which is capable of generating multiple acoustic trapping beams such as bottle beams (with three-dimensional trapping force) and quasi-Airy beams (with the capability to rotate trapped particles). The second design relies on a multi-foci linear Fresnel lens to generate large, cuboid-shaped trapping zones that can trap plastic and silicon chips with up to 5.3?0.67?0.51 mm? in size and 3.98 mg in weight. ? The other two types are both airborne acoustic transducers based on diaphragms vibrating in flexural mode. The third type is a piezoelectric micromachined ultrasonic transducer (PMUT) based on a hemispherical dome-shaped diaphragm driven by eight piezoelectric zinc oxide (ZnO) actuation elements designed to increase the sound output through releasing residual stress and transforming in-plane vibration to large radial deflection. The fourth type is an acoustic propeller capable of generating propulsion force from synthetic air jets generated by acoustic waves passing through small orifices on a thin plastic covering a mini membrane-type speaker. With different driving conditions and operating frequencies, the transducer that weighs 603 mg can rotate, jump, move, or lift objects, propelled by the air jets. The device could also be driven wirelessly with acoustic signals instead of electrical power. ? Finally, a summary of current research accomplishments, as well as future research directions will be presented.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Integrated wireless piezoelectric ultrasonic transducer system for biomedical applications

PDF

Audio and ultrasound MEMS built on PZT substrate for liquid and airborne applications

PDF

Magnetic spring in electromagnetic vibration energy harvester and applications of focused ultrasonic transducer

PDF

Development of high frequency focused transducers for single beam acoustic tweezers

PDF

Piezoelectric MEMS for acoustic sensing, contactless RF switching, and resonant mass sensing

PDF

Additive manufacturing of piezoelectric and composite for biomedical application

PDF

MEMS piezoelectric resonant microphone arrays and their applications

PDF

Quantification of cellular properties using ultra-high frequency single-beam acoustic tweezer

PDF

Fabrication of ultrasound transducer and 3D-prinitng ultrasonic device

PDF

Zero-power sensing and processing with piezoelectric resonators

PDF

Design and development of ultrasonic array transducers for specialized applications

PDF

Highly integrated 2D ultrasonic arrays and electronics for modular large apertures

PDF

High-frequency ultrasonic transducers for photoacoustic applications

PDF

High frequency ultrasonic phased array system and its applications

PDF

High-frequency ultrasound array-based imaging system for biomedical applications

PDF

Surface acoustic wave waveguides for signal processing at radio frequencies

PDF

2D ultrasonic transducer array’s design and fabrication with 3D printed interposer and applications

PDF

High frequency ultrasound elastography and its biomedical applications

PDF

Advanced applications of high frequency single beam acoustic tweezers in fluid and cell mechanics

PDF

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

Asset Metadata

Creator Tang, Yongkui (author)

Core Title Piezoelectric ultrasonic and acoustic microelectromechanical systems (MEMS) for biomedical, manipulation, and actuation applications

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Electrical Engineering

Degree Conferral Date 2021-08

Publication Date 07/21/2021

Defense Date 06/04/2021

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

Tag acoustic actuators,acoustic droplet ejection,acoustic lens,acoustic neuromodulation,acoustic propellers,acoustic transducers,acoustic trapping,acoustic tweezers,focused ultrasound (FUS),Fresnel acoustic lens,microelectromechanical systems (MEMS),microfabrication,OAI-PMH Harvest,piezoelectric micromachined ultrasonic transducers (PMUT),piezoelectric transducers,self-focusing acoustic transducers (SFAT),ultrasonic transducers,ultrasound cancer treatment

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Kim, Eun Sok (committee chair), Meng, Ellis Fan-Chuin (committee member), Wu, Wei (committee member)

Creator Email tangyk1992@gmail.com,yongkuit@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC15613323

Unique identifier UC15613323

Legacy Identifier etd-TangYongku-9735

Document Type Dissertation

Format application/pdf (imt)

Rights Tang, Yongkui

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu