SINGLE MASK, MULTIPLE LEVEL SINGLE CRYSTAL SILICON PROCESSES: MEMS OUT-OF-PLANE ACTUATORS AND 3-D WIRE ARRAYS A Dissertation Presented to the Faculty of the Graduate School of Cornell University in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy by Kanakasabapathi Subramanian May 2003 © 2003 Kanakasabapathi Subramanian SINLGE M ASK, MULTIPLE LEVEL SINLGE CRYSTAL SILICON PROCESSES: MEMS OUT-OF-PLANE ACTUATORS AND 3-D WIRE ARRAYS Kanakasabapathi Subramanian, Ph.D. Cornell University 2003 A multilevel process and a wire process have been developed in single crystal silicon with applications in Micro ElectroMechanical Systems (MEMS). Both processes use a single mask level and can be combined with high aspect ratio bulk micromachining methods such as SCREAM (Single Crystal Reactive Etching and Metallization). The multilevel process is used to design and fabricate a micro-mirror device with a resonance frequency of 50 kHz and 20o tilt about a torsional axis. Applications of the wire process include three-dimensional photonic bandgap structures, mechanical oscillators, three-dimensional fluid channel arrays and three-dimensional arrays of self aligned lateral tips. The multilevel process uses thermal oxidation to place structures with differing line widths on different levels without requirements of self-alignment. Structures with similar line widths are copied on to a lower level in a self-aligned fashion. The out-of- plane (Z) asymmetry created by this process is used to make a Z directional actuator. Total overlap Z drives employ a simple process but they have poor force and displacement characteristics and are shown to work for small displacements. Partial overlap drives are true Z comb drives and employ an innovative extension to the basic multilevel process. While detailed fabrication results are presented, motion is not proven on these actuators. A micro-mirror device is designed using the partial overlap Z drive concept. Electromechanical design, process design and fabrication results are described along with relevant trade-offs. The device has not shown any movement for lack of reliable back side alignment. The single crystal silicon wire process uses deposition-etch cycling in deep silicon etching to create released wires in each cycle. Large 3-D arrays of wires are demonstrated using this process. An alternative method of making wires using thermal oxidation is also presented. Wires less than 100nm in size and arrays as large as 9000 wires are shown. Three-dimensional cubic and hexagonal arrays have been demonstrated. Finally, it is proposed that the multilevel and wire processes can be combined to create a stepping actuator in the Z direction. BIOGRAPHICAL SKETCH K. Subramanian was born in Bombay in the west coast of central India in 1974. After a short stay with his grandparents, he moved to Podanur in the south where his father was working for the Indian Railways. In 1979, he started formal schooling and moved to Madras in the southeastern coast of India, home to the second longest beach in the world. This also coincided with the birth of a baby brother in Palakkad, Kerala, who is working on his PhD at UC Berkeley at this time. The next twelve years were spent schooling at Chinmaya Vidyalaya where he topped the class almost every year and was a personal favourite of the teachers. After graduating from high school in 1991, he took the Joint Entrance Examination for the Indian Institutes of Technology, reputed to be the best and the most difficult undergraduate admission test in the world. After having made it into the top 2500 or so in the country, he enrolled in Civil Engineering and later moved to Mechanical Engineering, despite his father’s aspirations of a medical career for his son. The American bug bit him in 1993 when he saw his older friends doing interesting research abroad. A successful project in robotics and vision systems and an excellent GPA set the path to Cornell University in 1995. Working in materials processing for a year, he found out about Micro ElectroMechanical Systems in a Scientific American article during the fall of 1995. This got him into Prof. Noel C. MacDonald’s research group, where he changed his major to Electrical and Computer Engineering. The years at Cornell helped shape him into a confident and outgoing person and helped him discover several of his current interests, including travel, cooking, music and philosophy. He followed his advisor to UC Santa Barbara in 2000. In August 2001, he married the most beautiful, talented and encouraging Vidya. Several trips to Ithaca and Stanford, a few patents and a lot of late nights and weekends later, he finally decided to defend his thesis and enter the real world. iii To Vidya, Rahul, Appa & Amma for always giving me their best iv ACKNOWLEDGEMENTS I would like to thank my mother and father for making me who I am. My sincere thanks to my thesis advisor, Prof. Noel MacDonald for everything I know in MEMS. I would like to thank him for his support, his undying drive towards focus and his keen intuition. I would also like to thank my other committee members, the late Prof. J. P. Krusius and Prof. Subrata Mukherjee for their help and constant support. Thanks to Prof. Amit Lal for joining my committee at the last minute. My time at Cornell was enriched by the entire MacDonald group research team. Trent Huang has been a friend, philosopher and guide. He has always been available to talk whenever there was a problem with research. He even helped me build my own computer. Special thanks to Trent for starting me off on the multilevel process. Dr. Jia Chen, Dr. Kim Turner, Dr. Mike Wolfson, Dr. Pete Hartwell, Dr. Yang Xu, Dr. Taher Saif, Dr. Madanagopal, Dr. Chris Lee, Dr. John Chong and Raji have been good friends. I would like to thank the CNF staff for helping with processing issues, especially associate director Dr. Alex Pechenik who stood by me. Scott Coldren in the ECE department has been an invaluable resource in taking care of paper work and other formalities. At UC Santa Barbara, I have had invaluable help from Masa, Anton, Garrett and Seth. Davide taught me the basics of optical testing before flying back to Italy to resume his PhD. Brainstorming sessions with Masa always recharged my sprits. Anton, Seth and Garrett contributed their share of ideas and were extremely useful in the lab. Additional thanks to Garrett for his suggestions in photonics, to Anton for providing the mechanics expertise, to Seth for the stepping motor simulations and to Masa for v coordinating everything and making things happen. Life in 2145 EII would not have been the same without Alok, Adam, Lori and Dave. Special thanks to Dave Bothman for helping with my job search and making sure that our equipment always ran well. Marco gave me my late night booster shots with his undying enthusiasm and prompt feedback on ideas and proofs. Brian Thibeault at the UCSB clean room was a store house of ideas and Jack Whaley has always been there to help. It is impossible to mention the long list of people who helped enrich my social life in the USA. The list begins with Sudhir, Seethu, Ramesh, Indira, Ram, Ganja, Reddy, Preeti, Madhu, Manu, Kirti, Bhavin, Parvathi, Mahesh, Kiddo, Venkat, Giri, Rukmini, Grammy, 2D, VT, Vasu, Maria and the rest of the talkies gang. Last but certainly not the least, this thesis would not have been possible without the encouragement from my wife Vidya and my brother Rahul. Special thanks to them and to God, Almighty. This work was performed in part at the Cornell Nanofabrication Facility (CNF), the clean room at UC Santa Barbara and the Stanford Nanofabrication Facility (SNF), all members of the National Nanofabrication Users Network. The work done at Cornell University until 2000 was supported by DARPA and several trips to the CNF in 2001- 2002 were supported by DARPA contract DABT63-95-c-0121, Cornell University subcontract to UCSB. The work done at UC Santa Barbara and at the SNF in 2002- 2003 was supported by the DARPA funded subcontract RF742404 from Ohio State University. vi TABLE OF CONTENTS CHAPTER ONE: INTRODUCTION 1 1.1Micromachining Technologies 1 1.2The SCREAM Process 3 1.3SCREAM: Scope for Improvement 7 1.4The Single Mask SCS Multilevel Process 8 1.5The Single Mask SCS Wire Process 9 1.6Stepping Electrostatic Actuators 9 1.7Outline of the Thesis 10 CHAPTER TWO: MULTIPLE LEVEL BULK MEMS 15 2.1 Introduction 15 2.2 Background 15 2.3 Process Description 18 2.3.1 Two Level Process 18 2.3.2 Three level and multilevel processes 20 2.4 Processing Example: Out of Plane Actuator 23 2.4.1 Principle 23 2.4.2 Actuator Design 24 2.4.3 Two Level Z Actuator Process Sequence 30 2.4.4 Isolation Requirements 34 vii 2.5 Problems and Issues with Z Actuator Total Overlap Design 36 2.5.1 Effect of Finger Gap 36 2.5.2 Effect of first RIE Etch Depth on Maximum Force 39 2.5.3 Effect of Lower Level RIE Etch Relative to RIE 1 41 2.5.4 Effect on In-Plane Finger Length 44 2.5.5 Effect of Voltage 45 2.5 Conclusions 46 CHAPTER THREE: TRUE COMB DRIVES IN Z – SHADOW MASKING AND ELECTRICAL ISOLATION 54 3.1 Introduction 54 3.2 Three Level Shadow Masking 56 3.3 Z Actuator with Three Level Shadow Masking 57 3.4 Vertical Gap 65 3.5 Vertical Isolation 66 3.6 Characteristics of Two and Three Level Z Comb Drives 67 3.6.1 Constant Force and Large Range 67 3.6.2 Effect of Finger Gap 68 3.6.3 Effect of In-Plane Finger Length 69 3.6.4 Effect of Voltage 70 3.6.5 Effect of Vertical Separation between Levels 71 3.7 Vertical Motion Approximation 79 3.8 Conclusions 81 viii