USF Libraries
USF Digital Collections

Nanocrystalline diamond for RF MEMS applications

MISSING IMAGE

Material Information

Title:
Nanocrystalline diamond for RF MEMS applications
Physical Description:
Book
Language:
English
Creator:
Balachandran, Srinath
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
NCD
Charging
High power
Capacitive switch
DC switch
Dissertations, Academic -- Electrical Engineering -- Doctoral -- USF   ( lcsh )
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: Nanocrystalline diamond (NCD) due its outstanding thermal, mechanical and tribological properties is an ideal candidate for MEMS/NEMS devices. NCD offers the possibility to increase the reliability and life time of RF-MEMS switches and by mitigating the problems of stiction, charge trapping, surface wear and cold welding found in traditional all metal MEMS devices. In this work, nanocrystalline diamond cantilever beams and bridges have been fabricated on a low resistive silicon substrate by using standard micromachining techniques. The diamond structures are then integrated onto alumina and aluminium nitride substrates upon which microwave transmission lines in the microstrip and coplanar waveguide (CPW) topology have been fabricated. The diamond actuators are integrated using a combined soldering and flip chip technique.The NCD bridges are thermally actuated wherein the difference in coefficient of thermal expansion between copper and diamond bends the diamond bridge thus moving the bridges to the actuated state. In the CPW topology, RF-MEMS switches and tunable planar inductors are realized using the micromachined devices. These devices are mounted on a 650 micrometer thick alumina substrate and the microwave characteristics are analyzed in the frequency range of 5-30 GHz. The switches yield a return loss of 15 dB and an insertion loss of 0.2 dB at 20GHz. An inductance ratio of 2.2 is achieved by the tunable inductors at 30 GHz. High power measurements are performed on the diamond actuators which utilize a dual actuation scheme which comprises of thermal and electrostatic actuation. The measurements are performed on the diamond actuators in the power range of 24-47 dBm for the mechanically actuated switches, and 24-40 dBm for electrically actuated switches.The measurements show an insertion loss of 0.2-03 dB in the entire power spectrum. NCD based RF-MEMS capacitive switches is also designed, fabricated and tested. The switches are fabricated on a high resistive silicon substrate and are electrostatically actuated. Small signal measurements are presented in the frequency range of 1-65 GHz. The measured insertion loss in the up-state is 1.1 dB at 50 GHz with 30 dB isolation in the down-state. Dielectric characterization is performed using the Corona-Kelvin technique and the standard I-V and C-V stress tests for nitride and diamond films. The leaky nature of the diamond films provides a potential solution to reliability issues related to dielectric charging.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2009.
Bibliography:
Includes bibliographical references.
System Details:
Mode of access: World Wide Web.
System Details:
System requirements: World Wide Web browser and PDF reader.
Statement of Responsibility:
by Srinath Balachandran.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 117 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002317427
oclc - 660062534
usfldc doi - E14-SFE0003042
usfldc handle - e14.3042
System ID:
SFS0027359:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 2200385Ka 4500
controlfield tag 001 002317427
005 20100901165512.0
007 cr cnu|||uuuuu
008 100901s2009 flua ob 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0003042
035
(OCoLC)660062534
040
FHM
c FHM
049
FHMM
090
TK145 (Online)
1 100
Balachandran, Srinath.
0 245
Nanocrystalline diamond for RF MEMS applications
h [electronic resource] /
by Srinath Balachandran.
260
[Tampa, Fla.] :
b University of South Florida,
2009.
500
Title from PDF of title page.
Document formatted into pages; contains 117 pages.
Includes vita.
502
Dissertation (Ph.D.)--University of South Florida, 2009.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
520
ABSTRACT: Nanocrystalline diamond (NCD) due its outstanding thermal, mechanical and tribological properties is an ideal candidate for MEMS/NEMS devices. NCD offers the possibility to increase the reliability and life time of RF-MEMS switches and by mitigating the problems of stiction, charge trapping, surface wear and cold welding found in traditional all metal MEMS devices. In this work, nanocrystalline diamond cantilever beams and bridges have been fabricated on a low resistive silicon substrate by using standard micromachining techniques. The diamond structures are then integrated onto alumina and aluminium nitride substrates upon which microwave transmission lines in the microstrip and coplanar waveguide (CPW) topology have been fabricated. The diamond actuators are integrated using a combined soldering and flip chip technique.The NCD bridges are thermally actuated wherein the difference in coefficient of thermal expansion between copper and diamond bends the diamond bridge thus moving the bridges to the actuated state. In the CPW topology, RF-MEMS switches and tunable planar inductors are realized using the micromachined devices. These devices are mounted on a 650 micrometer thick alumina substrate and the microwave characteristics are analyzed in the frequency range of 5-30 GHz. The switches yield a return loss of 15 dB and an insertion loss of 0.2 dB at 20GHz. An inductance ratio of 2.2 is achieved by the tunable inductors at 30 GHz. High power measurements are performed on the diamond actuators which utilize a dual actuation scheme which comprises of thermal and electrostatic actuation. The measurements are performed on the diamond actuators in the power range of 24-47 dBm for the mechanically actuated switches, and 24-40 dBm for electrically actuated switches.The measurements show an insertion loss of 0.2-03 dB in the entire power spectrum. NCD based RF-MEMS capacitive switches is also designed, fabricated and tested. The switches are fabricated on a high resistive silicon substrate and are electrostatically actuated. Small signal measurements are presented in the frequency range of 1-65 GHz. The measured insertion loss in the up-state is 1.1 dB at 50 GHz with 30 dB isolation in the down-state. Dielectric characterization is performed using the Corona-Kelvin technique and the standard I-V and C-V stress tests for nitride and diamond films. The leaky nature of the diamond films provides a potential solution to reliability issues related to dielectric charging.
538
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
590
Advisor: Thomas M. Weller, Ph.D.
653
NCD
Charging
High power
Capacitive switch
DC switch
690
Dissertations, Academic
z USF
x Electrical Engineering
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.3042



PAGE 1

Nanocrystalline Diamond for RF MEMS Applications by Srinath Balachandran A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Electrical Engineering College of Engineering University of South Florida Major Professor: Thomas M. Weller, Ph.D. Shekhar Bhansali, Ph.D. Jing Wang, Ph.D. Ashok Kumar, Ph.D. Sarath Witanachchi, Ph.D. Date of Approval: June 15, 2009 Keywords: ncd, charging, high pow er, capacitive switch, dc switch Copyright 2009 Srinath Balachandran

PAGE 2

Dedication This dissertation is dedicated to my late parents and late grand parents for the values they taught and instilled in me and also for helping me understand the importance of family and friends in life.

PAGE 3

i Table of Contents List of Tables................................................................................................................. ....iv List of Figures................................................................................................................ ......v Abstract....................................................................................................................... .........x Preface........................................................................................................................ .......xii Chapter 1 Introduction....................................................................................................... 1 1.1 Overview............................................................................................... .....1 1.2 Dissertation Organization..........................................................................5 1.3 Contributions..............................................................................................6 Chapter 2 RF-MEMS – An Overview of the Technology and its Reliability Issues.........7 2.1 Introduction ..............................................................................................7 2.2 Overview of RF-MEMS Switches............................................................8 2.2.1 Capacitive Switches.....................................................................10 2.3 Actuation Schemes in RF-MEMS..........................................................14 2.4 Failure Mechanisms of RF-MEMS Switches.........................................17 2.4.1 Dielectric Charging......................................................................17 2.4.2 Contact Material Issues................................................................22 2.5 Power Handling Capabilities..................................................................23 2.6 Summary.................................................................................................26

PAGE 4

ii Chapter 3 Nanocrystalline Diamond – Properties, Grow th and Characterization...........27 3.1 Introduction.............................................................................................27 3.2 Structure and Growth of Nanocrystalline Diamond (NCD)...................28 3.3 Characterization Techniques for NCD Films.........................................34 3.4 Mechanical Characteristics of NCD Films.............................................36 3.4.1 Young’s Modulus.........................................................................36 3.4.2 Intrinsic Stress..............................................................................38 3.5 Summary.................................................................................................43 Chapter 4 Thermal and Dual Actuation Nanocrystalline Diamond Bridges and Cantilevers...................................................................................44 4.1 Introduction.............................................................................................44 4.2 Design of the NCD Actuator...................................................................45 4.3 Fabrication..............................................................................................50 4.4 Small Signal Analysis of the CPW Integrated Switch............................55 4.5 Small Signal Analysis of the CPW Inductor...........................................56 4.6 Large Signal Measurements – 1st Generation.........................................61 4.7 Dual Mode Actuation of the NCD Switches...........................................65 4.8 Large Signal Measurements – 2nd Generation........................................67 4.9 Summary.................................................................................................70 Chapter 5 Nanocrystalline Diam ond Capacitive Shunt Switches....................................71 5.1 Introduction.............................................................................................71 5.2 Design and Simulation Results...............................................................72 5.3 Material Characterization and Fabrication..............................................77 5.4 Small Signal Measurements and Analysis..............................................80 5.5 Corona Kelvin Measurements.................................................................82 5.6 Stressed I-V and C-V Measurements – 1st Generation...........................84 5.7 Stressed I-V and C-V Measurements – 2nd Generation..........................89 5.8 Summary.................................................................................................95

PAGE 5

iii Chapter 6 Summary and Recommendations....................................................................96 6.1 Summary.................................................................................................96 6.2 Recommendations for Future Work........................................................98 References..................................................................................................................... ...101 Appendices..................................................................................................................... ..109 Appendix A Photolithography Procedures.................................................110 Appendix B Fabrication of the Diamond Actuator and Host Substrate........................................................................112 Appendix C Fabrication of the Capacitive Shunt Switch...........................116 About the Author...................................................................................................End Page

PAGE 6

iv List of Tables Table 2.1 Performance Comparis on Between FETs, PIN Diodes and RF-MEMS Switches..........................................................................................8 Table 2.2 State of Art in High Power RF-MEMS Switches............................................24 Table 3.1 Mechanical Properties of NCD Thin Films in Comparison to Materials Used in Microsystems Technology..................................................28 Table 3.2 Growth Recipe for NCD Films Using Hot Filament (HFCVD) and Microwave Plasma (MPECVD) Process..................................................31 Table 3.3 Measured Resonance Frequency and Young’s Modulus for Diamond Cantilevers.......................................................................................38 Table 3.4 Measured Intrinsic Stress of a 1.2 m NCD Film...........................................42 Table 4.1 Comparison of Young’s M odulus (E) and Coefficient of Thermal Expansion ( th) for Different Materials............................................49 Table 5.1 Comparison of Roughness of Metal + Dielectric Stack Before and After Deposition............................................................................78 Table 5.2 Comparison of the Lumped Element Values of the Capacitive Switch from Simulations and Measurement....................................................82

PAGE 7

v List of Figures Figure 1.1 Cross Section of the NCD Actuat or Integrated to a Host Substrate...............3 Figure 1.2 Design of the Capacitive S hunt Switch with NCD as a Dielectric.................4 Figure 2.1 Broadside MEMS Series Sw itch Utilizing Cantilever Beams [11]................9 Figure 2.2 Raytheon MEMS Capaci tive Shunt Switch Utilizing Fixed-Fixed Beam [11]...................................................................................9 Figure 2.3 Capacitive RF-MEMS Switc h in a Series Configuration.............................10 Figure 2.4 Equivalent Circuit of a Series Capacitive Switch.........................................11 Figure 2.5 S-Parameters of the Ca pacitive Series Switch in the OFF and ON States......................................................................................12 Figure 2.6 Capacitive RF-MEMS Sw itch in Shunt Configuration................................12 Figure 2.7 Equivalent Circuit of a Capacitive Shunt Switch.........................................13 Figure 2.8 S-Parameters of the Ca pacitive Shunt Switch in the ON and OFF States.......................................................................................13 Figure 2.9 Schematic Representation of the Actuation Mechanism in Capacitive RF-MEMS Switch.....................................................................18 Figure 2.10 Accumulation of Imag e Charges on the Electrodes During Actuation..........................................................................................19 Figure 2.11 Top View of the Fabr icated Schottky Contact RF-MEMS Switch [37]...................................................................................................21 Figure 2.12 Cross Section of DC Contact Switch in OFF and ON State [41].................22 Figure 2.13 SEM Image of DC Cont act Switch with Organic Deposits..........................23 Figure 3.1 Crystal Structure of Diamond Lattice [53]...................................................29 Figure 3.2 Varying Current Value in Time During the BEN Process............................31 Figure 3.3 SEM Image of a NCD Film Grown by the HFCVD Technique...................32

PAGE 8

vi Figure 3.4 SEM Images of Diamond Films Grown with Different Ar Ratios (a) 50% and (b) 98%..........................................................................35 Figure 3.5 Raman Spectra of Carbon Based Materials [54]...........................................36 Figure 3.6 SEM Image of Released Di amond Cantilever (a) Using HFCVD, and (b) Using MPECVD Technique.............................................................37 Figure 3.7 Released Structures Used fo r Measuring the Intrinsic Stress in Diamond Films..............................................................................................39 Figure 3.8 Cantilever Structure Used to Evaluate Intrinsic Stress.................................40 Figure 3.9 Geometry Used to Evaluate th e Rotation Angle to Calculate Strain ( ”).....40 Figure 3.10 SEM Image of the Released Cantilevers with Intrinsic Stress.....................41 Figure 3.11 Measured Intrinsic Stre ss with Varying Temperature and Pressure Conditions.....................................................................................42 Figure 3.12 Measured Compressive Di stribution Across a 4-inch Wafer.......................43 Figure 4.1 Design of the Thermally Actuated NCD Bridge...........................................45 Figure 4.2 Top View of the Diam ond Actuator with Different Copper Heating Elements.............................................................................46 Figure 4.3 Stages in a Straight Beam which is Compressively Stressed (a) EC is Greater than EBB and (b) EBB is Greater than EC.............................47 Figure 4.4 Simulated Force vs Distance of Separation Between Actuator and Host Substrate...............................................................................................48 Figure 4.5 Bi-Stable Layout of the Actuator with Individual Copper Heating Elements..........................................................................................50 Figure 4.6 Fabrication Procedure of the Nanocrystalline Diamond Actuator................52 Figure 4.7 Microphotograph of the Fabricated Diamond Actuator (a) Front View (b) Back Vi ew with the Silicon Frame.................................53 Figure 4.8 Diamond Actuator Integr ated onto a Host Substrate (Alumina, Aluminum Nitride) Using SOLID Process..................................53 Figure 4.9 Phase Transformation in the Cu-Sn SOLID Process....................................54 Figure 4.10 Integrated Switch with a Host Alumina Substrate in a Microstrip Topology.....................................................................................55

PAGE 9

vii Figure 4.11 S-Parameters of the CPW Integrat ed Actuator in the Non-Actuated State..56 Figure 4.12 S-Parameters of the CPW Integr ated Actuator in the Actuated State...........56 Figure 4.13 Design of the Integrated CPW Inductor and Diamond Actuator..................57 Figure 4.14 Change in Impedance a nd Effective Inductance Between the Up-State and Down-State of the Tunable Inductor.......................................59 Figure 4.15 Measured Return Loss (S11) of the Inductor in the Non-Actuated and Actuated State.................................................................59 Figure 4.16 Measured Insertion Loss (S21) of the Inductor in the Non-Actuated and Actuated State.................................................................60 Figure 4.17 Measured Inductance in Up and Down States and Inductance Ratio...........61 Figure 4.18 Diamond Actuator Integrat ed in Microstrip Topology for High Power Testing......................................................................................62 Figure 4.19 High Power Bench for Testing Diamond Switches......................................63 Figure 4.20 Measured Isolation of the Diamond Actuator in the Non-Actuated State with Varying Input Power............................................64 Figure 4.21 Measured Insertion Loss of the Diamond Actuator in the Actuated State with Varying Input Power....................................................65 Figure 4.22 Fabrication and Integration of the Dual Mode Actuation Scheme of the NCD Bridges......................................................................................66 Figure 4.23 Measured Insertion Loss at a CW Frequency of 1.9 GHz with Maximum Input Power Varying from 30 dBm to 40 dBm...........................68 Figure 4.24 Measured Isolation Loss Af ter Two Hours of Continuous Actuation of the Diamond Bridges................................................................................69 Figure 4.25 Measured Insert ion Loss After Two Hours of Continuous Actuation of the Diamond Bridges................................................................................70 Figure 5.1 Cross Section of the NCD Capacitive Switch...............................................72 Figure 5.2 Equivalent Circuit of the NCD Capacitive Shunt Switch.............................73 Figure 5.3 Top View of the Basi c and Inductively Tuned NCD Capacitive Switch.........................................................................................74

PAGE 10

viii Figure 5.4 Simulated Change in Resonant Frequency of the Inductively Tuned Shunt Switch in the Down-State for Varying Lengths of the High Impedance Line....................................................................................75 Figure 5.5 Simulated S11 and S21 for the Basic (Design 1) and Inductively Tuned (Design 2) Capacitive Switch in the Up-State...................................76 Figure 5.6 Simulated S11 and S21 for the Basic (Design 1) and Inductively Tuned (Design 2) Capacitive Switch in the Down-State..............................76 Figure 5.7 Comparison of Simulated and Modeled S21 in the Down-State...................77 Figure 5.8 Fabrication Procedure of the NCD Capacitive Switch.................................79 Figure 5.9 Microphotograph of th e Fabricated NCD Shunt Switch (a) Without Holes, (b) With Holes................................................................80 Figure 5.10 Comparison of Simulated and Measured S11 and S21 in the Up-State of the NCD Capacitive Shunt Switch............................................81 Figure 5.11 Comparison of Simulated and Measured S11 and S21 in the Down-State of the NCD Capacitive Shunt Switch.......................................81 Figure 5.12 Typical Setup for Corona Kelvin Measurement (CKM)..............................83 Figure 5.13 Voltage Decay of Nitride F ilm at Three Different Sites of Samples Through CKM Technique..............................................................83 Figure 5.14 Voltage Decay of NCD Film at Three Different Sites of Samples Through CKM Technique..............................................................84 Figure 5.15 Measured Leakage Current vs Voltage for Nitride Film with Different dv/dt Values...................................................................................85 Figure 5.16 Measured Leakage Curren t vs Voltage for NCD Film with Different dv/dt Values...................................................................................85 Figure 5.17 Stress Induced Leakage Cu rrent (SILC) for Nitride Films Stressed at 20 Volts for Different Periods of Time.......................................86 Figure 5.18 Stress Induced Leakage Cu rrent (SILC) for Nitride Films Stressed at 40 Volts for Different Periods of Time.......................................87 Figure 5.19 Stress Induced Leakage Current (SILC) for NCD Films Stressed at 20 Volts for Different Periods of Time.......................................88

PAGE 11

ix Figure 5.20 Stress Induced Leakage Current (SILC) for NCD Films Stressed at 40 Volts for Different Periods of Time.......................................88 Figure 5.21 Capacitance Measurement of the NCD Capacitors Before and After Stress Tests..........................................................................................89 Figure 5.22 Stress Induced Leakage Current (SILC) for Nitride MIM Capacitor with Dielectric Thickness of 1400 A0...........................................................90 Figure 5.23 Stress Induced Leakage Curre nt (SILC) for NCD MIM Capacitor with Dielectric Thickness of 1500 A0...........................................................90 Figure 5.24 Capacitance Measurement of 1500 A0 NCD MIM Capacitor......................91 Figure 5.25 Stress Induced Leakage Current (SILC) for Nitride MIM Capacitor with Dielectric Thickness of 5000 A0...........................................................92 Figure 5.26 Stress Induced Leakage Curre nt (SILC) for NCD MIM Capacitor with Dielectric Thickness of 5000 A0...........................................................92 Figure 5.27 Comparison of Stress Induced Leakage Current (SILC) in Log Scale for 5000 A0 NCD and Nitride Capacitors.....................................................93 Figure 5.28 Measured I-V Respons e for the NCD MEM Capacitor in the Up-State..............................................................................................94 Figure 5.29 Measured I-V Respons e for the NCD MEM Capacitor in the Down-State.........................................................................................94 Figure 6.1 Block Diagram for Switching Speed Measurement .....................................98 Figure 6.2 Layout of the MultiBit DMTL NCD Phase Shifter.....................................99

PAGE 12

x Nanocrystalline Diamond for RF MEMS Applications Srinath Balachandran ABSTRACT Nanocrystalline diamond (NCD) due its outstanding thermal, mechanical and tribological properties is an ideal candida te for MEMS/NEMS devices. NCD offers the possibility to increase the reliability and life time of RF-MEMS switches and by mitigating the problems of stiction, charge tr apping, surface wear and cold welding found in traditional all metal MEMS devices. In this work, nanocrystalline diamond cantilever beams and bridges have been fabricated on a low resistive silicon s ubstrate by using sta ndard micromachining techniques. The diamond structures are then integrated onto alumina and aluminium nitride substrates upon whic h microwave transmission lines in the microstrip and coplanar waveguide (CPW) topology have be en fabricated. The diamond actuators are integrated using a combined soldering and flip chip technique. The NCD bridges are thermally actuated wherein the difference in coefficient of thermal expansion between copper and diamond bends the diamond bridge thus moving the bridges to the actuated state. In the CPW topology, RF-MEMS switc hes and tunable planar inductors are realized using the micromachined devi ces. These devices are mounted on a 650 m thick alumina substrate and the microwave characteri stics are analyzed in the frequency range of 5-30 GHz. The switches yield a return loss of 15 dB and an insert ion loss of 0.2 dB at 20GHz. An inductance ratio of 2.2 is achieved by the tunable inducto rs at 30 GHz. High power measurements are performed on th e diamond actuators which utilize a dual actuation scheme which comprises of th ermal and electrostatic actuation. The measurements are performed on the diamond act uators in the power range of 24-47 dBm

PAGE 13

xi for the mechanically actuated switches, and 24-40 dBm for electrically actuated switches. The measurements show an insertion loss of 0.2-03 dB in the entire power spectrum. NCD based RF-MEMS capacitive switches is al so designed, fabri cated and tested. The switches are fabricated on a high resistiv e silicon substrate and are electrostatically actuated. Small signal measurements are pres ented in the frequency range of 1-65 GHz. The measured insertion loss in the up-state is 1.1 dB at 50 GHz with 30 dB isolation in the down-state. Dielectric ch aracterization is performe d using the Corona-Kelvin technique and the standard I-V and C-V stre ss tests for nitride and diamond films. The leaky nature of the diamond films provides a pot ential solution to relia bility issues related to dielectric charging.

PAGE 14

xii Preface I sincerely thank my major professor a nd advisor, Dr. Tom Weller for allowing me to work in this project. His continued support, encouraging words and the confidence he had in me helped me overcome various se tbacks throughout the project. His technical insight, patience and humility make him one of th e best advisors one could ask for. It was a pleasure working for and with him over th e years and I thank him for making me a better person both from a prof essional and personal front. I would like thank the committee members, Dr. Shekhar Bhansali, Dr. Jing Wang, Dr. Ashok Kumar and Dr. Sarath Witanachch i for taking the time to serve on my committee. Furthermore, I thank the dean for the college of engineering, Dr. John Wiencek for chairing my dissertation defens e. Along with Dr. Weller, Dr. Dunleavy was instrumental in setting up excellent measur ement facilities and I would like thank them for taking great efforts in making them availa ble to us. Many thanks to Dr. Bhansali and Dr. Kumar for providing good fabr ication facilities in the college and for their timely suggestions in the dissertation work. Dr. A ndrew Hoff was very helpful in sharing his expertise in charging measurements. I am very thankful to him for always being there to answer my numerous questions and offer fruitf ul suggestions. I would also like to thank Dr. Jeremy Muldavin for his timely and valu able inputs regarding charging experiments. The fabrication portion of the diamond act uator portion of this dissertation was carried out through research collaboration at th e University of Ulm. I am deeply indebted to Prof. Erhard Kohn from the institute of electron devices and circuits for allowing me to carry out this research with his group. During the two stints of my stay at the University, I had the great opportunity of learning and worki ng on state of art facilit ies in the institute. I am very thankful to Joachim Kusterer, Michele Dipalo and David Maier for all the help they had extended during my stay at the univers ity. I have indeed learned a lot from them and I am deeply indebted to them for acco mmodating me in their research projects.

PAGE 15

xiii Over the years, the staff at the Nanoma terials and Nanomanufacturing Center at USF has been extending round the clock support to us, the researchers. The tireless efforts of Mr. Robert Tufts, Mr. Rich Everly and others are sincer ely appreciated. I would personally like to thank Rich for going out of his way in helping me with many fabrication issues. Having spent more than seven years at USF, I have had the opportunity of working with some outstanding researchers. The former members of the WAMI program, Bala, Thomas, Chris, Saravana, Hari, Lester Jason, Eid and Suzette have been very helpful in various aspects of my work and I would like to sincerel y thank them for their help and friendship. Many thanks to the curren t members, Sergio, Bojana, Scott, James, Quenton, Evelyn, Tony and many others for all your support and friendship. Furthermore, I would like to thank the past and presen t members of Dr. Dunleavy’s group, Sathya, Alberto, Sriram and Siva for teaching and sharing your knowledge in the nuances of measurement. Shyam, Subbu, Puneet and Supr iya from Dr. Bhansali’s research group were a pleasure to work with and I would lik e to thank them for all the help and support they have given over the years to maintain a good facility. Harish and Humberto from Dr. Kumar’s research group were a pleasure to work with and they have been extremely helpful in sharing their insights in diamond growth and characterization. There are numerous people at USF who have become great friends over the years and it was an absolute pleasure knowing each and everyone. My past and present room mates, Madhan, Siva, Venkatesh, Balaji, Sarava na, Shyam and KarthikI consider them as family and it was a pleasure staying with you over the years and I would like to thank everyone for their unconditional love and suppor t. In addition to being my former room mate, Bala has and continues to be an excellent mentor. Apart from teaching many aspects of work, Bala was instrumental in instilling the importance and value of education and hard work. I like to thank him for his support, encouragement and friendship over the years. I would also like to thank Harish Sankara narayanan for all the help he has offered with understanding th e fundamental concepts in semiconductor physics and processing tools. He has also been a great friend and a me ntor over the years.

PAGE 16

xiv My friends from my high school, undergraduate studies, Houston, Germany and Indiamany thanks to each and everyone for your support and friendship. It was a pleasure knowing you all and I va lue your friendship a lot. Finally, I would like to thank my sister, Vidya and my uncles, Krishnan and Sekar for their continued support th roughout my education and also standing by me in all the tough times. I am for ever indebted for what they have done for me. My brother in law, Kumar for his encouragement and motivation, and the entire Sivadas and Munirathnam families for their love, affection and encouragement.

PAGE 17

1 Chapter 1 Introduction 1.1 Overview Micro-Electro-Mechanical Systems (M EMS) technology is a rapidly growing field and finds various applications that include RF/Microwave circuit design, including the development of switches and tuning elements such as capacitors a nd inductors. MEMS are integrated systems combining bot h electrical and mech anical components. They are traditionally in the microscale, but also extend to a few millimeters. MEMS devices can be used as miniature sensors, c ontrollers or actuators and have commercial uses which include pressure sensors, flow se nsors, accelerometers, and optical scanners [1-3]. Switches are fundamental compone nts that are used in Radio Frequency Integrated Circuits (RFICs) and other wire less front-end circuitr y. Switches are of different kinds and can be broa dly classified as active and pa ssive. Field effect transistors (FETs) and PIN diodes are generally used in active switches. RF-MEMS based devices predominantly use passive, metal contact or cap acitive type switches. In addition to being used in passive tuning circuits, inductors play a role in resonators for low phase-noise voltage controlled oscillators, filter components and reactive impedance elements in RF circuits. RF-MEMS devices are fabricated th rough bulk micromachining, surface micromachining, or LIGA techniques [1-4]. Bulk micromachining is the process of fabricating devices by directly etching into a wafer. In this technique both sides of the wafer can be patterned and etched and can be used to assemble three dimensional structures. The method is widely used in forming membranes, holes, beams and grooves [4]. The surface micromachining technique consists of building up micro electro mechanical structures in layers of thin films on the surface of a wafer (or any other suitable substrate). It is different from bul k processes as the devices are fabricated

PAGE 18

2 entirely out of the thin f ilm material. Unlike surface m illing, in surface micromachining layers of materials can be added or subtra cted [1, 4]. LIGA is a German acronym which stands for Lithography, Electropl ating and Molding and is a technique used to produce molds for the fabrication of micromachined components [5]. At present, MEMS components are domi nated by silicon technologies. Silicon is the only material to combine sensors and actua tors with passive and active electronics for signal readout, data conversion, data processing data storage, and so forth. Silicon wafer substrates or films as well as fabrication me thods for devices have reached a nearly ideal status in terms of quality and yield. Howeve r, for applications under extreme conditions silicon may not be suitable any more. Such harsh environments ar e for instance high temperatures, aggressive media, or high en ergy particle radiat ion. For these purposes semiconductors with a wide bandgap and ceramics are preferable. An excellent candidate in this respect is diamond. It is extremely hard and stiff, mechanically and temperature stable, chemically inert and corrosion resistant, piezoresistive, and has the highest thermal conductivity of all natural solids at room temperature. Its resistivity range can be varied by doping over about 1015 orders of magnitude from highly insulating to quasimetallic. Besides, its mechanic ally-relevant material properties such as Young’s modulus and hardness remain practically the same within a wide range of temperature. Also quasimetallic electrical conductivity is nearly temperature independent within typical operation regimes of MEMS. With particular reference to RF-MEMS switches, the superior mechanical and electrical properties of diamond can eliminat e, or at least mitigate, common failure mechanisms. For Ohmic switches these mechan isms include wear at contact surfaces, surface hardening and cold welding which can arise from electrical arching and resistive heating. The high hardness and thermal conduc tivity of nanocrystalline diamond (NCD) enables significant improvements in each of these areas. However, for Ohmic switches, if diamond is to be used as contact, it needs to have high electrical conductivity, which currently is only provided by the unique nitr ogen-doped ultrananocrystalline diamond (NUNCD) developed and patented at Argonne National Laboratories [6]. For the capacitive switches, an additional failure mechanism is charge trapping at the surface of the

PAGE 19

3 insulating material used on the bottom electrode (typically silicon nitride or SiO2); depending on the nature of th e trapping the switch may spontan eously release or remain closed when the bias voltage is removed. The dielectric propertie s of diamond films can be varied by controlling the amount of incorporated hydrogen in the plasma and thereby changing the carbon bonding c onfiguration at the grain boundary, providing a novel approach to producing a leaky dielectric, which is one approach to minimizing or eliminating the charging problem in RF-MEMS switches. This research is mainly focused on combining NCD films and RF-MEMS technology to generate tunable NCD DC cont act switches, inductors and capacitive shunt switches using diamond as a dielectric. The DC contact switches are fabricated on a low resistivity silicon substrate and a comb ination of surface and bulk micromachining techniques are used to realize tunable devi ces in CPW and microstri p configurations. The tunable switches and inductors are integrated using solid-liquid interdiffusion, which is a popular technique in assembling interconnects in the IC industry. Figure 1.1 shows the cross section of an integrated NCD actuator to a host substrate which carries the microwave transmission lines. Figure 1.1 Cross Section of the NCD Actuat or Integrated to a Host Substrate In order to fabricate these actuators for reliability, the mechanical characteristics of NCD thin films should be well understood. The first part of this research is focused on the growth process, material and mechan ical characterization of doped and undoped NCD ilicon Frame Doped NCD Actuator Microwave Transmission Line Host Substrate Silicon Frame Doped NCD Actuator Microwave Transmission Line

PAGE 20

4 films. The diamond actuators are designed to ope rate in a bi-stable c onfiguration. In order to understand the operation, mechanical simulations using ANSYSTM are carried out. The microwave performance of the switches and inductors is studied by measuring the Sparameters of the devices in the frequency range of 140 GHz. Fi nally, these actuators are tested for microwave performance and mechanical stability by performing large signal measurements at 1.9 GHz and 2.1 GHz with varying power levels. The next portion of this research is focused on developing capacitive shunt switches using NCD as a dielec tric. Figure 1.2 shows the top vi ew of the device. Prior to making these devices, the growth of NCD on metallic films is studied. In order to obtain reliable films, diamond re quires carbide forming materi als. For this purpose both tungsten and molybdenum deposited by s puttering techniques are used. NCD Dielectric Bottom Electrode Shunt Beam CPW TL Figure 1.2 Design of the Capacitive S hunt Switch with NCD as a Dielectric Electromagnetic simulations using Agile nts Advanced Design System (ADS) software and small signal measurements are carried out in the frequency range of 1-65 GHz to understand the microwave charact eristics of the swit ch. The charging characteristics of thin film NCD is studied using Corona Kelvin metrology (CKM), C-V and I-V measurements to demonstrate the l eaky and non-charging nature of the diamond films. These measurements are carried out for NCD based metal-insulator-metal (MIM) capacitors and MEMS switches.

PAGE 21

5 1.2 Dissertation Organization Chapter 2 presents the background of RF -MEMS technology. Different types of capacitive switches along with S-parameter simulation results and lumped element models will be presented. A comprehensive st udy of the limitations of capacitive and DC contact switches will be disc ussed. Furthermore, the vari ous microactuation techniques are discussed along with the parameters which have an influence on the actuation voltage. The mechanical aspects, forces which play a role in the bending of the beams, and their uses in the area of MEMS are discussed. In Chapter 3, the growth along with the seeding techniques to develop intrinsic and doped NCD films will be presented. The two popular schools of thought involved in diamond growth, hydrogen chemistry and argon ch emistry, are discussed in detail. Apart from micromachining techniques, measuremen t techniques to understand the mechanical properties that include Young’s modulus, intr insic stress and mechanical resonant frequency of NCD cantilevers ar e reviewed. Finally, the effect of process of parameters on the intrinsic stress during growth and it s distribution on a 4-inch wafer will be highlighted through measurements. The design, fabrication and measurement results of a thermally actuated NCD actuator are presented in Chapter 4. Mechan ical simulations are performed in ANSYSTM to achieve bi-stable mode of operation of the actuator and small signal simulations of the switch and tunable inductor are performed in Agilent’s ADS software. In addition to the integration techniques, high pow er measurements of the NCD actuator integrated in a microstrip topology will be presented. Chapter 5 deals with the design, fabr ication and measurement of NCD based capacitive shunt switches. Materi al characterization highligh ting the quality of diamond growth on metallic thin films will be presented. In addition to the simulation and measurement of S-Parameter results, charging properties of NCD as a dielectric will be discussed using Corona Kelvin metrology (C KM), I-V and C-V measurements. C-V and I-V measurements are performed on both MIM and MEM structures.

PAGE 22

6 Chapter 6 concludes the dissertation with a discussion of improvements and future directions for NCD based RF-MEMS devices. For completeness, appendices with fabrication details are presented. 1.3 Contributions The following contributions have been made to the RF-MEMS and diamond community through this dissertation research. A thorough analysis of the growth pro cess involved in developing intrinsic and doped diamond films. The two schools of thought, hydrogen chemistry and argon chemistry, and the various s eeding methodologies and their effect on the quality of the films are discussed. Although system dependent, the various process parameters which are i nvolved in controlling the stress in the NCD films have been studied. A thermally actuated NCD based actuat or is integrated using a SOLIDTM process to realize RF-MEMS based switc hes and tunable in ductors in a CPW topology. The switches and inductors ar e tested up to 40 GHz. High power measurements are performed on the micr ostrip switch in the power range of 24-47 with an insertion loss of 0.2-0.3 dB in the entire frequency range. To the best of the author’s knowledge this is the first demonstration a fully integrated diamond switch for microwave applications. Intrinsic NCD films are us ed as the dielectric layer in capacitive shunt switches. The switch is electro-sta tically actuated and small signal measurements are presented in th e frequency range of 1-65 GHz. The measured insertion loss in the up-state is ~ 1.1 dB at ~50 GHz with 30 dB isolation in the off-state. Dielectric characterization was performed using the Corona-Kelvin technique and standard I-V testing on comparison nitride and diamond test fixtures. The leaky nature of the diamond films provides a potential solution to reliability issues re lated to dielectric charging. To the best of the author’s knowledge this is the first RF-MEMS capacitive switch using NCD as a dielectric.

PAGE 23

7 Chapter 2 RF-MEMS – An Overview of the Technology and its Reliability Issues 2.1 Introduction Micro-Electro-Mechanical Systems (M EMS) technology is a rapidly growing field and finds various applications that in clude RF/microwave circuit design. MEMS are integrated systems combining both electr ical and mechanical components. MEMS structures are batch fabricated using standard integrated circuit processing techniques and can range in size from micrometers to millimeters. These devices have been in development since the early 90’s and have sinc e matured and attract significant attention in defense and commercial applications. Th eir commercial applications are in the wireless, automotive and biomedical indus tries and they have been used in accelerometers, pressure sensors and flow sensors. Micromachining provides a new dimension to the fabrication of high performan ce and low cost circuits when compared to the circuits fabricated using conventiona l MMIC processing. The field of MEMS and micromachining has been applied to RF (le ss than 2GHz) to millimeter wave frequency (3 to 300GHz) circuits to create high-perfo rmance passive components such as switches, phase shifters, high-Q varactors, tunable filters, matching netw orks and oscillators [7]. In this chapter, RF-MEMS switches w ith various topologies and actuation schemes will be presented. An overview of capacitive switches with small signal simulation results performed in Agilent’s Advanced Design System (ADS) will be presented. Finally the shortcomings of RF-MEMS that include power handling limitations for Ohmic contact switches and reliab ility issues due to charging in capacitive switches will be discussed in detail.

PAGE 24

8 2.2 Overview of RF-MEMS Switches Most of the RF-MEMS devices designed and fabricated to date use silicon, glass, or quartz substrates [8-10] and are monolithically integrated. These devices have been fabricated utilizing electrostatic [8], thermal [9] and piezoelectric [10] actuation schemes. RF switches are one of the most research ed and fabricated devices in the MEMS technology. These devices outperform their active counterparts that in clude field effect transistors (FETs) and pin-diodes in terms of loss at high frequencie s, linearity and power consumption. Although established since its introduction in the early 90’s, RF-MEMS devices still has reliability issues, slow switching speed, high packaging costs and high power handling limitations in comparison to the active devices. Recent developments have at least reduced the above menti oned limitations. Table 2.1 [7] presents a comparison of MEMS switches with PIN diodes and FETs. Table 2.1 – Performance Comparison Be tween FETs, PIN Diodes and RF-MEMS Switches Parameters RF-MEMS PIN FET Actuation Voltage (V) 20 80 +/-3 5 3 – 5 Power Consumption (mW) 0.05 0.1 5 100 0.05 0.1 Switching Time 1 300 s 5 100 ns 1 100 ns Isolation (1-10 GHZ) Very High High Medium Isolation (10-40 GHZ) Very High Medium Low Loss (1-40 GHZ) (dB) 0.05 0.2 0.3 – 1.2 0.4 – 2.5 Power Handling (W) <1 <10 <10 RF-MEMS switches can be fabricated us ing a floating cantilever (dive board design) or fixed-fixed membrane as show n Figure 2.1 and Figure 2.2, respectively. These moveable membranes are modeled as mechan ical springs with an equivalent spring constant, k [N/m]. The spring constant depe nds on the geometrical dimensions of the membrane or cantilever and on the Young’s modulus of the material used (Au, Al,

PAGE 25

9 nitride, etc.) and is 5-40 N/m for most RF -MEMS switch designs [7]. The switches have very low mass, around 10-10 to 10-11 Kg and, therefore, are not sensitive to acceleration forces. These switches can be fabricated in the microstrip and CPW topologies. In a CPW configuration, the anchors of the switch are directly connected to ground plane or signal line depending on the switch type. Via-holes or a quarter wavelength open stub are used to connect the anchors to the ground pl ane in the microstrip topology. RF-MEMS switches can be fabricated as DC contact or capacitively coupled switches. The following section presents the design and simulation results of series a nd shunt capacitive switches. Figure 2.1 Broadside MEMS Series Sw itch Utilizing Cantilever Beam s [11] Figure 2.2 Raytheon MEMS Capacitive Shunt Switch Utilizing Fixed-Fixed Beam [11]

PAGE 26

10 2.2.1 Capacitive Switches Capacitive RF-MEMS switches consist of a movable membrane that is used to realize a variable capacitanc e, and thereby a variation in impedance between the nonactuated (“OFF”) and actuated (“ON”) states. A DC voltage is applied between the bridge and the transmission line, causing the bridge to collapse on a dielectric layer. The increased capacitance value that arises when the bridge or beam is actuated connects the transmission line to the ground and acts as a short at microwave frequencies. The dielectric layer in capacitive switches is ge nerally formed using a PECVD growth process to achieve conformal deposition. The dielec tric thickness depends on the required capacitance ratio as per the a pplication. The popular dielectr ic materials used in RFMEMS switches are silicon nitride and sili con oxide. The bridge height is typically between 2-3 m, length of the moveable membrane is 250-400 m, and the width is between 30-150 m. As stated earlier, cap acitive switches can be designed in series and shunt configurations; Figure 2.3 shows a RF-MEMS capacitive switch in a series configuration. Figure 2.3 – Capacitive RF-MEMS Sw itch in a Series Configuration Relatively simple circuit schematics can be used to emulate the electrical response of a capacitive switch. The capacitive switch show n in Figure 2.3 can be represented as a series capacitor-inductor-resistor (CLR) ne twork. Figure 2.4 shows the circuit schematic of a series capacitive switch; in this circu it the series capacitance varies between 30fF in the non-actuated state to 3 pF in the actuated state. The in ductance contributed by the

PAGE 27

11 bridge is assumed to be 10 pH, and the cont act resistance contributed by the bridge is accounted for with the 0.5 resistor. Figure 2.4 – Equivalent Circuit of a Series Capacitive Switch One of the major disadvantages of a series capacitive switch is the low return loss associated with the non-actuated state. When the beam is the up-state, which is the RFOFF (non-actuated) state, the switch capacitance is small and hence most of the signal is reflected back. When actuated, i.e. when th e beam makes contact with the dielectric layer, the switch capacitance becomes high and th e switch is the ON state. In this state, the switch exhibits a high–pass characteristic with the cu t-off frequency. The cut-off frequency, which is controlled by the bridge inductance and capacitance values, defines the working range of the device. For the circuit in Figure 2.4, the device can be considered as a switch for a frequency less than 5 GHz wherein the isolation is greater than 20 dB in the OFF state and the insertion lo ss is less than 1 dB in the ON state. Figure 2.5 shows the S-parameters of the series swit ch in the OFF and ON states. For an ideal series switch designed to provide an isolati on and insertion loss better than 40-50 dB and 0.2dB in the K band frequency range, respective ly, the capacitance ra tio should be on the order of ~ 3000. This ratio is difficult to achieve in practice, often pr ompting the use of a DC contact switch in combination with a capacitive shunt switch [12].

PAGE 28

12 0 5 10 15 20 25Frequency [GHz] -1.2 -1 -0.8 -0.6 -0.4 -0.2 0S 1 1 U p & S 2 1 D o w n [ d B ] -50 -45 -40 -35 -30 -25 -20 -15 -10 -5S 2 1 U p & S 1 1 D o w n [ d B ] S11 (Up) S21 (Down) S21 (Up) S11 (Down) Figure 2.5 – S-Parameters of the Capacitive Series Switch in the OFF and ON States Figure 2.6 and Figure 2.7 show the design and equivalent circ uit of a capacitive shunt switch, respectively. Similar to the se ries switch, the capacitance varies between 30 fF in the OFF state to 3 pF in the ON state. The inductan ce and resistance values are assumed to be similar to th at of the series switch. Dielectric Moveable Membrane Figure 2.6 – Capacitive RF-MEMS Switch in Shunt Configuration

PAGE 29

13 Figure 2.7 – Equivalent Circuit of a Capacitive Shunt Switch In the shunt switch, when the beam is in the non-actuated state, the return loss is dominated by the 30 fF capac itor. Upon actuation, the si gnal goes through the high capacitance path to ground, thereby providi ng good isolation between the two ports. Figure 2.8 shows the simulated S-parameters of the capacitive shunt switch in the ON and OFF states. At high frequencies the combin ed effect of the inductance of the bridge and capacitance causes a resonance (minimum impedance to ground) in the isolation response. 0 5 10 15 20 25Frequency [GHz] -8 -6 -4 -2 0S 1 1 D o w n & S 2 1 U p [ d B ] -50 -40 -30 -20 -10 0S 1 1 U p & S 2 1 D o w n [ d B ] S11 (Down) S21 (Up) S21 (Down) S11 (Up) Figure 2.8 – S-Parameters of the Capacitive Shunt Switch in the ON and OFF States

PAGE 30

14 The capacitive shunt switch developed by C. Go ldsmith et al. [13] and the DC contact series switch developed by Linc oln laboratories [14] have been used as standards in their respective configurations. 2.3 Actuation Schemes in RF-MEMS Actuation refers to the act of affecti ng or transmitting mechanical motion, forces and work by a device or system on its surroun dings in response to the application of a bias voltage or current [15]. Considerable research has been directed at designing and fabricating RF-MEMS devices with a variety of different actuation schemes. The most popular actuation schemes used in MEMS processes are electr ostatic actuation, piezoelectric actuation, magnetic act uation, and thermal actuation Electrostatic actuation [1, 2], which is the simplest and the easiest to control, is the preferred actuation scheme in many MEMS applications. In this case, the positive and negative charges set by applied voltages on stru ctures elicit Coulom b forces which cause motion. This electrostatic force between a to p and bottom electrode when a DC potential is applied is given by 2 22 2 r dt g AV QE F (2.1) where V, g, and td are the voltage, electrode gap and th ickness of the dielectric layer, respectively. The threshold voltage which is the voltage at which the moveable membrane falls on to the bottom electrode, can be approximated to the value shown in equation 2.2 wherein Vth is the threshold voltage, E is the Young’s modulus, I is the geometrical moment of inertia, td is the inter-electrode distance, and l and w are the length and width of the structure. w EIt Vd th 4 0 35 18 (2.2) Electrostatic actuation is useful in applica tions where integration on a chip is easy from a fabrication point of view. Most of the MEMS beams which are easily fabricated require more than 15 volts for actuation, which is typically not compatible with low

PAGE 31

15 voltage CMOS applications. Also, there is no proper scheme for producing repulsive forces using this method of actuation. In addition to RF-MEMS switches, electrostatic actuation is also used in actuators for resonators and light modulators Piezoelectricity is a phenomenon in which a mechanical stress on a material produces an electrical polarization and, r eciprocally, an applied field produces a mechanical stress. In piezoelectric actuation [7] the applied voltage induces fields which change the dimension of struct ure and this dimensional change is used to cause motion. Here the electrically induced strain is a pproximately proportional to the applied electric field. The stress associated with this actuati on scheme is very high, which constitutes a high energy density. Piezoelectric actuation scheme has its share of advantages and disadvantages in comparison to electrostatic actuation. Although the actuation voltage is very less (w.r.t electrostatic ac tuation), integrating piezo materials with host substrate and the fabrication methodologies are complicated. Potential limitations of piezoelectric actuation include hysteresis beha vior, change in response ov er time [16]. Piezoelectric materials (ZnO, AlN) are characterized by the charge sensitivity coefficients, dij, which relate the amount of charge ge nerated at the surfaces of the material on the i axis to the applied force, F, on the j axis A d F d Qij j ij i (2.3) The voltage across the electrodes can be given as, A Qx C Q Vr 0 ; A x F d Vr j ij i 0 (2.4) The popular materials which exhibit piezoel ectric properties are barium titanate (BaTiO3), lead zirconium titanate (PZT), zinc oxide (ZnO) and aluminum nitride (AlN). These materials have been used in many RF-M EMS devices such as phase shifters, filters etc. [7] Magnetic actuation is not common in ME MS process because, apart from being incompatible with CMOS processes, the fabrication process is very tedious. In this actuation scheme, the magnet-induced or current-induced magnetic force produces

PAGE 32

16 motion. This force is caused when charge carriers travel through a perpendicular magnetic field thereby causing a deflection; this is popularly known as the Lorentz force [17]. The charges produce a voltage called the Ha ll voltage which is used in the actuation scheme. The hall voltage as shown in equatio n 2.5 is dependent on the hall coefficient (RH), current density (J), width of the st ructure (W), and magnetic flux density (BZ). Z X H HWB J R V (2.5) Magnetic actuation [2, 7] is difficult because ferromagnetic materials are required for focusing the magnetic flux needed for actuati on. On the positive si de, magnetic actuation in MEMS provides sufficient force and requi res very small voltages for actuation. Magnetic actuators find their a pplication in micro generators and in low-voltage/large – force /low –efficiency actuators. In the case of thermal actuation, the current or voltage applied causes an element to heat up and expand. This expansion re sults in a dimensional change used to communicate motion. The coefficient of thermal expansion, L, quantifies the relative dimensional change of an object that occurs for a change in temperature. The movement in the actuator is achieved when two dissimilar materials with different L are sandwiched together. This pr inciple is used in thermostat switches which use the bimetallic or thermal bimorph mechanism. Similar to magnetic actuation, thermal actuation requires very low volta ge for actuation. Power consumption, which is a major limitation of this scheme, can be avoided by us ing bi-stable structures [18] or dual mode actuation schemes [19]. Thermal actuation is used in gas pressure sensors, thermal flow sensors and in humidity sensors [2]. Apart from the aforementioned there ar e other actuation schemes that include electrostrictive actuation, magnetostrictive actu ation, ultrasonic actuation and chemical actuation. These methods are not widely pop ular and are used for specific purposes depending on the nature of the application.

PAGE 33

172.4 Failure Mechanisms of RF-MEMS Switches When first introduced, RF-MEMS devices had limitations with reliability, power handling, switching speed, high-voltage drive and packaging. Recent developments by researchers at MIT Lincoln Lab and the Univer sity of California at San Diego [14, 20] have successfully demonstrated long term re liability up to a few hundred billion cycles. These reliability experiments have been ca rried out for both DC contact and capacitive switches under hot and cold switching c onditions. The switching speed of RF-MEMS devices was considered slow in comparison to active competitors, however special fabrication techniques like the focused i on-beam (FIB) milled na no-switch [21] and “Mini-MEMS” [22] switches have show n switching speed in the hundreds of nanoseconds. Further improvements in these designs are in progr ess to improve the switch speed to less than 200 ns. Performanc e and reliability of the switches due to packaging have been partially addressed in terms wafer level and chip level packaging. Glass frit wafer bonding [23], metallic cap techni que [24] by Teravicta, low temperature thermosonic flip chip and low temperature glas s seal ring [25] by XC OM wireless are the popular packaging schemes available. Long term reliability and the high cost associated with these techniques are still being addressed. Apart from the above stated problems, ther e are two main issues when it comes to the reliability of RF-MEMS switches. In capacitive switches, reliability is limited by dielectric charging of the insulator layer. In DC contact switches, the reliability is affected by the metal contact used and by pow er handling capabilities. In the following section, an overview of charging and the me thods to improve the reliability of capacitive switches and power handling capab ilities for DC contact switch es will be presented. 2.4.1 Dielectric Charging Reliability of capacitive switches (low power) is reduced mainly due to the moveable membrane not releasing or comi ng back to the normal OFF state after the actuation voltage is removed. This failure is mainly due to stiction between the metal layer and the dielectric and is caused by charge accumulation due to injection and trapping inside the dielectric layer. Charge trapping has been studied extensively in the

PAGE 34

18 CMOS industry for years for transistor ap plications. Many groups [26-29] in the RFMEMS community have addressed this charging issue in multiple ways which deal with surface charging and bulk charging effects. Dielectric charging can be associated with stress which can be mechanical, electrica l or thermal. Typically during charging, electrons are trapped at low el ectric fields and get released or de-trapped at high fields, but holes are typically observed only at hi gh fields (~ 10 MV/cm). Apart from charge injection and trapping, the surface and interf ace of the dielectric layer and electrodes where defects are concentr ated, will be areas for charge accumulation. Figure 2.9 shows the schematic representa tion of an electro statically actuated capacitive switch. At a certain vo ltage, the electrostatic force (FE) becomes greater than the restoring spring force (Fspring) and causes the top plate to collapse. This voltage is called the pull-in-voltage (Vpi). (The pull-in voltage mentioned here is same as the threshold voltage discussed earlier.) Once actua ted, the beam continues to stay in the bottom state until the app lied voltage is lesser than the hold down voltage (Vpo). When actatuated repeatedly there is net amount of charge injected into the dielectric and this effectively changes or shifts the pull-in and pull-out voltages by a margin. This shift in voltage, Vshift changes continuously, leading to the eventual failure of the switch. Figure 2.9 – Schematic Representation of th e Actuation Mechanism in Capacitive RFMEMS Switch When the beam is actuated, the electric fiel d in the dielectric layer is typically on the order of few Megavolts/cm. Because of this high electric field, charge is continuously

PAGE 35

19 introduced into the dielectric layer thereby changing the electric field in the gap between the two electrodes and thus affecting the elect rostatic force. A trapped charge in the dielectric layer results in image charges in the top and bottom electrodes i.e. if the trapped charges are +ve in the dielectric, th at results in accumulation of ve charges in the electrodes as shown in Figure 2.10. This reduces the total amount of charge on the electrode if a voltage is applied. This vo ltage shift will eventually change the net capacitance contributed by the sw itch. As reported by Reid, a uniform trap density of 1012/cm2 is more than enough to cause a capacitive sw itch to fail in the actuated state. [7] Figure 2.10 Accumulation of Image Charge s on the Electrodes During Actuation Researchers from NXP semiconductors [30] have suggested two different techniques to study charge injection as a function of voltage and time. In the first technique, called the whole CV method, capacita nce is measured in terms of voltage until the pull-in voltage. The shift in voltage is measured from successive CV curves measured during the actuation cycle. This technique is probably the oldest and the most well known method to evaluate the charging mechanism. In the second technique, the center shift method, only the shift in the center part of the CV curve is measured. Through this shift the voltage at the lowest capacitance value is calculated by fitting a parabola through the center of the curve. This technique is faster than the simple CV and successive approximation methods, and therefore has less influence on the device under test. There are modifications introduced into this tech nique wherein RF measurements and voltage shift is determined by manually tuning the bias voltage for measuring the capacitance. Charging has also been studied by understanding the effects of dipolar and intrinsic space charge and in terfacial polarization on the RF-MEMS switches. In this technique, charging is studied in contactless mode wherein the membrane is actuated at low voltages (6-8 volts). Polarization which is caused when an electric field is applied to

PAGE 36

20 a dielectric layer is measured and the change in intrinsic polarization effects is monitored. This change in the polarization effects is dire ctly related to the ch ange in capacitance of the switch. Temperature dependent CV measur ements were also performed to understand the amphoteric nature of the traps and its eff ects. Furthermore, a co rrelation between the method of deposition of the dielectric and the thickness of the di electric in charging mechanism was deduced [31]. Apart from measuring MEM structures, re searchers have analyzed the charging mechanism with switches intentionally fa bricated in the dow n-state and on MIM structures. I-V and C-V measurements have b een done to see the shif t in leakage current and capacitance, respectively. In these measurements, the cu rrent measured from the I-V technique is comprised of the displacement current, trap charging cu rrent and steady state leakage current. The change in the leakage cu rrent and capacitance valu es have also been performed under stressed voltage and current conditions. The increasi ng voltage used in the stress testing causes defects in the dielectric layer, which lead to increased current and capacitance values [32]. The use of different dielectric materials to mitigate these charging issues has also been suggested by many research groups. PECVD silicon-dioxide has a lower trap density than silicon nitride films. Del Rio et al. [33] have demonstrated capacitive switches using alumina and zinc oxide alloys as the dielectric layer. The dielectric layer is deposited using atomic layer deposition (A LD) technique. The conformal growth of the dielectric layer makes it very reliable. Preliminary tests have shown that these films are capable of dissipating trapped charges and maximizing the ON state capacitance. Research groups from Sandia National Labs [34] and Purdue University [35] have suggested the use of amorphous and na nocrystalline diamond (NCD). Amorphous diamond films have been used in RF-MEM S switches and their non-charging behavior has also been reported. Nanocrystalline diam ond is a very new film and little to no research has been done in using it as a di electric film in capacitive switches. The complexity involved in the growth proce ss and micromachining has made NCD related MEMS a rarity. But considerable improveme nts in micromachining techniques and the superior electrical and microw ave properties is making NCD films attract lot of attention

PAGE 37

21 in the RF-MEMS area. These charge leaky films provide a conductive path for the charges during actuation, there by avoiding charge trapping a nd storage. In addition to these new films, carbon nanotube based RF MEMS switches [36] have been demonstrated. In these switches, double walled carbon nanotubes have been incorporated into the switch and the increased reliability due to the high density of the dielectric was demonstrated. Another standard method to reduce char ging is by using bipol ar voltage when actuating the beams. By using this scheme the electrostatic force which is proportional to the applied voltage is maintained constant through the actuation process. Although this technique does not entirely remove charge in jection into the dielectric, extended test results have shown switches having highe r reliability through this technique. Charging in RF-MEMS switches has been addressed by using a Schottky diode with a membrane and semiconductor in place of th e usual dielectric laye r. Pillans et al. [37] have demonstrated the use of Schott ky barrier contact based RF-MEMS switch. In this switch, n++ InGaAs is used as the bottom electrode and epitaxial InAlAs is used as the dielectric layer. The entire switch was fabricated on an InP substrate, which facilitates direct integration with solid state devi ces. The switch operates as a normal RF-MEMS switch under reverse bias conditions and once the charges accumulate the switch can be forward biased to effectively recombine any trapped charges. Figur e 2.11 shows the top view of the fabricated Schottky switch. Figure 2.11 – Top View of the Fabricated Schottky Contact RF-MEMS Switch [37]

PAGE 38

22 In addition to the above stated solutions charging problems can also be reduced by using low voltage actuation schemes (thermal, magnetic and piezoelectric). Fabrication complexities and integration issues with these low voltage actuation techniques are being addressed to make them viable in future applications. Research groups have suggested alternative solutions [38-40] which deal with charging due to Frenkel-Poole conduction, mechanical defo rmation and choice of substrate. These methods as claimed by the authors tend to offer better results for chargi ng related studies. 2.4.2 Contact Material Issues In DC contact switches the most prev alent failure mechanism is the damage created in the contact area of the beam or th e electrode as repeated actuation causes two metal surfaces to come in contact under lo w, medium or high power conditions. Damage can include pitting and hardening [7] of the metal at these contact areas. Over time the repeated contact also reduces the contact area thereby increasing the contact resistance of the switch which is typically between 1-2 under normal conditions. Over time, this value starts to increase and can become as large as 7-8 As demonstrated by researchers at Rockwell Scienc e Center, reliability can be increased by controlling the actuation voltage in electrostatically actuate d switches. By controlling the actuation voltage, the impact upon actuation can be reduced thereby reducing the pitting effects. Figure 2.12 – Cross Section of DC Cont act Switch in OFF and ON State [41] Contact resistance can be controlled by th e choice of material used in the fixed beams or cantilever structures. The contact material to be used is dependent on the force

PAGE 39

23 required during the actuation scheme. Gold is the most common material used in RFMEMS structures, particularly for low force designs, and the hardne ss of sputtered and electroplated gold is 3 GPa and 1 GPa, re spectively [7,44]. For high contact force designs, apart from gold the ot her popular metals are rheniu m and gold-palladium alloys. There is considerable inte rest among many to use doped NCD or diamond like carbon (DLC) in metal contacts. Although these ma terials are good choices because of their hardness, deposition of these materials on metals in low temperature conditions is still a challenge. Other failure mechanisms in DC contact switches are related to humidity, contamination in contact areas and organic de posits after the fabrication process. Figure 2.13 (a and b) shows the SEM image of a fabricat ed tunable inductor. It is seen that there are organic deposits around the contact area of the beam; this contamination will result in the failure of the cantilever structure duri ng actuation. Contamination related problems can be reduced by fabricating and packagi ng the switches in a clean environment. (a) (b) Figure 2.13 – SEM Image of DC Cont act Switch with Organic Deposits 2.5 Power Handling Capabilities Over the years the high power capab ility of RF-MEMS switches (DC or capacitive; shunt or series) has been lim ited by failures due to self actuation, electromigration, latching and Joule hea ting issues [42-44]. Although considerable research has gone into addressing this probl em, the long term reliabi lity of the switches under high power conditions is still an active area of study. At lower power levels Organic Deposits Deposits Near Contact Area

PAGE 40

24 (milliwatts) under nitrogen ambience, dry ai r and vacuum conditions little to no difference in the number of cycles under hot or cold switching condition is typically observed [7]. Table 2.2 shows the list of the current state of art in high power testing for capacitive and DC contact switches [45-48]. Table 2.2 – State of Art in High Power RF-MEMS Switches Research Group Switch Type Testing Conditions Lincoln Labs Capacitive 10 Watts cold switching & 1.7 Watts hot switching Raytheon Capacitive 4 Watts cold switching & 510 mW hot switching Radant MEMS DC Contact 10 Watts hot switching The power handling capability of an RF -MEMS switch depends on the type of switch (DC contact or capacitive) and the conf iguration (series or shunt) in which it is fabricated. In most the cases electrostatic actuation is the prefe rred scheme for high power RF-MEMS switches. The DC actuation voltage (Vact) of a bridge can be given by [42] A kg Vo act27 83 (2.6) Where k is the spring constant, g is the initial height of the bridge and A is the area of cross section. An RF signa l with a magnitude of VO generates an equivalent DC voltage of 2O eqV V (2.7) And the equivalent input power can be evalua ted in terms of the input voltage and the input impedance (ZO) O O inZ V P 22 (2.8)

PAGE 41

25 When the switch is the up-state, almo st zero power gets reflected and the minimum power [42] to actuate or rather “s elf actuate” the switch can be given as shown in equation 2.9. This input power causes a force on the metallic plates and if this power is sufficiently high, the induced force will be large enough to self actuate the moveable membrane. O O actAZ kg P27 83 (min) (2.9) Equations 2.6 – 2.9 apply for a shunt switch. For a series sw itch, the equivalent voltage (Veq) is 2VO. Assuming the same area and spring cons tant, series capacitive switches are capable of handling 1/4th of the power as the shunt switches. Recent improvements in designs have shown improvements for designs fabr icated in the series configuration, too. Another high power handling issue deal s with the high current which is propagated in the moveable membrane a nd transmission under high power conditions. Electromigration [43] is defined as the move ment of metal ions in a conductor upon an input electric current. The tr ansmission lines can exhibit this effect, thereby lowering the conductivity of the material a nd increasing the overall loss (degrading performance) of the switch. Furthermore, high power levels in capacitive switches also lead to charging related issues which can cause the switch to fail. In DC contact switches, the contact area typically dissip ates 0.5 % of the incident power [1]. This dissipation can result in loca lized heating near or at the contact area. Localized heating can subsequently lead to in creased contact resistance in the actuated state. The increase in contact resistance is due to the bilateral heat current [44] due to the high power. Therefore thermal conduction plays an important role in the stable operation and reliability of the switch. Thermal cons triction conductance (W/K) [44] as shown in equation 2.10 is the ability of the contact to dissipate heat through the contact and it is dependent on the thermal conductivity of the mate rial (k), the radius of the contact spot, and b the radius of the cathode. 11 b a F R Gcd ct (2.10) where Rcd can be defined in terms of the thermal conductivity (k) and a

PAGE 42

26 ka Rcd4 1 (2.11) and 33441 0 4098 1 1 b a b a b a F (2.12) As demonstrated by Hyman et al. [44] sa mples with large c ontact area and good thermal conductivity exhibit li ttle change in contact re sistance up to 50 mA, while switches fabricated with low thermal conduc tivity yield higher contact resistance at elevated current and power levels. Damage and material transfer also increase with contact force and current level which is the effect of heat conduction. 2.6 Summary RF-MEMS switches of different types and co nfigurations have been discussed. Lumped element circuit models and small si gnal simulations using Agilent’s ADS for series and shunt type capacitive switches we re presented. Different actuation schemes that include electrostatic, th ermal, piezoelectric and magnetic along with their advantages and disadvantages have been discussed in de tail. Charging which is the major reason in the failure of capacitive switches and the soluti ons to eliminate or mitigate this effect has been presented in detail. Similarly, reliability issues in DC contact switches that include material contamination and failure due to high power handling have been discussed. Although a multitude of solutions have been an alyzed and presented, choosing the best method is cost and application dependent.

PAGE 43

27 Chapter 3 Nanocrystalline Diamond – Properties, Growth and Characterization 3.1 Introduction Micro-electro-mechanical devices (MEMS) have been fabricated on substrates that include silicon, quartz and glass among others. But these devices are limited in applications which require reliability under high temperat ure and high power conditions. Furthermore, MEMS devices unlike integr ated circuits (IC’s) involve moveable components wherein the mechanical and tri bological properties of silicon limit its application. Because of these limitations othe r materials such as SiC, GaN and diamond are now under investigation. Diamond has been investigated for applic ation to active devices such as field effective transistors (FET) fo r high power electronics as well as for tools used in grinding, polishing, cutting, and dicing. Diam ond has the highest Young’s modulus, hardness and thermal conductivity and it is transparent from the UV to far IR region. Furthermore, its superior elec tronic properties make it suita ble for use in heat sinks, and radiation detectors [49]. Diamond is chemically inert, stable at hi gh temperature (10000C in vacuum) and is suitable for operation in harsh environments (except oxygen ambience) [50]. Because of these char acteristics diamond is a very good candidate for realizing reliable, high power and temperaturestable MEMS and microwave devices. Table 3.1 [51, 52] compares the mechanical properties of NCD films with other materials used in microsystems technology. Along with positive mechanical attributes NCD possesses low loss when used as a thin film at microwave frequencies.

PAGE 44

28 Table 3.1 – Mechanical Properties of NCD Thin Films in Comparison to Materials Used in Microsystems Technology Si (3c)SiC (6c) SiC (h)GaN Diamond Bandgap (ev) 1.12 2.2 2.9 3.45 5.45 Break down field (106 V/cm) 0.5 4-6 3 3-6 10 Young’s Modulus (Gpa) 170 450 390 1050 Fracture Strength (Gpa) 1.37 ~2.5 10.3 Thermal Conductivity(W/cm.K) 1.47 4.9 4.9 1.3 22 Thermal Stability (0C) 500 900 1300 650 1500 Thin film diamond can be classified into single crystal, microcrystalline (MCD), nanocrystalline (NCD) and ultrananocrystalli ne (UNCD) films. These films are grown on different substrates which is dependent on the respective application. In this chapter, discussions will be focused on the growth and characterization of NCD films. The popular techniques used to grow these thin films along with their chemistry of growth and seeding process will be discussed in deta il. Finally, mechanical characteristics that include Young’s modulus, mechanical resonanc e frequency and intrin sic stress of NCD films will be presented. 3.2 Structure and Growth of Na nocrsytalline Diamond (NCD) Diamond has a face centered cubic crystal (FCC) latti ce structure with a unique arrangement of carbon atoms with eight corner atoms, six face centered atoms and four other atoms from adjacent interpenetrating lattices offset by one-quarter of the body diagonal as shown in the ba ll and stick model of Figure 3.1 [53]. Each of the carbon atoms is covalently bonded to four nearest neighboring atoms by bonds resulting in a strong sp3 character. The (111) plan es of the diamond are along th e bond direction with a lattice constant (a0) of 3.567 and a bond length of 1.54 . Due to this unique chemical

PAGE 45

29 bonding, and atomic density of 1.76 X 1023 cm-3, diamonds possess several extraordinary material properties. Figure 3.1 – Crystal Structur e of Diamond Lattice [53] Chemical vapor deposition (CVD) involves the dissociation of chemical species in a vapor phase to form a coating or a thin film. Thin film diamond is generally grown through a CVD process. The growth process in diamond films can be initiated by adding one carbon atom to its initial template. Subs equent addition of these carbon atoms result in a tetrahedron bonded carbon network. CVD gr owth process of a diamond film can be broken down into the fo llowing steps [54]: A gas phase must be activated, either by a high temperature (ex: hot-filament CVD) or by plasma excita tion (ex: microwave CVD). The gas phase must contain carbon-cont aining species such as hydrocarbon, carbon dioxide or carbon monoxide. A sufficiently high con centration of atomic hydrogen to etch graphite and suppresses gaseous graphite pr ecursors must be provided. The substrate must be seeded to in itiate the nucleation and growth of diamond from the vapor phase. A driving force must exist to transport the carbon-containing species from the gas phase to the surface of the subs trate. In most CVD methods, the temperature gradient acts as a dr iving force for the motion of diamondproducing species via diffusion.

PAGE 46

30 Thin film NCD is grown through a chemical vapor deposition (CVD) process wherein the growth occurs by the deco mposition of carbon containing precursor molecules (typically methane) in eith er a pure hydrogen, or hydrogen and argon environment. NCD growth is done through a thermal (hot filament) [55], plasma (microwave or RF) [56] activation or use of a combustion flame (oxyacetylene). Of the three, hot filament and microwave plasma methodologies are the most popular techniques used for thin film diamond growth. Prior to diamond growth, the wafer needs to go through a seeding step which aids in the growth of the thin film. Seedi ng is popularly done through three different techniques: Mechanical polishing of the wafer: In this technique nanometer sized diamond powder is sprinkled on the silicon wafer and the wafer is mechanically scratched. By this the diamond powder is spread uniformly across the wafer and this acts as a seeding layer in the CVD system. Seeding through this method results in a nucleation density of 107cm-2 [57]. Ultrasonication: Here a silicon wafer is suspended in slurry of nanometer sized powder with acetone or methanol for 20-30 minutes. Through this process the surface of the wafer is damaged and seeded with the diamond powder for the subsequent growth process. Nucleation density of 106 – 1010 cm-2 is achieved through this method [58]. Bias Enhanced Nucleation (BEN): Although the first two processes are popular and result in good diamond films, nucleation density is best in the BEN process [59]. In the microwave plasma enhanced CVD (MPECVD) process, prior to growth, in the BEN stag e the substrate is negatively biased at around 250 volts resul ting in a starting current va lue of 10mA. The current increases and saturates at 100mA (F igure 3.2) in a half hour seeding procedure, beyond which the current tend s to decrease with time. Nucleation density achieved through this procedure is around 1015 cm-2.

PAGE 47

31-5051015202530 20 40 60 80 100 120 140 Bias Current (mA)Bias Time (min) Figure 3.2 – Varying Current Value in Time During the BEN Process In this work, diamond films grown by th e hot filament (HFC VD) method are used in the MEMS actuators and the microwave plasma (MPECVD) films are used in the shunt switches. The growth recipes used at the University of Ulm (hydrogen chemistry) and at the University of S outh Florida (argon chemistry) for intrinsic NCD are given below: Table 3.2 – Growth Recipe for NCD F ilms Using Hot Filament (HFCVD) and Microwave Plasma (MPECVD) Process Gas Flow (sccm) Growth Process Ar H2 CH4 Total Pressure (Torr) Substrate Temp. (0C) Power (kW) Hot Filament CVD (HFCVD) – 200 3 12 – 15 850 2.4 – 2.8 Microwave Plasma CVD (Ulm MPECVD) – 400 – 500 4 – 8 15 800 – 850 2.4 Microwave Plasma CVD (USF MPECVD) 788 8 4 135 725 1.8

PAGE 48

32 Diamond films grown in the HFCVD tec hnique use bias enhanced nucleation (BEN) as the seeding method. The HFCVD tec hnique is based on the heating of metal filaments up to 2000-2200C in order to break molecular hydrogen and carbon compounds and to form CyHx free radicals; these free radicals move toward the substrate by a temperature gradient. On the substrate surface both diamond (sp3) and graphite (sp2) bonds are formed; the graph ite bonds are etched by atomic hydrogen allowing diamond growth on the substrate surface. Figure 3.3 shows the SEM image of a diamond film grown in the HFCVD reactor. Figure 3.3 – SEM Image of a NCD Film Grown by the HFCVD Technique In the MPECVD technique, a C2 dimer-based growth mechanism that would result in nanocrystalline st ructure was proposed [60]. In the films deposited using 5% CH4 and 95% Ar, the C2 dimers resulted in the inclusion of an amorphous carbon or graphitic carbon [61]. Such non-diamond fo rms of carbon are due to the homogenous nucleation resulting from a high ratio of hydrocarbon to carbon dimers. But on the other hand, during the deposition of nano-diamond f ilms, the heterogeneous nucleation rate (>1010 cm2sec-1) increases due to highly reactive C2 species, resulting in the smaller grain size of the diamond films [63]. According to the proposed model, the feed gases methane and argon disassociate and favor the formation of (C2H2)+ at a low ionization potential. The positively charged acetylene radical attrac ts an electron to form a highly reactive carbon dimer and hydrogen. Hydrogen is then desorbed away while the carbon dimers nucleate at the reconstructed surface. As th e reaction continues, the number of carbon

PAGE 49

33 dimers in the plasma increases and they join th e previously hybridi zed carbon atoms. In this way, a closely hybridized sp3 network of carbon atoms forms a continuous film of nanocrystalline diamond. The limited applications of microcryst alline diamond (MCD) films have been surpassed by synthesizing a new class of ma terial known as “nanocrystalline diamond” (NCD) films. The nanocrystalline diamond films can be grown by altering the CVD process [64, 65]. Unlike MCD, NCD films consist of small grains on the order of 20-50 nm and a low surface roughness of ~20 nm. Recently, “ultra-nanocrystalline diamond” (UNCD) films having smaller grain size (3-5 nm) than NCD have been developed [66]. The growth of NCD/UNCD f ilms opened wide windows of a pplications ranging from tribology, MEMS, optics, RF applications and field emission devices [67-69]. Typically, MCD films are deposited in CH4 (1%)/H2 (99%), NCD films are deposited in CH4 (1%)/Ar (98%)/ H2 (1%) and UNCD films are deposited in CH4 (1%)/Ar (99%) gas chemistries. Microcrystalline diamond films c onsist of large grains (grain size: ~ 5-10 m) and rough surfaces (mean surface roughne ss: ~ 300-700 nm) thereby limiting their application to cutting tools, abrasive coatings and heat sinks [14-16]. The intrinsic diamond films are electrically insulating with resi stivity on the order of 10131016 ohm-cm. Electrical conductivity can be achieved by doping the films during the deposition. The diffusion of dopants into the diamond films is not a practical method of doping as the surface is not diffusive to most of the impurities. Though there are a few reports on the ion-implantation of diamond film s [70], it is an expensive technique and can damage the surface. Therefore, dopants such as boron (p-type), nitrogen, phosphorous and sulphur (n-type) are incorporated in the gas chemistry during the growth [71]. The most widely used dopa nts are boron (ptype ) and nitrogen (n-type), as these are readily soluble with diamond. It was observed th at the quality of the films improve with the incorporation of trace am ounts of boron by reducing the point defects. But excess concentration of boron promotes graphitiz ation due to the incorporation of boron interstitial sites. In the case of single crystal or micr ocrystalline diamond, p-type conductivity can be easily achieved. But, it is difficult to obtain n-type conductivity at room temperature in these films as nitrogen forms a d eep donor (~1.7 eV). Nitrogen

PAGE 50

34 forms a shallow donor level (~ 0.4 eV) [72] in NCD/UNCD films and results in high ntype conductivity (~143 -1 cm-1). The drastic reduction in the grain size of the diamond films from several microns to few nanomet ers by changing the gas chemistry suggests that the growth mechanism of nanocrysta lline diamond films is different from conventional CVD diamond films. Boron is the most common and preferred p-dopant and the doping mechanism are well established both in the HFCVD and MPECVD techniques [73]. Nitrogen doped diamond film s are available using MPECVD technique whereas HFCVD films with nitrogen doping are still in the process of being matured. 3.3 Characterization Techniques for NCD Films Unlike the microcrystalline diamond film s, NCD films deposited in hydrogen poor gas chemistry have a complex grain bounda ry structure with grain size on the order of few nanometers. These differences in the gr ain structure result in different mechanical and electrical propert ies of NCD films. The structur al, mechanical and electrical properties of these films have been st udied by several analytical and metrology techniques. Scanning electron microscopy (SEM) has be en a very useful technique in the characterization of diamond thin films. The microstructure of diamond films changes dramatically with the conti nued addition of Ar to reacting gas mixtures during CVD process. The transition from microto nanocry stalline by systematically adding argon to hydrogen-rich plasma has been characterized by SEM micrographs as a function of argon content shown in Figure 3.4. Different combinat ions of gas mixtures have been used.

PAGE 51

35 (a) (b) Figure 3.4 – SEM Images of Diamond Films Grown with Different Ar Ratios (a) 50% and (b) 98% Raman spectroscopy [74] is a powerful tec hnique to determine the chemical and structural properties of liquid or solid materials by a simple non-destructive and noncontact method of measurement. In the ca se of Raman spectroscopy of carbon based materials, the scattering is a bout 50 times more sensitive to -bonded amorphous carbon and graphite than to the phonon band of diam ond. Hence, this method can be used to establish the crystall ine quality of diamond thin film s by estimating the amount of sp2 bonded carbon in the films. Lin et al. [75] performed the analysis of diamond films grown with Ar/CH4/H2 plasmas with different gas mixt ures. For films grown without Ar, a sharp diamond characteristic peak is observed at 1332 cm-1. No scattering can be found in the range from 1400 to 1600 cm-1 suggesting that the diamond f ilm contains very little sp2 bonded carbon. With addition of argon to th e reactant gas up to 92%, a sharp diamond peak still exists indicating the presence of microcrystalli ne diamond grains. The typical spectrum of a single crystal diamond, highl y ordered pyrolytic graphite (HOPG), microcrystalline diamond and nanocrystall ine diamond are shown in Figure 5 (a-d) respectively. Apart from Raman spectroscopy, near edge X-ray abso rption fine structure (NEXAFS) is another popular technique for characterizing di amond films. This technique is used in identifying the percentage of sp2 and sp33 bonded carbon in the thin film.

PAGE 52

36 Figure 3.5 – Raman Spectra of Ca rbon Based Materials [54] 3.4 Mechanical Characteristics of NCD Films In order to improve the reliability of mi cromachined devices, it its very important to understand the mechanical pr operties of the thin film used to fabricate them. Amongst others, Young’s modulus, intrinsic stress and the parameters affec ting it, and fracture strength determine the performance of th e devices. In this se ction, the different characterization techniques used to measure these properties will be discussed in detail. 3.4.1 Young’s Modulus According to solid mechanics, Young’s m odulus (E) is a measure of the stiffness of an isotopic elastic material [76]. It is de fined as the ratio of the uniaxial stress over uniaxial strain in the range of stress in which the Hooke’s law is valid. In electrostatically actuated RF-MEMS switches, th e actuation or threshold volta ge is dependent on the spring constant of the material. For thin, long cantilevers with small air gaps, the spring

PAGE 53

37 constant is associated with a small reset fo rce which in turn affects the release of the structure when switched off. Under such circ umstances, it is preferable to use a stiff material with a high Young’s modulus. Sticti on of the moveable membrane due to dielectric charging in capacitive switches ca n be mitigated by using such stiff material. Theoretically, diamond possesses the highest Young’s modulus among all solid materials. The values of 1150 GPa and 960 GPa have been reported for NCD and UNCD films, respectively [77]. Furthermore, the Young’s modulus of diamond is very stable at elevated temperature (7000C) which makes it far superior to other materials. Young’s modulus can be measured using different techniques and in this work it is determined by cantilever resonance meas urements [77]. In this technique, diamond cantilevers of different lengths and widths are fabricated and the first order mechanical resonant frequency is measured using a piezo oscillator. Firs t, boron doped NCD film is grown on a low resistive silicon substrate to a thickness of 1.2 m. Titanium is patterned and used as a hard mask to etch the NCD film. Diamond is th en etched in a RIE system and subsequently released by et ching the silicon layer by CF4 plasma. The entire wafer is then diced wherein individual diamond cantilev ers with varying lengths are stuck to a commercially available piezo crystal oscilla tor using epoxy. Figur e 3.6 (a) shows the SEM image of the diamond cantilevers of different lengths grown in the HFCVD technique and Figure 3.6 (b) shows the SEM image of the released diamond cantilever grown in the MPECVD technique. It is eviden t from these SEM images, that films grown using both techniques have an in trinsic stress that builds up du ring the growth process. (a) (b) Figure 3.6 – SEM Image of Released Diamond Cantilever (a) Using HFCVD, and (b) Using MPECVD Technique

PAGE 54

38 A function generator is used to measur e the first order mechanical resonant frequency. The released diamond cantilevers are attached to a commercial piezo crystal which is in turn connected to the function gene rator. The first order resonant frequency is observed through a microscope as the tip of the cantile ver starts to vibrate at the highest center frequency value. The m echanical resonant frequency is dependent on the length and width of the structure and on the thickness of the film and can be evaluated given in equation 3.1 [77] 2 2* 3 * 4 L t E Fi RES (3.1) In this experiment, the measured resonant frequency is used to evaluate the Young’s modulus of the diamond film. Table 3.2 give s the details of the measured resonant frequency and evaluated Young’s modulus for cantilevers whose width is 40 m and thickness is 1.2 m. Table 3.3 – Measured Resonance Fre quency and Young’s Modulus for Diamond Cantilevers Length of Cantilever Measured Resonant Frequency Young’s Modulus from Measurement 100 m 326 KHz 1015 GPa 300 m 82 KHz 1010 Gpa 3.4.2 Intrinsic Stress Reliable MEMS devices can be fabricat ed by understanding the stress conditions that develop during the deposition (sputtering, evaporation and CVD) process. Stress in NCD films can vary widely (-500MPa to + 700 MPa) depending on the growth parameters that include pressure, substrate temperature and gas ra tio. There are various techniques which are available to measure the compressive and tensile stress in the film [78-80]. In this work, the rota tion tips are used to measure the compressive stress in NCD films wherein the strain is converted into a rotation angle which is directly proportional to the strain caused in the material. Although this technique can be used for measuring both

PAGE 55

39 kinds of stresses (tensile and compressive), our discussion is only focused on compressively-stressed NCD structures. Me asured values show compressive stress ranging from 140 MPa to 557 MPa for a 1.2 m thick film is measured. Figure 3.7 shows the SEM images of the rotation tip stru ctures used for measuring the stress. Figure 3.7 – Released Structures Used for Measuring the Intrinsic Stress in Diamond Films Micromachined structures tend to elonga te or shorten depending on the stress which builds up during the growth process. The cantilever structures studied in this work are compressively stressed resulting in an elongation of the short beams upon release. This elongation on either si de causes the beam to cha nge its angle (rotation). The difference in the distance between the two long beams is used to evaluate the strain using equation 3.2 [80]. Considering the rotational points to be ideal, the structure can be represented as shown in Figure 3.8. As shown in the figure, LA and LB correspond to the length of the sta tionary beam, LC is the length of the moveable beam, W is the width of the moveable beam O is the dist ance between the turning points.

PAGE 56

40 LALB O Figure 3.8 – Cantilever Structure Used to Evaluate Intrinsic Stress Figure 3.9 shows the geometry used to ca lculate the rotation angle for measuring the strain in the film. Equation 3.2 is used to evaluate the strain with respect to the stationary arm. This equation is derived under the assumption that the width of the moveable arms is much smalle r than the stationary arms (LA and LB). B AL L O tan *" (3.2) Figure 3.9 – Geometry Used to Evaluate th e Rotation Angle to Calculate Strain ( ”)

PAGE 57

41 The deflection caused by the moveable arm LC can be defined in terms of the rotation angle as shown in equation 3.3. O L yC2 1 tan (3.3) From equations 3.2 and 3.3, the overall st rain can be represented in terms of the deflection as shown in equation 3.4 O L L L y OC B A2 1 * (3.4) Figure 3.10 shows the SEM image of the moveable arms which have vertical displacement due to the compressive stress in the growth process. Figure 3.10 – SEM Image of the Released Cantilevers with Intrinsic Stress This equation, which shows the linear rela tionship between strain and deflection, can be used for both tensile and compressive stress measurements. The only limitation with this technique is that th e distance between the turning points (O) must be not smaller than a value which results in stiffness that thwarts the movement of the arms. Table 3.3 shows the measured stress value (a s a function of st rain) of a 1.2 m intrinsic diamond film for various lengths of the stationary and moveable arms. The Young’s modulus for these measurements is assumed to be ~ 1020 GPa.

PAGE 58

42 Table 3.4 – Measured Intrinsic Stress of a 1.2 m NCD Film LA & LB ( m) LC ( m) O ( m) 2Y ( m) Stress ( ’) MPa 33 76 8.27 0.305 243 33 76 8.27 0.35 279 50 95 11 0.51 285 50 95 11 0.47 262 Stress during the growth process is depe ndent on many parameters like pressure, temperature and gas mixture ratio. In the HFCVD technique, films with high compressive stress can be achieved at highe r pressure and temperature. Although a thorough mathematical model cannot be deri ved, Figure 3.11 shows the compressive stress that can be achieved with vary ing temperature and pressure conditions. Furthermore, stress distribution is generally uniform over a 1 inch radius of a 4 inch wafer. The uniformity in the stress distributio n is a direct result of the homogeneity of the diamond film during the growth process. Figure 3.12 shows the stress distribution in the diamond film for a 4 inch wafer. 0 20 40 60 Pressure [torr] 650 750 850 Temperature [C] -200 0 200 400Stress [MPa] -200 0 200 400Stress [MPa] Figure 3.11 – Measured Intrinsic Stress w ith Varying Temperature and Pressure Conditions

PAGE 59

43 0 1 2 3 4 5 X [cm] 0 1 2 3 4 5 Y [cm] 200 250 300 350 400 Stress [MPa ] 200 250 300 350 400 Stress [MPa] Figure 3.12 – Measured Compressive Distribution Across a 4-inch Wafer 3.5 Summary Growth of NCD films using MPECVD and HFCVD techniques has been presented. The effect of using different seed ing techniques on the quality of the diamond films have been presented w ith examples. The growth pro cess involved in diamond films using hydrogen and argon chemistry has been di scussed in detail. Mechanical properties of the diamond films that include Young’s mo dulus, mechanical resonant and intrinsic stress have been measured. Furthermore, the effect of process parameters (temperature, gas ratio and pressure) on the mechanical pr operties of NCD films was described with a mathematical model wherein intrinsic stress m easurements and the f actors affecting this intrinsic stress were studied in detail.

PAGE 60

44 Chapter 4 Thermal and Dual Actuation Nanocrystal line Diamond Bridges and Cantilevers 4.1 Introduction Radio frequency micro-electro mechanic al systems (RF-MEMS) technology has been growing rapidly in the last few years and has found a multitude of applications in the automotive, defence and communicatio n industries. Due to their outstanding properties and excellent performance at high frequencies, RF-MEMS devices have started replacing their solid state counterparts in filters [81], phase shifters [82] and antennas [83]. According to a 2006 market study, the global market for MEMS, with RFMEMS being a major part will reach $12 b illion by 2010 [84]. Although they are small in size and exhibit low parasitic losses, the mono lithic configuration confines the device to a substrate common to the entire system. Furt hermore, power handling [38] capabilities of RF-MEMS devices have been limited due to us ing all metal structures in the devices. Diamond has long been used in active devices such as FETS for high power electronics [85]. Also its stabil ity under high temperature makes it a very good candidate for realizing reliable, high power and temperatur e stable RF-MEMS and microwave devices. The outstanding mechanical properties of na nocrystalline diamond (N CD) thin films and its low loss at microwave frequencies can be used to produce mechanically stable and high power RF-MEMS devices In this chapter, the life cycle of a thermally actuated NCD bridges and cantilevers will be presented. The diamond actuators are designed to operate in a bi-stable mode. Design equations realizing the thermal actua tion scheme and simulation results which facilitate the bi-stable operat ion will be discussed. The ove rall design of the actuator and the choice of material will also presented. Fabrication st eps along with th e solid-liquid interdiffusion (SOLID) process used for integrating the actuator will be discussed in detail. Small signal measurements are carried out in the frequency range of 1-30 GHz and

PAGE 61

45 compared with the simulation results, which are performed using Agilent’s Advanced Design System (ADS). Finally, high power measuremen ts of the switch integrated in the microstrip topology will be presented. In addi tion to thermally actuated switches, a dual mode switch which combines thermal and electrostatic actuation schemes is also discussed. The fabrication and measurement re sults of this switch are also presented. 4.2 Design of the NCD Actuator The NCD actuator is made of boron dope d diamond film which is grown through a hot filament CVD process. Figure 4.1 s hows the design of the NCD actuator. The diamond bridge is 1200 m long, 300 m wide and 1.5 m thick. The diamond film used in the actuator is grown to purposefully achieve compressive stress of ~ +300 Mpa. The parameters involved in achieving these stress va lues have been discussed in Chapter 3. The resistivity of the bor on doped NCD film is ~ 1m -cm. In addition to the fixed-fixed bridges, NCD based cantilevers (500 m, 1000 m in length) were also designed. Figure 4.1 – Design of the Thermally Actuated NCD Bridge The bridges are thermally actuated using a bi-metal actuation scheme [86]. Compared to electrostatic actuation, therma l actuation has the advantage of having a lower actuation voltage and a higher contac t force. The main drawback of thermal actuation is the static power consumption, which can be avoi ded by using a designing the actuator to operate in a bi-stable mode. Coppe r which is used to facilitate the thermal actuation scheme is deposited on top of the doped diamond thin film. In addition to the

PAGE 62

46 actuation pads, copper is also used in the cont act pads of the actuato r. The difference in the coefficient of thermal expansi on between the two materials (0.8-6/K for diamond and 13-6/K for copper) results in re sistive heating of the doped areas. This resistive heating effect causes the di amond actuator to bend thereby switching to the actuated state. The pull-in voltage (and current) to switch the bridge depends on the geometry of the diamond heating elements. Figure 4.2 (a) and Figure 4.2 (b) show the top view of the different actuator designs. These designs were optimized in ANSYSTM to achieve the best performance in terms of bending moment, deflection and ease of integration. Gold wirings are included in the design to provide a DC electrical pa th to the contact pad at the center of the diamond bridge. The contact pads are 100 m x 100 m in dimension. Figure 4.2 Top View of the Diamond Actuator with Different Copper Heating Elements As explained in Chapter 3, the buckling effect in the beam is obtained by optimizing pressure, gas composition and temperature during the NCD growth. This buckling effect can also be explained through a mathematical formula. According to theory of beam mechanics analysis, a straig ht beam can be axially compressed to yield buckled stable positions dependi ng on the nature (axial or la teral) and position (extreme or center) of the load. A straight beam which is subjected to an axial load p can be represented as [79]

PAGE 63

47 02 2 4 4 dx wd p dx wd EI (4.1) where w is the lateral displacement of the beam E is the Youngs modulus of the beam, and I is the moment of inertia of the beam. By assuming fixed boundary conditions at both ends and normalizing the axial for ce equation 4.1 can be represented as DCx l Nx B l Nx Aw cos sin (4.2) where A, B, C and D are constants. The comp ressive force generated can be categorized into various modes and the buckling caused in the beam can vary depending on the force generated (F) at each mode. Initially, the beam is straight, if the compressive energy (EC) generated is less than beam bending energy (EBB). As the compressive force changes (F=F1), EBB is greater than EC, and this results in buckling of the beam. The buckling effect in the first mode increases the ove rall length of the beam thereby lowering the compressive force. Figure 4.3 shows this buckling effect of the beam. (a) (b) Figure 4.3 Stages in a St raight Beam which is Compressively Stressed (a) EC is Greater than EBB and (b) EBB is Greater than EC Mechanical simulations we re performed in ANSYSTM to optimize the length of the beam to achieve the bi-sta ble condition. Figure 4.4 show s the simulated contact force for double anchored cantilever s of different lengths vers us the separation distance between the actuator and integrated substrate.

PAGE 64

48 0 1 2 3 4 5 6 7Distance [ m] 0 0.05 0.1 0.15 0.2 0.25 0.3F o r c e [ mN ] 800 microns 1000 microns 1200 microns A B C Figure 4.4 – Simulated Force vs Distance of Separation Between Actuator and Host Substrate Many aspects have to be taken into account while designing the bridge. Intrinsic stress, thickness of the diamond layer and the length of the beam determine the buckling effect. Furthermore, these parameters togeth er with the separation between the beam and the base plate influence the magnitude of the force, which is generated against the base plate. At the point A (near 6 m), the force is zero, since the beam is completely released and just touches the base plate. When the separation is reduced, the cantilever is in contact and generates a center force on the base plate. This force increases with decreasing separation and r eaches a maximum, at 3.4 m (point B), before it drops immediately back to zero. At point C (~ 3 m) the beam can no l onger reach the lower stable state; the beam flips back and becomes mono-stable. For very long beams the center force decreases steadily from its maximum, because an s-shaped quasi-stable interstate appears be fore it becomes monostable. Although both 1000 m and 1200 m long beams gave satisfactory results, the longer beam was chosen to accommodate the copper heating elements for better actuation. In thermal actuation, the amount of thermal expansion for a material can be given by

PAGE 65

49 dv vT T th th1 0) ( (4.3) where T0 and T1 are the lower and uppe r temperature limit respectively, and th is the coefficient of thermal expansion. The expansion th and the strain produced varies as a function of temperature wherein T is the change in temperature, x is the stress and E is the Young’s modulus. th and strain ( ”) can be represented as T l lth th (4.4) Ex (4.5) In the bridge or cantilever structure, the generated thermal strains in both materials (NCD and copper) are converted in to intrinsic stress leading to a bending moment. This in turn leads to the deflecti on of the actuator. Acco rding to Hooke’s law [87], stiff materials that posse ss high Young’s modulus lead to high stress values. For this a proper figure of merit for thermal st ress generation is the product of E and th. Upon heating, the bending moment caused by the ma terial is dependent on its Young’s modulus (E), thickness, length and wi dth. To achieve the maximu m bending moment the product of the E and th of NCD should be differe nt (lower or higher) th an the other material. Table 4.1 shows the E, th and their product for different mate rials as compared to that of NCD. It is seen from this table that a comb ination of copper and NCD is nearly ideal for the thermal actuation scheme, second only to the copper-Ni combination. Table 4.1 – Comparison of Young’s Modulus (E ) and Coefficient of Thermal Expansion ( th) for Different Materials Property NCDCu Ni Au Al Young’s Modulus [GPa] 1050 130 210 78 70 CTE ( th) [ppm/C] 0.85 17 13 14 23 E* th 840 2210 2730 1092 1610

PAGE 66

50 In the first iteration of the design, nickel was used as the heating element. Although the product of E and th is higher in nickel than copper, Cu was preferred for because fewer fabrication steps are required and the integration process is simplified. Figure 4.5 show the operation of the actuator in the down-stat e and up-state, respectively. The up and down movement of th e actuator is dependent on th e heating of the individual copper heating element placed on top of the doped diamond film. The corner two blocks are used to push the bridge dow n and the center block is used to bring the bridge back to the up-state. As mentioned earlier, the actuator is designed to be bi-stable, and hence the design must accommodate heating elements for transforming the bridge between both states. Figure 4.5 – Bi-Stable Layout of the Actuator with Individu al Copper Heating Elements 4.3 Fabrication The diamond bridges are fabricated on a 500 m thick low resistivity silicon wafer. Prior to the 2nd generation design a diamond actuato r was fabricated in which the bridge was composed of intrinsic diam ond with areas of se lectively grown doped diamond. The doped diamond bridge was chosen so that resistance of the heaters can be lower without actually making the bridge thicker and stiffer. Furthermore, the first generation switches included ni ckel to facilitate the th ermal actuation scheme. The fabrication steps as shown in Figure 4.6 are as follows: The silicon wafer is nucleated by BEN (bias enhanced nucleation) and an intrinsic diamond layer of 1500 thickne ss is grown thro ugh a microwave plasma assisted CVD process. Boron doped diamond (p-type) is later grown

PAGE 67

51 with HFCVD (hot filament CVD) to a thickness of 8500 . This boron doped diamond is the heart of the micromachined actuator. Intrinsic diamond is selectively grown using a SiO2 mask. The 4000 thick diamond layer is used for electrical is olation of the contact areas while actuating the bridges. A Cr/Au seed layer of 700 is deposite d using an ion beam reactor after which a 1 m thick copper film is deposited by electroplating to serve as the bi-metal for thermal actuation. Copper pads which are used to inte grate the diamond switches onto the host substrate are electroplat ed to a thickness of 12 m. The RF contact areas are also formed by electropl ating in this step. The previously deposited seed layer is patterned to provide electrical continuity to actuate the bridges. 400 of platinum is patterned over th e copper contact ar ea using lift-off technique. Diamond bridges are then etched in a RIE system using titanium as the hard mask Finally, using patterned silicon dioxide as a backside hard mask, diamond structures are released from the silicon wafer through a DRIE process resulting in a free stan ding diamond bridge that is embedded in a silicon frame.

PAGE 68

52 Figure 4.6 – Fabrication Procedure of the Nanocrystalline Diamond Actuator Figure 4.7(a) and 4.7(b) show s the front view and the back view of the fabricated diamond actuator. The overall size of the entire chip is 1600 m long and 900 m wide.

PAGE 69

53 Figure 4.7 Microphotograph of the Fabricated Diamond Actu ator (a) Front View (b) Back View with the Silicon Frame After the release process, each actuator (with the frame) is individually diced using a dicing saw. To avoid an y damage the released actuators are glued onto a 3-4 inch silicon substrate. After dicing, they are clean ed in acetone and methanol to remove the glue. These diamond actuators, being embedde d in the silicon frame, are substrate independent and can be integrated onto any microwave substrate th at can withstand the high-temperature SOLID process [88]. Figure 4.8 shows the integration process of the diamond actuator to the host substrate in the CPW topology. In addition to CPW switches, the actuators are in tegrated to realize tunabl e CPW inductors and tunable switches in the microstrip topology. Figure 4.8 Diamond Actuator Integrated onto a Host S ubstrate (Alumina, Aluminum Nitride) Using SOLID Process Contact Area Copper Bi -metal Gold wiring Silicon Frame Copper Pads Diamond Bridge Silicon Frame(a) (b )

PAGE 70

54 In addition to the copper pads in the actuator frame, copper pads along with tin should be included in the host substrate for integration. Solid–liquid interdiffusion occurs between the two phases, resulting in a phase transformation of the liquid component to a higher melting point material which is strong enough to serve as a bond and withstand elevated temp eratures. Copper-Tin has been considered a SOLID [88] couple where tin acts a melting phase and copper as a solid phase. While integrating, the copper pads on the actuator are kept on top of the copper-tin stack and heated till 2500C (melting point of tin). During the diffusion process one of the inter-metallic compounds forms the bond between the two structures and is stable till 6000C. The alignment is carried out in a flip-chip bonding test setup wherein an accuracy of + 5 microns is achieved during the integration. Heating fo r the solder process is supplied from the actuator part to prevent alloying of the pads that are located on the host substrate. Figure 4.9 shows the phase transformation in the SO LID process. Apart from Cu/Sn the other popular metal combinations are Ag/In, Au/Pb and Au/Sn. liquid solid solid new melting point 640C CuxSnysolid state diffusion reaction Tsolder= 250 C CuSn T high melting phase liquid solid solid new melting point 640C CuxSnysolid state diffusion reaction Tsolder= 250 C CuSn T high melting phase Figure 4.9 – Phase Transformati on in the Cu-Sn SOLID Process In the second generation switches, tin was in cluded in the actuator part instead of the host substrate. This approach was chosen for better yield in the tin electroplating

PAGE 71

55 process due to fabrication limitations. Figure 4.10 shows a microphotograph of the integrated switch with the host substrate. Th e four outer rectangles are copper blocks used for biasing the switches. Figure 4.10 – Integrated Switch with a Host Alumina Substrate in a Microstrip Topology 4.4 Small Signal Analysis of the CPW Integrated Switch The diamond actuators are used to realiz e RF-MEMS DC contact type switches in CPW and microstrip topologi es. The CPW transmission lines are designed on a 650 m thick alumina substrate ( r=9.9, tan =0.0002). The transmission lines are 3000 m long with a center conducto r width (W) of 100 m and slot width (G) of 50 m. The center conductor of these lines is purposefully in terrupted in the middle resulting in two transmission lines which are 1475 m long. During actuation, the contact pad in the diamond bridge closes this gap resulti ng in a continuous transmission path. Small signal measurements were done in the frequency span of 1-30 GHz using an Anritsu Lightning VNA. A bias tee was used to protect the VNA test ports from DC current. Before measuring the structures a probe-tip SOLT calibrati on is performed on a commercial GGB CS-9 calibration. The diamond bridges are thermally actuated at 2 volts and upon actuation the platinum coated copper pad makes c ontact with the CPW line. Small signal simulations were performed using ADS Momentum, a 2.5 D electromagnetic simulator. Figure 4.11 and Figure 4.12 show the comparison between the simulated and measured S11 and S21 in the non-actuated and actuated states for the CPW integrated switch, respectively. The return lo ss and insertion loss in the actuated state are

PAGE 72

56 20 dB and 0.2 dB, respectively, at 20GHz. It is evident from the S-parameters that in the actuated state, the diamond bridge makes a very good contact with the transmission line with a contact resistance of ~ 0.8 ohms. 0 5 10 15 20 25 30 35 40Frequency [GHz] -1.2 -1 -0.8 -0.6 -0.4 -0.2S 1 1 [ d B ] -50 -40 -30 -20 -10S 2 1 [ d B ] Simulated S11 Measured S11 Simulated S21 Measured S21 Figure 4.11 – S-Parameters of the CPW Integr ated Actuator in the Non-Actuated State 0 5 10 15 20 25 30 35 40Frequency [GHz] -50 -40 -30 -20 -10S 1 1 [ d B ] -1.2 -0.9 -0.6 -0.3 0S 2 1 [ d B ] Simulated S11 Measured S11 Simulated S21 Measured S21 Figure 4.12 – S-Parameters of the CPW Inte grated Actuator in the Actuated State 4.5 Small Signal Analysis of the CPW Inductor Inductors are integral com ponents in RF front end ar chitectures that include filters, matching networks and tunable circuits such as phase shifters. The most common

PAGE 73

57 inductor topologies include planar spirals, air-core, and embedded solenoid designs [7]. In comparison to capacitors, however, relativel y few tunable inductor configurations have been published, and among those presented many are hybrid approaches that employ (MEMS) switches to activate di fferent static inductive s ections. Furthermore, less attention has been paid to desi gns that enable control in the sub-nH range as is potentially desirable for matching purposes in applicatio ns that use distributed loading of small capacitances, e.g. in loaded-line phase shifters [7]. From transmission line theory, narrow or thin high impedance transmission lines are analogous to inductors. The inductance valu e is dependent on the width and length of the line. Generally speaking th e net inductance value increases with an increase in length and decrease in width. In addition to impl ementing the devices as a simple RF switch, tunable inductors were reali zed wherein the non-actuated a nd the actuated-s tate of the bridges yield different net induc tance values [89]. The inductor circuits fabricated on the alumina substrate are 400 m long. Figure 4.13 shows the inductor layout along with the integrated diamond bridge. Figure 4.13 Design of the Integrated CPW Inductor a nd Diamond Actuator The geometry of the transmission line is a crucial factor in determining the net inductance value in the non-actuated (Lhigh) state and in the actuated state (Llow) of the inductor. The inductors are designed such that the characteristic im pedance of the feed lines is close to 50 Ohms. The impedance valu e of these feed lines is a function of the center conductor width and slot width. In th is work, the center conductor width (w) and

PAGE 74

58 slot width (s) is 270 m and 115 m, respectively. The overall length of the inductor circuits is approximately 400 m. Although the inductance va lue can be increased by making the center conductor na rrower, processing limitations restricted the minimum width to 30 m. These CPW lines are designe d in a way that the effective inductance value (Leff) is much greater in comparison to the ca pacitance in the transmission line. The inductance value for all the designs is contro lled by the fixed-fixe d beams, wherein the Leff values differ with the actuated and non-actua ted state of the beam. The inductors are designed to have a high inductance value when the beams are up and low inductance value in the actuated state of the beams. Wh en the beams are in the non-actuated (up) state, the circuit offers high impedance (Zh) to the input signal. Since the length of the device is electrically small (0.075 at 25GHz) the topology effectively emulates a lumped inductor with high inductance value (Lhigh). In the down-state the inductance value (Llow) reduces due to the decrease in the effective characteristic impedance (Zl). The inductance ratio (Lratio) is directly related to the change in the impedance states (Zh/Zl) and is defined as (Lhigh/Llow). For a short electrical length (< /4) a high impedance transmission line section emulates a series inductor as given by equation 4.6. Tunabl e operation is achieved by changing the effective width of the slot a nd/or the center conductor by using the diamond actuator. 12;hl duZZ LL qq ww == (4.6) In the non-actuated state of the actuator the inductor re presents a high impedance state which translates into a high inductan ce value. When the bridges are actuated the contact pads makes a DC short with the transmission lines, leading to low impedance and in turn a low inductance state. Figure 4.14 (a) and (b) shows the top view the tunable inductor wherein the change in width of the transmission line correspond to the change in inductance between the two states.

PAGE 75

59 Lhigh Llow Beam in Up-State Be am in Down-State (a) (b)Figure 4.14 Change in Impedance and Effective Inductance Between the Up-State and Down-State of the Tunable Inductor The inductors were measured in the fr equency range of 1-30 GHz in the same setup as that of the RF switches. Figure 4.15 and Figure 4.16 shows the return loss and insertion loss, respectively of the tunable inductor in the non-actua ted and the actuated state of the diamond bridge. 0 5 10 15 20 25 30Frequency [GHz] -50 -45 -40 -35 -30 -25 -20 -15 -10 -5S 1 1 [ d B ] Up-State Down-State Figure 4.15 Measured Return Loss (S11) of the Inductor in the Non-Actuated and Actuated State

PAGE 76

60 0 5 10 15 20 25 30Frequency [GHz] -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0S 2 1 [ d B ] Up-State Down-State Figure 4.16 – Measured Insertion Loss (S21) of the Inductor in the Non-Actuated and Actuated State The effective inductance (Leff) of the circuit is extrac ted by numerically shorting port 2 of the inductor and is re lated to the input impedance (Zin) by equation 4.7. freq Z Lin eff* 2 } Im{ (4.7) The measured inductance in the two states and the i nductance ratio (Lratio) are shown in Figure 4.17. An inductance ratio of 2.2 was achieved at 30 GHz with 1.2 being the maximum inductance value.

PAGE 77

61 0 5 10 15 20 25 30Frequency [GHz] 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2I n d u c t a n c e [ n H ] 1.2 1.4 1.6 1.8 2 2.2I n d u c t a n c e R a t i o Up-State Down-State Figure 4.17 – Measured Inductance in Up and Down States and Inductance Ratio 4.6 Large Signal Measurements – 1st Generation Power handling capabilities of RF-MEM S devices have been previously investigated for series [38] and shunt capac itive switches [90] at X-band frequencies up to an input power of 7 watts. In this work, the micromachined diamond air-bridge integrated in the microstrip topology is test ed at 2.1 GHz until 45 watts. Prior to testing, the microstrip substrate with the diamond brid ge is modified for the measurement setup. The alumina wafer is solder attached to a brass carrier. Two 250 mil long, 50 microstrip lines are fabricated on a 31 mil FR 4 substrate and solder mounted on the either side of the alumina wafer; this is done to connect the SMA coaxial adapte rs on either side of the carrier for testing. Bond wires which ar e 3-mil in diameter are used to connect the 50 lines with the microstrip line on the alum ina substrate. Figure 4. 18 shows the details of the coaxial test fixture us ed for high power measurement.

PAGE 78

62 Figure 4.18 – Diamond Actuator Integrated in Microstrip Topology for High Power Testing The setup for high power measuremen t is shown in Figure 4.19. Maury Microwave ATS 400 software was used to extract the data. An Anritsu 68169B signal generator is used as the RF source. Th e frequency and input power level for the measurements are 2.1 GH and 15 dBm, respectivel y. In the next stage the input signal is passed through a 100 watt amplifier with a gain of 50 dB. A three port circulator is used as an isolator with the thir d port terminated by a 100 watt load, in order to isolate the signal source and the amplifier from potentia l high reflected power. The isolator and the brass carrier are mounted on a heat sink to avoid therma l issues. The signal source, amplifier and the isolator constituted the input side of the high power setup. The output signal from the diamond actuator (DUT) is passed through stages of high power attenuators before measuring the power leve l in an Anritsu power meter. Prior to measurement, a two port thru calibration is pe rformed wherein the reference planes were shifted to the inner edges of the 31 mil FR4 transmissi on line. Measurements are performed in the power range of 24-47 dBm at 2.1 GHz. For the initial test circuits, electrical actuation of the diamond actuators did not result in intimate contact of the copper pads with the host subs trate due to issues associated with the flip-chip bonding. To work around this issue, the diamond bridges were mechanically actuated in addition to the applied DC voltage.

PAGE 79

63 Figure 4.19 – High Power Bench for Testing Diamond Switches The switches are tested fo r self actuation and isolat ion in the up-state. The measured insertion loss (or isolation) in th e non-actuated state is shown in Figure 4.20. As seen from graph, the measured isolation is around 16-17 dB at 2.1 GHz in the entire power range. Unlike most of the RF-MEM S switches, the bias pads and the RF transmission lines are isolated from each ot her in the integrated diamond actuator. This isolation prevents the bridge s from self actuation problems when an input RF power is applied. Furthermore in the initial high power tests, some variation in the device performance was noticed and complete co ntact was not obtaine d using electrical actuation; this can be attributed to the he ight differences in the flip-chip mounted structures. Therefore the devices used in th e high power testing required such mechanical actuation, which was achieved using a needle probe attached to a micro-positioner.

PAGE 80

64 15 20 25 30 35 40 45 50Power [dBm] -25 -20 -15 -10I s o l a t i o n [ d B ] Figure 4.20 – Measured Isolati on of the Diamond Actuator in the Non-Actuated State with Varying Input Power As shown in Figure 4.21, the insertion loss is around 0.2-0.3 dB throughout the entire power spectrum. Meas urements were made for mo re than one instance for repeatability and there was no significant difference in the insertion loss. The diamond bridge was stable at such power with little or no damage due to heating of the structure. Diamond which also is an excellent conduc tor of heat acts a good heat sink during measurement. As an improvement to this device, a dual actuation scheme based NCD bridges are being developed. In this scheme, the advantages of electrostatic and thermal actuation are used together to de velop more reliable NCD actuators.

PAGE 81

65 23 28 33 38 43 48Input Power [dBm] -0.6 -0.5 -0.4 -0.3 -0.2 -0.1I n s e r t i o n L o s s [ d B ] Cycle 1 Cycle 2 Figure 4.21 – Measured Insertion Loss of th e Diamond Actuator in the Actuated State with Varying Input Power 4.7 Dual Mode Actuation of the NCD Switches As discussed in Chapter 2, a potential di sadvantage of therma l actuation is that DC power may be required to hold a device in the ON-state or OFF-state. This problem was addressed by developing actuators which operated in a bi-stable mode, wherein the actuators remain down after removing the voltage. An alternate method to address this issue is to develop switches which can actua te at a low voltage and avoid the current consumption issue by adopting a latching appro ach. Saias et al. [19] has demonstrated this technique for medium to high power a pplications using sta ndard micromachining. The switches are fabricated on a silicon subs trate in a CPW topology and are actuated using an integrated 300 m X 300 m CMOS driver. Once the devices are in the on-state latching electrodes secure the MEMS beam and the drive voltage/current can be removed. In the second generation switches studied in this work the diamond actuators are integrated onto aluminum nitride (AlN) substrates. In comparison to alumina (Al2O3), aluminum nitride has higher thermal conduc tivity (140 W/mK for AlN and 18-20 W/mK for Al2O3). Furthermore, this integr ation process is also done to exhibit the substrate independent nature of the diamond actuators Furthermore, in addition to thermally

PAGE 82

66 actuating the switches, an electrostatic latch feature is used to hold the switch in the actuated state. In the absence of a bi-stable operation, this dual mode actuation (thermal + electrostatic) scheme will be useful to avoid the power consumption issues. The electrostatic latch is incorporated on the hos t substrate. High resistive silicon chrome (SiCr) is deposited using an e-beam evaporator to a thickness of 1500 A0. Silicon nitride which serves as an isolation layer is patter ned on top of the SiCr bias lines. After the switches are thermally actuated, SiCr and the doped diamond actuator are used as the two electrodes to facilitate the electrostatic latch mechanism. Figure 4.22 shows the description of this dual mode actuation scheme. The SiCr bias lines do not come in contact with the diamond bridges upon actuation and hence the i ssue of an electrical short can be avoided. In the CPW version of th e switches, the electrostatic latch can be implemented by positioning the SiCr lines in the slot region or by using the ground planes as the latch electrode. Figure 4.22 Fabrication and Integration of the Dual Mode Actuation Scheme of the NCD Bridges

PAGE 83

67 Figure 4.22 (Continued) 4.8 Large Signal Measurements 2nd Generation In the thermal scheme, the switches are first actuated at 3 volts with a current value of 20 mA. Then the electrostatic ho ld down voltage of 2830 volts is applied between the SiCr pads the NCD bridge. Furthermore, due to equipment limitations, measurements for the 2nd generation switches are performed at a CW frequency of 1.9 GHz. In order to check the power handling limit of the NCD bridges, the actuators are gradually tested at increasing power levels starting from 20 dBm to 40 dBm. This is done to better understand the perfor mance of the actuators and also avoid any damage at high power levels. Figure 4.23 shows the measured insertion loss as functi on of power. In this graph cycle 1 corresponds to a maximu m power level of 30 dBm and cycle 10 corresponds to the measured insertion loss at a maximum input power of 40 dBm. Cycle

PAGE 84

68 1 to cycle 10 increases in power level by 1 dBm. From the measured results we can conclude the increased power level at little to no effect on the in sertion loss. The high thermal conductivity of NCD film s prevents issues like welding and hot spot formation at these elevated power levels. The measurem ents are stopped at 40 dBm due to the limitations of the high power amplifier used. New measurements are under progress to test the switches at high power levels by in corporating small changes in the large signal measurement setup. 20 22 24 26 28 30 32 34 36 38 40Power [dBm] -1 -0.8 -0.6 -0.4 -0.2 0S 2 1 [ d B ] Cycle 1 Cycle2 Cycle 3 Cycle 4 Cycle 5 Cycle 6 Cycle 7 Cycle 8 Cycle 9 Cycle 10 Figure 4.23 – Measured Insertion Loss at a CW Frequency of 1.9 GHz with Maximum Input Power Varying from 30 dBm to 40 dBm In order to understand the performance of the NCD actua tors at high power levels for a longer period of time, measurements are al so carried out in the power span of 20-37 dBm. These measurements are carried out af ter actuating the NCD br idges continuously for a maximum of 2 hours using an Agilent 33120A function generato r. A square wave signal at a frequency of 1 KHz (value lesser than the first order mechanical resonant frequency) with a peak to peak (VPP) voltage of 4 volts is used for actuating the switches. Firstly the switches are actuated using the dual actuation scheme and th en insertion loss is measured in the power span of 20-37 dBm. After the first set of measurements, the switches are actuated continuously using the thermal scheme for a period of 30 minutes. The high power measurements are once again car ried out after that duration to see the

PAGE 85

69 depreciation, if any, in the performance of the switches. This is repeated for a period 30120 minutes. The above mentioned power level is chosen to ensure the actuators do not suffer any mechanical degradation over time. In addition to measuring the insertion loss, the switches are also tested for self actuation after the above mentioned 2 hours of testing. The actuators continued to exhibit no self actuation issues after actuating them continuously for 2 hours. Figure 4.24 and Figur e 4.25 show the measured isolation loss and insertion loss, respectively. It is seen that the actuators exhibited consistent performance with no degradati on in the mechanical stability and electrical performance of the switch. Further measur ements are under progress to te st the switches at high power levels for a longer period of time. 20 22 24 26 28 30 32 34 36 38Power [dBm] -30 -25 -20 -15 -10 -5S 2 1 [ d B ] 0 Minutes 30 Minutes 60 Minutes 90 Minutes 120 Minutes Figure 4.24 – Measured Isola tion Loss After Two Hours of Continuous Actuation of the Diamond Bridges

PAGE 86

70 20 22 24 26 28 30 32 34 36 38Power [dBm] -1 -0.8 -0.6 -0.4 -0.2 0S 2 1 [ d B ] 0 Minutes 30 Minutes 60 Minutes 90 Minutes 120 Minutes Figure 4.25 – Measured Insertion Loss After Two Hours of Continuous Actuation of the Diamond Bridges 4.9 Summary In this work, thermally act uated nanocrystalline diam ond bridges are presented. Intrinsic and boron doped NCD f ilms have been grown on low resistive silicon wafer. The films are intentionally grown with built in compressive stress and the measured values were in the range of 140560 MPa. Design, fabrication and integration of the compressively stressed bi-stable actuators are presented in detail. Experimental results of a RF switch and tunable inductors using na nocrystalline diamond bridges are presented. The switches were thermally actuated and the measured results show a return loss and insertion loss of 20 dB and 0.2dB, respectivel y, at 20GHz. Tunable inductors were also designed and measured in th e frequency range of 1-30GHz and the measured results show a inductance ratio of 2.2 at 30 GHz. In addition to the thermally actuated switches, the second generation switches are designed u tilizing both therma l and electrostatic actuation schemes. High power measurements are performed on the diamond actuators in the power range of 24-47 dBm for the mechan ically actuated sw itches, and 24-40 dBm for electrically actuated switches. The measur ements show an insertion loss of 0.2-03 dB in the entire power spectrum.

PAGE 87

71 Chapter 5 Nanocrystalline Diamond Capacitive Shunt Switches 5.1 Introduction RF-MEMS devices have been implemented in a broad range of designs as DC contact [89] and capacitive switches [13] in series and shunt configurations [11]. This development work has facilitated significan t maturing of the technology and technical challenges such as reliability [91], packagi ng [41] and high power operation [42] are now relatively well understood. In particular, the di electric charging mechanism [26] is one of the main factors that limits the reliability of capacitive switches. This charging mechanism has been addressed in detail by several researchers [26-29] and has been discussed in Chapter 2. The usage of bi-pol ar actuation voltage [92] is one of the accepted means of mitigating this phenomenon. Further reduction of charging effects can be realized with the use of leaky dielectric s, given that the associated microwave loss does not significantly impair the RF switch performance. The finite DC conductivity of the diamond layer provides a conductive path for possible trappe d charges [93]. In this work, a prototype design of a mm-wave shunt capacitive switch in coplanar waveguide, using NC D as the dielectric is presented. While the superior material properties of NCD films have been utilized in realizing electrostatic [94] and thermally [95] actuated switches, this is fi rst demonstration of RF-MEMS switches using NCD as the insulator. The devices are fabric ated on a high resistivity silicon substrate using standard lithography and surface micr omachining techniques. Small signal measurements are performed in the frequenc y range of 1-65 GHz, along with preliminary charging studies that have been performed using Corona-Kelvin metrology (CKM) and standard I-V techniques. Both techniques confirm the leaky nature of the NCD film, where the voltage time rate of decay obser ved with the CKM technique and the leakage current in the I-V technique are increased relative to the PECVD nitride films.

PAGE 88

72 5.2 Design and Simulation Results Figure 5.1 shows the geometry of the shunt switch. The switch is designed in a CPW topology and the dimensions of the CP W feedlines are determined using ADS LinceCalc [96] to be 100 m for the center conductor and 60 m for the slot, yielding a characteristic impedance of 50 ohms. The switches are 320 m in length and 50 m in width. The overlap area of the switch with the dielectric is 130 x 130 m. Figure 5.1 Cross Section of the NCD Capacitive Switch The capacitive switch is actuated usi ng the electrostatic actuation scheme, wherein a DC voltage is applied between the MEMS bridge and the CPW transmission line. This creates an electrostatic force resu lting in the collapse of the suspended bridge on the dielectric layer. In the down-state, the capacitance contributed by the metalinsulator-metal structure increases by a factor of 30-50. The increased capacitance connects the bridge and ground of the tran smission line thereby creating a short at microwave frequencies. Once the voltage is removed, the bridge returns to the normal position by the restoring force. Undoped NCD fi lm, which is the dielectric layer, is deposited using the MPECVD tec hnique to a thickness of 0.5 m. The relative dielectric constant of the dielectric NCD film is ~ 5. The movable membrane is 1.5 m thick and is suspended 2 m above the dielectric layer. The switch design can be electrically represented as a shunt LCR circuit. Figure 5.2 shows the equivalent circuit of the capacitive shunt switch. In this L, is the bridge

PAGE 89

73 inductance, CUp and CDown represents the parallel plate capacita nce in the non-actuated and actuated state, respectively. The seri es resistance “R” is contributed by the transmission line and the bridge. Although a be tter model can be derived to separate the individual resistance valu es, in this work, R accounts for both the components. TL Port 1 C R Port 2 TL L UpCDown/ Figure 5.2 – Equivalent Circuit of the NCD Capacitive Shunt Switch The shunt switch impedance is dependent on all three lumped elements and can be represented as C j L j R Z 1 (5.1) where C changes between CUp and CDown depending on the actuation state of the switch. The LC resonant frequency of the switch is given by LC Fres2 1 (5.2) The impedance of the switch changes de pends on the frequency at which the switch operates. Below the resonant frequenc y, the circuit operates as a capacitor and above the resonant frequency it operates as an inductor. At the resonance frequency, the model is dominated by the series resistance. The up-state capacitance is a combination of

PAGE 90

74 the parallel plate capacitance (Cpp) and the fringing field capacitance (Cff). This fringing field capacitance is generally 3040 % of Cpp, which can be represented as r ppd g wW C 0 (5.3) where 0 and r are the absolute and relative dielect ric constants, respectively. W and w are the widths of the transmission line and beam, respectively. g is the air gap between the transmission line and bridge and d is the thickness of the NCD dielectric layer. The down-state capacitance can be given by d wW Cr Down 0 (5.4) For the switch designed in this work, the i nductance L does not have a significant effect on the up-state loss within the intended frequency band of operation. Hence the equivalent circuit shown in Figure 5.2 can be modeled only as a capacito r in this state. In addition to the basic design, an induc tively tuned shunt sw itch (Design 2) has also been designed. Figure 5.3 shows the ba sic and inductively tune d designs. Inductively tuned designs are used to achie ve high isolation at lower fr equencies (X band etc). This can be done by increasing the shunt induc tance while maintaining the down-state capacitance of the switch Figure 5.3 Top View of the Basic and I nductively Tuned NCD Capacitive Switch

PAGE 91

75 Figure 5.4 shows the isolation of the MEMS switch with a down-state capacitance of 1.26 pF. The capacitance value is similar to the standard switch. The change in resonant frequency is contributed by the ch ange in inductance from the high impedance section transmission line between the br idge and ground plane. Simulations are performed for different values of L1, which is the length of the hi gh impedance line. As seen from Figure 5.4, the resonant frequenc y shifts from 30 GHz to 20 GHz with L1 changing from 50 m to 200 m. For convenience in this wo rk, L1 is chosen to be 150 m. 0 10 20 30 40 50 60 70Frequency [GHz] -45 -40 -35 -30 -25 -20 -15 -10 -5 0S 2 1 [ d B ] 200 Microns 150 Microns 100 Microns 50 Microns Figure 5.4 – Simulated Change in Resonant Frequency of the Induc tively Tuned Shunt Switch in the Down-State for Varying Lengths of the High Impedance Line The shunt switches are simulated in Agilent’s ADS Momentum which is a 2.5D electromagnetic simulator. Figure 5.5 and Figure 5.6 show the simulated return loss (S11) and insertion loss (S21) in the non-actuated state and act uated state, respectively, for both designs.

PAGE 92

76 0 10 20 30 40 50 60 70Frequency (GHz) -50 -40 -30 -20 -10 0S 1 1 ( d B ) -2.5 -2 -1.5 -1 -0.5 0S 2 1 ( d B ) S11 (Design 1) S11 (Design 2) S21 (Design 1) S21 (Design 2) Figure 5.5 – Simulated S11 and S21 for the Basic (Design 1) and Inductively Tuned (Design 2) Capacitive Switch in the Up-State 0 10 20 30 40 50 60 70Frequency (GHz) -20 -15 -10 -5 0S 1 1 ( d B ) -50 -40 -30 -20 -10 0S 2 1 ( d B ) S11 (Design 1) S11 (Design 2) S21 (Design 1) S21 (Design 2) Figure 5.6 – Simulated S11 and S21 for the Basic (Design 1) and Inductively Tuned (Design 2) Capacitive Switch in the Down-State From the simulated S21 response we can observe a sh ift in resonant frequency between the two designs. Although each design has nearly equal down-state capacitance (~ 1.28 pF), the inductively-tuned design ha s a much higher i nductance value. The

PAGE 93

77 modeled inductance value for design 1 and de sign 2 in the actuated state is 11.9 pH and 38.7 pH, respectively. Figure 5.7 shows the comparison of the simulated and modeled S21 in the actuated state for both the designs. Tabulated result s of the lumped element components from the equivalent circuit will be pr esented in the measurements s ection. As stated earlier, the high impedance T.L has little effect on th e up-state performance of the switch. 0 10 20 30 40 50 60 70Frequency [GHz] -60 -50 -40 -30 -20 -10 0D o w n S t a t e S 2 1 [ d B ] Design 2 (Sim) Design 2 (Mod) Design 1 (Sim) Design 1 (Mod) Figure 5.7 – Comparison of Simulated and Modeled S21 in the Down-State 5.3 Material Characterization and Fabrication The switches are fabricated on high resistivity silicon wafers ( r=11.8, > 2000 ohm-cm) that are cleaned using a standard RCA process. The fabrication steps as shown in Figure 5.8 are as follows: Molybdenum, which is a carbide formi ng material, is used as the bottom metal. Molybdenum is deposited by RF sputtering to a thickness of 0.7 m. An intrinsic NCD film is grown to a thickness of 5000 A0. Prior to the growth the substrate is ultrasonically seeded in a nano-diamond pow der dispersed in methanol. This film is later etched in O2:CF4 plasma using titanium as the etch mask. The typical etch rate for diamond is 300 A0/minute.

PAGE 94

78 CPW metal lines are fabricated using a lift-off process. The Cr/Au metal lines are deposited using a thermal evapor ator to a total thickness of 1 m. PMMA is used as the sacrificial layer and spun on to a thickness of 1.5 m. The moveable membrane is gold el ectroplated to a thickness of 1.3 m. Prior to plating, a seed layer of Cr/Au is deposited to a thickness of 1600 A0. The sacrificial layer is removed in 1165 solution and the MEMS structures are released using a cri tical point dryer. In this work, the NCD film is grown us ing the MPECVD technique. Prior to this, attempts were made to deposit the films th rough the hot filament technique. Apparently the very high temperature near the filaments of the reactor (~ 12000C) caused growth defects and this resulted in the peeling of the underlying carbide forming layer. Both tungsten (W) and molybdenum (Mo) were used and the peeling problem was noticed in both the cases. In the MPECVD technique NCD growth is good in both W and Mo. Unlike other dielectric films, the roughness of NCD on the under metal has a significant effect on the performance of the switch and in turn the capacitance upon actuation. Table 5.1 shows the comparison of the roughness of the under metal before and after the dielectric deposition. For clar ity the roughness of NCD films are compared to that of nitride films grown in the PECVD technique. Table 5.1 – Comparison of Roughness of Metal + Dielectric Stack Before and After Deposition Metal + Dielectric Combination Roughness before Dielectric (nm) Roughness after Dielectric (nm) Moly + NCD 1.5 9.4 Moly + Si3N4 1.5 4.8 Tungsten + NCD 0.9 8.7 Tungten + Si3N4 0.9 4.1

PAGE 95

79 (A) (B) (C)(D) (E) (F)Deposit Under Metal (Mo/W) Intrinsic NCD is grown and patterned in RIE CPW lines are evaporated with Cr/Au Sacrifical layer (PMMA) is spun to a thickness of 1.5 m Shunt beam is gold electroplated to 1.3 m Sacrificial layer is removed in 1165 and released in CPD Figure 5.8 Fabrication Procedur e of the NCD Capacitive Switch Figure 5.9(a) and Figure 5.9 (b) show mi crophotographs of the fabricated shunt switch. As seen in Figure 5.9 (b ) there are holes in the mova ble membrane. In addition to permitting the removal of the sacrificial layer during the release process, the holes allow faster operation of the switch by reducing the air damping. The holes are 10 m in diameter and spaced 10 m apart in a triangular lattice. The holes do contribute not change the capacitance in the up-state, but in the down-stat e they do decrease the total capacitance. In this work this effect is not taken into consideration.

PAGE 96

80 Bottom Electrode NCD Dielectric 1.5 m Shunt Beam Bottom Electrode NCD Dielectric 1.5 m Shunt Beam (a) (b) Figure 5.9 Microphotograph of the Fabricated NCD Shunt Switch (a) Without Holes, (b) With Holes 5.4 Small Signal Measurements and Analysis The switches are tested in the frequenc y range of 1-65 GHZ using an Anritsu Lightning VNA and 150 m pitch probes on a Karl Suss semi automatic probe station. A standard center of thru TRL ca libration routine is performed prior to the measurements. A Picosecond Pulse Labs bias-tee is used to protect the VNA ports from DC current and a Keithley 2400 source meter was used for th e actuation of the shunt beams. The beams actuated at 22 volts draw ing a current of 5-6 A. Figure 5.10 and Figure 5.11 show the comparison of the simulated (from ADS mo mentum) and measured return loss and insertion loss of a switch in the up-state and down-state, respectively. The up-state insertion loss of 1.1 dB at ~50 GHz is pr edominantly due to mismatch loss cause by beam imperfections, as verified by the high S11 value; the loss factor calculated from the measured S-parameters is below 0.15 over the entire frequency range. An equivalent CLR circuit (explained in sectio n 5.2) is used to extract the lumped element values in both the states from which the capacitance wa s found to be 50-55 fF in the up-state, and 0.74 pF in the down-state. The extracted i nductance and resistance are 11.8 pH and 0.83 respectively. The switch was designed to yield a down-state cap acitance of 1.26 pF and the discrepancy in the measured value can be attributed to roughness in the NCD dielectric film and possibly reduction in th e actual contact area due to imperfect beam flatness, which reduces the net capaci tance upon actuation. Table 5.2 shows the

PAGE 97

81 comparison of the extracted lumped element values from the simulation and measurement. In addition to the roughness, the difference in overlap area during measurements is another reason for the difference in capacitance values. 0 10 20 30 40 50 60 70Frequency (GHz) -50 -40 -30 -20 -10 0S 1 1 ( d B ) -3 -2.5 -2 -1.5 -1 -0.5 0S 2 1 ( d B ) Measured S11 Simulated S11 Measured S21 Simulated S21 Figure 5.10 – Comparison of Simulated and Measured S11 and S21 in the Up-State of the NCD Capacitive Shunt Switch 0 10 20 30 40 50 60 70Frequency (GHz) -20 -15 -10 -5 0S 1 1 ( d B ) -60 -50 -40 -30 -20 -10 0S 2 1 ( d B ) Measured S11 Simulated S11 Simulated S21 Measured S21 Figure 5.11 – Comparison of Simulated and Measured S11 and S21 in the Down-State of the NCD Capacitive Shunt Switch

PAGE 98

82 Table 5.2 – Comparison of the Lumped Elemen t Values of the Capacitive Switch from Simulations and Measurement Design Name L (pH) R (ohms) Cup (fF) Cdown (pF) 10.2 0.06 50 1.26 Design 1 (Simulated) Design 1 (Measured) 11.8 0.8 55 0.75 40.8 0.06 50 1.26 Design 2 (Simulated) Design 2 (Measured) 41.2 0.8 55 0.78 5.5 Corona Kelvin Measurements The Corona Kelvin measurement (CKM) technique [97] is a non-destructive dielectric characteriza tion method that does not require a top electrode and can therefore be used as an in-line process step. In th is method, ions are gene rated at atmospheric pressure from air by putting a hi gh potential on a short tip. Th ese ions diffuse through air and are deposited on the substrate under test. The non-contact voltage is measured with a vibrating Kelvin probe rela tive to the grounded bottom el ectrode. In addition to the measured potential, this technique also co mputes the amount of ions deposited. Figure 5.12 shows the setup for CKM measurement. CK M is a well established technique in the CMOS industry and is accurate to 0.1mV under low voltage conditions (20 volts). Figure 5.13 and Figure 5.14 show the CKM-measured vo ltage decay for the nitride and diamond films, respectively, taken at three different s ites across the wafers. It is evident from the data that both nitride and diamond are leak y, but the rate of voltage decay for the diamond film is considerably faster. The CKM method yields a voltage drop from 35 volts to 10 volts in ~2 minutes for the nitride films. The diamond films, due to the leaky nature have a voltage drop from 8 volts to te nths of a volt in approximately 10 seconds.

PAGE 99

83 + + + Corona Dose Generation Kelvin Probe Electronics Motorized Arm Personal Computer Moving Chuck Wafer Dielectric Film Vibrating Kelvin Probe Corona Ions Light Diode Discharge Electrode Figure 5.12 Typical Setup for Corona Kelvin Measurement (CKM) 0 20 40 60 80 100Time (seconds) 5 10 15 20 25 30 35V o l t a g e ( v o l t s ) Site 1 Site 2 Site 3 Figure 5.13 Voltage Decay of Nitride Film at Three Different Sites of Samples Through CKM Technique

PAGE 100

84 0 10 20 30 40 50 60 70 80 90 100Time (sec) 0 0.05 0.1 0.15 0.2V o l t a g e ( v o l t s ) 7 7.05 7.1 7.15V o l t a g e ( v o l t s ) Site 1 Site 2 Site 3 Figure 5.14 – Voltage Decay of NCD Film at Three Different Sites of Samples Through CKM Technique 5.6 Stressed I-V and C-V Measurements – 1st Generation In addition to CKM measurements, standard I-V and C-V tests are also performed to establish the non-charging na ture of NCD films. For this test, metal-insulator-metal (MIM) and MEM structures using silicon nitride and NCD films are fabricated. The capacitor test structures were fabricated on high resistivity silic on substrates with molybdenum as the bottom electrode and the dielectric films were etched into small islands using standard etching recipes. Th e MEM devices did not work fine due to processing irregularities and hence the discus sion is focused on the MIM structures. The MIM structures have a 0.7 m thick bottom electrode made of molybdenum. The insulator material was grown to a thickness of 0.5 m and the top metal was electroplated to a thickness of 1 m. The overall capacitance area of the MIM structure was 90 x 90 m. I-V measurements are carried out usin g a 4140B picoammeter. Measurements are performed at dv/dt values of 0.25 volts/sec, 0.5 volts/sec and 1 volt/sec. Figure 5.15 and Figure 5.16 show the leakage current for di fferent dv/dt values for the nitride and diamond film, respectively.

PAGE 101

85 5 10 15 20 25 30 35 40Voltage [volts] 0 0.1 0.2 0.3 0.4C u r r e n t [ n A ] 0.25 dv/dt 0.5 dv/dt 1 dv/dt Figure 5.15 – Measured Leakage Current vs Volt age for Nitride Film with Different dv/dt Values 5 10 15 20 25 30 35 40Voltage [volts] 0 100 200 300 400 500C u r r e n t [ n A ] 0.25 dv/dt 0.5 dv/dt 1 dv/dt Figure 5.16 – Measured Leakage Current vs Vo ltage for NCD Film with Different dv/dt Values

PAGE 102

86 The leakage current (I) in the MIM cap acitors is dependent on the capacitance (C), instantaneous rate of vo ltage (dv/dt), applie d voltage (V), and co nductance (1/R) as given in equation 5.5 R V dt dv C I (5.5) Although both nitride and diamond films e xhibit a non-linear I-V behavior, the strong dependence on the dv/dt term for the n itride film allows the capacitance value (~3.4 pF) to be extracted. For the diamond films the 1/R term dominates due to the conductive path provided by the diamond grain boundaries in the cluster, leading to leakage current in the A range. In addition to basic IV measurements, leakage current testing under stress conditions is also performed. Figure 5.17 and Figure 5.18 show the stress induced leakage current (SILC) measurem ents for nitride films stressed at 20 volts and 40 volts, respectively. These tests are carried out different periods of time (20 mins, 2 hours and 5 hours). The tests perf ormed at 20 volts have little to no effect on the leakage current value. This is evident from the graph in Figure 5.17 wherein the leakage current is almost constant after the SILC testing. 5 10 15 20 25 30 35 40Voltage [volts] 0.12 0.16 0.2 0.24 0.28C u r r e n t [ n A ] No Stress 20 Minutres 2 Hours 5 Hours Figure 5.17 – Stress Induced Leakage Current (S ILC) for Nitride Films Stressed at 20 Volts for Different Periods of Time

PAGE 103

87 In Figure 5.18, the capacitors stressed at 40 volts show an increase in current after the SILC tests. This increase in current can be attributed to trapped charges which lower the effective stress voltage of the dielectric layer. Increase in current can be a result of barrier lowering due to electron de-trapping or the effect of SILC [98]. Unlike barrier lowering, the SILC process creates a permanen t defect on the dielectric layer which is irreversible. Stress testing creat es a high resistance breakdown s pot in the dielectric layer and is very hard to detect during in-situ measurements. The initial defect created during the SILC testing gets propagated overtime resulting in a dielectric breakdown. 5 10 15 20 25 30 35 40Voltage [volts] 0 0.4 0.8 1.2C u r r e n t [ n A ] No Stress 20 Mins 2 hours 5 hours Figure 5.18 – Stress Induced Leakage Current (S ILC) for Nitride Films Stressed at 40 Volts for Different Periods of Time Figure 5.19 and Figure 5.20 s how the SILC results for diamond films which are stressed at 20 volts and 40 volts respectively. It is evident from the graphs that stress induced charge does not affect or increase the leakage current before and after the tests. Unlike nitride films where the charges create defects, the leaky natu re of NCD films does not allow the charge to be stored on the surface or in the bulk of the material. NCD capacitors were also te sted under stress conditions of 60 volts but this re sulted in the increase of leakage current to milliamps. Unlike in silicon nitride and oxide where milliamps of leakage current correspond to dielectric break down, the dielectric

PAGE 104

88 properties of NCD films are not altered. C-V measurements ar e done to confirm this that the NCD films retain their dielectric propertie s. The results of the C-V measurements are shown in Figure 5.21. 5 10 15 20 25 30 35 40Voltage [volts] 0 100 200 300 400 500C u r r e n t [ n A ] No Stresss 20 Minutes 2 Hours 5 hours Figure 5.19 – Stress Induced Leakage Current (SILC) for NCD Films Stressed at 20 Volts for Different Periods of Time 5 10 15 20 25 30 35 40Voltage [volts] 0 100 200 300 400 500C u r r e n t [ n A ] No Stress 20 Minutes 2 Hours 5 hours Figure 5.20 – Stress Induced Leakage Current (SILC) for NCD Films Stressed at 40 Volts for Different Periods of Time

PAGE 105

89 0 5 10 15 20 25 30Voltage [volts] 0.4 0.5 0.6 0.7 0.8 0.9C a p a c i t a n c e [ p F ] 20 Volts 40 Volts 60 Volts Figure 5.21 – Capacitance Measurement of th e NCD Capacitors Before and After Stress Tests 5.7 Stressed I-V and C-V Measurements – 2nd Generation In the second generation switches, tungs ten is used as the bottom electrode instead of molybdenum. Tungsten is sputtered wi th Ti as a seed layer. The thickness of the heterostructure is deposited close to 9000 A0. Silicon nitride and NCD films are once again deposited through CVD technique followe d by a top electrode of gold. In this version, the MEM structures are also fabri cated. For the MEM structures, after patterning the dielectric layer, PMMA, which is used as the sacrificial layer is patterned to a thickness of 1 micron. The top electrodes ar e gold electroplated to a thickness of 1 m. For this set of measurements dielectric layers (both Si3N4 and NCD) of different thicknesses are used. This is done for the MIM and MEM structures to understand the effect of thickness on the I-V and C-V meas urements under stress conditions. Although the growth conditions were closely m onitored a small difference of 100-200 A0 in the final thickness is seen. Although this difference is crucial in many tran sistor applications, for this work the effect due to the difference in thickness is ignored. Figure 5.22 and Figure 5.23 show I-V measurements for the MIM nitride and NCD capacitors. The dielectric thickness of the nitride ca pacitors is 1400 A0 and for

PAGE 106

90 NCD is 1500 A0. The area of the capacito rs being tested is 90 m X 90 m. In order to avoid dielectric break down the capacitors ar e tested by gradually increase the voltage. For the nitride films, the l eakage current values are low initially up to 15 volts beyond which the current increases and dielectric break down occurs 30 volts. In the second generation tests, all stress measuremen ts are performed for duration of 1 hour. 5 10 15 20 25 30Voltage [volts] 0 100 200 300 400 500C u r r e n t [ n A ] No Stress 20 V 30 V Figure 5.22 – Stress Induced Leakage Current (SILC) for Nitride MIM Capacitor with Dielectric Thickness of 1400 A0 0 10 20 30 40Voltage [volts] 0 10 20 30C u r r e n t [ m A ] No Stress Figure 5.23 – Stress Induced Leakage Current (SILC) for NCD MIM Capacitor with Dielectric Thickness of 1500 A0

PAGE 107

91 The growth time for the diamond film is too short (~ 1 hour) causing many defects, openings and voids in the film which result in a short upon application of an electric field. As the measured leakage cu rrent is in the order of mA in the first measurement, no stress tests are performed fo r the NCD capacitors. This claim can be confirmed by the very high capacitance value (~ 400 pF) as shown in Figure 5.24. Surface defects in the diam ond film lead to a through conductive path between the bottom and top electrode. In this work undope d NCD films exhibits dielectric behavior for thicknesses above 0.3 m. 0 5 10 15 20 25 30Voltage [Volts] 300 350 400 450 500C a p a c i t a n c e [ p F ] Device 1 Device 2 Figure 5.24 – Capacitance Measurement of 1500 A0 NCD MIM Capacitor MIM capacitors with 0.5 m thick dielectric layers are used for the I-V stress tests. These tests, similar to the first generation switches, are carried out for both the nitride and diamond capacitors. Figure 5.25 an d Figure 5.26 show the I-V response for the nitride and NCD capacitors, respectively. Th e MIM structures are stressed at different voltages for more than 1 hour. Th e nitride based capacitors ex hibit an increase in leakage current after the stress tests. At higher fiel ds (> 65 volts), the nitride capacitors are dominated by the Fowler-Nordheim [99] tunneli ng effect. This effect is not seen in the NCD based capacitors, wherein the higher con ductivity in the diamond film dominates the increase in current in comparison to the nitride films.

PAGE 108

92 0 10 20 30 40 50 60 70 80 90 100Voltage [volts] 0 100 200 300 400 500 600C u r r e n t [ n A ] No Stress 20 Volts 40 Volts 60 Volts Figure 5.25 – Stress Induced Leakage Current (SILC) for Nitride MIM Capacitor with Dielectric Thickness of 5000 A0 0 10 20 30 40 50 60 70 80 90 100Voltage [volts] 0 1000 2000 3000 4000 5000C u r r e n t [ n A ] No Stress 20 Volts 40 Volts 60 Volts Figure 5.26 – Stress Induced Leakage Current (SILC) for NCD MIM Capacitor with Dielectric Thickness of 5000 A0 For comparison purposes, Figure 5.27 show s the leakage current measurements for the nitride and NCD capacitors in the log scale. The graph shows the leakage current

PAGE 109

93 value for the capacitors under no stress and 60 volts of stress. It is evident from the results, that there is a shift observed in the nitride capacitors where as the value does not alter in the case of NCD capacitors. 0 20 40 60 80 100Voltage [volts] 0.1 1 10 100 1000 10000C u r r e n t [ n A ] No Stress (Nitride) 60 Volts (Nitride) No Stress (NCD) 60 Volts (NCD) Figure 5.27 – Comparison of St ress Induced Leakage Current (SILC) in Log Scale for 5000 A0 NCD and Nitride Capacitors In comparison to the first generation measurements, the new set of SILC tests are carried for different thickness for both the ni tride and NCD films and they show a better performance in terms of low leakage curren t for the films. From the measurements, tungsten is better suited as the under metal in place of molybdenum. The SILC tests for NCD based MEM capacitors are carried out by ob serving the shift in leakage current and actuation voltage, both as a function of time The MEM capacitors actuate at ~ 35 volts. Figure 5.28 shows the I-V res ponse of the MEM capacitor in the non actuated state. There is no shift in the current response upto 30 volts which equates to no change in actuation voltage, after 2 hours of testing. When the membrane actuates and falls on the dielectric layer the current value in creases from nA in the up-state to A in the downstate as shown in Figure 5.29. In comparison to the popular dielectric (nitride or oxide) based capacitive switches, repeated actuation of the switches will lead to a shift in the actuation voltage. This problem al ong with the charge accumulation in the dielectric layer

PAGE 110

94 will eventually lead to the failu re of the switches. Further progress is underway to test the switches for a longer time to check its reliability. 5 10 15 20 25 30Voltage [volts] 0.1 0.15 0.2 0.25C u r r e n t [ n A ] 0 Minutes 30 Minutes 60 Minutes 90 Minutes 120 Minutes Figure 5.28 – Measured I-V Response for th e NCD MEM Capacitor in the Up-State 30 40 50 60 70 80 90Voltage [volts] 0 7 14 21 28 35C u r r e n t [A ] 0 Minutes 30 Minutes 60 Minutes 90 Minutes 120 Minutes Figure 5.29 – Measured I-V Response for th e NCD MEM Capacitor in the Down-State

PAGE 111

955.8 Summary Capacitive shunt switch es using undoped NCD film have been designed, fabricated and tested till 65 GHz. The switc hes are electrostatically actuated and the measurements show an insertion loss of 1.1 dB and an isolation of 30 dB at ~50 GHz, respectively. The measured capacitance in the down-state of the beam is 0.75pF in comparison to 1.26 pF from simulation resu lts obtained from ADS The difference in capacitance can be attrib uted to the roughness of the dielec tric layer and the difference in overlap area when the beam is actuated. Ch arging studies are performed for the nitride and NCD films using CKM metrology and stressed I-V measurements. SILC measurements are carried out for the capac itors using both Mo and W as the bottom electrodes for various dielectric thicknesses. The leaky nature of the NCD films is proven through these measurements. A lthough static power consumption is an issue with NCD capacitors, they can be used in applications which demand little to no degradation in performance and allow microwatts of power consumption.

PAGE 112

96 Chapter 6 Summary and Recommendations 6.1 Summary The work presented in this disserta tion was focused on developing RF-MEMS and microwave devices based on NCD thin f ilms. The research was multi-disciplinary and included thin film development and an alysis (material science), mechanical characterization of NCD to rea lize compressively stressed f ilms (mechanical engineering) and developing high power RF-MEMS devices based on NCD actuators and capacitive shunt switches with NCD as a dielectric (m icrowave engineering). Furthermore, this research presented the life cycle of a NCD based RF-MEMS starting from growth, proceeding to the design, fabrication and integration of the device and finishing with the measurement and modeling of the fabricated device. Growth recipes for developing intr insic and doped NCD films along with different seeding techniques using microwav e plasma and hot filament CVD techniques have been presented in detail. Mechanical characterization techniques to measure the Young’s modulus, first order resonant frequenc y and intrinsic stress of the grown NCD films have been demonstrated. A Young’s modulus of 1020 GPa wa s measured using a non-destructive technique. The films were in tentionally grown with built in compressive stress and the measured values were in the range of 140560 MPa. Measured results for the effect of different growth parameters that include temperature, pressure and gas ratio on the intrinsic stress developed in the films have been presented. NCD based RF-MEMS actuators were de signed for high power applications wherein diamond based devices were designed, fabr icated and tested to generate switches and inductors in CPW and mi crostrip topologies. The ther mally actuated diamond bridges operated in a bi-stable mode and the design to generate this bi-stable operation was studied using mathematical formulae fr om beam analysis. Mechanical and

PAGE 113

97 electromagnetic simulations were performed using ANSYSTM and ADSTM, respectively to understand mechanical and electrical pr operties of the diamond actuators. Small signal S-parameter measurement was performed in the frequency range of 1-40 GHz on the integrated switches. Measured results show a return loss and insert ion loss of 20 dB and 0.2dB, respectively, at 20GHz. Furthermore, tunable inductors were designed and fabricated in a CPW topology. Measurements we re carried out in the frequency range of 1-30GHz and the measured results show an inductance ratio of 2.2 at 30 GHz. Large signal (high power) measurements were pe rformed on both mechanically-actuated diamond switches in the power range of 24-47 dB m and on electrically actuated switches (that use a dual actuation scheme) in the ra nge of 24-40 dBm. The measurements show an insertion loss of 0.2-0.3 dB in the entire pow er spectrum. The switches did not show any mechanical degradation and depreciation in the RF performance for the limited period in which the tests were performed. NCD films were also used as insulators in capacitive shunt switches to address reliability issues due to charging. Prior to the design, material characterization studies were performed to grow intrinsic NCD films on metal thin films. A carbide forming layer was necessary to facilitate NCD growth. Diam ond films were successfully grown on Mo and W thin films with mode rate roughness (~ 8-9 nm). Th e electrostatically actuated switches were simulated in Agilent’s ADS a nd an equivalent circ uit was generated to extract the lumped element valu es of the switch model. The switches were fabricated on a high resistive silicon substrate using standa rd surface micromachining techniques. Small signal measurements were performed in the frequency range of 165 GHz with insertion loss of 1.1 dB and an isolation of 30 dB at ~50 GHz, respectively. Preliminary charging studies were performed using the CKM and I-V techniques for nitride and NCD films and both methods demonstrate the relatively leak y nature of the diamond dielectric film. Despite the finite conductivity of the NCD film, however, th e measured loss of the switch is dominated by mismatch loss due to im perfections in the MEM beam. These NCD based switches can be used in applications where microwatt of power consumption is permitted without any depreciation in performa nce for a longer duration of time. This

PAGE 114

98 balance between microwave loss and DC c onductivity may prove to be an effective method to mitigate dielectric charging. 6.2 Recommendations for Future Work In order to realize switches with better performance, the shunt switches can be fabricated with using tungste n as the bottom electrode. This could possibly account for the difference in the down-state capacita nce value during actuati on. Furthermore, a thorough theoretical analys is can be carried out to better understand the different regions in the leakage current measurements. Reli ability measurements for both the diamond actuators at higher power levels and the cap acitive switches should be performed for a longer period of time. The fabricated NCD actuators and NCD capacitive switches could be tested in switching speed measurements Figure 6.1 shows the typical set up which will be used for the switching speed measurements. The 10 GHz signal source will be modulated and the diode detector converts the RF signal into DC. An oscilloscope will be used to compare the modulated and driving waveform to measure the switching speed. Figure 6.1 – Block Diagram for Switching Speed Measurement Temperature-based studies on the di amond actuators will highlight the performance of the switch which includes microwave measurements and mechanical stability under low and elevated temper ature conditions. The outstanding thermal

PAGE 115

99 conductivity of NCD thin films can be useful in their oper ation. As an improvement to the current design, the contacts pads which ar e currently of platinum can be substituted with doped diamond; the low temperature UNCD films introduced by Argonne National Laboratories can be used. After electroplatin g copper, a thin film Mo or W can be deposited over which doped UNCD films can be grown. Repeated actuation under hold and cold switching conditions will have lit tle to no effect on the contact due to the outstanding mechanical and material characteristics of diamond films. NCD based distributed MEMS transmissi on line (DMTL) phase shifters can be designed to operate in the K band. Diamond actuators designed and tested in this work will be used to generate th ese phase shifters. Figure 6.2 s hows the layout of the phase shifter. In this design 10-12 switches will be fabricated on a single silicon frame and actuated using the dual (thermal + electrostatic) actuati on scheme. The first generation 9 section phase shifter that were successfu lly fabricated. But due to the excessive compressive stress in each individual switches resulted in the switches breaking from the silicon frame. In designing such NCD based DM TL phase shifters, it is very important to understand and control the stress build up during the diamond growth process. Figure 6.2 Layout of the Mult i-Bit DMTL NCD Phase Shifter Silicon Frame Contact Pads Diamond Actuators Bi-metal (Cu) Pads Copper Pads

PAGE 116

100 Ferroelectric films are widely used in th e microwave community as thin films to generate tunable phase shifters, varactors, ci rculators etc. But their performance at high frequencies is limited due to high inser tion loss and single degree of freedom in capacitive tuning. Combining UNCD and ferroel ectric films will he lp overcome these limitations. Barium strontium titanate (BST) w ill be the ferroelectric material used for investigation. BST films need to be annealed during the deposition process or post annealed after deposition. UNCD films which offer outstanding thermal properties will be ideal for depositing BST. Sputtering, PLD and sol-gel techniques could be used for this investigation. Characterization tools such as SEM, AFM, XRD and Raman spectroscopy will be helpful used to investigat e these heterostructures. This proposed thin film research will be the first of its ki nd to merge ferroelectric and diamond films. RF-MEMS/NEMS devices based on UNCD and BST heterostructures can be developed. Tunable MEMS capacitors with BS T and UNCD will provide great range in tunability and also controlling charging issues and parasitic losses thereby increasing the overall Q at microwave frequencies. High pow er terminations and attenuators based on doped UNCD films can be developed. These w ill be a good candidate to replace the high power attenuators which are made of bulk substrates like aluminum nitride and beryllium oxide. The tunable switches could be furt her integrated with the UNCD high power attenuators to realize phase and amplitude controlled circuits. Finally true time delay (TTD) phase shifters integrating BST and UN CD films wherein the dual mode actuation scheme will be used to facilitate a multi bit 10-12 section UNCD actuator will be integrated to a host substrate with th e heterostructure dielectric layer.

PAGE 117

101 References 1. Hector J.DeLos Santos, “Introduction to microelectromechanical microwave Systems”, Boston Artech House, 1999. 2. Varadan V.K, “RF MEMS and their applications”, John Wiley & Sons, Inc. 2003. 3. Larry E. Corey, “Radio fr equency micro-electro-mechan ical switches (RF MEMS) Improvement Program,” Pre-soli citation Conference, Oct 19 2001. 4. Gregory T.A.Kovacs, “Micromachined transducers sourcebook”, McGraw-Hill Publishers, 1998. 5. Source: http://newton.ex.ac.uk/research/qsystems/people/sque/ 6. S.Bhattacharyya et al., “Synthesis and char acterization of high ly-conducting nitrogendoped ultrananocrystalline diamond films” Applied Physics Letters, Vol.79, No.10, pp.1441-1443, 2001. 7. G.M.Rebeiz, “RF MEMSTheory, design and technology”, John Wiley and Sons Publication, 2003. 8. V.Milanovic et al., “Batch transfer integra tion of RF microrelays”, IEEE Microwave and Guided Wave Letters, Vol.10, No.8, pp.313-315, August 2000. 9. P.Blondy et al., “Packaged mm-wav e thermal MEMS switches”, 31st European Microwave Conference, pp.1-4, September 2001. 10. H.C.Lee et al., “ Piezoelectrically actuated RF MEMS DC contact switches with low voltage operation”, IEEE Microwave a nd Guided Wave Letters, Vol.15, No.4, pp.202-204, April 2005. 11. G.M. Rebeiz and J.B. Muldavin, “RF ME MS switches and switched circuits,” IEEE Microwave Magazine, Vol.2, No.4, pp.59-71, December 2001.

PAGE 118

102 12. B.Schauwecker et al., “Serial combinat ion of ohmic and capacitive RF MEMS switches for large broadband applications ”, IEEE Electronic Letters, Vo.40, No.1, January 2004. 13. C.L. Goldsmith et al., “Performance of low-loss RF MEMS capacitive switches,” IEEE Microwave and Guided Wave Letters, Vol.8, No.8, pp.269-271, August 1998. 14. H.A.C.Tilmans et al., “Wafer level packag ed RF-MEMS switches fabricated in a CMOS fab”, International Electron Devices Meeting, pp.41.4.1-41.4.4, December 2001. 15. Larry E. Corey, “Radio Frequency MicroElectro-Mechanical Switches (RF MEMS) Improvement Program,” Pre-soli citation Conference, October 2001. 16. Electronic Reference: http://en.wikipedia.org/wiki/Piezoelectric 17. Electronic Reference: http://en.w ikipedia.org/wiki/Lorentz_force 18. P.Schmid et al., “Diamond switch using new thermal actuation scheme”, Diamond and Related Materials, Vol .12, No.3-7, pp.418-421, March-July 2003. 19. D.Saias et al., “An above IC MEMS RF switch”, IEEE Journal of Solid State Circuits, Vol.38, No.12, pp.2318-2324, December 2003. 20. B.Lakshminarayanan et al., “High power high reliability sub-microsecond RF MEMS switched capacitors”, IEEE Internati onal Microwave Symposium, pp.1801-1804, June 2007. 21. T. Ketterl et al., “MEMS Series Switch w ith Nanometer Wide Gaps in Suspended Coplanar Waveguide Transmission Li nes”, IEEE International Microwave Symposium, pp.255-258, June 2006. 22. B.Lakshminarayanan et al., “High reli ability miniature RF MEMS switched capacitors”, IEEE Transactions on Microw ave Theory and Techniques, Vol.56, No.4, pp.971-981, April 2008. 23. D.Sparks et al., “Reliable Vacuum Packaging Using NanoGettersTM and Glass Frit Bonding”, Proceedings of SPIE, pp.5343-5346, January 2004. 24. Technological developments by Teravicta Technologies Inc. 25. Technological developments by XCOM Wireless.

PAGE 119

103 26. J.R.Reid et al., “Measurements of chargi ng in capacitive microelectromechanical switches,” Electronic Letters, Vol.38, No.24, pp.1544-1545, November 2002. 27. W.M.van Spengen et al.,“A comphrensive model to predict the charging and reliability of capacitive RF MEMS sw itches“, Journal of Micromechanical. Microenginerring, No.14, pp.514-521, Jan 2004. 28. X.Yuan et al., “Initial obs ervation and analysis of diel ectric charging effects on RF MEMS capacitive switches”, IEEE Inte rnational Microwave Symposium, pp.19431946, June 2005. 29. G.J.Papaioannou et al., “On the dielectric polarization effects on capacitive RF MEMS switches”, IEEE International Microwave Symposium, pp.761-764, June 2005. 30. R.W.Herfst et al., “Center-shift method for th e characterization of dielectric charging in RF MEMS capacitive switches“, IEEE Transactions on Semiconductor Manufacturing, Vol.21, No.2, pp.148-153, May 2008. 31. G.Papaioannou et al., “Temperature study of the dielectric polar ization effects of capacitive RF MEMS switches”, IEEE Tran sactions on Microwave Theory and Techniques, Vo.53, No.11, pp 3467-3473, November 2005. 32. H.San et al., “Using metal-insulator-semiconduc tor capacitor to inve stigate the charge accumulation in capacitive RF MEMS switche s”, 3rd IEEE International Conference on Nano/Micro Engineered and Molecular Systems, pp.1048-1052, 2008. 33. F.W.DelRio et al., “Atomic layer deposition of Al2O3/ZnO nano-scale films for gold RF MEMS”, IEEE International Micr owave Symposium, pp.1923-1926, June 2004. 34. J.R.Webster et al., “Performance of amorphous diamond RF MEMS capacitive switch“, IEEE Electronic Letters, Vol.40, No.1, pp.221-223, January 2004. 35. J.Chee et al., “DC-65 GHz characterization of nanocrystalline diamond leaky film for reliable RF MEMS switches”, 35th European Microwave Conference, pp.581-584, October 2005. 36. C.Bordas et al., “Carbon nanotube based dielectric for enhanced RF MEMS reliability”, IEEE International Mi crowave Symposium, pp.375-378, June 2007.

PAGE 120

104 37. B.Pillans et al., “Schottky barrier contact based RF MEMS switch”, IEEE Journal of Microelectromechanical Systems, Vol.17, No.6, pp.1439-1446, December 2008. 38. S.Melle et al., “Reliability modeling of capacitive RF ME MS”, IEEE Transactions on Microwave Theory and Technique s, Vol.53, No.11, pp.3482-3488, November 2005. 39. R.W.Herfst et al., “Identifying degradat ion mechanisms in RF MEMS capacitive switches”, IEEE MEMS Conf erence, pp.168-171, January 2008. 40. P.Czarnecki et al., “New insights into charging in capacitive RF MEMS switches”, 46th Annual International Reliability Physics Symposium, pp.496-505, April 2008. 41. D.Hyman et al., “GaAs-compatible surf ace micromachined RF MEMS switches”, IEEE Electron Letters, Vol.35, No.3, pp.224-226, 1999. 42. D.Peroulis et al., “RF MEMS switches with enhanced power-handling capabilities”, IEEE Transactions on Microwave Theory and Techniques, Vol.52, No.1, pp.59-68, January 2004. 43. B.Ducarouge et al., “Power capabilities of RF MEMS”, 24th International Conference on Microelectronics, Vol.1, pp.65-70, May 2004. 44. D.Hyman et al., “Contact physics of gold mi crocontacts for MEMS switches”, IEEE Transactions on Components and Packaging Technology, Vol.22, No.3, pp.357-364, September 1999. 45. J.Muldavin et al., “Power ha ndling and linear ity of MEM capacitive switches”, IEEE International Microwave Sym posium, Vol.3, pp.1915-1918, June 2003. 46. B.Pillans et al., “RF power handling of capacitive RF MEMS switches”, IEEE International Microwave Symposium, pp.329-332, June 2002. 47. J.D.Martinez et al., “Surface and bulk mi cromachined RF MEMS capacitive series switch for watt-range hot switching ope ration”, IEEE European Microwave Conference, pp.1237-1240, October 2007. 48. C.Goldsmith et al., “Update on RF MEMS sw itch reliability”, Technical Report 2009. 49. P. Bergonzo et al., “Improving diamond de tectorsA device case”, Diamond and Related Materials, Vol.1 6, No.4-7, pp.1038-1043, April-July 2007. 50. W.P.Kang et al., “Nanodiamond lateral VF EM technology for harsh environments”, IEEE Transactions on Nuclear Science, Vo.54, No.4, pp.1061-1065, August 2007.

PAGE 121

105 51. K.J.Gabriel et al., “In situ friction and wear measuremen ts in integrated polysilicon mechanisms”, Sensors and Actuators A, pp. 84-188, February 1990. 52. E Kohn et al. “RF MEMS based on diamond”, Workshop presentation, 35th European Microwave Conference, October 2005. 53. Electronic reference: http://newton.ex.ac.uk/research/qsystem s/people/sque/diamond/structure/ 54. I.Ahmed, “Deposition of textured diamond using hot-filament assisted chemical vapor deposition method”, Masters Thesis University of South Alabama, 1998. 55. M.Amaral et al., “Growth rate improvement s in the hot-filament CVD deposition of nanocrystalline diamond”, Diamond and Related Materials, Vol.15, No.11-12, pp.1822-1827, November 2006. 56. M.Miyake et al., “Characteristics of nano-crysta lline diamond films prepared in Ar/H2/CH4 microwave plasma”, Diamond and Related Materials, Vo.51, No.9, pp.4258-4261, March 2007. 57. X.Zheng et al., “Investigation on the etchi ng of thick diamond film and etching as a pretreatment for mechanical polishing” Diamond and Related Materials, Vol.16, No.8, pp.1500-1509, December 2006. 58. Y.K.Liu, et al., “Comparative study of nucleation processes for the growth of nanocrystalline diamond”, Diamond a nd Related Materials, Vol.15, No.2-3, pp.234238, February 2006. 59. J.C.Arnault et al., “Comparis on of classical and BEN nucl eation studied on thinned Si (111) samples: a HRTEM study”, Applied Surface Science, May 2003. 60. D. M. Gruen, “Nanocrystalline Diamond Films” Annual Review of Material Science, Vol.29, pp.211-259, 1999. 61. W. Zhu et al., “Effects of noble gases on diamond deposition from methane-hydrogen microwave plasmas”, Journal of Applied Physics, Vol.68, No.4, pp. 1489-1496, 1990. 62. M.Belmonte et al., “Wear resistant CVD diam ond tools for turning of sintered hard metals”, Diamond and Related Materi als, Vol.12, No.37, pp.738-743, March-July 2003.

PAGE 122

106 63. S.Turchetta et al., “CVD diamond insert in stone cutting”, Diamond and Related Materials, Vol.14, No.3-7, pp.641-645, March-July 2005. 64. A.Glaser et al., “Deposition of uniform and well adhesive diamond layers on planar tungsten copper substrates for heat spread ing applications”, Ma terials Science and Engineering: B, Vol.127, No.2-3, pp.186-192, February 2006. 65. D.M.Gruen et al., “Deposition and Charact erization of Nanocrystalline Diamond Films”, Journal of Vacuum Science and Technology A, Vol.12, No.4, pp.1491-1495, 1994. 66. D. Zhou et al., “Control of diamond film microstructure by Ar additions to CH4/H2 microwave plasmas”, Journal of Applied Physics, Vol.84, No.4, pp.1981-1989, 1998. 67. O. Auciello et al., “Materials science and fabrication processes for a new MEMS technology based on ultrananocrystalline di amond thin films”, Journal of Physics: Condensed Matter, Vol.16, No.16, pp.539-552, 2004. 68. M.Kubovic, et al., “Surface channel MESFETs on nanocrystalline diamond” Diamond and Related Materials, Vo.14, No.3-7, pp. 514-517, 2005. 69. N. S. Xu et al., “Study of field electron emission from nanocrystalline diamond thin films grown from a N2/CH4 microwave plasma”, Journal of Physics D: Applied Physics, Vol.33, No.3-4, pp.1421-1427, 2000. 70. U.A.Palnitkar et al., “Adhesion propert ies of nitrogen ion implanted ultrananocrystalline diamond films on silicon substrate”, Diamond and Related Materials, Vol.17, No.4-5, pp. 864-867, 2008. 71. R.Haubner et al., “Comparison of sulfur boron, nitrogen and phosphorus additions during low-pressure diamond deposition”, Di amond and Related Materials, Vol.14, No.3-7, pp.355-363, 2001. 72. S.Bhattacharyya et al., “Synthesis and char acterization of high ly-conducting nitrogendoped ultrananocrystalline diamond films” Applied Physics Letters, Vol.79, No.10, pp.1441-1443, 2001. 73. J.Zhang et al., “Characterization of bor on-doped microand nanocrystalline diamond films deposited by wafer-scale hot filament chemical vapor deposition for MEMS applications”, Diamond and Relate d Materials, Vol.17, No.1, pp.23-28, 2008.

PAGE 123

107 74. D.A.Long, “Raman Spectroscopy”, McGraw-Hill, 1977. 75. T. Lin et al., “Compositional mapping of the argon–methane–hydrogen system for polycrystalline to nanocrystalline diamond film growth in a hot -filament chemical vapor deposition system”, Applied P hysics Letters, Vol.77, No.17, pp.2692-2694, 2000. 76. Electronic reference: http://en.w ikipedia.org/wiki/Young%27s_modulus 77. O.Auciello et al., “Are diamonds a ME MS’s best friend?”, IEEE Microwave Magazine, Vol.8, No.6, pp.61-75, 2007. 78. B.P.van Drieenhuizen et al., “Compari son of techniques for measuring both compressive and tensile stress in thin films“, Sensors and Actuators A, Vol.38, pp.756-765, 1993. 79. M.G.Allen et al., “Microfabricated structur es for the insitu measurement of residual stress, Young’s modulus and ultimate strain of thin films”, Applied Physics Letters, Vol.51, pp.241-243, 1987. 80. H.Guckel et al., “Fine grained polysilicon films with built-in tensile strain”, IEEE Transaction on Electron De vices, Vol.35, No.6, pp.800-801, 1988. 81. J.Brank et al., “RF MEMS-based tunabl e filters”, International Journal RF Microwave CAE, Vol.11, pp. 276-284, September 2001. 82. N.S.Barker et al., “Distributed MEMS true -time delay phase shifters and wide band switches”, IEEE Transactions on Microwav e Theory and Techniques, Vol.46, No.11, pp.1881-1890, November 1998. 83. C.W.Baek et al., “2D mechanical beam steering antenna fabricated using MEMS technology”, IEEE MTT-S International Microwave Symposium, pp.211-214, June 2001. 84. J.Bryzek et al., “Marvelous MEMS”, IEEE Circuits and Devices Magazine, Vol.22, No.2, pp.8-28, March-April 2006. 85. A.Aleksov et al., “First diamond FET RF power measurement on diamond quasisubstrate”, 60th Device Research Conference, pp.181-182, June 2002. 86. P.Schmid et al., “Diamond switch using new thermal actuation scheme”, Diamond and Related Materials, Vol .12, No.3-7, pp.418-421, March-July 2003.

PAGE 124

108 87. Electronic Reference: http://en.w ikipedia.org/wiki/Hooke%27s_law 88. P.Benkart et al., “3D chip stack technol ogy using through-chip interconnects”, IEEE Design and Test of Comput ers, pp.512-518, November 2005. 89. S.Balachandran et al., “MEMS tunable plan ar inductors using DC contact switches”, 34th European Microwave Conference, Vol.12, pp.713-716, October 2004. 90. J. B. Muldavin, “Design and analysis of series and shunt MEMS switches,” Ph.D. dissertation, Department of Electrical Engineering and Computer Science, The University of Michigan at Ann Arbor, Ann Arbor, MI, 2001. 91. C.L.Goldsmith et al., “Lifetime characteriz ation of capacitive RF MEMS switches”, IEEE MTT-s International Microwave Symposium, Vol.1, pp.227-230, May 2001. 92. Z.Peng et al., “Dielectric Charging of RF MEMS capacitive switches under bipolar control-voltage waveforms”, IEEE In ternational Microwave Symposium, pp.18171820, June 2007. 93. X.Yuan et al., “ Modeling and characterization of di electric-charging effects in RF MEMS capacitive switches”, IEEE Inte rnational Microwave Symposium, pp.753756, June 2005. 94. M.Adamschik et al., “Diamond microw ave micro relay”, Diamond and Related Materials, Volume 12, No.3-7, pp.342-346, July 2003. 95. S.Balachandran et al., “Thermally actua ted nanocrystalline diamond micro-bridges for microwave and high-power RF appli cations”, IEEE International Microwave Symposium, pp.367-370, June 2007. 96. Agilent Advanced Design System Linecalc, Version 2006A. 97. P. Edelman et al., “Non-contact charge-volta ge method for dielectr ic characterization on small test areas of IC product wafers”, Materials Science in Semiconductor Processing”, Vol.9, No.1-3, pp.252-256, June 2006. 98. T. Wang et al., “Hot hole stress induced l eakage current (SILC) transient in tunnel oxides”, IEEE Electron Device Lette rs, Vol.19, No.11, pp.411-413, November 1998. 99. S.M.Sze, “Physics of semiconductor devices”, John Wiley and Sons, 1981.

PAGE 125

109 Appendices

PAGE 126

110 Appendix A Photolithography Procedures The photolithography steps for 3000 PY negative resist is: Clean sample with Acetone and Methanol. Blow dry with Nitrogen Spin coat the sample with Futu rrex 3000PY at 3000 rpm for 40 seconds Soft bake sample at 1550C for 60 seconds on a hot plate Expose the sample for 17 seconds in UV-light. A mask aligner is used for this purpose Hard bake at 1100C for 60 seconds. Develop the exposed photoresist in RD-6 for 25 seconds. Rinse the sample with DI water thoroughly and blow dry with Nitrogen. Typical thickness : 3.4 3.6 m Three kinds of ShipleyTM resists were used for vary ing thickness which includes SC 1813, 1818, and 1827. The photolithogra phy steps for SC 18xx resist is: Clean sample with Acetone and Methanol. Blow dry with Nitrogen Spin coat HMDS followed with the photoresist at 2500 rpm for 30 seconds. HMDS is used for adhesion Soft bake at 1100C for 70 seconds. Expose in the mask aligner for 25 seconds Develop in MF-319 for 80 seconds. Rins e thoroughly in DI water and blow dry with nitrogen. The photolithography steps for AZ 5214 resist is: Spin PMGI SF 11@ 2500rpm for 45 secs Bake it @ 180oC for 5 minutes in the hot plate Repeat this process twice to yiel d a thickness of close to 2 microns After cooling down, spin AZ 5214 @ 4000rpm for 45 secs

PAGE 127

111 Appendix A (Continued) Bake it @110oC for 90 seconds in the hot plate Expose using the mask aligner (Channel 2) for 6 seconds Develop the photo-resist in AZ 726 for 35 seconds Expose the PMGI in the Deep UV syst em for 50 minutes. This is done to expose the PMGI completely Develop in 101 developer for 4 minutes. The photolithography process for AZ 9260 resist is: Spin AZ 4533 thinner 1:50 (adhesion promoter) @ 6000rpm for 45 secs Bake it @ 180oC for 5 minutes in the hot plate After cooling down, spin AZ 9260 @ 6000rpm for 45 secs Bake it @110oC for 3 minutes in the hot plate Expose using the mask aligner (Channel 2) for 27 secs Develop the photo-resist in AZ 400K: H2O {1:4}for 3 minutes Rinse in water thoroughly for 2-3 minutes The photolithography process for maN 490 resist is: Spin maN @ 3000rpm for 45 secs Bake @ 100oC for 10 minutes Expose in CI2 for 45 secs Develop in mA-332 for 1 -1.5 minutes Oxygen plasma for 2 minutes

PAGE 128

112 Appendix B Fabrication of the Diamond Actuator and Host Substrate A two inch low resistive silicon substrate is thoroughl y cleaned with acetone and methanol and blow dried with nitrogen An intrinsic diamond layer is grown in th e big plasma system to a thickness of 1500 A0 Boron doped (p-type) nanocrystalline di amond is grown to a thickness of 1.4 m The silicon wafer with the nanocryst alline diamond is cleaned in Piranha solution (1 part of H2O2 + 2 parts of H2SO4) and then subjected to oxygen plasma for the diamond to be oxygen terminated. Oxygen terminated is necessary for good adhesion of metal or di electric film with nanocrystalline diamond. The recipe of the oxygen plasma is: o O2 gas : 6 sccm ; Pressure: 100 mT ; Power: 100 watts ; Time: 2-3 mins Silicon dioxide is grown using a PECVD process to a thickness of 300nm. This oxide layer is used as the mask ing layer for growing intrinsic diamond on top of the doped diamond layer A positive tone lithographic process is carried to pattern the silicon dioxide. After the lithography step, si licon dioxide is etched in CF4 plasma in the RIE system. The recipe for etching the oxide is: o CF4 gas: 45 sccm ; Pressure: 40 mT ; RF power: 600 watts; Time: 20 mins The photo-resist, which is used as the etch mask for silicon dioxide is removed by dipping in the above menti oned piranha solution for close to 2 minutes. The photo-resist is completely removed in this process and the wafer is thoroughly rinsed with DI water Intrinsic diamond is grown in the two samples using the hot filament CVD technique to a th ickness of 300 nm After the intrinsic diamond growth, remaining SiO2 is etched in the RIE for 30 minutes

PAGE 129

113 Appendix B (Continued) Backside of the wafer was etched in the RIE system. During intrinsic diamond growth, traces of diamond are deposite d on the backside. The wafers are treated in pure oxygen plasma to etch this diamond. The recipe is: o O2 gas: 45 sccm; Pressure: 40 mT; Power: 800 watts; Time: 30 mins The wafer is cleaned in acetone a nd IPA and treated in oxygen plasma A seed layer of Ti (60 A0) and Au (500 A0) is deposited using an e-beam evaporator. This seed layer is required for electroplating in the subsequent steps Lithography is done to open the areas of the bi-metal (copper) using a positive resist. The mask is a dark field mask, so 5214 resist is used for the lift off process Copper which is used for the thermal actuation scheme (bi-metal pads) is deposited using a thermal eva porator to a thickness of 2 m After the bi-metal (Cu) layer, the cent er contact pads along with the solder pads are to be electroplated. A si ngle spin lithography using AZ 9260 is carried out to yield a thickness of 8-9 m Copper is electroplated to a thickness of 5 m with a current setting of 3 mA and voltage setting of 0.25 volts. Electropl ating is slowly done to ensure no stress build up in the copper films The silicon wafer is cleaned with acet one and methanol and ashed in oxygen plasma Tin solder pads are included in the act uator portion in th e second generation switches. A two spin lithography using AZ 9260 is carried out to yield a thickness of 14 m Tin is electroplated to a thickness of 2-3 m with a voltage setting of 0.16 volts Photo-resist is removed by rinsing the samples thoroughly in 1 methyl 2 pyrilidone (1M2P). This is followed by ri nsing the wafers in acetone and isopropanol.

PAGE 130

114 Appendix B (Continued) Photo-resist is used as the mask to etch the seed layer. A two spin lithography using AZ 9260 is carried out to yield a thickness of 14 m. The seed layer is etched in the RIE system and the recipe is: o Ar: 45 sccm; Pressure: 40 mT; Power: 1200 watts; Time: 10 mins for etching gold o CF4: 45 sccm; Pressure: 40 mT; Power: 600 watts; Time: 30 mins for etching titanium Titanium (Ti) is used as the hard mask for etching diamond. Hence close to 380nm of titanium is deposited in the ion beam. This is a time consuming and slow process and also the samples are kept in two distinct positions in order to facilitate conformal coati ng of Ti throughout the wafer. This is a very crucial step, as non-conformal deposition of Ti will mean etching diamond from places where it should be protected. During deposition 5-7 minute break is given after every 30 minutes of deposition A single spin lithography using AZ-9260 is carried out to pattern the Ti layer. Ti is etched in the RIE using the following recipe: o Pressure: 40mT ; Gas: CF4 (45 sccm); Power:600 watts ; Time: 45mins After removing the photo-resist using acetone and methanol, Ti is used as the hard mask to etch diamond in the RIE system. The recipe for etching diamond is: o Pressure: 10 mT; Gas: O2 (8 sccm), Ar (3 sccm); Power: 1000 watts; Time:10 minutes o Pressure: 10mT; Gas: CF4 (10 sccm); Power:500 watts; Time:10 secs o These two steps are repeated for a total of 3 hrs to etch the diamond After removing the photo-resist used to pattern the Ti layer, a single spin lithography process using ma N 490 is carried to deposit platinum (Pt) on the copper contact pad. Ti (50 A0) and Pt (500 A0) are deposited using an e-beam evaporator

PAGE 131

115 Appendix B (Continued) To release the diamond beams, silicon is etched in a DRIE system using aluminum (Al) as a hard mask. Al is de posited using an e-beam evaporator to a thickness of 4000 A0. Prior to this, a single sp in lithography using AZ 5214 is done. The host substrates (alumina, aluminum nitride) are cleaned using acetone and methanol. A seed layer of Ti (100 A0) and Au (1200 A0) is deposited using an e-beam evaporator. A single spin li thography using AZ 9260 is carried out to open the transmission line areas. The transmission lines are gold elec troplated to a thickness of 3-4 m. The copper solder pads are electropl ated to a thickness of 2-3 m using the same procedure. The diced diamond actuators are integrat ed onto the host substrate using the SOLID process with a flip chip bonder.

PAGE 132

116 Appendix C Fabrication of the Capacitive Shunt Switch A one inch high resistive silicon substrate is thoroughl y cleaned with acetone and methanol and blow dried with nitrogen Molybdenum (Mo), which is the bottom el ectrode, is sputtere d to a thickness of 0.9 m. For the second generation switches tungsten is used as the bottom electrode. Tungsten (W) is de posited to a thickness of 1 m after depositing a seed layer of Ti which is 200 A0 thick. Both Mo and W are etched in an RIE system. Prior to etch process, a single spin 1827 lithography process is carried out The etch recipe for W and Mo is: o CF4 gas: 40 sccm; O2 gas: 5 sccm; Pressure: 100 mT; Power: 150 watts; Time: 15 mins Nanocrystalline diamond ((NCD) is grow n using the MPECVD technique to a thickness of 0.5 m Titanium is used as the hard mask to etch diamond. Titanium is deposited using lift-off technique to a thickness of 4000 A0. Prior to the deposition, a single 3000 PY lithography process is carr ied out. The etch recipe for NCD film is: o O2 gas: 50 sccm; CF4 gas: 5 sccm; Pressure : 50 mT; Power: 350 watts; Time: 30 mins Gold transmission lines are patterned using a sing le spin 3000 PY lithography process and lift-off technique to a thickness of 1 m PMMA which is used as the sacrificia l layer is spun to a thickness of 2 m. After spinning PMMA, titanium is pa tterned using 3000 PY lithography process and lift-off techni que to a thickness of 300 A0. This titanium layer is used as the hard mask for etching PMMA. The etch recipe for PMMA is: o O2 gas: 25 sccm; Pressure: 100 mT; Power: 100 watts; Time: 10 -15 mins

PAGE 133

117 Appendix C (Continued) Titanium is removed in 1:20 (HF:H2O) and cleaned thoroughly in DI water. A seed layer of Ti (200 A0) and Au (2000 A0) is deposited using an e-beam system. A single spin lithography process is ca rried out using 18 27 photo-resist to open up the beam areas The beam along with the pedestal is gol d electroplated to a thickness of 2 microns. In electroplating, the gold soluti on is heated to a temperature of 600C in a hot plate with a stir rer set to 2000 rpm. The cu rrent rating is set to 0.06 mA. The thickness of the platted gold is measured using a profilometer. The procedure is repeated if the thickness is not equal to the desired value. Prior to measuring the thickness, the samples ar e rinse using DI wa ter and blow dried with nitrogen. Then using a bright-field beam mask, the beam and pedestal layers are protected with a positive resist, pref erably 1827, using the positive tone process The seed layer is etched throughout the sample apart from the protected beam areas. Gold is etched at 250C at a rate of 25 A0/sec and chrome is etched at 400C at a rate of 40 A0/sec in gold and chrome etchant respectively Finally the samples are kept in b eaker of 1165 soluti on heated to 800C overnight and later rinsed in DI water and isop ropyl alcohol. Prior to placing the samples in the critic al point dryer (CCP), they are kept in a beaker of high purity methanol. Follo wing the procedure given, the samples are dried in the CCP, which is fill ed with the high purity methanol. If photo-resist is still sticking on to the sample, it can be removed by ashing the same in a plasma etcher. The typical settings are: o O2 gas: 40 sccm; Pressure: 250 mT; Power: 100 watts; Time: 2 mins

PAGE 134

About the Author Srinath Balachandran received his Bachelor of Engineering in 2000 from the University of Madras, India. He started his M.S program in fall 2001 and worked with Dr. Thomas Weller from the summer of 200 2 specializing in the area of RF-MEMS. Srinath completed his M.S degree in spring 2004. Upon completion, he enrolled into the Ph.D. program starting from fall 2004. His resear ch interests are in the area of RF-MEMS techniques for microwaves, application of micromachining for mm-wave circuits.