1 I NVESTIGATION OF MICROSCALE DIELECTRIC BARRIER DISCHARGE PLASMA DEVICES By JUSTIN C. ZITO A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 201 3
2 201 3 Justin C. Zito
3 To my family and Lauren
4 ACKNOWLEDGMENTS thank Dr. Arnold, my advisor and committee chair, for all of his guidance and supporting me throughout graduate school He has instilled in me a strong skill set and provided the training needed to contribute at a professional level in the field of engineering. I also thank my committee members : Dr s Subrata Roy, Mark Sheplak and Y.K. Yoon. I am proud to have a strong, prestigious committee that provide d knowledge and feedback that assisted me to carry out my dissertation work. Dr. Roy fo r providing a lab full of plasma diagnostic equipment and allowing me full access to these tools Ryan Durscher deserves a personal shout out as well, especially for his assistance within the fluidic domain. I thank the technicians of the NRF cleanroom. Namely, Mr. Bill Lewis, M r. David Hays and Mr. Al Ogden, who have helped me diagnose and correct several processing issues. I acknowledge my girlfriend Lauren for all her support especially when I need it most She is always there for me. I thank my family for their continuous support throughout my entire education I also thank my colleagues from both IMG and APRG.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 ABSTRACT ................................ ................................ ................................ ................... 15 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 17 1.1 Plasma Basics ................................ ................................ ................................ .. 18 1.2 Dielectric Barrier Discharge Devices ................................ ................................ 20 1.3 Application of DBD Devices for Flow Control ................................ .................... 21 1.4 Research Objectives ................................ ................................ ......................... 23 2 BACKGROUND AND MOTIVATION ................................ ................................ ...... 26 2.1 DBD Plasma Device Overview ................................ ................................ ......... 26 2.1.1 State of the Art ................................ ................................ ........................ 27 2.1.2 DBD Plasma Devices Scaling Trends ................................ .................. 29 2.1.3 DBD Plasma Devices Limitations ................................ ......................... 30 2.2 Potential Benefits of Microscale DBD Plasma Devices ................................ ..... 31 2.2.1 Increased per Volume Fluid Control Authority ................................ ......... 31 2.2. 2 Reduced Power Requirements ................................ ................................ 33 2.2.3 Improved Manufacturability ................................ ................................ ..... 35 2.3 Summary of Related Research on Microscale Plasma Devices ....................... 35 3 MICROSCALE DESIGN AND FABRICATION ................................ ........................ 40 3.1 Geometry and Materials ................................ ................................ .................... 40 3.2 Fabrication Process ................................ ................................ .......................... 42 3.3 Fabrication Challenges ................................ ................................ ..................... 48 4 EXPERIMENTAL CHARACTERIZATION METHODS ................................ ............ 51 4.1 Electrical Measurements ................................ ................................ ................... 52 4.1.1 Alternative Power Measurement Methods ................................ ............... 53 4.1.2 Considerations for Power Measurements ................................ ................ 55 4.2 Fluidic Measurements ................................ ................................ ....................... 60 4.2.1 Particle Image Velocimetry Measurements ................................ ............. 60 4.2.2 Pressure based Velocity Measurements ................................ ................. 64 4.3 Mechanical Measurements ................................ ................................ ............... 66 4.3.1 Direct Force Measurement ................................ ................................ ...... 67
6 18.104.22.168 Torsion balance ................................ ................................ ............. 68 4.3.2 Velocity based Force Measurement ................................ ........................ 74 22.214.171.124 Control volume analysis ................................ ................................ 74 126.96.36.199 Spatial body force estimation ................................ ......................... 79 188.8.131.52 Investigation of 2D flow ................................ ................................ .. 80 184.108.40.206 Investigation of fluid continuity ................................ ....................... 82 4.4 Thermal Measurements ................................ ................................ .................... 85 5 EXPERIMENTAL RESULTS: POLYMER DIELECTRICS ................................ ....... 89 5.1 Power Consumption ................................ ................................ .......................... 90 5.2 Ve locity Data ................................ ................................ ................................ ..... 92 5.3 Thrust and Plasma Force Results ................................ ................................ ..... 99 5.3.1 Direct Thrust Measurements ................................ ................................ ... 99 5.3.2 Velocity based Force Measurements ................................ .................... 102 5.4 Thermal Results ................................ ................................ .............................. 106 5.5 Failure Analysis ................................ ................................ .............................. 107 6 EXPERIMENTAL RESULTS: CERAMIC DIELECTRICS ................................ ...... 112 6.1 Power Consumption ................................ ................................ ........................ 112 6.2 Velocity Data ................................ ................................ ................................ ... 115 6.3 Thrust and Plasma Force Results ................................ ................................ ... 121 6.4 Thermal Results ................................ ................................ .............................. 126 7 CHARACTERIZATION METRICS AND COMPARISON WITH MACROSCALE DBD ACTUATORS ................................ ................................ ............................... 128 7.1 Materials, Geometries and Performance Data ................................ ................ 128 7.2 Thrust Metrics ................................ ................................ ................................ 130 7.3 Velocity Metrics ................................ ................................ ............................... 132 7.4 Actuator Efficiency ................................ ................................ .......................... 133 8 SUMMARY AND FUTURE WORK ................................ ................................ ....... 139 8. 1 Summary and Conclusions ................................ ................................ ............. 139 8.2 Research Contributions ................................ ................................ .................. 143 8.3 Future Work ................................ ................................ ................................ .... 144 APPENDIX A DATABASE OF VELOCITY CONTOURS ................................ ............................. 152 A.1 Devices with 10 m Polyimide Dielectric ................................ ........................ 152 A.2 Devices with 5 m Silicon Dioxide Dielectric ................................ .................. 155 A.3 Devices with 10 m Silicon Dioxide Dielectric ................................ ................ 158 B UNCERTAINTY ANALYSIS ................................ ................................ .................. 162
7 B.1 Power Measurement ................................ ................................ ...................... 162 B.2 Thrust Measurement ................................ ................................ ...................... 163 C LIST OF PUBLICATIONS ................................ ................................ ..................... 170 LIST OF REFERENCES ................................ ................................ ............................. 171 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 177
8 LIST OF TABLES Table page 4 1 Parameters for computing the force balance moment of inertia. ........................ 74 7 1 Device dimensions and material properties used in analysis of micro and macroscale DBD actuator performance metrics. ................................ .............. 129 7 2 List of materials used in performance comparisons and corresponding material densities. ................................ ................................ ............................. 129 7 3 Characterization data used in analysis of micro and macroscale DBD actuator performance metrics. ................................ ................................ .......... 130 7 4 Summary of various thrust metrics comparing micro and macroscale actuator performance. ................................ ................................ ................................ .... 131 7 5 Summary of various velocity metrics comparing micro and macroscale actuator performance. ................................ ................................ ...................... 132
9 LIST OF FIGURES Figure page 1 1 Three common types of plasma discharge. ................................ ........................ 20 1 2 DBD Configurations. ................................ ................................ ........................... 21 1 3 Schematic view of a single dielectric barrier discharge (DBD) plasma actuator. ................................ ................................ ................................ ............. 22 2 1 DBD demonstrating flow re attachment (Roth 2 003). ................................ ......... 27 2 2 Time averaged velocity flow field from a standard macroscale DBD plasma actuator (Durscher and Roy 2012). ................................ ................................ .... 28 2 3 Plasma synthetic jet actuator (Benard et al. 2008). ................................ ............ 29 2 4 Electric field, charge and force density plotted as a function of electrode separation distance. ................................ ................................ ........................... 32 2 5 law (Torres and Dhariwal 1999). ................................ ................................ ......... 34 2 6 Millimeter scale DBD actuator for flow control (Okochi et al. 2009). ................... 36 2 7 Investigation of breakdown voltages for microfabricated silicon electrodes (Ono et al. 2000). ................................ ................................ ................................ 37 2 8 A chemical vapor detector based on microdischarge excitation (Mitra and Gianchandani 2008). ................................ ................................ .......................... 37 2 9 Microfabricated inductively coupled plasma for emission spectroscopy ( Minayeva and Hopwood 2002) ................................ ................................ ......... 38 3 1 Microscale DBD actuator design. ................................ ................................ ....... 41 3 2 Cross section diagrams of device fabrication steps. ................................ ........... 43 3 3 First generation of fabricated microscale DBD plasma devices with polyimide dielectric material; showing linear and serpentine electrode geometries. ........... 46 3 4 Profilometer measurement of SiO 2 dielectric layer topology, scanned over bottom electrode contact pad ................................ ................................ ............ 47 3 5 Images showing several of the process steps involved in the fabrication of microscale DBD plasma actuators (process shown for polyimide dielectric). ..... 48
10 3 6 SEM image showing 2 to 4 m wide craters observed on a platinum anode due to electrode sputtering (Ono et al. 2000). ................................ .................... 49 3 7 Regions of delaminated silicon dioxide over 1 mm wide copper features. ......... 50 4 1 DBD device showing orientation of x,y,z coordinate axes. ................................ 52 4 2 Schematic of typical power supply and electrical measurement probes for DBD plasma generation. ................................ ................................ .................... 52 4 3 Voltage (blue) and current (green) waveforms plotted with time for different oscilloscope current resolution settings. ................................ ............................. 56 4 4 Average power as a function of current waveform resolution for two microscale DBD actuators. ................................ ................................ ................. 57 4 5 Average power as a function of sampling rate for a macroscale actuator. ......... 58 4 6 Influence of number of periods on average power value. ................................ ... 5 9 4 7 PIV data showing time averaged induced velocity flow field from a microscale DBD plasma actuator. ................................ ................................ ........................ 61 4 8 Schematic of experimental setup for velocity measurements indica ting the field of view used with reference to the location of the electrical contact points. ................................ ................................ ................................ ................. 63 4 9 Velocity convergence plots ove r 300 image pairs for the x component of velocity. ................................ ................................ ................................ .............. 64 4 10 Configuration for pressure measurements. ................................ ........................ 65 4 11 Schematic indicating the shear and body forces acting on an actuator and the imparted plasma force produced by a DBD actuator. ................................ ... 66 4 12 DBD thrust measurements (Thomas et al. 2009). ................................ .............. 67 4 13 Torsional force balance schematic. ................................ ................................ .... 69 4 14 Displacement measurements f rom torsional force balance plotted against time. ................................ ................................ ................................ .................... 71 4 15 Zoom extracting the balance spring constant. ................................ ................................ ..................... 71 4 16 Graphical representation of the superposition of the mass inertia for each compon ent of the torsional balance. ................................ ................................ ... 73
11 4 17 Control volume analysis for computing actuator thrust (Durscher and Roy 2012). ................................ ................................ ................................ ................. 75 4 18 Comparing horizontal thrust component from two measurement techniques (Durscher and Roy 2012). ................................ ................................ .................. 77 4 19 Schematic view of the control volume used for extraction of thrust data from PIV measurements. ................................ ................................ ............................ 78 4 20 Spatial plasma body force for a microscale DBD actuator. ................................ 80 4 21 Top view velocity contour, shown with device overlaid to indicate its location relative to the flow field. ................................ ................................ ...................... 81 4 22 Variation of the velocity contour along the span (z direction) of the microscale actuator. ................................ ................................ ................................ ............. 82 4 23 Divergence of velocity for devi ce with 50 100 100 m geometry and 10 m thick silicon oxide dielectric, operated at 4 kVpp and 1 kHz. .............................. 84 4 24 Infrared image of d ielectric surface temperature taken at t = 120 seconds. ....... 87 4 25 Evolution of the surface temperature of the DBD dielectric layer during two minutes of operation. ................................ ................................ .......................... 88 5 1 First generation of microscale DBD actuators designed for device testing and characterization. ................................ ................................ ................................ 89 5 2 Average normalized power consumed for four microscale DBD device geometries plotted against applied volt age. ................................ ....................... 91 5 3 Velocity contour plots for microscale DBD actuators. ................................ ......... 93 5 4 Velocity profiles of two microscale DBD actuator geometries having a 10 m thick polyimide dielectric, taken at x = 3mm downstream. ................................ .. 94 5 5 Velocity profiles of two microscale DBD actuator geometries having a 10 m thick polyimide dielectric, taken at x = 3mm do wnstream. ................................ .. 95 5 6 Plot of maximum velocity (x component) as a function of ground electrode width for var ious top electrode geometr ies. ................................ ........................ 96 5 7 Velocity profiles for two microscale DBD actuators having a 10 m thick polyimide dielectric. Comparing 1 mm and 5 mm ground electrodes. ................ 97 5 8 Pressure based velocity measurement. ................................ ............................. 98 5 9 Displacement measurements from the torsional balance for a microscale DBD actuator having a 20 m thick SU 8 dielectric. ................................ ......... 100
12 5 10 Thrust values measured for four microscale DBD actuators. ........................... 101 5 11 Thrust values computed based on the control volume analysis for devices with 10 m polyimide dielectric. ................................ ................................ ........ 102 5 12 Plasma force computed from the spatial body force estimation for devices with 10 m polyimide dielectric. ................................ ................................ ........ 103 5 13 Average temperature increase plotted against applied voltage for devices with different electrode geometries. ................................ ................................ .. 107 5 14 Polyimide failure analysis. ................................ ................................ ................ 109 5 15 SEM Images of polyimide dielectric surface after discharge. ........................... 110 5 16 Optical inspection of DBD actutaors at 10x magnification. ............................... 111 5 17 Failed microscale DBD actuator due to sputtering of the metal electrode from the polyimide surface. ................................ ................................ ....................... 111 6 1 Average power consumed plotted against input voltage for devices with 5 m thick silicon dioxide dielectric layer. ................................ ............................ 113 6 2 Average power consumed plotted against input voltage for devices with 10 m thick silicon dioxide dielectric layer. ................................ ............................ 114 6 3 Velocity profiles of four microscale DBD actuator geometries having a 5 m thick dielectric layer taken at x = 3mm downstream. ................................ ....... 116 6 4 Velocity profiles of four microscale DBD actuator geometries having a 10 m thick dielectric layer, taken at x = 3mm downstream. ................................ 117 6 5 Plot of maximum velocity (x component) as a function of voltage input for various microscale DBD actuator geometries. ................................ .................. 118 6 6 Velocity contour plots for microscale DBD actuator having 1000 m ground (left column) and 1000 m ground (right column). ................................ ............ 119 6 7 Frequency dependency of microscale DBD actuators. ................................ ..... 121 6 8 Thrust values computed based on the control volume analysis for devices with 5 m SiO 2 dielectric. ................................ ................................ .................. 122 6 9 Plasma force computed from the spatial body force estimation for devices with 5 m SiO 2 dielectric. ................................ ................................ .................. 123 6 10 Thrust values computed based on the control volume analysis for devices with 10 m SiO 2 dielectric. ................................ ................................ ................ 124
13 6 11 Plasm a force computed from the spatial body force estimation for devices with 10 m SiO 2 dielectric. ................................ ................................ ................ 124 6 12 Comparison between control volume estimation and direct integration of spatial body force. ................................ ................................ ............................ 125 6 13 Streamwise thrust component plotted against frequ ency for devices with 100 m ground geometry and 10 m SiO 2 dielectric, operated at 4.0 kVpp. ........... 126 6 14 Average temperature increas e plotted against applied voltage for devices with different electrode geometries. ................................ ................................ .. 127 8 1 Array of three microscale DBD actuators with 5 mm spacing. .......................... 145 8 2 Array of five microscale DBD actuators with 5 mm spacing. 50 100 100 m geometry with 10 m thick polyimide dielectric. ................................ ............... 146 8 3 Array of three microscale DBD actuators with 10 mm spacing. 50 100 50 m geometry with a 10 m thick polyimide dielectric. ................................ ............ 147 8 4 Serpentine DBD actuators. ................................ ................................ ............... 148 8 5 Microscale DBD actuators with serrated anode and 1 mm wide ground electrodes. ................................ ................................ ................................ ........ 148 8 6 Plasma synthetic jet actuators. ................................ ................................ ......... 148 8 7 Microscale DBD actuators shown d uring operation and under test conditions. 149 8 8 Microscale DBD actuator with serrated anode; base to height ratio r = 2. ........ 150 8 9 Comparison between serrated and standard DBD actuators .......................... 151 A 1 Velocity fields for microscale DBD actuators ................................ ................... 152 A 2 Velocity fields for microscale DBD actuator with 500 m wide ground electrode geometry, shown for various electrode sizes. 5 kVpp, 1 kHz input. .. 153 A 3 Velocity fields for microscale DBD actuator with 1000 m wide ground electrode geometry, shown for various electrode sizes. 5 kVpp, 1 kHz input. .. 154 A 4 Velocity fields for microscale DBD actuator with 50 100 50 m geometry, shown for voltages ranging from 2 to 3.5 kVpp, 1 kHz input. ............................ 155 A 5 Velocity fields for microscale DBD actuator with 50 100 100 m geometry, shown for voltages ranging from 2 to 4 kVpp, 1 kHz input. ............................... 155 A 6 Velocity fields for microscale DBD actuator with 50 100 500 m geometry, shown for voltages ranging from 2 to 4 kVpp, 1 kHz input. ............................... 156
14 A 7 Velocity fields for microscale DBD actuator with 50 100 1000 m geometry, shown for voltages ranging from 2 to 4 kVpp, 1 kHz input. ............................... 157 A 8 Velocity fields for microscale DBD actuator with 50 100 50 m geometry, shown for voltages ranging from 2 to 7 kVpp, 1 kHz input. ............................... 158 A 9 Velocity fields for microscale DBD actuator with 50 100 100 m geometry, shown for voltages ranging from 2 to 7 kV pp, 1 kHz input. ............................... 159 A 10 Velocity fields for microscale DBD actuator with 50 100 500 m geometry, shown for voltages ranging fro m 2.5 to 6.5 kVpp, 1 kHz input. ......................... 160 A 11 Velocity fields for microscale DBD actuator with 50 100 1000 m geometry, shown for voltages ranging from 2.5 to 7 kVpp, 1 kHz input. ............................ 161
15 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy INVESTIGATION OF MICROSCALE DIELECTRIC BARRIER DISCHARGE PLASMA DEVICES By Justin C. Zito August 2013 Chair: David P. Arnold Major: Electrical and Computer Engineering This dissertation presents research performed on reduced scale dielectric barrier discharge (DBD) plasma actuators. A first generation of microscale DBD actuators are design ed and manufactured using polymeric dielectric layers, and successful l y demonstrate operation at reduced scales. The actuators a re 1 cm long and vary in width from tens of microns to several millimeters. A thin film polymer or ceramic material is used as the dielectric barrier with thicknesses from 5 to 20 microns. The devic es are characterized for their electrical, fluidic and mechanical performance. With electrical input of 5 kVpp, 1 kHz, the microscale DBD actuators induce a wall jet with velocity reaching up to 2 m/s and produce 3 .5 mN/m of thrust, while consuming an aver age power of 20 W/m. A 5 mN/m plasma body force was observed, acting on the surrounding air. Failure of the microscale DBD actuators is investigated using thermal measurements of the dielectric surface in addition to both optical and scanning electron micr oscopy. The cause of device failure i s identified as erosion of the dielectric surface due to collisions with ions from the discharge. A second generation of microscale actuators is then designed and manufactured using a more reliable dielectric material, namely silicon dioxide. These actuators
16 demonstrate a significant improvement in device lifetime compared with first generation microscale DBD actuators The increase in actuator lifetime allowed the e lectrical, fluidic and mechanical characterization to be repeated over several input voltages and frequencies At 7 kVpp, 1 kHz, the actuators with SiO 2 dielectric induced velocities up to 1.5 m/s and demonstrated 1.4 mN/m of thrust while consuming an average power of 41 W/m. The plasma body force reached up to 2.5 mN/m. Depending on electrical input, the induced velocity and thrust span an order of magnitude in range. C omparisons are made with macroscale DBD actuator s performance and power co nsumption with the mass and volume of the actuator design. The small size and of microscale DBD actuators reduces its weight and power requirements, making them attractive for portable or battery powered applications (e.g., on UAVs).
17 CHAPTER 1 INTRODUCTION Portions of this dissertation have been published in the papers reported in Appendix C efforts to improve aircraft flight control systems for increased stability and e fficiency. Fixed wing aircraft use a n array of control mechanisms. Some of these include the throttle for control of the speed of the aircraft and a system of moveable flight surfaces (ailerons, elevators, rudd er s etc.) for control of the orientation. Additional devices such as flaps, slats, spoilers and/or trim tabs offer additional aerodynamic control. The unifying goal in the design of all aero control mechanisms is to control the flow over the aircraft in o rder to affect the flight dynamics. These flight control systems typically aim to improve some aspect of flight performance, such as stability and maneuverability, with minimum penalty in cost, weight, and complexity. Flow control systems may be separated into two categories: active and passive. Passive control devices include flaps, slats, slots and vortex generators that interact with the flow witho ut requiring an applied input. Active control systems use actuators that interact with the flow only when there is an external input applied Some examples of active flow control methods include zero net mass flux actuators ( such as synthetic jets), circulation control systems (valves and combustion systems that induce suction and/or blowing into the flow), and vibrating ribbons or oscillating wires that induce local fluid motion / p erturbations ( Cattafesta and Sheplak 2011) These a ctive control techniques convert electrical signals into desired physical effects that act on the fluid.
18 Another recent interest for active flow control is the use of pl asmas to induce flow effects. The most a ttractive features for using plasma act uators for flow control are that they provide very fast response time with no mechanically moving parts. 1.1 Plasma Basics Plasma s ha ve been studie d for over a century. The discovery of plasma was by Sir William Cr ooke in 1879, who invented the tube which consisted of a glass cylinder with m etal electrodes on either end (Crookes 1879) radiated matter inside the glass vessel when a high voltage was appli ed across the electrodes. The radiated matter was identified in 1897 as a cathod e ray made up of free electrons b y Sir Joseph Thompson (Thomson 1897) In 1928, about 30 years later, American scientist Irving Langmuir coined the term plasma ( Langmuir 1928 ) Lang muir is best known for his invention of the Langmuir probe which is an electrostatic probe that enabled measurements of the electron density and electron temperature within th e plasma. in the context of the study of the universe Plasma research has since spread to include topics from energy/fusion physics and chemistry to materials science and semiconductor/ electronics processing. A more recent surge of research has re surfaced in the last f ew decades with attention paid towards the use of plasma discharge in flow control applications. Generally speaking, plasma is a collection of charged and neutral particles. More specifically, plasma is a gaseous assembly of electrons, ions, and neutral molecules residing in electric and mag netic fields ( Hutchinson 2002). Plasma is similar to a gas, but uniquely different as it is a good conductor of electricity and also affected by a
19 magnetic field. th state of matter over 99% of matter in the universe ( Chen 1984 ) Plasmas occur naturally in outer space phere as lightning and auroras Plasma may be artificially created to make use of i ts con sistently ionized state. Example applications of man m ade plasma are the hall effect thruster (space propulsion), etching (electronics manufacturing), ozone generation (water purification), sterilization (medical devices, toxic waste treatment) and chemical spectroscopy (material analysis). Plasma d ischarge occurs w hen a sufficiently high electric field between conductors result s in electrical breakdown of the med ium between them. Typically a non conductive gaseous medium is us ed between the two electrodes The term discharge refers to any flow of electric current through an ionized gas; or any process of ionization of the gas by the applied electric field ( Raizer 1991 ) There are several types of plasma discharges, the most common being corona discharge, arc discharge and glow discharge Corona discharge (Figure 1 1A) occurs when the electric field is high enough to ionize the gas surrounding a conductor. Typically, a sharp point or thin wire is used to create a concentrated electric field at the conductor generate ozone, to create ionic wind (cooling applications), and to ionize sampl es for spectroscopic analysis. With further increase in the electric field, arc discharge occurs where the highly concentrated field ph ysically arcs between the conductor and another surface. Arc discharge (Figure 1 1B) is commonly used for welding and cutting ( plasma machining ), and is also used in arc lamps for lighting applications. In contrast, glow
20 discharge (Figure 1 1C) is a more uniform discharge that is characterized by having low curre weakly ionized Weakly ionized plasma discharge s are categorized as hav ing the following features: (1) they are driven electrically; (2) charged particle collisions with neutral gas molecules are important; (3) there are boundaries at which surface losses are important; (4) ionization of neutrals sustains the plasma in the steady state; and (5) the electrons are not in the rmal equilibrium with the ions ( Lieberman and Lichtenberg 2005 ) More generally, a plasma is considered weakly ionized when the ratio of ionized particle density to neutral particle density is on the order of 10 6 10 8 (or less), equivalently, when one in 1 to 100 million particles are ionized within the discharge A B C Figure 1 1. Three common types of plasma discharge A) corona discharge B) A rc discharge (reprinted with permission, Mircea Madau 2005) C) G low discharge. 1.2 Dielectric Barrier Discharge Devices A dielectric barrier discharge (DBD) is a particular glow type plasma discharge that includes a dielectric mater ial in between two electrodes. As the electric field is increased the tendency for an arc discharge to occur becomes more likely. The dielectric acts as a barrier limiting the current a nd preventing arcing from occurring The DBD architecture maintain s the plasma in a glow discharge state and allow s for large
21 electric fields to be applied between the conductors that would otherwise arc. The discharge can be configured to provide either a volume or surface discharge. The difference between a surface and volume discharge is that a surface discharge occurs over an area just above the dielectric, whereas a volume discharge occurs in a gaseous region be tween two electrodes in space. Figure 1 2 provides a schematic of both volume and surface discharge configurations A B Figure 1 2. DBD Configurations A ) DBD volume discharge B ) DBD surface discharge The thickness of the dielectric and separation between the electrodes affect the required voltage to ionize t he gas for discharge to occur. The breakdown voltage depends on the distance separating the electrodes, as it is the electric field strength (V/m) that must be suff iciently large for ionization. For example, air at atmospheric p ressure and 20 C requires ~30 kV/cm for electrical breakdown. It should be noted that the breakdown voltage does not only depend upon the distance between electrodes; it is also affected by temperature, pressure, and properties of the medium separating th e electrodes, such as relative permittivity (Kogelschatz 2003) (and applies for either gaseous or solid dielectric mediums). Further details regarding the properties affecting electrical breakdown are provided in Section 2.2.2. 1.3 Application of DBD Devic es for Flow Control D ielectric barrier discharge (D BD ) plasma devices have been shown to provide unique and complex electronic control over fluidic behavior. These devices offer simple
22 robust construction (no moving parts) and provide near instantaneous t emporal response (plasma time scales much faster response than the typical fluidic time scales). These attributes are appealing for many applications, e specially active flow control. A typical DBD device used in flow control applications consist s of two el ectrodes place d asymmetrically on either side of a dielectric material (Roth et al 2000) A schematic of a standard DBD actuator i s shown in Figure 1 3. A high voltage time varying signal is applied across the two electrodes, creating a stron g electric fi eld between them. With sufficiently high applied voltage, the electric field causes the gas just above the dielectric surface to become weakly ioniz ed creating a plasma discharge. This creates a surface discharge in the region between the two electrodes Figure 1 3. Schematic view of a single dielectric barrier discharge (DBD) plasma actuator. The plasma discharge provides an electrohydrodynamic (EHD) body force on the device and imparts momentum into the gas. The induced flow forms as a wall jet along the surface of the dielectric downstream from the powered electrode A corresponding reaction force (thrust) is also imparted on the actuator. The reaction force on the actuator is equal to the force imparted to the fluid less the viscous force acting over the device surface.
23 In most cases it is common to encapsulate the bottom electrode with an insulating tape or epoxy in order to suppress discharge from occurring on the bo ttom surface of the dielectric. If both the top and bottom electrodes were exposed, the discharge would occur on either surface of the dielectric creating opposing plasmas, eliminating the directional, single sided discharge that is desired for using DBD in flow control applications. Most prior investigations on DBD plasma actuators hav e focused on macroscale devices, having characteristic device dimensions ranging from millimeters to 10s of centimeters (discussed further in S ection 2.1 ). However d espite their advantages of speed and simplicity, employment of DBD devices as flow control actuators in actual flight vehicles has been limited by modest fluidic control authority and the requirement for lar ge high voltage power supplies. Additionally, a d hoc, often unrepeatable fabrication approaches have furthe r hampered their use. The limita tions of DBD actuators are further discussed in Section 2.3. 1.4 Research Objectives T he objective of th is dissertation is to fabricate and systematically characterize DBD devices having micrometer scale dimensions Devices are fabricated using wafer level microfabrication processes yielding DBD plasma devices with dimensions reaching down to 5 m. C omprehensive experimental methods are then used to characteriz e their electrical, f luidic and mechanical behavior. These experiments add to scientific understanding of plasma dynamics, particularly microscale effects. There are numerous key moti vations for the performed work. First, m iniaturization of the DBD architectures to microscale dimensions addresses important system considerations namely (i) to increase the per volume actuator authority, (ii) to reduce
24 the voltage and power requirements, and (iii) to improve the manufacturability. The the plasma discharge can c ouple into the surrounding gas, typically measured as the The high voltages and power consumed by conventional, macroscale actuators require a large, massive power supply. Using m icroscale electrode geometries enable discharge at significantly reduced voltages, providing a corresponding reduction i n the re quirement of the power supply size and mass Manufacturing of macroscale devices is performed using hand assembly and offers very limited precision. Microscale actuators take advantage of semiconductor fabrication technologies and allow precise dimensional control of device geometries with excellent alignme nt of features between layers. In addition the systematic and comprehensive experimental measurements will provide a unique experimental benchmark database for reference by other researchers. For example, the data can be used to validate numerical plasma a nd electrohydrodynamic models. The core intellectual merit of this project lies at the intersection of e lectromagnetics, electrohydrodynamics, and microtechnology. In the end, t hese experiments will characterize microscale DBD actuator performance and provide a database that can be leveraged for future application systems such as aerodynamic flow control, a erospace propulsion, microfluidic pump ing ozone generation, water purification, medical sterilization surface treatment/activation, plasma machining, chemical detection, emission spectroscopy and plasma display panels For these other (non flow control applications), another primary benefit of using microscale devices is
25 sterilization of tools, and plasma machining.
26 CHAPTER 2 BACKGR OUND AND MOTIVATION This chapter presents current research on DBD plasma actuators including some of the advantageous fluidi c effects that make them attractive for flow control applications A review of research efforts focused on increasing actu ator perfo rmance is presented. Several trends are discussed regarding the electrical, fluidic, and mechani cal behavior of these devices. Some of the major limitations of macroscale DBD actuators are presented followed by a section on the benefits of microscale DBD d evices, which address some of the macroscale limitations. T he chapter concludes with a summary of research related to microscale discharge devices, highlighting relevant results. 2.1 DBD Plasma Device Overview DBD actuators have been most widely studied for active flow control of relatively low Reynolds number (Re) flows. The Reynolds number is a parameter relating the ratio of internal forces with viscous forces, defined as where L is the length scale of the flow, u o is the flo w velocity, and is the kinematic viscosity of the fluid. The Reynolds number is also used to characterize laminar and turbulent flow regimes. Hence, t he combination of small length scale s and induced velocities make DBD actuators attractive for flow cont rol applications having low Reynolds number. Example applications include prevention or promotion of separation over an airfoil, drag re duction, and lift enhancement. Figure 2 1 wind tunnel (Roth 2003) showing the flow separation over the airfoil before the discharge is applied, and the flow re attachment along the airfoil when t he plasma discharge is present
27 A B Figure 2 1 DBD demonstrating flow re attachment (Roth 2003). A) Airfoil in wind t unnel without plasma actuation (NACA 0015 airfoil, 2.85 m/s free stream velocity, 12 angle of attack). B) Demonstrating flow re atta chment with DBD plasma actuators 2.1.1 State of the Art Many efforts have been made to increase the control authority (fluidic impact) of DBD actuators ( Roth 2003; Enloe et al. 2004b; Roth and Dai 2006 ; Corke et al. 2007 ; Forte et al. 2007 ; Jolibois and Moreau 2009; Thomas et al. 2009) while simultaneously reducing the power req uirements. Paramet ric trends have been stud ied ranging from input voltage amplitude and frequency, waveform shape (sinusoidal, pulsed, saw tooth, triangular, etc.), material pr operties, and device geometry. Waveforms with steep slopes, e.g., square waves and saw tooth waves, were reported to result in large current values which contribute poorly to the induced velocity (Jolibois and Moreau 2009). Typical DBD actuators produce a wall jet with velocity of 3 6 m/s occurring about 0.5 1 mm above the dielectric surface. Maximum induced velocities hav e been reported up to 7 m/s as experimentally measured ( Forte et al. 2007), and has been numerically predicted to saturate at approximately 10 m/s ( Likhanskii et al. 2010). Figure 2 2 displays the average velocity profile measured for a standard macroscale DBD actuator using particle imaging velocimetry measurement techniques ( Durscher and Roy 201 2 )
28 Figure 2 2 Time averaged velocity flow field from a standard macroscale DBD pla sma actuator ( Durscher and Roy 201 2 ) The detailed geometries of the electrodes have been found to play an important role. Thomas et al. (2009) found a reduction in the required voltage when using electrodes with a serrated edge as compared to a straight edge electrode configuration. They also indicated that the thrust was increased by over 100% for input voltage amplitude below 15 kV (30 kVpp) Abe et al. ( 2008 ) also investigated the electrode geometry and found an increase in the momentum transfer to th e fluid for thinner electrodes. The increase is believed to occur from having a stronger local electric field near th e edge of a thinner electrode. In further investigation by Abe et al. (2008) a metal mesh was used for the electrodes in order to provide concentrated local electric field poi nts where the mesh is cut off. Again, an increase in the thrust was found, up to 150% at atmospheric pressure. In addition to basic parallel stripe electrodes, DBD actuators have been made in a number of configuration s to induc e flow in multiple directions. Both circular ( Santhanakrishnan and Jacob 2007 ) and parallel ( Benard et al. 2008 ) sets of actuator s ha ve been designed which are used to pinch the fluid normal to the actuator surface, acting as a plasma jet as shown schematically in Figure 2 3 A Directional plasma jets
29 may be induced through control of the relative voltages applied to individual parallel actuator pairs (Figure 2 3 B). Serpentine or oscillatory electrode geometries have been investigated induc ing both pinching and spreading effects on the fluid, creating three ( Wang and Roy 2009 a ; Durscher and Roy 2011 ) A B Figure 2 3. Plasma synthetic jet actuator (Benard et al. 2008). A) Schematic of two actuator co nfiguration for creating a plasma jet. B) Induced flow showing directional control capability with voltage biasing. The two numbers in the images indicate the voltage applied to the left and righ t electrode pairs, respectively. 2. 1. 2 DBD Plasma Devices Scaling Trends Extensive reviews of plasma actuators for use in flow control ( Enloe et al. 2004a ; Enloe et al. 2004b ; Moreau 2007 ; Thomas et al. 2009; Corke et al. 2 010 ; Cattafesta and Sheplak 2011 ) summarize several performance trends that have been vali da ted repeatedly in experiments. For example, t he power dissipated by DBD actuators is found to be exponentially proportional to the sinusoidal input voltage amplitude, as (Enloe et al. 2004b). The re is also dependency on the power with input frequency, al though this has been shown to be small in comparison to the exponent ial dependency upon amplitude. Furthermore, both the induced velocity and the resulting thrust are also reported to scale exponentially with voltage, having similar expone nts (ranging betw een 3 and 4). However, the velocity and thrust are known to saturate a t some input voltage or power (Thomas et al. 2009; Durscher et al. 2012) In this
30 saturation regime, increasing the input no longer increases the velocity or thrust. In fact, it was rece ntly shown by Durscher et al. (2012) that in the saturation mode the measured velocity and thrust both decrease in comparison to their values before saturation. Saturation mode can be identified visually by the onset of filamentary discharge events and als o by a large increase in power consumption. In addition, the thrust and power nominally increase l inearly with electrode length and are typically reported as normalized by the electrode length 2. 1. 3 DBD Plasma Devices Limitations As described above, DBD plasma devices have been widely st udied for active flow control (Roth et al. 2000; Roth 2003; Enloe et al. 2004a; Baughn et al. 2006; Roth and Dai 2006; Corke et al. 2007; Forte et al. 2007; Enloe et al. 2009; Thomas et al. 2009; Versailles et al. 2010 ; Kriegseis et al. 2013b ) Despite a decade of research by many research groups, their primary application has so far been largely limited t o low speed flow modification. One dominant reason is the l imited flow control authority. The momentum transfer to the fluid can be related to the force density of the plasma discharge, where the force density is defined as the plasma body force acting on the fluid per unit volume. At the macro scale, the force density acting on the fluid is on the order of 1 10 kN/m 3 It will be shown in the following sections that the force density of microscale plasma discharge may be much higher, on the order 1 M N/m 3 as predicted numerically. Another limiting criterion for utilization in other applications such as aero propulsion is the excessively large size and weight o f the required power supplies. In order to create the tens of kilovolts required for ionization of macroscale plasma discharge, large amplifiers and/or transfo rmers are typically necessary. Reduction of the
31 voltage levels and/or total power requirements could enable new applications. This would require an on board power supply of reasonable size and weight. It will be shown that the power requirement (per unit length) for microscale discharg e is dramatically reduced, from ~100 W/m for macroscale devices down to ~20 W/m. Most prior research on macroscale DBD plasma actuators have been manufactured by hand, typically by cutting strips of adhesively backed copper and placing them on either sid e of the dielectric layer. This handmade construction limits the dimensional precision of the electrode geometry as well as tolerances in the assembly, e.g. manually aligning the asymmetric electrodes. These manufacturing variations lead to inconsistency i n the resulting device performance M icroscale actuators can be made with excellent dimensional control and with precision alignment through the use of semiconductor fabrication techniques. 2. 2 Potential Benefits of Microscale DBD Plasma Devices Reduction of the geometric dimensions of a DBD device is anticipated to introduce interesting physical phe nomena and scale dependencies. In addition to the intellectual merit of fundamental discovery and exploration of these reduced scale physics, there are a numbe r of potential practical benefits of microscale plasmas including: Increased per volume fluid control authority Reduced power requirements Improved manufacturability These concepts are further described below. 2. 2 .1 Increased per Volume Fluid Control A uthority Recent experiments and numerical predictions ( Longwitz 2004; Wang and Roy 2009 b ) give evidence of potential increases in fluidic control authority for microscale plasma devices. The aforementioned papers have considered a volume plasma
32 discharge between two electrodes in air, where the separation distance ranged from 200 m down to 10 m. The experimental measurements of the breakdown electric field from Longwitz ( 2004) (black) showed excellent agreement with the numerical predictions of Wang and Roy (2009 b ) (blue) as illustrated in Figure 2 4. Specifically, the electric field ( E ) is shown to increase expo nentially with decreasing gap. The numerical predictions are found using a first principles approach solving a coupled system of hydro dynamic plasma equations and Poisson equation for ion density, electron density, and electric field distribution ( Wang and Roy 2009 b ) Figure 2 4 Electric field, charge and force density plotted as a function of electrode separation distance. Numerical results (blue circles) ( Wang and Roy 2009 b ) compared with experimental results (black squares) ( Longwitz 2004) of the breakd own electric field (left axis). The charge density (red triangles) and plasma f orce density (orange diamonds) are also plo tted (right axis). The numerical models enable prediction of parameters that are di fficult to physically measure. Interestingly, while the electric field is predicted to increase exponentially (matching the experiments), t he numerical model also indicates a nominally constant net charge density, q (red), where q is the difference between the ion and electron densities inside the plasma. T he product of charge density and electric field, qE is the Lorentz force density (assuming no external magnetic field) which
33 imparts force on the fluid The numerical model indicate s that this force density (orange) scales up by 2 3 orders of magnitude. If true, t his Lorentz force density increase should correlate with increased fluidic actuation (i.e. fluidic control auth ority regarding a flow control actuator) It should be emphasized that the results described above are for a volume discharge, and not for a DBD surface discharge. However, the electric field trend and fundamental physical explanations can be extended to dielectric barrier discharges The physics behind this increased force density are described as follows. Following the electric field in Figure 2 4, a strong exponential increase in the field strength occurs when the electrode gap reaches below ~50 m. It is in this geometrical regime that the increase in the force density is predicted. Hence, b y using thin dielectric barriers, on the order of 10 m in thickness, the microscale DBD actuators attempt to leverage the increase in electric field strength. 2. 2 .2 Reduced Power Requirements In addition to the potential increased flow control authority described above, microscale plasma devices are anticipated to have dramatic ally lower power requirements. A major consideration for use a plasma device is the break down voltage, the voltage required to cause ionization of the gas For relatively large electrode separation (millimeter or larger), the breakdown voltage generally decreases with closer electrode spacing since the field requi red for bre akdown is relatively constant. This reduction in voltage results in a reduction in the required power supply si ze. A lower voltage power supply is generally smaller and lighter than a higher voltage supply with equivalent power rating. The decr ease in breakdown voltage with micron scale gaps is somewhat debated. ( Paschen 1889) is a mathematical formula predicting the breakdown
34 voltage for parallel plate electrodes as a function of both the gas pressu re and the electrode distance. Paschen found that, for a constant pressure, the breakdown voltage decreases with decreasing electrode gap distance, but then increases rapidly for electrode gaps below some threshold distance (typic ally on the micrometer scale). In r (1959) experimentally tested the breakdown voltage of metal sphere electrodes in air and vacuum and found devi Torres and Dhariwal (1999) repeated similar experiments, again using metal sphere electrodes, and also found deviat io The experimental results from Torres and Dhariwal (1999) are shown in Figure 2 5 More recently, i n 2004, Longwitz (2004) studied breakdown voltages for microfabricated thin strip electrodes, and he also observed differen t trends from the predictions of Paschen. Figure 2 5 Measurements of breakdown voltage show ing law ( Torres and Dhariwal 1999 ). electrode geometries, as deviations were shown for both spherical and planar s trip electrode configurations. And in fact, where Paschen found an increase in the required voltage for micron scale gaps, more recent research has indicated a continued
35 decrease in the re quired breakdown voltage. This decrease in the voltage requirement consumption. If discharge can be obtained using voltages on the order of 10 100 volts, rather than the kilovolts required at the macro scale, the power supply size and complexity for operating DBD actuators c ould significantly be reduced. 2. 2 .3 Improved Manufacturability In addition to potential improvements in performance, the manufacturability of plasma devices m ay be improved by considering microscale fabrication Fabrication of microscale DBD plasma devices will employ semiconductor like, batch fabrication techniques, which affords several key benefits. First, this approach allows precise dimensional control for device geometries, as well as precise alignment b etween layers of electrodes. Thin film deposition and photolithography processes are used to create and align the e lectrodes with great accuracy. Multiple layers of complex 2 D electrode geometries can be f abricated via a series of chemical and mechanical processing steps. Compared to the macroscale fabrication, often using a razor blade to hand cut and place electrodes, lithographic fabrication techniques eliminate the u ncertainties of hand assembly. Anoth er advantage of using semiconductor fabrication is that of batch processing, which enables the manufacturing of many actuators in p arallel on a single substrate. The potential for arrays of numerous microscale actuators packed densely on the device combine s the advantages of increased force density with those of batch fabrication. 2. 3 Summary of Related Research on Microscale Plasma Devices There ha ve been previously reported microscale discharge generating devices ranging in application from pressure sensing (Wright and Gianchandani 2009) mass
36 spectrometry (Taylor et al. 2000) and optical emission spectroscopy (Marcus and Davis 2001) to flow control actuators (Okochi et al. 2009) and microthrus ters (Guman and Nathanson 1970) However, to knowledge there is no known body of work on surface discharge DBD devices with micrometer size geometries The most similar DBD research is by Okochi et al. (2009) who created DBD actuators with el ectrodes as narrow as one millimeter (Figure 2 6 A). The electrodes are deposited on either side of a 525 m thick Pyrex wafer which acted as the dielectric barrier. Velocity measurements were made using la ser Doppler velocimetry (LDV). Velocities up to 2 m/s were measured using 1 mm wide electrodes with no gap or displacement between the electrodes (Figure 2 6 B). The actuators were operated at 10 kV pp and 10 kHz. A B Figure 2 6 Millimeter scale DBD actuator for flow control (Okochi et al. 2009). A) DBD actuator used for velocity measurements made using a glass dielectric and chrome covered electrodes. B) Horizontal velocity component measured at several downstream locations. Bass et al. (2001) designed a capacitive coupled dielectric barrier volum e discharge using helium gas within a 200 m wide quartz channel. The discharge is used to ionize gases for optical emission spectroscopy. Ono et al. ( 2000 ) investigated breakdown voltages for microfabricated silicon electrodes with gaps ranging from 2 5 0 m (Figure 2 7 ) Their research focuses on corona discharges be tween two electrodes in space. A CCD camera was used to capture photon emissions and found that small
37 amounts of discharge occu r below the breakdown voltage. From Figure 2 7B, microdischarge is observed via light emission below the predicted breakdown voltage for both 2 m and 5 m electrode gaps. A B Figure 2 7 Investigation of breakdown voltages for microfabricated silicon electrodes (Ono et al. 2000). A) SEM image of planar silicon el ectrodes. B) Plot of light intensity vs. applied voltage Mitra and Gianchandani (2008) used microfabrication methods to create a handheld chemical vapor detector using pulsed corona discharges between electrodes separated by as little as 200 m. Results demonstrate detection of 100 ppm isopropyl alcohol and 320 ppm acetone in the presence o f air at atmospheric pressure. Figure 2 8 A provides an image of the microdischarge device. The images in figures B and C demonstrate the emission spectra for air and a mixture of acetone in air, showing detection of carbon compounds in the acetone. A B C Figure 2 8 A chemical vapor detector based on microdischarge excitation (Mitra and Gianchandani 2008). A) Fabricated device, microdischar g e occurs at the center of t he electrodes. B) Detection of atmospheric air. C) Dete ction of acetone in air
38 Hopwood et al. ( 2000; 200 2 ) also utilized MEMS fabrication techniques in creating an inductively coupled plasma generator consisting of a 3 turn, 5 mm diameter planar s piral inductor (Figure 2 9 ). The discharge is shown to operate in argon and air at pressures from 0.1 to 10 torr. Minayeva and Hopwood (2002) demonstrate d detection of sulfur dioxide as low as 3 ppm using emission spectroscopy Surprisingly, they we re able to initiate plasma using only 1.5 W of p ower with +/ 12 V at 450 MHz. Gianchandani et al. ( 2009 ) present ed a review of portable microdischarge devices specific ally for sensing and detecting applications. The review focuses on sensors capable of liquid, g as and radiation detection with emphasis on compact size and low power consumption Several of the detectors are able to operate in ambient air at atmospheric pressure. A B Figure 2 9 Microfabricated inductively coupled plasma for emission spectroscopy ( Minayeva and Hopwood 2002) A) Top view of device showing the inductor coil and two capacitors for impedance matching B) Side v iew of plasma discharge operating in argon Of the several configurations of microscale discharge summarized above, none have investigated micrometer geometries with DBD surface plasma generation. However, the
39 devices reviewed above show several confirmatory results that support the work perform ed here. Some key noted observations are the demonstration of microdischarge in atmospheric pressure in air, the ability to initiate and sustain discharge with low power requirements, and the capability to transfer momentum to the flow.
40 CHAPTER 3 MICROSCALE DESIGN AN D FABRICATION This chapter presents the device manufacturing steps, beginning with a brief description of the physical dimensions and composition of the plasma devices and followed with a step by step explanatio n of the fabrication process. T he devices are created layer by layer using additive thin film processes on a glass substrate. 3.1 Geometry and Materials The goal of this research is to investigate the behavior and flow characteristics of D BD plasma devic es with microscale dimensions. The se microscale DBD plasma devices will operate at standard atmospheric pressure and temperature, utilizin g air as the discharge medium. Figure 3 1 shows a schematic of the microscale device geometry. First generation devic es were designed to mimic macro scale plasma actuators but with scaled down geometries. These devices used a polymer based dielectric layer from 1 0 2 0 m in thickness. Polyimide, commonly known by its brand name Kapton, has been used wid ely as th e discharge barrier for DBD actuators (Enloe et al. 2004a; Enloe et al. 2004b; Moreau 2007; Thomas et al. 2009) The rrier for the plasma discharge; it must withstand strong electric fields and be c hemical ly inert from reacting with the discharge In addition to high dielectric strength, low relative permittivity is also desired in order to maximize the electric field lines through the gas locally above the actuator (higher permittivity causes the electric fi eld to take preference of traveling to ground through the barrier material instead of the surrounding air). Another polymer, SU 8, is an epoxy based material that is commonly used in microelectromechanical systems ( MEMS ) manufacturing as a structural laye r with good
41 chemical stability. An advantage of using SU 8 is that it is a photosensitive polymer thus lending itself to photolithographic patterning. For the second generation of microscale DBD actuators, a ceramic dielectric material is used instead of p olymers. Specifically, silicon dioxide (SiO 2 ) is chosen for its proven use in microelectronics as an insulation layer, and for it s similarity to glass (glass is primarily compromised of SiO 2 ), which is also frequently used for macroscale DBD actuators ( Pon s et al. 2005; Forte et al. 2007; Gregory et al. 2007; Moreau 2007; Thomas et al. 2009). Figure 3 1 Microscale DBD actuator design. A) Top view schematic. B) C ross section schematic indicating actuator dimensions Thin film copper electrodes (typically 0. 5 m thick) range in width from 1 0 5 0 00 m and are offset by a lateral gap ranging from 0 500 m The e lectrodes are 10 mm in length. The discharge created is a surface discharge, as compared to a volume discharge, imparting a wall jet wit h maximum velocity occurring 0.5 1 mm above the dielectric surface. T he discharge may provide 2 D or 3 D flow actuation, controll ed by the electrode geometry. S traight parallel electrodes are shown in the schematic. Throughout this document, the device g eometry is often identified using three numbers,
42 e.g., 50 100 100 m; these values represent the width of the exposed electrode the electrode gap width and ground electrode width (units of microns). 3.2 Fabrication Process A 500 m thick glass ( specif ically soda lime glass) substrate is used on whi ch to fabricate DBD actuators. Glass was used as the substrate to avoid omnidirectional discharge that resulted when using a silicon substrate (due to the semiconducting properties of silicon). The individu al steps of the fabrication process are illustrated in Figure 3 2 cleanrooms A standard photolithographic process (Fairchild Corporation 1979; Campbell 1996; Madou 1997; Senturia 2001) is used to transfer the first electrode p attern to the glass substrate. Beginning with a clean glass wafer, photoresist is spin deposited (Suss Delta 80 Spinner) and patterned with th e bottom electrode structures. AZ 9260 (MicroChemicals) positive tone thick film photoresist is used in all lithography steps throu ghout the fabrication process. After spinning the photoresist, it is soft baked for 1 8 0 seconds at 110 C to evaporate any resi dual solvent. The pattern is exposed using a Karl Suss MA 6 Contact Aligner equipped with an ultra violet lamp in hard contact mode, having approximate intensity of 5 W/cm 2 at a wavelength of 365 nm (i line) An approximate dose of 75 0 mJ/cm 2 is used to expose a 7 m t hick target photoresist layer. This d ose is 30 % larger than that recommended for use with a silicon substrate to account for the transparency of glass (less light is reflected from the glass into the photoresist). The exposed photoresist is developed in a 3:1 ratio of deionized water and a potassium hydroxide based developer solution (MicroChemicals AZ 400K).
4 3 Figure 3 2. Cross section diagrams of device fabrication steps. Copper electrodes are then deposited via sputter deposition (K urt J. Lesker CMS 18) for a target thickness o f 0.5 m. Prior to sputtering, the substrate is exposed to oxygen for 30 seconds which cleans and roughens the deposition surface to increase adhesion of the sputtered metal. Following the oxygen, a rgon is pum ped into the chamber for another 30 seconds to create an inert low pressure environment before beg inning the sputtering process. Before depositing copper, a 50 nm titanium layer is sputtered to improve adhesion between the copper and glass substrate. Copp er is then
44 sputtered at 250 watts and 5 mTorr pressure, providing a deposition rate of 4 /sec. After sputtering, the photoresist and excess metal are removed technique, in which the wafer is soaked in a heated bath (70 C) of the organic solvent NMP (N Methyl 2 pyrrolidone) for approximately 10 minutes First generation devices use polyimide ( HD MicroSystems PI 261 1 ) as the dielectric material. The polyimide is spin deposited fo r a target thickness of 10 m. The polyimide is dispensed onto the wafer and ~15 seconds are given to allow the polyimid e to relax. The wafer is spun at 5 00 rpm for 7 seconds and then quickly ramp ed to 3500 rpm for 45 seconds. The polymer is th en soft baked by placing on a hotplate at 90C for 90 seconds and then immediately transferred to a second hotplate at 1 2 0C for another 90 seconds. This spin coat/soft bake process creates an approximately 5 m thick polyimide layer, and is repeated twice to achieve the desired 10 m thickness. After the final iteration, the polyimide layer is cured by ramping a hot plate from 150C to 350C at a rate of 4 C/min, and letting the wafer bake at 350C for 30 minutes The wafer is then allowed to cool to room temperature while on the hotplate t o minimize stress. Once the polyimide is completely cured it i s insoluble and reaches its optimal mechanical and electrical properties. For the devices with silicon dioxide dielectric, the SiO 2 is deposited using plasma enhanced chemical vapor deposition ( PECVD ) In this method, the substrate is placed into a chamber at 300 C and plasma is used to excite the reactive gases causing dissociation of precursor molecules for the chemical reaction th at follows. The SiO 2 is deposited at a rate of 42 nm/min, requiring a 4 hour deposition for a 10 m layer.
45 After the dielectric layer has been deposited, it is necessary to etch regions to enable electrical contacts to the bottom electrodes that are covere d by the dielectric Both polyimide and silicon dioxide are etched using a reactive ion etch (RIE), a low pressure dry chemical etch process The RIE system has an inductively coupled plasma (ICP) module that assists in generating and confining ions, while the platen potential is used to control the accelerati on of ions toward the substr ate during the etching process. A 20 m thick photoresist layer is used as an etch mask during the RIE step. To create the etch mask, a nother photolithography process is per formed (spin coat a 2 layer photoresist, exposing the pattern with UV lamp, and developing the exposed resist) to reveal th e contact regions in the etch mask so they are exposed for the following etch The wafer then undergoes the reactive ion etch The polyimide is etch ed using O 2 gas pumped into the chamber at a flow rate of 60 sccm (standard cubic centimeters per minute) using 2 00 W of platen power and 600 W for the ICP unit. To etch silicon dioxide, both CHF 3 and O 2 gas es are used with flow rates of 2 5 and 2.5 sccm respectively, with 50 W of platen power and 200 W for the ICP. To help dissipate heat from the surface, a substrate cooling system flows 12 sccm of helium onto the backside of the wafer throughout the process The etch rate for both polyimid e and silicon dioxide is approximately 500 n m/min using these parameters. The etch is allowed to run for an extra 10 minutes to help metal electrode / dielectric boundary. Each wafer has four continuity test structures patterned on the bottom electrode layer. A digital multimeter is used to check for continuity and confirms whether the dielectric has been completely etched through. After etching through the
46 dielectric material the photoresist mask is stripped in the NMP bath which completes the patterning of the dielectric barrier The final processing step is creatin g the top layer of electrodes. This process is very similar to manner in which the bottom electrodes were deposited: photolithography followed by sputter deposition and then lift of f. The top electrodes are patterned directly onto the dielectric insulation layer, and a titanium adhesion promoting layer is again used between the copper and dielectric surface. Following the lift off of the top electr odes, t he wafer is cleaned off a final time (acetone/methanol/DI water) which completes the fabrication process. First generation de vices are shown in Figure 3 3 with polyimide dielectric The straight configurations are known as linear DBD devices, and th e curved ones are referred to as having serpentine geometry. Arrays with 3 6 devices connected in parallel were also f abricated. Figure 3 3 First generation of fabricated microscale DBD plasma devices with polyimide dielectric material ; showing linear and serpentine electrode geometries.
47 After completing the device manufacturing, m easurements a re made to verify the dielectric layer thickness. A contact profilometer i s used to make a one dimensional line scan across the device topology. T he profilometer is a Dektak (model 150) diamond tipped stylus profilometer with sub nanometer resolution. The tip of the stylus is 12 m in diameter and applies 10 mg of force as it scans along the dielectric surface. The profile shown in Fig ure 3 4 corres ponds to a scan across an etched region in the dielectric to provide an opening to the lower electrode for electrical contacts. The scan verifies that the 10 m target thickness was achieved. Figure 3 5 illustrates several of the processing steps that were described in the fabrication of microscale DBD plasma actuators. General equipment used for manufacturing is shown along with several pictures of the wafer throughout the fabrication process. The illustrations correspond to the process for actuators havin g polyimide dielectric material as indicated in parts G and H. Figure 3 4 Profilometer measurement of SiO 2 dielectric layer topology, s can ned over bottom electrode contact pad (see insert) for device with a target dielectric thickness of 10 m
48 A B C D E F G H I J K L Figure 3 5. Images showing several of the process steps and equipmenet involved in the fabrication of microscale DBD plasma actuators (process shown for polyimide dielectric) A) photoresist spin coater, B D ) mask aligner, wafer load and alignment, E ) sputter machine F ) lift off, G ) polyimide spin coater, H ) polyimide heat cure, I J ) RIE for polyimide etch and wafer load, K ) post RIE ( etch mask removed), L ) completed devices (after patterning top / exposed el ectrodes). 3.3 Fabrication Challenges One issue that has been raised by prior microscale investigations is erosion of the electrodes. Ono et al. ( 2000 ) found that both metal and silicon electrodes undergo sputtering/erosion during device operation which decreases the breakdown threshold voltage and reduces the lifetime of the device before failure. It was shown for metal coated electrodes that the anode suffers from sputtering and craters were observed from 2 4 m in diameter under m agnified ima ging (Figure 3 6 ). Device failure is an important issue for microscale DBD devices A long with electrode sputtering during operation, intensified electric fields form at corners and sharp points of conductors wher e the current is concentrated. These local regions of concentrated electric field can breakdown the dielectric surface causing a short circuit and inevitable device failure The adhesion or bonding of the device layers is also critically important. Adhesion issues are notoriously common in the m anufacturing o f microelectronics in general. For
49 Figure 3 6. SEM image showing 2 to 4 m wide craters observed on a platinum anode due to electrode sputtering (Ono et al. 2000). titanium is used. Copper has poor adhesion to bare silicon and may separate from the silicon. A thin layer of titanium, which has excellent adhesion to silicon, is used in betwee n the copper and bare silicon. The seed layer is typically on the order of 10s of nanometers. The copper can then be sputtered onto the titan ium with better bond strength. The adhesion issue is also pertinent when cho osing the dielectric material. Polyimide, which has been used successfully in the fabrication of the first generation devices, allows for good bo nding with sputtered titanium. On the other hand, a dielectric such as PDMS, having desirable material properties (high dielectric strength, resists ozone and UV degradation, ability to deposit thick films), has extremely poor adhesion with metals being deposited onto it. Adhesion issues limit the available material set that can be utilized for possible enhanced device performance and/or device lifetime. Another fabrication challenge arises in depositing the SiO 2 dielectric barrier. The chamber is held at 300 C and following the deposition the substrate is removed into the ambient room temperature. The mismatch in the thermal coefficients of expansion
50 (TCE) of the SiO 2 and the glass substrate induces stress as th e wafer cools to room temperature In addition typical SiO 2 layers used to insulate between layers of conductive wires in semiconductor devices are on the order of angstroms to nanometers whereas several microns are used in the se DBD devices. This deposi tion takes a significant amount of time in which the substrate is subject to fairly high temperatures. Cracking was observed in the dielectric primarily over metal regions of the largest electrode features (e.g., over the large ground electrode geometry). Figure 3 7 displays images of devices with 1 mm ground electrodes where the SiO 2 has delaminated. Electrode features having dimensions of 100 m and below typically did not suffer from stress induced cracking of the SiO 2 dielectric Figure 3 7. Regions of delaminated silicon dioxide over 1 mm wide copper features.
51 CHAPTER 4 EXPERIMENTAL CHARACT ERIZATION METHODS Experimental characterization of microscale plasma devices is complicated by the fine spatial dimensions and short temporal scales of the fluidic response, as well as the small magnitude (nano to micro Newtons) of the mechanical force response. Accurate measurement of these quantities is of critical importance for improving fun damental understanding of the microscale physics, validating numerical models, and fueling new applications for these plasma devices. As such, characterization of microscale DBD actuators requires accurate data acquisition equipment along with maintaining sound engineering methods throughout the experimental setup and measurements collection. In order to use microscale DBD actuators for flow control applications, their electrical, fluid thermal and mechanical performance must be quantified, providing repea table results with minimal variation between devices The methods and equipment used for device characterization are presented in this section Figure 4 1 indicates the orientation of the actuator with respect to the coordinate axes. This orientation is us ed throughout the characterization of both velocity and force data. The x direction corresponds to the streamwise flow, the z direction to the spanwise flow, with the y The electrical voltage and current ar r consumption may be computed. The fluidic wall jet flow and the reaction force (thrust) on the device due to the induced fluid momentum must also be measured to quantify ature of the DBD actuator is also measured to assist with failure analysis. The following sections describe each of the characterization methods in detail.
52 Figure 4 1 DBD device showing orientation of x,y,z coordinate axes. 4.1 Electrical Measurements The electrical setup is shown in Figure 4 2 A sinusoidal input voltage (typically ~kHz frequency) is synthesized by a function generator (Tektronix AFG 3022B). The voltage is then amplified to the kilovolts range using a high voltage power amplifier ( Trek model 30/20 ), which is connected directly to the DBD actuator terminals to create the plasma discharge. Figure 4 2 Schematic of typical power supply and electrical measurement probes for DBD plasma generation. For the electrical charac terization, both instantaneous and average electrical quantities are of interest. T he voltage across the device terminals is measured using a high voltage probe (Tektronix P6015A), and the current entering the device is measured with a n inductive coil cur rent monitor (Pearson 2100). A digitizing oscilloscope (Tektronix DPO3014) captures both of these signals capable of sampling rates up to 2 GSa/s (giga samples per second) with a record length of 1 million points.
53 The time average power co nsumption is computed by averaging the instantaneous power over an integer number of periods using MATLAB software. Using the discrete set of sampled data captured by the oscilloscope, the time average power is found by averaging the voltage current product over the to tal number of data samples, N according to (4 1) Note that the number of samples, N should correspond to an integer number of periods of data for Equation 4 1 to be valid. An uncertainty analysis for the power measurement is provided in Appendix B The error computed using this analysis is used to propagate the error bars reported in the average power data. 4.1.1 Alternative Power Measurement Methods The current can be monit ored a variety of ways depending on the available equipment. An inductive coil is commonly used within the DBD research community to measure the alternating current flowing into the DBD actuator. Alternatively, a small resistor is sometimes used when a cur rent monitor is not available. The resistor is placed in series between the DBD actuator and ground. The resistance is typically 100 Ohms or smaller such that the voltage drop across it is much smaller than that across the actuator load. This allows for a standard voltage probe (i.e., does not require a high voltage probe) to measure the resistor voltage, and the current is then computed via The current signal is most challenging to resolve. Large spikes corresponding to conduction current are superimposed on a smaller sinusoidal component or displacement current signal. These current spikes can be more than two orders of
54 magnitude in amplitude compared with the amplitude of the displacement current, which follows the periodic frequency of the A C input. In order to capture the large current sinusoidal component of the current to become compressed greatly and results in very poor resolution of the displacement current component. Inversely, if resolution is set finer to better resolve the displacement current, the current spikes are clipped. An investigation of this tradeoff in the current signal resolution is presented in the following section. Lissajous figures are used as an alternative method to compute the average power dissipated in a DBD ( Kogelschatz 2003; Pons et al. 2005; Hoskinson et al. 2008; Hoskinson and Hershkowitz 2010; Kriegseis et al. 2011 b; Ashpis et al. 2012; Kriegseis et al. 2013 a ) Instead of plotting the voltage and current waveforms with respect to time, the signals are plotted against each other, creating a Lissajous figure. The resulting plot is elliptical shaped in which the aspect ratio of the ellipse is related to the phase shift between the two signals. In this case a small capacitor, referred as a measurement capacitor measurement capacitor (~nF) is chosen to be a few orders of magnitude larger than the capacitance of the actuator (~pF) such that the load capacitance is dominated by the plasma device. The voltage across the measurement capacitor is monitored and used to compute the charge accumulation Q on the capacitor, where Q = CV The charge voltage characteristi c curve is integrated to find the average power. In this method, the voltage across the measurement capacitor is typically noisy following the conduction
55 across the measurement capacitor is averaged while recording. The averaged signal is used in the integration of the average power, and suppresses many of the discharge events. Ashpis et al. (2012) implemented a non linear compression circuit between the DBD actuato r and ground. This circuit is used to suppress the large current spikes without attenuating the sinusoidal component in order to better resolve both the sinusoidal current as well as the current spikes simultaneously. In addition to the co mpression circuit Ashpis et al. (2012) provides a thorough overview and comparison of the different methods available for computing the power dissipated in a dielectric barrier discharge. The reader is referred to this reference for details in these power measurement meth ods. In addition, the electrical performance of DBD actuators is thoroughly investigated as reported by Kriegseis et al. (2011b). In this work, the power consumption is related to the capacitance of the discharge, the thrust force and also the streamwise e xtent of the plasma discharge region. 4.1.2 Considerations for Power Measurements There are several nuances to consider when computing the power consumed for DBD actuators. These apply to both the micro and macroscale. Variables such as the oscilloscope sampling rate, the current amplitude resolution, and the number of point or periods recorded can affect the average power value. As mentioned in the previous section, there is a trade off in resolving the components of the current signal. The current resol ution is first investigated. A microscale actuator is tested (50 100 100 m geometry, 10 m thick silicon dioxide dielectric) at 3.5 kVpp and 1 kHz input, and 100 periods of the voltage and current waveforms are downloaded for each of three current resolut ion settings. The oscilloscope is set to 20, 50 and 100 mA/division for
56 consecutive recordings ( Figure 4 3 ). The sequence of these three current resolutions is repeated four times, again recording 100 periods during each download and integrated to compute the average dissipated power. Figure 4 3. Voltage (blue) and current (green) waveforms plotted with time for different oscilloscope current resolution settings. A) 20 mA/div. B) 50 mA/div. C) 100 mA/div. The average power values corresponding to the waveforms in Figure 4 3 are reported in Figure 4 4 The average power increases when using a larger range, i.e., with decreased current resolution. The conduction current spikes are not clipped when using a larger dynamic range, and the large amplitudes of the current spikes are captured and contribute to the average power calculation. With increased resolution, the spikes are clipped more heavily and they do not contribute accurately in the average power calculation. The power v alues in blue circles correspond to the largest geometry actuator, and vary from 0.14 0.23 W, indicating that nearly half of the dissipated power is not captured when clipping the current in these data. S imilar power data is reported in Figure 4 4 for sm aller device geometry (red diamonds) The smaller device consumes considerably less power, and the variation (percentage) between the power values based on the current reso lution is even greater The current spikes correspond to the discharge events, contr ibuting to the conduction current which pertains to the real power
57 consumed. The sinusoidal component represents the displacement current that travels back and forth in and out of the load, and pertains to the reactive power due to the capacitance of the D BD actuator. While the sinusoidal component contributes some real power (amount depending on the phase between current and voltage), it is more important to capture the spikes in the current which contribute most to the real power dissipated in the DBD. Figure 4 4. Average power as a function of current waveform resolution for two microscale DBD actuators. Both devices have a 50 m wide anode, 100 m gap, with a 5 m thick silicon dioxide dielectric and are operated at 3.5 kVpp, 1 kHz. The following ex ample is discussed to further clarify these effects. There are eight divisions over the vertical range of the oscilloscope, and there are 8 bits dedicated to the vertical resolution corresponding to 256 discrete amplitude levels. A typical value for the am plitude of the sinusoidal current component for the microscale actuator is ~1 mA. At 20 mA/division resolution, the current signal achieves 8*(20mA)/2 8 = 0.625 mA resolution. This corresponds to only three or four values for which the sinusoidal current
58 wi ll be discretized. In this case, there is poor resolution over the periodic component of the current, and the spikes will be clipped at +/ 80 mA. With increased current resolution, the displacement current may be better resolved at the cost of increased a ttenuation of the conduction current signal, resulting in misleading low power values. The next variable investigated is the sampling rate, or time resolution, of the oscilloscope. This test was performed using a macroscale DBD actuator, having 5 mm wide copper strip electrodes and a 3 mm thick PMMA dielectric. The device was operated at 14 kHz and 1 million points were recorded. The sampling rate is varied from 5 MSa/s up to 250 MSa/s with eight total increments (5, 6.25, 8.33, 12.5, 25, 50 125 and 250), and the average power is computed for each case over an integer number of periods (depending on the sampling rate). The results are shown in Figure 4 5 for several input voltages. The average power has little dependency on the sampling rate, with a slight decrease in the power for the slowest sample rate (5 MSa/s). From 6.25 250 MSa/s the power values are consistent and independent of the sampling rate. Figure 4 5. Average power as a function of sa mpling rate for a macroscale actuator
59 Instead of the sampling rate, the number of periods recorded is of more importance for an accurate power measurement. The same macroscale actuator was used for this experiment. The frequency was kept at 14 kHz and the sampling rate was set to 250 MSa/sec, providing 56 p eriods of data for each recording. The data was segmented into individual periods and a running average was performed over the 56 periods. A B Figure 4 6. Influence of number of periods on average power value. A) Average power over n periods shown with standard deviation error bars. B) Standard deviation of the average power over n periods. Figure 4 6 plots the mean power as a function of the number of periods used for averaging. In Figure 4 6a the error bars correspond to the stand ard deviation in the mean power value over n periods, while in Figure 4 6b only the standard deviation is plotted over n periods. From these plots, over 10 periods are required before the mean power begins to stay within the standard deviation. Acquiring e ven more periods reduces the standard deviation and provides more accurate average power data, with a tighter confidence interval.
60 To summarize, the mean power is affected most by the resolution of the current amplitude. A sufficient number of periods shou ld be recorded and included in the mean power computation. The specific number of periods may vary based on the actuator geometry and electrical inputs, and so a running average analysis is recommended to ensure enough periods are captured. Finally, the sa mpling rate shows little influence on the power data, so long as the signals are sufficiently resolved in time. A low end figure of merit for the sampling rate is to capture at least 500 points per period. 4.2 Fluidic Measurements A spatial velocity map of the induced wall jet is of primary interest for fluidic characterization. Flow visualization techniques may be used qualitatively to view and compare the profile of the micro DBD induced flow with that of the macro DBDs. For quantitative velocity measurem ents, the fluid velocity may be measured using a number of methods. Non intrusive test methods are preferred for two primary reasons; first, the large electric field for plasma discharge requires that no metal probes be used nearby, eliminating the possibi lity for utilizing hot wire velocimetry. Second, the induced velocity occurs in close proximity to the surface making it challenging to use physical probes without disturbing the flow N on metallic pitot or stagnation probes are a possible option, although they are limited in near wall resolution due to restrictions in the probe diameter. A non intrusive method, namely two component particle imaging velocimetry (PIV), will be used to make the velocity measurements. Pitot measurements will be made to compare with the PIV results for validation purposes. 4.2.1 Particle Image Velocimetry Measurements Two component PIV measure ments provide a two dimensional cross sectional image of the flow. A PIV image of the induced flow from a macroscale DBD act uator is
61 illu strated in Fig ure 4 7 (Durscher and Roy 2012). The vectors indicate the flow is being entrained near the upper electrode and pushed along the surface in the downstream direction, revealing the profile of the induced wall jet. Using a standard macro lens in combination with typical optical teleconverters, an adjustable field of view (FOV) from ~20 60 millimeters (streamwise) may be obtained. When used with a high resolution CCD camera, it provides a spatially resolved detailed image of the induced flow for the DBD actuators. The profile of the flow field for microscale actuators follows as a scaled version of that observed with macroscale DBD actuators, where the maximum velocity occurs just above the dielectric surface. Figure 4 7 PIV data showing time averaged induced velocity flow field from a m i croscale DBD plasma actuator. The actuators are tested directly on the glass wafer substrate on which they were fabricated. A large acrylic chamber is used to prevent the induced flow from being af fected by ambient fluctuations in the laboratory creating a quiescent test environment The chamber is 0.61 m square by 1.22 m tall, and also helps to keep large amounts of ozone from circulating into the lab oratory
62 A dual cavity pulsed Nd:YAG laser (New Wave Research Solo PIV II 30) is used to generate a light sheet along the centerline of the actuator in the direction of the induced flow (normal to the span). In order to ensure that the laser sheet is centered and perpendicular to the device span, refere nce indicators were included in the fabrication design which allow the light sheet to be accurately aligned. The laser sheet is adjusted using attached on axis optics to achieve a 1 mm thick beam waist. The test chamber is seeded with flow tracing particl es. Ondina oil is used for the seeding material, which is vaporized using a TSI atomizer (Model 9302) using 25 psi of pressure. Using these settings, the atomizer produces seed particles with a mean diameter of ~0.8 m (TSI 2000) Durscher and Roy (2012) p reviously showed reasonable agreement between PIV data obtained using o ndina oil as the seeding material and pitot static measurements implying the ondina particles are negligibly affected by the electrostatic forces nea r the high voltage electrodes. A sim ilar experiment is discussed in 4.2.2, further verifying the use of ondina to test microscale DBD actuator velocities. For each PIV test, 300 image pairs are taken at a repetition rate of 7.2 Hz. The time between laser pulses ( dt) is adjusted based on th e induced velocity. The dt is set to maintain a particle displacement of 5 to 7 pixels between image pairs for optimal data correlation. A LaVision camera (ImagerPro X 4M 2048 x 2048 pixels 2 ) is used to capture the PIV images and is fitted with a 105 mm l ens In some cases a teleconverter lens (1.4x or 2x) is used in addition to the primary lens, reducing the field of view and increasing the spatial resolution and post process the PIV images. Prior to data collection, image calibration is
63 performed using a 40 mm square, two tier calibration plate. Data post processing begins by computing the average intensity of each image frame. The average intensity is then subtracted from ea ch raw image in order to increase the signal to noise ratio of the image A particle intensity correction is applied locally over a window of 3 to 5 pixels allowing for smaller particles to be more effectively included in th e image correlation (TSI 2000). The correlation is then applied using a multi grid / multi pass process. The first pass applied a 32 x 32 pixel window with 50 % overlap, followed by two refining passes using a 16 x 16 pixel window and 50 % overlap. The resulting velocity flow field has a vector resolution of 110 m for a 3 0 mm wide x 20 mm high field of view, and ~220 m for a larger field of view (46 mm x 25 mm). of the wall jet during the PIV measurements The induced flow structure begins near the edge of the exposed electrode and extends downstream, taken as the direction toward the grounded electrode (positive x direction). A schematic of the velocity measurement set up is provided in Figure 4 8 indicat ing the location of the velocity interrogation window with respect to the electrical contacts. Figure 4 8 Schematic of experimental setup for velocity measurements indicating the field of view used with reference to the location of the electrical contact points.
64 The convergence of the time averaged velocity is investigated in Fig ure 4 9 in order to determine whether 300 images provide a statistically sufficient number of image pairs. The data in Fig ure 4 9 A displays the x component of the velocity measured at x = 3 mm and y = 0.5 mm, and in Fig ure 4 9 B for x = 8 mm and y = 1.5 mm. These point are chosen to in vestigate the region in the core of the wall jet (x = 3 mm) and also the downstream diffused region of the induced flow (x = 8 mm). The velocity remains fairly constant at 3.0 and 3.5 kV pp with fluctuations (standard deviation / average velocity) within 2. 1 % of the mean velocity. At 4 kV pp the data at x = 3mm varies slightly more, within 2.4 % of the mean velocity value, although at x = 8 mm ( Fig ure 4 9 B ) the velocity variation is 1.3 %. The overall variation in the averaged velocity is within 3.0 % and p ermits confidence in the time averaged velocity measurements. A B Figure 4 9. Velocity convergence plots over 300 image pairs for the x component of velocity A) x = 3.0 mm, y = 0.5 mm. B) x = 8.0 mm, y = 1.5 mm 4.2.2 Pressure based Velocity Measuremen ts In order to validate the velocity measurements from the PIV experiments, a stagnation probe is used to collect pressure data for a second independent measurement. From the pressure measurements the velocity may be computed according to
65 (4 2) where is the density of air (kg/m 3 ), v is the fluid velocity (m/s), and p and p atm are the total and ambient pressures (Pa), respectively. The velocity is computed assuming steady and incompressible fluid properties and also neglecting viscous losses. The stagnation probe is constructed out of a modified glass pipette with an inner an d outer diameter of 1.0 mm and 1.5 mm, respectively. Pressure measurements are made using a Furness Controls FCO332 differential pressure transducer calibrated to 9 Pa with a 10 V output, providing 0.01 Pa resolution with accuracy within 0.5 %. Each da ta point represents the average of 100 voltage readings recorded at a sampling rate of 15 Hz using a National Instruments data acquisition module (PCI 6133). The differential pressure, P measurements were converted to velocities using Equation 4 2 where the air density value was taken as 1.18 k g / m 3 A two axis motorized traverse (Velmex PK268 03B ) is used to make accurate adjustments for the locations of the stagnation probe. An image of the experimental setup is shown in F igure 4 10. The probe is shown m ounted onto a two axis motorized traverse for accurate control of the measurement locations. The results from the pressure measurements will be used primarily to indicate some level of confidence in the PIV data. A B Figure 4 10 Configuration for pressure measurements. A) Top view of device under test, the probe is mounted to the traverse (in foreground ). B) S ide view.
66 4.3 Mechanical Measurements The mechanical response of a macroscale DBD actuator is typically characterized by its induced thrust ( Baughn et al. 2006; Gregory et al. 2007; Porter et al. 2007 ; Abe et al. 2008; Enloe et al. 2008; Enloe et al. 2009; Font and McLaughlin 2010; Kotsonis et al. 2011; Durscher and Roy 2012; Kriegseis et al. 2013 c ; Neumann et al. 2013). When a plasma actuator imparts momentum to a fluid, there is an equal and opposite net reaction force (thrust) acting on the device, as illustrated in Figure 4 11. Here the term thrust describes the net reaction of the actuator due to pressure, shearing, and plasma induced force s. Characterization of th ese force s will e nable verification of numerical studies of the force density acting on the fluid since the net fluidic response must be balanced by a reaction force on the DBD device less the viscous losses on the plate surface. Q uantifying the net thrust provided by the actuator is useful to identify possible aero propulsion or space thruster applications. Figure 4 11. Schematic indicating the shear and body forces acting on an actuator and the imparted plasma force produced by a DBD actuator. The thrust may be directly measured using a force balance or computed indirectly from velocity measurements. For a direct measurement, a torsional balance can be used to measure the actuator thrust in opposition to two torsion springs. Here, an optical displacement sensor is used to measure the angular displacement of a beam when the DBD actuator is turned on. For the indirect measurement the force can be extracted by applying a control volume analysis or spatial body force analysis to the measured
67 velocity data. These methods are investigated for the measurement of the DBD body force and are discussed in d etail in the following sections. 4.3.1 Direct Force Measurement The thrust produced by a macroscale DBD actuator is typically on the order of m N/m and can be measured directly using a digital force balance. Here the actuator is mounted on the balance such that the plasma is directed away from the balance such that the reaction force is applied downward onto the balance. An example of macroscale actuator thrust measurements is shown in Figure 4 12 ( Thomas et al. 2009 ). In Figure 4 12 A the thrust is plotted against input voltage for va rious dielectric materials and thicknesses. The thrust measured for Teflon is over 20 mN/m (normalized per meter length of electrodes) for both thicknesses that were tested. In Figure 4 12 B the thrust is again plotted against input voltage for various inpu t frequencies. The largest thrust measured here reaches almost 150 mN/m and show generally greater values with lower frequency. A B Figure 4 12. DBD thrust measurements (Thomas et al. 2009). A) Normalized thrust measured using a digital balance for d ifferent thickness dielectrics plotted vs. input voltage. B) Normalized thrust plotted vs. peak to peak voltage
68 However, f or microscale DBD devices, the thrust is on the order of 0.1 1 mN/m ( Zito et al. 2012 ) The electrode length is 10 mm, indicating t hat the measured force before normalizing the electrode length is on the order of 1 10 N ( 0.1 1 mg ) The resolution of typical balances used to directly measure the DBD force is 1 mg. Hence, the microscale DBD force is, at best, on the order of the ba signal. For the smaller of the microscale DBD geometries (i.e., 50 100 50 m geometry), the thrust is even smaller and cannot be measured on a typical scale balance. In addition, the thrust acts in plane with the DBD and is on th e order of micro Newtons present ing some dem anding measurement challenges. Extremely sensitive force measurement capabilities based on micro electromechanical systems (MEMS) have been developed with pico Newton force resolution ( Chandrasekharan et al. 2009; Chandrasekharan 2010 ; Heinrich and Waugh 1996 ). Capacitive (comb fingers), piezoresistive or piezoelectric means of detecting the displacement are MEMS based sensors that could be considered. The challenge with these methods of detection is t hat they all are subject to electromagnetic interference (EMI). These sensors need to be mounted in close proximity to the actuator itself, and are likely to be affected by the strong electric field surrounding the discharge. Alternatively, an optical bas ed measurement is capable of det ecting sub micron displacement ( Chen et al. 2010 ; Horowitz et al. 2004 ; Gamero Castano 2003 ) and is desirable for its immunity to EMI. This method has been implemented to measure the displacement of a torsion force balance. 220.127.116.11 Torsion b alance Direct thrust measurements are made using a custom built torsional force balance, which measures the angular deflection of a beam acting against torsion springs. The
69 balance is designed similar to that reported by Gamero Castano (2003). The actuator is mounted at the end of a beam moment arm such that the thrust displaces the balance away from an optical displacement sensor. The balance rotates upon a vertical axis defined by an aluminum beam mounted with two torsion springs (one at each end), and an aluminum moment arm deflects horizontally as the axis rotates. A schematic of the torsional force balance is shown in Figure 4 13 A B Figure 4 13 Torsiona l force balance schematic. A) Top view schematic of torsional balance showing location of DBD actuator and the optical displacement sensor. B) Side view schematic showing the horizontal beam, torsion springs and axis of rotation. The induced thrust from t he actuator produces a torque on the balance, which is related to the rotational spring constant (or torsion coefficient) of the torsion springs as well as the angle of deflection about its rotational axis, (4 3) where T is the thrust force (N) acting on the balance, l is the length (m) of the moment arm, k is the rotational spring constant (Nm/rad), and is the angle of deflection (rads) of the balance arm. The angular displacement is measured using a reflectance base d optical displacement sensor (PhilTec D63 ). The displacement sensor has 50 nm resolution when operated using a minimum of 256 averages per sample.
70 The balance is calibrated using logarithmic decrement analysis to extract the rotational spring constant fo r an underdamped system This method is based only on the reaction of the balance to an initial displacement (the displacement amount does not frequency ( o ) and mass ine rtial term ( M I ), according to (4 4) The mass inertial term ( MI ), a.k.a. moment of inertia or rotational inertia, is dependent on the geometry and materials of the balance. The natural frequency ( o ) can be extracted using the frequency of the damped oscillations ( d ) as well as the damping ratio ( ), according to (4 5) To compute the natural frequency, the damped frequency and damping ratio must first be extracted. The damped frequency ( d ) is extracted graphically using the period Similarly, the damping ratio is also extracted graphically using the relative amplitudes of the damped oscillations. Figur e 4 14 shows displacement measurements from the torsional force balance. The deflections are caused by the electrostatic force between two parallel plate circular electrodes, and measured using the optical displacement sensor. One electrode is mounted on t he horizontal beam of the balance, while the second electrode is mounted to a separate pillar that is brought in close proximity (~1 mm) to the beam using a
71 micrometer (opposite the side of the optical displacement sensor). In this case, a 120 V DC potenti al is applied across the electrodes, causing a deflection of ~9 m. Figure 4 14 Displacement measurements from torsional force balance plotted against time. The voltage is applied for about 15 20 seconds and then turned off, and the response of the b alance is given enough time to settle back to its zero position. Analysis of the transient response after the input is removed provides the data necessary to extract the natural frequency (i.e., damped frequency and damping ratio). Figure 4 15 provides a c lose up view of the transient response of the first pulse in Figure 4 14 The peaks of the first three oscillations (after the input is removed) are recorded for the analysis. Figure 4 15. Zoom used for e xtracting the balance spring constant
72 From these data, the damped frequency is computed using the time between the oscillations: (4 6) where n = 1, 2, 3, etc. are the number of oscillation peaks recorded ( n = 3 in this case). For n oscillations, there are n 1 data points provided for each calibration pulse. A total of ten pulses are recorded for statistical averaging. Next, the damping ratio ( ) is extracted using the amplitudes of the decaying oscillations via the logarithmic decrement, The logarithmic decrement is equal to the natural log of the ratio of amplitudes of the decaying oscillations, according to (4 7) where y m is m periods away from y o For m =1, the log decrement is equal to the natural log of the ratio of two successive peaks. From the logarithmic decrement, the damping ratio is calculated as (4 8) Now that the damped frequency and damping ratio have been extracted, the natural frequency of the balance may be computed according to Equation 4 5. The next step is to calculate the mass inertial term for the torsional f orce balance. The mass inertia is found by superimposi ng the mass inertia for each component of the balance about the same axis, as illustrated in Figure 4 16. The balance is made up of a vertical aluminum bar which serves as the axis of rotation, an aluminum horizontal cross bar, and a cylindrical stainless steel (SS) counterweight. The horizontal cross bar
73 and counterweight both have a center of mass which does not lie on the axis of rotation, and so the parallel axis theorem is used to compute the moment of inertial about the physical axis of rotation. The condition for using the parallel axis theorem is such that the moment of inertia before applying the theorem must be computed about an axis that is parallel to the axis of rotation. Figure 4 16. Graphical representation of the superposition of the mass inertia for each component of the torsional balance. As shown in Figure 4 16, the moment of inertia for each of the three components Table 4 1 provides the parameters of the balance used for the mass inertia calculation. The width, depth and radius parameters are measured with digital calipers with micron resolution, and the uncertainty represents the variation in the measurement at various loca tions of each component. The offset from the axis of rotation is measured with a standard ruler, and the uncertainty is within 1 mm (based on the resolution of the ruler). The mass is weighed using a digital balance (Royal Model dS3) having 1.0 gram resol ution and a 1.3 kg weight limit. The moment of inertia is calculated for each component and then summed for a total balance mass inertia equal to 0.0392 kgm 2
74 Table 4 1. Parameters for computing the force balance moment of inertia. Vertical Aluminum b ar Horizontal Aluminum bar Stainless steel Counterweight Width (m) 0.0325 0.0005 0.4580 0.0005 NA Depth (m) 0.0325 0.0005 0.0325 0.0005 NA Radius (m) NA NA 0.0192 0.0001 Offset (m) 0 0.075 0.001 0.130 0.001 Mass (kg) 0.297 0.001 1.187 0.001 0.682 0.001 MI 5.228 x 10 5 2.753 x 10 2 1.165 x 10 2 TOTAL MI = 0.0392 kgm 2 Having computed the moment of inertia and extracted the natural frequency, the rotational spring constant ( k ) can now be calculated from Equation 4 4 Finally, with knowledge of the spring constant and the measured angular displacement, the DBD force can be computed using Equation 4 3 A complete uncertainty analysis for the direct thrust measurement is provided in Appendix B. 4.3.2 Velocity based For ce Measurement In addition to a direct measurement, the net reaction thrust and/or plasma induced body force may be inferred from the velocity field (Hoskinson et al. 2008; Albrecht et al. 2011; Kotsonis et al. 2011; Durscher and Roy 2012; Kriegseis et al. 2013 c ). Two methods are investigated to do so: first, a control volume analysis is used to compute the net thrust based on the difference in momentum flux through a set of boundaries. The second method utilizes the Navier Stokes equations to solve for the body force term spatially over all points within the velocity field. The total force may then be determined through integration. The details of these two methods follow. 18.104.22.168 Control v olume a nalysis In the first method, the thrust acting on the device may be estimated via a control volume (CV) analysis using the measured velocity data. T hree recent publications (Hoskinson et al. 2008; Durscher and Roy 2012 ; Kriegseis et al. 2013 c ) have compared
75 control volume extracted thrusts with direct measurements o btained using a digital balance T he results showed relatively good agreement. The following presents results from Durscher and Roy (2012), which provides a guideline for an appropriate analysis in order to accurately compute the force. In this work, a con trol volume was applied to PIV measurements to investigate the effect of varying the height and width of the control volume on the force computed. The thrust results were found to have a strong dependence on the boundary locations of the control volume. Sh own in Figure 4 17 A is a plot of the horizontal force component calculated for three different input voltages while varying the width (W) and height (H) of the applied control volume boundaries ( Figure 4 17 B ). If the window is too short in length, the thru st is significantly over predicted (roughly doubling in value for the 15 and 19 kV pp data). The force values converge in agreement when the width has sufficient downstream extent and the height is ample to include the thickness of the induced wall jet A B Figure 4 17. Control volume analysis for computing actuator thrust (Durscher and Roy 2012). A) Horizontal component of thrust force calculated from control volume analysis plotted against varying control volume width for three different input voltages The effect of varying the height is also shown (via line styles). B) Control volume used to calculate reaction forces induced by the plasma discharge.
76 Direct thrust measurements were also made using a replica actuator to compare with the control volume a nalysis. The length of the plate over which the induced flow acts upon was varied to investigate the effect of the actuator plate length. It was found that the thrust measured converged only when the plate was sufficiently long. When the plate was shortene d such that it was just long enough to cover the ground electrode the thrust measured is shown to increase by 20%, indicating increased discrepancies with applied voltage. This increase in the measured thrust is believed to stem from having insufficient length for the flow to fully develop and not allowing saturation of the viscous drag to occur. With at least 10 cm of downstream length the thrust measurements converge. As important it is for the control volum e to accurately predict the thrust inferred from PIV data measurements, it is of equal importance to make sure the direct thrust measurements also provide accurate data. Figure 4 18 shows the comparison of the direct thrust measurements with the thrust com puted using the control volume analysis (Durscher and Roy 2012) The two methods provide similar results with the larger control volume. With the shorter CV window, the force calc ulated from the CV method is 25 % larger. The direct force measurements showed close agreement with the control volume analysis only when there is ample downstream distance included in the control volume boundary B y applying the appropriate boundaries to the microscale DBD induced flow PIV measurements, the thrust provided by the m icroscale actuators may be extracted. This method will be used to characterize the device thrust and to compare with the direct force values measured from the torsional balance.
77 Figure 4 18. Comparing horizontal thrust component from two measurement techniques (Durscher and Roy 2012). T he direct thrust measurement (red triangles) is shown to match well with the thrust calculated using the larger control volume (black squares). The thrust predicted using the narrower control volume (green diamonds) is shown to over predict the values, especially at higher voltages. As shown in Figure 4 1 9 a rectangular control volume is applied to the flow field. The individual components of the thrust, T x and T y are computed using the conservation of momentum equatio ns, assuming two dimensional flow (negligible spanwise variations), time independence (time averaged data is used) and negligible pressure gradients (constant pressure). The analytical expression for the individual x and y thrust components follow, normalized by the electrode length (units of N/m): ( 4 9 A ) ( 4 9 B )
78 In the above eq uations, the density of air, is taken as 1.18 kg/m 3 corres ponding to a temperature of 25 C and atmospheric pressure and is assumed spatially con stant throughout the analysis. Constant density is assumed in similar works (Hoskinson et al. 2008; Durscher and Roy 2012) and a form al investigation is presented by Enloe et al. (2006), where a 2 % increase in the fluid density was observed 1 mm above the actuator surface near the exposed electrode edge. Figure 4 19. Schematic view of the control volume used for extraction of thrust data from PIV measurements. To extract the thrust results from the PIV data, the integrals in Equations 4 9 are applied to the time averaged velocity data from the PIV measu rements using MATLAB software. A second order trapezoidal method is used to numerically integrate the data using the MATLAB embedded trapz function. A s shown in Equation 4 9 A the x component of thrust is a combination of the integrated plasma body force and fluidic shear, therefore by estimating the viscous shear the net plasma body force F plasma (units of N/m) may be recovered The net shear force, F shear (units of N/m) ac ting on the plate may be estimated by integrating the viscous shear component, f shear (units of N/m 2 ) given by (4 10)
79 where the dynamic viscosity of air, is taken as 1.86 x 10 5 Pa s. The first row of velocity values above the actuator surface are used for the computation of the shear force, following the same method as reported by other groups (Versailles et al. 2010; Albrecht et al. 2011; Kriegseis et al. 2013 c ). The estimation of the net plasma body force enables comparis on with predictions from numerical studies 22.214.171.124 Spatial b ody f orce e stimation In the second method, the entire (two dimensional) flow field is used to estimate the spatial plasma body force f plasma (units of N/m 3 ) This method has been reported by o ther groups ( Albrecht et al. 2011; Kotsonis et al. 2011; Kriegseis et al. 2013 c ) Compared with the control volume analysis, which only provides integrated quantities the body force may be computed at all points within the flow field After applying the aforementioned assumptions (time averaged, negligible pressure gradient, two dimensional flow), the incompressible Navier Stokes momentum equations reduce to ( 4 11 A ) ( 4 11 B ) wher e, as mentioned in the previous section, the density of air, is taken as 1.18 kg/m 3 and the dynamic viscosity of air, is taken as 1.86 x 10 5 Pa s. Equations 4 11 provide the spatial plasma body force distribution over the two dimensional flow field. The first and second spatial derivatives of the velocity are computed using the built in MATLAB gradient function. The results from this method may be compared to the results of the control volume analysis by integrating the spatial plasma body force over a prescribed area. Figure 4 2 0 plots the x component of the
80 spatial body force as computed from Equations 4 11 A for a device with 50 100 1000 m geometry and a 10 m thick silicon dioxide dielectric layer. Estimating the plasma body force spatially provides data for comparison with num erical simulations of DBD actuators which commonly compute the plasma body force acting on the fluid. Figure 4 2 0 Spatial plasma body force for a microscale DBD actuator with 50 100 1000 m geometry and a 10 m t hick SiO 2 dielectric layer, operated at 7 kVpp and 1 kHz. 126.96.36.199 Investigation of 2D f low The assumption that the plasma induced flow is primarily two dimensional i s investigated in this section. Two experimental configurations are used in this analysis. First, the actuator is mounted parallel to the laser plane to allow top view PIV measurements at different heights above the actuator surface. A single axis manual traverse (Velmex model A 1503 P40 S1.5) allows the floor of the test chamber to translate to the laser, which is fixed in position to keep focus with the camera throughout the experiment. Figure 4 2 1 shows a top view velocity contour plot taken just above the surface of the actuator (y = 0.5 mm) for a device with 50 100 1000 m geometry and a
81 10 m thick silicon dioxide dielectric, operated at 6 kVpp and 1 kHz. An image of the device is superimposed over the velocity contour to provide the relative location of the DBD actuator with respect to the flow field. Figure 4 2 1 Top view velocity contour, shown with device overlaid to indicate its location relative to the flow field. Displaying x z plane at y = 0.5 mm for actuator with 50 100 1000 m geometry and a 10 m thick dielectric layer, operated at 6 kVpp and 1 kHz. Over the span of the actuator the flow is primarily in the x direction. Near the edge of the electrodes (z = 5 mm), the z component of the velocity becomes significant. However, over the range of z from 4 to 4, corresponding to the spanwise length of the actuator neglecting the edges (indicated with dotted white lines in Figure 4 2 1 ), the z component is minimal. In the second experiment PIV measurements are taken across the span of the actuator The device for this test had 50 100 100 m geometry and a 10 m thick silicon dioxide dielectric, operated at 4 kVpp and 1 kHz. Starting in the center of the actuator (z = 0 mm), measurements are repeated at z = 2, 4 and 6 mm. Recall the actuator is 10 m m in span, ranging in z from 5 to + 5 mm. The data at 0, 2 and 4 mm lie within the plasma generation region, where at z = 6 mm the laser plane is 1 mm past
82 contribute in t his region. Figure 4 2 2 Variation of the velocity contour along the span (z direction) of the microscale actuator. Bottom right insert shows top view; dashed lines indicate z locations of four spanwise velocity contours. Figure 4 2 2 displays the results from the spanwise tests. The insert in the bottom right indicates the locations corresponding to the four contours shown. The velocity fields show good agreement among the flow from z = 0 to 4 mm. By z = 6 mm the plane of illumination is beyon d the end of the electrodes indicating reduced velocity values at the electrode ends. The top view velocity contour in addition to these spanwise varying contours provides some justification for the two dimensional flow assumption. 188.8.131.52 Investigation o f f luid c ontinuity The conservation of linear momentum is investigated next. This analysis is performed on the velocity data from the device used in the previous investigation on 2D flow (50 100 100 m geometry with 10 m thick SiO 2 dielectric 4 kVpp, 1 kHz) For a
83 steady state flow, the continuity equation states that mass entering the system must be equivalent to that leaving the system. In differential form the continuity equation is written as (4 12) Assuming incompressi bility Equation 4 12 reduces to ( 4 13 ) indicating that the divergence of velocity is exactly zero at every point within the flow field. For an ideal 2D incompressible flow, the z component of velocity is zero and Equation 4 1 3 holds. In reality there is some non zero value to Equation 4 13. This non zero value can be attributed to one, or more likely a combination, of three conditions: (1) the flow is 3D, (2) the flow is compressible, or (3) velocity measurement uncertainty. A ssuming incompressible flow and negligible velocity error, Equation 4 13 can be rewritten to account for the unknown velocity component, as (4 14) which implies that the non zero value of the 2D convergence may be attributed to the change in the z component of velocity across the thickness of the laser sheet. Figure 4 2 3 plots the divergence of velocity corresponding to two planes along the span of the device. In the local discharge region, the divergence values are largest as the velocity changes most rapidly. For z = 0 mm, the profile corresponding to the middle of the electrode span, the divergence values range in magnitude from 0 to 20 s 1 in the downstream region corresponding to the wall jet. From Equation 4 14 and following the logic that this value represents the unknown velocity component across the laser sheet,
84 t he change in velocity across the thickness of the laser is computed by multiplying this value by the thickness of the laser plane (0.5 mm). This indicates an estimated 0.01 m/s change in the z component of velocity across the laser sheet The primary or x component of the induced velocity reaches 0.24 m/s in this case, such that the z component is 4 % in comparison. At z = 6 mm, end effects are captured showing an increase in the divergence values where the flow is expected to become three dimensional. Near the electrode edges, the divergence values are more than double that compared with the profile at z = 0 mm, reaching 60 s 1 in the wall jet region. The plasma induced flow is treated as two dimensional with support from these analyses. Figure 4 2 3 Di vergence of velocity for device with 50 100 100 m geometry and 10 m thick silicon oxide dielectric, operated at 4 kVpp and 1 kHz Divergence shown for profiles at z = 0 mm (through middle of actuator) and at z = 6 mm (near electrode edge). These assumpt ions are made to provide a mathematical representation to model and extract force information from the fluid data. They assume ideal conditions, though in reality the conditions of spatially and temporally constant pressure (and density) as well as the 2D flow assumption do not necessarily hold. As mentioned in section 184.108.40.206, Enloe et al. (2006) reported a 2 % increase in the fluid density in the discharge region. The plasma induced force term arises from complex electrohydrodynamic (and
85 magnetic) interac tions within the discharge, and the local pressure near the plasma region does not remain constant. In regions away from the discharge, which includes the majority of the flow field, the pressure gradient may be considered negligible for analysis. It shoul d be noted that the dynamic viscosity and density term s are most likely not constant, at least inside of the plasma region. The viscosity term is used to estimate the spatial body force and also the shear loss along the actuator surface. Away from critical pressure and temperature, the viscosi ty is dependent on temperature, as inferred from may not be the case locally near the discharge Based on the maximum observed increase in the surface temperature of the DBD actuators from infrared measurements (discussed in the S ection 4.4 bound for the viscosity term. An 8% increase was computed for the dynamic viscosity, leading t o a 4.8% increase in the plasma body force values based on Equations 4 11 and an 8% increase in shear force from Equation 4 10 The infrared measurements provide the actuator surface temperature which is not the same temperature as inside the plasma and su rrounding fluid. These values are discussed to provide a general range of the uncertainty associated with the viscosity term. However, since plasma is not an ideal gas and the temperature of the fluid is unknown, these values are not used for error bars in the plasma body force data. 4.4 Thermal Measurements down. There are several causes of device failure, including sputtering/erosion of the dielectric due to areas of conc
86 field strength, and excessive thermal heating due to poor thermal conductivity of the compare the performance of t he different dielectric materials used as the barrier layer. Surface measurements are made using a FLIR A325 infrared camera and accompanying FLIR ExaminIR Max software. The camera has a 14 bit temperature resolution and a spatial resolution of 320 x 240 pixels, and operates in the far infrared spectrum ( = 7.5 to 13 m). The camera can be operated in either of two temperature ranges: 20 to 120 C or 0 to 350 C. The accuracy in measured temperatures is the larger of 2 C or 2 %. The software is capable of capturing either single images or continuous recordings up to 30 frames per second. The emissivity of the surface that is being measured must be taken into account in nergy. This affects the temperature measured by the infrared camera; in general, the more reflective a material is, the lower its emissivity value. A perfect black body would have an emissivity of 1 (100% absorbance, 0% reflected radiation); all real mater ials have an emissivity between 0 and 1. In addition to emissivity, other parameters that must be taken into account for an accurate temperature reading include the atmospheric temperature and relative humidity, distance between the camera and object, and temperature of the camera optics. The emissivity of the dielectric material is computed using a hot plate and two surface thermocouples (type J). The dielectric is heated to a steady state temperature and the two thermocouples are placed on the surface in close proximity. The thermocouples are read using a National Instruments thermocouple module (model
87 USB 9211). Within the FLIR thermal software, a point is chosen on the dielectric surface between the two thermocouples and the emissivity is adjusted until the camera value matches with the thermocouples. The resulting emissivity values are 0.95 0.04 for polyimide 0.92 0.05 for silicon dioxide. Figure 4 2 4 The image is the final frame recor ded during the two minute long data acquisition for a device with 5 m thick SiO 2 dielectric (50 100 1000 m geometry, 2.5 kVpp at 1 kHz input). The convergence of the surface temperature is investigated in order to determine the length of time required f or the actuator surface temperature to reach steady state. Figure 4 2 5 shows the thermal evolution of two different actuators. The location at which the temperature is extracted is in the central region of the plasma discharge where the Figure 4 24. Infrared image of dielectric surface temperature taken at t = 120 seconds. 5 m SiO 2 dielectric, 50 100 1000 m geometry, 2.5kVpp, 1kHz input. maximum temperatures are observed (region indicated by the data cursor in Figure 4 2 4 above). In Figure 4 2 5 A the actuator has a 1000 m wide ground electrode and is operated at 2.5 kVpp, while in Figure 4 2 5 B the actuator has a 100 m wide ground electrode and is operated at 4 kVpp. All other geometries are equivalent between the
88 two actuators (50 m wide powered electrode, 100 m gap, 5 m thick SiO 2 dielectric layer, 1 kHz frequency). After ~ 80 seconds the dielectric temperature reaches a quasi steady state. A B Figure 4 2 5 Evolution of the surface temperature of the DBD dielectric layer duri ng two minutes of operation. Both devices have a 5 m thick silicon dioxide dielectric. The actuator is turned on at 5 seconds into the recording. A) 50 100 1000 m geometry with 2.5kVpp, 1kHz input. B) 50 100 100 m geometry with 4kVpp, 1 kHz input.
89 CHAPTER 5 EXPERIMENTAL RESULTS : POLYMER DIELECTRICS First generation devices have been fabricated using microfabrication techniques and successfully demon strated their basic operation. The devices utilize a 1 0 m thick dielectric material made from a low stress, high dielectric strength polyimide (HD MicroSystems PI 2611) having a relative permittivity of 2.9 (HD MicroSystems 2008) The ground electrode was allowed to extend up to 5 mm in order to test the effects of the device performance with varying e lectrode width. All of the devices have a 100 m gap between the electrodes while the anode and ground electrodes are varied in size A sample of m icroscale DBD actuators used for these tests is shown in Figure 5 1 The devices were designed specifically for PIV measurement compatibility; the electrodes are extended to one side (upstream from the induced flow) such that they would not block any of the field of view of interest for the velocity measurements. Figu re 5 1. First generation of microscale DBD actuators designed for device testing and characterization. The nomenclature regarding the three numbers in the images indicates the exposed electrode width electrode gap and ground electrode width in microns respectively. The following results are reported for devices primarily having 10 m thick polyimide dielectric barriers. However, for the direct thrust measurements reported here, another dielectric material, SU 8, is used in addition to the polyimide ba sed DBD
90 actuators. SU 8 was investigated for its ability to deposit thick layers that are also photodefinable. This boils down to a more simpl ified fabrication process: a single spin coat achieves a 20 m thick layer, and SU 8 does not require the dry etch step to expose the ground electrode contacts. The power, velocity, and force results are presented for polyimide, with the additional data from the SU 8 devices only reported in the force results using the torsional balance. 5.1 Power Consumption For elec trical measurements, the oscilloscope captures the voltage and current waveforms at a sampling rate of 100 MSa/sec (million samples per second). The devices are excited at a frequency of 1 kHz with sinusoidal input voltage ranging from 1 5 kV pp (peak to peak voltage amplitude). One million data samples are recorded providing a total of 10 periods of data. The waveforms are downloaded from the oscilloscope 10 consecutive times, providing a total of 100 periods over which the power is averaged using Equati on 4 1. This computation provides the amount of real power that is consumed by the device (as opposed to the reactive power that is reflected back tow ard the supply at the actuator load ). Figure 5 2 reveals the power consumed for microscale DBD devices w ith varying electrode widths operated with sinusoidal input at 1 kHz. Specifically the power per unit electrode length is plotted as a function of peak to peak voltage amplitude. The power consumption shows little dependency on the exposed electrode width. However, slight differences are observed for the grounded electrode width; a wider ground electrode slightly increases the power dissipation. The ability to dissipate power can be related to the area of the ground electrode. This was shown indirectly by E nloe et al. ( 2004b ) where the maximum induced flow velocity was limited by the area of the grounded
91 electrode, but no longer increased above some saturation voltage (dependent on device geometry) when the ground electrode width was extended. Figure 5 2 Average normalized power consumed for four microscale DBD device geometries plotted against applied voltage The frequency is held cons tant at 1 kHz in all cases. The nomenclature indicates the width of the powered / grounded electrode, respectively. The data in Figure 5 2 is fit with a power law curve to examine the dependency of power on the applied voltage. The fit lines indicate the power scales exponentially with the voltage, with the exponent ranging from 3.1 to 3.5 in these data. This matches cl ose with the exponential dependency observed for macroscale actuators for thin dielectrics shown by Enloe et al. (2004a) as Note that for devices with thick dielectric layers, the exponential dependency is observed as where V o is a threshold voltage and 2 < n < 3 ( Jolibois and Moreau 2009 ) considered to be at least 1 mm in these results.
92 The average power consumed (per unit length of e lectrode) reaches 20 W/m at 5 kVpp and 1 kHz. In contrast, m acroscale plasma devices often consume over 100 W/m or more at 25 kVpp (depending on geometry and frequency of operation). The microscale devices offer a substantial reduction in the power requirement ; about an order of magnitude in this case 5.2 Velocity Data Throughout the PIV measurements, the actuators are operated using 5 kV pp input voltage at a frequency of 1 kHz. The goal of these experiments was to investigate the effect of geometric variations on the velocity, and to de monstrate the ability for microscale DBD actuators to induce a wall jet similar to that induced from macroscale actuators. Time averaged PIV data is shown in Fig ure 5 3 for six different microscale DBD actuator geometries indicating the microscale devices do in fact induce a wall jet similar in profile to those induced using macroscale actuators In the left column, the powered electrode is 10 m wide while the actuators in the right column have a 50 m wide powered electrode The ground electrode width v aries with each column. The materials and dimensions of all other features (100 m electrode gap, 10 m polyimide dielectric) are the same as they were batch fabricated together on the same substrate. The induced flow show very similar flow profiles; however the device with the narrower exposed electrode results in consistently greater induced velocit ies The maximum velocity from the actuators with the 10 m wide electrode (2.0 m/s) is nearly double that induced from the actuators with the 50 m wide electrode (1.1 m/s) This trend support s the findings of Hoskinson et al. (2008 ; 2010 ) who showed increased momentum transfer for narrower exposed electrodes by comparing electrodes made of various gauge wires. The in duce d flow is surprising strong, indicating maximum velocities of 2
93 m/s, which is on the order of the macroscale devices ( Roth et al. 2000; Roth 2003; Enloe et al. 2004b; Pons et al. 2005; Roth and Dai 2006; Forte et al. 2007; Moreau 2007; Abe et al. 2008; Jol ibois and Moreau 2009; Thomas et al. 2009; Corke et al. 2010; Neumann et al. 2013 ) Figure 5 3 Velocity contour plots for microscale DBD actuators A) Devices having 10 m powered electrode B) Devices having 50 m powered electrode The ground electr ode varies from top to bottom. All devices operated at 5 kVpp, 1 kHz and have a 10 m thick polyimide dielectric. Velocity p rofiles are shown in Figure 5 4 to help compare the performance between the different actuator geometries. The profiles in Figure 5 4 A correspond the
94 velocity contours in the left column of Figure 5 3, for actuators having a 10 m wide powered electrode Similarly, the profiles in Figure 5 4 B corresponds to the velocity contours shown in the right column in Figure 5 3 for the devices having a 50 m wide electrode The induced velocity increases with the ground electrode size. This follows with the size of the discharge region as well, and as expected a larger plasma region provides a stronger fluidic impact. The thick ness of the wall jet is approximately 1 mm for the three devices with a 10 m wide electrode and slightly thicker for the three devices with a 50 m wide electrode geometry. A B Figure 5 4. Velocity profiles of two microscale DBD actuator geometries h aving a 10 m thick polyimide dielectric taken at x = 3mm downstream. A ) 10 m wide electrode B) 50 m wide electrode Input: 5 kVpp 1 kHz. Velocity profiles are plotted in Figure 5 5 comparing devices with different top electrode sizes. In Figure 5 5 A the actuators all have a 1 mm wide ground, while in Figure 5 5 B the actuators have a 500 m wide ground. Similar trends are observed here: the induced velocities are greatest when using larger ground electrodes and narrower top electrodes. The wall jet thi ckness is also about 1 mm for the devices with a
95 1 mm wide ground electrode, and slightly thicker for the devices with a 5 mm wide ground electrode. A B Figure 5 5 Velocity profiles of two microscale DBD actuator geometries having a 10 m thick polyimide dielectric taken at x = 3mm downstream. A ) 1000 m wide ground electrode B) 500 m ground Input: 5 kVpp 1 kHz. Maximum velocity values are extracted from the PIV data and reported in Figure 5 6 for several microscale DBD actuator geometries. These data indicate the maximum streamwise component of the induced velocity over the entire flow field. These plots are useful since the velocity profiles are extracted at a specific downstream location and may not capture the absolute max velocity. The b lue circles correspond to the PIV data from the devices in the left column of Figure 5 3 having a 10 m wide top electrode and the green squares correspond to the PIV data in the right column with a 50 m wide electrode Notice in Figure 5 6 that the maxi mum velocity only slightly increases when the ground electrode size is extended from 1 mm to 5 mm. In this situation, the size of the plasma discharge region does not extend over the entire 5 mm ground electrode. There is a limit to how far the ground elec trode can effectively increase the streamwise extent
96 of the discharge. This was also observed with macroscale DBD actuators as reported by Moreau (2007), where the maximum induced velocity plateaus when the ground electrode width reaches about 20 mm, corre sponding to the limit of the streamwise plasma extent. Figure 5 6 Plot of maximum velocity (x component) as a function of ground electrode width for various top electrode geometries. Devices have 10 m thick polyimide, operated at 5 kVpp, 1 kHz. For anode sizes of 10 m and 50 m, t he induced velocit ies reach ~2.0 m/s and 1.1 m/s, respectively, for both 1 mm and 5 mm ground electrodes. T he device with the larger ground (5 mm) extends the streamwise length over which the actuators can demonstrate fluidic impact This may be attributed to the increase in power consumption for devices with wider ground electrodes. This is best observes using velocity profile data, which is presented in Figure 5 7. The velocity profiles are used to compare actuators w ith 1 mm vs. 5 mm grounds for both 10 m and 50 m top electrode sizes. In Figure 5 7 A the profiles are extracted at x = 2 mm, through the core of the induced wall jet. It is shown that the profiles are nearly id e ntical between the 1mm and 5 mm ground ele ctrodes in this region near close to the discharge region. However, at 10 mm
97 downstream there are differences in the induced velocity profiles (Figure 5 7 B ) The effect of the larger ground electrode is observed downstream, as the actuator with larger grou nd geometry indicates stronger fluidic impact. A B Figure 5 7 Velocity profiles for two microscale DBD actuator s having a 10 m thick polyimide dielectric Comparing 1 mm and 5 mm ground electrodes. Input: 5 kVpp 1 kHz. A) At x = 2 mm downstream B ) At x = 10 mm downstream. To validate the results of the PIV data, pressure measurements were m ade 5 mm downstream from the exposed e lectrode using a stagnation probe. The horizontal component of the induced velocity is computed from the differential press ure using Equation 4 2 The first measurement is made with the probe just touching the surface of the substrate. The center of the probe diameter is 0.75 mm from the bottom of the probe, which limits the near wall resolution that can be measured. Each addi tional measurement is shifted vertically by 0.25 mm using the motorized traverse for precision increments. Figure 5 8 compares the results from the pressure based velocity measurements with an extracted profile of the PIV data at the same downstream locati on ( x = 5 mm). The stagnation probe measurements show good agreement with the PIV data, with slight discrepancy in the vertical location. The maximum velocity at
98 this downstream location match well between data, indicating 0.8 m/s, and both data show simil ar flow profiles. At ~3 mm above the actuator surface the horizontal velocity component is reduced to nearly zero velocity. The offset between the PIV and stagnation probe data is likely attributed to the 1 mm diameter over which the pressure measurement i s averaged. Error in the initial manual alignment of the probe may also contribute to the offset in the vertical location of the data points. A B Figure 5 8 Pressure based velocity measurement. A ) Stagnation probe measurements at x = 5 mm downstream compared with extracted PIV profile for a DBD actuator with 1 0 m wide powered electrode and 1 mm grounded electrode. B) PIV data shown indicating the downstream location of the extracted profile. The transducer used to measure the differential pressure has resolution of 0.01 Pa. From Equation 4 2 this corresponds to a minimum detectable velocity of 0.13 m/s (for = 1.18 kg/m 3 ). For devices operated at lower voltages or with smaller geometries, the induced velocities c an be well below 0.1 m/s. Hence the pressure based velocity measurement is no longer feasible. It does provide, however, confidence in the PIV measurements.
99 5.3 Thrust and Plasma Force Results Results from both direct and indirect force measurements are pr esented in this section. Recall from section 220.127.116.11, for the direct thrust measurement, an optical displacement sensor is used to measure the angular displacement of a torsion balance. The thrust acts in opposition to two torsion springs causing the displ acement of the torsion beam. Thrust is then computed using the balance s spring constant. In addition, the thrust and plasma body force is extracted from PIV measurements in order to compare with the direct force measurements. 5.3.1 Direct Thrust Measurem ents Measurements from the torsional balance are reported in this section. Calibration is performed before and after each actuator is tested in order to verify the spring constant of the system with the device in situ. As mentioned in the introduction to t his chapter, actuators with two different dielectric materials are used in these tests. One set of microscale DBD actuators uses a 10 m thick polyimide dielectric, while the other set uses a 20 m thick SU 8 polymer layer. From the direct thrust thrust is measured in opposition to the plasma force. S hear losses due to friction alo ng the surface of the actuator are inherently captured in the direct thrust measurement. Therefore, the plasma body force cannot be retrieved from these tests M easurements from the torsion balance are plotted in Figure 5 9. These data correspond to a device with 10 100 1000 m geometry and has a 20 m thick SU 8 dielectric barrier. The plots show the optically measured displacement as a function of time for several input voltages. Each voltage is applied for approximately 10 seconds before the actuator is switched off for an additional 10 seconds.
100 Figure 5 9. Displacement measurements from the torsional balance for a microscale DBD actuator hav ing a 20 m thick SU 8 dielectric. The voltages and corresponding thrust force are indicated for each discharge cycle. The displacement data is measured for several microscale DBD actuators with both polyimide and SU 8 dielectric materials. The thrust is t hen computed from the angular displacement values according to Equation 4 3 The t hrust data is presented in Figure 5 10 for actuators with both dielectric materials and two geometries The devices have either 10 m polyimide or 20 m SU 8, and a 10 m or 1 mm ground electrode. These devices all have a 10 m wide powered electrode and 100 m wide electrode gap. The observed trends are as expected, in which devices with 1 mm wide ground electrodes produce more thrust at a given voltage than those with 10 m wide ground electrodes. In addition, the devices with polyimide dielectric (blue data) show consistently larger thrust values than from the devices with SU 8 (red data).
101 Figure 5 10 Thrust values measured for four microscale DBD actuators Two device geometries (10 m and 1 mm ground widths) and two dielectric materials (10 m polyimide and 20 m SU 8) are compared. Similar to the power dependency with applied voltage, the thrust follows a similar power relationship with the input voltage. Erro r bars are shown in the thrust data based on an analysis performed to estimate the uncertainty in the thrust measurement from the torsional balance. The error bars are significant is size due to conservative values chosen in the uncertainty analysis, speci fically regarding the mass inertia of the balance which is used to compute the spring constant (Equation 4 4). The thrust also shows a power dependency upon input voltage. For the polyimide actuators, = 6.5 and 3.4 for the larger and smaller geometries, respectively; for the devices with SU 3.5 and 2.2 for the larger and smaller geometry, respectively. These data show similar, larger and smaller exponential dependencies on input voltage as compared with macroscale DBD actuators, with va riations based upon both geometry and actuator material
102 5.3.2 Velocity based Force Measurements Following the methods described in section 4.3.2, the thrust and plasma body force terms are extracted from PIV data and reported in this section. The PIV res ults reported in S ection 5.2 are used for this analysis which were operated at 5 kVpp and 1 kHz First, the x and y components of thrust force are presented in Figure 5 11. The blue circles and green squares correspond to the devices with 10 m wide and 50 m wide top electrodes respectively. The thrust is plotted as a function of the ground electrode geometry. Parallel thrust values reach up to 3.5 mN/m, corresponding to the device with the largest ground elec trode and most narrow anode. The wall normal thrust (y compon ent) reaches 0.7 mN/m, about 20 % of the primary thrust component. The thrust data follows the same geometrical trends as observed for the induced velocity val ues: a large ground electrode and narrow anode provide the best performance. A B Figure 5 11 Thrust values computed based on the control volume analysis for devices with 10 m polyimide dielectric A) x component B) y component The plasma body force was also of interest in order to compare with the body force predicted in numeri cal works. Recall that the plasma force is estimated by integrating the spatial body force and also from a control volume analysis includ ing the
103 shear force. The x and y component of the plasma force results are plotted in Figure 5 12 for both methods. Th ere is significant difference in the values as computed from the two methods, specifically for the data corresponding to the actuators having a 10 m top electrode as the ground geometries increase. This is observed for both components of the plasma body f orce. The values match closer between the three devices having 50 m electrode geometry. Furthermore, from these data there is not consistency between which method provides larger or smaller force values. The three devices with 10 m wide anodes consistent ly show larger plasma force from the CV analysis, while for the devices with 50 m wide top electrodes the integrated spatial body force provides larger values. A B Figure 5 12 Plasma force computed from the spatial body force estimation for devices with 10 m polyimide dielectric. A ) x component B ) y component. Using both methods provides a reasonable estimate of the plasma force; however the method of integrating the spatia l body force is believed to provide a better plasma force estimate. The argument for this is based on the thrust as it varies dependent upon the location of the downstream boundary chosen for the CV analysis. As shown by Durscher and Roy (2012), if the do wnstream boundary of the control volume is chosen
104 close to the discharge region, the thrust values are significantly larger than if the boundary is chosen downstream where the induced flow is diffused. Similarly, Hoskinson et al. (2008) reported that the f orce predicted using a control volume analys is was reduced by as much as 47 % when the downstream location of the CV boundary was doubled from 8 mm to 16 mm downstream. These smaller force values provided much better agreement between their control volume a nalysis results and the direct force measurements. Incorporating the full downstream region where the flow acts allows a larger region for frictional losses to occur. These losses are significant in both microscale and macroscale DBD actuators. Integrating the spatial body force over the entire flow field incorporates all of the data from the induced flow in order to compute the plasma force. In opposition, the control volume analysis only uses data at the boundaries of the CV. If, for example, the downstre am boundary is chosen very far away from the actuator (approaching infinity), the amount of force computed at that boundary will approach zero. The downstream boundary of the CV must be carefully chosen to capture the induced flow before it completely diff uses, but substantially downstream such that the thrust is not grossly over predicted. Integrating the spatial body force is not as sensitive to the downstream boundary for the integrated flow field: if the downstream boundary is far from the actuator it d oes not reduce the plasma force since the data in the region of the induced wall jet is still included in the integration. Including the entire flow field in the plasma force estimate is believe to provide a more accurate estimation since fluid events at a ll points in the flow field contribute to the plasma net body force.
105 A comparison of thrust measurement s is also warranted between the torsional balance and CV analysis The device used for this comparison has 10 100 1000 m geometry and a 10 m thick pol yimide dielectric barrier. T he measured thrust from the torsional balance reached 3 mN/m at 5 kVpp, 1 kHz input The CV based estimate of the thrust from PIV data was 3.5 mN/m for the same actuator and input parameters There is reasonable agreement in the parallel thrust component between the direct force measurement and control volume analysis. A few points should be noted regarding the plasma force as computed using the CV analysis. In this method the shear force is computed in order to reconstruct the net body force. The near wall region of the flow field is important to resolve in order to correctly estimate the shear force, providing a challenging measurement. In addition, the cross correlation analysis used in post processing the PIV data suffers mos t in accuracy at any boundary. This is due to there being fewer particles to correlate with at a boundary since there is no fluid inside of these physical boundaries. For example, at the bottom wall, there are no seed particles below the wall surface that can be used to correlate with. Furthermore, the substrate upon which the microscale DBD actuators are built has a non level surface. There is a slight bow to the wafer that is induced from several thermal cycles that are performed during the fabrication process. The non level substrate adds further uncertainty to the accuracy of the first row of velocity data which can affect the value of the net shear force. Hence, it is useful to also compute the plasma force from the method of integrating the spatial body force for co mparison with the CV analysis.
106 5.4 Thermal Results Thermal measurements of the dielectric surface temperature are reported in this section. In these experiments, two minutes of data were recorded for each test case at a frame rate of 2 Hz, providing a tot al of 240 frames. The actuator is kept at a distance of 0.254 meters (10 inches) from the infrared camera. The discharge is turned on after 5 seconds (about the 10 th frame) and the evolution of the surface temperature is recorded. These measurements were taken for three actuator geometries, all having 10 m thick polyimide for the dielectric barrier. The average surface temperature is plotted in Figures 5 13 Devices with polyimide dielectric material showed a slight temperature increase: = 14 C for th e larger geometry and = 2 C for the smallest geometry. The overall surface temperature reaches just 38 C (largest device geometry) and does not exceed 30 C for either of the smaller devices. The surface temperature of the dielectric shows a consisten t trend with the device geometry, with larger geometries reaching higher temperatures. The extent of the plasma region is at least 10 times larger for the actuator with the 1000 m ground electrode compared with the smaller device geometries. The surface temperature was investigated as whether it is the cause of dielectric breakdown when the actuator fails. Soda lime glass, the substrate material, has a glass transition temperature above 500 C. PI 2611 polyimide has a glass transition temperature of 360 C (HD MicroSystems 2008). Hence, with surface temperatures reaching just 38 C, local surface heating is ruled out as the primary cause of dielectric breakdown. Further investigation is required to determine the primary cause of actuator failure.
107 Figure 5 13 Average temperature increase plotted against applied voltage for devices with different electrode geometries. Actuators have a 10 m thick polyimide dielectric layer. 5.5 Failure Analysis These first generation microscale DBD actuators revealed a short lifetime of operation before they failed. Breakdown of the dielectric material occurred within a few minutes of operation, and sometimes within just seconds of initiating discharge. In order to get the maximum performance from these devices, they were test at relatively high voltage levels since they did not generally last long enough to test at several voltages. For example, polyimide has a dielectric strength of at least 2 MV/cm (HD MicroSys tems 2008), which corresponds to 2 kV (4 kVpp) for a thickness of 10 m. Correspondingly the devices with polyimide were operated at 5 kVpp in most of these tests. Using lower voltages slightly increased the device lifetime before failure, however the per formance (induced velocity and thrust) significantly diminishes with reduced input voltages. In addition, the actuators still failed after a few minutes of operation when lower voltages were used.
108 A visual investigation was performed regarding the cause of device failure since t he results of the surface temperature tests suggest that local heating of the polyimide is not a primary contributing factor. Both optical and scanning electron microscope (SEM) images were used to analyze the dielectric topology. Regions of polyimide are examined from actuators that were and were not operated and also regions where failure has occurred. A progression of the general failure investigation is illustrated in Figure 5 14 Optical images of the actuator before and aft er failure are shown in Figure s 5 14 A and 5 14 B respectively. Figure 5 14 C provides an SEM image of the device surface in the region where the discharge has occurred, with a red box highlighting the location where There is notic eable erosion of the polyimide, especially near the front edge of the anode from where the discharge originates. This erosion is due to the collisional processes that sustain the discharge, which constantly bombard the dielectric surface causing material t o sputter from the surface. Eventually, the polymer punches entirely through the dielectric, creating a low resistance path for current to travel. The failed region is magni fied in Fig ure 5 14 D ; the polymer around the point of failure becomes charred due to the increase in current which occurs almost instantly after the barrier fails. The topology of the polyimide was examined in downstream regions further away from the edge of the anode. An SEM image of a microscale DBD actuator that has been operated is presented in Figure 5 15 A The anode is shown horizontally across the image, with the downstream direction toward the bottom of the image. The highlighted
109 A B C D Fi gure 5 14. Polyimide failure analysis. A B) Optical images of device during operation and after filure. C) SEM image of faied device, burn out region highlighted with red box. D) Close up SEM image of burn out region at front edge of the powered electrode regions in Figure 5 15 A correspond to the approximate locations of the images shown in Figures 5 15 B E moving further downstream with each successive image. Observe that the density of these markings as well as their apparent depth is greatest near the anode, and diminishes downstream. Away from the discharge region, the surface of the polyimide looks fairly clean of markings and uniform in color (Figure 5 15 E ). It is believed these markings are a result of erosion due to sputtering of the dielectric fro m ion bombardment resulting from the collisional processes that sustains the plasma discharge. A second analysis was performed using an optical microscope for comparison Neighboring DBD actuators from the same substrate are used to compare devices that
110 Figure 5 15. SEM Images of polyimide dielectric surface after discharge. A) View of powered electrode (horizontal near top of image), the area below the powered electrode is the primary region of discharge. The four colored boxes estimate the locations of images B E), starting near the electrode edge and moving downstream. were operated to devices that were nev er operated at all. Figure 5 16A shows the region where the actuator failed (between the two electrodes), while Figure 15 6 B shows a region of p olyimide near the edge of the ground electrode. In both figures there are clear markings on the surface of the dielectric corresponding to the region where the discharge is sustained. Figure 15 6 C shows the neighbor DBD actuator that has never been excited for plasma discharge. The polyimide is clean, clearly indicating the markings on the surface were made during to the discharge process. In addition to dielectric breakdown, the electrode material is subject to erosion due to the same sputtering process th at erodes the dielectric surface. Figure 5 17 A shows optical images of a microscale DBD actuator in which the anode has sputtered away from the dielectric surface. A close up view of the sputtered anode region is shown in Figure 5 17 B ; a faint outline of the electrode is left from where it has detached.
111 Figure 5 16. Optical inspection of DBD actutaors at 10x magnification. A B) Markings in polymer dielectric visible aroundpowered electrode after discharge. C) Unused devices have clean polymer surface. In Figure 5 17 C a region of the powered electrode is further magnified and pieces of re deposited metal can be observed on the dielectric surface. Recall from section 3.3, sputtering of the electrode material was also reported by Ono et al. (2000). Figure 5 17. Failed microscale DBD actuator due to sputtering of the metal electrode from the polyimide surface.
112 CHAPTER 6 EXPERIMENTAL RESULTS : CERAMIC DIELECTRICS Results are presented in this chapter for the actuators having silicon dioxide for the dielectric barrier. Changing the dielectric material from polymer to ceramic improved the lifetime of the microscale actuators significantly. This reduced the possibilit y of devices failing during the experiment and enabled testing to be conducted over much wider range s of voltages or frequencies. Two dielectric thicknesses (5 m and 10 m) and f our actuator geometries are focused upon in these tests; all having a 50 m w ide anode and 100 m wide electrode separation. The four geometries vary in the ground electrode widths: 50 m, 100 m, 500 m and 1000 m. Using these geometries, the streamwise size of the microscale DBD actuators range from 200 m to 1.15 mm. 6 .1 Power Consumption Average power data is plotted in Fig ure 6 1 for devices with a 5 m thick silicon dioxide dielectric layer. The devices are operated at 1 kHz frequency in these tests. The smaller two device geometries (50 m and 100 m ground electrodes) show very little consumed power, partly due to their geometry confining the downstream extent of discharge. For these geometries, the devices consume less than 3 W/m of power at 4 kVpp input voltage. For the larger geometry devices (50 0 m and 1 mm ground electrodes), significantly more power is consumed, as the plasma region is 5x 10x larger in area as compared with the small geometries. At 4 kVpp, the largest device dissipated 28 W/m on average, while the device with a 500 m wide g round dissipated 15 W/m on average. The average power data is fitted using a power fit of the form For thin dielectric layers ( 1 mm), the exponent reported for macroscale DBD ac tuators (Enloe et al. 2004a) and is consistently greater
113 than 2 For the microscale data in figure 6 1, the values of from largest to smallest geometry, are found to be 3.1 2. 9 2.3 and 1. 5 The larger two device geometr i es follow close to the trend o bserved with macroscale DBD actuators; however the smaller two actuator geometries show values closer to 2. Figure 6 1. Average power consumed plotted against input voltage for devices with 5 m thick silicon dioxide dielectric layer. Four different device geometries are shown. Figure 6 2 reports power data for devices with a 10 m thick silicon dioxide dielectric layer. At 7 kVpp, the largest device dissipated 41 W/m on average, while the device with a 500 m wide ground dissipated 20 W/m on average. At 5 kVpp, the device with 1000 m ground geometry consumed 19.4 W/m. The values of from largest to smallest geometry, are found to be 2.3 1.8 1. 8 and 2.3 for these data. These values are lower than that obs erved with macroscale DBD actuators indicating that these ultra thin dielectrics allow s the discharge / induced flow to be generated at a reduced cost in power Compared with the first generation devices with 10 m thick polymer dielectric ( S ection 5.1), of which the largest similar geometry consumed an
114 average power of 20 W/m at 5 kVpp and 1 kHz, the actuators with silicon dioxide show similar power consumption. Figure 6 2. Average power consumed plotted against input voltage for devices with 10 m t hick silicon dioxide dielectric layer. Four different device geometries are shown. Comparing the power between the 5 m and 10 m silicon dioxide dielectric devices, nearly twice the power is dis sipated from the devices with 5 m thick SiO 2 at a given input voltage. At 4 kVpp, devices with 10 m SiO 2 consumed 12.5 W/m as compared to 28.5 W/m consumed for the devices with 5 m SiO 2 Similarly, the 500 m geometry device consumed 8.2 W/m and 14.8 W/m at 4 kVpp for the 10 m and 5 m thick diel ectrics, respectively. This increase is due to the electric field having larger magnitude across the thinner dielectric. The maximum input voltage is limited by the dielectric strength of the SiO 2 layer. At 1 kHz, the actuators with 5 m thick dielectric w ere able to withstand up to 4 kVpp, while the actuators with 10 m thick dielectric could operate up to 7 kVpp.
115 6.2 Velocity Data The velocities induced from the microscale DBD actuators span a range from less than 0.1 m/s to over 1 m/s. Figure s 6 3 and 6 4 show velocity profiles for each of the four geometries for devices with 5 m and 10 m thick silicon dioxide dielectric layer taken at x = 3 mm These plots illustrate the relative growth in the velocity with increasing input voltage (frequency kept at 1 kHz). Starting with the devices having 5 m dielectric (Figure 6 3), at low voltages (2.0 2.5 kVpp range) the plasma is able to induce velocities which are less than 0.1 m/s. Here the fluidic impact is rather weak; instead of a downstream directed wal l jet, the plasma is only able to very locally interact with the fluid. The flow filed in this case shows a small downstream effect; just down stream from the plasma region the fluid is no longer attached to the wall but instead begins to separate from the wall and recirculate above the actuator surface. With increasing voltage the plasma region is able to impart more momentum into the fluid, creating the characteristic wall jet that is similar to the macroscale actuators. Increasing the input voltage provid es a corresponding increase in the induced wall jet velocity with larger device geometries inducing greater velocities. Comparing similar geometries with different dielectric thickness, the induced velocities are in close agreement. At 4 kVpp, the 1000 m ground geometry shows 0.45 m/s for both dielectrics. Similar agreement is indicated for devices with 500 m and 100 m geometries, with slightly higher velocity produced for the device with the 5 m thick dielectric. The actuator with a 1 mm wide ground electrode creates velocities up to 1.5 m/s when operated at 7 kVpp for the 10 m dielectric case (Figure 6 4) While t he smallest device, having a 50 m wide ground, shows a maximum velocity as low as 0.04 m/s when operated at 2 kVpp, and reaches up to 0.4 1 m/s at 7 kVpp. The
116 Figure 6 3. Velocity profiles of four microscale DBD actuator geometries having a 5 m thick dielectric layer taken at x = 3mm downstream. Input v oltage varies from 2 4 kVpp, 1 kHz frequency A ) 1000 m wide ground electrode B ) 500 m ground C ) 100 m ground D ) 50 m ground. geometry, specifically the width of the ground electrode, governs the extent of the plasma discharge region. Larger actuators provide greater regions of discharge which in turn are able to impart more mo mentum into the surrounding fluid. The thickness of the wall jet can also be inferred from the velocity profile data. The geometry of the DBD actuator affects the height of the fluid jet as well as the velocity, but to a lesser extent. The wall jet thickne ss is generally around 2 mm, but varies up to ~3 mm for smaller geometries at lower input voltages. The larger devices, providing the
117 Figure 6 4. Velocity profiles of four microscale DBD actuator geometries having a 10 m thick dielectric layer taken at x = 3mm downstream. Input v oltage varies from 2 7 kVpp, at 1 kHz frequency. A ) 1000 m wide ground electrode. B ) 500 m ground. C ) 100 m ground. D ) 50 m ground. strongest momentum coupling between the plasma discharge and surrounding air, create thinner and longer wall jet regions as compared with smaller device geometries. Another useful way to plot the velocity data is to plot the absolute maximum x component of velocity as shown in Fig ure 6 5 This is similar to looking at the maximum velocity from each of the velocity profile plots in Fig ures 6 3 and 6 4 (above) but accounts for all x locations while the velocity profiles are extracted at a specific
118 downstream location ( x = 3 mm ) It also provides a clear visual interpretation of the max veloc ity trend with voltage. Similar to the power consumption, the maximum induced velocity follows a power law with input voltage, with In these data the values of from largest to smallest geometry, are found to be 4.3, 2.8, 1.8 a nd 1.8 for the devices with 5 m thick SiO 2 and 3.2, 2.4, 1.9 and 1.7 for the devices with 10 m thick SiO 2 The power dependency of microscale DBD actuators demonstrate values both similar to and below that observed in macroscale actuators. A B Figure 6 5. Plot of maximum velocity (x component) as a function of voltage input for various microscale DBD actuator geometries. A ) Devices with 5 m thick SiO 2 dielectric. B ) D evices with 10 m thick SiO 2 dielectric. A third way to analyze the velocity data is to examine the entire flow field. Figure 6 6 plots velocity contours of two device geometries for voltages ranging from 3 7 kVpp at 1 kHz frequency. The left column corresponds to a device with a 100 m ground electrode and the right column for a device with a 1000 m ground, and both having 10 m thick dielectric barriers. Starting with the larger geometry shown in the right column, with increasing voltage the induced wall jet becomes thinner and has more downstream effect. From 2.5 to about 4 kV pp, the induced velocity is fairly weak (< 1 m/s) and the flow begins to dissipate away from the wall by ~ 10 mm downstream. For
119 Figure 6 6. Velocity contour plots for microscale DBD actuator having 100 m ground (left column) and 1000 m g round (righ t column) For better clarity, the contours have independent scaling as indicated below each column. these voltages, the y location of the max velocity value is above 1 mm from the surface. At voltages of 4.5 kVpp and above, the induced velocity resembles the characteristic
120 wall jet produced by macroscale actuators. The location of the max velocity value is consistently at y = 0.86 mm, and varies in x location from 1.5 1.7 mm, occurring just past the downstream extent of the discharge. The velocity reach es 1 m/s when the voltage is 5 kVpp for this device geometry. In the left column of Figure 6 6, for the device with a 100 m wide ground electrode, the induced velocities reach 0.48 m/s at 7 kVpp, and just 0.05 m/s at 2 kVpp. From 2 4 kVpp the induced velocity begins to dissipate by ~ 10 mm downstream, similar to the larger geometry. Between 4 5 kVpp the velocity transitions into a wall jet with a more narrow and traditional profile. For these volta ges, the y location of the max velocity is between 1.4 1.6 mm, and the x location of the max velocity is pushed further downstream, between 5 6.5 mm. Even though the velocity values are relatively small, the ir fluidic effect extends over 25 mm downstre am. The total width of the device geometry in the streamwise direction is only 0.25 mm (top electrode width, gap, and ground electrode width); showing that the reduced size actuator can have significant downstream fluid effects, especially for its size. Th e dependence of the induced velocity on the input frequency was also investigated. T he device tested in this experiment had 100 m ground geometry and a 10 m thick silicon di oxide dielectric layer. The input voltage was held constant at 4 kVpp while the f requency was varied from 100 Hz up to 20 kHz. This range covers two orders of magnitude for the applied frequency, which corresponds to the range of frequencies that are possible with our lab equipment. The velocity profiles are plotted in Figure 6 7 A for the range of frequencies from 100 Hz to 20 kHz. The results are similar in trend with that observed from varying the
121 voltage input: with higher frequency there is greater induced velocity but to a lesser extent Figure 6 7 B plots the maximum induced veloc ity as a function of the input frequency. Both axes are plotted in logarithmic scale; here, the slope of the fit line ( = 0.54) indicates the power law relationship between the induced velocity and signal frequency A B Figure 6 7. Frequency depend ency of microscale DBD actuators. A) Velocity profiles plotted for various frequenc ies taken at x = 3 mm downstream; device input vo ltage held constant at 4.0 kVpp B) Maximum x component of velocity vs. frequency plotted with logarithmic axes 6.3 Thrust and Plasma Force Results In this section, data is reported for the extracted thrusts and net plasma body forces. Starting with the thinner dielectric case, the x and y components of the thrust are plotted in Figure 6 8 for devices with 5 m SiO 2 T hese devices indicate streamwise (x component) thrust values up to 0.20 mN/m and wall normal (y component) values reaching just 0.05 mN/m. The streamwise thrust component is on average about four times that of the wall normal component. The actuators with 50 and 100 m ground electrodes show similar thrust values for both x and y components. For these two geometries, the downstream extent of the plasma is 150 m and 200 m respectively
122 (including the 100 m gap between electrodes as well as the ground ele ctrode width). There is little difference in the induced velocities and thrust forces for these geometries. For these data the values of from largest to smallest geometry, are found to be 5.2, 6.3, 5.8 and 4.5. The power depend ency of microscale DBD act uator thrust demonstrate s values larger that observed in macroscale actuators ( The y component of thrust is typically an order of magnitude lower than the x component, although demonstrates a similar power dependency. A B Figure 6 8. Thrust values computed based on the control volume analysis for devices with 5 m SiO 2 dielectric. A ) x component. B ) y component. The net plasma force acting on the fluid is shown in Figure 6 9. This data is computed by integrating the spatial body force at all points within the two dimensional flow field. The streamwise plasma force reaches 0.35 mN/m while the y component reaches 0.05 mN/m. The y components of the plasma force match fairly well with the y component from the thrust data. In contrast, the plasma f orce shows significantly larger values for the x component as compared with the thrust. The difference in these values For the x component of
123 the plasma force the values of from largest to sma llest geometry, are found to be 5.5, 5.1, 4.8 and 3.3. A B Figure 6 9. Plasma force computed from the spatial body force estimation for devices with 5 m SiO 2 dielectric. A ) x component B ) y component. Similar results are reported next for the device s having a 10 m thick SiO 2 dielectric barrier. Thrust values reach 1.4 mN/m in the streamwise direction and 0.3 mN/m in the wall normal direction for the largest device geometry (Figure 6 10). These devices are able to sustain plasma at higher voltages co mpared with the devices having a thinner dielectric layer, enabling higher forces to be achieved due to the power law relationship of the thrust upon the input voltage. For the x component of thrust the values of from largest to smallest geometry, are f ound to be 4.9, 3.7, 3.6 and 5.6. These values demonstrate a stronger dependency on the thrust with input voltage compared with macroscale DBD actuators. The plasma force is also plotted for the devices having 10 m SiO 2 in Figure 6 11. Here, the plasma f orce reaches 2.3 mN/m for the largest actuator geometry at 7 kVpp. The device having 50 m ground geometry produces a plasma force of 0.22 mN/m at 7 kVpp, indicating a range of force values spanning an order of magnitude. The wall
124 A B Figure 6 10. Thrust values computed based on the control volume analysis for devices with 10 m SiO 2 dielectric. A) x component. B) y component. normal or y component of the plasma force reaches up to 0.13 mN/m, and as low as 0.05 mN/m at 7 kVpp. For the x component of the plasma force the values of from largest to smallest geometry, are found to be 4.4, 3.4, 3.4 and 3.8. Similar to the data from the actuators with 5 m thick dielectric, the plasma force values are consistently larger than the thrust values. Recall f rom Equation 4 9 A the shear force may be estimated by integrating the viscous shear component. Adding the A B Figure 6 11. Plasma force computed from the spatial body force estimation for devices with 10 m SiO 2 dielectric A) x component. B) y compon ent.
125 net shear force to the thr u st enables a method to estimate the plasma body force. Figure 6 12 plots the plasma force from integrating the spatial body force for comparison with the thr u st while including the additional shear force. There is reasonble agreement between these two methods of computing the plasma force. Two of the four actuator geometries are shown for clarity (1000 m and 100 m ground electrodes) for the two dielectric thi cknesses investigated. The device with 100 m geometry shows nearly equal values between the two methods, while the values vary more for the larger geometry. Thrust values are plotted in Figure 6 13 over the range of frequencies investigated for the device with 100 m ground geometry and a 10 m thick dielectric layer. At 100 Hz the parallel thrust value indicates 0.005 mN/m as the maximum velocity at this electrical input (4 kVpp, 100 Hz) reaches just 0.04 m/s. At 1 kHz, the thrust computed is 0.05 mN/m an d by 20 kHz the thrust reaches 0.50 mN/m. Here the frequency and corresponding thrust values both cover two orders of magnitude. The trend of increasing thrust with frequency is apparent from these plots, however the dependency A B Figure 6 12. Compari son between control volume estimation and direct integration of spati al body force. A ) 5 m SiO 2 B ) 10 m SiO 2
126 of both velocity and force values are largest with respect to increasing input voltage. The power fit in Figure 6 12 indicates value of 1.05 for the frequency dependency of microscale DBD actuators, quite similar to the value of 1.1 observed for macroscale actuators. A B Figure 6 13. Streamwise thrust component plotted against frequency for device s with 100 m ground geometry and 10 m SiO 2 dielectric, operated at 4.0 kVpp. A ) Linear scale. B ) L og scale. 6.4 Thermal Results The thermal properties of the silicon dioxide actuators were also investigated ( Figure 6 14 ) They show a significant temperature increase for the devices having 500 and 1000 m wide ground electrodes ( = 32 C), but very little increase for the devices having 50 and 100 m wide electrodes ( = 5 C). These devices with silicon dioxide dielectric show larger overall surface temperatures (up to 58 C) compared with the a ctuators with polyimide dielectric ( Figure 5 13 ) However, at a given voltage, the electric field is larger across the devices with silicon dioxide since the dielectric layer is half as thick as the polyimide layer. More energy is dissipated across the thi nner devices (higher average power consumed) which contributes to the larger temperatures observed.
127 Figure 6 14 Average temperature increase plotted against applied voltage for devices with different electrode geometries. Actuators have a 5 m thick si licon oxide dielectric layer.
128 CHAPTER 7 CHARACTERIZATION MET RICS AND COMPARISON WITH MACR OSCALE DBD ACTUATORS The power, velocity and force data are collected and summarized here for an overall look at the performance of microscale DBD actuators. The following tables relating to device power consumption and actuator design. The tables are used to make side by side comparisons between the measured performance of microscale and macroscale DBD actuators. 7.1 Materials, Geometries and Performance Data The tables include results from F orte et al. (2007), Abe et al. (2008) and Thomas et al. (2009) for comparison with standard macroscale DBD actuator s In these publications, the power consumption, induced velocity, and thrust data are all reported in detail along with device geometries. T he dimensions and material properties used in the different studies of micro and macro DBD actuators are provided in Table 7 1, while Table 7 2 reports a list of materials and their densities which are used in the calculation of the actuator mass Two micr oscale DBD actuators are used in these comparisons; one with polyimide and the other with a silicon dioxide dielectric, both having 10 m thick dielectric layers. The device used for the polyimide data has 10 100 1000 m geometry while the actuator with si licon dioxide has 50 100 1000 m geometry. A few words are necessary in order to clarify the derivation of the performance metrics. In these calculations of the device volume and mass, the area used for the the dielectric surface in the streamwise direction (it is typically not reported in most publications). Also, the thickness, or height of the adhesive backed electrodes is not
129 Table 7 1 Device dimensions and mate rial properties used in analysis of micro and macroscale DBD actuator performance metrics. Actuator w idth* (m) Dielectric h eight (m) Substrate h eight (m) Actuator v olume* (m 3 ) Actuator m ass* (g) Micro DBD with polyimide 1.11 x 10 3 10 5 5 x 10 4 5.67 x 10 7 1.42 Micro DBD with SiO 2 1.15 x 10 3 10 5 5 x 10 4 5.87 x 10 7 1.49 Forte et al. (2007) 10 2 2.0 x 10 3 2.10 x 10 5 26.3 Abe et al. (2008) 3.1 x 10 2 1.8 x 10 3 5.70 x 10 5 97.2 Thomas et al. (2009) 7.3 x 10 2 6.35 x 10 3 4.67 x 10 4 1256 does not include streamwise length of dielectric included in the calculation of the actuator volume or mass for the macroscale DBD actuators (electrode thickness is included in the volume of microscale actuators). Table 7 2 List of materials used in performance comparisons and corresponding material d ensities. Material Density (kg/m 3 ) Aluminum 2700 Copper 8940 Glass (soda lime) 2525 Polyimide (PI 2611) 1400 PMMA (Plexiglas) 1180 Quartz 2650 Silicon dioxide (PECVD) 2650 Titanium 4506 Table 7 3 reports thrust velocity and power data reported from the respective research groups. The data extracted from different research groups all correspond to actuators operated at 1 kHz frequency with sinusoidal input. This alleviates the difficulty in comparing data with different combinations of input volt age and frequency. These
130 Table 7 3 Characterization data used in analysis of micro and macroscale DBD actuator performance metrics. AC i nput (kVpp / kHz) Power c onsumption (W/m) Thrust (mN/m) Induced v elocity (m/s) Micro DBD with polyimide 5 / 1 15 2 .0 2 Micro DBD with silicon dioxide 7 / 1 41 1.4 1.5 Forte et al. (2007) 24 / 1 25 2 Abe et al. (2008) 20 / 1 20 3.9 1.4 Thomas et al. (2009) 74 / 1 590 120 values are used with the device geometry and material parameters to populate the following tables for comparison of micro and macro device performance. 7.2 Thrust Metrics Table 7 4 presents the performance metrics for micro and macro DBD actuators. The first column in the table reports the actuator thrust ef fectiveness (force produced per consumed power). One of the reported actuators (Abe et al. 2008) indicates similar thrust and power values to the microscale actuator, while the other macroscale device (Thomas et al. 2009) indicates significantly larger thr ust production and power consumption. However, the microscale DBD actuators with polyimide dielectric demonstrate similar thrust effectiveness (thrust per consumed power) to the mac roscale actuators. The microscale actuator with a SiO 2 dielectric barrier s how s about an order of magnitude reduction in thrust effectiveness.
131 Thrust density is reported in the second column of Table 7 4. Note here that the thrust density represents the thrust force per device volume This should not be confused with the plasma force density (acting on the fluid). Microscale actuators having both dielectric materials show an order of magnitude improvement compared with thrust density computed for Abe et al. (2008) and Thomas et al. (2009). Similarly, the thrust per actuator mass (third column) shows order of magnitude larger values for both microscale actuators in comparison with macroscale data. This is clearly due to the significant size reduction of the microscale actuators. Table 7 4 Summary of various thrust metrics compar ing micro and macroscale actuator performance. Thrust e (mN/W) Thrust / actuator volume (mN/m 3 ) Thrust / actuator m ass (unitless, g/g) Micro DBD with polyimide 0.13 3.53 x 10 6 1.41 x 10 1 Micro DBD with silicon dioxide 3.41 x 10 2 2.39 x 10 6 9.40 x 10 2 Abe et al. (2008) 0.20 6.84 x 10 4 4.10 x 10 3 Thomas et al. (2009) 0.20 2.57 x 10 5 9.7 x 10 3 R ecall, the microscale DBD actuators are manufactured on a glass substrate which is only necessary as a handle on which to fabricate the devices. Hence, these devices could be manufactured on a variety of [thinner, lower density] substrates. For example, a thin polymer or sacrificial layer could be deposited on the handle substrate prior to device fabrication, and the actuators could be released from substrate after the fabrication is complete creating a thin and flexible sheet of microscale DBD actuators. T his is not only theoretically possible but easily realizable (this would require an
132 additional layer of dielectric under the bottom electrode during the fabrication process in order to provide mechanical support and also encapsulate the bottom electrode to prevent discharge on the bottom side of the dielectric ). The thrust density and thrust per mass metrics both increase two additional orders of magnitude when neglecting the size and weight of the glass substrate. 7.3 Velocity Metrics Table 7 5 presents a list of performance metrics similar to Table 7 4 in this case relating to the maximum induced velocity that the actuators produce. Parameters of interest relate the maximum value of the induced velocity to the size, weight and power requiremen t needed to produce a wall jet with such velocity. The velocities produced by the microscale actuators are on the order of the macro DBD actuators as shown in the corresponding velocity results for each actuator type. Table 7 5 Summary of various velocity metrics comparing micro and macroscale actuator performance. Velocity / p ower (m/s) / (W/m) Velocity / actuator v olume (m/s) / m 3 Velocity / actuator m ass (m/s) / g Micro DBD with polyimide 0.13 3.53 x 10 6 1.41 Micro DBD with silicon dioxide 3.66 x 10 2 2.56 x 10 6 1.01 Forte et al. (2007) 0.08 9.52 x 10 4 7.60 x 10 2 Abe et al. (2008) 0.07 2.46 x 10 4 1.44 x 10 2 The first column in table 7 5 relates the induced velocity to consumed power. The microscale actuators show similar values to the macroscale data reported by both Forte et al (2007) and Abe et al. (2008). Specifically, the microscale actuator with polyimide
133 indicates 54% and 62% higher values than the two macroscale devices, while the microscale actuator with SiO 2 indicates slightly lower velocity effectiveness The velocity per actuator volume and velocity per actuator are reported in the final two columns. The velocity per volume and velocity per mass both indicate two orders of magnitude improvement due to the small size of the microscale actuators. Without the glass substrate, these values could reach up to four orders of magnitude improvem ent over macroscale actuators 7.4 Actuator Efficiency The energy conversion efficiency of the DBD actuator can be computed as the ratio of the mechanical power output to the electrical power input, according to (7 1 ) where the mechanical ( P me ) and electrical power ( P e l ) are normalized per unit length (both having units of W/m) The electrical power is computed from Equation 4 1. However, t here are several methods that may be used to compute the mechanical output power for determining actuator efficiency. One simple method is to use the peak velocity ( units of m/s) and thrust (units of N/m) values, in which the mechanical power is computed as (7 2 ) In this case the result represents the best case actuator efficiency. From here, the actuator efficiency can be computed as the ratio of mechanical output power to electrical input power from Equation 7 1.
134 A deficiency or downside of this method for computing the actuator efficiency is that the mechanical output power is taken as the product of the thrust and maximum velocity valu es T h is value may provide an upper boundary for the best possible efficiency; however, the maximum integrated thrust and peak velocity do not necessarily occur at the same location. Furthermore, a single data point of velocity does not necessarily represent the global velocity field nor d oes the thrust capture all of the details of the body force. Another method to compute the DBD actuator efficienc y is to use the kinetic energy density flow rate. In this case, t he mechanical output po wer per unit length is computed according to (7 3) This method has been used by other research groups to define the DBD actuator mechanical power (Moreau 2007; Jolibois and Moreau 2009), and was also suggested by Cattafesta an d Sheplak (2011). An advantage of this method is that variations in the velocity profile are incorporated in the mechanical power computation. This method also has one drawback; it is susceptible to variability based on the d ownstream choice of location of the velocity profile used in the calculation A third option for computing the mechanical power is based on the volume velocity and dynamic pressure. The volumetric flow rate, Q is found by integrating the velocity flux through a given downstream surface in the y z plane (perpendicular to the primary flow direction). In this case of two dimensional data, the volumetric flow rate is computed in the wall normal or y direction, and normalized per unit length in the spanwise or z direction (units of m 2 /s)
135 (7 4 ) In this case the volumetric flow rate is compute d in the wall normal or y direction, and normalized per unit length in the spanwise or z direction (units of m 2 /s) The dynamic pressure, q is found as follows (units of N/m 2 ): (7 5) where (7 6) The value of is chosen as 1 mm to represent the wall jet thickness. The mechanical output power is then found as the product of dynamic pressure and volumetric flow rate according to (7 7 ) This method also suffers from variability based on the downstream choice of location of the velocity profile used in the calculation. The fourth and final method proposed t o comput e the efficiency is to i ntegrate the velocity body force product over a domain of interest. This method includes all of the force and velocity events within the entire domain, instead of picking single data points to represent the mechanical power. Recall, the body force f plasma is computed from Navier Stokes equations (Equations 4 11). In this case, the mechanical output power is computed as (7 8 )
136 Values are summarized in Table 7 6 for the different methods of computing DBD actuator efficiency based on the four methods presented for computing the mechanical output power Table 7 6 Summary of various methods for computing DBD actuator efficiency. Method 1 Equation 7 2 Method 2 Equation 7 3 Method 3 Equation 7 7 Method 4 Equation 7 8 Micro DBD with polyimide 2.67 x 10 4 7.70 x 10 6 2.46 x 10 5 2.74 x 10 5 Micro DBD with silicon dioxide 5.12 x 10 5 3.33 x 10 6 3.74 x 10 5 1.65 x 10 5 This efficiency based on Equation 7 2 can be computed only if both velocity and thrust data are reported. Abe et al. (2008) reports thrust and velocity parameters, providing one macroscale actuator reference to compare with the microscale actuators. The efficiency computed for reported macrosc ale data is 2.73 x 10 4 (Abe et al. 2008), while the efficiency for microscale actuators is computed as 2.67 x 10 4 for the polyimide based device. Here the microscale actuator indicates equivalent energy conversion efficiency to the macroscale actuator ba sed on this efficiency parameter For the microscale device with SiO 2 dielectric, the efficiency is 5.12 x 10 5 demonstrating reduced energy conversion of the consumed electrical input to mechanical output performance. The efficiency computed using Equati on 7 3 is significantly lower than that from the first method (using max thrust and velocity values ). The efficiency is 2.44 x 10 5 for polyimide based actuators and 1.68 x 10 5 for actuators having SiO 2 dielectric. Since this method is inclusive of variations along the velocity profile, it is expected to produce
137 a more accurate value of the overall actuator efficiency as compared to the efficiency computed from Equation 7 2. From Equation 7 7, the DBD actuator effic iency is based on the product of dynamic pressure and volumetric flow rate. The microscale actuator efficiency is found to be 6.48 x 10 5 and 1.33 x 10 5 for actuators with polyimide and SiO 2 dielectric materials, respectively. These values are similar to those using the previous method in which variations in the velocity profile are incorporated for the mechanical power calculation. Using the fourth method (Equation 7 8) based on integrating the velocity body force product the microscale DBD actuator efficiency is found to be 2.74 x 10 5 and 3.74 x 10 5 for actuators with polyimide and SiO 2 dielectric materials, respectively. Interestingly, t his method indicates the actuators with SiO 2 dielectric demonstrate better efficiency despite their producing lo wer thrust and velocity values as compared with polyimide based actuator s. Unlike the previous two methods, which use a single downstream velocity profile to compute the mechanical power, this method accounts for all of the force and velocity data within t he domain of integration. However, the values computed from this method are also consistent with the two previous efficiency results. Furthermore, Kriegseis et al. (2011a; 2013b) has reported two publications specifically on the topic of DBD actuator effic iency. In his more recent publication relating the input power to the power delivered to the DBD actuator, the fluid mechanic efficiency relating the thrust force t which relates the amount of power used to the amount that physically makes sense as
138 control application. The parameters are call ed effectiveness instead of efficiency ratios since both of these parameters can theoretically reach larger than one in magnitude, and the fluid mechanic ratio has units of N/W (instead of being unitless). For further details on DBD actuator efficiency, th e reader is referred to these discussions by Kriegseis et al. (2011a; 2013b). In summary, microscale DBD actuators show promising results, inducing flow with significant velocity and similar profile to macroscale devices with a drastic reduction in d evice size and power consumption. Four methods are suggested for computing the DBD actuator efficiency as a ratio of mechanical power output to electrical power input. The results indicate potential for microscale DBD actuators to outperform macroscale act uators; they provide increased control authority in several performance metrics when compared with data reported from several macroscale devices. The primary advantage comes from scaling the actuator geometry, resulting in drastic reductions in the actuato r volume and mass. In addition, reduction in the power requirement gives rise to the potential for portability: a primary advantage that macroscale actuators are not able to capitalize upon.
139 CHAPTER 8 SUMMARY AND FUTURE W ORK This chapter begins with a sum mary of the work performed and presented throughout this document on the design, fabrication and characterization of DBD plasma actuators with microscale dimensions. Contributions to the science research community regarding DBD actuators are identified alo ng with proposed applications where microscale DBD actuators may be utilized. Results from preliminary work on non standard DBD actuator designs are presented as future work, in which further investigation on the actuator geometry is required to find the optimal design parameters. Among these nontraditional DBD actuators include arrays of DBD actuators, act uators with serrated anodes, and also serpentine and plasma synthetic jet designs intended to pinch the fluid normal to the actuator surface. 8.1 Summary and Conclusions A first generation of microscale DBD plasma actuators was fabricated and demonstrated successful operation. These devices primarily used polyimide for the dielectric barrier material. Device characterization focused on three categories: power consumption, induced flow velocity, and force/thrust. The thrust force was measured both directly and indirectly, while the plasma force can only be inferred from velocity measurements. Analyses of the data was performed and reported over numerous microscale DBD actuators ranging in various dimensions, materi als and electrical excitation. Polyimide bas ed actuators produced up to 3 .5 mN/m of thrust and 2 m/s induced velocity while consuming 20 W/m of electrical power. The net plasma body force reached 5 mN/m.
140 It was found that the first generation actuators consistently failed during operation. The shor t device lifetime made it challenging to collect data The cause of the failure of the polyimide barrier was investigated and ultimately attributed to the surface being eroded due to sputtering of the material caused by ions within the discharge bombardin g the polyimide surface. A new generation of devices was designed with silicon dioxide for the dielectric barrier material. SiO 2 was chosen as it is a popular thin film insulation material for electronics manufacturing and is also used for macroscale DBD actuators. The SiO 2 improved the lifetime of the microscale DBD actuators, enabling devices to operate for 10s of minutes. These devices were characterized with the same methods as the first generation of devices. However, the extended lifetime of the actu ators with SiO 2 enabled them to be tested at several voltages and/or frequencies in order to observe trends in the device performance with electrical input. At 7 kVpp and 1 kHz, d evices with SiO 2 dielectric barriers produced up to 1.4 mN/m of thrust and de monstrate 1.5 m/s induced velocity while consuming 41 W/m of electrical power. The net plasma body force reached 2.5 mN/m. The microscale DBD actuator performance was summarized and compare d with reported macroscale data sed to compare thrust and velocity when normalized by power consumption; the microscale actuators with polyimide dielectric demonstrate similar thrust effectiveness with macroscale devices and indicate 54 % and 62 % improvement in velocity effectiveness as compared with two macroscale devices. The devices with silicon dioxide, however, showed about an order of magnitude reduction in thrust effectiveness compared with macroscale actuators, and
141 slightly lower velocity effectiveness as well. Although, o n a per volume or per mass basis, the microscale DBD actuators show an order of magnitude improvement over macroscale actuators for both thrust and velocity metrics. In the case where force and velocity data were both reported, the efficiency is computed as a rat io of the mechanical output power to electrical input power. Based on max thrust and velocity data, t he microscale actuator with polyimide demonstrated equivalent energy conversion efficiency compared with the macroscale actuator Three additional methods have been suggested in order to compute the efficiency of DBD actuators for flow control applications. These other methods incorporate integrated values of the velocity and force data and are expected to provide a more realistic representation of the overa ll actuator efficiency. In general the microscale DBD actuator induced velocity, thrust, and power consumption scale favorably with size reduction. Microscale DBD actuators demonstrated induced velocities on the order of that induced from macroscale actu ators with a significant reduction in power requirements. Utilizing semiconductor fabrication techniques improves manufacturing tolerances and repeatability between devices compared to handmade actuators. The compact size and low mass of the micro actuator s make them implementable with minimal weight penalty. Overall, the actuators with silicon dio xide provide better reliability but show lower performance. Actuators with p olyimide show ed larger velocity and thrust values but has low reliability due to their short lifetime before the dielectric fails Challenges in the fabrication process may limit the available materials or possible actuator geometries. For example, using PECVD to deposit a 10 m layer of SiO 2 is difficult as thermal
142 stresses induce cracking or delamination over regions of large metal features. Material selection may also become limited due to adhesion compatibility between materials; PDMS, an excellent polymer with good dielectric strength and excellent chemical resistivity, does not adhere to sputtered metals and was not investigated as the dielectric barrier. Recall from Section 2.2, t here were three benefits of scaling DBD actuators which provided motivation for this work: 1) increase in thrust density, 2) reduction in power requirement, and 3) improv ement in manufacturing. Unfortunately, the first benefit of increased force density was not demonstrated using microscale DBD actuators. From the spatial body estimate (Figure 4 20), the plasma force density acting on the fluid is on the order of 100 N/m 3 much lower than the predicted force density indicating values up to MN/m 3 The large discrepanc y between the numerical prediction and experimentally observed values of the force density are most likely due to the difference in actuator configurations as well as the assumptions made in both cases. Specifically, the numerical simulation used a volume D BD configuration with a dc input voltage of 500 V across two parallel plate electrodes. Furthermore the discharge medium was modeled in a pure nitrogen environment. The two other benefits, namely the reduction in power and improvement in manufacturing, wer e both achieved in this work. The microscale actuators demonstrate an order of magnitude reduction in power as observed in the results data (Sections 5.1 and 6.1). Using semiconductor fabrication techniques also improve the manufacturing of DBD actuators i n comparison with handmade actuators. Photolithography patterning provides precise feature geometry
143 and accuracy in electrode alignment. Using these processing methods enables complex electrode patterns that would otherwise be challenging to create by hand 8.2 Research Contributions Investigation of microscale plasma systems offers the opportunity for fundamental scientific understanding of tru ly multidisciplinary research. The study of plasma incorporates aspects of electrical, chemical, optical, mechani cal and aerospace engineering along wi th complex physics principles. U nique microscale measurements were made of p hysical variables to quantify microscale DBD actuator performance The quantitative and comprehensive physical measurements provide valuable p hysical insight for better understanding of microscale plasma actuator performance, and also provide body force data for comparison and validation with numerical studies The completed contributions of th is research include: Development and demonstration o f microfabrication methods for realizing DBD actuators with electrode dimensions from 1 0 5 000 m using 5 2 0 m thick dielectric barriers with an electrode gap of 1 00 m Thorough and systematic characterization of the electrical, fluidic, thermal and me chanical behavior of microscale DBD plasma devices, with particular interest in the actuator thrust as well as the net body force acting on the fluid for comparison with numerical simulations Exploration of microscale DBD actuator configurations for maximu m induced flow velocity and correspondingly imparting maximum momentum transfer into the flow while consuming the lowest possible power requirement Development and dissemination of a benchmark experimental database of microscale DBD device performance for reference by other researchers These efforts shall serve as the foundation toward future application systems, such as aerodynamic flow control, aerospace propulsion, microfluidic pumps, ozone generation, water purification, and medical sterilization. A gre at advantage of microscale discharge devices is the possibility of creating portable systems that could sterilize
144 medical devices out in the field, or perhaps even a portable water purification kit that could be used in cities with no water infrastructure to provide clean drinking water. 8.3 Future Work recorded. Arrays of DBD actuators have been studied for macroscale DBD actuators ( Roth 2003; Roth and Dai 2006; Forte et al. 2007; Thomas et al. 2009; Likhanskii et al. 2010) to investigate whether the induced velocity can be increased or extended using sequential actuators. Both Forte et al. (2007) and Thomas et al. (2009) report a slight increase in the velocity w ith multiple actuators, about 1 m/s increase shown between 1 and 2 actuators, but show diminishing returns with additional actuators. Forte et al. (2007) reported maximum velocity when there is 2 cm spacing between actuators. These values were measured usi not been reported PIV data for arrays of DBD actuators, possibly due to the large spacing required and a trade off with PIV resolution with using a large field of view. Arrays of microscale DBD actuators having either 5 mm or 10 mm separation between electrode pairs were tested using two component PIV The devices used in these experiments have either 10 m polyimide or 10 m silicon dioxide for the dielectric barrier. Figure 8 1 provides PIV data for a microscale DBD array of three actuators with 5 mm spacing. The device has 20 100 40 m geometry, a 10 m thick polyimide dielectric and is operated at 3.5 kVpp and 1 kHz. The device geometry is indicated in Figure 8 1 under the velocity field for reference. A maximum induced
145 Figure 8 1. Array of three microscale DBD actuators with 5 mm spacing. Device has 20 100 40 m geometry with a 10 m thick polyimide dielectric barrier, operated at 3.5 kVpp and 1 kHz. A schematic of th e device geometry is shown below for reference. velocity of 0.55 m/s occurs downstream from the last actuator in the DBD array. The first two actuators in the array induce velocities that are 0.3 0.4 m/s The spacing of the actuators cannot be too small such that the anode of one device is close to the in the unwanted direction. If the actuator pairs are too close, the electric field lines will lose their directional ity reducing momentum transfer to the surrounding fluid. An array of five actuators with 5 mm spacing is shown in Figure 8 2 for voltages ranging from 2.5 to 4 kVpp Similar trends are observed as with Figure 8 1: the maximum velocity occurs downstream of the last actuator in the array at 4 kVpp. The first four actuators produce similar effects, inducing velocities about 80 % of the maximum velocity that is produced further downstream. At lower voltage input, the induced flow does not connect from one actu ator to the next. At 2.5 kVpp the induced
146 Figure 8 2. Array of five microscale DBD actuators with 5 mm spacing. 50 100 100 m geometry with 10 m thick polyimide dielectric. flow is very local, and quickly diffuses downstream before the next actuator can entrain the flow to continue its momentum down the array. Instead, each actuator produces a local fluid perturbance which recirculates in the region above the electrode pair. By 3 kVpp and above, the induced flow from neighboring actuators reaches the next pair and the flow is pushed downstream as intended by the array. Measurements from a rrays with 10 mm spacing are presented in Figure 8 3. Three actuators with 50 100 100 m geometry and a 10 m thick SiO 2 dielectric layer are
14 7 Figure 8 3. Array of three microscale DBD actuators with 10 mm spacing. 50 100 50 m geometry with a 10 m thick polyimide dielectric. operated with voltages ranging from 2.5 5 kVpp. The induced flow is not able to connect between neighboring actuators with 10 mm spacing, regardless of the input voltage. Each actuator creates a recirculation region above the surface; the streamwise effect on the surrounding fluid is very local around the electrodes, diffusing within about 5 mm downstream except for the last actuator in the array. Similar to the previous array results, the last actuator in the array produces the largest velocity and shows the greatest downstream velocity effects.
148 Figures 8 4, 8 5 and 8 6 show pictures of the microscale designs for serpen tine, serrated, and synthetic jet actuator configurations. Some challenges were encountered upon testing of these designs due to the small features used in the actuator geometries. Figure 8 4. Serpentine DBD actuators. 100 m wide powered electrode 100 m electrode gap, with various ground electrode geometries. Figure 8 5 Microscale DBD actuators with serrated anode and 1 mm wide ground electrodes. From left to right, the ratio r of the serration base to height is 2, 4, 5 and 10. Figure 8 6 Plasma synthetic jet actuators. Ground electrode dia meter, from left to right: 1, 2 and 3 mm. There are limitations to the dimensions of the features of these DBD designs. The laser sheet used to illuminate the plane of interest for PIV measurements has a
149 thickness of about 0.5 mm. Thus, the period of the serpentine has a minimum such that the laser does not illuminate too large of a region of the actuator. For example, if the period of the serpentine actuator was 1 mm or smaller, the laser would illu minate at least half of the serpentine period and the regions of the peaks and valleys could not be resolved with the laser sheet. Similarly, the size of the teeth of the serrated actuators and the diameter of the synthetic jet ha s lower limits as well. Fi gure 8 7 illustrates these limits in the test equipment. The images in Figure 8 7 A show a plasma synthetic jet A B Figure 8 7. Microscale DBD actuators shown during operation and under test conditions. A) Plasma synthetic jet actuator. B) Serpentine actuator. actuator with a 1 mm ground diameter in operation (left side of Figure 8 7A) and during the PIV test (right side of Figure 8 7A ). The diameter of the laser is large such that a plane through the center of the plasma synthetic jet cannot be well r esolved with relation to the actuator geometry. The serpentine geometry suffers from the same limitations, as shown in the images of Figure 8 7 B Data from these two actuators was unclear and disregarded for analysis.
150 Results from a microscale DBD actuator with a serrated anode are shown in Figure 8 8. This device has a 10 m thick SiO 2 dielectric barrier and a 1 mm wide ground electrode. For this actuator, t he teeth geometry has a base of 200 m and height of 100 m, such that the base to height ratio r of the teeth is 2. Joussot et al. (2013) recently reported that using a small value of r (< 1) is beneficial to induce higher velocity to increase thrust while using large values of r (> 2) is beneficial to induce vortices, creating a 3D flow. For the micros cale geometry having a base of 200 m, the plane of illumination in the PIV test spans two entire teeth of the serrated anode. Hence the resulting velocity field includes combined fluid effects at the peaks and valley of the anode serrations. Figure 8 8 Microscale DBD actuator with serrated anode; base to height ratio r = 2. Device has a 1 mm wide ground electrode and a 10 m thick SiO 2 dielectric. The induced velocity from the serrated actuator has a significant y component and does not represent the typical wall jet induced from standard DBD actuator geometries. Pinching of the fluid between the teeth give s rise to the increased y component of the induced velocity. The results from the serrated actuator are compared in Figure 8 9 with the linear or standard DBD design having the same dielectric and electrode geometry (besides the serrations). Two differences stand out: first the maximum velocity is greater
151 for the standard actuator design. This does not agree with results reported by Thomas et al. (2009), where maximum thrust is reported for a teeth ratio r = 0.25 and is 50 % greater than the thrust measured from a standard DBD actuator. Second, the flow Figure 8 9. Comparison between serrated and standard DBD actuators. A ) Serrated DBD actuat or. B) Standard DBD actuator Both actuators have 1 mm wide ground electrodes and a 10 m thick silicon dioxide dielectric induced from the serrated actuator is no longer primarily in the streamwise direction; the x and y components of the induced veloci ty are on the same order, as indicated in the PIV data. In order to resolve the fluid effects at the peaks and valleys of the serrations, either the geometry must be increased such that these regions can be measured independently of each other, or the equipment used fo r characterization must change. Techniques such as laser Doppler velocimetry (LDV) provide a non intrusive velocity measurement with high spatial resolution. LDV also allows for near wall veloci ty measurements (although may be challenging). Compared with PIV, which measures velocity over the entire two dimensional field of view LDV measurements correspond to a single point within the flow field.
152 APPENDIX A DATABASE OF VELOCITY CONTOURS This da tabase shall provide measurements and performance results for microscale DBD actuators over a range of [microscale] dimensions and materials. It is useful in determining the appropriate applications for which microscale DBD actuators may be utilized, and a lso provides reference data for numerical modeling. The appendix begins with data for devices with 10 m thick polyimide dielectric barriers, follow ed with velocity fields for devices with 5 and 10 m thick silicon dioxide dielectrics. In this appendix t here is not a discussion following each figure; the reader is referred to the sections where the corresponding velocity results are discussed. A.1 Devices with 10 m Polyimide Dielectric Figure A 1. Velocity field s for microscale DBD actuator s. A) Devices with 10 m wide powered electrodes. B) Devices with 50 m wide powered electrodes S hown for three ground electrode sizes with 5 kVpp, 1 kHz input.
153 Figure A 2. Velocity field s for microscale DBD actuator with 500 m wide ground electrode geometry shown for various electrode sizes 5 kVpp, 1 kHz input.
154 Figure A 3. Velocity fields for microscale DBD actuator with 1000 m wide ground electrode geometry, shown for various electrode sizes. 5 kVpp, 1 kHz input.
155 A.2 Devices with 5 m Silico n Dioxide Dielectric Figure A 4. Velocity fields for microscale DBD actuator with 50 100 50 m geometry, sh own for voltages ranging from 2 to 3. 5 kVpp, 1 kHz input. Figure A 5. Velocity fields for microscale DBD actuator with 50 100 100 m geometr y, sho wn for voltages ranging from 2 to 4 kVpp, 1 kHz input.
156 Figure A 6. Velocity fields for microscale DBD actuator with 50 100 500 m geometry, sho wn for voltages ranging from 2 to 4 kVpp, 1 kHz input.
157 Figure A 7. Velocity fields for microscale D BD actuator with 50 100 1000 m geometry, shown for voltages ranging from 2 to 4 kVpp, 1 kHz input.
158 A.3 Devices with 10 m Silicon Dioxide Dielectric Figure A 8. Velocity fields for microscale DBD actuator with 50 100 5 0 m geometry, shown for voltages ranging from 2 to 7 kVpp, 1 kHz input.
159 Figure A 9. Velocity fields for microscale DBD actuator with 50 100 100 m geometry, shown for voltages ranging from 2 to 7 kVpp, 1 kHz input.
160 Figure A 10. Velocity fields for mi croscale DBD actuator with 50 100 500 m geometry, shown for voltages ranging from 2 .5 to 6.5 kVpp, 1 kHz input.
161 Figure A 11. Velocity fields for microscale DBD actuator with 50 100 1000 m geometry, shown for voltages ranging from 2.5 to 7 kVpp, 1 kHz input.
162 APPENDIX B UNCERTAINTY ANALYSIS This appendix provides the uncertainty analysis for power and thrust measurement techniques. These values are used to propagate the error bars reported in the power and direct thrust data. B.1 Power Measurem ent The uncertainty in the power measurement is a combination of the bias error introduced from the current and voltage probes in addition to the statistical error based on the variations in repeated measurements. The uncertainty of the instantaneous power measurement bias error (from only the measurement probes) is computed as (B 1) where the symbols V and I represents the uncertainties in the measured voltage and current, respectively. The power and current probe errors are reported as a percentage of the measured values: 3% for voltage probe and 1% for the current monitor. By discretely averaging, the power measurement bias error may be expressed as (B 2) where the average power, P avg is computed from Equation 4 1. There is also some error associated with statistically averaging sets of power measurements. The statistical error of the average power is computed based on the standard deviation and applying a 95% (double sided) confidence interval. T he statistical error is computed as (B 3)
163 where is the standard deviation of the mean power averaged o ver N data samples, and the value of the scalar t is based on a t distribution table. The total error in the time averaged power measurement is found by combining Equations B2 and B3, providing (B 4) B.2 Thrust Measurement The direct thrust measured from the torsion balance requires a more complex uncertainty analysis. Recall from Equation 4 3, the thrust is a function of the rotational spring constant, the displacement angle and length of moment arm. The displacement angle is measured using an optical sensor which has some associated uncertainty as does the length of the moment arm. However, the uncertainty in the rotational spring constant is based upon uncertainties in balance calibration which propagates through from the logarithmic decrement to the damping ratio and natural frequency parameters (as described in Section 18.104.22.168). Equation 4 3 provides an analytical expression for the thrust as a function of the displacement angle Based on the moment arm length, l = 0.0 285 m, the deflection angle is less than 1 mrad for beam displacements up to 25 m, which is greater than the maximum displacement produced by the microscale DBD actuators. This justifies the use of the small angle approximation to replace the displaceme nt angle with the ratio of the perpendicular displacement ( x ) to the moment arm length, x / l With this substitution, Equation 4 3 reduces to (B 5) The measurement uncertainty of the direct thrust measurement is computed as
164 (B 6) The optical displacement sensor has 200 nm resolution when operating using 256 averages per sample. With increased averaging, the resolution improves to its best case of 40 nm using 4096 averages per sample. The recorded data uses at least 512 averages per sample, so an uncertainty of 200 nm is used in this analysis as a worst case value for x The uncertainty in the moment arm length is based on a physical measurement, and 1 mm is c hosen as thought to be a fair value for l Recall from Equation 4 4, the rotational spring constant ( k ) is a function of the natural frequency ( o ) and moment of inertia ( MI ) of the torsion balance. Hence, t he uncertainty in the rotational spring constan t is found by propagating the uncertainties of the displacement measurement through the calculation of natural frequency as well as the uncertainty in the moment of inertia according to (B 7) where the partial derivates of k with respect to o and MI are and respectively. Plugging these terms into Equation B 7 the uncertainty in the rotational spring constant is computed as (B 8)
165 The follo wing analysis outlines the uncertainty propagation through the calibration of the torsional balance in order to extract the terms in Equation B 8 which is then used to determine the total uncertainty of the thrust measurement. Regarding the calibration of the torsional balance, which is based on the response of the system to an initial displacement as described in Section 22.214.171.124 and illustrated in Figures 4 14 and 4 15. The measured displacements are recorded in time, and the time values are presumed to b e the only known value which the measured data is referenced upon. This implies zero uncertainty in the time data in which is used to extract the damped frequency ( d ). The uncertainty in the natural frequency ( o ) stems from the uncertainty in the amplitudes of the measured displacements which propagates through the logarithmic decrement, onto the damping ratio, and finally to the natural frequency. The logarithmic decrement ( ) depends on the measured amplitudes from the uncertainty in is weighted by the individual uncertainties from each variable in Equation 4 7, according to (B 9 ) Taking the partial derivatives of with respect to y o and y m we have and respectively. As mentioned above, the uncertainty in measured displacement amplitude is 200 nm, or y ,o = y ,m = y = 2 x 10 7 m. Plugging these values into Equation B 9 we compute the log decrement uncertainty as
166 (B 10 ) From here, the uncertainty in the damping ratio ( ) is computed in a similar fashion: (B 11 ) Taking the derivative of with respect to we get Plugging this into Equation B 11 the uncertainty in the damping ratio is computed as (B 1 2 ) The uncertainty in the damping frequency ( d ) is based upon the time data which are presumed exactly known in value. Hence, x,n = x,n+1 = x = 0 sec, so the uncertainty in the damping frequency is zero. For completeness of the analysis at hand, the equation for the uncertainty in d is computed as (B 1 3 ) where x n and x n+1 are the time values at which two consecutive peaks of the damped oscillations are extracted. The partial derivatives of d with respect to x n and x n+1 are and respectively. Plugging these into Equation B 1 3 the uncertainty in d is found as (B 1 4 )
167 From these values, the uncertainty in the natural frequency o can be computed as (B 1 5 ) Taking the partial derivatives of o with respect to and d gives and respectively. Plugging these into Equation B 1 5 along with the uncertainty values for both and d we arrive at the uncertainty in o according to (B 1 6 ) the uncertainty in o which is based on the measured displacement data, the uncertainty in the moment of inertia is based solely on the geometry and materials of the balance, resulting in a constant value. The uncertainty values for both the geometry and mass of the balance are provided in Table 4 1 based on the resolution of the measuring tools and digital scale. The moment of inertia for the force balance propagates from the uncertainty in each of the three components, according to (B 1 7 ) The equations for the uncertainty for each component of the balance follow, starting with the vertical aluminum beam:
168 (B 1 8 ) where the partial derivates of MI with respect to the mass, width and depth of the beam are , and respectively. Plugging these back into Equation B 1 8 the uncertainty in the vertical aluminum beam is found as (B 1 9 ) Similarly, for the horizontal aluminum beam, the uncertainty equation is (B 20 ) where the partial derivates of MI with respect to the mass, width, depth and axial offset of the beam are , and respectively. Plugging these back into Equation B 20 the uncertainty in the horizontal aluminum beam is found as (B 21 ) Lastly, the uncertainty in the moment of inertia for the cylindrical stainless steel counterweight is (B 2 2 )
169 where the partial derivates of MI with respect to the mass, radius and axial offset of the cylinder are , and respectively. Plugging these back into Equation B 2 2 the uncertainty in the cylindrical counterweight is computed as (B 2 3 ) Equations B 19 B 21 and B 2 3 may be substituted back into Equation B 1 7 to calculate the total uncertainty in the force balance moment of inertia. At last, the uncertainty in the rotational spring constant k may now be computed from Equation B 8 Finally, the measurement uncertainty in the direct thrust measurement can be computed from Equation B 6.
170 APPENDIX C LIST OF PUBLICATIONS The following publications have reported portions of the work presented in this document. Zito JC, Arnold DP, Durscher RJ, Roy S ( 2013) Exploration of Ceramic Dielectrics for Microscale Dielectric Barrier Discharge Plasma Actuators. In: 44 th AIAA Plasmadynamics and Lasers Conference AIAA 20 13 2495 Zito JC Arnold DP Houba T Soni J Durscher RJ Roy S (2012) Microscale dielectric b arri er discharge plasma actuators: p erformance c haracterization and n umerical comparison. In : 43rd AIAA Plasmadynamics and Lasers Conference AIAA 20 12 0 391 Zito JC Durscher RJ S oni J, Roy S, Arnold DP (2012) Mechano f luidic c haracterization of m icroscal e dielectric b arrier d ischarge p lasma a ctuators In: Proceedings of the Hilton head workshop on solid state sensors, actuators, and microsystems. Zi to JC, Durscher RJ, Roy S, Arnold DP (2012) F low and f orce in ducement u sing m icron s ize d ielectric ba rrier d ischarge a ctuators Appl Phys Lett 100 : 193502 Zito JC, Arnold DP (2010) Fabrication and e lectrical c haracterization of m icroscale d ielec tric barrier discharge devices. In: Proceedings of the Hilton head workshop o n s olid state sensors, a ctuators, and micr osystems.
171 LIST OF REFERENCES Ab e T, Takizawa Y, Sato S, Kimura N ( 200 8 ) Experimental s tudy for m omentum t ransfer in a d ielectric b a rrier d ischarge p lasma a ctuator. J AIAA 46 ( 9 ): 2248 2256 Albrecht T, Weier T, Gerbeth G, Metzkes H, Stiller J (2011) A method to estimate the planar, instantaneous body force distribution from velocity field measurements. Phys Fluids 23:021702 Ashpis DE, Laun MC, Griebele r EL (2012) Progress t oward accurate measurements of power c onsum ptions of dbd plasma a ctuators. In: 50 th AIAA aerospace sciences meeting. AIAA 2012 0823 Bass A, Chevalier C, Blades MW (2001) A c apacitiv f ormed in a channel in a quartz wafer. J Anal At Spectrom 16:919 921 Baughn JW, Porter CO, Peterson BL, McLaughlin TE, Enloe CL, Font GI, Baird C (2006) Momentum transfer for an aerodynamic plasma actuator with an imposed boundary layer. In: 44 th AIAA aerospace sciences meeting. AIAA 2006 168 Benard N, Jolibois J, Moreau E, Sosa R, Artana G, Touchard G (2008) A erodynamic plasma actuators: a directional micro jet device. J Thin Fi lm Solids 516:6660 6667 Benard N, Moreau E (2012) Role of the electric waveform supplying a dielectric barrier discharge plasma actuator. Appl Phys Lett 100:193503 Campbell SA (1996) The science and engin eering of microelectronic fabrication. Oxford U niversity p ress New York, NY Cattafesta LN, Sheplak M (2011) Actuators for active flow control. Annu Rev Fluid Mech 43:247 272 Chandrasekharan V, Sells J, Arnold DP, Sheplak M (2009) Cha racterization of a mems based fl oati ng element shear stress sensor. In: 47t h AIAA aerospace s ciences meeting. Chandrasekharan V (2010) A m icroscale d ifferential c apacitive w all s hear s tress s ensor. Ph.D. thesis University of Florida Chen FF (1984) Introduction to p lasm a physics and controlled fusion. 2 nd Ed Springer Los Angeles, CA Chen T A, Mills D, Chandras ekharan V, Zmuda H, Sheplak M (2010) Optical m iniaturization of a mems b ased f loating e lement s hear s tress s ensor with m oi r. In: 48th AIAA aerospace sciences m eeting Corke TC, Post ML, Orlov DM (2007) SDBD plasma enhanced aerodynamics: c oncepts, optimization and applications. Prog Aero Sci 43:193 217
172 Corke TC, Enloe CL, Wil kinson SP (2010) Dielectric barrier d isch arge plasma actuators for flow c ontro l. Annu Rev Fluid Mech 4 2: 505 529 Crookes W (1879) On the illumination of lines of molecular pressure, and the trajectory of molecules. Phil Trans R Soc Lond 170:135 164 Debien A, Benard N, David L, Moreau E (2012) Unsteady aspect of the electrohydrodynamic force produced by surface dielectric barrier discharge actuators. Appl Phys Lett 100:013901 Durscher RJ, Roy S (2011) Induced flow from serpentine plasma actuators acting in quiescent a ir In: 49th AIAA a erospace sciences m eeting Durscher R, Stanfield S, Roy S (2012) Characterization and manipulation of the discharge actuator. Appl Phys Lett 101:252902 Durscher RJ, Roy S (2012) Ev aluation of thrust measurement t echniques fo r dielectric barrier discharge a ctuators E xp Fluids 53:1165 1176 Enloe CL, McLaughlin TE, VanDyk en RD, Kachner KD, Jumper EJ, Corke TC (2004a) Mechanisms and r e sponses of a single dielectric barri er plasma actuator: plasma m orpho logy. J AIAA 42(3):589 594 Enloe CL, McLaughlin TE, VanDyk en RD, Kachner KD, Jumper EJ, Corke TC, Post M, Haddad O (2004b) Mechanisms and r e sponses of a single dielectric b arr ier plasma actuator: geometric e ff ects. J AIAA 4 2(3):595 604 Enloe CL, McLaughlin TE, Font GI, Baughn JW (2006) Parameterizatio n of temporal structure in the s in gle d ielect ric barrier a erodynamic plasma a ctuator J AIAA 44(6):1127 1136 Enloe CL, M cHarg MG, McLaughlin TE (2008) Time c orrelated force production m easurements of th e dielectric barrier discharge p lasma aerodynamic actuator. J Appl Phys 103:073302 Enloe CL, McH arg MG, Font GI, McLaughlin TE (2009) Plasma induced f orce and s elf induced drag in the dielectric barrier discharge aerodynamic pl asma a ctu a tor In: 47 th AIAA aerospace sciences m eeting Fairchild Corporation (1979) Semiconductor and integrated circuit fabrication techniques. Reston Reston, VA Font GI, McLaughlin TE (2010) Pla sma volumetric effects of the f orce p roduction of a p l asma actuator. J AIAA 48 (9) :1869 1874
173 Forte M, Jolibois J, Pons J, Moreau E, Touchard G, Cazalens M (2007) Optimization of a d ielectric barrier discharge actuator by s tationary a nd non stationary m easurements of the induced flow velocity: a pp lication to airflo w control. J Exp Fluids 43:917 928 Gamero Castano M (2003) A torsional balance for the characterization of microNewton thrusters. Rev Sci Inst 74 (10) :4509 4514 Germer LH (1959) Elec trical breakdown between close e lectrodes in air. J Appl Phys 30(1): 46 51 Gianchandani YB, Wright SA, Eun CK, Wilson CG, Mitra B (2009) Exploring microdischarges for p orta ble sensing applications. Anal Bio anal Chem 395:559 575 Gregory JW, Enl oe CL, Font GI, McLaughlin TE (2007) Force production m echanisms of a d iele ctric b ar rier discharge plasma actuator. In: 45 th AIAA aerospace s cien ces meeting. Guman WJ, Nathanson DM (1970) Pulsed plasma microthrusters propulsion system for synchronous obit satellite. J Spacecraft 7(4):409 415 HD MicroSystems (2008) Technical information: PI 2600 series low stress polyimides. Heinrich V, Waugh RE (1996) A p ico N ewton force transducer and its a pplication to measurement of the bending s tiffness of p hospholipid m embr anes. Ann Biomed Eng 24 : 595 605 Hopwood JA (2000) A m icrofabrica ted inductively coupled plasma g enera tor. J MEMS 9:309 313 Horowitz S, Chen T A, Cattafesta, LN, Sheplak M Nishida T, Chandrasekaran V (2004) Optical f low s ensor u sing g eometric m oir i nterferometry. In: 34th international conference on environmental systems. Hoskinson AR, He rshkowitz N, Ashpis DE (2008) Force m easurements of single and double barrier dbd p lasm a a ctua tors in quiescent air. J Phys D Appl Phys 41:245209 Hoskinson AR, Hershkowitz N (2010) Differences between dielectric barrier discharge plasma actuators wit h cylindrical and rectangular exposed electrodes. J Phys D Appl Phys 43:065205 Hutchinson IH (2002) Principles of p lasma d iagnostics 2 nd Ed University Press Cambridge, UK
174 Jolibois J, Moreau E (2009) Enhancement of the e lectromech anical p erformance of a s ingle dielectric barrier d ischarge actuator. IEEE Trans Dielectr El ectr Insul 16:758 767 Kogelschatz U (2003) Dielectric b arrier d ischarges: t heir h istory, d ischarge p hysics, and i ndustrial a pplications Plasma Chem Plasma Process 23 :1 46 Kotsonis M, Gh aemi S, Veldhuis L, Scarano F (2011) Measurement of the body force fi el d of plasma actuators. J Phys D Appl Phys 44:045204 Kriegseis J, Schroter D, Grundmann S, Tropea C (2011a) Online characterization of dielectric barrier discharge plasma actuators for o ptimized efficiency of aerodynamical flow control applications. J Phys Conf Series 301:012020 Kriegseis J, Grundmann S Tropea C (2011 b ) Power c onsumption, d ischarge c apacitance and l ight e mission m easurements for t hrust p roduction of d ielectric b arrier di scharge plasma actuators. J Appl Phys 110:013305 Kriegseis J, Schroter D, Barckmann K, Duchmann A, Tropea C, Grundmann S (2013a) Closed loop p erformance control of dielectric barrier d isch arge plasma actuators. J AIAA 51 ( 4 ) :961 966 Kriegseis J, Moller B, G rundmann S, Tropea C (2013b) On performance and efficiency of dielectric barrier discharge plasma actuators for flow control applications. Intl J Flow Control 4(3):125 132 Kriegseis J, Schwarz C, Tropea C, Grundmann S (2013c ) Velocity i nformation based force term estimation of d iele ctric barrier discharge plasma actuators. J Phys D Appl Phys 46:055202 Langmuir I (1928) Oscillations in ionized gases. In: Proc Nat Academy of Science USA 14( 8 ): 627 637 Lieberman MA, Lichtenberg AJ (2005) Principles of plasma d isc harges and materials p rocessing 2 nd Ed Wiley Hoboken, NJ Likhanskii AV, Shneider MN, Opaits DF, Miles RB, Macheret SO (2010) Limitations of the dbd effect on the external flow. In: 48th AIAA aerospace sciences m eeti ng. AIAA 2010 0470 Longwitz RG (200 4) Study of gas ionization in a glow d ischa rge and development of a micro g as ionizer for gas detection and a nalysis. Ph.D. t hesis, Swiss Federal Institute of Technolo gy Madou M (1997) Fundamental of microfabrication. CRC p ress New York, NY Manley TC (1943) The electric characteristics of the ozonator d ischarge J Electrochem Soc 84 (1) :83 96
175 Marcus RK, Davis WC (2001) An atmospheric pressure glow discharge optical emission source for direct sampling of liquid media. Anal Chem 73(13):2903 2910 Mina yeva OB, Hopwood JA (2002) Emission s pectroscopy using a microfabricated inductively coupled plasma on a chip. J Ana l At Spectrom 17:1103 1107 Mitra B, Gianchandani YB (2008) The d etection o f chemical vapors in air using o p tical emission spectroscopy of p ulsed microdischarges from two and t hree electrode microstructures. J IEEE Sensors 8( 8 ): 1445 1454 Moreau E (2007) Airflow control by non thermal plasma a ctuators J Phys D Appl Phys 40: 605 636 Neumann M, Friedrich C, Czarske J, Kriegseis J, Grundmann S ( 2013 ) Determination of the phase resolved body force produced by a d iele ctric barrier discharge plasma a ctua tor. J Phys D Appl Phys 46:042001 Okochi S, Kasagi N, Suzuki Y, Ito S (2009) Development of m icro p lasma a c tuator for a ctive f low c ontrol In: 7 th world conf e xp heat transfer, fluid mechanics and thermodynamics. Ono T, Sim DY, Esashi M (2000) Micro discharge and electric b reakdow n in a micro gap. J Micromech Microegn 10:445 451 Opaits DF, Likhanskii AV, Neretti G, Zaidi S, Shneider M N, Miles RB, Ma cheret SO (2008) Experimental investigation of dielectric b arrier d isc harge plasma a ctuators driven by r epetitive high v oltage n anosecond p ulses with dc or low frequency sinusoidal bias. J Appl Phys 104: 043304 Paschen F (1889) Ueber die zum f unkenbergang in l uft, w asserstoff und k ohlensure bei verschiedenen d rucken erforderliche p otentialdiffe renz. Annalen der Physik 273:69 96 Pons J, Moreau E, Touchard G (2005) Asymmet ric surface dielectric barrier d ischarge i n air at atmospheric pressure: e lectrical properties and induced airflow c haracteristics J Phys D Appl Phys 38:3635 3642 Porter CO, Baughn JW, McLaughlin TE, Enloe CL, Font GI (2007) Plasma actuator force measurements. J AIAA 45(7):1562 1570 Raizer YP (1991) Gas d ischarge p hysics Springer Roth JR, Sherman DM, Wilkinson SP (2000) Electrohydrodynamic flow control with a glow discharge surface plasma. J AIAA 38(7):1166 1172 Roth JR (2003) Aerodynamic flow acceleration using paraelectric and peristaltic electrohydrodynamic effects of a One Atmosphere Uniform Glow Discharge Plasma. Phys Plasmas 10(5):2117 2126
176 Roth JR, Dai X (2006) Optimization of the a erodynamic p lasma a ctuator as an e lectrohydrodynamic ( ehd ) e lectrical d evice In: 49 th AIAA a erospace sciences m eeting and e xhibit Santh anakrishnan A, Jacob JD (2007) f low c ontrol with p lasma s ynthetic j et a ctuators. J Phys D Appl Phys 40 : 637 651 Senturia SD (2001) Microsystem design. Kluwer academic publishers Norwell, MA Taylor S, Srigengan B, Gibson J, Tindall D, Syms R, Tate T, Ahman M (2000) In: Proceedings of SPIE, chemical and biological sensing. Thomas FO, Corke TC, Iqbal M, Kozlov A, Schatzman D (2009) Optimization of d iele ctric barrier discharge plasma a ctuators for a ctive a erodynamic f low c ontrol. J AIAA, 47 ( 9 ): 2169 2178 Thomson JJ (1897) Cathode Rays. Phil Mag Series 5(44):293 316 Torres JM, Dhariwal RS (1999) Electric f ield b reakdown at m icrometre s eparations in a ir and v acuum. J Microsys Tech 6 : 6 10 TSI (2000) Model 93 02 atomizer, i nstruction m anual TSI Inc, St Paul, MN Versai lles P, Gingras Gosselin V, Vo HD (2010) Impact of pressure and temperature on the performance of plasma actuators. J AIAA Tech Notes 48(4):859 863 Wang C C, Roy S (2009a) Flow shaping using three dimensional microscale gas discharge. Appl Phys Lett 95:081 501 Wan g C C, Roy S (2009 b ) Microscale plasma a ctuato rs for improved thrust d ensity. J Appl Phys 106 :013310 Wright SA, Gianchandani YB (2009) Discharge based pressure sensors for high temperature applications using three dimensional and planar microstructu res. J Microelectromech Syst 18(3):736 743 Zito JC, Arnold DP (2010) Fabrication and e lectrical c haracterization of m icroscale d ielec tric b arrier d ischarge d evices. In: Proceedings of the Hilton h ead workshop on s olid state sensors, a ctuators, and microsystems. Zi to JC, Durscher RJ, Roy S, Arnold DP (2012) F low and f orce in ducement u sing m icron s ize d ielectric ba rrier d ischarge a ctuators Appl Phys Lett 100 : 193502
177 BIOGRAPHICAL SKETCH Justin C. Zito was born in Queens, New York City, in November of 1982. He received his Bachelor of Science and Master of Science in e lectrical e ngineering from the University of Florida, Gainesville, FL, in 2006 and 2009, respectively. He earn ed his D octor of P hilosophy from the Department of Electrical and Computer Engineering in 2013 also from the University of Florida. From 2008 2013 he was a Graduate Research Assistant and members of both the Interdisciplinary Microsystems Group (IMG) and the Applied Physics Research Group (APRG) within the University of Florida. He has served as Lab Manager and member of the IMG safety committee from 2010 2012. Mr. Zito is the auth or of seven conference publications and two journal articles, his research group to the student having the most hours logged annually for working in the Nanosc al e Research Facility (NRF) cleanroom at UF. The author most enjoys spending time with his girlfriend Lauren and their two dogs, and also playing disc golf. He will be joining Intel Corporation upon the completion of his doctorate degree.