Design, Fabrication, and Characterization of Carbon Nanotube Field Emission Devices for Advanced Applications

Home , John Ambrose Fleming

Design, Fabrication, and Characterization of Carbon

Nanotube Field Emission Devices for Advanced Applications

Erich Justin Radauscher

Department of Electrical and Computer Engineering Duke University

Date:______Approved:

______Jeffrey T. Glass, Supervisor

______Brian R. Stoner

______Adrienne D. Stiff-Roberts

______Aaron D. Franklin

______Martin A. Brooke

Dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Electrical and Computer Engineering in the Graduate School of Duke University

2016

ABSTRACT

Design, Fabrication, and Characterization of Carbon

Nanotube Field Emission Devices for Advanced Applications

Erich Justin Radauscher

Department of Electrical and Computer Engineering Duke University

Date:______Approved:

______Jeffrey T. Glass, Supervisor

______Brian R. Stoner

______Adrienne D. Stiff-Roberts

______Aaron D. Franklin

______Martin A. Brooke

An abstract of a dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Electrical and Computer Engineering in the Graduate School of Duke University

2016

Abstract

Carbon nanotubes (CNTs) have recently emerged as promising candidates for electron field emission (FE) cathodes in integrated FE devices. These nanostructured carbon materials possess exceptional properties and their synthesis can be thoroughly controlled. Their integration into advanced electronic devices, including not only FE cathodes, but sensors, energy storage devices, and circuit components, has seen rapid growth in recent years. The results of the studies presented here demonstrate that the

CNT field emitter is an excellent candidate for next generation vacuum microelectronics and related electron emission devices in several advanced applications.

The work presented in this study addresses determining factors that currently confine the performance and application of CNT-FE devices. Characterization studies and improvements to the FE properties of CNTs, along with Micro-Electro-Mechanical

Systems (MEMS) design and fabrication, were utilized in achieving these goals.

Important performance limiting parameters, including emitter lifetime and failure from poor substrate adhesion, are examined. The compatibility and integration of CNT emitters with the governing MEMS substrate (i.e., polycrystalline silicon), and its impact on these performance limiting parameters, are reported. CNT growth mechanisms and kinetics were investigated and compared to silicon (100) to improve the design of CNT

emitter integrated MEMS based electronic devices, specifically in vacuum microelectronic device (VMD) applications.

Improved growth allowed for design and development of novel cold-cathode FE devices utilizing CNT field emitters. A chemical ionization (CI) source based on a CNT-

FE electron source was developed and evaluated in a commercial desktop mass spectrometer for explosives trace detection. This work demonstrated the first reported use of a CNT-based ion source capable of collecting CI mass spectra. The CNT-FE source demonstrated low power requirements, pulsing capabilities, and average lifetimes of over 320 hours when operated in constant emission mode under elevated pressures, without sacrificing performance. Additionally, a novel packaged ion source for miniature mass spectrometer applications using CNT emitters, a MEMS based Nier-type geometry, and a Low Temperature Cofired Ceramic (LTCC) 3D scaffold with integrated ion optics were developed and characterized. While previous research has shown other devices capable of collecting ion currents on chip, this LTCC packaged MEMS micro-ion source demonstrated improvements in energy and angular dispersion as well as the ability to direct the ions out of the packaged source and towards a mass analyzer.

Simulations and experimental design, fabrication, and characterization were used to make these improvements.

Finally, novel CNT-FE devices were developed to investigate their potential to perform as active circuit elements in VMD circuits. Difficulty integrating devices at

micron-scales has hindered the use of vacuum electronic devices in integrated circuits, despite the unique advantages they offer in select applications. Using a combination of particle trajectory simulation and experimental characterization, device performance in an integrated platform was investigated. Solutions to the difficulties in operating multiple devices in close proximity and enhancing electron transmission (i.e., reducing grid loss) are explored in detail. A systematic and iterative process was used to develop isolation structures that reduced crosstalk between neighboring devices from 15% on average, to nearly zero. Innovative geometries and a new operational mode reduced grid loss by nearly threefold, thereby improving transmission of the emitted cathode current to the anode from 25% in initial designs to 70% on average. These performance enhancements are important enablers for larger scale integration and for the realization of complex vacuum microelectronic circuits.

Dedicated to my dearest wife, Rebecca

To my beloved parents, Erich and Sandra, and brother, Matthew

And

To my amazing baby boy, Aiden

vii

Table of Contents

Abstract ...... iv

Table of Contents ...... viii

List of Tables ...... xiv

List of Figures ...... xv

List of Symbols and Abbreviations ...... xxii

Acknowledgements ...... xxiii

1. General Introduction ...... 1

1.1 Electronic Materials for Cold Cathode Field Emission ...... 2

1.2 Vacuum Microelectronics Overview ...... 7

1.3 Motivation and Objectives of this Research...... 11

1.4 Dissertation Outline ...... 12

2. Background - Carbon Nanotubes (CNT), Field Emission (FE), FE Vacuum Microelectronic Devices (FE-VMDs), and Mass Spectrometry (MS) ...... 15

2.1 Introduction ...... 15

2.2 Carbon Nanotubes ...... 15

2.2.1 Significance of CNT Research ...... 16

2.2.2 Structure ...... 17

2.2.3 Growth Mechanisms of CNTs ...... 19

2.2.4 Properties of CNTs ...... 21

2.2.4.1 Mechanical ...... 21

2.2.4.2 Electronic ...... 22

viii

2.2.5 CNT Applications ...... 24

2.3 Field Emission Fundamentals ...... 25

2.3.1 Electron Emission ...... 26

2.3.1.1 Probability of Escape ...... 26

2.3.1.2 Emission Processes ...... 28

2.3.1.3 Emission Metrics Specific to FE ...... 31

2.3.2 The Fowler-Nordheim Theory ...... 32

2.3.2.1 Correction to the F-N Equation ...... 37

2.3.3 Field Emission from CNTs ...... 38

2.3.3.1 Unique FE Properties of CNTs ...... 38

2.4 Field Emission Devices ...... 42

2.4.1 FE Device Classifications ...... 43

2.4.1.1 Spindt Cathode ...... 43

2.4.1.2 Silicon Field Emitters ...... 44

2.4.1.3 Carbon Field Emitters...... 45

2.4.2 VMD Operating Modes ...... 47

2.4.2.1 Vacuum Diodes ...... 47

2.4.2.2 Vacuum Triodes ...... 48

2.4.3 Present VMD Challenges ...... 48

2.4.4 Duke/RTI MEMS VMD ...... 49

2.4.4.1 Vertical and Lateral VMDs ...... 50

2.4.4.2 History and Previous Work ...... 50

2.4.4.3 Microelectromechanical Systems (MEMS) ...... 52

2.4.4.4 Device Assembly ...... 53

2.4.4.5 Device Testing ...... 54

2.5 Mass Spectrometry (MS)...... 55

2.5.1 Introduction to Basic Concepts ...... 56

2.5.2 Ionization Processes ...... 57

2.5.2.1 Electron Ionization ...... 57

2.5.2.2 Chemical Ionization ...... 58

2.5.2.3 CNT Emitters for Mass Spectrometry ...... 59

3. Chemical Ionization Mass Spectrometry Using Carbon Nanotube Field Emission Electron Sources ...... 61

3.1 Introduction ...... 63

3.2 Experimental Methods ...... 66

3.3 Results and Discussion ...... 70

3.3.1 Ex-Situ Characterization ...... 70

3.3.2 Integration into the Desktop Mass Spectrometer ...... 74

3.3.3 Evidence for Chemical Ionization ...... 76

3.4 Conclusion ...... 80

4. Integrating Carbon Nanotube Forests into Polysilicon MEMS: Growth Kinetics, Mechanisms, and Adhesion ...... 82

4.1 Introduction ...... 84

4.2 Experimental ...... 87

4.2.1 Substrate Preparation ...... 87

4.2.2 Carbon Nanotube Growth ...... 89

4.2.3 Characterization Techniques ...... 90

4.2.4 Adhesion Testing ...... 91

4.3 Results and Discussion ...... 92

4.3.1 CNT Growth Kinetics and Activation Energies ...... 92

4.3.2 Diffusion Mechanisms ...... 98

4.3.3 Causes of Reduced Diffusivity for CNTs grown on Polysilicon ...... 105

4.3.4 Improving CNT Adhesion to Polysilicon ...... 108

4.4 Conclusions ...... 111

5. Improved Performance of Field Emission Vacuum Microelectronic Devices for Integrated Circuits ...... 113

5.1 Introduction ...... 114

5.2 Experimental Methods ...... 118

5.3 Results and Discussion ...... 121

5.3.1 Determination of Proximity Effects ...... 121

5.3.2 Simulation Guided Analysis and Redesign ...... 124

5.3.3 Simulation for Electron Transmission Analysis and Redesign ...... 128

5.3.4 Experimental Confirmation ...... 130

5.4 Conclusion ...... 133

6. Low Temperature Co-fired Ceramic Packaged MEMS Micro-ion Source for Miniature Mass Spectrometry Applications ...... 135

6.1 Introduction ...... 137

6.2 Design and simulation ...... 140

6.2.1 Components of a Nier-type Ion Source ...... 140

6.2.2 High Performance Ion Source Characteristics ...... 141

6.2.2.1 Ionization Volume Impact on Resolution ...... 142

6.2.2.2 Parallel Equipotential Surfaces ...... 143

6.2.3 Proposed Solution ...... 144

6.2.3.1 MEMS ...... 144

6.2.3.2 LTCC ...... 146

6.3 Design Details ...... 149

6.3.1 MEMS ...... 149

6.3.1.1 Observed Failure Mechanisms and Mitigation Techniques ...... 150

6.3.1.2 Cathode and Grid ...... 154

6.3.1.3 Repeller (i.e., Floor) ...... 155

6.3.1.4 Wire Bonding Pads ...... 156

6.3.1.5 Latches ...... 156

6.3.2 LTCC ...... 157

6.3.2.1 LTCC Layer Specifics ...... 157

6.3.2.2 Assembly Details ...... 160

6.3.2.3 First Generation ...... 161

6.3.2.4 Second Generation ...... 162

6.4 Characterization of Packaged Micro-ion Source ...... 165

6.4.1 Electrical Characterization ...... 165

6.4.1.1 Electrons ...... 165

xii

6.4.1.2 Ions ...... 168

6.4.2 Dispersion Measurements ...... 169

6.4.2.1 Simulation ...... 169

6.4.2.2 Experimental ...... 170

6.5 Conclusion ...... 173

7. Summary and Recommendations for Future Work ...... 174

7.1 CIMS CNT Electron Source ...... 174

7.1.1 Recommendations for Further Investigation ...... 175

7.2 CNTs and Poly-Si: Kinetics, Mechanisms, Adhesion ...... 178

7.2.1 Recommendations for Further Investigation ...... 179

7.3 Improved Performance of FE-VMDs for IC ...... 180

7.3.1 Recommendations for Further Investigation ...... 181

7.4 LTCC Packaged MEMS Micro-ion Source ...... 183

7.4.1 Recommendations for Further Investigation ...... 185

Appendix A...... 187

Appendix B ...... 193

References ...... 197

Biography ...... 227

xiii

List of Tables

Table 1: Comparing threshold electric field requirements for different materials; reproduced from [98]...... 42

Table 2: Primary FE Device Applications ...... 43

xiv

List of Figures

Figure 2.1: Bonding structure of a carbon nanotube. The sp2 hybridized bond (s bond) and the out of plane orbital p bond are shown. Diagram derived from [43]...... 18

Figure 2.2: (a) TEM image of the tips of one MWCNT with bamboo structure, reproduced from [58]; (b) schematic diagram of the 915 MHz MPECVD system used to grow vertically aligned MWCNTs in this study, reproduced from [57]...... 21

Figure 2.3: A free electron (red sphere in diagram) possesses a lower energy inside a metal than in vacuum, the work function is the amount of energy needed by the electron to escape into vacuum. Adapted from [88] with relevant modifications...... 27

Figure 2.4: Various electron emission processes; adapted and modified from [91]...... 29

Figure 2.5: CNT field emitter characterization as reported by [98]. Electron emission current versus applied field characteristics for a CNT field emission sample (inset). The F-N plot, as derived by equation 10, is shown...... 37

Figure 2.6: (a) schematic showing the spatial distribution of an electric field potential around a sharp tip, (b) SEM of high aspect ratio CNT emitters with tips shown in the inset, (c) tungsten field emitter from a SEM reproduced from [108]...... 39

Figure 2.7: Potential-energy diagram illustrating the effect an applied electric field has on the energy barrier of a planar emitter versus a micro-tip emitter...... 40

Figure 2.8: (a) Spindt-type field emitter. SEM of single emitter shown; reproduced from [2]; (b) Deep reactive ion etched (DRIE) silicon field emitter arrays; reproduced from [117]...... 44

Figure 2.9: (a) SEM of a microfabricated nanodiamond lateral FE diode reproduced from [15]; (b) SEM of a CNT emitter array utilized in this work...... 46

Figure 2.10: Schematic diagrams representing electrode type and placement in a simple diode and triode configuration...... 47

Figure 2.11: SEM’s showing evolving development of the Duke/RTI VMD platform; (a) original concept as fabricated with MEMS platform and CNT emitters, (b) adoption of CNT FE-VMD as an ion source with ion collection on chip laterally, (c) adoption of the CNT FE-VMD for dual-charge carrier operation; reproduced from [145, 146, 148, 151]. . 52

Figure 2.12: (a) PolyMUMPS process flow with assembly; (b) device on test board...... 54

Figure 2.13: UHV test system used for field emission device testing...... 55

Figure 2.14: Schematic diagram of a mass spectrometer...... 57

Figure 2.15: Schematic diagrams showing the ionization process in both, (a) electron ionization, (b) and chemical ionization...... 59

Figure 3.1 Images of the CNT electron source components (a) SEM image of CNT field emitters as grown on the substrate, (b) photograph of the Duke/FLIR CNT electron source platform set inside its wiring harness mount, (c) photograph of CNT field emitters mounted into the platform. The source is shown before and after placement of extraction mesh...... 68

Figure 3.2 Electrical characteristics of the carbon nanotube electron source (a) I-V characteristics showing three consecutive and increasing voltage sweeps, (b) Fowler- Nordheim plot of I-V curve #3 with nonlinearity of the slope attributed to a nonlinear field enhancement factor, (c) Lifetime analysis operating at a pressure of 1x10-2 Pa (1x10-4 Torr) of room air and a fixed 5 µA CNT emission current...... 71

Figure 3.3 Integration of the carbon nanotube electron source into the explosive trace detection mass spectrometer. (a) Schematic of the ionization region showing source location, (b) photograph of the desktop mass spectrometer, (c) a magnified photograph of the ionization region showing the CNT field emission electron source installed...... 75

Figure 3.4: Trace explosive chemical ionization spectra collected using both the CNT field emission (left column) and the filament thermionic emission (right column) sources in the explosive trace detection desktop MS. The spectra for TNT are shown in (a) and (d), the spectra for RDX are shown in (b) and (e), and the spectra for PETN are shown in (c) and (f)...... 80

Figure 4.1 Morphology of carbon nanotubes simultaneously grown on a) Si (100) and b) polysilicon imaged by SEM. c) TEM micrograph of a CNT with bamboo-type inner walls, with inset depicting a catalyst iron nanoparticle embedded within the CNT base; d) schematic of field emission test chamber to determine poor CNT adhesion...... 90

xvi

γ-Fe catalyst nanoparticles. The slope in the parabolic region differs between CNTs on Si (100) and polysilicon, leading to different activation energies for the diffusion-limited phase of growth...... 95

Figure 4.3: CNT forest height kinetics at 850 °C. The solid curves correspond to the fit to the Deal and Grove expression, with A and B values listed...... 97

Figure 4.4: CNT forest height as a function of growth temperature on polysilicon MEMS devices. All CNT forests were grown for 150 s...... 100

Figure 4.5: Raman spectra of CNTs grown on Si (100) and polysilicon substrates. The crystallite size is 66 Å for CNT on Si (100) and 201 Å for CNTs on polysilicon...... 102

Figure 4.6: Plan view SEM micrograph of graphenated CNTs and b) higher magnification image of a g-CNT. Inset: TEM micrograph of a few-layered graphene foliate terminating in 3-5 graphene layers...... 104

Figure 4.7: Catalyst nanoparticle arrays produced by dewetting at 850 °C in an ammonia plasma environment on a) Si(100) and b) polysilicon. A typical particle size of 65 nm was observed on Si (100) and at the top of polysilicon grains, and smaller nanoparticles of approximately 25 nm diameter were observed at the polysilicon grain boundaries. 107

Figure 4.8: (a) Si (100) and (b) poly-Si. The CNTs on poly-Si appear to display a higher degree of tortuosity, which could explain why the diffusivity is markedly lower for growth on polysilicon and the parabolic activation energy is larger...... 108

Figure 4.9: Average reflectance from the underlying substrate as measured by spectrophotometer for CNTs grown using various metallic interlayers. The error bars represent standard deviation. An adhesion improvement of approximately 1.3x was observed for Mo interlayers, and approximately 2.5x for Ti interlayers compared with the control sample of CNTs grown on Fe-coated poly-Si...... 111

Figure 5.1: (a) schematic diagram of dual CNT FE-VMD illustrating the physical layout of panel electrodes with device #1 (left) and device #2 (right), location of CNT emitters, and the biasing method. The labeled IN’s and OUT represent the potential inputs/output for logic circuit realization; (b) SEM image of the fabricated device with electrode panels rotated up and locked in place as described in [147, 271]. An SEM image of the CNT field emitter arrays as grown on the cathode are shown in the inset...... 120

xvii

Figure 5.2: Current-voltage characteristics of a dual CNT FE-VMD revealing the undesired proximity effects; (a) both devices are operating in the ON state (0 V on the cathode, 200 V on the anode, and extraction potentials on the grids swept) with no adverse effects; (b) device 1 in ON state with ground potential on the cathode, 200 V on the anode, and the extraction potential on the grid swept from 0 V to 95 V, device 2 (inset) is in the nominally OFF state with 0 V on the cathode and 0 V on the grid; the signal observed at the anode of device 2 is due to crosstalk (i.e. current from device 1’s emitters being pulled across the vacuum channel)...... 124

Figure 5.3: A schematic diagram illustrating the potentials on each electrode and the SIMION® simulation demonstrating the proximity effects experienced by the dual CNT FE-VMD; (a) device 1 is biased for field emission (ON state) and device 2’s grid is grounded, rendering it in a nominally OFF state; (b) simulation results show that even though device 2 is OFF, its anode is receiving 10-15% of the electron current from device 1 (similar to experimental results) resulting in an undesired ON state signal at anode 2...... 126

Figure 5.4: SIMION® simulation modeling with optimized center shielding electrode to eliminate the crosstalk between neighboring devices; (a) two proximal devices with the shielding electrode off and crosstalk taking place from the ON device 1 (left) to the OFF device 2 (right); (b) same design with the shielding electrode grounded, effectively shielding the devices and their corresponding vacuum channels from one another. Operating the shielding electrode at ground potential provided the best shielding without interfering with individual device performance...... 127

Figure 5.5: SIMION® simulation modeling of electron transmission properties of the CNT FE-VMDs; (a) Simulated electron trajectories in the original microtriode device with an average transmission of 25% (electrons passing through grid and collecting at anode); (b) simulation showing equipotential lines of new geometry designed to create pre-lens effect around CNT emitter bundles thereby enhancing focusing properties; (c) simulated electron trajectories of new geometry showing an increased grid transmission of 80%...... 130

Figure 5.6: (a) schematic diagram illustrating new physical layout of panel electrodes with device #1 (left) and device #2 (right), location of CNT emitters, biasing method, and new potential inputs/outputs for logic circuit realization with decoupled cathode and extraction grid; (b) SEM image of redesigned fabricated device with electrode panels rotated up and locked in place and center shielding electrode flat on the substrate between the devices...... 131

xviii

Figure 5.7: Center shielding electrode effect on proximity effects. Left device is ON and the right device is OFF; (a) with shielding electrode grounded, crosstalk current on the order of 10-15% is measured at the anode of the OFF device; (b) with the shielding electrode electrically grounded the crosstalk current measured at the anode of the OFF device is decreased by an order of magnitude (1-2% on average)...... 132

Figure 5.8: (a) original 3-panel microtriode operated with cathode = 0V, extraction grid swept (0-140 V), and anode = 150 V. Pie chart inset shows average distribution of emitted current at maximum extraction potential (25% reached the anode, 58% intercepted by the grid, and 17% lost); (b) 4-panel geometry driven in the new operating mode with cathode = 0V, screening electrode = 20 V, extraction grid swept (0-125 V), and anode = 150 V. On average, 67% of the current reached the anode, 23% was intercepted by the grid, and 10% was lost...... 133

Figure 6.1: Schematic diagram of a Nier-type geometry ion source...... 140

Figure 6.2: schematic showing the two important characteristics of a high performance ion source; (a) ions should be formed in a region of small potential difference, or zero electric field, in order to reduce energy dispersion; (b) the equipotential field lines above the ionization region should be flat in order to minimize angular dispersion...... 144

Figure 6.3: SEM failure analysis of microtriodes; (a) spring failure due to overcurrent, (b) panel snapping from high fields or mechanical failure, (c) loss of CNT adhesion and cathode-grid shorting, (d) latch failure with panel snapping...... 151

Figure 6.4: (a) Top-down schematic of new micro ion source; (b) SEM of fully assembled symmetric micro ion source; CNT field emitters (inset) are synthesized to specific lengths directly on cathode prior to assembly; (c) SEM image of precise alignment between CNT emitter bundles and grid holes; (d) SEM image showing thicker latch arms (current carriers) and improved electrical contact point...... 155

Figure 6.5: (a) Exported CAD image showing exploded view of the LTCC layers with labels indicating various components, (b) cutaway view of the assembled LTCC package...... 158

Figure 6.6: Exploded view of the 2nd generation of the LTCC scaffold with labels ...... 159

Figure 6.7: (a) Magnified image of the 1st layer with MEMS micro-ion source secured within the cavity, SEM images of source given; (b) cutaway view of the complete

xix

scaffold with locations of apertures shown, SEM images of extraction and coded aperture shown...... 160

Figure 6.8: Optical images of the 1st generation of the packaged miniature ion source; (a) device secured in the cavity and wire bonded to the 2nd layer of the LTCC, (b) bottom half of LTCC secured onto test board, (c) after capping with final two layers, packaged device is mounted upside down on PCB and bonded for testing, (d) a copper plate is used to collect emitted ions, (e) the UHV testbed used...... 162

Figure 6.9: Optical image of the 2nd generation of the LTCC scaffold prior to the 1st and 2nd stage of electrical characterization; a) before capping; b) after capping...... 163

Figure 6.10: simulation results of 2nd generation miniature ion source; (a), (c), and € provide a side view; (b), (d), and (f) provide a view from the center of the floor looking towards a grid; (a) and (c) show the device with no field lines or ions plotted; (c) and (d) show the field lines plotted; (e) and (f) show field lines and simulated ion trajectories plotted; (g) and (h) are top down views with field lines and ion spatial mapping through the coded spatial filter, respectively...... 164

Figure 6.11: Electrical characterization performed on a device prior to capping; (a) IV characteristics; (b) Triode characteristics; (c) Fowler-Nordheim model fitted to IV data to verify field emission...... 166

Figure 6.12: These three plots illustrate the process of burn-in on a device; (a) An initial I- V curve is taken showing a VTO of approximately 160 V, (b) a one hour burn-in is conducted where the VTO is reduced, (c) improved I-V curve after burn-in...... 167

Figure 6.13: Electrical characterization plots of a packaged MEM micro-ion source; both turn-on and threshold voltage I-V analysis plots are shown...... 168

Figure 6.14: Ion current measurements from a packaged micro-ion source. Chamber pressure was increased until ion current was detected. Repeller and extraction aperture voltages were then cycled to establish a periodic measurement of ions...... 169

Figure 6.15: SIMION energy and angular dispersion results for the packaged device. Ion trajectories, plotted in tan, are focused vertically off-chip and through the series of apertures and through the outlet which would lead to a mass analyzer. The energy dispersion is 14 +/- 4 eV. Angular dispersion is approximately 13 degrees...... 170

Figure 6.16: Dispersion measurements of the packaged micro-ion source; (top) schematic diagram showing arrangement of electrodes on the MEMS device, apertures, and orientation with the detector, and camera picture of the packaged micro-ion source secured in the aluminum mount; (middle) magnified camera packaged source directed at the detector for dispersion measurements, and image of the ion exit point showing the strike points on the detector during angular dispersion measurements; (bottom) angular dispersion data on the detector showing the average angular dispersion of 13 degrees, and energy dispersion data showing ion current on detector on the left y-axis, sweeping retarding grid potential on the x-axis, and the derivative of the stopping potential data on the right y-axis. Average energy of the ions was 22 eV with an average energy dispersion of 6 eV...... 172

Figure 7.1: (top left and right) Camera and SEM images of several embodiments of the preliminary device fabricated on Si substrates; (bottom left and right) electrical characterization shows performance capabilities of 8-10x better than previous design. 178

Figure 7.2: (a) Schematic diagram of logic OR gate configuration for CNT-FE VMD; (b) microscope image of designed/fabricated quad logic device test platform; (c) SEM image of designed/fabricated oscillator circuit composed of four CNT-FE VMDs...... 183

Figure 7.3: (top left) Schematic diagram of ion trajectories leaving the packaged ion source and traveling a cycloidal path within the electric sector due to the orthogonal uniform electric and magnetic field; (top right) CAD image of electric sector, cutaway view shown; (bottom left) CAD image of electric sector situated in its guide with the detector in place; (bottom right) entire assembly placed inside a vacuum manifold; the magnet (not shown) wraps around the vacuum manifold...... 186

B.1: PFTBA spectra collected in the explosive trace detection mass spectrometer with the CNT field emission source (left column) and the filament thermionic emission source (right column). Three modes are illustrated, including an initial electron ionization (EI) mode to test for proper source operation, followed by both negative and positive chemical ionization (NCI and PCI) modes. (a) EI spectra with CNT source (b) NCI spectra with CNT source (c) PCI spectra with CNT source (d) EI spectra with filament source (e) NCI spectra with filament source (f) PCI spectra with filament source. Analysis shows spectra are nearly identical for both EI and NCI modes, whereas PCI mode is visually different, with several indications that chemical ionization is occurring. For example, signal peaks at m/z 502 exceed that of both m/z 131 and m/z 69...... 196

B.2: A calibration curve of desorbed TNT is shown and is linear, as expected. Trace explosive amounts were consistent with trace standards in the security field...... 196

xxi

List of Symbols and Abbreviations

CNT: Carbon Nanotube

SWNT: Single-walled Carbon nanotube

MWNT: Multi-walled Carbon Nanotube

VMD: Vacuum Microelectronic Device

MEMS: Micro-Electro-Mechanical Systems

FE: Field Emission

F-N: Fowler-Nordheim

MS: Mass spectrometry

EI / CI: Electron Ionization / Chemical Ionization

CIMS: Chemical Ionization Mass Spectrometry

UHV: Ultra High Vacuum

I-V: Current-Voltage

MUMPs®: Multi-User-MEMS-Processes

Vto / Vth: Turn-On Voltage / Threshold Voltage

LTCC: Low Temperature Cofired Ceramic

CAD: Computer-Aided Design

xxii

Acknowledgements

First and foremost I would to wholeheartedly thank my advisor, Dr. Jeffrey Glass for believing in me and giving me the opportunity to work with him and the amazing research group he has developed. The laboratory environment that he has fostered with brilliant peers, mentors, and state-of-the-art facilities is clearly the reason for my personal accomplishments. His proficiency to always be available for motivation, encouragement, and support, as well as his ability to impart unvarying positive work energy and excitement, has instilled in me values that I will carry throughout my entire career and life. I am a better scientist and professional having worked for him. He has truly been a great mentor and friend, and I will certainly miss my time in the group.

I would like to thank Dr. Charles Parker for being so supportive throughout my time at Duke. He consistently went above and beyond the role of lab manager and his mentorship, both professionally and personally, is equally as important to my success at

Duke University. I appreciate every opportunity and all the advice he provided me when constructing manuscripts and proposals, and mentoring students. Most importantly, I will miss all of our conversations. Whether it consisted of professional recommendations, personal and family advice, or even the friendly banter of Android versus Apple. Because of him, I leave the NTFL with three things instilled in my mind:

1) never mix fonts on a presentation, 2) I will never have a better cup of coffee than one brewed by Charles Parker, 3) there is an RSS feed for everything, yes, even Florida.

xxiii

I would like to thank Dr. Brian Stoner, Dr. Jeffrey Piascik, and Dr. Kristin

Gilchrist of RTI International, Engineering and Applied Physics Division. They provided me with an amazing opportunity to join RTI as an intern upon first arriving here at

Duke University. The professional development and growth that I gained from my experience working under their advisement and mentorship is priceless. Not a day passed that I did not feel blessed to have the opportunity as a graduate student to simultaneously gain industry experience and work with such a talented group of scientists. I thank them for every meeting, every minute of their time assisting with publications, and the opportunity to learn from them all.

I would like to thank Dr. Jason Amsden and Dr. Matt Kirley for always being available for inspirational conversation and ideas throughout the time we spent together in the NTFL. Their assistance with project and personnel management, as well as project organization and success, is more than appreciated.

I am certainly very grateful to all of the members of my Ph.D. committee: Dr.

Brian Stoner, Prof. Adrienne D. Stiff-Roberts, Prof. Aaron Franklin, and Prof. Martin

Brooke. They were all individually important to my success here at Duke. I am truly thankful for all of their teachings, professional and research input, and your passion to help me grow as a scientist and make the most out of my time as a graduate student.

I would like to thank my entire research group and all of the individuals, both past and present, within and outside our lab that I have had the pleasure of working

xxiv

with. I leave Duke having grown not only academically and professionally, but also with many friends that I will certainly never forget. Whether inside the lab, on the soccer field, or “off-site,” I value all of the time we had together. I truly believe I am a better person and leave having learned many things from my colleagues and friends.

I would like to thank Steve Hall at RTI International. It was an absolute pleasure to work alongside Steve on so many amazing projects. Not a day went by that Steve was not willing to assist me with anything I needed for a project, publication, or conference.

We shared many great conversations, both professional and personal, and I value all of the advice and input I received from him over our years working together.

I would also like to thank Dr. Stephen Saddow, Dr. Christopher Locke, and Dr.

Christopher Frewin of the University of South Florida, my alma mater. Dr. Saddow gave a young sophomore engineering undergraduate the opportunity to volunteer in his research lab, and it forever changed my outlook on research and my career. I can’t thank him enough for allowing me to join his research team and gain valuable experience when I needed it the most. That experience, without a doubt, gave me the chance to come to Duke. Dr. Locke and Dr. Frewin were two of the most amazing mentors that an aspiring researcher could ever ask for. They spent many hours working by my side, training me, letting me work with them during long hours in the clean room or growing

SiC thin films, and of course, accompanying them to Mr. Dunderbak’s to “discuss science.”

xxv

Last, but certainly not least, I would like to say thank you, from the bottom of my heart, to my family for always being there for me. Words cannot begin to describe how important they are to my success as a scientist, and as the person I am today. The support I received throughout my entire academic career is more than any one person deserves. My dearest wife, Rebecca, who has stood by my side from the very beginning and supported me through all of the wonderful, and challenging, times. I love her very much for all that she has done to make this possible. And words cannot explain the love

I have for the son that she has blessed me, it is my goal to provide him with the most amazing life possible. Rebecca has been my tower of strength through it all, and I know that our growing family will only benefit from the work that “both of us” put into making this a reality. To my amazing mother and father, I will never forget the sacrifices that they made to get me to this point in my life. From every science project they stayed up all night perfecting with me, to each second they spent pushing me to be the best student, and person, that I could be. I could spend a lifetime remembering all of the important events, which led up to this point, that were a direct result of their constant support, motivation, and desire to create the best opportunities they could for Matthew and I. This achievement is shared by both of them. And to my amazing brother, who has been there for me more than he will ever know. He has been my best friend for my entire life, and he has inspired me in so many ways. From every game of golf, to every hug of support, I cannot thank him enough for everything he has done.

xxvi

1. General Introduction

Nanotechnology is transforming the way we live. Not only are electronic devices and computers growing in number and capability but other fields of science are benefiting from our ability to understand and work at the atomic level. Where established technologies see boundaries, nanostructured materials and the advanced devices they produce, bridge the gap between molecular science and applied physics.

With new and present day micro and nanostructures continuing to demonstrate their impressive properties, it is important we remember to combine continuous progression in discovery with focused research on existing concepts; thereby leading to improved understandings and applications.

While much research is dedicated to the continued development of nanostructured materials, understanding and developing methods towards integration with existing technology is also of great importance. Whether it be for microelectronic development or another unique employment, if nanostructured materials are to be realized in practical applications, the challenges associated with integrating materials, and the work required to integrate them, cannot be disregarded. Nevertheless, the field of microelectronics has already shown, and is primed to continue benefitting significantly, from nanostructured material contribution.

One category in particular, cathode development, has seen an influx of exciting and novel advancements through the incorporation of nanomaterials. Not only have

existing technologies been enhanced through their incorporation, but innovative and novel nanostructured materials have been used to realize new possibilities in cathode development. As a result, new application space has been developed and long-standing technologies are remerging with new capabilities.

This chapter begins with a brief introduction into the electronic materials currently investigated for cathode development, specifically in field emission microelectronic device development, a subject that will be discussed in more detail in section 2.3. Next, the emergence of the field of vacuum microelectronics, whose roots are ultimately derived from the traditional vacuum tube, and a brief overview of vacuum microelectronics and governing principles will be discussed. Additionally, this chapter introduces the motivation and objectives of the research herein and concludes with a brief outline and arrangement of the work.

1.1 Electronic Materials for Cold Cathode Field Emission

Undergraduate electronic materials courses teach us that there are three basic classes of materials: conductors, insulators, and semiconductors. Conductors permit the flow of electrons, insulators block the flow of electrons, and semiconductors fall in between as a “part-time” conductor whose conductivity can be controlled. Electronic materials have been investigated as source materials in vacuum electronics since their development. The traditional vacuum tube, developed from conductive metal electrodes, operates on the premise of thermionic emission. In this process, electrons

overcome the surface potential barrier (work function) of a material and are emitted into vacuum because their internal thermal energy is raised via the application of very high temperatures. Cold cathode field emitters do not provide emitted electrons through the application of high temperatures; instead they use electric fields to allow the electrons to overcome the surface potential barrier. Early cold cathode field emitters were developed from basic conductive metals, similar to vacuum tubes. They have since evolved to utilize a wide variety of electronic materials. Characteristically field emitters are now developed from a micro/nanostructured material with excellent electronic and additional properties. The process of generating electrons using electric fields is referred to as field emission; and while at its core the process is fundamentally dissimilar to thermionic emission, both processes accomplish the same goal of extracting electrons to move through the device. The fundamentals and specifics of electron field emission as proposed by Fowler and Nordheim will be discussed in detail in Chapter 2. Briefly, the intense electric field is applied across a vacuum on a solid to extract the electrons [1, 2].

The electrons within the solid undergo the quantum mechanical process of tunneling through the surface barrier of the material/vacuum interface and are emitted into the vacuum channel. When compared to thermionic emission, field emission is able to generate higher current densities which leads to its higher power applications. Electrons are extracted from a material when the electric field is approximately one gigavolt per meter (or 0.5 volts per angstrom) at the surface of a metal [1, 2]; there is also a strong

dependence on the geometrical properties and work function of the material as well. It is this large field requirement that renders field emission from a flat surface impossible for practical applications. However, the smaller the emitting material is fabricated (a sharp needle like structure for example), the more efficient it becomes in terms of the required applied voltage. This is because the field is concentrated, or enhanced, at the sharp tip which is otherwise known as the emission site. As a result, smaller tips require substantially less applied voltages [1, 2]. The specifics of enhancements experienced by nanostructures will also be discussed in detail in Chapter 2.

It is motivating to see that cold cathode field emitters do currently exist in some commercial products, primarily in the display market with field emission displays

(FEDs) [3-5]. However, like most technologies, a number of challenges still remain to further commercial applications. Present day research tends to focus on new field emission materials or cathode structures. Field emitting materials require specific properties, and often times the more easily fabricated emitters suffer from a collection of unsatisfactory properties. For example, there are often instability concerns with the emitted current in most materials due to the comparatively larger field emitted currents.

Additionally, lifetime and failure concerns vary with material. While several materials have been explored to date, there are still concerns with many of them and a perfect material has not yet emerged. Simply put, present day emitter materials require high operating voltages and their emission current is far from perfectly stable and reliable.

These factors are a large barrier for practical applications and thus much research into controlled fabrication techniques and development of low work function materials have been conducted. It is these reasons that have led to the very limited commercial applications of field emission based instruments [6].

Early field emission research began with metal structures. However, large voltage requirements caused by the inability to develop ideal geometries, high material work functions, and regular impurity adsorption on metal surfaces leading to large current instabilities rendered their development limited. From there we saw the development of the Spindt cathode [7], which employed the most common material available in the semiconductor fabrication industry as a platform, silicon. Well understood IC fabrication techniques were used to conduct microfabrication and develop sharp tip field emitters made out of molybdenum. Over the years, as microfabrication processes improved and developed into nanofabrication possibilities, the structures became smaller with greater aspect ratios. Consequently, field emission properties of the structures improved. Fast forwarding in time, numerous materials have been developed and investigated for their field emission properties, ranging from silicon fibers and nanowires [8, 9], gallium arsenide (GaAs) [10], gallium nitride (GaN)

[10, 11], boron nitride (BN) [10, 12], silicon carbide (SiC) [13, 14], and diamond [14-17].

While they all have their benefits, these materials as electron field emitters have had difficulty reporting performance metrics suitable for promising practical application.

The ideal field emitter needs to offer several key properties: high durability and lifetime

(i.e., resistance to sputtering by residual gases in the emitting environment), low turn-on voltages (i.e., the threshold voltage required to initiate electron field emission), the capability of producing large current densities if the application requires it, and an overall stable and low noise operating ability.

Recently, field emission from another class of nanostructured carbon materials including carbon nanofibers, both single-wall and multi-wall carbon nanotubes

(SWCNT and MWCNTs respectively), as well as graphitic like carbon [18-22] have been investigated in great detail. In short, the considerable extent of research on these structures have shown that carbon nanostructures encompass many of those attractive qualities discussed in the previous paragraph. On the subject of durability, the sputter yield for carbon is the lowest among all solids making it much less susceptible to ambient gas bombardment damage. Because of the very high aspect ratios found in carbon nanostructures, the field enhancement at their tips is very strong resulting in very large local electric fields and ultimately low turn-on voltages. And in regards to current density, the top current-density values to date for carbon based field emitters, held by a CNT emitter array as reported by Fursey [2], are approaching the free-electron limit (i.e., when the work function barrier is completely compensated by the external field) at a value of nearly 1011 A/cm2 (at an area of 1013 cm2). This assemblage of reasons

and achievements have revealed the potential of carbon nanotube field emitters, thus recognizing them as one of the most promising field emitters currently available.

1.2 Vacuum Microelectronics Overview

Emerging vacuum microelectronic devices continue to be a prime candidate for applications where advanced micro and nanostructures are integrated with novel platform materials, such as field emission cathodes, and typically benefit from novel packaging solutions. Their interest, widely due to their numerous unique properties, is also driven by the wide variety of practical applications. The original thermionic vacuum tube, invented in 1904 by John Ambrose Fleming [23] and respectively termed the Fleming Diode, or Fleming Valve, led to the establishment of several advanced technologies throughout the early 1900’s. Three years after Fleming introduced his tube, primarily used for rectifiers in tube amps, Lee DeForest invented the triode vacuum tube

[24] and introduced the concept of amplification. It took some time for the amplification properties of the triode device to become well understood. Shortly following, electronic device development in radio and telephone, television, and even more complex applications including microwave radar, analog, and digital computer were transformed considerably [25]. The vacuum tube offered these technologies a path forward to practical application along with the ability to scale and spread globally.

These early electronic vacuum devices weren’t without their shortcomings, however. The mere fact that a strict vacuum and a cathode heated to 1000 °C were basic

requirements simply to generate electrons meant that power consumption and overall efficiency weighed heavy on the technology. Additionally, their physically large size and high operating voltage requirements added to the longing for a better solution.

Relatively speaking, it wasn’t long thereafter that many believed the inadequacies of the vacuum tube were slowly being overcome with the invention of the solid-state transistor in 1947 [26], and soon after the integrated circuit (IC) [27, 28]. The once frail, sizable, and not very efficient vacuum tube now stood in the shadow of the small, high-speed, and mostly reliable solid-state counterpart.

It goes without saying that the solid-state electronics era has certainly brought forth many advancements in a wide variety of applications and fields. However, the vacuum tube lives on. In fact its “rebirth”, if you will, can indeed be attributed to the advancements in processing techniques that were developed for solid-state electronic production. There are, however, certain properties, or new applications, where solid- state unfortunately falls short. For example, the need for high frequency and/or high power operation; or applications that require transistor-like performance while maintaining a small form factor, but require environmental immunity (e.g., extreme temperature or radiation hardness). An example would be certain aeronautical and industrial applications where the ideal location for sensor placement would be in, or on, a high temperature structure [29, 30]. A common problem with using a solid-state device is the undesirable generation of thermally induced carriers in the channel; an issue that

is not of concern in a vacuum channel. Modern day practice sacrifices accuracy for reliability by locating sensors at ideal distances and making connections with wiring

[31].

When radiation heavy regions are of interest for sensor and circuit placement, a great deal of expense is invested in radiation hardening the devices. This process not only ads significant costs, but also sacrifices efficiency as rudimentary circuits must be utilized to gain maximum radiation protection [29, 32]. Even still, radiation hardening simply delays the inevitable demise of the device to a variety of failure modes related to ion irradiation and subsequent damage to the channel in the SSDs; this again, is not a concern when vacuum channels are utilized. Additionally, the operating speeds of solid- state devices is directly proportional (i.e., constrained) by the carrier saturation velocity in the material. The lifetime of SSDs is also problematic.

Utilizing a vacuum channel device as an active circuit element brings about advantages in speed and lifetime due to the ballistic transport in the vacuum channel [2,

33]. The elimination of signal scattering in the channel material means the devices can be operated at elevated frequencies with much higher efficiency. Therefore it is no surprise that the renewed interests in the utilization of vacuum devices to meet these growing needs has grown steadily in the last 50-60 years, spawning a new field of study centered on vacuum microelectronics. Vacuum microelectronic devices (VMDs) are typically fabricated using advanced semiconductor microfabrication processes and can

operate at room temperatures. VMDs demonstrate amplification, switching, and control because ultimately they are analogous to the conventional vacuum tube. Because they too incorporate the vacuum channel as the medium for carrier transport, similar to the vacuum tube, they retain the temperature and radiation tolerance, as well as the longer lifetimes and slightly higher reliability [2]. Additionally, higher operating speeds are attained not only because of the vacuum channel, but also just by the nature that the electrons are having to transmit over much shorter distances [33]. Developing a

“junction-less” small form factor vacuum device, immune to temperature and radiation concerns, and with the added benefit of longer lifetimes and higher operating speeds, has been the goal for researchers in the field.

Additionally, VMDs have shown great potential as the electron source within an ion source; in mass spectrometry applications for example. Traditional mass spectrometer ion sources typically utilize filament thermionic emitters which required high temperatures to provide enough electron current to generate significant ion current. Cold cathode field emitters offer many advantages over their thermionic emitter counterparts. Specifically, cold cathodes can be cycled on and off rapidly without the degradation disadvantage seen with thermionic emitters. This is advantageous for power concerns in an instrument. In the absence of thermal distortions the hot thermionic emitting cathodes, the extraction grids can be placed close to the cathode which gives further operating speed and energy benefits. Field emission sources can also

be operated in pulsed mode. Thermionic emitters do not traditionally scale down in size well, a fundamental limit if miniaturization of the ion source is desired. Finally, field emission sources provide benefits in the emitted energy spread of the electrons.

Typically field emission energy spread is an order of magnitude less than thermionic emitters. This is why field emission sources are often used in scanning electron microscopes (SEM) to achieve improved resolution.

1.3 Motivation and Objectives of this Research

The united ambition to arrive at a scalable, batch-fabricated, micron-sized miniature vacuum tube-like device is undoubtedly shared among all researchers in the field. Despite all of the research in today’s innovative vacuum microelectronic platforms and field emission based cathode structures, there exists a need to better understand the challenges experienced with integrating these two key components. It is clear that the application space for VMDs, and specifically FE-VMDs, is plentiful. Continued development in either of the two key areas, emitter improvement and integrable packaging, could see progress in many application categories: electron and ion sources, power amplifier and IC’s, as well as novel instrumentation that has yet to be investigated. Application driven research towards utilizing the community established

CNT field emitter arrays alongside existing technologies that allow for packaging and implementation is needed. As is the case in most situations, addressing the fundamental lack of information on dependent aspects is also important when moving towards

complete device realization; it is these knowledge gaps that illustrate the demand for more studies. The primary goal of this work is to address the important challenges related to emitter lifetime, emitter and substrate/package integration difficulties, the ability to incorporate VMDs into instrumentation, and undesired effects that limit performance in specific applications. Addressing these key topics is important if the technology is to progress towards achieving a widely adopted commercial application goal. It is this primary interest that sets the precedence for the research presented here.

1.4 Dissertation Outline

The work herein is organized as follows:

Chapter 2: The fundamental background on carbon nanotubes and carbon nanotube emitters, the basic and advanced principles of electron field emission, necessary background on field emission vacuum microelectronic devices (FE-VMDs), and the required background details of mass spectrometry will be discussed.

Chapter 3: Even though carbon nanotubes, and the FE arrays they produce through integration with complementary components, have established themselves as a primary candidate for next-generation cathode structures, many challenges with the

CNT emitters themselves and their integration into a packaged and scalable device remain. Chapter 3 addresses the lifetime concerns of the CNT field emitter array.

Growth, fabrication, and testing processes towards improving overall emitter and device lifetime will be discussed. Additionally, the ability to integrate the improved CNT

emitters into a scalable platform was investigated. Finally, the commercial viability of the fabricated CNT-FE devices was investigated through their utilization as an electron/ion source in a commercially available portable mass spectrometer where they were compared against a standard filament thermionic emission source in an instrument. The devices were fully characterized and mass spectra was collected and analyzed to investigate the benefits of the source.

Chapter 4: Microelectromechanical systems (MEMS) offer unique packaging and platform options for emerging CNT field emission VMDs. Polysilicon is the primary material used to fabricate MEMS structures, and while many groups have investigated

CNT growth on single crystal silicon, there has been little to no analysis on CNT emitter growth on polysilicon substrates. In order to better understand and control emitter fabrication, it is important to understand the growth and interaction properties between emitter and substrate. Chapter 4 discusses these integration challenges and provides insight into improving control of CNT emitters grown on polysilicon. Details on the substrate preparation, CNT deposition, characterization, and CNT-to-substrate adhesion are analyzed. The growth kinetics, activation energies, diffusion mechanisms, and causes for reduced diffusivity for CNT emitters grown on polysilicon substrates are also investigated.

Chapter 5: The integrated circuit abilities of CNT FE-VMDs through the development of MEMS microfabricated active circuit element structures are discussed.

The design is derived from work previously investigated by our group where devices were used in other operating modes. The work reported here designs, characterizes, fabricates, and evaluates improved adoptions, and assesses two key integrated circuit performance restrictors. The first entails an investigation into crosstalk that takes place between devices placed in close proximity, a must for VMD integrated circuit considerations. Crosstalk was investigated and mitigated using both simulations and experimental validation. The second was overall transmission of the emitted electrons from the cathode (i.e., CNTs) to the anode. Grid loss reduction, charged particle trajectory focusing, and resulting performance metrics are discussed.

Chapters 6: The fabrication of a novel drop-in packaged MEMS micro-ion source for miniature mass spectrometer applications is investigated. The integration of the three key ion source components is discussed: MEMS platform, CNT field emitters, and a low temperature co-fired ceramic (LTCC) 3D scaffold with integrated ion optics. This chapter focuses around geometrical designs, failure mitigation, and operating modes for the CNT based MEMS VMDs. Additionally, a complete packaging option with integrated extraction and focusing apertures was designed and tested experimentally to validate theoretical simulation work. The fully packaged MEMS micro-ion source was tested and characterized.

Chapter 7: A summary of contributions and conclusion with brief discussions of future work to continue the research outlined in this dissertation will be discussed.

2. Background - Carbon Nanotubes (CNT), Field Emission (FE), FE Vacuum Microelectronic Devices (FE- VMDs), and Mass Spectrometry (MS)

2.1 Introduction

This chapter provides background and supporting information for the four key constituents of the research presented in this work. Section 2.2 gives insight into the importance of carbon nanotube (CNT) research and introduces essential structure, growth mechanisms, properties, and applications. Section 2.3 introduces the theory of field emission and relevant concepts including: work function and Fermi level significance, emission processes, the Fowler-Nordheim (F-N) model, and emission from various electronic materials (including CNTs). Section 2.4 focuses on history and operating dynamics of vacuum microelectronic devices, explicitly the field emission vacuum microelectronic device (VMD). Additionally, essential literature outlining the device platform that the work in chapters 5 and 6 are built upon is summarized, including work done in collaboration with RTI International. Finally, section 2.5 briefly introduces the basic concepts of mass spectrometry with emphasis on the ionization source component of the instrument. This is relevant to chapters 3 and 6 where the fabricated device targets mass spectrometry applications.

2.2 Carbon Nanotubes

With thousands of publications released annually on carbon nanotubes, the continuous, and increasing interest in the subject by researchers worldwide is clear.

Recent years have seen a sharp rise in work discussing fabrication methods, efforts to fully characterize the chemical and physical properties of the material, and the search to find the most attractive applications [34-38]. CNTs exemplify an appealing area of study for many micro and nanotechnology applications as they embody many desired properties of a material: stability (thermal, chemical, and mechanical), ideal electrical properties, and their extremely small size.

2.2.1 Significance of CNT Research

With only around 25 years since their discovery in 1991 and 1992 by S. Iijima [39] and Ebbesen [40], and even fewer years for the discovery of some of their properties and applications, CNTs are just beginning to emerge in the application space. The realization of the importance of CNT research began to emerge when researchers demonstrated that

CNTs exhibited many of the same desired properties that existed in graphite, specifically intra-plan graphite [39, 41, 42]. The added benefit of CNTs oriented in a near ideal one- dimensional structure, as well as their ability to be either semiconducting or metallic, offers further considerations for research and development. To date, CNTs have been integrated into a variety of industries. Report statistics [35] pertaining to forecasts of the technology application predict a market increase to over $2.8 billion by 2023. Areas of current and predicted application include: photovoltaics, sensors, semiconductor devices, displays, conductors, smart textiles and energy conversion devices (e.g., fuel cells, harvesters and batteries) [35]. And while the technology application opportunities

and their market drivers are plentiful, the challenges and pathway to commercialization are also important considerations.

2.2.2 Structure

The structure of a CNT is well understood by considering the structure and bonding of graphene as the CNT can be considered to be a sheet of graphene seamlessly rolled up on itself. With six electrons (2 core 1s orbital and 4 valance electrons residing in the 2s and 2p orbitals), carbon is the basis for numerous organic compounds. The hexagonal structure of graphene is a consequence of three sp2 hybridized bonds (s bonds) giving it the traditional honeycomb shape. A hybridized sp2 bond occurs when the 2s electrons(s) are raised to the 2p energy level thereby hybridizing with the existent

2p state electrons. This occurs in other carbon based structures as well. Diamond for example, is made up of four tetrahedral bonds which are derived from four equal sp3 hybrid orbitals. With graphene, we also see a weak out of plane p bond derived from the fourth valence electron that is utilized in the 0.34 nm interlayer bond in graphite. The

C-C triple bond at 0.14 nm length is considered to be the strongest and shortest single bond in nature [20]. Similar to graphene, CNTs are also comprised of sp2 hybridized bonds and the out of plane orbital p bond. However, unique to CNTs, the curvature of the structure gives rise to a combination of sp2 and sp3 bonds in order to reduce the overall stress along the curves. It is this unique bonding, along with their high aspect

ratio, that give CNTs some of their exceptional properties. Figure 2.1 illustrates the covalent bonds within a carbon nanotube structure. The s and p bonds are shown.

Figure 2.1: Bonding structure of a carbon nanotube. The sp2 hybridized bond (s bond) and the out of plane orbital p bond are shown. Diagram derived from [43].

CNTs are typically categorized by the number of walls that exist in the structure.

Classifications include: single wall carbon nanotubes (SWCNT) with a diameter on the order of 0.4 to 3 nm [44], multi wall carbon nanotubes (MWCNT) which include several concentric SWCNTs with an average inter-wall spacing of 0.34 nm (identical to that of graphene) [45, 46] and a range of diameters, and sometimes few walled carbon nanotube

(FWCNT) which is a term for structures with only two to a few walls [47]. In the case of

MWCNTs, the inter-wall spacing can be an inverse function of diameter, increasing as diameter decreases as a result of wall curvature [48]. The electronic properties of the

nanotube can vary depending on the number of walls thereby shifting the material into a metallic (finite value of carriers in the density of states at the Fermi energy) or semiconducting state (no charge carriers in the density of states at the Fermi energy)

[49]. A defect free 1D CNT could theoretically benefit from ballistic transport (i.e., conductive at room temperature with electrons moving freely across the structure without scattering effects). However, during synthesis it is common for defects in the walls to emerge. These defects can alter the properties of the CNT, which can be detrimental or beneficial depending on the application and has been researched in detail

[49-51].

2.2.3 Growth Mechanisms of CNTs

Carbon nanotubes are catalytically grown structures and have been studied in great detail [52]. To summarize the important growth mechanisms utilized in their synthesis, three primary methods are available. The arc-discharge method, which uses electric arcs at very high temperatures (approx.. 4,000 °C) to evaporate graphite electrodes was the method used by Iijima in their initial discovery [39]. The process produces well crystalized but highly impure (large content of amorphous carbon and metallic constituents) CNTs. Laser-vaporization, another method developed subsequent to arc-discharge [53], utilizes high temperatures and high powered lasers to evaporate high-purity graphite. Laser-vaporization is also a high purity process, however the overall yield of production is very low. Cassell, et al [54], first reported chemical vapor

deposition of CNTs, currently the most popular method of growth due to its low-cost and scalability. The catalyst-assisted process utilizes thermal energy to achieve decomposition of the hydrocarbon precursors. CNTs grow on the catalyst which is typically loaded onto a substrate (Si for example). Numerous variations of the process have been explored that alter the growth, such as; type of hydrocarbon precursor species, growth pressure, temperature used to dissociate the hydrocarbons, flow-rate of the hydrocarbons, and time. Variations of CVD also exist. Plasma enhanced CVD

(PECVD), the synthesis method used in this work, employs a plasma to assist in dissociation of the hydrocarbons. CNTs can be synthesized with vertical alignment as they are aligned with the electric field lines present in the plasma sheath. PECVD is the most common synthesis method for vertically aligned nanotubes (VA-CNTs) and the process has been heavily studied [55]. A microwave PECVD (MPECVD) was explicitly used for the growth of the carbon nanotube field emitters in this work; a schematic diagram of the 915 MHz MPECVD system used in this study is given in Figure 2.2b. The aligned MWCNTs that were utilized in this study were grown following a process and model similar to that reported and proposed by B. Stoner and colleagues [56]. The

MWCNTs grown were of periodic bamboo-like structure with repeating hollow conical segments similar to the structure shown in Figure 2.2a. Additional details on the growth process and structure of the MWCNTs has been reported previously [57].

2.2.4 Properties of CNTs

2.2.4.1 Mechanical

The mechanical properties of CNTs are of great interest. Widely studied, both experimentally (when possible due to their small size) and also theoretically, CNTs have reported Young’s modulus, tensile strength values, and fatigue strength several times larger than steel with nearly four times lower mass density. However, these properties are measured per unit area value and thereby remains relative and thereby they cannot always be scaled for practical applications. Nonetheless, the mechanical properties have proved themselves valuable in applications where composites are fabricated for structural reinforcement [59]. Comparison of theoretically calculated stress and strain values of SWCNTs with varying diameters show that large gains are possible with small changes in the CNT diameter. This is seen when comparing micrographs of SWCNTs,

whose structures tend to have high tortuosity (very twisted), to those of MWCNTs who remain relatively straight on a relative nano scale [37]. Tensile strength measurements, again both theoretically and experimentally, have also shown that CNTs exhibit values fifty times those of steel [37, 60-62]. Not only does this place CNTs as the strongest and stiffest material in existence, it indicates why CNTs offer superior performance when used in applications that apply large tensile forces. An example of this is field emission from CNTs, where electric fields administer large forces on the structures. It has been shown previously [37, 63-66], as well as in this work, that the mechanical failure associated with CNTs as emitters is mostly failure at the CNT-substrate adhesion point, and not along the CNT structure itself.

2.2.4.2 Electronic

Full details on the electronic properties have been reported and studied in detail

[34, 67, 68]. Certainly the most unique property of CNTs is their ability to exist as semiconducting or metallic. The distinction is determined solely by their chiral structure.

Their conductivity is understood through modification of graphene’s band structure.

The aforementioned π orbitals overlap along the hexagonal planes resulting in very high mobility. Graphene has a zero band gap, whereas CNTs electronic properties are realized by modifying graphene’s available electronic states in the radial direction [37].

The enormous aspect ratio of the CNT structure results in a multitude of available electronic states in the axial direction, with almost no electronic states along their

circumference. It is true that the curvature of CNT walls, and the resulting sp3 hybridization does have effects on the band structure, however, it has been shown that this only affects CNTs with diameters less than 20 nm (i.e., SWCNTs) [37]. Also, the band gaps of CNTs are highly sensitive to the environment. It has been shown that the transition from n-type, to intrinsic, to p-type due to the absorption of impurities, gases, and trapped charges, can have an impact on the conductivity by orders of magnitude

[46, 69]. Early work demonstrating large current capacity in CNTs without substantial heating by de Heer, et al. showed evidence of the one dimensional quantum conduction properties of CNTs. The charge transport in CNTs was found to be ballistic in nature, as opposed to conventional conductors whose length would directly impact scattering, electron flow resistance, and thus the energy dissipated in the conductor [70, 71].

Furthermore, degradation by electromigration, a common unfavorable affliction to lifetime in conventional metal conductors, is mostly nonexistent in CNTs [37]. It is these reasons that literature has reported current carrying capabilities of CNTs three orders of magnitude higher than copper (109 A/cm2 versus 106 A/cm2). Specifically to MWCNTs, the aforementioned work by de Heer, et al. [70] and others have shown that in the case of MWCNTS, the current is primarily conducted by the outermost walls. Furthermore, traditional metal conductors display continuous degradation when evaluating current carrying lifetime. However, MWCNTs have shown their electrical breakdown is dependent on the sequential failure of the individual walls [72]. The ability to exist as

semiconducting or metallic, adjust band gap energy without the need for doping, and possess superb conductive properties, provides an understanding of why CNTs are under strong consideration for various electronic applications.

2.2.5 CNT Applications

It is a widely accepted theory that fundamental scaling limits of silicon technology and silicon devices are nearly at an end [73]. Reduction in size has been the driving force for increased performance (and reduced power consumption) since the early works of integrated circuit development. Much research has been dedicated to new materials that can supplement, or replace, silicon; CNTs being one of them. CNTs have numerous applications in the electronics industry. The properties discussed in section 2.1.4, including their ballistic transport and current carrying capacity, inspire continued development as these properties correlate directly to those properties desired in electronic devices (i.e. low power dissipation and scalability). CNTs have also been investigated for related applications such as interconnects and thermal management structures [74-77]. Energy applications have also received interest due to the respectable chemical stability in combination with the electrical conductivity. Fabrication of electrodes from CNTs, as well as their integration in batteries, photovoltaics, and fuel cells have also been investigated in detail [78-83]. One popular research direction has been CNT sensor development [84, 85]. Section 2.1.4 eluded to the capabilities to transition a CNT from n-type to p-type via the absorption of impurities and gases. This

absorption has an impact on the conductivity and this shifting ability can be tapped for sensor development by exploiting the CNTs ability to yield a conductance change when absorbing environmental gases. Additionally, the low power consumption characteristic is ideal for a portable instrument, or an instrument that desires low operating energy overhead. Finally, CNTs have shown to be an excellent candidate as an electron field emitter in field emission applications. This is primarily due to their very large aspect ratio and small diameter tips. Additionally, their ability to support large currents, withstand high temperatures, and resist degradation better than traditional field emission materials further supports their consideration to supplement or replace metal field emitters. CNTs as field emitters, and pertinent references, are discussed in detail in section 2.3.3.

2.3 Field Emission Fundamentals

Field emission is a growing area of research, and it is beginning to gain ground in commercial applications establishing itself as a significant field of research [2, 4, 22,

86]. A wide literature search including all related terminology (i.e., field emission, field emitters, etc.) reveals a consistent increase year after year in research publications. Some of the more well established and required instruments in scientific research and commercial fabrication (i.e., semiconductor fab) rely on field emitters as their source of electrons to achieve the desired performance (e.g., scanning electron and transmission electron micrographs). Historically, thermionic emission (or the Richardson-Dushman

effect) and cold electron field emission were independently studied from the late 1800’s to the early 1900’s (approx.. 1880 to 1930). Both emission processes, important in their own respect, pioneered the development of numerous modern day electronics and instrumentation. This section will introduce the key concepts of electron emission, give a brief overview of additional electron emission mechanisms, discuss the governing F-N theory for field emission, and close with a discussion about field emission from CNTs.

2.3.1 Electron Emission

At its core, the term electron emission is associated with a process that allows electrons to leave the surface of a solid or liquid and transfer into another phase

(typically vacuum). For free electrons, this behavior is best described by the electron gas or Drude model [87]. The free electrons inside a condensed phase have a lower energy than the vacuum and therefore the emission of electrons from a surface is only possible if sufficient energy is provided.

2.3.1.1 Probability of Escape

To understand the energy of the electrons more specifically, we use a quantum mechanics enhanced theory of solids which shows us that the multitude of energy bands in a metal will actually overlap to form a single energy band [88]. Within that energy band there exists several discrete energy levels that become occupied by the valence (i.e., free) electrons. If a material were to be held at absolute zero, all energy levels up to the

Fermi level (EF) would be filled. The energy difference from this Fermi level to vacuum

level is referred to as the work function (Φ) of the material and it is equal to the energy required to excite the electron from Fermi to vacuum so it can leave the surface/material.

The work function is the surface potential barrier that holds the electrons within the material. Figure 2.3 gives a schematic energy diagram for metals, the work function is shown.

The aforementioned electron energy state distribution is fully described by the density of states (DOS) equation. Because the energy at the bottom of the band (Ebottom) is equal to zero, the states that exist from the bottom of the band to the middle can be represented by the following DOS equation:

8휋√2 푔(퐸) = 푚∗3/2√퐸 [1] ℎ3 where Planck’s constant (h) equals 6.6.3 x 10-34 J.s and m* is the electron mass. Similarly, if we take equation 1 and replace √퐸 with √(퐸푡표푝 − 퐸) than we have the DOS equation representing states from the top of the band to the center. The number of states that exist

in the energy interval described (E to E+dE) is found by multiplying g(E) by dE. In order to find the probability that an electron will reside at a specific energy level, we use the

Fermi-Dirac distribution function:

1 푓(퐸) = 퐸−퐸 [2] 1+exp⁡( 퐹) 푘푇 where Boltzmann’s constant (k) is 1.38 x 10-23 J.K-1 and EF is the Fermi energy.

Multiplying g(E) by f(E) gives the number of electrons per unit energy. The energy required for electron emission, which we have defined above as the work function, is the specific energy that must be provided to each electron within the material.

2.3.1.2 Emission Processes

The capacity and process by which electron emission takes place relies heavily on the properties of the material from which the electrons are emitting. Geometry, electronic structure, and work function are examples of a few of those properties and must be considered in each case [1]. The following are brief summaries of the four principal methods of obtaining electron emission: thermionic/photo emission, secondary emission, Schottky emission, and field emission. In thermionic emission, sufficient energy is provided through heat (typically greater than 2,500 °C) in order to overcome the surface potential barrier [89]. The quantity of electrons emitted is directly proportional to the temperature applied. In photoemission, the incidence of electromagnetic waves, or more specifically the impinging photons, transfer energy larger than the work function allowing electrons to emit from the surface. The number of

electrons emitted is proportional to the intensity of the light on the surface [89, 90].

Secondary emission is more an after effect that occurs when the surface of the material is struck by a high energy particle (electron or ion) and sufficient energy (i.e., more than the work function) is transferred to the electrons on the material surface. Schottky emission is slightly different in the sense that energy is not transferred to the electrons

[90]. Instead, the effective barrier is lowered by a specific amount (ΔΦ) through a combination of vacuum level bending due to an electric field, and what is known as the image charge effect [90]. With the lowered barrier, electrons with sufficient energy can escape (emit into vacuum). Figure 2.4 illustrates the emission mechanisms with a schematic diagram comparing thermionic/photo emission, Schottky emission, and field emission.

Figure 2.4: Various electron emission processes; adapted and modified from [91].

Field emission, the final electron emission mechanism, will be discussed in more detail as it describes the emission process utilized in this work. Fundamentally, electron

field emission is a quantum-mechanical phenomenon where the electrons at the Fermi level tunnel through the surface potential barrier due to the application of an external electric field to thin the vacuum level barrier. A higher applied field correlates to a thinner barrier. In order to generate electron emission from a flat metal surface, a local field on the order of 3-5 x 107 V/cm is required. Using a simple metal with work function

Φ, we can resolve the minimum barrier width (d) and additional required conditions needed for tunneling (i.e., field emission to occur). If we assume the electron is given kinetic energy equal to that of its work function, we can calculate its uncertainty of momentum and using Heisenberg’s uncertainty principle we find the uncertainty in position to be:

ℏ Δ푥 = [3] 2(2푚Φ)1/2

In the presence of an applied field (F), the following equation

푃퐸(푥) = 퐸퐹 + Φ − 푒퐹 [4] represents the potential energy just outside the metals surface, thereby allowing the potential barrier (d) to be given by the following equation:

Φ 푑 = [5] 푒퐹

This simply shows that by applying high electric fields, the barrier width can be reduced. More specifically, by nature of the wave mechanics of electrons, if the barrier width is smaller (or equal) to the calculated uncertainty position from equation 3, it is possible for an electron to tunnel through the barrier creating field emission. It can also 30

be shown that materials with higher work functions require higher applied electric fields to enable field emission. Recalling that Schottky emission takes place due to the bending of the vacuum level enough so that electrons can “spill” over the top, we can now set an upper limit on the applied field before tunneling current dominates the total contribution to emitted current density. It has been reported that this field strength is approximately 106 V/cm [92]. This indicates that there is a transition region where both

Schottky and field emission will be contributing to the total emission current, but specific details are beyond the scope of this work. Additional details into the fundamentals of field emission are discussed in section 2.3.2, F-N theory, the governing theory that describes the field emission phenomena.

2.3.1.3 Emission Metrics Specific to FE

In the field of electron emission, and more specifically, electron emitter characterization and application, there exists a list of common specifications and performance metrics that emitters are characterized to. While these measurements are certainly case and process dependent, it is appropriate to define the most common as reported in literature for field emission as it is the primary process utilized in this work.

Because field emission is dependent on an applied electric field in order to generate emission, it is common to report two efficiency measurements related to this applied electric field. These two metrics were developed in the electronic display industry where the current densities needed to turn on and saturate a pixel were

referred to as the turn-on and threshold voltage, respectively. Therefore, for field emission, turn-on voltage (VTO) is the voltage needed to commence emission. This is usually recorded when the emitted current is measured at 1 nA, or potentially 1 nA/cm2 if reporting in current density. Depending on instrument resolution, it could be taken when the I-V characteristic plot becomes exponential. The threshold voltage (VTH), is the applied voltage required to reach a specific emission current value or density. This value varies widely in literature, but for the purposes of this work, VTH is the voltage needed to emit 1 µA of current. Reporting VTO and VTH provide a means to compare performance between different emitting structures or devices. Current density in field emission is analogous to current density reporting in similar fields of research. It is defined as the current output from the emitter per unit area of the emitter.

The lifetime and stability of the emitter are the final two metrics of interest.

Lifetime is defined as the length of time the emitter can continue to perform. Lifetime testing is often conducted using constant or pulsed measurements. Stability is a measurement of the fluctuation on the emitted current over time given a fixed voltage.

Degradation leads to changes in stability and ultimately failure.

2.3.2 The Fowler-Nordheim Theory

Schottky in the 1920’s first reported experimentally that electron emission could occur under the presence of large electric fields [93, 94]. However, it was Fowler and

Nordheim in 1928 who utilized wave mechanics and some approximations to develop a

theory for the emitted current leaving a flat metal surface. They proposed that the current is established from the probability of an electron successfully tunneling through the potential barrier. Calculating the barrier transparency using the Wentzel-Kramers-

Brillouin (WKB) approximation [95] and incident electron flow, and then integrating, allows for theoretical current density calculations. The Fowler and Nordheim theory uses the following assumptions:

 the free electron model using Fermi-Dirac statistics is utilized by the

emitter in consideration

 the emitting material is atomically smooth and clean

 any and all potentials within the emitter are constant and are not affected

by the external electric fields; and the electric field is uniform in the

vacuum gap

 the temperature is 0 °K (field emission is weakly temperature dependent)

 the WKB approximation is used to determine the electron tunneling

probability.

If these conditions are met, the current density can be described by the following equation:

∞ 퐽 = 푒 푛(퐸 )퐷(퐸 , 퐹)푑퐸 [6] ∫0 푥 푥 푥

Here, J is the current density, e is the charge on the electron, Ex is the kinetic energy of the electron, F is the electric field that is applied, and n(Ex) is the number of electrons

whose energy falls between Ex and Ex + dEx. Utilizing the aforementioned WKB approximated barrier transparency, the classic F-N formula is derived with constants k1 and k2 (1.54 x 10-6 AV-2 eV and 6.83 x 107 eV-3/2Vcm-1 respectively) [96], J is the emission current density (A/cm2), Φ is the work function of the surface from which the emission is taking place (eV), and E is electric field between the anode and the emitting surface; or

V/d (V/cm):

3 퐸2 Φ2 퐽 = 푘 exp⁡(−푘 ) [7] 1 Φ 2 퐸

From equation 7 we can see the exponential relationship between the emission current and the applied electric field, with the work function also having a strong impact. It should be clear that equation 7 is true in the case of a flat metal cathode, as was derived by F-N. Since current is actually measured during field emission experiments, the equation can be simply converted to current by multiplying through by an approximation of the emission area.

Both theoretical and experimental validations of the F-N equation describing electron field emission from cold cathodes of a flat surface have been reported [2, 86, 97].

However, fields greater than 3 x 108 V/m are required in order to generate field emission from a typical flat metal surface. Therefore, early experiments sought to machine metal emitters into sharp tips in order to take advantage of the concentrated fields that take place on regions of high curvature. In other words, the local electric field at a tip will be significantly larger due to the effect that gradients of electric field lines increase with 34

decreasing radius of curvature. Smaller applied voltages will result in much higher local electric fields in the vicinity of the sharp tips; therefore smaller applied voltages are needed to achieve field emission from the structure. This is known as field amplification and is often represented by the field enhancement factor (β) using the following equation:

훽푉 퐸 = [8] 푑

A consequence to using the sharp tip is that the electric field will be at its highest at the apex of the tip, but it will decrease quickly moving off axis. Therefore, the traditional F-

N equation (Equation 7) cannot be precisely applied to obtain the emission current from a sharp tip as it violates the assumption that there is a uniform electric field in the vacuum gap region. However, modification of the F-N equation by replacing the uniform electric field (E) with βV/d will allow for its use in calculating emission currents from tips.

To further understand and represent experimental data, we can convert current to current density, and use the localized electric field expression to derive an equation that allows us to express the measured current:

3 (훽푉)2 Φ2 I = 푛훼푘 exp⁡(−푘 ) [9] 1 Φ 2 훽퐸

where n represents the number of micro-tip emitters in a given field emitting array, and

α represents the emitting area per tip in cm2. It is common practice to reorganize

Equation 9 to present the equation in what is referred to as F-N coordinates:

3 퐼 푘 푛푎훽2 푘 Φ2 ln ( ) = ln ( 1 ) − 2 [10] 퐸2 Φ 훽퐸

This enables the creation of the F-N plot, which displays a linear relationship between ln(I/V2) on the y-axis and the reciprocal of the extraction voltage 1/E on the x-axis. Figure

2.5 shows measured I-V characteristics and the calculated F-N plot for a carbon nanotube field emitter as reported by [98]. The slope (m) of the linear relationship can be calculated and used along with known information to calculate the field enhancement factor using the following equation:

−푘 Φ3/2 푚 = 2 [11] 훽

The F-N equation and F-N plot is widely used to validate electron field emission from experimental data. Extraction of the field enhancement factor also provides a metric for comparing field emitters to one another. Furthermore, a linear F-N plot verifies field emission is taking place and not another emission process. A deviation from the linear relationship described above can indicate another emission mechanisms is contributing.

Additionally, deviations can give insight into energy band structure and voltage-to- barrier-field conversion factor of the emitting material [99].

2.3.2.1 Correction to the F-N Equation

While 2.3.2 introduced the work of Fowler and Nordheim as a means for comprehending experimental field emission data, it is important to use care when applying the F-N theory. Due to assumptions used in the theory, and intended applications to flat surfaces, gross overestimates of parameters can be easily made. It is possible for the assumptions in the F-N equations to lead to current density approximations that are off by 3-8 orders of magnitude for modern nanostructure emitters [100]. Many have worked to modify the F-N theory for use with nanostructures such as carbon nanotubes [2, 97, 101]. Experts in the field, such as Forbes, have worked to clarify and correct the misuse of the F-N equation. Additionally, they have reported

on improved conditional tests (Forbes orthodoxy test for example [100-103]) to validate application of the theory to extract desired parameters. The modern versions of the F-N equation add correction factors for several components of the equation, and while important to the overall science, the detailed explanations are beyond the scope of this work. It is important to note that modern F-N equations, as well as sophisticated algorithms designed to follow and apply the work of Forbes and others, was utilized in the work presented here.

2.3.3 Field Emission from CNTs

This section will focus on field emission from the type of structure utilized in this work (i.e., metallic MWCNTs), and will not cover FE from semiconducting or SWCNTs which has been reviewed in detailed elsewhere [96, 104]. As previously mentioned briefly in chapter 1, carbon nanotubes encompass many of the attractive qualities of the ideal field emitter including durability, low power operation, and acceptable current densities. It is for these reasons CNTs have been recognized as one of the most promising field emitters to date. Specific details on their field emission capabilities and methods of emitter design from CNTs will be discussed in this section.

2.3.3.1 Unique FE Properties of CNTs

First reported as an electron field emitter in 1995 [18, 19, 105], CNT emission properties have been well studied and have proven themselves to be a viable candidate for electron sources in various applications [106, 107]. CNTs have several advantages as

FE electrodes. The most apparent advantage of CNT field emitters is their very large aspect ratio and small radius of curvature at the tip. The diagram in Figure 2.6a shows how the spatial distribution of the electric field potential is concentrated near a sharp tip; this is especially true in CNTs. The large concentration at the tip equates to very large field enhancement factors (i.e., lower applied voltages required). Figure 2.6b and Figure

2.6c provide SEM images of CNT field emitters and a standard tungsten filament used in a field emission SEM.

Figure 2.7 gives a potential-energy diagram for a CNT emitter (i.e. represented by a micro-tip metal) illustrating the advantages to the micro-tip emitter in terms of tunneling probability under an applied field. The solid green line represents the surface potential barrier seen by the electrons in the CNT/metal. The solid blue line represents

an applied electric field; -eE(x). The electric field applied to a planar emitter will only reduce the barrier slightly. The dashed blue line represents the “reduced” surface potential barrier seen in the planar emitter case; V(x) = -e2/4x – eE(x). In the case of a micro-tip emitter, the applied potential is multiplied by the enhancement factor; solid red line –βeE(x). Therefore, the barrier is reduced dramatically increasing the tunneling probability (i.e., tunneling current or emission current) significantly; represented by

V(x)= -e2 – βeE(x).

Figure 2.7: Potential-energy diagram illustrating the effect an applied electric field has on the energy barrier of a planar emitter versus a micro-tip emitter.

Another advantage is the activation energy for surface migration of atoms on CNTs is significantly larger (17 eV) than tungsten (3.2 eV) and rhenium (2.7 eV) (e.g. standard materials used for commercial field and thermionic emitters) [106]. CNTs also possess a high melting temperature when compared to refractory metals such as tungsten. 40

Additionally, CNTs have good conductance, extremely large Young’s modulus (high stiffness), and high tensile strength (allowing the tip to withstand the extremely strong fields (several V/nm) needed for FE) [106]. It is these characteristics that give CNT field emitters their lower threshold fields when compared to conventional field emitters [109].

It has been shown that CNT field emitters can also endure large current densities [21,

110]; and Fursey et al. [2] has reported emission current densities approaching the free- electron limit at a value of nearly 1011 A/cm2 at an area of 1013 cm2. While lifetimes of various times from minutes to hours have been reported, a CNT emitter lifetime is dependent on its application and several factors: temperature, ambient pressure, surrounding gas species, CNT structure, and substrate [36, 38, 111]. Finally, it has been shown that CNT field emitters can be described by the well-known F-N relationship, given the inclusion of a few modifications, with acceptable accuracy [112, 113]. The afore mentioned factors, as well as several others including CNT growth conditions, alignment, inter-tube spacing, and CNT doping, have all been shown to affect the field emission characteristics of CNTs. However, their study exceeds the scope of this work and are only mentioned to offer a complete understanding. It is easily shown, however, that the unique advantages provided by CNTs directly impacts their performance as a field emitter. Table 1 provides a comparison of the threshold electric field required to achieve an emission current of 10 mA/cm2 as reported by [98]. It is clear that CNT

emitters will provide comparable emission current at significantly less applied voltages when compared to other traditional field emitter materials.

Table 1: Comparing threshold electric field requirements for different materials; reproduced from [98].

Threshold Field (V/µm) for a Cathode Material current density of 10 mA/cm2 Mo tips 50 – 100 Si tips 50 -100 p-type diamond 160 Defective CVD diamond 30-120 Amorphic diamond 20-40 Cesium-coated diamond 20-30 Graphite powders 10-20 Nano-diamond 3-5 (unstable > 30 mA/cm2) Assorted carbon nanotubes 1-2 (stable > 4,000 mA/cm2)

2.4 Field Emission Devices

A large challenge in the adoption of field emitter electron sources is the difficulty that arises with integration and packaging. The growth of nanometer scale field emitter arrays is typically dependent on specific substrates (e.g., silicon). Therefore, applications are presented with engineering challenges to integrate the emitters to benefit from their advantages in a way that will not impact the application parameters (i.e., reliability, cost, etc.) negatively. Designing VMDs requires knowledge of the emitters themselves, the platform on which they are grown on, as well as any integration challenges or system level performance concerns that could come about from the integration of the field emitter. To date, FE-VMDs have been utilized in a handful of applications with the most

prominent being flat low-voltage displays; several others are listed in Table 2. This section will briefly introduce the more prevalent and modern VMDs under development, the two primary operating modes (i.e., diode and triode), a brief discussion of present VMD challenges and the motivation towards laterally oriented devices, and finally specific background details related to previous work on the

Duke/RTI developed monolithic VMD platform.

Table 2: Primary FE Device Applications

FE Device Application 1.) Flat low-voltage displays of high brightness and high resolution 2.) Electron optics systems of super high resolution (e.g., SEM, TEM, electron lithography, Auger electron spectrometer, tunnel microscopy) 3.) Active elements for integrated circuits (e.g., diodes, transistors, etc.) 4.) Employment in microwave devices 5.) Variety of pressure gauges, magnetic field gauges, and other sensors 6.) New types of ion sources 7.) Realization of new very efficient production technologies for microelectronics

2.4.1 FE Device Classifications

2.4.1.1 Spindt Cathode

It is well agreed upon in literature that Spindt and coworkers pioneered the field emitter array in the late 1960’s using field emission cathodes now referred to as the

Spindt cathode [7]. The device leveraged present day microfabrication technology to fabricate sharp molybdenum cone shaped field emitters inside a small etched cavity through a Si wafer with an oxide layer on top. A counter electrode was deposited on top surrounding the etched cavity to provide an extraction gate to generate field emission 43

from the Mo cone. Because of their small form factor, many Spindt tips, as they were called, could be populated on the same substrate in close proximity to fabricate field emitting arrays; an SEM of one Spindt-type emitter is given in Figure 2.8a. Spindt cathodes do have drawbacks in the sense that the field emitters themselves, the metal

Mo cones, are subject to degradation during normal operation and high currents [114].

Sputtering of the tips is also possible which further degrades the tips [115]. Small undesired deviations from the originally fabricated sharp tips lead to fast field enhancement factor declines and ultimately lower emission and degraded efficiency.

The emitters are also subject to chemical reactions with ambient environment which changes the electronic structure of the Mo [115, 116].

Figure 2.8: (a) Spindt-type field emitter. SEM of single emitter shown; reproduced from [2]; (b) Deep reactive ion etched (DRIE) silicon field emitter arrays; reproduced from [117].

2.4.1.2 Silicon Field Emitters

Silicon based VMDs, such as Si nanowires or pillars grown on substrates for example, are another heavily studied type of FE-VMD. The devices are typically

developed by integrating the wafer based emitters with an arrangement of on or off chip grids. Both bottom up (i.e., typically growth) and top down (i.e., etched using methods like lithography and reactive ion etching) are used for fabricating the emitters.

Fabrication of the devices can involve many processing steps, but they can also achieve sufficient current densities with low turn-on voltages. Failure of Si emitter based devices has also been studied in detail [118-121]. Similarly to Spindt cathodes, thermal failure and ion bombardment are two of the more common failure mechanisms due to the nature of the emitting material (i.e., Si). It has also been shown that Si emitters are prone to damage by a combination of joule heating and what is referred to as the Nottingham effect, which is quantum-mechanical energy exchange process that contributes heavily to thermal imbalances during field emission [2]. Failure of the emitter results in ultimate failure of the VMD. High-current and uniform emitting capable silicon structures have been reported [117]; Figure 2.8b shows an example of an array of Si emitters from the

Akinwande group.

2.4.1.3 Carbon Field Emitters

The last class of modern VMDs are based on carbon based emitters, either diamond or CNTs. Carbon based FE cathodes and VMDs have been shown to be more robust than both Spindt type cathodes and Si emitter based cathodes. This is mostly attributed to the robustness of the emitters themselves as was discussed in section

2.3.3.1. Fabrication of VMDs based on CNT and diamond emitters has been

demonstrated in a variety of geometries; both vertical and lateral. VMDs that employ nanodiamond emitters and are patterned in finger-like emitter structures have shown strong performance in the realm of integrated circuit application via both suspended and self-aligned anode designs [15, 122-126]. There are many similarities between the two materials as emitters. The variation between the two is mostly realized through the methods and geometries fashioned to create the VMDs. One advantage that both share is the significantly reduced effects of sputtering and ion bombardment that is experienced by the aforementioned VMDs based on Spindt and Si emitters. Present day challenges with VMDs based on diamond and CNT are mostly related to lifetime and integration [66, 127-129], and these two will be discussed in detail in section 2.4.3. Figure

2.9a gives an example of a microfabricated nanodiamond lateral FE diode as reported by

[15]. Figure 2.9b gives an example of a carbon nanotube field emitter array grown and used in this work.

Figure 2.9: (a) SEM of a microfabricated nanodiamond lateral FE diode reproduced from [15]; (b) SEM of a CNT emitter array utilized in this work.

2.4.2 VMD Operating Modes

The primary types of field emission devices were eluded to in the previous section on modern VMDs. The bulk of research with VMDs fabricates either ungated diode or gated triode structures; very similar to early vacuum tube development.

2.4.2.1 Vacuum Diodes

The field emission diode is the most basic field emission VMD. The emitters (and whatever electrode that might be holding them) is referred to as the cathode, and an anode is present to apply the necessary field to generate field emission. The anode also acts as the primary collector of the emitted current. Typically the cathode is held at ground, and the anode is set to a positive voltage. However, depending on application, the cathode can be negatively biased and the anode held at ground or a small positive voltage. Ideally the area of the anode is greater than the square of the distance between the cathode and anode, allowing us to disregard any edge effects.

Figure 2.10: Schematic diagrams representing electrode type and placement in a simple diode and triode configuration.

2.4.2.2 Vacuum Triodes

In a triode device, three electrodes are utilized for either amplification purposed, or in the case of vacuum field emission devices, to achieve better control over emitted electrons. In addition to the cathode and anode, a grid electrode is added in between in order to modulate the electron current. This is typically the case in vacuum tubes where small changes in the grid voltage allow for large changes in the anode current. This is why the original triode vacuum tube was termed an amplifier circuit. In the case of vacuum field emission devices, the grid in combination with the cathode establishes the field needed for electron emission. In a vacuum field emission device, the electrons will be emitted from the cathode and some will pass through the grid to continue towards a positively biased anode. A triode design will result in lower current density reaching the anode as majority of the electrons will be lost at the grid [130]. However, this is typically offset by a larger field enhancement. Figure 2.10 shows a geometrical comparison between the aforementioned diode and the triode.

2.4.3 Present VMD Challenges

There are several challenges in VMD development, and ultimately their path to commercialization. Some of the main challenges are related to lifetime of the cathode

(i.e., the emitters), the ability to integrate the nano and micro scale emitters at a system level (i.e., into an instrument), and the loss of electrons at a grid which correlates directly to power loss and the need to run the devices at higher operating levels. Lifetime of

CNT field emitters is the largest challenge of the three. Improved lifetime of the emitters will ultimately improve overall VMD development, has been investigated in great detail

[38, 65, 111, 128, 129], and even though much has been discovered on the failure mechanisms, lifetime improvement has seen little progress. It is understood that several factors play an important role in the failure of CNT emitters including: mechanical stresses causing emitter loss and gradual degradation, ion bombardment (CNTs are more resilient, but not immune), and CNT shortening which ultimately increases the applied voltage to maintain the required field for emission. Additionally, it has reported that high temperatures caused by the large flow of current are also failure mechanisms

[34, 93, 131, 132]. In regards to electron loss, it is apparent that any improvement in reducing the number of electrons that are lost at the grid and are able to make their way to the anode is an improvement to a VMD. This is because in almost all VMD development, the current at the anode is the desired signal. These three challenges are all addressed in this work. Improvement to the lifetime of CNT emitters is discussed in chapters 3 and 4, the transmission of electrons in a VMD is discussed in chapter 5, and novel packaging methods are discussed in chapters 3 and 6.

2.4.4 Duke/RTI MEMS VMD

This section will briefly introduce the two primary geometrical layouts utilized in VMD development. This will serve as a passage to introduce the platform developed

by our current research team over the past ten years. This platform served as the foundation for which the work outlined in chapters 5 and 6 were built upon.

2.4.4.1 Vertical and Lateral VMDs

The orientation chosen for device development impacts a variety of factors.

Primarily there have been two orientations explored due to practical application potential of both in vacuum micro/nanoelectronic devices. Devices can be developed in a vertical orientation where the cathode, grid(s), anode, and supporting structures are stacked vertically in some respect; the Spindt cathode is one such example. Other devices are developed laterally, where the elements are all fabricated in the same plane along the substrate. It has been shown that laterally developed devices offer many advantages, specifically related to high-frequency and high-speed applications [133-144].

Lateral devices still maintain a short electron path length as microfabrication allows for precision placement of all electrodes; this ultimately improves speed. Additionally, one of the more detrimental factors to a devices speed is the capacitance experienced between the electrodes and the substrate. A laterally constructed device reduces capacitance due to the small amount of overlapping area that exists.

2.4.4.2 History and Previous Work

In 2002, the research team consisting of Duke University and RTI International proposed a micromachined vacuum triode fabricated using a microelectromechanical systems (MEMS) platform and integrated carbon nanotube cold cathodes [145, 146]. The

device was the first demonstration of a MEMS and carbon nanotube based technology designed for on-chip power-amplifying vacuum devices. The fully integrated on-chip vacuum microtriode combined solid-state fabrication technology with high-performance nanomaterials to offer respectable current densities, transconductance values, and low power requirements. The triode structure provided a means to parameterize characterization and establish insight into the performance and design fundamentals of the device before more sophisticated devices were explored. Figure 2.11a shows the originally developed MEMS CNT FE-VMD. In 2007 - 2010, the utilization of the microtriode device as an ion source was explored [147-150]. The MEMS fabrication provided an easily amendable platform that allowed for the inclusion of additional panels to serve as ion collector electrodes on-chip. The CNT emitters provided an electron source that could be used for electron ionization (EI) of an analyte. Ion collection was achieved on the added electrodes. The ion collection efficiency of the device established its potential to perform as a miniature ion source that benefits from reduced size and cost, while still maintaining core ion source requirements. Figure 2.11b shows the MEMS CNT FE-VMD operated as an ion source with the ion collector on chip.

And in 2011 the device was investigated for its ability to expand beyond single-charge carrier operation [151]. Utilizing both electrons and ions as charge carriers, the device was able to demonstrate the first microelectronic device able to utilize both positive and negative charge states similar to how a complimentary metal-oxide-semiconductor can

be configured as a p-channel or n-channel device. The second charge state leads to the realization of additional circuit applications, in particular digital logic. Figure 2.11c shows the MEMS CNT FE-VMD operating as a bipolar vacuum microelectronic device.

The two lateral panels orthogonal to the cathode are the bipolar collectors.

2.4.4.3 Microelectromechanical Systems (MEMS)

MEMS is a technology that allows for the development of miniaturized mechanical and electro-mechanical elements using microfabrication techniques similar to modern day IC processing. Using MEMS, the field of miniaturized sensors, actuators, and VMDs has grown significantly. The MEMS devices in this work have been developed using Multi-User MEMS Processes (or MUMPs®) which is a well-established commercial program that allows for “cost-effective access to MEMS prototyping and a streamless transition into volume manufacturing” [152]. The PolyMUMPs process

integrates consecutive layers of polysilicon and sacrificial oxides, 7 physical layers total, with masking and etching steps to create 3D structures out of conductive polysilicon.

2.4.4.4 Device Assembly

The details of the microtriode fabrication process have been outlined in detail previously [145-147]. Briefly, MEMS devices received from the foundry are released from their sacrificial oxide using an acetone, isopropyl alcohol, hydrofluoric acid, DI water, and methanol process. Devices are then transported to a critical point CO2 dryer to ensure a stiction-free surface. Devices are then masked for Fe catalyst deposition using integrated MEMS masks that contain an array of patterned holes. MEMS masks are manually flipped over and secured on top of each device using optical adhesive/epoxy that is then cured with UV exposure. Devices are secured onto 4” stainless steel discs with laser etched metal masks placed over the MEMS mask where the array is located. 50 Å of Fe catalyst is deposited using electron-beam deposition.

CNTs are then grown following a similar process as described in [56]. Finally, devices are masked again using metal masks that expose the MEMS contact pads and electron- beam deposition is used to deposit 1500/5000 Å Ti/Au to be used for wire bonding.

Devices are then assembled using a probe station and micropipettes attached to the micromanipulator arms. MEMS latches are lifted up with one micropipette while the electrode panels are flipped with the other. All electrodes are locked in using this process. Finally, devices are mounted to a PCB test board using carbon tape and wire

bonds are secured. Figure 2.12a contains a process flow graphic illustrating the steps involved in device prep, and Figure 2.12b shows a device mounted on a test board.

Figure 2.12: (a) PolyMUMPS process flow with assembly; (b) device on test board.

2.4.4.5 Device Testing

Devices were tested in a custom built UHV chamber equipped with 12 electrical feedthroughs and a base pressure of 1 x 10-7 Torr. Figure 2.13 provides a picture of the test setup. Keithley model 4200-SCS Parameter Analyzer (hereinafter referred to as

Keithley 4200) equipped with six internal source measure units (SMU) was used to electrically characterize the performance of the field emission microtriode. When additional configuration options were needed, developed LabVIEW code was used to integrate the Keithley 4200 with up to seven Keithley model 2410 SMUs. Developed

LabVIEW code is available in Appendix A; programs to perform I-V and lifetime characterization are included. The Keithley 4200 provided sourcing voltages up to ± 210

V with low current measurement options. The accuracy of the current measurements

was 10-5 times the current range measured which was typically 1-10 µA, providing an accuracy of 0.1 nA. Parallel channel measurements was possible using the six source- measure units (SMU) available, which included one high accuracy preamplifier equipped SMU capable of fA accuracy. Device test protocols were developed to collect transfer and output curves from the devices.

Figure 2.13: UHV test system used for field emission device testing.

2.5 Mass Spectrometry (MS)

Chapters 3 and 6 focus on the utilization of carbon nanotube field emission based devices as the electron source for use in mass spectrometry applications. Therefore, this section will provide a brief introduction to the fundamentals of mass spectrometry.

2.5.1 Introduction to Basic Concepts

MS is an analytical technique that produces spectra of the masses of the atoms or molecules comprising a sample of material. The spectra are used to determine the molecular, elemental, or isotopic signature of a sample, the masses of particles and of molecules, and to elucidate the chemical compositions of molecules, such as peptides and other chemical compounds. MS works by ionizing chemical compounds to generate charged molecules or molecule fragments and measuring their mass-to-charge ratios

[153]. It is the role of the ion source to convert a portion of the sample into ions.

Extracted ions of varying mass are moved through a mass analyzer, sorted by their mass-to-charge ratio, and carried forward to make contact with a detector. There are several types of ion sources that are used in mass spectrometry ranging from thermionic and field emitters, to radioactive sources [153, 154]. Mass analyzers can also vary; examples include single and double focusing magnetic sectors which separate ions by mass and time of flight (TOF) mass analyzers which separate ions by time [153-156]. A simple schematic of the three main system components, ion source, mass filter, and detector, is shown in Figure 2.14.

Figure 2.14: Schematic diagram of a mass spectrometer.

2.5.2 Ionization Processes

“Ionization is the process by which an atom or a molecule acquires a negative or positive charge by gaining or losing electrons to form ions, often in conjunction with other chemical changes [157].” There are several ionization processes used in modern day mass spectrometry [153, 155]. The fundamentals of the two utilized in this work are briefly provided here.

2.5.2.1 Electron Ionization

In electron ionization MS (EI-MS), samples are introduced in the vapor state and are bombarded by energetic electrons moving quickly across an ionization region. A rapidly moving electron will not be captured by a molecule, instead however, they impact the molecule and knock away one of its electrons. The molecule will lose one of its least tightly bonded electrons, i.e. one of the electrons located in the highest occupied

molecular orbit (HOMO) [158]. The typical order electrons are lost upon EI is the lone pair first, then the π-bonded pair, then the s-bonded pair. EI is a relatively harsh technique due to the excess amount of energy typically delivered to the molecule. While the specific energy required for removal of an electron is dependent on what orbital it was occupying, the approximate energy required is 7 eV. The excess energy can be delivered to the molecule, causing excess vibrational energy, and results in breaking up of the molecule, also known as fragmentation [158]. Fragment ions, those formed from decomposed molecular ions, are abundant in EI which can mean significantly less intense molecular ion peaks [153, 155].

2.5.2.2 Chemical Ionization

In chemical ionization (CI), ions are formed by a process called ion-molecule reaction. These reactions take place when a reagent gas, typically ammonia or methane, is ionized by EI and then undergoes a chemical reaction with an introduced sample molecule. In the process an electron, proton, or another charged species is transferred between the ionized reagent gas and the sample [159]. Because samples are ionized by protonation, the results are typically the production of [M+1]+ ions ,where M is the analyte introduced. There is little excess energy in that process which results in much less fragmentation. In the CI process both positive and negative ions can be formed and analyzed providing improved identification of the sample. Schematic diagrams

illustrating the ionization process in both electron ionization and chemical ionization are shown in Figure 2.15.

Figure 2.15: Schematic diagrams showing the ionization process in both, (a) electron ionization, (b) and chemical ionization.

2.5.2.3 CNT Emitters for Mass Spectrometry

For the work in chapters 3 and 6, CNT field emitters as ionization sources in mass spectrometry (MS) applications, there are additional benefits contributing to the use of CNT field emitters over conventional emitters. During MS operation ions impact electron sources, thereby sputtering the electrode and decreasing its lifetime [160].

CNTs are composed of carbon, which has a low sputter coefficient [161], making them robust field emitters for applications involving ion creation (e.g. MS). Many groups have shown that CNT field emitters are capable of emission current exceeding 4 A/cm2 [21, 59

162, 163], demonstrating they can provide sufficient electron emission current for applications such as MS, which usually require current levels in the 1-2 mA/cm2 range

(10s to 100s of µA emitted from a point on the filament). Finally, the low turn-on fields for CNT field emitters discussed earlier provide benefits in not only enabling low power operation, but also allowing for control of the ionization energy and, in turn, an increase in the ionization rate, i.e. the ionization probability in unit time.

3. Chemical Ionization Mass Spectrometry Using Carbon Nanotube Field Emission Electron Sources

In the last five years, extensive investments have been made by the Department of

Homeland Security (DHS) into improving national security [164]. One of the eight key

Security Technologies receiving these investments is in the area of explosives trace detection (ETD) [164]. In 2011 alone, DHS awarded $15 million for the purchase of 400 fixed ETD units, and an additional $39 million to purchase 800 portable ETD machines

[164]. Quoting TSA acting administrator, Gale Rossides, “Explosive Trace Detection technology is a critical tool in our ability to stay ahead of evolving threats to aviation security……Expanding the use of this technology at checkpoints and at departure gates greatly enhances security to keep the traveling public safe [165].” As a result to this demand, the need for robust, reliable, affordable, and technologically capable instrumentation (e.g. ETD’s) is essential to the field. Mass spectrometry (MS) is a standard laboratory technology for identifying trace levels of chemical compounds in complex environments. The selectivity and specificity of chemical identification via MS across a wide range of species exceeds that of other more commonly fielded instruments, such as Raman spectroscopy and FTIR. Because the rapid and accurate identification of trace explosives is so important, mass spectrometry is of great interest in the realm of early threat detection

This chapter investigates the development of a novel packaged CNT field emission electron source for use in a commercially available mass spectrometer 61

instrument. The source was designed to integrate nanostructures with straightforward and functional packaging options to enable the efficiency of nanostructures to be realized in field deployable and desktop instrumentation. CNT emitters were grown directly on a readily available substrates and integrated into scalable packages to achieve the required emission geometries. The electron source was fully characterized and methods to improve the lifetime of the CNT emitters was also explored. The packaged source aimed at improving explosives trace detection instrumentation in homeland security circumstances (e.g. security screening for the Transportation Security

Administration, TSA). The source was designed to replace the current electron source, i.e., a filament thermionic emitter, in a commercially available instrument, i.e., a FLIR

Systems, Inc. ETD mass spectrometer. The instrument primarily operates as a chemical ionization mass spectrometer (CI-MS) due to the benefits of using CI-MS’s in explosive trace detection. The developed CNT field emission CI-MS source was evaluated by doing a direct comparison to a standard filament thermionic electron source in the same system under the same operating conditions.

The research aims to optimize device parameters including electrical performance, lifetime at operating pressure, power requirements, and the collection of mass spectra and its related parameters including resolution, collision energy, and selectivity. The research demonstrates the first reported use of a CNT-based ion source capable of collecting chemical ionization mass spectra in both positive and negative

modes; furthermore, design validation is achieved through mass spectra collection of a standard MS calibration compound (PFTBA) as well as trace explosives pertinent to homeland security applications (i.e. TNT, RDX, and PETN).

3.1 Introduction

Mass spectrometry is a standard laboratory technology for identifying trace levels of chemical compounds in complex environments. The selectivity and specificity of chemical identification via MS across a wide range of species arguably exceeds that of other more commonly fielded instruments, such as Raman spectroscopy and Fourier transform infrared spectroscopy (FTIR). For example, the accurate and rapid identification of trace explosives by MS is of great interest in the realm of early threat detection for homeland security and defense [166, 167]. The purpose of the work reported here was to evaluate the use of carbon nanotube field emitters as the electron source in a mass spectrometer designed for security checkpoint screening and reports the first utilization of a CNT field emission source for chemical ionization mass spectrometry.

Ionization of the analyte of interest is a critical step in the identification process during mass spectrometry [153]. Electron ionization and chemical ionization are both used when identifying chemical threats with MS. CI-MS is advantageous for the selective identification of explosives as they can be negatively ionized more easily than most interferents due to the strong electron affinity of the nitro group of these

compounds [168]. However, gas requirements, power consumption, and inadequate lifetime are often limiting factors for the use of conventional ion sources in portable mass spectrometers. Thus, the development of a low cost, replaceable CI mass spectrometer field emission electron source would be beneficial for use in mass spectrometers for portable applications, such as security checkpoint screening.

Field emission electron sources have several advantages for portable applications such as very small size, low power consumption, and minimized thermal degradation of the materials. Since first being reported as an electron field emitter in 1995 [18, 19, 105] carbon nanotubes have proven themselves to be a viable candidate for electron sources in various applications [106, 107]. CNTs have a high aspect ratio, small tip radius of curvature, good conductance, extremely large Young’s modulus (high stiffness), and high tensile strength allowing the tip to withstand the extremely strong fields (several

V/nm) needed for field emission [106]. Mass spectrometers usually require current levels in the 1-2 mA/cm2 range (10s to 100s of A emitted from a filament). Many groups have shown that CNT field emitters are capable of emission current exceeding 10 mA/cm2 [21,

162, 163], demonstrating they can provide sufficient electron emission current for MS.

Turn-on fields for CNT field emitters are also very low, enabling low power operation.

Wang et al. reported turn-on fields as low as 0.64 V/µm at a current density of 10 µA/cm2 from functionalized CNT structures [169]. Furthermore, the highly ordered covalently bonded crystalline structure of the CNTs provide increased stability when compared to

other metal field emitters commonly used such as tungsten and rhenium. The activation energy for surface migration of atoms on CNTs is significantly larger (17 eV) than tungsten (3.2 eV) and rhenium (2.7 eV), thus CNTs are less sensitive to surface migration

[106]. CNTs also possess a high melting temperature when compared to refractory metals such as tungsten. Additionally, ions impact electron sources during MS operation, thereby sputtering the electrode and decreasing its lifetime. CNTs are composed of carbon, which has a low sputter coefficient [161], making them robust field emitters for mass spectrometry. With so many benefits, it is not surprising that the number of CNT field emitter applications have increased in the last decade. However, while CNTs have been used as field emission sources in numerous applications [170], including EI-MS [147, 171-173], there are to the authors’ knowledge no reports that have examined CI using a CNT field emission electron source.

In this research, we evaluate a custom packaged CNT field emission electron source that allows for direct replacement of an existing filament thermionic emission electron source in a FLIR Systems, Inc. (“FLIR”) explosives trace detection desktop mass spectrometer for use in CI-MS mode. To the authors’ knowledge, the present research demonstrates the first use of a CNT field emission electron source for CI-MS. Electrical characteristics, power requirements, and lifetime for these devices have been evaluated.

Both negative and positive chemical ionization (NCI and PCI, respectively) are

examined. Mass spectra for MS calibrants and trace explosives have been obtained with this novel CI source.

3.2 Experimental Methods

Fabrication of the CNT field emission electron source involved a two-step process: CNTs were grown on prepared substrates and then mounted onto a specially designed printed circuit board (PCB) field emission platform. CNT substrates were prepared by performing a 250 µm shallow cut (or trench) with a dicing saw into 500 µm thick conductive silicon wafers to define 3.6 mm2 die sizes. The trench allowed for separation after growth of the CNT emitters by carefully snapping the wafer along this trench. A 50 Å film of iron was deposited onto the substrates by electron beam evaporation to serve as a CNT catalyst. CNTs were prepared using microwave plasma- enhanced chemical vapor deposition (MPECVD) as previously reported [56]. Briefly, the substrates were loaded into the growth chamber and treated with a 100 sccm (standard cubic centimeters per minute) ammonia heat-up cycle for 120 s to reach pretreatment conditions of 850 °C, 21 Torr, and 2.15 kW. Next, a pretreatment step consisting of 100 sccm of ammonia for 360 s de-wet the iron film into catalyst nanoparticle templates with diameters of 50-100 nm. Vertically aligned multiwalled nanotubes (VA-MWCNTs, hereafter CNTs) approximately 30 µm in height were grown using a 3:1 gas ratio of 150 sccm methane and 50 sccm ammonia for an additional 360 s. After growth, the 3.6 mm2

dies were separated by snapping along the trenches for mounting and testing. A scanning electron microscope (SEM) image of the CNT forest is shown in Figure 3.1a.

A custom PCB platform (Figure 3.1b) was designed to allow for direct fit and replacement of the existing filament thermionic emission source installed in the prototype desktop mass spectrometer. A single die with CNTs was mounted with conductive silver paste to the center electrode; the thickness of the paste is approximately 20 µm. A surrounding square electrode was then fitted with a custom 750

µm thick copper standoff fabricated by photochemical machining (Fotofab, Inc.). An extraction grid of tungsten woven wire mesh with 36% open area was attached to the top of the copper standoff, providing an extraction grid for electron field emission.

Thinner meshes were investigated, but did not provide sufficient mechanical stability.

Figure 3.1c shows devices with and without the extraction mesh above the CNT die. The

CNT tip to mesh distance was in the range of 195 µm to 205 µm with an average of 200

µm due to the thickness of the conductive silver paste.

Prior to testing in the mass spectrometer, the fabricated devices were characterized in a test bed using two Keithley 2410 source-measure units and a custom

LabVIEW program. This test bed provided a controlled environment and geometry for device characterization prior to mounting in the desktop mass spectrometer. The test bed also allowed for failure analysis and examination of the effects of controlled pressure variation. Current-Voltage (I-V) curves were measured to determine basic

device parameters and model the electrical behavior (hereafter referred to as ex-situ characterization to distinguish it from the data taken while in the mass spectrometer). I-

V curves were taken at two pressures, a low vacuum level of approximately 1x10-4 Pa

(1x10-6 Torr) to verify device operation, and 1x10-2 Pa (1x10-4 Torr) of room air, the approximate operating pressure and reagent gas used in the explosives trace detection desktop mass spectrometer during CI mode.

During the ex-situ characterization, the turn-on voltage (Vto) and threshold voltage (Vth) were measured for all devices. Vto is defined as the applied voltage when the emitted current reaches 1 nA, and Vth is defined as the voltage required to emit 1 µA of electron current. Devices were then subjected to lifetime testing at a pressure of 1x10-2

Pa of lab air until failure. Emitter current was held constant by using a feedback control loop to vary the bias applied to the extraction grid to maintain 5 µA. I-V measurements from the electrodes were collected every second. Devices that were used in the desktop

mass spectrometer tests underwent one hour of burn-in [98] at 1x10-2 Pa to condition the

CNT emitters before installing them into the mass spectrometer. Initial testing of the sources in the mass spectrometer was performed by operating the instrument with the ion source outlet flow restrictors removed so that the source was effectively performing electron ionization rather than CI. Once device performance was confirmed, flow out of the ion volume was again restricted to achieve source pressures suitable for CI.

The ion source conditions were identical when using either the CNT field emission electron source or the filament thermionic emission source. A short capillary restrictor admits the sample into an ion region heated to 180 °C with a volume of about 1 cm3. Two small orifices communicate with the bulk vacuum chamber of the mass spectrometer. A 200 µm diameter orifice in a 6 mm diameter stainless steel plate serves to admit electrons from either the filament or the CNT source. A 400 µm diameter orifice in another plate allows ions created in the ion source to enter into electrostatic lenses that focus the ion beam into the mass analyzer. The pressure in the ion source, while not measured directly in these experiments, can be calculated in the 13 to 133 Pa (0.1 to 1

Torr) range based on the conductance’s of the inlet capillary and the orifices that let gas out of the ion volume into the main vacuum chamber. This pressure range is consistent with most CI source pressures. The only gas admitted into the ion volume was lab air from directly above the sample ticket that also contains the analytes in the gas phase; therefore the “chemical ionization gas” consisted primarily of nitrogen and oxygen.

Mass spectra were acquired over the m/z range of 40 to 700 for the standard MS calibration compound perfluorotributylamine (PFTBA). Headspace vapor was leaked into the vacuum system under control of a flow restrictor and solenoid valve. Spectra for trace explosives were acquired by dosing aliquots of a solution containing the explosive in an organic solvent (typically methanol) onto a surface-sampling ticket, allowing the solvent to dry, and then analyzing the ticket via thermal desorption and transport of the explosive analyte into the CI volume. Mass analysis was performed with an ion trap analyzer coupled to the CI volume with electrostatic lenses for ion transport. Ion detection was via a conversion dynode/electron multiplier assembly.

3.3 Results and Discussion

3.3.1 Ex-Situ Characterization

Figure 3.2a displays typical ex-situ I-V characteristics for the CNT field emission electron source. The I-V curve provides both a turn-on voltage and operating bias for desired emission current levels. This information is essential to choosing operating conditions for the explosives trace detection mass spectrometer during testing. The three

I-V plots shown in Figure 3.2a were obtained with an increasing sweep voltage applied between the CNTs and the extraction grid electrode. The devices maintained reliable emission current throughout an operating range of 1 nA to a saturation level of approximately 500 µA. The saturation level is 14 mA/cm2, which is more than sufficient.

The Fowler-Nordheim (FN) plot shown in Figure 3.2b was generated from the 0

V to 450 V I-V curve of sweep #3 in Figure 3.2a. To create this plot, the FN equation was re-written in FN coordinate form (equation 1) where J denotes the emission current density, ϕ is the work function of the CNT emitters, E is the applied electric field, 훽 is the field enhancement factor, and the parameters a and b are universal constants defined in literature, 1.54 x10-6 AeVV-2 and 6.83 x103 eV-3/2 Vµm-1 respectively [174]:

J 2 3/2 ln( ) ln(ab ) [12] EE2 

The slopes of the generated FN plots indicate that our CNT field emission electron sources exhibit the non-linearity that is characteristic of CNT field emitters

[174]. This non-linear phenomenon is not clearly understood. Researchers have attributed it to space charge effects and varying electron transport tunneling mechanisms, with the transition point represented by the “knee” in the F-N plot [175-

177]. The device characterized in Figure 3.2 has a knee between the low and high field behavior at a field of 1.575 V/µm, which closely matches values reported in literature

[178]. Using the FN plot, slope coefficients were extracted and field enhancement factors

(β1 and β2) were calculated for both regions using the following equation:

 3/2  b [13] slope

For the device whose data is shown in Figure 3.2, and using a typical MWCNT value of 4.8 eV for ϕ, we observed slope coefficients of -16.67 and -33.49 for the high and low regions respectively, which yielded β values of 4309 and 2144 respectively. We observed that the computed β can diverge from the reported values by +/- 18%. This is due to the variation of the distance from the CNTs to the extraction mesh caused by the silver paste thickness. The reported β values, inclusive of the variation range, also closely coincide with literature as expected [98].

Figure 3.2c shows the ex-situ emission lifetime measurements of a device at a steady state pressure of 1x10-2 Pa (1x10-4 Torr). The device functioned for more than 320 hours of continuous operation. Assuming 10 hours of continuous operation per day, this 72

indicates a required emitter replacement period of approximately once per month.

However, assuming a reasonable duty cycle that accounts for the sample collection and traveler interactions that would occur at a security checkpoint as 25% on-time, then this

320 hour lifetime would equate to 4 months of service. Another benefit of using the

CNTs is the ability to cycle the emitter on and off with minimal degradation. A standard tungsten filament could not be cycled quickly without dramatically shortening its lifetime. Cold field emission cathodes, including these CNT field emitters, can be quickly cycled between on and off states to conserve power without affecting emitter lifetime [179]. We found that a 10 s on/off cycle had little effect on CNT cathode performance or lifetime in the test bed. Specifics of pulsing performance will be the focus of a future publication. It should be noted that an air jet directed at the CNTs during CI operation in the explosives trace detection mass spectrometer can reduce on- time, however modifications could be made to the system to correct for this.

Nevertheless, the ability to power cycle without a lifetime penalty, coupled with its low power and low temperature characteristics, make the CNT field emitters attractive options for further study and a viable candidate for CI and EI ion sources.

The increase in the voltage required to maintain 5 µA emission over time suggests gradual CNT field emitter degradation (Figure 3.2c). This degradation effect has been reported previously [132, 160, 180-182] and was related to several distinct causes, including shortening of the CNTs from ion bombardment and damage to the

sidewalls from joule heating at the CNT tip during electron field emission. The voltage spikes observed in Figure 3.2c can be attributed to sudden changes in emission current forcing fast instrument correction to maintain the set current. These sudden changes in emission current could be explained by CNTs undergoing deflection during the emitting process. It has been reported [182] transients in emission current can be caused by electrodynamic interactions between field emitting CNTs. Deflected CNTs experiencing these interactions can undergo an abrupt pull, resulting in the sudden current changes.

3.3.2 Integration into the Desktop Mass Spectrometer

The fabricated CNT field emission electron sources were incorporated into the explosives trace detection mass spectrometer. The schematic in Figure 3.3a illustrates the position of the device with respect to the ionization region, repeller electrode, sample inlet capillary, and the aperture leading to the mass analyzer. As described above, CNTs were grown on a 3.6 mm2 die, which was then mounted to the PCB with conductive silver paste. This PCB was designed to be the same size and have the same electrical connections as the filament assembly used in the desktop mass spectrometer, allowing the CNT field emission electron source to be a drop-in replacement. Figure 3.3b shows a wide view image of the explosive trace detection desktop mass spectrometer with the device installed; the wiring harness mount protrudes from the top for electrical connection. Figure 3.3c shows the CNT field emission electron source integrated into the

system just above the ionization region with electrical extensions from the harness board extending down to it.

Power requirements of the CNT field emission electron source installed in the mass spectrometer were minimal. Operating voltages varied between 500 to 1500 V with

5 to 50 µA of emission current, equating to a maximum source power requirement of 75 mW. The mass spectrometer allowed 10 ms between the time the voltage was applied to the source and the time ions were accepted from the source by the mass analyzer for emission of electrons to begin. The minimum response time of the CNT source was not interrogated, but in all cases emission was observed after this 10 ms period in the form of ion signal. Thus 10 ms serves as a maximum bound for the turn-on time. The kinetic energies of electrons emitted from the CNT field emission electron source was not directly measured, but are initially expected to be similar to the bias voltage applied before being cooled by the gas in the source.

3.3.3 Evidence for Chemical Ionization

The chemical ionization conditions for this work differ from more traditional CI investigations. The technique is normally performed using methane, isobutane, or ammonia as the CI gas [183-185]. In those cases, protonated molecules and molecular fragments tend to dominate the positive ion chemical ionization mass spectra. For the data presented here, the chemical ionization gas (lab air) does not donate a proton, but rather the following simplified mechanism dominates, according to Hunt and Crow

[186].

   N22 e70 eV  N 2 e [14]

 NMMN22   [15]

Mass spectra collection was conducted with the filament thermionic emission source operating at its standard setting of 200 µA and the CNT field emission source at

50 µA. The latter was limited by the dimensions of the CNT source and the instrument capabilities. Nonetheless, the lower current is beneficial for mobile instruments for portable operation due to its contribution to lower power. The differences in electron current could have an effect on the intensity of the mass spectral signal if the electrons are the limiting reagent in the ionization reaction. Thus, this work makes no assertion about the MS signal intensity resulting from the CNT vs the filament source. However, the current is not expected to affect the peaks that appear for each analyte nor their relative abundance. Thus, the nearly identical peak structure of the resulting mass

spectra indicate that the effect of the electrons from the CNT array, once emitted and cooled by the CI gas, are essentially identical to the effect of the electrons emitted by the filament.

Prior to collection and examination of mass spectra for explosive analytes, the instrument was calibrated using a standard MS calibration gas (PFTBA) in EI mode, as well as both positive and negative chemical ionization modes. Analysis and brief discussion of the PFTBA spectra can be found in Appendix B and provides further evidence for chemical ionization resulting from each source. After spectra were collected using the calibration gas, amounts consistent with trace standards in the security field of trinitrotoluene (TNT), Research Department explosive (RDX), and pentaerythritol tetranitrate (PETN) were examined. Identical amounts of trace explosives were introduced during spectra collection for each source. The same emission currents used for PFTBA spectra collection were also used with the trace explosives (50µA for the CNT field emission source and 200 µA for the filament thermionic source).

The normalized mass spectra for TNT collected using the CNT field emission source and filament thermionic emission source are shown in Figure 3.4a and Figure

3.4d, respectively, and correlate well with each other as well as literature spectra collected with filament thermionic emitters [185, 187]. Gas molecules in relatively high pressure ion sources can collisionally cool electrons to kinetic energies near 0 eV. These energies are low enough for attachment to molecules of compounds that have positive

electron affinities while causing little bond breakage, forming molecular anions. In the case of TNT, the process is accurately referred to as electron capture (EC). However, the

EC process in a gas that has been intentionally added to collisionally cool electrons is often referred to as negative chemical ionization (NCI) in the literature [188]. This latter convention will be utilized in the present manuscript. The mass locations and assignments for the fragments of TNT include m/z = 197 [(TNT-NO)-], m/z = 210 [(TNT-

OH)-], and an intense peak appearing at m/z = 227, as expected for NCI. The presence of the molecular anion is strong evidence of chemical ionization. Electron transfer from the reagent ions and electron capture can both result in molecular ions [187].

The normalized mass spectra for RDX collected using the CNT field emission source and filament thermionic emission source are shown in Figure 3.4b and Figure

3.4e. The data shows adduct ions that are clearly the result of ion-molecule reactions. For example, m/z 268, observed in both the CNT and filament source data, seems likely to be the result of the reaction:

  RDX NO22   RDX  NO  [16]

If EI dominated the ionization mechanism, one would expect no significant signal beyond m/z 222, which corresponds to the molecular ion of RDX. The source of the NO2- is likely RDX itself, although some NOx production in an ion source that contains both N2 and O2 may be reasonably expected. The mass locations and assignments for additional fragmentations include m/z = 102 [(C2H4N3O2)-], m/z = 129

[(C3H5N4O2)-], and m/z = 324 [(RDX+102)-]. The above peaks are consistent between the two sources as well as those observed with NCI literature spectra using filament thermionic emitters [185, 187, 189]. Finally, the small peak at m/z 176 could be attributed to [(RDX-NO2)-] [C3H6N5O4]- [185].

The normalized mass spectra for PETN collected using the CNT field emission source and filament thermionic emission source are shown in Figure 3.4c and Figure

3.4f. Similar to RDX, the ion signal for PETN at m/z 378 probably results from the following reaction:

  PETN NO32   PETN  NO  [17] as the molecular weight of PETN is 316 g/mol, so a peak at m/z 378 cannot be the result of electron ionization and must be the product of an ion-molecule reaction. Both of these product ions are clearly shown in previously published, more traditionally generated CI spectra [185]. The mass locations and assignments for additional fragments include m/z

= 62 [(NO3)-], m/z = 315 and 317 [(PETN-H)- and (PETN+H)-], and m/z = 333

[(PETN+OH)-]. The peaks are consistent between the two sources and with those in literature using filament thermionic emitters [185, 190]. The calibration curves were generally linear as well. In the case of TNT, the linear plot (which is included in supplemental information for reference) had a slope of 991, a y-intercept of -4,518, and a

R2 of 0.963. The CNT field emission electron sources show similar overall sensitivity to the standard system using filament thermionic emission.

A custom designed, fully packaged CNT field emission device has been developed and used in a prototype commercial mass spectrometer as an electron source

to obtain chemical ionization mass spectra in both positive and negative modes.

Although further study is needed, this source is a candidate for CI and EI mass spectrometry applications. The CNT field emission source demonstrated low power, low temperature, and attractive lifetime. A constant-on lifetime of 320 hours was demonstrated for the CNT devices. This lifetime equates to 4 months of operation, assuming a 25% on-time in real-time security checkpoint deployment. The CNT emitters also demonstrated the ability to cycle and thus avoid continuous-on operation, the details of which will be the subject of future publication. Spectra were collected for a standard MS calibration material (PFTBA) as well as trace amounts of explosives (TNT,

RDX, and PETN). The spectra collected showed similar peaks for both the CNT field emission source and a traditional filament thermionic emission electron source in the same system. Furthermore, the EI and CI spectra collected with the CNT field emission source closely matched those observed in literature that were produced with traditional filament thermionic emission sources. Sensitivities for the CNT field emission source were similar to filament thermionic emitters used in the same mass spectrometer.

4. Integrating Carbon Nanotube Forests into Polysilicon MEMS: Growth Kinetics, Mechanisms, and Adhesion

Chapters five and six of this work discuss development of next-generation vacuum microelectronic devices for integrated circuit and micron scale mass spectrometer ion source applications. Both technologies are built upon a previously investigated platform that encompasses two proven technologies: Micro-Electro-Mechanical Systems (MEMS) platforms and CNT field emitters. Where the research focuses on both functionality and performance improvement, as well as additional operation modes, a core fundamental concern has yet to be investigated. The electrodes fabricated within the MEMS platform are fabricated explicitly from polycrystalline silicon (poly-Si); and the carbon nanotube emitters are grown directly on the poly-Si electrodes. While growth of carbon nanotubes on single crystalline silicon has been well studied, the growth of CNTs and CNT field emitters on poly-Si has not been explored. More specifically, a general understanding of the growth mechanisms and the kinetics involved with CNT growth on poly-Si has not been investigated.

Fundamentally, it is important to have a good understanding of all material interfaces and interactions in a system. This will reduce undesired effects at a systems level. On a device performance concern, it is very important to be able to control specific properties of the CNT emitters on the device. Two important CNT emitter characteristics for device performance are the length of the emitters and the uniformity of the CNT diameters. The benefits of scaling down VMDs, specifically the ability to place the 82

cathode and the grid near each other, are easily lost if the length of the CNT emitters cannot be controlled. Too short and the required voltages are too large for reasonable performance. Too long and the CNT emitters will short the electrodes, resulting in complete device failure. Because field emission from CNTs is directly related to the field enhancement factor, and the field enhancement factor is directly related to the geometrical properties of the emitter, having control over the CNT diameter is also of great important. Additionally, the CNT emitters have a specific adhesion strength when grown on poly-Si. Adhesion is also critical to device performance. Poor attachment of the CNT emitter under the high potentials needed for field emission can result in extraction of the emitter and lead to shorting of the electrodes and device failure.

This chapter will focus on the growth of CNTs on poly-Si. The growth mechanisms, kinetics, and adhesion of the CNTs on poly-Si will be discussed and contrasted with single crystalline silicon (Si). We have shown that the lifetime and operation characteristics of devices based on those CNT emitters, are influenced by the substrate variation. These results are not only important for improving VMD designs that incorporate CNT emitters, but also for the general scientific community and all applications involving interaction between MEMS and CNTs. The results have improved the general understanding of the integration of these two enabling technologies. More specifically, the results have led us to improved development of the next generation VMD devices which will allow expansion in application.

4.1 Introduction

Vacuum microelectronic devices (VMDs) and sensors [7, 191-195] present an attractive alternative to solid-state technology in applications where solid-state devices are unreliable or do not provide adequate performance. Situations that require high frequency and high power device performance, or operation in harsh environments such as high temperature or radiation, are ideal for vacuum microelectronics. While charges flowing through semiconducting channels tend to scatter, resulting in a loss of power and signal quality, electrons in a vacuum tube are unaffected by scattering loss [195].

Despite the performance advantages in certain applications, the use of vacuum electronics has been limited due to inadequate development of a suitable cathode that can be integrated into the common vacuum microelectromechanical systems (MEMS) platform. MEMS devices are fabricated by silicon based microfabrication techniques

[196], are constructed from almost entirely polycrystalline silicon (polysilicon or poly-Si) to achieve conductive structures, and can be micromachined into desired architectures.

MEMS, a technology presently used as a platform for many types of navigation, automotive, and consumer microelectronic devices, offers a versatile and reliable microscale platform that can allow for integration of a large number of vacuum circuit elements on a single substrate.

Carbon nanotube (CNT) cold-cathode field emitters have been shown to be excellent candidates for integration with MEMS [197-199] owing to their outstanding

electrical and mechanical properties, allowing for the design and fabrication of complex and innovative integrated VMDs. To date, however, the factors involved with integrating CNTs as field emitters, specifically in situ growth of CNTs on polysilicon substrates, have not been well studied. A scalable and reliable technology requires precise control over all components and conjoined materials. For CNT field emitters, precise control of the distance between the cathode and extraction electrode is critical to controlling power requirements and to avoid electrode shorting and device failure.

Understanding CNT growth mechanisms, kinetics, and adhesion on MEMS substrate materials will enable seamless integration of CNT technology with existing standard

MEMS processing techniques. CNT diameter, forest density, and length all influence field emission characteristics. Here, we will focus on the growth kinetics as a means to developing fine control over these parameters.

While no studies have been performed on the kinetics and mechanisms of CNT growth on polysilicon to our knowledge, significant research has been performed studying growth kinetics and mechanisms for CNTs grown on monocrystalline silicon by both thermal and plasma-enhanced chemical vapor deposition (CVD). In early experiments by Baker et al. [200] on thermal growth of carbon nanofibers with catalyst particles fixed at the tips (tip-growth mode), bulk carbon diffusion through the catalyst was identified as the rate-limiting step due to the similarity of the growth activation energy with carbon diffusion. Similarly, Chhowalla et al. [201] reported a bulk-

diffusion-limited growth process in tip-growth by plasma-enhanced (PE) CVD, finding an activation energy of 1.4 eV and a growth rate that varied inversely with catalyst nanoparticle diameter [202]. Several other groups have reported activation energies in the range of 1.3 – 1.6 eV [203-205], and others have found the energy barrier to CNT growth closer to 2.2 – 2.4 eV [206-208]. Conversely, in the low-temperature regime (T =

150 – 500 °C), Hofmann et al. [209] reported a very low activation energy of 0.25 eV for tip-mode PECVD growth, attributing the low value to surface diffusion of plasma- dissociated precursors across the catalyst particle. As such, bulk diffusion through the catalyst is not required for successful CNT growth. While most growth trends are nearly linear, especially for tip-mode thermal CVD deposition techniques, parabolic growth trends have been observed as well by several groups [210-213] in thermal CVD.

This effect was attributed to gas phase diffusion-limited deposition. Wirth et al. found a

p0.6 pressure dependence of and activation energy < 1 eV for tip-growth thermal CVD

[214], attributing observations to both dissociation of acetylene on the catalyst surface and a rate-limiting step of diffusion in the catalyst. Surface-diffusion-based growth mechanisms have been described theoretically by Louchev et al. [215, 216] involving carbon precursor species traveling along the length of growing CNT forests. In addition to a lack of literature addressing CNT growth kinetics on polysilicon substrates, there is a lack of existing research studying higher temperature (800 – 900 °C) PECVD base- growth kinetics and mechanisms, especially considering a mechanism involving a rate-

limiting step of surface diffusion along the length of the growing CNT forest to the catalyst.

Despite significant efforts to understand CNT growth kinetics on monocrystalline Si, little is known about growth details on poly-Si substrates. Therefore, we study base-growth of CNTs by PECVD (catalyst particles fixed at the substrate surface) on Si (100) and polysilicon substrates, designed identically to those fabricated in the industry-standard Polysilicon Multi-User MEMS Process, or PolyMUMPs, with the goal of modeling the same type of growth that occurs on MEMS devices. We observe a parabolic growth trend between CNT forest height and deposition time, separate activation energies corresponding to different phases of growth, and present evidence in favor of a growth mechanism dominated by surface diffusion of carbon precursor along the outer surface of the CNTs, along the length to the catalyst fixed at the base of the forest. Data is then presented to explain the causes of growth differences between CNTs deposited on Si (100) and CNTs deposited on poly-Si. Finally, we engineered thin film interlayers between the catalyst and the polysilicon substrate to improve the adhesion of

CNT field emitters to the substrate.

4.2 Experimental

4.2.1 Substrate Preparation

The monocrystalline silicon substrates used for the growth of CNTs in this study were N-type conductive (100) silicon. The polycrystalline silicon wafers were likewise

fabricated at RTI International using a similar process to the standard PolyMUMPs to replicate the substrate conditions for CNT growth on a MEMS device. A 1 µm-thick polysilicon film was deposited on a 500-µm thick N-type (100) silicon wafers via low- pressure CVD (LPCVD) followed by a 200 nm thick sacrificial layer of phosphosilicate glass (PSG) to serve as a phosphorus dopant source. The wafer was then annealed at

1050 °C for 1 hour in argon to N-dope the polysilicon as well as release stress in the polysilicon film formed during the LPCVD process. Next, the substrates were cleaned in a buffered oxide etch (BOE), rinsed with DI water, and dried with N2 gas. All CNT growth substrates were coated with an iron catalyst layer 5 nm in thickness at RTI

International using a CHA electron beam evaporation system.

Improvements in CNT adhesion to polysilicon substrates were made possible through the use of 2.5 nm thick metallic layers introduced between the polysilicon substrates (prepared as described in Section 2.1) and the thin film CNT catalyst. It has also been shown [217-219] that these interlayers can result in better control of the final CNT film morphology by reducing catalyst diffusion into the poly-Si substrate during growth, but this was not directly investigated in this study. In order to compare the adhesion between CNT films with and without a metallic interlayer, samples of identical dimensions were fabricated for each of the interlayers, as well as a control sample that contained no interlayer. A dicing saw was used to precisely define

6.25 mm2 substrates. Next, a brief BOE rinse was performed to remove any residue and

then the substrate was loaded into an electron beam evaporator to deposit the refractory metal layers and iron catalyst.

4.2.2 Carbon Nanotube Growth

Microwave PECVD was employed for the growth of CNT forests on each type of substrate, which has been described in detail in previous publications [220-225]. In brief, the process occurs during two steps: heat up and growth. The heat up step involves raising the substrates to the desired deposition temperature, which is regulated by a thermocouple located directly under the quartz sample stage, followed by igniting and tuning a reducing plasma. The plasma is struck under a 100 sccm flow of NH3 at the base reactor pressure, and tuned to approximately 2.1 kW microwave power as the pressure in the reactor rises to 21 torr. During this initial temperature and plasma ramp, the iron film dewets into catalyst nanoparticles. Following pretreatment, CH4 is introduced as the carbon-containing precursor at 150 sccm flow rate, and the NH3 flow is decreased to 50 sccm.

CNTs were grown for different time intervals (1, 2, 3, 4, and 5 minutes) at temperatures of 825, 850, and 875 °C on both Si (100) and polysilicon for the purpose of identifying activation energies associated with different stages of growth. An extended range of deposition times was performed on each substrate at 850 °C in order to identify the diffusivity values of carbon feedstock through the growing CNT film. CNTs grown on each substrate possess bamboo-type inner walls as seen in Figure 4.1, grow by the

base-growth mode, and CNTs grown on polysilicon substrates frequently display a region of lower density, smaller diameter CNTs past the tips of the bulk CNT film.

4.2.3 Characterization Techniques

Scanning electron microscopy (SEM) was performed with a FEI XL30 SEM-FEG microscope to measure the CNT film thickness, and a Tecnai G2 Twin instrument was used to collect transmission electron microscopy (TEM) images. Raman spectra were

measured using a Horiba Jobin Yvon LabRam ARAMIS Raman microscope with a 633 nm HeNe laser. Adhesion tests were performed using reflectance mode of a Shimadzu

UV-3600 UV-Vis-NIR Spectrophotometer. Field emission testing to determine poor adhesion was performed in a custom UHV test chamber with Keithley 2410 SMU’s.

4.2.4 Adhesion Testing

Adhesion of the CNT emitters to the substrate was examined using the standard

American Society for Testing and Materials (ASTM) test methods for measuring adhesion of thin films [226], which assesses the adhesion of films to substrates by applying and removing pressure-sensitive tape over the film (3M Magic Tape 810D).

This procedure was combined with spectrophotometer measurements performed iteratively after each tape test in order to measure the reflectance from the underlying sample. More than ten samples from each interlayer set were tested with 50-60 spectrophotometer measurements performed on each sample for statistical relevance.

The reflectance correlated directly to the CNT density remaining on the sample after each adhesion test, which provided quantitative measurements of the degree of adhesion of the CNT film to the substrate. Each sample was mounted onto a custom 3D- printed sample holder that was designed to be compatible with the spectrophotometer.

The instrument was operated in reflectance mode, scanning wavelengths between 200 and 700 nm. After the baseline measurements, the sample holder was removed from the instrument and the first tape test was performed. Constant pressure was applied using

fixed weights that were moved uniformly across the film surface, and tape removal speed was carefully monitored; however, it has been shown [227] that variability in the speed at which the tape is removed, as well as the variability in applied pressure, do not strongly impact adhesion evaluation. After the tape was removed, the sample was loaded back into the spectrophotometer for another set of reflectance measurements.

This process was repeated until the reflectance measurements saturated, indicating that all possible CNTs were removed.

4.3 Results and Discussion

4.3.1 CNT Growth Kinetics and Activation Energies

In order to understand the differences between CNT growth on crystalline silicon substrates using PECVD and the corresponding growth on a polysilicon substrate, modeling was performed of the growth kinetics on each substrate. A model based on Fick’s first law for a thin membrane [228] was used in a similar way to the Deal and Grove model of thermal oxidation of silicon [212, 213, 229]. In short, the flux of carbon species through the boundary consisting of the tip of the CNT forest is given by

Fick’s first law, and the diffusion constant for carbon through this boundary layer as well as the thickness of the boundary are assumed constant in space and time. The steady-state assumption is also in effect, whereby the flux of carbon species impinging upon the growing CNT array is assumed equal to the flux of carbon species through the

CNT array, and also equal to the flux through the catalyst particle [230]. If N is the

number of carbon species that diffuse into the iron catalyst and are subsequently incorporated into the CNT film and F is the overall flux, the growth rate of the forest may be expressed as

dL dt = F N (18)

The general solution to this expression takes the form

2 B() t AL L (19) gr where A and B may be expressed in terms of the diffusivity D, which is proportional to exp(-Ea/kBT), kB is Boltzmann’s constant, T is the growth temperature, and tgr is the growth time [229, 230]. For short growth times, hereafter referred to as the “linear” region, equation (2) may be approximated as 퐵(푡𝑔푟) ≈ 퐴퐿, and thus the linear rate constant is B/A. For longer growth times, equation (2) may be approximated as

2 퐵(푡𝑔푟) = 퐿 which will be referred to as the “parabolic” growth regime. From the relationship between the linear rate constant and the activation energy associated with the physical process that occurs during this period (i.e. nucleation of CNTs), the activation energy may be calculated via the slope of the corresponding Arrhenius plot.

Likewise, the activation energy corresponding to the physical process occurring during the parabolic region (i.e. the diffusion of carbon feedstock from the gas phase to the catalyst through the growing nanotube forest) may be determined in the same way using the parabolic rate constant B [230]. The termination growth regime, existing for growth times exceeding the time steps considered for the parabolic region, is not

discussed here, although reports of CNT growth termination mechanisms may be found in the literature [231-234].

As discussed in the Experimental Section, to determine the activation energies for each phase of growth, growth temperatures of 825, 850, and 875 °C were used for deposition times of 1, 2, 3, 4, and 5 minutes. As such, each point in Figure 4.2 is a composite of five growth data points. During each growth experiment, Si (100) and polysilicon substrates were loaded simultaneously in the PECVD reactor, and the film thickness for each was measured by SEM. Figure 4.2 is the Arrhenius plot displaying the natural logarithm of both the linear and parabolic rate constants as a function of inverse temperature. The slope (m) of the Arrhenius curve is related to activation energy by 퐸푎 = −푘푏푚. The results of this analysis indicate activation energies during nucleation

Si Si 푆𝑖 corresponding to diffusion of carbon feedstock into the catalyst of Ea, lin Ea, lin 퐸푎,푙𝑖푛 =

푝표푙푦−푆𝑖 1.61⁡푒푉 on the Si (100) substrates, and 퐸푎,푙𝑖푛 = 1.54⁡푒푉 on polysilicon substrates.

These values closely match the literature values for bulk diffusion of carbon into γ-Fe, which have been reported in the range of 1.54-1.6 eV [235]. While γ-Fe typically does not form below a temperature of 912 °C [236], the existence of residual carbon in the

PECVD reaction chamber could allow for the formation of γ-Fe rather than α-Fe or a mixed phase. On the other hand, activation energies from the parabolic growth regime

푆𝑖 푝표푙푦−푆𝑖 were found to be 퐸푎,푝푎푟푎 = 1.90⁡푒푉 and 퐸푎,푝푎푟푎 = 3.69⁡푒푉 for CNTs grown on Si (100) and polysilicon substrates respectively.

Figure 4.2: Arrhenius plot used to determine activation energies corresponding to the linear and parabolic regions of CNT growth. The slope of the linear region for both substrates is quite similar and corresponds to bulk diffusion of carbon precursor into the γ-Fe catalyst nanoparticles. The slope in the parabolic region differs between CNTs on Si (100) and polysilicon, leading to different activation energies for the diffusion-limited phase of growth.

To understand the physical processes responsible for the increased energy barrier during CNT growth compared with bulk diffusion occurring during nucleation, as well as to understand parabolic activation energy differences between growth on each substrate, the theory of diffusion developed by Deal and Grove [229] was again employed. The overall growth rate expression used to fit growth rate data is

 4Bt (gr ) R( t ) 0.5 A 1   1 (20) A2 

where R is the height of the CNT forest, tgr is the growth time, and A and B are composite terms previously discussed, and defined as

B 2 Dn n A 2 D k , 0 s (21)

D is the diffusivity, k is the effective rate constant for the conversion of feedstock into nanotubes, n0 is the concentration of feedstock at the tips of the CNTs as estimated by the ideal gas law [237], and ns is the density of the deposited nanotubes (~0.02 g cm-3)

[212].

A more robust data set was needed to effectively fit equation (3), and thus another growth series was performed at 850 °C with growth times of 20, 30, 40, 50, 60,

120, 240, and 480 s. Equation (3) was fit to the above set of data treating A and B as constants to determine through the fit. The Levenberg-Marquardt nonlinear least squares algorithm was used to calculate values of A and B, which were subsequently used to extract diffusivity values (Figure 4.3). The diffusivity values calculated by this method were 3.5 x 10-4 cm2/s and 7.9 x 10-5 cm2/s for CNT growth on Si (100) and polysilicon substrates, respectively.

Figure 4.3: CNT forest height kinetics at 850 °C. The solid curves correspond to the fit to the Deal and Grove expression, with A and B values listed.

To determine the degree to which the growth process is diffusion-limited, a quantity referred to as the diffusion-limiting factor α was calculated in a similar fashion to Puretzky et al. [212]. The time-dependent term found under the radical in equation (3) clearly plays a crucial role in the curvature of the length vs. deposition time curve for a

CNT growth reaction, and is defined as

22 (t ) 4 Bt A 2 n k t Dn (22) gr gr0 gr s

Based on our fit of the Deal and Grove expression, the corresponding values of the diffusion-limiting factor as a function of growth time are 훼푆𝑖 = 0.012푡𝑔푟 and 훼푝표푙푦−푆𝑖 =

0.834푡𝑔푟⁡for growth at 850 °C. In order for the process to be considered diffusion-

limited, 훼 ≫ 1⁡[212]. Therefore, it is clear that growth of CNTs on polysilicon becomes diffusion-limited about 10 times faster compared with CNTs grown on Si (100), resulting in a CNT growth rate reduction earlier in the growth process on poly-Si compared to Si

(100).

4.3.2 Diffusion Mechanisms

Considering the above description of CNT growth kinetics on Si (100) and polysilicon substrates, two closely related issues have arisen. First, the activation energy for the parabolic region of growth is higher than the bulk activation energy calculated for the linear region associated with nucleation, and the activation energy found for the parabolic region of growth on polysilicon is significantly larger than the equivalent energy barrier for growth on Si (100). Second, the diffusivity for growth on polysilicon is significantly lower than the diffusivity for growth on Si (100). To understand the factors giving rise to these observations, we must first consider the physical diffusion mechanisms of the growth process under our PECVD growth conditions.

The first potential mechanism to consider is that of precursor transport through the channels in between adjacent CNTs in the growing forest, or Knudsen diffusion.

The Knudsen number, the figure of merit determining whether transport is governed by

Knudsen diffusion, is the ratio of the pore size (r) to the mean free path of the carbon- containing precursor (λMFP). Namely, if this ratio is less than 0.1, Knudsen diffusion may occur [213, 238, 239]. Mean free path may be calculated as [240]

 RT2 d2 N p (23) MFP A where R is the ideal gas constant, T is the temperature, d is the molecular diameter, NA is

Avogadro’s number, and p is the chamber pressure. Considering methane as the carbon-carrying species and the growth parameters enumerated in the Experimental

Section, 휆푀퐹푃 ≈ 7.8⁡휇푚. If the inter-CNT pore size is on the order of tens of nanometers, then the above ratio is much less than 0.1. Therefore, it is possible that some Knudsen diffusion may occur during growth. It seems unlikely, however, that this is the primary diffusion mechanism, as most carbon-containing species exist in the form of reactive radicals due to the presence of the plasma. Additionally, the CNTs possess a high defect density due to ion bombardment, which may serve as adhesion sites for reactive radicals and hinder Knudsen diffusion.

There are several pieces of supporting evidence that favor surface diffusion along the length of the CNTs as the primary diffusion-limiting mechanism during steady growth. In the case of iron-catalyzed growth in this PECVD reactor, the growth proceeds from the iron catalyst nanoparticles, which remain anchored at the substrate- nanotube interface. A growth mechanism dominated by surface diffusion must allow for diffusion of carbon species from the gas-nanotube interface to the catalyst particle at the substrate surface, so surface diffusion must occur along the length of the CNTs.

Figure 4.4: CNT forest height as a function of growth temperature on polysilicon MEMS devices. All CNT forests were grown for 150 s.

Across a wider temperature range (700 – 900 °C), we observe a general inverse relationship between CNT forest height and growth temperature on MEMS devices

(Figure 4.4). This is also reflected in the results of several other mechanistic studies [56,

201, 216, 241], and we initially attributed this effect to diffusion of the catalyst into the polysilicon substrate; however, it is unlikely that a significant amount of iron silicide exists in the catalyst nanoparticles, at least during the linear phase of growth, due to the earlier conclusion that the nanoparticles are austenite, which does not form readily if a significant amount of silicon has become incorporated into the nanoparticle [242].

Another explanation of the trend of decreasing CNT length as a function of temperature could stem from the assumption that the majority of the diffusion of carbon

100

species from the gas phase to the catalyst particle occurs by the surface diffusion mechanism along the CNT sidewalls. The average distance that a carbon species may travel along the length of a CNT is given by [216]

SDa0 exp E ads E SD 2 K B T (24)  where a0 is the distance between adsorption sites on the CNT surface, Eads is the energy barrier to adsorption, and ESD is the activation energy of surface diffusion along the length of the CNT. Under these experimental conditions and using literature values to approximate the different energy barriers, 휆푆퐷 is on the order of tens of micrometers, and this characteristic length is reduced strongly with increasing temperature. Therefore, it is possible that the reduction in CNT length with increasing temperature on polysilicon

MEMS devices is primarily due to the limitation in surface diffusion length.

In order to elucidate the surface diffusion mechanism as well as understand the large activation energy for the parabolic growth region on polysilicon substrates, Raman spectroscopy was used to measure the mean distance between defect sites on CNTs grown on each substrate. Through the Tuinstra-Koenig relationship [243], it is possible to calculate the mean crystallite size (La) of CNTs by studying the relative intensities of the D and G Raman bands. This expression was modified by Matthews et al. to take into account the wavelength of the incident laser [244]. As seen in Figure 4.5, there is a significant difference in La between CNTs grown on Si (100) and polysilicon, which were found to be 66 and 201 Å respectively. This difference in La may arise from the presence

101

of smaller catalyst nanoparticles on poly-Si substrates, discussed below in Section 4.3.3, which could produce smaller diameter and more crystalline CNTs. If it is assumed that

La is a good approximation to a0 in equation (7), the distance between adsorption sites that act as a pathway to surface diffusion along the length of the CNTs is larger for

CNTs grown on polysilicon. This may explain the large Ea for the parabolic region of growth on polysilicon (3.69 eV) compared to CNTs grown on Si (100) (1.90 eV), as the energy barrier for transport between adsorption sites should increase with increasing distance between adsorption sites. In addition, the parabolic activation energy for CNTs grown on Si (100) is close to the reported value for carbon adsorption on CNT sidewalls

[216, 245].

Figure 4.5: Raman spectra of CNTs grown on Si (100) and polysilicon substrates. The crystallite size is 66 Å for CNT on Si (100) and 201 Å for CNTs on polysilicon.

102

More support of surface diffusion as the primary diffusion mechanism in this system comes from a hybrid carbon nanostructure consisting of few-layered graphene covalently bonded to the sidewalls of CNTs referred to as graphenated carbon nanotubes (g-CNTs). As reported elsewhere [220-224, 246], this hybrid nanostructure

(Figure 4.6) usually requires an elevated temperature (~1050 °C) in order to form the leaves of few-layered graphene, referred to as “foliates.” The growth mechanism of these foliates has been proposed as either a stress-buckling mechanism [221] where varying growth rates between concentric nanotube walls causes buckling and fracture, allowing growth of the foliate at the fracture site, or an ion bombardment mechanism

[222] where energetic ions from the plasma environment cause defects in the outer nanotube walls, allowing foliates to nucleate and grow. In either case, defect sites in the outer walls of the CNTs exist at high spatial frequency, which is also the pathway by which surface diffusion-driven growth occurs. However, at the elevated temperatures used to grow g-CNTs, growth typically terminates at about 10 μm, after which point growth of foliates begins [221, 222]. This lower terminal length is in agreement with a surface diffusion mechanism that limits the maximum forest length as temperature is increased. If the temperature during growth remains lower (~850 °C), these g-CNT structures may also form, but only when the growth time is much longer [223]. These lower temperature g-CNTs also form foliates only after the termination length of the forest is reached, which is typically about 40-50 μm at 850 °C.

103

Figure 4.6: Plan view SEM micrograph of graphenated CNTs and b) higher magnification image of a g-CNT. Inset: TEM micrograph of a few-layered graphene foliate terminating in 3-5 graphene layers.

There are two heat sources during the growth reaction: the first is the substrate heater underneath the growth substrate, and the second is the plasma above the growing CNT array. The heat generated in the CNT film by ion- and reactive radical- bombardment from the plasma, powered by 2.1 kW microwave power, should be higher than that of the substrate heater, so a thermal gradient exists in the downward direction.

As a result, the CNT tips are the hottest regions of the film. After growth terminates due to the limitation in surface diffusion length, the tips continue to experience reactive carbon-containing radicals impinging upon them. As the tips are at a higher temperature than the lower regions of the CNTs due to the aforementioned thermal gradient as well as heat loss occurring from conductive losses along the length of the tube and convective loss from gas molecules in the region between CNTs (due to some limited Knudsen diffusion), the surface diffusion length is smallest for this region. As 104

carbon species adsorb onto the CNT tips, surface diffusion may occur over a very small length scale, especially at higher growth temperatures, and nucleate foliates at the defect site where it comes to rest. In this way, surface diffusion seems the most likely mechanism in light of the evidence presented for this hybrid nanostructure.

As a final point on the diffusion mechanism, it has been observed that when the temperature is increased further (1100 °C or higher), CNTs or g-CNTs are no longer present in the resultant film. Instead, vertically oriented sheets of few-layered graphene are deposited on the substrate without any nanotube structure present [222]. As the

CNTs grown by this PECVD reaction have bamboo-type inner CNT walls, it is possible that the surface diffusion length is even smaller for these conditions than the length of a single bamboo segment, preventing the growth of nanotubes. Bamboo-type growth has also been associated with a surface diffusion mechanism [215].

4.3.3 Causes of Reduced Diffusivity for CNTs grown on Polysilicon

Thus far, we have discussed the kinetics of CNT growth on polysilicon as compared to growth on Si (100), examined the physical interpretation of the associated activation energies, and considered diffusion mechanisms. Here, we hope to understand the reason why growth on polysilicon has lower diffusivity and becomes diffusion-limited earlier in the growth process compared with CNTs grown on Si (100).

It is well known that the diameter of the catalyst nanoparticles is closely related to the diameter and growth rate of the resultant CNT [201, 247, 248]. Catalyst

105

nanoparticle diameter and CNT diameter are directly proportional, and CNT diameter and CNT growth rate are inversely proportional. It is therefore important to examine the way in which the iron catalyst film dewets on polysilicon in relation to Si (100).

Figure 4.7 shows SEM micrographs of catalyst nanoparticles forming on each substrate at 850 °C, prepared by terminating the CNT growth process before introduction of the carbon precursor. From Figure 4.7, it is clear that nanoparticles that dewet on polysilicon are less uniform, with the highest frequency nanoparticle diameter between

20 and 30 nm, and form a bimodal distribution. In contrast, the nanoparticles that form on Si (100) substrates possess more uniform diameters and are larger on average.

Additionally, the nanoparticles that form on top of the grains on polysilicon substrates tend to be larger (~60-70 nm) and smaller particles aggregate near the grain boundaries.

The lower mode in our bimodal distribution is attributed to particles formed around the grain boundaries, where the crystal facets are expected to vary. This is in agreement with literature [249], in which smaller particles form on the Si (111) crystal face (25-35 nm) and larger particles (55-65 nm) form on Si (100) due to differences in surface energy.

While catalyst thickness and pretreatment time impact the resultant nanoparticle dimensions [250, 251], these parameters were very similar in the referenced study compared to the present experimental conditions.

106

As mentioned above, the catalyst nanoparticle diameter influences the resultant

CNT diameter as well as growth rate. Since the nanoparticle diameters are dispersed across a range of values more so on polysilicon compared to Si (100) substrates, there is also a dispersion of growth rate of the nanotubes catalyzed by these particles. Due to the close proximity of CNTs to one another during the growth process, van der Waals forces [252, 253] may cause neighboring CNTs with differing growth rates to collectively bend and become tortuous. Since the dispersion of nanoparticle diameters is higher for

CNTs grown on polysilicon due to the variety of surface energies displayed by the various exposed crystal faces, the tortuosity of these films is expected to be higher. CNT films with a higher degree of tortuosity have a lower apparent film thickness, assuming equal CNT length, due to the compression of CNT length along the vertical direction

107

[254]. Therefore, an increase in tortuosity would lead to a reduction in apparent growth rate and a reduction in diffusivity, as is observed for CNT growth on polysilicon compared to growth on Si (100). This discrepancy would be eliminated if real CNT length were used for kinetics calculations rather than film thickness. Additionally, from

SEM analysis (Figure 4.8), CNTs grown on polysilicon possess a much higher degree of tortuosity compared with CNTs grown on Si (100). This may also help to explain why the diffusivity is markedly lower for growth on polysilicon and the parabolic activation energy is larger, as the effective growth rate disparity is reduced when taking tortuosity into account.

4.3.4 Improving CNT Adhesion to Polysilicon

A practical requirement for the industrial viability of a packaged carbon nanotube field emission vacuum microelectronic device (CNT FE-VMD) is the ability to integrate a VMD platform with cathodes that can provide high current density and long 108

lifetime. There have been numerous studies [38, 255, 256] on the failure mechanisms of stand-alone CNT field emitters, and there are several possible causes of failure.

Examples include shortening of the CNT emitter length during operation over time due to enhanced oxidative ablation of the tube caused by local resistive heating (i.e. thermo- mechanically activated fracture) [255, 257], degradation due to ion irradiation/bombardment and the concurrent degradation of the CNT [160, 258], and extraction of the CNT from the substrate due to either electrodynamic force-activated failure or increased resistive heating at the substrate-CNT interface [255, 256]. In the case of CNT emitters grown on MEMS polysilicon-based substrates, poor adhesion between the CNT emitter and its substrate is the primary limiting factor for overall device lifetime. A loss in adhesion between the emitter and the substrate will not only render the emitter unable to contribute to the total emission current density, but it could lead to catastrophic device failure if that emitter bridges electrically isolated components

(e.g. the cathode and the extraction grid). A common method to improve adhesion of thin films is to use an adhesion layer between the substrate and the film, and because

CNTs nucleate from a thin film catalyst, we investigated the integration of an adhesion layer between the polysilicon substrate and catalyst to improve adhesion of the CNT film. Many groups have explored adhesion layers for CNT films on crystalline silicon and metallic substrates [259-261]. Titanium and molybdenum adhesion layers were specifically chosen based on the best literature results on crystalline silicon, and we

109

compared them to a control sample without an adhesion layer. The titanium interlayer produced the best adhesion compared to the control and the molybdenum interlayer samples (Fig. 9). The reflectance for the as-prepared films and the average reflectance after each subsequent tape pull were directly compared to the other interlayer samples by measuring against control samples for each metal interlayer without a CNT forest.

The titanium interlayer samples required approximately 2.5 times more adhesion tests to reach saturation when compared to the control sample, whereas the molybdenum showed only a small improvement of approximately 1.3 times more than the control sample. It was also seen throughout all the samples tested that the titanium interlayer samples had a final reflectance saturation point of approximately 35%, whereas the control and molybdenum samples averaged approximately 40%. This could indicate that more CNT emitters remain permanently adhered (i.e. have an adherence strength larger than the removal force of the tape) after all adhesion tests were performed when compared to the control and molybdenum samples. This data indicates that using a titanium interlayer for CNT emitter growth on MEMS structures can offer improved adhesion properties, which could correspond to improved device lifetime. Furthermore, additional interactions could exist at the interface between the CNT emitter and the polycrystalline substrate that could reduce contact resistance and improve field emission properties of the emitter, while also reducing resistive heating at the substrate-CNT

110

interface during operation of the emitter. These interactions will be the subject of future research.

4.4 Conclusions

Growth of CNTs on polycrystalline silicon substrates has been studied to improve understanding of growth mechanics for applications in MEMS technology. The kinetics of CNT growth on polysilicon was elucidated using the model of Deal and

Grove [229] to understand the activation energies and mechanisms for base-mediated

111

growth using microwave PECVD. These results were compared to CNT growth on crystalline (100) silicon substrates, finding nucleation-stage activation energies that match literature values for bulk carbon diffusion into austenite catalyst nanoparticles.

Parabolic diffusion-limited growth was observed on each substrate, with activation energies for the diffusion-limited growth phase of 1.90 and 3.69 eV for growth on Si

(100) and polysilicon, respectively, in the temperature range of 825 – 875 °C. A difference in overall diffusivity of carbon species through the growing CNT network was found, with values of 3.5 x 10-4 cm2/s for growth on Si (100) and 7.9 x 10-5 cm2/s for growth on polysilicon at 850 °C. In addition, deposition became diffusion limited earlier in the growth process using polysilicon substrates. Evidence was presented in favor of a growth mechanism involving surface diffusion of carbon species along the length of the growing CNTs from the gas phase to the catalyst at the base of the CNT forest, but some limited Knudsen diffusion may also occur. Possible reasons for this difference in diffusivity and the activation energy differences were explored through analysis of the catalyst nanoparticle dewetting process as well as subsequent changes in CNT tortuosity and alignment, as measured by SEM. Finally, interlayer addition techniques were presented to improve CNT adhesion to polysilicon substrates for applications as cold- cathode field emission sources for MEMS integration.

112

5. Improved Performance of Field Emission Vacuum Microelectronic Devices for Integrated Circuits

Vacuum electronics based on thermionic emission and cold cathode emission have been a focus of research and development for many years. It was the vacuum tube amplifier that pioneered the use of vacuum electronics in numerous applications.

Eventually solid-state electronics consumed the market due to their improved efficiency, versatility, and reliability when compared to the original vacuum tube. However, vacuum electronics remained significant in select applications, specifically when high power and high frequency were desired. As micro and nanofabrication technology began to grow in the late 1960’s to 1980’s, so did the electronic device communities interest in vacuum microelectronics. In addition to existing markets where vacuum electronics already exist, there are others that could potentially benefit from high performance and miniaturized forms of the existing vacuum technology. The feasibility of developing reliable vacuum electronics on the micron or nano scale has continued to grow with advances in micromachining techniques. Versatile vacuum microelectronic devices that can serve as compliments, or even replacements, to solid-state electronics currently face several roadblocks preventing their widespread adoption. It is these challenges that have inspired much research in the field in the last decade.

Our research team from Duke University and RTI International, has dedicated over a decade to the development of a novel vacuum microelectronic platform based on

MEMS technology and carbon nanotube cold cathode field emitters [109, 145-151]. The 113

platform has been explored for a variety of applications from a basic three terminal active circuit element for integrated circuits, to a micro-ion source for mass spectrometer and sensor development. The work presented in this chapter aims to improve the performance of the previously investigated three terminal device. Specifically, this study employs simulation and experiment to investigate two key concerns related to the utilization of vacuum microelectronic devices (VMDs) in integrated circuits. Whereas recent work in the community has shown great promise towards a device technology for harsh environments, important issues with their performance on an integrated platform have not been addressed. This chapter examines integrated structures to develop a solution to device cross-talk caused by densely packing VMDs on a single substrate. This is an important development for the realization of the integrated vacuum circuit.

Additionally, this chapter investigates another key performance concern with VMDs; electron transmission (i.e., electron grid loss). A solution using novel designs and operating modes to reduce grid loss by nearly 3x is demonstrated.

5.1 Introduction

The vacuum tube amplifier pioneered the communication industry and early computing technology. However, the invention of the solid-state transistor in the late

1940’s [26], and the development of the integrated circuit in the late 1950’s [27, 28] resulted in vacuum tubes being replaced in nearly all modern computing and related fields. However, vacuum microelectronics, a field of science witnessing the return of

114

vacuum technology with micron-scale vacuum devices, has experienced incredible growth recently. The development can be widely attributed to the continued advancements in microfabrication processing technology as well as the pioneering work of Spindt, Brodie, and many others in the 1960’s through 1990’s [7, 262, 263]. The promising performance of vacuum microelectronic devices (VMDs) along with limited ability of solid-state semiconductor microelectronics to perform in harsh environments, such as severe temperatures or intense radiation, has inspired investigation of their use in a variety of applications including: high-power/speed amplifiers and switches, field emission displays, novel MEMS sensors, communication system components for aviation, Hall Effect thrusters, electrodynamic tethers, traveling wave tubes, space satellite communication and various types of radar), and finally key elements for VMD based integrated circuits [4, 22, 122-124, 264-266]. Therefore, vacuum microelectronic review articles are consistently published every few years capturing the essential breakthroughs in the field [264, 267, 268]. Recently, researchers [269, 270] have even shown that nanoscale dimension VMDs are possible.

High performance VMDs are accomplished via the integration of field emission cathodes, often referred to as cold cathode field emitters, as well as leveraging the rising maturity of semiconductor processing to fabricate field emission vacuum microelectronic devices (FE-VMDs) [86, 264]. The physics of field emission in VMDs has been described in detail previously [271]. Briefly, field emission is the process by which

115

electrons are generated (i.e., emitted) via the application of an electric field between a cathode material and an anode. This is in contrast to thermionic emitters which require high temperatures to “boil” off electrons, thus creating the need for high input power and inefficient operation. The use of a vacuum channel in the field emission cathode reduces power loss and degradation of signal quality caused by electron crystal-lattice scattering in solid state devices (SSD), whose mobility is dominated by collisions.

Therefore, these “junction-free” vacuum devices benefit from current densities greater than 100 A/cm2, uninterrupted ballistic electron transport in a vacuum channel with a velocity 1,000 times that of a SSD, less energy loss, and no carrier freeze-out [195, 272].

In regards to the aforementioned radiation benefits of FE-VMDs, a vacuum channel is immune to defect generation (e.g. lattice displacement) and other ionization effects caused by ionizing radiation and charged particles. Those two fundamental damage mechanisms in crystalline material based microelectronics are the primary cause of performance degradation (e.g. increased electronic noise, signal spikes, and minority carrier depletion) and overall failure (e.g. a latchup or an ionizing dose surpassing the devices threshold) that necessitate radiation-hardening techniques. With copious potential benefits in circuit applications, a next logical step is to move the technology towards integration of multiple VMDs, which others have started to explore [125, 126,

270]. However, the ability to densely pack VMDs on-chip without sacrificing individual device or overall circuit performance, is an important concept that has not been explored

116

in detail. Additionally, improving the performance of the individual VMD is also still of concern.

Our previous work [146-150, 273, 274] presented the integration of MEMS-based vacuum devices with a unique application of cold cathodes employing carbon nanotube

(CNT) electron field emitters to achieve a more reliable platform for FE-VMD circuits.

The devices are fabricated using the well-known PolyMUMPS [152] process, allowing for seamless integration of three-dimensional conductive polysilicon electrode panels that are locked in position orthogonal to the substrate for improved performance. To date, we have investigated the development of an active vacuum circuit component consisting of a cathode, extraction grid, and anode. The design is amenable to integration, however, the effects of tightly packing FE-VMDs on the performance have not previously been investigated. Furthermore, loss of current at a grid (i.e., poor transmission properties) in a VMD is a performance concern that limits the capability of integrated vacuum circuits. For example, devices with only 8% of the emission current transmitting to the anode and 89% grid loss were utilized in a recent NASA and Air

Force Institute of Technology mission [275] launched in late 2013 to study carbon nanotube emitter array efficiencies in space [130, 276]. The overall effectiveness of their

CNT electron source was limited due to most of the electrons being captured by the gate.

Our previous microtriode designs averaged approximately 75% grid loss. Improving

117

transmission through the vacuum channel by reducing the loss of electrons to the grid will improve overall device performance.

This work presents a systematic study of proximity effects and transmission efficiency in integrated FE-VMD devices as circuit components using both experimental data and simulations. Experimental data reveals that electron transmission in the vacuum channel is negatively influenced by neighboring devices with crosstalk on the order of 10-14% in the absence of isolation structures. Electrostatic field and particle simulation software was used to design optimal structures and modified device geometries to nearly eliminate these proximity effects. Furthermore, simulations were used to design geometries that improved electron transmission to the working anode by more than 3x, from 25% to 80%. Finally, the improved devices were fabricated and tested to validate the simulation results

5.2 Experimental Methods

The FE-VMD devices consist of a series of lateral polysilicon panels, and the fabrication has been described in detail elsewhere [146, 273, 274]. Briefly, the devices were fabricated in a well-known polysilicon MEMS micromachining process

(PolyMUMPS®, Memscap, Inc.). Iron catalyst for CNT growth was selectively deposited on the cathodes using integrated MEMS masks, and CNT emitters were grown in a 915

MHz microwave plasma enhanced chemical vapor deposition (PECVD). The devices were assembled by rotating and locking the electrodes into a vertical position. Because

118

the fabrication of high-performance integrated vacuum microelectronic circuits such as logic circuits and oscillators requires that multiple active devices operate in close proximity to eliminate internal resistances, we investigated devices separated by less than 100 µm. A modified dual microtriode design consisting of a shared cathode panel was fabricated to allow for two devices to be placed in close proximity on-chip. A schematic diagram depicting the geometry layout and biasing process for this design can be seen in Figure 5.1a; the labeled IN and OUT represent the potential inputs and output for logic circuit realization. A SEM image of the overall fabricated device is seen in Figure 5.1b with an inset SEM of the CNT field emitter arrays. The versatility of the fabrication process allowed us to demonstrate the addition of new structures and significant geometry changes without increasing fabrication complexity (e.g. additional photolithography steps). We designed new devices to investigate methods for reducing device crosstalk and overall transmission (i.e., reduce grid loss). Two key designs were fabricated: 1) a dual three panel traditional microtriode with center shielding electrode, and 2) a new dual four panel microtetrode to investigate transmission benefits and ultimately serve as a high-performance building block for a vacuum integrated circuit.

119

Testing was conducted in a custom vacuum chamber including twelve electrical feedthroughs for a variety of electrode biasing options. All measurements were performed at room temperature with a base pressure of 1 x 10-7 torr. Custom LabVIEW software controlling a multichannel semiconductor analyzer (Keithley 4200-SCS), and four SourceMeter instruments (Keithley 2410s), was used to electrically characterize the devices. The charged particle simulation software SIMION® 8.2 was used to evaluate three-dimensional models of the vacuum circuits. Those models enabled rapid and accurate simulation of numerous geometries to optimize device designs. A feedback

120

loop was implemented between simulations and experiments to improve device performance. The CNT emitter length was controlled through our growth recipes, and determined the distance between the cathode and the grid, thereby influencing the emission field requirements and overall DC device characteristics. A uniform surface emission model was chosen for the simulations in which electrons were uniformly emitted from the surface of the CNT emitter array bundles. This best represents the emission taking place on the device as the CNT emitter arrays, as grown, consist of a range of protruding emission tips across the bundle face which can be seen in Figure 5.1.

5.3 Results and Discussion

5.3.1 Determination of Proximity Effects

It is important to both understand and control the electron current that is emitted from the CNT emitters, passes through the grid, and propagates across the vacuum channel in order to establish viable transistor performance (e.g. overall power gain, clear cutoff, saturation regions, and transconductance). Therefore, we measured cathode emission current (Ic) as well as the current collected by the grid (Ig) and anode (Ia) as a function of extraction grid voltages (Vg). Measurements were repeated for several devices to validate statistical relevance. Fowler-Nordheim emission tunneling theory

[277] predicts increasing grid voltage will result in an exponential increase in the cathode emission current. Furthermore, there is a threshold voltage (Vth) that must be reached in order to achieve the necessary field strength to begin electron emission. In

121

addition to forward biasing the cathode and extraction grid to achieve field emission, the anode is biased so that electrons that pass through the grid continue towards the anode, i.e. the output of the three-terminal device. If we consider biasing all three panels as a device in ON mode, we observed that two ON devices in close proximity had little effect on one another. Even though there was grid loss in the devices, adequate signal was observed at both anodes. Figure 5.2a displays current-voltage (I-V) characteristic curves to show the overall current collected from each of the two grids and two anodes for one of the devices tested. The visible difference in turn-on voltage is due to small differences in CNT emitter lengths between device 1 and device 2. For all tests, the shared cathode was grounded and potentials above the threshold voltages were applied to the grids and anodes to generate current through both devices. The cathode was held at ground throughout all the tests and the varying parameter was the potential on the grids and anodes. Each grid was incremented to a value that well surpassed their threshold voltage in order to generate substantial emission current from the corresponding CNT emitter arrays. We typically observed sizeable grid loss during characterization experiments of the microtriodes. This means there was more current registering on the grid than on the anode, and this undesired performance characteristic was the basis for the work reported in later sections. Overall, however, we observed no concerns with the dual operating mode as both devices performed as they should with sufficient current reaching the anodes. Additionally, if we reverse biased the cathode

122

and grid for each device, turning both of them OFF, we did not measure a current at either anode as expected.

The difficulty arises when we attempt to operate one device ON and one device

OFF as shown in Figure 5.2b. This mode should register anode current only on the device whose cathode and grid are in forward bias conditions. We observed, however, unintended signal in the nominally OFF device when taking I-V curves from device 1. In this operating mode the grid of device 2 was grounded to ensure no field emission was attained from the CNT emitters of device 2. An additional parameter was also added into the test; the potential on the anode of nominally OFF device 2 was incremented by 1

V after each I-V sweep of device 1; 160 V was used as a stopping point for that OFF anode. The currents measured on the grid and anode of device 1 were relatively equal for all I-V sweeps performed. The data is plotted such that the main plot displays the current measured on the grid and anode of the ON state device (#1). The inset displays the current measured on the grid and anode of the nominally OFF device (#2). We observed that even though the current registering at the grid of the nominally OFF device (#2) was essentially zero, there is significant current being collected at the anode of device 2. This indicates a portion of the electrons from the ON state device are being collected at the anode of the nominally OFF device. The same experiment was repeated in the opposite configuration (device 1 OFF and device 2 ON) with similar results. On average, 10-15% of the electron current from the ON device reached the nominally OFF

123

device’s anode. This is an undesirable outcome if we are attempting to read a low signal

(or no current) at the nominally OFF anode. Efforts to mitigate the undesired crosstalk between densely packed devices was then explored using simulation based modeling.

5.3.2 Simulation Guided Analysis and Redesign

The SIMION® 8.2 software package was used to confirm the source of the observed proximity effects as well as investigate and optimize the MEMS geometry to mitigate the problem without compromising device performance. The software

124

numerically solves the Laplace equations for the given potentials; a lattice with constant mesh size is used to approximate the derivatives by finite differences with optimized linear time solving. The MEMS electrodes were precisely defined using geometry files with a mesh size of two voxels per micron to allow for extraction of the electric fields and potentials at all points around the electrodes. Therefore we were able to systematically model the electric potential distributions and electron trajectories around each MEMS structure, and plot trajectories visibly and numerically to reveal all electron paths in the device. Initial models were created to replicate the exact geometries and potentials in the dual device setup. Support structures are included in the simulations to establish realistic conditions as these structures have an impact on electron trajectories.

The simulation results given in Figure 5.3 provide insight as to why we measured current on the nominally OFF device’s anode (Figure 5.3). The cathode, extraction grid, and anode panels, being only 2 microns thick, are very thin in comparison to the support structures and therefore have been accentuated with black lines in Figure 5.3 and Figure

5.4. The electric fields concentrated around the structures of the nominally OFF device penetrate into the vacuum channel of the ON device altering the electron trajectories; the electrons are therefore attracted towards and collected at the anode of the nominally

OFF device as shown in Figure 5.3. A portion of the red colored electrons (current from the ON device) can be seen being pulled and collected at the nominally OFF device, whereas we clearly observe all of the blue colored electrons (current from the nominally

125

OFF device) are stopped by the grid of the nominally OFF device and are not contributing to the current at the nominally OFF device’s anode. The simulation results confirm the experimental data observed from earlier dual device tests; undesired anode current magnitudes (i.e., the amount of current reaching the nominally OFF device’s anode) were approximately 10-15% of the total CNT electron emission current in both the experimental and simulation results.

SIMION® was further employed to investigate additional geometry designs, placement, and operating potentials to mitigate the crosstalk between devices without adding significant complexity to the device. Simulations revealed that a dividing strip or

“shielding electrode” on the substrate produced sufficient isolation, or shielding, to the

126

vacuum channels of neighboring devices. Simulation-based optimization was employed to determine thickness, spacing, and operating potential that eliminated proximity effects without interfering with the particle trajectories of the active devices. Figure 5.4 summarizes the effect of the shielding electrode. Figure 5.4a shows a scenario with one

ON and one nominally OFF device and the shielding electrode voltage off. Here we recreate the crosstalk effects as seen in the previous design. Figure 5.4b uses the same simulation parameters except with the shielding electrode grounded, thereby isolating the vacuum channels of each device and eliminating the undesired current on the anode of the nominally OFF device. A floor based electrode is simple in design and does not add complexity during assembly; it is easily fabricated alongside the existing structures during the MEMS process steps. The proposed redesigns were fabricated and tested experimentally to validate the simulation results.

127

the OFF device 2 (right); (b) same design with the shielding electrode grounded, effectively shielding the devices and their corresponding vacuum channels from one another. Operating the shielding electrode at ground potential provided the best shielding without interfering with individual device performance.

5.3.3 Simulation for Electron Transmission Analysis and Redesign

As mentioned previously, prior work has reported sizeable grid loss in the

MEMS microtriodes. Because electron field emission from the cathode is directly coupled to the voltage on the grid, we inevitably cannot use the grid to focus the electron beam which results in poor electron transmission to the anode. Therefore devices must be operated at larger emission currents to compensate, resulting in more mechanical and electrical stress on the device, as well as higher power requirements.

Furthermore, coupling of the grid and cathode gives rise to poor transconductance, an undesirable quality for designing high-performance integrated circuits. To date the best measured transconductance for the MEMS microtriode is 2 µs [198]. SIMION® was again used to investigate methods for improving the overall electron transmission through the device. By adding additional polysilicon electrodes to optimally shape the equipotential curves in the vacuum channel and around the device, we reduced the angular and energy dispersion of the emitted electrons, thus improving the collimations of the electron beam. The geometries from the original microtriode produced an average transmission (i.e. electrons moving from the cathode, through the grid, and to the anode) of 25% [147, 273] which we replicated in SIMION® simulations. Figure 5.5 shows the large number of electrons lost at the grid and not transmitting to the anode.

128

SIMION® was used to determine the number, spacings, and potentials on the added electrodes to optimize transmission. With the added electrodes, we were able to mimic the effect of a Wehnelt cylinder lens used in electron optics [278]. The screening and extraction electrodes were designed to create a pre-lens electric field effect around the

CNT emitter bundles to dramatically enhance the focusing properties of the electron beam by 3x. The 25% grid transmission efficiency of the old geometries was improved to

80% with these new designs. The simulation results can be seen in Figure 5.5. The pre- lens effect around the CNT emitters is seen by plotting the equipotential curves between the electrodes (Figure 5.5b). The pre-lens effect is further supported by the flat equipotential curves on the back side of the screening grid which significantly reduces angular dispersion, resulting in improved beam focusing (Figure 5.5c). Another benefit to this design is that we have now decoupled the extraction grid from the cathode.

Additional simulations have revealed that the electron beam emitted from the CNTs can be left in constant emission mode and fully retarded via small changes in the screening electrode voltage, similar to tetrode-style vacuum tubes. This can lead to improved device transconductance, and these experiments will be the focus of future work.

129

5.3.4 Experimental Confirmation

The simulations were verified experimentally with devices including the shielding electrode and devices including the additional electrodes for improved focusing. A schematic diagram depicting the geometry layout and biasing process for the microtetrode design can be seen in Figure 5.6a; the labeled IN’s and OUT’s represent the new potential inputs and output for logic circuit realization. An SEM image of the newly fabricated microtetrode containing the center shielding electrode can be seen in

Figure 5.6b. The redesigned microtriodes with the center shielding electrode (not shown) are similar just without the addition of the extra grid.

130

In order to compare and validate improvements of the dual microtriode device, initial tests were performed with the center shielding electrode off to mimic the original devices. Again, crosstalk was observed in similar percentages (10-15%), the results from a device can be seen in Figure 5.7a. Again, the main plot displays the currents collected at the grid and anode of the ON device, and the inset displays the currents collected at the gird and anode of the nominally OFF device. Results show current collected at the

OFF anode without current at the OFF grid. The tests were repeated with a ground potential applied to the shielding electrode and the current collected at the anode of the nominally OFF device was observed to decrease from the 15% to 1-2% on average. This is an order of magnitude decrease is crosstalk current. The results of the significantly reduced anode current at the nominally OFF device are plotted in Figure 5.7b. These results show a grounded center electrode properly isolate the vacuum channels. 131

Potential transmission benefits were investigated using one side of the dual microtetrode design. The current reaching the anode was approximately 2.6 to 2.8 times more than in the microtriode device. Transmission values of 65-70% on average were measured. Figure 5.8 compares the results from an old microtriode device to a new microtetrode. Figure 5.8a shows most of the current being intercepted by the grid in the old microtriode design with an average of 25% of the emitted current reaching the anode. Figure 5.8b shows much less grid loss in the new 4 panel microtetrode design with 65-70% of emitted current reaching the anode. The failure to reach the 80%

132

transmission predicted by the simulation may be due to small misalignment between the

CNT emitters and the grid holes.

Densely packed FE-VMDs using a novel MEMS and CNT emitter based platform were fabricated and tested. Undesired proximity effects in the form of current crosstalk between devices caused by the biasing potentials required for single device operation was observed. On average, 10-15% of the emission current from one device is inadvertently collected by the nominally OFF device. It was determined these proximity

133

effects could not be neglected in the operation of high-performance vacuum microelectronic device circuits. Simulation modeling was used to understand the root cause, design structural solutions, and improve overall device performance by methods that could be easily integrated into the device fabrication process. Simulation modeling was also used to investigate the reduction of grid loss and improve the overall transmission of electrons from the cathode to the anode from 25% to 80%, an important parameter improving power gain in the key active vacuum circuit component. This improvement was achieved by adding an additional electrode panel that decoupled the cathode and grid and improved the focusing of the emitted electrons. Finally, devices were fabricated and tested to validate the simulations. The crosstalk mitigation structure effectively reduced current crosstalk by an order of magnitude from 10-15% to 1-2% on average. For transmission improvements, experimental data revealed an increase in transmission from 25% to 65-70% on average. The additional understanding gained from this work, combined with the improved device performance, will help to advance the level of circuit integration and scalability of VMDs.

134

6. Low Temperature Co-fired Ceramic Packaged MEMS Micro-ion Source for Miniature Mass Spectrometry Applications

In chapter four, the interface between CNT electron emitters and MEMS platforms was investigated and improved. Chapter five utilized this knowledge and improved active circuit elements from the CNT field emission MEMS devices. The device can also be utilized as an ion source in mass spectrometry. This has been demonstrated by generating and collecting ion current on-chip using integrated panel electrodes [147, 150,

151]. While this was a successful proof of concept, it is limited in its application due to large energy and angular dispersion of the ions and the limited ability to direct the ions to a mass analyzer. If CNT FE-MEMS devices are to be realized as practical components in miniaturized mass spectrometer development, these concerns must be addressed.

Thus, this chapter investigates improving the technology through a redesign of device geometry and the introduction of new operating modes to achieve orthogonal ejection of ions relative to the source, thus allowing them to be directed into a mass analyzer.

Additionally, this chapter introduces the development of a novel packaging solution that provides a 3D scaffold for the MEMS micro-ion source, as well as integrated apertures needed for ion focusing and spatial filtering.

Device improvements and redesign were accomplished through extensive parametric characterization, simulation of new design concepts, and failure analysis of existing devices and test structures to determine performance limiting elements. The

135

development involved three key areas: mechanical, electrical, and materials. The mechanical properties investigated structural integrity of the panel electrodes during operation, locking mechanisms and latch configuration, hinging mechanisms, alignment of electrodes, and contamination during materials integration (i.e., during CNT growth).

The electrical studies investigated the current carrying capacity of all structures, operating modes, failure by electrostatic discharge and the electrical overstress susceptibility of the MEMS structures, electric field edge-gradient effects, and current- voltage characteristics. Finally, the materials, or more specifically, the nanomaterials properties that were investigated, focused primarily on the CNT emitters and methods to improve their performance within the device. This included array size and placement, as well as control over the length of the CNT emitters between cathode and grid. The spacing between CNT tip and grid establishes the field for electron emission, therefore

CNT length correlates directly to the potentials required to operate at a specific current.

Failure analysis (FA) was conducted and mitigation techniques were proposed, evaluated, and implemented in new designs.

Novel packaging for the MEMS micro-ion source was realized by utilizing low temperature cofired ceramic (LTCC). Layer by layer fabrication allows for placement of required bonding pads, electrical traces, and connector pads in order to house the source, apertures, and micro-connectors. Design considerations centered on assisting ion current extraction from the micro-ion source, as well as focusing of the ions through a

136

spatial filter based on coded apertures [279, 280]. Gas inlet and exhaust ports were integrated into the LTCC to allow for delivery of analytes to the ion source. Simulations of micro-ion source operation within the package, as well as fabrication techniques and assembly methods are discussed in this chapter.

6.1 Introduction

Miniature mass spectrometers have a plethora of potential applications in environmental monitoring, security and defense, point of care medicine, and space exploration. The ion source, a critical component in any mass spectrometer, has been developed in a variety of classes including electron ionization, chemical ionization, electrospray ionization, field ionization, fast atom bombardment, photoionization, and laser desorption/ionization. The use of microelectromechanical systems (MEMS) components in mass spectrometer miniaturization is particularly attractive due to their small size and scalability [281, 282]. MEMS have been investigated for use in several subcomponents from ion sources [283-288], to mass analyzers [289-295] and detectors

[296-300]. The MEMS fabrication process combines micron scale lithography and thin- film deposition methods to consistently outperform conventional machining of small- scale geometries. MEMS provide a path towards wafer-scale full system development.

Carbon nanotubes (CNTs) have attracted much attention as a high-performance electron field emission source [301-303]. They offer several benefits over traditional thermionic filament emitters in terms of power use, lifetime, and current density [304].

137

Additionally, Low Temperature Cofired Ceramic (LTCC), a technology used to package integrated circuits in multichip ceramic modules for nearly 20 years, has been recognized as one of the most significant recent developments for packaging and integrating sensors and microelectronics [305, 306]. LTCC materials have been used to construct ion mobility spectrometers [307] and arrays of ion trap mass analyzers [308], as well as for packaging solutions for a variety of microelectronic devices [309-311].

Our group has pioneered the development of miniature electron ionization sources and VMDs combining the versatile Polysilicon Multi-User MEMS Processes

(PolyMUMPs) with integrated CNT cold-cathode field emitters [146-150, 198, 273, 274].

To date, ion currents have been measured on-chip using integrated MEMS ion collectors for various analytes over a wide pressure range [147, 149]. However, until now they have been of limited use due to both large energy and angular dispersion and limited ability to direct the ions to a mass analyzer. Fabrication of an ion source that provides better control over electron energies and their corresponding ionization cross-sections, as well as improved focusing properties of electron and ion trajectories, would allow further development and integration of the current MEMS ion source platform in a complete MS instrument. The development of a scalable, non-restrictive packaging solution that could integrate the ion source with the necessary apertures to move ion collection off-chip, would allow for a drop-in source solution for miniaturized

138

instruments. Together, the integration of the microfabricated ion source and package would be of both scientific and technological importance.

To address the limitations of current miniature ion sources, this chapter discusses the development of a Nier type electron ionization source [156, 312] combining a MEMS chip and integrated field emission cathode with a low-temperature cofired ceramic

(LTCC). The LTCC provides a 3D substrate and ion optics for the MEMS device. The new MEMS geometry, field emission cathode, and LTCC scaffold together create a fully packaged miniature ionization source. In the ion source design, the MEMS chip contains the CNT electron source, grid, repeller, and anode, while the LTCC provides a scaffold for making electrical connections to the MEMS chip as well as positioning the extraction aperture, spatial filter, and sample inlet. The ion source focuses ions formed at the

MEMS substrate vertically and uses electrode structures to minimize undesired energy and angular dispersion. In the following sections we detail the design and simulation, fabrication, and experimental characterization of this new miniature ion source.

139

Figure 6.1: Schematic diagram of a Nier-type geometry ion source. 6.2 Design and simulation

6.2.1 Components of a Nier-type Ion Source

A traditional Nier-type ion source (Figure 6.1) is composed of various electrodes including electron beam inlet, an electron source (or cathode and grid), anode, extraction aperture, and repeller arranged in a cubical geometry. The potential on each electrode and geometry of the electrode is chosen to direct ions through the extraction aperture towards the spatial filter with minimal energy and angular dispersion. Ion trajectories along a non-uniform equipotential surface and through a non-uniform equipotential volume results in an undesired spread in energy of the ejected ions. Uniform electric fields in an ion source that introduces minimal deviation from as-formed m/z ions leads to reduced distortions in peak shape and loss of resolution [153, 183]. Therefore, electrode placement and operating potential arrangements are especially important in

140

order to achieve a uniform and parallel field across the ionization volume. For miniaturized sources, where the ions are created in a smaller volume resulting in shorter trajectory path lengths, this is especially important.

Proper placement of extraction apertures and electrical connections are equally important to control electric fields, ion trajectories, and injection and ejection efficiencies.

Several additional apertures can be added after the extraction aperture to collimate the ion beam. A spatial filter is typically placed at the exit of the ion source and entrance of the mass analyzer, and the sample inlet is a small hole in the ionization volume.

Typically the ionization volume is constructed using metal plates for the electrodes separated by insulating spacers. The electrical connections are made by attaching wires to the various electrodes. The conventional method of constructing the ion source and ionization volume described above is adequate for large instruments; however, when miniaturizing, it becomes prohibitively difficult to assemble and electrically connect the various electrodes.

6.2.2 High Performance Ion Source Characteristics

A high performance ion source has two main characteristics. First, the ions are created over a small potential difference. This results in an ion beam with a small energy dispersion. Second, electrodes should be positioned such that the equipotential lines are as parallel to the substrate as possible. This results in minimized angular dispersion.

141

6.2.2.1 Ionization Volume Impact on Resolution

The 1st important characteristic is energy distribution of ions. The goal of the ion source is to create ions that can travel through the mass analyzer and onto a detector.

The mass analyzer provides the resolving power of the instrument; and the resolution depends inversely on the dispersion at the analyzer outlet [312]. Since analyzers separate ions based on mass + charge, a broader range of energies will lead to broader “peaks” in the analyzer, thus a narrow range is better. This is very important to both mass resolution and accuracy. Therefore, reducing the energy and angular dispersion at the source will assist in achieving the best resolution at the mass analyzer output. Energy dispersion is dictated by the properties and efficiency of the ionization volume. Ideal performance, i.e., low energy dispersion, dictates that the ions are created in a volume with a small potential gradient [312], or zero electric field ideally as illustrated in the schematic of Figure 6.2a. The volume between the grid electrodes and floor, which constitutes the ionization region, must remain small such that the ions need only to cover a very short distance. The small potential gradient minimizes energy dispersion. If the potential gradient of the ionization region is large, the ions are accelerated to varying kinetic energies and therefore will not follow the same trajectories through the magnetic fields [312]. By utilizing microstructures, the ionization volume can be designed such that the potential gradient is very small.

142

6.2.2.2 Parallel Equipotential Surfaces

The 2nd important characteristic the angular dispersion, or the diverging velocity properties of the ions. Angular dispersion is dictated by the electric field and the trajectory the ions take through that field. For the analyzer, all ions should be traveling in the same direction, or collimated, to improve resolution and accuracy as seen in the schematic of Figure 6.2b. The equipotential field lines are shaped by the electrodes and the extraction apertures in the source. The electric fields will guide the ions throughout the volume. To minimize angular dispersion, ions should move along a path of parallel equipotential field lines (Figure 6.2b), in respect to the substrate, as they move towards the mass analyzer. Any divergence imparted onto them from non-parallel field lines could increase once they are inside the mass analyzer, resulting in large angular dispersion and lower resolution. In order to improve collimation of the ion beam, and reduce angular dispersion, the electric field can be engineered in such a way to focus the ions towards the apertures.

143

6.2.3 Proposed Solution

Each component of the packaged miniature ion source was evaluated and designed to minimize dispersion by maintaining a small ionization volume and parallel equipotential surfaces, with respect to the substrate, throughout the device. The structures considered were as follows: cathodes, grids, ion repeller (i.e., floor), latches, extraction aperture, spatial filter, sample inlet, and electrical connections/traces.

6.2.3.1 MEMS

Cathode and Grid. The cathode and grid are the primary contributors to the electric fields that enable field emission from the CNT emitters. The absence of thermal

144

load present in a thermionic emission source means that the grid can be placed close to the cathode, reducing the voltage required to achieve the electric fields necessary for electron emission. Previously, we have investigated ion source performance with a triode arrangement consisting of a cathode, grid, and anode. For this design, the anode was eliminated and an additional cathode/grid pair was placed in its location. Starting from the left side of the chip, the electrodes moving left to right now consist of a cathode, grid, floor, another grid, and the final cathode. Having two symmetric cathode/grid pairs as seen from the center of the device looking out, creates relatively parallel and equal equipotential field lines across the device. However, in standard operating mode, the cathode is held at a negative potential with respect to the grid. This creates a strong negative field on the order of tens of volts that distorts the parallel equipotential field lines in the ionization region. Therefore, the grids were designed to be taller than the cathode to shield the negative fields from penetrating the ionization volume.

Repeller (i.e. Floor). In traditional Nier-type sources, a repeller electrode is fixed within the ionization region in order to direct ions and move them towards the mass filter [153]. In order to accomplish similar goals, we utilized the MEMS process to fabricate a floor electrode on the substrate. This floor electrode, located beneath the ionization region, completes the cuboidal ionization volume and, along with the extraction aperture, directs ions orthogonally away from the substrate in the z-direction.

145

This is similar to previous work with the MEMS micro-ion source where ions were pulled towards side panels fabricated on chip where ion current was collected and measured [149, 151]. However, the change here is the direction (off-chip instead of remaining on-chip) and the distance the ions travel (hundreds of microns instead of

10’s). The H-shape of the repeller electrode is important to maintain field uniformity for ion extraction. Without the proposed shape, the potentials from the cathodes, grids, and supporting structures dominate the far field boundary conditions and negatively impact the ion trajectories. The H-shape minimizes the impact of other electrodes that would otherwise reduce performance. Together with both grids, a symmetric cuboid-like ionization space is created. The vertical extraction approach allows for placement of required apertures parallel, and above, the substrate; thus creating a vertically stacked device above the symmetric ionization region.

6.2.3.2 LTCC

Utilizing a multilayer LTCC package option allows for placement of the MEMS chip in the desired location, as well as the placement of extraction and focusing apertures at controlled distances from the substrate to enhance uniformity in ion extraction, creating an ideal ionization volume. Several sets of LTCC layers containing cavities, vias, and electrodes for electrical connections were used. The combined layers form a 3D scaffold for assembly of the electron source, ionization volume, sample inlet, and ion optics. 3D CAD design software SOLIDWORKS® was used to design the LTCC

146

parts for the ion source package. LTCC starts as a semi-flexible substrate called tape.

Vias can be punched, conductive material screen printed, and patterns can be cut.

Various numbers of layers can be stacked and cavities created where layers have different cut patterns. Cavities were cut in some of the layers to hold the electron source, ionization volume, collimation optics, spatial filter, and sample inlet. Electrodes were screen printed using gold, and vias were punched and filled to enable electrical connection to the electron source, ionization volume, collimation optics, and spatial filter. Fabrication and firing of the LTCC layers utilizing the ion source package designs were completed by DuPont USA and OMEGA Micro Tech. The extraction apertures and spatial filter (coded apertures) that were inserted and electrically connected to the LTCC were machined by National Aperture, Inc.

Extraction Aperture. In order to assist the ion movement into the mass filter, an extraction aperture was also considered. Using a negative potential on an extraction aperture allows positive ions to gain kinetic energy and follow trajectories off-chip in the z-direction. If a large hole was used as an extraction aperture, the parallel equipotential field lines in the ionization region would be distorted and would increase dispersion of the ions. To minimize increasing the energy or angular dispersion of the ions, the extraction aperture was designed with large open area, using small bar widths at a large pitch. This resulted in flat field lines around the aperture.

147

Spatial Filter. In order to allow miniaturization without a commensurate decrease in performance, aperture coding was integrated into the design of the ion source. Aperture coding breaks the historical trade-off between throughput and resolution and enable smaller volumes of miniature instruments to maintain the throughput of a much larger instruments. It has been described for mass spectrometry in detail previously [280]. Throughput gains of more than 10x with no loss in resolution over conventional slits have been reported in magnetic sector mass spectrometry [279].

Sample Inlet. An important component in a mass spectrometer instrument is the inlet system which allows the analyte of interest to enter the ionization source. The inlet system can be as simple as a port through which the sample is injected or drawn due to a pressure differential. A port was integrated into the LTCC by patterning cavities in each of the corresponding layers. Pumping speed and range was considered when designing an appropriate port to facilitate proper delivery of the sample to the ionization source.

Electrical Connections. The connections to the ion source, and its corresponding electrode panels, is an important part of the system. They are integrated onto the LTCC using screen printed gold and wire bonds were used to connect from the LTCC to the

MEMS. All electrical traces contribute to the electric field gradients in the system and must be considered in order to avoid adverse effects on the ion trajectory. Because the desired path of the ions is in the vertical z-direction, electrical connections were also

148

designed to be equal on all sides of the device, thereby minimizing distortions to the parallel equipotential field lines. Connections from wires to the electrode traces on the substrate was done through wire bonding, which also has an effect on the fields as the wire bonds are large relative to the other components on the micro-ion source.

Therefore, designs extended the metal traces far away from the ionization region. This allowed bonding to be done far enough away so as to have no observable effect on the ion trajectories.

6.3 Design Details

6.3.1 MEMS

The MEMS micro-ion source utilized in this work is derived from previously reported sources [109, 146, 147, 149, 151, 273, 274]; therefore, the fabrication process is similar. Briefly, the devices were fabricated with a well-known polysilicon MEMS micromachining process (PolyMUMPS®, MEMSCAP, Inc.). Iron catalyst for CNT growth was selectively deposited on the cathode substrate using integrated MEMS masks, and CNT emitters were then grown on the substrate in a 915 MHz microwave plasma enhanced chemical vapor deposition. The devices were assembled by rotating and locking the electrodes into a vertical position. The charged particle simulation software SIMION® 8.2 was used for modeling, evaluating, and redesigning the micro- ion source. Design considerations were separated by component. A 2D schematic of the

149

redesigned MEMS micro-ion source and legend providing component identifiers is given in Figure 6.4a.

6.3.1.1 Observed Failure Mechanisms and Mitigation Techniques

Micro-ion sources were originally designed with MEMS springs that served two purposes: electrical contact (i.e. current carrier) from bond pad to electrode panel, and restoring force for vertical alignment of panel. A common failure point observed was overheating and damage to the spring or spring contacts due to overcurrent as can be seen in Figure 6.3a. This occurred as a result of the joule heating of the polysilicon springs, which is resistive in comparison to a metal. This failure mechanism was observed in two situations: 1) devices were operating around 150 µA according to device statistics, and 2) a temporary direct short was made between the cathode and grid or grid and anode electrodes. The latter is caused by high fields between the cathode and grid, resulting in the snapping together of panels in Figure 6.3b.

150

Direct shorts were also observed due to CNT emitters that extended across the extraction region between the cathode and grid Figure 6.3c. High fields, as well as CNT damaging effects discussed earlier, can result in loss of adhesion between the emitter and MEMS cathode. In the event of panel overextension, such as the case with panel snapping, structural integrity of the latches would fail as is seen in Figure 6.3d. Often temporary voltage spikes occurred, as is the case with current controlled voltage lifetime testing, which resulted in momentary snapping of the panels. The period in which panels were touching in this case was too short for overcurrent and burning of the spring to occur, however, latch failure resulted in panels falling and rendering the device destroyed. 151

It has been observed [313] that grain boundary size and dopant atoms effects phonon scattering and therefore thermal conductivity. In heavily doped thin films on the order a few microns, which is the case with our devices 2 µm MEMS polysilicon, smaller grain sizes are present due to shortened annealing times for dopant activation

[314]. These grain sizes, along with increased phonon scattering sites due to phosphorus dopants, lowers the thermal conductivity of the polysilicon leading to variability and poor tolerance to transient overloading. Furthermore, it is well known that contact resistance between two MEMS layers is large due to band alignment and mobility change at the junction. Calculated values from run data provided by MEMSCAP foundry is 3060 Ω/μ2 for poly0-poly1, 3150 Ω/μ2 for poly0-poly2, and 1530 Ω/μ2 for poly1-poly2. We believe this explains the structural failures. Using sheet resistance and additional run data attained from MEMSCAP, and 1st principle calculations based on standard resistivity equations, structures were redesigned to larger dimensions and have been tested to currents as high as 250 µA without failure. Latches were also modified to improve mechanical stability and serve as current carrying structures in order to remove the springs from the devices all together. The curved latches act as a bent spring with a specific spring tension. Primary forces considered in the redesign include the spring tension, electrostatics, and friction as gravity and weight are not substantial. The spring tension was increased by making the legs thicker and by doubling the thickness by adding a second poly layer through the use of poly1-poly2 via

152

etching layers. Latch legs were designed to be effective current carrying conductors and an additional bond pad connection path was added thus supplying two electrical contacts to each electrode panel. This divides the current between structures and provides a backup conductor in the event the spring fails. For structural redesign, theoretical calculations were conducted to approximate specifications. Then, a series of simple test structures were designed, fabricated, and tested before being permanently included in subsequent runs.

A failure mechanism related to the device as a whole was observed and was associated to device shorting due to the existence of a thin conductive film on the surface. It was detected using I-V characterization combined with SEM analysis and 4- point probe testing. Devices that visually appeared ok would occasionally experience shorting between electrodes. SEM analysis would confirm that panels were not in contact, and often reveal the existence of a thin film on device surfaces. 4-point probe would confirm the existence of a conducting film that was not desired. It was determined the film was of organic nature and most certainly a thin carbon film that was deposited during growth on the device surface. A process using a matrix plasma asher and a short duration, low power O2 plasma was developed to remove the film, and restore open circuit function, while not harming the CNT field emitters.

153

6.3.1.2 Cathode and Grid

The cathode panels were fabricated with a height of 55 microns and a width of

460 microns. The CNT emitters were grown in a 2x16 array of 5x5 micron bundles as shown in Figure 6.4b. The width of the CNT regions were 360 microns. The CNT arrays were patterned this way to create a flat and wide electron beam for ion generation. The small variance in height is such that ions are created on approximately the same plane above and parallel to the substrate. The grids were placed 40 microns away from the cathodes and oriented such that the CNT emitter bundles were center aligned with the grid holes as seen in Figure 6.4c. The designed height of the grid was 90 microns and the width was 735 microns. The symmetric cathode and grid pair were spaced 300 microns apart and parallel to each other to create, in conjunction with the floor, a cuboid ionization region. The cathodes have a thickness (polycrystalline silicon film thickness) of 2 microns. The grids also have a thickness of 2 microns.

154

6.3.1.3 Repeller (i.e., Floor)

The repeller (floor) electrode was designed to be 160 microns wide, 800 microns in length, and spaced 60 microns away from each grid to allow for adequate spacing and avoid any shorts between electrodes. The geometry and placement of the floor electrode is an important consideration in order to best shape the equipotential surfaces seen by the ions along their path from the substrate to the extraction aperture. The floor is held at a similar potential as the grids to create the cuboid ionization volume at the center of the device. Because the floor must be biased, it needs an electrical trace connecting it to a bond pad. However, extending a trace down one side only would distort the symmetry

155

in the device. So the floor was extended down each side of the chip to maintain symmetry and balance the equipotential surfaces; shown in Figure 6.4a. The polycrystalline silicon floor has a thickness of 1.5 microns, slightly thinner than the panel electrodes as it is fabricated in a separate processing step in the MEMS fabrication.

6.3.1.4 Wire Bonding Pads

The device requires biasing to operate, and therefore gold wire bonding is required for all MEMS micro-ion sources. Because each electrode is independent from one another, multiple wire bonding pads are included on chip. However, due to the micron scale of the device, the thin bond wires still have an impact on the fields surrounding the device due to their comparatively large size. Therefore, the pads were spaced out far from the device using electrical traces as shown in Figure 6.4b. All traces were fabricated directly in the MEMS fabrication with no additional processing steps.

The wire bonding electrodes were nominally1 1 mm in length and 150 microns wide.

Wire bonds were attached at the outermost edge of the electrodes to minimize their impact on the equipotential field lines.

6.3.1.5 Latches

The latches, also fabricated using MEMS poly-Si layers, have two roles in the ion source. The first is to supply a mechanical locking mechanism to hold the panel vertical.

1 nominally is used to indicate that the dimensions are the designs dimensions of the device and do not account for minor processing errors in the dimensions. 156

The second is to bias the panels which is accomplished at the electrical contact point as seen in Figure 6.4d. The latch arms serve as the current carriers and therefore must withstand the total electrode current to avoid failure and breakage due to overcurrent.

The latches geometrical dimensions were designed to accommodate well over the maximum current requirements of the device (approx. 150 micro amps). Test structures were fabricated and used to validate the current carrying ability of the structures.

6.3.2 LTCC

Design of the LTCC package was accomplished by evaluating each of the integrated components individually. Each component was simulated at multiple stages in the design. Two iterations (i.e., generations) of the LTCC designs were fabricated and evaluated. Design changes were made after experimental experience was gained on the first generation.

6.3.2.1 LTCC Layer Specifics

Figure 6.5a shows an exploded view of the 1st generation of the LTCC scaffold.

The “MEMS device” reference indicates the location where the MEMS chip, and the heart of the ionization volume, is bonded into the LTCC using epoxy. The assembly is comprised of 4 LTCC layers; each layer is comprised of 5 pieces of LTCC tape, or sublayers.

157

Figure 6.5: (a) Exported CAD image showing exploded view of the LTCC layers with labels indicating various components, (b) cutaway view of the assembled LTCC package.

The first layer includes electrical connections and vias to the additional layers as well as a cavity for the MEMS device and a channel for sample outlet. The second layer includes a cavity for the ionization volume as well as a larger cavity which exposes screen printed electrical traces where electrical connection can be made to the ionization volume and electron source. The third layer includes a cavity to enclose the electrical connections in the ionization volume and electron source. The third layer also contain a smaller cavity as well as a screen printed trace for electrical connection to an extraction aperture. In this design, the extraction aperture is a single layer of metallized LTCC with a small hole cut in it. The area around and inside the hole is covered with screen printed conductor. The third layer also includes a cavity for the sample inlet. The fourth layer contains the cavity where the spatial filter (i.e., coded aperture) is placed, along with electrical traces to bias it. In this 1st generation, the spatial filter is a metal plate with a series of holes,

158

placed over a cavity in the LTCC slightly smaller than the metal plate. Figure 6.5b (right) shows a cutaway view of the fully assembled scaffold with device inside. Not shown in the CAD image is gold that was deposited on as much of the surface as possible to shield the mass analyzer from the electric fields inside the packaged device. Therefore, some gold was applied post fabrication using manual masking and electron beam deposition.

Figure 6.6: Exploded view of the 2nd generation of the LTCC scaffold with labels

Figure 6.6 shows an exploded view of the 2nd design iteration (i.e., 2nd generation) of the LTCC. Image (a) gives a top down view, where image (b) provides a bottom up view to show the clearance cavities, vias, and micro-connectors that bias the electrical traces. Layers are arranged similar to the first generation, but cavities, electrical traces, and vias are modified to accommodate insertion of large MEMS chips, and to allow for better centering of the device under the apertures. The layers have been reduced to three total, with the 1st layer containing 10 sublayers and the 2nd and 3rd layer containing only

5 sub layers. Additionally, traces and contact points between layers were reduced 159

significantly to improve assembly of the multi-layer stacks. In the 1st generation, more than 12 pads needed to be aligned and secured with conductive epoxy in order to have a functional scaffold. In this new embodiment, only 6 were needed, greatly improving the yield during assembly. Additionally, micro-connectors were added to provide simple biasing to the traces without the need to bond to a test PCB. In this generation, the extraction aperture is a metal sheet with a hole replacing the LTCC with a hole in the first design. Figure 6.7 (a) shows a magnified image of layer one with the newest design iteration of the MEMS chip secured within. Additionally, a cutaway view of the complete structure is shown indicating the location of the extraction and coded aperture.

SEM images of the apertures and the MEMS ion source are included.

Figure 6.7: (a) Magnified image of the 1st layer with MEMS micro-ion source secured within the cavity, SEM images of source given; (b) cutaway view of the complete scaffold with locations of apertures shown, SEM images of extraction and coded aperture shown.

6.3.2.2 Assembly Details

The vias of LTCC layers 1 and 2 were fitted with conductive silver epoxy via a screening process and joined manually. Precision alignment was performed using a Suss 160

Microtech FC150 Flip Chip Bonder in order to properly align the top and bottom vias.

An assembled MEMS chip (details provided in section 6.2) was secured into the exposed cavity using Loctite Black Max 380 Cyanoacrylate Adhesive. The MEMS chip was wire bonded to the bond pads exposed on layer 2. Each bonding pad on the MEMS chip was routed to a separate bonding trace on the layer. After initial device testing to verify proper operation, the 3rd LTCC layer, with extraction aperture secured within, was attached onto the 2nd layer. After initial electrical testing (details provided in section 6.4), the 4th layer, with coded aperture secured within, was attached onto the 3rd layer.

6.3.2.3 First Generation

Figure 6.8 shows several optical images of the 1st generation packaged miniature ion source. Image (a) shows the 1st generation of the MEMS chip secured into the 1st layer. Wire bonds are used to secure the device to the 2nd layer of the LTCC. Care is taken to ensure the wire bonds do not touch the edge of the substrate, and also to ensure they do not extend too high and make contact with the 3rd layer when secured on top.

Images (b) and (c) show the packaged ion source mounted on a PCB test board with the exit outlet of the ions placed directly over a hole in the PCB. Image (d) shows the mounted device fitted with a copper plate over the hole on the backside of the PCB to act as a collector for ions during testing. A wiring harness was used to connect the board to the electrical feedthroughs of the vacuum chamber. Image (e) is the vacuum testbed used for device testing.

161

6.3.2.4 Second Generation

Figure 6.9 gives an image of the 2nd generation of the packaged source in two different stages. Image (a), “before capping,” shows a MEMS chip secured within the cavity and wire bonded to the LTCC. This is the stage at which initial electrical characterization is performed by plugging the micro-connectors into the harness of the vacuum test chamber. Image (b), “after capping,” shows the state of the device once the final two layers have been secured on top. This is the state at which ion testing is performed; ion current can be measured at the outlet of the top aperture (i.e., coded aperture). Additional details on electrical characterization are given in section 6.4.

162

Figure 6.9: Optical image of the 2nd generation of the LTCC scaffold prior to the 1st and 2nd stage of electrical characterization; a) before capping; b) after capping.

The 2nd generation of the LTCC and MEMS packaged miniature ion source was able to maintain a very small potential gradient in the ionization region, as well as maintain relatively parallel equipotential field lines in the region of the ion trajectories.

The slotted extraction aperture also assists in maintaining the flat field lines. This results in improved energy and angular dispersion. This can be seen in the simulation results shown in Figure 6.10. The images give two different vantage points of the device. The left column is the view from the side showing both cathode/grid structures on the left and right side. The right column shows a cross-section view down the middle of the device from the center of the floor towards a grid. These two perspectives provide good visual understanding of the electric field lines. Images (a) and (b) show the device with no field lines plotted. Images (c) and (d) show the field lines plotted with no ions simulated. Images (e) and (f) show the plotted field lines with simulated ions indicating their orthogonal path from the ionization region to the extraction aperture. Images (g) gives a top down view of the device with plotted field lines showing the cubic ionization

163

volume. Finally, image (h) shows the distribution of simulated ions passing through the spatial filter. The red and green regions indicate more ions than the purple.

164

6.4 Characterization of Packaged Micro-ion Source

Device testing was conducted in a custom vacuum chamber with the device either on a PCB test board or plugged directly into a harness. Bias was applied through twelve electrical feedthroughs which allowed for a variety of electrode biasing options.

All measurements were performed at room temperature with a base pressure of 7 x 10-7 torr. Custom LabVIEW software controlling a multichannel semiconductor analyzer

(Keithley 4200-SCS), and SourceMeter instruments (Keithley 2410s) was used to electrically characterize the devices. In order to perform angular and energy dispersion measurements, a 1704 channel capacitive transimpedance amplifier array detector from the University of Arizona [315] was utilized. The detector included custom CMOS integrated circuits with a 3.3 V, 0.35 micron process. Each detector finger has a separate preamplifier, logic, and sample and hold on the custom IC. Characterization of the

MEMS micron-ion source was performed in two stages. The first stage involved characterization of the electrical properties of the device prior to complete packaging.

The second stage involved characterization of the fully packaged device.

6.4.1 Electrical Characterization

6.4.1.1 Electrons

Both stages of electrical characterization were similar but the addition of ion collection was added during the second stage. In the 1st generation of the packaged source, the LTCC was wire bonded to a test board to provide electrical bias to all

165

contained components. In the 2nd generation, micro-connectors secured to the LTCC were plugged into a wiring harness connected to the vacuum chamber testbed. Electrical characterization was performed in a similar matter to what has been previously reported

[149-151, 273]. Briefly, IV characterization was performed between each cathode and grid to determine turn-on voltage and threshold voltage. These voltage settings were then used once the device was completely sealed in a package. Additionally, triode characterization curves were taken by operating the opposite grid as an anode. This provided information into the ability of the electrons to transmit across the ionization region where ions are created. The Fowler-Nordheim model was fitted to the I-V data to verify field emission was occurring. Figure 6.11 shows the results from a device that has undergone the three tests performed during electrical characterization.

A burn-in process was developed to improve the performance of devices before using them in application. Constant current emission from the CNTs reduced amorphous carbon and other absorbates from CNT tips through joule heating. The 166

developed current controlled voltage lifetime (i.e. constant emission current) LabVIEW program is used to drive new emitters at a constant emission of 1 µA for one hour. It was observed that devices subjected to burn-in operated with better performance than those without. An example of an I-V curve taken from a device is shown in Figure 6.12a.

That same device was then immediately subjected to the developed burn-in process

Figure 6.12b. Finally, another I-V curve was taken and is shown in Figure 6.12c. The results demonstrate an improvement in turn-on and threshold voltage. The burn-in shows the reduction in required voltage to maintain the desired current after one hour.

Figure 6.12: These three plots illustrate the process of burn-in on a device; (a) An initial I-V curve is taken showing a VTO of approximately 160 V, (b) a one hour burn- in is conducted where the VTO is reduced, (c) improved I-V curve after burn-in.

After characterization, IV sweeps were conducted between each set of electrodes in the system to verify no shorting of electrodes was present, i.e., cathode-grid, grid- floor, cathode-floor, grid(1)-grid(2), cathode-extraction aperture, grid-extraction aperture, and all electrodes to ground. I-V analysis of devices at various stages of build confirmed performance. Initial I-V curves and burn-in tests were used to condition the samples, resulting in I-V performance with improved stability and control. Tests of

167

various magnitudes of current emission were performed to establish threshold voltages for various emission current magnitudes. Figure 6.13 shows two I-V characterization curves taken for a packaged device (i.e., after capping). Plot (a) gives the initial I-V test performed to determine the turn-on voltage (VTO) of the device. This is why the device was only biased until emission was achieved and the y-axis indicates nano-amps of current. For this device, the VTO was approximately 85 V. Plot (b) reveals the threshold voltage I-V analysis performed after the VTO sweep. The VTH for this device was approximately 130 V to emit 1 micro-amp of current.

Figure 6.13: Electrical characterization plots of a packaged MEM micro-ion source; both turn-on and threshold voltage I-V analysis plots are shown.

6.4.1.2 Ions

Ion current measurements were conducted at the outlet of the coded aperture.

Pressure was increased in the chamber until ion current was detected on the collector.

Ion extraction transient experiments were performed by cycling potentials on the electrodes that govern ion extraction (e.g. floor and extraction apertures). By cycling

168

these control voltages OFF and ON, we can see the ion current respond accordingly, as shown in Figure 6.14. With increased CNT emission current and increased sample inlet pressure, ion current was increased.

6.4.2 Dispersion Measurements

6.4.2.1 Simulation

Simulations used to design and evaluate performance of the packaged source were also used to evaluate ion beam collimation and evaluate the energy and angular dispersion. An example of a SIMION simulation on a packaged source can be seen in

Figure 6.15. Equipotential lines, electrons, and ions were plotted to give insight into the trajectories and the field lines. Energy and angular dispersion values were then

169

extracted from the simulations. Average energy and angular dispersion were 14 +/- 4 eV and 13 +/- 1 degree, respectively.

6.4.2.2 Experimental

After packaging, the dispersion of the packaged miniature ion source was investigated experimentally. Experimental measurements of energy dispersion were conducted through stopping potential experiments. A retarding grid was placed between the ion exit point of the packaged device and the detector. Ion current from the device was established and measured on the detector. Early stopping potential experiments presented difficulties in obtaining accurate measurements due to an

170

abundance of electron current interfering with ion measurements. The experimental setup was modified to include a new aluminum mount for the packaged device and retarding grid. The mount included magnets on each side of the device to create a magnetic field to collimate the electrons and prevent their escape from the ionization regions, thus eliminating their interference with the ion current. In testing, the retarding grid potential was swept from 0 V up to 40 V to successfully stop the ions from reaching the detector. Approximately 10 sets of detector data per retarding grid step size were averaged to establish a single data point on the stopping potential plot shown in Figure

6.16. A derivative of the data was then taken and the full width half max was calculated to reveal the energy of the ions as well as the energy dispersion. Average energy of the ions was 22 eV +/- 6 eV. For angular dispersion measurements, the packaged device secured in the aluminum mount was oriented so that the ion exit point (i.e., the coded aperture) was pointed directly at the detector as seen in Figure 6.16. The device was operated at its threshold voltage to establish ion current. The pressure inside the chamber was held at 5x10-5 torr using lab air. The ion current was collected and its peak width was measured at the detector as seen in the bottom left plot in Figure 6.16. The ions strike the detector at a specific pixel range, centered at the 1125th pixel, providing the exit angle of the ions. This angle is calculated using the peak data and is recorded for each sample. The average angular dispersion from the micro-ion source was approximately 13° which matches values shown in simulation (Figure 6.15).

171

6.5 Conclusion

A packaged miniature ion source was designed using a MEMS platform with integrated CNT field emission cathodes and a LTCC 3D scaffold with integrated ion optics. The packaged ion source was designed to operate similar to a Nier-type ion source. The MEMS chip contained a CNT electron source, grid, repeller, and anode, while the LTCC provided the scaffold for making electrical connections to the MEMS chip, as well as positioning the extraction aperture, spatial filter, and sample inlet. Ion current was generated from electron ionization that occurred in the ionization region located close to the MEMS substrate. The design permitted generated ions to be extracted vertically off-chip. The ion source focused the ions and used electrode structures to minimize undesired energy and angular dispersion. The packaged miniature ion source was evaluated using both simulation and experimental results.

Simulation was used to model the energy and angular dispersion of the packaged device. Experimentally, the device was characterized electrically and angular dispersion of the source was determined to be 13 °. Energy dispersion was not successfully measured experimentally and will be the focus of future publication.

173

7. Summary and Recommendations for Future Work

In this research, novel field emission vacuum electronic and microelectronic devices using integrated carbon nanotube field emitters for a variety of advanced applications have been developed. Additionally, details on their fabrication, characterization, and application specifics have been discussed. The summary and conclusions per chapter will be presented in order of their appearance in this work.

7.1 CIMS CNT Electron Source

A fully packaged CNT field emission device was developed and employed in a prototype commercial mass spectrometer as an electron source to obtain chemical ionization mass spectra in both positive and negative modes. The electron source demonstrated performance benefits over standard filament thermionic emitter sources, establishing itself as a candidate for CI and EI mass spectrometry applications. A summary of key conclusions is as follows:

 Substrate and growth procedures were developed to improve device performance;

specifically device assembly techniques (important for scaling) and lifetime at

operating pressures (important for commercial viability when utilized in an

instrument) were addressed

o Substrate partial pre-dicing was developed to target final CNT emitter die size to

reduce CNT emitter damage when assembling devices. Undesired film liftoff

was a primary reason for device shorting and failure.

174

o Thin film deposition of polycrystalline on top of the low resistive single

crystalline silicon substrate was investigated to reduce contact resistance

between the device scaffold bias electrode and the base of the CNT emitter.

Utilizing a conductive silver epoxy spread to the poly-Si region during assembly

reduced turn-on voltages and improved lifetime

 Constant-on lifetimes of 320 hours was demonstrated for the CNT electron sources

during ex-situ characterization. If operated at a 25% on-time cycle as expected

during real-time security checkpoint deployment, this equates to approximately 4

months of operation without replacement required.

 Cycling benefits (ability to turn the source on and off with succession) over standard

filament thermionic sources were demonstrated. This equates to reduced power

consumption and improved lifetime as cycling does not degrade performance of

cold cathode emitters, as is the case with thermionic.

 Device functionality was demonstrated and compared side by side to thermionic

emitters in the same mass spectrometer instrument. No loss in performance or

sensitivity was seen, and collected EI and CI spectra closely matched those collected

with the thermionic source, as well as those in literature.

7.1.1 Recommendations for Further Investigation

The following recommendations to enhance the electron source have been developed from this research:

175

 By design, device extraction grids were positioned at a relatively large distance away

from the CNT emitters, approximately 200 µm. This was a parameter limited by the

electrode stand-off used to support and bias the grid. As a result of the distance, the

voltage requirements to achieve desired current values were relatively large. The

standard filament thermionic emitters were typically operated at 200 µA. Due to

voltage requirements from the CNT device, 50 µA was typically used. Whereas low

current is beneficial in portable instruments, with improvements to the emission

properties of the CNTs, the same low current could be achieved at lower voltages.

This would result in lower power requirements.

 The addition of focusing lenses and a small magnetic field integrated on the device

could assist in columniation of the electron beam. This would improve the total flux

into the ionization region.

 A proposed method for achieving both of the previously mentioned

recommendations is an embodiment that incorporates the emitters into a recessed

cavity in a silicon substrate which can then be directly secured to a PCB scaffold for

integration into the instrument. The extraction grids and option focusing grids could

then be deposited lithographically with oxide buffer layers to electrically isolate the

grids and cathode. Biasing could be achieved similarly to the design discussed in the

work, but turn-on voltages would reduce as a function of the separation, which

would be small given the distance is a factor of the insulating oxide layer.

176

Investigations into this method were initiated, and minor electrical characterization was performed for proof of concept. Cavities were etched using photolithography methods and deep reactive ion etching (DRIE). Catalyst was deposited into the cavities and CNT growth was controlled such that the emitters remained well underneath the surface. Further investigation is needed to demonstrate viability as an instrument source. Figure 7.1 shows several versions of the preliminary device,

SEMs of the devices with magnified and cross section views of the emitters in the cavity, and electrical characterization showing performance. Preliminary results show capabilities of the device to reach emission currents nearly 8-10 times greater than previous designs at similar voltages. Placement of one chip directly on the previously developed PCB platform could result in a better performing electron source.

177

The growth of CNTs on polycrystalline silicon substrates was studied to improve the design of CNT cold-cathode field emission sources when integrated with microelectromechanical systems, specifically in vacuum microelectronic devices applications. The kinetics of CNT growth on polysilicon were compared to growth on Si

(100) substrates using the model of Deal and Grove to understand the activation energies and mechanisms for base-mediated growth using microwave PECVD.

Diffusivity values for growth on poly-Si vs Si were also investigated. Finally, methods to

178

improve adhesion of CNT films during operation as field emission sources were investigated. Primary conclusions include:

 Activation energies for CNT growth in the temperature range of 825 – 875 °C on

polysilicon and on Si (100) substrates were discovered to be 1.61 and 1.54 eV for the

nucleation phase of growth, respectively, and 1.90 and 3.69 eV for the diffusion

limited phase on Si (100) and polysilicon, respectively.

 Diffusivity values for growth on polysilicon were notably lower with a value of 7.9 x

10-5 cm2/s at 850 °C than the corresponding values for growth on Si (100) at a value of

7.9 x 10-5 cm2/s. The growth process became diffusion-limited earlier in the case of

poly-Si. Evidence favors a surface diffusion growth mechanism involving diffusion

of carbon precursor species along the length of the CNT forest to the catalyst at the

base.

 Methods to integrate thin film interlayers between the CNTs and the poly-Si

substrate to improve adhesion were investigated. Addition of a thin film of titanium

equivalent to the catalyst thickness (i.e., 5 nm in this work) showed improved

adhesion benefits of approximately 2.5 times when compared to the control samples

without adhesion interlayers.

7.2.1 Recommendations for Further Investigation

The following recommendations to enhance the CNT growth have been developed from this research:

179

 Results in this work revealed improvement in adhesion through the use of titanium

interlayers, however, small effects in CNT growth uniformity and electrical

properties (i.e., field emission performance) were also seen. Future work would

investigate new materials that could be deposited to serve as an adhesion layer, but

also improved electrical properties by reducing contact resistance between the CNT

emitter and the poly-Si substrate.

 Additionally, depositing materials that would double as a diffusion barrier during

growth would provide improved control of the CNTs. The length of the CNT

emitters, an important parameter when growing emitters between electrodes of a

fixed spacing, is proportional to the catalyst during growth. Any loss of catalyst into

the substrate during growth reduces the overall control of desired parameters.

 The use of intermetallics, such as TiN (titanium nitride), have been reported recently

as ideal materials for both adhesion and diffusion barriers. Additionally, a variety of

intermetallics have allowed for investigation of CNT growth on conductive metal

substrates, such as copper, further reducing the contact resistance between emitter

and substrate.

7.3 Improved Performance of FE-VMDs for IC

Particle trajectory simulation and experimental characterization were used to improve device performance of a MEMS CNT-FE VMD as an active circuit element. The performance enhancements serve as important enablers for larger scale integration and

180

for the realization of complex vacuum microelectronic circuits. Solutions for the operation of multiple devices in close proximity and for enhancing transmission (i.e., reducing grid loss).

 It was determined that proximity effects between closely packed VMDs, specifically

crosstalk current active devices, cannot be neglected in the operation of the devices

as high performance integrated circuits.

 Isolation structures were developed from integrated MEMS floor electrodes. Their

incorporation into revised design geometries reduced crosstalk between neighboring

devices from 15% on average, to nearly zero. This mitigation technique allows for

dense placement of VMD active circuit elements on a common substrate, a necessity

for integrated vacuum circuits.

 Poor transmission (i.e., large electron loss at the grid) was improved.

o Innovative geometries and a new operational modes reduced grid loss by nearly

3x, thereby improving transmission of the emitted cathode current to the anode

from 25% from previous designs to 70% on average.

7.3.1 Recommendations for Further Investigation

The following recommendations to continue development of the CNT-FE VMD’s have been developed from this research:

 The next logical step for this work is the integration of multiple active circuit

elements (i.e., CNT FE-VMDs). The knowledge gained in this work towards densely

181

packing devices can be used to explore more complex circuits (e.g. oscillators, logic circuits) o Preliminary work towards development of lateral field emission logic devices

was explored via the design and fabrication of chips including four compact

CNT field emission vacuum microelectronic devices with integrated shielding

electrodes. This all-inclusive device, consisting of four sub-devices, can be wired

on or off chip to explore OR, XOR, and NAND gate logic. Figure 7.2a and Figure

7.2b provide a schematic diagram and microscope image of the multi logic

device in its planar state (not assembled). Image ‘a’ gives an example of a logic

OR configuration if both anodes are tied together. The device is equipped with a

multitude of bonding pads which will allow electrodes to be electrically

connected together in the desired state. Figure 7.2c shows a fabricated oscillator

circuit comprised of four CNT-FE VMD active circuit elements. Characterization

has not yet been investigated.

182

7.4 LTCC Packaged MEMS Micro-ion Source

Microelectromechanical systems, carbon nanotube field emitters, and low temperature cofired ceramic were utilized to develop a Nier type electron ionization source for use in miniature mass spectrometer applications. MEMS fabrication was used to outperform conventional machining of small-scale geometries and enable wafer-scale full system development. Integrated CNT field emitters were utilized to achieve a high- performance and reliable electron source with several benefits over traditional filament thermionic emitters in terms of power use, lifetime, and current density. LTCC was used as a 3D scaffold to integrate the MEMS device, integrated ion optics (i.e., extraction

183

apertures), and spatial filters (i.e., coded apertures). Specific work summaries and conclusions include:

 New micro-ion source designs were envisioned to produce a complete symmetric

Nier type EI source. Symmetry was achieved in both geometry and equipotential

surfaces. Integrated MEMS repeller electrodes on the substrate were developed, and

together with the symmetric equipotential surfaces, vertical extraction of generated

ion current was attainable.

 Improvements to support structures (i.e., panel electrodes, latches, etc.) through

failure analysis and iterative design and test was conducted. Improved structures

were implemented to improve mechanical and current carrying stability of the

MEMS ion source.

 LTCC scaffold was designed such that all ion source components, i.e. gas inlet and

exhaust, electron source, ion optics, and spatial filter, could be integrated into a small

form factor package.

 Characterization of the packaged ion source was conducted and verified using both

simulation and experimental results. Electron and ion current measurements were

conducted for numerous devices to fully characterize the performance of the

packaged ion source. Angular dispersion measurements of the emitted ion current

from the packaged was measured.

184

7.4.1 Recommendations for Further Investigation

The following recommendations to continue development of a miniature mass spectrometer have been developed from this research:

 To complete electrical and performance relevant characterization of the packaged ion

source, energy dispersion measurements must be completed. A full understanding

of the energy and angular dispersion of the ion source would allow for improved

understanding of the complete miniature mass spectrometer instrument.

o Preliminary attempts to measure energy dispersion were performed a similar

mount used for angular dispersion. A retarding grid was fixed between the ion

current exit point (i.e., coded aperture) and the detector. Ion current was

generated and the retarding grid potential was swept in an attempt to retard the

ion current and measure their energy. Inconsistencies in measurements due to

the presence of escaping electron current and the experimental setup presented

inconclusive results. Methods to surround the electron source portion of the ion

source with a magnetic field, to collimate the electron beam and reduce electron

escapement, was underway at the time of this writing.

 The next logical step after full characterization of the packaged ion source is

integration with the complete miniature mass spectrometer. Development of the

miniature mass spectrometer instrument is a combined effort and the result of the

entire research team. The packaged ion source was designed to integrate directly

185

with a cycloidal mass analyzer with orthogonal electrical and magnetic field,

vacuum components, and additional miniature components for full system

realization. The packaged ion source fits inside the electric sector and the emitted

ions follow a cycloidal path as seen in the schematic diagram of Figure 7.3(top left).

CAD images of the electrical sector, its integration with the detector, and the

complete integration into the vacuum manifold can be seen in Figure 7.3. The

magnet is fitted around the vacuum manifold. This work will be the focus of future

publications.

186

Appendix A

Current-Voltage Characterization: Program developed to control two (2)

Keithley 2410 SourceMeter Source Measure Unit (SMU) Instruments for the purposes of measuring the current response due to an applied voltage on each electrode (the sample and the counter electrode.) Keithley low level VIs provided by the manufacture were integrated to set both instruments in source voltage, measure current mode. The initialization was abstracted into a separate VI titled “Initialize Keithley for IV” which presets all of the instruments before the main loop of the program is initiated. The program also sets the measurement speed, compliance, range, and the voltage start, stop, and step levels. The data is plotted on screen and saved to an excel file.

187

188

189

Lifetime Characterization: Program developed to control two (2) Keithley 2410

SourceMeter Source Measure Unit (SMU) instruments for the purposes of characterizing the lifetime capabilities of a sample. The instruments are setup such that one electrode is placed into a voltage controlled current mode in order to source a fixed current from the sample by automatically modulating the voltage required to maintain that current. The other instrument is set into a voltage bias mode in order to maintain a specific voltage bias as a counter electrode. The instruments set the compliance and range values, as well as the bias levels and hours to test. The plots measure both the applied voltage on the cathode as well as the current.

190

191

192

Appendix B

Calibration Gas Spectra Discussion. Normalized mass spectra of perfluorotributylamine (PFTBA) are shown in Figure A1. Electron ionization (EI) spectra for both the carbon nanotube (CNT) field emission and filament thermionic emission sources (Figure A1a and Figure A1d, respectively) closely match each other as well as NIST reference EI spectra. PFTBA spectra were also collected in both negative chemical ionization (NCI) (Figure A1b and Figure A1e) and positive chemical ionization

(PCI) (Figure A1c and Figure A1f) modes using both the thermionic emission source as well as the CNT field emission source.

The PFTBA fragment ion peaks in NCI spectra taken with the CNT field emission source (Figure A1b) reveals nearly identical peaks to that of the filament thermionic emission source (Figure A1e). Strong peaks are observed at m/z 452 and 633 which is consistent with the reference spectra of PFTBA [316] and indicate chemical ionization conditions occurring with the CNT field emission source. The PCI mode revealed some differences, including the absence of the m/z 264 peak (a common PFTBA fragmentation peak) in the CNT field emission source that appears in the filament thermionic emission source spectra. Additionally, it should be noted that the spectra taken with the filament thermionic source in PCI mode stopped at m/z 600, so the m/z 654 peak that appears on the CNT field emission source spectra is not visible in the filament source spectra. Also, arguably the PFTBA is a much larger molecule (molecular weight 671 g/mol) with far

193

more fragmentation pathways than the smaller nitro compounds (eg. TNT, molecular weight 227 g/mol). Thus, small differences in electron energies during the ionization step might have a much greater influence on the PFTBA spectra and could have contributed to the small differences in the spectra collected.

Because the ionization energy of N2 is greater than that of many typical organic analytes, Hunt and Crow [186] state that “it is not surprising that nitrogen CI and conventional EI spectra are essentially identical.” Oxygen, also present in significant abundance in our ion source, similarly has an ionization energy greater than most organics and should behave like nitrogen in Reactions 3 and 4. We would not assert that our PCI PFTBA data taken with the CNT source is identical to the EI data, but it is similar. Therefore, the existence of additional peaks present in the PCI data taken with the CNT source spectra (Figure A1c), such as m/z 40 and m/z 69, could be explained by some unavoidable EI taking place. Additionally, the higher pressure in the source region under CI conditions seems to favor the production of higher m/z ions, likely because of increased collisional cooling of the high energy electrons injected into the region. While the CNT and filament PCI spectra in Figure A1c and Figure A1f, respectively, are not identical, they are much more similar than spectra collected under EI conditions.

Moreover, the EI data is far more similar to the NIST reference EI spectra than either PCI spectrum. The PCI PFTBA spectrum of both sources (CNT and filament) show an

194

increase in higher m/z ion intensity relative to the lower m/z ion intensity when collected in PCI mode versus EI mode, consistent with previous observations [317].

Calibration curves were collected and analyzed for the trace explosives. The calibration curves were generally linear, as expected. The data for the TNT calibration curve can be seen in Figure A2.

195

B.2: A calibration curve of desorbed TNT is shown and is linear, as expected. Trace explosive amounts were consistent with trace standards in the security field.

196

References

1. Jensen, K.L. and J.G. Webster, Field Emission — Fundamental Theory to Usage, in Wiley Encyclopedia of Electrical and Electronics Engineering. 1999, John Wiley & Sons, Inc.

2. Fursey, G.N., Field emission in vacuum micro-electronics. Applied Surface Science, 2003. 215(1–4): p. 113-134.

3. Chung, D.S., S.H. Park, Y.W. Jin, J.E. Jung, Y.J. Park, H.W. Lee, T.Y. Ko, S.Y. Hwang, J.W. Kim, N.H. Kwon, M.H. Yoon, C.G. Lee, J.H. You, N.S. Lee, and J.M. Kim. Field emission display using self-aligned carbon nanotube field emitters. in Vacuum Microelectronics Conference, 2001. IVMC 2001. Proceedings of the 14th International. 2001.

4. WB Choi, D.C., JH Kang, HY Kim, YW Jin, IT Han, YH Lee, JE Jung, NS Lee, GS Park, JM Kim, Fully sealed, high-brightness carbon-nanotube field-emission display. Applied Physics Letters, 1999. 75(20): p. 3129-3131.

5. Ajluni, C., FED Technology Takes Display Industry by Storm. Electronic Design, 1994. 42(22): p. 55-66.

6. Gaertner, G., Historical development and future trends of vacuum electronics. J. Vac. Sci. Technol. B, 2012. 30(060801).

7. Spindt, C.A., A Thin‐ Film Field‐ Emission Cathode. Journal of Applied Physics, 1968. 39(7): p. 3504-3505.

8. Matsuo, T., H. Matsuda, and I. Katakuse, SILICON EMITTER FOR FIELD DESORPTION MASS-SPECTROMETRY. Analytical Chemistry, 1979. 51(1): p. 69-72.

9. Temple, D., Recent progress in field emitter array development for high performance applications. Materials Science & Engineering R-Reports, 1999. 24(5): p. 185-239.

10. Xu, N.S. and S.E. Huq, Novel cold cathode materials and applications. Materials Science & Engineering R-Reports, 2005. 48(2-5): p. 47-189.

197

11. Fang, X., Y. Bando, U.K. Gautam, C. Ye, and D. Golberg, Inorganic semiconductor nanostructures and their field-emission applications. Journal of Materials Chemistry, 2008. 18(5): p. 509-522.

12. Iannazzo, S., A SURVEY OF THE PRESENT STATUS OF VACUUM MICROELECTRONICS. Solid-State Electronics, 1993. 36(3): p. 301-320.

13. Wu, R., K. Zhou, C.Y. Yue, J. Wei, and Y. Pan, Recent progress in synthesis, properties and potential applications of SiC nanomaterials. Progress in Materials Science, 2015. 72: p. 1-60.

14. Willander, M., M. Friesel, Q.U. Wahab, and B. Straumal, Silicon carbide and diamond for high temperature device applications. Journal of Materials Science- Materials in Electronics, 2006. 17(1): p. 1-25.

15. Davidson, J.L., W.P. Kang, K. Subramanian, and Y.M. Wong, Forms and behaviour of vacuum emission electronic devices comprising diamond or other carbon cold cathode emitters. Philosophical Transactions of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, 2008. 366(1863): p. 281-293.

16. Terranova, M.L., S. Orlanducci, M. Rossi, and E. Tamburri, Nanodiamonds for field emission: state of the art. Nanoscale, 2015. 7(12): p. 5094-5114.

17. Yu, Y., L. Wu, and J. Zhi, Diamond Nanowires: Fabrication, Structure, Properties and Applications, in Novel Aspects of Diamond: From Growth to Applications, N. Yang, Editor. 2015. p. 123-164.

18. Chernozatonskii, L.A., Y.V. Gulyaev, Z.J. Kosakovskaja, N.I. Sinitsyn, G.V. Torgashov, Y.F. Zakharchenko, E.A. Fedorov, and V.P. Val'chuk, Electron field emission from nanofilament carbon films. Chemical Physics Letters, 1995. 233(1– 2): p. 63-68.

19. de Heer, W.A., A. Châtelain, and D. Ugarte, A Carbon Nanotube Field-Emission Electron Source. Science, 1995. 270(5239): p. 1179-1180.

20. Ajayan, P.M. and T.W. Ebbesen, Nanometre-size tubes of carbon. Reports on Progress in Physics, 1997. 60(10): p. 1025-1062.

21. Zhu, W., C. Bower, O. Zhou, G. Kochanski, and S. Jin, Large current density from carbon nanotube field emitters. Applied Physics Letters, 1999. 75(6): p. 873-875. 198

22. Saito, Y. and S. Uemura, Field emission from carbon nanotubes and its application to electron sources. Carbon, 2000. 38(2): p. 169-182.

23. Fleming, J.A., Instrument for converting alternating electric currents into continuous currents. 1905, Google Patents.

24. De Forest, L., Space telegraphy. 1908, Google Patents.

25. Adams, M., Lee de Forest: King of Radio, Television, and Film. 1 ed, ed. L. Springer Science+Business Media. 2012: Copernicus.

26. Bardeen, J. and W.H. Brattain, The Transistor, A Semi-Conductor Triode. Physical Review, 1948. 74(2): p. 230-231.

27. Kilby, J.S., Invention of the integrated circuit. Electron Devices, IEEE Transactions on, 1976. 23(7): p. 648-654.

28. Noyce, R.N., Semiconductor Device-and-Lead Structure, Reprint of U.S. Patent 2,981,877 (Issued April 25, 1961. Filed July 30, 1959). Solid-State Circuits Society Newsletter, IEEE, 2007. 12(2): p. 34-40.

29. Czichy, R.H., Aerospace Sensors, in Sensors. 2008, Wiley-VCH Verlag GmbH. p. 365-411.

30. Gerblinger, J., K.H. Haerdtl, H. Meixner, and R. Aigner, High-Temperature Microsensors, in Sensors. 2008, Wiley-VCH Verlag GmbH. p. 181-219.

31. Pratt, N.K. and R. Jones, Sensors in Manufacturing and Quality Assurance, in Sensors. 2008, Wiley-VCH Verlag GmbH. p. 525-538.

32. Jones, K.W., Environmental Sensors, in Sensors. 2008, Wiley-VCH Verlag GmbH. p. 451-489.

33. Anatoliy Evtukh, H.H., Oktay Yilmazoglu, Hidenori Mimura, Dimitris Pavlidis, Vacuum Nanoelectronic Devices - Novel Electron Sources and Applications. 2015: Wiley.

34. Ado Jorio, G.D., Mildred S. Dresselhaus, Carbon Nanotubes: Advanced Topics in the Synthesis, Structure, Properties and Applications. Vol. 111. 2008: Springer.

199

35. Das, R., Carbon Nanotubes (CNT) for Electronics & Electrics 2013-2023: Forecasts, Applications, Technologies, IDTechEx, Editor. 2013.

36. Eletskii, G.S.B.a.A.V., Theory of Carbon Nanotube (CNT)-Based Electron Field Emitters. Nanomaterials, 2013. 3: p. 393-442.

37. Harris, P.J.F., Carbon Nanotube Science: Synthesis, Properties and Applications. 2011: Cambridge University Press 2009. 314.

38. Williams, L.T., V.S. Kumsomboone, W.J. Ready, and M.L.R. Walker, Lifetime and Failure Mechanisms of an Arrayed Carbon Nanotube Field Emission Cathode. Electron Devices, IEEE Transactions on, 2010. 57(11): p. 3163-3168.

39. Iijima, S., Helical microtubules of graphitic carbon. Nature, 1991. 354(6348): p. 56-58.

40. Ebbesen, T.W. and P.M. Ajayan, Large-scale synthesis of carbon nanotubes. Nature, 1992. 358(6383): p. 220-222.

41. Iijima, S. and T. Ichihashi, Single-shell carbon nanotubes of 1-nm diameter. Nature, 1993. 363(6430): p. 603-605.

42. Bethune, D.S., C.H. Klang, M.S. de Vries, G. Gorman, R. Savoy, J. Vazquez, and R. Beyers, Cobalt-catalysed growth of carbon nanotubes with single-atomic-layer walls. Nature, 1993. 363(6430): p. 605-607.

43. Pastore, R. Allotropes of Carbon, Part 4 - The Carbon Nanotube. 2015 [cited 2016 1/22/2016]; Available from: http://learningchemistryeasily.blogspot.com/2015/03/allotropes-of-carbon-part-4- carbon.html.

44. Tara Spires and R. Malcolm Brown, J., High Resolution TEM Observations of Single-Walled Carbon Nanotubes. Department of Botany, The University of Texas at Austin, Austin,Tx., 78713, 1996.

45. Lucas, A.A., F. Moreau, and P. Lambin, Colloquium: Optical simulations of electron diffraction by carbon nanotubes. Reviews of Modern Physics, 2002. 74(1): p. 1-10.

46. Dai, H.J., Carbon nanotubes: opportunities and challenges. Surface Science, 2002. 500(1-3): p. 218-241.

200

47. Hou, Y., J. Tang, H. Zhang, C. Qian, Y. Feng, and J. Liu, Functionalized Few- Walled Carbon Nanotubes for Mechanical Reinforcement of Polymeric Composites. ACS Nano, 2009. 3(5): p. 1057-1062.

48. Nagai, H., Y. Okazaki, S.H. Chew, N. Misawa, Y. Yamashita, S. Akatsuka, T. Ishihara, K. Yamashita, Y. Yoshikawa, H. Yasui, L. Jiang, H. Ohara, T. Takahashi, G. Ichihara, K. Kostarelos, Y. Miyata, H. Shinohara, and S. Toyokuni, Diameter and rigidity of multiwalled carbon nanotubes are critical factors in mesothelial injury and carcinogenesis. Proceedings of the National Academy of Sciences, 2011. 108(49): p. E1330–E1338.

49. Charlier, J.C., Defects in Carbon Nanotubes. Accounts of Chemical Research, 2002. 35(12): p. 1063-1069.

50. Dresselhaus, M.S., A. Jorio, A.G. Souza Filho, and R. Saito, Defect characterization in graphene and carbon nanotubes using Raman spectroscopy. Philosophical Transactions of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, 2010. 368(1932): p. 5355-5377.

51. Narlikar, A.V. and Y.Y. Fu, Oxford Handbook of Nanoscience and Technology. Vol. 1. 2010: Oxford University Press. 856.

52. Kumar, M., Carbon Nanotube Synthesis and Growth Mechanism, in Carbon Nanotubes - Synthesis, Characterization, Applications, S. Yellampalli, Editor. 2011, InTech. p. 528.

53. Thess, A., R. Lee, P. Nikolaev, H. Dai, P. Petit, J. Robert, C. Xu, Y.H. Lee, S.G. Kim, A.G. Rinzler, D.T. Colbert, G.E. Scuseria, D. Tománek, J.E. Fischer, and R.E. Smalley, Crystalline Ropes of Metallic Carbon Nanotubes. Science, 1996. 273(5274): p. 483-487.

54. Cassell, A.M., J.A. Raymakers, J. Kong, and H. Dai, Large Scale CVD Synthesis of Single-Walled Carbon Nanotubes. The Journal of Physical Chemistry B, 1999. 103(31): p. 6484-6492.

55. Meyyappan, M., D. Lance, C. Alan, and H. David, Carbon nanotube growth by PECVD: a review. Plasma Sources Science and Technology, 2003. 12(2): p. 205.

56. Cui, H., O. Zhou, and B.R. Stoner, Deposition of aligned bamboo-like carbon nanotubes via microwave plasma enhanced chemical vapor deposition. Journal of Applied Physics, 2000. 88(10): p. 6072-6074. 201

57. Natarajan, S., A study of field emission based microfabricated devices, in Department of Electrical and Computer Engineering. 2008, Duke University.

58. Cui, H., Nucleation and growth of nanoscaled one-dimensional materials. 2001, University of North Carolina at Chapel Hill.

59. Kepple, K.L., G.P. Sanborn, P.A. Lacasse, K.M. Gruenberg, and W.J. Ready, Improved fracture toughness of carbon fiber composite functionalized with multi walled carbon nanotubes. Carbon, 2008. 46(15): p. 2026-2033.

60. Bhushan, B., Springer handbook of nanotechnology. 2010.

61. Nardelli, M.B., B.I. Yakobson, and J. Bernholc, Mechanism of strain release in carbon nanotubes. Physical Review B, 1998. 57(8): p. R4277-R4280.

62. Yu, M.F., O. Lourie, M.J. Dyer, K. Moloni, T.F. Kelly, and R.S. Ruoff, Strength and breaking mechanism of multiwalled carbon nanotubes under tensile load. Science, 2000. 287(5453): p. 637-640.

63. Shearer, C.J., J. Yu, K.M. O'Donnell, L. Thomsen, P.C. Dastoor, J.S. Quinton, and J.G. Shapter, Highly resilient field emission from aligned single-walled carbon nanotube arrays chemically attached to n-type silicon. Journal of Materials Chemistry, 2008. 18(47): p. 5753-5760.

64. Kim, J.W., J.W. Jeong, J.T. Kang, S. Choi, S. Park, M.S. Shin, S. Ahn, and Y.H. Song, Great improvement in adhesion and uniformity of carbon nanotube field emitters through reactive nanometer-scale SiC fillers. Carbon, 2015. 82: p. 245- 253.

65. Wei, W., Y. Liu, Y. Wei, K.L. Jiang, L.M. Peng, and S.S. Fan, Tip cooling effect and failure mechanism of field-emitting carbon nanotubes. Nano Letters, 2007. 7(1): p. 64-68.

66. Williams, L.T., V.S. Kumsomboone, W.J. Ready, and M.L.R. Walker, Lifetime and Failure Mechanisms of an Arrayed Carbon Nanotube Field Emission Cathode. Ieee Transactions on Electron Devices, 2010. 57(11): p. 3163-3168.

67. Zhifeng Ren, Y.L., Yang Wang, Aligned Carbon Nanotubes: Physics, Concepts, Fabrication and Devices. 2012: Springer.

202

68. Martín, D.M.G.N., Carbon nanotubes and related structures : synthesis, characterization, functionalization, and applications. 2010: Wiley.

69. Avouris, P., Carbon nanotube electronics. Chemical Physics, 2002. 281(2–3): p. 429-445.

70. Frank, S., P. Poncharal, Z.L. Wang, and W.A. de Heer, Carbon nanotube quantum resistors. Science, 1998. 280(5370): p. 1744-1746.

71. Robertson, J., Growth of nanotubes for electronics. Materials Today, 2007. 10(1- 2): p. 36-43.

72. Schatz, J.C.K.-H.L.J.L.D.T.G.C., Multi-walled carbon nanotubes experiencing electrical breakdown as gas sensors. Nanotechnology, 2004. 15(11): p. 1596- 1602.

73. Waldrop, M.M. The chips are down for Moore’s law. Nature - International Weekly Journal of Science, 2016.

74. Biercuk, M.J., M.C. Llaguno, M. Radosavljevic, J.K. Hyun, A.T. Johnson, and J.E. Fischer, Carbon nanotube composites for thermal management. Applied Physics Letters, 2002. 80(15): p. 2767-2769.

75. Kreupl, F., A.P. Graham, G.S. Duesberg, W. Steinhogl, M. Liebau, E. Unger, and W. Honlein, Carbon nanotubes in interconnect applications. Microelectronic Engineering, 2002. 64(1-4): p. 399-408.

76. Li, J., Q. Ye, A. Cassell, H.T. Ng, R. Stevens, J. Han, and M. Meyyappan, Bottom-up approach for carbon nanotube interconnects. Applied Physics Letters, 2003. 82(15): p. 2491-2493.

77. Naeemi, A., R. Sarvari, and J.D. Meindl, Performance comparison between carbon nanotube and copper interconnects for gigascale integration (GSI). Ieee Electron Device Letters, 2005. 26(2): p. 84-86.

78. Claye, A.S., J.E. Fischer, C.B. Huffman, A.G. Rinzler, and R.E. Smalley, Solid- state electrochemistry of the Li single wall carbon nanotube system. Journal of the Electrochemical Society, 2000. 147(8): p. 2845-2852.

79. Shimoda, H., B. Gao, X.P. Tang, A. Kleinhammes, L. Fleming, Y. Wu, and O. Zhou, Lithium Intercalation into Opened Single-Wall Carbon Nanotubes: Storage

203

Capacity and Electronic Properties. Physical Review Letters, 2001. 88(1): p. 015502.

80. Landi, B.J., S.L. Castro, H.J. Ruf, C.M. Evans, S.G. Bailey, and R.P. Raffaelle, CdSe quantum dot-single wall carbon nanotube complexes for polymeric solar cells. Solar Energy Materials and Solar Cells, 2005. 87(1–4): p. 733-746.

81. Kymakis, E. and G.A.J. Amaratunga, Electrical properties of single-wall carbon nanotube-polymer composite films. Journal of Applied Physics, 2006. 99(8): p. 7.

82. Che, G., B.B. Lakshmi, E.R. Fisher, and C.R. Martin, Carbon nanotubule membranes for electrochemical energy storage and production. Nature, 1998. 393(6683): p. 346-349.

83. Wang, C., M. Waje, X. Wang, J.M. Tang, R.C. Haddon, and Yan, Proton Exchange Membrane Fuel Cells with Carbon Nanotube Based Electrodes. Nano Letters, 2004. 4(2): p. 345-348.

84. Sinha, N., J.Z. Ma, and J.T.W. Yeow, Carbon nanotube-based sensors. Journal of Nanoscience and Nanotechnology, 2006. 6(3): p. 573-590.

85. Trojanowicz, M., Analytical applications of carbon nanotubes: a review. Trac- Trends in Analytical Chemistry, 2006. 25(5): p. 480-489.

86. Xu, N.S. and S.E. Huq, Novel Cold Cathode Materials and Applications. 2005.

87. Wang, Z.M., One-Dimensional Nanostructures. Springer.

88. Kasap, S., Principles of Electronic Materials and Devices. 3rd ed. 2005: McGraw-Hill Education. 768.

89. Modinos, A., Field, Thermionic and Secondary Electron Emission Spectroscopy. 1984: Springer US.

90. Chen, N.P., Handbook of Light Emitting and Schottky Diode Research. 2010: Nova Science Publishers.

91. YAGHOOBI, P. and A. NOJEH, ELECTRON EMISSION FROM CARBON NANOTUBES. Modern Physics Letters B, 2007. 21(27): p. 1807-1830.

204

92. de Bruyne, N.A., The action of strong electric fields on the current from a thermionic cathode. Proceedings of the Royal Society of London Series a- Containing Papers of a Mathematical and Physical Character, 1928. 120(785): p. 423-437.

93. Loiseau, A., Launois-Bernede, P., Petit, P., Roche, S., Salvetat, J.-P, Understanding Carbon Nanotubes: From Basics to Application. 2006, Springer.

94. Schottky, W., Concerning cold and hot electron discharge. Zeitschrift Fur Physik, 1923. 14: p. 63-106.

95. N Froman, P.O.F.a., JWKB Approximation Contributions to the Theory. 1965.

96. Groning, P., P. Ruffieux, L. Schlapbach, and O. Groning, Carbon nanotubes for cold electron sources. Advanced Engineering Materials, 2003. 5(8): p. 541-550.

97. Murphy, E.L. and R.H. Good, THERMIONIC EMISSION, FIELD EMISSION, AND THE TRANSITION REGION. Physical Review, 1956. 102(6): p. 1464-1473.

98. Cheng, Y. and O. Zhou, Electron field emission from carbon nanotubes. Comptes Rendus Physique, 2003. 4(9): p. 1021-1033.

99. Al-Tabbakh, A.A., M.A. More, D.S. Joag, I.S. Mulla, and V.K. Pillai, The Fowler−Nordheim Plot Behavior and Mechanism of Field Electron Emission from ZnO Tetrapod Structures. ACS Nano, 2010. 4(10): p. 5585-5590.

100. Forbes, R.G., J.H.B. Deane, and Ieee, Progress with developing theory for Fowler-Nordheim plot interpretation, in 2013 26th International Vacuum Nanoelectronics Conference. 2013.

101. Forbes, R.G., Extraction of emission parameters for large-area field emitters, using a technically complete Fowler-Nordheim-type equation. Nanotechnology, 2012. 23(9).

102. Forbes, R.G., Development of a simple quantitative test for lack of field emission orthodoxy. Proceedings of the Royal Society a-Mathematical Physical and Engineering Sciences, 2013. 469(2158).

103. Forbes, R.G., Use of a Spreadsheet to Test for Lack of Field Emission Orthodoxy. 2014 Tenth International Vacuum Electron Sources Conference (Ivesc), 2014.

205

104. Anantram, M.P. and F. Leonard, Physics of carbon nanotube electronic devices. Reports on Progress in Physics, 2006. 69(3): p. 507-561.

105. Rinzler, A.G., J.H. Hafner, P. Nikolaev, P. Nordlander, D.T. Colbert, R.E. Smalley, L. Lou, S.G. Kim, and D. Tománek, Unraveling Nanotubes: Field Emission from an Atomic Wire. Science, 1995. 269(5230): p. 1550-1553.

106. Jonge, N.d. and J.-M. Bonard, Carbon Nanotube Electron Sources and Applications. Philosophical Transactions: Mathematical, Physical and Engineering Sciences, 2004. 362(1823): p. 2239-2266.

107. Nojeh, A., Carbon Nanotube Electron Sources: From Electron Beams to Energy Conversion and Optophononics. ISRN Nanomaterials, 2014. 2014: p. 23.

108. Williams, D.B. and C.B. Carter, Transmission Electron Microscopy. 2 ed. 2009: Springer US.

109. Bower, C., O. Zhou, W. Zhu, A.G. Ramirez, G.P. Kochanski, and S. Jin, Fabrication and field emission properties of carbon nanotube cathodes. Materials Research Society: Amorphous and Nanostructured Carbon, 2000.

110. Shiffler, D., O. Zhou, C. Bower, A. LaCour, and K. Golby, A high-current, large- area, carbon nanotube cathode. Ieee Transactions on Plasma Science, 2004. 32(5): p. 2152-2154.

111. Fennimore, A.M.R., D. H.; Wilson, G. A.; et al., Enhancing lifetime of carbon nanotube field emitters through hydrocarbon exposure. Appl. Phys. Lett., 2008. 92(21): p. 213108.

112. Liang, S.-D. and L. Chen, Generalized Fowler-Nordheim Theory of Field Emission of Carbon Nanotubes. Physical Review Letters, 2008. 101(2): p. 027602.

113. Yuasa, K., A. Shimoi, I. Ohba, and C. Oshima, Modified Fowler–Nordheim field emission formulae from a nonplanar emitter model. Surface Science, 2002. 520(1–2): p. 18-28.

114. Li, X., G. Bai, H. Li, M. Ding, J. Feng, F. Liao, and Ieee, A Potential Failure Reason for Spindt Cathodes in High Current Applications. Ieee International Vacuum Electronics Conference, 2014: p. 509-510.

206

115. Chalamala, B.R., R.M. Wallace, and B.E. Gnade, Poisoning of Spindt-type molybdenum field emitter arrays by CO2. Journal of Vacuum Science & Technology B, 1998. 16(5): p. 2866-2870.

116. !!! INVALID CITATION !!! [32, 33].

117. Velasquez-Garcia, L.F., S.A. Guerrera, Y. Niu, and A.I. Akinwande, Uniform High-Current Cathodes Using Massive Arrays of Si Field Emitters Individually Controlled by Vertical Si Ungated FETs—Part 2: Device Fabrication and Characterization. Electron Devices, IEEE Transactions on, 2011. 58(6): p. 1783- 1791.

118. Meassick, S. and H. Champaign, Influence of fill gases on the failure rate of gated silicon field emitter arrays. Journal of Vacuum Science & Technology B, 1996. 14(3): p. 1914-1917.

119. Huang, Z., N.E. McGruer, and K. Warner, 200-NM GATED FIELD EMITTERS. Ieee Electron Device Letters, 1993. 14(3): p. 121-122.

120. Aplin, K.L., C.M. Collingwood, and B.J. Kent, Reliability tests of gated silicon field emitters for use in space. Journal of Physics D-Applied Physics, 2004. 37(14): p. 2009-2017.

121. Ancona, M.G., Modeling of thermal effects in silicon field emitters. Journal of Vacuum Science & Technology B, 1996. 14(3): p. 1918-1923.

122. Ghosh, N., W.P. Kang, and J.L. Davidson, Fabrication and implementation of nanodiamond lateral field emission diode for logic OR function. Diamond and Related Materials, 2012. 23: p. 120-124.

123. Subramanian, K., K. Weng Poo, and J.L. Davidson, A Monolithic Nanodiamond Lateral Field Emission Vacuum Transistor. Electron Device Letters, IEEE, 2008. 29(11): p. 1259-1261.

124. SH Hsu, W.K., S Raina, JH Huang, Nanodiamond vacuum field emission device with gate modulated triode characteristics. Applied Physics Letters, 2013. 102(20): p. 203105.

125. Kang, W.P., S.H. Hsu, N. Ghosh, J.L. Davidson, J.H. Huang, and D.V. Kerns. Nanodiamond vacuum field emission integrated devices. in Vacuum Nanoelectronics Conference (IVNC), 2012 25th International. 2012.

207

126. Shao-Hua, H., K. Weng Poo, J.L. Davidson, J.H. Huang, and D.V. Kerns, Nanodiamond Vacuum Field Emission Integrated Differential Amplifier. Electron Devices, IEEE Transactions on, 2013. 60(1): p. 487-493.

127. Groning, O., R. Clergereaux, L.O. Nilsson, P. Ruffieux, P. Groning, and L. Schlapbach, Prospects and limitations of carbon nanotube field emission electron sources. Chimia, 2002. 56(10): p. 553-561.

128. Bonard, J.M., C. Klinke, K.A. Dean, and B.F. Coll, Degradation and failure of carbon nanotube field emitters. Physical Review B, 2003. 67(11).

129. Mahapatra, D.R., N. Sinha, and R.V.N. Melnik, Degradation and Failure of Field Emitting Carbon Nanotube Arrays. Journal of Nanoscience and Nanotechnology, 2011. 11(5): p. 3911-3915.

130. Sanborn, G., S. Turano, P. Collins, and W.J. Ready, A thin film triode type carbon nanotube field emission cathode. Applied Physics A, 2012. 110(1): p. 99-104.

131. Kaiser, M., M. Doytcheva, M. Verheijen, and N. de Jonge, In situ transmission electron microscopy observations of individually selected freestanding carbon nanotubes during field emission. Ultramicroscopy, 2006. 106(10): p. 902-908.

132. Wang, Z.L., R.P. Gao, W.A. de Heer, and P. Poncharal, In situ imaging of field emission from individual carbon nanotubes and their structural damage. Applied Physics Letters, 2002. 80(5): p. 856-858.

133. Milanovic, V., L. Doherty, D.A. Teasdale, S. Parsa, and K.S.J. Pister, Micromachining technology for lateral field emission devices. Ieee Transactions on Electron Devices, 2001. 48(1): p. 166-173.

134. Lee, C.S., J.D. Lee, and C.H. Han, A new lateral field emission device using chemical-mechanical polishing. Ieee Electron Device Letters, 2000. 21(10): p. 479-481.

135. Mei, Q., S. Zurn, and D.L. Polla, PROCESS CHARACTERIZATION AND ANALYSIS OF SEALED VACUUM MICROELECTRONIC DEVICES. Journal of Vacuum Science & Technology B, 1994. 12(2): p. 638-643.

136. Lee, S.B., A.S. Teh, K.B.K. Teo, M. Chhowalla, D.G. Hasko, W.I. Milne, G.A.J. Amaratunga, and H. Ahmed, Fabrication of carbon nanotube lateral field emitters. Nanotechnology, 2003. 14(2): p. 192-195.

208

137. Lee, J.H., M.B. Lee, S.H. Hahm, J.H. Lee, H.I. Seo, D.H. Kwon, J.S. Kim, and K.M. Choi, Lateral field emitter arrays with high emission currents and wide operation region by high field activation. Journal of Vacuum Science & Technology B, 2003. 21(1): p. 506-510.

138. Lee, J.H., K.R. Kwon, H.J. Lee, M.B. Lee, H.I. Sea, D.H. Kwon, J.S. Kim, K.M. Choi, S.H. Hahm, and J.H. Lee, Lateral field emitter arrays with higher emission currents and wider operation region by high field activation. Ivmc 2000: Proceedings of the 14th International Vacuum Microelectronices Conference, ed. C.E. Hunt, et al. 2001. 149-150.

139. Park, S.S., D.I. Park, S.H. Hahm, J.H. Lee, H.C. Choi, and J.H. Lee, Fabrication of a lateral field emission triode with a high current density and high transconductance using the local oxidation of the polysilicon layer. Ieee Transactions on Electron Devices, 1999. 46(6): p. 1283-1289.

140. Itoh, J., K. Tsuburaya, S. Kanemaru, T. Watanabe, and S. Itoh, FABRICATION AND CHARACTERIZATION OF COMB-SHAPED LATERAL FIELD-EMITTER ARRAYS. Japanese Journal of Applied Physics Part 1-Regular Papers Short Notes & Review Papers, 1993. 32(3A): p. 1221-1226.

141. Kanemaru, S. and J. Itoh, FABRICATION AND CHARACTERIZATION OF LATERAL FIELD-EMITTER TRIODES. Ieee Transactions on Electron Devices, 1991. 38(10): p. 2334-2336.

142. Oro, J.A. and D.D. Ball, LATERAL FIELD-EMISSION DEVICES WITH SUBTENTH-MICRON EMITTER TO ANODE SPACING. Journal of Vacuum Science & Technology B, 1993. 11(2): p. 464-467.

143. Gray, H.F., J.L. Shaw, A.I. Akinwande, and P. Bauhahn. Film edge emitters: the basis for a new vacuum transistor. in Electron Devices Meeting, 1991. IEDM '91. Technical Digest., International. 1991.

144. Busta, H.H., LATERAL COLD-CATHODE TRIODE STRUCTURES FABRICATED ON INSULATING SUBSTRATES. Institute of Physics Conference Series, 1989(99): p. 29-32.

145. Bower, C., D. Shalóm, W. Zhu, D. López, G.P. Kochanski, P.L. Gammel, and S. Jin, A Micromachined Vacuum Triode Using a Carbon Nanotube Cold Cathode. IEEE Transactions on Electron Devices, 2002. 49(8): p. 1478-1483.

209

146. Bower, C., W. Zhu, D. Shalom, D. Lopez, L.H. Chen, P.L. Gammel, and S. Jin, On-chip vacuum microtriode using carbon nanotube field emitters. Appl. Phys. Lett., 2002. 80(20).

147. Bower, C.A., K.H. Gilchrist, J.R. Piascik, B.R. Stoner, S. Natarajan, C.B. Parker, S.D. Wolter, and J.T. Glass, On-chip electron-impact ion source using carbon nanotube field emitters. Applied Physics Letters, 2007. 90(12).

148. Gilchrist, K.H., C.A. Bower, M.R. Lueck, J.R. Piascik, B.R. Stoner, S. Natarajan, C.B. Parker, and J.T. Glass. A Novel Ion Source and Detector for a Miniature Mass Spectrometer. in Sensors, 2007 IEEE. 2007.

149. Natarajan, S., C.B. Parker, J.R. Piascik, K.H. Gilchrist, B.R. Stoner, and J.T. Glass, Analysis of 3-panel and 4-panel microscale ionization sources. Journal of Applied Physics, 2010. 107(12): p. 124508.

150. Srividya Natarajan, K.H.G., Jeffrey R. Piascik, Charles B. Parker, Jeffrey T. Glass and Brian R. Stoner, Simulation and testing of a lateral, microfabricated electron- impact ion source. Applied Physics Letters, 2008. 94(044109).

151. Stoner, B.R., J.R. Piascik, K.H. Gilchrist, C.B. Parker, and J.T. Glass, A Bipolar Vacuum Microelectronic Device. Electron Devices, IEEE Transactions on, 2011. 58(9): p. 3189-3194.

152. MEMSCAP. MUMPs®. 2011-2012 Feb. 2015]; An explanation of the MUMPS process]. Available from: http://www.memscap.com/products/mumps.

153. Gross, J.H., Mass Spectrometry: A Textbook. 2nd ed. 2011, Heidelberg: Springer.

154. Borman, S. A Brief History of Mass Spectrometry Instrumentation. 1998.

155. Glish, G.L. and R.W. Vachet, The basics of mass spectrometry in the twenty-first century. Nat Rev Drug Discov, 2003. 2(2): p. 140-150.

156. El-Aneed, A., A. Cohen, and J. Banoub, Mass Spectrometry, Review of the Basics: Electrospray, MALDI, and Commonly Used Mass Analyzers. Applied Spectroscopy Reviews, 2009. 44(3): p. 210-230.

157. A. D. McNaught, A.W., Compendium of Chemical Terminology+. 2nd ed. 1997: International Union of Pure and Applied Chemistry (IUPAC).

210

158. Anderson, R.J., D.J. Bendell, P.W. Groundwater, E.W. Abel, A.G. Davies, D. Phillips, and J.D. Woollins, Organic Spectroscopic Analysis. 2006.

159. Todd, J.F.J., Recommendations for nomenclature and symbolism for mass- spectroscopy -(including an appendix of terms used in vacuum technology). International Journal of Mass Spectrometry and Ion Processes, 1995. 142(3): p. 211-240.

160. Bocharov, G.S. and A.V. Eletskii, Degradation of a carbon nanotube-based field- emission cathode during ion sputtering. Technical Physics, 2012. 57(7): p. 1008- 1012.

161. Paulmier, T., M. Balat-Pichelin, D. Le Quéau, R. Berjoan, and J.F. Robert, Physico-chemical behavior of carbon materials under high temperature and ion irradiation. Applied Surface Science, 2001. 180(3–4): p. 227-245.

162. Teo, K.B.K., M. Chhowalla, G.A.J. Amaratunga, W.I. Milne, G. Pirio, P. Legagneux, F. Wyczisk, D. Pribat, and D.G. Hasko, Field emission from dense, sparse, and patterned arrays of carbon nanofibers. Applied Physics Letters, 2002. 80(11): p. 2011-2013.

163. Chen, Z., Q. Zhang, P. Lan, B. Zhu, T. Yu, G. Cao, and D. den Engelsen, Ultrahigh-current field emission from sandwich-grown well-aligned uniform multi-walled carbon nanotube arrays with high adherence strength. Nanotechnology, 2007. 18: 265702(26): p. 1-6.

164. Department of Homeland Security Appropriations Act, D.o.H. Security, Editor. 2011-2015: Congress.gov.

165. Administration, T.S., TSA Expands Use of Explosive Trace Detection Technology at Airports Nationwide. 2010: Official website of the Department of Homeland Security.

166. Alberici, R.M., R.C. Simas, G.B. Sanvido, W. Romao, P.M. Lalli, M. Benassi, I.B. Cunha, and M.N. Eberlin, Ambient mass spectrometry: bringing MS into the “real world”. Analytical and Bioanalytical Chemistry, 2010. 398(1): p. 265-294.

167. Harris, G.A., A.S. Galhena, and F.M. Fernandez, Ambient Sampling/Ionization Mass Spectrometry: Applications and Current Trends. Analytical Chemistry, 2011. 83(12): p. 4508-4538.

211

168. Xu, X., A.M. van de Craats, and P.C.A.M. de Bruyn, Highly Sensitive Screening Method for Nitroaromatic, Nitramine and Nitrate Ester Explosives by High Performance Liquid Chromatography--Atmospheric Pressure Ionization--Mass Spectrometry (HPLC-API-MS) in Forensic Applications. Journal of Forensic Sciences, 2004. 49(6): p. 1171-1180.

169. Wang, Z., Y. Zuo, Y. Li, X. Han, X. Guo, J. Wang, B. Cao, L. Xi, and D. Xue, Improved field emission properties of carbon nanotubes decorated with Ta layer. Carbon, 2014. 73: p. 114-124.

170. Lan, Y., Y. Wang, and Z.F. Ren, Physics and applications of aligned carbon nanotubes. Advances in Physics, 2011. 60(4): p. 553-678.

171. Kyusung, H., L. Yuri, J. Dohan, L. Soonil, J. Kwang Woo, and Y. Sang Sik, Field Emission Ion Source Using a Carbon Nanotube Array for Micro Time-of-Flight Mass Spectrometer. Japanese Journal of Applied Physics, 2011. 50(6S): p. 06GM04.

172. Getty, S.A., T.T. King, R.A. Bis, H.H. Jones, F. Herrero, B.A. Lynch, P. Roman, and P. Mahaffy. Performance of a carbon nanotube field emission electron gun. in Micro (MEMS) and Nanotechnologies for Defense and Security. 2007. Proc. of SPIE.

173. Evans-Nguyen, T., C.B. Parker, C. Hammock, A.H. Monica, E. Adams, L. Becker, J.T. Glass, and R.J. Cotter, Carbon Nanotube Electron Ionization Source for Portable Mass Spectrometry. Analytical Chemistry, 2011. 83(17): p. 6527- 6531.

174. Lu, X., Q. Yang, C. Xiao, and A. Hirose, Nonlinear Fowler–Nordheim plots of the field electron emission from graphitic nanocones: influence of non-uniform field enhancement factors. Journal of Physics D: Applied Physics, 2006. 39(15): p. 3375.

175. Zhang, J., C. Yang, W. Yang, T. Feng, X. Wang, and X. Liu, Appearance of a knee on the Fowler–Nordheim plot of carbon nanotubes on a substrate. Solid State Communications, 2006. 138(1): p. 13-16.

176. Feng, Y. and J.P. Verboncoeur, Transition from Fowler-Nordheim field emission to space charge limited current density. Physics of Plasmas (1994-present), 2006. 13(7).

212

177. Barbour, J.P., W.W. Dolan, J.K. Trolan, E.E. Martin, and W.P. Dyke, Space- Charge Effects in Field Emission. Physical Review, 1953. 92(1): p. 45-51.

178. Bai, R. and H. Kirkici. Nonlinear Fowler-Nordheim plots of carbon nanotubes (CNTs) in vacuum and partial pressures. in Power Modulator and High Voltage Conference (IPMHVC), 2012 IEEE International. 2012. San Diego, CA, USA: IEEE.

179. Leberl, D., R. Ummethala, A. Leonhardt, B. Hensel, S.F. Tedde, O. Schmidt, and O. Hayden, Characterization of carbon nanotube field emitters in pulsed operation mode. Journal of Vacuum Science & Technology B, 2013. 31(1).

180. Dean, K.A., T.P. Burgin, and B.R. Chalamala, Evaporation of carbon nanotubes during electron field emission. Applied Physics Letters, 2001. 79(12): p. 1873- 1875.

181. Wei, Y., C. Xie, K.A. Dean, and B.F. Coll, Stability of carbon nanotubes under electric field studied by scanning electron microscopy. Applied Physics Letters, 2001. 79(27): p. 4527-4529.

182. Mahapatra, D., N. Sinha, and R. Melnik, Degradation and Failure of Field Emitting Carbon Nanotube Arrays. Journal of Nanoscience and Nanotechnology, 2011. 11(5): p. 3911-5.

183. Field, F.H., Chemical Ionization Mass Ssectrometrv. Accounts of Chemical Research, 1968. 1(42).

184. Westmore, J.B. and M.M. Alauddin, Ammonia chemical ionization mass spectrometry. Mass Spectrometry Reviews, 1986. 5(4): p. 381-465.

185. Yinon, J., Analysis of explosives by negative-ion chemical ionization mass spectrometry. Journal of Forensic Sciences, 1980. 25(2): p. 401-407.

186. Hunt, D.F., G.C. Stafford, F.W. Crow, and J.W. Russell, Pulsed positive negative ion chemical ionization mass spectrometry. Analytical Chemistry, 1976. 48(14): p. 2098-2104.

187. McLuckey, S.A., G.L. Glish, and P.E. Kelley, Collision-activated dissociation of negative ions in an ion trap mass spectrometer. Analytical Chemistry, 1987. 59(13): p. 1670-1674.

213

188. Gregor, I.K. and M. Guilhaus, Comparative study of the influence of the electron- energy moderating gas in electron-attachment reactions under NCI conditions. International Journal of Mass Spectrometry and Ion Processes, 1984. 56(2): p. 167-176.

189. Yinon, J., D.J. Harvan, and J.R. Hass, Mass spectral fragmentation pathways in RDX and HMX. A mass analyzed ion kinetic energy spectrometric/collisional induced dissociation study. Organic Mass Spectrometry, 1982. 17(7): p. 321-326.

190. Boumsellek, S., S.H. Alajajian, and A. Chutjian, Negative-ion formation in the explosives RDX, PETN, and TNT by using the reversal electron attachment detection technique. Journal of the American Society for Mass Spectrometry, 1992. 3(3): p. 243-247.

191. Spindt, C.A., C.E. Holland, A. Rosengreen, and I. Brodie, Field-emitter arrays for vacuum microelectronics. Electron Devices, IEEE Transactions on, 1991. 38(10): p. 2355-2363.

192. Brodi, I., Physical considerations in vacuum microelectronics devices. Electron Devices, IEEE Transactions on, 1989. 36(11): p. 2641-2644.

193. Utsumi, T., Vacuum microelectronics: what's new and exciting. Electron Devices, IEEE Transactions on, 1991. 38(10): p. 2276-2283.

194. Han, J.-W. and M. Meyyappan, Introducing the Vacuum Transistor: A Device Made of Nothing. IEEE Spectrum, 2014.

195. Stoner, B.R. and J.T. Glass, Nanoelectronics: Nothing is like a vacuum. Nat Nano, 2012. 7(8): p. 485-487.

196. MEMSCAP. PolyMUMPs. [cited 2015; Available from: http://www.memscap.com/products/mumps/polymumps.

197. Natarajan, S., C.B. Parker, J.T. Glass, J.R. Piascik, K.H. Gilchrist, C.A. Bower, and B.R. Stoner, High voltage microelectromechanical systems platform for fully integrated, on-chip, vacuum electronic devices. Applied Physics Letters, 2008. 92(22): p. 224101.

198. Gilchrist, K.H., J.R. Piascik, B.R. Stoner, E.J. Radauscher, J.J. Amsden, C.B. Parker, and J.T. Glass. Platform for integrated vacuum microelectronic circuits. in Vacuum Electronics Conference, IEEE International. 2014.

214

199. Radauscher, E., A. Keil, M. Wells, J. Amsden, J. Piascik, C. Parker, B. Stoner, and J. Glass, Chemical Ionization Mass Spectrometry Using Carbon Nanotube Field Emission Electron Sources. Journal of The American Society for Mass Spectrometry, 2015. 26(11): p. 1903-1910.

200. Baker, R.T.K., P.S. Harris, R.B. Thomas, and R.J. Waite, Formation of filamentous carbon from iron, cobalt and chromium catalyzed decomposition of acetylene. Journal of Catalysis, 1973. 30(1): p. 86-95.

201. Chhowalla, M., K.B.K. Teo, C. Ducati, N.L. Rupesinghe, G.A.J. Amaratunga, A.C. Ferrari, D. Roy, J. Robertson, and W.I. Milne, Growth process conditions of vertically aligned carbon nanotubes using plasma enhanced chemical vapor deposition. Journal of Applied Physics, 2001. 90(10): p. 5308-5317.

202. Hofmann, S., M. Cantoro, B. Kleinsorge, C. Casiraghi, A. Parvez, J. Robertson, and C. Ducati, Effects of catalyst film thickness on plasma-enhanced carbon nanotube growth. Journal of Applied Physics, 2005. 98(3): p. 034308.

203. Lee, Y.T., J. Park, Y.S. Choi, H. Ryu, and H.J. Lee, Temperature-Dependent Growth of Vertically Aligned Carbon Nanotubes in the Range 800−1100 °C. The Journal of Physical Chemistry B, 2002. 106(31): p. 7614-7618.

204. Kim, K.-E., K.-J. Kim, W.S. Jung, S.Y. Bae, J. Park, J. Choi, and J. Choo, Investigation on the temperature-dependent growth rate of carbon nanotubes using chemical vapor deposition of ferrocene and acetylene. Chemical Physics Letters, 2005. 401(4–6): p. 459-464.

205. Liu, K., K. Jiang, C. Feng, Z. Chen, and S. Fan, A growth mark method for studying growth mechanism of carbon nanotube arrays. Carbon, 2005. 43(14): p. 2850-2856.

206. Puretzky, A.A., D.B. Geohegan, S. Jesse, I.N. Ivanov, and G. Eres, In situ measurements and modeling of carbon nanotube array growth kinetics during chemical vapor deposition. Applied Physics A, 2005. 81(2): p. 223-240.

207. Vinten, P., J. Lefebvre, and P. Finnie, Kinetic critical temperature and optimized chemical vapor deposition growth of carbon nanotubes. Chemical Physics Letters, 2009. 469(4–6): p. 293-297.

208. Zhu, L., J. Xu, F. Xiao, H. Jiang, D.W. Hess, and C.P. Wong, The growth of carbon nanotube stacks in the kinetics-controlled regime. Carbon, 2007. 45(2): p. 344-348. 215

209. Hofmann, S., C. Ducati, J. Robertson, and B. Kleinsorge, Low-temperature growth of carbon nanotubes by plasma-enhanced chemical vapor deposition. Applied Physics Letters, 2003. 83(1): p. 135-137.

210. Zhong, T. Iwasaki, J. Robertson, and H. Kawarada, Growth Kinetics of 0.5 cm Vertically Aligned Single-Walled Carbon Nanotubes. The Journal of Physical Chemistry B, 2007. 111(8): p. 1907-1910.

211. Noda, S., K. Hasegawa, H. Sugime, K. Kakehi, Z. Zhang, S. Maruyama, and Y. Yamaguchi, Millimeter-thick single-walled carbon nanotube forests: Hidden role of catalyst support. Japanese journal of applied physics, 2007. 46(5L): p. L399.

212. Puretzky, A.A., G. Eres, C.M. Rouleau, I.N. Ivanov, and D.B. Geohegan, Real- time imaging of vertically aligned carbon nanotube array growth kinetics. Nanotechnology, 2008. 19(5): p. 055605.

213. Zhu, L., D.W. Hess, and C.-P. Wong, Monitoring Carbon Nanotube Growth by Formation of Nanotube Stacks and Investigation of the Diffusion-Controlled Kinetics. The Journal of Physical Chemistry B, 2006. 110(11): p. 5445-5449.

214. Wirth, C.T., C. Zhang, G. Zhong, S. Hofmann, and J. Robertson, Diffusion-and reaction-limited growth of carbon nanotube forests. ACS nano, 2009. 3(11): p. 3560-3566.

215. Louchev, O.A., Transport-kinetical phenomena in nanotube growth. Journal of Crystal Growth, 2002. 237–239, Part 1: p. 65-69.

216. Louchev, O.A., T. Laude, Y. Sato, and H. Kanda, Diffusion-controlled kinetics of carbon nanotube forest growth by chemical vapor deposition. The Journal of Chemical Physics, 2003. 118(16): p. 7622-7634.

217. Wang, L., T. Chen, T. Feng, Y. Chen, W. Que, L. Lin, and Z. Sun, Effect of sputtered Cu film’s diffusion barrier on the growth and field emission properties of carbon nanotubes by chemical vapor deposition. Applied Physics A, 2007. 90(4): p. 701-704.

218. Sharma, H., A.K. Shukla, and V.D. Vankar, Effect of titanium interlayer on the microstructure and electron emission characteristics of multiwalled carbon nanotubes. Journal of Applied Physics, 2011. 110(3): p. 033726.

216

219. García-Céspedes, J., S. Thomasson, K.B.K. Teo, I.A. Kinloch, W.I. Milne, E. Pascual, and E. Bertran, Efficient diffusion barrier layers for the catalytic growth of carbon nanotubes on copper substrates. Carbon, 2009. 47(3): p. 613-621.

220. Stoner, B.R., A.S. Raut, B. Brown, C.B. Parker, and J.T. Glass, Graphenated Carbon Nanotubes for Enhanced Electrochemical Double Layer Capacitor Performance. Applied Physics Letters, 2011. 99(18): p. 183104.

221. Parker, C.B., A.S. Raut, B. Brown, B.R. Stoner, and J.T. Glass, Three- Dimensional Arrays of Graphenated Carbon Nanotubes. Journal of Materials Research, 2012. 27(07): p. 1046-1053.

222. Ubnoske, S.M., A.S. Raut, B. Brown, C.B. Parker, B.R. Stoner, and J.T. Glass, Perspectives on the Growth of High Edge Density Carbon Nanostructures: Transitions from Vertically Oriented Graphene Nanosheets to Graphenated Carbon Nanotubes. J. Phys. Chem. C, 2014. 118(29): p. 16126-16132.

223. Ubnoske, S.M., A.S. Raut, C.B. Parker, J.T. Glass, and B.R. Stoner, Role of Nanocrystalline Domain Size on the Electrochemical Double-Layer Capacitance of High Edge Density Carbon Nanostructures. MRS Commun., 2015. 5(02): p. 285-290.

224. Ubnoske, S.M., Q. Peng, E.R. Meshot, C.B. Parker, and J.T. Glass, A Protocol for High Sensitivity Surface Area Measurements of Nanostructured Films Enabled by Atomic Layer Deposition of TiO2. J. Phys. Chem. C, 2015: p. Just Accepted Manuscript.

225. Brown, B., C.B. Parker, B.R. Stoner, and J.T. Glass, Growth of vertically aligned bamboo-like carbon nanotubes from ammonia/methane precursors using a platinum catalyst. Carbon, 2011. 49(1): p. 266-274.

226. ASTM D3359-09e2, Standard Test Methods for Measuring Adhesion by Tape Test. 2009: West Conshohocken, PA.

227. Lee, S.W., K.K. Kim, Y. Cui, S.C. Lim, Y.W. Cho, S.M. Kim, and Y.H. Lee, Adhesion test of carbon nanotube film coated onto transparent conducting substrates. Nano, 2010. 5(03): p. 133-138.

228. Fick, A., V. On liquid diffusion. Philosophical Magazine Series 4, 1855. 10(63): p. 30-39.

217

229. Deal, B.E. and A.S. Grove, General Relationship for the Thermal Oxidation of Silicon. Journal of Applied Physics, 1965. 36(12): p. 3770-3778.

230. Cui, H., Nucleation and Growth of Nanoscaled One-Dimentional Materials, in Applied and Materials Science. 2001, University of North Carolina at Chapel Hill: Chapel Hill.

231. Baker, R.T.L. and M.A. Barber, in Chemical Physics and Carbon, P.L. Walker and P.A. Thrower, Editors. 1978, Dekker: New York.

232. Geohegan, D.B., A.A. Puretzky, I.N. Ivanov, S. Jesse, G. Eres, and J.Y. Howe, In Situ Growth Rate Measurements and Length Control During Chemical Vapor Deposition of Vertically Aligned Multiwall Carbon Nanotubes. Applied Physics Letters, 2003. 83(9): p. 1851-1853.

233. Stadermann, M., S.P. Sherlock, J.-B. In, F. Fornasiero, H.G. Park, A.B. Artyukhin, Y. Wang, J.J. De Yoreo, C.P. Grigoropoulos, O. Bakajin, A.A. Chernov, and A. Noy, Mechanism and Kinetics of Growth Termination in Controlled Chemical Vapor Deposition Growth of Multiwall Carbon Nanotube Arrays. Nano Letters, 2009. 9(2): p. 738-744.

234. Bedewy, M., E.R. Meshot, H. Guo, E.A. Verploegen, W. Lu, and A.J. Hart, Collective Mechanism for the Evolution and Self-Termination of Vertically Aligned Carbon Nanotube Growth. The Journal of Physical Chemistry C, 2009. 113(48): p. 20576-20582.

235. Holstein, W.L., The Roles of Ordinary and Soret Diffusion in the Metal-Catalyzed Formation of Filamentous Carbon. Journal of Catalysis, 1995. 152(1): p. 42-51.

236. Askeland, D.R., The Science and Engineering of Materials. Second Edition ed. 1989, Boston: PWS-KINT Publishing Company.

237. Pathria, R.K., Statistical Mechanics. Second Edition ed. 1996, Oxford: Butterworth-Heinemann.

238. Youngquist, G.R., Symposium on Flow through Porous Media: Diffusion and Flow of Gases in Porous Solids. Industrial & Engineering Chemistry, 1970. 62(8): p. 52-63.

239. Scott, D.S. and F.A.L. Dullien, Diffusion of ideal gases in capillaries and porous solids. AIChE Journal, 1962. 8(1): p. 113-117.

218

240. Hines, A.L. and R.N. Maddox, Mass Transfer: Fundamentals and Applications. 1985, Englewood Cliffs, NJ: Prentice Hall.

241. Rostrup-Nielsen, J. and D.L. Trimm, Mechanisms of carbon formation on nickel- containing catalysts. Journal of Catalysis, 1977. 48(1): p. 155-165.

242. Baker, R.T.K., J.J. Chludzinski Jr, and C.R.F. Lund, Further studies of the formation of filamentous carbon from the interaction of supported iron particles with acetylene. Carbon, 1987. 25(2): p. 295-303.

243. Tuinstra, F. and J.L. Koenig, Raman Spectrum of Graphite. J. Chem. Phys., 1970. 53(3): p. 1126-1130.

244. Matthews, M.J., M.A. Pimenta, G. Dresselhaus, M.S. Dresselhaus, and M. Endo, Origin of dispersive effects of the Raman D band in carbon materials. Physical Review B, 1999. 59(10): p. R6585-R6588.

245. Lee, Y.H., S.G. Kim, and D. Tománek, Catalytic Growth of Single-Wall Carbon Nanotubes: An \textit{Ab Initio} Study. Physical Review Letters, 1997. 78(12): p. 2393-2396.

246. Henry, P.A., A.S. Raut, S.M. Ubnoske, C.B. Parker, and J.T. Glass, Enhanced electron transfer kinetics through hybrid graphene-carbon nanotube films. Electrochemistry Communications, 2014. 48(0): p. 103-106.

247. Choi, Y.C., Y.M. Shin, Y.H. Lee, B.S. Lee, G.-S. Park, W.B. Choi, N.S. Lee, and J.M. Kim, Controlling the diameter, growth rate, and density of vertically aligned carbon nanotubes synthesized by microwave plasma-enhanced chemical vapor deposition. Applied Physics Letters, 2000. 76(17): p. 2367-2369.

248. Ren, Z.F., Z.P. Huang, J.W. Xu, J.H. Wang, P. Bush, M.P. Siegal, and P.N. Provencio, Synthesis of Large Arrays of Well-Aligned Carbon Nanotubes on Glass. Science, 1998. 282(5391): p. 1105-1107.

249. Choi, H., J. Gong, Y. Lim, K.H. Im, and M. Jeon, Effects of the electrical conductivity and orientation of silicon substrate on the synthesis of multi-walled carbon nanotubes by thermal chemical vapor deposition. Nanoscale Research Letters, 2013. 8(1): p. 110.

250. Huang, Z., D. Wang, J. Wen, M. Sennett, H. Gibson, and Z. Ren, Effect of nickel, iron and cobalt on growth of aligned carbon nanotubes. Applied Physics A, 2002. 74(3): p. 387-391. 219

251. Bower, C., O. Zhou, W. Zhu, D.J. Werder, and S. Jin, Nucleation and growth of carbon nanotubes by microwave plasma chemical vapor deposition. Applied Physics Letters, 2000. 77(17): p. 2767-2769.

252. Bedewy, M., E.R. Meshot, and A.J. Hart, Diameter-dependent kinetics of activation and deactivation in carbon nanotube population growth. Carbon, 2012. 50(14): p. 5106-5116.

253. Bedewy, M. and A.J. Hart, Mechanical coupling limits the density and quality of self-organized carbon nanotube growth. Nanoscale, 2013. 5(7): p. 2928-2937.

254. Bedewy, M., E.R. Meshot, M.J. Reinker, and A.J. Hart, Population Growth Dynamics of Carbon Nanotubes. ACS Nano, 2011. 5(11): p. 8974-8989.

255. Bonard, J.-M., C. Klinke, K.A. Dean, and B.F. Coll, Degradation and failure of carbon nanotube field emitters. Physical Review B, 2003. 67(11): p. 115406.

256. Mahapatra, D.R., N. Sinha, and R. Melnik, Degradation and Failure of Field Emitting Carbon Nanotube Arrays. Journal of nanoscience and nanotechnology, 2011. 11(5): p. 3911-3915.

257. Collins, P. and P. Avouris, Multishell conduction in multiwalled carbon nanotubes. Applied Physics A, 2002. 74(3): p. 329-332.

258. Deng, J.-H., X.-G. Hou, L. Cheng, F.-J. Wang, B. Yu, G.-Z. Li, D.-J. Li, G.-A. Cheng, and S. Wu, Irradiation Damage Determined Field Emission of Ion Irradiated Carbon Nanotubes. ACS applied materials & interfaces, 2014. 6(7): p. 5137-5143.

259. Lahiri, I. and W. Choi, Interface control: A modified rooting technique for enhancing field emission from multiwall carbon nanotube based bulk emitters. Acta Materialia, 2011. 59(14): p. 5411-5421.

260. Ominami, Y., Q. Ngo, M. Suzuki, A.J. Austin, C.Y. Yang, A.M. Cassell, and J. Li, Interface characteristics of vertically aligned carbon nanofibers for interconnect applications. Applied physics letters, 2006. 89(26): p. 263114.

261. Lim, S.C., H.K. Choi, H.J. Jeong, Y.I. Song, G.Y. Kim, K.T. Jung, and Y.H. Lee, A strategy for forming robust adhesion with the substrate in a carbon-nanotube field-emission array. Carbon, 2006. 44(13): p. 2809-2815.

220

262. Brodie, I. and C.A. Spindt, The application of thin-film field-emission cathodes to electronic tubes. Applications of Surface Sceience, 1979. 2(2): p. 149-163.

263. Neidert, R.E., P.M. Phillips, S.T. Smith, and C.A. Spindt, Field emission triodes. Electron Devices, IEEE Transactions on, 1991. 38(3): p. 661-665.

264. Gaertner, G., Historical development and future trends of vacuum electronics. Journal of Vacuum Science & Technology B, 2012. 30(6): p. 060801.

265. Singh, L.A., G.P. Sanborn, S.P. Turano, M.L.R. Walker, and W.J. Ready, Operation of a Carbon Nanotube Field Emitter Array in a Hall Effect Thruster Plume Environment. Plasma Science, IEEE Transactions on, 2015. 43(1): p. 95- 102.

266. Yuchi, C., C. Haitian, G. Hui, L. Jia, L. Bilu, and Z. Chongwu, Review of carbon nanotube nanoelectronics and macroelectronics. Semiconductor Science and Technology, 2014. 29(7): p. 073001.

267. Nagao, M. and T. Yoshida, Fabrication of gated nano electron source for vacuum nanoelectronics. Microelectronic Engineering, 2015. 132: p. 14-20.

268. Iannazzo, S., A survey of the present status of vacuum microelectronics. Solid State Electronics, 1993. 36(3): p. 301-320.

269. Srisonphan, S., Y.S. Jung, and H.K. Kim, Metal-oxide-semiconductor field-effect transistor with a vacuum channel. Nat Nano, 2012. 7(8): p. 504-508.

270. Jin-Woo Han, J.S.O., M. Meyyappan, Vacuum nanoelectronics: Back to the future?—Gate insulated nanoscale vacuum channel transistor. Applied Physics Letters, 2012. 100(213505).

271. Fursey, G., Field Emission in Vacuum Microelectronics. Microdevices. 2005: Springer.

272. Meyyappan, J.-W.H.M., Introducing the Vacuum Transistor: A Device Made of Nothing. 2014: IEEE Spectrum.

273. Natarajan, S., C.B. Parker, J.T. Glass, C.A. Bower, K.H. Gilchrist, J.R. Piascik, and B.R. Stoner. High voltage MEMS platform for fully integrated, on-chip,

221

vacuum electronic devices. in Vacuum Electronics Conference, 2008. IVEC 2008. IEEE International. 2008.

274. Bower, C., D. Shalom, W. Zhu, D. Lopez, G.P. Kochanski, P.L. Gammel, and S. Jin, A micromachined vacuum triode using a carbon nanotube cold cathode. Electron Devices, IEEE Transactions on, 2002. 49(8): p. 1478-1483.

275. Grayzeck, D.E. ALICE - NSSDC/COSPAR ID: 2013-072F. 2014 8/26/14 [cited 2016 2/9/16]; Available from: http://nssdc.gsfc.nasa.gov/nmc/spacecraftDisplay.do?id=2013-072F.

276. Technology, A.F.I.o., AFIT’s first satellite slated to launch by end of year. 2013: www.afit.edu.

277. Fowler, R.H. and L. Nordheim, Electron Emission in Intense Electric Fields. Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, 1928. 119(781): p. 173-181.

278. Ohno, T., Energy Distribution of Electrons from Thermionic Cathodes with a Wehnelt Cylinder. Journal of the Physical Society of Japan, 1972. 32(6): p. 1587- 1594.

279. Chen, E.X., Z.E. Russell, J.J. Amsden, S.D. Wolter, R.M. Danell, C.B. Parker, B.R. Stoner, M.E. Gehm, J.T. Glass, and D.J. Brady, Order of Magnitude Signal Gain in Magnetic Sector Mass Spectrometry Via Aperture Coding. Journal of The American Society for Mass Spectrometry, 2015. 26(9): p. 1633-1640.

280. Russell, Z.E., E.X. Chen, J.J. Amsden, S.D. Wolter, R.M. Danell, C.B. Parker, B.R. Stoner, M.E. Gehm, D.J. Brady, and J.T. Glass, Two-Dimensional Aperture Coding for Magnetic Sector Mass Spectrometry. Journal of The American Society for Mass Spectrometry, 2014. 26(2): p. 248-256.

281. Snyder, D.T., C.J. Pulliam, Z. Ouyang, and R.G. Cooks, Miniature and Fieldable Mass Spectrometers: Recent Advances. Analytical Chemistry, 2016. 88(1): p. 2- 29.

282. Richard, R.A.S. and W. Steven, MEMS mass spectrometers: the next wave of miniaturization. Journal of Micromechanics and Microengineering, 2016. 26(2): p. 023001.

222

283. Kornienko, O., P.T.A. Reilly, W.B. Whitten, and J.M. Ramsey, Field-Emission Cold-Cathode EI Source for a Microscale Ion Trap Mass Spectrometer. Analytical Chemistry, 2000. 72(3): p. 559-562.

284. Velasquez-Garcia, L.F., B.L.P. Gassend, and A.I. Akinwande, CNT-Based MEMS/NEMS Gas Ionizers for Portable Mass Spectrometry Applications. Journal of Microelectromechanical Systems, 2010. 19(3): p. 484-493.

285. Tassetti, C.M., R. Mahieu, J.S. Danel, O. Peyssonneaux, F. Progent, J.P. Polizzi, X. Machuron-Mandard, and L. Duraffourg, A MEMS electron impact ion source integrated in a microtime-of-flight mass spectrometer. Sensors and Actuators B- Chemical, 2013. 189: p. 173-178.

286. Licklider, L., X.-Q. Wang, A. Desai, Y.-C. Tai, and T.D. Lee, A Micromachined Chip-Based Electrospray Source for Mass Spectrometry. Analytical Chemistry, 2000. 72(2): p. 367-375.

287. Chen, L.Y., L.F. Velasquez-Garcia, X. Wang, K. Teo, A.I. Akinwande, and Ieee, A micro ionizer for portable mass spectrometers using double-gated isolated vertically aligned carbon nanofiber arrays. 2007 Ieee International Electron Devices Meeting, Vols 1 and 2. 2007. 843-846.

288. Han, K., Y. Lee, D. Jun, S. Lee, K.W. Jung, and S.S. Yang, Field Emission Ion Source Using a Carbon Nanotube Array for Micro Time-of-Flight Mass Spectrometer. Japanese Journal of Applied Physics, 2011. 50(6).

289. Hauschild, J.P., E. Wapelhorst, and J. Müller, Mass spectra measured by a fully integrated MEMS mass spectrometer. International Journal of Mass Spectrometry, 2007. 264(1): p. 53-60.

290. Badman, E.R. and R. Graham Cooks, Miniature mass analyzers. Journal of Mass Spectrometry, 2000. 35(6): p. 659-671.

291. Sillon, N. and R. Baptist, Micromachined mass spectrometer. Sensors and Actuators B: Chemical, 2002. 83(1–3): p. 129-137.

292. Geear, M., R. Syms, S. Wright, and A. Holmes, Monolithic MEMS quadrupole mass spectrometers by deep silicon etching. Journal of Microelectromechanical Systems, 2005. 14(5): p. 1156-1166.

293. Freidhoff, C.B. and R.M. Young, Miniaturized mass filter. 1996, Google Patents.

223

294. J. A. Diaz, C.F.G., and W. R. Gentry, Sub-miniature E x B sector-field mass spectrometer. J. Amer. Soc. Mass Spect, 2001. 12: p. 619-632.

295. Syms, R.R.A., Advances in microfabricated mass spectrometers. Analytical and Bioanalytical Chemistry, 2008. 393(2): p. 427-429.

296. Tang, F., X. Wang, L. Zhang, and Z. Yan, Study on simulation and experiment of array micro Faraday cup ion detector for FAIMS. Science China-Technological Sciences, 2010. 53(12): p. 3225-3231.

297. Antila, J., Miniaturized Spectrometer Technologies, in American Workshop on Information Optics. 2010: Helsinki, Finland.

298. Das, A.N., J. Sin, D.O. Popa, and H.E. Stephanou. Design and Manufacturing of a Fourier Transform Microspectrometer. in Nanotechnology, 2008. NANO '08. 8th IEEE Conference on. 2008.

299. Riehl, P.S., K.L. Scott, R.S. Muller, R.T. Howe, and J.A. Yasaitis, Electrostatic charge and field sensors based on micromechanical resonators. Journal of Microelectromechanical Systems, 2003. 12(5): p. 577-589.

300. Zhu, Y., J. Lee, and A. Seshia, System-level simulation of a micromachined electrometer using a time-domain variable capacitor circuit model. Journal of Micromechanics and Microengineering, 2007. 17(5): p. 1059-1065.

301. Saito, Y., Carbon nanotube field emitter. Journal of Nanoscience and Nanotechnology, 2003. 3(1-2): p. 39-50.

302. Calderon-Colon, X., H.Z. Geng, B. Gao, L. An, G.H. Cao, and O. Zhou, A carbon nanotube field emission cathode with high current density and long-term stability. Nanotechnology, 2009. 20(32).

303. de Jonge, N. and J.M. Bonard, Carbon nanotube electron sources and applications. Philosophical Transactions of the Royal Society a-Mathematical Physical and Engineering Sciences, 2004. 362(1823): p. 2239-2266.

304. Milne, W.I., K.B.K. Teo, G.A.J. Amaratunga, P. Legagneux, L. Gangloff, J.P. Schnell, V. Semet, V. Thien Binh, and O. Groening, Carbon nanotubes as field emission sources. Journal of Materials Chemistry, 2004. 14(6): p. 933-943.

224

305. Golonka, L.J., Technology and applications of Low Temperature Cofired Ceramic (LTCC) based sensors and microsystems. Bullentin of the Polish Academy of Sciences Technical Sciences, 2006. 54(2): p. 221-231.

306. Peterson, K.A., R.T. Knudson, E.J. Garcia, K.D. Patel, M. Okandan, C.K. Ho, C.D. James, S.B. Rohde, B.R. Rohrer, F. Smith, L.R. Zawicki, and B.D. Wroblewski. LTCC in microelectronics, microsystems, and sensors. in Mixed Design of Integrated Circuits and Systems, 2008. MIXDES 2008. 15th International Conference on. 2008.

307. Plumlee, D., J. Steciak, and A. Moll. Development of a micro-nozzle and ion mobility spectrometer in LTCC. in Microelectronics and Electron Devices, 2004 IEEE Workshop on. 2004.

308. Van Amerom, F.H.W., A. Chaudhary, M. Cardenas, J. Bumgarner, and R.T. Short, MICROFABRICATION OF CYLINDRICAL ION TRAP MASS SPECTROMETER ARRAYS FOR HANDHELD CHEMICAL ANALYZERS. Chemical Engineering Communications, 2007. 195(2): p. 98-114.

309. Kopola, H.K., J. Lenkkeri, K. Kautio, A. Torkkeli, O. Rusanen, and T. Jaakola. MEMS sensor packaging using LTCC substrate technology. 2001.

310. Karioja, P., K. Kautio, J. Ollila, K. Ker, x00E, nen, M. Karppinen, V. Heikkinen, T. Jaakola, and M. Lahti. MEMS, MOEMS, RF-MEMS and photonics packaging based on LTCC technology. in Electronics System-Integration Technology Conference (ESTC), 2014. 2014.

311. Peterson, K.A., Patel, K. D., Ho, C. K., Rohde, S. B., Nordquist, C. D., Walker, C. A., Wroblewski, B. D., and Okandan, M., Novel Microsystem Applications with New Techniques in Low-Temperature Co-Fired Ceramics. International Journal of Applied Ceramic Technolog, 2005. 2(345-363).

312. Edmond de Hoffmann, V.S., Mass Spectrometry: Principles and Applications. 3 ed. 2007: John WIley & Sons, Ltd.

313. McConnell, A.D., S. Uma, and K.E. Goodson, Thermal Conductivity of Doped Polysilicon Layers. Journal of Microelectricalmechanical Systems, 2001. 10(3): p. 360-369.

314. Bihan, R.L., F. Karray, and B. Cambou, Grain size and resistivity of polycrystalline silicon films. Poly-Micro-Crystalline and Amourphous Semiconductors, Strasbourg, France, 1984. 225

315. Felton, J.A., G.D. Schilling, S.J. Ray, R.P. Sperline, M.B. Denton, C.J. Barinaga, D.W. Koppenaal, and G.M. Hieftje, Evaluation of a fourth-generation focal plane camera for use in plasma-source mass spectrometry. Journal of Analytical Atomic Spectrometry, 2011. 26(2): p. 300-304.

316. Scientific Instrument Services, S., FC-43 (Perfluorotributylamine) Calibration Compound for Mass Spectrometers. p. Catalog A65.

317. Hübschmann, H.-J., Handbook of GC/MS: Fundamentals and Applications. 2008, Wiley-VCH Verlag GmbH & Co. KGaA. p. 7-292.

226

Biography

Erich Justin Radauscher was born on August 28, 1984 in Brooklyn, NY. He moved to Florida when he was 4 and grew up in beautiful Sarasota/Bradenton Fl. He obtained his Bachelor’s degree in Electrical Engineering in May of 2012 from the University of South Florida, Tampa, FL. He then moved to NC to pursue his doctoral degree in Electrical and Computer Engineering at Duke University in June of 2012. He concurrently worked as a research assistant at RTI International, Engineering and Applied Physics Division throughout his time in the Ph.D. program. His research interests include investigation of material properties and applications involving thin film materials and micro/nano electronic technologies. Specifically, vacuum microelectronics, carbon based materials growth and applications in field emission, MEMS devices, and miniaturized instruments and sensors.

Publications:

1. Gilchrist, KH and Radauscher EJ; Piascik, JR; Stoner, BR; Amsden, JJ; Parker, CB; Glass, JT; (2014). “Platform for Integrated Vacuum Microelectronic Circuits.” IEEE International Vacuum Electronics Proceedings, IVEC. 155-156 2. Radauscher, EJ; Keil, AD; Wells, M; Amsden, JJ; Piascik, JR; Parker, CB; Stoner, BR; Glass, JT; (2015). “Chemical Ionization Mass Spectrometry Using Carbon Nanotube Field Emission Electron Sources.” Journal of The American Society for Mass Spectrometry. 26(9): 1633- 1640 3. Radauscher, EJ; Gilchrist, KH; Di Dona, ST; Russell, ZE; Piascik, JR; Parker, CB; Stoner, BR; Glass, JT; "Eliminating proximity effects and improving transmission in field emission vacuum microelectronic devices for integrated circuits," 2015 IEEE International Electron Devices Meeting (IEDM), Washington, DC, 2015, pp. 33.2.1-2.4. 4. Radauscher, EJ; Gilchrist, KH; Di Dona, ST; Russell, ZE; Piascik, JR; Amsden, JJ; Parker, CB; Stoner, BR; Glass, JT; “Improved Performance of Field Emission Vacuum Microelectronic Devices for Integrated Circuits,” (submitted for publication) 5. Ubnoske, SM; Radauscher, EJ; Meshot, ER; Parker, CB; Stoner, BR; Glass, JT; (2015- 2016). “Integrating Carbon Nanotube Forests into Polysilicon MEMS: Growth Kinetics, Mechanisms, and Adhesion.” (submitted for publication) 6. Radauscher, EJ; Amsden, JJ; Di Dona, ST; Herr, P; Gilchrist, KH; Hall, SD; Carlson, JB; Parker, CB; Stoner, BR; Glass, JT; “Low Temperature Co-fired Ceramic Packaged MEMS Micro-ion Source for Miniature Mass Spectrometry Applications” (submitted for publication)

227