Vela Russell Thesis

ELECTRICAL BREAKDOWN OF SF6
AT HIGH PRESSURE
by
RUSSELL VELA, B.S.E.E.
A THESIS
IN
ELECTRICAL ENGINEERING
Submitted to the Graduate Faculty

of Texas Tech University in
Partial Fulfillment of
the Requirements for
the Degree of
MASTER OF SCIENCE
IN
ELECTRICAL ENGINEERING
Approved
Andreas Neuber
Chairperson of the Committee
Hermann Krompholz
Accepted
John Borrelli
Dean of the Graduate School
August, 2007
© 2007
Russell Vela
Texas Tech University, Russell Vela, August 2007
ACKNOWLEDGEMENTS
I would like to thank the committee members for all of their time and guidance. I
would especially like to extend my gratitude to Dr. Andreas A. Neuber, the committee
chair, for his assistance and patience. When I have wavered under the weight of my own
ambitions, he has given me the support to succeed. I would also like to thank Dr.
Hermann Krompholz for his valuable advice and contributions to issues that arose
throughout this project. Many thanks also go to all of the Pulsed Power professors for
their interest, questions, and invaluable teachings.
Thanks are also owed to Dr. John Krile and Kim Morales for the passing of their
knowledge and experiences in this field unto me. John, know that as you begin to guide
the eager young minds of tomorrow, patience is a companion of wisdom. Thanks are also
owed to all the staff at the Center for Pulsed Power and Power Electronics at Texas Tech.
Specifically, I would like to thank Danny Garcia, Shannon Gray, Dino Castro, Elmer
Thornton, and Marie Byrd. Appreciation also goes to all of my fellow colleagues and
friends. I cannot give enough thanks for their support, humor, willingness to help and
most importantly friendship.
Thanks to my sisters, Vanessa and Valerie for showing the “old guy” their humor,
compassion and love during all of this. Lastly, I would like to thank my mother and father
for their love, patience and for teaching their children how to fly. I hope all of you will
take pride in knowing that you have helped me become the man I am today. I can only
hope our paths will cross again.
ii
TABLE OF CONTENTS
ACKNOWLEDGEMENTS................................................................................................ ii
ABSTRACT....................................................................................................................... vi
LIST OF TABLES............................................................................................................ vii
LIST OF FIGURES ......................................................................................................... viii
CHAPTER
1. INTRODUCTION .......................................................................................................... 1
1.1 Motivation............................................................................................. 2
2. BACKGROUND ............................................................................................................ 5
2.1 Electron Emission Mechanisms............................................................ 5
2.1.1 Field Electron Emission......................................................... 6
2.1.2 Photoelectric Emission........................................................... 8
2.1.3 Secondary Electron Emission ................................................ 9
2.2 Ionization and Electron Loss Processes.............................................. 11
2.2.1 Ionization Processes............................................................. 12
2.2.1.1 Impact Ionization .................................................. 12
2.2.1.2 Photoionization ..................................................... 15
2.2.2 Electron Loss Processes....................................................... 16
2.2.2.1 Diffusion................................................................ 16
2.2.2.2 Recombination ...................................................... 17
2.2.2.3 Electron Attachment.............................................. 18
2.3 Electrical Breakdown Mechanism ...................................................... 19
2.3.1 Townsend Mechanism ......................................................... 20
2.3.2 Streamer Formation ............................................................. 23
2.3.3 Secondary Electron Emission Avalanche ............................ 26
2.3.4. Paschen’s Law .................................................................... 27
2.3.5 Static Breakdown of SF6 ...................................................... 29
2.4 Surface Flashover................................................................................ 31
3. EXPERIMENTAL SETUP........................................................................................... 33
iii
3.1 Experimental Test Chamber ............................................................... 35

3.2 Charging System................................................................................. 37
3.2.1 Primary Setup....................................................................... 39
3.2.1.1 Power Supplies...................................................... 39
3.2.1.2 Marx Generator .................................................... 40
3.2.1.3 Magnetic Switch.................................................... 41
3.2.2 Secondary Setup................................................................... 44
3.2.2.1 Power Supplies...................................................... 44
3.2.2.2 Marx Generator .................................................... 45
3.3 Volume Breakdown UV Source ......................................................... 45
3.4 Environmental Control........................................................................ 46
3.5 Diagnostic Setup ................................................................................. 47
3.5.1 Current Sensors.................................................................... 47
3.5.2 Voltage Measurements......................................................... 48
3.5.3 Breakdown Event Imaging .................................................. 49
3.5.4 Fiber Optics.......................................................................... 51
3.5.4.1 Spectrograph......................................................... 55
3.6 Data Manipulation .............................................................................. 56
3.7 Electrode Design................................................................................. 57
3.7.1 Primary Setup Surface Flashover Design ............................ 58
3.7.2 Primary Setup Volume Breakdown Design......................... 60
3.7.3 Secondary Setup Surface Flashover Design ........................ 60
3.8 Procedure ............................................................................................ 61
3.8.1 Primary Setup Volume Breakdown Mode........................... 62
3.8.2 Primary Setup Dielectric Surface Flashover Mode ............. 64
3.8.3 Secondary Setup Surface Flashover..................................... 65
4. ELECTRICAL BREAKDOWN RESULTS................................................................. 67
4.1 Reference Waveforms......................................................................... 67
4.2 Volume Breakdown ............................................................................ 69
4.2.1 Volume Breakdown Images................................................. 69
iv
4.2.2 Volume Breakdown Spectroscopy Analysis........................ 73

4.2.3 Solid By-Products of Volume Breakdown .......................... 77
4.3 Dielectric Surface Flashover............................................................... 80
4.3.1 Surface Flashover Images .................................................... 80
4.3.2 Surface Flashover Spectroscopy Analysis ........................... 86
4.3.3 Post Surface Flashover Analysis.......................................... 89
5. CONCLUSION............................................................................................................. 96
5.1 Volume Breakdown in SF6 ................................................................. 96
5.2 Dielectric Surface Flashover in SF6 .................................................... 97
REFERENCES ............................................................................................................... 100
APPENDIX
A PASCHEN’S LAW WITH REFERENCE TO SF6 .................................................... 104
B SPECTRAL ANALYSIS OF PULSED VOLUME BREAKDOWN IN SF6 AT HIGH
PRESSURES................................................................................................................... 106
C PULSED DIELECTRIC SURFACE FLASHOVER IN AN SF6 ENVIRONMENT. 118
v
ABSTRACT
This project was conducted to assist the upgrade of the Z-machine at Sandia
National Laboratories. At Sandia, the 5.5 Megavolt Laser Triggered Gas Switch (LTGS)
developed surface flashover within the PMMA envelope. This surface flashover occurs
initially during the regular cascade switching process without affecting the switch’s
electrical performance. However, the switch will pre-fire in the successive shot. The
damage to the envelope has to be repaired before normal switch operation can resume.
One of the more probable failure modes is that the existing electric field on the surface of
the envelope with the addition of the UV from the volume breakdown event initiates
surface flashover.
A 340 kV system was designed which replicated conditions found in the LTGS
between two adjacent backbone electrodes. Besides measurement of flashover delay
times for different dielectric materials (Lexan, Acrylic, Epoxies, etc.), emphasis was put
on the detailed characterization of the flashover phenomenology. Hence, in addition to
voltage and current diagnostics, fast optical imaging of the flashover along with the
acquisition of optical emission spectra (~200 nm to 700 nm) was employed.
This paper will investigate individually the two events that are occurring within
the chamber, surface flashover and volume breakdown. The conclusions will serve to lay
the groundwork for experiments to follow this paper which will directly investigate the
impact of volume breakdown on flashover across surfaces in the vicinity of the volume
breakdown in high pressure SF6. Understanding the fundamental physical mechanisms
and the extent to which these processes influence discharge behavior is vital to
characterizing and modeling electrical breakdown in high pressure SF6.
vi
LIST OF TABLES
2.1: Common insulators used in experimentation and their corresponding work

function, eΦ [4]........................................................................................................6
2.2: Variables for Fowler-Nordheim equation.....................................................................8
2.3: Variables for photoelectric emission equation 2.2........................................................9
2.4: Variables for the Gerald Dionne model......................................................................10
2.5: Summation of typical collision processes, where A, B are atoms, A*, B* are
excited atoms, A+, B+ are positive ions, e is an electron, A+, B+, e are ions and
electron with appreciable kinetic energy , ν is the frequency, and h is Planck’s
constant [12]...........................................................................................................11
2.6: Variables for equation to solve Townsend’s first ionization coefficient ....................15
2.7: Variables for describing electron amplification in Equation 2.10 ..............................21
2.8: Variables for describing current due to electron amplification, Equation 2.11 ..........21
3.1: Overall delivery system component characteristics....................................................38
vii
LIST OF FIGURES
1.1: LTGS design [2]. Purple wave indicates light emission or UV....................................3
1.2: Example of normal operation of backbone electrodes (right) [2].................................3
1.3: Example of failure in backbone electrode section [2] ..................................................4
2.1: Illustration of the modified potential barrier due to an applied field [5] ......................7
2.2: Graph of total SEE coefficient versus the energy of incident electron.......................10
2.3: Impact ionization cross section for oxygen (red) and nitrogen (black) [14] ..............13
2.4: Total electron-impact ionization cross section for SF6 as a function of electron
energy [13] ............................................................................................................14
2.5: Selected ionization constants and ranges of applicability. T = 20 ºC. [15] ................15
2.6: Total photodissociation σpdis,t(λ), and total photoionization σpi,t(λ), cross sections
of SF6 as a function of photon wavelength [13] ....................................................16
2.7: Recombination coefficient (ion-ion) in air at 20 ºC [15]............................................18
2.8: Direct and dissociative attachment cross section of SF6 as a function of energy
[17].........................................................................................................................19
2.9: Current vs. voltage relationship developed by Townsend [15] ..................................20
2.10: Diagram of the electron avalanche structure and effect of space charge [20]..........22
2.11: Diagram of the stages of streamer formation [5]......................................................24
2.12: The transition of a Townsend (generation) type of breakdown to a streamer

breakdown [5] ........................................................................................................25
2.13: Diagram of secondary electron emission avalanche.................................................26
2.14: Measured and calculated Paschen curves for atmospheric air [20] ..........................28
2.15: Minimum breakdown voltages for various gases [15]..............................................29
viii
2.16: Effective ionization coefficient (α- η)/p and weight function 1 – (η/α) describing
the probability that an electron creates further avalanche electrons, for SF6
[20].........................................................................................................................30
2.17: (α-η)/p ~ E/p relationship in SF6 [15] .......................................................................31
3.1: Experimental setup for pulsed volume and surface flashover testing ........................34
3.2: Overall circuit diagram of testing setup......................................................................34
3.3: Secondary setup surface flashover electrode configuration .......................................35
3.4: View of chamber internals with both primary and secondary gaps installed .............36
3.5: Example of surface flashover to chamber wall...........................................................36
3.6: Maxwell 3D simulations showing original field enhancements (left) at chamber

entrance and the field reductions (right) after improved feedthrough design.
Applied voltage = 320 kV......................................................................................37
3.7: Charging setup of overall system................................................................................38
3.8: Primary setup 320 kV Marx generator design. 2.7 nF Capacitors (2 in parallel per
stage), 6 kΩ per stage and 1 MΩ on HV charging.................................................40
3.9: Magnetic switch core configuration and center bobbin dimension ............................42
3.10: Final magnetic switch assembly in a reinforced oil vessel.......................................43
3.11: T-junction connecting magnetic switch to main chamber path, 20 psig SF6 ............44
3.12: Gas flow configuration for experimental setup ........................................................47
3.13: Cross-sectional rendering of a capacitive voltage divider ........................................49
3.14: Triggering setup used for testing ..............................................................................51
3.15: Absorption cross sections for molecular oxygen and sulfur hexafluoride [2]......... 52
3.16: Light transmission for molecular oxygen and SF6 through 2.75 inches of gas at a
pressure of 40 psig [2]................................................................................52
3.17: Fiber optics setup for spatially resolving light emission in the volume gap.............53
3.18: Fiber optics setup for spatially resolving light emission in the flashover gap..........53
ix
3.19: Top view of fiber optic lens layout...........................................................................54
3.20: Demonstration of difference in imaging and non-imaging spectrographs,

imaging spectrographs have specially designed mirrors to maintain vertical
resolution................................................................................................................55
3.21: Schematic of multi-track spectrograph operation.....................................................56
3.22: Ring electrode geometry in the LTGS at Sandia National Laboratories [2] ............57
3.23: Dielectric sample geometry and dimensions used for testing...................................58
3.24: Drawing of angled electrodes used for surface flashover/erosion experiments .......59
3.25: Maxwell 3D simulation of electric field magnitude for the angled electrode
geometry in a cut plane normal to the dielectric surface with 72 kV applied .......59
3.26: Electrodes used for volume breakdown in the high pressure chamber. The
electrodes are designed to mimic outside edge of the “vertebrae” in the z20
rimfire switch .........................................................................................................60
3.27: Secondary setup electrode configuration..................................................................61
3.28: Side view of camera positions with respect to breakdown event .............................63
3.29: Overall primary setup layout. Gas and charge lines are excluded............................64
3.30: Overall primary and secondary setups in dual firing. Gas and charging lines
excluded .................................................................................................................66
4.1: Representative set of waveforms typical of both surface flashover and volume
breakdown ..............................................................................................................68
4.2: Time integrated image of volume breakdown (~ 1 kA) between two stainless
steel electrodes in SF6 at 10 torr. ND = 0. Charged electrode is on the left ..........70
4.3: Time integrated image of volume breakdown (~ 2 kA) between two stainless
steel electrodes in SF6 at 20 psig. ND = 2. Charged electrode is on the left .........71
4.4: Time integrated image of volume breakdown (~ 10.2 kA) between two stainless
steel electrodes in SF6 at 20 psig. ND = 4. Charged electrode is on the left .........71
4.5: Front view of time integrated image of volume breakdown (~ 10.2 kA) between
two stainless steel electrodes in SF6 at 20 psig. ND = 4. Charged electrode is
on the left ...............................................................................................................72
x
4.6: Front view of ICCD image of volume breakdown (~ 10.2 kA) between two
stainless steel electrodes in SF6 at 20 psig. ND = 4. Charged electrode is on
the left ....................................................................................................................72
4.7: Volume breakdown voltage and delay time of a 9.52 mm gap in SF6 at
increasing pressures with ~ 50 ns risetime voltage pulse. Delay time is
defined as the FWHM from the rise of the voltage pulse to the drop due to
breakdown ............................................................................................................73
4.8: Spectra collected from a volume discharge in SF6 at high pressures. Spectra
gathered from low current breakdowns .................................................................75
4.9: Scaling curve for all collected spectra, accounting for internal loss in the collection
apparatus, spectrograph, and imaging camera .......................................................75
4.10: Spectra collected from a volume discharge in SF6 at high pressures and at various
current levels. Spectra gathered from higher current breakdowns ........................76
4.11: Spectra comparison of low current and high current breakdowns. High current
is approximately 10.4 kA, and low current is roughly 1 kA. Low current
breakdown at high pressures still maintains black body radiation pattern ............77
4.12: White powder deposit noticeable within the testing chamber. Powder is
prominent on the electrode holder, corona ring and within the RG220 chamber
entrance ..................................................................................................................78
4.13: Hair like fibers found on the electrode surfaces with electrostatic effect. Metal
fluorine deposits can be seen by the orange region ...............................................80
4.14: Time integrated image of surface flashover of a Lexan (polycarbonate) surface

in SF6 at 10 torr. Charged electrode is on the left. Low current (< 2 kA)
discharge ................................................................................................................81
4.15: Time integrated image of surface flashover of a Lexan (polycarbonate) surface

in SF6 at 40 pisg. Low current (< 2 kA) discharge ................................................82
4.16: Time integrated image of high current surface flashover at 20 psig of SF6
on Lexan (polycarbonate) sample. ND = 4. High current (~ 10.2 kA)
discharge. Side view ..............................................................................................82
4.17: Time integrated image of high current surface flashover at 20 psig of SF6 on
Virgin Teflon sample. ND = 4. High current (~ 10.2 kA) discharge. Side view...83
xi
4.18: ICCD image of high current surface flashover at 20 psig of SF6 on Virgin
Teflon sample. ND = 3. High current (~ 800 A) discharge. 500 ns exposure
time. Top view .......................................................................................................83
Teflon sample. ND = 3. High current (~ 2 kA) discharge. 500 ns exposure
time. Top view .......................................................................................................83
Teflon sample. ND = 4. High current (~ 10.2 kA) discharge. 500 ns exposure
time. Top view .......................................................................................................84
Teflon sample. ND = 4. High current (~ 8.6 kA) discharge. 500 ns exposure
time. Top view .......................................................................................................84
4.22: Dielectric material comparison of the average time delays. Samples arranged
by field strength (asterisk indicates one sample only tested). Surface flashover
event at atmospheric SF6 and at low current (< 2 kA) breakdown. Only one
sample of Plexiglas ................................................................................................85
4.23: Dielectric material comparison of the average time delays. Samples arranged
by time delay. Atmospheric breakdown with low current (< 2 kA) and 20 psig
with high current (~10.2 kA) .................................................................................86
4.24: Comparison of spectra collected from a volume breakdown and a surface

flashover of a Lexan (polycarbonate) surface in an SF6 environment at 20 psig.
Both are low current breakdown events.................................................................87
4.25: Spectra collected from a surface flashover on Lexan in SF6 at 20 psig and at
various current amplitudes.....................................................................................88
4.26: Spectra collected from a surface flashover on Teflon in SF6 at 20 psig and at
various current amplitudes.....................................................................................88
4.27: Spectra comparison of low current and high current surface flashover events on
Lexan. High current is approximately 10.4 kA, and low current is roughly
1 kA. Low current breakdown at high pressures still maintains black body
radiation pattern .....................................................................................................89
4.28: Smooth surface dielectric sample 826/D400/KF-865 cured at 80ºC for 3 days
after 38 discharges (1 atm SF6, 30 µC) ..................................................................90
4.29: Smooth optical grade Lexan after 10 discharges (1 atm SF6, 30 µC).......................91
xii
4.30: SEM image of smooth surface dielectric sample 826/D400/KF-865 cured at

80ºC for 3 days after 38 discharges (1 atm SF6, 30 µC). Distinct difference
between flashover region and the unaltered Epoxy surface...................................91
4.31: SEM image from dielectric material 826/D400/KF-865 used for X-ray
fluoroscopy analysis (1 atm SF6, 30 µC) ...............................................................92
4.32: X-ray fluoroscopy results for crosshair location in Figure 4.30. X-ray
fluoroscopy indicated sulphur deposit ...................................................................92
4.33: Lexan image used for X-ray spectroscopy (40 psig SF6, 108 µC). Cross hair
denoted point at which elemental composition was performed.............................93
4.34: X-ray fluoroscopy results for crosshair location in Figure 4.32. X-ray fluoroscopy
indicates sulphur deposit........................................................................................93
4.35: Material surface degradation comparison. Lexan, Teflon, and Rexolite from left
to right respectively. Top row is high current discharges (301 mC), with
1, 6 and 2 surface discharges respectively. Bottom row is low current
discharges at 100 shots per sample (108 µC).........................................................95
B.1: Cross-sectional view of the LTGS with the laser triggered gap on the right
and the self-break cascading section on the left. The cascade section is where
envelope failures occur [2]...................................................................................107
B.2: Overview of the high pressure chamber (25.4 cm x 25.4 cm x 25.4 cm).
Additional electrodes and flashover surfaces are omitted for clarity ..................108
B.3: Current and voltage waveform of SF6 breakdown across a 9.52 mm gap at
40 psig. The voltage signal is not corrected for the voltage sensor droop...........109
B.4: Cross-section of the optical emission collection apparatus.....................................109
B.5: Measured spectral response of the TTU spectroscopy system to a broadband

Xe high pressure lamp for transmission through 12.7 cm of SF6 or N2,
Respectively .........................................................................................................110
B.6: Calculated light transmission for air (21% molecular oxygen) and SF6
through 70 mm of the denoted gas at a pressure of 40 psig.................................110
B.7: Scaling curve for all collected spectra, accounting for internal loss in
the collection apparatus, spectrograph, and imaging camera ..............................111
xiii
B.8: Time integrated image of volume breakdown between two stainless steel
electrodes in SF6 at 10 torr. No Neutral Density used. Charged electrode is
on the left .............................................................................................................112
electrodes in SF6 at 20 psig. Image attenuated by Neutral Density filter
of ND = 2. Charged electrode is on the left .........................................................112
electrodes in SF6 at 40 psig. Image attenuated by Neutral Density filter of
ND = 3. Charged electrode is on the left .............................................................112
B.11: Breakdown voltage and delay time of a 9.52 mm gap in SF6 at increasing
pressures with ~ 50 ns risetime voltage pulse. Delay time is defined as the
FWHM from the rise of the voltage pulse to the drop due to breakdown ...........113
B.12: Spectra peak identification using calculated atomic and diatomic spectra.
Measured spectra collected in SF6 environment at 100 torr ................................114
B.13: Optical emission spectra at 50 Torr with high power microwave excitation
at 2.85 GHz. Top – 10:1 mixture of SF6 and N2, bottom – control pure N2 ........114
B.14: Spectra collected from a volume discharge in SF6 at various pressures.

The 100 Torr spectrum exhibits more N2 contamination.....................................115
B.15: Spectra collected from a volume discharge in SF6 at intermediate pressures........115
B.16: Spectra collected from a volume discharge in SF6 at high pressures.....................116
C.1: Overview of the atmospheric chamber setup. The dotted lines represent
the removable outer conductor and the dashed lines represent fiber optic
cables....................................................................................................................119
C.2: Overview of the high pressure chamber (25.4 cm x 25.4 cm x 25.4 cm).
Additional electrodes and flashover surfaces are omitted for clarity ..................120
C.3: Cross-section of the optical emission collection apparatus...........................................121
C.4: Maxwell 3D simulation of electric field vector for the angled electrode geometry
in a cut plane normal to the dielectric surface. Left electrode charged to 72 kV,
right electrode grounded. Gap spacing of 7.87 mm..................................................121
C.5: Dielectric sample geometry and dimensions used for testing. All units in
Millimeters ..................................................................................................................122
xiv
C.6: Representative waveforms of current and voltage signals for flashover in

SF6 at 90.6 kPa. Dashed lines represent the beginning and end of the camera
Exposure .....................................................................................................................123
C.7: Grooved Teflon sample at 90.6 kPa of SF6. Gap = 10.53 mm, Tdelay = 92 ns,
Vpulse ≈ 91 kV. White lines indicate the location of the surface and
Electrodes.............................................................................................................123
C.8: Grooved Teflon sample at 90.6 kPa of SF6. Gap = 9.48 mm, Tdelay = 307.3 ns,
Vpulse ≈ 90 kV. White lines indicate the location of the surface and
Electrodes.............................................................................................................123
C.9: Smooth surface dielectric sample 826/D400/KF-865 after 38 discharges

(1 atm SF6, 30 µC). Similar to all epoxy mixtures tested. Gap distance
of ~9.9 mm...........................................................................................................124
C.10: Grooved Teflon after 10 discharges (1 atm SF6, 30 µC). No damage

visible other than groove machining. Gap distance of ~10.5 mm ......................125
C.11: SEM image of smooth surface dielectric sample 826/D400/KF-865

after 38 discharges (1 atm SF6, 30 µC). The flashover region is denoted
by the dark area in the center of the image which lacks significant
machining marks. The region appears to be altered compared to the
surrounding epoxy surface. Imaged area is ~2.5 mm wide .................................125
C.12: SEM image of smooth surfaced optical grade polycarbonate. View is

into the hole for the electrode, the upper surface is the where the flashover
occurred. The cracks visible in the surface are located on the edge of the
surface corresponding to the electrode/dielectric interface. Discharges
occurred 40 times at eight various pressures
(Varied pressure SF6, 108 µC).............................................................................126
C.13: SEM Image from dielectric material 826/D400/KF-865 used for X-Ray
Spectroscopy analysis (90.6 kPa SF6, 30 µC) ......................................................126
C.14: Dielectric material comparison of average time delays. Samples arranged

by field strength (asterisk indicates one sample only tested)...............................127
C.15: Spectra collected from a surface flashover of a polycarbonate surface in an

SF6 environment at 6.7 kPa. Spectra are collected from three points along
the discharge path, as denoted ....................................................................................128
C.16: Spectra peak identification using calculated atomic spectra. Measured

spectra collected in SF6 environment, across a polycarbonate sample, at
6.7 kPa. The calculated spectra are given for a temperature of 30,000 K ...........128
xv
C.17: Calculation of remaining spectra after identified atomic spectra lines

are removed. Measured spectra collected in SF6 environment, across a
polycarbonate sample, at 6.7 kPa. The calculated spectra is given for
an temperature of 30,000 K .................................................................................129
C.18: Spectra collected from a surface flashover of a polycarbonate surface in

an SF6 environment at various pressures. Spectra are collected from
middle of the gap at each noted pressure .............................................................129
C.19: Comparison of spectra collected from a volume breakdown and a surface

flashover of a polycarbonate surface in an SF6 environment at 365.4 kPa.
Spectra are collected from middle of the gap in both cases.................................130
xvi
CHAPTER 1
1. INTRODUCTION
With the new requirements in switching voltage, the present Z-switch design is
deemed inadequate due to switch failure occurring after only a few shots. The primary
failure mode in the Laser Triggered Gas Switch (LTGS) is surface flashover in the high
pressure gas (SF6, > 40 psig) on the inside of the PMMA envelope. This surface flashover
occurs initially during the regular cascade switching process without affecting the
switch’s electrical performance. However, the switch will pre-fire in the successive shot.
The damage to the envelope has to be repaired before normal switch operation can
resume.
In the following experimental results of pulsed surface flashover, different
dielectric materials and volume breakdowns in SF6, primarily at high pressure, are
presented. Results of flashover and volume breakdown in SF6 at pressures from 10 Torr
to atmospheric pressure can be found in Appendix B. Besides fast voltage and current
monitoring of the breakdown event, an increased emphasis was put on imaging the event,
as well as, gathering optical emission spectra (~200 nm to 700 nm) from it. The role of
UV emission due to a volume SF6 arc, potentially a flashover hold-off reduction cause, is
discussed. The small scale experiments were designed to reproduce at least part of the
conditions as they are found in the large 5.5 MV Z-switch.
1
1.1 Motivation
The motivation for this project was to assist in the upgrade of the LTGS for the Z-
machine at SNL. A flashover event in high pressure SF6 (~40 psig) frequently occurred
on the inside of the PMMA envelope when the switch was pushed to a voltage exceeding
5.5 MV. This appeared to be one possible switch failure mode, and as such, warranted an
investigation. Identifying the underlying physics of surface flashover in the switch
environment; plus, the determining of key processes was necessary. The Center for
Pulsed Power and Power Electronics at Texas Tech University has performed many
experiments on surface flashover behavior over the past two decades. Recent research has
investigated the effects of UV on a surface flashover event [1, 38].
An example of the LTGS at SNL can be seen in Figure 1.1. First, the trigger gap
of the LTGS is broken down by a laser. This over-voltages the backbone electrodes, thus
causing a cascaded breakdown over the backbone electrodes, see Figure 1.2. In failure
mode, an additional breakdown path develops across the surface of the PMMA envelope,
see Figure 1.3. The hypothesis of why the surface flashover event was occurring
encompasses the entire system encased in the PMMA envelope. Dynamic electric field
simulations had shown that the electric field on the PMMA envelope is not equal to zero
during the closing process [39, 40]. This, plus the light and UV emitted from the volume
breakdown of the backbone electrodes, was thought to be enough to initiate the surface
flashover event on the envelope. SNL also requested surface flashover testing on various
materials. These tests were mainly concerned with surface erosion, material composition
changes, time delays and identifying possible dielectrics which exhibited better voltage
hold off than the PMMA.
2
Figure 1.1: LTGS design [2]. Purple wave indicates light emission or UV.
Figure 1.2: Example of normal operation of backbone electrodes (right) [2].
3
Figure 1.3: Example of failure in backbone electrode section [2]
In order to be able to identify how UV produced by a volume breakdown shapes a
surface flashover event, both events had to be investigated in an environment closely
mimicking the LTGS. This paper presents the foundations created for both volume
breakdown and surface flashover in high pressure SF6. Optically emitted spectra (~ 200
nm to ~700 nm), imaging of the event, fast current and voltage monitoring of surface
flashover and volume breakdown will be discussed.
4
CHAPTER 2
2. BACKGROUND
In order to better understand the processes involved in surface flashover, it
is helpful to look at the known underlying fundamental mechanisms. Surface flashover is
a complex situation, since the event contains a triple point junction. This is the point at
which the dielectric material, electrode and gas meet. To better understand surface
flashover, the junction can be investigated in two distinct categories. The first is from the
interaction of the electrode material and the dielectric which results in the electron
emission mechanisms. The second is from the gaseous electronics, which result in the
ionization and electron loss mechanisms. The other processes which will be discussed are
the Townsend Breakdown and streamer development.
2.1 Electron Emission Mechanisms
The emission of electrons from the surface is one dielectric surface flashover
event. The main processes for electron emission when dealing with dielectric surface
flashover are photoelectric and secondary electron emission. These processes are similar
due to the fact that they must overcome the work function, eΦ, of a particular material in
order to extract an electron. The electron work function, eΦ, is simply a measure of the
minimum energy required to extract an electron from the surface of a solid [3]. Insulators,
such as polycarbonate and nylon 66, have work functions of 4.26 eV and 4.08 eV,
respectively [4]. It should be important to note that eΦ is dependent on the cleanliness of
the surface and as such measurements reported in literature may often vary.
5
Table 2.1: Common insulators used in experimentation and their corresponding work
function, eΦ [4].
Polycarbonate (Lexan) 4.26 eV
Nylon 66 4.08 eV
Polytetrafluoroethylene (PTFE, or Teflon) 4.26 eV
Polystyrene “Rexolite” 4.22 eV
Polyimide “Kapton” 4.36 eV
Polyethylene Terephthalate “PVC” 4.85 eV
2.1.1 Field Electron Emission
Field electron emission is directly related to the electric field strength and is
relevant to the case at hand. For field emission to occur, the effective potential barrier at
the surface of the material must become sufficiently low. For an electron leaving the
surface, the field between can be approximated as a point charge and an equipotential
plane. The potential energy from this field is found using Coulomb’s law for the force on
the electron and then integrating the expression from ∞ to x [5]. The result yields the
potential barrier without an external field, see Figure 2.1, curve 1. Application of an
electric field to the cathode yields the dashed line in Figure 2.1, curve 2. Thus, combining
the energies represented by 1 and 2 produces a lowered potential energy curve, identified
as curve 3 in Figure 2.1.
6
Figure 2.1: Illustration of the effective potential barrier due to an applied field [5].
The modified potential barrier now has a height of finite value. However,
electrons in the conduction band, range between eΦ and Wa in Figure 2.1, still do not
have enough energy to overcome the barrier. Nevertheless, a quantum mechanical
phenomenon, called “tunneling”, allows or increases the probability that an electron will
be able to penetrate through a barrier of finite height and thickness [6].
The emission current due to the applied field is described by the Fowler-
Nordheim equation; see Equation 2.1 and Table 2.2 [5]. Currents of only a few
microamperes require field intensities on the order of 108-109 V/cm [7]. The currents at
lower fields are enhanced by microscopic protrusions and dielectric inclusions present on
the material surface. These factors can increase local electric fields to intensities
sufficiently high for field emission to occur [7]. The Fowler-Nordheim equation is
7
modified to include the influence of micro-protrusions by introducing β, the field
enhancement factor. Typical values for β range somewhere from 10 to several 100 for
vacuum environments [8] or from 1 to 250 for atmospheric air [9]. Thus, macroscopic
fields much lower than 108 and 109 V/cm, such as those in the 106 V/cm range, can lead
to significant field emission currents [7].
j= f⋅
(βE ) 2
⋅e
− g ⋅Φ 2 ⋅u
φβE
(2.1)
eΦ
Table 2.2: Variables for Fowler-Nordheim equation.

Variable (units) Variable Definition
2
j (A/m ) current density
E (V/m) macroscopic electric
field
f 1.54 x 10-6
g 6.83 x 109
eΦ (V) work function
u 1-√(1.4 x 10-9 βE/Φ2)
2.1.2 Photoelectric Emission
Photoelectric emission is the process by which electrons are released from
liberated from a material by irradiating it with photons. The energies of incident photons,
hν, are imparted upon the surface electrons. If the electron gains sufficient energy to
overcome the work function, photoelectric emission occurs. With an energy hν greater
than the work required to liberate an electron, eΦ, the remaining energy is found in the
form of kinetic energy in the liberated electron. The energy necessary for electron
emission (non-relativistic case) is expressed by [5]:
8
1 2
⋅ me ⋅ u e = hv = eΦ (2.2)
2
Table 2.3: Variables for photoelectric emission equation 2.2.

me (kg) mass of electron
ue (V/m) electron velocity
h (kg m2/s) Planck’s constant
v (1/s) photon frequency
eΦ (V) work function
A quantity used for the characterization of photoelectric emission is the
photoelectric yield γ. The photoelectric yield is the number of electrons per incident
photon. Significant photoelectric yield, ~10-2 to 10-1, typically occurs in the far UV or
below the 200 nm range [7]. A list of work functions for various materials is located in
Table 2.1. For a detailed explanation of photoemission see Reference [10].
2.1.3 Secondary Electron Emission
Secondary electron emission is the process of releasing a secondary electron from
a dielectric surface by impacting it with another electron called the primary. For every
primary electron incident on the dielectric’s surface there is a certain probability that a
secondary electron will be release from the surface. The yield, called δ, is the ratio of
secondary electrons emitted for every primary electron. An important factor in the release
of secondary electrons is the energy of the incident primary electrons, as well as, the
penetration depth of the primary electron. Other factors in determining the yield are the
material absorption coefficient and probability of escape. These factors are shown by the
Gerald Dionne model [11]:

9
1
⎛ B ⎞ ⎛ An ⎞
( )
n
⋅ (αd )
1
δ = ⎜⎜ ⎟⎟ ⋅ ⎜ ⎟ (n −1) ⋅ 1 − e −αd . (2.3)
⎝ξ ⎠ ⎝ α ⎠
Table 2.4: Variables for the Gerald Dionne model

δ SEE Yield
B Secondary escape probability
ξ Secondary-Electron Excitation Energy
A Primary Electron Absorption Constant
α Secondary-Electron Absorption Constant
n Power-Law Exponent
d Maximum Penetration Depth
The curve shown in Figure 2.2 shows the yield of secondary electrons vs. the
energy of the primary electrons. The curve has two “cross-over” points where the yield is
equal to 1. The reason that the curve has a maximum at a certain level is that the
electrons with too large an energy will penetrate deep into the surface and the secondary
electrons will have a low probability of reaching and escaping the surface.
Figure 2.2: Graph of total SEE coefficient versus the energy of incident electron
10
2.2 Ionization and Electron Loss Processes
Electrical breakdown cannot occur without the presence of ionization and
electron losses. In this experiment, a large contributor to ionization is the electric field
between the electrodes. This field serves to accelerate free electrons and thereby
increases the probability of impact ionization. The high field strength also can contribute
electrons to this process by means of field electron emission. Photoelectric emission is
another contributing factor due to UV radiation. It is essential to know exactly how
charged particles are created and possibly lost within a gas that normally consists of
neutral molecules only. Key types of ionization and de-ionization processes will be
presented. Typical collision processes is presented in Table 2.5.
Table 2.5: Summation of typical collision processes, where A, B are atoms, A*, B* are
excited atoms, A+, B+ are positive ions, e is an electron, A+, B+, e are ions and electron
with appreciable kinetic energy , ν is the frequency, and h is Planck’s constant [12].
A + e → A* + e Excitation by electron impact.
A + e ← A* + e A collision of the second kind.
A + e → A+ + e + e Ionization by electron impact.
A + hv → A* Photo excitation (absorption of light).
A + hv ← A* Emission of light.
A + hv → A + + e Photo-ionization.
A + hv ← A + + e Radiative recombination.
A + B → A* + B Excitation by atom impact.
A + B → A+ + e + B Ionization by atom impact.
11
A* + B → A + B * Excitation by excited atoms.
A+ + B → A + B + Change of charge.
A+ + B → A+ + B + + e Ionization by positive ion impact.
2.2.1 Ionization Processes
The process of liberating an electron from a gas particle with either the
simultaneous production of a positive ion, or with the increase of positive ion charge, is
called ionization. Electron impact ionization is the most prominent gaseous mechanisms
for increasing the number of liberated electrons within a gas. A sufficiently large applied
electric field is the cause of ionization by accelerating free electrons and increasing the
probability of collision and energy transfer with surrounding gas molecules. Free
electrons for ionization mechanisms are also provided from field and photoelectric
electron emission mechanisms, all of which contribute to final breakdown of the gas.
2.2.1.1 Impact Ionization
Electron impact ionization is a significant factor leading to breakdown in gases.
Impact ionization results from the collision between an electron and a neutral atom.
Dissociative ionization is a type of ionization dominant in molecular gases resulting in
the disassociation of the molecule due to an electron collision [5]. A symbolic
representation of impact ionization can be seen in Table 2.5.
Collisions may also free additional electrons or excite an atom or molecule to a
higher state. A key factor in impact ionization is the application of an external electric
field. In the presence of an applied field, electrons will experience a force parallel to the
12
field direction, thus accelerating free electrons. Electrons gain kinetic energy along the
mean free path while being accelerated, increasing the probability that a collision will
lead to ionization. The impact ionization cross sections for oxygen and nitrogen are
shown in Figure 2.3 and for comparison the total ionization cross section of SF6 is shown
in Figure 2.4.
Figure 2.3: Impact ionization cross section for oxygen (red) and nitrogen (black) [14].
13
Figure 2.4: Total electron-impact ionization cross section for SF6 as a function of
electron energy [13].
It is of great importance to quantify the number of ionizing collisions a single
electron undergoes over a certain distance, including the distance from the cathode to the
anode. The number of electrons per unit length in the field direction is the called the
ionization coefficient, α, or Townsend’s first ionization coefficient. The first ionization
coefficient, α, is also the number of ions produced due to the fact that each ionization
event produces a secondary electron and positive ion. The equation for α at high values
of E/p and moderate energy levels can be approximated to [5]:
14
− ⎛⎜ B p ⎞⎟
α = Ap e ⎝ E⎠
(2.4)
Table 2.6: Variables for equation to solve Townsend’s first ionization coefficient
α(cm-1) Townsend’s First Ionization Coefficient
A(cm-1·torr-1) Constant Related to Temperature
B(V/cm·torr) Constant Related to Type of Gas
p(torr) Atmospheric Pressure
E(V/m) Electric Field
Figure 2.5: Selected ionization constants and ranges of applicability. T = 20 ºC. [15]
2.2.1.2 Photoionization
Ionization may occur due to absorbtion of energy from either an external radiation
source or background radiation gas. In the case of an incident photon on a gas molecule,
ionization occurs if hν ≥ Wi, where Wi is the ionization energy of the gas molecule and hν
is the energy of the photon. The relationship can be expressed in the form:
ch
≥ Wi (2.5)
λ
Ultraviolet light, in the range of 125 nm, has been shown experimentally to have
sufficient energy to ionize a majority of gases [5]. Thus, any VUV emitted during the
15
breakdown process, or provided by an external UV source, will have a significant impact
on the importance of photoionization leading to flashover. The total photodissociation
and total photoionization cross sections for SF6 is shown in Figure 2.6
Figure 2.6: Total photodissociation σpdis,t(λ), and total photoionization σpi,t(λ),

cross sections of SF6 as a function of photon wavelength [13].
2.2.2 Electron Loss Processes
Diffusion, recombination and electron attachment are loss mechanisms that
reduce the number or energy of free electrons in the background gas. These processes
occur simultaneously along with electron amplification mechanisms, such as ionization.
Gaseous breakdown can only occur if the rate of ionization is greater than the rate of de-
ionization. Symbolic examples of de-ionization are seen in Table 2.5.
2.2.2.1 Diffusion
Diffusion is a process by which the particles of one type (in this case charged
particles) spread throughout a volume of particles of a second type (neutral particles)

16
until the concentration difference of both particle densities in the volume in the absence
of any external forces present, goes to zero. The primary factor contributing to diffusion
is the natural thermal chaotic motion of all particles in a gas. The process of diffusion
removes charges carrying particles from a potential breakdown channel.
The constant of proportionality between the rate of flow and the concentration
gradient is known as the diffusion coefficient, D, and is given by
1−−
D = λυ m2 /sec (2.6)
3
with υ as the average velocity and λ as the mean free path of the charged particles [5]. It
is important to note that D is proportional to p-1.
2.2.2.2 Recombination
Whenever there are positively and negatively charged particles present,
recombination takes place. The potential energy and the relative kinetic energy of the
recombining electron-ion is released as quantum of radiation. Symbolically the reaction
can be seen in Table 2.5.
At high pressures, ion-ion recombination takes place. The rate of recombination
in either case is directly proportional to the concentration of both positive and negative
ions. For equation concentrations of positive ions, n+ and negative ions n- the rate of
recombination
dn+ dn−
= = − βn+ n− (2.7)
dt dt
where β is a constant known as the recombination rate coefficient.
17
Since n+ ≈ n- = ni and if it assumed at time t = 0: ni = nio and at time t: ni = ni(t),
then equation (2.7) becomes
nio
ni (t ) = (2.8)
1 + nio βt
The variation of the recombination rate coefficient β with pressure in air is shown in
Figure 2.7. The recombination process is particularly important at high pressures for
which diffusion is relatively unimportant [15].
Figure 2.7: Recombination coefficient (ion-ion) in air at 20 ºC [15].
2.2.2.3 Electron Attachment
In electronegative gases (gases with unfilled outer electron shells) such as SF6, the
attachment of electrons to gas particles is a common deionization process. This type of
attachment is also known as dissociative electron attachment [5]. Like deionization,
attachment results in decreasing the amount of free electrons within the gas and reducing
18
the likelihood of breakdown. The probability of attachment [16], h, is sometimes given as
β, which is analogous to the first ionization coefficient α and is given by
υa
β= (2.9)
µE
with υa as the frequency of attachment and E as the applied field.
Figure 2.8: Direct and dissociative attachment cross section of SF6 as a function of
energy [17].
2.3 Electrical Breakdown Mechanism
Electrical breakdown is a colloquial term used to describe the process by which a
nonconducting medium becomes conductive through the application of a sufficiently
strong electric field. There exist many comprehensive literature sources describing the
characteristics of electrical breakdown mechanisms, including but not limited to books by
Nassar, Raizer, Meeks and Loeb [5, 7, 12, 18].The mechanisms leading to gaseous
19
breakdown, studied by Townsend, Raether, and Nasser, will be discussed in the following
sections.
2.3.1 Townsend Mechanism
As the voltage between electrodes in a gas increases, the electrode current at the
anode increase in accordance. Until at some point there is a sudden transition from dark
current to a self sustaining discharge, Figure 2.9. J.S. Townsend observed that an applied
voltage results in a linear increase to point V1. Past V1, the current remains constant until
a threshold point is reached V2. The increase in current beyond V2 Townsend ascribed to
ionization of the gas by electron collision. As the field increases, electrons leaving the
cathode are accelerated more and more between collisions until they gain enough energy
to cause ionization on collision with gas molecules or atoms [15].
Figure 2.9: Current vs. voltage relationship developed by Townsend [15].
This observation led Townsend to introduce a quantity known as Townsend’s first
ionization coefficient, α. The ionization coefficient expresses the number of ionizations

20
produced by an electron per unit length of path in the direction of the applied electric
field. Equation 2.4 is the empirical formula used to derive Townsend’s first ionization
coefficient, Section 2.2.1.1.
For the point beyond V2, Townsend developed Equation 2.10 to quantify the
number of electrons at distance x from the cathode in the direction of the applied field
[15]. A loss coefficient, η, accounts for deionization and attachment mechanisms leading
to a reduction in the number of unbound electrons. Equation 2.10 can also be expressed
in terms of the current leaving the cathode, I0, see Equation 2.11 [15].
n = no ⋅ e (α −η )⋅d (2.10)
Table 2.7: Variables for describing electron amplification in Equation 2.10.

n (m-3) electron density
no (m-3) electron density at cathode
α (m-1) Townsend’s first ionization
coefficient
η (m-1) loss coefficient
d (m) distance from cathode
I = I o ⋅ e (α −η )⋅d (2.11)
Table 2.8: Variables for describing current due to electron amplification, Equation 2.11.
I (A) current
Io (A) current leaving cathode
21
The exponential growth of electrons, expressed by Equations 2.10 and 2.11, is
called an electron avalanche. Avalanches develop in the direction opposite of the applied
electric field. High mobility electrons advance towards the anode leaving heavier positive
ions behind at the cathode [5]. The relatively immobile ions stay behind and form the
avalanche tail, while the mobile electrons form the spherical avalanche head, Figure 2.10.
Compared with the applied field, the field behind and ahead of the avalanche is increased
by the space charge and reduced between the electron and ion clouds. The variation in the
field around the avalanche is shown in Figure 2.10.
According to Raether, [19], UV light emitted in recombination and de-excitation
events creates electrons by photoionization ahead and behind the avalanche, initiating
further avalanches; which finally forms a conducting bridge between the anode and
cathode. Because of the higher fields at the ends of the avalanche, electrons created there
can cause ionization more efficiently. This kind of discharge is called the streamer
mechanism and will be discussed in the following section [20].
Figure 2.10: Diagram of the electron avalanche structure and effect of space charge [20].
22
2.3.2 Streamer Formation
As mentioned above, streamer formation is the event in which the channel is
formed between the anode and cathode via multiple electron avalanches. An explanation
of streamer formation is taken from Nasser [5] see Figure 2.11. In part (a) of Figure 2.11,
the developing avalanche creates a space charge on the order of the externally applied
electric field. The associated high amplitude fields allow the electrons in the head of the
avalanche to acquire higher energy, leading to increased UV emission. In part (b),
absorption of the UV radiation by neutrals leads to seed electrons for a second generation
of avalanches. As the positive space charge bridges the gap as shown in (c), the
secondary avalanches are attracted to the positive space charge trunk. In part (d), the
accumulation of positive ions at the cathode has increased by the addition of two
auxiliary streamers, and an ionization channel is forming between the anode and cathode.
Two subsequent streamer branches grow into incoming avalanches in (e) whereas one
dies out due to the lack of incoming avalanches in (f). If the process continues, the
streamer will eventually bridge the gap as shown in part (g), and the ionization channel
becomes the breakdown path.
23
Figure 2.11: Diagram of the stages of streamer formation [5].
24
The transition from a Townsend breakdown to a streamer breakdown is illustrated
in Figure 2.12. In part (a), an avalanche emits photons to produce photoelectrons at the
cathode. Part (b), the photoelectrons then produce avalanches of varying size with
electrons drifting at the same velocity. Next, positive space charge is accumulated at the
anode enhancing the field and augmenting the avalanche size to reach the critical size, in
part (c). Finally, a streamer grows from the avalanche toward the cathode by the
oncoming avalanches, part (d).
Figure 2.12: The transition of a Townsend (generation) type of breakdown to a streamer

breakdown [5].
25
2.3.3 Secondary Electron Emission Avalanche
Secondary Electron Emission Avalanches (SEEA) are one of the most widely
accepted mechanisms that contribute to surface flashover in vacuum [21]. In SEEA,
electrons are emitted from the surface of a dielectric via secondary emission due to the
impact of some primary ion or electron incident upon the surface. Once free from the
surface, the electrons experience acceleration parallel to the surface in the direction of the
applied electric field. The process of liberating electrons from the dielectric surface
leaves a net positive charge on the surface, and thus, a DC restoring field appears. This
restoring field accelerates the “freed” electrons back towards the surface. If the applied
field is large enough, the freed electrons will impact the surface of the dielectric with
enough energy to liberate more electrons, having gained this energy from the applied
electric field in the time of flight between departing from and returning to the surface.
The continuation of this process leads to a SEEA illustrated in Figure 2.13.
Figure 2.13: Diagram of secondary electron emission avalanche
An important parameter in the development of SEEA is the “hopping distance” of
the electrons, i.e. the lateral range of the electron trajectory [44]. For Lexan with a gap
distance of 1 cm and an applied breakdown field of 20 kV/cm, the average value for
26
hopping distance is 22 µm and about 2 µm for the maximum height above the surface.
Since the mean free path in SF6 at high pressures is much less than 22 µm (< 100 nm), an
SEEA will not develop under the conditions given by the described high pressure
flashover experiments.
2.3.4. Paschen’s Law
The second Townsend coefficient, γ, allows for a discharge to become self-
sustaining. In the case of SF6, γ is most likely due to photo emission, and is generally
represented by
γ (e αd − 1) = 1 (2.12)
meaning that each initial electron must produce at least one secondary electron to make
the discharge self-sustaining. When discussing the Townsend breakdown mechanism, it
can be shown that the αd in equation 2.12 is equivalent to ηVb [22, 16]. η is the ionization
proportionality constant related to the applied electric field and is given by:
α
η= (2.13)
E
Substituting these into equation 2.12 and rearranging shows that the breakdown voltage
can be written as:
Bpd
Vb = (2.14)
⎡ ⎤
⎢ ⎥
⎢ Apd ⎥
ln
⎢ ⎛ 1 ⎞⎥
⎢ ln⎜⎜1 + ⎟⎟ ⎥
⎣ ⎝ γ ⎠⎦
27
A and B values for various gases can be found in Figure 2.5. This method of
relating the breakdown voltage as a function of pd is known as Paschen’s Law. From
equation 2.14, it can be seen that at large and small values of pd, the breakdown voltage
Vb increases. Consequently, Vb must go to a minimum which is called the Paschen
minimum. Calculation of this minimum breakdown voltage gives:
⎛ 1⎞B
Vb (min) = 2.718 ln⎜⎜1 + ⎟⎟ (2.15)
⎝ γ⎠A
Similarly, the minimum value of pd is found to be:
⎛ 1⎞
ln⎜⎜1 + ⎟⎟
γ⎠
pd b min = 2.718 ⎝ (2.16)
A
With a voltage lower than the Vbmin, it is impossible to cause the breakdown of a gap with
a uniform distribution, no matter what the pressure or spacing [20].
Figure 2.14: Measured and calculated Paschen curves for atmospheric air [20].
28
Figure 2.15: Minimum breakdown voltages for various gases [15].
In electronegative gases, in which attachment is significant, the Paschen law holds
in general and specifically within a limited range around the breakdown minimum [23].
There are many literature sources for Paschen’s law with respect to SF6 [24, 25]. The
Paschen curve with respect to SF6 can be found in Appendix A.
2.3.5 Static Breakdown of SF6
To describe the breakdown properties of electronegative gases such as SF6, the
attachment coefficient η must be considered. η defines the attachment probability of an
electron per unit path length. Similarly to α/p, η/p is a unique function of E/p. The
effective ionization coefficient αe is defined as the difference between α and η:
⎛E⎞
α e = α − η = pf ⎜⎜ ⎟⎟ (2.17)
⎝ p⎠
29
Electron avalanche formation becomes possible only if αe is greater than 0. For SF6, this
is the case if E/p is greater than 8.84 kV/ (mm*bar). (α- η)/p can be expressed as a linear
function of E/p as seen in Figure 2.16 [20].
(α − η ) ⎡E ⎛ E ⎞ ⎤
= k ⎢ − ⎜⎜ ⎟⎟ ⎥ (2.18)
p ⎢⎣ p ⎝ p ⎠ o ⎦⎥
For SF6, k = 27.7 kV and (E/p)o = 8.84 kV/ (mm*bar). Using these parameters and the
linear relationship (2.18) to determine the static breakdown voltage Vb for a uniform field
the following can be obtained:
Vb = pd 8.84[kV / mm * bar ] + 0.5kV (2.19)
Figure 2.16: Effective ionization coefficient (α- η)/p and weight function 1 – (η/α)
describing the probability that an electron creates further avalanche electrons, for SF6
[20].
30
Figure 2.17: (α-η)/p ~ E/p relationship in SF6 [15].
2.4 Surface Flashover
Dielectric surface flashover in vacuum has been studied rather extensively and the
processes involved have been documented. It is upon these processes that the theoretical
basis for flashover at high pressures has been derived. The prevalent model divides this
type of surface flashover into three stages: (a) initiation, (b) growth and (c) gaseous
ionization along the surface of the dielectric [21].
It is generally accepted that the initiation is triggered by the field emission of
electrons at the cathode triple point [21, 26]. The triple point is a term which describes
the point at which the dielectric, electrode and vacuum intersect. The most commonly
accepted mechanism of the second stage is a SEEA, see Section 2.3.3. The final stage of
31
surface flashover in vacuum is generally accepted as an event which involves gaseous
ionization near the surface due to atoms or molecules desorbed by the SEEA [21].
Given these processes at the vacuum level, one could speculate that at higher
pressures, it is more probable that a discharge develops as a result of impact ionization
versus desorption as seen for vacuum flashover. In addition, since molecules are more
abundant at high pressures, the spark channel is not limited to just developing near the
dielectric surface.
32
CHAPTER 3
3. EXPERIMENTAL SETUP
As stated earlier, the experimental chamber accommodated volume breakdown,
surface flashover and a system that would allow for the testing of volume breakdown
effects on surface flashover. To develop this environment within the chamber, two
separate HV pulse systems were created as seen in Figure 3.1, 3.2. The primary system
allowed for a high voltage/high current discharge which could also be adjusted to
administer a high voltage low current setup. Interchangeable electrodes made both
volume and surface flashover tests possible. The secondary system was designed to be
used in testing the effects of an electric field on a surface and UV produced from a
volume gap. As such, the electrode configuration was strictly a surface flashover design,
see Figure 3.3.
33
Figure 3.1: Experimental setup for pulsed volume and surface flashover testing.
Figure 3.2: Overall circuit diagram of testing setup.
34
Figure 3.3: Secondary setup surface flashover electrode configuration.
3.1 Experimental Test Chamber
The experimental testing chamber was designed to house essentially two separate
electrode systems, allow for maximum accessibility and have visibility of the electrodes
for imaging, see Figure 3.4. The chamber is composed of approximately 0.1 inch
stainless steel and has external dimensions of 10x10x10 inches. The chamber has a 1 inch
Lexan (polycarbonate) lid which seals via an o-ring to the top flange. There is also a one
inch thick epoxy lid used for sealing the bottom of the chamber. A plastic compression
flange allows the secondary system to enter the chamber from the epoxy lid. For an added
viewport, the front of the chamber has a 1 in thick quartz window with a 3 inch diameter
and a top surface accuracy of λ 1 and a polished bottom surface from Edmund
10
Industrial Optics (Stock # 32640, Serial # 361). The chamber also has ports for the
vacuum system, SF6 pressure adjustment, over pressure valve, fiber optic lines for spectra
imaging and a relief system which will all be discussed in the following sections.
35
Figure 3.4: View of chamber internals with both primary and secondary gaps installed.
The electric field at the coaxial line chamber entrance caused surface flashover
issues from the electrodes to the chamber, see Figure 3.5.
Figure 3.5: Example of surface flashover to chamber wall.
To remedy the field enhancement found at the RG220 entrance, static electric field
simulations using Maxwell 3D and 2D were used to help minimize field strength, see
Figure 3.6. After many designs, a toroidal corona ring was designed which attached to the
chamber wall and a plastic compression flange was used to replace a previously existing
36
metal one. The combination of these two fixes through simulations showed a reduction in
electric field strength, see Figure 3.6.
Figure 3.6: Maxwell 3D simulations showing original field enhancements (left) at

chamber entrance and the field reductions (right) after improved feedthrough design.
Applied voltage = 320 kV.
3.2 Charging System
The experimental setup had two independent charging systems. The first was
used to charge the primary Marx generator, the HV capacitor and the trigger pulsers, see
Figure 3.7. The second system had only the secondary Marx generator to charge. Each of
which will be discussed thoroughly in their respective sections. For a list of component
specifications, see Table 3.1.
37
Figure 3.7: Charging setup of overall system.
Table 3.1: Overall delivery system component characteristics
Primary Marx Generator Erected capacitance: 675 pF

Charging capacitance: 43.2 nF
Maximum stored charge: 229.5 µC
Maximum voltage: 340 kV
Maximum energy: 39 J
Secondary Marx Generator Erected capacitance: 337.5 pF
Charging capacitance: 21.6 nF
Maximum stored charge: 114.8 µC
Maximum voltage: 340 kV
Maximum energy transfer: 19.5 J
HV Capacitor Voltage: 40 kV
Capacitance: 17 µF
Maximum stored energy: 13.6 kJ
Magnetic Switch Saturation determinant: 24.5 – 26 mVs
Saturation time @ 320 kV: 76 – 81 ns
Saturated inductance: < 17 µH
38
3.2.1 Primary Setup
The high voltage and high current that allowed for the breakdown of the primary
gap is supplied by a Marx generator and a high voltage capacitor, see Figure 3.2. First,
the Marx generator and high voltage capacitor are both charged. The Marx generator is
then triggered while the high voltage provided breaks down the primary gap. While the
arc is on, the magnetic switch saturates; reducing the inductance in the system causing a
decrease in voltage through the switch and an increase in current. This supplies the
desired high current to the breakdown event. The primary setup involves three major
components. The first, power supplies, contained all the high voltage power supply
systems for the components of the primary system, and the triggering system for the
primary Marx generator. The second will be the Marx generator used in breaking down
the main gap. The third will be the magnetic switch used in the delivery of the high
current.
3.2.1.1 Power Supplies
The primary setup had two identical high voltage power supplies. The high
voltage DC power supplies are from Hipotronics and each are capable of delivering up to
50 kV DC at a maximum of 5 mA. One, set to 40 kV DC, was used to charge the Marx
generator. The other was used to charge the high voltage capacitor to 10 kV.
Triggering of the high voltage Marx generator was accomplished via a solid state
high voltage pulse generator. An initial pulse of ~1.2 kV was generated by the pulser and
then input into a pulse transformer with a turn ratio of 1:44. The pulse generator had an
39
electronic trigger system which allowed it to be synchronized with the secondary Marx
generator.
3.2.1.2 Marx Generator
The voltage necessary to breakdown the primary electrodes was supplied by a 320
kV Marx generator, see Figure 3.8. Housed in an aluminum casing, the 8 stage Marx
generator had 5.2 nF capacitance per stage and a charging voltage of < 40 kV. Typically,
the pulse width was ~150 to 200 ns and had a rise time of ~80 to 100 ns. To monitor the
transient voltages a capacitive voltage-divider integrated onto the RG220 coaxial line was
used, see Section 3.5. If this Marx generator is charged with a positive voltage, the output
will be a negative voltage, see Figure 3.8.
Figure 3.8: Primary setup 320 kV Marx generator design. 2.7 nF Capacitors (2 in parallel
per stage), 6 kΩ per stage and 1 MΩ on HV charging.
40
3.2.1.3 Magnetic Switch
In the design presented above, a method of delivering a high amount of current
(10-50 kA) had to be designed. The scaled current amplitudes were chosen with respect
to the amplitude found in the LTGS (600 kA). It was decided that a charged high voltage
capacitor could deliver the current to a formed arc via some means of a closing switch.
Due to the high voltages involved in this design, the list of closing switches soon became
scarce. A magnetic switch was chosen as the solution to this problem since it could be
designed to handle both high voltages and currents, and still have a high lifetime [27, 28].
Using the fundamental equations for a magnetic switch, as given in equation 3.1, a basis
for a design which could hold off the 320 kV pulse for the duration of the arc lifetime
was formed. The formation time of an arc in SF6 at 20 psig was taken from experimental
waveforms dealing with SF6 at these pressures [29].
Setting the constants into the left side of equation 3.1 (a), results in what is called
a saturation determinant. The saturation determinant used in the design was a 24.5-26
mVs with a saturated inductance of less than 17 µH. This meant that the switch could
41
hold off a 320 kV pulse for 76-81 ns, which was more than sufficient for arc formation.
Three switches acting as one were used to lower the saturation time of the switch and the
total inductance. Two AMCC1000 Metglas c-cores were used due to their high magnetic
field saturation of 1.56 Teslas, see Figure 3.9. The core edges were grinded to a chamfer
to reduce the field strength. To avoid breakdown issues caused by the high voltage levels,
high density polyethylene was used to create bobbins for the AMCC1000 cores. Their
wire wrappings consisted of 15 kV DC 12 gauge corona resistant Teflon wire (CRT 12-3)
from Daburn. The outer bobbins were wrapped counterclockwise, while the center was
wrapped clockwise. This ensured that the magnetic fields flowed in a continuous pattern,
thus working together to saturate the cores. The bobbins were given a tapered effect on
each side to decrease the points at which the electric field magnitude was at its highest.
After wrapping the outer and inner bobbins with 22.5 turns of the CRT wire, the switch
was submerged in transformer oil and then out-gassed over night to insure maximum
voltage hold off of the oil, see Figure 3.10.
Figure 3.9: Magnetic switch core configuration and center bobbin dimension.
42
Figure 3.10: Final magnetic switch assembly in a reinforced oil vessel.
A pressurized T-style junction was developed in order to integrate the magnetic
switch into the setup, see Figure 3.11. The junction was pressurized because previous
experience at these high voltages had exhibited corona problems. The junction was made
of clear PVC and pressurized to 20 psig with SF6. Compression interconnects were
utilized to ensure a current carrying path.
43
Figure 3.11: T-junction connecting magnetic switch to main chamber path, 20 psig SF6.
3.2.2 Secondary Setup
The purpose of the secondary electrodes in the chamber is to be excited to the
point of breakdown, as well as, under breakdown. To accomplish this, a second high
voltage Marx generator was incorporated. The secondary surface flashover electrodes in
the chamber are excited by a separate system from the primary. Hence it can be
synchronized to the primary Marx generator. The secondary setup is very similar to the
primary setup, except that it is consistent of only a Marx generator and high voltage
power supply. Since only breakdown was required, a high current pulse was not
necessary.
3.2.2.1 Power Supplies
The Marx generator used in the secondary setup is charged by high voltage DC
power supply from Glassman High Voltage, Inc (model # PS/WR125P2.0-11) and is
capable of providing a DC voltage of up to 125 kV at a maximum current of 2 mA. A
44
high voltage resistor is put between the output of the power supply and the Marx
generator to limit current and to help protect against transients.
The triggering of the Marx generator is performed via a 1.2 kV solid state pulse
generator identical to the one described in section 3.2.1.1. The pulse transformer used for
this design has a ratio of 1:37. Both the primary and secondary setup pulse transformers
are wound on bobbins on AMCC100 Metglas cores. The primary windings consist of
0.1” diameter, 20 AWG wire with FEP insulation from Reynolds Industries coated wire.
The secondary is a quadruple build magnet wire 22 AWG from Magnetic Wire Co.
3.2.2.2 Marx Generator
The Marx generator used in exciting the secondary setup surface flashover
electrodes is essentially identical to the Marx generator used in the primary setup. The
physical arrangement is similar except that each stage has half the capacitance of the
primary Marx generator stages, see Table 3.1. Another varying feature is that this Marx
generator is triggered by a triple electrode configuration. Trigger differences between the
two Marx generators, pressure differences and closeness to the Marx generator self-break
point all cause variances in the time to Marx generator erection.
3.3 Volume Breakdown UV Source
For the experiments involving the testing of an applied UV source on a surface
flashover event, both breakdown gaps were used. The primary setup was put into the
volume breakdown configuration and a surface was placed underneath the electrodes on
the secondary setup. The high current volume arcs produced a larger contribution in the
45
UV region (≤320 nm) than previous low current experiments. The synchronization of the
two Marx generators was not completed at the time of thesis publication. Results will be
presented in a future report to Sandia.
3.4 Environmental Control
Since the voltage levels of the system were greater than the breakdown strength of
air, the gas sulphur hexafluoride (SF6) was chosen to be used as high voltage insulation.
The magnetic switch T-junction and both Marx generators where filled with SF6 (20/20/9
psig respectively) as a high voltage insulator. The testing chamber required SF6 for
experimenting in high pressure environments was equipped with gas lines, pressure
release valves and, as a safety precaution, an automatic over pressure release valve set to
50 psig, see Figure 3.12.
Since the chamber had an operating pressure ranging from 10-4 torr to 50 psig
(2586 torr), a vacuum pressure system was incorporated. A Varian roughing pump
(model # PTS03001UNIV, serial # LPE80462) was used to evacuate the chamber. Once a
set point was reached, the Varian turbo-V550 MacroTorr pump (model 969-9050, serial #
75046) was activated and pulled the chamber to the 10-4 torr range. The vacuum
pressures where recorded by three separate measurement devices. An MKS 999-Quattro
ion gauge measured pressures low 10-4 torr pressures while two MKS BARATRON
gauges recorded 0-100 torr and 100-1000 torr respectively.
46
Figure 3.12: Gas flow configuration for experimental setup.
3.5 Diagnostic Setup
Since the breakdown processes occur in the nanosecond regime, instrumentation
with a high temporal resolution is required. In addition to a high temporal resolution,
consideration had to be placed on the physical placement of the measurement devices in
order to minimize the signal reflections and distortions. The signals used for
characterizing both the surface flashover event and the volume breakdown are the current
and voltage waveforms. Three different styles of probes where chosen and used to obtain
these signals.
3.5.1 Current Sensors
Originally the current measurements were obtained from a traveling wave current
sensor incorporated into the RG220 transmission line entering the testing chamber. When
dealing with signals in the nanosecond regime, the current sensor produced minimal
signal distortion and proved suitable for the measuring of the current signals produced at
47
both low current breakdown events with a resolution of 0.1 V/A [30]. However, with the
incorporation of the magnetic switch and high voltage capacitor, the time regime
changed. Since the magnetic switch had a saturation time in the microseconds, due to the
lower voltages being applied to it, the traveling current sensor was deemed unsuitable for
use. Thus, two commercially available Pearson coils (Model: 101, 100 ns rise time, 0.01
V/A; Model: 4418, 200 ns rise time, 0.001 V/A) were incorporated into the setup to
monitor the current. While one was used to record the current, the other was used to
trigger the pulse generator for the ICCD camera and spectrograph. Both sensors were
placed at the exit of the transformer oil filled magnetic switch housing that entered the T-
junction, see Figure 3.11.
3.5.2 Voltage Measurements
Voltage waveforms are obtained by a capacitive voltage divider embedded in to
the RG220 transmission line at the entrance of the testing chamber, see Figure 3.1. The
reason for choosing the capacitive voltage divider is that they are easily incorporated into
the coaxial transmission line and provide low nanosecond rise-times. Capacitor, C1, of
the divider consists of a piece of copper tape and the inner conductor with the
transmission line insulation as the dielectric. C2 of the divider includes the copper tape
and the outer conductor with a piece of Kapton tape placed in between. Figure 3.13 is a
simple cross-sectional diagram of a capacitive voltage divider. An aluminum shell with a
built in type N connector, for signal transmission, is placed over the divider with a center
pin that extends through the Kapton to make contact with the copper tape. The particular
divider utilized for the surface flashover experiments has a sensitivity of 259 mV/kV and
48
a rise time better than 1 ns. For a more detailed explanation of operation, see reference
31.
Figure 3.13: Cross-sectional rendering of a capacitive voltage divider.
As for the high voltage DC used in charging the capacitor, primary Marx
generator and secondary Marx generator, high voltage DC probes by Fluke where used.
Each Fluke probe (model # 80K-40), rated up to 40 kV, has a sensitivity of 1000:1 and
required an individual multi-meter.
3.5.3 Breakdown Event Imaging
An Andor Technology Intensified CCD (charge couple device) DH-734 GenIII
camera with nanosecond temporal resolution is used to capture the flashover images.
Intensified CCD (ICCD) cameras are either lens coupled or fiber optically coupled to the
49
microchannel plate shutter. This allows for the nanosecond gating times required to
capture surface flashover events. The camera is fitted with a zoom lens and is mounted
on a fully adjustable camera stand. The camera is typically positioned either directly in
front of the chamber viewing through the quartz window or 45 degrees from normal to
the breakdown event looking through the Lexan lid. Since the fiber optic lines are
positioned directly above the breakdown event within the chamber, the camera had to be
given the slight angled offset. The camera is either focused on the dielectric gap space
between the electrodes or the volume between the electrodes when used for volume
breakdown imaging. The output of the .01 V/A Pearson coil current senor is used to
trigger a trigger-delay generator that sends a TTL gate pulse to a second trigger-delay
generator which then sends a 14 V pulse to both the ICCD camera and the spectrograph,
see Figure 3.14. An optical gate time of 500 ns was utilized for both breakdown
experiments. Along with the ICCD camera, a Nikon D200 camera was used to take time
integrated photos of the breakdown events. With the high current breakdown events, the
intensity necessitated a neutral density filter on both the ICCD camera and the Nikon
camera. Usually a neutral density filter (ND) of 4 was taped to the Nikon and a ND of 4
or 5 was placed in front of the ICCD camera.
50
Figure 3.14: Triggering setup used for testing.
3.5.4 Fiber Optics
Since imaging of the discharge provides just a single picture of the developing
discharge, an approach utilizing fiber optics was chosen to capture spatially resolved light
emissions from the discharge gap, see Figure 3.15, 3.16. This spatially resolved light is
used to gain a temporal picture of the arc formation in the gap or to gather spatially
resolved spectra. A special system is constructed to gather the light from the breakdown.
For a given distance, the expected light transmission can be calculated from the cross
section using Beer-Lambert Law,
T = I / I o ∗ exp(−σ ∗ n ∗ d ) (3.2)
with the transmission T, intensity of the incident/transmitted light Io or I, respectively,
absorption cross section σ [cm2], n molecule density (2.5 x 1019 cm-3 for STP), and d
distance [cm]. The resulting transmission for the switch conditions, d = 2.75 inches, n
corresponding to 40 psig of gas, reveals that light above 160 nm is easily transmitted in
an SF6 atmosphere, while this limit is roughly 180 nm for molecular oxygen, see Figure
3.13.
51
Figure 3.15: Absorption cross sections for molecular oxygen and sulfur hexafluoride [2].
Figure 3.16: Light transmission for molecular oxygen and SF6 through 2.75 inches of gas
at a pressure of 40 psig [2].
The light gathering setup has a specially designed lens holder that supports three
cylindrical rod lenses shown in Figure 3.17, 3.18. These rod lenses are made of
Suprasil®; a highly pure, synthetically-fused silica material that does not attenuate light
52
above 175nm and can be purchased from J. R. Cumberland Inc. The lenses are 3.0mm in
diameter and 5.0mm long. They are mounted parallel to one another and are 4.95 mm
apart center to center.
Figure 3.17: Fiber optics setup for spatially resolving light emission in the volume gap
Figure 3.18: Fiber optics setup for spatially resolving light emission in the flashover gap
53
This places the field of focus of the outside lenses just in front of the two
electrodes. The center lens is focused between the electrodes. Behind each lens is a quartz
fiber optic cable into which the lens focuses any collected light. The optical fibers used
are Superguide G fibers from Fiberguide industries. The fibers have a 200 µm core
diameter and a 220 µm cladding including the diameter. The fibers contain a high
hydroxyl ion content which provides high transmission efficiency from UV through
visible and into near IR. The fibers are fed out of the removable, outer conductor through
a small hole that connects to a series of feed-throughs located on a specially designed
leak-proof flange on the side of the test chamber. Connected to the outside of the feed-
throughs is a spectrograph. The entire lens assembly is held in place above the gap by a
sliding Lexan mounting bracket attached to an aluminum support base, see Figure 3.19.
Figure 3.19: Top view of fiber optic lens layout.
54
3.5.4.1 Spectrograph
In order to further investigate the processes contributing to the flashover, a
spectrograph is utilized to analyze the emissions from the arc. The spectrograph used is
an Oriel MS 257 ¼ meter imaging spectrograph. It is equipped with a 4 grating turret
and automatic grating switching. The MS 257 is a multi-track or imaging spectrograph.
While normal spectrographs are designed for spectral (horizontal) resolution at the
expense of vertical resolution, the MS 257 has specially shaped mirrors designed to allow
multiple vertical points to be diffracted at one time as illustrated in Figure 3.20 and
Figure 3.21. Because the F-number of the fiber optic cables, which is 1.67 (much smaller
than the 3.9 F-number of the spectrograph), an F-number corrector is attached to the input
of the spectrograph. The F-number corrector also holds the output of the three optical
fibers in place.
Figure 3.20: Demonstration of difference in imaging and non-imaging spectrographs,

imaging spectrographs have specially designed mirrors to maintain vertical resolution
55
Figure 3.21: Schematic of multi-track spectrograph operation
The second ICCD camera is mounted to the output of the spectrograph. The
spectrum collected is saved as an intensity map for later analysis [32].
3.6 Data Manipulation
The data collected by the oscilloscopes, plus the spectrograph, must each be
corrected to account for the variances within the test setup. In the case of the waveforms
collected by the oscilloscope, each had to be shifted to account for the differences in
lengths found in the coaxial lines. Since 20 dB attenuators were used for both the current
and voltage pulses, the amplitudes had to have the attenuation factors accounted for. Each
row created by the individual fiber optic lines (20 pixels high) had to be converted into a
usable form of data (nm). In order to identify the peaks, the individual rows of pixels
were integrated and scaled to show the relative peak intensities. This integration is done
by a script program written in Andor MCD. The rest of the conversions and all of the
graphs used in this paper are done using PlotIt.

56
3.7 Electrode Design
Since the goal of the design was to mimic the breakdown occurring within the
LTGS at Sandia National Laboratories, the electrodes used had to resemble the electrodes
in the switch, see Figure 3.22. Since the volume breakdown was occurring on the outer
edge of the backbone electrodes used in the rimfire switch, the volume breakdown
electrodes were designed with a similar geometry. The surface flashover experiments
used a geometry that was employed in previous surface flashover experiments at Texas
Tech University. Finally, the electrodes on the secondary setup which also hold the UV
test surface in place will be discussed.
Figure 3.22: Ring electrode geometry in the LTGS at Sandia National Laboratories [2].
57
3.7.1 Primary Setup Surface Flashover Design
The angled electrode design for the surface flashover and surface erosion tests
came from an earlier design which was used in arc formation tests, see Figure 3.24. The
tests used grooved surface dielectric samples which tested whether the arc was more
prone to follow the surface or the electric field lines, see Figure 3.23. If the arc would
follow the lines parallel to the surface and the breakdown over the groove, it adhered to
the electric field lines. The cut groove in the surface increased the distance anode to
cathode. If the arc followed the surface, it would take the longer path and breakdown in
the groove.
Figure 3.23: Dielectric sample geometry and dimensions used for testing.
58
Figure 3.24: Drawing of angled electrodes used for surface flashover/erosion

experiments.
The electrode design allowed for the creation of a triple point junction
(electrode/dielectric/gas), a grasp of the electrode and strong electric field components
normal to the dielectric surface, see Figure 3.25.
Figure 3.25: Maxwell 3D simulation of electric field magnitude for the angled electrode
geometry in a cut plane normal to the dielectric surface with 72 kV applied.
59
3.7.2 Primary Setup Volume Breakdown Design
As explained above, the breakdown environment of the test chamber was to
mimic that found in the rimfire switch. As such, the volume breakdown electrode
geometry came from the outer edge geometry of the backbone electrodes found in the
Z20 rimfire switch, see Figure 3.26, 3.22. Slight variations in machining created some
minor distinctions in dimensions.
Figure 3.26: Electrodes used for volume breakdown in the high pressure chamber. The
electrodes are designed to mimic outside edge of the “vertebrae” in the z20 rimfire
switch.
3.7.3 Secondary Setup Surface Flashover Design
The design for these electrodes had to be one that allowed for a solid connection
to a flat surface and produce strong field components parallel to the surface. Since the
goal of this setup was to create an electric field on the surface, care was taken to ensure
that the breakdown region was from triple-point to triple-point of each electrode. From
previous experiments performed at Texas Tech University, an electrode design was
chosen which ensured a triple point where the electrode met the dielectric surface, but not
60
any other regions under the electrode, see Figure 3.3. A method of compressing the
dielectric surface with the electrodes was installed in the design. For positioning, a
sliding stand was made, see Figure 3.27. It should be noted, that although it is not visible
in Figure 3.27, half of each electrode was then cut in half with a tapered effect, see Figure
3.3.
Figure 3.27: Secondary setup electrode configuration.
3.8 Procedure
Most of the components used in both modes of the primary setup remain the same
except when conducting experiments in a low current fashion. When a low current mode
(≤2 kA) is desired, the magnetic switch and capacitor are grounded. This allows only the
Marx generator to deliver the current. The secondary has only one mode, the low current
mode. The oscilloscopes used to collect the data from the voltage and current sensors,
trigger-delay generators and spectrograph remain the same for high and low current tests.
61
3.8.1 Primary Setup Volume Breakdown Mode
As mentioned above, the two types of breakdown are either the low current or
high current (>2 kA) configurations. To initiate the volume breakdown mode, first the
volume breakdown electrodes are cleaned with methanol and then again with
cyclohexane to ensure a clean surface. The electrodes are available in both brass and
stainless steel. For the experiments discussed in this paper only the stainless steel
electrodes were used. Once installed on the electrode connections inside the chamber, the
fiber optic stand was aligned over the gap. Care is taken to ensure that the fiber optic
lines are equally spaced throughout the gap.
After positioning the fibers, the chamber lid was sealed and the vacuum
components were activated. While the system pressure was dropping to vacuum, the
ICCD camera was typically aligned above the breakdown event and the Nikon camera
was given a side view via the quartz window at the front of the chamber, see Figure 3.28.
Literature has shown that small amounts of Nitrogen dominate the spectra of breakdowns
in SF6, so the chamber was evacuated during each test. Once the chamber reached ~
5x10-5 torr, the gate valves were sealed and the chamber was back filled with SF6 up to
the desired testing pressure. In this paper testing pressure is assumed to be 20 psig unless
otherwise stated. The next step after ensuring the chamber was at the desired pressure is
to pressurize the T-junction and primary Marx generator to 20 psig and 19 psig of SF6
accordingly. Once pressurized, the system was ready to be charged. The Marx generator
was charged to 40 – 41 kV, and when the high current mode was desired, the capacitor
was charged to 10 kV. Making sure that the oscilloscope was armed, plus the trigger-
delays set accordingly, the solid state pulse generator was armed. When ready to fire, the
62
manual trigger on the solid state pulse generator was pushed. Once fired, the system
would recharge itself and be ready to fire again if need with a matter of minutes. An
overall system layout can be seen in Figure 3.29.
Figure 3.28: Side view of camera positions with respect to breakdown event.
63
Figure 3.29: Overall primary setup layout. Gas and charge lines are excluded.
3.8.2 Primary Setup Dielectric Surface Flashover Mode
The dielectric surface flashover mode on the primary setup was essentially similar
to the volume breakdown mode. Rather than the volume breakdown electrodes, the
angled surface flashover electrodes were used along with a dielectric sample of your
choice.
The electrodes were cleaned with the same method as the volume breakdown
electrodes. After the installation of the electrodes, the dielectric sample surface was
cleaned using cyclohexane and a chemical wipe cloth. The electrodes were then inserted
into the slots in the sample and great care was taken in making sure the sample and
electrode made a tight seal. This helped avoid any air gaps between the electrode and
dielectric sample which could lead to field enhancements. Henceforth, the procedure is
the same as that listed in the volume breakdown section (3.8.1).
64
3.8.3 Secondary Setup Surface Flashover
Since the electrodes in the secondary setup were essentially fixed, the procedure
started by cleaning the electrodes with methanol followed by cyclohexane. After this, the
0.1 in thick Lexan surfaces had their protective sticker removed. The surface was then
cleaned with cyclohexane to remove any residual residue. After positioning the surface
underneath the anode and cathode electrodes, the high voltage connections in the
chamber were checked. The process from here on is the same listed in the volume
breakdown section (3.8.1); except that along with the primary setup, the secondary Marx
generator was charged as well. The charging voltage depended on whether the test
required creating an electric field capable of breaking down along the surface, or a field
which lies just under the critical field intensity. Once the voltage is applied, the secondary
solid state pulse generator is armed along with the primary one. At this point, both pulse
generators were setup to be triggered off of the same source. Typically a BNC model 510
trigger-delay generator was used and set to 14 volts output.
65
Figure 3.30: Overall primary and secondary setups in dual firing. Gas and charging lines
excluded.
66
CHAPTER 4
4. ELECTRICAL BREAKDOWN RESULTS
Electrical breakdown experiments where conducted at various pressure and
current conditions. The results presented in this paper deals with the higher pressure
results, but at times will compare to previous low pressure results only to emphasize the
differences. In the following sections, the results for volume breakdown at various
current amplitudes and pressures will be presented. An explanation of the optical
emission spectra gathered from the breakdown event will be presented as well. In the
latter, the results for surface flashover events at varying currents and pressures will also
be presented. The results for varying dielectric materials have been tested and cataloged
and will be presented.
4.1 Reference Waveforms
Typically, the waveforms collected were the camera gate, current and voltage.
These waveforms were saved in a comma delimited file format which allowed for
importing into graphing software such as PlotIt. Small deviations occurred from time to
time in voltage amplitude, time scale offsets due to triggering and coaxial line lengths,
and current amplitude. The current amplitude deviations occurred mainly with surface
flashover due to fluctuations of the arc resistance. A representative set of waveforms can
be seen in Figure 4.1.
67
Figure 4.1: Representative set of waveforms typical of both surface flashover and volume
breakdown.
The camera gate is a TTL gate pulse with a width that was set to 500 ns. After
experimenting with the camera gate delay times, four times were chosen that gave a good
representation of the current distribution. Each of the four current amplitudes resulted in a
distinctly different spectrum and ICCD image. Although the initial breakdown of the gap
occurs within the early ~ 200 ns regime, the magnetic switch does not go into saturation
until ~ 9 µs. When the saturation point is reached by the switch, the current from the HV
capacitor is then allowed to enter the formed plasma channel. It should be noted that
previous data shows that the oscillations occurring in the current last for approximately
1.2 ms. This means that the arc is staying active for 1.2 ms. The typical voltage waveform
is characterized by a sharp rise in amplitude as the voltage pulse is applied to the gap;
68
followed by a distinct drop in voltage as the breakdown event occurs. The bipolar
appearance of the voltage pulse is due to the droop of the fast capacitive voltage divider
incorporated in the transmission line, Figure C.6. It was confirmed with a voltage probe
of slower rise time, but much larger droop time, that the gap voltage is constant after the
initial rise or charging until breakdown occurs. The collapse of the gap voltage due to
electrical breakdown is also indicated by the simultaneous rise in current. This also
occurs during the initial breakdown event in Figure 4.1, but due to the time scales cannot
be seen.
4.2 Volume Breakdown
For the volume breakdown experiments, the primary setup was used with stainless
steel electrodes and a gap spacing of 9.25 mm. A detailed explanation of the electrodes
and operating procedure when in volume breakdown mode can be found in section 3.8.1.
As mentioned previously, the spacing between the electrodes is identical to that found
between the backbone electrodes of the LTGS. Spectroscopic descriptions of the various
environments and amperages, and corresponding image analysis of the arc formations are
shown below. Also, an investigation into the solid by-products of electrical breakdown in
SF6 will be discussed.
4.2.1 Volume Breakdown Images
The results of electrical breakdown in SF6 at various pressures were recorded. The
influence of the various current intensities was also investigated. As far as the imaging
was concerned, the difference was the intensity of the arc. As the delay time of the
69
camera gate shifted from lower to higher currents, a larger attenuation was required from
the neutral density filters. Typically a neutral density filter of 5 was required in the 10 kA
range. To better show the effect of the higher current on the breakdown, a series of
breakdown events from lower to higher pressures will be shown. Keep in mind that at the
lower pressures, sub 0 psig, the influence of gases, such as Nitrogen due to a small leak,
can be noticed by the variation of arc color. Nitrogen for example is known to have a
pinkish color when electrically broken down, see Figure 4.2.
Figure 4.2: Time integrated image of volume breakdown (~ 1 kA) between two stainless
steel electrodes in SF6 at 10 torr. ND = 0. Charged electrode is on the left.
70
Figure 4.3: Time integrated image of volume breakdown (~ 2 kA) between two stainless
steel electrodes in SF6 at 20 psig. ND = 2. Charged electrode is on the left.
Figure 4.4: Time integrated image of volume breakdown (~ 10.2 kA) between two
stainless steel electrodes in SF6 at 20 psig. ND = 4. Charged electrode is on the left.
A noticeable difference in the arc behavior becomes apparent from Figure 4.3 to
Figure 4.4 as the current amplitude is increased. The plasma length and width develop as
large as the diameter of the electrodes as seen in Figure 4.5.
71
Figure 4.5: Front view of time integrated image of volume breakdown (~ 10.2 kA)
between two stainless steel electrodes in SF6 at 20 psig. ND = 4. Charged electrode is on
the left.
At first glance, the breakdown at 20 psig of SF6 seems to be a wide and large arc.
Yet when viewed with an ICCD camera, the main arc can be seen within the plasma
breakdown. In Figure 4.6, it should be pointed out that the inner arc is still more
prominent that an arc formed at 20 psig of SF6 with a lower current.
Figure 4.6: Front view of ICCD image of volume breakdown (~ 10.2 kA) between two
stainless steel electrodes in SF6 at 20 psig. ND = 4. Charged electrode is on the left.
72
Figure 4.7: Volume breakdown voltage and delay time of a 9.52 mm gap in SF6 at
increasing pressures with ~ 50 ns risetime voltage pulse. Delay time is defined as the
FWHM from the rise of the voltage pulse to the drop due to breakdown [41].
In Figure 4.7, the breakdown voltages by corresponding pressure and the average
delay time per pressure can be seen. The pressures listed are from 10 Torr to 2069 Torr.
Figure 4.7 shows that as the hold-off voltage increases, the corresponding delay time
increases. This is expected, and can be attributed to increased electron energy loss
processes, such as inelastic collisions, at higher pressures due to the decrease in mean
free path. It should be noted that the values in Figure 4.7 correspond to a low current (≤ 2
kA) volume breakdown event and that in the pressure regime below 20 psig (1034 Torr),
the breakdowns occur on the rising edge of the incident pulse.
4.2.2 Volume Breakdown Spectroscopy Analysis
The optical emissions of the breakdown event had been gathered and analyzed for
various pressures and current amplitudes. An interesting observation between the spectra
73
gathered along the discharge path at low currents is that they are almost identical. For a
look at the spectra gathered from low pressure breakdowns, see Appendix B. The spectra
presented in this section were gathered by the center fiber optic lens. The fiber optics
close to the electrodes showed similar spectroscopic results; except they showed iron
lines from the steel electrodes.
The identification of prominent UV emission from volume breakdowns in SF6 at
high pressures was crucial. Previous experiments at Texas Tech University have shown
that UV below 320 nm largely affects the behavior of a surface flashover event [1]. Since
photoelectric emission from the surface increases with an increased photon energy, or
decreased wavelength [34, 35], it is obvious that a higher UV output from the volume
spark will liberate an increased number of photoelectrons from the LTGS PMMA
envelope surface. As seen in the Figure 4.8 and 4.10, a larger contribution of UV in the <
320 nm region of the spectra is apparent and as such reaffirms the initial hypothesis of
UV influencing the initiation of surface flashover in the LTGS envelope. The broad
features in the spectra at 20 psig are compared to black body radiation in Figure 4.8,
using equation 4.1. The spectral sensitivity of the collection apparatus drops off
significantly at wavelengths approaching 200 nm, Figure 4.9, and is perhaps responsible
for the drop in the spectra around 220 nm. The other possibility for the drop around 220
nm is self absorption.
74
Figure 4.8: Spectra collected from a volume discharge in SF6 at high pressures. Spectra
gathered from low current breakdowns [41].
Figure 4.9: Scaling curve for all collected spectra, accounting for internal loss in the
collection apparatus, spectrograph, and imaging camera.
75
400000
50 µs Delay (8.57 kA)
200000
Relative Intensity [a.u.]

400000
200000
0
200000 10 µs Delay (2 kA)
100000
0
80000
140 ns Delay (800 A)
40000
0
200 300 400 500 600 700
W avelength [nm]
Figure 4.10: Spectra collected from a volume discharge in SF6 at high pressures and at
various current levels. Spectra gathered from higher current breakdowns.
According to previous surface flashover investigations, illumination of the surface
with a UV pulse significantly decreases the delay time by up to 50% from pulsed
excitation to flashover, and decreased variations in the delay time from shot to shot [42].
The UV also had a profound effect on the discharge path, resulting in the discharge
following the surface as opposed to the electric field line [43]. This effect could easily
contribute to the LTGS failure modes.
Comparison of the gathered spectra pattern to a black body radiation pattern was
accomplished using Plank’s equation. In Figure 4.10, Plank’s equation:
2 ⋅ h ⋅ c2 1
ρ λ (T ) = ⋅ (4.1)
λ5 (
exp hc
kTλ
)−1
at 16,800 K, is used to model the broad features, which a high degree of correlation (h:
Plank’s constant, c: speed of light, λ: wavelength, T: blackbody temperature). It is
76
assumable that the higher the pressure, the closer the spectral output will be to a black
body radiator. However, the spark plasma is still not completely thermalized at the 20
psig pressure as indicated by spectral line features superimposed onto the broad
background.
Nevertheless, the estimated temperature of ~ 16,800 K for the SF6 spark falls
within the range of centerline temperatures measured for a stationary arc at a pressure of
120 psig (15,500 K for 400 A and 27,000 K for 13,000 A) [35], see Figure 4.11.
400000
20 psig (High Current)
200000
0
20 psig (Low Current)
40000
Black Body (16800 K)
20000
0
200 300 400 500 600 700
W avelength [nm]
Figure 4.11: Spectra comparison of low current and high current breakdowns. High
current is approximately 10.4 kA, and low current is roughly 1 kA. Low current
breakdown at high pressures still maintains black body radiation pattern.
4.2.3 Solid By-Products of Volume Breakdown
After approximately 60 volume discharges at high currents several changes to the
inside of the chamber were noticed. The most noticeable was a fine white powder like
77
substance found to be covering all components within the chamber. A thin film covered
items located greater than 5 inches horizontally from the breakdown event. An example
of the thin white layer found in the chamber can be seen in Figure 4.12.
Figure 4.12: White powder deposit noticeable within the testing chamber. Powder is
prominent on the electrode holder, corona ring and within the RG220 chamber entrance.
According to literature [36], X-Ray analysis has found that the dominant element
in the powder to be sulphur. The analysis also revealed a complex array of atomic species
– fluorine, sodium, aluminum, silicon, sulphur and potassium with traces of calcium and
zinc. The fluorine and sulphur were produced by SF6 dissociation; silicon resulted from
the etching of the large chamber windows found in their experiment and the other atoms
were associated to the chamber aluminum-alloy walls. Except for the chamber walls
being stainless steel in our experiment, the design is similar in nature that found in the
literature reference. One could easily instigate that the powder found in the experiments
presented in this paper are similar in material composition. Identification of element
composition is still required.
78
The second most noticeable byproduct was that found on the stainless steel
electrodes. Asides from pitting of the electrode surface, a build up on the electrode
material was apparent. A literature search revealed that the deposit and accompanied
pitting effect were caused by the fluorine disassociated from the SF6 gas. The fluorine
reacts with the surface of the electrodes to produce metal fluorides such as FeF2.
Although further identification is still required in our experiments, the same paper
confirmed their findings through X-ray analysis. An example of the build up and deposit
can be seen in Figure 4.13.
The last effect seen in the chamber was again found in conjunction to the
electrodes. Hair-like fibers approximately 2.5 mm in length were found on both the anode
and cathode, see Figure 4.13. Because of the method primary system design, when the
capacitor is charged a DC voltage is put on the primary electrodes in the chamber. After
firing of the primary setup, there is still a residual voltage left of the capacitor. At one
such time, a voltage was left with ~ 80 Volts and the hairs became electrostatic. Once the
capacitor was discharged, the electric field was removed and the hairs then drooped.
Further investigation is required to identify the elements in the fibers, but they could
contain either a metal fluoride or iron from the stainless steel electrodes.
79
Figure 4.13: Hair like fibers found on the electrode surfaces with electrostatic effect.
Metal fluorine deposits can be seen by the orange region.
4.3 Dielectric Surface Flashover
For the surface flashover experiments, the primary setup was used and the angled
surface flashover electrodes were installed as well, see section 3.7.1. For a detailed
explanation of the operation of the primary setup see section 3.8.2. The purposes of the
surface flashover testing was to characterize various materials found within the LTGS,
and then determine which characteristics would make a material more suitable for
breakdown prevention and quantify the results. Asides from waveform analysis, each
breakdown event was documented through spectroscopy. Post-surface flashover analysis
will also be presented in this section.
4.3.1 Surface Flashover Images
Surface flashover testing occurred with various pressures of SF6 and at varying
current levels. The results for the higher pressures of SF6 will be presented in this section,
80
while the lower pressure data can be found in Appendix C. A lower pressure breakdown
image will be shown to emphasize the difference between high and low current
breakdowns. The conditions of the surface were also varied to view the effects it had on
the arc behavior at lower current levels. The three surfaces used are roughed (sand
papered surface), grooved (notch cut tangential to the electrode slot) and smooth (cleaned
factory smooth surface). This section will be limited to the results for a smooth surface.
The data for the roughed and grooved surfaces at low currents can be seen in Appendix
C. Although many material types were tested, Teflon and Lexan (polycarbonate) were
materials of most interest and as such they will be presented. The distance between the
electrodes averaged at 10 mm, but can vary ± .5 mm due to machining error.
Another noticeable difference in the surface flashover arc from lower to higher
pressures was the color. As in the volume breakdown events, at lower pressures, the
discharges had a pinkish hew, see Figure 4.14. Nitrogen is known to give a pinkish color
when excited, and the change in color can be associated to a small leak in the chamber at
the lower pressures.
Figure 4.14: Time integrated image of surface flashover of a Lexan (polycarbonate)

surface in SF6 at 10 torr. Charged electrode is on the left. Low current (< 2 kA) discharge.
81
Figure 4.15: Time integrated image of surface flashover of a Lexan (polycarbonate)

surface in SF6 at 40 pisg. Low current (< 2 kA) discharge.
Figure 4.16: Time integrated image of high current surface flashover at 20 psig of SF6 on
Lexan (polycarbonate) sample. ND = 4. High current (~ 10.2 kA) discharge. Side View.
82
Figure 4.17: Time integrated image of high current surface flashover at 20 psig of SF6 on
Virgin Teflon sample. ND = 4. High current (~ 10.2 kA) discharge. Side View.
Figure 4.18: ICCD image of high current surface flashover at 20 psig of SF6 on Virgin
Teflon sample. ND = 3. High current (~ 800 A) discharge. 500 ns exposure time. Top
view.
Teflon sample. ND = 3. High current (~ 2 kA) discharge. 500 ns exposure time. Top
view.
83
Teflon sample. ND = 4. High current (~ 10.2 kA) discharge. 500 ns exposure time. Top
view.
Teflon sample. ND = 4. High current (~ 8.6 kA) discharge. 500 ns exposure time. Top
view.
In addition to the imaging of the surface flashover event, the time delay for
breakdown was also recorded and was most revealing for the pulsed voltage hold-off
capabilities of different materials. For the various materials tested at atmospheric pressure
(Lubbock, TX: ~680 torr ), a comparison table was created and lists the average time
delays sorted by field strength, see Figure 4.22. The materials are sorted by field strength
due to minor gap adjustments (± 1 mm) originating from the machining process. Despite
the varying field strengths, Figure 4.22 indicates that Teflon had a longer breakdown
delay time even at a higher applied field when compare to the Lexan and the epoxies. The
one sample of Plexiglas (material taken directly from an unused Z switch envelope)
exhibited the best performance at 1 atmosphere of SF6, but must be treated with caution
84
since only a single sample was tested. The reason for this discussion of the lower pressure
results is due to a hypothesis of breakdown time delay and the dielectric constants.
Lexan, Teflon and Plexiglas have dielectric constants between ~ 2 to 3, while the epoxy
based materials have a relative permittivity of ~ 5, which could help explain their poor
performance. A higher dielectric constant could lead to increased electron emission at the
triple point.
Figure 4.22: Dielectric material comparison of the average time delays. Samples arranged
by field strength (asterisk indicates one sample only tested). Surface flashover event at
atmospheric SF6 and at low current (< 2 kA) breakdown. Only one sample of Plexiglas.
At the higher current levels, the same trend between the Teflon and Lexan
samples was noticed, see Figure 4.23. Yet again, the Teflon held off at higher field
strengths for a longer amount of time.
85
120
1 atm SF
84 kV/cm 6
110 ns 20 psig SF
6
80 kV/cm
98 ns
80
Delay Time [ns]
40
243 kV/cm 178 kV/cm

21 ns 18 ns
0
Teflon Lexan
Figure 4.23: Dielectric material comparison of the average time delays. Samples arranged
by time delay. Atmospheric breakdown with low current (< 2 kA) and 20 psig with high
current (~10.2 kA).
4.3.2 Surface Flashover Spectroscopy Analysis
In order to understand the effects an applied UV excitation has on a surface
flashover event, the surface flashover event must first be investigated and understood.
The spectroscopic analysis of surface flashover events over various materials was
documented and will be presented. For spectra collected at lower pressures, see Appendix
C.
Overall, an important similarity was found between the surface flashover events
and that of the volume breakdowns. At the lower current breakdowns, the spectra
resembled the radiation pattern which was similarly found in the volume breakdown
events, see Figure 4.24. A slight difference is the presence of carbon peaks found in the
surface flashover due to the addition of the dielectric surfaces, see Figure 4.27. Easily
noticeable are the carbon peaks at ~ 250 nm and at ~ 660 nm. Another interesting
86
observation is that since the composition of Teflon is ~76 % fluorine, a larger
contribution of fluorine is found in the spectra when compared to volume breakdowns at
high current amplitudes; see Figure 4.26, 4.10. A similar situation occurs in the Lexan
whose composition is ~ 75.6 % carbon. Some identified peaks can be seen in Figure 4.27.
Figure 4.24: Comparison of spectra collected from a volume breakdown and a surface
flashover of a Lexan (polycarbonate) surface in an SF6 environment at 20 psig. Both are
low current breakdown events.
87
120000
80000
40000
0
400000

200000
0
120000
10 µs Delay (2 kA)
80000
40000
0
40000 140 ns Delay (800 A)
20000
0
200 300 400 500 600 700
Wavelength [nm]
Figure 4.25: Spectra collected from a surface flashover on Lexan in SF6 at 20 psig and at
various current amplitudes.
40000 50 µs Delay (8.57 kA)
20000
0

80000
0
80000
10 µs Delay (2 kA)
40000
0
20000 140 ns Delay (800 A)
10000
0
200 300 400 500 600 700
Wavelength [nm]
Figure 4.26: Spectra collected from a surface flashover on Teflon in SF6 at 20 psig and at
various current amplitudes.
88
As the current is increased in the arc, the UV contribution is larger than at the
lower current levels. This is most likely attributed to the higher excitation states of the
disassociated SF6 molecules.
Figure 4.27: Spectra comparison of low current and high current surface flashover events
on Lexan. High current is approximately 10.4 kA, and low current is roughly 1 kA. Low
current breakdown at high pressures still maintains black body radiation pattern.
4.3.3 Post Surface Flashover Analysis
At the sponsor’s request, post flashover analysis was also performed on the
dielectric materials. The analysis helped in assessing the damage or tracking produced by
the discharge in association with the material. Damage to the surface results in
conditioning and as such the subsequent flashover event and material performance are
both affected. Low level analysis was performed using an optical microscope. This level
documented visual surface damage. The areas of interest were then investigated by means
of a scanning electron microscope (SEM). The SEM proved useful for identifying
89
damage that could not be seen or was vague under a light microscope. At times, surface
contaminations required elemental identification, and as such analysis is further
performed via X-ray fluoroscopy. Most of the SEM analysis at this point was performed
on the low current discharges. Sulphur deposits at the low current discharges, for
example, were not as prominent in the higher current breakdowns. At the higher
breakdowns, the thin white sulphur film was easily collected via a cotton swab or finger.
The difference in the surface erosion with respect to the breakdown current can be seen in
Figure 4.35.
Figure 4.28: Smooth surface dielectric sample 826/D400/KF-865 cured at 80ºC for 3 days
after 38 discharges (1 atm SF6, 30 µC).
90
Figure 4.29: Smooth optical grade Lexan after 10 discharges (1 atm SF6, 30 µC).
Figure 4.30: SEM image of smooth surface dielectric sample 826/D400/KF-865 cured at
80ºC for 3 days after 38 discharges (1 atm SF6, 30 µC). Distinct difference between
flashover region and the unaltered Epoxy surface.
91
Figure 4.31: SEM image from dielectric material 826/D400/KF-865 used for X-ray
fluoroscopy analysis (1 atm SF6, 30 µC).
Figure 4.32: X-ray fluoroscopy results for crosshair location in Figure 4.31. X-ray
fluoroscopy indicated sulphur deposit.
92
Figure 4.33: Lexan image used for X-ray spectroscopy (40 psig SF6, 108 µC). Cross hair
denoted point at which elemental composition was performed.
Figure 4.34: X-ray fluoroscopy results for crosshair location in Figure 4.33. X-ray
fluoroscopy indicates sulphur deposit.
Although SEM analysis showed sulphur being disassociated from the SF6, the gas
was not the only material being excited. At the high current surface flashovers, carbon
was being pulled from the surface of the dielectrics. In the top row of Figure 4.35, the
black marks on the surface are from the carbon found in the various materials (Lexan,
93
Teflon, Rexolite). The carbon removal is more prominent in the Lexan and Rexolite than
in Teflon. Most likely due to the fact that Teflon is composed of only ~ 24% carbon, a
factor of 3 less than that found in Lexan (~75% carbon) [37].
The amount of carbon removed from the various materials is related to the quantity
found in the material and directly impacts the behavior of the surface flashover event. An
effect of surface flashover was found in the Lexan and Rexolite while charging the HV
capacitor for a successive shot. The carbon pulled to the surface of the materials after a
single surface flashover event was enough to create a carbon resistor (~50 kΩ). This new
resistance allowed the current used to charge the HV capacitor a second path with lesser
resistance and therefore did not allow for charging of the capacitor. Only one shot was
allowed on the Lexan samples before cleaning of the surface was required. The same
surface degradation occurred for the Rexolite, except that two flashover events were able
to be performed rather than one as in the Lexan. The excitation of the carbon on the
surface of the Teflon can be noticed by small black spheres appearing directly under the
breakdown event. Not enough carbon was pulled from the material to cause an effect as
seen with the Rexolite and Lexan. As a side note, a week after discharging across the
Teflon sample, a yellow hew developed on the surface outside of the flashover region.
Further investigation is required to identify the source of this phenomenon.
94
Figure 4.35: Material surface degradation comparison. Lexan, Teflon, and Rexolite from
left to right respectively. Top row is high current discharges (301 mC), with 1, 6 and 2
surface discharges respectively. Bottom row is low current discharges at 100 shots per
sample (108 µC).
95
CHAPTER 5
5. CONCLUSION
The phenomenology of volume breakdown and surface flashover in high pressure
SF6 has been experimentally investigated. While future work will focus on the direct
impact of volume breakdown on a nearby surface under high electric field stress just
below the surface flashover point, the following will summarize the results gained from
each independent electrical breakdown type with a brief mentioning of a possible
interaction between the two.
5.1 Volume Breakdown in SF6
Photoemission of electrons by UV photons impacting the envelope is one of the
suspected reasons that can initiate the unwanted tracking along the inside envelope
surface in the Sandia LTGS. It is shown that a significant amount of UV light is produced
by the volume spark in high pressure SF6. The more intense light output is in the
wavelength regime below 320 nm; a regime that has been identified in previous research
as having a distinct effect on the spark path of flashover [1]. The majority of species
contributing to the observed optical emission spectra between 200 and 700 nm has been
identified, see Appendix B. It was possible to estimate the volume spark’s electronic
temperature (~35,000 K) for the lower pressures (50 torr) using lines of atomic spectra.
However, for breakdown in 20 psig SF6 or more, the broadband features mask distinct
atomic or molecular lines. Assuming black body radiation the temperature is estimated to
be of the order of 17,000 K at 20 psig (~2 kA) and similar to that at 40 psig (~2 kA).
96
The absence of metal lines in the low current breakdown spectra is a result of the
low coulomb transfer and low current amplitude (~ 2 kA); however metal lines do show
up in the optical emission spectra at higher currents (~10 kA). As expected, the spectral
shape changes with the current amplitude and should continue to change as the current is
increased to the 100 kA regime in future research. Overall, the volume breakdown results
support the hypothesis of UV radiation contributing to the initiation of switch failure in
the sense that UV radiation has been detected at wavelengths that a.) will cause
photoemission of electrons from the surface and b.) easily propagates with little
attenuation within radial distances found in the Sandia LTGS switch.
5.2 Dielectric Surface Flashover in SF6
Several aspects of surface flashover along dielectric insulators have been studied
in SF6 environments. Imaging of the discharge along a dielectric surface revealed a
material dependence on the discharge path. Previous low current experiments on grooved
surface materials have shown that even with an electric field leading away from the
surface, discharges continued to preferentially follow the surface for most materials; i.e.
polycarbonate and the epoxies, indicating a strong interaction between the surfaces and
the discharge. This interaction was believed to involve the productions of photoelectrons
by the surface while exposed to UV radiation generated by the forming discharge, in
which existence was confirmed by the collected optical emission spectra. Teflon showed
the weakest interaction with the surface, following the surface only 50% of the time. This
indicated that Teflon, or a material with similar or lower quantum yields for photo
emission, would be the preferred choice for minimizing undesired envelope flashovers
97
[29].
In addition to imaging of the discharge, post-mortem imaging of the samples
made the differences in surface conditioning apparent. For instance, the epoxy, Rexolite
and polycarbonate exhibited very distinct visible surface damage, as opposed to Teflon
that exhibited less visible damage. As surface conditioning reduces flashover fields of
successive flashover events, Teflon was again shown superior to the other materials. This
was further supported by the flashover delay time measurements, thus indicating that
Teflon would hold off flashover longer under greater field stress than the epoxies,
polycarbonate and Rexolite. It should be noted, however, that thus far only one Plexiglas
sample was tested, and it too exhibited a long flashover delay time, and non-visible
surface conditioning at low current surface flashover.
Comparing volume breakdown and surface flashover optical emission spectra
revealed the presence of strong carbon lines in the UV for surface flashover at 20 psig.
Since these line were absent in the volume breakdown spectra, they serve as an in-situ
indicator for surface erosion for hydrocarbon based materials. This also implies that a
measure of caution should be taken when using hydrocarbon based materials in envelope
construction. As shown in section 4.3.3, the higher the composition of carbon in a
material, the more damage to the surface; also possibly due to variances in the bond
energies of the various materials.
The occurrence of sulfur on the dielectric surface, as recorded in literature [36],
was consistent with the prevalent appearance of sulfur ionic lines in the flashover optical
emission spectra (SII and SIII). The question remains as to whether these deposits are a
cause or result of the poor hold-off characteristics of the materials. Future testing with
98
secondary surfaces placed in the chamber to measure hold-off potential, after collecting
any deposits created by an initial flashover event, should help clarifying the role of the
sulfur deposits.
99
REFERENCES
[1] K. Morales, “ Physics of Pulsed Unipolar Dielectric Surface Flashover at

Atmospheric Conditions”, Thesis for the Degree of Master of Science in Electrical
Engineering, Texas Tech University, Lubbock, Texas, 2006.
[2] Sandia Report: SAND2007-0217, “Fundamental Science Investigations to Develop

a 6-MV Laser Triggered Gas Switch for ZR: First Annual Report”, March 2007.
[3] D. Lide, CRC Handbook of Chemistry and Physics, CRC Press, 81st Ed., 2000-
2001.
[4] D. K. Davies, “Charge generation on dielectric surfaces,” Brit. J. Appl. Phys., Ser.
2, Vol. 2, 1969.
[5] E. Nasser, Fundamentals of Gaseous Electronics, Wiley-Interscience, New York,

NY, 1971.
[6] B. G. Streetman and S. Banerjee, Solid State Electronic Devices, Prentice Hall,
Upper Saddle River, NJ, 2000.
[7] Y.P. Raizer, Gas Discharge Physics, Springer-Verlag, New York, NY, p. 71, 1991.
[8] D. Alpert, D.A. Lee, E.M. Lyman, and H.E. Tomaschke, “Initiation of Electrical
Breakdown in Ultrahigh Vacuum,” Jour. Vac. Sci. & Tech., Vol. 1, Iss. 2, pp. 35-
50, November 1964.
[9] P.G. Slade and E.D. Taylor, “Electrical Breakdown in Atmospheric Air Between
Closely Spaced Electrical Contacts,” IEEE Trans. Components and Packaging
Technologies, Vol. 25, No. 3, pp. 390-396, September 2002.
[10] Feuerbacher, Fitton, Willis, Photoemission and the Electronic Properties of

Surfaces, John Wiley & Sons, 1978.
[11] Dionne, G. F., “Origin of Secondary-Electron-Emission Yield-Curve Parameters,”

J. Appl. Phys., Vol. 46, No. 8, August 1975, pp. 3347-3351.
[12] Meek, J.M. and Craggs, J.D., Electrical Breakdown of Gases, John Wiley and Sons,
1978.
[13] Christophorou and Olthoff, “Electron Interactions with SF6”, Journal of Physical
and Chemical Reference Data, 29 (3): 267-330, May-Jun 2000.
100
[14] J. T. Krile, “Surface Flashover Under RF and Unipolar Excitation at Atmospheric

Conditions,” Dissertation for the Degree of Doctor of Philosophy in Electrical
[15] E. Kuffel, W. S. Zaengl, and J. Kuffel, High Voltage Engineering: Fundamentals,

Newnes, New York, NY, 2000.
[16] Brown, Sandborn C., Introduction to Electrical Discharges in Gases, John Wiley &
Sons Inc., New York, 1966.
[17] G.G. Raju, Gaseous Electronics, CRC Press, 2006.
[18] L B. Loeb, Basic Processes of Gaseous Electronics, University of California Press,

1955.
[19] H Raether, Electron Avalanches and Breakdown in Gases, Butterworths, London,

1964.
[20] H Bluhm, Pulsed Power Systems: Principles and Applications, Springer, 2006.
[21] Miller, H., “Surface Flashover of Insulators,” IEEE Transactions on Electrical

Insulation, Vol. 24, No 5, pp. 765-786, October, 1989.
[22] Papoular, R., Electrical Phenomena in Gases, American Elsevier pub. Co. Inc., New
York, 1965.
[23] S. Schreier, On the Beakdown Voltages of Some Electronegative Gases at Low

Pressure, IEEE Trans. PAS, 83, 468, 1968.
[24] I. W. McAllister, “Electric Fields and Electrical Insulation”, IEEE Trans. On

Dielectrics and Electrical Insulation, Vol. 9, No. 5, Oct. 2002.
[25] I. W. McAllister, and G. C. Crichton, “The Concept of Paschen’s law with

Reference to SF6”, J. Phys. D: Appl. Phys. 20, 1537-1539, 1987.
[26] R. A. Kishek, Y. Y. Lau, L. K. Ang, A. Vafells, and R. M. Gilgenbach, “Multi-

pactor discharge on metals and dielectrics: Historical review and recent theories”,
Physics of Plasmas, Vol. 5, No. 5, pp. 2120-2126, May, 1998.
[27] S. E. Black, “Nonlinear Magnetic Switches for Pulse Generation”, Thesis for the
Degree of Master of Science in Electrical Engineering, Texas Tech University,
Lubbock, Texas, 1980.
[28] G. T. Coate, and L. R. Swain Jr., High-Power Semiconductor-Magnetic Pulse

Generators, The M.I.T. Press, 1966.
101
[29] J. K. Krile, R. Vela, A. A. Neuber, H. G. Krompholz, “Pulsed Dielectric Surface

Flashover in an SF6 Environment”, IEEE Trans. on Plasma Sciences, submitted for
publication.
[30] H. Krompholz, K. Schoenbach and G. Schaefer, “Transmission Line Current

Sensor,” Proceedings of the ITMC-IEEE Instrumentation and Measurement
Technology Conference, pp. 224-227, March 1985.
[31] S. T. Pai and Qi Zhang, Introduction to High Power pulse Technology, World
Scientific Publishing Co. Pte. Ltd., 2003.
[32] J. Krile, “Physics of Dielectric Surface Flashover at Atmospheric Pressures,” Thesis

for the Degree of Master of Science in Electrical Engineering, Texas Tech
University, Lubbock, Texas, 2003.
[33] J. Tom, H. A. Verhaart, A. L. Verhage, and C. S. Vos, “Photo-emission of Charged

Insulators in Insulating Gases,” Proceedings of the 2nd International Conference on
Conduction and Breakdown in Solid Dielectrics, pp. 301-307, Erlangen, 1986.
[34] A. A. Guzhov, and Yu. A. Shuba, “The Phototemission of Some Massive Insulators
in the Vacuum Ultraviolet,” Optical Technology, Vol. 38, No. 4, pp. 198-199, April
1971.
[35] J.J. Lowke and R.W. Liebermann, "Predicted Arc Properties in Sulfur
Hexafluoride," J. Appl. Phys., Vol. 42, 3532-3539, 1971.
[36] T. C. Mortensen, Y. Bunganaen, G. B. Scelsi, G. A. Woolsey, “Gaseous and solid

by-products of SF6 discharges”, Properties and Applications of Dielectric materials,
1994., Proceedings of the 4th International Conference on, Vol. 1, Iss., 3-8 Jul 1994.
pp. 135-138.
[37] NIST Atomic Spectra Database Lines Form, online at:

http://physics.nist.gov/PhysRefData/ASD/lines_form.html
[38] K. P. Morales, J. T. Krile, A. Neuber, and H. Krompholz. “Pulsed Dielectric

Surface Flashover at Atmospheric Conditions,” Pulsed Power Conference, 2005
IEEE, Vol., Iss., June 2005, pp 1147-1150.
[39] J. Corley, et. al., “Tests of 6-MV Triggered Switches on APPRM At SNL”,
PPCS2001, pp 1778, IEEE 0-7803-7120-8, 2002.
[40] J. Corley, et. al., “Development/Tests of 6-MV Triggered Gas Switches at SNL”,
PPC2003, pp 875, IEEE 03CH37472C, 2003.
102
[41] J. K. Krile, R. Vela, A. A. Neuber, H. G. Krompholz, “Spectral Analysis of Pulsed

Volume Breakdown in SF6 at High Pressures”, IEEE Trans. on Plasma Sciences,
submitted for publication.
[42] J. Krile, A. Neuber, J. Dickens, and H. Krompholz, “DC and Pulsed Dielectric
Surface Flashover at Atmospheric Pressure,” IEEE Trans. On Plasma Sci., 1149-
1154, Aug 2005.
[43] K. P. Morales, J. T. Krile, A. A. Neuber, and H. G. Krompholz, “Pulsed Dielectric

Surface Flashover in Atmospheric Conditions,” IEEE Trans. on Dielectrics and
Electrical Insulation, vol. 13, no. 4, pp. 803-809, Aug. 2006.
[44] F. Hegeler, “Dielectric Surface Flashover in a Simulated Low Earth Orbit

Environment,” Dissertation for the Degree of Doctor of Philosophy in Electrical
103
APPENDIX A
A PASCHEN’S LAW WITH REFERENCE TO SF6
A sf6 := 3.74kV
kV
Bsf6 := 6.96
mm⋅ bar
Vb ( x) := x⋅ 8.84⋅ ⎛⎜
kV ⎞
+ .5kV [1] Assuming a uniform electric field - requires adjusting to
⎝ mm⋅ bar ⎠ desired electrode configuration. See discussion below.
[2] See discussion below.
Vb2( x) := A sf6 + Bsf6⋅ x
300
200
V [kV]
100
0
0 500 1000
pd [torr*m]
Curve according to reference 1
Curve according to reference 2
104
Discussion of curves:
The linearity of the paschen curve for SF6 has been experimentally proven for ranges of
p
.25 ≤ ≤ 4 [2]. The electrode geometry was a point plane configuration, a hemispherically-
bar
ended rod of tip radius 8 mm with a gap spacing of 5 mm [2]. As stated above, the equation
from bluhm assumes uniform electric fields. Equations are available which will adjust Bluhm's
breakdown equation for various electrode geometries [1].
It should be noted that the paschen curve for SF6 deviates from linearity (tails off) at the ends
[8].
According to literature [3,4] there is a E/p limit below which no sparking should be possible
regardless of the magnitude of p*d to take note of for SF6 :
V
Predicted: 117
cm⋅ mmHg
V
Observed: 117.5 [3]
cm⋅ mmHg
V
103 [4]
cm⋅ mmHg
References:
[1] Bluhm, Pulsed Power Systems, Springer, pp. 13-15

[2] McAllister, Crichton, The Concept of Paschen's law with reference to SF6, J. Phys. D:
Appl. Phys. 20 (1987) 1537-1539.
[3] M. S. Bhalla and J. D. Craggs, Measurement of Ionization and Attachment Coefficients in
Sulphur Hexafluoride in Uniform Fields, Proc. Phys. Soc., 1962, vol. 80.
[4] Geballe, R., and Reeves, M. L., A condition on Uniform Field Breakdown in Electron-
Attaching Gases, 1953, Phys. Rev., 92, 867
Selected papers for more on SF6 Paschen curve issues:
[5] Berger, Senouci, The role of impurities on the deviation from paschen's law of SF6, J.
Phys. D: Appl. Phys. 19 (1986). Printed in Great Britain.
[6] McAllister, Electric Fields and Electrical Insulation, IEEE Transactions on Dielectrics and
Electrical Insulation, Vol. 9, No. 5, October 2002.
[7] McAllister, Crichton, The concept of paschen's law with reference to SF6, J. Phys. D: Appl.
Phys. 20 (1987) 1537-1539. Printed in the UK.
[8] Osmokrovic, krivokapic, Krstic, Mechanism of Electrical Breakdown left of Paschen
minimum, IEEE, Transactions on Dielectrics and Electrical Insulation, Vol.1, No.1,
February 1994. pp77-81.
105
APPENDIX B
B SPECTRAL ANALYSIS OF PULSED VOLUME BREAKDOWN IN SF6 AT HIGH

PRESSURES
John T. Krile, Member, IEEE, Russell Vela, Student Member, IEEE, Andreas A. Neuber,
Senior Member, IEEE, and Hermann G. Krompholz, Senior Member, IEEE
Abstract—The Z-machine, located at Sandia National Laboratories, is currently undergoing refurbishment to

increase the output drive current. Due to increased switching voltage requirements, some switch failure modes
have been identified with the laser trigger gas switch design, including envelope surface flashover. In order to
improve the performance and life time of these switches a basic understanding of the underlying physics of the
failure mechanisms is required. A small-scale experimental setup has been constructed to approximate
conditions within the switch. The possible impact of the SF6 volume spark between the switch electrodes on the
envelope surface flashover is investigated. Measured optical emission spectra of the SF6 volume spark in a wide
pressure range, rough vacuum to 40 psig, are analyzed regarding their potential to contribute to switch failure.
Index Terms—Breakdown, SF6, Spectroscopy
I. INTRODUCTION
P REMATURE switch failure has become a problem in the refurbishment of the Z-machine located at
Sandia National Laboratories [1], [2]. The Laser Triggered Gas Switch, LTGS, typically operated at
above atmospheric SF6 pressures, such as 40 psig, has evolved to accommodate significantly higher voltage
requirements over the years. The generic switch design, with 305 mm envelope diameter and 165 mm
vertebrae diameter, can be seen in Fig. 1. In the present design, a primary failure mode is surface flashover
in the high pressure gas (SF6, > 40 psig) on the inside of the PMMA envelope. This surface flashover
occurs initially during the regular cascade switching process without affecting the switch’s electrical
performance. However, the switch will pre-fire in the successive shot. The damage to the envelope has to
be repaired before normal switch operation can resume. It is believed that damage to the envelope occurs
prior to the failure. Under normal operation the envelope is exposed to optical radiation (UV-VIS)
generated by the volume breakdown of the cascade section of the switch. In order to determine what effect
this illumination might have on the envelope, a smaller scale pulsed SF6 volume discharge setup was
constructed. Optical emission spectroscopy is preformed on the radiation emitted by the volume spark and
the emission spectrum is analyzed.
A literature survey focused on SF6 emission spectra for the conditions found in the LTGS
uncovered little information. Most research was done on stationary SF6 arcs or for the power distribution
industry, with the goal of detecting corona discharges in insulating SF6. This is used primarily as a method
of detecting emerging failures in SF6 insulated power systems [3]-[5]. A reoccurring theme in the literature
is the pronounced effect of nitrogen contamination on the observed optical emission spectra. Having as
little as a 1% fraction of N2 in SF6 will show up prominently as the N2 second positive band in the
otherwise SF6 dominated spectrum. There are several shortcomings in the literature, for instance, only the
raw spectra are typically shown. This makes evaluating the plasma temperature from the spectrum difficult
if not impossible. In some cases the temperature is inferred from spectra generated by the nitrogen
contamination in the system; however, this is again done without adequate sensitivity correction. It is
Manuscript received October 20, 2006. This work was supported Sandia National Laboratories.
The authors are with Texas Tech University, Lubbock, TX 79409 USA (e-mail: jkrile@ieee.org;
russell.vela@gmail.com; hermann.krompholz@ttu.edu; andreas.neuber@ttu.edu;).
106
imperative for accurate temperature determination to correct the raw spectra with the apparatus’ spectral
sensitivity. In addition to plasma temperature, identification of spectral line, specifically sulfur and fluorine
from the dissociated SF6, is difficult using merely the raw spectra only. By identifying these lines in the
corrected spectra the plasma temperature can be more accurately determined.
Fig. 1: Cross-sectional view of the LTGS with the laser triggered gap on the right and the self-break
cascading section on the left. The cascade section is where envelope failures occur [2].
In the following, experimental results of pulsed volume breakdown in SF6 at pressures from 10
torr to 40 psig are presented. Besides fast voltage and current monitoring of the breakdown event, an
increased emphasis was put on imaging the event as well as gathering optical emission spectra (~200 nm to
700 nm). The role of UV emission due to a volume SF6 spark as potentially flashover hold-off reducing
cause is discussed. As much as possible, the small scale experiment was designed to reproduce at least
partly the conditions as they are found in the large 5 MV Z switch. The physical dimensions of importance
for the impact of volume breakdown on surface flashover are the backbone electrode gap distance of 9.5
mm and the distance between electrodes and envelope, 69.85 mm.
II. Experimental Setup

A. Test Chamber
In order to accurately simulate the conditions within the LTGS a test chamber has been constructed that
can support pressures from 10 torr absolute to 40 psig, see Fig. 2. In order to achieve this pressure range the
chamber is constructed of stainless steel. The chamber contains a variety of access ports to allow for the
various diagnostics as well as pump-down and fill line access. The high voltage connection to the chamber
inside is made by a custom fabricated feed through using an RG-220 cable fed through a compression port.
The cable acts as a holder for one of the electrodes, while a sliding assembly attached to the chamber holds
the ground electrode. This assembly enables the changing of the electrodes as well as the adjustment of the
gap spacing.
107
Fig. 2. Overview of the high pressure chamber (25.4 cm x 25.4 cm x 25.4 cm). Additional electrodes and
flashover surfaces are omitted for clarity.
The chamber also contains features to study the effects of volume discharge on adjacent dielectric
surfaces. Dielectric sample holders, complete with secondary excitation apparatus and electrodes, are
located above and below the primary discharge electrodes. The sample holders ensure the distance from the
primary gap to the dielectric is similar to that found in the LTGS. Future studies are aimed at investigating
the direct impact of volume breakdown on surface flashover initiation, as well as the effect on hold-off
voltages for various dielectric materials.
B. Electrical Diagnostics
Measurement of the breakdown requires instrumentation with high temporal resolution. For this reason, a
capacitive voltage divider with a sensitivity of 109 mV/kV and a rise time better than 1 ns was incorporated
on to the 52 ohm transmission line. Traveling wave current sensors are utilized for the current
measurements and have a sensitivity of 0.1 V/A with a ~1 ns rise time. All signals were then recorded using
a 500 MHz bandwidth, 2 GSa/s, Agilent Infinium oscilloscope.
C. Excitation
Transient excitation is achieved via an eight stage, 360 kV maximum pulse, Marx generator. The Marx
generator is remotely charged with a HV DC power supply capable of generating up to 125 kV at a
maximum of 2 mA. The Marx generator is triggered by a 40 kV high voltage pulser. When triggered, the
voltage pulse from the Marx generator travels down a 1 m transmission line, through a 52 Ω water resistor
(cupric sulfate), and then to the electrodes in the breakdown chamber. The Marx output voltage was
typically set to 300 kV, which corresponded to a coulomb transfer of ~200 µC. Fig. 3 shows a typical
voltage and current signal for volume breakdown.
108
Fig. 3: Current and voltage waveform of SF6 breakdown across a 9.52 mm gap at 40 psig. The voltage
signal is not corrected for the voltage sensor droop.
D. Optical Diagnostics
Imaging of the flashover event is achieved using a digital SLR camera with open shutter exposure and
neutral density filters with values ranging from 1 to 3. The SLR is primarily used with the high pressure
setup due to the EMI generated by the discharge that could damage more sensitive equipment. The camera
is fitted with a zoom lens and is positioned directly outside the breakdown chamber. The camera can either
be focused on the gap through the quartz view port or through a smaller viewport, in the top flange of the
chamber. The second method is utilized to allow for both imaging of the gap and collection of optical
emission spectra concurrently.
In order to help determine some of the processes contributing to the breakdown, a spectrograph is
utilized to analyze the emissions from the discharge. The spectrograph is an Oriel MS 257 ¼ meter imaging
spectrograph. It has toroidal mirrors designed to allow multiple vertical points, or fiber optic inputs, to be
diffracted simultaneously. This feature enables measuring spectra concurrently from multiple points. The
light emitted by the volume spark is collected utilizing cylindrical quartz lenses that focus the optical light
emission from three spatially distinct areas between the electrodes into three fiber optic cables leading back
to the spectrograph, see Fig. 4. In this way, spatially resolved light emission can be collected concurrently.
The spectrum is recorded with an intensified CCD camera located inside a screen room.
Fig. 4 Cross-section of the optical emission collection apparatus
In order to verify that UV is easily transmitted through high pressure SF6, a broadband UV source was
utilized to measure the spectral transmission through 12.7 cm of SF6 and N2, both at 40 psig, see Fig. 5.
Overall, the SF6 and the N2 reference curve match well, clearly indicating very little, if any, absorption for
109
SF6 in the range from 250 to 800 nm for the conditions found in the LTGS switch. It should be noted that
the signal below ~ 250 nm becomes very noisy, which is due to the decreasing light output of the
broadband source at shorter wavelength and the also smaller spectral response of the utilized spectroscopy
system at shorter wavelength.
The lack of UV absorption by SF6 is confirmed utilizing absorption cross section data available in the
open literature [6], [7]. Using Beer’s law for light absorption, it is calculated that 40 psig of SF6 will
transmit UV easily down to 160 nm, see Fig. 6. Note that the cut-off wavelength of air at around 180 nm is
well known and due to dissociation of the O2 molecule.
Since there is presently no known absorption resonance of SF6 between 165 nm and 250 nm, it should
hence be assumed that virtually all light between 160 and 800 nm is transmitted easily through up to 70 mm
of SF6 at 40 psig.
Fig. 5: Measured spectral response of the TTU spectroscopy system to a broadband Xe high pressure lamp
for transmission through 12.7 cm of SF6 or N2, respectively.
Fig. 6. Calculated light transmission for air (21% molecular oxygen) and SF6 through 70 mm of the denoted
gas at a pressure of 40 psig.
Previous studies revealed that the surface flashover path can be affected by
external UV light illumination via photoemission of electrons from the surface if the
wavelength range is extending down to ~ 320 nm and below (no noticeable impact for
110
light in wavelength range > 400 nm) [8]. Although these tests were done in a nitrogen
environment at atmospheric pressure across a polycarbonate surface (Lexan), there is
little reason to doubt that the wavelength range between 160 nm and 320 nm will play an
important role in surface flashover initiation under the conditions found in the LTGS. The
primary difference is expected from the photoemission yield that varies from material to
material.
While the above clearly shows that UV (~ 180…360 nm) and even VUV (< 180 nm) light can easily
propagate in the conditions found in the LTGS it still needs to be confirmed that the volume spark actually
produces light emission in the spectral range of interest, i.e. between 160 and 320 nm.
The spectral sensitivity of the spectroscopic system is primarily determined by the spectrograph grating,
the fiberoptics (UV grade fused silica fiber), and the camera’s photocathode quantum efficiency. An
inverse apparatus function accounts for these internal losses, and can be multiplied by the collected data to
yield the actual spectra. This function is determined using two calibration sources with known outputs. The
first source is a deuterium lamp (#63162 bulb from Newport) and the second is a high pressure xenon lamp
(#6254 bulb from Oriel). Both sources have a relatively well defined output in specific wavelength ranges
and the combined response was used to calculate the apparatus function. Any raw data recorded by the
ICCD is typically in counts, while the scaled data, which account for the losses in the collection system, is
labeled as relative intensity in arbitrary units. All spectra shown in the following were corrected using the
calibration curve given in Fig. 7.
Fig. 7: Scaling curve for all collected spectra, accounting for internal loss in the collection apparatus,
spectrograph, and imaging camera.
III. Results
A. Imaging
For accurate modeling of the vertebrae section of the LTGS, stainless steel electrodes with similar
geometries to those in the switch are utilized with 9.25 mm gap spacing. The gap is imaged concurrent to
the collection of spectra. These images are time integrated over the entire discharge. During testing the
camera shutter is held open and the aperture is set to a minimum. Flashovers at lower pressures yield
dimmer and more diffuse spark channels, with multi-channel sparks, as in Fig. 8, occurring roughly a fifth
of the time. As the pressure is increased the discharge becomes more intense and multi-channel formations
are still observed, as seen in Fig. 9 and 10. Multi-channel discharges are desired, as they decrease the
switching time of the gap.
111
Fig. 8: Time integrated image of volume breakdown between two stainless steel electrodes in SF6 at 10 torr.
No Neutral Density used. Charged electrode is on the left.
Fig. 9: Time integrated image of volume breakdown between two stainless steel electrodes in SF6 at 20
psig. Image attenuated by Neutral Density filter of ND = 2. Charged electrode is on the left.
Fig. 10: Time integrated image of volume breakdown between two stainless steel electrodes in SF6 at 40
psig. Image attenuated by Neutral Density filter of ND = 3. Charged electrode is on the left.
B. Breakdown Voltage
An additional effect of the increased pressure can be seen in the breakdown voltage, as shown in Fig. 11
below. The voltage at breakdown is recorded for pressures ranging from 10 Torr to 40 psig. It should be
noted that in the pressure regime below 20 psig, the breakdowns occur on the rising edge of the incident
pulse. As expected the increased hold-off voltages correspond to an increase in the delay time from pulse
application to breakdown of the gap, also shown in Fig. 11. Increased hold-off and delay times can be
attributed to increased electron energy loss processes, such as inelastic collisions, at higher pressures due to
the decrease in mean free path.
112
Fig 11: Breakdown voltage and delay time of a 9.52 mm gap in SF6 at increasing pressures with ~ 50 ns
risetime voltage pulse. Delay time is defined as the FWHM from the rise of the voltage pulse to the drop
due to breakdown.
C. Optical Emission Spectra

The spectra collected at different points along the discharge path are almost identical. Given the
similarities, only the spectrum from the middle fiber will be shown at each pressure. There are several
distinct lines present in the spectra, especially at lower pressures. In order to identify these lines the spectra
of various elements are calculated and compared to the experimentally collected spectra. In order to
calculate the spectra certain values are needed, such as the wavelength of the species specific spectral lines,
the level degeneracy, and the associated energy levels and Einstein coefficients for spontaneous emission.
This information is gathered primarily from the National Institute of Standards and Technology (NIST)
atomic spectra database [9]. From this information the relative line intensities are calculated with the
plasma temperature as the principal parameter. Finally, the lines are convolved to account for the effects of
the aperture of the spectrograph and natural spreading.
A test element such as fluorine, which was found to be present in the spectra, is used to determine the
temperature of the collected spectra. This is done by varying the temperature until the relative intensity of
the lines is identical to that of the measured spectra. The determined temperature is then used to calculate
the spectrum of various other elements. Once calculated, the various elemental spectra are compared to the
collected data.
Spectral line identification, as shown in Fig. 12, indicates the presence of sulfur and fluorine, as expected
in a SF6 discharge, in addition to strong nitrogen lines. It should be noted that the spark plasma is strongly
non-thermal at the lower pressures as the calculated spectra in Fig. 12 matched the experimental spectra for
an electronic temperature of Te = 30,000 K, N2 vibrational temperature of Tv = 3,000 K, and a N2 rotational
temperature of Tr = 1,000K. The calculated spectra will change significantly, i.e. perceivable with the
naked eye, if any of these temperatures is changed by ± 20%.
It should be noted that breakdown studies with high power microwave excitation of SF6 at 1 atmosphere
have revealed that, as expected, small amounts of nitrogen (few percent) in SF6 already show up
prominently in the emission spectrum of a spark, see Fig. 13.
113
Fig 12: Spectra peak identification using calculated atomic and diatomic spectra. Measured spectra
collected in SF6 environment at 100 torr.
Fig. 13: Optical emission spectra at 50 Torr with high power microwave excitation at 2.85 GHz. Top – 10:1
mixture of SF6 and N2, bottom – control pure N2.
Following the intensity of the N2 lines with respect to background pressure in Fig. 14 thru Fig. 16 reveals
that the N2 lines fade as the pressure increases, and are almost indistinguishable at 500 Torr. These lines
can be attributed to small leaks in the system that allows the outside environment, specifically air, into the
system, primarily at lower pressures. Even with a fixed trace amount of air in the system, the ratio between
SF6 to air will increase with the absolute SF6 pressure, thus causing fading of the N2 line features at the
higher pressures.
Another trend, especially at higher pressures, is the increasingly broadband nature of the spectra, with a
strong contribution in the UV region. This is significant as external UV illumination of the surface in
surface flashover experiments has been show to have a profound effect on surface flashover [8].
Illuminating the surface with a UV pulse significantly decreased the delay time, by up to 50%, from pulsed
excitation to flashover, in addition to decreasing variations in the delay time from shot to shot. The UV also
had a profound effect on the discharge path, resulting in the discharge following the surface as opposed to
the electric field line. This effect could easily contribute to the switch failure modes.
The broad features in the spectra at higher pressure are compared to black body radiation. In Fig. 16,
Plank’s equation,
114
2 ⋅ h ⋅ c2 1
ρ λ (T ) = (1)
λ 5
⋅
(
exp hc
kTλ
)−1
at 18,000 K, is used to model the broad features, with a high degree of correlation (h: Planck's constant, c:
speed of light, λ: wavelength, T: blackbody temperature). Hence, one might assume that the higher the
pressure the closer the spectral output will be to a black body radiator. However, the spark plasma is still
not completely thermalized at the 40 psig pressure as indicated by spectral line features superimposed onto
the broad background.
Nevertheless, the estimated temperature of ~18,000 K for the SF6 spark falls within the range of
centerline temperatures measured for a stationary arc at a pressure of 120 psig (15,500 K for 400 A and
27,000 K for 13,000 A) [10].
Fig. 14: Spectra collected from a volume discharge in SF6 at various pressures. The 100 Torr spectrum
exhibits more N2 contamination.
Fig. 15: Spectra collected from a volume discharge in SF6 at intermediate pressures.
115
Fig. 16: Spectra collected from a volume discharge in SF6 at high pressures.
Photoelectric emission from the surface increases with an increased photon energy, or decreased
wavelength [11], [12]. As the spectrum broadens, higher intensity is seen in the lower wavelength ranges
below 320 nm. Obviously, a higher UV output from the volume spark will liberate an increased number of
photoelectrons from the envelope surface.
Due to the transient nature of the switching process one must assume that the electric field amplitude at
the envelope is still considerably high even if the volume sparks in the cascade have formed. Hence, the
increasing number photoelectrons with the non-zero field at the envelope in the early moments of switching
are likely contributing to the observed LTGS envelope failure.
IV. Conclusions
Photoemission of electrons by UV photons impacting the envelope is one of the suspected reasons that
can initiate the unwanted tracking along the inside envelope surface in the LTGS. It is shown that a
significant amount of UV light is produced by the volume spark in high pressure SF6. The more intense
light output is in the wavelength regime below 320 nm; a regime that has been identified in previous
research as having a distinct effect on the spark path of flashover. The majority of species contributing to
the observed optical emission spectra between 230 and 700 nm has been identified. While it was possible to
estimate the volume spark’s electronic temperature (~35,000 K) for the lower pressures (50 torr) using
individual spectral line, at 40 psig the broadband features mask distinct atomic or molecular lines.
However, assuming black body radiation the temperature is estimated to be of the order of 18,000 K.
The absence of metal lines in any of the observed spectra is a result of the rather low coulomb transfer
and low current amplitude (~ 2 kA) in the test setup. It should be assumed that the spectral shape will
change with the current amplitude increased into the 100 kA regime in future research. Overall, our results
support the hypothesis of UV radiation contributing to the initiation of switch failure.
ACKNOWLEDGMENT
The authors would like to thank John Maenchen and Roy Jorgensen in the Electromagnetics and Plasma
Physics Analysis Department at Sandia National Laboratories for their support.
REFERENCES
[1] J. P. Corley, et. al., “Tests Of 6-MV Triggered Switches On APPRM At SNL”, Digest of Technical Papers in Pulsed Powere
Plasma Science, vol. 2, pp. 1778 – 1781, 2002.
[2] J. P. Corley, et. al., “Development/Tests of 6-MV Triggered Gas Switches at SNL”, Proc. 14th Pulsed Power Conference, vol. 2,
pp. 875 -878, 2003.
116
[3] A. M. Casanovas, J. Casanovas, V. Dubroca, F. Lagarde, and A. Belarbi, “Optical detection of corona discharges in SF6, CF4,
and SO2 under dc and 50-Hz as voltages,” J. Appl. Phys., vol. 70, pp. 1220-1226, 1991.
[4] A. Lemzadmi, N. Bonifaci, A. Denat, and M. Nemamcha, “Light emission from corona discharge in SF6 and SF6/N2 gas
mixtures at high pressure,” Eur. Phys. J. Appl. Phys., vol. 33, pp. 213-219, 2006.
[5] K. Fujii, M. Yamada, A. Tanaka, and K. Kurosawa, “Emission spectrum of partial discharge light in SF6 gas,” Proc. IEEE Intl.
Symp. on Elect. Insul., June 1992.
[6] C. Zetzsch, "UV absorption cross sections of sulfur hexafluoride and acetonitrile," presented at the International Ozone
Symposium, 1988, published in: Ozone in the Atmosphere (R.D. Boikov and P. Fabian, Eds.), DEEPAK Publishing (1989), 685-
689.
[7] J.E. Frederick and J.E. Mentall, "Solar irradiance in the stratosphere: Implications for the Herzberg continuum absorption of O2,
" Geophys. Res. Lett. 9, 461-464 (1982).
[8] K. P. Morales, J. T. Krile, A. A. Neuber, and H. G. Krompholz, “Pulsed Dielectric Surface Flashover in Atmospheric
Conditions,” IEEE Trans. on Dielectrics and Electrical Insulation, vol. 13, no. 4, pp. 803-809, Aug. 2006.
[9] NIST Atomic Spectra Database Lines Form, online at: http://physics.nist.gov/PhysRefData/ASD/lines_form.html
[10] J.J. Lowke and R.W. Liebermann, "Predicted Arc Properties in Sulfur Hexafluoride," J. Appl. Phys., vol. 42, 3532-3539, 1971.
[11] J. Tom, H. A. Verhaart, A. L. Verhage, and C. S. Vos, “Photo-emission of Charged Insulators in Insulating Gases,” Proceedings
of the 2nd International Conference on Conduction and Breakdown in Solid Dielectrics, pp. 301-307, Erlangen, 1986.
[12] A. A. Guzhov, and Yu. A. Shuba, “The Phototemission of Some Massive Insulators in the Vacuum Ultraviolet,” Optical
Technology, Vol. 38, No. 4, pp. 198-199, April 1971.
John T. Krile (S’98–M’06) was born in Terra Haute, Indiana. He graduated from Texas Tech University in
Lubbock, TX in 2006 with a Ph.D. degree in electrical engineering. He worked as a Research Assistant from
2002 to 2006 at the Center for Pulsed Power and Power Electronics. After graduating he has continued at
Texas Tech as a Senior Research Associate. His research interests include surface flashover physics, high-
power microwaves, explosively driven pulsed power, and compact pulsed power systems. He has published
more than 12 journal articles and conference papers.
Russell Vela (S’01) was born in McAllen, TX. He received his B.Sc. degree in electrical engineering from
Texas Tech University in Lubbock, TX in 2006. He is currently working as a Research Assistant for the Center
for Pulsed Power and Power Electronics at Texas Tech University while pursuing his M.Sc. degree in electrical
engineering.
Andreas A. Neuber (M’97-SM’03) was born in Aschaffenburg, Germany. He received the Dipl.-Phys. and
Ph.D., ME, degrees from the Darmstadt University of Technology, Germany, in 1990 and 1996, respectively.
He was a full time scientific employee at the institute of Energy- and Power Plant Technology, Darmstadt
University of Technology, from 1990 through 1996, in the area of nonlinear laser spectroscopy and chemical
reaction kinetics in combustion. He joined the Texas Tech University, Lubbock, in 1996 and is currently
Associate Professor in Electrical Engineering. His present research interests are high-power microwaves,
unipolar surface flashover physics, and explosive-driven pulsed power. He has published more than 90 journal
articles and conference proceedings paper.
Hermann Krompholz (SM ’84) received the Ph.D. degree in physics from Technical University Darmstadt,
Germany, in 1977. He was research associate at TU Darmstadt from 1977 to 1982, working on
nonequilibrium phenomena in high energy density plasmas. From 1982 to 1985, he joined Texas Tech
University, Lubbock, with activities in the areas of diffuse discharge opening switches and spark gap erosion.
After a brief stay at the Fraunhofer Institute for Laser Engineering and Technology in Aachen, Germany, he
rejoined the faculty at Texas Tech University in 1987, where he is now Professor of Electrical and Computer
Engineering/ Physics. His research interests include several aspects of pulsed power physics and technology,
with emphasis on the physics of electrical breakdown in gases, liquids, and along surfaces. He has published
about 110 journal articles and conference proceedings papers.
117
APPENDIX C
C PULSED DIELECTRIC SURFACE FLASHOVER IN AN SF6 ENVIRONMENT
John T. Krile, Member, IEEE, Russell Vela, Student Member, IEEE, Andreas A. Neuber,
Senior Member, IEEE, and Hermann G. Krompholz, Senior Member, IEEE
Abstract— A recently upgraded Laser Triggered Gas Switch, LTGS, at Sandia National Laboratories has
developed a failure mode resulting in the breakdown spark tracking to the inside of the containment envelope.
These breakdowns along the surface, or surface flashovers, degrade the performance of the overall switch,
causing the switch to pre-fire in the successive shot. In the following, experimental results of pulsed surface
flashover across different dielectric materials in SF6 primarily at atmospheric pressure as well as flashover and
volume breakdown in SF6 at pressures from 1.3 kPa to 365.4 kPa are presented. In addition to fast voltage and
current monitoring of the breakdown event, an increased emphasis was put on imaging the event as well as
gathering optical emission spectra (~200 nm to 700 nm) from it. As much as possible, the small scale experiments
were designed to reproduce at least partly the conditions as they are found in the large 5 MV switch. An effort
was made to determine what changes could be made to reduce the occurrence of surface flashovers, in addition
to some broadly applicable conclusions on surface flashovers in an SF6 environment.
Index Terms—Flashover, Dielectric materials, Electric breakdown, Ultraviolet radiation effects
I. INTRODUCTION
A S part of an extensive overhaul of the Z-machine, located at Sandia National Laboratories, several
Laser Triggered Gas Switches, LTGS, have been upgraded to operate in the 5 MV range [1]. With the
new requirements in switching voltage, the present LTGS design has been deemed inadequate due to
switch failure occurring after only a few shots. The primary failure mode was surface flashover in the high
pressure gas (SF6, > 365 kPa) on the inside of the polymethyl methacrylate, or PMMA, envelope. This
surface flashover occurred initially during the regular cascade switching process without affecting the
switch’s electrical performance. However, the switch tended to pre-fire in the successive shot once
flashover has occurred. The damage to the envelope has to be repaired before normal switch operation can
resume.
It has been shown previously that a significant amount of UV radiation was generated during
normal switch operation due to the volume breakdown between the cascade trigger electrodes [2]. In this
investigation, the spectral output of dielectric surface flashover was analyzed over a wide pressure range
(1.3 kPa to 365.4 kPa) and specifically compared to volume breakdown optical emission spectra taken at
365.4 kPa.
If an envelope flashover cannot be avoided, triggered due to, for instance, emission of
photoelectrons from the surface, it is prudent to identify materials that exhibit the least damage to a surface
flashover spark. In the present investigation, several electric material properties were compared, such as
flashover delay time and the likelihood of spark formation along the surface under fields with a non-
negligible component normal to the surface. The spectral output of a surface flashover along various
surfaces in SF6 was observed and high resolution imaging of the surface was utilized for surface damage
assessment. The surface was also examined for possible sulfur deposits, as seen in previous research [3].
Manuscript received January 8, 2007. This work was supported Sandia National Laboratories.
The authors are with Texas Tech University, Lubbock, TX 79409 USA (e-mail: jkrile@ieee.org;
russell.vela@gmail.com; andreas.neuber@ttu.edu; hermann.krompholz@ttu.edu).
118
II. Experimental Setup

A. Atmospheric Pressure Apparatus
While the emphasis of this research was to closely mimic the conditions within the LTGS, which is
typically pressurized to 365 kPa of SF6, a much wider pressure range was covered to reveal possible
pressure dependent tendencies. For instance, some initial testing was performed in a small surface flashover
chamber, capable of maintaining an SF6 environment at atmospheric pressures. This chamber as well as a
second higher pressure chamber were excited with an 8 stage Marx generator with up to 45 kV charging
voltage per stage. For this atmospheric chamber, the Marx output voltage was typically set to 90 kV with a
50 ns rise time, which led to a charge transfer of 60 µC.
The atmospheric testing setup was an adaptation from a previous project in which extensive field
simulations and impedance matching between the transmission line and electrode feed-throughs was done
to ensure minimal signal distortion and system robustness [2]. The design utilizes a coaxial geometry of 52-
Ω impedance throughout the setup, see Fig. 1. The chamber was constructed of half inch thick
polycarbonate to allow for easy optical access. Dotted lines in Fig. 1 represent a removable outer conductor
which contains a small hole for imaging of the gap. Two ports, located at the base of the chamber, allowed
for a continuous flow of SF6. The chamber was kept at a slight overpressure during testing to prevent
outside contamination.
Fig. 1. Overview of the atmospheric chamber setup. The dotted lines represent the removable outer
conductor and the dashed lines represent fiber optic cables.
B. High Pressure Test Apparatus

A high pressure chamber was utilized that enabled adjusting the pressure from the kilopascal range up to
365.4 kPa, see Fig. 2. In this case, the Marx output voltage was increased to 320 kV (~ 220 µC charge
transfer) for the high pressure chamber to facilitate reproducible flashover.
119
Fig. 2. Overview of the high pressure chamber (25.4 cm x 25.4 cm x 25.4 cm). Additional electrodes and
flashover surfaces are omitted for clarity.
For the presented research, the second chamber was configured with a set of electrodes identical to the
atmospheric setup. The chamber was pressure tested up to 500 kPa, and a minimum pressure in the
millipascals range was achieved. The ability to evacuate the chamber was useful in eliminating the possible
effects of environmental contaminations, specifically nitrogen gas, on the flashover processes. The chamber
has been equipped with several ports to allow for easy access by the various diagnostics. The gap was
excited with the same Marx generator discussed previously.
C. Electrical Diagnostics
Measurement of the breakdown required instrumentation with high temporal resolution. For this reason,
a capacitive voltage divider with a sensitivity of 109 mV/kV and a rise time better than 1 ns was
incorporated on to the 52-Ω transmission line. Traveling wave current sensors were utilized for the current
measurements and have a sensitivity of 0.1 V/A with a ~1 ns rise time. All signals were then recorded using
a 500 MHz bandwidth, 2 GSa/s, Agilent Infinium oscilloscope.
D. Optical Diagnostics
The flashover event was imaged using an Andor Technology Intensified CCD DH-734 camera with
nanosecond temporal resolution. This camera was fitted with a zoom lens and positioned directly outside
the flashover chamber. In the atmospheric chamber the camera was focused on the dielectric gap through a
small hole in the detachable outer conductor. Optical gate times of 3 to 300 ns were utilized for the
flashover experiments. In the high pressure chamber images of the flashover gap could either be taken
through the quartz view port or through a secondary view port in the upper flange of the chamber. The
second method was utilized to allow for both imaging of the gap (time integrated with digital camera) and
collection of optical emission spectra concurrently.
Spatially resolved optical emission spectra were collected using a specially designed apparatus, see Fig.
3, consisting of three quartz (Suprasil) cylindrical rod lenses. These lenses were positioned to collect light
from three rectangular shaped regions in the gap and focus it into three fiber optic cables which were fed
into a spectrograph, Oriel MS 257. This spectrograph was an imaging spectrograph that had toroidal
mirrors designed to allow multiple vertical points, or fiber optic inputs, to be diffracted at one time. This
120
feature was used to collect spectra from different positions along the discharge path at the same time.
Fig. 3: Cross-section of the optical emission collection apparatus.
For quantitative optical emission spectroscopy, the spectral response of the complete spectral apparatus
has to be considered. In this case, the spectra collected by the CCD will be a function of the spectral
sensitivity of the entire system, which is primarily determined by the spectrograph grating, the fiber optics
(UV grade fused silica fiber), and the camera’s photocathode quantum efficiency. Two light sources with
known spectral output were utilized to determine the inverse spectral response of the apparatus, which was
given in detail elsewhere [2]. Note that all spectral data shown in this manuscript was corrected with the
spectral response.
In addition to the sensitivity of the collection apparatus, the attenuation of the environment must also be
considered. However, the expected absorption due to SF6 for distances within the chamber or LTGS
conditions has been shown negligible above 160 nm through the visible [2]. Hence, no correction due to
“cold” gas absorption has been made for the wavelength range between 200 nm through 700 nm.
E. Electrode, Dielectric Geometry

The electrode geometry, see Fig. 4, was chosen to produce an electric field with a strong electric field
component normal to the surface. Both electrodes are embedded approximately 13.7 mm into the dielectric
material. Electrostatic field simulations were done on the electrode geometry using Ansoft’s Maxwell 3D
simulation program.
Fig. 4. Maxwell 3D simulation of electric field vector for the angled electrode geometry in a cut plane normal to
the dielectric surface. Left electrode charged to 72 kV, right electrode grounded. Gap spacing of 7.87 mm.
Two different surface geometries were tested. Some dielectric samples had surfaces that were not altered
or machined except for the electrode slots. These samples were utilized for optical spectra collection. In
order to increase the surface path length between the electrodes a groove was machined along the surface of
the sample, see Fig. 5.
121
Fig. 5. Dielectric sample geometry and dimensions used for testing. All units in millimeters.
In both cases the overall dimensions of the dielectric samples remained the same. Due to different
machining methods, the groove depth varied slightly from sample to sample, however, the basic field shape
was the same for all grooved samples.
Preparation of the samples involved removing the protective film used during machining and gently
cleaning off any residual adhesive with cyclohexane and lint-free wipes. The sample was then inserted into
the test chamber with powder-free latex gloves in order to prevent contamination of the surface. Before
testing, the chamber was evacuated to 0.01 Pa to efficiently remove background gases (humidity) from the
chamber and at least to some extend from the chamber walls, and the dielectric surface. In a final step, the
chamber is backfilled with SF6 to the desired pressure.
III. Results and Discussion

A. Reference Waveforms
For each flashover event, the camera gate (when applicable), current, and voltage waveforms were
acquired and recorded. A representative set of waveforms for a typical surface flashover event in the
atmospheric chamber is depicted in Fig. 6. The typical flashover voltage waveform was characterized by a
rise in voltage as the voltage pulse is applied to the gap, followed by a sharp drop in voltage as flashover
occurs. The collapse of the voltage due to flashover was accompanied by the simultaneous rise in current.
Note that the first portion of the current (or voltage waveform) remained unchanged from shot to shot. That
is, the small hump in current is due to charging of the capacitance associated with the gap and the
supporting structure. This charging is considered complete when the current returns to zero, indicated by
the first dashed line in Fig. 6. All recorded flashovers occurred after fully charging of the gap, that is after
the first dashed line. The Marx output a negative voltage pulse and had a maximum coulomb transfer of
~220 µC at maximum output voltage. The breakdown delay time was defined as the full width at half
maximum from the voltage rise to the voltage drop due to breakdown. The output voltage was adjusted to
ensure no flashover events occur during the voltage rise. The voltage and current signals typical in the high
pressure chamber were similar to Fig. 6; however, since the gap has a higher breakdown voltage at higher
SF6 pressure, the Marx output voltage was significantly increased to reliably achieve a breakdown.
122
Fig. 6. Representative waveforms of current and voltage signals for flashover in SF6 at 90.6 kPa. Dashed lines
represent the beginning and end of the camera exposure.
B. Imaging the Discharge

For imaging of the flashover event with the ICCD camera, the surface was slightly angled to provide a
side-on view of the flashover path. Given that the electric field leads away from the surface it was desired
to know whether the discharge would form along the field lines, as in volume breakdown, or along the
surface. Following the surface would indicate the surface played some role in the spark development and
the potential for increased surface erosion. If the surface was important for the flashover path then varying
the dielectric material could have an impact on the performance of the switch by limiting undesirable
surface tracking. Fig. 7 and 8 show examples of the varying degrees of surface tracking observed.
Fig. 7: Grooved Teflon sample at 90.6 kPa of SF6. Gap = 10.53 mm, Tdelay = 92 ns, Vpulse ≈ 91 kV. White
lines indicate the location of the surface and electrodes.
Fig. 8: Grooved Teflon sample at 90.6 kPa of SF6. Gap = 9.48 mm, Tdelay = 307.3 ns, Vpulse ≈ 90 kV. White
lines indicate the location of the surface and electrodes.
It became apparent that surface tracking varies widely depending on the dielectric material, see Table 1.
From the table, it can be seen that Teflon produced the largest percentage of flashover liftoffs, which would
make it the preferred choice in applications where surface tracking or erosion is undesirable.
The epoxies listed in Table 1 were selected by Sandia in order to determine the feasibility of producing
the envelope material in-house. The epoxies consist of an Epon 826 resin with either a D-400
123
(polyoxypropylene diamine) or T-403 (polyoxypropylene triamine) curing agent. The second material also
incorporated a degassing agent, KF-865.
While data on quantum yields for photoemission is relatively scarce, especially for custom epoxy
compounds, it is generally accepted that Teflon’s photoemission coefficient is smaller than most dielectrics
[4]-[5]. As stated previously, a decrease in surface liftoffs indicated a lesser interaction between the
discharge and the surface. It is believed that, at least in part, this interaction involves the UV radiation
emitted by the spark itself producing photoelectrons from the surface [6]. As such, the quantum yield of the
envelope should be minimized either through careful material selection or the application of a UV
absorbing coating to prevent surface flashover.
Table 1. Various dielectrics and there percentages of liftoff.

Materials Lift-off Percentages
Teflon 50%
Polycarbonate 0%
Epon 826/T-403 13%
Epon 826/D-400/KF-865 18%
C. Dielectric Sample Degradation

In addition to imaging the discharge, the dielectric samples were examined to assess the damage or
tracking produced by the discharge. Any damage to the surface could affect subsequent flashover events.
This phenomenon is known as conditioning the surface. Each dielectric was photographed using an optical
microscope. This process documented any visible damage on the surface of the material. Next a scanning
electron microscope (SEM) was utilized to achieve better resolution in areas of interest. This proved useful
when damage could not be seen or was vague under a light microscope with its limited magnification.
Surface tracking damage was clearly visible using the optical microscope for some materials, as seen in
Fig. 9. Fig. 10 shows that, although Teflon flashover did follow the surface (≈ 50% of the time), the
flashover did not leave visible tracking on the surface, as opposed to the other materials.
Fig 9: Smooth surface dielectric sample 826/D400/KF-865 after 38 discharges (1 atm SF6, 30 µC). Similar
to all epoxy mixtures tested. Gap distance of ~9.9 mm.
124
Fig 10: Grooved Teflon after 10 discharges (1 atm SF6, 30 µC). No damage visible other than groove
machining. Gap distance of ~10.5 mm.
While no notable damage was found on Teflon with the SEM, the epoxy, see Fig. 11, exhibited
significant changes. Since the SEM was sensitive to both topography and composition, the observed pattern
could have been simply due to deposits on the surface or due to changes in the material, including charging
of the material just below the surface.
Fig 11: SEM image of smooth surface dielectric sample 826/D400/KF-865 after 38 discharges (1 atm SF6,
30 µC). The flashover region is denoted by the dark area in the center of the image which lacks significant
machining marks. The region appears to be altered compared to the surrounding epoxy surface. Imaged
area is ~2.5 mm wide.
The same materials were tested in the high pressure chamber and displayed similar damage characteristic
as in the atmospheric chamber. Again, Teflon showed no signs of visible damage, unlike the epoxy and
polycarbonate samples. Examination of the polycarbonate sample with the SEM revealed significant
damage at the electrode surface interface, as seen in Fig. 12. The polycarbonate was flashed 40 times, five
times at each of the following pressures: 1.3 kPa, 2.7 kPa, 6.7 kPa, 13.3 kPa, 33.3 kPa, 90.6 kPa, 227.5 kPa,
and 365.4 kPa. Since changes in the structure of the material would show up in the SEM images as well, it
could again be speculated that the internal cracking and discoloration of the material was due to a change in
the composition of the material.
125
Fig 12: SEM image of smooth surfaced optical grade polycarbonate. View is into the hole for the electrode,
the upper surface is the where the flashover occurred. The cracks visible in the surface are located on the
edge of the surface corresponding to the electrode/dielectric interface. Discharges occurred 40 times at
eight various pressures (Varied pressure SF6, 108 µC).
In addition to the damage observed, spherical deposits were observed on the surface of the epoxy, as
seen in Fig. 13. Utilizing the SEM, X-Ray spectroscopy analysis was performed on the deposits revealing
that they consisted primarily of sulfur. Subsequent analysis showed that sulfur deposits also were present in
the crack in the polycarbonate surface, seen in Fig. 12. The sulfur deposits were found imbedded in the
surface as opposed to on the surface, as a settled byproduct of the discharge would be. These discharge
byproducts were eventually found for all dielectric materials tested with the exception of Teflon, although
in some of the materials sulfur deposits took the form of flakes, rather than spheres. It is possible that sulfur
exist on the Teflon as well, however locating it is difficult due to the nearly identical coloration assigned to
sulfur deposits and Teflon surface by the SEM. These deposits further contribute to the surface
conditioning and showed that the background gas could also play a role conditioning.
Fig 13: SEM Image from dielectric material 826/D400/KF-865 used for X-Ray Spectroscopy analysis (90.6
kPa SF6, 30 µC).
D. Flashover Delay Time

The study of the time delay for breakdown was most revealing for the pulsed voltage hold-off
capabilities of different materials. All materials were tested under identical environmental conditions, 90.7
kPa of SF6. In the ideal case the field strength of all tested samples would be identical, however, due to
minor gap adjustments (± 1.0mm) necessary to fall within the atmospheric breakdown chamber’s voltage
limits (90kV), the field strength varied between samples. As a result, a direct comparison of the time delays
was not straightforward and had to be approached with caution. The average delay times of the tested
materials are depicted in Fig. 14, sorted by applied field strength. All delay times have been averaged over
more than 60 shots per material type. The figure can be interpreted such that if a material exhibits a longer
breakdown delay time at a larger applied field than another material, its hold-off strength can be considered
superior.
126
Fig 14: Dielectric material comparison of average time delays. Samples arranged by field strength (asterisk
indicates one sample only tested).
Despite the varying field strengths, Fig. 14 indicates that Teflon had a longer breakdown delay time even
at a higher applied field when compared to the Lexan and the epoxies. The one sample of Plexiglas
(material taken directly from an unused Z switch envelope) exhibited the best performance at 1 atmosphere
of SF6, however, this value has to be treated with some caution as only a single sample could be tested here.
Nevertheless, the change of flashover delay time of Plexiglas was very pronounced, and in general was
much larger than any of the sample to sample variations for the other materials. Polycarbonate, Teflon, and
Plexiglas have dielectric constants between ~ 2 to 3, while the epoxy based materials fell out of this range
with a relative permittivity of ~ 5, which could help explain their poor performance. A higher dielectric
constant could lead to increased electron emission at the triple point. Given their propensity for surface
conditioning and low hold-off times, the epoxy based materials were not suited for high voltage switch
envelopes.
E. Optical Emission Spectra

Using the optical emission collection apparatus described earlier, spectra was collected from three points
along the spark channel, see Fig. 15. The first lens collected light from just in front of the negative high
voltage (charged) electrode, the second observed the middle of the gap, and the third observed just in front
of the grounded electrode. In general, the spectra from different positions in the gap were quite similar over
the entire length of the gap, as shown in Fig. 15. However, one notable difference was the increased UV
content just in front of the cathode. This was significant given the suspected interaction of the UV radiation
and the dielectric surface.
Most spectral peaks have been identified, including sulfur, fluorine, hydrogen and carbon, see Fig. 16. A
temperature of around 30,000 K resulted in the best fit between calculated and experimental spectra. The
carbon and hydrogen was of particular significance because it indicates the discharge was removing
material from the surface and exciting it. Note that the hydrogen could also be attributed to moisture, not
fully removed while under vacuum, being desorbed from the surface. In either case the discharge was
altering or conditioning the surface and potentially lowering its flashover hold-off ability.
127
Fig 15: Spectra collected from a surface flashover of a polycarbonate surface in an SF6 environment at 6.7 kPa.
Spectra are collected from three points along the discharge path, as denoted.
Fig. 16: Spectra peak identification using calculated atomic spectra. Measured spectra collected in SF6
environment, across a polycarbonate sample, at 6.7 kPa. The calculated spectra are given for a temperature
of 30,000 K.
In order to isolate any peaks that have not yet been identified, and to illustrate the underlying, broadband
spectra, the identified peaks were added together and subtracted from the original signal. The convolution
profiles used in the calculation of the atomic spectra were adjusted to match those of the actual setup. Fig.
17 shows the result of this subtraction. It can be seen that most of the peaks in the 450 nm to 700 nm range
have been identified and removed leaving an approximation of the broadband spectra, which was most
likely due to contributions from excited SF6 molecules and recombination.
128
Fig. 17: Calculation of remaining spectra after identified atomic spectra lines are removed. Measured
spectra collected in SF6 environment, across a polycarbonate sample, at 6.7 kPa. The calculated spectra is
given for an temperature of 30,000 K.
.
A series of shots were performed while varying the pressure of SF6 in the chamber, Fig. 18 shows
selected results of the 1.3 kPa to 365.4 kPa series of tests. Due to the consistency of the spectra along the
discharge path, only the center of the gap is shown at each pressure. All presented spectral data has been
scaled by the correction factor to account for internal losses in the collection system discussed previously.
At increasing pressures the intensity of the singly ionized fluorine relative to the other atomic/ionic lines
was suppressed followed by the weakening of all lines compared to the broadband features and the singly
ionized carbon line.
Fig. 18: Spectra collected from a surface flashover of a polycarbonate surface in an SF6 environment at
various pressures. Spectra are collected from middle of the gap at each noted pressure.
Overall, the pressure dependence of the flashover spectra was similar to what was previously measured
for volume breakdown [2]. That is, the measured optical emission spectra start out at low pressures as line
spectra that are, with increasing pressure, increasingly masked by broadband radiation, as indicated by Fig.
18. The only spectral line in flashover that shows up dominantly at 365.4 kPa was the singly ionized
Carbon line at 250 nm, see bottom of Fig. 19. This clearly shows the erosion of the surface as no carbon
line was present at the higher pressures for volume breakdown under similar conditions, as shown in Fig.
19.
129
Fig. 19: Comparison of spectra collected from a volume breakdown and a surface flashover of a
polycarbonate surface in an SF6 environment at 365.4 kPa. Spectra are collected from middle of the gap in
both cases.
IV. Conclusions
Several aspects of surface flashover along dielectric insulators have been studied in SF6 environments.
Imaging of the discharge along a dielectric surface revealed a material dependence on the discharge path.
Even with an electric field leading away from the surface discharges continued to preferentially follow the
surface for most materials, i.e. polycarbonate and the epoxies, indicating a strong interaction between the
surfaces and the discharge. This interaction was believed to involve the productions of photoelectrons by
the surface while exposed to UV radiation generated by the forming discharge, which existence was
confirmed by the collected optical emission spectra. Teflon showed the weakest interaction with the
surface, following the surface only 50% of the time. This indicated that Teflon, or a material with similar or
lower quantum yields for photo emission, would be the preferred choice for minimizing undesired envelope
flashovers.
In addition to imaging of the discharge, post-mortem imaging of the samples made the differences in
surface conditioning apparent. For instance, the epoxy and polycarbonate exhibited very distinct visible
surface damage, as opposed to Teflon that exhibited almost no visible damage. As surface conditioning
reduces flashover fields of successive flashover events, Teflon was again shown superior to the other
materials. This was further supported by the flashover delay time measurements, which indicated that
Teflon would hold off flashover longer under greater field stress than the epoxies.
Comparing volume breakdown and surface flashover optical emission spectra revealed the presence of a
strong carbon line in the UV for surface flashover at 365.4 kPa. Since this line was absent in the volume
breakdown spectra, it serves as an in-situ indicator for surface erosion for hydrocarbon based materials.
This also implies a measure of caution should be taken when using hydrocarbon based materials in
envelope construction.
The occurrence of sulfur on the dielectric surface, as recorded in literature, was consistent with the
prevalent appearance of sulfur ionic lines in the flashover optical emission spectra (SII and SIII). The
question remains as to whether these deposits are a cause or result of the poor hold-off characteristics of the
materials. Future testing with secondary surfaces placed in the chamber to measure hold-off potential after
collecting any deposits created by an initial flashover event should help clarifying the role of the sulfur
deposits.
ACKNOWLEDGMENT
The authors would like to thank John Maenchen and Roy Jorgensen at Sandia National Laboratories for
their support.
130
REFERENCES
[1] J. P. Corley, et. al., “Development/Tests of 6-MV Triggered Gas Switches at SNL”, Proc. 14th Pulsed Power Conference, vol. 2,
pp. 875 -878, 2003.
[2] J. K. Krile, R. Vela, A. A. Neuber, H. G. Krompholz, “Pulsed Volume Breakdown of SF6 at High Pressures,” IEEE Trans. on
Plasma Sciences, submitted for publication.
[3] H. R. Hiziroglu, J.M. Kokosa, and A. Przyjazny, “Discharge By-Products of SF6 in the Presence of Polymeric Materials,”
Annual Report of IEEE Conference on Electrical Insulation and Dielectric Phenomena, pp. 877, October 1994.
[4] J. Tom, H. A. Verhaart, A. L. Verhage, and C. S. Vos, “Photo-emission of Charged Insulators in Insulating Gases,” Proceedings
of the 2nd International Conference on Conduction and Breakdown in Solid Dielectrics, pp. 301-307, Erlangen, 1986.
[5] A. A. Guzhov, and Yu. A. Shuba, “The Phototemission of Some Massive Insulators in the Vacuum Ultraviolet,” Optical
Technology, Vol. 38, No. 4, pp. 198-199, April 1971.
[6] K. Morales, J. Krile, A. Neuber, and H. Krompholz, “Dielectric Surface Flashover at Atmospheric Conditions with Unipolar
Pulsed Voltage Excitation,” IEEE Trans. Dielectr. Electr. Insul., August 2007, to be published.
John T. Krile (S’98–M’06) was born in Terra Haute, Indiana. He graduated from Texas Tech University in
Lubbock, TX in 2006 with a Ph.D. degree in electrical engineering. He worked as a Research Assistant from
2002 to 2006 at the Center for Pulsed Power and Power Electronics. After graduating he has continued at
Texas Tech as a Senior Research Associate. His research interests include surface flashover physics, high-
power microwaves, explosively driven pulsed power, and compact pulsed power systems. He has published
more than 12 journal articles and conference papers.
Russell Vela (S’01) was born in McAllen, TX. He received his B.Sc. degree in electrical engineering from
Texas Tech University in Lubbock, TX in 2006. He is currently working as a Research Assistant for the Center
for Pulsed Power and Power Electronics at Texas Tech University while pursuing his M.Sc. degree in electrical
engineering.
Andreas A. Neuber (M’97-SM’03) was born in Aschaffenburg, Germany. He received the Dipl.-Phys. and
Ph.D., ME, degrees from the Darmstadt University of Technology, Germany, in 1990 and 1996, respectively.
He was a full time scientific employee at the institute of Energy- and Power Plant Technology, Darmstadt
University of Technology, from 1990 through 1996, in the area of nonlinear laser spectroscopy and chemical
reaction kinetics in combustion. He joined the Texas Tech University, Lubbock, in 1996 and is currently
Associate Professor in Electrical Engineering. His present research interests are high-power microwaves,
unipolar surface flashover physics, and explosive-driven pulsed power. He has published more than 90 journal
articles and conference proceedings paper.
Hermann Krompholz (SM ’84) received the Ph.D. degree in physics from Technical University Darmstadt,
Germany, in 1977. He was research associate at TU Darmstadt from 1977 to 1982, working on
nonequilibrium phenomena in high energy density plasmas. From 1982 to 1985, he joined Texas Tech
University, Lubbock, with activities in the areas of diffuse discharge opening switches and spark gap erosion.
After a brief stay at the Fraunhofer Institute for Laser Engineering and Technology in Aachen, Germany, he
rejoined the faculty at Texas Tech University in 1987, where he is now Professor of Electrical and Computer
Engineering/ Physics. His research interests include several aspects of pulsed power physics and technology,
with emphasis on the physics of electrical breakdown in gases, liquids, and along surfaces. He has published
about 110 journal articles and conference proceedings papers.
131
PERMISSION TO COPY
In presenting this thesis in partial fulfillment of the requirements for a master’s
degree at Texas Tech University or Texas Tech University Health Sciences Center, I
agree that the Library and my major department shall make it freely available for research
purposes. Permission to copy this thesis for scholarly purposes may be granted by the
Director of the Library or my major professor. It is understood that any copying or
publication of this thesis for financial gain shall not be allowed without my further
written permission and that any user may be liable for copyright infringement.
Agree (Permission is granted.)
Russell Vela_____________________________________ 07/27/07__________

Student Signature Date
Disagree (Permission is not granted.)
_______________________________________________ _________________
Student Signature Date

Vela Russell Thesis

Uploaded by

Copyright:

Available Formats

Vela Russell Thesis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vela Russell Thesis

Uploaded by

Copyright:

Available Formats

ELECTRICAL BREAKDOWN OF SF6

RUSSELL VELA, B.S.E.E.

Submitted to the Graduate Faculty

their interest, questions, and invaluable teachings.

most importantly friendship.

hope our paths will cross again.

3.1 Experimental Test Chamber ............................................................... 35

4.2.2 Volume Breakdown Spectroscopy Analysis........................ 73

2.1: Common insulators used in experimentation and their corresponding work

2.2: Variables for Fowler-Nordheim equation.....................................................................8

2.3: Variables for photoelectric emission equation 2.2........................................................9

2.4: Variables for the Gerald Dionne model......................................................................10

2.7: Variables for describing electron amplification in Equation 2.10 ..............................21

3.1: Overall delivery system component characteristics....................................................38

1.2: Example of normal operation of backbone electrodes (right) [2].................................3

1.3: Example of failure in backbone electrode section [2] ..................................................4

2.7: Recombination coefficient (ion-ion) in air at 20 ºC [15]............................................18

2.9: Current vs. voltage relationship developed by Townsend [15] ..................................20

2.11: Diagram of the stages of streamer formation [5]......................................................24

2.12: The transition of a Townsend (generation) type of breakdown to a streamer

2.13: Diagram of secondary electron emission avalanche.................................................26

2.15: Minimum breakdown voltages for various gases [15]..............................................29

2.17: (α-η)/p ~ E/p relationship in SF6 [15] .......................................................................31

3.2: Overall circuit diagram of testing setup......................................................................34

3.3: Secondary setup surface flashover electrode configuration .......................................35

3.5: Example of surface flashover to chamber wall...........................................................36

3.6: Maxwell 3D simulations showing original field enhancements (left) at chamber

3.7: Charging setup of overall system................................................................................38

3.10: Final magnetic switch assembly in a reinforced oil vessel.......................................43

3.12: Gas flow configuration for experimental setup ........................................................47

3.13: Cross-sectional rendering of a capacitive voltage divider ........................................49

3.14: Triggering setup used for testing ..............................................................................51

3.19: Top view of fiber optic lens layout...........................................................................54

3.20: Demonstration of difference in imaging and non-imaging spectrographs,

3.21: Schematic of multi-track spectrograph operation.....................................................56

3.23: Dielectric sample geometry and dimensions used for testing...................................58

3.27: Secondary setup electrode configuration..................................................................61

4.14: Time integrated image of surface flashover of a Lexan (polycarbonate) surface

4.15: Time integrated image of surface flashover of a Lexan (polycarbonate) surface

4.24: Comparison of spectra collected from a volume breakdown and a surface

4.30: SEM image of smooth surface dielectric sample 826/D400/KF-865 cured at

B.4: Cross-section of the optical emission collection apparatus.....................................109

B.5: Measured spectral response of the TTU spectroscopy system to a broadband

B.14: Spectra collected from a volume discharge in SF6 at various pressures.

B.15: Spectra collected from a volume discharge in SF6 at intermediate pressures........115

B.16: Spectra collected from a volume discharge in SF6 at high pressures.....................116

C.3: Cross-section of the optical emission collection apparatus...........................................121

C.6: Representative waveforms of current and voltage signals for flashover in

C.9: Smooth surface dielectric sample 826/D400/KF-865 after 38 discharges

C.10: Grooved Teflon after 10 discharges (1 atm SF6, 30 µC). No damage

C.11: SEM image of smooth surface dielectric sample 826/D400/KF-865

C.12: SEM image of smooth surfaced optical grade polycarbonate. View is

C.14: Dielectric material comparison of average time delays. Samples arranged

C.15: Spectra collected from a surface flashover of a polycarbonate surface in an

C.16: Spectra peak identification using calculated atomic spectra. Measured

C.17: Calculation of remaining spectra after identified atomic spectra lines

C.18: Spectra collected from a surface flashover of a polycarbonate surface in

C.19: Comparison of spectra collected from a volume breakdown and a surface

In the following experimental results of pulsed surface flashover, different