Вы находитесь на странице: 1из 9

IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 34, NO.

2, APRIL 2006

467

Conduction and Breakdown Mechanisms


in Transformer Oil
Michael Butcher, Student Member, IEEE, Andreas A. Neuber, Senior Member, IEEE,
Michael D. Cevallos, Member, IEEE, James C. Dickens, Senior Member, IEEE, and
Hermann Krompholz, Senior Member, IEEE

AbstractWith a fast coaxial test setup using high speed electrical and optical diagnostics, prebreakdown current pulses and
shadowgraphy images are measured for direct current (dc) breakdown in Univolt 61 transformer oil. Also, dc currents across the
gap are measured using a high sensitivity electrometer. The conduction and breakdown mechanisms in transformer oil as function
of applied hydrostatic pressures are quantified. Together, this information provides data on the development of current flow in the
system. We have identified three stages in the conduction process
prior to breakdown for highly nonuniform fields. Stage 1 is characterized by a resistive current at low fields. Increasing the applied
electric field lowers the effective barrier at the metal/dielectric interface allowing a tunneling mechanism to begin, leading to the
rapid rise in the injection current observed in stage 2. In stage
3, at high fields, the current reaches space charge saturation with
an apparent mobility of 3 10 3 cm2 V s prior to breakdown.
The processes of final breakdown show a distinct polarity dependence. A strong pressure dependence of the breakdown voltage is
recorded for negative needle/plane breakdown; a 50% reduction
in breakdown voltage is observed when the hydrostatic pressure is
lowered from atmospheric pressure to hundreds of mtorr. Positive
needle discharges show a reduction of only about 10% in breakdown voltage for the reduced pressure case. Weak pressure dependence indicates the breakdown mechanism does not have a strong
gaseous component. We will discuss possible links between conduction current and dc breakdown.
Index TermsBubble formation, conduction mechanisms, liquid
breakdown, mobility, pressure effects, temperature effects.

I. INTRODUCTION

OR more than 50 years, studies of the mechanisms involved with the development of liquid breakdown have
lead to a deeper understanding while still not achieving a
definitive model explaining the initial breakdown process.
The amount of research in the area of liquid dielectrics has
increased in the last 1015 years. Insulating fluids such as
transformer oils are critical components to high voltage and
pulsed power systems. The desire for compact pulsed power
systems drives research to develop a model that is capable of
describing the basic processes of liquid breakdown, similar to
the promotion of the model for gases almost a century ago.
Manuscript received March 23, 2005; revised January 31, 2006. This work
supported by the Compact Pulsed Power MURI Program funded by the Director
of Defense Research and Engineering (DDRE) and managed by the Air Force
Office of Scientific Research (AFOSR).
The authors are with the Center for Pulsed Power and Power Electronics,
Electrical Engineering Department, Texas Tech University, Lubbock, TX
79409 USA (e-mail: mbutcher@ieee.org; andreas.neuber@ttu.edu; mic.cevallos@p3e.ttu.edu; jdickens@coe.ttu.edu; hermannk@coe.ttu.edu ).
Digital Object Identifier 10.1109/TPS.2006.872487

Improvements over the last decade, increased energy density


capacitors and solid state switch gear, have further fueled interest in liquid dielectrics. Higher relative permittivity of liquid
dielectrics, compared to gases, is of advantage for use in these
higher energy density systems. Also, the ability to conform to
complex geometries and maintaining the self-healing/recovery
capabilities typical of gas systems makes liquids suitable for
use as switching medium. Therefore, a thorough understanding
of liquid dielectrics including mechanisms of charge conduction and processes leading to breakdown is of vital interest to
advancing the technology of high voltage and pulsed power
systems.
Liquid breakdown involves a unique level of complexity compared to gas or solid dielectric breakdown. Physical characteristics, such as fluid viscosity, electro-convection, temperature,
density, and pressure complicate the analysis and modeling of
the breakdown and conduction mechanisms. Most liquids have
characteristics more similar to solids than gases, such as high
permittivities, high densities, and weak molecular order, while
maintaining the flowing quality of a gas.
Research on conduction in liquid dielectrics began many
years ago because of its importance for applications as an
insulator. The effects of age, contamination, and gas content
impact the overall quality of the oil, affecting the conduction
and breakdown mechanisms [1][6]. Characterizing breakdown of liquids is important for optimizing the use of materials.
Identifying the failure modes or critical conditions required for
the liquid to fail leads to understanding how breakdown will
develop.
Several models of liquid breakdown have been proposed
over the years, commonalities between many of them are the
presence of an electron avalanche and bubble formation. These
avalanches are observed experimentally as current spikes or
partial discharges (PD) and precede the formation of a gas
bubble or low density region as verified using optical diagnostics. Most of the energy of the partial discharge is used in the
vaporization of the liquid forming a small bubble [7][12]. The
main aspect of the bubble model is that a vapor or low density
region forms in front of the needle and leads to a breakdown
that develops in the gas phase.
This paper discusses experiments on the basic conduction
mechanisms associated with transformer oil for nonuniform and
uniform applied fields. Past studies have done similar tests but at
much smaller gaps or only for a single geometry [4], [5], [10].
The variation of electrode material and geometry will test the
conduction current as a result of varying metalliquid interfaces.

0093-3813/$20.00 2006 IEEE

468

IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 34, NO. 2, APRIL 2006

Fig. 1. Sketch of conduction test chamber for transformer oil at reduced


temperature and pressure.

Also, changes to the temperature and pressure will clarify these


mechanisms. The data leads to the development of an explanation of the conduction current and how it evolves with applied
voltage.
This paper also discusses experiments on the investigation
of the basic breakdown mechanisms as a function of reduced
hydrostatic pressure. Optical experiments and fast electrical
measurements strongly support the gas bubble model mentioned above for cathode initiated breakdown. Data presented
here reveals that cathode initiated breakdown is strongly
pressure dependent, while anode initiated breakdown is less
dependent upon the applied pressure.
II. EXPERIMENTAL TECHNIQUES
The test setup consists of a coaxial transmission line with a
test gap on the inner conductor that is impedance matched for
fast risetimes and minimal reflections. Two unique chambers are
used: the first is for testing low current, conduction measurements of Univolt 61 transformer oil at reduced pressure or temperature and the second is for optical studies of the breakdown
of transformer oil at room temperature as a function of applied
hydrostatic pressure.
A. Setup for Conduction Tests
1) Insulated Test Chamber: The primary test chamber consists of a cylindrical stainless steel chamber 20 cm in diameter
and 40 cm in height. The chamber is heavily insulated with approximately 15 cm of thermal insulation, 10 cm of spray insulation and 5 cm of Styrofoam on the sides. A fast coaxial
setup with an impedance of 52 is used for conduction and
breakdown studies. Fig. 1 is a sketch of the test chamber and
the charging system. Two RG-19 transmission lines connect to
feedthroughs on either side of the chamber. The transmission
lines, as connected, can be charged to a maximum of 100 kV per
cable with opposite polarity, for a maximum of 200 kV across
the gap. Single sided charging, a requirement for conduction
tests, rarely exceeded 50 kV.
In the center of the chamber is the test gap, which connects
to either side of the transmission lines center conductor. A
cylindrical mesh connects the outer conductors maintaining the
impedance of the circuit. The test gap is easily accessible and

various electrode materials and geometries are used to investigate the conduction and breakdown mechanisms. Electrode materials tested in the chamber include copper, tungsten, and stainless steel. Three geometries are used to explore multiple field
configurations. A highly nonuniform field is created using tungsten electrode with a conical top (needle/point) with a radius of
curvature of about 20 m and a 1-cm-diameter copper plane. A
slightly less, nonuniform field is created using either a copper or
stainless steel electrode (pin) with a hemispherical tip of 1-mm
radius and a corresponding copper or stainless steel plane 1 cm
in diameter. The pin and plane are both the same materials, either copper/copper or stainless/stainless. The uniform field case,
plane/plane, consists of 1-cm-diameter copper planes.
The connection of the chamber to a vacuum system allows
for the adjustment of the hydrostatic pressure on the liquid. Reducing the hydrostatic pressure allows the trapped gas to be removed from the transformer oil. The adjustment of the oil temperature requires a heat exchanger submerged in the oil. A flow
of liquid nitrogen through the heat exchanger reduces the temperature to a minimum of 240 K. A pressure of 100 mtorr is
applied during the cooling process to eliminate condensation inside the chamber.
2) Oil Filtration System: Oil tests in this chamber required
some modification to the existing chamber [13]. Gravity feeds
the magnetic drive pump, which circulates the oil through a
3- m, cellulous, in-line filter back into the top of the tank. The
oil sprays into the vacuum region above the liquid. Since the
system is sealed, a simultaneous vacuum is applied; then the oil
filters and degasses at the same time. Any gas trapped within the
liquid stream rapidly expands and is removed via the vacuum
pump. Also, trapped water in the oil is removed since the applied pressure is below the vapor pressure of water. The oil is
degassed, dehydrated, and filtered for at least 24 h before experiments are undertaken.
3) Conduction Current: Conduction current tests use the
Keithley 6514 electrometer to measure the direct current (dc)
current from the picoampere to the microampere level. A protection circuit limits the voltage and current, at the electrometer
input during transient or breakdown events. The electrometer
is used only for single sided charging tests. The breakdown
voltage is determined by slowly charging the transmission
lines at a rate of 12 kV/s until the gap self-breaks. The peak
dc voltage is measured using a 1000:1 high-voltage probe, 80
K-40 from Fluke. Once the voltage is recorded it is set 13
below the breakdown threshold. During and
kV or
after breakdown, the conduction current is recorded using an
oscilloscope. The current is recorded for 50 s and then the mean
and standard deviation is calculated for the test. The voltage
is stepped down from near breakdown to a minimum of 1 kV.
After each series of tests, the oil preparation process is repeated.
B. Setup for Breakdown Tests
1) Optical Test Chamber: The optical test chamber was designed specifically for high speed electrical and optical diagnostics of breakdown events in transformer oil. It uses a coaxial test
configuration similar to the one described above. The discharge
chamber is constructed with two 15-cm viewports, on either side
of the gap. The windows are inset and 108 mm away from the

BUTCHER et al.: CONDUCTION AND BREAKDOWN MECHANISMS IN TRANSFORMER OIL

469

Fig. 3. Graph of FN plot for positive and negative polarity. Open symbols:
negative. Closed symbols: positive. Gap = 10 mm, Point/plane. Stage I
resistive. Stage II tunneling. Stage III space charge saturation.

III. EXPERIMENTAL RESULTS


Fig. 2. Sketch of shadowgraphy setup used for imaging data.

A. Conduction Current

gap for optical resolution [14]. The oil preparation is similar as


for the previous tests. After each breakdown, a vacuum is reapplied to the chamber to remove gas produced during breakdown.
A minimum of 10 min of pumping between tests is required
to remove the visible gas, in the form of bubbles suspended
in the liquid, from the system. The next test is performed when
the oil is observed to no longer be outgassing trapped gas from
within the liquid.
2) Shadowgraphy Configuration: Fig. 2 is a sketch of the
shadowgraphy setup used for all the imaging data in this article.
The system consists of a 1-W solid state diode laser with a 650
nm wavelength that is triggered by a pulse/delay generator. The
laser is expanded for illumination of the entire test gap. A lens
that images the electrode region through a 650-nm bandpass
filter, 10 nm bandwidth, and onto the intensified charge-coupled
device (ICCD) camera collects the light.
A breakdown event alters the local density changing the index
of refraction and the laser light is affected accordingly as it
passes through the test gap. Shadowgraphy image intensity is
of
proportional to changes in the second derivative
the index of refraction [15]. Therefore, it is sensitive to large
changes in the index i.e., bubble/vapor formation. This technique allows measurement of spatial variations in the index of
refraction across a probing beam. These variations provide information on index of refraction changes in the area of interest,
which correspond to the structure of the breakdown event.
3) Electrical Diagnostics: High speed electrical diagnostics
consist of traveling wave current sensors (TWCS) [16] and capacitive voltage dividers. Current is monitored on both coaxial
lines using the TWCS, which have subnanosecond risetimes and
0.1 V/A sensitivities into 50 . The transient voltage is measured simultaneously on the charging line and load line on either
side of the gap. The capacitive dividers have a subnanosecond
rise time and sensitivities of 0.01 V/V. A wide bandwidth amplifier is used with one of the voltage dividers to trigger the optical gate. For breakdown tests, the transmission line is slowly
charged at a rate of about 12 kV/s until breakdown.

Previous work on conduction mechanisms did not include the


results from both varying geometries and electrode materials.
El-Suliman et al. used a stainless steel needle with a submillimeter gap and reduced the hydrostatic pressure to a minimum
of 50 torr [5]. Swan discusses the effects of various oxide layers
on several electrode materials for dc uniform fields, where the
effect of oxide layers on the electrodes was found to affect the
barrier height more than the work function of the material [4].
The work presented here varies the pressure down to 100 mtorr
and reduces the temperature of the oil. We also discuss the effect of uniform and nonuniform fields on the conduction mechanisms.
1) Geometry and Materials: Measurement of the dc current
across the gap was made for three geometries and various electrode materials. The data is graphed using a FowlerNordversus
for interpretation
heim (FN) plot, i.e.,
of the mechanisms. Fig. 3 represents a typical data set of current and voltage for point/plane geometry displayed in a FN plot
for positive and negative polarity at atmospheric pressure. Three
regimes can be clearly distinguished from the graph for the negative polarity test case. Typically, starting from the left with the
kV (i.e.,
kV), and
high field region,
. For 0.1 kV
the current is proportional to
kV
kV
kV , the slope of the curve is
linear and negative which validates the FowlerNordheim relakV
kV , the curve
tion. Finally, for
is characteristic for ohmic conduction. The transition between
kV for the posregime II and III occurs near,
itive case. While the positive case contains fewer data points in
region III, it does indicate that the current appears proportional
.
to
Calculating the mobility in the high field region, stage 3,
will reveal the conduction mechanisms for this region. It is
calculated assuming a space charge limited current, and that
the point/plane can be modeled by a simple geometry such as
concentric spheres. This approximation was used by Halpern
and Gomer [17] to approximate the point/plane configuration.
Using the simplified geometry and solving Poissons equation

470

IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 34, NO. 2, APRIL 2006

Fig. 5. FN plot of uniform field case for two successive tests. Plane/Plane
geometry. Gap = 1:55 mm.
Fig. 4. Graph of current versus square of the applied voltage for point/plane
geometry. Test gap 10 mm.

and charge conservation equations with appropriate boundary


conditions yields (1) [17]. Assuming the carrier mobility, is
constant in the region away from the tip then the velocity, ,
of the charge carriers will be proportional to the electric field,
. The relative permittivity of the oil is
,
gap
(solid angle factor) and
is the potential
width,
drop; the current varies as the square of the potential drop
(1)
Plotting the current versus the square of the voltage yields a
straight line with the slope proportional to the mobility of the
for the
primary charge carrier. Fig. 4 is a graph of versus
mm, nonuniform field. The apparent mopoint/plane,
cm
bility for nonuniform fields was found to be 2.9
in the space charge saturation region far from the tip using the
slope from Fig. 4. It is well documented that the application of
electric fields will cause electro-hydrodynamic motion (EHD)
in a dielectric liquid [18][22]. EHD mobility, a fluid characteristic, is defined as the square root of the permittivity divided
for uniform fields. This is
by the density of the liquid,
an upper limit for the mobility of the liquid due to convective
charge transport [20]. EHD is a blending of electric and hydrodynamic forces that describes the mechanical effects of electric
fields acting on a dielectric fluid. Charge carriers in the liquid,
ions, or impurities, experience a net force from the electric field
that causes them to drift in the field. As the ions drift, they collide with liquid molecules and lose momentum and transfer energy to the fluid. The transfer of momentum causes a net drift
of the liquid in the direction of ion motion. EHD motion aides
in carrying the injected charge away from the injection point.
cm
for transThe calculated EHD mobility is 1.5
former oil in a uniform applied field. It is a factor of two less
than the apparent mobility in the high field region mentioned
above. The difference in mobility can be explained by the increase in the fluid velocity, which is a function of the applied
field (nonuniform). As the field enhancement increases near the
needle tip, electro-hydrodynamic turbulence [12] increases with
the fluid velocity, further enhancing the apparent mobility. This
leads to the formation of a buildup of space charge near the metal
tipliquid interface.
For the case of uniform fields the apparent mobility is
cm
, which is 33% higher than the expected
2.0
EHD mobility, consistent with apparent mobilities collected by

McCluskey et al. [12] for various liquids. Since the apparent


mobility is greater than the EHD upper limit, then other factors
contribute to the final apparent mobility as well. The presence
of particles in the oil enhances the mobility either by increasing
the number of charge carriers or by altering the local field.
Other charge carriers such as electrons, holes or ions in the
liquid could contribute to the mobility. Mirza et al. showed
,
that even in uniform fields with currents of less than
liquid motion is induced in hexane, with the fluid and charge
carriers having comparable velocities [22]. Fig. 5 is a graph of
the FN plot for uniform field, plane/plane geometry. The third
stage, space charge saturation, is not present in this case. Since
the field is uniform, the magnitude of the fluid velocity at the
electrodes is less than for the nonuniform field (needle/plane
geometry), therefore, the liquid velocity is smaller hence the
mobility is less.
Ion mobility without the enhancement of liquid motion is significantly smaller than the EHD mobility limit. The mobility
to 1.0
cm
for
rates of ions ranges from 1.2
typical liquids [23]. The low field region of stage 1 is initially
expected to have similar mobilities. Ions formed in the liquid
from electron attachment or field ionization will drift in the applied field adding to the apparent mobility of charge carriers.
Charge carriers in stage 1 are ions, which dominate the mobility
due to fluid motion caused by the interactions of ions and impurities with the liquid.
Stage 2 represents the linear region of the FN plot and it is assumed to be associated with a tunneling mechanism. The basic
process of electron injection into the liquid is via field emission
described by the FN equation, or tunneling through a Schottky
barrier at the metal liquid interface [24], [25]. Both of these
mechanisms have the same functional dependence of the emission current on the applied voltage expressed in (2) [13]
(2)
Assuming tunneling through a Schottky barrier,
constants where is given [25]

and

are
(3)

where
electron affinity barrier,
barrier thickness,
gap width,
. Calculating the slope from the data in Fig. 3
the barrier height can be determined. The potential barrier exists
at the metalliquid interface. The permittivity of the barrier is
assumed to be the same as the liquid and the thickness of the barrier is estimated to be a few nanometers when tunneling begins.

BUTCHER et al.: CONDUCTION AND BREAKDOWN MECHANISMS IN TRANSFORMER OIL

471

TABLE I
BARRIER CALCULATIONS FOR VARIOUS GEOMETRY AND MATERIAL

Then the barrier heights for the negative and positive polarities
are calculated as 0.30 and 0.39 eV, respectively. Similar values
are determined for each of the electrode materials. These values
are smaller than estimated values of (23 eV) found for similar
processes described by Lewis [23], which is explained by the
100 nm barrier thickness used in those calculations compared
to the thickness of a few nanometers used here. Table I lists the
calculated barrier height values determined for each geometry
and material tested. The table indicates that there is a clear difference in the barrier as a function of polarity.
An important feature seen from these data is that the anode
barrier is always larger than the cathode barrier except for the
uniform field case where anode and cathode geometry are identical. The polarity effect indicates a difference in the mechanism
of conduction. Negative polarity leads with electrons tunneling
from the metal through the barrier into the liquid. The metal
has an abundance of electrons available for tunneling that will
be able to transition across the barrier into the liquid. The barrier for the positive case is different and is discussed by Lewis
[23]. Any quasi-free electrons arriving at the anode from the
liquid are likely to be trapped within the thin oxide barrier at the
liquid interface. The charge carriers are holes, which are generated by electrons tunneling from the valence band of the insulator through the barrier leaving a hole. The process requires
the electron density near the anode to be capable of providing
the magnitude of tunneling current. Since the mobility of holes
is significantly less than electrons, a buildup of positive space
charge around the anode occurs. A net positive space charge
alters the potential from the applied field resulting in a higher
effective barrier. The effectiveness of the process depends upon
the presence of electrons at the interface layer between liquid
and metal.
Identifying the transition point between resistive current and
tunneling current flow is important to the final model of conduction. The critical point where tunneling begins is a function of the applied field. The tests with used oil, gap resistance
, indicate that a voltage of 3 kV alters the barof
rier to the point where tunneling into the liquid can begin for
the point/plane geometry, 10-mm gap. Similar results for the
mm) and plane/plane (7 kV
pin/plane (8 kV,
mm) results were observed. The applied voltage for the
uniform case is higher since there is no geometrical field enhancement only enhancement due to microprotrusions on the
surface of the electrodes. The critical field is around 40 kV/cm
for the initiation of tunneling for uniform fields. At the transition
point, the current rises rapidly from a few nanoamps to hundreds
of nA before breakdown.
A simple model of the first two stages can be generated using
the parameters obtained from the experiment. The measured
and calculated current data for the point/plane configuration is

Fig. 6. Simulation of resistive and tunneling model of conduction current.


Solid line: experiment. Dotted line: calculated. Stage I resistive. Stage II
tunneling. Stage III space charge saturation. Negative polarity.

shown in Fig. 6. The curve matches the data through stages 1


and 2 fairly well. Although the match at the first transition point
is higher than the measured value the calculation is still reasonable. Using (2) and (3) with values above for the barrier height
and appropriate field enhancement factors for each geometry
the Schottky barrier model is capable of explaining the results.
The field enhancement factor for point/plane, pin/plane, and
plane/plane geometries used for the calculation is 216, 10, and
13 respectively, as determined using field simulations or surface roughness.
2) Temperature and Pressure Effects: Investigating the dynamics of conduction in transformer oil requires altering parameters that are expected to have an effect. Variations in the
hydrostatic pressure and local temperature provide more information about how charge flows prior to conduction. Several
mechanisms affect the conduction process such as field emission, EHD, liquid viscosity, field ionization, temperature, and
pressure. Changing the viscosity of the liquid by altering the
temperature and pressure will support our model of conduction.
The observed changes due to varying temperature and pressure
of the liquid support our model of conduction.
Previous research on conduction and breakdown [5] showed
that for small gaps the effect of reducing the pressure had little
effect down to 300 torr. Fig. 7 is a graph of the characteristics as a function of applied pressure for a cathode point/plane
gap. The data indicates that there is not a measurable pressure
dependence on the conduction mechanism in transformer oil. It
indicates that all the conduction occurs in the liquid state prior
to breakdown.
Reducing the temperature of the oil affects the viscosity of the
oil strongly. Since the velocity of the liquid is proportional to the
applied field, changing the velocity of the liquid will affect the
apparent mobility and hence the space charge saturation for the
point/plane geometry. At room temperature, the fluid velocity
prior to breakdown ranges from 10 to 40 cm/s, this increases
the mobility significantly, above the mobility for ions, discussed
earlier, in a liquid without EHD. Cooling the oil and increasing

472

IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 34, NO. 2, APRIL 2006

Fig. 9. F-N plot of maximum and minimum temperature curves for the
point/plane geometry. Negative polarity.

Fig. 7. Graph of IV characteristics for varying hydrostatic pressure.


Point/plane geometry. Negative polarity.

TABLE II
BARRIER CALCULATION FOR VARYING PRESSURE AND TEMPERATURE

changed because the enhanced motion of the oil is not present


at reduced temperature.
Determination of the barrier height from the temperature and
pressure data is displayed in Table II for point/plane geometry.
The data in the table shows that the barrier height is approximately constant, within the error, for each polarity and establishing that the tunneling mechanism is little affected by the
change in temperature and pressure.
B. Breakdown Mechanism

Fig. 8. I V curves for varying temperature. Negative polarity. Point/plane


geometry. Gap = 10 mm.

its viscosity compared to the room temperature value, reduces


the maximum velocity of the oil by a factor of hundred. At the
lowest temperature, 240 K, the oil is almost solid. Fig. 8 is a
graph of the characteristics as a function of temperature
for cathode point/plane geometry.
The peak conduction current prior to breakdown is reduced
from around 1.0 A at 300 K to about 100 nA at 240 K, caused
by changes in the carrier mobility. Fig. 9 is a FN plot of the
temperature data for negative polarity. The reduced current due
to the temperature change shifts the plot down but maintains the
shape and transition points for stages 1 and 2.
The space charge saturation region is no longer observable
which supports a fluid velocity effect. Increasing the viscosity
reduces the fluid velocity but does not alter the mechanisms of
conduction in stages 1 or 2. The resistivity of the oil increases
with lower temperatures due to reduced ion mobility. The tunneling regions initiates at a similar field amplitude as the electrons traverse the barrier. The space charge saturation region

Data collection for the conduction current as a function of


pressure indicated that the breakdown voltage for the cathode
initiated case was strongly pressure dependent. In previous
research, most experiments were done at atmosphere or much
higher applied pressures [9], [11], [12], [26][29]. In these
cases, modest pressure dependence is observed since the bubbles formed will be smaller than at reduced pressure. With the
assumption of a gas phase required for breakdown the next
step is to emphasize the gas phase by reducing the applied
pressure. This has the advantage of being easy to accomplish
while simultaneously varying the pressure by several orders of
magnitude.
The first step in quantifying the pressure effect was to determine the breakdown voltage for both cathode and anode
initiated breakdown as a function of applied hydrostatic pressure. Fig. 10 shows the breakdown voltage versus pressure for
point/plane geometry. The breakdown voltage at 100 mtorr
is 50% less than at atmosphere for negative needle tests. For
anode initiated breakdown, the voltage varies by about 10% at
100 mtorr, which is relatively weak pressure dependence. Typical current waveforms for cathode initiated breakdown show
multiple current pulses prior to final breakdown [13], [14].
Each pulse is capable of producing a bubble and shadowgraph

BUTCHER et al.: CONDUCTION AND BREAKDOWN MECHANISMS IN TRANSFORMER OIL

473

Fig. 12. Shadowgraphy image of cathode initiated breakdown with


self-luminosity due electron impact ionization. Pressure = 0:01 torr,
2.5-mm gap.
Fig. 10. Breakdown versus pressure curves for positive and negative polarity.
Point/plane geometry. Gap = 2:5 mm. Open symbols = positive. Closed
symbols = negative.

most of the gap. Assuming the bubble obeys the ideal gas law
then a decrease in the pressure corresponds to an increase in
the volume. Bubbles can expand from a few micrometers to
millimeters at reduced pressure. Bubble lifetime is very long at
low pressures and can be estimated using the classical Rayleigh
model for bubble generation and collapse [30]. The Rayleigh
is the maximum bubble
lifetime is given by (4), where
liquid density,
applied hydrostatic pressure
radius,
[30], [31]. Typical lifetimes for a bubble at 680 torr are a few
microseconds and it increases to milliseconds at 0.01 torr with
a maximum bubble radius of about 1 mm
(4)

Fig. 11. Shadowgraph images of bubble expansion for negative polarity test
gap. (a) 300 torr. (b) 30 torr. (c) 0.01 torr. Gap = 2:5 mm. Exposure time 60 ns.

images confirm the correlation between number of bubbles and


number of current pulses.
Images of prebreakdown current pulses show that a PD precedes the formation of a low density/bubble region in front of
the electrode. The mechanism of formation of the low density
region is most likely due to energy injection into the liquid. The
injection of electrons from the PD causes the temperature of
the liquid to rise very rapidly because the energy excites the
molecules of the oil. The energy injection is sufficient to cause
a phase change in the oil or to disassociate a part of the oil
[10]. The
molecule liberating simpler molecules, such as
newly formed vapor expands rapidly to reach equilibrium with
the surrounding liquid. At the reduced pressure, expansion of
the bubble can be significant. Fig. 11 shows typical images of
the bubble/vapor for various pressures. The high pressure image
shows a chain like structure as predicted from chains of current
pulses.
Investigations of the pressure dependence show several
important facts about the mechanisms. Self-breakdown at low
pressures reveals that single prebreakdown events occur up
to milliseconds before final breakdown. This means that a
large partial discharge or PD will generate a bubble that fills

The breakdown mechanism for cathode initiated breakdown


is gaseous in nature given the pressure dependence and images
discussed here. A vapor bubble or series of bubbles bridge the
gap between cathode and anode allowing for the development
of a gas phase breakdown, which occurs at voltages lower than
breakdown of the liquid. At 100 mtorr, a bubble is formed in
front of the needle tip associated with a current pulse. Then the
next current pulse injects charge carriers into the bubble and
will initiate current amplification and breakdown through electron impact ionization. Images of final breakdown, as shown
in Fig. 12, show the bubble becoming fully luminous with increasing intensity closer to the anode, which is characteristic of
breakdown in gases.
In the case of anode initiated breakdown, current pulses are
not present in the waveform. Instead a slow rising current over
tens to hundreds of nanoseconds leads to breakdown [13], [14].
The slow rise in current does not inject high energy particles into
the liquid so there is not a phase change in the liquid. The breakdown voltage is about 10% lower at 100 mtorr than at atmosphere. Images of positive breakdown in oil show thin, branch
like structures emanating from the anode tip. These streamer
like structures show no measurable expansion at reduced pressure. These observations are specific to Univolt 61 transformer
oil and will vary with liquid properties.
Fig. 13 is a typical image of positive prebreakdown at reduced
pressure, which indicates no expansion of the channels. The data
supports a process triggered by a buildup of charge in the liquid
leading to a very rapid final breakdown of the gap. Positive ions

474

IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 34, NO. 2, APRIL 2006

Fig. 13. Shadowgraphy image of positive pre-breakdown at reduced


hydrostatic pressure. Pressure = 0:01 torr, 2.5-mm gap.

or holes are the likely charge carriers and will tend to form channels of charge through the liquid [23]. These channels have a
higher conductivity than the rest of the oil and the local space
charge field will field ionize the local molecules. This process
will increase the length and width of the channel and build up
the channel. Finally, the process will become self-sustaining and
breakdown through the liquid can occur. The measured current
will rise exponentially as the process continues.
IV. CONCLUSION
In this paper, the conduction and breakdown mechanisms in
transformer oil are studied as a function of applied pressure and
temperature. Stage 1 begins with a low current phase, where the
conductivity of the liquid controls the current with the charge
carried by ions formed from electron attachment. Electrohydrodynamic effects dominate the mobility of charge carriers in
transformer oil. Stage 1 mobilities should be controlled by the
cm
ion mobility, which is usually about (110)
[23]. It has been observed that even in a uniform field picoamp
currents can produce liquid motion and alter the apparent mobility [22]. The apparent mobility for uniform fields is around
cm
due to EHD fluid motion as well as the pres2
ence of impurities that could not be filtered out of the oil. The
particles increase the number of charge carriers or the local field
in the oil.
Stage 2 begins when the uniform applied field reaches a critical value of about 40 kV/cm and tunneling through the potential
barrier at the metalinsulator interface begins. For the case of a
positive barrier, a local space charge in front of the anode alters
the barrier height. The onset of tunneling through the potential
barrier at the metalinsulator interface and out into the liquid is
unaffected by changes in temperature or pressure. The barrier
.
height is usually a few tenths of an eV,
The space charge saturation, stage 3, occurs in the high
field region near final breakdown where fluid motion enhances
cm
were measured for
the mobility, values of 3
point/plane geometry. Reduction of the applied hydrostatic
pressure has no effect on the conduction mechanisms. Decreased temperature increases the viscosity and reduces the
fluid velocity without altering the primary mechanisms. Stage
3 is only present at room temperature for highly nonuniform
fields where the fields near the needle tip generate large fluid
velocities increasing the apparent mobility of charge carriers.
Bubbles have not been observed in the gap during dc conduction prior to breakdown at reduced pressure. The lack of bubbles or low density region during the dc conduction phase indicates that transformer oil is still a liquid across the gap. The

data has not shown a strong link between conduction and breakdown. A transition from pure conduction to an electron amplification process has not been established. The observed changes
due to varying temperature and pressure of the liquid support
our model of conduction.
For cathode initiated breakdown, the onset of prebreakdown
pulses allows the formation of localized bubbles in the oil, due
to highly localized energy injection. Varying the applied pressure alters the bubble size and lifetime. For the 0.01 torr case,
the maximum bubble size is of the order of the gap spacing,
12 mm, with a lifetime of milliseconds. After the bubble has
formed, the next current pulse injected into it leads to an immediate breakdown and at a 50% lower voltage than at atmosphere. Final breakdown is initiated from within the gas phase.
The case of positive breakdown indicates minimal pressure dependence correlating to a process that develops within the liquid
state. The breakdown voltage at 0.01 torr is about 10% lower
than at atmosphere and images of the events indicate no expansion of the event. Breakdown initiates from channels of higher
conductivity in the liquid which grow exponentially as they develop into a self-sustaining arc across the gap.
REFERENCES
[1] R. Tobazeon, Charges and discharges of conducting particles in a
plane oil gap subjected to a dc voltage, in Conf. Record IEEE 12th Int.
Conf. Conduction Breakdown Dielectric Liquids, Rome, Italy, 1996,
pp. 309312.
[2] M. Krins, H. Borsi, and E. Gockenbach, Influence of carbon particles
on the breakdown and partial discharge inception voltage of aged mineral based transformer oil, in IEEE 7th Int. Conf. Dielectric Materials,
Measurements, Appl., Sep. 2326, 1996, pp. 251254.
[3] J. Sheshakamal and J. D. Cross, The influence of impurities on electrohydrodynamic motion in insulating oils, in Proc. IEEE Conf. Electr.
Insul. Dielectric Phenomena, Oct. 1989, pp. 338345.
[4] D. W. Swan, A recent review of investigation into electrical conduction
and breakdown of dielectric liquids, Brit. J. Appl. Phys., vol. 13, pp.
208218, 1962.
[5] A. A. El-Sulaiman, A. R. M. Alamoud, A. S. Ahmed, and M. I. Qureshi,
Effect of sub-atmospheric pressure on the conduction current in transformer oil, in Conf. Record IEEE 10th Conf. Conduction Breakdown
Dielectric Liquids, Sep. 1990, pp. 176180.
[6] Y. Toriyama, T. Sato, and H. Mitsui, Dielectric breakdown of insulating
liquids, Brit. J. Appl. Phys., vol. 15, pp. 203204, 1964.
[7] F. M. J. McCluskey, F. Huynh, A. Denat, N. Bonifaci, and R. Kattan,
Bubble formation in simple and complex liquids, in Conf. Record
IEEE Int. Symp. Electrical Insulation, Pittsburg, PA, Jun. 58, 1994, pp.
406409.
[8] W. G. Chadband, From bubbles to breakdown, or vice-versa, in Conf.
Record IEEE 11th Conf. Conduction Breakdown Dielectric Liquids, Jul.
1993, pp. 184193.
[9] R. Kattan, A. Denat, and O. Lesaint, Generation, growth, and collapse
of vapor bubbles in hydrocarbon liquids under a high divergent electric
field, J. Appl. Phys., vol. 66, pp. 40624066, Nov. 1989.
[10] E. J. Murphy, The generation of gases in the dielectric breakdown of
oil, J. Phys. D, Appl. Phys., vol. 3, pp. 917923, 1970.
[11] R. Kattan, A. Denat, and O. Lesaint, Formation of vapor bubbles in
nonpolar liquids initiated by current pulses, IEEE Trans. Electr. Insul.,
vol. 26, no. 4, pp. 656662, Aug. 1991.
[12] F. M. J. McCluskey and A. Denat, The behavior of small bubbles generated by electrical current impulses over a wide range of applied pressures, J. Appl. Phys., vol. 80, pp. 20492059, Aug. 1996.
[13] M. Butcher, A. Neuber, H. Krompholz, and J. Dickens, Pre-breakdown
current behavior in dc volume breakdown in transformer oil, in Proc.
14th IEEE Int. Pulsed Power Conf., vol. 1, Jun. 1518, 2003, pp.
289292.
[14] M. Cevallos, J. Dickens, A. Neuber, M. Haustein, and H. Krompholz,
Self electrical breakdown in biodegradable oil, in 14th IEEE Int.
Pulsed Power Conf., vol. 2, Jun. 1518, 2003, pp. 10361039.

BUTCHER et al.: CONDUCTION AND BREAKDOWN MECHANISMS IN TRANSFORMER OIL

[15] I. H. Hutchinson, Principles of Plasma Diagnostics. New York: Cambridge Univ. Press, 1987, pp. 112115.
[16] H. Krompholz, K. Schoenbach, and G. Schaefer, Transmission line
current sensor, in IEEE Instrum. Measure. Technol. Conf., Tampa, FL,
1985, pp. 224224.
[17] B. Halpern and R. Gomer, Field emission in liquids, J. Chem. Phys.,
vol. 51, no. 3, pp. 10311047, Aug. 1969.
[18] P. K. Watson, W. G. Chadband, and M. Sadeghzadeh-Araghi, The role
of electrostatic and hydrodynamic forces in negative-point breakdown of
liquid dielectrics, IEEE Trans. Electr. Insul., vol. 26, no. 4, pp. 543559,
Aug. 1991.
[19] G. Theodossiou, J. Nelson, M. J. Lee, and G. M. Odell, The influence
of electro-hydrodynamic motion on the breakdown of dielectric liquids,
J. Phys. D, Appl. Phys., vol. 21, no. 1, pp. 4550, 1988.
[20] R. E. Tobazeon, Streamers in liquids, in The Liquid State and Its
Electrical Properties, E. Kunhardt, L. Christophorou, and L. Luessen,
Eds. New York: Plenum Press, 1987.
[21] A. Castellanos, Coulomb-driven convection in electrohydro-dynamics, IEEE Trans. Electr. Insul., vol. 26, no. 6, pp. 12011215, Dec.
1991.
[22] J. S. Mirza, C. W. Smith, and J. H. Calderwood, Liquid motion and
internal pressure in electrically stressed liquids, J. Phys. D, Appl. Phys.,
vol. 3, pp. 580585, 1970.
[23] T. J. Lewis, An overview of electrical processes leading to dielectric
breakdown of liquids, in The Liquid State and Its Electrical Properties, E. Kunhardt, L. Christophorou, and L. Luessen, Eds. New York:
Plenum Press, 1987.
[24] R. V. Latham, The origin of pre-breakdown electron emission from
vacuum insulated high voltage electrodes, Vacuum, vol. 32, pp.
137140, 1982.
[25] C. S. Athwal and R. V. Latham, Switching and other nonlinear phenomena associated with pre-breakdown electron emission current, J.
Phys. D, Appl. Phys., vol. 17, no. 5, pp. 10291043, 1984.
[26] A. Beroual and T. Aka-NGnui, Influence of additives and hydrostatic
pressure on streamers initiation and dielectric strength of liquids,
in Proc. IEEE Conf. Electr. Insul. Dielectric Phenomena, 2002, pp.
248251.
[27] O. Lesaint and P. Gournay, On the gaseous nature of positive filamentary streamers in hydrocarbon liquids. I: Influence of the hydrostatic
pressure on the propagation, J. Phys. D, Appl. Phys., vol. 27, pp.
21112116, 1994.
[28] P. Gournay and O. Lesaint, On the gaseous nature of positive filamentary streamers in hydrocarbon liquids. I: Propagation, growth and collapse of gaseous filaments in pentane, J. Phys. D, Appl. Phys., vol. 27,
pp. 21172127, 1994.
[29] H. M. Jones and E. E. Kundhardt, Development of pulsed dielectric
breakdown in liquids, J. Phys. D, Appl. Phys., vol. 28, pp. 178188,
1995.
[30] L. Rayleigh, On the pressure developed in a liquid during the collapse
of spherical cavity, Philosophical Mag., vol. 34, pp. 9498, 1917.
[31] F. Jomni, F. Aitken, and A. Denat, Investigation of the behavior of
microscopic bubbles in insulating liquids: Transition from the inertial
regime to the viscous one, in Proc. 14th IEEE Int. Conf. Dielectric Liquids, Jul. 712, 2002, pp. 194198.

Michael Butcher (S01) was born in El Paso, TX.


He received the M.S. degree in applied physics, in
1998, from Texas Tech University, Lubbock, where
he is currently working toward the Ph.D. degree in
electrical engineering.
He is a Research Assistant at the Center for Pulsed
Power and Power Electronics at Texas Tech University while completing his doctoral work.

475

Andreas A. Neuber (M97SM03) was born in Aschaffenburg, Germany. He received the Dipl.-Phys.
M.E. and Ph.D. degrees from Darmstadt University
of Technology, Darmstadt, 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 to 1996,
in the area of nonlinear laser spectroscopy and chemical reaction kinetics in combustion. He joined 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 explosivedriven pulsed power. He has published more than 100 journal articles and conference proceedings paper.

Michael D. Cevallos (S02M04) was born in San


Antonio, TX, on October, 16, 1970. He received the
M.S. degree in electrical engineering, in 2002, from
Texas Tech University in Lubbock, TX, where he is
currently working toward the Ph.D. degree in electrical engineering.
He is a Research Assistant at the Center for Pulsed
Power and Power Electronics at Texas Tech University, while completing his doctoral work.

James C. Dickens (S89M95SM03) was born in


San Jose, CA. He received the B.S.E.E, M.S.E.E., and
Ph.D. degrees from Texas Tech University, Lubbock,
TX, in 1991, 1993, and 1995, respectively.
He is currently an Associate Research Professor
of Electrical Engineering at Texas Tech University.
His current research interests include electric space
propulsion, high-efficiency power electronics for
space applications, high-power microwaves, and
pulsed-power technology.

Hermann Krompholz (SM 84) received the Ph.D.


degree in physics from Technical University Darmstadt, Darmstadt, Germany, in 1977.
He was a Research Associate at Technical University 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, 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.