IEEE Transactions on Dielectrics and Electrical Insulation Vol. 13, No. 1; February 2006
1070-9878/06/$20.00 © 2006 IEEE
On Priorities of Cathode and Anode Contaminations in
Triggering the Short-pulsed Voltage Breakdown in Vacuum
A. V. Batrakov, S. A. Onischenko, D. I. Proskurovsky
Institute of High Current Electronics SD RAS,
2/3, Akademichesky Ave., Tomsk 634055, Russia
and D. J. Johnson
Sandia National Laboratories,
Albuquerque, New Mexico 87115, USA
Modern theoretical notations on electrical breakdown in vacuum consider cathode
triggering mechanisms to be most responsible on short-pulsed (< 1 ?s) breakdowns
while anode mechanisms to be responsible in a part on dc and long-pulsed breakdowns.
Following those notations, we tried to reveal conditions at which either mechanism
steps aside to another one. The study involved several experimental techniques
including the anode-probe surface scanning, pulsed electron-beam surface melting in
vacuum for surface cleaning, and intentional dust particle contamination of electrode
surfaces. Breakdown tests were performed using a pulser capable of producing 220 kV
quasi-square pulses that were adjustable to ~ 30 to 80 ns pulse length. Our experiments
showed that cathode emission sites are responsible for breakdowns at relatively low
hold-off fields. At higher electric fields of up to 1 MV/cm, the anode share in the
mechanism of triggering breakdown becomes probably more significant than the
Index Terms — Vacuum breakdown, vacuum insulation, vacuum measurement.
THE question about priorities of anode and cathode
mechanisms triggering electrical breakdown in vacuum gaps
arises first with the use of broad-area gaps. However the small
gaps used to study vacuum breakdown, which take place in the
majority of experimental works to simplify conditions under
study and to illuminate certain mechanisms of breakdown, lead to
an incomplete picture of the breakdown physics. As a result, the
priorities of the mechanisms in theoretical formulations depend
somewhat on the convenience of their study. Finally, theoretical
formulations that properly describe the breakdown of idealized
gaps have modest success in the description of the breakdown of
The above discourse refers to the absolute priority of the
cathode breakdown triggering mechanisms in modern theoretical
formulations on the subject, but they don’t explain why it is so.
However, this is because the bulk of data on breakdown in
vacuum, including the well-known experimental fact that the gap
plasma starts to glow on the cathode at both pulsed  and dc 
voltage breakdowns. The latter finally approved the certainty of
the concept that the field-emission sites are the basic cause of
breakdowns, based on earlier experiments [3, 4]. Further research
works were connected to a great extent with explaining of the
field emission initiated breakdown mechanisms. Those works
were based on experiments with either gaps preliminary
conditioned in situ with breakdowns or needle-to-plane gaps. In
situ breakdown pre-treatment produces a particular surface relief
that contains numerous sharp protrusions enhancing the gap
electric field enough to provide intense field emission as with
needle-to-plane gas. The treatment of voltage-current
characteristics of such gaps in coordinates of the Fouler-
Nordheim equation results in field enhancement factors, ?, being
comparatively low and equatable to geometrical factors. The
same experiments on fresh (non-conditioned) electrodes gave
unrealistic (for a field emitter) values of ? that were explained in
terms of emission from dielectric inclusions . Furthermore, the
absolute majority of emission sites on a fresh cathode surface
were found to correspond to dielectric inclusions rather than
protrusions, even if electrodes are made of high-purity-grade
In themselves cathode initiated breakdown mechanisms allow
describing most of the experimental data excluding the
dependence of the breakdown electric field, E, on the gap
distance, referred to as the total voltage effect. A cathode
breakdown mechanism implies a priori the local electric field to
be the sole parameter determining electric field strength on an
insulating gap, while the gap distance should affect only the
following development of the gap conductivity. Actually,
Manuscript received on 24 January 2005, in final form 14 April 2005.
A. V. Batrakov et al.: On Priorities of Cathode and Anode Contaminations in Triggering the Short-pulsed Voltage Breakdown in Vacuum 42
however, the breakdown electric field turns out to be depending
inversely on the gap distance to the square root of the gap length
, d. As a rule, this dependence is described in terms of the
macroparticle induced breakdown giving either E ∝ d-1/2  or
E ∝ d-5/8 . The particles are considered typically  to be of
micron and sub-micron sizes. The concept implies the importance
of both anode and cathode cleanness in the breakdown initiation
and look quite successful except for the fact that the local
breakdown field, which appears as the gap field multiplied by ?,
doesn’t depend on d . The latter fact implies that ? depends on
d. This fact continues to be not understood.
Anode initiated breakdown mechanisms, which rely on a
micron-sized particle to be accelerated across the gap and produce
plasma as a result of impact, are reasonable for long-pulsed and
dc voltages because time is adequate for this to happen. The
same concerns arise in gas liberation since a certain time is
necessary to produce high enough gas pressure in the gap. This is
why the observation of the total voltage effect in nanosecond
duration pulses was an unexpected fact  whose cause was
associated in first approximation to loosely bound nanoparticles
accelerated across the gap. This seems an exotic mechanism
because there is no information on features of nanoparticle
impacts. Nevertheless, we decided to consider the possible
influence of nanoparticle contamination on electrodes on hold-off
as one of our goals in the present work while the central idea is to
trace correlations between electrode contaminations and
breakdowns. An emphasis was also made in the investigation to
the separate influence of cathode and anode cleanness on hold-
The above motivations of the work were comprehended after
long-lasting development of the method for enhancement of the
vacuum insulation by preliminary treating the electrodes with the
pulsed electron beam surface melting mode  referred to as the
EBEST method. In this connection, the present work was
conceived to shed light upon the physical restrictions of further
development of the EBEST method.
2 EXPERIMENTAL TECHNIQUES AND
2.1 EXPERIMENTAL CONDITIONS
All the experiments were performed in conditions of high
oil-free vacuum of about 5×10-5 Pa (5×10-7 mBar). A couple
of 8 cm diameter electrodes formed a plane-parallel gap.
Manipulations with electrodes and their installation into the
chamber were performed in dust-free environment provided
by the air-filtering equipment based on HEPA H13 air filters.
The use of dust-free clothes and a mask and powder-free
gloves was the obligatory condition in manipulations with
electrodes. Electrode transportation between set-ups was
carried out with using once-through clean boxes.
2.2 ELECTRODES PERFORMANCE AND
All the electrodes were shaped as plane cylinders of 8 cm
diameter and were 1.25 cm thick with Chang-profile edges to
provide a uniform electric field in the gap. The shape of the
electrode profile curves was specified by the Chang formula
0.51 0.001 cosh
g is the anode-cathode (AK) gap. The electrode materials were
stainless steel 304L, 316L SCQ, and 12X18H10T (steel 321
equivalent). The electrodes were mechanically machined with
Ra = 0.63 mm. Some electrodes were treated further using the
electron-beam surface treatment (EBEST)  in the mode of
thin-layer surface melting in an oil-free vacuum.
Before installation into the vacuum chamber, the electrodes
were rinsed in an ultrasonic washer, followed by wiping with a
lint-free cloth and high-purity-grade solvent (ethyl alcohol,
isopropyl alcohol, or acetone). Some electrodes were not
cleaned with solvent nor wiped to determine whether
removable contamination was present on the electrodes and if
the electron-beam treatment facility produces additional
contamination of the surface under treatment.
Intentional contaminations of two types were used in a part
of the experiments. The first type was to skip the electrode
wipe cleaning. We speculate from evidence that any plasma-
based technology brings a problem of machine-produced
debris particles [13-15] that levitate in plasma and could be
delivered by plasma to the inner surfaces of a vacuum
chamber. We speculate that the same problem takes place in
our EBEST facility because the electron beam is generated
with a plasma filled diode.
Intentional contamination with nano-sized particles was
also used. Electrodes were contaminated by means of
spraying ~ 0.4 cm3 of alcohol-based suspension, containing
0.1 vol.% of nanoparticle powder, on the electrodes. In some
breakdown tests, 0.01 vol.% suspension was used. After
complete vaporization of the solvent, the electrodes were
blown on with air to remove particles whose weak adhesion
could lead to particle exchange between electrodes under
handling. Then, the electrodes were installed into the chamber,
and breakdown fields were measured. Copper, alumina, and
carbon nano-powders were used. The copper and alumina
powders were produced by means of the electrical explosion
of wires . The carbon powder was produced by harvesting
the inner-wall debris after burning a diffuse-mode vacuum arc
between graphite electrodes. Such debris is known to consist
of nanoparticles including fullerenes .
2.3 ANODE-PROBE SCANNING TECHNIQUE
The method of surface emission scanning is based on
measuring field emission currents in a vacuum gap formed by
a plane cathode and a needle anode. Herewith, the anode is
moved along the cathode surface. This technique was able to
observe a distribution of field emission sites on a surface
under the study. The method allows also measurement of the
Fouler-Nordheim current-voltage characteristics of separate
emission sites. These data could reveal the nature of emission
from separate emission sites. The method was successfully
used elsewhere [18, 6].
In this work, we have realized the principle of a disk drive
(Figure 1). The plane cathode is under rotation. The
manipulator carries the needle anode along the cathode radius
in parallel with the cathode surface. A positive voltage is
applied to the anode, which gives an opportunity to measure
the emission current. Emission site positions are recorded in
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 13, No. 1; February 2006
Figure 1. Schematic diagram of the anode-probe scanning technique.
Figure 2. Circuit diagram for measuring the leak currents in a vacuum gap.
Figure 3. Waveform of emission current (lower trace) and rotation phase
zero clock pulses from the optical receiver (upper trace). The emission
current waveform scale is of 25 nA. The emission current waveform testifies
to the emission site.
polar coordinates. The radial coordinate is produced by the
scale of the anode manipulator whilst the angular coordinate is
produced by means of an oscilloscope and the zero-phase
marker being recorded with the light source, the light receiver,
and a mirror at the cathode holder. As soon as an emission site
has been revealed, it’s current-voltage characteristic is
measured by variation of the anode potential. The accuracy of
emission current measurements is mainly determined by beats
of the cathode under rotation. The beating amplitude at the
electrode edge was typically about 3 ?m, while the typical
scanning gap was about 150 ?m. This produces <5 %
instability in electric field strength under the scanning.
After the scanning procedure the cathode was examined ex
situ from the vacuum chamber in the clean air environment
with a visible light microscope to recognize emission site
The circuit diagram for measuring the dark current is shown
in Figure 2. The 2200 pF capacitor supplying the vacuum gap
is grounded through 1 ΜΩ input of the high-sensitive
wideband amplifier of electrical signals. An amplified output
signal is registered by the Tektronix TDS-220 digital
oscilloscope. The amplifier has an accuracy of 5%. The
amplifier has the hardware possibility for filtration of a signal
in a user-defined range, which gave the opportunity to
minimize the net noise pickup. The capacitor was charged by
the BP-52M high-voltage (HV) dc power supply (instability
and pulsations do not exceed 0.1%) to voltages controlled
within 8 – 15 kV. The LED optical pair was used for
registration of the phase under the cathode rotation. The
cathode was rotated with ~10 Hz rotation frequency.
Instability of the period did not exceed 200 ?s. An example of
waveforms under the scanning is given in Figure 3.
2.4 BREAKDOWN TEST TECHNIQUE AND
The set-up is based on a 12-cascade Marx generator
producing voltage pulses of the amplitude being adjustable
within 220–230 kV and the pulse length being controlled by
the crowbar switch to give a FWHM of 30 to 80 ns. The
crowbar switch was adjusted to produce either 30±5 ns or
60±10 ns voltage pulses.
The breakdown test protocol started at a gap undoubtedly
wider than the gap which was expected to allow breakdown.
Then the electrode gap was closed between HV pulses by
100 ?m for gaps larger than 0.3 cm and 50 ?m for smaller
gaps. As a result, the maximal step of increase in electric field
strength did not exceed 33 kV/cm.
Two breakdown criteria were used in the work. The first is
the gap voltage and current waveforms and the second the
visibility of the plasma glow in the gap. The latter was of great
importance in a case of the short pulse since at shorter pulses
there was no time for discharge to develop up to high enough
currents. An existence of a gap plasma glow without reliably
detectable discharge current (i.e., the discharge current is
below the displacement current) was considered as a
breakdown event. The typical waveforms are shown in
A. V. Batrakov et al.: On Priorities of Cathode and Anode Contaminations in Triggering the Short-pulsed Voltage Breakdown in Vacuum 44
Figure 6. Histograms of ? for machined cathodes.
Figure 4. Typical waveforms of the gap voltage (upper traces, 75 kV scale)
and gap current (lower traces, 400 A scale) at longer ((a), (c), and (e); 25 ns
scale) and shorter ((b), (d), and (f); 10 ns scale) pulse durations, corresponding
to (the up row) absence of breakdown and breakdown of (the middle row)
lower current and (the bottom row) higher current.
We fixed gap lengths and waveforms. Knowing voltage
pulse amplitude, gap length, and waveforms behavior, we
recognized breakdown electric fields.
We performed statistical measurements of breakdown
electric fields. Each point was produced as a result of four to
six breakdown tests at the same conditions. First-breakdown
electric field values, Ebr
values gotten as a result of HV conditioning, Ebr
were measured. The experimental results are reported below
as bars of mean values with standard deviations.
Corresponding mean gaps, <g>, can be estimated as V/E
ratios, where V is about 220 kV.
1 (E_1), and highest breakdown field
3.1 EMISSION SITES APPEARANCE AND
The anode-probe scanning of cathodes revealed emission
sites being evenly distributed on surfaces under scans. The
field emission current-voltage characteristics produced quasi-
straight lines in coordinates of the Fauler-Nordheim equation,
so, calculations of ? were suitable.
Examination of surfaces under the scan with the light
microscope revealed large-size surface irregularities at local
areas corresponding to emission sites. Figure 5 shows typical
images of the sites. The probability of observation of such
irregularities under this examination was about 90 %. In the
rest of the cases, surface irregularities causing pre-breakdown
emission had smaller dimensions than the optical resolution of
Figure 6 shows all the ? values obtained in the experimental
run. It is remarkable that the ? values are comparable with
those observed with large-area gaps after long-lasting
electrical breakdown conditioning  and lower than those
observed at non-conditioned gaps . This can be tentatively
explained by the pulsed nature of the current-voltage
characteristics acquired in our experiments because the
cathode was rotated under the anode probe. Also, the anode-
probe needle probably acts to electrostatically remove dust
particles from the cathode being tested.
It was found experimentally that surface cleaning with
solvent and lint-free wipers drastically changed emission-site
maps. However, the appearance of the emission sites observed
visually with the microscope did not change with wiper
cleaning but some of the sites later became active and others
inactive emission sites. Therefore the emission sites observed
with the optical microscope are primarily traps for small
particles that are loosely bound to the surface of the cathode.
Figure 5. Typical cathode surface images corresponding to emission sites at a
machined surface and their parameters. Image (d) contains many craters
produced by breakdown.
The final goal of a prebreakdown study is the prediction of
a HV breakdown location and breakdown electric field value.
With such a test procedure we found a 70% probability of co-
occurrence of field emission sites and HV breakdown
locations. This correlates well with results  obtained in
another experimental condition. Furthermore, the HV
breakdown electric fields correlated also with fields of field-
emission switching, and the first was almost equal to the
second for all the electrodes tested.
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 13, No. 1; February 2006
Figure 8. Histograms of β for non-wiped EBEST cathodes.
3.2 EMISSION SITES APPEARANCE AND
Pre-breakdown emission sites under the scanning of EBEST
cathode surfaces were detected in the case of 12X18H10T
steel only, which contains many more impurities than 304L
and 316L steels. Also the current-voltage characteristics were
highly instable, which failed calculation of ?. Moreover, they
manifested hysteresis as described elsewhere .
Optical microscopic examination revealed emission sites on
EBEST surfaces that appeared often as pores associated with
craters that occur from explosive removal  of second-
phase FeS sulfide particles located at grain boundaries
(Figure 7). Pores are known to be a cause of emission
according to microdischarge mechanism  considering a
pore as a hole cathode. This mechanism for a pore-like
emission sites produces hysteresis in I-V characteristics which
agree with our results.
Figure 7. Typical cathode surface images corresponding to pore emission
sites (a), (b), and (c) and cracks (d) at EBEST cathode surfaces.
Lint free wiper re-cleaning of EBEST electrodes gave alike
changes of emission-site maps as cleaning polished electrodes.
This indicates the detected emission sites are traps for tiny
Experiments on cathodes made of 304L and 316L steels
demonstrated absence of detectable emission sites at anode-
probe scanner fields of up to ~ 1 MV/cm. This is an
unexpected result since mean values of Ebr
made of those materials were below 1 MV/cm as a rule, i.e.
there are no correlations between cathode emission sites and
breakdown events. Careful optical microscope examination of
complete cathode surfaces didn’t reveal pores, which explains
partially the absence of emission sites. Therefore, since the
HV breakdown fields were lower than the anode-probe
scanner switching fields it is necessary to recognize the direct
1 for electrodes
influence of the anode on breakdown initiation even for
nanosecond duration pulses.
3.3 PRE-BREAKDOWN CONDUCTIVITY AND
BREAKDOWN OF GAPS WITH NATURALLY
CONTAMINATED EBEST ELECTRODES
With the above experiments, loosely bound contaminants
were found to be a cause of prebreakdown electron emission
and breakdowns. As noted, emission from the cathode ceases
to play a dominating role in breakdown initiation influenced
by the electric field. With strong electric fields of up to
~ 1 MV/cm we suspect that the anode plays an important role
in breakdown initiation. To check this more precisely,
experiments with non-wiped EBEST electrodes were
performed. Doing this, one has to take into account that
plasma-based technologies including EBEST are unpleasantly
featured with dust particles being distributed among inner
walls of a facility, including parts under a treatment. This is a
great problem for semiconductor technologies [13, 15] that
imply a consequence of technological procedures within a
single cycle, but a lesser problem for EBEST HV hold-off
performance that can be fixed with the post-treatment lint-free
wiping. Thus, we ensure a certain reproducible contamination
of electrode surfaces. To determine which contaminated
electrode was responsible for breakdown we performed
experiments with a wiper cleaning of only one electrode.
First of all, technological contamination was investigated on
cathodes. Anode-probe scanning revealed high emission
activity of cathodes. It manifested itself in high values of ?
and low Eonset. Note emission activity of cathode surfaces in
this experiment was higher than that of machined and wiped
cathodes (Figures 8 and 9).
Microscopic examination of non-wiped EBEST cathodes
didn’t reveal observable surface irregularities corresponding to
emission sites, i.e. their sizes are beyond the microscope
In this experiment the breakdown electric fields were found
to be almost equal to the fields of emission switching. The
breakdown results are shown in Figure 10. Comparing bar 7 to
bar 6, bar 9 to bar 2, and bar 10 to bar 3, one can see the
unambiguously positive influence of pulse shortening on hold-
off. As well, comparing bar 10 to bar 8, one can see the
unambiguously positive influence of wiping the electrodes on
A. V. Batrakov et al.: On Priorities of Cathode and Anode Contaminations in Triggering the Short-pulsed Voltage Breakdown in Vacuum 46
Figure 10. Breakdown electric fields for 8-cm electrode gaps at short
(30 ns) and long (60 ns) pulses.
Figure 9. Eonset vs. β for wiped and non-wiped cathodes.
Figure 11. Breakdown electric fields of (1) non-wiped cathode non-wiped
anode gaps, (2) wiped cathode non-wiped anode gaps, (3) non-wiped
cathode wiped anode gaps, and (4) wiped cathode wiped anode gaps. All the
electrodes were EBEST treated preliminary. <g> is the corresponding mean
hold-off at longer high-voltage pulses. However, the positive
effect of wiping disappears at shorter pulses (please compare
bar 2 to bar 4). This means that the contamination left by the
EBEST facility is sensitive to high-voltage pulse duration in
breakdown initiation. The latter is an important fact in the
To probe the nature of contaminations lowering hold-off,
we performed experiments with only one electrode cleaned by
wiping. This approach allowed us to check which electrode is
most responsible for vacuum breakdown. Taking into account
that electrode cleaning with wipers has a positive effect at the
longer pulse duration only, we performed an experiment with
60 ns duration high voltage pulses. The results of this
investigation are presented in Figure 11. Comparing the bars
in Figure 11, it can be seen that the first breakdown electric
field is primarily determined by cleanness of an anode.
3.3 BREAKDOWN OF GAPS WITH INTENTIONALLY
CONTAMINATED EBEST ELECTRODES
Intentional contamination of previously wiped EBEST
electrodes is a reasonable continuation of experiments with
non-wiped electrodes. Taking into account the nature of
contamination by plasma-based technologies, we used nano-
powders as artificial contaminants.
measurements with applying 30 ns and 60 ns high-voltage
pulses. The results of the experimental run with the carbon
powder are presented in Figure 12.
The use of 60 ns voltage pulses gave results with similar
trends to those of naturally contaminated electrodes. Decrease
of pulse duration inversed correlations in breakdown fields.
Unlike the longer pulse duration, cathode contaminations
worsen vacuum insulation to a greater extent than anode ones
at shorter voltage pulses, so, the cathode cleanness becomes
most important. The inversion in correlations of breakdown
fields means a change in the breakdown triggering mechanism
with shortened pulse duration. This gives a reason to conclude
that a 60 ns pulse provides conditions for particles to initiate
breakdowns, while a 30 ns pulse is too short to do this.
Therefore, explosive cathode emission probably dominates the
breakdown mechanism at shorter voltage pulses. However,
this conclusion is only tentative because the difference in the
mean values of measured breakdown fields is less than the
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 13, No. 1; February 2006
Figure 12. Breakdown electric fields of dusty cathode dusty anode gaps
(1), clean cathode dusty anode gaps (2), and dusty cathode clean anode gaps
(3) at 60-ns (a, b) and 30-ns (c, d) high-voltage pulses. All the electrodes
were under EBEST treatment followed by carbon-powder intentional
Figure 13. Arrangement allowing direct recognition of an electrode whose
contamination is responsible on breakdown.
Figure 14. Breakdown electric fields of dusty cathode dusty anode gaps (1,
1’), clean cathode dusty anode gaps (2, 2’), dusty cathode clean anode gaps
(3, 3’) with higher (1, 2, and 3) and lower (1’, 2’, and 3’) concentrations of
Al2O3 nano-particles on electrode surfaces.
standard deviations of the mean values. A more substantive
conclusion is that the anode contamination plays a dominant
role in breakdown ignition and is more important than that of
the cathode contamination.
We performed also HV pulse experiments with partial
contamination of electrodes with powder as is shown in
Figure 13. For these tests 60 ns voltage pulses were applied to
the electrode gaps. This arrangement gave an opportunity to
check directly which electrode produces breakdown with
higher probability. We performed three tests with this
arrangement. In two tests the gap broke down due to the anode
dust and one test the gap broke down due to the cathode dust.
These data correlate well with breakdown electric fields at
gaps with only one of electrodes being contaminated.
The correlations in breakdown electric fields appeared to be
less certain when alumina powder was used as an intentional
contamination (Figure 14). Nevertheless, alumina-powder
contamination of the cathode surface tended to give a greater
probability of lowering the breakdown threshold than when
alumina-powder was applied to the anode of fresh (non-
conditioned) gaps. To proof the trend found in the experiment
with alumina contamination, we performed two tests on
breakdown of gaps with spottily contaminated electrodes
A. V. Batrakov et al.: On Priorities of Cathode and Anode Contaminations in Triggering the Short-pulsed Voltage Breakdown in Vacuum 48
Figure 15. Breakdown electric fields of dusty cathode dusty anode gaps
(1), clean cathode dusty anode gaps (2), dusty cathode clean anode gaps (3).
Dusty electrodes were contaminated with copper nano-powder. First
breakdown electric fields were about 0.4 MV/cm for all the gaps under the
revolved from each other by 180o. Both tests were recognized
to correspond to cathode-dust initiated breakdown.
Finally, an experiment with the copper nano-powder
contamination was performed under the same experimental
conditions. Copper powder contamination lowered breakdown
fields to the same level observed with other nano-powders
(Figure 15). In this experiment there is no difference in the
hold-off fields of gaps with contamination only on the anode
or cathode surface.
Thus, a certain set of experimental facts was gained by
carrying out the experiments. The most important facts are as
Prebreakdown electron emission appears at fields close
to the breakdown field if conditions on electrode surfaces
in a vacuum gap can sustain only moderately strong
(essentially below 1 MV/cm) short-pulsed (< 100 ns)
hold-off electric fields. There is also a spatial correlation
breakdowns in this case.
(ii) Prebreakdown emission activity from a cathode changes
it's topology after surface cleaning with pure solvent and
lint-free cloth while the visual appearance of emission
sites shows no changes. This gives a ground to believe
that emission sites observable in a light microscope
emission sites are traps for smaller objects. These objects
are probably tiny particles. Really, they could not be
oxide films since a simple wiping can't destroy them.
This could not be also organic films since electrodes
were cleaned carefully before the emission tests.
(iii) The spatial and electrical correlations disappear if the
conditions are improved so much so that the breakdown
fields are close to 1 MV/cm. In this case, the breakdown
electric field turn out to be weaker than the fields at
which detectable prebreakdown electron emission could
start to operate. This fact shows unambiguously that
cathode emission fails independent explanation of the
emission sites and HV
breakdown ignition, and the anode becomes more
influencing on electrical insulation of gaps. It is
reasonably to conclude that tiny particles play also the
key role in breakdown ignition when they situated at the
(iv) The latter assumption agrees with the data on separate
contamination of electrode surfaces with powders. Such
contamination turned out to be sensitive to high-voltage
pulse duration in a part of the experiments. This concerns
to carbon nanoparticles being the least dense particles
among those used. The remarkable fact is that for carbon
nano-powder applied only to the anode drops hold-off to
a greater extent in comparison with it applied to the
All these facts turn our attention toward nanoparticles as a
possible prime cause of breakdown ignition. In our
considerations, we have to bear in mind the fact that the
discharge plasma occurs at the cathode electrode in the first
place [1, 2]. That is why the macro-particles under
consideration could be only at the anode. They have to be
taken off and accelerated in a gap and to strike the cathode
with the initial plasma production.
Three criteria should be met to cause a breakdown event by
an impact, which are (i) ‘a macro-particle should be taken off
the surface due to some reason’, (ii) ‘a macro-particle should
have a time to cross a gap’, and (iii) ‘a macro-particle should
produce after-effects capable to initiate the breakdown’.
Criterion (ii) seems to be easiest for analysis; so, let us analyze
it in the first place.
Possibility for a macro-particle to cross a gap of a few
millimeters length for a time of 120 ns and Vmax(1 - cos ?t)
shape of high-voltage pulse is shown elsewhere . We
follow alike routine here. The simple equations for the particle
motion were used, which are as follows:
Here, v is the velocity of a macro-particle, E the electric
field strength, q the charge of a macro-particle, m the mass of
a macro-particle, x the distance from a surface at which a
macro-particle takes off, and ts the time of taking off to the
voltage pulse beginning. We follow the above approach and
use E(t) functions simulating the high-voltage pulses of the
breakdown test set-up and taking into account a certain hold-
off level for the gap under consideration.
( ) ( )
An electric signal of any shape could be described at
certain accuracy with the superposition of several harmonics.
We used the following approximations of E(t).
Equations (2) and (3) produce electric field waveforms
shown in Figure 16. The waveforms look like those ones
8 -0.7 1.005
( )1.5 10eSin
( )1.4 10eSin
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 13, No. 1; February 2006
Figure 16. Functions E(t) simulating gap electric field strength waveform.
There are shown time gates for (1) short pulse and (2) long pulse when a
macro-particle has a time to cross a gap.
observed in the experiments.
Solving equations (1) for 0.314 cm gap1 and the voltage
pulse amplitude of 220 kV, we have found maximal sizes of
macro-particles having a time to cross the gap. Doing this, we
consider speculatively that a macro-particle starts motion at
electric field strength of 0.5 MV/cm and has to reach an
opposite electrode 10 ns before the end of pulse. These
conditions produce the time gate within which the equations
(1) were solved. Results for the both pulse lengths are
presented in Table 1. Thus, just nano-particles being close to
molecular clusters could be considered in experiments on
short voltage pulse breakdowns.
Table 1. Result of the calculation of macro-particles parameters, having a
time to cross the 0.314-cm gap for different times. M is the number of
nucleons in the macro-particle and Q the number of elementary charges.
tp (FWHM), ns
Having analyzed the possible sizes of macro-particles, one
can estimate conditions at which a particle could take off the
surface due to an electrostatically inducted charge. Particle can
get charged if the charge becomes equal or exceed the
elementary charge. The charge induced on a spherical droplet
can be estimated according the expression :
Q r E
where r is the droplet radius. Under assumption of a carbon
particle and the macro-field E = 0.5 MV/cm, Q becomes equal
the elementary charge if there is the local enhancement of
electric field with factors of 38 for a 1860-nucleons particle
and of 104 for a 410-nucleons particle. Note the β = 38 is
probably realistic condition for a geometrical enhancement of
the electric filed at a sharp protrusion, while β = 104 could be
associated with some dielectric inclusion . Though, the
speculation about the nature of electric field enhancement
1 The reason for such a gap length is stipulated by the assumption of
maximum E-field to be of about 0.7 MV/cm at voltage pulse amplitude of
can’t be considered too seriously since there is no way to
check this experimentally.
Under strong enough electric fields, electron emission
processes could take place at a macro-particle. These
processes are expected to simplify the detachment of a particle
from a surface since they could make a particle charged at
local electric field predicted by equation (4). We skip an
analysis of such a possibility because of its complexity.
Another way for a particle to gain the charge is the
secondary electron emission charging a particle positively.
This process is most probable if a particle happens to be
situated on the anode opposed to an emission site on the
cathode. Herewith, a primary high-energy electron originated
from the cathode causes the avalanche of secondary electrons
from the anode, including Auger electrons, and x-ray quanta.
This avalanche seems to be the most probable reason for
charging the macro-particle, because target-emitting electrons
and x-rays have much lower energy in comparison with a
primary electron and they are numerous.
The final question is about possible mechanisms of the
breakdown development after the triggering by an impact of
an accelerated nano-particle to a surface. Unfortunately, there
are no data in the literature on the matter, so, we have to settle
for data and notations concerning to micron-sized particles.
Metallic vapor is considered to be a prime cause of the
breakdown. A macro-particle produces vapor due to either
direct impact to a surface (ablation)  or self-evaporation due
to an electron flow from an emission site at the cathode .
The prebreakdown electron flow ionizes the vapor and
produces the initial plasma. The above scenario implies
particles to be big enough to produce essential amount of
vapor. The amount of vapor would be negligible in a case of a
nano-particle. However, such particles have great velocities,
which produce just a plasma cloud as a result of the impact.
Table 2 contains result of simulations  of impact plasma
contents for a SiO2 particle at various particle velocities. Thus,
a nano-particle could produce 100%-ionized plasma when
impacts to a metallic surface at cosmic velocities.
Table 2. Plasma composition after SiO2 1-µm particle impact to SiO2 target
v, [km/s] Si+[%] Si++[%]
40 2.3 0
60 14 0
80 51 0.003
100 83 0.4
150 54 46
The further question arises whether the amount of plasma is
enough to provide sufficient current to sustain an arc
discharge. To estimate the amount of plasma let us take into
account data in Table 1 and assume that the total particle
energy is utilized for vaporization of the target material.
Furthermore, we consider all the evaporated target material to
be plasma. Such estimation gives the size of a hemispherical
plasma cloud of about 6 µm if the average plasma
concentration is of 1020 m-3. The problem of the gap current at
the plane-parallel geometry when there exists a hemispherical
emitter with the unlimited emission ability [plasma] has been
A. V. Batrakov et al.: On Priorities of Cathode and Anode Contaminations in Triggering the Short-pulsed Voltage Breakdown in Vacuum 50
solved elsewhere  and the solution is as follows (in SGS
where V is the applied voltage, d the gap length, R the emitter
radius, and i the gap current. The current would be about
0.2 mA in the conditions under consideration. Such a current
is of four orders of magnitude lower than the threshold level.
To rescue our speculations, we have to suppose that the impact
produces either some specific hollow hemispherical plasma
cloud accompanied with the positive space charge of ions in
plasma sheath or a contracted plasma jet similar to that
appeared in simulations of a cosmic dust particle impact to a
Finally, it seems that the hypothesis of possibility (more so,
domination in some conditions) of nano-particle induced
breakdowns could not be evaluated in details because of lack
of reference data, and further development of the concept calls
for the goal-directed experimentation.
In experiments with anode-probe scanning followed by 100-
ns-range breakdown tests, it has been proved that cathode
emission sites are responsible for breakdowns at relatively low
hold-off fields. At higher electric fields of up to 1 MV/cm, the
anode share in the mechanism of triggering breakdown
becomes probably more significant than the cathode
Cathode emission sites are associated with irregularities of
the surface relief. These irregularities are rather traps for tiny
loosely bound particles being direct cause of emission. Solvent
cleaning of electrode surfaces removes most of particles and
deposits the remaining particles among surface relief
irregularities. Such particles could be considered also as a
cause for breakdown triggering at extremely high hold-off
electric fields. In this case, such particles operate probably as
projectiles producing the primary cathode plasma as a result of
Intentional contamination of one of the electrodes results in
either an absence of correlation between the breakdown fields
and which electrode is under contamination or rather a slight
correlation. The same is true about the influence of pulse
length on the above correlations. All these observations
indicate that there are no gates in breakdown conditions within
which either mechanism gains a dominating role in breakdown
initiation. This is partially because of non-taken into account
factors whose presence could also affect breakdown fields.
Nevertheless, available data allow one to conclude that there
are conditions at which the priority of the anode mechanism of
triggering the electrical discharge in vacuum takes place at
high-voltage pulses with durations shorter than 100 ns.
The work was carried out due to the financial support by
Sandia National Laboratories (Contracts 18754, 119716, and
306919). The authors are appreciative of much scientific
interest and encouragement of Dillon McDaniel and Kenneth
Struve. We thank also Gregory Ozur, Constantine Karlik,
Valentine Sedoi, and Anatoly Ushakov for assistance and
participation in the work.
 G. A. Mesyats, S. P. Bougaev, D. I. Proskurovsky, V. I. Eshkenazy, and
Ya. Ya. Urike, “Study of initiation and development of short vacuum
gaps pulse breakdown in nanosecond range. Part I”, Proc. 3rd ISDEIV
(International Symposium on Discharges and Electrical Insulation in
Vacuum), Paris, France, pp. 200–205, 1968.
 D. I. Proskurovsky and Ya. Ya. Urike, “The spark current increase
during steady voltage breakdown in vacuum”, Proc. 4th ISDEIV,
Waterloo, Canada, pp. 84–87, 1970.
 W. P. Dyke, J. K. Trolan, E. E. Martin et al., “The field emission
initiated vacuum arc. I. Experiments on arc initiation”, Phys. Rev., Vol.
91, No. 5, pp. 1043–1054, 1953.
 D. Alpert, D. A. Lee, E. M. Lyman, and H. E. Tomaschke, “Vacuum
breakdown for broad area tungsten electrodes”, Proc. 1st ISDEIV,
Cambridge, Massachusetts, USA, pp. 3–15, 1964.
 R. Latham, High Voltage Vacuum Insulation: the Physical Basis,
Academic Press, London, UK, 1981.
 N. Pupeter, A. Göhl, T. Habermann, A. Kirschner, E. Mahner, G.
Müller, and H. Piel. “Influence of surface roughness and preparation,
bulk purity and heat treatment on electron Field emission from Nb and
Nb3Sn”, Proceedings of the 7th Workshop on RF Superconductivity,
pp. 67–77, Gif-sur-Yvette,France, 1995. Proceedings also published as
internal Saclay report CEA/Saclay 96 080/1.
 L. Cranberg, “About Electrical Breakdown in Vacuum”, J. App. Phys,
Vol. 23, pp. 518–522, 1952.
 I. N. Slivkov, “On the mechanism of electrical breakdown in vacuum”,
Soviet Phys. – Tech. Phys., Vol. 27, No. 2, p. 1928, 1957.
 E. W. Gray and R. W. Stinnet, “Energy Losses in Magnetically Insulated
Transmission Lines due to Microparticles”, J. Appl. Phys. Vol. 61, pp.
 D. J. Johnson, M. E. Savage, R. A. Sharp, A. V. Batrakov, and
D. I. Proskurovsky, “Investigation of Pulsed Breakdown between
Polished, Powder Coated, E-beam Treated Stainless Steel Electrodes
with 2 to 8 mm Gaps”, this issue.
 A.V. Batrakov, D.H. Mc Daniel, D.S. Nazarov, G.E. Ozur, S.A. Popov,
D.I. Proskurovsky, and V.P. Rotshtein, “Enhancement of the Electric
Strength of Vacuum Insulation by Pulsed Electron Beam Treatment”,
Proc. 11th Intern. Pulsed Power Conf., Baltimore, USA, pp. 519–524,
 T. Y. Chang, “Improved Uniform-Field Electrode Profiles for TEA
Laser and High-Voltage Applications”, Rev. Sci. Instrum., Vol. 44, pp.
 T. Kawai, M. Naitoh, M. Nogami et al, “PR-80 high current ion
implantation machine”, Nuclear Instrum. and Methods in Phys.
Research, Vol. B21, No. 2-4, pp. 239–244, 1987.
 K. Narihara, K. Toi, Y. Hamada et al, “Observation of dust particles by
a laser scattering method in the JIPPT-IIU tokamak”, Nuclear Fusion,
Vol. 37, pp. 1177–1182, 1997.
 F. Y. Huang, H. H. Hwang, and M. J. Kushner, “A model for transport
and agglomeration of particles in reactive ion etching plasma reactors”,
J. Vac. Sci. Technol. A, Vol. 14, pp. 562–566, 1996.
 Sedoi V.S., Valevich V.V., “Characterization of ultra-fine powders
produced by the exploding wire method”, Proc. 32nd International
Annual Confer. of ICT, Germany, Karlsruhe: Fraunhofer ICT., pp. 80-1
– 80-10, 2001.
 K Saïdane, M. Razafinimanana, H. Lange et al, “Fullerene synthesis in
the graphite electrode arc process: local plasma characteristics and
correlation with yield”, J. Phys. D: Appl. Phys., Vol. 37, pp. 232–239,
 N.S. Xu and R.V. Latham, “Electrical and spatial correlations between
direct current pre-breakdown electron emission characteristics and
subsequent breakdown events”, J. Phys. D: Appl. Phys.,Vol, 27, pp.
 M. Okawa, T. Shioiri, H. Okubo, and S. Yanabu, “Area Effect on
electric Breakdown of Copper and Stainless Steel Electrodes in
Vacuum”, IEEE Trans. Electr. Insul., Vol. 23, pp. 77–81, 1988.
 High Voltage Vacuum Insulation. Basic Concepts and Technological
Practice. Edited by R. Latham, London: Academic Press, 1995.
IEEE Transactions on Dielectrics and Electrical Insulation Vol. 13, No. 1; February 2006 Download full-text
 G. E. Ozur, D. I. Proskurovsky, V. P. Rotshtein, and A. B. Markov,
“Production and application of low-energy, high-current electron
beams”, Laser and Particle Beams, Vol. 21, pp. 157–174, 2003.
 N. V. Tatarinova and Yu. V. Grigoriev, “Total Voltage Effect and
Electrode Porosity ”, Proc. 16th ISDEIV, Moscow-St.Petersburg, Russia,
pp. 49–52, 1994.
 D. K. Davies and M. A. Biondi, “Detection of electrode vapor in gap
prior to current amplification in vacuum”, Proc. 3rd ISDEIV, Paris, pp.
 K. Hornung, Yu. G. Malama, and K. S. Kestenboim, “Impact
Vaporization and Ionisation of Cosmic Dust Particles”, Astrophysics and
Space Science, Vol. 274, pp. 355–363, 2000.
 S. Ya. Belomytsev, S. D. Korovin, and I. V. Pegel, “Effect of the
expanding Explosive-Emission Centers Plasma on the Impedance of a
High-Current Diode”, IEEE Trans. on Plasma. Sci., Vol. 27, pp. 1572–
 S. B. Bazarov and V. A. Skvortsov, “Cumulative Effect by High
Velocity Interaction”, Int. J. Impact Engng., Vol. 17, pp. 47–55, 1995.
This paper is based on a presentation given at the 21st ISDEIV Conference,
Yalta, Ukraine, September 27 - October 1 2004.
Alexander V. Batrakov was born in Tomsk, Russia
in 1963. He graduated from Tomsk State University
in 1984. He has received the Ph.D. (physics and
mathematics) degree in 1997 from the Institute of
High-Current Electronics Tomsk, Russia. Since
1984, he has been with the Institute of High-Current
Electronics. He has the senior research physicist
position now. His scientific interests are in physics
of prebreakdown phenomena in vacuum, vacuum
breakdown, and vacuum discharge.
Sergey A. Onischenko was born in Astana,
Kazakhstan in 1982. He received the B.Sc. degree in
physics from Tomsk State University, Tomsk,
Russia in 2003 and currently is studying for the
M.Sc. degree at the same university.
Dmitry I. Proskurovsky graduated from the Tomsk
Polytechnical Institute, Russia, in 1961. He received
the Ph.D. degree in 1969 and the Sc.D. degree in
physics and mathematics in 1982. He became a
Professor in 1990. Since 1970, he has worked at the
Institute of High-Current Electronics of the Russian
Academy of Sciences as a Senior Research Scientist
(up to 1980) and has been the Head of the Laboratory
up to now. He is an expert in the physics of vacuum
discharges, explosive-electron emission, generation of
low-energy high-current electron beams, and applications of such beams for
surface modification of materials.
David J. Johnson received the Ph.D. degree in
nuclear physics from the University of Iowa in 1970.
He did experimental work on the dense plasma focus,
z-pinch implosions, and electron beam diodes at the
Air Force Weapons Laboratory from 1970 to 1974
and the Naval Research Laboratory from 1974 to
1977. He worked on high power pulsed diodes for
ion beam fusion, multistage induction accelerators,
and ion beam surface treatment applications at Sandia
National Laboratories from 1977 to 1999. He worked
on nanosatellite development from 1999 to 2001 and HV breakdown from
2001 to 2005. He was a Distinguished Member of the Technical Staff at
Sandia National Laboratories until he retired from SNL in Feb. 2005.