IEEE Transactions on Electrical InsuLlation Vol. E1-20 No.

4, August 1985 b-9 1

PREBREAKDOWN CONDUCTION IN VACuum GAPS UNDER

SWITCHING IMPULSE ExcITATIONS

K. V. Srinivasa and G. R. Nagabhushana

Indian Institute of Science
Department of High Voltage Engineering
Bangalore, India

ABSTRACT
The present investigation concerns prebreakdown con-
duction in vacuum gaps under slow impulses. Hemispheri-
cal, plane, and needle electrodes have been used in a
vacuum emvironment of better than 0.1 mPa. The gap
spacings were 0.5, 1, and 1.5 mm. The front times
ranged from 200 ps to 5 ins. The results indicate that
prebreakdown currents were smaller than the highest
sensitivity of 10-7 Afor front times of about 1 ins.
For larger front times, breakdown currents were ob-
served, which increased with front time. The voltage-
current relationship appears to follow Fowler-Nordheim
theory but Fowler-Nordheim plots are branched. The
high magnitudes may be attributed to ion-assisted field
emission based on which a model has been developed,
that agrees substantially with experimental results.
Applicability of non-metallic type of emission is dis-
cussed.

INTRODUCTION plane electrodes of 30 mm diameter (edge rounded off to

a radius of 1. 5 mm), (2) needle-plane, needle tip diam-
Prebreakdown conduction in vacuum gaps have been ex- eter 0. 1 mm and plane as anodle and (3) needle-needle.
tensively studied under dc excitation, under clean con- In each case, the first mentioned is the low-voltage (LV)
ditions of vacuum and electrode surfaces. Though it is electrode. Gap spacings of 0.5, 1, and 1.5 mm have been
generally believed that Fowler-Nordheim [1] (F-N) studied. The hemispheres and plane electrodes were of
theory satisfactorily explains the process of prebreak- stainless steel Type 304. Needles of stainless steel
down (pbd) conduction, in the recent past Latham and 304 and also of a special alloy steel were used. The
co-workers [2] Hurley et al. [3] have shown that emis- special alloy steel needles get mildly rusted when kept
sion could be of the "non-metallic"' type. The non- immersed in water over a day. These needles were used
metallic type of emission also gives rise to an approx- to study the effect of such rusting on pbd currents.
imately linear F-N plot having a high field intensifi- The electrodes were polished to mirror finish with
cation factor ~3. Very few, studies have been reported micron-sized alumina powder and ultrasonically cleaned
on pbd conduction in vacuum gaps under ac and switching before use. Great care was taken to prevent any con-
impulse excitations [4-11]. These studies are more tamination during handling. Impulse voltages were
complicated because of the presence of gap displacement generated using a 4-stage Marx circuit (50 kV, 0.5 pF/
current which is dominant at low applied voltages. In stage).- A non-inductive resistor of 30 kG was used for
the present work pbd currents have been measured under limiting the current and minimizing the electrode damage
slow impulses for different front times, gap settings, on breakdown of the gap. A current sensing resistor of
and electrode configurations. 0.81 kG was used in series with the LV electrode. The
voltage and current were recorded with an oscilloscope.
The front times of the impulses were 200 750 ps, pis,

Authorized licensed use limited to: Northeastern University. Downloaded on May 27,2024 at 06:07:28 UTC from IEEE Xplore. Restrictions apply.
 IEEE Tnansa EEl,t
--tn
on cal Tnlnauiltion Vol. EI-2C0 No.4. Al! ugut 1985

reduced to 80% of the sparking value and maintained con- addition to the delay and asymmetry, the pbd currents
stant for 10 minutes. From experimental observations, under impulse voltages have been found to be at least
the maintenance voltages during ac conditioning were an order higher than under comparable 50 Hz ac stress.
found to be 18, 36, and 48 kV for 0.5, 1 and 1.5 mm It is worth noting that pbd current magnitudes under
gaps, respectively. A few experiments were also con- 50 Hz voltages are about two orders higher than under
ducted with electrodes subjected to 'spark conditioning"? dc voltages [7].
in which the gaps were subjected to 20 to 30 sparks with
the current on sparkover being less than 0.5 A (peak). 0 6 1 1 1
The experiment started with the application of im-
pulses at a low level and increasing in small steps.
Five impulses were applied at each level at intervals of
30 s. That voltage level at which there were at least E 0 4 -

two breakdowns (resulting in chopped waves as seen on

the CR0) was defined to be the breakdown voltage. Z -

The pbd current was obtained by subtracting the gap

displacement current (corresponding to the applied vol- * 0-2
tages) from the total current. The highest sensitivity w
of current measurement was 10-7 A.

0 I Il
RESULTS AND DISCUSSIONS 1 2 3 4 5
Prebreakdown currents have been measured for each of TIME TO FRONT IN ms
the above mentioned configurations. Additionally, some
experiments were conducted to study the effect of con- Fig. 1(b): Variation of tirne-Zag between voltage and
tamination. (1) in one set of experiments, liquid ni- current peaks as a function of time to front of
trogen was not used in the trap so that back-streaming appZied impulse voltages
of diffusion pump fluid would contaminate the electrodes.
This is designated "contaminated condition A", and (2)
in another set of experiments, the special alloy needle, INFLUENCE OF TIME TO FRONT ON PBD CONDUCTION
allowed to rust in water over a day, was used. This
is designated "contaminated condition "B". For any given gap configuration, pbd currents were
below perceptible levels (i.e. <<10-7 A) for front
times less than or equal to 750 ps. For 1.7 ms
fronted impulses, a well-defined pbd current waveform
developed occasionally. Over the regime of 200 to
750 ps front times, microdischarges were observed
frequently. These were typically rectangular pulses,
magnitudes ranging to a few tens of mA and duration
being around 1 to 2 ms. Sometimes, a micro-discharge
resulted in subsequent development of a well-defined
pbd waveform and at some other times a pbd current
culminated in a microdischarge. It was observed that
l___ \ ___ pbd currents would be conditioned off with repeated
application of impulses or by application of a higher
voltage impulse. For 0.5 mm gaps, more often than
not, a microdischarge resulted in a breakdown. With
3 and 5 ms fronted waves, pbd currents developed in
a regular manner. It is found that the current mag-
Fig. 1(a): Typical pbd current oscillogram of 3 ms nitudes increase as front time increases from 1.7 to
Front iapulse excitation
Upper beamr (voltage) 120 kV/div,
3 to 5 ms, with saturation becoming evident towards
5 ms (see Fig. 2).
Lower beam (current) 63 uiA/di-d,
Sweep 2 ms/div.
Peak has been indicated by an arrow. INFLUENCE OF CONTAMINATION

In every case, the pbd current waveform was "asym- Rusted needle cathodes (contaminated condition B)
metric"
metric" sensehatthecurretpeakalwaysgave
ntheinsnse
the tha the curent pek alway the highest pbd currents. These were almost two
orders of magnitude Citionparesute
higher compared to currents under
occurred later than the voltage peak, the time lag normal o niti. currentsu
ranging from 0.3 to 0.5 ms. The current magnitude in that w tice t e onlytaout
currentu
that were only about twice the current under normal
tailtailwaswashigher for aa given voltage.
the the Fig. 1(a)
voltage . Fi. a conditions .
shows a typical oscillogram. Fig. 1(b) shows time
lag between the voltage and current peaks as a func-
tion of front time of the applied impulse. The consis-
tent increase in the time lag with front time confirms INFLUENCE OF SPARK CONDITIONING
that the time lag is due to phenomena in the gap. It codtoigo.lcroe ie iet
Spr
has also been verified that the time lag due to the sytmicgohofpdureswihaealouh
measuring circuit is < 1 us, and therefore should be
precluded . Simi lar time lags between voltage and cur- hge nmgiue
rent peaks and asymmetry in the current waveform have
been observed also by earlier workers [6,7, 8, 9]. In

Authorized licensed use limited to: Northeastern University. Downloaded on May 27,2024 at 06:07:28 UTC from IEEE Xplore. Restrictions apply.
Sriniva4Sa and Nagabhushana: PrE-breakdo-wn conduct-ioin a-n vacuum gaps 69-

4X1064 W ; C a ~
7X10 ~5 t
~ ~~~~~~~~~~~ 7X 1-
0-

o -13 \ S B\ (73)
co~~~~~~~~~~~~~~~~~1
-3~~~~3
Fig f:pdcretai ueto ffottm f2\
_ \
i\\ t \A l
B

The voltage-current relationship was analyzed in 10 F^ \ \

terms of the classical F-N theory. The F-N plots are_I\\ 8^
shown for various conditions in Figs. 3 and 4. As nay C
be expected from the asymmetry of the current wave-
o
form, the F-N plot is a branched curve (in the clock- o\ ' (64)
wise direction) rather than a straight line, (see _|\.
Fig. 3). In Fig. 3 are also shown F-N plots to illus- (05
trate the influence of front time and spark condition- I C6O _
ing. Increase of front time from 1.7 to 5 ms has I Abet
resulted in higher currents and reduced the values of (64)
6, the field intensification factor. The latter is
unexpected and the reasons are also not very clear at A
C1
present. Spark conditioning may be seen to have
shifted the F-N plots upwards very significantly and (E
the curvature of the F-N plots may be seen to have
1cok
changed. The values of 6 have increased as a result 1*O 1.5 2.0 2.5 3.0
of spark conditioning. These are understandable in IV Xs105 (volf1)
the sense that a more rough surface microstructure 1
(due to spark conditioning) is compatible with higher
r values and hence higher emission currents. Also, Fig. 3: Typical F-N plots of hemispherical-plane
such a surface can give off vapors more readily to electrodes with 1.5 mmi gap A and B time to front =
enable space-charge formation. 5 is, A Normal conditioning, B Spark conditioning,
C Time to front = 1.? n s, Nnormt ac conditioning.
All quantities in bracket indic ate vyoltages in kV
DISCUSSION (peak) .
Most of the above observations support the view that In view of the above, a m 2del based on the I-F emis-
an anode phenomenon is active. It is believed that sion theory of Porotnikov and Rodnevich [13] has been
the high energy electron bean impinges on the anode developed that enables computation of pbd currents.
causing vaporization. The vapors get ionized and the
positive ions, on approaching an emitter sufficiently
closely, significantly enhance the emission, explain-MOE
ing the observed high currents. Anode time constants MODEL
have been reported to be about 1 ms [12]. If the Porotnikov and Rodnevich proposed the following
front time is just comparable or less than the anode expression for current density considering enhanced
time constant, one may expect the anode phenomenon emission due to ions (i.e. space charge)
not to be influential. The absence of perceptible
pbd currents for front times of 750 ibs and less and
erratic development of pbd currents for 1.7 ms front 1.5x106 62E2 e [ 85X107 01.5 (1)
impulses appear to be strong evidence of the anode
.

IF 0
phenomenon being a main cause of the enhanced pbd
currents. Furfher, the slowanode process es would re-
3xsuf7 0ic ie
sult in a delayed current peak. The longer the time
fl
-- -
to front, the longer the activity of the anode phenom- Here the usual notation is adopted and as is common
enon, thereby enhancing emission. Thus higher cur- the functions t2(y) and V(y) are equated to unity.
rents could be expected with larger wave fronts. Further aend/0«1 and is neglected.

Authorized licensed use limited to: Northeastern University. Downloaded on May 27,2024 at 06:07:28 UTC from IEEE Xplore. Restrictions apply.
694 IEEE Transactions on Electria-l Tnsulation Vol. EI-20 No.4. ugst 1935

13 _________________ for the computation. The F-N curves are plotted for
7 X 10 _ different voltages in kV (peak). They are shown in
Fig. 5. It is seen that experimental and theoretical
X \(37 5) curves have reasonable similarities, justifying the
model.

- ~ >\Apart from the reasonableness of the model discussed

r
\iOt(36) | above, some discrepancies deserve to be discussed.
The work of Latham et al. [2] and Hurley et al. [3]
seem to affirm a "non-metallic" type of emission that
\ ~vi\ 35) are F-N like. Thehigh F values of earlier investiga-
'\ t \tions have been attributed by them to field enhance-
ment by the presence of non-metallic particles, rather
1013 than microprojections. However, in the present ex-
periments under impulse voltages F-N plots have yielded
relatively lower F values than under dc or ac. Further-
more, "rusted needles" have yielded F-N plots with
still lower F values, the branches of the plots are
0 \ \\\ \\ much better straight lines and also are closer to-
> - B\;\gether. This seems to be in some conflict with the
XIT C 5<\ \ _ observations of Latham [2] and Hurley [3]. It is also
E: C *- \\ can
not clear how the waveshape of the applied voltagecoming
(37.5) influence F or if there are any other parameters
into the picture. Further work is in progress.

(30)' ~X (31)N ECQ.

4X105I
2 3 4 5 6 0
>
1/V x 165(VOLT')
Fig. 4: Typical F-N plots of needle-pZane eZectrodes (35) RI
with 1.5 mm gap i t ) (30' \
Front time = 3 ms (polarity is positive) bX
A Standard conditions, B Rusted needZe, and 168 (2 5)
C Without liquid nitrogen in trap.
AZZ quantities in bracket indicate voltages in kV _
6 x
(peak). 21 3 4 5
-8 -1
l/E X 108 (VOLT cm)
In Eq. (1) JIF is the current density (A/cm2), E is
the average electric field intensity (V/cm), 0 is the
work function (eV) and F'is the field intensification Fig. 5: F-N plots as per the theoretical model
factor. Eq. (1) does not incorporate temporal growth and 0 = 2 eV
of current but only gives the final stable value. an _ e
Since, in the present case, the applied voltage is (all
Pea
kq
time dependent, it becomes necessary to account for
the temporal growth. If the applied field intensity
El is changed to E2 in a time int3rval At, starting
at the instant t=0, it is now postulated that current CONCLUSIONS
density J(t) for t>0 is given by: Prebreakdown currents in vacuum gaps under slow im-
J(t) = J2 + (Jl-J2)exp (-t/a) (2) pulse excitations with Lront times from 200 Ps to 5 ms
have been studied. The results strongly indicate that
where Jl and J2 are current densities as given by Eq. an anode mechanism is at work to influence the current
(1), corresponding to: El and E2 respectively, and a is magnitude for front times of 1 ms or higher. This re-
the time constant of current growth. Thus even though sults in (1) current magnitudes one order higher than
the voltage is raised instantly to the value E2 the under comparable 50 Hz ac stress, '2) the current peak
current increase occurs gradually with a time constant occurring later than the voltage peak and (3) an
as. Thus in the case of impulses, the current continues "asymmetry" in the current waveform in that the current
to increase for a little while after the voltage has is higher in the tail for the same voltage. These
reached the peak. Therefore, the current peak occurs result in F-N plots that are branched. It has been
later than the voltage peak. By a similar process it possible to explain these on the basis of the ion-
may be seen that the currents will be different for the assisted field emission theory ( f Porotnikov and
same instantaneous voltage on the front and tail por- Rodnevich after incorporating time dependence in terms
tions of the wavre. A relatively low value (2 eV) of of the anode time constant. The agreement is con-
work function and a moderate F - lue (200) are assumed sidered satisfactory. Some features of the present

Authorized licensed use limited to: Northeastern University. Downloaded on May 27,2024 at 06:07:28 UTC from IEEE Xplore. Restrictions apply.
SrinivasS and Nagabht h-ana: Prbr-eakdcwn conduction in vacuum gaps 695

results, namely, adequacy of lower values of work [7] G. Parthasarathy and H. V. Gopalakrishna, "Pre-
function and S to explain the current magnitude seems breakdown phenomena in vacuum gaps subject to
to conflict with the "non-metallic" emission model of alternating voltages", Proc. IV Int. Symp. on
Latham et al. and Hurley et al. Disch. and Elect. Insul. in Vacuum, p 214, (1970).

[8] G. A. Farrall, "A study of emitters in broad area

ACKNOWLEDGMENTS electrodes using a phosphor screen", Report No.
71-C068, General Physics Laboratory, Feb (1971).
The authors wish to thank the authorities of the
Indian Institute of Science for supporting this work. [9] Bernd-D, Schmidt, "AC breakdown of vacuum gaps at
One of the authors (K.V.S) is indebted to University high voltages", III Int. Symp. on High Voltage
Grants Commission, Government of India for sponsoring Engineering, Milan, August (1979).
him under the Quality Improvement Program.
[10] T. Shiori, I. Ohshima, M. Honda, H. Okumara,
H. Takahashi and H. Yoshida, "Impulse voltage
REFERENCES field emission characteristics and breakdown de-
pendency upon field strength in vacuum gaps",
[1] R. H. Fowler and L. NQrdheim, "Electron Emission IEEE Trans., Vol. PAS-101, No. 10, p 4178,
in Intense Electric Fields", Proc. Roy. Soc., October (1982).
Vol. 119A, p 173, (1928).
[11] K. Tsuruta, "Prebreakdown field emission current
[2] N. K. Allen, B. M. Cox and R. V. Latham, "The and breakdown mechanism of a small vacuum gap",
Source of high 6 electron emission sites on broad IEEE Trans., Vol. EI-18, No. 3, p 204, June
area high voltage alloy electrodes", J. Phys.D: (1983).
Appl. Phys., Vol. 12, pp 969-977, (1977).
[12] F. Rohrbach, "Relation between spark time-lag
[3] R. E. Hurley, "Electrical phenomena occurring at spectra and the mechanisms leading to breakdown
the surface of electrically stressed metal cath- between plane Titanium electrodes in Ultra High
odes, I Electro luminescence and breakdown Vacuum", Proc. IV Int. Symp. on Discharge and
phenomena with medium gap spacings", J. Phys.D: Elect. Ins. in Vacuum, Waterloo, p 68, (1970).
Appl. Phys., Vol. 12, pp 2229-2245, (1979).
-jC. C. Erven, J. J. Wavre and R. G. Van Heeswijk, [13] A. A. Porotnikov and B. B. Rodnevich, "I-F Emis-
"60 Hz vacuum breakdown studies in a dynamically sion", Soy. Phys. Tech. Phys., Vol. 23, No. 6,
pumped system", IV Int. Symp. on Disch. and Elect. p 740, (1978).
Ins. in Vacuum, Waterloo, p 219, (1970).
This paper was presented at the InternationaZ Symposium
[5] R. Hackam, "Comparison between 50 Hz ac and dc
vacuum breakdown", J. Phys.D: Appl. Phys., Vol. 8, on Discharges and Electrical Insulation in Vacuum,
pp L53, (1975). BerZin, Germany DDR, 24-28 September 1984.
pp~~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~Mnsrp L5, (1975). Ma 185
[6] K.A.N. Kutty and G. R. Nagabhushana, "Prebreakdown Manuscript was received 2 May 1985.
current measurements in vacuum gaps stressed by
ac voltages", Proc. IEE, Vol. 123, No. 5, p 475,
(1976).

Authorized licensed use limited to: Northeastern University. Downloaded on May 27,2024 at 06:07:28 UTC from IEEE Xplore. Restrictions apply.