Solid Insulators in Vacuum: A Review

(Invited Paper)
R Hawley, C A Parsons & Co Ltd, Heaton Works, Newcastle upon Tyne 6

After reviewing the prebreakdown phenomena (leakage currents, outgassing and luminosity) that arise
across a solid insulator held between two electrodes in a vacuum gap, the factors that affect the
magnitude of the breakdown voltage (conditioning, electrode and insulator material and surface
finish, insulator length and shape, pressure, etc) the phenomena occurring in the region of the
cathode-insulator junction and the hypotheses put forward to explain the breakdown mechanism are
discussed.

As is well known if the electrodes in a vacuum gap are separated Current bursts that did not develop into breakdown took place
by a solid insulating material, breakdown over the insulator and on occasions the current suddenly changed to a new
takes place without its puncture, at a lower voltage than in a relatively steady value.
straightforward vacuum gap. Quite often breakdown along the Hamisch et al”* have alsoimade a comprehensive investiga-
surface of a piece of insulation can be the limiting factor in the tion of the prebreakdown phenomena. They found that weak
operation of vacuum insulated equipment. The breakdown current discharges occurred that caused only small voltage drops
mechanism across insulators is therefore of very great practical between the electrodes (100 to 300V at 30 kV total voltage with
importance because at the moment the operation of any high a series resistance of 10 megohms). Such discharges were as-
vacuum high-voltage device necessitates either constructing the sociated with glow phenomena and at a fixed voltage their fre-
vacuum chamber itself of an insulating material or insulating quency decreased with time. In fact these discharges were
the high-voltage conductor as it passes through the wall of a analogous to microdischarges in a plain vacuum gap. In addition
metallic vacuum chamber. In addition there is often the problem to these microdischarges there was a steady current component
of supporting the high-voltage conductor within the vacuum that fluctuated only very slowly.
chamber. Hamisch et al found that during the steady current stage the
Whilst a great deal of rather random information has been insulator surface exhibited a uniform weak glow. This they
published about the phenomenon of breakdown across in- took to indicate the surface was being uniformly bombarded
sulators, most of which is concerned with the fabrication and by electrons. This presupposes an electrical field directed to-
operation of high-voltage accelerator tubes, there have recently wards the insulator. If such a field is not originally present
been several systematic studies of this phenomenonl-lO. These Hamisch et al postulated it arises through the charging of the
recent studies have enabled the factors responsible for initiating surface in following manner. Field emitted electrons from the
the breakdown to be established and have therefore led to insulator-cathode junction region hit the insulator and release
methods of improving the electrical strength. secondary electrons. Thus the insulator becomes positively
charged so that there exists a field component, directed towards
the insulator, which brings back the released electrons. On
Prebreakdown phenomena colliding with the insulator new secondary electrons are re-
As the voltage is gradually increased from zero across an leased. The process is repeated and there results an electron
insulator several events occur. A prebreakdown leakage current, movement in small jumps towards the anode. Amplification of
which can be either in a relatively steady or pulsed form, starts this current does not occur because the charging process is
to flow as electrons are released into the vacuum from the such that a mean yield of secondary electrons of unity occurs
cathode-insulator junction. These electrons then cause charging at all points along the insulator. As the electrons which reach
of the insulator surface and also the release of x-rays from the the anode, in the main, do not jump from the insulator until
anode surface. Luminous areas appear on the insulator and the close to the anode they impinge onto the anode with a low
release of gas from the insulator surface can occur. Further energy_ Ham&h et al showed theoretically that the surface
increase in voltage produces increased leakage currents, out- charge on the insulator was proportional to the field strength
gassing and luminosity until ultimately streamers start to pro- and dependent on the angle of inclination of the insulator to the
pagate across the insulator surface and then breakdown takes cathode surface. With the sides of the insulator perpendicular
place to the cathode the charge is positive and as the angle of inclina-
The first comprehensive study reported of the prebreakdown tion increases the charge decreases. At some angle the surface
phenomena was that made by Gleichaufl*e who found that the charge is zero and as the angle increases further the charge
prebreakdown current decreased in magnitude with successive becomes negative.
voltage applications to the same insulator but that the results Hamisch et al measured the energy of the electrons reaching
exhibited a wide scatter. There appeared to be no critical break- the anode as a function of the distance away from the insulator
down current. Sometimes breakdowns occurred after a current of and found that the slow electrons impinged quite near to the
about lo-l1 amps and in other cases after acurrent of IO-‘amps. insulator and with an increasing distance from the insulator the

Vacuum/volume 18lnumber 7. Pergamon Press Lfd/Prinfed in Great Britain 383

 R Hawley: Solid insulators in vacuum : A review

total current density, after passing through a maximum, de- previous history of the insulator. As the time interval between
creased, but the proportion of rapid electrons making up the the successive series of tests was increased, the initial breakdown
diminishing current density increased. voltage on resumption of testing decreased but was usually of
By means of a probe Hamisch et al also measured the charge a higher value than the very first breakdown voltage for the
on the insulator left after stopping the discharge process. Figure insulator. The same loss of conditioning was found if the
1 shows the results obtained as a function of the capacitor insulator was taken from the continuously pumped vacuum
chamber and exposed to the atmosphere, the length of exposure
to the atmosphere being immaterial.
Multiple breakdowns can cause irreversible damage to an
insulator surface (by producing breakdown tracks or depositing
metallic vapour from the electrodes) or to the joints between
the insulator and the electrodes and thus lower rather than
increase the breakdown strength. The correct choice of a surface
coating on the insulator can inhibit the formation of breakdown
tracks and assist in improving the electrical performance.
Gleichaufl was also able to condition insulators by applying
a high-voltage to the insulator continuously. In this case break-
down occurred less and less frequently at the constant voltage.
Shannon et al8 have also conditioned insulators by maintaining
across them a voltage lower than the ffashover voltage. Under
these circumstances the prebreakdown current, the luminosity
and the quantity of gas released from the insulator decreased
with time and allowed the application of higher voltage. Such a
procedure has the obvious advantage that it minimizes the
possibility of permanent damage to the insulator.

Electrode material and surface tinish. The only deliberate in-

vestigation of the effect of varying the electrode material on the
breakdown voltage appears to be that made by Gleichauf2 who
obtained the results shown in Table 1. These show that under
the experimental conditions used the material of the electrodes
had no effect on the breakdown voltage.
Figure 1. Variation of surface charge with angle of insulator inclina-
tion (After Hamisch et aP)
Table 1. Effect of electrode material on the breakdown across solid
voltage for four different angles of inclination of the insulator insulators. (After Gleichauf3
surface. It is seen that the charge density is linearly dependent
Electrode Average Breakdown Voltage (kV)
on the angle of inclination. The positive charge at a zero angle
Material Pyrex Glass Fused Quartz
of inclination decreased linearly with increasing angle and
Stainless steei 41.0 67.5
attained a zero value at an angle of 31.5” and became negative
beyond this. Copper 44.5 66.5
Magnesium 37.01 65.5
Factors affecting the breakdown voltage Aluminium 39.0 67.0
The roles played by the various variables that affect the break-
* Improperly conditioned
down strength of a plain vacuum gap are not necessarily the Insulators in the form of rods, 22.5 mm long and 12-12.5 mm diameter,
same when a breakdown occurs across a solid insulation surface. butt-jointed to electrode.
In this latter case the effect of these general variables is often
outweighed by the physical properties of the insulator and by Material and surface condition of the insulator.
phenomena occurring at the cathode-insulator junction. Material. The breakdown voltage for an insulator of fixed shape
and length is strongly dependent on the material of the insulator.
Conditioning. As in a plain vacuum gap the phenomenon of This has been demonstrated by many investigators, in particular
conditioning takes place whereby the electrical strength im- Gleichauf* and Borovik and Batrako+ whose results are given
proves during successive breakdowns. GIeichaufl investigated in Table 2, the latter investigators obtaining much greater
the general behaviour of the conditioning process, the part strengths than Gleichauf for insulators of the same length and
played by the electrodes and the part played by the insulators material.
on the conditioning process and he also studied the effect on With a butt joint the field in the small voids at the cathode
conditioning of the circumstances under which the breakdown end of the insulator, caused by imperfect contact between the
arc extinguished. He found that with successive breakdowns insulator and the electrode, is increased in proportion to the
the breakdown voltage followed a genera1 trend towards higher dielectric constant of the insulator. Kofoid’ has obtained results
voltages, the rate of increase diminishing with time. When the that indicate that the voltage required to release electrons from
voltage was removed for a while, part of the conditioning was the cathode-insulator junction and to cause breakdown is
lost but the insulator subsequently reconditioned at a faster systematically lower for materials of higher dielectric constant.
rate. The degree of conditioning lost was dependent on the From the practical point of view the bulk properties of the
R Hawley: Solid insulators in vacuum : A review

TsbIe 2. Breakdown voltage for different insulator materials. (After down voltage of a completely smooth rod. When the roughened
Gleichauf2 and Borovik and Batrako+ end of the cylinder was adjacent to the cathode however the
Material Direct breakdown voltage rose to 68 kV.
_~ Breakdown Voltage (kV)
-____ Obviously the surface of any solid insulator should be care-
Results obtained bv Gleichaufa fully cleaned before use otherwise a breakdown along the
Fused quartz 65 contaminated surface will result in a permanently low resistance
Pyrex glass 4.5 track.
Pyrex covered in oil 56-73
Soda-lime glass 40 Length and shape of insulator. As in ‘ordinary vacuum gaps the
Steatite 50 breakdown voltage across an insulator does not increase in a
Rutile CTi0.J 40 linear manner for increasing length of insulator. The results for
Zirconium dioxide 40 different materials obtained by various investigators are shown
-
Polvstvrene
I _

___
75 .~~__ __
in Figure 2. These indicate the wide divergence of electrical
- Teflon 50 .~_____ strength possible for an insulator of given material and length
Sulphur - 45 and show the importance of the cathode-insulator junction, the
Results obtained by Borovik and Batrakovs _.____~
.__- conditioning and the type of applied voltage.
Plexiglass 170-186
Vinyl sheet --- 1So-200
Pressure. For short insulators variation of ambient pressure
Polystyrene 100 _ appears to have little or no effect on the breakdown voltage.
Ebonite 120-140
For longer insulators however Ramrn’” has shown there is a
Quartz 82
most definite dependence of the breakdown voltage on pressure.
Porcelain 170
He observed the variation in breakdown voltage with pressure
Molybdenum glass 100-130
across a 65 cm long bushing. In the curve he obtained (Figure 3)
All insulation samples were cylinders about 2.2 cm long, butt-joined to each measurement represented a condition of a low current and
the electrodes.
a few sparks per ten minutes. The breakdown voltage rose to
a maximum for pressure above lo-* torr. The limit of 600 kV
insulator material are important. For example it is pointless
using a material of low dielectric strength because the break- I

’ r80kvd2
’
down will then occur through the body of the insulator.
Srivastava” has pointed out that unglazed electrical quality I(c)
porcelain is unsuitable because it is porous and often has
cavitiesand inclusions close to the surface, from which absorbed
gases are released under the action of the electrical stress. Pyrex
glass is a good material, providing it is free from bubbles, but 4(c)
it is prone to mechanical damage. Hydrostatically pressed /
aluminium ceramic insulators can be made free from bubbles
and inclusions, but if extruded can still contain surface defects. /

Surface condition. GIeicha@ established that there was some

indication that the breakdown voltage increased with the surface
resistivity of the material. To check this effect he increased the
surface resistivity of a conductive glass sample by applying a
coating of silicone oil which was then baked onto the surface of
the insulator. The breakdown voltage increased by more than
50 per cent. Improvements in the breakdown voltage due to
surface coatings have been reported by other investigators.
Vierstrala for example increased the breakdown strength of an
insulator, from about 40 kV to 70 kV, by baking a coating of
chromic oxide in a silicate binder on to the surface. Fryszman
et al@improved the strength by coating the surface with a semi-
conducting film of ferric oxide.
Gleichaufa also found that roughening the surface of a glass
or a quartz insulator improved the breakdown strength by
about 40 per cent. The same increase was obtained for different
degrees of roughness. Kalyatskii and Kassirov’” improved the
impulse strength of a polished insulator by 40-70 per cent by
roughening the surface with fine emery cloth. It is more impor-
tant to roughen or coat the area adjacent to the cathode. For I L I L I I

example Gleichauf polished a 23 mm long glass rod except for 5 IO 15 20 25 30

a 3.5 mm ring at one end. When the roughened end of the Length of insulator, mm
cylinder was adjacent to the anode the breakdown voltage was Figure 2. Variation of breakdown strength of an insulator in vacuum
49.5 kV which was not significantly different from the break- with length (Key overleaf).
 R Hawley: Solid insulators in vacuum : A review

Investigator Material of Type of cathode- Voltage Remarks

Insulator insulator joint
~~ __~~~~~~_
la
-.- Finke13
~~~~~ ~~__~~~_~~ .~ ~~- ~ Alumina
~~~ (AlaOs)
~___ ___ Butt Direct
-_-
lb ,, 9, ,, ,, After extensive long
--pp term conditioning.
IC ,z 1, Butt Shielded Cathode-insulator
,,
junction shielded by
means of metallic
~- ____.. - -_ ~~-- ~~- ~~- sphere.
z Coenraads et al’” Recessed
3 Gleichaufz Pyrex glass Butt -
~~~_____ _I) ..____
4a Kalyatskii and Teflon ,, 1, -
Kassirov”’
4b ,. II 7, Impulse Pulse rise time 2.6 ,,s
4c ,T >, ,, 7, Pulse rise time 0.2 @s
_~
5 Leaderis Polythene ,, Direct

,Limit of high voltoge

SUPPlY
I Electrode

0.75 Insuiator
3
a,
L !
f

300-

200 I j I I I
IO* D4 10-3
Pressure, torr 0.25 -

Figure 3. Effect of pressure variation on the breakdown voltage

across an insulator (After Rammle).

01 / I , I
was set by the maximum voltage output obtainable from the IO+ 1O-3 0' IO-' I IO
generators used for the test. Similar results were obtained with Pressure, torr
air, nitrogen, argon and helium.
Figure 4. Breakdown stress versus pressure (After Smith”)
Insulator right cone 2.5 cm long oiled perspex with half angle of 25”.
Breakdown over insulators in a poor vacuum. As part of a pro- Voltage 30 ns pulses.
gramme for the construction of an x-ray tube to withstand
voltage pulses of several megavolts SmithL7 studied the break-
down strength of insulators, under pulse conditions, in a poor
vacuum (1O-4 to 10-l torr). The breakdown process was
extremely fast forming a discharge 50 cm long in 2 x 1Om8set,
the resistance of the breakdown path falling to a few ohms in
a few nanoseconds. The voltage generators Smith used produced
an output pulse which rose in S-10 ns, and lasts about 30 ns,
followed by a voltage reversal of similar duration but smaller
amplitude and subsequent decaying oscillations. Under these
conditions Smith found that the breakdown field was in-
pendent of (a) gap in the range 0.5 cm to 20 cm (b) diameter of
the test cell if this is greater than the gap, (c) electrode material
(d) type of residual gas (e) gas pressure, providing this was
lower than 0.03 torr and (f) existence of x-ray dose rates up to I I I
lOLo Roentgens/sec. When the pressure was raised above 0.03 -60” -300 0 + 30” + 600
torr the breakdown voltage began to decrease (Figure 4). The Half angle
same effect was noted for insulators 5 and IO cm long so that it
Figure 5. Variation of breakdown stress with half angle across
was a pressure and not a (pressure distance) effect. The pressure conical insulators (After Smith”).
at which the strength started to decrease was strongly dependent epoxy resin - - - - glass - - - perspex (oiled) -. - polethylene
on the identity of the gas present, being lower for the more
electronegative gases, insulator and the electric field if this was less than about 25”
Smith also investigated the effect of varying the half angle of (b) a given material exhibited a wide range of strengths (20 to
the cone shaped insulator for several different materials (Figure 300 kV/cm) (c) there was a marked polarity effect, (for example
5). He observed that (a) there was an extremely strong de- for a 20” epoxy resin cone this was as high as 8:l) (d) the
pendence of breakdown strength on the angle between the highest gradients were withstood on cones of semi-angle about
R Hawley: Solid insulators in vacuum : A review

W, higher angles eventually lead to erratic behaviour as well length and material was strongly dependent on the rise time of
as a reduction in strength, and (e) breakdown occurred most the applied voltage pulse (Figure 7). The,pulses they used had
readily if the half angle was small and negative. wavefronts varying between 0.1 and 3 ps. For lengths of
insulator up to 20 mm the breakdown voltage increased with
Particle bombardment. For fused quartz and uranium glass reduction in pulse duration, the increase being dependent on
insulators Gleichaufa showed the breakdown voltage remained length.
unchanged when the test samples were exposed to ultraviolet
radiations and also, 60 to 90 kV x-rays. Formative time lag. The formative time lag of the breakdown
Under the action of rf fields electron bombardment can result across an insulator has been investigated by Bugaev and
in multipactoring, whereby, if the secondary emission coefficient MesyatP. For 3 mm long insulators, subjected to an over-
of the insulator surface is greater than unity, the release of more voltage in the form of a very fast pulse, the formative time lags
than one secondary electron by the impingement of one primary varied between 7 and 20 ns depending on the insulator material.
electron can result in a cumulative process producing a WatsonZ2 has postulated that the formative time lag is governed
breakdown. by the thermionic emission of hot electrons from within the
dielectric.
High frequency breakdown. Several studies have been made of
the failure of ceramic insulators due to surface sparking at Phenomena in the region of the cathode-insulator junction
microwave frequencies. Using a frequency of 3,000 MC/S Walker Most of the hypotheses put forward to explain the flashover
and LewiP found that breakdown was liable to occur, across along insulators depend on assuming an initial stage whereby
a titanium dioxide ceramic disc, with electric fields of the order electrons are released from the region of the cathode-insulator
of lo4 V/cm normal to the surface of the disc. A coating, only junction. In addition careful design of this junction can have a
a few thousandths of an inch thick, on the disc, of a glaze, marked effect on the electrical strength of a given length of
consisting mainly of lead borate, increased the strength above insulator. It is therefore of interest to list the various observa-
3 x lo5 V/cm, the limit attainable with the power source. The tions made of phenomena occurring in this region and the
reasons for this improvement were investigated by Hayes and methods employed to reduce electron emission from the triple
WalkerI who came to the conclusion that the breakdown interface between cathode, insulator and vacuum.
strength was improved because the glaze provided a protective
coating. They found that plain titanic ceramic had a poor Visual phenomena. Kofoid4,5 and others have carried out
resistance to sparking and a single spark or any weak initial intensive investigations of the phenomena arising at the
discharges caused permanent damage to the material destroying insulator-cathode interface. Prior to the breakdown it is possible
the surface insulation. Coating the ceramic with a lead-borate to clearly see discharges more or less uniformly spaced around
glaze prevented this damage occurring and as a result any
initial discharge caused a clean-up process to take place con-
siderably raising the breakdown voltage.

Pulse breakdown. There is not a great deal of information

available on the impulse breakdown strength of insulators.
Kofoid4 found no significant trend in the crest voltage required
to release electrons from the cathode-insulator junction when
the impulse wave-shape was varied over a wide range. Work by
Thumwoodeo, who investigated the breakdown across glass
cylinders 18 inches long, has however indicated that the wave-
shape has some effect on the breakdown voltage. Thumwood
found that the breakdown voltage was dependent on the wave-
front duration (Figure 6) and for a long enough wavefront the

25-

a-

0 I 2
Rise time, ps
Figure 6. Variation of breakdown voltage versus impulse voltage
wavefront (After ThumwoodZo). Figure 7. Variation of breakdown voltage across insulators in
vacuum with rise time of applied voltage pulse. (After Kalyatskii
and Kassirov”‘).
strength approached the direct voltage strength (which in this
case was approximately 200 kV). This is the inverse of the effects A Teflon abed-breakdown
B Organic glass
observed by Kalyatskii and Kassirovlo who with much shorter all 9.5 mm long strength under
C Vinyl plastic direct voltage
insulators also found that the breakdown strength for a given D Epoxy resin

387
 R Hawley: Solid insulators in vacuum : A review

the cathode-insulator junction. The discharges appear to exist geometry of the electrodes and the insuiator permit it, is to
simultaneously in the form of a diffused blue glow and are completely bury the end of the metal electrode in the dielectric.
steady in location, intensity and shape. Increasing the voltage Kofoid measured the breakdown voltage for the configuration
causes the luminosity to extend further towards the anode. shown in Figure 8b. The breakdown voltage was 23 kV when
the butt joint was at the cathode but 98 kV when it was at the
Methods of reducing stress at cathode-insulator junction. When anode.
a cylindrical insulator simply rests with a butt joint on to the Where possible the field should be concentrated through the
surface of the cathode Kofoid’ has pointed out that since the volume of the insulator to remove stress from the edge by using
metal and dielectric surfaces cannot be made perfectly flat voids for example electrodes recessed into the body of the insulator.
will exist at the junction. Kofoid calculated that the existence Dielectric surfaces should not be exposed to the sharp metallic
of a void of 0.025 mm thickness would produce fields at the edges and care should be taken to obtain uniform field condi-
cathode surface greater than lo6 V/cm, that is fields of sufficient tions at the ends of the insulator. Figure 9 shows the results
magnitude to produce electron emission. Since the onset of obtained by Coenraads et al” when the end conditions for two
eIectron emission herafds the initiation of a flashover aiong the glass insulators were varied. Although there is some scatter in
insulator surface Kofoid made a comprehensive study of the results, they do indicate that Pyrex Glass is in general better
methods of minimising the release of electrons at the cathode- than soda-lime glass. Metallizing the ends of the Pyrex in the
insulator junction. recessed electrodes lowered the strength due to the removal of
The first approach Kofoid4 tried was to metal coat the ends the metal during conditioning.
of the insulator. By this method any voids between the insulator Shannon, Philip and Trump8 made a comprehensive study of
and the cathode were removed. He showed however that the methods of reducing the field intensity in the region of the
coating either the ends alone or even extending the coating from insulator-cathode junction. Insulators with poor dielectric-to-
the end part way up the insulator only caused a very small cathode contacts consistently gave inferior performance whilst
increase in the voltage required to release electrons compared similar poor dielectric-to-anode contacts showed little adverse
to the uncoated case. He found that electrons were now being effects. Soft metal films, such as aluminium, lead or indium,
released from the junction between the end edge of the coating produced good dielectric-cathode joints. Shannon et al however
and the insulator, probably due to voids now existing at that found it preferable to joint the insulator to the electrodes with
position. either a thermoplastic or a thermosetting dielectric.
Kofoid then studied various electrode shapes5 to arrive at
the optimum shape for reducing the field at the insulator-
0 ’
cathode junction. He found the optimum shape was the com-
I
bination of two approaches. The first was to sit the insulator O I
in a slot in the cathode. The field at the bottom of a slot in an
electrode is small compared to the field at the electrode surface. SO-
If a piece of insulation, narrower than the slot, is inserted into
the slot the field in any voids between the end of the insulation
and the bottom of the slot can be shown to be small. The second TO-
0
approach was to reduce the field at the junction by rounding
off any sharp edges at the top of the slot. To demonstrate the
improvements made possible by these approaches Kofoid 60- co 0 0

tested a Pyrex cylinder with the configuration shown in Figure

h 0
8a.

I 0
0
I
I
03
I
I

Breakdown voltage Breakdown voltage

23 kV 96 kV

(a) (b)

Soda lime-gloss Pyrex (glass

Figure 8. Effect of cathode-insulator junction on the breakdown
strength of insulators. (After Kofoid5).

With the glass cylinder simply resting in a butt joint on the

electrodes direct voltage breakdown occurred at 25 kV. With a
groove cut in the electrode surface, breakdown took place at
28 kV with the grooved electrode as anode but at 70 kV with it Pointed UWolllte Pointed Unpamted Parholly Fully
as cathode. Increasing the radius on the electrode at either side ends ends ends ends painted pointed
ends ends
of the slot from 0.08 cm to 0.65 cm increased the breakdown
voltage to 90 kV with the grooved plate as cathode. Figure 9. Variation in breakdown voltage with insulator-electrode
Another method of minimising junction phenomena, if the junction (After Coenraads et all*).

388
R Hawley: Solid insulators in vacuum : A review

d
8
9
+ ISO-

,ar
B
+ + 0 100 -
f + t 5

50-

t + + + + O-
LLI A2 A3 A4 A5 81
L

Figure 10. Geometries investigated by Shannon et alO. Figure 11. Effect of insulator geometry on breakdown voltage (After
Shannon et ala).
They pointed out the field intensity at the critical insulator-
cathode junction can be reduced in several ways. Some of those C4. Thus those designs that provide a barrier to surface dis-
are illustrated in Figure 10. Group A includes insulators shaped charges and are of a configuration that reduces the field at the
so that single or multiple steps are placed near the cathode to insulator-cathode junction will improve the strength tre-
collect electrons emitted from the cathode. The resultant space mendously. These designs, with the exception of Bl, also had
charge on the insulator surface should reduce the field at the a high initial breakdown voltage. Shannon et al also found that
cathode and hence the emission. Group B includes several the presence of a charge-trapping corrugation near the cathode
insulators shaped so as to preserve the beneficial effect of the had a profound effect on the magnitude of the currents flowing
emission-repressing corrugation near the cathode but with the across the insulator surface and also through the vacuum gap.
excess insulator material removed away from the cathode to For example the currents for the designs A2 and A5 were an
minimize the field-increasing effect due to the high dielectric order of magnitude lower than those for design Al. On a basis
constant of the insulator material. Group C consists of geo- of these investigations it would appear that by appropriate
metries in which the cathode is shaped to place the insulator- design and suitable conditioning insulators can be capable of
cathode junction in a low field position. Cl and CS however sustaining voltage gradients of 80 kV/cm. Even without con-
were not tested because these methods either increase the inter- ditioning gradients of 50 kV/cm can be satisfactorily insulated.
electrode gradient, due to the reduced distance between the
electrode surfaces, or require electrodes of considerable thick- Surface coatings
ness. C2 lowers the field at the junction by intensifying the field Recent investigations, notably those of Jedynakas, have shown
within the bulk of the insulator. In this case the cathode insert that significant improvements in breakdown strength can be
was formed by grinding a circular groove in the insulator, obtained by covering the cathode with an insulating film. Since
metallizing it and connecting the coating to the cathode. This however this is then breakdown through the insulator rather
method exploits the inherent high volume puncture strength of than surface flashover, this work will not be considered in the
the insulator glass for example having a puncture strength in present paper.
excess of 500 kV/cm avoids increased gradients in the gap and
allows the use of these electrodes. I-Iypotheses to explain the breakdown mechanism
Figure 11 shows the results Shannon et al8 obtained for the The experimental evidence so far reviewed indicates the im-
various insulator designs for two types of glass. For any given portant role played by electrons, released at the insulator-
insulator the lowest horizontal bar indicates the value of the cathode junction, in initiating a breakdown across a solid
first breakdown voltage. The second line is the breakdown value insulator in a vacuum gap. For example Kofoid4’” showed that
after ten conditioning sparks. The top bar indicates the highest electrons and negative ions were produced at the insulator-
insulating voltage that could be obtained. No results were cathode junction and he postulated that when these negative
quoted for designs B4, B5 and C5. particles impacted upon the solid insulator or the anode surface
Designs A4 and A5 were inferior to the simple butt contact they release x-rays which in turn release further electrons from
cylinder design of Al. This was thought to be due to the action the cathode and the solid dielectric thus leading to a cumulative
of the single space-charge step being insufficient to overcome breakdown process. That the release of electrons initiated the
the increase of field at the cathode proportional to the value of breakdown was confirmed by Kofoid when he designed the
the dielectric constant of the insulator. All the other designs usual butting insulator-cathode junction to inhibit the release of
were superior to the single insulator so that in general any design electrons until much higher fields were reached and found that
that attempts to reduce the gradient of the insulator-cathode the breakdown strength was also much increased.
junction should improve the performance of the insulator, Fryzman et al6 have postulated the following hypothesis for
A full-conditioned performance in excess of 180 kV across the breakdown mechanism. Free electrons emitted at the
the 25 mm gap was obtained with designs A3, Bl, B3, C3 and cathode-insulator junction bombard the anode, or insulator
R Hawley: Solid insulators in vacuum : A review

near the anode, and release secondary electrons. Thus the sur- ensuring low stress conditions between the cathode surface at
face of the insulator near the anode becomes positively charged the insulator, by the use of shields at this junction and by a
to a potential approaching that of the anode. The charged area proper choice of the geometrical shape and electrical properties
gradually moves closer to the cathode so that the stress at the of the insulator.
cathode eventually increases to a value sufficient to cause the
release of electrons and hence breakdown. Once the arc is
References
formed the insulator surface discharges, the field strength
1 P H Gleichauf, J Appf Phys 22, 1951, 535. Electrical breakdown over
decreases, the arc is extinguished and the cycle is repeated. insulators in high vacuum.
After many breakdowns the insulator becomes covered with a 2 P H Gleichauf, J Appf Phys 22, 1951, 766. Electrical breakdown over
insulators in high vacuum.
layer of evaporated metal, and leakage currents prevent the 3 E S Borovik and B P Batrakov. Soviet Phvsics-Technical Phvsics 3.
accumulation of charge on the insulator surface. Fryzman et al 1958, 1811. Investigation of breakhow” in vacuum.
found that by screening that part of the insulator surface near 4 M J Kofoid, Trans AIEE (LISA) Part III, 79, 1960, 991. Phenomena
at metal-dielectric junctions of high-voltage insulators in vacuum and
the cathode or covering this part with a semi-conducting layer magnetic field.
the breakdown voltage was raised by a factor of approximately 5 M J Kofoid, Tram AtEE (USA) Part I[[, 79. 1960, 999. Effect of
metal-dielectric junction phenomena on high-voltage breakdown over
2.5. These workers suggested that insulators having a small insulators in vacuum.
secondary emission rate and a low surface resistivity should also 6 A Fryszman, T Strzyz and M Wasinski, Bull Polonaise, 8, 1960, 379.
improve the breakdown voltage. Hamisch et al” who measured On a mechanism of breakdown in high vacuum.
’ H Hamisch, Fifth Internat Congressfor Electron Microscopy D-13 (1962)
the surface charge on insulators, also postulated a breakdown Academic Press Inc. Surface discharges across insulators in vacuum.
mechanism relying on surface charging of the insulator due to s H Boersch. H Hamisch and W Ehrlich, Z angew Phys, 15, 1963,
the release of secondary electrons. 518. Surface discharges across insulators in a vacuum.
9 J Shannon, S Philip and J G Trump, J Vnc Sci Tech, 2, 1965, 234.
Watsonzz on the other hand has proposed that electrons are Insulation of hieh voltage across solid insulators in vacuum.
emitted from the dielectric surface by thermionic emission at a I0 I I kalyatskii-and G M Kassirov, Soviet Physics-Technical Physics 9,
1965, 1137. Investigation of impulse flashover of some solid dielectrics
rate depending on the square of the electric field strength. Thus in vacuum.
the surface becomes positively charged to a degree sufficient to I1 K D Srivastava, Rutherford Lab Rept RH ELIR124, HMSO London,
draw any other electrons into it, where they will multiply by 1966. Support insulators for high voltage apparatus in vacuum.
lE A Vierstra, Lincoln Laboratory Semiannual Technical Summary Report
secondary emission to a breakdown condition. Hence whilst the “High-Power Tube Programme” ASTIA AD 270734, VIII, 1961, 47.
breakdown is produced by secondary emission the initial Voltage breakdown.
charging of the surface is governed by thermionic emission of I3 R C Finke, Proc Second International Symposium on Insulation of High
Voltages in Vacuum, MIT, 1966, 217. A study of parameters affecting
hot electrons from within the dielectric. the maximum voltages capabilities of shielded negative dielectric junction
vacuum insulators. _
I1 C Coenraads, A S Denholm, .I Lavelle and F McCoy, Space Power
Practical insulator and shields Systems Conference, San Monica, American Rocket Society Preprint
Since the breakdown strength does not increase in a linear 2555-62, 1962. Electrostatic power generators for space.
manner with insulator length, to obtain insulators of a reason- I5 D N Leader, Ph D Thesis, University of London, 1951. A study of
electrical breakdown in vacuum with particular reference to the develop-
able length that will support voltages in the megavolt range, it ment of an electron gun for projection TV.
is a well known technique to break the insulator surface up I6 C A Ramm, International Conf on Instrumentation for High Energy
Ph:, Tics, Berkeley, 1960. Some features of beam handling equipment for
into a series of parts each supporting a controlled potential.
the CERN PS.
Thus advantage is taken of the higher stresses held by the I7 I D Smith, Proc International Symposium on Insulation of High Voltages
shorter insulator lengths. Such an insulator capable of standing iw Vacuum, MIT, 1964, 261. Pulse breakdown of insulator surfaces in
a poor vacuum.
10” V across 30 cm has been described by Britton et alz4. I8 G B Walker and E L Lewis, Nature, 181, 1958, 38. Vacuum
Finke13 has shown that the use of a solid spherical shield breakdown in dielectric-loaded wave-guides.
around the cathode insulator junction improved the strength of 1!1 R Haves and G B Walker, Proc IEE (GE) 111, 1964, 600. Vacuum
breakdown at a glazed ceramic surface.
an unshielded epoxy feed through by 520 per cent. The use of z” R F Thumwood. PhD Thesis. Universitv of London, 1965. The
negatively biased or earthed grids around the bushing end or production of X-ray’ impulses at hjgh voltage. ”
81 S P Bugaev and G A Mesyats, Soviet Physics-Technical Physics, 10,
the chamber walls has been found to improve the breakdown
1966, 930. Nanosecond time development of a pulsed discharge at a
voltage across a feed-through bushingz5*26. Such grids probabIy dielectric-vacuum interface.
suppress secondary particles leaving the chamber wall, hitting e2 A Watson J Appl Phys, 38, 1967, 2019. Pulsed flashover in vacuum.
58 L Jedynai, .I Appl Phys, 35, 1964, 1727. Vacuum insulation of high
the bushing and causing a breakdown. voltages utilizing dielectric coated electrodes.
24 R B Britton, K W Arnold and A S Denholm, Rev Sci Instrum, 34,
Conclusions 1963, 185. Ability of a voltage-graded surface to support a high
voltage in vacuum and in a pressurised gas.
For combined vacuum-solid insulation arrangements there is 25 K W Arnold, Proc Second International Symposium on Insulation of
definite evidence to indicate that the breakdown is initiated at high voltages in vacuum, MIT, 1966, 73. Vacuum breakdown phenomena
at 1 million volts: A postscript.
the junction between the cathode and the insulator. In practice a6 F Rohrbach, LeVide, 121, 1966, 52. High voltage electrostatic
significant improvements in voltage holding can be achieved by separators.

390