NPA/Int. 67-5 19.4.67
HIGH VOLTAGE BREAKDOWN IN VACUUM *)
by
C. Germain QTI.d F. Roll~bnch
1. Introduction
The phenomeno. tho.t occur when high voltages o.re applied between electrodes
in n vo.cuum ho.ve been studied for m1::.ny ye'lrs. However, the progress in this
field ho.s only recently been speeded up under the spur of the increo.sing number
of teclmical o.pplicc,tions of v=:cuum insul,'.',tion. Dur:Lng the course of the lo.st
deco.de there has been a considerable incre'.'..se in the nuirrber of people involved
and, of course, in the number of papers they ho.ve 9ublished ~nd conferences
they kcve attended. As a result of this increG.sed effort o. progressively clern'er
picture is emerging, ct le0st on some '.~sr,ects of the lJroblem, but there is still
much to do before we can. obt::cin a coherent, clec.r c~nd. complete explo.nation of
all the relevc.nt observed phenomenn.
Let us first ex::;,mine whd we me::m by electric.
- 2 -
Except in very short pulses a high voltage ca1:not be maintained between
electrodes together with a high current since we no longer have a vacuum between
electrodes. We can thus understand how vacuum breakdovm and gas discharge are
related and there is an intermediate region where curious phenomena occur as
we shall see it later.
Since we know that residual gas molecules are not involved in the init-
iation of vacuum breakdovm, what kind of limitation are we going to encounter
with the electrodes upon raising the voltage be-tween them ? At first sight we
might expect an appreciable current to appear when field values of 107 to 108 V/cm
are reached because electron field emission from the cathode metal becomes
significant at this field intensity according to Fowler-Nordheim theory. At
about the saoe field intensity the mechanical stress on the electrodes, which is
proportional to the sq_u:ire of the field intensity 1 bec.omes comparable to the
bulk tensile strength of the electrode material. A mecl:anical stress of about
5 x 108 pascals (i.e. rv 50 kg/mrn2) is generated by a field intensity of 108 v/cm.
Therefore at a field of about 107 - 108 v/cm there should be a catastrophic
breakdown between electrodes for these reasons. Is this the limitation encountered
in eq_uipment in which high voltages are applied in a vacuum ? .Anyone who has
spent hours or event days trying to bring up the performance of an electrostatic
separator or of an accelerating column to its nominal value will tell you the .. t
there is a "performance gap" of about three orders of magnitude !!
In order to lmd.erstand tho reasons for this enorwous discrepancy we must
look again more closely into conditions at the electrodes
1) The average or mo..croscopic· field, which is voltage/ gap length, is the
releV'ant par2.:-neter for particle so:r;an:tor performa.rice but for determ-
ining the current between the electrodes tho surface field is the
critical factor. This true microscwp:Lc field on the electron emitting
cathode points is much larger than tho average field because of the
enhancement produced by the point gGometry. The enhancement factor may
reach one or two orders of magnitude.
2) The microgeometry of the electrodes, that is the shape of the points,
.is continuously modified under the combined action of the electric
field and of the surface curvature gradient; this modification is
PS/5902
.. .
- 3 -
facilitated when the local temperature is raised. Therefore perfect
polishing of the surface will not permanently reduce the field
enhancement factor. Sputtering of the electrode surface also has to
be taken into account in some cases.
3) In large scale practical devices such as separators, where ultrahigh
vacuum technology and high ternperaturebaking are not used, the
surface of the electrodes is not ~u.re metal but is contaminated by
all kinds ofl organic materials which have a much smaller work function
than perfectly clean metals. k.purities and dislocations in the metal
also enha.i.~ce electron field emission.
A fairly satisfactory theory of the electrical breakdown of small gaps
( ~ 1 mra), bnsed on a detailed analysis of the electron emission from the cathode
and on the current interaction wit:'.1 either electrode, has been developed by
1 1,2) .. th' several authors for the case of very cles.11 c8tc::.l e ectrodes • however, is
theory cl.oes not fit, in its presEmt fom., the experimental data accumulated on
the electrical breclcdown of large g:::.ps. Otl:er )henomena may become prominent in
large gaps and cause the onset of rree...~dmm at lower voltage values than expected
from this theory. We only mention briefly t~1e clunp hypothesis which we shall
discuss in more detail later. No satisfactory theory has yet been produced for
larga gaIJ brer.:kdoim .filld a completely new line of attack may perhaps be necessary
in order to develop a successful theory.
We shall start our review with 2 brief look at the theory of emission of
electrons from metals; then we shall c-:mside1· the energy exchanges between the
electrons and electrodes and discuss tbe anode versus ca.thode control of breakd01m.
We shall shoi-1 that this provides a rather satisfactory explanation of experimental
data for electrical brea_l\:down in s:.:ie.11 gs.:;is or relatively low voltages (i.e.
millimetre gap or hundred kilovolt regioE) whereas other hypotheses have to be
considered for larger gaps and voltages.
Since th8re is clearly not enough ti:ne to mention all the important
contributions in the literature, we have selected papers which are typical of the
main trends of the present state of knowledge. We also have to apologize for so·ne
over-simplifications which are inevitable for the sake of a clearer presentation.
PS/59J2
- 4 -
·2. Emission of Electrons from i'· etals
The theories cif thermionic and field erc.ission of electrons frora r'.letals
were worked out about half a century ago 3,4, 5,6 ), but lL'1til recently thsy were
considered as sepc::.rate theories, each one h'lving its own range of temperature
and field within 11hich the corrosp0nding ex-;;ression of the current i.ras valid.
However, I4urphy and Good ?) in 1956, gave a unified treat=-,ent and obtained a
general expression fo:r the cur:.r:ent as a function of electric field F, temperature T
and work-fuI1ction ~' in tl:s fora1 of a definite integral. Then they applied
approxirnating tecl:1niqi:ss for t:·,is integral e..'1d set up hro distinct formulae fOl'
thermionic and field cmfas:.on, 1d th a characteristic dependence on F, T and yl for each type of emission. Jm e.pprox:'..::,ation for :i..ow fields 2nd high temperatu:res
leads to an extension of the Richardson-Schottky formula for field assisted
thermionic efilssion. The 71.l:i_cl:'_ ty of the a~)proxlic.a tion determines the range of
temporo.ture ar;.d field within which it apylies. A si:r:il2r trGatnent of the
integr&l for high fields and low te:J.:_::eratures, gives an exter:si.on of tl1e
Fowler-Nordhei::i forrm~:ta for field emission, wi t~1 the corresponding r2nge of
temperature and field in which the o.pproxL1ation is valid. With e..nother approx-
imation they evaluated the integral in an :L'1ten:ediate region between the::'mionic
and field emission. Outside th0se three regions the emitted current is obtained
by numerical integration of the general e:z:pression for each set of values of
the paraneters.
Murphy and Good der~.ved their general ex9ression for the emitted current
from the well established. r::o:iel of e.. rr·etEcl : the Fermi-Dirac distribution for a
free electron gas in the m~tal and the clr~ssical iincEe force barrier at the
surface. This model provides c~n expression for E(T,
- 5 -
Without going through the mathematical details of this interesting paper,
we shall just examine the general expression for the emitted current density
3 o 5 I ( ) in figure 2. In the expression (2) y = (e F) • I WI and v y is a tabulated function. Withj_n the regions of validity of the approximations we have mentioned,
simpler expressions can be derived as shown in figure 2, expression (3) for
thermionic emission and expression (4) for field emission. From these expressions the well-known Richardson-Schottky and Fow1er-Nordheim formulae, written with
the system of units and the symbols used by Murphy and Good, are respectively
obtained within still more restrained regions of validity.
The boundaries for the three types of emission discussed by Murphy and
Good are represented in figure 3 for a work function of 4.5 eV, a representative
value for tungsten. At current densities of about 108 A/cm2, which are typically
required for excessive heating of tungsten protrusions the emission of electrons
is mostly due to tunneling Ui..rough the barrier, i.e. the emission increnses
rapidly 1,1ith the field and is relatively insensitive to temperature 8 ), .as cnn
be seen in figure 4.
3. Exchon&:es of Energy between Electrons Emd Electrodes --· ----·------------------------
· .l..J._~ons on the co.thode surfac~
Upon raising the voltage V between plane parallel electrodes a current is
emitted at very localized points on the c:1thode surface. The value F of the field
that determines the intensity I of the electron boom is not the average value of
the field in the gap d, F = V/d, but F = yF at the apex of the emitting point. 0 0
The unhancement foctor y is determined by the shape of the point and by its
height h expressed with respect to d. Even when there are several emitting points, 11) the factor y is determined, to a first approximation, by the sharpest one •
If the emitted current follows the :B'owler-Nordheim formula, then the expression
log (r/v2-) is a linear function of l/V ~ which the factor y enters in a simple way.. In fact the current measured as a function of V, in any experiment not
interrupted by a spark which would destroy the emitting point, appears generally
to follow· the Fowler-Nordheim formula quite well so that a stra:Lght line is obtained
in the F-JIT graph, that is when log (1/v2) is reprnsented versus l/V as in figure ~ From the slope of this straight line m =A ¢1. 5 d/y, where A is a known constant,
nn experimental determination of the value y of the emitting point is easily
obtained, provided yJ is known.
- 6 -
By making some assumptions on the geometrical shape of the protrusion,
for example by considering an ellipsoid or a cylindrical rod with a hemispher-
ical cap 81 9), the corresponding y values can be calculated from electrostatic
principles. Experimental values of y, deduced from F-N graphs, are currently
observed in the range 1 (for extremely small gaps) to a few hundreds (for
millimetre gaps) 9,lO)• When comparing the theoretical values with the exper-
imental values obtained as a function of the gap d the geometrical parameters
of protrusions are obtained : values of about 1 µm and 0.1 µm, within an order
of magnitude each, are generally obtained for the height and the diameter of the
protrusions respectively. These dimensions pertain to protrusions of simple
geometrical shapes but direct geometrical measurements by using electron.
microscopes have.confirmed that the dimensions thus estimated are of the right . 11 12)
order of magnitude ' •
When the emitted cu~rent density exceeds 107 P/cm2 electron space-charge effects reduce the true field F at the emitting surface appreciably below the
value calculated from electrostatics. The field emission formula is still Yalic.
provided the true value of the field is used. The space-charge effects at
current densities of about 108 A/ cm2 , which are often attained in practic«~, strongly reduce the influence of variations in the work function on the emitt'
surface 8 ).
In general the presence of protrusions on the surf ace cannot be avoided
and is primarily responsible for the limited voltage strength of vacuum gaps.
Even perfect polishing of the surface would not be of great help since the
surface condition can deteriorate quite quickly with time under the action.of
temperature, applied field and ion bombardment. It has also been shown by
:Little and Smith 13) that projections ~re removed from the anode. by the electric field and these produce craters and sharp field-enhancing protrusions on impact
with the cathode; The growth of the protrusions has been studied by Charbonnier 14) .. .
et al. , using Herring's theory, whose fundamental postulate is that the flu."?'.
of atoms b·eing transported past a certain point on the surface of a protrusion
is proportional to the product of a diffusion coefficient and the local gradient
of chemical potential at that point. ~lhen the average field exceeds a certain
value the protrusion will grow indefinitely and produce a breakdown. They have
also studied the formation and growth of protrusions by the action of sputterin;
PS/5902
-7-
J.L2 E}ectron-cathode energy exchanges
The field-emission induced heating of cathode protrusions has been studied 15) . t 1 8,1~) R . t· theoretically in detail by Lee and by Charbonn1er e a • • esis ive
heating alone cannot explain the stability of field emission at current densities
high enough to raise the tip temperature to 2000° K or higher : a balance of
Jouie'heating with thermal conduction and radiation losses is impossible, leading
to an unstable situation. Furthermore with a low work function coating on the
tungsten tip ?ne can draw a higher current density while maintaining a lower tip
temperature which is impossible with heating by Joule effect alone. For these
reasons the Nottingham effect clearly comes into account in the thermal balance
of the _protrusion tip.
The Nottingham effect is due to the difference between the average
energy < E ) of the emitted electrons and that < E 1 > of the replacement electrons which is assumed to be the Fermi energy in the simple Sommerfeld free
electron model of metals. At room temperature < E > is lower than the Fermi energy so that the cathode is heated by the Nottingham effect. As the temperature
is raised the energy distribution of the··~mitted electrons is modified and
becomes symmetrical with respect to the Fermi level at the transition temperatu::.-e,
At higher temperatures the Nottinghrun effect changes its sign as < E ) becomes higher than < E' > and cooling of the cathode results from the absorption of energy by the electron beam. At these temperatures the Joule and Nottingham
effects have opposite actions on the tip temperature. The transition temperatur-;
is Ti= 5.35 x 10-5 Fjr;{J• 5, that is Ti.= 1250 °K at F = 5 x 107 V/cm for tungsten (
- 8 -
1) T'11e pulse is longer than a thermal time constant \. Then the
steady-state distribution of the temperature is reached for the
protrusion standing on the cathode surface.
2) The pulse is very short : shorter than a smaller time constant t 2 •
This is the adiabatic case where no appreciable heat diffusion can
take place and the local temperature is governed by the local heat
generation.
3) The pulse duration is intermediate between t 1 and t 2 •
8,16) The values of t 1 and t 2 are given by the theory and for pulses
longer than t 1 or shorter than t 2 simple expressions for the critieal currents
as functions of the relevant parameters nre obtained. From these the corresponding
critical voltage may be calculated.
4. Breakdown across Small Gaps
A.J Cathode initiation of breakdow11
Field emission at high current density does not necessarily lead to
breakdoi:m as pointed out by Char bonnier 8 ), since de emission densities in
excess of 107 A/cm2 or pulsed emission densities in excess of 108 A/cm2 have
been maintained without instability for operating periods of several thousand
hours. However, excessive current density causes excessive heating of the tip
and evaporation of cathode material. This metal vapour has a high ·probability
of being ionized in the very high density, low-energy electron beam just outside
the emitting surface. Production of positive ions in front of the tip reduces
the electron space-charge effect and therefore increases the field erni tted
current. Furthermore the cathode tip is bombarded by these positive ions which
can liberate more electrons so that these t1vo regenerative processes lead to an
extremely rapid development of a full vacuum arc in times of the order of
10 nanoseconds. The Nottingham effect does not markedly affect these general
conclusions.
Experimental studies have shoi-m the onset of rapid instability when the
protrusion tip reaches a critical temperature corre8ponding to a metal vapour
pressure of 10-4 Torr. The critical current density jc does not differ widely
PS/5902
- 9 -
from one material to another and is about 108 A/cm2 for a typical conical
. t• 16) pro JGC J.on •
The rate of increase of the breakdovm voltage, as pulse length t . 0
becomes smaller, is quite low since, in the adiabatic.limit, the permissible
current density jM is proportional to
·08 AA/ 2 . . t 1 v3 1 /6
cm varies approxirna e y as
t- 1 6 • Therefore the voltage increase 0
duration is reduced by a factor ton •
t- 0.5 • Now since j in the region of 0
the breakdown voltage increases as
is only 40 to 50 o/o when the pulse
.4._.J_.lin.ode versus cathode control of bL,e_a..::l\:dow.n.
We have so far discussed the emission-induced heating of the cathode
protrusions. However, there is an additionnl l1eating effect : the field emitted
electrons, accelerated by the field across the gap, cause anode heating on
impact with the anode. Breakdo1m is likely to occur when the temper2ture at
either electrode oecomes high enough loce.lly to cause significant eva~oration
of electrode mderial into the path of the electron beam. Whether this critical
condition first occurs at the cathode or at the anode depends on the gap
geometry, on the field enh::illcement factor y, on the relative electrode materials 8)
and on whether the gap voltage is de or pulsed, as discussed by Charbon..nier
Chatterton 17 ) has pointed out that in the case of de voltages and
parallel plane electrodes it is virtually certain that excessive heating will
occur at the anode well before the electron current density reaches a level
sufficient for excessive ?r9trusion self-heating at the cathode.
From the radial spreading of the electron beam emitted from a protrusion
in the case of parallel plane electrodes the beD1n power density at the anode
may be calculated as a function of the current density end of the corresponding . . . 8 10 17 18) .
fie1d on the protrusion ' ' ' j the values of these two quantities should
be smaller than the values obtained from the conditions of cathode-controlled
breakdovm since 'de are now interested in anode-controlled breakdown. If the
values corresponding to cathode-controlled brea1..cdown are used in the expression
of the anode power density the maximum possible vive heating of the anode.
PS/5902
- 10 -
We may define two thermal time constant values t 1 and t 2, corresponding
respectively to the steady-state and adiabatic conditions of anode temperature
distribution, in order to assess the criticnl or maximum admissible anode power
density. The value of the voltage pulse duration t 0 relative to t 1 and t 2
determines which expression for critic al anode power density hclS to be Considered
· and then the corresponding criterion of anode-controlled breakdown is obtained
simply by stating that the m'UCimum possible VD.lue of the anode power density
exceeds the critical value 8119 ). This criterion turns out to be a 2imple
expression for the maximum value y," of the cathode field enhancement factor y. Fl
If no protrusion on the cathode has a y value _larger than Yr.r then the breakdmm
is induced by excessive heating of the anode. When the voltage pulse duration t 0
decreases the relevant value of y11 also becomes smaller : the breakdown which is
generally Qnode-controlled in de conditions is fina1ly cathode-controlled at
very short pulse durations.
10) Utsumi and Dalman have. studied theoretically and experimentally what
are the respective conditions for o.node versus cathode·control of de breakdown
between plane parnllel electrodes as a function of their separation. The type
of breakdown which is likely to occur can be determined from the dimensions
of the projection and the thermal andelectrical properties of the electrode
materials. Af3 the separation d is increased the clomin-::nt mode of breakdo11m is
first the anode type when the value of the enhancement factor y is small, then
the cathode type with which is associated a critic2l cu.rrer:.t of constant value,
E:nd finally, the anode type again when d reaches the 0.1 - 1 rnm region and the -·o 5 corresponding critical current varies as d • •
The nnalysis of vacuum breakdovm, which we hove just presented in
simplified form, is sufficient to provide estimates of the breakdown voltage
and indication of the necessary electrode conditions for initiahon, which are
0 f . d . t "1 lO, 19) ~ t 1 «. t f 1 c n irme experimen
- 11 -
we have just presented, which is.based solely on the heating power of the field
emission-induced electron beam. The large gap studies with plane parallel
electrodes have revealed several facts which cannot be explained by this theory
at the present time ; these are :
1) 'rhe breakdown field FOB is a decreasing function of the gap d. The
experDrrental data can be fitted to the following eCTpirical expression
- ex FOB = A d with O. 3 < ex ( 0.8
2) The steady current value just below the microdischarge threshold is
strongly dependent upon the gap d in the range 1 - 6 cm: Whereas
in a 0.5 cm gap between stainless steel electrodes in ultrahigh vacuum
conditions hundreds of microamps c~n be drawn without causing
breakdown, it is impossible to obtain more tkm a few nanoamps when
the gap is increased to 5 cm 24).
3) The breakdmm field increases with pressure when gas is admitted to
the vacuum chrunber up to the region where Paschen's gas discharge law
becomes applicable 1 which is usually in the milli tor range 27 ) • This
effect, which was first demonstr~ted clearly for separators at CERL~
in 1959, had been noted in the early fifties by a few authors 29).
T'ne effect of the gas pressure on the microdischarge threshold and on
the breakdown fields is stronger the larger the gap, (figure 6).
~l.hen discussing the cathode initiation of breakdo1m we have established
the existence, in this CB.se, of a critical value of the cathode emitted current
which is independent of the gap d. Since the value of the enhancement factor y
is independent of d for large gC>.ps, the as.sociatGd critical average field FOB
sho:;.ld .lso be independent of d,
The first and second points just m::ntion'-'d, ns well ns our
- 12 -
If we now consider the possibility of breakdown controlled by the
vaporization of a small amount of anode material under the action of the field-
emi tted electron beam impinging· on the anode, we have to remember, according.
to Utsumi and Dalman lO)' that in this case the critical current varies as
d- 0 •5. This condition is obtained by ir:lposing the condition that the temperature
of the heated anode spot centre be kept constant, at the melting point for
instance, when the separation between electrodes is increased. The experimental
results mentioned in noint (2) show th2t the microdischarge threshold current . . . varies by several orders of magnitude instecd of half 2.Il order when the gap is
increo.sed from0.5 to 5 cm, which rules out also this theory for rnicrodischnrges
initiation. This conclusion is valid also for bre&1\:down initiation, since
microdischarges develop into full discharges, or brea.1\:down, at a higher voltage
and can therefore be considered as being governed by the same law .
.2_._~ The micropc.rticle brea .... ~down ini ti2tion hypothesis
In 1952 Cr2nberg 28 ) put forwa:rd ~he hypothesis that breakdown is initiated by tiny bits of material or clumps, impinging on the electrodes under the effect
of the electric field. The brdakdown, in this clump theory, is the consequence
of the electrical gas discharge taking place in the v~pour generated when a high-
speed microp~rticle collides with one of the electrodes. Assuming that the
micro-particle gets detached from one electrode, it carries along an electrical
charge q, which is proportio~al to the field, and it has an energy W = q.V after crossing the gap when it impinges on the other electrode. A criterion for breakdown
is obtained by simply stating that breakdown will ta..~e place when this energy
is larger than a critical value W · . t. cri . ,
The breakdown field-gap relationship derived from this simple clump
theory, in the case of plane parallel electrodes, is : F0B2 x d =constant.
When comparing this formula with the exµ;rimental results mentioned in point (1).
above no discrepancy appears but the range of ex?erir:lentally possible values for
a is too broad .for this to be a strong confirmation of the clump theory.
Furthermore by refining the clump theory, the theoretical value obtained for a
can be IDQde to differ slightly from 0.5, so that some direct experimental confirm-
ation not only of the existence of clumps but also of their triggering action
in the breakdovm process is necessary. Nany attempts are made in that direction
PS/5902
- 13 -
and we only mention briefly a method.which was first proposed by C. Germain at
CERN for detecting clumps during their flight ncross the go.p in order to
establish directly their triggering action in the breakdown process. This method
is based on the detection of the light reflected by the clump when it crosses
at right angles a wide laser beam. in the gap between the electrodes.
·The microdischarges we ha.ve mentioned in the introduction are observed
at voltages lower than the breakdown voltage. 'l'hey are essentially self-
extinguishing_ bursts of current, that is they do not involve 2 complete colls.pse
of the voltage across the gap. On the basis of the clump hypothesis. a micro-
discharge 211d a full discharge can be caused by the same process, that is by a
more or less complete vaporization of the micro1xi.rticle when it impinges on
the opposite electrode. In the case of a microdischarge one can assUL~e that the
amount of material vaporized do not generate a bubble of va~our large enough to
induce a diverging gas discharge process.
In the clump theory the microdischarge and the full discharge are consiO.ered
as being unrelated to the steady current th0-t exists simultc.neously between
electrodes since no proposed clump extraction or generction process under the ::.ction
of this steady current can account for the very large variation of the micro-
dischnrge threshold current as a function of the gap in the centimetre range.
The clump can rather be considered as detached ~y the action of the electrostatic
pressure from the electrode to which it is loosely bound. The exact process of
this extraction is not yet known o.nd this is the first of a series of questions
which must be asked about the theory of brealcdovm across le.rge gaps. Little l '1)
and Smith _, have reproduced natural bre~ikdmm phenomena in a 0.25 rxn gap by
intentionally inserting nickel particles about 0.01 mm in size into a small hole
on the anode surfnce Ctncl they evaluated tho cr:i. tical clump energy to be a few
nanojoules. They demonstn1ted that in no.tu.ral breakdown clwnps had been extracted
from anode pi ts located at impurity centres near grain bou..'1daries Ei.nd had
produced sharp-field-enhancing protrusions on the cnthode surface. No sueh exper-
imental work has yet been perfo'rmed for larg"' gaps and high voltages.
A regenerative process is required for the creation of clumps, otherwise
it would be possible to condition electrodes once o.nd for all at a given voltage
provided the working voltage is kept lower than this value in a constant gap.
PS/59J2
14 -
The growth of protrusions under the action of the electrostatic pressure is a
possible regenerative process which is well known experimentally 12 ) but it
is only valid for small gaps since it is associated with a given field or current
level. Any proposed clump regeneration proces~, based on the effect of either
the current or the field, must account for the respective variations of the
critical values with the gap length.
The marked influence of the residual gas pressure on breakdown voltage . . 22 26)
mentioned ear~ier, which is of :i.mportance·in some practical applications ' ,
is also difficult to explain completely on the basis of the clump theory because
of- the g2.p dependence of both the microdischarge threshold current and the
breakdown field. The influence of the pressure on the breakdown voltage in small
gaps has been investigated recently 11129) and it can be explained by the
selective ion bombardment of the sharpest points of the cathode : the corresponding
values of the field enhanc8ment factor y are thus reduced by sputtering so that
a higher v2lue of the average field FOB can be sustained. Such.an expl:ination
was also attempted for large gaps but the mechanism of this.influence is not . 21)
yet completely understood
6. Conclusion
The initiation and the development of electrical discharges in vacuum
across small gaps hetween clean metal electrodes can be described satisfactorily
in terms of the gap geometry, cathode surface microgEometry and the physical
properties of the electrode material. The relevant theory is based on a detailed
analysis of· the electron field emissiorifrom the cathode and on the associated
current thermal interaction.with e:.ther. electrode. However this theory fails to
fit the experimental data obtained with large gaps: in this case the hypothesis
of breakdown initiation by microparticles which are accelerated across the gap
by the electric field, provides an explana.tion which has received some experimental
confirmation, but is not fully sntisfactory. Further investigations, both
theoretical and experimental, are required in order to develop a completely
quantitative theory of br0akdown across large gaps.
It is a pleasure to acknowledge the assistance of Dr. Colyn G. Norgan,
University College of Swansea, for revising this text.
/fv PS/5902
C. Germain F. Rohrbach
- 15 -
References
1) Proceedings of the first Internation:cl Symposium on the Ins_ulation of High Voltages in Vacuum, October 19 - 21, 1964, M.I.T. Boston, referred to as 11Proc. Itt in other referencas.
2) Proceedings of the second International Sympqsium on the _Insulation of High Voltages in Vacuum, September 7 - 8, 1966, M.I.T. Boston, referred to as 11Proc. IItt in other references.
3) O.W. Richardson.- The Emission of Electricity from Hot Bodies, (Long:nans Green o.nd Company, London, 1921).
4) W. Schottky - Physik: z. 15, 872 (1914) !
5) L.W. Nordheim - Proc. Roy. Soc. -(London) A 121, 626 (192$).
6) R.H. F'owler and L.\I. Nordheim - Proc. Roy. Soc. (London) A 119, 173 (1928).
7) E.L. Nurphy and R.E. Good - Phys. Rev. 102, 1464 (1956).
8) F.M. Charbonnier - Proc. I, 23 and 477.
9) D. Alpert, D.A. LBe, E.M. Lyman and H.E. Tomaschke - Proc. I, 1.
10) T. Utsumi and G.C. Balraan - Proc. II, 151.
11) H.E. Tomaschke, D. Alpert, D.A. L€e 3nd E.iI. Lyma.Yl, Proc. I, 13
12) R.P. Little and S.T. ~nith - Proc. I, 171.
13) R.P. Little 2nd S.T. Smith - Proc. II, 41.
14) F.M. Charbonnier, E.E. Hartin, R.W. Strayer l'cnd C.J. Bennette - Proc. I 169 and 507.
15) D.A. Lee - Proc. I, 69 and 453.
16) F.N. Charbonnier, L.W. Swanson and C.J. Bennette, Proc. II, 11.
17) P.A. Chatterton - Proc. I, 25.
18) F.E. Vibrans - Lincoln Lab. Technical Report No 353, 8 Ntiy 1964,
19) C.J. Bennette, L.W. Swanson and F.r1;, Charbonnier - P-..roc. II, 21.
20) C. Germ,-:cin et F. Rohrbach - IVeme Conf6rence Internationale sur les Phenomenes d'Ionisation dans le Vide - Paris 1963 - Section Va-To:::e II.
21) F. Rohrbach - Proc. I, 393.
PS/5902
- 16 -
22) J .J. Murray - UCRL report 9506 (1960).
23) K.W. Arnold - Proc. II, 73.
24) Ji,. Hohrbach - Proc. II,· 83.
25) W.A. Smith, T.R. Mason - P1·oc. II, 97.
26) C. Germain, L. Jeanneret, F. Rohrbach, D.J. Simon et R. Tinguely -Proc. II, 279.
27) C.M. Cooke - Proc. II, 181.
28) L. Cranberg - J. Appl. Phys. 23, ·51s·- 1952.
29) E.N. Lyman, D.A. Lee, H.E. Tomaschke arid D. Alpert, Proc. II, 33.
PS/5902
· ..
Energy
! I
0 x ~~~~--~~~~~~~~~~~~~~~~~~~~~-·~
~ I
'I max. f /'"-.............. , ·1/ ............ ,
' ' -!% ,~, I. ~ ~ ~ '( - -~,,,,~":'::_-
. ' - '" I '- > ·-- "'~~. i '°' ,, -~ -..... ~ t:tl="ormi lev·... j---- .,,..,.~% t''-l --.. • '• ___,_ -- ---- - --·i .,,,,, . --. ' -~--=-=--- -·{ '
-~- -~=-;~~J ~.~ -- ----------- ----- I "' - - "~ --~-~-__ ' ~ - _, - I --------
q)
! '.I +-v~a ''1
~
V (x) = Pote1nticll ener~IY of an electron near the rneto.l surface
X = Coordinate norrnal to the surface ~ = Ferr·ni en,:rgy
(b e
rJ .. ~ ~; • ~\ 2 "i n
\!;._fa - ~;J \! \.< Er.~.::r -'1·•. '"" _,,""\''""'>-.a. t• -- 10 c·~· ron ut:l..t~\·~' cu~ ::~:nan ~·t11.;.; t l
F
l"'l-oi -:.r~·,t • ~ h .. ah. •
r- . ··-~ nr:::rli"'~!~I':" ':!~ &~:.;,II'"''"'\,.;.~~ _,. •ff'~""' ,,,, ~'4..i3 u
.,t l
i:" ~.,, '{ .o. rr ,,..,,, r~· I i "/ f-,f'\ "''- l ~ i ~---""~ ll.j
1 \.! ~~ ( • 3 t- v:--· "'"f'\{"-(«J' -· ....., c::i. f• ~ 2 -~ i ~~"·· • ,._ J -
-··o)" p•ct,., !.,) l'"" I' ,.,. VII rt'-¥ L
f"""~\.li!,
1
in~.id110. thP mpta1l ~1o1 s"-~-;1:!:""1:~·.,.,¥ i:J ~ ..... "" il: '4'!t ,
rn ~ll'"""'""'O I' t~~,,k> ~ ~~~ t
~ t . !I'• if~! •CJ,(" rRf"' ~(2 • ..., 11.·..,, ~ ~ ...... ~:
Vrnax.
'l°' 0,, ,,
\!""'"""'. r~ i ~ . ~ """"'' ' l -1'.1; a "'"~ I!':·~~ - -- !i. i~, ...... n if_\'\ - -~Ii ' ··--·'"'""~~. I! ~ ~~ ''\>,.( ~,\lf'ilo,;) ~ \4~ ~M~ ~ iv~ ~','l
Th&rmionic e mi ssicn
Field emission
f \ ' J
Cil' rd l "'"
Current density j ( F, T, ~ ) as a function of electric field F, ternperature T and , Ferrni energy J 1n Hartree uni • according to r,l1urphy and Good 7> _________________ ,
3000
~ 0
2000 c
OJ I-::i --d l-{j) 1000 0. E ~
for a
• ion ( ~ ) i.en
'7)
log. 1 ... VL
-13
-14
--i~S
-16
~~,
- . -\'' -~ \-. -----,-1 ,ff\ (X)
I \/ I ~ ! \\, I . ·r \i~
I
Typi l ( D.
m::
[ n I' q' " \ " J
( kv) A Voltage V
500 -
300
200 -
100 _l
gap d ---t--------+--· ---+-----------+--·--, --~
2 3 4 5 (cm)
£lg~_§ __ Ty . VO fc t"' v-1 ... ;o· ·ue:-.,QI .,u,, a -ii current and t t