IEEE Trans. Electr.

Insul, Vol EI-13 No 3, June 1978 135

R E V I E W

BREAKDOWN MECHANISMS IN SULPHUR-HEXAFLUORIDE

N. H. Malik and A. H. Qureshi

Electrical Engineering Department
University of Windsor
Windsor, Ontario, Canada
N9B 3P4

ABSTRACT
The well known breakdown theories, Townsend's generation
mechanism and the streamer mechanism, are reviewed and
applied to the results of breakdown in the strongly elec-
tronegative gas, sulphur-hexafluoride. Experimental
results reported in the literature on the breakdown
behavior of sulphur-hexafluoride are examined in the
light of these theories. The breakdown theories are
used for the estimation of the breakdown voltages in
pure SF6. Other factors that may affect the breakdown
characteristics of SF6 have been discussed. Further
areas of work have been proposed in order to obtain a
better understanding of the breakdown mechanism.

1. INTRODUCTION 2. THE BREAKDOWN MECHANISMS IN SF6

It has been known for a long time that heavy gases be- A gas is normally an almost perfect insulator. How-
longing to the seventh group of the periodic table,
e.g., fluorine, chlorine, etc., have a considerably ever, some free electrons and ions are always present
higher dielectric strength compared to air under similar due to ionization. Under the influence of an applied
experimental conditions. The high breakdown strength electric field, free electrons may gain sufficient
depends mainly on their capability of taking up free energy to ionize a neutral gas molecule on collision.
electrons, thereby forming heavy negative ions. Gases The number of electrons produced in the path of a
having these properties are called electronegative. Of single electron travelling a distance of one centimeter
the many available electronegative gases, sulphur hexa- in the direction of applied electric field is known as
fluoride, SF6, has especially gained importance in the Townsend first ionization coefficient a. Several
recent years because of its chemical stability in addi- types of ionization processes in SF6 have been sug-
tion to its high breakdown strength. Therefore, it is gested [1]. One possibility is
being used increasingly in high pressure, high voltage
systems as insulation. Due to its effectiveness in the SF6 + e +SF5+ + F + 2e (1)
extinction of arcs it is used effectively in circuit In SF6 negative ions are formed when a low energy
breakers and switchgear with a pressure in the range of electron attaches itself to the neutral gas molecule
200 to 600 kPa.
resulting in either a direct attachment
SF6 + e SF6 (2)
Due to the increasing interest in SF6 as high voltage
insulation, many experimental and theoretical investi- or to one of the constituents of the molecule after
gations have been reported in the literature describing disassociation.
the breakdown behavior of SF6 under direct, alternating
and impulse voltages. In this review, the experimental SF6 + e -SF5 + F- (3)
results are discussed in the light of the theoretical
statements. We also have outlined possible areas of This later reaction is known as disassociation attach-
further investigations to clarify the breakdown ment. The attachment processes are described by the
mechanisms and the discharge phenomenon in compressed electron attachment coefficient i, which is the number
SF6. of attachments produced in the path of an electron
travelling a unit distance under the influence of the

0018-9367/78/0600-0135$00.75 (i 1978 IEEE

 J
136 IEEE Trans. Electr. Insul, Vol EIN13 No 3, June 1978

applied field. Both coefficients, a and n, are engineering problems. Furthermore, the Townsend
strongly dependent on the applied electric field. As mechanism is unable to explain the breakdown under
a result of the combined ionization and attachment steep voltage surges.
processes the increase dN in the number of electrons
along the path dx in the direction of applied electric In the streamer mechanism it is assumed that the
field is given by: growth of a single electron avalance becomes unstable
before reaching the anode. This results in the forma-
tion of fast moving streamers from the avalanche head.
dN = N(a - n) dx (4) These streamers form a highly conducting channel across
the gap, which ultimately causes the collapse of the
where N is the initial number of electrons. For a > , applied voltage. The basic mechanism behind the for-
equation (4) leads to an exponential growth of the mation of these streamers is assumed to be photoioniza-
electron population, eventually resulting in the break- tion in the gas. However, a satisfactory quantitative
down of the insulation. theory for streamer formation has not been formulated.
In spite of the criticism of the streamer theory in
Two types of breakdown mechanisms are generally used the literature, it is necessary to assume a single
to describe the breakdown processes in gases. Accord- avalanche mechanism to explain the breakdown character-
ing to the "Townsend mechanism", a succession of elec- istics in non-uniform field gaps subjected to steep
tron avalanches initiated at the cathode causes break- fronted impulse voltages [5]. Meek [6] developed the
down of the gas. However, in the "streamer mechanism" following equation for the breakdown of a non-uniform
a single electron avalanche can lead to a breakdown. field gap
Details of these mechanisms are given in most text-
books on gas discharges. Pedersen [2,3] has given
short summaries of the Townsend and streamer theories.
(X-nLx exp[fr(a-nf) dx] = k EX Fx 10 (8)
In an electronegative gas under non-uniform field P.
conditions, the prebreakdown current I compared to the
initial photoelectric current Io can be written as [3]
d x
In this equation x denotes the avalanche length at the
1 +
0
exp[JO (a-n) dx] adx moment when streamers are formed. In a non-uniform
field, x is only a fraction of the gap length d.
I (5) (a-n)x is the apparent Townsend first ionization co-
efficient at the avalanche head, p is the gas density
1- y J
exp[f0 (a-n) dx] adx while k is a constant. A similar equation was pro--
posed independently by Raether [7] who suggested that
where the secondary ionization coefficient y is ti 'ie a critical number of charges, = 108, is necessary to
number of secondary electrons produced per primar) transform an avalanche into a streamer. This number
electron as a result of ionizing processes such ass is believed to be independent of the gas pressure.
electron emission from the cathode by positive ior
bombardment or photoelectric emission from the-catthode Meek's equation was modified by Pedersen [2,3] to
and photoionization in the gas caused by photons ire- obtain the semi-empirical streamer criterion:
leased from excited atoms or from recombination piro-
cesses. From equation (5) the Townsend criterion for
breakdown in non-uniform fields is derived as
;OX (a-n) dx = k (9)
d
exp[jr (n-n) dx] adx - 1 (6)
Values of k equal to 18 and 10.5 have been reported in
the literature [3,8,9].
Equation (6) implies that an infinite amount of current
will flow as a result of breakdown of the dielectric
medium. A limit to the current flow is dictated by the 3. MEASUREMENTS OF DISCHARGE PARAMETERS
parameters of the electric circuit. Equation (6) can
be modified for applications to uniform fields as In the absence of detachment and secondary processes,
the current flowing in a uniform field gap can be
expressed as
XE (exp[t T] a - n) -1 (7)
a-n
I0 l (exp[(a-n) d] aJ
I

n~j
-
-n
(10)
The Townsend breakdown criterion (equation 6 or 7) is
of little use to the high voltage design engineer be-
cause the secondary ionization coefficient y is a very Experimentally, the current in a uniform field gap at
sensitive function of the electrode surface conditions fixed values of E/p and variable gap length d is
and gas purity. y is well known only under carefully measured. HereE is the applied electric field intensity
controlled test conditions. Moreover, measurements of and p is the gas pressure. Values of a and n in equa-
y have so far been possible only at pressures below tion (10) are obtained from semilogarithmic plots of
3.4 kPa [4] and such values are not valid for pressures I/Io against d.
of technical importance. It is therefore difficult to
apply the Townsend criterion in a meaningful way to
 Malik and Qureshi: Breakdown Mechanisms in Sulphur-Hexafluorine 137

Geballe and Harrison [10] and Bhalla and Craggs [4] Values of Townsend's second ionization coefficient y
measured values of a/p and r/p as function of E/p in a have been reported by Bhalla et al [4], Boyd et al [12]
the range of 0.6 E/p < 1.24 V/cmPa with
< pressures and Maller et al [13] in the range of 0.75 < E/p < 4.5
varying from 0.66 to 26.66 kPa. They concluded that up V/cmPa for pressures below 3.33 kPa. These values
to pressures of 54 kPa, the breakdown in SF6 is caused strongly depend on the gap separation. For example,
by the Townsend build-up mechanism. Dutton et al [11] with an increase in the gap length from 0.1 to 8 cm for
reported a strong pressure dependence of a/p and n/p a virtually constant field strength gap, y decreases
for pressure up to 53.32 kPa and E/p 0.88 V/cmPa. = by a factor of 109. This weakening of the secondary
They suggested that at such pressures one or more ionization processes is a consequence of the effect of
attachment processes operate which depend p2. How- on photon absorption in the gas medium. This effect is
ever, Boyd et al [12] who carried out similar measure- enhanced by an increase in the gap length [16].
ments over a range of 0.86 1.5 V/cmPa ruled out < F/p <
any pressure dependence of a/p and n/p in the pressure Measured values of a and n in equation (11), are
range of 0.66 < p 53.32 kPa. According to them, the used in calculating breakdown voltages based on
previous conclusions [11] concerning the occurrence of streamer or Townsend theories. All these values have
three-body attachment processes may be in error, and been obtained at relatively low pressures and it is
a/p and n/p are linear functions of E/p only. More generally assumed that these results can be applied to
recently Maller et al [13] also did not observe any higher pressures of technical importance. However,
pressure dependence of a/p and n/p in the of 0.82 range the validity of such an assumption has not been veri-
< E/p < 7.5 V/cmPa with gas pressure varying from fied. Similarly, values of y which have been obtained
133.3 to 1333 Pa at 20fC. Results of the various investi- under laboratory conditions may not be applicable to
gations in reasonable agreement with each other.
are practical cases.
Bortnik et al [14] measured a/po and n/po equal to
333.3, 653.2, 2133, and 4266 Pa in the range of 0.75
< E/po < 1.5 V/cmPa. Here po is the gas normal- pressure 4. MEASUREMENTS OF DISCHARGE VOLTAGES
ized to 0C. Their results show a fair agreement with IN UNIFORM FIELDS
those of Bhalla et at [4].
The maximum discharge voltage for any gaseous insula-
tion can be obtained when the electrodes are designed
The values of a/p and ni/p can be approximated by the for uniform field distribution in the gap. In high
following linear relations in the vicinity of the point voltage SF6 insulated equipment, one of the main condi-
of intersection of the two curves tions for obtaining minimal insulation spacing is,
wherever possible, to use uniform or slightly non-
uniform field gaps in which the ratio of the average
al/p = 23 (E/p) - 12. 34 cm I kPal (11) field strength to its maximum value is greater than 0.2.
Therefore, uniform field discharge data are of consider-
n/p = -4 (E/p) + 11.35 cm-1 kPa-1 able importance from a theoretical as well as a practical
point of view.
Thus a reasonable correspondence with the data over a Uniform field breakdown studies can best be represented
wide range of values is obtained by using the empirical by obtaining the Paschen curve and the ranges in which
expression Paschen's law is valid. Paschen's law states that the
breakdown voltage Vd of a gas in a uniform electric field
ap = k Elp -
(E/P)crit} cm1 kPa-1 (12) field gap is a simple function of the product pd at con-
stant temperature. Paschen curves for SF6 reported in
the literature were not always measured under the same
where k 27 and (E/p)or,t 877,5 V/cmt
test parameters and in many cases the test conditions
were not fully described. Therefore, only a limited
Here E [kV/cm] denotes the electric field strength and comparison of these results is possible, especially
p [kPa], the SF6 pressure referred to a temperature of in the pd range above 100 kPa cm. For SF6 direct vol-
200C. tage breakdown data is available for pd values up to
about 300 kPa cm [4,13,17-25]. The Paschen curve for
direct voltage is shown in Fig. 1. The minimum of the
It has been suggested that in order to obtain reason- curve occurs at about 3.S x 10-2 kPa cm and is near
ably accurate values of a and n from the prebreakdown 500 V as given by George et al [18]. A considerable
current measurements the variations in the field amount of information is needed to complete the curve
strength between the electrodes must be less than 0.1% at both ends, i.e., at high as well as at low values
while the accuracy in setting the gap distance d must of pd. Deviations from Paschen's law are observed by
be better than 0.1% [8,12,15]. According to Boyd et Maller et al [13] for small spacings with pd values
al [12] a pressure of about 100 kPa sets the upper near Paschen's minimum and by Dutton et al [17,27] and
limit in such investigations. Karlsson et al [8] gave Chalmers et al [20] at higher pressures.
an upper limit of less than 100 kPa cm for prebreak-
down ionization current measurements of a and n. Thus
while values of and n are very important at pressures
a

of technical importance, we may not be able to deter-

mine these values accurately at such pressures [12,8].
138 IEEE Trans. Electr. Insul, Vol EI-13 No 3, June 1978

I I I

1000 1000 a [30] ;a

1-

lW100 .129)
4J
]
2-
100 100 .+[32]
[31] /

v [34] X+
fa
.10
o 1 OO

5-

0..

0.1 s -
.

10-2 10- 1 10 102 103 10-2 10-' 1 10 102 103

Pd (kPa cm) Pd (kPa cm)

Fig. 1: i6aschen curve for SF6 for d$rect applied Fig. 2: Paschen curve for SF6 for alternating (50 or
voZtages. 60 Hz) applied voltages.

TABLE 1
ac Breakdown Voltages for SF6 In Uniform Field Gaps [25]

Temperature - 25QC

50 Hz Crest Breakdown Voltage kV

Distance mm

pd
kPa cm 1 2 3 5 8 10 15 20 25 30 40 50 60

10 9.5 9.5 9.5

9'5 9.5 9.5 9.5 95 9.5 9.5 935 95 9*5
40 26.5 35.5 35.5 35.5 35.5 35.5 35.5 35.5 35.5 35.5 35.5 35.5 35.5
80 42.5 55.5 64.5 10.5 70.5 70.5 70.5 70.5 70.5 70.5 70.5 70.5 70.5
100 50.0 60.0 74 89 89 89 89 89 89 89 89 89 89
200 98.0 103 120 150 168 170 170 170 170 170 170 170 170
300 144 155 190 217 240 253 253 253 253 253 253 253
400 177 185 22b 250 275 303 310 310 310 310 310 310
500 216 248 275 305 356 385 385 385 385 385 385
600 250 278 305 340 395 440 455 455 455 455 455
800 335 370 390 450 500 535 580 580 580 580
1000 395 430 450 505 555 595 635 730 730 730
1200 485 505 562 610 650 685 750 805 870
1400 535 565 620 668 715 748 808 865 (920)
1600 585 618 678 725 768 (805) (865) (920) (970)
1800 680 733 (785) (825) (862) (925) (975) (1025)
2000 735 (790) (840) (880) (920) (975) (1025) (1075)
Malik and Qureshi: Breakdown Mechanisms in Sulphur-Hexafluorine 139

Figure 2 contains the Paschen curve for alternating on the left hand side of equation (9), the field dis-
applied voltages (50 or 60 Hz) [25,26,29-35]. In this tortion caused by space charge in the avalanche head is
Figure, results are included of investigations with generally neglected. Because of the relatively rapid
values falling within the range for which Paschen's variation of a-n with field strength, equation (9) is
law is valid. In this range two rough limits are ob- satisfied even for a very small increase in field
served for pd, i.e., p < 200 kPa and d 0.3 mm. Other strength above the value corresponding to a-n = 0.
test conditions and parameters being the same, the The essential result is that in non-uniform field gaps,
Paschen curves for alternating and direct applied vol- like rod-plane configuration, the critical avalanche
tages almost coincide. Under carefully controlled test length x is a very small fraction of the total gap
conditions there is an extremely small scatter in the length.
values of breakdown voltages and the influence of dif-
ferent electrode materials, e.g., aluminum, silver, The streamer criterion has been extensively studied
nickel, stainless steel, and copper is negligible [25]. [2,2,9,36-46] and has been used to estimate discharge
At higher pressures the breakdown voltage depends on voltages for fields having varying degrees of non-
the type of voltage as well as electrode material. uniformity. It has been used successfully by Pedersen
With aluminum electrodes, breakdown voltage can be [9,38,47] and other [40-42,45] to explain some of the
up to 20% lower than with stainless steel electrodes. effects such as anomalous edge breakdown observed in
For this reason, it is necessary to take the well de- SF6 in uniform field electrodes and the reduction in
fined test conditions into account when comparing re- breakdown voltage due to electrode surface roughness.
sults of different investigations. The major limitation, at present, is the lack of
adequate knowledge of ionization and attachment co-
efficients for the range of pressures used in high
Table 1 gives alternating breakdown voltages for SF6 voltage SF6 insulated equipment [36]. The streamer
in uniform field gaps. The values within brackets are breakdown criterion may give reasonable agreement be-
extrapolated. The values below the heavy line lie out- tween the theoretical and experimental results at
side the range for which Paschen's law holds. No test relatively low pressures. However, at higher pressures
results are available for direct voltage breakdown the experimental and theoretical results differ sub-
values for electrode spacings larger than 2.5 cm. stantially [39,40]. In a recent paper [44] Rein et al
have proposed a theoretical model for the impulse
Departures from Paschen's law, which seem to be breakdown in SF6 using a statistical approach to the
unique for each gas, are observed notably for quite streamer breakdown criterion. Their calculations
high pressures and small spacings. The reasons for the agree with the experimental results for different
limitations of Paschen's law have not yet been fully electrode configurations covering a wide range of
clarified. However, it is interesting to note that pressures and electrode areas.
Paschen's law fails when the electric field strength
exceeds 200 kV/cm (crest). The threshold for these The streamer breakdown criterion, equation (9), can
deviations, which give breakdown voltages departing also be expressed as
from the Paschen curve, is variable, depending on the
electrode material and surface conditions and the x
presence of fine particles in the gas. These and other
factors are discussed in subsequent sections. f 0 (a-n) d - ln (NC) (13)
where Nc is the critical number of ions in the avalanche
when streamer formation starts. A constant value of k
in equation (9) gives a constant value of NC independent
of pressure. Recently Khalifa et al [48,49] have devel-
5. ESTIMATION OF MINIMUM DISCHARGE VOLTAGE oped a method for computing the discharge threshold
The estimation of minimum discharge voltage, i.e., voltage in SF6 at pressures up to 1200 kPa, In this
corona inception or breakdown voltage is of consider- method, the size of the primary avalanche at the onset
able importance from engineering point of view. A of discharge does not coincide with the streamer cri-
method capable of predicting discharge voltage with terion given in equation (9) having values of k = 18 or
reasonable accuracy can save considerable amount of 10.5. Its calculated size depends on the gap geometry
money required in testing new SF6 equipment and can [50]. In the uniform as well as non-uniform field
provide some insight in the breakdown mechanism itself. gaps of positive polarity, there is a critical pressure
As the discharge phenomenon in compressed SF6 is not above which the field emission from the cathode surface
fully understood, there is yet no general model which has been taken into account. This method predicts the
can be used successfully to determine discharge vol- discharge threshold voltages for uniform and non-
tages in SF6 insulated equipment, Existing methods uniform field gaps and explains the dependence of these
can broadly be classified into two categories: those voltages on pressure.
based on streamer theory and those based on concept
of critical field strength. In 1953 Geballe and Reeves [51] proposed a special
condition of self-sustaining discharge for electro-
The streamer criterion, as given in equation (9), negative gases on the basis of Townsend theory. In
offers a very promising approach for the solution of case of uniform fields with large values of pd this
engineering problems. This criterion states that criterion for self-sustaining discharge is
streamers are formed when a critical number of ions is
developed in SF6. These result in partial or total oc = n (14)
breakdown of the insulation depending on the degree of
non-uniformity of the electric field in the gap. In a which is satisfied at a threshold value of the field
uniform field gap, streamer formation will directly intensity. This value of the field intensity is
lead to a breakdown while in a non-uniform field it generally referred to as the critical field intensity
may not cause breakdown. WVh'ile calculating the terms Ecrit. Its value for SF6 at normal atmospheric pres-
sure is about 87 kV/cm. For a < -, electron avalanche
should not grow in electronegative gases and thus
140 IEEE Trans. Electr. Insul, Vol EI-13 No 3, June 1978

a = n sets the threshold condition for the growth of

electron avalanches. In a practical case an electron 6. FACTORS AFFECTING DISCHARGE VOTAGE IN SF6
avalanche will grow when a is slightly greater than n.
Thus, for uniform field gaps, breakdown gradients 6.1. Foreign Particles
should have a limiting value given by
It is now well known that any metallic particles
present in the gas can lower the corona onset and
Ecrit - 877.5 p V cm l (15) breakdown voltages considerably [62-71]. The breakdown
where p is pressure in kPa. However, anomalous break- voltage in contaminated SF6 can be as low as 10% of the
down below this limiting value of Ecrit has been ob- values in clean gas. The breakdown voltage depends on
served [4,14,21,22]. This has been attributed [38,47] the particle shape, size, material, location and motion,
to the field enhancement at the curved edges of the gas pressure and nature of the applied field as shown
in Figure 3 [69]. The effects of thes-e parameters have
electrodes with breakdown taking place outside the been studied by various authors to gain an understanding
uniform field region of the gap. ThIs "edge sparking"
is observed to occur for electrode separations well
within the maximum des igned gap spacings. 500 I I

The concept of threshold field intensity was modified

by Howard [52] to obtain the discharge voltage for non-
uniform field gaps. According to Howard discharge
inception takes place when a becomes equal to ri at 400
points having maximium stress in the gap. Using this (A
approach the discharge voltage can be expressed as
Vd = i d Ecrit (16) -P
o0) 300
Here i _ Ea/.Emax is the field utilization factor
[34,53-57]. Takuma et al 158] suggested a correction
term H in the above expression to obtain a better agree- fo
ment between the calculated and measured breakdown
voltages. The modified expres-sion for the minimum dis- co
200
charge voltage is

Vd i d (E rit +iH) (17)

Using this empirical expression, the calculated values 100
of the discharge voltage were found to be in reasonable .
0~ ~~ ____ 1.

agreement with measured values even for highly non-

uniform field gaps having i = 0.02. This means that
breakdown in SF6 is closely related to the maximum
field strength in the gap, at least as far as non- I I
uniform field configurations are concerned. This may 0 500 1000 1500
explain the relatively flat nature of the flashover 2000
voltage as a function of gap length for long non- Pressure kPa
uniform field gaps insulated with SF6 159,60],
Although the above-mentioned methods mar give reason- Fig. 3: Breakdown in SF6 initiated by free copper wire
able correspondence between the theoretical and experi- particZes in a 150 nn/250 coaxiaZ eZectrode
am

mental breakdown voltages in some cases, a general system: wire diameter is 0.4 [69]. am

agreement has not yet been reported in the literature.

None of these criteria account for the simultaneous
influence of factors like pressure, voltage waveform,
polarity, electrode material, area and surface condi- of the particle initiated breakdown in SF6. However,
tion, and gas purity. These methods only give the in compressed gases where particles play a dominant
corona inception voltages in cases where the breakdown role, a mechanism has not yet been proposed which gives
voltage is substantially higher than the corona onset a satisfactory explanation for this complex phenomenon.
voltage due to the stabilizing action of the positive It has been observed that gas breakdown usually occurs
ion cloud surrounding the highly stressed electrode when metallic particles are at or very near the elec-
[61]. Furthermore, ionization and attachment coeffi- trodes. According to the existing literature the break-
cients used in all these calculations are average values down starts at the particle tip as a result of the high
and the statistical nature of the breakdown phenomenon pulsed field which appears there when the particle
is not taken into consideration. In our view, a con- approaches an electrode and creates a microdischarge.
siderable amount of work is required in this area to Insulating particles such as glass, lucite, etc., have
understand the breakdown phenomenon in SF6, especially little or no effect on the breakdown voltage [63-65,68].
at higher pressures. However, dust particles can lower the breakdown vol-
tages by as much as 30% [65]. Ikeda et al [67] have
reported that the effect of dust particles is particu-
larly pronounced when alternating voltages are applied.
Thus whereas very small quantities of gaseous impuri-
ties or insulating particles do not cause any appre-
ciable reduction of the discharge voltage, free metallic
or dust particles can reduce the breakdown voltage in
SF6 considerably.
Malik and Qureshi: Breakdown Mechanisms in Sulphur-Hexafluorine 141

6.2. Electrode MateriaLc

For low value of pd ie, the range in which Paschen's u

EE 500 I

-0 Fine Finish
law is- valid, the electrode material has practically 1-
-.- Rough Finish
no influence on the breakdown characteristics of SF6
[72]. However, the electrode material definitely
= 400 600 kPa NS
affects the breakdown gradients when the field strength
cm
'N.
exceeds about 200 kV/cm. Figure 4 shows uniform field
breakdown gradients- in SF6 as a function of pressure
300
N.0~~~
for different electrode materials [73]. In the case of U_
400 .
parallel plane electrodes at gas pressure up to 800 kPa
the dielectric strength of SF6 was found to be a L 200 <- e.
function of the mechanical strength of the electrode
material, its melting temperature and work function 200
[72]. At higher pressures, similar dependence of L 100 _*
breakdown voltage on electrode materials have reported
in other gases and their mixtures with SF6 [24,65,74].
While in parallel plane electrodes, the breakdown vol-
tage depends on the electrode material, no such effect I a

10 2 3 4
has been observed in non-uniform fields such as coaxial 10 10 10

arrangement for pressures up to 400 kPa [75]. Area (cm2)

Fig. 5: Minimum breakdown field strength as a function

800 _ _ _ of electrode area and gas pressure [78].
700 - I - Stainless Steel

2 - Copper 6.3 Electrode Area

600
'600- 3 - Dural < < 2
The breakdown voltage in SF6 decreases with increasing
500 - electrode area when the field strength in the gap ex-
ceeds about 100 kV/cm [65]. Figure 5 shows the 0.5%
breakdown strength as a function of the electrode area
V)400 and pressure [78]. The influence of the electrode area
0 is more pronounced at higher pressures and therefore
at higher gradients. Similar results have been recently
300-
a)
5.-
reported by Rein et al [44]. The effect of electrode
area also depends on the waveshape and polarity of the
200- applied voltage. The negative switching impulse break-
down gradients are less influenced by the electrode
area as compared to the power frequency breakdown
100 gradients because of the relatively short duration of
the impulse voltages.
0
100 200 300 400 500 600 700 800
In SF6 the breakdown voltage at higher pressures obeys
Pressure kPa the extreme value distribution. This means-that the
breakdown is governed by the weakest points in the gap,
which have statistical properties in nature. The model
Fig. 4: Dielectric strength of SF6 as a function of value of the distribution of breakdown strength of the
gas pressure for various eZectrode materials [78] irregular gap may be higher than that of the ideal gap.
However, the irregular gap is the dominant gap for
large electrode areas at higher pres-sures, since the
characteristic of the gap is quite sensitive to the
The breakdown voltage primarily depends on the cathode
number of weak points. This number is expected to in
crease with the area of the electrodes.- Weak points
material and the influence of material increases with
pressure and field strength. Although the anode is of the gap are destroyed by breakdown s.parks, and thus
usually considered to have no influence on breakdown, the breakdown vQltage changes- by a ftconditioning effect'".
an anode effect has been reported under certain condi-
The number of conditioning s-parks and the relative
tions [65]. MacAlpine and Cookson [76] showed that an increase in breakdown voltage become larger as the
insulating film on the anode affects the breakdown electrode area is increas-ed 165].
voltage in compressed SF6. Similarly an increase in 6.4 Electrode Surface Conditiona
the breakdown voltage results if the cathode is covered
with a thin insulating film [77]. In general, insulated Electrode surface roughnes-s can cause a large reduc-
coverings markedly increase the breakdown strength. tion of the threshold for breakdown in GF6 insulated
With uniform field conditions they enable Paschen's
law to be obeyed for fields up to 130 kV/cm [76]. apparatus 19,31,41,42,45,46,65,78]. Surface roughness
can lead to the existence of localized microscopic
regions in the gas near the electrode with a field
strength much higher than the macroscopic average field.
 142 IEEE Trans. Electr. Insul, Vol EI-13 No 3, June 1978

Depending on gas pressure, such regions* of enhanced

field strength can result in a large reduction of the 1000
breakdown voltage. This may be one of the reasons that
sometimes breakdown gradients in SF6 are in the region
where a is less than n. 800
.If
Pedersen [9] has studi.ed the influence of the elec,
trode surface roughness using streamer breakdown cri. O
4-
terion. He has shown that the surface roughness will 4.) 600
not affect the threshold of breakdown if a,
va
1-

DRax O. 4 kPa cm (18) 0

0)
Eu
400
where Rmax is the maximum roughnes.s height, assuming a
semispherical protrusion. Figure 6 shows- the results 5-
of his caluclations. Cook 142] suggests that the break- to
down voltage decreases when the product of gas pressure 200
and the protrusion height above a flat electrode ex-
ceeds 0.8 kPa cm. Detailed investigations have been
carried out to understand the effect of pressure, shape
0V I I A a
)l
500 1000 1500 2000
Pd (kPa cm)

Fig. 7: Breakdown voZtage between 43 cm diameter

a.4
cL
Rogowski profiled electrodes as function of pd
U (corrected to 20C) [31].
_e

LI~ x
fE
-1

1000

4,
102 104 I c05 S"-' 800
(U
PRmax (Pa cm) a,

fO 600
0)

Fig. 6: Breakdown limitations in SF6 from surface

roughness. (Elp)max is the threshold value of E/p.
0)
ro
5-
Rmax is the maximum surface roughness and p is the a) 400
gas pressure [9].

200
and size of electrode surface protrusion and the field
configuration on the discharge characteristics of SF6.
Experimental observations generally support these
theoretical findings [9,40-42]. Kawagachi et al [31] 0 2 4 6 8 10
did not find any distinct effect of wavy unevenness of
up to 30 pm, caused by lathing, on the breakdown Gap Length (cm)
voltage of a coaxial system for pressures to about 500
kPa.
With the increase in electrode surface roughness, Fig. 8: 60 Hz breakdown voltage of SF6 in Long gap
the number of conditioning sparks increase many times region as a function of gap Length when electric
that required for smooth electrodes of the same area. field is uniform [80].
Moreover, with an increase in surface roughness,
deviations from Paschen's law occur at lower values of
pd.
Malik and Qureshi: Breakdown Mechanisms in Sulphur-Hexafluorine 143

6.5 Other Factors [10] L. B. Loeb, "Basic processes of gaseous electron-

Other factors such as pressures, gap length, field ics", California University Press, 1955, pp 415.
configuration, voltage waveform and polarity greatly [11] J. Dutton et al, Nature, vol. 227, August 1970,
affect breakdown voltage of SF6 [79]. For uniform field pp 702-703.
gaps, usually the variation due to pressure and spacing
can be combined togehter as in Paschen's law. In the [12] H. A. Boyd et al, Proc. IEE, vol. 118, no. 12,
range of pressure and gap spacing where Paschen's law 1971, pp 1872-1877.
holds, any decrease of one of these factors can be com-
pensated by an increase in the other if their product [13] U. N. Maller et al, Proc. IEE, vol. 123, no. 12,
is kept constant. However, at higher pressures where 1976, pp 107-108.
Paschen's law fails, an increase in pressure at constant
gap length usually results in a saturation tendency in [14] I. M. Bortnik et al, Soviet Physics - Technical
the breakdown voltage versus pressure curve, Figure 7 Physics, vol. 16, no. 4, 1971, pp 571-575.
[31]. Figure 8 gives the breakdown voltage of SF6 in
long gaps as a function of gap length when the electric [15] A. Pedersen et al, IEEE Trans. on P.A.S., vol.
field is uniform. This illustrates a similar saturation PAS-90, no. 5, 1971, pp 2175-2180.
caused by varying the gap length [80]. For uniform
fields the ratio of V/pd or E/p for breakdown is not [16] A. S. Perlin, Soviet Physics - Technical Physics,
constant, but it decreases with increasing gap length vol. 17, no. 5, 1972, pp 813-817.
towards an asymptotic value [25,51,79]. This limit is
reached when the attachment coefficient n equals the
ionization coefficient a. [17] J. Dutton et al, PrQc. IEE, vol. 118, no. 5, 1971,
pp 732-733.

In non-uniform fields, the effect Qf pressure and [18] D. W. George et al, Brit. J. Appl. Phys., vol. 2,
spacing depends also on the polarity of the highly Ser. 2, 1969, pp 1470-1471.
stressed electrode. For example, in the case of posi-
tive rod-plane gap, the breakdown voltage expressed as [19] H. A. Boyd et al, Proc. IEE, vol. 119, no. 2,
a function of pressure exhibits a maximum. This maxi- pp 275-276.
mum depends on the gap spacing, radius of curvature of
the rod and the waveform of the applied voltage [18, [20] I. D. Chalmers et al, Proc. IEE, vol. 118, no. 12,
37,81-84]. For negative rod plane gaps, the breakdown 1971, pp 1893-1894.
voltage is a linear function of pressure for smaller
gap spacings. However, negative slopes of breakdown [21] P. R. Howard, Proc. IEE, vol. 104, Part A, no. 13,
voltage versus pressure curve have been observed when 1957, pp 123-238.
the gap is sufficiently long [50,59,60,64]. Therefore,
one has to be careful in using test data for design [22] E. Kuffel et al, Proc. IEE, vol. 113, no. llj 1966,
purposes, as results obtained under one set of condi- pp 1863-1872.
tions may not be applicable to a different set of
conditions. [23] D. F. Binns et al, Proc. IEE, vol. 116, no. 11,
1969, pp 1962-1968.
[24] E. H. Cohen, Proc. IEE, vol. 103, Part A, no. 7,
REFERENCES 1956, pp 57-68.

[1] D. Kind, Trans. of SA Institute of Electrical [25] T. W. Dakin et al, Electra, no. 32, January 1974,
Engineers, vol. 65, part 12, 1974. pp. 61-82.

[2] A. Pedersen, IEEE Trans. on P.A.S., vol. PAS-86 [26] G. Oppermann, "Uber Die Gultigkeit des Paschen-
no. 2, 1967, pp 200-206. Gesetzes Fur Schwefelhexafluorid Bei Gleich,
-Wechsel-und Stoss-Spannungen". Ph.D. Dissertation,
[3] A. Pedersen, IEEE Trans. on P.A.S,, vol. PAS-89, 1974, Technical University of Berlin.
no. 8, 1970, pp 2043-2048.
[27] J. Dutton et al, 1970 International Conference on
[4] M. S. Bhalla et al, Proc. Phys. Soc., vol. 80, Gas Discharges, IEE Conf. Publication, no. 70,
1962, pp 157-160. pp 544-548.

[5] A. Rein, Electra, no. 32, 1974, pp 43-60. [28] A. M. Zalesskii et al, Elektrichestvo, no. 12,
1967, pp 6-9,
[6] L. B. Loeb, J. M. Meek, "The mechanism of electri-
cal spark", Stanford, California, Stanford [29] C. N. Works et al, Annual Report 1964 Conf. on
University Press, 1941. Elect. Insul., pp 69-72.

[7] H. Raether, "Electron Avalanches & Breakdown in [30] S. Schreier, IEEE Trans. on P.A.S., vol. PAS-83,
Gases", London, England, Butterworth, 1964. 1964, pp 468-471.

[8] P. W. Karlsson et al, IEEE Trans. on P.A.S., [31] Y. Kawaguchi et al, IEEE Trans. Qn P.A.S., vol.
vol. PAS-91, no. 4, 1972, pp 1597-1601. PAS-9U, no. 3, 1971, pp 1072-1078.
[9] A. Pedersen, IEEE Trans. on P.A.S., vol. PAS-94, [32] W. A. Wilson et al, Journal of Applied Physics,
no. 5, 1975, pp 1749-1754. vol. 21, March 1950, pp 203-205.
144 IEEE Trans. Electr. Insul, Vol EI-13 No 3, June 1978

[33] G. Camilli et al, Trans. of AIEE, vol. 74, Part I, [54] V. N. Borin, Elektrichestvo, no. 9, 1976,
Nov. 1955, pp 637-642. pp 51-53.

[34] S. Menju et al, IEEE Trans, on P.A.S., vol. PAS-93, [55] W. Mosch et al, Elektrie, vol. 28, H-3, 1974,
no. 5, 1974, pp 1706-1712. pp 152-156.

[35] G. Camilli et al, Trans. of AIEE, Vol. 71, Part II, [56] W. Mosch et al, Elektrie, vol. 26, H-9, 1972,
1952, pp 348-357. pp 250-253.

[36] J. Blackett et al, 1970 International Conference [57] W. Houschild et al, Elektrie, vol. 26, H-7, 1972,
on Gas Discharges, IEE Conf. Publication No. 70,
pp 198-202.
pp 293-297.
[58] T. Takuma et al, Electrical Engineering in Japan,
[37] R. Hazel et al, IEEE Trans. on P.A.S., vol. PAS-95, vol. 90, no. 4, 1970, pp 26-33.
no. 1, 1976, pp 178-186.
[59] T. Takuma et al, Electrical Engineering in Japan,
[38] A. Pedersen et al, IEEE Trans. on P.A.S., vol. vol. 90, no. 6, 1970, pp 227-234.
PAS-93, no. 6, pp 1820-1826.
[60] T. Takuma et al, Electrical Engineering in Japan,
[39] T. Nitta et al, IEEE Trans. on P.A.S., vol. PAS-90, vol. 90, no. 3, 1970, pp 166-176.
no. 3, 1971, pp 1065-1071.
[61] C. N. Works et al, Trans. of AIEE, vol. 72, Part I,
[40] I. M. Bortnik et al, IEEE Trans. on P.A.S., vol. 1963, pp 682-687.
PAS-91, no. 5, 1972, pp 2196-2203.
[62] A. H. Cookson et al, IEEE Trans. on P.A.S., vol.
[41] S.-Berger, IEEE Trans. on P.A.S., vol. PAS-95, PAS-91, no. 4, 1972, pp 1329-1338.
no. 4, 1976, pp 1073-1079.
[63] H. C. Doepken, IEEE Trans. on P.A.S., vQl. PAS-88,
[42] C. M. Cook, IEEE Trans. on P.A.S., vol. PAS-94, no. 5, 1969, pp 364-369.
no. 5, 1975, pp 1518-1523.
[64] A. Diessner et al, IEEE Trans. on P.A.S., vol.
[43] W. Mosch et al, Elektrichestvo, no. 5, 1974, PAS-89, no. 8 , 1970, pp 1970-1978.
pp. 50-54.
[65] A. H. Cookson, Proc. IEE, vol. 117, no. 1, 1970,
[44] A. Rein et al, IEEE Trans. on P.A.S., vol. PAS-96, pp 269-279.
no 3, 1977, pp 945-954.
[66] H. Kuwahara, IEEE Trans. on P.A.S.,?vol. PAS-93,
[45] S. Berger, IEEE Paper No. A75 494-5, IEEE PES no. 5, 1974, pp 1546-1555.
Summer Meeting, San Francisco, Calif., July 20-25,
1975. [67] E. Ikeda et al, Electrical Engineering in Japan,
vol. 91, no. 5, 1971, pp 67-74.
[46] S. Berger, IEEE Trans. on P.A.S,? vol. PAS-96,
no. 4, 1977, pp 1179-1189. [68] H. W. Graybill et al, IEEE Trans. on P,A.S., vol.
PAS-89, no. 1, 1970, pp 17-23.
[47] T. M. Nielson et al, Proceedings of Second
International Conference Qn Gas Discharges, [69] A. H. Cookson et al, IEEE Trans. on P.A.S., vol.
London, England, Sept. 11-15, 1972, pp 323-325. PAS-92, 1973, pp 871-876.
[48] M. Khalifa et al, IEEE paper No. F-76-324-4, PES [70] H. C. Doepken, Annual Report 1969 Conf. on Elect.
Sunmer Meeting, Portland, Oregon, July 18-23, 1976, Insul. & Dielectric Phenomenon, NAS. Pub. No. 1764,
pp 343-348. 1970, pp 7-11.
[49] M. Khalifa et al, Proceedings of International [71] C. M. Cook et al, IEEE Trans. on P.A.S., Vol.
High Voltage SympQsium, Zurich, Sept. 9-13, PAS-96, no. 3, 1977, pp 768-777.
1975, pp 343-348.
[72] I. M. Bortnik, Elektrichestvo, no. 12, 1974,
[50] Abdel-Salam, Discussion, IEEE Trans. on P.A.S., pp. 20-27.
vol. PAS-96, no. 3, 1977, pp 952.
[73] B. A. Goryunov, Soviet Physics - Technical
[51] R. Geballe et al, Phys. Rev., vol. 92, no. 4, Physics, vol. 20, no. 1, 1975, pp 66-67.
1953, pp 867-868.
[74] G. D. Trump et al, Electrical Engineering, vol. 69,
[52] P. R. Howard, Proc. IEE, vol. 104, Part A, no. 13, 1950, pp 961-964.
1957, pp 139-142.
[75] Y. Kawaguchi et al, Electrical Engineering in
[53] Y. Kawaguchi et al, IEEE Trans. on P.A.S., vol. Japan, vol. 90, no. 5, 1970, pp 197-204.
PAS-90, no. 3, 1971, pp 1079-1085.
[76] J.M.K. MacAlpine et al, Proc. IEE, vol. 117,
no. 3, 1970, pp 646-652.
 IEEE Trans. Electr. Insul, Vol EI-13 No 3, June 1978 145

[77] D. J. Chee-Hing et al, IEEE Trans. on Elect. [82] D. Berg et al, Trans. of AIEE, vol. 77, Part III,
Insul., vol. EI-10, no. 4, 1975, pp 119-124. 1958, pp 820-823.

[78] T. Nitta et al, IEEE Trans. on P.A.S., vol. PAS-93, [83] T. R. Foord, Nature, vol. 166, 1950, pp 688-689.
no. 2, 1974, pp 623-629.
[84] H. C. Pollock et al, Phys. Rev., vol. 56, July
[79] P. Narbut et al, Trans. of AIEE, vol. 78, Part III, 1939, pp 170-175.
1959, pp 545-551.
[80] K. Itaka et al, IEEE Trans. on P.A.S., vol. PAS-89,
no. 8, 1970, pp 1986-1994.

[81] A. A. Azer et al, IEEE Trans. on Elect. Insul., Manuscript was received 12 December 19?7; in revised
vol. EI-8, no. 4, 1973, pp 136-141. form 3 March 1978.

HIGH SPEED PHOTOGRAPHY OF SURFACE FLASHOVER

IN VACUUM

J. D. Cross

Dept. of Electrical Engineering

University of Waterloo
Waterloo, Ontario, Canada

ABSTRACT
The method and results of high speed streak photography
of surface flashover in vacuum are presented. It is
shown that the bright phase of the flashover arc bridges
a 12.5 mm gap in 0.15 ns. The streak records indicate
that the flashover arc is preceded by an intense electron
burst from the cathode-insulator junction.

INTRODUCTION EXPERIMENTAL PROCEDURE

In any system using vacuum as electrical insulation The photography of arcs under dc stress at high
there must be at some point a solid insulator separating voltages involves problems in coincidence and pro-
and supporting the conductors. The surface of such a tection. With a breakdown under dc stress the event
solid is a weak point in the insulation and therefore time cannot be predicted and yet the camera operation
its characteristics are of technical importance. The must coincide with the initial stage of the arc that
insulation fails by the development of an arc along lasts less than one nanosecond. In earlier work [3],
the vacuum-solid interface. This type of failure is the coincidence was achieved by triggering both the
referred to as surface flashover. It has been reported arc and the camera from a spark but with the camera
that the flashover is a very fast process [1,2,3,4], used in this study (Imacon 600) a more convenient
and in an earlier study at this laboratory we showed method of obtaining the coincidence is available. The
that the high luminosity phase of the flashover be- camera is an electronic image converter camera, used
tween uniform field electrodes bridged by a cylinder in this study in the streak mode, with a streak rate
of alumina propagated at a speed greater than 107 m/s; of 20 mm/ns. The camera will start operation with a
that is, beyond the limit of resolution of the camera small time delay after the arrival of a trigger signal.
available at that time. This paper describes a That time delay is 28 ns for the camera speed used in
successful attempt to measure the rate of arc propaga- this study, the time increasing as the speed of the
tion using a camera with much higher speed. camera is reduced. If the trigger is obtained from the

0018-9367/78/0600-0145$00.75 i 1978 IEEE