SF6 Decomposition in Gas-Insulated Equipment

IESEE Trarnsactions fl
on ectrical Insulatiorn Vc;s EI--l No.5, cJctober l?B6
SF6 DECOMPOSITION IN GAS-INSULATED EQUIPMENT
F. Y. Chu
Ontario Hydro Research Division
Toronto, Canada
ABSTRACT
The increasing application of SF6 as an insulating gas has
led to many studies on SF6 decomposition in gas-insulated
equipment. In the presence- of an electric arc, spark or
corona, SF6 decomposes to a wide variety of chemically active
products which possess completely different properties from
SF6. The accumulation of these decomposition products in the
equipment has caused concerns regarding personnel safety and
material compatibility problems. This paper reviews previous
research in SF6 decomposition relating to the operation of gas-
insulated switchgears, gas-insulated transmission lines, and
electrostatic accelerators. Results on the qualitative and
quantitative determination of the by-products and their forma-
tion rates in various modes of electrical discharges are sum-
marized. The mechanisms leading to the formation of transient
and stable products are described. In particular, the influence
of discharge energies and impurities on the formation of SOF2
and S02F2, the two dominant stable by-products, is discussed.
The effects of the by-products on personnel safety and equip-
ment dielectric integrity are presented. The application of
SF6 gas analysis as a tool for diagnosing the internal con-
dition of gas-insulated equipment is assessed. Based on the
results of many phenomenological observations, future research
activities are suggested to address the issues of safety, com-
patibility and equipment aging. More fundamental studies on
electron, ion, and neutral reaction rates in an SF6 discharge
are required to gain a better understanding of the decompositon
mechanisms and the influence of the products on equipment
operation.
1. INTRODUCTION
stances must be determined so that procedures can be
established to protect personnel during maintenance and
The increased applications of SF6 in gas-insulated clean-up of faulted equipment [21,146]. The require-
switchgear (GIS), gas-insulated transmission lines ments for long-term equipment reliability impose severe
(GITL), electrostatic accelerators, x-ray equipment and constraints on the material compatibility with the SF6
pulse power apparatus [9,20,50,110,176] have led to environment. The effects of SF6 decomposition products
growing interest in the mechanisms of SF6 decomposition on materials used in gas-insulated equipment have to be
and the properties of the decomposition products. In examined carefully to ensure that no degradation of the
the presence of an electric arc, spark or corona, SF6 materials' insulating, physical and chemical properties
decomposes into lower fluorides of sulphur which in turn would occur, leading to premature aging. In the develop-
react with the electrodes or gas impurities to form many ment of SF6 circuit breakers, the influence of decomposi-
chemically active products. Although SF6 is chemically tion products on the dielectric recovery after current
inert and environmentally acceptable, the decomposition zero is critical to the breaker's function. Much of the
products of SF6 are known to be toxic and corrosive earlier work in SF6 decomposition focused on the thermal
[23,30,45,65,83,85,165]. The motivation behind the phenomena during high-current interruption in a hot SF6
studies on SF6 decomposition is the concern for person- gas gap [73,108,125,127]. The insulating properties of
nel safety and material compatibility. The studies may the SF6 dissociation products in the predischarge stage
also lead to improved equipment diagnostic techniques. are reported to be critical in influencing the formation
of leader channels in the discharge process [139,140,150].
From a health and safety point of view, the toxicity On the more positive side, the identification of SF6 de-
of SF6 decomposition products must be clearly identified composition products in electrical equipment offers a
and the quantities generated under all possible circum- means to diagnose the equipment's internal condition.
C1Cu1B-9j57/B cr 1 ,c){c,
6c,
,}_j_-ff$f_ct wi9f,lO I_
- 8- IEE E
Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:03 UTC from IE Xplore. Restricon aply.
1 ^ . r t~ F _-_
+ r~"-
t r x- esct~>t: 1n, t st:Jr V. _ 1 N' n D. te- XI Z
Diagnostics by SF6 gas analysis is very similar to the process: power arc, spark and corona or partial dis-
well known technique of gas-in-oil analysis for detect- charge. Table 1 summarizes the types and character-
ing abnormal conditions in transformers. istics of discharges in gas-insulated equipment. Under
In order to address adequately the issues outlined
above, an in-depth knowledge of the SF6 environment, TABLE 1. Discharge In SF6 InsuZated Eqadpment
decomposition mechanisms, decomposition rates, proper-
ties of the decomposition products and the interaction
of the by-products with the equipment are required.
The objective of this paper is to review the major in- Types of Disdharge Discharge CharactrLstics
vestigations relating to SF6 decomposition which have
been carried out in the past 35 years. Due to the Powver Arc 3-100 kA, 50-150 ms
diversified nature of the work, which crosses many dis- (Current interruption, in- -10'fi- 10' J
ciplines such as chemistry, physics, electrical engi- service fault)
neering, material science, and industrial hygiene, it High instantaneous cur-
is impossible to give a comprehensive review covering Spark rent, As duration
10' - 10' J
topics ranging from the fundamental aspects of SF6 (Capacitive discharge, fault
during testing, disconnector
decomposition to practical issues. Instead, the empha- switching)
sis of this review is to extract information from the 10- 10' PC
many phenomenological observations and to outline the Partial Discharge or Corona 10- - 1O- J er pulse
possible factors influencing the formation of decom- (Stress enhancement, elec- 10' 10' Hz
position products. trically floating component,
free conducting particles)
Much of the earlier work on SF6 decomposition focused

on the qualitative analysis of the by-products formed
under conditions easily simulated in the laboratory normal operating conditions, SF6 can be decomposed in a
with a gas cell [30,64,78,165,178]. The limitations in power arc during current interruptions in circuit break-
those early days included the lack of sophisticated ers. The energy dissipated in the discharge depends on
analytical techniques to determine unambiguously the the arc current which ranges from several kA to 100 kA.
types of species formed. As newer techniques were The energy can be expressed as:
developed and researchers gained a better understanding
of the decomposition environment, research emphasis fr(t)I(t)dt (1)
shifted to the quantitative analysis of the by-products
and attempts were made to estimate the amounts formed where I is the arc current, U is the arc supporting
in a realistic environment such as encountered in oper- voltage in the range of several hundred V and t is the
ating GIS [14,129,179,193]. One of the objectives in duration of arcing [16,46]. Typical values of t are in
the earlier works was to determine the amount of ab- the range of 50 to 150 ms. The energy dissipated is in
sorbent material required inside the GIS to remove the the order of 105 to 107 J and the arc temperature at the
decomposition products. Boudene's work in the mid- axis can reach 20000 K [5,142,164]. In SF6 circuit
seventies signified a major step forward in understand- breakers, the electrode material is usually a Cu-W alloy
ing the complicated issue of health and safety and SF6 which is highly arc resistant and the hot plasma gas
decomposition [23]. As the development of SF6 circuit flows through a Teflon nozzle. Hot SF6 (T>2000C) is
breakers proceeded at a rapid pace, many studies were known to interact with metals, especially copper [36a,
carried out to determine the equilibrium concentration 93].
of the decomposition products in a high temperature arc
plasma [18,52,72,73,108,123,126]. The influence of Another type of power arc is generated during short
ions as well as neutrals in the plasma was recognized, circuits inside a gas compartment. The characteristics
and theoretical formulations of the equilibrium con- of such fault arcs are very similar to those for arcs in
centration of the by-products were developed. However, circuit breakers except for the absence of the cooling
the circuit breaker work mainly focused on the recom- process from the flow gas, and the electrodes are typi-
bination process without devoting too much attention to cally aluminum instead of copper alloys. The occurrences
the stable neutral species. With the increase in con- of fault arcs in gas-insulated equipment are less fre-
cern about safety, the formation and effects of the quent than in circuit breaker operations, but under ab-
stable neutral species started to receive more atten- normal circumstances such as an enclosure burnthrough
tion, and sophisticated techniques to trace the forma- or a pressure relief disc operation [41], the decomposi-
tion of these species were developed. Investigations tion products generated can be discharged into the am-
were carried out to address the various modes of forma- bient environment creating hazardous situations.
tion in high current arc, spark and corona [11,44,45,
100,116,147,159-161, 166, 180,184,187,188]. The effects Spark is the term used to describe capacitive dis-
of impurities on the formation rates were systematically charges across a gas gap of very short (usually ps)
investigated. In recent years, the equipment reliabil- duration. The energy level involved in the discharge
ity problems and the diagnostics issues have also rein GIS is in the order of 10-1 to 102 J per spark. The
ceived considerable attention [12,35,47,104,105,116,130, spark channel is much narrower than the arc channel de-
141,171,172,180]. scribed in previous paragraphs and the temperature dis-
tribution is highly localized. In electrical power
2, THE SF6 OPERATING ENVIRONMENT equipment, flashovers during equipment testing by a
resonant test device, where all the energy is that
2.1 Conditions for SF6 Decomposition stored in the equipment's capacitance and inductance,
can be classified as a spark. During a disconnector
In gas-insulated equipment, decomposition of SF6 by operation in SF6 insulated switches, arcing can occur
an electrical discharge is the most common mode of dis- across contacts as they separate or approach one another.
sociation. The discharges can be broadly divided into The breakdown magnitudes, durations, energies, and fre-
three types according to the energy dissipated in the quencies ef occurrence of this type of arcing (usually
C1h--u: de_:3foo6i-! ,_ -ul_ated
- n equi-pme1;
called restrikes or prestrikes) have been investigated expressed in gram per joule. In corona discharge, the
extensively [19,138]. Except for fast-acting switches, energy dissipated is sometimes difficult to obtain. In
normal arcing time in a disconnector operation lasts many cases, the unit of charge transport (C) is used
from 0.5 to 1.5 s depending on the disconnector's de- instead of the energy unit. The concept of mole per
sign. Within this operating period, there are a few coulomb or ,umol/C will be used to express the formation
hundred restrikes across the contacts with each re- rate of decomposition products in corona or partial dis-
strike lasting tens of ps. Because of the capacitive charge.
nature of the discharge, each strike can be classified
as a spark. In a typical disconnector operation, up 2.3 Impurities
to 500 J of energy can be dissipated in the discharge
process [45]. The quantities of SF6 decomposition The decomposition of SF6 is greatly influenced by gas
generated depend on the number of disconnector impurities. In industrial grade SF6 the typical im-
operations. purities are CF4, N2, 02, and H20. Table 2 shows the
new gas specifications, the American Society for Testing
The third type of discharge discussed in this review
is corona or partial discharge. Under abnormal operat-
ing conditions, partial discharge or corona discharge
can occur in SF6 insulated equipment due to electrically TABLE 2. Impricis in SF6
floating components, stress enhancements, metallic
particle discharges and voids inside solid insulators
[111,181]. Many electrostatic accelerators employ a New Gas ASTM Typical Values
series of continuous corona discharges over the length Spec 161 2472-S1i21 Operating GIS (10)
of the accelerator to establish a uniform gradient
between the HV terminal and the ground [9,96,143,156, Purity (wt %) 99.9 99.8
182]. Prolonged partial discharge or corona gradually Air (ppmv) 2500 2500 1,00010,000
decomposes the SF6 leading to the accumulation of high
levels of decomposition products inside the compartment. H,O (ppmv) 11 71 100500
The magnitudes of the partial dischage are normally ex- CP, (ppmV) 800 640 100500
pressed in pC per pulse and the energy dissipated in a Hydrolyzable 0.3 0.3 7,
partial discharge pulse varies from pJ to mJ depending Fluoride (ppmw)
on the discharge level and configuration. Since Oil (ppmw) 5 5 N/A
partial discharge can be a continuous process, the
levels of SF6 decomposition products formed in a com-
partment are time dependent. With pulse repetition
rates in the 100 Hz range, the total energy dissipated
Max. SOF, concentrations for operating circuit breakers
Max. SOF, concentrations for operating bus ducts
over a period of time can approach 10 kJ, resulting in
significant accumulation of arc by-products in the
compartment.
and Materials (ASTM) standards, and typical concentra-
In addition to the above three major mechanisms, SF6 tions found in operating GIS in North American utilities
can be decomposed thermally without the presence of an [2,6,100,101]. It is not surprising to find that in
electrical discharge [36a,93]. Poor contact resistance reality the impurity concentrations in operating GIS are
in high-current contacting points is the frequent cause much higher than the specifications. This is partly due
of overheating in power equipment which eventually re- to the introduction of impurities during filling and
sults in thermal decomposition of the SF6 insulating gas partly due to the desorption of moisture into the dry
[65,94,181]. Optical radiation can dissociate SF6 mole- SF6 after filling [48,57]. A survey of major North
cules, and in an arc, the ultra-violet radiation may de- American utilities indicated that the average air con-
compose SF6 located away from the arc zone. The use of centrations in SF6 compartments are 5000 ppmv and the
x-ray radiography to diagnose GIS and the use of ioniza- average moisture content is about 500 ppmv [133,175].
tion-type gas density monitors [155] may lead to SF6 In addition to impurities which exist in the gaseous
decomposition by the ionizating radiation [86]. When phase, there are adsorbed and dissolved impurities, main-
SF6 is used as an insulating gas in rf or microwave ly H20, in electrodes and solid insulators. During an
power cables, it can be dissociated in a rf discharge. electrical discharge, moisture and oxygen will be driven
Recently SF6 has been used in the production of fluo- out from the host matrix to react with the primary SF6
rine atoms in plasma etching of silicon or other semi- decomposition products. In practical GIS environments,
conductor materials [3,149]. the presence of impurities in the 100 to 1000 ppm range
is unavoidable.
2.2 Energy Units and Formation Rates
3. DECOMPOSITION OF SF6
Throughout this review, the term "energy per unit vol-
ume" will be employed in describing the formation rates 3.1.1 SF6 Decomposition in a Power Are
of many SF6 decomposition products. The discharge
energy is calculated according to the nature of the dis- This Section reviews the phenomenological observations
charge and is expressed in J or kJ. The volume of SF6 of SF6 decomposition in a power arc while Sections 3.2
at standard temperature and pressure (STP) conditions and 3.3 deal with decomposition in spark and corona con-
is used to obtain the value of energy per unit volume. ditions respectively. The first reported observation of
In most analytical techniques, the quantity of the de- decomposition of SF6 was made by Schumb et al. [165] in
composition products is expressed in microliter per a 20 kV corona discharge and in an ac arc. A portion of
liter (pl/1) or parts per million by volume (ppmv). In the arced gas was absorbed by potash lye, while the non-
order to adopt a uniform unit for comparison purposes, absorbed portion had oxidizing properties. The authors
the concept of mole will be used. Taking a mole as concluded, on the basis of the chemical reactions, that
22.4 1 at STP, the unit of mole per joule (mol/J) or the products were lower fluorides of sulfur. No further
10-9 mole per joule (nmol/J) will be adopted in this re- attempts to analyze the products were made. Camilli,
view. The unit for solid arcing by-products is Gordon and Plump [30] reported the observation of SF2
;nscE
JEEET _iCal
on Elec TnSUt
L.tion Vol. E T-21 N b5e
O 19- 5
and SF4 by IR spectroscopy in SF6 which had been expos- The formation of SF4 in a power arc was further con-
ed to a 5 kA arc discharge between copper electrodes firmed by Becher and Massonne [14] who carried out SF6
separated by 1.3 cm. Traces of C02, H2S, and HF were arcing experiments in a geometry similar to Edelson's.
also identified, but no S2FJO was found. A more thor- A 220 V ac breaking arc was established with a striking
ough investigation was carried out by Edelson using an frequency of 1 Hz. The arc current was estimated be-
IR technique to study the formation of SF6 arc by- tween 5 to 10 A with a duration of 0.23 s. The concen-
products in a 60 Hz discharge [64]. SF2 was identified tration of the by-products in the cell was monitored in-
as the main decomposition product, along with SOF2 and situ by IR spectroscopy. The results showed that SF4
SiF4. and SOF2 were formed preferably in arcs, whereas the
compounds of SOF4 and S02F2 were prevalent in the spark
Leeds, Browne and Strom [118] studied the formation discharges. S2Fl and S2FlOO were never detected during
of SF6 arc by-products in a commercial size 132 kV cir- SF6 decomposition in arcs while quantities in the range
cuit breaker and proved the effectiveness of activated of 100 ppmv of S2FlO were detected in spark experiments.
alumina in removing the by-products. A few kg of acti-
vated alumina was proven to be sufficient to remove all Tokuyama et al [179] studied the influence of 15 elec-
the by-products generated during 50 current interrup- trode materials on the amount of decomposition in SF6.
tions at 5000 A. The drop in pressure due to absorption Current interruptions in 3 atm SF6 at 800 A and 127 V
was about 2%. No test was made to determine the iden- produced up to 10% decomposition. Experiments were
tity of the products. carried out with full-sized circuit breakers. Empirical
formulae for the amount of decomposition as a function
Gerasimov and Sidorkina [78] carried out experiments of the arc current and duration were derived. For Cu and
with gas mixtures containing SF6 and air in a 500 W Cu-W electrodes the volume of the by-products is given
arc. IR analysis confirmed the formation of SOF4 as
well as S02F2 as the main breakdown products. The v = 0. 043I2; t (with Cu-W) (2)
authors' observations were quite different from many
other investigations which confirmed SOF2 as the main v = 0. 02018t (with Cu) (3)
decomposition product. Manion, Philosophos and Robin-
son [129] studied the stability of some electronegative where v is the absolute volume of the decomposed gas in
gases in low-energy discharges and arcs with copper liters, lis the interruption current in kA rms and t is
electrodes. In an 80 A arc, the formation rates of SF4 the duration in s. The gas, decomposed in the presence
and SOF2 were determined quantitatively by mass spec- of aluminum and zinc electrodes, contained 100 times the
troscopy as a function of arc duration. The SF6 decom- amount of arc by-products as gas decomposed in the pres-
position rates and the dielectric strength of the gap ence of silver or copper electrodes. The primary com-
are shown in Fig. 1. The dielectric withstand of the position of the gases identified in circuit breaker in-
terruptions by mass spectrometer were SF4, SOF2.
Boudene et al. [23] carried out a comprehensive study
of SF6 arc by-products covering formation mechanisms,
detection techniques and toxicities of the stable arcing
by-products. In experiments with arc parameters of 500
A, 60 V and 40 to 80 ms, the formation rates of various
arc by-products as a function of the input arc energy
per unit gas volume were determined. The arc by-product
concentrations were measured by gas chromatographic (GC)
and infrared techniques. A linear relationship was ob-
tained on the production rate of SOF2 and total SF6 de-
composition as a function of arc energy. The formation
of SOF2 in Cu-W electrodeswas determined to be 90x10-9
mol/J while the formation of S02F2 was estimated to be
about 150x10-9 mol/J. Fig. 2 shows the quantities of
SOF2 and S02F2 generated as a function of arc energy per
unit volume for cutene electrodes. The influence of
electrode materials, moisture, and oxygen on the forma-
tion of SF6 arc by-products was also investigated.
Other gaseous by-products including SOF4, S02, HF, and
WF6 were observed but only at trace quantities as com-
pared to SOF2.
500 1000 1500 2000 2500
ARC TIME -
ARC CURRENT PRODUCT (COULOMBS) The influence of moisture in the decomposition process
was investigated by Hirooka et al. [93] who used the ' F
Fig. 1: SF6 decomposition, decomposition products nuclear magnetic resonance (NMR) technique to observe
and dielectric strength as a function of arc time- the formation of arc by-products in a 9.2 kV, 30 kA arc.
arc current products. Conditions: arc current When the test vessel was dried before the arcing experi-
80 A, atmospheric pressure, arc gap 3.2 t129] mm. ment, lower fluorides of sulphur such as SF4, S2F2, SOF2
were detected. However, when the vessel was not dried
before the experiments, the major products were WF6,
test gap remained unchanged even after 40% of the SF6 S0F2, S02F2, CF4, and SF4. The compound Si(CH3)2F2 was
was decomposed into SF4 and SOF2 after arcing. The identified in the wet vessel experiment which the authors
authors also observed the interaction between SF4 and claimed was formed from the reactive arc by-products and
the pyrex wall of the 0.6 1 sample cell forming SiF4 the silicone vacuum grease.
and SOF2.
Ch: 3-F6 decomposition in gs r-in}lateJ e-quipment
compound S02F2 existed in far greater quantities than

SOF2. However, S02F2 was completely absent in arc dis-
charges. No SOF4 was detected in the GC/IR or GC/MS
analysis in either type of discharge. The by-product
formation rate was found to be proportional to I 2t where
4. I is the current and t is the time.
Baker et al. [11] performed high-current arcing ex-
periments in the 30 to 50 kA range to determine the
>0J}
formation of gaseous and solid arcing by-products in
conditions simulating a power arc fault in gas-insulated
E. equipment. In a 15 kA arc for a duration of 133 ms, the
>
level of SOF2 can reach 2% by volume in a 120 1 chamber.
_
E
0
., The authors reported a linear relationship between the

vq -
_
amount of decomposition products formed and the arc
CL ,
energy. Of the seven samples analyzed by GC and GC/MS,
Eo c
only one sample contained any detectable SO2F2 and its
w 0 concentration was 67 times lower than SOF2. The authors
-o observed up to 6000 ppmv of CF4 in a 41 kA experiment
involving polymeric spacers.
Tominaga et al. [181] examined the amount of decompos-
ed gas in an SF6 arc for aluminum, copper and steel
electrodes. The current ranged from 1 to 50 kA which is
typical of short-circuit currents in gas-insulated sub-
stations. In all cases of the high-current arc experi-
ments, the major by-product was SF4 which subsequently
was reported to convert to SOF2. The decomposition
product formation rate at 30 kA was almost 10 times that
of the rate reported by Boudene at 500 A. The author
suggested that the formation rate discrepancies may be
due to the differences in current between the two experi-
ments.
A series of experiments simulating fault arc conditions
were reported by Chu et al. [43,44,45]. The major by-
10 20 products generated in experiments with aluminum elec-
Energie dissipee par Litre de S F6 ( kJ/ t ) trodes and copper electrodes was SOF2. The formation
Dissipated energy by Liter of SF (kJ/L) rates were in the range of 300 to 450x10-9 mol/J.
Fig. 2: Formation of SOF2, 502F2, SOFa4 under the Latour-Slowikowska et al. [117] examined the influence
effect of arcs between,electrodes of cutene and of gas contaminants such as moisture in SF6 arc dis-
cunitene as a function of total energy dissipated charges. Experiments with 1 kA arc current generated
per liter of SF6. 02 content, 4000 ppmv; H20 SOF2 and S02 as the major by-products. Even at high
content 200 ppmv. Curve 1: SOF2, Cutene, Curve moisture contents of 2400 ppmv, the major species de-
2: SOF2, cunitene, Curve 3: S02F2, Curve 4: tected was still SOF2 with SO2 being the minor compound.
SOF4 [23]. No S02F2 or SOF4 was detected.
Ruegsegger et al. [153] employed a molecular beam set-

Bartakova [13] performed several experiments with arc, up to detect the decomposition species from a high-
corona, and dc spark discharges. Analysis of the by- pressure arc discharge in SF6. A 15 kA arc was generated
products was made with the GC technique. The major between graphite electrodes at 1 atm pressure. The in-
products identified in the high-current discharge were situ mass spectrometer detected SF4 and SOF2 in the time-
SF4 and SOF2- of-flight mass detection technique. The recovery time
of SF6 after the discharge was estimated to be in the
Kulseta et al. [109] measured the formation of SOF2 order of ms. The experimental geometry did not allow
in high current arcs (3.7 to 38 kA) for durations rang- enough time for the SF4 to react with moisture outside
ing from 60 to 200 ms. The electrode materials used the arc zone to form SOF2. Therefore, SOF2 could be
were aluminum and copper. The formation rate of alumi- formed in the neighborhood of the arc as a result of the
num electrodes were determined to be between 90 to reactions between moisture in the electrodes and SF4.
500x10-9 mol/J. The formation rate of by-products with
aluminum electrodes was 3 to 5 times higher than the 3.1.2 Discussion of Power Arc Results
rate with copper electrodes. The exothermic reactions
between aluminum and SF6 at high temperature is respon- Table 3 summarizes the gaseous decomposition products
sible for the higher by-product formation rates with identified after high-current arcing had taken place in
aluminum. Significant quantities of CF4 were found in SF6 for various electrode materials. The trend is quite
experiments in which the arc burned near epoxy spacers. clear that lower fluorides of sulphur, SF2 and SF4, were
identified in experiments using IR techniques where in-
Grasselt et al. [80] carried out high and low current situ determination of the by-products was possible. This
arc experiments in SF6. The arc by-products were sepa- observation indicates that these species are the primary
rated by gas chromatograph and the species were identi- products in the decomposition process. With only two
fied by means of IR spectroscopy for S02F2 and mass exceptions [30,78], SOF2 was identified in every investi-
spectrometry for CF4 and SOF2. The authors observed gation. This species also dominates in quantity in all
that in low-current discharges such as a spark, the high-current arc experiments. The formation rate of this
IEEE - t 1 1 - El
r' 2_ J,z.
N
1- t-
TABLE 3 Gaseous Decomposition Products Identified in Power Arcs at Various Laboratories
Decomposition Products Arc Condition Electrode Anal. Tech. Formation Rate Reference
(nmol/J)
SF2, SF4, C02, H2S, HF 8 kA Cu IR Camilli [30]
SF2, SOF2 High Current Cu-Au IR Edelson [64]
ac
S02F2 500 W ac arc ER Garasimov [78]

SOF2, S02F2, SOF 1-3kA ac MS Miyamaoto & Kamatani [161]
SF4, SOF2 80 A dc Cu MS 600 (SF4) Manion [129]
150 (SOF2)
SF4, SOF2, SOF4
S02F2 5-10 A ac Ag GC Becher and Massonne [14]
Al IR
SF4, SOF2, SF2, CF4,
S02F2, SOF4, SiF4 3 kA Cu-W MS 40 (Total) Tokuyama [179]
3 kA Al 130
SOF2, S02F2 SOF4*, HF
WF6 0.5 kA Cu-W GC 90 Boudene [23]
IR (SOF2)
SOF2, S02F2, CF4, SiF4 30 kA
HF, Si(CH3)2F2 (wet cell) Cu-W NMR Hirooka [93]
SF4, S2F2, SOF2, WF6
SiF4, COF2, CF4 (dry cell) Hirooka [93]
SF4, SOF2, S2F2 - GC Bartakova [13]
3.7 kA
SOF2, CF4, AIF3, 6 kA Al GC 100-500(SOF2) Kulseta [109]
3.7 kA Cu 60
Fe 48
SOF2, CF4 High Current (C 1% S9F2 Grasselt [80]
ac IR at 10 (A s)
MS
SOF2, CF4, S02F2, S02 13-49 kA Al GC 650 (SOF2) Baker [11]

GC/MS
SOF2, SF4, CF4 30 kA Al GC 630 (Total) Tominaga [181]

20 kA Cu-W NMR 650
Chu [44,45]
SOF2, SO2, S2F2, CF4 Al GC 600 (SOF2)
Si(CH3)2F2 8-25 kA /MS
SOF2, S02 1 kA GC Latour-Slowikowska [115]
SF4, SOF2 15 kA Graphite MS Ruegsegger [153]
SF4, SOF2, CF4, SOF4, 30-60 kA Al GC 8-32% (SOF2) Pettinga, Janssen
S02F2, S2F10, HF, 500 V MS 0.8-4.3% [147,103]
S02, H2S, S2F100 (Enrichment) (CF4)
SF4, SF2 1.8 kA Cu-W GC Suzuki [172]
molecule has been obtained in 6 investigations. Fig. 3 At low currents, the formation rate of SOF2 depends on
shows the quantities of SOF2 formed as a function of electrode materials with a higher formation rate asso-
arc energy for aluminum electrodes. Despite the scatter- ciated with aluminum electrodes. Other species such as
ed valued, the rate is about 600xlO-9 mol/J. There ap- S02F2, SOF4, Si(CH3)2F2 are found in trace quantities
pears to be a tendency that the formation rate depends as compared to SOF2. CF4 is formed when arcing is near
on arc current. The differences between Boudene's organic materials such as polymeric spacers. S02 is
value for Cu-W electrode [23] and Tominaga's work [181] also detected when SOF2 is present in significant
tend to support the theory that the formation rate is quantities. However, the quantity of S02 is always
current dependent.
Chu; SF6 decompositionian ggas-i.rtssulated euipm'-em-
A9.
I'U-
* ONTARIO HYDRO
A KULSETAS ET AL 0
A COULD-BROWN BOVIER -H
o TOMINAGA ET AL 4-)
w ~~~~~3000 A
SOF2 FORMATION RATES A 10-
to4 FOR
Al-Al ELECTRODES U)a) 5
0.0
-
z * A/
0
I o'H
fl 0.05
z
w a) N . I
u
z
0
0
U.
VIL
0
0 <
- ND 0
Nfn E LL <
O33
Oz< 0
Electrode Material
Fig. 4: The reZation between eZectrode consunption
and the electrode material for power arc in SF6
[179].
102LDi I
0.1 1.0
I
10
Reaction mechanisms between the metal and the SF6 were
0.
ARC ENERGY PER LITRE OF SFi (kJ/L)
proposed.
Boudene identified tungsten sulphide as the major
Fig. 3: Formation of SOF2 a function of as arc solid by-product in experiments with copper tungsten
energy for aluminum electrodes 144]. electrodes [23]. Tungsten hexafluoride which was also
formed during the experiments, existed in gaseous form
and was identified in the gaseous stream. Hirooka ob-
lower that SOF2 and it is suspected that S02 is formed served the formation of CuF2, CuOHF, CuF2*2H20, and
as a hydrolysis product of SOF2 and moisture. CuF2*CuOHF in 30 kA arcing experiment involving copper
electrodes [93]. Of these compounds, only CuF2 is be-
The results of prior work with power arcs can be sum- lieved to be formed during arcing, and the remainder
marized as follows: were probably formed after the powder was removed from
the test vessel by absorbing moisture from the ambient
(1) SOF2 is the major stable decomposition product air. Similar phenomena in experiments using copper and
formed in an SF6 power arc. iron electrodes were observed [115]. Many hydrated
phases of CuF2 and FeF3 were identified by x-ray dif-
(2) There is ample evidence to indicate that SOF2 is fraction analysis.
generated from the primary products SF4 and SF2.
Aluminum trifluoride AlF3 is the major solid arc by-
(3) The formation rate of SOF2 is about 600 nmol/J for product found in most GIS power arc faults. Kulseta
aluminum electrodes. The rate appears to decrease for and Chu have determined the formation rates of the com-
low arc current and copper alloy electrodes. pounds under high-current arcing conditions [39,45,109].
The exothermic reaction between the aluminum metal and
3.1.3 Formation of SoZid Arcing By-products in a SF6 was responsible for the increase in yield in alumi-
Power Arc num fluoride formation. Among the reactions between
various metals with SF6, the aluminum-SF6 reaction has
In addition to gaseous decomposition products, solid the highest heat of formation [107]. The solid by-
arcing by-products have been found in circuit breakers products generated in SF6 arcs exist in a finQ powder
or in faulted equipment [23,45,69,70,145]. After a 30 form having size distributions ranging from sub-pm to
kA, 3 cycle arc in an aluminum enclosure, Fukada found ,um sizes [39,40].
about 20 ml of white powder near the arc location. The
powder was identified as aluminum trifluoride [75]. The results of the solid by-product investigation are
Tokuyama [179] examined the erosion rates of 15 metals summarized in Table 4. In all cases the major by-prod-
which had experienced a 3000 A arc in SF6. The results uct formed is solid metal fluoride with the exception
are shown in Fig. 4. The major solid arcing by-products of tungsten electrodes where the metal fluoride WF6 is
identified in the electrode erosion experiments were a gas. This explains the observation that in circuit
metal fluorides. The authors suggested that the re- breaker chambers with Cu-W alloy arcing contacts, the
action rates were related to the oxidation and reduction amount of solid by-products is less than the amount
potential of the various metals. Vijh [191] reanalyzed formed in a fault arc normally supported by the alumi-
Tokuyama's data and concluded that the metal was attack- num conductor and enclosure. Copper sulphide has been
ed by the sulphur instead of the fluorine atoms. identified but in much lower quantities than metal
Elert-I+-inVl
Solid*Decomposon
TABLE4r- on-
TE-EE Tr-nsasctiom- -trat i'n. E-1N.J
Pout Gene m ccbe 1-9L
Er- co-. 0t-ber
TABLE 4 Solid DecompWton Products GeneTated in PoweT Arc
Solid Decomposition
Byproducts Arc Condition Electrode Anal Tech. Formation Rate Reference
AlF3 22-37 kA Al XRD 20 c.c. Fukada [75]

60 ms ac
FeF3, CuS, SnF Steel XRD Fukada [179]
AlF3 3000 A Al XRD 5 g/kAs Tokuyama [179]

CuF2, WF6 (gas) 3000 A Cu-W XRD 0.5 g/kAs Tokuyama [179]
Metal Sulphide and 500 A Cutene 3.6 mg/kJ Boudene [23]
Fluoride, v'VF6 (gas)
Metal Sulphide and 500 A Cunitene 1.9 mg/kJ Boudene [23]
Fluoride, WF6
CuF2, CuOHF, CuF2-CuOHF, 30 kA Cu-W XRD Hirooka [93]
CuF2-2H20
AlF3 3-25 kA Al XRD 50-100 mg/kJ Chu [45]
CuF2, CuS 3-25 kA Cu XRD 25 mg/kJ
AIF3 3.7-38 kA Al XRD 30-100 mg/kJ Kulseta [109]

Copper Fluoride 3.7-22 kA Cu 20 mg/kJ
Iron Fluoride 3.7 kA Fe 20 mg/kJ
CuF2, VF3, S 1 kA Cu XRD Latour-Slowikowska [115]

CuF2-H2, CuSiF64H2O
FeF3, FeF2, FeF3-3H20 1 kA Fe XRD Latour-Slowikowska [115]
AIF3 1 kA Al XRD Latour-Slowikowska [115]
A1F3 (anhydrous) 16.4 kA Al Xi Baker [11]

EL
CuF2-2H20 15 kA Cu Xi Baker [11]
fluoride. There is little information on the quantities place. Manion et al. [129] carried out spark experi-
of powder formed as a function of arc energy. The work ments in a 240 ml Pyrex cell with copper electrodes.
of Chu and Kulseta [39,109] gives a formation rate of The spark current was 0.75 A. The spark decomposition
50 to 100 mg/kJ for aluminum electrodes. The formation products formed were identified as SOF2, SiF4, and
rate depends on arc motion. For a stationary arc, the S2F2. No S02F2 was detected. The presence of silica
formation rate is about twice that of a moving arc in a in the glass cell was responsible for the formation of
coaxial geometry. Converting the results of Chu and SiF4 and SOF2 from SF4.
Kulseta yield a value of 1000 nmol/J for AlF3 and about
200 nmol/J for CuF2. Ermel [66] detected the formation of NOx, NF3, and
N2F2 in a 10 mA spark discharge containing 20% SF6 and
3.2.1 Spark Decomposition-of SF6 80% nitrogen. Bartakova [13] measured the decomposition
rate of SF6 in a 25 hA spark discharge to be 15 nmol/J
While the distinction between an arc, spark and which is eight times less than the decomposition rate
corona in operating GIS is obvious, the classification of SF6 in an arc. In addition to SF4 and SOF2, S2F10
of spark in this review is quite arbitrary. Most authors was detected by the gas chromatographic technique.
did not give sufficient information for the distinction
to be made. In this section, all prior SF6 decomposition Becher and Massonne [14] examined the spark decomposi-
work carried out in arc currents less than 1 A and duration products in a capacitive type of discharge in SF6.
tions in the ,us range for each discharge, is classified The spark energy was estimated to be 0.3 J with a pulse
as a spark. Thorburn [178] observed an increase in F+ repetition rate of 100 Hz. In contrast to the arc ex-
current in a mass spectrometer when he subjected SF6 to periment which the authors carried out with the same
spark discharge in a 30 ml cell. The discharge condi- cell, significant quanitites of S02F2 and SOF4 were de-
tions were 60 mA for 2 h for point-to-point tungsten tected in the spark which were completely absent in arc
electrodes. Traces of SiF4+, SiF3+, S02', and SF2+ were discharges. The authors reported the observation of
found. The author concluded that permanent molecular S2FlO and S2FlO by IR and enrichment GC techniques.
dissociation as a result of the discharge had taken Quantities up to 150 ppmv of S2FlO and S2FlO were
Chu: SF6 decofmipOsi in gas-iflWnsulated a-2lUipTe7m-t
detected after 30 min of discharge in the 2 1 cell. breakdown products as a function of time following the
The maximum S2FlO and S2FlOO formation rate is esti- discharge. Both positive ions and negative ions were
mated to be 0.2 nmol/J. This formation rate is about detected as a result of sparking in SF6. However, the
a factor of 80 less than the reported values of SOF2 ion chemistry of SF6 discharge is strongly dependent on
and S02F2 generated in sparks. impurities originating from the gas, the chamber and
the buildup of decomposition products. In pure SF6
Chu [44] investigated the formation of SF6 decomposi- (99.9%), the only important positive ion appears to be
tion products during disconnect switching, an operation S2F7+ which may act as a precursor to the other ions
which consisted of hundreds of restrikes or short- observed in dirty SF6. At lower pressure in pure SF6
duration sparks across the contacts as they separated. discharges, the vast majority of ions are deficient in
Both SOF2 and S02F2 were found in comparable quantities. fluorine when compared with the 6:1 atomic ratio in SF6.
The formation rate of SOF2 is estimated to be 5 nmol/J Ions such as S8g have been observed. The negative ion
for copper electrodes. Although the instanteous cur- chemistry appeared to be dominated by ions containing
rent in such discharge is very high, in the order of the electrode materials.
10 kA, the average current is only 250 mA. Copper
fluoride was found as a solid by-product after many The neutral decomposition products of spark SF6 have
switching operations. The formation rate of the solid also been studied by the Oak Ridge National Laboratory.
by-product was found to be 10 mg/kJ. The mass spectrometer was modified to detect SF4 after
the spark. Ordinary GC cannot distinguish between
Castonguay [33] reported a total SF6 decomposition SOF2 and SF4 because they elute at the same time. For
rate in a 60 W spark employing stainless steel elec- the tungsten electrode geometry, the formation rate of
trode to be about 7 to 8 nmol/J. This value is about SF4 was determined to be 500 nmol/J. This value is
3 times less than the values obtained by Boudene [23] much higher than the formation rate of SOF2 obtained
on arcs supported by stainless steel electrodes. from direct detection of spark products in a stainless
steel electrode spark cell. The formation rate of S0F2
An extensive study on spark decomposition of SF6 has with stainless steel was determined to be 1.8 nmol/J.
been carried out at the Oak Ridge National Laboratory In addition to SOF2, SOF4, S02F2, and SiF4 were detect-
by Sauers et al. [71,158-161]. Using a highly sophis- ed. The formation rates of these by-products are given
ticated spark source and quadruple mass spectrometer, in Table 5 for comparison purposes. The presence of
the group was able to study ion formation in sparks. SiF4 is due to the ceramic insulator used in the spark
The sparks were generated in a tungsten electrode cell. The characteristics of the spark used in the
chamber at pressures ranging from 20 to 60 kPa. Sparks neutral decomposition experiments were 2 ns pulses, with
of 0.1 mJ energy were initiated across the electrodes 50 V and 1 A as arc voltage and current respectively.
by HV breakdown. The ions generated were directed The authors observed that while the formation rates of
toward the high vacuum region for analysis by a quad- SOF2 and SiF4 exhibit pressure or energy dependence, the
ruple spectrometer. Fig. 5 shows the evolution of formation rates of S02F2 and SOF4 are pressure
independent.
I VERY LITTLE DATA ON PRECU RRSORS MOST DATA

ACQUIRED HERE
COfIPLE TE SULFUR-
OISSOCIATION F LUORIOES
I LONG LIVED BY-PRODUCTS
IONS I
NEUTRALS I SOF2
SxFyt
SO F4,
e.g. THE NEUTRAL SOF
S. S', S'o* . . . PRODUCTS F SF6 1 REACTIONS WITH
F. F' ,f . . . SF4 SF6. IIMPURITIES. SiF4
(F-). e OTHER BY-PRODUCTS CF4
2
30.000
H; .lI
HF
'.j
x-
UFJ
300
a 10-3 l0, 104

TIME (s)
Fig. 5: Schematic representation of the evoZution of breakdown products as a function of time foZZowing the
spark discharge; time regions for which data exist and exconpZes of species in the various regions [161].
the exception of the tungsten case, the copper and
TABLE 5 Yields of By-Products Formed in Sparked SF6 [1611 stainless steel formation rates are in the order of 2
to 5 nmol/J. The third difference is in the detection
of S2F10 and S2F100 in a spark. With the exception of
Pettinga's work [147] (Section 3.1.1) which detected
trace amount of S2F10, there was no report of its
By-Product G (nmole/J) formation in power arc discharges. In a spark, the de-
tection of S2Fl0 appears to be unambiguous and the
quantities are substantial. The by-product formation
SOF, 0.9 - 2.7a mechanisms in sparked SF6 will be reviewed in Section
SOFP 0.22 0.09b 4.3.
SiT 0.02 0.11a
SOF2, 0.016 0.010b 3.3.1 Decomposition in Corona or PartiaZ Discharge
SO2 -0.005
The first observation of SF6 decomposition in corona
was reported by Schumb et al. [165] who exposed SF6 in
a glass cell to long-term corona discharges with cur-
aMaximum yield over the pressure range 230-40 kPa. rents between 50 and 200 IA. After an exposure time of
200 h, about 30% of the SF6 had been decomposed into an
bBased on a least-squares fit to a straight line for the acidic material, easily absorbed by potassium hydroxide
solution. It was suggested that lower fluorides such as
yield versus total spark energy.
SF4, SF2, or S2F2 were formed during the discharge.
3.2.2 Discussion of Spark Resul ts Continuous corona discharge in electrostatic accel-
erators insulated with SF6 was investigated by Ashbaugh
et al. [9] to find ways to minimize the corrosion ef-
Table 6 summarizes the spark decomposition of SF6 fects on the accelerator's components. A corona dis-
from eight previous works. While SOF2 is still the charge was maintained between strips of stainless steel
major by-product observed, the quantities of S02F2 ap- screening and a flat copper plate. Materials in con-
pear to increase in comparison with the results from tact with the discharge included Lucite, mild steel,
arc discharges. Furthermore, SOF4 is detected. The polyethylene, aluminum and nylon. The vrarious ions
other major difference between decomposition in spark formed during corona discharge were identified by mass
and arc is that the spark decomposition rate appears spectroscopy. Positive ion peaks from S02F2', S0F2',
to be much lower than that of an arc discharge. With and SOF3' were observed. The authors concluded that
TABLE 6 SF6 Decomposition Products Identified in Sparks at Various Laboratories
Spark Decomposition Spark Formation Rate

Byproducts Condition Electrode Anal. Tech. (nmol/J) Reference
F+, SiF4+, SiF3+ 0.4 mA Tungsten MS Thorburn [178]
S02+, SF2+ 1 Hz
SOF2, SiP4, S2F2, CuF2 750 mA Copper MS 18% SOF2 after Manion [129]
1250 coulomb
NO, NOX, NF3, N2F2 10 mA ac MS Ermel [66]
80%/o N2
20%o SF6
SF4, SOF2, S02F2, SOF4 200 mA ac Al GC/MS Castonguay [33]
SOF2, S02F2 250 mA Cu GC/MS 5 (Total) Chu [44]
CuF2 10 mg/kJ
SF4, SOF2, SOF4. 5 J/spark Stainless Steel MS 1.8 (SOF2) Sauers [161]
S02F2, SiT4, SO2, VY6 0.2 mJ/spark w MS 500
SF4, SOF2, S02F2, SOF4 25 ,uA GC 15 (Total) Bartakova [13]
S2F0l S2F
SF4, SOF2, SOF4 12 kV Copper GC,IR Becker [14]
S02F2, S2F10 0.3 J/spark
S2F100 100 Hz
SOF2, SOF4, S02F2 12 kV Al GC, IR Becker [14]
S2FPo, S2F10O 0.3 J/spark
100 Hz
Chu: SFA- ec- p iGc i 'g Me,
i -l U. Ed C i!2p'`
S02F2, SOF2, and SOF4 were present in the gas which had TABLE 7 Limiting Constant Values for SOF2
been exposed to the corona discharge. The mass peaks Production Rates [188]
of SO2F2 appeared to be stronger than the SOF2 peak
suggesting that the concentration of SO2F2 was higher
than that of SOF2. Discharge Discharge Gas Production
Polarity Power Current Pressure Rates
(W) (HLA) (kPa) (nmoles/J) (Q.moles/C)
Waddington [193] generated a corona discharge in an Pos. 0.804 20.0 116 1.28 49.5
SF6 filled vessel simulating conditions in an SF6-in- 0.054 1.5 200 5.20 181
sulated current transformer. The energy of the dis- 0.335 8.0 200 4.69 195
charge across the point-to-plane electrode system was 0.777 16.0 200 3.80 180
estimated to be 3 pi per pulse. The IR analysis in- 0.198 4.0 300 4.06 187
dicated that at low energy discharges S02F2 was the 0.430 8.0 300 3.09 178
0.945 16.0 300 2.64 172
dominant product formed, and as the energy of the dis-
charge was increased, SOF2 was also identified. The Neg. 0.898 25.0 144 1.21 45.0
types of decomposition products observed in corona ap- 0.230 8.0 200 1.52 35.2
peared to agree with Ashbaugh's result [9], but dis- 0.586
0.821
16.0
20.0
200
200
0.93
0.68
34.7
34.1
agreed with the results of Waddington's own power arc 2.225 40.0 200 0.54 32.4
experiments. The dominant products generated in the 4.290 64.0 200 0.44 33.7
power arc discharge were SF4 and S2F2. The formation 0.764 16.0 300 0.73 31.2
of the decomposition products was estimated to be 50 1.140 21.0 300 0.41 23.9
ml at STP for 1 kJ of corona discharge energy. The 1.820 30.0 300 0.33 22.3
formation rate did not vary with the energy of the
single discharge.
Bartakova [13] investigated the formation of by- TABLE 8 Limiting Constant Values for SO2F2
products in corona and partial discharge. The species Production Rates [188]
identified included SF4, SOF2, SOF4, S2FlO, and S02*
The corona discharge was generated in a point-to-plane
geometry with a current of 25 pA. The decomposition Discharge Discharge Gas Production
Polarity Power Current Pressure Rates
rate was estimated as 0.07 ml in 1 kJ of discharge (W) (HxA) (kPa) (nmoles/J) (JLmoles/C)
energy. The formation rate of individual components Pos. 0.804 20.0 116 1.83 69.7
was not determined. The decomposition products of par- 0.054 1.5 200 4.51 151
tial discharge were very similar to that of the corona 0.154 4.0 200 4.07 158
discharge except that SO2F2, SOF4, and SO2 were not 0.335 8.0 200 3.27 147
identified. No explanation was given for the absence 0.777 16.0 200 2.10 109
of these species in a discharge which is very similar 0.198 4.0 300 2.32 119
0.430 8.0 300 2.03 113
to corona. 0.945 16.0 300 1.29 111
The most extensive investigation of corona discharge Neg. 0.898 25.0 144 0.52 18.0
was carried out by Van Brunt and his co-workers at the 0.230 8.0 200 0.62 16.0
0.586 16.0 200 0.47 16.9
National Bureau of Standards [166,167,184-188]. The 0.821 20.0 200 0.36 14.8
discharge used for the investigation was a highly lo- 2.215 40.0 200 0.25 13.9
calized, low current, point plane corona. The zone of 4.290 64.0 200 0.16 12.6
ionization was confined to the immediate vicinity of 0.764 16.0 300 0.38 17.9
the point electrode tip due to the highly divergent 1.140 21.0 300 0.29 16.9
field. The extent of the active region and the ion
drift volume were therefore small compared to the rel-
atively inactive volume of gas contained in the dis-
charge cell. Typical dc positive and negative corona TABLE 9 Limiting Constant Values for SOF4
currents ranged from 1.5 to 64 pA with the correspond- Production Rates [188]
ing power dissipation from 0.054 to 4.3 W. The by-
products generated from the corona in the stainless
steel electrode tip were analyzed with a GC/MS attached Discharge Discharge Gas Production
to the discharge cell. The most abundant, long lived Polarity Power Current Pressure Rates
stable gaseous species generated by corona discharge in (W) (pA) (kPa) (nmoles/J) (,umoles/C)
SF6 gas containing trace levels Of 02 and H20 are the Pos. 0.804 20.0 116 7.02 290
oxyfluorides SOF2, SO2F2, and SOF4. Absolute energy 0.054 1.5 200 7.04 260
and charge rates-of-production of these and minor prod- 0.335 8.0 200 8.60 373
ucts S02, OCS and C02 have been measured at pressure 0.777 16.0 200 8.45 418
levels from 100 to 300 kPa and for discharges of dif- 0.198 4.0 300 5.79 292
ferent current, power, and polarity. Tables 7, 8 and 9 0.430 8.0 300 5.92 343
0.945 16.0 300 7.08 431
show respectively the production rates for SOF2, S02F2,
and SOF4.for various experimental parameters. The most
interesting observation from the data in Tables 7-9 is Neg. 0.898 25.0 144 0.85 30.1
that the production rates for SOF2, SO2F2, and SOF4 are 0.230 8.0 200 0.92 23.8
of comparable magnitude-to within roughly a factor of 0.586 16.0 200 0.97 34.6
0.821 20.0 200 1.05 41.5
3. At higher pressures, the production rate of SOF4 is 2.215 40.0 200 0.90 49.7
higher than that of the other two species. The second 4.290 64.0 200 0.94 63.6
interesting observation is that the oxyfluoride produc- 0.764 16.0 300 0.44 20.6
tion rates under most circumstances increase by a factor 1.140 21.0 300 0.39 21.0
of 5 or more in going from the negative to positive
polarities. Despite the sensitivity of Van Brunt's
detection system, the author did not report any obser-
vations of S2FlO or S2FloO in the GC/MS analysis.
r~rv=- Electricei e I _ I-21 No. F5-tobei1- 5
Based on the massive experimental data, Van Brunt et

al. concluded that oxyfluoride production is not con-
trolled by the concentration of either 02 or H20 at LLJ
levels below about 1%, and the rate controlling factor C]
is the dissociation rate of SF6 in the discharge. In l
separate experiments, Van Brunt et al. [166] investi- 0
gated the decomposition of SF6 in the presence of trace
levels of N2. In addition to the major oxyfluorides U-
observed in SF6-02 mixtures, NO, and N02 were also in- Ui- I
observed. Cr
In an experiment similar to Van Brunt's point-to-plane
UJ 0
set-up, Chu et al. [44] determined the formation rate E
of SOF2 and S02F2 in an ac point-to-plane corona with N
aluminum electrodes. The 60 Hz corona has a discharge 0-
c3.
level of 103 pC and a repetition rate of 2 kHz. The
concentrations of SOF2 and S02F2 were very similar. 0 0.1
The total production rate of SOF2 and S02F2 from the
measurements were estimated to lie between 0.6 and 1.4 C)
nmol/J. Van Brunt's experiment with a dc stainless I
steel electrode yield a total production rate
(SOF2+SO2F2+SOF4) of 0.59 nmol/J. The higher produc- 0.01
tion rate observed by Chu may be due to aluminum which
is known to yield a higher formation rate in arc dis- I 10 100 1000
charges. Chu did not observe SOF4 in the GC/MS analy-
sis. This may be due to the conversion of SOF4 to TIME (h)
S02F2 during the sampling process. Fig. 6: ReZationship between concentration of
decomposed gas and duration of partial discharge
The interaction of SF6 decomposition products with [181].
H20 generates hydrogen fluoride, HF. Kusumoto [111]
determined the HF levels from a simulated partial dis-
charge experiment with a point-to-plane electrode geom- that found in arcs and to a certain extent, in sparks.
etry. HF was analyzed quantitatively by passing the This appears to be a unique characteristic of corona or
gas through an alkaline solution and the quantity of F- low-energy discharges. The pattern appears that as the
ions was measured by an absorptiometric technique. The discharge temperatue decreases (e.g. from arc to
results indicated that the amount of HF is independent corona), the concentration of S02F2 increases. S2F10
of pressure and electrode materials, but dependent on was only identified in one investigation while SOF4 has
the charge transport. The author determined the re- been identified in three laboratories. Van Brunt actu-
lationship between apparent charge x and the quantity ally observed higher concentrations of SOF4 than of the
of decomposed gases y(l0-8 ml as HF) is expressed by other two oxyfluorides. The decomposition rate given
the following equation: by Waddington is much higher than that given by
Bartakova and Van Brunt, and this rate approaches that
log(y) = 1.5 log(x) -A, A - 5.8 to 6.8 (4) of power arc conditions. The validity of Waddington's
measurements is in doubt because of the analysis techni-
In an electrode geometry in the presence of epoxy mate- que. The total formation rates of Chu, Van Brunt, and
rials, a 50 fold increase in the quantity of the decom- Kusumoto are within an order of magnitude in spite of
posed gas was observed. No attempt was made to deter- the different electrode materials and diversified ex-
mine the individual components of the arc by-products. perimental conditions. Since not all of the oxyfluo-
rides may convert to HF, the value given by Kusumoto is
The results of Kusumoto were used by Tominaga [181] to very similar to Van Brunt's value. The higher formation
estimate the amount of by-products generated by partial rate observed by Chu is probably due to the effect of
discharge in a full size GIS. Using a dynamic model to the electrode material. The estimated formation rate
calculate the amount of by-products produced in partial of the total sulphur oxyfluoride in corona appears to
discharge pulses with 1000 pC or 10000 pC levels at a be in the neighborhood of 600 pmol/C.
repetition rate of 1000 pulses per second and the re-
moval of by-products by a layer of absorbents, the re- 3.4 Thermal Decomposition of SF6
lationship between the concentration of decomposed gases
and the duration of partial discharge was obtained. The thermal and chemical stability of SF6 at normal
Fig. 6 shows the concentrations of the hydrolyzable operating temperature of the gas insulated equipment is
fluorides as a function of time for the two partial dis- an important issue because it affects the long-term
charge levels. After a period of 100 h, the concentra- relability and aging of the equipment. In comparison
tion of HF starts to level off because of absorption by to the electrical decomposition, thermal decomposition
the desiccants. In the worst case (0.5 m3 volume and of SF6 has received relatively little attention. The
104 pC partial discharge), the final level at equilib- early work of Camilli [30] showed that the critical
rium is about 10 ppm by wt of HF in SF6 which is equiv- temperature of SF6 was 6000C, above which SF6 was ob-
alent to 80 ppm of HF by volume. This level is easily served to decompose into SF4 and SF2. However, this
detectable by most analytical methods. value of 6000C critical temperature was disputed by
subsequent investigations which showed that at tempera-
3.3.2 Discussion of Corona Results tures as low as 1500C there is evidence of SF6 reaction
with silicon steel plates [94]. At temperatures above
Table 10 summarizes the results on corona decomposi- 2000C, many metals start to react with SF6 forming
tion of SF6 from eight inrvestigations. While SOF2 is metal fluorides and sulphides [36a,65,94]. In addition
still the dominant species formed in the decomposition to the formation of fluorides and sulphides on the
products, the concentration of S02F2 is much higher than metal enclosure's surface, gaseous decomposition products
Chu: SF6 decomposition in -as-ensulate- 2,L pme,5
TABLE 10 SF6 Decomposition Products Identified in Corona
Corona Decomposition Discharge Formation Rate

Byproducts Condition Electrode AnaL Tech. ,moles/C Reference
SF4, SF2 or S2F2 50-200 ,uA S.S. point Wet Chemistry Schumb [165]
Aluminum
plate
SOF2, S02F2, SOF4 Continuous S.S. copper MS Ashbaugh [9]
Corona
S02F2 (low energy discharge) 3 g.J/pulse IR 2000 nmol/J Waddington [193]
S02F2, SOF2 point to plane (total)
SF4, SOF2, S02F2 25 ,uA Point plane GC 1-3 nmol/J Bartakova [13]
SOF4, S2F2, S2F10 (total)
(corona)
SF4, SOF2, S2F2
(partial discharge)
SOF2, S02F2, SOF4, 1.5 ,tA S.S. GC/MS 180 (SOF2) Van Brunt [1881
S02, OCS, CO2 151 (So2F2)
260 (SOF4)
SOF2, SO 2F2 1.5 /iA ac Aluminum GC/MS 640-1400 Chu [44]
(total)
HF 104-105 S.S. Wet 220 Kusomoto [11]
pC Copper Chemistry (Total HF)
such as S02 were also identified. The influence of Between 700 and 9000C, Padma and Murthy [144] observ-
moisture on the formation of S02 was found to be signi- ed that SF6 decomposed to SF4 in stainless steel, copper
ficant [94]. Table 11 shows the stability of SF6 gas and quartz tubes containing SF6. In the case of quartz
in various materials of construction at 2000C and 2500C tubes, SF4 was observed to react with SiO2 to form SiF4
[65]. When SF6 is contained in a mild steel enclosure and SOF2. In copper and stainless steel tubes SF4 was
at 2000C, up to 2000 ppmv of SF6 can be decomposed in a the principal decomposition product and no major differ-
year. ences were found between the formation rate of SF4 in
these two metal tubes. Padma and Murthy's results ap-
Table 11 Stability of SF6 in Various Materials pear to confirm the thermodynamic calculation of Wilkins
of Construction [65] [196,197] that at temperatures up to 15000C, SF6 dis-
sociates principally to SF4. The subsequent conversion
of SF4 to other oxyfluorides will be governed by the
Dopositon, %/yr presence of impurities such as oxygen and H20 in the
Materials 20(YC 2S(0C gas and on the metal surfaces.
The thermal and chemical stability of SF6 in the pres-
Aluminum 0.006 ence of polymeric insulating materials was examined by
Copper 0.18 1.4 Dakin [54,201] to estimate the long-term service life
Silicon Steel 0.005 0.01 of various types of dielectric gases. The results of
Mild Steel 0.2 2 accelerated aging tests showed that in the presence of
Hydantoin epoxy resin, a material used as an insulator
in SF6 insulated equipment, about 1.3 and 5.5% of SF6
*Estimated were decomposed when the gas, resin, and aluminum en-
closure were heated to 200 and 2250C respectively for
28 days. At 1750C temperature and 97 day exposure, the
decomposition is less than.1%. Using the limited time
temperature data for SF6 decomposition, Dakin was able
The decomposition rate of SF6 increases markedly as to estimate from the Arrenhius relationship that at
temperature increases. At temperatures of 6500C, Chu 1400C it will take 10000 days to decompose 5% of SF6.
and Massey [42] reported the formation of SOF2, S02F2, Fig. 7 shows the temperature dependence for time to
and S02 when aluminum and copper were exposed to SF6 reach 5% decomposition. After the epoxy resin was ex-
for 90 h. No by-products were found in the case of posed to SF6 at 2000C for 28 days, the epoxy changed to
stainless steel. In aluminum, only S02F2 was found, a darker color. It should be pointed out that the SF6
while in copper about 37000 ppmv of S02 and 2700 ppmv decomposition data of Dakin was obtained with a gas
of SOF2 were identified. The observation of large volume to resin ratio lower than that encountered in
amounts of S02 is consistent with previously reported normal GIS equipment. Dakin showed that at higher gas
observations of S02 as the main decomposition product volume to resin ratio, the decomposition rate at 2250C
[94]. The surface-adsorbed H20 is suspected to play an for 28 days' exposure was less than 0.5%, thus demon-
important role in the production process. strating the excellent thermal and chemical stability
c10~ %Los-
10000
8000
6000
4000
2000
1000
800
600
400
80
20
10
6 _<
100 110 120130140150160 180 200

Temperature, C
Fig. 7: Temperature dependenee of time to reaeh

C
12
.- S-- - - - ::. - -
10% LOSS
220 240 2602

:L .r-- - n::
CF3NL2
=I
-
r,
300
_
-- 'I
- -
.- -. .- 4
.
. '.
- .1-
'T z
-- -
. -;r-
!:-.
.-, I-Lt-1- "- --I .r-.
SOF2 and S02F2 normally found in other types of decom-

positions. Instead COS was detected in 0.1% to 6.5%
SF6/air mixture experiments. The use of open flames
such as cutting torches or propane heaters in an at-
mosphere containing SF6 was simulated and only SOF2 was
identified.
3.5 rf & Photo Dissociation
Rf-decomposition of SF6 can occur in SF6-insulated rf
zables or in low-pressure plasma generators for plasma
etching. Bruno et al. [29] studied the decomposition
of SF6 under RF discharges at 3 kPa as a function of
power densities and gas flow rates. The discharge was
carried out in a quartz cell and the decomposition prod-
ucts were qualitatively and quantitatively analyzed by
IR and GC techniques. For certain power densities and
flow rates, almost a 100% conversion was achieved. A
typical distribution of the products, in terms of molar
fraction, is 0.06 SF6, 0.33 SiF4, 0.27 SOF4, 0.16
S02F2, 0.10 F2, and 0.07 SOF2. The presence of SiF4 as
a major reactant product is a clear indication of the
high concentration of atomic fluorine and/or SFX radi-
cals which rapidly react with the quartz reactor walls.
In a related experiment, the authors observed the mark-
ed increase in the IR absorption lines of S02F2 after
the introduction of moisture into the discharge, while
there was a corresponding drop of SOF4 intensity in the
spectrum. The observations indicated that SOF4 can be
hydrolyzed in the discharge by the presence of H20 to
r-. .
-t
ti
-,
--
-1 --
! "41 O ..
- =
f7;L C
- Ir
-'
-
q---. ;--. L--- - I ;D C:- .'
57 deeompoeition of SF6 [54 ]. form S02F2.

Edelson [63] observed the formation of S2F2 in an
of SF6 at normal operating temperature of under 1 00C . electrodeless discharge. No other products were re-
It should be noted that Dakin 's results were obtained ported by the author. Dielbold and McFadden [55] re-
only in the presence of thermal stress without any ported the formation of sulphur in an electron para-
electrical field. magnetic resonance experiment after SF6 was introduced
to suppress the noise of the equipment.
Other practical aspects of thermal decomposition of
SF6 have been examined by Chu [42] to include situations Agustino and Flamm [3] studied the application of RF
normally encountered in gas-insulated substations. In discharge of SF6 and SF6-02 mixture for the etching of
indoor or underground SF6 insulated substations, there silicon substrates. The authors reported that within
is always the concern that an accumulation of SF6 from the detection limits, RF discharge in pure SF6 without
s low equipment leakage may be decomposed during normal Si or SiO2 substrates yielded no stable reaction prod-
substation activities such as arc welding and propane ucts. With a quartz discharge tube and silicon sub-
torch heating. Dissociation of SF6 under arc welding, strates, SiF4 and F2 were observed and other sulphur
propane heating and cigarette smoking conditions was in- oxyfluorides such as S02F2, SOF4, and SOF2 were also
vestigated by laboratory experiments. The results show- identified. However, if an alumina tube was used in
ed that in gas-metal arc welding, the argon gas shields the discharge cell, no oxyfluoride was detected. The
the arc to minimize the direct reaction between the SF6 author studied the addition Of 02 into the discharge
and the arc. No by-products were detected in SF6 con- and concluded that as the 02 flow rate increases, the
centrations less than 3000 p1/1 in air. For typical formation of sulphur oxyfluorides also increases. This
welding conditions of 25 Vdc, 130 A and 45 s duration, observation of oxyfluoride increase as a function of
SOF1 formation in the range of 10 ,u1/1 was detected in oxygen concentration has been confirmed by Picard and
5000 p1/1 of SF6 in ambient air. In addition to SOP2, Turban [149]. Fig. 8 shows the variation of molar
S and S02F2 were also identified. Since the gas-metal fractions of neutral secondary products synthesized in
arc was shielded with argon, the decomposition of SF6 is a SF6-02 RF plasma in the presence of silicon. At low
mainly from the interaction between the hot aluminum 02 concentrations, the dominant product is SF4 and as
metal and SF6. The decomposition rate of SF6 in the the oxygen increases, various species of oxyfluorides
vicinity of the arc column could be accelerated by a also increase. The role of the oxygen is to react with
reaction between SF6 and the ozone generated during the fluorosulphur radicals, preventing them from recom-
welding. Wray and Feldman [202, 203] studied the oxida- bining to form SF6. As the feed rate of oxygen is in-
tion of SF6 by 2 and 0 at high temperature by shock creased, S02F2 increases with respect to SOF4 while the
tube studies of 5F6+O2 and 5F6+03. SF6 is not attacked F2 concentration increases first, then decreases. The
by 2 and 0 atoms even at high temperatures, but one of wealth of information which can be obtained from an RF
the SF6 pyrolysis by-products is rapidly attacked pro- discharge experiment indicates that is is a powerful
ducing 502F2 demanded by equilibrium considerations. tool to study the SF6 decomposition process in detail
The heated surface in the case of welding may play a without interference from electrode materials and ad-
role in the dissociation of SF6 with subsequent attack sorbed moisture on electrode surfaces.
by ozone .
The photo-dissociation of SF6 has been investigated
In passing a mixture of SF6 and air through burning in relation to the isotope separation of UF6. Multi-
cigarette tips, Chu and Massey [42] did not detect any photon decomposition of SF6 by powerful IR lasers has
t A id e*'oMPI _1 :e bs Say a7s l_ s_ ~-_ l_-l_.. Lt:i n -aC-i
-e
'e
' ,
m
column. The readers are referred to the excellent re-

view by Christophorou on electron molecule interactions
[37] and reviews by Pfender [148] on thermodynamic
equilibrium in arcs.
4.1 Basic Mechanisms
The decomposition of SF6 can be caused by three
fundamental processes: electronic, thermal and optical.
In high voltage gas-insulated equipment, the electronic
and thermal processes are the dominant decomposition
mechanisms affecting the formation of stable products.
While both electronic and thermal processes can occur
in discharges such as arcs or sparks, the decomposition
of SF6 in corona is mainly under non-thermodynamic
equilibrium conditions where the electron temperature
is higher than the gas temperature. The decomposition
of SF6 at room temperature under the influence of an
electric field will be dealt with first.
At room temperature and in the absence of an external
applied electric field, the only charged particles in
the otherwise neutral SF6 gas are generated from nat-
ural ionization as a result of cosmic rays or natural
radioactivity [194,195]. The electrons generated from
the natural ionization quickly attach to the SF6 form-
ing SF6 negative ions. The attachment of electrons to
SF6 can best be explained by the mechanisms represented
below.
SF6 + e -b. (SF6)* -> SF6- (4.1)
-> SF5- + F (4.2)
Electrons may associate with SF6 to form a metastable
association complex (SF6)*. The metastable complex can
be collisionally stabilized to SF6 or alternatively,
the complex can decay to form SF5 and F according to
reaction 4.2. The lifetime of the (SF6-j* against
autodetachment is about 25 us [38].
Ahearn and Hannay [4] reported that the formation of
SF- and SF5 ions occurs at an electron energy of less
than 2 eV by a resonance capture process. The forma-
tion of SF6 and SF5 ions at low electron energies of
Fig. 8: Variations of the molar fractions of less than 0.4 eV by a direct dissociative electron
neutral secondary products in a SF6-02 RF dis- attachment process has also been observed, and the peak
charge during the etching of silicon. Note of SFs formation occurs at essentially zero electron
the change in scaZe of SF4. [149] energy [36,67,90].
Fig. 9 shows the electron energy dependence of reac-
been reported [36]. However, in practical conditions, tion 4.2 over the temperature range 300 to 1100 K [36].
the optical radiation emitted in electrical discharges The SF- formation at near zero energy grows dramatical-
will seldom reach the power levels of a laser for the ly as the gas temperature is increased, while the at-
photo-dissociation of SF6 in the infrared region. How- tachment at 0.38 eV is relatively insensitive to the
ever, Smardiewski reported the vacuum uv photolysis of gas temperature. From the data shown in Fig. 9, the
SF6 in an argon/SF6 matrix [169]. Using the radiation activation energy of this channel of dissociative elec-
from a hydrogen discharge lamp, SF4 was identified as tron attachment has been determined as 0.2 eV. Fig. 10
the major photolysis product. shows qualitatively the potential energy curves for the
formation of SF6 from SF5 and F and SF6 from SF- and
4. DECOMPOSITIOON MECHANISMS F, as a function of internuclear separation [36]. The
low energy peak observed in Fig. 9 is the result of at-
The subject of SF6 decomposition mechanisms is a com- tachment of near-zero energy electrons for those vibra-
plicated one, and it is not the intent of this paper to tional-rotational states of SF6 = 0.2 eV or more above
give an exhaustive review. Rather, the approach is to the ground state, i.e. dissociative attachment to "hot"
give a qualitative description of the important process- SF6 molecules
es occurring under predischarge, corona, spark, and arc
conditions regarding the formation of sulphur oxyfluo- SF6(v,r = 0.2 eV) + e ( O ev) -> SF5 + F (4.3)
rides identified in the previous section. Central to
the understanding of the decomposition mechanism are The 0.38 eV peak is the result of vertical Franck-
the electron molecule interactions and the concept of Condon transition of SF6 to the repulsive negative ion
thermodynamic equilibrium which governs chemical equi- curve after capturing a 0.38 eV electron according to
librium and other excitation processes in the discharge this reaction:
t- a.- I I t __ _ ez
SF6(v=O) + e (-0.38 eV) -e SF- + F (4.4) field and N is the number density of the SF6 molecules.
As the electron energy increases or E/N increases, the
production of SF- also increases. Fig. 11 shows the
total and partial attachment cross sections as a func-
tion of electron energy and the total attachment rate
constant as a function of mean electron energy for SF6
-I [38]. One of the processes which McGeehan studied at
l4j
aL)
L-
C
0
u
c
_
12 L 0.5 1.0
Electron Energy, eV
Fig. 9: The ratio of the SF3/SF6 signaZ produced

by Zow energy eZectron impact on SF6. Each curve
is ZabeZed with the SF6 gas temperature [36].
0 2 4 6 8 10 42 14
ELECTRON ENERGY e (eV)
Potential
Energy Fig. 11: Cross section for parent and fragment
negative ions produced by eZectron impact on
SF5 + F SF6. Insets: TotaZ attachment rate constant
as a function or mean energy measured in N2
(e) and Ar (e). Swarm unfolded totaZ attach-
ment cross section ( ), cross section for
EA (SF5) the production of SF6- + SF5- (o) and SF5-
DI(SF5-Fl (.)--- for SF6 [38].
S5 S F5+ Fj
L - - -- - - l EA SF6)
higher E/N is the collision between the SF6- ion and
the SF6 neutral to form SFs according to the following
process
SF6 + SF6 -> SF5 + F + SF6 (4.5)
At even higher E/N, F- appears and its formation is due
to
Internuclear Distance
SF- + SF- -e F + SF4 + SF6 (4.6)
Fig. 10: Schematic potential energy curves of
SF5 + F and SF5 + F as a function of inter- or
nuclear separation. The indicated quantities
are the first bond strength of SF6 D (SF p-F), SF5 + SF6 -> F-+ SF5 + SF5 (4. 7)
the eZectron affinities EA (SF5) and EA (SF6),
and the activation energy Ea 0. 2 eV [36]. The formation of SF5-, and subsequently SF5 and F, is
the first sign of SF6 decomposition. It is interesting
to note that the decomposition (reaction 4.2) can occur
McGeehan et al. [132] studied the ion molecule inter- at electron energy less than 0.4 eV which is much lower
action in SF6 and reported the branching ratio of the than the SF5-F bond energy which has been determined in
autodetach process favors the SF6- by a factor of 22 to the range of 3.5 to 4.0 eV [10,128]. The results of the
1 at low E/N ratio, where E is the applied electric observations imply that even at low E/N and moderate
rhu: SF-6 deco itn i; rt s-imsu1n,t '- --_e
i . ,n nt
temperature, SF6 may dissociate into SF5- which may SF5 OH -> SOF4 HF
convert into other long lived species if impurities are
+ + (4.12)
present in the gaseous medium. SF5 0 -X SOF4 F
+ + (4. 13)
The formation of SF5 radicals from the ion molecule SF4 + 0 -e SOF2 + 2F
reaction of SF5- and SF6 as the temperature increases (4.14)
is the second important step for the SF6 to decompose SF3 0 -e SOF2 F
into other stable products. According to McGeehan
+ + (4.15)
[132], the processes SF2 0
+ -> SOF2 (4. 16)
SFR + SF6 -> F-+--SFx_1 + SF6 (4.8) SF4 + OH -> SOF2 + HF + F (4.17)
-e F- + SFX + SF5 (4.9) SF3 + 02 -e SO2F2 + HF (4. 18)
have been confirmed for E/N up to 360 Td which is just SF2 + 02 -> S02F2
below the threshold field for avalanche. As soon as (4. 19)
the SF5 radical is formed, it can react with itself to Although from a stability point of view, SF4 is more
form S2FlO. This process has been observed by Benson stable than SF5 as an arc by-product, the formation of
and Bott [15] on the S2F10 pyrolysis experiments which SFs from the negative ions of SF6 will undoubtedly be
shows that the reversible process can occur at very low also the dominant process in the discharge region. The
temperature, and the SF5, SF4 radicals are highly re- presence of SOF4 (reactions 4.12 and 4.13) and the sub-
active with impurities in the gas to form the stable sequent conversion of SOF4 to S02F2 may be responsible
oxyfluoride. The above scenarios can occur below the for the observation of a significant amount of these two
corona inception voltage in SF6 which implies that un- species in corona or in low energy sparks. The formation
der the combined effect of electric field and tempera- of SOF2 will be from reactions 4.14, 4.15, and 4.16. The
ture, SF6 decomposition is possible even in the absence formation of other ions and neutral species such as
of a discharge and the decomposition would proceed at a SOF3+, SO2F3+, SF502, SFsO has been observed and they
rate faster than thermal decomposition alone. Should may act as intermediates in the production of the ob-
this rate of decomposition be possible, the thermal served oxyfluorides [188].
aging data previously reported [54,94] may not be appli-
cable to operating SF6 insulating equipment because of Outside the corona discharge region, slower reactions
the presence of an electric field. Unfortunately, between the long-lived lower sulphur fluorides such as
there is very little data in the literature to address SF4 and stable oxyfluoride with the contaminants will
the aging aspect of SF6 insulated equipment under com- further lead to the formation of other types of compounds
bined electric and thermal fields. such as SO2, S02F2, and SOF2. The relative abundance of
these species will depend on the abundance of the parent
4.2 Mechanisms for SF6 Decomposition in Corona compounds and the rate constants of the various reactions.
Sauers [161] reported a reaction rate of 1.0 to 2.6x10-19
In a corona, the electron temperature is much higher ml/s at a temperature of 350 K for the reaction of SF4
than the gas temperature in which non-equilibrium con- and H20.
ditions prevail [79]. Therefore the electron-dominant
process discussed above would be valid in the case of 4.3 SF6 Decomposition in an Arc PZasma
corona which is a weakly ionized gas. Inside the dis-
charge the electron energy will be much higher as the Generally, the arc plasma can be assumed in local
result of the limiting E/N for corona inception. The thermodynamic equilibrium [148]. Therefore the equilib-
mean energy of the electrons in a SF6 corona has been rium concentration of various decomposition products of
reported in the range of 5 to 10 eV [56,106,204]. This SF6 can be calculated as a function of temperature from
electron energy already exceeds the SFs-F bond energy known thermochemical data of the SF6 and its decomposi-
of 3.5 to 4.0 eV and electron impact [49,152] will tion products. Such a determination of the equilibrium
dominate the decomposition process. The formation of concentration of the particle densities in SF6 plasma
negative ions of SF4 and other lower fluorides of sul- has been carried out [72,74,174,196,197]. More recent
phur eventually leads to the formation of various calculations of the equilibrium particle densities have
neutral molecules. taken into account the many atomic and molecular -ions
to reflect the true composition of the plasma [24,73,
The electron impact dissociation processes can lead to 127]. Fig. 12 shows the equilibrium particle density
the following reactions [3,163,188] calculated by Frie using similar techniques as described
above.
e + SF6 -> SFX + (6-x)F, x <5 (4.10)
As the temperature increases beyond 1500 K, there is
e + SFX -> SFx_l sharp drop in the SF6 concentration with a correspond- + F (4.11) a
ing increase in F and SF4 density. The equilibrium con-
Multistep dissociation might also contribute to the centration of SF6 and the decomposition products at tem-
formation of lower fluorides of sulphur such as SF4, peratures between 1000 and 3500 K have been verified by
SF3. In the absence of contaminants in the gas or on shock tube studies and the decomposition rates have been
surfaces, the products of decomposition of SF6 quickly obtained [22,136]. At 4000 K, the molecules are almost
recombine through relatively rapid processes to explain completely dissociated while most of the electrons are
the thermal conversion of SF4+F2 into SF6. However, in still bound due to the formation of negative ions. At
the presence of oxygen or water vapor, the recombination temperatures beyond 15000 K, the particles are mostly
process can be interfered with by the reaction between positive ions and electrons. At the center of a high-
the lower fluorides of sulphur and the contaminants to current arc, the temperature has been measured to be
form sulphur-oxyfluoride, HF, and metallic fluoride. about 20000 K [5].
Some of the reactions can be expressed as follow [188].
-r- - 1 . 4- 4 - - -- T -- '
N '--: E r-It r-- =
.0-1
.,-!'.-! L)
11. -- - - -i.! ----.---;.j:=. ',-, . .. I --
. - -
i-
I
As shown in Fig. 12, the most stable species is SF4

while SFs is not formed in appreciable quantities due
to its instability. The formation of SOF4 according
to reaction (4.13) will be much less likely because of
the lack of parent compounds. Therefore the formation
of S02F2 or SOF4 is very small in the case of a power
arc in comparison to corona or spark.
F2\2 The decomposition mechanisms of SF6 in a spark are
E s+ probably very close to that of an arc discharge. While
1015
05
g' 4
sT~2 eF
/0
e- the spark has very short duration, the spark en-
ergy may be sufficient to cause localized heating so
that in the spark channel the temperature is suffi-
ciently high to dissociate all the molecules within the
very small volume [161]. Upon the collapse of the
spark, the atoms recombine to form sulphur fluoride
molecules, of which SF6 is the most stable. The SF4
2)
formation rate probably depends on the removal of
1013 2~~~~~ 1
fluorine via reactions with metals in the immediate
vicinity of the spark. Thus the amount of SOF2 and
0.5 1 2 4 6 10 20 10^ other by-products depends strongly on the electrode
T K) materials and its susceptibility to attack by fluorine
Fig. 12: Equilbriwun particZe denssities in a under the conditions prevailing during or just after
SF6 plasma as a function of temperature [72]. the spark. Fig. 13 shows the most possible scenario
for the formation of the metal fluoride and the various
As the arc cools, recombination of sulphur and fluo- by-products from contaminants.
rine to form SF6 occurs rapidly, often in the ps scale.
Detailed description of the important reactions in a
decaying SF6 plasma is given by [10&]. In the presence
of oxygen, H20 and metal vapor as a result of electrode SF6
heating, the recombination process is altered which
leads to the formation of various arc by-products. Fig.
12 shows that at temperatures below 3000 K, the main
constituents are SF2, SF4 and F. As the arc cools from ( I) J SPARK
the initial high temperature, the stable SF4 products
will dominate at temperatures between 1500 and 2500 K.
Reaction between oxygen and SF4 will probably take place
at that temperature range to give some of the reactions SF4 +2F
outlined above. As the stable SF4 diffuses out from the
arc zone, reaction with H20 will dominate to form the
lower oxyfluoride. p1 (III)
H20
In addition to the dissociation of SF6 and reactions
with gas impurities near the plasma column, there is
the mechanism near the arc root responsible for the
formation of solid arcing by-products. In the region
near the electrodes, the arc is characterized by the
cathode and anode fall phenomena which are substantially
SOF2 + 2HF ME
different from the arc column. The electrode regions
have much higher electric field strengths, temperature
gradients, and current densities than occur in the arc (IV) 20 SiO2 (v)
column. Heating and fluid dynamic effects in the elec-
trode regions give rise to the plasma jet effect. Molt-
en vapor from the electrode surface is forced into the
arc column by the pressure gradients near the electrode
region. The metal vapor immediately reacts with the
So2 + 2HF SiF4 +2H20
free fluorine atoms to form the metal fluoride compound
which results in a fine powder [43,45]. The high-tem- Fig. 13: Reaction scheme for sparked SF6 Leading
perature environment near the electrode and the arc to formation of the Zong Zived decomposition
column provides excellent conditions for the following products SOF2, S02, SiF4, and MFn (where M
fluorine and secondary reactions: represents the electrode materiaZ) [161].
Al + 3F -> AlF3 (4.20)
Cu + 2F -> CuF2 (4.21) 4.4 The Formation of S2F1o

3SOF2 + A1203 -e. 2A1F3 + 3S02 its formation in arced (4.22) Due to the toxicity of S2FlO,
SF6 is of great interest to researchers. According to
For aluminum electrodes, the reaction (4.20) is exo- Tables 3,6, and 10, S2F10 was identified in all three
thermic which produced about 850 kJ per mole of aluminum. modes of discharge. In the case of a high-current arc,
For copper and steel electrodes, the corresponding re- the amount identified is only 88 ppb [147] and the au-
actions are much less exothermic [109]. The difference thors did not elaborate on the detection methodology
in energies produced for various electrode materials and the circumstances under which the molecule was
explains the influence of electrode materials on the identified. In the case of corona and spark, both
formation rates of the gaseous and solid arcing by- Bartakova [13] and Becher [14] reported the detection
products.
r.-[
-.
v -
-
C . der- r-I-;.
_. i_z :-1 =. - - - _- - -;,r. - , ';-. '! r, '11, .,z T--
- --i
-- -
::? -,- -
I-, -,;,-- .-.
of S2FlO. Becher reported a formation rate of 0.05 4.6 The Effects of Impurities
nmol/J. It appears that the likelihood of finding
S2Fl0 in a low energy discharge such as corona or spark From the above discussions, impurities such as oxygen
is higher than in a high power arc. Although no de- and H20 play a key role in the formation of the oxy-
tailed investigation of the S2F10 formation mechanism fluoride as the dominant SF6 decomposition by-product.
in an SF6 discharge has been reported, the most plaus- The effect of 02 and H20 in the formation of SOF2,
ible route may be from the parent compound of SFs via S02F2 and SOF4 have been investigated by Boudene [23].
the following reactions [15] In the case of SOF2, moisture from the electrode mate-
rials plays a much more important role than the in-
SF5 + SFS =S2F10 (4.23) trinsic moisture in the gas. Moisture in the gaseous
medium will determine the conversion of SOF4 to SOF2 or
SF6 + SF4 > S2FI0 (4.24) HF in secondary reactions. Sauers et al. [161] had
carried out many experiments involving elaborate gas pu-
Based on kinetic studies at temperatures about 450 K, rification, equipment baking and outgassing to drive the
Benson and Bott reported that reaction 4.23 is the fast impurities away. It is only after such elaborate proc-
reaction and the rate constant was given [15]. Since esses that they can obtain ion spectra free from the
S2FlO is almost completely decomposed at temperatures oxyfluoride interference. Such elaborate procedures
higher than 2000C [65], it is unlikely that this mole- are not possible in a practical environment or even in
cule can survive the heat from a high-current discharge laboratory conditions. Therefore the effect of impuri-
after its formation. In a low energy discharge, the ties will continue to dominate the by-product formation
temperature of the gas is more suitable for newly form- in a practical environment.
ed S2FlO to exist without decomposition. In addition,
the electronic process for the formation of SF6 and SFs 5. EFFECTS OF SF6 DECOMPOSITION PRODUCTS
ions in a cool discharge would tend to yield more SFs
molecules than the high-current arc during cooling which 5.1 Toxicity of SF6 Decomposition Products
producog more SF4. This could explain the differences
in yield between high-current discharge and low-energy With the increased concern in personnel health and
spark and corona. Further work is required to explain safety, there is a growing interest in the toxicity of
the formation mechanism of S2Fl and evaluate its forma- SF6 and its decomposition products in gas-insulated
tion rates under various discharge conditions. equipment. This tendency is reflected in many recent
utility sponsored studies, workshops and discussions on
4.5 SF6 Decomposition in the Presence of PoZymeric the subject [11,81,83,119,158,161,190,200]. Pure SF6
MateriaLs is a non-toxic gas. Its inertness has been reported by
Lester and Greenberg [120] who exposed rats for several
Polymeric materials such as epoxy, polyethylene, hours at atmospheres consisting of 80% SF6 and 20% 02
Teflon, and phenolic are frequently used as solid in- and noted no sign of intoxication or even irritation in
sulators inside SF6-insulated equipment. The most com- these animals. Lester's results were confirmed by oth-
mon type of support insulators in GIS are made of filled er animal experiments carried out by Specht and Brubach
epoxy. When a flashover occurs at the SF6 solid inter- [170]. SF6 is given a Threshold Limit Value (TLV)-Time
face (e.g. spacer surface flashover), the presence of Weighted Average (TWA) of 1000 ppmv in air [122]. The
polymeric materials in the discharge will influence the only personnel risk from exposure to SF6 is from the
formation of the by-products. The chemical analyses of asphyxiation effect.
the gases produced by the degradation of polyethylene
and epoxy resin maintained in the vicinity of an arc The toxicities of SF6 decomposition products were
burning in SF6 have been reported [11,23,47]. The con- first recognized in the early fifties during the devel-
centrations of CF4, C02, S02F2 and SOF2 were found to opment of SF6 switchgear [30,63]. Camilli et al. car-
be higher when the arc was burning near polymeric mate- ried out high-current experiments (8000 A) with copper
rials. In addition, CS2 and COS were identified in arc- electrodes and the arced gas was analyzed by IR spectro-
ing near spacers. A model has-been proposed to describe scopy [30]. SF2 and SF4 were identified but no S2F10
the interaction between the SF6 arc and the polymeric was found. The arced gas was tested at the Applied
materials in two steps [32]. In the first step, the Physiology Laboratory of Yale University for animal ex-
polymeric materials were pyrolyzed by the arc, liberat- posure effects. The results indicated that rats showed
ing pyrolysis products such as methane, C02, etc. The physiological effects to air with 10% arced gas. Below
second step involved a gas phase reaction between the 7.5%, no harmful effect was observed.
SF6 and its by-products with the volatile products gen-
erated by the arc degradation. Since polymeric mate- General considerations on the toxicity of dielectric
rials can adsorb and absorb moisture, the volatile prod- gases are described by Lester [121]. He pointed out
ucts contain significant quantities of H20 molecules to the involuntary nature of gaseous exposure compared to
influence the gaseous phase reactions. Experimental re- the easily controlled conditions for solid and liquid
sults indicated that while CF4 is the dominant products exposure. Of the four broad categories of hazardous
from arcing, gas products of the thermal degradation of gases, SF6 and its decomposition by-products are de-
epoxy and phenolic resins favor hydrolysis reactions of scribed to fall under the simple asphyxiants and irri-
the primary decomposition species of SF6 which then re- tant gases category respectively.
sults in the immediate formation of the two most abund-
ant corrosive products, namely HF and SOF2 [8,32]. It Boudene [23] carried out an exhaustive study on the
has been shown [38a] that SF6/perfluorocarbon mixtures physiological effects of SOF2 and S02F2 for mice, rats
have certain advantages in this respect in that the C and rabbits. SOF2 and S02F2 were chosen because these
atoms of the perfluorocarbon interact with the F atoms two are the dominant stable arc by-products detected.
of SF6 forming principally inert perfluorocarbon com- The results of the animal test indicated that SOF2 at-
pounds. tacks the lungs of the animals exposed to this gas.
,.=- - , '-!
' ;L- i;xt-
. .s E ., _s_ 4
srn
-1sl - Olt-1
N_ c v-,I 98Bh
Octobe
The lethal concentration for rats was determined to be toxicity of the solid by-products generated in the
about 100 ppmv for exposure times of sixty minutes. presence of other gaseous arc by-products.
SF4, SOF4, HF and WF6 were found to have similar pulmo-
nary effects. On the other hand, S02F2 was found to be 5.2 Discussion of Toxicity
a convulsing compound. The lethal concentration for
rats is determined by the authors to be 2000 ppmv for a Since SF6-insulated equipment is a completely sealed
sixty minutes exposure. S02F2 is odorless and it can- system, any decomposition products generated should be
not be detected. However the research has shown that contained inside the enclosure unless the enclosure is
it is always present with other pungent gases such as burnt through or a pressure relief valve operates dur-
SOF2. ing a fault, releasing the by-products into the ambient
air [16,46]. The greatest risk for personnel exposure
The toxicity of sparked SF6 was investigated by the to the by-products is during repair or maintenance of
Oak Ridge National Laboratory's Health and Safety Re- the compartment. In order to minimize the risk, an
search Division in an extended study involving mam- adequate protection system and handling procedure are
malian cells [62,82-85,158,161]. The study was carried required. The most common approach to establishing
out with Chinese hamster lung cells exposed to pure safe handling procedures is to obtain information about
SF6, sparked SF6, and synthetic mixtures of the SF6 de- the types of decomposition products formed, their forma-
composition by-products. The toxicities were determined tion rates and toxicities. This information was used
from the relative survival rates of the cells after ex- for the estimation of levels of decomposition products
posure. Typical arc energy used in the exposure experi- generated under various situations and energy levels.
ments was 16 kJ in a volume of 60 ml. The study shows These levels were then compared to existing allow-
that SF6 is non-toxic to the hamster lung cells even able exposure levels, e.g. Threshold Limit Values [122]
after exposure to high concentration of SF6 for rela- to establish procedures and protection systems.
tively long times. However, the study concluded that
sparked SF6 exhibits cytotoxic actions with increasing The above approach has been adopted by a number of
exposure time. The toxicities have been observed to in- organizations to establish safe working procedures [45,
crease with total spark energy. The authors also deter- 119,134,177]. The toxicities of the various decomposi-
mined the cytotoxicity of individual arc by-products, tion products are usually obtained from the ACGIH [1]
SO2, HF, SOF2, S02F2, SiF4, SOF4, SF4. A rough ranking and NIOSH [122] publications. The TLV limits were the
of the cytotoxicity of the major decomposition products most common guideline used in determining safe levels.
shows that S02F2 and SOF4 as roughly equivalent and more Table 12 is a summary of animal toxicity to SF6 and its
toxic than SiF4, which is more toxic than SOF2. Sur- decomposition products. Detailed description of the
prisingly, the results also show that SO2 and HF did gaseous and solid decomposition products properties can
not show any toxicity. These observations were contrary be found in [45].
to toxicity studies on mice and rabbits which show that
SOF2 is more toxic than S02F2 and that SO2 and HF are
toxic. The cytotoxicity of a synthetic mixture of the TABLE 12 Summwry of Animal Toxicty to SF6 and Potential Breakdown Products 11,1611
by-products in concentrations similar to the sparked
sample was also determined. The cytotoxicity of the
synthetic mixture was not nearly as strong as a sparked
SF6 sample. Furthermore, while individual components TLV-TWA LDLO Human LDLO RatMou
do exhibit cytotoxic activity, this cytotoxicity is
seen at concentrations higher than for these components SF, 1000 ppm 80,000 ppm
found in spark-decomposed SF6. Thus, the authors have SF, 0.1 ppm 19 ppm/4 h
not been able to account for the overall high degree of SOF, (a)
SOF, 2.5 mg/m' 260 ppm/i h
cytotoxicity exhibited by sparked SF6 on the basis of SO,F, 5 ppm 1200 ppm/i h
the measured cytotoxicities of the detected by-products So, 2 ppm 400 ppm/i min 6000 ppm/S h
0.025 ppm 130 mg/mJ/10 min
or by testing synthetic mixtures. A typical mixture (a)
SiF,
contained 0.2% S02F2 and SiF4, 2% SOF2, and 4% HF. This HF 3 ppm 50 ppm/30 min 260 mg/m'/7 h(c)
synthetic mixture was only mildly cytotoxic in the assay CF. 2.5 mg/m'
Cs, 10 ppm
system; other synthetic mixtures have exhibited no cy- A1F, 2.5 mg/m'
totoxic activity [81a]. The authors speculated that CuF, 2.5 mg/m'
there is the possibility of yet unidentified minor spark- FeF, 2.5 mg/m'
ed by-products contributing considerably to the measured FeS 1mg/mi'
cytotoxicity of sparked SF6. Further work is necessary
Si(CHj),F, 1 mg/m'(b)
to quantify the toxicity of sparked SF6 with reference
to that of better characterized gases, to establish the
cause of the toxicity and to develop ameliorative pro-
cedures. The authors have previously hypothesized [85] aThe number 2.5 mg/mi haa been given elsewhere [1,11,451 as a TLV for this substance. The
NIOSH (1982) registry gives this number for several substances (for example, SOF,) as a
that S2FlO could be a possible candidate for the minor generally recommended value for inorganic fluorides. No such recommendation is made,
component in sparked SF6 samples which may contribute to however, for the substance in question.
the majority of the cytotoxic activity. Recent experi- b Ref [45]
mental evidence [81a] seems to support this hypothesis.
c Value for rabbit
The toxicities of the solid arcing by-products were
examined by Chu [45]. Aluminum trifluoride and copper
fluoride were the dominant solid by-products formed in 5.3 The Effects of By-Products on GIS
most circumstances. The author concluded that the major
contributions of the solid by-product's toxicities are 5.3.1 Gaseous By-Products
from the adsorbed gases on the surface of the pm-sized
powder. Furthermore, the small size of the powder im- During normal or abnormal operation of gas-insulated
poses certain pulmonary effects on the lung. Although equipment, SF6 is decomposed, and the accumulation of
the TLV for AlF3 is 2.5 mg/m3, there is no data on the decomposition products in the enclosed compartment can
.,n: E, I z
.EI b .,pTen- _7 1 -I-
affect the system performance in two ways. The pres-

of the gaseous and solid by-products may lower the
ence
dielectric strength of the system. If the by-products id6' PTFE
are left in the compartment for an extended period of
time, corrosion of the system components may ensue. i d4 L- UIA-FILLED EPOXY
The dielectric strengths of several gaseous by-products (a)
were investigated by Manion [129] in testing the arc
0 P SURFACE RESISTIVITY
stability of electronegative gases. The results indi-
cate that even for arc by-products up to 30% by volume,
there is no degradation of the dielectric strength of
the gas gap. Fig. 1 shows insignificant change of the cr_ 10 SILICA-FILLED EPOXY
impulse flashover strength of SF6 after arcing. ZD
Manion's results were confirmed by other investigations UI106-0 20 40 60 80 100
on the dielectric strength of the individual decomposi-
tion products. The relative dielectric strength at uni- 1.0 TIME (h)
form field for SF6, SOF2, S02F2, S02 and CF4 are re-
spectively 1.0, 1.4, 0.75, 1.0, and 0.'42 [25,201]. In 0.8 SILICA
0.8 -~FILLED EPOX
view of the dielectric strengths of these arc by-prod-
ucts and their concentrations, it is not surprising
1-
0.A (b)
that there is little effect on the gas gap's insulating 0 MOISTURE CONTENT
strength. 0.4
Although the gaseous by-products have little effect E 0.2 ALUMINA-FILLED EPOXY
on the insulation strength, the gas can chemically at-
tack solid insulating materials which may ev-entually o
0
L PTFE, 0
lead to reduction of the dielectric strength. Flash- 20 40 G0 80 100
over on the surface of support insulators or spacers is ClE1.0 TIME (h)
a common cause of GIS in-service failure [20,113,114]
and much effort has been spent in preventing attack of
$'EI.O[ SILICA- F LLED EPOXY
the corrosive by-products on spacer surfaces by choos-
ing the proper materials, surface coatings and by the ( C)
removal of by-products by placing absorbents inside
circuit breakers. Since filled-epoxy insulators are ArA ON THE SURFACE
extensively used in GIS as support spacers, most mate- :D ALUMINA-FILLED EPOXY
rial compatibility studies were focused on this type of
material. The results of many investigations show un- 0 20 40 60 80 100
ambiguously that silica filler in the epoxy spacers is TIME (h)
susceptible to attack by the decomposition products
[12,32,35,47,141,157,180]. Blisters were observed to
form on the surface of silica-filled epoxy spacers af-
Fig. 14: Surface resistivity, moisture content and
ter the material was exposed to 6000 ppmv of decomposi-
hydrolyzabZe fZuoride surfaces of PTFE, aZumina
and siZica epoxy resin sampZes exposed to SF6
tion products for 72 h [47]. The exposure time and containing 3 voZ% of S.F4 [180].
concentration are typical of that found in a gas com-
partment after it has experienced a power arc fault.
Associated with the blister formation, there is a cor- Silica- Alumina-
responding drop in the surface resistance and the im- Filled Filled
pulse strength. A 50% drop in the impulse strength and Material Materidl
a corresponding drop of 5 orders of magnitude of the
surface resistance have been observed [47,171]. Fig. 1.2
{I
14 shows the changes in surface resistivities, moisture
content and hydrolyzable fluoride on the surfaces of 1.1 k
three materials after exposure to SF6 containing 3 vol%
of SF4. Fig. 15 shows the drop in impulse withstand
1.0 1-
voltage of silica filled materials after exposure to c
6500 ppmv HF vapor for 72 h.

} I Control Specimen
D 0.9 - A Sandblasted
A model of decomDosition products attack on silica- 4)
(6500 ppmv HF Vapour)
filled spacers has been proposed [47,157 ,171-,181]. The 0
model basically starts with the conversion of oxyfluo- 0.8 1- * 35 ppmv HF Vapour
E) * 350 ppmv H F Vapour
rides to HF according to reaction 5.1. The sources of
moisture are H20 in the SF6 gas or adsorbed moisture on 0.7 H * 6500 ppmv HF Vapour
the surface or dissolved moisture in the epoxy resin.
The diffusion of HF or other decomposition products 0.6 M-
through the resin to the filler site depends on the dif-
fusivity and solubility of the arc by-products in epoxy
resin. Calculations have shown that the diffusion can 0.5 i-
be very rapid and within one hour, the top 20 ,im layer
of the resin is almost saturated with HF [47]. When 0.4
the HF molecules encounter the silica filler, chemical First Flashover (Normalized)
reactions can occur according to reaction 5.2. The re-
action generates SiF4 which is a gas. The pressure Fig. 15: The change in surface impulse withstand
build-up leads to the formation of the raised blisters. voZtage of siZica-fiZZed epoxy sampZes after ex-
With A1203 or dolomite as fillers the reaction between posure to 6500 ppmv and 350 ppmv of HF vapor for
HF and the particles is much slower and no blisters are 72 h [171].
t-T l:
El t jnl'-4-j__I
l'laion - * 2 n Octhr lPP6
v~
formed. The resistance of alumina or dolomite fillers that the fluoride powder formed during arcing has a very
to arc by-product attack is much better than silica large surface area and the powder can adsorb significant
[35,47,98]. quantities of gaseous by-products. The reaction of the
surface adsorbed gaseous by-product and the solid in-
SF4 + H20 -> SOF2 + 2HF (5.1, sulating surface when the solid by-product is in contact
with the surface is responsible for the reduction of the
SOF2 + H20 e S02 + 2HF (5.2' solid insulator's dielectric strength. From a practical
point of view, the problems caused by the solid by-prod-
SiO2 + 4HF -e. SiF4 + 2H20 (5.3) ucts are more severe because they cannot be easily re-
moved from the system without opening the compartments.
SiF4 + 2HF + H20 -> H2SiF6 + H20 (5.4)
5.4 RemovaZ of SF6 Arced By-Products
The formation of blisters which contain highly elec-
trolytic materials such as H2SiF6, and the absorption Since SF6 arced by-products are corrosive and may af-
of HF gas in the epoxy resin greatly modify the fect organic materials in the gas insulated equipment,
surface resistance and ultimately lead to the ob- the accumulation of excessive amounts in the compartment
served drop in the impulse flashover voltage. The di- has to be removed. This is especially true for circuit
rect relationship between the drop in surface resistance breakers because of the amount of arced by-products
and the drop in electrical insulation strength has been generated. Since the early days of circuit breaker de-
confirmed by a number of experiments using artificial sign, absorbent materials have been placed inside circuit
contaminants and arc by-products related contaminants breakers to remove the gaseous arced by-products [6,59,
[17,28,172]. 99,165]. Schumb suggested the use of soda lime and
activated alumina to absorb the arcing by-products. The
The effects of arc by-products or resins have been practice of installing absorbents has been extended to
studied [47,97,98] with respect to electrical and chemi- disconnect switch chambers where low energy arcing oc-
cal aging. There is evidence to show that certain types curs periodically. The absorbents serve the dual pur-
of aliphatic amine-cured resin can become more conduc- poses of absorbing moisture as well as arc by-products
tive as a result of the formation of amine salt which [6].
can dissociate into separate ions [47,53]. The recent
work of Chu has indicated that both the fillers and the The most common materials used as absorbents are acti-
resins can be affected by the arc decomposition products. vated alumina and molecular sieves. Ushio [183] deter-
mined that the absorptive power of molecular sieve ex-
The effects of gaseous by-products on other organic ceeds that of activated alumina as measured for H20, SF4,
materials inside the gas-insulated equipment have been and SOF2. The quantities of gas absorbed per unit weight
investigated [92,99,173,183]. SF6 dissociation prod- of absorbent have been examined by Boudene [23] who mea-
ucts tend to attack lithium-based grease while Bentonite sured the absorption capacity of activated alumina and
grease and Urea grease suffer relatively little degrad- charcoal at ambient temperature. The absorbers were pre-
ation [173]. Silicone-based materials are also subject pared by a 12 h vacuum bake at 1 Pa pressure and 2000C
to attacks by the arc gas with mechanisms similar to the temperature. The volume of by-products adsorbed by one
one described above. Silicone grease, silicone rubber gram of activated alumina and activated charcoal at nor-
and 0-rings are found to be degraded by the corrosive mal pressure and temperature is given in Table 13. The
HF resulting in the formation of the toxic gas dimethyl- adsorptive power depends strongly on the preparation of
difluorosilane Si(CH3)2F2 [45,93,192]. In large accel- the adsorber. The capacity of the activated alumina is
erators, the attack of the corona induced SF6 by-prod- reduced roughly one half if the bake-out at 2000C is per-
ucts on the nylon chain of the van de Graaf generator formed in air instead of vacuum.
is the cause of frequent outages and chain replacements
[143,156]. Recently the surface degradation of dielec- TABLE 13 Absorptiw Power of Activated Alumna and Charcoal (231
tric materials such as Lexan, nylon, Lucite, Delrin
and boron nitride used in HV spark gaps containing SF6
has been studied [102]. In order to study the mecha-
nisms of attack, sophisticated techniques such as Elec- Volume Absorbed per Gram Volume Absorbed per Gram
of Activated Charcoal
Component of Activated Alumina
tron Spectroscopy for Chemical Analysis (ESCA) and Absorbed (cm') NPT (cmJ) NPT
Scanning Electron Microscopy (SEM) have been used to
probe the solid surface [26]. These powerful techniques 32 52
have yielded valuable information in identifying key SOF,
compounds formed on surfaces due to attack by the de- SO,F. 12 35
composition products. 35 48 SF,
5.3.2 Effects of SoZid By-Products SOF, 20 40
Fukuda [75] reported that after a high-current fault So, 38 68

in a bus duct, the adjacent spacers were covered with a
layer of white powder and the spacer did not exhibit
any drop in insulation strength if the gas condition The movement of arc by-products in an enclosed chamber
was kept dry. However, Hampton [87] in a discussion to in the presence of absorbents has been evaluated with
Fukuda's paper claimed tnat tests at CEGB showed a drop the diffusion equation [111,183]. The results show that
in the insulation strength of the spacer. Murakami and the diffusion time constants for the arc by-products to
Menu [137] claimed a 20% reduction of flashover strength reach the absorbent location a few meters away can be
when the insulator's surface was covered with by- tens of hours. Kusumoto's work [111] yielded an absorp-
products generated from arcing. Latour-Slowikowska tion time constant for partial discharge products in a
[115] investigated the degradation mechanisms of the simulated bus duct to be in the order of 10 to 100 h be-
fluoride powder and concluded that it is the acidity of fore there is a levelling of the by-product concentration
the powder which causes the damage. Chu [45] reported for a constant source. Without forced convection, it is
7i
Cnu: .-3 d"6 nns-inaulates
de-msi ltion -i 'e-t
quite clear that the gaseous by-products will take some comprehensive review on the subject of arc by-products
time to reach the absorbent, and during the transit time handling and the report recommended procedures for
damage could be done to the nearby materials. Amalric [7] dealing with solid as well as gaseous by-products [45].
reviewed the French experience that a circuit breaker Due to the small size of the solid by-products and the
designed for the cumulative interruption of 106 A cor- absorbed gas on the surface, special masks and clothing
responding to an arc energy of 104 kJ decomposes 30 1 were recommended for personnel protection during clean-
of SF6 gas. This requires 2 kg of activated alumina up of GIS. A full face mask with external air breathing
for absorption. apparatus was recommended for situations in which the
ambient SF6 decomposition levels exceeded the TLV
A Japanese survey [11] reports that activated alumina limits. Special vacuum cleaners were recommended for
is effective for the removal of SOF2, H2CO3, H2S, and the removal of the solid by-products. The interactions
SF2. S02F2 is absorbed at a very slow rate. Soda lime of common cleaning fluid and water with the solid by-
(CaO-NaOH) is more effective in removing the stable products were investigated and no violent reactions
gases. The soda lime absorbent is also effective in were observed.
removing gases such as SF4, SiF4, WF6, and S2F2. Allied
Chemical [6] specified that decomposition products can 6. DECOMPOSITION AS A DIAGNOSTIC TOOL
be removed by a 50/50 mixture of soda lime and activated
alumina. The suggested weight of absorbent corresponds 6.1 AnaZysis Techniques for Field Applications
to about 10% the weight of gas.
Gas analysis is a powerful tool in diagnosing the in-
In particle accelerators and power equipment, SF6 gas ternal condition of gas-insulated equipment. Many
from the compartment can be removed by a gas storage and utilities have already used the gas analysis technique
handling system [124]. The compressor of such a system in their GIS maintenance [51,61,130,133]. The major
will liquefy the gas for storage. Separation of the by- incentive for using gas analysis as a diagnostic tool
products during the compression process is possible and is that the technique itself is non-invasive and re-
gases such as N2 are usually removed by liquefaction quires no retrofitting of complex equipment to the GIS.
[135]. However most by-products have boiling points In many instances where the gas handling system permits,
similar to SF6 and an external absorbent chamber is at- gas analysis or sampling can be done on-line while the
tached to the gas handling system for the removal of GIS is still energized [162,181]. Such an on-line
moisture and by-products. An elaborate gas purification diagnostic tool offers significant economic benefits to
system employing activated alumina to remove corona by- utilities or equipment operators who cannot afford to
products in an accelerator was described by Ophel [143]. take the equipment out of service for analysis.
The author suggested periodic checking of the absorptive
power of the activated alumina and the soda lime com- The simplest diagnostic is fault location in GIS.
pound to prevent saturation of the mixture. Due to the Due to the enclosed nature of gas-insulated equipment,
large volume of accelerators, absorbents can become the precise location of a fault during testing or under
saturated, thus reducing the absorptive power. in-service conditions is often very difficult. In many
circumstances, no visible sign of problems can be de-
Although the absorptive power of ordinary desiccants tected from the outside by visual inspection. Very of-
can be regenerated by vacuum pumping and heating the ten, the workers rely on the "rotten egg" smell generat-
desiccants, the reactivation of alumina after the mix- ed by the by-products to locate the power arc fault to
ture is saturated with arc by-products has to be treated within a single compartment. However, gas sniffing is
with extreme caution. There have been observations of discouraged for health and safety reasons. At present
exothermic reaction between the warm activated alumina there are several techniques for analyzing the gas af-
and the arc by-products or SF6 during the reactivation ter a power arc for fault location purposes [31,41,199].
process [6,58]. The build up of excess heat during the The most common and low cost technique is the appli-
chemical reactions may rupture the metal container. The cation of chemical detection tubes to determine the
possible reactions between the hot alumina and the by- concentration of HF or S02 in the gas [27,34,111,181].
products can proceed as follows: Braun and Chu have exploited the availability of low
cost and highly sensitive S02 detector tubes to locate
6 HF + A1203 -+ 2A1F3 + 3H20 (5.5) faults in gas-insulated substations which consist of
many gas tight compartments [27]. The gas from a sus-
3 SOF2 + A1203 + 2A1F3 + 3S02 (5.6) pected chamber is passed through a glass tube contain-
ing chemicals sensitive to the presence of S02. About
3 SF4 + 2A1203 - 4A1F3 + 3S02 (5.7) 100 ml of gas was required to produce a detectable color
change atan S02 concentration of 1 ppmv. As the major
Due to the chemical reactions between the activated decomposition product in a power arc fault is SOF2, the
alumina and the by-products, reactivation may not com- reaction of this by-product with ambient moisture in the
pletely regenerate the absorptive power. In such cir- sampling system converts SOF2 to S02 which is in turn
cumstances, it may be necessary to replace the absorbent detected by the chemical tube. The authors have used
completely. No information was available on the inter- this technique for locating several faults.
action between common molecular sieves and SF6 arc by-
products. Castonquay [34] evaluated several types of chemical de-
tector tubes for detecting SF6 decomposition products in
5.5 HandZing Procedures faulted equipment. The results show that certain types
of tubes which are sensitive to S02 offer improved sen-
The handling of SF6 arc by-products is governed by the sitivity and minimize the amount of gas required to flow
toxicities of the by-products. Although various utili- through the sensor. Tominaga [181] uses the agent
ties and organizations have their own rules and regula- chromocresol purple to detect the presence of HF in
tions regarding the handling and removal of the by- faulted equipment. The sensitivity can reach 10 ppmv
products from the compartments during routine mainte- but the sensor requires 30 1 of gas to detect any color
nance or after a fault, there is no uniform standard at change. In a small chamber with volume less than 500 1
the national or international level [45,70,134]. The the requirement of 30 1 of gas is a major hurdle in ap-
Canadian Electrical Association had commissioned a plying such a technique.
i _'*= -,- i _ e _ llFI l-
Et1- r,
a a .F- ? O1P l . =:,, -' e- + ryr
t 1
r* _
19PL
The detector tube technique had been extended to gas produced fluoroboric acid which is much more strong-
locate faults during testing and to distinguish whether ly ionized in solution than the boric acid so that a
the fault occurs across the gas gap of the spacer [77, significant increase in conductivity resulted. Since
154]. Due to the non-self-restoring nature of certain the conductivity technique cannot discriminate individ-
spacers [47,76], the GIS test engineer often requires ual species and is difficult to use, its application
the location information to decide whether further ap- is limited to laboratory use. The field application
plication of voltages is appropriate. The equipment of this wet chemistry technique by utilities is very
will likely withstand another application of voltage if limited.
the previous fault occurred across a gas gap. However,
the equipment will likely break down again if the fault 6.2 Laboratory Techniques
was across a spacer and further application of voltage
will cause further damage. For faults occurring across Due to its low cost and wide availability, gas chro-
spacers, the S02 concentration is significantly higher matography is the most common by-product detection
than for a gas gap breakdown. This is due to the re- technique [23,151] for routine analysis. The standard
lease of H20 from the surface of the spacer during the GC for SF6 by-product analysis is usually equipped with
discharge which enhances the reaction between SOF2 and Porapak Q columns and a thermal conductivity detector.
moisture to produce S02. Although the technique was Quantitative analysis can be obtained by preparing
successfully tested in the laboratory with controlled standard samples of known concentrations. The sensi-
arc energy release experiments, its application in the tivity of the standard GC technique is about 50 ppmv
field remains to be investigated [154]. One of the of SOF2 and S02F2. The relative retention times of
difficulties is that the amount of by-product produced several common SF6 arc by-products are given in Table
in a test flashover is limited and the detector tube 14 for the Porapak materials. The retention times for
may not be sensitive enough to identify a flashover,
let alone differentiate between a gas gap and a spacer
flashover. TAMLE 14 Reteion Times and Detection Limits in Gas Chrouatograph. 1231
The application of the detector tube technique to
partial discharge detection is a more challenging task
and so far there is very little information on this ap- Cwnn PORAPAK Q Temperature: 100C
Length 3048 m Carrier Gas: heium (flow 60 mi/mn)
plication. Braun and Chu [27] used anSO2 tube to de- DLameter: 0.97 cn
tect a compartment with prolonged partial discharge and
obtained satisfactory agreement with laboratory GC/MS Detection LimiU
analysis. The concentration in that analysis was about Gaseous Components Reention time (in mn) (Content In % Vdwne)
25 ppmv. The authors proposed using anSO2 tube in the 2.6 10o-
Air
range of 0.1 ppm sensitivity with a gas throughput of CF, 3.1 1o-j
100 ml or more to determine the condition of the cham- CO, 4.6 2.10-'
bers. Since one of the major by-products in partial SF, 5.7 3.10-'
discharge is S02F2 which is a stable molecule, the ap- SOJF, 7.6 3.10'3
SOF, 8.2 3.10-'
plication of anSO2 detector tube is only a qualitative SOF, 11.2 3.10-'
analysis and the technique may not be sensitive enough SF, 11.2 3.10-'
HF 16.5 3.10-'
to give reliable results. SO, 24.5 4.10-'
SFe 60 5.10-3
One of the problems of the detector tube method is
that it cannot distinguish the species. In chambers
such as for disconnect switches which have an intrinsic
amount of by-products, the technique cannot discriminate SF4 and SOF2 are very similar and the GC technique can-
the partial discharge by-products and the by-products not distinguish between these two species. The close
from the disconnect switching arcs. proximity of the S02F2 peak and the SF6 reteption peak
also make the standard GC technique not suitable for
Other techniques to detect decomposition products in high-sensitivity detection.
the field have also been developed but with limited
success. The Canadian Electrical Association has de- To increase sensitivity, combined gas chromotograph
veloped a field instrument based on the infrared absorp- and mass spectrometry has been used for the analysis of
tion principle to detect H20, SOF2, and CF4 [95]. The gaseous by-products derived from the arc-induced de-
sensitivity of the instrument is 112 ppmv of H20, 1 ppm composition of SF6 with success [42,91,188,192]. The
of SOF2 and 78 ppm of CF4. Any anomalous increase of separation of the species by the GC column and the sub-
CF4 can be traced to partial discharge near the epoxy sequent determination from the species' mass spectra
spacer. greatly improve the technique's sensitivity to about
1 ppmv for gases such as SOF2, S02F2, SOF4, CF4 and
A portable gas chromatograph has been developed by GIS other less common species sucy as COS and Si(CH3)2F2.
equipment manufacturers to check the internal operating Recent advances in GC/MS techniques and the availability
condition of a compartment. The GC is usually equipped of on-line computerized compound identification greatly
with a thermal conductivity detector with a sensitivity enhance this technique's sensitivity and popularity for
of 50 ppmv of SOF2 and S02F2, and 120 ppmv of H20. A routine analysis. Despite its cost, GC/MS has rapidly
microprocessor-controlled GC with printouts and auto- become an acceptable equipment in utility chemical
matic display has been developed for field use. analytical laboratories for routine analysis of SF6 by-
products.
To detect SF6 by-products in accelerators, a much
higher sensitivity technique is required to identify the Gas chromatography with a Hall conductivity detector
corona discharge products. Ophel et al. developed a was evaluated [60] to replace GC/MS as a laboratory
continuous flow conductivity cell to measure the total technique for routine SF6 arc by-product analysis.
hydrolyzable fluoride content with a sensitivity of 0.1 The Hall electrolytic conductivity detector is a sensi-
ppmv [143]. SF6 was passed through a stripping solution tive and element-specific detector which can detect
of boric acid,and hydrolyzable fluoride present in the sulphur containing compounds to the parts per billion
:h -dEi Xins in g53-1> ua;ed
poszt; equIL prer., E 77 1 -7
level. Detection limits of 200 ppb of SOF2 and 20 ppb utilities to sample the gas for subsequent laboratory
of SO2 have been reported. However, the sensitivity analysis. Such analyses are normally required for
to S02F2 was only 20 ppm and this sensitivity was health and safety and diagnostic purposes. Due to the
achieved with temperature programming and SF6 venting. corrosive nature of the by-products, proper sampling
equipment is needed to prevent the disappearance and
Infrared techniques have been used by many research- conversion of the by-products during transit from
ers to identify the formation of many SF6 arc by-prod- field to laboratory. Stainless steel sampling cylinders
ucts [14,63,131,189]. However, due to the many absorp- have been used with satisfactory results [11,45]. The
tion bands of SF6 and the arc by-products in the in- 150 ml cylinders are fitted with proper valves for
frared region, the quantitative determination of the connection to the operating GIS and the cylinders can
by-products by IR spectrometer has met with limited withstand up to 7 MPa pressure. The sampling procedures
success [63,95]. IR techniques offer the distinct ad- can be found in [45]. The application of 1 1 aluminum
vantage to monitor the evolution of the primary and lined Mylar bags for sampling purposes has been in-
secondary arc by-products during the formation process. vestigated [42]. SOF2 was found to be stable in the
Many transient species such as SF4 were detected by IR bags for a period up to 24 h, but S02F2 and SO2 were
techniques in-situ. Recently the Fourier Transform in- found to disappear within two hours after sampling.
frared technique has been applied to monitor the re-
action chemistry in an SF6 cell after arcing. Fig. 16 6.4 The Detection of S2F10
shows the relative concentrations of several species of
SF6 arc by-products immediately after arcing. Other Due to the extremely low TLV of S2FlO (25 ppb), its
exotic techniques to analyze SF6 by-products in the detection to within the TLV limit poses a technical
laboratory include a mass spectrometer directly coupled challenge. Although S2F1O in the ppm range has been de-
to the arc cell and the application of a tunable IR tected [14], the detection of S2F10 at subppm levels
laser diode as a source for high resolution absorption calls for more sophisticated techniques. The difficult-
spectroscopic detection [161,189]. ies in handling such extremely toxic gas also pose
problems in employing GC or GC/MS as an analytical tool.
Unfortunately, the infrared spectrum of S2F10 is inter-
fered by many strong SF6 and SOF2 lines [112,168,198],
making the identification of S2FlO by an infrared techni-
que very difficult. The application of ultra-high re-
solution infrared spectroscopy such as the tunable diode
laser IR absorption technique may overcome the inter-
ference problem [189].
The adsorption-desorption gas chromatographic-IR
technique has been used to determine ppm levels of
S2F10 [88]. The concentration of S2FlO was enriched by
passing the SF6 gas through a cold trap whereby the
S2FlO was preferentially adsorbed. Subsequent heating
of the trap desorbed the S2F10 into a gas chromatograph
for quantitative analysis and then to an on-line IR
spectrometer for positive identification. Using a
similar enrichment technique with cryogenic traps,
Janssen [103] reported successful separation of S2F10
and S2F100 from the huge background of SF6. The esti-
mated sensitivity of the GC-electron capture dector with
the enrichment technique is about 10 to 30 ppb.
TIME AFTER ARC, 1000 sec
6.5 Solid By-product AnaZysis
Fig. 16: The evolution of the arc by-products
as a function of time as monitored by FT-IR In contrast to the many techniques available to
spectrometer [131]. analyze the gaseous components, the most commonly used
technique for the analysis of solid by-products is
x-ray diffraction [45,115]. The various phases of the
Using the 19F nuclear magnetic resonance technique, metal fluorides such as AlF3, CuF2 can be easily identi-
Hirooka and his co-workers [93] were able to detect SF6, fied by XRD with known reference spectra. Very often
S2F2, SF4, WF6, HF, SOF2, S02F2, CF4, SiF4, Si(CH3)2F2, the hydrated form of fluoride can be determined. X-ray
and COF2. The sample gas from the arcing vessel was fluorescence and atomic absorption analysis were often
liquefied by cooling with liquid nitrogen. The mnagnetic used to determine the elemental content, such as sulphur
resonance frequency of the 1 F was detected by a high- [23,45]. The size analysis of the solid by-products has
resolution NMR spectrometer. Due to the chemical shifts been carried out with scanning electron microscopy and
of the 19F in various host matrices, identification impact size samplers [45]. While the identification of
of the various arc by-products with the fluorine atoms pure metal fluoride by XRD can be easily achieved, the
in them was achieved. A list of the chemical shifts in identification of the compounds formed when arcing in
the 19F NMR frequencies with respect to the resonant SF6 takes place in the presence of polymeric materials
frequency of the CFC13 standard was given in reference or oil is very difficult. There was no reported techni-
[93]. The evolution of the by-products as a function of que to identify solid compounds formed in such conditions.
time and their chemical reactions with moisture and the
sample cell wall were also detected by the NMR technique. 7. FUTURE RESEARCH
6.3 Sampling Techniques 7.1 Toxicity
The lack of a cost effective technique to detect the Despite intensive investigations during the past 35
arc by-products in the field accurately, forces years on the subject of SF6 decomposition and the wealth
=T7.7 -r_ =-4- ; --- .I =It -;
rT- -.L 'No.5, October , ge6
T "-.;j --- -
I
..- . . i -; = -:- .-.L -; --
r,
-- I-lect!-
- i
--
.,- -
f-- I 1. 7.-. =t
--
I - -L 0 r-;
I .
of information obtained, there are still some unanswer- statistical tests. The study of surface flashover char-
ed questions regarding the issue of health and safety. acteristics and flashover initiation mechanisms as a
The-re is ample evidence to suggest that SOF2 is the result of arc by-product degradation would be a good
dominant by-product in all three modes of discharge first step.
and the toxicity of arced SF6 gas should be governed
by the toxicities of SOF2 since most other known by- 7.3 Diagnostics cnd Basic Mechanisms
products are less toxic than SOF2. In fact, some safe-
ty procedures were established on the basis of the level Gas analysis as a means to diagnose GIS has started
and toxicity of this compound. However, the investi- to receive attention from utilities. The non-invasive
gations of Oak Ridge National Laboratory [82-85] indi- nature of the technique is its biggest attraction. In
cated that the toxicity of sparked SF6 gas cannot be order to detect partial discharge inside the equipment,
accounted for by the toxicities of the commonly known a sensitive field instrument capable of detecting ppm
decomposition products. This means that there may be levels of SOF2 and S02F2 is required. The development
some highly toxic by-products present in minute quanti- of a cost effective instrument will enable direct meas-
ties which may be largely responsible for the cyto- urement of the gas in the compartment to assess in-
toxicity of sparked SF6. The most obvious candidate is ternal conditions without sampling and transportation
S2F1O [81a] which has a known TLV of 0.025 ppm, about to the laboratory for analysis. To aid in the inter-
40 times lower than the TLV of most known SF6 decom- pretation of data, further understanding of the forma-
position products. From a health and safety point of tion rates of the various decomposition products in an
view, the existence or absence of S2FlO in arced, ac corona or partial discharge is required.
sparked or corona discharged SF6 must be clarified. The
results of previous work appear to indicate that S2F10 On the issue of decomposition mechanisms, the funda-
is likel>v to be found in low energy discharges. Ilowever, mental processes leading to the dissociation of SF6
the recent finding of Pettinga [147] indicated that under a combined electric and thermal field will be of
trace amounts of S2FlO was also found in arced SF6 from continued interest because of the aging question. Al-
high- current discharges. though thermal aging of SF6 has been investigated ex-
In order to clarify the issue of S2FlO's existence, tensively, aging of SF6 and the associated components
further research is required to determine the quantity in a highly electrically stressed environment has not
of S2FlO formed in the three modes of discharge. To been investigated. The formation rates of SFs and SF5-
overcome difficulties in detecting very low concentra- from the SF6- need to be determined. Knowledge about
tions in the parts per billion range, further develop- the ion molecule interactions and ion chemistry are
ments in ultrasensitive detection techniques are re- required to determine the stability of SF6 at high EIN
quired. In addition to the well known GC/MS or GC/IR values as a function of time. The investigations of
techniques, 19F nuclear magnetic resonance may be a some basic electron, ion, and neutral processes may re-
good candidate technique since the NMR spectra and sult in useful information about the aging processes of
chemical shift of S2Fl0 have already been established gas-insulated equipment.
[68,168]. In conjunction with the search for S2F10,
the toxicity of arced SF6 should be thoroughly in- 8. ACKNOWLEDGEMENT
vestigated to verify the findings of the Oak Ridge
National Laboratory. Particular emphasis should be The author would like to thank Dr. L.G. Christophorou
given to SF6 subjected to corona discharge which, due and Dr. G.D. Griffin of the Oak Ridge National Labora-
to the non-equilibrium conditions, tends to generate a tory and Mr. M. Kurtz of Ontario Hydro for their val-
variety of chemically active products and radicals. uable discussions and comments during the preparation
of this manuscript.
Most previous work investigated the decomposition of
SF6 in the presence of impurities and electrode mate- 9. REFERENCES
rials. However, a majority of discharges in GIS occur
in the presence of organic materials such as epoxy [1] ACGIH, "-TLV' s
- Threshold Limit Values for Chemi-
spacers. On some rare occasions, SF6 discharges may cal Substances and Physical Agents inithe Workroom
occur in the presence of oil as a result of leakage at Environment with Intended Changes for 1981." Amer-
the SF6/oil cable interface. There is little informa- ican Conf of Government Industrial Hygienists
tion in the literature about the types of SF6 arced by- Assoc. Publ., Cincinnati, 1981.
products formed in the presence of organic materials.
Further research into the interaction between the dis- [2] ASTM, "Specification D 2472-81," Annual Books of
charge and the organic material would be desirable. ASTM Standards, 1984.
5.2 MvIate2rnia ComniatiitY5 [3] R. Agostino and D. L. Flamm, "Plasma Etching of Si
and SiO2 in SF6-02 Mixtures." J. Appl. Phys., 52
In the area of material compatibility, the question (1), p. 162, January 1981.
of long term aging in the presence of minute quantities
of SF6 decomposition products will continue to receive [4] A. J. Ahearn and N. B. Hannay, "The Formation of
considerable attention as utilities are seeking to Negative Ions of SF6," J. Chem. Phys., 21, No. 1,
stretch the service life of their equipment. Although pp. 119-124, January 1953.
the effects of decomposition products on solid insulat-
ing materials have been extensively investigated, most [5] D. R. Airey, "Axial and Radial Heat Transport in
of the research focused on the short term and large High Temperature SF6 Arc," J. Phys. D., 12, 1979.
dosage exposure situations. The extrapolation of the
short term results to the long term situations is not [6] Allied Corporation, "Sulphur Hexafluoride,"
obvious and may require further investigation. Although Technical Bulletin, 1984.
statistical tests and conventional aging experiments can
be carried out, a fundamental investigation on the de- [7] J. Amalric, "Practical Consequences of Research on
gradation mechanisms would be a more desirable approach the Decomposition by Arc of SF6,"' Revue Generale
because of the large number of parameters involved in de l'Electricite, pp. 45-78, June 1974.
Thu: SF6 decomposition in gas-insulated equ 7715e1Pm
-
[8] N. Aoki and K. Kumai, "Decomposition of SF6 by [22] J. F. Bott, T. A. Jacobs, "Shock Tube Studies of
Arci,Discharge - Influence of Water, Organic Matter SF6," J. Chem. Phys., Vol. 50, pp. 3850-3855, 1969.
and Electrode Materials upon Arc Decomposition
Products of SF6," Central Research Institute of [23] C. Boudene, J-L Cluet, G. Keib, and G. Wind,
Electric Power Industry Report E278004, Japan, "Identification and Study of Some Properties of
1978. Compounds Resulting from the Decomposition of SF6
under the Effect of Electrical Arcing in Circuit-
[9] P. G. Ashbaugh, D. W. McAdam and M. F. James, Breakers," Revue Generale Electricite - Special
"SF6 - Its Properties and Uses as a Gaseous In- Issue, June 1974.
sulator in Van der Graaf Accelerators," IEEE
Trans on Nuclear Science, NS-12, p. 266, June [24] K. P. Brand and J. Kopainsky, "Particle Densities
1965. in a Decaying SF6 Plasma," Appl. Phys., Vol. 16,
pp. 425-432, 1978.
[10] L. M. Babcock and G. E. Streit, "Ion Molecule
Reactions of SF6:Determination of I. P., A. P. [25] K. P. Brand, "Dielectric Strength, Boiling Point
and D(SF5-F) ," J. Chem. Phys.74 (10), p. 5700, and Toxicity of Gases - Different Aspects of some
May 1981. Molecular Properties," IEEE Trans. Elec. Insul.,
Vol. El-17, pp. 451-456, 1982.
[11] A. Baker, R. Dethlefsen, J. Dodds, N. Oswalt and
P. Vouros, "Study of Arc By-Products in Gas- [26] J.M. Braun, F. Y. Chu, and N. S. McIntyre,
Insulated Equipment," EPRI Report No. EL-1646, "Characterization of Degraded Epoxy Spacer Sur-
December 1980. faces by Electron Spectroscopy," CEIDP Annual
Report, 1984.
[12] A. Bargigia, et al., "Endurance Test on Solid
Insulating Materials Used for GIS Cast Insula- [27] J. M. Braun and F. Y. Chu, "Novel Low-Cost SF6
tors," Proc. Fourth Int. Symp. on High Voltage Arcing By-Product Detectors for Field Use in GIS, "
Engineering, Athens, September 1983. IEEE PES Summer Power Meeting, Vancouber, B.C.,
Paper 85 SM369-4, July 1985.
[13] B. Bartakova, J. Krump and V. Vosahlik, "Effect
of Electric Partial Discharge in SF6, " Electro- [28] R. Brockmann and R. von Olshausen, "Flashover
technichy Obzor, Prague, 67, p. 230, 1978. Voltage Dependence on the Tangential Field Strength
of Contaminated Insulator Surfaces in SF6,"
[14] W. Becher and J. Massonne, "Contribution to the Seventh International Conference on Gas Discharges
Study of the Decomposition of SF6 in Electric and Their Applications, pp. 280-283, September
Arcs and Sparks," ETZ-A, Vol. 91, No. 11. pp. 1982.
605-610, 1970.
[29] G. Bruno, P. Capezzuto, and F. Cramarossa, "In-
[15] S. W. Benson and J. Bott, "The Kinetics and organic Volatile Fluorides Obtained from Electrical
Thermochemistry of S2F10 Pyrolysis," Intl. J. Decomposition of Sulphur Hexafluoride in a Quartz
Chem. Kinetics, 1, p. 451, 1969. Tube," Journal of Fluorine Chemistry, Vol. 14,
pp. 115-129, 1979.
[16] G. Bernard, "Electrical Faults Mastery in High
Voltage SF6 Insulated Substations," Revue [30] G. Camilli, G. S. Gordon, and R. E. Plump,
Generale de l'Electricite, No. 4, pp. 231-249, "Gaseous Insulation for High-Voltage Transformers,"
April 1982. Trans. of AIEE, pp. 348-357, January 1952.
[17] T. R. Blackburn and R. W. Boswell., "Insulator [31] G. J. Carlson, J. M. Houston, W. D. Davis, M. P.
Surface Damage in SF6 Insulated Systems," Proc. Perry, and T. H. Rautenberg, "Fault Sensors for
Eighth Intl. Conf on Gas Discharge and their SF6 Equipment," Proceedings of the American Power
Applications, September 1984. Conference, Chicago, IL, Vol. 42, pp. 615-619,
April 21-23, 1980.
[18] J. P. Boeuf, P. Segur, M. Yousfi, E. Marode and
A. J. Davies, "Calculations of Transport Para- [32] J. Castonguay, A. Theoret, and R. Gilbert, "Arc
meters in SF6 for a Non Equilibrium Medium by Degradation of SF6 in the Presence of Polymeric
Monte-Carlo and Boltzmann Methods," Proc. Seventh Insulating Materials," Proc. IEEE Int'l. Symp.
International Conference on Gas Discharges and Electr. Insul., 1978.
Their Applications, pp. 347-349, September 1982.
[33] J. Castonguay, "Kinetics of the Arc Decomposition
[19] S. A. Boggs, F. Y. Chu, and N. Fujimoto, "Dis- of SF6-Gas Mixtures," Proc. IEEE Intl. Symp. on
connect Switch Induced Transients and Trapped Elect. Insulation, June 1980.
Charge in GIS," IEEE Trans. Power and Systems,
PAS 101, October 1982. [34] J. Castonguay and R. Larocque, "Low Cost Technique
for Field Analysis of SF6 Decomposed By-Products,"
[20] S. A. Boggs, F. Y. Chu, and M. Mashikian, "Gas- Proc. of Intl. Symp. on Gas Insulated Substations,
Insulated Substation Reliability: Present September 1985.
Status and Future Trends," EPRI Final Report
EL-4422, February 1986. [35] L. Centurioni, G. Camilli, C. Luccini, and P.
Vismara, "Behavior of Polymeric Insulating Mate-
[21] P. C. Bclin and M. D. Hopkins, "Constraints on rials Exposed to Arced SF6," Proc. Fourth Intl.
Fault Repair of GIS, " Proc. of International Symp. on High Voltage Engineering, Athens,
Symposium on Gas Insulated Substations, Toronto, September 1983.
September 1985.
T -ECE T ,ct1-X c EIectra
; I'n Vic,. ET-MA No5. , October- 1986
[36] C. L. Chen and P. J. Chantry, "Photo-Enhanced [49] R. N. Compton, P. W. Reinhardt, and C. D. Cooper,
Dissociative Electron Attachment in SF6 and Its "Collisional Ionization between Fast Alkali Atoms
Isotopic Selectivity," J. Chem. Phys., Vol. 71 and Selected Hexafluo Ride Molecules," J. Chem.
(10), p. 3897, November 1979. Phys., Vol. 68(5), p. 2023, March 1978.
[36a] L. G. Christophorou, (Private Communication, [50] A. H. Cookson, "Gas Insulated Cables," IEEE
1986). Trans. Elect. Insul., Vol. E1-20, No. 5, pp.859-
890, October 1985.
[37] L. G. Christophorou, Electron-Molecule Inter-
actions and their Applications, Academic Press, [51] J. H. Cooper, W. H. South, and J. F. Shimshock,
Ed: L. G. Christophorou, Vols. 1,2, 1984. "Field Test Experience and Diagnostic Techniques
on 1200 kV GIS Components," CIGRE, Paris, Paper
[38] L. G. Christophorou, D. L. McCorkle, and A. A. 23-07, 1984.
Christodoulides, "Electron Attachment Processes,"
Electron Molecule Interactions and their Appli- [52] 0. Coufal, A. Fidler, and M. Stevanova, "Composi-
cations, Ed. L. G. Christophorou, p. 478, 1984. tion of SF6 for Temperatures up to 6000 K and
Pressures from 0 to 1 MPa," ACTA Technica,
[38a] L. G. Christophorou, I. Sauers, D. R. James, (Ceskoslovensk Akademie Ved), Vol. 30, No. 5,
H. Rodrigo, M. 0. Pace, J. G. Carter and S. R. pp. 575-584, 1985.
Hunter, "Recent Advances in Gaseous Dielectrics
at Oak Ridge National Laboratory," IEEE Trans. [53] T. W. Dakin, "Application of Epoxy Resins in
Electr. Insul. Vol. El-19, pp. 550-566, 1984. Electrical Apparatus," IEEE Tran. Elect. Insul.,
vol. EI-9, 4, December 1974.
[39] F. Y. Chu and C. K. Law, "Effects of Power Arc
Faults in Gas Insulated Equipment," Proc. of [54] T. Dakin, "Thermal Aging of Dielectric Gas,"
Sixth Intl. Conference on Gaseous Discharges, Gaseous Dielectrics II, L. G. Christophorou (Ed.)
Edinburgh, IEE Publ. 189, 1980. Pergamon Press. pp. 283-293, 1980.
[40] F. Y. Chu, "Experience with Solid Arcing By-Prod- [55] G. Diebold and D. McFadden, "Noise Reduction in
ucts in SF6 Insulated Equipment," Transmission EPR Discharge Flow Studies," Rev. Sci. Instrum.,
and Distribution, pp. 54-58, April 1981. Vol. 50, p. 157, 1979.
[41] F. Y. Chu, "Fault Analysis in Gas Insulated [56] M. S. Dincer and G. R. Raju, "Monte Carlo
Equipment," EPRI Final Report, Vol. EL-2248, Simulation of the Motion of Electrons in SF6 in
February 1982. Uniform Electric Field," J. Appl. Phys., Vol. 54
(11), p. 6311, 1983.
[42] F. Y. Chu and R. M. Massey, "Thermal Decomposition
of SF6 and SF6-air Mixtures in Substation Environ- [57] J. J. Dodds, "Moisture Measurements in GIS,"
ments," Gaseous Dielectrics III, L. G. Doble Client Conference, Boston, 1985.
Christophorou (Ed.) Pergamon Press, pp. 410-417,
1982. [58] J. J. Dodds and L. E. Taylor, "550 kV Type ELK
Breaker Failure," Paper presented at the 1986
[43] F. Y. Chu and F. Lutz, "SF6 Decomposition in a Doble Client Conference, Boston, 1986.
Plasma Arc for Various Electrode Materials,"
Proc. of 6th Intl. Symp. Plasma Chem., p. 231, [59] M. Dolezal and F. Posmure, "Composition, Toxicity
July 1983. and Filtration of the Products of SF6 Generated in
the Arc Quenching Devices of Power Circuit
[44] F. Y. Chu, H. Stuckless, and J. M. Braun, Breakers," Electrotechnichy Obzor, Vol. 68, No. 4,
"Generation and Effects of Low Level Contamination pp. 216-220, 1979.
in SF6 Insulated Equipment," Gaseous Dielectrics
IV, L. G. Christophorou and M. 0. Pace (Eds.) [60] N. Dominelli, "Analysis of Gaseous SF6 Decomposi-
Pergamon Press, pp. 462-472, 1984. tion Products," Proc. of Intl. Symp. on Gas
Insulated Substations, September 1985.
[45] F. Y. Chu, "Procedures for Dealing with Solid
Arcing By-Products from Gas Insulated Substations," [61] F. J. Dwyer and K. Miners, "SF6 Gas Analysis - an
Can. Electr. Assoc. Final Report 128-T-239, Effective Means of Saving Time and Money," Proc.
April 1985. Southeastern Electric Exchange Annual Conference,
April 1982.
[46] F. Y. Chu, F. Lutz, J. M. Braun, and H. A.
Stuckless, "Effects of Power Arc Faults in Gas [62] C. E. Easterly, C. S. Dudney, G. D. Griffin, and
Insulated Substations," Proc. CIGRE Symp. on High P. J. Walsh, "Health Effects of Dielectric Gases:
Current in Power Systems, Brussels, 1985. Preliminary Report," Oak Ridge National Laboratory
Report TM-8353, August 1982.
[47] F. Y. Chu, J. M. Braun, and R. Seethapathy,
"Degradation Mechanisms for Epoxy Insulators Ex- [63] D. Edelson, K. B. McAffe, "Notes on the Infrared
posed to SF6 Arcing By-Products," Proc. IEEE Intl. Spectrum of SF6," J. Chem. Phys., Vols. 19, 10,
Symp. Electr. Insul., June 1986. p. 1311, October 1951.
[48] F. Y. Chu, J. M. Braun, R. Seethapathy, and L. [64] D. Edelson, et al., "Electrical Decomposition of
Gonzalex, "A New Approach to Moisture Measurement SF6," Industrial and Engineering Chemistry, Vol.
and Control in Gas Insulated Switchgear," Proc. 45, pp. 2094-2096, September 1953.
of Canadian Electrical Association Spring Meeting,
Toronto, March 1986.
Chu: SF6cdecomposittic. in g-as-i-;nIated ei pme-t
[65] R. E. Eibech and W. H. Nears, "Sulfur Hexa- [82] G. D. Griffin, B. A. Owen, C. E. Easterly, and
fluoride," Kirk-Othmer Encyc. Chem. Tech., 3rd P. J. Walsh, "In Vitro Toxicity Screening of
Ed., Vol. 10, p. 799, 1980. Dielectric Gases Using Mammalian Cells,"
Gaseous Dielectrics III, L. G. Christophorou
[66] M. Ermel, Elektrotech. Z-A, Vol. 96, p. 231, (Ed.) Pergamon Press, pp. 420-426, 1982.
1975.
[83] G. D. Griffin, I. Sauers, L. G. Christophorou,
[67] F. C. Fehsenfeld, "Electron Attachment to SF6," C. E. Easterly, and P. J. Walsh, "On the Toxicity
J. Chem. Phys., Vols. 53, 5, p, 2000, 1970. of Sparked SF6," IEEE Transactions on Electrical
Insulation, Vol. EL-18, No. 5, pp. 551-552,
[68] E. G. Finer and R. K. Harris, "High Resolution October 1983.
Nuclear Magnetic Resonance Spectroscopy of Penta-
flurosulphur Compounds III," Spectrochimica [84] G. D. Griffin, I. Sauers, C. E. Easterly, and
Acta, Pergamon Press, Vol. 24A, p. 1939, 1968. H. W. Ellis, "Decomposition of Sparked SF6 and
[69]
Cytotoxicity of SF6 Decomposition Products,"
T. N. Fischer, Jr., "Experience with Contaminated Gaseous Dielectrics IV, L. G. Christophorou and
SF6 Gas," Transmission and Distribution, M. 0. Pace (Eds.) Pergamon Press, pp. 261-272,
October 1975. 1984.
[70] T. N. Fischer, Jr., "Utility Experience with Gas [85] G. D. Griffin, C. E. Easterly, I. Sauers, H. W.
Insulated Equipment," EPRI Final Report, Vol. Ellis, and L. G. Christophorou, "Cytotoxic
El-80-8-LD, December 1980. Activity of Spark-Decomposed SF6 and Analysis of
Cytotoxic Contributions of Individual Spark
[71] L. C. Frees, I. Sauers, H. W. Ellis, and L. G. Decomposition Products," Toxicological and
Christophorou, "Positive Ions in Spark Breakdown Environmental Chemistry, Vol. 9, pp. 139-166,
of SF6," J. Phys. *D: Appl. Phys., Vol. 14, 1984.
pp. 1629-1642, 1981.
[86] R. Grob, J. Casanovas, M. Besbes, L. Dechamps,
[72] W. Frie, "Berechnung der Gaszusammensetzung und and J. Perret, "Ineraction Between SF6 and
der Materialfunktionen," Z. Phys., Vol. 201, p. Water Exposed to Gamma Radiation or to Partial
269, 1967. Discharge," Conf. Rec. IEEE Intl. Symp. Electr.
Insul., Vol. 5, pp. 147-150, 1984.
[73] L. S. Frost and R. W. Liebermann, "Compositions
and Transport Properties Of SF6 and Their Use in [87] B. F. Hampton, "Current-Carrying and Short
a Simplified Enthalpy Flow Arc Model," Proc. Circuit Tests on EHV Cables Insulated with SF6
IEEE, Vol. 59, pp. 474-484, 1971. Gas," IEEE Trans. PAS, PAS-88, No. 2, February
1969.
[74] S. Fukuda, T. Ushio, T. Ito, and T. Miyamoto,
"Recent Studies on Arc Behaviour Around Current [88] J. M. Hanrahan and A. R. Paterson, "Adsorption-
Zero in SF6 Gas," CIGRE, paper 13-03, 1968. Desorption Gas Chromatographic-Infrared Deter-
mination of Trace Disulfur Decafluoride in Sulfur
[75] S. Fukuda, "Current-Carrying and Short Circuit Hexafluoride," Journal of Chromatography, Vol.
Tests on EHV Cables Insulated with SF6 Gas," 193, pp. 265-270, 1980.
IEEE Trans. PAS, PAS-88, No. 2, pp. 147-156,
February 1969. [89] W. Hertz, H. Motschmann, and H. Wittle, "Investi-
gation of the Properties of SF6 as an Arc
[76] Ganger, J. Vigreux, "May the GIS SF6 Insulation Quenching Medium," Proc. IEEE, Vol. 59, pp. 485-
be Considered as Self-Restoring?," CIGRE WG 492, 1971.
15-03, July 1980.
[90] W. M. Hickam and D. Berg, "Negative Ion Formation
[77] P. Gayet, G. Bernard, D. Demontussaint, and J. and Electric Breakdown in Some Halogenated
Lalot, "Research into a Voltage Test Procedure Gases," J. Chem. Phys., Vols. 29, 3, p. 517,
for Site Testing of Metalclad Substations with 1958.
Maximum Efficiency," CIGRE, Paris, Paper 23-05,
1984. [91] D. L. Hildenbrand, "Mass Spectrometric Studies of
Some Gaseous Fluorides", J. Physical Chemistry,
[78] B. Gerasimov and T. Sidorkina, Zhurnal Prikl. Vol. 77, No. 7, 1973.
Khim, Vol. 37, p. 2063, 1964.
[92] T. Hirano, "Effects of High Pressure SF6 Dissolved
[79] M. Goldman and A. Goldman, "Corona Discharges," in an Insulating Oil in dc Conduction Currents and
Gaseous Electronics, Academic Press, Vol. 1, ac Partial Discharges," J. Appl. Phys., Vol. 55(1),
p. 219, 1978. p. 9, January 1984.
[80] H. Grasselt, W. Ecknig, and H. J. Polster, [93] K. Hirooka, H. Kuwahara, M. Noshiro, and Y.
"Applications of Gas Chromatography for the Jitsugiri, "Decomposition Products of SF6 Gas by
Development of SF6 Insulated Switchgear and High Current Arc and Their Reaction Mechanism,"
Equipment," Elektrie, Vol. 32, pp. 369-371, 1978. Electrical Engineering in Japan, Vol. 95, No. 6,
1975.
[81] H. W. Graybill, "Workshop on User Experience with
Gas Insulated Substations," EPRI Report EL-2189, [94] K. Hirooka and M. Shirai, "Thermal Characteristics
December 1981. of SF6," J. Chem. Soc. Japan, Vol. 2, pp. 165-
169, 1980.
[81a] G. D. Griffin (Private Communication, 1986).
bECE z~cta'. rt ElDe_t'L cl In_-dation Vol. ,-1Nso.. October- l9S6j
X
[95] A. Hollinger. "Analyzer for Decomposition Prod- [110] E. E. Kunhardt, "Basic Topics of Current Inter-
ucts and Moisture in SF6 Gas," Can. Electr. Assoc. est to Switching for Pulse Power Applications,"
Final Report 086-T-176, June 1984. Gaseous Dielectrics IV, L. G. Christophorou and
M. 0. Pace (Eds.) Pergamon Press, p. 213, 1984.
[96] R. W. Holmberg, "Problems with SF6 Decomposition
in Large Electro-static Accelerators," EPRI Report [111] S. Kusumoto, S. Itoh, Y. Tsuchiya, H. Mukae, and
EL80-8-LD, December 1980. K. Takahashi, "Diagnostic Technique of Gas In-
sulated Substation by Partial Discharge Detec-
[97] M. Honda, H. Aoyagi, M. Koya, N. Kobayashi, and tion," IEEE Transactions on Power Apparatus and
S. Nemju, "Deterioration of Eposy Mold Insulators Systems, PAS-99, p. 1456, July/August 1980.
due to Voltage Endurance," Gaseous Dielectrics
III. L. G. Christophorou (Ed.) Pergamon Press, [112] R. T. Lagermann and E. A. Jones, "The Infrared
pp. 322-326, 1982. Spectrum of SF6," J. Chem. Phys., Vol. 19,5, p.
534, May 1951.
[98] E. Hubler, et al., "Comparative Evaluation of
Epoxy Resin Systems for SF6 Insulated Substation [113] J. R. Laghari and A. H. Qureshi, "Surface Flash-
Components," Proc. IEEE Insul. Symp. Xian, China over of Spacers in Compressed Gas Insulated
Paper CH2115-4/85, 1985. Systems," IEEE Trans. Electr. Insul., Vol. El-16,
No. 5, pp. 373-387, 1981.
[99] A. A. Hudson, "Degradation of SF6 in Electrical
Equipment, Toxic and Corrosive Effect," Elec- [114] J. R. Laghari, "Spacer Flashover in Compressed
trical Research Assoc. Report. No. 5185, Surrey, Gases," IEEE Transactions on Electrical Insula-
England, 1967. tion, Vol. El-20, No. 1, pp. 83-92, February
1985.
[100] International Electrical Commission, "Guide
to the Checking of Sulphur Hexafluoride (SF6) [115] J.. Lampe, H. Latour-Slowikowska, and J.. Slowikow-
taken from Electrical Equipment," International ska, "Study on Metal Fluoride Products Formation
Electrical Commission - Publication 480, Caused by an Electric Arc in SF6,"1 Gaseous Di-
1974. electrics III L. G. Christophorou (Ed.) Perga-
mon Press, p. 433, 1982.
[101] R. B. Jackson and K. B. Miners, "SF6 Gas Analysis
Service - Analytical Data and Case Histories'," [116] H. Latour-Slowikowska, J. Slowikowska, L.
the Forty-Ninth Annual Intl. Conf. Doble Clients, Dybowski, and J. Lampe, "Some Test Results on the
Boston, April 1982. Effect of SF6 Decomposition Products on Super-
ficial Properties of Epoxy Resin Compositions,"
[102] G. Jackson, et al., "Surface Studies of Dielectric IEEE Proc. Insulation Symp., 1978.
Materials used in Spark Gaps," J. Appl. Phys.
Vol. 55, No. 1, pp. 262-268, 1984. [117] L. Latour-Slowikowska, J. Lampe, and J. Slowilow-
ska, "On Reactions Occurring in the Gaseous
[103] F. J. J. G. Janssen, "Measurements at the Sub-ppm Phase in Decomposed SF6," Gaseous Dielectrics IV,
Level of Sulphur-fluoride Compounds Resulting L. G. Christophorou and M. 0. Pace (Eds.)
from the Decomposition of SF6 by Arc Discharge," Pergamon Press, pp. 286-291, 1984.
KEMA Scientific and Techmical Reports, Nether-
lands, Vol. 2, No. 2, 1984. [118] W. Leeds, F. Browne, and A. Strom, "The Use of
SF6 -Cor High Power Arc Quenching," AIEE Trans.,
[104] J. Jarmula, "Effect of SF6 and its By-Products Vol. 76, p. 906, 1957.
Evolved by Electric Arc on Metals and Insulating
Materials," Proc. Inst. Electrotech. (Poland), [119] D. A. Leep and R. J. Van Brunt, "Safety Con-
Vol. 23, No. 88, pp. 97-102, 1975. siderations for Handling SF6 Used in Electrical
Equipment," National Bureau of Standards
[105] J. Jarmula, "Laboratory Testings of Electrical Washington, D.C., NBSIR 80-2063, June 1980.
Insulating Materials Surface Resistance to
Hydrogen Fluoride Vapors as Selection Criteria [120] D. Lester and L. A. Greenberg, "The Toxicity of
for SF6 filled Circuit Breaker Insulators," SF6," Arch. of Ind. Hyg. Occupational Med., Vol.
Ann. Rep. CEIDP, pp. 457-462, 1981. 2, p. 348, 1950.
[106] L. E. Kline, D. K. Davies, C. L. Chen, and [121] D. Lester, "General Considerations on the Toxicity
P. J. Chantry, "Dielectric Properties for SF6 and of Gaseous Dielectrics," AIEE Trans., pp. 1183-
SF6 Mixtures Predicted from Basic Data," J. Appl. 1185, December 1957.
Phys., Vol. 50, pp. 6789-6796, November 1979.
[122] R. J. Lewis and R. L. Tatken, "Registry of Toxic
[107] W. Kolbe, "Effects of Arcing Faults on SF6 Metal Effects of Chemical Substances," National Inst.
Clad Equipment," Dr. Dissertation, Technical for Occup. Safety and Health Publication, 1982.
University, Darmstadt, Germany, 1976.
[123] R. W. Liebermann and J. J. Lowke, "Radiation
[108] J. Kopainsky, "Influence of the Arc on Breakdown Emission Coefficient for SF6 Arc Plasma," J.
Phenomena in Circuit Breakers," Current Inter- Quant. Spectrosc. Radiat. Transfer, Vol. 16, pp.
ruption in High Voltage Networks, Plenum Press, 253-264, 1976.
p. 329, 1978.
[124] G. D. Lockyer, "Clean Up of SF6 Gaseous By-Prod-
[109] J. Kulseta, A. Rein, and P. A. Holt, "Arcing in ucts," EPRI Final Report, EL80-8-LD, 1980.
SF6 Insulated Equipment, Decomposition Products
and Pressure Rise," Proc. Nordic Insulation
Symposium, NORD-IS, Vol. 78, June 1978.
Chu: SF5 decomposition in gas-in-suated equipment
[125] J. J. Lowke and R. W. Liebermann, "Predicted Arc [142] S. Okuda, Y. Ueda, Y. Murai, T. Miyamoto, Y. Doi,
Properties in SF6," J. Appl. Phys., Vol. 42, p. and C. Uenosono, "Spectroscopie Approach to the
3532, 1971. Analysis of High Current Arcs in SF6," IEEE
Transactions on Plasma Science, PS-8, No. 4, pp.
[126] J. J. Lowke and R. W. Liebermann, "Negligible 395-399, December 1980.
Effect of Negative Ions on the Constriction of
Arcs in SF6," J. Appl. Phys., Vol. 43, pp. 1991- [143] T. R. Ophel, D. C. Weisser, A. Cooper, L. K.
1994, 1972. Fifield, and G. D. Putt, "Aspects of Breakdown
Product Contamination of SF6 in Electrostatic
[127] J. J. Lowke, "Radiative Energe Transfer in Cir- Accelerators," Nuclear Instruments and Methods,
cuit Breaker Arcs," Current Interruption in High Vol. 217, pp. 383-396, 1983.
Voltage Networks, K. Ragaller (Ed.), Plenum
Press, 1978. [144] D. K. Padma and A. R. Murthy, "Thermal Decomposi-
tion of SF6," J. Fluorine Chem., Vol. 5, pp. 181-
[128] J. L. Lyman, "A Model for Unimolecular Reaction 194, 1975.
of SF6," J. Chem. Phys., Vol. 67, p. 1868,
September 1977. [145] J. Pankuch, "Hlandling of Arc Contaminated SF6
Gas in a Failed Circuit Breaker," Duquesne Light
[129] J. Manion, J. Philosophos, and M. Robinson, "Arc Company Report, 1975.
Stability of Electronegative Gases," IEEE Trans.
Electrical Insulation, El-1, Vol. 1, 1967. [146] H. Patrunky and W. Olsen, "Service Experience and
Maintenance Procedures," Presented to the Inter-
[130] M. Mastroianni, "SF6 Analysis is the Key to national Symposium on Gas Insulated Substations,
Maintenance," Electrical World, December 1980. Toronto, September 1985.
[131] R. S. McDonald, "Analysis of SF6 Arc By-Products [147] J. A. J. Pettinga, "Full Scale High Current Model
by FT-IR," EPRI Report, EL80-8-LD, December Tests on Busbar Constructions for CIS," Proc. of
1980. CIGRE Symp. High Current in Power System. pp.
506-585, 1985.
[132] J. P. McGeehan, B. C. O'Neill, A. N. Prasad, and
J. D. Craggs, "Negative Ion/Molecule Reactions [148] E. Pfender, "Electric Arcs and Arc Gas Heaters,"
in SF6," J. Phys. D., Vol. 8, p. 153, 1975. Gaseous Electronics, Academic Press, Vol. 1,
p. 291, 1978.
[133] K. B. Miners and M. Mastroianni, "SF6 Gas Analysis
of In-Service GIS, Results and Interpretations," [149] A. Picard and G. Turban, "Plasma Etching of Re-
Proc. Workshop on User Experience with GIS, EPRI fractory Metals and Silicon in SF6 and SF6-02.
Report EL-2189, December 1981. An Analysis of the Reaction Products," Plasma
Chemistry and Plasma Processing, Vol. 5, No. 4,
[1341 R. E. Minkwitz, "Some Experience with SF6 Insula- p. 333, December 1985.
ted Equipment," EPRI Final Report, EL80-8-LD,
December 1980. [150] F. Pinnekamp and L. Niemeyer, "Qualitative Model
of Breakdown in SF6 in Inhomogeneous Gaps," J.
[135] G. Mitchel, J. Castonguay, and N. G. Trinh, Phys. D: Appl. Phys. (UK), Vol. 16, pp. 1293-
"Practical Thermodynamics of SF6 Recovery from 1302, 1983.
SF6/N2 or SF6/Air Mixtures," Presented to the
International Symposium on Gas Insulated Sub- [151] M. C. Pohl, "Analytical Procedures for Detecting
stations, Toronto, September 1985. the Decomposition Products of SF6," EPRI Report,
EL80-8-LD, December 1980.
[136] A. P. Modica, "Shock Tube Study of SF6 and SF5CI
Equilibrium Decomposition," J. Phys. Chem., Vols. [152] D. Rapp and P. Englander-Golden, "Ionization by
77, 23, p. 2713, 1973. Electron Impact I," J. Chem. Phys., Vol. 43, p.
1464, 1965.
L137] Y. Murakami and S. Menju, "Fundamental Techniques
in Gas-Insulated Apparatus," Toshiba Review, Vol. [153] W. Ruegsegger, F. K. Kneubuhl, and H. J. Schotzen,
117, p. 5, September 1978. "Mass Spectrometry of High Pressure Arcs in Air
and SF6," Appl. Phys. B., Vol. 31, pp. 9-13,
[138] J. B. Neilson, "Energy Measurement of Transient 1983.
Ground Rise Pulses," Presented to the Internation-
al Symposium on Gas Insulated Substations, [154] H. M. Ryan, D. Milne, and C. W. Powell, "Site
Toronto, September 1985. Testing and the Evaluation of a Technique to
Differentiate Between a Gas or Spacer Flashover
[139] L. Niemeyer and F. Pinnekamp, "Surface Discharge in SF6 GIX," Proc. of Intl. Symp. on Gas Insulat-
in SF6," Gaseous Dielectrics III, L. G. ed Substations, September 1985.
Christophorou (Ed.) Pergamon Press, p. 379, 1982.
[155] E. W. Saaski and J. C. Hartl, "A Solid State Gas
[140] L. Neimeyer, "Leader Inception in SF6," Gaseous Density Monitor for Gas Insulated Substation
Dielectrics IV, L. G. Christophorou and M. 0. Equipment," IEEE Transactions on Power Apparatus
Pace (Eds.) Pergamon Press, p. 156, 1984. and Systems, PAS-101, No. 7, pp. 1921-1927, July
1982.
[141] T. Nitta, et al., "Factors Controlling Surface
Flashover in SF6 Gas Insulated Systems," IEEE [156] 0. Sala and G. Spalek. "The Nuclear Structure
Trans. Power Appar. Syst., PAS-97, No. 3, pp. Facility at the Univer-sity of Sao Paolo," Nuclear
959-968, May 1978. Instruments and Methods, Vol. 122, pp. 213-235,
1974.
?ct *r- kl = Vo 1 . E-2
I 19 H,J
Ot r
[157] R. Samuelsson, "Corrosion and Surface Resistivity [171] H. A. Stuckless, J. M. Braun, and F. Y. Chu,
Changes in Different Insulating Materials Caused "Degradation of Silica-Filled Epoxy Spacers by
by Decomposed SF6 and Moisture," ASEA Journal, Arc Contaminated Gasses in SF6 Insulated Equip-
vol. 52, No. 6, pp. 133-136, 1979. ment," IEEE Trans. Power App. and Systems, Vol.
PAS-104, p. 359, December 1985.
[158] I. Sauers, L. G. Christophorou, L. C. Frees, and
H. W. Ellis, "Aspects of Environmental Effects [172] T. Suzuki, S. Nokayama, and T. Yoshimitsu,
of Dielectric Gases," Gaseous Dielectrics III, "Degradation of Insulating Materials Including
L. G. Christophorou (Ed.) Pergamon Press, p. 387, SiO2 due to SF6 Gas Dissociation Products,"
1982. IEEE Trans. Electr. Insul., EI-15, pp. 53-58, 1980.
[159] I. Sauers, H. W. Ellis, L. C. Frees, and L. G. [173] T. Suzuki, H. Yoshida, A. Koyama, and S.
Christophorou, "Studies of Spark Decomposition Tomimuro, "Degradation Process of Grease Due to
Products of SF6 and SF6 Perfluorocarbon Mix- SF6 Dissociation Products," IEEE Trans. Power
tures," Gaseous Dielectrics III, L. G. Christo- Appar. Syst., PAS-101, No. 8, pp. 2805-2809,
phorou (Ed.) Pergamon Press, pp. 387-401, 1982. 1982.
[160] I. Sauers, L. G. Christophorou, and S. M. Spyrou, [174] P. Swarbrick, "Compositions and Properties of
"Negative Ion Formation in SF6 Spark-By-Prod- an SF6 Arc Plasma," B. J. Appl. Phys., Vol. 18,
ucts," Gaseous Dielectrics IV, L. G. Christo- pp. 419-426, 1967.
phorou and M. 0. Pace (Eds.) Pergamon Press,
pp. 292-309, 1984. [175] V. Tahiliani, W. J. Lannes, and K. B. Miners,
"Diagnostic Techniques to Detect Abnormal Operat-
[161] I. Sauers, H. W. Ellis, L. G. Christophorou, ing Conditions in GIS," CIGRE Paper 23-08, 1984.
G. D. Griffin, and C. E. Easterly, "Spark De-
composition of SF6; Toxicity of By-Products," [176] D. Tedford, L. C. Campbell, and J. T. Pender,
Oak Ridge National Laboratory Report ORNL/TM- "Modern Research and Development in SF6 Swtich-
9062, October 1984. gear," Electronics and Power, Vol. 29, No. 10,
p. 719, Ocotber 1983.
[162] J. L. Schmidt and W. E. Harper, "SF6 Gas Condi-
tioner/Analyzer Keeps Mini-Subs On-Line," [177] A. Theoret and R. Gilbert, "Occupational Hazards
Allis-Chalmers Engineering Review, Vol. 39, and Other Problems Relating to the Use of SF6 in
No. 2, 1974. Metalclad Substations," Can. Electr. Engr. Journ.,
Vol. 4, pp. 24-28, 1976,
[163] W. F. Schmidt, H. Jungblut, D. Hansen, and
H. Tagashira, "Ionic and Electronic Processes [178] R. Thorburn, "Permanent Dissociation of SF6 in
in SF6," Gaseous Dielectrics II, L. G. Christo- Corona Discharge," Nature, Vol. 175, p. 423,
phorou (Ed.) Pergamon Press, pp. 1-11, 1980. 1955.
[164] E. Schulz-Guide, "Temperature Determination for [179] S. Toku3ama, Y. Yushioka, F. Nakazima, and Y.
Arcs in SF6 Accounting for Demixing," J. Phys. Arikawa, "Analysis of SF6 Gas Decomposition due
D., Vol. 13, pp. 793-803, 1980. to Current Interruption," 2nd Intl. Conf. on Gas
Discharges, IEE Conf. Publ., Vol. 90, p. 200,
[165] W. C. Schumb, J. Trump, and G. Priest, "Effect 1972.
of High Voltage Electrical Discharge on SF6,"
Ind. Engr. Chem., Vol. 41, pp. 1348-1351, 1949. [180] S. Tominaga, H. Kuwahara, and K. Hirooka, "In-
fluence of Arc-Decomposed SF6 Gas on Cast Epoxy
[166] M. C. Siddagangappa and R. J. Van Brunt, "Decom- Insulators for Gas Insulated Switchgear," IEEE
position Products from Corona in SF6/N2 and Trans. PAS, PAS-98, pp. 2107-2114, 1979.
SF6/02 Mixtures," Proc. 8th Intl. Conf. on Gas
Discharges and Their Applications, Oxford, UK, [181] S. Tominaga, H. Kuwahara, K. Hirooka, and T.
September 1985. Yoshioka, "SF6 Gas Analysis Techniques and its
Application for Evaluation of Internal Conditions
[167] M. C. Siddagangappa and R. J. Van Brunt, "In- in SF6 Gas Equipment," IEEE Trans. PAS, PAS-100,
fluence of Oxygen on the Decomposition Rate of No. 9, pp. 4196-4206, September 1981.
SF6 in Corona," Proc. of IEEE Elec. Insul. Symp.,
June 1986. [182] P. R. Tunnicliffe, "Accelerator Research and
Engineering," AECL Symp. on Engr. Development in
[168] R. R. Smardzewski, R. E. Noftle, and W. B. Fox, Nuclear Power, AECL, Chalk River, Canada, pp.
"Argon Matric Raman and Infrared Spectra of 67-74, 1971.
SF5Cl, SF5Br and S2Flo," Journal of Molecular
Spectroscopy, Vol. 62, pp. 449-457, 1976. [183] T. Ushio, I. Shimura, and S. Tominaga, "Practical
Problems on SF6 Gas Circuit Breakers," IEEE
[169] R. R. Smardzewski and W. B. Fox, "Vacuum Ultra- Winter Power Meeting, New York, Paper 71 TP 64-
violet Photolysis of SF6 and its Derivatives in PWR, January 1970.
Argon Matrices. The Infrared Spectrum of
S2FlO," J. Chem. Phys., Vol. 67, 5, p. 2309, [184] R. J. Van Brunt and D. A. Leep, "Corona-Induced
September 1977. Decomposition of SF6," Gaseous Dielectrics IV,
L. G. Christophorou and M. 0. Pace (Eds.)
[170] H. Specht and F. Brubach, "Inhalation of Sulphur Pergamon Press, N.Y., pp. 402-409, 1982.
Hexafluoride," Science, Vol. 114, p. 662, 1951.
Chu; SF6 decomposition i_n gas-i #sl*ated equipment
[185] R. J. Van Brunt and M. Mashikian, "Mechanisms [196] R. L. Wilkins, Theoretical Evaluation of Chemical
for Inception of DC and 60-Hz AC Corona in SF6," Propellents, Prentice Hall, p. 100, 1963.
IEEE Trans. Electrical Insulation, EL-17, No. 2,
p. 106, 1982. [197] R. L. Wilkins, "Thermodynamics of SF6 and its
Decomposition and Oxidation Products," J. Chem.
[186] R. J. Van Brunt and M. Mashikian, "Role of Phys., Vol. 51, pp. 853-854, 1969.
Photodetachment in Initiation of Electric Dis-
charges in SF6 and 02," J. Appl. Phys., 54(6), [198] J. K. Wilmshurst and H. J. Bernstein, "The In-
pp. 3074-3079, June 1983. frared and Raman Spectra of Disulphur Deca-
fluoride," Can. J. Chem., Vol. 35, p. 191, 1957.
[187] R. J. Van Brunt, T. C. Lazo, and W. E. Anderson,
"Production Rates for Discharge Generated SOF2, [199] J. K. Wittle, J. M. Houston, G. J. Carlson,
S02F2 and S02 in SF6 and SF6/H20 Mixtures," W. D. Davis, A. M. Itani, and J. G. Jewell,
Gaseous Dielectrics IV, L. G. Christophorou and "Fault Detection Sensors for Gas Insulated
M. 0. Pace (Eds.) Pergamon Press, pp. 276-285, Equipment," EPRI Report No. EL-2249, February
1984. 1982.
[188] R. J. Van Brunt, "Production Rates for Oxy- [200] S. Wock, "Discussion: On the Toxicity of SF6
fluorides SOF2, S02F2 and SOF4 in SF6 Corona Insulating Gas," IEEE Trans. Electr. Insul.,
Discharge," J. of Reserach, National Bureau of EI-19, p. 156, April 1984.
Standards, Vol. 90, No. 3, pp. 229-253, May 1985.
[201] R. E. Wootton, et al., "Gases Superior to SF6
[189] T. Van Slambrouck, "Use of Tunable Diode Lasers for Insulation and Interruption," EPRI Final
for the Determination of Low Level Gas By-Products Report EL-2620, September 1982.
in Arced SF6," Spectra Physics Report (unpublish-
ed), September 1983. [202] K. L. Wray and E. V. Feldman, "Shock Tube Study
of the Decomposition Kinetics of S02F2, " J.
[190] A. van Roggen, P. F. Pugh, and K. P. Brand, "On Chem. Phys., Vol. 34, No. 8, pp. 3445-3449,
the Toxicity of SF6 Insulating Gas," IEEE Trans. April 1971.
Electr. Insul., EI-18, p. 93, 1983.
[203] K. L. Wary and E. V. Feldman, "The Pyrolysis and
[191] A. K. Vijh, "The Nature of Metal Electrodes/SF6 Subsequent Oxidation of SF6," Proc. of Third
Reactions in SF6 Decomposition Due to Direct- Int. Conf. on Combustion, pp. 229-238, 1971.
Current Interruption under Simulated Circuit-
Breaker Conditions," IEEE Transactions on Elec- [204] T. Yoshizawa, Y. Sakai, H. Tagashira, and S.
trical Insulation, EL-il, No. 4, pp. 157-160, Sakamoto, "Boltzmann Equation Analysis of the
December 1976. Electron Swarm Development in SF6," J. Phys. D:
Appl. Phys., Vol. 12, p. 1839, 1979.
[1921 P. Vouros, "The Analysis of Arc By-Products by
SF6 by Combined Gas Chromatography or Mass Regis tered Trademark
Spectrometry," EPRI Final Report, EL80-8-LD,
December 1980. Manuscript was received 25 April 1986, in revised
form 29 May 1986.
[193] F. B. Waddington and J. Heighes, "SF6 Insulation
for Current Transformers," AEI Engineering, Vol.
7, No. 4, pp. 196-199, 1967.
[194] M. W. Watts, "Cosmic Rays and Breakdown in SF6,"
Central Electricity Research Laboratories Report
No. RD/L/N58/79, March 1979.
[195] N. Wiegart, et al., "Inhomogeneous Field Break-

down in GIS," Paper submitted for publication
in IEEE PES Winter Power Meeting, 1987.

SF6 Decomposition in Gas-Insulated Equipment

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

SF6 Decomposition in Gas-Insulated Equipment

Uploaded by

Copyright:

Available Formats

IESEE Trarnsactions fl

on ectrical Insulatiorn Vc;s EI--l No.5, cJctober l?B6

SF6 DECOMPOSITION IN GAS-INSULATED EQUIPMENT

Much of the earlier work on SF6 decomposition focused

compound S02F2 existed in far greater quantities than

., The authors reported a linear relationship between the

Ruegsegger et al. [153] employed a molecular beam set-

TABLE 3 Gaseous Decomposition Products Identified in Power Arcs at Various Laboratories

S02F2 500 W ac arc ER Garasimov [78]

SOF2, CF4, S02F2, S02 13-49 kA Al GC 650 (SOF2) Baker [11]

SOF2, SF4, CF4 30 kA Al GC 630 (Total) Tominaga [181]

TABLE 4 Solid DecompWton Products GeneTated in PoweT Arc

AlF3 22-37 kA Al XRD 20 c.c. Fukada [75]

AlF3 3000 A Al XRD 5 g/kAs Tokuyama [179]

AIF3 3.7-38 kA Al XRD 30-100 mg/kJ Kulseta [109]

CuF2, VF3, S 1 kA Cu XRD Latour-Slowikowska [115]

AIF3 1 kA Al XRD Latour-Slowikowska [115]

A1F3 (anhydrous) 16.4 kA Al Xi Baker [11]

I VERY LITTLE DATA ON PRECU RRSORS MOST DATA

a 10-3 l0, 104

TABLE 6 SF6 Decomposition Products Identified in Sparks at Various Laboratories

Spark Decomposition Spark Formation Rate

Based on the massive experimental data, Van Brunt et

TABLE 10 SF6 Decomposition Products Identified in Corona

Corona Decomposition Discharge Formation Rate

100 110 120130140150160 180 200

Fig. 7: Temperature dependenee of time to reaeh

220 240 2602

SOF2 and S02F2 normally found in other types of decom-

57 deeompoeition of SF6 [54 ]. form S02F2.

column. The readers are referred to the excellent re-

Fig. 9: The ratio of the SF3/SF6 signaZ produced

As shown in Fig. 12, the most stable species is SF4

Cu + 2F -> CuF2 (4.21) 4.4 The Formation of S2F1o

affect the system performance in two ways. The pres-

6500 ppmv HF vapor for 72 h.

Fukuda [75] reported that after a high-current fault So, 38 68

[195] N. Wiegart, et al., "Inhomogeneous Field Break-

You might also like