Академический Документы
Профессиональный Документы
Культура Документы
March 1997
Declaration
Dragan Vukovic
II
Acknowledgments
I wish to express my thanks and appreciation to my supervisor Associate Professor G. A.
Woolsey, for his encouragement and assistance at all stages of this work.
I would like to acknowledge the support of the Australian Government, through an Australian
Postgraduate Research Award scholarship.
Much of the laboratory equipment has been acquired through grants from the Australian
Research Council and the Australian Electricity Supply Industry Research Board.
Finally, I wish to express my thanks to my fellow research students and all the other members
of the Physics Department who have contributed to this work by discussion and interest.
1
Contents
Chapter 1: Introduction 1
1.1. Refractivity measurement 2
1.2. Sulphur deposition on an optical fibre 3
1.3. CO2 laser light absorption 3
Appendix 1: The estimation of the SOF4 and SO2F2 refractive indices 107
Appendix 2: Data acquisition programs for the refractive index measurement 109
A2.1. FILL_IN program 110
A2.2. PUMP_OUT program 113
Appendix 4: Evanescent field of totally reflected light from a flat interface 120
References 122
1
Chapter 1
Introduction
switchgear. As there is a relation between SF6 degradation and the damage to electrodes and
insulators it is possible to estimate the degree of the damage if the amount of the remaining
SF6 is known (Belmadani et al., 1991B), provided there are no competing sources for loss, the
most common being gas leakage. This information is useful for determining when switchgear
should be serviced. Some of the by-products, such as SF2 and SF4 , are corrosive. They react
with the different materials in the switchgear and change their mechanical and electrical
properties. Also, the newly produced gases will change the dielectric properties of the gas
mixture (Gänger and Vigreux, 1981).
It is clear therefore that a monitoring technique, which provides a measure of
the level of SF6 or dissociation by-products in a high voltage system, would allow an informed
decision to be made on when the SF6 should be replaced, in order to maintain the desired level
of insulation. As such systems involve high voltages and a high level of electromagnetic
interference, an optical monitoring technique appears to be the most appropriate choice. In
this work a number of optical options have been examined and this thesis deals with
measurements that may lead to an optical monitoring system for the degradation of SF6 .
Three methods have been investigated for determining the level of SF6
degradation: 1) measuring the refractive index of the residual gas, 2) measuring sulphur
deposition on an unclad optical fibre placed in the discharge chamber and 3) measuring the
absorption of CO2 laser light in a silver halide optical fibre placed in the discharge chamber.
investigated the refractivity of SF6 at two wavelengths; 632.99 nm, using a helium-neon laser,
and 1300 nm, using an infra-red laser diode. In addition, we have investigated the refractivity
of thionyl fluoride (SOF2) at the same wavelengths. SOF2 is produced in SF6 glow and corona
discharges (Van Brunt, 1985; Woolsey and Grey,1989). It is produced by reactions of SF2 ,
SF3 and SF4 with OH, O and H2O, and, for example, if there is glass within the system, by the
process:
SF6 has a strong absorption line around 10.6 µm, which is the lasing line of the
CO2 laser. In fact, this absorption is so strong that light from a 4 W CO2 laser is absorbed by a
length of 10 cm in SF6 at 2 Torr (267 Pa). When a CO2 laser beam is launched into a silver
halide optical fibre placed in an SF6 atmosphere, light is absorbed via the evanescent field, and
this can be used for measuring the pressure of SF6 from 9 Torr (1200 Pa) up to several
atmospheres.
Chapter 1: Introduction 4
4
Chapter 2
The d orbitals of the M shells are saturated by 10 electrons. SF6 has a tendency to capture
other electrons required to fill these orbitals and because of this attaching capability is used for
quenching arcs.
The electronic structure of sulphur in the ground state is 1s22s22p63s23p23p13p1.
Promotion of an s and a p electron to d orbitals gives the structure 1s22s2 2p63s13p13p13d13d1.
The one 3s, three 3p, and two 3d orbitals hybridise to form six hybridised sp3 d2 orbitals, Each
orbital points to the corner of an octahedron, and is available for electron-pair sharing with a
fluorine atom.
The physical constants of SF6 are:
Molecular weight, 146.06.
Dissociation energy, 3.9 eV for the S-F bond (Lifshitz et al., 1980).
Melting point, -50.8 oC (Schumb, 1947).
Sublimation point, -63.7 oC (Schumb, 1947).
Critical temperature, 45.55 oC (Miller et al., 1951).
Critical pressure, 3.73 MPa (Miller et al., 1951).
The velocity of sound at atmospheric pressure is 133 ms-1 at low
frequencies and 147 ms-1 at high frequencies (Kaye and Laby, 1973).
Dielectric strength (relative to N2 =1) at 60 Hz - 1.2 MHz; 2.3 - 2.5
(Eibeck and Mears, 1980).
Dielectric constant at 25 oC and 1 atm, 1.002049 (Eibeck and Mears, 1980).
Refractive index at 632.99 nm, 25 oC and 1 atm, 1.00070296; at 1300 nm,
25 oC and 1 atm, 1.000694 (this work, see chapter 4).
2.2. Manufacturing
2.3. Applications
Figure 2.2. A cross section of the circuit breaker filled with SF6 (type 70-SF-500 GCB. The pressure in the high
pressure section is 15 atm and in the low pressure section, 3 atm (Yamauchi et al., 1985)
Ratings: voltage 84 kV
normal current 1200 A
breaking capacity 5000 MVA
break time 5 cycles
The main practical use of SF6 is in high voltage switchgear. SF6 is used in
circuit-breakers for switching voltages between 30 kV and 800 kV, and currents up to a few
thousand amperes. SF6 has high dielectric strength, more than double that of air or CO2 , and
because of its electronegative nature, it has excellent arc-quenching properties. Although SF6
dissociates during the arc that occurs on switching, self-healing takes place following the arc,
as the dissociation products recombine. This is the ideal situation, but in reality, a small
amount of SF6 is lost during each switching process. Indeed, this loss of SF6 occurs during
any electrical discharge; an arc, a glow or a corona.
Chapter 2: Properties and applications of SF6
Circuit breakers which use oil or gas as their dielectric medium are now being
replaced by SF6 - filled circuit breakers. The latter can operate at higher voltages than those that
use oil or air, so that in order to switch a particular voltage, fewer SF6 circuit breakers are
needed in series. Furthermore, SF6 circuit breakers have fewer parts and hence lower failure
rates.
In general, two types of SF6 circuit breaker are used: the gas-blast breaker
shown in figure 2.2 and the puffer breaker shown in figure 2.4. The former uses higher
pressures and so can operate at higher voltages. However, because the SF6 puffer circuit
breaker is cheaper and less complicated, it is often cost effective to use a few such units in
series rather than a single gas-blast unit.
Figure 2.3 illustrates the operation of an SF6 puffer circuit breaker. Four stages
in the breaking process are depicted. In the closed position the electrodes are in contact to
form the conducting path. In the compression stage, movement of the puffer cylinder
compresses the gas while the electrodes remain in contact. At the moment of separation of
the
electrodes, an arc is formed but the compressed gas rapidly extinguishes the arc. In the open
position with the arc extinguished, the circuit is broken.
Underground transmission lines, now being used in Europe, Japan and the
United States for electricity transmission up to 1100 kV are filled with high pressure SF6
because of its excellent insulating characteristics.
Linear accelerators and Van de Graaff generators can be designed for very high
voltages if they are filled with SF6 .
Silicon-wafer etching
While it is very inert and easy to handle under normal conditions, SF6 is an
excellent source of fluorine atoms in an RF plasma discharge. The fluorine atoms readily react
with a silicon wafer through the process
Si + 4F → SiF4
Chapter 2: Properties and applications of SF6
where SiF4 is a gas. This process is used for etching silicon wafers as one of the steps in the
manufacture of semiconductor devices. In the manufacture of semiconductor devices this
process is one of the steps used for etching silicon wafers: the etching gas may be either ClF4
or SF6 .
Waveguide transmission
The power transmitted through a waveguide can be greatly increased if the
waveguide is filled with SF6 . At 3 GHz it is 10 times that for an air-filled waveguide. For
example, the breakdown field for 1 atm SF6 is 4 times that for dry air at 1 atm, using 0.6 µs
pulses (Meek and Craggs, 1978).
Loudspeakers
The velocity of sound in SF6 is about 3 times lower than that in air. This,
together with its low toxicity and resistance to hydrolysis, makes SF6 a good filling gas for
sealed loudspeakers. Compared to a similar sized loudspeaker filled with air, an SF6 - filled
loudspeaker has a cut-off frequency which is 1.24 times lower. This provides a significant
improvement in bass response.
Magnesium casting
Magnesium alloys used for die casting burn in air at temperatures above
600 o C. To prevent this, the whole process is carried out in a protective atmosphere such as air
or CO2 mixed with 0.22 vol % of SF6 , susequently reducing slag (Kirk-Othmer Encyc. Chem.
Tech. 1980; Fruehling and Hanawalt, 1969).
2.4. Switchgear
be able to perform these operations several thousand times without failure. This means that
contacts should not be damaged at opening or closing. These are demanding criteria,
considering that interrupted currents can be higher than 100 kA and voltages are up to
1000 kV.
During the opening of the contacts of a switchgear unit, an arc plasma is
developed between the contacts. Even in a domestic light switch an arc occurs when it is
switched off. Although it may appear that the development of a plasma should be avoided, a
switchgear unit actually relies on plasma production because it is the plasma that absorbs the
energy stored in the circuit (network), that is, in the magnetic field. The stored energy is given
by
Li 2 where L is the circuit inductance and i is the current flowing through the circuit.
2
When the contacts of the switchgear unit open, an over-voltage is produced according to the
relation
V = L di .
dt
If the circuit is opened too quickly, the over-voltage can be very high because
di → −∞. The
dt
fault currents that are interrupted by circuit breakers are of the order of 100 kA and the
over-voltage could destroy the equipment. Fortunately, the arc that develops between contacts
at opening absorbs the inductive energy.
Let us consider first the operation of a basic vacuum circuit breaker in a DC
circuit. When the contacts move apart, there is a moment when only a small area of one of the
contacts touches the other. The resistance, R, of this contact point is high and the heat
dissipation at the point is very high, given by Ri 2 . The heat produced in that spot melts a
small amount of contact metal, and the evaporated metal atoms are ionised to produce an arc
plasma. This arc continues to conduct the current and absorb the stored energy. The energy is
lost by radiation and by further heating and evaporation of contacts. As the contacts move
further apart, the heat continues to be dissipated, the electric field becomes weaker and the
ionised atoms and electrons diffuse to the walls. The magnetic field (B) of the current (J)
through the plasma creates J x B forces which cause the arc to strike the circuit-breaker wall
Chapter 2: Properties and applications of SF6
and thus lose energy. These effects lead to the extinction of the arc and opening of the circuit.
This occurs when the contacts are in a vacuum. When a gas surrounds the contacts, the arc
energy, created during the first moments of opening, is absorbed by the gas and the gas
becomes ionised. Further ionisation occurs within the plasma and conduction of the current is
maintained. Apart from losing energy through radiation, the inductive energy is used for
further ionisation of the gas. For effective DC switching the gas in the circuit breaker should
have good thermal conductivity and high ionisation energy (Rieder, 1970). The arc usually
extinguishes faster in a gas-filled circuit breaker than in a vacuum unit.
Switchgear for alternating current (AC) are based on the fact that the current
goes to zero every 10 ms (8.3 ms in USA and some parts Japan). An arc is developed in the
same way as for a DC current, but as the current approaches zero, the arc intensity decreases.
The arc extinguishes when the current goes to zero. Ideally, when the voltage increases again,
Chapter 2: Properties and applications of SF6
the contacts are far enough apart that another arc is not initiated. In practice with short contact
separation, residual ionisation causes the arc to re-ignite and it usually takes a few cycles for
an arc to be extinguished.
Important features of the gas filling are the dependence of thermal and
electrical conductivity on the temperature of the gas, which is in the plasma state. Figure 2.4
shows the variation of the thermal conductivities of SF6 and N2 with plasma temperature, at
one atmosphere.
The ideal medium for an AC circuit breaker would have low ionisation energy
and thermal conductivity in the high current stage; that is, at high temperatures (Rieder, 1970).
Thus, the plasma electrons conduct the current and the energy of the arc is absorbed weakly
by the plasma. If the ionisation energy in the high current stage is high the energy of the arc is
absorbed quickly and the arc may extinguish before the current reaches zero. This would
cause an over-voltage which would produce a new arc. When the current approaches zero, the
temperature of the plasma decreases. At this time, the thermal conductivity should be
relatively high so that the arc energy is dissipated quickly, and this is the case for SF6 at
temperatures between 2000 K and 4000 K (figure 2.4). Hence, when the current passes
through zero the arc is extinguished. When the voltage increases again the distance between
contacts is large enough to prevent re-ignition of the arc. The distance necessary to prevent
re-ignition of an arc depends on the pressure and dielectric strength of the filling gas.
SF6 has a higher dielectric strength than most other gases, so this distance is
smaller than for other gases at a particular pressure. As a result, circuit breakers filled with SF6
are mechanically simpler than others.
SF6 is now generally used for high voltage switchgear, but older systems and
some lower voltage units use other media such as air (N2 + O2), oil and vacuum. In oil the arc
generates hydrogen from hydrocarbons and hydrogen acts as the quenching medium.
All switchgear media gradually degrade and have to be replaced from time to
time. SF6 has a relatively long life time due to reassociation of SF6 molecules following arc
quenching. Nevertheless, some loss of SF6 molecules does occur, both during switching and
as a result of unwanted corona discharges, and due to leakage.
Chapter 2: Properties and applications of SF6
N, the nature of the applied voltage (AC, DC, or pulsed), electrode materials, the electrode
surface conditions and the electric field produced by the applied voltage. Depending on these
parameters, a spark, arc, glow or corona can occur.
positive ions is equal to the sum of negative ions and electrons. However, the number of ions,
both positive and negative, is much greater than the number of electrons because electrons are
attached to neutral atoms and produce negative ions. Attachment can increase the initial
breakdown voltage of the gas (George and Richards, 1969).
of the cathode which provides a strong field between the cathode dark space - negative glow
boundary, and the cathode. This accelerates the positive ions and they strike the cathode to
produce secondary electrons. The electrons accelerate away from the cathode, and while still
near the cathode surface where their energy is relatively low, they excite gas atoms and
molecules, as well as atoms of material sputtered from the cathode surface by positive ions, to
provide the cathode-glow emission (Wronski et al., 1996). In the cathode dark space their
energy is high and they make some ionising collisions with little excitation. In the negative
glow, where the electric field falls to zero, they cause ionisation and excitation. The electrons
lose much of their energy in the negative glow and in the Faraday dark space little ionisation or
excitation occurs. Electrons are accelerated by the weak field of the positive column, causing
some excitation and ionisation in the positive column. In front of the anode there is an
increase in the electric field due to the existence there of an electron space charge. This
produces the anode glow. Sometimes, regular bright and dark striations are created in the
positive column, because of the establishment of space-charge layers (Emeleus and Woolsey,
1970).
Figure 2.5. Potential distribution between the cathode and anode in a glow discharge.
Chapter 2: Properties and applications of SF6
At low pressures (below 10 Torr for electropositive gases), the cross section of
the positive column has the same area as the electrodes. Such a positive column is called
diffuse and is controlled by charged-particle diffusion to the wall. Ions and electrons diffuse
radially at the same rate and this is called ambipolar diffusion. At higher pressures, the
negative regions move towards the cathode and the column may become constricted.
(Mouwen and Claassens 1970). There are several effects that can lead to a constricted glow
discharge: volume recombination, stepwise ionisation and gas-heating (Hatori and Shioda,
1976). Constriction also depends on the current and appears when the current is high. At high
currents the constricted positive column is controlled by three-body recombination (Hatori
and Shioda, 1976). At even higher currents a transition to an arc occurs.
discharge, in terms of a radial temperature gradient leading to a radial gradient in E/N (electric
field/gas density). The electrons produced on the axis diffuse radially into the outer, lower
E/N, regions where attachment occurs, so that electrons, and hence ionisation and excitation
are concentrated on the axis.
Chapter 2: Properties and applications of SF6
For the glow discharges described in this thesis, currents were around 10 mA,
with the initial pressure of SF6 being around 1 Torr, and the positive column was constricted.
In electropositive gases, the positive ions created around the point cathode
move to the cathode due to the high electric field and collide with the cathode to produce
more electrons. The electrons travel towards the anode creating a steady glow around the
cathode as a result of excitation within the high field region.
In electronegative gases some of the electrons created around the point cathode
produce positive ions while others are captured by the gas molecules to produce negative ions.
These negative ions move away slowly from the cathode and a cloud of negative ions builds
up around the cathode causing the discharge to extinguish. The discharge starts again when
the negative ions have moved away. The negative corona in electronegative gases therefore
consists of pulses called Trichel pulses (Meek and Craggs, 1978).
The SF6 corona experiments in this thesis use positive point - plane electrode
geometry at pressures from 90 kPa to 200 kPa and currents of up to 102 µA. SF6 is an
electronegative gas and the resultant positive coronas are of the burst-pulse glow-type reported
by Van Brunt and Leep (1981) and Macgregor et al. (1986).
20
Chapter 3
The cross-section of the low pressure SF6 glow discharge chamber is shown in
figure 3.1. The chamber is 16 cm high with an internal diameter of 17 cm. The internal volume
is 3.6 litres. Two identical electrodes are made of stainless steel. Each has a diameter of
25.4 mm and a thickness of 5 mm, and has rounded edges. Their vertical position is variable
but the distance between them was maintained at 3 cm for this work. The electrodes are
supported on stainless steel rods. Both electrodes are electrically insulated from the chamber,
so that their polarity can be changed. In this work, the upper electrode is held positive. The
upper electrode is separated from the top plate of the chamber by a teflon®
(polytetrafluoroethylene or PTFE) spacer. The lower electrode is isolated from the chamber
with teflon insulators, and inside the chamber the cathode support rod is covered with an
alumina tube. The cathode is connected to the negative pole of the power supply via a
moving-coil ammeter. The chamber is grounded. High voltage feedthroughs (Vacuum
Generators) are used to support the anode and cathode.
For the evanescent-wave absorption experiment the quartz window on the side
of the glow discharge chamber is replaced by a side arm connected to the system using a
"Conflat" gasket and 7 cm non-oxygenated copper seal. Connected to the end of this arm and
normal to it, there is a 25 cm long cylinder with the silver halide optical fibre placed along its
length. This stainless steel cylinder contains a concentric teflon cylinder so that the fibre
cannot touch the metal wall. If a silver halide fibre touches metal it causes corrosion and the
fibre becomes "dissolved".
Chapter 3: Experimental discharge systems 21
The vacuum system shown in figure 3.2 is divided into low vacuum and
ultra-high vacuum systems. There are two connections to the low vacuum rotary pump. One is
used for fast pumping and the second one for extracting gas samples from the chamber into a
Chapter 3: Experimental discharge systems 22
sample cylinder. The sample cylinder has NaCl windows so that the gas content of the
cylinder can be analysed using an infrared spectrometer. With this low vacuum system it is
possible to produce a base pressure of 10-3 Torr. Then the high-volume stop valves (Vacuum
Generators CR38) in front of the turbo-molecular pump and ionisation gauge (3 and 4 in
figure 3.2) are opened and the system is evacuated via the high vacuum system. In this way
the system is evacuated during baking and purging. During a discharge the stop valve (3) in
front of the ionisation gauge is closed.
The high vacuum section is linked to the chamber by a "Conflat" flange
interface, using a standard 70 mm non-oxygenated copper seal. Solid soldering wire is used to
seal the top lid and allows a base pressure of 5 x 10-7 Torr to be achieved. A similar seal is used
for the bottom plate. The pressure in the chamber is measured with a "Baratron" pressure
gauge with resolution of 0.04 Torr and 10 Torr maximum reading. The "Baratron" pressure
gauge is used to measure the gas pressure during the glow discharge. An Edwards Pirani
vacuum gauge is attached at one of the interfaces: its maximum resolution and the maximum
reading are 10-4 Torr and 0.5 Torr, respectively. For the evanescent-wave absorption
experiment this pressure gauge is replaced by a VAISALA barometer with a resolution of
20 Pa in the range from 60 kPa to 110 kPa. A 5 cm diameter quartz window is placed on the
side of the chamber for visual monitoring of the glow discharge and optical fibre. An o-ring is
used to seal the window. There are two other apertures on the sides of the chamber at the
same level and opposite to each other. They are used for input and output of an optical fibre
for sulphur deposition experiments. Otherwise these apertures are blocked. All optical fibres
are held in place using disposable 6 mm-diameter brass cylinders filled with Torr Seal®
adhesive. The cylinders themselves are sealed to the system using o-rings.
Chapter 3: Experimental discharge systems 23
Pirani
SF6 N2 pressure gauge
mass
spectrometer
glow discharge analyser tube
chamber
6
2
5
4 6
3 7
8
2
7
9
1
1 0
0
rotary vane
pump
gas sample 3
5
cylinder
5
4 6
3 7
8
2
9
1
1 0
0
Figure 3.2. Vacuum system associated with the glow discharge chamber
discharge, valve (3) is closed and the turbo-molecular pump evacuates the mass spectrometer
only; the ionisation gauge then measures the pressure in the mass spectrometer. The
turbo-molecular vacuum pump (Balzers Pfeiffer TSU 170) operates in association with a
Pfeiffer DUO 1.5 rotary vane backing pump. Water cooling for the turbo-molecular pump is
provided by a Balzers Pfeiffer TZK 350 cooling unit. The maximum flow rate of the
turbo-molecular pump is 170 l/s for nitrogen and the maximum flow rate of the backing pump
is 1.5 m3/h. With this assembly it is possible to achieve a pressure of 10-8 Torr in the mass
spectrometer.
60 k Ω 150 W
High-voltage probe
+
1:1000
digital High
v o l tm e t e r
Voltage
DC
Power
Supply
4 5 6
3 7
8
4 5 6 2
3 7 1 9
2 8 0
1 0
1 9
0 10
Ammeter
The SF6 and N2 inlets are through needle valves and the common stop valve (6).
Chapter 3: Experimental discharge systems 25
The discharge circuitry used for glow discharges is shown in figure 3.3. The
high voltage floating DC power supply source is operated at around 1500 V to provide a
10 mA current in the SF6 discharge. In order to maintain a constant current of 10 mA, the
discharge voltage has to be adjusted slightly from time to time. A 60 kΩ resistor is placed in
series between the anode and the power supply. It is a ballast resistor that provides a net
positive incremental load to the power supply to control the discharge current, and thus
prevents the current "running away" and an arc forming.
to the chamber. The input voltage to the plane electrode is via a BNC connector mounted on
the aluminium blank. The centre lead of the BNC connector is connected to the electrode.
For this work, the bottom electrode is used as cathode and the top electrode is
used as anode. Its support rod is of stainless steel. The insulator (2) for the anode support rod
is 300 mm long. The bottom section of this insulator is exposed to the discharge environment
and is made of teflon, while the top section is of nylon. The discharge electrode itself is a
stainless steel tip screwed to the end of the support rod. The tip has an angle of 6o. Vertical
movement of the support rod allows the anode position to be varied. On top of the rod,
outside the chamber, a large smooth aluminium dome prevents external coronas and sparks.
The dome is not shown in figure 3.4.
9 MΩ
4 5 6
3 7 4 5 6
8 3 7
2 2 8
0
1 9 1 9
0 1 0 0 1 0
Constant current
high voltage DC
power supply
4 5 6
3 7
2 8
1 9
0 10
Ammeter
The schematic diagram of the circuit for the corona discharge is shown in figure
3.5. The 9 MΩ ballast resistor is included in order to prevent arcing. A ballast resistor provides
Chapter 3: Experimental discharge systems 28
a net positive incremental load to the power supply to control the discharge current, and thus
prevents the current "running away" and an arc forming.
Routine monitoring of SF6 and its discharge by-products is carried out using a
mass spectrometer which is a part of the high vacuum system connected to the glow discharge
chamber. The mass spectrometer analyser tube shown in figure 3.2 is an A.E.I. MS10
(Associated Electric Industries) single focusing, magnetic-sector spectrometer of medium
resolution, having a 180o ion-path deflection of 2" (50.8 mm) radius. The top and bottom
flange plates are removable. The bottom plate and gas inlet port are sealed with preformed
gold wire gaskets. Due to the need for regular maintenance, tin wire is used for the top flange
sealing as it is less expensive than gold wire. Heating bands are fitted around the flanges to
allow the analyser tube to be heated to a maximum of 400 oC. As tin is used for the top flange,
the voltage for the heater is adjusted so that the temperature of the top flange does not exceed
200 oC. The spectrometer tube is 203 mm high, 149 mm long and 50.8 mm wide, with an
internal volume of 0.5 litres. The electron source is a filament wire made of rhenium, with
0.178 mm diameter and 0.5 Ω resistance.
For the operation of the mass spectrometer, a magnet is placed around the
analyser tube. Two magnets are available. One of them is a standard "horse-shoe" permanent
magnet with a magnetic field induction B of 0.184 Tesla. It is possible to measure ions of 2 to
200 amu using this magnet. Both negative and positive ions can be detected by reversing the
magnet. The other magnet is an A.E.I. MS 10 electromagnet controlled by an A.E.I. MS 10
control unit. It is used for positive ions and the mass detection range with this magnet can be
extended beyond 400 amu. For the measurements reported in this thesis the magnetic field is
set to 0.184 Tesla.
The mass spectrometer analysing tube is controlled using a VSS (Vacuum
Scientific Services) RGA 10 control unit. For the work of this thesis it is set to an automatic
Chapter 3: Experimental discharge systems 29
scan from 0 to 200 amu; the scan speed to 0.3 amu/s; the electron acceleration voltage to
+70 V; the ion repeller voltage to -1 V; and the trap current to 50 µA. The pressure in the
analyser tube is between 2 x 10-7 mbar and 10-6 mbar while the full-scale deflection of ion
current is between 1 and 30 pA. The pressure in the mass spectrometer and the ion-current
range were chosen to provide maximum sensitivity.
30
Chapter 4
Refractivity measurement
The most abundant gaseous products of an SF6 discharge are SOF2 and SO2
(Mortensen et al., 1994; Belmadani et al., 1991A). Their sum (SOF2 + SO2) increases during a
discharge and their ratio depends on the initial water content within the discharge chamber or
switchgear device (Belmadani et al., 1991A). SO2 is produced by the reaction
Both gases (SOF2 and SO2) have an index of refraction lower then SF6 : 1.00062166 for SOF2
(this work, 101325 Pa, T=25 o C, 632.99 nm) and 1.000686 (Gray, 1972) for SO2 , (101325 Pa,
0 oC, 589.6 nm) while SF6 has a value of 1.00070296 (this work, 101325 Pa, T=25 oC,
632.99 nm). This means that the refractive index of the gas mixture in the discharge chamber,
measured at 101325 Pa and 25 oC, decreases with discharge time. As SF6 molecules are
dissociated during a discharge and other molecules produced, the pressure in the chamber
rises. The refractive index of the resultant gas mixture at that high pressure is higher, but the
index of refraction of the same gas mixture measured at normal conditions (101325 Pa and
25 oC) decreases, because the concentration of SF6 decreases. All other gases produced in the
discharge chamber have lower indices of refraction then SF6 except SOF4 and SO2F2 . There
are no data available for these gases but it can be estimated (see Appendix 1) that their
refractive indices are 1.000818 and 1.000877, respectively. However, the concentrations of
these gases produced in the discharge are much smaller than those of SO2 and SOF2 . Because
SOF2 is an important product of the SF6 discharge, and therefore is a significant contributor to
the mean refractive index of the SF6 discharge gas mixture, and because no data on its
Chapter 4: Refractivity measurement 31
refractive index are available in the literature, a measurement was made of the refractive index
of SOF2 .
The relation between refractive index n and gas density ρ is given by the
Lorentz - Lorenz formula (St-Arnaud and Bose, 1979)
n 2 − 1 = A ρ + B (T) ρ 2 (4.2)
R R
n2 + 2
where AR and BR(T) are the first and second refractive virial coefficients, respectively.
p
= ρ + B p (T) ρ 2 + ... (4.3)
RT
where Bp(T) is the second virial coefficient of state. The refractivity (n - 1) can be written in
the form:
p p 2
n−1= A + B(T) (4.4)
RT RT
In order to find the relations between the coefficients A and B(T) and the coefficients AR and
A R ρ + B R (T) ρ 2 = n 2 − 1 ≈ 2 (n − 1) − 1 (n − 1) 2 − 4 (n − 1) 3 + ...
2
(4.6)
n +2 3 9 27
or
Equation 4.9 reveals that the coefficients A and B(T) are related to the coefficients A R and
BR by
AR = 2 A (4.10)
3
and
In equations 4.2, 4.3 and 4.4, the higher order terms are neglected because their influence only
becomes important for SF6 and SOF2 pressures above 10 atm; and in this work pressures were
always below 1.1 atm.
Chapter 4: Refractivity measurement 33
The gas under study was maintained in an invar tube placed in one arm of the
interferometer. The 7 mm diameter tube was thermally insulated by surrounding it with a
Chapter 4: Refractivity measurement 34
thick layer of insulwool and by using double quartz windows on each end. Six chromel/alumel
thermocouples were positioned along the length of the invar tube to monitor spatial and
temporal variations in temperature. Variations in temperature within the tube were found to
be less than 0.1 K during all series of measurements. The internal length of the tube was
486.42 mm. The gas pressure in the tube was measured using a VAISALA PTB200 A digital
barometer, connected to a computer through a serial port. The VAISALA PTB200 A digital
barometer uses a BAROCAP® silicon capacitive absolute pressure sensor. The barometer has
3 outputs: RS 232C full duplex serial interface, TTL level bidirectional serial output and a
pulse output. The RS 232C serial output gives the best precision. The total accuracy is 20 Pa in
the range 60000 Pa to 110000 Pa. The purities of the gases used were SF6 ; 99.9%, and SOF2 ;
98.5%. Light sources used were a He-Ne laser for 632.99 nm (the vacuum wavelength) and a
laser diode for 1300 nm. A UDT 10DP silicon p-i-n photodiode was used for measurements at
632.99 nm and an EG&G C30642E InGaAs photodiode for those at 1300 nm. Photodiode
signals were amplified and stored on computer after A/D conversion. The A/D converter is
part of the acquisition card CONTEC ADC-10. The A/D converter has a 12 - bit resolution and
maximum speed of 30000 samples per second. The whole interferometer arrangement was
shielded to minimise convective air flow and to improve thermal stability. With this
arrangement, it was possible to measure changes in optical path equivalent to 0.01 fringe at
632.99 nm and 0.05 fringe at 1300 nm.
needed. For measurements made as the gas was removed from the tube, initially a nitrogen
trap was used. However, the vacuum produced in the system between the needle valve NV1
(in figure 4.2) and the rotary vane pump by the same pump, was enough to lower the gas
pressure in the tube below 60 kPa even when the rotary vane pump had been switched off.
The procedure for the pumping out measurement was the following: the needle valve NV1
was closed and the rotary vane pump allowed to evacuate the volume between the pump and
the needle valve, the pump was then switched off and the valve NV1 opened. This allowed the
gas from the invar tube to slowly bleed into the evacuated part of the system. In this way the
pressure in the invar tube varied from 120 kPa to 55 kPa.
Pirani NV1
pressure
VAISALA head
PTB200A
barometer
nitrogen trap
Pirani
pressure
head
diffusion
Figure 4.2. The vacuum section of the experimental arrangement for measuring refractive index.
An appropriate computer program and driver for the A/D converter were
written for acquisition during the filling of the tube and a second program written for
Chapter 4: Refractivity measurement 36
evacuation. The programs are described in appendix 2. Fringe shifts were converted to
refractivities using the equation
mλ 0
n−1= (4.12)
L
where m is the experimental fringe shift, λ 0 is the source wavelength in vacuum and L is the
length of gas intercepted by the interferometer beam. The refractivity as a function of pressure
and temperature can be fed into equation 4.4 to provide values for the coefficients A and
B(T). Then a curve of the form of equation 4.4 was fitted to a plot of refractivity as a function
of pressure to provide values for the coefficients A and B(T) for comparison with published
data. AR can be determined from equation 4.10, and BR(T) from equation 4.11. In order to
calculate BR(T) , a value is required for Bp(T) at 25 oC, and this was determined by linear
interpolation of the experimental data of Hamann et al. (1953) obtained at 20 oC and 30 oC.
Refractivities and refractive indices for different temperatures and pressures
were calculated using equation 4.4. Although measurements were made at temperatures
between 20 oC and 25 oC, all refractivity values were converted to 25 oC and one atmosphere,
for comparison with published data. Equation 4.2 was used over this temperature range with
B(T) assumed constant. B(T) does depend on temperature according to the relation (Thomas
and Tayag, 1988)
where T* is a constant for a specific gas, and for SF 6 , is 4.20. However, the total variation in
B(T) between 20 oC to 25 o C, is only 3%, and because the contribution of the second term on
the right-hand side of equation 4.4 is 100 times less than the contribution of the first term,
B(T) can be assumed to be constant over this temperature range. Indeed, the overall
Chapter 4: Refractivity measurement 37
uncertainty in the measured refractivities resulting from this assumption is 2.5 x 10-7, which is
less than the precision of the experiment.
632.99n refractivity x106 refractivity x106 the first refractive coefficient A coefficient B
@101325 Pa at 25 oC virial coefficient AR (10-5m3/mol) (10-10 m6/mol2)
m @101325 Pa (10-6m3/mol)
SF6 708.05+0.5 702.96+0.5 11.336+0.008 1.7005+0.0012 47.3 @ 22.9 oC
@ 22.9 o C
SOF2 626.2+0.5 621.66+0.5 10.002+0.007 1.5004+0.0011 50.2 @ 22.8 oC
@ 22.8 oC
TABLE 4.1. Refractivity and refractivity virial coefficient at 632.99 nm for SF 6 and SOF2
1300nm refractivity x10 6 refractivity x106 the first refractive coefficient A coefficient B
@101325 Pa at 25 oC virial coefficient (10-5m3/mol) (10-10m 6/mol2)
@101325 Pa AR (10-6m3/mol)
SF6 698.7+1 694.0+1 11.18+0.016 1.676+0.0024 52 @ 21.8 oC
@ 21.8 oC
SOF2 618.9+1 614.7+1 9.934+0.015 1.490+0.023 34 @ 22.8 oC
@ 22.8 oC
TABLE 4.2. Refractivity and refractivity virial coefficient at 1300 nm for SF 6 and SOF2
coefficient B(T), because, even with the degree of precision achieved, the uncertainties in the
values of B(T) are estimated to be as high as 100%. This is essentially because B(T) is around
100 times smaller than A. However, since the nonlinear term in equation 4.4 is not significant
at pressures around 1 atm and room temperature, the value of refractivity depends little on
B(T). Values for BR(T) are not presented in Tables 4.1 and 4.2, because, apart from
depending on B(T), BR(T) depends on Bp, values for which vary by more than 10% among
different investigators (Hamann et al., 1953; Clegg et al., 1955; MacCormack and Schneider,
1951).
refractivity x 106 coefficient A the first refractive the second refractive coefficient B
at 25 oC (10-5m3/mol) virial coefficient virial coefficient (10-10m6/mol2)
@101325 Pa AR (10-6m3/mol) -12 6
BR (10 m /mol ) 2
704
702
700
Refractivity ( x10-6)
698
696
694
692
690
688
686
0 5 10 15 20 25 30
Time (days)
Under these ideal conditions, a gas mixture may be considered as being made
of layers of constituent gases, as illustrated in figure 4.4. for a three - gas mixture. In figure
4.4.(a), a mixture of 3 gases is assumed to exist in a dielectric cylinder contained between a
pair of metal electrodes. If the cross-sectional area of the cylinder is S and the length is L, the
electrical capacitance between electrodes is
ε0 ε m S
Cm = (4.14)
L
where ε0 is the dielectric constant of vacuum and εm is the relative dielectric constant of the
gas mixture.
mixture
a)
b) 1 2 3
L1 L2 L3
Figure 4.4. For linear gases the capacitance of a cylinder filled with mixture of 3 gases is the same as the
capacitance of 3 capacitors in series where the gases from the mixture are separated in 3 capacitors.
1 = 1 + 1 + 1 (4.15)
C m C 1 C2 C 3
Chapter 4: Refractivity measurement 42
where C1 , C2 and C3 represent the capacitances of the separate layers of gas, and
ε 0ε1S
C1 = (4.16)
L1
where ε1 is the relative dielectric constant of gas 1, and there are similar relations for C2 and
L = L1 + L2 + L3 (4.17)
gives
1 = K1 + K2 + K3 (4.18)
ε m ε 1 ε 2 ε3
V p Lp S
Kp = = (4.19)
Vm L S
where Vp and L p are the volume and the length of gas p, and Vm is the overall volume. For
gases
ε = n2 (4.20)
1 = K1 + K2 + K3 (4.21)
n 2m n 21 n 22 n 23
If npq is the refractive index of gas p at wavelength q, a system of linear equations can be
written such as
Chapter 4: Refractivity measurement 43
1 = K1 + K2 + K3 (4.22)
n 2m1 n 211 n 221 n 231
1 = K1 + K2 + K3 (4.23)
n 2m2 n 212 n 222 n 232
K1 + K2 + K3 = 1 (4.24)
n11 , n21 , n31 and nm1 are the refractive indices of gases 1, 2 and 3 and the gas mixture in the
cylinder, respectively, at the first wavelength.
n12 , n22 , n32 and nm2 are the refractive indices of gases 1, 2 and 3 and the gas mixture in the
cylinder, respectively, at the second wavelength.
Hence the values of the gas concentrations K can be determined from a series
of measurements of the refractive index of the gas mixture at different wavelengths,
nm1 , nm2 , ......, if the refractive indices of the individual gases are known at the same
wavelengths. Measurements must be made at (g-1) wavelengths, where g is the number of
gases in the mixture. Because the differences between refractivities are small, the refractive
indices are very close to 1, and the changes in refractive index with wavelength for one gas are
small, measurements must be made with a precision of at least 1x10-7 with, for example, a
refractometer of the type described by Kerl and Jescheck (1982). With this interferometer it is
possible to obtain the refractive indices of a gas in the whole visible spectrum during one
measurement. Also, calculations must be made with high precision. On some computers,
where floating-point calculations are carried out with 12 significant digits, which is equivalent
to 40 bits, an error can be accumulated during the calculation. For example, there would be a
difference between
n 11n 21 + n 11 n 31 + n 21n 31
n 11 n 21 n 31 (4.25)
and
1 1 1
n 11 + n 21 + n 31 (4.26)
Chapter 4: Refractivity measurement 44
The second calculation is less accurate than the first one because more divisions are involved.
Larger errors occur in calculations involving division than those involving multiplication,
subtraction and addition.
The measurement procedure described here for a mixture of three gases can be
extended to any number of gases in the mixture.
As discussed earlier, a monitoring procedure for concentrations of individual
species based on the measurement of refractive index at different wavelengths requires
refractive indices to be measured with a precision of 1 x 10-7. The precision available with the
measuring system used here is restricted to 5 x 10-7 at 632.99 nm. It was decided therefore not
to extend the refractivity analysis in this manner as part of the present program, although it is
an approach that could be considered in future, if a more precise system becomes available for
the measurement of refractivity.
used as an optical isolator to prevent feed-back into a laser, but with the optical components
reversed.
1 0
exp −i π (4.27)
4 0 i
and for a linear polariser, with transmission axis at 45o with respect to the y-axis, the Jones
matrix is
1 1 1
(4.28)
2 1 1
When these two plates are in series with the light first passing through the
quarter-wave plate, the resultant matrix is
1 exp −i π 1 i
2 4 1 i (4.29)
The Jones vector for a linear polarised beam at an arbitrary angle α with respect
to the y-axis (see figure 4.5) is given by
sin α
E (4.30)
cos α
In order to find the output light, the Jones matrix of the plates (equation 4.29)
has to be multiplied by the input beam matrix (equation 4.30). The result is:
1 1
E 1 = 1 E (sin α + i cos α) exp −i π = 1 E exp −i π − α (4.31)
2 4 1 2 4 1
α+π/2 and the equation for the orthogonally polarised beam with respect to the beam
described by equation 4.31 is obtained by substituting α with α+π/2. For an equal amplitude
beam it gives
1 1
E 2 = 1 E (cos α − i sin α) exp −i π = 1 E exp −i π + α (4.32)
2 4 1 2 4 1
input polarisation
is
beam
α
input polarisation
plane of the
horizontal beam
quarter-wave
plate
fast-axis direction
of the quarter-wave plate
Figure 4.5. Input orthogonally polarised beams and their positions with
respect to the fast axes of the quarter-wave plate.
Chapter 4: Refractivity measurement 47
The beams represented by E1 and E2 have the same amplitude and are linearly
polarised at 45o to the x-axis. There is a phase difference of π/2 between them. The two beams
will interfere to provide an interference pattern. It is clear that maximum contrast occurs when
the two beams have equal amplitudes, and this is achieved by rotating the laser source. When
the quarter-wave plate is omitted, maximum contrast requires appropriate rotation of the laser
and linear polariser.
4.5. Summary
Chapter 5
The developments in optical fibre sensing during the past decade show that
optical fibre sensors have great potential. They have better sensitivity than many existing
sensing techniques and they can be made very compact and insensitive to their environment.
Their dielectric characteristic makes them ideal for use in situations involving high voltage,
electromagnetic and electric noise, and corrosive materials. Another important advantage is in
multiplexing. It is relatively easy to combine, transmit and extract information from several
measuring points through the same fibre. They can be used to replace almost all classical
sensors. Up to the present, commercial exploitation has been limited, but a few have achieved
commercial success, such as the Luxtron 1000 temperature sensor based on temperature
-dependent phosphorescence (Grattan, 1990).
Optical fibre sensors can be divided into two types: intrinsic and extrinsic. In an
intrinsic sensor, light does not leave the fibre and the sensing signal is the result of modulation
of the light while it travels through the fibre due to changes produced in the fibre core or
cladding (or surrounding medium if there is no cladding) by changes in the quantity being
measured. An example of an intrinsic sensor is an evanescent-field optical fibre sensor where
the medium that surrounds the core (figure 5.2) causes attenuation of the light that propagates
through the fibre. Fibres in extrinsic sensors are used to transmit and receive light that is
modulated by the measurand when it is external to the fibres. In some cases, the light can be
produced in some other way and the fibre is used simply to collect the light and transmit it to
the detector. An example of an extrinsic sensor is shown in figure 5.1. The chemical
Chapter 5: Review of chemical optical fibre sensing 49
compound in the reflector changes its optical properties when it reacts with the measurand.
Light that is collected in the detecting fibre depends on these changes. Thus, the detected light
is modulated by the chemicals that react with the reflector.
According to which component of the light-wave function (equation 5.1) is
modulated by the quantity being measured, optical fibre sensors can be divided into
interferometric (phase), polarisation - modulated, frequency - modulated and amplitude -
modulated sensors. The electric vector of an electromagnetic wave can be written as:
E = E0 sin (ω t − k x + ϕ) (5.1)
The optical fibre sensing techniques used in this study are chemical. Many of
the established chemical sensing and analysing techniques that use bulk optics can be
employed with optical fibres; these include fluorescence, phosphorescence,
chemiluminescence, evanescent-wave absorption, internal reflection spectroscopy, ionic
indication, absorption, opto-mechanical effects and FTIR (Fourier Transform Infrared
Spectroscopy). The process used here for sulphur measurement relies on evanescent-wave
absorption, but it is useful to review some of the other optical methods that have been applied
to chemical sensing using optical fibres.
5.2.1. Luminescence
When energy is absorbed by a substance it can be re-emitted, usually with less
energy than the input energy. The input energy excites atoms or molecules which may radiate
when they return to a lower energy state. Otherwise they can lose the energy in thermal
collisions. Excitation may occur through photoluminescence (excitation by light),
thermoluminescence (excitation by heat), chemiluminescence (excitation by chemical
reaction) or electroluminescence (excitation by electric field or current).
Energy re-emission may take place through fluorescence or phosphorescence,
which are essentially special forms of light scattering: light comes from one direction and is
re-emitted in random directions with a longer wavelength. The difference between these two
processes is in the way in which the atoms or molecules return to the lower energy state. The
time between absorption and emission for fluorescence is less than 10-8 s (Dakin and Culshaw,
1988), while for phosphorescence the time is substantially longer.
Chapter 5: Review of chemical optical fibre sensing 52
In fluorescence, an excited atom, molecule or ion emits the light shortly after
absorption. When a molecule is in an excited state it may exist in one of many very close
vibration states within the first excited state. It can lose a small amount of energy in thermal
collisions by moving from one vibration state to another before it returns to the ground state.
Because of that, the peak of the emitted spectrum is shifted towards a longer wavelength
relative to the absorption wavelength.
Phosphorescence represents a similar effect. A condition for the existence of
phosphorescence is that there is a triplet state, close below the first excited singlet state, into
which the molecule can fall. Then a transition from this triplet state to the lower energy,
usually ground state, occurs. It involves a forbidden triplet → singlet jump, and because this
transition is not absolutely forbidden, the transition does finally occur but the process is
relatively slow. Because of the low probability of the transition, the molecule remains for a
relatively long time in this so-called metastable state before radiative emission occurs. So, for
phosphorescence, the time between absorption and emission is much longer (milliseconds)
than for fluorescence.
These mechanisms can be employed for optical fibre sensing in many ways.
The Luxtron 1000 temperature sensor, for example, is based on the temperature-dependent
phosphorescence of rare-earth phosphors (Dakin and Culshaw, 1988). The emission intensity
varies with time according to
I = I 0exp − τt (5.2)
light, t is time and τ is a temperature-dependent constant. For this sensor a piece of rare earth
phosphor is placed on the tip of an optical fibre. The same fibre delivers UV light to the piece
of phosphor and collects the phosphorescent light. The Luxtron 1000 measures the intensities
of the UV light and phosphorescent light. The temperature is determined from the ratio of
these two values. The later version Luxtron 750 does not measure the intensity of the UV light
source. For this measurement, the intensity of the phosphorescent light as a function of time is
Chapter 5: Review of chemical optical fibre sensing 53
recorded and the decay constant (τ) is calculated. The temperature is determined from this
temperature-dependent constant. The light source is pulsed. The phosphorescent material used
in the Luxtron sensor is magnesium fluorogermanate (Grattan, 1990).
During fibre manufacture, fluorescent material can be added. Such a fibre can
be used, for example, to detect a corona discharge inside a high-voltage device filled with SF6
(Farenc et al., 1994). Light produced by a corona discharge in SF6 has peak emission between
460 nm and 500 nm (Casanovas et al., 1991). A phosphorescent material with absorption
within this range is mixed with polystyrene to produce a plastic optical fibre. When the optical
fibre is placed in a high voltage device filled with SF6 , light produced by a corona is absorbed
by the optical fibre. Phosphorescent light is produced within the optical fibre and this light is
radiated in all directions. Some of the light is trapped in the optical fibre and transmitted along
the fibre out of the device to a detector. The output signal depends on many parameters: the
nature of the phosphorescent material, its concentration, the length of the optical fibre, the
quality of the fibre end, the sensitivity of detector, and the position of the fibre in relation to
the corona source. It is reported that with carefully chosen parameters it is possible to produce
useful signal after the photodetector of up to 50 mV from a 10 cm length of fibre exposed to
light from an AC corona discharge, 2.5 cm away from the corona source, which is a
point/plane electrode system in SF6 atmosphere (Farenc et al., 1994).
An optical fibre can be coated with a fluorescent or phosphorescent substance
using a sol-gel process. A sol-gel derived coating is porous and traps molecules, which
increases the concentration of measurand species (gas or liquid) in the coating. This makes a
sensor more sensitive. A compound that reacts with the measurand is mixed with the coating
during the sol-gel process. Light is absorbed in the coating via the evanescent field and a
fraction of the resultant fluorescent or phosphorescent light is coupled into the fibre. The
measurand species that surrounds the fibre is trapped in the sol-gel coating, and reacts with the
substance embedded in the coating. This reaction produces the optical effect that can be
monitored to provide quantitative information of the concentration of measurand species. For
example, oxygen quenches the fluorescence of a ruthenium complex and it is used for oxygen
sensing. The ruthenium complex fluoresces yellow when it is irradiated with blue light. The
Chapter 5: Review of chemical optical fibre sensing 54
ruthenium complex can be a part of sol-gel coating. In this case a part of the emitted light is
coupled into the measuring fibre and transmitted along the fibre. It can be measured if the
strong incident blue light is blocked with a cutoff filter at 505 nm. Oxygen quenches this
fluorescence and the light intensity at the output of the fibre decreases as the concentration of
oxygen increases. The light source can be a laser (MacCraith et al., 1993) or an LED
(MacCraith et al., 1994).
When light is absorbed in a semiconductor, electrons from the valence band
move to the conducting band, so that electron-hole pairs are generated. These electrons can
lose their energy by thermal collisions or by a radiative transition (photoluminescence). When
an n-type semiconductor is coated with a metal, a Schottky barrier junction is formed. A
Schottky barrier junction is similar to a p-n junction except that it is a metal - n-type
semiconductor junction where the metal acts as an anode and the semiconductor as a cathode
(Sah, 1991). An important feature of this junction is that the depletion region is much thinner
than in a p-n junction. For optical purposes, it is made by coating a semiconductor with a thin
metal (usually gold) film. The depletion region is just under the metal surface, so that light can
penetrate through the metal layer, reach the semiconductor underneath and generate
electron-hole pairs. The carriers produced in the depletion region do not emit light because the
generated electrons and holes are driven away by the internal electric field of the junction, and
the probability for their recombination is greatly reduced. Electrons are pushed into the
semiconductor and holes into the metal. Only electrons and holes generated in the
semiconductor, rather than in the depletion region, can cause photoluminescence. When the
width of the depletion region is large, fewer electrons are generated outside this region and
photoluminescence is weaker.
A sensor made from a Schottky barrier photodiode can be used for measuring
the concentration of fluorine. A Schottky barrier photodiode made up of several monolayers
(1 x 10-9 mol cm-2) of redox-active sites on n-GaAs (Van Ryswyk and Ellis, 1986) and
illuminated with argon laser light at 457.9 nm radiates light with a peak at 865 nm. When the
sensor is exposed to volatile oxidants such as the halogens (F2 , Cl2 , Br2 or I2) the depletion
width increases and hence the intensity of the 865 nm peak decreases. The photoluminescence
Chapter 5: Review of chemical optical fibre sensing 55
regains its original intensity when the sensor is flushed with N2H4 . In order to make an optical
fibre sensor for the study of the fluorine distribution in an SF6 discharge, a Schottky barrier
photodiode of this type could be bonded to an optical fibre with adhesive. The fibre would
deliver laser light and receive photoluminescent light. Another possibility is to grow gallium
arsenide on a sapphire optical fibre.
reflector
to detector
Figure 5.1. An optical fibre absorption sensor. Light is transmitted from a source to the reflector where it is
absorbed. The reflected light is coupled and transmitted by a second fibre to a detector. The source can be
monochromatic or white light.
Chapter 5: Review of chemical optical fibre sensing 56
5.2.3. Biosensors
In general, biosensors can be classified into those groups previously described
for chemical sensors. However, they do have some special characteristics. Immunochemicals
(antibodies) and enzymes, used in optical fibre biosensors, are very selective and this, together
with high optical sensitivity and small probe dimensions, makes optical fibre biosensors the
most selective of all chemical sensors. Enzymes catalyse chemical reactions that involve
measurands. Products of the reaction affect the signal in the optical fibre. The change in the
output signal can be due, for example, to a change in refractivity, absorption and
luminescence. Antibodies are more selective than enzymes. They attract corresponding
antigens that can be detected in many different ways. Immunochemicals and enzymes can be
covalently attached to the core of the fibre after removing the cladding (Anderson et al., 1994),
or the core can be coated with enzymes or antibodies using a sol-gel process, or they can be
placed inside the reflector of figure 5.1. The problem with antibody fibre optic sensors is that
they become irreversibly saturated and so does the measured signal (Janata, 1989). The
solution is to measure the rate of binding antigens with antibodies; that is, the differential of
the optical signal. Another problem with biosensors is their short-term stability, as they can be
used only for a few days. For example, the sensor made by Kunz et al. (1996) for detection of
fatty acids can be used only for four days.
and stretching. The core and cladding become thinner and the total diameter can be reduced to
less than 1 µm. As the diameter of the core decreases in the tapered region, the waveguide
parameter of the fibre, or the V -number, defined as V = 2π ρ n 2co − n 2cl (see Appendix 3)
λ
decreases, and the ratio of light power in the core to that in the cladding decreases. Depending
on the shape of the tapered region one or more modes can be coupled into the cladding
(Payne and Mackenzie, 1991). The intensity of the electric field of the light is depicted in figure
5.2 as a red net. If the evanescent field interacts with the molecules of the surrounding
medium, the intensity of the light at the output of the fibre is reduced.
Figure 5.2. Tapered single mode optical fibre: blue is the cladding; green is the core; the
amplitude of the red net represents the strength of the electric (and magnetic) field.
Light travelling through a tapered optical fibre, such as that depicted in figure
5.2 can be very sensitive to the surrounding medium. However, tapered fibres are very fragile
and difficult to handle because the diameter of the core in the tapered region is typically
around 2 µm (Henry, 1994A; Hale and Payne, 1994). The tapered region is usually a few
Chapter 5: Review of chemical optical fibre sensing 58
millimetres long and it is difficult to make it longer. Increasing the length of the tapered region
improves the sensitivity, but this is difficult because of the fragile nature of such a fibre.
The sensitivity of a tapered optical fibre sensor based on evanescent - field
interaction can be further increased by using a sol-gel-derived coating as described in section
5.2.1.
An alternative to the tapered optical fibre is the D-fibre (Stewart and Culshaw,
1994; Henry, 1994A) which is depicted in figure 5.3. The cladding is removed from one half of
the fibre along the axis so that a section of the core is exposed to the surrounding medium.
This configuration is not as sensitive (per unit length) as the tapered fibre but the mechanical
strength is similar to that of a standard optical fibre. The sensitivity can be increased by
extending the length of the fibre. A methane sensor using D-fibre has been developed using
the methane absorption line at 1.66 µm (Muhammad and Stewart, 1992).
The sensitivity of sensors made with multimode optical fibres can be increased
by launching the light with a high incident angle with respect to the fibre axis, as explained in
section 5.2.3.2. For example, a tunable lead salt laser diode at around 10.6 µm was used as a
light source to couple the light into a silver halide optical fibre in order to sense SF6 (Messica
Chapter 5: Review of chemical optical fibre sensing 59
et al., 1994). When the light is launched at a high angle, a significant amount of light is
reflected from the fibre end where the light is coupled in, so that the actual launching angle is a
compromise. Messica et al. (1994) have found that, for the optimal result, the light should be
launched at an angle of 20o. The output absorbance as a function of the pressure of SF6 is
shown in figure 5.4. The absorbance is the negative natural logarithm of normalised
transmitted power.
Figure 5.4. Evanescent wave absorption by SF6 . The absorbance is the negative natural logarithm of normalised
transmitted power. It is shown as a function of the pressure of SF6. A silver halide optical fibre without cladding
was used and its diameter was 900 µm. The lead salt laser diode was tuned to 10.562 µm, which is the absorption
peak of SF 6 (from Messica et al., 1994).
The typical output power of a lead salt laser diode is around 1 mW. These
diodes operate at cryogenic temperatures and are produced for wavelengths up to 30 µm.
where
planar waveguide o
θ1 70
n co= 2.1
o
45
air, ncl=1
Figure 5.5. Light beam coupling (from CO2 laser) into a 1 mm thick planar dielectric (silver halide) waveguide
in air. The wavelength is 10.6 µm. The refractive indices of the silver halide and air at 10.6 µm are 2.1 and 1
respectively. The electrical vector of the light is perpendicular to the paper. (A meridional ray in the 1 mm
diameter fibre propagates in the same way as for a planar waveguide.)
For the situation depicted in figure 5.5. the light from a CO2 laser (10.6 µm) is
launched at an angle of 45o to the waveguide axis. The ray is refracted at an angle of 20o and
then follows the zigzag path with the incident angle at the boundary of 70o. The thickness of
the waveguide is 1 mm. In this example, the electric vector is perpendicular to the paper, so
the light propagates as the TE mode (q=1), and from equation 5.3, the power that propagates
as the evanescent wave is 1.06% of the total power. If the light was launched as a TM mode, q
Chapter 5: Review of chemical optical fibre sensing 62
would be 3.84 (from equation 5.5) and the power propagating as the evanescent wave would
be 0.17% of the total power. It is seen that the power in the evanescent field for the slab
waveguide (and thick optical fibres), and hence the degree of interaction with the surrounding
medium, is maximum when light is launched as a TE mode with the incident angle θ1 as close
to the critical angle as possible.
This calculation was carried out for an optical fibre (or a planar waveguide)
with a surrounding lossless gas instead of cladding. When it is used as a sensor, the
surrounding gas is not lossless, but the conclusion is the same: the degree of interaction with
the surrounding medium is maximum when the light is launched as a narrow s-polarised (TE
mode) beam with the incident angle θ1 as close to the critical angle as possible.
where L is the length of the sensing part of the fibre and γ is the effective fibre attenuation
induced by the absorbing cladding or surrounding medium, which here is gaseous SF6.
with a microscope objective. To distinguish these two cases, γN and γW will be used for the
fibre attenuations for narrow-beam (N) and wide-beam (W) sources, respectively.
Chapter 5: Review of chemical optical fibre sensing 63
diameter greater than 0.5 mm. This means that the V-number is much larger than 1 and the ray
approach can be used to described the propagation of light within the fibre.
When this fibre is unclad and placed in a slightly absorbing medium (ni << ncl)
and when a narrow laser beam is launched into the fibre, the attenuation of the light is given
by (Schnitzer et al., 1990)
where
Chapter 5: Review of chemical optical fibre sensing 64
αs + αp
α≡ , where αs and αp are the effective absorption coefficients for
2
s (perpendicular) and p (parallel) polarisation with respect to the plane of incidence and
a cl C
αs ≡ tan 2θ c ; V is the V-number defined in appendix 3;
V
and a cl is the specific absorption of the cladding. In the literature a cl can be found with
different units: cm-1Pa-1 or cm2mol-1. An appropriate value for C has to be taken: in the first
case it is the partial pressure in Pa, in the second it is the concentration in mol/cm 3.
n 2
θ c = arcsin n cl is the critical angle and α p ≡ α s(2 − sin θ c ) .
co
The condition ni << ncl holds for all gases under normal conditions. For
methane, at 1 atm and room temperature, ni = 2.2427 x 10-4 at 3.392 µm (Stewart and
Chapter 5: Review of chemical optical fibre sensing 65
Culshaw, 1994). For dry air at 1 atm and room temperature, ni = 3.173 x 10-11 at the
wavelength of the P(16) CO2 laser line (Wolfe and Zissis, 1978). Even for SF6 at 10.548 µm
(very close to the P(16) CO2 laser line), 1 atm and 25 oC, ni is 4.036 x 10-3 (Chapados and
Birnbaum, 1988, from the graph in figure 4): 10.548 µm is one of two fundamental peaks in
the absorption spectrum of SF6 (see figure 7.1). Methane too has a strong absorption peak at
3.392 µm (Tai et al., 1987). These values were calculated using the equation (Stewart and
Culshaw, 1994)
ni = a C λ (5.9)
4π
and known values for the specific absorptions. In this equation, C is the concentration and a
is the specific absorption.
From equations 5.7 and 5.8 it can be seen that γ for narrow-beam
coupling depends linearly over a larger range than γ for wide-beam coupling, with the
operates in the linear regime under all conditions, because the quantity 8αsLΘδ is much less
than 3. Indeed, a value of 1.26 x 10-4 is obtained when the experimental parameters, listed
below are substituted in
8α sLΘδ (5.10)
where
α s = 1 a SF 6 C tan 2 θ c
V
V = 149765 ,
a SF 6 = 119 m2/mol ,
C = 52.426 mol/m3 , which is the concentration at 130 kPa and 25 oC,
Chapter 5: Review of chemical optical fibre sensing 66
L = 0.25 m ,
Θ = 19.677o = 0.34343 rad ,
δ = 0.015 rad, which is the divergence of the 1.5 mm diameter beam after passing through a
biconvex lens of 4.25" (107.95 mm) focal length,
metal εm is smaller than the dielectric constant of the glass εg . The dielectric constant of the
glass also has to be higher than the dielectric constant of the substrate εs (Richards et al.,
1993). The surface plasmons, which are surface electromagnetic waves (Dakin and Culshaw,
1988), propagate along the metal-substrate boundary with the surface plasmon propagation
constant given by (Janata, 1989):
ω 1
k sp = c (5.11)
1 1
εm + ε s
Chapter 5: Review of chemical optical fibre sensing 67
where c is the velocity of light and ω is the angular frequency of the incident light. εm is
frequency dependent and is given by
ωp 2
ε m (ω) = 1 − ω (5.12)
ρ
where ω p = 4πe 2 m is the plasma frequency for a free electron gas,
kp = εg ω
c sin θ (5.13)
E
H
θ
glass, ε g
metal, ε m
substrate, ε s
A resonance occurs when kp is equal to ksp . For a wavelength that satisfies this
condition there is strong attenuation of the incident light that can be detected through the
attenuation of the reflected light (Swalen et al., 1980). The intensity of the electric field of the
evanescent wave in the dielectric substrate is two orders of magnitude higher with the metal
layer on the glass plate than without the metal layer (Culshaw and Dakin, 1989). The substrate
can be a solid, liquid or gas material. Using plasmon surface resonance, refractive indices of
gases, liquids and solid substances can be measured very precisely. The technique is sensitive
enough to allow temperature measurement to be made from temperature-induced changes in
refractive index (Chadwick and Gal, 1993). Optic fibre sensors based on plasmon surface
resonance also can be designed. They can be made compact and structurally simple (Homola
and Slavik, 1996), although they are somewhat less sensitive than their bulk optic counterparts
(Kunz et al., 1996).
Optical fibres have been developed mainly for telecommunications. They are
usually silica-based or plastic-based with a small amount of dopant. Communications fibres
have minimum light attenuation at 0.9 µm, 1 µm, 1.2 µm, 1.3 µm and 1.55 µm (Gowar, 1984).
These fibres can be used for sensing but optimum performance is obtained with fibres
designed specifically for sensing applications. These include birefringent fibres, D-fibres, fibres
made of plastic mixed with luminescent material, fluoride fibres, silica fibres doped with
germanium, silver halide fibres, polymer clad silica (PCS) fibres, chalcogenide fibres, fluoride
fibres doped with heavy metals, hard plastic clad fibre, hollow fibres and nitrogen-doped silica
fibres.
Here, some optical fibres designed specifically for chemical sensing will be
described. High attenuation is not the problem for optical fibre sensors that it is for
communications, and for example, silica fibres are used for methane detection at 3.392 µ m
(Tai et al. 1987) in spite of the high absorption of silica at this wavelength. However, low
Chapter 5: Review of chemical optical fibre sensing 69
attenuation is sometimes an advantage even for sensing applications. For example, silver
halide fibres have a transmission range of up to 16 µm and this makes them ideal for infrared
spectroscopy (Schnitzer et al., 1990). They can be used with or without cladding. Those
without cladding essentially use the surrounding air as cladding. Chalcogenide optical fibres
have better mechanical properties (Busse et al., 1996) but their transmission range extends
only to 11 µm which is still good enough to allow them to be used for infrared spectroscopy.
Silver halide optical fibres will be described in more detail in the experimental section 5.3.2 on
SF6 sensing.
Silica fibres can be used for evanescent-field absorption measurements as some
diatomic molecules have absorption peaks in the visible spectrum: for example, Cl2 has an
absorption peak at 530 nm (Safaai-Jazi and Petersen, 1994). Polymer-clad silica fibres and
hard-plastic clad fibres are suitable for this application. Polymer-clad silica fibres have a silica
core, and cladding made of silicone. This makes the cladding easy to remove and the core can
be readily exposed to the surrounding medium. The cladding of hard plastic-clad fibres is
made of hard polymer that can be removed easily with chemicals without etching the core.
Optical fibre sensors which rely on phase or polarisation modulation use
single-mode fibres and polarised light. However, conventional single-mode fibres with circular
cross-section permit the propagation of two nearly degenerate modes with orthogonal
polarisation, and coupling between these modes results in the input polarisation not being
retained as light propagates through the fibre. This problem is overcome by using one of
several types of optical fibre which permit light to pass through whilst retaining its state of
polarisation. These polarisation- maintaining fibres can be classified into two general groups,
which are termed high-birefringence (Hi-Bi) and low birefringence (Low-Bi) (Dakin and
Culshaw, 1988). In one type of Hi-Bi fibre, the circular symmetry of the core is removed so
that the two modes have significantly different propagation constants: this hinders coupling
between modes so that light polarised along a principal axis of the fibre maintains its
polarisation. Fibres of this type include elliptical-core, bow-tie and PANDA. Another type of
Hi-Bi fibre allows only one polarisation mode to propagate by imposing a cut-off condition on
the other mode. Bow-tie and PANDA fibres can be designed in this way. On the other hand,
Chapter 5: Review of chemical optical fibre sensing 70
the residual linear birefringence within a conventional fibre can be substantially reduced by
introducing a high degree of circular birefringence. Methods of doing this, and thus producing
Low-Bi fibres, include twisting a conventional single-mode fibre after manufacture, and
rotating the glass preform during the fibre drawing process to produce spun fibre.
Polarisation-maintaining optical fibres do not find many applications in
chemical sensing. However, when a fibre is used to deliver light to a sensor based on surface
plasmon resonance, polarisation is important, and polarisation-maintaining fibres are used for
such devices (Richards et al., 1993).
Silver halide fibres are made by extrusion from a monocrystal of AgBrxCl1-x but
the extruded fibre has a polycrystalline structure. The size of the crystal grains in the fibre
increases with time, and attenuation increases due to this ageing effect as shown in figure 5.8.
Ultraviolet and blue light cause the formation of colloidal silver centres in silver halide fibres.
This changes the colour of the fibre from yellow to dark blue, and increases fibre losses at all
wavelengths. While the ageing process is not so important when these fibres are used for CO2
laser beam delivery, it is very important when the fibres are used for sensing. This increase in
attenuation must be taken into account in sensor design. The increase depends on many
factors, but mainly on the manufacturing procedure. To avoid any error in measurement due
to the ageing process there should be a reference fibre arm in the measurement system. The
reference fibre can be linked with the measuring fibre through a Y coupler, and it is relatively
easy to make a Y coupler with silver halide fibres (Eyal et al., 1994). Another possibility is to
send two beams at different wavelengths into the same fibre, one at the absorption peak and
another one far from that. Changing the intensity of the non-absorptive beam allows a
correction factor to be calculated.
Chapter 5: Review of chemical optical fibre sensing 72
72
Chapter 6
Optical fibres can be used for indirect monitoring of SF6 in an SF6 discharge.
During an SF6 discharge, dissociation processes cause free sulphur to be deposited on the wall
of the discharge chamber (Bunganaen, 1993). The technique applied here is based on the
measurement of the free sulphur deposited on an optical fibre placed in a discharge chamber.
A method has been developed for measuring the amount of deposited sulphur and it is
described in this chapter.
The first experiment for investigating sulphur deposition from an SF6 discharge
was carried out using a white light source. The experimental arrangement is depicted in figure
6.1. The light source is an ORIEL Q housing lamp with a 250 W, 24 V quartz tungsten halogen
bulb. Lenses made of BK7 glass are used to collimate and launch the light from the lamp into
the fibre. The fibre passes through a glow discharge chamber, as shown in figure 6.1. The
sealing of the chamber, where the fibre goes in and out of the chamber, is made with
Torr-Seal® adhesive (manufactured by Varian). The output end of the fibre is mounted to the
input of a SPEX MINIMATE-2 (SPEX Industries Inc., USA) 1/4 m monochromator with a
1200 lines/mm grating and 0.25 mm slits. The detector attached at the output of the
monochromator is an EMI 9558QB photomultiplier. The monochromator is equipped with a
scanning controller. Both the controller and the photomultiplier are connected to a National
Instrument acquisition board as part of a Macintosh computer. The spectrum from the output
Chapter 6: Sulphur deposition experiment 73
of the fibre is scanned and stored in the computer for further analysis. The output signal from
the photomultiplier is amplified with a photodiode current amplifier and the output signal is
set to between 0 and 5 V which is the range of the acquisition card.
glow discharge
focusing lens
chamber
lamp
optical
fibre
high voltage
power supply
photomultiplier
photodiode
current
amplifier
monochromator monochromator
servo motor
controller
Figure 6.1. Experimental arrangement for the measurement of sulphur deposition on an uncladed PCS optical
fibre, using a white light source.
reduced by removing the cladding and jacket. The core is made of silica, and with a thickness
of 0.6 mm, it is very brittle in the absence of the cladding and jacket. Since for unclad regions
longer than 5 cm the fibre needs to be handled very carefully, it was preferable to strip the
cladding chemically using Lumer optical fibre stripper "S". The main components of this
stripper are methylene chloride and sulphuric acid. For the first experiment with white light,
15 cm of the fibre was unclad and placed inside the chamber. Both cladding and jacket were
removed with the Lumer optical fibre stripper "S". The chamber was filled with SF6 to
0.96 Torr and a glow discharge was maintained for 7 hours. Spectra were taken every
30 minutes. At the same time as a spectrum was taken, a sample of the gas mixture from the
chamber was analysed using the mass spectrometer.
2000
1800
transmitted light (arbitrary units)
1600
1400
1200
1000
800
600
400
200
0
200 300 400 500 600 700 800 900
wavelength (nm)
Figure 6.2. The spectrum of the light transmitted through the fibre
discharge. Three samples of transmission spectra are shown in figure 6.4: the top curve is at
the beginning, before the discharge starts; the middle curve after 2 hours of discharge and the
bottom one, after 3 hours and 10 minutes. Absorption spectra are derived from the
relationship
transmission spectrum
absorption spectrum = 1 − top curve in figure 6.4
(6.1)
Two absorption curves are shown in figure 6.5, together with details of discharge times and
total charge transferred.
transmitted light (arbitrary units)
2000
1800
1600
1400
1200
1000
800
600
400
200
0
200 300 400 500 600 700 800 900
wavelength (nm)
Figure 6.3. The same spectrum as in figure 6.2 but filtered from noise.
From figure 6.5 it is seen that after 2 hours there is a relatively higher
absorption in the blue than over the remainder of the spectrum, but later, when a thicker layer
of sulphur builds on the optical fibre, the higher absorption in the blue region is less
pronounced. In later experiments, the white light was replaced with a 10 mW polarised He-Ne
laser (figure 6.6). A microscope objective (magnification = 60, numerical aperture = 0.85) was
Chapter 6: Sulphur deposition experiment 76
used to launch the laser light into the fibre. After a few experiments it was found that the best
sensitivity was obtained when the system was aligned as follows. The single-mode optical
fibre coupler that holds the fibre was first adjusted until the maximum reading at the output
was obtained; at that time the output of the fibre was a bright spot. Measurements were made,
however, with the fibre tilted until a distinctive doughnut pattern was observed at the output;
this time the output power was about 20% less than the maximum reading. This alignment
arrangement maximises the interaction between the laser light transmitted through the fibre
and the deposited sulphur. A silicon photodiode was used as detector and the signal was
recorded on a chart recorder after being amplified. A 3 cm length of cladding was removed
from the PCS 600N PUV fibre. The fibre was placed in the glow discharge chamber with the
unclad region close to the chamber wall. The glow discharge was maintained for 7 hours at
10 mA current, which gave a total charge transfer of 252 C. The intensity of the light
transmitted through the fibre was recorded on a chart recorder. The results of the absorption
measurements are discussed in section 6.5.
2000
1800
transmitted light (arbitrary units)
1600
1400
1200
1000
800
600
400
200
0
200 300 400 500 600 700 800 900
wavelength (nm)
Figure 6.4. The top curve is the starting one. The middle curve was obtained after 2 hours of
the glow discharge, and the bottom curve after 3 hours and 10 minutes of discharge.
Chapter 6: Sulphur deposition experiment 77
0.5
0.3
0.2
2 hours after
the discharge started
0.1
0.0
200 300 400 500 600 700 800 900
wavelength (nm)
Figure 6.5. Absorption spectra after 2 hours ( 72 C ) and 3 hours and 10 minutes ( 114 C ).
optical
fibre
chart
recorder
photodiode
transimpedance
amplifier
photodiode
Figure 6.6. Experimental arrangement for the measurement of sulphur deposition on an uncladed PCS optical
fibre, using a helium-neon laser.
Chapter 6: Sulphur deposition experiment 78
Electron microscope probe analysis (with X-rays) reveals that the deposit on
the fibre is sulphur. A piece of the 15 cm length of unclad PCS 600N PUV fibre, which was
placed in the glow discharge chamber for 7 hours, was used for probe analysis with the
electron microscope. The analysis requires the fibre to be coated with carbon. A piece of the
fibre close to the chamber wall, complete with sulphur deposit, was analysed in the electron
microscope. Figure 6.7 (b) shows the result of the analysis in which only sulphur, silicon,
n
oxygen and carbon were detected. Carbon comes from the coating, while silicon and oxygen
o
are components of silica (SiO2). Figure 6.7 (a) was obtained for reference using an unused
c
piece of PCS fibre core.
n
i
o
c
O x y g e n
l
i
i
l
i
S
S
Carbon
negO
yx
D
Carbon
r uhp l uS
C
a) b)
Figure 6.7. Electron microscope probe data: (a) fibre core (silica) in absence of a discharge (b) fibre core (silica)
following a 7-hour glow discharge. Quantitative comparisons of pulse amplitudes cannot be made because of the
nature of the electron-probe technique.
A part of the unclad fibre that was between the electrodes also was examined.
Probe analysis shows that there are other materials in the coating of this sample, such as iron
and chromium. They are released from the electrodes. The coating is cracked and uneven due
to the high temperature, up to 1000 oC, in the discharge centre (Ogle and Woolsey, 1987).
Electron micrographs of the fibre placed near the discharge wall and in the discharge centre,
are shown in figures 6.8 (a) and (b), respectively.
From figure 6.8 (b) it can be seen that the thickness of deposited sulphur for a
fibre from the centre of the discharge varies widely, since the high temperature of the
Chapter 6: Sulphur deposition experiment 79
discharge causes cracks and flaking of the sulphur deposits. This suggests that measurements
of sulphur deposition on a centrally positioned fibre are unlikely to be useful. The following
studies are concentrated therefore on measurements made using fibres positioned near the
wall of the chamber. Even in this region a few cracks do occur in the sulphur coating on the
fibre. These can be seen in figure 6.9 which has a magnification of 30 times that of figure
6.8 (a).
a) b)
Figure 6.8. a) An SEM (scanning electron microscope) picture of the fibre placed near the wall of the glow
discharge chamber. The width of the picture is 880 µm and the SEM accelerating voltage is 20 kV. b) An SEM
picture of the fibre in the centre of the glow discharge. The picture width is 880 µm, and the accelerating voltage
is 20 kV.
Chapter 6: Sulphur deposition experiment 80
a) b)
Figure 6.10. The cross section of the fibre with deposited sulphur: a) after 7 hours of discharge (the picture width
is 880 µm, SEM accelerating voltage 20 kV); (b) after exposure to four separate glow discharges, each of
seven-hours duration (the picture width is 13.2 µm, SEM accelerating voltage 20 kV)
was made. This suggests that the sulphur deposition following one discharge is around
600 nm.
6.5. Experimental results and discussion
The curve shown in figure 6.11 was obtained when the He-Ne laser was used in
the transmission experiment. In this figure there is a distinctive step that occurs as a sulphur
layer builds up on the optical fibre. The intensity of the output light starts to fall after 75 C of
charge has been transferred through the glow discharge (point 1). The mass spectrometer
analysis shows that at that time there was approximately 25% of the initial concentration of
SF6 remaining in the chamber. This leads to the conclusion that the sulphur deposited on the
chamber wall and fibre is produced from product gases rather than directly from SF6. The
concentration of SF6 left in the chamber following the transfer of 140 C (point 2) is around
10% of the initial concentration.
6
light transmission (arbitrary units)
5
1
4
3 2
0
0 50 100 150 200 250 300
charge (Coulombs)
Figure 6.11. The output from the optical fibre when sulphur was being deposited on it.
The curve of figure 6.11 was obtained each time the fibre was cleaned and a
further SF6 discharge generated using a fresh filling of SF6 . When a further discharge was run
without removing sulphur from the fibre there was no further change in the light transmission.
Chapter 6: Sulphur deposition experiment 82
Also, it was observed that light escapes along the length of the unclad fibre following
deposition of sulphur. This is the scattered light that is coupled into the sulphur layer by
refraction or via the evanescent field.
There are several mechanisms which may account for the observation of output
signal attenuation and emission along the length of the unclad fibre.
Sulphur is a biaxial material and its refractive indices along three axes are 1.95,
2.043 and 2.24 (Gray, 1972). These values are higher than the refractive index of the core of
the fibre and so total internal reflection should not occur. However, the refractive index of the
deposited sulphur layer may be lower than these measured values because the pressure in the
chamber is high (1 Torr) and the deposited sulphur has a sponge-like structure as shown in
figure 6.10. (b). If the refractive index of the sulphur layer is higher than the refractive index of
the core, the light will escape mainly by refraction and point 2 in figure 6.11 will be reached
when the exposed core becomes fully covered by sulphur.
There is also the possibility of the existence of a thin gap between the core and
the sulphur layer. In this case frustrated total reflection will occur and the light will propagate
in the same direction through the sulphur as it would in the absence of a gap, but with a lower
intensity.
Another possible mechanism involves penetration of the evanescent field of the
light in the core into the sulphur layer. As shown below, the penetration depth is expected to
be around 100 nm. In order to estimate how deep the evanescent wave spreads around the
fibre, it is assumed that the refractive indices of the core (fused silica) and the cladding
(polymer) are 1.457 (Malitson, 1965) and 1.400 (Gowar, 1984, page 86), respectively. Because
the light is launched into the fibre at an angle, as described in section 6.3, in order to maximise
the evanescent field, and because the core is very thick (0.6 mm diameter), the propagation
can be treated as propagation in a planar waveguide. In that case the critical angle for the
glass/polymer boundary is 73o 55'. In the unclad region the critical angle for the glass/vacuum
boundary is 43o 20'. When the light reaches the unclad region only the rays between 73o 55'
and 90o are present. The other rays have been refracted into the cladding before they reach the
Chapter 6: Sulphur deposition experiment 83
unclad region. Using the wavelength of the He - Ne laser, 633 nm, and 73o 55' for the angle of
incidence, the equation for the penetration depth (see appendix 4)
λ0
dp = (6.2)
2
2π n 1 sin 2 α − n 2
n
1
gives 103 nm for the penetration depth (dp), where λ 0 is the vacuum wavelength. α is the
angle of incidence to the normal of the interface (core - gas in this case) and n1 , n2 are the
refractive indices of the core and surrounding SF6 gas, respectively (n1 = 1.457 and n2 = 1). dp
is the distance over which the evanescent field amplitude falls to e-1 of its value at the
boundary when the core is in a vacuum or gas. This is in acceptable agreement with the
measured thickness of the deposited sulphur: 600 nm (see section 6.4). The derivation of
equation 6.2 is presented in Appendix 4.
6.6. Conclusion
The result suggests that this optical fibre method of sulphur deposition
measurement could be appropriate for the detection of SF6 dissociation in high-voltage SF6
switchgear. Conditions in the present study are very different to those encountered in a
practical gas-insulated high voltage device. In this experiment the glow discharge chamber was
filled with 0.96 Torr of SF6 . Sulphur deposition on the optical fibre was detectable when 75%
of the initial SF6 concentration in a glow discharge at 0.96 Torr was dissociated: this
corresponds to 0.72 Torr (0.00095 atm) partial pressure. The pressure in a high-voltage
switching device on the other hand, can be as high as 10 atm. A pressure of 0.72 Torr is only
0.0095% of 10 atm. This suggests that the optical fibre deposition method could provide a very
sensitive method of measuring the degradation of SF6 in a high-voltage switch.
The sensitivity of the sensor could be tailored to suit the particular switch by
Chapter 6: Sulphur deposition experiment 84
Chapter 7
Figure 7.1. The SF6 infrared absorption spectrum (from Lagemann and Jones, 1951)
The experimental arrangement for measuring the amount of SF6 in the corona
discharge chamber is shown in figure 7.3. The laser used was a SYNRAD 48G-2-28w CO2
Chapter 7: Experiment on CO2 laser light absorption 87
grating laser. A circular aperture of 1.5 mm diameter was placed in front of the laser output. A
ZnSe beam splitter divided the beam in the ratio 2:1, with two thirds of the power entering the
monochromator. One third was incident on the ZnSe biconvex lens, which collimated the
light, so that it was launched into the fibre at an angle of 45o to the fibre axis. The fibre passed
through a 25 cm long cylinder connected to the glow discharge chamber as described in
chapter 3. The glow discharge chamber was connected to the corona chamber via a 1.5 m long
tube as described in chapter 3. A pyrodetector (Murata IRA-E600S0) detected the infrared
light at the output of the fibre. The signal from the pyrodetector was amplified, extracted with
the lock-in amplifier ORIEL model 70707, and recorded on a chart recorder. The reference
signal for the lock-in amplifier was supplied by the digital signal generator that modulates the
laser. The signal from a pyrodetector ORIEL model 70841 attached to the monochromator
was amplified with an ORIEL amplifier/readout model 70701, extracted with the lock-in
amplifier PAR model 124, and recorded on a chart recorder.
The signal from this pyrodetector is proportional to the power of the laser and
is used to compensate any fluctuation in the laser power. The pressure and temperature in the
corona chamber were monitored. These values are used to calculate the pressure of the same
Chapter 7: Experiment on CO2 laser light absorption 88
density at 25 oC and the percentage of SF6 in the corona discharge chamber. During a corona
discharge the pressure rises as dissociation occurs, because more molecules are produced.
Dissociation of SF6 molecules in the discharge environment can lead to several events. These
include sulphur deposition on the chamber wall, reactions which produce sulphur
oxyfluorides, and production of SiF4 as a result of reaction between fluorine and glass
windows. The net effect of all such events is to produce a significant number of additional
molecules and a consequent rise in the total gas pressure in the corona chamber.
Z n S e
Timer/Counter split ter
micrometers
Digital
signal
generator
grating
Pyrodetector
Corona Murata
chamber
(see Fig. 7.5)
Lock-in Pyroelectric
amplifier amplifier
detector head Lock-in amplifier
ORIEL
Figure 7.3. Experimental arrangement for measuring evanescent-field absorption during an SF 6 corona discharge.
These cooling conditions are provided by an attached cooler NESLAB RTL-211D. Two
micrometers on top of the laser are used to select the lasing line. The large one is used to
change the angle of the grating. It selects the laser spectral line. The smaller one is used to
adjust the length of the optical cavity. Using this micrometer it is also possible to change the
wavelength within a few tens of megahertz around the spectral line. This also happens if the
temperature of the cooling water is not constant. If the temperature is not constant the laser
stays within the defined line, but drifts a few megahertz on either side. At maximum output
optical power, the grating and cavity length are set for the required spectral line. TEM mode
changes were observed during laser warm-up prior to experimental run. During the
experiment, mode changes were not observed.
Figure 7.4. The absorption coefficient of SF6 around the CO2 laser spectral
lines (P12, P14, P16, P18 and P20). The width of the x-axis for one spectral
line is estimated to be between 40 and 50 MHz (from Shimizu , 1969). The
arrows show the centres of the spectral lines.
The centres of these lines are (Maki et al., 1994):
P12 - 10.5131216772 µm
P14 - 10.5320883339 µm
P16 - 10.5513950296 µm
P18 - 10.5710454539 µm
P20 - 10.5910434604 µm
7.1.2. Source-fibre coupling
Chapter 7: Experiment on CO2 laser light absorption 90
The coupling of laser light into the fibre is not very efficient but this is not a
problem. Indeed, the power after the ZnSe lens must be maintained below 60 mW. The reason
for this is that, in the same way that the SF6 absorbs light via the evanescent field, other
materials in contact with the fibre also absorb light. Torr-Seal®, a two-component adhesive for
vacuum applications, is used to seal the fibre at ends of the cylinder, and where the Torr-Seal®
contacts the fibre it absorbs CO2 laser light and thus heats up. This heat is transferred to the
fibre and softens it at those points. At higher power the fibre melts at the point closer to the
input.
The limit of 60 mW was set for a launching angle of the beam into the optical
fibre of 19o. When the launching angle increases this limit increases, because more power is
reflected from the input end of the fibre. Equation 7.1 can be used to show that the intensity of
the light reflected from the fibre input face increases (when the light is coupled into the fibre)
when the incident angle with respect to the fibre axis is increased. Also, it will be used to
estimate the power absorbed by the Torr-Seal®. The reflection coefficient from a flat dielectric
surface, defined as the ratio of the reflected and incident electric field wave amplitudes, is
given by (Collin, 1991) :
ZT − ZI
R= (7.1)
ZT + ZI
where
constant, Z0 is the impedance of vacuum and n is the refractive index. There are two sets of
equations for R, ZT and ZI . One is for s-polarisation and one for p-polarisation. However,
only s-polarisation will be considered here. In figure 7.3, the fibre lies in the plane of the page
and the laser beam is polarised in the direction normal to the page. Consequently, the laser
Chapter 7: Experiment on CO2 laser light absorption 91
beam enters the fibre as an s-polarised beam and propagates through the fibre as mainly TE
mode. This provides better sensitivity in an FEFA sensor than TM mode coupling (see the
calculation in section 5.2.3.1). For s-polarisation,
ZS = E (7.2)
H cos θ
which finally provides an expression for the electric field reflection coefficient of the
s-polarised light
n I cos θ I − n T cos θ T
RS = . (7.3)
n I cos θ I + n T cos θ T
of 38 mW was obtained using 5.8 µs wide pulses. This width was measured with a digital
counter (see figure 7.3). Then the power meter was removed and the pulse width, as
monitored by the digital counter, was maintained at around 5.8 µs. For all measurements; at
different incident angles and during the corona discharge, no power meter was used. This
method of monitoring the laser power was preferred to the use of a beam splitter to direct
some fraction of the laser beam to a power meter, since the introduction of another optical
component adds noise to the system.
7.1.3. Pyrodetectors
9V G
47K Ohm
Output
Figure 7.5. The electric circuit for the mid-infrared light detector used for the measuring
the light at the output of the fibre.
After adjustment of the laser power level, the power meter at the output of the
fibre was replaced by the Murata IRA-E600S0 pyrodetector which is used for all further
measurements of the output light from the fibre because the precision and the resolution of the
power meter is much lower than those of the pyrodetector. This detector incorporates a
piezoelectric crystal that generates electrical charge when pressure is applied to it. A
temperature change produces a similar effect. The power of the light from the fibre raises the
temperature of the crystal. A constant heat input to the detector does not generate a signal, for
which reason the light beam must be chopped. Here, the laser beam was modulated with a
Chapter 7: Experiment on CO2 laser light absorption 93
17 Hz, 50% duty cycle square-wave signal from a digital signal generator. The piezocrystal is
encapsulated in a transistor-like case together with a FET transistor. A silicon window in the
case transmits light from 1 µm to 20 µm. To obtain a signal from the detector, the circuit
shown in figure 7.5 was assembled.
The monochromator used in the experiment (see figure 7.3) was equipped with
an ORIEL pyrodetector model 70841 with a KBr window, which operates on the same
principle as the Murata unit.
into the previously evacuated glow chamber. The chamber was filled slowly, while the fibre
output signal, Sf , was recorded on a chart recorder. The signal from the monochromator for
monitoring power was recorded on another chart recorder, Sp . Chamber pressure was
measured with the VAISALA PTB200 A digital barometer described in chapter 4. The
temperature of the chamber was measured with a temperature probe attached to the wall of
the chamber. Pressures were normalised to 25 oC. Values of Sf / Sp were recorded as a
function of pressure, with Sf / Sp equal to one for zero pressure. Data were obtained both
when SF6 was slowly admitted to the chamber and when it was slowly pumped out of the
chamber. Measurements for different incident angles were made in order to check the validity
of equation 5.7.
During these experiments it was important that no leakage of SF6 occurred into
the laboratory, as can happen from exhaust lines. If SF6 is released into the laboratory, it can
remain for a long time, since the heavy SF6 molecules replace the air at ground level. This
"atmosphere" of SF6 absorbs CO2 laser light and even a small number of randomly moving
SF6 molecules in the path of the laser beam can cause fluctuations in the final detector signal.
The effect was observed in the present work, before careful sealing of the exhaust line was
completed.
7.2.2. SF 6 monitoring
A typical corona - monitoring experiment took nine days. The initial pressure
of SF6 was set at 90 kPa and the distance between electrodes (point and plane) at 10 mm. The
current was maintained between 85 µA and 102 µA at a voltage of 35 kV. The current was
recorded on a chart recorder. The total amount of charge transported during the nine-day
discharge period was 63 Coulombs.
During this long time period the room temperature fluctuates and so the laser
wavelength varies around the lasing line. This can lead to substantial errors as SF6 absorption
varies significantly over a frequency range of a few megahertz (figure 7.4). From observation
of the chart recorder it is clear when the laser is unstable: large fluctuations in power occur.
Measurements obtained during laser instability were not used for data analysis.
Chapter 7: Experiment on CO2 laser light absorption 95
The partial pressure of SF6 during the discharge was calculated in the following
manner. An interpolation formula was first established to fit a calibration plot of absorbance as
a function of SF6 pressure (section 7.4.1). This formula has the form
2
pressure SF 6 (kPa) = A + B ln(signal) + C ln (signal) (7.4)
where signal is the normalised signal obtained from the fibre output and the coefficients A, B
and C are calculated for the pressure range between 50 kPa and 90.1248 kPa. This range was
chosen because the partial pressure of SF6 during the discharge fell from 90 kPa to around 70
kPa. Equation 7.4 with these coefficients was then used to calculate the partial pressure of SF6
during the discharge from the measurements of fibre output.
The pressure and temperature in the chamber were monitored at all times
during the experiment because the total pressure in the chamber continuously rises because of
SF6 dissociation. The pressure and temperature values were used to calculate the
concentrations (as ratios) of the SF6 and discharge by-products in the chamber at 1 atm and
25 oC. The partial pressure of the by-products is determined by subtracting the normalised
partial pressure of SF6 from the normalised total pressure.
Data for two incident angles θ (14o and 29o) were obtained for the transmitted
1.6
1.4
Absorbance (arbitrary units)
o
launching angle = 29
1.2
1.0
0.8
0.6
launching angle = 14o
0.4
0.2
0.0
0 20 40 60 80 100 120 140
Pressure (kPa) at 25 oC
Figure 7.6. Calibration curves: evanescent-wave absorbance of CO2 laser P(16) line versus pressure of SF6 at
25 oC for silver halide optical fibre, at launching angles of 29o and 14o, with the lines of the best linear fit.
After these measurements were completed the incident angle was set to 45o in
order to increase sensitivity further. For this setting it was necessary to slightly bend the fibre.
Unfortunately, this introduced a nonlinearity into the absorbance curve (figure 7.7). The
mechanical characteristics of a silver halide optical fibre are somewhat similar to those of
copper wire. The fibre is soft and has low elasticity, so that attempts to straighten the fibre
were unsuccessful. As a result, the measurements for 16o and 30o incident angles are also
non-linear. The three curves are shown in figure 7.7.
The bend introduces higher order modes. The effect is similar to that of
launching two laser beams into the fibre at different angles. In that case the absorption
coefficient becomes (see equation 5.7)
γN L = Lα Θ 21 + δ 21 + Lα Θ 22 + δ 22 (7.5)
The effect can be observed in figure 7.7, for the 45o curve, where it appears that
the bend creates more than one additional mode.
Chapter 7: Experiment on CO2 laser light absorption 97
1.6
1.4
1.2 launching angle = 30o
1.0
0.8
0.6
0.4 launching angle = 16o
0.2
0.0
0 20 40 60 80 100 120 140
Pressure (kPa) at 25 oC
Figure 7.7. Calibration curves: evanescent-wave absorbance of CO2 laser P(16) line versus pressure of SF6 at
25 oC for silver halide optical fibre, at launching angles of 45o, 30o and 16o, with the lines of the best linear fit.
The red and green triangles were recorded at the same incident angle. The red correspond to gas filling the
chamber, and the green to the gas being pumped out of the chamber.
From both figures 7.6. and 7.7 it can be seen that the highest sensitivity was
reached when the CO2 laser beam was coupled to the fibre with an incident angle of 45o. This
is in agreement with theory and equation 5.7. Further increase in the incident angle increases
the sensitivity, but because of mechanical constraints it was not possible to go above 45o.
100 40
90 36
80 32
70 28
60 24
50
0 10 20 30 40 50 60 70
Charge transported (Coulombs)
Figure 7.8. The amount of SF 6 during a corona discharge. The solid line is the best linear fit. The corona current
was between 85 µA and 102 µA. The distance between point-plane electrodes was 10 mm. The applied voltage
was around 35 kV. The initial pressure of SF 6 was 90.1248 kPa at 25 oC.
Energy (kJ)
0 500 1000 1500 2000
100
90
80
SF6
70
Concentration (%)
60
50
40
30
Discharge by-products
20
10
0
0 10 20 30 40 50 60 70
Charge transport (Coulombs)
Figure 7.9. Concentration of SF6 and discharge by-products in a corona chamber. The solid lines are the best
linear fits. Same corona conditions as figure 7.8.
This work shows that an FEFA sensor can be used to monitor the degradation
of SF6 in a discharge. The fluctuations of the experimental points in figures 7.8 and 7.9 are
probably due mainly to variations in the equilibrium of the laser. The experimental data on
each of figures 7.8, 7.9 and 7.10 depict an overall linear trend and lines of best fit have been
drawn to illustrate this. On the other hand, a plateau region on each curve may possibly exist
in the range 20 C to 40 C. The fact that this trend is observed in both the present work and that
of Mortensen et al. may be a coincidence but it is something that warrants further study. It
should be possible to reduce laser fluctuations using a more stable CO2 laser, and this might
Chapter 7: Experiment on CO2 laser light absorption 100
help determine whether or not a plateau region does in fact exist. Furthermore, it may be that
a lower power laser is more appropriate for these long-term measurements, since the output
power of such lasers is easier to control and the power dissipation is lower.
The chalcogenide fibres have the advantage of showing no ageing effect. Also, their durability
is better.
Another disadvantage of the FEFA sensor for SF6 monitoring is the high cost
of CO2 lasers. However, in an electrical plant containing several high-voltage insulating
devices, sensing fibres could be placed in each, and the CO2 laser radiation delivered from a
single laser.
Since a CO2 laser is much more powerful than a laser diode, pyrodetectors can
be used rather than the more sensitive, but expensive, cryogenic antimonide or HgCdTe
detectors.
The problem of wavelength drift can be solved in several different ways. For
example, the small micrometer that is used to tune the resonant cavity length of the laser
described in section 7.2.1 could be replaced with a piezocrystal that would vary the cavity
length and lock onto the lasing line P(16). Otherwise the piezocrystal could be used to scan the
frequencies around the lasing line, and measure the output signal from the fibre only when the
laser goes through the maximum point corresponding to the lasing line P(16). Such a device
was used in the experiments of Shimizu (1969).
102
Chapter 8
In the electrical power industry, there is a need for a simple and reliable method
of measuring the degradation of SF6 in switchgear and other SF6-filled high voltage devices.
Such information can be useful for determining service-times for switchgear, and when the
SF6 should be replaced. At the moment there appears to be no sensor appropriate for such an
application. Samples can be taken from the switchgear and analysed with laboratory
instruments such as FTIR, but this is not cost effective, particularly when examining power
systems in remote areas.
In the present work three methods have been devised for monitoring SF6
degradation, and each has the potential to be made simple and compact enough to be built
into switchgear systems. The three methods involve measurements of SF6 refractivity, sulphur
deposition, and evanescent-field absorption of SF6 at 10.6 µm.
The last two methods, based on optical fibre sensing, may be closer than the
first one to implementation as a final product for field measurement. Both can be used in high
voltage devices. A sensor based on sulphur deposition can be made extremely simple, in that
it requires only a source such as a red laser diode, a length of PCS optical fibre with a declad
section and a detector. In order to avoid interference from the high voltage device, the source
and detector can be placed remote from the device and the light can be delivered to and from it
with optical fibres. This sensor is irreversible but this should not be a problem, since it needs
to be replaced only when the SF6 is replaced, and at that time maintenance involving
dismantling of the device is likely to take place. The interferometer described in chapter 4 for
refractivity measurement is neither simple nor compact and is not appropriate for field
monitoring. However, the data obtained with the interferometer have shown that refractivity
Chapter 8: Summary and conclusions 103
measurement of the gas inside a high voltage device can provide useful information about the
insulating condition of SF6.
SF6 monitoring is important in other devices that use SF6 insulation. For
example, when the amount of SF6 by-products reaches 0.1 ppm in the ANU 14UD accelerator
at the Australian National University in Canberra, ACT, the SF6 has to be purified (Ophel et
al., 1983). The sampling method used for measuring this concentration of by-products is not
practical for switchgear because a few litres of SF6 are released during each measurement.
According to Ophel the concentration of by-products in a circuit breaker should be detected
with at least 100 ppm accuracy. This value is likely to vary for different switchgear systems:
circuit breakers are known to operate efficiently with contaminant levels up to 0.2%, such as
the one shown in figure 2.2 (Yamauchi et al., 1985).
The sulphur deposition method is the most sensitive of the three techniques.
The experiment described in chapter 6 was carried out at a pressure of 0.96 Torr of SF6 , and
changes in deposition were readily detected. Indeed, in a high pressure device the deposition
sensor could be too sensitive, with saturation occurring soon after the first fall in optical
transmission (figure 6.11). However, the distribution of sulphur through the system is not
uniform, and by choosing the right position in the chamber or by placing a shield around the
fibre it should be possible to tune the sensor for appropriate SF6 pressure and sensitivity.
The sensor based on evanescent-field absorption by SF6 at 10.6 µm is less
sensitive than the refractivity sensor. From the 45o calibration curve of figure 7.7 it can be seen
that the SF6 absorption sensor detects SF6 pressure changes down to 0.67 kPa (~5 Torr), at
filling pressures of around 100 kPa, this limit being set by the sensitivity of the lock-in
amplifier. Increasing the incident angle of the CO2 laser light into the fibre can increase the
sensitivity further. Increasing the length of the fibre and choosing a more sensitive detector,
such as an indium antimonide or HgCdTe photoconductive detector operating at 77 K,
increases the sensitivity. In the long-term measurements on the corona discharge, the expected
sensitivity of 0.67 kPa was not achieved because the CO2 laser did not remain stable over long
periods of time. With a more appropriate laser or laser diode it would be possible to achieve
the expected value of sensitivity.
Chapter 8: Summary and conclusions 104
In a circuit breaker changes in refractive index are larger than those experienced
in the present work, because of the higher pressures used in the former. This means that a
refractometer based on surface plasmon resonance could be considered as a simple solution
(Jorgenson and Yee, 1994). Its sensitivity of 5 x 10-5 may be enough for monitoring high
pressure systems. For low pressures, a bulk-optic device based on surface plasma resonance
has a resolution of 4 x 10-6 (Kunz et al., 1996). One of these refractometers together with a
precise pressure meter and thermometer would be able to reach the sensitivity of 10-4 required
for circuit-breaker monitoring. The arrangement chosen needs to be resistant to the high
electric and magnetic fields and the electromagnetic radiation that exist in switchgear under
usual working conditions.
Evanescent-field absorption at 10.6 µm is a direct method and it does give
information about the remaining SF6 , which is the information required. The refractive index
measurement and sulphur deposition methods require separate calibration for each
high-voltage device and if there is some defect in the device, such as a high level of water
vapour in the fill of SF6 , a wrong measurement can result. Because of this, precautions must
be taken and sensors for other parameters may have to be included. The SF6 absorption sensor
is almost immune to such problems, because it provides information about the SF6 wherever it
is placed. However, at the moment it would be the most expensive option, mainly due to the
high cost of the 10.6 µm radiation source. Furthermore, a CO2 laser is bulky, a lead salt
infrared laser diode requires a cryogenic cooler, and the life time of a high power tungsten
filament halogen light bulb or Nernst glower is short. The cost of a lead salt laser diode for
10.6 µm is around US$4000, but it needs a cryogenic cooler which costs a further US$15 000.
Liquid nitrogen can be used to cool some of these diodes. Another solution could be an
infrared light-emitting diode, operating at room temperature. Work on mid-infrared LEDs at
room temperatures is still in the experimental stage and the maximum peak output power
achieved so far is 10 µW (Tang et al., 1995B). These are currently under development in a few
laboratories.
The work presented here has demonstrated the feasibility of the three optical
sensors. The sulphur deposition method has enough sensitivity to be employed for SF6
Chapter 8: Summary and conclusions 106
monitoring in switchgear, while the other two are close to the targeted sensitivity of 10-4, but
more work is needed to produce a practical sensor. Indeed, more work is required to improve
all three techniques.
107
Appendix 1
For this calculation the equation for molar refractivity A (or atomic refractivity
in the case of monatomic substances) will be used. It is defined as (Born & Wolf, 1980):
A = 4π N m α = RT n2 − 1 (A1.1)
3 p n2 + 2
where:
refraction have been neglected; that is, it is assumed that coefficient B(T) (equation 4.4) is
equal to zero. This assumption matches reality for pressures below 1 atm at room
N2 atoms of molar refractivity A2 , then the molar refractivity of the molecule is (Born & Wolf,
1980)
A = A 1N 1 + A 2 N 2 (A1.2)
Appendix 1: The estimation of the SOF4 and SO2F2 refractive indices 108
Suppose that, in order to calculate the refractive index of SOF4 , we assume that
it is the combination of one SOF2 molecule and one F2 molecule. The index of refraction of F2
is 1.000214 at 101325 Pa and 0 oC measured at 589.4 nm (Landolt - Börnstein, 1962). As molar
refractivity is constant for a particular gas and does not depend on temperature and pressure,
equation (A1.1) can be used to calculate the index of refraction at 25 oC when it is known at
0o C. This gives the refractive index of 1.000196 at 25 oC and 101325 Pa. Also, 589.4 nm is
close to 632.99 nm and it can be assumed that the index of refraction is the same at both
wavelengths. The index of refraction of SOF2 is 1.00062166 at 101325 Pa and 25 oC (this
work, chapter 4). Using these values and equations (A1.1) and (A1.2), the value of 1.000818 is
obtained for the refractive index of SOF4 .
A similar approach is used to calculate the refractive index of SO2F2 ; as the
combination of a molecule of SO2 and a molecule of F 2 . The refractive index of SO2 is
1.000628 at 101325 Pa and 25 oC, calculated from 1.000686 at 101325 Pa and 0 oC at 589.6 nm
(Gray, 1972). This gives 1.000877 for the refractive index of SO2F2 at 101325 Pa and 25 oC.
In the determination of the refractive indices of SOF4 and SO2F2 , it has been
assumed that the gases are linear. Any non-linear contributions to the refractive indices are
likely to be small: for example, in the case of SF6 the contribution of B(T) is only 8x10-6,
although SF6 is a very non-linear gas in terms of refractive index and compressibility in
comparison to other gases. Therefore, although the values of 1.000877 and 1.000817 for the
refractive indices of SO2F2 and SOF4 , respectively, may not be exact, the conclusion that their
refractive indices are higher than the refractive index of SF6 , which is 1.00070296 at 1 atm and
25 oC at 632.99 nm, should be valid.
109
Appendix 2
Programs, "fill_in" and "pump_out", were written for the acquisition of data
during the refractive index measurement when the gas is bled in and out of the invar tube.
These programs control the A/D converter, receive data from the A/D converter, store the data
and receive data from the VAISALA barometer via a serial port. The A/D converter is in a
CONTEC ADC-10 card. It is a 12-bit A/D converter with a maximum speed of 30 ksamples/s.
With this program the sampling speed is 15 ksamples/s on an IBM PC compatible computer
with INTEL 386SX-16 processor, math cooprocessor and 4 MB memory. This speed is
sufficient, because the amplified signal from the photodiode which is a part of the polarisation
interferometer (see figure 4.1) is around 1 Hz. The programs were compiled with Borland
Turbo C++ ver. 3.0 for DOS. The VAISALA PTB 200A digital barometer sends ASCII string
data via the serial port. An interrupt 0x14 from BIOS was used to receive this string.
The photodiode signal I from the interferometer is given by (Born & Wolf,
1980):
I = A cos 2 δ (A2.1)
2
where A is a constant that depends on beam intensities and the sensitivities of the
photodiode, δ is the phase difference between the two beams of the interferometer, and δ
changes with time as the pressure in the invar tube changes. The programs monitor the
maxima and minima of this function. When the signal passes through a quadrature point the
110
computer acquires the pressure information from the VAISALA barometer. If the pressure is
not recorded at the quadrature point the program makes the required correction. The fringe
number and pressure are stored on the hard disk.
The flow chart in figure A2.1 represents the program for fringe counting and
pressure. The fringe-intensity signal is the signal from the photodiode (see figure 4.1)
measured with a 12-bit analog/digital converter CONTEC in the computer. The program for
controlling this converter is part of the FILL_IN program. It also displays the signal and the
pressures. When the measuring system is evacuated, the program is started and the valve is
opened.
A driver for the A/D converter was not available for Turbo C++, so a driver was
written and is shown in the listing A2.1.
end
Figure A2.1. The flow chart of the program for refractivity data acquisition.
Appendix 2: Data acquisition programs for the refractive index measurement 112
The pressure from the VAISALA PTB-200A barometer via the serial interface
is read using a BIOS interrupt 0x14. The subroutines for retrieving data from this port are
shown in listing A2.2. The data from the barometer are sent as ASCII data. These are taken
from the computer at a time determined by the main program (FILL_IN). The subroutine
takes 40 successive characters and the actual pressure is extracted in the main
program.
Listing A2.2. The subroutines for receiving a character from the serial interface and for receiving 40 successive
characters from the same port.
Appendix 2: Data acquisition programs for the refractive index measurement 113
This program collects data when gas is pumped out of the measuring cylinder
of the interferometer. The program is similar to the program FILL_IN except that it runs in the
opposite direction: it starts at 115 kPa and stops at 57 kPa.
114
Appendix 3
of the core and cladding are ε1 and ε2 respectively, then ε1 > ε2 . The radius of the fibre is ρ
and it can be assumed that the radius of the cladding is infinite and both regions are perfect
insulators. If a cylindrical coordinate system is chosen with the z - axis lying along the axis of
the fibre, then the solutions for the electric and magnetic components of the electromagnetic
wave travelling in the fibre are (Snitzer, 1961):
where ω is the angular frequency; β = 2π /λg is the propagation constant along the axis of
the fibre and λg is the wavelength along the fibre (see figure A3.1); ϕm and ψm are phase
the tangential components of the field are the same on both sides of the boundary. Jm(h1r)
and Km(h2r) are the Bessel function and the modified Hankel function of the first kind,
respectively. The propagation constants k1 and k2 , and h1 and h2 are defined by (k1 and k2 are
constants for a fibre, but h1 and h2 have different values for each mode)
k 2 = ω 2µε (A3.5)
h 21 = k 21 − β 2 (A3.6)
Appendix 3: Equations for light propagation in an optical fibre 115
h 22 = β 2 − k 22 (A3.7)
Cladding
n2
Core θ
n1
λg
z - axis
λ /n1
Figure A3.1. Considering geometrical optics propagation, the wavelength λg along a fibre (z - axis) of a certain
mode is
λ ; λ is the free-space wavelength.
n 1 sin θ
in terms of E z and Hz (Stratton, 1941 Chapter V). After taking the boundary conditions
(4 equations) into account, expressions for the field components in the core are obtained:
E z = J m (h 1 r) F c (A3.8)
β m Jm
Er = i Jm − P Fc (A3.9)
h1 h1 r
β m Jm
Eθ = i P Jm − Fs (A3.10)
h1 h1r
β
H z = − µω P J m F s (A3.11)
Appendix 3: Equations for light propagation in an optical fibre 116
k 21 β 2 m Jm
Hr = − i P 2 Jm − Fs (A3.12)
µ ω h 1 k1 h1r
k 21 h 2 m Jm F
Hθ = i Jm − P 2 s (A3.13)
µ ω h1 k 1 h 1r
where
µω B m cos(mθ + ψ m )
P=−
β A m sin(mθ + ϕ m )
−1
1 1 J m (h 1ρ) Km (h 2 ρ)
=m + 2 h ρ J (h ρ)
+ (A3.14)
( h 1 ρ )
2
( h 2 ρ ) 1 m 1 λ 2 ρ K m (h 2 ρ)
and J m is the first derivative of the Bessel function with respect to its argument. For a
multimode fibre there are many values for β (propagation constant along the fibre axis) that
satisfy these equations. Each of them represents a propagation mode. For every cross-section
geometry and refractive-index profile it may be stated that (Snyder and Love, 1983):
n cl 2π < β ≤ n co 2π (A3.17)
λ λ
where ncl and nco are the refractive indices of the cladding and the core, respectively. The
β = k2 (A3.18)
Appendix 3: Equations for light propagation in an optical fibre 117
where λ is the free-space wavelength and h1mj is the value of h1 at cutoff; that is, the jth root
of the cutoff condition involving the mth order Bessel function (Snitzer, 1961). The right-hand
term in equation (A3.19) is called the waveguide parameter, the waveguide frequency or the
V-number of a fibre, and is written as (Snyder and Love, 1983):
For multimode fibres V >> 1 and for step-profile circular single-mode fibres V < 2.405.
Figure A3.2. Left: Schematic picture of bound (within red semi-cone) rays, refracting (within magenta cone) rays
and a sample tunnelling ray. Right: Actual picture within the fibre showing the cones of bound and refracting rays
touching along a common edge line. The geometry of the fibre restricts the bound rays to a small section of cone.
In reality, the critical angle θc , and hence the angle of the magenta cone, is larger than depicted in the figure.
Appendix 3: Equations for light propagation in an optical fibre 118
The rays that propagate through an optical fibre are called bound rays. Other
rays are leaky rays. When a ray hits a point on the boundary core-cladding it will be a bound
ray if it comes within a red semi-cone depicted in figure A3.2. Rays in the red cone satisfy the
condition
0 ≤ α < π − θ c (A3.21)
2
where θc is the critical angle for the core-cladding interface, that is, θ c = sin-1(nco / ncl). A
refracting ray in the magenta cone in figure A3.2 is defined within the range
0 ≤ θ < θc . (A3.22)
All other rays are called tunnelling rays and they are defined within the ranges
π − θ ≤ α ≤ π
2 c and θc ≤ θ ≤ π (A3.23)
2 2
Rays in an optical fibre can also be divided into meridional and skew rays. The
meridional rays pass through the axis of the fibre while the skew rays do not pass through the
axis. Refracting and bound rays can be either meridional or skew rays, while tunnelling rays
are always skew rays.
When light is launched into a fibre all three types of ray (bound rays, tunnelling
rays and refracting rays) are present. Refracted rays irradiate their energy into the cladding and
the space around the fibre. According to Fresnel's equations, only a small fraction of a
refracted ray is reflected back into the fibre. The power carried by refracted rays can be
ρ2 2 πρ 2
N ≈ k 1 − k 22 = 2 n 2co − n 2cl (A3.24)
2 λ
If incoherent light is launched into a multimode fibre, all modes will be excited
and will carry equal amounts of power. In this case the power that travels in the cladding is
given by (Gloge, 1971)
Appendix 4
When light arrives at the interface of two materials, it is reflected and refracted
according to Snell's law, as shown in figure A4.1.
n2 θ2
x
n1
θ θr
1
we get:
2π n1 2 2
E 2 = E 0exp − i x n 1 sin θ 1 ± z 1 − n 2 sin θ 1 (A4.3)
λ0
Since for total reflection the term in the square root is negative, equation A4.3 needs to be
written as
n 2
E 2 = E 0exp − i 2π x n 1 sin θ 1 ± iz n 1 sin 2θ 1 − 1 (A4.4)
λ0
2
or
n 2
E 2 = E 0exp − i x 2π n 1 sin θ 1 exp ± z 2π n 1 sin 2θ 1 − 1 (A4.5)
λ0 λ0 2
The solution with negative sign in the second exponential term is the real
solution. It shows that the electric field decays exponentially in the z direction; that is, the
direction perpendicular to the interface. This field is called the evanescent field. The first
exponential term shows that this field travels along the x-axis in the same direction and with
the same phase as the wavefront in the waveguide. The depth of the evanescent-field
penetration is defined as the distance over which the amplitude of the field falls to 1/e of its
value at the interface, and is given by
λ0
dp = . (A4.6)
2
sin θ 1
2π −1
sin θ c
122
References:
Anderson G. P., Golden J. P. and Ligler F. S. "An Evanescent Wave Biosensor - Part I:
Fluorescent Signal Acquisition from Step-Etched Fiber Optic Probes" IEEE
Transactions on Biomedical Engineering vol. 41 (1994) pp. 578-584
Belmadani B., Casanovas J. and Casanovas A. M. "SF6 Decomposition under Power Arcs -
Chemical Aspects" IEEE Transactions on Electrical Insulation vol. 26
(1991A) pp. 1177-1182
Belmadani B., Casanovas J., Casanovas A. M., Grob R. and Mathieu J. "SF6 Decomposition
under Power Arcs - Physical Aspects" IEEE Transactions on Electrical
Insulation vol. 26 (1991B) pp. 1163-1176
Bobin B., Bordé C. J., Bordé J. and Bréant C. "Vibration-rotation molecular constants for the
ground and (ν3=1) states of 32SF6 from saturated absorption spectroscopy"
Journal of Molecular Spectroscopy vol. 121 (1987) pp. 91-127
Born Max and Wolf Emil Principles of Optics, Electromagnetic Theory of Propagation,
Interference and Diffraction of Light sixth edition, Pergamon Press 1980
Bucaro J. A., Dardy H. D. and Carome E. "Fiber optic hydrophone " Journal of the
Acoustical Society of America vol. 62, (1977) pp. 1302-1304
Busse L. E., Moon J. A., Sanghera J. S. and Aggarwal I. D. "Chalcogenide fibers deliver high
IR power" Laser Focus World vol. 32 (1996) pp.143-150
Butler M. A. "Optical fiber hydrogen sensor" Applied Physics Letters vol. 45 (1984)
pp. 1007-1009
Casanovas A. M., Casanovas J., Dubroca V., Lagarde F. and Belarbi A. "Optical detection of
corona discharges in SF6 , CF4 , and SO2 under dc and 50-Hz voltages" Journal
of Applied Physics vol. 70 (1991) pp. 1220-1226
Chadwick B. and Gal M "An Optical Temperature Sensor Using Surface Plasmons" Jpn. J.
Appl. Phys. vol. 32 (1993) pp.2716-2717
Chapados Camille and Birnbaum George "Infrared Absorption of SF6 from 32 to 3000 cm -1 in
the Gaseous and Liquid States" Journal of Molecular Spectroscopy vol. 132
(1988) pp. 323-351
Clegg H. P., Rowlinson J. S. and Sutton J. R. "The Physical Properties of Some Fluorine
Compounds and Their Solutions, Part 1. - Sulphur Hexafluoride" Transactions
of the Faraday Society vol. 51 (1955) pp. 1327-1333
Cole J. H., Johnson R. L. and Bhuta P. B. "Fiber optic detection of sound" Journal of
the Acoustical Society of America vol. 62 (1977) pp. 1136-1138
Collin Robert E. Field Theory of Guided Waves The Institute of Electrical and Electronics
Engineers, Inc. (1991)
Culshaw Brian and Dakin John Optical Fiber Sensors, Volume two: Systems and
Applications Artech House (1989)
Dakin John and Culshaw Brian Optical Fiber Sensors, Volume one: Principles and
Components Artech House (1988)
Eyal O., Shalem S. and Katzir A. "Silver halide midinfrared optical fibre Y coupler"
Optics Letters vol. 19 (1994) pp. 1843-1845
References 124
Farenc J., Mangeret R., Boulanger A., Destruel P. and Lescure M. "A fluorescent plastic
optical fiber sensor for the detection of corona discharges in high voltage
electrical equipment" Review of Scientific Instruments vol. 65 (1994)
pp.155-160
Francis Gordon "The Glow Discharge at Low Pressure" from Handbuch der Physik
vol. 22-Gasentladungen II, edited by Flügge S., Springer-Verlag (1956)
Frost L. S. and Liebermann R. W. "Composition and Transport Properties of SF6 and Their
Use in a Simplified Enthalpy Flow Arc Model" Proceedings of the IEEE
vol. 59 (1971) pp. 474-485
Gänger B. and Vigreux J. "May the GIS SF6 insulation be considered as self-restoring"
Electra No. 75 (1981) pp. 31-39
Gerrard A. and Burch J. M. Introduction to Matrix Methods in Optics John Wiley & Sons
(1975)
Gloge D. "Weakly Guiding Fibers" Applied Optics vol. 10 (1971) pp. 2252-2258
Grattan K. T. V. "Fibre Optic Sensors for Environmental Monitoring: can they meet the task?"
Environmental Monitoring editor Cashel E. M., IOP Publishing, Bristol, UK
(1990)
Hatori Shoichi and Shioda Susumu "Constriction of a Positive Column in an Argon Glow
Discharge" Journal of the Physical Society of Japan vol. 40 (1976)
pp. 1449-1455
Henry W. "Evanescent field devices: a comparison between tapered optical fibres and
polished or D-fibres" Optical and Quantum Electronics vol. 26 (1994)
pp. S261-S272
Homola J. and Slavik R. "Fibre-optic sensor based on surface plasmon resonance" Electronics
Letters vol. 32 (1996) pp. 480-482
Jorgenson R. C. and Yee S. S. "Control of the Dynamic Range and Sensitivity of a Surface
Plasmon Resonance Based Fiber Optic Sensor"
Sensors & Actuators A vol. 43 (1994) pp. 44-48
Kaye G.W.C. and Laby T.H. Tables of Physical and Chemical Constants, 14th Edition
Longman (1973) p. 63
Kerl K. and Jescheck M. "Equipment for precise measurement of the complex refractive index
of gases as a function of wavenumber, temperature and pressure"
Journal of Physics. E. Scientific Instruments vol. 15 (1982) pp. 955-960
Kildal Helge and Deutsch Thomas F., "Infrared Third - Harmonic Generation in Molecular
Gases", IEEE Journal of Quantum Electronics vol. QE-12 (1976)
pp. 429-435
Kogelnik H. and Weber H. P. "Rays, stored energy, and power flow in dielectric waveguides"
Journal of the Optical Society of America vol. 64 (1974) pp.174-185
Kunz U., Katerkamp A., Renneberg R, Spener F. and Cammann K. "Sensing fatty acid
binding protein with planar and fibre-optical surface plasmon resonance
spectroscopy devices" Sensors and Actuators B vol. 32 (1996) pp. 149-155
References 126
Lorrain Paul and Corson Dale R. Electromagnetic Fields and Waves Freemann (1970)
Ludman J. E., Ludman J. J., Callahan H., Caulfield H. J., Watt D., Sampson J. L.,
Robinson J., Davis S. and Hunt A. "Interferometric atmospheric
refractive-index environmental monitor" Applied Optics vol. 34 (1995)
pp. 3267-3273
MacCraith B. D., McDonagh C. M., O'Keeffe G., Keyes E. T., Vos J. G. O'Kelly B. and
McGilp J. F. "Fibre Optic Oxygen Sensor Based on Fluorescence Quenching
of Evanescent-wave Excited Ruthenium Complexes in Sol-Gel Derived Porous
Coatings" Analyst vol. 118 (1993) pp. 385-388
MacCraith B. D., O'Keeffe G., McDonagh C. and McEvoy A. K. "LED-based fibre optic
oxygen sensor using sol-gel coating" Electronics Letters vol. 30 (1994)
pp. 888-889
Maki A. G., Chou Che-Chung, Evenson K. M., Zink L. R. and Shy Jow-Tsong "Improved
Molecular Constants and Frequencies for the CO2 Laser from New
High-J Regular and Hot-Band Frequency Measurements", Journal of
Molecular Spectroscopy, vol. 167 (1994) pp. 211-224
Meek J. M.and Craggs J. D. editors Electrical Breakdown in Gases John Wiley & Sons
(1978)
Messica A., Greenstein A., Katzir A., Schiessl U.and Tacke M. "Fiber-optic evanescent wave
sensor for gas detection" Optics Letters vol. 19 (1994) pp. 1167-1169
Muhammad F. A. and Stewart G. "D-shaped Optical Fibre Design for Methane Gas Sensing"
Electronics Letters vol. 28 (1992) pp. 1205-1206
Ogle D. B. and Woolsey G. A. "Diffuse and constricted glow discharges in SF6" Journal of
Physics D: Applied Physics vol. 20 (1987) pp. 453-461
Ophel T. R., Weisser D. C., Cooper A. Fifield L. K. and Putt G. D., "Aspects of Breakdown
Product Contamination of Sulphur Hexafluoride in Electrostatic
Accelerators" Nuclear Instruments and Methods vol. 217 (1983) pp. 383-396
Park J. S., Yun S. H., Ahn S. J. and Kim B. Y. "Polarization- and frequency-stable fiber laser
for magnetic-field sensing" Optics Letters vol. 21 (July 15, 1996)
pp. 1029-1031
References 128
Pask C. "On the Neglect of Refracting Rays in Optical Fibre Power Calculations" Optical and
Quantum Electronics, Short Communication vol. 7 (1975) pp. 428-429
Rabinowitz Paul, Keller R. and LaTourrette J. T. " 'Lamb Dip' Spectroscopy Applied to SF6"
Applied Physics Letters vol. 14 (1969) pp. 376-378
Richards J. D., Garabedian R., Gonzalez C., Knoesen A., Smith R. L., Spenser R. and
Collins S. D. "Surface-plasmon excitation using a polarization-preserving
optical fiber and an index-matching fluid optical cell" Applied Optics vol. 32
(1993) pp. 2901-2906
Sah Chih-Tang Fundamentals of solid state electronics World Scientific Publishing (1991)
Sáinz C., Jourdain P., Escalona R. and Calatroni J. "Real time interferometric measurements
of dispersion curves" Optics Communications vol. 111 (1994) pp. 532-641
Sanghera J. S., Kung F. H., Pureza P.C., Nguyen V. Q., Miklos R. E. and Aggarwal I. D.
"Infrared evanescent-absorption spectroscopy with chalcogenide glass fibers"
Applied Optics vol. 33 No. 27 (20 September 1994) pp. 6315-6322
Schnell U., Dändliker R. and Gray S. "Dispersive white-light interferometry for absolute
distance measurement with dielectric multilayer systems on the target" Optics
Letters vol. 21 (1996) pp. 528-530
Schnitzer I., Katzir A., Schiessl U., Riedel W. J. and Tacke M. "Evanescent Field IR
Spectroscopy using Optical Fibres and Tunable Diode Laser" Material Science
and Engineering B vol. B5 (1990) pp. 333-337
Shimizu Fujio "Absorption of CO2 laser lines by SF6" Applied Physics Letters vol. 14 (1969)
pp. 378-380
Simhony S., Kosower E. M. and Katzir A. "Novel attenuated total internal reflectance
spectroscopic cell using infrared fibre for aqueous solutions" Applied Physics
Letters vol. 49 (1986) pp. 253-254
Snyder A. W. and Love J. D. Optical waveguide theory Chapman and Hall (1983)
Stewart G. and Culshaw B. "Optical waveguide modelling and design for evanescent field
chemical sensors" Optical and Quantum Electronics vol. 26 (1994)
pp. S249-S259
Sun H. C. and Wittaker E. A. "Real-time in situ detection of SF6 in a plasma reactor" Applied
Physics Letters vol. 63 (1993) pp. 1035-1037
Swalen J. D., Gordon J. G., Philpott M. R., Brillante A., Pockrand I. and Santo R. "Plasmon
surface polariton dispersion by direct optical observation" American Journal of
Physics vol. 48 (1980) pp. 669-672
Tai H., Tanaka H. and Yoshino T. "Fiber-optic evanescent-wave methane-gas sensor using
optical absorption for the 3.392-µm line of a He-Ne laser" Optics Letters
vol. 12 (1987) pp. 437-439
Tang P. J. P., Pullin M. J., Chung S. J., Yuen W. T., Phillips C. C., Stradling R. A.,
Norman A. G., Li Y. B. and Hart L. "Room temperature light emitting diodes for
the 3-12 µm IR band", (1995) Internal Report, Physics Department,
Imperial College, London
References 130
Thomas M. E. and Tayag T. J. "Refractive index of He, SF6 , and CO2 at 0.63299 µm as a
function of temperature and pressure" Applied Optics vol. 27 (1988)
pp. 3317-3318
Turk A., Edmonds S.M. and Mark H. L. "Sulfur Hexafluoride as a Gas Tracer"
Environmental Science and Technology vol. 2 (1968) pp. 44-48
Van Brunt R. J. "Production Rates for Oxyfluorides SOF2 , SO2F2 and SOF4 in SF6 Corona
Discharges" Journal of Research of the National Bureau of Standards vol. 90
(1985) pp. 229-253
Van Brunt R. J. and Leep D. "Characterization of point-plane corona pulses in SF6" Journal of
Applied Physics vol. 52 (1981) pp. 6588-6600
Wang D. N., Chen S., Grattan K. T. V. and Palmer A. W. "A low coherence 'white light'
interferometric sensor for eye length measurement" Review of Scientific
Instruments vol. 66 (1995B) pp. 5464-5468
Wang D. N., Ning Y. N., Palmer A. W., Grattan K. T. V. and Weir K. "An Alternative to
White Light Interferometric Sensing" Journal of Lightwave Technology
vol. 13 (1995A) pp. 961-966
Waters R. T. and Stark W. B. "Characteristics of the stabilized glow discharge in air" Journal
of Physics D: Applied Physics vol. 8 (1975) pp. 416-426
Weissler Gerhard L. "Positive and Negative Point-to-Plane Corona in Pure and Impure
Hydrogen, Nitrogen, and Argon" Physics Review vol. 63 (1943) pp. 96-107
Woolsey G. A. and Grey D. J. "Some Chemical Reactions in SF6 Glow Discharges" Proc. XIX
Int. Conf. Phen. Ioniz. Gases, Belgrade (Univ. Belgrade, ed. Labat J. M.)
vol. 3 (1989) pp. 534-535
Wronski Z., Sielanko J. and Sullivan J. L. "Studies of light emission from cathode material in
the plasma state of a glow discharge" Journal of Physics. D. Applied Physics
vol. 29 (1996) pp. 1509-1514
Yamauchi T., Onuma A. and Okuno M. "Operation Experience of 10000 Times in 15 Years
with the First GCB in Japan" Electric Energy Conference 1985, Newcastle,
Australia published by IEAust (1985) pp. 285-289
Zhujen Z. and Seitz W. R. "Optical Sensor for Oxygen Based on Immobilized Hemoglobin"
Analytical Chemistry vol. 58 (1986) pp. 220-222
References 132