Evaluation of arc quenching characteristics of various gases using power semiconductors

This paper presents the arc quenching abilities of various gases studied using a power semiconductor switching technique. The technique uses an insulated gate bi-polar transistor (IGBT) to inject current and apply voltage to the plasma arc. Using this technique, arcs in free recovery condition after a 50 A steady state condition were investigated in SF6, CO2, O2, N2, air and Ar gas flows. Furthermore, at a specific time high voltages of about 1.1 kV and 1.7 kV μs−1 were applied to the residual decaying arcs by the IGBT to elucidate the arc re-ignition process and recovery properties. These systematic experiments further enabled us to estimate the interruption probability versus the voltage application time. From these results, the voltage application time for 50% successful interruption was estimated for various gases, with the results showing a direct relation to the interruption capabilities of the respective gases. These results were then compared to the electron density measurement results and numerical simulation results to confirm their validity. All data obtained from the experiments and simulation is expected to be useful for elucidating the arc quenching physics and also for the practical application of arc quenching phenomena.


Introduction
A gas circuit breaker uses SF 6 gas as an arc quenching medium because it has quite a high arc-quenching and current interruption ability. However, it has also been specified as a greenhouse gas because it has a global warming potential (GWP) that is 22 800 times greater than that of CO 2 . Consequently, it is necessary to reduce the amounts of SF 6 released into the atmosphere by reducing the amount of it that is used. One solution for a reduction in the use of SF 6 is to develop alternative gases for SF 6 . Some alternative candidates have been investigated to date, such as high pressure CO 2 , CO 2 -CF 3 I, CO 2 -CH 4 and N 2 -H 2 [1][2][3][4]. Recently, a fluoronitrile-CO 2 mixture and a fluoroketone-CO 2 mixture were proposed as alternatives to SF 6 [5,6]. Other approaches to reducing the use of SF 6 include reducing the GCB size, with less deterioration of the interruption ability. To substitute SF 6 with alternative materials or to reduce the GCB size, a detailed understanding of the residual arc properties in the current interruption process is fundamentally important.
The authors have been investigating the decay process of gas-blast arcs, fundamentally using both numerical [7,8] and experimental approaches. For the experiment, we have developed a simple arc device. In the device, an arc plasma can be established between the moving electrode and the fixed electrode in the nozzle space. As a power source, a dc current was set up, with the option of switching the output current using a power semiconductor, e.g. an insulated gate bipolar transistor (IGBT). The IGBT can be used to control the arc current injection or voltage application accurately within a microsecond. By adopting time-accurate control of the current injection and voltage application, we measured the electron density in arcs around zero current and under free recovery conditions in a gas flow using laser Thomson scattering (LTS) [9,10] and Shack-Hartmann-type laser wavefront sensors [11]. In addition to these, we developed a new fundamental invest igation technique for arc behaviour during the re-ignition process [12][13][14]. This technique uses IGBTs to control not only the arc current but also the applied voltage with high accuracy in the time domain. Using this technique, transient voltage can be applied intentionally between the electrodes at a specified time under the free recovery condition. This applied voltage is called 'quasi-transient recovery voltage (quasi-TRV)' because it is generated artificially. By application of this quasi-TRV, we can study arc re-ignition processes and recovery properties in a thermal mode.
In this work, we systematically obtained the arc quenching properties of SF 6 , CO 2 , O 2 , N 2 , air and Ar using the developed technique [12]. A direct current 50 A arc was ignited in each gas flow at a given flow rate in the nozzle space and sustained in a steady state. Subsequently, the arc current was commutated by IGBT from the arc plasma to the IGBT to produce the arc under a free recovery condition. The arc decay process under a free recovery condition is the most basic property of arc-interruption phenomena. Arc behaviour in a steady state and in a decaying phase was observed using a high-speed video camera. Furthermore, we found the systematic data on the interruption probability of each gas to be a function of the timing of the quasi-TRV application through more than 10-30 shots, which can be regarded as the recovery properties in thermal mode. Electron density measurements were also taken at the nozzle throat using LTS to study the decaying arc processes in various gas flows. In addition, numerical simulation was undertaken to investigate the temperature decay process of the arc under a free recovery condition for various gases. All data obtained from the experiments and simulation will be useful in elucidating the arc quenching physics and also for the practical application of arc quenching phenomena. Figure 1 depicts the electrical circuit used for this work. The electric circuit has a direct current (dc) source, an IGBT element and an arc device. The dc current source is an inverter type rated at 150 A and 250 V. This dc current source is connected in series to the arc device to sustain an arc discharge between the electrodes in the device. An IGBT element (IGBT p ) is connected to the electrodes in parallel. Switching on this IGBT p can commutate the arc current from an arc discharge between the electrodes to the IGBT p . Doing this creates a free recovery condition between the electrodes, in which the arc discharge decays in time without either current injection or voltage application. In addition, at the specified time during the arc decay, switching off the IGBT p means the source voltage can again be applied to the decaying arc discharge between the electrodes. At this point, a steep voltage from the current source is applied respectively between the electrodes with a peak and a rise rate of 1.1 kV and 1.7 kV μs −1 , respectively. This applied voltage is called 'quasi-transient recovery voltage (quasi-TRV)' because it is similar to regular TRV, but it is artificially generated. One feature of the use of IGBT elements is that it can provide high time accuracy to create current commutation for a free recovery condition and the application of quasi-TRV. The IGBT element has a short response time and a jitter of less than 1.0 μs. Therefore, we can apply the quasi-TRV to the decaying arc discharge at the specified delay time of t d from the initiation of the arc decay.

Electric circuit for current injection and voltage application
When the quasi-TRV is applied during the arc decay, the arc plasma is re-ignited in some conditions. This arc re-ignition is regarded as a current interruption failure. If the arc plasma continues to decay in spite of the quasi-TRV application and the current converges to zero, then this situation is regarded as a successful interruption. The probability of the successful interruption depends on the delay time t d for the quasi-TRV application. In addition, this successful interruption probability versus t d is strongly dependent on the gas species, the gas flow rate, and other factors that determine the arc quenching ability. The voltage between the electrodes was measured using a CR potential divider, and the source current, the arc current and the IGBT current were measured respectively using current transformers. Figure 2 depicts the configuration of the arc device used in this work. The arc device consists of a vacuum chamber, a moving electrode and a fixed electrode, as well as a nozzle for gas flow. The outer diameter of the chamber is 200 mm and the upper electrode, which works as the anode, is a moving electrode driven by a compressed air cylinder. The lower electrode works as a cathode and is the fixed electrode; these electrodes are surrounded by a gas-blast nozzle. Figure 3 presents a schematic diagram of the nozzle and the electrodes as well as a photograph of the nozzle made of transparent polymethylmethacrylate (PMMA). The electrode tips are made of copper tungsten (70%Cu-30%W). The moving electrode has a 6 mm body diameter and 3 mm tip diameter. The body diameter and the tip diameter of the fixed electrodes are, respectively, 30 mm and 10 mm. The gap separating these electrodes measures 50 mm at the full open position, and the electrodes are surrounded by a nozzle. The gas inlet diameter is 40 mm, whereas that of the outlet is 18.75 mm. The nozzle has a throat with a 10 mm diameter and a 10 mm length, and was made of polytetrafluoroethylene (PTFE) for the arc tests. For observation of the arc behaviour in the nozzle only, a nozzle made of transparent PMMA was used.

Experimental arrangements for the laser Thomson scattering method
Laser Thomson scattering (LTS) is widely used for the precise measurements of electron density in arc plasmas. When a laser light is injected into an arc plasma, free electrons in the arc plasma are excited by the electromagnetic field of laser light, resulting in light scattering known as Thomson scattering. One of the characteristics of LTS measurements is that it can measure the local values of the electron density and electron temperature without assuming the local thermodynamic equilibrium (LTE) [15]. Details for LTS measurements have already been described for this system in our previous literature [9].
The electron density in the arc near the nozzle throat is a key physical parameter for ascertaining arc interruption phenomena. The electron density at the nozzle throat was measured using the LTS method. One characteristic of the LTS measurement is that it can measure local values of electron density and electron temperature without assuming the local thermodynamic equilibrium (LTE) [15]. For the LTS measurement, the second harmonic of a Nd:YAG laser at a wavelength of 532 nm was used. The pulse width of the laser beam was 10 ns and the laser energy was 200 mJ. Figure 4 shows an image of the PTFE nozzle for the LTS measurement and schematic cross sections of the nozzle. As this figure shows, the PTFE nozzle has a hole of 3.5 mm × 3 mm and a hole of 10 mm × 3 mm for the LTS measurement. These holes are positioned at the nozzle throat, which is at a height of 28 mm from the lower electrode surface. These two holes are arranged in a horizontal direction and are mutually perpendicular. The smaller hole is for the incident laser beam path, and the larger one is for the observation of 90 scattered laser light. The laser beam has a focused spot diameter of less than 0.5 mm. However, the laser beam size is spread intentionally in the horizontal direction using a cylindrical lens with a 4.5 mm width. A horizontally spread laser beam is used because it can be irradiated towards the arc column, even if the arc axis is shifted from the nozzle axis. The arc plasmas in SF 6 or CO 2 flow are well known to be unstable and to fluctuate. The larger hole was covered with glass plates to prevent hot gas ejection from the holes and covering the hole with glass plates reduced the influence of the hole on arc behaviour.

Experimental condition and procedure
We used gases SF 6 , CO 2 , O 2 , N 2 , air and Ar with a 100 l min −1 flow rate which corresponds to a gas flow velocity of 1.768 m s −1 at the nozzle inlet, and SF 6 and CO 2 with a 50 l min −1 flow rate which corresponds to a gas flow velocity of 0.884 m s −1 at the nozzle inlet. The electric arc current was fixed at dc 50 A. The pressure inside the chamber and the nozzle at the initiation of arc discharge was set to 0.1 MPa. The arc behaviour was observed using a high-speed video camera with the frame rate set to 300 000 frames s −1 and an exposure time of 10 μs. The PMMA nozzle was only used for high-speed video camera observation. In other experiments, a PTFE nozzle with holes was used for the LTS measurement and also for evaluation of the arc quenching ability of the gases. We measured the output current of the dc current source (i source ), the current between the electrodes (i arc ), the current in IGBT p (i IGBT ), the voltage between the electrodes (V arc ), the IGBT p gate voltage and the electrode driving voltage.
We measured the arc current, the IGBT current, the source current, and the voltage between the electrodes. Figure 5 presents an example of the measured current and voltage waveforms through one experiment. The experimental procedure is explained below.   In this time region, the arc is observed using a high-speed video camera. (vii) At t = 0 s, the IGBT p is switched-on to commutate the source current from the arc plasma to the IGBT p . As a result, the arc current and arc voltage smoothly decrease respectively to 0 A and 0 V. leading to the so-called free recovery condition. (viii) After a specified delay time t d from the IGBT p being switched on, the IGBT p is switched off again. The switching-off of the IGBT p applies a quasi-TRV to the electrodes. If the arc current increases between the electrodes after quasi-TRV application, then we judge the arc re-ignition and the current interruption failure. If the arc current continues to decrease to 0 A, then we judge the successful current interruption. In this work, t d can be set from 10 μs to 500 μs. (ix) After 1.0 ms from the initiation of quasi-TRV application, the IGBT p is switched on again to commutate the current from the arc plasma to the IGBT. (x) The current source is turned off to halt the experiment.
This experimental procedure provides (I) a gas blast arc sustained in a steady state, (II) a free recovery condition generated using the IGBT control, (III) the application of quasi-TRV to the electrodes, and (IV) judgment between the success or failure of current interruption.

Application of quasi-transient recovery voltage to gas-blast arcs under a free recovery condition
Figures 6 and 7 respectively present a conceptual diagram of the source current, the arc current and the voltage between the electrodes, in the cases of successful interruption and arc reignition. These figures only include the waveforms around the current commutation and quasi-TRV application. Successful interruption is recognized when the electric current continues to decrease, even after quasi-TRV application. However, if the current increases through the residual arc after quasi-TRV application, it is recognized to be an interruption failure or arc re-ignition. From the waveforms of the current and voltage between the electrode, one can judge successful interruption and interruption failure clearly. Furthermore, changing the delay time t d enables us to investigate the dielectric recovery property between the electrodes. At the same time, arc behaviour was observed using a high-speed video camera to study the arc decay and the position of the electrical breakdown between the electrodes.   It is necessary to fix the quasi-TRV waveform for comparison. Figure 8 presents the prospective waveform of quasi-TRV. The vertical axis shows the voltage between the electrodes. The horizontal axis presents the time from the initiation of a quasi-TRV application. The prospective quasi-TRV is an oscillating voltage governed by inductance in the dc current source and a strayed impedance in the circuit. The first peak of the quasi-TRV reaches 1.1 kV. Its rise rate is 1.7 kV μs −1 . In this work, the quasi-TRV is applied to an arc plasma under a free recovery condition.

Arc shape and arc voltage in a steady state before the free recovery condition
Before describing arc decaying processes in free recovery conditions, we show the arc behaviour in a steady state condition before the free recovery condition, because arc stability under a steady state condition with a gas flow is fundamental and is then also related to arc behaviour under a transient condition. Figure 9(a) depicts arc behaviour in a steady state in SF 6 with an inlet flow velocity of 1.768 m s −1 (= a gas flow rate of 100 l min −1 ) and 0.884 m s −1 (=50 l min −1 ). The shape of the nozzle and the electrodes is drawn with white lines, and the gas inlet is located at the bottom of the figure. Each image is captured by a high-speed video camera with a sensitivity for visible light in the wavelength range from 400 to 800 nm. Originally, the radiation of visible light from the arc is expressed as a monochrome image signal with a depth of 8 bits = 256 levels. We have converted this monochrome image signal to a coloured image signal according to the magnitude of radiation intensity. Here, the red colour corresponds to high intensity and the blue indicates low intensity. The images shown here are colour maps on a logarithmic scale, and include the timings of the photographs at the bottom of the figure. The time base is the equivalent to that of figure 5. As the pictures show, the SF 6 arc has a frequently changing shape in the nozzle space, even with spiral shapes for 1.768 m s −1 and 0.884 m s −1 . The shape instability of the arc might have originated from the turbulent flow because of the characteristics of SF 6 . Generally, SF 6 has a heavy mass density and a high Reynolds number (Re), which can be written as where ρ is the mass density of the gas, u is the velocity of the gas flow, μ is the viscosity of the gas, and L is a characteristic length, like the size of the gas flow inlet. The higher Reynolds number of the SF 6 gas flow with a higher mass density is thus subject to the production of turbulent flows. Furthermore, SF 6 can be dissociated thermally to SF 4 , SF 2 , SF, respectively, at 1800 K, 2000 K and 2200 K, involving local expansion and also causing turbulent flow. These dissociations also result in the high specific heat and thermal conductivity of SF 6 , leading to a high arc quenching ability, which can also make the arc discharge unstable and fluctuate. The arc current and voltage in the steady state are presented in figure 10(a) for a SF 6 gas inlet flow velocity of 1.768 m s −1 and 0.884 m s −1 . The upper panel presents the arc current, the lower panel presents the arc voltage. As shown, the arc current was almost constant around dc 50 A. The arc voltages fluctuate considerably, however, especially for the 1.768 m s −1 inlet flow velocity. The arc voltage fluctuation is attributable to the arc length fluctuation by turbulent flow. In spite of this, we can sustain these SF 6 arcs at a 50 mm length. The arc behaviour in CO 2 gas flow was also measured as depicted in figure 9(b) for an inlet flow velocity of 1.768 m s −1 and 0.884 m s −1 . The CO 2 arc is also bent with 1.768 m s −1 inlet flow, which is not that much compared to the SF 6 arcs. However, the CO 2 arc with a 0.884 m s −1 inlet has a straight stable shape. This stability can be confirmed in the measured arc voltage, as presented in figure 10(b). As the figure shows, the arc voltage for 0.884 m s −1 inlet flow velocity is absolutely stable around 120 V, whereas the arc voltage with 1.768 m s −1 inlet flow velocity shows random fluctuation. CO 2 has a relatively higher density than air and N 2 , but it is quite low compared to that of SF 6 . Only the higher inlet flow velocity can cause the appearance of turbulent effects for CO 2 arcs.
Arc behaviour in a steady state is presented in figure 11 in O 2 , N 2 , air and Ar with 1.768 m s −1 inlet flow velocity. As shown in this figure, arcs in these gas flows have quite a straight stable shape. Moreover, it is well controlled with almost no fluctuation. The main difference between these gas arcs is the diameter of the arcs. The Ar arc has the largest diameter among the gases, the second largest arc is the O 2 arc, while the N 2 and air arcs seem to have similar diameters. The Ar arc has the largest diameter because Ar is a noble gas without dissociations leading to energy consumption and then the shrinkage of the arc plasma. Figure 12    As the pictures show, the SF 6 arc discharge decays with time, keeping its shape at time 0 μs. The radiation intensity from the SF 6 arc decreases almost uniformly along it, roughly speaking, although some parts, between the lower electrode and the nozzle throat inlet for example, decline more rapidly. As a result, at t = 20 μs, the radiation from the arc becomes very weak, remaining downstream of the nozzle throat in this case. Furthermore, at t = 30 μs, the arc plasma has weak radiation in the nozzle space. If we use SF 6 with the inlet flow velocity of 0.884 m s −1 , then the arc behaviour is distinctive, as depicted in figure 13(b). In this case, the arc plasma also decays over the entire part of the nozzle space with a slightly lower decay rate than that at 1.768 m s −1 . However, one distinct feature is the local rapid decay in the arc plasma along the arc. In spite of this, rapid decay in the radiation intensity between the lower electrode and the nozzle throat is obtainable. Such a reduction and local rapid decay along the arc plasma is attributed to the high arc quenching ability of SF 6 and the turbulent effects.
In the case of CO 2 , local decay from the gas flow around the nozzle throat is apparent. Figure 14 presents the arc behaviour in a decaying phase in CO 2 with an inlet flow velocity of 1.768 m s −1 and 0.884 m s −1 . For both cases, the CO 2 arc decay was initiated near the nozzle throat. Fast decay around the nozzle throat is apparent, especially in the case of 1.768 m s −1 . At t = 6.7 μs, the radiation intensity around the upper side of the nozzle throat has already become much weaker than the other parts. At t = 30 μs, the intensity at the lower side of the nozzle throat also decreases with time. As this figure shows, the CO 2 gas flow enhances the arc plasma decay around the nozzle throat, which implies its importance for CO 2 arc decay. This tendency is also apparent in the case of 0.884 m s −1 inlet flow velocity, but the decay rate at 0.884 m s −1 is much lower than that at 1.768 m s −1 .
The arc behaviour in a decaying phase in O 2 , N 2 , air and Ar is obtainable in figure 15 at an inlet flow velocity of 1.768 m s −1 . The arc decaying process markedly depends on the kind of gas. The O 2 arc still has a high radiation intensity at t = 10 μs, whereas the radiation intensity from the N 2 and air arcs decreases faster than that of the O 2 arcs. In the test, the nozzle burned with a flame upon the introduction of O 2 gas. It would thus be possible for the combustion of ablation vapour to restrain the decrease in radiation intensity from the O 2 arcs. Similarly, the radiation intensity from the air arc is comparably higher than that of the N 2 arc at 30 μs. For O 2 , N 2 and air arcs, the arcs decline remarkably around the nozzle throat. The Ar arc shows a gradual and slow decay in radiation intensity.
High-speed video camera observations confirm that the nozzle throat plays an important role in arc decay in most cases. At the nozzle, cold gas from the inlet is blown directly at the arc plasma, which results in a high convection loss reducing the arc diameter; high thermal conduction loss is then present. However, the effects on the decaying arcs in each gas mutually differ. Such behaviour is obtainable by numerical simulation, as described in a later section.

The probability of successful interruption versus quasi-TRV application
One important feature of our experimental tests is to estimate arc interruption ability by the application of quasi-TRV in fixed conditions. In free recovery conditions, quasi-TRV was applied by switching the IGBT off again at a specified delay time of t d . This test was applied to investigate the recovery properties of the space between the electrodes. Figure 16 presents the IGBT signal, the arc current and the arc voltage for a SF 6 arc with 1.768 m s −1 inlet flow velocity at t d = 20 μs. The arc current is commutated to the IGBT at t = 0 μs by switching it on. In this case, the arc current has a finite current decay rate di/dt, because it requires a finite time for the transition of IGBT from an off-state to an on-state as well as circuit inductance. At t d = 20 μs, the IGBT is switched off again to apply a source voltage of 1.1 kV with a rise rate of 1.7 kV μs −1 between the electrodes. As panel (a) shows, the arc current continues to decrease to 0 A with time, which is regarded as successful interruption. Panel (b) depicts the waveforms in the interruption failure case, although the same delay time t d = 20 μs for quasi-TRV application was used. In this figure, the arc current again increases to 50 A through arc re-ignition at an applied voltage of 0.75 kV. This interruption failure seems to be a thermal failure or a failure in the thermal mode, not in the dielectric mode, because the arc current increases    gradually with time for 8 μs. Generally, in a dielectric mode, the current jumps up with a voltage drop at the arc re-strike. In this way, successful interruption occurs with a certain probability. If the delay time t d for quasi-TRV application is long enough, then the probability of successful interruption reaches 100%. When t d is short, the arc is always easily reignited between the electrodes. Figure 17 presents the experimentally obtained probability of successful interruption in a thermal mode versus the delay time t d for quasi-TRV application. This figure includes the results for an inlet flow velocity of 1.768 m s −1 . Each of the probabilities was obtained using 10-40 shots for each condition. As this figure shows, SF 6 requires a delay time t d less than 28 μs for quasi-TRV application to prevent arc re-ignition in thermal mode. In other words, a shorter t d than 20 μs engenders arc re-ignition. In addition, an important matter for SF 6 is that the interruption probability increases a lot with t d from 20 to 30 μs. The point above implies an extremely high arc interruption ability in thermal mode and the remarkably rapid recovery of SF 6 against voltage application. In other words, a SF 6 residual arc in free recovery conditions can lose its electrical conductivity at a point or points along the arc discharge between the electrodes. As a result, the delay time t d50% = 28 μs obtains 50% successful interruption,  where the t d50% is defined as t d , with which the interruption probability becomes 50%.
For other gases, a much longer t d is necessary for quasi-TRV application without arc re-ignition, compared to t d for SF 6 . The CO 2 and O 2 shows similar results for the interruption probability versus t d . Therefore, these gases have similar recovery properties in thermal mode. For CO 2 and O 2 , the quantity t d50% can be estimated as 108 μs. Air and N 2 gases require a longer t d for successful interruption in thermal mode. The delay times for 50% successful interruption t d50% were estimated as 190 μs for air and as 230 μs for N 2 . It is reasonable for air to show intermediate results between N 2 and O 2 because air is a mixture of 78%N 2 and 22%O 2 . However, Ar has an extremely long t d50% of 480 μs, which indicates the very slow recovery of Ar for electrical conductivity. This result is consistent with the fact that Ar is used widely as working gas for plasma applications for stable plasma sustainment.
The successful interruption property is related to the recovery properties in thermal mode, which depend on the decay rate in the electrical conductivity. The decay rate in the electrical conductivity is directly connected to the decay in the electron density and the temperature. The temperature decay at the nozzle throat is again influenced by the gas flow convection loss with a high specific heat at low temperature ρC p u · ∇T and the thermal conduction ∇ · κ∇T, where ρ is the mass density, C p denotes the specific heat, u represents the gas flow velocity, κ is the thermal conductivity, and T is the temperature. The gas SF 6 has a higher specific heat ρC p and a thermal conductivity κ at low temperatures of about 1800 K, 2000 K and 2200 K, arising from dissociation-association reactions from SF 6 to SF 4 , SF 4 to SF 2 and SF 2 to S and F. This higher specific heat and thermal conductivity is favourable for the decay of the temperature and electrical conductivity.
Such a successful interruption probability versus t d can be adopted for different gas conditions such as gas flow rates. Figure 18 presents the probability of successful interruption versus t d for SF 6  Open squares with a cross correspond to SF 6 gas introduction at 0.884 m s −1 . Similarly, results for CO 2 gas introduction at 1.768 m s −1 are shown by open triangles. Triangles with a cross are for CO 2 gas introduction at 0.884 m s −1 . For SF 6 gas flow, the increased rate in the successful interruption probability versus t d is rather gradual at 0.884 m s −1 compared to 1.768 m s −1 . As a result, t d50% is estimated as 40 μs for SF 6 at 0.884 m s −1 , which is 1.4 times as long as t d50% at 1.768 m s −1 . This result suggests a quantitative reduction in the interruption ability in thermal mode by a decrease in the gas inlet flow velocity for SF 6 . The interruption ability of CO 2 was also degraded by a decrease in the gas inlet flow velocity. Actually, the quantity t d50% becomes 150 μs for CO 2 at 0.884 m s −1 , which is 1.4 times as long as t d50% at 1.768 m s −1 . This result might imply that a decrease in the gas inlet flow velocity from 1.768 to 0.884 m s −1 reduces the recovery rate similarly for SF 6 and CO 2 .
The successful interruption probability versus the delay time t d for voltage application was obtained systematically.
It is noted that the relationship between interruption probability versus t d in figures 17 and 18 is just a relative relationship of the conditions created by different kinds of gas and gas flow rates at a fixed TRV application. If we use a higher TRV application, the results of figures 17 and 18 will shift to the side of longer delay times. This means that a longer time is needed for the space between the electrodes to recover enough to withstand higher TRV application. Nevertheless, we could compare the relative relationship for different kinds of gas and different gas flow rates. SF 6 would only need a short delay time for current interruption compared to the other gases because of its rapid recovery property. This data includes useful information and contributes to the elucidation of arc interruption phenomena for fundamental and practical issues.

Waveforms of current and voltage on interruption failure
The experiment results revealed some interruption failure cases at the critical t d . Figure 19 presents the current and voltage waveforms of interruption failure for gas introduction at 1.768  All waveforms were acquired as a 'failure case' with quasi-TRV application at nearly t d50% , i.e. t d , which provides 50% probability of successful interruption for each gas. Figure 19(a) shows the current and voltage after quasi-TRV application at a t d of 30 μs for SF 6 for gas introduction at 1.768 m s −1 . In this case, the arc current started increasing from t = 2 μs. The voltage dropped, which means that the SF 6 arc was almost re-ignited at t = 2 μs. In this way, SF 6 arc re-ignition mostly occurs before the first peak of quasi-TRV, which might be attributable to the fact that the residual SF 6 arc channel remains at a high temperature and becomes conductive again because of joule heating by a small current injection. Otherwise, arc re-ignition is avoided by rapid quenching of SF 6 . Therefore, the thermal mode determines the success or failure of current interruption at t d = 30 μs for SF 6 .
Arc re-ignition occurs in CO 2 gas at 1.768 m s −1 inlet flow velocity, as presented in figure 19(b) for t d = 110 μs. In this case, the arc was re-ignited more than 5 μs after quasi-TRV application initiation. Many cases show such microsecond arc re-ignition for t d50% TRV application. In this figure, a current of less than 5 A can be confirmed to flow immediately before the rapid increase in current. This result suggests that the residual CO 2 arc becomes conductive gradually because of joule heating caused by a small current after quasi-TRV application in thermal mode. Figure 19(c) depicts the result for N 2 gas at 1.768 m s −1 and t d = 240 μs. In this case, the arc was re-ignited more than 10 μs after quasi-TRV application. Furthermore, the arc current increases and the voltage drops slowly after re-ignition because the electrical conductance change in the N 2 arc was very slow in this case. From the results described above, it was also inferred that SF 6 gas has extremely high arc-quenching performance and high post-arc current withstanding capability, and that CO 2 gas has medium quenching ability and post-arc current withstanding capability, compared to N 2 .

Electron density measured using laser Thomson scattering
Through the current interruption tests, the electron density of the arc in the decaying process was measured using the laser Thomson scattering (LTS) method [9,10]. This section presents a description of part of the measured electron density for the consideration of the current interruption capabilities of SF 6 , CO 2 and N 2 .  At t = 0 μs, the electron density n e is 6.0 × 10 22 m −3 for SF 6 and CO 2 , whereas n e is estimated as 10.0 × 10 22 m −3 . After the current decreases to 0 A at t >0 μs, the electron density decreases rapidly with time. In N 2 gas, the electron density decreases gradually with time to reach 0.3 × 10 22 m −3 . SF 6 and CO 6 cause a rapid decrease in the electron density with time. At t = 20 μs, the electron densities for SF 6 and CO 2 become 1.1 × 10 22 m −3 . Additionally, it is worth noting that the electron density in SF 6 could not be determined from LTS at t > 20 μs, possibly because it had decayed to the lower limitation of <10 21 m −3 detectable by LTS. This fact suggests the remarkable decay ability of SF 6 for an electron density of t >20 μs compared to CO 2 and N 2 .
The electron density decay results show good agreement with the results of the radiation intensity in arcs observed by high-speed video cameras. As shown in figure 13, at t = 0 μs, the shape of the SF 6 arc tends to be a spiral. After the current has gone down to 0 A at t >0 μs, the entire part of the residual SF 6 arc quickly decays simultaneously. Then, the overall decay with some local decay in the residual arc can also be obtained at t > 20 μs. The quick decay of the SF 6 arcs agrees with the rapid decay in the electron density shown in figure 20. However, the shape instability of the CO 2 arcs is not as high as that of the SF 6 arcs, as shown in figure 14. The CO 2 arc decays more rapidly from the vicinity of the nozzle throat, which is the LTS measurement point of the electron density. Figure 20 shows that the characteristics of this CO 2 arc might cause rapid decay in the electron density at the nozzle throat. Finally, the N 2 arc is extremely stable, and decays slowly with time, as depicted in figure 15(b). Therefore, the obtained electron density is greater and decays with time more slowly than either SF 6 or CO 2 , as presented in figure 20.
As described in this section, the results of the electron density measurement agree fairly well with the arc behaviour observed using a high-speed video camera. In addition, there is no contradiction with the results of the current interruption tests. Nevertheless, the results also show that current interruption is not just determined at one point of the nozzle throat, especially in SF 6 . Further experiments must be conducted to elucidate the arc interruption phenomena.

Transient temperature distribution of various gas arcs in nozzle by numerical simulation
In several gas conditions, the arcs decay from near the nozzle throat. To investigate the effects of the nozzle throat on arc decay, a numerical simulation of the arc was conducted in the flows of different gases [7]. Numerical simulation uses the LTE model, with the following assumptions: (1) The calcul ation domain is axisymmetric. (2) The arc plasma is in LTE conditions. All temperatures, such as that of the electron, heavy particle and excitation, are equal. In addition, all reactions including dissociation/association and ionization/recombination take place under equilibrium conditions. (3) The flow is laminar; therefore, the turbulent effect is neglected. (4) The arc plasma is optically thin. (5) Phenomena on the electrode surface, such as electron emission and ion bombardment, are neglected. (6) The electric field only has an axial direction component. (7) We neglect density fluctuations caused by pressure fluctuations in a steady state, but in a transient state it is included. (8) The ablation effect of electrodes and the nozzle are neglected. (9) We consider heat conduction inside the electrodes and the nozzle. Based on the assumptions presented above, we solved the conservation equations of mass, momentum and energy as well as electromagnetic fields, together with the equation of state. The thermodynamic and transport properties of each gas were calculated in advance as functions of temperature using the calculated equilibrium composition of each gas. The program is a hand-made code based on the all-speed SIMPLE algorithm reported by Patankar. Using this model, the two-dimensional transient temperature and gas flow fields were calculated for the SF 6 , CO 2 and N 2 arcs. Figure 21 presents the calculated two-dimensional temperature distributions of CO 2 , N 2 and SF 6 arcs in a free recovery condition after a dc 50 A steady state condition. The temperature is expressed by a logarithmic colour scale, and the configuration of the nozzle and electrodes is the same as that in the experiment, as presented in figure 3. Figure 21(a) depicts results for a CO 2 arc with gas introduction at 0.884 m s −1 . It is readily apparent that the arc temperature decreases with time. Furthermore, the decrease rate in the arc temperature is greater around the nozzle throat, especially just at the nozzle throat inlet. Figure 14(b) shows that this calculated result agrees well with the time evaluation in the radiation intensity of CO 2 arcs. The rapid decay in the temperature around the nozzle throat inlet is attributable to a high convection loss of ρC p u · ∇T there, because the gas flow velocity u is high and the radial temperature gradient ∇T is higher than at other positions. This high radial temperature gradient arises from the cooling gas flow supplied to the nozzle throat inlet. In addition, the calculated results for 1.768 m s −1 CO 2 gas introduction are found in figure 21(b). In this case, the arc temperature starts decreasing around the nozzle throat, exhibiting a more rapid decrease in the temperature at 1.768 m s −1 than at 0.884 m s −1 . It is further shown that the experimentally obtained results of the radiation intensity from a CO 2 arc with gas introduction at 1.768 m s −1 in figure 14(a) seems to decay more rapidly than those predicted by the present numerical simulation in figure 21(a). Perhaps this is true because turbulent flow effects are not considered in the calculation model. Figure 21(c) presents the calculated results for a N 2 arc with gas introduced at 1.768 m s −1 . The calculation model also predicts that a N 2 arc will decay from the nozzle throat inlet. This calculated transient temperature distribution closely resembles the experimentally observed result depicted in figure 15(b), from the viewpoint of the decaying spatial position of the arc. Therefore, the present LTE model can validly predict arc behaviour because N 2 arcs are quite stable during both steady state and decay without turbulent effects, etc. in the experiment. However, a difference exists between the calculated result and the observed result for SF 6 arcs. Figure 21(d) represents the results for a SF 6 arc with an inlet flow velocity of 1.768 m s −1 . The simulation predicts that the SF 6 arc will decay remarkably fast around the nozzle throat inlet because of the strong convection loss by cold and heavy SF 6 gas. This rapid decay in the SF 6 arc temperature is attributed to the high convection loss caused by high specific heat, and high thermal conduction loss. The high specific heat and high thermal conductivity of SF 6 are involved equivalently in the dissociation reactions of mainly SF 6 to SF 4 , SF 4 to SF 2 and SF 2 to SF. This simulated temperature distribution shows the reasonably fast decay of the SF 6 arc, compared with the observation result presented in figure 13(a). However, a difference remains between them. In the experiment, SF 6 actually produces strong turbulent flow, which makes the arc fluctuate. Furthermore, the actual SF 6 arc process is in nonequilibrium chemical conditions in which dissociation/ association reactions take finite time. Such turbulent effects and nonequilibrium effects should be considered in future work. In other words, the present comparison between the experimentally obtained results and the LTE simulated results are useful for ascertaining the importance of these effects.

The influence of holes in the nozzle for the LTS on arc behaviour
The LTS measurements were made using a nozzle with a laser path hole and an observation hole. The observation hole was covered by a glass plate to prevent hot gas ejection from the nozzle throat, as described in the previous section. However, the current interruption test was conducted using nozzles with and without the holes. It is therefore necessary to confirm the influence of the presence of the holes in the nozzle on the arc behaviour. Figure 22 shows the average arc voltage in a steady state in various gas conditions for the two types of nozzle. One bar stands for the results of the nozzle without holes, the other one shows the results for the nozzle with holes for the LTS measurement. The arc voltage measured from the time −5 ms to 0 ms was averaged for each gas condition. Each error bar is produced by at least six samples, except the case of O 2 without a hole. As this figure shows, the effect of the hole on arc behaviour was not significant for any of the gases examined in this study, and except for SF 6 , the arc voltage is slightly lower with holes present; consequently, the influence of holes has been neglected in the present work.

Conclusions
This paper presents the current interruption capabilities of various gases quantified using a newly developed method. The method uses an insulated gate bi-polar transistor (IGBT) as a current and voltage control device. SF 6 , CO 2 , O 2 , N 2 , air and Ar arcs in a dc current of 50 A were ignited in a gasflow nozzle, after which they decayed because of the current commutation to the IGBT connected in parallel to the arcing electrodes. After a specified delay time of t d from the initiation of the arc decay, a steep voltage was intentionally applied to the electrodes. The case in which the arc did not re-ignite is regarded as a successful interruption. The probability of successful interruption was investigated statistically for every gas. Consequently, SF 6 was estimated as needing 28 μs to have a 50% probability of successful interruption (t d50% = 28 μs) making it the fastest of the gases. Also, the t d50% of CO 2 was 108 μs and the t d50% of N 2 was 240 μs. In this way, the current interruption capabilities of various gases can be evaluated quantitatively using the developed method. Furthermore, arc behaviour in a steady state and decay phase was observed using a high-speed video camera. The electron density of the residual arc in the decay phase was measured using the LTS method. A comparison between the arc behaviour and the electron density showed good agreement. No contradiction was found with the results of the current interruption tests.
The results obtained in this work will support the consideration of the fundamental characteristics of arc discharge and the development of alternatives to SF 6 . Additional approaches must be made to ascertain the fundamental properties of cur rent interruption phenomena. We are planning further invest igation, such as the influence of TRV peak voltage and rise rate using advanced techniques with power semiconductors.