Experimental Investigation of the Breakdown Voltage of CO 2 , N 2, and SF 6 Gases, and CO 2 –SF 6 and N 2 –SF 6 Mixtures under Different Voltage Waveforms

: This paper is devoted to a comparison study of the breakdown voltage of CO 2 , N 2 , and SF 6 gases, and CO 2 –SF 6 and N 2 –SF 6 mixtures under different voltage waveforms, namely AC, DC, and lightning impulse voltages, in point–plane and sphere–sphere electrode arrangements. The inﬂuence of pressure, voltage polarity, and percentage of SF 6 in CO 2 and N 2 were studied, and equivalencies between the breakdown voltage of SF 6 and those of the considered mixtures were analyzed. It is shown that the breakdown voltage of SF 6 is the highest, whatever the applied voltage waveforms. Similarly, for a given voltage waveform, the breakdown voltage of SF 6 is the highest. The AC breakdown voltage is the lowest for all gases. The addition of small amounts of SF 6 to CO 2 and N 2 signiﬁcantly improved the breakdown voltages of both natural gases. For a given breakdown voltage, the ratio between the pressure of CO 2 to that of SF 6 was generally lower than the pressure of N 2 to SF 6 , whatever the voltage waveforms.


Introduction
SF 6 possesses excellent insulating properties, it can interrupt the current of electric arcs and interfere with their performance, has good heat transfer capacity, and chemical stability. These factors mean that sulfur hexafluoride gas is the most widely used gas in high-voltage (HV) components integrated in the transmission and distribution of electric power (HV circuit breakers (GCB), gas insulated switchgears (GIS), transmission lines (GIL), etc.) for more than half a century. SF 6 is one of the best insulating gases known to date. However, despite its outstanding properties, SF 6 gas is an aggravating agent of the greenhouse effect, and has a global warming potential (GWP) 23,900 times that of CO 2 . It is for this reason that since the Kyoto Protocol (COP3, 1997), SF 6 use and emission into the atmosphere started to be controlled, and intense research has been undertaken to find substitutes for SF 6 that have a lesser impact on the environment, as well as comparable dielectric and current interrupting capabilities. Unfortunately, all considered gases present a high liquefaction temperature [1][2][3][4][5], in addition to certain gases having excessive prices. To overcome this problem, and to decrease the liquefaction temperature, these gases are mixed with N 2 and CO 2 ; both of these natural gases are used as buffers. Thus, promising new mixtures of complex fluids using natural gases (N 2 , CO 2 , dry air) as buffers have been developed these last years, and some have been introduced in mean and high voltage apparatus [2][3][4][5][6][7].
Another solution would be to dilute small amounts of SF 6 in N 2 or CO 2 ; as these natural gases (N 2 and CO 2 ) are freely available in the atmosphere. Their impact on the environment is weak; their ozone depletion potential (ODP) is low compared to SF 6 . With a suitable choice of parameters such as SF 6 concentration, pressure, and electrode gap, it would be possible to get mixtures that had a reduced impact on the environment (i.e., a low greenhouse effect) and would present a good compromise between insulating properties and minimum operating temperature. Many investigations have been conducted on SF 6 -N 2 mixtures [7,8].
This paper is devoted to a comparison between the breakdown voltage (BDV) of CO 2 , N 2 , and SF 6 gases, and CO 2 -SF 6 and N 2 -SF 6 mixtures under three types of voltage waveforms, namely AC, DC, and lightning impulse voltages, in two different electrode geometries, namely point-plane and sphere-sphere electrode arrangements, at different pressures. The influence of pressure, voltage polarity, and percentage of SF 6 in CO 2 and N 2 are investigated, and equivalencies between the breakdown voltage of SF 6 and those of mixtures are discussed.

Experimental Section
The tested gases and gas mixtures were SF 6 , N 2 , and CO 2 , and SF 6 -N 2 and SF 6 -CO 2 , respectively. The breakdown voltage was measured for three voltage waveforms: AC, positive, DC negative, and positive lightning impulse voltages using two different electrode arrangements.
We first compared the three gases (CO 2 , N 2 , and SF 6 ) in point-plane electrode arrangement for three different electrode gaps (5, 10, and 20 mm). Then, we compared the breakdown voltages of those gases and their mixtures in a sphere-sphere electrode arrangement.
The point was made of tungsten and its radius of curvature was 10.0 ± 0.5 µm. This value may have varied slightly with the succession of tests due to tip erosion, the tip was replaced as soon as its profile became different from the original; its height was 18 mm. As for the plane electrode, it had the shape of a plane disk of which the edges were profiled according to Rogowski. The disc had a diameter of 49.5 mm, a thickness of 6 mm, and a periphery of radius curvature of 3 mm. The sphere electrode had the same height as the tip electrode and a radius of 5 mm. Note that the surface state of the electrodes were controlled, and that the electrodes were changed/replaced as soon as a defect was detected. In both configurations, the two opposite electrodes were axisymmetric and their axis was vertical. Figure 1 shows a diagram of the experimental setup, a general view of the test cell, and details on the dimensions of the electrodes.  Note that we limited our measurements to 0.4 MPa (4 bar). This pressure is generally the operating gas pressure in high voltage apparatus filled with pure SF 6 (gas insulated switchgear (GIS), for instance).

High Voltage Supplies
The AC high-voltage supply was a 200 kV, 50 Hz, 60 kVA transformer (Hipotronics type). A voltage regulator enabled the operator to increase voltage at a chosen rate. The maximum and effective values of the output voltages of the test transformer were measured using a 60 MHz Maxtron 60 MHz oscilloscope connected to a capacitive voltage divider 1:10,000 (BER 1423/1212-A314, 0.01 µF/100 pF-200 kV).
Two direct high-voltage supplies were used: A 200 kV, 400 W, 2 mA DC generator of positive polarity (Spellman type), and a 150 kV, 225 W, 1.5 mA DC generator (SAMES type) of negative polarity connected to a resistive divider.
The lightning impulse (LI) voltage was supplied either by a 200 kV-2 kJ Marx generator or a 1 MV-50 kJ Marx generator (both generators are Haefely type) each generator was connected to a dedicated capacitive divider; the ratio of dividers was 1:10,000 for the first generator and 1:11,000 for the second generator. The waveform of LI voltage used had a 1.2 µs ± 30% front time and a 50 µs ± 20% tail time. The lightning impulse voltage breakdown was measured using the up-and-down method according to International Electrotechnical Commission (IEC 60060-1:2010).

Filling the Test Cell and Breakdown Voltage Measurement Procedures
Before each series of breakdown tests, we emptied the test cell using a vacuum pump in order to attain a minimum pressure of approximately 10 mbar, which meant we could proceed to completely flush the test cell in accordance with the American Standards for Testing and Materials (ASTM-D-2477-2005); filling the test cell with the gas to be tested, followed by emptying of the test chamber. The gas was then introduced into the cell before applying the voltage. For a given electrode gap, the measurements were performed from high to low pressure. The pressure was decreased in steps of 0.5 bar. Regarding the realization of gas mixtures, we used Dalton method that takes into account the coefficient of compressibility of gases [9].
After a number of measurement series the cell was completely disassembled and cleaned, and the electrodes were treated (cleaning and sharpening/changing the tip if necessary). Throughout the study, a halogen gas detector allowed us to prevent and/or remedy any gas leaks. Control of the intake/emptying pressure was ensured by a system of taps, equipped with a pressure gauge, connected to the gas bottles using polyurethane pipes.
For the 50 Hz AC tests, the applied voltage was raised to breakdown at a rate of 0.5 kV/s ± 20%, according to ASTM-D-2477-2005. The time interval (waiting time) between two successive applied voltages was approximately 2 min. At least ten voltage applications were considered to estimate the mean breakdown value for each series of measurements. Each breakdown point plotted on a figure is the average of five sets of measurements; this was done to ensure their reproducibility.
Regarding the DC breakdown voltage measurements, the tests were carried out by progressively increasing the voltage stepwise until breakdown occurred. Each level of applied voltage was sustained for 1 min; each point plotted on a figure is the average of at least twenty measurements.
For the impulse voltage tests, we first performed preliminary tests to roughly determine the voltage at which breakdown occurs. We then started a series of measurements using the up-and-down method, according to the International Electrotechnical Commission (IEC 60060-1:2010). The intermediate stand time between a shot and the following shot was 90 s. The number of shot series we applied to establish the U 50% breakdown voltage was 40. This voltage is visualized by means of a display screen (Haefely Digital Impulse Analysis System (DIAS) 730) connected to the output of a capacitive divider. An oscilloscope was also used to record these breakdown voltages, and to check the concordance of the results with those of the display screen.

Influence of Voltage Waveform and Polarity on Breakdown Voltage of Gases in a Point-Plane Electrode Arrangement
Figures 2-5 depict the BDV voltages of SF 6 , N 2 , and CO 2 gases as a function of the gas pressure and the inter-electrode gap, respectively, for a point-plane electrode arrangement under different voltage waveforms. These characteristics show that the breakdown voltage increases with the pressure and/or the inter-electrode distance d. By drawing straight lines of tendency, we noted that some experimental data deviates from the linear tendency. Note that for comparison purposes, AC breakdown voltages in all figures are peak values.
Under AC, CO 2 BDV is clearly higher than that of N 2 . While the BDV of SF 6 is more than three times higher than that of both CO 2 and N 2 for the three investigated electrode gaps (5, 10, and 20 mm). In contrast, under a positive LI voltage, the BDV of N 2 was somewhat higher than that of CO 2 . However, as expected, the SF 6 breakdown voltage was always the highest when compared to that of N 2 and CO 2 , because of its well-documented superior insulation performance. Note that for an inter-electrode gap of 5 and 10 mm, the visible differences observed in N 2 at low pressures decreased when pressure was increased. If the slope of growth is maintained, the breakdown voltage at 10 mm, 4 bar would be approximately 20 kV, while on Figure 2 it is approximately 17 kV, representing a difference of 15%, which is not negligible. This could be due to the dispersion of measurements in N 2 . The dispersion of the measurements of breakdown voltages in N 2 has also been reported by others [10].
Note that a linear variation in the breakdown voltage of N 2 and CO 2 has been observed by E. Onal [11] using a rod-plane electrode system with an electrode gap of 25 mm and a pressure between 1 to 5 bar; the tip radius of the rod was 1 mm.
Under DC, the BDV of N 2 and CO 2 were higher at negative polarity for the three electrode gaps. Similar to the trend observed for SF 6 with an electrode gap of 10 mm, we first observed, at 0.5 bar, that the positive DC BDV was higher than the negative DC BDV, and then we observed the inverse: that the negative DC BDV became higher (Figure 6). Such a variation has been reported in previous work in SF 6 under LI voltages [12]. This inversion phenomenon has also been observed with a 3.7% Fluoronitriles-96.3% CO 2 mixture [4]. This is likely due to space charges.
Note that the point-plane electrode geometry is generally investigated to simulate defects or triple points in high-voltage equipment (GIS for instance).

Influence of Voltage Waveform on the Breakdown Voltage of Gases and Mixtures in Sphere-Sphere Electrode Arrangement
Generally, in order to investigate the dielectric strength of a given material (gas here) one uses an electrode geometry that provides a uniform or quasi-uniform electric field. In this study, we consider a sphere-sphere electrode arrangement whose axis is vertical. The spheres have a radius of 5 mm and the inter-electrode distance was set to 10 mm (Figure 7). This geometry provides an electrode utility factor η of 0.42 (η being the Schwaiger form factor [13]). Regarding the mixtures, the amounts of SF 6 diluted in N 2 or CO 2 were limited to 20%. Because of what we stated earlier, beyond this value mixtures with SF 6 become less interesting, both ecologically and economically. The tests were conducted according to standards IEC 60060-1:2010.

Comparison of Breakdown Voltages of N 2 , CO 2 , and SF 6
The BDV is clearly dependent on the voltage waveform whatever the considered gas, as shown in Figure 8. It is noted that the AC and DC breakdown voltages of N 2 and CO 2 vary quasi-linearly with pressure, which is not the case with LI BDV, where the variation is clearly non-linear. On the other hand, the DC and LI BDVs were always higher than the AC BDVs. The LI BDVs of CO 2 and N 2 were at first higher than the negative DC BDVs, however, following an inversion, they were lower. This inversion is sharper in N 2 ; occurring between 1.5 and 2.0 bar, and in CO 2 between 2.0 and 2.5 bar. Beyond 2.0 bar, the LI and negative DC voltages were very similar for CO 2 . While in N 2 , DC BDV remained higher than LI BDV. In SF 6 , LI BDV was always the highest. This result makes it possible to envisage the replacement of SF 6 by nitrogen, especially in medium HV equipment, at a moderate pressure.  Figure 9 shows a comparison of the BDVs of each of the three investigated gases for a given voltage waveform. It is noted that the BDV of SF 6 is always higher than that of N 2 and CO 2 , whatever the voltage type. The lowest breakdown voltage of the three gases was observed when using AC. Under AC, the BDV of CO 2 was lower than that of N 2 , while under DC and LI voltages, the BDVs of both gases were very similar.  The addition of SF 6 to CO 2 and N 2 improves the breakdown voltages of these natural gases. This improvement depends both on the concentration of SF 6 and the type of voltage waveforms. The comparison between the three gases (N 2 , CO 2 , and SF 6 ) and their mixtures (N 2 -SF 6 and CO 2 -SF 6 ) can either be carried out at a given common breakdown voltage (U BDV ) of the three waveform voltages, or at a fixed common pressure. In the following, we compared the BDVs of mixtures at a given pressure, namely 2 bar, with the BDVs of N 2 , CO 2 , and SF 6 for the three voltage waveforms at this pressure (Table 1). Figure 10 depicts the variation in BDV versus the percentage of SF 6 for the three types of voltages at 2 bar. We observed that the AC BDV of CO 2 increased by 15% and 25% at 5% and 10% SF 6 , respectively. The AC BDV of N 2 increased by 12% and 18% at 5% and 10% SF 6 , respectively. For DC, the BDV of CO 2 increased by 20% and 29% at 5% and 10% SF 6 , respectively. While for N 2 , BDV increased by 18% and 22% at 5% and 10% SF 6 , respectively.  The LI BDVs of CO 2 and N 2 increased by 25% and 44%, respectively, at 5% SF 6 . While the LI BDVs of CO 2 and N 2 increased by 20% and 46%, respectively, at 10% SF 6 . The fact that the BDV of a 95% N 2 5% SF 6 mixture is higher than that of 90% N 2 -10% SF 6 mixture seems illogical at first glance. This is also the case with N 2 , where the LI BDV at 1.5 bar is higher than at 2 bar.
Note that the LI BDV of SF 6 at 2 bar (approximately 126 kV) is obtained with an 80% N 2 -20% SF 6 mixture.
In the range of investigated SF 6 concentrations, the DC BDVs of N 2 -SF 6 mixtures increase with the percentage of SF 6 and/or pressures, as depicted in Figure 11. We noticed an equivalency between pure SF 6 and 82% N 2 -18% SF 6 at 3 bar; the BDV was approximately 125 kV. We observed a linear tendency when increasing the SF 6 concentration; this is not the case under LI. Indeed, beyond 5% SF 6 , we even noted a decrease in BDV, evidencing a non-linearity, as shown in Figure 10.
This non-linearity could result from space charge and mechanisms involved in discharge phenomena. Juhre and Kynast [14] reported a similar observation when analyzing the BDV of SF 6 -CF 4 in a non-homogeneous field. They attributed this phenomenon to the stabilization mechanism of corona discharges. Onal [10] also reported this non-linearity of SF 6 -N 2 and SF 6 -CO 2 mixtures in the rod-to-plane electrode system. According to this author, the non-linearity is due to the charge density and development of streamers. The use of natural gases can be considered, particularly in medium voltage apparatus, when pressures remain moderate and do not present a great danger to people. Regarding breakdown, N 2 and CO 2 are more or less equivalent. However, the choice of which gas to use must integrate other parameters. CO 2 is usually preferred over N 2 because of its better repeatability of dielectric performance when compared to N 2 [15][16][17], and its better current interruption capabilities. These advantages justify the choice of CO 2 as a buffer gas in the development of mixtures with new gases that are potential candidates for the substitution of SF 6 in high-voltage apparatus [1,5,18].

Conclusions
This work shows that the breakdown voltage of SF 6 is the highest of the studied gases, whatever the applied voltage waveform. Similarly, for a given voltage waveform, the breakdown voltage of SF 6 is the highest. The AC breakdown voltage is the lowest voltage for all gases. Additionally, the dielectric strengths of natural gases (CO 2 and N 2 ) are very close, especially under lightning impulse voltage.
Thus, pure CO 2 and N 2 and their mixtures with SF 6 constitute potential gases for HV applications. Their insulating properties can be improved by increasing pressure and/or size of apparatus. With apparatus of the same size, the use of N 2 or CO 2 requires a working pressure about three times higher than that of SF 6 . However, it should be emphasized, that if increasing the working pressure of GIS, a new dimensioning of the mechanical structure would be required to withstand the higher pressure; one must also be very careful to remain below the security pressure. Of course, economic aspects must also be considered. In case of surface imperfections and defects in equipment what is simulated by point-plane electrodes geometry, a significant decrease breakdown voltage was observed.
On the other hand, the addition of small amounts of SF 6 to CO 2 and N 2 significantly improved the breakdown voltages of both natural gases. However, the amount of SF 6 has to remain at a reasonable value so as to not compromise the benefits of such mixtures or negatively impact the environment; the economic aspect (especially the SF 6 price) must be taken into account as well. Also, the use of CO 2 would be preferable to N 2 in the creation of mixtures used in HV apparatus, because, (1) CO 2 better interrupts the electrical arc than N 2 ; (2) the fact that for a given breakdown voltage, the ratio between the pressure of CO 2 to that of SF 6 is generally lower than that between the pressure of N 2 to SF 6 , whatever the voltage waveforms; and (3) compared to N 2 , CO 2 shows less dispersion in experimental measurements of breakdown voltage.