Optical properties and decomposition mechanisms of SF₆ at different partial discharge determined by infrared spectroscopy Open Access

Sulfur hexafluoride (SF₆) has seen wide application in power industry as an insulation material due to its high dielectric strength and outstanding arc-extinguishing performance.¹ In particular, pressurized SF₆ gas is extensively used in gas insulated switchgear (GIS) to ensure safe and reliable operation. Despite its chemical inertness at room temperature, SF₆ can decompose into lower sulfur fluorides under discharge induced by insulation defects in GIS.² Further reactions of these compounds with electrode materials and gas impurities in GIS generate products of significantly higher chemical reactivity and lower dielectric strength, such as CF₄, SO₂F₂, SOF₂, and SO_2, and increase the risks of electrical breakdown^3–5 Thus, it is of great importance to detect the decomposition products of SF₆ and thoroughly investigate the underlying mechanisms from both scientific and practical points of views.

Partial discharge (PD) has been accepted as an early evidence of insulation degradation inside GIS and one of the essential causes of degradation rate acceleration.⁶ The types and concentrations of decomposition products are strongly affected by factors such as discharge time, gas pressure, and discharge voltage. In spite of the extensive study on related topics in recent years, many details of SF₆ decomposition under PD remain controversial. Many detection techniques have previously been applied for the analysis of SF₆ decomposition products, such as gas chromatography, gas detection tubes, electrochemical methods, and spectral methods. However, the concentration of SF₆ and the development mechanisms still remain a controversial topic. Spectral techniques is a nondestructive and powerful technique to investigate the optical characteristics of materials.⁷ There are some spectral techniques to determine decomposition components and optical properties of SF₆, such as ultraviolet transmission spectroscopy, photoacoustic spectroscopy, and Fourier transform infrared spectroscopy (FTIR). Among them, FTIR has the merit of being nondestructive and easily accessible for the determination of decomposition components and optical properties of SF₆.⁸ Therefore, it is desirable to carry out a delicate study regarding the essential properties and their relation to SF₆ and its decomposition products using FTIR techniques.

In this paper, the optical properties of SF₆ and its decomposition products under different partial discharge conditions have been studied in the spectral range of 500-4300 cm^-1 using FTIR, and a detailed analysis of the influence of discharge time, gas pressure, and discharge voltage on the decomposition products has been performed.

For a GIS in standard operating conditions, the GIS chamber is filled with SF₆ gas, and it works with the SF₆ gas pressure of 0.3 MPa to 0.6 MPa. Therefore, the gas pressure dependence of the decomposition concentrations was investigated at 0.3 MPa, 0.35 MPa, and 0.4 MPa, respectively. Based on the operating voltage values, the GIS can be divided into three types: (1) medium voltage GIS (1 kV-66 kV); (2) high voltage GIS (66 kV-550 kV); (3) ultra-high voltage GIS (>550 kV). In this work, optical properties and decomposition mechanisms of SF₆ under partial discharge were researched based on the medium voltage GIS in standard operating conditions.

To obtain the infrared spectra of SF₆ and its main decomposition products, gas calibration standards with SF₆ as the balance gas were selected to simulate the actual state in GIS. Infrared transmittance spectra were measured over the spectral range of 500-4300 cm^-1 using an FTIR spectrometer (Bruker Vertex 80 V) equipped with a 10 cm optical sample cell and a HgCdTe detector. Pure SF₆ (>99.999%) was used for discharge experiments. Discharging time (up to 96 hours), gas pressure (0.3 MPa, 0.35 MPa, and 0.4 MPa), and discharging voltage (20 kV, 43 kV, and 46 kV) were varied to investigate their influence on the decomposition products. It should be noted that all samples were measured at room temperature and no mathematical smoothing was performed on the experimental data.

PD is the most common discharge modes in electrical equipment and usually occurs in the reaction zone of the gas chamber such as glow region, ion-drift region, and main gas volume. PD-induced electron collisions result in ionization, excitation, and decomposition of SF₆ molecules, generating intermediate products such as F atoms and lower sulfur fluorides such as SF₂. These highly reactive chemical species further react with other substances present in the gas chamber, such as carbon atoms (from the stainless steel electrodes and/or polymer materials), trace O atoms or O₂ molecules, forming various chemically stable products.^3–5 For example, the SOF₂ and SO₂F₂ molecules can be formed when the SF₂ molecules continue to react with O atoms or O₂ molecules. It has been suggested in previous research work that CF₄, SOF₂, and SO₂F₂ are the main decomposition products in the case of metal protrusion defects or creeping discharge faults, and the concentration ratios between them may provide crucial information on detailed fault type. Thus, these stable and characteristic compounds are systematically investigated in this work.^9–11 The main chemical reactions involved are as follows:

For the structure of SF₆ molecule, Six F atoms and one S atoms are set to describe the ground state SF₆ molecule, which shows an octahedral symmetry structure. The average S-F bond length is about 1.581 Å, which is similar to that predict by Fu et al¹⁰ and Tang et al.¹² For the structure of SOF₂ molecule, two F atoms and one O atoms surround one S atom. A trigonal pyramidal structure has been identified for SOF₂, with an average S-F bond length of 1.602 Å and a S-O bond length of 1.433 Å, both in good agreement with previously reported values.¹³ 

Fig. 1 shows the full infrared spectrum of SF₆ gas measured at the discharging time of 96 hours. The gas pressure and discharging voltage is fixed at 0.3 MPa and 46 kV, respectively. In order to exclude the effects from H₂O and CO₂, the photo frequency range of 600–1600 cm-1 was selected for SF₆ breakdown products analysis. In addition, the characteristic peaks of SF6 decomposition products are mainly distributed at this range. From the Fig. 1, we can observe the typical features of CF₄, SOF₂, and SO₂F₂ decomposition products. It indicates that FTIR spectra can be used to investigate the decomposition products and mechanisms of SF₆. The full infrared spectrum of SF₆ under partial discharge is similar to that reported from the previous reports.^1,14

The main absorption band of CF₄ molecules is located in the spectral range of 1280-1290 cm^-1.¹⁵ Fig. 2(a) shows the infrared spectra of CF₄ generated under PD, with varying discharging time between 30 and 96 hours. Three spectral features A, B, and C, originated from vibration transitions of CF₄, can be observed at about 1281 cm^-1, 1283 cm^-1, and 1285 cm^-1, respectively. This is related to the CF₄ vibration transition that is accompanied with the changing of the dipole moment. The intensities of all absorption bands increase with increasing the discharging time. The absorption coefficient α of CF₄ products can be obtained by fitting the infrared spectra with the certain physical model. By analyzing the absorption spectra of CF₄ decomposition products, the absorption coefficient values increase while the discharging time is increased. It suggests that the concentrations of CF₄ products increase while the discharging time is increased.

In order to further investigate the time dependence of CF₄ concentration, the intensities of peaks were fitted with respect to the discharging time in Fig. 2(b). The fitting results reveal a largely linear drop in transmittance for all features, indicating a monotonic linear increase in CF₄ concentration up to 96 hours. Similar dynamics have been observed by other groups using gas chromatography.^4,16 CF₄ has been reported as a characteristic product of discharge involving carbon atoms in steel and epoxy resin composites, and as a typical decomposition product for insulation defect rather than floating potential defect.¹⁵ Therefore, the increase in CF₄ concentration can be attributed to the increase of carbon-related defects. Any obvious deviation from a linear increase in CF₄ concentration over time may indicate a change in the condition of PD in GIS.

The characteristic absorption peaks and decomposition products of SOF₂ are mainly distributed at 1330-1340 cm^-1. The infrared spectra of SOF₂ collected at different discharging time between 30 and 96 hours are shown in Fig. 3(a). The peak at about 1330 cm^-1 can be ascribed to the feature peak of SOF₂ products. No obvious absorption peak can be observed at the beginning, so the spectra data measured at the discharging time less than 30 hours are not showed in this figure. Two characteristic absorption peaks emerge at 1330-1340 cm^-1 after 30 hours, indicating a relatively late start of the SOF₂ formation.

Fig. 3(b) shows the change in infrared spectra of SO₂F₂ over the same time period of discharging. Three characteristic peaks at 863 cm^-1, 873 cm^-1, and 902 cm^-1 can be identified from the data. With increasing the discharge time from 30 to 96 hours, we can observe that the intensities of the three characteristics increase again. Therefore, one can also conclude that the concentrations and absorption coefficient of SO₂F₂ decomposition products will also increase with the discharge time. The detailed concentration of SO₂F₂ products under different partial discharging conditions will be calculated later.

In order to investigate the effects of discharge time, discharge voltage, and gas pressure on the concentrations of decomposition products, the concentrations need to be calculated by fitting the infrared spectra. The absorption intensity is proportional to each component concentration and optical path length. Therefore, the concentrations of SOF₂ and SO₂F₂ at different discharging time were determined based on the Lambert-Beer law, which can be expressed as follow:¹⁵ 

Where T is the infrared transmission of the sample, α is the absorption coefficient, b is the length of the sample, and c is the gas concentration. The gas sample absorbs a specific frequency of infrared light when a beam of light passes through the gas sample.

Then we can further estimate the concentrations of the decomposition products by partial-least square calibration procedure.^17,18 The characteristic spectral range of each decomposition product was selected as calibration sample for quantitative analysis. The optimal number of component in PLS was determined by cross-validation.³ The best-calibration procedure was chosen by optimizing simultaneously the comparison between the experimental and calculated spectra. The root mean square error (RMSE) is an error measure for how well the model and calibration procedure performs. The fitting is a process of minimizing RMSE with the optimized values of the fitting parameters. In this work, the RMSE value of the calibration for SOF₂ product is 0.84%, while the RMSE value of the calibration for SO₂F₂ product is 1.67%. It indicates that a good agreement is obtained between the experimental and calculated spectra in the measured photon frequency range.

Fig. 4(a) shows the SO₂F₂ concentrations measured under a discharging voltage of 20 kV, 43 kV, and 46 kV, respectively, with the gas pressure fixed at 0.3 MPa. It turns out that under all discharging voltages, the concentrations increase over time, while a higher voltage always leads to a higher concentration for a specific discharging time. The similar trend has been observed by Ding et al.¹⁹ They found that the SOF₂ and SO₂F₂ decomposition products increase linearly with the accumulation of partial discharge energy. During the discharge, the strong interaction of high-energy electrons with SF₆ causes their rapid deceleration to the lower energy of electron capture and dissociative attachment.²⁰ Therefore, the SF₆ molecules are more likely to break down at higher electric field strengths.

Finally, Fig. 4(b) shows the influence of gas pressure on the SO₂F₂ formation. SO₂F₂ concentrations were measured under a fixed discharging voltage of 46 kV. It is interesting to see that the SO₂F₂ concentration decreases with increasing gas pressure for any specific discharging time. Under higher pressure, both the electron drift velocity and the volume of active region are reduced, lowering the overall rate of the decomposition reaction.¹⁶ Increased concentration of O₂ can also promote the formation of SO₂F₂ products.¹⁰ More explanations about the physical and chemical mechanisms need to be carried out by first-principle calculations. We can expect that the optical absorptions of SO₂F₂ decomposition products will decrease by increasing the gas pressure.

The influence of discharging time, gas pressure, and discharging voltage on SF₆ decomposition products have been evaluated by Fourier transform infrared spectroscopy at room temperature. The infrared spectra of decomposition products such as CF₄, SOF₂, and SO₂F₂ have been obtained and the origins of the molecules have been discussed. It can be found that the CF₄, SOF₂, and SO₂F₂ concentrations increase with increasing discharging time. The concentration of SO₂F₂ increases under higher discharging voltages, but decreases at higher pressure owing to the reduction in the electron drift velocity and the volume of the active region. This work establishes a relationship between product concentrations and PD conditions.

This work was financially supported by the Science and Technology Foundation of China Southern Power Grid (GXKJXM20152027).