3.1. Magnetic probe signal

In general, the plasma produced by a laser is a source of spontaneous magnetic field^[ Reference Stamper^38]. Perhaps the first measurement of this field was performed in a spark plasma by Korobkin and Serov^[ Reference Korobkin and Serov^39], and a simple model of laser spark plasma formed by the mechanism of delayed breakdown in the focal cone of a focused ns laser beam in gas was proposed by Raizer^[ Reference Raizer^40]. Regarding the self-generating electric and magnetic fields around a laser spark, Rohlena and Mašek^[ Reference Rohlena and Mašek^41] presented an assessment of various models of spark formation and their comparison with experimental findings.

In our experiment, the spontaneous magnetic field was detected 10 cm from the laser focus using a loop magnetic probe type RS H 400-1 with a diameter of 25 mm. Its orientation allowed one to record the time derivative of the azimuthal magnetic field ${\dot{B}}_{\varphi }$ induced by the pulsed electric current flowing through the plasma core, as shown in Figure 4. Please note that ${\dot{B}}_{\varphi }$ is a part of the total detected $\dot{B}$ , which also includes the contribution of the EMP field also detected by the horn antenna 2.5 m from the spark, as described in Section 3.2. Figure 4(a) compares the time course of six probe signals S _RS together with the time course of the laser intensity, I _L, where I _L is synchronized with the S _RS in such a way that the I _L peak matches the highest positive peak of S _RS appearing at time 1 ns. The dominance of the second positive peak of S _RS is evident. Once the interaction of the laser pulse with the gas stops, the magnetic field quickly disappears, causing a rapid reduction in the energy supply to the probe, as shown in Figure 4(b). This energy was calculated using the following relationship:

(1) $$\begin{align}{E}_\mathrm{RS}=\int {S}_\mathrm{RS}^2/R\;\mathrm{d}t,\end{align}$$

Figure 4 Electromagnetic radiation detected with the use of the magnetic probe localized 10 cm from the laser spark induced by a 350 ps, 125 J laser pulse focused into a cell filled with 53.3 kPa SF₆. (a) Correlation of the relative laser intensity (right-hand Y scale) with the first positive peak of S _RS signals of the magnetic probe, which were recorded at different shots. (b) Energy absorbed by the loop probe at shot 61812. (c) STFT of S _RS for shot 61812. (d) S _RS frequency spectra for shot 61812.

where R is the load impedance given by the coaxial cables and oscilloscope input. Please note that E _RS(t) reaches a level of 46% of the maximum value approximately in 0.5 ns.

The short-time Fourier transform (STFT) with a Hanning window and an overlap of 22 of the S _RS(t) signal presented in Figure 4(c) shows that the probe signal oscillates from 0.2 to 5 GHz, with three frequency bands dominating around 1.4, 2 and 3.2 GHz.

The used window size of ≈1/3 ns, corresponding to the FWHM of the laser intensity reveals a broadband spectrum in the range from 1 to 3 GHz with a maximum of ≈2 GHz at ≈0.5 ns. This time correlates with the maximum of the laser pulse, as shown in Figure 4(a), and therefore the frequency of ≈2 GHz could be considered as the frequency of the magnetic field generated primarily around the laser spark during the interaction of the laser pulse with the gas. This magnetic field is generated by the standard mechanism of crossed electron density (n _e) and temperature (T) gradients ( $\partial \overrightarrow{B}/\partial t\sim {n}_\mathrm{e}^{-1}\left(\mathit{\operatorname{grad}}\;T\times \mathit{\operatorname{grad}}\;{n}_\mathrm{e}\right)$ ), while the electric field is supposed to be created by the polarization of the plasma due to its radial expansion across the self-generated magnetic field^[ Reference Rohlena and Mašek^41]. Thus, the magnetic probe signal reveals two sources of the electric and magnetic fields. One is the laser pulse itself, which, through electron density and temperature gradients, generates the observed azimuthal magnetic field, B _φ, winding around the spark plasma, whereas the electric field with E _φ = 0 is induced by a charge separation, that is, polarization of the plasma streaming radially across the self-generated magnetic field. We note that the B _φ field is detected just during the interaction of the laser with the gas; see Figures 4(a) and 4(c).

Assuming that the magnetic flux density B _φ is constant within the probe’s effective area A _ef, the output voltage can be calculated as follows:

(2) $$\begin{align}{S}_\mathrm{RS}=-{A}_\mathrm{ef}\;G\frac{\mathrm{d}B}{\mathrm{d}t},\end{align}$$

where G is the attenuation of the probe, specified in the manufacturer’s datasheet^{[ 42]} as −25 ± 5 dB (a factor of ≈ 0.056) within the frequency range of 200–1000 MHz. The current inducing the magnetic field in the magnetic probe can be evaluated by the integration of the near-field probe signal S _RS, which is proportional to the time derivative of B _φ:

(3) $$\begin{align}{J}_\mathrm{RS}(t)=-\frac{1}{M}\int {S}_\mathrm{RS}(t)\mathrm{d}t,\end{align}$$

where M is the conversion coefficient representing the mutual inductance between the magnetic loop and the laser spark (conducting plasma kernel). By combining Equation (3) with Equation (2) and Ampère’s circuital law, we obtain the following:

(4) $$\begin{align}{J}_\mathrm{RS}(t)=-\frac{2\pi r}{\mu_0}\;\frac{1}{A_\mathrm{ef}G}\int {S}_\mathrm{RS}(t)\;\mathrm{d}t,\end{align}$$

where r is the distance of the B-field probe from the laser spark. An example of the time course of ${\int}_0^t{S}_\mathrm{RS}(t)\mathrm{d}t$ is shown in Figure 5. As shown, the magnetic flux density peak reaches several tens of μT.

Figure 5 S _RS signal and its integral ${\int}_0^t{S}_\mathrm{RS}(t) \mathrm{d}t$ . SF₆ pressure is 26.7 kPa.

Using the PALS laser, estimated values of the magnetic field intensity near the tip of the plasma can reach 0.1 T values and the intensity of the accompanying electric field can be around 100 V/cm^[ Reference Rohlena and Mašek^41]. Another source of electromagnetic fields is the gradient of the laser sparks, which persists in a short-term self-sustaining plasma inside the spark without being actively driven by the laser pulse, as will be elucidated in Section 3.2. Please note the frequency band around 1.4 GHz, which later appeared in the STFT only after the occurrence of 2 GHz at 1 ns (see Figure 4(c)), is the only one that dominates the far zone where the horn antenna detects the EMP; see the S _HA signal shown in Figure 6.

Figure 6 EMP induced by a single 350 ps, 125 J laser pulse focused into a cell filled with 53.3 kPa SF₆. (a) Correlation of the relative laser intensity, I _L, with the first positive peak of S _HA signals recorded at different shots. (b) Energy absorbed by HA for shot 61812. (c) STFT of S _HA for shot 61812. (d) S _HA frequency spectra for shot 61812.

Please note that in this case the probe signal is induced not only by non-radiating currents, but also by emitted EMP. Since the value of M is time independent, it can be deduced that the integral of the bipolar signal S _RS(t) results in a bipolar time course of the current J _RS(t). We note that only exception was an almost unipolar J _RS(t) waveform obtained. According to Equation (3) and magnetic field B(t), shown in Figure 4, the peak of the current J _RS(t) reaches tens of amperes and a carried charge of a few nanocoulombs.

3.2. Horn antenna signal

Sparks produced by a single near-infrared (NIR) laser pulse carrying an energy of 126 ± 4 J focused into a cell filled with SF₆ at a pressure of 10.7–101.3 kPa emit EMPs, as shown in Figure 6. The horn antenna was placed 2.5 m from the laser spark at an angle of 55° to the laser vector. Figure Figure 6 shows a series of signals, S _HA, obtained by repeating firing into a SF₆-filled cell. Figure 6(a) also shows the time course of a laser pulse, where the first positive peaks of the signals S _HA of the horn antenna match the peak of the laser intensity. The time course of the EMP emission can be characterized by the variation of the energy absorbed by the horn antenna, as shown in Figure 6(b) for shot 61812. This energy was calculated using Equation (1). The time evolution of E _HA shows that the duration of the EMP exhibiting only a few oscillations is shorter than 3 ns and the corresponding decay time τ _dec is of the order of a nanosecond. While E _HA(t) reaches 46% of its maximum value in ≈3 ns, E _RS(t) reaches it in just 0.5 ns. Figure 6(c) shows an example of the STFT of S _HA detected at shot 61812, where the dominant frequency f _c ≈ 1.1 GHz. This also shows that the EMP reaches its peak about 1 ns after the arrival of the maximum laser intensity. The frequency spectrum presented in Figure 6(d) reveals the dominant frequency lying in the range of the 0.8–1.5 GHz band. Frequencies from the 1.5–2.2 GHz band occur only in some shots.

The HA signals prove that the occurrence of a laser spark is accompanied by an EMP pulse. This microwave EMP is caused by time-varying currents originating from various sources such as ions, runaway electrons and slow electrons. The runaway (hot) electrons from the laser kernel move into the surrounding cold gas, where they produce secondary electrons in collisions with gas molecules. The runaway electrons thus cause the formation of an electrical double layer at the interface between the laser kernel and the surrounding gas, where the positive charge is located on the surface of the laser kernel and the negative charge is in the layer of gas touching the kernel. From a phenomenological point of view, the double layer can be considered a potential well. The positive charge of this potential well is created by the escape of hot electrons from the laser kernel, while the negative charge is created by these runaway electrons being captured by the cold gas near the core. Thus, the boundary between the kernel and the cold gas could be termed as a spark double layer (SDL). Eliezer and Hora^[ Reference Eliezer and Hora^43] approximated the hydrodynamic bounce frequency, ω_PW, in such a potential well by the following relationship:

(5) $$\begin{align}{\omega}_\mathrm{PW}=\frac{2\pi }{l}\sqrt{\frac{e{\phi}_0}{m}},\end{align}$$

where l is the dimension of the potential well, that is, the SDL thickness comparable to the mean free path of electrons, ϕ ₀ is the SDL potential and m is the electron mass. The electric field could be found by solving Poisson’s equation for the electrostatic potential in combination with equations for the density of low-energy electrons and positive and negative ions, including the runaway electron mechanism. Please note that the temporary experimental technique does not allow us to measure this potential.

It is reasonable that the knowledge of the different periods of the EMP (frequency spectrum) and corresponding decay times on a short time scale can provide basic experimental information about the mechanism of ion production in the vicinity of the spark kernel. We note that no natural frequencies or resonant modes affected the frequency spectrum of the EMP emitted by the laser-produced spark in the gas.

The mechanism of runaway electrons in a laser spark is fundamentally different from the mechanism of runaway electrons that are produced when a laser pulse interacts with a solid target placed in the vacuum. Firstly, many orders of magnitude more electrons are generated during solid target ablation than in the laser spark mode. Secondly, the electron flux escaping from the spark is therefore many times smaller and is stopped by collisions with particles of the surrounding gas, while in the case of solid targets irradiated in a vacuum, the runaway electrons pass a long way to the walls of the vacuum chamber. Therefore, the intensity of the EMP generated in a laser spark discharge is less than in the case of solid targets irradiated in a vacuum, as shown in Figure 7. The presented comparison shows that the energy absorbed by the HA was about 4000 times higher by detecting the EMP emitted by the interaction of the laser with the copper target than by the spark produced in the SF₆ by a laser delivering the same energy. However, in this case, the HA detected the EMP from the Cu plasma at 4.5 m from the target, while in the experiment with SF₆ it was located only 2.5 m from the spark. We can conclude that the EMP gain from a laser spark is a thousand times smaller than when the laser interacts with solid particles placed in vacuum.

Figure 7 Comparison of energy absorbed by HA detecting EMPs produced by the interaction of a 150 J laser pulse with a 1 mm thin Cu target and SF₆ gas at a pressure of 101.3 kPa.

Although the mechanism of broadband emission from a laser spark is not yet fully understood, there is some similarity in the EMP spectra, both when using ultrafast laser pulses and nanosecond pulses. It turns out that an important parameter is pressure, expressed, for example, in the density of the irradiated gas. At the same or higher density of gas irradiated with 30 fs laser pulses, the frequency spectra had a range of around 0.8–2 GHz^[ Reference He, Wang, Deng, Feng, Xia, Hu, Zhu, Xie, Yuan, Zhang, Lu, Yang, Cheng, Li, Yan, Fang, Li, Zhou, Li, Chen, Lin and Yan^44], as in our experiment. In contrast, with decreasing gas pressure, not only does the amplitude of the electromagnetic field increase, but also frequencies are higher than the 3 GHz registered by us, as shown by the experiment of Engelsbe et al. ^[ Reference Englesbe, Elle, Reid, Lucero, Pohle, Domonkos, Kalmykov, Krushelnick and Schmitt-Sody^45]. At a pressure of 66 Pa, the 50 fs laser generated approximately a hundred times more intense EMPs than at a pressure of 84 kPa. However, 10 GHz and higher frequencies were well pronounced. In the case of PALS experiments with solid targets, frequencies higher than 3 GHz are also generated^[ Reference Cikhardt, Bradford, Ehret, Agarwal, Alonzo, Andreoli, Cervenak, Ciardiello, Consoli, Davino, Dostal, Dudzak, Klir, Krasa, Krupka, Kubes, Malir, Mendez, Munzar, Novotny, Renner, Rezac, Ruiz, Santos, Sciscio, Singh, Valdova, Juha and Krus^46]. These experimental results support our idea that the SDL could be a source of EMPs.

3.3. Laser spark visualization

A photograph of a glowing spark can provide a basic insight into the plasma distribution in the cell, as shown in Figure 8. Please note that this is a time-integrated image of transient laser spark emission taken in the visible spectrum. The laser pulse arrives from the right-hand side after passing through a lens with a focal length of 300 mm and then an input cell window. Luminescence of the laser-irradiated SF₆ occurs as soon as the laser pulse enters the cell, where it reaches an intensity of about 5 × 10⁹ W/cm². As the image shows, the luminescence appears on microislands that are irregularly distributed in the cell. The life cycle of a spark can reach up to tens of microseconds^[ Reference Harilal, Brumfield and Phillips^47, Reference Sun, Chang, Rong, Wu and Zhang^48].

Figure 8 Passive laser spark imaging due to LIDB plasma optical emission. The longitudinal (Hline) and radial profiles (VL) of the spark’s luminosity were extracted at the locations marked by the yellow lines in the diagram. Red arrows indicate the focused breakdown laser beam.

Figure 8 shows that the spark length is approximately 2.3 cm. This dimension can be considered the limiting dimension of a spark emitting visible radiation. However, the dimensions of the kernel with the released electrons will probably be smaller. Regardless of the mechanism of the magnetic field generation in the kernel that affects the trajectories of the boundary secondary electrons, it is evident that the number of these boundary electrons is very small. Assuming that the signal of the magnetic probe localized in the near zone is induced only by the electron current flowing through the laser spark, a rough estimate of the peak current value indicates a current of only up to 100 mA. It corresponds to a current only of about 10⁸ electrons. Please note that the first estimate of the magnetic and electric field intensities did not consider the possibility of laser spark generation in an electronegative gas^[ Reference Rohlena and Mašek^41].

The laser creates a kernel consisting of fully ionized fluorine and almost fully ionized sulphur as well as electrons. Please note that although we did not measure the mass spectra of the kernel-forming ions, we can estimate the degree of ionization of the S and F elements from the spectra of polytetrafluoroethylene (PTFE) ionized with an equivalent laser energy of approximately 150 J also delivered by the PALS beam, which identified F⁹⁺ ions far from the target^[ Reference Krása, Velyhan, Jungwirth, Krouský, Láska, Rohlena, Pfeifer and Ullschmied^49]. The kernel formed by multiple ionized S and F ions expands into the background gas, creating a ‘bubble’ around the focus. In addition, this kernel is a source of UV and X-ray radiation, which are absorbed by SF₆. This produces a secondary plasma core within SF₆, in which the separation of electrons and ions also occurs.

3.4. Single-shot optical emission spectra

To help characterize a laser spark, a fibre optic spectrometer is used to analyse the emissions from the spark. The recorded spectra contain both the continuum background originating from the hot plasma kernel and the visible spectrum of recombining ions and excited atoms during laser spark quenching, as Figure 9 shows. This spectrum has a similar profile to the spectra of laser-generated sonoluminescent bubbles, which exhibit the characteristic blackbody spectrum (background continuum) observed in both laser-produced and HV discharge plasmas^[ Reference Kappus, Khalid, Chakravarty and Putterman^50– Reference Krile, Vela, Neuber and Krompholz^52]. The observed emission spectra are affected by the dramatically changing temperature determining blackbody radiation that occurs during the early phase of spark production. For these reasons, it was not possible to perform the best fit of the continuum background using the Planck law function of blackbody radiation. Although the background spectrum has been reported in several papers, its nature is still unclear. Therefore, we simulated the background signal using three bi-Gaussian functions with central wavelengths of 412.7, 500.7 and 561.2 nm.

Figure 9 Optical emission spectra of laser sparks produced in SF₆ (101.3 kPa, 133 J) and fit of sulphur lines: (a) S II lines at T _e = 0.9 eV and n _e = 1 × 10¹¹ cm^–3, (b) S III lines at T _e = 2.2 eV and n _e = 1 × 10¹⁷ cm^–3, fluorine lines F I at T _e = 0.9 eV and n _e = 1×10¹⁴ cm^–3, and the background signal (BS).

The presented spectroscopic lines data of sulphur and fluorine are from the National Institute of Standards and Technology (NIST) atomic spectra database^{[ 53]}. These were superimposed on the background signal so that the resulting spectrum matched the observed spectrum. Due to the varying temperature, T _e, and density, n _e, of electrons during spectrum recording, no average values of these parameters were used to exactly specify spectral peaks of S and F ions and atoms. However, the spectrum was divided into three wavelength ranges of 300–460, 460–600 and 600–720 nm to help clarify the origin of the spectral lines. In the first wavelength range of 300–400 nm, the S III spectral lines dominate, which approximately correspond to T _e = 2.2 eV and n _e = 1 × 10¹⁷ cm^–3. The decisive parameter is the electron temperature. When T _e drops to about 1 eV, lines from the second range appear, which correspond mainly to S II lines. The fluorine lines F I dominate for T _e = 0.9 eV and n _e = 1 × 10¹⁴ cm^–3 in the third range of 600–720 nm. Although we did not measure optical emission spectra (OESs) as a function of time and, thus, did not obtain time-resolved values of T _e and n _e, the values estimated in our experiment are like those obtained in the experiment with a laser spark produced in air with a very low pulse energy of 40–150 mJ performed by Yalçin et al. ^[ Reference Yalçin, Crosley, Smith and Faris^54]. The value T _e ≈ 1 eV obtained by Yalçin et al. was detected more than 1 μs after the end of the laser interaction. However, T _e in the kernel should reach much higher values (hundreds of eV). Here, T _e was estimated from the OES of plasmas produced in a SF₆ gas-puff target by a focused beam of a neodymium laser delivering 15 J of energy in 1 ns^[ Reference Bartnik, Dyakin, Parys, Skobelev, Faenov, Fiedorowicz and Khakhalin^55].

Although the purity of the SF₆ used was 99.9%, the impurity content of 0.1% in the gas and impurities absorbed on the inner surface of the cell allowed detection of the H_α line with a wavelength of 656.28 nm, as Figure 10 shows. Due to the low impurity content, the amplitude of the H_α line is small. However, the estimation of the H_α line width is crucial because it is assumed to be reciprocally correlated with the electron density of the plasma due to the Stark broadening^[ Reference Yalçin, Crosley, Smith and Faris^54, Reference Jasik, Heitz, Pedarnig and Veis^56– Reference Hussain Shah, Iqbal, Ahmad, Khandaker, Haq and Naeem^58].

Figure 10 Detail of the optical emission spectra of laser sparks produced in SF₆ at 101.3 kPa. The arrow indicates the wavelength of the H_α spectral line.

Stark broadening can lead to asymmetric line profiles, which can be analysed by fitting a Voigt function to the observed H_α line profile to determine the FWHM, w _FWHM, of the spectral line. Using this technique, which is widely used in various types of plasmas, the magnitude of the electron density, n _e, can be calculated using the following relationship^[ Reference Gigosos, Gonzalez and Cardenoso^57]:

(6) $$\begin{align}{n}_\mathrm{e}={\left(\frac{w_\mathrm{FWHM}}{1.098}\right)}^{1.4713}\times {10}^{17}\;\mathrm{c}{\mathrm{m}}^{-3}.\end{align}$$

Using the relationship (Equation (6)) we obtain n _e ~ 10¹⁷ cm^–3. This value correlates with the estimated value for the S III wavelength range shown in Figure 10.

The contribution of the Gaussian profile width to the total line is comparable to the Lorentzian width for other spectral lines, for example, the S V triplet at 703 nm. This triplet S V and the three fitted Voigt functions are shown in Figure 11. Fitting gave w _G = 0.146 and w _L = 0.337 being shared for all functions. The S V triplet was analysed using the Voigt function with PeakFit software.

Figure 11 Detail of the optical emission spectra of laser sparks produced in SF₆ at 26.7 kPa. The peak corresponds to the S V triplet at 702.76, 703.00 and 703.45 nm.

3.5. Shot-to-shot reproducibility

Like other experiments dedicated to the interactions of high-power laser pulses with solids, the presented experiment exhibits significant shot-to-shot fluctuations. Fluctuations relate not only to the emission of electrons, ions and possibly products of fusion processes, but also to the emission of EMPs and the continuum background. The range of EMP fluctuations during the interaction of an approximately 370 GW laser pulse with gas is shown in Figure 12.

Figure 12 Time course of the energy of the horn antenna signal induced by EMPs emitted from laser spark produced in SF₆ at 26.7 kPa, E _L ~ 125 ± 7 J.

The time-resolved dependence of the energy absorbed by the horn antenna shows that fluctuations occur already during the first phase of the laser–SF₆ interaction, that is, during the first 200 ps. The energy absorbed by the horn antenna fluctuates from 0.25 to 1.8 nJ during this period, while the fluctuations of the delivered laser energy are less than ±5%. The efficiency of converting laser energy into EMP energy, which is related to the flow of electrons escaping from the spark kernel, therefore indicates significant fluctuations in the absorption of laser radiation by the gas, as in the case of laser–solid interaction. However, please note that there is a significant difference between the two interactions, namely that the electron concentration, n _e, in the case of SF₆ cannot reach values higher than or equal to the critical value, n _c. However, in the case of fully stripped F and S atoms, the electron density can reach values higher than n _c/4, and the absorption of laser radiation can be affected by the instability of two-plasmon decay (TPD)^[ Reference Seka, Myatt, Short, Froula, Katz, Goncharov and Igumenshchev^13, Reference Cristoforetti, Antonelli, Atzeni, Baffigi, Barbato, Batani, Boutoux, Colaitis, Dostal, Dudzak, Juha, Koester, Marocchino, Mancelli, Nicolai, Renner, Santos, Schiavi, Skoric, Smid, Straka and Gizzi^59].

Shot-to-shot fluctuations also occur in the emission of visible light, as Figure 13 shows. The curves show the average intensity recorded by the camera during three series of shots. Fresh SF₆ charge was used in the first and second series, while the charge from the second series was used in the third series, but the break between them was 19 h (the average is shown by the red line labelled as ‘1st shots’). The luminosity of the first shots fluctuated around 5%. It always dropped by 20% with the second shot (see the black line labelled ‘subsequent shots’). Starting with the second shot, the luminosity fluctuated within ±10%. The subsequent shots were repeated with a period of approximately 30 min.

Figure 13 Intensity profile along the caustic line evaluated from spark photographs. The profiles shown are averages of the intensities obtained over three series of shots at SF₆ pressure of 26.7 kPa, E _L ~ 125 ± 7 J.

The shot-to-shot fluctuations in spark luminosity are different from fluctuations in the EMP emission and continuum background. While fluctuations in the luminosity of the sparks produced by the second and subsequent shots are steadily at the level of 10%, the EMP fluctuations are more pronounced, as shown in Figure 12. However, both phenomena, emission of EMPs and visible radiation, are detected at different stages of spark production and extinction and are therefore driven by different mechanisms.

3.6. Chemical change initiated by laser sparks

From a chemical point of view, the interaction experiments reported here can be divided into two groups. In the first one, gaseous SF₆ contained naturally admixed moist air (N₂/O₂/CO₂/H₂O) so that the conditions corresponded to those that can be expected in the normal use of SF₆ as an industrial gaseous dielectric. In the second series of experiments, SF₆ contained admixtures of dry air only (N₂/O₂/CO₂). Moisture (H₂O) was excluded from all gases as well as gas handling systems and procedures used. Final products formed in reactions initiated by PALS-produced plasmas have been in both series analysed using GC-MS detection.

In samples containing the moist air, gas chromatograms reveal only three products of SF₆ reactions, that is, thionyl fluoride (SOF₂), sulphuryl fluoride (SO₂F₂) and thionyl tetrafluoride (SOF₄). Yields of all three products are low (~1%). Their abundances depend on the pressure in the cell. At p(SF₆) = 10.7 kPa, that is, the lowest pressure in the cell, there are SOF₂ and SOF₄ formed (Figure 14). If we increase the pressure of SF₆ to 26.7 kPa, SOF₂ and SO₂F₂ appear as products. At even higher SF₆ pressures, that is, 53.3 and 101.3 kPa (atmospheric pressure), only one product (SOF₂) is indicated. The overall (stoichiometric) reactions leading to the above-mentioned final products could be expressed as follows:

$$\begin{align*}{\mathrm{SF}}_6+{\mathrm{H}}_2\mathrm{O}\to {\mathrm{SO}\mathrm{F}}_2+2\mathrm{HF}+{\mathrm{F}}_2,\\ {}{\mathrm{SF}}_6+2{\mathrm{H}}_2\mathrm{O}\to {\mathrm{SO}}_2{\mathrm{F}}_2+4\mathrm{HF},\\ {}{\mathrm{SF}}_6+{\mathrm{H}}_2\mathrm{O}\to {\mathrm{SO}\mathrm{F}}_4+2\mathrm{HF}.\end{align*}$$

Figure 14 The gas chromatogram of SF₆ (containing traces of moist air) chemically altered by LIDB plasmas induced by four laser pulses focused into the gas cell (the total pressure in the cell is 10.7 kPa) shot by shot.

However, neither HF/F₂ nor SiF₄ molecules have been indicated in gas chromatography records. The silicon tetrafluoride should be formed in reactions of hydrogen fluoride and/or molecular fluorine with SiO₂ in the cell wall.

Contrary to the samples contaminated by moist air, the samples that do not contain traces of water vapour exhibit quite poor laser–plasma chemical reactivity. In addition to remaining SF₆ and air, only sulphuryl fluoride (SO₂F₂) has been found in the gas cell after four and eight laser shots accumulated at initial total pressures varied from 10.7 to 101.3 kPa. Therefore, SO₂F₂ molecules represent a single stable product that testifies to the key role of molecular oxygen (and transient species formed from its molecule upon LIDB conditions) in the reaction mixture because water vapour is not present in these runs. Considering the situation just stoichiometrically, the overall reaction SF₆ + O₂ → SO₂F₂ + 2F₂ provides sulphuryl fluoride. However, the reaction gives molecular fluorine as another product, which was not registered by GC-MS.

Notable is the absence of SiF₄, which is thought to be formed by reactions of fluorine, hydrogen fluoride and other fluorine-containing reactive species with SiO₂ in the glass body of the cell and its window. The SiF₄ formation was observed in a narrow (1.44 cm in diameter) cell filled with SF₆ in which the focused pulsed CO₂ laser produced LIDB plasmas^[ Reference Lin and Ronn^60]. In the experiment reported here, a much larger inner diameter of the cell and the use of a long focal length lens ensure that fluorine reactants do not penetrate the mass of ambient, unirradiated gas towards the cell wall and the beam entrance window. Therefore, silicon tetrafluoride is not formed under these conditions. Both larger dimensions of the cell and a strong reduction of the gas mixture contact with carbon-containing components (e.g., O-rings, vacuum grease and flanges made of plastics) and contaminants (hydrocarbons from vacuum line) are likely responsible for an absence of CF₄ formation. This product has also been frequently reported in SF₆ subjected to electrical discharges^[ Reference Mahdi, Abdul-Malek and Arshad^61] and laser sparks^[ Reference Lin and Ronn^60] when fluorine-rich products and transients interact with carbon-containing species and surfaces in the gas cell.

In conclusion, we can say that the fully reproducible chemical change can be registered and quantified under the irradiation conditions applied here. Conversion efficiency and formation yields of the initial substance (SF₆) and final products (SO _x F _y ), respectively, are both low. In addition to that, the number of products is very low, that is, only one or two products, depending on water vapour content and the pressure of the initial substance. The absence of any S _x F _y molecule (e.g., SF₄ and/or S₂F₁₀) among products leads to the assumption that the product-forming reactions here are not stepwise unimolecular decompositions starting from SF₆^[ Reference Mahdi, Abdul-Malek and Arshad^61– Reference Miletic, Neškovic, Veljkovic and Zmbov^65], but rather bimolecular reactions of the initial substance with oxygen species (e.g., molecular and atomic oxygen, hydroxyl radicals and water molecules). All the above-mentioned findings contrast with those (especially a wide variety of products) obtained in electrical discharges between electrodes (see, for example, Refs. [Reference Kurte, Heise and Klockow14,Reference Mahdi, Abdul-Malek and Arshad61] and the references cited therein), conventional pyrolysis in a resistively heated reaction tube^[ Reference Padma and Murthy^66] and laser sparks induced in a narrow gas cell^[ Reference Lin and Ronn^60, Reference Eletskii, Klimov and Legasov^67], where an interaction of plasmas (including radiation, particles and reactive species liberated from the plasma^[ Reference Kurte, Heise and Klockow^14, Reference Lin and Ronn^60, Reference Mahdi, Abdul-Malek and Arshad^61, Reference Miletic, Neškovic, Veljkovic and Zmbov^65] or hot gas^[ Reference Padma and Murthy^66]) with solid surfaces can take place. Under our experimental conditions, the influence of solid surfaces on plasma chemical SF₆ decomposition patterns seems to be significantly reduced.