Time-Discretized Extreme and Vacuum Ultraviolet Spectroscopy of Spark

Home , Emission spectrum, Photoionization

Journal of Physics D: Applied Physics

PAPER Related content

- Photoionization capable, extreme and Time-discretized extreme and vacuum ultraviolet vacuum ultraviolet emission in developing low temperature plasmas in air J Stephens, A Fierro, S Beeson et al. spectroscopy of spark discharges in air, N2 and O2 - A nanosecond surface dielectric barrier discharge at elevated pressures: time- To cite this article: D Trienekens et al 2016 J. Phys. D: Appl. Phys. 49 035201 resolved electric field and efficiency of initiation of combustion I N Kosarev, V I Khorunzhenko, E I Mintoussov et al.

- Optical emission spectroscopy study in the View the article online for updates and enhancements. VUV–VIS regimes of a developing low- temperature plasma in nitrogen gas A Fierro, G Laity and A Neuber

Recent citations

- Guided ionization waves: The physics of repeatability XinPei Lu and Kostya (Ken) Ostrikov

- Practical considerations for modeling streamer discharges in air with radiation transport J Stephens et al

- Photoionization capable, extreme and vacuum ultraviolet emission in developing low temperature plasmas in air J Stephens et al

This content was downloaded from IP address 129.118.3.25 on 05/09/2018 at 18:26 IOP

Journal of Physics D: Applied Physics

J. Phys. D: Appl. Phys. J. Phys. D: Appl. Phys. 49 (2016) 035201 (5pp) doi:10.1088/0022-3727/49/3/035201

49 Time-discretized extreme and vacuum

2016 ultraviolet spectroscopy of spark

JPAPBE D Trienekens1,2, J Stephens2, A Fierro2, J Dickens2 and A Neuber2

035201 1 Elementary Processes in Gas Discharges, Eindhoven University of Technology, Eindhoven, The Netherlands 2 Texas Tech University Center for Pulsed Power and Power Electronics, Lubbock, TX 79407, USA D Trienekens et al E-mail: [email protected]

Received 28 July 2015, revised 21 October 2015 Accepted for publication 2 November 2015 Published 9 December 2015 Printed in the UK Abstract JPD In this paper we present time-discretized spectra of spark discharges in air, N2 and O2. In previous work, a system for temporally resolved spectral analysis of extreme ultraviolet (EUV) and vacuum ultraviolet (VUV) emission from spark discharges was presented, along 10.1088/0022-3727/49/3/035201 with some initial results. As was noted in this paper, statistical variances and the lacking of an apparatus sensitivity profile limited the usability of the data obtained. We have investigated the cause of these variances and improved the setup to reduce their effect. We also investigated Paper the apparatus sensitivity profile to correct the intensity of measured lines. Newly obtained spectra in dry air, N2 and O2 are presented. Air and N2 show high emission in the vicinity of 100 nm, where direct photoionization of molecular oxygen is possible, in the first 250 ns of 0022-3727 the discharge. We conclude this emission originates from nitrogen, which has several intense molecular transitions in this region. This finding is confirmed by our experimental results which show the emission in this region is much lower in oxygen. 3 Keywords: VUV spectroscopy, high voltage breakdown, photoionization (Some figures may appear in colour only in the online journal)

I. Introduction (referred to as a puff valve), a pressure gradient is introduced in a discharge chamber, such that a spark discharge takes Photoionization plays an important role in the early stages place in near atmospheric conditions, whereas the rest of the of streamer and spark discharges, as becomes evident from system remains under vacuum. The latter allows for the prop- experiments on discharges in N2/O2 mixtures [1–3]. Discharge agation of EUV/VUV radiation generated by the discharge. models incorporating photoionization generally assume Removing the transparent dielectric barrier separating the the necessary photoionization capable emission originates discharge chamber from the spectrograph (as was used previ- from molecular bands of nitrogen in the vacuum ultraviolet ously in [13, 14]) enables studying EUV radiation down to (98–102.5 nm) [4–9]. This assumption is based on spectra 60 nm, which would otherwise be absorbed by the dielectric obtained by measuring the spectra resulting from electron barrier. impact in N2 [10, 11]. However, experimental proof that this In [12], it was noted that statistical variances significantly emission also takes place in the early stages of streamer and affected the measured spectra, leading to physically unresolv- spark discharges is lacking. able negative values for background corrected intensities. In a previous paper [12], a system was presented that is Also, it was noted that spectral lines often appeared to be very capable of performing spectroscopic analysis of spark dis- broad. Although it was not made clear whether this was due to charges in the EUV/VUV region (60–160 nm in this case). the apparatus profile or due to physical line broadening mech- Using an electrically actuated fast pneumatic gate valve anisms, it is reasonable to assume that the apparatus profile

Reflectance (%) 20 0 60 80 100 120 140 160 Wavelength (nm)

Figure 1. Grating reflectance (%) as a function of wavelength (nm) for the Princeton #1200 coated grating [16, 17].

is the dominant contributor to measured spectral line widths. Thirdly, the apparatus sensitivity profile was unknown, making it impossible to directly compare intensity of lines from different positions in a given spectrum. In general, earlier results obtained using this system have not yet provided us with a clear view of the temporal devel- opment of the emission spectrum of spark discharges in the EUV/VUV region. Measurements in air and air-relevant gases and a more extensive analysis of the system are therefore needed.

II. System analysis

To be able to compare line intensities, the obtained spectra have to be corrected using an apparatus sensitivity profile. The fluorescent quantum efficiency of the sodium salicylate scin- tillator used has been assumed constant throughout the measurement range, as suggested by literature [15]. The grating reflectance was also taken into account. It was found that the percentage of light that is reflected by the grating varies from 5% to 89% through the measurement range (see figure 1) [16, 17]. The grating reflectance is assumed to be the dominant mechanism influencing the sensitivity of the system, and is used to correct the obtained intensity of the lines. This finding may explain why the spectra presented in [12] show relatively low intensity for lower wavelengths. It was found before that spectral lines obtained by the system are relatively broad, mostly due to the apparatus profile. This may lead to an overestimation of some lines, especially in the steep parts of the reflectance curve. Therefore, the reflectance curve is convoluted with a triangular profile, Figure 2. ICCD images of the discharge without the Kel-F spacer. yielding a more realistic sensitivity correction. As can be seen, the discharge would sometimes propagate along the In order to reduce statistical variances in the obtained puff valve cap above ((a) and (c)) or the Kel-F surface beneath the needles (b) initially. Later on, a spark discharge between the needles spectra, we investigated discharge behavior for a large number would form (d). Needle–needle distance is approximately 4 mm. of shots using an Andor DH734-25U-03 ICCD camera. Analysis of the obtained images showed the discharge would for geometry details. The close proximity to the discharge often propagate along the puff valve cap or on the dielec- ruled out the use of a metal entrance slit. Kel-F (PCTFE) tric surface beneath the electrodes in its early stage, as can (polychlorotrifluoroethene (CF2CC(F)n) was used to construct be seen in figure 2. Figures 2(a) and (c) show the discharge the entrance slit due to its opacity and high machinability. In propagating upwards from the electrodes and along the puff both cases, the discharge largely avoids the entrance slit to valve cap. Figure 2(b) shows the discharge propagating along the spectrometer chamber, which is located directly between the Kel-F surface beneath the electrodes, see reference [12] the needles. In a later stage, after ~50 ns, the discharge would

2 J. Phys. D: Appl. Phys. 49 (2016) 035201 D Trienekens et al

0 -0 -2 0 -2 -2 0 -4 -2 0 -6 -2 120 nm

Corrected PMT signal (AU) -8 107 nm ) 0 (V -2 0 100 200 300 400 500 600 PMT Time (ns) U 0

-2 Figure 4. Averaged PMT signal for 107 and 120 nm. The signal is 0 corrected by dividing by the grating reflectance percentage. -2 An example of two averaged PMT signals that are cor- 0 rected for grating reflectance is shown in figure 4. -2 In order to reduce the effect of statistical variance, we 0 reject shots where the time difference between the start of the -2 PMT signal and the voltage pulse is larger than 50 ns. Also, shots were the PMT signal stays below a threshold value of 0 1 V are taken as zero to suppress noise. -2 Combined, the improvement made on the setup and the 0 100 200 300 400 500 acquisition of an apparatus sensitivity profile is expected to Time (ns) result in more reliable obtained spectra.

Figure 3. Example of PMT signals for air at 130 nm. Shown are III. Experiments 10 shots over which averaging takes place. Average correlation between two shots is 67%. Spectra in the 60–160 nm region were obtained for air, O2 and ultra high purity (uhp) N2. All spectra were averaged over 10 propagate between the needle electrodes. This behavior how- shots and have a spectral resolution of 0.5 nm. Background ever did not always occur and was perceived as stochastic in measurements were performed using the method described in nature. [12] and averaged over 5 shots. This results in a total of 3015 Due to this stochasticity in the early stage of the discharge, shots per spectrum. Although the exact pressure is unknown, it is obvious this behavior is a cause of statistical variance it was shown in [12] that for a high voltage pulse delay of in the captured signal. In order to increase reproducibility of 350 μs (with respect tot the puff valve trigger) a pressure of the discharge, we placed a ~2 mm Kel-F spacer underneath 200–500 Torr is to be expected. This value was derived by the needles. This forces the discharge along the spacer surface comparing the discharge morphology at varying high voltage (and across the entrance slit), greatly reducing the statistical pulse delays to the discharge morphology at varying static variance observed before. pressures. Figure 3 shows the obtained PMT signals for 10 shots at Obtained spectra are shown in 50 ns timesteps. Although 130 nm, all taken under the same conditions and with the Kel-F severe artificial broadening of the lines still complicates inter- spacer beneath the needles. Statistical variation was greatly pretation of the data, spectral content can clearly be distin- reduced by the introduction of the Kel-F spacer. However, guished. The comparison between air, uhp N2 and O2 clearly as can be seen, a significant amount of variation still exists shows a difference in the 90–110 nm region, especially in between subsequent shots, most likely due to the stochastic the first ~250 ns of the discharge. Moreover, this range is nature of the discharge. No discharge took place at the last known to correspond to strong emission from the b1, b′1 and shot, therefore it is neglected in the averaging process. Note c′4 singlet states of N2 (i.e. the Birge-Hopfield I and II, and there is also some jitter in the start of the PMT signal, a result Carroll-Yoshino bands) [10, 11]. Note that for wavelengths of of discharge inception jitter. This was corrected for when below 102.7 nm direct photoionization of the oxygen mol- averaging over the 10 shots. ecule is possible. Photoionization is always assumed to play In general, we found that correlation between shots was an important role in the early stages of discharge inception lower for lower wavelengths. This is most likely a result of the and propagation [1–9, 18–20]. Molecular nitrogen, having a increased necessity of background corrections and the lower higher ionization potential than molecular oxygen, is capable grating reflectance, leading to relatively higher noise. of emitting photons with the energy required for this process

3 J. Phys. D: Appl. Phys. 49 (2016) 035201 D Trienekens et al

O 0-50 ns 2 Experiment Air Spectraplot uhpN2 50-100 ns Intensity (AU) 100-150ns

90 100 110 120 130 140 150 160 150-200ns Wavelength (nm)

Figure 6. Comparison between the experimentally observed spectrum for air (100–150 ns) and the theoretical spectrum calculated using Spectraplot [13]. 200-250ns AU)

( degradation of the C–Cl bond in the Kel-F spacer under the needles, is the most likely reason for the 107–109 nm peak [23, 24]. A peak at 107.5 nm was observed by Stephens et al in additional experiments on the same setup [25].

Intensity 250-300ns A comparison between the experimentally obtained spectrum in air (90–160 ns) and the theoretical spectrum obtained by Spectraplot is shown in figure 6. An electronic tempera- 300-350ns ture of 1.6 eV was assumed for this stage of the discharge [25]. Note that only atomic and singly ionized nitrogen and oxygen were taken into account. As can be seen, several of the experimentally observed spectral features are present in 350-400ns the theoretical spectrum as well. The difference in intensities may be explained by the fact that we expect our plasma not to be optically thin, an assumption that is used in SpectraPlot. This might for instance cause the lower intensity for the 120.0 nm line in the experimentally obtained spectrum. This 400-450ns is a transition to the ground state of atomic nitrogen, meaning self-absorption should play an important role here. Also, the assumption of local thermodynamic equilibrium (LTE) used in SpectraPlot can be questioned, as was discussed in [25], 90 100110 120 130 140150 160 and may lead to discrepancies between the experimental and Wavelength (nm) theoretical spectrum. The most important difference between the theoretical and experimental spectrum however can be observed in the Figure 5. Spectra of air, uhp N2 and O2 in the 90–160 nm region in 95–105 nm region. Optically thick or not, we observe strong 50 ns timesteps. emission in this region, where atomic and ionic nitrogen and oxygen are not expected to contribute to emission to [21, 22]. Our results show that spectra for air and uhp N2 are this extent [24], as is inferred from the spectrum simulated very similar in the 90–110 nm region in the first 150 ns, as can be seen in the upper three panes of figure 5. Also, it is evident with SpectraPlot. We attribute this emission to N2 singlet transitions, as was stated earlier. This supports our claim that emission for O2 is much lower in this region. This clearly that molecular nitrogen is the most important source of this shows that for air and uhp N2, emission is taking place in the region where direct photoionization of molecular oxygen is photoionization capable emission. Follow-up measurements possible, whereas this emission is much lower for a discharge indeed proved this hypothesis [25]. in O2. This implies emission is indeed mostly due to the pres- ence of nitrogen. IV. Conclusions In the later stage of the discharge, a large peak at ~107–109 nm can be seen in all three spectra. Although the purity of the We have experimentally obtained time-resolved spectra for oxygen is not known exactly, the fact the peak arises in spark discharges in O2, uhp N2 and air in the 90–160 nm range. all three spectra (air, O2 and uhp N2) suggests an external Although the broad apparatus profile complicates line identifica- factor. Emission from Cl, introduced in the discharge gap by tion, our results clearly show significantly higher emission in the

4 J. Phys. D: Appl. Phys. 49 (2016) 035201 D Trienekens et al

95–105 nm range in air and uhp N2 compared to O2 in the early [8] Naidis G V 2006 Plasma Sources Sci. Technol. 15 253–5 stages of discharge formation. We conclude this emission origi- [9] Pancheshnyi S 2015 Plasma Sources Sci. Technol. 24 015023 nates from N singlet transitions. In this range direct photoion- [10] Ajello J, James G, Franklin B and Shemansky D 1989 Phys. 2 Rev. A 40 3524 56 ization of oxygen molecules becomes possible. This leads us to – [11] Heays A N, Ajello J M, Aguilar A, Lewis B R and Gibson S T the conclusion that this photoionization capable emission origi- 2014 Astrophys. J. Suppl. Ser. 211 28 nates from molecular nitrogen, a claim that is further supported [12] Ryberg D, Fierro A, Dickens J and Neuber A 2014 Rev. Sci. by comparison with spectra calculated using SpectraPlot. These Instrum. 85 103109 simulated spectra show atomic and singly ionized nitrogen [13] Fierro A, Laity G and Neuber A 2012 J. Phys. D: Appl. Phys. 45 495202 and oxygen are not expected to contribute heavily to emission [14] Laity G, Neuber A, Fierro A, Dickens J and Hatfield L 2011 capable of directly ionizing molecular oxygen. IEEE Trans. Dielectr. Electr. Insul. 18 946–53 [15] Samson J A R 1967 Techniques of Vacuum Ultraviolet Spectroscopy (Lincoln, NE: Pied) Acknowledgments [16] Princeton Instruments Spectroscopy Group 2015 Acton VM Series, Rev. N1, Available www.princetoninstruments. One of the authors (D T) was supported by project 12119 of com/Uploads/Princeton/Documents/Datasheets/ Dutch Technology Foundation STW, partly funded by ABB. Princeton_Instruments_Acton_VM-502_VM-504_rev_ N1_11-21-13.pdf [17] Fischer B 2014 private communication References [18] Pancheshnyi S V, Starikovskaia S M and Starikovskii A 2001 J. Phys. D: Appl. Phys. 34 105 [19] Liu N and Pasko V P 2004 J. Geophys. Res. 109 A04301 [1] Chen S, Heijmans L C J, Zeng R, Nijdam S and Ebert U 2015 [20] Penney G W and Hummert G T 1970 J. Appl. Phys. 41 572 J. Phys. D: Appl. Phys. 48 175201 [21] Zipf E C and Mclaughlin R W 1978 Planet. Space Sci. [2] Nijdam S, Van de Wetering F M J H, Blanc R, Van 26 449–62 Veldhuizen E M and Ebert U 2010 J. Phys. D: Appl. Phys. [22] Morgan H D 1983 J. Chem. Phys. 78 1747 43 145204 [23] Radziemski L J Jr and Kaufman V 1974 J. Opt. Soc. Am. [3] Nudnova M M and Starikovskii A Y 2008 J. Phys. D: Appl. 64 366 Phys. 41 234003 [24] Kramida A, Ralchenko Y, Reader J and NIST Atomic Spectra [4] Zhelezniak M B, Mnatsakanian A K H and Sizykh S V 1982 Database Team 2014 NIST Atomic Spectra Database High Temp. Sci. 20 357–62 (version 5.2), [Online] National Institute of Standards and [5] Kulikovsky A A 2000 J. Phys. D: Appl. Phys. 33 1514–24 Technology, Gaithersburg, MD Available: http://physics. [6] Wormeester G, Pancheshnyi S, Luque A, Nijdam S and nist.gov/asd Ebert U 2010 J. Phys. D: Appl. Phys. 43 505201 [25] Stephens J, Fierro A, Beeson S, Laity G, Trienekens D, [7] Bourdon A, Pasko V P, Liu N Y, Celestin S, Segur P and Dickens J and Neuber A 2015 Plasma Sources Sci. Technol. Marode E 2007 Plasma Sources Sci. Technol. 16 656 submitted