**3. Results**

In order to avoid doubling, some more general results that were already described and discussed in the accompanying paper will not be repeated here. That relates to electrical waveforms, temporal evolution of pressure and plasma composition in the nozzle, and video observation by HSC with and without filtering. Additionally, only selected moments from the overview video spectroscopy were shown that are mandatory for the discussion of molecule emission and absorption. It should be noted that, for easier comparison, all points in time are given with respect to the current zero crossing.

First, experiments in the MCB (setup (b)) with the sine-like current up to 5.3 kA are considered. The arc voltage was around 200 V (after peak caused by the explosion of ignition wire) until the arc extinction peak some hundred μs before CZ. The total pressure in the nozzle started from a filling pressure of 1.0 bar to a maximum of 3.5 bar close to peak current and decreased to about 2.0 bar at current zero. After ignition, an arc discharge in CO2 atmosphere was observed, also containing copper from ignition wire and electrodes. Within the next few hundreds of microseconds, the ablation of the PTFE (C2F4) wall material started to dominate the discharge, blowing the CO2 out of the nozzle. In the following, a long and stable period was observed that was dominated by ablation. Another reversal of flow was found about 2 ms before CZ: With decreasing arc current, the wall ablation and thus the pressure in the nozzle decreased to values below that in the heating volume. Hence, relatively cold gas from the heating volume with a high fraction of CO2 flowed back into the nozzle. In the last ms, only emission from O I was observed, indicating a plasma composition completely dominated by CO2.

### *3.1. Analysis of C*2 *Swan Bands*

An example of a two-dimensional spectrum is shown in Figure 3. It was acquired with setup (b) shortly before peak current (7.3 ms to CZ). On the left side, an image of the HSC observation area (grey scale image) including the OES axis (yellow dashed line) is shown. The vertical axis represents the position along the observation slit in the nozzle, cf. dashed yellow line in the HSC image on the left side; the horizontal dimension is given by the wavelength in the spectral range ∼480–625 nm. The arc discharge was dominated by wall-ablation at that point in time; no emission from copper or oxygen but lines from atomic fluorine F I and atomic and ionic carbon lines C I, C II could be

observed. This radiation was mainly emitted in a broad distribution over the arc cross-section with the highest intensities in central positions, as it is typical for the wall-stabilized arcs with broad and flat temperature profile [9]. However, an additional structure can be recognized with a different lateral distribution: A dense pattern of lines with increasing intensities and numbers towards higher wavelengths with abrupt breaks at positions near 516 and 564 nm, spread over the whole nozzle diameter and partly even with maxima close to the wall. This structure has been attributed to the Swan band system originating from transitions between the electronic states d<sup>3</sup> <sup>Π</sup>*g* and a3 Π*<sup>u</sup>*. Four cases of appearance of Swan bands in the discharge will be presented in the following.

Firstly, the Swan bands occurred at the outer edges of the arc preferably close to the nozzle walls as shown in Figure 3. Generally, this can be regarded as typical behavior for cases of moderate PTFE influence, i.e., when current density is not too high and the temperature close to the wall is rather low, allowing the existence of carbon dimers.

**Figure 3.** (**left**) Photo (grey) of observation window. (**right**) Two-dimensional optical emission spectroscopy frame at 7.3 ms before current zero (CZ).

Secondly, other Swan band pattern ws observed over the full vertical axis of the side-on 2D spectra. The example shown in Figure 4 was acquired with setup (b) about 6 ms before CZ, i.e., shortly before the peak current. A grating of 1800 L/mm was applied to obtain higher spectral resolution. A good agreemen<sup>t</sup> was found of the 1D-spectrum taken in central position with spectra shown by Camacho in OES investigations on plumes produced by laser ablation of graphite targets [18]. The weaker continuum and stronger C II lines compared to [18] hint on rather high plasma temperatures at least in the arc center with higher current density than near to the wall.

Exemplarily, one of the lines of the C2 Swan band near 562.8 nm was analyzed; the carbon ionic line at 566.2 nm was used for comparison, cf. yellow arrows in the 2D spectrum. The side-on radiances are shown in the lower-left part of Figure 4: whereas the ionic line has its maximum in the center, the Swan band emission is spread more homogeneously over a wide side-on positions between center and 4 mm but has a distinct maximum near to 5 mm, i.e., near to the wall. Since both emissions showed good symmetry in relation to the center, this axis was used for symmetrization and as the central side-on position "0 mm". Then, the radial profile of the emission in the arc can be analyzed by Abel inversion of the side-on radiances. Results are shown in the lower right part of Figure 4. The C II 566.2 nm ionic line is emitted as expected mainly in the center; the emission coefficient decreases to 20% within radial positions of 2 mm. The C2 Swan band, however, has a sharp peak of less than 2 mm FWHM with a maximum emission coefficient below 1 mm to the wall. It should be noted that although the nozzle diameter is 12 mm some intensity was detected at side-on position above 6 mm due to experimental limitations like quartz plate connection and refraction at the windows. The algorithm of inverse Abel transformation is limited in case of very low emission from the central position, therefore

the C2 emission coefficient in the center is not plotted for values below 10% (radial positions < 3.5 mm). To summarize even in the case of Figure 4 the Swan bands are emitted only in a thin sheath at the wall.

**Figure 4.** (**Top**): 2D spectrum over the full arc cross-section (upper part) together with the corresponding 1D spectrum from central arc position (lower part) in the spectral range around the C2 Swan band head at 563.7 nm-Swan band on the left side and C II lines on the right side. (**Bottom left**): Spectrally integrated line intensities of the carbon ion line C II 566.2 nm and of the C2 Swan band line at 562.8 nm, labeled by yellow arrows. (**Bottom right**): Inverse Abel transformation carried out for these intensities to reveal the origin of emission.

A third example of the occurrence of C2 Swan bands is shown in Figure 5. It was only observed with setup (a) providing higher pressure and strong wall ablation due to peak currents of 8 kA (100 Hz). With the single long PTFE nozzle the current was not switched off and multiple current zero transitions were observed. Except the ignition phase and few hundred μs around CZ, all spectra are dominated by pronounced emission of the Swan bands. The band heads of the Swan bands are located at 473.7 nm, 516.5 nm, and 563.6 nm; they are indicated by red arrows in the two-dimensional spectrum in the upper part of Figure 5. The wavelength range chosen here does not include the band head at 438.2 nm but also contains the C II lines at 564.06 nm, 564.81 nm, and 566.26 nm as well as the C I atomic lines at 476.2 nm, 477.0 nm, 493.2 nm, 505.2 nm, and 538.0 nm. Weak or non-visible ionic and atomic carbon lines in comparison with the Swan bands give the first hint to rather low temperatures in the

center of the arc. Furthermore, it was observed that the occurrence of carbon lines drastically changes approaching current zero. Within some 100 μs, first the ionic and then the atomic lines disappear; after CZ they reappear in reversed order. In fact, disappearance of the atomic lines cannot be observed for first and second, but for the third CZ crossing.

**Figure 5.** (**top**) Spectrum acquired 300 μs before CZ with setup (a) and 8 kA peak current. It is completely dominated by molecular radiation of C2 Swan bands (band heads labeled by arrows). (**bottom**) Spectrally integrated line intensities (**left**) and emission coefficient obtained from Abel inversion (**right**) of the C2 Swan band emission at 562 nm.

The Swan band pattern has a much higher intensity in the central position, although the emission is extended to the side-on positions of the nozzle wall. The origin of emission is further analyzed using the band head around 563 nm as shown in the lower part of Figure 5. The side-on profile (left) and the emission coefficient obtained by inverse Abel transformation (right) reveal a different occurrence in comparison to the plasma in Figure 4. The Swan bands were emitted with the highest intensities in the center of the arc, continuously decreasing towards the nozzle walls. Thus, it can follow that the arc plasma is completely dominated by the PTFE material and it is characterized by rather low temperatures even in the arc center. It should be mentioned that this third case of Swan band appearance is the most extreme and could not be achieved with setup (b) with two nozzles separated by the heating channel even when the peak current was doubled to 10 kA.

The fourth example was typical for setup (b): Here, the Swan bands can be recognized by their characteristic absorption pattern at currents >4 kA, cf. example in Figure 6. The spectrum was taken shortly after the current maximum (4.7 kA, 4.6 ms before CZ). Emission from the hot plasma in the arc center served as an internal background radiator that was absorbed by the much cooler carbon dimers near to the nozzle wall. As in Figure 5, the Swan band heads are indicated by red arrows. Additionally to the absorption pattern, some emission lines can be found. These are all ionic carbon lines, e.g., at C II 564.06 nm, 564.81 nm, and 566.26 nm. They are preferably emitted in the arc center, i.e., at higher temperatures. In the third and fourth cases with most intense ablation and material transport towards the electrodes, which is probably the reason why no copper lines were observed at the slit position. Fluorine lines were not available in that spectral range.

**Figure 6.** Spectrum acquired with setup (b) during high-current phase. The typical structure of the C2 Swan bands was found as an absorption pattern with the plasma in the arc center serving as background radiator. The band heads are labeled by arrows.

### *3.2. Optical Absorption Spectroscopy around Current Zero*

The phase of current zero-crossing is of the highest importance for an understanding of the switch-off process and the dielectric recovery of the electrode gap region. Hence it is of special interest to extend experimental knowledge as close as possible to CZ and even beyond. However, even tapping the full potential of optical emission spectroscopy, e.g., by application of OES with intensified cameras as described above, the analysis based on optical emission spectroscopy is limited to times about 10 μs before CZ due to reduced energy input by the arc [9]. Consequently, absorption techniques were required for further investigation of the current zero-crossing and the immediately following time period. Since the majority of atoms are in the ground state in case of the lower temperatures near CZ, it will be necessary to mainly analyze lines going to ground or very low levels by optical absorption spectroscopy (OAS). However, most of the relevant lines are in deep UV regions far below 300 nm. From the experimental point it is extremely demanding to investigate such radiation under switching-relevant conditions since all components of the setup including high-pressure vessel and model circuit chamber have to be transparent for these wavelengths. With the actual setup even resonant lines that might be more suitable could not be detected due to limited spectral sensitivity of the cameras such as C I at 296 nm or Cu I at 324 nm and 327 nm. The few resonant lines in the available wavelength range above 340 nm have very low transition probabilities, e.g., C I at 462 nm and O I at 630 nm. However, it might be possible that some lines might be occupied around CZ and could be detected by OAS that are characterized by relatively low energy levels and medium transition probabilities, e.g., Cu I at 510 nm, 570 nm, and 578 nm with E*u* = 1.39 eV and 1.64 eV or O I at 557 nm with E*u* = 1.26 eV. Additionally, molecules are possible candidates for absorption, e.g., the C2 molecule since its Swan bands were observed in emission until few 100 μs before CZ and even in absorption during the high-current phase as shown above.

Broadband optical absorption spectroscopy (OAS) was carried out around CZ using the pulsed high-intensity xenon lamp as an external wide-band background illumination. Two examples are shown in Figure 7, comprising the wavelength ranges about 440–600 nm (left) and 640–800 nm (right column). In the upper panel only the emission from the Xe lamp is given, i.e., through model circuit breaker including all windows but without discharge. Broadband continuum can be seen in both spectral ranges. It should be noted that the spectra are not calibrated concerning absolute intensity. The edges of the nozzle can be recognized by sharp transitions from the bright stripe (white/red) from the Xe lamp illumination and the dark regions (green). The spatial distribution within the nozzle slit is quite homogeneous, showing smooth illumination by the background source. In the middle

panel, the OES spectra were taken at CZ (exposure time 50 μs) with the Xe lamp continuum passing the remainder of arc discharge in the nozzle. Patterns of horizontal stripes were sometimes observed. Similar experiments using HSC instead of ICCD camera revealed that these stripes did not change from one video frame to the next. Thus, it was reasonable to assume a deposition on the quartz glass sealing the slits, e.g., by particles. In the left spectrum a certain structure was found below 500 nm whereas the right spectrum did not show any peculiarities. From spectra in top and middle panel a transmission could be calculated, cf. lower panel of Figure 7. For an improvement of the signal-to-noise ratio and better visualization of the intensity ratio, spatial integration was carried out for the determination of the transmission. It revealed that there was only one significant absorption peak around 493 nm. This absorption was clearly accorded to the CuF molecule as will be discussed below. Beside this CuF peak, no hint on any absorbing lines or other features could be detected around current zero, even with the intensified camera with high sensitivity and dynamic range. Even the C2 Swan bands could not be observed before or after CZ in OAS with the Xe lamp as background radiator although they were detectable in OES up to a few hundred μs before CZ. Moreover, a closer look onto the emission spectra (cf. Figure 3) showed that the CuF absorption at 493 nm could also be found during the high-current phase of the discharge, though this effect was rather weak compared to the intense line emission. This will be further investigated in Section 3.3.

**Figure 7.** Optical absorption spectroscopy. (**Top**): Overview emission spectra of Xe lamp only. (**Middle**): Xe lamp with discharge of 5 kA peak current and setup (b), acquired at CZ with 50 μs exposure time. (**Bottom**): Spatially integrated transmission calculated from above spectra showing absorption around 500 nm.

Other species for absorption with maximum around 493 nm could be excluded in detailed spectral analysis, including all relevant elements as Cu and W from electrodes, C, O, and F from filling gas and nozzle, and even H as possible contamination. As an example, a prominent candidate might have been the carbon atomic line C I at 493.20 nm, although its lower energy level of 7.7 eV is rather high. However, this line was not detected in emission like other atomic carbon lines with similar upper level of about 10 eV and comparable transition probabilities in the range of several 10<sup>6</sup> s<sup>−</sup>1, e.g., C I 505.21 nm and 538.03 nm (cf. Figure 3). Moreover, these C I lines were still observed in emission 0.7 ms before CZ, while at 493 nm an absorption could be seen even during discharge.

No absorption spectra were found in the literature for the CuF molecules. Thus, in Figure 8, an emission spectrum from Cheon et al. [25] (black curve) was added to the calculated absorption spectrum (dashed blue) for comparison. Considering different experimental conditions and methods, a compelling agreemen<sup>t</sup> was found (OES with higher spectral resolution will be shown below). Basic data of the CuF emission are listed in Table 1.

**Figure 8.** Spectral absorption measured around CZ (dashed blue) compared with CuF emission from [25] (black line).


**Table 1.** Basic data of CuF emission lines [26,27].

In the following, the CuF molecular absorption after current zero should be analyzed in more detail. A series of time-resolved spectra is shown in the upper part of Figure 9. The transmission was calculated based on the division of the measured spectra (plasma plus xenon lamp) by a xenon lamp spectrum without discharge. A higher spectral resolution was obtained by the grating with 1800 L/mm. Thus, the peak structure including maxima at 493.2 nm, 492.7 nm, and 493.2 nm can be clearly recognized in agreemen<sup>t</sup> with the emission spectrum of CuF molecules from [25] shown in Figure 8. The overlaying periodic structure is not caused by the plasma in the nozzle since the same structure was also observed for the xenon lamp itself. Probably it was caused by an interference effect of glass plates in the detector. The background intensity increases although the xenon lamp is in its plateau phase.

For a quantization of the temporal evolution of the absorption, the area under the curve (AUC) was determined from the difference between the "background" (average of levels extracted at wavelengths aside the CuF absorption, i.e., around 490 nm and 496 nm) and the transmission, integrated over the spectral range. The AUC is plotted as the curve with black open circles in the lower part of Figure 9. Additionally, a normalization of the AUC was carried out by division by the (temporally increasing) intensity of the background signal. The normalized AUC is represented by the curve with red filled squares. The spectrum 0.2 ms after CZ was chosen as the starting point and the according value was set to 1 for better comparison. Within half a millisecond, the AUC decreases by 60%. The decrease of the normalized AUC is even more significant, namely down to 20% (factor of 5). Few other shots that were carried out confirmed this result. However, due to the exponential nature of this decrease, the absolute values are sensitive to the starting point. Summarizing it can be stated that the CuF absorption and thus, also the CuF density decreases after current zero on a timescale of several hundreds of microseconds.

**Figure 9.** (**top**) Series of detail transmission spectra after current zero. (**bottom**) Temporal development of absorption peak at 493 nm calculated as area under curve (AUC, open circles) and AUC normalized by background intensity (filled squares).

### *3.3. CuF during the High-Current Phase*

As mentioned above, overview OES spectra in Figure 3 gave hint on a possible absorption of the CuF molecule even during the arc discharge though the effect might be considerably lower in absolute intensity than the atomic line emission. Thus, the spectral range around the 493 nm-peak was investigated with video OES of higher resolution (grating 1800 L/mm instead of 150 L/mm, exposure time 98 μs). An example is shown in Figure 10 acquired with setup (b) and 5.3 kA peak current.

**Figure 10.** Temporal evolution of the spectral range around 493 nm with CuF absorption (Time to CZ is from (**bottom**) to (**top**), the color scale reaches from black for lowest emission to blue, green, and red for highest intensities).

In the upper part, all about 110 optical emission spectra (spatially integrated) are plotted line by line vs. time to CZ, forming a two-dimensional contour plot. In the lower part of Figure 10 a selection of four instants of time with characteristic spectral features are shown. Additionally, several atomic and ionic lines were labeled that helped for fine adjustment and control of exact wavelength positions. The ignition phase the spectra were dominated by atomic and ionic copper line emission. At first, i.e., 10.7 ms before current zero, an atomic copper line Cu I 486.61 nm was observed which was followed by several ionic copper lines (cf. spectrum 10.2 ms before CZ). As known from overview OES spectra, the ablation of PTFE was usually initialized about 1 ms after ignition. In the spectral range of Figure 10 no atomic fluorine lines can be observed. However, the occurrence of the hydrogen line <sup>H</sup>*β* at 486.13 nm (starting about 9 ms before CZ) can be regarded as an early sign of nozzle ablation, probably caused by a thin remaining water film on the nozzle surface. Thus, one may expect the occurrence

of CH molecules. That should be detectable by the CH(A-X) band with maxima around 430 nm. However, we did not find such an absorption pattern even in a detailed analysis of the corresponding spectral range. Within several hundred microseconds the spectrum is changed from being dominated by copper lines (more probable originating from the W–Cu electrodes than from the ignition wire) to being ablation-dominated. CuF absorption pattern is observed during the full high-current phase, at least from about 8 to 2 ms before CZ. In the spectral range below 500 nm this is basically visible by a broad continuum, starting about 2–3 ms after ignition (or 8 ms before CZ) and lasting at least until about 2 ms before CZ. As can be clearly seen from the top spectrum in Figure 10 the characteristic absorption pattern of CuF can be observed for these 5–6 ms, i.e., during the whole high-current phase of the discharge. Comparable to the case of absorption of the C2 Swan bands, the CuF absorption is enabled by background continuum from the arc plasma. The observed temporal fluctuation of the intensity has been found to be caused mainly by changes of background intensity, e.g., fluctuation of transmission or reflection due to droplets. Hence, even during the high-current phase a considerable amount of absorbing CuF molecules must be existent in the plasma at nozzle position, i.e., 8 mm away from the electrode. This might be unexpected but leads to the conclusion that the gas flow into the heating channel is strong enough to pull electrode material into the region of optical investigation.
