Properties and characteristics of the nanosecond discharge developing at the water–air interface: tracking evolution from a diffused streamer to a spark filament

The characteristics of nanosecond discharge propagating along the water–air interface in a unique dielectric-barrier discharge (DBD)-like configuration with coplanar electrodes submerged in deionised (DI)/tap water are studied by combining ultrafast imaging and emission spectra with electrical characteristics. Time-resolved images provide a clear signature of streamer channels excited on the water surface at either side of the blade (insulated plastic separating the anode/cathode) and propagating perpendicularly away from it towards the anode/cathode with different velocities. Later on, the streamer channels convert into a few discrete and bright discharge channels due to ionisation instability (spark phase). There is no distinctive dependence on water conductivity in the streamer phase, as the optical emission spectroscopy and images of discharges only showed an increase of the luminosity and no significant changes in morphology. However, in the spark phase, more numerous, brighter, and thicker filaments form in tap water. The time-resolved emission spectra reveal the dominance of the first and second positive system of N2 molecular bands in the streamer phase, followed by the appearance of atomic lines of hydrogen, nitrogen, and oxygen in the spark phase. The emission spectra are utilised to estimate several important parameters (gas temperature, reduced electric field ( E/N ), and electron density ( ne )). The streamer phase is characterised by a low gas temperature and a peak E/N amplitude between 700 and 850 Td. On the other hand, the subsequent spark phase is characterised by a gas temperature of ∼400/1100 K and a free electron density up to ne∼1017 –1018 cm−3 in DI/tap water.


Introduction
Electrical discharges in contact with or in liquids have been studied for decades due to their enormous application potential in various fields [1][2][3][4].These discharges can be produced using very different electrode arrangements, DC/AC/pulse power supplies and power density conditions to produce specific discharge regimes and, therefore, plasmas with distinct properties.In the case of an electrical discharge produced by repetitive fast-rise time high-voltage (HV) pulses of nanosecond/microsecond duration, the driving processes responsible for the formation of the plasma phase often remain unclear and require further detailed investigation [5][6][7][8].
A specific category of plasma sources is given by configurations leading to a gas-phase discharge occurring at the gasliquid interface, where the discharge is at least partially in contact with the water surface.The most common case of a discharge that is partially in contact with the water surface is a pin-to-plate electrode geometry with an HV pin electrode located near the water surface against a grounded plate electrode immersed in the liquid [9][10][11].
Hoffer et al [9] investigated the characteristics of long plasma filament produced in needle-to-water surface geometry and propagating along the bounded surface of the water excited by voltage pulses of microsecond duration.Time and space-resolved characteristics of plasma-induced emission (PIE) have been performed and reveal that the early emission is dominated by second positive system (SPS) emission of molecular nitrogen, which is replaced by the visible and near-infrared atomic line emission of oxygen and hydrogen.The plasma channel length was found highly dependent on the width of the voltage pulse and the authors concluded that the production of short-length discharges produced by the shorter voltage pulses is more efficient when it comes to applications.Akishev et al [10] investigated the propagation characteristics of positive streamers produced in pin-to-plate geometry on the surface of tap water using optical-electrical characteristics and numerical simulations.They have found that the streamer length and diameter are highly dependent upon the depth of the water basin and strongly influence the electric field in the air around the streamer.Hamdan et al [11] investigated nanosecond discharges produced by negative pulse voltage in air which were in contact with water where tungsten pin acts as anode and steel plate acts as cathode.The time-resolved study reveals the transition of the homogeneous disc-like ionisation front into discharge channels with increasing HV amplitude.The discharge characteristics are distinctive from the discharge initiated by positive pulsed voltage, indicating the dynamics is highly dependent upon the space charge formation.
The disadvantage of all the configurations based on a powered metallic electrode located above the water surface is the possible erosion of the electrode surface with subsequent contamination of the water.This point has recently been addressed by designing a DBD-like configuration in which all the electrodes are immersed in a liquid so that there is no direct contact of the metal surfaces with the active plasma [12].In this geometry, the discharge propagates over the water surface along the air-water interface and delivers nitrogen/oxygen radicals to the water surface, leading to effective nitrogen fixation in the water.A preliminary analysis of the plasma-activated water (PAW) resulting from such a discharge revealed a significant dependence of dissolved products on the discharge repetition frequency and the initial conductivity of the water.
We recently [13] demonstrated that the discharge in the case of deionised (DI) water develops as a bidirectional diffuse ionising wave followed by the formation of highly luminous discharge channels.In this work, we investigate, both in DI and tap water, the detailed characteristics of the discharge developing at the water-air interface in this unique DBD-like electrode geometry by combining electrical characteristics with ultrafast imaging and time-resolved emission spectroscopy.
This paper is structured as follows.Experimental setup and methods in the section 2 include the details of the used reactor, a description of emission spectroscopy and imaging techniques and information about data processing and analysis.The results in section 3 incorporate the measured electrical characteristics, four-channel imaging and emission spectra of the discharge on the surface of DI and tap water.The plasma parameters obtained from the imaging, such as filament thickness, rotational temperature, electric field determination, plasma density and temperature, estimated from the PIE spectra, are presented and discussed in sections 4 and 5.

Experimental setup
The experimental setup used in this work is shown in figure 1.The reactor consists of a 3D-printed plastic vessel containing inlet/outlet ports that allow water to circulate and three compartments fully separated by plastic blades.Each compartment contains a flat brass electrode attached to the bottom.The middle electrode is at HV while the two side electrodes are grounded to prevent HV transfer to the water circuit pipes.The DI/tap water is fed into the reactor from a reservoir through a needle valve to maintain the flow rate of the fluid (∼20 ml min −1 ).The water flows from the left to the right and flows out after passing through the glass felt, which maintains the necessary stable water level in the reactor (approx. 1 mm above the edge of the insulating plastic blades).More details about the reactor and its geometry can be found in [12].Here, we have performed the experiments in tap water with an initial conductivity of ∼450 µS cm −1 as well as in the DI water with an initial conductivity of ∼1.2 µS cm −1 .
The HV electrode is connected to a pulsed power supply (Suematsu Electronics MPC3010S-50PL, 0-30 kV) via a 3 nF capacitor to prevent the supply from DC current.The repetition frequency of 2 Hz was used for four-channel imaging measurements, while 20 Hz was used for emission spectroscopy (due to requirements on higher statistics).The pulse power is bipolar, where a positive cycle comes first, followed A Tektronix DPO 5204B oscilloscope was used to record the voltage and current, and Tektronix TDS 3054 C oscilloscope to measure the photomultiplier (PMT) waveforms and to control the timing of the discharge with respect to the detectors.The PMT (Photek PMT210) was used to measure the time evolution of PIE.The PMT signal was also fed to a forming circuit (Schmitt trigger) that produces transistor-transistor logic which synchronises diagnostic equipment (intensified charge couple device (ICCD) and a four-channel framing camera).To analyse PIE in the spectral range 200-900 nm, we have mounted the optical fibre just above the edge of the right insulating blade to collect the light from adjacent areas (see figure 1) and thus collected the space-integrated time-resolved emission using ICCD spectrometer.The optical bandpass filters were also placed in front of the slit opening to select specific wavelength interval.We used the Jobin-Yvon iHR-320 spectrometer (typical input slit width of 100 µm) with dispersion gratings (UV 150 G mm −1 , NIR 150 G mm −1 to produce overview spectra at low resolution and vis-NIR 1200 G mm −1 to produce high-resolution partial spectra suitable for estimation of plasma parameters).Emission spectra were registered using an Andor DH740i-18U-03 iStar ICCD camera.The wavelength is calibrated using an Hg-Ar pen lamp (Oriel No. 6035), and intensity is corrected using a deuterium-tungsten halogen UV-Vis-NIR light source (Ocean Optics DH-2000).All emission spectra reported in this work were obtained by combining the accumulation mode with the kinetic series technique, where the PIE from 500 triggered events is accumulated to obtain a single spectrum at a certain time interval (defined by the position of the multi-channel plate (MCP) gate relative to the onset of the discharge).Furthermore, to follow the evolution of discharge morphology during a single event, we used a four-channel XXRapidFrame camera (Stanford Computer Optics, Inc.).It is essentially a four-channel ICCD imaging system based on UV-extended mirror-based image-splitting optics and four Picos ICCD cameras.The image splitter splits the arriving flux of photons equally among the four channels (Ch1-Ch4), each registered by one ICCD.This device enables the registration of four/eight synchronised images employing single/double frame mode, all with high-temporal resolution and with spatial resolution 29 µm per pixel.The relative positions of the MCP gates of the four ICCD detectors are controlled using individual internal delay generators of the respective cameras.The ultra-fast imaging is recorded during a single discharge event.

Electrical characteristics
Electro-optical characteristics such as voltage, discharge current, and PMT waveforms are depicted in figure 2. As mentioned before, our power supply is bipolar in nature, and figures 2(a) and (b) shows the discharge characteristics in DI and tap water for both positive and negative polarity.One can clearly see that there are more peaks in the discharge current and PMT waveforms in the case of DI water.The repetitive discharges have a smaller conductive current and less intense plasma emission in comparison to tap water.In the case of tap water, there are only two prominent discharges, i.e. one at positive polarity and the second at negative polarity.Figures 2(c) and (d) shows the time evolution of instantaneous and integral power.There are clearly visible negative pulses in the power waveform in figure 2(c) These pulses result from the load capacitive character.When the voltage at the discharge chamber suddenly decreases, the current sign changes, although the voltage polarity remains unchanged (figure 2(a)).Therefore, there is energy flux back into the HV supply in these instances.Figures 2(e) and (f) zooms the discharge characteristics in DI and tap water up to 600 ns in order to see the shape of the primary HV pulse in more detail.
Figure 2(a) shows the measured waveform of HV pulse, current, and PMT up to the positive pulse when the discharge is initiated in DI water, and figure 2(b) shows the same for the case of tap water.The shape of the voltage pulse depends on the power supply loading, and hence the HV (blue) pulse profile changes from a rectangular-like waveform in the DI water case (low liquid conductivity) to a saw-tooth waveform with exponential decay in the tap water case (high liquid conductivity).
The current waveform (red) also shows a significant difference.In the case of DI water, the liquid electrical characteristics are closer to dielectrics (DBD) with the purely capacitive current.However, in the case of tap water, the conductivity is higher; thus, it draws purely conductive current with a large Figure 3.Time evolution of PMT of second positive system (SPS) (black), first negative system (FNS) (dashed blue), Hα (yellow circles), H β (red triangles) obtained by putting various interference filters (see table 1) measured for (a) DI and (b) tap water.total area and large value as compared to DI water.The delay of the peak in current with the peak of HV is also seen in DI and tap water.Interestingly, the current flow in DI water quickly increased to its peak value at 40 ns as compared to tap water, whose peak value is at 100 ns.This is because, due to the lower conductivity of water, some current flows through the water before the breakdown, and energy is dissipated in the liquid.Hence the rate of current rise is slower, and the difference in the delay in the peak of HV and current is greater in tap water.The PMT profile (black) shows the optical emission (integrated over the wavelength range enabled by the PMT's photocathode) from the plasma, and it also follows a trend similar to the current profiles for both DI and tap water.The emission decays slowly and covers a large area under the curve in tap water.However, the emission from the plasma is completely quenched after 80 ns in DI water.In DI water, the discharge is again initiated at ∼450 ns due to a sharp fall of HV, which is not seen in tap water.
Figures 3(a) and (b) shows the emission profiles of SPS(0,0), first negative system FNS (0,0), H α , H β measured by PMT in DI and tap water, respectively, by placing an interference filter (table 1) in front of PMT.The details of the bandwidth and central wavelength of the interference filter used are provided in table 1.Both figures show that SPS and FNS initially dominate, trailed by the atomic lines of hydrogen.Interestingly, the atomic line emission of H α and H β peaks early in tap water compared to DI.The early quenching of SPS and FNS emissions, followed by atomic lines of hydrogen, which also cease before 70 ns, are detected in DI water.However, a slow fall of molecular band emission is observed in tap water.The emission of atomic hydrogen lines, even after 100 ns, is seen in tap water.It is also seen that emission from H α and H β regions are also observed at the initial time in both DI and tap water, and this emission comes due to overlapping molecular emissions.The emission from the first peak of H α trace is due to the contribution from first positive system FPS (7,4) (654.4 nm) and FPS(6,3) (662.4),whereas in H β it comes from SPS(2,8) (481.4 nm) and SPS(1,7) (491.7 nm).

Four-channel imaging
The time-resolved imaging using a four-channel ICCD XXRapidFrame camera provides important information about the discharge onset and evolution e.g.expansion velocity, dimensions of the plasma structures, discharge channels branching, dynamics of the filaments, etc [14].Here, the area of interest of the captured images (view from above) covers only part of the water surface centred around the right plastic blade (see figure 1).The most important obtained results are summarised in figures 4-9.The location of the blade separating the HV and grounded electrodes is indicated by the horizontal dash-dot yellow line in the top left image.The upper half is HV electrode area and the lower part is the grounded electrode area.Figure 4 shows a series of four consecutive frames captured during the first few ns of HV pulse in DI water.Four successive images taken during one event allow monitoring of the dynamics of discharge formation.They were obtained using identical multi-channel plate (MCP) gate width (2 ns) for all channels, and delayed with respect to the onset of the HV pulse according to the time scale depicted in figure 4(a).Events 1-3 were selected from all captured single discharge events to illustrate the basic PIE characteristics during the first tens of nanoseconds.In the case of 'Event 1', the MCP gate of channel 1 was set just before the onset of the voltage pulse.Figure 5 shows a series of four consecutive   frames during the first few tens of ns in tap water.Here the gate width was set to 1 ns.In principle, we can observe very similar dynamics and morphology of the discharge compared to the DI case.The only observable difference is the higher intensity of PIE, which allowed the use of a shorter MCP gate time at a similar signal-to-noise ratio.In the case of 'Event 3', the MCP gate time of Ch1 coincided with the onset of the voltage pulse and we were able to capture the very early PIE which is visible in figure 5 as a small light island above the blade (indicated by the vertical yellow arrow).Due to the known time shift between Ch1 and Ch2, we can easily estimate the average propagation velocity of both luminous fronts.The estimated expansion velocity of luminous fronts in the anode area is ∼6 × 10 5 m s −1 , and nearby to the cathode area is ∼9 × 10 5 m s −1 The cathode-directed ionisation wave propagates faster than anode-directed wave, which has also been reported earlier in streamer discharges in air or liquidair interface [15].
Figures 6 and 7 show a kinetic series made of Ch2 images selected from a set of registered equivalent events similar to those shown in figures 4 and 5 (500 registered events in total for both DI and tap water).The image in the upper row captures the first light from the discharge, and the corresponding MCP delay is denoted as T 0 .It shows the initiation of plasma emission with streamer channels (having more ionisation fronts) present on both sides of the blade.The upper part near to anode area is more intense and broader than the cathode part.The horizontal asymmetry in luminosity could be due to the uneven height of the liquid layer above the blade.This could be due to structural imperfection of the blade surface or vibration of the liquid layer caused by the periodic discharge.The images in the middle row demonstrate the expansion of the streamer channels and their conversion into discrete filaments.Next, (T 0 + 6 ns), an anode-directed wave develops and starts to expand away from the blade (in a perpendicular direction), forming densely packed parallel filaments.A cathode-directed wave is darker and with bright spots at its front.
After that, the expansion of ionisation fronts stops (T 0 + 12 ns and T 0 + 18 ns), and the streamer channels are converted into thin filaments.Later, these thin filaments become more intense and distinct, with diffuse-like structures at both ends.The images in the last row show the formation of prominent discrete filaments.It is clear from the figure that the emission becomes more intense after (T 0 + 30 ns), which is the signature of the transition of streamer to spark phase.The spectrometric characteristics evidencing the nature of both phases will be described in more detail in section 3.3.The emergence of distinct patterns and growing intensity of PIE in the filaments could be the reason for ionisation instability [16].
In the case of tap water, the appearance of discharge and its expansion is similar to DI water (see figure 7).Again we find here multiple streamer channels near the blade, which expand away from the blade with different velocities in opposite directions.The only difference is the intensity of light emission, size of filaments, and number of prominent filaments at a later stage (T 0 +30 ns).The filaments are more intense and thicker in the case of tap water.There are only two prominent filaments that survived to a later stage in DI water, compared to about ten prominent filaments developing in tap water.The whole mechanism can be linked to the conductivity of water as it depends upon the propagation of discharges along the water's surface.The higher conductivity of the water enables larger discharge currents, which results in brighter discharge channels and thicker and shorter branches.
To inspect the characteristics of the discharge on a time scale covering the entire HV pulse (positive and negative peaks), we used a double-frame mode allowing the acquisition of two frames from one ICCD channel per trigger, i.e. a  total of eight synchronised frames during one event, as shown in figures 8 and 9. Figure 8 shows the snapshots of discharge evolution up to 4 µs when the discharge is initiated in DI water.The first frame captures the discharge during the positive HV; thus, the integration time is set to 150 ns, and each channel is delayed to 150 ns.The second frame captures the images from all the discharges created at later times, and hence the integration time and delay from each other are different for each channel as specified in the figure.As already mentioned, at initial times up to 30 ns, there is no distinctive difference between the discharge initiation, expansion, and filamentation in tap and DI water.The only difference (after 30 ns) is that the discharge in the case of DI water is less intense, with the formation of thin and less prominent filaments.However, at later times the discharge shows different dynamics in tap and DI water due to the different shapes of the HV profile.It has already been shown that with DI water, the shape of the HV profile is rectangular with a sharp fall, which produces a capacitive current profile.This effect can be seen clearly in the first frame of the images, where the ionisation fronts excited by HV quickly disappear at 450 ns.Also, with DI water, due to the pulse rectangular shape, the plasma discharges are created every time HV falls from peak to some lower value, which can also be seen from the peaks in the current and PMT profile at 0.5, 2, and 2.8 µs.The first two images from the second frame show the discharge initiated due to the fall of HV, and the last two images are due to the negative polarity of HV.In our case, the discharge initiated by negative HV polarity shows more current and optical emission.The dependence of the discharge dynamics on the voltage polarity is out of the scope of this paper and will be reviewed in the future.
Figure 9 shows the snapshots of plasma emission in tap water up to 4 µs.The first set of frames captures the emission up to 1 µs from the onset of discharge, the gate integration time is set to 250 ns, and the delay in each channel is also set to 250 ns with respect to each other.The snapshots have been captured to see the entire evolution of discharge during the positive HV phase.It has already been shown that streamer channels excited in the anode area start to develop ionisation fronts, which later expand and become thin filaments with tree-like structures at the bottom.Here also, as can be seen in the first image from the 1st frame of images, the intense thick ionisation fronts start to expand and become thicker at 500 ns (second image from top row) and turn into thin filaments with more branching at the top and bottom.The filaments become thin and less intense at 750 ns (third image from the first row) and prevail up to 1 µs shown in the last snapshots from the first row.The 2nd frame of images shows the disappearance of these tree-like tiny filaments, and only less intense ionisation fronts exist, which completely disappear as time progresses.In tap water, there is no secondary discharge (discharge other than the peak of HV) as the profile of HV is exponential; no additional discharge is created at later times up to 3 µs except the discharge created from the peak of HV, whereas in the case of DI water, there are four discharges created during the positive HV pulse due to rectangular shape of HV, which creates plasma discharges every time HV falls from peak to some lower value.

Emission spectroscopy
We have mapped the optical emission spectra to obtain the characteristics of the discharge phases revealed in section 3.2 and estimate important parameters such as gas temperature, reduced electric field (E/N), electron density (n e ) and electron temperature (T e ).The kinetic series of PIE has been recorded, starting from the onset of discharge up to 2 µs using different gate widths and time steps depending upon the signal intensity.The gate width was set to 2 ns and time step 2 ns to collect the emission up to 40 ns, gate width 10 ns and time step 10 ns to collect up to 200 ns, and gate width 100 ns and time step 100 ns to collected the emission from 100 to 2000 ns (each individual Figure 10.Overview of time evolution of optical emission spectra ranging 300-500 nm and 550-900 nm recorded for plasma discharge in (a) DI water and (b) tap water, respectively.Here, the grating used 150 lines mm −1 .The development of the molecular nitrogen band for the first tens of ns and the appearance of neutral atomic lines of hydrogen, oxygen, and nitrogen in the emission spectra shows the clear transition of the streamer phase to the spark phase transition as time progresses.The spectra were accumulated for 500 pulses in order to increase the signal-to-noise ratio.
spectrum sampled for a given MCP gate was obtained as an average over 500 discharge events to obtain sufficient signalto-noise ratio).
Figure 10 shows the time-resolved PIE from 300 to 500 nm and 550 to 900 nm wavelength range, respectively, when the discharge is initiated for the case of DI (a) and tap water (b).Here, we are showing the background subtracted normalised intensity to visualise all important emission lines and bands.Figure 10(a) shows that spectra are dominated by the SPS of nitrogen N 2 (C 3 Π u → B 3 Π g ) in the range of 350-450 nm for the first 20 ns.The FPS of nitrogen N 2 (B 3 Π g → A 3 Σ + u ) molecular bands is dominant in the range of 600-800 nm.The other atomic lines, such as H α (656.2) and O I (777.3nm), N I (746.8nm) are also observed in the range of 650-790 nm after 80 ns.These atomic lines survive till 2 µs.The transition from spectra characterised by intense molecular bands of nitrogen to spectra with dominating atomic lines of hydrogen, oxygen and nitrogen indicates the transition of the discharge from the cold streamer phase to the hot spark phase with a significantly increased degree of ionisation.In the case of tap water, similar to imaging, the dynamics of plasma are almost identical at earlier times, i.e. the PIE spectra (see figure 10(b)) also show the domination of the nitrogen SPS system at an earlier stage but with more intense bands.The atomic lines of H α , O I , N I are also observed in the 650-900 nm range after 50 ns (line intensities in tap water are approximately 100 times stronger compared to DI).However, H β (484.3 nm) and H γ (434 nm) are also observed in the case of tap water, which is not seen in DI water.When we compare the absolute values of the peak intensity at time interval 50-60 ns, we observe that the peak intensity is approximately ten times higher in the case of tap water than for DI water (H α -2800 vs. 175, O I (777.3nm)-3000 vs. 340, N I (822 nm)-1730 vs. 150 in arbitrary units for tap and DI water, respectively).Similar scaling of the atomic lines then continues up to 200 ns.
In both cases, the first detectable light comes from the molecular bands of nitrogen-SPS in 300-500 nm wavelength range and FPS in 550-900 nm wavelength range.The excitation of SPS, FPS bands signifies that nitrogen gas from the air in contact with the plasma's electrons on the surface excites the nitrogen molecular bands by electron impact excitation.After a few ns (80 ns in the case of DI water and 50 ns in the case of tap water), the light emitted is entirely dominated by atomic lines of hydrogen, oxygen, nitrogen and ionic line of nitrogen (not shown here), indicating the enhanced dissociation of nitrogen and oxygen molecules.At this stage, a high degree of ionisation is likely reached as compared to the streamer phase.
Molecular radicals and atomic species are essential for post-discharge chemistry and converting the water into PAW [12].Figure 11 shows typical emission spectra in the 300-340 nm wavelength range during the first tens of ns (a) and first microseconds (b), respectively.It has already been mentioned and shown in figure 10 that SPS nitrogen bands dominate during the first tens of ns characterising the streamer phase.Here, also the other bands of SPS dominate in the initial phase.The first few microseconds shown in figure 11(b) show that OH emission is excited in both DI and tap water, clearly depicting the production of molecular radicals.Interestingly, NH UV emission is also present in tap water, which is not observed in DI water.The NH species could be produced by N( 2 D, 2 P) + H 2 O → OH + NH [17] or through recombination of atomic nitrogen and hydrogen N + H + M → NH + M [18].In the case of tap water, the higher electron densities and higher gas temperatures lead to higher densities of NH precursors and, therefore higher production of NH than in the case of DI water.Nevertheless, our information about NH formation is quite limited.We observe that the radiative state NH(A 3 Π) is formed during the afterglow, and we do not observe NH emission at 335 nm during the streamer phase.
Figures 12(a) and (b) show the time evolution of emission spectra in the 380-410 nm wavelength range for the discharge initiated in DI and tap water, respectively.The molecular emission of SPS nitrogen bands observed in both cases during the first tens of nanoseconds characterises the streamer phase of the discharge.The presence of FNS(0,0) band at 391 nm depicts the higher electric field during the initial phase, with electrons exceeding an energy threshold of 19 eV.The emission intensity of each band first increases and then decreases, and this depletion of bands is mainly due to E/N variations during the discharge initiation [19].Time-resolved spectrometric signatures allow other characteristics to be obtained, such as the evolution of gas temperature, electric field, electron density and electron temperature, as detailed in the next section.

Parameters obtained from images
We have estimated the filament's diameter and plotted it as a function of distance from the blade in figure 13.The zero value corresponds to the approximated position of the blade, and positive and negative indicate the bottom (grounded electrode) region and top (near to HV) region with respect to the position of the blade.Here, the discharge is initiated on the surface of the water, and filaments expand above the blade on both sides.The thickness of the filament was evaluated from the images acquired at later times (shown in figures 8 and 9) because, at that time, the filament becomes prominent and fully developed.Later, these filaments thin out and develop more branches.
Figure 13(a) shows the diameter of one of the prominent filaments when the discharge is initiated in tap water for three different times.It is clear from the figure that the filament propagated at the bottom side is thicker than the top side, and the cause of this inhomogeneity could be related to the asymmetric evolution during the initial diffusive phase.At t ∼ 250 ns, the thickness of the filament is less as compared to t ∼ 500 ns.As time progresses, the branching is dominant, and the main filament becomes narrower, and hence the thickness decreases at t ∼ 1000 ns.However, the behaviour on both sides of the blade is similar.Figure 13(b) shows the comparison of the thickness of the filament in tap and DI water.One can see that the filaments formed in DI water are thinner as compared to tap water.

Parameters obtained from N 2 molecular emission
4.2.1.Determination of gas temperature.Gas temperature is a significant parameter in determining the discharge's underlying physical processes.The molecular rotational temperature  T rot of the electronically excited state is often used to estimate the gas temperature, and N 2 SPS bands are often used for this purpose [12,[20][21][22][23]. Figures 14(a) and (b) shows the N 2 (SPS) normalised spectra at different times for the case of DI and tap water respectively.All spectra were acquired using a 2 ns MCP gate and normalised to the SPS(0,2) band maximum (band head at 380.49 nm).One can clearly see from figure 14(a) that no big difference is observed in the spectra for the case of DI water as time progresses.However, in tap water, the tail of the SPS(0,2) band increases with time relative to the band head (see figure 14(b)), which is a clear evidence of an increase in N 2 (C 3 Π u ) state rotational temperature or gas heating.Synthetic spectra generated for T rot between 300 K and 6000 K were used to estimate rotational temperature from experimental spectra.These simulations were created by an open-source code [24,25].The measured spectra collected from PIE were compared with each simulated spectra using the least-square principle, and the accuracy of fit was defined by the chi-squares (χ 2 ) method.The least-squares method can estimate the parameters by minimising the contribution of χ 2 .The comparison of simulated and experimental spectra of N 2 -SPS for T rot = 500 K is shown in figure 14(c).The whole sequence is fitted well with simulated data.
Figure 15(a) shows the temporal evolution of the emission profile of N 2 -SPS(0,2) measured from the recorded spectra in both DI and tap water.Figure clearly depicts the peak of the emission intensity at t = 12 ns in both types of water.However, the emission intensity is lower in DI water than in tap water.Figure 15(b) shows the temporal evolution of rotational temperature in DI and tap water estimated by comparing synthetic and measured spectra.The initial rotational temperature is 330 K for both DI and tap water, peaks at 16 ns with T rot ∼390 K in DI water and T rot ∼500 K in tap water, and keeps on increasing to 500 K in DI water and 1100 K in tap water.The peak of the emission intensity (12 ns) occurs before the first peak in rotational temperature (16 ns) and this probably indicates the production of oxygen atoms by the quenching of excited states of N 2 molecule resulting in the heating of the gas [26].The linear increase of rotational temperature is the signature of the spark phase.The increase in rotational temperature in DI water has been reported by Wang et al [27].They have reported the temperature increase from 355 to 430 K with the increase in pulse voltage due to the formation of transition of transient spark phase when the voltage crosses the threshold value.In our case, the heating can be also attributed to the transition of the streamer phase to the spark phase.The more dominant spark phase is formed in tap water than in DI water, and hence, the gas temperature is higher in tap water as compared to DI water.Figures 15(c) and (d) show the calculated vibrational distribution functions (VDFs) in DI and tap water, respectively.The initial VDFs in both cases (DI and tap water) coincide with the typical VDF in air/wet air [21,28,29] due to electron impact excitation from zero vibrational level of N 2 .The VDF is approximately 1:(0.6 ± 0.2):(0.15± 0.05):(0.04± 0.02):(0.010± 0.005).The VDF reaches an equilibrium rapidly since the effective lifetimes of the C 3 Π upper levels is in the air approximately 0.5 ns or lower.During the fitting of the experimental data, we had to carefully choose the basis and adjust their choice according to the current rotational temperature.In both cases (DI and tap water), we observed the increase of the vibrational temperature of the N 2 (C 3 Π u , v) manifold in time.This increase refers to the excitation of the N 2 (C 3 Π u , v) manifold by the electron impact excitation from the higher vibrational levels of molecular nitrogen ground state (e + N 2 (X 1 Σ + g , v ⩾0) → e + N 2 (C 3 Π u , v ′ = 0-4)) and therefore to the increase of the vibrational temperature of the gas in the discharge, see figure 4 of [28].Obviously, the population of the N 2 (X 1 Σ + g , v >0) should be higher in the case of tap water.

Determination of reduced electric field.
The electric field is another quantifiable parameter to determine the plasma kinetics as it determines the mean energy of the electron energy distribution function (EEDF), which governs the fundamental properties of plasma.The electric field is often determined from the intensity ratio of spectral bands.Particularly the ratio of N + 2 -FNS and N 2 -SPS bands due to its very different ionisation/excitation threshold is very sensitive to reduced electric field (E/N) [19,[30][31][32].
The dependence R(E/N) relating the intensity ratio and reduced electric field is given by: where I FNS,SPS are the measured intensities of respective spectral bands and τ FNS,SPS eff are their effective lifetimes at air are τ FNS eff = (0.045 ± 0.010) ns, τ SPS eff = (0.5 ± 0.1) ns.In this work, we use the intensity ratio of FNS(0,0) and SPS (2,5) to determine the electric field.
Figures 16(a) and (b) show the time series spectra of N 2 bands obtained when the plasma is initiated in DI and tap water, respectively, and the central wavelength of the spectrometer was set to 390 nm.The time series of spectra are obtained with a step of 2 ns and an MCP gate of 2 ns.For the first 10 ns, the intensity of the peak of FNS(0,0) is slightly higher than the peak of SPS (2,5), and later on, the peak intensity of SPS(2,5) keeps on increasing.This is a typical signature of initially high E/N.The same scenario happens in DI water as time progresses, but the bands have less intensity as compared to tap water.
The experimental calibration curve relating the ratio of FNS and SPS to the reduced electric field has been determined in Townsend discharge by Paris et al [32] for various pressures.They have also formulated the relation in terms of the equation, which has the following form at atmospheric pressure: Here, a = 46.0,b = 89.0.Later, the uncertainty of R(E/N) was investigated using a kinetic model by Obrusnik et al [30].
The uncertainty appears due to scatter in the measured rate constants of the main processes (electron impact excitation, radiative and collisional quenching processes) and these processes define the R(E/N) relation.More details can be found in [19,30].
The same strategy to estimate E/N can be used also in our case.The calibration curves depicting lower and upper bounds of the uncertainty band as shown in figure 16(c) were determined using a kinetic model based on the following dominating processes: electron impact excitation of N 2 and N + 2 , spontaneous emission and quenching reactions depopulating the upper states by collisions with O 2 , N 2 and H 2 O.Moreover, the upper and lower bounds are fitted with the equation having the form of the equation (2).The estimated R(E/N) obtained from the measured spectra is used in these equations, and mean reduced electric field is determined.The temporal evolution of a reduced electric field in DI and tap water is shown in figure 16(d).The reduced electric field increases to reach the maximum value and then decreases with time.Recently [12], in this geometry, electric field distribution in the vicinity of one insulating blade has been modelled for the liquid conductivity similar to tap water using COMSOL Multiphysics predicting a peak value of 1200 Td (E > 3 × 10 7 V m −1 ).The reduced electric field in DI water shows a similar time evolution but with a lower peak value compared to tap water.

Electron density and temperature obtained from atomic lines
The knowledge of electron density (n e ) and electron temperature (T e ) is required to develop an accurate plasma chemical model.To determine these parameters, the line shapes of selected atomic lines were collected and analysed.The n e was determined from the Stark broadening of H α , H β since hydrogen atoms are most sensitive to Stark broadening, and estimated density is compared with density calculated from the Stark broadening of strongest atomic lines of O I (777.3nm).In addition, the electron temperature can be estimated by the intersection method.The basis of this method is to determine simultaneously n e and T e using the Stark broadening of two or more lines under the same working conditions [33,34].
Figures 17(a) and (b) show the evolution of the normalised intensity of H α and O I line profiles for the case of tap water.The figure shows that as time progresses, the width of both the lines first increases up to 70 ns and then decreases after 1000 ns.After that, the width becomes constant, and at that time, it just reflects instrumental broadening.The instrumental broadening for our spectrometer is 0.19 nm (grating 1200 G mm −1 , entrance slit 100 um).The instrumental broadening was measure using an Ar pen ray light source and fitted with Gaussian to estimate the full width at half maximum (FWHM).The maximum width determined from the spectra for H α , H β , O I is 20 nm, 4 nm, and 0.7 nm, respectively.Although the oxygen line is weak and insensitive towards any broadening mechanism compared to hydrogen atoms but is also employed to calculate n e for measuring the deviation and error bar in calculations.
Figure 17(c) (data points) shows the H α line shape measured in the experiment for the delay of 70 ns after the beginning of HV.Interestingly, the line is fitted with a single Lorentz function shown as a black dotted line, and the peak is shifted to 0.7 nm.To confirm the assumption that the observed broadening is mainly due to the Stark effect, the shifts of H α lines are estimated for various times.The theoretical shifts are also estimated from the stark shift model of Griem et al [35,36] and the Van der Waals shift model of Konjevic et al [37].The experimental shifts shown in figure 17(d) closely match H α lineshape broadening due to the Stark effect and not the Van der Waals pressure broadening (black dotted lines).This analysis has not been performed for atomic oxygen lines as the shifts of O are too small to be measured, and H β does not exhibit a shift.
The temporal evolution of electron densities determined from the line shape analysis of H α is shown in figure 18.In order to support the electron densities estimated from H α , the electron densities obtained from H β and O I have also been plotted in the same figure where the error bars come from the fitting.The electron densities increase to 70 ns and then follow an exponential decay.The densities measured from the three lines are in good agreement with each other up to 100 ns for oxygen triplet and 500 ns for H β .After 100 ns, the electron  densities are below 10 17 cm −3 , lower than the sensitive limit of the oxygen triplet line.However, for the case of H β , the width quickly reaches instrumental broadening, and hence the estimation of densities is not done.The electron density keeps on decreasing with time, indicating the electron loss due to either electron-ion recombination reaction or electron attachment reaction.Liu et al [38] suggested that electron attachment reaction is significantly larger than electron-ion recombination reaction with gases having higher water content.Recently, Sainct et al [39] performed the spatial and temporal electron number density measurement in water vapour at atmospheric pressure.They concluded that the dominant process for electron loss is via electron attachment reactions.Hence, we can conclude that in our system, the plasma density is falling due to an electron-attachment reaction.
Another application of Stark broadening is that it depends upon both the plasma density and electron temperature.This dependence can be exploited to estimate n e as well as T e simultaneously from the method called cross-point or intersection method [33,34].However, the sensitivity of the crossing point is not very high.There are several works that have used this method to estimate plasma parameters by using H I lines, e.g.H α , H β and H γ .Yubero et al [40] used this method in microwave discharge at atmospheric pressure to determine plasma parameters from the broadening of H α , H β and H γ lines at atmospheric pressure.They have estimated the uncertainty in n e is 10% and 40% in plasma temperature.Dobrynin et al [41] used this technique for hydrogenated   Despite inaccuracies, the intersection method is employed to estimate the temporal evolution of plasma parameters in the spark phase, the region where atomic lines are excited.The dependence of T e with n e for a fixed value of Stark width of H α , H β using the GC model [42,43] and O I [44] neutral is shown in figure 19(a).The intersection points are also shown in the same figure.The estimated value for fixed Stark width with n e ∼ 2.99 × 10 17 ± 5% cm −3 and T e ∼ 1.04 ± 29% eV. Figure 19(b) and (c) show the temporal evolution of plasma density and temperature when the values are estimated using the Stark cross method.This method also shows that plasma density decreases with time, as observed in figure 18 when line shape analysis is performed for density estimation.The plasma density decreases due to the reattachment reaction of electrons.Interestingly, electron temperature also decreases quickly from 4 eV to 0.5 eV.This could be due to loss of energy from electron collision with ions, water, and neutral atoms and electron attachment reaction with hydrogen and oxygen molecules.
The temporal evolution of plasma density estimated from the width of Stark broadened H α atomic line shape analysis has been shown when the discharge is initiated in tap and DI water and plotted in figure 20.The plasma density first increases and peaks at 70 ns and then keeps on decreasing with time.However, in the case of DI water, it peaks at 80 ns and then decreases at a higher rate than tap water.The mean plasma density is lower in DI water as compared to tap water.Here, the conductivity of water plays a critical role and creates more electrons in tap water than in DI water.In order to estimate the lifetime of electrons, we have fitted the decay in density with exponential fit and found that the effective lifetime is 70 ns in tap water and 30 ns in DI water.

Applicability of the intensity ratio method
The intensity ratio method can be used to estimate the electric field reliably only if the conditions for the applicability of this method are sufficiently verified [19].As can be seen from the temporal evolution of intensity, the band of FNS(0,0) is present up to 16 ns, and during this time, the rotational temperature ranges from 330-450 K for the case of tap water and 330-380 K for the case of DI water which can also be seen from figure 15(b).Furthermore, we evaluated the VDF of SPS during the first 15 ns, which verified the dominant process of electron impact excitation from ν = 0 as shown in figures 15(c) and (d).However, the VDF analysis of FNS is difficult to proceed as the intensity of other bands of FNS is too weak to determine the VDF at atmospheric pressure.Instead, we have assumed that the dominant excitation process of N + 2 is also an electron impact excitation.We have determined the band peak intensity of FNS and SPS from the measured spectra and converted the band head intensity to band integral intensity according to the formulation described in [31], and then the R(E/N) is determined.

Applicability of the Stark broadening method
The shape of an emission line can be affected by various broadening mechanisms, such as natural broadening, Van der Waals broadening, and Stark broadening, which corresponds to the Lorentzian shape.Doppler and instrumental broadenings are characterised by a Gaussian profile.The gas temperature in the discharge rises from 300 K to 1000 K and from 300 K to 450 K (figure 15) for tap water and DI water, respectively, and hence the Doppler broadening FWHM has a maximum value is 0.008-0.013nm for H α , 0.006-0.009nm for H β and 0.003-0.004nm for O I .Thus the Doppler broadening width is well below instrumental broadening and does not contribute to the observed lineshape.The Van der Waals broadening is pressure dependent.In our discharge, gas pressure can increase due to the abrupt heating of the gas.Even assuming an isochoric rise in temperature from 300 K to 1000 K, the gas pressure in the discharge filaments could not exceed 3 bars.The corresponding maximum pressure broadening FWHM would be in the range 0.269-0.135nm, 0.204-0.103nm, 0.109-0.055nm for H α , H β and O I , respectively, which is again well below the measured values.Thus, the observed line shapes can be fully explained by Stark broadening after subtracting the instrumental component.

General findings
The aim of this work is to reveal the general properties and parameters of a nanosecond discharge occurring at the waterair interface with electrodes fully immersed in water.The discharge consists of a streamer and transient spark phase.The streamer phases (t < 15 ns) in the DI and tap water differ only in the luminosity, but the morphology, propagation velocity, and reduced electric field are very similar.A significant difference in luminosity at comparable E/N means that the discharge in tap water is characterised by a higher electron density already during the streamer phase, which leads to increased production of the excited molecular states (N 2 (A 3 Σ + u , B 3 Π g , C 3 Π u )) as well as atomic species (N, O, H).During the first 40 ns, we also observe very rapid increase of N 2 (C 3 Π u ) state rotational temperature (from 300 to 1020 K) in tap water, which is not present in DI water (increases from 300 to 500 K).Fast increase of the rotational temperature on the nanosecond time scale is very likely due to dissociative quenching of the excited states by molecular oxygen [45]: This process is connected with heat release (several eVs) stored in the kinetic energy of O atoms.The atoms are thermalised in collisions with neutral gas, thus, increasing its temperature.We have compared also VDF of N 2 (C 3 Π u , v = 0-4) and VDFs are the same during the streamer phase for DI and tap water.However, the VDFs considerably differ at t = 15-40 ns.In tap water we observe an increase in the vibrational temperature of the N 2 (C 3 Π u , v) manifold in time, indicating also the increase of the vibrational temperature of the gas.This is very probably due to electron vibrational processes (e ).Moreover, during t = 15-40 ns some of the streamer channels are converted to spark channels (more bright and well-separated than streamers).Obviously, in tap water, more streamer channels survive and transform into sparks.The first measurements of electron density using Stark broadening were performed at t = 50 ns, when n e reaches values 2.2 × 10 17 cm −3 and 5.4 × 10 17 cm −3 for DI and tap water, respectively.Note that at t = 50 ns the spark phase is already well developed, see figures 6 and 7 and panels T 0 + 48 ns.The electron temperature at t = 50 ns in tap water was determined to be 4.2 eV (corresponding to E/N ∼ 150 Td in air), later the electron temperature decreased to approximately 0.65 eV (E/N ∼ 10 Td) at 180 ns.The difference between the DI and tap water in the spark phase is illustrated using table 2 listing the main parameters obtained through spectral analysis.The r and N f denote the average luminous radius and number of the spark channels, respectively.The estimated luminous radius (figure 13) corresponds to 0.4 mm and 0.18 mm at t = 250 ns in tap and DI water, and the number of filaments is approximately 20 and 8 in tap and DI water, respectively.Note that in the reactor, there are two grounded electrodes (therefore N f is twice as shown in fourchannel images see the first image of figures 8 and 9).The ε denotes the energy deposited in the discharge per one pulse, see figures 2(c) and (d).The ε ps is the energy per one species.Nevertheless, we can provide only a very rough estimate of ε ps based on the average properties of a typical spark channel as: where l is length of spark channel and p is pressure (760 Torr).
In both cases, the resulting ε ps is comparable, however, encumbered by 70%-90% relative error.Note that the calculation is based on the projected luminous dimensions of filaments, which do not consider the vertical dimensions of the discharge.The energy released in the discharge is consumed by the ionisation, production of excited states, atomic species and changes in vibrational distribution.1.0 ± 0.9 1.0 ± 0.7

Summary and conclusions
In this work, we study the properties and characteristics of filamentary discharges propagating along the surface of a thin water layer in a special surface coplanar DBD-like electrode geometry.The electrode system consists of two dielectric blades separating one HV and two grounded electrodes.Here in this unique configuration, the discharge filaments are in contact with only the water surface, which reduces contamination of the water with the electrode material.We have recently shown that this geometry enables the production of PAW with high energy efficiency [12], and this work aims to elucidate as much as possible the reasons for this high efficiency.The study uses ultrafast imaging and spectrometric diagnostics to reveal the fundamentals of the discharge on timescales from nanoseconds to microseconds for two different initial conductivities (1.2 and 450 µS cm −1 ) of water injected into the reactor.High-speed imaging reveals an initial bi-directional ionising avalanche/wave that propagates across the liquid surface, followed by the formation of discrete discharge channels.The filaments probably arise as a result of ionisation instability [16], leading to the formation of several highly luminous filaments, implying the transition of the streamer phase to the transient spark phase.The filaments re-ignite in the same positions on the liquid surface both for DI and tap water over several successive (mainly capacitive) current pulses (∼1 µs apart), which result from discontinuities in the driving voltage.Both the initial phase and the streamer-spark transition phase were further studied using time-resolved optical emission spectroscopy using the technique of ICCD kinetic series.The initial emission spectra are exclusively composed of molecular bands (N 2 /N + 2 ) and weak NIR O I atomic lines.The streamer-spark transition and the filamentary phase reveal a remarkable increase in both the intensity and broadening of the H I , O I , and N I atomic lines, a clear indication of the increased degree of dissociation and ionisation in the discharge channels.The evolution of gas temperature was estimated from the SPS of nitrogen molecular bands and we found that the temperature increases from 300 K to about 450 K and 1000 K in DI and tap water, respectively.The electric field was also estimated from the ratio of the SPS to the FNS of nitrogen molecular bands only in the streamer phase.The initial reduced electric field E/N appears to be comparable for both conductivities, the peak value reaches approximately 850 and 700 Td in tap and DI water, respectively.The plasma density was estimated from the broadening of hydrogen atomic and oxygen neutral lines during filamentary phase, and maximum electron density ∼ 10 18 cm −3 is observed.The electron temperature was also estimated from the Stark cross method.The temporal evolution of electron density and temperature is estimated for both types of conductivity of water.
The results provide fundamental insight into the underlying mechanisms of a novel DBD-like discharge at the air-water interface, which has interesting potential as an efficient source of PAW.However, few things are still not fully understood because all the spectra analysed in this work were obtained symmetrically around the dielectric blade without spatial resolution.Further extensive work is therefore needed to understand, for example, the specifics of anode versus cathode directed ionisation waves or the diffusive nature of the end parts of filaments.Appropriate experiments are underway to obtain spatially resolved spectrometric characteristics of the initial bidirectional streamer phase, as well as a detailed chemical analysis of the water collected at the reactor outlet.

Figure 1 .
Figure 1.Schematic of the experimental setup of nanoseconds discharge on water-air interface.

Figure 2 .
Figure 2. Waveforms of pulsed voltage, discharge current, and photomultiplier tube (PMT) of (a) DI (b) tap water.(c) and (d) shows the power and energy as a function of time for the case of DI and tap water, respectively.(e) and (f) shows the pulsed voltage, discharge current and PMT up to positive (main) HV pulse.The PMT signal was obtained by integrating the intensity in the wavelength region 300-800 nm.

Figure 4 .
Figure 4. XXRapidFrame image sequences obtained from single discharge events in DI water.Horizontal bars in (a) labelled as Ch1-Ch4 illustrate the timing of the Ch1-Ch4 MCP gates with respect to the onset (t = 0 ns) of the luminous signal.Images in panel (b) show the evolution of three selected (typical) events during the primary HV pulse acquired by using equal Ch1-Ch4 MCP gates of 2 ns.

Figure 5 .
Figure 5. XXRapidFrame image sequences obtained from single discharge events in tap water.Horizontal bars in (a) labelled as Ch1-Ch4 illustrate the timing of the Ch1-Ch4 MCP gates with respect to the onset (t = 0 ns) of the luminous signal.Images in panel (b) show the evolution of three selected (typical) events during the primary HV pulse acquired by using equal Ch1-Ch4 MCP gates of 1 ns.

Figure 6 .
Figure 6.XXRapidFrame image sequences obtained in DI water acquired by using MCP gate (CH2) to visualise the evolution of discharge morphology during the HV pulse.

Figure 7 .
Figure 7. XXRapidFrame image sequences obtained in tap water acquired by using MCP gate (CH2) to visualise the evolution of discharge morphology during the HV pulse.

Figure 8 .
Figure 8. XXRapidFrame image sequence obtained by single discharge event in DI water.Horizontal bars in (a) labelled as Ch1-Ch4 illustrate the timing of the Ch1-Ch4 MCP gates with respect to the onset (t = 0 ns) of the luminous signal.Images in panel (b) show the whole evolution of plasma utilising the double frame acquisition mode.

Figure 9 .
Figure 9. XXRapidFrame image sequence obtained by single discharge event in tap water.Horizontal bars in (a) labelled as Ch1-Ch4 illustrate the timing of the Ch1-Ch4 MCP gates with respect to the onset (t = 0 ns) of the luminous signal.Images in panel (b) show the whole evolution of plasma utilising the double frame acquisition mode.

Figure 11 .
Figure 11.The optical emission spectra in the 300-340 nm wavelength range recorded for the discharge in DI and tap water show the emission (a) during the first tens of ns (b) during microseconds.The figure shows that SPS molecular bands dominate in the initial phase and the excitation of OH emission at a later stage in both water conductivity.The NH emission is also observed in tap water, which is not present in DI water.

Figure 12 .
Figure 12.Time evolution of optical emission spectra in 380-410 nm range recorded for the discharge in (a) DI water and (b) tap water showing the initial evolution of the molecular SPS and FNS nitrogen bands during the first tens of nanoseconds and characterises the streamer phase.Taken in kinetic series mode using a 1200 G mm −1 grating and 100 µm entrance slit width.

Figure 13 .
Figure 13.Radiative radius of one of the filaments as a function of distance from the blade for (a) three different times in tap water and (b) shows the same at t = 250 ns in tap and DI water.

Figure 14 .
Figure 14.Time evolution of N 2 -SPS emission spectra (∆ν = −2 sequence) acquired at discharge for (a) DI and (b) tap water with integration time 2 ns.The spectra were obtained using 1200 G mm −1 with an entrance slit of 100 µm.(c) Shows a comparison of synthetic best-fit spectrum calculated by using Trot = 500 K and experimental data of N 2 SPS emission.

Figure 15 .
Figure 15.Time evolution of (a) emission intensity of N 2 SPS(0, 2) band profile and (b) rotational temperature obtained from the synthetic best-fit spectrum of N 2 SPS emission for tap and DI water, (c) and (d) shows the vibrational distribution of SPS of N 2 in DI and tap water.

Figure 16 .
Figure 16.Time series of SPS and FNS spectra obtained for grating 1200 G mm −1 with an entrance slit of 100 µm for (a) DI (b) tap water.(c) shows the calibration curves showing the relation of the ratio of FNS(0,0) and SPS(2,5) emission with the reduced electric field for synthetic air at atmospheric pressure.(d) temporal profile of reduced electric field for tap and DI water.

Figure 17 .
Figure 17.Typical Hα and O I emission line profiles ((a) and (b) respectively) produced by the surface discharge in tap water from 50 ns to 1.5 µs.A comparison of Hα line and simulated line fitted with Lorentz profile is shown in (c).A comparison of experimentally measured and theoretically estimated shifts of Hα line is shown in (d).

Figure 18 .
Figure 18.Temporal evolution of (a) width of Hα, H β and O(I) estimated by fitting the line emission and (b) electron density estimated from the stark broadening width and parameters.

Figure 19 .
Figure 19.(a) Dependence of electron density (ne) and electron temperature (Te) for fixed values of stark broadening corresponding to the width of Hα (2.07 nm), H β (10.03 nm) and O(I) (0.55 nm) using the GC model.Temporal evolution of electron density in (b) and electron temperature in (c) determined from the Stark crossing method.

Figure 20 .
Figure 20.Temporal evolution of electron densities estimated from the Stark width in tap and DI water.

H
α and non-hydrogenated O I neutral atomic line for a nanoseconds pulsed discharge in water.

Table 1 .
The table provides the details of central wavelength (lc), the bandwidth of interference filter and part no.used to capture the optical emission in the specified region of interest (ROI).

Table 2 .
Comparison of discharge parameters determined in tap and DI water.