DBD-like and electrolytic regimes in pulsed and AC driven discharges in contact with water

The interaction of an ambient air plasma with a water surface in a pin–water electrode configuration is presented in a polydiagnostic study. A discharge was generated by applying different high-voltage (HV) waveforms to a metallic pin electrode, positioned 2 mm above the water surface of a Petri dish filled with demineralized water. For pulsed discharge operation, a clear distinction is observed between a dielectric barrier discharge regime featuring a transient discharge at the rising as well as at the falling slope of the HV pulse, while a steady discharge is present in the gap during the complete HV pulse for the electrolysis regime. The occurrence of those two regimes is coupled to the increasing conductivity of the water over time, which additionally results in a quick rise of the dissipated discharge power and an increase of the gas temperature. The AC driven discharges exhibit only the electrolysis regime and do not significantly evolve over the treatment time. The resulting water conductivity was found to be a function of the total dissipated energy, irrespective of the discharge driving mode. Additionally, the resulting water conductivity shows a strong correlation with the total transferred charge in the gas phase. The total dissipated energy can potentially be used as a global measure to compare different experiments involving plasma–water interaction across different setups in different research groups.


Introduction
The interaction between plasmas and liquids has been investigated at an increased rate in the last decades [1,2].While the type of plasma interacting with liquids studied in literature varies from dielectric barrier discharges (DBDs), atmospheric pressure plasma jets and gliding arcs to sparks, the research of plasma-treated water (PTW) has already shown its value in many fields, e.g. in the field of agriculture, where it stimulates seed germination and plant growth [3][4][5][6][7][8].Nutritious properties of nitrates generated in the water during plasma treatment contribute to PTW becoming a potential alternative for environmentally harmful fertilizers [9], while disinfecting properties are promising for PTW becoming an ecologically-friendly pesticide.In the biomedical field, the disinfecting qualities of PTW are studied for wound treatment, cancer treatment and sterilization of medical equipment.Here the PTW has a role in stimulating wound healing and inactivating bacteria and cancer cells [10][11][12].The interaction of water with an electrical discharge leads to a complex chemistry at the surface region connecting the gas phase chemistry with the chemistry in the liquid.This results in a water solution containing relatively long-lived reactive species with a lifetime of days to months [7].The use of plasmas for water purification has been investigated as an alternative to other oxidation processes like treatment with UV or ozone as well [13].In order to bring this research field to applications, a better understanding of the physical and chemical processes is required for a controllable activation process.As the energy efficiency and the large-scale production of PTW are important for many applications, the influence of plasma dynamics and processes on the generated species in the water needs to be better understood.
The large range of different aspects investigated on plasmawater interaction in other works [7,[14][15][16][17][18][19][20][21][22] demonstrates the complexity of this topic.Therefore, knowledge on the interdependency of plasma and water properties is of importance for a better understanding of underlying processes.This is why a combined study on this dynamic system is relevant to relate the discharge processes to the changes that occur in the water.This article is therefore dedicated to the investigation of the general effects of changing liquid properties on plasma characteristics in dependence on the high-voltage (HV) waveform (pulsed and AC), without considering specific species.
In this work, the results of a polydiagnostic study on an ambient air discharge in a pin-water configuration are presented, using unipolar HV pulses of both polarities as well as AC (sinusoidal) operation.Different discharge modes have been investigated using a full electrical characterization, fast imaging and gas temperature determination by optical emission spectroscopy (OES) at various positions in the gap.These measurements have been combined with the determination of water properties like conductivity and pH value as a function of the treatment time.These results provide a comprehensive insight into the discharge behavior, which is changing over time.In addition, the existence of two regimes of operation are presented: the DBD regime, which occurs when the water behaves as a dielectric, and the electrolysis regime, which occurs when the water becomes more conductive.

Experimental setup
To investigate the interaction between an atmospheric pressure plasma and a water surface, an electrical discharge was generated in ambient air in a pin-plane electrode configuration, schematically shown in figure 1.An HV potential was applied to a needle electrode (tungsten welding electrode), positioned 2 mm above a Petri dish, initially filled with ≈80 ml of demineralized water with a conductivity of 0-5 µS cm −2 .A stainless-steel electrode was embedded in the water on the bottom of the dish and is connected to the ground through a hole in the Petri dish.Measurements of discharge characteristics and water properties were performed after operating the discharge in contact with water ('treatment time') ranging from 0 to 60 min.The demineralized water was refreshed before each measurement to start with comparable initial conditions.
The plasma was driven by either positive or negative HV pulses or by AC excitation.An HV pulse generator (DEI, PVX-4110) was used to apply unipolar square pulses 1.5 µs long with a repetition frequency f of 5 kHz and 6 kV in amplitude.The HV was generated by an HV amplifier (Trek, 10/10B-HS) that magnifies the voltage created by a dual power supply (Delta Elektronika, E018-0.6D).An AC HV waveform generator (custom made) was used to generate a sine waveform with a frequency of 30 kHz with a peak-to-peak potential of 12 kV.
To examine the electrical properties of the system during a period, the voltage and current signals were measured.An HV probe (Tektronix P6015A) was connected to the needle electrode to measure the applied potential.The current was measured with a current probe on the HV side (Pearson model 6585).The voltage and current waveforms (V(t) and I(t), respectively) with a period T were used to derive the averaged consumed electrical power by the setup, using the following relation The charging of the water surface was measured using the same HV probe on which a pin was mounted.The pin was positioned to touch the water surface.The measurements were performed 3 mm away from the central point where the discharge touched the water surface, only for discharges generated by negative HV pulses.Measurements closer to the discharge as well as measurements for the positively pulsed discharges could not be performed as the presence of the probe tip disturbed the discharge.
For pH value and conductivity measurements of the water, a pH-conductivity meter (Mettler Toledo, SevenExcellence S470-Std-K) was used.The pH and conductivity of the treated water were measured after the discharge was switched off and the water in the Petri dish was stirred before performing the measurement to create a homogeneous solution representative of the bulk of the water.The temperature of the water was measured simultaneously using the sensor built in the conductivity probe.
To observe the morphology and the temporal development of the discharge, fast imaging of the discharge was performed by use of an iCCD camera (4Picos Stanford Optics), which was synchronized with the applied HV waveform.When operating the camera with gating widths of several nanoseconds, the intensity of a single discharge event was not sufficient to obtain useful images in the described setup, therefore the displayed results show an accumulation of multiple events.
OES was performed by focusing the discharge emission on a vertical slit (50 µm width) at one end of a monochromator (Jobin-Yvon, HR 1000).The discharge channel was imaged on the entrance slit and the different axial positions were taken from a single image by selecting different heights in the image.The diffraction grating inside the monochromator contains A schematic overview of the experimental setup used for the results presented in this report.The discharge between the needle and water surface is generated either (i) by applying a unipolar pulse (+ or −) with an amplitude of 6 kV, pulse length of 1.5 µs and frequency of 5 kHz or (ii) by applying an AC HV sine waveform with 12 kV peak-to-peak voltage and a frequency of 30 kHz. 1200 grooves mm −1 and the diffracted light was captured by an iCCD camera (Andor, iStar).Images obtained by the camera cover a wavelength range of ≈10 nm centered at 355 nm with a spectral resolution of ≈10 pm.The images contain spatial information in the vertical direction.The rotational temperature was determined from the emission of the ∆ν = −1 rotational band of the second positive system (SPS) of nitrogen around 357 nm.The choice of this transition over the transitions with the band head at 337 nm was made because of an emission signal from the NH band at 336 nm.To perform a fit on the spectral data the intensity at a small vertical range is integrated (z ± 0.1 mm) to obtain a spectral profile.The confidence intervals of the fitting procedure are typically below 5%.The gate width of the detector is 50 ns.
The majority of the emission of the DBD-like discharges in air has a duration of about 2 ns (see e.g.[23]).Consequently, the 5 ns image of the rising slope shown in figure 4 at 30 ns resembles the overall discharge emission structure quite well.On the other hand, the OES captures reasonably well the rising slope of the discharge.

Characterization of the PTW as reference for gas-phase measurements
In the discharge gap, the plasma is a source of reactive oxygen and nitrogen species (RONS), being species such as ozone (O 3 ), hydrogen peroxide (H 2 O 2 ), nitrous acid (HNO 2 ) and nitric acid (HNO 3 ), or radicals such as hydoxyl (.OH), nitric oxide (.NO) and nitrogen oxide (.NO 2 ) [24].During treatment time (time during which plasma interacts with water), RONS are partially dissolved in the water to an equilibrium point.The transport of chemical species from the surface to the bulk of the water occurs due to a convective flow which is initiated by discharge processes.The bulk of the water contains mainly the long-lived species, since short-lived species will be lost in reactions before they are transported.The dominant long-lived species present in plasma treated water (PAW) are hydrogen peroxide (H 2 O 2 ), nitrite (NO − 2 ) and nitrate (NO − 3 ) [1].The conductivity of the treated liquid reflects the abundance of ions in the liquid phase.
In figures 2(a) and (b) the results of the pH and conductivity measurements of the bulk of the water are presented as a function of the treatment time.Each data point represents a separate measurement, where a Petri dish is refilled with demineralized water (initial conductivity ⩽ 5 µS cm −1 ) and treated for the displayed time.
During the plasma treatment, the pH value decreases and the conductivity increases, as is to be expected, since an increase in conductivity is directly related to the generation of ions in the water.This is in agreement with the effect of plasma treatment on the acidity and conductivity reported in published literature, e.g.Thirmudas et al [3] for different plasma sources.
The production of ions in the water affects its electrical properties over the treatment time.Consequently, for low conductivity the water typically behaves as a dielectric, while the increase in conductivity due to the production of ions in the water allows for an electrical current to be passed through the liquid, which is referred to as electrolysis.This will be discussed in more detail after the electrical data have been presented, in section 5.5.
While sinusoidal operation leads to a linear increase in the conductivity of the PTW, pulsed operation leads to a much faster increase.With pulsed operation, higher conductivity values (above 200 µS cm −1 ) are reached within a much shorter treatment time (15 min for pulsed discharges and 60 min for the AC-driven discharges).
This suggests the formation of nitrite and nitrate ions at a much faster rate when using pulsed discharges.The consequences of these differences in water conductivity on the discharge morphology and the discharge modes are discussed in more detail in sections 4 and 5, respectively.

Discharge morphology obtained by (fast) imaging
To visualize the average discharge morphology in the gas gap and on the water surface, long-exposure images of the discharges for the three different operation modes were recorded by a NIKON D3100 photo camera at approximately 20 • angle, with exposure time of 1/60 s thus showing approximately 83 discharges in the pulsed mode and 500 discharges in the AC mode.The images are shown in figure 3.For three cases, the discharge features a constricted channel (filament) in the gas gap, which is fairly spatially stable, widens up near the water surface and spreads over it 3 .Although these images are averaged, the discharge morphology on the water surface is clearly different for the positive pulsed operation compared to the other two cases.For the positively pulsed discharge, single discharge channels are recognizable on the water surface (see figure 3(a)), while there are no such channels visible for the negatively pulsed and AC powered discharges (figures 3(b) and (c)).These structures occurring for the positive operation are similar to the well-known Lichtenberg figures occurring on dielectric surfaces after streamer impact [25,26].It is often assumed that these pronounced and branched channels are a consequence of photoionization [25,27], which is more important for positive streamer propagation [28].Moreover, these findings for the different morphologies of positive and negative discharges on dielectrics are in accordance with the literature, see e.g.[27,[29][30][31].
For the visualization of the transient discharge emission structure for multiple phases of the applied HV waveforms, iCCD imaging was performed for two different treatment times for each of the three investigated cases.Please note that these iCCD images shown in figures 4 and 5 were recorded side-on (i.e. at 0 • angle in contrast to the the images shown in figure 3).Figures 4(a) and (b) show positively and negatively pulsed discharges.The top row of each figure shows the discharge at the beginning of treatment time (0 min) and the bottom row shows the discharges after 8 min of treatment time.The images were obtained using a 5 ns gate width showing approximately 800 accumulated discharge events.The iCCD gate was temporally related to the HV pulse to visualize the discharge emission during the rising slope (at 30 ns after the voltage rise), in the middle of the plateau (at 750 ns) and at the falling slope of the HV pulse (at 1530 ns).The exact timings during the slopes were determined by synchronizing the iCCD gate with the discharge currents peaks (see figure 6).The water surface is indicated as a white dashed line.The intensity visible below this dashed line is a reflection of the discharge emission on the water surface and should not be confused with a penetration of the discharge into the water.At the beginning of the treatment time (0 min), a discharge is ignited only during the rising and the falling slopes of the HV pulse.During the rise of the applied potential a discharge develops between the electrode and the water surface.Subsequently charge accumulates at the water surface, generating an electric field in the direction opposite to the external applied field, eventually extinguishing the discharge.When the external field drops to zero during the second slope, the remaining field generated by the accumulated charges ignites another discharge between the water and the needle electrode.This is analogous to the working principle of a DBD.Additionally, the discharge at the ending slope is weaker than the discharge at the initial slope, similar to the results reported in [32] for single-filament DBDs driven by asymmetric HV pulses like the ones used for this study.The discharge at the subsequent slope is weaker for the positively pulsed operation than for the negatively pulsed one (which is in accordance with the discharge current measurements, see figure 6).This is most likely a consequence of the different drift times of electrons and ions, i.e. after the initial discharges the positive ions need about 2 µs to move 1 mm under these conditions (which is longer than the HV pulse width of 1.5 µs), while electrons will reach the water surface much faster due to their higher mobility [33].Since the water surface is the cathode for the positively pulsed case, the very weak discharge near the powered anode and no emission near the water surface are an indication that the positive charge accumulation on the water has not completely compensated the electric field in the gap.This is similar to a specific discharge regime reported in [32,33], which was related to electric field rearrangement in the gap due to the residual positive ion density in the volume.On the other hand, for the negatively pulsed case, the water surface is the anode, i.e. the electric field strength compensation by surface charge accumulation on the water is done by electrons, which can cross the gap much faster.This could be an explanation for the stronger second discharge at 1530 ns for the negatively pulsed operation, see figure 4(b) for 0 min treatment time.Moreover, the positively pulsed discharge shows a lower spatial stability, resulting in occasional discharges on the side of the main channel.
As the treatment time (and therefore the water conductivity) increases (here at 8 min), additional emission is observed during the plateau of the HV pulse both for positively and negatively pulsed operation, shown in the bottom rows of figure 4. While the intensity of the long discharge in the middle of the pulse is much weaker, it remains visible during the pulse and its intensity increases during the treatment time.This can be interpreted as the complete reduction of surface charge accumulation after the initial discharge, since the charges approaching the water surface are transported away from the surface due to the increased conductivity of the water.Therefore, there is a conductive channel present during the ontime of the applied HV, which is leading to a stable discharge during the complete HV pulse plateau (in accordance with the current measurements, see figure 6).The spreading on the water surface seems to be larger for the positively pulsed discharges, like visible also in figures 3(a) and (b).Furthermore, the more intense discharge at the end of the pulse for the negatively pulsed operation at 0 min treatment time disappeared, which is another indication in favor of the aforementioned argument, since a higher water conductivity will reduce the negative charge accumulated on the water surface.
The imaging of the AC powered discharge at the beginning of the treatment (top) and after 15 min of operation time (bottom) is shown in figure 5 for the beginning, the middle and the end of the discharge occurring only at the negative half-cycle of the applied sinusoidal HV waveform.It was performed with a gate time of 100 ns.The behavior does not significantly change over the treatment time, suggesting that the discharge does not change modes over time, remotely related to a DC driven atmospheric pressure glow discharge [17].The discharge is present between the charged electrode and the water surface with a minimal discharge spreading on the water surface.A similar morphology was reported during the negative cycle of an AC powered discharge above a water surface [34,35].
The imaging shows that the morphology of the pulsed operated discharge significantly evolves over the treatment time, whereas the morphology of the AC operated discharge does not.This is the case for other discharge characteristics as well, which will be discussed next.

Pulsed discharge
Figure 6 shows the measured current (in color) and voltage (black) during the plasma treatment of initially demineralized water, for a positive and a negative unipolar pulsed discharge.In this article the colors red and blue are used for pulsed positive and negative discharges, respectively.The color purple is used for the AC powered discharge, which will be discussed in the next subsection.The current and voltage traces shown in this article are an average over 100 discharges.They do, however, represent well the signals recorded in a single discharge.
A discharge current peak is measured at the rising and falling slope of the pulse.This DBD-like behavior is present for all treatment times.At the falling slope the positively and the negatively pulsed discharges exhibit differences in accordance with observations from the imaging (figure 4).At the beginning of the treatment time, the positive discharge shows a minimal conducting current at slope, in accordance with only faint light emission observed.This indicates that the electric field generated by the surface charge on the water is not sufficient to start a full back-discharge, but only a weak reigniof the afterglow (i.e.volume preionization) from the previous discharge at rising slope [32].The negatively pulsed discharges show a significant conductive current during the voltage slope at the end of the pulse, correlated with the light emission observed in imaging.Different discharge behavior for different discharge polarities indicates a significant difference in the charge decay on the water surface depending on the polarity, where the negative charges are more abundant at the end of the pulse.Focusing for a moment only on the processes in the gas phase, as known from basic discharge physics, this effect can be expected because of the higher mobility of electrons with respect to ions, i.e. electrons cross the discharge gap much faster than the ions, which cannot make it in 1.5 µs across the complete gap distance.
At the beginning of the treatment time a current signal is not present during the pulse, however, already during the first minute of treatment it starts appearing, as visible in figure 6.This is the onset of the electrolysis regime, as will be further discussed in section 5.5.It is characterized by a current that rises in amplitude over time and light emission on the water surface during the discharge (figure 4).This discharge and the accompanying current signal last during most of the pulse, approximately 1.4 µs.A similar observation was also reported in [36] and [22].Here we have to distinguish between the different behavior of positive and negative operation, i.e. linear increase of current flow during positive operation (electrons move toward the metal pin), while almost constant current flow during negative operation (electrons move toward the water surface).

AC powered discharge
Figure 7 shows the measured current during the plasma treatment for the AC-powered discharge.At the ignition of the discharge, the applied waveform loses its sine shape, which is an effect induced by the properties of the power supply.
The discharge only appears in the negative half-cycle of the applied potential for these operating conditions.This is an effect that has been observed previously, with the explanation that unlike typical discharges between conductive (metallic) electrodes in which the counter electrode served as a cathode, in this case the water does not provide a sufficient amount of secondary electrons to sustain the discharge, due to its significantly lower secondary electron emission coefficient when compared to metallic electrodes [17,37].The voltage required for breakdown and generation of a discharge is therefore lower in the case of negative voltage polarity, which explains why discharges occur only at the negative half cycle of the voltage waveform in this study.Discharges in both polarities were obtained for higher applied voltage, however that mode is not included in this study.The long duration of the discharge of almost 10 µs, as visible in the I-V profile, supports this observation.
While the current signal (discharge duration) is much longer than for the pulsed discharge, the peak current is much lower, having a value not higher than 10 mA throughout the treatment time.Comparable currents are reported in the work of Diamond et al [34], where a discharge current is measured at the positive and negative cycle of a 10 kHz, 10 kV peak-topeak AC-powered discharge.This work reported current peaks with a length of ≈30 µs and amplitudes of ≈20 mA.Such long discharges point toward the interaction of plasma and water in which the water is effectively a resistor, the system thus promoting electrolysis.
The pulsed and the AC discharges evolve differently over the treatment time, the former exhibiting a transition between regimes and the latter not.In the following subsections other electrical properties will be examined for both pulsed and ACpowered discharges.

Average dissipated power and transferred charge
In figure 8, the average dissipated power (equation ( 1)) is plotted as function of the treatment time of the water.The position of data points with respect to treatment time should be taken with an error bar of ±15 s, caused by uncertainties due to the synchronization of the measurements.
A clear difference is visible between the pulsed and the AC discharge operation.The AC discharge settles on an almost constant average power (therefore also energy per pulse) from the beginning of the treatment time, showing a slow evolution of the system without shifts in regimes of operation.The power dissipation of the pulsed discharge is increasing faster than linearly during the treatment time, to a point where a saturation of the dissipated power is observed.This is the effect of the limited power that the power supply can deliver, which leads to a drop in the applied voltage as the current during the pulse increases.Part of the reason why the power dissipated by the AC discharges does not significantly change over time could also be in the design of the power supply.As the treatment time progresses, the voltage supplied by the source drops from 6 kV amplitude at the start to 3 kV amplitude at min 1 to  2 kV amplitude at min 60, as the discharge current increases from 5 mA at min 1 to 10 mA at min 60.Between min 1 and 60 the power is constant.
For pulse-driven discharges, when separating the energy dissipated during the DBD-like discharge (at the beginning and the end of the pulse) and dissipation during the pulse, the average power dissipation originating from the discharge at the rising slope increases approximately linearly during the treatment but remains below 1 W throughout the treatment time.The additional energy dissipated during the discharge occurs during the long discharge in the middle of the pulse that features a rising current throughout the treatment time, which becomes the largest contribution to the total dissipated power in the discharge gap.
Figures 9 and 10 show the absolute value of charge per voltage pulse (period) and the integrated charge transferred to the water target in pulsed and AC discharges.The charge per voltage pulse (period) has been obtained by integrating the discharge current over the voltage pulse (period) and the integrated charge is the charge per pulse (period) accumulated over the treatment time.
In accordance with the dissipated power , the AC driven discharges show a stable, low (in the range of 10 nC) value of charge exchanged per period during the entire treatment time, making the integrated charge a linear function of the treatment time.
The pulsed discharges both show a non-linear increase in the transferred charge per pulse, just as measured for the dissipated power.The biggest contributor to this large amount of charge transferred per pulse (order of 100 nC) is the long discharge in the middle of the HV pulse, not the DBD-like behavior during the rising and the falling slopes of the pulse.The integrated charge consequently also rises in a non-linear fashion, reaching higher levels within a few minutes than the AC discharge does over one hour of operation.
Significant differences were presented between the pulsed and the AC-powered discharges.However, the subsection 5.6 will show how they can still be compared.

Charging of the liquid surface
The transferred charge in figure 9 changes as a function of the treatment time, raising the question about the behavior of charge once it reaches the water surface.
Figure 11 shows the measured potential at the surface probed at a distance of 3 mm from the center of the surface discharge when negative pulses are applied to the powered electrode.The radial position of 3 mm from the axis of symmetry of the discharge was chosen to prevent the formation of a direct discharge between the needle and the HV probe, to not affect the discharge.The measurement proved to be more difficult for the positively pulsed discharges, where the surface discharge channels connect with the HV probe tip, thus influencing the overall discharge.The analysis of data includes subtracting the capacitive potential due to the applied pulse from the measured The graph shows the evolution of the surface potential in a single voltage pulse, for four different treatment times.The magnitude of the measured surface potential is several hundreds volts at the beginning of the treatment time to kilovolts at 8 min of treatment time.During the treatment time the water surface area covered by the discharge grows as well, by a factor of approximately 3 between min 3 and min 8.
The charge deposition on the interface happens mostly during the long discharge in the middle of the voltage pulse, rather than during the short discharges at the beginning and the end of the pulse.It can, thus, be concluded that this potential buildup plays an important role in the plasma-driven electrolysis, especially for longer treatment times.The data on surface potential during plasma-water interaction is scarce [38] and the question still remains what is the effective potential that should be taken into account when trying to calculate electrolysis-driven chemistry, as the surface potential exists only for the length of the pulse when operating the discharge at 5 kHz repetition frequency (1.5 µs over 200 µs in this case).It is, however, clear that the water surface charges and discharges periodically with the applied voltage, that the buildup of charge over multiple voltage cycles has not been observed and that the resulting surface potential is within the order of magnitude of the applied voltage.
When observing the surface potential after the applied pulse (t > 1.5 µs), the timescales at which the surface potential falls back to its 'uncharged' state (exponential time constant), decreases as the treatment time increases from 5 µs at the start, to 250 ns after 10 min of treatment.The faster charging and discharging with the passing treatment time could be related to the increased conductivity and the decreased charge relaxation time of the water.

Transition from DBD-like to electrolysis regime
The Maxwellian relaxation time (equation ( 2)) is a property of materials that indicates how long charge in that material needs to respond to the externally applied electric field.It is given by where ε 0 is the vacuum permittivity, ϵ r the relative permittivity of the medium and κ the conductivity of the medium [39].This equation describes the time required for excess charge carriers to distribute the charge to a value of 1/e in a material.Excellent insulators, like glass, have Maxwellian relaxation time in the range 1-10 3 s (depending on the type of glass), while the same property for excellent conductors like copper is approximately 10 −19 s. Figure 12 shows the Maxwellian relaxation time for the water used in the experiments in this work, as a function of the treatment time.It has been calculated taking into account the changing relative permittivity with the increasing water temperature according to [40].The dashed line represent the duration of the DBD-like event at the rising slope of pulsed discharges in this work (100 ns).After approximately 1 min of treatment, the Maxwellian relaxation time is lowered to the duration of the applied voltage pulse.The AC-driven discharges are longer than the Maxwellian relaxation time almost immediately after the start of the treatment.
The Maxwellian relaxation time of the water in this work falls between the typical values for good conductors and good dielectrics.In our system both a DBD-like regime and the electrolysis regime were observed: the DBD-like regime is present for phenomena that occur on short timescales, i.e. during the discharges at the rising and falling slopes of the HV pulses, while the electrolysis regime occurs for processes with typical timescales larger than the charge relaxation time, i.e. the AC driven discharge and, after some minutes of treatment, the long-duration discharge during the HV pulse for pulsed operation, as seen in figures 6 and 7.
The incomplete transition to the conductive regime of the treated water is also exhibited in the measurements of surface charging in section 5.4.When interacting with conductors, wide surface discharges are not typically formed.However, in our case even for large treatment times a wide surface discharge on the water surface still forms, concluding that the surface resistivity is still quite substantial.

The interplay between water conductivity and discharge properties
The pulsed and the AC discharges in this work operate at different average dissipated power, their behavior over time is different and finally they exhibit different regimes of interaction with water.However, figure 13 shows that their resulting effect on the treated water can be scaled with the total dissipated energy in the discharge, regardless of the form of the driving voltage.This is one way in which it may be possible to compare different experiments involving plasma-water interaction across different setups in different research groups.This also suggests that the densities of relevant gas species to be dissolved in water (HNO x ) are governed by the dissipated energy.Janda et al [41] have indeed shown that the density of HNO 2 is linearly correlated with the energy density in the discharge.
Another possible scaling property across differently driven discharges is the total transferred charge, as shown in figure 14.The values for the total transferred charge are the same data as shown in figures 9 and 10.The conductivity of the water scales very well with the total transferred charge regardless of the polarity or the driving voltage form.
Both instances of scaling of the conductivity across the different regimes of plasma-water interaction suggest that for the resulting density of active species in the water it is far more important which species are produced in the gas phase than how plasma interacts with water.

Gas temperature
Since the relevant quantities like power and charge change on a timescale of minutes, it is essential to analyze the temporal evolution of the gas temperature during the treatment time.To this end, the rotational temperature was determined from the emission from the SPS (C, ν ′ → B, ν ′ ′ ) of nitrogen.The rotational temperature from nitrogen molecules obtained in this way can be considered as an upper limit of the gas temperature because, especially for atmospheric pressure discharges, the lifetime of excited nitrogen molecules in the SPS band is typically longer than the time required for thermalization of the rotational states [42].The following figures at times miss data points of the rotational temperature.This is due to insufficient signal/noise ratio to be able to fit the data with acceptable accuracy.
For pulsed discharges the rotational temperature was measured throughout the treatment time at the rising slope of the pulse and 800 ns after the start of the pulse, shown in figure 15.In AC discharges the rotational temperature was measured at three moments during the discharge: 0.5 µs, 1.5 µs and 3.5 µs after the beginning of the discharge, see figure 16.
For both the negatively and positively pulsed discharge the temperatures increase as the treatment time of the water increases.This holds for the discharge at the start of the pulse and the discharge during the pulse well.The increasing temperature corresponds to the rise in average power dissipation from figure During the plasma treatment of the water, the temperature of the water increases as While the bulk of the water increases from ≈20 • C to ≈40 • C during treatment4 , evaporation visible above the water surface suggests that the temperature near the surface is higher.Different behaviors of the temperature profiles are observed during the short discharges at the beginning of the pulse when the discharge behaves like a DBD (figure 15(b)) and the temperature measured in the middle of the pulse, when the electrolysis regime is dominant (figure 15(b)).The spatial distribution of temperature along the axis between the tip of the charged electrode and the liquid surface is different in those two cases-during short discharges the temperature distribution seems approximately uniform axially, while for the long discharges during the majority of the pulse the coldest point is at the water surface.This is expected, as the water acts as a heat sink.
The temperatures measured during the pulse reach higher values (up to 2500 K in the middle of the gap) than the temperature measured from the emission from the discharge at the start of the pulse.This is in line with the power measurements, as the majority of the power is dissipated during the pulse.In the work of Hamdan et al [36], temperatures are reported on a pulsed discharge in contact with a water surface, which increases from 700 K to 1300 K for a positively pulsed discharge during the 20 min treatment of the water.The negatively pulsed discharge showed an overall lower temperature, which increased from 550 K to 1100 K.
In other work of Verreycken et al, Titov et al and Bruggeman et al [37,43,44], OES has been performed on a setup with a comparable geometry, by applying a positive DC voltage to a pin electrode above a water surface.Temperatures in the gap are measured between 2000 K and 3000 K.While a constant voltage is applied, the amplitude of the discharge current reported is 20 to 65 mA, which would result in values for the average dissipated power with a similar order of magnitude as shown in figure 8.
For the AC-powered discharge, the temperature stabilizes to a nearly constant value from the beginning of the treatment time, as can be observed in figure 16.This complies with the power measurement as well.
Importantly, the temperature near the water surface is lower than near the needle electrode, which is in line with the results in the pulsed discharges when measuring the temperature during the long discharge in the middle of the pulse.This is understandable, as the water acts as a heat sink independently of the applied voltage form.
A general conclusion is that the gas temperature rises during treatment time for pulsed discharges, while it stagnates for AC discharges.This behavior correlates with the trends in the dissipated power for AC and pulsed discharges and the behavior of the current flown through the plasma.This suggests that the heating is mostly resistive, which is further corroborated by the dependence of the rotational temperature on the total dissipated energy, as shown in figure 17.Another possible source of heating is an increased amount of water vapor in the gas phase, as the water heats up during treatment.The work of Komuro and Ono [45,46], shows that the gas temperature increases as the humidity increases, which is a mechanism related to the accelerated rate of vibrational relaxation.

Conclusions
A polydiagnostic study is performed on plasma interacting with water.A discharge is generated in ambient air above a static water surface by applying positive or negative HV pulses or an HV AC waveform to a needle electrode 2 mm above the water surface.Treatment of the water reduces the pH value of the water to ≈3 and the conductivity of the treated water increases up to 200 to 400 µS cm −1 .
The pulsed and the AC driven discharges show different operation regimes.When using pulsed operation, the interaction regime between plasma and water evolves over time, from a DBD-like regime to a hybrid regime comprising of DBD events at the beginning and the end of the pulse and the electrolysis regime in the middle of the pulse.This is accompanied by a fast rise of the dissipated power over the treatment time, where the dominant contribution comes from the electrolysis regime.In pulsed operation (at the start, middle, and end of the pulse), the discharge current is 1-2 orders of magnitude larger than the discharge current for AC operation.The AC driven discharges exhibit only the electrolysis regime but do not significantly evolve over the treatment time.
The two pulsed and the AC mode of operation evolve differently over time, but it was found that the gas temperature as well as the resulting water conductivity are correlated with the total dissipated energy for all three operation modes.This is one way in which it may be possible to compare different experiments involving plasma-water interaction across different setups in different research groups.For the gas temperature this indicates that gas heating is primarily resistive.For the water conductivity, this means that the gas phase production of relevant species (HNO x ) to be dissolved in water depends primarily on the dissipated energy.A strong correlation was also found between the resulting water conductivity and the transferred charge in the gas phase.In future research, the correlations between total dissipated energy, charge and conductivity as a global measure for PTW reported in this work have to be put to test using different plasma sources.

Figure 1 .
Figure 1.A schematic overview of the experimental setup used for the results presented in this report.The discharge between the needle and water surface is generated either (i) by applying a unipolar pulse (+ or −) with an amplitude of 6 kV, pulse length of 1.5 µs and frequency of 5 kHz or (ii) by applying an AC HV sine waveform with 12 kV peak-to-peak voltage and a frequency of 30 kHz.

Figure 2 .
Figure 2. pH (a) and conductivity (b) as function of the treatment time of the water, measured after switching of the HV and stirring the treated water.Each data point represents an individual measurement.(+) denotes the results when positive pulses are applied to the charged electrode, (−) is for the negative pulses and AC for AC-driven discharges.In the graphs the data points for treatment times from 20 min to 60 min are shown on a different timescale than the data for the first 20 min.

Figure 3 .
Figure 3. 'Long exposure' pictures of pulsed (a), (b) and AC powered (c) operation taken with the camera under a slight angle (i.e.not side-on) to visualize the discharge emission on the water surface as wells as the channel in the 2 mm air gap.A reflection of the emission in the volume is visible on the water surface (especially pronounced in (b)).

Figure 4 .
Figure 4. Averaged iCCD images for a positive (a) and negative (b) pulsed discharge obtained with 5 ns gate width (approximately 800 accumulations per image) and at 860 V gain (out of 1000 V maximum).The images are obtained at 30 ns (left, rising slope of the HV pulse), 750 ns (middle, plateau of the HV pulse) and at 1530 ns (right, falling slope of the HV pulse) after the start of the applied HV waveform, at the start of the treatment (top) and after 8 min of plasma treatment (bottom).The I-V characteristic in figure 6 can provide more clarity on the exact shape of the pulse.The white dotted and dashed lines represent the contours of the needle electrode and the water surface, respectively.The intensity is shown in arbitrary units and in a logarithmic scale.

Figure 5 .
Figure 5. Averaged iCCD images captured with 100 ns gate width for 0.2 µs (left), 2 µs (middle) and 7 µs (right) after the start of the current peak of an AC powered discharge at the beginning (top) and after 15 min of treatment (bottom).The gain was 860 V out of 1000 V maximum.Each image represents approximately 4800 discharges.The white dotted and dashed lines represent the contours of the needle electrode and the water surface, respectively.The intensity is scaled logarithmically and is given in counts.

Figure 6 .
Figure 6.Applied voltage and measured current for different treatment times for a positive (top) and negative (bottom) pulsed powered discharge.

Figure 7 .
Figure 7. Applied voltage and the current measured at the high voltage electrode for different treatment times for an AC powered discharge.

Figure 8 .
Figure 8.Average power dissipated as function of treatment time, determined from the measured electrical signals.

Figure 9 .
Figure 9. Transferred charge per voltage pulse or period in positive, negative and AC discharges.The value for AC discharges is approximately 10 nC and constant throughout the treatment time.

Figure 10 .
Figure 10.The integrated charge transferred to the water target in positive, negative and AC discharges.

Figure 11 .
Figure 11.Electric potential of the liquid surface for a negative pulse measured at 3 mm distance from the center of the discharge gap, for different treatment times.6 kV is applied to the needle electrode between 0 µs and 1.5 µs.

Figure 12 .
Figure 12.Charge relaxation time for the pulsed and AC powered discharge as function of treatment times.Values are based on conductivity and temperature of the water.The dashed line indicates the length of the DBD-like discharge current peak at the beginning of a pulsed discharge.The duration of the current signal for the AC-powered discharge of several µs lies outside this graph.

Figure 13 .
Figure13.The dependence of the resulting conductivity of the treated water (data from figure2) on the total dissipated energy, for pulsed and AC driven discharges.

Figure 14 .
Figure 14.The dependence of the resulting conductivity of the treated water (data from figure 2) on the total transferred charge, for pulsed and AC driven discharges.

Figure 15 .
Figure 15.Rotational temperature obtained for the pulsed discharge from a theoretical fit of the excited N 2 (C, ν ′ → B, ν ′ ′ , ∆ν = −1) transition through spectral data at different positions with respect to the needle electrode (z = 0).The figure displays the result from the initial discharge emission (a) and the discharge during the pulse measured 800 ns after the start of the pulse (b).z = 0 mm denotes the area next to the high voltage pin, while z = 2 mm is just above the water surface.

Figure 16 .
Figure 16.Rotational obtained for the AC powered discharge from a theoretical fit of the excited N 2 (C, ν ′ → B, ν ′ ′ , ∆ν = −1 ) transition through spectral data at different times after the start of the current peak.z = 0 mm denotes the area next to the high voltage pin, while z = 2 mm is just above the water surface.

Figure 17 .
Figure 17.Rotational temperature as a function of the total dissipated energy for pulsed and AC driven discharges.