Electron density in a non-thermal atmospheric discharge in contact with water and the effect of water temperature on plasma-water interactions

In this study the electron density of an atmospheric plasma generated between a pin electrode and a water surface is measured by determining the Stark broadening of H α and H β emission lines. Comparable values for the electron density are achieved using the H α and H β broadening obtained in separate measurements. During the temporally evolving system, increasing electron densities are measured of 0.5−3⋅1015 cm−3 during the plasma treatment of 5–10 min. The effect of the water temperature on plasma-water interaction is investigated by heating the water to ≈70∘ C prior to the measurements. This resulted in higher gas temperatures during the discharge up to 2500 K and 4000 K for positively and negatively pulsed discharge, respectively. Furthermore, an earlier increase of electron density and conductivity of the water is measured for the preheated experiments. The humidity of the gas is likely to be an important parameter causing the observed results.


Introduction
In the last decades, the interplay between plasmas and water has been studied extensively [1,2].Dissipating energy in ambient air by generating a plasma creates a gaseous medium with reactive radicals and charged species.While the chemistry in a plasma itself is already complex, the presence of a water, or liquid in general, makes it even more so.As reactive oxygen and nitrogen species in the air interact with the liquid, * Author to whom any correspondence should be addressed.
Original Content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
the chemistry of the water becomes interconnected with the gas phase chemistry.Although plasma chemistry occurs on a short timescale, the interaction leads to a water solution containing long-lived species with lifetimes of days to months [3].Plasma can thus be used as a tool to modify the composition of the water, without adding chemical components other than the species from the ambient air.This plasma treated water (PTW), or alternatively plasma activated water, finds potential in many fields.The presence of nitrates in the PTW adds nutritious properties to the water, making it interesting in agriculture where it can be used to stimulate seed germination and plant growth [3][4][5][6][7][8].Here the absence of environmentally harmful fertilizers makes PTW potentially an ecologically friendly alternative to current methods [9].Another aspect of interest in PTW are the disinfecting properties, which not only allows it to serve as potential pesticide in agriculture, but this quality finds its uses in other fields as well.In the biomedical field, PTW is studied as a tool for sterilization of medical equipment and inactivation of bacteria and cancer cells [10][11][12][13].As the different applications make use of different aspects of PTW, understanding the physical and chemical processes, in the liquid but in the gas as well, are required to achieve a controllable method for creating PTW with the desired qualities.While in my other work the liquid chemistry is primarily studied, this work focuses on the gas dynamics.
There are many different configurations investigated where a gas discharge directly or indirectly interacts with water.The results presented in this work continue on a setup presented in earlier published work [14], where a polydiagnostic study on an ambient air discharge in a pin to water configuration shows a transition of the system, from a DBD-like (dielectric barrier discharge-like) configuration where the water serves as insulator, to an electrolysis driven system as the conductivity rises during the treatment of the water.An important parameter in plasma dynamics and chemistry is the density of electrons.Knowing this parameter allows to make a better comparison with models or other experiments.
The focus of this work is the electron density during the plasma-water interaction, where the system evolves in time.Spatially resolved electron density is obtained from the Stark broadening effect of the Balmer α and β emission lines.Additionally, the effect of the temperature of the water in this work is studied by heating the demineralized water prior to the experiments.While the temperature change of ≈40 • C is small with respect to the temperature changes during the discharge event, it does have significant impact on the charge density.

Stark broadening
In cold plasmas electrons have a much higher temperature than ions and are moving much faster than other charged species present.The density of electrons in a plasma is an important parameter to characterise the dynamics of the discharge and is therefore useful to obtain.
In this study, a non-evasive method for the measurement of electron density is used.Optical emission spectroscopy (OES) is used to determine the Stark broadening on the Balmer H α and H β emission lines.The emission of these lines lines correspond to the electrons transition of excited hydrogen states with n = 3 → 2 (visible at a wavelength of 656.27 nm) and n = 4 → 2 (visible at a wavelength of 486.14 nm), respectively.This broadening mechanism is the result of charged particles interacting with the emitting particle.The Stark effect is caused by the electric field of perturbing particles and affects the energy levels of the emitting atoms, breaking the degeneracy of their atomic levels, and hence emit light at a slightly different wavelength.The combined effect of many emitting particles which slightly perturbed energies results in the broadening of the emission line.It is assumed that the typical duration of a collision of the charged particle and the emitter is much faster than the correlation time of the dipole moment of the emitting particle.This results in a stronger effect on the dipole moment when multiple collisions occur, and hence the broadening is dependent on the frequency of the collisions.Since the electrons have much more mobility than other charged species in the discharge, the Stark broadening is almost completely caused by the electrons.
The relation between the electron density and Stark broadening is different for the H α and H β lines.In the review of Nikiforov et al [15] the Stark broadening of hydrogen lines is studied of an atmospheric pressure plasma jet.In this work, n e as function of the Stark broadening is given as for the H α line, and as n e = 10 16 (∆λ S /0.9467) for the H β line.Note that the former expression uses, ∆λ A S , which is the full width at half area (FWHA) in nm, while ∆λ S uses the FWHM in nm.However, in this report the width of the Stark broadening is determined by deriving the Lorentzian component of a pseudo-Voigt profile fitted through the data.For a Lorentzian profile the FWHA is equal to the FWHM.

Other broadening mechanisms
Besides Stark broadening there are other broadening mechanisms that affect the emission lines as well.Important broadening mechanisms that might need to be considered besides Stark broadening are Doppler broadening, collisional broadening, resonance broadening, van-der-Waals broadening and instrumental broadening.These mechanisms are briefly explained in this section since their FWHM is required for the fitting procedure.To illustrate the relevance of the individual mechanisms in our setup, estimated values for the different broadening widths during the experiments in this work are presented in table 1.
The Doppler broadening is the broadening effect caused by thermal movement of the emitting particle, resulting in a slightly different emission wavelength.The combined effect of the emitting particles leads to an emission line with a Gaussian distributed wavelength.The FWHM of an intensity peak broadened by the Doppler mechanism is given by [16], with k B the Boltzmann constant, m the mass of the emitter, c the speed of light in vacuum, T the gas temperature and λ 0 the wavelength in nm, corresponding to the energy difference of the transition.Collisional broadening of a line arises due to inelastic collisions of particles in the gas with the emitting particle.This broadening of the emission line can be expressed as For transitions that are dipole-coupled to the ground state, interaction between ground state particles and emitting particles leads to resonance broadening of the emission line.This broadening mechanism scales with the number density of ground state particles in the gas.However, the fraction of ground state hydrogen particles in the gas is sufficiently low that this results in a negligible contribution of the resonance broadening to the total width of the peak.
Van-der-Waals broadening is the effect induced when the emitting particle is perturbed by van der Waals forces from surrounding particles.The FWHM of this Lorentz broadening is given by [17] ∆ vdW = 5.925 with pressure p in mbar, gas temperature T in K and coefficients K l and K p , which depend on the line and perturber, respectively.The line dependent coefficient K l used to estimate the width are 2.695 • 10 −8 for H α and 2.757 • 10 −8 for H β .The perturber dependent coefficient K p = α 2/5 /µ 3/10 , depends on the mean polarizability of the neutral interacting particle and on the reduced mass of the atom-perturber in amu.
The instrumental broadening is considered as a Voigt profile, which is a combination of a Gaussian and Lorentzian profile.This results in an emission line with a combined Gaussian, ∆ instr,G , and a Lorentzian, ∆ instr,L , broadening component, which is given in section 3.
As can be read from table 1, especially the van-der-Waals broadening and Doppler broadening take proportions that should not be neglected.In addition to this, the instrumental broadening is of a comparable magnitude.For the FWHM of the Stark broadening of H α to exceed the other broadening mechanisms in size, the electron density needs to be > 1.5 • 10 15 cm −3 for T = 500 K and > 0.5 • 10 15 cm −3 for T = 2000 K.This threshold is lower for the H β line, but the signal-to-noise ratio for this emission line is also lower, which is another bottleneck of this method.

Fitting profile
In this study, the electron density is determined by fitting a pseudo-Voigt function rather than a normal Voigt profile through the H α and H β emission lines.While a normal Voigt profile requires a convolution between the Gaussian and Lorentzian part of the function, which complicates the fitting procedure, the pseudo-Voigt function is a linear combination of a Gaussian and Lorentzian function which approximates this Voigt profile, as given in equation ( 6) [18], (6) with f G and f L the Gaussian and Lorentzian function, respectively, and ∆ pV the FWHM of the pseudo-Voigt function, f pV .This ∆ pV is determined by the Lorentzian and Gaussian FWHM, ∆ L and ∆ G , respectively, by expression 7, The η term used in formula ( 6) is given by The [18] states a maximum of 1.2% deviation between the pseudo-Voigt with respect to an normal Voigt function.In the fitting process of the pseudo-Voigt, the FWHM of the Lorentzian part of the pseudo-Voigt profile is optimized, while keeping the Gaussian part fixed.Since the combined Gaussian FWHM of the broadened line is expressed as this is possible when the gas temperature is known.In this work, the temperature is determined by a series of OES measurements on the rotational band structure of a nitrogen transition, which is presented later in section 4.1.The obtained Lorentzian FWHM of the broadened line is sum of the individual Lorentzian broadening components, from which the Stark broadening can finally be isolated.

Experimental setup
The interaction of a discharge in contact with water is investigated by generating an electric discharge between a tungsten needle electrode and a water surface.A schematic representation is shown in figure 1.The water is contained in a Petri dish with ≈80 ml volume and radius of 87 mm, with an stainless steel electrode embedded ≈13 mm under the water surface.This plate electrode is connected to the ground through a hole in the bottom of the Petri dish.The needle electrode is positioned 2 mm above the water surface and a discharge is formed in the gap between the electrode and the water surface in ambient air.
In this study, a discharge is generated by applying unipolar positive or negative pulses with an amplitude of 6 kV, a pulse length of 1.5 µs and a repetition rate of 5 kHz to the needle electrode.The applied pulses are generated by a high voltage pulse generator (DEI, PVX-4110).The function of the pulse are supplied by an arbitrary waveform generator (KEYSIGHT 33 500B) and the high voltage is supplied by a high voltage amplifier (Trek, 10/10B-HS).This setup is similar to the pulsed discharge investigated in earlier work [14].In figure 2  Prior to generation of the discharge, the Petri dish contains demineralized water.For the investigation of the water temperature on the discharge dynamics, the demineralized water is preheated in a microwave oven.At the start of the measurements, the temperature of the water is ≈20 • C for the unheated conditions and ≈60 • C -70 • C for the preheated conditions.This results in four different measurement conditions displayed in this report; a positively pulsed discharge using unheated water (1), a positively pulsed discharge using preheated water (2), a negatively pulsed discharge using unheated water (3) and a negatively pulsed discharge using preheated water (4).
For temperature, pH and conductivity measurements of the water, a pH-conductivity meter (Mettler Toledo, SevenExcellence S470-Std-K) is used.After an experiment, the water in the Petri dish is stirred and a sample is taken from the treated water, from which the pH, conductivity and temperature are measured.Thus, each obtained data point shows a separate measurement and represents the values for the bulk of the water.Furthermore, since the conductivity of water depends on the water temperature, the values in this report are corrected such that they display the value at 25 • C. To determine spatially and temporally resolved values for the gas temperature and electron density, an OES setup is used.The discharge emission is focused on a vertical slit of 50 µm on on end of a monochromator (Jobin-Yvon, HR 1000).The monochromator contains a grating with 1200 grooves/mm.On the other end of the monochromator and iCCD camera (Andor, iStar) is positioned to measure the diffracted light.The images obtained by the camera cover a spectral range of ≈ 10 nm and the spatial axis covers the complete discharge gap.The pixel resolution of this setup (depending slightly on the investigated emission line) is ≈0.01 nm on the spectral axis.The FWHM of the instrumental broadening is determined by fitting a Voigt profile through the light of a Mercury-Argon calibration lamp (Avantes), where the other broadening mechanisms are assumed negligibly small.The FWHM of this pseudo-Voigt function is 0.04 nm, which can be considered as the spectral resolution of the spectrometer.The Gaussian and Lorentzian components of this instrumental broadening are 0.016 nm and 0.034 nm, respectively, which are used in the analysis to determine the Stark broadening of the hydrogen lines.The values of the gas temperature and electron density are measured at either the start of the pulse (0 ns) or 400, 800 or 1200 ns after the start.
To obtain the gas temperature, the rotational band structure of the emission of the N 2 (C, ν ′ → B, ν ′ ′ , ∆ν = −1) transition of the second positive system of nitrogen, visible around 357 nm, is measured.A theoretical line is fitted through the obtained spectra to determine the rotational temperature of nitrogen, which is an approximate representation of the gas temperature.An example of a fitted spectrum of the rotational band through measured data can be seen in figure 3.
To obtain the electron density, the OES setup is used to investigate the broadening on hydrogen lines.For the electron density measurements, a camera gate of 50 ns is used.The spectral images contain the accumulated emission of multiple discharges, ranging from ×2500 (0.5 s) to ×20 000 (4 s), time, varying several minutes, for the positively pulsed discharge to form a stable discharge between the needle and the water surface.depending on the intensity of the hydrogen line.In this work, the electron density is measured by fitting a pseudo-Voigt function through the Balmer H α and H β lines, as described in section 2. The emission of these lines lines correspond to the electrons transition of excited hydrogen states with n = 3 → 2 (visible at a wavelength of 656.27 nm) and n = 4 → 2 (visible at a wavelength of 486.14 nm), respectively.In figure 4 the fitting of this profile is performed on the emission of the hydrogen α and β emission lines.As the intensity of other peaks in the proximity of the hydrogen lines is not negligible, the spectral range used for the fitting of the line is manually restricted to limit the influence they have on the fitting procedure.

Results
While the electron density measurements are the main focus of this work, the interpretation of the results benefit from a more comprehensive depiction of the discharge operating in this setup.As the experiments are performed on a similar setup used in earlier published work [14], which is referred to multiple times, this section begins with a brief summary of the experiments and results presented in this citation.At the start of the experiments, the Petri dish is filled with distilled water, which serves as a dielectric layer covering the grounded electrode.In ≈10 minutes of plasma treatment, the conductivity of the water increases to ≈200 µS cm −1 for this pulsed discharge, and electrolysis will take place in the water with surface charges acting as one of the electrodes.This transport of the surface charges allows a plasma to be formed during the pulse, which contradicts the characteristic behaviour of a DBD plasma.At short timescales (<50 ns) the water can be described as a dielectric barrier, but at longer timescales it behaves more like a resistor, albeit a rather bad one.This can be observed in the current signals presented in figure 2. At the start of the treatment no current is measured during the pulse, demonstrating the DBD-like behaviour, while after a few minutes of treatment a non-negligible current is present during the pulse.This current increases during the pulse when the conductivity increases, which results in increasing power dissipated in the system.A drop in the applied voltage is measured when the power source reaches its power limit.This limit can be observed in figure 5, where the average power dissipated in the system is displayed for the different operation modes.All four modes show a limit in the dissipated power at the same level around 16 W.Besides the similar power limits that can be observed, this result shows a discrepancy between the positive and negative modes, but also between the normal and preheated conditions.This will later be discussed in section 5.
While the system can still operate for several minutes with the lower voltage of the applied pulse before the discharge stops, the discharge dynamics are affected by this voltage drop.
Before continuing with the results of the electron density measurements in section 4.2, this section will first the results for the measurements gas temperature and the conductivity, pH value and temperature of the throughout the measurements.The to measure the water properties, is that they affect the gas-phase discharge.The gas temperature measurements are shown in this work since they are required to make an estimate of several line broadening mechanisms needed for the analysis of the electron density measurements.As the conductivity of the water increases during the plasma treatment, the discharge dynamics are evolving in time throughout the experiments, which is why the results will be presented as a function of the treatment time.This time is defined as the amount minutes the HV pulses are applied to the setup.

Tgas, Conductivity, pH and Twater
A series of OES measurements is performed on the nitrogen emission line at 357 nm to determine the gas temperature at the start of the pulse and during the The experiment is performed with and without heating the demineralized water prior to the plasma treatment, for a positively (red △) and negatively (blue ▽) pulsed discharge.The result of this measurement can be seen in figure 6.
Generally, the gas temperature increases with the treatment time.However, this increasing manner stops when the power limit of the HV pulse supply is reached.Typically higher temperatures are measured during the pulse, which can correlated with the conducting behaviour of the water.When a discharge is present during the pulse, a conductive path is created in the gap during which the majority of the power dissipation of the system takes place, resulting in higher temperatures.Figure 6 also shows an increased gas temperature measured in the discharge for the experiment where the water is preheated when compared to the unheated cases.Especially for the discharge during the pulse in figure 6(b), there is a clear difference between the different starting conditions.This discrepancy will be discussed later in section 5.
Properties of the water such as conductivity, temperature and pH are measured for different treatment times.In figure 7, the value for the conductivity, water temperature and pH are presented for the unheated and preheated conditions of the water.
When a plasma is formed in the discharge gap, reactive nitrogen and oxygen species are transported into the liquid phase.This is well known in literature and this phenomenon is used in agriculture, medicine, and environmental applications [6,12].The conductivity of the treated water is predominantly caused by positive H + , and negative NO − 2 and NO − 3 ions.The concentration of these species can be roughly estimated based on the conductivity [19].The results in figure 7 show a faster rise in conductivity and a faster reduction of the pH value for the preheated conditions with respect to the unheated conditions.This suggests a faster rise in the concentration of ionic NO − x species in the water.
During the treatment, the water temperature increases for unheated conditions.The non-linear increase can be correlated to the increase in power dissipation during the experiment.For preheated conditions, the decrease in water temperature can be related to the cooling down to the environment.Although the results display the water temperature of the bulk of water, the local temperature at the water surface is assumed to be higher.This is supported by the observation of steam above the water surface during the treatment.The high gas temperature during the discharge together with the observation of steam makes it plausible that the water is partly heating due to thermal conductivity between the gas and the water.However, as current measured at the grounded side and the HV side are approximately the same, Ohmic heating due to the conducting current is likely as well, considering that the temperature and power dissipation increase accordingly during the treatment of the water.

Electron density
With the temperature being measured, the Gaussian broadening of the H α and H β profiles could be determined before the fitting of the pseudo-Voigt profiles through the hydrogen lines.For a positively and negatively pulsed discharge, figure 8 shows the electron densities during the pulse in the middle of the discharge gap as function of the treatment time, based on the fitting of H α and H β profiles.The spectra of the two hydrogen lines are obtained in separate measurement sets.
The fitting of the H α profiles lead to a very similar result as the fitting of the H β profiles, which gives a form of validation of the used method.However, the results show of the different lines still display discrepancies showing that the error between different measurements is likely to be larger than the displayed error bars.The displayed error bars are based on a combination of the confident intervals of the fit and an induced error due to other broadening mechanisms when assuming a 20% error in the temperatures.As the H β emission generally wider than the H α emission line, the error due to this temperature variation remains small.However, since the intensity of the H β emission line is usually lower, the rest of the electron densities shown in this section are determined from the H α emission.
Using Stark broadening it is possible to measure temporally and spatially resolved values for the electron density, if the intensity of the hydrogen line allows it.The ICCD imaging on this system presented in [14] shows a bright DBD-like streamer discharge propagation from the needle electrode to the surface at the developing in the first 20-50 ns of the applied pulse.However, the intensity of the H α and H β emission lines during this part of the pulse is still weak.The discharge dissociates water in the gas phase, resulting in the intensity becoming high enough to measure a distinctive peak during the pulse, which increases further throughout the duration of the treatment.
In figure 9 the electron densities are plotted against the treatment time, obtained from the H α emission from the middle of the gap, 800 ns after the start of the pulse.As the water properties are changing in time, so does the discharge.This can be observed in the previous section and in other measurements, which are published in earlier work [14], but in the electron densities as well.For all four presented discharges, the electron density increases as function of the treatment time.This aligns with the rising electrical current during the pulse.As can be seen in the figure, the preheated measurements resulted in an earlier increase of the electron density than for the unheated cases.In the preheated measurements, electron densities up to 3 • 10 15 cm −3 are measured, before the power source shut down.For the unheated cases maximum densities of 2.5 • 10 15 cm −3 and 1.7 • 10 15 cm −3 are measured for the positively and negatively pulsed discharge, respectively.Densities of the same order have been measured in a pin-to-water setup in the work of Yang et al [20], where a pulsed discharge is studied in contact with a NaCl solution  (5 mS cm −1 ) in a 1.8 mm discharge gap.In the work of Yang, the electron densities are determined from the broadening of the H β emission, where densities around 1.3 • 10 15 cm −3 and Figure 9. Electron density against the treatment time for a positively (red △) and negatively (blue ▽) pulsed discharge, for unheated (filled) and preheated (empty) demineralized water.Data is obtained from the Hα emission in the middle of the gap (z = 1 mm), 800 ns after the start of the pulse.0.5 • 10 15 cm −3 are measured, near the needle electrode and in the middle of the gap, respectively.
In figure 10, the electron densities are plotted as function of the gap position, with z = 0 the position of the needle electrode and z = 2 the area near the water surface.Since not all discharges have sufficient intensity at the start of the measurement, the different discharge operation modes are compared for a system where the water has obtained a conductivity of ≈100 µS cm −1 , hence the different treatment times of the presented data.In this result, all modes show a similar effect where the density is higher near the needle electrode, and start to drop after ≈1 mm.This can be supported by the increase in the channel width of the discharge closer to the water surface.This was observed in the ICCD imaging as well, presented in earlier work [14], where the channel width starts to increase ≈1 mm from the surface for the positively and negatively pulsed discharge.
Figure 11 shows the electron densities at different moments during the applied pulse.With a delay of 400, 800 and 1200 ns, the electron density could be determined as the intensity of the discharge during the pulse increased throughout the treatment.For the positively pulsed discharge, the electron density appears to be stable during the pulse.This is remarkable since the current during the pulse for the positively pulsed discharge is increasing until the end of the pulse, as is displayed in figure 2(a).For a negatively pulsed discharge, a decrease in electron density is observed throughout the pulse.Again, this is in contrast to the measured current, which stays rather constant during the pulse.The effect the electron density present during the DBD-like discharge at the start of the pulse, which could not be measured, may still have influence on the electron densities later during the pulse.For the negatively pulsed discharge, a significantly higher current is measured at the DBDlike discharge than at the discharge during the pulse.This might result in a higher density at the initial discharge, which is still partly present in the early stage of the pulse.For the positively pulsed discharge, the current measured during the pulse becomes a similar magnitude as the DBD-like discharge at the start.It might be that the decrease in electron density from this DBD-like discharge is compensated by the increasing current during the pulse, leading to a constant appearance in the electron density.From temporally resolved ICCD imaging of the gap, no increase in the width of the discharge channel is observed for the discharge during the pulse.However, around 400 ns after the start of the pulse, the discharge that is reignited in the gap is not yet completely formed.As the fully conducting channel may not yet be formed, the direct correlation between current and electron density might not yet be at this

Discussion-the effect of water temperature
Preheating the treated water by only 40 • C-50 • C has a large effect on the dynamics of the system.Compared to the gas 11.Electron density versus the moment of exposure during the pulse for a positively (a) and negatively (b) pulsed discharge for unheated conditions.The gating of the camera is 50 ns with a delay 400, 800 or 1200 ns with respect to the start of the pulse.Data displayed is obtained in the middle of the gap (z = 1 mm).
of several hundreds to a few thousands Kelvin, a difference in water temperature at the start of the measurement between preheated and unheated conditions would seem quite small.However, when considering the humidity above the water, this temperature difference is more significant.The saturation vapor pressure in ambient air is roughly 2.3 kPa and 31 kPa for a temperature of 20 • C and 70 • C, respectively.When assuming that the gas temperature the water surface without plasma present is approximately the same as the water temperature, then the gas can contain ≈13 times as much water molecules when fully saturated.
In the modelling work of Barni et al [21], other things the evolution of the chemical composition of DBD discharge is studied for dry and humid air conditions (0% and 2.5% water concentration).The electron density for the humid air discharge does not show a significant difference compared to the dry air conditions for this DBD discharge in the timescale of 10 −6 s, which is the timescale of the pulse duration.However, the amount of negatively charged particles present in the gas at timescales of 10 −4 s does show orders of magnitude difference, which is order of magnitude of the time between separate pulses in this work.For example, an O − density of 10 4 cm −3 was modelled for dry air, compared to 10 8 cm −3 for humid air conditions.Furthermore, an OH − density of 10 9 cm −3 was present in the humid air model as well, while absent in the dry air model.Reflecting this to the experiments presented in this report, the main effect of the humidity on the discharge is not significant during a single discharge.With a period of 2 • 10 −4 s between pulses, it does have a significant effect on the amount of negatively charged species for subsequent pulses.During the streamer formation on the next pulse, remaining negatively charged particles in the air can play an important role as a source of electrons that are easily detached, which can become part of the discharge.The difference in humidity also supports the increased gas temperatures measured for the preheated conditions in 6(b).It is known that humidity plays a significant role in the thermalization of the air [22], which is why this effect can be observed especially during the pulse.In the work of Liu and Becerra [23], the effect could hardly be observed during the streamer-to-leader transition, which supports the much smaller difference between the preheated and unheated conditions observed at the start of the pulse in figure 6(a).
The results of the electron density measurements in figure 9 show a clear difference between the unheated and preheated measurement conditions, and between the positively and negatively pulsed discharge.However, they show a similar rate at which they increase.Perhaps a better way to study the role that the electron density plays in the treatment process is to correlate it to other parameters obtained in the experiments.In figures 12 and 13, the electron density during the pulse (800 ns after the start of the pulse) is plotted against the measured current at that point in time and conductivity of the water.
From figure 12 it seems that the electron density in the middle of the pulse can be correlated quite well with the measured current at the same point in time.The positively and negatively pulsed discharges the results show similar results within the error bars.However, there is still a difference between the preheated and unheated conditions.One aspect that may contribute to this is the width of the discharge channel.An attempt to determine the channel width is made using ICCD imaging.However, since the images capture the accumulated light emission from multiple discharge events, the spatial instability of the discharge lead to difficulties in determining the precise channel width, especially for the positively pulsed discharge.Still, for the negatively pulsed discharge, which is more stable than the positively pulsed discharge, the FWHM of the emission profile of the channel during the pulse was typically between 0.24 and 0.30 mm for the unheated measurements, compared to values between 0.17 and 0.24 mm for the preheated measurements, giving a difference of 30%-40% between the two modes.This suggests a difference in the area of the channel of approximately a factor  2. However, the effect of the humidity on the spatial stability is neglected in this comparison.The fact that the humidity affects the discharge channel can be supported with the modelling work on streamers of Malagón-Romero et al [24] and Starikovskiy et al [25], where an increase in humidity lead to an decrease in the channel width of a positive streamer.This finding supports the difference found between the unheated and preheated results in figure 12.
The difference in channel width can be correlated to the differences in the measured gas temperatures.High temperatures in the discharge column lead to lower densities, which results in higher reduced electric fields E/N.This phenomenon is known in literature and leads to the constriction of glow discharges [26].For the heating of the gas in the discharge, it would take ≈300 ns to affect the density over 100 µm.The temperature change within this time would result in a pressure difference within the channel.For the preheated measurements, the relative gas temperature change within this time is higher compared to the unheated measurements, which also suggests a higher pressure during the discharge.This could also contribute to the higher electron densities for the measured current.
In earlier work we observed an approximately linear trend in the conductivity when plotted versus the dissipated energy [14].In figure 14 this is plotted as well, including the preheated measurements.While there is not much difference between the positively and negatively pulsed discharge of the unheated conditions, the conductivity increase per dissipated energy is higher for the preheated measurements.The lower values for the total dissipated energy are due to the shorter treatment times.Although Henry's Law generally suggests a lower solubility for higher temperatures, it can still be used to explain the faster increase in conductivity for the preheated conditions.The NO − x species in the water are likely to originate from x species in the gas phase, which solvate easily the water.Humid conditions will lead to more HNO x in the gas phase, and thus a faster increase of the conductivity.In the work of Janda et al [27], who investigated the role of HNO 2 in plasma-activated water for a transient spark discharge in air.This showed that humid air conditions lead higher concentrations of HNO 2 in the gas phase, while the sum of HNO 2 and NO 2 remained constant.It also showed that the presence of HNO 2 resulted in higher NO − 2 concentrations in the water.

Conclusion
In this work, the discharge dynamics in the gas phase of a time evolving pin to water system is investigated.This study provides the results of electron density measurements based on Stark broadening of the hydrogen emission lines of the Balmer α and β emission line.In separate measurements, very similar results are achieved for the obtained electron density from the H α and H β emission line, showing self-consistency of this method.This method allows to obtain spatially resolved values for the electron density above 0.5 • 10 15 cm −3 in the discharge gap, provided that the intensity of the hydrogen emission is sufficient.Typically densities in the order of ×10 15 cm −3 are measured in this work.
For positively and negatively pulsed discharge, a approximately constant electron density is measured between the needle and middle of the gap.In the region between the middle and the water surface, the electron density reduces.This goes hand in hand with the increase of the channel width of the discharge observed in earlier work [14].As the system evolves in time, the electron density increases during the treatment of the water.This increase matches the electric current at the moment that the emission is recorded.
Additionally, the effect of the water temperature on the plasma dynamics is studied.The treatment process is significantly affected by heating the demineralized water by ≈ 40 • C prior to the measurement.Comparing the preheated with the unheated conditions, the former led to higher gas temperatures during the pulse and a faster increase of the conductivity of the water with respect to the latter.This has led to an earlier increase of the electron density for the preheated conditions.
In earlier work [14], a relation between the total energy dissipated by the gas discharge and the conductivity (an indication of number of NO x ) of the water was observed.Here, for the preheated conditions, the conductivity versus dissipated energy rate is significantly higher with respect to the unheated conditions.The effect of humidity is discussed as being an important parameter causing the observed differences.A higher water temperature, and thus a higher gas temperature in the period between pulses, will increase the saturation point of water in the gas phase.This increase in humidity affects the discharge chemistry and likely increases the formation of HNO x in the gas phase, leading to a faster increase in the density of dissolved species.

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 (1) by applying a unipolar pulse (+ or −) with an amplitude of 6 kV, pulse length of 1.5 µs and frequency of 5 kHz.
the voltage and current signals are shown for a positively (a) and negatively (b) pulsed discharge measured at the powered electrode.The plasma current measured at the grounded electrode (2(c)) is very similar in amplitude, shape and phase to the plasma current at the powered electrode, showing the conductive nature of the plasma-water system.

Figure 2 .
Figure 2. Voltage and current signal for a positively 1 (a) and negatively (b) applied pulse measured at different times during the experiment.To show the similarity between the plasma current measured at HV side and the grounded side, figure 2(c) shows the two signals after subtraction of the capacitive current for a single measurement of a positively powered discharge after 4 min of treatment.

Figure 3 .
Figure 3. Example of fitted spectrum through measured date for a positively pulsed discharge operating on preheated water.

Figure 4 .
Figure 4. Example of a pseudo-Voigt profile f pV fitted through experimental data of the Hα (a) and H β (b) emission lines, generated by a negatively pulsed discharge.

Figure 5 .
Figure 5. Average power dissipated in the setup as function of treatment time for a positively and negatively pulsed discharge, under normal and preheated conditions of the water.

Figure 6 .
Figure 6.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 from the emission in the middle of the gap, 1 mm from the needle electrode.The figure displays the result from the discharge at the start of the pulse (a) and the discharge during the pulse measured 800 ns after the start of the pulse (b).

Figure 7 .
Figure 7. Conductivity (a), water temperature (b) and pH value (c) of the plasma treated water as function of the treatment time.The data for the preheated (empty symbols) conditions are added to the data for unheated conditions (filled symbols), presented in earlier work[14].

Figure 8 .
Figure 8. Electron densities determined from the fitting of the Hα and H β line for the positively (a) and negatively (b) pulsed discharge at a position of z = 1.5 mm from the needle electrode.

Figure 10 .
Figure10.Electron density plotted versus the gap position with z = 0 mm at the needle electrode.Data is shown for measurements when the conductivity of the water is ≈100 µS cm −1 , for a positively and negatively pulsed discharge, operating on unheated and preheated demineralized water.

Figure 12 .
Figure12.Electron density versus measured current at 800 ns after the start of the pulse, for positively (red △) and negatively (blue ▽) pulsed discharge, for room temperature (filled icons) and preheated (empty icons) demineralized water.

Figure 13 .
Figure 13.Electron density versus conductivity of the water, for positively (red △) and negatively (blue ▽) pulsed discharge, for room temperature (filled icons) and preheated (empty icons) demineralized water.

Figure 14 .
Figure 14.Conductivity against the total dissipated energy during the treatment of the water, for positively (red △) and negatively (blue ▽) pulsed discharge, for room temperature (filled icons) and preheated (empty icons) demineralized water.

Table 1 .
FWHM values ×10 −3 nm for different broadening mechanisms, displayed for the Hα and H β transitions for a temperature range of 500-2000 K, for a gas composition of 80% N 2 and 20% O 2 at atmospheric pressure.