Chemical kinetics and density measurements of OH in an atmospheric pressure He + O2 + H2O radiofrequency plasma

This work presents experiments and modelling of OH densities in a radio-frequency driven atmospheric-pressure plasma in a plane-parallel geometry, operated in helium with small admixtures of oxygen and water vapour (He + O2 + H2O). The density of OH is measured under a wide range of conditions by absorption spectroscopy, using an ultra-stable laser-driven broad-band light source. These measurements are compared with 0D plasma chemical kinetics simulations adapted for high levels of O2 (1%). Without O2 admixture, the measured density of OH increases from 1.0 × 1014 to 4.0 × 1014 cm−3 for H2O admixtures from 0.05% to 1%. The density of atomic oxygen is about 1 × 1013 cm−3 and grows with humidity content. With O2 admixture, the OH density stays relatively constant, showing only a small maximum at 0.1% O2. The simulations predict that the atomic oxygen density is strongly increased by O2 addition. It reaches ∼1015 cm−3 without humidity, but is limited to ∼1014 cm−3 beyond 0.05% water content. The addition of O2 has a weak effect on the OH density because, while atomic oxygen becomes a dominant precursor for the formation of OH, it makes a nearly equal contribution to the loss processes of OH. The small increase in the density of OH with the addition of O2 is instead due to reaction pathways involving increased production of HO2 and O3. The simulations show that the densities of OH, O and O3 can be tailored relatively independently over a wide range of conditions. The densities of O and O3 are strongly affected by the presence of small quantities (0.05%) of water vapour, but further water addition has little effect. Therefore, a greater range and control of the reactive species mix from the plasma can be obtained by the use of well-controlled multiple gas admixtures, instead of relying on ambient air mixing.

This work presents experiments and modelling of OH densities in a radio-frequency driven atmospheric-pressure plasma in a plane-parallel geometry, operated in helium with small admixtures of oxygen and water vapour (He + O 2 + H 2 O). The density of OH is measured under a wide range of conditions by absorption spectroscopy, using an ultra-stable laser-driven broad-band light source. These measurements are compared with 0D plasma chemical kinetics simulations adapted for high levels of O 2 (1%). Without O 2 admixture, the measured density of OH increases from 1.0 × 10 14 to 4.0 × 10 14 cm −3 for H 2 O admixtures from 0.05% to 1%. The density of atomic oxygen is about 1 × 10 13 cm −3 and grows with humidity content. With O 2 admixture, the OH density stays relatively constant, showing only a small maximum at 0.1% O 2 . The simulations predict that the atomic oxygen density is strongly increased by O 2 addition. It reaches ∼10 15 cm −3 without humidity, but is limited to ∼10 14 cm −3 beyond 0.05% water content. The addition of O 2 has a weak effect on the OH density because, while atomic oxygen becomes a dominant precursor for the formation of OH, it makes a nearly equal contribution to the loss processes of OH. The small increase in the density of OH with the addition of O 2 is instead due to reaction pathways involving increased production of HO 2 and O 3 . The simulations show that the densities of OH, O and O 3 can be tailored relatively independently over a wide range of conditions. The densities of O and O 3 are strongly affected by the presence * 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.

Introduction
Non-equilibrium atmospheric-pressure plasmas are known to be efficient sources of reactive species [1][2][3][4][5][6][7][8][9][10][11][12][13] and have been investigated for several different applications, in particular surface processing, environmental applications and plasma medicine [7,. One of the challenges is tailoring the plasma gas-phase chemistry to each specific application. Tailoring the chemistry is usually achieved by adjusting external parameters such as: the gas composition [6,7,[14][15][16], the electric field distribution (by modifying the source design), or the applied voltage characteristics (direct or pulsed voltage, rise time, amplitude, frequency) [17,18,[35][36][37]. The latter two approaches can require modification of the plasma hardware, limiting their value for process control and operation for multiple different processes. Modifying the gas composition generally provides the most operational flexibility. Atmospheric pressure plasmas generally operate in rare gases with complex admixtures, arising from deliberate molecular gas additions, ambient air entrainment and/or impurities. The full plasma chemistry needs to be understood in order to effectively manipulate the density of desired reactive species.
For biomedical applications the emphasis has been on understanding and optimising the production of reactive oxygen and nitrogen species due to their high oxidative power and their natural roles in major biological functions (in particular in the immune system, or as signalling molecules in cellular functions [1,3,4,16,23,24]). In the presence of water, which can come from the biological target, itself or may be intentionally introduced to the feed gas, hydrogen peroxide (H 2 O 2 ) attracts particular attention. It is a relatively long-lived and powerful oxidising species that can interact with distant targets, and can be generated in significant concentrations [3,38,39]. One of the main mechanisms for the production of hydrogen peroxide in plasmas is the three-body recombination of two hydroxyl (OH) radicals. Therefore, understanding the production and destruction pathways of hydroxyl radicals is necessary to optimise and control hydrogen peroxide for applications.
Molecular oxygen (O 2 ) is often added to the gas mixture to produce O-based reactive species such as atomic oxygen and ozone, which also play important roles in disinfection and oxidation. Ozone is used in a wide range of applications, including treatment of municipal and waste water, and for food processing [5,8,9,40]. It is a powerful germicide and also efficiently neutralises organic odours. In the context of complex admixtures, the control of both OH/H 2 O 2 and O/O 3 is important. RF plasma sources in He + O 2 + H 2 O mixtures (the focus of this paper) are excellent sources of both H-species and Ospecies, and some tailoring of gas chemistry has been demonstrated. Several studies have investigated how changing the gas composition allows optimisation for a given application [14,41]. In particular, numerical simulations have elucidated the effect of either oxygen or water vapour addition on the generation of the ROS, and the relative contribution of different production and loss reactions.
These studies have helped to begin unravelling the mechanisms of ROS formation and their interaction with biological samples. The production of O-based species is greatly enhanced by the addition of O 2 to the gas mixture, due to additional reaction pathways. Conversely, the addition of water vapour to a discharge already containing O 2 appears to hinder the production of O-based species. On the other hand, the addition of both water and of O 2 increases the density of Hcontaining species [14,[41][42][43]. Therefore, the ROS composition in the gas phase can be controlled by varying the gas mixture. Remarkably, the total ROS concentration was observed to be little changed when the ratio [H 2 O]/[O 2 ] was varied over four orders of magnitude (for similar dissipated energies) [41]. This is despite dramatic changes in the dominant chemical reactions, indicating that both O 2 and H 2 O are efficient additives for ROS production.
In the specific case of hydroxyl, it was shown that its density mostly depends on the concentration of water vapour, and is relatively insensitive to O 2 addition (over the range of 0.1%-1% of O 2 ). However, the dominant reactions for both production and loss change dramatically in the presence of O 2 [14]. In more detail, when only water is present, OH is mostly produced by electron impact dissociation of H 2 O and lost by recombination of two OH radicals in a three-body process. In the presence of oxygen, the dominant OH production and loss reactions are driven by O( 1 D) and O, respectively. Under these conditions, the net contribution of oxygen to the formation of OH (as defined below) is not clear, since both production and loss pathways are enhanced. In the presence of high concentrations of water vapour, the complex role of lower density secondary species needs to be identified in more detail in order to fully understand the plasma chemistry. Such secondary species (for example, HO 2 ) are formed through two-step (or more) processes, and generally have low densities for low water vapour concentration, but start to significantly affect the kinetics at high water concentrations. In the literature, studies usually present the percentage contribution of individual reactions towards the production (or loss) of hydroxyl relative to the total production (or loss) rate. This enables the dominant reactions for production and loss to be determined. However, this representation does not show the net contribution of a given species to the generation of hydroxyl, i.e. the effective contribution of a given species, or group of species, to the total production rate of hydroxyl, defined as the rate of production of hydroxyl from reactions of the given species minus the rate of loss of hydroxyl in collisions with that species. In this paper, an analysis of the reaction rates is used to identify the net contribution of each given species. A detailed analysis of the kinetics of the generation of OH and other ROS is performed for a range of admixture ratios The previous studies cited above present simulations, whereas this paper presents new experimental data, expanding on the limited experimental results that can be found in the literature [9,[44][45][46][47][48]. The OH densities were measured for various concentrations of O 2 and H 2 O, up to 1% in each case. They were obtained using a broad-band UV-absorption spectroscopy technique using an ultra-stable broad-band light source [47,49]. This light source has excellent temporal intensity stability, leading to an absorption baseline variability lower than 2 × 10 −5 over the range of the OH(X)→OH(A) transition (306-311 nm) [49]. This setup allows a detection limit one order of magnitude lower compared to the typical limit (about 10 −3 ) that can be achieved with the more commonly used UV-LEDs [9].

Plasma source
The RF-driven plasma source used here was described in detail in [5] and is illustrated in figure 1. Briefly, a quasihomogenous plasma is generated between two plane-parallel electrodes with a 1 mm gap. The length of the electrodes is 30 mm and the width is 10 mm offering a surface-to-volume ratio of 2 mm −1 . Two MgF 2 windows close the reactor in the direction perpendicular to the plane electrodes. An RF generator (Coaxial-Power, ) and an L-matching network (Coaxial-Power, MMN 150) are used to drive the discharge at a frequency of 13.56 MHz and a dissipated power of 15.0 ± 1.5 W cm −3 . Voltage and current are measured with a PMK-14KVAC Tektronix probe (1000:1) and an Ion Physics Corp. CM-100 l probe (1 V A −1 ) from which the dissipated power was calculated following the method of [50].
The total gas flow is 5 slm. Helium with a purity of 99.996% was used. Most of the gas lines are made of stainless-steel to limit the impurities in the gas. Water vapour and oxygen are admixed to the gas flow using three mass flow controllers. A defined content of water vapour is added to the feed gas by guiding a fraction of the total helium flow through a glass bubbler filled with distilled water. The amount of water vapour is calculated using the vapour pressure at room temperature and the flow rate through the bubbler as in [47]. The output gas of the reactor is guided to an exhaust several meters downstream. It is centred at 1.5 cm of the plasma channel. RF: radio frequency, LDLS: laser driven light source. Reprinted from [5], with the permission of AIP Publishing.

Absorption spectroscopy
The hydroxyl densities are measured by absorption spectroscopy using an ultra-stable laser-driven broad-band light source (LDLS, Energetiq EQ-99) and reflective optics to limit chromatic aberrations (figure 2). The divergent light coming out of the light source is focused into the central plane of the discharge, collected and focussed onto the spectrograph's entrance slit by four parabolic mirrors (UV enhanced aluminium MPD-F01 Thorlabs). The focal beam waist is about 2 mm, so that the measured density of hydroxyl is spatially averaged over the 1 mm gap between the electrodes. The optical absorption path of 11 mm comprises the width of the electrodes and both 0.5 mm gaps between windows and electrodes. This path is perpendicular to the feed gas flow. The transmitted spectrum is recorded by a 0.5 m Czerny-Turner spectrograph (Andor SR500i) equipped with a 2400 l mm −1 grating that is blazed for λ = 300 nm and a CCD camera (Andor, Newton DU940P-BU2, 2048 × 512 pixels of 13.5 × 13.5 µm 2 size). The spectrometer entrance slit width was set to 100 µm, resulting in an instrumental broadening of 0.03 nm (see figure 3). Additional apertures in the beam path were used to reduce unnecessary optical plasma emission reaching the entrance slit. The OH ground state density is determined from the intensity of the integrated OH (X 2 Π i , ν ′′ = 0) → OH (A 2 Σ + , ν ′ = 0) rotational envelope in the wavelength range 306-310 nm.
As previously described in several papers [9,47,49], the absolute density of the hydroxyl radical, n OH , is obtained from analysis of the measured plasma absorbance spectrum A(λ) in the spectral range from 306 to 311 nm, assuming the Beer-Lambert law (equations (1) and (2)). Four signals are measured successively to correct for background and plasma emission. The first signal is recorded with both the plasma and the light source on (S P,L ), the second records the intensity of the light source only (S L ), the third records the plasma emission only (S P ) and the last signal is a background with both plasma and light source off (S B ). Each signal is integrated over a time period of 2.5 s after an equally long plasma stabilisation time.
The observed absorbance is equal to the product of the absorption cross section σ(λ), the absorption path length L, and the line-of-sight averaged absorber density n OH . A homemade spectrum simulation and fitting programme (see below) is used to find the best least-squared fit of the measured absorbance spectrum, yielding the density and rotational temperature of the lower OH (X 2 Π i , ν ′′ = 0) state. The majority of the population of the electronic ground state lies in the lowest vibrational level, ν ′′ = 0; no absorption band originating from the ν ′′ = 1 level was observed. The accuracy of the density measurements depends on the knowledge of the absorption length, here 11 mm with 5% uncertainty. The OH(X) rotational temperature is not necessarily in equilibrium with the gas translational temperature [46].
The absorption spectrum is simulated in the following way. The ro-vibronic energy levels of the upper OH(A 2 Σ + , ν ′ = 0) state are calculated according to [51], with Hund's coupling case (b) since Λ = 0, and the lower OH(X 2 Π i , ν ′′ = 0) state, with intermediate case (a) to case (b) coupling. The rotational line strength factors for the 12 electric dipole-allowed branches are calculated using the expressions in [52]. These factors are converted, firstly to relative Einstein coefficients for spontaneous emission, then to absolute ones using the value of 0.698 µs for the radiative lifetime of the lowest rotation level of the OH(A 2 Σ + , ν ′ = 0) state [53], and finally to spectrally integrated absorption cross sections for each individual rotational transition. The relative thermal population distribution among the rotation levels of the OH(X 2 Π i , ν ′′ = 0) state is calculated according to the rotational Boltzmann factor normalised to the partition sum. For all rotational transitions a common spectral Gaussian line profile is assumed, which represents the resolution of the spectrograph. The intrinsic Doppler (0.93 pm at 304 K) and pressure (0.66 pm in helium at 1 atm [54]) broadening are negligible under the given experimental conditions. Figure 3 shows an example of a measured absorbance spectrum together with the best fit simulation. All measurements were taken at the centre of the reactor, as illustrated by the position of the LDLS beam in figure 1.

Numerical model for 0D plasma chemistry
In this work, the GlobalKin code is used to describe the RF quasi-homogenous discharge. It is a 0D global plasmachemical kinetics model which is described in detail in [55]. Briefly, it consists of a reaction chemistry and transport module, a Boltzmann equation solver and an ordinary differential equation solver. In the ordinary differential equation solver, the mass continuity equations for each charged and neutral species are solved as a function of time, accounting for surface and gas phase production and consumption processes: Here, N represents the number density of a species i in the gas phase, S V is the surface area to volume ratio of the source, Λ D is the diffusion length, D the diffusion coefficient, γ the surface loss probability, f the return fraction of species from surfaces and S i represents gas phase production and consumption processes. Here, the diffusion length Λ D , is calculated as in [43] for a plasma confined in a rectangular chamber, giving a value of 0.0317 cm. Surface losses and gains are calculated using user defined surface loss probabilities as defined in [47]. In addition, the electron energy equation is also solved: Here, n e is the electron density, k B the Boltzmann constant, T e the electron temperature and p d represents the plasma power. The second and third terms on the right hand side of the equation represent electron energy changes through elastic and inelastic collisions, respectively. In these terms, ν mi is the electron momentum transfer collision frequency with a species i, m e the electron mass, M i the mass of the collision partner, T i the temperature of the collision partner, k the electron impact rate coefficient and ∆ε i the electron energy changes due to the inelastic collision. A pseudo 1D plug flow model is used to convert the temporally varying densities and temperatures obtained from equations (3) and (4) into spatially varying quantities along the channel of the plasma source using the gas flow velocity. The electron energy distribution function, mean electron energy, electron transport coefficients, and electron impact rate coefficients, are obtained regularly during the simulation, by solving the two-term Boltzmann equation, in this case, every 0.1002 cm. At each point the Boltzmann equation is solved for a range of reduced electric fields and the results tabulated. Values of the electron impact rate coefficients and transport coefficients corresponding to the instantaneous mean electron energy calculated by equation (4) are then used to calculate the relevant source terms in the ordinary differential equations. For all cases, the gas temperature is set to 305 K. Simulations are carried out for plasma electrodes of 3 × 1 × 0.1 cm 3 but a plasma volume of 3 × 1.1 × 0.1 cm 3 to account for both 0.5 mm gaps between windows and electrodes where the plasma expands. The power density is 15 W cm −3 and the flow rate is 5 slm.
The reaction mechanism includes 46 species and 577 reactions and is described in the appendix. It is largely based on the reaction mechanism for He + H 2 O mixtures presented in [47]. The reaction set was extended to account for higher O 2 concentrations by adding O-based ions and by refinements to some reaction rate coefficients based on the reaction mechanisms for He/O 2 mixtures presented in [9,11,12]. The species included in the model are listed in table 1.

Results and discussion
In a helium-oxygen-water vapour gas mixture, the density of hydroxyl reaches an approximately constant value within a very short distance in the channel, about 2 mm in our conditions [47]. The results presented hereafter are obtained in the quasi steady-state region of the gas phase, in the middle of the electrodes, 15 mm from the entrance.

Effects of water vapour and oxygen addition on OH density
To study the effect of water vapour and oxygen on the generation of hydroxyl, various concentrations of water vapour and oxygen are added to the feed gas. The concentration of each added species ranges from 0% to 1%. At higher contents the discharge cannot be sustained. The density of OH measured by absorption spectroscopy is presented in figure 4 (as squares). The OH density first increases rapidly with humidity content, up to a few tenths of a percent of water vapour, after which the increase is slower. This transition has already been studied [47,56,57] and was attributed to the change of electron density and temperature with increasing humidity content, affecting the reaction rate coefficient for OH production. In the absence of O 2 , OH is mainly produced by electron impact dissociation of water molecules. However, the water vapour dissociation frequency passes through a maximum, due to competition between an electron temperature increase (increasing the rate coefficient for electron impact dissociation) and electron density decrease (decreasing rate of electron impact dissociation) as the humidity content increases. Figure 4 also shows the numerical results for similar conditions (as solid lines). The experimental and simulation results agree within 50%. The trends are very similar, although the simulation results are somewhat higher by a nearly constant factor. This could be due to an overestimation of the power density (1-2 W cm −3 , compatible with the uncertainty of the power density measurement indicated in section 2.1) and/or uncertainties in reaction sets and rate coefficients. In addition, in helium-water vapour only, the OH density increase is stronger at high humidity content compared to the experimental results. Despite these discrepancies, the comparison with experiment demonstrates very similar trends and comparable absolute densities, giving confidence in the accuracy of the numerical model.
When molecular oxygen is introduced into the feed gas, it becomes the new dominant source of O-based species, in particular atomic oxygen and ozone. The simulated densities of atomic oxygen are presented in figure 5(a). These are in good agreement with the experimental results presented in [8,45,47]. In [8] and [45], nitrogen was also included in the gas mixture at a 4:1 ratio. This comparison may suggest that 0.1% nitrogen has a moderate impact on the atomic oxygen density in dry conditions. However, this comparison should not be considered as a validation of the simulation but simply as a coherent comparison in relatively analogous gases. The density of atomic oxygen increases by one to two orders of magnitude in the presence of 0.1% molecular oxygen compared to He + H 2 O only. This is due to the change of the main oxygen-atom production reaction already pointed out in [14]: On the other hand, as soon as few hundred ppm of water is introduced in a He + O 2 gas admixture, the oxygen-atom loss is mainly due to OH + O → O 2 + H, for any concentration of oxygen.
The density of atomic oxygen is predominantly affected by the concentration of molecular oxygen. Nevertheless, it should be noted that a steep decrease in the O density is observed with increasing water vapour concentration up to about 0.05%, for cases where the O 2 admixture is over 0.025%. Such densities of water vapour could be in the range of impurities or diffusion of ambient humidity in the case of an open plasma source. Over 0.05% water vapour concentration, the O density stabilises progressively and is only slightly affected by further water vapour addition. This observation indicates that the introduction of known amounts of additive species to the feed gas gives better control of ROS production than relying on mixing and diffusion of the effluent into the surrounding environment in the case of open plasma sources [58].
The density of ozone is strongly related to the addition of oxygen to the feed gas ( figure 5(b)). It increases by three orders of magnitude with addition of 0.1% O 2 . At very low water vapour concentrations, ozone also shows a sharp decrease with water vapour addition, similar to oxygen atoms but not quite as pronounced. Over 0.05% H 2 O, the ozone density is only weakly affected by increasing humidity content.
Overall, the simulation results suggest that OH, O and O 3 can be controlled quite independently by varying both water and oxygen concentrations in the feed gas over the range 0%-1% for H 2 O and 0.05%-1% for O 2 . The OH density depends almost entirely on the water vapour content, while the O and O 3 densities are significantly affected by presence of water vapour, but this effect is relatively independent of the concentration of water vapour for values above 0.05%. Conversely, the O and O 3 densities are more strongly affected by varying the O 2 content. More precisely, the O and O 3 densities drop by approximately an order of magnitude following the addition of up to 0.05% water vapour to a He + O 2 feed gas. If high densities of O and O 3 are required without OH species, then dry conditions appear to be better.

Analysis of pathways and reaction rates at low water vapour concentration
A detailed analysis of the hydroxyl reaction pathways in the presence of 0.025% water vapour and various concentrations of oxygen is drawn from the chemical kinetics simulation. The dominant reactions for production and consumption of OH, with and without 0.1% O 2 , are presented in tables 2 and 3. The analysis is performed at 15 mm into the plasma channel, corresponding to the measurement position. Table 6 presents the corresponding net production rates of the major species as defined in the introduction.
In the absence of molecular oxygen, electron impact reactions generate almost 62% of the hydroxyl radicals. 54% of the OH molecules are produced by dissociation of water molecules by electron impact, and 74% are lost by the recombination of two OH molecules to produce hydrogen peroxide (R(10)) or water plus atomic oxygen (R(11)). The other significant reaction for production of hydroxyl involves the H 2 O + ion, responsible for about 30% of OH production. H 2 O + is almost entirely produced by two-and threebody Penning ionisation reactions of He(2 3 S) with water ((He) + He(2 3 S) + H 2 O → (He) + He + H 2 O + + e). At low humidity content, the density of secondary species, such Simulations are only carried out for cases where the level of admixture is low enough to enable plasma ignition. Experimental measurements from [47] are obtained in very similar conditions except at a distance of 1.2 cm from the gas inlet instead of 1.5 cm. Experimental measurements from [8] are obtained in helium flux 10 slm, N 2 /O 2 (4:1) admixture 0.1%, and similar driving voltage characteristics (voltage of 234 V). (b) Simulated ozone densities under similar conditions as in (a). Table 2. Dominant reactions for production of OH in He + 0.025% H 2 O (+0.1% O 2 ) and associated percentage of production and reaction rates (the symbol 'v' corresponds to the reaction rate in cm −3 s −1 ).      as HO 2 , is low and the contribution of these species remains small. In particular, HO 2 contributes to only 3% of hydroxyl production and loss. These results are similar to those detailed in [47]. As soon as some oxygen is added to the mixture, the pathways for OH generation change dramatically, even though the resulting OH density remains little changed (see figure 6). The addition of 0.1% O 2 increases the density of atomic oxygen by more than one order of magnitude ( figure 5(a). Consequently, the reaction rates of O-induced reactions are strongly enhanced: the reaction rate of reaction R (6): increases by more than two orders of magnitude, becoming the dominant production reaction of hydroxyl. Loss mechanisms are also modified. Reaction R(13), OH + O → O 2 + H, involving O is greatly enhanced and becomes responsible for 72% of hydroxyl losses. Thus, atomic oxygen is responsible for most of the production as well as the loss of hydroxyl molecules. The calculation of the net production rate of OH by processes involving atomic oxygen, i.e. production reactions R(6) and R(8) minus consumption reaction R(13) (see table 6), shows the net effect O has on the OH density. It reveals that it actually only makes a small alteration to the density of OH at 0.025% H 2 O and 0.1% O 2 . More precisely, R (6,8,13) have an effective rate of −2.08×10 17 cm −3 s −1 which is −8% of the total production rate of 2.53 × 10 18 cm −3 s −1 . Though dramatically impacting the pathways for OH formation, the O and O( 1 D) production and consumption contributions nearly cancel out, leading to a small effective (and negative) contribution to the OH density. This observation remains valid for higher added O 2 concentrations, explaining the nearly constant OH density with O 2 addition (presented in figure 4) for a given low humidity content.
Further analysis of the reaction pathways for OH formation reveals that molecular oxygen has an indirect impact on the OH density. Apart from the O/O( 1 D) reactions, the other main production reactions are through the seemingly minor reactions R(5) and R(9): The density of atomic hydrogen is very similar with and without oxygen admixture, while the densities of HO 2 and O 3 Table 6. Net production rates for OH by major species or groups of species under conditions of low (0.025%) and high (0.75%) water content. The percentage represents the ratio of the net reaction rate of the species to the total production rate of OH.  table 6). The role of reaction (5) on the OH kinetics was also identified in a recent study of a He + O 2 + H 2 O pulsed plasma jet at 1 kHz and 10 kV [59]. For relatively low concentrations of oxygen, the role played by ozone is minor. However, it grows with the increasing oxygen addition since ozone is dominantly produced by: He + O 2 + O → He + O 3 . Subsequently, through reaction (6), it produces non-negligible amounts of OH radicals, representing 2% of the OH formation at 0.1% O 2 (effective rate 5.6 × 10 16 cm −3 s −1 ) and about 10% at 0.5% O 2 (reaction rates at 0.5% O 2 are not detailed here).

Analysis of pathways and reaction rates at high water vapour concentration (0.75% H 2 O)
Now let us consider a similar analysis of the hydroxyl reaction pathways in the presence of 0.75% water vapour and various concentrations of oxygen. The dominant reactions for the production and consumption of OH with and without 0.1% O 2 are presented in tables 4 and 5. The corresponding net production rates of the major species can be found in table 6. In the absence of molecular oxygen, at the highest water concentration studied (0.75% H 2 O), the reaction rates of all OH related reactions increase significantly and the relative importance of the different production or loss reactions of OH change compared to low humidity conditions (0.025% H 2 O). The density of most H-containing species increases dramatically, in particular for the secondary species H 2 O 2 and HO 2 . The H density increases to a lesser extent, while OH and O show a similar increase. Conversely, the positive ions H 2 O + and OH + decrease by up to an order of magnitude, due to a decrease in electron density, affecting the density of their precursor species, He(2 3 S) which are created through electron impact reactions.
Three main points can be made. The dissociation of H 2 O by electron impact remains the dominant reaction for OH production. But as mentioned above, due to the decrease in electron density as the water vapour concentration increases, OH production is reduced above a certain water vapour content. Also, with the larger increase of secondary species densities compared to other species, reactions R (5,8,14,15) are promoted. In particular, reaction (5) now accounts for more than a fifth of the production of OH, against 2.7% with only 0.025% H 2 O. Among the loss mechanisms of OH, reactions involving secondary species (H 2 O 2 , HO 2 ) also play a more significant role. The recombination of two OH molecules now only accounts for half of the losses, against three-quarters at 0.025% H 2 O. The effective reaction rate of HO 2 has an increased role in OH production; it goes from having a negligible contribution at low water vapour content to making +14% of OH formation (see table 6). It is noteworthy that reaction R(2) has a decreased relative contribution due to the significant decrease in H 2  In the presence of 0.1% molecular oxygen and 0.75% water vapour, the plasma chemistry does not tend towards a pure water vapour chemistry, in contrast to the low humidity case (0.025%). The effect of high amounts of both water vapour and oxygen leads to the development of complex pathways for OH formation. Thus, the chemistry exhibits strong characteristics of both water vapour and oxygen-induced pathways. In this way, OH is produced approximately equally by electronimpact dissociation of water and by the O( 1 D)-induced reaction R(6). The following comments compare a gas mixture with 0.1% O 2 at low humidity (He + 0.025% H 2 O + 0.1% O 2 ) to a high humidity case (He + 0.75% H 2 O + 0.1% O 2 ). The latter reaction R(6), largely benefits from the increase of water density: the reaction rate of R(6) increases by 193% from low to high humidity, while the reaction rate of R(13) remains almost unchanged. This means that large concentrations of water vapour are necessary for the net contribution of atomic oxygen R (6,8,13) to increase the production of OH. Nevertheless, O and O( 1 D) contributions remain low at high humidity, about +5% of OH production rate (3.83 × 10 17 out of 7.37 × 10 18 cm −3 s −1 ). In addition, the secondary species HO 2 which accounted for a quarter of OH formation at low humidity content has a negligible impact on OH formation (−2%) at high humidity. This is due to the strong drop of O density with higher water concentration, affecting the R(8) reaction rate.
These complex interactions show that even with only two additives, the chemistry cannot be easily tailored, and a detailed analysis of reaction rates must be undertaken for that purpose.
Focusing on O and O( 1 D), their generation mechanisms are very different, and change according to the gas mixture.
If water vapour is the only admixture, both the production and loss mechanisms of ground-state atomic oxygen depend dominantly on OH and do not change significantly with increasing water content. Consequently, the density of O increases almost proportionally with OH (+490% from 0.025% to 0.75% added water vapour) and the density of O may be inferred from the density of OH if it is known in one condition. The generation mechanisms of O( 1 D) are different, but again they do not change significantly with increasing water content. O( 1 D) production depends on low density secondary species (mostly O 2 , O 2 * and O) and a little on water vapour. Therefore, the O( 1 D) density does not increase significantly from low to high humidity content.
In Diffusion to the walls is typically the second most important loss process but it only represents a few percent of the overall losses. However, when some water vapour is introduced, OH and HO 2 become largely responsible for the loss of O. O( 1 D) is very efficiently quenched by water to form 2 OH, so that the decay to O becomes less significant. O losses are significantly higher in the presence of water vapour causing the density of O to drop by an order of magnitude from dry to low humidity conditions. As the humidity content increases from 0.05% to 1%, the variations are much smaller and the O density decreases by at most a factor 3.
Finally, figure 6 illustrates the variations of the densities of the major ROS and the electron density for the most representative admixtures. It also includes the densities simulated in He + 0.025% O 2 + 0.5% O 2 in order to accentuate the role of O 2 -induced reactions. It highlights that adding small (0.1%) or large (0.5%) concentrations of O 2 does not significantly affect the density of H-species (apart from HO 2 ). However, large concentrations of O 2 (0.5% O 2 ) significantly increase the production of O 3 . From 0.1% to 0.5% added oxygen, the ozone density rises by a factor 4 (1-4 × 10 13 cm −3 ).

Conclusion
This work presents measurements of the density of OH radicals in a He + O 2 + H 2 O RF plasma at atmospheric pressure. The measurements were made using an ultra-sensitive broadband absorption spectroscopy setup. The results are mostly in agreement with previous numerical analyses [14,41]. The plasma chemistry is also analysed using 0D plasma-chemical kinetics simulations. The simulated and experimental results are compared in order to understand the kinetics of the major ROS, in particular OH, O, O( 1 D) and HO 2 , over a wide range of water and oxygen concentrations. The analysis of reaction rates enabled the identification of the net contributions of each species to the generation or loss of OH and O.
In He + H 2 O, OH is created principally by electron impact dissociation of water vapour, and lost by three-body recombination of two OH molecules to form H 2 O 2 . The density of OH increases from 1.0 × 10 14 to 4.0 × 10 14 cm −3 for H 2 O admixtures from 0.05% to 1%. The density of atomic oxygen is of the order of 1 × 10 13 cm −3 . When water vapour is the only molecular admixture in the gas, the density of O increases nearly proportionally with OH over the studied range and may be inferred from the density of OH, and inversely. The experimental measurements and simulation results of OH and O densities are in good agreement. The discrepancies in OH densities between the two approaches are within 50%.
In the presence of molecular oxygen, the OH formation pathways are strongly modified. Nevertheless, the OH density is not significantly affected because the dominant O and O( 1 D)-induced production and consumption contributions to OH formation nearly cancel out. However, the introduction of around 0.1% molecular oxygen increases the density of O-species (O, O( 1 D), O 3 ) by several orders of magnitude. This also favours HO 2 production, through H + O 2 + He → HO 2 + He, which plays a significant role in OH formation at low water content. The density of the other H-species such as OH are not strongly affected by the addition of molecular oxygen as long as the total amount of impurities remains below 1%. The introduction of larger concentrations of oxygen only has a strong impact on the ozone density.
Therefore, it can be concluded that OH, O and O 3 can be tuned relatively independently by varying the concentration of water and oxygen in the feed gas. The OH density is not strongly affected by the oxygen content, but is instead determined by the water vapour content. The O and O 3 densities, on the other hand, are strongly affected by O 2 addition whether or not water vapour is added, but the effect is relatively independent of the concentration of water vapour as long as it is above 0.05%, with their densities then mainly controlled by the O 2 content. Below 0.05% water vapour, the O and O 3 densities are strongly decreased by water addition, and independent control of OH, O and O 3 is not possible. A perfectly dry environment is preferable if higher amounts of O and O 3 are required without OH species.

Data availability statement
The data that support the findings of this study are openly available at the following URL/DOI: https://doi.org/10.15124/ 6ebd6e18-17e5-4cf1-80cc-04379f58a1e7.    [98,141] a In s −1 , m 3 s −1 and m 6 s −1 for one-, two-and three-body reactions, respectively. b Value is estimated in reference. c Effective rate coefficients calculated from pressure dependent rates as described by Sieck et al [119] for 1 atm and a temperature range 280-350 K. d Background gas is (humid) air in given reference. Gas efficiency factors for He background gas are not known for these reactions, but could potentially change calculated reaction rate coefficients if taken into account. e Rate coefficients are estimated by extrapolating the coefficients k 0 300 and A given by Sieck et al [119] using an exponential fit, and using constant values n = 16, B = 5000, and kL = 10 −24 . f Value is listed as a lower limit in reference. g Estimated branching ratio.  [11,141,177] (Continued.)