Multi-hollow Surface Dielectric Barrier Discharge: Production of Gaseous Species Under Various Air Flow Rates and Relative Humidities

An evaluation of the gaseous species production by the discharge, i.e., discharge chemical activity, is very important for determining its potential for practical applications. In this work, production of gaseous species by the multi-hollow surface dielectric barrier discharge generated in a perforated ceramic substrate with the air-exposed electrode is investigated under conditions of various discharge powers (1–5 W), air flow rates (0.25–2.4 L/min) and air relative humidities (0–80%). Production of ozone O3, nitrous oxide N2O, nitric oxide NO, nitrogen dioxide NO2, dinitrogen pentoxide N2O5 and nitric acid HNO3 is evaluated in terms of concentration (ppm), production yield (g/kWh) and production rate (mg/h). The work demonstrates a critical impact of both air flow rate and relative humidity on prevailing discharge mode (“O3 mode” vs. “NOx mode”) and, thus, on overall composition and concentration of produced gaseous species. For low discharge power, the discharge operates in the “O3 mode”, when O3, N2O, N2O5 and HNO3 are dominant gaseous products. With the increasing power, the discharge transfers into the “NOx mode”, when N2O and HNO3 along with NO and NO2 are mostly produced. In dry air, transition from “O3 mode” to “NOx mode” is found for the specific input energy of 1000–1100 J/L. With an increase of air relative humidity from 20 to 80%, the transition gradually decreases from approximately 600 to 450 J/L, respectively.


Introduction
The nonthermal plasma (NTP), as it is very well known, can create a high-reactive environment by initiating and maintaining various chemical reactions at ambient conditions of temperature and pressure.This feature represents one of the most important benefits of the NTP in contrast to the thermal plasma.The most frequently employed NTP sources for effective plasma chemistry are electrical discharges which have been utilised in numerous applications including surface treatment of materials [1,2], air flow modification and control [3,4], air and water pollution control [5,6], bio-decontamination [7,8], medical [9,10] and agricultural applications [11].Among the discharges, the dielectric barrier discharges (DBDs) have a special position due to their simplicity, scalability, and the availability of reliable, efficient, and affordable power supplies [12].Out of them the surface dielectric barrier discharges (SDBDs) are getting an increasing attention because of their versatility, small dimensions, and possibility for electrodes encapsulation into the dielectric, what prevents their aging and damage.To optimise utilisation of the discharges in selected applications, their proper chemical characterisation (i.e., their chemical effects) must be known in detail especially under ambient conditions.
Chemical effects of NTP are in general mainly associated with high-energy electrons with energy of the order 10 eV or even higher.The electrons collide with bulk gas molecules (such as N 2 , O 2 , H 2 O) leading to their dissociation and production of short-lived high-reactive species, also known as radicals (e.g., • O, • N, • OH, etc.).Apart from molecular dissociation, ion formation may also occur (e.g., O + , O 2 + , N + , N 2 + , etc.).In all reactions, an excess of energy is transferred not only to kinetic energy of reaction products, but also to excitation of atoms and molecules (e.g., N 2 * , O * , etc.).In addition, the electronic or vibrational excitation of species can also take place in direct reactions with electrons if they do not possess enough energy for dissociation or ionisation [13].Reactive species can further undergo mutual post-discharge reactions [14] or react with bulk gas molecules and form more complex long-lived secondary species.
Evaluation of production of gaseous long-lived species by different discharges under various conditions, i.e., the discharge gas-phase chemical activity, have been already studied by many authors experimentally [15,16] and by numerical modelling [17,18].In the air-like mixtures, ozone O 3 and nitrogen oxides NO x (mainly nitric oxide NO, nitrogen dioxide NO 2 and nitrous oxide N 2 O) represent the most frequent long-lived products of atmospheric pressure air discharges [15].The production of the gaseous species generally depends on discharge parameters (e.g., applied voltage), and working conditions (e.g., composition of a gas mixture) and their formation can be used as a measure of the discharge chemical activity.
In this manuscript, a relatively novel type of discharge, the multi-hollow SDBD (also known as micro-hollow SDBD [19][20][21][22][23]), was investigated.The discharge was generated in a perforated ceramic substrate with the air-exposed electrode.A unique geometry forces the carrier gas to pass through holes (hollows) of the ceramic substrate inside which the discharge is formed.Although the multi-hollow SDBD has been already studied in relevant applications [20,[24][25][26][27][28][29] and its physical characteristics investigated by several authors [20][21][22]30], detailed evaluation of gaseous species produced by the discharge have not been examined yet except for ozone O 3 production [31][32][33].Moreover, the NO x and HNO 3 production promoted by nonthermal plasma have been identified as promising and emerging technology for N 2 fixation under ambient air conditions [34].In this sense, evaluation of NO x and HNO 3 produced by the discharge under various working conditions is also important.Therefore, the main objective of this work was to evaluate the gas-phase chemical activity of the multi-hollow SDBD.This research follows our previous work [35] in which the electrical and optical characteristics of the discharge were investigated.Out of the gaseous species produced by the discharge, the production of ozone O 3 , nitrogen oxides NO x (N 2 O, NO, NO 2 , N 2 O 5 ) and nitric acid HNO 3 were evaluated under conditions of various discharge powers (1-5 W), air flow rates (0.25-2.4 L/min) and air relative humidities (0-80%).The effect of air flow rate was examined under dry air conditions, while the effect of air relative humidity was studied for two air flow rates (0.5 and 1 L/min).In addition, the gas as well as surface temperature of the ceramic substrate were also monitored.

Experimental Setup and Methods
The experimental setup is depicted in Fig. 1.The multi-hollow SDBD was generated in a perforated ceramic square substrate (KD-EB2B10, Kyocera) with the dimensions of 50 × 50 × 1 mm (Fig. 2a).It consisted of 170 holes (hollows) with an inner diameter of 1.5 mm and two electrodes-one was embedded inside the ceramic while the other one (made of Ni/Au alloy) was printed on the ceramic surface (air-exposed electrode).More details about the substrate and its parameters can be found in our previous work [35].The substrate was powered by AC high voltage (HV) power supply consisting of function generator (GwInstek SFG-1013), signal amplifier (Omnitronic PAP-350) and HV transformer.
The HV was connected to the air-exposed electrode while the embedded electrode was grounded.Upon application of the AC HV, the discharge was generated at the edge (circumference) of each single hole (Fig. 2b).The waveform of the applied AC HV was measured by HV probe (Tektronix P6015A) and the discharge current by a current probe (Pearson Electronics 2877) both connected to a digital oscilloscope (Tektronix TBS2104).The power consumption of the discharge was evaluated using the Lissajous figure method with an 82 nF capacitor and a voltage probe (Tektronix P2220).Synthetic air (purity 5.0) supplied from a pressure tank was used as the carrier gas and its flow rate was controlled by mass flow controllers (MFC) (Bronkhorst El-Flow Prestige FG-201CV).The air was alternatively enriched by water vapours by passing it through a water cell.The air relative humidity (RH) was monitored by a capacitive humidity sensor (Arduino).Then the air was led into the reactor chamber from the bottom side of ceramic substrate, it passed through the substrate holes, and it exited the substrate by its top side with the air-exposed electrode and left the reactor chamber.At the outlet of the reactor chamber, a thermocouple was placed to measure the gas temperature after passing the discharge zone.Surface temperature of the ceramic substrate was measured by infrared thermal camera (Workswell WIC2) (Fig. 2c).Gas-phase chemical activity of the discharge was evaluated by means of FTIR spectroscopy (Shimadzu IR-Affinity 1S) with a resolution of 0.5 cm −1 inside a gas cell with an optical path length of 10 cm equipped with CaF 2 windows.Main gaseous species (O 3 , N 2 O, NO, NO 2 , N 2 O 5 , HNO 3 ) were identified in the FTIR spectra, and their concentrations were evaluated using absorption bands as follows: O 3 at 1055 cm −1 ; N 2 O at 2236 cm −1 ; NO at 1900 cm −1 ; NO 2 at 1602 cm −1 ; N 2 O 5 at 1245 cm −1 ; HNO 3 at 1325 cm −1 .The absolute concentrations were determined based on calibrations with commercial gas mixtures (NO, NO 2 ), UV absorption (O 3 ) or by modelling the spectra (N 2 O, N 2 O 5 , HNO 3 ) using a set of absorption lines from HITRAN database (more details can be found in [36]).The approximate detection limits of evaluated species were as follows: 70 ppm for O 3 ; 5 ppm for N 2 O; 70 ppm for NO; 15 ppm for NO 2 ; 5 ppm for N 2 O 5 and 15 ppm for HNO 3 .In addition to FTIR measurements, O 3 concentration was also evaluated by a homemade analyser based on UV absorption at 254 nm using a mercury lamp and the fiber optic emission spectrometer (Ocean Optics SD2000) in a 12.5 cm gas cell.The presence of NO was also monitored by an electrochemical sensor (Membrapor NO/SF-1000).
In addition to discharge power, the performance of the multi-hollow SDBD was also evaluated with the help of following variables: • Specific input energy (SIE) (i.e., energy density or discharge energy per gas volume): where P and Q represent discharge power and total air flow rate, respectively.• Production yield (PY) (i.e., the amount of produced gaseous species per SIE): where c represents the concentration of gaseous species produced by the discharge.PYs of all evaluated species were calculated for room temperature of 22 °C.
• Production rate (PR) (i.e., the amount of produced gaseous species per unit of time): (1) The experiments were performed with applied voltage amplitude in a range of 3.5-6.5 kV at a frequency of 1 kHz resulting in discharge power of 1-5 W. Gas-phase chemical activity of the multi-hollow SDBD was evaluated under conditions of air flow rates in a range of 0.25-2.4L/min, air RH in a range of 0-80% and the SIE of 25-1200 J/L.Upon application of applied voltage, the temperature of the ceramic substrate gradually increased.Since a production of gaseous species is generally temperature-dependent, their concentrations were measured once the gas temperature and O 3 concentration at the outlet of the reactor chamber were fully stabilised (usually 15-30 min).

Results and Discussion
Figure 3 depicts typical infrared absorption spectra of gaseous species produced by the discharge operated in two distinct modes: "O 3 mode" and "NO x mode" for the same SIE of 600 J/L (5 W, 0.5 L/min).As we will present in "The Effect of the Air Relative Humidity" section, the presence of humidity in the air can significantly alter the prevailing discharge mode.In the spectra the following absorption bands of the species are denoted: O 3 (1055 and 2125 cm −1 ), nitrous oxide N 2 O (1270 and 2236 cm −1 ), nitrogen dioxide NO 2 (1627 cm −1 ), nitric acid HNO 3 (1325 and 1711 cm −1 ) and dinitrogen pentoxide N 2 O 5 (1245 and 1720 cm −1 ) (Fig. 3).

The Effect of the Air Flow Rate
The gas residence time in the discharge zone can be controlled by varying of gas flow rate.Variation of gas flow rate has a strong impact on gas heating, residual humidity, and gas pre-ionisation and consequently strongly influences the production of gaseous species [20, (3) PR(mg∕h) = Qc.Fig. 3 Infrared absorption spectra of gaseous products of the discharge operated in "O 3 mode" (dry air) and "NO x mode" (air RH = 40%; SIE = 600 J/L; P = 5 W; Q = 0.5 L/min) 37].It also affects mixing of produced species and heat exchange in the reaction zone [38].Therefore, the effect of air flow rate on gas-phase chemical activity of the multi-hollow SDBD was investigated in a range of 0.25-2.4L/min by using dry air as a carrier gas.The following "Ozone O 3 and Nitrous Oxide N 2 O" section is devoted to the results obtained for O 3 and N 2 O.In the "Dinitrogen Pentoxide N 2 O 5 and Nitric acid HNO 3 " section, we present results for N 2 O 5 and HNO 3 since they were observed under dry air conditions for all tested air flow rates.Finally, the results for NO and NO 2 can be found in both sections within the discussion on other species, as they were not detected under all conditions tested, but only under the specific ones.

Ozone O 3 and Nitrous Oxide N 2 O
Ozone O 3 represents one of the most common gaseous products formed by atmosphericpressure air DBDs.Production of O 3 is strongly dependent on atomic oxygen species that may be produced by following reactions [18,39]: Then, O 3 is produced by a well-known three-body reaction: where M can be nitrogen N 2 or oxygen O 2 molecule or oxygen atom • O [40].
Figure 4a shows the O 3 concentration as a function of discharge power for various dry air flow rates.Since the SIE (energy density) is a very important parameter determining the chemical activity of the discharge and is not constant for a given discharge power and various air flow rates, the concentration of O 3 is also depicted as a function of the SIE (Fig. 4b).The results showed a common feature: an increase of O 3 concentration with a decrease of the air flow rate for a given discharge power (except for air flow rate of 0.25 L/ min).The same result for O 3 production by multi-hollow SDBD was also reported by Nayak et al. [32].
Dependence of O 3 concentration on discharge power (or SIE) for a given air flow rate was found nontrivial (Fig. 4).For the air flow rates of 1 and 2.4 L/min the O 3 concentration firstly increased and then culminated, whereas for air flow rates of 0.25 and 0.5 L/min, the increase and the culmination were followed by a subsequent decrease of O 3 concentration (Fig. 4).The same trend of O 3 concentration with an increase of DBD power (i.e., the initial increase, culmination, and the following decrease) was also reported by many authors [41][42][43].Šimek et al. found a maximum O 3 concentration of 750 ppm for the SIE ~ 110 J/L [43], while we observed the maximum concentration ~ 1840 ppm for much higher SIE of 480 J/L (i.e., 2 W and 0.25 L/ min).As is well known, the DBD-based reactors can be operated in two distinct modes.The "O 3 mode" is typical for low SIEs with O 3 dominating among the products.On the other hand, the "NO x mode" prevails for high SIEs with nitric oxide NO and nitrogen dioxide NO 2 among the products [44,45].
Production of NO is given by the following set of reactions known as the Zeldovich mechanism [46,47]: NO 2 is then formed by subsequent oxidation of NO [48,49]: Substantial drop of O 3 concentration to zero for very high SIEs when using dry air flow rate of 0.25 L/min (Fig. 4b) is related to the transition of the discharge from "O 3 mode" to "NO x mode".The transition was observed between the SIE of 1000 and 1100 J/L and was characterised by a complete depletion of O 3 associated with an increase of NO and NO 2 [45,48].This effect is also known as "discharge poisoning" [50,51] and will be discussed in detail later.
In addition to O 3 , NO and NO 2 , nitrous oxide N 2 O represents another frequent gaseous product of atmospheric pressure air discharges.Its production and decomposition are predominantly affected by the first excited state of molecular nitrogen N 2 (A) via following reactions [52,53]: Other reaction pathways leading to N 2 O formation are reactions of excited nitrogen atom N( 2 D) with NO and nitrogen radical • N with NO 2 [54,55]: On the contrary, decomposition of N 2 O may happen via reaction with excited oxygen atom O( 1 D) [53,54]: Unlike O 3 concentration, the N 2 O concentration as a function of discharge power (or SIE) and air flow rate was found monotonic when operating the discharge in "O 3 mode" (i.e., SIE < 1000 J/L) (Fig. 5).It increased almost linearly with discharge power and SIE, and significantly decreased with air flow rate.Maximum concentration of 121 ppm was reached for 4 W and 0.25 L/min (i.e., the SIE of 960 J/L).The linear increase of N 2 O concentration with the discharge power was also observed in [41,51] and is probably linked with an increase of the electron energy, what leads to higher production of reactive nitrogen species important for N 2 O formation [56].Tang et al. observed an increase of N 2 O concentration with an increase of SIE up to 1500 J/L in a coaxial cylindrical DBD reactor [53].Above this level a saturated concentration of N 2 O was obtained that is related to an equilibrium between the processes of N 2 O formation and depletion that are mediated mainly by excited state of molecular nitrogen N 2 (A) (Eqs.11, 12) and excited oxygen atom O( 1 D) (Eqs.15,16).Similar trend in N 2 O monotonic increase up to a specific discharge power and following N 2 O saturation was also reported in [52].When using the lowest dry air flow rate (0.25 L/min) and, thus, the highest SIE (> 1000 J/L), a slight decrease of N 2 O concentration was observed (Fig. 5b).This effect may be associated with a transition of the discharge from "O 3 mode" to "NO x mode".A similar observation was also reported by Kogelschatz and Baessler [51].
Production yields (PYs) of O 3 and N 2 O were also calculated (Fig. 6).In contrast to O 3 and N 2 O concentrations, they decreased monotonically with the discharge power for all tested air flow rates (except for PY of N 2 O for 2.4 L/min that showed a nontrivial trend).The effect of increase of PY of O 3 with an increase of the air flow rate for a given discharge power was also observed by other authors [31,57].This may be a result of more efficient cooling of the ceramic substrate.Maximum PYs of O 3 and N 2 O were 77.6 and 1.0 g/kWh for the SIE of 25 and 240 J/L, respectively.
To assess "energy effectiveness" of O 3 production by the discharge, the results of our work are compared to those of other authors who used various other types of DBD operated at atmospheric pressure in dry air.Figure 7 presents PY of O 3 as a function of the SIE for all these DBD arrangements.All results presented have the same trend -the PY of O 3 decreases with an increase of SIE.Homola et al. investigated the multi-hollow SDBD with both electrodes embedded inside the ceramic and obtained much higher PY ( 16) of O 3 (> 150 g/kWh) than we obtained with the same discharge in configuration with the air-exposed electrode (Fig. 7) [31].Their PR of O 3 (~ 90 mg/h for PY of 205.5 g/kWh) was comparable with our results (~ 80 mg/h for PY of 77.6 g/kWh).Maximum PR of O 3 we obtained was ~ 220 mg/h for PY of 43.8 g/kWh.The other SDBD reactors are also compared: coplanar SDBD showed higher [42], while the other selected SDBD reactors showed comparable [58] or lower PYs of O 3 [43,56] than our discharge (Fig. 7).Finally, cylindrical coaxial DBD showed also almost identical results than our discharge [57] (Fig. 7).Lower PY of O 3 obtained with our multi-hollow SDBD compared to its another configuration (embedded electrodes) as well as coplanar SDBD may be related to arrangement of the ceramic substrate itself.The metal electrode exposed to the flowing air could contribute to the catalytic decomposition of O 3 , thereby reducing the PY of O 3 [59].On the other hand, higher PY of O 3 obtained with our multi-hollow SDBD compared to other SDBD-based reactors could be attributed to the reactor geometry that enables efficient air mixing with generated NTP.
In addition to O 3 , PY of N 2 O was also evaluated by Šimek et al. for the SDBD similar to coplanar SDBD and reached ~ 0.8 g/kWh for 250 J/L [43] that can be well compared to our results (~ 1.0 g/kWh for 240 J/L).Besides, maximum PR of N 2 O in our experiment reached ~ 3.7 mg/h.
Our results clearly showed that the air flow rate plays an essential role in both O 3 and N 2 O production by the discharge.Upon increasing the SIE up to ~ 350-500 J/L, O 3 production dominated over O 3 decomposition (Fig. 4b).With further increase of the SIE, the O 3 concentration started to gradually decline implying an increasing influence of O 3 Pekárek and Mikeš [58] and Kim et al. [60] reported the direct electron impact reaction (Eq.17) as a dominant reaction leading to O 3 decomposition at a gas temperature below 77 ºC.Kim et al. [60] further stated that at higher gas temperatures, the recombination of O 3 with O 2 becomes dominant (Eq.18).Another important O 3 loss reaction especially at higher temperatures is the reaction with excited oxygen atom O( 1 D) (Eq.19) [61].An increasing effect of both reactions (Eqs.18,19) lies in an exponential dependence of their reaction rates on temperature [62,63].However, many authors who observed a decrease of O 3 concentration for higher discharge power explained their results referring to the term "O 3 thermal decomposition" without further explanation.One can also find that reported gas temperature threshold, when O 3 thermal decomposition starts to dominate, varies significantly among the authors.For example, Homola et al. [42] reported a gas temperature of around 80 °C, Osawa and Yoshioka [64] a temperature of 120 °C and Jodzis even temperature above 200 °C.Here, it must be noted that in addition to gas temperature, the temperature in the microdischarge channels should be also considered as it is significantly higher than a gas temperature [61].In our experiment, we monitored the temperature of both gas and ceramic substrate surface.For the SIE of 600 J/L, the maximum temperatures were almost equal and not too high (~ 70 °C) [35], so we suppose that temperature is not a dominant factor affecting O 3 decomposition in our case.
Besides the reactions of direct O 3 decomposition presented above, the consumption of • O atoms that are essential for O 3 production must be also taken into account.At higher gas temperatures the • O atoms can be consumed by reactions with N 2 and O 2 molecules (denoted as M): that may also contribute to the overall decline of O 3 concentration [40,56].Based on the discussion above, we suppose that an observed gradual decrease of O 3 concentration from the SIE of 450-500 J/L is given by combination of several factors.Since concentrations of O and N species increase with an increase of the SIE, their reactions with O 3 (Eqs.19,20) may play important roles.Other influencing factors are increasing temperature and its effect on reaction rates, as well as increased consumption of • O radicals via reactions with N 2 and O 2 molecules (Eq.21).
As we have already mentioned, when the SIE exceeded ~ 1000-1100 J/L, the discharge transition from "O 3 mode" to "NO x mode" was accompanied by a significant change in the composition of produced gaseous species.According to Eliasson and Kogelschatz [48], the transition happens when the NO x concentration reaches the level at which • O atoms react faster with NO and NO 2 than with O 2 to form O 3 .As a result, an enhanced recombination of • O atoms takes place by the following reaction cycle [50,51]: In addition, the previously formed O 3 is effectively consumed by the reaction cycle involving both NO and NO 2 as follows [41,50]: As the SIE increases, the vibrationally excited N 2 species start to play a dominant role in formation of NO, which is then a major quencher of O 3 (the first reaction of the cycle Eq.23) [44,65].Since NO and NO 2 concentrations increase with an increase of the SIE and gas temperature [43,48], the two reaction cycles considerably affect the O 3 concentration especially under conditions of high energy densities (SIE > 1000 J/L).Whereas the first cycle (Eq.22) contributes to consumption of • O radicals, the second cycle (Eq.23) to consumption of O 3 itself.This confirms the role of the reactions Eqs.19 and 21 in the observed decline of O 3 concentration in dry air under high energy densities (SIE) (Fig. 4b), as the contribution of these reactions is enhanced by a presence of NO and NO 2 .
The maximum concentrations of NO and NO 2 in dry air were 130 and 395 ppm with corresponding PYs of 0.5 and 2.3 g/kWh and PRs of 2.4 and 11 mg/h for 1200 J/L, respectively.It is also important to note that when the discharge was operated in "O 3 mode" (i.e., SIE < 1000 J/L) regardless of working conditions, NO and NO 2 were not detected in the FTIR spectra.

Dinitrogen Pentoxide N 2 O 5 and Nitric Acid HNO 3
Until now, we have discussed the formation of O 3 , N 2 O, NO and NO 2 .However, in a strongly oxidising environment, NO 2 can further undergo the oxidation processes promoted by O 3 and • O as follows [18,51]: Consequently, N 2 O 5 can be formed via reaction of NO 2 with NO 3 : Figure 8 show the N 2 O 5 concentration as a function of discharge power and SIE for various dry air flow rates.Here it must be noted that N 2 O 5 production was not observed for the lowest air flow rate of 0.25 L/min.The maximum concentration of N 2 O 5 was 23 ppm (5 W, 1 L/min, 300 J/L).N 2 O 5 production was found to increase almost linearly with an increase of discharge power (SIE) (Fig. 8).Unlike the other evaluated compounds in "O 3 mode" (O 3 , N 2 O, HNO 3 ), a maximum N 2 O 5 concentration was observed for a moderate air flow rate of 1 L/min (Fig. 8a).N 2 O 5 production by the discharge is generally dependent on several factors including not only discharge power and air flow rate, but also the molar ratio of O 3 /NO, geometrical parameters, and temperature of the reactor zone [51,66].These factors may explain the observed nontrivial trends of N 2 O 5 concentration.
In humid air-like mixtures, nitric acid HNO 3 can be also produced by gradual oxidation of NO x mediated by • H, • OH and HO 2 • radicals formed from dissociation of water molecules.The following set of selected reactions can be considered [18,67]: Linear trend of HNO 3 production as a function of discharge power and air flow rate (Fig. 9) was similar to that of N 2 O production (Fig. 5).In our dry air conditions, only residual (trace) moisture could serve as a source of the radicals ( • H, • OH and HO 2

•
) that mediate HNO 3 production.The same effect of the presence of HNO 3 traces in dry air conditions was also observed by Braun et al. [41].Upon transition of the discharge from "O 3 mode" to "NO x mode", concentration of HNO 3 further slightly increased and reached a maximum of 148 ppm (5 W, 0.25 L/min, 1200 J/L) (Fig. 9a).
PYs of N 2 O 5 and HNO 3 are presented in the Fig. 10.Similar to PY of O 3 , the PY of N 2 O 5 followed the same increasing trend with an increase of air flow rate for a given discharge power (Fig. 10a).A maximum of 3.2 g/kWh was reached for 1 W and 2.4 L/ (27) min (SIE of 25 J/L).It is worth mentioning that for air flow rates of 0.5 and 1 L/min, the PY of N 2 O 5 showed almost constant values with an increase of discharge power.Unlike other evaluated compounds, the PY of HNO 3 during discharge operation in "O 3 mode" increased with an increase of discharge power for all tested air flow rates.The maximum of 1.9 g/kWh was found for 5 W and 1 L/min (SIE of 300 J/L) (Fig. 10b).Besides PYs, the PRs of both N 2 O 5 and HNO 3 were also evaluated with maxima of ~ 9.9 and ~ 9.5 mg/h, respectively.
Here, it must be noted that in order to explain the presence of N 2 O 5 and HNO 3 in the FTIR spectra, formation of NO and NO 2 in the "O 3 mode" of the discharge must be considered as they are both precursors for N 2 O 5 and HNO 3 formation.As we have already noted, NO and NO 2 were detected only when the discharge was operated in the "NO x mode".However, it does not mean they did not form in the "O 3 mode".We suppose that at least traces of NO and NO 2 were produced in the discharge and their absence in the FTIR spectra indicates either their concentrations below the detection limits or, more probably, their fast consumption by O 3 which is a dominant product of the discharge in the "O 3 mode" (Eqs.10, 24) [43,56].The second effect could be even more pronounced as the FTIR spectrometer was placed in some distance from the reactor chamber.Once the air came to the gas cell, NO and NO 2 were already consumed very probably by O 3 .Additionally, to monitor NO the electrochemical sensor was placed just behind the reactor chamber.However, even the sensor did not detect any NO, thus confirming its fast consumption in the discharge in the "O 3 mode".Consequently, NO 2 and NO 3 may further react and form N 2 O 5 (Eq.26) or even HNO 3 (Eq.35) if at least a trace humidity is present.Braun et al. reported that when a large surplus of O 3 is present especially at low gas temperatures, the equilibrium of Eq. 26 is shifted almost completely towards N 2 O 5 without detection of NO 2 and NO 3 what corresponds with our results [41].The same absence of NO and NO 2 was also observed by Abdelaziz et al. when using SDBD-based reactor [56].Hence, we can conclude that the gaseous species produced by the multi-hollow SDBD operated in the "O 3 mode" include O 3 , N 2 O, N 2 O 5 and HNO 3 .When the discharge is transitioned to the "NO x mode", in addition to N 2 O and HNO 3 , NO and NO 2 are also present.The observed concentrations of O 3 , N 2 O and N 2 O 5 in the "O 3 mode" agree with the typical feature of this mode reported by Eliasson and Kogelschatz [48]: the O 3 concentration was roughly two orders of magnitude higher than the concentrations of N 2 O and N 2 O 5 .

The Effect of the Air Relative Humidity
Humidity of a gas mixture affects the discharge from many aspects.It influences not only the physical processes of discharge formation, propagation, and its characteristics [68,69], but also it substantially affects and alters the plasma chemistry [67,70].To evaluate the effect of air humidity on gas-phase chemical activity of the multi-hollow SDBD, experiments were performed for various air relative humidity (RH) levels of 20, 40, 60 and 80% at atmospheric pressure with a constant room temperature of ~ 22°C, so the absolute humidity of the air (water content) was 3.9, 7.8, 11.7, 15.6 g/m 3 , respectively.The effect of humidity was studied for a fixed air flow rate of 1 L/min and discharge powers of 1-5 W, so the maximum SIE was 300 J/L.Under these conditions, the discharge was operated only in the "O 3 mode".Decreasing the air flow rate to 0.5 L/min allowed the transition of the discharge to "NO x mode" with the maximum SIE of 600 J/L.

Ozone O 3 and Nitrous Oxide N 2 O
Figure 11a shows the O 3 concentration as a function of discharge power for various air RHs.Our results confirmed a well-known fact that production of O 3 by the discharges significantly decreases with any impurity in the gas mixture including water H 2 O, hydrocarbons, etc. [57,71].A decrease of O 3 concentration was found non-linear.It means that the rate of O 3 concentration decrease is greater for low than for high levels of air RH (Fig. 11a).A non-linear decrease of 3 production was also reported by Chen and Wang when investigating DC corona discharge in humid air [70].The significant decrease of O 3 concentration when water vapour is present can be explained by the consumption of • O radicals triggered by HO 2 • and • OH radicals formed from dissociation of H 2 O and O 2 molecules: or by reaction with water (Eq.28) [57,70].Thus, consumption of • O radicals accelerates with an increase of air humidity leading to lower O 3 production.Decline in formation of • O radicals can also be caused by a reduction in electron density due to an electronegativity of water molecules [67].In addition, humidity also accelerates the relaxation rate of the vibration-to-translation of N 2 and O 2 molecules leading to the decline of O as well as N excited species [67].Another possible pathway for O 3 consumption may be by direct reactions of O 3 with • OH and • H radicals [70,72]: When the reaction of • H with O 3 (Eq.39) is followed by the reaction of • OH with • O (Eq. 37), giving the net reaction of • O with O 3 (Eq.19), the • H and • OH radicals can act as a "catalyst" to reduce O 3 concentration (similar to NO x in the reaction cycle Eq.23) [72].
Figure 11b shows N 2 O concentration as a function of discharge power for various air RHs.For low discharge powers (< 2 W), the maximum concentration of N 2 O was obtained in dry air, while for higher discharge powers (> 4 W), it was reached in air with RH of 20% (Fig. 11b).In general, production of nitrogen excited species N 2 (A) important for N 2 O production decreases with an increase of air humidity.This is due to very efficient quenching of N 2 (A) species by water molecules [49].This explains a decrease of N 2 O concentration when increasing the air RH from 20 to 80%.However, an increase of N 2 O concentration with air RH from 0 to 20% for 4 and 5 W may indicate that other reactions leading to N 2 O formation may also play important roles under these specific conditions.Abdelaziz et al. [67] and Matsui et al. [73] observed an increase of NO and NO 2 concentrations upon increasing the air humidity, so the reactions of N 2 O formation moderated by NO and NO 2 (Eqs.13, 14) may dominate.
In addition to O 3 and N 2 O concentrations, their corresponding PYs were also evaluated (Fig. 12).Similar to O 3 concentration, the PY of O 3 followed the same trend with an increase of air RH: it significantly decreased for a given discharge power (Fig. 12a).Moreover, in air with RH higher than 40%, the increasing humidity had no effect on PY of O 3 for the highest tested discharge powers (4 and 5 W).In addition, the PY of N 2 O followed a trend similar to the concentration reaching maximum values in dry air (for 1 and 2 W) or in air with RH of 20% (for 4 and 5 W) (Fig. 12b).

Dinitrogen Pentoxide N 2 O 5 and Nitric Acid HNO 3
In humid air, N 2 O 5 was not detected in the FTIR spectra regardless of discharge power and air flow rate.Similar observations were also reported by other authors [41,51].This (36) can be explained by the complete consumption of N 2 O 5 by reaction with H 2 O leading to the formation of HNO 3 (Eq.35).
Figure 13a shows the HNO 3 concentration as a function of discharge power for various air RHs.It initially increased with an increase of air RH from 0 to 40%, but with its further increase, the concentration of HNO 3 declined.The initial increase of HNO 3 concentration can be easily explained, as precursors to its production (radicals • H, • OH, HO 2 • ) are expected to rise with an increase of humidity [49].In this sense, the decline of HNO 3 concentration at air RH > 40% seems counterintuitive.Similar trend of HNO 3 concentration increase followed by a decline was also observed by Matsui et al. [73] with a maximum concentration at air RH of 30%.They suggested that at higher air RH, HNO 3 produced by the discharge can be easily absorbed by the water which is previously adsorbed on the inner surface of the connecting tubes due to its high miscibility.If HNO 3 is absorbed by the water inside the experimental system before it enters the gas cell, it may cause a decline of HNO 3 in the measured spectra [67].In addition to concentration, the PY of HNO 3 followed a similar trend (Fig. 13b).The highest concentration of HNO 3 was 142 ppm corresponding to PY of 4.4 g/kWh (air RH 40%, 300 J/L).

Transition of the Discharge from "O 3 Mode" to "NO x Mode" at Higher Humidities
In a previous section, the effect of air RH production of gaseous species by the discharge was evaluated for air flow rate of 1 L/min and maximum SIE of 300 J/L, i.e., under "O 3 mode" conditions.To investigate the effect of air RH on transition of the discharge from "O 3 mode" to "NO x mode", the SIE must be further increased.Therefore, the air flow rate was decreased to 0.5 L/min obtaining the SIE up to 600 J/L.
The results showed that the air RH has a significant impact on the SIE value for the transition between the two discharge modes.Figure 14 show concentrations of O 3 , N 2 O, HNO 3 and NO 2 as a function of discharge power (or SIE) in air with various relative humidities (20-80%).According to these results it can be concluded that a presence of humidity substantially decreases the SIE value for transition between the two discharge modes.In dry air, the transition from "O 3 mode" to "NO x mode" was found between 1000 and 1100 J/L.In air with RH of 20 and 40%, the transition was observed between 500 and 600 J/L (Fig. 14a, b), while in air with RH of 60 and 80%, it was found around 450 J/L (Fig. 14c, d).In humid air under "NO x mode" conditions, NO was not found in the FTIR spectra.Although we did not observe it, it does not mean that it was not produced.In humid air, the radicals formed by water dissociation ( • OH, HO 2

•
) can mediate formation of NO through the following reactions [67]: These reactions can compensate for decreasing production of NO through reactions where • N, • O and N 2 (A) are involved [49].Moreover, NO can further undergo an oxidation to NO 2 or HNO 3 promoted by • O, O 3 and HO 2 • (Eqs.9, 10, 31, Therefore, we suppose that absence of NO and presence of NO 2 and HNO 3 in the FTIR spectra in humid air under x mode" conditions may indicate fast consumption of NO to higher oxidation products NO 2 and HNO 3 probably mediated by • OH and HO 2 • radicals.The maximum NO 2 concentrations were found in a range of 350-402 ppm for the SIE of 600 J/L (Fig. 14) with corresponding PYs of 4-4.6 g/kWh and PRs of 20-23 mg/h, respectively.Concentration of HNO 3 reached maximum of 229 ppm at air RH of 40% for 480 J/L (Fig. 14b) with PY of 4.5 g/kWh and PR of 17.9 mg/h.These results indicate that humid air promoted a production of NO 2 and HNO 3 compared to their production in dry air for the same air flow rate (0.5 L/min).A similar effect was also observed by Baerdemaeker et al. [74] using multihollow SDBD with both electrodes embedded inside the ceramic.
To explain decreasing SIE value for transition between the two discharge modes with an increase of air RH, one must consider the O 3 /NO x ratio.In the "O 3 mode", the O 3 / NO x ratio is typically very high ( ≫ 1) as O 3 dominates the products.On the other hand, in the "NO x mode" the ratio is typically low (< 1) due to very low O 3 concentration.Our results showed that the O 3 concentration substantially decreased upon increasing the air RH.Therefore, the increase of humidity can significantly decrease the O 3 /NO x ratio and, thus, may promote the transition between the discharge modes for lower SIE in contrast to dry air.

Conclusions
Evaluation of the gaseous species production by the discharge, i.e., the discharge chemical activity, is one of the key factors for its effective employment for practical applications.In this manuscript, the gas-phase chemical activity of the multi-hollow surface dielectric barrier discharge generated in a perforated ceramic substrate with an air-exposed electrode was presented.The unique discharge geometry allowed the carrier gas to pass through holes (hollows) of the substrate and, thus, enabled efficient production of gaseous species.The production of ozone O 3 , nitrous oxide N 2 O, nitric oxide NO, nitrogen dioxide NO 2 , dinitrogen pentoxide N 2 O 5 and nitric acid HNO 3 was evaluated in terms of concentration (ppm), production yield (PY in g/kWh) and production rate (PR in mg/h) under conditions of various discharge powers (1-5 W), air flow rates (0.25-2.4 L/min) and air relative humidities (RH; 0-80%).The effect of air flow rate was examined under dry air conditions, while the effect of air RH was studied for two air flow rates (0.5 and 1 L/min).
The work demonstrated a critical impact of both the air flow rate and the relative humidity on the prevailing discharge mode ("O 3 mode" vs. "NO x mode") and, thus, on production and composition of the gaseous species.Whereas the air flow rate determines the gas residence time in a discharge zone and gas heating, the air relative humidity influences the discharge from physical (processes of discharge formation, propagation, and its characteristics) as well as chemical (production of • H, • OH, HO 2 • radicals) aspects.When the discharge was operated in the "O 3 mode", O 3 , N 2 O, N 2 O 5 and HNO 3 were observed among the gaseous products, while in the "NO x mode", N 2 O, HNO 3 , NO and NO 2 were present.
In dry air, the results showed that the concentrations of O 3 , N 2 O and HNO 3 increase with a decrease of the air flow rate for a given discharge power, except for the lowest tested air flow rate (0.25 L/min) due to the discharge mode transition.On the contrary, the concentration of N 2 O 5 was found the highest for the moderate air flow rate of 1 L/min (23 ppm with corresponding PY of 1.2 g/kWh for 300 J/L).Moreover, concentrations of N 2 O, HNO 3 and N 2 O 5 monotonously increased an increase of discharge power in "O 3 mode", while concentration of O 3 showed a nontrivial trend for 0.25 and 0.5 Upon increasing the discharge power or specific input energy (SIE), the O 3 concentration firstly increased, culminated, and then gradually declined, or eventually dropped down to zero (for 0.25 L/min).A complete depletion of O 3 was related to the discharge transition from "O 3 mode" to "NO x mode" associated with an increase of NO and NO 2 with maximum concentrations of 130 and 395 ppm with corresponding PYs of 0.5 and 2.3 g/kWh for 1200 J/L, respectively.The maximum O 3 concentration was 1840 ppm with corresponding PY of 27 g/kWh for 480 J/L, whereas the maximum PY of O 3 (78 g/kWh) was reached at concentration of 272 ppm for 25 J/L.For a comparison, the maximum N 2 O concentration was 121 ppm with corresponding PY of 0.8 g/kWh for 960 J/L.
In humid air, the O 3 concentration substantially decreased with an increase of air RH for a given discharge power and a rate of reduction was larger for low than for high levels of air RH.The N 2 O concentration showed more complicated trend: for low discharge powers (< 2 W), the maximum concentration was obtained in dry air, while for higher discharge powers (> 4 W), it was reached in air with RH of 20%.Further, the maximum HNO 3 concentration (142 ppm) was observed at air RH of 40% with corresponding PY of 4.4 g/kWh for 300 J/L.On the other hand, N 2 O 5 was not detected at all under humid air conditions.
Overall, the results showed that a presence of air humidity substantially decreases the SIE (energy density) value for discharge transition from "O 3 mode" to "NO x mode".In dry air, a transition was found between 1000 and 1100 J/L, while with an increase of air RH from 20 to 80%, it gradually decreased from approximately 600 to 450 J/L, respectively.We suppose that upon increasing the humidity a decrease of O 3 /NO x ratio may be responsible for the observed effect.
The work demonstrated that the multi-hollow surface dielectric barrier discharge is capable of production of various gaseous species in a wide range of concentrations that can be controlled by proper discharge operating conditions.Therefore, we suppose the discharge may eventually find many possible applications, particularly in plasma pollution control and biomedicine.

Fig. 1 Fig. 2
Fig. 1 Experimental setup including discharge reactor and systems for electrical diagnostics and chemical analysis

Fig. 4
Fig. 4 Concentration of ozone O 3 as a function of a discharge power P and b specific input energy (SIE) for various air flow rates Q (dry air)

Fig. 5
Fig. 5 Concentration of nitrous oxide N 2 O as a function of a discharge power P and b specific input energy (SIE) for various air flow rates Q (dry air)

Fig. 6 Fig. 7
Fig. 6 Production yield (PY) of a ozone O 3 and b nitrous oxide N 2 O as a function of discharge power P for various air flow rates Q (dry air)

Fig. 8
Fig. 8 Concentration of dinitrogen pentoxide N 2 O 5 as a function of a discharge power P and b specific input energy (SIE) for various air flow rates Q (dry air)

Fig. 9 Fig. 10
Fig. 9 Concentration of nitric acid HNO 3 as a function of a discharge power P and b specific input energy (SIE) for various air flow rates Q (dry air)

Fig. 11
Fig. 11 Concentration of a ozone O 3 and b nitrous oxide N 2 O as a function of discharge power P for various air relative humidities (RH) (Q = L/min)

Fig. 12 Fig. 13 a
Fig. 12 Production yield (PY) of a ozone O 3 and b nitrous oxide N 2 O as a function of discharge power P for various air relative humidities (RH) (Q = 1 L/min)