CO2 Dissociation in Barrier Corona Discharges: Effect of Elevated Pressures in CO2/Ar Mixtures

The formation of carbon monoxide, oxygen and ozone in a barrier corona discharge (BCD) operating in pure carbon dioxide (CO2) and binary mixtures of CO2 and argon is studied. The asymmetric electrode configuration of the BCDs allows plasma operation at pressures exceeding 1 atm, up to 6 bar, at moderate high-voltage amplitudes below 15 kV. Charge–voltage plots and an equivalent circuit model are employed to characterize the electrical parameters at different pressures and gas compositions. Depending on these conditions and the voltage amplitude, full or partial coverage of the electrodes with plasma is obtained. The existence of an optimum pressure for power dissipation for each given operation voltage amplitude and gas composition can be confirmed and explained by the equivalent circuit model. Increasing the CO2 concentration in the working gas increases the mean reduced electric field strength E/N while pressure reduces it in the BCD. The CO2 conversion shows a maximum efficiency of about 4% at 1.5 bar for the gas mixture Ar/CO2 = 1:1 and a voltage amplitude of about 10 kV. The calculation of thermodynamic equilibrium parameters reveals that a relatively small increase in pressure can affect both, the equilibrium parameters and the reaction rates. As a result, the specific required energy for the reaction (ΔH/SEI\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta \mathrm{H}/\mathrm{SEI}$$\end{document}) shows an optimum, but only 8% of the electrical input energy is spent for CO2 dissociation at these optimum conditions.


Introduction
The mitigation of greenhouse gases has become one of the most challenging issues in the world today [1].Carbon dioxide (CO 2 ) contributes up to 65% of the greenhouse emissions produced in about 50 Gt yearly from anthropogenic activities.The increase of greenhouse gas concentrations in earth atmosphere is responsible for global warming [2].In consequence, finding solutions for CO 2 reduction is a priority research topic [3].On the other hand, carbon monoxide (CO), a likely product of CO 2 decomposition, can be considered a valuable chemical feedstock and enable Power-to-X technologies [4,5].
The effects of different operation parameters on DBD's conversion efficacy have been widely studied.These parameters are the specific energy input (SEI, which is determined by discharge power and gas flow rate [38]); reactor geometry (e.g.discharge gap width, dielectric material and its thickness) [39]; dilution of CO 2 in other gases (e.g.nitrogen, argon, helium) [40,41]; gas temperature [42,43]; application of external magnetic fields [44], change of the pressure [45].While Ponduri et al. [46].concluded that there should be no significant effect by pressure in plane parallel volume DBDs, Bak et al. reported a higher conversion rate and lower energy efficiency of CO 2 splitting by increasing the pressure (from 2.36 to 5.08 atm) in nanosecond-pulsed DBDs [47].Belov et al. reported that pressure increases not only tune the discharge current of the single microdischarges forming the plasma (and possibly the underlying plasma chemistry), but also triggers CO 2 conversion efficiency [48].DBDs operating at overpressures in air or oxygen [49][50][51] demonstrated the existence of an optimal pressure for ozone synthesis, namely 1.4 bar in air [51] and 1.5-2 bar for oxygen [50].Furthermore, studies on nanosecond pulsed surface DBDs operated at pressures between 1 and 6 bar in air gave evidence that the discharge dynamics changed with the pressure [52,53].It was also demonstrated that overpressure could influence plasma-assisted ignition systems [54] and the vibrational distribution function in pure N 2 discharges.There are a few works on CO 2 -containing discharges, mostly on CO 2 reforming with methane and hydrocarbons synthesis from syngas (CO + H 2 ) (Fischer-Tropsch process) using non-thermal discharges at very high pressures [55,56].Recently, Hosseini Rad et al. demonstrated the enhancement of the energy yield for CO formation by pressure increase from 1 to 2 bar in an argon-CO 2 gas mixture (ratio Ar:CO 2 = 4:1) [57].This study aims to develop a barrier discharge reactor for CO 2 conversion at pressures higher than 2 bar, but lower argon concentration in the gas.
Barrier corona discharges (BCDs) are a combination of a one-sided DBD and corona discharge, i.e. is the bare metal electrode that contains one or more sharpened pins to enhance the electric field strength.BCDs are employed in many applications such as surface modification and exhaust gas treatment [58].Beside the study of such applications, there are numerous investigations in single filament BCDs, i.e. between one metal pin and a dielectric-covered plane or semi-spherical electrode, see e.g.[59,60].The enhanced electric fields strength around the pins reduces the voltage amplitude required for breakdown.Since a higher voltage amplitude is expected for increasing pressure due to Paschen's law (for gap times distance g × p > 2 bar mm), a pin or brush electrode should enable to operate plasma at higher pressure with still moderate high voltage amplitudes.
To explore the effect of pressure increase (p > 1 bar) in a BCDs the electrical discharge parameters, such as the discharge voltage, plasma spreading over reactor volume and discharge power are studied by applying an equivalent circuit model.Furthermore, the influence of pressure on the mean reduced electric field strength E/N is of high interest, as this similarity parameter relates to the energy gained by electrons between two collisions.Thus, it determines the electron energy distribution function (EEDF).The EEDF determines the rate coefficients of the electronic processes and thus, has an impact on the plasma chemistry.For example, a lower E/N (< 100 Td in pure CO 2 ) is supposed to enhance the contribution of vibrational excited molecules to CO 2 dissociation, while a higher E/N (> 100 Td) favors direct electronic dissociation, which has a higher threshold energy [61].In DBDs, the mean E/N is in the range 100-200 Td, thus direct electronic dissociation dominates.However, the question of whether E/N can be influenced and whether this has an impact on conversion in DBDs remains open and will be addressed in this contribution.

Experimental Setup
The conversion of CO 2 was studied in a stainless steel, pressure-tight reactor.The electrodes were arranged in an asymmetric coaxial configuration.The high voltage electrode, as schematically shown in Fig. 1a, was a steel brush consisting of a large number of sharp metallic pins.A glass tube with a copper tape attached to the outer surface was used as the dielectric and grounded electrode, respectively.Table 1 contains the main parameters and its values.The inner diameter of the glass tube and diameter of the brush electrode resulted in discharge gap of about 2 mm.
The pressure-tight chamber with the electrode arrangement was equipped with a front quartz glass window for side observation of the BCD.The reactor was further equipped with two lateral connections for gas inlet and outlet, as illustrated in the experimental setup scheme in Fig. 2. The total gas flow through the reactor was set to 200 sccm by mass flow controllers (Bronkhorst, EL-FLOW Prestige).Gas cylinders of Ar (AirLiquide, AlphaGaz 1, purity 99.999 vol.%) and CO 2 (AirLiquide, Lasal 2, purity 99.995 vol.%) were used.Different gas compositions (Ar/CO 2 ) were supplied by setting the individual gas flows.A pressure valve regulator (Swagelok, Serie BP) was used to adjust the pressure inside the chamber, which could be varied from 1 to 6 bar.When the pressure valve was open the system flushed against the actual pressure in the laboratory which was p = (1015 ± 37) mbar during the period of our measurements.Since the mean value is near to 1013.25 mbar = 1 atm were refer to this condition as "atmospheric" or "1 atm" in the text and figures.
Before each experiment, the chamber was purged with gas for 15 min.The pressure inside the chamber was monitored during the experiments using an analogue pressure gauge (Riegler standard mano, 110.46-KDE).
The discharge was operated by a sinusoidal power supply (Chroma, Mod.61604) connected to a high voltage transformer (Bremer Transformatoren GmbH, output 0-9 kV rms ).The voltage amplitude was up to 24 kV pp while a frequency of 1 kHz was used in all experiments.Electrical measurements were performed with fast voltage (Tektronix, P6015A) and current probes (Pearson, 2877) at the high voltage (HV) and grounded electrode, respectively.Both signals were recorded with a digital phosphor oscilloscope (Tektronix, DPO4104B, 1 GHz bandwidth, 5 GS/s).Furthermore, the charge was recorded by a 1:10 voltage probe across a measuring capacitor (0.47 μF) between the copper electrode and its grounding point.
The outlet gas was analyzed with an ozone analyzer (BMT, 963) installed about 50 cm after the pressure valve and a Micro Gas Chromatograph (Inficon GmbH, 3000 Micro GC).
The ozone analyzer also destroys all ozone.The sampling point of the Micro GC was about 8 m behind the ozone analyzer outlet.Micro GC was equipped with a thermal conductivity detector to measure the CO 2 and CO concentrations and has been calibrated for each gaseous component using standard gas cylinders.

Evaluation of Chemical Conversion
To examine the efficiency of the BCD on CO 2 dissociation the degree of CO 2 conversion (χ), CO production, energy efficiency (η), and energy yield (EY) were determined.The concentration of CO, [CO], is measured by the Micro GC directly based on the device calibration, while the conversion of CO 2 is calculated by Eq. (1).
where n in CO 2 and n out CO 2 are the molar amounts of CO 2 in the mixture before and after the reactor, respectively by assuming that the total amount of argon and CO 2 before the reactor is equal to 1 mol.The molar amounts n in and ) are derived from the measured concentration [CO] and CO 2 as calculated by the balances of the amounts being formed.This procedure considers the gas expansion by CO 2 splitting to CO, O 2 and O 3 similar as described in [62].
In the case of argon dilution, the effective conversion should be considered as follows: The CO 2 concentration in the gas mixture was varied from 10 vol.% to 100 vol.%, i.e. the molar fraction x CO 2 = 0.1 … 1.0.
(1) The specific energy input SEI is determined by the discharge power and the gas flow rate (at standard temperature and pressure).
Apart from the CO 2 conversion, energy efficiency is the most important index to evaluate and compare the performance of the BCD with other processes and reactors.By considering the reaction enthalpies of CO 2 decomposition to CO and 0.5 O 2 (Δ r H = 279.8kJ/ mol), the energy efficiency is defined as: In Eq. 4 χ eff is related to the reaction CO 2 → CO + 0.5O 2 and therefore, = n CO .Furthermore, the energy yield of CO production, as defined in the fol- lowing Eq. ( 5), was calculated to evaluate the efficacy regarding the formation of the target molecule CO where n CO and x CO are the molar amount and fraction of CO being produced, M W is the molar mass of CO, and V m is the molar volume.The second term accounts for the gas expansion effect (1.5 molecules formed from one CO 2 molecule).It neglects the formation of ozone, but this acceptable since most of the oxygen atoms form molecular oxygen.

Thermodynamic Considerations of CO 2 Splitting
The input data for calculating thermodynamic parameters were extracted with Factsage software (version 6.1) for all species based on their heat capacity and for the reaction temperature at T = 300-320 K.The thermodynamic considerations are characterized by the reaction variables Δ r G°, Δ r S° and reaction heat Δ r H°.The relation between them is given by: The individual standard reaction values are formed from the standard formation values: n i is the molar amount and i the stoichiometric coefficient of each compound of the reaction.
At ambient conditions, CO 2 splitting is an endergonic and nonspontaneous (Δ r g < 0) reaction.It can only take place by adding reversible work Δ r g and reversible heat TΔ r s.Here Δ r g and Δ r s are, in contrast to Δ r G and Δ r S, not the molar values, but refer to any chemical conversion.Gas flow rate L ∕ min In this paper, only the gas compositions upstream and downstream of the reactor were used for the thermodynamic considerations.The calculations were carried out by taking into account the reaction temperature, molar amount, and partial pressure of reactants and products, while elementary processes and individual reaction steps like vibrational and electronic excitations are not taken into account.
In this paper the Gibbs free energy of reaction (Δ r g), is calculated by the initial amount of 1 mol Ar and CO 2 and the difference between the respective sums of the formation Gibbs free energies Δ r g i of CO, O 2 , O 3 , Ar and CO 2 and their molar amounts and partial pressures before (n i,0 , p i,0 ) and after (n i , p i ) the reactor: The indices r and f correspond to the reaction and formation of each species, and R and T represent the molar constant for gases (R = 8.31 J K −1 ⋅mol −1 ) and gas temperature, respectively.To measure the gas temperature inside the plasma zone in a pressure-tight chamber is problematic.However, the temperature measured at the surface of the electrode by an IR thermometer was taken for an estimation.

Results and Discussion
The results section offers a description of the electrical, chemical, and thermodynamic characteristics of CO 2 plasma discharge under various gas mixtures and pressures.

Interpretation of Electrical Measurements
The measured voltage U(t) and charge Q(t) signals are used to elucidate the discharge voltage U D , the mean reduced electric field strength E∕N and the coverage fraction of the elec- trode surface .The charge-voltage plots (Q-V plots) such as shown in Fig. 3b show the form of a parallelogram.Its area is the energy dissipated into the plasma per HV cycle.Furthermore, the minimum sustaining voltage U min and the capacities of the discharge arrangement can be elucidated from it, as given in the figure.
The interpretation of the Q-V plots is based on the equivalent circuit in Fig. 3a which was introduced by Peeters and van de Sanden [63].The simplest equivalent circuit consists of two capacitances in linear arrangement.The first, C diel , represents the dielectric barrier, the second, C gap , the discharge gap.Both capacities form the total capacity , C cell .When the applied voltage amplitude U pp is below the threshold U min , the voltage across the discharge gap is not sufficient to cause electrical breakdown; and the Q-V plot will consist of a line with the slope C cell .If full breakdown happens for U pp ≫ U min , the parallelogram will form and its sides have the slopes C cell and C diel .
However, in our BCD reactor, the discharge appears non-uniform in the discharge gap.The plasma itself is filamentary, i.e., consisting of many individual discharge channels, which develop between the pin electrode and the inner surface of the dielectric; and spread over the surface [59,64].The diameter of the microdischarge foot on the dielectric is much higher than the diameter of the streamer in the volume (some mm vs. 100 µm).The microdischarges do not appear simultaneously and do not penetrate the full volume between the electrodes or cover the entire dielectric surface.From a more microscopic point of view, the pins of the brush do not have identical length, shapes and orientations or can be contaminated by dust or other surface coverage.Consequently, there are locations with a lower or a higher voltage threshold, see Fig. 1b.Therefore, at low overvoltage, the electric field strength is not sufficiently high at all pins and microdischarges will only appear at the longest pins.When the applied voltage is high enough, the electric field strength is sufficient at more pins, the plasma will appear in the entire discharge region volume.This somewhat less defined breakdown voltage manifests in the lower right and upper left corner of the Q-V parallelogram.They are not sharp, but smooth.Furthermore, the steepest slope of the parallelogram, named effective dielectric capacity diel , will be lower than C diel in case of partial surface discharging.The equivalent circuit in Fig. 3a considers this by dividing the electrode area into a non-discharging (characterized by parameter ) and a discharging part (parameter ) with + = 1 .These parameters are calculated as follows.
The effective dielectric capacitance is then calculated as diel = C cell + C diel .As higher the voltage amplitude U pp as lower is the value of .For full coverage (obtained at U pp ≫ U min ) we see → 1 and = 0 .Figure 4b shows this for a BCD in The result (based on average measured curves for U(t) and Q(t) ) is also presented in the Fig. 5 (purple curve).As seen there, the gap voltage proceeds the applied voltage (due to the electric field of the residual surface charges) and is almost constant during the active discharge phase.The plateau value of the gap voltage is the discharge voltage U D , a macroscopic or averaged discharge parameter for a DBD with several simultaneous microdischarges.
The discharge voltage U D can be determined from the Q-V plot also, either from U min or from the voltage difference ΔU as follows [63,65].
Thus, U D and ΔU are not necessarily the same, in particular at low overvoltage and par- tial surface discharging [70].It can be underestimated if this effect is not considered.However, as shown in Fig. 6, U min increases with voltage amplitude, U pp .The calculation of U D as the voltage threshold for the discharge gap includes the actual capacities (see Fig. 4a), but shows a similar trend to that of U min .For the conditions in this figure, the saturated value for the discharge voltage corresponds to an electric field strength of about 1.9 kV/ mm.The lowest value of U D at 10 kV pp gives a voltage difference of 0.9 kV which cor- responds to a gap width uncertainty of about Δg = 50 µm.However, these differences can be also due to different shapes and orientations of the pins or by impurities such as dust particles or layer on the pins.Here r is the mean radius of the discharge gap r = r out +r in 2 and N A , p are Avogadro's number and the pressure inside the reactor.The non-SI Unit of E/N is 1 Td (Townsend), equal to 10 -21 V•m 2 .
The energy per HV cycle E and the discharge power P can be determined from the area of the Q-V plot: or calculated by Manley's formula, which was developed for the characterization of ozonisers, as reported by Kogelschatz [68].In its latter form, it includes the partial surface discharging effect.It converges into the original equation for full discharging with ζ diel = C diel [57,69].
Figure 7 shows the measured power for the BCD in pure CO 2 at 1 atm (black symbols).The blue line is the power calculated with the original Manley equation with ζ diel = C diel = 150 pF and the lowest minimum sustaining voltage of 7.7 kV pp .The red symbols are calculated power taking into account the partial surface discharging effect, i.e. ζ diel and the actual measured values of U min (see Fig. 6) are taken into Eq.( 16).There is a fair agreement between the data extracted from the Q-V plot area and the calculation with the adapted Manley formula, while the original equation (blue line) overestimates the power at lower voltage amplitudes.The minimum sustaining voltage derived from the interpolation of the symbols to P = 0 W results in a higher value (9.3 kV pp ) than obtained in the ( 14) experiments.However, the 7.7 kV pp only represents the ignition at the longest metal pins, while at higher voltage amplitudes shorter pins also contribute to the discharge.

Influence of Gas Composition and Pressure on Plasma Operation
Figure 8a shows front photos of BCDs operating in different binary mixtures of Ar and CO 2 with 10-100 vol.% of CO 2 at 14 kV pp .The microdischarges attach to the tips of the brush electrode and spread over the inner surface of the dielectric tube.The corresponding discharge power is P = (6.0 ± 0.2) W in all figures.Figure 8e shows the corresponding current measurements.The filamentary character of the plasmas is resembled by the current spikes in the active discharge phases with a duration of about a quarter period of the high voltage (250 µs).The appearance of different groups of microdischarges, as obtained in the singlefilament BCD arrangement in [59,64], cannot be confirmed here.The microdischarges are distributed throughout the entire discharge volume.The discharge appears brighter as the Ar concentration increases with is also resembled in the current signals; with increasing Ar the more filaments are formed.As plotted in Fig. 8b, U D and U min increase with CO 2 concentration.This is in agreement with the known Paschen curves and can be explained by the molecular nature as well as the electronegativity of CO 2 [57].Argon reduces the breakdown voltage.Thus, the value of mean E/N also increases linearly with CO 2 concentration.The difference between U D and U min is smaller for full discharging, as explained with Eq. ( 13) due to the higher diel .These values are presented in Fig. 8c and compared with the calculated value of C diel (black dashed line).The partial surface discharging is affected by both the gas composition and the applied voltage amplitude.For low CO 2 concentration the discharge voltage is the lowest and even 10 kV pp are sufficient to cause full breakdown.The effective capacity exceeds the calculated C diel due to the spreading of the discharge beyond the electrode area.However, a much higher voltage amplitude of 14 kV pp is required to achieve nearly complete discharging in pure CO 2 , where propagation beyond the electrode area is not as pronounced as with argon.
The discharge power obtained from Q-V plots as a function of U D for different gas compositions but constant U pp show an inverted quadratic behavior (dashed lines), see Fig. 8d.These power parabolas are well depicted by the Manley Eq. ( 16) with P ∝ U min U pp − U 2 min and U D ∝ U min , see Eq. ( 13).This equation also explain the shift of the maximum power to higher values with the voltage amplitude.The calculated power parabolas (dashed lines) fit better for lower U D (i.e. higher Ar) because of less pronounced partial discharging.For higher CO 2 concentrations the original Manley equation overestimates the power, as obtained also in Fig. 7.
So, for a constant voltage amplitude there is an optimal gas composition, fulfilling the condition U pp = 2U min = 2U D diel ∕ diel − C cell , see [68].In case of limited high volt- age amplitudes, it is possible to optimize the power by the gas composition.In our case, at U pp = 10 kV pp the maximum discharge power is obtained for the gas mixture CO 2 :Ar = 3:7, while for 14 kV pp it is at CO 2 :Ar = 1:1 and for 18 kV pp at CO 2 :Ar = 4:1 (marked by the arrows).
First, the effect of pressure on electrical characteristics was investigated for Ar:CO 2 = 1:1 at U pp =19.5 kV pp .Under these conditions, full discharging is achieved.As shown in Fig. 9a, U D and U min increase with pressure linearly, as predicted by the Paschen law (for g × p > 2 bar × mm).Although both electric field strength E and density N depend on pressure linearly, the mean E/N decreases with pressure from 75 Td at 1 bar to about 40 Td for 4 bar.This could be explained by an offset of the voltage threshold.Extrapolating the fitting lines gives a nonzero value of U D and thus U D ≈ U D,0 + c × p (with c as a constant) and finally E∕N ∝ U D,0 ∕p + c .Since U D varies with pressure, the plot of power vs. U D in Fig. 9b shows the same inverted parabola as obtained for the gas composition.For the amplitude of 19.5 kV pp , the optimum pressure would be about 1.6 bar, which corresponds to U D = 3.9 kV and U min = 9.7 kV pp , fulfilling the condition U pp = 2U min .In agreement with this, the corresponding current slopes in Fig. 9c show lower discharge activity with increasing pressure.The current pulses appear weaker and less intense; the active discharge phases become shorter.The decrease in the number of microdischarges with pressure is in agreement with Belov et al. [48].As is also evident from the corresponding Q-V plots, the transferred charge decreases with pressure.
The impact of discharge activity by increased pressure is also visible in the front photos in Fig. 10a.For 3 and 4 bar, fewer microdischarge channels are obtained than for 1 and 2 bar.The plasma power and the specific energy can be adjusted by the voltage amplitude.In Fig. 10a and b the voltage was set to obtain SEI = 2 or 3 kJ/L.Since the number of microdischarges decreases while the power remains constant, the energy per microdischarge must increase with pressure.Belov et al. also reported higher current amplitudes for higher pressures and voltage amplitudes [48].The luminosity of the plasma correlates with the areal fraction coefficients α and β as shown in Fig. 10b.The coefficients were calculated with Eqs. ( 12) and ( 13) from the experimentally measured capacities shown in Fig. 10c.For the lowest pressure, the β coefficient slightly exceeds 1, i.e., the plasma covers little more than the entire electrode area.For pressures above 2 bar the voltage amplitude of 19.5 kV pp is not sufficient for complete discharging, and at 4 bar only 62% of the area is covered.
Based on the results of the electrical characterization, the operational parameters for the chemical investigations were chosen.The role of gas composition was first investigated at atmospheric pressure.Then, the pressure was varied in the mixture Ar:CO 2 = 1:1.In these experiments, the voltage amplitude was set to values that ensure almost surface discharging, but kept constant, i.e., the plasma power and SEI were different, which has to

Effect of Ar Dilution
Figure 11 shows power, CO 2 conversion, CO production, energy efficiency, and energy yield as a function of CO 2 concentration in argon (inlet concentrations of 10, 25, 50, 75 and 100 vol.%) for a constant applied voltage amplitude of 14 kV pp .At this amplitude the power shows little variation with gas composition and is almost constant at P = (6.0 ± 0.2) W (see Figs. 8d and 11a).Graphs (b) and (c) demonstrate that absolute CO 2 conversion decreases drastically with addition of CO 2 , while effective conversion and CO concentration increase significantly.Ramarkers et al. [70] and Lindon et al. [71] discussed CO 2 splitting in Ar:CO 2 mixtures and found a higher absolute conversion with the addition of Ar.Schiorlin et al. also reported a higher CO concentration for CO 2 mixed with Ar at a constant SEI in a coplanar DBD [34].They explained this effect by stronger microdischarges or higher electron density and mean electron energy and, thus, a higher rate coefficient for dissociation processes.However, the effective CO 2 conversion drops with the addition of Ar, due to the lower CO 2 content in the gas mixture and the lower reaction rates.Therefore, since SEI is almost constant, energy efficiency and energy yield of CO are maximal in pure CO 2 , namely at 12% and 26 g CO /kWh.These values are comparable to those obtained in other DBDs in this SEI range [69].The fact that the filaments penetrate a slightly smaller volume of the plasma reactor for higher CO 2 concentrations (see Fig. 8a and e) has no consequence on efficiency.It should be mentioned that a considerable portion of the gas flows along the brush pins, thus not being penetrated by filaments.This is different from most other coaxial reactors in which the entire gas flows through the active discharge zone.Since CO formation is mainly due to electronic dissociation, which is only present in microdischarges, this is a limiting factor for the effectiveness and efficacy of the BCD reactor.From this point of view, the achieved energetic yield is promising.

Effect of Pressure Increase
As mentioned above, the pressure increase was studied in two different experimental campaigns.First, the gas mixtures Ar:CO 2 = 3:1, 1:1 and 0:1 were used, while the voltage amplitude remained constant.This resulted in different plasma powers, SEI and mean E∕N values (see Fig. 9).In the second set of experiments, the power and thus, SEI were kept constant by adjusting the voltage to the gas composition and pressure.It should be kept in mind that the pressure increase did not influence the SEI , since the reactor was operated with the same gas flow rates (given in liters at standard pressure and temperature, L n ).In fact, the total number of gas particles exposed by the discharge increases with pressure, but is also flowing slower through the reactor enhancing the mean residence time of the gas in the plasma zone.Therefore, the calculation of EY is also not affected by pressure, as long as the gas flow rate under standard conditions is considered.
As shown in Fig. 12b, c, for Ar:CO 2 = 1:1 at 19.5 kV pp , CO 2 conversion and CO generation show a maximum at about 1.5 bar.The produced concentration CO at 1 bar is (15,100 ± 700) ppm, about three times higher than the produced at 4 bar with (5300 ± 400) ppm.Similarly, CO formation for the Ar:CO 2 = 3:1 mixture at the same experimental conditions increases from about 10,000 ppm at 1.5 bar to 11,000 ppm at 3 bar and then decreases to 7,800 ppm at 5 bar.There is a similar inverted parabola for pure CO 2 , where the CO concentration decreases from about 17,500 ppm at 1 bar to 3900 ppm at 3 bar.Based on interpolation, the parabola's maximum was observed at 1 bar or even lower.Belov et al. also reported a maximum conversion at 1.5 bar in a coaxial DBD with pure CO 2 [48].However, it must be considered that the power differed significantly in these experiments, as shown in Fig. 12a, because of the significant increase in breakdown voltage with pressure.In fact, the same inverse parabolic behavior of power is obtained for chemical conversion.Consequently, and EY are almost constant at (5 ± 1) % and (17 ± 2) g CO / kWh, respectively, for all selected gas compositions.As mentioned in [72], SEI is the most crucial parameter for CO 2 conversion.For SEI < 10 4 J/L a proportionality with CO formation and SEI is obtained.In order to keep the SEI constant, the voltage amplitude was adjusted.The absolute conversion and CO concentration for SEI= 2 kJ/L ( P= 6.7 W) in Ar:CO 2 = 1:1 is presented in Fig. 13a.The applied voltage amplitudes are shown at the top of the diagram.At first glance, these results do not reveal significant differences.However, and EY show an optimum at 1.5 bar and then slightly decrease with the pressure; CO concentration drops from about 8,000 ppm to 6,830 ppm at 4 bar.
These results are interesting in several aspects.First, the pressure rise should increase the collision probability between CO 2 molecules and energetic electrons or other reactive species.On the other hand, Le Chatelier's principle for the gross reaction 2CO 2 ↔ 2CO + O 2 predicts a shift in equilibrium toward the side with the least amount of molecules, namely CO 2 at higher pressure.However, neither a significantly improved efficacy of CO formation nor a decrease is obtained.In [57] a nearly threefold increase of EY was obtained in a coaxial DBD reactor by doubling the pressure to 2 bar, but in a gas mixture Ar:CO 2 = 4:1.To investigate whether the role of pressure is different for the different binary Ar:CO 2 gas mixtures, we studied the CO generation, CO 2 conversion and EY for the different gas mixtures at 1 bar and 3 bar for a constant SEI = (1.8 ± 0.2) kJ/L by adjusting the voltage amplitude.As shown in Fig. 14a, CO production at 1 bar exhibits an upward trend as the CO 2 content increases from 10 vol.% (Ar:CO 2 = 9:1) to pure CO 2 .A remarkable difference in CO formation is observed between 1 and 3 bar for 10 vol.%CO 2 in Ar.At 3 bar, CO formation experiences a slight increase, but reaches a plateau at 11,000 ppm for all other gas composition.A lower absolute conversion, saturating at about 1% as illustrated in Fig. 14b, is achieved because of the CO 2 concentration.For 10 vol.%CO 2 in Ar, pressure enhances the energy yield by 50% to about 24 g CO /kWh.As the CO 2 content increases, the pressure effect becomes smaller, and for pure CO 2 it cannot be concluded that pressure has a significant effect; the energy yield remains at about 24 g CO /kWh.These findings are somewhat consistent with [57].
In Fig. 15, the results of the BCD are compared with the literature.The data for comparison were obtained in pure CO 2 at atmospheric pressure in empty plane parallel volume DBDs (a-c) or coaxial DBDs (without packed bed, d-g), and a coplanar DBD (i).Higher pressure was studied in one of the works of Belov et al. (e), reference [48].The work (h) includes the addition of argon [70].Only papers that present CO concentrations or CO 2 conversion together with a carbon balance or the selectivity of CO were considered for the figure.The approximate ranges of the literature results are visualized by the colored areas in the diagram.The experiments were operated in the power range P = 1...150W at differ- ent gas flows and thus, the SEI spans several orders of magnitude.Therefore, similar as in [48] a double logarithmic scaling was chosen for the presentation of the data.The dashed lines represent values of constantEY ; its theoretical maximum was marked by the full red line.The stoichiometric limit of the gross reaction (CO 2 → CO + 1/2 O 2 ) was indicated by the horizontal full red line.Indeed, it neglects the ozone formation; but, its concentrations were much lower than for O 2 (see next sub-section).
For all of the data, the values are found in an " EY corridor" between 3 and 30 g/kWh and the data obtained with the BCD for 1, 2 and 3 bar are within this corridor.So, the Fig. 15 Comparison of results with the literature: a-c planar DBDs with a [73], b [74], c [75]; d-h coaxial DBDs with d [38], e [48,76], f [77], g [78], h [70]; i coplanar DBD [34] BCD is competitive with the planar and coaxial reactors.However, a significant impact by the presence of the pin electrodes or the pressure increase cannot be concluded from this graph.Similar as in (h) the addition of Ar reduces the EY slightly.The SEI is a sole scaling parameter like in the other below about 10 4 J/L n .For a higher SEI the reten- tion time of the gas in the active discharge zone becomes important [38].
Obviously, looking at Fig. 14, the pressure impact is related to the presence of argon.According to the most recent reaction kinetic model for non-thermal plasma at atmospheric pressure in pure argon [79], the electron impact leads to ionization (R1) and excitation (R2) of argon atoms.Among all excited argon states, its metastable excited species (denoted as Ar* with E exc = 11.55 eV; 11.72 eV here) play a key role in these plasmas, as they form excimers (R3) or can more easily be ionized (R4, R5).Excimers and argon ions contribute to the formation of Ar 2+ ions by electron-excimer recombination (R6) and 3-body charge transfer (R7).The Ar 2+ is by far the most important ion under these conditions.Both ions undergo electron-ion recombination (R8).
In DBDs in CO 2 , the dominant dissociation process is direct electron impact (R9).Other processes are dissociative electron attachment and electronic excitation (R10, R11).Electronic dissociation via vibrational excited levels has a smaller contribution than (R9) [38].
Electron collisions lead to electron impact ionization (R12) and the formed ion can contribute to CO formation via dissociative electron-ion recombination and ion-ion recombination (R13, R14).The main loss mechanisms for CO molecules are 3-body collisions with oxygen atoms and collisional detachment of O − (R15, R16).
The addition of argon has a significant effect on the electron kinetics; it enhances the high-energy tail of the electron energy distribution function (EEDF) and thus, the electron temperature.This has a direct impact on the rate coefficients of processes (R9-R12).The analysis of the excitation and dissociation of CO 2 mixed in argon or helium in a radiofrequency discharge revealed significantly higher vibrational temperatures of CO 2 and CO molecules as well as higher conversion efficiencies for argon [80].This was explained by the different EEDF and more efficient creation of low-energy electrons for better vibrational excitation of CO 2 in argon.Taking into account, that Ar admixture decreases the mean E∕N (see Fig. 8b) a lower electron temperature and thus, lower rate of electronic dis- sociation should be expected.This trend is inline with the higher EY and being obtained in Fig. 11.
In case of Ar/CO 2 gas mixtures, charge transfer processes (R17) and (R18) effectively form CO 2 + ions.The enhanced formation of CO 2 + ion formation can promote CO formation via dissociative electron-ion recombination (R13) [70].In addition, argon metastable have enough energy to dissociate CO 2 (R19).Argon excimers emit VUV photons around 126 nm.The corresponding energy of the photons (9.84 eV) is sufficient for the photodissociation of CO 2 (R20).
Figure 16 summarizes the processes mentioned here.As outlined by the blue arrows, the processes (R3) and (R7) are 3-body processes.Their contribution increases with increasing pressure.On the other hand, CO recombination (R15) is also a 3-body process, which would be enhanced by increasing pressure as well.
The mean E∕N was found to decrease slightly with pressure in the BCD.From this, a lower electron temperature and rate coefficient for the electronic dissociation processes and higher fractional losses of energetic electrons to vibrational levels of CO 2 would be expected.It can still be assumed that direct dissociation (R9) or electron-ion recombination (R12) e − + CO 2 → 2e − + CO + 2 (E ion = 13.8eV) R14) will be predominant.Therefore, the enhancement of EY by pressure under Ar addi- tion cannot be correlated with the trend of the mean E∕N .The mean E∕N is not a direct measure of the reduced electric field strength in the microdischarge channel.It is known that its development is characterized by the streamer mechanism with local E∕N values exceeding the ignition electric field strength significantly [81].This calls into question the role of the mean E∕N for the plasma chemistry.To learn more about the oxygen by-products, we also analyzed these species, as discussed in the next sub-section.

Temperature, Ozone Generation and Oxygen Balance
Indeed, the discharge properties and chemical processes depend on the reaction temperature.It influences the mean electron energy, the density in the plasma reactor and the rate coefficients [67].In particular, ozone generation is affected by gas temperature [82].To get an estimate of temperature, the temperature of the inner electrode surface, which is in contact with the gas passing through the reactor, was measured with an infrared pyrometer (Fluke, 561 HVAC).It increased from room temperature to around 55°C within the first 30 min of discharge operation at a power of 10 W and then remained constant.
The ozone concentration at the outlet increased significantly within the first 2 min of the experiment to about 1770 ppm (see Fig. 17a), and then decreased to about 1100 ppm within about 15 min.The time scales of heating and ozone formation are in the same range, and ozone is known not to be heat stable [83].Switching off the plasma, the ozone concentration decayed within several minutes.Thus, the increase is related to the gas exchange and sampling, while the decrease can be associated with the temperature increase.After a constant gas temperature was reached, the ozone concentration remained almost stable.
To analyze in detail the effect of pressure (1-3 bar) on ozone formation, its concentration was measured after at least 10 min of discharge operation.The plasma was operated at SEI = (3.0 ± 0.1) kJ/L in pure CO 2 .The measured values in Fig. 17b show that more O 3 was generated when the pressure increased.For control, the monitored inner electrode wall temperature is shown.Pressure did not affect it, since the same flow was used in all experiments.The rate coefficient of this reaction depends on the reaction temperature.Collision with a third body (M = CO 2 , CO, O , Ar) is required to dissipate the surplus energy of the reaction between O and O 2 as translational energy and for impulse balance [84].Therefore, the reaction rate will increase with pressure, as obtained in the experiment (see Fig. 17b).
For the gross reaction 2CO 2 → 2CO + O 2 , due to stoichiometry, the ratio [CO]∕ O 2 should be equal to 2. Figure 17c shows significantly higher values which is related to a decrease in the O 2 molecule formation due to O 3 formation.Taking into account the O 3 formation, i.e., forming the ratio [CO]∕( O 2 + 3∕2 O 3 ) the value 2 is obtained as expected and for all pressures.Therefore, significant amounts of other oxygen-containing species such as carbonates, can be excluded.

Thermodynamic Aspects of Plasma-Based CO 2 Splitting
Several studies have performed thermodynamic analysis for CO 2 conversion [57,77,[84][85][86].Equilibrium thermodynamics cannot be applied to nonthermal plasmas.This is because Δ G ≠ 0 as a result of adsorption of reversible work Δ r g> 0. The thermodynamic calculations in this contribution are considering the concentration of different species in the ground state before and after the plasma reactor.Figure 17 shows several trends of Gibbs free energy (reversible work, Δ r g ), reversible heat ( TΔ r s ) and enthalpy ( Δ r h ) of the gross reaction CO 2 → CO + 0.5 O 2 calculated from the measured concentrations after the BCD together with the corresponding trends of SEI and wall temperature.Ozone formation was neglected due to its relatively low amounts.
The results of tree different experimental campaigns were analyzed.Figure 17a and b illustrate the influence of the gas mixture at fixed SEI and wall temperature.Figure 17c and d refer to different pressures, but constant voltage amplitude.In Fig. 17e and f, the plasma power and SEI remained constant for the different pressures due to high voltage amplitude adaptation.
The thermodynamic calculations show the conversion of electrical energy into heat in the system.The Δ r h is the amount of energy needed to split the corresponding amount of CO 2 .The difference between SEI and H is the energy loss that is converted to heat.The parameters for the chemical composition of the gases upstream and downstream of the reactor are used for the calculation of Δ r h = Δ r g + TΔ r s .Therefore, this value depends directly on CO production, increases with CO 2 content and decreases with pressure (compare Fig. 18b and Fig. 18 Electrical input energy and wall temperature as well as thermodynamic parameters (Gibbs free energy Δ r g, reversible heat TΔ r s, enthalpy Δ r h) for CO 2 splitting using BCD at a and b atmospheric pressure and diluting Ar in the mixture at 14 kV pp equal to (1.8 ± 0.1) kJ/L, c and d in a mixture of Ar:CO 2 1:1 at 19.5 kV pp and 1-4 bar and e and f at constant discharge power of (6.7 ± 0.2) W (error margins in some graphs smaller than symbol's size) d with Figs.14a and 12b).The influence of the temperature is negligible.In case of Fig. 18f, where pressure exhibits independence at constant power and Ar:CO 2 ratio Δ r g , r s , and Δ r h remain relatively constant since the CO production is also nearly constant.
While the thermodynamic parameters are directly determined by the change in gas composition, the SEI remains independent at the same voltage (Fig. 18a).This relationship is clearly illustrated in Fig. 19.The ratio Δ r h∕SEI indicates how much of the energy used (SEI) is spent for the actual reaction.The specific energy requirement, SER (or Δ r h ) is the theoretical energy required for the conversion of CO 2 to CO and O 2 .Figure 19 shows this fraction of energy as a function of SEI.In the case of pure CO 2, it is less than 8% in the BCD experiments.Obviously, more energy is absorbed in pure CO 2 for the dissociation, than in Ar/CO 2 (black points versus red points).A significant part of the energy is mostly converted into heat by the argon species.In addition, Δ r h∕SEI decreases with SEI for both cases, correlating with the tempera- ture increase as heat in the system.Therefore, from the thermodynamic point of view, it seems more promising to operate the discharge at lower SEI in pure CO 2 .

Conclusion and Outlook
A barrier corona discharge (BCD) is capable to operate a discharge in CO 2 at pressures above 1 bar, up to 6 bar, but with voltage amplitude less than 15 kV.Besides the field enhancement at the pin electrodes of the BCD arrangement, the addition of argon helps to reduce the required high voltage amplitude.The discharge voltage increases linearly with pressure and CO 2 concentration, while the power input follows the Manley equation extended by the surface partial discharging effect caused by the non-uniform discharge gap of the BCD.The discharge power scales linearly with voltage amplitude at full discharging, but inversely parabolic with discharge voltage and thus pressure.
The BCD enabled the operation at higher pressures and CO 2 fractions than in the coaxial DBD reactor studied in [57].Increasing the pressure has no dramatic negative impact on the energy yield in pure CO 2 .In a similar manner as observed in [57] a higher argon addition increases the CO 2 splitting efficiency with pressure, potentially due to increased 3-body collisions and the dominant role of dissociative electron-ion recombination.The mean E/N, determined from the discharge voltage, does not show a systematic influence on the chemistry and the efficacy of the CO 2 splitting process.The main disadvantage of the Fig. 19 Theoretical specific energy requirement to specific energy input ratio, as a functions of SEI extracted from Fig. 18 BCD compared to coaxial DBDs is that a certain amount of the gas flows through the pin electrodes without being penetrated by plasma.This reduces the overall performance of the Avoiding this, could lead to a higher efficacy.The comparison of the reaction enthalpy and the SEI shows that most of the energy is not used for CO production, but converted to heat.The efficiency of CO 2 splitting becomes higher when the SEI is smaller and the CO 2 inlet concentration increases.From a thermodynamic viewpoint the most promising operation is at lower SEI in pure CO 2 .
These findings advance our understanding of BCD plasma in CO 2 and its potential for CO 2 splitting applications, offering promising directions for future research and practical implementations in plasma chemistry.However, without a comprehensive analysis of the microdischarge parameters, the electron kinetics and the plasma chemistry in the whole, including oxygen and ozone, by simulations, it is impossible to provide a more sound explanation of the effect of pressure increase on the energy yield of argon-diluted CO 2 .Beside these aspects other chemical processes chemistry, e.g. the CO 2 reforming of methane are planned for future research.

Fig. 1 a
Fig. 1 a Schematic illustration of BCD, b coverage of the dielectric surface by microdischarges for two different overvoltage

Fig. 2
Fig. 2 Schematic diagram of the experimental setup

Fig. 6
Fig. 6 Discharge voltage and minimum sustaining voltage as a function of voltage amplitude in pure CO 2 at atmospheric pressure

Fig. 7
Fig. 7 Discharge power vs. voltage amplitude extracted from Q-V plot area and calculated with the Manley equation

Fig. 8
Fig. 8 Effect of Argon addition on discharge morphology and electrical characteristics.a Front view images of BCD operating in different dilutions of Ar in CO 2 at U PP = 14 kV pp (corresponds to P = (6.0 ± 0.2) W and SEI = (1.8 ± 0.1) kJ/L).b U D , U min, and mean E/N versus CO 2 concentration, measured at 14 kV pp .c ζ diel versus CO 2 concentration at different U PP .d Power vs. U D for different U PP ; lines are inverted parabolic fits based on Manley's formula.e Temporal development of discharge current at U PP = 14 kV pp

Fig. 9
Fig. 9 Effect of pressure increase on the electrical characteristics of BCD operating at U pp = 19.5 kV pp and CO 2 :Ar = 1:1.a U D , U min and mean E/N vs. pressure.b Power parabola.c Temporal development of discharge current as pressure increases

Fig. 10 a
Fig. 10 a Front photos of BCD operating in Ar: CO 2 = 1:1 for p = 1-4 bar at constant SEI of 2 and 3 kJ L −1 (corresponds to P = 6.7 W and 10 W). b Surface coverage coefficients (α, β) vs. pressure at U PP = 19.5 kV pp .c Related cell, gas, and effective capacitances vs. pressure

Fig. 11 a
Fig. 11 a Discharge power (with quadratic fitting curve), b absolute CO 2 conversion and CO concentration, c effective CO 2 conversion, and d energy efficiency of CO 2 splitting and energy yield of CO formation at atmospheric pressure by changing the CO 2 concentration in the mixture with Ar at 14 kV pp (equal to SEI = (1.8 ± 0.2) kJ/L)

Fig. 12 a
Fig. 12 a Dissipated power, b CO concentration, c CO 2 conversion, and d energy efficiency and energy yield of CO 2 splitting at a U = 19.5 kV pp versus pressure in Ar:CO 2 = 3:1 and 1:1 and for pure CO 2

Fig. 16
Fig. 16 Simplified scheme of dominant processes in DBDs in CO 2 /Ar gas mixtures

Fig. 17
Fig. 17O 3 formation and O 2 balance in CO 2 splitting.a Temporal evaluation of O 3 production from pure CO 2 , during 30 min plasma on at 10 W followed by a 2 min of post-discharge (plasma off).b Reactor wall temperature and [O 3 ] concentration at SEI = (3.0 ± 0.1) kJ/L constant, and c Oxygen balance with and without considering O 3 formation Fig. 17O 3 formation and O 2 balance in CO 2 splitting.a Temporal evaluation of O 3 production from pure CO 2 , during 30 min plasma on at 10 W followed by a 2 min of post-discharge (plasma off).b Reactor wall temperature and [O 3 ] concentration at SEI = (3.0 ± 0.1) kJ/L constant, and c Oxygen balance with and without considering O 3 formation