Impact of catalysis on n-butane oxidation in an RF atmospheric pressure plasma

The plasma catalytic oxidation of n-butane diluted in a helium oxygen RF plasma jet is used to study volatile organic compound removal to unravel plasma catalytic synergisms. The plasma conversion is tested for a stoichiometric n-butane oxygen mixture for varying plasma power and using a manganese oxide catalyst. It is shown that the interplay between plasma and catalyst is very complex. The catalyst enhances the power coupling, but also serves as a sink for oxygen atoms due to surface recombination. The surface processes are dominated by reactions of radicals and excited species from the plasma. The oxidation of n-butane at the catalyst surface is slightly enhanced. In total, however, n-butane oxidation without the catalyst is more efficient than with the catalyst, which constitutes an anti-synergism.


Introduction
Plasma-based conversion of molecules is of major importance for electrifying the chemical industry.With the demand for CO 2 neutral processes and the advent of renewable energies, many chemical processes that are based on using heat as an energy source have to be converted to use electricity.Plasmas are the most direct method to use electricity for species conversion [1,2], and many processes are currently investigated such as nitrogen fixation [3,4], CO 2 conversion [5][6][7][8] or plasma-based methane pyrolysis as a source of turquoise hydrogen [9].The disadvantage of plasma processes is their lack of selectivity in most cases, which led to the emergence of the field of plasma catalysis, where a catalyst is combined either in a plasma process (in-plasma-catalysis, IPC) or after the plasma (post-plasma-catalysis, PPC) to control the product spectrum [10].The simplest combination is PPC, where the product spectrum of plasma is converted and modified by a standard thermal catalyst downstream.Both aspects of the process, the plasma conversion and the thermal catalysis step, are reasonably well understood.For example, in plasmadriven removal of volatile organic compound (VOC) from a gas stream, a post-plasma catalyst can convert any residual CO to CO 2 [11].The IPC systems are more complex due to the influence of the catalyst on the plasma and the plasma species on any catalytic reaction.For instance, many experiments using packed bed DBD reactors show a positive contribution of the catalyst to species conversion.However, directly identifying the plasma catalyst synergisms in these systems is extremely challenging.DBD plasmas can be homogeneous or filamentary, but the latter mode dominates when coupled with a packed bed.The direct coupling between a catalyst nanoparticle in such a packed bed and a plasma filament is difficult to define.The dielectric properties of the catalyst can modify the plasma performance, and the plasma may contribute to an involuntarily heating of the catalyst so that an apparent plasma catalysis synergism may just constitute plasmabased heating.When the plasma acts only as a heat source, one would not regard this as plasma catalysis but still as thermal catalysis.Such a mode of operation may also often not be desired because there are many more affordable options to heat the catalyst than using a plasma.Remarkable progress has been made, especially by combining packed bed plasma experiments and global chemistry models in combination with microkinetic modeling of surface processes [4] also revealing the role of direct reactions by radicals [12].The reasoning regarding plasma catalysis mechanisms, however, remains very indirect due to the inherent complexity of these systems.
In this paper, we follow a different approach by using a much simpler reaction system and measuring as many species as possible in situ to provide enough constraints for a minimal data-driven model.Thereby, a plasma catalysis synergism can be isolated.Recently, we introduced a dedicated plasma channel device [13], consisting of an RF plasma in helium with admixtures of the molecules of interest.The species are analyzed by using infrared spectroscopy and molecular beam mass spectrometry.Due to the dilution of the precursor gases in a noble gas, secondary reactions are suppressed.Hence, the reaction chemistry is vastly simplified.The dilution in the noble gas also controls the temperature of the species, with a rotational temperature of the molecules close to the ambient temperature and a vibrational temperature potentially higher due to the balance between electron-induced excitation and vibrational quenching via noble gas collisions [14].Thereby, a non-equilibrium temperature distribution is maintained.
This system has been used to study CO 2 conversion [14] and methane oxidation [15] using a manganese oxide catalyst.In CO 2 conversion [14,16], no plasma catalysis synergism has been found up to the upper temperature limit of the setup of 200 • C [13].In the case of methane oxidation, a very small positive impact of the oxidation rate has been found due to the presence of the catalyst on the surfaces [15].Such a barely visible effect is conceivable because methane oxidation by a catalyst usually requires very high surface temperatures of 1000 K to allow methane adsorption to compete with dissociative oxygen adsorption.In this paper, this hydrocarbon oxidation is now tested for the much larger molecule n-butane, which is also a typical molecule to evaluate any VOC removal technique.Indeed, n-butane is a gaseous alkane, rather nonreactive because of its strong C-C and C-H bonds and more difficult to dissociate compared to aromatic compounds.Therefore, n-butane is a good test molecule for the evaluation of VOC removal techniques.Since n-butane is much larger than methane, it is expected that n-butane may adsorb and interact with the catalyst surface at much lower surface temperatures compared to methane.As a catalyst, we selected MnO 2 , which is known to oxidize CO to CO 2 at 450 K.By using this setup, we analyze the impact of plasma catalysis on n-butane oxidation

Plasma channel device
The plasma channel device used in this study has been described previously [13,14].It is based on a plasma slab of 1 mm thickness, confined by two glass plates that cover the copper electrodes.The top electrode is connected to a 13.56 MHz RF generator via an impedance-matching network.The bottom electrode is grounded.The gas is injected at one side and flows through the chamber before being exhausted at the end of it.The whole system is enclosed in an aluminum body, in which large quartz or KBr windows are inserted to allow for optical access from the four sides of the plasma slab.The plasma volume is 26 mm × 13 mm × 1 mm.A schematic of the plasma chamber is shown in figure 1(a).
Two variations of the plasma channel device are used: (i) an FTIR (Fourier Transform Infrared) setup with KBr windows, focusing an infrared beam into the center of the plasma channel to monitor species via their rotational-vibrational absorption bands, and (ii) an MBMS (molecular beam mass spectrometer) setup, where all windows are from quartz, but a macor block replaces one side window to house a quartz sampling capillary to guide species into the MBMS.The end of the quartz capillary is located at the center of the plasma channel, see figure 1(b).

Operation parameters
The gas mixture comprises 250 sccm of helium admixed with 0.31 sccm (0.125%) of n-butane and a variable amount of oxygen.The reference oxygen admixture is 2.03 sccm (0.8125%), as it corresponds to the stoichiometric ratio for full oxidation of n-butane to CO 2 and H 2 O: The residence time at these flow parameters is given by τ residence = V/ϕ with the plasma volume of V = 2.6 cm × 1.3 cm × 0.1 cm.This yields τ residence = 81 ms.
MnO 2 is coated on the glass plates covering the electrodes.MnO 2 is a well known catalyst for VOC oxidation [17,18].The α-MnO 2 catalyst is prepared by precipitation via the comproprotionation of KMnO 4 and Mn(NO 3 ) 2 .The samples were then calcinated at 460 • C, and the crystal structure was verified by XRD.The specific surface was determined by the BET method, yielding 37 m 2 g −1 .The catalyst was then applied to the glass plates via spray coating, yielding a loading of 3 mg cm −2 .For further details regarding the catalyst preparation and characterization, see [19,20] The measurement of the power dissipated in the plasma is performed using VI probes included in the electrical circuit of the chamber following: U RMS and I RMS correspond to the effective voltage and current and ∆ϕ to the phase shift between voltage and current.At first, a reference phase shift is measured without the plasma to account for the impedance of the setup.This is then subtracted from the phase shift measured when the discharge is on to quantify the absorbed power by the plasma.
The plasma power should predominantly scale linearly with the electron density, as can be deduced from global plasma models.However, the proportionality depends on the gas mixture and the design of the plasma setup since loss processes and the surface-to-volume ratio might differ.Consequently, comparing the data from the FTIR setup and the MBMS setup based on the same electron densities cannot directly be related to the same absorbed plasma power.Therefore, we analyzed the plasma composition measured in both setups at room temperature without the catalyst and observed the identical plasma composition if the absorbed plasma power in the FTIR setup of 6 W is matched with the data from the MBMS setup at 10 W. Apparently, the power coupling in the MBMS setup is less efficient because more power needs to be absorbed to reach the same plasma composition.This might be due to the exchange of a side window with a macor block, but most important, due to the insertion of the capillary into the center of the discharge, which is very invasive.The sampling through the capillary is expected to induce a pressure gradient in front of it and a local gas rarefaction.When comparing the electrical characteristic curve of the same chamber equipped with and without a sampling system, a shift toward higher plasma powers is observed with the sampling system.Consequently, to compare all data, we re-scaled the plasma powers from the MBMS setup by 0.6 to compare the species densities in both setups on the same plasma power scale.

Infrared diagnostic
The species densities in the plasma are analyzed by infrared spectroscopy in absorption using a Bruker Vertex 70 V FTIR spectrometer with an external MCT detector and an evacuated optical path focused in the center part of the plasma channel using KBr side windows.100 scans at a wavenumber resolution of nominal 0.3 cm −1 are taken for each spectrum.The IR absorption spectra are quantified based on the Einstein coefficients A ij , taken from the HiTRAN and HiTEMP database [21][22][23], and employing individual distribution functions for the population of each degree of freedom of the molecules, such as vibrations and rotations, to calculate the line strengths S ij .Details can be found in [16,24].

Molecular beam mass spectrometry diagnostic
A side window has been replaced by a homemade sampling system, already introduced in [13] to carry out the measurements.It consists of a 4 cm long capillary with a 100 µm inner diameter (Agilent type with phenyl methyl deactivation fused silica used for gas chromatography) embedded in an adapter to connect it to the differential pumping stage of the mass spectrometer.The capillary is connected to a triple-stage pumping system housing a HIDEN EPIC mass spectrometer with a chopper inside the first stage to create a very large beam-tobackground ratio in the signal [25,26].The entrance side of the capillary is positioned in the middle of the plasma chamber.

N-butane oxidation by the plasma without the catalyst at room temperature
At first, n-butane oxidation by the plasma alone is evaluated in the MBMS and the FTIR setup at room temperature without the catalyst but at varying plasma power.The measured densities of n-butane, CO, CO 2 , O 2 , and H 2 O in the MBMS setup and the measured densities of n-butane, CO, and CO 2 in the FTIR setup are shown in figure 2(a).There are no data points below 2 W for the MBMS setup because it is impossible to ignite a homogeneous discharge, presumably due to the invasive nature of the capillary in the center of the gas channel.An exact match of both datasets can be found when the absorbed plasma power in the MBMS setup is scaled by a factor of 0.6.The data show a depletion of O 2 and n-butane and a corresponding increase in CO, H 2 O, and CO 2 densities.The CO and H 2 O densities exhibit an intermediate character since their densities decrease again at very high plasma powers.The decrease of CO can be explained by its further oxidation to CO 2 .The decrease of H 2 O could constitute a back reaction at very high electron densities to H 2 and O 2 .As the oxygen balance is close to unity, one can assume that non-observable species like OH, which may be created in the plasma, are lost via recombination processes inside the sampling capillary, leading to H 2 O.This also shows that the complete oxidation of n-butane in the plasma does not necessarily produce only CO 2 and H 2 O, but a significant fraction of H 2 .This is already evident when considering the hydrogen mass balance: the C and H densities at high powers yield a sum of the CO and CO 2 density, almost twice the H 2 O density.This is measured C:H ratio of 1:1 in the products, although the C:H ratio in the feed gas is 4:10.The actual determination of the H balance is difficult because the quantification of H 2 by the molecular beam mass spectrometer exhibits a large error.The depletion of O 2 and n-butane also agrees with a simple rate equation model discussed below, shown as dashed and solid lines in figure 2(a).
The oxygen and carbon mass balances of all species measured with the MBMS setup and normalized to the initial density in the gas mixture are shown in figure 2(b).One can see that the balances are conserved, indicating that the diagnostic also detects all species produced.This results from the large helium dilution, simplifying the plasma chemistry.The balances might decrease slightly at high plasma powers.This could be due to the little plasma heating causing the overall densities to decrease according to the ideal gas law.

Influence of O 2 admixture
The influence of the O 2 admixture to the gas flow is analyzed in the MBMS setup at a constant absorbed plasma power of 4.8 W. The measurements are carried out without a catalyst.The densities of CO and CO 2 measured at varying O 2 admixture are shown in figure 3 The mass peaks at 28 and 44 may also originate from hydrocarbons such as C 2 H 4 and C 3 H 8 , respectively.Except at the lowest O 2 admixtures (0.2% and 0.4%), the carbon balance, calculated from the n-butane, CO, and CO 2 densities, is conserved, assuring that any contributions from C 2 H 4 and C 3 H 8 are neglectable.In addition, no characteristic cracking patterns at masses below 28 and 44 have been observed.
The density of n-butane without the plasma constitutes 3 • 10 16 cm −3 .At 0% oxygen admixture, a depletion towards 1 • 10 16 cm −3 is observed due to electron induced dissociation.At an admixture of 0.2% of oxygen (1/4th of the hypothetical stoichiometric concentration necessary for full oxidation of nbutane to carbon dioxide and water), mainly CO is formed.With increasing O 2 admixture, the concentration of carbon species formed eventually shifts to CO 2 , as expected.
If we add the CO, CO 2 , and n-butane densities at very small O 2 admixtures, a small deficit in the carbon balance is observed compared to the carbon balance at high O 2 admixtures (not shown).This can be explained by the assumption that a carbon source is not considered, such as hydrocarbon fragments directly being produced by electron impact, or that some peaks in the mass spectrum are not properly identified.For example, at very small O 2 admixture conditions, the peak at mass 28 can correspond to the participation of CO and C 2 H 4 ) combined and the peak at 44 by the ones of CO2 and C 3 H 8 .
A maximum CO 2 concentration is found at around 0.6% O 2 admixture.This is typical for helium-diluted RF jets.At higher oxygen admixture, the creation of negative ions reduces the power coupling into the plasma, and thus the creation of O atoms becomes less efficient [27].The effect of reduced power coupling at very high O 2 admixtures can also be seen in the variation of the educts n-butane and O 2 .Up to an O 2 admixture of 0.6% to the helium flow, n-butane is completely depleted, and n-butane is converted into CO 2 .At higher admixtures, the dissociation efficiency of O 2 decreases due to the change in plasma performance.However, a constant amount of O 2 is still depleted to convert a fraction of n-butane into CO 2 .

N-butane oxidation by the plasma with and without the catalyst at 450 K
The plasma catalytic conversion is performed at 450 K, where the catalyst is expected to convert CO into CO 2 .The data presented in figure 4 show the measured species densities as concentrations (relative to the gas density at room temperature of 2.4 • 10 25 m −3 ) plotted versus generator power (a) and absorbed plasma power (b), respectively.The initial concentration of O 2 and n-butane is 0.8125% and 0.125% if referenced to the room temperature helium gas density.Since the experiments are performed at 450 K, these concentrations are changed to 0.8125% • 300 K/450 K and 0.125% • 300 K/450 K to be in correct relation to the room temperature helium gas density.In addition, the densities are also re-scaled based on the measured rotational temperatures to a density at 450 K.This is especially important for very high plasma powers where rotational temperatures of 480 K are reached.The temperature of the species can be deduced from the analysis of the rotational bands of CO and CO 2 in the infrared spectra.This yields rotational temperatures of 450 K consistent with the heating temperature of the setup.Only at a very high absorbed plasma power of 10 W, a small contribution of plasma-induced heating up to 480 K be observed.The vibrational temperatures of the molecules for the n-butane oxygen gas mixture are close to the rotational temperature.In principle, one could also expect much higher vibrational temperatures, as has been found for CO 2 admixtures [14].However, due to the large degrees of freedom of n-butane molecules, a quick vibrational-translational relaxation may compensate for any electron-induced vibrational excitation of all reactive molecules colliding.
One can see that the experiments using the catalyst show a reduced conversion compared to the experiments without a catalyst.This is already visible in the plotted densities versus the generator power in figure 4(a).This difference is even more significant when plotted against the absorbed plasma power.The absorbed plasma power with the catalyst is much larger than without the catalyst.This could be due to the different dielectric properties of the catalyst deposited on the electrodes and/or due to a higher secondary electron emission from the MnO 2 catalyst surface.
The detrimental effect of using the catalyst in an IPC process for VOC removal is consistent with findings in the literature [19,28,29] where the coating of a DBD electrode with MnO 2 reduced the plasma performance.In our setup, however, we use an RF plasma, where an additional very thin dielectric layer such as the MnO 2 coating should not affect the RF current through the system.
It is also interesting to note that the presence of the catalyst creates a higher CO density at intermediate powers compared to the experiments without the catalyst.The catalyst can enhance CO formation in the plasma.Such a difference might not be visible at very high powers since the very high oxygen atom density will eventually lead to the complete conversion to CO 2 irrespective of a possible contribution of a catalytic process.
One could also see that the depletion of n-butane at low powers is much stronger for the experiments without the catalyst compared to the experiment with the catalyst.This is supported by figure 4(c).It shows the sum of carbon and oxygen atoms within the species measured with the FTIR setup as non-normalized concentrations.At low powers, the carbon balance is slightly higher in the experiments with the catalyst compared to the experiments without the catalyst.One may argue that, without the presence of the catalyst, some of the hydrocarbons are converted into deposits or other hydrocarbons that are not accounted for in the carbon balance from the FTIR data.With the presence of the catalyst, these reactions are more selective, and n-butane is more readily converted into CO and CO 2 , both accounted for in the mass balance.At very high plasma power, the carbon balances of both experimental series with and without catalysts are identical.
The oxygen balance in figure 4(c) shows a strong variation with the absorbed plasma power because the FTIR data do not contain the densities of O, H 2 O, and O 2 .One can see that the O balance for the experiments without the catalyst saturates at a much lower power than the experiment with the catalyst.This is consistent with the better conversion in the experiments without the catalyst.

N-butane oxidation by the plasma with and without the catalyst at room temperature
The plasma catalytic conversion is also analyzed at room temperature, where the catalyst is expected to be inactive.The data presented in figure 5 show the measured species densities as concentrations in relation to the room temperature gas density of 2.4 • 10 25 m −3 plotted versus absorbed plasma power.The initial concentration of O 2 and n-butane is 0.8125% and 0.125%, respectively.One can see that the experiments using the catalyst also show a reduced conversion at room temperature compared to those without a catalyst.The impact of the catalyst on n-butane conversion is dominated by non-thermal effects such as altering the plasma power coupling, affecting surface coverages by Eley-Rideal reactions, or by ion and metastable impact at the plasma-catalyst interface.This will be discussed in the following.

Discussion
The data show that the presence of the catalyst has an adversarial effect on n-butane oxidation.This is in contrast to the knowledge of PPC, where the presence of the MnO 2 catalyst downstream from the plasma shifts the equilibrium from CO to CO 2 .However, the IPC seems to be strongly affected by other reactions, such as the direct impact of radicals at the surface.
To assess the reactivity of the catalyst, thermal catalysis schemes for methane oxidation over nickel and platinum catalysts are analyzed in appendix.Any models for n-butane oxidation on a MnO 2 catalyst cannot be found in the literature.Nevertheless, the well-known schemes for methane oxidation can be used to estimate the dominant surface coverages because the plasma produces CO, CO 2 , CH 4 and H 2 O and the adsorption of these species is well contained in these thermal catalysis models, see also figure A1.In principle, hydrocarbon conversion requires very high surface temperatures for the hydrocarbons to compete with oxygen adsorption.Above 1000 K, the conversion of methane to CO becomes efficient.At lower temperatures, the adsorption of O 2 and CO and CO 2 produced by plasma oxidation dominates.Thereby, the CO coverage is expected to dominate at room temperature, whereas the surfaces are expected to be O covered at 450 K. Based on this expectation from thermal catalysis, we now analyze our data as follows.
The power dependence of the plasma catalytic conversion of n-butane is analyzed using a simple plasma chemistry model based on a few equations only.A large global chemical model usually involves hundreds of species and thousands of reactions.The significant advantage of global models is their ability to cover different operation parameter ranges where dominant reaction pathways may also change.In the case of n-butane, many equilibrium reaction rate coefficients are missing, making the development of large global models for n-butane oxidation most challenging.Therefore, we choose a different approach.We devise the most simple chemical model, being able to describe our data set consistently.This can be regarded as a data-driven model development.In a simplified model, one must first estimate the dominant reaction pathways.As shown below, this strategy's validity can only be judged a posteriori when the model can describe the data correctly and the used rate coefficients are close to literature values.
The main characteristic of our experiment is the significant dilution in a noble gas by using only molecular admixtures below the percent range.Thereby, secondary reactions are minimized, and reactions with the surface are emphasized.A large surface area in relation to the plasma volume is essential to observe the impact of the catalyst on plasma conversion.Therefore, packed bed DBD configurations are preferred in most plasma catalysis studies.The drawback of DBD systems, however, is that the plasma is very inhomogeneous and filamentary, which makes identifying the actual reaction site (such as the footprint of the filament or the afterglow or a single filament, etc) very difficult.The design of our experiments ensures a dominant contribution of surface processes, as discussed in the following: • surface reactions versus convection: The time constant τ surface for species to reach the surface is given by the diffusion against the He background using a diffusion constant D of typically 1 cm 2 s −1 [30] and a typical travel distance of l = 0.5 mm, as the distance from the center of the plasma to the surfaces.Therefore, the time constant τ surface yields: The time constant for convection is determined by the standard flow of 250 sccm and the plasma volume, which yields τ convection = 81 ms.τ surface is 32 times smaller than τ convection , so species typically undergo 32 wall collisions during their residence time in the plasma channel.

• Volume recombination versus surface recombination:
The time constant for surface reactions τ surface is now compared to the time constant for volume recombination.A recombination reaction in the volume is a reaction of third order since a third collision partner is necessary to account for the energy and momentum balance of the reaction.This yields typical values for the rate coefficient of 10 −47 m 6 s −1 (i.e.reaction 2 below: CO + O + M → CO 2 + M).Since the recombining species are in a percent concentration and the buffer gas helium is at atmospheric pressure at a density of n 0 = 2.4 • 10 25 m −3 , the time constant for recombination of a radical with a collision partner at 1% density mediated by a third species can be estimated as: This shows that the time constant for volume recombination τ recombination is much longer than τ surface , indicating that surface reactions can have a significant impact on recombination rates.This comparison also depends on the crosssection for the particular surface recombination reaction.For example, if surface recombination exhibits a small crosssection, then the rate for volume recombination dominates.Consequently, it seems reasonable to choose tabulated volume recombination rates k literature from databases but to allow their modification in the fitting of the data by an adjustment factor f to postulate a rate coefficient in the model as This factor f also accounts for the fact that most recombination rates are not necessarily given with helium as the third collision partner.• Volume dissociation versus surface processes: A similar comparison can be performed to compare second-order rate coefficients in the plasma volume with surface reactions.The time constant for a typical reaction rate for dissociation is of the order of k dissociation = 10 −22 m 3 s −1 (i.e.reaction 1 * below: CH 4 + O → CH 3 + OH).For a radical concentration of 1%, this yields a time constant of Table 1.Reactions and corresponding rate coefficients at 450 K from the literature, correction factors fnocat and fcat are multiplied with these rate coefficients to yield a good fitting of the model with the data.The literature values are taken from the NIST database for the rate coefficients k 2 and k 5 , with the third collision partner M being either M = N 2 (k 2 ) or M = Ar (k 5 ).The NIST value for reaction 1 with rate coefficients k * 1 is a proxy for the reaction with rate coefficient k 1 .The rate coefficient k 4 is derived from the cross sections and converted into a rate coefficient for an estimated electron temperature of T e = 2.3 eV.The rate coefficient k 3 is taken from literature [32].

Model (no cat)
Model (cat) which is much longer than the transport time to the surface τ surface .It is conceivable that surface reactions may contribute significantly to any 2nd order volume dissociation reaction with a similar volume rate coefficient.This is again accounted for by a factor f to estimate In principle, invoking a detailed surface model to describe the results would also be possible.However, using a standard model from thermal catalysis is not straightforward because, in the IPC case, many active species interact with the adsorbates, for example, in Eley-Rideal reactions with unknown cross sections.This would introduce many free parameters, and we tried to keep the model as simple as possible.As will be shown below, despite its simplicity, literature values for a few reactions already describe all data very well.Future research will be devoted to unraveling the surface chemistry model more directly.
Electron impact excitation leads also to the creation of helium ions and helium metastables.Due to the much higher threshold energies for electron reactions with helium compared to reactions with the admixed molecules, helium excitation may not dominate.Nevertheless, these excited helium species eventually transfer their energy to molecular species in various quenching reactions so that, in the end, energy from the electron is transferred to the dissociation of admixed molecules.
Based on these arguments, we regard the following simplified scheme of five reactions using rate coefficients either from the NIST chemical kinetics database, from literature values, or from calculating the rate coefficients using cross sections from the LXCat database [31] and applying correction factors f when necessary.The selected reactions are listed in table 2 with (1) the oxidation of n-butane, (2) the oxidation of CO, (3) the electron-induced dissociation of CO 2 , (4) the electroninduced dissociation of O 2 , and (5) the recombination of O.The rate equations are then: The model assumes a linear relationship between electron density n e and applied plasma power P P as n e = 1.25 • 10 17 m −3 • P P [W].This scaling is derived from a simple global model, as already discussed in [13].In addition, detailed Helium metastable measurements in a pure helium discharge yield a similar scaling for the metastable density, as will be presented elsewhere.Since metastables are a good proxy for electron density, the scaling between electron density and absorbed plasma power seems reasonable.
The calculation starts at initial species densities given by the gas mixture and integrating the rate equations to 43.68 ms according to the exposure of the species to the plasma along the gas flow until they reach the middle of the plasma channel, where they are detected by FTIR.The rate coefficients are selected from the literature, as shown in table 1, but are adjusted by fitting parameters f to fit the data.Most parameters f are tried to be kept at unity, indicating that the literature values for the rate coefficients are already sufficient to describe the experiment.f values close to unity for the recombination reactions also indicate that the reaction rate for the volume process is already enough to describe the data and that the impact of surface reactions is small.The fit of the model to the data with and without the catalyst is shown in figure 6.A very good agreement is found despite the simplicity of the model.The choice of fitting factors f cat and f no cat for the different rate coefficients in the experiments with and without the catalyst is motivated as follows: • n-butane oxidation: Reaction 1 denotes a sum reaction of the conversion of n-butane into CO.In detail, it is expected that O atoms react at first with n-butane to create OH species, as exemplified by reaction 1 * for the oxidation of methane.After that, O atoms may remove further hydrogen atoms until eventually CO is formed, exemplified by reaction 1 * * with a high rate coefficient.Therefore, a reaction similar to k * 1 is the rate-limiting step.As a result, CO, OH, and H 2 are being formed, which is consistent with the mass spectrometry data at low powers.The sequence of reactions 1 * and 1 * * is condensed into one effective reaction 1 as a proxy.The rate coefficient k 1 is higher compared to k * 1 , because n-butane is a larger molecule than methane.The conversion is enhanced by the catalyst because f cat > f no cat .It is expected that n-butane may adsorb and react with adsorbed oxygen to form CO at the surface.At 450 K, incident CO should swiftly recombine with adsorbed O at the surface to form CO 2 , which desorbs.The desorption process, however, occurs in a plasma environment where surface-bound CO molecules may also directly desorb by electronic transfer to an unbound state due to the de-excitation of incident He metastables upon surface impact.Reaction 1 does not account for the generation of H 2 O, since water is not considered in the reaction scheme, because its density is also not measured in the FTIR setup.The mixture in our experiment is chosen so that complete oxidation of n-butane to CO 2 and water molecules occurs.The stoichiometric correct version of the complete oxidation of nbutane would correspond to C 4 H 10 + 13 O → 4 CO 2 + 5 H 2 O.In principle, it would also be possible to use this hypothetical higher-order reaction, but we tried to keep the reaction scheme simple and to restrict ourselves to 2nd or 3rd-order reactions only.Since reaction 1 is used only as a proxy for a more complicated reaction sequence, this simplification seems justified.Reaction 1 is a substantial simplification of the oxidation pathway of n-butane.In our experimental parameter range, however, the oxygen admixture is relatively high so that at small plasma powers, the gas phase is already dominated by CO and CO 2 , and a sum reaction for the direct conversion of n-butane to CO is sufficient.Consequently, the model cannot describe n-butane oxidation at small reactive oxygen densities.The straightforward model can explain the trends in the data at 450 K rather well, and all rate coefficients can be motivated by minor adjustments to literature values.The impact of the catalyst on the plasma chemistry is three-fold: (i) the presence of the catalyst leads to a reduced oxygen atom density since it acts as a sink for O via recombination at the surface, (ii) the catalyst enhances CO formation because n-butane is also oxidized at the surface to CO, which is then desorbed by the impact of plasma activated species such as helium metastables, (iii) the power coupling is much better for the surface with a catalyst, which compensates partly, the detrimental effect of the catalyst-induced surface recombination of oxygen.However, the efficiency of oxidizing n-butane to CO 2 is reduced by the presence of the catalyst.This is opposite to expectations from thermal catalysis.This model is now also applied to the data at room temperature, and a good fit can be reached as well, as shown in figure 7 using the rate coefficients as tabulated in table 2. All literature values are now calculated for a temperature of 300 K, and most factors f stay at unity or show a similar deviation, as already discussed in the model for the 450 K data.The only differences are the rate coefficients for reaction 1, which is close to the literature value at 450 K, because the rate coefficient for the oxidation of methane at room temperature (k * 1 ) is very small.Apparently, the oxidation of n-butane to CO remains an efficient process and the reaction rate for a thermal process at room temperature does describe plasma-based oxidation very well.At room temperature, the conversion of CO with O to CO 2 exhibits a higher rate than anticipated from the rate equations.From the analysis of thermal catalysis, it is expected that the surface contains a significant fraction of adsorbed CO.It is therefore reasonable to assume that incident O atoms very likely desorb CO in an Eley-Rideal reaction leading to the desorption of CO 2 .This process is less efficient at higher temperatures because the surface coverage of CO is expected to be much smaller, and the volume recombination rate of CO and O is much higher so that volume processes dominate over surface processes at 450 K.
It is important to mention that this model is only a first step to identifying IPC mechanisms because the reasoning depends sensitively on the accuracy of all literature values for the known rate coefficients.The impact of the catalyst changed the species densities by 30% at most, which lies within the error margin of all input parameters.A more advanced fitting of these reaction schemes using the rate coefficients from literature only as priors for a Bayesian-type analysis will be the subject of a future publication.
These experiments also show that identifying a plasma catalysis mechanism in detail is extremely challenging due to the mutual impact of the plasma on the surface and vice versa.In addition, most tabulated reaction rates are for reactions in thermal equilibrium and do not regard species that are excited in a non-equilibrium state.An example could be the presence of electrically excited species such as singlet delta oxygen, which might be present at a small density, but exhibit a much higher reactivity in the system.This is an inherent challenge in the field, but the presented experiments illustrate how these questions can be addressed using well-controlled simplified experiments and reaction schemes.
In summary, one may conclude that IPC, at least in the case of n-butane oxidation on a MnO 2 , can be dominated by direct reactions of excited plasma species (radicals, ions, metastables, vibrationally excited species) with the surface for species adsorption, but also for species desorption.A similar dominant contribution of radicals for species formation is consistent with findings in the literature on ammonia systems [12,36,37].The surface physics of the catalyst controls the concentrations of the different adsorbed species, expressed as surface coverages.Still, the final reaction product is not determined by a thermal desorption threshold but rather by a direct plasma-induced transfer of surface species to gas phase species.This is very different from a traditional thermal catalysis reaction sequence.

Conclusion
The oxidation of n-butane by IPC is investigated in a dedicated helium-diluted RF plasma setup by measuring n-butane, CO, CO 2 with FTIR at room temperature and 450 K and comparing experiments with and without a MnO 2 catalyst on top of the electrodes.The variation of species densities with absorbed plasma power can be described by a simplified chemistry model consisting of five dominant reactions.It is postulated that the presence of the catalyst leads to better power coupling and a reduction of the oxygen atom density in the gas phase due to surface recombination.Both effects almost compensate for each other.In addition, we postulate that the catalyst leads to CO formation at the surface at room temperature.These CO species are released from the surface by the impact of plasma-activated species, which is much more dominant at room temperature because the thermal catalytic conversion is inefficient.The accuracy of modeling this simplified reaction scheme is still limited, but using literature values for most of the reactions can vastly reduce ambiguities.Nevertheless, more detailed experiments, especially regarding the state of surface coverages, are urgently needed to generate more constraints for the global chemistry models and identify plasma catalysis synergisms more uniquely in the future.

Appendix. Thermal catalysis examples
Data on thermal catalysis for n-butane conversion by a MnO 2 catalyst are not available.Therefore, we assess the possible contributions of a catalytic surface process by analyzing thermal catalysis schemes for steam and dry reforming.Steam reforming of methane and water over a nickel catalyst to CO and H 2 has been studied by Maier et al [38].We adopted this set of rate equations to analyze the possible conversion of a methane oxygen mixture to CO 2 and H 2 (dry reforming).We solved the set of rate equations by integration to a time of 43 ms to account for the transport in our plasma channel device with its given plasma volume.As the initial gas mixture, we used a ratio of CH 4 and O 2 of 1:3 at a partial pressure of 0.01 bar for methane.The temperature dependence of the plasma densities and surface coverages is shown in figure A1(a).One can see that no CO 2 and H 2 is created at low temperatures.Only at very high temperatures above 1000 K, the oxygen surface coverage starts to decrease to allow for methane adsorption and the subsequent oxidation of adsorbed methane to CO and OH at the surface.At these high temperatures, CO and H 2 O desorb.In the catalytic oxidation of hydrocarbons, the initial adsorption step of an incident methane molecule is rate-limiting since the surface is covered by strongly bound oxygen.Only at very high temperatures, oxygen starts to desorb back to O 2 and the catalytic cycle can start.In the case of n-butane as the initial hydrocarbon, one might expect that larger hydrocarbons can more easily chemisorb.
Most important, however, is the fact that the plasma conversion of n-butane with oxygen may quickly lead to CO and CO 2 species in the gas phase, which can more easily chemisorb at the catalyst surface.This is analyzed using simpler CH 4 oxidation schemes and using only CO, CO 2 and O 2 as input species for a Ni [39] and a Pt [40] catalyst.An equal mixture of CO, CO 2 and O 2 with a partial pressure of 0.01 bar each is used as input mixture.The rate equations are integrated to 43 ms and the temperature dependence of the species and the surface coverages is shown in figure A1(b) for the Pt catalyst and in figure A1(c) for the Ni catalyst.One can clearly see, that the surfaces are predominantly covered by CO at low temperatures.In the range between 400 K to 500 K, the surface conversion to CO 2 sets in and the surface becomes predominantly oxygen-covered.In summary, one would expect that the surfaces are covered with CO at room temperature and with mainly O at the temperature of 450 K.

Figure 1 .
Figure 1.Schematic representation of the plasma chamber: (a) along the gas flow in the MBMS and FTIR setup and (b) perpendicular to the gas flow, depicting the MBMS sampling system additionally.

Figure 2 .
Figure 2. (a) Densities of CO, CO 2 , O 2 , and H 2 O (solid points) measured at room temperature without a catalyst by the MBMS setup (solid points) and by the FTIR setup (open points) as a function of absorbed plasma power.The absorbed plasma power in the MBMS setup is being re-scaled by 0.6 to compare the data for the same electron densities; for details, see text.The dashed line denotes the predicted depletion of O 2 using the rate coefficient for electron-induced dissociation.(b) Mass balance of carbon atoms and oxygen atoms in the measured products as a function of the absorbed plasma power.
(a), the densities of O 2 and n-butane in figure 3(b).The measured and hypothetical O 2 densities for the plasma off case are also shown as open symbols and dashed lines in figure 3(b).

Figure 3 .
Figure 3. (a) Absolute densities of the products: CO and CO 2 as a function of the O 2 admixture.The plasma power is maintained at 4.8 W (b) Absolute densities of the reactants: n-butane and O 2 as a function of the O 2 admixture.The initial densities of O 2 : theoretical (dashed line) and estimated from the signal acquired when no discharge is ignited are also displayed.

Figure 4 .
Figure 4. Quantified species densities of n-butane, CH 4 , CO, CO 2 at 450 K as a function of generator (a) and plasma power (b).Carbon and oxygen balance in all measured species by FTIR (c).

Figure 5 .
Figure 5. Quantified species densities of n-butane, CH 4 , CO, CO 2 at room temperature as function of plasma power.

Figure 6 .
Figure 6.Comparison of the data (a) measured by FTIR at 450 K with a simple rate equation model showing the variation of species densities (b) at varying plasma power.

Figure 7 .
Figure 7.Comparison of the data (a) measured by FTIR at room temperature with a simple rate equation model showing the variation of species densities (b) at varying plasma power.

Figure A1 .
Figure A1.Methane oxidation, dry reforming over a Ni catalyst (a) and CO oxidation on a Pt (b) and a Ni catalyst (c) by thermal catalysis.Gas phase species are denoted by dashed lines, surface coverages by solid lines.
• CO oxidation: Reaction 2 denotes the recombination of CO with O to form CO 2 , M denotes a third collision partner.Factors f cat and f no cat are unity; therefore, the volume recombination dominates, and the surface process does not affect the recombination rate.• CO 2 dissociation: Reaction 3 denotes electron induced CO 2 dissociation.The rate coefficient is taken from Morillo et al [32] as 10 −17 m 3 s −1 assuming a reduced electric field of Reaction 5 denotes the recombination of O atoms to O 2 .Here we use very different factors f cat and f no cat for fitting the data with and without a catalyst to account for the fact that the probability for surface recombination of O atoms on a SiO 2 surface is 100 times smaller than for a MnO 2 surface [35] with γ O on SiO2 = 0.003 and γ O on MnOx = 0.3 at 450 K.In principle, one would expect that the factors f cat and f no cat should also differ by a factor 100.However, these factors describe only a partial contribution of surface processes to a volume recombination rate.The absolute values for f are very sensitive to the scaling between power and electron density because reaction 5 is counteracting the electroninduced dissociation of O 2 .Hence, any inaccuracy in the correlation of the plasma power with electron density directly translates into these correction factors.More important is the difference between factors f cat and f no cat indicating that surface recombination is primarily enhanced when the catalyst is present.

Table 2 .
[32]tions and corresponding rate coefficients at room temperature from the literature, correction factors fnocat and fcat are multiplied with these rate coefficients to yield a good fitting of the model with the data.The literature values are taken from the NIST database for the rate coefficients k 2 and k 5 , with the third collision partner M being either M = N 2 (k 2 ) or M = Ar (k 5 ).The NIST value for reaction 1 with rate coefficients k * 1 is a proxy for the reaction with rate coefficient k 1 .The rate coefficient k 4 is derived from the cross sections, as tabulated in LXCat, and converted into a rate coefficient for an estimated electron temperature of Te = 2.3 eV.The rate coefficient k 3 is taken from literature[32]for very low reduced electric fields as typical for AP-RF jets.