Microwave plasma-based dry reforming of methane: Reaction performance and carbon formation

We investigate atmospheric pressure microwave (MW) plasma (2.45 GHz) conversion in CO 2 and CH 4 mixtures (i.e., dry reforming of methane, DRM) focusing on reaction performance and carbon formation. Promising energy costs of ~2.8 – 3.0 eV/molecule or ~11.1 – 11.9 kJ/L are amongst the best performance to date considering the current state-of-the-art for plasma-based DRM for all types of plasma. The conversion is in the range of ~46 – 49% and ~55 – 67% for CO 2 and CH 4 , respectively, producing primarily syngas (i.e., H 2 and CO) with H 2 /CO ratios of ~0.6 – 1 at CH 4 fractions ranging from 30% to 45%. Water is the largest byproduct with levels ranging ~7 – 14% in the exhaust. Carbon particles visibly impact the plasma at higher CH 4 fractions ( > 30%), where they become heated and incandescent. Particle luminosity increases with increasing CH 4 fractions, with the plasma becoming unstable near a 1:1 mixture (i.e., > 45% CH 4 ). Electron microscopy of the carbon material reveals an agglom-erated morphology of pure carbon nanoparticles. The mean particle size is determined as ~20 nm, free of any metal contamination, consistent with the electrode-less MW design.


Introduction
Plasma-based conversion of the key greenhouse gases, CO 2 and CH 4 , so-called dry reforming of methane (DRM), offers promise to drive 'carbon circularity' in chemical and fuel production systems, which is important in the transition to a net zero carbon economy [1,2].Plasma processes typically have very short start-up times, so are readily amenable to intermittent renewable electricity supplies [3], which continue to make up an increasing share of energy grids worldwide [4].
Plasma-based gas conversion of hydrocarbons has gained renewed interest recently [2,[5][6][7], including innovations to address 'hard-to-abate' sectors where de-carbonisation and de-fossilisation are particularly challenging (e.g., aviation [8] and construction [9]).Plasma-produced 'syngas' (i.e., H 2 and CO mixture) can be used as a feedstock in synthesis methods (i.e., Fischer-Tropsch [10,11]) to produce, for instance, liquid hydrocarbons, such as kerosene, or key chemical building blocks, such as methanol [2,12].End products could be sequestered as a non-gaseous product or combusted (i.e., releasing CO 2 again, but with generally a low to net zero carbon budget possible for the overall process).The 'circularity' of the enabled process will also depend on the source of CO 2 and CH 4 utilised, which can range from fossil natural gas, captured carbon from combustion, or biogenic sources (i.e., biogas) produced via anaerobic digestion of biomass [13].
MW surface wave plasmas, under atmospheric pressure conditions, offer the advantage of a high degree of ionisation (i.e., high levels of gas activation) and warm plasma conditions (favouring the endothermic chemistry of DRM).Furthermore, they operate without metal electrodes (i.e., advantageous in limiting energy loss to and erosion of metallic electrodes required in other reactor designs), and the plasma is formed in a simple quartz tube.In addition, employing a swirling gas flow with a surface wave plasma avoids plasma-quartz interactions, increasing the reactor's lifetime.Finally, they allow a quite large throughput due to the fairly high gas flow rates required to sustain a swirl flow (i.e., > 5 L/min) when compared with other lab-scale plasma reactor designs, such as gliding arcs, atmospheric pressure glow discharges, and spark configurations [7].
In this report, we explore MW plasma-based DRM into syngas, measuring key performance indicators, including conversion and energy cost, selectivity towards syngas and by-products, as well as the plasma exhaust temperature.One important aspect, which we believe has been overlooked to date [15,16], is understanding the role of the small (but typically impactful) solid carbon by-product resulting from the dissociation of the CH 4 fraction.Small particles ("dust") in a plasma can act as a sink or source for (plasma) charged species, with the degree of coupling Fig. 1. i) Illustration of our MW plasma reactor consisting of a solid-state MW power supply, circulator, auto-tuner, and tapered waveguide section terminated by a sliding short.The plasma is ignited inside a quartz tube where a swirling flow is injected.Sample analysis of the exhaust gas is carried out using an NDIR and a luminescence O 2 sensor.ii) Photo of the reactor in operation for 1 kW power and 10 L/min flow rate, for a CO 2 /CH 4 ratio of 55/45 with significant downstream carbon deposition on the quartz tube.Ignition takes place in a tapered section of the 2.45 GHz WR340 waveguide where a plasma is suspended at the centre of the tube.iii) In-waveguide photos of the plasma at different CO 2 /CH 4 ratios (i.e., A-C) at 10 L/ min.The camera viewpoint (position shown with arrow in ii)) is looking towards the quartz tube inside the waveguide.The onset of carbon formation with increasing CH 4 fraction is visible in iii) B & C, from the increased light emission at the edges surrounding the plasma filament (see text).
S. Kelly et al. depending on the size, abundance, and temperature of the particulates [66][67][68].Particulates can also become heated, directing energy away from plasma sustainment.In addition to its importance in maintaining a plasma, the success of any future downstream coupling of the plasma exhaust with catalyst beds (i.e., to exploit the high temperatures in the MW plasma exhaust [18]) will largely depend on the extent of 'coking' deactivation, a common hurdle for catalyst operation in DRM.Indeed, carbon formation plays a central role in industrial-scale plasma arc technologies such as the 'Hüls process' [69,70], to convert methane into hydrogen and acetylene, and the 'Kvaerner process' [70,71] for methane pyrolysis to generate carbon black and hydrogen.Therefore, in this paper, we study DRM in atmospheric pressure MW plasma, with special focus on carbon formation.Our work gives original insight into the impact of solid carbon formation on plasma-based DRM, which has not been the focus of previous research [2,[14][15][16][17][18].Further to this, we include characterisation of the produced carbon, as well as a novel quantification of water formation (a key by-product) in MW plasma-based DRM via an atom balance analysis which, to the best of our knowledge, has not previously been carried out.

Experimental details and analysis
In Fig. 1i) we present a schematic of our experiment.A photo is shown in Fig. 1 ii).The plasma reactor is powered by a solid-state 2.45 GHz MW source composed of a collection of power amplifiers (based on laterally diffused metal-oxide semiconductor (LDMOS) technology) from which the output powers are combined in a waveguide (WR340).This waveguide is connected via an isolator and an auto-tuner to a tapered waveguide section, including a 16 mm inner diameter quartz tube.The latter is mounted perpendicularly through a 20 mm coupling hole (i.e., < 1/4 wavelength in diameter), where the plasma ignition takes place.An auto-tuner, impedance analyser, and adjustable short are used to tune the electric field to optimal conditions for electrical breakdown and to sustain a continuously powered plasma with minimum reflected power (< 1%).The measurement of the absorbed plasma power, a key input parameter, is carried out using the Homer series autotuner from S-TEAM labs, using a six-port-reflectometer (SPR) method [72,73].Gas injection ports coupled with a helical insert allow a swirl or vortex flow within the quartz discharge tube.Upon ignition, a surface wave sustained mode [74,75] is generated, with an axially elongated warm plasma filament located at the tube center, isolated from the tube walls (see Fig. 1 iii)).The temperature of the exhaust is monitored using a k-type thermocouple [76] and is found to be in a steady state after approximately 15 min post-ignition at which point gas analysis is performed (i.e., the exhaust concentrations are not changing after this time).
Analysis of the exhausted CO 2 , CH 4 , CO, and C 2 H 2 is performed using non-dispersive infra-red (NDIR) absorption spectrometry (Rosemount X-STREAM XEGP Continuous Gas Analyser) [77].The NDIR detector divides the gas sample between absorption and reference cells.The physical measurement is based on an opto-pneumatic technique where the IR absorption results in gas heating, which causes gas to outflow through a thin detection channel connecting the two chambers, and this is monitored by a micro-flow sensor providing the output voltage [77].For H 2 measurement, a thermal conductivity detector (TCD) incorporated with our NDIR detector is used, which exploits the large differential of H 2 over other gases to conduct heat, along with signals from each NDIR channel for selectivity.The TCD is based on a Wheatstone bridge arrangement [77].For O 2 detection, we use a sensor based on the luminescence of a dye excited with IR light, where O 2 quenches the luminescence and predictably changes its intensity for real-time monitoring [78].Gas sampling is performed once the reactor reaches a steady state in terms of the exhaust temperature [79].In addition to the internal calibration of each sensor, we perform external calibration using pre-mixed standards for several points across each sensor range.
We calculate the individual conversion of CO 2 and CH 4 considering their mole fractions y i=CO2,CH4 (Eqs.E1-E2 below) with and without the plasma ignited (i.e., plasma ON/OFF).This is equivalent to the species mole fractions entering the plasma region (i.e., in = plasma OFF) and exhausted from the plasma (i.e., out = plasma ON).The percentage conversion χ CO2,CH4 (%) for each feed gas is then given by: Here, α accounts for the changes in the gas composition and molar flow rate.Indeed, the conversion of CO The measurement of α is performed using a real-time O 2 sensor [78] where a small O 2 admixture of dilution ratio β (e.g., ~10% dilution of the gas inlet mass flow with β = 0.1) is added to the cooled exhaust for plasma ON and OFF conditions (i.e., y probe = y O2 , Eq. E3).Here the measured ratio y probe (OFF) y probe (ON) corresponds to the molar flow changes from chemical reactions while the term 'β ( y probe (OFF) y probe (ON) − 1 ' accounts for molar flow changes from the (probing) dilution.Notably, for small (probing) dilutions (i.e., β≪1) Eq.E3 is approximated as α ∼ y probe (OFF) y probe (ON) .The use of a diluting gas probe to measure α has been previously reported in the context of plasma-based gas conversion [7,52,80].No oxygen is found in the exhaust mixture for the conditions investigated in our work.Notably, the addition of O 2 to the cooled exhaust (i.e., removing any ignition source) is confirmed to be 'chemically inert' in the same way as probing gases such as N 2 or Ar have been previously employed [7,52,80].Owing to the real-time nature of the gas sampling in this work, the gas expansion measurement is taken separately/sequentially to the other gas species measurements and not simultaneously as in earlier reports [7,80].This allows for a more straightforward analysis (detailed below).The total (or weighted) conversion allows for a direct comparison of performance between various DRM reports [81], particularly those employing different CO 2 :CH 4 ratios.The total (or weighted) conversion, χ tot (%), is defined as the conversion for each reactant weighted over their molar fraction in the inlet gas mixture: The energy cost (EC, either in eV/molecule or kJ/L) is expressed in terms of the total conversion χ tot , the ratio of plasma power P plasma (kW) to inlet gas flow rate φ IN (nL/min) (i.e., a ratio known as the specific energy input, SEI), and the molar volume of gas under standard reference conditions (i.e., 24.05 [L/mol] [7,82]).These are given as: EC tot Of particular interest for DRM is the energy cost of syngas production given by: EC syngas EC syngas ) . 24.05 This value should be interpreted as the amount of energy needed to form a certain amount of syngas [83].
The mole fraction of H 2 O formed by the plasma conversion is not directly measurable in our setup, as the liquid condenses out along the exhaust lines, with the exhaust sample dried before measurement.In order to indirectly estimate H 2 O, we can consider the atom balances [52].For the oxygen balance: We consider b O ∼ 1 − y H2O (out) in Eq. (E10), where y H2O (out) is the mole fraction of H 2 O which condenses from the plasma exhaust before measurement.Notably, the approximation of H 2 O here as the dominant liquid product is in line with earlier reports where sampling of the liquid fraction condensed from a similar CO 2 /CH 4 plasma is found to be primarily H 2 O with only faint traces of oxygenates [7].Further, we found that O 2 is not present in the exhaust, also in line with earlier reports [7,52].This leaves only CO and H 2 O as the primary routes for the conversion of oxygen atoms.By re-arranging Eq.E10 above, we can express the H 2 O molar fraction in terms of known values: Selectivity provides an overview of the distribution of conversion products.The selectivity is defined as the number of atoms that end up as a product containing said atom, with respect to the number of atoms that are available from our reactants CO 2 and CH 4 .The atoms in question are C, O and H.We calculate the selectivity of each atom (i.e., C, O and H) for our primary products CO, H 2 and H 2 O along with minor product C 2 H 2 as follows: Given its importance in both the 'upstream' plasma dynamics and 'downstream' utilisation of the exhausted syngas, we characterise the solid carbon generated during conversion.To determine particle size distribution and the basic morphology, we employ scanning electron microscopy (SEM) with energy dispersive X-ray (EDX) spectroscopy (FEI Quanta 250 FEG) [84].The carbon deposits are collected from the reactor wall and subsequently dispersed in chloroform by sonication.A drop of this suspension is left to dry either on an SEM stub prior to SEM imaging or on double sided carbon tape for EDX spectroscopy analysis.Further, Raman spectra were obtained on a Micro-Raman Horiba (Xplora Plus Microscope) equipped with a green 532 nm laser.The measurements were done at room temperature and in air atmosphere in the range of 50-4500 cm -1 Raman shift.The laser power was varied from 1% to 100% (not shown), while multiple spots of the sample were analysed (not shown).Finally, thermogravimetric analysis (TGA) was performed on a Mettler Toledo TGA-DSC 3 + analyser.The measurements were performed in O 2 with a flow rate of 80 mL/min., while the temperature ramped with 10 • C/min.until 1000 • C was reached.A sample of graphite powder (particle diameter of ≤ 20 µm, Fluka Inc.) was also measured as a benchmark for crystalline carbon.

General considerations and plasma operating conditions
Syngas production is given by the following overall reaction: A competitive pathway to reaction R1 is the formation of, and (eventual) condensation to, liquid water, which results from the reaction of O atoms (formed from the splitting of CO 2 ) with H atoms (from the splitting of CH 4 ).Possible overall reaction pathways for water formation include the well-known methanation and reverse water gas shift reactions: Other by-products of note include C 2 H x molecules, where acetylene (C 2 H 2 ) is typically the most prominent, especially at high CH 4 fractions.Solid carbon is typically a small (< 1%) but impactful by-product in terms of its physical effects on the plasma and the utilisation prospects downstream.A key pathway for solid carbon production is due to pyrolysis of CH 4 via stepwise formation of C 2 H 6 (and H 2 ) and subsequent dehydrogenation of C 2 H 6 to C 2 H 4 and of C 2 H 4 to C 2 H 2 which finally dissociates to form solid carbon (and H 2 ) [85].The overall reaction is given by: Another possible pathway for solid carbon production is the disproportionation of CO given by: We investigate a range of inlet mixtures approaching 50/50 CO 2 / CH 4 at a fixed power of 1 kW at 2.45 GHz and a total mass flow rate of 10 L/min with a specific energy input of 6 kJ/L.Corresponding photographs of the discharge inside the waveguide center (i.e., at the ignition point) are shown in Fig. 1 iii) The visible appearance of the plasma is clearly sensitive to the mixture fraction.At 70/30 CO 2 /CH 4 , the filament, suspended at the tube's center, is visually consistent with CO 2 MW plasmas previously reported [19,23,24].The plasma region is contracted (radially) to approximately one skin depth (i.e., the characteristic penetration depth of MW energy), consistent with the plasma's dielectric properties in the surface wave plasma mode [24].At the 60/40 mixture, we begin to see increased light emission around the filament, including an orange incandescence at the far edges.Further, carbon particles are collected more visibly in the downstream walls of the quartz tube for this mixture (as indicated in Fig. 1 ii).The increased light intensity at higher CH 4 fractions is therefore correlated with the onset of significant carbon production (resulting from the CH 4 pyrolysis side-reaction).The 'white' incandescence is associated with solid carbon temperatures exceeding 1500 K, while orange hues are associated with incandescence from cooler particulates at the edge regions surrounding the plasma filament [86].Notably, much of carbon is observed to be 'pushed out' from the central (plasma) region, depositing on the quartz tube walls downstream [87].At the highest stable CH 4 mole percentage (i.e., 55/45 CO 2 /CH 4 in Fig. 1 iii) the (visible) light emission intensifies considerably and masks the visibility of the central plasma region.There is also a substantial increase in the downstream deposition of carbon on the quartz tube walls for this mixture (see Fig. 1 ii).At higher CH 4 fractions, e.g., 50/50 CO 2 /CH 4 mixtures, the discharge becomes very luminous, unstable, and eventually extinguishes before any steady state (> 15 mins) can be reached.
Notably, previous works [15,16] on MW DRM used a higher specific energy input (i.e., 12 kJ/L [15] vs. 6 kJ/L used in this work) to achieve (stable) ignition/conversion at 50/50 CO 2 /CH 4 .Unfortunately, these operational conditions are beyond our reactor capabilities at present.It is clear, however, that the impact of solid carbon formation is significant on the DRM performance in MW plasmas, with a considerable increase in the SEI seemingly required to form a stable ignition for only a small change in CH 4 fraction (i.e., + 5%).Investigation of the produced carbon will be described later in this paper, but it is known that the plasma dynamics change in the presence of solid carbon, which can act as macroscopic charge carriers [66,67].Indeed, aggregates/agglomerates of carbon particles typically acquire a negative charge in a plasma due to the high mobility of electrons, which reduces the electron density, and this could weaken the (electron) power coupling, which sustains the discharge.Further to this, carbon is known to be absorbent of MW radiation [88] which could weaken the power coupling to charge carriers (i.e., where energy is directed to heating the carbon) leading to difficulties sustaining a plasma.In our experiment, these key factors, related to onset of conditions with significant carbon formation, are possible explanations for the plasma extinguishing when the CH 4 fraction is only incrementally changed from 55/45 CO 2 /CH 4 to 50/50 CO 2 /CH 4 .
CO and H 2 are the main gaseous products formed, with H 2 O as a significant by-product.The CO and H 2 mole percentages found in the exhaust for the 70/30 CO 2 /CH 4 mixture are 41.0 ± 0.3% and 23.0 ± 0.1%.Small mole percentages of C 2 H 2 , ranging from 0.70 ± 0.16% to 1.58 ± 0.10% for 70/30 CO 2 /CH 4 to 55/45 CO 2 /CH 4 are measured as well (and other C 2 H x (x = 4,6) products may also be present, but in still smaller quantities).Notably, the exhaust is found to be free of O 2 (i.e., 0%) for all conditions investigated.These trends are generally in agreement with earlier reports [7,52].
We observe significant condensation (water) in the exhaust.This is attributed to the methanation and reverse water gas shift pathways (see reactions R2-4 above).The mole percentages of H 2 O are estimated here by considering the O atom balance (see Eq. E11 above), with values of 13.6 ± 3.2% at 70/30 CO 2 /CH 4 and 12.1 ± 3.3% at 65/35 CO 2 /CH 4 , while at higher CH 4 fractions, the H 2 O mole percentage is reduced to 7.2 ± 4.7% at 60/40 CO 2 /CH 4 and 6.7 ± 5.2 at 65/45 CO 2 /CH 4 .As a sidenote, the hydrogen and carbon atom balances are potentially useful for estimation of other unknown (i.e., unmeasured) species mole fractions, such as the gaseous C2 hydrocarbons (i.e., C 2 H 4 & C 2 H 6 ) and solid carbon.These species, however, constitute a small fraction of byproducts.Due to levels of uncertainty in the measured and inferred values (i.e., H 2 O) needed to estimate them, an accurate determination is not currently possible.Indeed, the quantification of the solid carbon directly could be challenging as this would involve the collection of all the material deposited in the reactor for a particular run time.Further, measurements of the C2 hydrocarbons (e.g., using NDIR/UV, FTIR or GC) should consider appropriate calibration at low ranges (i.e., to minimise uncertainty).
In Fig. 2 ii) (left axis) the corresponding energy cost is given.We obtain promising energy costs ranging from 2.8 ± 0.1 eV/molecule to 3.0 ± 0.1 eV/molecule, or from 11.1 ± 0.4 kJ/L to 11.9 ± 0.5 kJ/L, for total (weighted) conversion levels of ~50-54% (i.e., 49.6 ± 2.1% to 53.5 ± 1.7%).The corresponding energy costs for syngas production (Eqs.E8 and E9) range from 1.9 ± 0.1 eV/molecule to 1.7 ± 0.1 eV/ molecule, or from 7.8 ± 0.1 kJ/L to 6.9 ± 0.1 kJ/L.These total energy costs here are quite good when compared to previous works on DRM in MW plasmas (see Table 1), although at clearly lower conversion levels than [15,16].Indeed, Zhang et al. [14] investigated atmospheric pressure MW plasma conversion using a submerged 'wire' electrode arrangement.The power was pulsed and limited to low values (~120 W); also the (mass) flow rates were limited to 200 mL/min, achieving CH 4 and CO 2 conversions of 71% and 69%, respectively, for an energy cost of 6.5 eV/molecule [2,14].Chun et al. [15] investigated MW plasma-based DRM in a 'torch' design [89] with swirling flows of 30 L/min and powers of 6 kW (i.e., SEI ~12 kJ/L) in 50/50 CO 2 /CH 4 mixtures, achieving 68% and 97% conversion for CO 2 and CH 4, respectively.This yielded an energy cost of ~3.6 eV/molecule.Under similar SEI conditions (i.e., also 12 kJ/L), Sun et al. [16] employed a surface wave MW plasma reactor design with a swirling flow arrangement at 10 L/min and 2 kW power, reporting 91% and 96% conversion for CO 2 and CH 4 , respectively, and yielding an energy cost of ~3.2 eV/molecule.Another investigation by Chun et al. [17], using a different MW torch design, achieved conversion levels of 33% and 43% for CO 2 and CH 4 with swirling flows of 30 L/min and powers of 3 kW (i.e., SEI ~6 kJ/L) in 50/50 CO 2 /CH 4 mixtures (without a catalyst bed).This yielded an energy cost of ~3.9 eV/molecule (see Table 1).Combinations of MW plasma and downstream catalyst beds for DRM have also been investigated.Chun et al. [17], added a Ni catalyst bed adjacent to the discharge tube, comparing this hybrid design with the non-catalytic case.Considerable gains were found with the inclusion of the downstream catalyst, as the conversion increases from 33% to 68% and from 43% to 97% for CO 2 and CH 4 , respectively, compared to the non-catalytic scenario under similar conditions.This yielded an energy cost of ~2.0 eV/molecule (see Table 1).Notably, the authors reported significant issues with carbon deposition blocking the pores of the Ni nano-catalyst, leading to deactivation.Cho et al. [18] also applied a Ni catalyst, and achieved a CO 2 and CH 4 conversion of 68% and 60%, respectively, for 50:50 CO 2 /CH 4 mixtures at reduced pressures (~0.04 bar) and low (mass) flow rates (100 mL/min) for powers up to 1.5 kW.The authors reported much higher energy costs in the range of 93-343 eV/molecule across different mixture ratios [2,18].Notably the authors again reported issues with carbon deposition, leading to deactivation of the Ni catalyst employed.
In terms of energy cost, we reach slightly lower values of 2.8-3.0 eV/ molecule compared to the best performance of ~3.6 eV/molecule reported by Chun et al. [15] and ~3.2 eV/molecule reported by Sun et al. [16], although typically using a lower SEI (i.e., 6 kJ/L here vs 12 kJ/L by both Chun et al. [15] and Sun et al. [16]).Notably, our energy costs are considerably lower than the earlier MW plasma DRM reports of Zhang et al. [14] consuming 6.5 eV/molecule for an SEI of 36 kJ/L.
Comparing with plasma-catalytic approaches our energy cost is considerably lower than Cho et al. [18] consuming 93-343 eV/molecule for an SEI of 900 kJ/L, but higher than Chun et al. [17] consuming just 2.0 eV/molecule for an SEI of 6 kJ/L.Indeed, in this context, the incorporation of a downstream catalyst is an intriguing prospect to achieve better performance, however, as noted by Chun et al. [17] the deactivation of catalyst by carbon deposition remains a key hurdle for the viability of this approach, further motivating our study of carbon particle formations.
Snoeckx and Bogaerts [2] provided in 2017 an extensive literature overview of the state-of-the-art in the field of plasma-based DRM.They illustrated the energy cost vs.total conversion for DRM in many plasma reactors reported in the literature until that point.It should be noted that the effects of gas expansion are not described in detail in many papers, and this can lead to largely overestimated values towards conversion if not properly considered [83] (see section Experimental details and analysis above).Nevertheless, this was already corrected for in the review by Snoeckx and Bogaerts.Recently, our group has updated this overview with additional data points [7].Here we again add our best metrics for this work, as shown in Fig. 3.Note that the y-axis is reversed, with the lowest (= best) values of the energy cost shown at the top of the   [2] and updated based on more recent literature by Wanten et al. [7].Note that the y-axis is reversed (and in log scale).The efficiency target is defined as the energy cost which should be reached to be competitive with classical DRM and emerging technologies for producing syngas (dashed line).
Our own best result of 2.8 ± 0.1 eV (at 30% CH 4 , 10 L/min, 1 kW) is added as an orange star.
axis.The efficiency target defined by Snoeckx and Bogaerts [2] reasoned that any approach achieving an energy cost under 4.27 eV would be competitive against classical DRM and competing alternatives (shown as a dashed line in Fig. 3).Notably, this target is based on the 60% energy efficiency target, defined by Snoeckx and Bogaerts for plasma technology to be competitive with the classical DRM process or other emerging technologies, for the general stoichiometric DRM process, yielding 4.27 eV per molecule [2].Note however, that besides the energy cost, a major criterion for plasma-based DRM to become competitive with the classical DRM is how to avoid carbon deposition.Hence, in this work, we also focus on the carbon formation.Our performance is indeed clearly below this target, as all our energy cost values are between 2.8 and 3.0 eV/molecule, in the entire range of CO 2 /CH 4 mixtures investigated (i.e., our data points are all above this dashed line).Evidently from Fig. 3, there are many reactor configurations, especially the dielectric barrier discharge (DBD) reactors, which are still far above this threshold.Some of the best performing DBD reactors for DRM include reports by Wang et al. reaching an experimental conversion of 66% [90] and Snoeckx et al. reaching an energy cost at 18 eV/molecule [91].There are several non-MW reactor types shown in Fig. 3 which do achieve promising lower energy costs.For instance, Cleiren et al. [52] achieved an energy cost of 2.5 eV/molecule for gas mixtures of CO 2 and CH 4 for a gliding arc (GA) type reactor (so-called gliding arc plasmatron (GAP) [92]), but that work only considered lower CH 4 fractions (i.e., < 30% CH 4 ).Liu et al. [62] studied an AC pulsed 'tornado' type GA reactor and reported energy costs as low as 1.6 eV/molecule for higher CH 4 fractions, i.e., a 2/3 CH 4 /CO 2 ratio.Li et al. [53] combined the GA approach with a catalyst bed for DRM, achieving an energy cost < 2 eV/molecule.Small volume reactor types that also incorporate metallic electrodes, such as spark reactors [93] and atmospheric pressure glow discharges [7,55] also achieved low energy costs.Notably, these reactors are all electrode-dependent designs, a characteristic that our MW approach avoids.This electrodeless characteristic gives our approach a distinct advantage with potential benefits in terms of the longevity of operation (i.e., avoiding electrode erosion [94]) and energy costs improvements for conversion by avoiding energy loss to the reactor walls.

Product selectivity
In Fig. 4i)-iii) we present the O-, C-and H-based selectivity towards the conversion products for the different inlet mixtures (see Eqs. E12-E18 above).Syngas (i.e., CO and H 2 ) is the primary destination for O, C, and H atoms. Fig. 4 shows that the selectivity for syngas production improves with increasing CH 4 fraction, as the H 2 O mole fractions in the exhaust mixture gradually drop.
In Fig. 4i) we observe that at 70/30 CO 2 /CH 4 , the O-atom selectivity towards CO and H 2 O is S O,CO = 0.72 ± 0.05 and S O,H2O = 0.20 ± 0.05.As H 2 O levels drop with increasing CH 4 fraction, we observe a slightly lower selectivity towards water formation and slightly higher selectivity toward CO production, with values of S O,CO = 0.85 ± 0.10 and S O,H2O = 0.14 ± 0.11 measured at 55/45 CO 2 /CH 4 .Note that the values are quite close, given the large error bars.Across this range, the cumulative selectivity is ~1 (considering the relatively large error) with values of 0.92 ± 0.07 at 70/30 CO 2 /CH 4 , 0.98 ± 0.13 at 60/40 CO 2 /CH 4 and 0.99 ± 0.15 at 55/45 CO 2 /CH 4 .This is logical, given the way the H 2 O mole fraction was estimated; it implies that the O-atoms are distributed primarily between CO and H 2 O, except for the traces of oxygenates that are produced, considered quasi negligible, as previously discussed.

Carbon formation
The production of small quantities of solid carbon, observed in the plasma region and coated on the quartz tube downstream, especially for the higher CH 4 fractions studied (see Fig. 1 ii and iii), has a pronounced impact on the plasma stability (i.e., dictating a minimum SEI for sustainment) and has consequences for downstream utilisation of the product stream (e.g., coking of catalyst bed, in case the latter would be added, in so-called post-plasma catalysis).Further, the carbon formation will eventually require downstream separation, and it is, therefore, very important to understand the size of the particulates and their basic morphology.
SEM and EDX is employed to determine the particle size distribution, purity and the basic structure and morphology of the produced carbon.The form factor of prepared carbon samples consists of nm-sized primary particles (or "nodules") arranged into sub-micron aggregates, as shown in Fig. 5i) [95].The sampled material is found to be metal-free, as illustrated by the EDX spectrum in Fig. 5 ii).The main peak in the spectrum is coming from C, with smaller peaks attributed to O and Al from the support, Si and Al from the detector and microscope hardware and Cl from the solvent used.Indeed, this is expected given the 'electrodeless' characteristics of our surface MW reactor design (i.e., the plasma does not contact any metal surface).Manual SEM image analysis estimates a mean primary particle size of the smallest entities in the collected aggregates of ~21.0 ± 5.9 nm (see Fig. 5 iii).
Raman spectroscopy is an efficient and information-rich analytical technique when carbon-based materials are under investigation.In short, the most important bands are the D and G bands, respectively centred at around 1330 cm -1 and 1580 cm -1 , as depicted in Fig. 6i).The D band in Raman spectroscopy corresponds to the breathing modes of carbon atoms in disordered or defect-induced structures, and it is associated with structural defects.At the same time, the G maximum in a Raman spectrum of a carbon-based material corresponds to the in-plane vibrations of carbon atoms and is associated with C-C bond stretching.The similar intensity for the D and G bands could indicate the presence of amorphous carbon, disordered graphite, or a highly defective carbon structure.Moreover, the absence of significant intensity difference between the D and G peaks suggests a lack of long-range order or a higher degree of structural disorder in the carbon black obtained material.The very weak 2D band also suggests the formation of amorphous or disordered carbon material at 2650-2700 cm -1 , characteristic of crystalline carbon materials.In many cases of amorphous or disordered carbon materials, the 2D band may be entirely absent or appear as a very weak band [96][97][98].
Thermogravimetric analysis of the residual carbon shows two distinct mass weight loss regions, as depicted in Fig. 6 ii.The first one, below 200 • C, may be correlated with the presence of limited volatile molecules (roughly 2%), most likely water.The second and most important one describes the thermal decomposition and oxidation processes of the carbon material (approx.93%).The temperature interval of this consistent mass loss was observed between 400 • C and 650 • C, after which around 5% of the sample is left.The remaining carbon is stable until 1000 • C. It is rather difficult to pinpoint the nature of this residual, which can be in the form of crystalline carbon, carbonaceous particles, or other carbon-based structures that are more stable at high temperatures.Considering the relatively lower decomposition temperature of the carbon material (below 650 • C), we can conclude that the carbon is Fig. 5. i) SEM image of the carbon particles.Small nanoparticles are clustered together forming larger agglomerates.ii) Elemental analysis of the carbon particles using EDX.The main peak in the spectrum is coming from C, with smaller peaks attributed to O and Al from the support, Si and Al from the detector and microscope hardware, and Cl from the solvent used.The sampled carbon is found to be free of metal contaminants.iii) Particle size distribution of the carbon particles based on SEM images.The average particle size is 21.0 ± 5.9 nm.amorphous and contains many structural defects [99].Typically, crystalline carbon, e.g., in the form of graphite, is expected to burn at higher temperatures, even above 650 • C, as is confirmed in the TGA profile (see Fig. 6 ii).The amorphous structure of carbon makes it more reactive and susceptible to oxidation or combustion compared to crystalline carbon, which has a more ordered and stable structure.The increased reactivity of amorphous carbon is due to its higher surface area and structural defects, such as dangling bonds and disordered arrangements of carbon atoms [100,101].
In summary, the morphology and particle size distribution as observed by SEM and the largely amorphous structure of the carbon as indicated by the Raman and TGA measurements all support the hypothesis that the plasma-produced carbon particles resemble a 'carbon black' type material [102].
Carbon black is widely used as a filler in rubber and plastics where it promotes mechanical strength, durability and electrical conductivity [103].Recovery/removal of particulates in gas flows can be carried out in a variety of ways, which include flow separation (e.g., cyclones) and electrostatic precipitators [104].Notably, both methods are mature technologies for particulate separation at high temperatures (e.g., combustion flue gases) which could be advantageous to maintain the plasma exhaust thermal energy for downstream utilisation (e.g., heat exchange, catalyst bed).Evidently, the separated carbon black by-product could be a valuable commodity, lowering the overall cost of our MW plasma-based DRM process (e.g., offsetting the cost of carbon removal).The primary particle size is an important characteristic for the commercial attractiveness of carbon black materials, where (generally) smaller (primary) particle sizes are favoured as a filler for improving tensile strength, conductivity, specific surface area [105].The range of particle sizes found here (i.e., 10-40 nm as shown in Fig. 5 iii)), for example, is consistent with a premium grade 'carbon black' used to reinforce rubber in abrasion-resistant automotive tyres [106].

Exhaust temperature
MW plasmas direct a considerable amount of the energy to heat (i.e., so-called warm plasma conditions) in addition to chemical energy used in converting the feed mixture.Utilisation of this 'waste' heat could lead to further improvement in the energy cost of the process.The utilisation of this 'hot exhaust' also strongly motivates investigation of coupling with downstream catalyst beds and heat/energy recovery configurations.Notably, effective strategies for carbon removal or coke-resistant catalysts will be required, with work on-going [107,108].Such a study is beyond the scope of this work, however, to assess the potential for future progress, we recorded gas temperatures at a short distance outside the plasma region.
K-type thermocouples were placed downstream, in the exhaust plume, at ~15 cm from the outlet.The exhaust reaches a steady-state after approximately 15 min operation [79] when data is acquired.At a 70/30 CO 2 /CH 4 inlet mixture, the gas temperature reaches 726 ± 9 K; at 60/40 CO 2 /CH 4 , it reaches 687 ± 5 K; and for 55/45 CO 2 /CH 4 it reaches 677 ± 8 K.This shows a slight downward trend with increasing CH 4 fraction (and the increasing production and heating of solid carbon), with typical (average) values of ~700 K across the range.These conditions are at the lower end of activation temperatures required for thermal DRM catalysts [108].Evidently, considerably higher exhaust temperatures occur at < 15 cm (e.g., observation of the white incandescent carbon particulates characteristic with temperatures > 1500 K), however, to avoid strong (electrical) coupling between the plasma and our thermocouple, we used a 'safe distance' here.Higher SEI, thermal insulation, and greater proximity (< 15 cm) to the plasma would clearly benefit full exploitation of the heated exhaust for future plasma-catalytic reforming.However, this should be balanced against any possible disruptions to the swirling flow, plasma, and carbon production via the inclusion of a nearby catalyst bed.

Conclusion
In this paper, we investigate atmospheric pressure MW plasma conversion of CO 2 and CH 4 (i.e., dry reforming of methane, DRM) for different mixture fractions.This work gives original insight into the role of solid carbon on MW plasma-based DRM, a critical aspect which has not been studied in detail before.Further to this, for the first time, an estimation of the H 2 O by-product levels in the exhaust is carried out via an atom balance analysis in the context of MW plasma-based DRM.
We achieve a total (weighted) conversion of ~50-53%, with a promising energy cost of ~2.8-3.0 eV/molecule (or ~11.1-11.9kJ/L) with a syngas ratio of ~1 achieved at 45% CH 4 .This performance is among the best when compared to the current state-of-the-art for plasma-based DRM for all types of plasma reactors reported in the literature, demonstrating the importance of the MW approach, which has the key benefit of being "electrodeless" (i.e., advantageous in limiting energy loss to, and erosion of, metallic electrodes, which may happen in other reactor designs).
Syngas (i.e., CO and H 2 ) is observed as the primary product of Fig. 6. : i) Raman spectrum of the plasma produced carbon material.Measurements performed using a green 532 nm laser at room temperature and in air atmosphere in the range of 50-4500 cm -1 Raman shift.ii) TGA profile of the plasma produced carbon material including a comparison with graphite powder (Fluka Inc., particle diameter of ≤ 20 µm).The measurement is performed in O 2 with a flow rate of 80 mL/min and temperature increase of 10 • C/min.until 1000 • C was reached.

Fig. 2 .
Fig. 2. i) Total conversion (weighted according to inlet fraction of CO 2 and CH 4 ) along with individual CH 4 and CO 2 conversion (in %) for different CH 4 content in the CO 2 /CH 4 feed mixture, for an inlet flow rate of 10 L/min at 1 kW power.ii) Corresponding energy cost (in kJ/L and eV/molecule) shown in the left axis and syngas ratio (i.e., H 2 /CO ratio) present in the exhaust shown on the right axis.

Fig. 3 .
Fig.3.Overview of energy cost (eV/molecule) vs total conversion (%), for many different plasma reactors, collected originally from literature by Snoeckx and Bogaerts[2] and updated based on more recent literature by Wanten et al.[7].Note that the y-axis is reversed (and in log scale).The efficiency target is defined as the energy cost which should be reached to be competitive with classical DRM and emerging technologies for producing syngas (dashed line).Our own best result of 2.8 ± 0.1 eV (at 30% CH 4 , 10 L/min, 1 kW) is added as an orange star.
ii) the C-based selectivity towards CO and C 2 H 2 for the different inlet mixtures is presented.The C-based selectivity towards CO drops slightly with increasing CH 4 fraction, from S C,CO = 0.91 ± 0.04 at 70/30 CO 2 /CH 4 to S C,CO = 0.85 ± 0.06 at 55/45 CO 2 /CH 4 .For C 2 H 2, the C-based selectivity is low, i.e., S C,C2H2 = 0.03 ± 0.01 at 70/30 CO 2 / CH 4 , and S C,C2H2 = 0.09 ± 0.01 at 55/45 CO 2 /CH 4 .The cumulative selectivity is ~1 (considering the relatively large error) across the range, with values of 0.94 ± 0.04 at 70/30 CO 2 /CH 4 , 0.97 ± 0.06 at 60/40 CO 2 /CH 4 and 0.94 ± 0.06 at 55/45 CO 2 /CH 4 .This indicates that only a small fraction of the C atoms is likely directed towards other C 2 H x species (e.g., C 2 H 4 , C 2 H 6 ) or solid carbon production, which were unfortunately not available for measurement with our diagnostic setup.

Fig. 4 .
Fig. 4. Selectivity of O-, C-and H-based atoms towards the main DRM products (see Eqs. E12-E18) measured for the different CH 4 fractions in the CH 4 /CO 2 inlet mixtures (x-axis) at 10 L/min total inlet mass flow rate and 1 kW power: i) selectivity of O atoms towards CO and H 2 O. ii) selectivity of C atoms towards CO and C 2 H 2 .iii) selectivity of H atoms towards H 2 , H 2 O and C 2 H 2 .

Table 1
Comparison of typical performance metrics for MW plasma-based DRM reports to date, including this work (across different CO 2 :CH 4 inlet mixture ratios).§ inclusion of a catalyst bed.