Introduction

Non-thermal plasma at atmospheric pressure is produced by the partial ionization of a gas by electron impact using high AC or pulsed high voltages in the kV range. The temperature of the electrons in such plasmas is typically 104–105 K, whilst the heavier species such as the ions and gas molecules remain at temperatures close to ambient. This generates radicals, ions and vibrationally and electronically excited species not normally present in thermal equilibrium at room temperature [1, 2].

To date, it is generally perceived that plasma-driven catalytic (plasma catalytic) syntheses are too expensive in energy terms to compete with standard, thermal processes. However, one highly topical potential application where energy efficiency is not a primary driver is the use of electricity from renewable sources to generate chemical products. Matching the production of electricity from renewable energy to demand is very difficult, and often the surplus is simply spilled: the storage of this surplus energy chemically by plasma catalysis is seen as a possible solution to this problem. The exciting prospect that plasma-catalysis may generate different, and potentially novel products, when compared to the conventional thermal process also offers a firm foundation for the development of this pioneering technology.

Plasma catalysis has a number of potential advantages: thus, in principle, almost 97% of the discharge power of such plasmas can be converted into vibrational excitation of the gas feed molecules [3, 4]. Adding a catalyst can speed up the chemical processes in a plasma and/or offer alternative mechanistic pathways with lower activation energies: thus, reactive species are produced and interact with the catalyst at temperatures at which most thermally-driven processes would be slow. Further, vibrationally excited molecules can have markedly different sticking probabilities and different chemistries on the surface of catalysts in contact with plasma [5] and it should be possible to enhance targeted reactions and suppress others by controlling the number density of electrons and electron and gas temperatures.

As a consequence of the potential of plasma catalysis, many systems have been studied, including [6]: ozone generation [7], the conversion of CH4 to H2 and C2–C4 derivatives [8], toluene to phenol and cresols [9], CO2 and H2O to syngas and synfuels [10], the treatment of Volatile Organic Compounds (VOCs) in waste streams [11], decontamination and disinfection [12], the deep desulfurization of diesel fuels [13] and the treatment of flue gas [14]. Typically, the reactors employed are based on cylinders of alumina or quartz with one high voltage electrode wrapped around the outer surface of the cylinder and a metal rod, mounted along the axis of the cylinder, as the second electrode [15, 16]. Catalyst pellets can be placed in the reactors [15]. Of direct relevance to the work described in this paper is the plasma-driven dry reforming of CO2 and CH4 [17]. In general, CO2 and CH4 are employed without a third feed gas, and the products observed are syngas (CO + H2) with small amounts of C2H2, C2H4 etc. and very small amounts of oxygenated compounds such as ketones and alcohols.

Whilst plasma driven systems find very wide (and commercially-viable) application in industries including waste gas treatment (electrostatic precipitation [18]), ozone generation (primarily for large scale water treatment [19]) and surface treatment/etching [20], these do not involve plasma/catalyst interactions, and this is where both the challenge and opportunity lies. In essence, the brake on the development of plasma catalytic technology has been the limited understanding of the chemistry taking place in plasma and the absence of knowledge and understanding of plasma/solid (i.e. dielectric and/or catalyst) interactions (so much so that the US Government regards the interactions of plasmas with solid surfaces as one of the six critical challenges that define the plasma research frontier [21]).

The inadequate understanding of plasma chemistry probably lies in the lack of analytical studies able to provide molecular information: thus UV Vis absorption and emission spectroscopies have been employed for a number of years in the study of plasma, as may be expected given the inevitable emission of light from plasma. Such studies give useful information on excited species (see, for example, [22,23,24,25,26]). However, as an analogy, the field of electrochemistry underwent a paradigm shift in the 1980s with the advent of, in particular, in situ FTIR spectroscopy with its ability to provide molecular information and hence determine the identity of adsorbed and solution intermediates and products [27, 28]. It is generally recognised that emission spectroscopy and the standard analytical techniques currently employed in NTP research need to be complemented by methods able to provide molecular information such as FTIR spectroscopy [1]. Thus, the field of NTP plasma catalysis is at a similar stage of development as electrochemistry was in the mid-1980s: whilst the plasma/catalyst surface has not been investigated extensively with in situ FTIR, such studies have started to appear. There are a number of studies on the downstream analysis of the exhaust from NTPs, see for example [29], but actual studies of the plasma glow with IR spectroscopy [30] or of the catalyst surface in contact with a plasma is only a recent phenomenon. Thus, the quality of the information that can be obtained from in situ FTIR is exemplified by the work of Li et al. on the deposition of Si from hexamethyldisiloxane plasma [31], Rivallan and colleagues on the conversion of IPA at Al2O3 [32], Stere et al. [33] on the hydrocarbon-assisted NOx removal from simulated diesel fuel over silver-based catalysts, Rodriguez et al. [34] on the conversion of IPA at Al2O3, CeO2 and TiO2 and Jia and Rousseau on the plasma-assisted reaction of acetone at CeO2 [35]. These studies employed the Diffuse Reflectance approach [33, 34], transmission through the catalyst as a wafer [32, 35] or passed the IR beam through the plasma glow above the solid surface [30]. Rivallan et al. [36] employed step scan FTIR spectroscopy to study the gas phase reduction of CO2 in a tube reactor (i.e. they did not study the catalyst/plasma interface): however, they saw no reaction products or intermediates just the loss of CO2, although they did achieve a time resolution of ca. 400 μs.

Recently, we published in situ FTIR studies comparing the plasma- and thermally-driven reaction of CO2 at Macor, the first such direct comparison of plasma and thermal chemistries [37]. We showed that, in both the thermal and plasma experiments, rotational excitation of the CO2 (CO2*) took place in the plasma glow. In the plasma system, rotationally-excited CO (CO*) was produced by the reduction of the CO2, with conversion up to 9% and energy efficiencies of 1–2%, in line with the literature. In contrast, no reaction of CO2 was observed in the thermal experiments.

This paper extends the initial work to the in situ FTIR investigation of the dry reforming of CO2 and CH4 at Macor. In situ FTIR spectra of both the plasma and of the catalyst surface during operation enable detailed analysis of the potential benefits of plasma catalysis and explores further the apparent formation of liquid product films in these processes. The paper is structured as follows: firstly the experimental plasma catalytic reactor and measurement configurations are described; secondly, the initial characterisation of the system after a short run time followed by a detailed study of the changes observed after longer run time is described. Finally the unexpected results are discussed and possible mechanisms are proposed which could pave the way for exploiting the major, untapped potential of plasma catalysis.

Experimental

The Non-thermal Plasma Cells

The transmission NTP cell [37], see Fig. S1(a), was designed and fabricated in–house and was made from a 15 cm long, cylindrical polytetrafluoroethylene (PTFE) tube with an outer diameter of 6.3 cm and an inner diameter of 2.5 cm. The high voltage electrodes were in the form of two plungers, sealing against the inner walls of the PTFE via rubber ‘O’ rings. One of the plungers was hollow down its axes (a 4 mm diameter hole) and the feed gas was delivered via this channel and removed via an outlet in the top of the cell. The electrodes were covered with 5 mm thick Macor caps [Goodfellow Cambridge/quartz (35–50%), magnesium oxide (15–20%), aluminium oxide (15–20%) and fluoride (1–5%)] fitted snugly over each disc electrode, one of which had a hole aligned with the central gas delivery channel. Two infrared transparent CaF2 windows (25 mm diameter, 3 mm thick, Crystran) were glued into the PTFE cell. The distance between the windows (path length) was 5.1 cm. The windows were positioned such that their centres aligned with the centre of the gap between the electrodes. The plasma volume was 7.4 cm3 and the residence time at a flow rate of 200 cm3 min−1 was 2.2 s.

The electrodes were connected to a NeonPro lamp transformer, NP100000-30 (Hyrite, China) which provided an output up to 10 kV at a constant frequency of 24 kHz. A voltage controller (Carroll & Meynell) was used to control the input power to the HV transformers. The input power to the plasma was monitored using a N67FU Gadget 13A Power Meter (Maplin, UK); the input powers quoted below were those obtained by subtracting the input power observed with plasma to the reading obtained with the system switched on but with no plasma initiated (4 W). The temperature of the cell was monitored using a thermometer against one of the windows during runs without spectral data collection and found to be ca. 20–25 °C over the duration of the experiments detailed below: however, the observation of hot CO2 and CH4 (see below) suggests that the actual temperature of the gas species in the plasma was higher.

The reflectance NTP cell [37] is shown in Fig. S1(b): the IR beam passed through a 3 mm CaF2 plate window and a 5 mm gap before being incident on the surface of a 3 cm × 3 cm × 0.5 cm Macor plate covered by a Ti mesh (50% open area) electrode at an angle of 24 degrees. The other electrode was a 2.3 cm × 2.2 cm × 0.007 cm stainless steel foil mounted on an alumina plate behind the Macor. The same power supply etc. was employed as for the transmission cell. In contrast to the transmission cell, the plasma was in contact with a titanium mesh electrode as well as Macor. The plasma volume was 2.0 cm3 and the residence time at a flow rate of 30 cm3 min−1 was 4.0 s.

The design of the transmission cell did not allow for cooling of the electrodes apart from the 200 cm3 min−1 flow of the feed gas: the reflectance cell had dedicated cooling channels for nitrogen gas, as shown in Fig. 1b.

Fig. 1
figure 1

In situ FTIR spectra (8 cm−1 resolution, 100 spectra per scan set, and 1 min per spectrum) of the 8.5% CO2 + 13.1% CH4 + 78.4% N2 feed gas in the plasma transmission cell at a total flow rate of 200 cm3 min−1 and a temperature of ca. 25 °C: (i) before initiating plasma and (ii) after 2 min operation at 24 W. In both cases, the reference spectrum was collected using nitrogen gas as the feed and without plasma

Full size image

The Plasma FTIR System

An Agilent FTS7000 FTIR spectrometer with a Deuterated TriGlycine Sulfate (DTGS) detector was employed. The IR beam was passed through the plasma via the two CaF2 windows to the detector via an Amtir-1 filter, 25 mm × 2 mm (Spectra-Tech, USA) to remove visible light.

CO2 (R744/100%) and CH4 (N4.5/100%) were supplied by BOC Industrial gasses, UK. Dinitrogen was obtained from a NITROSource generator. The gasses were mixed as appropriate and the composition controlled using flow meters. The compositions of the gas feeds employed in the experiments described below are summarized in Table 1; the nominal composition intended for the gas feeds for the CO2 + CH4 + N2 experiments was 10% CO2 + 10% CH4 + 80% N2, with three exceptions (5% CO2, Ar instead of N2 and 54% CO2). The input powers of 20 and 28 W correspond to Specific Input Energies (SIEs) of 6.0 and 8.4 kJ dm−3, respectively defined as [38]:

$${\text{SIE}} = {{{\text{Discharge}}\;{\text{power}}\;\left( {\text{kW}} \right)} \mathord{\left/ {\vphantom {{{\text{Discharge}}\;{\text{power}}\;\left( {\text{kW}} \right)} {{\text{Total}}\;{\text{gas}}\;{\text{flow}}\;{\text{rate}}\;\left( {{\text{dm}}^{3} {\text{s}}^{ - 1} } \right)}}} \right. \kern-0pt} {{\text{Total}}\;{\text{gas}}\;{\text{flow}}\;{\text{rate}}\;\left( {{\text{dm}}^{3} {\text{s}}^{ - 1} } \right)}}$$
(1)

These correspond to average electron energies of 3.9–5.5 eV.

Table 1 The feed gas compositions employed in the work reported in this paper, and a summary of the data obtained from the spectra collected after 2 min during the various experiments
Full size table

The cell was first flushed with N2 at a flowrate of 200 cm3 min−1 for 120 min, and a single beam spectrum taken (100 co-added and averaged scans at 8 cm−1 resolution, 60 s per scan set), after which the chosen feed composition was admitted to the cell at a total flow rate of 200 cm3 min−1 and a second single beam spectrum collected, both without plasma. Sample spectra, SS, were then taken as a function of time after the high voltage–power supply was switched on, and at regular intervals thereafter, up to 20 min. By using the first single beam (of the nitrogen gas with no plasma) as the reference spectrum (SR) all of the infrared active species present in the plasma were observed; using the second single beam as the reference spectrum resulted in difference spectra, showing only the changes incurred on initiating the plasma. The flow rate employed with the reflectance cell was 30 cm3 min−1 to allow sufficient residence time.

The Thermal FTIR System

In-situ FTIR experiments as a function of temperature were carried out using a Varian 670-IR spectrometer equipped with a ceramic air-cooled infrared source, a DLaTGS detector and a Specac Environmental Chamber and diffuse reflectance unit [39]. A 12.5 mm diameter, 2 mm thick Macor disc was employed as the catalyst. The Specac reflectance accessory allows IR spectra to be collected under controlled atmosphere conditions from room temperature to 600 °C and pressures from vacuum to 34 atm. The IR beam was incident on the sample in the cell at angles from 20° to 76° with respect to the horizontal plane via a ZnSe window.

Data Manipulation

The spectra obtained in the thermally-driven and plasma-driven experiments are presented as:

$${\text{Absorbance}},\;{\text{A}} = { \log }_{10} \left( {{\text{S}}_{\text{R}} / {\text{S}}_{\text{S}} } \right)$$
(2)

This results in difference spectra in which peaks pointing upwards (i.e. to + absorbance) represent a gain in absorbing species at SS with respect to SR, and peaks pointing down (to − absorbance) represent the loss of absorbing species. The concentrations of the various species observed in the plasma experiments were calculated using the Beer–Lambert law:

$${\text{A}} =\upvarepsilon{\text{cL}}$$
(3)

where ε is the molar decadic extinction coefficient (M−1 cm−1), c concentration (M) and L optical path length (cm).

The conversion of CO2 to CO is:

$$\% {\text{CO}} = 100 \times {{\left[ {\text{CO}} \right]_{\text{glow}} } \mathord{\left/ {\vphantom {{\left[ {\text{CO}} \right]_{\text{glow}} } {\left[ {{\text{CO}}_{2} } \right]_{\text{glow}} }}} \right. \kern-0pt} {\left[ {{\text{CO}}_{2} } \right]_{\text{glow}} }}$$
(4)

The energy efficiency for the production of CO is:

$${{\eta_{\text{CO}} } \mathord{\left/ {\vphantom {{\eta_{\text{CO}} } \% }} \right. \kern-0pt} \% } = 100\Delta {\text{H}}^{0} / {\text{E}}_{\text{p CO}}$$
(5)

where ∆H0 is the enthalpy (279.8 kJ mol−1) of the process:

$${\text{CO}}_{ 2} \to {\text{CO}} + \raise.5ex\hbox{$\scriptstyle 1$}\kern-.1em/ \kern-.15em\lower.25ex\hbox{$\scriptstyle 2$} {\text{O}}_{2}$$
(6)

at 300 K [40]. Ep CO is the energy expended in producing 1 mol of CO and assuming that the IR spectra record an average picture over the 1 min data collection time.

The errors inherent in the calculations involving the integrated areas of the CO2 etc. features are detailed in the previous paper [37].

Results

Figure S2 shows a spectrum taken before the 24 W run of the 8.5% CO2 + 13.1% CH4 + 78.4% N2 feed gas without plasma at a total flow rate of 200 cm3 min−1, referenced to N2 gas at the same flow rate, at 25 °C. The intense feature between 2300 and 2400 cm−1 is due to the P and R bands of the CO2 asymmetric stretch (ν3) fundamental absorption [41, 42] and the features between 3500 and 4000 cm−1 are CO2 combination bands (ν1 + ν3 and 2ν2 + ν3) [43]. It is clear from the figure that the fundamental bands are saturating and hence cannot be employed for any quantitative measurements; hence it was decided to employ the integrated areas of the CO2 combination bands between 3491 and 3769 cm−1 to estimate the CO2 conversion, the absorption coefficient employed was 6.5 × 105 cm mol−1 [37]. Using the thermal FTIR system, it was found that this value remained constant over the temperature range from 25 to 150 °C [37].

The bands with Q branches centred at 3017 and 1302 cm−1 may be attributed to the P, Q and R branches of the ν3 and ν4 fundamentals of CH4, respectively [42]. Using a 1 cm pathlength gas cell [37] and 100% CH4, the extinction coefficients at 3086 and 1346 cm−1 were determined as 3.5 ± 0.4 and 6.4 ± 0.1 M−1 cm−1, respectively. These values were checked using a 0.6 cm pathlength cell.

Figure 1 shows the spectrum collected after 2 min at 24 W and the spectrum collected of the same CO2/CH4/N2 feed gas with no plasma, both using the nitrogen single beam spectrum as reference. Bands due to HCN and CO were observed and are discussed below. After 2 min plasma operation, although the features other than HCN and CO appear to be due to room temperature CO2 and CH4 bands they are all marginally broader and the Q branches of the ν4 and ν3 bands of CH4 are also shifted slightly. This can be seen more clearly in Fig. S3(a) which shows the spectrum in Fig. 1 using the single beam spectrum collected with no plasma but with the same feed gas (8.5% CO2, 13.1% CH4 78.4% N2), i.e. Fig. S3(a) is a difference spectrum. The broadening of the various features in Fig. 1 compared to the room temperature absorptions of the feed gases can be seen in Fig. S3(a) as the gain of “wings” either side of the CO2 and CH4 loss features, the much reduced fine structure on the ν4 band and, in the case of the Q branches of the CH4 bands, as significantly reduced absorptions and the bipolar nature of the ν3 Q branch. The ν3 band of CH4 and the asymmetric CO2 stretch were simulated using Spectralcalc and the difference spectra obtained by subtracting the room temperature spectra from those at 396 K are shown in Figs. S3(b) and (c): thus we assign the gain features in Fig. 1 for CO2 (CO2*) [30, 44] and CH4 (CH4*) to those due to higher temperatures in the plasma leading to increased rotational excitation.

There is no evidence of non-thermal vibrational excitation enhancement. This is supported by the experiment carried out using the thermal FTIR system, see below.

Figure 2 shows spectra collected after 20 min running with plasma at 20, 24 and 28 W, all using the spectra collected with the same gas feed but without plasma as the reference spectra. As can be seen, in addition to CO2* and CH4*, other species were clearly produced in both liquid and gas phases: with respect to the former, a pale brown liquid was found to coat the internal walls and windows of the cell and subsequent experiments (see below) showed this to be composed of a number of species. It was decided to obtain a spectrum of the liquid: thus Fig. S4 shows the spectrum collected after 20 min at 24 W in Fig. 2 along with the spectrum collected at the end of the same experiment after the cell had been purged with N2 to remove all gas phase species: the latter spectrum was thus of the liquid film on the CaF2 windows. The spectrum taken after 20 min operation had that taken after 2 min subtracted: the reason for this can be seen in Fig. 2, where the peaks between 2000 and 2250 cm−1 in Fig. 2 are overlying the P and R bands of gas phase CO (see below) which remain essentially unchanged with time, indicating that the CO rapidly attains a steady state concentration: this is also the case for HCN, these observations are discussed further below. By subtracting the spectrum taken after 2 min, the features between 2000 and 2250 cm−1 were clearer, and this approach was adopted below. The IR spectra obtained at an input power of 24 W were typical of those obtained at 20, 22, 26 and 28 W, differing only in the relative intensities of the various features.

Fig. 2
figure 2

In-situ FTIR spectra (8 cm−1 resolution, 100 spectra per scan set, and 1 min per spectrum) collected after 20 min operation with plasma as a function of input power. The gas compositions are as shown on the figure at a total flow rate of 200 cm3 min−1 and a temperature of ca. 25 °C; the reference spectrum was collected under the same conditions but with no plasma. Sample spectra were collected every 2 min up to 20 min

Full size image

The 1655 cm−1 band is directly associated with one or more of the components of the liquid film: as can be seen in Fig. 2, the exact position of the band maximum and the peak shape varied from experiment to experiment suggesting contributions from more than one species however, for clarity, it is referred to as the 1655 cm−1 band in the discussion below. As may be seen from Fig. S4, as well as the liquid film (represented by the 1655 cm−1 band), gaseous products were observed, one of which showed a strong absorption at 2152 cm−1 which consistently decreased in intensity when the cell was sparged with N2. We attribute this absorption to ketene [45, 46], a carbon chain oxide; we also attribute the band at 2210 cm−1 in Figs. 2 and S4 to another carbon chain oxide, C5O2 [47,48,49] which is a liquid at room temperature, hence its presence in the deposit [50]. This is an important observation: the production of neither ketene nor C5O2 has been observed previously in NTP experiments irrespective of the nature or composition of the catalyst or dielectric, and this was achieved by using a catalyst comprised of earth-abundant elements. Further, ketenes and their dimers are important reactants that find use in the production of a wide range of chemicals [51,52,53] in a diverse range of industries including pigments, pharmaceuticals and agrochemicals and as intermediates for the paper industry. The 2152 and 2210 cm−1 bands are in a distinctive spectral region that is populated by relatively few functional groups: in addition, their intensities do not, in general, track those of the other bands in Figs. 2 and S4, lending support to their identification as chain oxides. Further, ketene has been observed in a matrix FTIR study of products of benzene transformations in a pulsed glow discharge at low pressure in highly diluted mixtures of benzene with argon in the presence of small amounts of oxygen [54]. As an interesting aside, the observation of the production of ketene and C5O2 in a non-thermal plasma catalysed by common elements has direct relevance to the origin of life: e.g. C5O2 is a powerful tracer of the temperature history of formerly carbon monoxide rich ices in molecular clouds and star-forming regions [55] and ketene is one of the Complex Organic Molecules (COMs) formed in prestellar cores [56].

It is interesting to note from Fig. 2 that the relative yield of ketene and C5O2 depends strongly on the input power, whilst the production of the liquid film does not appear to be so dependent; this suggests one method of potentially controlling the product distribution from the plasma process.

The Spectra Collected After 2 min

Figure 3 shows the spectra collected after 2 min of plasma operation at 28 W in the experiments shown in Fig. 2 and analogous experiments carried out at the same input powers with varying gas feed compositions. The bands at 3334 and 3286 cm−1 in Fig. 3 were attributed to the P and R branches of HCN based on the work of Choi and Barker [57]; the paper also allows an estimate of the extinction coefficient of the 3334 cm−1 band of 2.8 M−1 cm−1.

Fig. 3
figure 3

The spectrum collected after 2 min of plasma operation at 28 W in Fig. 2, along with analogous experiments using gas feeds of various composition, as shown on the figure. The insets show the spectral regions where formate and ketene absorb in the experiment carried out using 5.0% CO2 + 12.7% CH4 + 82.3% N2

Full size image

The features at 2167 and 2116 cm−1 in Fig. 3 may be attributed to CO [37, 42, 58]: the frequency of the former, in some cases, shifted by the presence of an underlying absorption, as shown by the inset of the spectrum from the experiment using 5.0% CO2 + 12.7% N2 + 82.3% N2 in the figure. Fig. S5 shows the CO region of the spectrum collected at 2 min and 24 W in the experiment in Fig. 3, as well as that of a spectrum collected after 2 min operation at 24 W in 13.4% CO2 86.6% N2, the latter was reduced by a factor of 1.3 to render the 2116 cm−1 bands the same intensity. The figure also shows the spectrum resulting from the subtraction of the CO2 + N2 spectrum from that of CO2 + CH4 + N2: clearly, the CO bands are of the same intensities and frequencies, with the difference due to the presence of the 2152 cm−1 ketene feature. Using the 1 cm pathlength cell, the integrated absorption coefficient of the CO band between 2002 and 2225 cm−1 was found to be 7.0 × 105 cm mol−1 in agreement with the work of Bolis and et al. [59].

Figure S6 which shows the spectra collected after 2 min in experiments utilising the feed gas compositions and input powers shown in rows 1a, b, 3a, b and 5a, b of Table 1 (i.e. nominally 10% CO2 + 10%CH4 + 80%N2 at 20, 24 and 28 W) and that taken after 2 min during the experiment using 5.0% CO2 + 12.7% CH4 + 82.3%N2 as feed gas (row 16), with the analogous spectra obtained using CO2 + N2, scaled and subtracted to annul the CO absorptions in each spectrum; the intensities of the ketene band so obtained are presented in the last column of Table 1 which summarizes the data obtained from the experiments in Fig. 3 along with data obtained in analogous experiments carried out as blanks (omitting CO2 or CH4), replacing N2 with Ar and to probe the effect of input power, etc. The spectra from the blank experiments are not included in this work for brevity: the experiments using CO2 alone as the feed gas have been fully described in the previous paper [37], and those using CH4 will be described in detail in later publications. It is very clear from the table that the production of ketene (and indeed the liquid products seen at longer times, see below) requires all three gases. Further, at fixed composition, the gain of ketene increases with input power. This may be due to the formation of the methyl radical from CH4 as the first intermediate in the production of ketene (see Scheme 1): CH3 is known to be an important intermediate for example in the formation of acetic acid from CO2 and CH4 [60].

Scheme 1
scheme 1

Possible mechanism for the formation of ketene at Macor

Full size image

The second inset on Fig. 3 shows the spectral range between 1640 and 1900 cm−1 of the experiment using 8.3% CO2 + 13.5% CH4 + 78.2% Ar: the feature may be attributed to the P, Q and R branches of the ν2 fundamental absorption of formaldehyde [42, 61]. The integrated extinction coefficient of formaldehyde over the range 1660–1820 cm−1 is ca. 1.25 × 10−17 cm mol−1 [62] and this was employed to calculate the concentration of formaldehyde in the plasma glow region.

It is clear from Fig. 3 and Table 1 that there is some conversion of CO2 and CH4 to CO, HCN and formaldehyde after 2 min, but that the major fraction of the reactant gases remain unreacted and present at an increased gas kinetic temperature with broadening due to higher rotational excitation as a result of the higher gas temperature: this is a reasonable assumption for atmospheric pressure. In the absence and presence of CH4, the conversion of CO2 to CO (rows 1–5 and 11–15 in Table 1) appears to be largely independent of input power, which is unexpected as the mean electron energy increases with power, and hence more conversion would be expected. The conversions in CO2/N2 are comparable to those in CO2/CH4/N2 suggesting that CH4 played no part in the reduction of CO2 to CO. This is because the main process is a single step electron dissociation of CO2.

Increasing the dilution of the CO2 by N2 and N2 + CH4 (rows 16, 5a, b, 15 and 19) reduced the conversion of CO2 to CO by a factor of 4: i.e. from 15.5% at 5.0% CO2 to 4% at 54.0% CO2. Xu et al. [63] interpret such behaviour in terms of an increase in the number density of higher energy electrons in CO2 non-thermal plasma on dilution with N2 or Ar. The mean electron energy per molecule ranges from 3.9 to 5.5 eV at the lowest and highest input energies employed by us: given that the threshold energies for the excitation and ionization of CO2, N2, CH4 and Ar are all significantly above these values [64,65,66,67], see Table 2, this suggests that electrons in the high energy tail of the distribution must play a significant role.

Table 2 Electron impact excitation and ionization threshold energies [63,64,65, 67]
Full size table

When N2 was replaced by Ar, the input power had little or no effect (rows 17 and 18), neither did changing the feedgas at 20 W. However, at 28 W, the conversion to CO was reduced by ca. 36%. The latter effect is unsurprising as the ionization energy of Ar is high, 15.76 eV [68] and it is generally accepted that, at the high concentrations of N2 employed (70–90%), metastable N2 states will be primarily responsible for the conversion of both CO2 and CH4 [64, 68]:

$${\text{N}}_{2} \left({\text{A}}^{3} \sum_{\text{u}}^{ + } \right) \, + {\text{CO}}_{2} \to {\text{CO}} + {\text{O}} + {\text{N}}_{2}$$
(7)

and the Penning dissociation of CH4:

$${\text{N}}_{2} \left({\text{A}}^{3} \sum_{\text{u}}^{ + } \right) + {\text{CH}}_{4} \to {\text{CH}}_{3} + {\text{H}} + {\text{N}}_{2}$$
(8)

Table 3 shows the calculated mean electron energies for different compositions of CO2/CH4/N2/Ar assuming a reduced electric field, Td, of 200 and a temperature of 300 K using BOLSIG + [69]. The first excitation threshold of N2 is 7.7 eV (see Table 2) and the electron energy distribution of our system possesses a significant population with energy greater than this and so this mechanism is feasible in our plasma cell.

Table 3 Mean electron energies for different percentage compositions calculated from BOLSIG + [69] at a reduced electric field of 200 Td and a temperature of 300 K
Full size table

In the absence of CO2, 9–12% of the CH4 (rows 6–10) was converted to HCN. Although HCN is not a desired product, the presence of HCN shows that nitrogen fixation has occurred and it is produced in comparable quantities to CO, so its origin is of interest: as would be expected, CH4 is clearly essential for its formation, but a comparison of rows 1–5, 6–10, 16 and 19 strongly suggests that CO2 suppresses this process, due to the production of reactive O atoms favouring the oxidation pathways. The production of HCN increased with input power (see rows 1–5) in the presence of CO2 and CH4, but showed no dependence upon power in the experiments using just CH4 and N2 (rows 6–10). The study of the plasma driven conversion of CH4/N2 mixtures is topical due to their relevance to the atmospheres of Titan and Triton (and the postulate that these are models for the early development of Earth) [70,71,72] and also with respect to natural gas conversion [73]. Horvath et al. [72] have observed HCN at similar concentrations using the same SIE as here: thus the authors employed a cylindrical, co-axial quartz reactor packed with borosilicate beads and supplied with 10%CH4/N2 at SIEs of 1–30 kJ dm−3. At an SIE of ca. 8.4 kJ dm−3 the authors observed 10,500 ppm HCN in the exhaust from the reactor, ca. 5 × 10−4 M and comparable to the 6.0 × 10−4 M observed by us in the plasmaglow at the same SIE. Horvath et al. postulated that the formation of HCN occurs via the reaction of CHx radicals with N atoms: however, the dissociation energy of N2 is 9.756 eV [74] and the authors observed HCN at SIE as low as 1 kJ dm−3, equivalent to an average electron energy per molecule of 0.25 eV for a uniform electric field. However, in the packed bed plasma configuration enhanced electric fields are likely to arise between the beads leading to higher electron energies [60]. The high energy part of the electron energy distribution function is thus likely to be able to directly dissociate N2.

The gas phase production of HCN in the plasma region is most likely produced by the following mechanism [72]. the initial step is electron impact reaction with CH4 which requires electron energies greater than 10 eV. The CHx radicals can then react further with N to produce HCN:

$$\begin{array}{*{20}l} {{\text{CH}}_{4} + {\text{e}}^{ - } \to {\text{CH}}_{3} + {\text{H}}} \hfill \\ {{\text{CH}}_{4} + {\text{e}}^{ - } \to {\text{CH}}_{2} + {\text{H}}_{2} } \hfill \\ \end{array}$$

Followed by:

$$\begin{aligned} {\text{CH}}_{3} + {\text{N}} \to {\text{HCN}} + {\text{H}}_{2} \hfill \\ {\text{CH}}_{3} + {\text{N}} \to {\text{H}}_{2} {\text{CN}} + {\text{H}} \hfill \\ {\text{CH}}_{2} + {\text{N}} \to {\text{HCN}} + {\text{H}} \hfill \\ {\text{CH}}_{2} + {\text{N}} \to {\text{CN}} + {\text{H}}_{2} \hfill \\ {\text{H}}_{2} {\text{CN}} + {\text{N}} \to {\text{HCN}} + {\text{NH}} \hfill \\ {\text{CN}} + {\text{CH}}_{4} \to {\text{HCN}} + {\text{CH}}_{3} \hfill \\ {\text{H}}_{2} {\text{CN}} + {\text{H}} \to {\text{HCN}} + {\text{H}}_{2} \hfill \\ \end{aligned}$$

In contrast, Snoeckx et al. [73] in their cylindrical reactor without packing observed HCN at concentrations an order of magnitude lower than ourselves and Horvath: e.g. at 6 kJ dm−3 using N2 concentrations from 1 to 87% Snoeckx et al. observed HCN at concentrations from 4.0 × 10−7 M to 1.7 × 10−5M, respectively. The authors attributed the very low HCN production to the fact that the SIE was lower than required to generate N2+ which the authors and others have postulated ([73] and refs therein) are the intermediates in the formation of HCN [72], e.g.:

$${\text{N}}_{2} \left({\text{X}}^{1} \sum_{\text{g}}^{ + } \right) + {\text{e}}^{ - } \left( {15.63\;{\text{eV}}} \right) \to {\text{N}}_{2}^{ + } \left({\text{X}}^{2} \sum_{\text{g}}^{ + } \right) + 2{\text{e}}^{ - }$$
(9)

However, it is not clear from the cited papers how N2 is an intermediate in the formation of HCN.

As stated above, in our case, there was no clear relationship between the concentration of HCN in the plasma glow and the SIE: this and the lower dissociation energy of N2 favours the mechanism proposed by Horvath et al.

Interestingly, when N2 was replaced by Ar (Table 1, rows 17 and 18), a new absorption appeared with branches at 3315 and 3263 cm−1 which we attribute to acetylene [42], see Fig. S7.

In general, ηCO was ca.1–3%, which is in broad agreement with the literature where efficiencies are generally quoted as ≤ 10% [65]: in addition, the cell was designed to allow spectroscopic monitoring of the plasma, rather than to maximize conversion efficiencies.

We estimate that bands due to O3, NO2 and N2O at 1051, 1628 and 2237 cm−1 [44] having absorbances of 0.02 would be easily discerned in the spectra in Fig. 2: using the reported extinction coefficients for these species [44], we estimate therefore that if they are present it is at concentrations < ca. 10−5M.

The Spectra Collected at Later Times

In order to maximize the peak intensities observed, the spectra collected during the experiment carried out at 28 W were selected for further study. In addition, given that the HCN and CO bands in general did not change with time in any of the experiments, the spectrum collected after 2 min at 28 W was subtracted from the spectra collected at later times, and the results are shown in Fig. 4. The loss and gain features in the CH4 and CO2 regions of the figure reflect the increased gas kinetic temperature of these species with time: in other words, the loss features are due to the less rotationally excited species. The HCN and CO bands are clearly absent, showing that these species had reached steady state concentrations in the plasma by 2 min. Plots of the various features in Fig. 4 are shown in Figs. 5a and b, and the analogous plots for the experiments carried out at the other input powers, are shown in Figs. S8(a)–(d). The absorbance at 2295 cm−1 on the wing of the CO2* feature was taken as an indication of the increasing rotational excitation with temperature. Figure 5a shows plots of the band intensities as a function of operation time and Fig. 5b the corresponding plots of the intensities normalised to their maximum values. The plots of CO intensity at 2116 cm−1 were calculated with respect to the reference spectrum taken with no input power, all other features were calculated from the difference spectra, with the spectrum collected at 2 min subtracted: the plots of the normalised intensities of the CO band can also be taken as representative of HCN.

Fig. 4
figure 4

In-situ FTIR spectra (8 cm−1 resolution, 100 spectra per scan set, and 1 min per spectrum) collected as a function of time at a fixed input power of 28 W. The gas composition was 12.3% CO2 + 14.4% CH4 + 73.3% N2 at a total flow rate of 200 cm3 min−1 and a temperature of ca. 25 °C; the reference spectrum was collected under the same conditions but with no plasma. Spectra were collected every 2 min up to 20 min. The spectrum collected at 2 min was subtracted from those taken at longer times, see text for details

Full size image
Fig. 5
figure 5

a Plots of the intensities of the various features in Fig. 4 as a function of time. b The plots in a normalised to their maximum values

Full size image

It is clear from Figs. 4, 5a, and S8(a)–(d) that the intensities of the various features observed show that the species responsible show three types of behaviour: HCN and CO rapidly attain a steady state irrespective of the formation of the liquid film, suggesting these take place entirely in the plasma region. The CO2, CO2* and 2152 cm−1 ketene bands show a rapid initial increase which tails away at longer times, but their concentrations in the plasma continue to grow: in contrast, the 2210 cm−1 C5O2, 1655 and 3303 cm−1 bands take longer to develop and then increase fairly steadily. This suggests that the three types of behaviour are due to species produced or consumed by three distinct processes although links between them may also be present. From Fig. 4, it appears that the broad gain feature between 2500 and 3500 cm−1, which is distorted somewhat by ν3 CH4* absorptions, is due to the same species as the 1655 cm−1 feature and this was confirmed by plotting the intensities of the two features against each other, see Figs. S9(a)–(c) for the experiments conducted at input powers of 20–28 W (only the data for 20, 24 and 28 W are shown for clarity). The same procedure confirmed that the 1512 cm−1 band was also due to this species. A clue to the identity of this species was provided by the lack of any intense bands due to C-H stretches in Figs. 1, 3, 4 etc., along with the broad absorption between 2500 and 3500 cm−1 which appeared to comprise two bands. On the basis of the literature [42, 75, 76], we attribute the broad feature between 2500 and 3500 cm−1 to the N–H stretch of acetamide, the weak 2880 cm−1 band to the asymmetric C–H stretch and the 1655 and 1512 cm−1 bands to the amide I (C=O stretch) and amide II (N–H in plane deformation), respectively. As well as the carbon chain oxides, acetamide is also important to the study of prebiotic life as it is the largest interstellar molecule bearing the critical amide functionality [77].

The 1655 cm−1 feature also has contributions from the in-plane N–H bend, C-N stretch and C–C–N deformation. This is clearly not the only species present in the liquid film which shows a plethora of clearly defined (but as yet unidentified) bands below 2000 cm−1 as well as a significant broad underlying absorption suggesting a number of overlapping bands. Further, the liquid film is a yellowish colour suggesting, for example, conjugated C=C and C=N bonds. The postulate that the 1655 cm−1 band is a composite of a number of absorptions is supported by close inspection of the spectra collected after 4 and 20 min in Fig. 4, where it can be seen that the absorption around 1655 cm−1 in the 4 min spectrum comprises a number of peaks.

Figure 6 shows preliminary data obtained using the reflectance cell shown in Fig. S1(b) during a repeat of the experiment in Fig. 2 experiment in which spectra were collected every 2 min after initiating the plasma at an input power of 24 W. The figure shows the spectrum collected after 20 min plasma operation and the spectrum of the deposit which remained after 10 min flushing with N2. As a result of the shorter pathlength of the reflectance cell (1 cm), the CH4 ν4 bands are much weaker, revealing underlying absorptions due to the components of the liquid film: thus there are additional features at 1300 cm−1 (the 1296 cm−1 band in Fig. 4 perturbed by the loss features due to CH4*) and 1016 cm−1. In contrast to all the spectra observed using nominally the same feed gas composition in the transmission cell, where the 1752 cm−1 feature is relatively weak, this band is comparable in intensity to the feature at 1655 cm−1 in Fig. 6. The spectrum collected using the reflectance cell includes information from both the window and the Macor, in contrast to the transmission cell spectrum which only shows the deposit on the windows of the cell. Further, the reflectance cell has one exposed Ti mesh electrode. This suggests that the 1752 cm−1 feature may be attributed to a species present in a higher amount at the Macor/Ti mesh than on the CaF2 window, and that this may be produced as a result of contact of the plasma with the mesh. That the Macor and/or Ti are contributing to the spectrum in Fig. 6i is shown by the bands at 1150 and 1219 cm−1. As may be seen, both features decreased in intensity as the cell was flushed with N2: however, the latter band ceased to decrease, and changed structure. This was due to the presence of two bands, one at ca. 1228 cm−1 (correlating with the 1234 cm−1 feature in Fig. 4) due to the liquid film and a second, at 1210 cm−1 that increased in intensity in concert with the 1150 cm−1 band. The 1150 and 1210 cm−1 bands are associated with the Ti/Macor surface, as they appear in experiments using pure N2 as the feed gas, see Fig. S10. They were induced by the plasma and relaxed slowly, i.e. over many minutes, when the plasma is turned off. We have been unable to identify the species responsible for these absorptions to date and, whilst interesting, it is not clear that these species participate in the processes observed and hence are not relevant to the aim of this paper. Further work is in progress to elucidate this chemistry, and will be reported in a future publication.

Fig. 6
figure 6

(i) Spectrum of the liquid deposit formed on the CaF2 windows of the transmission following a repeat of the experiment at 24 W in Fig. 2. At the end of the experiment, the cell was flushed thoroughly with N2 and a spectrum collected using the spectrum taken of the cell containing nitrogen prior to the experiment as the reference. (ii) Spectrum of the liquid deposit on the CaF2 window of the reflectance cell following an analogous experiment at 24 W to that depicted in Fig. 2. See text for details

Full size image

Figure 7 compares the spectra obtained after 20 min at an input power of 28 W with the corresponding spectra collected after 2 min subtracted using feed gas compositions selected to investigate the effect of CO2 and N2. It is clear that excess CO2 or replacing N2 with Ar both prevent the production of the chain oxides and the liquid film, despite the presence of methane in both cases. Thus the feed gas composition has a major effect upon the products observed. The relative absorbances of the various hot CO2 and hot CH4 features essentially reflect the initial concentrations of the ground state species in the feed gas.

Fig. 7
figure 7

In-situ FTIR spectra (8 cm−1 resolution, 100 spectra per scan set, and 1 min per spectrum) collected after 20 min operation at input power of 28 W as a function of the gas compositions shown on the figure. In each case, the corresponding spectrum collected after 2 min operation was subtracted. Total flow rate was 200 cm3 min−1 the temperature ca. 25 °C

Full size image

A summary of all the assignments is presented in Table 4.

Table 4 A summary of the assignments of the various features in Figs. 2, 3, and 4. See text for details
Full size table

Thermal Experiments

Figure S11 shows the spectrum collected at 100 °C during an experiment in which the reference spectrum was collected at 25 °C and the temperature increased in steps up to 600 °C, with further spectra collected at each step. The spectra were collected from the Macor disc in 43% CH4 21% CO2 36% N2 in the sealed system (i.e. batch mode). It is clear from the figure the same features are observed in the ν3 and ν4 regions of the CH4 absorption and either side of the fundamental absorption of CO2 as in the plasma experiments, supporting the assignments to CH4* and CO2*. No other products were observed up to 600 °C, clearly showing that catalyst selection for plasma driven systems cannot be based solely on materials that work in thermally-activated processes.

Discussion

Both plasma cells are flow systems i.e. reactants were constantly supplied and products removed: however, the plots in Figs. 4, 5a, b and Figs. S8(a)–(d) show that the concentrations of CO2* and CH4* change relatively little with time, whilst the concentration of ketene in the plasma, and C5O2 in the liquid film increase. Given that the input power and flow rate did not vary during the experiments, this suggests that the rate of formation of these products increased with time, whilst the rates of formation of HCN and CO (and possibly formaldehyde) remained constant. As stated above, this suggests parallel and separate reaction pathways: in addition, the data indicate that some aspect of the system is changing and steadily increasing the rate of formation of ketene etc. A logical postulate would be the increasing build-up of reactive intermediates on the Macor surface: e.g. magnesium carboxylate and magnesium nitride, see Schemes 1 and 2. The higher chain oxide C5O2 could then be produced via the reaction of methane and ketene according to:

$${\text{CH}}_{4} + 2{\text{H}}_{2} {\text{C}} = {\text{C}} = {\text{O}} \to {\text{C}}_{5} {\text{O}}_{2} + 4{\text{H}}_{2}$$
(10)

However, in all the 20 min experiments discussed above, the reactions near to the Macor proceeded and indeed accelerated, as the liquid film formed. In contrast, by the end of a two-hour experiment, using essentially the same feed gas composition as that in Fig. 4 and 28 W input power, the production of all the species observed in Fig. 4 had ceased, presumably as the Macor was completely covered, or covered beyond a maximum thickness of the liquid film: the latter possibility arises due to the fact that in all the experiments employing the nominal gas feed composition to the naked eye both the Macor caps were covered with the liquid film after only 20 min plasma operation. Thus, it does not seem unreasonable to postulate that, at thicknesses lower than some critical maximum, the film facilitates the production of the ketene and C5O2: for example, it has been reported that the penetration depth of electrons into an aqueous solution is ca. 12 nm [78]: in our case, the third reaction zone could be a layer within the liquid droplets defined by the penetration depth of the electrons, analogous in some ways to the catalysis taking place at the boundary between metal particles and oxide substrates [79]. Above the critical maximum thickness, defined presumably by the dielectric constant of the liquid film, the electric field decreases below that necessary to sustain the plasma.

Scheme 2
scheme 2

Possible mechanism for the formation of acetamide

Full size image

Our data thus suggest the presence of at least two reaction zones: one at the Macor/plasma interface and one that is the plasma bulk. A third zone involves the liquid, perhaps at its boundaries. This model is a development of the two-zone model proposed by Kim et al. [80] in which the active species in the plasma such as OH radicals and ground state O atoms occupy a thin layer perhaps 50 μm thick above the catalyst and are available for reaction at the catalyst: above this layer, species produced in the plasma react in the same way as in the absence of catalyst. Our model is supported by an experiment in which isoprene was added to 11.0% CO2 + 15.0% CH4 + 74.0%N2 in an attempt to trap ketene via a cycloaddition reaction [81]: instead, whilst CO2*, CO and HCN were still observed, wholly different liquid products were produced, the spectrum of which was identical to that reported by Scarduelli et al. [82], see Fig. S12, and the isoprene simply polymerized on the Macor caps. Scarduelli et al. employed a conventional NTP tube reactor fed with a 60:40 mixture of CH4 and CO2 (no nitrogen); both electrodes of the reactor were isolated from the plasma by silica glass such that the plasma contacted only silica. Our data and the results of Scarduelli et al. strongly suggest that the polymerization of the isoprene on the Macor essentially masked its catalytic properties and the reaction zone was thus confined to the plasma itself. In other words, the presence of the Macor is essential for the specific liquid film composition obtained in our experiments discussed above, nitrogen is required to activate the Macor and the liquid film then has a direct influence on the production of ketene and C5O2, for example either by providing a liquid reaction zone where reactants/reactive intermediates can be concentrated, or by acting as a catalyst.

The chemical species observed here, namely: CO, HCN, HCHO, CH2CO, C2O5, and acetamide are very much a subset of the product species expected in the dry reforming reaction. H2 is one of the main products but could not be observed by our FTIR detection methods. However it would have been expected to observe hydrocarbons: C2H6, C2H4 and C2H2 which were either obscured in the spectra and/or below our detection limits. The one-dimensional fluid model for a typical DBD plasma reactor for CO2/CH4 by de Bie et al. [16] predicts these species as products but in our case with added N2 the routes to C2 species may be diverted by reaction with N atoms to preferentially form HCN. It is interesting that de Bie et al. also predict ketene (CH2CO) but did not include C2O5 in the calculation. As far as we are aware, this is the first time that it has been observed in dry reforming reactions.

Conclusions

Macor contains the oxides of the earth-abundant elements Mg, Al and Si: despite its relatively simple composition it catalyses the production of a rich variety of products from the non-thermal plasma conversion of CH4 and CO2 in nitrogen. The processes observed include the fixing of nitrogen to both HCN and acetamide, and the production of the chain oxides C5O2 and ketene. CO and formaldehyde are produced, in addition to unreacted but rotationally-excited CO2 and CH4, and a multicomponent liquid film. The Macor is essential to the formation of the liquid film, as is nitrogen gas, and the liquid film, in turn, plays an important role in the production of C5O2 and ketene. The formation of liquid products has been observed by others (see, for example, [16, 83]) but the possible wider significance of this has not hitherto been recognised: for example, by careful modification of the liquid film with added solvent and/or by careful choice of model liquids, it could be possible to ensure that the plasma-induced chemistry takes place at and close to the plasma/liquid interface as modelling of the potential distribution using slabs of dielectrics shows that the discontinuity in the electric field will be highest at this interface. Hence we believe this approach may offer a new dimension to plasma catalytic chemistry, one that helps to transform the technology into a viable option for large scale chemical syntheses through a paradigm shift in plasma catalysis and reactor design.

Finally, it is clear that using Macor as a catalyst in the plasma-driven conversion of CO2 + CH4 + N2 results in radically-different chemical processes to those observed in the analogous thermal experiments, supporting the contention that the two processes demand different, tailor-made catalysts identified using in-depth investigations of mechanism and kinetics.