Ammonia generation in Ns pulse and Ns pulse/RF discharges over a catalytic surface

Plasma-catalytic ammonia synthesis in a ns pulse discharge and a ‘hybrid’ ns pulse/RF discharge in plane-to-plane geometry is studied by Fourier Transform infrared absorption spectroscopy. The data are taken in a preheated H2–N2 mixture, with and without Ni/γ-Al2O3 or Co/γ-Al2O3 catalyst placed in the discharge section. The measurement results are taken using two different approaches. The first is a ‘single-stage’ process, where a ns pulse discharge in the H2–N2 mixture is sustained continuously. In this case, the ammonia yield increases slowly, over a period of tens of minutes. The second is a ‘two-stage’ process, where the catalyst is first activated by the ns pulse discharge sustained in pure nitrogen, and then the activated catalyst is exposed to the H2–N2 flow, with or without the discharge. In this case, a strong overshoot of the NH3 number density at the reactor exit is detected, by over a factor of two compared to the single-stage process. After the initial overshoot, the ammonia yield gradually decreases to the ‘single stage’ value (with the discharge on), or to near zero (with the discharge off). The results demonstrate that the ammonia yield in the plasma-catalytic reactor is controlled by the N atom accumulation on the catalyst surface, which reacts with H atoms thermally dissociated on the catalyst or generated in the plasma. The results also show that the plasma-catalytic ammonia yield is significantly higher compared to that in the ns pulse discharge without the catalyst. The accumulation of H atoms on the catalyst, with their subsequent reactions with N atoms generated in the plasma, is of relatively minor importance at the present conditions. An additional series of measurements was made with a sub-breakdown RF waveform overlapped with the ns pulse discharge train, to enhance the vibrational excitation of nitrogen. The ammonia yield measured with the RF waveform added is approximately 20% higher compared to that at the baseline ns pulse discharge conditions, both with and without the catalyst. This effect is weaker compared to that of the catalyst activation by N atoms. Additional data are necessary to isolate the possible effect of the vibrationally excited N2 molecules on the ammonia synthesis in the plasma catalytic reactions.

Plasma-catalytic ammonia synthesis in a ns pulse discharge and a 'hybrid' ns pulse/RF discharge in plane-to-plane geometry is studied by Fourier Transform infrared absorption spectroscopy. The data are taken in a preheated H 2 -N 2 mixture, with and without Ni/γ-Al 2 O 3 or Co/γ-Al 2 O 3 catalyst placed in the discharge section. The measurement results are taken using two different approaches. The first is a 'single-stage' process, where a ns pulse discharge in the H 2 -N 2 mixture is sustained continuously. In this case, the ammonia yield increases slowly, over a period of tens of minutes. The second is a 'two-stage' process, where the catalyst is first activated by the ns pulse discharge sustained in pure nitrogen, and then the activated catalyst is exposed to the H 2 -N 2 flow, with or without the discharge. In this case, a strong overshoot of the NH 3 number density at the reactor exit is detected, by over a factor of two compared to the single-stage process. After the initial overshoot, the ammonia yield gradually decreases to the 'single stage' value (with the discharge on), or to near zero (with the discharge off). The results demonstrate that the ammonia yield in the plasma-catalytic reactor is controlled by the N atom accumulation on the catalyst surface, which reacts with H atoms thermally dissociated on the catalyst or generated in the plasma. The results also show that the plasma-catalytic ammonia yield is significantly higher compared to that in the ns pulse discharge without the catalyst. The accumulation of H atoms on the catalyst, with their subsequent reactions with N atoms generated in the plasma, is of relatively minor importance at the present conditions. An additional series of measurements was made with a sub-breakdown RF waveform overlapped with the ns pulse discharge train, to enhance the vibrational excitation of nitrogen. The ammonia yield measured with the RF waveform added is approximately 20% higher compared to that at the baseline ns pulse discharge conditions, both with and without the catalyst. This effect is weaker compared to that of the catalyst activation by N atoms. Additional data are necessary to isolate the possible effect of the vibrationally excited N 2 molecules on the ammonia synthesis in the plasma catalytic reactions. * Authors to whom any correspondence should be addressed.
Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.

Introduction
Energy efficient generation of ammonia, an essential product used in the chemical industry, using plasma-assisted catalysis has been the focus of intense experimental and modeling studies over the last few years [1][2][3]. This process has been considered as a potential chemical energy storage, for the use of the distributed electrical power generated using renewable energy sources, such as wind and solar energy [4][5][6]. The main thrust of this work has been on the ammonia synthesis from nitrogen and hydrogen, although the use of nitrogen and water as the reactant mixture has also been analyzed [7]. Two critical issues that need to be overcome in the development and eventual commercialization of this process are the energy efficiency and scalability to high yields, up to ∼150-200 g-NH 3 kWh -1 (∼3-4 eV/molecule-NH 3 ) and ∼10 ton-NH 3 /day (∼100 g-NH 3 s -1 ), considering the efficiency of the Haber-Bosch process as a benchmark [1]. To achieve this goal, the efficiency of the plasma-catalytic process needs to be improved by at least an order of magnitude compared to the highest values demonstrated so far [8]. The scalability of the process also necessitates the plasma operation at atmospheric pressure, e.g. using AC dielectric barrier discharges (DBDs), without the use of a vacuum system.
The recent kinetic modeling studies suggest that the plasma-catalytic generation of ammonia is driven by the surface reactions of N atoms and H atoms accumulated on the catalyst surface (N(s) and H(s)) [9], or by the reactions of H(s) with the NH radicals generated in the plasma [10]. Direct evidence of the accumulation of NH x radicals on the Ni catalyst surface have been obtained recently, by using inelastic neutron scattering [11]. Isolating and quantifying the respective roles of N and H atoms, at well-characterized conditions, would provide further insight into the dominant reaction mechanism. It has also been suggested than vibrational excitation of nitrogen molecules in the plasma may enhance the rate of N 2 dissociation on the catalyst and the accumulation of N atoms on the surface [12]. However, the experimental evidence for this hypothesis is indirect, and is based in part on the optical emission spectra of molecular nitrogen in AC DBD plasmas, which do not provide accurate data on the vibrational temperature of N 2 molecules in the ground electronic state. Also, the analysis of the underlying molecular energy transfer processes [12][13][14] is mostly qualitative and therefore not entirely convincing. Specifically, the quantitative assessment of the effectiveness of the vibrationally excited N 2 molecules is complicated by the difficulty of their measurement, especially in low-pressure plasmas, where employing coherent anti-stokes Raman scattering (CARS) and spontaneous Raman scattering is challenging. The second difficulty is the lack of selectivity in the generation of the excited species, e.g. vibrationally excited nitrogen vs. N atoms, which complicates identifying and isolating the dominant kinetic mechanisms. The interpretation of the experimental results based on the parametric studies may be inconclusive [15,16], or indicate the prevalence of one of several possible factors, such as H atoms recombination on the catalyst [17] or suppression of the reverse plasma chemical reactions [18]. The theoretical analysis, such as based on the density functional theory (DFT) predictions [19] is also less than conclusive, suggesting the likely kinetic pathways without identifying them with confidence.
If confirmed, the dissociation of vibrationally excited N 2 molecules on the catalytic surface may reduce significantly the energy cost of breaking the critical N-N bond. This would improve the energy efficiency of the entire process and lead to higher NH 3 yields, since a large fraction of the input energy in low-temperature N 2 -H 2 plasmas goes to vibrational excitation of molecular nitrogen [20]. However, recent experiments in radio-frequency (RF) atmospheric pressure plasma jets, where the N and H atoms, as well as N 2 (v) and NH 3 molecules, have been detected by molecular beam mass spectroscopy [21], concluded that the role of nitrogen vibrational excitation on the ammonia formation was negligible compared to that of atomic species. This is consistent with the results of the kinetic modeling predictions in a low-pressure RF discharge [9] and DFT analysis [19] but at variance with several previous studies [12][13][14]18]. To isolate the vibrationally enhanced N 2 surface dissociation mechanism conclusively, the ammonia yield in the plasma needs to be measured at the conditions where the vibrational temperature of nitrogen can be varied independently of the other plasma parameters, such as the electron density and the number densities of the atomic species. This capability has been demonstrated recently in the 'hybrid' ns pulse/RF discharge developed in our previous work [22,23], where the N 2 (X 1 Σ g + ,v) vibrational populations in N 2 and N 2 -H 2 plasmas were enhanced by applying a sub-breakdown RF electric field between the high-voltage ns pulses which generate the ionization. The hybrid ns/RF discharge, as well as a conceptually similar ns/DC discharge [24], also exhibit much better stability compared to the self-sustained RF or DC discharges and generate diffuse nonequilibrium plasmas at high pressures (up to 1 atm) and high power densities (up to several hundred W cm −3 ).
The main objective of the present work is to study the plasma-catalytic NH 3 generation in N 2 -H 2 mixtures, using the electrode geometry and discharge excitation waveforms conducive to the generation of stable and diffuse plasmas, at controlled and reproducible conditions. This approach is used to analyze the respective roles of the plasma and the catalyst; identify the dominant kinetic processes involved, such as the species accumulation on the catalyst surface; and isolate the possible effect of N 2 vibrational excitation in a 'hybrid' ns pulse/RF plasma on the ammonia generation.

Experimental
The heated plasma flow reactor used in the present work, shown in figure 1, is a modification of the experimental apparatus employed in our previous work [25], where it was used for N and H atom measurements in the plasma. A rectangular cross section quartz channel (22 mm × 10 mm, 10 cm long), fused to a circular quartz tube 1 inch in diameter at both ends, is used to generate a plasma in a heated flow of nitrogen and hydrogen, in the presence of a catalyst. The length of the tube, with the rectangular cross section in the middle, is 60 cm. The tube is attached to two endpieces, quartz tubes 1 inch in diameter and 8 cm long, via UltraTorr vacuum fittings. Each endpiece has a fused silica optical access window at Brewster's angle at the end. A 10% H 2 -N 2 mixture flows through the cell at the total flow rate of 100 sccm and pressure of 190 Torr. This pressure was selected to prevent the RF breakdown and discharge selfsustaining (which occurred below 150 Torr at the present conditions), and to optimize vibrational excitation by the RF field (P = 100-200 Torr for the reduced RF electric field of 15-30 Td [20]). The leak rate of the cell, delivery lines, and exhaust lines is approximately 5 Torr hr −1 (3 × 10 −5 SLM), which is over four orders of magnitude lower compared to the flow rate. The estimated upper bound O 2 and water vapor impurity in the flow is ∼10 −5 and ∼10 −6 mole fraction, respectively.
The entire assembly is placed into a tube furnace (Thermcraft, Ltd, with a 6 inch diameter, 10 inch long heated section), as shown in figure 1. Two quartz coils on the gas supply side of the channel, with the total length of 2.5 m, are used to preheat the flow of reactants. The Ultratorr fittings with the endpieces, as well as the flow inlet and exit are located outside the furnace and remain at near room temperature due to the low thermal conductivity of quartz. A macor ceramic receptacle, filled with Ni or Co catalyst on alumina ceramic powder (Riogen, Inc. particle diameter range 10-80 µm, Brunauer-Emmett-Teller surface area ∼170 m 2 g −1 , 10% catalyst by mass, total mass of 2.0 g), or with the 'blank' alumina powder without the catalyst, is placed into the plasma, such that the flow of reactants is brought into contact with the catalytic surface. Prior to being placed into the channel, the receptacle was filled with a powder mixed with distilled deionized water and then dried. This is done to keep the powder in the receptacle and avoid its removal by the gas flow during the experiment, especially during pumping to vacuum. Before the experiment, the catalyst was reduced in a 10 sccm flow of a 50% H 2 -50% Ar mixture at T = 773 K over 30 min, followed by a 10 sccm flow of Ar for an additional 30 min, without the plasma. This was done to remove the residual H atoms adsorbed on the catalyst surface, which was verified by the fact that no ammonia was detected during the subsequent catalyst activation process in a nitrogen plasma.
Although in the present arrangement the area of the catalytic surface exposed to the plasma is relatively small, this approach enables a direct comparison between the plasma/catalytic surface/inert surface ammonia generation, in a simple geometry, at controlled and reproducible conditions. The exhaust flow from the cell is sent through an external absorption cell made of glass (1 inch in diameter, absorption path 1.2 m long) and equipped with ZnSe windows, to measure the number density of ammonia generated in the plasma, as shown in figure 1(a). The long absorption path is critical for the detection of ammonia, present in the exhaust flow in relatively low concentrations, of the order of tens of ppm. A calibrated blackbody source (Infrared Systems IR-564), heated to 1273 K, is used to generate the broadband IR emission, directed through the absorption cell and focused into the emission port of an Fourier transform infrared (FTIR) spectrometer (Varian 660-IR), operated at the spectral resolution of 0.5 cm −1 , using a pair of off-center parabolic mirrors.
The plasma in the channel is generated between two parallel rectangular plate stainless steel electrodes 60 mm long and 12 mm wide, mounted to the top and bottom walls of the channel, as shown in figure 1(a). The electrodes are placed into the recesses machined in alumina ceramic plates, held together by ceramic screws. The plasma is sustained by a custombuilt ns pulse generator [26] (peak voltage up to 12 kV, pulse duration ≈ 100 ns) operated continuously at a pulse repetition rate of 1 kHz. Between the pulses, additional energy is coupled to the plasma by a sub-breakdown (1 kV peak voltage) 13.56 MHz waveform produced by RF waveform generator (AG 0113 R, T&C power conversion), operated in a repetitive burst mode with the duty cycle of 10%-40%. The RF voltage was kept low, to prevent additional ionization, dissociation, and electronic excitation by electron impact between the discharge pulses and to isolate the possible effect of vibrational excitation of N 2 and H 2 on the ammonia generation in the plasma. Previous measurements of the number density of a metastable electronic state of molecular nitrogen, N 2 (A 3 Σ u + ), at similar conditions indicate the difference of 20%-30% between ns pulse and ns/RF plasmas [22].
Both the ns pulse train and the RF waveform are applied to the same pair of electrodes external to the channel. This is done to isolate and quantify the effect of the additional vibrational excitation of the reacting mixture on the ammonia generation in the plasma, with and without the catalyst. The RF generator is connected in series with the ns pulse generator (see figure 1), such that each power supply uses its counterpart as the ground, i.e. the ns pulse discharge current flows through the RF generator and vice versa. The choice of a ns pulse is based on its circuit design, which should not be affected by the RF current. The pulse voltage and current waveforms are measured by custom-made, high bandwidth capacitive voltage probes and shunt current probes [27]. Broadband plasma emission images are taken through the side wall of the rectangular section and through one of the Brewster angle windows, using a gated PI-MAX 3 Intensified charge-coupled device (ICCD) camera with a UV lens. The RF voltage and current waveforms are measured by the Tektronix P-6015 high voltage probe and Pearson 2877 current monitor. Due to the slow flow rate of the reactants through the channel and a significant volume of the absorption cell, the estimated cell flush time is rather long, ∼100 s. Because of this, the characteristic time for the ammonia concentration in the cell to reach the steady state, when the discharge is operated without the catalyst, is several minutes. When the receptacle with the catalyst is placed into the channel, the time to reach the steady state increases considerably, by up to an hour. Therefore the steady state operation of the discharge for extended periods of time is critical. This also necessitated the use of a relatively low discharge pulse repetition rate, 1 kHz. Operation at higher pulse repetition rates would result in overheating of the pulse generator. The time interval between taking the absorption spectra in the cell during the operation varied from 1 min early in the run (during the first 5-10 min) to 3-5 min for the runs 60-90 min long. Between the runs, the catalyst was reduced, first by flowing a 50%-50% H 2 -Ar mixture through the heated reactor for 30 min, and then argon for an additional 30 min, with the discharge turned off.

Results and discussion
The ns pulse and RF voltage waveforms used in the present work are shown in figure 2(a). Both waveforms are plotted on a long time scale of 1.5 ms, to better illustrate their timing. In addition, two insets in figure 2(a) show the ns pulse voltage and current, as well as the RF voltage on a shorter time scale of several hundred ns. In figure 2(a), the RF discharge is operated in burst mode, with bursts 100-400 µs long initiated with a delay after the ns discharge pulse, at the burst repetition rate of 1 kHz, same as the ns pulse repetition rate (see figure 2(a)). The two waveforms are separated in time, to minimize their possible distortion, since the two power supplies are not electrically isolated from each other and form a single external circuit, as shown in figure 1. In the present experiments, the delay between the ns pulses and RF bursts is varied from 5 µs to 200 µs, and the RF voltage rise time is approximately 1 µs. During the hybrid plasma experiments, the RF voltage amplitude was maintained at 1 kV, and the forward RF power was 50 W. The reflected power could not be measured due to the strong electromagnetic interference (EMI) interference from the ns pulse discharge. The rectangular bars in figure 2(a) indicate schematically the timing of the ICCD camera gates used to take the ns pulse and RF plasma emission images (the gates shown are not to scale). The instantaneous power and coupled energy waveforms for a ns pulse discharge, measured in a 10% H 2 -N 2 mixture at P = 190 Torr and T = 573 K, are plotted in figure 2. These measurements were repeated when the receptacle was filled with the 'blank' alumina powder and with Ni/alumina catalyst, showing very close and reproducible results. In both cases, the coupled pulse energy was approximately 1.5 mJ pulse -1 . Although the RF discharge current was not measured directly, the RF energy coupled to the decaying plasma generated by the ns pulse discharge was estimated based on the residual electron density and electric field in the plasma predicted by the kinetic model [28], ∼1 mJ over a 400 µs burst, i.e. comparable to the ns pulse energy per pulse.
Single-shot, broadband plasma emission images taken in a ns pulse discharge and in a decaying RF plasma at the conditions of figure 2 are shown in figure 3. Panels (a) and (b) show the single-shot images in a ns pulse discharge with an empty receptacle in the reactor and with a receptacle filled with 10% Ni/γ-Al 2 O 3 catalyst powder, respectively. It can be seen that adding the catalyst powder improves the plasma uniformity. A similar effect of adding a metal catalyst to the alumina powder on the plasma uniformity has been observed previously in AC DBD plasmas [29], and attributed to the enhancement of the surface discharge propagation by the conducting metal particles. The RF plasma images taken at the same conditions, shown in panels (c) and (d), appear completely dark, with no optical emission detected anywhere in the discharge gap, including the sheaths. In our previous work on characterization of a ns pulse/RF hybrid discharge [28], emission from the sheaths of the decaying RF plasma was detected, but at a significantly higher peak applied reduced electric field, (E/N) RF ≈ 40 Td. At the present conditions, the peak reduced electric field is estimated to be (E/N) RF ≈ 20 Td. Note that the camera gate used to take the RF plasma images is over three orders of magnitude longer compared to the gate used for the imaging of the ns pulse discharge plasma, 450 µs vs. 250 ns, respectively. In addition, the ns pulse discharge images were taken at a lower camera gain. This illustrates that the upper bound optical emission intensity from the decaying RF plasma is several orders of magnitude lower compared to that in the ns pulse discharge and indicates that the electron impact excitation of electronic states of N 2 is insignificant at the present conditions. As stated in section 2, the area of the catalytic surface exposed to the plasma in the present experiment is relatively small and limited by the size of a single catalyst powder receptacle. An attempt to enhance the catalytic surface area, by placing a 'blank' or catalyst-activated alumina/cordierite honeycomb monolith (400 square cross section cells per square inch) into the channel, was unsuccessful. In this case, the plasma was generated outside of the honeycomb, most likely due to the surface charge accumulation on the monolith resulting in applied electric field shielding. Figure 4 shows a typical NH 3 FTIR absorption spectrum, compared with best fit synthetic spectrum calculated based on the high resolution transmission (HITRAN) spectroscopic data [30] and the FTIR instrument function. The uncertainty of the ammonia number density inference from the synthetic spectra and the sensitivity limit of the present measurements Time-resolved NH 3 yield data, taken at different operating conditions, are plotted in figure 5. Specifically, figure 5(a) demonstrates that no ammonia was detected in the reactor exhaust at the baseline flow conditions (P = 190 Torr, T = 573 K, 10% H 2 -N 2 mixture), when a receptacle with 10% Ni/γ-Al 2 O 3 catalyst was placed in the reactor but the discharge was off. The same figure plots the ammonia yield measured in an 'empty reactor' (i.e. when the receptable was removed), with a 1 kHz ns pulse discharge turned on. It can be seen that the time for the NH 3 yield to reach the steady-state is approximately 5 min, due to the long absorption cell flush time. When an empty receptacle is placed in the reactor, the plasma volume decreases, but the surface area slightly increases, resulting in a marginally higher ammonia yield compared to that measured in an empty reactor, by less than 10%. These results show that the ammonia generation by the thermal surface catalytic reactions is below the detection limit, and establish the reference ammonia yield at the baseline conditions of the empty reactor. Figure 5(b) compares the effect of using different catalytic surfaces on the ammonia yield with the results obtained in the empty reactor (labeled as 'plasma only'), at the same flow and plasma conditions. It is readily apparent that placing the receptacle filled with alumina powder, with or without the catalyst, (i) increases the time to reach the steady-state ammonia concentration in the external absorption cell, and (ii) increases the ammonia yield, compared to the empty reactor conditions (see figure 5(b)). A longer transient rise of the ammonia concentration, approximately 20 min for the 'blank' alumina powder and over 40 min for the catalyst powder, at the same flow and plasma conditions indicates the effect of the surface reactions and suggest that the species generated in the plasma may accumulate on the packed powder surface. Also, even the use of the nominally 'inert' Al 2 O 3 surface enhances the NH 3 yield in the surface reactions, although replacing it with Co/γ-Al 2 O 3 or Ni/γ-Al 2 O 3 powder enhances the yield further, by up to a factor of 2 when Ni catalyst is used. This suggests that the observed yield enhancement is due to a combination of a larger surface area exposed to the flow (alumina powder in the receptacle vs. reactor walls) and a catalytic effect of Co or Ni. Note that this enhancement is detected at the conditions when the surface area of the packed powder in the receptable is small, and it may well be increased if the discharge geometry is modified to enhance the discharge surface area/volume ratio.
The effect of adding the RF waveform to the 1 kHz ns pulse train, such as shown in figure 2(a), is illustrated in figures 5(c) and (d). These data are taken at two sets of conditions, figure 5(c) in the empty reactor, and figure 5(d) in the reactor with the Ni/γ-Al 2 O 3 catalyst, at the same flow and ns pulse discharge parameters. The RF pulse bursts are applied with a delay relative to ns discharge pulses, ranging from 5 µs, when the residual electron density is still relatively high, n e ≈ 3 × 10 11 cm 3 s −1 according to the kinetic modeling predictions of [28], to 200 µs, when the electron density is significantly lower, n e ≈ 1 × 10 10 cm 3 s −1 . The results indicate that turning the RF voltage on in the empty reactor increases the ammonia yield by up to 15%, for the delay time of 5 µs. This effect is reproducible and has been observed in multiple runs. Increasing the RF delay time to 200 µs, i.e. reducing the residual ionization by over an order of magnitude, reduced the NH 3 yield enhancement to within the experimental uncertainty, about 5 ppm. When the ns pulse repetition rate was increased from 1 kHz to 3 kHz (still without the catalyst receptacle in the reactor), adding a 100 µs long RF burst with a 5 µs delay enhanced the yield by approximately 25% (see figure 5(c)). In this case, a 20% yield enhancement was observed even when the RF burst delay was increased to 200 µs, indicating that the electron density decay between the ns discharge pulses produced at a higher repetition rate becomes slower. This trend is consistent with the effect of associative ionization in collisions of excited metastable nitrogen atoms, N( 2 P) [31], which may accumulate over multiple discharge pulses.
When the catalyst receptacle was placed into the reactor, a 1 kHz ns pulse discharge was operated along for 70 min, to reach a nearly steady-state ammonia yield (see figure 5(d)). Similar to the results plotted in figure 5(b), the yield was higher compared to that in the empty reactor. After the ammonia yield reached the plateau, the RF waveform (400 µs long bursts applied 5-200 µs after the ns pulses) was turned on, and the yield was measured again after several minutes of operation. The largest difference between the yields measured with and without the RF voltage, is approximately 20% (see figure 5(d)), similar to the results obtained in the empty reactor. Again, this effect is reproducible and similar results have been measured in multiple runs.
To understand the underlying kinetic mechanisms involved in the plasma-catalytic ammonia synthesis, we used the results of time-resolved measurements of NH 3 yield at different operating conditions. These results are summarized in figure 6, which shows the data taken in three different set of experiments, while maintaining the same pressure, temperature, and flow rate through the reactor. The first is the plasma-catalytic process in H 2 -N 2 with Ni catalyst, referred to in figure 6 as a Single-stage process. These are the same data as plotted in figure 5(d), taken in a 10% H 2 -N 2 mixture at P = 190 Torr and T = 573 K, excited by a 1 kHz ns pulse discharge in the presence of the Ni/γ-Al 2 O 3 catalyst. As discussed above, the time for the ammonia yield to reach the quasi-steady-state is nearly an hour, an order of magnitude longer compared to the measurements in the empty reactor, approximately 5 min (see figure 5(c)). In the second experiment, referred to as a Twostage, plasma OFF process in figure 6, the Ni/ γ-Al 2 O 3 catalyst in the reactor is first 'activated' by operating a 1 kHz ns pulse discharge in pure N 2 flow for 30 min (resulting in no detectable NH 3 yield, as expected), and then exposed to the 10% H 2 -N 2 flow while the discharge was turned off. In this case, the ammonia yield exhibits a strong overshoot, exceeding the quasi-steady-state value measured in the single-stage process, and then decreases by more than a factor of 6 over the next 30 min (see figure 6). In the third experiment, referred to as a Two-stage, plasma ON process in figure 6, the catalyst is 'activated' for 30 min again by a 1 kHz discharge in pure N 2 , and then exposed to a 10% H 2 -N 2 flow while the discharge is kept on. In this case, the initial NH 3 overshoot when the nitrogen flow is switched to H 2 -N 2 is even higher, after which the yield decreases significantly over the next 40 min and approaches the quasi-steady-state yield of the single-stage process (see figure 6). Finally, when the discharge is turned off at that point, the yield decreases by about a factor of 6 over the next 30 min, similar to the Two-stage, plasma OFF case.
Based on the measurements and kinetic modeling predictions of N and H atoms number densities in a ns pulse discharge in N 2 -H 2 in our previous work [25], as well as the coupled pulse energy in the present work (a factor of 2 lower compared to that in [25].), the number densities of atomic nitrogen and hydrogen are estimated to be ∼10 14 cm −3 and ∼10 15 cm −3 , i.e. ∼30 ppm and ∼300 ppm, respectively. The results of the present NH 3 yield measurements, combined with the estimated concentrations of atomic species, suggest the following interpretation. During the catalyst 'activation' by a nitrogen plasma, N atoms, which are generated by the ns discharge pulses and accumulate in the flow [25], are adsorbed to the catalyst surface. When the discharge is turned off and the surface is exposed to the H 2 -N 2 flow (Two-stage, plasma OFF case), thermal dissociation of H 2 on the heated catalyst surface, which produces adsorbed H atoms, is followed by the surface reaction of N(s) and H(s) resulting in the formation of NH(s), NH 2 (s), and eventually NH 3 , which is desorbed back to the gas phase [9]. This explains the initial overshoot of the ammonia yield observed in this case (see figure 6), as well as its subsequent decay as N atoms accumulated on the catalyst surface during the N 2 plasma activation state are gradually depleted.
Exposing the activated catalytic surface to the H 2 -N 2 flow when the ns pulse discharge is kept on (Two-stage, plasma ON case) provides an additional continuous source of N and H atoms generated in the plasma for adsorption on the catalyst surface. In this case, the NH 3 yield still exhibits an overshoot when the N 2 plasma flow activating the catalyst is switched to H 2 -N 2 , and decays as the N atoms adsorbed during the catalyst activation are depleted (as illustrated in see figure 6). However, in this case the yield decays not to zero but to a quasi-steady-state value, controlled by the N atoms generation/adsorption on the catalyst, balanced by their depletion in the surface reactions with H atoms. The gradual decay of the ammonia yield at these conditions may also be due to the occupation of the surface sites by H atoms generated in the plasma, which limits the number of sites available to N atoms, as discussed in [11]. Note that H atoms are the dominant product of the plasma chemical reactions in a ns pulse discharge in H 2 -N 2 mixtures [25]. Turning the discharge off after the quasi-steady-state conditions are reached leads to the NH 3 yield decay, similar to the one observed without the discharge (Two-stage, plasma OFF case), again due to the depletion of N atoms adsorbed on the catalyst surface. This scenario also explains the very slow rise of the ammonia yield detected in the single-stage process, which is controlled by the rate of accumulation of N and H atoms generated in the plasma on the catalyst surface. As expected, the asymptotic steady-state ammonia yield is approximately the same for both Single-stage and Two-stage, plasma ON processes (see figure 6).
The effect of the catalyst activation is illustrated further in figure 7, which plots the time-resolved NH 3 yield measured at the same conditions (10% H 2 -N 2 mixture, T = 573 K, P = 190 Torr), over the Ni/γ-Al 2 O 3 surface activated by a 1 kHz nitrogen plasma, which is then turned off. The results are plotted for different catalyst activation periods, ranging from 5 to 30 min. These runs are separated by 30 minute periods when pure argon was flowing through the reactor, to minimize the effect of the catalyst initial conditions. It is readily apparent that a longer catalyst activation results in a more pronounced ammonia yield overshoot, demonstrating again that increasing the catalyst exposure time to N 2 plasma (i.e. greater accumulation of N atoms on the surface) is a major factor controlling the ammonia yield.
Nitrogen plasma activation of different packed powder surfaces is compared in figure 8. This figure shows the results of the time-resolved ammonia yield measurements after the 'blank' alumina, 10% Co/γ-Al 2 O 3 , and 10% Ni/γ-Al 2 O 3 Figure 7. Effect of the catalyst activation time on the ammonia yield in a 10% H 2 -N 2 mixture at T = 573 K, P = 190 Torr, and total flow rate of 100 sccm over nickel/alumina catalyst. The activated surface is exposed to the H 2 -N 2 flow without the plasma, after the catalyst activation by a 1 kHz ns pulse discharge in pure N 2 . Each run is separated by a 30 min Ar flow over the catalyst, without the plasma. surfaces were activated by a 1 kHz ns pulse discharge in pure N 2 (each for 30 min), and then exposed to a 10% H 2 -N 2 flow after the plasma was turned off. It can be seen that the ammonia yield over the activated alumina surface remains below the detection limit, while the yields measured over the Co and Ni catalyst surfaces exhibit similar overshoot and decay trends. The yield measured over the Ni catalyst surface is significantly higher, consistent with the rest of the present measurements (e.g. see figure 5(b)). This suggests that N atoms generated in the plasma accumulate on Ni and Co catalyst surfaces more efficiently, compared to the alumina surface. We note that the ammonia yield measured in H 2 -N 2 flows over the plasmaactivated alumina surface, after the discharge was turned off, was always below the detection limit. We conclude that the thermal dissociation of hydrogen on alumina, which is the only source of adsorbed H atoms at these conditions, is very inefficient at the present conditions. Because of this, there are very few adsorbed H atoms available for the surface reactions with N atoms adsorbed during the plasma activation stage, and ammonia is not detected in the reactor exhaust. This conclusion is supported by the additional measurements, when the N 2 plasma activated 'blank' alumina powder surface was exposed to H 2 -Ar mixtures, while the ns pulse discharge was kept on. Figure 9 shows that in this case, ammonia was detected in the external absorption cell, and a significant overshoot was observed when the H 2 fraction in the mixture was increased from 10% to 50%. This demonstrates that (i) N atoms are adsorbed on the 'blank' alumina surface during the plasma activation stage, and that (ii) hydrogen dissociation in the plasma provides the source of H atoms necessary for the surface reaction with N atoms. Thermal dissociation of hydrogen on alumina (in the H 2 -N 2 mixture with the plasma off, see figure 9) appears to be insignificant compared to its plasma  Since ammonia is detected with no nitrogen present in the flow through the reactor, it is generated only in the surface reactions with N atoms adsorbed during the surface activation stage. Also, a kinetic model of a ns pulse discharge in N 2 -H 2 mixtures which includes only the gas-phase processes [25] predicts the ammonia yield several orders of magnitude lower compared to the present measurements, which again indicates the dominant role of the surface reactions.
The effect of thermal dissociation of hydrogen on the Ni/alumina catalyst activated by a 1 kHz ns pulse discharge in nitrogen plasma, is illustrated in figure 10. The ammonia yields measured when the activated catalyst surface is exposed to 10% N 2 -H 2 and 10% H 2 -Ar flows, without the plasma, are close to each other and follow the same trend, indicating that the yield is not affected by nitrogen in the mixture and controlled by H atoms generated by the thermal dissociation on the heated catalyst surface. Turning the discharge on in the H 2 -Ar mixture results in a higher ammonia yield, most likely due to the additional generation of H atoms in the plasma.
Finally, the Ni/alumina catalyst activation by flowing a 10% H 2 -Ar mixture through the reactor, with or without a 1 kHz ns pulse discharge, and its subsequent exposure to a nitrogen flow, again with or without a 1 kHz ns pulse discharge, did not exhibit the detectable effect of H atoms accumulation on the surface on the ammonia synthesis. The ammonia yield measured at these conditions was significantly lower compared to the rest of the measurements, below 50 ppm, and did not show a significant variation with time. We are led to the conclusion that at the present conditions the accumulation of H atoms on the catalyst, with their subsequent reactions with N atoms generated in the plasma, is of relatively minor importance for the ammonia generation.

Summary
In the present work, we used the time-resolved, ex situ FTIR absorption measurements of the ammonia yield in the exhaust of a heated plasma flow reactor to elucidate the kinetic mechanism of ammonia generation by plasma-enhanced surface catalysis. The plasma is generated by a train of high-voltage, ns duration pulses, in a simple geometry which lends itself to the kinetic modeling analysis. The discharge electrodes are external to the reactor. The receptacle with a catalyst on a packed alumina powder (γ-Al 2 O 3 ) is placed directly into the plasma. The measurement results lead to the following conclusions. First, comparison of the NH 3 yield in the empty plasma flow reactor with the yield measured with the 'blank' alumina, Co/γ-Al 2 O 3 , and Ni/γ-Al 2 O 3 catalysts placed in the reactor indicates that the yield is enhanced due to (i) increasing the surface area available for the reactions of adsorbed species (alumina), and (ii) surface catalytic effect (Co, Ni). In all cases, the time for the yield to reach the steady-state value increases, by up to an order of magnitude compared to that in the empty reactor, which is attributed to the slow adsorption of the plasma-generated species on the alumina (catalyst) surface.
Measurements of the ammonia yield after the 'activation' of the catalyst in pure N 2 plasma, with its subsequent exposure to the H 2 -N 2 flow (with or without the plasma) demonstrate conclusively that the plasma-catalytic ammonia synthesis at the present conditions is controlled by the adsorption of N atoms generated in the plasma on the catalyst surface during the activation stage. Ammonia is produced in the surface reactions of the adsorbed N atoms with the adsorbed H atoms, produced by either thermal dissociation of H 2 on the catalyst surface (without the plasma) or by the plasma chemical reactions (in the plasma). This leads to a significant overshoot of the ammonia yield, during the depletion of N atoms accumulated on the catalyst during the activation stage. This result is complementary to that of [11], where NH x species were detected on the catalyst surface by the neutron scattering, and consistent with the kinetic mechanism proposed in [9].
The accumulation of N atoms on the catalyst surface is shown to increase over time (during the catalyst activation in N 2 plasma), and subsequently decays over time (during the exposure of the activated catalyst to the H 2 -N 2 flow without plasma). The accumulation of N atoms on Ni catalyst surface at the present conditions appears to be more significant compared to Co catalyst. Measurements of the ammonia yield over the plasma-activated alumina surface (with and without plasma) demonstrated that in this case the thermal dissociation of H 2 on the surface is ineffective, such that the yield measured without the plasma was below the detection limit. However, when H atoms are generated in the plasma (in H 2 -Ar mixtures without nitrogen), ammonia is detected in the reactor exhaust, demonstrating again that NH 3 is formed in the surface reactions between N atoms adsorbed during the catalyst activation stage and H atoms adsorbed from the H 2 -Ar plasma. The effect of the thermal dissociation of hydrogen atoms on a nickel catalyst activated by a nitrogen plasma is further demonstrated by comparing the ammonia yields measured when the activated catalyst is exposed to N 2 -H 2 and H 2 -Ar flows, without the plasma, which are close to each other. On the contrary, the accumulation of H atoms on the catalyst, with their subsequent reactions with N atoms generated in the plasma, is of relatively minor importance at the present conditions.
A sub-breakdown RF waveform was added to the ns pulse train, to isolate the possible effect of the enhanced vibrational excitation of N 2 on its dissociation on the catalyst surface. The enhancement of the N 2 vibrational temperature in a 'hybrid' N 2 -H 2 plasma has been demonstrated in our previous work [22]. Adding the RF voltage resulted in a moderate increase of the NH 3 yield, measured both in the empty reactor and over Ni catalyst, by approximately 20% compared to the ns pulse discharge operating alone. This increase is consistent and reproducible. It would be tempting to attribute this effect to the reactions of vibrationally excited nitrogen. However, it remains uncertain whether these reactions occur in the gas phase or on the catalyst surface. The ammonia yield increase may also be affected by the enhanced transport of the species generated in the plasma toward the catalyst surface, driven by the RF-induced drift oscillations of the plasma electrons. Comparison of the plasma emission images taken in the ns pulse and ns pulse/RF discharges [22] exhibits better uniformity and greater extent of the ns/RF plasma, possibly due to the enhanced transport.
The most likely reason why the effect of adding the RF field is rather marginal at the present conditions is the low ns pulse/RF burst repetition rate, 1 kHz, used to enable discharge operation over extended periods of time (up to 90 min). This limits the estimated N 2 vibrational temperature to T v ≈ 1200 K, due to the relatively rapid electron density decay after the ionizing pulses (on a time scale of ∼10 µs). This estimate is based on the previous CARS measurements of T v (N 2 ) at the pulse repetition rate of 10 kHz and kinetic modeling predictions [22]. In the future work, additional data will be taken at higher pulse repetition rates, up to 10 kHz, and higher peak RF voltage, which would increase the RF power coupled to the plasma and enhance the vibrational temperature. Catalyst activation by the hybrid ns pulse/RF plasma in pure N 2 , followed by thermal exposure to H 2 , will also be employed. This will further increase the N 2 vibrational temperature in the plasma, due to the slower vibrational relaxation, as shown in [22].

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).