Ammonia cracking for hydrogen production using a microwave argon plasma jet

Ammonia (NH3) is a promising hydrogen carrier that effectively connects producers of blue hydrogen with consumers, giving rapid conversion of ammonia to hydrogen a critical role in utilizing hydrogen at the endpoints of application in an ammonia-hydrogen economy. Because conventional thermal cracking of NH3 is an energy intensive process, requiring a relatively longer cold start duration, plasma technology is being considered as an assisting tool—or an alternative. Here we detail how an NH3 cracking process, using a microwave plasma jet (MWPJ) under atmospheric pressure, was governed by thermal decomposition reactions. We found that a delivered MW energy density (ED) captured the conversion of NH3 well, showing a full conversion for ED > 6 kJ l−1 with 0.5-% v/v NH3 in an argon flow. The hydrogen production rate displayed a linear increase with MW power and the NH3 content, being almost independent of a total flow rate. A simplified one-dimensional numerical model, adopting a thermal NH3 decomposition mechanism, predicted the experimental data well, indicating the importance of thermal decomposition in the plasma chemistry. We believe that such a prompt thermal reaction, caused by MW plasma, will facilitate a mobile and/or non-steady application. A process combined with the conventional catalytic method should also effectively solve a cold start issue.


Introduction
The concerns of climate change have been instigated by increased greenhouse gas emissions, particularly those attributable to carbon dioxide (CO 2 ) from fossil fuel utilization.To stimulate nations and industries to replace fossil fuel energy 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.
with green energy from renewable sources, policies like carbon taxing and carbon pricing have been proposed and conducted.In this regard, hydrogen-driven by its high energy density (∼120 MJ kg −1 ) and eco-friendly emission-is considered as a clean alternative for combustion or a direct feed for fuel cells for an anticipated carbon-neutral society [1].On the flip side, explosion hazards, poor transport compressibility and storage issues [2] have slowed the application of hydrogen.
Recently, ammonia emerged as the most feasible hydrogen carrier for long-distance transportation and as a carbon-free fuel to burn directly without emitting CO 2 [3].This was greatly stimulated by a blue hydrogen concept in the oil industry: Hydrogen from steam reforming of methane, capturing emitted CO 2 [4,5].Although the toxicity and low reactivity of ammonia [6] are limiting factors for its terminal application as a fuel, the low liquefied pressure (8.6 bar at 20 • C) and high hydrogen content of ammonia (17.8% by weight) make it a good hydrogen carrier-solving the storage issue.Thus, transporting hydrogen in the form of ammonia, cracking back to hydrogen at the terminal point of use, seems to be a feasible strategy, making rapid and efficient transformation of ammonia to hydrogen of great importance for the entire ammonia supply chain.
The decomposition reaction of ammonia is endothermic and-in accordance with thermodynamics-the ammonia conversion ratio is almost 100% at 400 • C under atmospheric pressure [7].Thus, thermal cracking-with or without a catalyst-has traditionally been the most common technique for ammonia decomposition [8][9][10].However, due to the warm-up period for cold start, conventional thermal cracking is good for steady and stationary operation.To mitigate the cold start issue, atmospheric pressure plasmas can be promising alternatives [11,12], because of their positive synergies with catalysts and readily available plasma-induced reactions.
Plasmas can provide a reacting environment characterized by electron temperature (T e ) and gas temperature (T g ), depending on the specific plasma source.In general, plasmas with T e > T g and T e = T g are referred to as nonthermal plasmas and thermal plasmas, respectively [13].In the thermal plasmas (e.g.arc plasma with both T e and T g ∼ 1 eV (∼10 4 K)), conventional thermal reactions due to high T g prevail over electron-induced reactions [14].On the other hand, because of the highly energetic electrons in nonthermal plasmas (e.g.T e ∼10 eV for dielectric barrier discharges, DBDs, or nanosecond repeatedly pulsed discharges, NRPDs), the electron-induced reactions play a significant role in producing chemically active species (electrons, radicals, and metastable excited states) [15][16][17], although a final product composition is governed mostly by thermal reactions for a given T g [18,19].For this reason, nonthermal plasmas were widely used in plasma chemical kinetic studies with a particular focus on electron-induced reactions [15][16][17][18][19][20][21][22].Recently, Bang et al [23] published the first comprehensive plasmachemical kinetic mechanism for ammonia cracking based on DBD, which included both electron-induced and conventional thermal reactions.Their results indicated that significant collective efforts from the related field are required to fill the gap in much fundamental data (e.g.cross section data with an electron).However, due to the relatively low gas temperature of DBDs [24][25][26][27], overall conversion efficiency of NH 3 and the H 2 yield was not generally feasible.Therefore, these nonthermal plasma methods were usually combined with heating or catalysts [23,[28][29][30][31].
To take advantage of plasma-chemical reactions for the ammonia cracking process, a hotter plasma source, in terms of T g , should be considered, since an almost full conversion of ammonia can be achieved for T g > 1370 K [32].In this regard, microwave plasma could be a viable plasma source for ammonia cracking.A microwave plasma jet (MWPJ) is known to have a relatively high gas temperature (2000-6000 K) with relatively lower electron energy (<1 eV) [13].Moreover, MWPJ is free from electrode erosion and corrosion (therefore also free from contamination from electrodes), due to its intrinsic method of building an electric field (remotely generated microwave from a magnetron is transmitted to a quartz tube carrying a gas stream).For these reasons, MWPJ has attracted considerable attention in the area of fuel transformation [33][34][35][36][37][38][39][40].
However, the MWPJ is usually operated using noble gases to facilitate the ionization process-particularly in a low power range (∼order of 10 2 W).Using the production of H 2 from ethanol as an example, Jimenez et al [36,38] and Rincon et al [37] studied the performance of MWPJ using 200 W and 200-500 W microwave power, respectively, by investigating the effect of Ar flow on T g and H 2 yield.Nevertheless, the effect of T g in the process remains unclear.Also, since the literature lacks a study on ammonia cracking by MWPJ, investigating its application potential toward NH 3 conversion and the critical factors influencing ammonia decomposition is worth exploring.
To this end, the present study targeted the potential of the MWPJ in the process of ammonia cracking for hydrogen production, seeking fundamental clarification of crucial factors governing the process and a practically feasible direction.Experiments were conducted under atmospheric pressure with systematically designed experimental conditions: MW power, a total gas flow rate, and an ammonia content.Optical emission spectroscopy and Fourier transform infrared spectroscopy were used to analyze the physical characteristics of the MWPJ and the ammonia conversion, respectively.By considering related thermal reactions, a simplified one-dimensional numerical model was established to understand the ammonia conversion process.The numerical result was compared with the experimental data to support the rationale behind the physicochemical mechanism for the MWPJ-induced NH 3 cracking process.

Experiment
The experimental setup consisted of a MW plasma generator, a reactant supply system, and measurement systems, shown in figure 1.A solid-state MW source (Sairem, GMS200W, 2.45 GHz) transmitted MW to a surfatron device (Sairem, Surfatron60), equipped with a 240-mm long quartz tube (inner and outer diameters are 4 and 6 mm, respectively) at the center and a built-in igniter to generate the MWPJ [14,41,42].The surfatron was cooled and maintained at room temperature using a water cooler operated at 20 • C. To avoid overheating during the discharge, the quartz tube was also cooled using a nitrogen flow.The incident MW power was adjustable from 50 to 150 W, while the reflected power was limited to less than 10% of the incident power.
To facilitate MW discharge, Argon (Ar, 99.999% purity) was chosen as a baseline feed gas.Once the MW discharge was stabilized with Ar flow through the quartz tube, ammonia (NH 3 , 99.98% purity) was added to the Ar stream.Two mass flow controllers (Brooks, SLA5850) controlled each flow rate of Ar and NH 3 .A tested range of the total flow rate-a sum of the flow rates of NH 3 and Ar-was 1-5 slpm.An initial NH 3 A spectrometer (Princeton Instruments, SP2750), equipped with a fiber-optic cable, a grating (900 grooves mm −1 ), and an intensified charge-coupled device (ICCD) camera (Princeton Instruments, PI-MAX3), was utilized to obtain spectral data emitted from the MW plasma.The fiber-optic cable collected emissions through a window at the bottom of the surfatron.A digital camera (Nikon, D500) captured the images of the MWPJs.
A Fourier transform infrared spectroscopy (FTIR, Thermo Fischer Scientific, Nicolet iS10) was used to analyze the concentration of NH 3 in the product, and thus the NH 3 conversion (X NH3 ) was defined as follows: where C NH3,f represent the final NH 3 concentrations.The term, C NH3,i C NH3,f in the denominator, appears to correct the molar increase for C NH3,f due to the conversion of NH 3 assuming 2NH 3 → N 2 + 3H 2 .Note that no absorption signal could be identified other than that of NH 3 in the FTIR measurement (600-4000 cm -1 ), confirming that only N 2 and H 2 were the final products (see figure S1 in the supplementary information, SI).Therefore, a H 2 production rate (Q H2 ) and H 2 production efficiency (η H2 ) were calculated as follows: where Q t is the total flow rate; P MW is the net delivered MW power; η H2 (= 0.09 g l −1 ) is the density of H 2 at the standard temperature and pressure.NH 3 conversion efficiency can be similarly defined as where η NH3 = 0.76 g l −1 .

General characteristics of MWPJ
To understand the physical characteristics of the MWPJ, the morphological structure of the MWPJ was investigated, depending on operating parameters such as Q t , P MW , and C NH3,i .First, pure Ar flows were considered; the resulting images of the MWPJs are displayed in figure 2 for various P MW and Q t .As the total flow rate is maintained at ), the height of the MWPJ (H P ) is increased almost linearly with P MW in a range of 50-150 W.However, for cases with fixed P MW = 100 W (figure 2(b)), the increase in Q t (up to 3 slpm) results in the slight decrease in H P , demonstrating a more or less unchanged feature for further increased Q t .
The visible height of the MWPJ was mainly attributed to a combined effect of energy density (ED = P MW /Q t ), producing light emitting species (e.g.excited states of Ar), and the advective traveling distance of the light emitting species, due to a bulk flow speed along the quartz tube.Because the advective distance should be proportional to Q t , the height of the MWPJ could be hypothesized to be proportional to ED and Q t : thus, H P ∼ ED(=P MW /Q t )× Q t ∼ P MW .This explained the observed trends of H p in figure 2 very well.The small decrease in H p for the two-fold increase in Q t (from 1 to 3 slpm) could be attributed to the dynamic behavior of the MWPJ.As Q t increased, the MWPJ became unstable and tore into two or more branches, rotating along the tube wall (see Supplementary Video).Emission spectra were measured for 5 different Q t as in figure 2(b) at P MW = 100 W; no discernable changes in the intensities and profiles of OES could be observed (see figure S2 in the supplementary information).Recalling that the collected light intensity should be proportional to an overall light emitting volume (see section 2), this indirectly evidenced that the overall plasma volume reasonably maintained for the tested range of Q t even with the branching phenomenon.It should be mentioned that previous literature reported the unaffected height of Ar MWPJ in a Q t range of 0.5-1.5 slpm [38].The cause of this difference can be attributed to the fact that the diameter of the present quartz tube was two times larger than that in [36,38], which could provide sufficient space for the dynamic behavior noted above.In addition, Reynolds numbers (Re) for all tested conditions were confirmed to be smaller than a critical Re for a laminar to turbulent transition inside a circular tube (Re ∼ 2300), indicating that all flow conditions were in a laminar regime.

Addition of ammonia
We could find a significantly reduced height and a change in the emission of the MWPJ, as NH 3 was added.Figure 3 shows the images of the MWPJ with added NH 3 (C NH3,i = 0.5% v/v) for a wide range of P MW and Q t , demonstrating significantly reduced H p .Similar observation was reported when  ethanol was added to an Ar MWPJ [38].Because the emission is related to the relaxation process from the excited states of Ar (Ar * ), the shortened H p with the added NH 3 could be attributed to (i) electron-energy-loss to NH 3 [38], which resulted in less production of Ar * and (ii) the quenching of Ar * due to collision with NH 3 .
Although the height was reduced drastically because of the added NH 3 , figure 3(a) also shows an increased trend of H p for the increased P MW , like those cases with pure Ar.However, the rate of increase slowed as P MW increased.For the fixed P MW = 100 W (figure 3(b)), the height is relatively unchanged compared to the big increase in Q t : a 400% increase in Q t results in only a small increase in H p .This might be attributed to more stable plasma dynamics (the plasma column was considerably contracted radially, demonstrating neither branching nor rotating behavior) than the case with pure Ar.
To understand experimental regimes that depend on initial NH 3 content, we investigated the effect of C NH3,i on MWPJ's stability in terms of P MW and Q t .Generally, as C NH3,i increased, the MWPJ height was reduced drastically, along with increased reflected MW power.For further increased C NH3,i , an extinction of the MWPJ occurred eventually.To systematically determine a critical (maximum) C NH3,i , over which the MWPJ was extinguished, for each tested combination of P MW and Q t , we used a limit of the reflected power (10% of the incident power) as a criterion.When the reflected power reached this limit, C NH3,i at the moment was considered as the critical C NH3,i for a given condition.As a result, figure 4 exhibits the critical C NH3,i representing the extinction of the MWPJ with P MW as an independent variable for various controlled Q t .
It was apparent that the MW power mainly governed the critical C NH3,i , affecting an almost linear increase of the critical C NH3,i with P MW , whereas the total flow rate was secondary, demonstrating relatively small negative impact on the critical C NH3,i for a given P MW (figure 4).MW discharge has been known to operate in a relatively low electric field, having difficulties in igniting by itself, and thus it requires an external source for seed elections to breakdown a gaseous media [42][43][44][45].As ammonia was added, not only some portion of the total electron energy was transferred to NH 3 but also the excited states of Ar lost their energy due to the collisions with NH 3 , resulted in less ionization of Ar [38].If the energy loss fraction toward non-ionization process exceeded a certain level (due to increased NH 3 content), the MWPJ could not be sustained by a given P MW .Therefore, the linearly increased critical C NH3,i shown in figure 4 was attributed to this energy loss fraction toward non-ionization collisions.The negative effect of Q t on the critical C NH3,i might be understood based on the advection of electrons from the root of the MWPJ.When the characteristic ionization time was not sufficiently shorter than the increased characteristic flow time (due to the increased Q t ), the MWPJ could no longer be sustained.As shown in figure 4, this effect is secondary for the tested ranges of P MW and Q t compared to that of P MW .
It was the inevitable disadvantage of a low power surfatron device (order of 10 2 W) to be operated with only a small portion of a target species.However, as a high-powered magnetron and a waveguide coupling device was employed, a practically feasible NH 3 content could be treated, even without the use of noble gases.

Ammonia conversion and hydrogen yield
A better NH 3 conversion was obtained using a higher P MW or a lower Q t .Figure 5(a) shows the NH 3 conversion with a controlled C NH3,i = 0.5% v/v for various P MW (50-150 W) and Q t (1-5 slpm).Overall, the conversion increases for increased P MW at a fixed Q t , and it decreases as Q t increases at a fixed P MW .This should be attributed to the increased reaction volume (height of the MWPJ) for the increased P MW (figure 3(a)) and the decreased residence time due to the increased Q t in a reasonably maintained reaction volume (figure 3(b)), respectively.The linearly increased trend of X NH3 for Q t = 5 slpm is changed as Q t decreases.For Q t = 2, 3, and 4 slpm, the rate of increase in X NH3 slows as P MW increases from 125 to 150 W. For the smallest Q t tested (1 slpm), X NH3 shows full conversion from 100 W onward.
Interestingly, as the result is replotted with the production rate of H 2 (figure 5(b)), Q H2 demonstrates relatively insensitive behavior to Q t , showing linear increase only with P MW .However, for P MW > 125 W, due to the reduced increasing rate of X NH3 with Q t = 2, 3, and 4 slpm compared to those for P MW < 125 W, the rate of increase in Q H2 also decreases, resulting in somewhat scattered Q H2 at P MW = 150 W. The saturated Q H2 with Q t = 1 slpm from P MW = 100 W was attributed to the full conversion of NH 3 caused by the limited supply of NH 3 .
Conversions of NH 3 for various combinations of P MW and Q t for the controlled C NH3,i = 0.5% v/v (figure 5(a)) could be correlated successfully with the energy density, as illustrated in figure 6.Like other plasma-assisted chemical processes [18,46], the energy density (input energy per unit volume of the treated gas) accurately characterized the NH 3 cracking process in the MWPJ.The conversion trend in figure 6 can be divided into two regimes: a linear regime and a full conversion regime.As a result of linear data fitting in the linear regime, 5.7 kJ L −1 was found to be a critical ED for the full conversion.
Although C NH3,i was varied in a range of 0.5-1.1% v/v, the conversion rate was reasonably maintained for a given P MW and Q t .Figure 7(a) shows X NH3 along C NH3,i at P MW = 100 W for Q t = 1, 2, and 4 slpm, and relatively unchanged X NH3 can be found for the three Q t .However, as the hydrogen production rates (figure 7  Q H2 is obtained as C NH3,i increases, demonstrating reasonably similar Q H2 for a given C NH3,i even with three-fold increase in Q t (1-4 slpm).
Therefore, we could conclude that a minimized Q t would be the best operating condition toward a cost-effective production of H 2 , a maximized C NH3,i , because the minimized Q t also reduced the consumption of ammonia with the maximized production rate of hydrogen.
We compared the NH 3 conversion and the H 2 production efficiency (η H2 , equation (3) in the present study with the literature (table 1).The result of the present MWPJ, shown in table 1, was obtained for the case of Q t = 2 slpm with C NH3,i = 2.32% v/v at P MW = 150 W, which showed the best η H2 among the conditions tested (detailed η H2 and η NH3 can be found in figure S3 in SI).Considering the compared DBD cases [23,31,[47][48][49][50] in table 1, the present study showed better η H2 .This could be attributed to the relatively high gas temperature (T g ) in the MWPJ, facilitating endothermic reactions such as the abstraction of H from NH 2 and NH 3 by H atoms to form H 2 : It should be noted that the absolute value of η H2 in the present study was still low, considering other plasma sources having comparable gas temperature.Previous studies using nonthermal arc plasma [28] and gliding arc plasma [51] showed significantly higher η H2 even with lower conversion of NH 3 as compared to the present result.This could be attributed to the very low C NH3,i in the present study.Therefore, to improve H 2 production efficiency, increasing the critical C NH3,i in a MWPJ system will be essential.

Optical emission spectroscopy: gas temperature
Motivated by the endothermic nature of the dehydrogenation reactions of ammonia (equation ( 4) and ( 5)), we conducted an optical emission spectroscopy to estimate the gas temperature of the MWPJ, which was further utilized in a numerical simulation in section 3.5.Figure 8 shows the resultant spectra of the MWPJ for cases with Q t = 1 slpm and C NH3,i = 0.5% v/v at P MW = 50, 100, and 150 W, identifying the spectral lines of Ar (690-850 nm) [52], NH (A-X: ∼336 nm, 2nd order: ∼672 nm) [42], H α (656.3 nm) [52], H β (486.1 nm) [28], N 2 (C-B: 337-380 nm) [53], and N 2 + (B-X: 391.9 nm) [49].The collected spectral intensity increases along the increased P MW , demonstrating no discernable changes in spectral lines.This was attributed to the integration of the emission along the jet axis, which was related to the height of the MWPJ shown in section 3.2 (H p ∼ P MW ).Thus, the relative intensity might be proportional to a luminous length.
To estimate the gas temperature, the rotational temperature of N 2 + was determined with Specair software (SpectralFit) using the N 2 + (B-X) line [54].Since the rotational-translational relaxation process should occur quickly under the MW plasma condition of interest, both rotational and gas temperatures could be reasonably similar [55].Figure 9 shows a typical fitting result of an experimental spectrum collected at P MW = 100 W for the case with Q t = 1 slpm and C NH3,i = 0.5% v/v.The synthetic spectrum at the rotational temperature (T rot ) of 4,500 K agreed well with the experimental spectrum, and thus, T g was considered to be 4,500 K in this case.
We varied P MW (50-150 W) and Q t (1-5 slpm) for controlled Q t = 1 slpm and P MW = 100 W, respectively, to examine their effects on T g for the cases at fixed C NH3,i = 0.5% v/v. Figure 10(a) shows the trends of T g for the two sets of data: read red-circle symbol along Q t (lower x-axis) and blue-square along P MW (upper x-axis).For the given Q t = 1 slpm, T g = 4,260 K at P MW = 50 W and slowly increases to   T g = 4,870 K at P MW = 150 W. For the controlled P MW at 100 W, T g = 4,560 K for Q t = 1 slpm and gradually decreases to T g = 4,150 K for Q t = 5 slpm.Although those two trends were reasonable, the changes observed in T g were very small, considering the three-and four-fold variations in P MW and Q t , respectively.Similar results, showing relatively insensitive T g , have also been reported in previous studies [34,52], and a reasonable explanation could be attributed to the fact that the radial cross-section of the quartz tube used in the present study was not completely filled with the luminous plasma (see figure 3).This indicated that a portion of the mixture was bypassed without passing through the luminous plasma zone.In combination with the result in section 3.2 (figure 3), we could conclude that the increased P MW resulted mainly in the increased plasma volume (height or diameter), while the increase in T g was minimal.
As the data is replotted in terms of ED (figure 10(b)), a trend overlapping well between those two data sets can be found, indicating a close correlation between ED and T g .Recalling the result in figure 6 (showing good correlation between X NH3 and ED), this also implied the importance of T g on the conversion of NH 3 .Thus, a hypothesis could be proposed to explain the result shown in figure 6: Energy density influenced the ammonia conversion process mainly through released heat characterized by T g due to the MW discharge.

Numerical simulation: thermal decomposition of NH 3
The MWPJ in this study was characterized by high gas temperature (>4000 K), while the electron temperature has been known to be relatively low (<1 eV) in general, compared to that in DBDs (∼order of 10 eV) [13].This implied that the thermal effect of the MWPJ on the ammonia conversion was likely to dominate over the electron-induced reactions.A previous plasma chemical kinetic study showed that the electroninduced reactions could initiate the decomposition of NH 3 in a relatively low-temperature regime; however, these were substantially ineffective as compared to the thermal reaction at higher temperature [23].The comparison in table 1 also shows that DBDs' (characterized by high electron energy) H 2 production efficiencies are poorer than hotter plasma sources (MW plasma, nonthermal arc, and gliding arc), indicating ineffectiveness of the electron-induced reactions.Since the endothermic dehydrogenation reactions (equations ( 4) and ( 5)) govern the decomposition reaction of ammonia in a hightemperature regime, almost all ammonia could be converted to hydrogen in chemical equilibrium at ∼700 K atmospheric pressure condition [7] and in a plasma chemical kinetic study at ∼1700 K [23].Thus, in the present study (T g > 4000 K), we believed that the role of Ar * should be secondary, while the thermal decomposition of NH 3 should be a dominant chemical pathway.In this regard, a thermal decomposition mechanism was used to model the present NH 3 cracking reaction.
To investigate the thermal effect on the conversion, a onedimensional (1-D) plug flow reactor (PFR) in Ansys Chemkin-Pro was adopted.A thermal NH 3 decomposition mechanism, developed by Bang et al [23] (which included 47 reactions with 13 species), was employed without considering electronimpact reactions.To be compatible with the 1-D PFR simulation, we modeled the MWPJ as a region with high temperature, and MW power affected the length of the high temperature region (axial length of the PFR).As discussed previously, H p was only dependent upon P MW, regardless of Q t ; thus, the length of the PFR for each P MW was determined based on the observation in figure 3, noting that the root of the MWPJ was assumed to locate at the axial center of the surfatron (figure 11).As Q t varied, the residence time in the PFR was naturally adjusted for a given length of the PFR (i.e.given P MW ).For simplicity, T g = 4,500 K (average T g in figure 10(b)) was consistently applied in all cases.The initial ammonia content was set at C NH3,i = 0.5% v/v.
As a result, we found that the ammonia conversion predicted by the simple 1-D PFR modeling captured the general trend of the conversion well in terms of ED. Figure 12 shows the comparison of X NH3 between the simulation and experiment for various Q t and P MW with a controlled C NH3,i = 0.5% v/v.Although the predicted tendency matches well with the experimental one, the simulation overpredicted X NH3 for ED < 3.75 kJ l −1 , while it underestimated X NH3 for ED > 6 kJ l −1 .This could be partially explained by the used average T g , which was higher than those of the cases for ED < 3.75 kJ l −1 and lower than those of the cases for ED > 6 kJ l −1 (see figure 10(b)).Note that the main purpose of this simple 1-D simulation was to briefly support the dominant role of T g in the MWPJ-induced cracking of NH 3 ; hence, heat loss and the radial size of the MWPJ were not considered, nor were electron-induced reactions.Nevertheless, it was considered sufficient to show that the NH 3 cracking process occurring in the MWPJ was driven by thermal reactions, rather than electron-induced reactions.It also indicated that this method can be improved by optimizing the use of heat released from the plasma.

Conclusions
This work investigated the potential of a microwave plasma jet (MWPJ) in ammonia (NH 3 ) cracking for hydrogen production.The MWPJs were formed by a surfatron device, using argon (Ar) as a base feeding gas; the experiment was conducted under various ranges of an input microwave power (P MW ), a total flow rate (Q t ), and an initial NH 3 content (C NH3,i ) under atmospheric pressure.A simplified one-dimensional (1-D) thermal reaction model was proposed for the NH 3 cracking and the simulation result agreed well with the experimental result.Based on the results and analysis, the following details could be summarized: (1) Conventional thermal decomposition reaction played a primary role in the MWPJ-induced NH 3 cracking process.Using gas temperature obtained in the MWPJ (∼4500 K), the 1-D thermal reaction model predicted the conversion of NH 3 well, indicating that electron-induced reactions were secondary in the MWPJ-induced NH 3 cracking process.Comparing the cost of the H 2 production, the result from this work was significantly better than using dielectric barrier discharges (even combined with heated gas or a catalyst).(2) An energy density (ED = input MW energy per unit flow volume) best described the NH 3 conversion (X NH3 ), regardless of P MW , Q t , and C NH3,i , showing a linear relationship between ED and the conversion for the ranges of the parameters tested herein.A full NH 3 conversion would be expected at ED = 5.7 kJ L −1 in the present setup.(3) A hydrogen production rate (Q H2 ) linearly increased with C NH3,i , showing minimal dependence of Q t .Therefore, for practical consideration, a minimized Q t and maximized C NH3,i would be more conducive to effective hydrogen production.
(4) The height of the MWPJ was proportional to P MW , while it was insensitive to Q t , which was supported by experiment and simple scaling analysis, based on ED and advection.Due to the addition of NH 3 , the height was significantly reduced, and a critical C NH3,i (over which the MWPJ was extinguished) was observed.
We found that hydrogen production rate and energy efficiency were significantly limited by low C NH3,i and heat loss.Also, argon was not a practical working gas for large-scale NH 3 cracking.In this regard, future work will focus on: (i) optimizing the reactor design and heat management to increase hydrogen production rate and efficiency, and (ii) replacing the Ar stream with pure NH 3 , or a N 2 diluted NH 3 stream, at the least.

Figure 2 .
Figure 2. Images of pure Ar MWPJs at exposure time of 1/100 s: (a) for various MW powers (P MW ) at Qt = 1 slpm and (b) various flow rates (Qt) at P MW = 100 W.

Figure 3 .
Figure 3. Images of NH 3 added MWPJs at exposure time of 1/320 s: with fixed initial NH 3 concentration (C NH3,i ) at 0.5% v/v.(a) for various MW powers (P MW ) at Qt = 1 slpm and (b) various flow rates (Qt) at P MW = 100 W.

Figure 4 .
Figure 4. Critical ammonia content for MWPJ's extinction along MW power (P MW ) for various flow rates (Qt).

Figure 5 .
Figure 5. Ammonia conversion (a) and corresponding hydrogen production rate (b) with C NH3,i = 0.5% v/v for various total flow rate (Qt) and MW power (P MW ).

Figure 6 .
Figure 6.Ammonia conversion as a function of energy density for cases with C NH3,i = 0.5% v/v.

Figure 7 .
Figure 7. Ammonia conversion (a) and corresponding hydrogen production rate (b) at fixed P MW = 100 W for various initial NH 3 (C NH3,i ) contents and total flow rates (Qt).

Figure 9 .
Figure 9.Comparison between synthetic spectrum of N 2 + (B-X) at Trot = 4,500 K (rotational temperature) and experimental spectrum collected at MW power (P MW ) of 100 W and total flow rate (Qt) of 1 slpm.

Figure 10 .
Figure 10.Estimated gas temperature (Tg) for various total flow rates (Qt) and MW powers (P MW ): (a) effect of P MW and Qt and (b) effect of energy density on Tg.

Figure 11 .
Figure 11.Normalized axial length of PFR as a function of MW power (P MW ).

Figure 12 .
Figure 12.Comparison of ammonia conversion (X NH3 ) between experiment and simulation using a simple 1-D plug flow reactor (PFR) model adopting a thermal NH 3 decomposition mechanism [23].

Table 1 .
Comparison of ammonia conversion (X NH3 ) and hydrogen production efficiency (η H2 ) between present MWPJ and other DBD plasma reactors.