Plasma-catalytic ammonia decomposition using a packed-bed dielectric barrier discharge reactor

(cid:1) The plasma was found to achieve an NH 3 conversion of 5% at ambient conditions. (cid:1) Introduction of dielectric materials with ε ¼ 4 e 30 improved conversion. (cid:1) MgAl 2 O 4 ( ε ¼ 8.3) yielded the highest conversion found of 15%. (cid:1) A linear correlation between conversion and number of micro-discharges was found. (cid:1) The decomposition primarily proceeds in the gas phase by collision with electrons.


Introduction
Hydrogen (H 2 ) is considered an important fuel of the future, because it is a clean energy source with water as the only product from its combustion [1,2].Certain small-scale power generation units, such as fuel cells, use H 2 as fuel for production of green electricity.However, delivery and storage of H 2 is hindering the commercialization of H 2 fuel cell-based processes, along with catalyst and reactor stability of a fuel cell system [3].On-site H 2 generation via ammonia (NH 3 ) is currently investigated as a solution to the delivery and storage issue, because NH 3 can be condensed at room temperature (25 C) and a relatively low pressure of ca. 10 atm, which yields a higher volumetric energy density than pressurized H 2 [3e6].Furthermore, NH 3 decomposition in principle only yields nitrogen (N 2 ) as by-product (see Reaction 1).
To use NH 3 as a H 2 carrier, effective conversion processes that avoid formation of byproducts, such as hydrazine [7], should be developed.Several heterogeneous catalysts, such as Ru, Ni, Fe, and Co, are capable of promoting the reaction, but require reaction temperatures above 500 C to achieve a reasonable conversion level [8e14].Lendzion-Bielun et al. [8] found that Co promoted with small amounts of Al 2 O 3 , CaO, and K 2 O could reach an NH 3 conversion of 50%, while Li et al. [13] found that Ru/SiO 2 could achieve 96.5% conversion and Ni/SiO 2 only could reach 34.6%, all at 550 C. Furthermore, Silva et al. [14] reached an NH 3 conversion of >99% with a FeeNi/g-Al 2 O 3 egg-shell catalyst at a temperature of 650 C.However, the thermodynamic equilibrium (Fig. 1) shows that a conversion of >99% can be achieved at a temperature of 375 C. Further development of inexpensive metal catalysts that can reach a reasonable conversion at lower temperatures, or alternative processes, are therefore required to improve the H 2 generation via NH 3 decomposition.
Recently, the use of plasma-catalytic processes have shown an enhanced ability for NH 3 decomposition where both diluted [15e17] and pure NH 3 feeds have been investigated [18e23].Akiyama et al. [15] examined NH 3 decomposition in a DBD reactor with wool-like inner electrodes consisting of different metals (Ag, Al, Cu, Fe, and Ti) in a range of applied power.No external heating was used and the gas mixture contained 4.87% NH 3 in N 2 [15].The decomposition rate was found to be independent of the electrode metal employed and linearly dependent on power in the range 0e50 W, with full decomposition achieved at approximately 50 W. El-Shafie et al. [16] investigated the effect of Al 2 O 3 particle size (1e2 mm) with a 0.5% NH 3 -Ar feed and found that the smaller particle size (1 mm) yielded a higher NH 3 conversion independent of feed flow rate (0.1e1 l/min) and plasma voltage (12e18 kV).In another study by El-Shafie et al. [17] the effect of using different packing materials (Ru/Al 2 O 3 and SiO 2 ) was explored.Here an increased energy efficiency was reported when using Ru/Al 2 O 3 compared to SiO 2 when operating with a plasma voltage of 18 kV.Wang et al. [21] found that when introducing a commercial Fe-based NH 3 -synthesis catalyst in a DBD reactor at 410 C, an NH 3 conversion of 99.9% could be achieved, while plasma alone and catalyst alone only reached 7.8% and 7.4% conversion, respectively.Hence, the plasmacatalytic system reached a conversion close to the thermodynamic limit (see Fig. 1).In another study, Wang et al. [20] showed that a Co/SiO 2 catalyst yielded an NH 3 conversion of 20% at 450 C and in combination with plasma a conversion of 99.2% was obtained.In both studies, pure NH 3 gas at a flow rate of 40 ml/min was examined [20,21].Hayakawa et al. [23] utilized a membrane DBD reactor to investigate the decomposition of NH 3 , where the plasma zone was 30 cm long.In the study they found that the membrane reactor increased the H 2 rate by ca.87% compared to a reactor without a membrane [23].This indicates that an equilibrium can occur in the plasma and that removing H 2 will improve the NH 3 decomposition.
As evident from the few studies of plasma-driven NH 3 decomposition in the literature, only little is known about the influence of solid, potentially catalytic, packing materials in the plasma zone, and more fundamental research is needed.In this study, decomposition of NH 3 was studied and the influence of utilizing different dielectric materials (alone or impregnated with various metals) in the discharge zone were Fig. 1 e Equilibrium composition of NH 3 , H 2 , and N 2 as function of temperature at 1 atm.
examined.In addition, the fresh and spent catalysts were characterized using X-ray diffraction (XRD).

Dielectric barrier discharge setup
A coaxial packed-bed DBD reactor of quartz was used for the plasma-catalytic ammonia decomposition as shown in Fig. 2. The outer electrode was a stainless steel mesh tightly wrapped around the reactor with a length of 5 cm, while the inner electrode was a stainless steel rod with an outer diameter of 10 mm.The inner diameter and wall thickness of the reactor were 19 mm and 1.5 mm, respectively, yielding a discharge gap of 4.5 mm and corresponding discharge volume of 10.2 cm 3 .The outer electrode was connected to a high voltage amplifier (TREK, model 20/20C-HS), which was controlled by a function generator (Tektronix, AFG1022).A high voltage probe (Tektronix, P6015A) was connected to the outer electrode to measure the applied voltage, while a Rogowski coil (Pearson, 4100) measured the current on the grounded line.Moreover, a low voltage probe (Picotech TA150) was used to measure the generated charges (Q) over an external capacitor (10 nF) placed in series on the grounded line.A digital oscilloscope (Picotech, Picoscope 6402C) was used for continuous recording of voltage, current, and charges, and calculation of plasma power.
The decomposition of NH 3 was mostly carried out at a plasma power of 21 W and a frequency of 3.0 kHz, except for experiments utilizing KM1R and TiO 2 (see further below) as packing material.Here a maximal plasma power of only 15 W could be achieved due to limitations of the amplifier.Complete filling of the discharge volume with packing material was utilized for the plasma catalysis experiments, equivalent to ca. 10 cm 3 of material.Glass wool was placed outside of the discharge zone on each side of the bed to hold the catalytic material in place.All experiments were conducted at atmospheric pressure without external heating.A flow of N 2 was passed through the reactor for 10 min to remove any oxygen, after which an NH 3 feed flow (99.999% purity) was maintained for 20 min to remove the N 2 before ignition of the plasma.After plasma ignition, a period of 2 h was used to stabilize the plasma.
NH 3 was fed to the reactor at a flow rate of 75 Nml/min through a Bronkhorst mass flow controller and the effluent was bubbled through a dilute aqueous solution of sulfuric acid to remove unconverted NH 3 .This solution was placed in a cooling bath and kept at a temperature of 4 C to avoid heating from the acid/base reaction and limit evaporation of water.The remaining flow (H 2 þ N 2 ) was measured on a flow meter (Gilian, Gilibrator 2, with a range of 1e250 cm 3 /min).The NH 3 conversion (X NH3 ) could then be determined from the inlet flow (Q in ) of NH 3 and the product flow of H 2 þ N 2 (Q out ) according to Equation (1).A factor of 2 is included in Equation ( 1) because two moles of NH 3 yield four moles of product (H 2 and N 2 ) (see Reaction 1).Variations in the measured product flow was used to determine the uncertainty of the conversion.
The energy input normalized H 2 yield was calculated according to Equation (2), where the ratio 3/2 is included due to the stoichiometry of Reaction 1, NH 3in is the molar flow rate of NH 3 fed to the reactor, M H2 is the molar mass of hydrogen, and P is the plasma power.
The specific energy input (SEI) is defined as the plasma power over the total flow rate (Equation (3)).In addition, we defined an energy efficiency (h NH 3 ) as the energy required for Fig. 2 e Packed-bed DBD reactor with analytical equipment used for NH 3 decomposition.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 3 2 0 8 1 e3 2 0 9 1 Reaction 1 relative to the actual energy input, which is calculated from the SEI and the reaction enthalpy (D r H) of the NH 3 decomposition (Equation ( 4)).A constant SEI of 16.8 J/ml was used for all experiments, except for KM1R and TiO 2 where a value of 12.0 J/ml was used.SEI ðJ = mlÞ ¼ Plasma power Total feed flow rate After determining the flow rate, the composition of the product stream was determined by an online gas chromatograph (Thermo Fisher Trace 1300 GC) equipped with a thermal conductivity detector (TCD).The Brunauer-Emmett-Teller (BET) surface area was determined for the mentioned dielectric materials by N 2 physisorption performed at À196 C (Quantachrome instruments, NOVA touch 2LX).The samples were degassed at 350 C for 4 h prior to physisorption.Pore volumes were determined by the water uptake of the dried materials (16 h at 120 C).The measured surface areas and pore volumes are listed in Table 1, along with the relative dielectric constants of the materials.

Plasma-catalytic material
Experiments were also made using MgAl 2 O 4 pellets impregnated with different metals.Incipient wetness impregnation was utilized to impregnate the pellets using nitrate salts as the metal precursors, after which the samples were dried and calcined at 450 C. The resulting materials used were 10 wt% Fe/MgAl 2 O 4 , 10 wt% Ni/MgAl 2 O 4 , 10 wt% Co/ MgAl 2 O 4 , and 5 wt%Fe-5 wt.%Ni/MgAl 2 O 4 , which was impregnated in two steps, first iron followed by nickel.
Furthermore, an Fe-based commercial NH 3 -synthesis catalyst (named KM1R, Haldor Topsoe A/S) was examined.In a standard experiment, the solids were crushed and sieved to a particle size of 0.85e1.4mm.
Prior to use, the M/MgAl 2 O 4 (M ¼ Fe, Ni, Co, FeeNi) materials were reduced in H 2 at 525 C for 8 h in a muffle furnace at a flow rate of 300 Nl/min.The samples were then passivated at room temperature with 1% O 2 in N 2 and transferred to the plasma reactor.The samples were then re-reduced before use in a 20 vol% hydrogen/argon plasma with a flow rate of 75 Nml/min for 2 h at a plasma power of 25 W (15 W for KM1R) with a frequency of 3 kHz.Previous work has shown that supported metal oxide catalysts such as Ni/Al 2 O 3 can be reduced in a low temperature hydrogen-argon plasma [29,30].

Electrical characterization
A simplified Q-U Lissajous diagram is shown in Fig. 3, which can provide information about the electrical behavior of the plasma.Four phases, two capacitive phases (AB and CD) and two discharge phases (BC and DA), occurring during one voltage period can be defined on the figure.The capacitive phases represent the periods where the reactor resembles a capacitor and no activity is present in the reaction.In contrast, the discharge phases represent the transformation of the gas volume to a plasma where a net flow of electrons is present.The capacitance of the reactor during a specific phase can be found from Fig. 3 from the related slope, dQ/ dU ¼ C [31,32].
Table 2 shows the electrical characterization of the experiments conducted with the dielectric materials.It can be seen that even though the approximately same peak-to-peak voltage (U pp ) (see Fig. 3) was recorded, the capacitance of the cell (reactor þ gas) (C cell ), the effective capacitance (z diel ), the  displaced charges (Q d ), and number of micro-discharges per half period (N MD ) changes with the introduced material.Furthermore, the effective capacitance (discharge phase) is observed to be less affected by the dielectric material than the capacitance of the cell (capacitive phases), with the nonpacked system yielding the highest effective capacitance and the lowest cell capacitance.HZSM-5, TiO 2 , and BaTiO 3 differ from the other materials very clearly by having a high cell capacitance, a high number of displaced charges, and a low number of micro-discharges per half period.The number of micro-discharges were determined by examining the current profile (see Fig. 4).In Fig. 4, a positive half period of the current and charge profile is shown.Here, it is observed that the current profiles for HZSM-5, TiO 2 , and BaTiO 3 show almost no spikes.In contrast to these three materials, the remaining dielectric materials tested and plasma-only showed several (þ20) intense current spikes, which is considered to occur due to a micro-discharge.A micro-discharge was therefore counted if the current value increased above the standard deviation.Eight half periods of logged data were used to determine the number of micro-discharges, and N MD is determined as an average per half period over these 8 half periods.

Effect of dielectric material and particle size
After 2 h of operation, the product flow rate was measured over a period of 15 min and the average was used to calculate the conversion according to Equation (1).The resulting conversions obtained from the plasma-assisted ammonia decomposition with the dielectric materials are shown in Fig. 5 with plasma-only as a reference, indicated by the dashed line.Tests with varying flow rate suggest that a higher conversion can be achieved at lower flow rates/longer residence times (Fig. S4).Consequently, the results in Fig. 5 should reflect reaction kinetics and not an equilibrium situation.A significant increase in the NH 3 conversion compared to plasma-only (5.16%) was achieved when implementing dielectric materials with a relative dielectric constant between 4 and 30 in the plasma zone, with MgAl 2 O 4 (dielectric constant of 8.3) yielding the highest conversion of 12.6%.Similarly, Wang et al. [20] observed that implementing SiO 2 and Al 2 O 3 increased the NH 3 conversion compared to plasma-only.It was further reported by Wang et al. [20] that utilizing packing materials with a higher dielectric constant, such as TiO 2 , resulted in a lower NH 3 conversion, which is similar to the results shown in Fig. 5.In the present results, the conversion falls below 1%, i.e. less than for plasma-only, when packing materials with a dielectric constant of 80e1000 are used.
In thermal heterogeneous catalysis, a support material is used to optimize parameters such as surface area, thermal stability, acidity, etc.However, it has previously been shown for other reactions performed in plasma, that implementing different dielectric materials in the plasma zone changes the conversion and electrical characteristics of the system [33,34].Fig. 4 e Measured current and charge profile for a positive half period.
Hence, the electrical properties of the support could similarly play an important role in plasma catalysis.Generally, a decrease in the electron density in the gas phase is observed when utilizing materials with a high dielectric constant in an electric field [28,35].This is due to accumulation of charges on the surface and loss of electrons to the walls, which is caused by the electric field enhancement at the contact points of these materials and could be a reason for the large decrease in the conversion observed for TiO 2 and BaTiO 3 (Fig. 5).This furthermore correlates to the high value of displaced charges observed (Table 2).The high electric field enhancement accelerates the electrons to the surface of the particles, which causes the electrons to lose their capability to move from void to void, hence the plasma is not able to extend from one void to a neighboring void and leads to a shift from a full gap discharge to local discharges [35].
Interestingly, using a material with a very low dielectric constant such as HZSM-5, also results in a reduced conversion and similar electrical measurements as TiO 2 and BaTiO 3 (high value of displaced charges and low number of microdischarges), although such a material with a low dielectric constant should promote a full gap discharge with a low electric field strength [28,35].When using materials with a relative dielectric constant around 10e20, an optimum in the conversion is achieved.These materials must therefore yield a more optimal balance between the extent of the volume of discharges and the strength of the electric field, which promotes electrons with sufficient energy for breaking an NeH bond (4.5 eV) and stepwise dissociate NH 3 , or for excitation to an unstable state that more easily leads to dissociation.This stepwise dissociation through an excited state has been reported for molecules such as CH 4 and CO 2 [36], and is likely the same for NH 3 .
From analysis of the Lissajous curves (Fig. S1), it can be observed that the materials showing increased conversion (SiO 2 , MgAl 2 O 4 , g-Al 2 O 3 , m-ZrO 2 , and t-ZrO 2 /La 2 O 3 ) promote a lower charge transfer, while also having higher intensity current spikes (Fig. 4).The electrical properties of the materials are thereby shown to be of great importance in plasma catalysis.Furthermore, no significant change in conversion was observed between MgAl 2 O 4 and g-Al 2 O 3 even though they were found to have different BET surface areas and pore volumes (Table 1), which suggests that surface reactions are unimportant for the enhanced conversion in the presence of these solid packing materials.
The absence of a dependence on surface area (and hence internal porosity within the particles) could be due to the absence of plasma streamers in the pores of the dielectric material.Zhang et al. [37] concluded that the pore size has to be in the order of or bigger than the Debye length for plasma streamers to penetrate into the pores of the structure.For a micro-discharge in a DBD plasma with a spike electron density and temperature of 10 20 m À3 and 20 eV (232000 K) [38], respectively, the Debye length is in the order of 3 mm, which is far greater than the actual pore size of the materials.It is therefore expected that no conversion is occurring in the pores due to the absence of plasma streamers.
The primary role of the solid packing material appears to be in regulating the number of micro-discharges.This is evident from Fig. 6, which shows a correlation between the number of micro-discharges per half period and the NH 3 conversion.Here it is observed that the conversion increases linearly with the number of discharges.Hence, the materials' ability to produce micro-discharges is the central factor for the NH 3 conversion, which to the best of our knowledge have not been reported in literature previously.This suggests that the decomposition reaction occurs in the gas phase and is enabled by the high energy electrons formed during the microdischarges.
The effect of packing the plasma zone with different particle sizes of MgAl 2 O 4 on NH 3 conversion is shown in Fig. 7.The conversions show a slightly decreasing trend with increasing particle size.A similar observation was made by El-Shafie et al. [16] where decreasing the particle size from 2 mm to 1 mm also resulted in a slightly higher conversion.However, when taking the experimental uncertainty into account, the statistically significant change is relatively weak.The electrical characterization is found to be similar for the different particle sizes (Table 3), hence no obvious change to the plasma can be found from these data, which is in good agreement with the relatively similar performance.The smaller particle size yields smaller voids between the particles, hence shorter distances for the electrons to jump.This could yield more paths for the electrons to go through the reactor, thereby increasing the probability of an electronmolecule collision, which in turn could increase the conversion.On the other hand, the smaller particle size leads to a weaker electrical field that lowers the energy of the electrons, making the collisions less energetic and thus not capable of breaking or destabilizing the NH 3 molecules [35].These phenomena seem to largely balance out, leading to the relatively weak size dependence in NH 3 conversion (Fig. 7).

Introduction of metal-containing materials
Impregnating MgAl 2 O 4 with metals resulted in a decreased conversion relative to the bare support, as shown in Fig. 8.The catalyst containing 10 wt% Ni still showed an improved conversion (7%) compared to the plasma-only, while the catalysts containing Fe, Co, and Fe þ Ni showed lower conversions of only ca.1%.Compared to the finding of Wang et al. [20] and Yi et al. [22], which used similar catalytic metals (Fe, Co, and Ni) and reported conversions above 35%, the observed conversions on Fig. 8 are far lower.However, in both their studies fumed SiO 2 was impregnated with metals and different operating conditions were also used, such as higher SEI.
Interestingly, nickel-containing catalysts have been shown to be effective in other plasma-catalytic systems where production of H 2 is desired such as reforming of hydrocarbons [39,40].In thermal catalysis, nickel-containing catalysts are used in natural gas steam reforming for H 2 production [41].Nickel thereby seems to be effective both in thermal catalysis and plasma catalysis when formation of H 2 is desired.
To better understand the influence of the impregnated MgAl 2 O 4 samples, the electrical parameters of these were analyzed (Fig. S3).It was observed that the catalysts (Fe, Co, and Fe þ Ni) with low conversion showed a high charge transfer number and low intensity current spikes, similarly to the dielectric materials providing low conversion (see Figs. 4  and 5).The number of micro-discharges per half cycle was found to be 3, 4, and 5 for FeeNi/MgAl 2 O 4 , Fe/MgAl 2 O 4 , and Co/ MgAl 2 O 4 , respectively, which all showed conversions of ca.1%, while Ni (7% conversion) was determined to promote 30 micro-discharges.These results are in good agreement with the trend seen in Fig. 6 (see Supporting Information Fig. S5).Hence the metal-containing solids affect the ammonia conversion by regulating the frequency of micro-discharges in the gas phase rather than through metal-catalyzed surface reactions under the used conditions.
Ideally, a metal is a perfect conductor and therefore has a net electric field inside of zero, which yields a dielectric constant of infinity.This could explain the similar electrical parameters between TiO 2 /BaTiO 3 and the metal-impregnated MgAl 2 O 4 .However, in contrast to dielectric materials with high dielectric constants, the metals will not have surface charging due to their conductivity [42].Furthermore, the use of an industrial ammonia synthesis catalyst, KM1R, with bulk iron as the primary phase also led to a decrease in the NH 3 conversion.In contrast to the experiments with the impregnated MgAl 2 O 4 , KM1R was only tested at 15 W. Furthermore, the composition of the KM1R also varies from the remaining catalysts, as it mainly consists of iron, while the other metal-containing materials tested mostly consist of a dielectric material (support).This could be the reason why KM1R deviates from the linear relation between conversion and micro-discharge frequency seen for all other experiments (see Fig. S5).
In thermal catalysis, transition metals provide the highest activity for ammonia decomposition since they have the most appropriate nitrogen binding energy [43].The absence of significant benefits from the transition metals is therefore another indication that the decomposition mainly occurs in the gas phase by interaction with the high energy electrons instead of on the surface of the catalysts.

Effect of insulating the reactor
A second set of NH 3 decomposition experiments was conducted with the same catalytic materials, where the reactor was insulated with woven ceramic fibers, hence no repacking was conducted between the first and second set of experiments with a specific packing material.The resulting temperatures with the insulation material tightly wrapped around the reactor are shown in Table 4.The temperatures were measured on the outside of the reactor wall with a thermocouple placed between the reactor and the insulation material, immediately after turning off the plasma.It can be seen that insulating the reactor increases the reactor temperature by approximately 100 C. The effect on NH 3 conversion of insulating the reactor is shown on Fig. 10.
A negative effect on the conversion is observed when adding the insulation for the modes with a conversion above 2%, with plasma-only decreasing from 5.0% to 4.4%, MgAl 2 O 4 from 12.6% to 9.9%, and Ni from 7.0% to 5.9%.For the remaining catalysts, the effect was within the experimental uncertainty.In spite of the increased activity that comes from the rise in temperature, the gas temperature is still too low for a thermal effect with the catalysts tested, even though the equilibrium prescribes a 90% conversion at 225 C. Wang et al. [18] observed a large decrease in NH 3 conversion (from ca.95%e12%) when the temperature was decreased from 400 C to 300 C when using Co/fumed SiO 2 .A similar observation was reported in another study by Wang et al. [19] where an Fe catalyst was used.Here, the plasma alone succeeded the NH 3 conversion of the plasma þ Fe catalyst at 300 C. Furthermore, Yi et al. [22] also reported a decrease in NH 3 conversion for a FeeNi catalyst (from ca.30%e11%) when decreasing the temperature from 400 C to 340 C.These observations correlate with the low conversions observed on Fig. 10 where a maximum temperature of 245 C was achieved, indicating that the surface reactions are not yet sufficiently fast at the achieved temperatures.
The number of micro-discharges was not affected by the temperature increase.However, the increase in gas temperature that comes from insulating the reactor, increases the volumetric flow rate and lowers the residence time of the gas in the plasma zone, which in turn lowers the probability of an electron-molecule collision.This change in gas residence time within the plasma zone is therefore the most likely reason for the lowered NH 3 conversion.This is in agreement with the observation of a declining conversion with increasing residence time when the residence time was altered through the feed flow rate (see Fig. S4).
It is worth noting, that a lower applied voltage was needed to reach the plasma power of 21 W when insulation was placed around the reactor.This gives a weaker electrical field, thereby causing the plasma to deliver a lower acceleration of the electrons, but the same energy per time.A lower energy transfer is therefore expected in the electron-molecule collisions, which in turn leads to fewer collision induced reaction events.Several parameters could therefore have caused the decrease in conversion.

Hydrogen formation rate and energy efficiency
From the gas chromatograph measurements, it was determined that the product flow from the experiments with a conversion above 2% contained a H 2 :N 2 fraction of 3:1 as prescribed by Reaction 1, with no NH 3 remaining after the absorption in sulfuric acid.The experiments with a conversion below 2% produced too low flow rates after absorption for accurate measurements on the gas chromatograph, hence a H 2 :N 2 fraction of 3:1 could not be determined, but still some H 2 was detected.The "make up" gas in these small samples was air, as an increased content of N 2 was detected, along with an impurity of O 2 .Hydrazine has been reported as a by-product from studies diluting NH 3 with Ar [7,44], but not from studies using pure NH 3 [18e22] or dilution with N 2 [15].In this work, pure NH 3 was converted into H 2 and no hydrazine was detected on the GC measurements.However, the hydrazine would also have been removed as the product stream was bubbled through the acidic solution.
The calculated energy input normalized H 2 yields (R H2 ) of selected packing materials are shown in Table 5, along with the energy efficiency and conversion.As expected from Equations ( 2) and ( 4), the H 2 yield and energy efficiency increase with conversion, giving that using MgAl 2 O 4 with a particle size of 0.300e0.425mm for NH 3 decomposition results in the highest energy input normalized H 2 yield (4.08 g/kWh) and energy efficiency (1.73%).For comparison, the theoretical value of R H2 based on the reaction enthalpy is 235 g/kWh.
To our knowledge, only few publications have investigated plasma-catalytic ammonia decomposition [15e23].Interestingly, a hydrogen formation rate of 32 g/kWh (98% NH 3 conversion) was reported by Wang et al. [18], i.e. about 10 times higher than reported in this study.However, this was at a measured reaction temperature of 380 C generated by the plasma and substantial thermal conversion must also have contributed.A lower flow rate, smaller plasma volume, and higher power input could be the reason for the higher temperature and NH 3 conversion achieved by Wang et al. [18], which resembles the parameters also used in Refs.[19e21].

Conclusion
It has been shown that a plasma is able to decompose NH 3 at atmospheric pressure and with no external heating and produce clean hydrogen from sustainable power.A decomposition of 5% NH 3 for plasma-only was achieved at the applied conditions.The effect of applying a packing material (pure dielectric material or impregnated with metal) in the plasma zone was tested.In general, dielectric materials with a relative dielectric constant in the range 10e30 were found to give strong synergies with the plasma, leading to enhanced NH 3 conversion.Interestingly, an optimum in the conversion was found when utilizing materials with a relative dielectric constant around 10 and implementing MgAl 2 O 3 (dielectric constant of 8.3) in the plasma zone increased the conversion by a factor of ca.2.5, to 12.6%.Furthermore, the effect of particle size of MgAl 2 O 4 was found to have only minor influence on the NH 3 conversion, with a maximum of 15.1% achieved with a particle size of 300e425 mm.
Impregnating MgAl 2 O 4 with metals (Fe, Ni, FeeNi, and Co) was found to decrease the conversion due to the fast removal of electrons caused by the metals' high conductivity and significantly lower number of formed micro-discharges.When insulating the reactor, causing a temperature increase of ca. 100 C, a decrease in conversion was observed because of a decreased residence time of the gas, without observations of an enhancement in thermally induced conversion.
In addition, there was found a linear correlation between the number of micro-discharges and the conversion, which indicates that the decomposition of NH 3 in plasma primarily proceeds in the gas phase by collision with high-energy electrons.Furthermore, the packing materials were observed to mainly alter the plasma parameters when the temperature is Different dielectric materials, commonly used as support materials in thermal catalysis, were tested in this work, including MgAl 2 O 4 (Haldor Topsoe A/S), HZSM-5 (Si/Al ¼ 25) (Zeolyst CBV5524G), SiO 2 (Saint-Gobain SS 61138), g-Al 2 O 3 (Saint-Gobain SA 6176), m-ZrO 2 (Saint-Gobain SZ 31163), t-ZrO 2 /La 2 O 3 (8.2wt%) (Saint-Gobain SZ 61156), TiO 2 (Saint-Gobain ST 61120), and BaTiO 3 (Sigma Aldrich 1120480250).

Fig. 3 e
Fig. 3 e Schematic Q-U Lissajous figure indicating typical measured values of the plasma.

Fig. 5 e
Fig. 5 e NH 3 conversion as function of the relative dielectric constant of different dielectric materials.Results obtained after 2 h in a 21 W (15 W for TiO 2 ) plasma with a frequency of 3 kHz, discharge gap of 4.5 mm, and flow rate of 75 Nml/min.

Fig. 6 e
Fig. 6 e NH 3 conversion as function of the number of microdischarges per half period.

Fig. 7 e
Fig. 7 e NH 3 conversion as function of MgAl 2 O 4 particle size (21 W, 3 kHz, 4.5 mm gap, and 75 Nml/min).Weight of packing was 10 g for all particle sizes.

Fig. 9
Fig. 9 shows the XRD patterns of the spent samples measured ex situ.The primary contribution to the diffraction pattern comes from the MgAl 2 O 4 support, which shows indications of both MgAl 2 O 4 , and Al 2 O 3 phases that are also seen in M/ MgAl 2 O 4 (M ¼ Co, Fe, FeeNi, and Ni) samples.The XRD pattern of the spent Co/MgAl 2 O 4 sample shows two weak reflection peaks at 43.8 and 51.1 that correspond to metallic Co, and the spent Ni/MgAl 2 O 4 sample shows reflections at 44.5 and 51.9 indicating metallic Ni.For Fe/MgAl 2 O 4 a reflection is present at 44.7 , which is consistent with metallic Fe.These samples thus verifiably contained metallic particles during the reaction, since metallic phases are even preserved in these ex situ analyses.If metal-catalyzed surface reactions were important in the plasma-assisted ammonia decomposition, these samples should thus have caused an effect.The diffraction pattern for the spent FeeNi/MgAl 2 O 4 sample does not differ strongly from the support.Hence, Fe and Ni must be present as either very small metal nanoparticles or as amorphous oxide phases, but given the ex situ nature of the

Table 1 e
Physicochemical properties of dielectric materials used.
a Supplier information.b Determined as a weighted average of the pure oxide components.

Table 2 e
Measured data from the oscilloscope and generated Lissajous figures (see Fig.S1in Supporting Information (SI)) for plasma-only and the dielectric materials.

Table 3 e
Measured data from the oscilloscope and generated Lissajous figures (Fig. S2) for the different particle sizes of MgAl 2 O 4 .

Table 5 e
H 2 formation rate and energy efficiency for selected packing materials.e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 7 ( 2 0 2 2 ) 3 2 0 8 1 e3 2 0 9 1