Nitrogen oxide emissions analyses in ammonia/hydrogen/air premixed swirling flames

Ammonia/hydrogen fuel blends have gathered interest as a promising solution for the development of a hydrogen economy, with advantages in storage cost or combustion properties compared to pure hydrogen or pure ammonia, respectively. In that pursuit, the present work reports the trends of nitrogen oxide emissions for ammonia/hydrogen blends at atmospheric conditions. NO, NO 2 and N 2 O productions/consumptions are approached in detail in combination with unburnt ammonia. All cases are measured in a turbulent, swirl-stabilised flame configuration across hydrogen fuel fractions from 0% to 25% and equivalence ratios from 0.55 to 1.30. A detailed chemistry analysis was conducted using a chemical reactor network (CRN) employing detailed reaction chemistry. The results show that NO and NO 2 emissions peaks around Φ = 0.8, whereas considerable amount of N 2 O is generated at very lean conditions, Φ ≤ 0.65. Availability of OH radicals and O/H pools in the flames contribute towards fuel NO formation, which in turn produces NO 2 and N 2 O. However, very lean conditions lead to lower temperatures that ensure the survival of N 2 O. The results identified Φ = 1.05 – 1.2 as the optimum equivalence ratios for reduced NO X emissions in ammonia/hydrogen blends, with further understanding of the flame chemistry responsible behind these emissions.


Introduction
Since the recent demonstration of ammonia and ammonia/methane combustion in a 50 kW micro gas turbine combustor by Kurata et al. [1], ammonia has seen a surge of renewed interest as an alternative fuel to establish a hydrogen economy, which led to detailed reviews by Valera-Medina et al. [2], Elishav et al. [3], and Kobayashi et al. [4]. Historically, combustion of pure ammonia has faced numerous challenges which have delayed its use as an industrial-scale fuel vector [5]. These challenges include a low heat release rate [4], low laminar burning velocity (maximum ~7 cm/s for NH 3 vs ~37 cm/s for CH 4 ) [6] high NO emissions under near-stoichiometric conditions attributed to increased fuel-NO [7] and a narrow range of stability [8]. Some solutions that can enhance the flame intensity of ammonia include fuel preheating [1], oxygen enrichment [9], or doping with more reactive fuels like syngas [10], methane [11], hydrogen [12] and both methane and hydrogen [13]. Hydrogen is a particularly promising additive, given its higher heat release rate and combustion intensity. Furthermore, within a combustion system, ammonia can be cracked at relatively low temperatures in-situ to produce hydrogen, as has previously been demonstrated in internal combustion engine systems [14]. However, research has shown that storage and transportation of hydrogen can be both challenging and expensive such that a binary fuel of higher ammonia/lower hydrogen content are a compromise between cost and combustion characteristics, [15].
Recent studies by Pugh et al. [16], Zhang et al. [17], and Franco et al. [18] have explored the product gas trends of ammonia/hydrogen flames in industrially relevant, swirl burner flame configurations. Furthermore, research of ammonia/methane flames in a tangential swirl burner configuration by Valera-Medina et al. [19] found that traditional, injection configurations are not suitable for ammonia-based blends, while other studies found that an approximate 30%/70% blend of H 2 /NH 3 or CH 4 /NH 3 is beneficial for combustion stability, as both high hydrogen [8] and high ammonia [20] blends can promote instability. Sorrentino et al. [21] showed the applicability of MILD combustion for pure ammonia with good stability and emissions performance when the reactor temperature was above 1300 K. Subsequent study by Ariemma et al. [22] showed that the addition of water in MILD ammonia combustion has the potential to improve both NO X emissions and stability regions, especially under fuel-lean conditions. Experimental studies in an industrial micro gas turbine system propose the use of two stage rich-lean combustion configuration, for both pure ammonia [23], ammonia/methane blends [24] and more recently, for liquid ammonia [25] combustion. Recent numerical study by Bozo et al. [26] on ammonia/hydrogen rich-quench-lean burners showed comparative efficiency to fossil fuel based power plants. These studies show that the NO emissions of the first (rich) stage should be optimised based on equivalence ratio to achieve low emissions at the outlet. However, other approaches have been suggested to use single stage combustion under very lean conditions [27]. The work successfully achieved stable, ultra-low NO combustion, although emphasising that other NO X emissions (particularly N 2 O) had not been quantified for these blends at the time. However, N 2 O must not been overlooked in the evaluation of product gas trends, as N 2 O has a 100-year global warming potential (GWP) that is 310 times that of CO 2 [28]. This means that even relatively low levels of N 2 O emission can nullify the benefits of using a carbon-free fuel, in this case, ammonia and hydrogen.
Existing literature mentions the production of N 2 O under certain conditions. For example, Lee et al. [29] studied numerically the effects of NH 3 addition on H 2 /air flames, showing that increasing NH 3 content in lean flames is a driving factor for N 2 O production. Okafor et al. [24] also showed the production of N 2 O in the lean region of ammonia/methane flames with high ammonia fraction. A further study of wall temperature influence on swirl flames suggests that lean combustion conditions in combination with low temperatures can also be a driving factor for N 2 O production in ammonia/air flames [30]. Despite this, very few studies provide experimental data for the full trends for N 2 O and other product gases in H 2 /NH 3 flames at industrial conditions, hence providing the motivation for this experimental campaign.
Recently, Ferraroti et al. [31] studied the uncertainty of modelling turbulent ammonia/hydrogen flames in a flameless burner, comparing the experimental data to chemical reactor network (CRN) data and computational fluid dynamics (CFD) model results. This study found that within the examined conditions, discrepancy in NO emissions caused by uncertainty of reaction rate constants of the detailed reaction chemistry model were significantly higher than the uncertainty derived from the numerical models. To tackle this challenge, commendable research efforts have been placed in studying the underlying chemistry of these blends and the establishing reaction mechanisms optimised for ammonia/hydrogen combustion. Experimental studies based on laminar burning velocity of ammonia [9] and ammonia/hydrogen blends [32] have noted NH (leading dominantly to production of HNO) as the key intermediary for the production of NO, with increasing production and consumption of NO from Zeldovich reactions as H 2 concentration increased. For oxygen enriched ammonia/air flames, Mei et al. [9] was also able to show that NH 2 and NH radicals predominantly formed H i NO (where i = 0,1,2) species, leading to contributing to NO formation, while deNOx reactions of NO with NH x to form N 2 O and N 2 H 2 acted as major consumption pathways. Furthermore, N 2 O doped flames have been studied by Han et al. [35] showing that an increase in N 2 O-doping enhances the laminar burning velocity through increased OH production, the main pathway featuring N 2 O being its production from NO and consumption to N 2 . Finally, in swirling flames, Mashruk et al. [12] showed correlations between N 2 O, NO locations with NH 2 , while Mei et al. [33] has established a correlation between NO and NH × H for blends of partially cracked ammonia/air. However, the ability of kinetic reaction mechanisms to predict N 2 O has not yet been established and further work is necessary to explore the underlying chemistry that drives NOx emission trends in H 2 /NH 3 flames. Therefore, the present work evaluates the production of N 2 O, NO, NO 2 and unburnt ammonia in a turbulent, swirl-stabilised flame configuration in ammonia/hydrogen blends. Effects of changing equivalence ratios on emissions productions, spectral profiles, operability and transient profiles are investigated. This is supported by numerical modelling to describe the underlying chemistry coupled with reaction rate diagrams and sensitivity analysis.

Experimental set-up
An industrial scale tangential swirl burner with optical access and a geometric swirl number of S g = 1.05, shown in Fig. 1, was employed to run different ammonia/hydrogen blends at a wide range of equivalence ratios (Φ) under atmospheric conditions, as shown in Table 1. The burner set-up was detailed elsewhere [34] and only summarised here. The burner was supplied using Bronkhorst mass flow controllers (±0.5% within a range of 15-95% mass flow). Ammonia and air were injected at the mixing chamber, while hydrogen was injected through 6 equispaced holes (1.5 mm diameter), located 4 cm below the burner exit, angled at 45 • , directly into the swirler for premixing with ammonia and air. Table 1 shows the various conditions assessed to determine the impact of ammonia mole fraction on NO X formation, for a wide range of Φ. The details of the data acquisition process and numerical simulation input requirements are illustrated in Fig. 2. The ammonia fraction in the binary fuel of ammonia/hydrogen, X NH3 , was defined by Eq. (1), where [Y] represents mole fraction of species Y. Time-averaged flame images were recorded with two intensified cameras (LaVision Imager intense) running simultaneously, fitted with different bandpass filters chosen to capture NH* (336 nm; A 3 Π-X 2 Σ − system [35]) and NH 2 * (630 nm; single peak of the NH 2 α band [36]) excited radicals. LaVision Davis v10 was used to gather 500 frames for each data point, which were then temporally averaged, and

Fig. 2.
Step-by-step experimental methodology and inputs for numerical simulation. post-processed using a bespoke MatLab script [37] designed to conduct Abel Deconvolution following a 3 × 3 pixel median filter. Colourmaps of the chemiluminescence images have been normalized to the maximum intensity of each image to display the changes in species distributions for each flame. A UV/visible-capable optical fiber head (Stellernet Inc DLENS with F600 fiber optic cable) was installed 3 cm above the burner's exit and 10 cm away from its central axis. The other end of the optical fiber was connected to a UV/visible-capable spectrometer (Stellernet Inc BLUE-Wave) featuring a 100-mm focal length and a 25-μm wide entry slit. The spectrometer was equipped with a 600-grooves/mm grating and a Si-CCD detector (Sony ILX511b) featuring 2048 effective pixels of size 14 × 200 μm 2 , yielding a spectral resolution of 0.5 nm. The detector's exposure time was set to 1 s and 20 scans were averaged to improve the signal-to-noise ratio (SNR). The Y-axis of the spectrometer was calibrated using a standard light source (SL1 Tungsten Halogen).
Temperature profiles were obtained via K and R type thermocouples feeding a data logger with a frequency of 1 Hz. Thermocouple data were taken for 120 s for each point and averaged. The thermocouples were calibrated showing an average error of 3% of reading. The acquired temperature data were corrected as per [38] to account for the convective and radiative heat transfer of the thermocouples with their surroundings, as well as conductive heat transfer between the thermocouple bead and the connecting thermocouple wires. These corrected temperature data for all the test points at location 2, Fig. 1  O) were measured using a bespoke quantum cascade laser analyzer (Emerson CT5100) operating at 463 K with a sampling frequency of 1 Hz (±1%, 0.999 linearity). An isokinetic funnel with an intake diameter of 30 mm was fixed at 50 mm above the quartz confinement's exit to capture homogeneous samples from the exhaust for selected operating conditions. All the emissions data reported here were recorded and averaged over a period of 120 s. All the emissions data presented here are normalized to 15% O 2 [39].

Chemical kinetic modelling
A chemical reactor network (CRN) model, Fig. 3, developed elsewhere [12] was adapted to simulate the chemistry of the experimental flames of Table 1. Inlets were used to provide fuel and air flows as well as the inlet conditions from experimental conditions set in Table 1, with four perfectly stirred reactors (PSR) to model the pre-mixing, flame (FZ), central recirculation (CRZ) and external recirculation zone (ERZ). The recirculation strength was determined by previous experimental campaigns that employed comparable industrial scale swirl burners [12]. The outlet from the flame zone fed a Plug Flow Reactor (PFR) with one dimensional length of 15 cm to simulate reactions in the post-flame zone. The background physics and rationale in the development of this network are detailed in Ref. [37] and summarised in the supplementary material, Section 1. The model was configured with a representative combustor geometry, and averaged residence times were estimated from empirical flow data. Heat losses for PSRs were calculated from the corrected thermocouple data and was fixed for all simulated cases with initial values compared to empirical results at a representative condition (X NH3 = 0.75 and Φ = 1.00) to give favourable agreement with sampled concentrations. All flows/conditions were then changed in an equivalent way to experiments, with no adjustment to other variables. It should be emphasized that the system developed was primarily used for qualitative analysis of the changing flame chemistry using the observed trends. The model employed the reaction mechanism developed by Stagni et al. for NH 3 -H 2 -air mixtures [40], with 31 chemical species and 203 reactions. This mechanism has shown good performance for NH 3 /H 2 /air blends in recent studies [41].
The final emissions from the simulation were collected at the endpoint of the Post-Flame Zone PFR. To evaluate the contribution of elementary reactions to the formation and consumption of N 2 O, absolute rate of production (ROP) values were used for flame zone as 0-D PSR was used to model the flame zone and integrated values of the production rate, I R , were evaluated for PFZ, as defined in Eq. (2), where ω R,i is the rate of production of the species, i, through the reaction, R. The length of the integration domain, L, was set to 15 cm, as per the defined length of PFZ.

Results and discussions
Figs. 4 and 5 illustrate Abel-Deconvoluted NH* and NH 2 * excited radicals, respectively, when changing X NH3 and Φ. Note that the origin corresponds to the burner centreline. Increase in flame brush thickness is observed with increasing X NH3 , while with increasing Φ, flame lifts off gradually due to the balance between flame speed and inlet flow speed. Due to ammonia's narrow flammability limit [13], pure ammonia could not be ignited below 0.8 equivalence ratio. Measured NH* intensities found to be significantly lower than NH 2 * across all flames which is also observed in the acquired chemiluminescence spectra by the spectrometer, Fig. 6. Fig. 6 (a) and (b) show chemiluminescence spectrums for constant X NH3 = 0.75 and Φ = 0.65, respectively. Even though the proportionality between ground state species and electronically excited species is complicated due to various effects such as quenching and radiation losses, previous work [42] has assumed proportionality between ground state and electronically excited species in NH 3 disassociation processes. In this study, a similar positive correlation between ground state and emitting species has been applied, thereby more emitting radicals suggest increased ground state population [43][44][45][46].

NH 3 oxidation
The oxidation process of NH 3 starts by reacting with OH radicals to produce NH 2 . NH 2 then further reduced to NH by reacting with H, O and OH radicals. NH 2 and NH convert to HNO by reacting with O and OH radicals, respectively. HNO is the main source of NO production at the flame zone. Some of this NO converts to N 2 O through the reaction NH + NO ↔ N 2 O + H. Most of this N 2 O produced in the flame zone convert to N 2 by reacting with H radicals. Unburnt NH 3 emissions measured at the exhaust are plotted in Fig. 7 which also shows the prediction by the CRN at X NH3 = 0.75 (blue solid line). Unburnt NH 3 emissions were observed at Φ < 0.60 due to flame instabilities and at rich conditions due to unavailability to oxidise caused by the reduced oxygen content. Emissions between this range are properly captured by the reaction mechanism employed. However, the CRN predictions do not capture NH 3 Fig. 8. However, the reaction involving N 2 O and H radical shows positive sensitivity even though OH radical is being produced. This is due to the fact that the reaction diminishes the availability of H radicals which is vital for OH production through the reaction O 2 + H ↔ O + OH (R2). Interestingly, reaction R1 has its highest absolute ROP values at Φ = 0.8, followed by Φ = 1.2, 1.0 and 0.6 for both flame zones and post-flame zones. This can be attributed to the sensitivity of the reaction R2 at these equivalence ratios, Fig. 8, which follows the exactly similar trend. It was showed in our previous study [48] that reaction R2 is the main source of OH productions in ammonia/hydrogen/air flames. The other significant source of NH 3 oxidation is when NH 3 reacts with H radicals to produce molecular H 2 and NH 2 radicals. This reaction shows its highest ROP at the rich condition due to the availability of H radicals. These results are in line with NH 2 * intensities observed in Fig. 6(a), where increasing NH 2 * production increases with Φ. However, NH* production decreases at rich conditions, Fig. 4, due to reduced availability of OH, H and O radicals. Recombination reactions to produce ammonia display very low ROP, confirming NH 3 emissions captured at rich conditions are mainly unburnt ammonia due to reduced availability of oxidising agents.    3. The solid line shows NO production predictions by the CRN network at X NH3 = 0.75. NO production at the exhaust for certain ranges of lean conditions across different blends could not be captured as the values exceeded the measurement range of the gas analyser. However, the predictions from the CRN network for X NH3 = 0.75 (blue solid line, Fig. 10) give an indication of NO emissions at these ranges, peaking at around Φ = 0.80, a trend already seen in other works [20]. Measured NO emissions drop as X NH3 increases up to Φ > 1.1, point at which NO emissions for the 75/25 VOL.% blend drop nearly to zero. However, the other blends considered here still show some NO emissions until Φ = 1.3, where they all drop to near zero. Fig. 11 displays the normalized sensitivity coefficients for overall NO concentrations at the flame zone while Fig. 12 (a) and (b) show the absolute and integrated ROP of [NO] at the flame and post-flame zones, respectively, at X NH3 = 0.75 and changing Φ. From the ROP graphs, HNO is the main source of NO production at both flame and post flame zones through the reaction HNO + H ↔ NO + H 2 (R3). And from the sensitivity graph in Fig. 10, the major source of HNO production is identified as NH + OH ↔ HNO + H (R4). From Fig. 6(a), OH* and NH* productions were the highest at Φ = 0.8, subsequently producing most NO. Also, from the first row of Figs. 4 and 5, NH* and NH 2 * radicals distributions cover a larger area as Φ increases. As stated earlier, reaction R2 is responsible for producing important radicals which help oxidising ammonia, thus producing fuel NO through the route NH 3 → NH 2 → NH → HNO → NO. Reaction R2 displays negative sensitivity for NO at the rich condition due to unavailability of O 2 . Therefore, from the sensitivity graph in Fig. 11, any reaction consuming OH and H radicals display negative sensitivity for [NO]. Other sources of NO are when atomic N reacts with OH radicals (extended Zeldovich) and through HNO dissociation reaction to NO and H radicals.
Further, NO converts to N 2 O and N 2 by reacting with NH and molecular N, respectively. Also, the ROP of these two reactions are dependent on NO produced through reaction R3. As most NO was predicted to be produced at Φ = 0.8, followed by 1.0, the overall ROPs for NO production are the highest at these two equivalence ratios. NO reduction to N 2 O will be analysed at the latter part of this investigation. Fig. 6(b) shows increased NH 2 * intensity with increasing Φ. NH 2 radicals are well documented for their de-NOxing characteristics through the chain branching reactions NH 2 + NO ↔ NNH + OH (R5) and the chain terminating reaction NH 2 + NO ↔ H 2 O + N 2 (R6) [48][49][50][51]. Also, these two reactions (R5 and R6) contribute mostly towards NO consumption at rich conditions, Fig. 12(b), a point reflected in the captured minimum NO emissions at Φ = 1.2, Fig. 10. Fig. 13 (a) and (b) show the normalized sensitivity coefficients and absolute ROP of [NO], respectively, at the flame zone for Φ = 0.65 and changing X NH3 . With the availability of plenty of excess O 2 at very lean condition, HNO converts to NO by reacting with molecular O 2 (R5). Similarly, another significant source of NO is through the reaction N + O 2 ↔ NO + O (Zeldovich, R6) at this lean condition. From Fig. 6(b), all the radicals (NH, OH and NH 2 ) decrease with increasing ammonia content. H, O and OH radicals production decreases with decreasing H 2  content in the blends, thereby affecting the productions of NH and NH 2 radicals as well, and thus reducing NO formation.

NO 2 emissions
Fig. 14 plots the measured NO emissions at the exhaust for 0.75 ≤ X NH3 ≤ 1.00 and 0.55 ≤ Φ ≤ 1.3. The solid line shows NO production predictions by the CRN network at X NH3 = 0.75. Even though the numerical predictions for NO 2 follow the experimental trends, they underpredict NO 2 by a substantial margin, another point of interest for modellers. Measured NO 2 peaks between 0.7 and 0.8 equivalence ratio for the blends considered here. Steep rise in NO 2 production was observed from the very lean conditions up to the peak production, followed by reduction as Φ increases, and finally dropping to zero at around Φ = 1.2. Fig. 15 displays the normalized sensitivity coefficients for overall NO 2 concentrations at the flame zone while Fig. 16 (a) and (b) show the absolute and integrated ROP of [NO 2 ] at the flame and post-flame zones, respectively, at X NH3 = 0.75 and changing Φ. NO mainly converts to NO 2 through two reactions at the flame zone: NO + HO 2 ↔ NO 2 + OH (R7), and the third body reaction NO + O + M ↔ NO 2 + M (R8). ROPs for both these reactions are substantially higher at Φ = 0.8, compared to other equivalence ratios considered here. This observation can be attributed to the increased availability of NO at Φ = 0.8. However, most of this NO 2 reacts with H and O radicals to revert to NO. From Fig. 16(b), reaction R8 is the most significant NO 2 production route at the post-flame zone. Recent study by Sabia et al. [52] concluded that ammonia can have high third body efficiency which has not been

N 2 O emissions
N 2 O emissions measured at the exhaust for different ammonia/ hydrogen blends and equivalence ratio are plotted in Fig. 17 which also shows the prediction by the CRN at X NH3 = 0.75 (blue solid line). Minimum N 2 O productions were observed at 0.8 > Φ > 1.0 for all the blends considered here. However, N 2 O productions at the lean conditions were more significant than the rich region. Numerical simulations at X NH3 = 0.75 were able to predict the emissions at the lean region but not so much for rich conditions. For the rich region, N 2 O seems to keep increasing for X NH3 = 0.75 and 0.85, whereas N 2 O emissions peak at Φ   duplicated reaction in the mechanism with E A of 4550. All the ROP diagrams with R10 show combined ROP values of the duplicated reactions. Reaction R11 is very sensitive to temperature due to its high activation energy requirements, whereas reaction R9 has the lowest temperature dependency among these three reactions. Fig. 20 shows the temperature dependency of these three reactions in terms of rate constant. At the flame zone, Fig. 19(a), N 2 O formed rapidly at Φ = 0.8, 1.0 and 1.2 in decreasing order via reaction R9. Further, the emission is mostly consumed through reaction R10 due to the presence of H radicals in the flame zone. This observation corresponds to the NH* intensity trend with changing Φ seen in Fig. 6(a). At the post-flame zone, Fig. 19 (b), more N 2 O gets consumed, rather than produced, due to the decreased presence of H radicals across all equivalence ratios considered here. Due to the low reaction rates at Φ = 0.6, it is hard to analyse N 2 O formation/consumption and the impact of other reactions. Therefore, [N 2 O] analysis has been carried out next for three blends (X NH3 = 0.75, 0.85, 0.95) at Φ = 0.65. This slightly higher Φ is chosen due to stable flames where experimental data were taken. From Fig. 6(b), NH* decreases with increasing ammonia content in the fuel mixtures and thus limits the NO conversion to N 2 O via reaction R9. According to Nakamura and Shindo [53], with the increasing ammonia content in the fuel blends, ammonia chemistry plays the dominant role in flame chemistry and becomes more sensitive to heat losses. As reaction R11 is very sensitive to temperature, N 2 O consumption reduces at the flame due to reduced temperature. Chemical time scale increases relative to the fuel residence time with increasing ammonia content in the fuel, thus contributing to reduced N 2 O production as flame changes from 85% to 95% ammonia content. Fig. 21 (a) and (b) show the normalized sensitivity coefficients and absolute ROP of [NO], respectively, at the flame zone, and (c) shows integrated ROP at the post-flame zone for Φ = 0.65 and changing X NH3 . From the sensitivity graph, influence of reaction R10 decreases with increasing X NH3 due to reduced availability of H radicals, whereas sensitivity to reaction R11 increases due to reduced flame temperature at high ammonia contents. From the ROP graphs at the flame and post-flame zones, Fig. 21 (b) and (c), N 2 O production rate via reaction R9 reduces significantly at X NH3 = 0.95, reaffirming the observations earlier.
Relative ROPs for reaction R11 increases compared to reaction R9 at the post-flame zone, thus N 2 O is being reduced considerably at the post-flame zone. Similarly, observed N 2 O emissions at rich conditions can also be attributed to decrease in flame temperature.

Conclusions
Different NH 3 /H 2 blends were analysed at a range of equivalence ratios for a constant thermal power of 8 kW using a generic industrial scale swirl burner with a combination of exhaust emissions measurements, spatially resolved NH* and NH 2 * chemiluminescence, spectrometry analyses and numerical simulations with Chemkin-Pro employing Stagni's mechanism.
Ammonia emissions captured at very lean conditions were attributed to flame instability, whereas unavailability of oxygen at the rich conditions contribute towards ammonia slip. Oxidation of ammonia was found to be dependent on OH radicals availability, thus highly dependent on the reaction O 2 + H ↔ O + OH.
NO emissions increased initially with increasing equivalence ratio, peaking around Φ = 0.8 and then being reduced significantly to nearly zero at rich conditions. HNO was identified as the main contributor of fuel NO formation, whereas NO was found to be reduced to N 2 O, NO 2 and N 2 by reacting with NH, HO 2 and atomic N, respectively. NO production reduces with increasing ammonia mol fraction due to decrease in H, O and OH radicals' availability.
Similar to NO emissions, NO 2 emissions also peak between 0.7 and 0.8 equivalence ratio for the blends considered here. A 75/25 VOL.% NH 3 / H 2 blend was found to be producing considerable quantities of NO 2 , whereas a 95/05 VOL.% NH 3 /H 2 blend had less NO 2 emissions. Most of the NO 2 converts back to NO by reacting with H radicals. N 2 O emissions were the highest at very lean conditions and some emissions were also observed at rich regions. These emissions were attributed to low flame temperatures at those conditions. Considerable amount of N 2 O revert to NO at the flame zone by reacting with H radicals. The third body reaction N 2 O + M ↔ N 2 + O + M was found to be very sensitive to temperature and was identified as the reason for reduced conversion of N 2 O to NO at comparatively lower flame temperatures.
Finally, it was noted that there are still discrepancies between experiments and models, especially when it comes to boundary conditions of stability (very rich and very lean) for species such as ammonia and NO 2 , topic of interest for further research.