Insights into the NO removal mechanism by hydrazine

The combustion of fossil fuels is the main source of nitrogen oxide (NO_x) emissions^[1,2], which have led to serious environmental problems. Recently, ammonia (NH₃) has been proposed and used to mitigate the impact of greenhouse gas (GHG) as a promising carbon-free fuel, which further exacerbates nitrogen oxide emissions^[3]. To control NO_x pollution, selective catalytic reduction^[4,5] (SCR) and selective non-catalytic reduction^[6] (SNCR) stand out prominently for the treatment of flue gas after combustion. Both methods can convert NO_x to harmless N₂ through chemical reactions. Compared with SCR, SNCR does not require the use of expensive catalysts and therefore has a relatively low operation cost.

Since the SNCR of NO by NH₃ was proposed by Lyon in the 1970s^[7], DeNO_x chemistry has been intensively studied over a long period. NH₃ and urea have been widely used as reducing agents^[8−10]. Extensive experimental and kinetic studies have confirmed that NO removal by NH₃ and urea relies mainly on the reactions of NH₂ + NO. However, the traditional SNCR has a relatively high and narrow temperature window (1,100−1,500 K), which limits its application at lower temperatures. Moreover, fluctuations in the oxygen concentration have significant adverse effects on the DeNO_x efficiency and further elevate the temperature window^[9].

Hydrazine (N₂H₄), another reducing agent, is widely used in rocket fuels, antioxidants, and pesticide materials^[11,12]. Many studies have focused on the decomposition of N₂H₄^[13−17], whereas few have reported NO removal by N₂H₄. In 1967, Sawyer & Glassman^[18] first reported the gas-phase reactions of N₂H₄, NO₂, and NO in a flow tube under oxygen conditions. They proposed that the cleavage of the N-N bond is a prerequisite to react with NO, which confirmed the ability of N₂H₄ to remove NO. In 1985, Azuhata et al.^[19] explored the applicability of N₂H₄ in NO_x removal and reported that N₂H₄ can remove NO in the temperature range of 773−873 K. Similarly, Lee & Kim^[20] reported that the optimum reaction temperature for NO removal by N₂H₄ was 873 K, and the best normalized stoichiometric ratio (NSR) was 4.0. Based on the above studies, it is proven that N₂H₄ can lower the temperature window and could be a better reducing agent than ammonia.

However, unlike DeNO_x studies of NH₃, kinetic modeling studies on the NO removal by N₂H₄ are insufficient, which hinders practical applications. Konnov & Ruyck^[21] developed a kinetic model of N₂H₄ to describe its thermal decomposition and flame chemistry. They indicated that the initial decomposition reaction of N₂H₄ and the subsequent reaction of NH₃ + NH₂ had high sensitivities to the overall decomposition rate of N₂H₄ under highly diluted conditions. Later, Hong et al.^[22] reported that N₂H₄ has a four-stage NO removal phenomenon and established a kinetic model to predict it. Moreover, their study revealed that the NO removal efficiency of N₂H₄ can bear an increase in oxygen concentration within a certain range. However, previous models^[21,23] cannot accurately predict the NO removal efficiency, and a detailed explanation of DeNO_x chemistry by N₂H₄ is also inadequate.

In this work, a kinetic model of N₂H₄/NO/O₂ is developed based on recent NH₃ models to predict the experimental data^[23] of N₂H₄/NO/O₂ over 673−1,523 K. Through the kinetic modeling analysis, the four-stage NO removal phenomenon observed in the experiments is thoroughly explained, and the competitive relationship between the NO formation and DeNO_x pathways is revealed. In addition, the effects of the oxygen content on NO removal by N₂H₄ are also discussed in detail.

This study reports a detailed kinetic model for the N₂H₄/NO/O₂ system, which is based on a recent NH₃/H₂ model from Zhu et al.^[24]. Zhu et al. recently developed their model based on a widely used H₂/O₂ model from NUIGMech1.3^[25], and the NH₂ related mechanisms, including pyrolysis, oxidation, and DeNO_x reactions, were evaluated and thoroughly incorporated^[26−28]. In addition, N₂H₄ and NO_x-related reactions were included in their model, which is based on the work of Glarborg et al.^[29] and Sahu et al.^[30], respectively. Their model is applicable for predicting the validation targets at 500−3,200 K and 0.08−75 atm. N₂H₄ can generate both N1 and N2 pathways. Specifically, the N1 pathway involves primarily NH₂ and NH radicals, whereas the N2 pathway includes a variety of N2 radicals (e.g. N₂H₃, N₂H₂, and H₂NN). The present model updated the rate constant of some critical reactions related to those radicals based on previous studies^[21,31−35]. Important reactions discussed later are summarized in Table 1.

Important reactions related to the N2 pathway

The important reactions of the N2 pathway include the unimolecular decomposition reaction of N₂H₄ and bimolecular reactions, such as H-abstraction reactions of N₂H₄ and other N2 radicals. Among them, the rate constant of R1 has been investigated by several groups^{[16,17,21,29,36−39]}. Konnov & Ruyck^[21] estimated the rate constants of N₂H₄ decomposition from updated thermodynamic data and successfully predicted the experimental data of N₂H₄ decomposition. Therefore, the present model adopted the rate constant of R1 from theirs. The H-abstraction reaction of N₂H₄ by OH (R2) was adopted from the kinetic model of Chen et al.^[35] to better predict the DeNO_x process by N₂H₄. Gao et al.^[40] measured the rate constant of R3 through laser photolysis/laser-induced fluorescence (LP-LIF) and computationally investigated the rate constant via quantum chemistry, which was adopted in this model. Additionally, the rate constants of H-abstraction reactions of N₂H₃ by different radicals (i.e. R4−R6) were adopted from Dean & Bozzelli^[41] via direct hydrogen transfer (DHT) estimation. Meanwhile, the isomerization of H₂NN (R7) was also adopted from Dean & Bozzelli^[41] via quantum Rice-Ramsperger-Kassel (QRRK) calculations. The decomposition reaction of H₂NN (R8) was adopted from Hwang & Mebel^[34] via theoretical calculations. The rate constants for other reactions associated with H₂NN, N₂H₃, N₂H₂, and NNH were adopted from previous studies^[10,32,42].

Important reactions related to the DeNO_x process

The reactions of NH₂ and N₂H₂ with NO play crucial roles in NO removal by N₂H₄. Through R9 and R10, NH₂ converts NO to NNH and N₂, respectively, and NNH can further be converted to N₂^[44] (i.e. R11 and R12), effectively leading to NO removal. Song et al.^[43] measured the rate constants of R9 and R10 through shock tube experiments between 1,716 and 2,507 K, and fitted their measurements to cover 300−2,500 K based on previous results at lower temperatures. The final fitted results were adopted in this model to predict the experimental data under wide temperature conditions. Unlike NH₃, N₂H₄ proceeds through N2 pathways to produce N₂H₂, which combines with NO to produce N₂O. R13 was adopted from the evaluated results of Dean & Bozzelli^[41]. N₂O then reacts with H to ultimately produce harmless N₂^[31] (i.e. R14). The above reactions constitute the DeNO_x mechanism in the present model.

Important reactions related to the NO_x formation process

NH₂ consumes NO₂ via the chain-propagation reaction R15 and simultaneously generates NO and H₂NO, which can further be converted to NO via H-loss reactions (i.e. R16−R18). Glarborg^[32] re-evaluated the rate constant of this reaction, which aligns with the theoretical calculation results of Klippenstein et al.^[33], and hence was adopted in this model. Stagni & Cavallotti^[46] investigated a series of H-abstraction reactions from H₂NO (i.e. R16 and R17) via CASPT2 calculations and their calculated results were adopted in this model. The reaction between H₂NN and oxygen to form NO₂ (i.e. R19) was referred to the comprehensive review by Dean & Bozzelli^[41]. The rate constant of R20 was adopted from theoretical calculation results from Klippenstein et al.^[47], which are consistent with experimental measurements^[47,49]. In addition, the NO_x formation pathways also involve the oxidation of NH under high-temperature conditions (i.e. R21 and R22). The rate constant of R21 was adopted from the evaluated value from Elishav et al.^[3].

Simulation method

In this study, the perfect stirred reactor (PSR) module in ANSYS CHEMKIN-PRO^[50] software was used for the simulations. The simulation conditions are consistent with those of Guan et al.^[23]. The detailed simulation conditions are summarized in Table 2. Since NO_x mainly comes from the combustion of hydrocarbon fuels, which simultaneously produces H₂O and CO₂ as major products. CO₂ and H₂O, accompanied with NO_x were adopted to stimulate flue gas components. The complete kinetic mechanism comprises 43 species and 313 reactions, and the kinetic, thermodynamic, and transport files of this model can be found in the supplementary information.

To better interpret the NO removal efficiency by N₂H₄, the simulated mole fractions of NO are converted to the NO removal ratios (η). It is calculated via Eq. (1), where NO_inlet and NO_outlet represent the NO inlet and outlet concentration, respectively. Figure 1 shows the comparison between the measured and predicted NO removal ratios of the present model, as well as those of previous models^{[23,24,32,51,52]}. Among previous models, three of them^[23,24,32] include recent kinetic models of NH₃ oxidation and DeNO_x chemistry by NH₃ or N₂H₄. Guan et al.^[23] conducted experimental and simulation results for NO removal by N₂H₄. Chen et al.^[51] developed a kinetic model to predict the laminar burning velocities of NH₃/O₂/Ar. Glarborg^[32] updated critical reactions in NH₃ ignition and N₂O formation and presented a model for the NH₃/NO₂/O₂ system. Thomas et al.^[52] developed an NH₃ model to predict the extinction of NH₃/H₂ flames and the formation of NO. Zhu et al.^[24] developed a comprehensive NH₃/H₂ kinetic model over a wide range of temperatures and pressures. These literature models are named as Guan 2019, Konnov 2023, Glarborg 2023, Shrestha 2023, and Zhu 2024, respectively. The experiment results of Guan 2019 and the simulation results of the present model show that NO removal by N₂H₄ is dependent on temperature, resulting in a four-stage NO removal phenomenon.

Figure 1. 

NO removal ratios of N₂H₄/NO/O₂ under condition 1. The symbols are experimental data from previous work^[23]. The solid, dash dot dotted, dashed, dotted, short dashed, and dash dotted lines are the predicted results by the present, Guan 2019^[23], Konnov 2023^[51], Glarborg 2023^[32], Shrestha 2023^[52] and Zhu 2024^[24] models, respectively.

Compared with other models, the model proposed in this study demonstrates greater accuracy in capturing this non-monotonous trend, as shown in Fig. 1. Specifically, recently published kinetic models related to ammonia oxidation, such as Konnov 2023^[51], Glarborg 2023^[32], Shrestha 2023^[52], and Zhu 2024^[24], show different trends from experimental data. In the temperature range of 800−973 K, the Konnov 2023 model predicts negative NO removal ratios, which is the opposite of experimental observations. To better present and compare the NO removal efficiency of different models, the NO removal ratios below −15% are not shown in Fig. 1. Compared with the Guan 2019 model^[23], the present model performs better in predicting the peak NO removal efficiency. Additionally, simulations are conducted to predict the experimental data with different oxygen concentrations, as shown in Figs 2 and 3. The results indicate that compared with previous models^{[24,32,51,52]}, the present model can better predict the four-stage NO removal behavior, although it still needs to be improved over 1,100−1,300 K. Therefore, recent comprehensive kinetic models for NH₃ oxidation are not applicable in the investigation of NO removal by N₂H₄, since the N₂H₄ sub-mechanism and its interaction with NO have not been well tested. The uncertainties of the present model may come from both the kinetic parameters and simulation methods. The rate constants of N2-related reactions are mainly from estimations. Future efforts involving experiments and theoretical calculations are highly needed. In addition, the simulations are based on the PSR module, while the experimental process^[23] makes it difficult to achieve the complete mixing state. Therefore, the perfect-stirred assumption may introduce extra uncertainties to the simulation results. Future work on improving the simulation methods is required.

Figure 2. 

NO removal ratios of N₂H₄/NO/O₂ under condition 2. The symbols are experimental data from previous work^[23]. The solid, dashed, dotted, short dashed, and dash dotted lines are the predicted results of the present, Konnov 2023^[51], Glarborg 2023^[32], Shrestha 2023^[52], and Zhu 2024^[24] models, respectively.

Figure 3. 

NO removal ratios of N₂H₄/NO/O₂ at condition 3. The symbols are experimental data from previous work^[23]. The solid, dashed, dotted, short dashed, and dash dotted lines are the predicted results of the present, Konnov 2023^[51], Glarborg 2023^[32], Shrestha 2023^[52], and Zhu 2024^[24] models, respectively.

Reaction kinetics of the four-stage NO removal behavior

The NO removal efficiency is the result of the competition between NO formation and DeNO_x pathways^[53]. To elucidate the mechanism of NO removal by N₂H₄, rate of production (ROP) analyses were conducted in this work. As shown in Fig. 4a, N₂H₄ proceeds via both the N1 and N2 pathways to generate NH₂ and N₂H₃ (yellow box), respectively. These two radicals further participate in both the DeNO_x (green box) and NO formation (red box) pathways. The dashed arrows represent the pathways that are important only under high-temperature conditions. The mole fractions of critical species related to the DeNO_x and NO formation pathways are shown in Fig. 4b & c. Both the experimental and simulation results of the present model reveal that NO removal exhibits a four-stage process. Based on the model analyses, the NO removal mechanism of each stage is analyzed in this section. Detailed reaction pathways of N₂H₄ at four typical temperatures corresponding to each stage are presented in Supplementary Figs S1−S4.

Figure 4. 

(a) Reaction pathways of N₂H₄ under condition 1. The green and red boxes highlight reaction pathways involved in the DeNO_x and NO_x formation processes, respectively. The dashed arrows represent the pathways that are important only under high-temperature conditions. (b), (c) Simulated mole fractions of critical species in the N₂H₄/NO/O₂ under condition 1.

NO removal mechanism in Stage I

Stage I (i.e. 673−848 K) can be considered as a DeNO_x region, since the mole fraction of NO decreases. In this stage, more than 75% of the N₂H₄ is consumed, while N₂H₂ and NH₂ radicals reach their first peak values, as shown in Fig. 4b. To study the NO removal chemistry of this stage, ROP analyses were conducted at 798 K. The results reveal that N₂H₄ is consumed mainly through the unimolecular decomposition reaction R1 to produce NH₂, as well as H-abstraction reactions by OH (i.e. R2) and NH₂ (i.e. R3) to generate N₂H₃. In this stage, the NH₂ radical mainly reacts with NO to produce NNH and N₂ via R9 and R10, respectively. While its reaction with NO₂ to produce NO via R15 is not competitive, since the mole fraction of NO₂ in this stage is very low, as seen in Fig. 4c. Therefore, the NH₂-related pathway mainly results in the DeNO_x-process, which can be confirmed by the sensitivity analysis results shown in Fig. 5. In contrast, N₂H₃ produced via R4 and R5 can be consumed by OH and HO₂ to produce H₂NN and N₂H₂, respectively. Both pathways are important in Stage I. N₂H₂ mainly reacts with NO to proceed the DeNO_x pathways via R13, while H₂NN is almost completely converted to NO₂ via R19 in this stage. Therefore, the N₂H₃-related pathway contributes to both the DeNO_x and NO_x formation process. Combining the NH₂ and N₂H₃ related pathways, the DeNO_x pathway are more favorable than the NO_x formation pathway in Stage I, leading to the increase of NO removal ratios, as seen in Fig. 1. In addition, since most of the NH₂ radicals from N₂H₄ decomposition contribute to the DeNO_x process, while only half of the fuel radicals N₂H₃ proceed DeNO_x pathways via N₂H₂, as seen from Supplementary Fig. S1. DeNO_x reactions related to NH₂ (R9 and R10) present larger negative sensitivity coefficients than those of N₂H₂ (R13) for NO formation in this stage, as seen in Fig. 5. This finding indicates that the N1 pathways via NH₂ are more effective than the N2 pathways in Stage I.

Figure 5. 

Sensitivity analysis of NO at 798, 898, 1,198, and 1,348 K, corresponding to Stage I to IV, under condition 1.

NO removal mechanism in Stage II

As the temperature increases to Stage II (i.e. 848−1,048 K), the mole fraction of N₂H₂ starts to decrease, while those of H₂NN and NO₂ greatly increase (see Fig. 4c). The NO removal ratios decrease in this stage, as seen in Fig. 1. Compared with Stage I, significant changes in the reaction pathways of NH₂ and N₂H₃ radicals occur in this stage. For the NH₂ pathways, the branching fraction of R15 (NH₂ + NO₂ = H₂NO + NO) increases, promoting the formation of NO. Meanwhile, H₂NO can be converted into NO through a series of H-loss reactions (i.e. R16−R18). Therefore, R15 has a strongly positive sensitivity coefficient for NO formation in this stage, indicating that this reaction plays a key role in NO_x formation, as shown in Fig. 5. For the N₂H₃ pathways, it is more inclined to generate H₂NN via R4 rather than N₂H₂ via R5. As mentioned above, N₂H₂ and H₂NN are two functionally distinct radicals. The former tends to participate in the DeNO_x process, while the latter is more inclined to react with oxygen and generate NO_x. Therefore, the reaction produces H₂NN via R4 exhibits positive sensitivity coefficients for NO formation, as seen from Fig. 5. It is surprising that R9 (NH₂ + NO = NNH + OH) presents positive sensitivity coefficient for NO formation in this stage. This is because this reaction produces OH radical, which is critical for converting N₂H₃ to H₂NN via R4 (N₂H₃ + OH = H₂NN + H₂O). Since the flux of R4 increases significantly, the mole fraction of H₂NN dramatically increases, which simultaneously promotes the formation of NO₂ in this stage via R19 (H₂NN + O₂ = NH₂ + NO₂), as shown in Fig. 4c. In summary, in this stage, the reaction sequence N₂H₄ → N₂H₃ → H₂NN → NO₂ → NO and R15 prevail, leading to an increase in the NO_x concentration.

NO removal mechanism in Stage III

In Stage III (i.e. 1,048−1,248 K), N₂H₄ is completely consumed, and an obvious downward trend in the NO content can be observed again. Besides, the mole fractions of H₂NN and NO₂ present decreasing trends, while that of NH₂ starts to increase again as shown in Fig. 4b & c. In Stage II, H₂NN is completely converted to NO₂ via R19. However, in Stage III, the branching fraction of H₂NN = N₂H₂ increases, as shown in Fig. 4a. This change leads to less NO₂ formation (see Fig. 4c), and further reduces NO formation via R15 (NH₂ + NO₂ = H₂NO + NO). Therefore, the flux of H₂NN →NO₂ → NO decreases. Meanwhile, the content of NH₂ increases so that the DeNO_x process via R9 and R10 becomes favorable. Therefore, R1 as the main source of NH₂, shows negative sensitivity coefficient again compared with Stage II, as shown in Fig. 5. On the other hand, this reaction successfully converts H₂NN to N₂H₂, which can effectively promote NO removal via R13 (N₂H₂ + NO = N₂O + NH₂). The sensitivity analysis further demonstrates that R7 (N₂H₂ = H₂NN) has a negative sensitivity coefficient for NO formation and thus promotes NO removal. Therefore, the DeNO_x pathway plays a dominant role again.

NO removal mechanism in Stage IV

In Stage IV (i.e. 1,248−1,523 K), the NH₂ radical tends to consume, while the NH radical greatly increases, as shown in Fig. 4b & c. Under these high-temperature conditions, NH₂ prefers to combine with OH to produce NH, which further reacts with O₂ to generate NO via R20−R22. Therefore, R20 (NH₂ + OH = H₂O + NH) has a strongly positive sensitivity coefficient for NO formation, as shown in Fig. 5. For N2 species shown in Fig. 4c, H₂NN remains and continues to be consumed. Compared with Stage III, the H₂NN tends to produce N₂ via the unimolecular decomposition reaction R8 (H₂NN = N₂ + H₂), as shown in Fig. 4a (the dashed lines). Other pathways of H₂NN, such as the formation of N₂H₂ and NO₂, are less important, indicating that the N2 pathways in this stage have little influence on NO formation. Therefore, in this stage, NO formation is favorable due to the reaction sequence of NH₂ → NH → NO.

Above all, compared with NH₃, N₂H₄ can achieve SNCR of NO over wider temperature windows. The chain-initiation reaction of NH₃ occurs at high temperatures (above 1,100 K) due to the strong N-H bond dissociation energy. However, much weaker N-N and N-H bonds in N₂H₄ make chain-initiation reactions of N₂H₄ easier to occur, resulting in a much lower DeNO_x temperature window (i.e. 673−848 K). In addition, the special N2 pathways in N₂H₄ have two opposite effects: the production of N₂H₂ promotes the DeNO_x process via R13, which is not found important in the NH₃ DeNO_x system; and the production of H₂NN inhibits the DeNO_x process due to the formation of NO₂ via R19. The enhanced formation of NO₂ further promotes NO formation via R15 (NH₂ + NO₂ = H₂NO + NO).

Effects of the oxygen content on NO removal

The increased oxygen content enhances the oxidation reactions by increasing the mole fractions of oxygenated species such as O₂, OH, and HO₂. These radicals play important roles in the removal and formation process of NO^[54]. The effects of the oxygen content on NO removal by N₂H₄ are simulated by the present model, and the results are compared in Fig. 6.

Figure 6. 

Effect of O₂ content on the removal of NO at [NO]_initial = 500 ppm, NSR = 4.0 at 1 atm and τ = 0.3 s. Symbols are experimental data from Guan 2019^[23], while lines are the predicted results by the present model.

Both the experimental and simulation results suggest that the NO removal ratio is not strongly affected by the oxygen content. However, at approximately 1,350 K, the experimental results show that an increase in the oxygen content leads to a significant decrease in the NO removal ratio, while the simulation results indicate that this effect is minimal. This could be attributed to the experimental uncertainties since only one point presents a large discrepancy. Based on model analyses, N₂H₄ initially generates NH₂ radical via its unimolecular decomposition reaction R1 without oxidative pathways, while in the NH₃ DeNO_x system, NH₂ is generated mainly via H-abstraction reactions by OH, O, and HO₂ radicals^[55]. Since NH₂ is the main radical for NO removal in both systems, the N₂H₄ system is less influenced by the oxygen content. Nevertheless, the formation of N₂H₂ is related to oxygenated radicals such as OH and HO₂, so the influence of oxygen content still exists. Above all, compared with NH₃, it is testified that N₂H₄ has a wider range of oxygen fractions for NO removal during the SNCR process. Therefore, another advantage of using N₂H₄ for the SNCR of NO is that its NO removal efficiency is relatively less affected by fluctuations in the oxygen content. This characteristic enables N₂H₄ to play a more stable role in the SNCR process, which fits better with the practical flue gas treatment environment after combustion.

In this work, a kinetic model for NO removal by N₂H₄ was developed. The simulation results of this model were reasonably consistent with the experimental data over the temperature range of 673−1,523 K. Compared with recently published kinetic models, the present model was more accurate in predicting the temperature window and peak values of the NO removal efficiency. It has also been demonstrated that recent NH₃ oxidation models are not accurate enough for predicting NO removal by N₂H₄, although they have been comprehensively validated against wide experimental targets of NH₃. Both the experimental and simulation results of this model present a four-stage NO removal trend by N₂H₄.

Based on the model analysis, it revealed that N₂H₄ can generate N1 and N2 pathways to achieve DeNO_x and the NO formation process. In the N1 pathway, NH₂ can undergo the reactions of DeNO_x and NO formation simultaneously. In the N2 pathway, N₂H₃ produces N₂H₂ and H₂NN. N₂H₂ tends to participate in the DeNO_x process, while H₂NN tends to react with O₂ to generate NO_x. In Stage I (673−848 K), N₂H₄ generates NH₂ and N₂H₂ through N1 and N2 pathways, respectively, both of which can promote NO removal via R9, R10, and R13, with N1 pathway being more effective. Since the chain-initiation reactions in the N₂H₄ DeNO_x system have lower energy barriers than that in NH₃, the DeNO_x process by N₂H₄ occurs at lower temperatures. In Stage II (848−1,048 K), the reaction sequence N₂H₄ → N₂H₃ → H₂NN → NO₂ → NO and R15 (NH₂ + NO₂ = H₂NO + NO) prevailed, increasing the content of NO_x and decreasing the efficiency of NO removal by N₂H₄. When the temperature increased to Stage III (1,048-1,248 K), the flux of H₂NN = N₂H₂ increased, leading to less NO₂ and more N₂H₂ being produced, increasing NO_x removal efficiency again. When the temperature exceeds 1,248 K, i.e. Stage IV, NH₂ prefers to produce NH via H-abstraction reactions, and NH further reacts with O₂ to generate NO. Furthermore, the simulation results suggested that NO removal by N₂H₄ is less affected by the fluctuations of oxygen content. The main reason is that N₂H₄ can be converted to NH₂ directly via the unimolecular decomposition reaction R1, whereas in the NH₃ DeNO_x system, NH₂ is generated mainly via H-abstraction reactions by OH, O and OH₂. These findings are highly important for better understanding the chemistry of NO removal by N₂H₄.