The rate constant of the reaction NCN + H 2 and its role in NCN and NO modeling in low pressure CH 4 /O 2 /N 2 -flames

Bimolecular reactions of the NCN radical play a key role in modeling prompt-NO formation in hydrocarbon flames. The rate constant of the so-far neglected reaction NCN + H 2 has been experimentally determined behind shock waves under pseudo-first order conditions with H 2 as the excess component. NCN 3 thermal decomposition has been used as a quantitative high temperature source of NCN radicals, which have been sensitively detected by diﬀerence UV laser absorption spectroscopy at ~ n = 30383.11 cm (cid:2) 1 . The experiments were performed at two diﬀerent total densities of r E 4.1 (cid:3) 10 (cid:2) 6 mol cm (cid:2) 3 and r E 7.4 (cid:3) 10 (cid:2) 6 mol cm (cid:2) 3 (corresponding to pressures between p = 324 mbar and p = 1665 mbar) and revealed a pressure independent reaction. In the temperature range 1057 K o T o 2475 K, the overall rate constant can be represented by the Arrhenius expression k /(cm 3 mol (cid:2) 1 s (cid:2) 1 ) = 4.1 (cid:3) 10 13 exp( (cid:2) 101 kJ mol (cid:2) 1 / RT ) ( D log k = (cid:4) 0.11). The pressure independent reaction as well as the measured activation energy is consistent with a dominating H abstracting reaction channel yielding the products HNCN + H. The reaction NCN + H 2 has been implemented together with a set of reactions for subsequent HNCN and HNC chemistry into the detailed GDFkin3.0_NCN mechanism for NO x flame modeling. Two fuel-rich low-pressure CH 4 /O 2 /N 2 -flames served as examples to quantify the impact of the additional chemical pathways. Although the overall NCN consumption by H 2 remains small, significant diﬀerences have been observed for NO yields with the updated mechanism. A detailed flux analysis revealed that HNC, mainly arising from HCN/HNC isomerization, plays a decisive role and enhances NO formation through a new HNC - HNCO - NH 2 - NH - NO pathway.


Introduction
Nitrogen oxides (NO x ) are atmospheric pollutants formed as byproducts in combustion processes. Especially under fuel rich conditions, NO is favorably formed over the so-called prompt-NO pathway, which is initiated by the reaction of small hydrocarbon radicals with molecular nitrogen from the combustion air. According to the traditional Fenimore mechanism, 1 it has been assumed for a long time that prompt-NO formation is mainly initiated by the reaction CH( 2 P) + N 2 ( 1 S + ) -N( 4 S) + HCN( 1 S + ). (1a) Even though the formation of the products N + HCN is spinforbidden and despite the fact that theoretical estimates 2,3 for the rate constant k 1a turned out to be inconsistent with the experimentally determined high temperature rate constants for the overall CH + N 2 reaction, 4,5 reaction (1a) is still used in some flame modeling studies. Already in the year 2000, based on quantum-chemical and RRKM calculations, Moskaleva et al. 6 suggested the alternative spin-conserved reaction channel CH( 2 P) + N 2 ( 1 S + ) -H( 2 S) + NCN( 3 S g À ) (1b) and predicted NCN + H to be the main product of the reaction. In the meantime it has been experimentally proven that NCN radicals are in fact formed in flames [7][8][9] and that NCN is the main product of the reaction CH + N 2 . 10 Consequently, flame mechanisms for NO modeling have been updated with regard to NCN high temperature chemistry. Current versions of NCN submechanisms are the Konnov0.6 11 and the GDFkin3.0_NCN 12,13 mechanisms. The performance of the latter has been extensively validated and improved over the last few years by quantitative measurements and modeling of CH, NO and NCN concentration profiles in low-pressure CH 4 /O 2 /N 2 and C 2 H 2 /O 2 /N 2 flames of various fuel/air equivalent ratios. 12,14 Recently, corresponding NCO, CN and HCN profiles have been measured as well. 13 Whereas early versions of NCN submechanisms relied on rate constant estimates of Glarborg et al., 15 the more recent implementations are based on extensive rate constant data from the theoretical work performed by the M.C. Lin group. They reported RRKM and TST studies for the most important bimolecular NCN reactions NCN + O 2 , 16 O, 17 OH 18 and H 19 as well as for the NCN forming reaction CH + N 2 . 6,19 From the experimental point of view, most direct rate constant measurements of NCN reactions have been accomplished over the past few years in the shock tube laboratory of the authors (N.F. and G.F.). Beside our studies on NCN + O, H, M, NCN, NO, and NO 2 , [20][21][22] only two other shock tube studies have been performed. Vasudevan et al. 10 measured NCN absorption profiles during the pyrolysis of ethane/N 2 mixtures and Busch et al. 23,24 investigated the unimolecular decomposition reaction NCN + M -C + N 2 + M by C atom resonance absorption spectroscopy (ARAS). Ongoing work of one of the authors (N.L.) is concerned with the implementation, testing, and validation of the expandingeven though not yet complete -experimental database on NCN chemistry for flame modeling. Here, in order to assess the potential influence of the reaction NCN + H 2 on NO x flame modeling, we rely on the GDFkin3.0_NCN submechanism as a well-validated starting point. Surprisingly, despite rather high H 2 concentrations in the flame front, the title reaction NCN + H 2products (2) has not yet been implemented into flame mechanisms. Seemingly it was considered to be rather slow and therefore dispensable. However, neither experimental nor theoretical studies have been performed so far to confirm or falsify this assumption. In contrast, the rather slow NCN + O 2 reaction is included both in the Konnov0.6 and the GDFkin3.0_NCN mechanisms. Starting from early and too high rate constant estimates, 15 this reaction was initially believed to be crucial for NCN modeling, 25 but later theoretical calculations of Zhu and Lin 16 clearly showed that the reaction is activation controlled, slow, and therefore plays a less important role in NO x formation in flames. As it turns out below, the rate constant of the reaction NCN + H 2 is about two orders of magnitude higher than for NCN + O 2 . Therefore, an accurate rate constant determination and a thorough analysis of its impact for NO x modeling as reported in this work is overdue.

Experimental
The used shock tube apparatus has been described in more detail elsewhere. 26 Briefly, the experiments have been performed in an overall 8.3 m long stainless steel shock tube with a 4.4 m long electropolished test section that could be evacuated to pressures of p E 10 À7 mbar by a combination of an oilfree turbo-molecular drag and a diaphragm pump. The test section and the driver section were separated by aluminum membranes of 30, 80 or 100 mm thickness. Hydrogen or hydrogen/nitrogen gas mixtures have been used as driver gas. The experimental conditions behind the incident and reflected shock waves were calculated from pre-shock conditions and the shock wave velocity, which was measured using four fast piezoelectronic sensors (PCB Piezotronics M113A21), by using a frozen-chemistry code. Storage gas mixtures of 500 ppm to 1000 ppm NCN 3 in argon were prepared using the partial pressure method. The reaction mixtures were prepared using calibrated mass flow controllers (Aera, FC-7700CU; 10, 50, and 1000 sccm). Pure H 2 (Air Liquide, Z99.999%) and the NCN 3 mixtures were further diluted with argon (Air Liquide, Z99.999%) in a flow system and were passed into the test section of the shock tube. The section was flushed with the test gas mixture for about 5 min to minimize possible gas adsorption effects on the shock tube walls. High H 2 mole fractions of up to 9.6% were necessary in the reaction gas mixtures to achieve a fast NCN consumption due to reaction (2). At such high mole fractions, vibrational relaxation effects may compromise a simple frozen-chemistry calculation of the experimental conditions. Equilibration of the Boltzmann population of H 2 in its n = 0 and n = 1 vibrational states may cause a noticeable, gradual decrease of the temperature behind the shock wave. However, on the one hand the vibrational relaxation time of H 2 in argon is known to be rather short, about 9 ms at p = 700 bar and T = 1750 K, 27 and hence close to the time resolution of the experiments. On the other hand, due to the high vibrational quanta of H 2 , the fraction of H 2 in the n = 1 state and with it the overall temperature effect remains small. At a typical experimental temperature of T = 1750 K, the ratio H 2 (n = 1)/H 2 (n = 0) = 0.03. Calculation of shock wave conditions assuming fully and non-relaxed H 2 showed that even at the highest experimental temperatures and H 2 concentrations used in this work, the maximum expected temperature effect was DT o 5 K. This is within the 1% uncertainty of the temperature calculation from the shock wave velocity. Consequently, due to the fast relaxation and the overall small temperature effect, it could be safely assumed that H 2 relaxation did not interfere with the rate constant measurements.

NCN precursor
It has been shown by Dammeier et al. 28,29 that the thermal decomposition of cyanogen azide (NCN 3 ) serves as a quantitative source of NCN radicals. NCN 3 thermal decomposition yields NCN in its first electronically excited singlet state, which is rapidly converted to the triplet ground state by collision induced intersystem crossing (CIISC).
Rate constants for reaction (3) and (4) have been adopted from previous work and are listed in Table 1. As it is known that the CIISC process (4) strongly depends on the collision partner and hence reaction mixture composition, its rate constant was allowed to vary within the error limits reported in ref. 29. The extremely explosive and toxic precursor NCN 3 cannot be purified by freeze-pump cycles. It has therefore been synthesized directly in high purity in the gas phase, according to BrCN(g) + NaN 3 (s) -NCN 3 (g) + NaBr(s), using a method described in detail in ref. 22. After an 8 h reaction time the remaining BrCN impurities were usually o3% according to FTIR analysis. The pure NCN 3 was diluted in argon and was used within three days since NCN 3 tends to slowly form solid polymers. The actual initial NCN 3 concentrations in the reaction gas mixtures were determined from the NCN absorption signal plateaus behind the incident shock waves (for experiments behind the reflected shock wave) or by fitting the maximum of the NCN concentration profile (for experiments behind the incident shock wave). In all cases, the determined concentration was consistent with the concentration calculated from the expected NCN 3 mole fraction in the storage gas mixture.

NCN detection scheme
The narrow-bandwidth laser absorption setup for time-resolved radical detection behind shock waves has been described in detail elsewhere. 22 Briefly, about 1 mW of UV radiation was generated by intra-cavity frequency doubling of a frequencystabilized continuous-wave ring-dye laser (Coherent 899) operated with DCM-special as dye and pumped using 8 W at l = 532 nm from a Nd:YVO 4 solid state laser (Coherent Verdi V10). The UV laser beam was split into a detection and a reference beam using a 50 : 50 beam splitter plate. The detection beam was focused, passed the shock tube through two quartz windows, and was coupled into an optical fiber connected to a balance photo-detector and amplifier (Thorlabs PDB 150A-EC). The reference beam intensity could be precisely adjusted using a variable neutral density filter to match the intensity of the detection beam. The resulting difference signal (DI) and the monitor signal of the detection beam (I 0 ) were stored by an analog input board (Measurement Computing PCI-DAS4020/12, 12 bit, 20 MHz) for further data processing.
Triplet NCN has been detected atñ = 30381.11 cm À1 (l = 329.1302 nm) on the maximum of an absorption band stemming from the superposition of the 3 P 1 sub-band of the Ã 3 P u (000)-X 3 S g (000) transition with the Q 1 band head of the vibronic 3 S + (010)-3 P(010) transition. 30 NCN concentrationtime profiles were calculated using the previously reported temperature-dependent absorption cross section, which has been measured with an accuracy of AE25% using the same apparatus at similar temperatures and pressures as used in this work. 28 As the (010) vibrational state becomes significantly populated at combustion temperatures, the absorption cross sections were comparatively high (e.g., s(1500 K, base e) = 4.5 Â 10 7 cm 2 mol À1 ). Hence, with a detection limit of 4 Â 10 À12 cm 3 mol À1 (corresponding to 1.5 Â 10 À3 absorption at T = 1500 K, p = 500 mbar, and an electronic time-resolution of Dt E 1 ms), NCN could be detected with high signal-to-noise ratios even at NCN mole fractions as low as a few ppm.

Numerical methods
Numerical simulations of NCN concentration-time profiles from the shock tube experiments were performed using the Chemkin-II program package 31 in combination with GDFkin3.0_NCN as a detailed background mechanism. In order to be consistent with a previous shock tube work, rate constants for NCN reactions have been replaced or added according to our directly measured rate constant data set. [20][21][22] The most important reactions for modeling the shock tube experiments are listed in Table 1. For sensitivity analysis, the sensitivity coefficient s(i,t) for reaction i at time t was normalized with respect to the maximum concentration [NCN] max over the time history, s(i,t) = 1/[NCN] max Â (q[NCN]/q ln k i ).
Flame modeling was performed with the Chemkin/Premix code 31,32 and the detailed mechanism GDFkin3.0_NCN. 12 As will be further outlined below, the mechanism has been modified to take into account reaction (2) and subsequent chemistry of HNCN and HNC. Rate-of-production (ROP) and the N-atom flux analyses have been accomplished at the NCN peak locations using a homemade post processor that relies mostly on the Chemkin subroutines. 31 Atom flux analysis has been performed with the program Kinalc, and the reaction fluxes were plotted using the included FluxViewer visualisation tool. 33 Thermodynamic data were adopted from GDFkin3.0 12 with updated NCN thermochemistry as described in ref. 13. Hence, in agreement with recent experimental work, 22 the controversial value of the enthalpy of formation of NCN 34 was set to D f H 298K ¼ 450:2 kJ mol À1 . For HNCN and HNC, thermodynamic data were extracted from the Goos/Ruscic database. 35 Quantum chemical calculations were carried out using the Gaussian 09 suite of programs. 36 The transition state of the reaction NCN + H 2 -HNCN + H was located and verified by Table 1 Arrhenius parameters for the rate constants of all the included NCN reactions as used for simulating the experimental NCN profiles behind shock waves. k i = AT n exp[ÀE a /RT], given in units of cm 3 , mol À1 , s À1 and kJ. Except for reactions (3) and (4), all reactions have been duplicated for 1 NCN to take 1 NCN chemistry approximately into account. In addition to the listed reactions, the GDFkin3.0 mechanism 12 was used as a background mechanism using a synchronous transit-guided quasi-Newton method (QST3 option) and an intrinsic reaction coordinate as follows.

Results and discussion
Shock tube experiments The rate constant of reaction (2), NCN + H 2 , has been measured behind incident and reflected shock waves with reaction gas mixtures containing 3-27 ppm NCN 3 and 0.8-9.6% H 2 in argon. k 2 values have been obtained in the temperature and pressure ranges of 1057 K r T r 2475 K and 324 mbar r p r 1665 mbar, at two total densities of r E 4.1 Â 10 À6 and 7.4 Â 10 À6 mol cm À3 . Under these experimental conditions the reaction NCN + H 2 was always the most important reaction for NCN consumption. Experimental temperature limits were set by the thermal decomposition of NCN, which becomes the dominant reaction above 2500 K, and by the rate of reaction (2) itself, which becomes too slow at temperatures T o 1000 K to be measured with sufficient sensitivity. Fig. 1a shows a typical NCN concentration-time profile behind the incident shock wave at T = 1582 K and a total density of r = 4.13 Â 10 À6 mol cm À3 . The NCN profile reveals a rather slow NCN consumption with a half-life of t 1/2 E 200 ms.
Since there are no studies on the possible reaction channels of the reaction NCN + H 2 , numerical simulations of the experimental NCN profiles have been performed assuming different sets of reaction products. Potential reaction products of reaction (2) include: The formation of the thermodynamically most favorable products H 2 NCN (2a) and HNCNH (2b) is spin-forbidden and hence their formation constitutes a presumably unimportant pathway. All other increasingly endothermic channels are spinallowed and may become accessible at combustion temperatures. Except for reaction (2e), the assumed product sets require the formation of an unlikely collision complex on the triplet surface followed by several rearrangement steps. This is in particular the case for reaction (2c), where a feasible reaction pathway can hardly be imagined. Actually, reactions (2a) and (2g) may become important for the corresponding 1 NCN reaction as singlet radicals are known to prefer insertion reactions. For the triplet radical, however, reaction (2e) constitutes the by far most probable reaction channel. Next to the decomposition of a 3 HNCNH intermediate, this reaction can take place as a direct activation controlled H abstraction reaction as well. It is known that H abstraction channels often become the dominating pathway at high temperatures even if complex-forming pathways are accessible. 26,37 Therefore, the most reasonable reaction products HNCN + H have been assumed for the target reaction NCN + H 2 in the first round of data evaluation. The effect of assuming different product sets will be further discussed below. The experimental NCN concentration-time profiles have been simulated based on a detailed mechanism assembled from our previous work (Table 1). Except for reactions (3) and (4), all reactions have been duplicated for 1 NCN to take 1 NCN chemistry approximately into account. Moreover, the reaction model was complemented by the extensive GDFkin3.0 mechanism. 12,13 Subsequent HNCN chemistry is only partly accounted for. Next to the reverse of reaction (2e), H + HNCN, only the thermal decomposition of HNCN has been included in the mechanism by the reverse of reaction (11). For the latter, the rate expression used for atmospheric pressure has been adopted from the recent work of Teng et al.; 19 a formerly reported rate expression for the low pressure limit of the unimolecular decomposition reaction (À11) by Moskaleva and Lin, 6 which is implemented into the Konnov0.6 mechanism, turned out to yield unrealistic high rate constant values. Other rate constant data for bimolecular HNCN loss reactions such as HNCN + C/CN/N are not available in the literature, however, these reactions are not expected to play significant roles for modeling the shock tube experiments.
In Fig. 1a, the solid curve represents the best numerical simulation of the NCN profile using k 2e = 1.8 Â 10 10 cm 3 mol À1 s À1 . Variation of k 2e by a factor of two yields the two dotted curves that fail to reproduce the experiment. The sensitivity analysis in Fig. 1b reveals that reaction (2e) is by far the most important reaction for NCN consumption. Only at reaction times t 4 250 ms, the reaction NCN + H gains some influence. Its high temperature rate constant and branching ratio, however, have been directly measured recently. 22 The sole other sensitive   reaction, which is important to model the initial NCN formation at short reaction times, is the CIISC process (4). It has been studied in some detail by Dammeier et al. 29 and its rate constant value is dependent on the composition of the reaction gas mixture. Here, k 4 has been used as an adjustable parameter and was varied within the error limits reported in ref. 29 in order to model the initial increase of the NCN signal. Finally, an alternative rate constant determination by fitting the NCN decay assuming a simple pseudo-first order exponential decay without applying any reaction mechanism has been performed as well. As expected for negligible secondary chemistry, very similar rate constant values are obtained. For example, for the experiment shown in Fig. 1 a rate constant of k 2 = 1.9 Â 10 10 cm 3 mol À1 s À1 has been obtained over the interval 25 ms r t r 250 ms, which is very close to the k 2e = 1.8 Â 10 10 cm 3 mol À1 s À1 value from the numerical simulation mentioned above. In order to analyze a possible influence of the assumed products of reaction (2), the experimental NCN concentrationtime profiles have been simulated using different product sets. Fig. 2a illustrates an experimental NCN profile at a reflected shock wave temperature of T = 2123 K. Assuming reaction channel (2e) with the radical products HNCN + H yields a rate constant of k 2e = 1.3 Â 10 11 cm 3 mol À1 s À1 (red curve). The sensitivity analysis of the chosen high temperature experiment reveals that the influence of secondary chemistry is more pronounced than for the previously discussed T = 1582 K experiment (Fig. 1). For example, the thermal decomposition of NCN, NCN + M -C + N 2 + M starts to play a significant role. Its rate constant has been measured in two independent studies and highly consistent values have been reported. 21,24 Hence, a sensitive determination of k 2 is still possible. Next, the same value for the rate constant k 2 but the alternative recombination product HNCNH of channel (2b) has been used instead of the products HNCN + H. The resulting simulated curve predicts a somewhat too slow NCN decay (black curve). As no additional HNCNH chemistry has been included into the mechanism and hence HNCNH has been treated as a stable species, the difference between the two simulations reflects the impact of the secondary reactions resulting from the radical products formed in the case of channel (2e) products. Similar results are obtained when assuming reaction channel (2a). In contrast, simulations with the other potential radical forming reaction channels (2c), (2d), (2f) or (2g) yielded more or less the same k 2 values as for channel (2e). In Fig. 2a, the blue curve represents the simulation using the products of channel (2f), HNC + NH, as an example. Within error limits it is identical to the simulation using channel (2e).
Total rate constant values for reaction (2) have been extracted from 36 shock tube experiments. The experimental conditions of all the experiments are listed in Table 2, an Arrhenius plot of the obtained k 2 values is given in Fig. 3. The symbols correspond to the results assuming HNCN + H as the reaction products. Within the scattering, the obtained data for the two different total densities of r E 4.1 Â 10 À6 mol cm À3 (open squares) and 7.4 Â 10 À6 mol cm À3 (star symbols) agree, showing that the reaction is not significantly pressure dependent.
The data points can be nicely represented by a two-parameter Arrhenius expression. The error bars in Fig. 3 exemplify the cumulative uncertainty of k 2 resulting from different error sources. Simulations show that a pessimistic AE25% error estimate for the initial NCN 3 concentration, which arises from the 25% error of the used NCN absorption cross section, 28 results in a AE6% uncertainty in k 2 . Varying the most important background reactions within their error limits sum up to AE8%. Finally, the uncertainty of the H 2 concentration was AE2% and the statistical error of the Arrhenius fit of the scattered data amounts to AE8% (2s standard error of the mean). A reasonable total error estimate in the middle of our temperature range at T E 1750 K is therefore AE24%, corresponding to Dlog k 2 = AE0.11. The dashed line in Fig. 3 corresponds to the Arrhenius expression k 2 /(cm 3 mol À1 s À1 ) = 6.5 Â 10 13 Â exp(À105 kJ mol À1 /RT) that has been obtained by assuming the unlikely formation of a stable reaction product (i.e., channel (2a) or (2b)). Such an evaluation yields data (not shown) that start to deviate from the evaluation assuming radical products (i.e., channels (2c)-(g)) at temperatures T 4 1700 K with a maximum deviation of +30% at T = 2480 K.
A comparison with a similar reaction and quantumchemical calculations show that the obtained activation energy is roughly consistent with a reaction that is dominated by an H abstraction pathway according to channel (2e). On the one hand, a comparable H abstraction reaction of the N-centered triplet species, 3 NH, NH + H 2 -NH 2 + H, exhibits an activation energy, which is 33 kJ mol À1 higher than the corresponding reaction enthalpy of D r H 298K ¼ 32 kJ mol À1 . 38 Accordingly, an activation energy of E a E 84 kJ mol À1 + 33 kJ mol À1 = 117 kJ mol À1 would be expected for reaction (2e). On the other hand, an estimate of the activation enthalpy of channel (2e) based on quantum-chemical calculations using G4 level of theory yields a similar result. This method provides a reaction enthalpy for channel (2e) of D r H 298K ¼ 86 kJ mol À1 , which is close to the value of 84 kJ mol À1 from thermodynamic data taken from the literature. The calculated energy of the H 2 Á Á ÁNCN transition state (H 2 is bonded to one of the N atoms and is oriented essentially perpendicular to the slightly bended NCN moiety) yields DH ‡ (T = 1750 K) = 101 kJ mol À1 . Accordingly, taking into account the simple transition state theory expression E a E DH ‡ + 2RT, an activation energy of about 130 kJ mol À1 can be estimated for channel (2e) at T = 1750 K. However, a more detailed comparison with theory should be based on more advanced multireference quantum-chemical and kinetic calculations including tunneling corrections as well as a complete RRKM/master equation analysis of the possible role of additional complex-forming reaction pathways. Given that accurate energy calculation of NCN related species turned out to be very challenging, 34 such an analysis would have been beyond the scope of this paper.

Flame modeling
The reaction NCN + H 2 turns out to be comparatively fast. With a rate constant of k 2 = 3.2 Â 10 10 cm 3 mol À1 s À1 at a typical flame temperature of T = 1700 K, it is about a factor of 300 faster than the reaction NCN + O 2 . Therefore, in order to assess the potential influence of reaction (2) on NO x formation in flames, reaction (2) as well as other NCN reactions have been implemented into the GDFkin3.0_NCN 12,13 flame mechanism. Arrhenius parameters of all reactions that have been added to the original mechanism are listed in Table 3. Simulations have been performed with HNCN + H, HCN + NH, and HNC + NH as the respective sole products of reaction (2).
Consideration of HNCN + H as main products implies an update of the mechanism with respect to HNCN reactions as well. HNCN species may rapidly react with O atoms (reactions (12)) generating HNC species. Consequently, two new blocks of reactions have been added to account for possible HNCN and HNC chemistry. In order to be coherent with the experimental rate constant determination, some additional NCN consumption reactions, (6) to (11), have also been considered in the detailed mechanism. In the following, this updated mechanism  Fig. 3 Arrhenius plot of the rate constant for the reaction NCN + H 2products. Experimental data at two different total densities r E 4.1 Â 10 À6 mol cm À3 (open squares) and r E 7.4 Â 10 À6 mol cm À3 (star symbols) and corresponding Arrhenius fit (solid line) for an evaluation assuming radical products (channel (2c)-(g)) are shown. The dashed line depicts an Arrhenius fit obtained for an alternative data evaluation (corresponding data points are not shown) assuming stable reaction products (channels (2a) and (2b)).
is named up-GDFkin3.0_NCN (up-GDF for short) in contrast to the original mechanism GDFkin3.0_NCN (GDF for short).
Calculations were performed to simulate species profiles in selected low pressure premixed flames where the reaction NCN + H 2 may play an important role in the prompt-NO pathway. Two fuel rich flames of CH 4 /O 2 /N 2 have been considered numerically. Both flames were simulated at low pressure (5.0 kPa) with the same total volumetric flow rate (300 L h À1 , under the standard conditions of temperature and pressure) and nitrogen dilution ratio (60%), but a different richness equal to f = 1.3 and f = 1.5, respectively. Imposed temperature profiles were identical for each flame. The temperature in the burned gas was limited to remain lower than 1850 K, hence thermal-NO contribution was reduced and prompt-NO formation was promoted.
Considering the original GDF mechanism, simulated temperature and species profiles of NO, NCN and H 2 are reported in Fig. 4. In the burned gases, NO mole fractions are equal to 24.5 and 38.3 ppm for f = 1.3 and f = 1.5, respectively. As shown in Fig. 4a, the NO mole fraction in the burned gases is hardly affected at f = 1.3, but is increased by 8% at f = 1.5 when the calculations are performed with the up-GDF mechanism instead. H 2 and NCN profiles are reported in Fig. 4b using the up-GDF mechanism. The profiles show that at the NCN peak location (height above burner, HAB (f = 1.  Table 4. Only NCN losses in the direction of the prompt-NO formation pathway (forward flux) are included. As expected, the reaction NCN + H -HCN + N is the most important NCN loss reaction that dominates the NCN forward flux, followed by the reaction with O atoms (yielding CN + NO). Nevertheless, about 1.8% of the NCN radicals are consumed through reaction (2) at  f = 1.3 and 3.5% at f = 1.5. The so-far neglected reaction with H 2 ranks third place and its contribution is even three orders of magnitude higher than for the reaction NCN + O 2 . Note that the N-flux analysis reveals that next to reaction (2) also reaction (8), NCN + C, which was omitted in the original mechanism as well, becomes more important than other, already implemented bimolecular reactions such as NCN + OH/HO 2 /NO/M that are negligible under the assumed flame conditions. Although it turned out that the implementation of reaction (2) is important, its rather low contribution to the forward flux is in seemingly contrast to the mentioned significant change in the overall NO yield. Obviously, other reactions added to the updated mechanism must be responsible for this prominent effect. A complete reaction pathway diagram for the f = 1.5 flame using the up-GDF mechanism is shown in Fig. 5. New reaction pathways that are absent in the corresponding diagram using the original GDF mechanism (not shown) are highlighted in red color. These pathways include the formation and loss reactions of the newly included species HNCN and HNC. Once formed, HNCN reacts quickly with O-atoms through reaction (12). According to theoretical calculations, the products of the reaction HNCN + O are mostly HNC + NO. 39 HNC then reacts with OH radicals yielding HNCO + H, and HNCO reacts with H atoms yielding NH 2 + CO. However, it becomes clear from the flux diagram that HNC radicals are formed primarily from HCN and only secondly from HNCN. In fact, the HCN/HNC isomerization according to reactions (15), which is close to its lowpressure limit, and the H atom initiated isomerization reaction (À16), HNC + H -HCN + H, represent 12% and 20% of the HCN consumption at f = 1.3 and f = 1.5, respectively. This finding is also reflected in the HCN profiles shown in Fig. 6 for the f = 1.5 flame. HCN mole fraction peak values obtained with the updated mechanism decrease by 20%. According to the ROP, HCN is mainly consumed through the reaction HCN + O " NCO + H and reaction (À16), both being three times more important than the reactions HCN + OH " HOCN + H and HNC (+M) " HCN (+M). The combined effect of HNC formation from reactions (À16) and (12a) results in a substantial increase of HNCO by a factor of two (Fig. 6). This increase is followed by a comparable increase of the peak values of NH 2 as well.
To sum up, although the implementation of reaction (2) has indeed a significant impact on NO formation through a new HNC -HNCO -NH 2 -NH -NO pathway, it is in fact the HCN/HNC isomerization and not the reaction NCN + H 2 that is mainly responsible for the distinct changes obtained with the up-GDF mechanism. Note that the rate constant of the H initiated isomerization reaction (16), H + HNC -HCN + H, is based on a theoretical QRRK estimate 42 and, to the best of our knowledge, has not been experimentally confirmed yet. Therefore, our interesting preliminary finding calls for a more detailed analysis to better constrain and verify the role of the HNC initiated pathway for NO formation in flames.
The possible influence of the assumed products of reaction (2) for the NO formation in flames has also been analyzed. Whatever channel (d-f) of reaction (2) was assumed, simulated NO profiles are identical to each other. However, considering channel (2d), implementation of reactions (12)(13)(14)(15)(16)(17)(18)(19) could be removed since HCN and NH species are already declared in GDF. 13 In that case, simulated NO profiles would be identical to those obtained with the original GDF mechanism. Again, this clearly highlights the potential importance of the HCN/HNC isomerization pathway through reactions (15) and (À16).

Conclusion
The rate constant of reaction (2), NCN + H 2products, has been measured for the first time. Shock wave experiments with time-resolved NCN radical detection by narrow-bandwidth laser UV absorption were carried out under nearly pseudo-first order reaction conditions with H 2 as the excess component. The total rate constant has been measured at temperatures 1057 K r T r 2475 K and can be represented by the Arrhenius expression k 2 = cm 3 mol À1 s À1 À Á ¼ 4:1 Â 10 13 exp À 101 kJ mol À1 RT ; D log k 2 ¼ AE0:11: No pressure dependence could be observed between p = 324 mbar and p = 1665 mbar. Quantum-chemical calculations show that the activation energy is roughly consistent with the formation of the products HNCN + H, hence the most likely direct abstraction pathway. Still, detailed quantum-chemical calculations in combination with RRKM/TST/ME modeling are desirable to further assess the role of complex-forming reaction pathways.
The detailed reaction mechanism GDFkin3.0_NCN has been updated to assess the potential influence of the so-far neglected reaction (2) on prompt-NO formation in flames. Next to reaction (2), HNC and HNCN submechanisms have been implemented as well to describe the subsequent chemistry of the reaction products. Two fuel-rich low-pressure methane flames served as a model case and it was shown that the contribution of reaction (2) on the overall NCN loss in the direction of prompt-NO is on the order of a few percent. Of course, the reaction may become even more important for other flame conditions and, therefore, needs to be included in detailed flame mechanisms. The flame simulation also highlights the fact that the reaction NCN + C " CN + CN, which was also omitted in the original GDFkin3.0_NCN mechanism should be considered in the future as well. As yet the rate constant for NCN + C has only been roughly determined experimentally, improved direct measurements are required. Moreover, as a pronounced impact of HCN/HNC isomerization on NO formation has been found in this study, a critical assessment and experimental verification of the role of this new HNC -HNCO -NH 2 -NH -NO prompt-NO formation pathway is important. A thorough validation of the updated mechanism going along with the implementation of other new experimental rate constant data that recently have become available for several bimolecular reactions of NCN is currently underway.