Combustion and Flame 185 (2017) 16–27

Contents lists available at ScienceDirect

Combustion and Flame

journal homepage: www.elsevier.com/locate/combustflame

Kinetic modeling of ammonia/air weak flames in a micro flow reactor

with a controlled temperature profile
Hisashi Nakamura a,b,∗, Susumu Hasegawa a, Takuya Tezuka a
a
Institute of Fluid Science, Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai 980-8577, Japan
b
Japan Science and Technology Agency, PRESTO, 4-1-8 Honcho, Kawaguchi, Saitama 332-0012, Japan

a r t i c l e i n f o a b s t r a c t

Article history: Ammonia is considered to be one of the promising energy carriers in the future and reliable chemical
Received 13 March 2017 kinetics to accurately predict ignition characteristics of ammonia/air mixtures is necessary for develop-
Revised 10 May 2017
ing ammonia combustors. However, ignition characteristics of ammonia/air mixtures at low temperatures
Accepted 29 June 2017
have not been well studied. The present study employed weak flames in a micro flow reactor with a con-
Available online 17 July 2017
trolled temperature profile, which have been extensively employed to examine ignition characteristics of
Keywords: hydrocarbons, to investigate ignition characteristics of ammonia/air mixtures at low temperatures. Species
Ammonia measurements for weak flames of ammonia/air mixtures at atmospheric pressure and equivalence ratios
Nitrogen oxides (NOx ) of 0.8, 1.0 and 1.2 under a maximum wall temperature of 1400 K were conducted using a mass spec-
Microcombustion trometer and profiles of the NH3 , O2 , H2 O NO, and N2 O mole fractions were obtained. Chemical kinetic
Energy carrier modeling was conducted with extensive updates mainly for the N2 Hx chemistry because N2 Hx species
Chemical kinetics
were expected to be produced from the NH2 + NH2 reactions at low temperatures. The mechanism devel-
oped in the present study well predicted species profiles of NH3 , O2 and H2 O for weak flames measured
in experiments. The present mechanism also well predicted the final values of the NO and N2 O mole
fractions behind the reaction zone of weak flames but overestimated these mole fractions in the reaction
zone of weak flames. To confirm the existence of N2 Hx species in the reaction zone of weak flames, sig-
nals from N2 H4 were distinguished from measured signals. Sensitivity analysis and reaction flux analysis
were conducted and the importance of the N2 Hx chemistry in the reaction zone of weak flames at low
temperatures was identified. Validation of the present mechanism with literature data on ignition delays
and flame speeds were conducted and reasonable agreements with literature data were confirmed. For
N2 O and NO in the reaction zone of weak flames, however, there was discrepancy between measured and
computational mole fractions and further improvements of chemical kinetics related to ammonia ignition
are still necessary.
© 2017 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

1. Introduction combustion technologies combined with heat regeneration and

swirl burners enables to overcome the low reactivity of ammo-
Ammonia (NH3 ) is considered to be one of promising energy nia. For further development of large-scale, high-efficiency ammo-
carriers [1,2] and ammonia direct combustion has a great potential nia combustors, reliable reaction mechanisms validated with var-
as an energy utilization method without carbon emission to com- ious ammonia combustion characteristics are necessary to be de-
pensate rapid and huge variation of energy demand due to renew- veloped.
able energy sources. Although the reactivity of ammonia is much Ammonia reaction mechanisms have been developed with ex-
lower than those of hydrocarbons, AIST (National Institute of Ad- tensive validation using experimental and theoretical studies on
vanced Industrial Science and Technology) has successfully demon- oxidation and pyrolysis reactions of ammonia [4–9] as well as
strated power generation using a small gas turbine of ammo- NOx reduction reactions with ammonia [10–13] and NOx forma-
nia/air combustion [3] and it has been shown that recent advanced tion reactions from ammonia [14–17] because the NOx reaction
system is a sub-mechanism of the ammonia oxidation. The de-
veloped ammonia reaction mechanisms have also been validated
∗
Corresponding author at: Institute of Fluid Science, Tohoku University, 2-1-1
with flame and ignition characteristics. As flame characteristics,
Katahira, Aoba-ku, Sendai 980-8577, Japan. laminar flame speeds [18–22], flammability limits [19,22], flame
E-mail address: [email protected] (H. Nakamura). structures [23] and extinction stretch rates [24] of ammonia/air

http://dx.doi.org/10.1016/j.combustflame.2017.06.021
0010-2180/© 2017 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
 H. Nakamura et al. / Combustion and Flame 185 (2017) 16–27 17

mixtures have been obtained. As ignition characteristics, shock

tube experiments for highly-diluted ammonia/oxygen/argon or
ammonia/oxygen/nitrogen mixtures at high temperatures (mostly
higher than 1800 K) have been conducted [25–30]. However, there
have been no studies on ignition delay times of ammonia/air mix-
tures at lower temperatures. Ignition characteristics at low temper-
atures are practically important in the base of premixed turbulent
jet flames [31]. Longer residence time by introducing swirl and re-
circulation near the base of premixed turbulent jet flames success-
fully attained stable combustion with ammonia/air mixtures [3].
To examine ignition characteristics of ammonia/air mixtures at low
temperatures (around 1300 K), weak flames in a micro flow reactor
with a controlled temperature profile were employed in our previ-
ous study [32].
The micro flow reactor was originally employed to investigate
flame dynamics in a heated microchannel of a premixture and
weak flames were observed in low flow velocity conditions [33].
A weak flame branch was found to be on the ignition branch
of the Fendell curve in a theoretical study [34] and weak flames
have successfully been applied to investigate ignition character-
istics of given premixtures. The transient, two-stage ignition of Fig. 1. Wall-temperature profile measured in experiment and used in computation.
high hydrocarbons has been observed as steady, three-stage weak
flames for dimethyl ether [35] and n-heptane [36]. Fuel reactiv-
ity has been evaluated using the weak flame position as an index, indicate those around 20 0 0 K relevant to temperatures of shock
namely, weak flames of lower/higher reactivity fuels are placed in tube experiments.
a higher/lower temperature region of the reactor for syngas [37],
C1–C4 alkanes[38], C2–C5 alkenes [39], gasoline primary reference 2. Experimental setup and computational method
fuels (PRFs) [40] and diesel PRFs [41]. The gas-phase temperature
profile is nearly equal to the given wall-temperature profile even in The same experimental setup and computational method as
the reaction zone in the weak flame regime [42] and the steady, 1- those employed in a previous study [32] were employed in the
D numerical model of the micro flow reactor has enabled the vali- present study. Details can be found in [32] and only a short sum-
dation of the employed reaction mechanisms under a well-defined mary is given here.
temperature profile by comparison with experimental weak flame A quartz tube with an inner diameter of 2 mm was used as
responses [38,40]. the reactor channel and heated by a hydrogen/air flat-flame burner
One of the important characteristics of the micro flow reactor to give a stationary temperature ramp along the inner surface of
is its capability for investigating ignition characteristics of low re- the reactor channel in the flow direction. Hereafter, “wall temper-
activity premixtures in low temperature conditions [43] because ature” means the temperature on the inner surface of the reac-
of long residence time and the well-defined temperature profile. tor channel. The present study set the maximum wall tempera-
Weak flames of ammonia/air mixtures, which have quite low reac- ture of 1400 K and measured wall temperature profile is shown in
tivity, were successfully observed in our previous study and species Fig. 1. Ammonia/air mixtures with an inlet mean flow velocity of
measurements in the stoichiometric condition showed complete 10 cm/s were supplied from the low temperature side of the reac-
consumptions of ammonia and oxygen and the formation of near- tor channel at atmospheric pressure and weak flames were formed
equilibrium water vapor in the reaction zone [32]. However, most in the high temperature region of the reactor channel. A quartz
existing reaction mechanisms did not predict the formation of micro-probe (inner diameter: 0.1 mm; outer diameter of the edge:
weak flame. Only the reaction mechanism developed by Konnov 0.25 mm; outer diameter of the root: 1.6 mm) was fused on the
[44] predicted the formation of weak flame but its position was side wall of the reactor channel and a T-shaped reactor was fab-
on the far upstream side (low temperature side) compared with ricated. Gas in the reactor channel was sampled from the micro-
the experimental weak flame position. These results indicated that probe and introduced to a quadrupole mass spectrometer (QMS)
the Konnov mechanism predicted too high reactivity, whereas the through a fused-silica capillary. The temperature of the sampling
other mechanisms predicted too low reactivity, which were also line was kept at 373 K by electric heaters to prevent condensation
confirmed by ignition delay simulation at relevant temperatures of water vapor. This setup attained rapid reductions of pressure
(around 1300 K). On the other hand, all the mechanisms showed and temperatures of sampled gas at a sufficiently low flow rate
similar ignition delay times at high temperatures (around 20 0 0 K). so that the disturbance of the flow in the reactor channel by the
Therefore, modeling studies based on species profiles of ammo- sampling and chemical reaction in the sampling line were negli-
nia/air weak flames are expected to develop a more reliable am- gibly small [32]. The sampled gas was analyzed by the QMS, and
monia reaction mechanism, especially for the validity of ignition NH3 , O2 , H2 O, N2 O and NO were identified and quantified. Elec-
characteristics at low temperatures. tron ionization of 13 eV was chosen to minimize the fragmentation
The objective of the present study is to develop an ammo- effects. Voltage of an electron multiplier was set to be 10 0 0 V for
nia reaction mechanism which correctly predicts species profiles NH3 , O2 and H2 O, whereas that was set to be 1400 V for N2 O and
of ammonia/air weak flames. Numerical simulations with the de- NO. Overall error of the measured mole fraction was estimated to
veloped mechanism for predicting ignition delay times and lam- be 10–15%. All experiments were conducted at atmospheric pres-
inar flame speeds were conducted and the comparisons with sure in the present study.
literature data were made to confirm the comprehensive per- To simulate weak flames in the micro flow reactor, the flow
formance of the developed mechanism. Hereafter, low tempera- field in the reactor channel was modeled as a one-dimensional
tures indicate those around 1300 K relevant to temperatures of steady-state reactive flow without a boundary layer. Computations
weak flames in the present experiments and high temperatures were conducted using PREMIX in ANSYS Chemkin-Pro v17.2 and a
18 H. Nakamura et al. / Combustion and Flame 185 (2017) 16–27

heat convection term between the gas and the reactor wall was The final step of improvements was to update individual rate
added to the gas-phase energy equation. The wall-temperature constants, as shown in Table 1. As mentioned above, N2 Hx chem-
profile used in computations is shown in Fig. 1. The gas-phase istry would be important in ammonia/air ignition characteristics
temperature is almost equal to the wall temperature even in the at low temperatures. However, N2 Hx chemistry has not been well
reaction zone in the weak flame regime [42] and the validity of studied compared with other nitrogen chemistry as well as hydro-
this 1-D modeling for weak flames has been confirmed for various carbon chemistry. Therefore, improvements in this step mainly fo-
hydrocarbons [35–43]. Since the radical quenching effect on the cused on reactions related to N2 Hx species and their rate constants
quartz surface was confirmed to be negligibly small in the heated were updated based on available literature data.
microchannel [45,46], the radical quenching mechanism was not R1–15 are reactions related to H2 NN taken from a review by
considered in the present study. Dean and Bozzelli [60]. H2 NN was not included in the mecha-
nisms of Miller and Bowman, Mathieu and Petersen, and Konnov.
Dean and Bozzelli examined that the H2 NN formation from hy-
3. Chemical kinetic model drazine (R1) is not significant, whereas H2 NN is one of the ma-
jor products of the NH2 + NH2 reaction (R13) at low temperatures.
The present study chose the nitrogen chemistry developed by H2 NN is also produced from HNOH (R12) and N2 H3 (R14 and
Miller and Bowman [47] as a starting mechanism and four steps R15). HNOH is formed from the isomerization reaction of H2 NO
of improvements were made. The first step of improvements was which is mainly produced from the NH2 + HO2 reaction (R30, dis-
to adopt the H2 /NHx /N2 O/NO2 /NNH chemistry studied by Math- cussed later). Hydrazine is also one of the major products of the
ieu and Petersen [48]. They employed the H2 /CO chemistry de- NH2 + NH2 reaction at low temperatures and gives N2 H3 through
veloped by Kéromnès et al. [49], the NHx /HCN chemistry devel- H-atom abstraction reactions with radicals. R2 is the isomeriza-
oped by Dagaut et al. [50], N2 O chemistry developed in their pre- tion reaction of H2 NN with N2 H2 , R3 is the decomposition reac-
vious work [51], NO2 chemistry developed by Dayma and Dagaut tion of H2 NN, and R4–11 are reactions of H2 NN with active species.
[52] and Sivaramakrishnan et al. [53], and the NNH chemistry R16–20 are N2 Hx + NOx reactions which are not included in the
developed by Klippenstein et al. [54]. As a result, the proposed mechanisms of Miller and Bowman, Mathieu and Petersen, and
model significantly improved predictions on ignition delay times Konnov. NOx reactions are important sub-mechanisms in ammo-
of highly-diluted NH3 /O2 /Ar mixtures at high temperatures [48]. nia oxidation and N2 Hx species will be formed from recombina-
In the present study, these updates were employed but reactions tion reactions of rich NH or NH2 radicals in ammonia/air com-
related to carbon-containing species were eliminated. bustion. Daimon et al. conducted high-level ab initio computations
The second step of improvements was to adopt the chemistry and calculated rate constants of N2 Hx + NO2 reactions (R16–19)
of N2 Hx except NNH, namely, reactions related to N2 H4 , N2 H3 and for simulations of hydrazine ignition with NO2 –N2 O4 mixtures in
N2 H2 studied by Konnov [44]. As mentioned earlier, only the reac- bipropellant systems [61]. It was also shown that reaction chan-
tion mechanism developed by Konnov predicted the formation of nels through trans-N2 H2 and trans-HONO are predominant com-
weak flames in the micro flow reactor, whereas the other mecha- pared with those through cis-N2 H2 and cis-HONO and the present
nisms did not show any reaction zone [32]. There is a major dif- study adopted those through trans-N2 H2 and trans-HONO. A re-
ference in modeling of hydrazine reactions between the Konnov lated reaction of HONO with NH2 (R21) was added to the present
mechanism and the other mechanisms. In addition to ammonia de- model from the estimation by Dean and Bozzelli [60], which was
composition [55] Konnov and De Ruycy investigated hydrazine de- also employed by Daimon et al. [61]. Rate constants of NNH + NO
composition and developed N2 Hx chemistry [56]. They also inves- and N2 H2 + NO reactions were estimated by Miller and Bowman
tigated the impact of N2 Hx chemistry on NO formation and identi- [47] and that of N2 H3 + NO reaction was estimated by Konnov [44],
fied the importance of N2 H3 reactions in NO formation at low tem- whereas N2 H4 + NO reaction was not included. The present study
peratures (below 1500 K) [57]. The Konnov mechanism have been employed rate constant of the N2 H4 + NO reaction (R20) calculated
developed incorporating fundamental findings on N2 Hx chemistry by Raghunath et al. [62].
in order to improve predictions of NOx formation. Another im- Reactions in R22–26 were included in the Konnov mechanism
portant fact is that most NOx studies have employed low con- but the present study employed different rate constants, as shown
centration of ammonia (typically below 0.1%), while ammonia/air in the table. Konnov and De Ruyck modeled N2 H3 = N2 H2 + H
mixtures contains much higher concentration of ammonia (approx. (R22) as a third-body reaction [56] and the low-pressure rate
22% at stoichiometry). This indicates the presence of rich NHx rad- constant was estimated based on the bond dissociation energy
icals in ammonia/air combustion, leading significant recombina- [63,64] since they conducted hydrazine decomposition experi-
tion reactions to form N2 Hx species [32]. Therefore, in the present ments at low pressures. They also pointed out that this esti-
study, N2 Hx chemistry was considered to play an important role in mation has large uncertainty at atmospheric and elevated pres-
ammonia/air ignition at low temperatures and employed reactions sures. The present study employed the rate constant calculated
related to N2 H4 , N2 H3 and N2 H2 studied by Konnov [44]. by Dean and Bozzelli [60]. The Konnov mechanism employed
The third step of improvements was to adopt an accurate and the rate constant of N2 H2 = NNH + H (R23) estimated by Miller
consistent set of thermochemical properties calculated by Bugler et al. [4]. The present study employed a new rate constant
et al. [58]. Thermochemical properties are quite important to eval- which was calculated with theoretical computations by Dean and
uate rates of reverse reactions, especially for isomerization reac- Bozzelli [60]. The Konnov mechanism employed rate constants of
tions, and it has been demonstrated for hydrocarbons that a mi- N2 H2 + X (H, OH, NH2 ) calculated with high-level ab initio com-
nor difference in thermochemical properties causes a major differ- putations by Linder et al. [65]. Subsequently, Chuang and Truhlar
ence in ignition delay times at low temperatures even with the reported an improvement of rate constant for the N2 H2 + H reac-
same chemical kinetics [59]. Bugler et al. also reported thermo- tion (R24) by taking into account for multidimensional tunneling
chemical properties of nitrogen-containing species have a signif- effects [66]. The present study employed the updated rate constant
icant disparity within the literature and they conducted accurate for R24 obtained by Chuang and Truhlar. Konnov and De Ruyck
and consistent evaluations of thermochemical properties for vari- modeled N2 H2 + N2 H2 = N2 H3 + NNH (R25) as a reverse reaction
ous nitrogen-containing species using high-accuracy ab initio com- [56] based on the estimation by Hanson and Salimian [67]. Since
putations [58]. The present study replaced thermochemical prop- thermochemical properties of N2 H3 show a significant disparity
erties of all nitrogen-containing species by those in [58]. within the literature [58], the present study employed the original
 H. Nakamura et al. / Combustion and Flame 185 (2017) 16–27 19

Table 1
List of reactions in the fourth step of improvements (cm, mol, s, cal).

Index Reaction A n E Source

R1 N2 H4 = H2 NN + H2 [60]
PLOG 0.1 4.0E44 −9.85 71,353
PLOG 1 5.3E39 −8.35 69,307
PLOG 10 2.5E39 −8.19 69,664
R2 N2 H2 = H2 NN [60]
PLOG 0.1 9.2E38 −9.01 67,727
PLOG 1 2.0E41 −9.38 68,452
PLOG 10 1.3E45 −10.13 70,757
R3 H2 NN = NNH + H [60]
PLOG 0.1 7.2E28 −7.77 50,758
PLOG 1 3.2E31 −6.22 52,318
PLOG 10 5.1E33 −6.52 54,215
R4 H2 NN + O2 = NH2 + NO2 1.5E12 0.00 5961 [60]
R5 H2 NN + H = N2 H2 + H 1.8E10 0.97 4471 [60]
R6 H2 NN + H = NNH + H2 4.8E08 1.50 −894 [60]
R7 H2 NN + O = NH2 + NO 3.2E09 1.03 2702 [60]
R8 H2 NN + O = NNH + OH 3.3E08 1.50 −894 [60]
R9 H2 NN + OH = NNH + H2 O 2.4E06 2.00 −1192 [60]
R10 H2 NN + NH2 = NNH + NH3 1.8E06 1.94 −1152 [60]
R11 H2 NN + HO2 = NNH + H2 O2 2.9E04 2.69 −1600 [60]
R12 HNOH + NH2 = H2 NN + H2 O 4.6E19 −1.94 1945 [60]
R13 NH2 + NH2 = H2 NN + H2 [60]
PLOG 0.1 2.4E20 −2.91 2136
PLOG 1 1.2E21 −3.08 3368
PLOG 10 2.3E19 −2.54 4183
R14 N2 H3 + OH = H2 NN + H2 O 3.0E13 0.00 0 [60]
R15 N2 H3 + NH2 = NH3 + H2 NN 3.0E13 0.00 0 [60]
R16 N2 H4 + NO2 = N2 H3 + HONO 1.3E01 3.36 11,346 [61]
Duplicate reaction 3.4E00 3.27 8018
R17 N2 H3 + NO2 = HNO2 + N2 H2 7.2E07 1.05 −2759 [61]
R18 N2 H2 + NO2 = NNH + HONO 1.1E-3 4.47 7183 [61]
R19 NNH + NO2 = HONO + N2 1.7E01 2.84 1673 [61]
R20 N2 H4 + NO = N2 H3 + HNO 6.4E01 3.16 30,485 [62]
R21 HNOH + NH2 = N2 H3 + OH 6.7E06 1.82 715 [60,61]
R22 N2 H3 = N2 H2 + H [60]
PLOG 0.1 2.3E43 −9.55 64,379
PLOG 1 3.6E47 −10.38 69,009
PLOG 10 1.8E45 −9.39 70,141
R23 N2 H2 = NNH + H [60]
PLOG 0.1 1.6E37 −7.94 70,765
PLOG 1 2.6E40 −8.53 72,931
PLOG 10 1.3E44 −9.22 77,084
R24 N2 H2 + H = NNH + H2 8.1E05 2.32 −956 [66]
R25 N2 H2 + N2 H2 = N2 H3 + NNH 1.0E13 0.00 10,014 [67]
R26 N2 H2 + HO2 = NNH + H2 O2 1.5E13 0.00 20 0 0 See text
R27a NNH +M = N2 + H +M 1.0E13 0.50 3060 See text, [60]
R27b NNH = N2 + H 3.0E08 0.00 0 See text, [60]
R27c NNH + O2 = N2 + H + O2 Removed
R28 NH2 + NO = N2 + H2 O 1.3E16 −1.25 0 [54]
Duplicate reaction −3.1E13 −0.48 1180
R29 NH2 + O = HNO + H 4.6E13 0.00 0 [60,73]
R30 NH2 + HO2 = H2 NO + OH 1.0E13 0.00 0 See text

estimation by Hanson and Salimian. Konnov and De Ruyck esti- and Petersen mechanism. The value of this rate constant multi-
mated the rate constant of N2 H2 + HO2 = NNH + H2 O2 (R26) to be plied by molar concentration of oxygen molecule in air at atmo-
1.0 × 1013 exp(−20 0 0/RT) in cm, mol, s, cal. Since R26 showed high spheric pressure and 10 0 0 K is 1.3 × 108 s−1 , which is also slower
sensitivity to overall reactivity in the micro flow reactor, as shown than the NNH lifetime. The present study considered the rate con-
in Fig. 3 later, an empirical correction with a factor of 1.5 was stant of the NNH dissociation proposed by Dean and Bozzelli [60]:
adopted to fit experimental results. 3 × 108 + 1 × 1013 T0.5 exp(−3060/RT) [M] s−1 . They modeled this
Reactions in R27–30 were included in the Mathieu and Pe- rate constant as the summation of dissociation due to tunneling
tersen mechanism but the present study employed different rate and thermal activation. This rate constant yields 1.1 × 109 s−1 at
constants, as shown in Table 1. R27a–c show three different for- 10 0 0 K and 1.6 × 109 s−1 at 20 0 0 K in air at atmospheric pressure,
mulations for the dissociation reaction of NNH. Klippenstein et al. which are consistent with the NNH lifetime estimated by Klippen-
estimated the lifetime of NNH to be shorter than 1 × 10−9 s [54]. stein et al. [54]. To adopt this rate constant in CHEMKIN format,
The rate constant of the NNH dissociation was estimated to be the present study employed rate constants of R27a and R27b in
6.5 × 107 s−1 in the R27b formula by Miller and Glarborg [68], the table and removed R27c included in the Mathieu and Petersen
which was employed in the Mathieu and Petersen mechanism. mechanism.
However, this rate constant is too slow compared with the esti- R28–30 are updates related to NH2 . Mathieu and Petersen
mation of the NNH lifetime. Glarborg et al. also modeled the dis- adopted the rate constant determined by Klippenstein et al.
sociation reaction of NNH with an oxygen molecule as a third-body [54] for NH2 + NO = NNH + OH but did not adopt that for
in R27c formula [69,70], which was also included in the Mathieu NH2 + NO = N2 + H2 O (R28). Since rate constants of these two
20 H. Nakamura et al. / Combustion and Flame 185 (2017) 16–27

Fig. 2. Comparison between measured and computational species profiles of weak flames for ammonia/air mixtures at φ = 0.8, 1.0 and 1.2 (symbols: experiments; solid
lines: computations with present mechanism; dash-dotted lines: computations with Konnov mechanism; broken lines: wall-temperature profile).
 H. Nakamura et al. / Combustion and Flame 185 (2017) 16–27 21

reactions were determined to reproduce literature data of the

branching ratio, the present study adopted the rate constant de-
termined by Klippenstein et al. for R28 to keep consistency of the
modeling. The rate constant of NH2 + O = HNO + H (R29) was orig-
inally modeled by Miller et al. [4] with negative temperature de-
pendence to be 6.63 × 1014 T−0.5 cm3 mol−1 s−1 and was employed
in the mechanisms of Miller and Bowman and that of Mathieu and
Petersen. Tian et al. [71] proposed the temperature-independent
rate constant for R29 to be 6.6 × 1013 cm3 mol−1 s−1 because mea-
surements of the rate constant by Inomata and Washida did not
show temperature dependence [72]. The present study employed
4.6 × 1013 cm3 mol−1 s−1 proposed by Dean and Bozzelli [60] for
R29 based on experiments by Dransfeld et al. [73]. The rate
constant of NH2 + HO2 = H2 NO + OH (R30) was estimated to be
2.5 × 1013 cm3 mol−1 s−1 by Glarborg et al. [74] and increased to
be 5.0 × 1013 cm3 mol−1 s−1 in [69]. This higher rate constant was
employed in the Mathieu and Petersen mechanism. On the other
hand, Sumathi and Peyerimhoff found potential intermediates for
the NH2 + HO2 reaction and estimated the rate constant for R30
with a multi-channel potential energy surface [75], which is a fac-
tor of 4–6 lower than the increased rate constant estimated by
Glarborg et al. [69]. Although there has been discussion on the fate
of the NH2 + HO2 reaction system, especially at high pressures [9],
the present study employed 1.0 × 1013 cm3 mol−1 s−1 for the rate
constant of R30, which is close to the estimation by Sumathi and
Peyerimhoff in the present condition. The developed mechanism in
the present study consists of 33 species and 232 reactions.

4. Results and discussion

4.1. Validation with weak flames

Figure 2 shows measured and computational species profiles of

weak flames at equivalence ratios, φ , of 0.8, 1.0 and 1.2. Since only
the Konnov mechanism predicted the formation of weak flames in
our previous study [32], computational results with the Konnov
mechanism are also shown in the figure. The gas-phase temper-
ature is almost equal to the wall temperature even in the reaction
zone in the weak flame regime [42].
The measured species profiles show distinct decreases of NH3
and O2 and an increase of H2 O around T = 1270 K (x = 6.1 cm)
for the three equivalence ratios. O2 and NH3 are seen to re-
main behind the reaction zones for φ = 0.8 and 1.2, respectively.
The present mechanism well reproduces these experimental re-
sults. The Konnov mechanism predicts the reaction zones around
T = 1200 K (x = 5.8 cm) for the three equivalence ratios, which are
approximately 70 K lower than the experimental results. This indi-
cates that the Konnnov mechanism significantly overestimates re-
activity of ammonia in the present temperature condition since the
previous study showed a 40 K difference in the weak flame posi-
tion between n-heptane and iso-octane [40]. Note that the Math-
ieu and Petersen mechanism did not predict the formation of weak
flames [32], indicating significant underestimation of the reactivity
of ammonia in the present temperature condition.
The measured N2 O profiles show that N2 O is formed in the re-
Fig. 3. First-order, A-factor sensitivity coefficients of (a) H2 O, (b) NO and (c) N2 O
action zone as an intermediate for the three equivalence ratios. fractions.
The N2 O mole fraction approaches zero behind the reaction zone
for φ = 1.0 and 1.2 while it approximately remains 600 ppm at
x = 7.5 cm for φ = 0.8. The present and Konnov mechanisms well φ = 1.2, on the other hand, the NO mole fraction decreases behind
predict these final values of the N2 O mole fraction but overes- the reaction zone and approaches zero, which is expected to be the
timate the peak values of the N2 O mole fraction in the reaction De-NOx process due to the unburned NH3 . The present mechanism
zone. The measured NO profile shows that a small amount of NO quantitatively well predicts the final values of the NO mole fraction
is formed in the reaction zone for the three equivalence ratios. The for the three equivalence ratios, whereas the Konnov mechanism
NO mole fraction gradually increases behind the reaction zone and significantly overestimates the final values of the NO mole fraction
approaches a certain value for φ = 0.8 and 1.0. The final value of at φ = 0.8 and 1.0. However, the present and Konnov mechanisms
the NO mole fraction at φ = 0.8 is higher than that of φ = 1.0. At overestimate the NO mole fraction in the reaction zone. These
22 H. Nakamura et al. / Combustion and Flame 185 (2017) 16–27

Fig. 4. Reaction pathway in weak flames at φ = 0.8, 1.0 and 1.2 and 30% fuel conversion. Numbers in brackets [] are percent contribution to consumption of the species on
the source side of the arrow.

results show that the present mechanism significantly improved only important to the ammonia reactivity but also to the NO and
prediction of the ammonia reactivity with air at low temperatures N2 O formation in the reaction zone.
and indicate that further improvements will be necessary for pre- Results of the sensitivity analysis also indicate the importance
diction of NO and N2 O formation, especially in the reaction zone. of the NH2 + HO2 and NH2 + NO reactions for H2 O, NO and N2 O
To identify important reactions for ammonia reactivity and NO and formation. R30 shows high sensitivity for H2 O and N2 O formation
N2 O formation in the reaction zone, sensitivity analysis and reac- but is not listed in Fig. 3b (sensitivity for NO). On the other hand,
tion flux analysis were carried out. an alternative product channel of R30, NH2 + HO2 = NH3 + O2 (S7),
The first-order, A-factor sensitivity coefficients of H2 O, NO shows high sensitivity for NO formation. R28 shows the highest
and N2 O fractions were computed with the present mechanism. sensitivity for H2 O, NO and N2 O. An alternative product channel of
H2 O was selected as a target species to evaluate the ammonia R28, NH2 + NO = NNH + OH (S5), also shows very high sensitivity
reactivity. Figure 3 shows results of sensitivity analysis at the for the H2 O formation. R28 shows very high, negative sensitivity
three equivalence ratios. The sensitivity coefficients were taken for the H2 O formation while S5 shows very high, positive sensitiv-
where 30% of inlet ammonia were consumed (x ≈ 6.0 cm) to ex- ity for H2 O formation. This indicates that the branching ratio be-
amine species formation in the reaction zone. The top 10 re- tween R28 and S5, which was employed from [54] in the present
actions with the highest sensitivity coefficients are shown in study, is one of the key parameters to control the ammonia reac-
the figure but hydrogen-oxygen reactions are not shown here tivity.
so as to focus on the nitrogen chemistry. Reactions with in- Figure 4 shows the reaction pathway in weak flames where 30%
dexes of Rx are listed in Table 1 and indexes of Sx are pre- of inlet ammonia was consumed. The thickness of arrows shows
sented here for discussion. Figure 3 shows that many reac- the absolute rate of production and the numbers in brackets []
tions related to N2 H2 have high sensitivity for H2 O, NO and are percent contribution to the consumption of the species on the
N2 O formation. R24–26 in the present improvements as well source side of the arrow in the order of the equivalence ratios of
as N2 H3 + O2 = N2 H2 + HO2 (S2) and N2 H2 + NH2 = NH3 + NNH 0.8, 1.0 and 1.2 from left to right. The H-atom abstraction reaction
(S3), which were taken from the Konnov mechanism, shows of NH3 with OH is dominant in formation of NH2 radicals. There
high sensitivity for H2 O, NO and N2 O. The NH2 + NH2 reac- are many branchings of the reaction pathway from NH2 radicals.
tions of two product channels, NH2 + NH2 = N2 H2 + H2 (S6) and One of the major products of the NH2 reactions is N2 H2 through
NH2 + NH2 = NH3 + NH (S8), show high sensitivity for NO and N2 O S6. Subsequently, H-atom abstraction reactions of N2 H2 with HO2
formation, indicating that the branching ratio between H2 H2 and (R26), NH2 (S3) and H (R24) radicals and N2 H2 itself (R25) proceed
NH3 products from the NH2 + NH2 reactions are important to to form NNH. R25 also produces N2 H3 radicals which mainly react
NO and N2 O formation. H-atom abstraction reaction of hydrazine with O2 to form N2 H2 again (S2). Fluxes in S2 show large depen-
with HO2 , N2 H4 + HO2 = N2 H3 + H2 O2 (S1), which does not include dence on equivalence ratio. The consumption of NNH is dominant
N2 H2 , shows high sensitivity for H2 O formation. These results ex- in the decomposition through R27. These results indicate that the
amine that the N2 Hx chemistry, especially related to N2 H2 , is not major reaction pathway of nitrogen in fuel strongly depends on
 H. Nakamura et al. / Combustion and Flame 185 (2017) 16–27 23

the N2 Hx chemistry. The second major reactions of NH2 radicals

are reactions with NO to form NNH (S5) and N2 (R28). These two
reactions show large contribution for the consumption of NO, indi-
cating the significant importance in both ammonia reactivity and
NO consumption. NO is mainly produced from reactions of HNO
with O2 , H and N2 H2 . The reaction of HNO with N2 H2 also pro-
duces N2 H3 radicals but the arrow to N2 H3 is not shown in the fig-
ure to avoid complicated intersections. These results indicate that
N2 H2 shows significant contributions to both ammonia reactivity
and NO formation in weak flames. Another important intermedi-
ate in weak flames, N2 O, is mainly produced from a reaction of NO
with NH radicals. The N2 Hx chemistry is not directly related to the
N2 O formation but shows large sensitivity for the N2 O formation,
as shown in Fig. 3. Dean and Bozzelli [60] have reported that one
of the major products of the NH2 + NH2 reactions is H2 NN at low
temperatures but results of sensitivity analysis and flux path anal-
ysis show H2 NN is not a major contributor to ammonia reactivity
and NO and N2 O formation.
To experimentally validate the existence of N2 Hx intermediates
in weak flames, measurements of N2 Hx intermediates were at-
Fig. 5. Comparison of measured N2 H4 signal with computed N2 H4 mole fraction at
tempted in the present mechanism. Since concentrations of N2 Hx
φ = 1.0.
intermediates were expected to be low and molecular and frag-
ment ions of N2 Hx were difficult to distinguish from those of other
species, only N2 H4 was considered to be a detectable N2 Hx inter- signal shows a distinct peak in the reaction zone as shown in the
mediate in the present study. Only N2 H4 and O2 were species of computed profile of the N2 H4 mole fraction. The present study in-
m/z 32 in the present study and their ionization energies are 8– dicated experimental evidence of the N2 H4 intermediate in ammo-
10 and 12.1 eV, respectively [76]. Since the present study employed nia/air oxidation at low temperatures.
electron ionization, there must be a signal from O2 molecular ions The present mechanism overestimated the NO and N2 O mole
even with a smaller ionization energy. To distinguish the N2 H4 sig- fractions in weak flames. Therefore, further kinetic modeling will
nal, SN2 H4 , from the measured signal, the following formula was be necessary in the future. Since extensive studies have been made
employed in the present study: for the NHx chemistry compared with the N2 Hx chemistry, results
of sensitivity analysis and flux path analysis suggest the necessity
SN2 H4 = Sm − CO2 × XO2
of further chemical kinetic studies on production and/or consump-
where Sm is the measured signal of m/z 32 at an ionization energy tion of NO and N2 O related to N2 Hx species (especially N2 H2 ). For
of 8 eV and an electron multiplier of 1400 V, CO2 is a conversion this, quantitative measurements of N2 Hx species with photoioniza-
factor of the O2 mole fraction to the signal, and XO2 is the O2 mole tion will be valuable for future examination.
fraction. XO2 was used from Fig. 2 which was obtained with an ion-
ization energy of 13 eV and an electron multiplier of 10 0 0 V. CO2 4.2. Validation with ignition delays and flame speeds
was measured with standard gas at an ionization energy of 8 eV
and an electron multiplier of 1400 V in advance of experiments. To examine effects of the present modifications on combustion
This method would not be appropriate for quantitative compari- and ignition characteristics in other systems, the present mech-
son with model predictions because of large uncertainty (approx. anism was validated with literature data on ignition delays and
50%) but it would be useful for qualitative comparison. Figure 5 flame speeds. Since the present study focused on weak flames at
shows qualitative comparison of the measured N2 H4 signal with atmospheric pressure, literature data near atmospheric pressure
the computed N2 H4 mole fraction at φ = 1.0. The measured N2 H4 were selected. Figure 6 shows the validation for N2 O and NO2

Fig. 6. Comparison of ignition delay times between predictions and measurements for (a) H2 /O2 /N2 O/Ar and (b) H2 /O2 /NO2 /Ar mixtures. Experimental data for (a) and (b)
were taken from [51] and [77], respectively. Note that broken lines are almost coincident with solid lines.
24 H. Nakamura et al. / Combustion and Flame 185 (2017) 16–27

Fig. 7. Comparison of half-life times in N2 H4 pyrolysis between predictions and

measurements. Experimental data were taken from [78]. Fig. 8. Comparison of ignition delay times between predictions and measurements
for a stoichiometric NH3 /O2 mixture diluted in Ar at an average pressure of 1.4 atm.
Experimental data were taken from [48].

sub-mechanisms. This validation was also made by Mathieu and

Petersen [48]. Figure 6a shows ignition delay times of a 0.01 pyrolysis at the total gas density of 7.5 × 10−5 mol/cm3 measured
H2 /0.01 O2 /0.0032 N2 O mixture diluted in Ar at an average pres- by Michel and Wagner [78] and predicted with the present mecha-
sure of 1.9 atm. Experimental results obtained by Mathieu et al. nism and the Konnov mechanism. The present mechanism predicts
[51] and computational results with the present mechanism and very slightly shorter half-life times than the Konnov mechanism
the Mathieu and Petersen mechanism are shown in the figure. (2–10%) but still shows satisfactory agreement with experimental
Figure 6b shows ignition delay times of a 0.01 H2 /0.01 O2 /0.0016 results.
NO2 mixture diluted in Ar at an average pressure of 1.6 atm. Ex- Mathieu and Petersen measured ignition delay times of NH3 /O2
perimental results obtained by Mathieu et al. [77] and computa- mixtures diluted in Ar at high temperatures and developed
tional results with the present mechanism and the Mathieu and the mechanism through validation with the measured data. The
Petersen mechanism are shown in the figure. Figure 6 shows that present mechanism has been developed through validation with
the present mechanism well predicted the experimental results weak flames at low temperatures. The validation with ignition
and that predictions of the present mechanism are in excellent delay times at high temperatures was conducted by the present
agreement with those of the Mathieu and Petersen mechanism. mechanism. Figure 8 shows ignition delay times for a stoichiomet-
It was confirmed that the present modifications do not affect the ric NH3 /O2 mixture in Ar at an average pressure of 1.4 atm mea-
original performance of the Mathieu and Petersen mechanism for sured by the Mathieu and Petersen and computed by the present
N2 O and NO2 sub-mechanisms. mechanism and the Mathieu and Petersen mechanism. The present
Ignition delay data for N2 H4 /air or N2 H2 /air mixtures near at- mechanism predicts a factor of 1.5 shorter ignition delay times
mospheric pressure have not been previously obtained because than experiments. Ignition delay times computed with the present
of difficulties inherent in such experiments. Available data of mechanism are in good agreement with those computed with
N2 H4 pyrolysis were employed to validate the N2 Hx chemistry the Mathieu and Petersen mechanism around 104 K/T = 4 (2500 K),
in the present study. Figure 7 shows the half-life time in N2 H4 whereas the discrepancy of the ignition delay time between the

Fig. 9. Ignition delay times of a stoichiometric NH3 /air mixture at atmospheric pressure computed by the present mechanism (pw), the Konnov mechanism [44], GRI 3.0
[79], the Mathieu and Petersen mechanism [48] and the Tian mechanism [71].
 H. Nakamura et al. / Combustion and Flame 185 (2017) 16–27 25

Fig. 10. Comparison of flame speeds between experiments and computations for
ammonia/air mixtures at atmospheric pressure and room temperature. (: Za-
kaznov et al. [20]; ×: Takizawa et al. [21]; ࢞: Pfahl et al. [19]; ◦: Ronny [22]; ●:
Hayakawa et al. [18]).

two mechanisms becomes large with an increase of inverse tem-

perature (with a decrease of temperature). Results indicate that
the present mechanism significantly improved predictions of weak
flames with less impact on ignition delay at high temperatures.
However, effects of the N2 Hx chemistry on ignition delay at high
temperature need to be further investigated for accurate predic-
tions of both weak flames and ignition delays.
Figure 9 shows ignition delay times of a stoichiometric NH3 /air
mixture at atmospheric pressure computed with several mecha-
nisms. Since the Konnov mechanism predicted weak flames in the
lower temperature region than experiments while the Mathieu and
Petersen mechanism did not form weak flames in our previous
study, actual ignition delay times at low temperatures relevant to
temperatures of weak flames (104 K/T = 7.5) were expected to be
results between the Konnov and the Mathieu and Petersen mech-
anisms [32]. Ignition delay times computed by the present mech-
anism is quite close to those by the Mathieu and Petersen mecha-
nism at high temperatures and those by the Konnov mechanism at
low temperatures. The present mechanism bridges ignition delay
times at high temperatures predicted by the Mathieu and Petersen
mechanism and those at low temperatures predicted by the Kon-
nov mechanism.
In addition to the validation with ignition characteristics, vali-
dation with flame characteristics was also conducted using avail-
able data of flame speeds. Figure 10 shows flame speeds of am-
monia/air mixtures at atmospheric pressure and room temperature
in experiments and computations. The Konnov mechanism signifi-
cantly overestimates flame speed at all equivalence ratios (a factor
of 2.5 faster than experiments at the stoichiometric condition). The Fig. 11. Sensitivity coefficients of (a) N2 H4 lifetimes, (b) ignition delays and (c)
Mathieu and Petersen mechanism underestimates of flame speed flame speeds.
near stoichiometric conditions but well predicts in very lean and
rich conditions. Flame speeds computed by GRI 3.0 agree well with
experimental results at all equivalence ratios but GRI 3.0 shows flame speeds close to experimental results obtained by Pfahl et al.
quite different ignition characteristics from the other mechanisms, [19] in lean conditions.
as shown in Fig. 9. The Tian mechanism predicts slightly higher Figure 11 shows results of sensitivity analyses for N2 H4 life-
flame speeds than experiments (15% at the stoichiometric condi- times, ignition delays and flame speeds. Reactions with indexes,
tion) but well predicted extinction stretch rates in counterflow ex- Rx or Sx, are also important for weak flames because of discus-
periments by Colson et al. [24]. The present mechanism predicts sion in Section 4.1. Figure 11a shows the first-order, A-factor sen-
flame speeds similar to those of the Tian mechanism. However, the sitivity coefficients of the N2 H4 fraction at the half-life time of
present mechanism shows slightly better agreement with experi- N2 H4 pyrolysis in the same mixture condition as Fig. 7. The reac-
ments than the Tiam mechanism in rich conditions and predicts tion of N2 H4 (+M) = NH2 + NH2 (+M) shows primary importance
26 H. Nakamura et al. / Combustion and Flame 185 (2017) 16–27

for N2 H4 pyrolysis. All reactions except S8 are related to N2 Hx [4] J.A. Miller, M.D. Smooke, R.M. Green, R.J. Kee, Kinetic modeling of the oxidation
species. Figure 11b shows the brute-force sensitivity coefficients of of ammonia in flames, Combust. Sci. Technol. 34 (1983) 149–176.
[5] M.F. Zabielski, D.J. Seery, High temperature measurements of the rate of the
ignition delay time in the same mixture condition as Fig. 9. The reaction of OH with NH3 , Int. J. Chem. Kinet. 17 (1985) 1191–1199.
analysis was performed by increasing and decreasing a rate con- [6] J.W. Sutherland, P.M. Patterson, R.B. Klemm, Flash photolysis-shock tube ki-
stant by a factor of two and the sensitivity coefficient was ex- netic investigation of the reaction of O(3 P) atoms with ammonia, J. Phys. Chem.
pressed using the following formula: lnln((τ2+//0τ.5−)) . For reactions with
94 (1990) 2471–2475.
[7] J.W. Bozzelli, A.M. Dean, Energized complex quantum Rice–Ramsperger–KasseI
indexes, R25, R26, R28 and S6, sensitivity coefficients at 1250 K are analysis on reactions of NH2 with HO2 , O2 , and O atoms, J. Phys. Chem. 93
higher than those at 20 0 0 K, indicating that important reactions (1989) 1058–1065.
[8] D.F. Davidson, K. Kohse-Höinghaus, A.Y. Chang, R.K. Hanson, A pyrolysis mech-
for weak flames are also important for ignition at low tempera- anism for ammonia, Int. J. Chem. Kinet. 22 (1990) 513–535.
ture with good correlation. Two reactions related to N2 Hx species, [9] Y. Song, H. Hashemi, J.M. Christensen, C. Zou, P. Marshall, P. Glarborg, Ammo-
NH3 + NH2 = N2 H3 + H2 and N2 H4 + OH = NH3 + H2 NO, also show nia oxidation at high pressure and intermediate temperatures, Fuel 181 (2016)
358–365.
higher sensitivity coefficients at 1250 K than those at 20 0 0 K. The [10] L.D. Smoot, S.C. Hill, H. Xu, NOx control through reburning, Prog. Energy Com-
other reactions show higher sensitivity coefficients at 20 0 0 K than bust. Sci. 24 (1998) 385–408.
those at 1250 K and are not related to N2 Hx species. Figure 11c [11] J.A. Miller, M.C. Branch, R.J. Kee, A chemical kinetic model for the selective
reduction of nitric oxide by ammonia, Combust. Flame 43 (1981) 81–98.
shows the first-order, A-factor sensitivity coefficients of mass flow [12] Ø. Skreiberg, P. Kilpinen, P. Glarborg, Ammonia chemistry below 1400 K un-
rate in the same temperature and pressure conditions as Fig. 10. der fuel-rich conditions in a flow reactor, Combust. Flame 136 (2004) 501–
Most reactions are not related to N2 Hx species but related to NH2 518.
[13] M. Koebel, M. Elsener, M. Kleemann, Urea-SCR: a promising technique to re-
or HNO.
duce NOx emissions from automotive diesel engines, Catal. Today 59 (20 0 0)
These validation results show that the present mechanism gives 335–345.
reasonable predictions of not only weak flames but also ignition [14] C.P. Fenimore, Formation of nitric oxide from fuel nitrogen in ethylene flames,
Combust. Flame 19 (1972) 289–296.
delays and flame speeds. However, as discussed above, there are
[15] T. Mendiara, P. Glarborg, Ammonia chemistry in oxy-fuel combustion of
still several weak points in the present mechanism. To overcome methane, Combust. Flame 156 (2009) 1937–1949.
them, further investigation will be necessary to develop chemical [16] I. Naruse, Y. Yamamoto, Y. Itoh, K. Ohtake, Fundamental study on N2 O for-
kinetic mechanisms to predict ignition, flame and emission charac- mation/decomposition characteristics by means of low-temperature pulverized
coal combustion, Symp. (Int.) Combust. 26 (1996) 3213–3221.
teristics of ammonia combustion. [17] H. Watanabe, T. Marumo, K. Okazaki, Effect of CO2 reactivity on NOx forma-
tion and reduction mechanisms in O2 /CO2 combustion, Energy Fuels 26 (2012)
938–951.
5. Conclusions [18] A. Hayakawa, T. Goto, R. Mimoto, Y. Arakawa, T. Kudo, H. Kobayashi, Laminar
burning velocity and Markstein length of ammonia/air premixed flames at var-
ious pressures, Fuel 159 (2015) 98–106.
Species measurements of ammonia/air weak flames at equiv- [19] U.J. Pfahl, M.C. Ross, J.E. Shepherd, K.O. Pasamehmetoglu, C. Unal, Flammabil-
alence ratios of 0.8, 1.0 and 1.2 and atmospheric pressure were ity limits, ignition energy, and flame speeds in H2 –CH4 –NH3 –N2 O–O2 –N2 mix-
tures, Combust. Flame 123 (20 0 0) 140–158.
made using a micro flow reactor with a controlled temperature [20] F.Z. Zakaznov, L.A. Kursheva, Z.I. Felina, Determination of normal flame velocity
and a mass spectrometer. Chemical kinetic modeling was con- and critical diameter of flame extinction in ammonia–air mixture, Combust.
ducted based on the Mathieu and Petersen mechanism and the Explos. Shock Waves 14 (1978) 710–713.
[21] K. Takizawa, A. Takahashi, K. Tokuhashi, S. Kondo, A. Sekiya, Burning velocity
Konnov mechanism with extensive updates of N2 Hx reactions to
measurements of nitrogen-containing compounds, J. Hazard. Mater. 155 (2008)
predict the obtained species profiles. The present mechanism well 144–152.
predicted the profiles of the NH3 , O2 and H2 O mole fractions and [22] P.D. Ronney, Effect of chemistry and transport properties on near-limit flames
at microgravity, Combust. Sci. Technol. 59 (1988) 123–141.
the final values of the NO and N2 O mole fractions behind the re-
[23] C. Brackmann, V.A. Alekseev, B. Zhou, E. Nordström, P.-E. Bengtsson, Z. Li,
action zone, but it overestimated the NO and N2 O mole fractions M. Aldén, A.A. Konnov, Structure of premixed ammonia + air flames at atmo-
in the reaction zone. Sensitivity analysis and reaction flux analy- spheric pressure: Laser diagnostics and kinetic modeling, Combust. Flame 163
sis identified the importance of the N2 Hx chemistry in the reac- (2016) 370–381.
[24] S. Colson, A. Hayakawa, T. Kudo, H. Kobayashi, Extinction characteristics of
tion zone of weak flames at low temperatures (around 1300 K). The ammonia/air counterflow premixed flames at various pressures, J. Therm. Sci.
validation of the present mechanism with literature data on igni- Technol. 11 (2016) JTST0048, doi:10.1299/jtst.2016jtst0048.
tion delays and flame speeds was conducted and reasonable agree- [25] T. Takeyama, H. Miyama, Reaction mechanism of ammonia oxidation in shock
waves, J. Chem. Phys. 42 (1965) 3737–3738.
ments with literature data were confirmed. For N2 O and NO in the [26] T. Takeyama, H. Miyama, A shock-tube study of the ammonia-oxygen reaction,
reaction zone of weak flames, however, there was discrepancy be- Symp. (Int.) Combust. 11 (1967) 845–852.
tween measured and computational mole fractions and further im- [27] H. Miyama, R. Endoh, Vibrational relaxation of nitrogen in shock-heated
NH3 –O2 –N2 mixtures, J. Chem. Phys. 46 (1967) 2011–2012.
provements of chemical kinetics related to ammonia ignition are [28] D.C. Bull, A shock tube study of the oxidation of ammonia, Combust. Flame 12
still necessary. (1968) 603–610.
[29] L.J. Drummond, High temperature oxidation of ammonia, Combust. Sci. Tech-
nol. 5 (1972) 175–182.
[30] N. Fujii, H. Miyama, M. Koshii, T. Asaba, A shock-tube study of the ammoni-
Acknowledgments
a-oxygen reaction, Symp. (Int.) Combust. 18 (1981) 873–883.
[31] S.H. Won, B. Windom, B. Jiang, Y. Ju, The role of low temperature fuel chem-
This study was supported by Precursory Research for Embry- istry on turbulent flame propagation, Combust. Flame 161 (2014) 475–483.
[32] H. Nakamura, S. Hasegawa, Combustion and ignition characteristics of ammo-
onic Science and Technology, Japan Science and Technology Agency
nia/air mixtures in a micro flow reactor with a controlled temperature profile,
(JPMJPR1343). Proc. Combust. Inst. 36 (2017) 4217–4226.
[33] K. Maruta, T. Kataoka, N.I. Kim, S. Minaev, R. Fursenko, Characteristics of com-
bustion in a narrow channel with a temperature gradient, Proc. Combust. Inst.
References 30 (2005) 2429–2436.
[34] S. Minaev, K. Maruta, R. Fursenko, Nonlinear dynamics of flame in a nar-
[1] S. Chiuta, R.C. Everson, H.W.J.P. Neomagus, P. van der Gryp, D.G. Bessarabov, row channel with a temperature gradient, Combust. Theory Model. 11 (2007)
Reactor technology options for distributed hydrogen generation via ammonia 187–203.
decomposition: A review, Int. J. Hydrog. Energy 38 (2013) 14968–14991. [35] H. Oshibe, H. Nakamura, T. Tezuka, S. Hasegawa, K. Maruta, Stabilized three-
[2] C. Zamfirescu, I. Dincer, Ammonia as a green fuel and hydrogen source for -stage oxidation of DME/air mixture in a micro flow reactor with a controlled
vehicular applications, Fuel Process. Technol. 90 (2009) 729–737. temperature profile, Combust. Flame 157 (2010) 1572–1580.
[3] O. Kurata, N. Iki, T. Matsunuma, T. Inoue, T. Tsujimura, H. Furutani, [36] A. Yamamoto, H. Oshibe, H. Nakamura, T. Tezuka, S. Hasegawa, K. Maruta, Sta-
H. Kobayashi, A. Hayakawa, Performances and emission characteristics of bilized three-stage oxidation of gaseous n-heptane/air mixture in a micro flow
NH3 –air and NH3 –CH4 –air combustion gas-turbine power generations, Proc. reactor with a controlled temperature profile, Proc. Combust. Inst. 33 (2011)
Combust. Inst. 36 (2017) 3351–3359. 3259–3266.
 H. Nakamura et al. / Combustion and Flame 185 (2017) 16–27 27

[37] H. Nakamura, H. Takahashi, T. Tezuka, S. Hasegawa, K. Maruta, K. Abe, Effects [58] J. Bugler, K.P. Somers, J.M. Simmie, F. Güthe, H.J. Curran, Modeling nitrogen
of CO-to-H2 ratio and diluents on ignition properties of syngas examined by species as pollutants: thermochemical influences, J. Phys. Chem. A 120 (2016)
weak flames in a micro flow reactor with a controlled temperature profile, 7192–7197.
Combust. Flame 172 (2016) 94–104. [59] J. Bugler, K.P. Somers, E.J. Silke, H.J. Curran, Revisiting the kinetics and thermo-
[38] T. Kamada, H. Nakamura, T. Tezuka, S. Hasegawa, K. Maruta, Study on combus- dynamics of the low-temperature oxidation pathways of alkanes: A case study
tion and ignition characteristics of natural gas components in a micro flow re- of the three pentane isomers, J. Phys. Chem. A 119 (2016) 7510–7527.
actor with a controlled temperature profile, Combust. Flame 161 (2014) 37–48. [60] A.M. Dean, J.W. Bozzelli, Combustion chemistry of nitrogen, in: W.C. Gardiner
[39] S. Kikui, H. Nakamura, T. Tezuka, S. Hasegawa, K. Maruta, Study on combustion (Ed.), Gas-Phase Combustion Chemistry, Springer, New York, 1999 (Chapter 2).
and ignition characteristics of ethylene, propylene, 1-butene and 1-pentene in [61] Y. Daimon, H. Terashima, M. Koshi, Chemical kinetics of hypergolic ignition in
a micro flow reactor with a controlled temperature profile, Combust. Flame N2 H4 /N2 O4 –NO2 gas mixtures, J. Propuls. Power 30 (2014) 707–716.
163 (2016) 209–219. [62] P. Raghunath, Y.H. Lin, M.C. Lin, Ab initio chemical kinetics for the N2 H4 + NOx
[40] M. Hori, A. Yamamoto, H. Nakamura, T. Tezuka, S. Hasegawa, K. Maruta, Study (x = 1–3) reactions and related reverse processes, Comput. Theor. Chem. 1046
on octane number dependence of PRF/air weak flames at 1–5 atm in a micro (2014) 73–80.
flow reactor with a controlled temperature profile, Combust. Flame 159 (2012) [63] B. Ruscic, J. Berkowitz, Photoionization mass spectrometric study of N2 H2 and
959–967. N2 H3 : N–H, N=N bond energies and proton affinity of N2 , J. Chem. Phys. 95
[41] S. Suzuki, M. Hori, H. Nakamura, T. Tezuka, S. Hasegawa, K. Maruta, Study (1991) 4378–4384.
on cetane number dependence of diesel surrogates/air weak flames in a mi- [64] J.A. Pople, L.A. Curtiss, The energy of N2 H2 and related compounds, J. Chem.
cro flow reactor with a controlled temperature profile, Proc. Combust. Inst. 34 Phys. 95 (1991) 4385–4388.
(2013) 3411–3417. [65] D.P. Linder, X. Duan, M. Page, Thermal rate constants for R + N2 H2 → RH + N2 H
[42] Y. Tsuboi, T. Yokomori, K. Maruta, Lower limit of weak flame in a heated chan- (R = H, OH, NH2 ) determined from multireference configuration interaction
nel, Proc. Combust. Inst. 32 (2009) 3075–3081. and variational transition state theory calculations, J. Chem. Phys. 104 (1996)
[43] T. Okuno, H. Nakamura, T. Tezuka, S. Hasegawa, K. Maruta, Ultra-lean combus- 6298–6307.
tion characteristics of premixed methane flames in a micro flow reactor with [66] Y.-Y. Chuang, D.G. Truhlar, Improved dual-level direct dynamics method for re-
a controlled temperature profile, Proc. Combust. Inst. 36 (2017) 4227–4233. action rate calculations with inclusion of multidimensional tunneling effects
[44] A.A. Konnov, Implementation of the NCN pathway of prompt-NO formation in and validation for the reaction of H with trans-N2 H2 , J. Phys. Chem. A 101
the detailed reaction mechanism, Combust. Flame 156 (2009) 2093–2105. (1997) 3808–3814.
[45] Y. Kizaki, H. Nakamura, T. Tezuka, S. Hasegawa, K. Maruta, Effect of radical [67] R.K. Hanson, S. Salimian, Survey of rate constants in the N/H/O system, in:
quenching on CH4 /air flames in a micro flow reactor with a controlled tem- W.C. Gardiner (Ed.), Combustion Chemistry, Springer-Verlag, New York, 1984
perature profile, Proc. Combust. Inst. 35 (2015) 3389–3396. (Chapter 6).
[46] Y. Saiki, Y. Suzuki, Effect of wall surface reaction on a methane-air premixed [68] J.A. Miller, P. Glarborg, Modeling the thermal De-NOx process: closing in on a
flame in narrow channels with different wall materials, Proc. Combust. Inst. 34 final solution, Int. J. Chem. Kinet. 31 (1999) 757–765.
(2013) 3395–3402. [69] P. Glarborg, M.U. Alzueta, K. Dam-Johansen, J.A. Miller, Kinetic modeling of
[47] J.A. Miller, C.T. Bowman, Mechanism and modeling of nitrogen chemistry in hydrocarbon/nitric oxide interactions in a flow reactor, Combust. Flame 115
combustion, Prog. Energy Combust. Sci. 15 (1989) 287–338. (1998) 1–27.
[48] O. Mathieu, E.L. Petersen, Experimental and modeling study on the high-tem- [70] J.A. Miller, P. Glarborg, Modelling the formation of N2 O and NO2 in the thermal
perature oxidation of ammonia and related NOx chemistry, Combust. Flame De-NOx process, in: J. Wolfrum, H.-R. Volpp, R. Rannacher, J. Warnatz (Eds.),
162 (2015) 554–570. Gas Phase Chemical Reaction Systems, Springer-Verlag, Heidelberg, 1996.
[49] A. Kéromnès, W.K. Metcalfe, K.A. Heufer, N. Donohoe, A.K. Das, C.-J. Sung, [71] Z. Tian, Y. Li, L. Zhang, P. Glarborg, F. Qi, An experimental and kinetic modeling
J. Herzler, C. Naumann, P. Griebel, O. Mathieu, M.C. Krejci, E.L. Petersen, study of premixed NH3 /CH4 /O2 /Ar flames at low pressure, Combust. Flame 156
W.J. Pitz, H.J. Curran, An experimental and detailed chemical kinetic modeling (2009) 1413–1426.
study of hydrogen and syngas mixture oxidation at elevated pressures, Com- [72] S. Inomata, N. Washida, Rate constants for the reactions of NH2 and HNO with
bust. Flame 160 (2013) 995–1011. atomic oxygen at temperatures between 242 and 473 K, J. Phys. Chem. A 103
[50] P. Dagaut, P. Glarborg, M.U. Alzueta, The oxidation of hydrogen cyanide and (1999) 5023–5031.
related chemistry, Prog. Energy Combust. Sci. 34 (2008) 1–46. [73] P. Dransfeld, W. Hack, H. Kurzke, F. Temps, H.GG. Wagner, Direct studies of
[51] O. Mathieu, A. Levacque, E.L. Petersen, Effects of N2 O addition on the ignition elementary reactions of NH2 -radicals in the gas phase, Symp. (Int.) Combust.
of H2 –O2 mixtures: experimental and detailed kinetic modeling study, Int. J. 20 (1984) 655–663.
Hydrog. Energy 37 (2012) 15393–15405. [74] P. Glarborg, K. Dam-Johansen, J.A. Miller, R.J. Kee, M.E. Coltrin, Modeling the
[52] G. Dayma, P. Dagaut, Effect of air contamination on the combustion of hydro- thermal DENOx process in flow reactors. Surface effects and nitrous oxide for-
gen-effect of NO and NO2 addition on hydrogen ignition and oxidation kinet- mation, Int. J. Chem. Kinet. 26 (1994) 421–436.
ics, Combust. Sci. Technol. 178 (2006) 1999–2024. [75] R. Sumathi, S.D. Peyerimhoff, A quantum statistical analysis of the rate con-
[53] R. Sivaramakrishnan, K. Brezinsky, G. Dayma, P. Dagaut, High pressure effects stant for the HO2 + NH2 reaction, Chem. Phys. Lett. 263 (1996) 742–748.
on the mutual sensitization of the oxidation of NO and CH4 –C2 H6 blends, Phys. [76] NIST Chemistry Webbook (2017). Available at http://webbook.nist.gov/
Chem. Chem. Phys. 9 (2007) 4230–4244. chemistry/, (accessed 2017).
[54] S.J. Klippenstein, L.B. Harding, P. Glarborg, J.A. Miller, The role of NNH in NO [77] O. Mathieu, A. Levacque, E.L. Petersen, Effects of NO2 addition on hydrogen
formation and control, Combust. Flame 158 (2011) 774–789. ignition behind reflected shock waves, Proc. Combust. Inst. 34 (2013) 633–640.
[55] A.A. Konnov, J. De Ruyck, Kinetic modeling of the thermal decomposition of [78] K.W. Michel, H.GG. Wagner, The pyrolysis and oxidation of hydrazine behind
ammonia, Combust. Sci. Technol. 152 (20 0 0) 23–37. shock waves, Symp. (Int.) Combust. 10 (1965) 353–364.
[56] A.A. Konnov, J. De Ruyck, Kinetic modeling of the decomposition and flames of [79] G.P. Smith, D.M. Golden, M. Frenklach, N.W. Moriarty, B. Eiteneer, M. Golden-
hydrazine, Combust. Flame 124 (2001) 106–126. berg, C.T. Bowman, R.K. Hanson, S. Song, W.C. Gardiner, Jr., V.V. Lissianski and
[57] A.A. Konnov, J. De Ruyck, A possible new route for NO formation via N2 H3 , Z. Qin (1999). Available at http://www.me.berkeley.edu/gri_mech/, (accessed
Combust. Sci. Technol. 168 (2001) 1–46. 2017).