air flames

Fuel 150 (2015) 394–407

Contents lists available at ScienceDirect

Fuel journal homepage: www.elsevier.com/locate/fuel

Experimental study and modelling of NOx formation in high pressure counter-flow premixed CH4/air flames L. Pillier ⇑, M. Idir, J. Molet, A. Matynia 1, S. de Persis ICARE – Institut de Combustion, Aérothermique, Réactivité, Environnement, UPR CNRS 3021, 1C Avenue de la Recherche Scientifique, 45071 Orléans Cedex 2, France

h i g h l i g h t s  NO mole fraction profiles measured by LIF in high pressure lean CH4/air flames. Ò

 Three kinetic mechanisms: GDFkin 3.0_NCN, GRImech2.11 and GRImech3.0 are compared.  Kinetic analysis to better understand the differences between the three mechanisms.  Inclusion of the new prompt-NO formation pathway in the GRImech3.0 mechanism.

a r t i c l e

i n f o

Article history: Received 6 August 2014 Received in revised form 27 January 2015 Accepted 28 January 2015 Available online 14 February 2015 Keywords: NOx formation High pressure flames Laser Induced Fluorescence Methane combustion Kinetic analysis

a b s t r a c t Nitric oxide (NO) is an atmospheric pollutant responsible for the destruction of the ozone layer and the creation of photochemical smog. As a result, NOx emissions from combustion sources are regulated in most industrialised countries. The need to control NOx emissions while also promoting more efficient use of fossil energy resources requires a better understanding of combustion processes, especially the chemical kinetics of NOx formation. NO formation in high-pressure flames is a research area of great practical interest as high pressure exists in practically all power-generation and propulsion engines and it is known that pressure influences the combustion chemistry. In the present work, NO mole fraction profiles were measured by Laser Induced Fluorescence in laminar high pressure (up to 0.7 MPa) counterflow lean CH4/air (E.R. = 0.7) flames. Inherent problems linked to the application of the NO LIF technique in high pressure environment were addressed. The experimental NO profiles were then compared with modelling using the OPPDIF code and the three detailed kinetic mechanisms: the GDFkinÒ3.0_NCN mechanism developed by Lamoureux et al. and the two mechanisms from the Gas Research Institute: GRImech 2.11 and GRImech 3.0. A kinetic analysis based on rate of production/consumption analyses was performed to better understand the differences between the three mechanisms. Finally, the GRImech3.0 mechanism was modified with three updated prompt-NO submechanisms proposed in the literature and the consequences on the N-containing species mole fractions predictions are discussed. Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction The combustion of fossil fuels (natural gas, coal, oil) leads to a number of pollutant emissions in the atmosphere. Among these pollutants (COx, SOx, soot, etc.), nitrogen oxides NOx (NO and NO2) emitted from combustion sources are regulated in most industri⇑ Corresponding author at: Université de Lille, PC2A – PhysicoChimie des Processus de Combustion et de l’Atmopshère, UMR 8522, 59655 Villeneuve d’Ascq Cedex, France. Tel.: +33 (0)3 20 33 64 66; fax: +33 (0)3 20 43 69 77. E-mail address: [email protected] (L. Pillier). URL: http://pc2a.univ-lille1.fr/ (L. Pillier). 1 Present addresses: Sorbonne Universités, UPMC Univ Paris 06, UMR 7190, Institut Jean le Rond d’Alembert, F-75005, Paris, France and CNRS, UMR 7190, Institut Jean le Rond d’Alembert, F-78210 Saint-Cyr l’École, France. http://dx.doi.org/10.1016/j.fuel.2015.01.099 0016-2361/Ó 2015 Elsevier Ltd. All rights reserved.

alised countries. The need to control NOx emissions requires a better understanding of combustion processes, especially the chemical kinetics of NOx formation during combustion. The study of NOx formation and destruction kinetics in high-pressure flames is a research area of great practical interest as high pressure exists in practically all power-generation and propulsion engines and it is known that pressure greatly influences the combustion chemistry. NOx measurements in industrial combustion conditions are usually performed in the exhaust gases using gas analysers (chemiluminescence for NOx). While this method provides global data, it does not enable the detailed study of NOx formation mechanisms during combustion. Hence, experimental investigations within the combustion chamber (engine) or in laboratory flames are required. Most of the studies related to NOx measurements in engines [1–6], or

L. Pillier et al. / Fuel 150 (2015) 394–407

high pressure flames [7–9] have been performed using the non-intrusive Laser Induced Fluorescence (LIF) technique. However, NO LIF measurements in high pressure environments are perturbed by various spectral features: interferences with O2 fluorescence bands (Schumann Runge B3R–X3R bands), collisional broadening and shifting, and absorption by CO2 and H2O (laser attenuation and trapping). Several strategies have been proposed in the literature to minimise those perturbations. In particular, Bessler et al. [9] presented a relevant comparative study, based on previous work by their group [10–12], on the different excitation/detection strategies using the A–X(v0 = 0, v00 = 0,1,2) vibrational bands of NO for LIF measurements in CH4/air flat flames stabilised at pressures up to 6 MPa. Their main conclusions were: (i) excitation on the A–X(0, 0) band followed by detection on the A–X(0, 1) and/or A–X(0, 2) bands present the best performances in terms of selectivity and signal to noise ratio as long as the absorption and trapping phenomena are weak. The authors recommend this scheme in high pressure flames with small diameters; (ii) excitation on the A–X(0, 1) band with collection over the A–X(0, 2) and/or (0, 3) bands gives relatively high fluorescence signals and has the advantage of reducing laser absorption and trapping effects, especially if the experiment requires long signal paths; (iii) excitation through the A–X(0, 2) band with collection on the A–X(0, 0) and/or (0, 1) bands has been mainly employed for NO measurements in engines. Even if signals are weak, attenuation and interference problems are strongly reduced. NO measurements can be achieved for relatively high concentrations (>1000 ppm). Most of the studies comparing experimental and modelled NO data in high pressure flames [8,13,14] use the well-known mechanisms from the Gas Research Institute: GRImech 2.11 [15] or GRImech 3.0 [16]. In the last decade, particular attention has been paid to the prompt-NO formation pathway and it has been demonstrated that the reaction CH + N2 = HCN + N (known to be spin forbidden) [17] has to be replaced by the reaction CH + N2 = NCN + H [18,19]. Recently, Lamoureux et al. revised the GDFkinÒ3.0_NCN mechanism [20] based on their extended experimental database (NO, CH, CN, NCN, HCN, NCO) in low pressure flames. To our knowledge, this mechanism has never been validated on NO in high pressure flames, except in our previous work [21] where its performance was tested by comparison with experiments from the literature [22] and other models (GRImech 2.11, 3.0 [15,16] and Konnov6.0 [23]). In the present work, NO mole fraction profiles were measured by Laser Induced Fluorescence in laminar high pressure (up to 0.7 MPa) counter-flow lean CH4/air (E.R. = 0.7) flames. Inherent problems linked to the application of the NO LIF technique in a high pressure environment were addressed. The excitation/detection scheme was carefully chosen, based on excitation and fluorescence spectra analysis, to limit interferences on NO LIF signals and a theoretical correction procedure is proposed to take into account the influence of spectral broadening and quenching on LIF signals when pressure increases. The experimental NO profiles were then compared with modelling using the OPPDIF code [24] and the three detailed kinetic mechanisms: GDFkinÒ3.0_NCN [20], GRImech 2.11 [15] and GRImech 3.0 [16]. To better understand the behaviour of these three mechanisms with respect to their predictions of NO formation in counter-flow flames at high pressure, a kinetic analysis was performed. The aim is to evaluate the relative contribution to the overall NO concentration from each of the four major NO formation pathways (thermal, prompt, N2O and NNH). For this purpose, a sub-mechanism subtraction technique as well as rate of production and consumption analyses were used. Finally, as the GRImech3.0 mechanism still does not include the appropriate channel for the initiation reaction of prompt-NO (CH + N2 = NCN + H), it was modified with three updated prompt-NO submechanisms proposed in the literature: Lamoureux et al. [20], Konnov [23] and Williams et al. [25]. The consequences

395

on the N-containing species mole fractions predictions are discussed. 2. Experimental set-up 2.1. High pressure facility The high pressure facility used in this work was detailed previously in [26]. An overview is presented here. It consists of two twin counter-flow converging burners placed in a high pressure vessel, equipped with optical access for laser diagnostics. The burners are mounted on a vertical translation system and the distance between the nozzles can be manually adjusted by moving the top burner with respect to the bottom one. In the present study, the distance between the burners was fixed at 10 mm. Each burner is composed of two co-annular nozzles of 7 mm and 13 mm diameter, which were aerodynamically shaped according to a modified empirical calculation from Rolon [27], resulting in a nearly uniform velocity profile on their exit. A nitrogen coflow around the burner isolates the flame from the surrounding gases. The burners are cooled using a closed loop water circulation at a fixed temperature between 30 and 50 °C depending on the flame conditions to avoid water condensation at the burner surfaces. The pressure within the vessel is controlled with a pressure transducer coupled with a control valve. Gas flows are monitored by Brooks mass flow meters through a Labview program. 2.2. Flame conditions Laminar lean (E.R. = 0.7) premixed CH4/air flames (dilution ratio X(N2)/X(O2) of 3.77) were studied in this work. The pressure was varied from 0.1 to 0.7 MPa. Flame conditions are summarised in Table 1 together with adiabatic temperatures and predicted NOx (NO + NO2) mole fractions computed at thermodynamic equilibrium using STANJAN [28], as well as flame temperatures and laminar flame velocities computed for a free flame configuration using PREMIX [29] with both GRImech2.11 [15] and GRImech3.0 [16] mechanisms as well as the GDFkinÒ3.0_NCN mechanism [20]. Table 1 confirms that NO2 mole fractions can be neglected compared to NO mole fractions. As mentioned in [26,30], depending on flame conditions, the gas velocity ratio between the upper and the bottom burners needs to be adjusted in order to keep the flames well centred between the burners. This ratio, noted b in [26,30] is equal to 1.05 for all the CH4/air lean flames studied here. 2.3. LIF experimental set-up The experimental set-up is presented in Fig. 1. The laser system consists of a frequency-doubled Nd-YAG pulsed laser (Quantel Brillant B, repetition rate 10 Hz, 6 ns pulses) pumping a dye laser (Quantel TDL+). For NO excitation, a wavelength around 226 nm was obtained by mixing the frequency-doubled output of the dye laser (mixture of Rhodamine 590 and 610) with the residual infrared radiation from the Nd-YAG laser. The resulting laser beam near 226 nm has a diameter of 6 mm, and delivers a few milli-joules per pulse. The beam linewidth is 0.06 cm1 (Quantel specifications). Measurements in the linear fluorescence regime were made by reducing the beam energy per pulse to 100 lJ with a variable attenuator composed of a half-wave plate and a Glan-Taylor prism. Part of the laser beam is collected by a fast photodiode (Newport, 818-BB-22 model) in order to monitor the laser beam energy fluctuations. The beam is focused with a f = 350 mm lens inside the high pressure chamber, on the centre axis between the burners. The fluorescence signal is collected at right angle through a

396

L. Pillier et al. / Fuel 150 (2015) 394–407

Table 1 CH4/air counter-flow premixed flame conditions. Equivalence ratio (E.R. = 0.7), Dilution ratio = 3.77. Pressure (MPa)

c

X(CH4)a

0.0685

X(O2)a

0.1952

X(N2)a

0.7363

Total flowrate (NL.min1)

0.1

303

0.3

313

2.69

0.5

323

3.19

0.7

323

4.30

a

1.68

Thermodynamic equilibriumb GRIMech2.11/ GRIMech3.0/GDFkinÒ3.0_NCN

Free flamec GRIMech2.11/ GRIMech3.0/GDFkin

Ta, K

X(NO) ppm

X(NO2) ppm

Tf, K

SL, cm s1 (T 00 = 298 K)

1840/1840/ 1840 1848/1848/ 1848 1856/1856/ 1856 1857/1857/ 1857

2398/2398/ 2545 2464/2464/ 2615 2525/2525/ 2679 2527/2527/ 2681

3/3/3

1824/1827/ 1830 1832/1832/ 1838 1829/1832/ 1839 1829/1829/ 1840

19.8/19.2/17.4

3/5/5 7/7/7 8/8/8

12.1/11.2/11.4 9.1/8.3/8.9 7.4/6.7/7.4

Experimental conditions for the bottom burner with T0: initial premixture temperature (K), X(CH4), X(O2), X(N2): initial mole fraction respectively for CH4, O2 and N2. STANJAN code [28] respectively with GRImech2.11, GRImech3.0 and GDFkinÒ3.0_NCN mechanisms. PREMIX code [29] respectively with GRImech2.11, GRImech3.0 and GDFkinÒ3.0_NCN mechanisms, with T0 0: initial temperature (K).

Dye Laser

Variable attenuator Mirror

Nd-YAG

a b

To (K)a

Photodiode

f=350 mm

Oscilloscope f=500 mm

Periscope

Counter-flow burner

Beam trap

PM

Sp ect ro

me

ter

f=300 mm

Mirror

PM : Photomultiplier Fig. 1. NO-LIF experimental set-up.

f = 500 mm lens and focused with a f = 300 mm lens on the entrance slit of a 500 mm focal spectrometer (Acton Spectrapro 2500i) equipped with a photomultiplier (PM) tube (Photonis XP2020Q). A 90° rotating periscope rotates the image of the probe volume and sets it parallel to the entrance slit of the spectrometer. The entrance slit is 100 lm in width and 4 mm in height, giving a probe volume of 160 lm by 6.4 mm in the flame, according to the magnification ratio of the optical collection system. The exit slit is adjusted to select the suitable fluorescence signal collection bandpass. Detected signals are sampled with a 1 GHz bandwidth oscilloscope (Tektronix, TDS5014B model).

2.4. Spectroscopic considerations: excitation/detection scheme High pressure NO LIF measurements are arduous to perform because of the influence of several spectroscopic interferences on the NO LIF signal. First, the chosen excitation line must be well isolated to limit the overlap with neighbouring NO absorption lines when pressure increases, as collisional broadening increases with pressure. The absorption coefficient must be sufficiently high to obtain a good signal to noise ratio, whereas laser absorption and trapping need to be limited. Interferences from the Schumann– Runge bands of O2 and absorption from combustion products

397

L. Pillier et al. / Fuel 150 (2015) 394–407

Fluorescence Signal (a.u)

O2

(a)

NO

Raman N2

NO O2

O2

228

233

238

243

248

Fluorescence Signal (a.u.)

1 P1(23.5),Q1+P21(14.5), Q2+R12(20.5)

0.9

λA

0.8 λA

0.6 0.5 0.4

λB

0.3 0.2 0.1 0 225.44

225.54

225.64

225.74

225.84

225.94

226.04

226.14

Wavelength (nm) Fig. 2. Experimental NO excitation spectrum (black line) measured in the burned gases of a stoichiometric CH4/air counter-flow flame at 0.1 MPa and calculated O2 excitation spectrum (red line) using the Lifsim code [33] (T = 2000 K, resolution = 0.001 nm). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

centred at 233, 238, 241, 245 and 249 nm). To identify the nature of these contributions (shown in Fig. 3), fluorescence spectra were also measured in a CH4/O2/Ar flame (not shown here) without any nitrogen source. Spectra in Fig. 3a and b clearly show that, at atmospheric pressure, an excitation via the P1(23.5), Q1 + P21(14.5), Q2 + R12(20.5) line in the A–X(0, 0) band with fluorescence signal collection through the A–X(0, 1) band (centred at 236 nm with a 2.8 nm bandpass) is the best compromise to maximise the NO LIF signal and minimise interferences in our flame conditions. Fig. 3c shows that O2 LIF interferences and background signal

NO

Raman N2

O2

O2 O2

228

253

(b)

NO

233

238

243

248

253

Wavelength (nm)

Wavelength (nm) NO

Fluorescence Signal (a.u.)

λB

Q2(26.5)

0.7

Fluorescence Signal (a.u)

(CO2 and H2O) need to be minimised. Finally, the variations of the Boltzmann fraction with temperature must be weak. In most studies relating NO measurements by LIF in high pressure flames within the A–X (0, 0) band, excitations via the Q2(26.5) line at 225.58 nm [31] or the P1(23.5), Q1 + P21(14.5), Q2 + R12(20.5) feature at 226.03 nm [32] are recommended. Fig. 2 represents an excitation spectrum measured in the burned gases of a stoichiometric CH4/air counter-flow flame at atmospheric pressure (the signal is collected over the A–X (0, 1) band). An O2 excitation spectrum was calculated using the Lifsim code [33] and superposed on the NO experimental spectrum. For each excitation feature (Q2(26.5) and P1(23.5), Q1 + P21(14.5), Q2 + R12(20.5)), the wavelength positions corresponding to maximum and minimum NO signals are depicted (kA for excitation line peak, kB for minimum NO signal). The ‘‘background’’ wavelengths kB were chosen following the recommendations in [31,32] and yield to a comparable intensity of O2-LIF compared to the excitation line peak position. In our flame conditions, the P1(23.5), Q1 + P21(14.5), Q2 + R12(20.5) candidate presents the highest signal to noise ratio and the lowest O2 interferences. Fluorescence spectra were measured in the burned gases of the lean (E.R. = 0.7) CH4/air counter-flow flame for both excitation features: Fig. 3a for the Q2(26.5) line at 0.1 MPa and Fig. 3b and c for the P1(23.5), Q1 + P21(14.5), Q2 + R12(20.5) feature, at 0.1 MPa and 0.7 MPa respectively. The lean flame was chosen to maximise the interferences from O2. The black spectrum is the fluorescence spectrum with laser excitation ‘‘on resonance’’ (kA) with the absorption line and the grey spectrum represents the fluorescence spectrum with laser excitation ‘‘off resonance’’ (kB) with a minimum contribution of NO fluorescence. The fluorescence spectra clearly show the contributions of interference sources (emission bands

(c)

O2

NO

Raman N2 O2

O2

228

233

238

243

248

253

Wavelength (nm) Fig. 3. Experimental fluorescence spectrum measured in the burned gases of a lean (E.R. = 0.7) CH4/air counter-flow flame: (a) excitation according to the Q2(26.5) line on resonance (kA, in black) and off resonance (kB, in grey) at 0.1 MPa; (b) and (c) excitation according to the P1(23.5), Q1 + P21(14.5), Q2 + R12(20.5) line on resonance (kA, in black) and off resonance (kB, in grey) at 0.1 MPa and 0.7 MPa respectively. Collection bandpass = 1.66 nm.

398

L. Pillier et al. / Fuel 150 (2015) 394–407

contribution to the A–X(0, 1) NO band increase with increasing pressure (Note that the NO LIF profiles measurements and their calibration were systematically performed on resonance (kA) and ‘‘off resonance’’ (kB)). The same conclusion was reached by Bessler et al. [9] in high pressure flat flames, as Naik et al. [34] found no particular improvement using the P1(23.5), Q1 + P21(14.5), Q2 + R12(20.5) feature compared to the Q2(26.5) line in their high pressure counterflow diffusion flames. 2.5. Analysis of the LIF signal The LIF signal was analysed with the same procedure as in Matynia et al. [26], employing a proportionality equation between the fluorescence signal Sf and the probed species density population N in the linear fluorescence regime. In the present case, the LIF measurements of NO were performed via the excitation of a set of different neighbouring lines: P1(23.5), Q1 + P21(14.5), Q2 + R12(20.5) in the A–X(0, 0) band, which are not exactly centred at the same wavelength. In this case, the fluorescence signal Sf can be expressed as:

Sf / NNO  EL

(

gf 

X

) 00

00

½g j ðY L ; Y NO ; dkj Þ  F b;j ðJ ; t ; TÞ 

BJj0 J00 

ð1Þ

j

where j represents each line of the set, EL is the laser energy per pulse, NNO is the NO density population, gf is the fluorescence quantum yield, g j ðY L ; Y NO ; dkj Þ is the spectral overlap function of the laser line with each absorption line with YL the spectral laser line shape, YNO the absorption line shape, dkj the spectral shifting between the laser line and each absorption centreline (considering that the laser line is centred on one of the absorption lines of the group), F B;j ðJ00 ; t00 ; TÞ is the Boltzmann fraction of each laser pumped rotational level (J00 ,t00 ) at the temperature T in K and BjJ0 J00 is the Einstein absorption coefficient (with J0 and J00 the upper and lower rotational levels, respectively) for each line j. The fluorescence quantum yield and the terms between brackets in Eq. (1) were calculated as shown in the following sections. 2.5.1. Spectral overlap function calculation The spectral overlap function gj(YL,YNO,dkj) was calculated for each line assuming a Gaussian line shape for the laser line YL with a Full Width at Half Maximum (FWHM) of 0.06 cm1 (Quantel specifications). Note that the spectral width of our laser is very thin compared to the work by Bessler et al. [10] (0.4 cm1). The absorption line shape YNO was calculated from the convolution product of a Lorentzian profile, due to collisional broadening, with a Gaussian one, due to Doppler broadening. The spectral overlap was determined by calculating the integral of the product of two line shapes YL and YNO with their areas normalised to unity. The laser line wavelength was centred on the Q1 + P21(14.5) line and the corresponding spectral shift dkj was taken into account for the P1(23.5) and Q2 + R12(20.5) lines. 2.5.2. Fluorescence quantum yield calculation P ðAJ0 J00 Þ obs The fluorescence quantum yield gf ¼ A þQ represents the eff

eff

ability of a molecule to emit fluorescence (i.e. spontaneous emission with Einstein coefficients AJ0 J00 ) in the experimental observation spectral window compared to the overall de-excitation (radiative and non-radiative) processes. In a multilevel system, these de-excitation processes are characterised by an effective spontaneous emission coefficient Aeff and by an effective quenching rate Qeff. According to this definition of the quantum yield, collisional population redistributions in the excited level such as Rotational Energy Transfers (RET) and Vibrational Energy Transfers (VET)

must be taken into account. However, the calculations of Aeff and Qeff are very complex due to the RET and VET effects and are strongly dependent on the collision rates of NO with neighbouring atoms and molecules. A few studies have been devoted to the effects of RET and VET on NO fluorescence measurements in atmospheric pressure and high pressure flames. Ravikrishna et al. [35] performed a comparison between laser-saturated fluorescence (LSF) and linear LIF measurements of NO in counter-flow diffusion flames at 0.1 MPa and showed that both techniques gave nearly identical results, suggesting that RET are small at atmospheric pressure. Driscoll et al. [36] applied picosecond-LIF to obtain spatial profiles of NO concentrations and effective fluorescence lifetimes in a counterflow, non-premixed CH4/air flame at atmospheric pressure. Their evaluation of picosecond-LIF signals using a four-level, densitymatrix model suggested that RET effects only become important at laser fluence higher than 100 mJ/mm2  pulse. Naik and Laurendeau [34] used a five-level model for NO molecular dynamics to investigate the effects of Rotational Energy Transfer (RET) on linear LIF measurements of NO at pressures up to 1.5 MPa. The results indicate that rotational relaxation effects are negligible under high-pressure conditions at low laser fluences, and thus do not need to be accounted for when measuring [NO] using linear LIF. According to Daily and Roth [37], VET are so slow in the X state that during the 10 ns laser pulse essentially no VET take place. Based on the above conclusions, the fluorescence quantum yield gf was quantified in this work assuming a simplified two-level system. In this simplified approach, the de-excitation collision rate QJ0 J00 (s1) can be expressed as follows:

Q J0 J00 ¼ Ntot 

X

vi ri v i

ð2Þ

i

where: i is summed among all the collision partners, vi is the corresponding molar fraction, ri is the quenching cross section (m2), vi the average relative velocity between NO and the collision partner i (m s1) and Ntot is the total density population (molecules. m3). We calculated the quenching rate by considering that the major colliding species are: CH4, CO2, O2, H2, CO, H2O, N2 and NO itself. Spatial profiles of temperature and concentration of those species along the centreline of the counter-flow flames were calculated with the OPPDIF code [24] and the GRImech 3.0 mechanism [16] in the case of adiabatic flames. The corresponding quenching cross sections were calculated as a function of temperature, based on the models of Paul et al. [38,39] and Settersten et al. [40] and compared to available experimental values from the literature. Fig. 4 represents the NO mole fraction profile (left-hand side scale) simulated by OPPDIF and the quenching rate profiles (right-hand side scale) calculated using the model of Paul et al. [38,39] and Settersten et al. [40] in the lean CH4/air flame (E.R. = 0.7) at 0.1 MPa and 0.7 MPa. Results show that the quenching rate varies essentially in the NO concentration gradients corresponding to the flame fronts where temperature and flame composition vary significantly. These variations are much weaker in the burned gases as temperature and gas composition are nearly constant in this area. The quenching rate increases proportionally with pressure. The differences observed between the model of Paul and that of Settersten are within 20%. As the model of Settersten gives quenching cross sections only for CO2, H2O and NO colliding species, the following calculations were done using the model of Paul. 2.5.3. Boltzmann fraction calculation The Boltzmann fraction was calculated as a function of temperature using the spectroscopic constants from [41,42] for each rotational number J00 = 23.5; 14.5 and 20.5. The temperature profile

399

8.0E+08

6.0E-06

7.0E+08 6.0E+08

5.0E-06 4.0E-06

(a)

5.0E+08

0.1 MPa

4.0E+08

3.0E-06

3.0E+08

2.0E-06

2.0E+08

1.0E-06

1.0E+08

0.0E+00

0.0E+00

0

0.2

0.4

0.6

0.8

1

Distance from the bottom burner (cm)

7.0E-06

6.0E+09

6.0E-06

5.0E+09

5.0E-06

4.0E+09

4.0E-06 3.0E+09

(b) 0.7 MPa

3.0E-06

2.0E+09

2.0E-06

1.0E+09

1.0E-06 0.0E+00

-1

7.0E-06

Quenching rate (s )

9.0E+08

NO mole fraction

8.0E-06

Quenching rate (s-1)

NO mole fraction

L. Pillier et al. / Fuel 150 (2015) 394–407

0.0E+00 0

0.2

0.4

0.6

0.8

1

Distance from the bottom burner (cm)

Fig. 4. NO mole fraction profile (left-hand side scale) simulated by OPPDIF [24] and quenching rate (s1) profiles (right-hand side scale) calculated using the model of Paul et al. [38,39] (solid line) and Settersten et al. [40] (dotted line) in a lean CH4/air flame (E.R. = 0.7) at 0.1 MPa (a) and 0.7 MPa (b).

was simulated by the OPPDIF code [24], with the GRImech3.0 mechanism [16], in free flame configuration. 2.5.4. Influence on LIF signal In order to evaluate the influence of the spectral overlap, the quantum yield and the Boltzmann fraction on the fluorescence signal profile, we compared both relative spatial profiles of the NO density population NNO, calculated by the OPPDIF code [24] in conjunction with the GRImech3.0 mechanism [16] with NNO multiP plied by Q 10 00  j ½g j ðY L ; Y NO ; dkj Þ  F b;j ðJ 00 ; v 00 ; TÞ  BJj0 J00 ; that represents

measured at two different excitation wavelengths kA and kB, which have maximum and minimum NO signal strengths respectively, as illustrated in Fig. 5. After measuring the NO LIF signal in the burned gases of each flame, at the two different wavelengths, for different NO seeding concentrations, the overall signal was linearly fit as a function of NO seeded for each of the excitation wavelengths. These two linear fits were extrapolated to their intersection, yielding the nascent NO concentration and the value of the baseline signal. This procedure was applied in each individual flame with an uncertainty of ±10% for the nascent NO concentration.

J J

an image of Sf/EL – see Eq. (1). The values of the absorption coefficients BJ0 J00 were taken from the LIFBASE database [41]. Our calculations (not presented here) showed that the combined influence of all three parameters is very weak along the profile between the flames. These observations lead to the conclusion that, in our conditions, the overall influence of the spectral overlap, quantum yield and Boltzmann fraction variations can be neglected across the flame. Consequently, at a given pressure, the fluorescence signal profile (Sf/EL) accurately reproduces the NNO population density profile. Hence, in order to determine the absolute concentration of NO, a calibration measurement of the fluorescence signal at only one point in the burned gases is sufficient. However, this calibration needs to be done for each flame condition (at each pressure). In the present work, this calibration phase was done by doping the flame with known quantities of NO and measuring a calibration plot. This procedure is detailed in the next section. Furthermore, we calculated the sensitivity of the fluorescence signal (i.e. Sf/EL) regarding the temperature variations in the burned gases of the lean CH4/air flame at 0.1 MPa and 0.7 MPa. We found that a variation of ±100 K around 1800 K induces variations of the fluorescence signal of ±4% at 0.1 MPa and ±1% at 0.7 MPa. Then, such uncertainty on the temperature does not have an important impact on the calculated fluorescence signal and consequently on the determined NO concentration. At higher pressure, this sensitivity appears to be almost negligible. 2.6. Calibration The nascent NO concentration in the flames was determined using a variable NO seeding (from 0 to 111 ppm of NO in the mixture) method [7,12]. It was assumed that doped NO does not react through the flame (no NO-reburning). This assumption was supported by computer modelling which indicated that reburning was below 3% for our lean CH4/air flames. The LIF NO signal was

3. Modelling Simulations were performed using the OPPDIF code [24] for the counter-flow flames, the PREMIX code [29] for laminar flame velocity calculations and the STANJAN code [28] for thermodynamic equilibrium calculations, from the CHEMKIN-II package [43]. Three detailed kinetic mechanisms were compared: the two versions of the Gas Research Institute Mechanism, GRImech2.11 [15] and GRImech3.0 [16], and the recently updated GDFkinÒ3.0_NCN [20] mechanism. OPPDIF [24] is a program that computes the steady-state solution for axisymmetric or planar premixed or diffusion flames between two opposing nozzles. The one-dimensional model predicts the species, temperature, and velocity profiles. The thermodynamic and transport property files provided with each mechanism were employed. Calculations were performed by solving the energy equation in the case of an adiabatic and isobaric system. Multi-component diffusion and thermal diffusion (Soret effect) options were taken into account for all the calculations presented here. The main input parameters for the OPPDIF code are the burner exit velocities of the fresh gases premixture flows, the temperature of the fresh gases, the radial gradient in inlet velocity for each burner (A set to 0 for all the present calculations) and reactant mole fractions. The adaptive mesh parameters GRAD and CURV were reduced to 0.1 for all simulations, resulting in an average number of 350 grid points and providing a grid-independent calculation. The PREMIX code [29], which allows the calculation of flame velocity, temperature profiles and species mole fraction profiles in premixed laminar flames, was also used. The present calculations were performed with the freely propagating flame option. Finally, STANJAN [28] was used to predict the thermodynamic equilibrium of an ideal gas in adiabatic and isobaric conditions. The GRImech3.0 mechanism [16], containing 53 species involved in 325 reversible reactions, is a reference mechanism

L. Pillier et al. / Fuel 150 (2015) 394–407

Fluorescence signal (a.u.)

400

λA

λB 0

20

40

60

80

100

120

X(NO) seeded (ppm) Fig. 5. Calibration method: LIF NO signals measured in the burned gases as a function of NO mole fraction seeded in the CH4/air lean (E.R. = 0.7) flame at 0,1 MPa, at the two wavelengths kA and kB.

largely employed in the literature concerning methane and natural gas combustion. The previous version of this mechanism, GRImech2.11 [15] (49 species, 277 reactions), was also tested, as it was previously shown [21] that better agreement is obtained when comparing simulated and experimental NO profiles in flat and counter-flow CH4/O2/N2 flames at atmospheric and high pressure from reference [22]. This point is also discussed in [44,45]. Some significant changes were made to GRImech3.0 compared to the 2.11 version, even though 203 reactions appear in both mechanisms with identical rate constants expressions. Four species (C3H7, C3H8, CH2CHO and CH3CHO) were added in GRImech3.0 in order to describe the combustion of natural gas. Five pressuredependent reactions were added in GRImech3.0 and some thirdbody reactions were replaced by pressure-dependent reactions. Concerning the NO reaction pathways, the main formation paths of NO in flames have been identified: – the thermal NO [46]

N2 þ O ¼ NO þ N

ðR:1Þ

N þ O2 ¼ NO þ O

ðR:2Þ

N þ OH ¼ NO þ H

ðR:3Þ

– the prompt-NO [17]

CH þ N2 ¼ HCN þ N

ðR:4Þ

– the NNH [47]

N2 þ H ¼ NNH

ðR:5Þ

NNH þ O ¼ NH þ NO

ðR:6Þ

– and the N2O routes [48]

N2 þ OðþMÞ ¼ N2 OðþMÞ

ðR:7Þ

N2 O þ H ¼ NO þ NH

ðR:8Þ

N2 O þ O ¼ 2NO

ðR:9Þ

The most significant changes with respect to GRImech2.11 concern the prompt-NO and the N2O routes. The NNH route is identical in both mechanisms. As shown previously, the main formation paths of NO in flames have been identified: the thermal NO [46], the prompt-NO [17], the NNH [47] and the N2O routes [48]. However, in the last decade, special attention has been paid to prompt-NO formation and it has been demonstrated that the reaction CH + N2 = HCN + N (known to be spin forbidden) has to be replaced by the reaction CH + N2 = NCN + H [18,19]. In 2010, Lamoureux et al. [49] reported an experimental and numerical study of the role of NCN in prompt-NO formation in low pressure (5.3 kPa) CH4–O2–N2 and C2H2–O2–N2 premixed flames. In their study, the authors obtained new experimental data and improved the understanding of prompt-NO formation by demonstrating the role of NCN in acetylene rich flames. Their large experimental database allowed them to revise the GDFkinÒ3.0_NCN

mechanism by varying the rate constants of sensitive reactions over their range of accuracy to obtain the best compromise. This mechanism is composed of 883 reactions and 119 species. The ability and limitations of this revised comprehensive kinetic mechanism GDFkinÒ3.0_NCN to predict NO formation in high-pressure flames were tested by comparing modelling to experimental NO measurements taken from the available literature [21]. Very recently, GDFkinÒ3.0_NCN was updated by Lamoureux et al. [20]: the NCN thermodynamic data were updated by considering the heat capacity (Cp) values reported by Goos et al. [50], but the heat of formation (450.2 kJ mol1) was conserved. The fuel oxidation part as well as the prompt-NO sub-mechanism (with the appropriate initiation reaction involving NCN: CH + N2 = NCN + H, see above) correspond to those validated at low-pressure in CH4/O2/N2 flames and two acetylene flames [20,49]. For our work, in order to reduce the calculation time and convergence problems, this mechanism was truncated and limited to the oxidation of C1–C3 hydrocarbons (+NOx chemistry), suitable for methane combustion. The good reproducibility of the results between calculations performed with the complete and the simplified mechanisms was checked. The GDFkinÒ3.0_NCN mechanism was employed in its atmospheric pressure version. To better understand the behaviour of the reaction mechanisms with respect to their predictions of NO formation in premixed counter-flow flames at atmospheric and high pressure, a reaction pathway analysis was performed. This kinetic analysis is essentially based on the rates of production and consumption as a function of the distance from the bottom burner (see Section 4.2). 4. Results and discussion 4.1. Comparison between experimental and calculated NO mole fraction profiles Fig. 6 compares the experimental and calculated NO mole fraction profiles (with the three detailed kinetic mechanisms used: GRImech 2.11 [15], GRImech 3.0 [16] and GDFkinÒ3.0_NCN [20]) in CH4/air flames (E.R. = 0.7) at P = 0.1–0.7 MPa. The experimental uncertainties reached ±20% for all flames. This value takes into account the following uncertainties: (i) uncertainties on the calibration procedure (around ±10%); (ii) uncertainties on the mass flow meters (implying uncertainties on the equivalence ratio and dilution ratio); (iii) uncertainties on the probe volume position; (iv) uncertainties on the flame temperature (estimated to be equal to the adiabatic flame with an uncertainty of ±5%); (v) the uncertainty on pressure inside the combustion chamber (±1%). As shown in Fig. 6, the NO mole fraction profiles present a bell shape due to thermal-NO formation (see Section 4.2.1) in the burned gases. The experimental maximum NO mole fraction increases slightly as the pressure increases from 6.9 ± 1.3 ppm at 0.1 MPa to 8.0 ± 1.6 ppm at 0.7 MPa. NO profiles simulated with the GDFkinÒ3.0_NCN and the two GRImech (2.11 and 3.0) mechanisms satisfactorily reproduce the experimental NO profile shapes for all the pressure conditions. Flame front positions are in good agreement with experiments for the bottom burner whereas one can note some discrepancies for the upper burner at high pressure. Maximum NO mole fractions are correctly predicted by the three mechanisms (within the error bars), with a slight overestimation for the GDFkinÒ3.0_NCN at pressure P > 0.3 MPa. Note that at 0.7 MPa, GRImech2.11 is in excellent agreement with experimental results. 4.2. Kinetic analysis In the present paper, a kinetic analysis was done in order to better understand the behaviour of each mechanism with respect to its prediction of NO formation in premixed counter-flow flames by determining the contribution of each sub-mechanism to the overall

401

10

-5

8.00x10

-6

0.1 MPa

NO mole fraction

NO mole fraction

L. Pillier et al. / Fuel 150 (2015) 394–407

6.00x10-6

4.00x10

-6

10

-5

8.00x10

-6

6.00x10-6

4.00x10

-6

2.00x10-6

2.00x10-6

0.00x100

0.00x100

0

0.2

0.4

0.6

0.8

0.3 MPa

1

0

Distance from the bottom burner (cm)

0.2

10

-5

8.00x10

-6

8.00x10

-6

NO mole fraction

NO mole fraction

-5

6.00x10-6

-6

4.00x10

0.00x100

0.00x100 0.4

1

-6

2.00x10-6

0.2

0.8

6.00x10-6

2.00x10-6

0

0.6

0.7 MPa

0.5 MPa 10

4.00x10

0.4

Distance from the bottom burner (cm)

0.6

0.8

1

Distance from the bottom burner (cm)

0

0.2

0.4

0.6

0.8

1

Distance from the bottom burner (cm)

Fig. 6. Comparison between experimental and calculated NO mole fraction profiles for the CH4/air flames (E.R. = 0.7) at pressures from P = 0.1 to 0.7 MPa. Symbols: experiments; modelling with black dashed line: GRImech 2.11; black solid line: GRImech3.0; and red solid line: GDFkinÒ3.0_NCN mechanism. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

NO concentration from each of the four major formation pathways (Zeldovich [46], prompt [17–19], NNH [47] and N2O [48]). As mentioned in Thomsen [22], this can be done by the substraction method or the addition method. The former consists in removing the initiation reaction for a given pathway and the latter in including only the relevant kinetics for a given pathway. Then, the effect of the sub-mechanism is deduced by comparing to the result given with the full kinetic model. These methods can only be carried out if the concentrations of other species are not affected by the subtraction or the addition method. This is due to inter-pathway dependencies. In our case, a preliminary study showed that it was only possible to conduct this type of analysis for the Zeldovich mechanism (thermal formation of NO) because it is the only sub-mechanism in which NO is formed directly from N2 without any interpathway. The result of this analysis is shown in the next section. In order to evaluate the contribution of the other sub-mechanisms (prompt, NNH and N2O), another way is to perform rate of production and rate of consumption analyses. This is shown in Section 4.2.2. 4.2.1. Contribution of thermal route to the formation of NO For this study, the thermal pathway contribution was determined in the following manner. The thermal NO contribution

was determined by removing the initiation reactions (R.1–R.3) in each reaction mechanism. Fig. 7 shows the comparison between the full kinetic model and the subtracted model (i.e. without the thermal NO pathway) for the three mechanisms (GRImech2.11, GRImech 3.0 and GDFkinÒ3.0_NCN) for E.R. = 0.7 and pressures from 0.1 to 0.7 MPa together with the calculated temperature profile. Fig. 7 shows that this method causes no perturbation on the ‘‘bell shape’’ NO profile at 0.1 MPa. For higher pressure, the calculated profile shapes without the NO thermal sub-mechanism are modified (equilibrium is reached earlier, and the burned gases exhibit a flat profile, similar to the temperature profile). For each mechanism, removing the NO thermal sub-mechanism has an effect on the NO concentration. The peak NO mole fractions for the full and the subtracted model are shown in Table 2 for each reaction mechanism for P = 0.1– 0.7 MPa. The percentages of NO-thermal contribution to the total peak NO are given in brackets in Table 2. It shows that the contribution of the NO thermal sub-mechanism increases slightly with pressure for GRImech2.11 and GRImech3.0. For GDFkinÒ3.0_NCN, the contribution varies around an average value of 20%. As shown in Fig. 7, the shapes of the NO profiles are affected by the removal of the NO thermal sub-mechanism at high pressure. This method gives a good assessment of the contribution of the thermal NO

402

L. Pillier et al. / Fuel 150 (2015) 394–407

10

0.1 MPa

-5

2000

-6

4.00x10

-6

2.00x10

-6

0.00x10

1200

800

8.00x10

400

0

0.2

0.4

0.6

0.8

-6

1600

6.00x10-6

4.00x10

1200

-6

800

2.00x10-6

0 0

NO mole fraction

NO mole fraction

6.00x10

1600

400

0.00x100

1

0 0

Distance from the bottom burner (cm)

0.2

-6

2000

10

1600

8.00x10

-6

800

-5

1600

6.00x10 -6

1200

4.00x10

-6

800

400

400

2.00x10 -6

0.00x100

0

0.00x10 0

0.4

0.6

0.8

2000

-6

2.00x10-6

0.2

1

1

Distance from the bottom burner (cm)

Temperature (K)

1200

0

0.8

T(K)

NO mole fraction

T(K)

6.00x10-6

4.00x10

0.6

0.7 MPa

-5

Temperature (K)

NO mole fraction

8.00x10

0.4

Distance from the bottom burner (cm)

0.5 MPa 10

2000

T(K)

Temperature (K)

-6

Temperature (K)

8.00x10

0.3 MPa

-5

10

T(K)

0 0

0.2

0.4

0.6

0.8

1

Distance from the bottom burner (cm)

Fig. 7. Comparison between calculated NO mole fraction profiles (left-hand side axis) for the full mechanism (solid line) and the subtracted model (dashed line) calculated for the CH4/air flames at E.R. = 0.7 and pressures from P = 0.1 to 0.7 MPa. Symbols: experiments; modelling with green line: GRImech 2.11; blue line: GRImech 3.0; red line: GDFkinÒ3.0_NCN mechanism and orange line: temperature profile (right-hand side axis). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 2 Calculated NO peak concentration in ppm (contribution of NO thermal sub-mechanism to total NO formation is given in percentage, %).

GRImech2.11 GRImech2.11 without NO thermal sub-mechanism GRImech3.0 GRImech3.0 without NO thermal sub-mechanism GDFkinÒ3.0_NCN GDFkinÒ3.0_NCN without NO thermal sub-mechanism

P = 0.1 MPa

P = 0.3 MPa

P = 0.5 MPa

P = 0.7 MPa

8.1 6.2 (24%) 7.2 5.9 (18%) 6.2 5.1 (18%)

8.4 6.0 (28%) 6.6 5.1 (24%) 8.1 6.8 (16%)

8.4 5.7 (32%) 6.5 4.7 (28%) 9.6 7.5 (22%)

8.5 5.4 (36%) 6.4 4.4 (31%) 10.2 8.1 (21%)

sub-mechanism; unfortunately it is impossible, with this technique, to evaluate the contribution of the other three sub-mechanisms (prompt, N2O and NNH) because removing one of the submechanisms will have an impact on the other ones. That is why rate of production and rate of consumption analyses were performed.

mechanisms GRImech3.0 and GDFkinÒ 3.0_NCN. In order to show the relative contribution to the overall NO formation by each of the four major NO pathways (Zeldovich, prompt, N2O and NNH), ROP and ROC analyses were integrated over the whole flame domain.

4.2.2. Rate Of Consumption (ROC)/Rate Of Production analyses (ROP) A kinetic analysis based on the rates of production (ROP) and consumption (ROC) as a function of the distance from the bottom burner was performed. Rates of production and consumption were computed for each species with CHEMKIN [43] for the two

4.2.2.1. Rate Of Consumption (ROC) analysis. The detailed ROC diagrams drawn for GRImech3.0 and GDFkinÒ3.0_NCN are shown in Fig. 8 for pressures varying from 0.1 to 0.7 MPa and E.R. = 0.7. For clarity reasons, the results obtained for GRImech2.11 are not shown: they are similar to those obtained with GRImech3.0 (the

403

L. Pillier et al. / Fuel 150 (2015) 394–407

(a) ROC GRI 0.1MPa/0.3MPa/0.5MPa/0.7MPa

100/100/100/100

HNCO

1/0/0/0

25/21/27/25

HCNN

2/2/3/4

N2O

42/37/37/38

86/84/83/83 3/5/7/7

100/100/100/100

30/23/20/18 24/14/10/7

NNH

43/55/61/65

18/10/7/5

N 5/2/4/4 93/97/94/94 1/1/1/1

NH

CN

NH3

NH2 58/54/53/47

39/48/50/49

39/41/44/49

N2

81/89/92/94

4/3/3/4

HOCN

1/1/1/1

62/58/55/ 52

HCN 4/5/6/9

100/100/100/100

H2CN

39/49/54/57

NCO

NO NO

HNO

NO2

ROC GDF 0.1MPa/0.3MPa/0.5MPa/0.7MPa

(b) 88/80/73/66

HOCN 11/20/27/34

HNCO

23/10/16/9

CN

3/4/5/5 7/11/13/15

62/56/51/47

HCN

12/20/23/25 70/65/62/60

0/2/2/1

13/22/25/26

96/98/99/99

NCN

NNH

H2NO N 0/0/1/1

100/100/100/100

54/65/73/74

15/14/14/15

NH

24/20/18/17

66/72/76/79

N2O 4/6/7/7

45/ 27/23/21

85/80/82/82

11/16/6/18

N2

41/52/56/56

3/3/3/3

NH2 19/21/20/17 100/100/100/100 NH3

47/63/68/70

H2CN

NCO 27/31/31/30

NO NO

NO2

HNO 4/6/5/6

Fig. 8. Integrated reaction pathways obtained from ROC for E.R. = 0.7 in premixed CH4/air counter-flow flames at four pressures (0.1–0.7 MPa) for: (a) GRImech3.0 mechanism; (b) GDFkinÒ3.0_NCN mechanism. Note that the percentages indicated on the arrows correspond to the consumption of the starting species. For clarity reasons, contributions lower than 1% are not represented.

values are not strictly the same, but the conclusion reached concerning the contribution of each mechanism is the same). These diagrams show how intermediate species leading to the formation of NO are consumed, starting from N2 to NO. If one considers N2 consumption, the relative contribution of each formation pathway can be calculated. The results are given in Fig. 9 as a function of pressure, where it can be seen that for both mechanisms, N2 is mainly decomposed through the N2O and NNH pathways.

Moreover, the contribution of N2O increases with pressure, the one concerning NNH decreases whereas the thermal one remains almost constant. The contribution of the prompt formation pathway decreases slightly for GRImech3.0 and remains constant for GDFkinÒ 3.0_NCN. This ROC analysis shows how the NO precursor in the chemical system decomposes, i.e. how N2 is consumed to finally form NO. This gives, however, no information on the production of NO.

404

L. Pillier et al. / Fuel 150 (2015) 394–407

(b) GDFkin3.0_NCN

(a) GRIMech3.0

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 0.1 MPa

0.3 MPa

0.5 MPa

0.1 MPa

0.7 MPa

prompt

N2O

thermal

0.3 MPa

0.5 MPa

0.7 MPa

NNH

Fig. 9. Relative contribution of the four NO formation pathways (the thermal NO in red, the N2O route in green, the prompt-NO in blue and the NNH in purple) obtained from N2 consumption rates, for E.R. = 0.7 as a function of pressure in premixed CH4/air counter-flow flames with: (a) GRImech3.0 mechanism; (b) GDFkinÒ3.0_NCN mechanism. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

(a) GRImech3.0

(b) GDFkin3.0_NCN

100

100

90

90

80

80

70

70

60

60

50

50

40

40

30

30

20

20

10

10 0

0 0.1 MPa

0.3 MPa

0.5 MPa

0.1 MPa

0.5 MPa

thermal

N2O

prompt

0.3 MPa

0.5 MPa

0.5 MPa

NNH

Fig. 10. Relative contribution of the four NO formation pathways (the thermal NO in red, the N2O route in green, the prompt-NO in blue, the NNH in purple) obtained from NO production rates, for E.R. = 0.7 as a function of pressure in premixed CH4/air counter-flow flames with: (a) GRImech3.0 mechanism; (b) GDFkinÒ3.0_NCN mechanism. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

4.2.2.2. Rate of production (ROP) analysis. In order to evaluate the relative contribution of each NO formation pathway, integrated rates of NO production were calculated and the contribution in percentage was computed. Only reactions directly involving NO were considered. For the thermal formation pathway, ROP of reactions (R.1) to (R.3) were considered for each mechanism. For the N2O formation pathway, ROP of reactions (R.8) and (R.9) were considered for each mechanism and for the NNH formation pathway, ROP of reactions (R.5) and (R.6) were considered for the GRImech3.0 mechanism and reactions (R.5) and (R.60 ): NH2 + NO = NNH + OH for GDFkinÒ3.0_NCN. This analysis shows that a great part of NO is due to NO2 reforming NO via reactions (R.10) and (R.11) (this can be clearly seen in Fig. 8, for the reverse process: 100% of NO2 gives NO back):

as follows: (1st) prompt; (2nd) thermal; (3rd) N2O and (4th) NNH. In the GDFkinÒ3.0_NCN mechanism, for all pressures, the prompt NO formation pathway dominates and the NNH formation pathway has no contribution (lower than 1%). The thermal formation pathway has the lowest contribution. These results show that the two mechanisms exhibit very different behaviours regarding NO production pathways contribution. The prompt-NO pathway is a major contributor in both mechanisms. It is important to note that the prompt-NO sub-mechanism implemented in GRImech3.0 takes into account the initiation reaction via HCN (spin forbidden) rather than via NCN. We then investigated the influence of the new prompt NO formation (via NCN) pathway in GRImech3.0. This is shown in the next section.

NO2 þ O ¼ NO þ O2

ðR:10Þ

4.3. Inclusion of the new prompt-NO formation pathway in the GRImech3.0 mechanism

NO2 þ H ¼ NO þ OH

ðR:11Þ

The contribution of NO2 was then removed and the prompt formation pathway was deduced by computing the difference between 100% and the sum of the other three (thermal, N2O and NNH) contributions. The results are shown in Fig. 10 for both mechanisms as a function of pressure. This rate of production analysis shows the different behaviour of each mechanism. In the GRImech3.0 mechanism, for all pressures, the contributions of each sub-mechanism are

GRImech3.0 was modified with three different prompt-NO submechanisms involving CH + N2 = NCN + H as initiation reaction, as NCN has been shown to be the appropriate product channel for the CH + N2 reaction [18,19]: (i) the Williams et al. (2009) submechanism (namely GRImech3.0_modif_Williams) [25]; (ii) the Lamoureux et al. (2014) submechanism (namely GRImech3.0_modif_Lamoureux) [20] and (iii) the Konnov6.0 (2009) submechanism (namely GRImech3.0_modif_Konnov) [23]. These mechanisms

405

L. Pillier et al. / Fuel 150 (2015) 394–407 Table 3 Reactions (prompt NO pathway) implemented in the modified version of GRImech3.0. Williams et al. (2009) submechanism [25]

Lamoureux et al. (2014) submechanism [20]

Konnov6.0 submechanism [23]

CH + N2 = HCN + N replaced by CH + N2 = NCN + H NCN consumption NCN + H = HCN + N NCN + O = CN + NO NCN + OH = HCN + NO NCN + O2 = NO + NCO CH2N2 production and consumption CH2(S) + N2(+M) = CH2N2(+M) CH2N2 + O = H2CN + NO CH2N2 + O = CH2O + N2 CH2N2 + OH = CH2OH + N2 CH2N2 + H = CH3 + N2 CH2N2 + O = HCNN + OH CH2N2 + OH = HCNN + H2O CH2N2 + H = HCNN + H2 Revised HCNN kinetics HCNN + O = CO + H + N2 HCNN + O = HCN + NO HCNN + H = CH2 + N2 Reactions added HCNN + OH = H2CN + NO HCNN + H = HCN + NH

CH + N2 = HCN + N and CH + N2(+M) = HCNN(+M) replaced by CH + N2 = NCN + H NCN consumption NCN + H = HCN + N NCN + O = CN + NO N + OH = HCN + NO NCN + O2 = NO + NCO

CH + N2 = HCN + N and CH + N2(+M) = HCNN(+M) replaced by CH + N2 = NCN + H NCN consumption CN + N2O = NCN + NO (DUP) CN + NCO = NCN + CO C2O + N2 = NCN + CO CH + N2 = HNCN HNCN + M = H + NCN + M HNCN + O = NO + HCN HNCN + O = NH + NCO HNCN + O = CN + HNO HNCN + OH = NCN + H2O HNCN + O2 = HO2 + NCN NCN = N + CN NCN = C + N2 NCN = CNN NCN + H = HCN + N NCN + O = CN + NO NCN + O = CO + N2 NCN + O = N + NCO NCN + N = N2 + CN NCN + C = 2CN NCN + OH = HCN + NO (DUP) NCN + O2 = NO + NCO NCN + CH = HCN + CN NCN + CN = C2N2 + N NCN + CH2 = H2CN + CN

1.40

1.20

1.20 1.00

0.80

0.80 0.60 0.60 0.40

0.40

0.20

XNOexperiment/XNO GRIMech3.0

XNOsubmechanism/ XNOGRIMech3.0

1.00

0.20 -

0.10

0.30

0.50

0.70

Pressure (MPa)

With Lamoureux sub-mechanism With Konnovsub-mechanism With Williams sub-mechanism Fig. 11. Maximum NO mole fraction (left-hand side axis) of each modified GRImech3.0 (with Lamoureux sub-mechanism in red, Konnov sub-mechanism in green and Williams sub-mechanism in purple) normalised by the original GRImech3.0 value as a function of pressure for E.R. = 0.7 in premixed CH4/air counter-flow flames. The black line gives the ratio XNO experiment on XNOGRIMech3.0 (right-hand side axis). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

together with their thermodynamic and transport data are given in the supplementary files in CHEMKIN format. Modifications to GRImech3.0 due to the inclusion of prompt-NO submechanisms via NCN product are presented in Table 3. The Williams et al. submechanism [25] involves two additional species compared to GRImech3.0: NCN and CH2N2. Only NCN is added in the Lamoureux et al. submechanism [20] and five supplementary species are added in the Konnov et al. submechanism [23]: NCN, C2O, HNCN, CNN and C2N2. Laminar flame velocities were computed with PREMIX [29]

with each modified version of GRImech3.0. While no change was observed concerning the prediction of flame speeds, this was not the case for the species profiles (not shown here): the inclusion of the different sub-mechanisms in the GRImech3.0 mechanism entails changes mainly in the burned gases (maximum NO mole fraction). The profile shapes remain unchanged. Fig. 11 compares the maximum of NO mole fraction for each modified GRImech3.0 normalised by the original GRImech3.0 value for each pressure. The ratio of XNO experiment to XNOGRImech3.0 is also given. As

406

L. Pillier et al. / Fuel 150 (2015) 394–407

shown in Fig. 11, the same trend is observed for all pressure conditions: the predictions given by GRImech3.0_modif_Williams and GRImech3.0_modif_Lamoureux are greater than those given by the unmodified GRImech3.0, and GRImech3.0_modif_Konnov gives lower predictions. The inclusion of these sub-mechanisms involves changes not only in the NO mole fraction but also in the other species involved in or linked to the prompt-NO formation pathway. To illustrate this point, the mole fraction predicted by the modified versions of GRImech3.0 normalised by the mole fraction predicted by the initial GRImech3.0 was calculated (not shown here). Differences are observed for some N-containing species: N, NH, NH2, NH3, NO2, HNO, HCN, HCNO, HOCN, HNCO and the most strongly affected species are CN, NCO and H2CN. Moreover, for NCN, which is not included in the initial version of GRImech3.0, the three modified versions of GRImech3.0 give different results (e.g. at atmospheric pressure, the maximum mole fraction of NCN predicted by GRImech3.0_modif_Lamoureux is 1.1  108; the prediction by GRImech3.0_modif_Konnov is 3.5  1010 and the prediction by GRImech3.0_modif_Williams is 6.6  109). Finally, a rate of production analysis of NO was performed on the modified versions of GRImech3.0 (not shown here) to evaluate the consequences of including a new prompt-NO submechanism on the contribution of each NO formation pathway. Results show that the inclusion of the new prompt-NO increases the contribution of the prompt-NO formation pathway while it decreases that of the thermal formation pathway for all sub-mechanisms. In the case of the GRImech3.0_ modif_Williams and GRImech3.0_modif_Lamoureux submechanisms, both NNH and N2O formation pathways are nearly unchanged. For the GRImech3.0_modif_Konnov, the NNH formation pathway remains unchanged whereas the N2O formation pathway is divided by more than a factor of 2. As shown previously, including submechanisms with the appropriate prompt-NO initiation reaction provides promising results on predictions of NO emissions in lean CH4/air flames at high pressure. A deeper study to compare the accuracy of the different submechanisms would require further experimental results on N-containing species other than NO (NCN, HCN, CN, etc.) at atmospheric and high pressure.

5. Conclusions In this study, NO mole fraction profiles were measured by Laser Induced Fluorescence in laminar counter-flow lean (E.R. = 0.7) CH4/ air flames stabilised at pressures up to 0.7 MPa. Analysis of the experimental excitation and fluorescence spectra showed that excitation via the P1(23.5), Q1 + P21(14.5), Q2 + R12(20.5) feature in the A–X(0, 0) band of NO followed by collection of the signal through the A–X(0, 1) band presents the best compromise to maximise the NO LIF signal and minimise interferences in our high pressure flame conditions. However, high pressure LIF measurements are greatly complicated by the variations in pressure and temperature dependent parameters which locally modify the ratio between the fluorescence signal and NO concentration: the spectral overlap function, the quenching rate and the Boltzmann fraction. We then evaluated theoretically the influence of those parameters on the fluorescence signal and showed that, in our conditions, the variations could be neglected, at a given pressure, across the flame. Thus, individual calibration of the fluorescence signal in each flame was performed using the NO seeding method, which yielded the absolute nascent NO concentration with an uncertainty of  ± 10%. The experimental NO profiles were then compared with modelling using the OPPDIF code and three detailed kinetic mechanisms: the GDFkinÒ3.0_NCN mechanism developed by Lamoureux et al. [49] and recently updated [20] and the two mechanisms from

the Gas Research Institute: GRImech 2.11 [15] and GRImech3.0 [16]. NO mole fraction profiles were satisfactorily predicted by the three mechanisms (within the error bars), with a slight overprediction for the GDFkinÒ3.0_NCN mechanism at a pressure P > 0.3 MPa. A kinetic analysis was then performed to understand the behaviour of each mechanism with respect to its prediction of NO formation when pressure increases, by determining the contribution of each NO sub-mechanism (thermal, prompt, NNH and N2O) to the overall NO concentration. For the thermal pathway, we used the subtraction method showing that the thermal-NO contribution can reach more than 30% for the GRImech mechanisms and 20% for the GDFkinÒ 3.0_NCN at high pressure. Rate of production/consumption analyses (integrated over the whole flame domain) were also performed. The analysis of rates of consumption for N2 (precursor of NO) showed that N2 mainly decomposes through the N2O and NNH pathways for both mechanisms. The analysis of rates of production of NO showed that the mechanisms exhibit different behaviours regarding the NO production pathways contribution, with a substantial contribution of the promptNO pathway for all mechanisms. Finally three modified versions of the GRImech3.0 mechanism were proposed, including the appropriate CH + N2 = NCN + H prompt-NO initiation channel (rather than the spin forbidden CH + N2 = HCN + N) and based on the work of Lamoureux et al. [20], Williams et al. [25] and Konnov [23]. While the three modified versions of GRImech3.0 still give correct predictions for NO in comparison with our experimental results, large differences are observed for some N-containing species predictions, notably for CN, HCN, NCN, NCO and H2CN. Acknowledgements This work was supported by the French ANR program BLAN-080130 (NO-mecha), by the French Ministry of Research and by the Région Centre. Appendix A. Supplementary material Supplementary material associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.fuel. 2015.01.099. References [1] Jamette P, Ricordeau V, Deschamps B, Desgroux P. Laser-induced fluorescence detection of NO in the combustion chamber of an optical GDI engine with A– X(0, 1) excitation. SAE International paper. 2001-01-1926; 2001. [2] Verbiezen K, Klein-Douwel RJH, van Vliet AP, Donkerbroek AJ, Meerts WL, Dam NJ, et al. Attenuation corrections for in-cylinder NO LIF measurements in a heavy-duty diesel engine. Appl Phys B 2006;83:155–66. [3] Verbiezen K, van Vliet AP, Dam NJ, ter Meulen JJ. Absorption of NO laserinduced fluorescence by hot O2 and CO2. Combust Flame 2006;144:638–41. [4] Verbiezen K, Donkerbroek AJ, Klein-Douwel RJH, van Vliet AP, Frijters PJM, Seykens XLJ, et al. Diesel combustion: in-cylinder NO concentrations in relation to injection timing. Combust Flame 2007;151:333–46. [5] Verbiezen K, Klein-Douwel RJH, van Vliet AP, Donkerbroek AJ, Meerts WL, Dam NJ, et al. Quantitative laser-induced fluorescence measurements of nitric oxide in a heavy-duty Diesel engine. Proc Combust Inst 2007;31:765–73. [6] Schulz C, Yip B, Sick V, Wolfrum J. A laser-induced fluorescence scheme for imaging nitric oxide in engines. Chem Phys Lett 1995;242:259–64. [7] Thomsen DD, Kuligowski FF, Laurendeau NM. Background corrections for laser-induced-fluorescence measurements of nitric oxide in lean, highpressure, premixed methane flames. Appl Opt 1997;36:3244–52. [8] Thomsen DD, Kuligowski FF, Laurendeau NM. Modelling of NO formation in premixed, high-pressure methane flames. Combust Flame 1999;119:307–18. [9] Bessler WG, Schulz C, Lee T, Jeffries JB, Hanson RK. Strategies for laser-inducedfluorescence of nitric oxide in high-pressure flames. III. Comparison of A–X excitation schemes. Appl Opt 2003;42:4922–36. [10] Bessler WG, Schulz C, Lee T, Jeffries JB, Hanson RK. Strategies for laser-inducedfluorescence of nitric oxide in high-pressure flames. I – A–X(0, 0) excitation. Appl Opt 2002;41:3547–57.

L. Pillier et al. / Fuel 150 (2015) 394–407 [11] Bessler WG, Schulz C, Lee T, Jeffries JB, Hanson RK. Strategies for laserinduced-fluorescence of nitric oxide in high-pressure flames. II – A–X(0, 1) excitation. Appl Opt 2003;42:2031–42. [12] Schulz C, Sick V, Heinze J, Stricker W. Laser-induced-fluorescence detection of nitric oxide in high-pressure flames with A–X(0, 2) excitation. Appl Opt 1997;36:3227–32. [13] Naïk SV, Laurendeau NM. LIF measurements and chemical kinetic analysis of nitric oxide formation in high pressure counterflow partially premixed and nonpremixed flames. Combust Sci Technol 2004;176:1809–53. [14] Lee T, Jeffries JB, Hanson RK. Experimental evaluation of strategies for quantitative laser-induced-fluorescence imaging of nitric oxide in highpressure flames (1–60 bar). Proc Combust Inst 2007;31:757–64. [15] Bowman CT, et al., GRImech2.11 Mechanism, 1995; [16] Smith GP, et.al., GRImech3.0 Mechanism, 1999; [17] Fenimore CP. Formation of nitric oxide in premixed hydrocarbon flames. Proc Combust Inst 1971;13:373–9. [18] Moskaleva LV, Lin MC. The Spin-conserved reaction CH + N2 ? H + NCN: a major pathway to prompt no studied by quantum/statistical theory calculations and kinetic modelling of rate constant. Proc Combust Inst 2000;28:2393–401. [19] Vasudevan V, Hanson RK, Bowman CT, Golden DM, Davidson DF. Shock tube study of the reaction of CH with N2: overall rate and product branching. J Phys Chem A 2007;111:11818–30. [20] Lamoureux N, El Merhubi H, Gasnot L, Schoemaecker C, Desgroux P. Measurements and modelling of HCN and CN species profiles in laminar CH4/O2/N2 low pressure flames using LIF/CRDS techniques. Proc Combust Inst 2015;35:745–52. [21] Molet J, Matynia A, Pillier L, Idir M, de Persis S. Kinetic study of NOx formation in high-pressure flames. In: Proc. of the 5th European Combustion Meeting 2011 (ECM2011), Cardiff, Wales; 2011. [22] Thomsen DD. Measurements and modelling of nitric oxide formation in counterflow, premixed, CH4/O2/N2 flames. PhD thesis, Purdue University; 1999. [23] Konnov AA. Implementation of the NCN pathway of prompt-NO formation in the detailed reaction mechanism. Combust Flame 2009;156:2093–105. [24] Lutz AE, Kee RJ, Grcar JF, Rupley FM. OPPDIF: A fortran program for computing opposed-flow diffusion flames, Report no. SAND96-8243, Sandia National Laboratories; 1997. [25] Williams BA, Sutton JA, Fleming JW. The role of methylene in prompt NO formation. Proc Combust Inst 2009;32:343–50. [26] Matynia A, Idir M, Molet J, Roche C, de Persis S, Pillier L. Absolute OH concentration profiles measurements in high pressure counterflow flames by coupling LIF, PLIF and absorption techniques. Appl Phys B 2012;108:393–405. [27] Rolon JC. Etude théorique et expérimentale de la flamme de diffusion à contrecourant. PhD thesis Ecole Centrale Paris, France; 1988. [28] Reynolds WC. The element potential method for chemical equilibrium analysis: implementation in the interactive program STANJAN. Stanford University Report ME 270 HO no 7; 1986. [29] Kee RJ, Grcar JF, Smooke MD, Miller JA. PREMIX: A fortran program for modelling steady laminar one-dimensional premixed flames, Report no. SAND85-8240, Sandia National Laboratories; 1985. [30] Matynia A, Molet J, Roche C, Idir M, de Persis S, Pillier L. Measurement of OH concentration profiles by laser diagnostics and modelling in high pressure counter-flow premixed methane–air and biogas–air flames. Combust Flame 2012;159:3300–11.

407

[31] Klassen MS, Thomsen DD, Reisel JR, Laurendeau NM. Laser-induced fluorescence measurements of NO formation in high-pressure premixed methane flames. Combust Sci Technol 1995;110–111:229–47. [32] Battles BE, Hanson RK. Laser induced fluorescence measurements of NO and OH mole fraction in fuel-lean high pressure (1–10 atm) methane flames: fluorescence modelling and experimental validation. J Quant Spectrosc Radiat Transfer 1995;54:521–37. [33] Bessler WG, Schulz C, Sick V, Daily JW. A versatile modelling tool for nitric oxide LIF spectra. In: Proceedings of the Third Joint Meeting of the U.S. Sections of The Combustion Institute, 2003c; (Chicago, March 16–19, paper PI05), . [34] Naïk SV, Laurendeau NM. Spectroscopic, calibration and RET issues for linear laser-induced fluorescence measurements of NO in high-pressure diffusion flames. Appl Phys B 2004;79:641–51. [35] Ravikrishna RV, Cooper CS, Laurendeau NM. Comparison of saturated and linear laser-induced fluorescence measurements of nitric oxide in counterflow diffusion flames. Combust Flame 1999;117:810–20. [36] Driscoll JJ, Sick V, Schrader PE, Farrow RL. Measurements of NO distributions and fluorescence lifetimes in a non-premixed counterflow CH4/air flame using picosecond time-resolved laser-induced fluorescence. Proc Combust Inst 2002;29:2719–26. [37] Daily JW, Roth EW. Effect of laser intensity and of lower-state rotational energy transfer upon temperature measurements made with laser-induced fluorescence. Appl Phys B 1999;68:131–40. [38] Paul PH, Gray JA, Durant JL, Thoman JW. A model for temperature-dependent collisional quenching of NO A2r+. ApplP Phys B 1993;57:249–59. [39] Paul PH, et.al, Corrections for the NO A2 + (v0 = 0) electronic quenching crosssection. Sandia Report; 1995. [40] Settersten TB, Patterson B, Gray JA, Temperature-dependent cross sections for quenching of NO LIF by NO, H2O, and CO2. Proceedings of the Third Joint Meeting of the U.S. Sections of The Combustion Institute; 2003: 1–5. [41] Luque J, Crosley DR. LIFBASE: Database and spectral simulation program (version 1.5). SRI International Report MP 99-009; 1999. [42] Amiot C, Bacis R, Guelachvili G. Infrared study of the X2P v = 0,1,2 levels of 14 16 N O. Preliminary results on the v = 0,1 levels of 14N17O, 14N18O and 15N16O. Can J Phys 1978;56:251–65. [43] Kee RJ, Rupley FM, Miller JA. Chemkin-II: A fortran chemical kinetics package for the analysis of gas-phase chemical kinetics, Report No. SAND89-8009B, Sandia National Laboratories; 1989. [44] Cao RR, Pope SB. The influence of chemical mechanisms on PDF calculations of nonpremixed piloted jet flames. Combust Flame 2005;143:450–70. [45] Vreman AW, Albrecht BA, van Oijen JA, de Goey LPH, Bastiaans RJM. Premixed and nonpremixed generated manifolds in large-eddy simulation of Sandia flame D and F. Combust Flame 2008;153:394–416. [46] Zeldovich YB. The oxidation of nitrogen in combustion and explosions. Acta Physicochimica 1946;21:577–628. [47] Bozzelli JW, Dean AM. O + NNH: a possible new route for NOx formation in flames. Int J Chem Kinet 1995;27:1097–109. [48] Malte PC, Pratt DT. The role of energy-releasing kinetics in NOx formation: fuel-lean, jet-stirred CO–air combustion. Combust Sci Technol 1974;9:221–31. [49] Lamoureux N, Desgroux P, El bakali A, Pauwels JF. Experimental and numerical study of the role of NCN in prompt-NO formation in low pressure CH4–O2–N2 and C2H2–O2–N2 flames. Combust Flame 2010;157:1929–41. [50] Goos E, Sickfeld C, Maus F, Seidel L, Ruscic B, Burcat A, et al. Prompt NO formation in flames: The influence of NCN thermochemistry. Proc Combust Inst 2013;34:657–66.