PLIF measurements of non-thermal NO concentrations in alcohol and alkane premixed flames

Combustion and Flame 194 (2018) 363–375

Contents lists available at ScienceDirect

Combustion and Flame journal homepage: www.elsevier.com/locate/combustflame

PLIF measurements of non-thermal NO concentrations in alcohol and alkane premixed flames Myles D. Bohon a,b,∗, Thibault F. Guiberti b, William L. Roberts b a b

Chair of Fluid Dynamics, Technische Universität Berlin, 10623 Berlin, Germany Clean Combustion Research Center, King Abdullah University of Science and Technology, Thuwal 23955-6900, Saudi Arabia

a r t i c l e

i n f o

Article history: Received 20 February 2018 Revised 19 March 2018 Accepted 21 May 2018

Keywords: NO x Alcohols PLIF Non-thermal NOx

a b s t r a c t There exists interest in using alcohols as renewable, lower emission fuels. It has been observed that alcohol flames generally produce lower concentrations of NO emissions and the cause of these reductions is attributable to a number of mechanisms. This work therefore investigates the relative contributions to total NO formation in alcohol fueled flames, relative to comparably sized alkane flames. Measurements of quantitative NO PLIF were conducted in two common premixed configurations: a conical, Bunsen-type flames and a lower peak temperature burner-stabilized McKenna flat flame. Additionally, these flames were modeled using a detailed NO x chemical mechanism and investigated to understand the primary contribution pathways to non-thermal NO formation. From this analysis, it was observed that alcohol fueled flames produced as much as 50% less non-thermal NO than alkanes. However, under lean conditions the non-thermal contributions increased to about 80–90% of those observed in alkanes due to a greater contribution to non-thermal NO in both alcohol and alkane flames from non-hydrocarbon radical related mechanisms. © 2018 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

1. Introduction With increasingly stringent regulations on combustion emissions, increasing emphasis is placed on utilizing alternative sources of energy while also reducing overall emissions. One strategy which combines both features is the increased usage of alcohols as fuels. Available from carbon-neutral sources and renewable, alcohols have also been observed to produce lower emissions of NO x [1,2]. It is therefore important to gain a better understanding of the mechanisms through which NO x is produced in alcohol flames, how that production compares with more traditional alkane flames, and whether these reductions can be achieved in mixtures of the alcohols with other fuels. NO x formation has been studied for a relatively long time. In this time, a number of mechanisms have been identified as sources of NO x in flames [3]. In general, these sources of NO x can be attributed to four primary categories of mechanisms: Thermal, Prompt, Fuel, and Minor. A detailed description of each of these categories is beyond the scope of this work, especially as several excellent sources examining these mechanisms in detail already exist. Therefore this work will only present a high-level overview ∗ Corresponding author at: Chair of Fluid Dynamics, Technische Universität Berlin, 10623 Berlin, Germany. E-mail address: [email protected] (M.D. Bohon).

of these mechanisms and how they relate to the current work, while interested readers are directed to the selected list of works referenced below. The most broadly applicable mechanism to NO x formation is through the Thermal mechanism. This mechanism is controlled by the reaction of N2 with atomic oxygen, with the resulting nitrogen atoms further contributing to NO formation by reaction with other oxidizing radicals. Due to the high activation energy of the nitrogen triple bond, this reaction requires high temperatures and is typically active in the post-flame. Additionally, this mechanism is often considered relatively slow and decoupled from the fuel oxidation [3]. Consequently, this mechanism is largely independent of the fuel chemistry (apart from the flame temperature). The Prompt mechanism is so called due to the rapid rate at which NO was formed in the flame front which was faster than could be accounted for by the Thermal mechanism alone [4,5]. This mechanism is initiated through the fixation of nitrogen by hydrocarbon radicals, HC. Consequently, the Prompt mechanism is strongly coupled to the fuel oxidation process. It was ultimately shown that CH and CH2 are the primary initiating species for the Prompt mechanism [6] with CH as the most significant initiating radical through reaction below [7].

CH + N2  NCN + H

https://doi.org/10.1016/j.combustflame.2018.05.024 0010-2180/© 2018 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

(1)

364

M.D. Bohon et al. / Combustion and Flame 194 (2018) 363–375

The resulting NCN then forms NO or N through reactions [8]

NCN + O  CN + NO

(2)

NCN + OH  HCN + NO

(3)

NCN + H  HCN + N

(4)

NCN + O2  NCO + NO

(5)

In summary, the controlling issues of the Prompt mechanism are: 1) the CH concentration and how it is established, 2) the rate of nitrogen fixation, and 3) the rates of interconversion among fixed nitrogen fragments [3]. Accurately predicting these features remains a significant hurdle to predicting NO x formation, however work continues to improve the ability to predict NO formation through this pathway [9]. The Fuel mechanism is dominated by the presence of nitrogen within the fuel. Such fuel-nitrogen is commonly present in coal combustion [3], and can be a significant source of NO x in the combustion of such fuels. In this work however, there is a negligible amount of nitrogen in the fuel and this mechanism will therefore not be considered further. The final category is (perhaps poorly) called Minor. This category is grouped as a collection of mechanisms which can be significant contributors to NO x under certain conditions and negligible at others. The first mechanism considered here is the NO-HCN-Reburn mechanism by which NO x is recycled by reaction with HC back to HCN, the fate of which is either to reform NO or to be converted back to N2 [10–13]. The second mechanism involves the reaction of forward diffusing NO with the hydroperoxyl radical HO2 to form NO2 . Through reaction with H and O, NO2 is also converted back to NO [14–16]. The balance of this reaction is particularly important in alcohol flames due to an abundance of HO2 . The final mechanism of consideration here is ultimately a combination of reactions involving N2 O [17] and NNH [18–20]. These two species can be significant contributors to NO x , especially in the absence of significant HC radicals. A more detailed survey of this H/N/O reaction set is available in Klippenstein et al. [21]. As both N2 O and NNH sub-mechanisms proceed through a strongly coupled initiating process, the approach here will be to combine them into a single mechanism called NNX. One of the first works examining NO x formation specifically in alcohol fuels is from William’s group. Li and Williams [1] studied laminar, premixed methanol flames with a relatively small set of experimental measurements of NO concentrations. They also studied the influence of the addition of H2 O, N2 , CO2 , and Ar [2]. In both works, lower formation of NO was observed, with the majority of the NO formation attributed to the Thermal Mechanism. Saxena and Williams [22] later extended this work to include ethanol, however an experimental measurements remained limited. Chung et al. [23] conducted a comparison of iso- and n-butanol isomers with butane in stagnating premixed flames. A reduction in the levels of NO was observed for most butanol isomers, except for iso-butanol. These reductions were attributed to reductions in both the Prompt and Thermal mechanisms. Watson et al. [24,25] also investigated C1–C4 alkanes and alcohols in similar stagnating premixed flames. They measured simultaneous NO and CH PLIF and observed a scaling of the formation of NO with respect to the concentration of CH in the flame front, especially when the CH is appropriately scaled by the residence time in the flame front. The objective of this work is two-fold. The first objective is to understand and compare the formation of NO through laminar, premixed alcohol and alkane fueled flames in two common configurations. The complex interaction between the different formation mechanisms with the oxidation of the fuel is not yet clearly

understood for alcohols. Toward this aim, the different NO formation regimes within the flame will be observed and the relative contribution between the thermal and non-thermal mechanisms to total NO formation will be quantified. The use of two different burner configurations – one with nearly adiabatic conditions and another with significantly reduced peak temperatures due to heat loss – will further highlight the differences between the NO x formation regimes. The second objective is to provide an empirical dataset against which improvements in detailed NO x chemical mechanisms can be modeled. In addition to providing such a dataset, an analysis of the estimated contribution to non-thermal NO formation from several mechanisms is conducted in order to understand the mechanisms by which reductions of NO x formation in alcohol flames are achieved. 2. Experimental methods 2.1. Experimental set-up This work examines alcohol and alkane flames in two different premixed configurations: a Bunsen-type conical flame and a McKenna burner stabilized flat flame. The conical flame burner is composed of a contoured nozzle with a 10 mm diameter exit and a surrounding nitrogen shroud. The flat flame burner is composed of a 6 cm diameter McKenna burner with a stainless steel porous plug and a surrounding nitrogen shroud. The flames in the flat flame burner have previously been investigated by the authors [26]. This previous work measured the concentration of NO in the far field of the flat flame using probed gas sampling and modeled the flames using a temperature profile measured by 2λ OH PLIF combined with thermocouple measurements. The two lines for the 2λ thermometry used were the P1 (7) and Q2 (11) transitions of the A2  + ← X 2 (1, 0 ) OH band (285.088 and 285.157 nm, respectively), with a line separation energy of ε 12 of 2046 K. This pair provided a good temperature sensitivity over the temperature range of 120 0–20 0 0 K [27,28]. Additional information regarding the temperature measurements is available in Appendix A. The current work aims to expand on this previous work with nonintrusive NO PLIF and by comparison with a flame with less heat loss to the burner surface. It should be noted that these 2λ OH PLIF thermometry measurements were conducted separately from the NO PLIF measurements discussed in Section 2.2, and the two techniques were not applied simultaneously. Both burners are supplied with premixed fuel and air using an in-house pre-vaporizing system, shown in Fig. 1. The vaporization system was composed of a series of three heated and insulated steel mixing vessels. The primary vaporization occurs in the first vessel. Here, an air atomizing nozzle uses preheated air to spray the liquid alcohol into the vessel. The temperature and flow rate of the atomizing air varied, depending on the fuel type and flow rate. Generally, the total atomizing air flow represented around 15–20% of the total combustion air. This rich mixture is then ducted through a distributed series of tubes into another settling tank, into which the remainder of the combustion air is injected. Within this tank and the final following tank, the mixture is brought to its final equivalence ratio and temperature. The gaseous alkanes were injected into this final tank when used. From here, the mixture is directed to the burner through a temperature controlled line. Both burners were insulated and temperature controlled to the target temperature. The recirculating fluid in the McKenna burner was replaced with oil and recirculated through a heat exchanger to also maintain the target temperature. Both burners were able to maintain very steady and consistent operation for many hours during testing. Gas flow rates were controlled by Brooks thermal mass flow controllers, and the liquid fuel flow rate was controlled by a Brooks coriolis mass flow controller. All controllers had been

M.D. Bohon et al. / Combustion and Flame 194 (2018) 363–375

365

Fig. 2. Experimental setup for NO LIF with the conical flame.

2.2. NO LIF technique

Fig. 1. Pre-vaporizing system for the vaporization, temperature control, and mixing of fuel and air.

Table 1 Experimental conditions for conical and flat flame experiments. Property

Conical

Flat

φ

0.8–1.2 373 2.5

0.25

150.8 160.9 158.9 190.1 178.8 147.5 136.7

15.1 16.1 15.9 19.0 17.9 14.8 13.7

Temperature (K) Ratio of u to SL Exit velocity u at φ = 1.0 (cm/s) Methane Ethane Propane Methanol Ethanol n-Propanol i-Propanol

recently calibrated and controlled flow rates to within specification. This system allows for delivery of perfectly premixed fuel and air mixtures at final temperatures up to 373 K with a wide range of equivalence ratios. Table 1 summarizes the range of experimental conditions for both burner configurations. The equivalence ratio was varied between 0.8 and 1.2 while the exit gas temperature was held constant at 373 K. This work examines the C1 –C3 alkanes and alcohols, including both propanol isomers. Exit velocities, u, were selected to be a constant multiple of the laminar burning velocity SL at each equivalence ratio. The laminar burning velocities for each fuel were taken from experimental measurements in the literature: alkanes by Vagelopoulos and Egolfopoulos [29], methanol and ethanol by Gülder [30], and the propanol isomers by Veloo and Egolfopoulos [31]. Corrections for deviations in the initial temperature from the reported temperature were conducted using the methodology of Metghalchi and Keck [32,33]. By maintaining a constant velocity relative to the laminar burning velocity, the impact of variations in SL are mitigated. In the flat flames, exit velocities have a spread of only 5 cm/s at stoichiometric conditions, and therefore these flames should be much more dependent on the temperature than the exit velocity. In the conical flames, a fixed ratio to SL resulted in consistent flame shapes throughout the tests, improving consistency and facilitating the PLIF measurements.

NO concentrations through the flames were measured by planar laser induced fluorescence (PLIF) in the linear regime. A cartoon of the optical set up is shown in Fig. 2. A 10-Hz Nd:YAG (Continuum Powerlite DLS9010) pumped, frequency doubled Rhodamine dye-laser (Continuum ND60 0 0+UVT) with output near 287 nm, which is then mixed with the 1064 nm fundamental output of the Nd:YAG yielding a mixed output wavelength near 226 nm, serves as a laser source. The available output energy is approximately 1 mJ per pulse. Excitation of the A-X(0,0) band with selection of the P1 (23.5), Q1 +P21 (14.5), Q2 +R12 (20.5) overlapping rotational lines at 225.963 nm (in air) has been shown to maximize signal strength while minimizing interference from molecular oxygen LIF [34–36]. A spectral scan of the region around 225.96 nm, as shown in Fig. 3a, allowed for the identification of the targeted transition and the detuned wavelengths. The simulated spectra was calculated using the LIFBASE software [37]. Broadband emission of the A-X(0,1) band was collected through a custom manufactured Asahi Spectra narrow bandpass filter centered at 236 nm with 7.5 nm full width at half maximum, as shown in Fig. 3b, and was detected by a gated Princeton Instruments PI-MAX3 blue-enhanced ICCD. The camera was gated to a 100 ns exposure time, with a 1 on, 4 off duty cycle. Due to the low concentrations of NO, 200 shots were accumulated on the CCD chip at each condition to improve the signal-to-noise ratio. The laser was also detuned (shifted 0.011 nm to the red) to measure the background signal and oxygen LIF, which was then removed from the measurement. Corrections were made to account for fluctuations in the total laser energy as well as the spatial distribution of laser energy. Due to on-chip accumulations, it was not possible to correct for single shot fluctuations. However, the energy and spatial distribution of the shots composing a given set of exposures were averaged and the mean correction was applied. Operation in the linear regime was verified for both the tuned and detuned exposures as well as the profile correction fluorescence. Lastly, corrections for non-linearity in the pixel response in the spectral range of 236 nm was applied. Correlation of the NO LIF signal to a local NO concentration requires consideration of a number of factors. In the linear fluorescence regime, the fluorescent signal can be shown to be proportional to the number density of the probed species. However, consideration must be given to temperature and collisional quenching effects which are composition, temperature, and pressure dependent. Eq. (6) illustrates the correlation between measured LIF signal intensity and NO number density:

SNO = Copt τλ B12 IL fb (T ) (νL , T , p, Xk )(T , p, Xk )noNO (T , p, Xi ) (6)

366

M.D. Bohon et al. / Combustion and Flame 194 (2018) 363–375

Fig. 3. Wavelength scan of (a) NO rotational lines and simulation of transitions identifying the P1 (23.5) and off-line transitions and (b) NO emission spectra and transmittance of 236 nm bandpass filter.

where Copt is the combined constant for the collection optics, τ λ is the transmissivity of the collection optics, B12 is the Einstein coefficient of absorption, IL (EL , ν L ) is the spectrally resolved laser irradiance, fb (T) is the Boltzmann population distribution, (ν L , T, p, Xk ) is the normalized overlap integral of the incident and absorbing transition frequencies, (T, p, Xk ) is the fluorescent yield in the weak excitation limit, and noi (T , p, Xi ) is the number density of the target species. EL and ν L are the laser energy and the spectral width of the laser, respectively. Xk is the molar fraction of each species in the gas mixture, at the temperature T and pressure p. Copt and τ λ can be grouped together and determined empirically for the specific apparatus. The absorption coefficient B12 is a constant for a given species and transition. The laser irradiance IL is measured experimentally as discussed above. The Boltzmann fraction in the ground state fb is temperature dependent, but can be estimated using Boltzmann statistics. The overlap integral is introduced through the spectral width of the exciting laser [38] and is weakly temperature, species, and pressure dependent. The fluorescent yield  is temperature and composition dependent, with estimations of the quenching rate made following the work of Settersten et al. [39]. The number density noNO is proportional to the concentration described by the ideal gas law and is temperature and pressure dependent. The relations used to determine the above parameters are described in greater detail in the appendix. Eq. (6) can be simplified through the combination of several terms. Copt , τ λ , B12 , IL , and can be combined into Ccal which is independent of the local temperature or composition and is determined experimentally as described later in relation to Fig. 4 a. The Boltzmann fraction, the fluorescent yield (accounting for the quenching rate), and conversion from number density to mole fraction can be grouped into a temperature and composition

Fig. 4. (a) Calibration fits for each of the fuels, and (b) example of normalized temperature and composition dependent corrections to the NO LIF signal.

dependent term, β , as shown in equation below.

β (T , p, Xk ) = fb (T )

A21 1 A21 + Q21 (T , p, Xk ) T

(7)

The fluorescent yield in a simple two-level model is a function of the Einstein coefficient for the spontaneous emission of the excited species, A21 , and the rate of collision quenching, Q21 (T, p, Xk ) [34]. Finally, the LIF signal intensity SNO can be related to the mole fraction of NO through equation below

SNO = Ccal · β (T , p, Xk ) · XNO

(8)

From the above, however, it is apparent that knowledge of the local temperature and quenching species concentrations (N2 , H2 O, CO2 , CO, O2 , and NO) is necessary. Because experimental measurements of all these parameters would be difficult and expensive, it was determined that a decent approximation of these parameters could be achieved by modeling the flame. Fristrom [40] showed fairly early on that temperature and density profiles of the three dimensional, axisymmetric conical flame are well reproduced in the near field by a one-dimensional flame model. Later, Nguyen et al. [41] showed with Raman-LIF measurements that major species (including CO, OH, and NO) and temperature were able to be satisfactorily predicted by a one-dimensional freely propagating premixed flame model with finite rate chemistry. A number of other studies have indicated good approximations of most of the conical flame surface by one-dimensional flames [42–46]. Therefore, it was decided to proceed with a one-dimensional model and limit the analysis to regions very near the flame front with minimal dilation and buoyancy effects. The flat flames were modeled using a defined temperature profile which was measured through the combination of probed thermocouple and 2λOH

M.D. Bohon et al. / Combustion and Flame 194 (2018) 363–375

LIF thermometry techniques and the simulation of the flames as burner stabilized flat flames. Additional information of the measurement of the temperature profiles is available in previous work [26], and temperature profiles are included in Appendix A. Determination of the calibration constant Ccal follows the procedure described by several other works [11,24,36,47–49]. Here, known concentrations of NO were seeded into the unburned gas mixture of lean flames (φ = 0.9) for each fuel. The resultant change in NO LIF signal with doped NO concentrations is linear, the slope of which is equal to Ccal β . The primary assumption in this procedure is that only a minor consumption of NO through the flame front occurs, which has been shown to be a decent approximation under lean conditions [50]. The calibrations for all seven fuels are shown in Fig. 4a. The slope of each fit is within less than ± 10% of the average slope, with the variations introduced through differences in the amounts of NO consumed through the flame and differences in the quenching environments. Lastly, the Y-intercepts for each case are different due to differences in the initial unseeded NO concentrations. Finally, the corrections for all the temperature and compositionally dependent quenching rates can be applied to all the flames in this study. Figure 4b shows an example of the temperature dependence of the LIF correction terms, normalized by the correction at 300 K. A peak in the Boltzmann fraction can been seen around 800 K, with a nearly linear decay with increasing temperature. The collisional quenching is calculated using the equilibrium combustion products of a stoichiometric methane flame. The total temperature and composition dependent correction factor, β , shows a nearly linear response changing by a factor of approximately two over the range of temperatures typical of premixed flames (10 0 0–20 0 0 K). In this temperature region, an uncertainty in temperature of ± 50 K corresponds to an error in the temperature dependent quenching factor of less than 5%. 2.3. Mechanism and modeling The flames in this study were approximated as either onedimensional freely propagating premixed flames (for the conical flames) or one-dimensional burner stabilized flat flames (for the flat flames) in CHEMKIN-Pro [51]. The mechanism for this study was based on the AramcoMech1.3 [52] including the C1 and C2 alcohol high temperature chemistry. The C3 chemistry was sourced from the propanol sub-mechanism of Sarathy et al. [53]. The NOX chemistry set was provided by El Bakali et al. [54] with the rate for the prompt initiating reaction CH + N2  NCN + H updated to better reflect more recent predictions [55] below 20 0 0 K. Lastly, rates for reactions participating in the NNH formation sub-mechanism for NO were updated [21]. Additional details regarding the mechanism and modeling are available in previous work [26] and the mechanism files are available in the Supplementary material. 3. Results NO PLIF measurements were recorded over the range of fuels and equivalence ratios listed in Table 1 for both burner configurations. After applying the corrections for background, dark current, and O2 LIF interference, along with correcting for fluctuations in total energy and energy distribution, non-linearity in ICCD camera response, and spatial calibration, images such as those in Fig. 5 were produced for the conical flames and flat flames (not shown). Immediately, several trends can be identified. Specification of the exit velocity as a constant factor of the laminar burning velocity provides a nearly constant flame aspect ratio within the uncertainty of SL , allowing the flame to stay within the image frame. Correcting for the fluctuations and non-uniformity in the

367

Fig. 5. NO fluorescence images for propane (left) and n-propanol (right) under lean, stoichiometric, and rich conditions in the conical flame configuration. (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)

energy distribution does a good job accounting for the spatial variations in the excitation beam. Absorption of the laser through the flame can be observed in the slight attenuation of the fluorescence signal when comparing left and right hand sides of the images. NO fluorescence intensity increases from lean to rich conditions. The flame front is a relatively straight region once sufficiently far from the base or from the tip of the flame. Rapid formation of NO through the region of interest (shown within the blue rectangle in Fig. 5) in the first 2–3 mm can be seen. Additionally, in the region of interest, NO fluorescence in the lean flames shows a relatively flatter increase of signal intensity normal to the flame front, whereas the stoichiometric flames show a steeper rise, while the rich flames show the highest gradient in NO PLIF signal intensity. The region of interest (ROI) in each flame was identified and the LIF signal was averaged along the dimension parallel to the flame front. In the conical flames, the region of interest extends 2 mm into the unburned gases and 5 mm into the burned gases, however the LIF is only analyzed over the first 3 mm. The zero position was determined as the location at which the averaged LIF intensity increases by 5% of the span between the lowest and highest measured intensities. This approach helps to provide a systematic technique to identifying the beginning of the formation of NO in an otherwise noisy PLIF image. In the flat flames, the region of interest extends 5 mm above the burner surface. The quantification procedure described in Section 2.1 was then applied and the NO concentration profiles were computed for each flame. The profiles for several of these flames with C3 fuels are shown in Figs. 6 and 7 for the conical and flat flames, respectively. While only the C3 profiles are shown in these two figures, they are representative of trends observed in the C1 and C2 flames as well. Examining the conical flames in Fig. 6 first, several trends can be observed. First, a steady increase in NO concentration at 3 mm can be observed with increasing equivalence ratio; however the increase is diminished as the equivalence ratio becomes richer. Secondly, all flames exhibit a region of very rapid increase in

368

M.D. Bohon et al. / Combustion and Flame 194 (2018) 363–375

Fig. 6. NO concentration profiles for several C3 conical flames.

Fig. 7. NO concentration profiles for several C3 flat flames.

M.D. Bohon et al. / Combustion and Flame 194 (2018) 363–375

NO concentration within the first 0.5 mm. The extent of this increase is a function of the fuel and equivalence ratio. Thirdly, a region of approximately linear increase in NO concentration can be observed; however the rate of increase in this region peaks around stoichiometric equivalence ratios and flattens out as the mixture becomes leaner or richer. The different slopes (formation rates) of NO observed within a flame are indicative of the presence of different NO formation regimes throughout the flame. Comparing the experiments to the model predictions in the lean cases of Fig. 6a–c, it can be seen that the model over-predicts at the end of the ROI, however the overall agreement is quite good, especially considering the disparate sources for the mechanism. The model does however have difficulty capturing the transition between the two NO formation regimes and the transition is not as well defined as was observed in the experiments. Under stoichiometric conditions shown in panels d–f, the NO formation is not as well predicted. It appears that the rate of NO formation in the post-flame is greatly over-predicted by the model. Watson et al. [24] observed a similar over-prediction of the rate of NO formation in the post-flame, which they attributed to discrepancies in the rate of the thermal initiation reaction measured by isolated rate measurements as opposed to measurements in flames. These measurements would appear to corroborate their observations. In these measurements the worst agreement is observed in the rich flames in panels g–i, where NO concentrations are under-predicted by approximately 50%. The rate of formation in the post flame appears to be well captured, however, this is most likely due to starvation of the O atom necessary for thermal initiation rather than an accurate prediction of the reaction rate. Additionally, the model predicts a much faster rate of formation of NO in the flame front than was measured experimentally. The NO concentrations in flat flames, shown in Fig. 7, do not exhibit the steady formation of NO in the post-flame as observed in the conical flames due to the lower peak temperatures and suppression of the Thermal mechanism. They do however demonstrate the rapid formation of NO in the first millimeter above the burner surface. The concentration of NO measured in the post-flame (around 3 mm above the burner surface) is higher in all the flames than the values measured in the previous work [26]. However, as those measurements were conducted with extractive gas sampling through a quartz microprobe, it was expected that they would be prone to measurement error and were primarily focused on identifying trends between the fuels. Comparing the experiments with the model shows similarity with the conical flames. However, the model now consistently under-predicts the formation of NO, especially under rich conditions. The model correctly predicts the lack of a region of steady NO formation in the post-flame. Additionally, the model appears to capture the primary NO formation region being limited to the first millimeter. It appears that these flames have sufficiently low peak temperatures to limit the formation of NO to primarily contributions from non-thermal mechanisms. The underprediction of non-thermal NO formation can be attributed to two effects: the importance of the accurate prediction of intermediate NO contributing species (such as CH, which is the primary prompt initiating species [3]) which serve essentially as boundary conditions for the NOX mechanism, and the importance of accurate rate measurements and branching ratios for the reactions within the NOX mechanism. Versailles et al. [56] showed that CH radical concentrations are often under-predicted and the thickness of the CH region in the flame is not well captured by their models. Additionally, there remains uncertainty in the reaction rate for the prompt initiating reaction CH+N2 [55]. However, several groups are working to improve the estimation of these rates, which should ultimately improve the predictive capability of these mechanisms with regard to quantitatively modeling Prompt NO [9].

369

Fig. 8. Method for determination of demarcation between thermal and nonthermal contributions.

From these experimental measurements, it is possible to observe some of the relative contributions to the total NO formation from the Thermal and non-thermal mechanisms. Figure 8 illustrates the technique applied to demarcate the difference. A chord is drawn between the base of the NO profile and the concentration of NO measured at a distance downstream. At each position along the NO profile, the perpendicular distance from that point to the chord is measured and is plotted in Fig. 8a. The point at which this distance reaches a maximum represents a change in the regime of NO formation from the relatively rapid non-thermal mechanisms to the slower Thermal mechanism. From Fig. 8a, one can see that the non-thermal mechanism contributes 51.1 ppm to the formation of NO, while the Thermal mechanism contributes another 28.4 ppm (up to the 3 mm limit in the ROI). It should be noted that this technique does not distinguish the contribution of the Thermal mechanism in the early part of the flame from the non-thermal mechanisms. However, the actual contribution by the Thermal mechanism during this relatively short period will be seen in the model to be small. Applying this criteria to all the flames allows for the comparison of the non-thermal NO contribution shown in Fig. 9. Here, one can see the strong dependency of non-thermal NO formation on equivalence ratio, with a rapid increase as the flame becomes richer. The difference between the alkane and alcohol flames can also be clearly seen, whereby the alcohol flames are consistently lower in non-thermal contributions for both conical and flat configurations. However, as the equivalence ratio becomes leaner the difference in non-thermal NO formation between the alcohols and alkanes is reduced, which will later be shown to be likely attributable a greater proportion of the non-thermal contribution through the non-HC-related mechanisms. In previous work by the authors in [26], a technique for tracing the flux of nitrogen through the various NO formation sub-mechanisms was presented. This algorithm works by identifying and distinguishing the contributions to total NO formation (or consumption) due to several known NO sub-mechanisms, including: Thermal, Prompt, NO-HCN reburn, NO2 , N2 O, NNH, HNO, and several other minor mechanisms. Initiation reactions for each sub-mechanism are identified based on recommendations

370

M.D. Bohon et al. / Combustion and Flame 194 (2018) 363–375

millimeter, most of the non-thermal mechanisms have completed their total contribution, and the majority of the remainder is attributable to the thermal mechanism. The flat flames behave similarly to the conical flames, with a much smaller contribution from the thermal mechanism due to the lower temperature regime. In both configurations, the Reburn mechanism serves as a significant sink, peaking slightly before the Prompt mechanism. While the magnitude of the peak rate of consumption through the Reburn mechanism is greater in the conical flames, the relative rate of consumption in the flat flames indicates that the Reburn mechanism is more significant in the lower temperature flames. It is important to also consider, that while the alcohol flames exhibit lower rates of Prompt NO formation than the alkane flames, they exhibit lower rates of NO consumption through the Reburn mechanism. This complicates the attribute of reductions in NO formation observed in alcohol flames simply to lower hydrocarbon radical concentrations. Due to the competing effects of the Prompt and Reburn mechanisms, both controlled by the availability of hydrocarbon radicals, the effect of reduced CH concentration is expressed in both sub-mechanisms. This approach was taken one step further by calculating the temporal integral of the rate of production (ROP) for each submechanism determined above in order to yield an estimate of the predicted number density of the contribution of each submechanism to the total NO formation. XNO, k , the molar fraction of NO attributable to sub-mechanism k, can be estimated from the model as shown in equation below

XNO,k ≈ Fig. 9. Non-thermal NO contributions.

from literature and the flux of nitrogen through the mechanisms at each position in the flame is identified. In the previous work, the results from this algorithm were presented for the simulations of the flat flames measured here and greater than 97% of the total predicted NO rate of formation was accounted for by the algorithm. While more detail on the implementation of this algorithm is available in Appendix C and in the above cited work, here it should suffice to say that the approach considers a network of nitrogen containing species with an instantaneous molar flux between species predicted by the CHEMKIN model. The algorithm steps through each species at each position in the flame and totals the flux of nitrogen into and out of the target species. By identifying the NO sub-mechanism initiation reactions a priori, the flux of nitrogen atoms for each sub-mechanism can be flagged and traced to ultimately contribute to NO. The benefit of this approach is clear, for example, when one considers reactions forming NO from monotonic nitrogen N which participates in the majority of NO formation sub-mechanisms. The final output of this algorithm is a rate of production for each sub-mechanism, the sum of which yields the total predicted NO rate of production. The result of applying this algorithm is the attribution of the total molar flux of NO production to its corresponding source. An example of this can be seen in Fig. 10, which shows the contributions of each of the studied NO sub-mechanisms for stoichiometric C3 flames in both burner configurations. Also plotted in the circle markers is the predicted total rate of NO production determined by CHEMKIN as well as the summed contributions from the attributed sub-mechanisms plotted in the dashed line. One can see that the algorithm does a good job of attributing a large percentage of the total flux to the corresponding sub-mechanisms. One can see that in the conical flames, the large early peak in NO formation is primarily from the Prompt mechanism, followed by a shorter, broader contribution from NNX. Within the first

RuT p



ROPk dt

(9)

where Ru is the universal gas constant and T and p are the temperature and pressure through the flame. This estimation is somewhat limited by the fact that it does not account directly for diffusion. The CHEMKIN solution from which the ROPk terms are derived does consider diffusion, however this approach of integrating the rate of production only considers the NO source term and neglects the diffusion of NO upstream. The net effect of neglecting diffusion on the locally integrated NO concentration depends on the gas velocity, in that flames with a higher gas velocity (such as the conical flames in comparison with the flat flames) yield a higher Péclet number. However, if considering the integral through the entire flame, and not only to local regions in the flame, the integrated rate of production yields a sense of the number of molecules produced through the entire flame. As a consequence, the result of the integration extended into the post-flame should still be indicative of the net contributions of NO through the entire flame even if disagreements are present in local regions of the flame due to diffusive effects. Figure 11 shows a summary of the these integrated nonthermal NO contributions. In this figure, the results have been collected into two groups dependent on whether the sub-mechanism involved hydrocarbon (HC) species or not non-HC. Across the range of equivalence ratios, the non-thermal contributions for both HC and non-HC sub-mechanisms are lower for alcohols than for alkanes. Additionally, the similarity in the trends between the two burner configurations indicates that these non-thermal NO contributions are indeed less affected by the temperature than the Thermal mechanism. The contributions from HC related sub-mechanisms quite reasonably diverge as the equivalence ratio increases due to the greater availability of unmatched hydrocarbon species while under lean conditions the difference in the contributions is much less. On the other hand, the contributions from the non-HC sub-mechanisms exhibit the greatest difference around stoichiometric conditions and converge under richer and leaner conditions. One can also see that as the size of the fuel molecule increases, the difference between the contributions

M.D. Bohon et al. / Combustion and Flame 194 (2018) 363–375

371

Fig. 10. Demonstration of the nitrogen flux accounting scheme applied to stoichiometric propane, n-propanol, and i-propanol conical and flat flames.

from the HC-related sub-mechanisms decreases. The conventional wisdom concerning the lower formation of NO in alcohol flames was due to lower CH radical concentrations. However, from these figures, it is apparent that under lean to stoichiometric conditions, a significant portion of the reduction in non-thermal NO formation in these alcohol flames is also attributable to the non-HC related mechanisms. However as the equivalence ratio becomes richer, the significant difference in the availability of CH radicals in the alcohol flames results in a lower formation of NO. 4. Conclusions This work presented a series of measurements of NO concentration in two different premixed flame configurations: a conical Bunsen-type flame and a McKenna burner stabilized

flat flame. Measurements of NO concentration within the flame were conducted using quantified NO PLIF with excitation and emission around 226 and 236 nm, respectively. The quenching rate was based on the temperature and composition of major colliding species, the concentrations of which were determined from detailed chemical kinetic models of the flames. The temperatures in the flat flames were taken from OH-PLIF thermometry measurements previously conducted by the authors, while the conical flames used the calculated temperatures from the model. This work builds off of previous work by the authors by investigating the pathways to NO formation under two different flame temperature regimes. From these experiments, detailed profiles of NO concentrations were measured for alcohol and alkane C1–C3 flames, with a focus

372

M.D. Bohon et al. / Combustion and Flame 194 (2018) 363–375

Fig. 11. Non-thermal NO contributions separated by HC and non-HC related mechanisms.

on capturing the formation of NO early in the flames due primarily to non-thermal mechanisms. It was then possible to extract the relative contribution to NO formation from the combined nonthermal sub-mechanisms. Here it was seen that significantly less non-thermal NO formation was observable in the alcohol flames than in the alkane flames, especially when the equivalence ratio became increasing rich. As much as 50% less non-thermal NO was produced in the alcohol flames. Under lean conditions however, the difference between the fuel families was reduced, with alcohols producing about 80–90% of the NO as measured in the alkanes. Using the detailed chemical kinetic models, the nitrogen flux accounting procedure for tracing the pathway of nitrogen fixation and formation of NO was applied to both flames. This approach was extending further to integrate the molar rate of production into a net contribution of NO. An analysis of these contributions of the non-thermal sub-mechanisms grouped into HC-related and non-HC-related sets indicated that the reduction in NO observed in alcohol flames is not solely due to a reduced availability of CH radicals. While the scarcity of CH radicals certainly results in a reduction in the NO formation in rich alcohol flames, under stochiometric and lean conditions reductions in the non-HC related sub-mechanisms, such as the NNH and N2 O sub-mechanisms, also contribute to the overall lower formation of NO.

Acknowledgments The authors would like to acknowledge the financial support of King Abdullah University of Science and Technology (KAUST), Clean Combustion Research Center (CCRC), Thuwal 23955-6900, Saudi Arabia.

Fig. 12. Temperature profiles measured in the flat flames for lean, stochiometric, and rich sample conditions.

Appendix A. Temperature profiles for modeling and LIF quantification Figure 12 shows sample temperature profiles at lean, stoichiometric, and rich equivalence ratios used initially for the modeling and then later for the NO LIF quantification procedure. The temperatures in the flat flames were measured in previous work [26]. They are the combined result of temperature measurements using thermocouples as well as 2λ OH LIF thermometry. Per the discussion in McMillin et al. [57], the fluorescence ratio can be written as:

R12 ≡

Sf1  B1 E1 f B1 (T )g1 (N, T )(χi , N, T ) = C12 Sf 2 B2 E2 fB2 (T )g2 (N, T )(χi , N, T )

(10)

 is dependent on the collection optics, B is the Einstein where C12 coefficient for stimulated absorption, E is the laser energy, fB is the Boltzmann fraction, g is the absorption overlap integral, and  is the fluorescence yield. Assuming the primary temperature dependency comes from the Boltzmann fractions, Eq. (10) can be reduced to equation below.



R12 = C12 exp −

ε12 kT



(11)

As described in previous work, the 2λ OH LIF thermometry and thermocouple measurements are combined by selecting the measurement method which is expected to perform best in each region of the flame. Additionally, C12 from Eq. (11) is calibrated by the thermocouple measurements located 3 mm above the burner surface. The merging approach recognizes that thermocouple measurements have difficulties measuring the temperature in regions with strong temperature gradients, such as in the flame front or in very close proximity to the burner surface. However, they perform better in regions with reduced temperature gradients

M.D. Bohon et al. / Combustion and Flame 194 (2018) 363–375

373

Fig. 13. Temperature profiles computed in the conical flames for lean, stochiometric, and rich sample conditions.

such as the post-flame. Meanwhile, the OH PLIF thermometry technique requires regions of high OH radical concentration (such as in the flame front) to be effective and the technique suffers as the OH concentration decreases (such as the post-flame of rich flames). Therefore, the merging process chooses the technique which is most appropriate through the flame. Generally, the 2λ OH LIF thermometry measurements are used in regions with high OH concentrations and steeper temperature gradients. The thermocouple measurements are used in regions with less steep gradients and lower OH concentrations, such as in the post flame.

Figure 13 shows the temperature profiles determined from the one-dimensional freely propagating flame model in CHEMKIN-Pro. The model accounts for thermal radiation of H2 O, CO, CO2 , and CH4 through optically thin media, which improves the temperature profile in the post-flame. Appendix B. Determination of quenching rates This appendix provides additional detail on the determination of values for fluorescence quantification and quenching rates in the NO PLIF. The quantities in Table B.1 are used to determine the

Table B1 NO LIF quenching and calibration parameters. Term

(Units)

B12 fb A21 NO H2 O CO2 O2 CO N2

σk

Function

Constants c1

c2

c3

c4

Ref.

(m2 /(J · s)) (-) (1/s)

c1 ec2 /T + c3 e−c4 /T

2.38 × 109 −0.2822 5.72 × 106

−1799

0.2183

408.4

[58] [24] [58]

˚ (A) ˚ (A) ˚ (A) ˚ (A) ˚ (A) ˚ (A)

c1 ec2 /T + c3 e−c4 /T c1 (300/T )c2 + c3 e−c4 /T c1 ec2 /T + c3 e−c4 /T c1 ec2 /T + c3 e−c4 /T c1 ec2 /T + c3 e−c4 /T c1 ec2 /T + c3 e−c4 /T

37.3 121.2 38.0 22.0 4.23 1.88

11.7 0.676 173 59.1 128 −2130

60 100 46 4.3 17.5 84

0.011 0.010 0.0022 0.00195 0.00198 0.0121

[39] [39] [39] [39] [39] [39]

374

M.D. Bohon et al. / Combustion and Flame 194 (2018) 363–375

F=



Fj

(17)

Gk

(18)

j

G=

 k

At any given point in the flame, there is no reason to assume that the concentration of species A is steady and not increasing or decreasing. Therefore, attribution of the flow out of species A to specific mechanisms is dependent on the state of A, as defined by Eqs. (17) and (18). Then, by assigning the initiating reactions of the NO sub-mechanisms as the boundary conditions, the flow of nitrogen through the network can be traced according to equations below.



gi,k = Gk 

j i, j

Fig. 14. Demonstration of scheme for tracing the flow of nitrogen leading to NO formation (reprinted from [26]).

Einstein coefficient of absorption, B12 , the Boltzmann population fraction, fb , and the Einstein coefficient of spontaneous emission, A21 . Per Eq. (7) in the text, the fluorescent yield is a function of A21 and the collisional quenching rate, Q21 . This total quenching rate can be estimated through the mole fraction weighted sum of the collisional quenching rate by each species in the bath gas,

Q21 =



Xk Qk .

(12)

The individual quenching rates of each bath gas species can be determined by equation below



Q k = σk n

8kB T

π μk

(13)

where σ k is the thermally averaged cross-section of the quenching species as listed in the second half of Table B.1, n is total number density in the gas, kB is the Boltzmann constant, T is the bath gas temperature, and μk is the reduced mass as shown in below equation.

μk =

Mk MNO Mk + MNO

(14)

Appendix C. Nitrogen flux accounting This appendix provides additional details on the nitrogen flux accounting approach developed in previous work by the authors in [26] and applied in this text. The objective of this previous work was to develop a technique to trace the flow of nitrogen through the network of reaction paths leading to NO. To achieve this goal, the total flux between all nitrogen containing species is calculated at each position in the flame. Then, for each species in the network, the total flux into and out of that species is analyzed. A demonstration of this is shown in Fig. 14. In this demonstration, the objective is to determine the flux from species A to C1 which can be attributed to mechanism M1 (labeled gM1, 1 ). One can see that the flow of nitrogen into species A comes from a number of different sources (B1 through B3 ) and is due to two different mechanisms (M1 and M2 ). The index i denotes the specific submechanism. Index j denotes the network connect from species Bj to A. Index k denotes the network connection from species A to Ck . From this figure, several terms can be defined:

Fj =



fi, j

(15)

gi,k

(16)

i

Gk =

 i

gi,k =

 j

fi, j fi, j

,F > G

G fi, j  k , F < G k Gk

(19) (20)

One of the assumptions of this approach is a relatively large Dahmköhler number, such that the reactions bringing nitrogen into and out of the target species happen quickly relative to convective or diffusive transport to another location in the flame. Within this frozen framework, Eq. (19) assigns the flux in the case when the local concentration is increasing. In this case, the total outflow from the target species is apportioned according to the relative inflow. In the case where the target species is decreasing, Eq. (20) allocates only a fraction of the total outflow based on the inflow. Supplementary material Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.combustflame.2018. 05.024. References [1] S.-C. Li, F.A. Williams, Formation of NOx, CH4 , and C2 species in laminar methanol flames, Symp. (Int.) Combust. (1998) 485–493, doi:10.1021/ jp047340h. [2] S.-C. Li, F. A. Williams, Experimental and numerical studies of NOx formation in two-stage methane–air flames, ASME 1998 International Gas Turbine and Aeroengine Congress and Exhibition (1998b). 10.1115/98-GT-073 [3] J.A. Miller, C.T. Bowman, Mechanism and modeling of nitrogen chemistry in combustion, Prog. Energy Combust. Sci. 15 (4) (1989) 287–338, doi:10.1016/ 0360- 1285(89)90017- 8. [4] C.P. Fenimore, Formation of nitric oxide in premixed hydrocarbon flames, Symp. (Int.) Combust. 13 (1) (1971) 373–380, doi:10.1016/S0082-0784(71) 80040-1. [5] C.P. Fenimore, Studies of fuel-nitrogen species in rich flame gases, Symp. (Int.) Combust. 17 (1) (1979) 661–670, doi:10.1016/S0 082-0784(79)80 065-X. [6] J. Blauwens, B. Smets, J. Peeters, Mechanism of “prompt” no formation in hydrocarbon flames, Symp. (Int.) Combust. 16 (1) (1977) 1055–1064, doi:10.1016/ S0082- 0784(77)80395- 0. [7] L.V. Moskaleva, M.C. Lin, The spin-conserved reaction CH+N2 H+NCN: a major pathway to prompt NO studied by quantum/statistical theory calculations and kinetic modeling of rate constant, Proce. Combust. Inst. 28 (2) (20 0 0) 2393– 2401, doi:10.1016/S0 082-0784(0 0)80652-9. [8] P. Dagaut, P. Glarborg, M.U. Alzueta, The oxidation of hydrogen cyanide and related chemistry, Prog. Energy Combust. Sci. 34 (1) (2008) 1–46, doi:10.1016/ j.pecs.20 07.02.0 04. [9] N. Lamoureux, H.E. Merhubi, L. Pillier, S. de Persis, P. Desgroux, Modeling of NO formation in low pressure premixed flames, Combust. Flame 163 (2016) 557–575, doi:10.1016/j.combustflame.2015.11.007. [10] D. Stapf, W. Leuckel, Flow reactor studies and testing of comprehensive mechanisms for NOx reburning, Symp. (Int.) Combust. 26 (2) (1996) 2083–2090, doi:10.1016/S0 082-0784(96)80 032-4. [11] T. Etzkorn, S. Muris, J. Wolfrum, C. Dembny, H. Bockhorn, P.F. Nelson, A. AttiaShahin, J. Warnatz, Destruction and formation of NO in low pressure stoichiometric CH4 /O2 flames, Symp. (Int.) Combust. 24 (1) (1992) 925–932, doi:10. 1016/S0082-0784(06)80110-4. [12] P. Dagaut, F. Lecomte, S. Chevailler, M. Cathonnet, Experimental and detailed kinetic modeling of nitric oxide reduction by a natural gas blend in simulated reburning conditions, Combust. Sci. Technol. 139 (1) (1998) 329–363, doi:10. 1080/00102209808952093.

M.D. Bohon et al. / Combustion and Flame 194 (2018) 363–375 [13] L. Prada, J.A. Miller, Reburning using several hydrocarbon fuels: a kinetic modeling study, Combust. Sci. Technol. 132 (1–6) (1998) 225–250, doi:10.1080/ 00102209808952016. [14] G.M. Johnson, M.Y. Smith, Emissions of nitrogen dioxide from a large gasturbine power station, Combust. Sci. Technol. 19 (1–2) (1978) 67–70, doi:10. 1080/00102207808946865. [15] R.W. Coutant, E.L. Merryman, A. Levy, Formation of NO2 in range-top burners, Environ. Int. 8 (1–6) (1982) 185–192, doi:10.1016/0160- 4120(82)90027- 7. [16] C.P. Fenimore, The ratio NO2 /NO in fuel-lean flames, Combust. Flame 25 (1975) 85–90, doi:10.1016/0010- 2180(75)90071- 1. [17] S.M. Correa, A Review of NOx formation under gas-turbine combustion conditions, Combust. Sci. Technol. 87 (1992) 329–362, doi:10.1080/ 00102209208947221. [18] 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, doi:10.1016/0010- 2180(81)90008- 0. [19] J.W. Bozzelli, A.M. Dean, 0 + NNH: a possible new route for NOx formation in flames, Int. J. Chem. Kinet. 27 (11) (1995) 1097–1109. [20] J.E. Harrington, G.P. Smith, P.A. Berg, A.R. Noble, J.B. Jeffries, D.R. Crosley, Evidence for a new NO production mechanism in flames, Proc. Combust. Inst. 26 (2) (1996) 2133–2138, doi:10.1016/S0 082-0784(96)80 038-5. [21] S.J. Klippenstein, L.B. Harding, P. Glarborg, J.A. Miller, The role of NNH in NO formation and control, Combust. Flame 158 (4) (2011) 774–789, doi:10.1016/j. combustflame.2010.12.013. [22] P. Saxena, F.A. Williams, Numerical and experimental studies of ethanol flames, Proc. Combust. Inst. 31 I (2007) 1149–1156, doi:10.1016/j.proci.2006.08.097. [23] G.A. Chung, B. Akih-Kumgeh, G.M.G. Watson, J.M. Bergthorson, NOx formation and flame velocity profiles of iso-and n-isomers of butane and butanol, Proc. Combust. Inst. 34 (1) (2013) 831–838, doi:10.1016/j.proci.2012.06.114. [24] G.M.G. Watson, P. Versailles, J.M. Bergthorson, NO formation in premixed flames of C1–C3 alkanes and alcohols, Combust. Flame 169 (2016) 242–260, doi:10.1016/j.combustflame.2016.04.015. [25] G.M.G. Watson, P. Versailles, J.M. Bergthorson, NO formation in rich premixed flames of C1–C4 alkanes and alcohols, Proc. Combust. Inst. 36 (2016) 1–9, doi:10.1016/j.combustflame.2016.04.015. [26] M.D. Bohon, T.F. Guiberti, S.M. Sarathy, W.L. Roberts, Variations in non-thermal NO formation pathways in alcohol flames, Proc. Combust. Inst. 36 (3) (2017) 3995–4002, doi:10.1016/j.proci.2016.05.024. [27] J.M. Seitzman, R.K. Hanson, Two-line planar fluorescence for temporally resolved temperature imaging in a reacting supersonic flow over a body, Appl. Phys. B Photophys. Laser Chem. 57 (6) (1993) 385–391, doi:10.1007/ BF00357380. [28] J.M. Seitzman, R.K. Hanson, P.A. Debarber, C.F. Hess, Application of quantitative two-line OH planar laser-induced fluorescence for temporally resolved planar thermometry in reacting flows., Appl. Opt. 33 (18) (1994) 40 0 0–4012, doi:10. 1364/AO.33.0 040 0 0. [29] C.M. Vagelopoulos, F.N. Egolfopoulos, Direct experimental determination of laminar flame speeds, Symp. (Int.) Combust. 27 (1) (1998) 513–519, doi:10. 1016/S0082- 0784(98)80441- 4. [30] Ö.L. Gülder, Laminar burning velocities of methanol, ethanol and isooctaneair mixtures, Symp. (Int.) Combust. 19 (1) (1982) 275–281, doi:10.1016/ S0082- 0784(82)80198- 7. [31] P.S. Veloo, F.N. Egolfopoulos, Studies of n-propanol, iso-propanol, and propane flames, Combust. Flame 158 (3) (2011) 501–510, doi:10.1016/j.combustflame. 2010.10.001. [32] M. Metghalchi, J.C. Keck, Laminar burning velocity of propane-air mixtures at high temperature and pressure, Combust. Flame 38 (1980) 143–154, doi:10. 1016/0010- 2180(80)90046- 2. [33] M. Metghalchi, J.C. Keck, Burning velocities of mixtures of air with methanol, isooctane, and indolene at high pressure and temperature, Combust. Flame 48 (C) (1982) 191–210, doi:10.1016/0010-2180(82)90127-4. [34] B.E. Battles, R.K. Hanson, Laser-induced fluorescence measurements of NO and OH mole fraction in fuel-lean, high-pressure (1–10 atm) methane flames: fluorescence modeling and experimental validation, J. Quant. Spectrosc. Radiat. Transf. 54 (3) (1995) 521–537, doi:10.1016/0 022-4073(95)0 0 020-L. [35] M.D. Di Rosa, K.G. Klavuhn, R.K. Hanson, LIF spectroscopy of NO and O2 in high-pressure flames, Combust. Sci. Technol. 118 (4–6) (1996) 257–283, doi:10. 1080/00102209608951981. [36] W.G. Bessler, C. Schulz, T. Lee, J.B. Jeffries, R.K. Hanson, Strategies for laserinduced fluorescence detection of nitric oxide in high-pressure flames. I. AX(0,1) excitation., Appl. Opt. 41 (12) (2002) 3547–3557, doi:10.1364/AO.42. 002031.

375

[37] J. Luque, D.R. Crosley, LIFBASE: Database and Spectral Simulation Program (Version 1.5), 1999. SRI International Report MP 99-009. [38] W.P. Partridge, N.M. Laurendeau, Formulation of a dimensionless overlap fraction to account for spectrally distributed interactions in fluorescence studies., Appl. Opt. 34 (15) (1995) 2645–2647, doi:10.1364/AO.34.002645. [39] T.B. Settersten, B.D. Patterson, J.A. Gray, Temperature- and species-dependent quenching of NO A2S+ (v’=0) probed by two-photon laser-induced fluorescence using a picosecond laser, J. Chem. Phys. 124 (23) (2006) 0–14, doi:10. 1063/1.2206783. [40] R.M. Fristrom, Flame zone studies. II. Applicability of one-dimensional models to three-dimensional laminar Bunsen flame fronts, J. Chem. Phys. 24 (4) (1956) 888, doi:10.1063/1.1742628. [41] Q.V. Nguyen, R.W. Dibble, C.D. Carter, G.J. Fiechtner, R.S. Barlow, Raman-LIF measurements of temperature, major species, OH, and NO in a methaneair Bunsen flame, Combust. Flame 105 (4) (1996) 499–510, doi:10.1016/ 0 010-2180(96)0 0226-X. [42] B. Lewis, G. von Elbe, Stability and structure of Burner flames, J. Chem. Phys. 11 (2) (1943) 75, doi:10.1063/1.1723808. [43] G. Andrews, D. Bradley, Determination of burning velocities: a critical review, Combust. Flame 18 (1) (1972) 133–153, doi:10.1016/S0010- 2180(72)80234- 7. [44] D.D.S. Liu, R. MacFarlane, Laminar burning velocities of hydrogen-air and hydrogen-airsteam flames, Combust. Flame 49 (1–3) (1983) 59–71, doi:10.1016/ 0010-2180(83)90151-7. [45] T.C. Wagner, C.R. Ferguson, Bunsen flame hydrodynamics, Combust. Flame 59 (3) (1985) 267–272, doi:10.1016/0010-2180(85)90130-0. [46] F.J. Higuera, Aerodynamics of a slender axisymmetric Bunsen flame with large gas expansion, Combust. Flame 156 (5) (2009) 1063–1067, doi:10.1016/ j.combustflame.2008.10.023. [47] J.R. Reisel, C.D. Carter, N.M. Laurendeau, M.C. Drake, Laser-saturated fluorescence measurements of nitric oxide in laminar, flat, C2 H6 /O2 /N2 flames at atmospheric pressure, Combust. Sci. Technol. 91 (4–6) (1993) 271–295, doi:10. 1080/00102209308907649. [48] J.R. Reisel, C.D. Carter, N.M. Laurendeau, Measurements and modeling of OH and NO in premixed C2 H6 /O2 /N2 flames at atmospheric pressure, Energy Fuels 11 (5) (1997) 1092–1100, doi:10.1021/ef9700341. [49] W.P. Partridge, J.R. Reisel, N.M. Laurendeau, Laser-saturated fluorescence measurements of nitric oxide in an inverse diffusion flame, Combust. Flame 116 (1–2) (1999) 282–290, doi:10.1016/S0010- 2180(98)00030- 3. [50] W.G. Bessler, C. Schulz, T. Lee, J.B. Jeffries, R.K. Hanson, Strategies for laserinduced fluorescence detection of nitric oxide in high-pressure flames. III. Comparison of A-X excitation schemes., Appl. Opt. 42 (24) (2003) 4922–4936, doi:10.1364/AO.42.004922. [51] Reaction Design, CHEMKIN-PRO 15131 (2013). [52] W.K. Metcalfe, S.M. Burke, S.S. Ahmed, H.J. Curran, A hierarchical and comparative kinetic modeling study of C1–C2 hydrocarbon and oxygenated fuels, Int. J. Chem. Kinet. 45 (10) (2013) 638–675, doi:10.1002/kin.20802. [53] S.M. Sarathy, S. Vranckx, K. Yasunaga, M. Mehl, P. Oßwald, W.K. Metcalfe, C.K. Westbrook, W.J. Pitz, K. Kohse-Höinghaus, R.X. Fernandes, H.J. Curran, A comprehensive chemical kinetic combustion model for the four butanol isomers, Combust. Flame 159 (6) (2012) 2028–2055, doi:10.1016/j.combustflame. 2011.12.017. [54] A. El-Bakali, L. Pillier, P. Desgroux, B. Lefort, L. Gasnot, J.F. Pauwels, I. Da Costa, NO prediction in natural gas flames using GDF-Kin®3.0 mechanism NCN and HCN contribution to prompt-NO formation, Fuel 85 (7–8) (2006) 896–909, doi:10.1016/j.fuel.2005.10.012. [55] L.B. Harding, S.J. Klippenstein, J.A. Miller, Kinetics of CH + N2 revisited with multireference methods, J. Phys. Chem. A 112 (3) (2008) 522–532, doi:10.1021/ jp077526r. [56] P. Versailles, G.M.G. Watson, A.C.A. Lipardi, J.M. Bergthorson, Quantitative CH measurements in atmospheric-pressure, premixed flames of C1–C4 alkanes, Combust. Flame 165 (2016) 109–124, doi:10.1016/j.combustflame.2015.11.001. [57] B.K. McMillin, J.L. Palmer, R.K. Hanson, Temporally resolved, two-line fluorescence imaging of NO temperature in a transverse jet in a supersonic cross flow., Appl. Opt. 32 (36) (1993) 7532–7545, doi:10.1364/AO.32.007532. [58] J.R. Reisel, C.D. Carter, N.M. Laurendeau, Einstein coefficients for rotational lines of the (0,0) band of the NO A2Sigma-X2Pi system, J. Quant. Spectrosc. Radiat. Transf. 47 (1) (1992) 43–54.