Investigation of the stability, radiation, and structure of laminar coflow diffusion flames of CH4/NH3 mixtures

The stability, radiation, and structure of laminar axisymmetric CH 4 /NH 3 /air diffusion ﬂames have been studied using photographic images, spectrally resolved measurements of ﬂame radiation, and the spatial distribution of temperature and major species mole fractions obtained by spontaneous Raman scattering. The ﬁt procedure for the Raman spectra of NH 3 includes a hitherto unquantiﬁed overtone feature, whose inclusion in the ﬁt signiﬁcantly improves the NH 3 fraction obtained. Nitrogen is used to replace NH 3 to separate chemical effects of NH 3 addition from those due to dilution. The results show that NH 3 addition drastically reduces radiation from carbon-containing species, with progressively increasing strong chemiluminescence from excited NO 2 and NH 2 , indicating a substantial change in ﬂame chemistry. While the Rayleigh/Mie scattering from soot particles is still observed in the Raman spectra at 28% NH 3 addition, 46% NH 3 in the fuel is seen to suppresses soot formation effectively. The measured axial and radial pro-ﬁles of temperature and major species indicate a substantial contribution from radial transport from the reaction zone, seriously complicating the relation between composition, mixture fraction, and the corresponding equilibrium temperature and mole fractions.


Introduction
The drive to reduce the carbon intensity of the energy system has generated much interest in the use of carbon-free fuels in combustion systems, such as ammonia (NH 3 ).Ammonia can be produced from renewable hydrogen (H 2 ) and nitrogen (N 2 ), and a global distribution network already exists [1] , albeit for NH 3 as a feedstock.However, the use of NH 3 in the fuel supply also presents considerable technical challenges and public concerns, particularly associated with its toxicity and high reactivity with container materials and water (H 2 O) [2] .From the perspective of its use as a fuel, the combustion properties of NH 3 are substantially different from those of gaseous fuels such as methane (CH 4 ) and Hydrogen (H 2 ) [3 , 4] .The burning velocities of NH 3 /air mixtures are much lower than those of hydrocarbon fuels [5] (5 times lower than CH 4 ) and NH 3 does not autoignite in the temperature range at which CH 4 or H 2 does ignite [6 , 7] .This complicates the use of pure NH 3 as a "simple" replacement fuel in the current generation of combustion equipment and, in the near term, admixing NH 3 with other fuels, such as CH 4 , will facilitate its use, as discussed in [4] .
While much effort has concentrated on characterizing premixed systems, with an eye towards use in engines and turbines [4] , most industrial combustion equipment uses non-premixed burners.Axisymmetric laminar diffusion flames are commonly used as a model system to provide insight into the elementary physical/chemical processes occurring in flames of non-premixed burners.Comprehensive experimental and computational studies of the structures of these flames have been performed, particularly using CH 4 as a fuel (see, e.g., [8][9][10] ).However, less attention has been paid to the structure of the laminar diffusion flames with NH 3 in the fuel.Earlier work studied NH 3 -doped flames focused on understanding the conversion of NH 3 to NO x in coflow systems.Sullivan et al. [11] investigated NO x formation in laminar, ammonia-seeded, nitrogen-diluted, CH 4 diffusion flames.They measured the mole fractions of stable species by sampling the exhaust gasses and compared the results with those of simulations of the flame structure.Woo et al. [12] investigated the formation of nitrogen oxides in non-premixed coflow CH 4 jet flames, with variable oxygen fraction in the oxidizer, experimentally and numerically, also analyzing the NO x fraction in the exhaust.More detailed information on the conversion of NH 3 to NO was obtained by Bell et al. [13] , who measured the spatial distributions of temperature and NO distributions in laminar, nitrogen-diluted CH 4 diffusion flames seeded with NH 3 in the fuel using Laser-Induced Fluorescence (LIF).The good agreement between the measurements and 2-D numerical simulations lent credence to the conversion mechanism used.While the NH 3 fraction in the fuel in these "seeding" studies was generally below 15%, other studies considered NH 3 as a major fuel component.Um et al. [14] studied NH 3 addition (up to 90% of the fuel) to nonpremixed H 2 /air to mitigate safety issues associated with hydrogen combustion and examined the effects of NH 3 on the stability limits and nitrogen oxide emissions.As to be expected when admixing a slow-burning component with a fast-burning fuel, they observed a significant reduction in the stability limits, which they considered advantageous for H 2 safety.Montgomery et al. [15] examined CH 4 as a combustion enhancer for NH 3 and investigated the effect of up to 40% NH 3 in CH 4 on soot formation in coflow diffusion flames, using microprobe sampling to determine stable species fractions and pyrometry for soot, temperatures and volume fractions.To separate possible chemical effects from dilution, they compared the results of NH 3 /CH 4 mixtures with those of N 2 /CH 4 mixtures.They observed that NH 3 strongly suppressed soot formation relative to N 2 ; the measured soot volume fraction was reduced by a factor of 10 at fractions of 20% or higher.Comparison of the measurements with 2-D detailed computations indicated the need for improvement in the interaction between nitrogen species and C 3 and higher hydrocarbons in the chemical mechanism.We note that the strong decrease in soot volume fraction caused by NH 3 limited their temperature measurements to 20% NH 3 or lower.
Natural gas is an attractive stability-enhancing fuel component for large-scale use of NH 3 , since the existing infrastructure provides a ready supply to many residential, commercial and industrial combustion systems.Maximizing the NH 3 fraction in a NH 3 /natural gas system will be a tradeoff between acceptable flame stability, NO x emissions and other unwanted effects such as soot formation.In this paper, we report the stability, natural flame emission and flame structure (distribution of temperature and major species) of laminar coflow CH 4 /NH 3 diffusion flames, in which CH 4 is used as a surrogate for natural gas.
While the examination of the structure of pure NH 3 /air diffusion flames would provide unadulterated information on the physical-chemical processes occurring in this system, laminar NH 3 diffusion flames without additives do not appear to be stable with room-temperature reactants (see [14] and below).Thus, using stabilizing additives is a necessity for the analysis of diffusion flame structure.Here we investigate increasing NH 3 fraction on the structure of laminar-coflow, non-premixed CH 4 /air flames.For this purpose, we report spectrally resolved flame emissions and distributions of major species obtained using spontaneous Raman scattering.To distinguish chemical effects of NH 3 addition from those due to dilution, the measurements are performed in flames where N 2 is used as a fuel additive instead of NH 3 , similar to the method used in [15] .

Experimental methods
In the present work, the measurements are carried out using a burner setup similar to that described in [16] , which is composed of an inner tube (i.d.7 mm) carrying the fuel mixture surrounded by an air-coflow annulus (i.d.58 mm).The fuel/air coflow system is confined by a quartz tube to improve the stability of the flame and to prevent mixing with ambient air.The mass flow rates of CH 4 , NH 3 , N 2 , and air are measured by calibrated Bronkhorst mass flow meters (accuracy 1% of full scale).In all flames examined, the average velocities of fuel and coflow air are fixed at 13.9 cm/s.The visible appearance of flames is recorded using a digital photo camera.The burner is moved axially and radially by a positioner (Parker) with positioning uncertainty less than 0.1 mm.
The flame temperature and mole fractions of the major species (CH 4 , NH 3 , CO 2 , H 2 O, N 2 , CO, O 2 , H 2 ) are measured using spontaneous Raman scattering.The optical setup for these measurements is essentially identical to that described in reference [17] .In short, radiation from a Spectra Physics Empower Nd:YLF laser (527 nm, average power 30 W) is focused in the center of the burner by a lens with a focal length of 500 mm.The scattered radiation is collected at right angles by a lens (f/2.8,f = 300 mm) and, after passing through a long-pass filter, dispersed by a f/3.9 spectrometer (Acton Research Spectra Pro 2300i).The spectrometer is oriented such that its entrance slit is parallel to the direction of laser beam propagation.At the exit plane of the spectrometer, a PI-Max intensified 1024 × 1024 pixel CCD camera (Princeton Instruments, 13 μm pixel size) is installed.In the present study, 40 camera pixels are binned in the horizontal direction corresponding to a distance of roughly 1 mm along the laser beam, yielding the spatial resolution of the measurements of temperature and mole fraction.The vertical resolution is determined by the diameter of the laser beam and estimated to be roughly 1 mm.At every measuring point, the Raman spectrum is measured twice (see below).The spectrum is first measured in the spectral region 560 -680 nm with resolution ∼1 nm, to obtain the data for the species fractions and then the region 597 -604 nm is measured again at resolution ∼0.1 nm to obtain the temperature from the N 2 spectrum.Binning 2 camera pixels in the spectral direction allows increasing signal to noise ratio without deteriorating the spectral resolution.Further improvement in the signal-to-noise ratio is achieved by measuring the signal obtained by rotating the polarization of the laser radiation by 90 °using a half-wave plate and subtracting this signal from the measured Raman spectrum [17] .While effective at eliminating unpolarized background signals of moderate intensity, the properties of the reaction products in the flames studied here expose shortcomings of this method, as discussed in Section 4 , below.
Spectrally resolved flame radiation (both chemiluminescence and soot radiation) is measured using the collecting part of the Raman setup when the pumping laser is switched off.The emission is collected from the flame at heights from 10 to 30 mm above the fuel tube exit.The depth of focus of the collection system is relatively large ( ∼15 mm), which allows no spatial resolution in the radial direction.The measurements are performed in the spectral range 350-700 nm with a resolution of 1 nm.

Analysis of measured Raman spectra
The Raman spectra of N 2 , CO 2 , CO, O 2 , H 2 O and CH 4 are calculated according to the procedures described in reference [17] .In contrast to characterizing and quantifying the Raman spectra of NH 3 experimentally, as done in [18] , here, we follow the method used for CH 4 [17] and calculate the temperature-dependent Raman spectrum of NH 3 using the harmonic approximation [19] , while completely neglecting rotational structure.This approximation is sufficient for deriving the mole fraction of NH 3 (and CH 4 ) at low temperatures, where rotational structure cannot be resolved, while at elevated temperatures the fractions of these species are negligible.To measure mole fractions of all main species simultaneously, a wide spectral range from 560 -680 nm needs to be covered that can be only done by measuring Raman spectra in relatively low resolution.To avoid ambiguities in deriving the temperature from the low resolution spectrum [17] , we obtain the temperature from the Raman spectra of N 2 measured with moderate resolution as described above, allowing some resolution of the vibrational and rotational structure.Because of stability of laminar flames studied, repeated Raman measurements do not introduce substantial additional noise.The daily calibration of Raman spectra is performed by measurement of Raman spectra in air at room temperature to determine the proportionality factor and lineshape parameters necessary for fitting individual spectral features [17] .
The performance of the Raman measurements has been verified at room and flame temperatures using a McKenna Products flat-flame burner.At room temperature, pure N 2 , CH 4 , O 2 , and CO 2 were flowed through the burner head.At room temperature, the accuracy of the measured species fractions is ∼5%.The accuracy of the setup at high temperatures has been described previously [20] .Briefly, the accuracy is estimated based on the use of reference flames described in a previous investigation [17] .Premixed flames at phi = 0.8 and 1.3 are stabilized on a McKenna Products burner at high enough mass flux to be essentially free flames, i.e., so that there is no heat transfer to the burner (flow rates of the fuel/air mixture 46 l/min and 40 l/min for the lean and rich flames, respectively).As observed in [17] , the measured species fractions in the post-flame zone of McKenna burner were always within 10% of the expected equilibrium values for species with mole fraction ≥ 0.05, while the temperature was within ∼40 K of the equilibrium value.We also report that varying the temperature by varying the mass flux of the fuel/air mixture through the burner at ϕ= 0.8 gave experimental results that were within the same margins when compared with calculations using the PREMIX code in CHEMKIN package [21] using the GRI-Mech 3.0 chemical mechanism [22] .Consequently, we take 10% and 40 K as the accuracy of the Raman measurements of these species and temperature.
To assign the spectral features of NH 3 and CH 4 to be used here, the room-temperature spectra were obtained from mixtures having compositions 50% NH 3 /50% N 2 , 50% CH 4 /50% N 2 , and 100% N 2 .The measured Raman spectra in these mixtures in the spectral region 210 0-360 0 cm −1 are given in Fig. S1 in the Supplementary Material.In the NH 3 /N 2 mixture, two spectral features arising from NH 3 are observed in addition to the N 2 transition at 2331 cm −1 .One spectral feature at 3334.2 cm −1 is due to a totally symmetric Raman-active vibrational mode [23] and the other at 3255 cm −1 is from the overtone of a Raman-inactive non-totally symmetric vibration [24] .When neglecting the frequency dependence of spectral sensitivity of the detection system, a mixture with equal species mole fractions should show Raman intensities whose ratio is equal to the ratio of the species cross sections.The data in Fig. S1 show that the ratio of the NH 3 Raman intensity at 3334.2 cm −1 to that of N 2 is ∼4, while the ratio of the CH 4 peak at 2914.2 cm −1 to the N 2 peak is ∼6.These ratios are in very good agreement with the reported values of 3.6 for the NH 3 transition [25] and 6.0 for CH 4 [26] .To our knowledge, there is no literature data available for the cross section of the NH 3 overtone Raman feature at 3255 cm −1 .Comparison of the intensity of this peak with that of N 2 gives a value of 0.7 for the relative cross section of overtone of the NH 3 non-totally symmetric vibration mode, which is subsequently used in the fitting program to simulate the Raman spectra; including this feature in the fitting procedure significantly improves the accuracy of the NH 3 fraction.We note that the ratio of the intensity of the totally symmetric mode to that of the overtone (unassigned in [18] ) is in good agreement with that shown in [18] .Figure 1 shows an example of the measured and simulated spectra for a 80% NH 3 /20% N 2 mixture at room temperature.The NH 3 mole fraction obtained from the simulation is ∼0.75 and indicates the accuracy of ∼10% for the NH 3 Raman measurements.

Flame images and spectrally resolved flame radiation
For the conditions used here, stable flames are obtained when X NH3 in the CH 4 /NH 3 fuel varies from 0 to 0.46.Further increasing X NH3 results in flame extinction.In contrast, the CH 4 /N 2 fuel are stable up to X N2 ∼0.68.As can be seen from the images of flames shown in Fig. 2 , both CH 4 /NH 3 /air and CH 4 /N 2 /air flames have the typical conical shape of axisymmetric diffusion flames.Increasing the fractions of N 2 and NH 3 in the fuel mixtures from 0.0 to 0.46, decreases the height of the flames from ∼65 mm for the pure CH 4 fuel to ∼30 mm for the CH 4 /N 2 mixture ( Fig. 2 f) and ∼45 mm for CH 4 /NH 3 ( Fig. 2 l) flames.Since exit velocity of the mixtures is constant, from simple Burke-Schumann theory [27] we expect the decrease in the flame height to follow the reduction in the stoichiometric air requirement for the fuel, arising from the decreased mixing time to reach the stoichiometric mixture at the centerline.The relative decrease in flame height observed here is consistent with the ratios of the air requirement, going from 9.54 for pure CH 4 to 5.1 for 46% N 2 in CH 4 and 6.8 for 46% NH 3 in CH 4 .
As discussed in more detail below, the pure CH 4 /air flame has the typical bright yellow soot emission at axial distances above ∼10 mm, and the blue emissions characteristic of hydrocarbon oxidation close to the burner exit.Increasing X N2 in the CH 4 /N 2 fuel decreases the spatial extent of the yellow emission zone, while the visibility of the blue emission zone improves ( Figs. 2 b-f).In as much as the soot volume fraction in these flames has its maximum along the centerline [15] , we still observe yellow soot emissions in the flame tip at 46% N 2 .In contrast, increasing X NH3 in the CH 4 /NH 3 fuel mixture ( Figs. 2 j-l) rapidly decreases the extent of the yellow (soot) emission, while the increasing reddish-orange emission visible in the entire cone progressively overwhelms the blue emission at the base of the flame in the photographs.At X NH3 ∼0.28, the relatively bright spot at the flame tip suggests residual soot, while neither yellow nor blue emissions can be discerned in the image, indicating a substantial change in flame chemistry increasing NH 3 fraction.We also note the increasing intensity of the orange emissions from the reaction zone as the NH 3 fraction increases from 28% to 46%.
As suggested by the flame images, the spectrally resolved measurements of flame radiation show a strong dependence of the flame emission spectra upon addition of NH 3 and N 2 to fuel. Figure 3 presents the emission spectra in the wavelength range 350-700 nm at heights 10 mm and 30 mm above the burner exit, and mole fractions X N2 and X NH3 of 0.0, 0.28, and 0.46.As can be seen, the structure of the emission spectrum of the pure CH 4 /air ( Fig. 3 a) at 10 mm is significantly different from that at 30 mm.While at 30 mm the spectrum is dominated by broadband soot emission with progressively increasing intensity with increasing wavelength, the spectrum at 10 mm consists of sharp spectral features at 390.5 nm, 430.4 nm, 436.3 nm, 468.4 nm, 516.1 nm, and 563 nm superimposed on a broadband background with a maximum beyond 600 nm.Increasing X N2 in the CH 4 /N 2 fuel to 0.28 and 0.46 ( Figs. 3 b and c) results in decreasing intensity of the broadband background at 10 mm, with a shift in the maximum of the background to ∼450 nm, while the intensity of the narrow spectral features is nearly unchanged.We ascribe the background at 10 mm in Figs. 3 b and c to radiation from the excited state of the CO 2 molecule [28] .While this background may also be present in Fig. 3 a (pure CH 4 ), soot emission is the most likely source of the background observed: the strong decrease in its intensity with increasing N 2 dilution observed at 30 mm is consistent with the disappearance of the "reddish" background emission at 10 mm.The narrow spectral features observed are assigned (shown in Fig. 3 b) to the excited states of CH and C 2 radicals [29] , produced during the oxidation of the fuel in the reaction zone.The modest decrease in the intensity of the emissions from these radicals observed in Figs. 3 a-c is consistent with the increasing dilution by nitrogen.
In comparison, the admixture of NH 3 results in significant changes in structure of the emission spectra.We first observe that at 28% NH 3 in the fuel ( Fig. 3 d) the intensity of the broadband emission at 30 mm has decreased strongly, being comparable to that seen at 46% N 2 ( Fig. 3 c), while a few sharper features begin to appear.This observation is consistent with the substantial disappearance of soot reported in [15] , where, under the conditions of  their experiments, the soot emission was negligible as compared to chemiluminescence from excited radicals from the fuel > 20% NH 3 .However, as will be seen in Section 4.2 , below, the Raman measurements indicate significant quantities of soot in this flame.The sharper peaks are more easily identifiable in the spectrum at 10 mm in Fig. 3 d.In addition to peaks from CH * seen in Fig. 3 b, new narrow spectral features appear at wavelengths 543.6 nm, 571.3 nm, 604.2 nm, 630.2 nm, and 665.2 nm.As indicated in Fig. 3 d, the positions of these narrow features are consistent with the structure of the NH 2 ammonia α band [30] .The NH 2 ammonia α band are superimposed on a broad background that we assign to the chemiluminescent emission of Nitrogen Dioxide (NO 2 ) following the recombination of NO with O atoms [31][32][33][34][35][36][37] , which produces NO 2 in the excited A 2 B 2 , B 2 B 1 and α 2 A 2 states that subsequently decay to the X 2 A 1 ground state with emission of a photon [34 , 36] .We note that the absence of a soot background in the spectrum at 10 mm of the CH 4 /N 2 mixture at 28% N 2 in Fig. 3 b, where the soot emission at 30 mm is roughly 5 times more in-  tense than in the 28% NH 3 mixture at the same height above the burner, argues against any soot contribution at 10 mm in Fig. 3 d.At 28% NH 3 in the fuel, the peaks from C * 2 have all but disappeared.Increasing the NH 3 fraction to 46% ( Fig. 3 e) results in essentially identical spectra at both axial distances, including a suggestion of CH * at 10 mm.

Raman spectra
Raman spectra in the region 150 0-40 0 0 cm −1 , measured in the flames at the centerline, are shown in Fig. 4 at 10 mm and 30 mm above the burner and at X N2 and X NH3 of 0, 0.28, and 0.46 in fuel, corresponding to the flame emission data shown in Fig. 3 .The Raman spectrum of the undiluted CH 4 /air flame ( Fig. 4 a) at 10 mm consists of the transitions of N 2 (at 2330 cm −1 ), CH 4 (2914 cm −1 ), and H 2 O (3657 cm −1 ).As expected, increasing X N2 in the CH 4 /N 2 fuel to 0.28 and 0.46 ( Figs. 4 b and c) does not alter the form of the Raman spectrum, while upon addition of NH 3 to the fuel ( Figs. 4 d  and e) the Raman peak of NH 3 at 3334 cm −1 appears.
At 30 mm in Figs. 4 a, b and d, the Raman spectrum is obscured by (partially) polarized scattered laser light from soot particles, which frustrates the subtraction of the signals obtained using different polarizations of the laser radiation (outlined in Section 2 , above).This radiation is also intense enough to "leak" through the long-pass filter in front of the spectrometer.We note that the setup used in [15] could not detect soot at X NH3 > 0.2.Consistent with the observations obtained from the spontaneous emission spectra in Fig. 4 , above, and the conclusions from [15] , at 28% in the mix-  c and e), the background from soot scattering at 30 mm is essentially absent and only Raman transitions of N 2 , NH 3 and H 2 O are observed.We remark that the increased noise in the spectrum at 30 mm in Fig. 4 c is the result of a small residual signal from soot scattering; since the measurements at different polarizations of the laser are performed sequentially, noise in these spectra is not removed in the subtraction process.While less noisy than in Fig. 4 c, the measurements at 30 mm in the CH 4 /NH 3 mixture in Fig. 4 e also show increased noise.In this case, subtraction process is effective at removing the strong unpolarized chemiluminescence from excited NO 2 (and, to a lesser extent, NH 2 ), but at the expense of increased noise as observed for the soot scattering.Since these chemiluminescent emissions are in the desired spectral region for the Raman signals, they fall within the transmission window of the long-pass filter.Anticipating the discussion of the axial profiles, below, we note that at some axial/radial distances the background in the Raman spectra (from soot in the CH 4 /N 2 flames or from intense emission from excited NO 2 in the CH 4 /NH 3 mixtures) was so large as compared to the Raman signal that the noise after the background-subtraction process swamped reliable measurement of a number of species (see further in Section 4.3 ).

Temperature and main components mole fraction profiles
To describe the overall structure of the flames from these mixtures, the axial and radial profiles of temperature and species mole fraction have been measured in at 46% additive in the fuel for both N 2 and NH 3 .

Axial centerline profiles
The axial profiles of temperature and mole fractions of N 2, CH 4 , NH 3  As can be seen from Fig. 5 a, the temperature profiles in the CH 4 /NH 3 /air and CH 4 /N 2 /air flames at heights from 1 to 20 mm are nearly identical and increasing monotonically from ∼300 K to ∼1600 K.The relatively low temperatures in this region argue against substantial heat release from chemical reactions, with conduction from the reaction zone situated in the mixing layer at the radial boundary of the flows.We attribute the rising temperature further downstream in both flames to oxidation in the centerline region.The increased scatter in the temperature profile of the CH 4 /NH 3 flame ( Fig. 5 a) at axial distances 20-30 mm is from the strong excited NO 2 emissions, indicating intense chemical activity.At heights > 30 mm, the measured temperatures in the two flames begin to diverge, albeit by less than 150 K.While the temperature in the CH 4 /NH 3 /air flame increases progressively to ∼1960 K, the temperature in the CH 4 /N 2 /air flame decreases from ∼1870 K at 30 mm to ∼1800 K at 40 mm.These differences are outside both absolute and day-to-day uncertainties of the Raman measurements and are consistent with the simple estimates of flame height indicated in Section 4.1 , above, arising from the differences in sto-ichiometric air requirement.While the leveling of the centerline temperature of the CH 4 /NH 3 flame suggests proximity of the mixture to its stoichiometric value, the decreasing temperature in the CH 4 /N 2 flame indicates that the mixture is becoming lean.We note that the maximum temperatures, which are also measured in the radial profiles, below, are substantially lower than the adiabatic, stoichiometric values for these fuel mixtures, which are 2140 K and 2190 K for the CH 4 /N 2 /air and CH 4 /NH 3 /air flames, respectively.At the relatively low exit velocity here ( ∼14 cm/s) we expect rapid transport processes to make substantial contributions to the species and temperature profiles in the flames studied here [38][39] .The importance of accounting for differential (radial) transport when assessing the degree of non-equilibrium was recently considered in the context of MILD combustion [20] and the species profiles shown below indicate significant differential transport.However, we cannot exclude the possibility of non-equilibrium chemistry, such as non-equilibrium NO/N 2 (see, e.g., [40] ), based on the current data.Quantitative measurements of NO and flame structure calculations using detailed chemistry are needed to ascertain the contributions of these different processes.The measurements presented in Fig. 5 will help assess the veracity of such calculations.
The axial profiles of the major species shown in Fig. 5 reflect the transport and chemical processes occurring in the flames.As can be seen in Fig. 5 b, the measured mole fraction of N 2 rapidly increases from their initial values, corresponding to the fractions in the fuel mixtures, during the first 15 mm.Since the mixture temperatures ( Fig. 5 a) are still below 1500 K, and that the contribution to N 2 mole fraction from the fuel during oxidation is small (only ∼0.025 from NH 3 and ∼0.09 from N 2 as diluent), this increase arises from diffusion of N 2 from the reaction zone.Further downstream, the N 2 mole fraction remains almost constant at value of ∼0.7, which is close to the N 2 mole fraction of 0.72 and 0.69 in the equilibrium stoichiometric CH 4 /N 2 /air and CH 4 /NH 3 /air mixtures, respectively.We note that even in the absence of differential transport the N 2 fraction in and of itself has little utility as an indicator of local equivalence ratio/mixture fraction, since the dependence of it is relatively insensitive to local equivalence ratio and its variation around the stoichiometric value is within the uncertainty of the measurements shown in the figure .As can be seen from Fig. 5 c, for the region accessible for measurement (heights below 15 mm and above 30 mm), using N 2 instead of NH 3 in fuel mixture has no apparent influence on the measured CH 4 profiles.At z > 30 mm the measured CH 4 mole fractions are less than 0.01 in both flames.We note that the data provided in the Supplementary Material in reference [15] , but not discussed in [15] itself, suggest that under similar conditions (40% additive in the fuel) the last 10% of the CH 4 in the CH 4 /N 2 flame was oxidized somewhat more slowly than in the ammoniacontaining flame (i.e., the CH 4 fraction in the Nitrogen-containing flame reached zero at greater axial distance, despite having a lower exit velocity); the soot interference in the measurements reported here prevent adequate comparison.The measured NH 3 axial profile in the CH 4 /NH 3 /air flame shows the similar behavior to that of CH 4 (see Fig. 5 d), decreasing from its initial mole fraction in the fuel to ∼0.1 at z = 15 mm, primarily due to mixing with combustion products from the radial combustion zone.As mentioned above, the strong chemiluminescence from excited NO 2 (and NH 2 ) at z = 20-30 mm indicates oxidation of the fuel along the centerline.We recall that in CH 4 /air diffusion flames (e.g., [9] ) the fuel fraction decreases below 10% before significant heat release occurs at the centerline.
While the CH 4 profiles in the two flames in Fig. 5 c are indistinguishable, the H 2 O profiles in Fig. 5 e are not.Even in the first 15 mm or so, where mixing dominates the profiles, the H 2 O mole fraction is significantly higher in the CH 4 /NH 3 flame than in the CH 4 /N 2 flame, clearly the result of the additional H 2 O production from NH 3 diffusing radially inward from the reaction zone.As discussed above, background from excited NO 2 increased the scatter in the profile for H 2 O in the ammonia-containing flame.At heights above 30 mm, the H 2 O profiles are nearly constant with the measured H 2 O mole fractions in the CH 4 /NH 3 /air flame being ∼50% higher than in the CH 4 /N 2 /air flame.A salient feature of these measurements is that the H 2 O mole fractions are well below those at stoichiometric equilibrium.For the CH 4 /NH 3 fuel, the expected equilibrium mole fraction is nearly 22%, versus ∼13% measured, while for the CH 4 /N 2 mixture, the equilibrium indicates ∼17% as compared to the roughly 9% shown in Fig. 5 e.
The CO 2 mole fractions at the last 5 mm of the measuring domain ( Fig. 5 f) are also essentially constant in both flames, with the data from the ammonia-containing flame showing more scatter as seen in the other profiles.As expected from the stoichiometry of the combustion reaction, CO 2 mole fraction in the CH 4 /N 2 /air flame ( ∼16%) is higher (by roughly 30%) than those in the CH 4 /NH 3 /air flame ( ∼12%).While this difference corresponds to difference in calculated CO 2 equilibrium mole fractions in the respective stoichiometric mixtures, they are substantially above the equilibrium stoichiometric mole fractions themselves ( ∼8% for the N 2 additive and ∼6% for the NH 3 mixture).We emphasize that the differences observed between the measured species fractions and those expected for the stoichiometric equilibrium mixture for both H 2 O and CO 2 are well outside the uncertainty of the current measurements.Such large deviations suggest a strong impact of diffusion on the profiles presented here.Noteworthy is the observation that the measurements reported in [15] at 40% additive in CH 4 (in the Supplementary Material, but not discussed in the paper) also appear to deviate substantially from equilibrium, with H 2 O being well below, and CO 2 well above, the equilibrium mole fractions expected from the mixtures.Here too, the analysis of detailed flamestructure calculations is necessary to assess and quantify the relative importance of molecular transport and chemistry in these flames.

Radial profiles
Figure 6 shows the radial profiles of temperature and mole fractions of CH 4 and NH 3 measured between the radial distances (r) of −8 to 8 mm at z = 5, 10, and 20 mm above the fuel tube exit.As discussed for the axial profiles, chemiluminescent emissions from NO 2 or soot scattering diminished the quality of several of the radial profiles, particularly those of the H 2 O and CO 2 .They are thus not shown here.However, the quality of the H 2 O profiles at z = 5 and 10 mm is, albeit with a few missing radial points at the peak of the NH 3 flame, adequate for illustrative purposes, and are included in Fig. 6 .
The measured temperature profiles in the two flames ( Fig. 6 ad) are nearly identical.At z = 5 mm, the maximum temperature is just above 1800 K and occurs at radial distance ∼5 mm, which is somewhat outside radius of the fuel tube (3.5 mm), as observed in other low-velocity diffusion flames [16 , 39] .At z = 10 and 20 mm, Figs. 6 b and c, the maximum temperature exceeds 1900 K, in agreement with that observed on the centerline in Fig. 5 a.We also note the spreading of the high-temperature region with increasing axial distance.In addition to showing centerline temperatures consistent with the axial profiles in Fig. 5 a, the temperature profiles at 30 mm ( Fig. 6 d) show a "dip" (local minimum) in the radial profile around the centerline for the CH 4 /NH 3 flame, while the profile from the CH 4 /N 2 fuel is essentially flat in this region.This supports the conclusion drawn above that heat release is still occurring in the ammonia-containing flame at this axial distance.Noteworthy is that the presence of 46% NH 3 has no significant impact on the profiles of temperature and CH 4 , despite its relative unreactivity, even at 5 mm above the exit of the fuel tube ( Figs. 6 a and e).As remarked above, the low velocity of the fuel mixture provides residence times that allow strong radial transport.Thus, at 5 mm, heat conduction from the reaction zone has already raised the centerline temperature to ∼700 K. Also noteworthy is the effect of radial molecular diffusion, as evinced in the radial profiles of H 2 O at z = 5 and 10 mm ( Figs. 6 l and m, respectively).Comparing these profiles to those of temperature at the same heights ( Figs. 6 a and b), we observe that the local minima around r = 0 in the H 2 O profiles are much less deep than those for temperature; at z = 10 mm ( Figs. 6 b and m), the "dip" in the H 2 O profile has nearly vanished (the minimum mole fraction at r = 0 is ∼90% of the maximum), while strongly present in the temperature (temperature at the centerline is only ∼65% of the maximum).As discussed above, strong molecular transport complicates the analysis of the mixture in terms of degree of equilibrium.
Similar to the observation in the axial profiles, above, we note that the mole fractions of both CH 4 ( Fig. 6 e) and NH 3 ( Fig. 6 h) at z = 5 mm decrease to zero between r = 3-4 mm, where the temperature increases to roughly 1500 K.

Summary and conclusions
The stability, radiation, and structure of laminar axisymmetric CH 4 /NH 3 /air and CH 4 /N 2 /air diffusion flames have been studied using photographic images, spectrally resolved measurements of flame radiation and the spatial distribution of temperature and major species mole fractions obtained by spontaneous Raman scattering.The fit procedure for the Raman spectra of NH 3 includes a hitherto unquantified overtone feature, whose inclusion in the fit significantly improves the NH 3 fraction obtained.The results from the flame images show that, for the current geometry, stable CH 4 /NH 3 /air flames are obtained when the X NH3 varies from 0.0 to 0.46 while further increasing X NH3 results in flame extinction.As expected from dilution with N 2 in an alkane fuel, increasing X N2 in the CH 4 /N 2 /air flames reduces the spatial extent and intensity of the yellow soot emission zone, while increasing the extent of the visible blue emission zone (arising from carbon-containing species).In contrast, increasing X NH3 in the CH 4 /NH 3 /air flames drastically reduces both blue and yellow emission zones, with pro-gressively increasing strong orange chemiluminescence from excited NH 2 and NO 2 distributed throughout the reaction zone, indicating a substantial change in flame chemistry upon NH 3 addition.Depending on the mole fraction of N 2 or NH 3 , we observe interference in the Raman spectra in certain spatial regions of these flames, which are not entirely removed by the background-suppression methods used in the Raman setup [17] .Rayleigh/Mie scattering from soot particles overwhelms the Raman signals in regions of substantial soot fraction.In CH 4 /NH 3 /air flames, strong emissions from excited NO 2 , which occur at the same wavelengths as the Raman peaks used in the analysis, are effectively suppressed by the experimental method but result in extra noise in the measurements, confounding quantification of the Raman measurements in spatial regions of high NH 3 conversion.
The experimental results show that NH 3 addition suppresses soot formation as reported in the literature [15] .While earlier measurements [15] using pyrometry suggested negligible soot formation at 20% NH 3 in the fuel, scattering from soot is clearly observable in the Raman measurements reported here at 28% NH 3 .Quantification of the soot volume fraction using a more sensitive method will enable more detailed analysis of the impact of NH 3 on soot formation in these flames.As expected from the stoichiometry of the oxidation of the fuels, the axial profiles of H 2 O show substantially more H 2 O formation from the CH 4 /NH 3 fuel.The results show that the measured temperatures and mole fractions of H 2 O and CO 2 are substantially different from their values at adiabatic, stoichiometric equilibrium.While we cannot rule out a contribution from non-equilibrium chemistry (such as the degree to which NO or N 2 is formed [40] ), the spatial profiles of major species, particularly H 2 O, suggest a significant contribution from radial transport [20] .The radial profiles of temperature and H 2 O support this notion, since at z = 5 mm the profiles show more than 7% H 2 O at the centerline, where the temperature is only 700 K.As a result, the relation between composition, mixture fraction and the corresponding equilibrium temperature is different than in the case of equal transport of all species [20] .Computations with detailed transport and chemistry are needed to understand the structure of these flames and the effect of NH 3 addition on the processed
, H 2 O, and CO 2 along the centerline of CH 4 /NH 3 /air and CH 4 /N 2 /air flames are shown in Fig. 5 .The measurements are performed at heights z = 1 -40 mm above the exit of the fuel tube.In this region, the measured mole fractions of O 2, CO, and H 2 in both flames are too low ( < 0.01) to quantify reliably and are not shown in Fig. 5 .As discussed above, scattering of the laser light from soot increased the noise in the CH 4 /N 2 /air flame at heights z = 20 -30 mm, precluding the quantification of the temperature and mole fractions in this region from the Raman spectra.While 46% NH 3 has effectively suppressed soot formation, permitting measurement of temperature, N 2 , CH 4 and NH 3 at these axial distances, the strong chemiluminescence in the CH 4 /NH 3 /air flame described above has increased the noise in the profiles, as can be seen in the additional scatter in the N 2 profile in this region in Fig. 5 b.The relatively low fractions of H 2 O and CO 2 in the flames of both additives (compared to N 2 and the fuel molecules) and the scattering/chemiluminescence interference limited quantification to the axial distances shown in Figs. 5 e (H 2 O) and f (CO 2 ).