HOT PARTICLE IGNITION OF METHANE FLAMES

Proceedings of the Combustion Institute, Volume 29, 2002/pp. 1605 1612 HOT PARTICLE IGNITION OF METHANE FLAMES FOKION N. EGOLFOPOULOS, CHARLES S. CAMPBELL and M. GURHAN ANDAC Department of Aerospace and Mechanical Engineering University of Southern California Los Angeles, CA 90089-1453, USA Ignition by sparks, glowing embers, or other hot particulate materials presents serious safety hazards. This paper describes a detailed numerical study conducted on the bulk ignition of premixed and nonpremixed atmospheric methane flames by hot inert particles in opposed-jet configurations. The coupled conservation equations were solved for both phases along the stagnation streamline, with detailed descriptions of gas-phase chemical kinetics, molecular transport, and radiative heat transfer. Results showed the effects of independently varying the reactant concentrations, strain rate, particle injection orientation, number density, temperature, and velocity. For both premixed and non-premixed configurations, increasing the amount of methane was found to inhibit ignition under all conditions considered. Analysis showed that the H-production is sensitive to the CH 4 r CH 3 r CH 3 O r H path that is strengthened under fuel-lean conditions. While premixed ignition by hot particles was found to be similar to ignition triggered by impinging a mixture on heated air or O 2, differences were observed for non-premixed systems. For nonpremixed particle ignition, increasing the fuel concentration was found to always inhibit ignition, while in non-premixed gas-phase ignition, this inhibiting effect was found to reverse at very low fuel concentrations. This is a result of the noticeably different gas-phase temperature profiles that are induced during particle and gas-phase ignition. Finally, there is a major alteration in the ignition response of non-premixed particle ignition caused by moving the particle seeding from the fuel to the oxidizer side. This is attributed to the different thermal properties of the air and fuel that result in different heat transfer rates between the particles and air, in turn affecting the kinetic paths leading to the production of H radicals. Introduction The presence of dust can induce complex behaviors in a reacting environment. Stokes drag, phoretic, and gravitational forces affect the dynamic behavior of both gas and particle phases. Similarly, the thermal behavior of both phases can be affected by the substantial interphasial temperature differences that may develop due to the large thermal inertia of the particle phase. As for pure-gas studies, a counterflow configuration can be used for fundamental studies of dusty flows [e.g., Refs. 1 5]. Recent work [6 8] by the authors described the (inert) dust-induced extinction in counterflow configurations. In this paper, numerical studies are reported on the opposite problem of the ignition of flames by hot inert particles, and comparisons are made with ignition achieved by hot air impingement [e.g., Ref. 9] The bulk ignition of CH 4 /air mixtures by injection of very small (particle diameter d p 70 lm) hot particles was investigated in the present study. Larger particle sizes will cause local ignition near the particle surfaces [e.g., Refs. 10,11]. But for small particles, the residence time of the combustible mixture near the hot surface is much smaller than the ignition delay, t ign, (typically several milliseconds for CH 4 ). The residence time of a fluid element moving at a velocity U 1 m/s (at the low end of the cases studied here) past a 70 lm particle is 70 ls, which is 1 2 orders of magnitude smaller than the ignition delay. One might argue that deep within the boundary or in the particle wake, gas may reside near the particle longer than t ign, but in that case, the diffusion of radicals will prevent ignition. Suppose that radicals do form near the particle in a CH 4 /air mixture. The mass diffusivities, D, of the critical CH 3 and O radicals at the 2000 K surface temperature are D CH 0.4 cm 2 3,O, /s, while for the all-important H radical, D H 9cm 2 /s. Making the liberal assumption of t ign 1 ms, the corresponding diffusion distance (2Dt ign ) 1/2 is 280 lm for CH 3 and O, or four particle diameters. For the H radicals, the diffusion distances would be about 1100 lm or about 16 particle diameters. (Realistic t ign s would double or triple these numbers.) Now any particle wake would be of the order of a particle size, and the portion of the boundary layer, with sufficient residence time, will be much smaller. Consequently, one can expect radicals to diffuse in the fast-moving stream far from the particle and be swept away, long before ignition can occur. In order for ignition to occur near the surfaces of particles, two criteria must be satisfied. The first is that the residence time of the gas near the particle 1605

1606 LAMINAR FLAMES Ignition Fig. 1. Variation of Y H,max with n p,inj around the ignition point for twin premixed flames at various s. The simulations were conducted for d p 70 lm, u p,inj 400 cm/s, T p,inj 2000 K, and u g,inj 100 cm/s. (t res d p /U) must at least be of the same order as the ignition time, t ign, so that d p /(Ut ign ) must be of order 1 or larger. The second is that radicals must remain near the particles (say within a particle diameter) long enough for ignition to occur. Thus, the ratio of d p to the diffusion distance, (2Dt) 1/2, must be d p /(2Dt) 1/2 1 for single particle ignition to occur. Using t res as a scale, this can be couched in terms of the familiar Peclet number, (Ud p /D) 1/2 Pe 1/2 1orPe k 1. Note that meeting both criteria, d p /(Ut ign ) 1 and Pe Ud p /D k 1, implies that the particles are large. However, small particles will serve as energy sources and will cause ignition if sufficient energy is supplied to raise the bulk mixture to the ignition temperature. This study considers the limit of Pe 1, or d p /(Ut ign ) K 1, where ignition will occur in the bulk gas and not on the particle surface. While the problem of single particle ignition has received much study, this is the only study of which the authors are aware that addresses the opposite, that is, the small Pe and d p /(Ut ign ) limits. Numerical Approach The quasi-one-dimensional model developed in Ref. [6] was used, which is valid along the stagnation streamline of a counterflow established between two opposing burners. The model includes gas-phase equations similar to Ref. [12], but also accounts for both the dynamic and thermal interactions with a dilute dust phase. Additional equations describe the evolution of the particle energy and number density. The code was integrated with the CHEMKIN and Transport [13,14] subroutine libraries with gas kinetics described by GRI 3.0 [15]. Particles are injected from the lower burner only, as shown in the inset of Fig. 1. Note that in all figures, the lower and upper burners correspond to the left and right of the spatial domain, respectively, so that the particles always enter from the left. The ignition states were determined by setting the gas (u g,inj ) and particle (u p,inj ) phase injection velocities and particle phase temperature (T p,inj ) and then increasing the injection particle number density (n p,inj ) until ignition is achieved. The ignition response exhibits the turning-point behavior shown in Fig. 1. To capture this singular behavior, the original code [6] was modified by using one-point continuation [16,17]. A predetermined temperature increment was imposed at the location at which the temperature had maximum slope on the side from which the particles are injected [16,17]. This acts as a new boundary condition (replacing n p,inj ) against which the S curve is single valued. Results and Discussion The studies were conducted for atmospheric premixed and non-premixed CH 4 /air configurations. The particles are assumed to be chemically inert and not to melt. Values of T p,inj varied from 1500 to 2500 K. The nozzle separation was L 1.4 cm, d p 70 lm, and the properties of nickel were used for the particle heat capacity, thermal conductivity, density (q p 8.04 gm/cm 3 ), and emissivity ( p 0.25). The particles were always injected with sufficient momentum to reach the opposing burner in order to avoid numerical complications associated with particle flow reversal [6 8]. The values of u p,inj were varied from 200 to 800 cm/s, while u g,inj was varied from 50 to 200 cm/s based on the ambient air density. For simulations that included excess CH 4 and/ or gas preheating, the u g,inj was adjusted in order to maintain the same momentum at the nozzle exit, thus preserving the location of the gas stagnation plane (GSP). Unless stated otherwise, the simulations were conducted with T p,inj 2000 K, u p,inj 400 cm/s, and u g,inj 100 cm/s. A number of reacting configurations were considered, all with the reactants supplied at ambient temperature: Twin premixed flames by counterflowing two identical CH 4 /air jets (Fig. 1 inset) Single premixed flames by counterflowing CH 4 /air jets against (1) air jets and (2) O 2 jets, with the particles seeded on the mixture side Single premixed flames by counterflowing CH 4 /air jets against air jets with the particles seeded on the air side Non-premixed flames by counterflowing CH 4 /N 2 jets against air jets with particles seeded on the (1) fuel and (2) air side

HOT PARTICLE IGNITION OF METHANE FLAMES 1607 TABLE 1 Elementary reactions of relevance in premixed and non-premixed flame ignition Elementary Reactions R11 CH 4 O r CH 3 OH R52 CH 4 H r CH 3 H 2 R97 CH 4 OH r CH 3 H 2 O R118 CH 3 HO 2 r CH 3 O OH R154 CH 3 O 2 r CH 2 O OH R156 CH 3 CH 3 M r C 2 H 6 M R159 CH 3 CH 2 O r HCO CH 4 R56 CH 3 O M r CH 2 O H M R168 CH 3 O O 2 r CH 2 O HO 2 R165 HCO M r CO H M R166 HCO O 2 r HO 2 CO R37 H O 2 r OH O R73 C 2 H 5 M r C 2 H 4 H M R157 CH 3 CH 3 r C 2 H 5 H Comments CH 4 consuming CH 4 consuming, H consuming CH 4 consuming CH 3 consuming, CH 3 O producing CH 3 consuming CH 3 consuming CH 3 consuming, HCO producing CH 3 O consuming, H producing CH 3 O consuming, HO 2 producing HCO consuming, H producing HCO consuming, HO 2 producing H consuming (main branching) H producing H producing The reaction numbers correspond to those in the GRI 3.0 mechanism [15]. To help understand the chemical kinetics of flame ignition, Table 1 lists the reactions referred to below. Fig. 2. Normalized sensitivity coefficients for Y H,max at the ignition states of the twin premixed flames shown in Fig. 1. In order to compare the response of particle ignition to that of gas-phase ignition, selected studies were conducted without particle seeding: Premixed flame ignition by counterflowing ambient temperature CH 4 /air jets against (1) heated air jets and (2) heated N 2 jets Non-premixed flame ignition by counterflowing heated air jets against (1) ambient temperature CH 4 /N 2 jets and (2) heated CH 4 /N 2 jets Premixed Flame Ignition Figure 1 depicts the variation of the maximum H mass fraction (Y H,max ) with n p,inj for premixed twinflame ignition for three different equivalence ratios, 0.5, 1.0, and 1.5. For each, the maximum temperature at ignition, (T max ) ign, was determined. Note that leaner mixtures ignite easier, as indicated by the lower values of the injection number density at ignition, (n p,inj ) ign, and (T max ) ign at smaller. The retarding effect of CH 4 has previously been reported for pure gas-phase systems [e.g., Ref. 9]. To begin to understand the ignition response in Fig. 1, Fig. 2 depicts results of sensitivity analysis performed on Y H,max at ignition. The H radical was chosen because of its relevance to the main branching reaction R37, which is largely responsible for the jump from the slow-chemistry branch to the fastchemistry branch of the S-curve. The sensitivity results indicate the central role of CH 3. Its reactions with O 2 and HO 2 generate CH 3 O and CH 2 O, whose subsequent reactions are key to H production. Also note the retarding effects of the propagation reaction R52 and the recombination reaction R156. R52 consumes H radicals to form the less reactive CH 3 and H 2, while R156 recombines the CH 3 radicals into C 2 H 6, which does not contribute to the branching sequence [e.g., Ref. 9]. The heat transfer from the hot particles generates a spatially bell-shaped gas temperature distribution. Fig. 3a depicts the gas (T g ), and particle (T p ) temperature profiles at ignition for 0.5. As gas is

1608 LAMINAR FLAMES Ignition Fig. 3. Structure of the reacting layers at the ignition states of the twin premixed flames shown in Fig. 1. Spatial variations of (a) T g, T p, and normalized particle number density for 0.5 and (b) the rates of reactions R37 and R52 for 0.5 and 1.0. of R52 over R37 in consuming H as increases. Around the ignition point, the reaction rates of R52 and R37 are of the same order, so one reaction can be favored over the other simply by changing the concentrations of the main reactants, O 2 and CH 4. Thus, at low (high O 2 concentration), the main branching reaction, R37, is favored over R52, promoting H production. In summary, for premixed ignition, the H pool critically depends on the rate of formation (R118) and decomposition (R56) of CH 3 O and the relative importance of R37 and R52 that have a promoting and retarding effect, respectively, on ignition. This is further shown in Fig. 3b, which depicts the spatial distribution of the rates of R37 and R52 for the 0.5 and 1.0 mixtures. While the two rates are nearly equal for 0.5, the R52 rate noticeably exceeds that of R37 for 1.0 simply because of the higher CH 4 concentration. The role of R52 in retarding ignition is further emphasized by the fact that both R37 and R52 peak at the same location, contrary to what has been observed in non-premixed systems [9]. The effects of symmetry, injection velocities, and particle injection temperature are shown in Fig. 4. Figure 4a depicts the dependence of (n p,inj ) ign on the reactant configuration at various s. For all configurations, ignition is facilitated, that is, there is a lower required (n p,inj ) ign,as decreases. For a given injected at ambient temperature and particles are introduced from the left burner, the gas issued from the left burner reaches its maximum temperature (T max ) at the GSP. The particles then cross the GSP and heat the gas issuing from the right burner. Fig. 3a also depicts the normalized (by n p,inj ) particle number density, which decreases as the particles approach the right burner, showing that radial particle transport is active throughout the domain [6]. Integrated species consumption path analyses were conducted at ignition for all three s. CH 4 is consumed by reactions R97 (70%, 66%, 63%), R11 (17%, 14%, 12%), and R52 (12%, 19%, 22%). The numbers within parentheses indicate the percentage contribution of each reaction in CH 4 consumption for 0.5, 1.0, and 1.5, respectively. CH 3 is consumed by R118 (34%, 36%, 36%), R156 (23%, 22%, 21%), and R154 (20%, 16%, 15%). CH 3 O is produced largely by R118 (over 80%) and consumed by R56 (68%, 68%, 70%) and R168 (31%, 30%, 28%). H radicals are largely produced by CH 3 O decomposition, R56 (74%, 71%, 68%), and at smaller amounts ( 10%) by R73, R157, and R165. They are consumed through reactions with the main reactants O 2 and CH 4, that is, R37 (48%, 34%, 26%) and R52 (41%, 56%, 64%). Note the increasing importance Fig. 4. Dependence of injection particle number density at ignition, (n p,inj ) ign, for premixed flames on (a) for twin flames, a single flame against air with particles seeded on the air side (air seeding), a single flame against air with particles seeded on the mixture side (mixture seeding), a single flame against O 2 with particles seeded on the mixture side (mixture seeding), all with u g,inj 100 cm/s, u p,inj 400 cm/s, and T p,inj 2000 K; (b) u p,inj for 1.0 twin flames, with u g,inj 100 cm/s and T p,inj 2000 K; (c) u g,inj for 1.0 twin flames, with u p,inj 400 cm/s and T p,inj 2000 K; (d) T p,inj for 1.0 twin flames, with u g,inj 100 cm/s and u p,inj 400 cm/s.

HOT PARTICLE IGNITION OF METHANE FLAMES 1609 and N 2 was studied and showed an ignition response similar to that for hot particles. The boundary temperature of the air jet required for ignition, (T b,exit ) ign, was 1332, 1362, and 1382 K for 0.5, 1.0, and 1.5, respectively, showing the inhibiting effect of increased. Non-Premixed Flame Ignition Fig. 5. Non-premixed ignition by particles of a CH 4 /N 2 jet counterflowing against an air jet, both injected at ambient temperatures, with u g,inj 100 cm/s (based on air density), u p,inj 400 cm/s, and T p,inj 2000 K. (a) Dependence of (n p,inj ) ign on the CH 4 mole fraction in the fuel jet, X CH4, with particles seeded on the air side (air seeding) and on the fuel side (fuel seeding). Spatial variations of (b) the gas-phase temperature for air seeding and fuel seeding, (c) the heat loss per particle for air seeding and fuel seeding, and (d) CH 3 O and H mass fractions for air seeding and fuel seeding. All results correspond to the ignition state of the 0.5 case. X CH4, eliminating the fuel on one of the two jets (single air) noticeably facilitates ignition, and an even more favorable effect is realized by eliminating N 2 from the inert jet (single O 2 ). Also, adding the particles on the air side is slightly favorable to adding them on the mixture side. The results of Fig. 4a were explained by analyzing the ignition kernel [e.g., Ref. 9] the region within which the radicals are developing, its center taken to correspond to the maximum CH 3 mass fraction. The local CH 4 /O 2 equivalence ratio ( local ), at the center the kernel is sensitive to the configuration and drops significantly when only O 2 is injected from the opposing jet. For example, for ignition at 1.0, local 0.978 for twin flame, 0.476 for single air, and 0.150 for single O 2 configurations. For single O 2, additional O 2 diffuses upstream into the reaction zone, reducing local, which promotes ignition by diminishing the inhibiting effect of CH 4 (R52). Figure 4b shows that ignition is facilitated as u p,inj increases for fixed n p,inj and T p,inj, which increases the particle mass flow and thus the supplied sensible energy. Fig. 4c shows that ignition is inhibited as u g,inj increases, which increases the local strain rate around the ignition kernel and augments the convective radical losses. Finally, Fig. 4d shows that increasing T p,inj facilitates ignition by increasing the sensible energy supplied by the particles. For contrast, the particle-free ignition of ambient temperature mixtures by impingement on heated air X CH4 Figure 5 depicts non-premixed ignition of CH 4 /N 2 counterflowing against air. Particles are injected either with the air (air seeding) or fuel (fuel seeding). In all cases, the particles are injected from the left of the figures, so that in air seeding, the air comes from the left and fuel from the right, while for fuel seeding the situation reverses. Both jets are injected at ambient temperatures. Figure 5a depicts the variation of (n p,inj ) ign with the CH 4 mole fraction in the fuel jet, X CH4, for both air and fuel seeding. In both cases, (n p,inj ) ign increases with X CH4, implying that, as with premixed systems, increasing the fuel concentration inhibits ignition. Furthermore, for X CH4 0.1, air seeding noticeably facilitates ignition compared to fuel seeding. The ignition response shown in Fig. 5a can be understood by following the radical path. CH 4 produces CH 3, that reacts with HO 2 (R118) to form CH 3 O. Subsequently, CH 3 O produces H and CH 2 O (R56), and the CH 2 O further reacts with CH 3 to produce HCO (R159). Thus, the key species for H production are HO 2,CH 3 O, and HCO. For both air and fuel seeding, integrated species consumption path analyses were performed at ignition for 0.5. CH 4 is consumed by R97 (55%, 62%), R52 (28%, 21%), and R11 (11%, 13%) to form CH 3. The numbers within the parentheses indicate the percentage contribution of each reaction in the consumption of the species for the fuel-seeding and air-seeding cases, respectively. Compared to premixed ignition, R52 consumes a greater portion of the fuel. CH 3 is subsequently consumed by R118 (28%, 32%), R156 (23%, 23%), R154 (15%, 16%), and R159 (18%, 14%). As for premixed systems, CH 3 O is produced largely by R118 (over 80%) and consumed by R56 (75%, 72%) and R168 (22%, 26%). The H radicals are produced by R56 (53%, 63%), R73 (14%, 10%), R165 (14%, 11%), and R157 (11%, 8%) and consumed by R52 (75%, 62%) and the ignition-promoting branching reaction R37 (18%, 30%). The differences between fuel and air seeding arise from the chemical sequence that produces CH 3 O, whose decomposition reaction R56 produces the majority of H radicals. The contribution of R56 to H production noticeably increases for air seeding. Key to that process is R118, where HO 2 combines with CH 3 to produce CH 3 O. The CH 3 O-producing R118 requires HO 2 radicals produced by reactions R166 (57%, 55%) and R168 (23%, 27%). R166 requires HCO that is mainly produced by R159 (66%, 56%) and is consumed by R166 (73%, 81%) and R165 (26%, 18%).

1610 LAMINAR FLAMES Ignition Fig. 6. Spatial variations of the rates of reactions R37 and R52 at the ignition state of a CH 4 /N 2 /air non-premixed configuration with X CH4 0.5. Fig. 7. Non-premixed ignition of CH 4 /N 2 counterflowing against air with u g,inj 100 cm/s (based on air density). (a) Dependence of ignition temperature on the CH 4 mole fraction in the fuel jet, X CH4, for cold-fuel/heated-air and heated-fuel/heated-air reactant configurations. (b) Spatial variation of the reactant concentrations and temperature for cold-fuel/heated-air configuration at the ignition state of the X CH4 0.5 case. (c) The corresponding heated-fuel/ heated-air case. (d) The corresponding particle ignition case. Both CH 3 O and HCO are consumed either by thermal decomposition (R56 and R165, respectively) to produce H or by reactions with O 2 (R168 and R166, respectively) to produce HO 2. Higher temperatures favor the thermal-decomposition reactions, while higher O 2 concentrations favor the HO 2 - producing reactions. Figure 5b for fuel seeding depicts that the gas phase reaches higher temperatures to the left of the GSP compared to air seeding and achieves a higher T max. (This is partially explained by the larger (n p,inj ) ign required for fuel seeding.) The fuel side has a larger thermal conductivity (k g ), and a lower heat capacity (c p,g ), than the air side. While lower c p,g results in greater temperature for the same amount of heat transfer, greater k g results in more effective heat transfer between the two phases [6]. Figure 5c depicts the interphasial heat transfer per particle, to account for the different (n p,inj ) ign s involved. As expected, the fuel seeding results in more heat transfer on the particle injection side of the GSP. The seeding orientation has little effect on the gas-phase temperatures to the right of the GSP since for air seeding, the particle number densities are lower, which compensates for the higher heat transfer per particle and the lower c p,g of the CH 4 /N 2 jet. Thus, for air seeding, the temperature of the air jet increases with n p,inj, but to a lesser degree than fuel seeding. At these lower temperatures, the HO 2 - producing R166 and R168 reactions are enhanced relative to the thermal decomposition reactions R165 and R56. Thus, the concentrations of CH 3 O and HCO peak closer to the left burner, that is, at larger O 2 concentrations. The augmented HO 2 production further accelerates the forward progress of R118 to produce CH 3 O, and subsequently H, more effectively than fuel seeding. This is illustrated in Fig. 5d, in which the mass fractions of H and CH 3 O for the two cases are shown as functions of the oxygen mass fraction, Y O2. Note the close correspondence between the CH 3 O and H profiles. Note also that for air seeding, the distributions are shifted into regions of higher Y O2 (lower Y CH4 ) that favor the consumption of H by R37 over R52. R56 and R165 result directly in H, while R166 and R168 result in HO 2 (which then produces H through R118). However, R166 and R168 also consume O 2, enhancing thus the involvement of a main reactant into the chain-branching sequence. At ignition, nearly 40% of O 2 is consumed by R166 ( 27%) and by R168 ( 13%), thus promoting the overall oxidation process. Figure 6 depicts the spatial distributions of the R37 and R52 reactions for both fuel-seeding and air seeding. Note that the R52 rate is noticeably reduced for air seeding, while the main branching R37 rate is unchanged, which promotes ignition by shifting the H consumption toward branching. Figure 7 depicts the effect of X CH4 on particle and hot-gas ignition. The latter includes ignition of cold fuel by heated air (as in Ref. [9]) and heated fuel by heated air. For cold fuel/heated air, Fig. 7a (as in Ref. [9]) illustrates that for low X CH 4 s, the boundary air ignition temperature, (T b,exit ) ign, first decreases with X CH4 and then mildly increases. Fig. 7b depicts that the ignition kernel is located about the highest temperature well within the heated-air region. As explained in Ref. [9], the high activation energy of

HOT PARTICLE IGNITION OF METHANE FLAMES 1611 the ignition process causes the kernel to form at the highest possible temperatures even though it lies deep in the air side, which fuel can only reach by diffusion. Thus, for small X CH 4, sufficient fuel cannot reach the kernel, requiring larger (T b,exit ) ign s. At larger X CH 4, the inhibiting effect of CH 4 on the kinetics again causes a rise in (T b,exit ) ign. (Examination of the kernel structures bears out these speculations. As expected, local is larger for the particle than for the cold-fuel/hot-air cases.) To distinguish between thermal and concentration effects, additional simulations were performed with both jets having the same boundary temperature (hot fuel/hot air). Both streams possessed sufficient thermal energy so that ignition is achieved solely by the mixing of reactants. The ignition response is shown in Fig. 7a and is similar to that of particle ignition. Fig. 7c depicts the ignition state for X CH4 0.5. Note that the ignition kernel is located at lower O 2 concentrations (i.e., closer to the GSP), resulting in a higher local than in Fig. 7b. Thus, there is no diffusion limitation on the fuel, and the observed monotonic rise in (T b,exit ) ign with X CH4 is solely a result of the inhibiting effect of CH 4 on the kinetics. Fig. 7d depicts the corresponding particle ignition structure. Note the similarity to Fig. 7c; because of the bell-shaped temperature distribution, the kernel lies closer to the GSP, so that ignition is never diffusion limited. Concluding Remarks Detailed numerical simulations were used to investigate the bulk ignition of premixed and non-premixed methane flames by hot inert particles. The effects of reactant composition, strain rate, particle seeding orientation, particle injection temperature, and velocity were assessed. Comparisons were made with pure gas-phase ignition. The H production that feeds the main branching reaction follows the CH 4 r CH 3 r CH 3 O r H path. Counterintuitively, particle ignition of premixed flames is promoted with decreasing equivalence ratio. Ignition is further promoted by counterflowing a mixture against an air jet and even more so against an O 2 jet, indicating a strong preference for O 2 -rich environments. Results indicate that at low equivalence ratios, ignition is promoted by favoring the main branching reaction (R37) over a propagation reaction (R52). Results for non-premixed particle ignition revealed that, similar to the premixed case, ignition is inhibited as the fuel concentration is increased, even at very low fuel concentrations. This differs from previous observations of ignition of cold fuel by hot air (and independently confirmed in this study), that at low fuel concentrations, ignition is promoted by adding fuel. This results from the different temperature profiles produced by hot particles and heated air; for cold-fuel/hot-air ignition, the ignition kernel lies very close to the heated-air boundary and is accessible to fuel only by diffusion. A heated-fuel/ heated-air configuration eliminated this effect of temperature asymmetry, yielding an ignition response similar to particle ignition. Finally, the results demonstrated the sensitivity of the ignition process to small configuration changes. In addition to the observed differences between particle and hot-air ignition, a major difference for nonpremixed systems was observed by changing the particle injection from the fuel side to the air side. The gas-phase thermal response is different for the two cases, which affects the chemical paths leading to H- radical production. This analysis assumes a small Peclet number so that ignition occurs in the bulk gas phase for which the particles act only as heat sources. But just before ignition, the bulk mixture will be at a high temperature and full of radicals, and ignition will most likely occur near particle surfaces, bootstrapped by the additional radicals generated there. However, following the arguments in the introduction, such particle ignition becomes less likely as the particle size is further reduced (as the Pe is reduced). The present analysis contains an implicit assumption of zero particle size and thus represents a limiting case. Previous studies in which a single large particle is sufficient to ignite the whole mixture on its own [e.g., Refs. 10,11] represent the opposite, high Pe limit. The middle ground, where many particles serve as preheaters until a single particle ignites the mixture, has yet to be studied. Acknowledgments NASA (grant NAG3-1877) supported this work under the technical supervision of Dr. Ming-Shin Wu of the Glenn Research Center. REFERENCES 1. Continillo, G., and Sirignano, W. A., Combust. Flame 81:325 340 (1990). 2. Chen, N. H., Rogg, B., and Bray, K. N. C., Proc. Combust. Inst. 24:1513 1521 (1992). 3. Chen, G., and Gomez, A., Proc. Combust. Inst. 24:1531 1539 (1992). 4. Gomez, A., and Rosner, D. E., Combust. Sci. Technol. 89:335 362 (1993). 5. Sung, C. J., Law, C. K., and Axelbaum, R. L., Combust. Sci. Technol. 99:119 132 (1994). 6. Egolfopoulos, F. N., and Campbell, C. S., Combust. Flame 117:206 226 (1999). 7. Andac, M. G., Egolfopoulos, F. N., Campbell, C. S., and Lauvergne, R., Proc. Combust. Inst. 28:2921 2929 (2000). 8. Andac, M. G., Egolfopoulos, F. N., and Campbell, C. S., Combust. Flame 129:179 191 (2002).

1612 LAMINAR FLAMES Ignition 9. Fotache, C. G., Kreutz, T. G., and Law, C. K., Combust. Flame 108:442 470 (1994). 10. Silver, R. S., Philos. Mag. S. 7. 23 (156):633 657 (1937). 11. Sharma, O. P., and Sirignano, W. A., Combust. Sci. Technol. 1:95 104 (1969). 12. Kee, R. J., Miller, J. A., Evans, G. H., and Dixon- Lewis, G., Proc. Combust. Inst. 22:1479 1494 (1988). 13. Kee, R. J., Warnatz, J., and Miller, J. A., Sandia report SAND83-8209. 14. Kee, R. J., Rupley, F. M., and Miller, J. A., Sandia report SAND89-8009. 15. Bowman, C. T., Frenklach, M., Gardiner, W. R., and Smith, G., The GRI 3.0 Chemical Kinetic Mechanism, 1999, www.me.berkeley.edu/gri_mech/. 16. Nishioka, M., Law, C. K., and Takeno, T., Combust. Flame 104:328 342 (1996). 17. Egolfopoulos, F. N., and Dimotakis, P. E., Proc. Combust. Inst. 27:641 648 (1998). COMMENTS Pierre Q. Gauthier, Rolls-Royce, Canada. Can this work be used to obtain information about autoignition in lean premixed CH 4 -air gas turbine combustors? The results showing the role of H radicals in CH 4 ignition in lean CH 4 air mixtures are very interesting and should be followed up for autoignition problems. Author s Reply. The present investigation focuses on the mechanisms of ignition in the presence of hot inert particles. Indeed much was learned about the various thermochemical mechanisms that control the ignition process. We are currently performing an independent study on issues related to the autoignition of practical fuels ranging from CH 4 to iso-octane.