Experiments on standing and traveling edge flames around flame holes

Transcription

1 Proceedings of the Combustion Institute 30 (2005) Proceedings of the Combustion Institute Experiments on standing and traveling edge flames around flame holes Giuliano Amantini a, Jonathan H. Frank b, Alessandro Gomez a, * a Department of Mechanical Engineering, Yale Center for Combustion Studies, Yale University, P.O. Box , New Haven, CT 06520, USA b Combustion Research Facility, Sandia National Laboratories, Livermore, CA 94551, USA Abstract The behavior of edge flames surrounding an axisymmetric hole in CH 4 /O 2 /N 2 laminar counterflow diffusion flames is experimentally studied with appropriate spatial and temporal resolutions, using a combination of CO PLIF, OH PLIF, CO + OH reaction-rate imaging, and PIV. The experimental approach is able to capture the standing edge flame structure, identifying the trailing diffusion flame and lean premixed flame unequivocally, and the rich premixed flame less conclusively. By perturbing a steadily burning flame locally with two counter-propagating toroidal vortices injected simultaneously from opposite sides of the flame, a transient behavior is observed in detail. It encompasses the complete phenomenology from the onset of local extinction on the centerline to the opening of a hole in the flame and its eventual healing. Even in the extinction mode, which is the inevitable regime at the beginning of a sufficiently disruptive flame/vortex interaction, the edge flame does not exhibit a negative propagation velocity but appears to be initially advected by the mean radial flow. It then rapidly undergoes a transition from the extinction mode to the ignition mode, developing a positive propagation velocity with respect to the unburned mixture, but with values lower than the planar premixed flame speed corresponding to the composition of the mixture in the mixing layer. The propagation velocity eventually reaches a value of the order of the planar premixed flame speed and maintains such a value during the rest of the interaction, until the extinction hole is fully healed. Ó 2004 The Combustion Institute. Published by Elsevier Inc. All rights reserved. Keywords: Edge flame; Counterflow diffusion flames; Vortex flame interaction; Flame hole 1. Introduction Edge flames have recently attracted much attention because of their relevance to diffusion flame stabilization local extinction/ignition phenomena in highly strained turbulent flames. The subject has received significant attention at * Corresponding author. Fax: address: alessandro.gomez@yale.edu (A. Gomez). all levels, analytical, experimental, and computational, much of which was reviewed in [1] and, for space limitations, will not be repeated here. In the ignition mode, an edge flame typically exhibits a tribrachial structure, with a diffusion flame connected at a triple point to two premixed flames, one lean flame facing the oxidizer side, and one rich flame facing the fuel side, and a strong maximum of the reaction rate at the triple point [1]. Such flames are referred to as triple flames and are characterized by positive laminar flame speeds, that is, they propagate towards the /$ - see front matter Ó 2004 The Combustion Institute. Published by Elsevier Inc. All rights reserved. doi: /j.proci

2 314 G. Amantini et al. / Proceedings of the Combustion Institute 30 (2005) unburned mixture. The pioneering experiments in [2] were revealing since they showed that the velocity at the triple point was close to the ordinary laminar flame speed of a stoichiometric mixture, but significantly lower, by a factor of 2 or more, than the velocity further upstream because of flow divergence. As a result, a triple flame has intrinsic stabilization properties, since it appears to propagate at a speed substantially higher than the ordinary laminar flame speed with respect to the unburned mixture far upstream [3]. Under conditions of moderately large strain rate, the tribrachial structure may collapse into a more compact configuration, caused by the shedding of one of the premixed flame whiskers, as reported in traveling ignition edge flames in [4]. In the extinction (or failure) mode, the edge flames present a simple edge, that is, without any tribrachial structure. It has no rate maximum near the edge and appears to be receding away from the unburned mixture, as essentially a twodimensional diffusion flame [1]. In fact, a number of both analytical and computational studies suggested the existence of a significantly negative edge flame propagation velocity, that is, with the edge moving away from the unreacted mixture, with a velocity several-fold the value of the laminar flame speed [5 10]. To the best of our knowledge, measurements in this regime have yet to be reported, possibly because of the difficulties in capturing an abrupt and inherently unsteady event. The present study examines a counterflow diffusion flame that is perturbed by a pair of counterpropagating toroidal vortices, which are simultaneously injected from the fuel and the oxidizer sides of the flame. The aim is to disrupt the flame locally, yielding only local extinction, thereby allowing for repeated observations of the phenomenology and the accumulation of the necessary statistics. The arrangement also has a direct tie, via the flamelet approach, to multiplyconnected turbulent non-premixed flames, in which local fluctuations of the velocity field may induce sufficiently large scalar dissipation rates to bring about local extinction in the way of flame holes. The existence of a hole in counterflow flames was first reported in [11], without any reference to edge flames, which were an unknown concept at the time. In previous work, we reported on the edge flame propagation velocity in the ignition mode [4]. In that study, the edge flame propagation velocity in the ignition mode was measured mostly for standing triple flames at large Damköhler number (Da), and only a few data points were reported for propagating ignition edge flames at moderate Da. This contribution aims at examining details of the behavior ensuing the reversible creation of a flame hole with high temporal and spatial resolutions. 2. Experimental methods Experimental diagnostic methods included CO/OH reaction-rate imaging and particle image velocimetry (PIV), which were, respectively, performed in the Advanced Imaging and the Turbulent Combustion Laboratories at the Combustion Research Facility of Sandia National Laboratory Reaction-rate imaging Simultaneous measurements of single-photon OH LIF and two-photon CO LIF are used to determine a quantity that is proportional to the forward reaction rate of CO + OH fi CO 2 +H. This reaction is the dominant pathway for CO 2 production in CH 4 /air flames. A schematic diagram of the experimental system, consisting of two lasers, two cameras, and an axisymmetric counterflow burner, is shown in Fig. 1. The reaction-rate imaging technique is described in detail elsewhere [12,13], and only a brief overview is given here. The forward reaction rate, RR, is given by RR = k(t)[co][oh], where k(t) is the forward rate constant and T is the temperature. The product of the LIF signals from CO and OH can be approximated by f CO (T)f OH (T) [CO][OH], where the temperature dependence of the LIF signals is represented by f(t). The pump/detection scheme determines the temperature dependence of the LIF signals and can be selected such that f OH (T)f CO (T) µ k(t). When this relationship is achieved, the pixel-by-pixel product of the OH LIF and CO LIF signals is proportional to the reaction rate. For OH LIF, the frequency-doubled output from a Nd:YAG-pumped dye laser was tuned near 285 nm to pump the Q 1 (12) transition of the A X(1,0) band. An intensified CCD camera ( pixels) with an f/1.8 Cerco quartz camera lens was used to record the OH LIF signal with a projected pixel size of 93.6 lm 93.6 lm. The image intensifier was gated for 400 ns, bracketing the dye laser pulse. The OH LIF images were corrected for spatial variations in the laser sheet using acetone LIF to record the beam profile. Two-photon CO LIF was excited by pumping overlapped transitions in the B X(0,0) bandhead of the Hopfield Birge system of CO using the frequency-doubled output from a Nd:YAG-pumped optical parametric oscillator (OPO) (14 mj) near nm. The laser was tuned to maximize the CO LIF signal in a laminar non-premixed methane counterflow flame. Sheet forming optics were used to form an 11.5-mm high laser sheet. The average laser beam profile was measured using CO LIF from a mixture of CO in N 2 (0.1% CO by volume). The image intensifier was gated for 400 ns, bracketing the OPO laser pulse. The CO fluorescence was imaged onto an intensified

3 G. Amantini et al. / Proceedings of the Combustion Institute 30 (2005) Fig. 1. Experimental setup for simultaneous imaging of OH LIF and two-photon CO LIF. Insets: schematic of the axisymmetric counterflow burner and of the interaction between vortices and flame. CCD camera ( pixels) with a f/1.2 camera lens and an interference filter (Dk = 484 nm and Dk = 10 nm), which transmitted fluorescence from the B A(0,1) transition at nm and blocked out-of-band interference. The projected pixel size was 93.6lm 93.6 lm. The repeatable flow-flame interactions considered in this study permitted phase averaging over 10 shots to improve the signal-to-noise ratio of the two-photon CO LIF, a relatively weak process compared to single-photon LIF. Timing of the two laser pulses was controlled with digital delay generators. The OPO laser fired 600 ns after the dye laser was eliminating the possibility of cross-talk between the two diagnostics. The CO/OH LIF measurements were essentially instantaneous because the elapsed time for a single measurement was orders of magnitude less than the flow timescales. The reaction-rate imaging technique requires careful matching of the OH and CO LIF images. A precise image matching technique was used to obtain accurate registration between the two CCD cameras [12]. Images were matched with an eight parameter bilinear geometric warping algorithm, and the residual matching error was in the subpixel range PIV measurements Velocity field measurements were performed in the reactant regions of the unsteady flames using particle image velocimetry (PIV). The fuel, oxidizer, and vortex flows were seeded with oil droplets, which were fully evaporated at a temperature of about 570 K. The average droplet diameter was estimated at 1 lm, resulting in Stokes numbers sufficiently small for inertia effects to be negligible. The PIV system (TSI) consisted of two pulsed Nd:YAG lasers and a pixel CCD camera. The laser beams were formed into overlapping sheets that intersected the burner axis. The lasers were sequentially pulsed with a 70-ls time delay, and the particle scattering from each laser pulse was imaged onto a separate frame of the camera. Velocity vectors were determined using a cross-correlation analysis with pixel (0.675 mm mm) interrogation regions separated by 16 pixels. Uncertainties are estimated at 3 cm/s Burner Axisymmetric counterflow diffusion flames were established in the burner shown in the inset of Fig. 1. Each side of the burner consisted of a contoured 12.5 mm dia. nozzle, which produced a uniform velocity profile at the nozzle exit. The nozzles were surrounded by 76.2 mm dia. flanges to facilitate the stabilization of triple flames and to provide well-specified boundary conditions for future computational studies. The flanges were water cooled to maintain a constant wall temperature. The bottom portion of the burner was surrounded by a coflowing nitrogen shroud to isolate the flame from external disturbances. The nozzle separation distance was kept constant at 13 mm.

4 316 G. Amantini et al. / Proceedings of the Combustion Institute 30 (2005) In one series of experiments, a CH 4 and N 2 mixture flowed from the top nozzle, and a mixture of O 2 and N 2 flowed from the bottom nozzle. In a second series of experiments, the fuel and oxidizer sides were switched. The burner was designed to study both standing and propagating edge flames. To study propagating edge flames, a 2.1 mm tube was inserted along the axis of each nozzle to allow the simultaneous injection of a toroidal vortex from each side of the burner as shown in the inset of Fig. 1. The exit of the tubes was located 20 mm upstream of the nozzle exits. Each tube was connected to a plenum, which was supplied with the same gas composition as the main nozzle and was pulsed by a loudspeaker. Each loudspeaker was driven with the same waveform, which generated a single isolated vortex and minimized the suction of gases into the tube following the vortex generation. The rise time of the waveform was 1 ms with a 27 ms decay. The speakers were pulsed at 1 Hz to ensure that a steady-state flow condition was reestablished prior to generating a vortex. The speaker pulses were phase locked to the laser pulses by digital delay generators. The vortex flame interactions were highly repeatable, which allowed phase-averaging of the LIF measurements. 3. Results and discussion 3.1. Steady edge flames Fluorescence measurements were performed first on a standing edge flame to demonstrate the ability of the combined CO-OH PLIF to track key flame features. The oxidizer and fuel streams issued from the top and bottom nozzles, respectively. The inlet flows contained 18% methane, by mole, in the fuel stream, and 37% oxygen in the oxidizer stream, and are characterized by a global strain rate of 88 s 1. The CO, OH, LIF and reaction rate images are shown in Fig. 2. Only half of the domain is shown, with the right edge at the axis of symmetry of the burner and the left edge at the domain outlet. If we examine the CO-LIF image, inspecting it from the right edge towards the left, we find the first evidence of CO in two recirculation zones, a marked one near the top of the domain and a fainter one at the bottom, both in light purple color. The existence of such zones is expected and results from the sharp corners at the nozzle outlets. The edge flame is stabilized immediately downstream of the two recirculation zones, with a marked lean premixed branch and another possible rich premixed branch at the bottom. The latter seems to separate the lower recirculation from a region of very intense fluorescence near the outlet of the domain due to the trailing diffusion flame. In the outermost part of the domain, the flow is swept up by the nitrogen shroud. The OH image in Fig. 2 clearly shows the trailing diffusion flame but gives no evidence of the premixed branches. Finally, the lower strip of Fig. 2 is an image of the CO oxidation reaction rate. One can observe a small protrusion on the top, facing the oxidizer side. The fuzzy region at the bottom without a well-defined structure is likely to be a consequence of the noisy OH signal in that region. Previous numerical simulations of an unstrained methanol edge flame show a distinct peak in this reaction rate near the triple point [14]. In contrast, the present measurements show a gradual decay of the reaction rate from the edge into the trailing diffusion flame. We conclude that the application of the combined CO and OH PLIF has unequivocally identified the trailing diffusion flame and the lean premixed branch, and less conclusively identified the rich premixed branch. Figure 3 shows both the raw image from the PIV measurements and the processed PIV vectors under the same conditions as in Fig. 2. Since the PIV measurements were performed in a different laboratory than the PLIF measurements, minor variations in the reproducibility of the flow-flame interaction from day to day and lab to lab have prompted us to extract information on front propagation and gas velocity from the same set of velocity data. To that end, the position of the edge flame front is tracked through the droplet evaporation front, whereas the velocity field in the domain upstream is obtained with standard PIV processing. Fig. 3 shows the stagnation point flow with the two recirculation zones near the nozzle outlets and a curve on the left that bounds a region devoid of any particles. Since the oil droplets should be fully evaporated at a typical temperature of about 570 K, we can interpret the curved profile on the left as an approximate isotherm at that temperature. Such an isotherm is expected to be shifted with respect to the leading edge of the flame by an approximately constant characteristic length, that is of the order of the convectivediffusive zone of the premixed flame Unsteady edge flames Figure 4 shows the temporal sequence of the interaction of two vortices, which are sufficiently intense to temporarily induce local extinction in the central part of the flame. For such a flame, the fuel and oxygen mole fractions in the feed streams are 24.2% and 36.7%, respectively, with a baseline global strain rate of 80.5 s 1 and a peak strain rate of s 1, both measured by hot wire anemometry in the non-reacting flow. Note that the fuel is issued from the top nozzle and oxidizer from the bottom, unlike conditions in Figs. 2 and 3. The time evolution of the event as captured by

5 G. Amantini et al. / Proceedings of the Combustion Institute 30 (2005) Fig. 3. PIV vectors superposed to oil particles for the steady edge flame. Fig. 2. Images of CO LIF (top), OH LIF (middle), and reaction rate (bottom), in a standing edge flame (image dimensions: mm 2 ). PF, premixed flame; DT, diffusion tail; UR, upper recirculation; LR, lower recirculation. CO PLIF, OH PLIF, and the reaction rate of CO + OH is shown in the left, middle, and right columns, respectively. The first set of pictures corresponds to the undisturbed flame. The CO image Fig. 4. Unsteady edge flames: time evolution of the formation of a hole in a flame and its healing. CO LIF (left), OH LIF (middle), and reaction-rate (right) images at different times after the generation of the vortex (image dimensions: mm 2 ).

6 318 G. Amantini et al. / Proceedings of the Combustion Institute 30 (2005) has a complex morphology resulting from the merging of the mixing layer in the middle of the domain and two large recirculation regions anchored at the exit of the top nozzle. The maximum value of CO LIF signal is reached near the flanges, possibly because their temperature is kept low by water cooling. The CO detection scheme is biased towards low temperatures [12,13]. The OH and reaction-rate layers have a region of relatively uniform thickness centered on the domain centerline and facing the burner mouths. As discussed in [4], the square of the thickness of the mixing layer is related through the diffusivity to the scalar dissipation rate. The uniform thickness in the middle is a mere consequence of the relatively uniform strain rate and scalar dissipation rate. At larger radial distances, the thickness grows, and the local scalar dissipation rate decreases. Seven milliseconds into the interaction, we notice an indentation in the CO image, corresponding to the arrival of the vortex from the fuel side. Near the flame centerline one can notice the thinning of the CO and OH layers. This corresponds to an increase in the local scalar dissipation and forecasts imminent local extinction. In the subsequent set of pictures, both the OH- LIF signal and the reaction rate decrease until a local extinction hole develops at t = 8 ms. The CO flame image remains simply connected, with a faint thread separating the upper vortex from the lower one, which has reached the reaction layer. The nearly simultaneous arrival of the synchronized vortices keeps the mixing layer locked in the middle of the domain and enables the monitoring of the phenomenon with ease. Clearly, at this time the reaction has been quenched on the centerline, and the CO layer merely lingers a bit longer while it gets stretched further, with CO being removed radially, without any new chemical source along the centerline. As one can judge from the reaction-rate image, the edge flame is already formed, manifesting itself as a hole that increases in width and recedes from the flame axis. For subsequent times, a hole is present in all three images. While the propagation speed in the laboratory frame can be measured via the displacement of the rather sharp trailing edges in the reaction-rate images, the CO images are, in particular, revealing in the next two time sets. The CO edge, in fact, develops an indentation that can be attributed to vortex roll-up. Such an indentation ultimately leads to a bifurcation of the edge that persists through t = 14.5 ms. It possibly suggests that, in this time interval, the trailing edge of the reacting layer is merely advected radially at approximately the same speed as the stretched vortices. We will seek further confirmation of this interpretation in the PIV data which are discussed below. Subsequently, the hole reaches its largest radial dimension of approximately 5.2 mm at t = 16.5 ms and remains in that position as a standing ignition edge-flame for as long as 11 ms. By this time, the reacting layer edge is no longer trailing but is leading into the unreacted mixture as an ignition front. At this point, the velocity of the gas is expected to be equal and opposite to the positive propagation speed of the edge flame. This is effectively a quasi-standing edge flame. Afterwards, the edge begins to move inward, as a propagating ignition front. In the second half of the interaction set, the CO images develop a pronounced whisker on the lower side, facing the oxidizer inlet, which can be attributed to the lean premixed branch, and is similar to the standing edge flame displayed in Fig. 2. No similar feature is observed on the rich side nor in the OH, and reaction-rate images on either side of the trailing diffusion flame. The presence of this single whisker is consistent with the features detected by HCHO PLIF under similar circumstances in [4]. The ignition phase terminates at t = 45.5 ms, when all images become simply connected. In the ensuing time, the flame ultimately relaxes to its unperturbed condition, and the entire cycle repeats itself. Measurements in the extinction and ignition modes of the highly strained unsteady edge flames do not show a distinct peak in the reaction rate near the edge. While this behavior is expected for the extinction mode, it is somewhat surprising for the ignition mode [14] Velocity measurements As a result of space limitations, only velocity measurements pertaining to the unsteady edge flame will be discussed. It is of interest to determine the propagation velocity of the edge flame with respect to the unreacted gas. The precise demarcation of the reaction-rate edge suggests that by analyzing the position of this edge in successive images, obtained at well-defined time intervals, as in Fig. 5, one can deduce its velocity with respect to the laboratory frame of reference. However, to avoid potential reproducibility difficulties associated with registering PLIF images with PIV data obtained at different times and in a different laboratory, we tracked the edge flame front using the demarcation line between droplet-laden region and droplet-free region from the raw unprocessed PIV images of the scattering seeded droplets. Figure 5 shows such data, normalized with respect to a stoichiometric laminar premixed flame speed, properly diluted to reflect the composition of the mixing layer. Negative velocities are associated with outward propagation and positive velocities are associated with inward propagation. The reason for this sign selection in the ordinate is that inward propagating edges, i.e., edges moving towards the unburned mixture, are unambiguously propagation fronts, with which are associated positive speeds. This conforms to the convention of positive laminar flame

7 G. Amantini et al. / Proceedings of the Combustion Institute 30 (2005) Fig. 5. Edge flame propagation velocity, normalized with respect to the planar premixed flame speed S L = 60 cm/s, under the conditions of Fig. 4: velocity in the laboratory frame (triangles); velocity of the unreacted flow upstream of the edge flame (squares); and relative velocity of the edge flame with respect to the unburned mixture (diamonds). speeds for planar flames. Conversely, outward propagating edges, i.e., edges moving away from the unburned mixture, begin initially as extinction fronts, with which a negative speed is typically associated. At some point in their travel, they turn into ignition fronts that may still move outward if they are swept out by a relatively strong convective flow. Eventually, they reach a quasi-steady location, and reverse their path, as the flow perturbation decays. The gas velocity in the vicinity of the edge flame was measured with PIV, and the results are plotted in Fig. 5. This profile is almost a mirror image of the previous one. Adding algebraically the two values, with the appropriate sign convention, one obtains the relative velocity of the edge flame with respect to the unburned gas mixture. The third curve in the plot represents such a difference. The data suggest that there is an initial transient in which the extinction edge flame is essentially advected by the mean flow, as first suggested in the discussion concerning Fig. 5 for times between 8 and 14.5 ms. At later times, the flame turns into a propagating edge with a characteristic velocity reaching a value approximately equal to the corresponding planar premixed flame. Contrary to the conclusions of numerous studies conjecturing the large negative propagation velocity of extinction fronts [5 10], no such evidence was found. Data at earlier times, which are not reported in the figure, also suggest that the edge flame in the extinction mode is merely advected by the mean flow, as originally reported in a computational study [15]. The phenomenological interpretation is as follows: in the vicinity of the mixing layer stagnation point, the tangential (radial) velocity in the mixing layer is of the same order as the transversal velocity, normal to the layer. Under conditions of local extinction, this transversal velocity is of the order of the laminar flame speed. Farther away from the stagnation point, the tangential velocity increases dramatically and overwhelms the front propagation velocity, sweeping it along. Eventually, as the scalar dissipation rate decreases in the direction tangential to the mixing layer, the Damköhler number increases, and the extinction edge turns into a propagation edge flame. If we consider the stochastic generation of holes in turbulent flames as the result of local increases in the flame scalar dissipation rate due to velocity fluctuations, the present results are the monochromatic equivalent under laminar conditions, and, as such, are directly relevant to partially disrupted turbulent flames. 4. Conclusions The transient behavior of an edge flame undergoing a transition from extinction mode to traveling propagation mode, including a quasi-steady standing propagation mode, is triggered periodically by impinging counter-propagating toroidal vortices on a counterflow laminar diffusion flame. The behavior is monitored in unprecedented detail using a combination of CO PLIF, OH PLIF, CO + OH reaction-rate imaging, and PIV. No evidence of a negative propagation velocity is found. The extinction edge flame appears to be initially advected by the mean flow. It rapidly turns into a propagating edge flame, whose propagation velocity is of the order of the planar premixed flame speed. This velocity is maintained until the flame hole is completely healed. Acknowledgments This research is supported by NSF, Grant #CTS (Dr. Farley Fisher, Contract Monitor) and the US Department of Energy, Office of Basic Energy Sciences, Division of Chemical Sciences, Geosciences, and Biosciences. The technical assistance of N. Bernardo (Yale University) and R. Sigurdsson (Sandia National Laboratories) in the construction of the hardware and in setting up the optical diagnostic system, respectively, and technical discussions with Professor Amable Linan (Yale University) are gratefully acknowledged. The authors thank Drs. T.C. Williams, R.W. Schefer, and Robert Harmon for use of the PIV facility. References [1] J. Buckmaster, Prog. Energy Combust. Sci. 28 (2002) [2] H. Phillips, Proc. Combust. Inst. 10 (1965) [3] G.R. Ruetsch, L. Vervish, A. Liñán, Phys. Fluids 7 (6) (1995) [4] V.S. Santoro, A. Liñán, A. Gomez, Proc. Combust. Inst. 28 (2000)

8 320 G. Amantini et al. / Proceedings of the Combustion Institute 30 (2005) [5] J.W. Dold, D. Hartley, D. Green, in: P.C. Fife, A. Liñán, F.A. Williams (Eds.), Dynamical Issues in Combustion Theory, IMA Volumes in Mathematics and its Application, vol. 35, Springer Verlag, New York, 1991, pp [6] P.N. Kioni, B. Rogg, K.N.C. Bray, A. Liñán, Combust. Flame 95 (1993) [7] H.G. Im, J.H. Chen, Combust. Flame 119 (1999) [8] T. Takagi, I. Nakajima, S. Kinoshita, Proc. Combust. Inst. 29 (2002) [9] J. Ray, H.N. Najm, R.B. Milne, K.D. Devine, S. Kempka, Proc. Combust. Inst. 28 (2000) [10] H.G. Im, J.H. Chen, Combust. Flame 126 (2001) [11] A.E. Potter, J.N. Butler, ARS J. 29 (1959) [12] J.H. Frank, S.A. Kaiser, M.B. Long, Proc. Combust. Inst. 29 (2002) [13] J.E. Rehm, P.H. Paul, Proc. Combust. Inst. 28 (2000) [14] T. Echekki, J.H. Chen, Comb. Flame 114 (1998) [15] C.E. Frouzakis, A.G. Tomboulides, J. Lee, K. Boulouchos, Proc. Combust. Inst. 29 (2000) Comments Hong G. Im, University of Michigan, USA. In our experience [1], the edge flame speed depends significantly on the location at which it is measured. Have you attempted to investigate the sensitivity of the edge flame speed to the location along the stoichiometric line? Would you expect consistent results if different locations were chosen? Reference [1] C.S. Yoo, H.G. Im, Proc. Combust. Inst. 30 (2005) Reply. To measure consistently the edge flame propagation speed we followed this procedure: first, we considered the region where gaseous velocity measurements and flame speed in the laboratory reference frame were simultaneously available; second, we identified a point at the intersection between the droplet vaporization curve that tracked the flame motion and the estimated stoichiometric line. The rational for this selection is the following: since the oil droplets should be fully evaporated at a temperature of 572 K, we can interpret the curved profile on the left of Fig. 3 as an approximate isotherm at that temperature. Such an isotherm is expected to be shifted with respect to the leading edge of the flame by an approximately constant length, that is a small fraction of the convective diffusive zone of a premixed flame. Since the flame speed for planar flame propagation that is used in Fig. 5 for normalization purposes is evaluated with respect to cold conditions, it is also appropriate to choose a relatively cold location for the computation of edge flame propagation. We cannot attempt a sensitivity analysis of the edge flame speed to the location along the stoichiometric line, because our PIV experiments allow us to measure velocities only in the cold region. Computational modeling of these flames, that is currently underway, will shed light on the sensitivity of triple point propagation measurements to the transverse location with respect to the flame. d Suresh Aggarwal, University of Illinois at Chicago, USA. There has been comparison between measurements and simulations for 2-D flames. For example, Rolon and Katta and Puri and Aggarwal groups have reported comparison between counterflow and coflow flame configurations. Using reaction between CO and OH, it is very difficult to identify rich premixed and/or lean premixed reaction zones in edge flames. This reaction is essentially dominant in the non-premixed reaction zone. Reply. We are aware of some comparisons between experiments and calculations on triple flames in the literature. However, such comparisons are qualitative in nature and limit themselves mostly to the general morphology of these structures. Our own experience on the computation of these flames, which will be reported in the near future, shows high sensitivity of the position of triple flames to boundary conditions. As a result, any quantitative comparison between experiments and calculations will require a level of detail in either the experiments and/or the computation that, to the best of our knowledge, has not appeared yet in the literature. As to the suitability of using the CO + OH reaction imaging to identify the premixed branches, our data show that such an image is able to identify the lean branch of a standing edge flames, which is consistent with the computational result in [14 in paper]. d Jacqueline H. Chen, Sandia National Laboratories, USA. (a) Have you examined the correlation of your measured triple flame speed with the density ratio (measured or computed) across the flame? (b) How sensitive is the triple point speed measurement to the transverse location, i.e., does your choice of upstream location where the oil droplets vanish bias the results? In methanol/air triple flames [14 in the paper] the point of maximum heat release is close to the triple point. Your product image of CO and OH is proportional to heat release. Have you considered tracking the speed at this location and how would it compare with your results upstream?

9 G. Amantini et al. / Proceedings of the Combustion Institute 30 (2005) Reply. (a) We addressed this issue in some earlier work at Yale in which we reported on the edge flame propagation as a function of an estimated Damkhöler number. For large Damkhöler numbers, indeed it was found that the edge flame propagation velocity, normalized with respect to the flame speed for a planar stoichiometric premixed flame, approaches the square root of the density ratio [4 in paper]. (b) As for the location of the triple point, the response we provided to the first comment applies.