Turbulent partially premixed combustion is broad-ly utilized in practical applications, e.g., in gas tur-bines for power generation. Due to the expected de-pletion of fossil fuels as well as the issue of globalwarming, there is a strong demand to improve thecombustion efficiency and to reduce the pollutantemission in modern combustion devices. For this pur-pose a detailed understanding of the fundamentalprocesses taking place in partially premixed flames isrequired. It becomes particularly complicated whereturbulent combustion is concerned, owing to the com-plex interactions between the reaction chemistry andthe rapid heat and mass transfer. For the analysisof combustion processes, two prominent approachesare commonly applied, numerical simulations basedon either physically accurate equations or simplifiedmodels, and experimental methods. Often both ap-proaches are employed to complement each other,e.g., when numerical simulations are validated by ex-perimental results [1].
In the field of combustion diagnostics nonintru-sive optical methods are of the utmost importancesince they provide data from the unaffected systemwith high temporal and spatial resolution [2]. Forinstance, the flow field can be visualized by parti-cle image velocimetry (PIV) [3,4]; laser Rayleighscattering can provide temperature information [5,6];the laser-induced incandescence (LII) can be used tomeasure the distribution of nascent soot [7,8]. An-other crucial measurement is the concentration andlocal distribution of chemical species. While spon-taneous Raman scattering and coherent anti-StokesRaman scattering are normally used for the detec-tion of major species [9–16], resonant excitation isrequired for the detection of minor species such ascombustion intermediates. Among the resonant tech-niques, laser-induced fluorescence (LIF) is one of themost established ones [17,18]. In particular, planarLIF (PLIF) is often used to obtain two-dimensionalinformation such as the spatial distribution of differ-ent trace species in flames. In general, quite a fewintermediates have been investigated with PLIF in thepast. However, in turbulent flames the visualization ofone single species usually cannot provide sufficientinformation. Examples are radicals such as CH andHCO, which exist only in a very narrow region ofthe reaction zone due to their short lifetimes. FromPLIF images showing the spatial distribution of oneof these species, the flame propagation cannot be ex-
tracted. In these images it is hard to distinguish be-tween the burnt and unburnt regimes on both sides ofthe strongly wrinkled thin layer observed. Therefore,a lot of effort has been put into the development ofexperimental methods enabling simultaneous imag-ing of two or even more species, or one intermedi-ate and the temperature or velocity field. Ayoola etal. [19] and Pfadler et al. [20] have measured localheat release by simultaneous OH and formaldehydePLIF in premixed flames. Medwell et al. [21] havevisualized the temperature field using laser Rayleighscattering along with OH and formaldehyde distrib-utions in nonpremixed jet flames. CH PLIF and PIVhave been used to investigate local flame structuresin nonpremixed flames [22–24]. Petersson and co-workers have characterized a premixed swirl-flameusing simultaneous OH/PIV and OH/Rayleigh meth-ods [25,26]. However, there is one combination ofradicals that has been chosen more often than others,namely CH and OH.
The CH radical is an interesting species to be uti-lized as a flame front marker in spite of its low numberdensity [27]. This is because of the fact that it existsonly in a very narrow layer of the reaction zone takingpart in the decomposition process of the fuel mole-cules in hydrocarbon flames. Furthermore, it playsa key role in the formation of nitric oxide, which isan important pollutant in combustion [28]. Therefore,various techniques have been developed for the de-tection of CH, e.g., LIF [29,30], cavity ring-down ab-sorption spectroscopy [29], two-color resonant four-wave mixing [31], degenerate four-wave mixing [32],tunable diode laser absorption [33], and polarizationspectroscopy [34]. However, CH PLIF applicationshave been limited to stoichiometric, rich, or non-premixed flame conditions in most studies, since thenumber density of CH is strongly dependent on thelocal equivalence ratio [35]. Recently Li et al. haveshown that the use of a pulsed alexandrite laser en-abling multiple rotational line excitation (multiplex)facilitates CH PLIF imaging with a high signal-to-noise ratio [35] over a wide range of stoichiometricratios. This approach enables single-shot applicationsin turbulent lean premixed flames. In contrast, the OHradical exists in a much larger flame region, since itis only consumed in three-body collision reactionswithin the recombination zone [36]. Therefore it iswidely utilized as a reaction zone marker in com-bustion research (see, e.g., Ref. [37] and referencestherein). A particularly interesting feature of single-shot OH imaging is that it can provide information
804 J. Kiefer et al. / Combustion and Flame 154 (2008) 802–818
about the reaction progress variable and the flame sur-face density and, hence, the correlation of the twoparameters [38–41]. In recent years, CH and OH havebeen visualized simultaneously in a number of appli-cations, e.g., revealing local extinction in turbulent jetdiffusion flames [42]. Moreover, local structures suchas thickening of the flame front have also been thesubject of interest in nonpremixed flames at variousturbulence intensities [43,44]. In premixed combus-tion the technique has been used to investigate theresponse of methane/air flames to transient strain andvariations in the local stoichiometry [45]. PIV hasbeen applied additionally to CH/OH PLIF in non-premixed [24,46] as well as premixed [47] flamesin order to investigate the interaction of chemistryand turbulence. Furthermore, CH and OH PLIF, alongwith Raman spectroscopy and Rayleigh scattering,has been employed to study laminar partially pre-mixed flames [48,49].
The main objective of this study is to apply state-of-the-art two-dimensional laser diagnostic tech-niques, i.e., simultaneous CH and OH PLIF meth-ods, to investigate the microscale local structures ofa partially premixed jet flame under different flowconditions, ranging from a typical laminar flame toa highly turbulent flame with local flame extinctionand reignition. To generate highly turbulent flames,a high-speed air jet in the center of the burner isinstalled, which can provide a flow speed of up to120 m/s. The flame is stabilized in proximity tothe burner by a coaxial low speed flame. The influ-ence of the jet exit Reynolds number on local flamestructures is systematically investigated, providing in-sightful understanding of the flames for future modeldevelopment. To provide quantitative statistical in-formation about the flame structures, the single-shotCH and OH images are utilized to determine localflame surface densities. Moreover, mean and standarddeviation maps are generated to deliver qualitative in-formation about the local Reynolds-averaged radicalconcentration distributions. These data may be usedfor validation of numerical simulations.
2. Experimental
2.1. Burner assembly
A jet burner is employed in this experiment to gen-erate turbulent partially premixed methane/air flameswith variable turbulence intensities. Fig. 1 shows theburner assembly consisting of two coaxial tubes withdiameters 22 and 2.2 mm. Both tubes can be fed withfuel/air mixtures independent of each other. Throughthe outer tube a methane/air mixture is supplied withan equivalence ratio of φ = 1.51 and a flow speed of
Fig. 1. Schematic drawing of the jet flame burner assembly.
1.7 m/s, serving as pilot flame. In this work pure airis injected through the small inner tube at differentflow speeds of 0, 70, and 120 m/s and resulting in aglobal stoichiometry of φ = 1.51,1.02, and 0.83, re-spectively. In spite of the small size and low power(∼3 kW), high turbulence can be readily producedin the coaxial jet burner because of the high-speedshear flow introduced from the inner jet. These char-acteristics enable the operation of this burner in laserlaboratories and facilitate close access to highly tur-bulent flames with advanced laser techniques. Theflame conditions investigated are summarized in Ta-ble 1 and photographs of the corresponding flames areshown in Fig. 2. One can easily see the changes of theflame shape and the flame height as the flow velocityof the inner air jet is increased.
A similar burner has been studied by Lee andMitrovic [50,51]. In their setup the diameters of theinner and outer tubes have been 2 and 17 mm, re-spectively. Various conventional methods have beenemployed in their work to characterize the generatedturbulent jet flames. In detail, laser Doppler velocime-
J. Kiefer et al. / Combustion and Flame 154 (2008) 802–818 805
Note. v0, outer jet flow speed; vi , inner jet flow speed; φ0,equivalence ratio of the outer jet flow; φtot, overall equiva-lence ratio of flow from both the inner and the outer jet; Re,Reynolds number based on the inner jet for flames 2 and 3and the outer jet for flame 1.
Fig. 2. Photographs of the investigated flames: (a) flame 1,(b) flame 2, and (c) flame 3, corresponding to Table 1.
try has been employed for flow speed measurements.Laser tomography has been applied to visualize theflame structure, allowing the derivation of, e.g., flamecurvature. Laser Rayleigh scattering has been used tomeasure temperatures. In general, the flames stabi-lized on their burners have looked fairly similar to theones investigated in the present work. However, oper-ating conditions in terms of flow rates, stoichiometry,and the ratio of tube diameters are different and hencethe results cannot be compared quantitatively.
2.2. Measurement technique
Different excitation and detection approaches forOH LIF imaging have been discussed in litera-ture [18,37]. In this work the Q1(8) line of theA2Σ+–X2Π(1,0) transition at 283.6 nm is excitedusing the second harmonic radiation from a Nd:YAG-pumped dye laser (Rhodamine 6G) and fluorescencefrom the transitions A2Σ+–X2Π(0,0) and (1, 1) ataround 308 nm is imaged through a bandpass filter(Schott UG-11).
Regarding the CH radical, there are two electronictransitions, A2Δ–X2Π and B2Σ–X2Π , which areusually excited for LIF applications in the near ultra-violet and visible spectral range. A short summary ofdifferent excitation-detection schemes has been given
Fig. 3. CH signal strength in laminar Bunsen flames asa function flame stoichiometry; the diamonds represent themean value from 50 single-shot images and the standard de-viation is indicated by error bars.
in Ref. [35]. In the experiments presented here, exci-tation of the B–X(0,0) R-branch bandhead is madenear 387.3 nm, and broadband fluorescence from theB–X(0,1), A–X(1,1), and A–X(0,0) bands over-lapping at around 431 nm is collected. The fluo-rescence emission caused by population of the A
state is due to electronic energy transfer from the B
state [52,53]. This approach allows sufficient suppres-sion of the elastically scattered light by a simple col-ored glass filter (Schott GG-400) and the collectionof an ample amount of signal. As already stated inthe Introduction, the CH number density is a strongfunction of the local flame stoichiometry. To empha-size this, the peak CH signal intensity obtained in CHPLIF images of laminar Bunsen flames with varyingequivalence ratio is plotted in Fig. 3. There it turns outthat the strongest signal is observed in slightly richmethane/air flames and a steep CH signal decrease ispresent going to leaner and richer conditions. Flameswith mixtures leaner than φ = 0.6, which is alreadyclose to the lean flammability limit of methane/airmixtures, cannot be stabilized on the employed Bun-sen burner. However, since the signal intensity is rea-sonably good at φ = 0.6, it can be assumed that itis possible to detect CH in the full range of localequivalence ratios present in the partially premixed jetflames under investigation.
A schematic drawing of the experimental setupis shown in Fig. 4. The pulsed alexandrite laser (seeRef. [54]), with its remarkable features such as pulseduration of 150 ns and pulse energy around 200 mJat 774 nm, is operated multimode in order to achievemultiple line excitation in the B–X(0,0) system ofCH after second harmonic generation (SHG). Further
806 J. Kiefer et al. / Combustion and Flame 154 (2008) 802–818
Fig. 4. Experimental setup for simultaneous CH/OH PLIFmeasurements: BS, beam splitter; WM, wavelength meter;SHG, second harmonic generation; PB, Pellin–Broca prism;DC, dichroic mirror; OC, optical components; CH, detectionsystem for CH PLIF; OH, detection system for OH PLIF.
details on this excitation scheme have been discussedby Li et al. [35]. The frequency-doubled radiation(70 mJ at 387 nm) is overlapped with the second har-monic output from the Nd:YAG-pumped dye laser at283 nm (5 mJ) for OH PLIF. The laser pulses aretemporally delayed approximately 100 ns to avoid in-terference. A laser sheet of various heights, either 80or 15 mm, is formed using cylindrical and sphericallenses to image the whole flame and local flame struc-tures in detail, respectively. OH and CH are detectedseparately with two intensified charge-coupled device(ICCD) cameras of the same type (Lavision FlameStar) perpendicular to the laser pathway on oppositesides of the flame. For every flame condition 100 si-multaneous single-shot images are recorded.
2.3. Data evaluation
The data recorded are processed in different waysto gain various kinds of valuable information aboutthe combustion process. To provide data for the vali-dation of numerical simulations, the single-shot im-ages are analyzed by means of statistics. To do sofor every pixel of the camera, the signal is averagedover the recorded 100 single shots and, in addition,the signal standard deviation is determined for everypixel. The resulting averaged images contain infor-mation about the mean radical distributions and thestandard deviation images illustrate the radicals’ lo-cal number density fluctuations. These images show-ing the mean shape of the individual flames providedata for the comparison with numerical simulations,as they include the Reynolds-averaged values and inaddition their variation with time. However, it should
be noted that effects such as shot-to-shot fluctuationsof the laser pulse energy and camera noise also con-tribute to the standard deviation to a certain extent.Nevertheless, the main contribution is due to fluctua-tions caused by the flame turbulence.
For the characterization of stoichiometric and leanpremixed flames, the reaction progress variable andthe flame surface density (FSD) are often derivedfrom Rayleigh scattering or OH PLIF images. For thispurpose the hot or OH-containing areas are assignedto burnt and the other regions to unburnt gas. This isdone because, in this sort of flame, the main chemistrytakes place in a single thin reaction zone. In contrast,in rich partially premixed flames, two reaction zonescan be observed, as will be described in more detail inthe next section. Therefore it is not reasonable to dis-tinguish strictly between burnt and unburnt gas onlybecause of the presence of OH.
From the single-shot OH images, the local FSD isdetermined along the burner axis above the burner exitwith 1.6-mm spatial resolution. This means that the80 × 1.6 mm area at the burner center axis is dividedinto 1.6 × 1.6 mm squares in which the FSD is de-termined. For this purpose, the images are processedby assigning the value 1 to the burnt region corre-sponding to the presence of OH and 0 to the unburntregion, resulting in an instantaneous binary map ofthe OH probability in the partially premixed flamesinvestigated. To correct for shot-to-shot fluctuationsand local intensity variations (e.g., caused by inhomo-geneities of the laser sheet or differences in concentra-tion; note that the first item can in principle be avoidedby using a beam homogenizer [55]), a dynamic inten-sity threshold for distinguishing between burnt andunburnt gas is set. This is done in such a way thatat the flame front locations the threshold is assignedto the strongest signal gradient position. This pro-cedure has been compared in preliminary tests withroutines employing constant threshold values (see,e.g., Refs. [38,56–58] and references therein). Usinga fixed threshold, it has been found that the resultsvary significantly with the set value. Fig. 5 shows thecenter region of a raw OH PLIF image (left) recordedin flame 2. The intensity distribution along the dottedline is plotted in the middle part. The dynamic in-tensity thresholds at the interfaces between the burntand unburnt regions are indicated by arrows. One canclearly see that if a fixed threshold value, e.g., oneoptimized for the areas with the strongest signal inten-sity, were employed, other regions with lower signalintensity might not be identified correctly. To the rightin Fig. 5, the resulting binary image is illustrated.A similar procedure is used to process the CH im-ages. The binary maps are generated by employing adynamic threshold intensity value, ensuring that onlythe sharp CH layer is captured.
J. Kiefer et al. / Combustion and Flame 154 (2008) 802–818 807
Fig. 5. Evaluation procedure: raw image (left), intensity distribution along a single-pixel line (center), binary image (right).
The instantaneous FSD in three dimensions iscommonly defined as
(1)Σ = δA
δV
or alternatively in two dimensions as
(2)Σxy = δL
δA,
where δV is an infinitesimal volume and δA is theinfinitesimal flame surface area in δV . In the two-dimensional planar case δL represents the infinitesi-mal flame front length in the infinitesimal area δA. Byaveraging the instantaneous FSD derived from single-shot images, the Reynolds-averaged two-dimensionalFSD is calculated.
3. Results and discussion
3.1. Chemiluminescence and single-shot PLIFimages
The photographs of the chemiluminescence im-ages of the chosen flames are shown in Fig. 2. In thesephotos, the bright/white thin zone corresponds to CH∗radiation and the thicker gray layer corresponds to thehot product gas radiation. Flame 1 in Fig. 2a is a typi-cal laminar rich premixed Bunsen flame. Without theinner jet airflow, the cone-shaped reaction zone of theinner rich premixed flame is clearly shown. This re-action zone is very thin in the entire flame. With a
flow speed of 70 m/s from the inner air jet (flame 2),one can see in Fig. 2b that the shape of the flamechanges drastically in the proximity of the burner exit.Within one burner diameter, the bright/white layeris still very thin, showing a laminar flame structure.The angle of the laminar flame is larger than that offlame 1. The bright/white zone is much thicker furtherdownstream, showing a typical turbulent mean flamestructure. The thickened mean flame brush is due tothe rapid fluctuation of the flame fronts in this zone,where the flow has transited to turbulent. The hot gaszone outside of the inner reaction layer is smaller ow-ing to the entrainment of ambient air toward the flame,which “compresses” the flame and intensifies the mix-ing and reactions. This trend is kept as the inner airjet speed is increased further to 120 m/s (flame 3), asshown in Fig. 2c. The laminar flame height in prox-imity to the burner exit is further shortened, corre-sponding to an earlier transition of laminar flow toturbulence and enhanced entrainment of ambient airto the flame. The size of the turbulent flame is alsoshortened, owing to an increased combustion rate.
To visualize the detailed structures of the flames,single-shot CH/OH PLIF imaging has been per-formed employing laser sheets with height 80 mm andthickness 200 µm. By crossing the flame center withthe two laser sheets, qualitative CH and OH radicaldistributions across the jet axis have been recorded.Fig. 6 displays simultaneously recorded single-shotimages of OH and CH from flame 1. The CH radi-cals show up within a thin layer some 220 µm thickin the inner reaction zone and the signal intensity is
808 J. Kiefer et al. / Combustion and Flame 154 (2008) 802–818
Fig. 6. Single-shot PLIF images of flame 1: (a) CH radical and (b) OH radical.
approximately constant over the entire flame heightinvestigated. This thin CH layer corresponds to thethin bright/white zone in Fig. 2a. The OH radicals,in contrast, are distributed all over the product zone,with an explicitly visible area of maximum inten-sity located in the outer flame region. Similar flamecharacteristics have been identified by other groupsusing CH, OH, and NO PLIF along with Ramanand Rayleigh scattering measurements [48,49]. Inthose works a careful analysis of a number of par-tially premixed methane/air Bunsen flames has beenperformed, showing that in the inner reaction zonewhere CH is present, the intermediates CO and H2are produced. Subsequently, these intermediate fuelsare further converted to CO2 and H2O. The peak OHappears in the outer reaction zone near the stoichio-metric contour. Similar flame structures have beenshown in other types of rich premixed flames, e.g., byusing dual pump CARS [14,59,60]. To further under-stand the structure of the laminar flame, a numericalcalculation of the partially premixed flame has beencarried out in a one-dimensional counterflow config-uration with one stream injecting methane/air withequivalence ratio 1.51 and one stream injecting pureair [61]. The numerical simulation also shows thatthere are two distinct reaction layers in the flame.One is a rich premixed flame front where fuel is con-sumed and the CH concentration is high. CO and H2are produced in this rich premixed flame front. Theseintermediate species are transported toward the air
stream by diffusion and finally oxidized in a diffusionflame front where a high OH concentration is found.In between the two reaction zones, the OH concen-tration and thus the LIF signal is low due to the lowavailability of oxygen in this buffer zone. The numer-ical results demonstrate the usefulness of utilizing CHPLIF to monitor the fuel consumption layer and OHPLIF to monitor the oxidation layer of CO and H2in this type of flame. Furthermore, it is worthwhileto mention that although laminar, the flame is notsteady. This can also be observed from the instanta-neous PLIF images revealing a slightly wrinkled CHlayer, which can be attributed to the nonuniformityof the flow velocity of the incoming fuel/air mixture.Additionally, a large-scale oscillation of the outer dif-fusion flame front with a frequency of about 23 Hz isnoticed. The underlining mechanism for this oscilla-tion is Rayleigh–Taylor instability.
In Fig. 7 the instantaneous CH and OH imagesrecorded from flame 2 are illustrated. The strongestCH signal is observed from ∼10 up to ∼40 mm abovethe burner. Then the signal intensity decreases withincreasing height. This can be attributed to the factthat the CH number density is a strong function ofthe local flame stoichiometry. As shown in Fig. 3,a lean or rich mixture results in low CH LIF signalintensity in the reaction zone. Owing to the strongturbulence mixing with the inner air jet, as well asthe entrainment of surrounding ambient air, the richpremixed gas is diluted along the downstream direc-
J. Kiefer et al. / Combustion and Flame 154 (2008) 802–818 809
Fig. 7. Single-shot PLIF images of flame 2: (a) CH radical and (b) OH radical.
tion. The mixture gets slightly rich, first becomingmanifest with an increasing CH signal. Then the mix-ture becomes stoichiometric and finally lean along thedownstream direction, resulting in a lower CH num-ber density and thus a lower CH LIF signal at fardownstream positions. The flame front imaged by CHis a closed line up to some 60 mm above the burnerexit with an increasingly irregular, wrinkled turbulentflame structure along the flame height downstream.Beyond 60 mm above the burner, pockets of CH sep-arated from the main flame can be observed, but stillthese CH layers form closed lines. From the CH PLIFimage alone it is not possible to conclude whetherthese CH fragment cells are formed from burnt or un-burnt gas, and therefore the local flame propagationdirection cannot be extracted. The structures observedin the OH image can be unambiguously related to theCH layer. In the lower pilot part of the flame a com-paratively weak OH signal was recorded because of alow OH number density in the rich premixed flamesection. The signal intensity increases in the upperpart of the flame due to the production of the inter-mediate fuel components. From the simultaneous OHand CH PLIF images it is straightforward to under-stand the structure of flame 2. In the lower part of theflame (up to 20 mm above the burner exit), the flamestructure is essentially similar to that of the laminarpilot flame (cf. Figs. 6, 7). There are two distinct re-action zones, an inner premixed flame with a rathersmooth reaction front and an outer diffusion flame.
From 20 mm above the burner exit and further down-stream, the effect of the high-speed air jet from theinner tube is clearly visible: the inner flame front be-comes wrinkled by the turbulence eddies formed inthe shear layer of the inner air jet. The outer diffusionflame layer remains smooth and laminar-like. Entrain-ment of the surrounding air pushes the diffusion flameinward toward the burner axis. This trend is continuedto a flame height of about 40 mm, where the innerturbulent flame and the outer flame merge together.Further downstream, the fuel is combusted in a singlereaction zone, where the CH layer is still very thin butthe OH layer is much more thickened as compared tothe diffusion flame layer in the lower part of the flame.Beyond a flame height of 40 mm, it is clearly visiblefrom the simultaneous OH and CH PLIF that some airpockets from the inner jet stream are penetrating intothe burnt region. In this turbulent condition, the coldair from the inner jet and from the entrainment fromthe surroundings is still properly separated from theunburnt fuel/air mixture by the product gas, as indi-cated by the OH and CH PLIF.
Shown in Fig. 8 are the simultaneously recordedCH and OH PLIF images from flame 3. With theinner air jet of 120 m/s a small-scale highly turbu-lent flame is clearly observed. From the CH and OHPLIF, one can see smaller wrinkling scales in the in-ner surface of the flame between 30 and 40 mm flameheight (as compared to flame 2). This corresponds tothe smaller turbulence eddies existing in this higher
810 J. Kiefer et al. / Combustion and Flame 154 (2008) 802–818
Fig. 8. Single-shot PLIF images of flame 3: (a) CH radical and (b) OH radical.
speed turbulent jet flow. The most significant differ-ence between flame 3 and the low-speed flame 2 isobserved in the upper part of the flame. The inner airflow blows a hole into the tip of flame 3. In the regionfrom 50 to 55 mm above the burner exit, both the CHand OH LIF signals disappear, showing a breakageof the flame front by the air and the unburnt fuel/airmixture. In this flame region it is reasonable to callthe observed behavior local extinction. This mightbe due to the fact that the local mixture is too leanto burn (recall that the overall equivalence ratio ofthe mixture from the inner and outer jets is 0.83 forthis flame). However, it is furthermore most likely at-tributable to the fact that the local scalar dissipationrate (in the region from 50 to 55 mm above the burnerexit) is fairly high, so that local flame extinction oc-curs. The mixing of the methane/air mixture and thepure air jet is so fast that the required inner-layer tem-perature of the diffusion flame front is not reached.Further downstream in the flame, the OH PLIF imageindicates highly wrinkled OH pockets with high LIFsignal levels and weak CH LIF signals, revealing thereignition of the unburnt fuel/air mixture at the tip re-gion of this flame. Furthermore, continually burningseparated flame pockets are observed.
In addition to the PLIF images recorded from theentire flame, a small part of the flame, in a 15×15 mmwindow (as indicated by a square in Figs. 7 and 8), isimaged with the full ICCD chip, using the zoom op-
tics of the cameras to achieve a higher spatial resolu-tion. By doing so, the local structures can be mappedin more detail and even the very small-scale config-urations can be resolved. As an example, Fig. 9 dis-plays two pairs of simultaneously recorded CH andOH PLIF images from flames 2 and 3, showing theflame areas inside the boxes in Figs. 7 and 8, respec-tively. It is interesting to note that CH LIF signalsare observed in rather thin layers in general, althoughcertain low-level signals of CH are detected in mod-erately wide regions, in particular on the unburnt sideof the flame, where the OH LIF signal is absent. Bylooking carefully at the CH and OH PLIF pairs (e.g.,Figs. 9c and 9d), one can see that in some places(e.g., the circle zone), the CH LIF signal is nearlyinvisible but the OH LIF signal still exists. This canbe explained by the fact that the OH lifetime in thehigh-temperature region is much longer than that ofthe CH radical. Similar results showing a breakage ofthe CH layer have been reported in previous works.For instance, Vagelopoulos and Frank [27] revealedby numerical simulations as well as experiments thatdiscontinuities in the CH front may not be an indica-tion of local extinction in environments where vari-ations in reactant composition may exist. However,in the experiments shown there and in the referencestherein, the single-shot CH detection limit might havebeen comparatively limited. Therefore, in those ex-perimental works, it could not be proved whether the
J. Kiefer et al. / Combustion and Flame 154 (2008) 802–818 811
Fig. 9. Single-shot PLIF images of flames 2 and 3 at high spatial resolution: (a) CH radical in flame 2, (b) OH radical in flame 2,(c) CH radical in flame 3, and (d) OH radical in flame 3.
disappearance of the CH LIF signal has been due tocomplete disappearance of CH or just to the limitedsensitivity of the measurement technique. Employingthe CH PLIF approach used here, this uncertainty isclarified to a wider extent; i.e., it can be assumed thatCH is nearly quenched in the present case.
3.2. Flame statistics
The OH/CH PLIF images provide a qualitative de-scription of the flame structures and dynamics. Theseinstantaneous images may be insightful for modeldevelopment; however, to validate numerical simula-tions, it is necessary that statistical data be compared.It is almost impossible to compare numerical resultswith an instantaneous image from a highly turbulentflame, since it might represent a unique situation thatmay not be reproduced precisely in a simulation.
Fig. 10 shows the ensemble-averaged (Reynolds-averaged) CH and OH distributions, as well as theirstandard deviations calculated from the single-shotdata of flame 1. In this laminar flame the CH layerin the lower part of the flame is essentially found in arather narrow region; in the upper part the mean CHlayer is broadened toward the tip of the flame. Thebroadening of the mean CH layer is a result of theunsteadiness of the premixed flame, as already ob-
served in the chemiluminescence image displayed inFig. 2a. In contrast to the narrow mean CH layer, themean OH layer is much broader. The large-scale un-steady motion due to the Rayleigh–Taylor instabilityand the already thick OH laminar flame structure (cf.Fig. 6b) results in the broadening of the mean OHlayer. Due to the unsteadiness of the present flame,certain levels of standard deviation indicating fluctua-tions of CH and OH LIF signals are found in the meanflame zone. However, the level of these fluctuations islower than that in the turbulent flames. Note that in anideal steady laminar flame the standard deviations ofCH and OH signals are expected to be negligible.
Fig. 11 illustrates the CH and OH averaged im-ages and their fluctuations from flame 2. Here theoverall picture is rather different from that of flame 1.The highest CH mean signal intensity is found in aquite narrow region up to 30 mm above the burnerexit. Then the CH is broadly distributed over a ra-dial distance of approximately 10 mm. As it spreadsover a wide zone in the upper part of the flame, themean CH signal also has low intensity. This is a typ-ical turbulence broadening effect. The CH standarddeviation is high at the midflame height correspond-ing to the strong wrinkling of the CH layer in thisregion (cf. Fig. 7a). In contrast, at the flame tip, theCH standard deviation is low, due to its low LIF sig-
812 J. Kiefer et al. / Combustion and Flame 154 (2008) 802–818
Fig. 10. Statistically evaluated PLIF images from flame 1: (a) CH average, (b) CH standard deviation, (c) OH average, and(d) OH standard deviation.
nal level, which corresponds to the lean mixture at thetip of the flame (cf. Figs. 3 and 7a). In the lower partof the flame (below flame height 40 mm), the meanflame OH is naturally found in a wide region, owingto the wide radial distribution of OH in the presenttwo-layer flame. The OH standard deviation in thisregion is correspondingly lower. In the upper part ofthe flame (above flame height 40 mm), the mean OHlayer is broadened by turbulence and the peak mean
OH is found around 5–10 mm radial distance from thecenter axis of the burner. The peak OH standard de-viation is found around 0–5 mm radial distance fromthe axis, due to the higher wrinkling flame structurein this region (cf. Fig. 7b).
The CH/OH averaged images and standard devia-tions from flame 3 are displayed in Fig. 12. The meanCH structure in the lower part of the flame is nearlyidentical to flame 2 (Fig. 11a), except that the pilot
J. Kiefer et al. / Combustion and Flame 154 (2008) 802–818 813
Fig. 11. Statistically evaluated PLIF images from flame 2: (a) CH average, (b) CH standard deviation, (c) OH average, and(d) OH standard deviation.
part of the flame is more compact (shorter). This canbe explained by the increased flow speed of the in-ner jet resulting in an earlier transition to turbulenceand an increase of the entrainment flow of ambientair, which influences the pilot flame. Compared toflame 2, the mean CH in the upper part of the flame(above flame height 50 mm) is rather low, correspond-ing to the low-level CH due to the leaner mixture andpartial extinction in this region. Accordingly, the CH
standard deviation is fairly low in this region. The lo-cations where the main OH radicals appear are foundsignificantly closer to the burner exit, as compared toflame 2. Moreover, in the upper part of the flame, themean OH layer is much narrower in the radial direc-tion, indicating a more intensive reaction zone in thishighly turbulent flame condition. The peak OH stan-dard deviation is found in the inner region of flamedue to the highly wrinkled flame surface; and quite
814 J. Kiefer et al. / Combustion and Flame 154 (2008) 802–818
Fig. 12. Statistically evaluated PLIF images from flame 3: (a) CH average, (b) CH standard deviation, (c) OH average, and(d) OH standard deviation.
high OH fluctuation is observed in the entire meanOH layer (e.g., 15 mm radial distance from the axis atflame height 60 mm).
3.3. Flame surface density (FSD)
From the binary single-shot OH maps, the flamesurface density (FSD) defined in Eq. (2) is determinedand its evolution along the burner axis is illustrated
in Fig. 13 for the three flames discussed above. Ingeneral, the FSD distribution is an indicator of thelocations where the actual combustion takes place.Moreover, its absolute value corresponds to the lo-cal intensity of the combustion process (or the fuelconsumption rate per unit volume) as it is defined bythe flame surface area per volume. For the laminarflame 1, the flame surface density remains zero until60 mm above the burner exit. Then it increases toward
J. Kiefer et al. / Combustion and Flame 154 (2008) 802–818 815
Fig. 13. Evolution of the Reynolds-averaged flame surface density along the burner axis in flame 1 (solid line), flame 2 (dashedline), and flame 3 (dotted line).
its maximum at some 75 mm and decreases again.The profile for flame 2 also shows a well-definedreaction region between heights 30 and 60 mm. Asalready described for the mean OH field (Fig. 11),this region is shifted toward the burner exit as com-pared to flame 1. The combustion intensity, i.e., theflame surface density, is increased by a factor of ap-proximately 4 as compared to the laminar case. More-over, the FSD distribution is significantly broadenedin the turbulent case. Note that in an ideal steady lam-inar flame the spatial distribution of the FSD alongthe burner axis would show an infinitesimal narrowpeak.
The evolution of the FSD derived from flame 3shows even a further increase of the maximal flamesurface density. Compared to the laminar case, themaximum is increased by a factor of approximate-ly 4.5, and its peak location is further shifted up-stream toward the burner exit. Moreover, in contrastto flames 1 and 2, the flame surface density is signif-icantly higher in the downstream region. After reach-ing the maximum some 40 mm above the burner exit,it slowly decreases and not vanishes in the entire re-gion investigated. This corresponds to the quenchingand reignition flame structures as evidenced from thesingle-shot images.
Table 2Flame surface densities determined from CH and OH imagesat high spatial resolution
Σ from CH (m−1) Σ from OH (m−1)
Flame 2 2497 2533Flame 3 1926 2583
3.4. Flame surface density measurements from CHand OH PLIF images
As described in the Introduction, the OH radicaldistribution is commonly used for the determinationof the FSD. However, for this purpose, the length ofthe boundary of the OH regions must be determinedby means of computational image processing. An al-ternative and comparatively simpler approach couldbe the utilization of the CH distribution. Since CHexists in a very narrow layer of the reaction zone, itrepresents the flame surface itself. For comparison,we calculate the FSD from 100 single-shot imageswith high spatial resolution of flames 2 and 3 (as ex-emplarily illustrated in Fig. 9) using both OH and CHPLIF images. The flame surface density is evaluatedat the flame height 50 mm above the burner, wherequenching and reignition occur in flame 3. The resultsare summarized in Table 2.
816 J. Kiefer et al. / Combustion and Flame 154 (2008) 802–818
One can see that for the moderately turbulentflame 2 the derived values from OH and CH are al-most identical, while for the highly turbulent flame 3the CH-based FSD is about 25% lower than that fromthe OH images. As described earlier (Figs. 9c, 9d), theCH PLIF signal in flame 3 disappeared locally, whileOH is still present, and as a result, the flame surfacedensity varies. It appears that CH PLIF can be usedas a suitable measure to determine FSD in the caseof connected flame surfaces. However, care must betaken in the case where broken CH layers are presentin the single-shot images. This can be the case at highturbulence intensity and small-scale turbulent flames.
In the case where multiple reaction zones exist inthe flame, such as the two-reaction-zone structures inproximity to the burner in the present partially pre-mixed flames, care must be taken when the FSD inevaluated from the CH images. As shown in the re-sults above, the CH signal in the diffusion flame layeris nearly undetectable. The OH signal, on the otherhand, exists in both the premixed flame layer and thediffusion flame layer. In such cases, the CH-basedFSD corresponds to the flame surface density of thefuel consumption layers, whereas the OH-based FSDcan be used to evaluate the flame surface density inboth the fuel consumption layer and the CO/H2 oxi-dation layer.
4. Summary and conclusions
Simultaneous CH and OH planar laser-inducedfluorescence imaging has been applied to investigatepartially premixed methane/air jet flames. The mul-tiplex excitation scheme applied for CH PLIF witha long-pulse alexandrite laser system operated mul-timode facilitated CH detection in a wide range ofstoichiometric ratios. The partially premixed flamesare stabilized on a coaxial burner, where from theouter tube a rich premixed methane/air mixture hasbeen supplied and pure air at different flow speedshas been introduced from the inner tube. In particular,three different flames have been investigated. By sys-tematically varying the pure air jet flow speed, differ-ent flame dynamics have been characterized, includ-ing moderately-high-turbulence-intensity and high-turbulence-intensity flames with local extinction andreignition.
The single-shot data recorded have been used toanalyze local flame structures. In general, good cor-relation between the simultaneously recorded single-shot images of CH and OH has been found. In thecase of a laminar flame a distinct two-layer flamestructure has been identified. A rich premixed innerflame layer consumes fuel and forms intermediatespecies such as CO and H2. This layer is characterized
properly by CH PLIF and also OH PLIF. Intermedi-ate species such as CO and H2 are consumed in anouter diffusion-controlled flame where the OH signalis high. Large-scale oscillatory motion was observedin the outer diffusion flame layer. The oscillationfrequency has been determined to be approximately23 Hz and it has been attributed to Rayleigh–Taylorinstability. In the case of turbulent flames with high-speed central air jets, the flame structures in prox-imity to the burner exit are similar to those of thelaminar flame. At the upper part the premixed flameand the diffusion flame merge together. At moder-ate turbulence intensities, a closed CH layer has beenmonitored. However, small parcels of cold gas havebeen present downstream from the main flame, whichcould be identified by the simultaneous OH distribu-tions. In the highly turbulent flame, instantaneous lo-cal quenching of the flame has been observed, whereboth the CH and the OH signals have been absent.Further downstream, the CH and OH signals reap-peared, indicating a reignition of the fuel/air mixture.These observations about the local flame structuresmay be useful for the further development of com-bustion models.
The instantaneous single-shot images have beenprocessed computationally in order to provide statisti-cal data. For this purpose, Reynolds-averaged CH andOH distributions and the corresponding standard de-viation distributions have been calculated. From theseimages, flow domains of high and low LIF signalintensities (indicating the corresponding CH or OHconcentration) as well as high-signal-intensity fluctu-ations have been identified. The CH and OH statisticsdata provide the possibility of validating the meanflame shape of numerical simulations.
Furthermore, the flame surface density has beendetermined from the OH images along the burner cen-ter axis with a spatial resolution of about 1.6 mm inorder to provide quantitative information. The FSDquantitatively reveals the reaction zone structure, in-cluding the reignition behavior in the highly turbulentflame case. The CH PLIF images have also been uti-lized to derive flame surface densities in the flamesat moderate and high turbulence. In the first case,this leads to values almost identical to those derivedfrom the OH images, while at high Reynolds num-bers, the value determined from CH has been foundto be significantly lower. This could be explained bythe observation of broken CH layers in the single-shotimages.
The present study illustrates the advantages andlimitations of CH and OH PLIF for the investigationof turbulent premixed flames, as it can provide sub-stantial information for an insightful understanding oflocal flame structures, as well as important data forthe validation of numerical simulation results. How-
J. Kiefer et al. / Combustion and Flame 154 (2008) 802–818 817
ever, care must be taken when processing the imagesto obtain statistical data such as the flame surfacedensity. For the sake of completeness it should benoted that the present simultaneous CH and OH PLIFmeasurements could be even more valuable for under-standing the local flame structure if the local equiva-lence ratio were known.
Acknowledgments
This work has been financed by SSF (SwedishFoundation for Strategic Research) and the SwedishEnergy Agency through CECOST (Centre for Com-bustion Science and Technology) and VR (SwedishResearch Council). J. Kiefer gratefully acknowledgessupport from the European Union (Large Scale Fa-cility in Combustion, Contract No. 026136 (RITA),LUCC) and the Erlangen Graduate School in Ad-vanced Optical Technologies (SAOT).
References
[1] R.S. Barlow, Proc. Combust. Inst. 31 (2007) 49–75.[2] A.C. Eckbreth, Laser Diagnostics for Combustion
Temperature and Species, second ed., Gordon andBreach, 1996.
[3] S. Pfadler, M. Löffler, F. Dinkelacker, A. Leipertz, Exp.Fluids 39 (2005) 375–384.
[4] I. Grant, Proc. Inst. Mech. Eng. C J. Mech. Eng.Sci. 211 (1997) 55–76.
[5] D. Most, A. Leipertz, Appl. Opt. 40 (2001) 5379–5387.[6] R.B. Miles, W.R. Lempert, J.N. Forkey, Meas. Sci.
Technol. 12 (2001) R33–R51.[7] P.-E. Bengtsson, M. Aldén, Appl. Phys. B 60 (1995)
51–59.[8] S. Will, S. Schraml, A. Leipertz, Opt. Lett. 20 (1995)
2342–2344.[9] W. Meier, O. Keck, Meas. Sci. Technol. 13 (2002) 741–
749.[10] W. Meier, R.S. Barlow, Y.L. Chen, J.Y. Chen, Combust.
Flame 123 (2000) 326–343.[11] T. Seeger, J. Egermann, S. Dankers, F. Beyrau, A.
Leipertz, Chem. Eng. Technol. 27 (2004) 1150–1156.[12] F. Beyrau, T. Seeger, A. Malarski, A. Leipertz, J. Ra-
man Spectrosc. 34 (2003) 946–951.[13] M.C. Weikl, F. Beyrau, J. Kiefer, T. Seeger, A.
Leipertz, Opt. Lett. 31 (2006) 1908–1910.[14] M.C. Weikl, T. Seeger, R. Hierold, A. Leipertz, J. Ra-
man Spectrosc. 38 (2007) 983–988.[15] S. Roy, T.R. Meyer, R.P. Lucht, M. Afzelius, P.-E.
Bengtsson, J.R. Gord, Opt. Lett. 29 (2004) 1843–1845.[16] S. Roy, T.R. Meyer, M.S. Brown, V.N. Velur, R.P.
[25] P. Petersson, J. Olofsson, C. Brackmann, H. Seyfried,J. Zetterberg, M. Richter, M. Aldén, M.A. Linne, R.K.Cheng, A. Nauert, D. Geyer, A. Dreizler, Appl. Opt. 46(2007) 3928–3936.
[26] K.-J. Nogenmyr, P. Petersson, X.S. Bai, A. Nauert, J.Olofsson, C. Brackman, H. Seyfried, J. Zetterberg, Z.S.Li, M. Richter, A. Dreizler, M. Linne, M. Aldén, Proc.Combust. Inst. 31 (2007) 1467–1475.
[27] C.M. Vagelopoulos, J.H. Frank, Proc. Combust. Inst. 30(2005) 241–249.
[28] J.A. Miller, C.T. Bowman, Prog. Energy Combust.Sci. 15 (1989) 287–338.
[29] J. Luque, P.A. Berg, J.B. Jeffries, G.P. Smith, D.R.Crosley, J.J. Scherer, Appl. Phys. B 78 (2004) 93–102.
[30] J. Luque, R.J.H. Klein-Douwel, J.B. Jeffries, G.P.Smith, D.R. Crosley, Appl. Phys. B 75 (2002) 779–790.
