Stratified Jet Flames in a Heated (1390K) Air Cross Flow With Autoignition 2012 Combustion and Flame

aKeywords:CrossowAutoignitionTurbulentFlameletsDistributed reaction

d o-in-

was expected to fall into either the distributed reaction, thickened amelet, or shredded amelet

ain pro, whichportan

Several fundamental questions arise. First, if the air tempera-ture exceeds the autoignition temperature, as it may in practicaldevices, how is the fundamental structure of the ame changed?For air temperatures greatly exceeding the autoignition tempera-

velocity, so it may be possible for some regions to experienceautoignition while other regions may contain a ame.

A second issue is how the ame structure is altered when thecross-ow air velocity is increased. The non-dimensional aerody-namic strain rate is dened to be the Karlovitz number, and whenthis number is sufciently large the combustion is expected to en-ter the regimes of thickened amelets, shredded amelets or dis-tributed reactions [15,16]. For the present JICF experiment a verylarge Karlovitz number is achieved because of the large air velocity(468 m/s) and the small jet diameter (2.49 mm).

Corresponding author. Address: 3004 FXB Building 2140, Department ofAerospace Engineering, University of Michigan, Ann Arbor, MI 48109-2140, USA.Fax: +1 734 763 0578.

E-mail addresses: [email protected] (D.J. Micka), [email protected] (J.F. Dris-coll).

Combustion and Flame 159 (2012) 12051214

Contents lists available at

Combustion

evi1 Present address: Creare Inc., Hanover, NH 03755, USA.vices such as lean-premixed-prevaporized gas turbine combustors[1], ramjets and scramjets [2] and HCCI engines [3]. A number ofprevious studies have considered the JICF geometry, but have beenlimited to either non-reacting conditions [49] or to reacting con-ditions for small cross-ow velocities that were less than 10 m/s[1014]. The reason for the small values of cross-ow velocity isthat the ame will blow out unless the air temperature is raisedsignicantly.

mary nature of the combustion, ame or autoignition, is less clear.A ame is dened to be a combustion event for which the diffusionterms in the enthalpy balance are comparable to the reactionterms. This is an assumption that is invoked in conventional pre-mixed and non-premixed ame theories. For an autoignition pro-cess the diffusion terms in the enthalpy balance are smallcompared to the reaction terms. In a realistic JICF there are highlynon-uniform regions of fuelair ratio, gas temperature and gas1. Introduction

There is a need to document certtion of a fuel jet-in-cross-ow (JICF)physics problem. The JICF plays an im0010-2180/$ - see front matter 2011 The Combustdoi:10.1016/j.combustame.2011.10.013regimes. Fluorescence images of CH, OH and formaldehyde identied the ame structure. The jet-in-cross-ow is a unit physics problem that occurs in turbojets and scramjets. While scaling relations areknown for the non-reacting case, more information about the reacting case is needed, especially whenautoignition and strain rates become important. Three regions were identied. In the liftoff region autoig-nition reactions occur which create a strong formaldehyde PLIF signal. However, ames and heat releasedo not occur in the liftoff region since CH and CH signals were negligible. The second region is the liftedame base, which has the character of a premixed ame, as evidenced by a very rapid rise in the heatrelease rate as indicated by the CH and OH signals. The third region contains a turbulent non-premixedame and the CH images indicate the presence of thickened and shredded amelets. The 23 mm thick-ness of each CH layer is more than 10 times the laminar amelet thickness. In the third region the heatrelease rate decays slowly downstream, which is typical of a non-premixed ame. Because both upstreamautoignition and downstream thickened amelets were observed, we classify this combustion to be anautoignition-assisted ame. Flame lengths increase linearly with fuel mass ow rate, indicating thatmixing is controlled by the air velocity rather than the fuel velocity.

2011 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

perties of the combus-is a fundamental unitt role in propulsion de-

ture, the ignition delay time is negligible compared to uid mixingtime and fuel and air react as quickly as they mix. For air temper-atures only marginally greater than the autoignition temperature,the autoignition delay times may be large compared to the uidtime, particularly in high speed applications. In this case the pri-Accepted 12 October 2011Available online 1 December 2011

of 1390 K, which is above the autoignition temperature, and the air velocity was 468 m/s, which is muchlarger than values that were considered previously. Aerodynamic strain rates are so large that the ameStratied jet ames in a heated (1390 K)

Daniel J. Micka 1, James F. Driscoll Department of Aerospace Engineering, University of Michigan, USA

a r t i c l e i n f o

Article history:Received 21 June 2011Received in revised form 12 October 2011

a b s t r a c t

Measurements are reporteproperties of a reacting jet

journal homepage: www.elsion Institute. Published by Elsevierir cross-ow with autoignition

f the heat release proles, the ame lengths, ame structure and othercross-ow (JICF) for two fuels. The air was heated to a static temperature

SciVerse ScienceDirect

and Flame

er .com/locate /combustflameInc. All rights reserved.

A third issue is that measurements of the ame length and theheat release rate proles of a JICF are needed as a rst step towardthe validation of CFD models. In a previous study in our laboratory[17], two CFD codes were run for the conditions of the presentexperiment; while they both compute reasonable values for thelength of a simple jet ame, they fail to predict the length of a JICFame. Heat release distribution is a critical parameter for ramjetsand scramjets; the combustor area must be designed to divergein the proper manner as the heat is added in order to achieve (oravoid) thermal choking at the desired location. Very few measure-ments of heat release distributions are available.

A few previous studies have considered high temperature and/or high velocity cross-ows. Grout et al. [18] performed a directnumerical simulation (DNS) for an air static temperature andvelocity of 750 K and 55 m/s, respectively. Their DNS was limitedto a modest jet Reynolds number of 3980 and they computeimages of instantaneous ame index and scalar dissipation rates.Their computed ames were nearly attached; their ame basewas lifted by about 2 jet diameters. Ryan et al. [2] conducted ascramjet JICF experiment for which the stagnation temperaturewas 1388 K. Ben-Yakar and Hanson [19,20] and Gamba et al. [21]used a shock tube to burn hydrogen and ethylene jet ames atair static temperatures of 12901400 K and cross-ow air veloci-ties of 2360 m/s; their run time was limited to 270 ls.

Figure 1 is a schematic of three possible reaction zone struc-tures. In Fig. 1a a fuel jet burns in a low temperature, low speedcross-ow of air. Mixing in the liftoff region causes the ame base

Figure 1b indicates that changes can occur when the air streamis heated to a temperature above the ignition temperature and theair velocity is made sufciently large. This schematic is based onthe images that were recorded in the present work, which are de-scribed below. Where the heated air and fuel mix, some auto-igni-tion reactions occur which create intermediates such asformaldehyde in the gray shaded region. The primary reactionshave not yet occurred, so there is no heat release measured inthe liftoff region. The ame base is believed to have a triple amestructure containing both rich and lean premixed regions. Atdownstream locations the non-premixed ame is thickened andshredded. Figure 1c illustrates the different geometry of a amein a heated co-ow, which has been studied by Cabra et al.[22,23] and others [2426]. The co-owing Cabra burner has anair temperature of 1350 K and a relatively low air velocity of5.4 m/s. Autoignition chemistry has been shown to be importantin the Cabra burner, but the aerodynamic strain rates most likelyare too small to create the thickened and shredded reaction layersdrawn in Fig. 1b. In contrast to the current work, the presence ofOH in the autoignition region of Cabra burners indicates thesereactions involve heat release.

2. Experimental arrangement

The measurements were obtained by operating the Michigan

1206 D.J. Micka, J.F. Driscoll / Combustion and Flame 159 (2012) 12051214to have the structure of a stratied premixed (triple) ame. Thecenter of the ame base burns as a fuel-rich premixed ame, whilethe far edges burn as lean premixed ames. The fuel-rich productscome in contact with the air to create a non-premixed ame. Sincethe air velocity must be small to avoid blowout, the ames typi-cally fall within the thin amelet regime. To model this situation,a stratied premixed amelet approach is believed to be applicablefor the ame base, while a non-premixed amelet approach isapplicable for the downstream region.Fig. 1. Schematic of jet-in-cross-ow and jet in co-ow stratied ames, illustrating chanlow speed air case with continuous, thin amelets; (b) heated, high speed air (present wobelow; (c) co-owing Cabra burner case [22,23].JICF experiment that is shown in Fig. 2. Run conditions for thetwo cases considered are listed in Table 1. The facility is describedin Refs. [27,28]; it can provide either a supersonic or a subsoniccross-ow; all cases considered here are subsonic cross-ows(rammode). The heated air rst is accelerated to Mach 2.2 in a con-verging diverging nozzle and then it is decelerated by a series ofshock waves to a Mach number of 0.62 and a velocity of 468 m/s(Case 1), which is typical of ramjet conditions in a dual-mode de-vice. The test section height is 25.4 mm and the width is38.1 mm. A constant area section extends 402 mm upstream ofthe fuel jet. The fuel consists of either gaseous hydrogenethyleneges due to raising the air temperature above the auto-ignition temperature: (a) cold,rk) case with distributed auto-ignition and broken, thickened amelets, as described

Previously it has been shown that the CH PLIF signal is a goodmarker of the primary reaction zones in premixed [2931] and

n-cr

n anFig. 2. Michigan jet-i

Table 1Run conditions for the jet-in-cross-ow study.

Case 1 Case 2

Fuel composition by volume 50% C2H4, 50% H2 H2MWA Vitiated air mol. weight (kJ/kmol) 26.4 26.4cA Vitiated air ratio of specic heats 1.3 1.3TA Air static temperature (K) 1390 1413T0A Air stagnation temperature (K) 1470 1500TF Fuel static temperature (K) 255 247T0F Fuel stagnation temperature (K) 296 296qF Fuel density (kg/m3) 3.47 0.504qA Air density (kg/m3) 0.64 0.62UF Fuel velocity (m/s) 432 1198UA Air velocity (m/s) 468 487MA Air Mach number 0.62 0.64MF Fuel Mach number 1 1/ Equivalence ratio 0.42 0.36R qFU2F =qAU2A1=2 2.1 2.2dF Fuel jet diameter (mm) 2.49 2.49

D.J. Micka, J.F. Driscoll / Combustio(Case 1) or pure hydrogen fuel (Case 2); in both cases the fuel stag-nation temperature is 296 K. The fuel is injected at the sonic speedin a direction that is perpendicular to the air stream through a sin-gle port of diameter (dF) of 2.49 mm that is located in the test sec-tion wall. The fundamental scaling parameter for the JICF problemhas been shown by Hasselbrink and Mungal [5] to be the momen-tum ratio parameter (R). This quantity is dened to beqFU2F =qAU2A1=2 where q and U are gas density and velocity,and subscripts F and A denote properties of the fuel jet and air crossow; values of R are listed in Table 1. The values for the air prop-erties in Table 1 are given just upstream of the fuel injection. Dueto the connement of the test section walls, the air is acceleratedby the combustion heat release as it moves downstream. Detailsare given in Ref. [28].

The air was heated by both an electric heater and a hydrogen-fueled vitiated air heater in order to achieve air stagnation temper-atures up to 1500 K. Make-up oxygen was added to insure that theO2 mole fraction is 0.21. The addition of hydrogen to the ethylenefuel in Case 1 was necessary to achieve ame stabilization. Runtimes were limited to 3 s in order to prevent thermal damage tothe uncooled combustor. A downstream spark plug ignites theame. To perform PLIF imaging, fused silica windows 305 mm inlength were mounted in three sidewalls. The ow conditions anddata acquisition were controlled by a LabView program. For eachrun the air rst was heated to 450 K by the electric heater. The viti-ator oxygen ow was initiated, and 2 s later the vitiator heater wasstarted. Four seconds later the spark was turned on and 1 s laterthe fuel valve was opened. After 3 s of data acquisition, the fuelvalve was closed. The lower wall of the test section contains a cav-ity of height 12.7 mm and length 50.8 mm starting 44.5 mm down-stream of the fuel injection as shown in Fig. 2. This cavity hadnegligible effect on the ow streamlines in the free stream, sinceour previous PIV images showed that the upper boundary of thecavity is a shear layer which extends only 23 mm into the freenon-premixed [3234] turbulent combustion. That is, the CH PLIFsignal rises and falls rapidly at the approximate locations wherethe heat release rate also rises and falls. The CH PLIF diagnosticsare described in Appendix A; they are similar to that of Carteret al. [34]. Heat release rates were determined by recording thechemiluminescence from CH and OH with a digital camera andinterference lters. Previous calibrations have been performed byAyoola et al. [29] and others [3541] that showed that chemilumi-nescence is an accurate indicator of the heat release rates (to with-in an uncertainty of about 10%). The total chemiluminescence froma laminar Bunsen ame has been shown to be proportional to theknown total heat release rate, which is the product of the mass persecond of fuel burned and the heating value of the fuel. In the pres-ent work a similar calibration also was performed; results are de-scribed in Section 3.4. Simultaneous formaldehyde-OH PLIFdiagnostics also were used to image both the autoignition zoneand the upstream boundary of the primary reaction zone. Thesediagnostics also are described in Appendix A. The boundary offormaldehyde and OH signals indicates where the primary reac-tions begin [42]. Formaldehyde exists on the upstream (fuel-rich)side. It is formed as part of the initial fuel decomposition reactions[43] and is consumed in the primary reaction layer. OH is createdin the primary reaction layer of ames causing a sharp gradient inOH concentration at this location [41]. OH continues to exist in theproduct gases that are downstream of any ame, so only the up-stream boundary of the OH helps to indicate the ame location[44].

3. Results

3.1. Structure of the primary reaction zone (the ame)stream, which is comparable to the wall boundary layer thickness.Except for the case discussed in Section 3.3, all ames in the cur-rent work had a base located upstream of the cavity leading edge.

oss-ow experiment.

d Flame 159 (2012) 12051214 1207The primary reaction zone was imaged in two ways: directchemiluminescence from CH and CH PLIF signals were recorded.Figure 3 illustrates a CH image for the Case 1 conditions listedin Table 1. The air velocity has a static temperature of 1364 Kand a velocity of 480 m/s. Note that the CH chemiluminescencerst occurs at x/d = 15 in this lifted ame; there is no CH luminos-ity seen in the liftoff region. The CH luminosity extends to about x/dF = 40. CH imaging represents a line-of-sight measurement and isa good indicator of the time-averaged heat release distribution. It isnot a good indicator of the internal structure of the reaction layers,so CH PLIF imaging also was performed.

The structure of reaction layers determined using CH PLIF; someresults appear in Fig. 4. Four regions were recorded separately andat random times, and the images are placed next to each other. Thetime-averaged CH pattern in Fig. 4a looks similar to the CH chemi-

luminescence in Fig. 3; the brightest region in Fig. 4a ends at aboutx/d = 40, as it does in Fig. 3. However, the threshold level in Fig. 4a

CH layers are the high strain rates and the turbulence that may en-ter the reaction layers. If the air was not preheated then the highlystrained ame would blow out. In fact lowering the air tempera-ture by 200 K causes this ame structure to blowout in this exper-iment. Figure 4c is a schematic of the broken, thickened ameletsseen in Fig. 4b. These reaction layers form a coherent boundarythat is seen to enclose the fuel jet. The enclosed uid is expectedto consist of fuel, intermediates and products at these downstreamlocations. At other times, Fig. 4d and e illustrate that the structureis similar to that observed in Fig. 4b. Previously, Joedicke et al. [45]observed a similar ame structure for lifted turbulent jet ames ina cross ow, but in their case the cross ow was a low temperature,low speed air stream.

Experimental uncertainties can cause an articial broadening ofthe measured amelet thickness, but these uncertainties can be ru-led out as the cause of the 23 mm thickness of the present CH lay-ers. The laser sheet thickness was 300 lm, as discussed inAppendix A. Articial broadening occurs when the ame sheet isinclined at an oblique angle to the laser sheet, but the amount ofarticial broadening is typically the thickness of the laser sheet.Since the laser sheet thickness is only 1015% of the measuredCH layer thickness, it is concluded that possible articial broaden-ing does not affect the present results.

Fig. 3. Time-averaged chemiluminescence from the jet ame in a cross-ow. (a)Case 1, CH signal, fuel = H2C2H4. (b) Case 2, OH signal, hydrogen fuel.dF = 2.49 mm.

1208 D.J. Micka, J.F. Driscoll / Combustion and Flame 159 (2012) 12051214was set at a low level to record low levels of CH PLIF signal fartherdownstream.

The instantaneous structure of the downstream (non-premixed)ame is seen in Fig. 4b to be that of a set of broken (shredded) andthickened amelets. A amelet here is dened to be any reactionlayer that is consistently 34 times longer than it is wide. Thus gra-dients are much larger in one direction (normal to the layer) thanin the other two directions. It is not implied that the amelet is alaminar amelet. The CH signal in Fig. 4b, d and e is seen to be con-ned to thickened layers that are 23 mm thick. This is more than10 times thicker than a laminar amelet for these conditions. Twofactors that are believed to be responsible for the thickening of theFig. 4. Structure of the primary reaction zone (broken and thickened amelets) in thevelocity: 1364 K, 480 m/s. (a) time-averaged CH-PLIF signal (b, d, e) instantaneous CH-PLarrows indicate location of the fuel jet. White dashed lines indicate different elds of vi3.2. Structure of the autoignition (liftoff) zone

Figure 5 is a set of simultaneous formaldehyde-OH PLIF imagesfor Case 1. Fuel is injected at the location marked by the arrow, andthe liftoff region is the zone between the black arrow and the ver-tical dotted black line. This vertical dotted black line indicates thestart of the ame base, where the OH signal begins. Figure 5ashows that formaldehyde reactions occur in a broad, distributedregion that is in the liftoff region. The solid white line in Fig. 5and 5b marks the boundary of the formaldehyde region, whilethe OH PLIF signal regions in Fig. 5b indicate the upstream bound-ary of the primary reactions. It is concluded that formaldehyde isformed by homogenous distributed reactions that occur in thedownstream region. Case 1, H2C2H4 fuel, cross-ow air static temperature andIF signal at different times, (c) schematic of the amelet layers seen in (b). Verticalew which were imaged at different times.

liftoff region for this JICF ame. The creation of formaldehyde over10 mm upstream of the heat release zone must be uncorrelatedwith the heat release from combustion. If the air temperaturewas below the autoignition temperature, the formation of formal-dehyde far upstream of the ame base would be unlikely. It is log-ical to conclude that raising the air temperature to a value abovethe autoignition temperature creates formaldehyde in the liftoff re-gion due to autoignition reactions.

In Fig. 6 more simultaneous formaldehyde-OH images areshown. For clarity only the upstream boundary of the formalde-hyde signal is shown (as the white line); the colored regions repre-sent the OH signal. In each image it can be seen that theformaldehyde begins well upstream of the OH. Some images ofthe CH PLIF signal at downstream locations also are shown, butthey were not recorded simultaneously with the upstream OHimages. Cross-sectional views of simultaneous OH/formaldehydePLIF signals are shown in Fig. 7 at three axial locations. The loweredge of each image is the lower wall of the combustor. At the mostupstream location (x/dF = 17.9), the jet core is lled with formalde-hyde due to the initial autoignition reactions decomposing the fuel.OH always appears on the lower side of the jet-core (as shown) andsometimes surrounds the entire jet core. At the downstream loca-

D.J. Micka, J.F. Driscoll / Combustion antions of x/dF = 36.2 and 50.5 the formaldehyde signal is virtuallyzero. A jet core is still often apparent from the OH-PLIF imageswhich have a sharp boundary (marking a reaction layer) surround-ing a region of very low signal (the jet core).

These results tend to conrm that the schematic at the top ofFig. 6 is realistic. Formaldehyde is formed in the fuel jet upstreamof premixed ame base. The OH regions appear to be clustereddownstream of a conical-like region that is labeled the thickenedpremixed ame. The OH signal is more prevalent on the lowerside of the fuel jet than on the upper side in upstream locations.This is expected since the lower side is in the wake of the fueljet, where velocities are lower and static temperatures are higher.On the upper side of the jet the large aerodynamic strain rates ap-pear to cause large regions of ame extinction. By (x/dF = 36.2), theformaldehyde has disappeared, but the CH PLIF and heat releasedistribution results (Results C) indicate that all the fuel has notbeen consumed. This indicates that all the fuel in the jet-core haspassed through the inner, rich premixed ame cone (indicated inthe schematic of Fig. 6) which consumes the formaldehyde [41].

Fig. 5. One example of simultaneous OH and formaldehyde PLIF images. The blackarrow marks the fuel jet, the solid white line marks the outer edge of theformaldehyde signal, and the dotted line marks the ame base. The broad

formaldehyde autoignition region is well upstream of the ame base, which iswhere the OH rst is observed. Case 1, H2C2H4 fuel. Each PLIF image is 25.4 mm.High and 61.5 mm wide.The hot, fuel-rich products of the premixed ame then come incontact with the heated air stream, which creates the thickened,shredded CH layers in the non-premixed ame.

3.3. More evidence for an auto-ignition assisted ame

It was decided to compute the expected ignition delay distancesfor Case 1, in order to determine if autoignition is a reasonableexplanation for the observed regions of formaldehyde that are seenupstream of the main ame base in Figs. 5 and 6. Ignition delaycannot be predicted accurately with a 1-D model because the jetin a cross-ow has non-uniform regions of gas temperature, gasvelocity and fuelair ratio. However, the dashed lines plotted inFig. 8 are the estimated upper and lower bounds. These dashedcurves were generated by running CHEMKIN with the Laskinet al. mechanism [40] for the fuel mixture used in Case 1 (50% eth-ylene, 50% hydrogen). The horizontal axis covers the range of stag-nation temperatures of the experiment. Local equivalence ratio (/)was not known, so / was varied from 0.2 to 1.0 and local Machnumber was varied from 0.4 to 0.6 in the computations. The solidline plotted in Fig. 8 is the measured ame liftoff distance obtainedfrom ame luminosity images.

For a stagnation temperature of 1450 K the upper and lowerdashed lines indicate that the maximum and minimum expectedignition delay distances are 20 and 70 mm, while the measuredame liftoff distance falls between these two values and is35 mm. It is concluded that the measured liftoff distance is a dis-tance that is long enough to allow some autoignition to occur. Thusthe broad region of formaldehyde that was observed in Fig. 5 to liewithin the liftoff distance can reasonably be attributed to autoigni-tion reactions. A second observation is that the dashed lines andsolid lines in Fig. 8 have different slopes. All of the dashed lineshave approximately the same slope, and this slope is a measureof the temperature dependence of the autoignition process. Theslope of the solid line is an indicator of the temperature depen-dence of the measured liftoff distance of the ame base. The differ-ence in slopes is one indicator that location of the ame base is notdetermined solely from autoignition concepts, but is also deter-mined by the turbulent ame propagation speed.

The presented evidence suggests that an autoignition assistedame is the most likely reaction mechanism controlling the stabil-ization and propagation of the Case 1 (and 2) combustion. Howeverit is difcult to completely eliminate the possibility that autoigni-tion dominates the reaction mechanism, with ame propagationplaying a minor role. The dynamic behavior of the combustion dis-cussed in Ref. [27] provides additional evidence that ame propa-gation plays a major role. The combustion was found to be stableonly when the base was located within a narrow region. If any uc-tuation pushed the lifted ame base downstream of this region, itwould blow off entirely. This is expected behavior for stabilizationdue to a lifted premixed ame base. It is inconsistent with an auto-ignition controlled reaction. Any uctuations which push anautoignition controlled reaction downstream would only providethe reactants more time, and therefore make them more likely toautoignite.

Another compelling piece of evidence for an auto-ignition as-sisted ame comes from an entirely different reaction zone struc-ture which was found at lower air temperatures in previous work[27,28]. For these lower air temperatures, the lifted jet ame blewout and the ame was stabilized by the wall cavity downstream.Figure 9 shows images of average CH, instantaneous CH-PLIF,and simultaneous CH2O/OH PLIF for this cavity stabilized combus-tion. This reaction zone features a constant average spreading an-

d Flame 159 (2012) 12051214 1209gle (which depends on temperature and fuel composition) and arelatively continuous reaction layer originating from the top ofthe cavity leading edge. These are exactly the features that would

n a1210 D.J. Micka, J.F. Driscoll / Combustiobe expected for a premixed ame stabilized by the cavity recircu-lation zone and are inconsistent with an auto-ignition dominatedreaction. Figure 9c shows that formaldehyde, indicating initialauto-ignition reactions, was found well upstream of the reactionzone for cavity stabilized combustion as well. Therefore an auto-ignition assisted ame is concluded to be the reaction mechanismfor this cavity stabilized combustion. This cavity stabilized com-bustion occurred for conditions that are virtually identical to thelifted jet ame of Case 1, but with an air temperature approxi-mately 200 K lower.

3.4. Flame lengths and heat release rate proles

Measurements of ame length and heat release rate proles areuseful to assist in the design of combustors and for the assessmentof DNS and LES predictions. Figures 10 and 11 contain the mea-sured heat release rate proles for Cases 1 and 2. To determinethese proles, the images of the OH chemiluminescence wereintegrated along the direction that is normal to the ow, and thisvalue is the intensity prole IOH x. Previous studies have con-cluded that the heat release rate is proportional to the OH inten-sity [29,3538] so it follows that:

qx=Q IOH x=AOH 1The heat release rate q(x) is the number of kilojoules per second

released per unit length in the streamwise direction. The quantityQ is the integral of q dx from x = 0 to x = innity. Q is the known

Fig. 6. Additional simultaneous formaldehyde-OH PLIF images showing that broad distrwhich marks the ame base. Case 1, H2C2H4 fuel. Each image is 25.4 mm high and 61.5 medge of the formaldehyde signal.nd Flame 159 (2012) 12051214total kJ/s of heat released, which equals the product of the fuel owrate and the heating value of the fuel. For Cases 1 and 2 the fuelow rates are 7.9 g/s and 3.0 g/s, heating values are 55,105 kJ/kg,and 119,000 kJ/kg, and values of Q are 435 kJ/s and 357 kJ/s,respectively. The area under each of the measured curves ofIOH x was determined and this area is dened to be AOH . SinceQ is known and both IOH and the area AOH are measured, the heatrelease rate (q) was determined from Eq. (1) without the need forany calibration.

To determine if q is proportional to IOH and to test the validity ofEq. (1), a simple calibration experimentwas performed. The calibra-tion procedure was similar to one that has been used previously[29,3538]. If q is proportional IOH then the integrated value of q,which is Q, should be proportional to the integrated value of IOH ,which is AOH . Measured values of Q and AOH are plotted in Fig. 12and are found to be linearly related. The horizontal axis is the equiv-alence ratio, which is proportional to Q. The vertical axis is the area(AOH ) under the curve of IOH x; values are normalized by thosemeasured for Case 2. Some uncertainty in the calibration arises be-cause it is not known if 100% of the fuel has burned in the viewingregion. The OH signal had to be extrapolated to zero downstreamof the end of the window. Linear extrapolation of the curves inFig. 11 indicated that the amount of heat release in the viewing re-gion was 8598% of the total heat release. Figure 12 veries the lin-earity of Eq. (1) and the uncertainty is approximately 10%.

One conclusion that can be drawn is that all six heat release rateproles in Figs. 10 and 11 have the same shape and they show that

ibuted reactions (autoignition) occur upstream of the upstream boundary of the OHm wide. The black vertical arrow marks the fuel jet, the white line marks the outer

n anD.J. Micka, J.F. Driscoll / Combustiothere are three distinct regions of very different heat release ratesin the JICF. The rst region is the liftoff region (that extends tox = 25 mm in Fig. 9) in which there is negligible heat released.Formaldehyde-forming reactions were shown to occur in this re-gion, but no CH or OH PLIF signals were observed. All curves inFigs. 10 and 11 sharply rise, reach a plateau, and then begin to fallin the second region. This abrupt rise and fall is characteristic of aturbulent premixed ame brush. In the third region the curves dis-play a gradual decay that is characteristic of a non-premixed tur-bulent ame. Thus the heat release proles are consistent withthe PLIF images in that in both cases there are three distinct com-bustion regions that were identied.

Fig. 7. Typical simulataneous formaldehyde and OH PLIF images with the laser sheet pedF = 2.49 mm. Each image is 25.4 mm high and 38 mm wide. High intensity streaks in th

igni

tion

dela

y di

stan

ce (m

)

air temperature (K)

Fig. 8. Dashed lines: computed maximum and minimum expected ignition delaydistances; solid line: measured liftoff distances. Case 1.d Flame 159 (2012) 12051214 1211Each curve in Figs. 10 and 11 has a sharp drop at approximatelyx = 100mm that is a function of the wall connement, not a funda-mental change in the combustion. The owpasses through a thermalthroat around this locationwhich leads to a sharpdrop in the fuelairmixing rates. This is explained inmore detail in Ref. [46]. For the H2C2H4 fuel inFig. 10, thegradualdecrease in theheat releasewhich sig-nies mixing limited combustion begins shortly upstream of thethermal throat. For theH2 fuel in Fig. 11, the ame lift-off ismuch lessand the mixing limited combustion begins at around x = 35 mm.

Flame lengths also were measured and results are plotted inFig. 13 for various fuel ow rates. Flame length is dened to be thedistance fromthe fuel port to the locationwhere90%of the total heatrelease has occurred (using the extrapolated curves in Fig. 11). Usingthis denition, ame lengths from experiments and computationscanbe compared. It is seen that themeasuredame lengths increaselinearlywith the fuel owrate. This scalingdiffers fromthat of a sim-ple turbulent jet with no cross-ow,which has a ame length that isindependent of the fuel velocity. As the fuel velocity is increased forthe case of a simple jet, the turbulent mixing rate increases and theamount of fuel that must be mixed also increases proportionally.Thusame lengthdoesnot change. For the JICF theame lengthdoeschange. This is expected since the conditions fall in the wake-likemixing regime that is described byHasselbrink andMungal [5]. Thatis, in the upstream regions of a JICF the rotation rate of a turbulenteddy (in the shear layer at the edge of the fuel jet) is driven by thelarge fuel velocity on the fuel side of the shear layer. For example,for Case 2 the initial fuel velocity is 1198 m/s. Downstream the fuelvelocity has decayed to a small value, so eddy rotation is determinedinstead by the air velocity. For small cross-ow velocities, JICFames are jet-like and have relatively large values of the momen-tumratio (q) that is dened inTable 1. Thepresentamesare wake-like due to the large air velocity and relatively small values of q.

rpendicular to the ow direction at different axial locations. Case 1, H2C2H4 fuel.e x/dF = 50.5 CH2O image are from laser sheet reections, not PLIF signal.

n and Flame 159 (2012) 120512141212 D.J. Micka, J.F. Driscoll / CombustioSince the air velocity is xed for the conditions of Fig. 13, the wake-like mixing rate also is xed, and any increase in the fuel ow raterequires a longer ame in order to mix the fuel to its stoichiometricvalue.

4. Conclusions

1. A reacting jet-in-cross-ow (JICF) was studied for two differentfuels and for large values of air temperature (1364 K) andincoming air velocity (480 m/s). Heat release proles, ame

Equivalence ratio 0 0.1 0.2 0.3 0.4

A OH

*/AO

H* ,

case

2

norm

aliz

edar

ea u

nder

hea

t rel

ease

cur

ve1.0

0.5

0.0

Fig. 12. Calibration test showing that heat release rate is proportional to OH

chemiluminescence.

Fig. 9. Cavity stabilized premixed ame observed in our previous work at lower airtemperatures of 12001300 K [27,28]: (a) time averaged CH signal, (b) aninstantaneous CH PLIF image, (c) and (d) show instantaneous formaldehyde andOH PLIF images acquired at the same instant. The white line in (d) represents theouter contour of the formaldehyde signal. Vertical arrows indicate location of thefuel jet.

axial distance x (mm)

heat

rele

ase

rate

q (k

J m

m-1

s-1 )

Case 150% C2H4, 50% H2

0 50 100 150 200 250

4.0

2.0

0.0

Fig. 10. Measured heat release rate distributions from OH and CH images for Case1, ethylene-hydrogen fuel. Conditions are listed in Table 1. Total area under thecurves is Q, which is 402 kJ/s. The black vertical arrow marks the location of the fueljet.

Fig. 11. Measured heat release rate distributions from OH images for four differenthydrogen fuel ow rates are given as solid lines. Linear extrapolation of the datapast the end of the window is given as dashed lines. / = overall fuelair equivalenceratio. Case 2 is the upper curve. Values of Q (total heat release rate) are 350, 292,253 and 224 kJ/s for / = 0.36, 0.30, 0.26, 0.23, respectively. The black vertical arrowmarks the location of the fuel jet.

Flam

e le

ngth

(mm

)

Equivalence ratio

300

250

2000.25 0.30 0.35

Case2

Fig. 13. Measured ame lengths for various fuel ow rates, pure hydrogen fuel.Circled data point is Case 2; conditions are listed in Table 1.

4. The third region contains a turbulent non-premixed ame and

the high incoming air temperature allowed the ame to be sta-

To simultaneously image both the auto-ignition reaction zoneand the primary reaction zone, simultaneous formaldehyde-OH

n anbilized in a cross-ow having the very large velocity of 468 m/s.Upstream autoignition reactions formed formaldehyde in theliftoff region which likely affected the ame speed of the pre-mixed ame base downstream. Because upstream autoignition,a premixed ame base, and downstream thickened ameletswere observed together, we classify this combustion to be anautoignition-assisted ame.

6. Flame lengths were found to increase linearly with fuel massow rate, indicating that mixing is controlled by the air velocityrather than the fuel velocity. This corresponds to a wake-likemixing process, rather than a jet-like mixing, both of whichhave been identied in non-reacting JICF studies.

Acknowledgments

The authors wish to thank Drs. Cam Carter and Mark Gruber ofAFRL and Dr. Chad Rasmussen for their assistance. The rst authoracknowledges fellowship assistance provided by NSF and DoD.

Appendix A. CH, OH and formaldehyde PLIF laser diagnosticssystems

The CH PLIF system is similar to that of Carter et al. [34]. TheQ1(7.5) transition of the B2RX2P(0,0) band of the CH moleculewas excited by a sheet of 390.30 nm laser light. The 390.30 laserbeam was formed by mixing the 616 nm output from a Nd:YAGpumped dye laser (Sirah CSTR-D-24) with the 1064 nm beam fromthe same Nd:YAG laser (Spectra-Physics LAB-150) in a KD P mix-ing crystal. The dye selected was a mixture of Rhodamine 610 andRhodamine 640. The resulting 390.30 nm beam was separatedfrom the 616 nm and 1064 nm beams using a Pelin-Broca prism.The 390.30 nm beam was expanded using a 3:1 Galilean telescopeand a concave cylindrical lens with a focal length of 100 mm. Thecentral 40% of the beam was focused into a sheet using a convexspherical lens with a focal length of 1000 mm. The sheet had an en-ergy of 8 mJ/pulse, a height of 60 mm and a thickness of 300 lm

2the CH images indicate the presence of both thickened andshredded amelets. The 23 mm thickness of each CH layer ismore than 10 times the laminar amelet thickness. In this thirdregion the heat release rate decays slowly in the downstreamdirection, which is typical of a non-premixed ame.

5. The present study differs frommost previous JICF studies in thatlengths, ame structure and other properties were measuredusing CH, OH and formaldehyde PLIF, and CH and OH

chemiluminescence.2. Because the cross-ow air temperature exceeded the autoigni-

tion temperature, the structure of the combustion was foundto differ from previous experiments that employed low temper-ature air. Three distinct regions in this JICF ame were identi-ed. In the liftoff region autoignition reactions occur whichcreate a strong formaldehyde signal. Images indicate that thesereactions are distributed and thus fall in the distributed reac-tion regime of turbulent combustion. However, signicant heatrelease does not occur in this region since the CH, CH, OH, andOH signals were negligible.

3. The second region is the lifted ame base, which has the char-acter of a premixed ame, as evidenced by the PLIF imaging, arapid rise of the local heat release rate, and the previouslyobserved dynamic behavior of the reaction zone.

D.J. Micka, J.F. Driscoll / Combustio(where the intensity dropped to 1/e ).CH uorescence from the AX(1,1), AX(0,0), and BX(0,1)

bands was detected in wavelength range between 420 nm andPLIF diagnostics were used. OH uorescence was obtained by excit-ing the Q1(6) transition of the A2RX2P band with a 283.01 nm la-ser sheet. Fluorescence from the AX(1,1) and (0,0) bands wascollected near 310 nm. The excitation beam was created by fre-quency doubling the 566 nm beam from an Nd:YAG pumped dyelaser with Rhodamine 590 dye. The resulting 283 nm beam had apulse duration and energy of 10 ns and 9 mJ.

Formaldehyde uorescence was obtained by exciting the RR3transition of the 410 vibrational band in the A

1A2X1A1 electronicband near 352.48 nm, as suggested by Harrington and Smyth[39]. Fluorescence was collected over the range of 385470 nm.An Nd:YAG pumped dye laser with LDS 698 dye produced a705 nm beam which was frequency doubled. Each 10 ns pulse con-tained 12 mJ of laser energy at 352 nm. The 283 and 352 beamswere combined using a dichroic mirror before passing themthrough the same sheet forming optics. The sheet thickness wasapproximately 250 lm for the 283 nm beam and 350 lm for the352 nm beam. A pair of Andor Istar intensied CCD cameras wereplaced on opposite sides of the test section as shown in Fig. 5. The352 nm (formaldehyde) sheet was delayed by 150 ns with respectto the 283 nm (OH) sheet in order to minimize interference be-tween signals. To block laser scattered light and ame luminosity,a 310 nm bandpass interference lter was mounted on the OH-PLIFcamera and Schott GG-385 and BG-3 lters were mounted on theformaldehyde-PLIF camera.

References

[1] S. Dhanuka, J. Temme, J.F. Driscoll, J. Propul. Power 27 (3) (2011) 631641.[2] M. Ryan, M. Gruber, C. Carter, T. Mathur, Proc. Combust. Inst. 32 (2009) 2429

2436.[3] D.J. Cook, H. Pitsch, J.H. Chen, E.R. Hawkes, Proc. Combust. Inst. 31 (2007)

29032911.[4] S.H. Smith, M.G. Mungal, J. Fluid Mech. 357 (1998) 83122.[5] E.F. Hasselbrink, M.G. Mungal Jr., J. Fluid Mech. 443 (2001) 125.[6] L.K. Su, M.G. Mungal, J. Fluid Mech. 513 (2004) 145.[7] M.R. Gruber, A.S. Nejad, T.H. Chen, J.C. Dutton, J. Propul. Power 16 (3) (2000)

449457.[8] K.C. Lin, M. Ryan, C. Carter, M. Gruber, C. Raffoul, J. Propul. Power 26 (3) (2010)

503513.[9] L.L. Yuan, R.L. Street, J.H. Ferziger, J. Fluid Mech. 379 (1999) 71104.[10] E.F. Hasselbrink, M.G. Mungal Jr., J. Fluid Mech. 443 (2001) 2768.[11] D. Han, M.G. Mungal, Combust. Flame 133 (2003) 117.[12] L.W. Kostiuk, A.J. Majeski, P. Poudenx, M.R. Johnson, D.J. Wilson, Proc.

Combust. Inst. 28 (2000) 553559.[13] R.F. Huang, J.M. Chang, Combust. Flame 98 (1994) 267278.[14] M. Fairweather, W.P. Jones, R.P. Lindstedt, A.J. Marquis, Combust. Flame 84

(1991) 361375.440 nm. This results in a relatively high uorescence yield and suf-cient separation between the excitation and uorescence wave-lengths to lter out the excitation beam. Fluorescence wascollected by a 16-bit Andor Istar intensied CCD camera and aninterference lter that had center wavelength of 430 nm and aband pass of 10 nm (full-width at half-maximum). A KV418 lteralso blocked scattered light from the pump beam. A f/2.0, 50 mmNikkor lens with a 12 mm extension tube was used to obtain a eldof view of 75 mm by 47 mm. The CCD was binned 3 3 andcropped to obtain an array of super pixels 341 214 with eachsuperpixel covering 220 lm. To remove most of the interferencescaused by ame luminosity from the PLIF images, a second inten-sied camera viewed the test section from the opposite side. It alsowas gated at 15 ns and was triggered 50 ns before the arrival of thelaser sheet and thus recorded only ame luminosity. The ameluminosity from the second camera was subtracted from theimages acquired by the rst camera.

d Flame 159 (2012) 12051214 1213[15] N. Peters, Turbulent Combustion, Cambridge University Press, Cambridge,2000.

[16] J.F. Driscoll, Prog. Energy Combust. Sci. 34 (1) (2008) 91134.

[17] S.M. Torrez, J.F. Driscoll, M. Ihme, M.L. Fotia, J. Propul. Power 27 (3) (2011)545555.

[18] R.W. Grout, A. Gruber, C.S. Yoo, J.H. Chen, Proc. Combust. Inst. 33 (1) (2010)16291637.

[19] A. Ben-Yakar, R.K. Hanson, Proc. Combust. Inst. 27 (1998) 21732180.[20] A. Ben-Yakar, R.K. Hanson, American Inst. of Aeronautics and Astronautics,

AIAA Paper 1999-2453, 1999.[21] M. Gamba, M.G. Mungal, R.K. Hanson, American Inst. of Aeronautics and

Astronautics, AIAA Paper 2011-0319, 2011.[22] R. Cabra, J.-Y. Chen, R.W. Dibble, A.N. Karpetis, R.S. Barlow, Combust. Flame

143 (2005) 491506.[23] R. Cabra, T. Myrvold, J.-Y. Chen, R.W. Dibble, A.N. Karpetis, R.S. Barlow, Proc.

Combust. Inst. 29 (2002) 18811888.[24] C.N. Markides, E. Mastorakos, Proc. Combust. Inst. 30 (2005) 883891.[25] H. Wang, S.B. Pope, Combust. Theory Model. 12 (5) (2008) 857882.[26] C. Olbricht, F. Hahn, A. Sadiki, J. Janicka, Int. J. Heat Fluid Flow 28 (6) (2007)

1215.[27] D.J. Micka, J.F. Driscoll, Proc. Combust. Inst. 32 (2009) 23972404.[28] D.J. Micka, J.F. Driscoll, American Inst. of Aeronautics and Astronautics, AIAA

Paper 2008-5071, 2008.[29] B.O. Ayoola, R. Balachandran, J.H. Frank, E. Mastorakos, C.F. Kaminski,

Combust. Flame 144 (2006) 116.[30] S.A. Filatyev, J.F. Driscoll, C.D. Carter, J.M. Donbar, Combust. Flame 141 (2005)

121.

[31] J.O. Sinibaldi, J.F. Driscoll, C.J. Mueller, J.M. Donbar, C.D. Carter, Combust. Flame133 (2003) 323333.

[32] J.M. Donbar, J.F. Driscoll, C.D. Carter, Combust. Flame 125 (2001) 12391257.[33] A. Ratner, J.F. Driscoll, J.M. Donbar, C.D. Carter, J.A. Mullin, Proc. Combust. Inst.

28 (2000) 245253.[34] C.D. Carter, J.M. Donbar, J.F. Driscoll, Appl. Phys. B 66 (1998) 129132.[35] J.G. Lee, D.A. Santavicca, J. Propul. Power 19 (5) (2003) 735749.[36] R. John, M. Summerfeld, Jet Propul. 27 (1957) 169178.[37] J. Diederichsen, R.D. Gould, Combust. Flame 9 (1) (1965) 2531.[38] J.M. Samaniego, F.N. Egolfopoulos, C.T. Bowman, Combust. Sci. Technol. 109

(1995) 183203.[39] J. Harrington, K. Smyth, Chem. Phys. Lett. 202 (1993) 196202.[40] A. Laskin, H. Wang, C.K. Law, Int. J. Chem. Kinet. 32 (10) (2000) 589614.[41] D.J. Micka, Ph.D. Thesis, Department of Aerospace Engineering, University of

Michigan, 2010.[42] P.H. Paul, H.N. Najm, Proc. Combust. Inst. 27 (1998) 4350.[43] M. Richter, R. Collin, J. Nygren, M. Alden, L. Hildingsonn, B. Johansson, JSME Int.

J. 48 (2005) 701707.[44] R.S. Barlow, R.W. Dibble, J.Y. Chen, R.P. Lucht, Combust. Flame 82 (1990) 235

251.[45] A. Joedicke, N. Peters, M. Mansour, Proc. Combust. Inst. 30 (2005) 901909.[46] D.J. Micka, S.M. Torrez, J.F. Driscoll, American Inst. of Aeronautics and

Astronautics, AIAA-2009-7362, 2009.

1214 D.J. Micka, J.F. Driscoll / Combustion and Flame 159 (2012) 12051214

Stratified jet flames in a heated (1390K) air cross-flow with autoignition1 Introduction2 Experimental arrangement3 Results3.1 Structure of the primary reaction zone (the flame)3.2 Structure of the autoignition (liftoff) zone3.3 More evidence for an auto-ignition assisted flame3.4 Flame lengths and heat release rate profiles

4 ConclusionsAcknowledgmentsAppendix A CH, OH and formaldehyde PLIF laser diagnostics systemsReferences