The original publication is available at

Transcription

1 G.C. Gebel, T. Mosbach, W. Meier, M. Aigner, Laser-Inuce Blast Waves in Air an their Effect on Monoisperse Droplet Chains of Ethanol an Kerosene, Shock Waves 25 (2015) The original publication is available at

2 Laser-Inuce Blast Waves in Air an their Effect on Monoisperse Droplet Chains of Ethanol an Kerosene G. C. Gebel T. Mosbach W. Meier M. Aigner Institute of Combustion Technology, German Aerospace Center (DLR), Pfaffenwalring 38-40, Stuttgart, Germany G. C. Gebel Abstract Weak spherical blast waves in static air an their breakup of ethanol an Jet A-1 kerosene roplets were investigate. The blast waves were create by laser-inuce air breakowns at ambient temperature an pressure. In the first step of this stuy they were visualize with schlieren imaging, an their trajectories were tracke with high temporal resolution. The laser pulse energy was varie to create blast waves of ifferent strengths. Their initial energies were etermine by the application of a numerical an a semi-empirical blast wave moel. In the secon step monoisperse ethanol an kerosene roplet chains were injecte. Their interaction with the blast waves was visualize by the application of shaowgraph imaging. The perpenicular istance of the breakown origin towars the roplet chains was varie to stuy the effect on the fuel roplets as a function of the istance. Droplets within a few millimeters aroun the breakown origin were isintegrate into two to three seconary roplets. The blast-inuce flow velocities on the post-shock sie an the corresponing Weber numbers were calculate from the ata of a non-imensional numerical simulation, an a close look was taken at the breakup process of the roplets. The analysis showe that the aeroynamic force of the blast-inuce flow was sufficient to eform the roplets into isk-like shapes, but iminishe too fast to accomplish aeroynamic breakup. Due to the release of strain energy, the eforme roplets relaxe, stretche into filaments an finally isintegrate by capillary pinching. Keywors Blast wave, Laser-inuce breakown, Droplet breakup, Flow visualization, Schlieren, Shaowgraph 1 Introuction Avance combustion systems are neee for moern aviation gas turbines to achieve high efficiency an low emissions. Their evelopment can benefit significantly from tailor-mae numerical tools.

3 Compare with iterative prototype testing, numerical simulations allow parameter surveying an geometry optimization with faster turnaroun times an at much lower cost. For fuel spray ignition, however, numerical tools have not yet mature. Spray ignition is a complex process involving multiple mechanisms relate to the research fiels of flui ynamics, two-phase flows, heat transfer, evaporation, chemical kinetics, an plasma physics. With the limitations of present numerical tools, the transition from an ignition spark into a flame kernel is only computable if the real physics are strongly simplifie. Therefore, a better unerstaning of the involve mechanisms is necessary to evelop more accurate moels. Experiments in laboratory setups with well-efine bounary conitions can provie very important contributions. Combustion engines typically ignite by electrical spark ischarge. But spark plugs are inconvenient in laboratory setups for the following reasons. Shifting of the breakown position is restricte. Bounary conitions are complicate, because the plug acts as a heat sink an interacts with the flow fiel. Variation of the breakown energy is restricte, an the triggering accuracy is low. Ignition by non-resonant laser-inuce breakown is a goo alternative, as it is avantageous in all the state aspects above. Both ignition concepts feature similar physical properties, particularly the electron cascae mechanism [1], an involve the formation of a blast wave, as emonstrate in Figure 1. It shows schlieren images from a laserinuce breakown, an inuctive an a capacitive spark ischarge in air. The ensity graients at the blast wave shock fronts are clearly visible. In our previous publications we investigate the features of laser-inuce breakowns an the ignition of liqui fuels [2-4]. We foun that spherical blast waves forme by the rapi initial expansion of the breakown plasmas play an important role in the ignition process, because they inuce seconary roplet breakup [3]. Inee, the seconary breakup of roplets in high velocity flow fiels has been extensively investigate over almost seven ecaes, an reviews are given in the papers by Pilch an Erman [5], Gelfan [6] an Guilenbrecher et al. [7]. Experiments were conucte in ifferent types of test facilities, incluing shock tubes, rop towers an blowown win tunnels. The paper by Guilenbrecher et al. inclues a review on the specific characteristics of the ifferent types of test facilities. It is necessary to emphasize that the flow conitions in ifferent experiments varie significantly, an consequently ifferent roplet behaviors an breakup mechanisms were ientifie. The most isruptive breakup mechanisms were observe at supersonic flow conitions, see for instance the publications [8-12], where roplets were expose to steay blast-inuce flows for up to several millisecons. Although we also investigate roplet breakup in a blast-inuce flow in this stuy, the basic conitions are ifferent. In the Eulerian frame of reference, the blast-inuce flow on the post-shock sie of a spherical blast wave is not steay, but ecays rapily an eventually turns into a weak rarefaction wave. In the Lagrangian frame of reference, the expansion velocity of the shock front converges towars sonic velocity, the wave turns into an acoustic wave. Summarizing, roplets expose to spherical blast waves face highly transient flow conitions, an apart from our previous paper [3], we are not aware of any other stuy regaring the effect of spherical blast waves on roplets. Our previous paper provie microscopic images of kerosene roplets in the vicinity of a laser-inuce breakown. The breakup by the blast-inuce flow was observe 10 mm below the

4 breakown origin with a very high spatial an temporal resolution. The investigation presente in this paper continue our previous work. The focus was on the effect of laser-inuce blast waves on fuel roplets at ifferent raial istances from the breakown origin. The investigation was ivie into three parts, which are reflecte by the structure of the paper. In the first part, the trajectories of spherical blast waves were tracke with high temporal resolution by schlieren image sequences. Four ifferent methos to etermine the energies from the trajectories were teste. Two of them were selecte an valiate on blast waves of ifferent strengths. The investigation provie the initial blast wave energy E 0, which was require to calculate the roplet Weber numbers in the blast-inuce flow on the post-shock sie. In the secon part, the blast wave interaction with monoisperse roplet chains of ethanol an Jet A-1 kerosene was investigate with shaowgraph imaging. The blast wave strength was kept constant at a high level. Different geometrical configurations were investigate, incluing single roplet chains an five chains in a row at ifferent horizontal istances from the breakown origin. Finally, calculations of the blast-inuce flow velocities an Weber numbers were performe on the basis of the etermine blast wave energy. This allowe a comparison of the observe roplet breakup with conventional breakup moels. Enlarge views of roplets at ifferent horizontal istances from the breakown origin gave insight into the mechanism behin the observe breakup. Fig. 1 Exemplary Schlieren images, taken several microsecons after a laser-inuce breakown (left), an inuctive (center) an a capacitive spark ischarge (right) in ambient air. The blast waves are clearly visible. 2 Experimental setup 2.1 Test rig an injector Experiments were performe in a vertically arrange flow channel with a square cross section of 62 cm 2 an a length of 1 m. A schematic an a cutaway are shown in Figure 2. Three sies were equippe with winows with anti-reflective coating. The fourth sie was attache to a two-axis

5 traverse system for horizontal an vertical positioning. A ownwars irecte roplet chain injector moel TSI 3450 was installe at approximately one fourth height of the flow channel. Droplets were generate by the inuce Plateau-Rayleigh capillarity instability mechanism [13]: A piezoelectric actuator inuce a sustaine regular istortion on the fuel jets which streame out of the orifice holes. Two orifices were available, one with a centere 50 µm hole, an one with five coplanar 50 µm holes space at a centre-to-centre istance of 1 mm. Accoring to Rayleigh [13], the excitation wavelength to obtain a monoisperse roplet breakup is given by: 4.508D j. (1) In Equation (1), is the excitation wavelength an D j is the iameter of an unisturbe fuel jet. For incompressible fuels the jet iameter can be replace by the hole iameter D h. The excitation frequency is then given by u jet (2) 4.508D h with the fuel jet velocity u jet 4m f xd f 2 h. (3) In Equations (2) an (3), m f an f are the total fuel mass flow for all holes an the fuel ensity, D an the center-to-center respectively. The number of holes is given by x. The roplet iameter istance between two successive roplets can be calculate by geometrical means: D 6m f f x 1 3, (4) 2D. (5) 3D 3 2 h The fuel was supplie by a siphon bottle with an inner an a circumferential chamber. The inner chamber containe the fuel, the circumferential chamber was perfuse by a water flow with a controlle temperature. Thus, the fuel temperature coul be controlle with an accuracy of ± 0.2 K.

6 Inflowing air isplace the fuel an rove it to the injector. The flow channel coul be vente by a steay top own air flow, in orer to flush out fuel vapour to prevent unwante ignition an uncontrolle combustion. The air flow an fuel vapour were exhauste sieways at the bottom of the flow channel. Liqui fuel was capture with a funnel an collecte in a steel bottle at the very bottom. Fig. 2 Schematic of the flow channel, outsie view (left) an cutaway (right). 2.2 Laser-inuce air breakowns Breakowns were create by a flash lamp-pumpe, Q-switche an frequency-ouble N:YAG laser (moel InnoLas SpitLight ). It provie Gaussian (spatial an temporal) pulses of up to 300 mj an a trigger accuracy of ± 1 ns in the single pulse moe. The pulse energy was ajuste by means of the elay between the flash lamp an the Q-switch. The beam path is illustrate in Figure 3.

7 Laser pulses were steere to an optical bench with a periscope an expane to a iameter of approximately 40 mm with a Galilean telescope. Afterwars, they were irecte into the flow channel with a ownwar angle of 23 egrees an refocuse irectly in front of the inlet winow by a lasergrae oublet lens with a focal length of 120 mm. Non-resonant multiphoton ionization at the focal point generate electron cascae breakowns with high ionization levels. This process is escribe in more etail by Ronney [1]. The initial supersonic expansion of the breakown plasma cause the formation of a spherical blast wave, which etache from the breakown after approximately 1 to 2 µs. 2.3 Schlieren an shaowgraph setup A schematic of the experimental setup is given in Figure 3. The light beam from a halogen projector lamp was collimate with a plano-convex lens ( f 300 mm ) to a iameter of approximately 90 mm. It passe through the observation section insie the flow channel an was refocuse by another plano-convex lens ( f 600 mm ) on the opposite sie. A monochrome Photron / LaVision High Spee Star 6 high-spee CMOS camera recore schlieren images at a sustaine repetition rate of 25 khz, an exposure time of 1 µs an a resolution of pixels. The camera gate function is shown in Figure 4. It features a fast rising ege, a slower falling ege an an overall time span of 1.09 µs. The temporal jitter is approximately ± 40 ns. The 50 % effective exposure time is the time when half of the integral over the gate function has passe, which is 510 ns. This information is particularly important for the blast wave energy etermination in section 3.1. Two ifferent camera lenses were use: A Tokina 100 mm f/2.8 lens (for schlieren images of blast waves) an a Nikkor 180 mm f/2.8 lens (for shaowgraph images of roplet chains). The camera an the laser were externally triggere by a combination of a BNC 565 an a BNC 555 pulse generator. The BNC 565 gave the continuous trigger signals for the camera controller an for the 10 Hz laser flash lamp. The BNC 555 was synchronize with the BNC 565 an gave the singular trigger signals for the laser Q-switch an the camera start. The repetition rate of 25 khz was not sufficient to capture the very fast flui mechanical processes in a single recoring. Therefore, multiple recorings were performe, an the elay between the Q-switch trigger an the camera start was sequentially increase. The Schlieren effect was generate with an iris aperture, which was ajuste to cut with one lamella into the focal point from one sie, in orer to visualize vertical ensity graients. An OG570 longpass filter in front of the aperture suppresse chromatic aberration. The majority of the breakown raiation was blocke by the aperture, but the remaining raiation coul still saturate the camera s CMOS sensor. Therefore, broaban transmission filters were mounte in front of the camera lens for the recorings between 2 an 8 µs after the laser pulse. The transmission of the filters range from to 0.5. For shaowgraph imaging, the aperture an the filters were remove from the light beam path. The camera was shifte closer to the flow channel to obtain a greater magnification. In contrast to the

8 schlieren setup, the shaowgraph setup visualize fuel roplets with great sharpness, while blast waves were not clearly visible. The raiation from the breakown plasma was blocke with a small screen to protect the CMOS sensor from amage. The pressure, temperature an humiity of the ambient air were regularly measure uring the experiment. Variations were very small, the temperature range between 22.6 an 23.8 C, the pressure between an hpa, an the humiity between 27.5 an 28.4 %. Consequently, the calculate sonic velocity range between an m/s. The ata were applie in the calculations of the blast wave energies an blast-inuce flow velocities. Fig. 3 Experimental setup with laser beam path (green) an light beam path for schlieren an shaowgraph imaging (yellow). Fig. 4 Camera gate for a configure exposure time of 1 µs.

9 3 Measurements, post-processing, an analysis 3.1 Laser-inuce blast waves in air Laser-inuce blast waves in ambient air were investigate with the schlieren imaging technique. Three recorings were taken for every selecte laser pulse energy an elay, in orer to account for statistical scattering. The intention of this investigation was to ientify a metho to etermine accurate blast wave energies from the expansion trajectories. Blast waves of five ifferent strengths were create by varying the laser pulse energy. The pulse energy was ajuste by means of the elay between the flash lamp an the Q-switch trigger: A elay of 215 µs provie the maximum laser pulse energy, whereas a elay of 335 µs provie just enough energy to create laser-inuce breakowns at the focal point. The ajuste elays were 215, 255, 295, 315, an 335 µs. Only a fraction of the laser pulse energy was absorbe in the breakown. During the first few nanosecons of the pulse, many of the photons passe the breakown region, because the plasma was partially transparent an i not fully occupy the focal point, see Chen et al. for more etails [14]. Therefore, the breakown energy was etermine in a two-step measurement with a volume absorber type Gentec-EO UP19K-15S-VR- D0. Average energies over 2000 pulses were separately measure 20 mm in front of an behin the focal point. Energy losses ue to scattering an iffraction are negligible in laser-inuce breakowns [15]. Therefore, the ifference between the pulse energy E p an its transmitte fraction behin the breakown E t virtually equale the breakown energy E b. Table 1 gives the average pulse an breakown energies for five ifferent Q-switch elays t, along with the FWHM (full with at half maximum) pulse lengths an the stanar eviations s b of the breakown energies. Minor FWHM fluctuations between iniviual pulses occurre for Q-switch elays of 315 an 335 µs, which are inicate by the given accuracy of ± 1 µs. The small stanar eviations inicate that the total pulse energies were not significantly affecte by the fluctuations. Representative schlieren images for five ifferent breakown energies 10, 24, an 48 µs after the laser pulse are shown in Figure 4. Spherical blast waves are clearly visible as circular ensity graients. The white spot at the origin is the breakown plasma, which is visible for several microsecons. The breakown is surroune by a plume of compresse, hot gas. After the breakown has ecaye, the gas plume expans slowly an mixes with the surrouning air. Blast wave trajectories were etermine from the schlieren images with a MATLAB algorithm. It create synthetic blast wave images with various raii an cross-correlate them with the schlieren images. With regar to the high expansion velocities of the blast waves, the camera exposure time of 1 µs was relatively long. Consequently, the shock fronts appeare somewhat blurre in the schlieren images. This challenge was overcome by the interpolation of the raius at the 50 % effective exposure time from the cross-correlation peak istribution. Consequently, the blast wave raii were etermine with a theoretical accuracy of ± 0.05 mm, which was about ± ½ pixel of the image resolution. The resulting

10 trajectories are shown in Figure 5. They are averages over three atasets. Error bars are not isplaye, because the reproucibility was very goo. The low curvature of the trajectories inicates that the blast waves were observe uring their transitions from strong blasts into acoustic waves. Literature provies several expansion moels which allow the etermination of the initial energy E 0 from the trajectory. Four moels were applie to the blast wave trajectory of the mj breakown. A wiely use moel is the Taylor-Seov solution, which is given as [16]: r t Et K 0. (6) In Equation (6), r is the raius, t the time, an 0 the ensity of the unshocke air. The constant K is for 1.4 [17]. In the experiment, blast waves were investigate as they expane into ambient air at atmospheric conitions. The only effect that coul have heate the air in the vicinity of the breakown an therefore affecte is the absorption of plasma raiation emitte by the breakown. This effect is negligible at the investigate istances from the breakown surface, as emonstrate in our recently publishe numerical simulation of a laser-inuce air breakown, see Joarer et al. [4]. Consequently, 1.4 was applie in all analyses in the course of this stuy. Equation (6) is a self-similar solution to the blast wave problem. It assumes a very high pressure graient over the shock front, so that the pressure of the unshocke air p 0 is set to zero in the Rankine-Hugoniot relations [18]. This assumption is sufficient only for a short time. A spherical blast wave ecelerates with its increasing raius, an the pressure an temperature graients across the shock front ecrease. This is cause by the geometrical growth of the spherical shock front an the associate energy ensity ecrease, an by issipation. The latter is particularly relevant at the early stage, when the pressure graient is still high. Consequently, p 0 becomes relevant within several nanosecons, an Equation (6) preicts an unnatural ecrease of the shock front velocity into the subsonic regime. This is emonstrate by the blue graph in Figure 6: The Taylor-Seov solution was fitte to the measure trajectory with an in-house LabVIEW algorithm, applying the metho of least y -squares. The agreement was of poor quality, an the Taylor-Seov solution was obviously not applicable for the present measurements. Actually it is only accurate for strong blasts, such as nuclear etonations [17]. The laser-inuce blast waves create in our experiments were only strong for several nanosecons, an the measure trajectories covere the transition from strong blasts into acoustic waves. This is well illustrate in Figure 6 by the fact that the trajectory is nearly parallel to the line for a Mach number of Ma 1. Therefore, three self-similar moels were applie which were evelope for the transition regime: The moels by Seov [18] an Broe [19] are numerically erive, non-imensional trajectories. The moel by Jones [20] is a set of semi-empirical equations

11 for planar shocks an cylinrical an spherical blast waves. Jones equations for spherical blast waves are: xt 2 1, (7) t, (8) c0 r 0 xt r t, (9) r E 0 r0. (10) B p0 Here, is the non-imensional time, t is the imensional time, r is the raius, r 0 is a reference raius, c 0 is the sonic velocity, an is the isentropic exponent. B is a geometry parameter, values for ifferent cases are given in [21]. The moels were iteratively fitte to the measure trajectory for a breakown energy of mj with an in-house LabVIEW algorithm. The increment of the iterations was 0.01 mj, the best fits are isplaye in Figure 6. The qualities of the fits are inicate by the sums of the least y -squares. Seov s moel gave a blast wave energy of mj an a relatively high 2 y i,seov of mm 2. In contrast, Jones moel gave a 2 y i,jones of 7.73 mm 2 an Broe s moel 2 gave a y i,broe of only 4.1 mm 2. Consequently, only the moels of Jones an Broe were applie to the trajectories of the four weaker blast waves. The blast wave energy etermination from the expansion trajectory was highly sensitive to measurement inaccuracies, an therefore a careful error analysis was performe. Possible error sources were the time assignments an the calibrate scales of the schlieren images. The temporal accuracy was etermine to be better than ± 40 ns an the scale accuracy to be better than ± 0.1 %. The results of the energy etermination an error analysis are given in Table 2 an isplaye in Figure 7. Table 2 gives the blast wave energies together with their maximum eviations, the blast wave energy proportions in relation to the breakown energy an the sums of the least y -squares. Error bars in Figure 7 inicate the maximum eviations. Concluing, the blast waves consume a significant amount of the breakown energies, which is approximately 52 % at a breakown energy of mj an steaily increases to 77 % at a breakown energy of 11.6 mj. The moels of Broe an Jones feature a very goo agreement for all

12 investigate breakown energies an o not isagree by more than 4.2 % (etermine for E b = 89.6 mj). Therefore, we regare the two moels as successfully valiate within our experimental possibilities, an assume that a blast wave energy of 52 % for a laser pulse energy at ~246 mj was sufficiently close to reality to use it in our subsequent analysis, see section 3.3. Our results are in goo agreement with the stuy by Brieschenk et al. [22], who investigate the blast wave expansion from laser-inuce breakowns create by a Q-switche ruby laser an a focusing lens with a focal length of 100 mm. The laser pulse energy range between 100 an 2700 mj. Blast wave energy fractions were etermine by the numerical integration of the Rankine- Hugoniot relations using Dewey s empirical trajectory moel [23]. They were foun to be 50 to 65 % of the laser pulse energies. Moreover, our results are in very goo agreement with the stuy by Phuoc an White [24]. They investigate the blast wave expansion from laser-inuce breakowns create by a Q-switche N:YAG-laser at 1064 nm an a focusing lens with a focal length of 120 mm. They etermine blast wave energies uring the first few microsecons after the laser pulse using the Taylor-Seov solution. Breakown energies range between 15 an 50 mj. At 50 mj, the blast wave energy was 57 % of the breakown energy. Its proportion increase with ecreasing breakown energy, an 70 % were etermine at 15 mj. The tren an the orer of magnitue are very similar to our results. In particular, they etermine a blast wave energy ratio of 61.6 % for a breakown energy of 44 mj, while we etermine 62.6 % (Broe s moel) an 64.9 % (Jones moel) for a breakown energy of 46.1 mj. Our experiment an analysis presente in this section were require for the analysis of roplet breakups in section 3.3, because Phuoc an White applie maximum breakown energies of 50 mj, while we investigate roplet breakups at breakown energies of almost 250 mj. Tab. 1 Average laser pulse energies, estimate breakown energies, breakown energy stanar eviations over 2000 laser pulses an laser pulse FWHM length. Laser Q-switch t [µs] Laser pulse FWHM [ns] 14.3 ± 1 10 ± Laser pulse E p [mj] Air breakown E b [mj] Air breakown s b [mj]

13 Fig. 4 Representative schlieren images of laser-inuce blast waves in air for five ifferent air breakown energies. The black shape in the upper area is the holer of the oublet lens. The origin is at the focal point of the breakown. Fig. 5 Tracke blast wave trajectories for five ifferent breakown energies.

14 Fig. 6 Comparison of a measure blast wave trajectory with fitte trajectory moels. Tab. 2 Average breakown an blast wave energies, an sums of least squares of fitte trajectory moels. The energy proportions (100 E 0 E b ) are given in brackets. E b E 0,Broe 2 y i,broe [mm 2 ] E 0,Jones 2 y i,jones [mm 2 ] [mj] [mj (%)] [mj (%)] ± 3.2 (52.2 ± 1.3) ± 2.7 (51.0 ± 1.1) ± 2.5 (53.7 ± 1.4) ± 2.2 (54.2 ± 1.2) ± 1.4 (56.2 ± 1.6) ± 1.4 (58.2 ± 1.6) ± 0.9 (62.3 ± 2.0) ± 0.9 (64.9 ± 2.0) ± 0.4 (76.6 ± 3.4) ± 0.4 (77.7 ± 3.4) 5.13 Fig. 7 Average blast wave energies an proportions (100 E 0 E ) for five ifferent breakown energies. The error bars give the maximum errors within the accuracy of the measurements. b

15 3.2 Interaction of blast waves with roplet chains In the secon part of the experimental investigations, monoisperse ethanol an Jet A-1 kerosene roplets chains were injecte into a weak an steay airflow insie the flow channel. The fuel temperature was controlle to be 294 ± 0.2 K. The air flow was at ambient pressure an temperature, an its mass flow was set to be ± 10.3 g/min. The purpose of the airflow was to flush fuel vapour out of the measurement area. Although roplet evaporation was negligible at the applie moerate temperatures, the airflow was a precaution to prevent the formation of an ignitable mixture uring the experiment. The elay between the laser flash lamp trigger an the Q-switch trigger was kept at 215 µs, proviing the maximum laser pulse energy throughout all measurements. The interaction of the roplet chains with the blast waves was visualize with schlieren an shaowgraph imaging. To obtain a better temporal resolution than the 25 khz of the high-spee camera, recorings with ifferent elays between the camera start trigger an the laser Q-switch trigger were taken for every investigate experimental configuration. Schlieren an shaowgraph images were recore for several configurations. Single an five parallel roplet chains of ethanol an a commercial Jet A-1 kerosene (from the international airport of Stuttgart, Germany) were investigate, an the horizontal istance towars the breakown position was varie to stuy the effect on the fuel roplets as a function of the blast wave raius. The parameters of the applie roplet chains are given in Table 3. The require excitation frequency TSI of the injector an the jet velocity v jet were calculate with the equations (2) an (3). The roplet iameter D an the center-to-center istance between two successive roplets were calculate with the equations (4) an (5). Particularly revealing images were taken for five parallel roplet chains, which were positione between 5 an 9 mm horizontal istance from the breakown position. Figure 8 shows three representative schlieren images of kerosene roplet chains 20, 40 an 200 µs after the laser pulse. The horizontal istance between the focal point an the closest roplet chain was 5 mm. At 20 µs, the plasma raiation from the breakown is still visible. The laser-inuce blast wave has alreay passe the roplet chains. At 40 µs, the gas plume has a raius of approximately 4 to 5 mm an just approaches the first roplet chain. At 200 µs, the plume continues its expansion an isturbs sections of the first an secon roplet chain, marke by the white ashe box. A weak horizontal wave is visible, which is the reflection of the blast wave on the injector plane. The regimes, in which roplets flatten an eventually break up are marke by yellow ashe boxes. They will be in the focus of the forthcoming iscussion an analysis in this paper, see section 3.3. Figure 9 shows shaowgraph images of five parallel kerosene roplet chains recore 40, 140, 200, an 300 µs after the laser pulse. The experimental configuration is exactly the same as in Figure 8, but the fiel-of-view is smaller, an ensity graients are not visible. The expansion of the gas plume is inicate by a blue circle. All four images show the roplet chains after being struck by the blast

16 wave. At 40 µs, the roplets have eforme to isks. Subsequently they relax an turn into filaments, which are clearly visible at 140 µs. Filaments at the 5 to 8 mm horizontal positions show pinchings at 200 µs, which inicate the beginning breakup. Some roplets at the 5 to 7 mm positions show two pinchings, while the roplets below an roplets at the 8 mm position show only one pinching. At the 5 to 7 mm positions, roplets have isintegrate into two to three seconary roplets at 300 µs. No breakup occurs at the 8 mm position. The roplets have recovere from the pinching, an the upper roplets have contracte to isks. Droplets at the 9 mm position o not show any breakup, but relax from isks into filaments an eventually become spherical again. The ifferent observe eformations an breakups epen on the raial istance from the breakown origin: Breakup into three seconary roplets occurs within a raius of 7.4 ± 0.1 mm, an breakup into two seconary roplets occurs within 8.0 ± 0.1 mm. The raii were etermine from images of six iniviual events, taken 300 an 320 µs after the laser pulse. They are inicate by the ashe re circles in the right image of Figure 9. Their origin is 1.7 mm below the breakown position, which is the istance the roplets have roppe since the blast wave has passe. Breakup oes not occur in the upper part of the roplet chains at 5 to 7 mm istance, whereas it occurs at the lower part. The upper roplets are impacte by the flow fiel of the expaning gas plume. This apparently overries their breakup. Instea, it inuces istortion an partly coalescence or atomization. Figure 10 shows similar images, but with ethanol instea of kerosene. The same process is visible, with one exception: Not all roplets at the 8 mm position recover from the contraction at 200 µs but isintegrate into two seconary roplets, which are visible at 300 µs. Breakup into three seconary roplets occurs within a raius of 7.7 ± 0.1 mm, an breakup into two seconary roplets within 8.5 ± 0.1 mm. The greater raii in comparison to kerosene probably resulte from the lower surface tension of ethanol, which is 22.2 mn/m at 21 C [25]. The surface tension of Jet A-1 kerosene is not efine in the fuel specification DEF STAN [26], an literature offers ifferent values. For a temperature of 21 C, the CRC Hanbook of Aviation Fuel Properties provies 23.6 mn/m [27], while Rachner provies 22.8 mn/m [28]. Therefore, a sample of our Jet A-1 was sent to an external chemical analysis laboratory. The surface tension was measure to be 25.9 mn/m, which was greater than the two literature values. But it still remaine within the normal range for kerosenes an similar fuels. It was 16.7 % greater than the surface tension of ethanol, an thus kerosene roplets are more resistant towars aeroynamic breakup. Figure 11 shows shaowgraph image overlays of single roplet chains 300 µs after the laser pulse. Images were recore for ethanol an for kerosene. The ifferent colors emphasize that each image is an overlay of seven iniviual recorings. The intention of this part of the experiment was to examine if any slipstream effects occur in the configuration with five parallel chains: Droplet chains might iminish the blast-inuce flow or the gas plume expansion an consequently influence the breakup of the chains that lie behin. Regaring the gas plume, this was the case: In Figures 9 an 10 the breakup at 7 mm is not isturbe, contrary to Figure 11. There, the upper part of the chain only shows sporaic breakup, while most of the roplets stay intact an ae partly istorte. Hence, the breakup at the 7 mm position in the configuration with five chains was not isturbe, because the chains at 5 an

17 6 mm iminishe or even blocke any isruptions by the gas plume. In contrast, the effect of the blast waves on the roplets was the same as in the configuration with one chain. In Figures 10 an 11 breakup into two seconary roplets occurs at the upper part of the ethanol roplet chains at 8 mm. In Figures 9 to 11 breakup into three seconary roplets occurs in the lower part of the kerosene an ethanol roplet chains at 6 mm. In conclusion, the only relevant slipstream effect in the configuration with five chains was the iminishing of the expaning gas plume. Hence, the configuration with five chains was avantageous for the observation of blast wave effects on fuel roplets. The next section provies a closer look at the observe roplet breakup, along with an analysis of the unerlying mechanism. Tab. 3 Controlle ( m, TSI ) an calculate ( D, ) parameters of the applie roplet chains. f Jet A-1 Jet A-1 Ethanol Ethanol One roplet chain Five roplet chains One roplet chain Five roplet chains m [g/min] f [khz] TSI D [µm] [µm] Fig. 8 Schlieren images of five parallel kerosene roplet chains nearby a laser-inuce air breakown. The origin is at the focal point of the breakown.

18 Fig. 9 Shaowgraph image of five parallel kerosene roplet chains. The origin is at the focal point of the breakown. Fig. 10 Shaowgraph image of five parallel ethanol roplet chains.

19 Fig. 11 Overlays of shaowgraph image of single kerosene an ethanol roplet chains. The ifferent colors emphasize that each image is an overlay of seven iniviual recorings. 3.3 Analysis The shaowgraph images presente in section 3.2 emonstrate that the blast waves inuce roplet breakup. Droplets isintegrate into two to three seconary roplets within a raius of 7.9 to 8.4 mm aroun the breakown origin. As the raius increase, the effect iminishe. Above 8.4 mm roplets oscillate but no breakup occure. The Figures 12 an 13 provie enlarge views at 7, 8 an 9 mm horizontal istance from the breakown origin for times between 20 an 340 µs after the laser pulse. Please note: The image sequences show fuel roplets after the blast-inuce flow has passe. The analysis in this section shows that in the Eulerian frame of reference the flow ecays rapily. At 20 µs after the laser pulse aeroynamic forces, characterize by the Weber number We, are alreay too weak to have any impact on the roplets. The views in Figures 12 an 13 shift ownwar by 2 mm an leftwar by 0.1 to 0.2 mm over time. The shift compensates gravitation-inuce escent of the roplets an their leftwar motion initiate by the aeroynamic rag of the blast-inuce flow. Hence, the enlarge views only show roplets which were approximately at the height of the breakown when they were hit by the blast wave. The roplet eformation an breakup occur as escribe in section 3.2. Droplet breakup into three seconary roplets occurs at the 7 mm position. At the 8 mm position ethanol roplets break up into two seconary roplets, while kerosene roplets remain intact but oscillate linearly. At the 9 mm position, ethanol an kerosene roplets oscillate linearly but remain intact. Linear oscillation an roplet breakup take place along a nearly horizontal axis. This results from the local expansion irection of the blast wave, which hit the roplets from the left an proceee to the right. Several recorings with ifferent elays between the Q-switch trigger an the camera start were require to overcome the limite camera repetition rate. The sequences shown in Fig. 12 an 13 were put together from three iniviual recorings of the experiment. Although the reproucibility was

20 very goo, small ifferences coul occur. Please see for instance the roplet filaments at the 7 mm position at 200 µs in Fig. 13. Their tilt angles iffer from their neighbors at 180 an 220 µs (both taken in the same recoring) by 4.5 egrees. This oes not result from a rotational oscillation but from small, unientifie ifferences between the two recorings of the same experiment. However, no ifferences between the transient morphologies of the eformations an breakup processes were observe, an they are well represente by the sequences. The enlarge views assist the analysis of the unerlying mechanism. On the post-shock sie, blast waves feature strong, but quickly ecreasing blast-inuce flows. At present, there is no analytical metho available to characterize roplet eformation an breakup in such flows. However, the application of the methos evelope for steay flows can provie insightful information. Droplet breakup in steay flow fiels is commonly characterize by the Weber number We an the Ohnesorge number Oh. The impact of the latter is negligible if Oh 0.1 [7], which is true for the herein presente experiments. The Weber number is efine as 2 ug gd We. (11) Here, u g is the velocity of the gas flow fiel, g is the ensity of the gas, an is the roplet surface tension. It is very ifficult to measure blast-inuce flow velocities, but they can be numerically etermine when the blast wave energy is known. The energy etermination is reporte in section 3.1. It shows that the blast waves absorb approximately 52 % of the breakown energy at the energy regime applie in the experiments presente in section 3.2. One of the successfully applie blast wave moels in section 3.1 is the non-imensional numerical simulation by Broe [19]. Another simulation by Broe provies non-imensional ensities an velocities of blast-inuce flows [29]. The iagrams from his paper were igitize, an the transient flow velocities an Weber numbers at 5, 6, 7, 8 an 9 mm raial istance from the breakown origin were interpolate. A blast wave energy of mj was applie, which is the energy etermine with Broe s moel in section 3.1. Figure 14 gives the flow velocities for the five investigate raii, an Figure 15 gives the corresponing roplet Weber numbers. The latter were calculate for ethanol an kerosene roplet iameters of 94.5 an 94.3 µm, respectively, accoring to Table 3. Primarily ue to the ifferent surface tensions, kerosene Weber numbers are smaller by approximately 17 % compare to ethanol Weber numbers. The iagrams also show that the post-shock flow velocities ecrease towars zero within approximately 10 µs after the shock front has passe. The subsequent negative flow velocities inicate the rarefaction waves, which are consierably weaker an last for approximately 25 µs. Pilch an Erman [5] reviewe experimental stuies of roplet breakup in steay flow fiels. They ientifie five ifferent breakup moes, which are etermine by the magnitue of the Weber number. Accoring to Guilenbrecher et al. [7], the extremes are the vibrational breakup for We 11, where roplets isintegrate ue to resonant excitation, an the catastrophic breakup for We 350, where

21 roplets atomize into very small seconary roplets. Several breakup mechanisms exist in between, such as the bag breakup an the sheet thinning breakup. Experimental an numerical stuies consistently show that for We 11 roplet breakup in steay flows typically begins with a eformation into a flat shape, such as a isk, lens or bowl, see for instance the publications [10,12,30,31]. Extensive atomization of the surfaces of still spherical roplets occurs only for very high Weber numbers at the orer of A comparison of the known breakup mechanisms with Fig. 12 an 13 provies an unerstaning of the observe process. Table 4 provies supporting information. It gives the blast wave arrival time after the laser pulse an the Weber numbers 0.1 µs later. At all three positions, 7 to 9 mm, the Weber numbers are initially well above 11 an therefore, for a short time, aeroynamic forces are sufficient to eform the roplets into flat shapes. The result of this aeroynamic flattening can be seen clearly at 20 µs in Fig. 12 an 13. The flattening has proceee at 40 µs. At this time, the flow fiel has alreay turne into a weak rarefaction wave, accoring to Figure 14. In the associate Fig. 15 the Weber numbers are well below 1, aeroynamic forces are too weak to have any impact on the roplets. Thus, the proceeing flattening of the roplets results from previously transmitte kinetic energy from the blast-inuce flow, which now turns into strain energy. The following sequences in Fig. 12 an 13 are ifferent for the two fuels an the iniviual horizontal positions. At the 9 mm position both, kerosene an ethanol roplets o not isintegrate, but oscillate. The angular funamental frequencies of a spherical roplet surroune by a gaseous meium are [32] n1nn1n2 n n1 R f 0, n 3 g f 1 2, (12) where R is the roplet raius an n 2 is the funamental moe. For n 2 roplets feature a linear, axisymmetric oscillation at the orinary funamental frequency 0,2 6 f g 3 f R 1 2. (13) For the fuel ensities 3 f,kerosene = kg/m an 3 f an the air ensity of,ethanol = 789,8 kg/m 3 g 1.21 kg/m the orinary frequencies calculate to be 0,kerosene 0,ethanol = Hz. The frequencies correspon to perios of 0,kerosene T 0,ethanol = Hz an T = µs an = µs. At the 9 mm position in Fig. 12, kerosene roplets are flat at 40 µs, but subsequently expan into spheres at 80 µs. The expansion procees an is followe by a contraction.

22 Droplets again feature a spherical shape at 200 µs, followe by another flattening. Finally, at 340 µs they are almost spherical again. It seems that a secon linear expansion just begins, which means that the spherical shape is reache shortly before 340 µs. Shaowgraph images are available every 20 µs, therefore perios are etermine with an accuracy of ± 10 µs. A spherical roplet shape at 80, 200 an < 340 µs inicates a perioic length of approximately 120 ± 10 µs. The same analysis for ethanol roplets inicates a perioic length of 140 ± 10 µs with a spherical roplet shape at 80 an 220 µs. The agreement with the calculate funamental frequencies is very goo. It inicates the existence of a transient oscillation, which results from a preceing external impulse excitation. The image sequence from 180 to 220 µs at the 7 an 8 mm position in Figures 12 an 13 inicates that capillary forces play a crucial role for the roplet breakup. After having expane linearly, roplets feature filament-like shapes. The filaments have iameters of only several tens of microns an therefore are prone to capillary pinching. At the 8 mm position, single pinchings occur at the centers of the filaments. Kerosene roplets recover from the pinchings an continue to oscillate without isintegration. Due to their lower surface tension, ethanol roplets at the 8 mm position isintegrate an form two seconary roplets with a uniform iameter, see Fig. 13 at 340 µs. At the 7 mm position, both kerosene an ethanol roplet filaments feature two pinchings, resulting in the pinch-offs of two seconary roplets after 220 µs. A comparison of Fig. 12 an 13 at 200 µs shows that the ethanol filaments stretch more than the kerosene filaments. Again, this is a result of the lower surface tension of ethanol. Interestingly, the three seconary roplets of both, kerosene an ethanol appear to be of nearly uniform iameters. Highly resolve images of the isintegrating roplets are provie in our previous publication [3]. For an in-epth escription of roplet breakup by capillary pinching, the reaer is referre to the investigation by Tjahjai et al. [33] an the review paper by Stone [34]. In conclusion, the herein observe roplet breakup procees as follows: When roplets are expose to spherical blast waves from laser-inuce breakowns, they experience strong aeroynamic forces which initiate flattening. The flow velocities ecrease quickly. Accoring to Fig. 15, Weber numbers ecrease below 1 within 10 µs after the arrival of the blast wave. Therefore, breakups o not result from instantaneous aeroynamic forces, which is the case in steay flow fiels. When the Weber numbers have ecrease, roplets continue to flatten for several microsecons ue to kinetic energy previously transmitte by the flow fiels. As soon as the kinetic energy has completely turne into strain energy, the flattening terminates. Subsequently, the strain energy is release, the roplets relax an stretch into filaments. The stretching rate epens on the level of store strain energy, an therefore is a function of the integral strength of the flow fiel, which becomes weaker with increasing istance from the breakown origin. Consequently, at closer istances filaments are longer an thinner than at further istances. Thinner filaments are more prone to capillary pinching. Therefore, filaments within a certain istance from the breakown origin isintegrate by pinching in the center of the filament or by pinch-offs of two seconary roplets at the filament ens. Filaments at greater raial istances, which are loae with less strain energy, may also feature pinching, but subsequently contract an oscillate for several perios. In principle, this process is similar to the vibrational breakup

23 in steay flows at We 11 by the fact that roplets stretch into filaments an isintegrate by capillary pinching. The ifference is that stretching in steay flows results from the builup of a linear oscillation. Here, the critical stretching is reache after a single, impulse-like excitation. Tab. 4 Arrival time of the blast wave after the laser pulse an post-shock flow Weber numbers 0.1 µs after the arrival of the blast wave. 7 mm 8 mm 9 mm Blast wave arrival time t 1 [µs] We t kerosene 1 0.1µs ethanol 1 0.1µs We t Fig. 12 Enlargement of kerosene roplets in the shaowgraph images at 7, 8 an 9 mm horizontal position. The blast-inuce flow has passe from right to left before the images were taken. The origin is at the focal point of the breakown. The sequences are put together from three iniviual recorings of the same experiment.

24 Fig. 13 Enlargement of ethanol roplets in the shaowgraph images at 7, 8 an 9 mm horizontal position. The sequences are put together from three iniviual recorings of the same experiment. Fig. 14 Calculate post-shock flow velocities for ifferent raii aroun the breakown origin. Positive velocities stem from expansion behin the shock front, negative velocities stem from the rarefaction wave.

25 Fig. 15 Calculate post-shock roplet Weber numbers for ifferent raii aroun the breakown origin. 4 Summary an conclusion Blast waves were generate by laser-inuce air breakowns. They were visualize with schlieren imaging, an their expansion trajectories were tracke. Theoretical blast wave moels by Taylor, Seov, Broe an Jones were fitte to the trajectory of a blast wave create by a breakown with an energy of mj. The initial blast wave energy was etermine, an the qualities of the fits were evaluate by the sums of the least y -squares. The moels by Broe an Jones gave very goo fits an reasonable results. They were subsequently valiate over a large range of energies by their comparative application to five iniviual blast wave trajectories from breakowns with energies between 11.6 an mj. Both moels gave similar results: blast waves consume significant portions of the breakown energies, which increase with ecreasing breakown energy. Approximately 52 % was transferre into the blast waves at a breakown energy of mj, while 77 % was transferre at 11.6 mj. The effect of the blast waves on ethanol an kerosene fuel roplets was visualize with shaowgraph imaging. Droplets with a iameter of approximately 94.5 µm isintegrate into two to three seconary roplets within a raius of several millimeters aroun the breakown position. Ethanol roplets isintegrate into three seconary roplets within 7.7 mm an into two seconary roplets between 7.7 an 8.5 mm. Due to the higher surface tension, the raii were smaller for kerosene. Disintegration into three seconary roplets occure within 7.4 mm an into two seconary roplets between 7.4 an 8.0 mm. The given raii are averages from six iniviual images, the variation is ± 0.1 mm. The calculation of the transient post-shock flow velocities an Weber numbers showe that initial Weber numbers behin the shock front were of the orer of 100, but ecrease below 1 within 10 µs. Therefore the aeroynamic forces behin the shock front were strong enough to eform the roplets into isks an to initiate, but not complete, breakup. We conclue that the blast-inuce flow was only responsible for the roplet eformation into isks, but the subsequent breakup resulte from the release of strain energy. Disk-like roplets relaxe an stretche into filaments. Depening on the initial Weber numbers, e.g. the loae strain energy, roplets either oscillate over several perios or isintegrate within 220 µs. Oscillating roplets at 9 mm istance

26 from the breakown origin feature a perio of 120 ± 10 µs (kerosene) an 140 ± 10 µs (ethanol). Theoretical funamental perios were calculate to be µs (kerosene) an µs (ethanol), which was in very goo agreement with the measure perios an inicate the presence of a transient roplet oscillation after a single, impulse-like excitation. Disintegration into two an three seconary roplets of nearly uniform size resulte from pinching of the filaments ue to capillary forces. The experiments helpe to unerstan the role of blast waves in fuel spray ignition. Blast waves can inuce breakup of the fuel roplets into smaller seconary roplets within a raius of several millimeters aroun the spark location. Those seconary roplets can evaporate faster than the initial ones an therefore contribute to the generation of a combustible gas mixture, in which a spray flame kernel can grow an stabilize. To consier this effect in future numerical spray ignition simulations, further investigations of the herein presente breakup mechanism are recommene with both, experimental an numerical methos. Our herein presente experiments were performe with great thoroughness, an results may be use as valiation ata for numerical simulations. References [1] Ronney, P.D.: Laser versus Conventional Ignition of Flames. Opt. Eng. 33, (1994) [2] Gebel, G.C., Mosbach, T., Meier, W., Aigner, M.: Optical an spectroscopic iagnostics of laserinuce air breakown an kerosene spray ignition, Combust. Flame (2014). oi: /j.combustflame [3] Gebel, G.C., Le Brun, S., Mosbach, T., Meier, W., Aigner, M.: An Experimental Investigation of Kerosene Droplet Breakup by Laser-Inuce Blast Waves. J. Eng. Gas Turb. Power 135, (2013) [4] Joarer, R., Gebel, G.C., Mosbach, T.: Two-imensional Numerical Simulation of a Decaying Laser Spark in Air with Raiation Loss. Int. J. Heat. Mass Transf. 63, (2013) [5] Pilch, M., Erman, C.A.: Use of Breakup Time Data an Velocity History Data to Preict the Maximum Size of Stable Fragments for Acceleration-Inuce Breakup of a Liqui Drop. Int. J. Multiphase Flow 13, (1987)