EXPERIMENTAL INVESTIGATION OF MIXING AND IGNITION OF TRANSVERSE JETS IN SUPERSONIC CROSSFLOWS

Transcription

1 EXPERIMENTAL INVESTIGATION OF MIXING AND IGNITION OF TRANSVERSE JETS IN SUPERSONIC CROSSFLOWS a dissertation submitted to the department of mechanical engineering and the committee on graduate studies of stanford university in partial fulfillment of the requirements for the degree of doctor of philosophy Adela Ben-Yakar December, 2000

2 c Copyright 2001 by Adela Ben-Yakar All Rights Reserved ii

3 Abstract Ignition, flame-holding, and mixing enhancement are fundamental aspects of supersonic combustion and are critical to the development of hypersonic airbreathing propulsion engines. High velocities associated with supersonic/hypersonic flight speeds constrain the performance of propulsion systems because of the limited flow residence time inside the combustor. A useful hypervelocity propulsion system therefore requires enhanced mixing of fuel and air, injection with very low drag penalty, and effective distribution of fuel over the burner cross-section. One of the simplest approaches is the transverse injection of fuel from wall orifices. The interesting but rather complicated flow-field dynamics of transverse jets injected into a supersonic crossflow has been studied by many supersonic combustion researchers since 1960 s, but with limited freestream flow conditions. Most of the previous research was performed in conventional wind tunnels by accelerating cold air into supersonic conditions, namely in low velocity and low total enthalpy flow conditions. However, a real supersonic combustor environment at flight speeds beyond Mach 8 can only be simulated using impulse facilities due to the required high total enthalpies. Among various impulse facilities, expansion tubes are especially useful in providing high total enthalpy flows with the proper chemical composition, namely the absence of dissociated species. This research is focused on studying the near-field mixing and ignition properties of transverse fuel jets injected into realistic supersonic combustor flows. We use advanced flow visualization techniques, namely planar laser-induced fluorescence (PLIF) imaging of the hydroxyl radical (OH) and ultra-fast-framing-rate schlieren imaging. While schlieren indicates the location of shock waves, jet penetration and large scale flow features, OH-PLIF is used to map the regions of ignition. The first objective of the present work is to characterize the expansion tube facility iii

4 for three operating points, simulating flight Mach 8, 10 and 13 total enthalpy conditions. The ability of the expansion tube to provide a steady-flow test time of adequate duration and a core flow of sufficient size for 2 mm diameter jet-in-crossflow studies is verified. The second objective is to study the flow-field properties of hydrogen and ethylene jets, owing to their relevance to supersonic combustion. Visual observations of image data, supported by the results for the convection velocity and jet penetration, reveal significant differences between the hydrogen and ethylene injection cases with similar momentum flux ratio. Previously the momentum flux ratio was found to be the main controlling parameter of the jet penetration but the results here demonstrate the existence of an additional mechanism which alters the vortical structure, the penetration and the mixing properties of the jet shear layer. The thickness of the penetration band, used as the representation of the jet-shear-layer thickness is considerable in the ethylene injection case, due to the tilting-stretching-tearing mechanism and also due to the larger growth rate of the jet shear layer. Furthermore autoignition of an ethylene transverse jet is achieved at flight Mach 10 conditions despite the relatively long ignition delay times of ethylene (hydrocarbons), a key limitation for hydrocarbon-fueled scramjets. These results of higher penetration, larger jet shear layer growth rate and autoignition capability indicate that hydrocarbons might be a useful fuel in scramjets flying at Mach 10 conditions. The third objective is to investigate the stability of the jet shear layer at various speed ratios and density ratios via schlieren. The high shear stresses induced by the large velocity difference across the jet shear layer have a large effect on the structure of the layer. For the unstable case, we notice: 1) breakdown of Kelvin-Helmholtz structures with the tilting-stretching-tearing mechanism; 2) increased growth rates with decreasing values of jet-to-free-stream velocity ratio; 3) large intrusions of crossflow in between the eddies, and 4) additional shock waves and distortion of the bow shock around the large eddies. Stable layers show well-defined Kelvin-Helmholtz spanwise rollers. The results plotted in a density-effective velocity ratio (s λ) diagram demonstrate two separate regions of stable and unstable jet shear layers with a separation line at a critical effective velocity ratio. The final objective is to study the ignition and flame-holding capabilities of a hydrogen transverse jet injected into flight Mach 8, 10 and 13 total enthalpy conditions. The results demonstrate self-ignition in the near-field of the hydrogen jet for the high iv

5 total enthalpy conditions (flight Mach 10 and 13). OH-fluorescence is detected along the jet shear layer periphery in a continuous and very thin filament. For the low total enthalpy Mach 8 condition, however, the ignition is limited to a small region behind the bow shock and no OH fluorescence can be observed farther downstream. It is evident from the results that improved injection schemes for better flame-holding would be required for practical applications in scramjet engines, especially in the flight Mach 8 range. During the last few years, cavities have gained the attention of the scramjet community as a promising flame-holding device, owing to results obtained in flight tests and to feasibility demonstrations in laboratory scale supersonic combustors. In this thesis, we summarize the flowfield characteristics of cavities and research efforts related to cavities employed in low- and high-speed flows. Open questions impacting the effectiveness of the cavities as flame-holders in supersonic combustors are discussed. Preliminary studies on cavities with upstream injection are presented indicating selfignition inside and around the cavity. v

6 Contents Abstract iii 1 Introduction Background and Motivation Typical Scramjet Burner Entry Conditions Flow-Field Features of Jets in Supersonic Crossflows Ignition and Flame-Holding Strategies in Supersonic Combustion Thesis Objectives Experimental Aspects Critical Parameters in Supersonic Combustion Simulation Experimental Facility Expansion Tube Injection System and its Calibration Cavity/Injection Plate Test Flow Characterization in the Flight Mach 8-13 Range Flow Conditions Measurement of Flow Properties Boundary Layer Effects on Test Time Core-Flow Size Flow Establishment Time Summary Flow Visualization Techniques Ultra-Fast Framing Rate Schlieren vi

7 3.1.1 Previous and Current High Speed Imaging Efforts High-Speed Schlieren Imaging Components Timing and Synchronization Resolution Considerations Image Processing and Analysis OH-PLIF Excitation and Detection Strategy OH-PLIF Laser Source and Tuning OH-PLIF Imaging System and Its Spatial Resolution Interpretation of OH-PLIF Simultaneous Schlieren and OH-PLIF Time Evolution and Mixing Characteristics of Hydrogen and Ethylene Transverse Jets Introduction Results and Discussion General Flow-Field Features Large Scale Coherent Structures Convection Characteristics Penetration and Shear Layer Properties OH-PLIF Results Summary The Effect of Velocity and Density Ratio on Transverse Jets Effect of Jet Molecular Weight Flow Visualization Results Penetration and Shear Layer Thickness Convection Characteristics Characteristic Large Eddy Frequencies (Possible Transverse Jet Modes) Jet Compressibility Analysis Effect of Density and Velocity Ratios Flow Visualization Results vii

8 5.2.2 Definition of an Effective Velocity Ratio, λ Discussion on the Effect of the Curvature - Centrifugal Instability Analysis Autoignition and Flame-Holding Capability of a Hydrogen Transverse Jet Ignition and Flame-Holding Considerations Ignition and Flame-Holding Results Simultaneous OH-PLIF/Schlieren Results Top View OH-PLIF Images Comparison of Ignition in Flight Mach 8-13 Total Enthalpy Range Discussion of the Ignition Process Ignition Characteristics of Hydrogen Ignition in Supersonic Combustors Ignition of a Hydrogen Transverse Jet Ignition of Ethylene Transverse Jet Cavity Flame-Holders Review of Previous Research Cavity Flow-Field Characteristics Cavity in Reacting Flows Outstanding Questions Preliminary Cavity Results Visual Observation of Cavities Using Ultra-Fast Schlieren Preliminary Ignition Results of Injection/Cavity Schemes Concluding Remarks Summary of Major Results and Conclusions Experimental Aspects Flow Visualization Techniques Characteristics of Hydrogen and Ethylene Transverse Jets Density and Velocity Ratio Effects Ignition and Flame-Holding Capability of a Hydrogen Transverse Jet viii

9 8.1.6 Cavity Flame-Holders Recommendation For Future Work A Expansion Tube Equations 175 B Maps of Estimated Expansion Tube Test Conditions 178 Bibliography 184 ix

10 List of Tables 1.1 Advantages and disadvantages of expansion tubes relative to shock tunnels for hypervelocity combustion simulations Test gas (free-stream) flow properties simulating the burner entry conditions of three flight Mach numbers. The corresponding values are from Fig Summary of measured, ideal (inviscid 1-D) and predicted (based on Mirels solution) properties of test gas for Mach 10 and 13 flow conditions Jet exit flow properties The general flow exit properties of gaseous jets with different molecular weights The specific flow exit properties of gaseous jets used in the study of the jet molecular weight effect. The free-stream used in these experiments simulates the flight Mach 10 flow condition Summary of the different conditions used in the study of jet instability analysis Summary of cavity oscillation frequencies, f m, for different cavity length to depth ratios, L/D. The table includes the expected values based on Rossiter s formula and the ones measured in our experiments Recommended free-stream flow conditions for further ignition studies x

11 List of Figures 1.1 Typical scramjet burner entry conditions as a function of flight Mach number, calculated assuming adiabatic compression. a) The burner entry Mach number, M 3, for different temperature ratios, T 3 /T 0. b) The burner entry pressure, p 3, and the flight trajectories of constant dynamic pressure, q 0, of 50 and 100 kpa. In our experiments, total enthalpy flows (Mach number, M 3, and static temperature, T 3 ) simulating three nominal flight conditions (Mach 8, 10 and 13) were generated Schematic of an underexpanded transverse injection into a supersonic cross-flow, (a) instantaneous side view at the center-line axis of the jet; (b) 3-D perspective of the averaged features of the flow-field (Gruber et al. 1995) Flow-field schematics of traditional injection/flame-holding schemes for supersonic combustors. a) underexpanded fuel injection normal to the crossflow, b) fuel injection at angle, c) injection behind a sudden expansion produced by a step Expansion tube facility (12 m in length and 89 mm inner diameter) and imaging system Expansion tube distance-time (x-t) diagram calculated for flight Mach 13 condition. Method of characteristics was used to solve the flow gasdynamics properties assuming one-dimensional inviscid theory. Test time is defined as the time that the test gas has uniform flow quantities and determined by the time arrival of the contact surface to the tube exit, and that of the first subsequent rarefaction wave (reflected rarefaction head in our case of high total enthalpy simulations) xi

12 2.3 Schematic of the test section (27 x 27 cm cross section) where a rake of 4 pitot probes, instrumented with pressure transducers, was located 2.5 cm downstream of the tube exit. The flow history during the expansion tube operation was detected via pitot pressure information. Note that the inner diameter of the tube is 8.9 cm Optical set-up to measure the test gas velocity, assumed to be equal to the CS - contact surface velocity. IR emission from 5% CO 2 seeded in the test gas nitrogen is collected by an InSb IR detector at the viewing port located at cm from the end of the tube. The test gas velocity can then be calculated by considering its time of arrival at the viewing port and at the pitot rake Schematic of a) Injection system, b) cavity/injection plate system Schlieren visualization of an underexpanded gaseous injection into still air. (a)-(c) hydrogen (d)-(e) ethylene jets. The exposure time of the images was 3 musec. Mach disk height, y 1, was measured for different pressure ratios, P j /P eb, to calibrate the injection system Maps of estimated test gas (nitrogen) conditions (state 5) at the exit of an expansion tube: a) pressure and temperature, b) total enthalpy and velocity of the test gas are plotted for different initial driven and expansion section pressures. Calculations are performed using the inviscid 1D equations for a given driver pressure of P 4 = 600 psig (helium). Note that the effective filling pressure of the driver section is taken as P 4,eff = 686 psig, as its inner diameter (10.2 cm) is larger than that of the driven and expansion sections (8.9 cm). This area difference is accounted for in the curves presented above Example of IR emission, pitot pressure, wall pressure records and the Mach number variation based on the pitot-to-static pressure ratios, as a function of time for the Mach 10 flow condition. t = 0 represents incident shock arrival at the pitot probe, placed 2.5 cm downstream of the tube exit, while the wall pressure transducer and IR detector are positioned 40.6 cm and cm upstream of the tube exit, respectively (see Fig. 2.4). Note that the time scale of the static pressure trace is shifted by 235 µs to match the shock arrival at the pitot probe xii

13 2.9 Example of IR emission, pitot pressure and wall pressure traces as a function of time for the Mach 13 flow condition Example of IR emission, pitot pressure and wall pressure traces as a function of time for the Mach 8 flow condition Comparison of the measured contact-surface velocity (test gas velocity) with the shock-induced gas velocity estimated using the measured shock speeds in the expansion section X-t diagrams for the expansion section only. 1-D inviscid calculations are plotted in straight lines and results applying Mirels model to include the boundary layer effects are plotted in dashed lines. One can see the improved test time as a result of the contact surface (CS) acceleration due to the developing boundary layer behind the incident shock in the low pressure expansion section helium flow. The incident shock velocity was measured and assumed to be constant along the expansion section Useful core-flow size in flight Mach 13, 10 and 8 conditions, (a), (b) and (c), respectively, determined by measuring the radial variation of pitot pressure at different distances from the tube exit Schlieren imaging set-up Examples of schlieren images of jet issuing into quiescent air as obtained for different positions of the knife edge (razor blade) at the focal point. We use the set-up demonstrated in (d) where the knife edge cuts the focused light at an angle to enhance both the vertical and the horizontal density gradient effects Timing diagram of the high-speed rate imaging system and its synchronization with the expansion tube test flow time Examples of schlieren images with different integration/exposure times: a) 100 ns exposure time, resolving the instantaneous features of the flowfield, b) 200 ns exposure time, resulting in blurring of the image, c) 3 µs exposure time, averaging the general features while enhancing the weak shocks such as upstream separation shock wave and downstream recompression wave xiii

14 3.5 Timing diagram of the high-speed rate imaging system and its synchronization with the expansion tube test flow time a) Triggering diagram and timing connections of the imaging, the injection and the data acquisition systems. b) Timing diagram of simultaneous OH-PLIF and schlieren and their synchronization with the expansion tube test flow time Examples of hydrogen (a) and ethylene (b) injections into a supersonic crossflow (nitrogen). Exposure time of each image was 200 ns. The x-axis is normalized by the jet diameter d An example of schlieren image with 3 µs exposure time for hydrogen injection case. While the unsteady features (coherent structures) are averaged to zero, some of the weak shocks such as upstream separation shock wave and downstream recompression wave are emphasized (a) Bow shock position and its angle at the center-line of the jet as measured from the long exposure schlieren image shown in Fig (b) The free-stream velocity behind the bow shock and the flow turning angle based on the measured bow shock shape. For the calculations a calorically perfect gas has been assumed An example of 8 consecutive schlieren images of underexpanded hydrogen injection (d=2 mm) into a supersonic crossflow (nitrogen) obtained by high-speed-framing camera. Exposure time of each image is 100 ns and interframing time is 1 µs. Free-stream conditions are: U =2360 m/s, M =3.38, T =1290 K, p =32.4 kpa; and jet-to-free-stream momentum ratio is: J=1.4± The second example of 4 of 8 consecutive schlieren images of hydrogen injection into flight Mach 10 condition. Exposure time of each image is 100 ns and interframing time is 1 µs Time evolution of an ethylene jet in a supersonic crossflow (nitrogen) as observed from 8 consecutive schlieren images. Exposure time of each image is 100 ns and interframing time is 1.5 µs. Free-stream conditions are: U =2360 m/s, M =3.38, T =1290 K, p =32.4 kpa; and jet-tofree-stream momentum ratio is: J=1.4± xiv

15 4.7 The second example of an ethylene transverse jet flow-field in a supersonic crossflow as observed from 8 time correlated schlieren images. Exposure time of each image is 200 ns and interframing time is 1.2 µs Schematic of the three-dimensional shape (Ω shape) of the unsteady vortical structures formed intermittently (Brizzi et al. 1995) Development of a large-scale ethylene structure (eddy number -1 in Fig. 4.7) as it goes through the tilting and stretching processes. Four different parts of the eddy structure were independently tracked in the duration of the 8.6 µs flow visualization time Space-time trajectories of large-scale eddies present in the hydrogen jet shear layer. The center of the eddies are tracked from the 8 successive schlieren images shown (a) in Fig. 4.4 and (b) in Fig Space-time trajectories of ethylene large scale eddies as tracked from 8 time-correlated schlieren images: (a) x-t diagram of the example shown in Fig. 4.6, (b) x-t diagram of the example shown in Fig Convection features of coherent large scale structures present in the hydrogen jet/free-stream shear layer. The data were subtracted by analyzing the eddy displacement in 8 consecutive schlieren images of 2 experiments (images shown in Figs. 4.4 and 4.5). (a) the convection velocity of eddies in streamwise and transverse directions, U c,x and U c,y, respectively; (b) the convection angle of eddies Convection features of eddies present in the ethylene jet/free-stream shear layer. The data were subtracted by analyzing the eddy displacement in 8 consecutive schlieren images of 2 experiments (images shown in Figs. 4.6 and 4.7). (a) the convection velocity of eddies in streamwise and transverse directions, U c,x and U c,y, respectively; (b) the convection angle of eddies Measured convection velocity of large eddy structures in the hydrogen and ethylene jet shear layers. The results are compared with the estimated values of the free-stream velocity immediately behind the bow shock Schematic showing the low- and high-speed regions of the bow shockinduced free-stream velocity around the large-scale ethylene eddies xv

16 4.16 Transverse penetration data of (a) hydrogen jet and (b) ethylene jet. The data points were obtained by manually tracking the visually observable outer edge of the jet from 8 consecutive schlieren images for J = 1.4±0.1. Both of the figures include analysis of 2 experiments namely 16 images. For comparison, also shown in the figures is the penetration correlation given by other studies OH-PLIF results mapping the ignition regions at the jet center-line of: a) hydrogen injection into air, b) ethylene injection into air, c) ethylene injection into pure oxygen Examples of instantaneous schlieren images of jets with different molecular weights. Free-stream conditions are: U =2360 m/s, M =3.38, T =1290 K, p =32.4 kpa Jet transverse penetration along the axial distance, x/d. Data for four gases with different molecular weights are presented: a) M w = 2, J = 1.84, b) M w = 4, J = 1.72, c) M w = 8, J = 1.85, d) M w = 16, J = For comparison, empirical correlations suggested by Gruber et al. (1995) and Rothstein and Wantuck (1992) are also included for J = Convection velocity of large scale structures in the streamwise (M c,x ) and transverse (M c,y ) directions as a function of axial distance x/d. The results for each case (for each molecular weight of jet) are obtained from 4-5 experiments each including 8 consecutive schlieren images Formation frequency of the large scale structures and the corresponding preferred mode Strouhal number, St d = f j d/u j, as a function of the jet exit velocity. The data were collected from the time evolution observation of the jet from 8 consecutive schlieren images. Each data point was obtained by averaging 5-10 experiments with the error bars representing the deviation from the mean value Formation frequency of the large scale structures and the initial vortex shedding Strouhal number, St θj = f θj θ j /U j, as a function of the jet Reynolds number Flow-field schematics used in the jet compressibility analysis. Letters A, B and C indicate the zones of the jet shear layer xvi

17 5.7 Estimated convective Mach number in zone A, M A c, (refer to the schematic in Fig. 5.6) and the measured visible jet shear layer thickness, δ vis, at x/d 22 as obtained from penetration width measurements Estimated velocity fields for the jet and the free-stream in zones B and C Schlieren images at selected conditions given in Table Schlieren images at selected conditions given in Table Schlieren images at selected conditions given in Table Velocity vector field (U, U j ) for a skewed mixing layer and the effective velocity ratio, λ., described in the total velocity vector direction Jet-to-free-stream density ratio vs. velocity ratio. The number near the data points corresponds to the experimental conditions summarized in Table 5.3. Unstable flow jet is defined when the large structures lose coherence downstream of the injection port and significant distortions in the bow shock shape can be observed Jet-to-free-stream density ratio vs. the effective velocity ratio, λ. The number near the data points corresponds to the experimental conditions summarized in Table Schematics illustrating the stability regions based on a) Rayleigh criterion for centrifugal forces in the curved mixing layers as given in Eq where cons. = δ vis h max and b) current experimental results Simultaneous OH-PLIF and schlieren images visualizing hydrogen injection into supersonic crossflow. Free-stream conditions are M = 3.57, T = 1300 K, P = 0.32 atm, V = 2500 m/s. The jet-to-freestream momentum flux ratio is J = 1.4. a) Schlieren image, b) OH-PLIF image demonstrating the ignition and combustion regions of jet-in-crossflow at high enthalpy condition, c) Overlaid OH-PLIF and schlieren images Instantaneous top-view OH-PLIF images obtained at different height above the injection plate. Free-stream conditions are M=3.57, T=1300K, P=0.32atm, V=2500m/s. The jet-to-freestream momentum flux ratio is J=1.4. a) y/d=3, b) y/d=2.5, c) y/d=2, d) y/d=1 above the injection plate.119 xvii

18 6.3 Instantaneous OH-PLIF acquired at center-line axis of the hydrogen jet injected into flight Mach 10 and 13 conditions. The images are obtained by combination of 2 different instantaneous images: near the exit of the jet ( 5 < x/d < 1) and downstream of the jet (1 < x/d < 10) Two instantaneous OH-PLIF images acquired at center-line axis of the hydrogen jet injected into flight Mach 8 conditions Explosion limits of a stoichiometric hydrogen-oxygen mixture (after Sung et al., 1999) Variation of ignition delay times τ ign of a stoichiometric mixture of H 2 and air with temperature and pressure. Calculations are perfomed using Chemkin and the GRI mechanism. a) τ ign vs. T, b) pτ ign vs. T Variation of ignition time with fuel-air equivalence ratio, φ, for cold H 2 (T jet = 300 K) injected into hot air. The values of the ignition delay time are calculated for different air temperatures, T air The free-stream temperature and pressure (T 2 and P 2 ) behind the bow shock, measured from schlieren images as discussed in Section (see Fig. 4.3). Ignition delay times are calculated for several conditions of air assuming φ = 0.2. The free-stream flow properties simulate the flight Mach 10 conditions Comparison of ignition delay of a stoichiometric mixture of C 2 H 4 (ethylene) and air/oxygen at 1 atm with a stoichiometric mixture of H 2 and air. Two different reaction mechanisms are used to calculate the ignition delay times of C 2 H a) Variation of ignition delay of a stoichiometric mixture of C 2 H 4 (ethylene) and air/oxygen at various pressures xviii

19 7.1 Flow-field schematics of cavities with different length to depth ratios, L/D, in a supersonic flow. a) Open cavity flow for L/D < 7 10; shear layer reattaches to the back face while spanning over the cavity. Small aspect ratio cavities (L/D < 2 3) are controlled by transverse oscillation mechanism while in larger aspect ratio cavities longitudinal oscillation becomes the dominant mechanism. b) Closed cavity flow for L/D > 10 13; shear layer reattaches to the lower wall. The pressure increase in the back wall vicinity and the pressure decrease in the front wall results in large drag losses Typical longitudinal cavity oscillations are caused by the impingement of the free shear layer on the rear wall which generates travelling shocks inside the cavity. The shear layer spanning the cavity becomes unsteady as a result of these acoustic waves deflecting the shear layer up and down, and/or by the shock induced vortices generated at the front wall leading edge of the cavity. As a result unsteady waves emanate from the cavity Different concepts can be employed to suppress the cavity oscillations: a) Cavities with an angled back wall suppress the unsteady nature of the free shear layer by eliminating the generation of the travelling shocks inside the cavity due to the free-shear-layer impingement. b) In addition, small disturbances produced by spoilers or by the secondary jet injection upstream of the cavity can enhance the free-shear-layer growth rate. The thickening of the cavity shear layer alters its instability characteristics, such that its preferred roll-up frequency is shifted outside of the natural frequency of the cavity, and as a result the oscillations are attenuated xix

20 7.4 Instantaneous schlieren images with 200 ns of exposure time demonstrating the effect of the back wall angle on the flowfield structure of a cavity exposed to a supersonic flow. The free-stream was generated in an expansion tube to simulate Mach 10 total enthalpy conditions at the supersonic combustor entry: M = 3.4, U = 2360 m/s, T = 1290 K, p = 32 kpa. The boundary layer thickness at the trailing edge of the cavity is approximately 1mm. a) Cavity with L/D = 5 shows the unsteady nature of the shear layer at the reattachment with the trailing edge of the back wall. b) Cavity with slanted back wall (20 o ) stabilizes the shear layer reattachment process Effect of length-to-depth ratio, L/D on a) magnitude (root-mean-square) of pressure fluctuations on the bottom of the cavity (at x/d = 0.33), b) drag of the cavity at Mach 1.5 and 2.5 flows. The values were adapted from Zhang and Edwards (1990) Cavity-actuated supersonic mixing enhancement concepts: (a) Sato et al. (1999), studied the influence of acoustic waves, emitted from a cavity and impinging on the initial mixing layer. (b) Yu and Schadow (1994) used the same concept to enhance the mixing of supersonic reacting jets Axisymmetric combustor of the Scramjet engine which was flight-tested by Russian-CIAM/NASA joint program (1998). In this engine two cavities with angled-rear wall were used for flame-holding purposes. The dimensions are in mm (McClinton et al. 1996) Position of pressure transducers located at the bottom of the cavity to measure the history of the flow oscillations inside the cavity. Pressure transducer located farther downstream at x/d = 1.5 provided a more accurate oscillation frequency measurements Examples of cavity pressure traces in arbitrary units: a) L/D = 3, b) L/D = 5, c) L/D = 5 with upstream hydrogen injection, d) L/D = 7. t = 0 represents incident shock arrival at the cavity. The free-stream (N 2 ) conditions represent Mach 10 total enthalpy at the supersonic combustor entry: M = 3.4, U = 2360 m/s, T = 1290 K, p = 32 kpa xx

21 7.10 Schlieren images demonstrating the differences in the flow-field structure of cavities with different length-to-depth ratios and back wall angle. The depth of the cavities is constant and equal to D = 3 mm. The freestream was generated to simulate Mach 10 total enthalpy conditions at the supersonic combustor entry: M = 3.4, U = 2360 m/s, T = 1290 K, p = 32 kpa. The boundary layer thickness at the trailing edge of the cavity is approximately 1mm Schlieren images demonstrating jet interaction with different cavities. The hydrogen jet is injected into a non-reacting free-stream 3 mm upstream of the cavity from a d = 1 mm orifice. The injection is performed at angle of 30 o to the plate. The free-stream, N 2, represents the flight Mach 10 burner entry conditions Instantaneous OH-PLIF images demonstrating the ignition regions of a hydrogen jet interacting with a L/D = 3 cavity. The images are obtained from 5 single shots at the same conditions. Hydrogen is injected 3 mm upstream the cavity leading edge at an angle of 30 o. The free-stream (air) properties represent the flight Mach 10 burner entry conditions: M = 3.4, U = 2360 m/s, T = 1290 K, p = 32 kpa. Note that a schlieren image is also included to indicate the flow-field properties around the cavity Instantaneous OH-PLIF images demonstrating the ignition regions of a hydrogen jet interacting with a cavity with L/D = 3 and 30 o back wall. The images are obtained from 5 single shots at the same conditions. Hydrogen is injected 3 mm upstream the cavity leading edge at an angle of 30 o. The free-stream (air) properties represent the flight Mach 10 burner entry conditions: M = 3.4, U = 2360 m/s, T = 1290 K, p = 32 kpa. Note that a schlieren image is also included to indicate the flow-field properties around the cavity Flow-field schematics demonstrating different concepts of angled jet injection combined with cavity flame-holder. a) upstream injection, b) base injection, c) cavity injection Flow-field schematic of a shock-wave/jet interaction xxi

22 8.3 Schematic of the 25 o wedge to generate a shock wave above the injection plate (a) An oblique shock wave impinging the hydrogen jet as visualized using schlieren imaging. The shock was produced by a 25 o angled wedge mounted above the injection plate. Flight Mach 13 free-stream condition. (b) Combined OH-PLIF and schlieren images visualizing the effect of shock/jet interaction on OH number density B.1 Maps of estimated test gas (nitrogen) conditions at the exit of an expansion tube: a) pressure and temperature, b) total enthalpy and velocity of the test gas are plotted for different initial driven and expansion section pressures. Calculations are performed using the inviscid 1D equations for a given driver pressure of P 4 = 300 psig (helium) B.2 Maps of estimated test gas (Argon) conditions at the exit of an expansion tube: a) pressure and temperature, b) total enthalpy, velocity and Mach number of the test gas are plotted for different initial driven and expansion section pressures. Calculations are performed using the inviscid 1D equations for a given driver pressure of P 4 = 300 psig (helium) B.3 Maps of estimated test gas (Argon) conditions at the exit of an expansion tube: a) pressure and temperature, b) total enthalpy, velocity and Mach number of the test gas are plotted for different initial driven and expansion section pressures. Calculations are performed using the inviscid 1D equations for a given driver pressure of P 4 = 600 psig (helium) B.4 Maps of estimated test gas (Helium) conditions at the exit of an expansion tube: a) pressure and temperature, b) total enthalpy, velocity and Mach number of the test gas are plotted for different initial driven and expansion section pressures. Calculations are performed using the inviscid 1D equations for a given driver pressure of P 4 = 300 psig (helium) B.5 Maps of estimated test gas (Helium) conditions at the exit of an expansion tube: a) pressure and temperature, b) total enthalpy, velocity and Mach number of the test gas are plotted for different initial driven and expansion section pressures. Calculations are performed using the inviscid 1D equations for a given driver pressure of P 4 = 600 psig (helium) xxii

23 Chapter 1 Introduction 1.1 Background and Motivation The success of future hypersonic airbreathing propulsion systems will be largely dependent on efficient injection, mixing and combustion processes inside the supersonic / hypersonic combustion chamber. At flight speeds beyond Mach 6, air entering the combustor must be supersonic to avoid excessive dissociation of both nitrogen and oxygen gases. Consequently, the time available for fuel injection, fuel-air mixing and combustion is very short, of the order of 1 msec, which results in troublesome constraints on the combustion problem (Ferri 1973; Kumar et al. 1989) Typical Scramjet Burner Entry Conditions The combustor entry conditions (Mach number, static temperature and pressure) of hypersonic airbreathing propulsion systems depend on the flight conditions of the vehicle. In order to keep the density inside the combustor high for efficient combustion and the lift at reasonably high values, the flight Mach number, M 0, should increase as the altitude of the vehicle increases. Residence time is another issue that has to be considered for efficient performance of a high-speed propulsion system. The air must be compressed in the diffuser in order to reduce velocities and increase the flow residence time and therefore to allow a combustor of reasonable length. On the other hand, the reduced velocities at the combustor entry are restricted by the maximum allowable compression temperature (in 1

24 CHAPTER 1. INTRODUCTION 2 Burner Entry Mach Number, M Current experiments 1. Mach 8 (3 MJ/kg) 2. Mach 10 (4 MJ/kg) 3. Mach 13 (6 MJ/kg) T 3 /T 0 =6 7 8 M 3 = Burner Entry Pressure, p 3, atm Flight Mach Number, M 0 Flight Mach Number, M q 0 =50 kpa q 0 =100 kpa p 3 =1 atm Height, km (a) (b) FIGURE 1.1 Typical scramjet burner entry conditions as a function of flight Mach number, calculated assuming adiabatic compression. a) The burner entry Mach number, M 3, for different temperature ratios, T 3/T 0. b) The burner entry pressure, p 3, and the flight trajectories of constant dynamic pressure, q 0, of 50 and 100 kpa. In our experiments, total enthalpy flows (Mach number, M 3, and static temperature, T 3) simulating three nominal flight conditions (Mach 8, 10 and 13) were generated. the range of K (Heiser and Pratt 1994)) to avoid excessive dissociation in the exhaust flow. These constraints determine the expected values of combustor entry Mach number and temperature, M 3 and T 3, respectively. Considering the above issues, the expected values of flow conditions at the combustor entrance of an airbreathing propulsion system were estimated and plotted in Fig. 1.1 as a function of flight Mach number, M 0. Calculations were performed for different burner entry temperature to atmosphere temperature ratios (T 3 /T 0 ) assuming adiabatic compression (constant total enthalpy) throughout the diffuser according to Eq. 1.1: T 3 = 1 + γ0 1 T γ M M 2 0 (1.1) As shown in Fig. 1.1a, for hypersonic flights beyond Mach 6, (M 0 > 6) a supersonic combustion ramjet (scramjet) where the flow remains supersonic / hypersonic throughout the engine should be considered. Furthermore, to keep structural loads on the hypersonic vehicle at acceptable levels, namely, to keep the dynamic pressure, q 0 = 1/2ρ 0 V0 2, in the range of kpa, flight at high speeds is confined to altitudes of km. Consequently, the burner entry pressure, p 3, can be directly evaluated (see Fig. 1.1b) for fixed dynamic pressure, q 0,

25 CHAPTER 1. INTRODUCTION 3 of 50 and 100 kpa, compression efficiency, η c, of 0.9 and temperature ratio, T 3 /T 0, of 6 using Eq. 1.2 (Heiser and Pratt 1994): P 3 P 0 = ( T 3 ) γ c γc 1 T 0 T 3 T 0 (1 η c ) + η c where γ c is the average specific heat ratio; γ c = 1/2(γ 0 + γ 3 ). (1.2) Therefore, at flights beyond Mach 8, typical pressures at the entrance of supersonic combustors range from approximately 0.2 to 4 atm depending on the operating parameters for the flight mission, such as the Mach number and the altitude. Most supersonic combustion research in the open literature has focused on flight speeds of Mach 8 and below (Allen et al. 1993; McMillin et al. 1994; Gruber et al. 1995; Parker et al. 1995; Santiago and Dutton 1997), and there are relatively few works which have been performed for higher flight Mach numbers (Stalker 1989; Anderson et al. 1990; Bakos et al. 992a; Bakos et al. 992b; Bakos et al. 992c; Erdos 1994; Anderson 1994; Albrechcinski et al. 1995; Wendt and Stalker 1996; Bélanger and Hornung 1996; McIntyre et al. 1997; Erdos 1998; Ben-Yakar and Hanson 002a). Due to the large total enthalpies (greater than 3 MJ/kg) associated with high flight Mach numbers, only impulse facilities are capable of providing the required total temperature and Mach number to replicate a combustor environment. Expansion tubes and reflected shock tunnels are two possible types of impulse facilities for ground testing. Of concern for high stagnation enthalpy simulations is the chemical composition of the test gas. While in reflected shock tunnels significant amounts of dissociated species are formed, in expansion tubes the amounts of these species are negligible (?). Therefore, an expansion tube can provide a more correct simulation of the true flight combustion chemistry including ignition delay and reaction times. In general, however, expansion tubes have shorter test times than reflected shock tunnels. The principal advantages and disadvantages of expansion tubes as compared to other hypersonic test facilities, especially shock tunnels, are summarized in Table 1.1. In the present study, the Stanford expansion tube facility is used to generate total enthalpy conditions in the Mach 8-13 flight range. This facility is one of the few impulse-type facilities which can provide a wide range of total enthalpies. The freepiston reflected shock tunnel, T5, located at GALCIT (Bélanger and Hornung 1996),

26 CHAPTER 1. INTRODUCTION 4 Calspan reflected shock tunnel (Albrechcinski et al. 1995) and the HYPULSE expansion tube located at GASL (Bakos et al. 992a; Bakos et al. 992b; Erdos 1994) are three current examples of larger impulse facilities. In these facilities, generic combustor models with hydrogen injection have been tested using conventional measurement techniques such as pressure measurements along the combustor and flow visualization with differential interferometry. While most of the high enthalpy and high speed combustion flow-field studies in the open literature utilize these methods, modern laser-based diagnostics can provide flow-field and species information critical for fundamental research (Erdos 1994; Erdos 1998; Anderson et al. 1992; Rogers et al. 1992; Ben-Yakar and Hanson 998b) Flow-Field Features of Jets in Supersonic Crossflows Efficient performance of very high-speed combustor systems requires fuel and air mixing at the molecular level in the near-field of the fuel injection. One of the simplest approaches is the transverse (normal) injection of fuel from a wall orifice. As the fuel jet, sonic at the exit, interacts with the supersonic crossflow, an interesting but rather complicated flow-field is generated. Figure 1.2 illustrates the general flow-features of an under-expanded transverse jet injected into a cross-flow. As the supersonic crossflow is displaced by the fuel jet a 3-D bow shock is produced due to the blockage produced by the flow. The bow shock causes the upstream wall boundary layer to separate, providing a region where the boundary layer and jet fluids mix subsonically upstream of the jet exit. This region confined by the separation shock wave formed in front of it, is important in transverse injection flow-fields owing to its flame-holding capability in combusting situations, as has been shown in previous publications (Parker et al. 1995; Ben-Yakar and Hanson 998b). The recent experimental studies performed by Fric and Roshko (1994) provide a new insight into the vortical structure of a jet injected into a low speed crossflow. Their photographs, obtained using the smoke-wire visualization technique, illustrate four types of coherent structures: (1) the near-field jet-shear layer vortices; (2) the far-field counter rotating vortex pair (CVP); (3) the horseshoe vortex which wraps around the jet column; and (4) the downstream wake vortices originating from the horseshoe vortex. Figure 1.2 shows the presumed vortical structures for the jet in supersonic crossflow (which are

27 CHAPTER 1. INTRODUCTION 5 TABLE 1.1 Advantages and disadvantages of expansion tubes relative to shock tunnels for hypervelocity combustion simulations. Shock Tunnels Significant level of radicals such as O and NO are produced in the test gas affecting the combustion chemistry. In the reflected zone of the shock tube, air dissociates due to high temperatures and recombines only partially during the fast expansion process. The facility needs to contain the total pressures and temperatures of the flow it generates. As noted by Anderson (1994), at flight Mach numbers above 12, the total pressure requirements approaches a million psi or 68,000 atm, which can be produced only by expansion tubes. Free-stream Mach number is fixed by the nozzle geometry. Simulation of different conditions requires replacement of the nozzle with a new geometry. Nozzle can be damaged due to the high heat transfer rates at the throat and flying diagrams inside the tube. Boundary layer develops throughout the nozzle and can be thick compared to the dimensions of the injection port. It is usually required to eliminate the boundary layer by, for example, inserting a step before the fuel injection port (Parker et al. 1995) Test times 1 msec Test times are of the order of 1msec. However, a substantial part of it is wasted during the nozzle start-up time (of the order of 0.5 msec), required for the supersonic flow to be established. Test time decreases with increasing stagnation enthalpy. Longer test section because of the longer test time and larger core flow. However, side wall effects should be taken into account. Expansion Tubes Negligible amounts of radicals are produced. The working gas never stagnates, thus reduces the extent of dissociation. As a result the test gas reaches to the test section with more accurate chemical composition. Higher stagnation pressures and temperatures can be achieved in expansion tubes even for the same initial driver pressure and sound speed, as velocity is added to the flow through the unsteady expansion process without stagnating it. Variable Mach numbers and conditions can be easily obtained by just altering the initial filling pressures. High heat transfer rates are avoided in the absence of a sonic throat. However, the test object is prone to damage from flying diagrams arriving at the end of the expansion tube operation. A thin boundary layer is developed upstream of the injection port as the injection plate is placed in the free-jet exiting the tube. Test times msec No nozzle start-up time is required. In addition, the establishment of flow on the studied model begins during the expansion section gas flow prior to the test gas arrival. As a result, less useful test time is consumed during the flow establishment. Test section dimensions depend on the size of the core flow at the exit of the tube which is diminished by the boundary layer growth on the tube walls.

28 CHAPTER 1. INTRODUCTION 6 known to exist in subsonic jet-in-crossflow) as they were partially observed by numerous studies (Gruber et al. 1996; Gruber et al. 997a; Ben-Yakar et al. 998a). The origin of the jet vortical structures was studied by several researchers (Fric and Roshko 1994; Brizzi et al. 1995; Yuan et al. 1999). Among those studies, Yuan et al. (1999) performed a large-eddy simulation of transverse jets in subsonic crossflows. Their results revealed that the majority of the jet vortical structures arose from the Kelvin Helmholtz (K-H) instability of the jet-shear layer in the near-field. Interestingly, they do not observe the formation of vortex rings around the periphery of the jet as was assumed in previous studies. Instead they find two kinds of vortices originating from the jet exit boundary layer: 1) regularly formed spanwise rollers on the upstream and downstream edges (large scale jet shear layer vortices), 2) quasi-steady vortices, the socalled hanging vortices that form in the skewed mixing layers (mixing layers formed from non-parallel streams) on each lateral edge of the jet leading to the formation of the CVP. The near-field mixing of transverse jets is dominated by the so-called entrainmentstretching-mixing process driven by large scale jet-shear layer vortices. In the region near the injector exit, the injectant fluid moves with a higher velocity tangent to the interface than the free-stream fluid. As a result, large vortices are periodically formed engulfing large quantities of free-stream fluid and drawing it into the jet-shear layer (macromixing). These large scale vortices also stretch the interface between the unmixed fluids. Stretching increases the interfacial area and simultaneously steepens the local concentration gradients along the entire surface while enhancing the diffusive micromixing. Preliminary examinations (Gruber et al. 997a; Ben-Yakar and Hanson 002b) of the convection characteristics of these large-scale structures, developed in a sonic transverse jet injection into supersonic crossflows, determined that in the far-field the eddies tend to travel with velocities that are closer to the free-stream velocity. This indicates that in high speed free-stream conditions, these large coherent structures, where the fuel and air are mixed by slow molecular diffusion, will also travel at high speeds. Consequently the combustion process will be mixing controlled. High mixing efficiency, however, must be achieved in the near-field of the fuel injection for the success of hypersonic propulsion systems. Therefore, it is important to understand how these structures and their growth rates evolve as flow and jet conditions

29 CHAPTER 1. INTRODUCTION 7 (a) LARGE-SCALE STRUCTURES M >1 BOW SHOCK BARREL SHOCK MACH DISK BOUNDARY LAYER SEPARATED REGION RECIRCULATION ZONE INJECTANT (Hydrogen or Ethylene) RECIRCULATION ZONE M >1 (b) MACH DISK & BARREL SHOCK AVERAGE JET BOUNDARY 3-D BOW SHOCK COUNTER-ROTATING VORTEX PAIR (CVP) HORSESHOE-VORTEX REGION FIGURE 1.2 Schematic of an underexpanded transverse injection into a supersonic cross-flow, (a) instantaneous side view at the center-line axis of the jet; (b) 3-D perspective of the averaged features of the flow-field (Gruber et al. 1995).

30 CHAPTER 1. INTRODUCTION 8 are changed. Two types of fuel are being considered for use in supersonic combustion: 1) hydrogen and 2) hydrocarbon fuels. The large differences in the molecular weights of these fuels result in a big variation in injection velocities that might lead to a wide variation in the jet shear layer growth rate and the mixing properties. However, none of the previous jet penetration studies (Zukoski and Spaid 1964; Schetz and Billig 1966; Rogers 1971; Rothstein and Wantuck 1992; Papamoschou and Hubbard 1993; Gruber et al. 1995) found any dominant differences between jets with different molecular weights. Penetration was shown to be dependent primarily on the jet-to-free-stream momentum flux, J, expressed by: J = ( ρu 2 ) jet (ρu 2 ) (1.3) Most transverse jet-in-crossflow studies were, however, carried out in cold supersonic flows (namely low velocities) generated in blow-down wind tunnels. The free-stream temperatures and velocities in these facilities were usually lower than that expected in a real supersonic combustor environment. Comprehensive studies still need to be performed to determine the mixing properties of different type of fuels in a relatively accurate supersonic combustor environment. These observations gave rise for the following question: is there any other mechanism or controlling parameter which will alter the large eddy characteristics of the jet shear layer to enhance its near-field mixing in realistic conditions? We were therefore challenged to study the flow features of hydrogen and ethylene transverse jets exposed to high-speed supersonic free-streams at realistic conditions leading to high shear levels Ignition and Flame-Holding Strategies in Supersonic Combustion Different injection strategies have been proposed (Billig 1993; Tishkoff et al. 1997; Abbitt et al. 1993; Hartfield et al. 1994; Riggins et al. 995a; Riggins and Vitt 995b; Fuller et al. 1998) with particular concern for rapid near-field mixing. These injection strategies, both flush-mounted injectors and intrusive injectors, typically rely on the generation of strong streamwise counter-rotating vortices. As a result, mixing is enhanced both in macro-scale by entrainment of large quantities of air into the fuel

31 CHAPTER 1. INTRODUCTION 9 and in micro-scale due to stretching of the fuel-air interface. Stretching increases the interfacial area and simultaneously steepens the local concentration gradients thereby enhancing the diffusive micro-mixing. Micro-scale mixing is required for combustion since chemical reactions occur at the molecular level. However, efficient mixing of fuel and air does not directly initiate the combustion process. Ignition and flame-holding in supersonic flows (Huber et al. 1979; Miller 1994; Im et al. 1998; Sung et al. 1999; Ben-Yakar and Hanson 998b) are two other important factors that have to be addressed in the design of an injection system. Once the fuelair ignition is established, the combustion depends directly on the efficiency of the mixing. In order for self-ignition (and therefore combustion) to be accomplished in a flowing combustible mixture, it is necessary that four quantities have suitable values: static temperature, static pressure, fuel-air mixture, and the residence time at these conditions. The ignition is considered accomplished when sufficient free radicals are formed to initiate the reaction system, even though no appreciable heat has yet been released. When the conditions of spontaneous ignition exist, the distance l i at which it occurs in a medium flowing at a velocity U is: l i = Uτ i, where τ i is the ignition delay time. As the combustor velocity U becomes larger, the ignition requires longer distances. The primary objective of a flame-holder in a supersonic combustion, therefore, is to reduce the ignition delay time and provide a continuous source of radicals for the chemical reaction to be established in the shortest distance possible. In general, flameholding is achieved by three techniques: 1) organization of a recirculation area where the fuel and air can be mixed partially at low velocities, 2) interaction of a shock wave with partially or fully mixed fuel and oxidizer, and 3) formation of coherent structures containing unmixed fuel and air, wherein a diffusion flame occurs as the gases are convected downstream. These three stabilization techniques can be applied in a supersonic combustor in different ways. One of the simplest approaches is the transverse (normal) injection of fuel from a wall orifice (see Fig. 1.3a). As the fuel jet interacts with the supersonic crossflow a bow shock is produced. As a result, the upstream wall boundary layer separates, providing a region where the boundary layer and jet fluids mix subsonically upstream of the jet exit. This region is important in transverse injection flowfields owing to its flame-holding capability in combusting situations, as has been shown in previous

32 CHAPTER 1. INTRODUCTION 10 (a) Bow Shock M >1 Autoignition Zones Fuel (b) Smaller Recirculation Region Weaker Bow Shock (~ Mach Wave) Fuel (c) Bigger Recirculation Region Combined Bow and Step-Induced Shock Fuel FIGURE 1.3 Flow-field schematics of traditional injection/flame-holding schemes for supersonic combustors. a) underexpanded fuel injection normal to the crossflow, b) fuel injection at angle, c) injection behind a sudden expansion produced by a step. publications (Huber et al. 1979; Ben-Yakar and Hanson 998b; Ben-Yakar and Hanson 999a). However, this injection configuration has stagnation pressure losses due to the strong 3-D bow-shock formed by the normal jet penetration, particularly at high flight velocities. Another way of achieving flame stabilization is by means of a step, followed by transverse injection (see Fig. 1.3c). The step creates a larger recirculation area with the hot gases serving as a continuous ignition source. This approach can provide sustained combustion but, like the previously described method, has the disadvantage of stagnation pressure losses and increase in drag due to the low flow pressure base behind the step. On the other hand, it is possible to reduce the pressure losses associated with the

33 CHAPTER 1. INTRODUCTION 11 injection process by performing angled injection (e.g., 60 o or 30 o rather than 90 o ) so that the resulting bow shock is weaker (see Fig. 1.3b). In this approach, jet axial momentum can also contribute to the net engine thrust. Riggins et al. (995a) studied the thrust potential of a supersonic combustor at Mach 13.5 and Mach 17 flight conditions with 30 o flush wall injection of hydrogen and concluded that the major component of thrust potential gain is due to the jet momentum. In our previous work (Ben-Yakar and Hanson 998b; Ben-Yakar and Hanson 999a), autoignition of a hydrogen jet transversely injected into Mach flight enthalpy flow conditions was observed in the upstream recirculation region of the jet and behind the bow shock. However, different experiments (McMillin et al. 1994) performed for similar geometry but at much lower total-enthalpy flow conditions showed that ignition occurred only far downstream of the jet. Based on those observations, angled injection is likely to reduce or eliminate these forms of autoignition and stabilization especially at flight speeds lower than Mach 10. Therefore, it is likely that a new technique will be required to obtain autoignition and downstream combustion stabilization. In recent years, cavity flame-holders, an integrated fuel injection/flame-holding approach, have been proposed as a new concept for flame-holding and stabilization in supersonic combustors (Tishkoff et al. 1997). Cavity flame-holders, designed by CIAM (Central Institution of Aviation Motors) in Moscow, were used for the first time in a joint Russian/French dual-mode scramjet flight-test (hydrogen fueled) (Roudakov et al. 1993). Further experiments (Vinagradov et al. 1995; Ortweth et al. 1996; Owens et al. 1998) showed that the use of a cavity after the ramp injector significantly improved the hydrocarbon combustion efficiency in a supersonic flow. Similar flame stabilization zones, investigated by Ben-Yakar et al. (998a), have been employed within a solid-fuel supersonic combustor, demonstrating self-ignition and sustained combustion of PMMA (Plexiglas) under supersonic flow conditions. In November 1994, NASA contracted CIAM (Roudakov et al. 1996; McClinton et al. 1996) to continue exploring the scramjet operating envelope from dual-mode operation below Mach 6 to the full supersonic combustion mode at Mach 6.5. The proposed combustor design also included two cavity flame-holders (20 mm in depth by 40 mm in axial length and 30 mm by 53 mm). The performance predictions obtained by analytical solutions indicated that these cavities would be quite effective as autoignition and flame-holding devices. Indeed, the recent flight test of this combustor has been

34 CHAPTER 1. INTRODUCTION 12 successfully completed, encouraging further investigation of cavity flame-holders. It is worth noting that, although there is recent interest in cavity flame-holders for supersonic combustors, their application in subsonic combustors goes back to the 1950 s. Probably, the first published investigation of cavity flame-holders is due to Huellmantel et al. (1957), who studied various shapes of cavities to sustain combustion in low speed propane-air flames. The main purpose of this thesis is to summarize relevant known characteristics of cavities in supersonic flows and research efforts related particularly to cavities employed in low- and high-speed combustors. 1.2 Thesis Objectives The ultimate objective of this dissertation is to investigate near-field mixing and flame-holding characteristics of different gaseous fuels such as hydrogen and ethylene injected normally from a single orifice into a realistic supersonic combustor environment. We apply advanced non-intrusive flow diagnostic techniques such as Planar Laser-Induced Fluorescence of OH radicals (OH-PLIF) and schlieren imaging using an ultra-fast-framing rate digital camera. These techniques and the simulation of high speed and high temperature free-stream conditions enable unique observations that were not available in the previous studies. The thesis includes four primary elements: 1. The experimental approach: The goal is to generate a relatively accurate supersonic burner entry condition, namely a radical-free, high total enthalpy air flow. An expansion tube is used to generate three nominal free-stream conditions for flight Mach 8, 10 and 13 regimes. The experimental approach is discussed in Chapter 2 which includes descriptions of the critical parameters that have to be considered in the simulation of a supersonic combustor environment, the facility itself and the measurement techniques. The characterization of the test flow is then presented summarizing determination of the useful test time, core-flow size and boundary layer effects, issues that have to be addressed to fully characterize the flow generated in an expansion tube. The flow visualization techniques are discussed in detail in Chapter Mixing: In Chapter 4, we study the flow features of hydrogen and ethylene transverse jets exposed to high-speed supersonic free-streams at realistic conditions

35 CHAPTER 1. INTRODUCTION 13 leading to high levels of shear. Guided by the observations of these experiments, we continue in Chapter 5 with a more fundamental study looking into the origin of the observed phenomena. The outstanding questions that we investigate are: How do the jet shear layer vortices develop and which parameters control their stability and coherence? What is the contribution of the jet shear layer vortices to the near-field mixing? Does the penetration mechanism only depend on jet-tocrossflow momentum ratio as has been proposed for the last 40 years or is there any other mechanism leading to higher penetration and better mixing properties? 3. Ignition and flame-holding: The ignition and the flame-holding capabilities of a hydrogen jet in high total enthalpy flow conditions are presented in Chapter 6. We study the self-ignition regions in the near-field of the jet in flight Mach 8, 10 and 13 flow conditions using OH-PLIF flow visualization. We also compare the near-field ignition results of a hydrogen transverse jet with an ethylene transverse jet at flight Mach 10 conditions. 4. Cavity flame-holders: In Chapter 7, an extensive overview of cavities, which are considered as a promising flame-holding devices for supersonic combustion, is presented. Open questions impacting the effectiveness of the cavities as flameholders in supersonic combustors are then discussed. Preliminary experimental results are also summarized. The goal is to study the ignition capability of a jetcavity configuration and to observe the differences in the shock wave structures around cavities as the length-to-depth ratio and the geometry of the cavity back wall are changed.

36 Chapter 2 Experimental Aspects Our experimental approach includes the use of an expansion tube to provide a wide range of variability in the freestream conditions with relatively accurate chemical composition. The latter is critical for supersonic combustion studies in the high total enthalpy flows associated with hypersonic air-breathing propulsion systems. Efforts are focused on achieving three operating points, simulating flight Mach 8, 10 and 13 total enthalpy conditions at the entrance of a supersonic combustor. The ability of the expansion tube to provide a steady-flow test time of adequate duration and a core-flow of sufficient size for 2 mm jet-in-crossflow studies is verified. In the following sections, the important parameters that must be considered in the design of a supersonic combustion experiment are discussed and the facility and the test flow characterization techniques are then summarized. Additional test conditions are characterized for fundamental fluid mechanical studies and are presented in Chapter Critical Parameters in Supersonic Combustion Simulation An experimental simulation of a supersonic reacting flow requires the replication of 5 parameters (Heiser and Pratt 1994). These simulation parameters including pressure (p), temperature (T ), velocity (u), characteristic length of the model (L) and gas composition (ν i ) must be manipulated to provide the flight values of certain non-dimensional parameters such as: 14

37 CHAPTER 2. EXPERIMENTAL ASPECTS 15 Mach number: Reynolds number: Damköhler number: Re M u T (2.1) ρ pl u T T 3/2 (2.2) Da L uτ c (2.3) Damköhler number represents the ratio of flow residence time, L/u, through the combustor to chemical time, τ c, and must be larger than 1 to achieve flame-holding and a complete combustion process. For flame-holding considerations ignition delay time, τ i, replaces the chemical time in Damköhler number, τ c = τ i. For a hydrogen-air combustion process, the ignition delay time, varies inversely with pressure because of the two-body reactions and depends exponentially on temperature. As a result, Damköhler number can be related to basic parameters in the following form: Da pl u exp(θ/t ) (2.4) where θ is a characteristic temperature for the ignition time. Consequently, in order to preserve the values of these three non-dimensional parameters it is required to simulate all 5 basic parameters, including temperature, pressure, velocity, model length and the gas chemical composition. However, it is worth noting the following point: If the chemical composition of the flow, its velocity and temperature were to be duplicated, then a constant value of the product pl would satisfy the requirements for simulation of the three non-dimensional parameters. Therefore, from the standpoint of mixing and flame-holding studies a correct simulation of only 4 parameters is essential: chemical composition, temperature, velocity and the product pl. In our experimental approach, we replicate 3 of these 4 parameters: the required burner entry velocity and burner entry static temperature, u 3 and T 3, respectively, according to the values of burner entry Mach number, M 3, estimated in Fig. 1.1a. The use of an expansion tube enables acceleration of the air to total enthalpy conditions

38 CHAPTER 2. EXPERIMENTAL ASPECTS 16 (3-6 MJ/kg air ) corresponding to the Mach 8-13 flight range, without exposing it to high stagnation temperatures ( K). Therefore, the free-stream contains only negligible amounts of radicals, produced only by the incident shock wave. The test gas, first shocked to its maximum temperature ( K), is then accelerated and cooled to the required static temperature ( K). Through this unsteady expansion process, the test gas gains in total temperature and total pressure. Although in our experiments the free-stream flow composition, Mach number and static temperature correspond to typical scramjet combustor entrance values, its static pressure is somewhat below that of actual systems. Table 2.1 summarizes the three nominal test flow conditions, Mach 8, Mach 10 and Mach 13, achieved in the Stanford expansion tube facility. Furthermore, since the characteristic length scale in our experiments is small, about 2 mm (the diameter of the injection orifice), the parameter pl is not sufficiently high to replicate a real combustor environment. This might result in chemical kinetic limitations on the H 2 - air ignition and combustion process. On the other hand, this limitation can be circumvented if an elevated concentration of oxygen is used in the test gas to increase the collision rates as suggested by Bakos et al. (992b). Finally, in the current study we have shown that in high-enthalpy flows, ignition of hydrogen, injected transversely into a free-stream of air, can be achieved in the nearvicinity of the injector, even at low pl values. Therefore, the ignition will be guaranteed at higher pressures as the Damköhler and Reynolds numbers increase linearly with pl in realistic systems. In conclusion, the most important parameters that have to be replicated for supersonic combustion studies are chemical composition, temperature and velocity of the free-stream, and the less important parameter is the product pl. Variation in pressure affects the ignition time linearly, while variation in temperature has an exponential effect through the activation energy (and hence characteristic temperature ignition time, θ) in chemical kinetics.

39 CHAPTER 2. EXPERIMENTAL ASPECTS 17 TABLE 2.1 Test gas (free-stream) flow properties simulating the burner entry conditions of three flight Mach numbers. The corresponding values are from Fig Flight Simulation Mach 8 Mach 10 Mach 13 (1) (2) (3) Initial filling pressures Driven section, He, psig (2.17 MPa) (4.24 MPa) (4.24 MPa) Driven section, 95% N 2 + 5% CO 2, psia (3.10 kpa) (3.45 kpa) (1.04 kpa) Expansion section, He, torr (9.13 kpa) (2.67 kpa) (0.27 kpa) Free-stream conditions Total enthalpy, MJ/kg 2.9 ± ± ± 0.15 Mach number 2.40 ± ± ± 0.07 Static temperature, K Static pressure, atm (65.9 kpa) (32.4 kpa) (4 kpa) Velocity, m/sec (measured) 1800 ± ± ± 50 Test time, µsec (measured) 170 ± ± ± 10 Test slug length, m (velocity test time) Establishment length for laminar boundary layer at L 1 = 50 mm, m Maximum measured recirculation region length, L 2, (d jet = 2 mm) d jet 2 d jet 4 d jet Establishment time for the jet upstream recirculation region, m based on (30 70) L 2 Free-stream Reynolds number at the injection port, Re x = 50 mm Boundary layer thickness upstream of the injection port, mm Shock speeds in the expansion section, m/s (measured) Shock Mach number in the expansion section Maximum temperature that the test gas is exposed to, T 2, K 29,000 22,000 3,

40 CHAPTER 2. EXPERIMENTAL ASPECTS 18 IMACON 468 Ultrafast Framing Camera for Schlieren Imaging (inc. 8 ICCD modules, each 576 x 384) Mirror 2 Dichroic Mirror Focusing Mirror Knife Edge ICCD Camera for OH-PLIF Imaging 578 x 384 Array Driven/Expansion Diaphragm Expansion Section Driven Section Double Diaphragm Amplifiers & Interval Counters Driver Section 8 Channel Data Acquisition System Dump Tank Long duration Xenon Arc Light Source Mirror 1 YM 1200 Nd:YAG Laser HD 500 Dye Laser Image Acquisition Computers HT 1000 Frequency Doubler FIGURE 2.1 Expansion tube facility (12 m in length and 89 mm inner diameter) and imaging system. 2.2 Experimental Facility Expansion Tube The expansion tube facility with its dedicated lasers and optical arrangement is schematically illustrated in Fig The tube is 12 m in length (including dump tank) with an inner diameter of 89 mm, and includes three sections: driver, driven and expansion. The driver section is filled with high pressure helium gas and is separated by double diaphragms from the lower pressure driven section, which is filled with the desired test gas. Mylar film (6.35 µm thick) is used as the diaphragm material at the driven/expansion interface to separate the test gas from low pressure helium gas in the expansion section.

41 CHAPTER 2. EXPERIMENTAL ASPECTS 19 time, sec Quiescent Test Gas 2 Test Gas Behind Incident Shock 3 Expanded Driver Gas 4 Driver Gas 5 Expanded Test Gas 10 Expansion Gas 20 Expansion Gas Behind Incident Shock x-distance, m IR detector 4 Driver Section (He) 1 Driven Section (CO 2 /N 2 /O 2 ) 3 2 incident shock, s1 first disturbance arrival 10 5 reflected rarefaction head rarefaction tail rarefaction head 20 contact surface Expansion Section (He) test time Test Section FIGURE 2.2 Expansion tube distance-time (x-t) diagram calculated for flight Mach 13 condition. Method of characteristics was used to solve the flow gasdynamics properties assuming one-dimensional inviscid theory. Test time is defined as the time that the test gas has uniform flow quantities and determined by the time arrival of the contact surface to the tube exit, and that of the first subsequent rarefaction wave (reflected rarefaction head in our case of high total enthalpy simulations). The operating sequence of an expansion tube is best represented by the distance-time (x-t) diagram shown in Fig A run is initiated by bursting the double diaphragms, which generates a shock wave propagating into the test gas and producing flow of intermediate velocity with an increased pressure and temperature. The shocked test gas is then accelerated by an unsteady and constant area expansion process from the driven section into the lower pressure expansion section, while gaining total temperature

42 CHAPTER 2. EXPERIMENTAL ASPECTS 20 and total pressure. The test gas emerging from the downstream end of the expansion thus has both a higher stagnation enthalpy and higher effective stagnation pressure than the shock tube flow from which it originated. Further detail on the operating cycle of an expansion tube can be found in the review papers of Erdos (1994) and Anderson (1994). A square viewing chamber of cm cross section is mounted at the exit of the expansion tube (see Fig. 2.3). A rake of pitot tubes or an instrumented model with the injection system, is positioned in this test section, which is equipped with an opposed pair of square (13 13 cm) quartz windows for observation and a fused silica slot on top of the chamber for admission of the vertical laser sheet. Six piezo-electric pressure transducers are mounted along the driven and expansion sections for shock speed and wall pressure measurements. An additional transducer, mounted 20.3 cm downstream of the driven/expansion diaphragms, is used to monitor the unsteady expansion process at that location. The expansion section is also equipped with sapphire viewing ports for optical measurements during flow characterization experiments. In those tests, an InSb IR detector (Judson J-10 InSb equipped by a Perry model 720 amplifier) is mounted at a viewing port (see Fig. 2.4) to detect the arrival of the test gas (at the viewing port) through the emission of IR light by small amount of CO 2 (5%) seeded into the test gas (nitrogen). Also, for flow characterization tests, the injection system is replaced with a pitot rake consisting of four pressure transducers across the diameter of the tube as shown in Fig The test gas velocity can then be calculated by considering its arrival time at the viewing port and at the pitot rake. Data from these sensors are recorded at 1 Msample/sec on a PC-based, 8-channel (12-bit) computer-scope. The flow imaging techniques include Planar Laser-Induced Fluorescence of OH radicals (OH-PLIF) and schlieren imaging using an ultra-high-speed framing digital camera. Detailed description of these systems and their synchronization with the expansion tube operation are provided in Chapter Injection System and its Calibration The injection system is positioned right at the exit of the expansion tube inside the test section (Fig. 2.5a). The system consists of a flat plate with an attached high

43 CHAPTER 2. EXPERIMENTAL ASPECTS 21 FIGURE 2.3 Schematic of the test section (27 x 27 cm cross section) where a rake of 4 pitot probes, instrumented with pressure transducers, was located 2.5 cm downstream of the tube exit. The flow history during the expansion tube operation was detected via pitot pressure information. Note that the inner diameter of the tube is 8.9 cm. speed solenoid valve (less than 1 msec response time, General Valve Series 9, Iota One controller) which allows near-constant injection flow rates during the expansion tube test time period. For the results presented here, an under-expanded transverse jet of hydrogen with a 2 mm port diameter has been used. The jet port is located at a distance 30 mm downstream of the tube exit and about 50 mm downstream of the flat plate leading edge. At this location, the boundary layer thickness, developing on the flat plate, is approximately 0.75 mm for the conditions presented in this paper. Table 4.1 summarizes the jet flow exit properties. Calibration of the injected system was performed to determine the stagnation pressure losses through it. This was accomplished by comparing the Mach disk height of an underexpanded jet into still air with a well-known empirical correlation. Schlieren flow visualization (Fig. 2.6) was used to measure the Mach disk height for different pressure ratios. The expected jet Mach disk position, based on the correlation suggested

44 CHAPTER 2. EXPERIMENTAL ASPECTS 22 Sapphire Optical Port Globar Test Gas 5%CO / N 2 2 CS Wall Pressure Transducer Acceleration Gas He Pitot Probe Focusing Lens 40.6 cm 2.5 cm InSb IR Detector Band-Pass Filter m x 104 cm FIGURE 2.4 Optical set-up to measure the test gas velocity, assumed to be equal to the CS - contact surface velocity. IR emission from 5% CO 2 seeded in the test gas nitrogen is collected by an InSb IR detector at the viewing port located at cm from the end of the tube. The test gas velocity can then be calculated by considering its time of arrival at the viewing port and at the pitot rake. by Ashkenas and Sherman (1966) as a function of jet stagnation pressure (P tot,jet ) to effective back pressure (P eb ) ratio, is given by: y 1 d jet = 0.67 ( Ptot,jet P eb ) 1/2 (2.5) On the basis of this correlation, measurements indicated a stagnation pressure loss of 48% for hydrogen injection and 41% for ethylene injection during valve operation (note that the fuels were supplied from flow lines of different length). In addition, the valve actuation time and the tube firing have to be synchronized such that the jet is fully developed by the time the steady test flow conditions are obtained. Within that constraint, the time interval between the valve actuation and the test gas arrival should be short enough to avoid significant changes in the expansion section initial pressure. To determine the jet development time, schlieren imaging was used to observe the temporal development of the jet. This combined with the traces obtained using a fast response pressure transducer located at the jet exit, allowed the determination of the optimum valve actuation time ( 1.5 msec before start of test time).

45 CHAPTER 2. EXPERIMENTAL ASPECTS 23 (a) (b) FIGURE 2.5 Schematic of a) Injection system, b) cavity/injection plate system Cavity/Injection Plate The cavity/injection system (Fig. 2.5b) is designed with different cavity and jet inserts to systematically study the following configurations: 1) 90 o and 30 o flush wall gaseous fuel (hydrogen, ethylene) injection, 2) cavities with length-to-depth ratio of L/D=3,5,7 with 90 o, 60 o and 30 o rear-walls, 3) cavities with 30 o upstream fuel injection. In addition, a miniature pressure transducer (PCB dynamic piezo transducer, model 1105B12) was installed inside the cavity to measure the pressure oscillations and to monitor the flow establishment time. 2.3 Test Flow Characterization in the Flight Mach 8-13 Range There are several issues that have to be addressed to fully characterize the properties of a supersonic flow generated in an impulse facility. Of particular concern are the determination and characterization of flow conditions, the steady-flow test time, the

46 CHAPTER 2. EXPERIMENTAL ASPECTS 24 (a) (b) (c) (d) (e) Jet Boundary Mach Disk Barrel Shock y 1 P j, M =1 j FIGURE 2.6 Schlieren visualization of an underexpanded gaseous injection into still air. (a)-(c) hydrogen (d)-(e) ethylene jets. The exposure time of the images was 3 musec. Mach disk height, y 1, was measured for different pressure ratios, P j/p eb, to calibrate the injection system. core-flow size and the boundary layer effects on the flow properties Flow Conditions A variety of flow conditions can be easily achieved using an expansion tube by simply changing the initial filling pressures and the speed of sound of the gases at different sections of the expansion tube. In our facility, helium gas was used in the expansion and driver sections with a maximum filling pressure of 4.24 MPa (600 psig) in the driver section. The selection of correct initial pressures to achieve the required flow conditions, however, is not straight-forward as the expansion tube combines two shock tubes in tandem. We have, therefore, calculated the flow conditions that can be achieved in an

47 CHAPTER 2. EXPERIMENTAL ASPECTS 25 expansion tube as a function of initial filling pressures. Figure 2.7 presents maps of these flow conditions estimated using simple one-dimensional inviscid theory (see Appendix A for a given driver filling pressure of 4.24 MPa (600 psig). These maps provide guidelines in the selection of the initial pressures for different conditions of interest. Note the blank (forbidden) regions in the maps of Fig These are the regions where the shocked test gas cannot expand into the expansion section as the pressure in the expansion section is higher than the pressure of the shocked test gas. In a conventional shock tube, for a given driver initial pressure, one must decrease the driven section pressure to generate flows with higher velocities, temperatures and Mach numbers. As the driven section initial filling pressure decreases the shock-induced static pressure decreases as well. In expansion tubes, on the other hand, the flow velocity and static pressure do not vary significantly as the driven initial pressure is changed for a given driver and expansion section pressures. As shown in Fig. 2.7, different velocities can be achieved primarily by manipulating the expansion section initial pressure. For lower expansion section pressures, the flow accelerates to higher velocities and expands to lower pressures. In our experiments, we have used the maps in Fig. 2.7 to simulate the required burner entry conditions of flight Mach 8, 10 and 13 based on the estimated values from Fig Simulation of our flight Mach 10 condition, for example, corresponds to 4 MJ/kg of total enthalpy and a Mach number of about 3.5. Therefore, according to the maps in Fig. 2.7, initial filling pressures of the expansion and driven sections were determined to be 2.67 kpa (20 torr) and 3.45 kpa (0.50 psia), respectively, to simulate flight Mach 10 total enthalpy condition at the selected driver initial pressure of 4.24 MPa (600 psig). Initial pressures required to simulate flight Mach 8 and 13 were chosen in a similar manner and are marked on the maps of Fig As discussed in the previous section, our experimental approach includes correct replication of the velocity and temperature of the air entering the combustor. The static pressure in our experiments, on the other hand, is lower than the expected values in a supersonic combustor at those flight conditions. However, the correct replication of pressure is not as crucial as the simulation of the required total temperature and burner entry Mach number, in the basic study of ignition and flame-holding processes of different injection schemes. Pressure dependence of the ignition process can be extrapolated

48 CHAPTER 2. EXPERIMENTAL ASPECTS 26 (a) Driven Gas (Nitrogen) Initial Pressure, P 1 (psia) Driven Gas (Nitrogen) Initial Pressure, P 1 (psia) K 400 K 600 K 0.05 atm 0.1 atm Mach K 0.2 atm 0.3 atm 0.4 atm 0.6 atm Mach K 1800 K 2200 K 2400 K 0.8 atm 1 atm 1.4 atm Mach Expansion Gas (Helium) Initial Pressure, P 10 (torr) 3400 m/s 3200 m/s 3000 m/s 2800 m/s MJ/kg Mach MJ/kg (b) 2600 m/s 2400 m/s 2200 m/s 2000 m/s Mach 10 4 MJ/kg Expansion Gas (Helium) Initial Pressure, P 10 (torr) m/s Mach 8 3 MJ/kg FIGURE 2.7 Maps of estimated test gas (nitrogen) conditions (state 5) at the exit of an expansion tube: a) pressure and temperature, b) total enthalpy and velocity of the test gas are plotted for different initial driven and expansion section pressures. Calculations are performed using the inviscid 1D equations for a given driver pressure of P 4 = 600 psig (helium). Note that the effective filling pressure of the driver section is taken as P 4,eff = 686 psig, as its inner diameter (10.2 cm) is larger than that of the driven and expansion sections (8.9 cm). This area difference is accounted for in the curves presented above.

49 CHAPTER 2. EXPERIMENTAL ASPECTS 27 to higher pressures since the Damköhler number is approximately proportional to pressure in hydrogen-air combustion systems. Furthermore, the oxygen concentration can be increased in the test gas if the combustion process is rate-limited due to low static pressure. Table 2.1 summarizes the initial filling pressures of the three test conditions characterized in the Stanford expansion tube. Other factors, such as the maximum available injector pressure, were taken into account in determining these test flow conditions. The initial driver pressure for the Mach 8 condition, for example, was chosen to be 2.17 MPa (300 psig) instead of 4.24 MPa (600 psig) to provide reasonable penetration of the fuel jet. Penetration of fuels injected transversely into supersonic flows is known to be strongly correlated with the jet-to-free-stream momentum flux ratio (J) defined as J = ( ρu 2 ) jet (ρu 2 ) = ( γpm 2 ) jet (γpm 2 ) (2.6) where the subscript jet corresponds to the jet exit conditions and corresponds to free-stream conditions ahead of a bow shock. In our experiments, the jet-to-free-stream momentum flux ratio was typically J = Measurement of Flow Properties The test flow conditions (pressure, temperature, velocity, Mach number and test time), presented in Table 2.1, were determined using the combined data of wall pressure, pitot pressure, and IR emission. Briefly, the test flow characterization was performed in the following stages: 1. the shock speeds at the driven and expansion sections were measured using six piezo-electric pressure transducers mounted in the tube, 2. gasdynamic conditions of the post-shock test gas in the driven and expansion sections were obtained from the measured shock speeds by a 1-D frozen-chemistry code using standard thermochemical data, 3. the test gas temperature and sound speed at the exit of the tube were then calculated assuming isentropic expansion of the shocked test gas in the driven section to the expected value of the post-shock static pressure of the acceleration gas in the expansion section,

50 CHAPTER 2. EXPERIMENTAL ASPECTS the velocity of the contact surface was deduced by measuring the time interval between its arrival at the IR detector port and the pitot probe located at the exit of the tube, 5. test gas flow velocity was then estimated by equating the contact surface velocity with that of the test gas immediately after it. An example of IR emission from CO 2 seeded in the test gas, together with pitot and static pressure traces measured in flight Mach 10 flow simulation is given in Fig Note that in characterization experiments, instead of air, a mixture of 95% N 2 +5% CO 2 was used as a test gas, which provides an effective molecular weight (28.8 gr/mole) equivalent to that value of the air. Based on the time history of the pitot pressure at the tube exit, it is possible to identify the arrival of the shock wave, the period of expansion section helium flow, the helium/test gas contact surface, and the steady flow test time. It is evident that the contact surface is not a perfectly sharp boundary between helium and test gas. Instead test gas concentration seems to increase over a period of time. A similar trend can be observed in IR emission which can be detected with the arrival of the helium/test gas contact surface (CS) to the viewing port. The intensity of IR emission increases through the CS passage as the CO 2 concentration in the CS increases. At the end of the steady test gas the emission intensity decreases as the test gas cools down with the arrival of the expansion waves (reflected rarefaction tail). During the steady flow test time, we observe that the IR emission is, however, not as constant as the traces of the pitot and static pressures. Instead, an initial peak followed by a monotonic increase are observed in the Mach 10 flow characterization (see Fig. 2.8). In general, the intensity of IR emission increases with the volume of emitting gas, its temperature and concentration. Since the IR emission is the integrated emission across the tube including the wall boundary layer, it does not reflect the properties of the core-flow only. The monotonic increase of the intensity (subsequent to the drop from its initial peak value) can be explained as due to an increase in the volume of the hotter CO 2 in the growing boundary layer, while the initial peak signal can be attributed to the small hot region formed by the reflected shock during the break of the helium/test gas diaphragm.

51 CHAPTER 2. EXPERIMENTAL ASPECTS 29 IR Emission CS Arrival Test Gas Normalized Pitot Pressure Shock Wave Arrival 440±5 sec x=104 cm HeliumFlow Time ~455 sec CS CS Velocity: 2360±25 m/s Test Time ~270 sec Rarefaction Wave Normalized Static Pressure Normalized Mach number Shock Wave Arrival CS Test Gas Rarefaction Wave Shock Wave Arrival Test Gas Rarefaction Wave time, msec FIGURE 2.8 Example of IR emission, pitot pressure, wall pressure records and the Mach number variation based on the pitot-to-static pressure ratios, as a function of time for the Mach 10 flow condition. t = 0 represents incident shock arrival at the pitot probe, placed 2.5 cm downstream of the tube exit, while the wall pressure transducer and IR detector are positioned 40.6 cm and cm upstream of the tube exit, respectively (see Fig. 2.4). Note that the time scale of the static pressure trace is shifted by 235 µs to match the shock arrival at the pitot probe.

52 CHAPTER 2. EXPERIMENTAL ASPECTS 30 The steady test period was taken to be the time over which the pitot pressure changed by no more than ± 5% from the average value. Accordingly the measured test time was approximately 270 µs. The steady test period is always limited by the arrival of waves, either the unsteady expansion waves (rarefaction tail) or the reflected waves (reflected rarefaction head) from the intersection of the driver gas interface with the unsteady expansion waves. In any case, the arrival of waves is clearly identifiable as the test gas pressure rises sharply and the IR emission from cooling test gas begins to decrease. An average value of 2360 m/s for the test gas flow velocity was measured over the 104 cm length between the IR emission port and pitot rake (see stages 4 and 5 described above). The test gas velocity was also estimated from measured shock speeds in the expansion section using inviscid 1-D theory, resulting in 2130 m/s. The measured velocity (2360 m/s) therefore exceeds the estimated inviscid value based on shock speed by approximately 10%, as may be explained by the acceleration of the flow due to the growing boundary layer on the tube walls. The Mach number of the test gas was calculated using the measured contact surface velocity and calculated sound speed as described in stage 3. The results indicated an exit Mach number of 3.38±0.04 for the flight Mach 10 condition. Mach number variation in the test gas can also be obtained from the pitot to static pressure ratio as shown in Fig It is worth noting that while the static pressure of the test gas rose sharply after the arrival of the reflected rarefaction wave, the Mach number of the flow varied little over the following µs. Figure 2.9 presents an example of flight Mach 13 flow characterization traces which have features similar to those of flight Mach 10 traces. The characterization results show a relatively long test time of 400 µs in the case of flight Mach 13 simulation; this is much larger than the ideal values (180 µs). The test gas velocity was measured to be 3200 m/s, about 23% faster than the shock-induced ideal flow velocity. This acceleration of the test gas is again believed to be a result of the boundary layer developed on the tube walls. Shorter test times ( 170 µs) are observed for low total-enthalpy conditions as shown in Fig presenting an example of flight Mach 8 flow characterization traces. Note that the static pressure stays constant for a longer period of time although the pitot pressure increases significantly by the arrival of the first disturbances.

53 CHAPTER 2. EXPERIMENTAL ASPECTS 31 IR Emission CS Test Gas Normalized Pitot Pressure Normalized Static Pressure Shock Wave Arrival 325±5 sec x=104 mm CS Helium Flow Time ~180 sec CS Velocity: 3200±50 m/s Test Time ~400 sec Rarefaction Wave Shock Wave Arrival CS Test Gas Rarefraction Wave time, msec FIGURE 2.9 Example of IR emission, pitot pressure and wall pressure traces as a function of time for the Mach 13 flow condition. In Fig. 2.11, we compare the measured values of the free-stream velocity with the expected values estimated using the measured shock speeds in the expansion section. The results indicate that the measured velocities are always larger than the expected ideal values. In addition, as the flight Mach number increases and therefore the static pressure decreases the measured velocity approaches the shock velocity. These observations suggest that boundary layer effects are significant, causing the flow to be non-uniform. As a result of the viscous effects, the shock slows down and the contact surface (test gas) accelerates. At the limit of a sufficiently long expansion section, boundary layer effects would cause the shock and the contact surface to equilibrate to a constant velocity. Accordingly, an additional calculation, taking into account the viscous effects based on Mirels solution (Mirels 1963; Mirels 1966) for post-shock boundary layers, has been

54 CHAPTER 2. EXPERIMENTAL ASPECTS 32 IR Emission CS Arrival Test Gas Normalized Pitot Pressure Shock Wave Arrival Velocity: 1800±20 m/s 578±5 sec, x=104 cm HeliumFlow Time ~725 sec CS Test Time ~170 sec Rarefaction Wave Normalized Static Pressure Shock Wave Arrival time, msec CS Test Gas Rarefaction Wave FIGURE 2.10 Example of IR emission, pitot pressure and wall pressure traces as a function of time for the Mach 8 flow condition. performed, and will be discussed in the following section Boundary Layer Effects on Test Time The uniform flow time of test gas at the expansion tube exit is defined as the test time; this period begins with the arrival of the driven/expansion contact surface at the tube exit, and ends with arrival of the first rarefaction wave (in our case, the rarefaction head reflected from driver/driven contact surface as shown in the x-t diagram in Figs. 2.2 and Under ideal conditions, where no wall effects exist, the shock wave and the contact

55 CHAPTER 2. EXPERIMENTAL ASPECTS Measured contact-surface velocity Estimated shock-induced velocity Measured shock velocity Velocity, m/s Mach 8 P=0.65atm Mach 10 P=0.32atm Mach 13 P=0.04atm Expansion Section Shock Mach Number, M s,2 FIGURE 2.11 Comparison of the measured contact-surface velocity (test gas velocity) with the shockinduced gas velocity estimated using the measured shock speeds in the expansion section. surface would move with constant velocities and the flow between them would be uniform. However, in a real shock tube, the flow becomes non-uniform as the boundary layer develops at the tube walls. The presence of a wall boundary layer causes the incident shock to decelerate and the contact surface to accelerate. Consequently, the time duration of the flow between the shock and the contact surface (expansion section helium flow time) is reduced. In conventional shock tubes, therefore, the test time is reduced as a result of the boundary layer growth. In expansion tubes, on the other hand, the effect of the boundary layer in accelerating the contact surface can actually lead to an increase in test time as observed in our experiments. We have measured 400 µsec of steady test time at the Mach 13 condition, while only 180 µsec of test time is expected based on 1-D inviscid calculations (see x-t diagram in Fig. 2.12a). To study the boundary layer effects on the test time, we have performed an improved calculation taking into account the viscous effects based on Mirels boundary layer solution. The predicted contact surface velocity from this solution was implemented in x- t diagram calculations and the results are plotted together with 1-D inviscid calculations in Fig. 2.12a. The calculations (dashed lines in Fig. 2.12a) resulted in a test time of

56 CHAPTER 2. EXPERIMENTAL ASPECTS 34 time, sec (a) Mach 13 condition rarefaction head (measured test time = ~ 400 sec) Inviscid Solution Mirels' Model inc. boundary layer rarefaction tail reflected rarefaction head Incident Shock (Measured) Inviscid CS CS with Boundary Layer ideal test time 180 sec 450 sec x-distance, m time, sec (b) Mach 10 condition (measured test time = ~ 270 sec) Inviscid Mirels' Model inc. boundary layer Inviscid CS CS with Boundary Layer Incident Shock (Measured) ideal test time 140 sec 380 sec x-distance, m FIGURE 2.12 X-t diagrams for the expansion section only. 1-D inviscid calculations are plotted in straight lines and results applying Mirels model to include the boundary layer effects are plotted in dashed lines. One can see the improved test time as a result of the contact surface (CS) acceleration due to the developing boundary layer behind the incident shock in the low pressure expansion section helium flow. The incident shock velocity was measured and assumed to be constant along the expansion section.

57 CHAPTER 2. EXPERIMENTAL ASPECTS µsec, agreeing well with the measured value of 400 µsec. This effect can be explained by the fact that the contact surface, accelerated by viscous effects, arrives at the test section earlier than if viscous affects were negligible. Therefore, the test time duration is increased as the time of arrival of the first flow disturbance is delayed relative to the contact surface arrival. An additional interesting result can be observed from the x-t diagrams: the length of the current expansion section is seen to be optimum to achieve a maximum test time duration for flight Mach 13 simulation as the arrival of both the rarefaction wave and the reflected rarefaction head overlap at the end of the tube (see the dashed lines in Fig. 2.12a.) Any variation in the length of the expansion section will cause the test time to decrease. By contrast, the x-t diagram calculated for the Mach 10 condition (Fig. 2.12b) demonstrates that the expansion section length is longer than its optimum value. A 30% longer test time could have been achieved if the expansion section was 1.8 m shorter than its current length. The calculation based on Mirels solution also gives an improved estimation of the expansion section helium flow time of 170 µsec, which compares well with that inferred from the pitot pressure trace, 180 µsec (within 5.5% accuracy). The test gas velocity obtained from this calculation, 3300 m/s, shows improved agreement with the measured value of 3200 (within 3%). These results, summarized in Table 2.2, confirm the importance of boundary layer effects and demonstrates that the viscous calculations based on Mirels model provide good estimates of the test time and the test gas velocity at the low pressure flight Mach 13 condition. Mirels solution, performed for the flight Mach 10 condition (see Fig. 2.12b), results in a less accurate prediction of flow properties. As summarized in Table 2.2, the calculated test time based on Mirels solution overpredicts (by 41%) the actual test time. The reason for the overprediction is that the initial filling pressure in the expansion section is relatively high (20 torr), and therefore the wall boundary layer behind the shock wave is not fully developed as assumed in Mirels solution. In addition, we have assumed that the boundary layer properties of the test gas were similar to the properties of the expansion section helium gas, which can be significantly different as the pressure increases. We have shown that the flow test time is affected by the boundary layer development in the expansion section tube walls in such a way that the steady test time duration is

58 CHAPTER 2. EXPERIMENTAL ASPECTS 36 TABLE 2.2 Summary of measured, ideal (inviscid 1-D) and predicted (based on Mirels solution) properties of test gas for Mach 10 and 13 flow conditions. Test Time Test Gas Helium Flow Velocity Time µsec m/s µsec Measured 270 ± ± ± 2 Mach 10 Ideal Mirels Measured 400 ± ± ± 2 Mach 13 Ideal Mirels increased. Longer test times are important in supersonic flow experiments as the length of the test period determines the maximum model length for which steady flow can be fully established Core-Flow Size The other parameter which limits the maximum model length is the radius of the axially uniform flow called the useful core-flow. While the test time increases as the boundary layer on the tube walls develops, the useful core-flow size, on the other hand, becomes smaller. The thickness of the boundary layer developed in the acceleration section can be large in comparison to the tube radius because of the low filling pressures of the expansion section. In our experiments, we characterized the useful core-flow size by measuring the pitot pressure at different radial and axial locations. The results, plotted in Fig. 2.13, show that at 12.7 mm (0.5 ) away from the tube exit, pitot pressure varied only ±5% within a 25 mm core-flow diameter for the three conditions that we have studied. As we moved the pitot probe downstream of the tube exit, the core-flow size at the Mach 13 condition did not change significantly. On the other hand, the flight Mach 10 and 8 conditions resulted in a deceased core-flow size as we moved away from the tube exit. At the Mach 8 condition, the pitot pressure at the center line of the tube decreased by 15%

59 CHAPTER 2. EXPERIMENTAL ASPECTS 37 at 63.5 mm (2.5 ) downstream of the tube exit. This is expected for the flight Mach 8 condition, since the Mach number of the free-stream is only about 2.4, corresponding to a steeper Mach wave angle. In conclusion, the results indicate the existence of a 25 mm core-flow diameter that is suitable for near-field studies of transverse fuel jets injected from a 2mm diameter orifice. The injection plate is positioned as close as possible to the tube exit and 6.4 mm (1/4 ) below the centerline to allow a maximum field of study of the jet Flow Establishment Time Undesirably, part of the steady test time is consumed during the flow establishment process for the model under investigation. Correlations are available in the literature (Rogers and Weidner 1993; Jacobs et al. 1992; Davies and Bernstein 1969; Holden 1971) for predicting the flow establishment times for different flow features of the model. In general, the test slug length, defined as the distance traveled by the volume of test gas during the period of steady test time, must be larger than the characteristic length of the flow device or process by some multiple. The criterion for the establishment of flow over a flat plate requires that the test slug length, i.e., the product of test time and gas velocity, t u, satisfies (Davies and Bernstein 1969). 2 L 1, for turbulent boundary layer t u = 3 L 1, for laminar boundary layer (2.7) where L 1 is the distance from the edge of the plate. The flow establishment criterion for separated flows, on the other hand, is required to be tens of times larger than the characteristic recirculation region length L 2, as given by (Holden 1971) for the wake of sphere t u = for separated flows 30 L 2, based on pressure measurements 70 L 2, based on heat transfer measurements (2.8) Note that L 2 is the length of the recirculation region and therefore smaller than the characteristic distance L 1 by more than an order of magnitude, resulting in similar flow establishment times.

60 CHAPTER 2. EXPERIMENTAL ASPECTS 38 (a) 1.0 Flight Mach 13 condition Normalized pitot pressure useful core flow ~ 25 mm Distance from the tube exit 12.7mm (0.5") 38.1mm (1.5") 63.5mm (2.5") Free-stream flow radius, mm (b) Normalized pitot pressure Flight Mach 10 condition Distance from the tube exit 12.7mm (0.5") 38.1mm (1.5") 63.5mm (2.5") Free-stream flow radius, mm (c) Normalized pitot pressure Flight Mach 8 condition Distance from the tube exit 12.7mm (0.5") 38.1mm (1.5") 63.5mm (2.5") Free-stream flow radius, mm FIGURE 2.13 Useful core-flow size in flight Mach 13, 10 and 8 conditions, (a), (b) and (c), respectively, determined by measuring the radial variation of pitot pressure at different distances from the tube exit.

61 CHAPTER 2. EXPERIMENTAL ASPECTS 39 Based on these correlations, we have estimated two flow establishment times in our experiments and summarized the conclusions in Table 2.2. First, the boundary layer flow establishment time was calculated by assuming laminar flow, as the local Reynolds numbers at the injection port were relatively small. Second, the establishment time of the recirculation region upstream of the injection port was estimated. In general, the bow shock around the jet interacts with the approaching boundary layer and causes its separation. This recirculation region length confined with the separation shock wave was measured from schlieren images and used as the characteristic length, L 2, of the separation zone. The estimated flow establishment values indicated that the test slug (1.28 m) for the flight Mach 13 condition is fairly long as compared to the predicted flow establishment times (0.15 m for boundary layer and 0.56 m for the recirculation region heat transfer establishment time = 0.71 m in total). Therefore, about 0.60 m of a test slug or 235 µs of test time is still available for measurements. By contrast, at the flight Mach 8 condition, flow establishment seems to consume most of the steady test time. However, it is worth noting that the expansion section helium flow prior to the test gas arrival contributes to the establishment of jet flow even though it is not considered as a part of the test time. Furthermore, the useful test time in our experiments is determined using the pitot probe after the flow is established around it. Therefore, the estimated flow establishment times given in Table 2.2 are expected to be larger than the actual flow establishment times, so that even at the flight Mach 8 condition, part of the slug length is useful for measurements after the flow-field is established. To summarize, as the flight Mach number in our facility increases, a longer test slug length becomes available for measurements. This is a result of the long test times achieved with high-total enthalpy, low-pressure conditions and the fact that higher speed implies faster flow establishment. 2.4 Summary We used the Stanford expansion tube to generate high total enthalpy flow conditions in the Mach 8-13 flight range. A variety of supersonic flow conditions are also simulated in the facility for fundamental mixing studies of transverse jets. These conditions are summarized in Chapter 5. See also Appendix B for maps of ideal flow conditions that

62 CHAPTER 2. EXPERIMENTAL ASPECTS 40 can be achieved using an expansion tube. The maps are calculated for different freestream gas (Nitrogen, Helium and Argon) and for two different initial driver pressures (P 4 = 300 and 600 psig).

63 Chapter 3 Flow Visualization Techniques Much of our basic understanding of the behavior of flows has come from flow visualization. One of the best examples of this is the work of Brown and Roshko (1974), which shows the existence of large scale structures in a two-dimensional shear layer using shadowgraphs. A clear visualization of axisymmetric shear layers, as the one present in the periphery of transverse jets in crossflows, is more difficult to achieve due to three dimensional effects, especially when the jet characteristic length is as small as d=2 mm. The flow visualization of transverse jets in supersonic crossflow studied in this work was obtained using an ultra-high-framing-rate schlieren system and Planar Laser Induced Fluorescence (PLIF) of the hydroxyl radical (OH). The ability to image the flow with good temporal and spatial resolution is particularly important, because of the high velocities of the flows and also because it provides a means of examining the instantaneous turbulent structures that control the mixing and combustion processes. 3.1 Ultra-Fast Framing Rate Schlieren The ability to capture the time evolution of unsteady supersonic flows is critical to their understanding. Non-intrusive visualization techniques, such as schlieren and laserbased planar flow imaging, are powerful and commonly used optical methods. However, tracking the structural evolution of high-speed flows requires acquisition of images at fast (typically MHz) repetition rates. In addition, very short exposure times ( ns) are required to resolve instantaneous features. As the spatial resolution is increased to 41

64 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES 42 avoid the blurring of turbulent structures, the exposure time must be reduced and the repetition rate increased. It is challenging to fulfill the temporal resolution requirements of high-speed imaging while maintaining meaningful spatial resolution for supersonic flows. Before we introduce previous and current efforts towards the development of supersonic flow visualization at ultra-fast-framing rates, it is useful to present the motivation behind such efforts. Of great interest is the study of shear layers formed at the interface of two parallel or skewed fluid streams. Since Brown and Roshko (1974) demonstrated that large-scale coherent structures are dominant in subsonic shear layers and that their structure and evolution control the mixing process, many researchers have concentrated their efforts on measuring large-scale convection characteristics of shear layers in supersonic flows. However these studies were limited to double-pulse visualization techniques, such as double-pulse schlieren by Papamoschou (1991), double-pulse Mie/Rayleigh scattering by Elliott et al. (1995), double-exposure planar laser-induced fluorescence (PLIF) of acetone by Papamoschou and Bunyajitradulya (1997) and Fourguette et al. (1991) and double-pulse imaging using simultaneous aceton/oh-plif by Seitzmann et al. (1994). Those studies focused on measuring the convective velocity (U c ) of large-scale structures by capturing a maximum of two images that are temporally correlated. Of particular interest to our research is the shear layer formed when a gaseous jet interacts with a supersonic crossflow stream, an example of skewed shear layers. This is also a common fuel injection scheme in practical systems, such as a scramjet, and therefore fundamental study of its mixing process is important. Although there have been numerous studies of the shear layer properties of two parallel streams, there have been relatively few works on jets in supersonic crossflow. Among those studies only Gruber et al. (997a) studied the large-eddy convection characteristics of jets in supersonic crossflow, again by capturing two consecutive Rayleigh/Mie scattering images. Their results revealed the convection characteristics of helium and air jets injected into a Mach 2 crossflow. The highly compressible helium jet exhibited larger convection velocities in the near-field of the injection than the air jet of low compressibility. However, the accuracy of the velocity measurement was only about ±10%, as the minimum separation between the laser pulses was limited to 1 µsec. It is crucial to characterize the full life cycle of the flow-field. For example, the largescale eddies, formed periodically at the early stages of shear layer development, undergo

65 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES 43 structural evolution as they convect downstream while entraining, pairing, engulfing, stretching and tearing. Therefore, measurement of the formation frequency of these eddies and their evolution should aid understanding of the origin and formation of the jet vortical structure and, hopefully, provide tools in controlling its mixing properties. Such studies require application of an ultra-high-speed imaging system. In our experiments, we therefore selected a new fast-framing imaging system capable of recording 8 consecutive high-resolution schlieren images at rates up to 100 MHz. Our goal is to study the time evolution and convection properties of large-scale structures present in the shear layer of the jet/free-stream interface. It is important to understand how these structures and growth rate vary as flow conditions and fuel types are changed. Moreover, most of the earlier jet studies were carried out in blow-down wind tunnels where the free-stream conditions were usually of low temperature and therefore with relatively low speeds. Free-stream velocity in the experiments of (Gruber et al. 997a), for example, was about 515 m/s. In our experiments we employ an impulse facility which can generate realistic conditions of a typical supersonic combustor with high velocities ( m/s) and high static temperatures (1300 K) (Ben-Yakar and Hanson 2000). However, these facilities have short test times ( ms) during which there is only a brief opportunity to perform a flow-diagnostic measurement. As a result, application of MHz imaging in impulse facilities becomes almost necessary as we can obtain multiple images that are also time-correlated. In our facility, the amount of data obtained per experiment is, therefore, increased by at least 8 times by the new ultra-high-speed imaging system. In the following sections, we present the components of the ultra-fast schlieren imaging system and discuss issues of resolution, timing, synchronization and imageprocessing techniques Previous and Current High Speed Imaging Efforts High-speed imaging requires two components: a camera which can acquire at high framing rates and a light source with either long duration time or high pulsing rate. Early flow visualization at high-speed framing rates was performed using a rotating mirror camera. Mahadevan and Loth (1994), for example, utilized a rotating mirror camera to temporally resolve compressible mixing layer structures at approximately

66 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES khz using schlieren and Mie scattering. A Xenon flash system with square pulse durations up to 200 µs was used as a continuous light source. The quality of images was poor because of the low resolution and blurring of the images due to the convection of the flow structures. Patrie et al. (1993, 1994), on the other hand, performed 3-D snapshots of instantaneous flow structures using Mie scattering and laser-induced fluorescence of flames and turbulent jets. Their camera system included a high-speed image converter coupled by fiber optics to a CCD camera. Up to 20 sequential planar images were collected at the rate of 10 MHz. The output of the system was a single digital image with pixels, containing the sequence of 20 images, namely, each image was approximately pixels for a 10 image set. This imaging system required post-processing algorithms to correct the spatial distortion of the images caused by electrodynamic interaction of the photoelectric currents within the camera. Island et al. (1996) used the same imaging system to study the three-dimensionality of supersonic mixing layers. The planar illumination for a 3-D scan was provided from a 2 µs pulse duration of a flashlamp-pumped dye laser with a pulse energy of 3 J at 590 nm. The laser output was reflected from a high-rpm rotating mirror. The pixel resolution was approximately 1 mm resolving only the largest scales of mixing. Recently, several researchers including Huntley et al. (2000), Thurow et al. (2000) and Wu et al. (000b) reported their efforts toward MHz-rate digital planar flow visualization using a prototype CCD camera manufactured by Princeton Scientific Instruments, Inc. (PSI) with pixel format of or This camera has a 32 image storage buffer built onto the image sensor chip itself, and can frame at rates up to 1 MHz. The drawback of the PSI camera is the low pixel resolution of images (the pixel size is 50 µm 50 µm) and the low fill factor of the light sensitive area which is about 14%. Princeton Scientific Instruments, Inc has attempted to develop this unique high-speed framing CCD camera under a government SBIR contract. However, further development of this prototype camera is currently not possible because of the manufacturing difficulties of the chip. The PSI camera is typically paired with a laser working in a pulse burst concept (Wu et al. 000b). The burst train of pulses was formed by applying a high-speed Pockel Cell slicer to the long duration output (200 µs) from a continuous wave (CW) Nd:YAG ring laser. The energy of each of the individual pulses comprising the train was

67 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES mj, with minimum separations of 1ms (1 MHz). With this energy level of pulses the researchers succeeded in performing CO 2 -enhanced Filtered Rayleigh Scattering measurements in supersonic flows. As summarized above, since the early 90 s, several research groups attempted to perform high speed imaging of supersonic flows. Their results showed the excellent potential of high speed 2-D and 3-D measurements. However, their imaging systems were usually of low resolution and low light sensitivity. In the following sections, we present initial results with a new commercial imaging system that can capture images (free of distortions) at rates up to 100 MHz. The system includes a fast framing rate camera (IMACON 468) combined with a long duration, high intensity light source (Xenon flashlamp) to acquire 8 consecutive schlieren images of supersonic flows with pixel resolution. In this system, resolution and blurring of the images can be controlled by adjusting the image exposure time which can be as small as 10 ns. A similar fast framing camera system with a maximum repetition rate of 1 MHz is also being used by Kaminski et al. (1999) to study turbulent reacting flames. The system performance was demonstrated at 8 khz repetition rate using a custom-built unit (BMI) of four double-pulsed Nd:YAG lasers with an accompanying dye laser. This laser system allowed variable inter-framing times of 25 to 145 µsec High-Speed Schlieren Imaging Components The ultra-fast-framing schlieren system illustrated in Fig. 3.1 comprised three components: 1) a high-speed framing camera, (Imacon 468, manufactured by Hadland Photonics), 2) a long duration light source, (Xenon flash-lamp), and 3) mirrors and knife edge in a standard Z-arrangement. The Imacon 468 consists of 8 independent intensified CCD cameras for high-speed framing that can capture 8 consecutive images with variable exposure and interframing times down to 10 ns. The single optical input is divided uniformly and distortion-free by a special beam splitter into 8 different intensified CCD modules, each with a array of µm size pixels. The light source is a high intensity Xenon flash discharge unit (Hadland Photonics model flash system with an extension to 200 µs duration). The unit has three

68 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES 46 Xenon Flash Light Source Power Supply Mirror 1 Mirror 2 Test Section Knife Edge Focusing Lens IMACON 468 Input Trigger Control Computer & Image Monitor Camera Fiber Optic Link FIGURE 3.1 Schlieren imaging set-up. ranges providing 20 µs, 50 µs and 200 µs durations of the light source with intensity levels of each being 125 J, 375 J and 700 J per pulse, respectively. In the optical set-up, two f/10, 200 cm focal length concave mirrors are used to collimate the light through the test section, and then refocus it onto a knife edge (razor blade). This knife edge (KE) at the focal point of the second schlieren mirror is used to partially cut off the deflected rays for observing the schlieren effect (visualization of the density gradients). Blocking more of the light by moving the KE transverse to the optical axis makes the system more sensitive, showing more features of the jet (Merzkirch 1965). The KE, oriented horizontal or vertical with respect to the focused light, will emphasize the density gradients in the horizontal or vertical directions respectively. In Fig. 3.2, examples of schlieren images with different KE orientations are presented. These are the images of an underexpanded hydrogen jet issuing into quiescent air. Significant

69 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES 47 (a) (b) (c) (d) Knife edge Razor blade Light source at the focal point FIGURE 3.2 Examples of schlieren images of jet issuing into quiescent air as obtained for different positions of the knife edge (razor blade) at the focal point. We use the set-up demonstrated in (d) where the knife edge cuts the focused light at an angle to enhance both the vertical and the horizontal density gradient effects. differences in the details of the flow-field can be observed just by positioning the KE at different orientations. In our experiments, the KE is positioned at 45 o to emphasize both the vertical and horizontal gradients. The test object is then imaged with a single constant focal length lens onto the intensified CCD camera. Two different focal length lenses (an f/12.5, 100 cm focal length lens and an f/6, 49 cm focal length lens) were used to capture different sizes of the field of interest. For the images presented in this paper, de-magnification of 0.44 was required and obtained using a 100 cm focal length (f/12.5) lens Timing and Synchronization Flow establishment, timing and synchronization are important issues that have to be addressed very carefully in preparation of an experiment in an impulse facility. The imaging system must be synchronized with the facility operation and the delay times must be set to allow the data to be acquired during the short steady test time. Therefore, test flow arrival and its steady duration are first studied through characterization experiments (see Chapter 2) to determine the required delay times. The general approach is to replace the injection plate with a pitot probe and to trace the flow history at

70 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES 48 Driver Section Driven Section Expansion Section Pressure Transducers: P10 P9 P8 P3 P2 P1 Injection Valve P8 trig in Pulse Generator Valve opens ms ~ 1.1 ms Free-stream flow is established Trigger pulse to the imaging system P2 trig in Test time = 270 s Pitot Pressure Shock wave arrival Ultra-Fast Framing Camera Hadland Imacon 468 Xenon Flashlamp Light Source Exposure time typ ns Interframing time typ s 20, 50 or 200 s FIGURE 3.3 Timing diagram of the high-speed rate imaging system and its synchronization with the expansion tube test flow time. the injection location by analyzing the pitot pressure. An example of the pitot pressure trace is given in the timing and synchronization diagram described in Fig Based on the time history of this pitot pressure, it is possible to identify the arrival of the shock wave, the time period of expansion section flow, the pressure rise as the contact surface between the test gas and expansion section gas arrives, and finally the steady flow test time. The images can be acquired during the last 100 µsec of the 270 µsec window of the steady flow test time, after the free-stream flow around the jet is established. As described in Fig. 3.3, the imaging system and the injection plate are synchronized by the instantaneous pressure rise of one of the piezoelectric transducers located on the tube walls as the incident shock travels through. The injection system is triggered early enough to allow the injector valve to actuate and the underexpanded jet to be

71 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES 49 fully developed by the time the steady test flow conditions are obtained. Within that constraint, the time interval between the valve actuation and the test gas arrival should be short enough to avoid significant changes in the expansion section initial pressure. The delay times for the imaging system are then set to sample during the actual test time when both the free-stream (expansion tube flow) and the flow around the jet are established. The high-speed camera, IMACON 468, and the long-duration light source receive the delayed trigger pulse simultaneously. The gating/exposure times of the 8 intensifiers and their interframing times are then controlled by the image acquisition computer. The first ICCD is set to acquire after the build-up time of the light source as the uniform light intensity is achieved Resolution Considerations Each intensified CCD detector in the ultra-fast framing camera has an 8-bit dynamic resolution (0-255 gray scale) array with µm pixel size. For the results presented in Chapter 4, the field of view is about mm (de-magnification of 0.44), corresponding to a minimum spatial resolution of µm. Visualization without blurring from the flow velocity requires careful consideration of the gating/exposure time. First the characteristic length scales and velocities of the flow-field need to be known. In the experiments presented in Chapter 4, we study the time evolution of jets issuing from a 2 mm sonic orifice into a high speed (2360 m/s) free-stream. Of particular interest is the convection characteristics of the jet-shear layer large eddies with dimensions ranging between 1-2 jet diameters (2-4 mm). We expect these structures to travel at speeds between that of the jet at the injector exit and the free-stream flow, namely between m/s ( mm/ns) in the case of hydrogen injection. Therefore, to achieve 1 pixel spatial resolution (50 50 µm), the exposure time must be in the range of ns. Exposure time of schlieren images are determined by optimizing the following competing factors: 1) schlieren sensitivity, 2) spatial resolution, 3) dynamic range, and 4) signal to noise ratio. The sensitivity to detect the smallest density gradients in the flow, is controlled by the KE position when the light source intensity and the optical components are fixed. As the KE cuts off more deflected rays, schlieren sensitivity increases while

72 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES 50 less light reaches into the camera. For short exposure times, however, a high intensity light source is required to cover the full dynamic range (256 gray levels). Therefore, we performed an optimization between these counteracting effects to achieve the best performance available from the system. We chose an exposure time of 100 ns to visualize the flow-field of mm, even though shorter gating times as low as 10 ns are available in our ultra-fast-framing camera. During 100 ns, the large scale structures translate mm which is about 2-10 % of their thickness (2-4 mm). This corresponds to blurring of images by 2-5 pixels. In addition, the ICCD modules were set to intensify % of their maximum potential. We found that larger gain results in noise levels that confuse the details of the flow features. Figure 3.4 represents 3 examples of schlieren images captured with different exposure times, 100 ns, 200 ns and 3 µs. Schlieren images with a 100 ns exposure time (Fig. 3.4a) provide the instantaneous features of the flow-field with an optimized spatial resolution (2-5 pixels). Increase of the exposure time to 200 ns (Fig. 3.4b) results in a significant blurring of the features, as the flow structures translate larger distances (4-10 pixels). Instantaneous flow features are eventually diminished with further increase of the integration time to the order of microseconds (Fig. 3.4c). While, the instantaneous features of the flow are wiped out due to the integration during the long exposure time, these images, on the other hand, provide information on the average features of the flow-field. Note that the noise level in the long exposure image is very low as the intensifier is set to its minimum value. Average features with laser-based diagnostic techniques can be achieved only by capturing a large number of images as the integration time is fixed with the laser pulse width or the fluorescence time. As noted in the introduction it is very difficult to achieve multiple images in an impulse facility because of the short test times. Observations obtained from the schlieren images of Fig. 3.4 will be discussed in Chapter 4. Temporal resolution, or the interframing time, was chosen so that we could capture events occurring on very short time scales, less than or equal to the convection time of the flow through the region of interest. For the jet diameter (2 mm) and free-stream velocities up to 4000 m/s, studied in our investigation, framing rates of 2 MHz to 200 KHz (interframing times between µsec) are needed to resolve and to follow the development of the flow structures (convection of large eddies, fluctuation of shock waves around the jet.)

73 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES 51 (a) (b) (c) FIGURE 3.4 Examples of schlieren images with different integration/exposure times: a) 100 ns exposure time, resolving the instantaneous features of the flow-field, b) 200 ns exposure time, resulting in blurring of the image, c) 3 µs exposure time, averaging the general features while enhancing the weak shocks such as upstream separation shock wave and downstream recompression wave.

74 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES 52 t x2 x1 y2 y1 Ucx ; Ucy t t y 2 y1 arctan x2 x1 FIGURE 3.5 Timing diagram of the high-speed rate imaging system and its synchronization with the expansion tube test flow time Image Processing and Analysis Post image processing of the images was performed using several image processing software packages (IPLab, Matlab, Adobe Photoshop and Premiere). Background images were acquired just prior to each test and subtracted from the schlieren image to eliminate speckle from the imperfections in the test section windows. Normalization of the intensity levels and the gamma-factor were changed to improve the contrast of the images. Gamma-factor is a factor applied to intensity distribution of images to enhance the perception of the human eye which has a non-linear sensitivity. To compute the convection characteristics of the large-scale eddies each individual structure was tracked with the cross-correlation method using the fast Fourier transform (FFT). In this tracking procedure, we measured the displacement of a particular feature in the streamwise and transverse directions. Guidelines for a fully automated classical cross-correlation method can be found in Smith and Dutton (1999). We utilized an FFT in our cross-correlation method to decrease the image-processing time.

75 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES 53 The cross-correlation is calculated (see Fig. 3.5) from the complex conjugate multiplication of their Fourier transforms: R fg ( x, y, t) F ( x, y, t + t) G ( x, y, t) (3.1) The size of the interrogation window ( pixels for the results presented in this paper) is selected to be large enough to include the eddy and its later location while maintaining the maximum resolution. Figure 3.5 describes the cross-correlation procedure and presents a representative cross-correlation field of the two images shown in the same figure. The highest cross-correlation magnitude corresponds to the convection distance of the large-scale structure. Therefore, the correlation peak was detected by simply scanning the correlation plane R fg for the maximum correlation value R(i, j) and storing its integer coordinates (i, j) with uncertainty of a ±1/2 pixel. A subpixel accurate displacement estimate could be achieved for example by applying three point interpolation (Raffel et al. 1998). However, the largest uncertainty in our measurements originates in the determination of the interrogation region. In summary, displacement measurement accuracy in our technique is ±1 pixel which corresponds to a velocity uncertainty of ±50 m/s ( 2 % of the free-stream velocity) when the interframing time is 1 µs. Once the displacement from image to image is known, the large-scale convection velocity and the convection angle are determined using U c,x = x 2 x 1, U c,y = y 2 y 1 (3.2) t t ( ) x2 x 1 Φ = arctan (3.3) y 2 y 1 where t is the interframing time between images. An interframing time of t = 1 µs was chosen based on the residence time of the eddies in the field-of-view. It takes the coherent structures about 8 µs to travel 6 jet diameters near the injector port. interframing times larger than 3 µs, it becomes difficult to track the structures as they travel large distances. For 3.2 OH-PLIF PLIF imaging of reactive flows relied on OH, a naturally occurring combustion radical, as the fluorescent tracer. OH is an indicator of ignition and reaction zones. For

76 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES 54 the purpose of this thesis, OH visualization provides information on whether ignition is occurring at all, and if so, in which regions it does occur Excitation and Detection Strategy The OH-PLIF measurements are obtained by excitation of the A 2 + X 2 Π(1, 0) band of OH, near 283 nm combined with detection of the strong (1,1) band near 315 nm (ranged from nm). This excitation strategy, which can be performed using a frequency-doubled dye laser, was chosen to avoid fluorescence trapping, the absorption of the emitted OH fluorescence by other OH molecules (Seitzmann and Hanson 1993). In addition, the limited pulse energies ( 1-30 mj) available from the frequency-doubled dye laser sources provide a linear fluorescence measurement regime. The isolated Q 1 (7) transition at nm is selected to minimize signal dependence on temperature (i.e., ground state population). For J =7.5, the population term is only weakly sensitive to temperature over a wide range of temperatures (i.e., K). The relative uncertainty in number density for a nearly constant pressure region is approximately ±10% due to Boltzmann temperature variations (Parker et al. 1995) OH-PLIF Laser Source and Tuning The UV laser radiation for excitation of the OH molecule is provided by the frequencydoubled output of a dye laser pumped by a pulsed Nd:YAG laser (Lumonics models YM-1200, HD-500, HT-1000). The 532 nm output of the Nd:YAG laser ( mj) pumps the grating-tuned dye laser which provides approximately mj/pulse at a wavelength of 566 nm using Rhodamine 590 dye. The dye laser beam formed as a 0.5 cm wide sheet is frequency doubled using a KT*P crystal in a temperature controlled housing, which is angled-tuned with a stepper-motor system to provide maximum energy (8-10 mj/pulse at 283 nm). At the exit of the frequency doubler, a Pellin-Broca prism is used to separate the UV beam from the fundamental. The beam is raised and turned toward the test section using four UV-enhanced turning mirrors and it expands to almost 35 mm near the test section. The beam is then focused into a µm thick 35 mm wide sheet using a f = 50 cm spherical lens. The top test section window, through which the laser sheet reaches the injection plate, is made of fused-silica to provide a low UV-beam reflection.

77 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES 55 The laser tuning is performed using a fluorescence measurement in an atmospheric methane/air flame. The actual laser calibration (the dial offset) is determined by measuring the OH-fluorescence across multiple excitations in the wavelength range nm and comparing the results to the estimated spectra using the programs LINES and SPECTRA (Seitzmann 1991). The fluorescence is collected using a Hamamatsu photomultiplier tube (PMT) fitted with Schott UG11 and WG-305 filters, and the resulting PMT signal is integrated using a boxcar averager. The laser wavelength dial offset did not drift significantly from day to day and, therefore, it was not necessary to perform this tuning procedure before each experiment. The laser linewidth in the current experiments is broader than the absorption linewidth by almost an order of magnitude. The absorption linewidth for OH is around 0.04 cm 1 for the temperature and the number densities present in the current work, and the laser linewidth is approximately 0.3 cm 1. The assumption of broad excitation is then valid, minimizing the potential errors due to laser tuning and lineshape effects. A real-time laser sheet correction is performed using a beam splitter/dye-cell/ccd camera arrangement (McMillin, Seitzmann, and Hanson 1994). The laser sheet energy distribution is measured by imaging the visible fluorescence of a dye-cell containing a Rhodamine 6G dye/methanol mixture which is excited by the partial reflection of the laser sheet. A Cohu 4810 CCD camera with an f/1.8 Nikon (50 mm) lens is used to detect the visible fluorescence from the dye cell. Post-image processing, performed using commercially available software, includes remapping of the 1-D laser intensity profile into 2-D sheet and normalization of the OH-PLIF image by this 2-D intensity profile OH-PLIF Imaging System and Its Spatial Resolution The fluorescence is collected onto the pixel array of an ICCD (Princeton Instruments) camera using a UV lens system (Nikon 50 mm, f/4.5). UG-5 and WG-305 Schott glass filters (2 mm thick) are used to block elastic laser scattering and background emission which was minimal in the current experiments. For the jet-in-crossflow and the cavity experiments performed in this thesis, the imaged region was mm. The minimum flow feature that can be resolved with an ideal detection system is 51 µm corresponding to µm per pixel (based on array with 23 µm pixels, magnification=0.45). In a real system, however,

78 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES 56 the minimum spatial resolution is actually limited to about 3-4 pixels (McMillin 1993) because of the imperfections in the imaging system (focusing errors, limited spatial resolution of the intensifier). Thus, the gating time of the camera intensifier is set to collect photons for a maximum duration of 100 ns to prevent further resolution reduction by flow motion. In conclusion, the resulting resolution of the PLIF imaging system is µm (4-5 pixels) Interpretation of OH-PLIF The relationship between the fluorescence signal and number density or mole fraction has been described by Hanson et al. (1990). Briefly, the fluorescence is proportional to the laser energy, the fluorescence yield, the species number density, and the Boltzmann population fraction for the absorbing transition. Effective combustion visualization with OH PLIF requires that the variation in OH mole fraction in the region of interest affect the fluorescence signal more than changes in the other parameters described above. The most critical issue, typically, is the temperature dependence of the Boltzmann fraction of the absorbing state. At the combustion pressures in this work, the fluorescence signal can be modeled as (Hanson et al. 1990): [ ] fj S f = χ OH T (3.4) where χ OH is the OH mole fraction, and f J is the Boltzmann fraction of OH molecules in the absorbing state. For the absorption transition considered here - the Q 1 (7) transition of the A 2 + X 2 Π(1, 0) band of OH, located at nm - such effects play a relatively minor role in interpreting the signal in the regions observed to contain OH, and the fluorescence intensity can be qualitatively linked to OH mole fraction. 3.3 Simultaneous Schlieren and OH-PLIF Two intensified CCD cameras are used simultaneously to collect both schlieren and OH-PLIF images. The fluorescence signal is collected through the same exit window as that of the schlieren system. To separate both light signals, we mount a 5 cm diameter dichroic mirror at 45 o to the optical axis perpendicular to the exit window. The dichroic, designed for larger than 99% reflectivity between 300 and 320 nm, reflects the OH fluorescence but is transparent to the schlieren beam.

79 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES 57 Figure 3.6 shows the triggering and timing diagrams of the simultaneous schlieren and OH-PLIF imaging. A homemade firing box and two delay generators are used to synchronize the laser/xenon-lamp/cameras with the expansion tube operation. Because of the long rise time of the Xenon flash lamp and also because of its emission in UV, OH-PLIF was performed approximately 2 µs before schlieren to avoid background noise in OH fluorescence. The firing box is designed to keep the laser operating (i.e, pulsing at 10 Hz) until the initiation of the expansion tube operation to minimize shot-to-shot pump laser energy fluctuations. The laser pulsing is stopped when a firing pulse is sent to double diaphragms to initiate the test, and then, the laser is pulsed once more to acquire the OH-PLIF image during the useful test time. Operating the pump laser in this fashion prevented laser misfires.

80 CHAPTER 3. FLOW VISUALIZATION TECHNIQUES 58 (a) Injection Valve Controller Expansion Section P1 P2 P3 P8 trig in BNC pulse generator Pressure Transducers Driven Section P8 P9 P10 Firing box (home-made) P10 trig in Driver Section Gage-Scope Data acq sys P2 trig in Set image delay time Set delay time Schlieren light source Laser System for PLIF To YAG laser flashlamps From Q-switch PLIF camera ICCD-1 Sclieren camera ICCD-2 (b) Injection Valve P8 trig in Pulse Generator Valve opens ms ~ 1.1 ms Flow established Trigger pulse to the imaging system P2 trig in Test time = 270 s Pitot Pressure at the test section Shock wave arrival OH-PLIF Imaging Laser pulse typ. 10 ns OH-PLIF camera exp. time typ. 150 ns Schlieren Imaging ~2 s ~3 s Xenon Flashlamp Schlieren camera exp. time typ ns FIGURE 3.6 a) Triggering diagram and timing connections of the imaging, the injection and the data acquisition systems. b) Timing diagram of simultaneous OH-PLIF and schlieren and their synchronization with the expansion tube test flow time.

81 Chapter 4 Time Evolution and Mixing Characteristics of Hydrogen and Ethylene Transverse Jets In this part of the investigation, flow-field properties of hydrogen and ethylene jets injected into a supersonic flow are reported. The free-stream flow replicates a representative supersonic combustor environment associated with a hypersonic airbreathing engine flying at Mach 10. The structural evolution, the penetration and the convection characteristics of both jets are analyzed. 4.1 Introduction Early studies suggested that the jet-to-free-stream momentum flux ratio, J, is the dominant parameter which controls the transverse jet penetration while the mechanism for mixing is controlled mainly by the counter-rotating vortex pair. However, large scale coherent structures are dominant in the jet shear layer and their structural evolution might have a big influence on the jet near-field mixing process. In the studies of mixing layers of two parallel streams (Brown and Roshko 1974) the mixing process was found to be controlled by large scale vortical structures. It is therefore important to understand how these structures and their growth rates evolve with time in the case of transverse jets as the crossflow and jet conditions are changed. 59

82 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 60 TABLE 4.1 Jet exit flow properties. Jet exit conditions Hydrogen Ethylene M jet 1 1 U jet, m/s T jet, K p jet, MPa J 1.4± ±0.1 M w,jet, g/gmole 2 28 γ jet d jet, mm 2 2 ν jet Re djet =U jet d jet /ν jet 150, , Results and Discussion We have studied the flow-field properties of both hydrogen and ethylene transverse jets using non-intrusive diagnostic techniques. Time-correlated schlieren images provides information on the structural evolution and convection characteristics of the jet, and OH-PLIF maps the regions of ignition where the fuel and the crossflow (air or oxygen) are mixed and burn at the molecular level. The jet exit flow properties for both fuels are presented in Table 4.1. Note that the exit velocities of hydrogen and ethylene sonic jets are quite different because of the substantial difference in their molecular weights. Also included in Table 4.1 is the jet-to-free-stream momentum flux ratio (J) that is chosen to be identical (J = 1.4) for both cases and expected to result in similar penetration heights for each case, as was suggested in previous studies. In this chapter, we will present results obtained from only one value of J. It is worth noting that experiments with different values of J provide similar results. In the following sections, we will first present the global flow-field properties of a transverse injection into a supersonic crossflow. Then the characteristics of the large scale eddies, their convection and mixing properties, the jet penetration and finally the OH-PLIF results will be discussed.

83 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS General Flow-Field Features Schlieren imaging provides a visual observation of both instantaneous and average characteristics of the flow-field depending on the exposure time of the image. While a short duration schlieren image ( ns exposure time) reveals some of the instantaneous vortex and shock structure of the flow-field, a long duration schlieren image (3 µs exposure time) provides information on the average and more steady properties. Two instantaneous schlieren images related to hydrogen and ethylene injection cases are shown in Fig Free-stream fluid (nitrogen) flows from left to right, and the fuel jet (hydrogen or ethylene) enters from the bottom at x/d = 0. Several interesting features, such as the large-scale structures at the jet periphery and the bow shock are very apparent in those images. The large-scale eddies are periodically generated in the early stages of the jet/free-stream interaction. While those eddies exist in both cases, they demonstrate significant differences in their development as they convect downstream. In the hydrogen case, these structures preserve their coherence with distance while in the ethylene case they disappear beyond about 12 jet diameters downstream. This result is not a schlieren contrast issue, rather it might be related to the enhanced mixing characteristics of the flow-field. As will be discussed in the following section the schlieren contrast for ethylene injection is expected to be 3 to 4 times larger than the hydrogen case in the absence of mixing (hot nitrogen vs. cold ethylene). Since the molecular weight of ethylene is nearly similar to nitrogen, the schlieren contrast will diminish when the hot free-stream fluid begins to mix with the cold ethylene jet while creating a region of reduced density gradient. The ethylene structures are bigger and penetrate deeper into the crossflow. Besides the bow shock, additional weak shock waves are formed around the ethylene eddies indicating their subsonic motion relative to the free-stream. A detailed examination of these large scale structures is performed using high speed schlieren movies and will be discussed in the following sections. Figure 4.1 also demonstrates that the bow-shock is almost merged with the jet close to the injection location with a very small stand-off distance and curves sharply downstream. Its local shape appears to depend strongly on the large scale shear layer structures, especially close to the jet exit where the free-stream behind the steep bow shock is subsonic. As a result the bow-shock reveals local fluctuations in position, small in the hydrogen case but significant in the ethylene case. Figure 4.2 shows an example for the

84 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 62 (a) (b) FIGURE 4.1 Examples of hydrogen (a) and ethylene (b) injections into a supersonic crossflow (nitrogen). Exposure time of each image was 200 ns. The x-axis is normalized by the jet diameter d. hydrogen flow-field, visualized with a longer exposure time (3 µs). Additional features are emphasized and become visually observable: such as the upstream separation shock wave and the downstream reattachment shock. The small instantaneous fluctuations of the bow shock are observed to average into a smoother and slightly thicker one. The barrel shock and the Mach disk are, however, not very clear even in the long exposure schlieren images, most probably due to the unsteadiness of the shear layer

85 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 63 FIGURE 4.2 An example of schlieren image with 3 µs exposure time for hydrogen injection case. While the unsteady features (coherent structures) are averaged to zero, some of the weak shocks such as upstream separation shock wave and downstream recompression wave are emphasized. vortical structures. Only the Prandtl-Meyer expansion fan of the underexpanded jet is observable (the white region at the jet core) indicating that the jet is indeed underexpanded. We have therefore attempted to estimate the displacement of the first Mach disk for our experiments by substituting an effective back pressure term in the Ashkenas and Sherman correlation given in Eq The effective back pressure introduced in earlier works is a notion which permits an analogy between the very complicated flow-field of an underexpanded jet emerging into a supersonic crossflow and that for the simpler and well-understood case of a jet exhausting into a quiescent medium. Among those previous studies, Schetz and Billig (1966) suggested P eb = 0.8P, where P is the free-stream pressure behind a normal shock wave. Later on, Billig et al. (1971) developed a correlation to predict the height of the Mach disk, y 1, assuming that the effective back pressure is equal to two thirds of the free-stream stagnation pressure behind a normal shock: P eb = 2/3P tot,. More recently Everett et al. (1998) measured the pressure distribution around a sonic jet injected transversely into a Mach 1.6 free-stream using a pressure-sensitive-paint technique. Their averaged surface pressure resulted in P eb = 0.35P (for J<1.5) which differs greatly from the earlier works. This discrepancy was attributed to the larger jet-to-momentum flux ratios (J) used in the

86 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 64 earlier works. We have adopted the back pressure values of Everett (P eb = 0.35P ), as the value of J in our experiments is small. Using Eq. 2.5, the Mach disk height for the current experiments was estimated to be around y d jet which compares well with the jet bending location (see discussion below). The free-stream conditions behind the hydrogen bow shock could be estimated by measuring the average bow shock position. Figure 4.3 presents two plots; the first shows the measured bow shock position and its angle (β), while the second plot exhibits the bow shock-induced free-stream velocity (U 2 ) and its turning angle (θ). Calculations are performed assuming a calorically perfect gas. In the region of 10 jet diameters studied in this work, the bow shock starts almost at 90 o and weakens downstream as it angle decays continuously down to 20 o -25 o. Further downstream, the bow shock is expected to reach its minimum strength or a Mach wave with an angle of 17.2 o (M =3.38). The induced velocity of the free-stream behind the bow shock is subsonic upstream to the location of the critical bow shock angle (β cr 67.6 o ), defined as the minimum angle for an oblique shock to be attached to the wedge. It is interesting to see that the bow shock reaches this angle around jet diameters above the wall at the expected height of the upper side of the Mach disk. Since the Mach disk occurs at a rather high Mach number on the jet centerline, the jet loses most of its momentum and the subsequent trajectory of the jet turns nearly parallel to the free-stream direction. Consequently, beyond the critical angle, the bow shock curves sharply downstream and the shock-induced free-stream velocity becomes supersonic varying between approximately 1050 m/s to 2260 m/s at 9.5 jet diameters downstream (note that the free-stream velocity is U = 2360 m/s). In the following sections, this estimated free-stream velocity behind the bow shock will be compared to the measured convection velocity of the large-scale structures. Before that we will first discuss the temporal evolution of these structures Large Scale Coherent Structures The most interesting observations are related to the coherent structures easily identified in instantaneous schlieren images. The large scale jet-shear layer vortices are considered important because of their role in the near-field mixing. These intermittently formed eddies tend to enlarge and engulf free-stream fluid as they travel downstream with the flow. We therefore studied the temporal evolution of large eddies and their

87 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 65 (a) y/d critical bow shock angle: cr ~67.6 y/d x/d Bow Shock Angle, (deg) (b) Free-streamVelocity, U 2 (m/s) max = U subsonic supersonic x/d Bow Shock Angle, (deg) FIGURE 4.3 (a) Bow shock position and its angle at the center-line of the jet as measured from the long exposure schlieren image shown in Fig (b) The free-stream velocity behind the bow shock and the flow turning angle based on the measured bow shock shape. For the calculations a calorically perfect gas has been assumed.

88 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 66 properties for both hydrogen and ethylene jets utilizing the high-speed-framing rate camera. Examples of instantaneous schlieren images are presented in Figs. 4.4 and 4.5 for hydrogen injection and in Figs. 4.6 and 4.7 for the ethylene case. While largescale eddies are visible in the early stages of the jet/free-stream interaction, there are significant differences in their development for hydrogen and ethylene injection. Hydrogen large scale coherent structures survive long distances. Coherence of these shear layer eddies can be seen in Figs. 4.4 and 4.5, which constitute consecutive schlieren images from two different experiments. Close to the jet exit, the spanwise rollers rise periodically creating gaps in between the eddies. The evolution of these eddies occurs primarily through engulfment of the cross-flow fluid into the jet but also through merging/pairing of smaller eddies in the beginning of the shear layer (see eddy number 3 in Fig. 4.4). Beyond 3-4 jet diameters downstream, the separation between the eddies becomes constant and no further merging is visible. The energetic structures elongate in the transverse direction while the crossflow fluid fills the braid regions in between the eddies. Interesting features in the evolution of ethylene large-scale structures are demonstrated in Figs. 4.6 and 4.7 through two examples of 8 consecutive schlieren images. Larger structures appear in the near-field of the ethylene jet and persist until the jet bends with the crossflow. In the bending region, the large scale structures begin to tilt in the streamwise direction. Simultaneously, the shear between the accelerating crossflow and the jet increases, leading to the stretching of the large-scale structures. In the case of ethylene injection the jet exit velocity (315 m/s) is four times smaller than in the hydrogen case (1205 m/s). Therefore, for ethylene injection the eddies are exposed to very large velocity gradients across the shear layer. As a result, these large-scale eddies lose their coherence as they turn in the streamwise direction and break up into smaller eddies through a tilting-stretching-tearing mechanism. Further downstream, beyond 6-8 jet diameters, the jet shear layer is not visually observable by schlieren imaging anymore, as the vortical structures tear down into smaller scale turbulence. The flow visualization of large scale structures using schlieren is based on the principle of refraction of light. The contrast in schlieren imaging, defined as the relative change in the illumination, is expressed in terms of the optical index of refraction (n)

89 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 67 and parameters related to schlieren system (Saad 1985): I I = fl dn ny 1 dy (4.1) where f is the focal length of the focusing lens, L is the width of the test section and y 1 is the size of the image of the light source (where the knife edge is positioned to cut the deflected beam). For a given schlieren system, the parameters f, L and y 1 are constant. The contrast is, therefore, directly proportional to the gradient of the index of refraction in the flow: I I dn dy (4.2) The index of refraction of a gas is expressed as a function of density (ρ) and a constant characteristic of the gas (β), : n = 1 + β ρ ρ s (4.3) where ρ s is the density at standard conditions (273 K and atmospheric pressure). The density ratio for a specific gas is equal to: Substituting from Eqs. 4.4 and 4.3 gives: I I ρ = P T s ρ s P s T (4.4) d dy (β ρ ρ s ) d dy (β P T s P s T ) (4.5) Consequently, the flow visualization of large scale structures based on schlieren is a result of the differences in the pressure, the temperature and the characteristic β constant of the free-stream fluid and the jet fluid. As the jet turns in the streamwise direction the static pressure between the hot free-stream ( 1300 K) and the cold jet ( 300 K) approaches to equilibrium. The schlieren contrast between unmixed jet and free-stream fluids can therefore be expressed in terms of: I I (β T s T ) (β T s T ) jet (4.6) By substituting the values of β and T in Eq. 4.6 we found that the schlieren contrast between the ethylene jet and the free-stream nitrogen (or air) should be 3 to 4 times larger than the hydrogen jet case. The loss of the visibility of the ethylene jet shear layer structures can, therefore, be attributed to the loss of the coherence of the vortical structures and also to enhanced molecular-mixing. When the ethylene large structures burst

90 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 68 into smaller scale turbulent structures due to stretching, molecular mixing between the crossflow and the ethylene jet might be enhanced. As a result, the observable schlieren contrast degrades as the difference between β/t across the shear layer decreases. Although the large scale eddies seem to be two-dimensional, recall that they are part of the unsteady Kelvin-Helmholtz spanwise rollers wrapping around the jet. They are only the traces of three-dimensional transverse vortex tubes whose cores coiled up around the jet with their legs connected downstream of the jet exit. The schematic in Fig. 4.8 shows a diagram of the three-dimensional unsteady structures as adapted from Brizzi et al. (1995). Similar flow-field features were also observed by Fureby and Ben- Yakar (2000), where a similar geometry and conditions are being studied by large eddy simulation. In the simulation results for the hydrogen injection case, large Ω-shaped vortices develop that grow as they convect downstream. We suggest that the vortex tubes on the sides of the Ω-vortices are stretched by increased shear stresses in the regions of steep velocity gradient. Time evolution of the tearing mechanism of ethylene eddies can be easily followed in the sequence of schlieren images. For example, the temporal development of eddy number 0 in Fig. 4.6 is captured during the 10.6 µs of visualization time. This eddy, generated by merging of two individual smaller eddies, is an energetic structure which penetrates deep into the free-stream. The initially almost round eddy stretches in the transverse direction due to the increasing velocity gradients across the layer while it is tilting in the clockwise direction. In the 8 th image the eddy numbered 0 has almost entirely dispersed into smaller eddies as the side arms of the vortex tube cannot continue to sustain the large shear stresses. Eddy number -1 in Fig. 4.7 is another example for the tilting-stretching-tearing mechanism. We have plotted the evolution of this eddy in a y-x diagram shown in Fig. 4.9, by tracking different parts of its structure across the shear layer. While the bottom part of the eddy travels at the slower jet velocity, the upper part of it is exposed to higher crossflow velocities. The shear stress steepens further downstream as the crossflow behind the weaker bow shock accelerates. Consequently, the eddies begin to stretch in the transverse direction while continuously tilting towards the fast crossflow stream.

91 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 69 1) t = 0.1 µs 5) t = 4.1 µs 2) t = 1.1 µs 6) t = 5.1 µs 3) t = 2.1 µs 7) t = 6.1 µs 4) t = 3.1 µs 8) t = 7.1 µs FIGURE 4.4 An example of 8 consecutive schlieren images of underexpanded hydrogen injection (d=2 mm) into a supersonic crossflow (nitrogen) obtained by high-speed-framing camera. Exposure time of each image is 100 ns and interframing time is 1 µs. Free-stream conditions are: U =2360 m/s, M =3.38, T =1290 K, p =32.4 kpa; and jet-to-free-stream momentum ratio is: J=1.4±0.1.

92 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 70 a) (Image 1) t = 0.1 µs b) (Image 3) t = 2.1 µs c) (Image 5) t = 2.1 µs d) (Image 7) t = 3.1 µs FIGURE 4.5 The second example of 4 of 8 consecutive schlieren images of hydrogen injection into flight Mach 10 condition. Exposure time of each image is 100 ns and interframing time is 1 µs.

93 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 71 1) t = 0.1 µs 5) t = 4.1 µs 2) t = 1.1 µs 6) t = 5.1 µs 3) t = 2.1 µs 7) t = 6.1 µs 4) t = 3.1 µs 8) t = 7.1 µs FIGURE 4.6 Time evolution of an ethylene jet in a supersonic crossflow (nitrogen) as observed from 8 consecutive schlieren images. Exposure time of each image is 100 ns and interframing time is 1.5 µs. Free-stream conditions are: U =2360 m/s, M =3.38, T =1290 K, p =32.4 kpa; and jet-to-free-stream momentum ratio is: J=1.4±0.1.

94 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 72 1) t = 0.2 µs 5) t = 5.0 µs 2) t = 1.4 µs 6) t = 6.2 µs 3) t = 2.6 µs 7) t = 7.4 µs 4) t = 3.8 µs 8) t = 8.6 µs FIGURE 4.7 The second example of an ethylene transverse jet flow-field in a supersonic crossflow as observed from 8 time correlated schlieren images. Exposure time of each image is 200 ns and interframing time is 1.2 µs.

95 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 73 FIGURE 4.8 Schematic of the three-dimensional shape (Ω shape) of the unsteady vortical structures formed intermittently (Brizzi et al. 1995). Space-time trajectories of large structures: Following the sequential high-speed-framing rate schlieren images, space-time trajectories (x-t diagram) of the identifiable coherent structures have been traced. Figure 4.10 presents two x-t diagrams of hydrogen eddies as analyzed from the schlieren images of Figs. 4.4 and 4.5. The spacing between the core of the eddies varies with distance, eventually reaching an average value of almost 3 jet diameters. Occasionally, big gaps of the order of 4 to 5 jet diameters in dimension (see Fig. 4.10b) are created as the smaller eddies are amalgamated into the larger ones. Two x-t diagrams showing the trajectories of the identifiable ethylene eddies are plotted in Fig None of the coherent large scale eddies could be traced beyond 6-8 jet diameters downstream. The spacing between the initial eddies is larger than the ones in the hydrogen case because of the large amounts of the crossflow intrusion in between the eddies, and also because of the larger size of the eddies formed near the jet exit. Information on the eddy formation frequency can also be obtained from the x-t

96 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 74 stretching tilting y/d frame number: x/d FIGURE 4.9 Development of a large-scale ethylene structure (eddy number -1 in Fig. 4.7) as it goes through the tilting and stretching processes. Four different parts of the eddy structure were independently tracked in the duration of the 8.6 µs flow visualization time. diagrams. Only 2 eddies are formed during the 10.6 µs time evolution of the ethylene jet, while in the hydrogen case 4 eddies are formed in even a shorter time period of 7.1 µs. Experiments with different sonic jets (see Chapter 5) revealed that the eddy formation frequency scales linearly with the jet exit velocity Convection Characteristics Once the centers of the large scale eddy structures are identified (as shown in the x-t diagrams), their convection velocity and the angle of inclination may be computed. For that purpose, each individual structure was tracked from image to image using cross-correlation techniques, as explained in Chapter 3. The resulting large-scale convection characteristics are summarized in Figs and 4.13, for hydrogen and ethylene cases respectively. Data for each case were collected from 16 images (two experiments per case). Included also in the figures are the reference lines for the jet exit velocity and for the free-stream velocity. The uncertainty in the determination of the eddy displacement is ±1 pixel (±45 m/s) in the hydrogen case

97 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 75 (a) U c,x = x/ t 9 8 Frame Number eddy 4 eddy 3 eddy 2 x t Time, sec 1 0 eddy x/d (b) Frame Number eddy 4 eddy 3 eddy 2 eddy x/d Time, sec FIGURE 4.10 Space-time trajectories of large-scale eddies present in the hydrogen jet shear layer. The center of the eddies are tracked from the 8 successive schlieren images shown (a) in Fig. 4.4 and (b) in Fig. 4.5.

98 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 76 (a) 9 12 Frame Number t U c,x = x/ t x eddy 2 eddy 1 eddy Time, sec x/d (b) bottom part middle part 8 Frame Number eddy 2 upper part Time, sec 1 0 eddy 1 eddy 0 eddy x/d FIGURE 4.11 Space-time trajectories of ethylene large scale eddies as tracked from 8 time-correlated schlieren images: (a) x-t diagram of the example shown in Fig. 4.6, (b) x-t diagram of the example shown in Fig. 4.7.

99 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 77 (a) 2500 Convective Velocity, m/s U =2360 m/s U jet =1205 m/s U c,x U c,y 0 (b) x/d Convection Angle, (deg) x/d FIGURE 4.12 Convection features of coherent large scale structures present in the hydrogen jet/free-stream shear layer. The data were subtracted by analyzing the eddy displacement in 8 consecutive schlieren images of 2 experiments (images shown in Figs. 4.4 and 4.5). (a) the convection velocity of eddies in streamwise and transverse directions, U c,x and U c,y, respectively; (b) the convection angle of eddies.

100 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 78 (a) 2500 U c,x U c,y Convection Velocity, m/s U =2360 m/s (b) Convection Angle, (degrees) 0 U jet =315 m/s x/d x/d FIGURE 4.13 Convection features of eddies present in the ethylene jet/free-stream shear layer. The data were subtracted by analyzing the eddy displacement in 8 consecutive schlieren images of 2 experiments (images shown in Figs. 4.6 and 4.7). (a) the convection velocity of eddies in streamwise and transverse directions, U c,x and U c,y, respectively; (b) the convection angle of eddies.

101 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 79 (a) hydrogen injection 2500 Convection Velocity, U c (m/s) U =2360 m/s measured, U c measured free-stream velocity behind the average bow shock, U x/d (b) ethylene injection 2500 U =2360 m/s Convection Velocity, U c (m/s) U jet =315 m/s x/d U 2 FIGURE 4.14 Measured convection velocity of large eddy structures in the hydrogen and ethylene jet shear layers. The results are compared with the estimated values of the free-stream velocity immediately behind the bow shock.

102 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 80 and ±2 pixels (±62-71 m/s) in the ethylene case. It is important to note that some of the eddy positions were tracked manually, especially near the injector exit where the cross-correlation method was not able to identify the initial small eddies at the vicinity of the bow shock. According to the results of Fig. 4.12, the hydrogen eddies initially travel fast in the transverse direction with velocities close to the jet exit velocity. As the jet bends downstream, the eddies start to accelerate monotonically in the streamwise direction and achieve almost 90% of the free-stream velocity 9 jet diameters downstream. At this location, the jet moves at shallower angles to the crossflow direction (around 0 o -10 o ) with reduced transverse convection velocities (between m/s). This reveals that beyond 9 jet diameters the jet shear layer eddies are convected almost parallel to the free-stream while the transverse penetration of the jet is just slightly increasing. Convection properties of the ethylene eddies (Fig. 4.13) are somewhat different from those in the hydrogen case. A large scattering of the velocity both in the transverse and streamwise directions is visible. The convection characteristics were measured not only by following the coherent large structures but also by tracking parts of the eddies that had began to lose their coherence. We observe that the upper part of the eddies tend to travel at higher velocities in both streamwise and transverse directions than the lower part of the eddies (see also Fig. 4.9). The transverse velocity (y-component) of some eddies is higher than the jet exit velocity. As the eddies stretch due to the large velocity gradient across the jet shear layer, the transverse velocities, specially at the upper part of the eddy, becomes as high as 700 m/s. The convection velocity in the streamwise direction is, on the other hand, much lower than the free-stream velocity. A result that can be attributed to the stronger (steeper) bow shock present for ethylene injection as the eddies rise up higher into the crossflow. The convection angle of the ethylene eddies, shown in Fig. 4.13b, are larger than the hydrogen ones, again a result of the higher penetration of the energetic ethylene eddies in the transverse direction as will be discussed in the following section. The free-stream velocity behind the bow shock, U 2, is computed based on the average bow shock position measurements as explained in section The results for the hydrogen injection are plotted in Fig. 4.14a together with the measured total convective velocities. We observe that the convection velocities of the low density hydrogen eddies are mainly influenced by the free-stream, as most of the eddies follow the shock-induced

103 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 81 bow shock nitrogen U 2 high U 2 low U 2 free-stream entrainment large-scale eddies low U 2 ethylene FIGURE 4.15 Schematic showing the low- and high-speed regions of the bow shock-induced free-stream velocity around the large-scale ethylene eddies. free-stream velocity. For the ethylene case, it is not possible to compute an average shock-induced free-stream velocity because of the bow shock fluctuations. Instead, we have measured two instantaneous bow shock positions and plotted the corresponding shock-induced free-stream velocities in Fig. 4.14b together with the total convective velocity across the ethylene shear layer. We observe large fluctuations in the values of U 2 varying in a wide range, between 1400 m/s to 2300 m/s around a single eddy. Figure 4.15 illustrates the low- and high-speed regions of U 2. Large velocity variation in the values of U 2 around the ethylene eddies, as opposed to monotonic increase in the hydrogen case, might be contributing to the tilting and stretching of eddies while contributing to the mixing process. Also, the reduced convective velocity of ethylene eddies provides longer flow residence time, crucial for the completion of the mixing process in shorter distances Penetration and Shear Layer Properties The upper boundary of the jet is defined by the maximum penetration of its shear layer vortices while the penetration bandwidth can be related to the visible thickness

104 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 82 of the jet shear layer. By measuring the visually observable upper edge of the jet in schlieren images, jet maximum penetration and bandwidth data became available. Brown and Roshko (1974), in their mixing layer studies, have shown that the visible shear layer width, as would be measured in a schlieren image, corresponds to about 1% concentration of molecularly mixed fluid. The results are presented in Fig to quantify the penetration properties and to compare it to previous studies. We observe significant differences in the penetration data and its width between hydrogen and ethylene injection. While the hydrogen jet penetrates 5.5 jet diameters into the free-stream at about 10 jet diameters downstream of the injection port, the ethylene jet penetrates as much as 8 jet diameters at the same location. This result is not surprising after studying the jet large-scale structure development in the previous sections. It is very surprising, however, when it is compared to previous studies (Schetz and Billig 1966; Rogers 1971; Papamoschou and Hubbard 1993; Gruber et al. 1995). These earlier studies showed that the jet transverse penetration into the crossflow is mainly controlled by the jet-to-free-stream momentum flux ratio (J). Therefore, both jets studied here should have comparable transverse penetration into the crossflow as the two cases have essentially the same momentum flux ratio. However, it is very clear from the results that the transverse penetration height of ethylene jet is higher than the hydrogen jet case. A power law fit to the penetration data has been proposed by various authors (Mc- Daniel and Graves 1988; Rothstein and Wantuck 1992; Gruber et al. 1995) who found that the upstream boundary layer properties, that is laminar/turbulent and the boundary layer thickness play an important role in the penetration of the jet. The most comprehensive and recent study was performed by Gruber et al. (1995), who suggest a power law fit of the form of: ( ) y x 1/3 dj = c (4.7) dj where the constant c has the value of 1.23 for circular injection. Their measurement technique relies on Mie scattering from ice particles in the free-stream, and defines the jet penetration as the trajectory where the jet concentration is about 10%. The thickness of the approaching boundary layer (δ/d=1) and the range of the jet-to-momentum ratio (J = 1 3) of their experiments were similar to the ones in our experiments, so that a comparison can be made. Therefore, the above correlation is plotted for J = 1.4

105 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 83 (a) hydrogen injection y/d Rothstein & Wantuck McDaniel & Graves Gruber et al. present work 3d x/d (b) ethylene injection Rothstein & Wantuck McDaniel & Graves Gruber et al. present work 5 6d y/d x/d FIGURE 4.16 Transverse penetration data of (a) hydrogen jet and (b) ethylene jet. The data points were obtained by manually tracking the visually observable outer edge of the jet from 8 consecutive schlieren images for J = 1.4±0.1. Both of the figures include analysis of 2 experiments namely 16 images. For comparison, also shown in the figures is the penetration correlation given by other studies.

106 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 84 in Fig together with our results measured for J = 1.4 ± 0.1. Two additional empirical correlations suggested by McDaniel and Graves (1988) and Rothstein and Wantuck (1992) are also included in Fig for further comparison. The penetration band in our experiments lies on top of the expected 10% penetration trajectory based on Gruber s correlation. The measurement of the visible jet s penetration as measured in schlieren images corresponds to 1% of the jet concentration, while Gruber s results correspond to 10%. Therefore, it is reasonable that the penetration measurements based on schlieren are somewhat lower than the ones based on 10% concentration measurements. A better agreement is achieved with the correlation of Rothstein and Wantuck (1992) who used OH fluorescence to visualize the jet penetration. Their experimental conditions (hydrogen jet injected into a high temperature air crossflow) are similar to our hydrogen injection case. In summary, the penetration data for the hydrogen case agrees relatively well with the previous studies. The differences in the observed penetration between hydrogen and ethylene data are most probably due to the tearing mechanism explained above. The thickness of the ethylene shear layer (the penetration width) grows to 6 jet diameters, twice as much as the hydrogen case at the end of the field of view. The practical impact of this result is significant as it indicates a mechanism for enhanced fuel (jet) distribution. It might eventually be possible to enhance and control the fuel penetration based on the flow properties. It seems that there is an additional mechanism which controls the jet penetration besides the jet-to-free-stream momentum flux ratio. This mechanism is expected to be associated with the jet shear layer properties which control its growth rate and therefore the near-field mixing of the transverse jet. Jet-to-free-stream density and velocity ratios are the two main parameters which might influence the large scale vortical structure development of the jet. In the next chapter (Chapter 5), we discuss the effects of these two parameters on the penetration and the development of the jet OH-PLIF Results To gain further insight into the coherence and the mixing properties of the injection flow-field we have examined the ignition characteristics of hydrogen and ethylene jets

107 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 85 using Planar Laser Induced Fluorescence of OH radicals. The presence of OH, a naturally occurring combustion product, indicates that the fuel and the oxidizer are mixed at the molecular level and the conditions for ignition to occur are met. As the total enthalpy of the free-stream in our experiments is high ( 4 MJ/kg), namely the total temperature is about 4000 K, autoignition of a transverse fuel jet is achieved. Figure 4.17 contains three instantaneous side-view images of OH-PLIF captured at the center-line of hydrogen and ethylene transverse jets injected into a reacting crossflow. OH-PLIF is a two-dimensional visualization technique which maps the auto-ignition locations illuminated by a roughly 0.4 mm thick laser sheet. The first image (Fig. 4.17a), related to a hydrogen jet injected into air, demonstrates a continuous and a very thin filament along the jet shear layer periphery. The side-view 2-D visualization clearly shows the presence of the large scale shear layer vortices. Since a relatively cold hydrogen jet is injected into hot air, there will be a significant variation of temperature with equivalence ratio through the mixing layer around the jet. The ignition time is a strong function of the mixture temperature, which will be higher at low equivalence ratios (fuel lean). The self-ignition point is therefore on the lean side of the mixing layer around the jet. Namely, ignition is likely to occur as soon as a fuel particle meets with the high temperature oxidizer. Since OH appears only quite near the hydrogen jet we suspect that the mixing is only occurring in the finite-thickness interfacial diffusion region that separates the unmixed fluids. Figures 4.17b and 4.17c are related to an ethylene jet injected into air and pure oxygen crossflows, respectively. Due to longer ignition delay times associated with ethylene, self-ignition could only be achieved when a higher concentration of oxygen was used in the crossflow. In contrast to hydrogen, OH radicals in the ethylene case could be detected in a wide region distributed across the jet. This is most likely a result of the enhanced molecular-mixing related to stretching-tilting-tearing mechanism discussed above. An additional interesting observation is related to the intense OH signals taking place in the vicinity of the Mach disk. This is the region where ethylene is self-igniting even when it is injected into air (see Fig. 4.17b). At this location, the ethylene jet becomes subsonic behind the Mach disk and begins to lose its transverse momentum letting the high temperature crossflow intrude deep inside the jet. Santiago and Dutton (1997), have also shown that the regions of high turbulent kinetic energy (TKE) exist in the jet shear layer near the Mach disk leading to better mixing properties.

108 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 86 (a) (b) x/d (c) x/d x/d FIGURE 4.17 OH-PLIF results mapping the ignition regions at the jet center-line of: a) hydrogen injection into air, b) ethylene injection into air, c) ethylene injection into pure oxygen.

109 CHAPTER 4. EVOLUTION OF HYDROGEN AND ETHYLENE JETS 87 In conclusion, OH-PLIF results demonstrated that significant differences exist in the near-field ignition properties for ethylene and hydrogen injection. These results support the tearing mechanism suggested to enhance the near-field mixing properties of the ethylene jet. 4.3 Summary In this part of the thesis, we have summarized results that are related to hydrogen and ethylene fuel jets because of their relevance to supersonic combustion. Significant differences related to the development of large-scale coherent structures were found to be present in the jet shear layer. The results demonstrate features not observed in previous studies where the free-stream conditions were limited to low velocities and low temperatures. In the current effort, the use of an impulse facility made it feasible to achieve high temperature and high velocity conditions relevant to a realistic supersonic combustor environment. The application of supersonic flow visualization at ultra-fastframing rates enabled a detailed study of the temporal evolution of the fuel jets. Further investigations of the experimental results are presented in the next chapter, which examines the dominant influences of different parameters on the stability and structural characteristics of the shear-layer formed in the jet periphery.

110 Chapter 5 The Effect of Velocity and Density Ratio on Transverse Jets In the previous chapter, we studied the temporal evolution of the shear layer structures of ethylene and hydrogen jets because of their relevance to supersonic combustion. The large difference in the molecular weights of these two gases revealed two important observations: 1) a tilting-stretching-tearing mechanism which caused the ethylene large-scale structures to lose their coherence and to burst into smaller eddies and 2) higher transverse penetration of the ethylene jet for a similar momentum flux ratio, J, as the hydrogen jet due to highly energetic ethylene eddies which penetrate deep into the free-stream. In the current chapter, we report our study of the fundamental origin of these phenomena. The parameters expected to be influential in the stability, and the structural characteristics, of the jet shear layer are the jet-to-crossflow density and velocity ratios. These two parameters are, however, coupled through the molecular weight of the jet. An increase in the jet s molecular weight reduces the exit velocity and increases the density. Therefore, to understand the role of each parameter on the instability of the large-scale eddies, we must decouple the two parameters and study them independently. The first part of this chapter discusses the effect of systematically changing the jet molecular weight. The second part presents the independent influence of the jetto-crossflow density and velocity ratios on the flow features. To decouple these two parameters, a variety of free-stream conditions are used. Also we discuss possible flow 88

111 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 89 instabilities, such as centrifugal instabilities associated with curved shear layers which could potentially cause the spanwise rollers to lose their coherence. 5.1 Effect of Jet Molecular Weight Gaseous jets with five different molecular weights (M w,jet =2, 4, 8, 16 and 28 g/mole) are studied. Tables 5.1 and 5.2 summarize the composition and the properties of these jets. The experiments are designed for similar jet-to-crossflow momentum flux ratio in the range J = Note the high Reynolds numbers associated with the jet exit properties. Also included in Table 5.2 are the jet-to-free-stream velocity ratio (r), the density ratio (s) and a convective Mach number parameter (M A c, see section 5.1.5) defined to estimate the compressibility level of the jet shear layer for the cases studied here Flow Visualization Results r = U jet U (5.1) s = ρ jet ρ = (M w/t ) jet (M w /T ) (5.2) Examples of instantaneous schlieren images obtained using the ultra-fast framing camera are presented in Fig. 5.1 for visual observations of the flow-field. A systematic increase in the jet molecular weight gradually changes the structural characteristics of the jet shear layer. Large-scale structures, which dominate the jet shear layer in all cases, become larger with the increase of the molecular weight. The heavier jets penetrate deeper into the free-stream. Consequently, the shape of the bow shock changes as it wraps around the large eddies. This process starts to be visible for M w =16 and becomes very pronounced for M w =28. The coherent structures occur less frequently with increasing M w, which results in a larger spacing between the eddies. This is followed by large intrusions of the crossflow between the eddies. In the case of M w =28, the fast crossflow stream sweeps the jet as the eddies lose their coherence. In the case of N 2 injection shown in Fig. 5.1e (M w =28) the large-scale structures are not visible beyond x/d In Chapter 4 we postulated that the reason the large-scale structures lose their coherence is related to the tiltingstretching-tearing mechanism which is due to the large velocity differences between

112 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 90 TABLE 5.1 The general flow exit properties of gaseous jets with different molecular weights. Jet Gas M w,jet γ e T e U e = a e µ e g/mole K m/s Pa s (T=T e ) 1) H 2 - Hydrogen ) 85.7% H % CH ) 57.2% H % CH ) CH 4 - Methane ) N 2 - Nitrogen TABLE 5.2 The specific flow exit properties of gaseous jets used in the study of the jet molecular weight effect. The free-stream used in these experiments simulates the flight Mach 10 flow condition. Jet M w,jet p e,eff J ν = µ/ρ Re d r s M A c g/mole atm m 2 /s (T=250 K) 1) ± , ) ± , ) ± , ) ± , ) ± , the jet and the crossflow. We will provide more evidence supporting this hypothesis in section Penetration and Shear Layer Thickness Figure 5.2 includes plots of the transverse penetration height and the penetration width of jets with different molecular weights but with almost identical jet-to-momentum flux ratios (J = ). The data are obtained by identifying the visually observable upper edge of the jet shear layer in the instantaneous schlieren images. The penetration profiles clearly demonstrate an increasing improvement in the penetration height for increasing values of the jet molecular weight. The same trend can also be observed for the growth rate of the jet shear layer. In Chapter 4 we suggested a relationship

113 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 91 a) M w = 2 g/mole b) M w = 4 g/mole c) M w = 8 g/mole d) M w = 16 g/mole e) M w = 28 g/mole Free- FIGURE 5.1 Examples of instantaneous schlieren images of jets with different molecular weights. stream conditions are: U =2360 m/s, M =3.38, T =1290 K, p =32.4 kpa.

114 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 92 between the jet shear layer thickness and the width of the maximum jet penetration. Accordingly, the visible jet shear layer thickness, δ vis, ranges between 4d-7d at 22 jet diameters downstream where the thicker shear layer is associated with the larger M w. The peak transverse penetration in the low M w (2 g/mole) case is about y/d=6.5 while in the high M w (16 g/mole) case is about y/d=10. The penetration height and therefore its width are expected to increase further downstream especially for the higher M w cases. These observations are not in agreement with other studies which state that identical penetration heights should be observed independent of the jet molecular weight. During the last years of studies, the jet penetration height was believed to scale with the momentum flux ratio J. For comparison with previous results two empirical correlations proposed by Gruber et al. (1995) and Rothstein and Wantuck (1992) are also plotted in Figs. 5.2a- 5.2d for J = Gruber et al. measured the penetration of hydrogen and air jets injected into a cold free-stream flow based on 10 % of the jet concentration. Rothstein and Wantuck, on the other hand, used OH fluorescence imaging to detect the hydrogen penetration injected into a high temperature reacting crossflow. The penetration measurements obtained in this work are higher than those observed in previous studies for all cases except the hydrogen case (M w =2 g/mole) which agrees reasonably well with Rothstein s correlation. Also, note that penetration profiles of Rothstein and Wantuck are almost 20 % higher than those reported by Gruber et al. The variation between these works are likely due to differences in experimental conditions and measurement techniques as we will discuss next. We may explain the agreement between our hydrogen injection results and the measurements of Rothstein and Wantuck through similarities in our experimental approaches. The free-stream velocity and temperature in both works are high (r < 1 in both experiments) and the measurement techniques are quite similar. In our reacting experiments (see Chapter 6), we demonstrated that OH radicals lie on the periphery of the hydrogen jet shear layer observed through a simultaneous OH/PLIF and schlieren imaging. Therefore, measurements based on OH emission and schlieren imaging should provide similar penetration profiles. Experimental procedure may also explain the discrepancy between the results of Gruber et al. and Rothstein and Wantuck. The free-stream velocity and temperature in the experiments of Gruber et al. are lower (r > 1, s < 1) than those in the work of

115 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS Rothstein 1992 a) M w =2 Gruber et al y/d x/d vis 4d 10 8 b) M w =4 y/d x/d vis 5d 10 8 c) M w =8 y/d x/d vis 6d 10 8 d) M w =16 y/d x/d vis 7d FIGURE 5.2 Jet transverse penetration along the axial distance, x/d. Data for four gases with different molecular weights are presented: a) M w = 2, J = 1.84, b) M w = 4, J = 1.72, c) M w = 8, J = 1.85, d) M w = 16, J = For comparison, empirical correlations suggested by Gruber et al. (1995) and Rothstein and Wantuck (1992) are also included for J = 1.75.

116 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 94 Rothstein and Wantuck. As will be argued in the second part of this chapter, the velocity ratio, r, is an important factor in the development and growth rate of the jet shear layer structures and, therefore, directly affects the penetration of the jet. An increase in the jet molecular weight decreases the jet exit velocity and hence the velocity ratio, r. In our experiments, higher penetration is observed for lower values of r. The penetration results of Lee et al. (1995) are an example which supports our findings. They measured jet penetration using PLIF (Planar Laser-Induced Fluorescence) measurements of NO of a non-reacting injectant (N 2 ) and OH of a reacting injectant (H 2 ) in a high temperature supersonic flow. The penetration height indicated by NO fluorescence was significantly larger than that indicated by OH fluorescence for the same J. Their reasoning for this phenomena, however, was not complete. They related the differences between N 2 and H 2 injections to the frictional losses in the H 2 injector and to the consumption of H 2 near the stoichiometric ratio. First, the frictional losses can only drop the penetration height by 5 % at most. Second, combustion between the cold hydrogen and the hot oxygen takes place in the lean region of the mixture where the mixture temperature is high and not near the stoichiometric value, as Lee et al. suggested. We propose, therefore, that the velocity ratio differences between the two cases (r = 1.1 for H 2 vs. r = 0.3 for N 2 ) cause the different penetrations observed by Lee et al. (1995). Namely, the smaller the r, the better the penetration. Another noteworthy observation was recently recorded by Mathur et al. (1999) in a high temperature (850 K), high velocity (1750 m/s) free-stream flow. Flame spreading angles of ethylene gas (M w =28) injected from a flush wall injector were measured and found to be twice as large as the predicted spreading. Mathur et al. concluded that some other mechanism, in addition to the transverse momentum of the fuel, must cause the rapid spreading of the flame around the ethylene jet. Indeed, their jet-to-free-stream velocity ratio was r = 0.18, similar to the case with M w =16 g/mole studied in our investigation. Here we suggest that variables other than the jet-to-free-stream momentum flux ratio, J, influences the penetration trajectory. We can achieve higher penetration heights as the molecular weight of the jet is increased systematically. Increase in the jet s molecular weight decreases the jet exit velocities, namely the r. Our findings are also consistent with previous investigations where similar experimental conditions were studied.

117 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 95 We can provide an improved explanation for their observations and for the discrepancies shown in the previous works by taking into account the velocity ratio, r, in addition to the momentum flux ratio, J Convection Characteristics Figures 5.3a - 5.3b summarize the convection velocity of large scale eddies in the transverse and streamwise directions. Also included in the figures are the velocity of the jets at the exit of the injector and the velocity of the free-stream. The convection characteristic of nitrogen eddies (M w =28 g/mole) could not be measured beyond x/d=8, as they lose their coherence through the bursting mechanism explained in the previous chapter. From consideration of the y-momentum equation, the pressure rise behind the Mach disk implies that the jet fluid loses its momentum in the y-direction. As the jet loses its y-momentum, the shear layer eddies accelerate in the free-stream direction because of the crossflow which applies drag forces on the jet. Further downstream, the eddies eventually convect with velocities that are closer to the free-stream velocity independent of the jet s molecular weight. The only difference between the cases is the distance where the eddies achieve their maximum velocity. Low density eddies (low M w ) could follow the free-stream velocity earlier than the denser ones with larger M w. For example in the case of M w =2, the maximum convection velocity was achieved around x/d=7, while in the case of M w =16 the plateau was reached just after x/d=16. This result is consistent with the fact that drag forces are the main cause of the convection of the eddies in the streamwise direction. For the same drag force, acceleration of heavier eddies is slower than the acceleration of lighter eddies Characteristic Large Eddy Frequencies (Possible Transverse Jet Modes) Since coherent structures include concentrated regions of vorticity which are the main mechanisms for the entrainment of the crossflow into the jet shear layer, understanding the origin of vorticity is essential for understanding the structures. We have, therefore, measured the characteristic formation frequency of the large scale eddies in the jet shear layer using the ultra-fast-framing schlieren system. Before presenting

118 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 96 a) M w =2 U c,x U c,y b) M w =4 U c,x U c,y Convection Velocity, m/s ~2100 m/s U =2360 m/s U jet =1205 m/s Convection Velocity, m/s ~2000 m/s U =2360 m/s U jet =850 m/s x/d x/d c) M w =8 U c,x U c,y d) M w =16 U c,x U c,y Convection Velocity, m/s ~1950 m/s U =2360 m/s U jet =600 m/s Convection Velocity, m/s ~2050 m/s U =2360 m/s U jet =420 m/s x/d x/d FIGURE 5.3 Convection velocity of large scale structures in the streamwise (M c,x ) and transverse (M c,y ) directions as a function of axial distance x/d. The results for each case (for each molecular weight of jet) are obtained from 4-5 experiments each including 8 consecutive schlieren images.

119 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 97 the results, we first summarize the possible instability modes of free-jets and jets-incrossflows. Free-jets generally have two dominant instability frequencies associated with different sizes of vortices. The first is originated by the instability of the shear layer at the jet orifice. The initial vortex shedding frequency, also called the most amplified frequency, f θj, scales with the initial shear layer momentum thickness, θ j, and jet exit velocity, U j. The corresponding initial vortex shedding Strouhal number, St θj = f θ j θ j U j (5.3) is found to be scattered from 0.01 to (Gutmark and Ho 1983). The second dominant jet instability mode is related to larger scale structures present downstream of the jet potential core. The characteristic frequency of this mode is referred to as the preferred mode frequency, f j of the jet. The initial vortices of the shear layer grow by merging and entrainment as they convect downstream. At the end of the jet potential core the dominant frequency is governed by the jet column instability (Crowe and Champagne 1971). The preferred mode frequency, f j, scales with the jet exit diameter and velocity, d and U j to yield the preferred mode Strouhal number St d = f jd U j (5.4) While many researchers confirmed the existence of a preferred mode, their experimental results surveyed by Gutmark and Ho (1983) revealed that the value of St j varies widely between 0.24 to These discrepancies were attributed to the various initial conditions of different facilities. To determine the origin of the large scale structures of sonic jets in crossflows, an analogy between this and the case of a free-jet has been presented. Using the concept of the preferred mode frequency, we obtain a dominant preferred Strouhal number for the transverse injection case. Figure 5.4 represents the measured dominant frequency of the eddy formation in the beginning of the jet shear layer and the corresponding Strouhal number as measured for different jet exit velocities. The measurements are obtained by analyzing 5-10 different experiments for each case. Each experiment includes 8 time correlated schlieren images. The appropriate interframing time between the images allows us to identify the newborn

120 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS St j f j, khz U jet, m/s FIGURE 5.4 Formation frequency of the large scale structures and the corresponding preferred mode Strouhal number, St d = f j d/u j, as a function of the jet exit velocity. The data were collected from the time evolution observation of the jet from 8 consecutive schlieren images. Each data point was obtained by averaging 5-10 experiments with the error bars representing the deviation from the mean value. eddies during the flow visualization time (ranging between µs). The results show that the characteristic frequency of the eddy formation is scaled linearly with the jet exit velocity. This suggests that the shear layer eddies are associated with the preferred instability mode of the jet. The corresponding preferred mode Strouhal number can therefore be calculated based on Eq The results indicate that this preferred mode corresponds to a Strouhal number of about St d = 1 for all the cases studied in this work. A comparison with previously published results is not possible due to the lack of such measurements in supersonic crossflows. The only comparison that can be made is with the results of Fric (1990) performed in subsonic flows. From his measurements, the preferred mode frequency of the subsonic transverse jet was found to decrease with distance along the jet. Merging of the eddies was the main reason for the decrease in the frequency, changing the Strouhal number from St d 1 2 down to 0.2 around x/d 5.

121 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 99 St j / St j, min f j, khz 1.5x x x x x x x x x x x x10 5 Re jet FIGURE 5.5 Formation frequency of the large scale structures and the initial vortex shedding Strouhal number, St θj = f θj θ j /U j, as a function of the jet Reynolds number. St d 0.2 is smaller than the values quoted for free jets. In our experiments, the number of eddies are counted just before the bending of the jet as they become identifiable. Further downstream, the vortex merging does not happen very often. St d 1 can therefore be presented as the preferred mode Strouhal number for the transverse jets in supersonic crossflows for the geometry examined in this work. This Strouhal number is larger than those quoted for free-jets. It is also larger than the ones measured for transverse jets in subsonic flows. We can also calculate the Strouhal number based on the initial shear layer momentum thickness, θ j. However, the appropriate data are not available to estimate the momentum thickness of the jet shear layer exiting the injector port. We therefore normalized the St θj with its minimum value assuming that θ j = Re jet. Figure 5.5 summarizes the findings showing that the formation frequency is not associated with the initial vortex shedding frequency. St θj appears to increase with increasing Re jet.

122 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 100 These results suggest that the shear layer eddies are independent of the exit shear layer instability characteristics, because the characteristic length scale of the mode has been shown to be the jet diameter and not the thickness of the exit shear layer. The most amplified frequency of the jet column instability is the origin of its large-scale structures Jet Compressibility Analysis In studies of mixing layers of two parallel streams, it has been shown that compressibility levels control the growth rate of the mixing layer which decreases with increasing compressibility. We have therefore attempted to analyze the effect of compressibility on transverse jets. Convective Mach number, M c, is a key compressibility parameter which is based on the velocity of the large scale structures relative to either free-streams. Papamoschou and Roshko (1988) defined a convective Mach number for each stream: M c1 = U 1 U c a 1 ; M c2 = U c U 2 a 2 (5.5) where U 1 and U 2 are the high and low speed streams, respectively (refer to Fig. 5.6). Assuming that entrained fluid stagnates isentropically in the frame of the structures, the pressure-matching condition at the stagnation point yields: M c1 = γ2 γ 1 M c2 (5.6) U c = 1 + r s U s (5.7) assuming γ 1 = γ 2 M c1 = M c2 = M c = U 1 U 2 a 1 + a 2 (5.8) At low compressibilities (M c < 0.5) the structure of a non-reacting shear layer is two-dimensional and the large eddies travel with the average convective Mach number, M c. As the compressibility increases, this theory seems to fail as the central mode stabilizes and a transition from a two- to three-dimensionality occurs. Beyond M c =0.5, experimental results by Papamoschou (1991 and 1997) and by Dimotakis (1991) showed that convective velocities follow the stream selection rule which is based on the existence of fast and slow modes. If one stream is supersonic and the other subsonic, the convective velocity of structures will be closer to that of the fast stream (fast mode),

123 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 101 (a) (b) Bow Shock Induced Free-Stream Jet Boundary U c U 2 C M >1 P =? eb A 2 B M > ' M >1 1 ' y 1 U -U 2 c U 1 Stationary Frame P M=1 j, j x 1 Convective Frame U -U c 1 FIGURE 5.6 Flow-field schematics used in the jet compressibility analysis. Letters A, B and C indicate the zones of the jet shear layer. and when both streams are supersonic the convective velocity will favor the slow speed stream (slow mode). One could basically adapt the convective Mach number concept to scale the compressibility of transverse jets. However, because of the three-dimensional complexity of the flow-field and lack of the velocity field information, there is an additional difficulty in predicting the convective Mach number of transverse jets in supersonic crossflows. Neither crossflow nor jet-flow properties are uniform. The underexpanded jet accelerates inside the barrel shock and eventually gets compressed through the Mach disk, while the crossflow changes its properties continuously behind the curved bow shock. We, therefore, suggest separating the jet flow-field into three different zones to analyze the compressibility effects. The schematic of the proposed regions are illustrated in Fig. 5.6 and indicated with letters A, B and C. Zone A is close to the jet exit where the y-component of the free-stream velocity is negligible. The convective Mach number in zone A (Mc A ) can therefore be calculated using the equation 5.8 and by assuming that U 1 = U jet and U 2 = 0. Figure 5.7 shows the estimated values of Mc A together with the visible thickness of the jet shear layer, (δ vis ) at x/d 22 as obtained from penetration

124 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS Convective Mach Number, M C A M c A vis Shear Layer Thickness, vis /d U j (m/s) FIGURE 5.7 Estimated convective Mach number in zone A, M A c, (refer to the schematic in Fig. 5.6) and the measured visible jet shear layer thickness, δ vis, at x/d 22 as obtained from penetration width measurements. width measurements (see Section 5.1.2). A decrease in δ vis is observed for increasing values of Mc A, indicating that the compressibility properties in the initial region of the jet shear layer affect its growth rate. For zone B, the convective Mach number model, given by Gruber et al. (997b), guided our calculations. Gruber et al. suggested quantifying the compressibility level of transverse jets at a single point in the jet shear layer near the Mach disk (zone B in our case). For this calculation, the jet velocity just upstream of the Mach disk (U 1 ) is estimated. Figure 5.8a presents the estimated values of U 1 with error bars obtained for a range of assumed pressure ratios (P 1 /P ). Also included in this plot is the range of the free-stream velocity downstream of the bow shock, which is determined using the oblique shock relations for a range of shock angles (45 o < β < 65 o ) measured in section 4.2.1). Evident in Fig. 5.8 is that the free-stream velocity behind the bow shock is faster than the heavy jets (M w = 28 and 16) and slower than the light jets (M w = 2 and 4). In the experiments of Gruber et al. the free-stream velocity was slower than both the heavy and the light jets and therefore a comparative convective Mach number could be identified to quantify the compressibility level of jets. In the

125 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 103 MaximumJet Velocity, U 1 (m/s) (a) B-zone U =2360 m/s U behind the bow shock U j (m/s) Jet Velocity Behind the Mach disk U 1' (m/s) (b) C-zone U =2360 m/s U behind the bow shock U j (m/s) FIGURE 5.8 Estimated velocity fields for the jet and the free-stream in zones B and C. current experiments, however, it is difficult to find a clear dependence of δ vis on M B c. We, therefore, suggest the use of the convective Mach number in zone A (M A c ) as an indicator for the jet compressibility level. Figure 5.8b presents an estimate for the expected velocities in zone C. The value of jet velocity U 1, immediately after the Mach disk, is calculated using the normal shock and isentropic relations. The free-stream velocity behind the bow shock is again estimated using the measured bow shock angle as described above. The plot indicates that the supersonic free-stream is significantly faster than the jet which becomes subsonic behind the Mach disk. It is possible that in this region the stream selection rule mentioned above applies. According to this rule the convective velocity of structures will be closer to that of the fast stream (fast mode) as one stream is supersonic and the other subsonic. In the previous section, the jet shear layer eddies in zone C and beyond are indeed demonstrated to convect with velocities close to the fast free-stream values. In summary, we propose the convective Mach number in zone A as an indicator of the jet compressibility level. The growth rate of the shear layer scales with M A c. The convective velocity of the eddies, on the other hand, seems to follow the stream selection rule as observed in the shear layers of two parallel streams.

126 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS Effect of Density and Velocity Ratios In this part of the investigation, we pursue the decomposition of the molecular weight effect into its constituent parameters, velocity and density. For a given pressure and temperature of the jet: γ U j and ρ M w,j (5.9) M w,j Therefore, by changing the jet molecular weight, its density and velocity are also changed. By using different free-stream conditions we can control the jet-to-free-stream velocity and density ratios independent of each other. For that purpose, 24 different experimental conditions, summarized in Table 5.3, are studied. The stability of the jet shear layer eddies are analyzed by the aid of the flow visualization Flow Visualization Results Having examined the effect of jet molecular weight, we showed that the coherence of the large scale eddies is significantly affected by the increase of the jet molecular weight. The eddies burst into smaller structures by the tilting-stretching-tearing mechanism discussed in the previous chapter. This bursting mechanism will henceforth be referred as the instability of the eddies. We observed the stability of the eddies by visual observations based on schlieren imaging. In such work, the eddies are identified as unstable when the structures lose coherence and significant distortions in the bow shock shape can be observed. Examples of schlieren images for selected experiments are shown in Figs. 5.9, 5.10 and The first set of experiments numbered as 1 to 5 in Table 5.3, is performed in the flight Mach 10 condition. These experiments are already discussed in detail in the studies of jet molecular weight effect (see Section 5.1). Here we will discuss the rest of the experiments, numbered from 6 to 24. The first goal is to eliminate the density ratio effect by simulating conditions with a similar velocity to the Mach 10 case, but with much lower static temperature so that the density ratio varies minimally when the jet molecular weight is changed. Therefore, the free-stream conditions (600/9/6) for the second set of experiments (experiments no. 6-8) have static temperature of only 243 K. Results of these experiments show that the jet becomes unstable again as the velocity ratio is decreased to below 0.17 (see Figs. 5.9b

127 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 105 TABLE 5.3 Summary of the different conditions used in the study of jet instability analysis. No Free-stream M w, γ U p T M M w,j r s Condition m/s atm K U j /U ρ j /ρ a /0.5 b /20 c (flight Mach 10) /9/ /4/ /0.3/ /0.45/ (flight Mach 8) /4/ /10/ /10/ a /5 b / The free-stream conditions are represented by the expansion tube initial filling pressures with nomenclature referring to Fig a. P 4, psig, b. P 1, psia, c. P 10, torr. and c) even though the density ratio stays at low values at which the Mach 10 condition showed stable flow structures. The bow shock which stays smooth in the hydrogen injection case (Fig. 5.9a) is distorted in the case of methane and nitrogen injections. Additional shock waves are generated around shear layer structures as the velocity of the free-stream becomes supersonic relative to the convection of the eddies. Tilting of the eddies is visible in the region close to the wall. Note also the shock waves generated from leading edge of the plate. These are actually weak Mach waves which are clearly visible owing to the high sensitivity of the schlieren system. The same flow features

128 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 106 are obtained in the third set of experiments (experiments no. 9 an 10) presented in Figs. 5.9d and e, as the flow conditions (600/4/6) are similar to the second set. The second goal is to eliminate the velocity ratio effect by generating conditions of high density ratio while keeping the free-stream velocity low. For that purpose we run the facility in a shock tube mode, namely without the second diaphragm between the expansion and driven sections (experiments no. 21 to 24). As shown in Figs. 5.11b and 5.11d, the jet eddies stay coherent for high values of s (experiments no. 22 and 24) at which jet eddies in the Mach 10 condition lose their stability. The eddies stay coherent without large intrusions of air into the jet. It is worth noting that these span the range of velocity conditions at which most of the previous studies were performed. Therefore, not many significant differences in the flow field could be observed in those studies. Finally, note the shock waves emitting from the injection location (see Figs. 5.11a and c). In these experiments the jet exit velocity is larger than the free-stream velocity, r > 1. Note also the large stand-off distance of the bow shock because of the low Mach numbers. The results are presented in a density versus velocity ratio (s-r) diagram shown in Fig Separate regions of stable and unstable flows can clearly be identified. The plot indicates that the density ratio, s, has a little effect on the instability of the jet shear layer structures while the velocity ratio, r, is the main controlling parameter. We therefore propose to use an effective velocity ratio, λ, which will be presented in the following section Definition of an Effective Velocity Ratio, λ We propose to define an effective velocity ratio, λ, as a measure of the magnitude of the velocity difference across the jet shear layer. We assume that near the jet exit port, the jet issues in the y-direction at velocity U j and the free-stream flows in the x-direction at velocity U. This configuration, shown schematically in Fig. 5.12, is a special case of skewed mixing layers with a skewing angle of 90 o between high- and lowspeed streams. The jet fluid is expected to convect in the direction of the total velocity vector. Therefore, the effective shear will be proportional to the velocity difference in that direction. The components of U j and U in the direction of the total velocity

129 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 107 a) H 2 injection in 600/9/6 (No.6) d) H 2 injection in 600/4/6 (No.9) b) CH 4 injection in 600/9/6 (No.7) e) CH 4 injection in 600/4/6 (No.10) c) N 2 injection in 600/9/6 (No.8) FIGURE 5.9 Schlieren images at selected conditions given in Table 5.3.

130 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 108 a) H 2 injection in 600/0.3/60 (No.11) d) CH 4 injection in 300/10(Ar)/15 (No.18) b) CH 4 injection in 600/0.3/60 (No.12) e) N 2 injection in 300/10(Ar)/15 (No.19) c) N 2 injection in 600/0.3/60 (No.13) f) N 2 injection in 300/10(He)/15 (No.20) FIGURE 5.10 Schlieren images at selected conditions given in Table 5.3.

131 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 109 a) H 2 injection in 80/0.5 (No.21) c) H 2 injection in 100/0.5 (No.23) b) N 2 injection in 80/0.5 (No.22) d) N 2 injection in 100/0.5 (No.24) FIGURE 5.11 Schlieren images at selected conditions given in Table 5.3.

132 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 110 U j U cos( ) U j sin( ) U U U = U j sin( ) - U cos( ) FIGURE 5.12 Velocity vector field (U, U j) for a skewed mixing layer and the effective velocity ratio, λ., described in the total velocity vector direction. vector are: U j = U j sin(α) and U = U cos(α) (5.10) The velocity difference U is therefore equal to: U = U j sin(α) U cos(α) (5.11) U U by substituting tan(α) = U j /U, the effective velocity ratio, λ, can then be calculated as the ratio of the velocity difference U to the free-stream velocity U λ = U U = 1 r2 1 + r 2 (5.12) Therefore, when λ = 0, there is no shear, and when λ = 1, the shear is a maximum as the free-stream velocity is much higher than the jet velocity. Figure 5.14 presents the experimental results in a new velocity-density diagram where the effective velocity ratio λ is used instead of r. Again two distinct regions of stable and unstable flows are identifiable. Clearly, a critical value for the effective velocity ratio exists in the vicinity of λ cr 0.94 beyond which the jet flow becomes unstable as the tilting-stretching and tearing mechanism becomes important Discussion on the Effect of the Curvature - Centrifugal Instability Analysis Certain configurations possessing curvature are prone to centrifugal instabilities. Curvature introduces centrifugal forces which tend to produce streamwise vortical structures of the Taylor-Görtler type. Rayleigh (1880) showed that the stability condition

133 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 111 s j 9 Unstable Stable 8 13 Stable 7 20 Unstable U j r U FIGURE 5.13 Jet-to-free-stream density ratio vs. velocity ratio. The number near the data points corresponds to the experimental conditions summarized in Table 5.3. Unstable flow jet is defined when the large structures lose coherence downstream of the injection port and significant distortions in the bow shock shape can be observed. 9 Stable Unstable j s 8 Stable 13 7 Unstable U U 2 1 r 2 1 r cr FIGURE 5.14 Jet-to-free-stream density ratio vs. the effective velocity ratio, λ. The number near the data points corresponds to the experimental conditions summarized in Table 5.3.

134 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 112 (a) Rayleigh-Synge criterion for centrifugal stability: s 2r cons. (b) Experimental results s unstable s stable stable unstable r r s j U r U j FIGURE 5.15 Schematics illustrating the stability regions based on a) Rayleigh criterion for centrifugal forces in the curved mixing layers as given in Eq where cons. = δ vis h max and b) current experimental results. for an inviscid flow depends on the gradient of angular momentum. Synge (1933) gave a stability criterion taking into account that a density difference may also produce an unstable layer. The associated instability mechanism is therefore known as the Rayleigh-Synge criterion and is given in the form of: for dr > 0, this criterion can be written as follows: d ( ρu 2 r 2) > 0 (5.13) dr dρ ρ + 2dU U + 2dr r > 0 (5.14) by substituting dρ = ρ ρ j, du = U U j and dr/r = δ vis /h max we obtain: ρ j ρ + 2 U j U < δ vis h max (5.15) where the curvature radius is assumed to be the maximum jet penetration height, h max. Eq (s + 2r < const), therefore, provides a criterion for the stability of the curved jet shear layer in case the centrifugal forces play an important role on the stability. However, comparison of Equation 5.15 with the results presented in Fig indicates the opposite trends. As shown in Fig. 5.15, the regions where a curved jet shear layer

135 CHAPTER 5. VELOCITY AND DENSITY RATIO EFFECTS 113 is stable are actually the regions of the unstable jet shear layer as observed in our experiments. Therefore, we can conclude that instabilities associated with curved shear layers do not contribute to the instabilities in the case of a normal injection into a crossflow.

136 Chapter 6 Autoignition and Flame-Holding Capability of a Hydrogen Transverse Jet This chapter describes the experimental efforts in characterizing the ignition and the flame-holding capabilities of a transverse jet injected into high total enthalpy supersonic crossflows. The use of an expansion tube provides a correct simulation of true flight combustion chemistry, including ignition delay and reaction times. The experiments are designed to map the hydroxyl radical (OH) in the near-field of an underexpanded hydrogen jet injected into flight Mach number 8, 10 and 13 total enthalpy flow conditions. 6.1 Ignition and Flame-Holding Considerations The stabilization of flames in supersonic flow is a difficult issue for two primary reasons: First since flow times are short, ignition delays can result in significant travel of the jet plume through the combustor. Second, strain rates tend to be high in compressible flows which can suppress combustion. Transverse injection is a commonly used flame stabilization scheme in supersonic combustors. It provides flame stabilization: 1) by organization of an upstream recirculation zone, 2) by formation of coherent structures containing unmixed fuel and air, where a diffusion flame occurs as the gases are convected downstream, provided the 114

137 CHAPTER 6. IGNITION OF HYDROGEN JET 115 strain rates are not too high. Transverse injection schemes have two main points where the ignition is likely to occur: the region behind the jet bow shock where high temperatures and pressures are obtained, and the recirculation regions ahead of and behind the base of the jet where long residence times and high temperatures exist. The residence time of the hydrogenair mixture in the bow shock region is short, since the mixture expands around the jet flow-field immediately after compression in the bow shock. For the higher Mach number flows, however, ignition may still be initiated in the bow shock region due to the relatively high static temperature, though a more likely place for ignition to occur is in the recirculation region upstream the jet exit as will be shown in the results section. In scramjet combustors, where relatively cold hydrogen is injected into hot air, there is a significant variation of temperature with equivalence ratio (φ) through the mixing layer around the jet. Since the temperature of the mixture will be higher at low equivalence ratios, and since ignition time is a strong function of the mixture temperature, it is expected that the self-ignition point will be on the lean side of the mixing layer around the jet (φ = 0.2 (Huber et al. 1979)). The ignition delay times associated with hydrogen-air mixtures will be discussed in Section In order for self-ignition (and therefore combustion) to be accomplished in a flowing combustible mixture, it is necessary that four quantities have suitable values: static temperature, static pressure, fuel-air ratio, and the residence time at these conditions. In a reacting system, ignition is considered accomplished when sufficient free radicals are formed to initiate the reaction, even though no appreciable heat has yet been released. When the conditions of spontaneous ignition exist, the distance l i at which it occurs in a medium flowing at a velocity u is: l i = u τ i (6.1) Since the ignition delay time τ i varies inversely with pressure (because of the two body reactions involved in the ignition chemistry of hydrogen and air) the product τ i p is effectively constant for a given temperature and fuel-air equivalence ratio. This allows the use of the binary scaling law for the ignition of hydrogen-air mixtures in the following form, pl i u constant (6.2)

138 CHAPTER 6. IGNITION OF HYDROGEN JET 116 This means that for a given combustor entry temperature (typically in the range of K, Heiser and Pratt 1994), the ignition lengths are directly proportional to flight velocity, V 0, by l i u p 1 V 3 0 (hyrogen fuel) (6.3) Remember that supersonic burner entry pressures and velocities are scaled with the flight velocity according to: p 3 1/V 2 0 and u 3 1/V 0 (see Chapter1). Equation 6.3 indicates that the ignition lengths in a hydrogen-fueled scramjet become very large at high flight speeds. The ignition lengths are even larger if hydrocarbon fuels are employed. Because of the larger dependence of the ignition delay time on pressure, the ignition length in a hydrocarbon-fueled scramjet has a larger dependence on 1/V 0 as: l i u p n 1 V 2n+1 0 (ethylene fuel) (6.4) where n > 1. This is one of the reasons why hydrocarbon fuels are not likely to be used for high-flight Mach numbers. In the supersonic combustion area, a general consensus is: storable JP-type hydrocarbon fuels can be used up to Mach 6-8. Liquid methane could be used to somewhat higher Mach numbers, but speeds in excess of about Mach 10 requires liquid hydrogen. Mainly because hydrogen fuel is characterized by fast reaction rates and high heat release per kilogram of fuel (120 MJ/kg fuel ). On the other hand, hydrocarbon fuels have significant shortcomings in supersonic combustion when compared to hydrogen. The hydrocarbon fuels have relatively long ignition delays and limited cooling capability. Furthermore, at low speed flight conditions (flight Mach below 8), the total temperature of the free-stream is lower introducing difficulties on flame stabilization inside the supersonic combustor. These considerations will require some innovative thinking (see Chapter 7) especially when the fuel of choice is a heavy hydrocarbon (JP-7, decane, etc.) The main objective of this study is therefore to investigate the combustion characteristics of a hydrogen transverse jet in order to obtain a picture of its near-field autoignition and flame-holding capability. The second objective is to compare the hydrogen jet autoignition capability with that of the ethylene jet.

139 CHAPTER 6. IGNITION OF HYDROGEN JET Ignition and Flame-Holding Results Simultaneous OH-PLIF/Schlieren Results Figure 6.1 demonstrates an example of simultaneous schlieren and a side view OH- PLIF image overlaid in a single image (Fig. 6.1c). These images are obtained at the flow conditions simulating flight Mach 10. Apparent in the images are the regions containing OH molecules, indicating the location of the reaction zone. The structural evolution of the reaction zone is in good agreement with the jet position determined by the schlieren imaging although there is a small shift between their position due to the fact that schlieren image was taken 2 µs after the PLIF image. A significant and fairly uniform level of OH along the outer edge of the jet plume attached to the recirculation area upstream of the injector is visible. However, farther downstream a decrease of OH fluorescence is obtained as the mixture expands around the jet flow-field. As explained in Section 3.2.4, the fluorescence signal in this case is essentially proportional to the OH mole fraction. Therefore the decrease in signal level observed in the sheet corrected PLIF images is a direct indication of the decrease in OH mole fraction Top View OH-PLIF Images In order to achieve a more complete picture of the combustion, we have also obtained top PLIF views of the jet. A set of 4 instantaneous top view images collected at 4 different heights above the jet exit is shown in Fig. 6.2 (the white dots in the images indicate the center of the jet exit.) The results show OH around the jet while the center of the plume has no OH-formation. The bottom image at 1-jet diameter above the plate shows two main features: 1) the jet spreads very quickly in the lateral direction (up to about 8-jet diameters) assuming that the flame is around the jet-air interface, and 2) the OH concentration downstream of the jet remains nearly constant. By contrast, the other three images obtained at 2, 2.5 and 3-jet diameters height demonstrate the same tendency of the side views, namely that the OH signal level decreases as the jet moves downstream.

140 CHAPTER 6. IGNITION OF HYDROGEN JET 118 (a) (b) (c) Air Region Illuminated by PLIF Sheet x/d FIGURE 6.1 Simultaneous OH-PLIF and schlieren images visualizing hydrogen injection into supersonic crossflow. Free-stream conditions are M = 3.57, T = 1300 K, P = 0.32 atm, V = 2500 m/s. The jet-to-freestream momentum flux ratio is J = 1.4. a) Schlieren image, b) OH-PLIF image demonstrating the ignition and combustion regions of jet-in-crossflow at high enthalpy condition, c) Overlaid OH-PLIF and schlieren images.

141 CHAPTER 6. IGNITION OF HYDROGEN JET 119 (a) y/d=3 (b) y/d=2.5 (c) y/d=2 (d) y/d=1 FIGURE 6.2 Instantaneous top-view OH-PLIF images obtained at different height above the injection plate. Free-stream conditions are M=3.57, T=1300K, P=0.32atm, V=2500m/s. The jet-tofreestream momentum flux ratio is J=1.4. a) y/d=3, b) y/d=2.5, c) y/d=2, d) y/d=1 above the injection plate.

142 CHAPTER 6. IGNITION OF HYDROGEN JET Comparison of Ignition in Flight Mach 8-13 Total Enthalpy Range Figure 6.3 compares examples of OH-PLIF imaging conducted at the center-line of the hydrogen jet injected in flight Mach 10 and 13 conditions. Each image includes 2 separate subimages from different tests: the first has been acquired near the jet exit while the second has been acquired farther downstream. Apparent in the images is the isolated thin filament along the outer edge of the plume. The center of the plume itself has no OH-signals indicating poor mixing of the air with the core of the hydrogen jet. The OH radicals are primarily produced in the hot separation region upstream of the jet exit and behind the bow shock and convected downstream with the shear-layer vortices. Ignition in our experiments is likely due to the hot and radical-rich separation region upstream of the jet exit where the boundary layer and jet fluids mix subsonically. Although these recirculation kernels in general are small in volume, in high enthalpy flows the temperature of these zones can be as high as the stagnation temperature of the bulk flow. In Mach 10 condition, a low signal level of OH in the recirculation zone/boundary layer upstream of the injector is visible. In Mach 13 condition, on the other hand, a fairly uniform and intense OH signal level is observed even when air is used as the freestream gas. The high concentration of OH radicals in the upstream recirculation region can be attributed to high recovery temperatures associated with high total enthalpy flows. The ignition process will initiate instantaneously in the upstream recirculation region as the delay times effectively approach zero ( 1 5 µs) at the Mach 13 condition. Figure 6.4 presents two instantaneous OH-PLIF images obtained at the center-line of the hydrogen jet injected at Mach 8 flight conditions. Limited amounts of OH are visible on the leading edge of the jet, mainly behind the steep regions of the bow shock. The ignition is quenched farther downstream. It is evident from these results that improved injection schemes for better flame-holding would be required for practical applications in scramjet engines flying at low Mach numbers (below 10).

143 CHAPTER 6. IGNITION OF HYDROGEN JET 121 a) Mach 10 Air M = 3.4 V = 2360 m/s P = 0.32 atm T = 1290 K H 2, J=1.4 x/d b) Mach 13 Oxygen M = 4.7 V = 3200 m/s P = 0.04 atm T = 1250 K H 2, J=5 x/d FIGURE 6.3 Instantaneous OH-PLIF acquired at center-line axis of the hydrogen jet injected into flight Mach 10 and 13 conditions. The images are obtained by combination of 2 different instantaneous images: near the exit of the jet ( 5 < x/d < 1) and downstream of the jet (1 < x/d < 10).

144 CHAPTER 6. IGNITION OF HYDROGEN JET 122 Mach 8 Air M = 2.4 V = 1800 m/s P = 0.65 atm T = 1400 K H 2 x/d H t, = 2.9 MJ/kg x/d H 2 J=2.3 FIGURE 6.4 Two instantaneous OH-PLIF images acquired at center-line axis of the hydrogen jet injected into flight Mach 8 conditions.

145 CHAPTER 6. IGNITION OF HYDROGEN JET Discussion of the Ignition Process Ignition Characteristics of Hydrogen Ignition of hydrogen and air in high-temperature turbulent flow fields associated with supersonic combustors can be characterized purely by radical runaway as opposed to thermal runaway (Im et al. 1994; Im et al. 1998; Sung et al. 1999). Thus, the heat release of combustion is not essential for the propagation of the ignition/combustion process. The well-known explosion limits of a hydrogen-oxygen system are plotted in Fig. 6.5 (after Sung et al. 1999). Nishioka and Law (1997) have shown that the state of the second explosion limit is an important boundary in the ignition response of a hydrogenair laminar mixing layer. The crossover temperature, T c, for the second explosion limit is defined as the temperature at which the rate of the main chain-terminating reaction H + O 2 + M HO 2 + M (R1) equals that of the rate-controlling branching reaction H + O 2 OH + O (R2) Reaction (R2) is a two-body, temperature-sensitive branching reaction with an activation energy of 68.8 kj/mole (16.44 kcal/mole), while (R1) is a three-body, temperature insensitive terminating reaction because HO 2 is relatively inactive. Thus increasing temperature promotes (R2) and hence the overall ignitability. When the mixture temperature in the flow, T mix, is larger than T c, the system response is controlled by (R2). The chain-branching ignition in coflow mixing layers leads to a continuous growth of the radical pool. If the temperature is sufficiently above crossover, then the effect of three body recombination reactions, responsible for most of the heat release in H 2 -O 2 combustion, is negligible, and the two streams mix and react initially without significant chemical heating. This gives rise to a thermally frozen branched-chain explosion (Sanchez et al. 1997). As a point of reference, we first calculated the ignition delay times, τ ign, for hydrogenair mixtures. The chemical kinetics software package Chemkin, developed by (Kee et al. 1980), was used for the delay time calculations. The reaction mechanism was taken from GRI-Mech 2.12 (Bowman et al.). The ignition times based on the time when [OH] = 1/2 [OH] max are plotted in Fig If the pressure and fuel-air mixture are held constant, the effect of temperature on ignition time can be readily shown in this

146 CHAPTER 6. IGNITION OF HYDROGEN JET 124 FIGURE 6.5 Explosion limits of a stoichiometric hydrogen-oxygen mixture (after Sung et al., 1999). figure. Note that the ignition time has a strong exponential dependence on temperature. Over the range 1000 K to 1700 K, τ i varies by about a factor of 100. At temperatures between K (depending on pressure) the ignition time approaches infinity and self-ignition cannot occur. We have earlier discussed that at temperatures above the crossover temperature, T c, the hydrogen-air system response is controlled by the two-body reaction (R2). Thus the product pτ i forms a single curve in Fig. 6.6b for all the pressures at high temperatures. We however observe a deviation from the binary scaling law at a critical temperature depending on pressure. This critical temperature is lower for low pressures consistent with the second explosion limit discussed above.

147 CHAPTER 6. IGNITION OF HYDROGEN JET H 2 -air for =1 Ignition Time ( sec) p=0.3atm 1atm 2atm 4atm /T(K) (a) 10 3 H 2 -air for =1 4atm 2atm 1atm 0.3atm p ign (atm- sec) /T(K) (b) FIGURE 6.6 Variation of ignition delay times τ ign of a stoichiometric mixture of H 2 and air with temperature and pressure. Calculations are perfomed using Chemkin and the GRI mechanism. a) τ ign vs. T, b) pτ ign vs. T

148 CHAPTER 6. IGNITION OF HYDROGEN JET 126 Ignition Time ( sec) T mix =1010K T mix =1120K T air =1300K T mix =1425K T mix =1265K T air =1500K T air =1800K T mix =1650K T mix =1527K T mix =1720K T air =2400K T air =3200K Equivalence Ratio, FIGURE 6.7 Variation of ignition time with fuel-air equivalence ratio, φ, for cold H 2 (T jet = 300 K) injected into hot air. The values of the ignition delay time are calculated for different air temperatures, T air Ignition in Supersonic Combustors In scramjet combustors, auto-ignition is achieved by injecting a relatively cold fuel into hot air. Therefore there is a significant variation of temperature and equivalence ratio, φ, inside a supersonic combustor. In order to approximate the variation of temperature with φ, a simplified one-dimensional enthalpy balance between the cold hydrogen and the hot air is used. The gases are assumed to be nonreacting, thermally and calorically perfect and ideally mixed without compressibility effects. Based on this model, the enthalpy balance between the two streams can be written as n air C p,air (T air T mix ) = n H2 C p,h2 (T mix T H2 ) (6.5) where n is the number of moles which is related to the equivalence ratio φ; C p is the specific heat at the mixed mean temperature. The mixture temperature, T mix, can then

149 CHAPTER 6. IGNITION OF HYDROGEN JET 127 Free-streamTemperature, T 2 (K) T 2 P 2 ign =1 s ign =2 s ign =5 s ign =10 s ign =43 s x/d Free-streamPressure, P 2 (atm) FIGURE 6.8 The free-stream temperature and pressure (T 2 and P 2) behind the bow shock, measured from schlieren images as discussed in Section (see Fig. 4.3). Ignition delay times are calculated for several conditions of air assuming φ = 0.2. The free-stream flow properties simulate the flight Mach 10 conditions. be calculated by rearranging Eq. 6.5 as T mix = T air + αt H2 1 + α (6.6) where α = n H 2 C p,h2 n air C p,air. Representative ignition delay times are plotted in Fig. 6.7 as a function of φ for various values of T air. The results indicate that OH is likely to be formed in the hot, lean regions of the mixing layer (φ = 0.2) where the temperatures are the highest. For air temperatures below 1800 K the ignition delay times are much larger than 10 µs, which exceeds the residence time of the hydrogen within the imaged region studied here (based on the hydrogen large-scale convection velocities of 2 mm/µs and 10 jet diameters) Ignition of a Hydrogen Transverse Jet The transverse jet interacts strongly with the crossflow, producing a bow shock and localized highly three-dimensional flow-field around it. Therefore, it is very difficult

150 CHAPTER 6. IGNITION OF HYDROGEN JET 128 to formulate the ignition characteristics of this field. We can however try to estimate characteristic time scales of the ignition process in an attempt to explain the OH-PLIF results. The OH-PLIF images shown in Figs. 6.3 and 6.4 are taken at the center-line of the jet. At this location, the average free-stream flow properties behind the bow shock were estimated using schlieren images (see Section for more details). Figure 6.8 shows these estimated values of static temperatures and pressures of free-stream behind the bow shock, T 2 and p 2 respectively, at the flight Mach 10 condition. Considering the values of T 2 and p 2 and using the Eq. 6.6 we calculated the ignition delay times at several locations along the x-axis of the jet. We assumed that the cold hydrogen (T H2 = 300 K) interacts with hot air at the lean side of the jet shear layer where φ=0.2. The results shown in Fig. 6.8 indicate that an instantaneous autoignition can be achieved close to the jet exit as the ignition delay times are of the order of 1-2 µs. Further downstream beyond x/d 6, the ignition delay times become larger (τ ign > 10 µs) exceeding the maximum flow residence time of the imaged region. As the air temperature behind the weaker bow shock begins to decrease below 1800 K (T mix < 1120 K for φ = 0.2) OH radicals can only be formed downstream of the imaged region, beyond x/d = 10. This might be the reason for the reduced OH mole fractions observed downstream the injection location at the flight Mach 10 condition (see Fig. 6.3a). OH radicals generated near the jet exit are convected downstream with the shear-layer vortices. However, since new radicals are not formed instantaneously downstream of the injector the OH concentration dilutes and therefore the OH-PLIF signal is reduced Ignition of Ethylene Transverse Jet In scramjet engines the hydrocarbon fuel may be partially cracked to C 1 -C 3 species through its use as a coolant before reaching the combustor. It is anticipated that this partially reacted fuel will burn as well as ethylene, namely with combustion properties (such as ignition delay) approaching that of ethylene. We therefore studied the nearfield ignition characteristics of ethylene jet as it is a good simulant for hydrocarbons in scramjets and the OH-PLIF results were summarized previously in Section Here, we calculated the ignition delay times for ethylene-air mixtures using two different chemical reaction systems. Due to the large number of elementary reactions

151 CHAPTER 6. IGNITION OF HYDROGEN JET 129 Ignition Time ( sec) H 2 -air C 2 H 4 -air using GRI-mechanism C 2 H 4 -air using LLNL-mechanism C 2 H 4 -O 2 using LLNL-mechanism 10 1 p=1atm, = T(K) FIGURE 6.9 Comparison of ignition delay of a stoichiometric mixture of C 2 H 4 (ethylene) and air/oxygen at 1 atm with a stoichiometric mixture of H 2 and air. Two different reaction mechanisms are used to calculate the ignition delay times of C 2H 4. involved in the ignition process of ethylene, a large number of data for reaction rates is required. This makes the ignition delay times of ethylene fuel difficult to calculate. The most comprehensive data is available in the GRI Mechanism 3.0 (Smith et al.). The latest version of GRI-Mech was recently updated by Marinov et al. (1998) at LLNL (Lawrence Livermore National Labs.). The calculated ignition delay times are plotted in Fig Note that the LLNL chemical mechanism which is optimized for ignition processes estimates longer ignition delay times than the GRI mechanism. As shown in Fig. 6.9, longer ignition delay times are associated with the ethylene-air mixtures compared to the values of hydrogen-air. Indeed, when we performed reacting experiments where ethylene fuel was injected into air simulating flight Mach 10 conditions, no ignition could be observed in the near-field of the transverse jet (see Fig b).

152 CHAPTER 6. IGNITION OF HYDROGEN JET 130 Ignition Time ( sec) C 2 H 4 - air for =1 p=1 atm p=2 atm p=0.3 atm C 2 H 4 - air C 2 H 4 - O T(K) FIGURE 6.10 a) Variation of ignition delay of a stoichiometric mixture of C 2 H 4 (ethylene) and air/oxygen at various pressures. Although, in our experiments the air composition, Mach number and static temperature corresponds to typical scramjet combustor entrance values, its static pressure (0.32 atm) is somewhat below that of the actual systems ( atm). This limitation can be circumvented partially by using higher concentration of oxygen in the test gas. As plotted in Fig. 6.10, the ignition delay times of a stoichiometric ethylene-oxygen mixture at p = 0.3 atm are smaller than those for an ethylene-air mixture at similar conditions. Indeed, in our experiments when ethylene was injected into pure oxygen crossflow (see Fig c), autoignition in the near-field of ethylene transverse jet could be achieved. Eventough ethylene-oxygen delay times are smaller than the ones for ethylene-air, still they are almost an order of magnitude larger than these for hydrogen-air. The reason for ethylene autoignition in our experiments can be attributed to the longer residence time of ethylene large-scale eddies which have slower convective velocities than the hydrogen ones. OH-radicals could be observed across the jet shear layer as opposed to

153 CHAPTER 6. IGNITION OF HYDROGEN JET 131 thin-filament observed around the hydrogen jet. The improved mixing properties of high molecular weight transverse jets might be the primary reason for these results (see more explanation in Chapters 4 and 5. Note that ethylene has higher temperature or activation energies for ignition than hydrogen. Hydrocarbons have effective activation energy for ignition in excess of kj/mole compared with about 60 kj/mole for hydrogen. While autoignition of ethylene fuel seems to require higher temperatures and pressures, hydrogen fuel combustion is more challenging because of the lower reaction rates. From experiments presented in this thesis, we therefore learned that autoignition of an ethylene transverse jet could be achieved at flight Mach 10 conditions. This result indicates that hydrocarbons might be a useful fuel in scramjets flying at Mach 10 conditions. This is an important result as the long ignition delay time of ethylene (hydrocarbons) relative to hydrogen is a key limitation for hydrocarbon-fueled scramjets. However, once ignition in hydrocarbon fuels is achieved, the combustion process would be completed faster than the hydrogen fueled scramjet.

154 Chapter 7 Cavity Flame-Holders This chapter describes ongoing research efforts in the scramjet community, in general, as well as our preliminary study on cavity flame-holders, a concept for flame-holding and stabilization in supersonic combustors. During the last few years, cavities have gained the attention of the scramjet community as a promising flame-holding device, owing to results obtained in flight tests and to feasibility demonstrations in laboratory scale supersonic combustors. However, comprehensive studies are needed to determine the optimal configuration which will yield the most effective flame-holding capability with minimum losses. In this chapter, we summarize the flowfield characteristics of cavities and research efforts related to cavities employed in low- and high-speed flows. Open questions impacting the effectiveness of the cavities as flame-holders in supersonic combustors are discussed (Ben-Yakar and Hanson 2001). 7.1 Review of Previous Research Cavity Flow-Field Characteristics Supersonic flow over cavities has been extensively studied (McMillin et al. 1994; Roudakov et al. 1993; Vinagradov et al. 1995; Ortweth et al. 1996; Owens et al. 1998; Ben-Yakar et al. 998a; Angus et al. 1993; Roudakov et al. 1996; McClinton et al. 1996; Huellmantel et al. 1957) for many years because of their relevance to aerodynamic configurations. A cavity, exposed to a flow, experiences self-sustained 132

155 CHAPTER 7. CAVITY FLAME-HOLDERS 133 oscillations which can induce fluctuating pressures, densities and velocities in and around the cavity, resulting in drag penalties. This problem motivated many experimental and computational studies which have been directed toward improving the understanding of the physics of cavity flows and the means to control their nature. Cavity flow regimes: In general, cavities can be categorized into two basic flow regimes depending primarily upon the length-to-depth ratio, L/D (see Fig. 7.1). In all cases, a shear layer separates from the upstream lip and reattaches downstream. For L/D < 7 10 the cavity is termed open as the upper shear layer reattaches to the back face. The high pressure at the rear face as a result of the shear layer impingement, increases the drag of the cavity. For L/D > the cavity is termed closed as the free shear layer reattaches to the lower wall resulting in significantly increased drag (see Fig. 7.1b). The critical length-to-depth ratio, at which a transition between different cavity flow regimes occurs, depends also on the boundary layer thickness at the leading edge of the cavity, the flow Mach number and the cavity width. Cavity oscillations: The cavity pressure fluctuations consist of both broadband small amplitude pressure fluctuations typical of turbulent shear layers, as well as discrete resonances whose frequency, amplitude, and harmonic properties depend upon the cavity geometry and external flow conditions. Experimental results reviewed by Zhang and Edwards (1990) found open cavities to be dominated either by longitudinal or transverse pressure oscillations (Fig. 7.1a) depending on L/D ratio and the Mach number (M ). In the short cavity filled by a single large vortex, the oscillation is controlled by a transverse mechanism, while in the long cavity filled by vortices, the oscillation is controlled by a longitudinal mechanism. The transition from transverse oscillation to longitudinal oscillation has been found to occur near L/D = 2 at Mach 1.5 and between L/D =2 to 3 at Mach 2.5. There are currently two primary models used to explain the longitudinal cavity oscillation process (Fig. 7.2). The unsteady motion of the shear layer above the cavity is the paramount mechanism for cavity oscillations and results in mass addition and removal at the cavity trailing edge (rear wall). The shear layer impinging on the rear wall causes free-stream flow to enter the cavity. As a result of the impingement, the cavity pressure increases and creates an acoustic wave (compression wave) which propagates upstream at the local sound speed and impacts the front wall. The first model proposes that this acoustic wave induces small vortices at the leading edge of the front wall which

156 CHAPTER 7. CAVITY FLAME-HOLDERS 134 (a) OPEN CAVITY FLOW (L/D < 7-10) Transverse Mechanism transition at L/D ~ 2-3 Longitudinal Mechanism D L C p 0 (b) CLOSED CAVITY FLOW (L/D > 10-13) C p 0 FIGURE 7.1 Flow-field schematics of cavities with different length to depth ratios, L/D, in a supersonic flow. a) Open cavity flow for L/D < 7 10; shear layer reattaches to the back face while spanning over the cavity. Small aspect ratio cavities (L/D < 2 3) are controlled by transverse oscillation mechanism while in larger aspect ratio cavities longitudinal oscillation becomes the dominant mechanism. b) Closed cavity flow for L/D > 10 13; shear layer reattaches to the lower wall. The pressure increase in the back wall vicinity and the pressure decrease in the front wall results in large drag losses. grow as they are convected downstream. Due to instabilities, the shear layer deflects upwards and downwards resulting in a shock/impingement event on the rear wall of the cavity. The second model, on the other hand, assumes that the acoustic wave reflection from the front wall, rather than the shedding vortices, is the cause of the shear layer deflection and therefore the impingement event on the rear wall. The oscillation loop is closed when the instability (caused either by vortex shedding or a reflected acoustic wave) propagates downstream and the mass added in the beginning of the loop is ejected at the trailing edge again. Typically, the frequency of the longitudinal oscillations is expressed in terms of the

157 CHAPTER 7. CAVITY FLAME-HOLDERS 135 Shear layer impingement at the rear wall Shedding vortices and reflected acoustic waves FIGURE 7.2 Typical longitudinal cavity oscillations are caused by the impingement of the free shear layer on the rear wall which generates travelling shocks inside the cavity. The shear layer spanning the cavity becomes unsteady as a result of these acoustic waves deflecting the shear layer up and down, and/or by the shock induced vortices generated at the front wall leading edge of the cavity. As a result unsteady waves emanate from the cavity. Strouhal number based on the cavity length (impingement length, L). S L = f ml U (7.1) Multiple peaks of comparable strength in unsteady pressure spectra were observed in compressible flow-induced cavity oscillations. These resonant frequencies can be predicted using the Rossiter s semi-empirical formula (Rossiter 1964), developed based on the coupling between the acoustic radiation and the vortex shedding (model 1), f m = m α M + 1 k U L (7.2) M and U are the free-stream Mach number and flow speed, respectively; f m is the resonant frequency corresponding to the m th mode, and a and k are empirical constants. While k represents the ratio of the speed of the convection of the shear layer vortices to the free-stream flow speed (U ), α is the phase shift between the acoustic waves and the shear layer instability. This equation is modified by Heller and Bliss (1975) for

158 CHAPTER 7. CAVITY FLAME-HOLDERS 136 compressible flows by taking into account the effect of the higher sound speed within the cavity, which is approximately equal to the free-stream stagnation sound speed. Their model assumes that the pressure fluctuations are a result of the interaction of the shear layer with the reflected acoustic waves (model 2), f m = m α M γ 1 M 2 2 k U L (7.3) where γ is the ratio of specific heats. Heller and Delfs (1996) determined from their experiments that a = 0.25, k = 0.57 for cavities with L/D ratio of 4 or more and estimated the difference between the formula and experiments as ±10%. Therefore, the oscillatory frequency of a particular mode in a shallow cavity decreases with increasing length or L/D ratio of the cavity. However, the dominant oscillatory mode (the mode with the largest amplitude) jumps from a lower mode to a higher mode as the L/D ratio increases. Stabilization techniques for cavity oscillations: Several passive (Perng and Dolling 1998; Zhang et al. 1998) and active (Sarno and Franke 1994; Vakili and Gauthier 1994; Lamp and Chokani 1997) control methods have been proposed and developed to suppress the cavity oscillations (Fig. 7.3). Since the shear layer interaction with the rear cavity wall is the main factor for fluctuations as discussed above, the stabilization or control of the shear layer can ultimately suppress the cavity oscillations. Passive control methods, which are usually inexpensive and simple, utilize mounted devices such as vortex generators and spoilers upstream of the cavity or a slanted trailing edge that modifies the shear layer so that the reattachment process does not reflect pressure waves into the cavity. These methods are found to be very effective in suppressing the cavity oscillations. However, since those are permanent devices, the performance of a cavity at different conditions may actually be worse than the performance of a cavity without passive control. A visual observation of a cavity flow-field stabilized by an oblique rear wall is shown in Fig This figure contains two instantaneous schlieren images from our recent experimental efforts demonstrating the stabilizing effect of a slanted back wall upon the shear-layer reattachment with the back wall. While the cavity with a 90 o back wall (Fig. 7.4a) emits shock waves at the trailing edge as the pressure increases due to shear layer impingement and recompression of the flow, the angled back wall shown in

159 CHAPTER 7. CAVITY FLAME-HOLDERS 137 (a) Reduced Cavity Oscillations Angled Back Wall (No Reflected Acoustic Waves) (b) Small Upstream Disturbances Enhanced Shear Layer Growth Injection FIGURE 7.3 Different concepts can be employed to suppress the cavity oscillations: a) Cavities with an angled back wall suppress the unsteady nature of the free shear layer by eliminating the generation of the travelling shocks inside the cavity due to the free-shear-layer impingement. b) In addition, small disturbances produced by spoilers or by the secondary jet injection upstream of the cavity can enhance the free-shear-layer growth rate. The thickening of the cavity shear layer alters its instability characteristics, such that its preferred roll-up frequency is shifted outside of the natural frequency of the cavity, and as a result the oscillations are attenuated. Fig. 7.4b leads to a steady shear-layer reattachment process. Active control methods, on the other hand, can continuously change to adapt to different flow conditions. Forcing of the shear layer can be accomplished by various mechanical, acoustical or fluid injection methods. The use of steady or pulsating mass injection upstream or at the leading edge of the cavity is one of the most commonly studied techniques. Various researchers (Sarno and Franke 1994; Vakili and Gauthier 1994; Lamp and Chokani 1997) have examined the feasibility of this technique. Vakili and Gauthier (1994) observed significant attenuation of cavity oscillations with upstream mass injection. This was attributed to the thickening of the cavity shear layer, which altered its instability characteristics, such that its preferred roll-up frequency was shifted outside of the natural frequencies of the cavity. Cavity Drag: Two components produce pressure drag in the cavity. First, the pressure in the backward facing step may be lower than the free stream pressure. This

160 CHAPTER 7. CAVITY FLAME-HOLDERS 138 (a) M=3.5 M>1 Reflected Acoustic Waves Shear Layer D=3mm Trailing Edge Vortex L = 15 mm C p 0 Front Wall Floor Back Wall (b) Recirculation Zone Oblique Shock Wave D=3mm 20 9 mm L = 17.2 mm C p 0 Front Wall Floor Back Wall FIGURE 7.4 Instantaneous schlieren images with 200 ns of exposure time demonstrating the effect of the back wall angle on the flowfield structure of a cavity exposed to a supersonic flow. The freestream was generated in an expansion tube to simulate Mach 10 total enthalpy conditions at the supersonic combustor entry: M = 3.4, U = 2360 m/s, T = 1290 K, p = 32 kpa. The boundary layer thickness at the trailing edge of the cavity is approximately 1mm. a) Cavity with L/D = 5 shows the unsteady nature of the shear layer at the reattachment with the trailing edge of the back wall. b) Cavity with slanted back wall (20 o ) stabilizes the shear layer reattachment process. results in a net force in the positive x-direction (drag force) acting on the base area (base pressure higher than freestream would result in a thrust force). Second, the reattachment of the shear layer at the back wall produces a region of high pressure that imparts a force in the positive x-direction acting on the forward facing area. In Fig. 7.5, the magnitude of pressure fluctuations on the floor of the cavity and the drag coefficient for different L/D are given, as adapted from Zhang and Edwards (1990). Their experimental results demonstrate a sharp rise of the oscillatory level and the drag when the oscillatory mode inside the cavity changes from a transverse mode to a longitudinal one. The magnitude of the fluctuations decreases gradually with the

161 CHAPTER 7. CAVITY FLAME-HOLDERS 139 (a) P rms / ( U 2 /2) M=1.5 M=2.5 (b) M=1.5 M=2.5 L/D Cavity Drag Transverse Mode Longitudinal Mode L/D FIGURE 7.5 Effect of length-to-depth ratio, L/D on a) magnitude (root-mean-square) of pressure fluctuations on the bottom of the cavity (at x/d = 0.33), b) drag of the cavity at Mach 1.5 and 2.5 flows. The values were adapted from Zhang and Edwards (1990). increasing L/D of the cavity, while the average drag coefficient, however, rises significantly. As the L/D of the cavity increases, the shear layer thickens at the reattachment point damping the oscillations and simultaneously increasing the pressure on the back wall of the cavity. Subsequently, the time-mean pressure on the upstream wall of the cavity drops as a result of the momentum diffusion across the shear layer. These combined effects of increasing pressure in the back wall of the cavity and decreasing pressure in the upstream wall of the cavity, increase the drag of the cavity. The drag penalties become larger as the cavity L/D ratio reaches a critical value at which the closed cavity flowfield is established. The drag coefficient of an open cavity is affected greatly by the cavity back wall

162 CHAPTER 7. CAVITY FLAME-HOLDERS 140 geometry. Gruber et al. (999a) studied the drag penalties of open cavities with θ = 16 o and 30 o angled back wall, where θ is defined as the angle relative to the horizontal wall (see Fig. 7.3). They concluded that the drag coefficient increases for shallower back wall angles. First, the small back wall angles lead to the formation of an expansion wave (rather than a compression wave) at the cavity leading edge that reduces the pressure on the backward facing step adding drag. Second, the shear layer deflects farther into the cavity which results in a larger area of recompression on the angled back wall, again increasing the drag. In contrast to Gruber et al. (999a) findings, numerical calculations of Zhang et al. (1998) resulted in a reduced average drag coefficient as the back wall angle is decreased from θ = 90 o to θ = 67.5 o and 45 o. The observations from these two references agree, however, that the pressure on the upstream face of the cavity decreases with decreasing back wall angle. It is possible that in the 67.5 o and 45 o cases studied by Zhang et al. (1998), the compressive nature of the separation wave at the upstream corner of the cavity actually keeps the shear layer from deflecting into the cavity and could result in lower levels of pressure drag than the 16 o case that Gruber et al. (999a) studied. In a different study, Samimy et al. (1986) used a cavity with a 20 o of back wall angle to create an undisturbed free shear layer. This geometry was chosen such that the wall pressure across the cavity would stay unchanged, thereby minimizing the drag losses associated with the shear layer deflection inside the cavity. These observations suggest that there might be a critical back wall angle (between θ = 45 o to 16 o ) at which the drag penalties of a cavity are minimal. A qualitative description of the pressure distribution along the back wall surface of cavities with and without an angled wall is plotted in Fig In a rectangular cavity, below the shear layer reattachment point, the trailing edge vortex accelerates the flow and causes a pressure decrease in the middle of the back wall. On the other hand, in the cavity with the angled wall, the high pressure at the corner of the cavity disappears and a monotonic increase of pressure takes place behind the reattachment point. The drag coefficient depends strongly on the back wall pressure distribution as it is altered by the cavity geometry. Further comprehensive studies are required to complete our understanding of cavity geometry, particularly the effect of the back wall angle on the drag penalty. Cavity Residence Time: Residence time, τ, of the flow inside a cavity is a direct

163 CHAPTER 7. CAVITY FLAME-HOLDERS 141 function of the mass exchange rate in and out of the cavity. In the open cavities, mass and momentum transfer mechanism are controlled by the longitudinal oscillations and the vortex structure inside the cavity. Computational visualizations of Gruber et al. (999a) demonstrate the existence of one large vortex stationed near the trailing edge of the cavity and a secondary vortex near the upstream wall. The mass exchange of the cavity is controlled by the large trailing vortex which interacts with the unstable shear layer. The mass exchange between the vortices inside the cavity, on the other hand, is relatively small, and therefore as the trailing edge vortex occupies larger volume inside the cavity, the mass exchange increases and flow residence time inside the cavity decreases. Consequently, the steady-state numerical calculations showed that the flow residence time in a large cavity (L/D = 5) is smaller than the value in a small cavity (L/D = 3), in contrast with expectation. Although the volume of the cavity increases (increases τ) with increasing length, the mass exchange rate, on the other hand, increases even more (decreases τ), resulting in a decreased residence time. However, it is not yet clear how the flow residence time inside a cavity is affected by the unsteady nature of the cavity. The steady state computations (Gruber et al. 999a) mentioned above, estimated that 1 msec is the order of magnitude of residence time in an L/D = 5 cavity with 9 mm depth size in a Mach 3 cold flow. This value decreases for slanted wall cavities. As summarized above, the cavity is a basic fluid dynamic configuration which generates both fundamental and practical interest. A cavity is often characterized by a strong oscillation inside the cavity driven by the instability of the shear layer. Hence, these oscillations may be controlled and suppressed by the stabilization of the shear layer Cavity in Reacting Flows In the past few years, the use of cavities has been considered as a means of performance improvement in a supersonic combustor. Basically there are two main directions in which several research groups have focused their efforts: 1) cavity-actuated mixing enhancement, 2) trapping a vortex within the cavity for flame-holding and stabilization of supersonic combustion. Some recently performed studies, investigating the above concepts are summarized in the following sections.

164 CHAPTER 7. CAVITY FLAME-HOLDERS 142 Cavity-Actuated Supersonic Mixing Enhancement It is known that the normalized growth rate of the mixing layer between supersonic air and gaseous fuel in a scramjet combustor decreases as the convective Mach number increases due to compressibility effects (Papamoschou and Roshko 1988). Researchers suggested that cavity flow oscillations can actually be used to provide enhanced mixing in supersonic shear layers. A shear layer develops instability waves in its initial region. This long wavelength Kelvin-Helmholtz instability which leads to large rollers are suppressed at high convective Mach numbers. As a method to enhance the K-H instability, Kumar et al. (1989) suggested using oblique oscillating shock waves of high frequency, and Yu and Schadow (1994) concluded that for the required frequency excitation, transverse acoustic waves emanating from cavities are powerful enough to affect mixing in a significant manner. Yu and Schadow (1994), therefore, suggested use of cavities to enhance the mixing of supersonic non-reacting and reacting jets, where the cavity was attached at the exit of the jet circular nozzle, Fig. 7.6b. When the cavity was tuned for certain frequencies, large scale highly coherent structures were produced in the shear layer substantially increasing the growth rate. The spreading rate of the initial shear layer with convective Mach number M c = 0.85 increased by a factor of three, and for jets with M c = 1.4 by 50%. Finally, when the cavity-actuated forcing was applied to reacting supersonic jets, 20 30% reduction in the afterburning flame length with modified intensity was observed. Sato et al. (1999) also studied the effect of an acoustic wave, emitted from a cavity and impinging on the initial mixing layer, Fig. 7.6a. Their results revealed that the mixing was enhanced by the acoustic disturbance and the rate of the enhancement was controlled by the cavity shape while the total pressure losses were negligibly small. This novel use of cavity-induced oscillations in turbulent compressible shear layers to control the mixing rate, which have been demonstrated in the experiments mentioned above, encourages the use of unstable cavities in high speed propulsion applications. However, before implementing such techniques, one must consider and evaluate the potential thrust loss and noise generation associated with the technique.

165 CHAPTER 7. CAVITY FLAME-HOLDERS 143 (a) CAVITY acoustic wave M>1 mixing layer fuel injection (b) SUPERSONIC NOZZLE initial mixing layer supersonic jet CAVITY cavity actuated mixing layer FIGURE 7.6 Cavity-actuated supersonic mixing enhancement concepts: (a) Sato et al. (1999), studied the influence of acoustic waves, emitted from a cavity and impinging on the initial mixing layer. (b) Yu and Schadow (1994) used the same concept to enhance the mixing of supersonic reacting jets. Cavity as a Flame-Holder While an unstable cavity can provide enhancement in the turbulent mixing and combustion as discussed above, a stable cavity can be used for flame-holding applications. In an effort to reduce the combustor length required for efficient high speed combustion, during the past few years, the scramjet community has proposed the use of wall cavities to stabilize and enhance supersonic combustion. The main idea is to create a recirculation region inside the cavity with a hot pool of radicals which will reduce the induction time, such that autoignition of the fuel/air mixture can be obtained. However, for a stable combustion process, the cavity recirculation region has to be sufficiently stable to provide a continuous ignition source (pilot flame). As discussed above, it is possible to control the self-sustained oscillations occurring in cavities either by proper design of

166 CHAPTER 7. CAVITY FLAME-HOLDERS 144 the cavity or by a passive/active control system. In the following sections, we will first discuss the literature for low-speed and then the recent advances in high-speed combustors which utilize cavity flame-holders. Cavity TV - Trapped Vortex concept in low-speed flows: Recently, cavities have been employed in low-speed flows to stabilize combustion utilizing the so-called trapped-vortex concept (Hsu et al. 1998). In this concept, a stationary vortex is established or trapped inside the cavity by optimal design of the dimensions, namely, by optimal cavity length to depth ratio (L/D). It is known that a vortex will be trapped in the cavity when the stagnation point is located at the downstream end of the cavity which also corresponds to the minimum drag configuration (Heller et al. 1970). Based on this evidence, Hsu et al. (1998) designed an experimental cavity to investigate the low-speed flame stability characteristics of a trapped-vortex combustor, while Katta and Roquemore (998a, 998b) performed numerical calculations for this geometry. Their results showed that a vortex is locked in a short cavity (L/D < 1). However, when a vortex is trapped in the cavity, very little fluid is entrained into the cavity, resulting in very little exchange of the main flow and cavity fluid. When flame stabilization is a consideration, a continuous exchange of mass and heat between the cavity and the main flow is required. To overcome this problem, it has been suggested to directly inject both fuel and air into the cavity in a manner that reinforces the vortex and increases mass transfer of the reactive gases with the freestream. The main conclusions revealed from low-speed cavity flame-holder studies can be summarized as follows: 1. In non-reacting flows, a stable cavity flow was observed at an optimal dimension (L/D = 0.6) that produces minimum drag, namely, minimum pressure drop. This was also the optimal cavity length which provided the most stable flame. 2. A sufficient amount of fuel and air must be injected directly into the cavity to obtain good performance characteristics of a combustor with a trapped-vortex cavity. 3. The fluid injection inside the cavity had a strong impact on the stability of the vortex inside the cavity. When jets were injected in such a way that they reinforced the vortex, the flame stabilization capability of the cavity was enhanced.

167 CHAPTER 7. CAVITY FLAME-HOLDERS The optimum size (L/D) for steady flow should be larger in the case of cavities with fluid injection than for cavities with no injection. Cavity flame-holders in high-speed flows: In the scramjet community, there is a growing interest in the use of cavity flame-holders. In a 1997 Air Force/NASA workshop (Tishkoff et al. 1997), an integrated fuel injector/cavity flame-holder was mentioned as one of the new concepts that may provide potential performance gain in a scramjet engine. It was indeed very encouraging to see this new concept employed and flight-tested in the scramjet engine by the Central Institute of Aviation Motors in Moscow (Roudakov et al. 1993; Vinagradov et al. 1995; Ortweth et al. 1996; Owens et al. 1998; Roudakov et al. 1996; McClinton et al. 1996). The combustor of the axisymmetric scramjet engine, illustrated in Fig. 7.7, included three fuel injection stages, two with cavity flame-holders (D = 20 mm by L = 40 mm and D = 30 mm by L = 53 mm) and one with a step flame-holder (D = 17 mm). The injection of the fuel (hydrogen) was performed within the cavity flame-holders from the front-facing wall at 30 o to the engine axis and just upstream of the step at 45 o. With this integrated injection/cavity flame-holder approach, numerical studies (McClinton et al. 1996) showed that autoignition and flame-holding within the cavity could be obtained at Mach 6.5 flight, even without the spark ignition plugs. Their analysis also revealed that without the cavity, the ignition is unlikely due to the small injector dimension (d j = mm) and low combustor operation pressure (p 0.4 atm) as estimated previously by Huber et al. (1979). Finally, the joint Russian / U.S. effort demonstrated in the flight test performed on February 12, 1998 that a positive thrust from the scramjet engine could be successfully achieved. One can find several recent studies investigating cavities for flame stabilization of a supersonic combustor. Some of these works, performed for different kinds of fuels (solid, liquid and gaseous fuels), are summarized as follows. The combustion of kerosene in a scramjet requires additional ignition and flameholding elements because of the long ignition times and reduced reaction rates as compared to hydrogen. Owens et al. (1998) tried to determine the flame stability of kerosene injected upstream of a cavity flame-holder with Mach 1.8 free-stream conditions. Due to the low stagnation temperatures of 1000 K, ignition was provided by pilot hydrogen fuel injected into the cavity. Flame-holding could be achieved only when large flow rates

168 CHAPTER 7. CAVITY FLAME-HOLDERS Injector Step Injector 30 Cavity 1 20 x Injector Cavity x 53 FIGURE 7.7 Axisymmetric combustor of the Scramjet engine which was flight-tested by Russian- CIAM/NASA joint program (1998). In this engine two cavities with angled-rear wall were used for flame-holding purposes. The dimensions are in mm (McClinton et al. 1996). of hydrogen were used. In this case the enlargement of the recirculation region led to entrainment of additional quantities of fresh air contributing to the flame stability. An additional investigation of scramjet combustors operating on kerosene was performed by CIAM (Vinagradov et al. 1995). In their configuration, the combustion was sustained by a row of hydrogen fuel injectors placed in front of a cavity. The use of cavities as flame-holders in solid fuel supersonic combustors has been also studied (Ben-Yakar et al. 998a; Angus et al. 1993). In the experiments of Ben-Yakar et al. (998a), self-ignition and sustained combustion of PMMA (Plexiglas) solid fuel with no external aid (such as reactive gas injection or a pilot flame) was demonstrated under supersonic hot-air flow conditions. This was accomplished by a recirculation region formed inside a cavity which was positioned at the entrance of the combustor. Typically, in a subsonic solid fuel ramjet, a step is used for flame-holding purposes, and it is known that larger step heights (leading to bigger recirculation zones) can provide better flame stabilization. However, in supersonic flows where a large step is required, the free-stream flow velocity would increase as well by the sudden expansion, deteriorating the flame-holding capability. Under those considerations, a cavity consisting of a step followed by an angled wall was chosen as a flame-holder in the supersonic solid fuel experiments mentioned above. The results revealed that both the cavity length (L) and the step height (D) have significant effect in sustaining the combustion. While short L caused flameout even for relatively large D, the inverse, namely small D, did not permit sustained combustion even though L was quite long. Ultimately, cavity length-to-depth ratio between 1.7 < L/D < 2 showed a regime of sustained combustion. Besides the use of cavities in liquid and solid fueled supersonic combustors, there are

169 CHAPTER 7. CAVITY FLAME-HOLDERS 147 other research groups (Yu et al. 1998; Yu et al. 1999; Niioka et al. 1995; Mathur et al. 1999; Davis and Bowersox 997a; Davis and Bowersox 997b) concentrating on characterization of cavity flame-holders in gaseous supersonic combustors. Initial experimental efforts were performed by Yu et al. (1998, 1999). They analyzed flow stability and flame-holding characteristics of several wall cavities with various sizes and aspect ratios (L/D = 0.5, 1, 2, 3 and inclined cavity) in a Mach 2 air-stream. Pressure oscillations, observed in cold flow experiments, were diminished in reacting flow, when the thin shear layer above the cavity disappeared by three fuel jets injected at 45 o upstream of the cavity. Typically, small aspect ratio (1 < L/D < 3) cavities appeared to be good flameholders, which is consistent with the trapped vortex concept discussed above. The narrow cavities (L/D = 0.5) provided very steady flame-holding, however they had relatively little effect on the downstream emission characteristics. With the inclined cavity, which was also the longest cavity tested (L/D = 5), no flame-holding was observed. Additional experiments were conducted by Niioka et al in Mach 1.5 airflow. They achieved flame stabilization using two struts and by injecting hydrogen gas in the interval between the two parts. They showed that flame stability could be controlled by the cavity length which controls the competition between the mass transfer rate and the chemical reaction rate, i.e., the Damköhler number. Wright-Patterson Air Force Research Laboratories have also initiated a program (Mathur et al. 1999; Davis and Bowersox 997a; Davis and Bowersox 997b) to examine the effectiveness of cavities in supersonic flows. Experiments on a cavity with upstream ethylene fuel injection were performed in the supersonic combustor facility operating at conditions that simulate flight Mach numbers between 4 and 6. Initial results demonstrate flame-holding and large flame spreading in the cavity vicinity. In parallel, Mach 3 cold flow research is also in-progress to study the fundamental aspects of cavities. The results showed that: 1. The cavity geometry had an effect on mass entrainment rate and residence times. A decrease in cavity residence time was observed in cavities with longer length and slanted walls. 2. In general, the length of the cavity determined the mass entrainment, while the cavity depth determined the cavity residence time.

170 CHAPTER 7. CAVITY FLAME-HOLDERS Larger cavities (L/D = 7) had significantly higher drag coefficients than the smaller cavities (L/D = 3). Reduction of the back wall angle below 90 o resulted in additional drag penalties. 4. Cavities with offset ratios larger than 1 (upstream wall height is larger than the back wall height) caused the cavity base to experience lower pressures and therefore larger drag penalties. In addition, Davis and Bowersox (997a, 997b) used a combined CFD/perfectly stirred reactor methodology as a design guide for sizing of the cavity. He recommends that initial cavity size can be estimated based on the minimum residence time required to obtain ignition by assuming a perfectly stirred reactor cavity flow. Similar to Gruber et al. (999a), he concluded that cavity depth, D, which mainly controls the residence time, can be estimated using their numerically obtained empirical equation: D = τ r U/40, where τ r is the required residence time for ignition and U is the free-stream velocity Outstanding Questions As discussed above, during the last few years, cavities have gained attention as promising flame-holding devices. However, comprehensive studies still need to be performed to determine optimal configurations which yield the most effective flame-holding capability with minimum losses. We can pose the following questions concerning the effectiveness of the cavities as stable flame-holders in supersonic combustors: 1. Can the TV concept be used in supersonic combustors? Several investigators have recognized the aerodynamic advantages of trapping vortices inside small aspect ratio cavities (L/D < 1) both as a means of reducing the drag penalties of cavities and also obtaining stable flame-holding in a low-speed combustor. Stable small aspect ratio cavities may possibly be adapted to provide sustained combustion in supersonic flows. However, the cavity flow residence time associated with high-speed flows will be smaller than in low-speed flows and might eliminate its flame-holding capability. Therefore, stable cavities may possibly be adapted to

171 CHAPTER 7. CAVITY FLAME-HOLDERS 149 provide sustained combustion in supersonic flows as long as the Damköhler number is larger than unity (namely, the residence time inside the cavity is sufficient to initiate the ignition process). For example in the flight tested scramjet engine, designed by CIAM and NASA, fuel was injected within the cavity flame-holder to provide autoignition and flame-holding (McClinton et al. 1996). Otherwise, autoignition was unlikely due to the low total enthalpies of the Mach 6 flight condition, and small injector dimensions and the low combustor pressures of the design point. 2. What are the cavity dimensions and its geometry? Open cavities with L/D < 7 10 are good candidates for flame-holding owing to their reduced drag coefficients relative to the closed cavities. The dimensions of an open cavity have to be derived from ignition and flame-holding considerations. The cavity depth can be determined according to the required residence time to initiate ignition. The cavity length, on the other hand, has to be chosen to provide sufficient volume of radicals to sustain the combustion further downstream. 3. Can an unstable cavity be used to establish flame-holding? While a stable cavity is preferable to sustain continuous and stable combustion, an unstable cavity can be used to enhance mixing and ignition by the shock waves emitted as a result of strong cavity oscillations. However, unstable cavities are unlikely to provide a continuous flame-holding region inside the cavity as will also be shown in our preliminary ignition experiments (see Section 7.2). 4. How does fuel injection affect the cavity flow-field? Jet injection upstream or inside the cavity can alter the shear layer specifications (its thickness and stability) directly, and therefore, the cavity performance. Raman et al. (1999), for example, have found that jet interaction with a cavity can produce different oscillation frequencies. 5. How does the cavity flow-field affect a fuel jet injected upstream? Shock waves emanating from a cavity can enhance the mixing of fuel jets injected upstream of the cavity. As shown by several researchers, the acoustic waves of an unstable cavity can be used to enhance mixing. On the other hand, a stable cavity can also enhance mixing. As the jet reaches to the back wall it interacts with

172 CHAPTER 7. CAVITY FLAME-HOLDERS 150 the strong trailing edge shock wave of the cavity. It is known that an obliqueshock-wave-jet interaction enhances the molecular mixing between supersonic air and gaseous fuel by the vorticity generated due to the baroclinic torque. This might have immediate significance to the spreading rate of the jet and mixing enhancement of the fuel-air, resulting in enhanced combustion efficiency. 6. Is local wall heating inside the cavity a problem? High total temperatures of air stagnating inside the cavity can result in excessive heat transfer to the walls. However, the transpiration technique of mass addition from a porous surface can be used as a way to cool the cavity surfaces. This method can, furthermore, decrease the skin friction losses on the cavity floor surface and reduce the drag losses associated with the shock wave structure of the cavity (Castiglone et al. 1997). Fuel mass bleeding inside the cavity can alter the shear layer bending towards the cavity by increasing the cavity pressure distribution. In this way, the strong trailing edge reattachment shock wave can be eliminated or reduced in strength. Therefore, an optimized transpiration cooled cavity may also be designed to improve the pressure losses and the drag penalties. 7. At which flight conditions can a cavity flame-holder be effective? At high flight Mach numbers, beyond Mach 8, the velocity and the total enthalpy of air entering the combustor is high. In this hypersonic flight regime, hydrogen fuel is preferred owing to its reduced combustion characteristic times. Ignition of the hydrogen-air system can be purely characterized by radical runaway without the need for thermal feedback (substantiated by direct numerical analysis of Im et al. (1998)). Therefore, for a hydrogen-air system, a cavity flame-holder, in which the high-stagnation temperatures will initiate ignition by radical runaway, can be designed even though no appreciable heat has yet been released. As we move into lower flight speeds, below Mach 8, application of a flame-holder becomes crucial. In this supersonic flight regime, the selection of a cavity flame-holder is required to achieve longer flow residence times inside the cavity because of the reduced total enthalpies and longer ignition delay times associated with hydrocarbon fuels, which are the candidate fuels for supersonic flight below Mach 8. Consequently, cavities can be utilized in a wide range of flow conditions, in both supersonic and hypersonic airbreathing propulsion systems.

173 CHAPTER 7. CAVITY FLAME-HOLDERS 151 L/D=3 x/d=1.5 x/d=0.5 L/D=5 P P cavity pressure transducers FIGURE 7.8 Position of pressure transducers located at the bottom of the cavity to measure the history of the flow oscillations inside the cavity. Pressure transducer located farther downstream at x/d = 1.5 provided a more accurate oscillation frequency measurements. 7.2 Preliminary Cavity Results Here, we will summarize our preliminary results, where the primary objective is to demonstrate the feasibility of the experimental set-up to provide information for cavity flame-holder studies. These appear to be the first cavity experiments performed in such high total enthalpy flows. The pressure is measured within the cavity to monitor the time dependent acoustic field. The fast response pressure measurements are performed at 1.5 D downstream the forward face of the cavity as illustrated in Fig Established cavity oscillations are observed and a sequence of oscillation cycles could be captured during the limited test time ( 300 µs) of the flow facility. Figure 7.9 shows examples of the cavity pressure development at x/d = 1.5 during a typical expansion tube run. These pressure traces are obtained for 4 different geometries: a) L/D = 3, b) L/D = 5, c) L/D = 5 with hydrogen injection upstream of the cavity and d) L/D = 7. After the arrival of incident shock wave, different flow regimes exist including: a time during which helium is flowing; a period during which the contact region is passing; and a window in which there is a steady flow of test gas before the arrival of rarefaction waves. These flow periods can clearly be visualized in cavity pressure measurements where the pressure rises by the arrival of contact surface. After

174 CHAPTER 7. CAVITY FLAME-HOLDERS 152 (a) 1.0 L/D = 3 Test Gas 0.5 P cavity, AU 0.0 (b) (c) P cavity, AU P cavity, AU ~18 s t, sec L/D = 5 Test Gas ~38 s t, sec L/D = 5 with Test Gas 30 o injection ~38 s (d) P cavity, AU t, sec L/D = 7 Test Gas ~53 s t, sec FIGURE 7.9 Examples of cavity pressure traces in arbitrary units: a) L/D = 3, b) L/D = 5, c) L/D = 5 with upstream hydrogen injection, d) L/D = 7. t = 0 represents incident shock arrival at the cavity. The free-stream (N 2) conditions represent Mach 10 total enthalpy at the supersonic combustor entry: M = 3.4, U = 2360 m/s, T = 1290 K, p = 32 kpa.

175 CHAPTER 7. CAVITY FLAME-HOLDERS 153 TABLE 7.1 Summary of cavity oscillation frequencies, f m, for different cavity length to depth ratios, L/D. The table includes the expected values based on Rossiter s formula and the ones measured in our experiments. L/D Rossiter s formula Present data Difference f m, khz T m, µs T m, µs % ± ± ± ± ± ± 2 a period of flow establishment time (about 200 µs), several cycles of distinct pressure oscillations are observed during the steady flow test time. The measured frequencies, averaged from several experiments, together with the expected frequencies estimated from Rossiter s empirical formula are summarized in Table 7.1. A good agreement within 6 % could be achieved for the frequency measurement of L/D = 3. For larger L/D ratios, the agreement is weaker and the difference between the measured and the expected frequencies reaches up to 20 %. This difference might be due to the difference in total temperature ( 4000 K) of the flow tested in our experiments. Note that the formula of Rossiter s contains two empirical parameters derived from cold flow experiments. However, before any conclusion can be made, we have to analyze the flow-field dynamics of different cavity geometries in more detail. These results show that 2-D cavities with the geometry chosen here (D = 3 mm, L/D =3, 5, 7) can be studied in an expansion tube facility despite the limited test time. However, a longer test time is required to achieve accurate values of the acoustic field frequencies inside/around the cavity Visual Observation of Cavities Using Ultra-Fast Schlieren Flows around two-dimensional cavities are investigated to show the effects of variations in length-to-depth ratio (L/D) of the cavity. For this purpose cavities with L/D =3, 5 and 7 are tested. In addition, a 30 degrees angled rear wall was tested for the cavities with L/D =3 and 5. All the cavities have the same depth of 3 mm. The different length-to-depth ratios are formed by removable back wall inserts. Figure 7.10 summarizes examples of instantaneous schlieren images (200 ns exposure

176 CHAPTER 7. CAVITY FLAME-HOLDERS 154 time) obtained for cavities with L/D = 3, 5, and 7 that we have tested. In all cases, the boundary layer separates from the upstream lip and reattaches downstream. As the boundary layer separates from the leading edge of the cavity, a free shear layer forms. Depending upon the pressure inside the cavity the shear layer deflects upwards or downwards producing a compression or an expansion wave consequently. For the cavity with L/D = 3 (Fig. 7.10a) a compression wave appears at the leading edge of the cavity. As the cavity length is increased to L/D = 5 (Fig. 7.10b) this compression wave weakens. Eventually, for the cavity with L/D = 7 (Fig. 7.10c), it diminishes and an expansion wave at the leading edge takes place instead of the compression wave. The strongest shock wave appears from the trailing edge of the cavities which is consistent with the numerical observations of Zhang et al. (1998). The shear layer deflects downward near the trailing edge of the cavity and creates a high pressure stagnation point on the downstream face (the dark regions as can be seen in schlieren images in Figs. 7.10a and 7.10b). While this strong shock wave seems to be attached to the trailing edge of the cavities with L/D = 3 and 5, it moves upstream off the back wall for the cavity with L/D = 7. There are generally two basic flow regimes that a cavity can yield depending upon the length-to-depth ratio, L/D (see Fig. 7.1). A cavity is termed open if the shear layer reattaches to the back face while the drag of the cavity is small (Fig. 7.1a). Beyond some critical L/D ratio, the cavity is termed closed as the free shear layer reattaches to the lower wall resulting in significantly increased drag. Based upon this description, in our experiments cavities with L/D = 3 and 5 demonstrate open cavity flow-field features, while a cavity with L/D = 7 seems to be in the transition regime between open and closed cavity. These results are also consistent with the numerical calculations of Baurle and Gruber (1998). Their results showed that larger cavities (L/D = 7) have significantly higher drag coefficients than the smaller cavities (L/D = 3). Figures 7.10d and 7.10e demonstrate instantaneous flow-field structure of cavities with a back wall angled at 30 degrees. These schlieren images reveal that the leading edge compression waves observed for the cavities with L/D = 3 and 5 do not exist with an angled aft wall cavity. However, the strong trailing edge shock wave still exits as the shear layer reattaches at the angled back wall. Additional experiments are performed in order to study the influence of upstream

177 CHAPTER 7. CAVITY FLAME-HOLDERS 155 (a) L/D = 3 (d) L/D = 3 with 30 o angled back wall (b) L/D = 5 (e) L/D = 5 with 30 o angled back wall (c) L/D = 7 FIGURE 7.10 Schlieren images demonstrating the differences in the flow-field structure of cavities with different length-to-depth ratios and back wall angle. The depth of the cavities is constant and equal to D = 3 mm. The free-stream was generated to simulate Mach 10 total enthalpy conditions at the supersonic combustor entry: M = 3.4, U = 2360 m/s, T = 1290 K, p = 32 kpa. The boundary layer thickness at the trailing edge of the cavity is approximately 1mm.

178 CHAPTER 7. CAVITY FLAME-HOLDERS 156 (a) L/D = 3 (b) L/D = 3 with 30 o back wall and upstream injection (c) L/D = 5 with upstream injection FIGURE 7.11 Schlieren images demonstrating jet interaction with different cavities. The hydrogen jet is injected into a non-reacting free-stream 3 mm upstream of the cavity from a d = 1 mm orifice. The injection is performed at angle of 30 o to the plate. The free-stream, N 2, represents the flight Mach 10 burner entry conditions.

179 CHAPTER 7. CAVITY FLAME-HOLDERS 157 fuel injection onto the cavity flow field. The results shown in Fig include 30 degrees hydrogen injection from a 1 mm diameter orifice positioned at 3 mm upstream of the cavity leading edge. Those schlieren observations present no significant differences in the cavity flow field structure due to the upstream injection. On the other hand, some cavity influence on the jet can be observed. The jet seems to be disturbed as it propagates over the cavity for the case of L/D =3 and 5 by the compression waves at the leading edge. As this leading edge shock wave diminishes for the cavities with an angled wall the jet is not disturbed as it spans over the cavity until it reaches to the trailing edge. As the jet reaches to the back wall it interacts with the strong trailing edge shock wave emitting from the cavity. This shock wave - jet interaction at the trailing edge of the cavity might have an important role in reacting cases. It is known that an oblique shock wave - jet interaction enhances the molecular mixing between supersonic air and gaseous fuel. The vorticity generated when a shock wave interacts with a shear layer (due to the baroclinic torque) has immediate significance to the mixing enhancement in supersonic flows resulting in enhanced combustion efficiency. Furthermore, the trailing edge shock is expected to direct the fuel jet towards the airflow, increasing the fuel penetration, the static pressure, the static temperature, and therefore the reaction rates. Since the fuel jet penetration is expected to decrease as the angle of injection decreases, an improved transverse penetration and lateral spreading of the fuel might be achieved by interaction of the jet with the cavity leading edge shock wave Preliminary Ignition Results of Injection/Cavity Schemes The ignition of a hydrogen jet interacting with a cavity is studied using OH-PLIF. Several instantaneous OH-PLIF images obtained at the center-line of the jet are presented in Fig for cavity with L/D = 3 and in Fig for cavity with a 30 o angled back wall. Hydrogen is injected 3 mm upstream the cavity leading edge at an angle of 30 o and the free-stream (air) properties represent the flight Mach 10 burner entry conditions: M = 3.4, U = 2360 m/s, T = 1290 K, p = 32 kpa. For a cavity with L/D = 3, five images (see Fig. 7.12) are obtained from different experiments. The unsteady nature of the jet/cavity interaction is apparent as the intensity level of OH radicals change from experiment to experiment. Figure 7.12a shows

180 CHAPTER 7. CAVITY FLAME-HOLDERS 158 an ignition along the jet/cavity shear layer with a little OH fluorescence signal inside the cavity. As the jet interacts with the hot cavity air an auto-ignition can be achieved at the interface. The jet/cavity shear layer impinging on the back wall causes jet and free-stream flows to enter the cavity. Figures 7.12b-e show significant levels of OH concentration inside the cavity. The intensity differences between the images might be a result of the breathing motion of the cavity. However, quenching of OH is observed downstream of the cavity for all of the five experiments, indicating that this cavity configuration can not provide flame-holding. The OH-signal levels from shot to shot are more uniform inside the cavity with an angled back wall as the cavity oscillations are suppressed (see Fig. 7.13). This cavity configuration provides a continuous ignition downstream of the cavity. Also visible in Fig. 7.13a is the ignition on the upper side of the jet, most likely induced by the shock wave attached to the cavity back wall. In summary, the few OH-PLIF images indicate no ignition around the jet even in the high total enthalpy conditions of flight Mach 10 due to the shallow injection angle. Addition of a cavity downstream the injection port does provide ignition in the nearfield. However, the flame-holding capability of these cavities need to be examined in more detail.

181 CHAPTER 7. CAVITY FLAME-HOLDERS 159 (a) (b) (c) (d) (e) FIGURE 7.12 Instantaneous OH-PLIF images demonstrating the ignition regions of a hydrogen jet interacting with a L/D = 3 cavity. The images are obtained from 5 single shots at the same conditions. Hydrogen is injected 3 mm upstream the cavity leading edge at an angle of 30 o. The free-stream (air) properties represent the flight Mach 10 burner entry conditions: M = 3.4, U = 2360 m/s, T = 1290 K, p = 32 kpa. Note that a schlieren image is also included to indicate the flow-field properties around the cavity.

182 CHAPTER 7. CAVITY FLAME-HOLDERS 160 (a) (b) (c) FIGURE 7.13 Instantaneous OH-PLIF images demonstrating the ignition regions of a hydrogen jet interacting with a cavity with L/D = 3 and 30 o back wall. The images are obtained from 5 single shots at the same conditions. Hydrogen is injected 3 mm upstream the cavity leading edge at an angle of 30 o. The free-stream (air) properties represent the flight Mach 10 burner entry conditions: M = 3.4, U = 2360 m/s, T = 1290 K, p = 32 kpa. Note that a schlieren image is also included to indicate the flow-field properties around the cavity.

183 Chapter 8 Concluding Remarks 8.1 Summary of Major Results and Conclusions Experimental Aspects We have demonstrated the feasibility of using an expansion tube to generate clean (radical-free), high-total enthalpy supersonic flows of air associated with hypervelocity combustors. Characterization experiments with test gas mixtures of 95%N 2 + 5%CO 2 were performed to study the facility s performance such as the free-stream flow properties, the useful test time and the core-flow size available for mixing and combustion studies of a 2 mm jet in crossflow. Our experimental approach included simulation of the required total enthalpy (3-6 MJ/kgair) of flight Mach 8, 10 and 13 conditions by simulating the required burner entry Mach number, burner entry static temperature, and consequently the burner entry velocity. As the burner entry static temperatures were chosen to be in the range of K, velocities of 1800, 2360 and 3200 m/s were generated for the flight Mach 8, 10 and 13 conditions, respectively. On the other hand, simulated static pressures were below the desired values because of the limited maximum pressures available in the current driver section and the limited maximum pressures at which the jet injector valve could operate with sufficient speed. The steady test time duration and the core-flow size of the expansion tube flow 161

184 CHAPTER 8. CONCLUDING REMARKS 162 were characterized. These are important parameters which define the model dimensions where a fully established flow can be achieved. In addition, primary effects of the boundary layer on expansion tube flow were observed. In particular, test times ranging between µs, higher than the ideal values, were observed. In contrast with shock tunnels, the expansion tube allows an increased duration of useful test times, as the corresponding flight Mach number of the test flow increases. The boundary layer developed on the tube walls increases the contact surface velocity and therefore delays the arrival of the first disturbance wave. However, in our experiments the static pressure of the free-stream was low for high enthalpy flows, causing the boundary layer effects to increase the test time. Simulation with higher pressures might result in shorter test times, and therefore shorter useful test slug lengths would be available. We have demonstrated that Mirels solution for boundary layer effect implemented in x-t diagrams is a useful tool for prediction of test time and for optimizing the expansion section length to achieve the maximum test duration. Pitot pressure surveys at the exit of the expansion tube have identified an inviscid test core of approximately 25 mm diameter over which the pitot pressure is constant to within ±5%. While the core-flow size in Mach 10 and 13 conditions did not diminish significantly at 6.35 cm downstream of the tube exit, the core-flow size for the Mach 8 condition dropped by half. Since the Mach wave angle is steeper for small free-stream Mach numbers, the boundary layer information at the tube exit reaches to the centerline in a smaller distance at the flight Mach 8 condition. Finally, compared to large facilities that can average a limited number of experiments per day, many experiments per day in this facility can be performed with the effort of just one student. Therefore, this facility provides a useful tool for basic study of near-field features of different fuel injection configurations that have potential for future application in scramjet engines.

185 CHAPTER 8. CONCLUDING REMARKS Flow Visualization Techniques We used two non-intrusive flow diagnostic techniques: Planar Laser-Induced Fluorescence of OH radicals (OH-PLIF) and schlieren imaging with an ultra-fast-framing-rate digital camera. While schlieren showed the location of shock waves and jet penetration, OH-PLIF mapped the regions of combustion. Ultra-Fast Framing Rate Schlieren We have presented the first demonstration of an ultra-fast flow visualization system (at framing rates up to 100 MHz) based on schlieren imaging. High-temporal and high-spatial resolutions allowed both qualitative and quantitative study of supersonic flows. The system included a fast-framing camera (IMACON 468), capable of acquiring 8 full resolution images in a pixel format with interframing times and exposure times down to 10 ns and a Xenon flashlamp system capable of providing up to 200 µs duration of a uniform light source. Supersonic movies were obtained by assembling the consecutive images. These movies basically slow the flow motion by one million times, elucidating the instantaneous unsteady features. For example, the pulsating nature of periodically formed eddies causing the bow shock to fluctuate is very apparent and can be easily followed. Qualitative flow observations as well as quantitative measurements such as velocity, propagation angle and formation frequency of large-eddy structures, jet penetration and the width of the jet shear layer were obtained. A cross-correlation technique using fast Fourier transforms has been employed to analyze the 8 consecutive images. Among those measurements, the spatial-temporal development (x-t diagram) of unsteady structures and their frequency of formation were part of the unique measurements available only by analysis of time correlated multiple images. For example, due to pairing and stretching, the spatial gap between eddies is not necessarily a measure of the eddy formation frequency. Therefore, only high-speed framing rate imaging could provide such information, necessary for understanding the origin of the jet shear layer vortical structures, which are the dominant mechanism in the near-field mixing of jets.

186 CHAPTER 8. CONCLUDING REMARKS 164 We have shown optimum exposure and inter-framing times which require the optimization of four main factors including the schlieren sensitivity, the spatial resolution, the dynamic range and the signal-to-noise ratio. The image area was mm and the exposure time was 100 ns. In other words, the spatial resolution was µm in the near-field (x/d < 2) and µm in the far field (x/d > 8) depending on the large-scale structure movement during the exposure time. The corresponding resolving power was 2-5 pixels, achieved by the optimization of the four factors discussed above. Resolution considerations became important with increasing velocities and decreasing region of interest. In our experiments, we could achieve very high quality schlieren images even though light deflections were minimum since the jet diameter was only 2 mm. The intensified CCD cameras and the high intensity light source allowed us to control the sensitivity of the imaging system and achieve the required quality for quantitative data analyses. The application of a high-speed-framing rate imaging system became even more important as it was combined with an impulse facility operation. The unique free-stream conditions with high-speed and high-temperatures studied in this investigation could only be generated for very short times. Therefore, the use of the ultra-fast schlieren system was crucial because it increased significantly the amount of data available from a single experiment. OH-PLIF and Simultaneous Measurements The OH-PLIF measurements were obtained by excitation of the A 2 + X 2 Π(1, 0) band of OH, near 283 nm, and by the detection of the (1,1) band near 315 nm. A broader excitation assumption was valid as the linewidth of the laser beam, provided from a frequency-doubled dye laser source, was broader than the absorption linewidth. The isolated Q 1 (7) transition at nm was selected to minimize signal dependence on temperature. Therefore, the fluorescence intensity could be related directly to OH mole fraction. Simultaneous OH-PLIF and schlieren imaging could be implemented using two intensified CCD cameras and a dichroic mirror to separate the OH fluorescence from the schlieren beam.

187 CHAPTER 8. CONCLUDING REMARKS Characteristics of Hydrogen and Ethylene Transverse Jets We studied the flow-field properties of hydrogen and ethylene jets injected into flight Mach 10 conditions at similar jet-to-free-stream momentum flux ratio. The results reveal significant differences in the development of large-scale coherent structures present in the jet shear layer. Previously, the momentum flux ratio was found to be the main controlling parameter of the jet penetration; the results here demonstrated the existence of an additional mechanism which altered the vortical structure, the penetration and the mixing properties of the jet shear layer. These new observations became possible by the simulation of high velocity and high temperature free-stream conditions which could not have been achieved in the facilities that have been widely used in previous studies. The details of the main results can be summarized in the following points: Visual observations, supported by the qualitative measurements of the convection velocity and jet penetration, reveal large differences between the hydrogen and ethylene injection cases. Special attention was given to the large scale coherent structures present at the jet/free-stream interface. Instantaneous images provided a well-resolved representation of the coherent structures at the jet periphery. While the hydrogen eddies persisted for long downstream distances, in the ethylene case the eddies dissipated quickly. It is conjectured that increasing stresses due to the steep velocity gradient across the shear layer are responsible for this change. The large variation in the molecular weight between hydrogen and ethylene leads to significantly different exit velocities at the sonic orifice. Because of the low jet exit velocity of ethylene (315 m/s), the shear layer vortical structures tilt and stretch in the direction of the fast crossflow (2360 m/s). The large structures eventually became unstable and were torn apart by the stretching of the vortical structures. The above observations were supported by PLIF imaging of OH radicals which maps the regions of auto-ignition. These ignition regions can be related to homogenously mixed regions since molecular mixing is required before the fuel and the oxidizer react. Ethylene injection demonstrated high concentration of OH radicals across the jet while in the hydrogen case only a thin flamelet could be observed around the large eddy structures. Clearly, molecular mixing of the ethylene jet

188 CHAPTER 8. CONCLUDING REMARKS 166 was dramatically altered during or after the onset of the tilting-stretching-tearing process. Eddy convection characteristics and jet transverse penetration were also different between the two cases. Hydrogen structures tended to travel with velocities ( 2200 m/s) that were closer to the free-stream velocity as they align with the free-stream flow in the far-field (x/d > 9). The convection velocity of ethylene structures were slower than the hydrogen eddies due to the low jet exit velocities. Tracking different parts of the ethylene large eddies, a wide convection velocity distribution was shown to exist across the shear layer ranging between m/s. The differences in the bow shock steepness could result in the observed differences in the convection velocity between the cases. The properties of the free-stream (the bow shock shape and the shock-induced flow properties) were directly influenced by the convection characteristics of the large-scale eddies. The ethylene jet penetrates deeper into the free-stream than the hydrogen jet. This was an unexpected result as all of the previous studies showed that the jetto-free-stream momentum flux ratio (J) was the primary penetration controlling mechanism. We therefore expected to observe identical penetration heights as the J was identical for both cases in our studies. This interesting and surprising result could again be attributed to the evolution of the jet shear layer under large velocity gradients. The thickness of the penetration band, used as the representation of the jet-shear-layer thickness was considerable in the ethylene injection case, due to the tilting-stretching-tearing mechanism and also due to the larger growth rate of the jet shear layer Density and Velocity Ratio Effects Following the observations of the previous section, we investigated the stability of the jet shear layer at various speed ratios and density ratios via flow visualization (schlieren). The high shear stresses induced by the large velocity difference across the jet shear layer had a large effect on the structure of the layer. For the unstable case, we noticed: 1) loss of Kelvin-Helmholtz structures with the tilting-stretchingtearing mechanism, 2) increased growth rates with decreasing values of jet-tofree-stream velocity ratio, 3) large intrusions of crossflow in between the eddies,

189 CHAPTER 8. CONCLUDING REMARKS 167 4) distortion of the bow shock around the large eddies. Stable layers showed welldefined Kelvin-Helmholtz rollers. An effective velocity ratio parameter, λ = (1 r 2 )/ 1 + r 2 was suggested. The results plotted in a density-effective velocity ratio (s-λ) diagram demonstrated two separate regions of stable and unstable jet shear layers Ignition and Flame-Holding Capability of a Hydrogen Transverse Jet The problem of hydrogen transverse injection and its flame-holding capability was studied in very high-speed, high-total-enthalpy flow conditions. The experiments applied simultaneous OH-PLIF and schlieren imaging to map the regions where combustion occurs relative to the jet position. The main results are summarized as follows: At the flight Mach 10 condition, OH fluorescence was mainly observed along the outer edge of the jet plume. Simultaneous OH-PLIF/schlieren revealed that the structural evolution of the reaction zone is in good agreement with the jet shear layer position determined by the schlieren imaging. The ignition was initiated in the recirculation region upstream of the jet. The ignition product OH, convected downstream along with the large eddies, was mainly detected along the jet shear layer periphery in a continuous and very thin filament. At the flight Mach 13 condition, OH-PLIF demonstrated high signal levels of OH fluorescence starting in the upstream recirculation region and along the jet shear layer. The ignition delay times in this condition were effectively zero ( 1 5µs) due to the high total enthalpies (namely high total temperatures) of the freestream. At the low total enthalpy Mach 8 condition, the ignition was limited to a small region behind the bow shock and no OH fluorescence could be observed farther downstream Cavity Flame-Holders In this part of the thesis, we have first provided a review of cavities in supersonic flows and their use for flame-holding in supersonic combustors. Second, we have performed a

190 CHAPTER 8. CONCLUDING REMARKS 168 preliminary investigation, where the primary objective was to demonstrate the feasibility of the experimental set-up to provide information for cavity flameholder studies. These appear to be the first cavity experiments performed in such high total enthalpy flows. In the first part of the review, the basic flow-field features of cavities studied by various researchers are summarized, including: different flow regimes of cavities based upon the length-to-depth ratio (open and closed), oscillations, techniques to suppress these oscillations, drag penalties for different cavity geometries, and flow residence time inside a cavity which is crucial to initiate the ignition. Both experimental and numerical studies still need to be performed to answer some of the contradictory results that have been observed by different investigators (drag penalties of angled back wall cavities, amplitude of pressure fluctuations, flow residence time inside an unsteady cavity). In the second part of the review, studies demonstrating the feasibility of cavities to achieve ignition and to enhance flame-holding in subsonic and supersonic combustors, are described. Finally, we have introduced several questions followed by comments that need to be addressed in the development of cavities for practical combustors. Through a combination of simultaneously performed fast response pressure measurements, established cavity oscillations were observed and a sequence of oscillation cycles were captured during the limited test time ( 270 µs) of the flow facility. The results demonstrated that short duration pulse facilities can be used to study gasdynamic aspects of cavities, though with small dimensions (depth of D = 3 mm), in hypersonic flows. In the first part of the preliminary study, flows around 2-D cavities (D = 3 mm, L/D =3, 5 and 7) in a supersonic flow were investigated. Significant changes in the shock wave structure around the cavities were observed as the length-todepth ratio were systematically changed. Leading edge shock waves diminished in cavities with large L/D. In all cases a strong reattachment shock wave at the trailing edge of the cavity was observed. A transition from an open cavity flow to a closed cavity flow was obtained for L/D = 7. An angled back wall reduced the leading edge shock strength. Schlieren movies of the cavities reveal the shock

191 CHAPTER 8. CONCLUDING REMARKS 169 wave fluctuations around the jet are caused by the pressure oscillations inside the cavity. In the second part of the preliminary study, ignition properties of a 30 o hydrogen jet combined with a downstream cavity (L/D = 3) were investigated. While both cavities (with and without an angled back wall) provided an autoignition in and around the cavity, flame-holding seems to require an improved cavity design. Recommended configurations are presented in the following section. 8.2 Recommendation For Future Work Recommended future work can be summarized in the following topics: Extension of Ignition Measurements of Transverse Jets We have shown that a fundamental supersonic combustion study can be performed using an expansion tube, providing realistic free-stream conditions with relatively accurate chemical composition. We therefore recommend the study of the near-field ignition mechanism of transverse jets in more detail by taking advantage of the various freestream conditions. For example, this study can be performed using a set of free-stream conditions with reasonably similar velocities and total enthalpies but with significantly different static temperatures. Table 8.1 summarizes a set of recommended conditions which were actually characterized in the expansion tube. Maps of Fig. 2.7 guided us in determining the appropriate initial pressures. Observation of the ignition processes for decreased static temperatures, while keeping the other flow parameters constant, can reveal information about the source of the ignition and the parameters controlling the ignition process. Ignition in Cavity Flame-Holders The effects of the jet in crossflow in conjunction with a cavity can easily be studied in the current facility setup. Fig. 8.1 shows three possibilities to produce a reacting jet in crossflow which is stabilized by a cavity flame-holder. A preliminary examination of a 30 o angled hydrogen injection upstream of the cavity showed autoignition inside the cavity. However, a detailed investigation is still required to reveal the flame-holding

192 CHAPTER 8. CONCLUDING REMARKS 170 TABLE 8.1 Recommended free-stream flow conditions for further ignition studies. Condition U P T M H tot m/s atm K MJ/kg 600/0.5/ ± ± /1/ ± ± /2/ ± ± /3/ ± ± /4/ ± ± properties of the configuration. Injection in the upstream wall of the cavity might provide a larger volume of combustible mixture of fuel and air within the cavity flow. Injection further downstream might provide a pool of radicals both upstream, as well as downstream of the jet. It is however impossible to predict the interaction of jet/cavity shear layers in those configurations. A detailed experimental study is therefore needed for a better understanding of the coupling between the jet and the cavity flow. Finally, it should be noted that for hydrocarbon fuels with chemical kinetic rates notably slower than hydrogen, a cavity may be the only viable means for flame stabilization. Shock Wave/Jet Interaction Oblique shock wave impingement into the jet is one method known to enhance the molecular mixing between supersonic air and gaseous fuel. Waves are also unavoidable in SCRAMJETS and often originate in the inlet isolator leading to the combustion zone or from ramps and the bow-shock in front of the jet in crossflow. Thus, we propose to use wedge generated waves to interact with the jet injection flow-field. A schematic showing the expected features of shock impingement into a transverse jet is drawn in Fig It is well known that vorticity is generated when a shock wave interacts with a shear layer due to the baroclinic torque (in general, the pressure gradient caused by the shock and the density gradient in the shear layer will not be colinear). Amplified turbulence and vorticity have immediate significance to the mixing enhancement in supersonic flows. Furthermore, the shock directs the airflow towards the fuel jet, increasing the air entrainment rate, the static pressure, the static temperature, and therefore the reaction

193 CHAPTER 8. CONCLUDING REMARKS 171 (a) Jet (b) Injectant Free Shear Layer Jet Free Shear Layer (c) Shear Layer - Jet Interaction Jet FIGURE 8.1 Flow-field schematics demonstrating different concepts of angled jet injection combined with cavity flame-holder. a) upstream injection, b) base injection, c) cavity injection. rates. There are few works (Hwanil and Driscoll 1996; Menon 1989; Marble 1994) which have studied the effect of shock-wave/shear-layer interaction on mixing enhancement in a supersonic combustor. The results indicated that the spreading rate of the shearlayer may be enhanced by the shock impingement, resulting in enhanced mixing and combustion efficiency. Shock wave jet interaction can easily be studied using the current facility/injection system. A 25 o angled wedge is actually designed to generate an oblique shock wave as illustrated in Figure 8.3. The wedge, positioned 22 mm above the injection plate, has a window made of sapphire to pass the OH-PLIF laser sheet into the imaged region. A typical schlieren image demonstrating the oblique shock wave jet interaction is shown in Fig The relatively strong shock, generated by the 25 o wedge, directs the air flow inward toward the hydrogen jet. In a different experiment in which simultaneous

194 CHAPTER 8. CONCLUDING REMARKS 172 SHOCK GENERATOR PLATE M >1 OBLIQUE SHOCK BOW SHOCK SHOCK-ENHANCED COMBUSTION BARREL SHOCK MACH DISK BOUNDARYLAYER SEPARATED REGION RECIRCULATION ZONE INJECTANT (Hydrogen) RECIRCULATION ZONE FIGURE 8.2 Flow-field schematic of a shock-wave/jet interaction. FIGURE 8.3 Schematic of the 25 o wedge to generate a shock wave above the injection plate.

195 CHAPTER 8. CONCLUDING REMARKS 173 OH-PLIF and schlieren was applied (see Fig. 8.4b), no change in the OH signal level was observed behind the shock impingement; meanwhile the thin filament of OH was directed along the jet contour.

196 CHAPTER 8. CONCLUDING REMARKS 174 PLIF Sheet 25 o Slit for OH-PLIF Laser Sheet M = 4.7 V = 3200 m/s P = 4 kpa T = 1250 K Air Jet: Hydrogen, d jet = 2 mm, M jet =1, J=3 y/d jet Region Illuminated by PLIF Sheet y/d jet FIGURE 8.4 (a) An oblique shock wave impinging the hydrogen jet as visualized using schlieren imaging. The shock was produced by a 25 o angled wedge mounted above the injection plate. Flight Mach 13 free-stream condition. (b) Combined OH-PLIF and schlieren images visualizing the effect of shock/jet interaction on OH number density.