Planar imaging measurements to study the effect of spanwise structure of upstream turbulent boundary layer on shock induced separation

44th AIAA Aerospace Sciences Meeting and Exhibit 9-2 January 26, Reno, Nevada AIAA 26-324 Planar imaging measurements to study the effect of spanwise structure of upstream turbulent boundary layer on shock induced separation B. Ganapathisubramani, N. T. Clemens and D. S. Dolling Center for Aeromechanics Research, University of Texas at Austin, TX 7872 The relationship between the upstream boundary layer and the unsteadiness of the separated flow in a Mach 2 compression ramp interaction is investigated by performing wide-field PIV and PLS measurements in streamwise-spanwise planes. Presence of spanwise strips of elongated regions of uniform momentum is detected in the upstream boundary layer. These long coherent structures exhibit strong similarities to those that have been found in incompressible boundary layers, which suggests an underlying similarity between the incompressible and supersonic regimes. At a wall normal-location of y/δ =.2, the upstream envelope of the separation region is found to oscillate between = -3 and (where = is the ramp corner). The instantaneous separation line (in the spanwise direction) is found to respond to the elongated regions of uniform momentum. The separation region exhibits an undulated envelope which conforms to the low- and high-speed regions in the upstream boundary layer. The low frequency unsteadiness of the separation region/shock foot observed in numerous previous studies can be reconciled by a turbulent mechanism that includes these elongated regions of uniform momentum. Nomenclature x Streamwise direction z Spanwise direction y Wall-normal direction U Streamwise velocity (m/s) W Spanwise velocity (m/s) V Wall-normal velocity (m/s) u Streamwise velocity fluctuation (m/s) w Spanwise velocity fluctuation (m/s) U Streamwise freestream velocity (m/s) u τ skin friction velocity M Mach number δ Boundary layer thickness θ Momentum thickness δ Displacement thickness U m Mean streamwise velocity at any given location in the boundary layer σ u r.m.s streamwise velocity at any given location in the boundary layer ω y Wall-normal vorticity T o Stagnation temperature T Static temperature C p Specific heat at constant pressure I. Introduction Shock wave/boundary layer interactions (SWBLI) remain an active research topic owing to the problems they present to the design of supersonic aircraft, missiles and projectiles. Sufficiently strong interactions cause severely turbulent and highly unsteady separation of the boundary layer. 3 This turbulent separation can cause rapid mechanical fatigue of aero-structures owing to the presence of large fluctuating pressure loads and high heat transfer rates. Postdoctoral fellow, Center for Aeromechanics Research, University of Texas at Austin. Member Professor, Department of Aerospace Engineering and Engineering Mechanics, University of Texas at Austin. Associate fellow Professor, Department of Aerospace Engineering and Engineering Mechanics, University of Texas at Austin. Fellow Copyright 26 by B. Ganapathisubramani, N. T. Clemens American and D.S. Dolling. Institute Published of Aeronautics by the American andinstitute Astronautics of Aeronautics and Astronautics, Inc., with permission. of 6

Figure. Schematic of the observed relationship between upstream boundary layer and the location and scale of the separation region. Reproduced from Beresh et al. 4 The unsteadiness of SWBLIs includes a wide range of frequencies and scales. However, previous studies 5 indicate that the low frequencies are more dominant and have posed the biggest challenge to theoretical and computational models. Typically, the lowest frequency of the motion of shock foot is at least an order of magnitude lower than the nominal boundary layer frequency based on boundary layer thickness (δ). This low-frequency motion of the shock foot is a consuming aspect of shock-induced separation due to the difficulty in associating such low frequencies with typical turbulent mechanisms. The low frequency motion of the shock foot seems to be related to the low-frequency motion of the separated flow. 6 Recent studies have focused on investigating the physics responsible for the unsteadiness of the separated flow at frequencies that are an order of magnitude smaller than the typical boundary layer frequency (δ/u ). In this quest, investigation of the upstream turbulent boundary layer is an important step. There is growing evidence in the literature that the unsteady nature of the separation shock wave (in interactions by various bluff body geometries and compression ramps) and in turn the unsteady oscillation of the separation bubble upstream of the bluff body is related to the structure of the upstream turbulent boundary layer. However, the exact nature of this dependence is not known. There is emerging evidence that the low-frequency unsteadiness is driven by low-frequency turbulent fluctuations in the upstream turbulent boundary layer. Recent particle image velocimetry (PIV) studies 4, indicate that the motion 4, 7 9 of the shock foot responds to a large-scale thickening and thinning of the upstream boundary layer. It was found that the shock foot moves upstream/downstream if the lower part of the upstream boundary layer is slower/faster and overall it is locally thick/thin. Consequently, the separation region extends farther upstream (the scale is larger) if the upstream boundary layer is slower and the separation region is smaller if the boundary layer is faster as seen in figure. To date it has been difficult to associate the low frequency motion of the shock to any particular turbulent mechanism, since typical turbulent structures are believed to be of scale δ (boundary layer thickness). It would require structures with scales that are an order of magnitude larger than δ to reconcile the low frequencies observed. Studies in incompressible turbulent boundary layers by various researchers using PIV have shown the existence of long coherent regions of streamwise velocity fluctuations in the log region 3 (extending to 2 of 6

lengths greater than 2δ). Other studies using hotwire measurements seem to indicate that these long lowand high speed regions extend as long as 2δ. 4 Recently, Ganapathisubramani et al 5 performed wide-field PIV measurements in streamwise-spanwise planes in the log region of a Mach 2 supersonic boundary layer (the same facility utilized in the present study, without the compression ramp). The authors identified the presence of long- low and high-speed regions that extended to lengths greater than 8δ and speculated that these low frequency events could explain the low frequency unsteadiness of the shock foot/separation region. This study can be considered a continuation of the work by Ganapathisubramani et al 5 as we aim to investigate the relationship between the upstream boundary layer and the unsteadiness of the separated flow scale in a Mach 2 compression ramp interaction. The primary diagnostics used are wide-field particle image velocimetry (PIV) and cinematographic planar laser scattering measurements (PLS) in streamwise-spanwise planes. The objective of the current work is to apply planar imaging techniques to study the relationship between these long, low-frequency, large-scale uniform momentum regions and the unsteady motion of the shock-induced turbulent separation region. II. Experimental facility and Details Figure 2. Experimental setup (a) Perspective view and (b) Top view. A. Wind tunnel facility All of the experiments were conducted in the Mach 2 blowdown tunnel of the University of Texas at Austin. The constant-area test section was 6 in (5.2 cm) wide by 6.3 in (6 cm) high and had a length of 3 in (76.2 cm). Removable side doors allowed access to an instrumented floor section. The side walls of the test section were equipped with fused silica windows which are 6 in (5.2 cm) long by 2 in (5. cm) high by.75 in (.9 cm) thick to facilitate to facilitate the passage of laser sheets used as a source of illumination in PIV. The top wall of the tunnel is fitted with an acrylic window for optical access of size 8 in (2.3 cm) long, 2.5 in (6.4 cm) and.75 in (.9 cm) thick. A total of about 4 m 3 of compressed air was provided by a Worthington HB4 four-stage compressor and stored in external tanks at a pressure of about 25 psia. The stagnation chamber pressure and temperature for the present experiments were approximately 26 ± 7 kpa and 292 ± 5 K, respectively. For these stagnation conditions, stable run times of up to 4 seconds could be obtained. The freestream conditions were as follows: Mach number M = 2 and velocity U = 5 m/s. The incoming turbulent boundary layer underwent natural transition and developed under approximately adiabatic wall temperature conditions. The boundary layer properties were as follows: boundary layer thickness δ = 2.5 mm, momentum thickness θ =.9 mm, displacement thickness δ* = 2.6 mm, skin friction velocity u τ = 8.5 m/s, Reynolds number based on momentum thickness, Re θ = 35,. Finally, the shock boundary layer interaction was generated using a 2 compression ramp that spanned the entire width of tunnel. B. Particle image velocimetry (PIV) The wide field PIV system used pulsed laser sheets from a dual-cavity flashlamp pumped Nd:YAG laser (Spectra Physics PIV 4) separated in time by 2 µs and directed through one side window and oriented parallel to the tunnel floor as shown in figure 2. The laser pulse energy was 4 mj, and the thickness of the sheets was about.7 mm. Sets of digital images were captured by three Kodak Megaplus CCD cameras 3 of 6

(24 24 pixels) lined up next to each other in the streamwise direction as shown in figure 2 to provide a wide field of view with an area of approximately 38 38 mm 2 ( 9δ 3δ). Nikon MicroNikkor 5mm f/2.4 lenses were used with all three cameras. The system was synchronized such that each camera simultaneously imaged the laser sheets. Three pulse-delay generators (Stanford Research Systems DG535s) were used to synchronize the PIV system and to trigger the individual components. Data were acquired at two wallnormal locations, one in the logarithmic region (y/δ =.2) and the other in the outer region (y/δ =.7). A total of 75 images were acquired at each station for statistical convergence of mean and r.m.s statistics of the velocity components. Titanium dioxide (TiO 2 ) with a manufacturer-specified nominal particle diameter of.2µm was used as seed particles for PIV. The particles were seeded upstream of the stagnation chamber of the tunnel by using a two-stage fluidized-bed seeder followed by a cyclone separator. The particle seeders were driven by compressed nitrogen. Previous studies have shown that this seeding system produces excellent seeding densities for PIV. Hou 6 discusses in detail the measurement of the response time of the PIV seed particles used in this study. Time response measurement of the particles though a normal shock indicated that the response time τ p was about 2.6µs. From the response time, the effective (agglomerated) diameter was estimated to be d p.26 µm. Samimy & Lele 7 suggest that for particles to faithfully track the velocity fluctuations in a turbulent mixing layer, the Stokes number defined as St = τ p /τ f (with τ f = δ/u, the characteristic flow time scale) must be less than about.5. The Stokes number based on the outer time scale of boundary layer (δ/u =26µs) is about.. This clearly shows that the particles easily track the large scale velocity fluctuations. The PIV images were processed using TSI Insight 6. software which recursively refined the interrogation window from 28 28 pixels down to a size of 32 32 pixels. The linear resolution of the final interrogation window is.28 mm. A 5% overlap was used to provide a vector field of size 8 6 vectors. The average streamwise pixel displacement in the upstream boundary layer was about 9 pixels. A standard 3 3 neighborhood median filter with a tolerance of 5 pixels was utilized to remove erroneous vectors. Any missing vectors were interpolated using a 3 3 local mean technique. The number of spurious vectors was less than 2% in the dataset. Figure 3. Schematic of the compression ramp interaction. C. Planar laser scattering (PLS) The flow was also visualized with high repetition rate planar laser scattering where the scattering medium was condensed acetone fog. The setup for PLS imaging was similar to that used for the PIV measurements except that a single laser and camera were used. A diode-pumped Nd:YLF laser (Coherent Evolution-9, output wavelength of 527 nm) at a repetition rate of khz was used as the light source. This laser was optimized to produce approximately 9 W of power which results in about 9 mj per pulse at khz. The fog was generated by seeding the flow with acetone droplets by using two fine spray atomizing nozzles 6m upstream of the stagnation chamber. The droplets evaporate before reaching the nozzle and then re-condense into fine fog owing to the isentropic cooling through the nozzle. The light scattered by the 4 of 6

acetone fog was imaged with one high-framing rate CMOS camera (Photron FASTCAM-Ultima APX) with a Nikon MicroNikkor 5mm lens with an aperture setting of f/2.8. The camera resolution was 52 256 pixels (streamwise spanwise), and the field of view was 76 34 mm 2. Further details on the PLS system can be found in Bueno et al. 8 Planar laser scattering data was acquired at two wall-normal locations y/δ =.2 and.7 (same locations as the PIV data). III. Results and Discussion In all results presented in this section, x, y and z are streamwise, spanwise and wall-normal directions respectively. The coordinates are normalized by the 99% boundary layer thickness (δ). The origin in the streamwise direction ( = ) is located at the corner of the ramp as shown in figure 3. The origin in the spanwise direction ( = ) is along the centerline of the test section. A. Planar laser scattering results Figure 4. Planar laser scattering image at y/δ =.2. The flow is from left to right. A PLS image reveals the concentration of acetone droplets in the flow. High and low signal intensity regions correspond to regions of high and low acetone droplet concentration respectively. The droplets tend to evaporate in regions with relatively high static temperature (T ) and so such regions exhibit low signal levels. If we assume a constant stagnation temperature (T o ) and use the energy equation (C p T o = C p T +.5U 2, where C p is specific heat at constant pressure, T is the local static temperature and U is the local streamwise velocity), then the regions of high static temperature are also regions of low velocity. Therefore, regions of low concentration of acetone fog tend to be associated with regions of low velocity. Similarly, regions of high concentration of acetone fog are regions of high velocity. We caution that this relationship between PLS signal and velocity is only approximate because the finite times for evaporation/condensation will obviously affect this interpretation. Figure 4 shows a sample PLS image at a wall-normal location of y/δ =.2. The separation region behind the shock is a region of low momentum and therefore the signal levels are low. The upstream boundary layer reveals a remarkable spanwise organization of low and high intensity regions. This indicates the presence of spanwise strips of low and high momentum regions. The presence of long streamwise structures has been previously observed with the instantaneous visualization experiments in the upper part of the boundary layer (y/δ =.49 and.65). 9 This signature is similar to the velocity fluctuation signature observed in low Reynolds number incompressible boundary layers, 2 and in Mach 2 supersonic boundary layer. 5 Interestingly, the separation region seems to be farther upstream where the upstream boundary layer is slow. Conversely, the separation region is pushed downstream if the upstream boundary layer is fast. This 5 of 6

observation is consistent with the other studies in the literature. 4 This observation is further discussed in section B where quantitative PIV data reinforces this observation. Figures 5 and 6 show a time-sequence of images at y/δ =.2 &.7 respectively. The time between each image in the sequence is µs during which the flow structures travel across about half the image length. At the wall-normal location of y/δ =.2 the images clearly reveal streamwise strips of regions of high and low signal intensity. Also, the separation region seems to extend to a distance of 2δ upstream of the ramp foot. However at y/δ =.7, the streamwise strips are no longer clear. This could be due to the fact that the velocity fluctuations at this height are much smaller than at y/δ =.2. Therefore the variations in static temperatures are smaller and consequently so are the differences in the acetone fog concentrations. Readers must note that PLS image sequences can reveal important information on the interaction dynamics, when viewed as a movie. This is particularly the case for low-frequency events that cannot be represented in sequences of images. Therefore some of the observations that are discussed in this paper were made from viewing the movies and therefore may not be apparent in the image sequences. The time separation between successive images in figures 5 and 6 is µs. The turbulent structures nominally travel a distance of about 45 mm (6/ th of the frame) in this time (assuming a convection velocity of.9u ). Hence, the structures can be tracked from one frame to the next, but the displacements are too large for the motion to appear continuous. In contrast, the separation region (identified as the large region of low signal) exhibits clear unsteadiness and does not move appreciably in the µs between frames (this point is clear only on watching the movies). In fact, the movies are very useful for demonstrating the wide range of time-scales that characterize the interaction. In particular, when the movies are played at a reasonably high framing rate (> fps), the envelope of the separation region is seen to respond to the presence of low- and high-speed regions in the upstream boundary layer. At lower framing rates, the separation region does not seem to be correlated to the upstream boundary layer. This is evidence that the upstream boundary layer structures that drive the large-scale motion of the separation region are characterized by frequencies that are less than khz. This conclusion is consistent with previous studies performed with cylinder interactions in a Mach 2 flow. 8 Such low frequencies can be reconciled with a turbulent mechanism analogous to the very-large scale motion model proposed by Kim & Adrian 2 for incompressible boundary layers. The authors proposed that very long meandering regions of low- and high-speed regions can be interpreted as groups of hairpin packets 2, 2 traveling together with a common convection velocity. Other interpretations for the presence of this low frequency oscillations include the presence of Görtler vortices, 8 which may be wind-tunnel dependent. This discussion is further elaborated in section B where quantitative PIV data seem to point to the presence of signatures similar to very-large scale motions in the upstream boundary layer. The low-frequency motion of the separation region led to computing locally averaged (low-pass filtered) PLS images. The image sequences were low-pass filtered by locally averaging each image with its ten nearest neighbors (in time). This filter will have the effect of blurring the high-frequency motions and leaving structures that occur at frequencies of less than khz. Figure 7 shows a ten-frame sequence at y/δ =.2, where the effective time between images is ms. It is seen from this sequence that the PLS images are indeed smoothed, but the low frequency component of the upstream boundary layer, namely the large streamwise strips of low- and high-intensity regions survive the averaging. For example, figure 7a still reveals long strips of low- and high-intensity regions indicating that these low- and high-speed regions are indeed very long in the streamwise direction (at least 3δ, since the structures move about 3δ between consecutive images and successive images were averaged). Also, the separation region exhibits an undulated envelope which conforms to the low- and high speed regions in the upstream boundary layer. B. Wide field particle image velocimetry results. instantaneous results Sample PIV vector and contour plots are shown in figure 8. The flow structures observed in the samples are representative of those found in many realizations. The coordinates are normalized by the boundary layer thickness (δ). The origin in the streamwise direction is located at the ramp corner as shown in figure 3. Figures 8a and 8b show the streamwise-spanwise velocity vector fields at y/δ =.2 &.7. The two figures reveal the extent of the separation region at both wall-normal locations. Figure 8a indicates that the separation region extends upstream up to a distance of -2δ (can be seen from the reverse flow as marked in the figure). However, the separation region is not obvious at y/δ =.7 (figure 8b), which indicates that the 6 of 6

Figure 5. A time-sequence of planar laser scattering images at y/δ =.2. Subsequent images are separated by µs. Flow is left to right. 7 of 6

Figure 6. A time-sequence of planar laser scattering images at y/δ =.7. Subsequent images are separated by µs. Flow is left to right. 8 of 6

Figure 7. A sequence of locally averaged planar laser scattering images at y/δ =.2. averaged images are separated by ms. Flow is left to right. 9 of 6 Subsequent locally

(b) (a) Separation region -9-8 -7-6 -5-4 -3-2 -9-8 -7-6 -5-2 47 476 482 488 494 5 56 52 58 524 53 35 36 37 38 39 4 4 42 43 44 45-9 -8-7 -6-5 -4-3 -2-9 -8-7 -6-5 -4-3 -2 (f) (e) -5 5 75 2 225 25 275 3 325 35 375 4 425 45 475 5 525 5 2 25 3 35 4 45 5-3 (d) (c) -9-8 -7-6 -5-4 -3-2 -9-8 -7-6 -5-4 -3-2 (g) (h) -5-4 -3-2 2 3 4-3 -24 8 2-6 5 6 2 8 24 3-4 -9-8 -7-6 -5-4 -3-2 -9-8 -7-6 -5-4 -3-2 Figure 8. Instantaneous velocity vectors and contours. Velocity vectors at (a) y/δ =.2 and (b) y/δ =.7. Streamwise velocity (U ) with a color scale that emphasizes the upstream boundary layer (c) y/δ =.2 and (d) y/δ =.7. Streamwise velocity with a color scale that emphasizes the separation region (e) y/δ =.2 and (d) y/δ =.7. Spanwise velocity (W ) contours at (g) y/δ =.2 and (f ) y/δ =.7. Note that figures (a), (c), (e) and (g) are from the same instantaneous velocity field at y/δ =.2. Similarly, figures (b), (d), (f ) and (h) are from the same instantaneous field at y/δ =.7. of 6

plane of measurement (laser sheet) is above the separated flow and is probably in the shear layer above the separation bubble (see schematic in figure 3). Figure 8c and 8d show the streamwise velocity contours (corresponding to the vectors in figure 8a and 8b) with a color scale that emphasizes the structure of the upstream boundary layer. The streamwise velocity (U) plot shown in figure 8c reveals the existence of strips of uniform high- and low-speed regions in the upstream boundary layer. These strips exhibit characteristic widths of approximately.25.5δ and they extend a large distance in the streamwise direction. The streamwise extent of these coherent regions is at least 7δ. Similar experiments without the ramp in the flow revealed that these long regions extend beyond the full field of view (i.e., > 8δ). 5 The strips also exhibit a spanwise gradient in U that is relatively large as the velocity changes by about m/s (.2U ) over a distance of quarter of a boundary layer thickness (i.e., 3 mm). It should be emphasized that these long structures are randomly distributed in space (i.e., not stationary) as they disappear when several vector fields are averaged (can be seen in figure ). Figure 8c also shows the envelope (upstream extent) of the separation region and the influence of the upstream boundary layer on the structure of the separation region. The envelope is undulated and conforms to the presence of high- and low-speed regions in the upstream flow. The separation region seem to extend far upstream (at the bottom of the figure) when the upstream boundary layer has a long low-speed region. Similarly, the separation is pushed downstream (middle of the figure), when there is high-speed region in the boundary layer. These findings are consistent with the PLS results discussed in section A. Figure 8d reveals a plot of U at y =.7δ. The overall structure at this location is similar to the lower wall-normal location, however, the spanwise scales are larger. Also, the range of velocities are much smaller than the range at y/δ =.2. This is understandable since the location is practically at the edge of the boundary layer. The velocity scale of 47 m/s to 53 m/s was chosen to reveal the structure of the upstream boundary layer and therefore the downstream edge gives a false impression that the separation region extends upstream to about -2δ. In reality the lowest velocity in the whole field was 8 m/s, which is much higher than the lowest velocity at y/δ =.2 (where reverse flow could be identified). Figures 8e and 8f show contours of streamwise velocity (depicted in figures 8c and 8d) with an alternate color scale that emphasizes the separation region. These figures are included to clarify the above mentioned point that the separation region is only apparent at y/δ =.2 where reverse flow (with velocity -5 m/s) is observed. Due to the large range in velocities, it is necessary to view these fields with two different color scales in order to deduce any relationship between the boundary layer and the separation region. Figure 8e together with figure 8c reveals a correlation between the presence of low momentum region in the upstream boundary layer and existence of reverse flow in the separation region at y/δ =.2. At y/δ =.7, only a gradual decrease in the velocity is noted. Since the measurement plane at this location is above the separation bubble (as shown in figure 3), the velocity fields do not reveal any correlation between the upstream boundary layer and the shear layer region. Figure 8g is a plot of the spanwise velocity (W ) at y/δ =.2. The velocities are in the range ±5 m/s. This plot reveals that spanwise velocity is far less coherent along both the streamwise and spanwise directions. The figure shows that the spatial scale of W contours are mostly short, compact and seems to appear in spanwise patches of alternating positive and negative velocity. Figure 8h shows a plot of W at y/δ =.7. This plot indicates that W is more compact than U, but the W velocity signatures are longer and wider farther from the wall. This suggests an increasing trend in the representative length scale of the spanwise velocity in both streamwise and spanwise directions. The presence of long streamwise structures in the upstream boundary layer is in agreement with the instantaneous visualization experiments where the upper part of the boundary layer (y/δ =.49 and.65) was seen to be populated with elongated longitudinal structures. 9 The long streamwise structures are also consistent with the results of Ünalmis & Dolling8 who made measurements of fluctuating wall and Pitot pressures in a Mach 5 turbulent boundary layer that developed on the wind tunnel wall. They computed cross-correlations of the pressure data and concluded that the boundary layer exhibits streamwise vortical structures. They interpreted these structures as likely resulting from Görtler type vortices generated in the upstream nozzle. It is not known at this time whether the structures observed in figure 8 are remnants of a Görtler instability, but some evidence suggests they may not be. For example, it might be expected that Görtler vortices would exhibit vorticity that is primarily streamwise, in which case the contours of W should also exhibit the presence of elongated structures with alternating positive or negative spanwise velocity. The patterns observed in figure 8 seem to be more consistent with the velocity fields from instantaneous PIV measurements made in incompressible boundary layers., 2 The hairpin packet model 2 was used by of 6

those studies to explain the existence of instantaneous uniform momentum zones. The region between the legs of the hairpin vortices possesses negative velocity fluctuations and the zones on either side of the legs of the hairpin vortices have positive velocity fluctuations (This is analogous to ejections between the legs and sweeps on either side of the legs). Adrian et al 2 hypothesized that hairpin packets can contain up to individual vortices propagating as a coherent entity that extends to streamwise distances greater than 2δ. However, hotwire based studies showed that these uniform momentum regions extended to much larger 4, 2 streamwise distances (> 5δ). Therefore, Kim & Adrian 2 proposed a model for these very large scale motions (VLSM) based on the hairpin packet model. In this model hairpin packets group together to form a larger-scale structure (VLSM). Kim & Adrian 2 also noted that the very large scale motion (VLSM) model is a hypothesis that avoids asserting that the observed large-scale coherence constitutes a new type of turbulent motion and simply uses the existing hairpin packet model as a basic building block. Since it is hypothesized in this model that a group of hairpin packets convect together, the streamwise velocity fluctuations within such a group would have similar magnitudes and can be observed as zones of uniform low- and high-speed regions in an instantaneous velocity field. The above mentioned signature is remarkably similar to the patterns found in figure 8c. Therefore, the very large-scale motions model could be used to understand the instantaneous flow patterns observed in the upstream boundary layer. (a) -5-4 -3-2 2 3 4 5 (b) -4-32 -24 6-8 8 6 24 32 4-9 -8-7 -6-5 -4-3 -2-9 -8-7 -6-5 -4-3 -2 Figure 9. (a) Streamwise velocity fluctuations normalized by upstream boundary layer skin friction velocity (u/u τ ) (b) Wall-normal vorticity normalized by δ and u τ (ω y δ/u τ ). Figures (a) and (b) are taken from the same instantaneous field depicted in figures 8(a), (c), (e) and (g). Figure 9 shows the effect of the ramp shock on the turbulent structures present in the boundary layer. Figure 9a reveals streamwise velocity fluctuations u normalized by upstream boundary layer skin friction velocity u τ at y/δ =.2. The figure clearly, shows that the fluctuations are amplified downstream of x = 2.2δ. The fluctuations downstream of the shock (in the separation region) are at least 2-3 times higher than in the upstream boundary layer. This phenomenon of amplification of turbulent fluctuations behind the shock in the boundary layer has been observed in both numerical and experimental studies., 22 24 Figure 9b shows contours of wall-normal vorticity normalized by boundary layer thickness and skin friction velocity (ω y δ/u τ ). The contours indicate amplification in vorticity behind the shock (in the separation region). The readers must be aware that the grid resolution of the present study is not suitable for computing statistics of vorticity. However, it can be used to highlight a general point that the shock acts as an amplifier of turbulent structures. 2. Statistical results Figure a and b show the mean streamwise velocity at y/δ =.2 &.7 respectively. The mean and r.m.s statistics of the streamwise velocities in the upstream boundary layer are comparable to various other studies in the literature. 3, 6, 25 27 Figure a shows a decrease in streamwise velocity (relative to the upstream boundary layer) starting at = -2.2 at y/δ =.2. This decrease in streamwise velocity at y/δ =.2 can be interpreted as the effect of the separation region and its extent in the streamwise direction. At y/δ =.7 (figure b), the decrease in streamwise velocity is observed at =.4. However the lowest streamwise velocity at this higher wall-normal location is much greater than the lowest velocity at y/δ =.2. This is likely due to the fact that the plane of measurement is above the separation bubble and is in the shear layer above it (see figure 3). 2 of 6

(a) 45 9 35 8 225 27 35 36 45 45 (b) 45 9 35 8 225 27 35 36 45 45-9 -8-7 -6-5 -4-3 -2-9 -8-7 -6-5 -4-3 -2 (c) 2 3 4 5 6 7 8 9 2 (d) 2 3 4 5 6 7 8 9 2-9 -8-7 -6-5 -4-3 -2-9 -8-7 -6-5 -4-3 -2 Figure. Mean and r.m.s statistics. Mean streamwise velocity at (a) y/δ =.2 and (b) y/δ =.7. Root-meansquare velocity at (c) y/δ =.2 and (d) y/δ =.7. location (y/δ) U m (m/s) σ u (m/s) Condition (U < U m -4σ u ).2 397 35 U < 257.7 52 24 Separation point does not exist at this height Table. Mean and r.m.s statistics of the upstream boundary layer and the conditions employed to identify the separation region Figure c and d show the r.m.s fluctuations of streamwise velocity at y/δ =.2 and.7. At the wall-normal location of y/δ =.2, the r.m.s values in the region between = -2 and are 4-5 times higher than the r.m.s in the upstream boundary layer. This region of elevated r.m.s corresponds to the intermittent region of the spanwise separation line where the unsteadiness of the separation region leads to higher fluctuations. The increased r.m.s values can also be influenced by the amplification of turbulent fluctuations by the shock. At y/δ =.7 (figure d), the r.m.s values steadily increase downstream of =.8 to a maximum of about 3-4 times higher than that of the upstream boundary layer. The steady increase in the r.m.s can be attributed to the shear layer that is present above the separation bubble. To further quantify the location of the separation region a threshold based technique is utilized to identify the upstream extent of the separation point. This algorithm is applied to the dataset obtained at y/δ =.2. The data at y/δ =.7 is not used in this analysis, since it is clear from various instantaneous velocity fields that the separation region does not extend (vertically) to this height. The measurement plane at y/δ =.7 was located above the separation bubble as shown in figure 3. The location of separation point was identified as the streamwise point at which the instantaneous streamwise velocity is less than U m -4σ u (where U m and σ u are the mean and r.m.s streamwise velocities of the upstream boundary layer). This streamwise point is identified for every spanwise location. Table lists the mean and r.m.s velocities at the two wall-normal locations and the condition that defines the separation point. Figure a shows the probability distribution of the streamwise location of the separation point at y/δ =.2. The most likely location of the separation point at y/δ =.2 is found to be at =.8. Figure a indicates that the range of motion of the separation point at y/δ =.2 varies from = -3 to =. This is consistent with the intermittent region (range of motion of the shock foot) as inferred from wall-pressure measurements in previous studies in the same flow. 6 Hou 6 identified the intermittent region in a Mach 2 3 of 6

(a).5 (b) 475 p.d.f.25.75.5.25 End of field of view y/δ =.2 Streamwise velocity (U m/s) 45 425 4 375 35 U m = 397 m/s 5 5 2 25 3 35 4 45 5-3 -2.5-2.5 -.5 Separation location () 325-3 -2.75-2.5-2.25-2.75.5.25 -.75 Separation location () Figure. (a) p.d.f of the location of the separation point at y/δ =.2. Joint p.d.f between the location of the separation point and the mean streamwise velocity in the upstream boundary layer along the line of the separation point at y/δ =.2. ramp interaction (similar to the present setup) using wall-pressure measurements and showed that the range of motion of the shock foot was between = -3.5 and.5. Since the measurement plane in the current study is at y/δ =.2, it expected that the separation region would be nominally farther downstream of the shock foot. The range of motion of the shock foot was found to be approximately 2δ 6 ( 3.5 < <.5). Figure a shows that the range of motion of the separation region is also about 2δ ( 3 < < ). This consistency in the range of motion of the shock foot and the separation region reinforces the conclusion by Erengil & Dolling 6 that the unsteadiness of the shock foot is related to the unsteadiness of the separation region (maybe controlled by the same mechanism). Joint probability densities between the location of the separation region and the streamwise velocity in the upstream boundary layer were computed to understand the role of upstream boundary layer in the oscillation of the instantaneous separation point. The mean streamwise velocity along a streamwise line upstream of an identified separation point is computed for every spanwise location. The joint p.d.f. between this mean streamwise velocity and the streamwise location of the separation point is computed. Figure b shows the joint p.d.f. between the local mean streamwise velocity and the streamwise location of the separation point at y/δ =.2. The separation point location is nominally at =.8 when U in the upstream boundary layer is close to U m (mean upstream boundary layer velocity). This suggests that this location is the neutral position for the separation region at this height. The contours are elliptical and the major axis of the ellipse is seen to be inclined. This indicates that when U < U m in the upstream boundary layer, the separation point is located upstream; whereas, if U > U m, then the separation point is pushed downstream. The velocity ranges between 35 m/s (.7U, when the separation point is upstream) to 45 m/s (.8U, separation point downstream) resulting in a net fluctuation ±5 m/s (.U ) between the two extreme positions. This range in velocity fluctuations is comparable to previous studies 6 where the upstream boundary layer was found to be slower by about 3-4 m/s when the shock foot was located at its most extreme upstream position; similarly the boundary layer was faster by the same amount when the shock was located downstream. The results from the probability distributions and the instantaneous PIV fields in liaison with previous studies in the literature 4, 6, 6 indicate that the long low- and high-speed regions in the upstream boundary layer are responsible for the low frequency unsteadiness of the shock foot and the separation region. 4 of 6

IV. Conclusions Wide field particle image velocimetry and cinematographic planar laser scattering measurements were performed in the streamwise-spanwise planes of a shock-wave turbulent boundary layer interaction generated by a 2 ramp in a Mach 2 flow. PLS imaging of the upstream boundary layer revealed elongated regions of low- and high-speed region that extended to a maximum of 3δ (length computed based on Taylor s hypothesis). In addition, the separation region was found to respond to these low- and high speed regions. The upstream/downstream presence of the separation location was found to be correlated with the presence of low-speed/high-speed region in the upstream boundary layer. Wide field PIV measurements of the interaction were also made using three cameras capturing a large part of the upstream boundary layer and a considerable portion of the separation region. Measurements were made at two wall-normal locations (y/δ =.2 &.7). Instantaneous velocity fields show the presence of long regions of uniform momentum (both high- and low-momentum) regions in the upstream boundary layer. The velocity signature is remarkably similar to the signature observed in incompressible boundary layers. The uniform momentum regions are analogous to the very-large scale motions model proposed by Kim & Adrian. 2 The instantaneous spanwise separation line is seen to be undulated, conforming to the presence highand low-speed regions in the upstream boundary layer. Amplification of turbulent fluctuations and vorticity is observed behind the ramp shock (in the separation region). Statistical analysis utilizing joint p.d.fs and observations of various instantaneous velocity fields indicate that the location of separation is pushed back towards the ramp foot if the upstream boundary layer is faster. Conversely, the separation region is far upstream if the boundary layer is slower. Therefore, based on the instantaneous PLS and PIV fields and the probability density distributions, it can concluded that the low frequency unsteadiness of the separation region can be explained by the presence of long strips of low- and high-speed regions in the upstream boundary layer. Acknowledgements The authors wish to thank Pablo Bueno, Edward J. Zihlman Jr, Justin Wagner and Jeff Searcy for all the help in the data acquisition phase of the study. The authors gratefully acknowledge the support of the Air Force Office of Scientific Research under grant FA955-4-387. References Andreopoulos, Y., Agui, J. H., and Briassulis, G., Shock WaveTurbulence Interactions, Annu. Rev. Fluid Mech., Vol. 32, 2, pp. 39 345. 2 Dolling, D. S., Fifty years of shock-wave/boundary-layer interaction research: what next? AIAA Journal, Vol. 39-8, 2, pp. 57 53. 3 Smits, A. J. and Dussauge, J. P., Turbulent shear layers in supersonic flow, American Institute of Physics Press, 996. 4 Beresh, S. J., Clemens, N. T., and Dolling, D. S., Relationship Between Upstream Turbulent Boundary-Layer Velocity Fluctuations and Separation Shock Unsteadiness, AIAA Journal, Vol. 4(2), 22, pp. 242 2422. 5 Dolling, D. S., High-speed turbulent separated flows: Consistency of mathematical models and flow physics, AIAA Journal, Vol. 36-5, 998, pp. 725 732. 6 Erengil, M. E. and Dolling, D. S., Physical causes of separation shock unsteadiness in shock wave/turbulent boundary layer interactions, AIAA paper # 993-334, 993. 7 Andreopoulos, J. and Muck, K. C., Some new aspects of the shock-wave boundary layer interaction, J. Fluid Mech., Vol. 8, 987, pp. 45 428. 8 Ünalmis, O. H. and Dolling, D. S., Experimental study of causes of unsteadiness of shock induced turbulent separation, AIAA Journal, Vol. 36-3, 998, pp. 37 378. 9 Hou, Y. X., Ünalmis, O. H., Bueno, P. C., Clemens, N. T., and Dolling, D. S., Effects of Boundary-Layer Fluctuations on Unsteadiness of Blunt-Fin Interactions, AIAA Journal, Vol. 422, 24, pp. 265 269. Hou, Y. X., Clemens, N. T., and Dolling, D. S., Wide-Field PIV Study of Shock-Induced Turbulent Boundary Layer Separation, AIAA paper # 23-44, 23. Tomkins, C. D. and Adrian, R. J., Spanwise structure and scale growth in turbulent boundary layers, J. Fluid Mech., Vol. 49, 23, pp. 37 74. 2 Ganapathisubramani, B., Longmire, E. K., and Marusic, I., Characteristics of vortex packets in turbulent boundary layers, J. Fluid Mech., Vol. 478, 23, pp. 35 46. 3 Ganapathisubramani, B., Hutchins, N., Hambleton, W. T., Longmire, E. K., and Marusic, I., Investigation of large-scale coherence in a turbulent boundary layer using two-point correlations, J. Fluid Mech., Vol. 524, 25, pp. 57 8. 5 of 6

4 Hutchins, N., Ganapathisubramani, B., and Marusic, I., Dominant spanwise fourier modes and existence of very large scale coherence in turbulent boundary layers, 5 th Australasian fluid mechanics conference, December 37, Sydney, Australia, 24. 5 Ganapathisubramani, B., Clemens, N. T., and Dolling, D. S., Large scale motions in a supersonic turbulent boundary layer, J. Fluid Mech., Vol. to appear, 26. 6 Hou, Y. X., Particle Image Velocimetry Study of Shock Induced Turbulent Boundary Layer Separation, Ph.D. thesis, Department of Aerospace Engineering and Engineering Mechanics, The University of Texas at Austin, Texas, USA, 23. 7 Samimy, M. and Lele, S. K., Motion of Particles with Inertia in a Compressible Free Shear Layer, Phy. Fluids A, Vol. 3-8, 99, pp. 95 923. 8 Bueno, P. C., Ganapathisubramani, B., Clemens, N. T., and Dolling, D. S., Cinematographic Planar Imaging of a Mach 2 Shock Wave / Turbulent Boundary Layer Interaction, AIAA paper # 25-44, 25. 9 Samimy, M., Arnette, S. A., and Elliot, G. S., Streamwise structures in a supersonic turbulent boundary layer, Phys. Fluids, Vol. 6(3), 994, pp. 8 83. 2 Kim, K. C. and Adrian, R. J., Very large-scale motion in the outer layer, Phys. Fluids, Vol. (2), 999, pp. 47 422. 2 Adrian, R. J., Meinhart, C. D., and Tomkins, C. D., Vortex organization in the outer region of the turbulent boundary layer, J. Fluid Mech., Vol. 422, 2, pp. 53. 22 Agui, J. H., Shock wave interactions with turbulence and vortices, Ph.D. thesis, City University, New York, 998. 23 Lee, L., Lele, S. K., and Moin, P., Direct numerical simulation of isotropic turbulence interacting with a weak shock wave, J. Fluid Mech., Vol. 25, 993, pp. 533 562. 24 Loginov, M., Adams, N., and Zheltovodov, A., Large-eddy simulation of shock-wave/turbulent boundary layer interaction, 2 st ICTAM, August 5 2, Warsaw, Poland, 24. 25 Elena, M. and Lacharme, J. P., Experimental study of a supersonic turbulent boundary layer using a laser doppler anemometer, J. Theoretical & Applied Mechanics, Vol. 7(2), 988, pp. 75 9. 26 Spina, E. F., Organized structures in a supersonic turbulent boundary layer, Ph.D. thesis, Department of Mechanical and Aerospace Engineering, Princeton University, NJ, USA, 988. 27 Kistler, A. L., Fluctuation measurements in supersonic turbulent boundary layer, Phys. Fluids, Vol. 2(3), 959, pp. 29 296. 6 of 6