In this study, we analyse the statistics of both individual inertial particles and inertial particle pairs in direct numerical simulations of homogeneous isotropic turbulence in the absence of gravity. The effect of the Taylor microscale Reynolds number, $R_{{\it\lambda}}$, on the particle statistics is examined over the largest range to date (from $R_{{\it\lambda}}=88$ to 597), at small, intermediate and large Kolmogorov-scale Stokes numbers $St$. We first explore the effect of preferential sampling on the single-particle statistics and find that low-$St$ inertial particles are ejected from both vortex tubes and vortex sheets (the latter becoming increasingly prevalent at higher Reynolds numbers) and preferentially accumulate in regions of irrotational dissipation. We use this understanding of preferential sampling to provide a physical explanation for many of the trends in the particle velocity gradients, kinetic energies and accelerations at low $St$, which are well represented by the model of Chun et al. (J. Fluid Mech., vol. 536, 2005, pp. 219–251). As $St$ increases, inertial filtering effects become more important, causing the particle kinetic energies and accelerations to decrease. The effect of inertial filtering on the particle kinetic energies and accelerations diminishes with increasing Reynolds number and is well captured by the models of Abrahamson (Chem. Engng Sci., vol. 30, 1975, pp. 1371–1379) and Zaichik & Alipchenkov (Intl J. Multiphase Flow, vol. 34 (9), 2008, pp. 865–868), respectively. We then consider particle-pair statistics, and focus our attention on the relative velocities and radial distribution functions (RDFs) of the particles, with the aim of understanding the underlying physical mechanisms contributing to particle collisions. The relative velocity statistics indicate that preferential sampling effects are important for $St\lesssim 0.1$ and that path-history/non-local effects become increasingly important for $St\gtrsim 0.2$. While higher-order relative velocity statistics are influenced by the increased intermittency of the turbulence at high Reynolds numbers, the lower-order relative velocity statistics are only weakly sensitive to changes in Reynolds number at low $St$. The Reynolds-number trends in these quantities at intermediate and large $St$ are explained based on the influence of the available flow scales on the path-history and inertial filtering effects. We find that the RDFs peak near $St$ of order unity, that they exhibit power-law scaling for low and intermediate $St$ and that they are largely independent of Reynolds number for low and intermediate $St$. We use the model of Zaichik & Alipchenkov (New J. Phys., vol. 11, 2009, 103018) to explain the physical mechanisms responsible for these trends, and find that this model is able to capture the quantitative behaviour of the RDFs extremely well when direct numerical simulation data for the structure functions are specified, in agreement with Bragg & Collins (New J. Phys., vol. 16, 2014a, 055013). We also observe that at large $St$, changes in the RDF are related to changes in the scaling exponents of the relative velocity variances. The particle collision kernel closely matches that computed by Rosa et al. (New J. Phys., vol. 15, 2013, 045032) and is found to be largely insensitive to the flow Reynolds number. This suggests that relatively low-Reynolds-number simulations may be able to capture much of the relevant physics of droplet collisions and growth in the adiabatic cores of atmospheric clouds.

The behaviour of passive-scalar fields resulting from mean scalar gradients in each of three orthogonal directions in homogeneous turbulent shear flow has been studied using direct numerical simulations of the unsteady incompressible Navier-Stokes equations with 128 × 128 × 128 grid points. It is found that, for all orientations of the mean scalar gradient, the sum of the pressure-scalar gradient and velocity gradient-scalar gradient terms in the turbulent scalar flux balance equation are approximately aligned with the scalar flux vector itself. In addition, the time derivative of the scalar flux is also approximately aligned with the flux vector for the developed fields (corresponding to roughly constant correlation coefficients). These alignments lead directly to a gradient transport model with a tensor turbulent diffusivity. The simulation results are used to fit a dimensionless model coefficient as a function of the turbulence Reynolds and Péclet numbers. The model is tested against two different passive-scalar fields in fully developed turbulent channel flow (also generated by direct numerical simulation) and is found to predict the scalar flux quite well throughout the entire channel.

The theory of the shear-induced small deformation of viscous drops at zero Reynolds number is reviewed. The general result for arbitrary shear and surface tension is presented, and the asymptotic forms for weak flow and for high internal viscosity are derived from it. In the latter case, numerical solutions are compared with the experiments of Torza, Cox & Mason (1972).

The velocity correlation function of a Brownian particle in a viscous compressible fluid is studied in the limit of very small compressibility. The main effect of compressibility is an initial rapid change of the particle mass from a real inertial mass to a virtual mass.

Water waves that propagate over a non-rigid bed are attenuated as a result of energy dissipation within the bed. This paper describes the analysis, from the basis of small-amplitude wave theory, of the coupled interaction between the bed, which responds in both an elastic and viscous manner, and an overlying layer of inviscid fluid. A dispersion relation is derived, from which rates of wave attenuation and sea bed deflexions are calculated.

The results are relevant wherever the sea bed, in its response to water waves, is non-rigid, as is often the case in coastal waters. Depending on the elasticity and viscosity of the sea bed, the wave attenuation can be of the same or of a larger order of magnitude than that due to bottom friction or percolation in a permeable bed. Where waves propagate over a soft viscous bed, for example as is the case at certain mud flats off the south-west coast of India, exceptionally high rates of attenuation are possible whereby waves are almost completely damped within several wavelengths.

A mathematical model is developed to study the behaviour of an oscillating wave energy converter in a channel. During recent laboratory tests in a wave tank, peaks in the hydrodynamic actions on the converter occurred at certain frequencies of the incident waves. This resonant mechanism is known to be generated by the transverse sloshing modes of the channel. Here the influence of the channel sloshing modes on the performance of the device is further investigated. Within the framework of a linear inviscid potential-flow theory, application of Green’s theorem yields a hypersingular integral equation for the velocity potential in the fluid domain. The solution is found in terms of a fast-converging series of Chebyshev polynomials of the second kind. The physical behaviour of the system is then analysed, showing sensitivity of the resonant sloshing modes to the geometry of the device, which concurs in increasing the maximum efficiency. Analytical results are validated with available numerical and experimental data.

Planar flow in the interfacial region of an open porous medium is investigated by finding solutions for Stokes flow in a channel partially filled with an array of circular cylinders beside one wall. The cylinders are in a square array oriented across the flow and are widely spaced, so that the solid volume fraction ϕ is 0.1 or less. For this spacing, singularity methods are appropriate and so they are used to find solutions for both planar Couette flow and Poiseuille flow in the open portion of the channel. The solutions, accurate to O(ϕ), are used to calculate the apparent slip velocity at the interface, Us, and results obtained for Us are presented in terms of a dimensionless slip velocity. For shear-driven flow, this dimensionless quantity is found to depend only weakly on ϕ and to be independent of the height of the array relative to the height of the channel and independent of the cylinder size relative to the height of the channel. For pressure-driven flow, Us is found to be less than that under comparable shear-flow conditions, and dependent on cylinder size and filling fraction in this case. Calculations also show that the external flow penetrates the porous medium very little, even for sparse arrays, and that Us is about one quarter of the velocity predicted by the Brinkman model.

Experiments were conducted in a wind tunnel in which a turbulent boundary layer was naturally grown over flat beds of three types of nearly mono-disperse spherical particles with different diameters, densities and coefficient of restitution (r) (snow, 0.48 mm, 910 kg m−3; mustard seeds, 1.82 mm, 1670 kg m−3, r = 0.7; ice particles, 2.80 mm, 910 kg m−3, r = 0.8–0.9). The surface wind speeds (defined by the friction velocity u∗) were varied between 1.0 and 1.9 times the threshold surface wind speed (defined by u∗t). The trajectories, and ejection and impact velocities of the particles were recorded and analysed, even those that were raised only about one diameter into the flow.

Measurements of the average horizontal flux of saltating particles per unit area, f(z), at each level z above the surface showed that, for u∗/u∗t [les ] 1.5, f(z) is approximately independent of the particle density and decreases exponentially over a vertical scale length lf, that is about 3 to 4 times the estimated mean height of the particle trajectories 〈h〉. Numerical simulations of saltating grains were computed using the measured probabilities of ejection velocities and the mean velocity profile of the air flow, but neglecting the direct effect of the turbulence. The calculated mean values of the impact velocities and the trajectory dimensions were found to agree with the measurements in the saltation range, where u∗/u∗t < 1.5. Similarly, in this range the simulations of the horizontal flux profile and integral are also consistent with the measurements and with Bagnold's u∗3 formula, respectively.

When u∗/u∗t [ges ] 1.5, and u∗/VT [ges ] 1/10, where VT is the settling velocity, a transition from saltation to suspension occurs. This is indicated by the change in the mean mass flux profile which effectively becomes uniform with height (z) up to the top of the boundary layer. An explanation is provided for this low value of turbulence at transition relative to the settling velocity in terms of the random motion of the particles under the action of the turbulence when they reach the tops of their parabolic trajectories. The experiments show that, as u∗/u∗t increases from 1.0 to 1.9 the normalized mean vertical impact velocity 〈V3I〉/u∗ decreases by nearly 60% to about 0.6, which is less than 50% of the value for fluid particles. There is also a decrease in the vertical and horizontal component of the ejection velocity to values of 0.8 and 2.3, which are much less than their values in the saltation regime. We hypothesize that at the transition from saltation to suspension the ejection process changes quite sharply from being determined by impact collisions to being the result of aerodynamic lift forces and upward eddy motions.

The fingering instabilities in vertical miscible displacement flows in porous media driven by both viscosity and density contrasts are studied using linear stability analysis and direct numerical simulations. The conditions under which vertical flows are different from horizontal flows are derived. A linear stability analysis of a sharp interface gives an expression for the critical velocity that determines the stability of the flow. It is shown that the critical velocity does not remain constant but changes as the two fluids disperse into each other. In a diffused profile, the flow can develop a potentially stable region followed downstream by a potentially unstable region or vice versa depending on the flow velocity, viscosity and density profiles, leading to the potential for ‘reverse’ fingering. As the flow evolves into the nonlinear regime, the strength and location of the stable region changes, which adds to the complexity and richness of finger propagation. The flow is numerically simulated using a Hartley-transform-based spectral method to study the nonlinear evolution of the instabilities. The simulations are validated by comparing to experiments. Miscible displacements with linear density and exponential viscosity dependencies on concentration are simulated to study the effects of stable zones on finger propagation. The growth rates of the mixing zone are parametrically obtained for various injection velocities and viscosity ratios.

A numerical method is developed for solution of the full nonlinear equations governing irrotational flow with a free surface and variable bed topography. It is applied to the unsteady motion of non-breaking water waves of arbitrary magnitude over a horizontal bed. All horizontal variation is approximated by truncated Fourier series. This and finite-difference representation of the time variation are the only necessary approximations. Although the method loses accuracy if the waves become sharp-crested at any stage, when applied to non-breaking waves the method is capable of high accuracy.

The interaction of one solitary wave overtaking another was studied using the Fourier method. Results support experimental evidence for the applicability of the Korteweg-de Vries equation to this problem since the waves during interaction are long and low. However, some deviations from the theoretical predictions were observed - the overtaking high wave grew significantly at the expense of the low wave, and the predicted phase shift was found to be only roughly described by theory. A mechanism is suggested for all such solitary-wave interactions during which the high and fast rear wave passes fluid forward to the front wave, exchanging identities while the two waves have only partly coalesced; this explains the observed forward phase shift of the high wave.

For solitary waves travelling in opposite directions, the interaction is quite different in that the amplitude of motion during interaction is large. A number of such interactions were studied using the Fourier method, and the waves after interaction were also found to be significantly modified - they were not steady waves of translation. There was a change of wave height and propagation speed, shown by the present results to be proportional to the cube of the initial wave height but not contained in third-order theoretical results. When the interaction is interpreted as a solitary wave being reflected by a wall, third-order theory is shown to provide excellent results for the maximum run-up at the wall, but to be in error in the phase change of the wave after reflection. In fact, it is shown that the spatial phase change depends strongly on the place at which it is measured because the reflected wave travels with a different speed. In view of this, it is suggested that the apparent time phase shift at the wall is the least-ambiguous measure of the change.

Bubble dynamics near a rigid boundary are associated with wide and important applications in cavitation erosion in many industrial systems and medical ultrasonics. This classical problem is revisited with the following two developments. Firstly, computational studies on the problem have commonly been based on an incompressible fluid model, but the compressible effects are essential in this phenomenon. Consequently, a bubble usually undergoes significantly damped oscillation in practice. In this paper this phenomenon will be modelled using weakly compressible theory and a modified boundary integral method for an axisymmetric configuration, which predicts the damped oscillation. Secondly, the computational studies so far have largely been concerned with the first cycle of oscillation. However, a bubble usually oscillates for a few cycles before it breaks into much smaller ones. Cavitation erosion may be associated with the recollapse phase when the bubble is initiated more than the maximum bubble radius away from the boundary. Both the first and second cycles of oscillation will be modelled. The toroidal bubble formed towards the end of the collapse phase is modelled using a vortex ring model. The repeated topological changes of the bubble are traced from a singly connected to a doubly connected form, and vice versa. This model considers the energy loss due to shock waves emitted at minimum bubble volumes during the beginning of the expansion phase and around the end of the collapse phase. It predicts damped oscillations, where both the maximum bubble radius and the oscillation period reduce significantly from the first to second cycles of oscillation. The damping of the bubble oscillation is alleviated by the existence of the rigid boundary and reduces with the standoff distance between them. Our computations correlate well with the experimental data (Philipp & Lauterborn, J. Fluid Mech., vol. 361, 1998, pp. 75–116) for both the first and second cycles of oscillation. We have successively reproduced the bubble ring in direct contact with the rigid boundary at the end of the second collapse phase, a phenomenon that was suggested to be one of the major causes of cavitation erosion by experimental studies.

To locate noise sources in high-speed jets, the far-field sound pressure fluctuations $p^\prime $ were correlated with each of density $\rho $, axial velocity $u$, radial velocity $v$, $\rho uu$ and $\rho vv$ fluctuations measured from various points in jet plumes. Detailed surveys were conducted in fully expanded, unheated plumes of Mach 0.95, 1.4 and 1.8. The velocity and density fluctuations were measured simultaneously using a recently developed non-intrusive point measurement technique based on molecular Rayleigh scattering. The technique uses a continuous-wave narrow line-width laser, Fabry–Perot interferometer and photon counting electronics. Laser light scattered by air molecules from a 1.06 mm long region on the narrow beam was collected and spectrally resolved by the interferometer. It was observed that the fluctuation spectra for air density inside the plume were in general similar to those of axial velocity spectra, while the radial velocity spectra were somewhat different. For the correlation study, microphone polar angles were varied from 30$^\circ$ to 90$^\circ$ to the jet axis. The sound pressure fluctuations $p^\prime $ at the shallowest 30$^\circ$ angle provided the highest correlation with turbulent fluctuations. The correlations sharply decreased as the polar angle was increased to 60$^\circ$, beyond which all data mostly fell below the experimental noise floor. Among all turbulent fluctuations $\langle\rho uu; p^\prime\rangle$ correlations showed the highest values. Correlation with the first-order terms $\langle\rho ^\prime \bar{u} \bar{u}; p^\prime \rangle$, $\langle\skew3\bar\rho \bar{u}u^\prime; p^\prime\rangle$ and third-order terms $\langle\rho ^\prime u^\prime u^\prime ;p^\prime\rangle$ was higher than that from the second-order terms $\langle\skew3\bar\rho u^\prime u^\prime ;p^\prime \rangle$ and $\langle\bar u\rho ^\prime u^\prime ; p^\prime \rangle$. Both $\langle v^\prime ; p^\prime \rangle$ and $\langle\rho vv; p^\prime \rangle$ correlations with the 90$^\circ$ microphone signal fell below the experimental noise floor, while that from the shallow 30$^\circ$ microphone showed weaker values. By moving the laser probe to various locations in the jet, it was found that the strongest noise source lay downstream of the end of the potential core and extended many diameters beyond. Correlation measurements from turbulent fluctuations along the lip shear layer showed a Mach-number dependency: significant values were measured in supersonic jets, while correlations fell below the noise floor for subsonic jets. Various additional analyses showed that fluctuations from large coherent structures mostly contributed to the measured correlation, while that from small-scale structures fell below the noise floor.

The Korteweg-de Vries (KdV) equation and the finite-depth equation of Joseph (1977) and Kubota, Ko & Dobbs (1978) both describe the evolution of long internal waves of small but finite amplitude, propagating in one direction. In this paper, both theories are tested experimentally by comparing measured and theoretical soliton shapes. The KdV equation predicts the shapes of our measured solitons with remarkable accuracy, much better than does the finite-depth equation. When carried to second-order, the finite-depth theory becomes about as accurate as (first-order) KdV theory for our experiments. However, second-order corrections to the finite-depth theory also identify a bound on the range of validity of that entire expansion. This range turns out to be rather small; it includes only about half of the experiments reported by Koop & Butler (1981).

In this study, the onset of cellular structure on oblique detonation surfaces is investigated numerically using a one-step irreversible Arrhenius reaction kinetic model. Two types of oblique detonations are observed from the simulations. One is weakly unstable characterized by the existence of a planar surface, and the other is strongly unstable characterized by the immediate formation of the cellular structure. It is found that a high degree of overdrive suppresses the formation of cellular structures as confirmed by the results of many previous studies. However, the present investigation demonstrates that cellular structures also appear with degree of overdrive of 2.06 and 2.37, values much higher than ${\sim }$1.8 suggested previously in the literature for the critical value defining the instability boundary of oblique detonations. This contradiction could be explained by the use of differently shaped walls, a straight wall used in this study and a custom-designed curved wedge system so as to induce straight oblique detonations in previous studies. Another possible reason could be due to the low and possibly insufficient resolution used in previously published studies. Hence, simulations with different grid sizes are also performed to examine the effect of resolution on the numerical solutions. Using the present results, analysis also shows that although the characteristic lengths of unstable surfaces are different when the incident Mach number changes, these length scales are proportional to tangential velocities. Hence, the interior time determined by the overdrive degree is identified, and its limitation as the instability parameter is discussed.

The three-dimensional incompressible flow past a rectangular two-dimensional shallow cavity in a channel is investigated using large-eddy simulation (LES). The aspect ratio (length/depth) of the cavity is $L/D\,{=}\,2$ and the Reynolds number defined with the cavity depth and the mean velocity in the upstream channel is 3360. The sensitivity of the flow around the cavity to the characteristics of the upstream flow is studied by considering two extreme cases: a developing laminar boundary layer upstream of the cavity and when the upstream flow is fully turbulent. The two simulations are compared in terms of the mean statistics and temporal physics of the flow, including the dynamics of the coherent structures in the region surrounding the cavity. For the laminar inflow case it is found that the flow becomes unstable but remains laminar as it is convected over the cavity. Due to the three-dimensional flow instabilities and the interaction of the jet-like flow inside the recirculation region with the separated shear layer, the spanwise vortices that are shed regularly from the leading cavity edge are disturbed in the spanwise direction and, as they approach the trailing-edge corner, break into an array of hairpin-like vortices that is convected downstream the cavity close to the channel bottom. In the fully turbulent inflow case in which the momentum thickness of the incoming boundary layer is much larger compared to the laminar inflow case, the jittering of the shear layer on top of the cavity by the incoming near-wall coherent structures strongly influences the formation and convection of the eddies inside the separated shear layer. The mass exchange between the cavity and the main channel is investigated by considering the ejection of a passive scalar that is introduced instantaneously inside the cavity. As expected, it is found that the ejection is faster when the incoming flow is turbulent due to the interaction between the turbulent eddies convected from upstream of the cavity with the separated shear layer and also to the increased diffusion induced by the broader range of scales that populate the cavity. In the turbulent case it is shown that the eddies convected from upstream of the cavity can play an important role in accelerating the extraction of high-concentration fluid from inside the cavity. For both laminar and turbulent inflow cases it is shown that the scalar ejection can be described using simple dead-zone theory models in which a single-valued global mass exchange coefficient can be used to describe the scalar mass decay inside cavity over the whole ejection process.

Stingrays, in contrast with many other aquatic animals, have flattened disk-shaped bodies with expanded pectoral ‘wings’, which are used for locomotion in water. To discover the key features of stingray locomotion, large-eddy simulations of a self-propelled stingray, modelled closely after the freshwater stingray, Potamotrygon orbignyi, are performed. The stingray’s body motion was prescribed based on three-dimensional experimental measurement of wing and body kinematics in live stingrays at two different swimming speeds of 1.5 and $2.5L~\text{s}^{-1}$ ($L$ is the disk length of the stingray). The swimming speeds predicted by the self-propelled simulations were within 12 % of the nominal swimming speeds in the experiments. It was found that the fast-swimming stingray (Reynolds number $Re=23\,000$ and Strouhal number $St=0.27$) is approximately 12 % more efficient than the slow-swimming one ($Re=13\,500$, $St=0.34$). This is related to the wake of the fast- and slow-swimming stingrays, which was visualized along with the pressure on the stingray’s body. A horseshoe vortex was discovered to be present at the anterior margin of the stingray, creating a low-pressure region that enhances thrust for both fast and slow swimming speeds. Furthermore, it was found that a leading-edge vortex (LEV) on the pectoral disk of swimming stingrays generates a low-pressure region in the fast-swimming stingray, whereas the low- and high-pressure regions in the slow-swimming one are in the back half of the wing and not close to any vortical structures. The undulatory motion creates thrust by accelerating the adjacent fluid (the added-mass mechanism), which is maximized in the back of the wing because of higher undulations and velocities in the back. However, the thrust enhancement by the LEV occurs in the front portion of the wing. By computing the forces on the front half and the back half of the wing, it was found that the contribution of the back half of the wing to thrust in a slow-swimming stingray is several-fold higher than in the fast-swimming one. This indicates that the LEV enhances thrust in fast-swimming stingrays and improves the efficiency of swimming.

The present study consists of wind profile surveys, drift current measurements and water surface observations for a wide range of wind velocities in a wind–wave tank. It is confirmed that the velocity distribution essentially follows the logarithmic law near the water surface and the velocity-defect law toward the outer edge of the boundary layer. The wind stresses and surface roughnesses calculated from these distributions are divided into two groups separated by the occurrence of the wave-breaking phenomenon. For low wind velocities the surface roughness is dictated by ripples, and the wind-stress coefficient varies with U0−½, where U0 is the free-stream wind velocity. The surface roughness is proportional to the average height of the basic gravity wave at higher wind velocities; the stress coefficient is then proportional to U0. In addition, it is found that Charnock's expression (k ∝ u*2/g) holds only at high wind velocities, and that the constant of proportionality determined from the present experiment correlates very well with field observations. A new technique, involving the use of various-sized surface floats to determine the drift current gradient and the surface drift current, has been developed. A good agreement is shown between the gradients obtained from the measured currents and those determined from the wind stresses. Finally, the wind-stress coefficient is shown to be larger than the friction coefficient for turbulent flow along a solid rough surface; the difference is shown to be the wave drag of the wind over the water surface.

The stability of oscillatory channel flow to different classes of infinitesimal and finite-amplitude two- and three-dimensional disturbances has been investigated by direct numerical simulations of the Navier–Stokes equations using spectral techniques. All infinitesimal disturbances were found to decay monotonically to a periodic steady state, in agreement with earlier Floquet theory calculations. However, before reaching this periodic steady state an infinitesimal disturbance introduced in the boundary layer was seen to experience transient growth in accordance with the predictions of quasi-steady theories for the least stable eigenmodes of the Orr–Sommerfeld equation for instantaneous ‘frozen’ profiles. The reason why this growth is not sustained in the periodic steady state is explained. Two-dimensional infinitesimal disturbances reaching finite amplitudes were found to saturate in an ordered state of two-dimensional quasi-equilibrium waves that decayed on viscous timescales. No finite-amplitude equilibrium waves were found in our cursory study. The secondary instability of these two-dimensional finite-amplitude quasi-equilibrium states to infinitesimal three-dimensional perturbations predicts transitional Reynolds numbers and turbulent flow structures in agreement with experiments.

The effects of polymer stresses on near-wall turbulent structures are examined by using direct numerical simulation of fully developed turbulent channel flows with and without polymer stress. The Reynolds number based on friction velocity and half-channel height is 395, and the stresses created by adding polymer are modelled by a finite extensible nonlinear elastic, dumbbell model. Both low- (18%) and high-drag reduction (61%) cases are investigated. Linear stochastic estimation is employed to compute the conditional averages of the near-wall eddies. The conditionally averaged flow fields for Reynolds-stress-maximizing Q2 events show that the near-wall vortical structures are weakened and elongated in the streamwise direction by polymer stresses in a manner similar to that found by Stone et al. (2004) for low-Reynolds-number quasi-streamwise vortices (‘exact coherent states: ECS’). The conditionally averaged fields for the events with large contribution to the polymer work are also examined. The vortical structures in drag-reduced turbulence are very similar to those for the Q2 events, i.e. counter-rotating streamwise vortices near the wall and hairpin vortices above the buffer layer. The three-dimensional distributions of conditionally averaged polymer force around these vortical structures show that the polymer force components oppose the vortical motion. More fundamentally, the torques due to polymer stress are shown to oppose the rotation of the vortices, thereby accounting for their weakening. The observations also extend concepts of the vortex retardation by viscoelastic counter-torques to the heads of hairpins above the buffer layer, and offer an explanation of the mechanism of drag reduction in the outer region of wall turbulence, as well as in the buffer layer.

Measurements of a separating two-dimensional incompressible boundary layer with an airfoil-type pressure distribution are reported. Unique mean and fluctuation velocity measurements and the distribution of the fraction of the time γp during which the flow moves downstream were obtained in the separated region using a directionally sensitive laser anemometer. Linearized hot-film anemometer measurements of mean velocities, turbulent shearing stress and intensities, eddy speeds, spectra and dissipation were made for γp > 0·8. The wall shearing stress, bursting frequencies, wall speed and spanwise structure were obtained using flush-surface hot-film sensors. The turbulent/non-turbulent interfacial intermittency γ and the frequency of passage of turbulent bulges were determined using smoke as a turbulence marker and the laser anemometer system for illumination and signal detection.

Upstream of separation the velocity profile correlations of Perry & Schofield (1973) are supported within the uncertainty of the data. Normal-stress effects are very important, influencing $-\overline{uv}/\overline{q^2} $ and the dissipation length correlations, and directly providing sizable terms in the momentum and turbulence energy equations. The criteria of Sandborn for turbulent separation and fully developed separation are found to hold. Downstream of separation there is apparent similarity of $\overline{u^2}$, U and γp throughout the shear flow. The passive low velocity backflow near the wall apparently just serves to satisfy continuity requirements after the energetic outer-region flow has deflected away from the wall upon separation.

The wall bursting frequency nA scales on outer velocity and length scales, with U∞/δnA ≈ 10, or about twice the value observed for zero-pressure-gradient flows. The non-dimensional spanwise spacing of wall eddies is given approximately by the relation λzUM/v ≈ 100 upstream of separation, where $U_M = (- \overline{uv}_{\max})^{\frac{1}{2}}$. The speed of wall eddies is found to be about 14Uτ.