Cavitation experiments performed in the near field of a 50 mm diameter (D) jet at ReD = 5 × 105, showed inception in the form of inclined ‘cylindrical’ bubbles at axial distances (x/D) less than 0.55, with indices of 2.5. On tripping the boundary layer, cavitation inception occurred at x/D ≈ 2, as distorted ‘spherical’ bubbles with inception indices of 1.7. To investigate these substantial differences, the near field of the jet was measured using PIV. Data on the primary flow, the strength distribution of the ‘streamwise’ vortices and the velocity profiles within the initial boundary layers were obtained. The untripped case showed a direct transition to three-dimensional flow in the near field (x/D < 0.7) even before rolling up to distinct vortex rings. Strong ‘streamwise’ vortices with strengths up to 25% of the jet velocity times the characteristic wavelength were seen. Cavitation inception occurred in the core of these vortices. In contrast, in the tripped jet the vortex sheet rolled up to the familiar Kelvin–Helmholtz vortex rings with weak secondary vortices. Using the measured nuclei distribution, strengths and straining of the ‘streamwise’ structures, the rates of cavitation events were estimated. The estimated results match very well the measured cavitation rates. Also, the Reynolds stresses in the near field of the jet show similar trends and magnitudes to those of Browand & Latigo (1979) and Bell & Mehta (1990) for a plane shear layer.

From shallow-water gravity wave theories it is shown that the velocity field in the whole fluid domain can be reconstructed using an analytic transformation (a renormalization). The resulting velocity field satisfies the Laplace equation exactly, which is not the case for shallow-water approximations. Applying the renormalization to the first-order shallow-water solution of limited accuracy, gives accurate simple solutions for both long and short waves, even for large amplitudes. The KdV and Airy solutions are special limiting cases.

The weakly nonlinear theory for modelling flows away from the bifurcation point developed by the authors in their previous work (Suslov & Paolucci 1997) is generalized for flows of variable-density fluids in open systems. It is shown that special treatment of the continuity equation is necessary to perform the analysis of such flows and to account for the potential total fluid mass variation in the domain. The stability analysis of non-Boussinesq mixed convection flow of air in a vertical channel is then performed for a wide range of temperature differences between the walls, and Grashof and Reynolds numbers. A cubic Landau equation, which governs the evolution of a disturbance amplitude, is derived and used to identify regions of subcritical and supercritical bifurcations to periodic flows. Equilibrium disturbance amplitudes are computed for regions of supercritical bifurcations.

Based on amplitude expansions developed in Part 1 (Suslov & Paolucci 1999), we examine the mean flow characteristics of non-Boussinesq mixed convection flow of air in a vertical channel in the vicinity of bifurcation points for a wide range of temperature differences between the walls, and Grashof and Reynolds numbers. The constant mass flux and constant pressure gradient formulations are shown to lead to qualitatively similar, but quantitatively different, results. The physical nature of the distinct shear and buoyancy disturbances is investigated, and detailed mean flow and energy analyses are presented. The variation of the total mass of fluid in a flow domain as disturbances develop is discussed. The average Nusselt number and mass flux are estimated for supercritical regimes for a wide range of governing parameters.

The interaction between an initially laminar boundary layer developing spatially on a flat plate and wakes traversing the inlet periodically has been simulated numerically. The three-dimensional, time-dependent Navier–Stokes equations were solved with 5.24×107 grid points using a message passing interface on a scalable parallel computer. The flow bears a close resemblance to the transitional boundary layer on turbomachinery blades and was designed following, in outline, the experiments by Liu & Rodi (1991). The momentum thickness Reynolds number evolves from Reθ = 80 to 1120. Mean and second-order statistics downstream of Reθ = 800 are of canonical flat-plate turbulent boundary layers and are in good agreement with Spalart (1988).

In many important aspects the mechanism leading to the inception of turbulence is in agreement with previous fundamental studies on boundary layer bypass transition, as summarized in Alfredsson & Matsubara (1996). Inlet wake disturbances inside the boundary layer evolve rapidly into longitudinal puffs during an initial receptivity phase. In the absence of strong forcing from free-stream vortices, these structures exhibit streamwise elongation with gradual decay in amplitude. Selective intensification of the puffs occurs when certain types of turbulent eddies from the free-stream wake interact with the boundary layer flow through a localized instability. Breakdown of the puffs into young turbulent spots is preceded by a wavy motion in the velocity field in the outer part of the boundary layer.

Properties and streamwise evolution of the turbulent spots following breakdown, as well as the process of completion of transition to turbulence, are in agreement with previous engineering turbomachinery flow studies. The overall geometrical characteristics of the matured turbulent spot are in good agreement with those observed in the experiments of Zhong et al. (1998). When breakdown occurs in the outer layer, where local convection speed is large, as in the present case, the spots broaden downstream, having the vague appearance of an arrowhead pointing upstream.

The flow has also been studied statistically. Phase-averaged velocity fields and skin-friction coefficients in the transitional region show similar features to previous cascade experiments. Selected results from additional thought experiments and simulations are also presented to illustrate the effects of streamwise pressure gradient and free-stream turbulence.

Convection plays a major role in a variety of natural hydrodynamic systems. Those in which convection drives exchange flows through a lateral contraction and/or over a sill form a special class with typical examples being the Red and Mediterranean Seas, the Persian Gulf, and the fjords that indent many coastlines. The present work focuses on the spatial distribution and scaling of the density difference between the inflowing and outflowing fluid layers. Using a long water-filled channel, fitted with buoyancy sources at its upper surface, experiments were conducted to investigate the influence of the geometry of the strait and the channel as well as the magnitude of the buoyancy flux. Two different scaling laws, one by Phillips (1966), and one by Maxworthy (1994, 1997) were compared with the experimental results. It has been shown that a scaling law for which g′ = kB02/3x/h4/3 best describes the distribution of the observed density difference along the channel, where B0 is the buoyancy flux, x the distance from the closed end of the channel, h its height at the open end (sill) and k a constant that depends on the details of the channel geometry and flow conditions. This result holds for the experimental results and appears to be valid for a number of natural systems as well.

We have performed a numerical study on the transition of a cylindrical pipe flow under the influence of a localized disturbance in the form of periodic suction and blowing (PSB) applied at the pipe wall. We focus here on the so-called receptivity problem where the spatial evolution of this disturbance is studied as it travels downstream through the pipe. The study is carried out by means of two techniques: an eigenmode expansion solution (EES) and a full nonlinear direct numerical simulation (DNS). The EES is based on an analytical expansion in terms of the eigenfunctions of the linear operator which follows from the equations of motion expressed in a cylindrical coordinate system. The DNS is formulated in terms of a spectral element method.

We restrict ourselves to a so-called subcritical disturbance, i.e. the flow does not undergo transition. For very small amplitudes of the PSB disturbance the results of the EES and DNS techniques agree excellently. For larger amplitudes nonlinear interactions come into play which are neglected in the EES method. Nevertheless, the results of both methods are consistent with the following transition scenario. The PSB excites a flow perturbation that has the same angular wavenumber and frequency as the imposed disturbance itself. This perturbation is called the fundamental mode. By nonlinear self-interaction of this fundamental mode higher-order harmonics, both in the angular wavenumber and frequency, are generated. It is found that the harmonic with angular wavenumber 2, i.e. twice the wavenumber of the fundamental mode, and with zero frequency grows strongly by a linear process known as transient growth. As a result the (perturbed) pipe flow downstream of the disturbance region develops extended regions of low velocity, known as low-speed streaks. At large disturbance amplitudes these low-speed streaks show the development of high wavenumber oscillations and it is expected that at even higher disturbance amplitudes these oscillations become unstable and turbulent flow will set in.

Our result agrees (at least qualitatively) with the transition scenario in a plane Poiseuille flow as discussed by Reddy et al. (1998) and Elofson & Alfredson (1998).

We have experimentally examined the effects of bubble size (0.4 [les ] λ [les ] 2.0), inclination angle (0° [les ] α [les ] 90°), and tube material on suspended gas bubbles in flows in tubes for a range of Weber (0 [les ] We [les ] 3.6), Reynolds (0 [les ] Re [les ] 1200), and Froude (0 [les ] Frα [les ] 1) numbers. Flow rates and associated pressure differences which allow the suspension of bubbles in glass and acrylic tubes are measured. Due to contact angle hysteresis, bubbles which dry the tube wall (i.e. form a gas–solid interface) may remain suspended over a range of flows while non-drying bubbles remain stationary for a single flow rate depending on experimental conditions. Stationary bubbles increase the axial pressure gradient with larger bubbles and steeper inclination angles leading to the greatest increase in the pressure gradient. Both the suspension flow range and pressure difference modifications are strongly dependent upon gas/liquid/solid material interactions. Stronger contact forces, i.e. smaller spreading coefficients, cause dried bubbles in acrylic tubes to remain stationary over a wider range of suspension flows than bubbles in glass tubes. Bubble deformation is governed by the interaction of interfacial, contact, and flow-derived forces. This investigation reveals the importance of bubble size, tube inclination, and tube material on gas bubble suspension.

The response of a Gaussian vortex to a weak time-dependent external strain field is studied numerically. The cases of an impulsive strain, an on–off step function, and a continuous random strain are considered. Transfers of enstrophy between mean and azimuthal components are observed, and the results are compared with an analogous passive scalar model and with Kida's elliptical vortex model.

A ‘rebound’ phenomenon is seen: after enstrophy is transferred from mean to azimuthal component by the external straining field, there is a subsequent transfer of enstrophy back from the azimuthal component to the mean. Analytical support is given for this phenomenon using Lundgren's asymptotic formulation of the spiral wind-up of vorticity. Finally the decay of the vortex under a continuous random external strain is studied numerically and compared with the passive scalar model. The vorticity distribution decays more slowly than the scalar because of the rebound phenomenon.

The properties of gravito-inertial waves propagating in a stably stratified rotating spherical shell or sphere are investigated using the Boussinesq approximation. In the perfect fluid limit, these modes obey a second-order partial differential equation of mixed type. Characteristics propagating in the hyperbolic domain are shown to follow three kinds of orbits: quasi-periodic orbits which cover the whole hyperbolic domain; periodic orbits which are strongly attractive; and finally, orbits ending in a wedge formed by one of the boundaries and a turning surface. To these three types of orbits, our calculations show that there correspond three kinds of modes and give support to the following conclusions. First, with quasi-periodic orbits are associated regular modes which exist at the zero-diffusion limit as smooth square-integrable velocity fields associated with a discrete set of eigenvalues, probably dense in some subintervals of [0, N], N being the Brunt–Väisälä frequency. Second, with periodic orbits are associated singular modes which feature a shear layer following the periodic orbit; as the zero-diffusion limit is taken, the eigenfunction becomes singular on a line tracing the periodic orbit and is no longer square-integrable; as a consequence the point spectrum is empty in some subintervals of [0, N]. It is also shown that these internal shear layers contain the two scales E1/3 and E1/4 as pure inertial modes (E is the Ekman number). Finally, modes associated with characteristics trapped by a wedge also disappear at the zero-diffusion limit; eigenfunctions are not square-integrable and the corresponding point spectrum is also empty.

We study experimentally and theoretically the downward vertical displacement of one miscible fluid by another lighter one in the gap of a Hele-Shaw cell at sufficiently high velocities for diffusive effects to be negligible. Under certain conditions on the viscosity ratio, M, and the normalized flow rate, U, this results in the formation of a two-dimensional tongue of the injected fluid, which is symmetric with respect to the midplane. Thresholds in flow rate and viscosity ratio exist above which the two- dimensional flow destabilizes, giving rise to a three-dimensional pattern. We describe in detail the two-dimensional regime using a kinematic wave theory similar to Yang & Yortsos (1997) and we delineate in the (M, U)-plane three different domains, characterized respectively by the absence of a shock, the presence of an internal shock and the presence of a frontal shock. Theoretical and experimental results are compared and found to be in good agreement for the first two domains, but not for the third domain, where the frontal shock is not of the contact type. An analogous treatment is also applied to the case of axisymmetric displacement in a cylindrical tube.

It is shown that there is an abstract subgrid model that is in all senses ideal. An LES using the ideal subgrid model will exactly reproduce all single-time, multi-point statistics, and at the same time will have minimum possible error in instantaneous dynamics. The ideal model is written as an average over the real turbulent fields whose large scales match the current LES field. But this conditional average cannot be computed directly. Rather, the ideal model is the target for approximation when developing practical models, though no new practical models are presented here. To construct such models, the conditional average can be formally approximated using stochastic estimation. These optimal formulations are presented, and it is shown that a relatively simple but general class of one-point estimates can be computed from two-point correlation data, and that the estimates retain some of the statistical properties of the ideal model.

To investigate the nature of these models, optimal formulations were applied to forced isotropic turbulence. A variety of optimal models of increasing complexity were computed. In all cases, it was found that the errors between the real and estimated subgrid force were nearly as large as the subgrid force itself. It is suggested that this may also be characteristic of the ideal model in isotropic turbulence. If this is the case, then it explains why subgrid models produce reasonable results in actual LES while performing poorly in a priori tests. Despite the large errors in the optimal models, one feature of the subgrid interaction that is exactly represented is the energy transfer to the subgrid scales by each wavenumber.