This paper contains the details of an experimental study of the vortex formed in front of a piston as it moves through a cylinder. The mechanism for the formation of this vortex is the removal of the boundary layer forming on the cylinder wall in front of the advancing piston. The trajectory of the vortex core and the vorticity distribution on the developing vortex have been measured for a range of piston velocities. Velocity field measurements indicate that the vortex is essentially an inviscid structure at the Reynolds numbers considered, with viscous effects limited to the immediate corner region. Inviscid flow is defined in this paper as being a region of the flow where inertial forces are significantly larger than viscous forces. Flow visualization and vorticity measurements show that the vortex is composed mainly of material from the boundary layer forming over the cylinder wall. The characteristic dimension of the vortex appears to scale in a self-similar fashion, while it is small in relation to the apparatus length scale. This scaling rate of t0.85+0.7m, where the piston speed is described as a power law Atm, is somewhat faster than the t3/4 scaling predicted by Tabaczynski et al. (1970) and considerably faster than a viscous scaling rate of t1/2. The reason for the structure scaling more rapidly than predicted is the self-induced effect of the secondary vorticity that is generated on the piston face. The vorticity distribution shows a distinct spiral structure that is smoothed by the action of viscosity. The strength of the separated vortex also appears to scale in a self-similar fashion as t2m+1. This rate is the same as suggested from a simple model of the flow that approximates the vorticity being ejected from the corner as being equivalent to the flux of vorticity over a flat plate started from rest. However, the strength of the vorticity on the separated structure is 25% of that suggested by this model, sometimes referred to as the ‘slug’ model. Results show that significant secondary vorticity is generated on the piston face, forming in response to the separating primary vortex. This secondary vorticity grows at the same rate as the primary vorticity and is wrapped around the outside of the primary structure and causes it to advect away from the piston surface.

The stability of one-layer vortices with inhomogeneous horizontal density distributions is examined both analytically and numerically. Attention is focused on elliptical vortices for which the formal stability theorem proved by Ochoa, Sheinbaum & Pavía (1988) does not apply. Our method closely follows that of Ripa (1987) developed for the homogeneous case; and indeed they yield the same results when inhomogenities vanish. It is shown that a criterion from the formal analysis – the necessity of a radial increase in density for instability – does not extend to elliptical vortices. In addition, a detailed examination of the evolution of the inhomogeneous density fields, provided by numerical simulations, shows that homogenization, axisymmetrization and loss of mass to the surroundings are the main effects of instability.

This work reports results of numerical simulations of viscous incompressible flow past a sphere. The primary objective is to identify transitions that occur with increasing Reynolds number, as well as their underlying physical mechanisms. The numerical method used is a mixed spectral element/Fourier spectral method developed for applications involving both Cartesian and cylindrical coordinates. In cylindrical coordinates, a formulation, based on special Jacobi-type polynomials, is used close to the axis of symmetry for the efficient treatment of the ‘pole’ problem. Spectral convergence and accuracy of the numerical formulation are verified. Many of the computations reported here were performed on parallel computers. It was found that the first transition of the flow past a sphere is a linear one and leads to a three-dimensional steady flow field with planar symmetry, i.e. it is of the ‘exchange of stability’ type, consistent with experimental observations on falling spheres and linear stability analysis results. The second transition leads to a single-frequency periodic flow with vortex shedding, which maintains the planar symmetry observed at lower Reynolds number. As the Reynolds number increases further, the planar symmetry is lost and the flow reaches a chaotic state. Small scales are first introduced in the flow by Kelvin–Helmholtz instability of the separating cylindrical shear layer; this shear layer instability is present even after the wake is rendered turbulent.

Channel flow, initially fully developed and two-dimensional, is subjected to mean strains that emulate the effect of rapid changes of streamwise and spanwise pressure gradients in three-dimensional boundary layers, ducts, or diffusers. As in previous studies of homogeneous turbulence, this is done by deforming the domain of a direct numerical simulation (DNS); here however the domain is periodic in only two directions and contains parallel walls. The velocity difference between the inner and outer layers is controlled by accelerating the channel walls in their own plane, as in earlier studies of three-dimensional channel flows. By simultaneously moving the walls and straining the domain we duplicate both the inner and outer regions of the spatially developing case. The results are used to address basic physics and modelling issues. Flows subject to impulsive mean three-dimensionality with and without the mean deceleration of an adverse pressure gradient (APG) are considered: strains imitating swept-wing and pure skewing (sideways turning) three-dimensional boundary layers are imposed. The APG influences the structure of the turbulence, measured for example by the ratio of shear stress to kinetic energy, much more than does the pure skewing. For both deformations, the evolution of the Reynolds stress is profoundly affected by changes to the velocity–pressure-gradient correlation Πij. This term – which represents the finite time required for the mean strain to modify the shape and orientation of the turbulent motions – is primarily responsible for the difference (lag) in direction between the mean shear and the turbulent shear stresses, a well-known feature of perturbed three-dimensional boundary layers. Files containing the DNS database and model-testing software are available from the authors for distribution, as tools for future closure-model testing.

The structure and development of the scalar wake produced by a single line source are studied in decaying isotropic turbulence. The incompressible Navier–Stokes and the passive-scalar transport equations are solved via direct numerical simulations (DNS). The velocity and the scalar fields are generated by simulating Warhaft's (1984) experiment. The results for mean and r.m.s. scalar statistics are in good agreement with those obtained from the experiment. The structure of the scalar wake is examined first. At initial times, most of the contribution to the scalar variance is due to the flapping of the wake around the centreline. Near the end of the turbulent convective regime, the wake develops internal structure and the contribution of the flapping component to the scalar variance becomes negligible. The influence of the source size on the development of the scalar wake has been examined for source sizes ranging from the Kolmogorov microscale to the integral scale. After an initial development time, the half-widths of mean and scalar r.m.s. wakes grow at rates independent of the source size. The mixing in the scalar wake is studied by analysing the evolution of the terms in the transport equations for mean, scalar flux, variance, and scalar dissipation. The DNS results are used to test two types of closures for the mean and the scalar variance equations. For the time range simulated, the gradient diffusion model for the scalar flux and the commonly used scalar dissipation model are not supported by the DNS data. On the other hand, the model based on the unconditional probability density function (PDF) method predicts the scalar flux reasonably well near the end of the turbulent convective regime for the highest Reynolds number examined. The scalar source size does not significantly influence the models' predictions, although it appears that the time-scale ratio of mechanical dissipation to scalar dissipation approaches an asymptotic value earlier for larger source sizes.

Instabilities of a wake produced by a circular cylinder in a uniform water flow are studied experimentally when viscoelastic solutions are injected through holes pierced in the cylinder. It is shown that the viscoelastic solutions fill the shear regions and drastically modify the instabilities. The two-dimensional instability giving rise to the Kármán street is found to be inhibited: the roll-up process appears to be delayed and the wavelength of the street increases. The wavelength increase obeys an exponential law and depends on the elasticity number, which provides a ratio of elastic forces to inertial forces. The three-dimensional instability leading to the A mode is generally found to be suppressed. In the rare case where the A mode is observed, its wavelength is shown to be proportional to the wavelength of the Kármán street and the streamwise stretching appears to be inhibited. Injection of viscoelastic solutions also decreases the aspect ratio of the two-dimensional wake, and this is correlated with stabilization of the A mode and with changes in the shape of the Kármán vortices. The observations of this work are consistent with recent numerical simulations of viscoelastic mixing layers. The results suggest mechanisms through which polymers inhibit the formation of high-vorticity coherent structures and reduce drag in turbulent flows.

Vortex ring formation in a starting axisymmetric buoyant plume is considered. A model describing the process is proposed and a physical explanation based on the Kelvin–Benjamin variational principle for steady vortex rings is provided. It is shown that Lundgren et al.'s (1992) time scale, the ratio of the velocity of a buoyant plume after it has travelled one diameter to its diameter, is equivalent to the time scale (formation time) proposed by Gharib et al. (1998) for uniform-density vortex rings generated with a piston/cylinder arrangement. It is also shown that, similarly to piston-generated vortex rings (Gharib et al. 1998), the buoyant vortex ring pinches off from the plume when the latter can no longer provide the energy required for steady vortex ring existence. The dimensionless time of the pinch-off (the formation number) can be reasonably well predicted by assuming that at pinch-of the vortex ring propagation velocity exceeds the plume velocity. The predictions of the model are compared with available experimental results.

New laboratory experiments on different types of lock-exchange particle-driven gravity currents advancing into a flume of fresh water are presented. These include purely saline currents, monodisperse particle-laden gravity currents with both fresh and saline interstitial fluid, and bidisperse particle-laden currents. For each case a simple box model is developed. These agree well with the experimental data. We find that particulate gravity currents with saline interstitial fluid flowing into ambient fresh fluid are best described using a Froude number of 0.52 in the box model (cf. Huppert & Simpson 1980). However, particulate gravity currents with fresh interstitial fluid are best described using a higher Froude number of 0.67. The change in Froude number reflects the different shape and structure associated with the different density of interstitial fluid. For all experiments, box models provide accurate predictions for up to twenty lock-lengths.

Several recent papers discuss a viscous micropump consisting of Poiseuille flow of fluid between two plates with a cylinder placed along the gap perpendicular to the flow direction (e.g. Sen, Wajerski & Gad-el-Hak 1996). If the cylinder is not centred, rotating it will generate a net flow and an additional pressure drop along the channel, due to the net tangential viscous stresses along its surface. The research reported here complements existing work by examining the lubrication limit where the gaps between the cylinder and the walls are small compared to the cylinder radius. Lubrication analysis provides analytical relations among the flow rate, torque, pressure drop and rotation rate. Optimization of the flow parameters is shown in order to determine the optimal geometry of the device, which can be used by micro-electrical-mechanical systems designers. It is also shown, for example, that a device cannot be developed that achieves maximum flow rate and rotation simultaneously. In addition, since the Reynolds number can be smaller than 1, the Stokes equations are solved for this configuration using a numerical boundary integral method. The numerical results match the lubrication solution for small gaps, and determine the limits of validity for using the lubrication results.

When a ferrofluid layer is subjected to a uniform and vertically oriented magnetic field, an interfacial instability occurs, above a critical value of the magnetic field, giving rise to a hexagonal array of peaks. On increasing the magnetic field, a smooth morphological transition from the hexagonal array to a square array was observed above a second threshold. The hexagon–square transition phenomenology, in addition to the role of penta–hepta defects initially present in the hexagonal pattern, was investigated. Furthermore, the pattern and wavenumber selection was studied by two different procedures: first by imposing jumps in field intensity and second by varying the magnetic field in a quasi-static way. The results obtained were very different for the two procedures. They indicated that the square pattern was a metastable state induced by the compression of the hexagonal pattern on increasing the control parameter. This hypothesis was confirmed by performing an additional experiment where the pattern was isotropically compressed. In this experiment, the transition was induced at a constant magnetic field lower than the transition onset value. However, the theoretical values for stability domains of hexagons and squares proposed in the literature were found to not agree with the experimental values.

Burgers turbulence subject to a force f(x, t) = [sum ]jfj(x)δ(t − tj), where tj are 'kicking times' and the 'impulses' fj(x) have arbitrary space dependence, combines features of the purely decaying and the continuously forced cases. With large-scale forcing this ‘kicked’ Burgers turbulence presents many of the regimes proposed by E et al. (1997) for the case of random white-noise-in-time forcing. It is also amenable to efficient numerical simulations in the inviscid limit, using a modification of the fast Legendre transform method developed for decaying Burgers turbulence by Noullez & Vergassola (1994). For the kicked case, concepts such as ‘minimizers’ and ‘main shock’, which play crucial roles in recent developments for forced Burgers turbulence, become elementary since everything can be constructed from simple two-dimensional area-preserving Euler–Lagrange maps.

The main results are for the case of identical deterministic kicks which are periodic and analytic in space and are applied periodically in time. When the space integrals of the initial velocity and of the impulses vanish, it is proved and illustrated numerically that a space- and time-periodic solution is achieved exponentially fast. In this regime, probabilities can be defined by averaging over space and time periods. The probability densities of large negative velocity gradients and of (not-too-large) negative velocity increments follow the power law with −7/2 exponent proposed by E et al. (1997) in the inviscid limit, whose existence is still controversial in the case of white-in-time forcing. This power law, which is seen very clearly in the numerical simulations, is the signature of nascent shocks (preshocks) and holds only when at least one new shock is born between successive kicks.

It is shown that the third-order structure function over a spatial separation Δx is analytic in Δx although the velocity field is generally only piecewise analytic (i.e. between shocks). Structure functions of order p ≠ 3 are non-analytic at Δx = 0. For even p there is a leading-order term proportional to [mid ]Δx[mid ] and for odd p > 3 the leading-order term ∝Δx has a non-analytic correction ∝Δx[mid ]Δx[mid ] stemming from shock mergers.

Flows between ocean basins are often controlled by narrow channels and shallow sills. A multi-layer hydraulic control theory is developed for exchange flow through such constrictions. The theory is based on the inviscid shallow-water equations and extends the functional approach introduced by Gill (1977) and developed by Dalziel (1991). The flows considered are those in rectangular–cross-section channels connecting two large reservoirs, with a single constriction (sill and/or narrows). The exchange flow depends on the stratification in the two reservoirs, represented as a finite number of immiscible layers of (different) uniform density. For most cases the flow is ‘controlled’ at the constriction and often at other points along the channel (virtual controls) too. As with one- and two-layer hydraulics, controls are locations at which the flow passes from one solution branch to another, and at which (at least) one internal wave mode is stationary. The theory is applied to three-layer flows, which have two internal wave modes, predicting interface heights and layer fluxes from the given reservoir conditions. The theoretical results for three-layer flows are compared to a comprehensive set of laboratory experiments and found to give good agreement. The laboratory experiments also show other features of the flow, such as the formation of waves on the interfaces. The implications of the results for oceanographic flows and ocean modelling are discussed.

We model the propagation of turbulent gravity currents through an array of obstacles which exert a drag force on the flow proportional to the square of the flow speed. A new class of similarity solutions is constructed to describe the flows that develop from a source of strength q0tγ. An analytical solution exists for a finite release, γ = 0, while power series solutions are developed for sources with γ > 0. These are shown to provide an accurate approximation to the numerically calculated similarity solutions. The model is successfully tested against a series of new laboratory experiments which investigate the motion of a turbulent gravity current through a large flume containing an array of obstacles. The model is extended to account for the effects of a sloping boundary. Finally, a series of geophysical and environmental applications of the model are discussed.

A quantitative theory is described for the formation mechanism of sand bars under surface water waves. By assuming that the slopes of waves and bars are comparably gentle and sediment motion is dominated by the bedload, an approximate evolution equation for bar height is derived. The wave field and the boundary layer structure above the wavy bed are worked out to the accuracy needed for solving this evolution equation. It is shown that the evolution of sand bars is a process of forced diffusion. This is unlike that for sand ripples which is governed by an instability. The forcing is directly caused by the non-uniformity of the wave envelope, hence of the wave-induced bottom shear stress associated with wave reflection, while the effective diffusivity is the consequence of gravity and modified by the local bed stress. During the slow formation, bars and waves affect each other through the Bragg scattering mechanism, which consists of two concurrent processes: energy transfer between waves propagating in opposite directions and change of their wavelengths. Both effects are found to be controlled locally by the position of bar crests relative to wave nodes. Comparison with available laboratory experiments is discussed and theoretical examples are studied to help understand the coupled evolution of bars and waves in the field.