We consider the evolution of packets of water waves that travel predominantly in one direction, but in which the wave amplitudes are modulated slowly in both horizontal directions. Two separate models are discussed, depending on whether or not the waves are long in comparison with the fluid depth. These models are two-dimensional generalizations of the Korteweg-de Vries equation (for long waves) and the cubic nonlinear Schrödinger equation (for short waves). In either case, we find that the two-dimensional evolution of the wave packets depends fundamentally on the dimensionless surface tension and fluid depth. In particular, for the long waves, one-dimensional (KdV) solitons become unstable with respect to even longer transverse perturbations when the surface-tension parameter becomes large enough, i.e. in very thin sheets of water. Two-dimensional long waves (‘lumps’) that decay algebraically in all horizontal directions and interact like solitons exist only when the one-dimensional solitons are found to be unstable.

The most dramatic consequence of surface tension and depth, however, occurs for capillary-type waves in sufficiently deep water. Here a packet of waves that are everywhere small (but not infinitesimal) and modulated in both horizontal dimensions can ‘focus’ in a finite time, producing a region in which the wave amplitudes are finite. This nonlinear instability should be stronger and more apparent than the linear instabilities examined to date; it should be readily observable.

Another feature of the evolution of short wave packets in two dimensions is that all one-dimensional solitons are unstable with respect to long transverse perturbations. Finally, we identify some exact similarity solutions to the evolution equations.

We present a simple model of a steady barotropic eddy superimposed on a current, with varying Coriolis force, to study the structure of a nonlinear deformed eddy. We also develop the linearized time-dependent formulation of the problem. The resulting solutions are stable perturbations, with scales smaller than the eddy.

Free-flight and wind-tunnel measurements by previous investigators of the flat-plate autorotation phenomenon have been analysed. The variation of the autorotation characteristics with changes in the Reynolds number and the aspect ratio, thickness ratio and moment of inertia of the flat plate have been correlated. The interpretation of the role of the Reynolds number made in a previous investigation is shown to be incorrect. The tip-speed ratio, for the ranges of the dimensionless parameters investigated, is shown to be a function of only the plate aspect ratio, thickness ratio, and also the moment of inertia if the latter is sufficiently small. The lift and drag coefficients, and therefore the free-flight glide angle, are shown to be functions of the tip-speed ratio, the aspect ratio and the Reynolds numbers based on the chord and plate thickness.

Experimental investigations in the region following the passage of an isolated turbulent spot in a laminar boundary layer reveal the existence of a pair of oblique wave packets. These packets are swept at an angle of approximately 40°, and exhibit frequency and wave speed characteristics in agreement with predictions made for oblique Tollmien-Schlichting waves. No waves exist near the centreline of the spot.

Several observations of the breakdown of this ordered motion into a new turbulent spot are shown. This breakdown is accompanied by the appearance of an intense shear layer inclined to the wall.

A theoretical framework is given, upon which to examine the dispersion relation of random gravity waves. First a weakly nonlinear theory is developed to the third-order for a statistically stationary and homogeneous field of random gravity waves. Both the spectrum of forced waves and the nonlinear dispersion relation are expressed in terms of the spectrum of free waves under the assumption of the Gaussian process for the first-order surface displacement. Next a method is proposed by which to separate each of the spectra of free and forced waves from the measured spectrum. This gives practical and powerful means of investigating the statistical structure of wind waves.

The propagation of linearized waves on a non-uniform slowly varying potential current is studied by converting the equations of flow into a Schrödinger ordinary differential equation in the complex plane. This equation, which is solved by the WKB method, indicates the existence of current barriers which allow partial energy transmission while reflecting the complementary part. The classical result of total transmission (Longuet-Higgins & Stewart 1961) as well as that of complete reflexion (Peregrine 1976) are recovered as limiting cases by the present, more general approach.

The generation of internal waves by resonantly interacting surface waves is examined in the framework of spectral scattering theory in the random-phase approximation. Coupling coefficients are derived from Euler's equation of motion for arbitrary stratification. The spectral energy transfer is discussed for deep-water surface waves and a simple three-layer model of the stability frequency. Analytical and numerical evaluation of the transfer integral leads to a parametrization in terms of the basic model parameters. These are the depth, thickness and stability frequency of the thermocline and the scale parameters and bandwidth of the surface wave spectrum. Strong dependence on some of these parameters, in particular the surface wave energy and the ratio of surface and internal wave frequencies, indicates a large spatial and temporal variability of the transfer rate. The transfer to the internal wave field in the oceanic main thermocline is found to be negligible compared with the effect of other processes. High frequency waves in the seasonal thermocline may be generated very efficiently.

The higher-order theory of the Weissenberg effect is developed as a perturbation of the state of rest. The perturbation is given in powers of the angular frequency Ω of the rod and the solution is carried out to O(Ω4). The perturbation induces a slow-motion expansion of the stress into Rivlin-Ericksen tensors in combinations which are completely characterized by five viscoelastic parameters. The effects of each of the material parameters may be computed separately and their overall effect by superposition. The values of the parameters may be determined by measurement of the torque, surface angular velocity and height of climb. Such measurements are reported here for several different sample fluids. Good agreement between the third-order theory and measured values of the velocity is reported. Secondary motions which appear at are computed using biorthogonal series. The analysis predicts the surprising fact that secondary motions run up the climbing bubble against gravity and intuition.

Random waves are generated by wind in the first half of a wind-wave flume. The latter half of the flume is kept free from wind to measure the waves unaffected by the wind and wind-generated current. The random waves in the latter area are measured with a linear array of wave gauges, and their phase velocities and coherences are determined by a usual technique of the cross-spectral analysis. The measured results are compared with the nonlinear theory of two-dimensional random waves, which has been presented in part 1 of this paper (Masuda, Kuo & Mitsuyasu 1979). Agreement between the theory and the experiment is satisfactory, and observed characteristics of the phase velocity and coherence of the spectral components can be attributed to the effects of the nonlinearity and angular dispersion of the random waves.

The cross-stream migration of a deformable drop in a unidirectional shear flow of a second-order fluid is considered. Expressions for the particle velocity due to the separate effects of deformation and viscoelastic rheology are obtained. The direction and magnitude of migration are calculated for the particular cases of Poiseuille flow and simple shear flow and compared with experimental data.

The flow past a cylinder bounded by parallel planes in a rotating frame is treated in terms of a nonlinear Stewartson layer. It is shown that separation is strongly dependent on the ratio of the Rossby number to the square root of the Ekman number, covering the whole range from fully attached flow to the classical non-rotating separated flow past a cylinder. The results are in agreement with published experimental observations.

The equations for the rotation of non-axisymmetric ellipsoids in a simple shear flow at low Reynolds numbers are derived in terms of Euler angles. Numerical solutions of this third-order system of equations show a doubly periodic structure to the rotation, with a change in the general nature of the solutions when a certain planar rotation of the particle becomes unstable. Some analytic progress can be made for nearly spherical ellipsoids and for nearly axisymmetric ellipsoids. The near spheres show the same qualitative behaviour as the general ellipsoids. Quite small deviations from axial symmetry are found to produce large changes in the rotation.

The asymptotic theory for the laminar, incompressible, separating and reattaching flow past the bluff body is based on an extension of Kirchhoff's (1869) free-streamline solution. The flow field (only the upper half of which is discussed since we consider a symmetric body and flow) consists of two basic parts. The first is the flow on the body scale l*, which is described to leading order by the Kirchhoff solution with smooth inviscid separation, but with an $O(Re^{-\frac{1}{16}})$ modification to explain fully the viscous separation (here Re ([Gt ] 1) is the Reynolds number). The influence of this $O(Re^{-\frac{1}{16}})$ modification is determined for the circular cylinder. The second part is the large-scale flow, comprising mainly the eddy and the ultimate wake. The eddy has length scale O(Rel*), width O(Re½l*) and is of elliptical shape to keep the eddy pressure almost uniform. The ultimate wake is determined numerically and fixes the eddy length. The (asymptotically small) back pressure from the eddy acts (on the body scale) both in the free stream and in the eddy, and it has a marked effect at moderate Reynolds numbers; combined with the Kirchhoff solution, it predicts the pressure drag on a circular cylinder accurately, to within 10% when Re = 5 and to within 4% when Re = 50. Other predictions, for the eddy length and width, the front pressure and the eddy pressure, also show encouraging agreement with experiments and Navier-Stokes solutions at moderate Reynolds numbers (of about 30), both for the circular cylinder and the normal flat plate. Finally, an analysis in the appendix indicates that, in wind-tunnel experiments, the tunnel walls (even if widely spaced) can exert considerable influence on the eddy properties, eventually forcing an upper bound on the eddy width as Re increases instead of the O(l* Re½) growth appropriate to the unbounded flow situation.

Previous analyses of natural convection in a porous medium have drawn seemingly contradictory conclusions as to whether the motion is two- or three-dimensional. This investigation uses numerical results to show the relationship between previous contending observations, and demonstrates that there exists more than one mode of convection for any particular physical configuration and Rayleigh number. In some cases, a particular flow situation may be stable even though it does not maximize the energy transfer across the system.

The methods used are based on the efficient numerical solution of the governing equations, formulated with the definition of a vector potential. This approach is shown to be superior to formulating the equations in terms of pressure.

For a cubic region the flow pattern at a particular value of the Rayleigh number is not unique and is determined by the initial conditions. In some cases there exist four alternatives, two- and three-dimensional, steady and unsteady.

Numerical solutions for the irrotational flow of an incompressible fluid about a circular cylinder accelerated from rest below a free surface are presented. The usual restriction to linearized free-surface boundary conditions has been avoided. The transient period from the start to a local steady state or to the development of a very steep wave slope is investigated in terms of free-surface profiles and body-surface pressure distributions. Linear and nonlinear results are used to illustrate the transition from deep submergence when nonlinear effects are small to shallow submergence when linearized analysis is inaccurate.