The structure and development of turbulent plane jets in still air and moving streams are described. The nature of the small-scale turbulence cannot be accurately ascertained because of the difficulties inherent in the measurement of dissipation in highly turbulent flows. Although correlation measurements in a jet in still air indicate a large-scale structure which can best be described as ‘local flapping’, measurements in a jet in a moving stream do not reveal a similar structure. The development of the turbulence structure in a jet in a moving parallel stream is described and the properties of turbulent jets and wakes are shown to be reasonably well predicted by the use of a variable-eddy-viscosity formula together with the formal self-preserving properties of the equations of motion.

We study the time-dependent solutions of a nonlinear cascade model for homogeneous isotropic turbulence first introduced by Novikov & Desnyansky. The dynamical variables of the model are the turbulent kinetic energies in discrete wave-number shells of thickness one octave. The model equations contain a parameter C whose size governs the amount of energy cascaded to small wavenumbers relative to the amount cascaded to large wavenumbers. We show that the equations permit scale-similar evolution of the energy spectrum. For 0 ≤ C ≤ 1 and no external force, the freely evolving energy spectrum displays the Kolmogorov k power law, and the total energy decreases in time as a power t−w, where the exponent w depends on the value of C. Grid-turbulence experiments seem to favour a value of C in the range 0·3-0·6. In the presence of an external stirring force acting near a wavenumber k0, the model predicts, in addition to the Kolmogorov k spectrum for k > k0, a scale-similar flow of energy to wavenumbers k < k0. This backward energy flow falls off as a power law in time, and establishes a stationary energy spectrum for k < k0 which is a power law in k less steep than k. We discuss the similarity of the behaviour of the model for C > 1 to the behaviour of turbulent fluid for a spatial dimensionality near 2. The model is shown to approach the Kovasznay and the Leith diffusion approximation equations in the limit in which the thickness of the wavenumber shells approaches zero. However, the cascade model with finite shell thicknesses appears to behave in a more physically reasonable way than the limiting differential equations.

The force on a small sphere translating relative to a slow viscous flow is found to O(Re½) for two different fluid flows far from the sphere, namely pure rotation and pure shear. For pure rotation, the O(Re½) correction to the Stokes drag consists of an increase in the drag. For pure shear, the O(Re½) force contains a component perpendicular to the Stokes drag force.

The paper describes correlation measurements in both broad and narrow frequency bands of the longitudinal velocity fluctuations in fully developed pipe flow at four positions for a reference probe whilst a second probe was traversed radially from deep in the sublayer to a position near the axis with both longitudinal and transverse separations zero (Δx = Δz = 0). Such measurements require that both the Covariant (Co) and Quadrature (Quad) correlations be determined for each of the 15 frequencies used to constrain the wave component λx.

The new data demonstrate that low frequency, large scale turbulence fluctuations extend over the majority of the radial region and that these components are highly correlated. By using a similarity variable kxy, along with a normalized wall distance y/y REF, both correlation functions, i.e. the Co and the Quad components, are shown to collapse. The physical significance of this is discussed.

The broad-band data do not collapse because of the large range of wave sizes. However, the present experiment does show that strong radial correlations exist even when one probe is at y+ = 1. This conflicts with the earlier data of Favre, but agrees with the more recent work of Comte-Bellot. There is a significant amount of turbulent energy in frequencies less than 16 Hz (ω+ = 0·008) for turbulent flows of about 105 Reynolds number.

The spectral function ωΦ(ω) is also presented for a range of y+ values. Using this form for the power spectral density, along with the stochastic wave modelling and similarity arguments of this paper, it is shown how a consistent explanation for the behaviour of these spectra is obtained. In addition some preliminary results from cross-spectral analyses are presented and suggestions made as to their physical significance.