As two fluid particles separate in time, the entire spectrum of eddy motions is being sampled from the smallest to the largest scales. In large-scale geophysical systems for which the Earth rotation is important, it has been conjectured that the relative diffusivity should vary respectively as D 2 and D 4/3 for distances respectively smaller and larger than a well-defined forcing scale of the order of the internal Rossby radius (with D the r.m.s. separation distance). Particle paths data from a mid-latitude float experiment in the central part of the North Atlantic appear to support these statements partly: two particles initially separated by a few km within two distinct clusters west and east of the mid-Atlantic ridge, statistically dispersed following a Richardson regime (D 2 ∼ t 3 asymptotically) for r.m.s. separation distances between 40 and 300 km, in agreement with a D 4/3 law. At early times, and for smaller separation distances, an exponential growth, in agreement with a D 2 law, was briefly observed but only for the eastern cluster (with an e-folding time around 6 days). After a few months or separation distances greater than 300 km, the relative dispersion slowed down naturally to the Taylor absolute dispersion regime. IntroductionObservations of the separation of pairs of particles is one of the few experimental methods available to examine the spatial structure of geophysical turbulent flows. Richardson (1926) proposed that the relative diffusivity of an ensemble of pairs should scale as the 4/3 power of the (r.m.s.) separation distance. In his review of the subject, Corrsin (1962) emphasized the concept that turbulent eddies much smaller or much larger than the separation scale are relatively inefficient in further separation at the difference of eddies near the separation scale: the small eddies cause independent random walks of each member of the pair while the larger ones move the pair coherently as a single unit. As a result the relative velocity of the pair changes with the separation and is a non-stationary random variable. This accelerating property of relative diffusion has been used to infer properties of the energy wavenumber spectrum in the inertial range. The key quantity in the inertial range of three-dimensional turbulence is the energy dissipation rate. When the relative diffusivity is assumed to depend only on this energy dissipation rate and on separation, Richardson's law is recovered (Obukhov 1941;Batchelor 1952). The study of particle dispersion is important because of the interest in transport and mixing of chemicals in large-scale geophysical systems. While Taylor's (1921) single-particle dispersion theory relates to tracer dispersal from a fixed geographical origin and for very large times, two-particle dispersion relates to the spreading of a cloud of tracer from its centre of gravity (Batchelor 1952) and will

We investigate bubble dispersion in turbulent Taylor-Couette flow. The aim of this study is to describe the main mechanisms yielding preferential bubble accumulation in near-wall structures of the flow. We first proceed to direct numerical simulation of Taylor-Couette flows for three different geometrical configurations (three radius ratios η = R 1 /R 2 : η = 0.5, η = 0.72, and η = 0.91 with the outer cylinder at rest) and Reynolds numbers corresponding to turbulent regime ranging from 3000 to 8000. The statistics of the flow are discussed using two different averaging procedures that permit to characterize the mean azimuthal velocity, the Taylor vortices contribution and the small-scale turbulent fluctuations. The simulations are compared and validated with experimental and numerical data from literature. The second part of this study is devoted to bubble dispersion. Bubble accumulation is analyzed by comparing the dispersion obtained with the full turbulent flow field to bubble dispersion occurring at lower Reynolds numbers in previous works. Several patterns of preferential accumulation of bubbles have been observed depending on bubble size and the effect of gravity. For the smaller size considered, bubbles disperse homogeneously throughout the gap, while for the larger size they accumulate along the inner wall for the large gap width (η = 0.5). Varying the intensity of buoyancy yields complex evolution of the bubble spatial distribution. For low gravity effect, bubble entrapment is strong leading to accumulation along the inner wall in outflow regions (streaks of low wall shear stress). When buoyancy effect dominates on vortex trapping, bubbles rise through the vortices, while spiral patterns stretched along the inner cylinder are clearly identified. Force balance is analyzed to identify dominating forces leading to this accumulation and accumulation patterns are compared with previous experiments.

To measure the void fraction distribution in gas-liquid flows, a two-dimensional x-ray densitometry system was developed. This system is capable of acquiring a two-dimensional projection with a 225 cm 2 area of measurement through 21 cm of water. The images can be acquired at rates on the order of 1 kHz. Common sources of error in x-ray imaging, such as x-ray scatter, image distortion, veiling glare, and beam hardening were considered, and mitigated. The measured average void fraction was compared successfully to that of a phantom target and found to be within 1%. To evaluate the performance of the new system, the flow in and downstream of a ventilated nominally two-dimensional partial cavity was investigated and compared to measurements from dual tip fiber optical probes and high speed video. The measurements were found to have satisfactory agreement for void fractions above 5% of the selected void fraction measurement range.