Statistical descriptions of polydisperse turbulent two-phase flows

Disperse two-phase flows are flows containing two non-miscible phases where one phase is present as a set of discrete elements dispersed in the second one. These discrete elements, or 'particles', can be droplets, bubbles or solid particles having different sizes. This situation encompasses a wide range of phenomena, from nano-particles and colloids sensitive to the molecular fluctuations of the carrier fluid to inertia particles transported by the large-scale motions of turbulent flows and, depending on the phenomenon studied, a broad spectrum of approaches have been developed. The aim of the present article is to analyze statistical models of particles in turbulent flows by addressing this issue as the extension of the classical formulations operating at a molecular or meso-molecular level of description. It has a three-fold purpose: (1) to bring out the thread of continuity between models for discrete particles in turbulent flows (above the hydrodynamical level of description) and classical mesoscopic formulations of statistical physics (below the hydrodynamical level); (2) to reveal the specific challenges met by statistical models in turbulence; (3) to establish a methodology for modeling particle dynamics in random media with non-zero space and time correlations. The presentation is therefore centered on organizing the different approaches, establishing links and clarifying physical foundations. The analysis of disperse two-phase flow models is developed by discussing: first, approaches of classical statistical physics; then, by considering models for single-phase turbulent flows; and, finally, by addressing current formulations for discrete particles in turbulent flows. This brings out that particle-based models do not cease to exist above the hydrodynamical level and offer great interest when combined with proper stochastic formulations to account for the lack of equilibrium distributions and scale separation. In the course of this study, general results for dynamical systems influenced by colored or white noises are established which point to the necessity of developing statistical models for particles in random media from a new standpoint that breaks away from descriptions based only on particle kinetic variables. This is needed to obtain well-posed formulations of particle dispersion by non fully-resolved turbulent flows and suggests also to revisit classical ideas of particle–particle interactions in the presence of a turbulent flow which correlates the dynamics of colliding discrete elements.