Computational Investigation Of The Effects Of Turbulence, Inertia, And Gravity On Particle Dynamics
In this work, we examine the motion of particles which are subjected to varying levels of turbulence, inertia, and gravity, in both homogeneous and inhomogeneous turbulence. These investigations are performed through direct numerical simulation (DNS) of the Eulerian fluid velocity field combined with Lagrangian particle tracking. The primary motivation of these investigations is to better understand and model the dynamics and growth of water droplets in warm, cumulus clouds. In the first part of this work, we discuss the code we developed for these simulations, Highly Parallel Particle-laden flow Solver for Turbulence Research (HiPPSTR). HiPPSTR uses efficient parallelization strategies, timeintegration techniques, and interpolation methods to enable massively parallel simulations of three-dimensional, particle-laden turbulence. In the second, third, and fourth sections of this work, we analyze simulations of particle-laden flows which are representative of those at the edges and cores of clouds. In the second section, we consider the mixing of droplets near interfaces with varying turbulence intensities and gravitational orientations, to provide insight into the dynamics near cloud edges. The simulations are parameterized to match windtunnel experiments of particle mixing which were conducted at Cornell, and the DNS and experimental results are compared and contrasted. Mixing is suppressed when turbulence intensities differ across the interface, and in all cases, the particle concentrations are subject to large fluctuations. In the third and fourth sections, we use HiPPSTR to analyze droplet motion in isotropic turbulence, which we take to be representative of adiabatic cloud cores. The third section examines the Reynolds-number scaling of single-particle and particle-pair statistics without gravity, while the fourth section shows results when gravity is included. While weakly inertial particles preferentially sample certain regions of the flow, gravity reduces the degree of preferential sampling by limiting the time particles can spend interacting the underlying turbulence. We find that when particle inertia is small, the particle relative velocities and radial distribution functions (RDFs) are almost entirely insensitive to the flow Reynolds number, both with and without gravity. The relative velocities and RDFs for larger particles tend to weakly depend on the Reynolds number and to strongly depend on the degree of gravity. While non-local, path-history interactions significantly affect the relative velocities of moderate and large particles without gravity, these interactions are suppressed by gravity, reducing the relative velocities. We provide a physical explanation for the trends in the relative velocities with Reynolds number and gravity, and use the model of  to understand and predict how the trends in the relative velocities will affect the RDFs. The collision kernels for particles representative of those in atmospheric clouds are generally seen to be independent of Reynolds number, both with and without gravity, indicating relatively low Reynolds-number simulations are able to capture much of the physics responsible for droplet collisions in clouds. We conclude by discussing practical implications of this work for the cloud physics and turbulence communities and suggesting areas for future research.
isotropic turbulence; shearless mixing layer; inertia particles
Diamessis, Peter J.; Warhaft, Zellman
Ph.D. of Mechanical Engineering
Doctor of Philosophy
dissertation or thesis