Institut für Theoretische Physik, Ruhr-Universität Bochum (Germany)
Local Project ID:
HPC Platform used:
JUQUEEN of JSC
Using high resolution direct numerical simulations of a flow seeded with particles around a sphere, an international research team aimed at studying the hydrodynamic problem of collisions among particles in the potentially turbulent wake of a sphere. HPC system JUQUEEN of JSC served as computing platform for this challenging simulation project.
The aim of this project is to study the hydrodynamic problem of collisions among particles in the potentially turbulent wake of a sphere (see Illustration 1). An environmental application is a water drop (sphere) falling through a warm cloud seeded with small water droplets (particles). Some droplets might hit the water drop due to their inertia but others pass the drop and are exposed to its wake that can be laminar or turbulent depending on the size of the water drop. The passing droplets are accelerated by fluctuations in the wake that can trigger collisions among them. It is the influence of turbulent accelerations on inter-particle collisions that the scientists study in this project.
In the context of cloud physics a crucial quantity is the droplet size distribution (DSD). However, our today's knowledge about the evolution of the DSD is too limited to explain for example the observed timescales of rain formation. A reason is that it is very difficult to measure the DSD experimentally and numerically. It is known that it is first ruled by condensation of water vapor and later on by coalescence of droplets. However, today it is not known how falling raindrops modify the DSD. This project tries to give a detailed answer to this question
High resolution direct numerical simulations of a flow seeded with particles around a sphere is a challenging problem requiring the power of supercomputers. The non-linear equations of motions have to be solved in an extended three-dimensional domain that contains the major part of the turbulent wake. The resolution must be sufficient to resolve the thin boundary layer around the sphere and the small scale turbulent fluctuations in its wake. Additionally, the trajectories of millions of small particles have to be integrated and collisions faithfully detected. For this the scientists developed a massive parallel pseudo-spectral Navier-Stokes solver (called LaTu) adapted to the BlueGene/Q supercomputer JUQUEEN of the Jülich Supercomputing Centre.
The performed simulations revealed two important findings. First, the position and number of collisions strongly depends on the structure of the sphere wake. A turbulent wake triggers much more collisions than a laminar wake. The reason is that turbulent velocity fluctuations create particle concentrations and enhanced relative velocities. This implies that the turbulent DSD is much broader than the laminar one, meaning that a sphere is able to produce large condensates in its sphere. Another remarkable result is that the positions of particle collisions are much more spread in the turbulent than in the laminar case (see illustration 2).
Rainer Grauer (Ruhr-Universität Bochum, Germany), Holger Homann (Observatoire de la Côte d'Azur, Nice, France), and Jérémie Bec (Observatoire de la Côte d'Azur, Nice, France)
Laboratoire Lagrange UMR7293
Université de Nice Sophia Antipolis
CNRS/Observatoire de la Côte d’Azur
BP4229, 06304 Nice Cedex 4 (France)
e-mail: holger.homann [@] oca.eu