Fluid mechanics with Adaptively Refined Large Eddy simulations
|Principal Investigator(s)||Prof. Jens Niemeyer|
|Leading Institution||Universität Würzburg, Lehrstuhl für Astronomie|
|DEISA Home Site||LRZ|
Project summary and results
Turbulence in engineering applications and atmospheric sciences has frequently been modelled by large eddy simulations (LES). In LES, the dynamics of turbulent eddies is computed on large scales, while a subgrid scale model approximates the influence of smaller eddies. However, in astrophysics, phenomena such as supersonic turbulence in star-forming gas clouds challenge the LES approach. The self-similarity hypothesis employed in LES fails to be applicable over a wide range of disparate scales. Thus, Alexei Kritsuk from the University of California, San Diego, proposed to adopt a method called adaptive mesh refi nement (AMR). This method involves inserting computational grids of higher resolution into fl ow regions where turbulent structures such as eddies or shock fronts are forming. A major problem is to fi nd the criteria for the generation of refi ned grids based on various fl uid dynamic processes.
The computational resources granted by the DEISA Extreme Computing Initiative (DECI) to perform highly resolved AMR simulations of supersonic turbulence were used. Depending on the size of the computational grid, 16 to 126 CPUs of the SGI Altix supercomputer at SARA, Netherlands, were required for each simulation. The basic idea was to trigger the refinement by monitoring fl ow properties such as the vorticity (the rotation of the velocity fi eld) and the rate of gas compression (due to shocks or gravity). This is illustrated by the simulation of a threedimensional visualisation of the isosurfaces of the vorticity (refer to the picture). The refined grids (drawn as yellow boxes) are mostly generated in the vicinity of sheet-like structures arising from shocks. The tube-like structures indicate the centres of eddies. An extensive statistical analysis showed good agreement with the known properties of turbulence inferred from simulations without AMR. This makes the FEARLESS project team confident that AMR can be carried over to more complex scenarios involving thermal and chemical processes as well as self-gravity.
However, owing to the extreme range of different length scales, it is generally impossible to treat fully developed turbulence by means of AMR only. This is because a prohibitive number of refined grids would be necessary. For this reason, the project team is presently combining AMR with a subgrid scale model which links the notions of AMR and LES. They call this new method "Fluid mEchanics with Adaptively Refined Large Eddy SimulationS" (FEARLESS). Upcoming applications will encompass the star formation in the turbulent interstellar medium, the feedback from star formation onto the evolution of spiral galaxies and the dynamics of hot gas in galaxy clusters. The project team expects that FEARLESS will open new perspectives in astrophysics by the as yet unequalled level of sophistication in the treatment of turbulence.
3D Visualisation: Isosurfaces of vorticity