Towards Ultimate Turbulence

PhD defence: 10 November 2017, 14:45 Hours
Thesis title: Towards parallel-in-time-simulations of turbulent Rayleigh-Bénard convection.
Gijs L. Kooij

Large-scale computations of Rayleigh-Bénard Convection


Funded by:



Gijs Kooij


Bernard Geurts & Mike Bochev




A classic example of natural convection of a fluid is Rayleigh-Bénard convection. A layer of fluid is heated from below and cooled from above. When buoyant forces on the fluid are sufficiently large, the fluid convects the heat from the hot bottom to the cool top. Rayleigh-Bénard convection is encountered in various environments, such as the atmosphere on Earth or the interior of the Sun. In many cases, the buoyant forces are strong enough to create a turbulent flow.

The PhD project is mainly concerned with direct numerical simulations of turbulent Rayleigh-Bénard convection and the development of numerical methods for this application. Direct numerical simulations of strongly turbulent Rayleigh-Bénard convection present a considerable challenge at present. Simulations of turbulent flows are computationally demanding and unavoidably require parallel computing. Current numerical methods are however limited in the number of processors they can use efficiently.

We plan to develop a method that parallelizes in the time domain, in contrast to the traditional parallelization in the spatial domain. This parallel-in-time technique allows the additional use of many more processors in parallel. With this capability we can speed-up simulations that would take impossibly long computation time otherwise.



Rayleigh-Bénard convection in a cylindrical container