March 7, 2014
Speaker: Dr. Lucian Ivan, Computational Fluid Dynamics Postdoctoral Fellow and Lecturer, University of Waterloo
Title: High-performance computational methods for partial differential equations (PDEs) with application to space-physics flows
Abstract: Accurate, efficient and scalable computational methods are highly desirable for large-scale scientific computing applications, especially for problems exhibiting spatial and temporal multi-resolution scales, non-trivial geometries and complex boundary conditions (BCs). For global magnetohydrodynamics (MHD) modelling of space-physics problems, high-performance approaches could significantly reduce the grid requirements, thereby enabling more affordable yet accurate predictions of large-scale space-weather phenomena. Key challenges encountered relate to providing solenoidal magnetic fields, accurate discretizations on spherical domains, capturing of MHD shocks, and implementing accurate BCs.
This talk gives an overview of a high-order finite-volume discretization procedure in combination with a parallel solution-adaptive algorithm for the numerical simulation of physically complex flows. The focus is on the challenges, development and application to conservation laws pertaining to space physics, in particular to MHD for space plasma flows. Discretization of spherical geometries is performed with cubed-sphere grids, which have recently gained popularity for simulations in a variety of application domains such as climate and weather modelling, and magna/mantle dynamics. Numerical results to demonstrate the accuracy and capability of the multidimensional high-order, solution-adaptive, cubed-sphere computational framework are presented. Extensions to unstructured grids and other applications are also discussed.