Objective: To develop a fluid dynamics code that incorporates the effects of magnetic and radiation fields for massively parallel supercomputers and apply it to the study of the dynamics of astrophysical plasmas.
Approach: Standard finite-difference methods are used to evolve the equations of fluid dynamics. Special-purpose algorithms developed by the PI are used to evolve the magnetic and radiation fields. The code is written in Fortran using a data parallel paradigm.
Accomplishments: Fully 3D hydrodynamic algorithms including the effects of magnetic fields have been implemented on a variety of massively parallel supercomputers, including the Thinking Machines Corp. CM-2 and CM-5 and the MasPar MP-2. Performance on these machines varies from 2 to 20 times faster than on one CRAY Y-MP processor. The code now is being used to study the dynamics of magnetized accretion disks. The accompanying figure shows the turbulence that results in a 3D section of a weakly magnetized accretion disk from the development of magnetic instabilities in the flow. The data is taken from the largest simulation ever performed to date of these phenomena, run on a 512-node CM-5. The left panel shows the density in the disk on two surfaces of the 3D computational volume. Red denotes the highest densities, purple the lowest. Large-amplitude fluctuations are present throughout the disk. The right panel is a volumetric rendering of the magnetic pressure in the disk. The field is clearly organized into discrete blobs and filaments with a chaotic pattern indicative of magnetohydrodynamic turbulence.
Significance: Many astrophysical systems behave as fluids; thus, a theoretical description of their dynamics is given by solutions of the equations of fluid dynamics. However, astrophysical plasmas are complex because they are affected by a variety of physical phenomena, such as magnetic fields and radiation fields from nearby stars. By implementing numerical algorithms for magnetic fluids on massively parallel machines, the largest and most-detailed numerical simulations of the dynamics of astrophysical plasmas in a variety of contexts will be possible. This development not only directly benefits research in astrophysical plasmas, but many of the techniques developed here are applicable in a wide range of disciplines that rely on computational fluid dynamics.
Status/Plans: Both serial and parallel programs of the algorithms have been implemented and tested. Currently, we are investigating various techniques to incorporate adaptive block coding to improve the performance of our algorithms.
Point of Contact:
Dr. James Stone
Department of Astronomy
University of Maryland at College Park
jstone@astro.umd.edu
301-405-2103