Home » Technology » Scientific Computing » Simulating Galaxy Cluster Collisions

Simulating Galaxy Cluster Collisions

© Adam Hughes

Galaxy clusters, some of the largest structures in the universe, are a source of complex physical phenomena with tremendous potential value to astronomers searching for key information about the nature of all matter and existence.

Of particular recent interest are the large-scale motions caused by the collision of two galaxy clusters and the associated effects on the new, combined system. To investigate this and other problems related to clusters of galaxies, researchers at the University of Chicago have turned to state-of-the-art computer simulation techniques.

The code (COSMOS) developed to simulate galaxy clusters consists of multiple components linked together to achieve a realistic model of these galactic systems. The first piece is a hydrodynamical code similar to those that have been discussed here in the past. This program is linked to a particle mesh N-body code through a solver for the gas equations of motion. This gas model, which is gravitationally coupled to the masses in the mesh, such as dark matter, allows the researchers to model the flows in the core of the colliding galaxies. In particular, this model can be used to study what happens to the cooling flow in the interior of a galaxy during and after collision.

A closer look at the COSMOS code itself reveals many of the characteristics that we have come to expect from modern high- performance software packages. Written in Fortran 90, the modular design of COSMOS allows it to handle a variety of different physical conditions as input, and thus to simulate a diverse sampling of phenomena.

Utilizing MPI (the Message Passing Interface), COSMOS's parallelization takes the form of an explicit domain decomposition, wherein each computer processor carries out calculations on a distinct portion of the problem space. The decomposition appears to be fairly straightforward for the hydrodynamics portion of the code, and communication among processors is required at the beginning of each time step. While the mesh parallelization is more complicated and some tricky load-balancing issues are inherent to the problem, communication overhead isn't an incredibly important issue. As with most classical dynamics studies, the most computationally expensive piece of the simulation is the force calculation, making communication costs small by comparison.

Next week, we'll look at some of the methods employed in COSMOS, as well as some of the simulation results. In the meantime, the COSMOS home page is

http://flash.uchicago.edu/~ricker/resear...

The copyright of the article Simulating Galaxy Cluster Collisions in Scientific Computing is owned by . Permission to republish Simulating Galaxy Cluster Collisions in print or online must be granted by the author in writing.