ITHACA, N.Y.---Scientists have successfully modeled the evolution of a massive cluster of X-ray galaxies using the IBM RS/6000 Scalable POWERparallel Systems (SP) at the Cornell Theory Center (CTC). Because X-ray bright clusters of galaxies are extremely rare in nature, computer simulation is one of the few ways to study them. With CTC's SP, computational astrophysicist Paul Bode (Physics, MIT) and astronomer Renyue Cen (Astrophysical Sciences, Princeton University) are now able to explore the evolution of these large-scale structures in the universe with simulations of unprecedented complexity and resolution.
"While previous, lower-resolution simulations (based on 2703 points) provided some insights into the properties of X-ray galaxy clusters in different cosmological model universes, a minimum dynamic range of 5123 is required to start making quantitative statements," says Cen.
Increased computational capabilities have allowed the researchers to expand their methodological approach. Cen and Bode are able to incorporate both gravity and hydrodynamics in their model to simulate the collapse of a high-density region contained in a cube of 256 million light years on a side. A total of 134 million fluid elements and 18 million dark matter clouds is used to follow the motion of the two distinct yet interacting (through gravitational force) components from about 51 million years after the Big Bang to today (a period of 13 billion years).
According to Cen, until very recently the only method used to study the formation and evolution of complex, multiscale, multicomponent cosmic structures has been the N-body method, which deals only with the gravitational effect. Achieving accurate predictions of a cosmological theory that can be directly compared to the luminous observed universe requires detailed modeling of the dynamics of ordinary matter (the baryonic component) as well as of the dynamics of gravitationally dominant dark matter.
"Based on Fortran-77 plus message passing (MPI) code, our model follows collisionless dark matter (which interacts via gravity only) with the Particle-Mesh (PM) method, and it follows baryons (i.e., gas, affected by both gravity and gas pressure) with the Total Variation Diminishing (TVD) method," says Bode. "Basically the TVD code follows gas and dark matter over time, as matter collapses into galaxies and clusters of galaxies." Gas in the center of a cluster of galaxies is heated up to more than ten million degrees during the gravitational collapse making the cluster a very bright source in the X-ray spectrum.
The researchers have determined that the code speedup (wallclock time per number of points) scales almost linearly with the number of CPUs used, from 4 to 128 processors. They are still analyzing the output (about 24 gigabytes) of the simulation, which ran for 6 days on 64 SP nodes.
Cen and Bode are also examining the evolution of Lyman alpha clouds, clouds of neutral hydrogen gas that are identified as Lyman alpha absorption lines in the spectrum of high redshift quasars. "Doing the Lyman alpha simulations requires adding more microphysics to the gas," says Bode. In contrast to the X-ray clusters of galaxies, these clouds are very numerous and have been studied in detail by astronomers. Thus the researchers can make statistical comparisons with ground-based observations as well as observations from space, such as those by the Hubble Space Telescope. Earlier calculations of the Lyman alpha clouds with simulations based on 2883 points indicated that the researchers need a larger simulation box with finer resolution to ensure that they get both representative sampling and fully resolved structure. They expect that if they scale the simulation up to 7683 points and run it in parallel on the SP, they will be able to address such fundamental issues as the baryon content of the universe and the nature of the metagalactic ultraviolet radiation field.
Bode and Cen report that the scale of the simulations they are now conducting (using a mesh of a billion points) is unprecedented. "Soon we hope to begin a 10243-point simulation that should provide groundbreaking, accurate calculations of X-ray clusters of galaxies--only possible on the SP2 computer at CTC--that will enable us to discriminate between cosmological models when compared to observations." This model will use cosmic microwave background radiation (CMBR) simulation results from CTC user Edmund Bertschinger (Physics, MIT) to seed the evolutionary process out of the Big Bang. The CMBR results reproduce extremely small but important variations in the patterns of CMBR observed by NASA's COBE satellite that suggest an uneven distribution of the density of matter in the early universe.
Both Cen and Bode are members of the Grand Challenge Cosmology Consortium, one of the National Science Foundation's high-performance computing and communication (HPCC) Grand Challenge projects, that is devoted to harnessing the power of parallel computers to explore the origin of large-scale structure in the universe and the formation of galaxies.
CTC is one of four high-performance computing and communications centers supported by the National Science Foundation. Activities of the center are also funded by New York State, the Advanced Research Projects Agency, the National Center for Research Resources at the National Institutes of Health, IBM, and other members of CTC's Corporate Partnership Program.