Dark Matter Evolution

Both arms of Bertschinger's GC3 work bear on what is probably the key question in 1990s cosmology: dark matter. Galaxy mapping and other measurements over the last two decades have raised a perplexing problem. Luminous matter -- the visible stars and planets -- doesn't have enough mass, hence enough gravity, to prevent galaxies from flying apart. This leads cosmologists of many stripes to conclude that most of the universe, probably more than 90 percent, is unseen -- dark matter.

Various theories have been proposed, suggesting different kinds of not yet discovered particles pervading the universe. Until these particles are detected in laboratories, the only way to test the theories is cosmological modeling. The basic idea is to posit an initial configuration of matter and let the big bang happen until, billions of years later in the computer, galaxies form. One can then compare the structure produced by one theory against others and with reality.

The computational challenge is the incredible range of scales between galaxies, about 10,000 light years across, and the visible universe, at least 10 billion light years. One approach is to model a volume smaller than the observable universe, which can still give potentially useful results. Taking this path, Bertschinger's dark matter simulations on the Connection Machine, CM-5, at the National Center for Supercomputing Applications evolved 16 million particles starting less than a hundred million years after the big bang. The results show that the character of structure formed -- whether dense clumps or filament-like chains -- depends on the nature of the initial fluctuations. This modeling also suggests that a purely "cold dark matter" scenario -- one of various dark matter theories proposed -- forms too much structure in comparison with the real universe.

"The T3D opens up new possibilities," says Bertschinger, who is working to port his dark matter evolution code to Pittsburgh's T3D. "The long-term goal of our project, besides trying to answer these questions, is developing a portable, scalable parallel cosmology evolution code that includes everything from soup to nuts, from microwave background to galaxy formation and particle and gas dynamics. We're making substantial progress."

go back to the main page