INDUSTRY

By Karen Green, NCSA Public Information Officer -- Massive computer simulations are giving scientists clues to the nature of dark matter -- Think of today's astrophysicists as cosmic detectives. As better instruments make it possible to look at clusters of stars and galaxies billions of light years from earth, astrophysicists are able to see back in time to when the universe was relatively youthful. With this vision, they can compare their observations with theoretical simulations that attempt to explain cosmic events in the early universe. These glimpses of the past, as well as the computer simulations, are clues that will hopefully lead to answers to some of the most fundamental questions of science: Why is cosmic material dispersed in clumps rather than distributed evenly across the universe? What is the nature of that mysterious, unseen matter that makes up most of the universe's mass, commonly referred to as dark matter? Paul Bode, a research scientist at Princeton University, and Jeremiah Ostriker, a Princeton astronomy professor and a member of the Alliance Cosmology team who also holds the Plumian Professorship of Astronomy and Experimental Philosophy at the University of Cambridge, concern themselves with the latter two questions. They have so far logged 25,000 hours on the Alliance's Platinum Linux cluster at NCSA in an effort to describe the properties of dark matter in ways that fit logically with what other astronomers have observed in the universe. Scientists have postulated the existence of dark matter for more than 25 years, even though they have never been able to see it. In the observed universe, galaxies and clusters of galaxies spin fast -- so fast that they must contain more matter than what can be seen. The outer portions of spiral galaxies rotate around their galaxy centers with such speed that they would fly apart if only their visible matter were holding them together. Likewise, galaxies inside galaxy clusters move in relation to each other at speeds that are faster than what would be induced by the gravity of the galaxies' visible matter. There simply must be more matter, undetectable to human optics, holding together these spinning disks, scientists theorize. They coined the term dark matter to describe this invisible mass. Dark matter is believed to surround galaxies in invisible halos. The exact nature of dark matter, including its exact mass, how it interacts with other particles, how fast it moves, and how it obeys the general laws of physics, are questions that still need to be answered. Bode and Ostriker believe that the key is to figuring out the characteristics dark matter would need to create a universe that looks like the one we see. "We may think of these very large cosmological simulations as a means of test driving new fundamental theories of physics," explains Ostriker. "We put into the computer some new, specific but speculative theory, start with initial conditions given by microwave observations of the deep past, and then calculate forward. The test is to see if the final computed universe looks like the real one." "One of the questions we have is: How dense and how energetic would dark matter need to be in order for the universe to be the way it is?" adds Bode. "The density of the universe is not uniform, and we believe dark matter provides the extra gravitational pull to attract more matter together into the dense clumps of galaxies that we observe." The cold, the warm, and the hot Current theory describes dark matter in three ways. Cold dark matter (CDM) is highly interactive gravitationally with other matter and has no relativistic speed. That is, it is not moving at a percentage of the speed of light. Computer simulations show that CDM results in a universe much like the one we observe if the simulation is done on a very large, multigalactic scale. However, at smaller scales CDM doesn't fare quite as well, creating simulated galaxies with too many small dense clumps of matter. "Cold dark matter predicts many dwarf structures (small, rogue galaxies) between the large clumps in a galaxy," says Bode. "That is not what has been observed. " Another possible form of dark matter -- hot dark matter -- describes a particle like massive neutrinos, which are so weakly interactive that they pass right through regular matter. Scientists have been able to detect neutrinos using huge particle detectors, and neutrinos are believed to be a type of dark matter. However, if most dark matter were neutrinos, matter in the universe would be quite uniformly distributed, explains Bode, and researchers know that it is not. Ostriker and Bode simulate the formation of galaxy halos using the warm dark matter (WDM) model. WDM is a slight variation of CDM. Because particle velocities are lower, dark matter clumps more with its surroundings than do neutrinos but not as much as CDM particles. Recreating gravity's effects The research team's simulations started with a cube of WDM particles initially distributed almost uniformly across a theoretical portion of the universe. For the Linux cluster simulations, this cube consisted of 17 million particles in a grid with 512 cells on each side. The grid area was a randomly picked volume of space, 20 megaparsecs per side. An average galaxy in this grid would be made from a volume of three megaparsecs and would occupy about 10 kiloparsecs. The researchers use Tree Particle Mesh (TPM) code to calculate the gravitational force that the particles have on each other over a series of time steps. Over time, particles are attracted to each other and fall toward each other to form dark matters halos. Galaxies form within these dark halos, explains Bode. The 17 million-particle run took eight days using 128 processors full time. Early in 2002 the team began another set of simulations using 134 million particles in a billion-cell grid. That run is expected to take two months using 128 processors continuously. The TPM algorithm gives Bode and Ostriker a mathematical shortcut for looking at the gravitational force the WDM particles exert on each other. In a world of unlimited computing power, the most straightforward way to compute this force would be to calculate Newton's law of gravitation on each pair of particles. However, since the current computations involve roughly 10^16 pairs of particles -- that's 10 quadrillion pairs -- such calculations are impractical if not impossible. TPM breaks this massive problem down into many smaller ones. It uses a tree code to calculate short-range gravitational forces (forces within the many halos produced by gravity). Then it employs a fixed particle mesh to calculate long-range gravitational forces (forces exerted from outside the halos, including forces in the voids between halos). "Preliminary indications are that the warm dark matter model could provide a better description of the universe as we've actually observed it compared to the cold dark matter model, in which case it is a clue to what dark matter is," says Bode. "What exactly it is, we still don't know." The dark matter mystery is still unsolved, but, thanks to more detailed simulations, the clues are coming quickly. With a little more detective work, some answers should be revealed. This research is supported by the National Science Foundation and the National Computational Science Alliance. Relevant URLs --Access story: http://access.ncsa.uiuc.edu/Stories/wdm/ --Research website: http://astro.princeton.edu/~bode/