Similarly, the University of Washington researchers first simulate the universe from primordial lumpiness to the present, about 15 billion years later. In this first simulation, the universe is represented by 47 million particles moving within a 3D grid according to the laws of gravity. Each side of the grid represents about 3 million light-years although, as with the real universe, the size of the universe increases--that is, the grid gets bigger--with the passage of time.

Although 47 million particles may seem like a lot, they provide only a coarse representation of the universe. "Each particle is about size of a galaxy," Quinn says. "We can't follow the details of galaxy formation." The simulation is detailed enough, however, to pick out large clusters and voids. Even at this low level of detail, one run of this coarse universe simulation consumes more than 15,000 CPU-hours on the SGI Origin2000 supercomputer at NCSA.

A second simulation focuses on one of the clusters seen in the first simulation, with the rest of the universe fuzzed out. "Where the cluster forms I throw down many more (smaller) particles," Quinn says. That allows researchers to see the finer structures of a cluster that were blurred out in the bigger view.

Although the second simulation uses fewer particles--5 million--and focuses on a smaller region of space, it is just as computationally challenging. The detailed structure in the clusters form much more quickly, which means the simulation has much smaller timesteps. Smaller steps mean that more steps cover the same period of time and the simulation takes longer to run.

"Now we're talking a few hundred particles making a galaxy," Quinn says. "A cluster in our simulation looks like a cluster in the sky. We'll have simulations in which we'll be able to see galaxy distribution."

How realistic are their simulations? A poster-size image of the sky hanging on the wall of Lake's office looks as if it had been shot through a telescope. "We have people come in and they say, 'Is it a picture of the sky?'" Lake says. "Nobody has ever done simulations of galaxy clusters that have caused people to ask that before."

In one sense, all cosmological models are quite simple: there's basically one force that drives the formation of everything from stars to galaxies to clusters of galaxies. That force is gravity, the same garden-variety gravity that pulls apples down from trees.

What astronomers don't precisely know is how much mass is in the universe, the amount and type of dark matter--the unseen 90 percent of the universe whose gravity keeps galaxies from flying apart--and the strength of the cosmological constant where empty space literally pushes against itself. In each simulation run, the researchers vary the mass, dark matter, and the cosmological constant so that each variation produces different results.

For each run the researchers also calculate several statistical measures, such as the number of different-sized clusters and how close the clusters are to each other. They will be able to calculate these measures from the Sloan data as well. Presumably the simulation run that mostly closely matches the Sloan numbers is the one that contains the most nearly correct values for the universe's mass, dark matter, and cosmological constant.