Jupiter, the brightest point in a recent moonless night sky, faded to a pale milk drop as it rose alongside the Tribute in Light that soared from the World Trade Center site during March of 2002. The sentiment of the memorial was touching, but the bright columns, along with the city's multitudes of other glittering lights are an astronomer's bane. Yet on the Upper West Side of Manhattan astrophysicist Mordecai-Mark Mac Low is peering deep into faint gas clouds and nebulae without so much as a squint.

Astronomers routinely look deep into time - all of the way to the background radiation they believe was left by the Big Bang. But it's only inside supercomputers that researchers can fiddle with get an intimate look at the kinds of objects we see thousands of light years from Earth, and even experiment upon them by fiddling with the fundamental forces that shape them.

Specifically, Mac Low is using an enviable digital toolbox to solve a cosmological riddle: in galaxies like our own, a fraction of the original hydrogen and helium hasn’t yet collapsed to form stars. In places as familiar as the Orion Molecular Cloud, which the Orion Nebula peeks out of, new stars are being born at this moment. That's astonishing when you consider that by sheer gravity, those gas clouds should have collapsed in the first few hundred million years of the universe. Yet about 10 billion years later we creatures formed mainly from elements fused by long dead and recycled stars are still observing star formation involving that original gas.

The reason for this suspended evolution is that something is jostling matter around too much for gravity to pull it together densely enough to ignite a star, but what? "Stuff is thrashing around but we don't know what is putting energy in to keep these gases moving," Mac Low explains.

If the paleontologists have their barosaurus skeleton and the marine biologists have their 94' long fiberglass blue whale, the monster in the Museum closest to Mac Low's heart is the GRAPE 6 supercomputer cluster. Designed by Jun Makino of the University of Tokyo and Piet Hut of the Institute for Advanced Study in Princeton (and a Museum research associate), the GRAPE 6 is designed to crunch the numbers needed to simulate in three dimensions the motions of millions of interacting gravitational sources - imagine colliding galaxies, evolving star clusters, or tempest-tossed molecular clouds - at once.

In fact, the name "GRAPE" is an acronym for "gravity pipeline." The Museum now has three of these machines, plus a fourth prototype board that together can perform three and one-half teraFLOPS, matching 3,500 one-gigahertz Pentiums IIIs. (A "FLOP" is a floating-point operation, a calculation that deals with decimal points, rather than simpler binary integers.)

It is little wonder Mac Low needs such killer hardware. He's trying to calculate hypersonic (Mach 5 to Mach 50), turbulence of gas in clouds light years across over millions of years in increments of 10,000 years. In his largest models, volume is represented by a three dimensional grid of 134 million boxes. Each box affects its neighbors by having higher or lower pressure than the surroundings, so that gas flows into or out of the other boxes surrounding it.

"We won't get every wiggle in the flow but we can get a look at where there's turbulence, and integrated quantities like the average velocity," Mac Low explains. That data, however, was enough to loose a tremor beneath prevailing theory.

The consensus for decades was that magnetic field lines stretched like rubber bands across the partially ionized clouds smoothed hypersonic shocks into waves that did not lose energy, keeping that turbulence alive, as well as providing enough repellant force to overcome the gravitational attraction of the gas and slow or prevent star formation.

"We went in reading papers and thought, 'Oh yes, we've got new tools to confirm this and were actually really surprised," Mac Low says. "We pumped energy into our boxes and turned on self gravity," he says of his simulation. "We found that magnetic fields did not reduce how much energy was lost. We haven't ruled out magnetic fields but we explored if they are necessary, and to our surprise found that they weren’t. Meanwhile the observers are telling us that the preponderance of evidence is that magnetic fields are a little bit too weak to support big clouds against collapse."

So what else could be preventing the collapse? One candidate is the shear caused by our galaxy's rotation, but Mac Low questions if this mechanism has as much energy as others, though it might be important in galaxies where nothing else is going on. Some have posited that gravitational collapse itself might drive turbulence, and that would produce a feedback loop that keeps star formation in check, but later studies showed the energy of that motion dissipates too quickly to cause the enduring phenomenon observed. An exotic explanation might be that parsecs-long jets ejected from low mass protostars forming within the clouds is roiling up the region too much for other stars to form, but Mac Low doubts that turbulence can be "driven by sticking needles into molecular clouds."

The horse Mac Low's betting on is the brute force of supernovae pumping energy into the system. The exploding stars would also explain some of the "pollution" of heavier elements observed in spectral lines coming from the clouds, Mac Low adds.

In the simulations Mac Low diminished the magnetic field strength and then removed it entirely to emphasize his point - the clouds didn't act much differently. His team is confident in the fidelity of its model because he checked it against experiments with gas in his laboratory and analytic solutions to the equations of gas dynamics. More than that, what he was seeing on his computer screen appears consistent with what astronomers are seeing through their telescopes.

"Turbulence was spread across the box but it actually did not support gas against collapse everywhere. Little bits of the box were collapsing even though most of the box was not, which is what we see in most places and that was really exciting," he recalls. "This issue is now a subject of great controversy. For years astronomers like Richard Larson, who said magnetic field lines wouldn't do the trick, were voices in the wilderness but they didn't have numeric models to support them," Mac Low explains. "Now many people are taking this very seriously. We hope we're in the middle of a framework shift."