In 1967, a Cambridge University graduate student poring over data from a newly constructed radio telescope noticed something very odd–a radio signal blinking regularly from a far corner of the sky.
There was no natural source known to produce such a signal. Early explanations varied, with possible sources ranging from interference by television signals to communications from extraterrestrial beings. The facetious name given at first for this odd, regular blip, in fact, was LGM-1, with the LGM standing for “little green men.”
The signal, it turns out, did not come from space aliens; it came from a pulsar, the spinning remnant of a great stellar explosion. Pulsars are the enormously massive–smaller than the moon, heavier than the sun–spinning leftovers from core-collapse supernovas, the cataclysms that provided most of the elements on earth and made our own lives possible.
The leftover core is known as a neutron star, and a neutron star that spins is known as a pulsar. A pulsar appears to blink because radiation shoots out of its magnetic poles, which, as with the earth, can be tilted a little from its axis of spin. As a result, the pulsar behaves like a stellar lighthouse, pointing at an observer once with each rotation.
Nearly four decades later, a team of scientists using Oak Ridge National Laboratory (ORNL) supercomputers has discovered the first plausible explanation for a pulsar's spin that fits the observations made by astronomers. Their surprising results show that the spin of a pulsar is not just a continuation from the massive star that preceded it; in fact, the pulsar spin can be in the opposite direction.
Anthony Mezzacappa of ORNL and John Blondin of North Carolina State University explain their results in the January 4, 2007, issue of the journal Nature. According to three-dimensional simulations they performed at the Leadership Computing Facility, located at ORNL, the spin of a pulsar is determined not by the spin of the original star, but by the shock wave created when the star's massive iron core collapses.
That shock wave is inherently unstable, a discovery the team made in 2002, and eventually becomes cigar-shaped instead of spherical. The instability creates two rotating flows–one in one direction directly below the shock wave and another, inner flow, that travels in the opposite direction and spins up the core.
“The stuff that's falling in toward the center, if it hits this shock wave that is not a sphere any more but a cigar-shaped surface, will be deflected,” Mezzacappa explained. “When you do this in 3-D, you find that you wind up with not only one flow, but two counterrotating flows.”
The discovery comes at an opportune time, because astronomers did not have a workable explanation for how the pulsar gets its spin. The assumption to this point has been that the spin of the leftover collapsed core comes from the spin of the original star. Being much smaller, the pulsar would then spin much faster than the original star, just as a figure skater spins faster by pulling his or her arms in.
The problem with that approach is that it would explain only the fastest observed pulsars. The ORNL team, on the other hand, predicts spin periods that are in the observed range between 15 and 300 milliseconds.
“Our discovery came at a critical time,” Mezzacappa noted. “It came at a time when there was no description in the literature of how neutron stars are spun up and, therefore, how pulsars are born, that are consistent with observation. It was a crisis, if you will. Now our simulations come along and provide a way around that conundrum.”
The discovery is an outgrowth of the team's use of three-dimensional simulations and the advances in high-performance computing that made the simulations possible. The simulations performed for the Nature paper used the Cray X1E system at ORNL, known as Phoenix. That system boasts a peak performance of more than 18 teraflops and is currently the fastest vector computer in the United States. Later simulations done by the team made use of the center's Jaguar system, a Cray XT3 with a peak performance of more than 50 teraflops.
The counterrotating spins found by the ORNL team are not found in two-dimensional simulations, Mezzacappa noted, because two-dimensional simulations assume that the star, its core, and the shock wave are symmetrical. In particular, the two-dimensional simulations assume that the star is axisymmetric, meaning that if you take a slice through the center of the star from top to bottom, it doesn't matter where you take the slice, because the slices will all look the same.
“If you have a two-dimensional simulation, you're assuming some symmetry in the problem, Mezzacappa explained. “These counterrotating flows are the result of breaking that symmetry, which can happen in 3-D because you don't impose any such symmetry.
“When you really open up the degrees of freedom the way Mother Nature does, you discover new phenomena that were not otherwise possible because the previous models were constrained.”
The team used the VH1 code, developed by Blondin when he was a postdoctoral research associate at the University of Virginia. The largest of the simulations involved 1,250 grid points in each of three dimensions, or nearly 2 billion zones. It generated 40 terabytes of data, enough to fill more than 50,000 CDs.
At the time, the amount of data presented a challenge, Mezzacappa noted, because Blondin was performing the visual analysis in North Carolina.
“When it came time to move the simulation data to North Carolina State in order to render it for visual analysis and discovery, it became really problematic,” he said. “There was no discovery until we began working with networking experts to effectively move the data to North Carolina State. Then we had to create an analysis and visualization capability at North Carolina State, both in terms of hardware and software, to be able to complete the process.”
Mezzacappa noted that researchers are able now to perform visualizations remotely, without having to move the data off site, but at the time of their early three-dimensional simulations this capability was not in place.
He stressed also that the team is looking forward to further advances in high-performance computing that will be coming to ORNL. For example, the team's simulations have not incorporated the influence of nearly massless, radiation-like particles known as neutrinos and the star's magnetic field.
The real prize, though, for his and other teams is a complete explanation of how the collapse of a star's core leads to the explosion that ejects most of its layers. So far, that explanation has proved elusive.
“That is one of the most important unsolved problems in astrophysics,” Mezzacappa said. “Core-collapse supernovae are the dominant source of elements in the universe, and the mechanism has everything to do with that.”
Beyond the explosion mechanism, they are looking forward to using simulation to explain other aspects of a supernova, such as the elements produced and the kick with which the explosion sends the leftover neutron star flying through space.
“Basically,” he said, “to simulate core-collapse supernovae with all of the physics described with sufficient realism and detail, the threshold computing power needed is 1 petaflops. But ultimately to be able to predict the elements that are produced in the supernova in detail and make predictions for all of the supernova-associated observables, one needs sustained petaflops capabilities. That requires a multi-petaflops system.
“In a nutshell, this rapid advance in supercomputing technology will give us the tools to solve this problem and to make these important predictions and to understand these events and their role in our universe. This is a very, very exciting and very satisfying thing.”
Source: Oak Ridge National Laboratory