For the first time, a team of astronomers has succeeded in modeling in three dimensions the behavior of a massive star during the final seconds of its life, as it collapses, rebounds, and initiates a supernova explosion. Michael Warren and Chris Fryer (Los Alamos National Laboratory) reported at the American Astronomical Society meeting in Albuquerque that this simulation, which required a month of processing time on the world's fourth-most-powerful computer (an IBM RS/6000 SP), provides important confirmation of the basic mechanism of the explosion, which had previously been modeled only in two dimensions and with a simulation using far fewer particles. Because of the limitations of the earlier modeling, theorists had had little confidence in the relevance of its results.
In the earliest, one-dimensional simulations of a standard core-collapse (Type II) supernova, carried out in 1966 by Stirling Colgate and Richard White (then at Lawrence Radiation Laboratory), the rebound of the stellar material after its initial collapse usually "stalled" before it could actually produce an explosion — a fatal flaw in the simulations. Much later, in 1994, Willy Benz (then at the University of Arizona) and Marc Herant (Los Alamos), with Colgate and Fryer, were able to carry out two-dimensional simulations that reliably produced explosions. They concluded that the crucial difference was the convection process that allowed for a mixing of the infalling and the expanding material, preventing the problem that had stalled the one-dimensional models.
But three-dimensional behavior, especially of complex processes like convection, can often lead to very different outcomes, so the two-dimensional simulations were still not considered reliable indicators of supernova dynamics. "Modeling the collapse of a massive star represents one of the greatest challenges in computational physics," Warren said. Now, the 3-D simulations carried out by the Los Alamos team have provided a significant confirmation of the earlier 2-D results, agreeing to within 10 percent on the explosion energy, time scale, and the mass of the remaining neutron star. The simulations were carried out using 300 parallel processors and a smoothed-particle hydrodynamics model involving 3 million simulated particles rather than a fixed computational grid.
The initial runs, like the earlier simulations, were based on the idealized case of a nonrotating star — partly to ensure that the results could be compared directly with the earlier ones. Now that the model has passed that test, the next step will be to simulate the more realistic case of a rotating star, and then to add additional parameters such as a more realistic representation of the varying gravitational effects due to density fluctuations as the star's mass becomes disrupted. "This result is a milestone," said Adam Burrows (University of Arizona). "It’s a foretaste of what is to come." The new simulations, he added, should produce sufficiently detailed results to allow for direct comparison with ongoing studies of Supernova 1987A in the Large Magellanic Cloud and detailed predictions of the gravity waves that should be seen when the Laser Interferometer Gravitational-wave Observatory (LIGO) begins full operation.