INCITE Project Reports Unprecedented Full-Star Simulations

One of three computationally intensive large-scale research projects selected under DOE’s Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program has achieved unprecedented simulations of stars and supernovae using NERSC’s computing resources.

Called “Thermonuclear Supernovae: Stellar Explosions in Three Dimensions,” the project is led by Tomasz Plewa of the Center for Astrophysical Thermonuclear Flashes at the University of Chicago and is a collaboration between scientists at the university and at Argonne National Laboratory.

The group, which was awarded 2.7 million processor hours at NERSC, is studying the longstanding problem of thermonuclear flashes on the surfaces and interiors of compact stars. These phenomena are not only fascinating in themselves, but are also important for the light they shed on other fundamental questions in astrophysics: X-ray bursts for what they tell us about the masses and radii of neutron stars; classical novae for the contribution they make to the abundances of intermediate-mass elements in the galaxy, and for what they say about how the masses of white dwarfs change with time in close binary systems; and Type Ia supernovae for the contribution they make to the abundances of intermediate mass and heavy elements in the galaxy. Type Ia supernovae are also important for their crucial role as standard candles in determining the Hubble constant. Between January and March of 2004, the group used 168,887.59 CPU hours, just over 6 percent of the total allocation, as they tested their applications. The group, according to researcher Timur Linde, expects to typically use 512 to 1024 processors, then use 4,096 or more processors for their final calculations.

Although only in their testing phase, the group has already achieved significant results.

“We have calculated three Type Ia supernova explosion models, including one octant model with 8 km resolution,” Linde reported. “This was our test run on Seaborg, but it is worth mentioning that this run would already be state of the art in this field.”

The group has also calculated one full-star model with 8 km resolution and 30 km ignition regions and one full-star model with 8 km resolution and 50 km ignition regions.

“These last two runs are the first of this kind ever calculated  no group other than ours has produced full-star simulations before,” Linde said. “We are now moving to 4 km and 1km resolution models, which will require at least 512 processors and possibly the whole system at times.”