Stan Woosley’s group is seeing Type 1a supernovae in a new light, thanks to two combustion codes, MAESTRO and CASTRO, developed with support from DOE’s Scientific Discovery through Advanced Computing (SciDAC) program.
MAESTRO is a unique, low Mach-number code. This means it’s ideal for modeling not the supernova bang but the sub-sound-speed simmering that for a century precedes stellar detonation and thus defines its conditions.
“With MAESTRO we can follow this simmering flow for hours,” Woosley says. “It can study the ignition phase much better than anyone has done before, or anyone else in the world can do right now.”
When the star’s ready to blow, Woosley’s team turns to CASTRO. It’s a compressible hydrodynamics code with Adaptive Mesh Refinement (AMR) that enables efficient use of petascale computational resources by focusing on the physics of greatest interest – in this case the nuclear flame front.
When a white dwarf ignites near its core, the flame itself is only about a micron thick (a human hair is about 50 microns thick) in a star about 2,000 kilometers in radius.
“Turbulence can stir up that very thin flame sheet and make it have very complex topology,” Woosley says. “We don’t resolve all of that topology, but we have to resolve enough of it to get the progress at the flame front right.”
The team’s latest runs with CASTRO on Oak Ridge National Laboratory’s Titan computer achieved a peak resolution of 100 meters using a billion zones, the maximum available with the team’s computational budget.