The Flash Code was developed at the University of Chicago under the DOE ASCI Strategic Alliances Program to study the problem of thermonuclear flashes in white dwarfs and neutron stars. Astrophysical events to be studied include X-ray bursts, novae, and Type Ia supernovae. The code, which is written mostly in Fortran 90, solves the equations of compressible gas dynamics (the Euler equations) using the Piecewise-Parabolic Method. Energy generation rates are computed by solving a nuclear reaction network. An equation of state module provides self consistent thermodynamics for a gas of ions, relativistically degenerate electrons/positrons, and radiation. The code also contains modules for thermal conduction and Poisson solvers for self-gravity. Improved spatial resolution and efficiency are obtained by using PARAMESH, an adaptive mesh refinement package. The framework of the code was designed to provide ease of use, portability across a wide variety of computer architectures, and modularity, so that additional physics packages are easily added. Flash is fully parallel using MPI and scales well to thousands of processors. Sustained performance of .25 TFLOPS on 6000 processors of ASCI Red has been attained.
Two examples of simulations performed with the Flash code will be presented. The first studies the propagation of a detonation through an accreted layer of helium on the surface of a neutron star. Results show the excitation of surface waves and the creation of complex turbulent flows in the neutron star atmosphere. This calculation is relevant to models of X-ray bursts. The second simulation shows the detailed structure of a detonation front propagating through a pure carbon gas in a Type Ia supernova explosion. In this case, the results clearly show the formation of a complex cellular structure and the non-uniform distribution of fuel and ashes behind the front.