We study the properties of supernova (SN) driven interstellar turbulence with a numerical magnetohydrodynamic (MHD) model. Calculations were done using the RIEMANN framework for MHD, which is highly suited for astrophysical flows because it tracks shocks using a Riemann solver and ensures pressure positivity and a divergence-free magnetic field. We start our simulations with a uniform density threaded by a uniform magnetic field. A simplified radiative cooling curve and a constant heating rate are also included. In this radiatively-cooling magnetized medium, we explode SNe one at a time at randomly chosen positions with SN explosion rates equal to and 12 times higher than the Galactic value. The evolution of the system is basically determined by the input energy of SN explosions and the output energy of radiative cooling. We follow the simulations to the point where the total energy of the system, as well as thermal, kinetic, and magnetic energy individually, has reached a quasi-stationary value. From the numerical experiments, we find that: i) both thermal and dynamical processes are important in determining the phases of the interstellar medium, and ii) the power index n of the B-p n relation is consistent with observed values.

The advent of robust, reliable and accurate higher order Godunov schemes for many of the systems of equations of interest in computational astrophysics has made it important to understand how to solve them in multi-scale fashion. This is so because the physics associated with astrophysical phenomena evolves in multi-scale fashion and we wish to arrive at a multi-scale simulational capability to represent the physics. Because astrophysical systems have magnetic fields, multi-scale magnetohydrodynamics (MHD) is of especial interest. In this paper we first discuss general issues in adaptive mesh refinement (AMR), We then focus on the important issues in carrying out divergence-free AMR-MHD and catalogue the progress we have made in that area. We show that AMR methods lend themselves to easy parallelization. We then discuss applications of the RIEMANN framework for AMR-MHD to problems in computational astophysics.