Simulating the Universe with 100,000 cores

shapeimage_2
How do supermassive black holes form, grow and interact with their environments? What drives the X-ray spectral state changes in stellar mass black hole systems? How do planets form in accretion disks around young stars? The basic physics at work in all of these systems is that of gas dynamics in combination with magnetic fields, gravity, dust and radiation. The tightly coupled equations describing these processes are inherently multi-dimensional and non-linear, so while some insight can be gained into the physics at work using analytic techniques, numerical study of these systems is essential. Even so, the astrophysical phenomena in question are too complex to be attacked head-on using the largest supercomputers. Formation of supermassive black holes in the early universe through (for example) direct collapse takes place on scales from that of massive protogalaxies to sub-parsec scales. Accretion disks in Active Galactic Nuclei (AGN) extend from the black hole event horizon to that of the Bondi radius, similarly, relativistic jets launched from these same AGN extend from the black hole event horizon to size scales larger than that of the host galaxy. In Protoplanetary Disks, collagulation of dust on size scales much smaller than the thickness of the disk (through e.g. the streaming instability) plays an essential role in forming the building blocks of planets, while the evolution of these building blocks is determined by turbulent eddies within the disk on scales comparable if not greater than the disk scale height. Numerical investigations of astrophysical phenomena must therefore be broken down into manageable pieces, each piece being carefully designed by the scientist, executed using well-tested algorithms and only then analyzed in order to understand the physics at work. This is the realm of the computational astrophysicist.