Multi-Scale Supercomputer Simulations of the Birth of Stars
The process of star formation is crucial to cosmology and astrophysics. The properties of galaxies are determined by their stellar populations, specifically by the rate at which gas is converted into stars, and by the distribution of stellar masses. It is known that stars are formed by the collapse of dense gas clumps found within larger gas clouds, known as giant molecular clouds because they are composed primarily of molecular hydrogen. By observing such gaseous stellar nurseries, astronomers have discovered they are extremely complex. Their internal motions are supersonic and chaotic, a state of gas dynamics known as turbulence. The degree of complexity, resulting from the mutual interaction of magnetic fields, gravity, and supersonic turbulence, is such that no complete theory of star formation is available to date. The best way to tackle this problem is to use powerful supercomputers such as petascale system SuperMUC of LRZ to compute specific solutions of the fundamental physical laws.
A major challenge for these computer simulations is the huge range of scales in the problem. Gas in the Galaxy goes through a complex life-cycle known as the Galactic fountain: Massive stars explode as supernovae, driving shock waves that compress the gas into giant molecular clouds. Part of the hot supernovae gas leaks out of the disk into the Galactic halo, where it cools, condenses, and rains down on the disk. The turbulent motions of molecular clouds are the result of this energy injection by supernovae.
To study the evolution and fragmentation of clouds into stars, one must therefore model this large-scale energy sources. On the other hand, to follow the collapse of individual stars, it is necessary to zoom in to a very small scale. The main goal of this project, which was made possible through the Partnership for Advanced Computing in Europe, PRACE, was to develop supercomputer simulations that could cover the whole range of scales, describing the Galactic fountain, while at the same time modelling the formation of each individual star.
The international team of scientists achieved their goal by using a computational method known as “adaptive-mesh refinement”, which consists of focusing the computational resources on the most interesting regions. The large-scale Galactic-fountain flow is described at relatively low resolution, while the regions where dense cores collapse into stars are computed at much higher resolution (meaning with many more and much smaller computational cells). This method is very difficult to implement in large supercomputers, so part of the challenge consisted in developing a numerical code that could take advantage of the huge number of processors available in modern supercomputers.
The scientists were able to show that the mass distribution of massive stars is a natural result of turbulent flows driven by supernova explosions, and that the rate of star formation is comparable to what is observed in actual galaxies. The results of the simulations are also used by astronomers to carry out detailed comparisons with observations of interstellar gas obtained with space telescopes such as Herschel and Planck.
The project was a collaboration involving several research centers, primarily the Institute of Cosmos Sciences of the University of Barcelona and the Niels Bohr Institute of the University of Copenhagen.
Instituto de Ciencias del Cosmos (ICC), Universidad de Barcelona