The Aquila comparison Project: The Effects of Feedback and Numerical Methods on Simulations of Galaxy Formation

We compare the results of thirteen cosmological gasdynamical codes used to simulate the formation of a galaxy in the LCDM structure formation paradigm. The various runs differ in their hydrodynamical treatment (SPH, moving-mesh and AMR) but share the same initial conditions and adopt their latest published model of cooling, star formation and feedback. Despite the common halo assembly history, we find large code-to-code variations in the stellar mass, size, morphology and gas content of the galaxy at z=0, due mainly to the different implementations of feedback. Compared with observation, most codes tend to produce an overly massive galaxy, smaller and less gas-rich than typical spirals, with a massive bulge and a declining rotation curve. A stellar disk is discernible in most simulations, though its prominence varies widely from code to code. There is a well-defined trend between the effects of feedback and the severity of the disagreement with observation. Models that are more effective at limiting the baryonic mass of the galaxy come closer to matching observed galaxy scaling laws, but often to the detriment of the disk component. Our conclusions hold at two different numerical resolutions. Some differences can also be traced to the numerical techniques: more gas seems able to cool and become available for star formation in grid-based codes than in SPH. However, this effect is small compared to the variations induced by different feedback prescriptions. We conclude that state-of-the-art simulations cannot yet uniquely predict the properties of the baryonic component of a galaxy, even when the assembly history of its host halo is fully specified. Developing feedback algorithms that can effectively regulate the mass of a galaxy without hindering the formation of high-angular momentum stellar disks remains a challenge.