Supercomputer models of the formation and evolution of galaxies

In this thesis, we explore the use of semi-analytic galaxy formation models and related techniques as a means to investigate the physics of galaxy formation and evolution. We begin by investigating the ability of a highly cited semi-analytic model to reproduce the evolution of the observed galaxy population over the last 7 billion years of cosmic time. This is achieved by carrying out a detailed statistical calibration of the model's free parameters in order to simultaneously reproduce the observed galactic stellar mass function at z=0 and z≈0.8, as well as the z=0 black-hole bulge relation. In order to be successful, we are required to push the parameters of the model associated with supernova feedback to implausibly high values, suggesting that the current implementation of this physical prescription may be inadequate. Additionally, we suggest that some extra mechanism is required to preferentially increase the efficiency of star formation in the most massive galaxies at high redshift. In order to further explore the utility of semi-analytic models, we then present their novel use as a tool to investigate the current evolutionary status of the Milky Way and M31. The Milky Way is the most closely studied galaxy in the Universe. However, to understand the Milky Way's place in the broader landscape of galaxy evolution, we require a baseline population of galaxies against which to compare. Using a sample of analogue galaxies drawn from both observational data and semi-analytic models we find that both the MilkyWay and M31 may be 'green valley' galaxies undergoing an important evolutionary transition. Furthermore, using the histories of our model analogue sample, we investigate the possible physical mechanisms which could be driving this evolutionary change. Finally, we introduce a new, self-consistent model for connecting the growth of galaxies to the formation history of their host dark matter halos. This model dispenses with attempts to implement all of the dominant baryonic processes associated with galaxy evolution and replaces them with two simple, phenomenologically motivated equations that depend only on a single host halo property. We demonstrate the ability of our new formation history model to reproduce the observed red and blue stellar mass functions at z=0. Then, by adding a simple redshift dependence to the parameterisations, we show that it can also successfully match the observed global mass functions out to z=4. To conclude, we highlight the advantages of this model over the relevant alternatives, as well as highlight its general utility.