Phenomenological models of galaxy formation

Galaxy formation involves a diverse range of physical processes, including:

  1. the gravitational collapse of dark matter to form galaxy haloes,
  2. cooling of gas in haloes by emission of radiation,
  3. star formation in the cold, dense molecular gas at the centres of galaxies,
  4. reheating and ejection of gas by supernova explosions and active galactic nuclei (AGN),
  5. merging and tidal interactions between individual galaxies and
  6. the spectral evolution of stellar populations over time.
Figure: Galaxy formation model

Flowchart illustrating the ingredients of a typical galaxy formation model.

Phenomenological (or semi-analytic) models attempt to bring together all of these processes in a single, coherent model (see, e.g. the flowchart to the right), with the power to make predictions for statistical samples of galaxies.

The evolution of the dark matter distribution is well understood through the use of cosmological N-body simulations but, since we can't observe it directly, if we want to test models of the Universe we need some way to link the underlying dark matter distribution to the galaxies we can observe. By populating haloes with observable galaxies, phenomenological models provide the missing link between our understanding of the processes operating in galaxies and galaxy formation in a cosmological context.

Figure: Galaxies in an N-body simulation

Galaxies in an N-body simulation.

In our model, the sites of galaxy formation are identified in a series of ‘snapshots’ of an N-body simulation using a group-finding algorithm to identify groups of particles representing collapsed overdensities¹. By following these haloes between snapshots we build up a ‘merger tree’ for each one. Simple, phenomenological rules for gas dynamics, star formation, etc. are used to predict the properties (magnitudes, colours, sizes, metallicities, etc.) of the galaxies that form. The image on the left shows the result of this process: the dark matter is shaded grey, while the galaxies are shown as coloured dots, with the colour being related to (but not of course the same as) their actual colours and size in proportion to their luminosities.

The major advantage of phenomenological models over high-resolution N-body simulations is that is that they require fewer resources, enabling a large region of parameter-space to be explored with high efficiency.

¹ A common alternative is to use Monte-Carlo simulations of the extended Press-Schechter formalism to generate galaxy halo merger trees. N-body simulations have the advantage that they give information not only on the galaxies themselves, but also their environments and distribution relative to each other.