(Cassiopeia – Autumn/l’automne 2016)
By Sahar Rahmani
Thesis defended on August 23, 2016
Department of Physics and Astronomy, Western University
Thesis advisor: Dr. Pauline Barmby
Understanding the process of star formation is one of the key steps in understanding the formation and evolution of galaxies. In this thesis, I address the empirical star formation laws, and study the properties of galaxies that can affect the star formation rate.
The Andromeda galaxy (M31) is the nearest large spiral galaxy, and therefore, high resolution images of this galaxy are available. These images provide data from various regions with different physical properties. Star formation rate and gas mass surface densities of M31 have been measured using three different methods, and have been used to compare different star formation laws over the whole galaxy and in spatially-resolved regions. Using hierarchical Bayesian regression analysis, I conclude that there is a correlation between surface density of star formation and the stellar mass surface density. A weak correlation between star formation rate, stellar mass and metallicity is also found.
To study the effect of other properties of a galaxy on the star formation rate, I utilize an unsupervised data mining method (specifically the self-organizing map) on measurements of both nearby and high-redshift galaxies.
Both observed data and derived quantities (e.g. star formation rate, stellar mass) of star-forming regions in M31 and the nearby spiral galaxy M101 are used as inputs to the self-organizing map. Clustering the M31 regions in the feature space reveals some (anti)-correlations between the properties of the galaxy, which are not apparent when considering data from all regions in the galaxy. The self-organizing map can be used to predict star formation rates for spatially-resolved regions in galaxies using other properties of those regions.
I also apply the self-organizing map method to spectral energy distributions of high-redshift galaxies. Template spectra made from galaxies with known morphological type are used to train self-organizing maps. The trained maps are used to classify a sample of galaxy spectral energy distributions derived from fitting models to photometry data of 142 high-redshift galaxies. The grouped properties of the classified galaxies are found to be more tightly correlated in mean values of age, specific star formation rate, stellar mass, and far-UV extinction than in previous studies.