2011MNRAS.410.2137C

In order to use supernovae (SNe) as cosmological probes, a good understanding of their properties and diversity is necessary. Here we investigate whether principal component analysis (PCA) can be used to improve the calibration of Type Ia SNe. We apply PCA to two different cases: a small data set of supernova spectra taken at maximum light and a larger data set with more spectra taken at various epochs. On the SN Ia luminosity scale, the supernova SN 1991T appears at the upper end and SN 1991bg at the lower end. While 91bg-like SNe seem to form a distinct group, 91T-like SNe show a continuum of properties with normal SNe. The differences are mainly explained by line shifts in the spectra. However, we do not find that PCA is able to distinguish trends or subsets in the supernova data beyond what has already been found using specific spectral features.

The main utility of PCA will be as a tool for characterizing large sets of spectra. We show how the information in a data base of supernova spectra can be vastly simplified using PCA. This can be used to make a continuum of spectral templates from a discrete library of spectra, which may be useful in k-corrections and the training of supernova light-curve fitters.