2016MNRAS.460..373S

A new method to study the intrinsic colour and luminosity of Type Ia supernovae (SNe Ia) is presented. A metric space built using principal component analysis on a spectral series for SNe Ia between -12.5 and +17.5 d from the B maximum is used as a set of predictors. This metric space is built to be insensitive to reddening. Hence, it does not predict the part of the colour excess due to dust extinction. At the same time, the rich variability of SN Ia spectra is a good predictor of a large fraction of the intrinsic colour variability. Such a metric space is a good predictor of the epoch when the maximum in the B - V colour curve is reached. Multivariate partial least-squares regression predicts the intrinsic B-band light curve and the intrinsic B - V colour curve up to a month after the maximum. This allows us to study the relation between the light curves of SNe Ia and their spectra. The total-to-selective extinction ratio R_V in the host galaxy of SNe Ia is found, on average, to be consistent with typical Milky Way values. This analysis shows the importance of collecting spectra to study SNe Ia, even with a large sample publicly available. Future automated surveys, such as the Large Synoptic Survey Telescope, will provide a large number of light curves. The analysis shows that observing accompanying spectra for a significant number of SNe will be important even for normal SNe Ia.