Two populations of progenitors for Type Ia supernovae?

We use recent observations of the evolution of the Type Ia supernova (SN Ia) rate with redshift, the dependence of the SN Ia rate on the colours of the parent galaxies, and the enhancement of the SN Ia rate in radio-loud early-type galaxies to derive on robust empirical grounds, the delay time distribution (DTD) between the formation of the progenitor star and its explosion as an SN. Our analysis finds: (i) delay times as long as 3-4 Gyr, derived from observations of SNe Ia at high redshift, cannot reproduce the dependence of the SN Ia rate on the colours and on the radio-luminosity of the parent galaxies, as observed in the local Universe; (ii) the comparison between observed SN rates and a grid of theoretical 'single-population' DTDs shows that only a few of them are possibly consistent with observations. The most successful models are all predicting a peak of SN explosions soon after star formation and an extended tail in the DTD, and can reproduce the data but only at a modest statistical confidence level; (iii) present data are best matched by a bimodal DTD, in which about 50 per cent of SNe Ia (dubbed 'prompt' SNe Ia) explode soon after their stellar birth, in a time of the order of 10 8 yr, while the remaining 50 per cent ('tardy' SNe Ia) have a much wider distribution, well described by an exponential function with a decay time of about 3 Gyr.