Redshift evolution of the underlying type Ia supernova stretch distribution

The true nature of type Ia supernovae (SNe Ia) remains largely unknown, and as survey statistics increase, the question of astrophysical systematic uncertainties rises, notably that of the SN Ia population evolution. In this paper, we study the dependence with redshift of the SN Ia SALT2.4 lightcurve stretch, a purely intrinsic SN property, to probe its potential redshift drift. The SN stretch has been shown to strongly correlate with the SN environment, notably with stellar age tracers. We model the underlying stretch distribution as a function of redshift, using the evolution of fraction of young and old SNe Ia as predicted by Rigault et al. (2018), and assuming constant underlying stretch distribution for each age population made of Gaussian mixtures. We test our prediction against published samples chosen to have negligible magnitude selection effects, so that any observed change is indeed of astrophysical and not observational origin. We clearly demonstrate that the underlying SN Ia stretch distribution is evolving as a function of redshift, and that the young/old drifting model is a much better description of the data than any time-constant model, including the sample-based asymmetric distributions usually used to correct Malmquist bias. The favored underlying stretch model is the bimodal one derived from Rigault et al. (2018): a high-stretch mode shared by both young and old environments, and a low-stretch mode exclusive to old environments. The precise impact of the redshift evolution of the SN Ia population intrinsic properties on cosmology remains to be studied. Yet,the astrophysical drift of the SN stretch distribution does affect current Malmquist bias corrections and thereby distances derived from SN affected by selection effects. We highlight that such a bias will increase with surveys covering increasingly larger redshift ranges, which is particularly important for LSST.