Further evidence for significant luminosity evolution in supernova cosmology

Supernova (SN) cosmology is based on the assumption that the corrected luminosity of SN Ia would not evolve with redshift. Recently, our age dating of stellar populations in early-type host galaxies (ETGs) from high-quality spectra has shown that this key assumption is most likely in error. It has been argued though that the age-Hubble residual (HR) correlation from ETGs is not confirmed from two independent age datasets measured from multi-band optical photometry of host galaxies of all morphological types. Here we show, however, that one of them is based on highly uncertain and inappropriate luminosity-weighted ages derived, in many cases, under serious template mismatch. The other dataset employs more reliable mass-weighted ages, but the statistical analysis involved is affected by regression dilution bias, severely underestimating both the slope and significance of the age-HR correlation. Remarkably, when we apply regression analysis with a standard posterior sampling method to this dataset comprising a large sample ($N=102$) of host galaxies, very significant ($> 99.99 \%$) correlation is obtained between the global population age and HR with the slope ($-0.047 \pm 0.011$~mag/Gyr) highly consistent with our previous spectroscopic result from ETGs. For the local age of the environment around the site of SN, a similarly significant ($> 99.96 \%$) correlation is obtained with a steeper slope ($-0.057 \pm 0.016$ mag/Gyr). Therefore, the SN luminosity evolution is strongly supported by the age dating based on multi-band optical photometry and can be a serious systematic bias in SN cosmology.