Still Accelerating: Type Ia supernova cosmology is robust to host galaxy age evolution

TL;DR

The paper challenges the claim that host-galaxy age evolution markedly biases SN Ia cosmology by demonstrating that standard mass-based corrections and bias modeling already remove the purported age dependence. It argues that the SN Ia progenitor age does not track host age one-to-one due to a short-delay-dominated DTD, and that progenitor ages evolve modestly with redshift, leading to negligible redshift-dependent luminosity evolution. Using DES-SN5YR data and Wiseman 2022 simulations, the authors show no significant redshift evolution in the host-mass correction (gamma) and no evidence for a redshift-evolving mass step, implying a shift in w of less than 0.01 when such evolution is included. The work emphasizes careful, galaxy-by-galaxy modelling of SFHs, DTDs, and survey effects, reinforcing the robustness of current SN Ia cosmology in constraining dark energy.

Abstract

Type Ia supernovae are a cornerstone of modern cosmology, providing first evidence for cosmic acceleration and new tests of dark energy. Son et al. 2025 (S25) claim a strong redshift evolution in standardized supernova luminosities driven by supernova progenitor age, with dramatic cosmological implications: rapidly evolving dark energy, decelerating expansion, and a $9σ$ tension with $Λ$CDM. We show that the underpinning evidence required for this conclusion -- the supernova progenitor-age dependence, the redshift-dependent age difference, and their combined impact -- is either negligible or relies on effects already corrected for in modern supernova analyses. First, the S25 analysis omits the standard host-galaxy stellar mass correction that captures known environmental dependencies that also correlate with stellar age. Applying this correction to the S25 sample, we find no dependence of standardized supernova brightness on host age. Independent data also show no significant difference at low-redshift in standardized brightness between star-forming galaxies and several Gyr older quiescent galaxies of the same stellar mass. Second, the S25 scenario predicts strong redshift evolution of the host-mass effect. Data from the Dark Energy Survey supernova survey measure evolution of $-0.028 \pm 0.034~\mathrm{mag}\,z^{-1}$, consistent with zero and altering the dark-energy equation-of-state measurement ($w$) by $<$0.01 if included. Third, we demonstrate that the claimed $\sim5$~Gyr progenitor age difference between nearby and distant supernovae is overstated by factors of three to five largely due to a conflation of host galaxy age with supernova progenitor age. We conclude that type~Ia supernova cosmology remains robust for current measurements of dark energy.

Figures (8)

• Figure 1: Left: Host galaxy stellar mass versus galaxy age for the sample used by son_strong_2025. Galaxy ages are taken from chung_strong_2025 and are strongly correlated with host galaxy stellar mass. Centre: Hubble residual before bias correction and mass standardization, versus galaxy age. Hubble residuals are taken from the Pantheon+ analysis. We recover a somewhat smaller slope than chung_strong_2025, and with lower significance; Right: Hubble residuals after a bias correction and standardization for stellar mass. The relationship between Hubble residual and host galaxy age is smaller, and not significant.
• Figure 2: Hubble residuals of low-redshift SNe Ia as a function of stellar mass, split by the morphology of the host galaxies. Hubble residuals are from the DES-SN5YR compilation and have had both the stellar mass standardization and bias correction applied The mean difference in age of the two morphology categories is several Gyr, which based on the son_strong_2025 scenario of 0.03 mag Gyr$^{-1}$ would predict a $\sim$0.09--0.18 mag difference between the two populations, which is not seen in the mean of the data.
• Figure 3: The size of the redshift-evolving $\gamma$ (' mass step') correction measured with DES-SN5YR data (blue points). The blue line is the best fit to the DES data of the form $\gamma(z)=\gamma_0 + \gamma_1 z$ and the blue shaded area the range of evolution allowed based on the uncertainty from this fit. This is compared to the predicted redshift evolution from the age-bias model of son_strong_2025 (purple line). The important difference occurs at low-redshift where the observed $\gamma$ is much smaller than the prediction and little changed from high redshift.
• Figure 4: Relationships between galaxy stellar masses, stellar and SN Ia progenitor ages, and redshifts, according to the realistic models from Wiseman2022. Left: Mean galaxy age (gold), mean SN Ia host galaxy age (pink), and mean SN Ia progenitor age (blue) as a function of galaxy stellar mass at $z=0$. The age of galaxies, as well as the SN Ia progenitors, is clearly correlated with the stellar mass; Right: Mean galaxy age (gold), mean SN Ia host age (pink), and mean SN Ia progenitor age (blue) as a function of redshift. Due to the SN Ia delay time distribution (DTD), the SN age difference across the redshift range is half the galaxy age difference and 50 per cent smaller than the SN host age--redshift difference. In purple is the prediction from the DTD and cosmic SFH assumed by son_strong_2025.
• Figure 5: Evolution of the fraction of SNe Ia in high stellar mass galaxies as a function of redshift. Data are taken from the full DES-SN5YR dataset, including low-redshift surveys. The solid line is the prediction from a simulation with full treatment of galaxy quenching, starbursts, dust, and SN and galaxy selection effects from Wiseman2022. The dashed line is a prediction based on estimating how the fraction of SNe Ia in high mass galaxies would evolve with redshift given the large difference in SN Ia ages predicted by the DTD and cosmic SFH model used by son_strong_2025.