The Mean Type Ia Supernova Spectrum Over the Past 9 Gigayears

We examine the possibility of evolution with redshift in the mean rest-frame ultraviolet (UV; <4500A) spectrum of Type Ia Supernovae (SNe Ia) sampling the redshift range 0<z<1.3. We find new evidence for a decrease with redshift in the strength of intermediate-mass element (IME) features, particularly Si II and to a lesser extent Ca II “H&K” and Mg II blends, indicating lower IME abundances in the higher redshift SNe. A larger fraction of luminous, wider light-curve width (higher “stretch”) SNe Ia are expected at higher redshift than locally, so we compare our observed spectral evolution with that predicted by a redshift-evolving stretch distribution (Howell et al. 2007) coupled with a stretch-dependent SN Ia spectrum. We show that the sense of the spectral evolution can be reproduced by this simple model, though the highest redshift events seem additionally deficient in Si and Ca. We also examine the mean SN Ia UV-optical colors as a function of redshift, thought to be sensitive to variations in progenitor composition. We find that the expected stretch variations are sufficient to explain the differences, although improved data at z~0 will enable more precise tests. Thus, to the extent possible with the available datasets, our results support the continued use of SNe Ia as standardized candles.

💡 Research Summary

The authors set out to test whether the average rest‑frame ultraviolet (UV; λ < 4500 Å) spectrum of Type Ia supernovae (SNe Ia) evolves with redshift, a question of direct relevance to the use of SNe Ia as standardizable candles in cosmology. They assembled a heterogeneous sample of several hundred SNe Ia spanning 0 ≤ z ≤ 1.3, drawing from low‑redshift spectroscopic surveys, the Sloan Digital Sky Survey (SDSS), and high‑redshift observations with the Hubble Space Telescope (HST) and ground‑based facilities. After correcting each spectrum for Milky Way extinction, shifting to the rest frame, and normalizing in a common wavelength window, they constructed mean spectra in several redshift bins.

The most striking result is a systematic weakening of intermediate‑mass‑element (IME) absorption features—principally Si II λ 4130, the Ca II “H&K” blend, and Mg II λ 2800—as redshift increases. Quantitatively, the Si II feature depth declines by roughly 15 % between the lowest‑z bin (z ≈ 0.1) and the highest‑z bin (z ≈ 1.2). This trend suggests that, on average, higher‑z SNe Ia possess lower abundances of IMEs in their ejecta, or that the line‑forming regions are physically different (e.g., higher ionization).

To interpret this evolution, the authors invoke the well‑known correlation between light‑curve stretch (a proxy for luminosity and light‑curve width) and spectral properties. Higher‑stretch SNe Ia tend to have broader light curves, higher peak luminosities, and weaker IME absorption. Independent studies (Howell et al. 2007) have shown that the stretch distribution itself evolves with redshift: the fraction of high‑stretch events rises at larger z because of selection effects and possibly underlying population changes. The authors therefore constructed a simple model in which (i) the stretch distribution follows the Howell et al. parameterization as a function of redshift, and (ii) the SN Ia spectrum varies smoothly with stretch, based on empirical low‑z templates. By convolving the evolving stretch distribution with the stretch‑dependent spectral templates, they generated predicted mean spectra for each redshift bin.

The model reproduces the observed direction and approximate magnitude of the IME weakening, confirming that much of the spectral evolution can be explained as a demographic shift toward higher‑stretch SNe Ia at earlier cosmic times. However, the highest‑z bin shows a residual deficit in Si II and Ca II strength beyond the model prediction, hinting at an additional effect—perhaps a genuine metallicity evolution of the progenitor population or a change in explosion physics that is not captured by stretch alone.

The authors also examined UV‑optical colour evolution (e.g., U‑B, B‑V) because UV colours are sensitive to progenitor metallicity. They compared the observed colour trends with those expected from the stretch‑driven spectral changes. The colour differences across redshift are fully accounted for by the stretch‑induced spectral variations; no extra metallicity term is required given the current uncertainties. Consequently, the data do not demand a revision of the standardisation procedures based on colour corrections.

Limitations of the study are acknowledged. The low‑z UV sample is sparse, making it difficult to disentangle intrinsic metallicity effects from stretch‑related trends. High‑z spectra often have lower signal‑to‑noise, especially in the far‑UV, which inflates uncertainties on line depths. Moreover, the model assumes a linear, monotonic relationship between stretch and spectral features, which may oversimplify the true physics.

In conclusion, the paper provides compelling evidence that the apparent evolution of the mean SN Ia UV spectrum with redshift is largely driven by the changing mix of light‑curve stretch in the observed population. While a modest residual deficiency of Si and Ca at the highest redshifts could point to genuine progenitor evolution, the current data are insufficient to claim a statistically significant metallicity trend. Therefore, within the precision of existing datasets, SNe Ia remain reliable standard candles for cosmological distance measurements, though future large‑scale, high‑quality UV spectroscopic surveys at low redshift will be essential to tighten these constraints.