Confirmation of a Star Formation Bias in Type Ia Supernova Distances and its Effect on Measurement of the Hubble Constant

TL;DR

This study confirms a local star-formation environment bias in Type Ia supernovae by analyzing the Constitution/GALEX data and reproducing the brightness difference between SNe Ia in locally passive versus star-forming regions across SALT2 and MLCS2k2 standardization. By quantifying the bias and combining it with prior R13 results, the authors derive a universal SF-bias magnitude and propagate its effect to the direct H0 measurement, showing that the H0 value is overestimated by about 3.3% due to environmental differences between Cepheid-calibrated and Hubble-flow SN samples. The resulting H0 corrections bring SN-based estimates into closer agreement with CMB-derived values, reducing tension and highlighting the importance of accounting for local environments in SN cosmology. The work also links the SF bias to the host-mass step, suggesting that masking the root environmental dependence may resolve multiple systematic issues in SN distance scales.

Abstract

Previously we used the Nearby Supernova Factory sample to show that SNe~Ia having locally star-forming environments are dimmer than SNe~Ia having locally passive environments.Here we use the \constitution\ sample together with host galaxy data from \GALEX\ to independently confirm that result. The effect is seen using both the SALT2 and MLCS2k2 lightcurve fitting and standardization methods, with brightness differences of $0.094 \pm 0.037\ \mathrm{mag}$ for SALT2 and $0.155 \pm 0.041\ \mathrm{mag}$ for MLCS2k2 with $R_V=2.5$. When combined with our previous measurement the effect is $0.094 \pm 0.025\ \mathrm{mag}$ for SALT2. If the ratio of these local SN~Ia environments changes with redshift or sample selection, this can lead to a bias in cosmological measurements. We explore this issue further, using as an example the direct measurement of $H_0$. \GALEX{} observations show that the SNe~Ia having standardized absolute magnitudes calibrated via the Cepheid period--luminosity relation using {\textit{HST}} originate in predominately star-forming environments, whereas only ~50% of the Hubble-flow comparison sample have locally star-forming environments. As a consequence, the $H_0$ measurement using SNe~Ia is currently overestimated. Correcting for this bias, we find a value of $H_0^{corr}=70.6\pm 2.6\ \mathrm{km\ s^{-1}\ Mpc^{-1}}$ when using the LMC distance, Milky Way parallaxes and the NGC~4258 megamaser as the Cepheid zeropoint, and $68.8\pm 3.3\ \mathrm{km\ s^{-1}\ Mpc^{-1}}$ when only using NGC~4258. Our correction brings the direct measurement of $H_0$ within $\sim 1\,σ$ of recent indirect measurements based on the CMB power spectrum.

Figures (3)

• Figure 1: SN Ia redshifts and SALT2 standardized Hubble residuals ($\Delta M_B^{\mathrm{corr}}$) as a function of $\log(\Sigma_{\mathrm{SFR}})$. Cases where no counts were found in the GALEX FUV image are arbitrarily set to $\log(\Sigma_{\mathrm{SFR}})=-5.3$ dex. Upper panel: the $\log(\Sigma_{\mathrm{SFR}})$ distributions for the environmental subgroups. Each SN contributes to the amplitude of the open histogram according to its value of $\mathcal{P}\mathrm{(Ia\epsilon)}$, and the filled, green histogram according to its value of $\mathrm{\mathcal{P}({Ia}\alpha)}$. Main panels: Marker colors encode the value of $\mathcal{P}\mathrm{(Ia\epsilon)}$ for each SN. Those identified as having a globally passive host (Pa and $\sim$Pa, as defined in Section \ref{['sec:local-dust-correction']}), are highlighted with thick black marker contours (see the legend for details). The vertical dashed blue line shows our $\log(\Sigma_{\mathrm{SFR}})=-2.9$ dex star-formation surface-density threshold. Right panels, from top to bottom: Marginal distributions of redshift and $\Delta M_B^{\mathrm{corr}}$ for each subgroup. These bi-histograms follow the same color code and construction method as the $\Sigma_{\mathrm{SFR}}$ histograms. The weighted mean of the $\Delta M_B^{\mathrm{corr}}$ values of each H09 subsample is drawn over its respective marginal distribution in the lower panels. The transparent bands show the $\pm 1\,\sigma$ uncertainty on these means. Compare to Figure 6 of R13, and see Figure \ref{['fig:R13_H09_LHa_bias_MLCS']} for the MLCS2k2 results.
• Figure 2: Summary of the influence of the analysis choices made in the paper. The main analysis results are indicated in the upper left of their corresponding panels and drawn as horizontal lines. The shaded bands indicate the corresponding $\pm1\,\sigma$ errors. Lower and Middle panels: The SF bias as presented in Section \ref{['sec:H09_Ha_bias']}, measured using Hubble residuals from H09 based on SALT2 (lower) and MLCS2k2 (middle) lightcurve parameters. Upper panel: The $H_0$ bias, as presented in Section \ref{['sec:H_0andH_a']}, using Hubble residuals based on MLCS2k2 lightcurve parameters, as in SH0ES11. The main $H_0$ bias uses $\psi^{C}=$$7.0$%, but we also present two variants. We consider the case when SN 2007sr is assumed to be a $\mathrm{Ia}\alpha$, in which case $\psi^{C}= 0.8\%$, as well as the case where the Cepheid hosts are measured with angular resolution and signal-to-noise typical of the Hubble-flow sample, in which case $\psi^{C}=$$15.4$%. The summary results are reported in Table \ref{['tab:environmental_H0_bias']}.
• Figure A.1: SN Ia MLCS2k2 standardized Hubble residuals ($\Delta M_B^{\mathrm{corr}}$) as a function of $\log(\Sigma_{\mathrm{SFR}})$, from the bottom to the top:$R_V$ of 1.7, 2.5 and 3.1, respectively. This plot is constructed like Figure \ref{['fig:R13_H09_LHa_bias']}.