The need for accurate redshifts in supernova cosmology

Recent papers have shown that a small systematic redshift shift (Δ z ∼ 10 −5 ) in measurements of type Ia supernovae can cause a significant bias (∼1%) in the recovery of cosmological parameters. Such a redshift shift could be caused, for example, by a gravitational redshift due to the density of our local environment. The sensitivity of supernova data to redshift shifts means supernovae make excellent probes of inhomogeneities. We therefore invert the analysis, and try to diagnose the nature of our local gravitational environment by fitting for Δ z as an extra free parameter alongside the usual cosmological parameters. Using the Joint Light-curve SN Ia dataset we find the best fit includes a systematic redshift shift of Δ z = (2.6 +2.7 −2.8 ) × 10 −4 . This is a larger shift than would be expected due to gravitational redshifts in a standard Λ-Cold Dark Matter universe (though still consistent with zero), and would correspond to a monopole Doppler shift of about 100 km s −1 moving away from the Milky-Way. However, since most supernova measurements are made to a redshift precision of no better than 10 −3 , it is possible that a systematic error smaller than the statistical error remains in the data and is responsible for the shift; or that it is an insignificant statistical fluctuation. We find that when Δ z is included as a free parameter while fitting to the JLA SN Ia data, the constraints on the matter density shifts to Ω m = 0.313 +0.042 −0.040 , bringing it into better agreement with the CMB cosmological parameter constraints from Planck. A positive Δ z ∼ 2.6 × 10 −4 would also cause us to overestimate the supernova measurement of Hubble's constant by Δ H 0 ∼ 1 kms −1 Mpc −1 . However this overestimation should diminish as one increases the low-redshift cutoff, and this is not seen in the most recent data.