2004MNRAS.347.1011A

Using published Hubble Space Telescope (HST) Cepheid data from 25 galaxies, we have found a correlation between the dispersion in the Cepheid period-luminosity (P -L) relation and host galaxy metallicity, which is significant at the ~3σ level in the V band. In the I band the correlation is less significant, although the tighter intrinsic dispersion of the P -L relation in I may make it harder to detect such a correlation in the HST sample. One possibility is that low metallicity galaxies have smaller metallicity gradients than high metallicity galaxies; if the Cepheid P -L relation has a significant dependence on metallicity then this might explain the higher P -L dispersion in the higher metallicity galaxies. A second possibility is that the increased P -L dispersion is driven by metallicity dispersion but now due to a relation between metallicity and Cepheid colour rather than luminosity. A third possibility is that the increased P -L dispersion is caused by an increase in the width of the instability strip with metallicity.

Whatever the explanation, the high observed dispersions in the HST Cepheid P -L relations have the important consequence that the bias due to incompleteness in the P -L relation at faint magnitudes is more significant than previously thought. Using a maximum likelihood technique which takes into account the effect on the P -L relations of truncation by consistently defined magnitude completeness limits, we rederive the Cepheid distances to the 25 galaxies. In the case of the galaxies with the highest P -L dispersion at the largest distances, we find that the published distance modulus underestimates the true distance modulus by up to ~0.5 mag.

When both metallicity and magnitude incompleteness corrections are made, a scale error in the published Cepheid distances is seen in the sense that the published distance moduli are increasingly underestimated at larger distances. This results in the average distance modulus to the four galaxies in the Virgo cluster core increasing from (m -M)₀31.2±0.19 to (m -M)₀31.4±0.19 if the γ_VI-0.24 mag/dex metallicity correction of Kennicutt et al. is assumed. For the 18 HST galaxies with good Tully-Fisher (TF) distances and (m -M)₀> 29.5 the Cepheid-TF distance modulus average residual increases from 0.44±0.09 to 0.63±0.1 mag with γ_VI-0.24. This indicates a significant scale error in TF distances, which reduces the previous Pierce & Tully TF estimate of H₀85±10 km/s/Mpc to H₀63±7 km/s/Mpc, assuming γ_VI-0.24 and a still uncertain Virgo infall model. Finally, for the eight HST galaxies with Type Ia supernovae (SNIa), the metallicity and incompleteness corrected Cepheid distances marginally suggest there may be a metallicity dependence of SNIa peak luminosity in the sense that metal-poor hosts have lower luminosity SNIa. Thus, SNIa Hubble diagram estimates of both H₀and q₀may therefore also require significant corrections for metallicity, once the exact sizes of the Cepheid metallicity corrections become better established.