SIMBAD references

We analyze the redshift evolution of the rotation measure (RM) in the Taylor et al. data set, based on NVSS radio data at 21 cm, and compare with results from our previous work based on RMs determined at lower wavelengths, e.g., 6 cm. We find that, in spite of the same analysis, Taylor et al.'s data set produces neither an increase of the RM dispersion with redshift nor the correlation of RM strength with Mg II absorption lines that we found previously. We develop a simple model to understand this discrepancy. The model assumes that the Faraday rotators, namely the QSO's host galaxy and the intervening Mg II host galaxies along the line of sight, contain partially inhomogeneous RM screens. We find that this leads to an increasing depolarization toward longer wavelengths and to wavelength-dependent RM values. In particular, due to cosmological redshift, observations at fixed wavelength of sources at different redshift are affected differently by depolarization and are sensitive to different Faraday active components. For example, at 21 cm the polarized signal is averaged out by inhomogeneous Faraday screens and the measured RM mostly reflects the Milky Way contributions for low-redshift QSOs whereas polarization is relatively unaffected for high-redshift QSOs. Similar effects are produced by intervening galaxies acting as inhomogeneous screens. Finally, we assess the performance of RM synthesis on our synthetic models and conclude that the study of magnetic fields in galaxies as a function of cosmic time will benefit considerably from the application of such a technique, provided there is sufficient instrumental bandwidth. For this purpose, high-frequency channels appear preferable but not strictly necessary.