Recovering galaxy stellar population properties from broad-band spectral energy distribution fitting - II. The case with unknown redshift

In a recent work (Paper I), we explored the dependence of galaxy stellar population properties derived from broad-band spectral energy distribution fitting on the fitting parameters, e.g. star formation histories (SFHs), age grid, metallicity, initial mass function (IMF), dust reddening and reddening law, filter setup and wavelength coverage. In this paper, we consider the case that also redshift is a free parameter in the fit and study whether one can obtain reasonable estimates of photometric redshifts and stellar population properties at once. As in Paper I, we use mock star-forming as well as passive galaxies placed at various redshifts (0.5-3) as test particles. Mock star-forming galaxies are extracted from a semi-analytical galaxy formation model. We show that for high-redshift star-forming galaxies, photometric redshifts, stellar masses and reddening can be determined simultaneously when using a broad wavelength coverage (including the Lyman and the 4000 Å break) and a wide template setup in the fit. Masses are similarly well recovered (median ∼0.2 dex) as at fixed redshift. For old galaxies with little recent star formation (which are at lower redshift in the simulation), masses are better recovered than in the fixed redshift case, such that the median recovered stellar mass improves by up to 0.3 dex (at fixed IMF) whereas the uncertainty in the redshift accuracy increases by only ∼0.05. However, a failure in redshift recovery also means a failure in mass recovery. As at fixed redshift, mismatches in SFH and degeneracies between age, dust and now also redshift cause underestimated ages, overestimated reddening and underestimated masses. Stellar masses are best determined at low redshift without reddening in the fit. Masses are then underestimated by only ∼0.1 dex whereas redshifts are similarly well recovered. Not surprisingly, the recovery of properties is substantially better for passive galaxies, for which e.g. the mass is recovered only slightly worse than at fixed redshift (underestimated by ∼0.02 dex instead of ∼0.01, at fixed IMF) using a setup including metallicity effects. In all cases, the recovery of physical parameters is crucially dependent on the wavelength coverage adopted in the fitting because the redshift recovery depends on the wavelength coverage. As is well known, redshifts are best recovered for a wavelength coverage including the Lyman and 4000 Å break. As in Paper I, all effects from changing templates, the wavelength coverage and filters are quantified and scaling relations for the transformation of stellar masses obtained using different fitting parameters, including stellar population models, are provided.