We examine the dependence of the mass-to-light (M/L) ratio of large-scale structure on cosmological parameters, in models that are constrained to match observations of the projected galaxy correlation function w(rp). For a sequence of cosmological models with a fixed P(k) shape and increasing normalization sig8, we find parameters of the galaxy halo occupation distribution (HOD) that reproduce SDSS measurements of w(rp) as a function of luminosity. Using these HOD models we calculate mean M/L ratios as a function of halo mass and populate halos of N-body simulations to compute M/L in larger scale environments, including cluster infall regions. For all cosmological models, the M/L ratio in high mass halos or high density regions is approximately independent of halo mass or smoothing scale. However, the plateau value of M/L depends on sig8 as well as Omega_m, and it represents the universal mass-to-light ratio <M/L> only for models in which the galaxy correlation function is approximately unbiased, i.e., with sig8 ~ sig8_gal. Our results for cluster mass halos follow the trend M/L = 577(Omega_m/0.3)(sig8/0.9)^{1.7} h Msun/Lsun. Combined with Carlberg et al.s (1996) mean M/L ratio of CNOC galaxy clusters, this relation implies (sig8/0.9)(Omega_m/0.3)^{0.6} = 0.75 +/- 0.06. M/L ratios of clusters from the SDSS and CAIRNS surveys yield similar results. This constraint is inconsistent with parameter values Omega_m ~ 0.3, sig8 ~ 0.9 favored by recent joint analyses of CMB measurements and other large-scale structure data. We discuss possible resolutions, none of which seems entirely satisfactory. Appendices present an improved formula for halo bias factors and an improved analytic technique for calculating the galaxy correlation function from a given cosmological model and HOD. (Abridged)
Download