On the Luminosity Dependence of the Galaxy Pairwise Velocity Dispersion

(Abridged) We make predictions for the pairwise velocity dispersion (PVD) of galaxies with models that are constrained to match the projected correlation function and luminosity function of galaxies in the Two-Degree Field Galaxy Redshift Survey (2dFGRS). We use these data to constrain the halo occupation distribution (HOD), then calculate the PVD by populating the halos of a high resolution N-body simulation. We examine the luminosity and scale dependence of the predicted PVD. At r3 Mpc/h, we find that the PVD decreases with increasing galaxy luminosity. This result is mostly driven by the fraction of satellite galaxies f_sat, which is well-constrained by the correlation function. We find f_sat~25% for galaxies fainter than L_star, while for brighter galaxies the satellite fraction rapidly declines, creating the decrease in the PVD with luminosity. At r=1 Mpc/h, the PVD has no dependence on luminosity because satellite galaxies dominate the statistics for all objects. Recent measurements of the PVD in Fourier space using the "dispersion model" have reported a strong decline in PVD with increasing luminosity at k=1 h/Mpc. We test this method with our HOD models, finding that there is no consistent comparison between the PVD at a given k and the true dispersion at a given value of r. This results in a luminosity dependence in k-space that is stronger than in configuration space. The luminosity dependence of the HOD results in Fourier space are consistent with those measured at k=1 h/Mpc; thus the recent measurements of the PVD are fully explainable in the context of halo occupation models. The normalization of the PVD is lower than predicted by our fiducial model, and reproducing it requires a lower value of Omega_m (~0.2 instead of 0.3), a lower value of sigma_8 (~0.7 instead of 0.9), or strong velocity bias