On the scatter in the relation between stellar mass and halo mass: random or halo formation time dependent?

Abstract

The empirical HOD model of Wang et al. 2006 fits, by construction, both the stellar mass function and correlation function of galaxies in the local Universe. In contrast, the semi-analytical models of De Lucia & Blazoit 2007 (DLB07) and Guo et al. 2011 (Guo11), built on the same dark matter halo merger trees than the empirical model, still have difficulties in reproducing these observational data simultaneously. We compare the relations between the stellar mass of galaxies and their host halo mass in the three models, and find that they are different. When the relations are rescaled to have the same median values and the same scatter as in Wang et al., the rescaled DLB07 model can fit both the measured galaxy stellar mass function and the correlation function measured in different galaxy stellar mass bins. In contrast, the rescaled Guo11 model still over-predicts the clustering of low-mass galaxies. This indicates that the detail of how galaxies populate the scatter in the stellar mass -- halo mass relation does play an important role in determining the correlation functions of galaxies. While the stellar mass of galaxies in the Wang et al. model depends only on halo mass and is randomly distributed within the scatter, galaxy stellar mass depends also on the halo formation time in semi-analytical models. At fixed value of infall mass, galaxies that lie above the median stellar mass -- halo mass relation reside in haloes that formed earlier, while galaxies that lie below the median relation reside in haloes that formed later. This effect is much stronger in Guo11 than in DLB07, which explains the over-clustering of low mass galaxies in Guo11. Our results illustrate that the assumption of random scatter in the relation between stellar and halo mass as employed by current HOD and abundance matching models may be problematic in case a significant assembly bias exists in the real Universe.