The Evolution of Assembly Bias

We examine the evolution of assembly bias using a semi-analytical model of galaxy formation implemented in the Millennium-WMAP7 N-body simulation. We consider fixed number density galaxy samples ranked by stellar mass or star formation rate. We investigate how the clustering of haloes and their galaxy content depend on halo formation time and concentration, and how these relationships evolve with redshift. At $z=0$ the dependences of halo clustering on halo concentration and formation time are similar. However, at higher redshift, halo assembly bias weakens for haloes selected by age, and reverses and increases for haloes selected by concentration. The variation of the halo occupation with concentration and formation time is also similar at $z=0$ and changes at higher redshifts. In this case, the occupancy variation with halo age stays mostly constant with redshift but decreases for concentration. Finally, we look at the evolution of assembly bias reflected in the galaxy distribution by examining the galaxy correlation functions relative to those of shuffled galaxy samples which remove the occupancy variation. This correlation functions ratio monotonically decreases with larger redshift and for lower number density samples, going below unity in some cases, leading to reduced galaxy clustering. While the halo occupation functions themselves vary, the assembly bias trends are similar whether selecting galaxies by stellar mass or star formation rate. Our results provide further insight into the origin and evolution of assembly bias. Our extensive occupation function measurements and fits are publicly available and can be used to create realistic mock catalogues.