Where are z=4 Lyman Break Galaxies?: Results from conditional luminosity function models of luminosity-dependent clustering

Using the conditional luminosity function (CLF) - the luminosity distribution of galaxies in a dark matter halo - as a way to model galaxy statistics, we study how z = 4 Lyman Break Galaxies (LBGs) are distributed in dark matter haloes. For this purpose, we measure luminosity-dependent clustering of LBGs in the Subaru/XMM-Newton Deep Field by separating a sample of 16 920 galaxies to three magnitude bins in i' band between 24.5 and 27.5. Our model fits to data show a possible trend for more-luminous galaxies to appear as satellites in more-massive haloes; the minimum halo mass in which satellites appear is 3.9(-3.5)(+4.1) x 10(12), 6.2(-4.9)(+3.8) x 10(12) and 9.6(-4.60)(+7.0) x10(12)M circle dot (1 sigma errors) for galaxies with 26.5 < i' < 27.5, 25.5 < i' < 26.5 and 24.5 < i, < 25.5 mag, respectively. The satellite fraction of galaxies at z = 4 in these magnitude bins is 0.13-0.3, 0.09-0.22 and 0.03-0.14, respectively, where the 1s ranges account for differences coming from two different estimates of the z = 4 LF from the literature. To jointly explain the LF and the large-scale linear bias factor of z = 4 LBGs as a function of rest UV luminosity requires central galaxies to be brighter in UV at z = 4 than present-day galaxies in same dark matter mass haloes. Moreover, UV luminosity of central galaxies in haloes with total mass greater than roughly 10(12)M circle dot must decrease from z = 4 to today by an amount more than the luminosity change for galaxies in haloes below this mass. This mass-dependent luminosity evolution is preferred at more than 3 sigma confidence level compared to a pure-luminosity evolution scenario where all galaxies decrease in luminosity by the same amount from z = 4 to today. The scenario preferred by the data is consistent with the 'downsizing' picture of galaxy evolution.