We study the spatial distribution of faint satellites of intermediate redshift ( 0.1 < z < 0.8 ) , early-type galaxies , selected from the GOODS fields .
We combine high resolution HST images and state-of-the-art host subtraction techniques to detect satellites of unprecedented faintness and proximity to intermediate redshift host galaxies ( up to 5.5 magnitudes fainter and as close as 0 \farcs 5/2.5 kpc to the host centers ) .
We model the spatial distribution of objects near the hosts as a combination of an isotropic , homogeneous background/foreground population and a satellite population with a power law radial profile and an elliptical angular distribution .
We detect a significant population of satellites ( N _ { s } = 1.7 ^ { +0.9 } _ { -0.8 } ) that is comparable to the number of Milky Way satellites with similar host-satellite contrast .
The average projected radial profile of the satellite distribution is isothermal ( \gamma _ { p } = -1.0 ^ { +0.3 } _ { -0.4 } ) , which is consistent with the observed central mass density profile of massive early-type galaxies .
Furthermore , the satellite distribution is highly anisotropic ( isotropy is ruled out at a > 99.99 % confidence level ) .
Defining \phi to be the offset between the major axis of the satellite spatial distribution and the major axis of the host light profile , we find a maximum posterior probability of \phi = 0 and | \phi| less than 42 ^ { o } at the 68 % confidence level .
The alignment of the satellite distribution with the light of the host is consistent with simulations , assuming that light traces mass for the host galaxy as observed for lens galaxies .
The anisotropy of the satellite population enhances its ability to produce the flux ratio anomalies observed in gravitationally lensed quasars .