We identify a very deep sample of 198 extremely red objects ( EROs ) in the Chandra Deep Field South , selected on the basis of I _ { 775 } - K _ { s } > 3.92 , to a limit K _ { s } \simeq 22 
using the public ESO/GOODS survey .

The ERO number counts flatten from a slope of \gamma \simeq 0.59 to \gamma \simeq 0.16 at K > 19.5 , where they remain below the predictions for pure luminosity evolution , and fall below even a non-evolving model .
This suggests there is a significant decrease with redshift in the comoving number density of passive/very red galaxies .

We investigate the angular correlation function , \omega ( \theta ) , of these EROs and detect positive clustering for K _ { s } = 20.5 –22.0 sources .
The EROs show stronger clustering than other galaxies at the same magnitudes .
The \omega ( \theta ) amplitudes are best-fitted by models in which the EROs have a comoving correlation radius r _ { 0 } \simeq 12.5 \pm 1.2 ~ { } h ^ { -1 } Mpc , or r _ { 0 } \simeq 21.4 \pm 2.0 ~ { } h ^ { -1 } Mpc in a stable clustering model .

We find a 40-arcsec diameter overdensity of 10 EROs , centered on the Chandra X-ray source ( and ERO ) XID:58 .
On the basis of colours we estimate that about 7 , including XID:58 , belong to a cluster of EROs at z \simeq 1.5 .

The 942 ksec Chandra survey detected 73 X-ray sources in the area of our ERO sample , 17 of which coincide with EROs .
Of these sources , 13 have X-ray properties indicative of obscured AGN , while the faintest 4 may be starbursts .
In addition , we find evidence that Chandra sources and EROs are positively cross-correlated at non-zero ( \sim 2 –20 arcsec ) separations , implying that they tend to trace the same large-scale structures .

In conclusion these findings appear consistent with a scenario where EROs are the z > 1 progenitors of elliptical/S0 galaxies , some forming very early as massive spheroids , which are strongly clustered and may evolve via an AGN phase , others more recently from mergers of disk galaxies .