1999ApJ...514..535H

High-resolution, deep-imaging surveys are instrumental in setting constraints on semianalytical structure formation models in cold dark matter (CDM) cosmologies. We show here that the lack of unresolved B-band ``dropouts'' with V>25 mag in the Hubble Deep Field appears to be inconsistent with the number of ``miniquasars'' expected if massive black holes form with a constant universal efficiency in all CDM halos. To reconcile the models with the data, a mechanism is needed that suppresses the formation of active nuclei in halos with circular velocities v_circ≲50-75 km.s^–1. This feedback naturally arises because of the photoionization heating of the gas by the UV background. We consider several alternative effects that would help reducing the number of miniquasars and find that these cannot alone account for the observed lack of detections. If reddening by dust can be neglected at early epochs, consistency with the optical data also requires that the luminous extent of dwarf galaxies at high redshifts be larger than a few percent of the virial radii of their dark matter halos, in order not to overpredict the number of pointlike B-band dropouts. Future deep observations in the J and H bands with NICMOS might reveal several z≳5 objects per field or provide even stronger constraints on the models than existing B, V, and I data.