Using the results of recent optical surveys we conclude that the non -detection of quasars down to faint magnitudes implies a significant flattening of the high redshift ( z \sim 6 ) optical active galactic nuclei ( AGN ) luminosity function for M _ { 1450 } \gtrsim - 24.7 .
We find that all the data are consistent with a faint-end slope for the optical AGN luminosity function of \beta = -2.2 and \beta = -2.8 , at the 90 % and 99 % confidence level respectively , flatter than the bright-end slope of \beta ^ { \prime } \sim - 3.2 .
We also show that X-ray deep surveys have probed even fainter magnitudes than the optical ones yielding more significant constraints on the shallow faint-end slope of the optical luminosity function .
The inclusion of Type II AGN candidates , detected in the Chandra deep fields , hints towards an higher normalization for the total AGN luminosity function , if these sources lie at 5 \lesssim z \lesssim 6.5 .
We then discuss simple theoretical models of AGN formation and evolution in the context of cold dark matter cosmology .
The comparison with the total AGN luminosity function favors a redshift-dependent relation between black hole and dark matter halo masses of the type M _ { \bullet } \propto M _ { halo } ^ { \alpha } , with 1.3 \lesssim \alpha \lesssim 1.7 , compatible with independent studies from statistical analysis and rotation curve measurements .
Finally we compute the quasar contribution to reionization to be \lesssim 9 \% at z \sim 6 , up to \sim 30 \% when integrated within 5.5 \lesssim z \lesssim 6.5 , significantly smaller than that from galaxies .