4.5. Comparison of X-ray and Optical Surveys
So the question arises as to what is the difference between the X-ray and optically-selected samples? Since, as discussed above, there are strong selection effects in soft X-ray, UV, and optical samples, especially at luminosities less than 1044 ergs s-1, it is not a great surprise to found that the major differences between these two surveys occur at this luminosity and less (Steffen et al. 2003). Clearly, optical-selection techniques are much less sensitive to low-luminosity objects, where the effects of stellar dilution can be large (Moran et al. 2002; Moran, this volume).
Stellar dilution reduces the amplitude of optical variability. In one of the very few direct comparisons of optical variability versus X-ray selection, Dobrzycki et al. (2003) found a ratio of X-ray to variability-selected quasars of ~ 5 : 1 at V < 20.5. Similar effects are seen in a comparison of objects in the SSA13 field, where Cowie et al. (1996) found only three obvious quasars, and a moderately deep Chandra observation found > 20 active galaxies in the same solid angle. Similar results are obtained in the Hubble Deep Field-North (HDF-N) and Flanking Fields, where Liu et al. (1999) found an areal density of B < 21 quasars of 30 deg-2 (rather typical of optical quasar surveys), while there are > 1000 Chandra AGN deg-2 in the same field.
Detailed analyses of several of the "optically-dull" galaxies (including the famous first one, 3C264) show that dilution, while important (Watanabe et al. 2002), is often not enough to remove an obvious AGN signature (Severgnini et al. 2003), and the optical continuum and emission lines must be reduced in strength if these objects have the same ratio of H or B-band flux to X-rays as "normal" AGN. A large fraction of the "optically-dull" galaxies have hard X-ray spectra indicative of column densities > 1022 atoms cm-2, but this is primarily derived from hardness ratios, rather than directly from X-ray spectral fitting.
Alternatively, it is possible that the "optically-dull" galaxies are intrinsically weak in the optical band, as seen in many LINERs, which have the same properties: high X-ray-to-optical ratios, absent or very weak optical continua, no broad lines, and weak total emission-line flux (Ho 1999). This reviewer suspects that all three effects are important but that many of the Chandra "optically-dull" galaxies are higher luminosity versions of LINERs, which are not cataloged in the low-redshift universe, since such objects would probably only be found in the low-redshift universe via a large solid angle sensitive hard X-ray survey. There are recent indications from the SDSS that such objects are quite common (as indicated in the Ho et al. 1995 work), but without X-ray measurements, their true luminosities are difficult to estimate.
Because of the absence of measurable optical nuclear light in these objects, the bolometric correction factors are not known, and thus the contribution of these objects to the mass density of black holes is difficult to estimate (see, however, Cowie & Barger, this volume). The recent HST observations of Chandra sources (Grogin et al. 2003) indicate that the observed nuclear light is ~ 20 times less than anticipated on the basis of the X-ray flux and the X-ray-to-optical ratios of the ROSAT sources.
One of the unexpected features of the distribution of the Chandra sources was their strong concentration in large scale structures (Barger et al. 2003b; Gilli et al. 2003; Yang et al. 2003), in contrast to optically-selected samples, which have the same correlation functions as normal galaxies. I suspect that this can be understood as a matter of the higher space density of X-ray-selected AGN that allows them to be traced on smaller length scales not possible with optical samples.
Theoretically, this result is perhaps unexpected. It is believed that there is a strong correlation between the mass of the black hole and the mass of the galaxy. Therefore, luminous AGN should be in massive galaxies, which are more strongly clustered than other galaxies. However, the mean bolometric luminosity of the Chandra AGN sources is considerably less than that of objects in optically-selected surveys, especially at z > 0.2. Thus, one might naively expect X-ray-selected AGN to be more weakly correlated than optically-selected objects.
It is likely that the X-ray luminosity is strongly correlated with the black hole mass, with a scatter of roughly 50 (Grupe 2004). This sort of correlation is apparently not seen with the bolometric luminosity (Woo & Urry 2002). Thus, the correlation of X-ray luminosity with the optical luminosity of the host galaxy seen in the Chandra fields (Fig. 22 in Barger et al. 2003b) indicates that a very large fraction of the "medium luminosity" Chandra AGN lie in massive galaxies that are strongly correlated (Barger et al. 2001; Cowie et al. 2004b). I believe the true question should be: why does the optically-selected sample not show a stronger correlation function?