4.3. Pre-Chandra and XMM-Newton Hard X-ray Surveys
The most sensitive large solid angle 2 - 8 keV surveys have been obtained via serendipitous sources detected by ASCA (Akiyama et al. 2003; Nandra et al. 2003) and BeppoSAX (Fiore et al. 2000). These satellites had moderate fields of view and moderate sensitivities, reaching 2 - 10 keV flux limits of ~ 2 × 10-13 ergs cm-2 s-1. At this flux level, there are 5 - 10 sources per square degree (Cagnoni, della Ceca, & Maccacaro 1998; Ueda et al. 2001), or roughly one serendipitous source per ASCA or BeppoSAX pointing. Over its lifetime, ASCA obtained ~ 500 serendipitous sources over ~ 100 deg2, and BeppoSAX somewhat fewer. Because of their moderate ~ 1 - 3' FWHM angular resolutions and 40 - 100" angular accuracies (similar to those of the earlier Einstein soft X-ray serendipitous survey), optical follow-up has been tedious, and the number of identified sources is now only ~ 150 objects. The difficulty of optical, IR, and radio follow-up seriously delayed the results from these surveys, and they appeared after the initial Chandra results, despite the fact that these satellites were launched 5 - 7 years earlier. The angular resolution of these surveys limits the fluxes of the optical counterparts to R < 21 mag for the BeppoSAX HELLAS survey and R < 19 mag for the ASCA Large Area Sky Survey, in order to avoid confusion and speed up the identification process. This is about two magnitudes brighter than the ROSAT Deep Survey limits. At these levels, ~ 85% of the sources are identified with fairly high confidence.
The nature of the sources is rather different from the ROSAT sources, with approximately one-third not having prominent broad lines (La Franca et al. 2002; Akiyama et al. 2003), while in the bright and faint ROSAT samples (Lehmann et al. 2000; Appenzeller et al. 2000), more than 90% of the AGN identifications have broad lines. The hard X-ray surveys find very few narrow-line Seyfert 1 galaxies compared to the large fraction in the ROSAT surveys. There are only a very few sources at these hard X-ray flux levels that have "normal" galaxies as optical counterparts.
The range of optical-to-X-ray flux ratios is very large,
± 2 dex, which is much larger than in the ROSAT bands,
± 1 dex. It is believed that there are two main effects in
the sample differences: 1) obscuration, and
2) redshift. It has been noted in the deep ROSAT,
Chandra, and XMM-Newton samples
(see Section 4.4)
that there is a strong correlation
(Hasinger et al. 1999)
between R - K color
and optical magnitude, with the fainter optical counterparts of the
X-ray sources being systematically redder. Part of this is clearly
due to the increasing effect of dust on the rest-frame UV colors of
AGN (the same amount of dust is much more significant in the UV than
in the optical, and, of course, it is the UV which gets redshifted
into the rest-frame optical band for z > 1 objects), and part
seems to be due to the relative dominance of starlight in many
of the faint sources and the steep SED of stellar objects.
The mean X-ray spectral indices of the objects flatten as the sources
get fainter by about 
~ 0.3 over a factor of
103 in flux
(Ueda et al. 1999).
The properties of these sources indicate that a significant fraction of them have highly absorbed spectra, but the absorption is not "simple". The popular "leaky absorber model" (Turner et al. 1997), which fits many of the high signal-to-noise ASCA AGN spectra, produces a significant signal for highly absorbed objects in the soft (0.5 - 2 keV) band. Recently, results from many Chandra and XMM-Newton serendipitous fields are becoming available, which will allow a much larger database for the FX < 10-13 ergs cm-2 s-1 sources, but the solid angles covered will be too small to produce many sources in the FX ~ 10-12 -10-13 ergs cm-2 s-1 range, for which good X-ray spectra can be obtained.
It is entertaining to note that the recent deep Chandra hard X-ray surveys have more objects per pointing than the entire HEAO-1 or ASCA/BeppoSAX data sets.