3.1. Properties of the X-ray Sources
A broad diversity of X-ray sources is found in the recent Chandra and XMM-Newton surveys. This is apparent in even basic flux-flux plots such as that shown in Figure 3; at the faintest X-ray flux levels in the CDF-N, the extragalactic sources range in optical flux by a factor of 10, 000.
Figure 3. I-band magnitude versus 0.5-2 keV flux for extragalactic X-ray sources in the CDF-N. Sources with redshifts of 0-0.5, 0.5-1, 1-2, and 2-6 are shown as violet, blue, green, and red filled triangles, respectively (symbol sizes also increase with redshift). Small black dots indicate sources without measured redshifts. The slanted, dotted lines indicate constant values of log(fX / fI); the respective log(fX / fI) values are labeled. Adapted from D.M. Alexander, F.E. Bauer, W.N. Brandt, et al., 2003, AJ, 126, 539 and A.J. Barger, L.L. Cowie, P. Capak, et al., 2003, AJ, 126, 632.
Classification of the X-ray sources is challenging for several reasons. First of all, many of the sources are simply too faint for efficient optical spectroscopic identification with 8-10 m class telescopes (note the small dots in Figure 3). Intensive optical identification programs on the deepest Chandra and XMM-Newton fields typically have 50-70% completeness at best. Furthermore, many of the X-ray sources have modest apparent optical luminosities, and thus their host galaxies make substantial diluting contributions to the flux measured in a spectroscopic aperture. Finally, another challenge is an apparent "schism" between optical (type 1 vs. type 2) and X-ray (unobscured vs. obscured) schemes of classification; not all X-ray obscured AGN have type 2 optical spectra, and not all AGN with type 1 optical spectra are unobscured.
Considering X-ray, optical, and multiwavelength information, the primary extragalactic source types are found to be the following:
Unobscured AGN. Blue, broad-line AGN are found that do not show signs of obscuration at either X-ray or optical/UV wavelengths. They are found over a broad range of redshift (z 0-5), and they comprise a significant fraction of the brightest X-ray sources. At z 1.5 they also comprise a substantial fraction of all X-ray sources with spectroscopic identifications (certainly in part because these objects are the most straightforward to identify spectroscopically).
Obscured AGN with clear optical/UV AGN signatures. Some objects showing X-ray evidence for obscuration have clear AGN signatures in their rest-frame optical/UV spectra. Notably, such AGN can have both type 1 and type 2 optical/UV classifications. Most of these objects have z 1.5.
Optically faint X-ray sources. These sources have I 24 and thus usually cannot be identified spectroscopically. Many, however, appear to be luminous, obscured AGN at z 1-3 when their X-ray properties, optical photometric properties (including photometric redshifts), and multiwavelength properties are considered. Thus, these objects largely represent an extension of the previous class to higher redshifts and fainter optical magnitudes.
X-ray bright, optically normal galaxies (XBONGs). XBONGs have X-ray luminosities ( 1041-1043 erg s-1) and X-ray-to-optical flux ratios suggesting some type of moderate-strength AGN activity. Some also have hard X-ray spectral shapes suggesting the presence of X-ray obscuration. Optical spectra give redshifts of z 0.05-1, but AGN emission lines and non-stellar continua are not apparent. The nature of XBONGs remains somewhat mysterious. Some may just be Seyfert 2s where dilution by host-galaxy light hinders optical detection of the AGN, but others have high-quality follow up and appear to be truly remarkable objects. These "true" XBONGs may be (1) AGN with inner radiatively inefficient accretion flows, or (2) AGN that suffer from heavy obscuration covering a large solid angle ( 4 sr), so that optical emission-line and ionizing photons cannot escape the nuclear region.
Starburst galaxies. At the faintest X-ray flux levels in the deepest Chandra surveys, a significant fraction of the detected sources appear to be z 0-1.3 dusty starburst galaxies. They are members of the strongly evolving starburst population responsible for creating much of the infrared background. The observed X-ray flux appears to be the integrated emission from many X-ray binaries and supernova remnants.
"Normal" galaxies. Apparently normal galaxies are also detected in the deepest Chandra surveys out to z 0.2. The observed X-ray emission is again probably largely from X-ray binaries and supernova remnants; these objects and the starburst galaxies above are probably not distinct but rather constitute a single population of galaxies with star formation of varying intensity. Low-luminosity AGN are likely present in some cases as well. Some normal galaxies sport luminous X-ray sources clearly offset from their nuclei. At even fainter X-ray flux levels, normal and starburst galaxies should be the dominant class of extragalactic X-ray sources.
Most of the AGN found in X-ray surveys are "radio quiet" in the sense that the ratio (R) of their rest-frame 5 GHz and 4400 Å flux densities are R < 10.
Figure 4 shows some of the source classifications in the HDF-N, which is at the center of the CDF-N (see Figure 1) and thus has the most sensitive X-ray coverage available. Note, for example, that three of the brightest X-ray sources are XBONGs. These were not recognized as AGN prior to the Chandra observations, despite the many intensive studies of the HDF-N.
Figure 4. Chandra and HST images of the HDF-N. The 22 Chandra sources are circled on the HST image; the circles are much larger than the Chandra source positional errors. The numbers are source redshifts; redshifts followed by a "p" are photometric. Basic source type information for many of the sources is also given.
Luminosity and Redshift Distributions
The combined results from deep and wider X-ray surveys show that the sources comprising most of the XRB have X-ray luminosities comparable to those of local Seyfert galaxies, such as NGC 3783, NGC 4051, and NGC 5548 (e.g., see Figure 5). While a few remarkable obscured quasars have been found, these appear fairly rare and only make a small contribution to the XRB. Indeed, it appears that the fraction of obscured AGN drops with luminosity from 60-70% at Seyfert luminosities to 30% at quasar luminosities.
Figure 5. Luminosity in the 0.5-2 keV band (computed from the 0.5-2 keV flux assuming a power-law spectrum with a photon index of = 2) versus redshift for extragalactic sources in the CDF-N with spectroscopic redshifts. Sources with I = 16-22, I = 22-23, and I > 23 are indicated with filled circles, open circles, and stars, respectively. The dotted curve shows the approximate sensitivity limit near the center of the CDF-N. Also shown are the well-studied Seyfert 1 galaxy NGC 5548 (filled square) and Sloan Digital Sky Survey (SDSS) quasars from the SDSS Early Data Release with X-ray coverage in archival ROSAT data (small dots; the relevant solid angle covered by pointed ROSAT observations is 15 deg2). Note that NGC 5548 could have been detected to z ~ 10 in the CDF-N. Note also that the CDF-N and SDSS populations are nearly disjoint, as a consequence of the different solid angle coverages (a factor of ~ 120) and depths. Adapted from D.M. Alexander, F.E. Bauer, W.N. Brandt, et al., 2003, AJ, 126, 539; A.J. Barger, L.L. Cowie, P. Capak, et al., 2003, AJ, 126, 632; and C. Vignali, W.N. Brandt, & D.P. Schneider, 2003, AJ, 125, 433.
Most spectroscopically identified AGN in the deep X-ray surveys have z 2, although a significant minority have z 2-5. This is partly due to spectroscopic incompleteness bias, as is apparent by comparing the filled circles, open circles, and stars in Figure 5. However, as will be described further in Section 3.2, there is a real enhancement of AGN at z 1 relative to expectations from pre-Chandra AGN-synthesis models of the XRB. An impressive ~ 60% of the 2-8 keV XRB arises at z < 1.
AGN Sky Density
Most ( 70-100%) of the extragalactic X-ray sources found in both the deep and wider X-ray surveys with Chandra and XMM-Newton are AGN of some type. Starburst and normal galaxies make increasing fractional contributions at the faintest X-ray flux levels, but even in the CDF-N they represent 20-30% of all sources (and create 5% of the XRB). The observed AGN sky density in the deepest X-ray surveys is 6500 deg-2, about an order of magnitude higher than that found at any other wavelength. This exceptional effectiveness at finding AGN arises because X-ray selection (1) has reduced absorption bias and minimal dilution by host-galaxy starlight, and (2) allows concentration of intensive optical spectroscopic follow-up upon high-probability AGN with faint optical counterparts (i.e., it is possible to probe further down the luminosity function).
Completeness of X-ray AGN Selection
Are there significant numbers of luminous AGN that are not found even in the deepest X-ray surveys? This could be the case if there is a large population of AGN that are X-ray weak due either to absorption or an intrinsic inability to produce X-rays. This question can be partially addressed by looking for AGN found at other wavelengths that are not detected in X-rays. In the CDF-N, one of the most intensively studied regions of sky at all wavelengths, there are only 1-2 such AGN known. The most conspicuous is 123725.7+621128, a radio-bright ( 6 mJy at 1.4 GHz) wide angle tail source that is likely at z 1-2 (although the redshift of this source remains uncertain). This is one of the brightest radio sources in the CDF-N but has been notoriously difficult to detect in X-rays. Manual analysis of the 2 Ms Chandra data at the AGN position indicates a likely, but still not totally secure, detection (at a false-positive probability threshold of 3 × 10-5 using the standard Chandra wavelet source detection algorithm). The 0.5-2 keV luminosity is 5 × 1041 erg s-1. The only other known AGN in the CDF-N without an X-ray detection is 123720.0+621222, a narrow-line AGN at z = 2.45; its 0.5-2 keV luminosity is 2 × 1042 erg s-1.
Despite the spectacular success of X-ray surveys at finding AGN, appropriate humility is required when assessing the AGN selection completeness of even the deepest X-ray surveys. This is made clear by consideration of "Compton-thick" AGN, which comprise a sizable fraction ( 40%) of AGN in the local universe. Such AGN are absorbed by intrinsic column densities of NH >> 1.5 × 1024 cm-2, within which direct line-of-sight X-rays are effectively destroyed via the combination of Compton scattering and photoelectric absorption. Such AGN are often only visible via weaker, indirect X-rays that are "reflected" by neutral material or "scattered" by ionized material. (5) Many of the local Compton-thick AGN (e.g., NGC 1068, NGC 6240, Mrk 231), if placed at z 0.5-1.5, would remain undetected in even the deepest Chandra surveys. Thus, it appears plausible that 40% of AGN at such redshifts may have been missed (the number, of course, could be higher or lower if the fraction of Compton-thick AGN evolves significantly with redshift). Deeper observations with Chandra ( 10 Ms; see Section 4.1) may be able to detect the indirect X-rays from any missed Compton-thick AGN, and observations with Spitzer may be able to detect "waste heat" from such objects at infrared wavelengths.
Another way to address AGN selection completeness in X-ray surveys is to consider "book-keeping" arguments: can the observed sources explain the observed 20-40 keV XRB intensity, and can all the observed accretion account for the local density of supermassive black holes? The answer is plausibly "yes" in both cases, but with some uncertainty. In the first case, one must make a significant spectral extrapolation from 5-10 keV and worry about mission-to-mission cross-calibration uncertainties. In the second, significant uncertainties remain in bolometric correction factors, accretion efficiencies, and the local density of supermassive black holes. The current book-keeping arguments cannot rule out the possibility that a significant fraction of the AGN population (e.g., Compton-thick AGN) is still missed in X-ray surveys. Indeed, some book-keepers find better agreement with the local black-hole mass function after making a substantial correction for missed accretion in Compton-thick AGN.
5 In some "translucent" cases, where the column density is only a few × 1024 cm-2 (i.e., a few Thomson depths), direct "transmission" X-rays from a Compton-thick AGN may become visible above rest-frame energies of ~ 10 keV. For comparison, the column density through your chest is ~ 1 × 1024 cm-2; if you stood along the line-of-sight to an AGN, you could almost render it Compton thick! Back.