2.5. High-Energy Selection

The other two techniques used to find active galaxies are X-ray and -ray-selection. Like radio emission, luminous, compact X-ray and -ray emission is an almost certain indicator of the existence of an AGN and does not need "confirmation" by data in other wavelength bands. The relatively low signal in most X-ray surveys makes X-ray spectroscopic redshift determinations difficult, and for most objects one must rely on optical redshifts. However, in a few recent cases, X-ray redshifts have been determined (Hasinger 2004). The need to obtain optical redshifts has driven the entire field of "optical identification", in a fashion similar to that of the identification of radio surveys.

It was realized in the early 1970s (Pounds et al. 1975) that about one-quarter of the high-latitude X-ray-selected objects in the Ariel-V X-ray survey could be identified with previously known AGN. Given the large uncertainties in the positions of these objects (~ 0.5 - 2 deg²), this was quite surprising. Detailed follow-up work of the unidentified, high-latitude X-ray sources in the early surveys (Ward et al. 1980; Wilson 1979) discovered that most of them were previously unknown AGN with properties that "hid" them from optical surveys. This work was strengthened with the first accurate X-ray positions (Griffiths et al. 1979), which confirmed that objects with rather weak or narrow optical lines and no evidence for a non-thermal continuum could be luminous X-ray sources. The advent of X-ray imaging with the Einstein Observatory in 1979 vastly increased the efficiency and accuracy of X-ray surveys, but, with the exception of low-redshift objects, the positions were still not accurate enough to have an unique optical counterpart. This drove a very large program of identification (the Einstein Medium-Sensitivity Survey or EMSS; Gioia et al. 1990), which took several years to complete.

A major difference in the early X-ray and optical surveys was the redshift distribution, with optical selection criteria finding many objects over a wide redshift range out to z ~ 3, while X-ray samples were much more concentrated at z < 1. Even the early X-ray samples found very little correlation of redshift with flux.

So far, while -ray emission is certainly an almost unique feature of high-latitude active galaxies, the positions are too poor to allow the identification of the objects. Based on correlations with radio data, most of the identified AGN are blazars or flat-spectrum radio sources, and very few, if any, classical Seyfert 1 galaxies or quasars are found in the Compton Gamma-Ray Observatory surveys.

It has been known for over 35 years that the vast majority of high-latitude, point-like "hard" X-ray sources are AGN ⁴ (Pounds 1979), and this is one of the most efficient and least "error prone" ways of selecting AGN. At the present time, there are very few, if any, known AGN that are not also luminous X-ray sources. However, because of the wide variance in the SEDs of AGN, there are many objects which do not have sufficiently sensitive X-ray data to decide if they really are X-ray "quiet" objects (Leighly, Halpern, & Jenkins 2002; Gallagher et al. 2001). There are classes of AGN (in particular, broad absorption-line quasars and strong FeII objects; Lawrence et al. 1997) that have rather low soft X-ray-to-optical ratios. At present, this is thought to be due to the high column densities in these objects (see Section 3), but work is still proceeding on this. Recent results suggest that there is a class of IR-selected AGN with broad optical emission lines that are X-ray "weak" (Wilkes et al. 2002).

The efficiency of X-ray surveys is very high, finding considerably more AGN at a fixed optical magnitude than other techniques. The asymptotic limit of optical surveys is ~ 130 AGN deg^-2 at B < 23 mag (Palunas et al. 2000), while the Chandra surveys find ~ 1000 deg^-2 at r < 24 mag. This relative efficiency is true even for very optically-bright samples; for example, Grazian et al. (2000) finds three times more AGN with v < 14.5 in the ROSAT All-Sky Survey than does the color-selected Palomar-Green (PG) bright quasar survey (Schmidt & Green 1983). In the 2 - 8 keV band at fluxes between 10^-15 to 10^-10 ergs cm^-2 s^-1, essentially all of the point-like sources are AGN. At lower fluxes, the fraction of star-forming galaxies increases, but the X-ray counts are still dominated by AGN over the range to which Chandra has so far reached.

It is only recently with Chandra and XMM-Newton data that the X-ray/optical/near-IR distribution of normal galaxies has been measured, and thus an estimate of the AGN "contamination factor" derived. The bottom line is that all compact X-ray sources with a luminosity above ~ 10⁴² ergs s^-1 (2 - 10 keV) are considered to be AGN. The main interlopers are "ultraluminous X-ray sources" (ULXs; Colbert & Ptak 2002), which are objects with X-ray luminosities between 4 × 10³⁹ and 10⁴² ergs s^-1 and do not reside in the nuclei of the host galaxies. While easily recognized at low redshifts, Chandra imaging is required to separate these out at z > 0.1 (Hornschemeier et al. 2003). At X-ray luminosities below 10⁴² ergs s^-1, there is always the possibility that a nuclear source might really be an ULX rather than an AGN, as is apparently the case in M33 (Long, Charles, & Dubus 2002). However, it is not clear if the distinction is meaningful if the ULXs turn out to be supermassive black holes. Starburst galaxies have a considerably lower X-ray-to-optical flux ratio than AGN, and while they frequently harbor ULXs (Grimm, Gilfanov, & Sunyaev 2003), their total luminosities rarely reach 10⁴² ergs s^-1 and are well correlated with their IR luminosities (Ranalli, Comastri, & Setti 2003).

For "typical" soft X-ray-selected AGN, the X-ray-to-B-band flux ratio is ~ 1, with a full range of ~ 100 and a variance of ~ 6 (Mushotzky & Wandel 1989; Anderson et al. 2003). In the 2 - 10 keV X-ray band, there is a broader range of X-ray-to-optical flux ratios, with a significant tail of very high X-ray-to-optical ratios that extends out to > 10⁴ (Akiyama et al. 2003; Barger et al. 2003b; Comastri et al. 2003; see also Comastri, this volume).

Because of the broad wavelength coverage of the X-ray band (about a factor of 100 in wavelength), there is a significant difference between soft-band (0.2 - 2 keV), hard-band (2 - 6 keV), and very hard-band (5 - 10 keV) surveys. The soft-band surveys can suffer from the same extinction effects as UV and optical surveys, while in the harder bands, absorption is a much smaller effect. Based on ROSAT and Chandra data, the soft and hard X-ray selection tends to find different objects. The soft X-ray selection preferentially finds broad-line quasars and narrow-line Seyfert galaxies (these objects tend to have a bright, soft spectral component, in addition to a flat, high-energy, power-law spectrum). The hard X-ray selection finds, in addition to the classical Seyfert 1 galaxies and quasars, large numbers of objects with weak or absent optical emission lines and lacking non-thermal nuclei. It is believed that this selection effect may be related to the presence of absorbing material, but other possibilities exist (see Section 3). Large follow-up programs for X-ray-selected AGN find other selection effects. The median value of the soft X-ray-to-optical flux ratio changes by a factor of 2 - 3 over four orders of magnitude in optical flux (Anderson et al. 2003). Thus, high-luminosity objects are somewhat weaker in the X-ray band for a given optical luminosity. On the other hand, radio-loud AGN have a factor of three higher X-ray-to-optical flux ratio than radio-quiet objects (Worrall 1987). Thus, soft X-ray selection compared to optical selection is biased towards radio-loud, lower luminosity objects. While there is a fair range in the X-ray spectral slopes of AGN (the variance is ~ 0.2), there is no correlation of slope with luminosity or redshift (Vignali et al. 2003; Reeves et al. 1997).

Before the Chandra and XMM-Newton surveys, X-ray error circles were often large (e.g., > 15"), engendering large optical follow-up programs and allowing a bias in the selection of the optical "counterpart". However, Chandra and XMM-Newton follow-up of ROSAT identifications (McHardy et al. 2003) show that virtually all of them were "right", and, thus, this possible selection effect is small.

⁴ In a historical point, before the Ariel-V results, there were strong indications from Uhuru data of a class of "unidentified high galactic latitude X-ray sources" (Holt et al. 1974). Arguments - later proved false - were put forward that they could not be Seyfert galaxies. Back.