2.3. Radio Selection
Radio selection was the original way to find AGN. It is very powerful, since almost all luminous radio sources are AGN in our definition of radiating supermassive black holes - the only contamination is at lower luminosities, where there can be significant radio emission from rapidly star-forming regions. Thus, for most luminous objects, mere detection in the radio indicates the presence of an AGN. Radio surveys are also rather sensitive, and the positions are extremely accurate. The morphological information is also important, since all radio sources showing double-lobe or jet-like structures are AGN.
However, radio surveys are very incomplete, since less than 10% of AGN have luminous radio emission (White et al. 2000), the ratio of bolometric luminosity to radio luminosity varies by a factor of 105, and the fraction of the total energy radiated in the radio is small. While it is true that most "radio-quiet" AGN have some radio emission, the relative intensity is very low, making searches for these objects in the radio very difficult. Even the early radio samples found objects with a wide range of optical properties and a wide redshift distribution. The absence of a correlation of radio flux with redshift indicated a broad luminosity function.
It was noted in the 1960s that a significant fraction of the radio sources did not have broad lines or had no lines at all and that their optical colors were not unusual (Kristian, Sandage, & Katem 1974). To quote from Ulrich (1971), "Comparison of the optical and radio properties of nuclei of galaxies shows that galaxies with compact nuclear radio sources are more likely to have optical spectra with emission lines than are galaxies without central radio sources", thereby indicating the relative rareness of broad or strong lines in radio-selected AGN. This has been quantified recently in a combined study of the 2dF survey with a southern radio survey (Sadler et al. 2002; Magliocchetti et al. 2002), as well as in SDSS and FIRST (Faint Images of the Radio Sky at Twenty-cm) data (Ivezic et al. 2002). In these data, over 60% of nearby radio galaxies show no evidence for strong emission lines of any width, and their optical continua show weak or no evidence for non-thermal activity. This is also true at higher redshifts, where the radio sources are not even found to be X-ray sources (Cowie et al. 2004a). In the SDSS data, only 10% of the optical counterparts of radio sources show evidence for strong AGN-like emission lines. However, in several of the nearby objects for which no obvious nucleus or broad lines exist in ground-based data, HST imaging finds evidence for weak non-thermal continua (Chiaberge et al. 2003). In fact, the absence of strong lines in the radio galaxies made optical spectral identification so hard that it required more than 30 years to completely identify the first radio catalog (Spinrad et al. 1985). It is quite interesting that the vast majority of radio AGN would never be selected by optical techniques. This was the first indication that different selection techniques tend to find different objects. To rephrase this, while only 10% of optically-selected AGN are "radio loud", less than 30% of radio-selected AGN are "optically loud".
The "radio" colors of AGN allow discrimination against other sources of radio emission for luminous objects. Almost all flat-spectrum, compact radio sources are AGN. In addition, the strong correlation of mid-IR with radio flux found for star-forming galaxies allows the detection of AGN by exclusion, if one also has IR data (Yun, Reddy, & Condon 2001). That is, the objects with low IR-to-radio luminosity ratios are almost all AGN (Condon, Anderson, & Broderick 1995), and those objects that lie in a narrow range of high IR-to-radio luminosity ratios are starbursts (or radio-quiet AGN). This technique has been used to show that many of the objects thought to be starbursts in the ROSAT All-Sky Survey data were really previously uncataloged AGN (Condon et al. 1998). However, inspection of a large sample of X-ray-selected objects (the most famous of which is NGC1068) shows many of them with "star formation" IR-to-radio ratios, indicating that most of the IR and radio luminosities are not AGN related. It thus requires a good deal of care (Miller & Owen 2001) to select the AGN, which tend to have lower relative far-IR-to-radio luminosities and larger scatter in the correlation than those seen for star-forming galaxies.
In an unbiased survey of radio-selected objects using the SDSS (Ivezic et al. 2002), the number of starburst galaxies outnumbers those objects with AGN-like optical spectra by 5:2, and the number of objects without any unusual optical properties dominate the optically-active AGN by 12:1. It thus seems that deep radio surveys have a large contamination factor from non-AGN populations but discover lots of AGN that would not be selected with classical optical criteria.
Most radio data have fairly accurate positions (better than 45" for the largest radio survey, the NRAO VLA Sky Survey or NVSS, and better than 7" for the most sensitive large solid angle survey, FIRST), allowing counterparts in other wavelength bands to be readily identified. However, only ~ 8% of optically-selected AGN (Ivezic et al. 2002) are radio "loud" and selectable in radio surveys.