4. Specific classes
The FIRST radio survey is a factor of more sensitive than other large-scale radio surveys, deep enough to see transitional populations, such as quasars intermediate between typically `radio-loud' to `radio-quiet' objects.
Quasars, the largest individual group of objects in our sample, are traditionally divided into a radio-quiet and a radio-loud class, with apparently markedly different properties (Kellerman et al. 1989, B97, Yuan et al. 1998).
Fig. 11 shows the radio-loudness distribution of all quasars. The top line represents all quasars, while the hatched region denotes newly classified quasars. Classical radio-loud quasars typically have (see Fig. 16 of B97). The histogram shows that most RASS-FIRST quasars have values near the radio-quiet - radio-loud transition and there is no evidence for a radio-loud / radio-quiet bimodality. These results are consistent with those of the FIRST Bright Quasar Survey (FBQS; W00) which showed that the distribution of R for a large sample FIRST-selected quasars falls steadily from to and that when the insensitivity of the FIRST survey to quasars with is taken into account, the distribution of R should rise smoothly down to (see Fig. 15 in W00). Previously known quasars exhibited a bimodal distribution in R. W00 conclude that the FBQS result is not due to selection effects arising in the FIRST survey, but rather the sensitivity of the FIRST survey to sources with intermediate values of R. Our data show there are a large number of `previously known' (X-ray bright) quasars at , i.e., in the intermediate regime between radio-loud and radio-quiet. With a few exceptions, these objects have very low 1.4 GHz fluxes (1 - 3 mJy) and previously were not associated with any known radio source.
4.1.1. Spectral variations
The average X-ray power-law slopes of the RASS/FIRST quasars are steeper than those of radio-loud quasars, but flatter than those of radio-quiet quasars (Sect. 3.2.1). Previously, it was thought that the X-ray photon index correlates with radio-loudness R (Williams et al. 1992), however, in B97 it was argued that radio-loudness, core dominance, and radio spectral slope are, in the framework of unification schemes, similar indicators for the orientation of the source and thus the principal origin of the correlations between the X-ray spectra and the radio properties of the quasars remains unknown. A further complication is introduced by redshift-dependent effects, caused by different redshift distributions of the samples.
The present study's quasars constitute a low-redshift sample: 107 of the 146 total objects are at and only four are at (one of them is the z=4.71 object RXS J1430.4+4204). A regression analysis, disregarding six quasars without redshift information and the extreme z=4.71 object, shows that the slopes follow a redshift dependence like . The errors are at 95% confidence and the `no regression' hypothesis is rejected at the 97.4% confidence level. These values are very similar to the results found for radio-quiet quasars (, Yuan et al. 1998), however with larger errors due to the smaller sample size. Classical radio-loud quasars show a smaller value for (; B97).
If radio-loud quasars are physically different objects from radio-quiet quasars, the photon index should not depend on the orientation. A correlation between X-ray slope and radio-loudness, where the values of R cover the transition between radio-loud and radio-quiet objects, might indicate an intrinsic physical transition between these two classes. Indeed, as suggested from Fig. 12 there exits a correlation between these quantities: and the `no-regression' hypothesis can be rejected with 99.79% confidence, although the scatter of the photon indices as well as the errors of the individual slopes are rather large. A correlation between and R might be introduced by redshift effects: varies with redshift due to the intrinsic curvature of the spectrum and the radio-loudness correlates with z (, at 99.9% confidence). This is, at least partly, a result of the flux-limits of the radio selection such that at larger redshifts only the radio-loudest objects are detected. To assess quantitatively the extent of a spurious correlation introduced by distance effects, the partial linear correlation coefficient R (e.g. Hald 1962, Kembhavi et al. 1986) can be used. We determined the partial correlation coefficients R with the effect of redshift eliminated and found that the no-regression hypothesis is rejected at the 17.8% confidence level only, which means that the - correlation could be just a redshift dependent selection effect.
Finally, we performed a regression analysis for both the extreme ends of the - distribution, checking the radio-quiet and the radio-loud sample separately. For the objects with we obtain with the `no-regression' hypothesis rejected at the 64.9% confidence level only. For the radio-loud quasars () we get at a 5.1% confidence level. Both the low confidence levels as well as the large errors indicate that we are seeing the two different slopes of the radio-quiet and radio-loud quasar population, with nearly the same average values as found for the two classes previously (B97, Yuan et al. 1998).
4.1.2. Luminosity correlations
A knowledge of the exact form of the luminosity correlations and is required to relate the quasar statistics (e.g., evolution, luminosity function) in the different wave-bands and to understand the quasars' broad-band emission. Early studies usually found values for the slopes of , but follow-up programs produced results which differed markedly and also seemed to depend not only on the quasar selection criteria but also on the exact correlation method (see the discussion by Padovani 1992, Franceschini et al. 1994, and La Franca et al. 1995). B97 and Yuan et al. (1998) showed convincingly, by using orthogonal direction regression (ODR) analysis, that and that the value obtained for depends on the core dominance of the radio emission. For highly beamed sources, the X-ray luminosity correlated linearly with the core radio flux but in general both the core flux as well as the total radio flux contribute to the correlation. Further, using a generalized ODR analysis which takes into account measurement errors as well as allowing for intrinsic variances (Fasano & Vio 1988, (FV88)), the correlation yielded non-zero intrinsic variances in all cases, i.e., the data scatter (intrinsically) around the regression line.
Fig. 13 presents the correlation between the monochromatic (in erg s-1 Hz-1) X-ray and radio luminosities, where radio-loud objects ( are indicated by open symbols, and radio-quiet quasars ( by filled symbols. Applying the FV88 ODR analysis, and assuming a typical error for the radio flux of 5% and for the X-ray flux of 30% (B95), the radio-loud subgroup shows a correlation of the form , with an intrinsic variance . For the radio-quiet objects, this correlation is , with an intrinsic variance of . The slope for the radio-loud objects is very similar to that found in B97 for the correlation between the X-ray and total radio luminosity. This, as well as the large value of the intrinsic variance, indicates an inhomogeneous population of resolved and unresolved radio sources. The slope of is a strong indicator that in radio-quiet quasars the radio emission is a direct tracer of the nuclear activity, as in radio-loud quasars with high core dominance. This is consistent with the results of a VLA study of nearby low-redshift radio-quiet quasars where Kukula et al. (1998) find that the radio emission originates in a compact nuclear source, directly associated with the central engine of the quasar and that radio-quiet quasars generally have steep spectral indices ().
The most radio- and X-ray luminous object is the z=4.715 quasar B3 1428+422, a known radio-loud ROSAT source (B95) only recently classified as a quasar (Hook & McMahon 1998). It appears that at a given radio luminosity, radio-quiet quasars have a larger X-ray luminosity, however, at a given radio luminosity, a radio-quiet quasar is optically brighter than a radio-loud quasar. And indeed, the correlation (Fig. 14) indicates that the X-ray luminosity scales similarly with optical luminosity for both types of quasars (), with the well-known tendency that radio-loud quasars are X-ray brighter at a given optical luminosity (B97).
This raises the question which of the three energy bands provides the most direct measure of a quasar's activity. Table 5 lists the average monochromatic luminosities (in erg s-1 Hz-1) of the two quasar populations (radio-loud vs. radio-quiet). The optical luminosities differ by a factor of order unity, and the X-ray luminosities of radio-loud quasars are higher by a factor of 3 (and depend further on the the radio properties of the objects, B97). The radio-luminosities are differ greatly and seem to be the main separator of the two classes. Selection effects and the small sample sizes do not allow more quantitative conclusions.
Table 5. Average Monochromatic Quasar luminosities (in erg s-1 Hz-1)
The two classes of quasars populate distinct regions in the phase space - the dividing plane is for all . Any two-dimensional plot is thus only a projection of this distribution. Apart from the above mentioned different averages, the radio-loud quasars show a more `compact' distribution, while the radio-quiet objects have a larger dispersion.
The relatively large intrinsic dispersion of obtained from the FV88 ODR regression analysis for the correlation is directly visible in Fig. 14 and is mainly caused by low-luminosity radio-quiet objects. Most of the `optical outliers' at low X-ray luminosities in the figure are newly classified quasars. While it is possible there is a systematic overestimation of the source's optical brightnesses by as much as 0.3 mag (see W00), this alone cannot account for the large offsets. Other possible causes are incorrect K-corrections due to unusual spectral slopes, strong variability of the X-ray emission, or genuine luminosity differences. Previous ROSAT studies of quasars (B97, Yuan et al. 1998) also found a small fraction of objects which seem to deviate significantly from the usual trend.
The second largest group of objects in the RASS-FIRST correlation are the 99 galaxies. We lack complete information for all of these, and therefore limit our discussion to the 59 galaxies for which all fluxes and redshifts are available and believed to be reliable.
Fig. 15 shows the X-ray versus radio luminosity. Galaxies extend to lower luminosities from the quasar population (Fig. 14), with a general trend of lower X-ray luminosities at similar optical luminosities, although with some overlap. The most luminous X-ray galaxy is RGB J1514+366. The radio position coincides with the z=2.723 protogalaxy cB58 which would result in an extreme X-ray luminosity from the source of erg s-1. It appears likely that the dominant part of the X-ray emission is either from the positionally coincident cluster of galaxies MS 1512.4+3647 or, as proposed by Hamana et al. (1997), the X-ray flux from cB58 is amplified by gravitational lensing from the foreground cluster.
Two other galaxies with unusually high radio and X-ray luminosities are RXS J1625.5+2705 (87GB 1623+2712) and RXS J1317.3+3925 (B3 1315+396). Both are at high redshift indicating the galaxy classification could be in error. Classifications for both these sources come from NED, and we have no independent measurement of these objects' optical spectra. RXS J1625.5+2705 is classified by Stocke et al. (1991) as an `AGN' and B3 1315+396 seems to be a genuine quasar (Vigotti et al. 1990). Even after eliminating these three objects from consideration, the galaxy as well as the relations do not indicate any correlation. All galaxies (with the exception of the above three suspicious cases) are found at redshifts below z = 0.3. While the as well as the plots display cutoffs due to the flux limits of the corresponding surveys, the optical luminosities are unaffected by redshift related selection effects.
4.3. BL Lacs
A total of 71 objects are classified as BL Lacs or as possible BL Lacs. Redshift information is available for 41 of these, including 12 previously known BL Lacs. The newly discovered objects have generally lower X-ray and radio luminosities than the previously known sources. As suggested in Fig. 9, the flux ratios indicate that the `previously known' BL Lacs populate the traditional `X-ray selected' BL Lac region while the new objects belong to the `radio-selected BL Lac' branch and extend into the region of phase space traditionally occupied by galaxies.
This transitional nature is demonstrated in Fig. 16 where we show the distribution of the radio - to -X-ray spectral index, , of various well known BL Lac samples, including the `radio selected' 1 Jy BL Lacs (RBLs) and the `X-ray selected' EMSS BL Lacs (XBLs). The dashed line represents the classical division between X-ray and radio-selected BL Lacs (Padovani & Giommi 1996). The FIRST-RASS BL Lacs (thick line) are mainly `intermediate' objects like those of the RGB sample (Siebert et al. 1999, Laurent-Muehleisen et al. 1999) clearly demonstrating that at least part of the previously claimed bimodality of the BL Lac population must be attributed to selection effects. The soft X-ray spectral indices are found in the same range as the `classical' BL Lacs (Fig. 8 and Fig. 2 of Brinkmann et al. 1996). The luminosities of the current sample extend to lower values both in and than the XBL population in Fig. 3a of Brinkmann et al. (1996), while the optical luminosity is on average higher than the optical luminosities of previously known the XBLs (Fig. 3b of Brinkmann et al. 1996).
The remaining large subclass contains 117 AGN, categorized as `broad emission line AGN' (`BA' in Table 1: 48 objects), `narrow emission line AGN' (`A': 21 objects), and 28 previously classified Seyfert galaxies of various types, as found from NED. We also group the 20 starburst galaxies (`H' in the tables) into the present discussion.
The lowest X-ray luminosities ( erg s-1 Hz-1) are exhibited by the narrow-line objects RXS J0316.0-0226, RXS J0919.0+2616, RXS J1204.7+3110, and RXS J1258.6+2736. These sources are classified in NED as galaxies, with various peculiarities which would justify a more detailed study. In addition, the Seyfert 1 galaxy RXS J 1220.1+2916 has an X-ray luminosity below the lower plot boundary. The only other broad-line AGN with such a low X-ray luminosity in Figs. 17 or 18 is RXS J1140.2+2441 which is also classified as a `galaxy' in NED. However, all these objects show emission-line properties which indicate AGN-activity, but their low luminosity undoubtedly signals a very weak nucleus.
Another abberant group of objects are some extremely X-ray and radio luminous Broad-lined AGN, like RXS J0035.9-0912, RXS J0746.0+2226, and RXS J2228.9-0753, all at rather high redshifts . From their deduced absolute optical magnitudes they are borderline objects, still belonging to the AGN class, but with X-ray and radio luminosities like low luminosity radio-loud quasars.
There are clear correlations between the X-ray and the radio luminosities for all classes. Interestingly the starbursts ("H") are narrowly confined in phase space by about an order of magnitude in both luminosities while the other groups exhibit a much larger dispersion. The X-ray luminosity shows no dependence on optical luminosity for any of the classes (Fig. 18). The narrow emission line sources and the Seyferts span the entire optical luminosity range between whereas the broad-line objects and the starbursts are considerably optically fainter and exhibit a narrow distribution of optical luminosities , where the luminosities are given in erg s-1 Hz-1. However, the narrow-line objects have the on average lowest X-ray luminosities.
4.5. Unclassified sources
Nearly half of the objects in the sample are presently spectroscopically unclassified. The obvious reason for this can be seen immediately in Fig. 19, where we plot the E-magnitude versus the logarithm of the 1.4 GHz flux for all objects. Open circles are classified sources, solid circles are unclassified. Most of the unclassified objects are at faint optical magnitudes, directly indicating a selection bias. Some unclassified objects are rather bright: in most cases they are confused regions requiring a detailed identification effort or the positions are close to bright stars, complicating spectroscopic observations.
Interestingly, the flux ratios (Fig. 10) show that most of the unclassified objects belong to the region of phase space typically populated by BL Lacs. These objects' average X-ray spectral indices are also similar to those exhibited by BL Lacs. The large dispersion, however, indicates a mixture of object classifications. Finally, a certain number of the formally `classified' objects must strictly be regarded as `unclassified', for example, galaxies that might be members of clusters or objects classified as `galaxies' purely by their appearance on optical plates, without spectroscopic confirmation.
© European Southern Observatory (ESO) 2000
Online publication: April 10, 2000