Source Classification

The definition of the blazar class has varied since the 1978 Pittsburgh conference, where the terminology was first suggested. The original definition of the class (cf. Angel & Stockman 1980) emphasized the dominance of a highly polarized, variable, nonthermal continuum over other properties. But in the last twenty years, the definition of the class as a whole, as well as various subclasses, has varied, partly as a result of observational selection. A variety of names (e.g., HPQ or highly polarized quasar; OVV or optically violent variable) have been applied to some objects, usually based upon finding extreme values of one or more of the signal properties of the blazar class (§ 1). A more commonly used set of subclasses are based upon the character of the optical spectrum: FSRQ (flat-spectrum radio quasar, for objects with emission-line dominated spectra) and BL Lac (nearly lineless objects). Other authors have restricted the term ``blazar'' to those with emission-line spectra.

Further sub-divisions have been invented to describe objects found as a result of X-ray or radio surveys, or with certain broadband spectral shapes (e.g., X-ray selected BL Lacs or XBL, radio selected BL Lacs or RBL, low-energy peaked BL Lacs or LBL, high-energy peaked BL Lacs or HBL; see Padovani & Giommi 1995a and Urry & Padovani 1995). While the latter two are at least based upon a strictly defined spectral shape (see §6.2), all point out the difficulties inherent in defining the properties of a class based upon single-band surveys which cover fairly small ranges of flux in their survey band. Since the blazar population spans over seven decades of luminosity in the radio, optical and X-ray band, and over four decades in its ratio of X-ray to radio luminosity, single-band surveys are unable to representatively sample the blazar population, particularly when the dynamic range of fluxes being surveyed (i.e. F_max / F_lim) is less than 100.

The result is a confusing array of nicknames which are utilized with abandon in today's literature. The physical meaning of many of these divisions is not at all clear. For example, much has been written about the temporary appearance of broad H$\alpha$ with $W_\lambda \approx 6$ Å, in the spectrum of BL Lac (Vermeulen et al. 1995). Similar occurrences have been noted in other objects, including 0846+513 (Arp et al. 1979), 0537-441 (Peterson et al. 1976), and 0215+015 (Boisse & Bergeron 1988). Indeed, the recent results of Sambruna et al. (1996) and Scarpa & Falomo (1997) suggest that the separation between BL Lac and FSRQ may be rather ill defined and perhaps of questionable physical meaning. And while it is true that the properties of BL Lacs found in radio surveys (which mostly have values of $\alpha_{{\rm rx}} \lower.5ex\hbox{$\; \buildrel \gt \over \sim \;$}0.8$) are considerably different from those found in X-ray surveys (which mostly have lower values of $\alpha_{{\rm rx}}$), the explanation for this difference is controversial, and has been the subject of some debate in the literature (e.g., Padovani & Giommi 1995a; Fossati et al. 1997; Georganopoulos & Marscher 1997).

A variety of deeper, multiwavelength surveys for blazars (of which DXRBS is one) are currently underway. The X-ray based surveys, DXRBS (this paper), REX (Wolter et al. 1997; Maccacaro et al., in preparation), HQS/RASS (Nass et al. 1996), RC (Kock et al. 1996), RGB (Laurent-Muehleisen et al. 1997), and RASS/NVSS (Giommi, Menna & Padovani, in preparation) take as their starting point either the pointed ROSAT database (REX, DXRBS) or the all-sky survey (HQS/RASS, RGB, RC, RASS/NVSS), and make up their candidate list via cross-correlations with radio survey lists or other properties. The radio-based surveys emerging from the FIRST project take the radio-selected FIRST sample as their starting point, and use variability, polarization and optical colors to select candidates (Gregg et al. 1996; Laurent-Muehleisen et al. in prep). These projects are complementary, using different techniques to sample different regions of parameter space. These surveys will both sample the parameter space available to blazars more deeply and fill the holes left by previous, disjoint selection techniques. As they do so, we will gain the first complete picture of the range of properties encompassed by the blazar class. Given the confusing array of names currently in use (which may or may not be physically meaningful), these surveys (once completed) will need to resystematize the blazar definition, as well as those of its subclasses. However, neither this survey nor any of the other new X-ray or radio-based surveys can yet undertake the task of resystematizing the classification of flat-spectrum radio sources, as the identification of their samples are all incomplete. To do so at this time would risk not only confusion, but the real possibility of missing a population still extant within the unidentified objects.

For the present paper, we will adopt a form of the FSRQ-BL Lac dichotomy, basing our classifications solely upon the optical spectrum. We will apply the term ``blazars'' to both BL Lacs and FSRQs, since recent evidence (Fugmann 1988; Impey, Lawrence & Tapia 1991; Kühr & Schmidt 1990; Jannuzi et al. 1993, 1994) has shown that the properties outlined in § 1 are shared by both FSRQs and BL Lacs. We adopt the modified form of the BL Lac definition advocated by Marchã et al. (1996; see their Fig. 6) to classify BL Lacs and radio galaxies.

Their starting point is that the line luminosity seems to be independent of the observed continuum in blazars (see Koratkar et al. 1998 for an example of this behavior in 3C 279). It then follows that the Ca H & K break contrast, a measure of the presence of non-thermal continuum in a galaxy [defined by C = (f₊ - f_-) / f₊, where f₊ and f_- are, respectively, the flux redward and blueward of the Ca break], and equivalent width $W_\lambda$ will be correlated (the lower C, i.e., the higher the non-thermal contribution, the lower $W_\lambda$). Thus, changing the viewing angle and/or the luminosity of the BL Lac relative to that of the host galaxy will move an object on a diagonal trajectory in the contrast - equivalent width plane. Marchã et al. (1996) showed convincingly that objects in a triangular area limited by contrast C = 0.4 (breaks with $C \sim 0.5$ are typical of elliptical galaxies; Dressler & Schectman 1987) and the diagonal line shown in Figure 1 (which assumed the line and galaxian continuum emission of 3C 371 as its starting points and a smoothly decreasing AGN contribution) should still be called BL Lacs. Note that this expands upon the ``classical'' definition used by previous authors (Stickel et al. 1991; Stocke et al. 1991; Perlman et al. 1996a) of equivalent width $W_\lambda < 5 (1+z)$ Å  for all emission lines and Ca H & K break strength C < 0.25. The fuzziness of the previously used criterion is further illustrated by the occasional observation of broad H$\alpha$ lines in the spectra of several famous BL Lacs (among them Mkn 501 and BL Lac; see, for example Vermeulen et al. 1995). Objects which fall outside the traditional definition of the BL Lac class, but within the Marchã et al. definition which we adopt, will be discussed individually in § 4, and we will return to the subject in § 7 when selection effects are discussed.

Thus, objects which meet the Marchã et al. criteria are classified herein as BL Lac objects. Objects with higher-equivalent-width emission lines which are still narrow (FWHM $\lower.5ex\hbox{$\; \buildrel < \over \sim \;$}1000-2000$ km/s), or stronger Ca H & K breaks, are classified as radio galaxies, and are discussed individually in § 4, and as a group in § 7. All flat radio spectrum objects with higher equivalent widths and broad emission lines (FWHM $\lower.5ex\hbox{$\; \buildrel \gt \over \sim \;$}1000-2000$ km/s) are classified as FSRQs. It is important to note that this classification is being applied without regard to any other characteristic, such as redshift, presence or lack thereof of a stellar continuum, or physical extent. As a result, a number of objects which have been called broad-line radio galaxies by other authors are included as FSRQs within the previously-identified portion of our sample. We return to this last topic in § 7.

Due to the many commonalities shared by all blazars, we believe that a unified approach to these enigmatic objects is more helpful in helping us understand them, similar to that taken for radio galaxies by Baum, Zirbel & O'Dea (1995) and Zirbel & Baum (1995). Therefore, while we will use the classical definition to classify sources here (in order to ensure easy compatibility with past studies), it is our goal in future works to consider the equivalent width and luminosity of emission lines simply as additional variables in the analysis.