Conclusions

While the DXRBS sample is not yet completely identified (the objects discussed in this paper represent $\sim 60\%$ of our object list), this paper has detailed a number of interesting and exciting results from our deep survey. Most prominent among these results are:

1. A very high efficiency (95%) at finding FSRQs and BL Lacs once the list of radio-X-ray sources found by a cross-correlation of the ROSAT WGACAT with single-dish radio catalogs has been limited to serendipitous flat radio spectrum sources ($\alpha_{\rm r} \leq 0.70$).

2. The DXRBS sample has vastly expanded coverage of the low luminosity end of the luminosity function both for BL Lacs and FSRQs, compared to all previous samples of blazars. Twenty-eight of 135 DXRBS FSRQs are at $L_R < 10^{33.5} {\rm ~erg ~s^{-1} ~Hz^{-1}}$, compared to only 12 of 383 in the 1 Jy and S4 surveys combined. Among these 28 DXRBS objects, six are at $L_R < 10^{32.5}{\rm ~erg ~s^{-1} ~Hz^{-1}}$. These numbers are sure to increase as the remaining 40% of DXRBS objects (primarily optically faint) are identified.

For the BL Lacs, the increase is just as drastic, though restricted to objects with L_X/L_R < 10^-5.5 (i.e. LBL and intermediate BL Lacs). The DXRBS sample includes eight BL Lacs with $L_R < 10^{32} {\rm ~erg ~s^{-1} ~Hz^{-1}}$ and L_X/L_R < 10^-5.5. While a few such objects are probably also included in the ROSAT based samples of Nass et al. (1996) and Kock et al. (1996), their prevalence in these samples is difficult to evaluate because of the large fraction of objects in those samples which lack redshifts. However it must be smaller given the higher X-ray flux limits of the Kock et al. and Nass et al. surveys, which are an order of magnitude higher than DXRBS. This is confirmed by the more recent work of Bade et al. (1997), who have just published redshifts for all but a few of the Nass et al. sample; they find very few objects in their sample at $L_X < 10^{26}{\rm ~erg ~s^{-1} ~Hz^{-1}}$.

3. DXRBS has also filled large holes in our coverage of (L_X,L_R) parameter space, both for BL Lacs and FSRQs. The impact here is much more drastic for the FSRQs. Prior to DXRBS, only nine FSRQs within complete samples were known at values of L_X/L_R > 10^-6. Indeed, the continuity of LBL and FSRQ broad-band and X-ray spectral properties led Sambruna et al. (1996) to predict that no class of HBL-like FSRQs exists. Our results clearly refute this prediction. Thirty-two of the 135 (25%) DXRBS FSRQs so far identified fall in this category; the fraction is even larger (40%; 25 of 59) among the newly identified objects. These objects (whose numbers will surely increase as the remainder of the DXRBS sample is identified), which we term HFSRQs, exhibit clearly smaller (by nearly an order of magnitude) radio luminosities than lower L_X/L_R FSRQs. In the light of this finding, a re-examination of the broadband properties of FSRQs and indeed of the blazar class is in order. We intend to make this subject a priority in our future work.

For BL Lacs, DXRBS contains a large number of ``intermediate'' BL Lacs, objects with 10^-6.5<L_X/L_R<10^-5.5. Until very recently, this region of parameter space was almost completely unexplored. The Einstein Slew survey found the first such objects (Perlman et al. 1996a), and more recently two ROSAT based surveys (Kock et al. 1996, Nass et al. 1996) have found considerable numbers of such objects. However, due to the considerably fainter flux limit of DXRBS, our sample includes fainter objects in this region of parameter space than any previous sample.

EP acknowledges support from a USRA Visiting Scientist Fellowship while at Goddard Space Flight Center, and helpful discussions with G. Madejski and C. M. Urry. PP acknowledges financial support from MURST and ASI. PP and PG acknowledge S. Benetti and M. Turatto for their assistance at La Silla, and M. Della Valle and A. Fontana for their help in preparing for the ESO observing run. RS and LJ acknowledge the support of NRC Regular and Senior Research Fellowships while at Goddard Space Flight Center. This work would have not been possible without the availability of the many radio, optical, and X-ray databases quoted in this paper, namely the NVSS, PMN, GB6, NORTH20CM, TEXAS, APM, COSMOS, WGACAT. We thank all the persons involved in the making of these databases for their effort. This research has made use of the BROWSE program developed by the ESA/EXOSAT Observatory and by NASA/HEASARC and of the NASA/IPAC Extragalactic Database (NED), which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. The Australia Telescope Compact Array, a facility of the Australia Telescope National Facility, is funded by the Commonwealth of Australia for operation as a National Facility managed by CSIRO.

Figure Captions

Figure 1. Ca II break strengths C and rest-frame emission line equivalent widths of radio galaxies and BL Lacs are shown. Quasars are not graphed here because they fall too far to the right to be included (as does one radio galaxy, WGAJ0500.1-3040, which, despite the extremely large equivalent width of several of its emission lines, is a narrow-line object, as described in § 4). We have overplotted the traditional definition of the BL Lac class (dashed box) as used in Stickel et al. (1991), Stocke et al. (1991) and Perlman et al. (1996a), as well as the expanded definition of the BL Lac class advocated by Marchã et al. (1996) (the region in between the dot-dashed lines). Objects where both C and $W_\lambda$ could be measured are shown as squares. Objects where one or both of these figures are upper limits are denoted by diamonds. All error bars shown are 1 $\sigma$, and all upper limits shown are at the 2 $\sigma$ significance level.

Figure 2. Optical spectra of all 85 objects for which we announce identifications in this paper. All spectra have been dereddened and cleaned of cosmic rays as described in Section 3.

Figure 3. Redshift distribution for the DXRBS, S4, and 1 Jy samples of FSRQs (radio quasars with $\alpha_{\rm r} \le 0.7$).

Figure 4. Redshift distribution for the DXRBS, Slew, and 1 Jy samples of BL Lacs. The hatched areas represent lower limits. Redshift figures for 1 Jy BL Lacs have been taken from Stickel et al. (1994), while those for Slew BL Lacs have been taken from Perlman et al. (1996a), Bade et al. (1997), and Perlman, Schachter & Stocke (in preparation).

Figure 5. The X-ray and radio luminosities of FSRQs. Newly identified DXRBS FSRQs are shown as filled circles, while previously identified serendipitous DXRBS FSRQs are shown as filled squares. DXRBS objects identified as radio galaxies with broad emission lines are shown as crosses. The published complete samples of blazars (the 1 Jy [triangles] and S4 [squares]) cover the low-luminosity end very poorly: while still incomplete, the DXRBS blazar survey already includes higher numbers of faint FSRQs (over 3$\times$ the number in the 1 Jy and S4 combined). One in 4 FSRQs have high ratios of X-ray to radio luminosity L_X/L_R > 10^-6 (to the right of the dashed line). Previous radio surveys included very few objects in this region. See sections 5 and 6 for discussion. Radio data for the S4 and 1 Jy sources from Stickel & Kühr (1994) and Stickel et al. (1994); X-ray data from the multifrequency AGN database of Padovani et al. (1997b) and references therein. Note that X-ray data are available only for $\sim 53\%$ and $66\%$ of the S4 and 1 Jy FSRQs respectively.

Figure 6. The X-ray and radio luminosities of BL Lacs. Newly identified DXRBS BL Lacs are shown as filled circles, while previously identified serendipitous DXRBS BL Lacs are shown as filled squares. DXRBS objects identified as radio galaxies with narrow or no emission lines are shown as crosses. The 1 Jy, Slew, and EMSS BL Lacs are represented by triangles, circles, and squares respectively. Crosses represent the NLRGs in our sample. While DXRBS does not include extremely high L_x/L_r BL Lacs such as those found in the Einstein Slew Survey, it can be seen that prior to DXRBS, region of the graph between $-6.5 \lower.5ex\hbox{$\; \buildrel < \over \sim \;$}\log L_x / L_r \lower.5ex\hbox{$\; \buildrel < \over \sim \;$}-5.5$ (denoted by two dashed lines) was very poorly populated, a consequence of the disparate survey methods used. The high sensitivity and combined selection method of DXRBS reveals the previous ``zone of avoidance'' in this graph to be illusory. See Sections 4 and 5 for discussion. Most of the data come from the original papers; additional radio and X-ray data are from the multifrequency AGN database of Padovani et al. (1997) and references therein.

Figure 7. The X-ray-optical ($\alpha_{ox}$) and radio-optical ($\alpha_{ro}$) effective spectral indices of the BL Lacs in the DXRBS sample compared to those in the samples of Kock et al. (1996) and Nass et al. (1996). Newly identified DXRBS BL Lacs are shown as filled circles, while previously identified serendipitous DXRBS BL Lacs are shown as filled squares. Empty circles represent the Nass et al.'s objects, while empty squares indicate the Koch et al.'s sources. Crosses represent the NLRGs in our sample. The two dashed lines denote the loci of points with $\log L_x / L_r = -6.5$ ($\alpha_{\rm rx} \simeq 0.85$, upper line) and $\log L_x / L_r = -5.5$ ($\alpha_{\rm rx} \simeq 0.72$, lower line). Each of the three surveys covers different areas of parameter space, as shown. The spectral indices $\alpha_{ox}$ and $\alpha_{ro}$ are defined in the usual way and calculated between the rest-frame frequencies of 5 GHz, 5000 Å, and 1 keV.