The Ca H&K Break – Equivalent Width Plane

I now want to investigate if it is in general physically meaningful to separate BL Lacs and FSRQ using the Ca H&K break – equivalent width plane.

In the previous Chapter I have shown that the Ca H&K break value of BL Lacs and low-luminosity radio galaxies decreases with viewing angle. This, however, means that only the equivalent widths of narrow emission lines, believed to be radi-ated isotropically, are expected to decrease with Ca H&K break value. The broad emission lines, on the other hand, will be obscured by the putative circumnuclear dusty torus at relatively large viewing angles, and will come progressively into our line of sight as the viewing angle, and therefore the Ca H&K break, decreases. Si-multaneoulsy we also expect the continuum flux to increase (owing to the beamed jet component). This combination of emission line and continuum flux increase ren-ders the resulting equivalent width and its dependence on the Ca H&K break value difficult to quantify and the possibility that the equivalent widths of broad emission lines increase with Ca H&K break value cannot be excluded. Moreover, if the Ca H&K break – equivalent width plane is to be used to separate also radio galaxies and blazars (as was suggested by March˜a et al.) it is obvious that it can be applied to narrow emission lines only. Thus the question that I want to answer becomes:

Is it physically justified to separate BL Lacs and FSRQ in the Ca H&K break – equivalent width plane using their narrowemission lines?

For my studies I have first simulated the decrease in equivalent width with in-creasing contribution from non-thermal jet emission. For this purpose I have as-sumed at C= 0.5 starting equivalent width valuesW_λ= 1, 5, 10, 20, 50, 100, 200, 300, and 500 ˚A, and have increased the non-thermal jet emissionf_jet. I have then cal-culated the Ca H&K break valueCusing the relation logf_jet/f_gal=−3.74×C+0.43, where f_gal is the flux of the underlying host galaxy, assumed to be constant. This relation results from the simulations presented in Section 3.1 for a jet of optical

4.3 The Limitations of the Current Classification Scheme 71 spectral index α_ν = 1, assuming a jet spectrum of the form S_ν ∝ ν^−α (see also Fig. 3.2). The resulting correlations between Ca H&K break value and equivalent width are represented by the dotted lines in Fig. 4.2, left panel. Note that contrary to the results of March˜a et al. the above relation between Ca H&K break and jet emission gives a non-linear dependence between the Ca H&K break and equivalent width values.

These simulations show that the equivalent width range of objects with high Ca H&K break values determines the (maximum) equivalent width range of objects with low Ca H&K break values. (Note that it is the maximum range since an increasing jet emission can further decrease the equivalent width, although the Ca H&K break value has reached its minimum value of zero.) In other words, the equivalent width range of blazars will depend on the (intrinsic) equivalent width range of their parent radio galaxy populations. Then, a discrimination between BL Lacs and FSRQ using, e.g., one of the simulated dashed lines, would be physically justified only if there existed two populations of radio galaxies with significantly different [O II] and/or [O III] equivalent width distributions. This means that we need to observe a bimodal [O^II] and/or [O^III] line luminosity distribution intrinsic to the entire class of radio-loud AGN, which is expected to manifest itself as a bimodal equivalent width distribution at any given orientation. Note that the equivalent width, although orientation dependent in radio-loud AGN, is independent of redshift and so more appropriate than the line luminosity itself to quantify in a meaningful way such a bimodality.

In order to investigate the existence of such a bimodality we need a large number of radio galaxies and blazars that span a wide range of equivalent widths and Ca H&K break values. Therefore, I have additionally included in this study the radio galaxies from the DXRBS (listed in Table 4.3). However, given its radio spectral index cut (α_r ≤ 07) DXRBS selects against most radio galaxies. Therefore, I have included also the sample of radio galaxies and quasars from the 2 Jy survey presented by Tadhunter et al. (1993) and Morganti et al. (1997). The spectra of the 2 Jy sources were kindly made available to us by Clive Tadhunter. For these additional sources I have measured the equivalent widths of [O^II]λ3727 and [O^III]λ5007, and the Ca H&K break values. Upper limits on equivalent widths have been derived as described in Section 4.2. Note that the use of objects from radio surveys with different flux limits is warranted and should not introduce any bias since line flux was not part of the selection criteria.

Fig. 4.2, left panel, plots for radio galaxies and blazars from the DXRBS and the 2 Jy survey the Ca H&K break value versus the rest frame equivalent width

72 A New Classification Scheme for Blazars

Figure 4.2. Left panel: The Ca H&K break value versus the [O^II]λ3727 and [O III] λ5007 rest frame equivalent widths for DXRBS and 2 Jy ob-jects. Arrows indicate upper limits. Dotted lines represent the simulated rest frame equivalent width decrease with increasing non-thermal jet contin-uum for starting values (at C = 0.5) of W^λ = 1, 5, 10, 20, 50, 100, 200, 300, and 500 ˚A (from left to right). Right panel: The [O II] λ3727 and [O^III]λ5007 rest frame equivalent width distributions for objects from the left panel withC= 0, 0< C <0.25, andC ≥0.25. Arrows indicate upper limits. The dashed curves represent the best-fit Gaussian models.

of [O II] λ3727 and [O III] λ5007. The panel for the [O II] ([O III]) emission line includes 59 (48) and 39 (45) sources from the DXRBS and 2 Jy sample respectively.

Note that sources with errors on the Ca H&K break value>0.2 have been excluded from these studies. Fig. 4.2, right panel, shows the equivalent width distributions for these objects, binned into three groups of Ca H&K break valuesC= 0, 0 < C <0.25 and C≥0.25. These groups comprise 52, 25 and 21 objects respectively in the case of [OII], and 40, 25 and 28 objects respectively in the case of [O III].

In order to quantify a possible bimodality in these distributions I have used the

KMMalgorithm developed by Ashman et al. (1994). TheKMMalgorithm computes for a given univariate dataset the confidence level at which the single Gaussian model can be rejected in favour of a two Gaussian model. I have fitted homoscedastic groups (i.e., groups with similar covariances), and only if the resulting confidence level was below 95% I have also considered the heteroscedastic case. For the distributions in Fig. 4.2, right panel, I get a high confidence level for the rejection of the single Gaussian model in the case of [O III] for objects with C ≥0.25 (P = 98.4%) and

4.3 The Limitations of the Current Classification Scheme 73

Table 4.3. DXRBS Radio Galaxies

Name z [OII] [OIII] LNLR C σ

W^λ log L W^λ log L

[˚A] [erg/s] [˚A] [erg/s] [erg/s]

WGAJ0204.8+1514 0.833 49.1 42.35 85.7 42.93 43.77 0.21 0.37 WGAJ0247.9+1845 0.301 <5.8 <40.71 <1.8 <40.63 <41.82 0.47 0.26 WGAJ0500.0−3040 0.417 20.5 42.12 57.2 42.77 43.58 0.15 0.08 WGAJ0605.8−7556 0.458 39.8 41.21 32.9 41.18 42.33 0.03 0.24

WGAJ1120.4+5855 0.158 43.2 42.71 43.44

WGAJ1229.4+2711 0.490 74.5 42.69 53.0 42.57 43.78 0.42 0.53 WGAJ1420.6+0650 0.236 33.2 41.69 11.3 41.63 42.81 0.34 0.10 WGAJ1835.5−6539 0.554 <4.8 <40.46 42.5 41.53 42.36 0.11 0.23 WGAJ2131.9−0556 0.085 <9.2 <39.64 <2.3 <39.53 <40.74 0.43 0.20

WGAJ2205.2−0004 0.827 41.1 42.15 43.59 0.36 0.75

WGAJ2303.5−5126 0.426 11.9 41.70 18.9 42.06 42.99 0.12 0.19

0< C <0.25 (P = 94.7%). In all other cases a single Gaussian model is the better fit. The resulting best-fit Gaussians are overlaid as dashed lines in Fig. 4.2, right panel. Note that, although a relatively large number (9/28 or 32%) of objects with C ≥ 0.25 have only upper limits on their [O III] equivalent widths, the limits only increase the significance of the bimodality but do not cause it. If these are excluded, the bimodal [OIII] equivalent width distribution remains significant (P = 95.6%).

The single Gaussian models for the [O ^II] equivalent width distributions give mean values of about 20, 13, and 3 ˚A for objects withC≥0.25, 0< C <0.25, and C = 0 respectively. This observed decrease in equivalent width is well reproduced by the simulations that give atC= 0 a value of ∼5 ˚A if a starting value of∼20 ˚A at C = 0.5 is assumed. This confirms the use of the Ca H&K break as a statistical indicator of orientation. In the case of [O III] the two best-fit Gaussians give mean values of ∼3 and 100 ˚A for objects with C ≥ 0.25, and ∼4 and 80 ˚A for objects with 0< C <0.25. For a starting value of ∼100 ˚A the simulations predict a value of∼25 ˚A atC = 0. This is similar to the mean of∼20 ˚A observed for objects with C = 0. On the other hand, for a starting value of∼5 ˚A a value of∼1.5 ˚A atC= 0 is expected. This group of objects seems to be absent in the sample under study.

These could be the featureless BL Lacs with no available redshift. In the DXRBS there are 13 such objects (5 of which have been observed by us) that have been excluded from this analysis. Tadhunter et al. (1993) failed to identify recognizable

74 A New Classification Scheme for Blazars features for 11 sources out of their complete sample of 87 objects.

The presence of a bimodality for the [O ^III] emission line could indicate the existence of two physically distinct classes of radio-loud AGN, thus allowing a re-finement of the present classification scheme for blazars. This I want to investigate in the remainder of this chapter.