DETECTION OF REST-FRAME OPTICAL LINES FROM X-SHOOTER SPECTROSCOPY OF WEAK EMISSION-LINE QUASARS^*

We focus on WLQs at moderate redshift (z ∼ 1.5) here, a redshift where X-shooter covers a large range of emission lines (spanning C iv through Hα). We select targets from the lists of radio-quiet, weak-featured quasars in Table 6 of P10 (86 objects) and Table 5 of Collinge et al. 2005 (27 objects). These catalogs include quasars that pass optical spectroscopic criteria to be classified as BL Lac objects, namely, that all line emission have W_r < 5 Å, and the 4000 Å break is smaller than 40% if present (see P10, for details). To select X-shooter targets, we restrict the above lists to sources with declinations δ ≲ 20° (to be visible to the VLT), we require targets to have a reliable redshift (from weak emission features in their SDSS spectra, typically Mg ii, and sometimes C iii] and C iv) in order to avoid any contamination from stars, and to ensure that the Hβ line will fall in an NIR atmospheric transmission window, and we require sufficient signal-to-noise ratio (S/N) in <1 hr observing blocks. These cuts ween our target list to ∼15 sources.

For observations, we select six targets (1.4 < z < 1.7) with SDSS spectra that are representative of the larger ∼15 object sample (in terms of relative line strengths and continuum shapes). We also include the z = 1.7 WLQ SDSS J094533.98+100950.1 (hereafter SDSS J0945), which was discovered serendipitously by Hryniewicz et al. (2010). We include SDSS J0945 because it has slightly stronger Mg ii emission than the other targets, and it therefore probes a different part of WLQ parameter space (also see, e.g., Czerny et al. 2011; Laor & Davis 2011 for constraints this particular source has placed on quasar accretion disks). Observations of these seven WLQs were taken over two observing seasons (program IDs 088.B-0355 and 090.B-0438; PI Plotkin), and a summary of the observations is provided in Table 1. Throughout the text, we refer to each source by its SDSS designation truncated to the first four digits.

Table 1.  X-shooter Observation Log

Source Name				ipsfd	Mie		Exp. Time (min)
(SDSS J)	za	${{z}_{{\rm sys}}}$ b	αλc	(mag)	(mag)	Obs. Date	UVB	VIS	NIR
083650.86+142539.0	1.750	1.749	−1.1	18.72	−26.69	2012 Feb 25	38	40	40
094533.98+100950.1	1.671	1.683	−1.1	17.44	−27.80	2011 Dec 21	20	20	20
132138.86+010846.3	1.423	1.422	−1.7	18.80	−26.08	2013 Mar 4	38	40	40
133222.62+034739.9f	1.447	1.442	−0.7	17.94	−26.97	2013 Mar 3	20	22	20
141141.96+140233.9	1.753	1.754	−1.8	18.53	−26.84	2013 Feb 12	38	40	40
141730.92+073320.7	1.710	1.716	−1.7	18.40	−26.92	2013 Mar 4	38	40	40
144741.76–020339.1	1.431	1.430	−2.3	18.61	−26.37	2012 Mar 23	38	40	40

Notes.

^aRedshift obtained from Hewett & Wild (2010). ^bSystemic redshift (see Section 3.1). ^cBest-fit power-law index (${{f}_{\lambda }}\sim {{\lambda }^{{{\alpha }_{\lambda }}}}$) for the continuum of each X-shooter spectrum, fit between rest-frame 1680–1710, 1975–2050, 2150–2250, and 4010–4050 Å (see Section 3.2). ^dSDSS PSF i-band magnitude from Schneider et al. (2010). ^eAbsolute i-band magnitude, K-corrected to z = 2, from Shen et al. (2011) (using the same cosmology as adopted here). These absolute magnitudes include both continuum and emission line K-corrections, for which the latter may be incorrect for WLQs. We adopt these M_i values nonetheless, for consistency with comparison quasar samples for which we also utilize measurements from Shen et al. (2011). ^fA known lensed quasar; see Morokuma et al. (2007) and Section 2.2.1.

Download table as:  ASCIITypeset image

Similar instrument configurations and observing strategies were employed for all seven observations. X-shooter splits the incoming light beam into three segments or "arms" (termed UVB, VIS, and NIR) using two dichroics at 560 and 1024 nm, and the light is fed into three independently operated spectrographs. We used a 1.0 × 110 slit for the UVB arm and 0.9 × 110 slits for the VIS and NIR arms, providing spectral resolutions of R ∼ 4350, 7450, and 5300 in the UVB, VIS, and NIR arms, respectively. The detectors were binned by 2 pixels along the dispersion direction for the UVB and VIS arms, and no binning was used for the NIR arm. We acquired four exposures for each target, nodding the target along the slit in an ABBA pattern ($4\buildrel{\prime\prime}\over{.} 5$ separation between positions). We observed SDSS J0945 and SDSS J1332 for ∼20 minutes each, and each of the other five targets were observed for ∼40 minutes (see Table 1). To improve S/N in the NIR arm, the four observations from 2013 were taken with the K-band blocking filter in place (the blocking filter was not available at the time of the other three observations). For telluric corrections, an A0V star was observed before or after each WLQ at a similar airmass, with the same slit and instrument configuration. A spectrophotometric standard star was observed for flux calibration at the beginning or end of each night using a 5'' slit (the flux standard was taken on the following night for SDSS J1321 and SDSS J1417). Each X-shooter arm is reduced individually using the X-shoter pipeline (Modigliani et al. 2010), following standard procedures that are described in detail in the Appendix. The final flux calibrated spectra (including telluric corrections, and corrections for Galactic extinction) are presented in Figure 1.

Zoom In Zoom Out Reset image size

Absolute flux calibration of echelle spectroscopy is difficult, especially for an instrument like X-shooter with a very broad spectral range (see, e.g., Pita et al. 2014, for a recent discussion specific to X-shooter). We crudely estimate the accuracy of our flux calibration by comparing the luminosity at 5100 Å, ${{L}_{5100}}={{\left( \nu {{L}_{\nu }} \right)}_{5100}}$, in our flux calibrated X-shooter spectra to L₅₁₀₀ inferred from SDSS photometry (extrapolating from the psf i-band magnitude, and assuming a continuum following f_ν ∝ ν^−0.5; e.g., Vanden Berk et al. 2001). The X-shooter luminosity at 5100 Å is typically 20%–50% lower than in the SDSS, except for SDSS J0836 where X-shooter is brighter by ∼10%. We suspect that variability could contribute up to ∼15% of the difference in flux, based upon a comparison of the i-band psf magnitude from the SDSS photometric epoch to the synthetic i-band magnitude from the SDSS spectroscopic epoch for each source. Of course, we do not expect variability to systematically bias the flux calibration in the same direction for all sources. Regardless, we suspect that our X-shooter flux calibration is systematically biased toward lower values for most sources (although the magnitude of the bias is not well-determined). This bias could in turn influence measurements of line luminosities, black hole masses, and Eddington ratio estimates. We stress, however, that W_r and FWHM measurements are not affected by uncertainties in the flux calibration.

2.2.1. SDSS J133222.62+034739.9

Prior to these X-shooter spectra, our targets' redshifts were estimated from rest-frame UV emission lines in their SDSS spectra. The SDSS-derived redshifts are listed in Table 1, as taken from Hewett & Wild (2010). UV emission lines can be susceptible to blueshifts (relative to their laboratory wavelengths) due to winds or other non-viralized motion. From the X-shooter NIR spectra, we derive systemic redshifts (z_sys) from lower-ionization emission lines, which have profiles that are expected to be dominated be virialized motion. Systemic redshifts are determined, in order of preference, from the peak of narrow [O iii] (which is firmly detected only for SDSS J1321) or from the average of the peaks of broad Hα and Hβ. We estimate that we can determine the wavelength of the peak line flux density to within one resolution element (∼3 Å in the NIR), so that typically ${{\sigma }_{{{z}_{{\rm sys}}}}}\lesssim 0.0006$. The new systemic redshifts are similar to the SDSS redshifts from Hewett & Wild (2010) except for two sources, where the new systemic redshifts are slightly larger by Δz = 0.012 (SDSS J0945) and 0.006 (SDSS J1417). The X-shooter derived systemic redshifts are listed in Table 1 and are adopted as the redshift for each source throughout the rest of the text.

We fit a power law (${{f}_{\lambda }}\sim {{\lambda }^{{{\alpha }_{\lambda }}}}$) to each X-shooter spectrum in order to constrain the shape of the UV continuum. The power law is fit between the following (rest-frame) wavelength regions, 1680–1710, 1975–2050, 2150–2250, and 4010–4050 Å, by using a χ² minimization routine (the fit is performed to all measured flux densities within each spectral window, weighted by the uncertainty on each measurement). The above spectral windows are relatively free from contamination from emission lines (including blended iron emission), they are common among all seven X-shooter targets (given their redshifts), and they provide a wide dynamic range. The best-fit spectral indices are given in Table 1 and, excluding SDSS J1332, they are typical of other SDSS quasars (which have α_λ ≈ −1.56; Vanden Berk et al. 2001).

For each of the six WLQs, we perform spectral fits to the Hα, Hβ, Mg ii, and C iv complexes, fitting each complex separately by following a procedure similar to Trakhtenbrot & Netzer (2012). These are among the strongest emission lines in quasars, and the ones that can be measured most accurately given the S/N of our data. Our spectral model includes a local linear continuum, (up to three) Gaussians to model each emission line, and templates for broadened Fe ii and Fe iii emission (based on Boroson & Green 1992 in the optical and Vestergaard & Wilkes 2001 in the UV; see Appendix C of Trakhtenbrot & Netzer 2012 for details). Details on the spectral fitting (including how we associate uncertainties to best-fit parameters) are provided in the Appendix. For the Hβ complex of SDSS J1321 and SDSS J1447, we include narrow Gaussians to model the [O iii] λ4959, 5007 emission lines. These are the only two targets with sufficient spectral coverage to include [O iii] profiles in the spectral modeling (due to their relatively lower redshifts). We also attempted to include other narrow emission line species that fall within each complex, but doing so did not improve the quality of any other fits (see Appendix). The best-fit models to each complex are shown in Figures 2 and 3, and the corresponding spectral properties (i.e., W_r, FWHM, and line luminosities) are presented in Tables 2–4.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 2.  Rest-frame Equivalent Width Measures

Name	Hα	Hβ	Mg ii	C iv	Fe ii a	Fe ii[UV]b	He iiλ4686c	He iiλ1640d
	(Å)	(Å)	(Å)	(Å)	(Å)	(Å)	(Å)	(Å)
J0836	$162.2_{-13.7}^{+40.6}$	$27.9_{-4.5}^{+51.3}$	$9.6_{-0.7}^{+1.4}$	$4.2_{-0.5}^{+0.3}$	69.2 ± 16.9	14.2 ± 5.9	<8.0	<0.4
J0945	$210.6_{-18.5}^{+39.4}$	$31.8_{-3.6}^{+5.1}$	$18.3_{-0.5}^{+0.3}$	$2.9_{-0.6}^{+0.3}$	63.6 ± 9.5	35.7 ± 2.6	<1.3	<0.2
J1321	$236.3_{-18.9}^{+3.5}$	$37.5_{-4.2}^{+17.8}$	$16.1_{-0.4}^{+3.8}$	$18.7_{-1.6}^{+0.8}$	$32.6_{-7.8}^{+32.1}$	$27.6_{-0.9}^{+15.9}$	...	0.9 ± 0.1
J1411	$183.4_{-24.2}^{+1.2}$	32.3 ± 10.2	$6.3_{-0.4}^{+0.6}$	$3.8_{-0.2}^{+0.8}$	45.5 ± 17.0	7.1 ± 2.8	<7.5	<0.9
J1417	$151.8_{-1.3}^{+18.4}$	15.6 ± 2.4	$12.9_{-0.2}^{+0.5}$	$2.5_{-0.7}^{+2.1}$	25.7 ± 4.4	27.7 ± 1.9	<7.0	0.2 ± 0.07
J1447	$237.9_{-18.1}^{+6.8}$	$31.2_{-2.6}^{+16.0}$	$13.1_{-0.1}^{+2.5}$	$7.7_{-1.3}^{+0.2}$	$49.8_{-5.2}^{+21.4}$	$13.2_{-2.7}^{+13.3}$	...	0.8 ± 0.06
Other WLQs with Hβ in the Literature
SDSS J114153.34+021924.3e	...	$20_{-3}^{+4}$	...	0.4 ± 0.2	...	...	...	...
SDSS J123743.08+630144.9e	...	$35_{-5}^{+8}$	...	7.7 ± 1.1	...	...	...	...
SDSS J152156.48 + 520238.5f	...	30.7	...	9.1 ± 0.6	...	...	...	...
PG1407g	...	<40	...	$4.6\pm 2$	...	...	...	...
PHL1811h	...	50	...	6.6	...	...	...	...

Notes.

^aEquivalent width for optical Fe is calculated between 4434–4684 Å. ^bEquivalent width for UV Fe is calculated between 2250–2650 Å. ^cHe iiλ4686 is measured between 4665–4700 Å. Upper limits are quoted at the 3σ level (see Section 6.1.1). ^dHe iiλ1640 is measured between 1620–1650 Å. Upper limits are quoted at the 3σ level, and errors on detected He ii λ1640 are quoted at the 1σ level (see Section 6.1.1). ^eW_r(Hβ) is from Shemmer et al. (2010); W_r[C iv] is from DS09. Shen et al. (2011) report W_r[C iv] = 5.3 ± 2.0 Å for SDSS J123743.08+630144.9. ^fW_r(Hβ) is from Wu et al. (2011); W_r[C iv] is from Wu et al. (2011). Shen et al. (2011) report W_r[C iv] = 3.0 ± 0.2 Å. ^gLine measurements from McDowell et al. (1995). They present line measurements across two epochs for C iv, and we report the epoch closest in time (1992) to their Hβ observations (1994). The other C iv epoch has W_r = 3.6 ± 2.5 Å. ^hLine measurements from Leighly et al. (2007a).

Download table as:  ASCIITypeset image

Table 3.  FWHM Measures

Name	Hα	Fe ii[Hα]	Hβ	Fe ii[Hβ]	Mg ii	Fe ii[Mg ii]	C iv
	(km s−1)	(km s−1)	(km s−1)	(km s−1)	(km s−1)	(km s−1)	(km s−1)
J0836	$2298_{-412}^{+287}$	1575	$2880_{-1069}^{+1877}$	2775	$3284_{-210}^{+695}$	3850	$5423_{-853}^{+570}$
J0945	$3368_{-50}^{+112}$	3525	4278 ± 598	1800	$5695_{-199}^{+78}$	2950	$8018_{-1730}^{+1058}$
J1321	$2461_{-296}^{+92}$	...	$2551_{-435}^{+1173}$	2025	$3460_{-51}^{+510}$	1925	$2760_{-264}^{+488}$
J1411	$3137_{-473}^{+282}$	...	3966 ± 1256	1475	$2860_{-170}^{+420}$	1075	$6177_{-410}^{+2210}$
J1417	$1973_{-31}^{+140}$	2250	2784 ± 759	925	$4717_{-51}^{+222}$	1375	$9694_{-2980}^{+10305}$
J1447	$2000_{-209}^{+19}$	...	$1923_{-164}^{+933}$	1500	$2442_{-32}^{+618}$	1975	$4536_{-991}^{+1539}$

Download table as:  ASCIITypeset image

Table 4.  Line Luminosities and Blueshifts

Name	log L[Hα]	log L[Hβ]	log L[Mg ii]	log L[C iv]	Δv[Mg ii]	Δv[C iv]
	(erg s−1)	(erg s−1)	(erg s−1)	(erg s−1)	(km s−1)	(km s−1)
J0836	$44.29_{-0.06}^{+0.21}$	$43.71_{-0.19}^{+1.09}$	$43.56_{-0.06}^{+0.09}$	43.54 ± 0.07	197 ± 216	2266 ± 191
J0945	44.64 ± 0.05	44.02 ± 0.08	44.29 ± 0.02	43.75 ± 0.12	1281 ± 183	5485 ± 380
J1321	$43.89_{-0.04}^{+0.03}$	$43.32_{-0.09}^{+0.39}$	$43.48_{-0.02}^{+0.11}$	$43.95_{-0.06}^{+0.26}$	202 ± 188	396 ± 189
J1411	$44.03_{-0.06}^{+0.03}$	43.52 ± 0.20	43.34 ± 0.06	$43.60_{-0.06}^{+0.16}$	−136 ± 181	$3142_{-208}^{+370}$
J1417	$44.15_{-0.03}^{+0.05}$	43.45 ± 0.11	43.88 ± 0.02	$43.55_{-0.23}^{+2.38}$	624 ± 180	$5321_{-872}^{+4178}$
J1447	$43.99_{-0.03}^{+0.02}$	$43.41_{-0.07}^{+0.51}$	$43.56_{-0.02}^{+0.09}$	$43.89_{-0.15}^{+0.50}$	1 ± 185	$1319_{-381}^{+759}$

Note. Δv[Mg ii] and Δv[C iv] are line of sight blueshifts of the peaks of the Mg ii and C iv profiles, respectively. Δv is based on the observed wavelength of each line center (expected to be at rest-frame 2800 Å for Mg ii and 1549 Å for C iv) compared to the systemic redshifts in Table 1. Blueshifts are defined to be positive for approaching motions.

Download table as:  ASCIITypeset image

Only three other SDSS WLQs have NIR spectroscopy, all of which are at higher redshift than our X-shooter targets. These sources include SDSS J114153.34+021924.3 (hereafter SDSS J1141; z = 3.55) and SDSS J123743.08+630144.9 (hereafter SDSS J1237; z = 3.49) presented by Shemmer et al. (2010), and SDSS J152156.48+520238.5 (hereafter SDSS J1521; z = 2.238) presented by Wu et al. (2011). We note that Wu et al. (2011) identified SDSS J1521 as a PHL 1811 analog, and it is very X-ray weak (by a factor of 34.5) and extremely optically luminous (M_i = −30.19 mag; Just et al. 2007).

We also consider the quasar PG 1407+265 (z = 0.94), which we consider to be the "prototype" WLQ because it displays similar properties to the WLQ population later identified by the SDSS. PG 1407+265 was identified by McDowell et al. (1995) as having an unusual BELR, with extremely weak high-ionization emission lines but otherwise normal quasar properties. We also consider PHL 1811 (z = 0.19), another AGN with unusual BELR properties similar to WLQs (see Leighly et al. 2007a for a broad spectrum covering Lyα through Hα).

There is no uniform data coverage among the five quasars listed above. We therefore consider only their C iv and Hβ equivalent widths, which are included in Table 2 for reference. These five sources generally show relatively normal low-ionization emission lines, weak high-ionization lines, and often blueshifted C iv emission.

We create samples of other quasars, in order to compare our WLQs to the parent population. We start with the SDSS Data Release 7 quasar catalog (Schneider et al. 2010) and the emission line measurements provided by Shen et al. (2011). We restrict the SDSS quasar sample to only include quasars targeted for SDSS spectroscopy using the Richards et al. (2002) algorithm, which is flux limited to i = 19.1 mag, and we also exclude all objects identified as weak-featured quasars in Collinge et al. (2005) and P10. We only include quasars with absolute i-band magnitudes M_i > −28 mag (K-corrected to z = 2), to compare to quasars that are of similar luminosities as our X-shooter targets. We also require quasars to not be flagged as broad absorption line quasars (BALs) by Shen et al. (2011), and to not be radio-loud (as determined by Shen et al. 2011).

4.2.1. The "Intermediate-redshift" Sample

4.2.2. The "Low-redshift" Sample

4.2.3. The "Small" Sample

Broad Hα and Hβ emission lines are clearly detected in the NIR spectra of all six WLQs. One source (SDSS J1417) has W_r[Hβ] ≈ 16 Å, which is 4σ weaker than the mean of the SDSS-LZ sample; all other sources display W_r[Hβ] ≈ 28–38 Å, which is a rather typical value (1.8σ–2.6σ lower than the mean). Other WLQs with Hβ coverage in the literature also display similar W_r[Hβ], ranging from 20–50 Å (see Table 2). All six X-shooter targets also show blended Fe ii emission near Hβ. Following Boroson & Green (1992), we define the ratio R_{opt,Fe ii} = W_r[Fe ii]_opt/W_r[Hβ], where W_r[Fe ii]_opt is measured from 4434–4684 Å. For all six targets, R_{opt,Fe ii} lies toward the higher end of the expected distribution for the parent quasar distribution (i.e., Fe ii is enhanced relative to Hβ). That said, no WLQ shows an exceptionally large R_{opt,Fe ii} (see Figure 5). WLQs generally possess rest-frame optical properties that are consistent with other quasars that display enhanced Fe ii emission. For example, all six X-shooter targets have FWHM[Hβ] ≲ 4000 km s⁻¹, which is consistent with other quasars that have similar R_{opt,Fe ii} values (Boroson & Green 1992; Sulentic et al. 2000, 2007; Shen & Ho 2014, also see Figure 5).

Zoom In Zoom Out Reset image size

To quantify the above statements, we compare the R_{opt,Fe ii} distribution of WLQs and SDSS-LZ quasars by running Kolmogorov–Smirnov (K–S) and Mann–Whitney (M–W) tests. The K–S and M–W tests are both non-parametric tests that assess if two (independent) distributions are drawn from the same parent population. The K–S test achieves this by comparing the maximum deviation between the cumulative distribution functions of the two observed populations, while the M–W test uses rank-ordering to compare the median values of the two distributions (e.g., Sheskin 2011). Given the small number of objects in our WLQ sample, we opt to run both tests here as alternative ways to illustrate statistical conclusions. The R_{opt,Fe ii} distribution of WLQs is different than that of the SDSS-LZ quasars at the >99% level (p = 0.004 from a K–S test; p = 0.001 from a M–W test), indicating that WLQs as a population are indeed statistically more likely to display enhanced optical Fe ii emission. We again stress that any potentially unidentified WLQs in the SDSS-LZ sample are too small in number to bias the above statistics (and any bias would only strengthen our conclusions, since the bias would force the two distributions to appear more similar).

We expect quasars with large R_{opt,Fe ii} to also show weak [O iii] emission (e.g., Boroson & Green 1992). Unfortunately, [O iii] constraints are extremely limited for our X-shooter targets, as we only have [O iii] coverage for the two targets at z ≈ 1.4 (i.e., SDSS J1321 and SDSS J1447; the other four targets are at too high-redshift for 5007 Å to fall within an atmospheric transmission window). Indeed, only SDSS J1321 with the weakest Fe ii emission (R_{opt,Fe ii} = 0.9) has a firm [O iii] detection, but constraints on more objects are required to determine if WLQ [O iii] emission follows the trend expected from typical quasars.

The bluer sensitivity of the UVB arm compared to the SDSS provides our first constraints on C iv for two sources (SDSS J1321 and SDSS J1447, both are at z ∼ 1.4 and were included in the P10 sample because of their small W_r[Mg ii]). These two sources display the largest W_r[C iv] of all six targets. SDSS J1321 has W_r[C iv] ≈ 19 Å, which is only 1.7σ weaker than the mean W_r[C iv] of the SDSS-IZ comparison sample; SDSS J1447 has W_r[C iv] ≈ 8 Å, which is 3.7σ weaker than the mean. Their stronger C iv therefore implies that weak Mg ii does not guarantee very weak C iv emission (also see Wu et al. 2012, for a similar conclusion). It is noteworthy that both sources have W_r[C iv] > 5 Å, meaning that neither would have passed the P10 spectroscopic criteria for BL Lac objects if SDSS covered C iv.

The UV emission lines covered by each SDSS spectrum remain weak in each X-shooter observation. For example, the four X-shooter targets with C iv coverage in SDSS still have W_r[C iv] < 5 Å. We also apply our fitting routine to the SDSS spectra to measure W_r[Mg ii], and Mg ii is not highly variable between the two epochs.¹⁷ The fractional change between W_r[Mg ii] in the SDSS and X-shooter epochs are (W_r[Mg ii]_SDSS—W_r[Mg ii]_Xshooter)/—W_r[Mg ii]_Xshooter = 0.41 ± 0.18, $0.05_{-0.03}^{+0.06}$, 0.34 ± 0.29, $-0.14_{-0.24}^{+0.15}$, 0.19 ± 0.08, 0.11 ± 0.17 for SDSS J0836, SDSS J0945, SDSS J1321, SDSS J1411, SDSS J1417, and SDSS J1447, respectively.

The new systemic redshifts from the X-shooter data also allow improved insight into the rest-frame UV properties. With updated redshifts, we can more rigorously investigate potential velocity offsets between the peaks of high- and low-ionization emission lines. Five of our targets (SDSS J0836, SDSS J0945, SDSS J1411, SDSS J1417, SDSS J1447) display strong blueshifts in the peak wavelength of their C iv lines (Δv > 1000 km s⁻¹; calculated relative to 1549 Å), suggestive of a non-virialized component to the C iv emitting BELR gas. With the new systemic redshifts from X-shooter, we find that two of these sources (SDSS J0945 and SDSS J1417) also show significant Mg ii blueshifts (calculated relative to 2800 Å) as well (1281 ± 183 and 624 ± 180 km s⁻¹, respectively; see Table 4). These blueshifts are discussed in Section 6.1.

We use the fits from the Hβ complexes to estimate virial black hole masses (M_BH) and Eddington ratios (L_bol/L_Edd). Since the optical properties from our X-shooter spectra are not atypical compared to normal quasars, we likely can obtain representative M_BH estimates from the Hβ line (although see caveat below). We use the luminosity of the best-fit (linear) continuum at 5100 Å, L₅₁₀₀, and the best-fit FWHM[Hβ], along with the empirical BELR size–luminosity relation of Kaspi et al. (2005; as modified by Bentz et al. 2009):

$$\begin{eqnarray}&&\frac{{{M}_{{\rm BH}}}}{{{10}^{6}}\;{{M}_{\odot }}}=5.05{{\left[ \frac{{{L}_{5100}}}{{{10}^{44}}\;{\rm erg}\;{{{\rm s}}^{-1}}} \right]}^{0.5}}\;{{\left[ \frac{{\rm FWHM}({\rm H}\beta )}{{{10}^{3}}\;{\rm km}\;{{{\rm s}}^{-1}}} \right]}^{2}},\end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray}&&{{L}_{{\rm bol}}}/{{L}_{{\rm Edd}}}=0.13\;f(L){{\left[ \frac{{{L}_{5100}}}{{{10}^{44}}\;{\rm erg}\;{{{\rm s}}^{-1}}} \right]}^{0.5}}\;{{\left[ \frac{{\rm FWHM}({\rm H}\beta )}{{{10}^{3}}\;{\rm km}\;{{{\rm s}}^{-1}}} \right]}^{-2}},\end{eqnarray} \tag{ 2 }$$

where f(L) is a luminosity-dependent bolometric correction to L₅₁₀₀. We calculate f(L) from Equation (21) of Marconi et al. (2004). We adopt the above relations for consistency with Shemmer et al. (2010) and Wu et al. (2011), who presented Hβ measurements for two WLQs at z ∼ 3.5 and one PHL 1811 analog at z ∼ 2.2, respectively. The black hole mass and Eddington ratio estimates are listed in Table 5. We find black holes masses between log(M_BH/${{M}_{\odot }}$) ≈ 8–9, and Eddington ratios L_bol/L_Edd ≈ 0.3–1.3. The L_bol/L_Edd estimates are toward the high-end of the range typically observed for z ∼ 1.4–1.7 quasars, although we note that the uncertainties on each measurement are rather large (see Table 5).

Table 5.  Virial Black Hole Masses and Eddington Ratios

Name	f(L)a	logL5100b	log MBHc	Lbol/LEddc
		(erg s−1)	(${{M}_{\odot }}$)
J0836	5.96	$45.93_{-0.23}^{+0.10}$	$8.59_{-0.26}^{+0.64}$	$0.87_{-0.65}^{+1.36}$
J0945	5.85	46.17 ± 0.03	$9.05_{-0.08}^{+0.14}$	0.51 ± 0.15
J1321	6.27	$45.41_{-0.06}^{+0.03}$	$8.22_{-0.13}^{+0.47}$	$0.63_{-0.37}^{+0.23}$
J1411	6.11	45.64 ± 0.10	8.72 ± 0.24	$0.34_{-0.17}^{+0.42}$
J1417	5.97	45.91 ± 0.03	$8.55_{-0.17}^{+0.30}$	0.92 ± 0.50
J1447	6.16	$45.56_{-0.06}^{+0.02}$	$8.05_{-0.06}^{+0.51}$	$1.33_{-0.78}^{+0.17}$

Notes.

^aBolometric correction from Marconi et al. (2004). ^bLuminosity of the power-law continuum at 5100 Å. The error bars are determined in the same manner as in Section B.1. ^cQuoted errors on log M_BH and L_bol/L_Edd are based on propagating the errors determined by our spectral fitting (see B.1), and they do not include systematics or other uncertainties related specifically to the virial-based scaling technique (i.e., Equations (1) and (2)).

Download table as:  ASCIITypeset image

Our six WLQs are at higher redshift than reverberation mapped quasars to which virial black hole masses are calibrated; our WLQs also generally have larger R_{opt,Fe ii} and narrower Hβ than most reverberation mapped quasars. These caveats may introduce systematic biases into our M_BH and L_bol/L_Edd estimates (Richards et al. 2011). We thus consider our M_BH and L_bol/L_Edd estimates to be approximate, but still useful for a qualitative analysis. We do not attempt M_BH or L_bol/L_Edd estimates using either C iv or Mg ii because we have indications for non-virialized motion affecting those lines for some of our sources, which could bias the mass measurements (see, e.g., Baskin & Laor 2005; Denney 2012; Trakhtenbrot & Netzer 2012; Kratzer & Richards 2015).

If the WLQ phenomenon is not primarily driven by low gas content in the BELR, then the BELR is probably exposed to an unusually soft ionizing continuum. The following discusses soft ionizing continua in the context of "disk/wind" models of the BELR (e.g., Murray et al. 1995; Elvis 2000; Leighly 2004). In these models, lower-ionization potential lines (e.g., Mg ii, Hβ, Hα) are typically considered "disk" lines, because they are associated with gas with kinematics dominated by virialized motions (e.g., Eracleous & Halpern 2003). High-ionization lines (like C iv) are thought of as "wind" lines, because they can show a component with non-virialized motion, often interpreted in the context of a radiatively line driven wind. Some lines can have both a disk and wind component (see, e.g., Leighly 2004 and references therein). In order to drive a wind radiatively, the balance between the number of X-ray and UV photons is critical. UV photons are required to accelerate the wind; however, a large X-ray flux will over-ionize the BELR gas, stripping too many valence electrons from the metals for the line-driving to be effective. The above scenario is consistent with some well-known correlations between quasars, specifically relations between X-ray properties and the relative balance between disk and wind line properties (i.e., the so-called Eigenvector 1 correlates; see, e.g., Boroson & Green 1992; Laor et al. 1997; Wills et al. 1999; Sulentic et al. 2000, 2007; Baskin & Laor 2005; Shang et al. 2007; Gibson et al. 2009; Kruczek et al. 2011; Richards et al. 2011; Shen & Ho 2014).

Wu et al. (2011, 2012) discuss in detail the relationship between rest-frame UV emission lines (in particular C iv blueshift) and X-ray properties for WLQs/PHL 1811 analogs. The ratio of X-ray luminosity to the luminosity of the rest-frame UV continuum provides constraints on the ionizing continuum (as does the relative strengths of some UV emission species), while C iv blueshifts can be interpreted as a proxy for the strength of the disk wind (along the line of sight). Wu et al. (2011, 2012) find that PHL 1811 analogs (which were selected to have enhanced UV Fe emission, large C iv blueshifts, and weak W_r[C iv]) are very X-ray weak. On the other hand, WLQs that show weaker UV Fe emission (and typically also smaller C iv blueshifts) are generally X-ray normal.

To examine the properties of our X-shooter targets in the context of the above phenomenological picture, we show our targets in the C iv W_r–blueshift plane in Figure 8 (also see Figure 8 in Wu et al. 2012). The comparison SDSS-IZ quasars in Figure 8 have C iv blueshifts calculated relative to Mg ii, so we present C iv blueshifts for our X-shooter targets relative to both Mg ii (blue circles) and to the X-shooter derived systemic redshift (other filled symbols). SDSS J1321 is the only target that clearly occupies a similar parameter space as typical quasars. The other five sources lie in the "wind-dominated" quadrant of Figure 8 (see, e.g., Richards et al. 2011), albeit with generally much weaker W_r[C iv] than other quasars displaying similar blueshifts.¹⁹

Zoom In Zoom Out Reset image size

It is interesting that the two X-shooter targets with the strongest C iv blueshifts are the only two that also display blueshifted Mg ii emission. Mg ii is often thought of as a disk emission line, given its relatively low ionization potential (χ_ion ≈ 8 eV). However, Mg ii can still have a non-negligible cross section to UV radiation, and it sometimes has a wind component (e.g., Shang et al. 2007). From our X-shooter sample, it appears that a relatively strong wind is required to cause Mg ii to display an outflowing component. We also speculate that there could be a connection between the wind strength and the amount of Fe ii UV emission. We quantify the strength of the Fe ii UV emission as R_{uv,Fe ii} = W_r[Fe ii]_uv/W_r[Hβ], where W_r[Fe ii]_uv is calculated from 2250–2650 Å, which we show relative to FWHM[Hβ] in Figure 9. Similar to the rest-frame optical Fe ii emission (i.e., R_{opt,Fe ii}; Figure 5), none of the X-shooter targets displays abnormally large R_{uv,Fe ii}. However, the two WLQs showing Mg ii blueshifts are the only ones to display relatively enhanced R_{uv,Fe ii} (compared to all five sources with C iv blueshifts showing enhanced R_{opt,Fe ii}). Mg ii blueshift is therefore a newly identified parameter that could ultimately provide additional clues on the BELR properties of WLQs (also see Luo et al. 2015 for a discussion on UV Fe ii and X-ray weakness).

Zoom In Zoom Out Reset image size

6.1.1. Weak He ii Emission

6.1.2. Is the Ionizing Continuum Intrinsically Soft?

6.1.3. SDSS J1447

There is a final possibility that could explain the relatively normal Hβ and the weak/blueshifted C iv in our X-shooter spectra. Hryniewicz et al. (2010) propose a model where WLQs represent a short-lived phase (∼10³ yr) of an evolutionary sequence where black hole activity has just recently been (re)activated. They propose that WLQs are essentially "adolescent" quasars, where the BELR has not yet finished forming. Low-ionization species (like Hβ and Mg ii) that form close to the accretion disk should appear normal. However, the disk wind has not yet had sufficient time to populate regions farther above the disk, and higher-ionization "wind" species like C iv could appear exceptionally weak. We cannot effectively discriminate between adolescent quasars and soft ionizing continua from this X-shooter program alone, because each scenario predicts similar spectral properties. Testing this scenario would require stringent constraints on WLQ evolution (i.e., more WLQs would be expected at higher redshifts) and/or the WLQ luminosity function. There are intriguing indications that WLQs are indeed more common at higher redshifts, from studies on the fraction of WLQs at z ≈ 3–6 (DS09; Bañados et al. 2014). The systematics involved in WLQ selection at lower redshifts are not understood well enough to extend this type of work to z < 3 at the moment. However, such systematics are starting to be addressed, between this X-shooter program and other NIR spectroscopic campaigns on z < 3 SDSS WLQs (e.g., Shemmer et al. 2010; Wu et al. 2011), and also the recent efforts by Meusinger & Balafkan (2014) at WLQ selection (we note, however, that the P10 and Meusinger & Balafkan 2014 WLQ samples likely suffer from different sets of selection-based systematics, and as a result may not recover identical populations of rare quasars).

Meusinger & Balafkan (2014) recently applied machine learning techniques to the SDSS spectroscopic database to assemble a new WLQ sample, and they find that a substantial fraction of their WLQs have radio detections in FIRST (approximately one-quarter of their WLQ sample are radio-loud). In light of this work, and the fact that PG 1407+265 may also have a beamed (albeit weak) relativistic jet (Blundell et al. 2003; Gallo 2006), we revisit the prospect of relativistic beaming here. A beamed jet is still highly unlikely to cause the weak BELRs for our X-shooter targets (as well as for PG 1407+265), for the following reasons. First, all X-shooter targets were required to be radio-fainter, at the >3σ level, than the population of radio-loud SDSS BL Lac objects in P10. Therefore, any radio jet must be low luminosity and/or weakly beamed. Any beamed continuum would also be highly variable, so that we would expect variable W_r measures between the SDSS and X-shooter epochs (e.g., Ruan et al. 2014), which we do not observe. Furthermore, dilution from a relativistic jet would reduce the W_r of all lines, inconsistent with our observations of enhanced optical Fe ii (i.e., large R_{opt,Fe ii}). Finally, the C iv blueshifts cannot be explained by dilution from a beamed jet, and the magnitude of the blueshifts (i.e., ∼10³ km s⁻¹) for our X-shooter targets are uncommon for radio-loud quasars (Richards et al. 2011).

The uncertainties on best-fit line parameters are influenced by systematics, with the error budget dominated by uncertainties on the level of the continuum (i.e., there are degeneracies between the best-fit linear and Fe continuum levels and the parameters describing the line profiles). Since the uncertainties are not necessarily dominated by statistical fluctuations, we do not use the absolute value of the final reduced $\chi _{r}^{2}$ to assign a statistical confidence to the quality of each fit, nor do we trust the error bars on each best-fit parameter returned by the χ² minimization routine (which are likely underestimated). We instead employ an empirical scheme to estimate uncertainties on each best-fit parameter, which implicitly accounts for (the dominant) systematic and statistical uncertainties.

We begin with the best spectral fit to each line complex, and we record the corresponding $\chi _{{\rm best}}^{2}$ and degrees of freedom, ν_best. In the following, we adopt the same number of Gaussians to model the line profile as used for the above best-fit, and we include an Fe template only if one is included in the best-fit. We then create a 51 × 51 grid of linear continua. Each continuum in the grid is set by a fixed flux level within the adopted 20 Å continuum normalization bands, and we allow the flux level to vary through evenly spaced steps across the grid from (typically) 0.8–1.2 times the median observed flux density within each normalization band. The exact range across the grid varies with each source and complex, to ensure that we sample sufficient parameter space surrounding the best-fit continuum level determined earlier. We then systematically step through each continuum in the grid and refit each complex for each source (following the same procedure as described earlier, but keeping the linear continuum fixed during each fit). After each fit is performed, we calculate line properties (e.g., λ_c, FWHM, line flux, W_r, etc.), and we record $\chi _{i}^{2}$ (i.e., $\chi _{i}^{2}$ is for the fit using the ith continuum model in the grid). Since each fit within the grid contains the same number of degrees of freedom, we consider the relative χ_i² (i.e., ${\Delta }\chi _{i}^{2}=\chi _{i}^{2}-\chi _{{\rm best}}^{2}$) across the grid to estimate uncertainties. The 68% (∼1σ) confidence interval typically corresponds to continua providing fits with ${\Delta }\chi _{i}^{2}\sim 30-50$ (given ν ∼ 1900–6500, depending on the source and line complex). For each continuum providing a ${\Delta }\chi _{i}^{2}$ within the 68% confidence range, we determine the minimum and maximum value of each line property (i.e., λ_c, FWHM, W_r, etc.), linear continuum slope and intercept, and Fe continuum normalization (when an iron template is included in the fit). We use the minimum and maximum values of each parameter compared to the best-fit values derived earlier to determine approximate ±1σ uncertainties.

SDSS J0836—Narrow absorption systems for the C iv λλ1548,1551 and Mg ii λλ 2796,2804 doublets are detected at z = 1.733. When fitting the Hα complex, we mask 5800–6000 Å to avoid potential broad He i λ5877 emission blended with the continuum.

SDSS J0945—Narrow absorption systems for C iv λλ1548,1551 and Mg ii λλ 2796,2804 are detected at z = 1.590. These systems were also reported by Hryniewicz et al. (2010) from the SDSS spectrum.

SDSS J1321—[O iii] is detected in the NIR arm. There is a hint of narrow Hα emission when visually examining this spectrum, but including a narrow Hα component does not improve the quality of the fit.

SDSS J1411—We mask 5800–6000 Å when fitting Hα for this source. A telluric line falls near Mg ii. The affected pixels are given less weight during the spectral fitting, but this contamination may cause an additional source of systematic uncertainty on the measurements of the Mg ii line.

SDSS J1417—We mask 5800–6000 Å when fitting Hα for this source. A telluric line also falls near Mg ii (see notes above).

SDSS J1447—[O iii] is tentatively detected in the NIR arm. There is a hint of narrow Hα emission when visually examining this spectrum, but including a narrow Hα component does not improve the quality of the fit. There is also a spurious feature in the UVB arm toward the blue end of the broad C iv emission line, near 3600 Å in the observed frame. This feature is an artifact from the reduction cascade, resulting from the continuum calibration lamp switching from a halogen lamp to a D₂ lamp at λ ≲ 3600 Å. We therefore mask rest-frame 1485–1510 Å prior to fitting the C iv complex. This artifact is not present in the other targets' UVB spectra.