X-RAY AND MULTIWAVELENGTH INSIGHTS INTO THE NATURE OF WEAK EMISSION-LINE QUASARS AT LOW REDSHIFT

The redshift values (see Table 1) for our low-redshift WLQs are generally those from Hewett & Wild (2010) which are the best available measurements for large SDSS quasar samples. There are three sources lacking Hewett & Wild measurements. For two quasars (J1109+3736 and J1139−0201), the redshift values are taken from the catalog of Plotkin et al. (2010a). The redshift of the other source (J2115+0001; z = 2.4995 ± 0.0052) is measured based on a Lyα + C iv absorption system.¹⁰

To obtain accurate measurements of the weak emission lines, we manually measured rest-frame emission-line properties for C iv, Si iv, the λ1900 complex,¹¹ and Fe iii UV48 (see Table 2) following the method in Section 2.2 of Wu et al. (2011), which is summarized below. We first smoothed the SDSS spectra with a 5 pixel sliding-box filter, and manually interpolated over strong narrow absorption regions. We then fitted a power-law local continuum for each line between their lower and upper wavelength limits λ_lo and λ_hi (see Table 2 of Vanden Berk et al. 2001). After subtracting the local continuum, we measured the REW value for each line. The C iv blueshifts were calculated between the lab wavelength in the quasar rest frame (1549.06 Å; see Table 2 of Vanden Berk et al. 2001) and the observed mode of all pixels with heights greater than 50% of the peak height, where mode = 3×median − 2×mean. For comparison, we also include in Table 2 the corresponding measurements of the spectrum of PHL 1811 (Leighly et al. 2007a) and of the composite spectrum of typical SDSS quasars in Vanden Berk et al. (2001). The spectral measurements of PHL 1811 are included here because some of our low-redshift WLQ candidates show similar unusual UV/optical spectral properties to those of PHL 1811 (see Section 5.2). The Mg ii measurements from Shen et al. (2011) are also listed in Table 2. These measurements are reliable because the Fe ii component, which could affect the Mg ii strength measurement, was well modeled. These REW(Mg ii) values somewhat exceed the selection criterion of REW ≲ 5 Å for BL Lac candidates in Plotkin et al. (2010a). This discrepancy mainly originates from differences in measurement methods. For Plotkin et al. (2010a), it was impractical to define reference wavelengths to model the continuum in a uniform way for the entire large sample since many objects lack redshift measurements. The REW values in Plotkin et al. (2010a) were measured manually after defining the continuum by eye for most sources. While this method generally performed well for BL Lac objects, it did not properly model blended Fe emission for unbeamed objects.

Only two sources (J0945+1009 and J1612+5118) have high-quality C iv coverage in their SDSS spectra so that we are able to measure their C iv REW and blueshift values. Both sources have weak and highly blueshifted C iv lines. J1139−0021 has no clearly detectable C iv line in its SDSS spectrum; we could only obtain an upper limit on its REW. Therefore, we obtained follow-up UV spectroscopy for this source with the Low-Resolution Spectrograph (Hill et al. 1998) on the Hobby–Eberly Telescope (HET; Ramsey et al. 1998). The UV emission-line measurements based on the HET spectroscopy are also listed in Table 2. J1139−0021 has a weak and strongly blueshifted C iv line in its HET spectrum.

All of the sources having C iii] coverage show weaker C iii] semiforbidden lines than those of typical quasars. The Fe iii UV48 strength of our low-redshift WLQ candidates is generally similar to those of typical quasars. The SDSS spectrum of J1109+3736 does not have coverage of these rest-frame UV emission lines because of its much lower redshift, while the signal-to-noise ratio (S/N) of the SDSS spectrum of J2115+0001 is too low to make reliable emission-line measurements.

After measuring the strengths of the UV emission lines of our low-redshift WLQ candidates (see Table 2), we further investigated the REW distributions of prominent emission lines that are covered by the SDSS spectra of low-redshift quasars (such as C iv, Mg ii, and Hβ). This allowed assessment of the emission-line weakness of our low-redshift, radio-quiet WLQ candidates compared to typical SDSS quasars. Furthermore, our method of selecting low-redshift WLQ candidates is different from that for high-redshift WLQs because low-redshift quasars do not have coverage of Lyα + N v emission in their SDSS spectra. The REW distributions of emission lines such as C iv, Mg ii, or Hβ may give insights into universal selection criteria for WLQ candidates at different redshifts.

We utilized the REW measurements of C iv, Mg ii, and Hβ from the catalog of Shen et al. (2011) for SDSS DR7 quasars (see their Table 1). For the C iv and Mg ii lines, we adopted the REW values for the entire lines, while for Hβ we added the REW values of the broad and narrow components to obtain the REW of the entire line. Shen et al. (2011) reported the REW distributions of these lines for all SDSS DR7 quasars having applicable REW measurements (see their Figures 12–14). In this work, we selected unbiased samples of SDSS quasars with high-quality optical/UV spectra to study the REW distributions of these emission lines by imposing the following criteria.

• 1.  
  We only use the DR7 quasars selected with the final algorithm given by Richards et al. (2002) to maintain consistency with DS09.
• 2.  
  BAL quasars that were cataloged in Gibson et al. (2009) and Shen et al. (2011) were removed.
• 3.  
  We restricted the redshift ranges of the objects for each line (C iv: z = 1.55–4.67; Mg ii: z = 0.42–2.15; Hβ: z < 0.84) to ensure that the SDSS spectra of these objects cover the whole region of each emission line defined by λ_lo and λ_hi listed in Table 2 of Vanden Berk et al. (2001).
• 4.  
  To select objects with high-quality SDSS spectra, we calculated the S/N of continuum regions close to each emission line, SN₁₇₀₀ for C iv, SN₃₀₀₀ for Mg ii, and SN₅₁₅₀ for Hβ (see Table 3), following the method of Gibson et al. (2009). SN₁₇₀₀, SN₃₀₀₀, and SN₅₁₅₀ are calculated as the median of the ratio between the flux and the error (obtained from the SDSS pipeline) for all the spectral bins in the rest-frame 1650–1750 Å, 2950–3050 Å, and 5100–5200 Å regions, respectively. These wavelength regions are free of strong emission and/or absorption features, but are still close to the above emission lines in our study. We require each of the above continuum S/N to be greater than 7. The S/N values of the emission lines themselves were not utilized because that would introduce bias against objects with weak emission lines.
• 5.  
  We eliminated the objects that have large fractions of bad pixels in their SDSS spectra in the wavelength ranges of these emission lines (these lead to unreliable measurements). We imposed the following cuts on the numbers of pixels that were included in the fitting for each line given in the catalog of Shen et al. (2011): LINE_NPIX_CIV >250 for C iv, LINE_NPIX_MGII >300 for Mg ii, and LINE_NPIX_HB >150 for Hβ.

Applying the above cuts on redshift, continuum S/N, and the numbers of pixels in the fitting (criteria 3–5 above) removed 35%, 27%, and 37% of the objects in the REW distribution investigations for C iv, Mg ii, and Hβ, respectively. All the above restrictions and quality cuts are necessary because unreliable line measurements could significantly affect the REW distributions particularly in the tails with low or high REW values.

Figure 3 shows the REW distributions of the C iv, Mg ii, and Hβ lines for our selected samples of SDSS quasars. We also include the histogram for Lyα + N v from DS09 for comparison. DS09 also showed a histogram for C iv REWs of SDSS DR5 quasars, which has a similar profile as the C iv REW histogram in Figure 3. We fit the REW histogram of each line with a lognormal distribution (see the blue solid lines in Figure 3; also see the dotted and dashed lines for the 2σ and 3σ ranges of each lognormal model). The REW measurements for the quasars in our low-redshift WLQ sample (see Section 3.1) are also shown in Figure 3. All the C iv REW values of our sources are far below the negative 3σ deviation of the lognormal distribution. Most of the REW(Mg ii) values are also below the negative 3σ deviation of the lognormal distribution; the largest REW(Mg ii) value for our sample is close to the negative 2σ deviation (see the top right panel of Figure 3).

Zoom In Zoom Out Reset image size

The REW distribution of Lyα + N v in DS09 shows a prominent tail toward low REW values; this tail is the basis on which the high-redshift WLQs are defined. The C iv REW distribution shows similar behavior in that a more prominent skew tail toward low REW values exists, while there is no corresponding tail toward high REW values. However, the histogram of the Mg ii REWs appears more symmetric, with only small tails toward both the low end and high end of the REW distribution. The Hβ REW distribution is similar to that for Lyα + N v; a prominent tail toward low REW values exists, while the tail toward high REW values is much less significant. We randomly chose sets of sources in the tails with low or high REW values in the histograms for C iv, Mg ii, and Hβ and then visually examined their SDSS spectra and the quality assessment plots¹² for their individual spectral fits in Shen et al. (2011). These sources have good spectral-fit quality; their REW measurements should be reliable. It is worth noting that the quasars in the REW histograms for various lines have different ranges of redshift and luminosity, which may affect their REW distributions (e.g., via the Baldwin effect; Baldwin 1977). We test this hypothesis by comparing the C iv and Mg ii REW distributions for a set of SDSS quasars with 1.55  <  z  <  2.15 and 45  <  log νL₃₀₀₀  <  46 (νL₃₀₀₀ is the luminosity at rest-frame 3000 Å in erg s⁻¹, obtained from Shen et al. 2011). Figure 4 shows that both the C iv and Mg ii REW distributions show similar behavior to the distributions presented in Figure 3; the C iv histogram has a significant tail at the low REW end, while the Mg ii histogram is symmetric with weak tails. While the physical reason for the different emission-line REW distributions is unclear, the Mg ii line has low optical depth, while the C iv line has much higher optical depth (e.g., see Eracleous et al. 2009). In the context of a disk-wind model for the BELR (e.g., Murray & Chiang 1997; Proga et al. 2000), the C iv emission is considered to be mainly from the active galactic nucleus (AGN) wind, while the Mg ii line mostly originates from the accretion disk (e.g., Leighly 2004; Richards et al. 2011).

Zoom In Zoom Out Reset image size

The 3σ tail toward low REW values of the C iv histogram is defined by REW(C iv) ≲ 10 Å, which is consistent with that derived in DS09. Using this criterion one could perhaps hope to extend the redshift range of WLQ selection from z > 3 down to z > 1.5. However, this definition of WLQs has significant inconsistency with that in DS09 based on REW (Lyα + N v)  <  15.4 Å. Figure 5(a) shows the relation between the REW values of Lyα + N v and C iv, and the best-fit power-law model found using the IDL routine linmix_err.¹³ The shaded area shows the 90% confidence uncertainty range obtained via a nonparametric bootstrap method (Efron 1979). Although the REW values of these two lines are positively correlated, large scatter exists. Over half of the sources with REW (Lyα + N v)  <  15.4 Å have REW(C iv) >10 Å, and vice versa (see Figure 5(a)). The large scatter is likely to be intrinsic, since the REW measurement errors (see Figure 5 for typical error bars) are much smaller than the scatter. DS09 also suggested the inconsistency between sources with weak Lyα + N v and those with weak C iv emission. A total of 39 of the 74 WLQs cataloged in DS09 have REW(C iv) >10 Å based on the measurements in Shen et al. (2011).¹⁴ As stated in DS09, their measurements may underestimate REW(Lyα + N v) when strong intervening absorption exists; this is perhaps one reason for the inconsistency. The correlation between the REWs of C iv and Mg ii also has significant scatter, and so does the correlation between the REWs of Mg ii and Hβ (see Figures 5(b) and (c)). The red shaded histograms in Figure 4 show the REW(C iv) distribution for sources with weak Mg ii and the REW(Mg ii) distribution for sources with weak C iv. While the sources with weak Mg ii also tend to have weak C iv (below the negative 2σ deviation), many objects with weak C iv have fairly strong Mg ii (see the bottom panel of Figure 4).

Zoom In Zoom Out Reset image size

Given the results above, it is therefore difficult to find consistent criteria for WLQs at different redshifts using different emission lines even though the REW distributions of C iv and Hβ show similar behavior to that of Lyα + N v. Since there is no single line that is covered by SDSS spectroscopy for quasars at all redshifts between zero and six, it appears that there is not a straightforward, direct way to define universal selection criteria for WLQs at all redshifts solely based on SDSS spectroscopy. Therefore, we only choose our low-redshift WLQ candidates mainly from the catalog of Plotkin et al. (2010a), which has a strict criterion on all emission-line strengths (REW ≲ 5 Å). Future UV spectroscopy that covers the Lyα + N v and/or the C iv regions for low-redshift WLQs will provide insights toward a universal definition for WLQs.

The Δα_ox parameter (see Equation (2) for definition) is utilized to assess the X-ray brightness of a quasar relative to typical radio-quiet quasars with similar UV luminosity. We compare the Δα_ox distribution of our low-redshift, radio-quiet WLQ candidates (see Figure 6) to that of the 132 radio-quiet, non-BAL quasars in Sample B of Gibson et al. (2008),¹⁷ which represent typical radio-quiet SDSS quasars. All of the 132 Sample B quasars are X-ray detected. The Peto-Prentice test (e.g., Latta 1981), implemented in the Astronomy Survival Analysis package (ASURV; e.g., Lavalley et al. 1992), is used to assess whether our low-redshift WLQ candidates follow the same Δα_ox distribution as that for typical quasars (see results in Table 6). We prefer the Peto-Prentice test to other possible similar tests because it is the least affected by the factors of different censoring patterns or unequal sizes of the two samples which exist in our case. We also compare the Δα_ox distribution of high-redshift, radio-quiet WLQs in Shemmer et al. (2006, 2009) to that of our low-redshift, radio-quiet WLQ candidates and that of typical SDSS quasars (also see Figure 6 and Table 6).

Zoom In Zoom Out Reset image size

The Δα_ox distribution of our low-redshift, radio-quiet WLQ candidates is significantly different from that of typical SDSS quasars. The probability of null-hypothesis (two samples following the same distribution) is only 6.3 × 10⁻⁷. This result is mainly due to the presence of a skew tail of X-ray weak WLQs (see Figure 6). Seven out of the 11 objects in our sample of low-redshift, radio-quiet WLQ candidates have Δα_ox <−0.2, giving a fraction of X-ray weak objects of (64⁺³⁴ _{− 24})% (68% confidence level). The mean Δα_ox value for the low-redshift, radio-quiet WLQ candidates is −0.214 ± 0.078, calculated with the Kaplan–Meier estimator also implemented in the ASURV package, while that for the Sample B quasars is −0.001 ± 0.011. The Δα_ox distribution of the nine high-redshift, radio-quiet WLQs in Shemmer et al. (2006, 2009) is also different from that of typical radio-quiet SDSS quasars, but less significantly (the probability of null-hypothesis is 7.1 × 10⁻³).¹⁸ Five of the nine objects in the high-redshift, radio-quiet WLQ sample have Δα_ox  <  − 0.2, giving a fraction of X-ray weak objects of (56⁺³⁷ _{− 24})% (68% confidence level); we note that the fraction could be somewhat higher (6/9) owing to the weak X-ray upper limit for J1237+6301. The mean Δα_ox value for the high-redshift, radio-quiet WLQs is −0.144 ± 0.075. As expected, the combined low-redshift and high-redshift, radio-quiet WLQ sample (mean Δα_ox value of −0.187 ± 0.056) also follows a different Δα_ox distribution from that of typical SDSS quasars (null-hypothesis probability of 4.5 × 10⁻⁶). The Δα_ox distribution of low-redshift, radio-quiet WLQ candidates is consistent with that of high-redshift, radio-quiet WLQs (null-hypothesis probability of 0.35), though the sample sizes being compared are limited.

To investigate the multi-band properties of low-redshift, radio-quiet WLQ candidates, we plotted the sources of our sample in an α_ro–α_ox diagram (Figure 7) along with the high-redshift, radio-quiet WLQs in Shemmer et al. (2006, 2009), the BL Lac sample in Shemmer et al. (2009), and the Sample B quasar in Gibson et al. (2008). The low-redshift, radio-quiet WLQ candidates have similar multi-band properties to those of high-redshift, radio-quiet WLQs. They are generally much fainter in radio and X-rays than most of the BL Lac objects. The weak emission lines of low-redshift, radio-quiet WLQ candidates are therefore not likely due to the dilution by relativistically boosted continua as for BL Lac objects (see discussion in Section 4.1 of Shemmer et al. 2009). However, it is possible that a small percentage of the WLQ candidates actually belong to the radio-faint tail of the BL Lac population (see below).

Zoom In Zoom Out Reset image size

The population of radio-quiet WLQ candidates (both at low redshift and high redshift) has a wide dispersion of relative X-ray brightness and UV emission-line properties. Motivated by their Δα_ox distribution, their emission-line properties discussed below, and observations of related objects (e.g., Wu et al. 2011), we will discuss them in three groups.

The majority of WLQ candidates are not only X-ray weaker than BL Lac objects, but also weaker than typical radio-quiet SDSS quasars. These WLQ candidates may belong to the notable class of X-ray weak quasars termed "PHL 1811 analogs" which were recently studied in detail by Wu et al. (2011). The PHL 1811 analogs generally have weak and highly blueshifted high-ionization lines (e.g., C iv, Si iv), weak semiforbidden lines (e.g., C iii]), and strong UV Fe ii and/or Fe iii emission. Some of our low-redshift, radio-quiet WLQs have similar UV emission-line properties to those of PHL 1811, as listed below.

• 1.  
  J0812+5225 (Δα_ox = − 0.42) has weak C iii] and strong Fe ii emission.
• 2.  
  J0945+1009 (Δα_ox  <  − 0.34) has weak C iv and C iii] emission lines. Its C iv line is highly blueshifted (≈ − 7000 km s⁻¹).
• 3.  
  J1252+2640 (Δα_ox =−0.39) has weak C iii] and strong Fe ii emission.
• 4.  
  J1139−0201 (Δα_ox =−0.38) has weak and highly blueshifted (≈ − 2950 km s⁻¹) C iv emission, weak C iii] emission, and strong Fe iii emission.

A high-redshift, radio-quiet WLQ J1302+0030 (Δα_ox =−0.38) also has a weak and highly blueshifted C iv emission line (DS09; Wu et al. 2011). All of the above mentioned sources are X-ray weak by a factor of >7 (see Table 5). Figure 8 shows the distribution of our WLQ candidates in the Δα_ox–C iv blueshift–REW(C iv) parameter space. J0945+1009, J1139−0201 and J1302+0030 are similar to PHL 1811 analogs in this diagram. Based on the model in Section 4.6 of Wu et al. (2011), these PHL 1811 analogs may have high-ionization shielding gas with large column density and a large covering factor of the BELR which blocks most of the ionizing photons, resulting in weak high-ionization emission lines. If a quasar of this kind is viewed through the BELR and shielding gas, it would be an X-ray weak WLQ with weak and highly blueshifted high-ionization lines (e.g., C iv). Based on the estimate in Section 4.6 of Wu et al. (2011), PHL 1811 analogs should make up ≈30% of the total WLQ population. However, our sample of low-redshift, radio-quiet WLQ candidates appears to have a higher fraction (≳ 50%) of PHL 1811 analogs, which may indicate that our sample has some selection bias toward X-ray weak WLQ candidates. This bias could perhaps be the result of a stricter criterion on the strengths of emission lines for most sources in our sample (REW ≲ 5 Å). Quasars with weaker emission lines (e.g., C iv) are perhaps more likely to be weak in X-rays (e.g., see Section 4.5 of Wu et al. 2011). The apparently higher fraction of PHL 1811 analogs in our sample than that in Wu et al. (2011) may perhaps also be caused simply by small-sample statistics. A Fisher's exact test (Fisher 1922) gives an 11.1% probability for the different fractions of PHL 1811 analogs among these two samples under the null-hypothesis (i.e., the two samples have the same fraction of PHL 1811 analogs).

Zoom In Zoom Out Reset image size

Some of our WLQ candidates have similar X-ray brightness to that of typical radio-quiet quasars (−0.2 ≲ Δα_ox ≲ 0.2). Their high-ionization lines are also weak, but perhaps not highly blueshifted (e.g., see Figure 8 for J1408+0205). Some of them (e.g., J1612+5118) have very weak UV Fe ii and/or Fe iii emission. J1612+5118 does seem to have a highly blueshifted C iv line for which the reason is unclear. However, the C iv line of this source is close to the blue border of its SDSS spectral coverage. Further UV spectroscopy with better C iv coverage is needed to confirm its C iv blueshift. Based on the model in Wu et al. (2011), these sources are similar to PHL 1811 analogs physically, but they are viewed at different orientations. These sources are observed along lines of sight that avoid the shielding gas and the BELR. Therefore, they appear normal in X-rays. Their high-ionization lines are generally not highly blueshifted.

In our WLQ candidate sample, one source (J1109+3736) is remarkably strong in X-rays. It also shows similar UV/optical spectral properties to those of BL Lac objects. This source may belong to the radio-faint tail of the BL Lac population; we will discuss it further in Section 5.5. It is worth noting that the division of our radio-quiet WLQ candidates into the three groups discussed above (as shown in Figure 7) is somewhat arbitrary. We do have some "border-line" sources with Δα_ox ≈±0.2 (e.g., J1212+5341). It is difficult to classify these sources clearly based on current information.

For the purpose of investigating further the multiwavelength SEDs of our low-redshift, radio-quiet WLQs, we gathered photometry for our sample from the following bands: (1) near- and mid-infrared from Wide-field Infrared Survey Explorer (WISE; Wright et al. 2010), (2) near-infrared from Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006), (3) optical from the SDSS, (4) UV from Galaxy Evolution Explorer (GALEX; Martin et al. 2005), and (5) X-rays from this work. Figure 9 shows the SEDs of the five low-redshift WLQs in our sample which have the best multi-band coverage. We also include the SED of the radio-quiet BL Lac candidate J1109+3736 (see more discussion in Section 5.5). A key point to keep in mind is that these multi-band observations are non-simultaneous. The SEDs are therefore subject to potential distortions due to variability.

Zoom In Zoom Out Reset image size

We examined the WISE and GALEX image tiles by eye to identify potential cases of source blending, confusion, or incorrect matching caused by the low angular resolution of WISE and GALEX. None of the sources in our sample is subject to these kinds of problems. The 2MASS magnitudes in the SDSS DR7 quasar catalog were utilized; this catalog provided aperture photometry for additional sources detected down to 2σ (see Section 5 of Schneider et al. 2010). For the sources without detections at ⩾2σ, we adopted flux upper limits obtained following the same photometry procedure (C. M. Krawczyk & G. T. Richards 2011, private communication). The first five sources in Figure 9 are from the groups of X-ray weak and X-ray normal WLQ candidates discussed above. Their mid-infrared-to-UV SEDs are generally consistent with the composite SEDs of typical SDSS quasars in Richards et al. (2006), and they are significantly different from the SEDs of BL Lac objects (see the dotted parabolic line in the top-left panel of Figure 9; e.g., Nieppola et al. 2006). We also investigated the SEDs for the WISE-covered radio-quiet objects cataloged in Plotkin et al. (2010a) which do not have sensitive X-ray coverage (see the Appendix). The majority of them also have mid-infrared-to-UV SEDs consistent with those of typical radio-quiet quasars. Lane et al. (2011) obtained similar results for their high-redshift WLQs; the composite SED of their high-redshift WLQs is inconsistent with SEDs of BL Lac objects. For one source in our sample, J0945+1009, the flux in the UV band is lower than for typical SDSS quasars (see the GALEX data points in the top-right panel of Figure 9). The UV deficiency of this source may be caused by Lyα-forest intervening absorption. However, Laor & Davis (2011) argued that such intervening absorption is not significant (∼11% at most) for this source with z = 1.66. The near-infrared-to-UV SED of J0945+1009 can be well fitted with their local blackbody model for a cold accretion disk.

Most of the low-redshift, radio-quiet WLQ candidates do not have sufficient X-ray counts for an individual X-ray spectral analysis. We therefore investigate the average X-ray spectral properties for these sources via stacking analyses and joint fitting.

A stacked spectral analysis was performed for the six low-redshift, radio-quiet WLQ candidates with Δα_ox <−0.3 (J0812+5225, J0945+1009, J1139−0201, J1252+2640, J2115+0001, and J2324+1443). These sources are the weakest in X-rays among the full sample of low-redshift WLQ candidates. The detected sources have similar numbers of X-ray counts, so that any one of them will not dominate the stacking analysis. These six sources span a relatively wide range of redshift (z = 1.15–2.50), and thus the observed-frame bands of each source correspond to different energy ranges in the rest frame. However, under the assumption of a simple power-law spectral model, one can stack the X-ray counts to obtain the average effective power-law photon index. We added the X-ray counts of these sources in the observed-frame soft band and hard band, respectively. The numbers of total net counts are 14.4^+4.9 _{− 3.7} in the soft band and 5.2^+3.4 _{− 2.2} in the hard band (68% confidence level), and the resulting band ratio is 0.36^+0.27 _{− 0.18}. With the average Galactic neutral hydrogen column density of these sources (N_H = 3.50 × 10²⁰ cm⁻²), the band ratio was converted to an effective power-law photon index Γ = 1.66^+0.63 _{− 0.51}. The average X-ray spectrum of the X-ray weak low-redshift WLQs is perhaps somewhat harder than that for typical radio-quiet quasars (Γ ≈ 2), but it is consistent within the error bars. This average X-ray spectrum is likely softer than that of the PHL 1811 analogs at z = 2.19–2.38 (Γ = 1.10^+0.45 _{− 0.40}) in Wu et al. (2011) which was also obtained via a stacking analysis, but consistent within 2σ. Both stacking analyses suffer from large uncertainty due to limited X-ray counts. For a sample combining the six X-ray weak WLQs analyzed here (which are likely to be PHL 1811 analogs) and the radio-quiet PHL 1811 analogs in Wu et al. (2011), the average X-ray spectrum has a flat effective power-law photon index of Γ = 1.35^+0.33 _{− 0.31}. Deeper X-ray observations are necessary to give tighter constraints on the X-ray spectral properties of these X-ray weak WLQ candidates.

Two sources (J0945+1009 and J2115+0001) are undetected by Chandra. Adding the X-ray counts of these two sources cannot generate a stacked source that would be detected by Chandra because both of these two sources have zero X-ray counts. However, we are able to obtain a tighter average constraint on their X-ray brightness via stacking analysis. The upper limit on the soft-band count rate of the stacked source is 2.51 × 10⁻⁴ s⁻¹. Average values of redshift, Galactic N_H, and f_2500 Å are adopted in the following calculation. The upper limit on the average flux density at rest-frame 2 keV is 5.16 × 10⁻³³ erg cm⁻² s⁻¹ Hz⁻¹ under the assumption of the Galactic-absorbed power law with Γ = 2. The upper limits on α_ox and Δα_ox are calculated to be α_ox <−2.21 and Δα_ox <−0.50. Therefore, these two sources are X-ray weak by a factor of >20 on average.

For the two X-ray normal, low-redshift WLQs (J1604+4326 and J1612+5118), we performed joint fitting to study the average X-ray spectral properties of these sources. The X-ray spectra were extracted from apertures of 3'' radius centered on the X-ray positions of these sources via the standard CIAO routine psextract. The background spectra were extracted from annular regions with inner radii of 6'' and outer radii of 9'', which are free of X-ray sources. Two spectra were extracted individually for J1604+4326 from its two Chandra observations. Spectral fitting was performed with XSPEC v12.6.0 (Arnaud 1996). The C-statistic (Cash 1979) was used in the spectral fitting instead of the standard χ² statistic because the C-statistic is well suited to the limited X-ray counts in our analysis (e.g., Nousek & Shue 1989). We fit the spectra jointly using a power-law model with a Galactic absorption component represented by the wabs model (Morrison & McCammon 1983). We also used another model similar to the first, but adding an intrinsic (redshifted) neutral absorption component, represented by the zwabs model. Both sources were assigned their own values of redshift and Galactic neutral hydrogen column density; the Galactic column density was fixed to the values calculated with COLDEN (Column 4 of Table 5). The joint fitting results are shown in Table 7. The quoted errors or upper limits are at the 90% confidence level for one parameter of interest (ΔC = 2.71; Avni 1976; Cash 1979). The average X-ray spectral properties of these two sources are similar to those of typical radio-quiet quasars. They have an average photon index (Γ = 2.07^+0.31 _{− 0.30}), consistent with that of typical radio-quiet quasars (Γ ≈ 2). The average photon index is also consistent with those from their band-ratio analyses. We did not find evidence of strong intrinsic neutral absorption for J1604+4326 and J1612+5118 (N_H ≲ 1.58 × 10²² cm⁻²); the spectral-fitting quality was not improved after adding the intrinsic neutral absorption component.

Two sources (J1109+3736 and J1530+2310) have sufficient X-ray counts for individual spectral analysis. We will discuss the X-ray spectral properties of J1530+2310 here and leave the radio-quiet BL Lac candidate J1109+3736 for the next subsection. The X-ray spectrum of J1530+2310 was extracted following the same procedure as for J1604+4326 and J1612+5118 (see above). The spectrum of J1530+2310 was first grouped to have at least 10 counts per bin (see Figure 10). We used the same spectral models as those for the joint fitting above. The standard χ² statistic was utilized. The fitting results are also shown in Table 7. Figure 10 shows the X-ray spectrum, the best-fit power-law model with Galactic absorption, and the contour plot of the Γ–N_H parameter space for the spectral model with intrinsic neutral absorption for J1530+2310. The X-ray spectral properties of J1530+2310 are similar to those of J1604+4326 and J1612+5118. The photon index of J1530+2310 from the spectral fitting (Γ = 2.11^+0.37 _{− 0.34}) is consistent with that from the band-ratio analysis (see Table 4). J1530+2310 also shows no evidence of strong intrinsic neutral absorption (N_H ≲ 2.67 × 10²² cm⁻²). In summary, the X-ray spectral properties of X-ray normal low-redshift WLQs are consistent with those of typical radio-quiet quasars. Shemmer et al. (2009) found similar results for their high-redshift, radio-quiet WLQs. Their sources have a somewhat harder average X-ray spectrum (Γ = 1.86^+0.72 _{− 0.48}), perhaps because they fit both X-ray weak and X-ray normal WLQs jointly. Shemmer et al. (2009, 2010) also performed individual X-ray spectral analysis on two high-redshift, radio-intermediate WLQs (J1141+0219 and J1231+0138).

Zoom In Zoom Out Reset image size

J1109+3736 is a radio-quiet BL Lac candidate based on its multi-band properties. It is strong in X-rays by a factor of ≈6.3 (α_ox =−1.10, Δα_ox =0.31). Its α_ox value is similar to that of the majority of the BL Lac population (see Figure 7). X-ray spectral analysis was carried out for J1109+3736 following similar procedures as those in Section 5.4. The Chandra spectrum of J1109+3736 was grouped to have at least 20 counts per bin (see Figure 10). The best-fit parameters are listed in Table 7. The X-ray spectrum, the best-fit power-law model with Galactic absorption, and Γ–N_H contours for J1109+3736 are also shown in Figure 10. The best-fit photon index of J1109+3736 (Γ = 1.77 ± 0.14) is consistent with that from the band-ratio analysis. This photon-index value indicates a harder X-ray spectrum than those for the majority of the high-energy peaked BL Lac objects (HBLs),¹⁹ which have a mean photon index Γ ≈ 2.2, but it is still consistent with the broad distribution of HBL photon-index values (e.g., see the bottom panel of Figure 1 in Donato et al. 2005). Although this source has a high X-ray count rate, we do not expect strong photon pile-up effects because a 1/2 subarray mode was used for its Chandra observation. There is no evidence for significant intrinsic absorption in the J1109+3736 spectrum (N_H ≲ 1.4 × 10²¹ cm⁻²).

The SDSS spectrum of J1109+3736 (see the bottom-right panel of Figure 2) shows a strong power-law continuum without any detectable emission lines, in particular no Balmer lines. Note that some of the X-ray weak, low-redshift WLQ candidates (e.g., PHL 1811; see Leighly et al. 2007a) have fairly strong Balmer lines. The strength of its Ca ii H/K break²⁰ (C = 0.185) indicates a relatively small contribution to the SDSS spectrum from the host galaxy. J1109+3736 shows moderate X-ray variability between its Chandra observation and the epoch of the ROSAT All Sky Survey (RASS) in 1990 November (Voges et al. 1999). It was not detected by RASS. The upper limit for α_ox is estimated to be α_ox  <   − 1.29, showing a factor of >3 variation compared to its Chandra observation. The quasar should have been detected in the RASS if it had the same X-ray brightness as that in the Chandra epoch (expecting ≈20 counts in a 250 s RASS observation). There is also evidence for optical variability of J1109+3736. This source was optically identified during the Cambridge Automated Plate Measuring Machine (APM) scans of Palomar Observatory Sky Survey (POSS) plates (e.g., McMahon et al. 2002). We converted the APM magnitudes in the O and E bands (O = 20.76, E = 19.12) into a B-band magnitude (B = 20.57) using Equations (1)–(3) in McMahon et al. (2002). We also converted the SDSS magnitudes to a B-band magnitude (B = 18.88) using the equations in Table 1 of Jester et al. (2005). The B-band magnitude difference is 1.69, indicating a factor of ∼5 variability over a time span of ∼50 years. This source shows greater variability than most quasars during this time span (e.g., see Figure 24 of Sesar et al. 2006). Therefore, J1109+3736 is a BL Lac candidate on the radio-faint tail of the full BL Lac population.

The broadband SED of J1109+3736 is shown in Figure 9. It is relatively weak in the near-infrared compared to its SDSS brightness. However, the SED profile of this source could be strongly affected by its variability (see the comparison between its SDSS and POSS fluxes in Figure 9) given its possible BL Lac nature. Infrared photometry from WISE, which will be available in the WISE full data release,²¹ will be helpful because the infrared-to-UV SEDs of BL Lac objects are very different from those of typical quasars (e.g., Lane et al. 2011). Future polarization measurements of J1109+3736 will also be useful in understanding its nature, since BL Lac objects are usually highly polarized in the optical/UV band. However, polarimetry surveys of optically selected BL Lac samples (e.g., Smith et al. 2007; Heidt & Nilsson 2011) did not find any highly polarized radio-quiet BL Lac candidates, indicating this kind of source, if it indeed exists, should make up only a small fraction of the total population of radio-quiet WLQ candidates.

The broadband SEDs of WLQ candidates are able to provide useful insights into their nature. Lane et al. (2011) showed that the mid-infrared-to-UV SEDs of their high-redshift WLQs were consistent with those of typical quasars, but were significantly different from those of typical BL Lac objects. In Section 5.3, we have obtained similar results for the X-ray weak and X-ray normal WLQ candidates in our low-redshift sample (see Figure 9). However, it is possible to gain insights from SED measurements even when sensitive X-ray data are not available. Therefore, in this appendix we will further study the mid-infrared-to-UV SEDs of the radio-quiet weak-featured AGNs cataloged in Plotkin et al. (2010a) that do not have sensitive X-ray coverage.

We cross-correlated the objects in Table 6 of Plotkin et al. (2010a) to the WISE source catalog in its preliminary data release, obtaining 30 WISE-detected sources (not including the four sources already discussed in Section 5.3). The WISE coverage of the remaining sources was checked with the online image tile look-up tool.²³ An additional six sources with WISE coverage were identified. Among these six sources, two (J0901+3846 and J1409−0000) were identified as stars based on their SDSS spectra and proper-motion data; these two sources were removed from our SED study. For the remaining four sources, we examined their WISE image tiles. Two of them (J0755+3525 and J1448+2407) have WISE detections in all four bands that lie below the catalog limit. We performed aperture photometry (using a standard 825 aperture radius) and obtained their fluxes by scaling their counts in the aperture to those of nearby sources (within 60'' separation) appearing in the WISE catalog. The other two sources (J0857+2342 and J1541+2631) were not detected in their WISE images. Following the standard WISE photometry procedure, we calculated their flux upper limits at a 95% confidence level by adding the aperture flux measurement plus two times the uncertainty. For J0857+2342, we could only obtain a flux upper limit in the w3 band because there are nearby bright sources in its aperture in the other band images. The photometric data in the other near-infrared-to-UV wavebands were obtained following the same methods described in Section 5.3. The X-ray flux limits from the RASS are taken from Table 8 of Plotkin et al. (2010a).

The SEDs of the 34 objects are shown in Figure 11. The majority of these sources have SEDs consistent with those of typical radio-quiet quasars in Richards et al. (2006), showing that they are more likely to be WLQs in nature rather than BL Lac objects. However, there are also several sources with other kinds of SED profiles, which we discuss in more detail below.

Zoom In Zoom Out Reset image size

View extended figure (3 panels)

J0834+5112, J1556+3854, J1610+3039, and J1658+6118. These four sources have very red SEDs (strong in the infrared and weak in the optical/UV). They also have very red SDSS spectra. They are more likely to be absorbed quasars than bona fide WLQs.

J1421+0522. This source has an SED profile peaking in the near-infrared band, which is more similar to the SEDs of BL Lac objects.

J1522+4137. The SED of this source appears similar to that of J0945+1009, which can be fit with a cold accretion disk model as in Laor & Davis (2011). It is probably a high-redshift, radio-quiet WLQ.

J1633+4227. The SED of this source peaks in the near-infrared band and drops rapidly in both blueward and redward directions. This SED profile is similar to that of a radio-loud BL Lac object with significant host-galaxy contamination (e.g., J0823+1524 in Plotkin et al. 2012). J1633+4227 also has the same Ca ii H/K break value (C = 0.33) as that of J0823+1524. It is possible that J1633+4227 is a radio-quiet BL Lac object with substantial host-galaxy contamination, although its radio-loudness needs to be constrained more tightly (currently α_ro >−0.19 estimated from its FIRST coverage). The near-infrared peak of the SED of J1633+4227 could also perhaps be caused by extreme variability.