THE X-RAY PROPERTIES OF THE OPTICALLY BRIGHTEST MINI-BAL QUASARS FROM THE SLOAN DIGITAL SKY SURVEY

Absorption lines in quasar spectra provide useful probes of outflows in active galactic nuclei (AGNs). Broad absorption lines (BALs) are traditionally defined to have velocity widths ⩾2000 km s⁻¹ (e.g., Weymann et al. 1991) in transitions such as C iv λ1549, Si iv λ1400, Al iii λ1857, and Mg ii λ2799. The number of quasars with intermediate velocity width absorption lines (1000–2000 km s⁻¹), known as mini-BALs, is comparable to or greater than the number with traditional BALs (e.g., Trump et al. 2006; Gibson et al. 2009a, hereafter G09). BAL and mini-BAL outflows are powerful in luminous quasars, partly indicated by their high velocities, and they could be responsible for significant feedback into quasar host galaxies (e.g., Ganguly & Brotherton 2008; Brandt et al. 2009). These outflows are often envisioned as radiatively accelerated equatorial disk winds, and it appears that they must be significantly shielded from the soft X-ray emission from the central region in order to avoid overionization of the outflow, which would effectively prevent radiative acceleration (e.g., Murray et al. 1995; Proga et al. 2000). Multiwavelength observations of both BAL and mini-BAL quasars must be compiled in order to determine the physical characteristics and processes governing AGN outflows. Furthermore, multiwavelength data will also help determine whether mini-BALs are part of a continuum of absorption phenomena whose progenitor is the same as BALs and narrow absorption lines (NALs), or if mini-BALs instead constitute a different physical phenomenon.

It has been suggested (e.g., Knigge et al. 2008) that mini-BAL quasars are part of a population that is different from BAL quasars, but UV absorption-line strength and maximum outflow velocity appear to be correlated with relative X-ray brightness (i.e., an estimate of the level of X-ray absorption), with mini-BAL quasars generally being intermediate between BAL and non-BMB (meaning neither BAL nor mini-BAL) quasars (e.g., Gallagher et al. 2006; G09). The effective hard X-ray photon index, Γ, is also correlated with the relative X-ray brightness for combined radio-quiet BAL, mini-BAL, and non-BMB quasar samples. These results suggest a physical link between BALs and mini-BALs. On the other hand, there are also cases in which the outflow velocity for a mini-BAL is higher than expected for a given X-ray brightness, which indicates that some mini-BALs may not require strong X-ray absorption or may be launched in a different way than is typical (G09). Furthermore, the classical definition of "mini-BAL" using absorption index (AI, see Section 2) cannot fully address the complex absorption features in quasar spectra (e.g., G09).

In this paper, we study the X-ray properties of 14 of the optically brightest radio-quiet quasars whose spectra contain a C iv mini-BAL in the Sloan Digital Sky Survey (SDSS; York et al. 2000) Data Release 5 (DR5) quasar catalog (Schneider et al. 2007) using a combination of new Chandra snapshot observations and archival XMM-Newton observations. We seek to extend the X-ray coverage of mini-BAL quasars, to study nuclear outflows, and to evaluate the properties of mini-BAL quasars relative to BAL and non-BMB quasars. Compared to the mini-BAL quasars in G09, our more luminous sources cover a different portion of the luminosity–redshift plane (see Figure 1(a)) which has not been explored. Our sources have uniform, high-quality X-ray data and a 100% detection rate (see Section 3). Furthermore, our sources are among the optically brightest mini-BAL quasars (see Figure 1(b)), which are also generally likely to be the X-ray brightest. They may be the best objects for study in the X-ray band using future missions, e.g., the International X-ray Observatory (IXO; e.g., White et al. 2010), as well as in the optical band.

Zoom In Zoom Out Reset image size

Unless otherwise specified, a cosmology with H₀ = 70 km s⁻¹ Mpc⁻¹, Ω_M = 0.3, and Ω_Λ = 0.7 will be assumed. Furthermore, the term "HiBAL" refers to BAL quasars with high-ionization states only (e.g., C iv), while "LoBAL" refers to BAL quasars with low-ionization states (e.g., Al iii or Mg ii) and usually also high-ionization states (e.g., Gibson et al. 2009b). We define positive velocities for material flowing outward with respect to the quasar emission rest frame. Section 2 discusses our sample selection, Section 3 details our X-ray data reduction and analysis, and Section 4 discusses the implications of our results. Section 5 is a summary of our conclusions.

We sorted the SDSS DR5 quasar catalog (Schneider et al. 2007), which covers ≈5740 deg², by m_i and identified all of the mini-BAL quasars with m_i⩽ 17.5 and z ⩾ 1.9, which resulted in a list of 29 robustly identified quasars whose mini-BAL nature (as described further below) is insensitive to reasonable continuum placement. The redshift requirement ensures that all objects considered have high-quality C iv coverage in their SDSS spectra. The magnitude limit was chosen to provide a sample size that could be observed with a practical amount of Chandra observation time. The SDSS quasar catalog BEST photometric apparent i-band magnitudes were used wherever possible, but if the BEST m_i = 0, then the TARGET value of m_i was used. Also, we sought radio-quiet objects that were not detected by the FIRST survey (Becker et al. 1995). "Radio-quiet" is commonly defined as R < 10, where R is calculated by Equation (5) in Section 3. We required our potential targets to be covered, but undetected by the FIRST survey to minimize any X-ray emission contribution from possible jets associated with the quasars. Six of the 29 objects had FIRST detections and were therefore excluded. Furthermore, two of the radio-quiet mini-BAL quasars on our list already had X-ray coverage; we included these archival objects in our sample but did not propose new targeted Chandra observations for them. We proposed short "snapshot" (4–6 ks) Chandra observations of the resulting 21 mini-BAL quasars, generally prioritizing the targets by their brightness (m_i). We were awarded observing time in Cycle 10 for 12 of the brightest objects on the target list.

Our resulting sample thus consists of 14 of the optically brightest mini-BAL quasars in the SDSS DR5 quasar catalog, as shown in Figure 1 (12 new Chandra observations and two archival XMM-Newton observations). The XMM-Newton data for SDSS J142656.17+602550.8, also known as SBS 1425 + 606, were previously published in Shemmer et al. (2008). Relevant X-ray information for SDSS J160222.72+084538.4, observed by XMM-Newton at 170 off-axis, is located in the XMM-Newton Serendipitous Source Catalog (2XMMi; Watson et al. 2009). For both of the XMM-Newton observations, only data from the pn detector were used, because of its high sensitivity and number of counts with respect to the MOS detectors. An observing log is shown in Table 1. Compared to the mini-BAL quasar sample (48 objects, all of which have X-ray coverage) utilized in G09, nearly all of the objects in our sample are brighter in both apparent i-band magnitude, m_i, and absolute i-band magnitude, M_i. There are, however, two objects from G09 in Figure 1(b) that are comparably bright in m_i to our sample objects. One is the well-known gravitationally lensed quasar PG 1115+080 (e.g., Michalitsianos et al. 1996; Chartas et al. 2007). This object was not included in our sample because its redshift of 1.74 does not meet our z ⩾ 1.9 criterion. The other object, SDSS J231324.45+003444.5, likely has a high-velocity C iv BAL in its spectrum, but was included in the sample from G09 as an "ambiguous" mini-BAL source. See Section 2.1 in G09 for a more detailed discussion regarding this source.

The two quantities that were used in distinguishing mini-BALs from BALs and non-BMBs are the extended balnicity index (BI₀) and the AI. BI₀ is defined as

$$\begin{equation} \mbox{BI}_{0} = \int _{0}^{25,000} \left(1 - \frac{f(v)}{0.9}\right)C\ dv, \end{equation} \tag{ 1 }$$

where f(v) is the ratio of the observed spectrum to the continuum model for the spectrum as a function of velocity. The continuum model fit to the SDSS spectra was determined using the algorithm described in Section 2.1 of Gibson et al. (2008b). C is a parameter used to indicate BAL absorption. C = 1 if the spectrum is at least 10% below the continuum model for velocity widths of at least 2000 km s⁻¹; C = 0 otherwise. The integration limits here are the same as in G09, which differ from those used in Weymann et al. (1991), but allow characterization of BALs even at low outflow velocities.

AI is defined as

$$\begin{equation} \mbox{AI} = \int _{0}^{29,000} (1 - f(v))\ C^{\prime }dv, \end{equation} \tag{ 2 }$$

where C' = 1 when the velocity width is at least 1000 km s⁻¹ and the absorption trough falls at least 10% below the continuum; C' = 0 otherwise. The integration limits for AI are chosen such that the range of C iv outflow velocities is maximized and does not include the Si iv λ1400 emission line (Trump et al. 2006). BALs have AI>0 and BI₀ > 0; mini-BALs have AI>0 and BI₀ = 0; and non-BMBs have AI = 0 and BI₀ = 0. Thus, the objects in our sample were selected such that, for C iv λ1549, BI₀ = 0 and AI>0. All measurements of BI₀, AI, and other absorption parameters are made in the rest frame of the quasar as defined by the improved redshift measurement of Hewett & Wild (2010). One mini-BAL quasar in G09, SDSS J142301.08+533311.8, is classified as a non-BMB quasar using this improved redshift value because its only broad absorption trough lies above the velocity integration limit of 29, 000 km s⁻¹ in the AI definition.

The SDSS spectra for our sample are shown in Figure 2. These spectra are corrected for Galactic extinction according to Cardelli et al. (1989) and for fiber light loss. The light-loss correction was calculated as the average difference between the synthetic and photometric g, r, and i magnitudes for each quasar, as described in Section 3 of Just et al. (2007). The synthetic magnitudes are provided by the SDSS data pipeline and correspond to the integrated flux over the g, r, and i bandpasses in the SDSS spectrum. It is assumed that there is no variation in the flux between the SDSS photometric and spectroscopic epochs.

Zoom In Zoom Out Reset image size

The rest wavelength of C iv λ1549 is indicated in Figure 2, as well as the minimum (v_min) and maximum (v_max) outflow velocities. The wavelengths of absorption troughs can be converted to outflow velocities using the Doppler effect formulae. v_max (v_min) is the outflow velocity corresponding to the shortest (longest) wavelength associated with the C iv λ1549 mini-BAL troughs. For quasars with a single absorption trough, v_max and v_min mark the boundaries of the absorption trough where the C' parameter in Equation (2) equals unity. To keep consistency with previous works on BAL and mini-BAL quasars (e.g., Trump et al. 2006; G09; Gibson et al. 2009b), the rest wavelength used for C iv λ1549 is 1550.77 Å, which is the red component of the C iv doublet line (Verner et al. 1996). Relevant properties of the C iv mini-BALs are shown in Table 2.

Source detections were performed with the wavdetect algorithm (Freeman et al. 2002) using a detection threshold of 10⁻⁴ and wavelet scales of 1, $\sqrt{2}$, 2, $2\sqrt{2}$, and 4 pixels; all 12 Chandra targets were detected within 06 of the object's SDSS coordinates. Aperture photometry was performed for each object using IDL aper routine with an aperture radius of 3'' and inner and outer annulus radii of 6'' and 9'' for background subtraction, respectively. The resulting counts in the observed-frame soft (0.5–2.0 keV), hard (2.0–8.0 keV), and full (0.5–8.0 keV) X-ray bands are reported in Table 3. Errors on the X-ray counts were calculated using Poisson statistics corresponding to the 1σ significance level according to Tables 1 and 2 of Gehrels (1986). The band ratios and the effective power-law photon indices are also included in Table 3. The band ratio is defined as the number of hard-band counts divided by the number of soft-band counts. The errors on the band ratio correspond to the 1σ significance level and were calculated using Equation (1.31) in Section 1.7.3 of Lyons (1991). The band ratios for all of the Chandra objects can be directly compared with one another because they were all observed during the same cycle. The effective power-law photon index is the photon index of the presumed Galactic-absorbed power-law spectral model for each source, under which the band ratio would be the value calculated from the X-ray counts. It was calculated using the Chandra PIMMS ⁵ tool (version 3.9k), using the Chandra Cycle 10 instrument response to account for the decrease in instrument efficiency over time.

The key X-ray, optical, and radio properties of our sample are listed in Table 4.

Column 1: the SDSS J2000 equatorial coordinates for the quasar.

Column 2: the quasar's redshift. These values were taken from the improved measurements in Hewett & Wild (2010), which have significantly reduced systematic biases compared to the redshift values in the SDSS DR5 quasar catalog.

Column 3: the apparent i-band magnitude of the quasar using the SDSS quasar catalog BEST photometry, m_i.

Column 4: the absolute i-band magnitude for the quasar, M_i, from the SDSS DR5 quasar catalog, which was calculated assuming a power-law continuum index of α_ν = −0.5.

Column 5: the Galactic neutral hydrogen column density obtained with the Chandra COLDEN tool, in units of 10²⁰ cm⁻².

Column 6: the count rate in the observed-frame soft X-ray band (0.5–2.0 keV), in units of 10⁻³ s⁻¹.

Column 7: the Galactic extinction-corrected flux in the observed-frame soft X-ray band (0.5–2.0 keV) obtained with PIMMS, in units of 10⁻¹⁴ erg cm⁻² s⁻¹. An absorbed power-law model was used with a photon index Γ = 2, which is typical for quasars and is consistent with joint X-ray spectral fitting (see Section 4.1). The Galactic neutral hydrogen column density was used for each quasar (N_H, given in Column 5 of Table 4).

Column 8: the flux density at rest-frame 2 keV obtained with PIMMS, in units of 10⁻³² erg cm⁻² s⁻¹ Hz⁻¹. Our sources are generally brighter in 2 keV X-ray flux density than the mini-BAL quasars in G09 (see Figure 1(c)).

Column 9: the logarithm of the Galactic extinction-corrected quasar luminosity in the rest-frame 2–10 keV band obtained with PIMMS.

Columns 10 and 11: the flux density at rest-frame 2500 Å in units of 10⁻²⁷ erg cm⁻² s⁻¹ Hz⁻¹ and the logarithm of the monochromatic luminosity at rest-frame 2500 Å, respectively. The flux density was obtained using a method similar to that described in Section 2.2 of Vignali et al. (2003). It is calculated as an interpolation between the fluxes of two SDSS filters bracketing rest-frame 2500 Å. A bandpass correction is applied when converting flux into monochromatic luminosity.

Column 12: the X-ray-to-optical power-law slope, given by

$$\begin{equation} \alpha _{\rm ox} = \frac{{\rm log}(f_{\rm 2\;keV} / f_{2500\,{\rm \scriptsize \AA}})}{{\rm log}(\nu _{\rm 2\;keV} / \nu _{2500\,{\rm \scriptsize \AA}})}. \end{equation} \tag{ 3 }$$

The value of α_ox for SDSS J142656.17+602550.8 was taken directly from Shemmer et al. (2008).

Column 13: Δα_ox, defined as

$$\begin{equation} \Delta \alpha _{\rm ox} = \alpha _{\rm ox(measured)} - \alpha _{\rm ox(expected)}. \end{equation} \tag{ 4 }$$

The expected value of α_ox for a typical quasar is calculated from the established α_ox − L_2500 Å correlation given as Equation (3) of Just et al. (2007). The statistical significance of this difference, given in parentheses, is in units of σ, where σ is given in Table 5 of Steffen et al. (2006) as the rms α_ox for various luminosity ranges. Here, σ = 0.146 for 31 < logL_2500 Å < 32 and σ = 0.131 for 32 < logL_2500 Å < 33.

Column 14: the radio-loudness parameter, given by

$$\begin{equation} R = \frac{f_{\rm 5\;GHz}}{f_{4400\,{\rm \scriptsize \AA}}}{\rm.} \end{equation} \tag{ 5 }$$

The denominator, f_4400 Å, was found via extrapolation from f_2500 Å using an optical/UV power-law slope of α_ν = −0.5. The numerator, f_5 GHz, was found using a radio power-law slope of α_ν = −0.8 and a flux at 20 cm, f_20 cm, of three times the rms noise in a 05 × 05 FIRST image cutout at the object's coordinates. All the quantities of flux density here are per unit frequency. This value of f_20 cm was used as an upper limit because there were no FIRST detections. R is thus listed as an upper limit for all of the objects. All the upper limits are well below the commonly used criterion for radio-quiet objects (R < 10). Therefore any X-ray emission from jets can be neglected.

4.1. Joint X-ray Spectral Analysis

Constraining X-ray spectral properties, especially the power-law photon index and the intrinsic absorption column density, can shed light on the X-ray generation mechanisms and the nuclear environments of mini-BAL quasars. We investigate the X-ray spectra of the 12 mini-BAL quasars observed in the Chandra snapshot survey. Joint spectral fitting is performed for the 12 sources, most of which do not have sufficient counts for individual analysis. We only did spectral fitting individually for one source (SDSS J1051+4017) which has ≈210 counts. Excluding this source in the joint fitting does not significantly change the best-fit parameters, which indicates that this source does not introduce biases into the joint fitting. The X-ray spectra were extracted with the CIAO routine psextract using an aperture of 3'' radius centered on the X-ray position for each source. Background spectra were extracted using an annulus with an inner radius of 6'' and outer radius of 9''. All background regions are free of X-ray sources.

Spectral analysis was carried out with XSPEC version 12.5.1 (Arnaud 1996) using the C-statistic (Cash 1979). The X-ray spectrum for each source was grouped to have at least one count per energy bin. This grouping procedure avoided zero-count bins while allowing retention of all spectral information. The C-statistic is well suited to the low-count scenario of our analysis (e.g., Nousek & Shue 1989). Each source is assigned its own redshift and Galactic column density in the joint fitting. The following models were employed in the spectral fitting: (1) a power-law model with a Galactic absorption component represented by the wabs model in XSPEC (Morrison & McCammon 1983), in which the Galactic column density is fixed to the values from COLDEN (see Column 5 in Table 4); (2) a model similar to the first, but adding an intrinsic (redshifted) neutral absorption component, represented by the zwabs model. The errors or the upper limits for the best-fit spectral parameters are quoted at 90% confidence level for one parameter of interest (ΔC = 2.71; Avni 1976; Cash 1979).

Table 5 shows the X-ray spectral fitting results. For the individually fitted source, SDSS J1051+4017, the best-fit photon index is Γ = 1.99^+0.21_−0.21 without adding intrinsic absorption. This value is consistent with that from band-ratio analysis (see Table 3). After adding the intrinsic absorption component, the photon index changes insignificantly to Γ = 2.06^+0.37_−0.26. The power-law becomes slightly steeper. The constraint on the intrinsic column density is weak (N_H ≲ 1.8 × 10²² cm⁻²). The spectrum of this source is shown in Figure 3(a), binned to have at least 10 counts per bin for the purpose of presentation. For the joint fitting, the mean photon index is Γ = 1.90^+0.11_−0.11 in the model without intrinsic absorption. After adding the intrinsic absorption component, the mean photon index changes slightly to Γ = 1.91^+0.14_−0.11. This result is consistent with the absence of evidence for strong intrinsic neutral absorption on average (N_H ≲ 8.2 × 10²¹ cm⁻²). An F-test shows that adding an intrinsic absorption component does not significantly improve the fit quality (ΔC = 0.15 for SDSS J1051+4017; ΔC = 0.01 for the joint fitting). The XMM-Newton archival source, SDSS J1426+6025, has similar X-ray spectral properties (Γ = 1.76^+0.14_−0.13, N_H ≲ 2.3 × 10²² cm⁻²; Shemmer et al. 2008). The lack of substantial intrinsic absorption in these mini-BAL quasar spectra shows that mini-BAL quasars are more similar to non-BMB quasars in their X-ray absorption properties. Figure 3(b) shows a contour plot of the Γ–N_H parameter space for the joint fitting with intrinsic absorption added.

Zoom In Zoom Out Reset image size

Table 5. X-ray Spectral Analysis

Sources	Counts	Power Law		Power Law
	Total Full-band	with Galactic Absorption		with Galactic and Intrinsic Absorption
		Γ	C-Statistic (binsa)	Γ	NH(1022 cm−2)	C-Statistic (binsa)
SDSS J105158.74+401736.7	210	1.99+0.21−0.21	106.17(111)	2.06+0.37−0.26	<1.77	106.02(111)
All 12 Chandra snapshot survey objects	815	1.90+0.11−0.11	523.18(599)	1.91+0.14−0.11	<0.82	523.17(599)

Note. ^aThe number of bins is smaller than the number of total counts because of the grouping of the spectra (see Section 4.1).

Download table as:  ASCIITypeset image

We also investigated whether an ionized intrinsic absorber could better describe the X-ray spectra of the mini-BAL quasars. The absori model (Done et al. 1992), instead of the previously used zwabs model, was used to represent intrinsic absorption. The absori model has more spectral parameters and thus requires higher-quality X-ray spectra for the placement of useful spectral constraints. However, due to the limited counts of the sources in our sample, the X-ray spectral fitting could not provide a preference between the neutral and ionized intrinsic absorption models.

4.2. Relative X-ray Brightness

The Δα_ox parameter, defined by Equation (4), quantifies the relative X-ray brightness of a mini-BAL quasar with respect to a typical non-BMB quasar with the same UV luminosity. Figures 4 and 5 show the distributions of α_ox and Δα_ox, respectively, for our sample, as well as the non-BMB, mini-BAL and HiBAL quasars in G09. The mean values of α_ox and Δα_ox were calculated using the Kaplan–Meier estimator implemented in the Astronomy Survival Analysis (ASURV) package (e.g., Lavalley et al. 1992). A filled triangle with horizontal error bars is used to show the mean value of α_ox and Δα_ox for our sample combined with the G09 mini-BAL quasars. The mean values for α_ox and Δα_ox for the non-BMB and HiBAL quasars in the G09 sample only are shown using a filled square with horizontal error bars.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

After our sample is combined with the mini-BAL quasars in G09, the mean α_ox value, −1.68 ± 0.02, remains the same (with a smaller error bar) as the value for the mini-BAL quasar sample in G09 only. The mean Δα_ox value became slightly less negative once our sample was included: the mean Δα_ox value of our sample combined with the mini-BAL quasar sample from G09 is −0.03 ± 0.02, while it was −0.05 ± 0.03 for the G09 sample. Therefore, the mean Δα_ox of the combined mini-BAL quasar sample is closer to that of the non-BMB quasars in G09 (0.00 ± 0.01), which strengthens the trend that mini-BAL quasars are only slightly X-ray weaker than non-BMB quasars, while BAL quasars are considerably X-ray weaker than both mini-BAL and non-BMB quasars (the mean Δα_ox of the BAL quasar sample in G09 is −0.22 ± 0.04).

We have performed two-sample tests to assess if the Δα_ox values for mini-BAL quasars follow the same distribution as those for BAL quasars or non-BMB quasars. Gehan tests (Gehan 1965), also implemented in the ASURV package, were used on the distributions of Δα_ox values for the combined mini-BAL quasar sample, and the non-BMB and BAL quasar samples in G09. The Δα_ox distributions of mini-BAL and non-BMB quasars are not found to be different at a high level of significance; the probability of the null hypothesis from the Gehan test is 19.3%. In contrast, the Δα_ox distributions of mini-BAL and BAL quasars clearly differ, with a null-hypothesis probability of 0.01%. Sources with upper limits on Δα_ox are included in the two sample tests.

We have tested the normality of the distribution of Δα_ox for mini-BAL quasars. Deviations from a normal distribution, such as a skew tail, may indicate a sub-population of physically special sources (e.g., extremely X-ray weak sources) in our sample. Anderson–Darling tests (e.g., Stephens 1974) were performed on the Δα_ox values of the mini-BAL quasars. First, we examine the distribution of the 14 sources in our sample. The Anderson–Darling test cannot reject the hypothesis that the Δα_ox values follow a Gaussian distribution; the rejection probability is only 0.5%. Adding the mini-BAL quasars in G09, the sample size increases to 61. However, there are eight sources having upper limits on their Δα_ox values. Since the Anderson–Darling test (as well as other commonly used normality tests) can only treat uncensored data, we generate a sample of 40 sources in an unbiased way with all sources detected by considering only the mini-BAL quasars observed by Chandra with an exposure longer than 2 ks. One mini-BAL quasar in G09, SDSS J113345.62+005813.4, which is undetected in a Chandra exposure of 4.3 ks, is removed from this sample because this source likely has both intervening and intrinsic X-ray absorption (Hall et al. 2006) while we are investigating intrinsic X-ray absorption only. The probability of non-normality of the Δα_ox values for this sample is only 3.6%. Therefore, the Δα_ox values of mini-BAL quasars follow the Gaussian distribution to a reasonable level of approximation. Strateva et al. (2005) reported that for their non-BMB quasar sample (with only a few percent BAL quasar contamination), a Gaussian profile could provide a reasonable fit to the distribution of the Δα_ox values. Gibson et al. (2008a) also found that normality of the Δα_ox distribution for their radio-quiet non-BMB quasar sample could not be ruled out by the Anderson–Darling test. For our sample, we find no evidence that any significant subset of sources falls outside the Gaussian regime. Thus, mini-BAL quasars are similar to non-BMB quasars with respect to the distribution of relative X-ray brightness. We also fit the Δα_ox distributions using the IDL gaussfit procedure (see Figure 6). For the 14 sources in our sample, the best-fit parameters are μ = 0.04 ± 0.01,  σ = 0.07 ± 0.01. For the 40 mini-BAL quasars observed by Chandra with an exposure longer than 2 ks, the fitting results are μ = 0.03 ± 0.02,  σ = 0.12 ± 0.02. The measurement errors for the Δα_ox values for most mini-BAL quasars are ≈0.03, which is much smaller than the intrinsic spread of Δα_ox. In fact, the square roots of the Δα_ox variances for the two groups of mini-BAL quasars tested above are 0.086 and 0.137, respectively, which indicates that the measurement errors have smaller influence than the spread of Δα_ox.

Zoom In Zoom Out Reset image size

4.3. Mini-BAL Quasar Reddening

BAL quasars are found to be redder than non-BMB quasars (e.g., Weymann et al. 1991; Brotherton et al. 2001; Reichard et al. 2003; Trump et al. 2006; Gibson et al. 2009b), which is likely caused by increased dust reddening. Following Gibson et al. (2009b), we investigate the reddening of mini-BAL quasars by measuring the quantity R_red, defined as,

$$\begin{equation} R_{\rm red} \equiv F_{\nu }(1400 \ {\rm \AA})/F_{\nu }(2500 \ {\rm \AA}), \end{equation} \tag{ 6 }$$

where F_ν(1400 Å) and F_ν(2500 Å) are the continuum flux densities at rest-frame 1400 Å and 2500 Å, respectively. These two wavelengths approximately represent the broadest available spectral coverage for most of our mini-BAL quasars. The continuum flux densities were calculated by fitting the SDSS spectra using the algorithm described in Section 2.1 of Gibson et al. (2008b).

Figure 7 shows the distribution of R_red values for the 61 mini-BAL quasars from this work and G09. The median value of R_red is 0.70. Gibson et al. (2009b) studied the R_red values of non-BMB and BAL quasars in the redshift range 1.96 ⩽ z ⩽ 2.28, and found median R_red values for non-BMB, HiBAL, and LoBAL quasars of 0.75, 0.64, and 0.30, respectively. The median R_red of mini-BAL quasars is intermediate between those of non-BMB and HiBAL quasars. We performed Kolmogorov–Smirnov tests on the distribution of R_red values for mini-BAL quasars versus those for non-BMB, HiBAL, and LoBAL quasars in the common redshift range of 1.96 ⩽ z ⩽ 2.28. The probability of the R_red values for 27 mini-BAL quasars and for 6689 non-BMB quasars coming from the same population is 0.100, while those probabilities are 0.053 and 3.05 × 10⁻⁸ for the mini-BAL quasars compared to 904 HiBAL quasars and 82 LoBAL quasars, respectively. These results cannot distinguish whether the R_red distribution for mini-BAL quasars is more similar to that for non-BMB quasars or for HiBAL quasars.

Zoom In Zoom Out Reset image size

The "redness" of the mini-BAL quasars may be caused by either steeper intrinsic power-law slopes or dust reddening. We calculated the relative SDSS colors Δ(u − r), Δ(g − i), and Δ(r − z), to assess curvature in the optical continuum, which can be seen as the evidence of dust reddening (e.g., Hall et al. 2006). The relative color is defined as the difference between the source color and the modal color for SDSS quasars in the corresponding redshift bin, calculated following the method in Section 5.2 of Schneider et al. (2007). For a pure power-law continuum, these three relative colors should be equal within uncertainties, while for dust-reddened objects, we would expect curvature in the continuum, specifically, Δ(u − r)>Δ(g − i)>Δ(r − z). For our mini-BAL quasar sample, the median values⁶ of the three relative colors are 0.086 ± 0.025, 0.074 ± 0.018, and 0.032 ± 0.015, respectively. Kolmogorov–Smirnov tests show that the probability for Δ(u − r) and Δ(r − z) following the same distribution is 7.5 × 10⁻³, while those probabilities for Δ(u − r) versus Δ(g − i) and Δ(g − i) versus Δ(r − z) are 0.35 and 0.068, respectively. Therefore, we can see a moderate trend that Δ(u − r)>Δ(g − i)>Δ(r − z), indicating mild dust reddening in mini-BAL quasar spectra.

Assuming all quasars have a similar underlying continuum shape on average, different R_red values represent different dust-reddening levels. According to an SMC-like dust extinction curve (Pei 1992), the median R_red for mini-BAL quasars corresponds to E(B − V) ≈ 0.011, which is lower than the E(B − V) values of HiBAL quasars (≈0.023) and LoBAL quasars (≈0.14) in G09. Since the relative X-ray brightness Δα_ox largely represents the absorption in X-ray band, we tested for correlations between R_red and Δα_ox for the mini-BAL quasars using the Spearman rank-order correlation analysis. This analysis returns a correlation coefficient ρ and a corresponding probability P_S for the null hypothesis (no correlation). ρ = 0 indicates no correlation between the two variables tested, while ρ = ±1 means a perfect correlation. The correlation probability is 1 − P_S, e.g., P_S = 0.01 means a correlation probability of 99%. We find a marginal correlation between R_red and Δα_ox for the 61 mini-BAL quasars in the combined sample at an 86.2% probability (ρ = 0.190).

4.4. UV Absorption Versus X-ray Brightness

We define a high signal-to-noise ratio (S/N) sample of BAL and mini-BAL quasars to study correlations between Δα_ox and AI, v_max, v_min, Δv (≡|v_max − v_min|) (see Section 2 for the definitions of these quantities). Radio-quiet BAL and mini-BAL quasars in G09 with S/N₁₇₀₀ > 9, and the 14 mini-BAL quasars analyzed in this paper (which also have S/N₁₇₀₀ > 9), are included in the high S/N sample. S/N₁₇₀₀ is defined as the median of the flux divided by the noise provided by the SDSS pipeline for spectral bins between rest-frame 1650 Å and 1700 Å. Excluding low S/N sources enables the most-reliable measurements for the UV absorption parameters listed above; Gibson et al. (2009b) reported that high S/N spectra enabled more accurate identifications of broad absorption features. The values of the AI, v_max, and v_min parameters for the sources in G09 are updated according to the improved redshift measurements in Hewett & Wild (2010). There are 23 BAL quasars (four of them have upper limits on Δα_ox) and 33 mini-BAL quasars (two of them have upper limits on Δα_ox) in the high S/N sample.

Figure 8 shows the SDSS spectra for the 33 mini-BAL quasars in the high S/N sample in order of Δα_ox. The positions of mini-BAL troughs are shown in the figure. Some mini-BAL troughs show doublet-like features. However, for mini-BAL quasars only, no significant trends are apparent between Δα_ox and the positions or morphologies of mini-BAL troughs.

Zoom In Zoom Out Reset image size

In Figure 9, we plot the UV absorption properties against Δα_ox for the high S/N mini-BAL and BAL quasar sample. The correlation probabilities between Δα_ox and AI, v_max, v_min, Δv are determined using Spearman rank-order correlation analysis (see Table 6). Our correlations are consistent with those found in G09. Our enhanced correlation between AI and Δα_ox confirms the trend found in G09 that UV broad absorption strength decreases with the relative X-ray brightness. As in G09, v_max is correlated with Δα_ox, while there is no significant correlation between v_min and Δα_ox. The Δv–Δα_ox correlation shows that mini-BAL quasars with narrower absorption troughs are relatively X-ray brighter than BAL quasars. Some mini-BAL quasars have large Δv values because they have multiple absorption troughs.

Zoom In Zoom Out Reset image size

Table 6. Spearman Rank-order Correlation Coefficients and Probabilities

Correlation			G09 Sample		G09 Sample		G09 mini-BAL only
	G09 Samplea		+ Our Sample		+ Our Sample		+ Our Sample
	(1000 km s−1)b		(1000 km s−1)b		(500 km s−1)b		(500 km s−1)b
	ρ	1 − PS	ρ	1 − PS	ρ	1 − PS	ρ	1 − PS
AI vs. Δ αox	−0.474	99.76%	−0.519	99.99%	−0.599	>99.99%	−0.434	98.59%
vmax vs. Δ αox	−0.442	99.54%	−0.380	99.52%	−0.409	99.76%	−0.402	97.69%
vmin vs. Δ αox	−0.014	7.00%	−0.055	32.74%	−0.189	83.92%	−0.252	84.56%
Δv vs. Δ αox	−0.620	99.99%	−0.628	>99.99%	−0.463	99.94%	−0.422	98.30%
vwt vs. Δ αox	−0.191	78.99%	−0.205	87.24%	−0.252	93.85%	−0.260	85.87%

Notes. ^aThe correlation analysis for G09 sample sources only is repeated with the Spearman rank-order correlation analysis, using updated AI, v_max, v_min, and Δv values according to the improved redshift measurements in Hewett & Wild (2010). ^bVelocity width limit used when defining mini-BALs.

Download table as:  ASCIITypeset image

We have calculated the weighted average velocity, v_wt, which is defined in Trump et al. (2006) using the following equation,

$$\begin{equation} v_{\rm wt}=\frac{\int _{0}^{29,000}(1-f(v))\ v\ C^{\prime }dv}{{\mbox{AI}}}, \end{equation} \tag{ 7 }$$

where f(v) and C' have the same definition as in Equation (2). The values of v_wt for the 14 mini-BAL quasars in our sample are listed in Table 2 and also shown in Figure 8. The v_wt parameter represents the weighted average position of mini-BAL and BAL troughs. However, for some mini-BAL quasars with multiple absorption troughs, v_wt may fall between troughs where no absorption features are present. v_wt is marginally correlated with Δα_ox at a probability of 87.9% for the high S/N sample of mini-BAL and BAL quasars (see Figure 9(e)).

In the disk-wind model for BAL and mini-BAL quasars (e.g., Murray et al. 1995), the UV absorbing outflow is radiatively accelerated, while the material interior to the UV absorber shields the soft X-ray radiation to prevent overionization. This model may be supported by the correlations between UV absorption properties and relative X-ray weakness (see Table 6). As in G09, we also find some high velocity, relatively X-ray bright mini-BAL quasars in our newly analyzed 14 sources. These sources can perhaps be fit into the "failed" BAL quasar scenario discussed in Section 4.2 of G09, in which the high incident X-ray flux overionizes the outflow material so that BAL troughs cannot form along the line of sight.

4.5. Alternative Measures of UV Absorption

4.5.1. Effects of the Velocity Width Limit

From the rest-frame UV spectra of mini-BAL quasars (e.g., Figure 2), we can see that some likely intrinsic absorption features are missed by the definition of mini-BALs using the parameter AI (also see the "Notes" column of Table 2). When calculating AI, only absorption troughs with velocity widths greater than 1000 km s⁻¹ are included. This somewhat conservative velocity width limit was adopted because we wanted to spend valuable Chandra observing time only on sources with bona fide intrinsic absorption features. Here we now slightly modify the definition of AI by lowering the velocity width limit to 500 km s⁻¹ to form a new parameter AI₅₀₀. This velocity width limit is similar to the value in the original introduction of AI (450 km s⁻¹) in Hall et al. (2002). The limit of 500 km s⁻¹ can include more broad absorption features while still minimizing the inclusion of, e.g., complex intervening systems (see the discussion in Hall et al. 2002). The equation for calculating AI₅₀₀ is the same as that of AI,

$$\begin{equation} \mbox{AI}_{500}=\int _{0}^{29,000}(1-f(v))\ C^*dv, \end{equation} \tag{ 8 }$$

while C* here has a different definition from C' in Equation (2): C* = 1 when the velocity width is at least 500 km s⁻¹ and the absorption trough falls at least 10% below the continuum; C* = 0 otherwise. Five of the 14 mini-BAL quasars in our sample have new absorption troughs included using this new velocity width limit (see the blue spectra in Figure 8). The AI₅₀₀, as well as v_max, v_min, Δv, and v_wt also changes accordingly (see Table 7).

Table 7. C iv Mini-BAL Properties for Sources with New Absorption Troughs Included under the Definition of AI₅₀₀

Object Name (SDSS J)	AI500	vmin	vmax	vwt	No. of Troughs	Notesa
	(km s−1)	(km s−1)	(km s−1)	(km s−1)	Contributing to AI
Chandra cycle 10 objects
091342.48+372603.3	1136.8	3179	23014	10715	4	Feature at ≈1535 Å included
						Feature at ≈1480 Å included
105904.68+121024.0	2503.5	5378	23349	11717	6	Feature at ≈1470 Å included
151451.77+311654.0	692.7	24124	27003	25592	2	Feature at ≈1430 Å included
Archival XMM-Newton objects
142656.17+602550.8	478.3	4937	28324	21746	2	Feature at ≈1525 Å included
160222.72+084538.4	517.1	2136	15239	10006	2	Feature at ≈1540 Å included

Note. ^aThis column describes the newly included absorption troughs using the velocity width limit of 500 km s⁻¹. As discussed in the "Notes" column of Table 2, some of them failed the width criterion when the velocity width limit was taken as 1000 km s⁻¹.

Download table as:  ASCIITypeset image

New absorption troughs are also identified for some BAL and mini-BAL quasars in G09 using the AI₅₀₀ definition (e.g., see Figure 8 for mini-BAL quasars). From Figure 8, we can see that more relatively X-ray weak sources (with more negative Δα_ox) have new absorption troughs (indicated by the red horizontal bars) included compared to relatively X-ray bright sources. We repeated the correlation analysis between Δα_ox and UV absorption properties for the high S/N BAL and mini-BAL quasar sample; the results are listed in Table 6. The UV absorption properties under the new definition of AI are plotted against Δα_ox in Figure 10. Red data points show the quantities changed under the new definition. All UV absorption parameters, except Δv, have better correlations with Δα_ox, while the correlation probability between Δα_ox and Δv is still consistent with the previous result. v_min now has a marginal correlation (83.9%) with Δα_ox. If we only consider mini-BAL quasars in the high S/N sample (i.e., leaving out BAL quasars), these correlations still exist, although they are somewhat weaker (see the last two columns of Table 6). However, under the velocity width limit of 1000 km s⁻¹, the correlation between Δα_ox and AI disappears (1 − P_S = 60.0%) for mini-BAL quasars only, while the correlation between Δα_ox and v_max becomes much weaker (1 − P_S = 91.4%). We consider 500 km s⁻¹ to be a better velocity width limit than the previously used value (1000 km s⁻¹), in the sense that it can reveal somewhat clearer correlations between X-ray and UV absorption properties. Figure 8 shows that most of the newly included absorption troughs have similar profiles and velocities to those troughs with >1000 km s⁻¹ velocity widths. A Kolmogorov–Smirnov test does not distinguish the distributions of central velocities for the troughs with >1000 km s⁻¹ velocity widths and those with 500–1000 km s⁻¹ velocity widths (P = 0.46). Therefore, these newly included troughs are not likely to be caused by intervening systems.

Zoom In Zoom Out Reset image size

4.5.2. Investigating New UV Absorption Parameters

The Δv parameter measures the velocity span of the UV absorbing outflow for sources with a single mini-BAL or BAL trough. However, for sources with multiple absorption troughs, Δv also includes regions where no absorption features are present. We therefore define a new parameter, called "total velocity span (v_tot)," for the mini-BAL and BAL quasars. v_tot is defined by the following equation,

$$\begin{equation} v_{\rm tot}=\int _{0}^{29,000} C^*dv, \end{equation} \tag{ 9 }$$

where C* is the same as that in the definition of AI₅₀₀. The values of v_tot for the 14 mini-BAL quasars in our sample are listed in Table 8. The v_tot parameter represents the total velocity width of all absorption troughs up to 29, 000 km s⁻¹. v_tot is correlated with Δα_ox at a probability of >99.99% for the high S/N sample of BAL and mini-BAL quasars (see Figure 11(a)). This strong correlation also stands for mini-BAL quasars only (1 − P_s = 99.77%).

Zoom In Zoom Out Reset image size

Table 8. The Values of the v_tot, v_wt,st, and K Parameters for the 14 Mini-BAL Quasars

Object Name (SDSS J)	vtot	vwt,st	K
	(km s−1)	(km s−1)	(109 km3 s−3)
Chandra cycle 10 objects
080117.79+521034.5	1095.8	26641	4331.9
091342.48+372603.3	4958.8	11695	592.7
092914.49+282529.1	1165.7	24483	4183.3
093207.46+365745.5	1449.4	3871	21.9
105158.74+401736.7	1104.4	2383	3.1
105904.68+121024.0	8811.7	6435	813.3
120331.29+152254.7	1098.5	22118	3067.5
123011.84+401442.9	1242.5	2326	5.9
125230.84+142609.2	1449.5	2914	9.4
141028.14+135950.2	1725.7	2394	1.6
151451.77+311654.0	2329.8	26188	5002.6
224649.29-004954.3	1449.3	4858	53.6
Archival XMM-Newton objects
142656.17+602550.8	1716.1	27811	4475.0
160222.72+084538.4	2068.1	14689	487.2

Download table as:  ASCIITypeset image

As mentioned in Section 4.4, the weighted average velocity, v_wt, may represent regions without absorption features for spectra with multiple troughs. Therefore, we also calculated the weighted average velocity for the strongest absorption trough v_wt,st using the following equation similar to the definition of v_wt,

$$\begin{equation} v_{\rm wt,st}=\frac{\int _{v_{\rm min,st}}^{v_{\rm max, st}}(1-f(v))\ v\ dv}{{\rm AI}_{\rm st}}, \end{equation} \tag{ 10 }$$

where v_min,st and v_max,st are the minimum and maximum velocity for the strongest absorption trough, which has the largest AI value (denoted by AI_st) among those multiple troughs. The v_wt,st values for the 14 mini-BAL quasars in our sample are listed in Table 8. v_wt,st has a better correlation (98.9%) with Δα_ox than v_wt (see Figure 11(b)).

The kinetic luminosity of an AGN outflow, $L_{\rm kin}\ (=\dot{M}v^2/2)$, is proportional to N_Hv³, where $\dot{M}$ is the mass flux of the outflow and N_H is its column density (e.g., see Section 11.3 of Krolik 1999). We define another new parameter,

$$\begin{equation} K=\frac{\int _{0}^{29,000}(1-f(v))\ v^3\ C^*dv}{v_{\rm tot}}. \end{equation} \tag{ 11 }$$

K is a product of absorption strength, which is denoted by the factor 1 − f(v), and v₃, averaged over the velocity parameter space. It is heuristically motivated by kinetic luminosity, but it is not a proper measure of this quantity since, e.g., absorption strength is usually not proportional to N_H (cf. Hamann 1998; Arav et al. 1999). For the high S/N sample, the K parameter is correlated with Δα_ox at a probability of 97.1% (see Figure 11(b)). The values of K for the 14 mini-BAL quasars in our sample are also listed in Table 8.

For sources with a single broad absorption trough, v_tot and v_wt,st have the same values as Δv and v_wt, respectively. For sources with multiple troughs, v_tot and v_wt,st have the advantage that they represent absorption features only. Furthermore, v_tot and v_wt,st have better correlations with the relative X-ray brightness. Although v_tot and Δv have similar high correlation probabilities with Δα_ox (>99.99% versus 99.94%), Figure 11(a) versus Figure 10(d) shows that the correlation between v_tot and Δα_ox is stronger, which is also indicated by the Spearman rank-order correlation coefficients (ρ = −0.682 versus ρ = −0.463). We consider v_tot and v_wt,st to be more physically revealing parameters. The other newly introduced parameter K does not have a better correlation with Δα_ox than the simpler parameter AI does. Further investigations may find additional useful parameters which could better represent the physical properties of UV absorbing outflows.

4.6. Investigation of a Very High Significance Mini-BAL Sample

We present the X-ray properties of 14 of the optically brightest mini-BAL quasars from SDSS DR5; 12 have been observed in a new Chandra snapshot survey in Cycle 10 and 2 objects have archival XMM-Newton observations. All 14 sources are detected. We study correlations between the UV absorption properties and the X-ray brightness for a sample consisting of these 14 sources, as well as the BAL and mini-BAL quasars in G09. Our main results are summarized as follows.

• 1.  
  The mean X-ray power-law photon index of the 12 Chandra observed sources is Γ = 1.90^+0.11_−0.11 for a model without intrinsic absorption. Adding an intrinsic neutral absorption component only slightly changes the mean photon index to Γ = 1.91^+0.18_−0.11, consistent with the absence of evidence for strong intrinsic neutral absorption on average (N_H ≲ 8.2 × 10²¹ cm⁻²). This indicates that mini-BAL quasars generally have similar X-ray spectral properties to non-BMB quasars.
• 2.  
  The relative X-ray brightness, assessed with the Δα_ox parameter, of the mini-BAL quasars has a closer mean value and distribution to those of non-BMB quasars than to those of BAL quasars. An Anderson–Darling test finds no non-normality of the Δα_ox distribution of mini-BAL quasars.
• 3.  
  The reddening, assessed with the parameter R_red, of the mini-BAL quasars is intermediate between those of non-BMB and HiBAL quasars. Relative colors show curvature in the optical spectra of mini-BAL quasars, indicating mild dust reddening.
• 4.  
  Significant correlations are found between Δα_ox and AI, v_max and Δv. The weighted average velocity v_wt has a marginal correlation with Δα_ox. These correlations may support the radiatively driven disk-wind model where X-ray absorption is important in enabling UV line-absorbing winds.
• 5.  
  We find that more intrinsic broad absorption troughs are included if the velocity width limit in the definition of mini-BALs is lowered to 500 km s⁻¹. The UV absorption parameters AI, v_max, v_min, Δv, and v_wt have clearer correlations with the relative X-ray brightness under the new velocity width limit. We also propose three new parameters, v_tot, v_wt,st, and K, defined by Equations (9), (10), and (11), respectively. v_tot is the total velocity span of all the UV broad absorption troughs, while v_wt,st is the weighted average velocity for the strongest absorption trough. We consider v_tot and v_wt,st to better represent the physical properties of UV absorption features than Δv and v_wt. K is similar to, but not a proper measure of, the kinetic luminosity of the absorbing outflows.
• 6.  
  Testing shows that the complex factors in the selection of mini-BAL quasars, such as the placement of continua, do not significantly affect our main results on the relative X-ray brightness and its correlations with UV absorption properties.

Further accumulation of high-quality X-ray and UV/optical data will allow studies of even larger samples of mini-BAL and BAL quasars. The improved source statistics should enable the construction and testing of further measures of mini-BAL and BAL absorption that may provide additional insight in the exploration of the broad absorption region. Ultimately, high-resolution X-ray spectroscopy performed by future missions, such as IXO, should reveal the detailed physical and kinematic properties of the X-ray absorber. This will open a new dimension in studies of X-ray versus UV/optical absorption for quasar outflows.

We thank the anonymous referee for constructive comments. We gratefully acknowledge the financial support of NASA grant SAO SV4-74018 (G.P.G., Principal Investigator), NASA ADP grant NNX10AC99G (J.W., W.N.B., M.L.C.), NASA Chandra grant AR9-0015X (R.R.G.), and NSF grant AST07-09394 (R.R.G.). We thank Y. Q. Xue for helpful discussions.

Funding for the SDSS and SDSS-II has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the U.S. Department of Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, the Max Planck Society, and the Higher Education Funding Council for England. The SDSS Web site is http://www.sdss.org/.