Does the X-ray outflow quasar PDS 456 have a UV outflow at 0.3c?

Does the X-ray outflow quasar PDS 456 have a UV outflow at 0.3c? Abstract The quasar PDS 456 (at redshift ∼0.184) has a prototype ultra-fast outflow (UFO) measured in X-rays. This outflow is highly ionized with relativistic speeds, large total column densities log NH(cm−2) > 23, and large kinetic energies that could be important for feedback to the host galaxy. A UV spectrum of PDS 456 obtained with the Hubble Space Telescope in 2000 contains one well-measured broad absorption line (BAL) at ∼1346 Å (observed) that might be Ly α at v ≈ 0.06c or N v λ1240 at v ≈ 0.08c. However, we use photoionization models and comparisons to other outflow quasars to show that these BAL identifications are problematic because other lines that should accompany them are not detected. We argue that the UV BAL is probably C iv at v ≈ 0.30c. This would be the fastest UV outflow ever reported, but its speed is similar to the X-ray outflow and its appearance overall is similar to relativistic UV BALs observed in other quasars. The C iv BAL identification is also supported indirectly by the tentative detection of another broad C iv line at v ≈ 0.19c. The high speeds suggest that the UV outflow originates with the X-ray UFO crudely 20–30 rg from the central black hole. We speculate that the C iv BAL might form in dense clumps embedded in the X-ray UFO, requiring density enhancements of only ≳0.4 dex compared to clumpy structures already inferred for the soft X-ray absorber in PDS 456. The C iv BAL might therefore be the first detection of low-ionization clumps proposed previously to boost the opacities in UFOs for radiative driving. line: formation, quasars: absorption lines, quasars: general, quasars: individual: PDS 456 1 INTRODUCTION Accretion-disc outflows are an important part of the quasar phenomenon that might drive ‘feedback’ to regulate black hole growth and host galaxy evolution (e.g. Di Matteo, Springel & Hernquist 2005; Hopkins et al. 2008; Hopkins & Elvis 2010; Debuhr, Quataert & Ma 2012; Rupke & Veilleux 2013). The outflows are often studied in the rest-frame UV via blueshifted broad absorption lines (BALs) or their narrower cousins called ‘mini-BALs’ (with a nominal boundary near full width at half-minimum FWHM ∼2000 km s−1; Weymann et al. 1991; Korista et al. 1993; Crenshaw, Kraemer & George 2003; Hamann & Sabra 2004; Trump et al. 2006; Knigge et al. 2008; Gibson et al. 2009, and references therein). These features appear most often at moderate velocity shifts v < 0.1c, but relativistic BALs and mini-BALs at v ∼ 0.1–0.2c have been measured in a small but growing number of quasars (Hamann et al. 1997a, 2013; Rodríguez Hidalgo 2008; Rodríguez Hidalgo, Hamann & Hall 2011; Rogerson et al. 2016). The basic physical properties for these outflows can be difficult to determine due to limited wavelength coverage and line saturation that is masked by partial line-of-sight covering of the background light source(s). The strongest measured UV lines1 are typically C iv λ1549, N v λ1240, and O vi 1034, indicating moderate to high degrees of ionization. The maximum ionizations are not known, but there might be a wide range. O vi absorption tends to be as strong as or stronger than C iv (Section 3.4 below, also Baskin, Laor & Hamann 2013; Moravec et al. 2017; Herbst et al., in preparation) and higher ionization lines such as Ne viii λ774 and Mg x λ615 have been observed in a few cases with suitable spectral coverage (e.g. Hamann et al. 1997b; Telfer et al. 1998; Arav et al. 2001). Measurements of the low-abundance line P v λ1121 help to overcome the saturation issues to reveal generally large total hydrogen column densities, log NH(cm−2) ≳ 22.3, across a wide range observed BAL strengths (Hamann 1998; Leighly et al. 2009; Leighly, Dietrich & Barber 2011; Borguet et al. 2012; Capellupo, Hamann & Barlow 2014; Capellupo et al. 2017; Moravec et al. 2017, but see also Arav et al. 2001). X-ray observations have revealed another variety of ultra-fast outflows (UFOs) that reach relativistic speeds in luminous quasars (Chartas et al. 2002, 2009; Reeves et al. 2009; Tombesi et al. 2010, 2013; Gofford et al. 2013, 2015, and references therein). UFOs are also challenging to study because they are highly variable and highly ionized to the point where the only strong absorption features appear at X-ray wavelengths. They are often characterized by Fe K-shell absorption features with derived total column densities in the range log NH(cm−2) ∼ 22–24 (see references above). UFOs also appear to have generally very large kinetic energies, sufficient to drive feedback effects in the quasar host galaxies (Tombesi et al. 2012, 2013; Gofford et al. 2014, 2015; Reeves et al. 2014). The quasar PDS 456 (at redshift ze ≈ 0.184) has the best-studied example of a powerful, relativistic X-ray UFO (e.g. Reeves, O'Brien & Ward 2003; Reeves et al. 2009, 2014, 2016; Gofford et al. 2014; Nardini et al. 2015; Matzeu et al. 2017). It is the most luminous quasar in the local Universe, with bolometric luminosity L ∼ 1047 erg s−1 and estimated black hole mass ∼(1–2) × 109 M⊙ that together indicate an accretion rate relative to Eddington that is L/LEdd ≳ 0.3 and perhaps near unity (Nardini et al. 2015). The X-ray absorber is complex and highly variable, with two main components. The main component measured via Fe K-shell absorption has speeds in the range v ∼ 0.25–0.34c, very high degrees of ionization featuring Fe xxv and Fe xxvi, and large total column densities log NH(cm−2) ≳ 23 (Gofford et al. 2014, 2015; Reeves et al. 2014; Nardini et al. 2015). Its radial distance from the black hole is estimated at a few hundred gravitational radii (based on absorber variability; e.g. Nardini et al. 2015). The kinetic power of this outflow is also remarkably large, crudely ∼20 per cent of the bolometric luminosity, which is well above the threshold needed for feedback to the host galaxy (Gofford et al. 2014; Nardini et al. 2015, and references therein). The second outflow component measured in soft X-rays has somewhat lower speeds, v ∼ 0.17–0.27c, lower ionizations, and column densities in the range log NH(cm−2) ∼ 22–23 (Reeves et al. 2016). There is evidence for time-variable covering fractions in this absorber that might be indicative of small dense clumps or substructures embedded in the overall X-ray outflow (Matzeu et al. 2016). PDS 456 also has signatures of outflow in the UV. A UV spectrum obtained in 2000 by O'Brien et al. (2005) using the Space Telescope Imaging Spectrometer (STIS) on board the Hubble Space Telescope (HST) revealed a highly blueshifted C iv broad emission line at speeds near v ∼ 5000 km s−1 plus a single BAL that might plausibly be Ly α or N v at v ∼ 0.06c or ∼0.08c, respectively. These UV features provide further evidence for exotic mass-loss from PDS 456 across a wide range of spatial scales, from the X-ray UFO that is believed to originate near the black hole at radii of ∼20–30 rg (gravitational radii, corresponding to ∼0.001–0.0015 pc; Nardini et al. 2015; Matzeu et al. 2017) to an outflow-dominated C iv broad emission-line region that we place crudely at ∼0.3 pc [based on scaling relations with luminosity (Kaspi et al. 2005) for a 109 M⊙ black hole]. The UV BAL is an important component to this outflow picture, but its measured speed and physical nature depend critically on the line identification. In this paper, we re-examine the UV BAL in PDS 456 with the main result that it is likely to be C iv λ1549 at v ≈ 0.30c. This would be the fastest UV outflow line ever reported but similar in speed to the X-ray UFO in this quasar. Throughout this paper, we adopt a redshift for PDS 456 of ze = 0.183 75 based on the emission line [Fe ii] 1.6435 μm measured by Simpson et al. (1999). We describe two archival HST spectra of PDS 456 in Section 2 below. Section 3 presents our analysis and measurements of the observed BAL, comparisons to photoionization models, comparisons to UV BALs in other quasars, and a discussion of the plausible BAL identifications. Section 4 presents a summary and discussion of the results. 2 HST SPECTRA Fig. 1 shows the spectrum of PDS 456 obtained by O'Brien et al. (2005) in 2000 using the Hubble Space Telescopea Imaging Spectrometer (STIS) with the G140L and G230L gratings. We obtained this spectrum and another HST spectrum measured in 2014 (described below) from the Mikulski Archive for Space Telescopes. We present them here without further processing. The STIS G140L grating provided wavelength coverage from 1137 to 1715 Å at resolutions ranging from ∼310 to ∼200 km s−1, while STIS G230L covered 1580–3148 Å at resolutions from ∼640 to ∼300 km s−1 (see O'Brien et al. 2005 for more details). The spectrum plotted in Fig. 1 is combined from the two gratings by masking out the extreme ends of the wavelength coverage to avoid excessive noise and then calculating variance-weighted average fluxes at the remaining wavelengths of overlap. Figure 1. View largeDownload slide HST-STIS spectrum of PDS 456 from 2000 plotted as observed (bottom panel) and reddening-corrected using a Galactic extinction curve with E(B − V) = 0.45 (top panel). The grey dotted curves are the corresponding error spectra. The red dashed curves are an approximate fit to the continuum using a power law with index αλ = −1.68 shown with (bottom panel) and without (top) Galactic reddening. The red dotted curves in the bottom panel show the same power law reddened alternatively by E(B − V) = 0.42 and 0.48. The BAL we attribute to C iv at v ≈ 0.30c is labelled above the spectrum at 1346 Å. Some broad emission-line wavelengths are marked across the top. Galactic absorption lines (including strong damped Ly α at 1216 Å) are marked by green vertical lines near the bottom. See Section 2. Figure 1. View largeDownload slide HST-STIS spectrum of PDS 456 from 2000 plotted as observed (bottom panel) and reddening-corrected using a Galactic extinction curve with E(B − V) = 0.45 (top panel). The grey dotted curves are the corresponding error spectra. The red dashed curves are an approximate fit to the continuum using a power law with index αλ = −1.68 shown with (bottom panel) and without (top) Galactic reddening. The red dotted curves in the bottom panel show the same power law reddened alternatively by E(B − V) = 0.42 and 0.48. The BAL we attribute to C iv at v ≈ 0.30c is labelled above the spectrum at 1346 Å. Some broad emission-line wavelengths are marked across the top. Galactic absorption lines (including strong damped Ly α at 1216 Å) are marked by green vertical lines near the bottom. See Section 2. The UV spectrum of PDS 456 is substantially reddened by dust in our Galaxy (due to the quasar's sky position near the Galactic plane). Previous studies based on visible spectra and photometry indicate that the visual extinction towards PDS 456 is AV ∼ 1.5 mag, corresponding to selective extinction E(B − V) ∼ 0.48 for a standard Galactic reddening curve (Torres et al. 1997; Simpson et al. 1999). We obtain an independent estimate of the reddening by fitting the UV continuum with a single power law modified by the Galactic extinction curve from Cardelli, Clayton & Mathis (1989) with RV = 3.1. The fit is constrained by the median flux in wavelength windows between ∼1170 and ∼2400 Å that avoid strong emission and absorption lines. The dashed red curve in Fig. 1 shows our preferred fit using E(B − V) = 0.45 and power-law index αλ = −1.68 (for $$F_{\lambda }\propto \lambda ^{\alpha _{\lambda }}$$). An important feature of Galactic extinction at these wavelengths is the ‘bump’ at ∼2175 Å. We experimented with different fit parameters (e.g. the dotted red curves in Fig. 1), but the values of E(B − V) and the power-law slope are well constrained2 by the data because Galactic reddening curves with a strong 2175 Å bump suppress the flux at both ends of the spectral coverage shown in Fig. 1. The top panel in this figure shows the reddening-corrected spectrum of PDS 456 together with the best-fitting power law. Fig. 2 shows a previously unpublished spectrum of PDS 456 obtained in 2014 with HST Cosmic Origins Spectrograph (COS) using the G140L grating (PI: O'Brien). This spectrum (shown by the red curve) is plotted on top of the HST-STIS spectrum (black curve) from Fig. 1 with no reddening corrections. It is combined from separate exposures totalling 5186 s. The spectral resolution ranges from roughly 200 to 100 km s−1 from blue to red across the wavelengths shown in Fig. 2. A gap in the COS wavelength coverage from ∼1170 to ∼1265 Å avoids Galactic/geocoronal Ly α absorption/emission. The blue arrows in Figs 1 and 2 mark the positions of observed or expected absorption lines discussed in Section 3 below. Figure 2. View largeDownload slide HST spectra of PDS 456 obtained with STIS in 2000 (black curve) and COS in 2014 (red) plotted at observed wavelengths, uncorrected for reddening. The red and grey dotted curves are the corresponding error spectra (1σ uncertainties per pixel). The STIS spectrum is scaled vertically by a factor of ∼0.60 to match the COS spectrum approximately in the continuum. The dashed blue curve shows our fit to the STIS spectrum from Fig. 1. The BAL we attribute to C iv λ1549 at 0.30c is labelled below the spectrum at ∼1346 Å. Also marked are the expected positions of O vi λ1034 and C iv BALs if the observed BAL is attributed, instead, to N v λ1240 at 0.08c or Ly α at 0.06c (e.g. ‘if N v, Ly α’). We tentatively identify another weak BAL at ∼1513 Å as C iv at 0.19c. Dashed blue vertical lines mark the positions of Galactic absorption features. The narrow emission spike at 1304 Å is geocoronal O i. See Fig. 1 and Section 2 for additional notes. Figure 2. View largeDownload slide HST spectra of PDS 456 obtained with STIS in 2000 (black curve) and COS in 2014 (red) plotted at observed wavelengths, uncorrected for reddening. The red and grey dotted curves are the corresponding error spectra (1σ uncertainties per pixel). The STIS spectrum is scaled vertically by a factor of ∼0.60 to match the COS spectrum approximately in the continuum. The dashed blue curve shows our fit to the STIS spectrum from Fig. 1. The BAL we attribute to C iv λ1549 at 0.30c is labelled below the spectrum at ∼1346 Å. Also marked are the expected positions of O vi λ1034 and C iv BALs if the observed BAL is attributed, instead, to N v λ1240 at 0.08c or Ly α at 0.06c (e.g. ‘if N v, Ly α’). We tentatively identify another weak BAL at ∼1513 Å as C iv at 0.19c. Dashed blue vertical lines mark the positions of Galactic absorption features. The narrow emission spike at 1304 Å is geocoronal O i. See Fig. 1 and Section 2 for additional notes. 3 ANALYSIS 3.1 The UV BAL at 1346 Å The HST-STIS spectrum in Fig. 1 clearly shows a BAL at ∼1346 Å (observed), as reported by O'Brien et al. (2005). Fig. 3 plots this spectrum again after normalizing by the reddened power law in Figs 1 and 2. We fit the BAL in this normalized spectrum using a simple Gaussian optical depth profile of the form   \begin{eqnarray} \tau _{\rm v} \ = \ \tau _{\rm o}\, {\rm e}^{-{v}^2/b^2}, \end{eqnarray} (1)where τo is the line-centre optical depth, v is the velocity shift from the line centre, and b is the Doppler parameter that sets the line width. The BAL fit shown by the magenta curve in Fig. 3 yields τo = 0.35 ± 0.01, b = 5135 ± 203 km s−1, and observed line-centre wavelength λo = 1345.7 ± 0.6 Å (where the errors quoted are 1σ uncertainties returned by the line-fitting routine). The full width at half-minimum of this fitted profile is FWHM ≈ 8550 ± 340 km s−1. In the quasar frame, the fit has rest equivalent width REW = 10.8 ± 0.6 Å. Figure 3. View largeDownload slide HST-STIS spectrum of PDS 456 in the observed frame, normalized by the continuum fit shown in Fig. 1. The smooth magenta and orange curves show our fit to the observed BAL at 1346 Å and then that fit transposed to the wavelengths of O vi and C iv assuming that the measured BAL is Ly α (magenta) or N v (orange). The weaker profiles drawn at the C iv positions show the transposed fit at 1/3 its measured strength to illustrate approximate upper limits to the actual absorption there. The absence of absorption at the transposed fit positions argues against the Ly α and N v identifications for the observed BAL (Sections 3.5 and 3.6). Other labels are the same as Figs 1 and 2. Narrow Galactic absorption lines of O i λ1304 and C ii λ1335 have been removed by interpolation for clarity. Figure 3. View largeDownload slide HST-STIS spectrum of PDS 456 in the observed frame, normalized by the continuum fit shown in Fig. 1. The smooth magenta and orange curves show our fit to the observed BAL at 1346 Å and then that fit transposed to the wavelengths of O vi and C iv assuming that the measured BAL is Ly α (magenta) or N v (orange). The weaker profiles drawn at the C iv positions show the transposed fit at 1/3 its measured strength to illustrate approximate upper limits to the actual absorption there. The absence of absorption at the transposed fit positions argues against the Ly α and N v identifications for the observed BAL (Sections 3.5 and 3.6). Other labels are the same as Figs 1 and 2. Narrow Galactic absorption lines of O i λ1304 and C ii λ1335 have been removed by interpolation for clarity. Our study concerns identification of the ∼1346 Å BAL trough. One possibility favoured by O'Brien et al. (2005) is Ly α 1216 at velocity shift v ≈ 0.06c (18 000 km s−1). Another is N v λ1240 at v ≈ 0.08c. These identifications seem reasonable because the observed BAL is on the blue side of the Ly α–N v emission-line blend and the inferred velocities are in a normal range for BAL outflows in other quasars. However, we will argue below (Sections 3.5 and 3.6) that the Ly α or N v identifications are problematic because they are not accompanied by C iv and O vi λ1034 BALs at the same velocity shifts. The absence of these other lines is illustrated by the other orange and magenta curves in Fig. 3, which show our fit to the observed BAL transposed to the positions expected for accompanying C iv and O vi BALs. At the C iv positions, the transposed BAL fit is shown twice – once at full strength and once at 1/3 of its measured strength at 1346 Å. We estimate that the 1/3 scalings represent approximate upper limits to C iv BALs that might accompany an N v or Ly α BAL at 1346 Å. We will argue below that the absence of O vi and C iv absorption at these wavelengths suggests that the BAL at ∼1346 Å is actually C iv λ1549 at v ≈ 0.30c. We test these line identifications using photoionization models (Section 3.3) and comparisons to BALs/mini-BALs observed in other quasars (Section 3.4). First, we estimate ionic column densities for each possible identification. These estimates follow from the fitted Gaussian optical depth profile by   \begin{eqnarray} N_{\rm i} \ = \ {{m_{\rm e}\, c}\over {\sqrt{\pi }\, e^2}} {{b\,\tau _{\rm o} }\over {f\lambda _{\rm r}}}, \end{eqnarray} (2)where Ni is the ionic column density, f and λr are the oscillator strength and rest-frame wavelength of the transition, and we assume implicitly that the ground-state column density equals the ionic column density. Thus, we find that, if the observed BAL is Ly α, it corresponds to a neutral hydrogen column density of log NHI(cm−2) ≈ 15.35. If the BAL is N v or C iv, then the column density is $$\log N(\rm{N\,{\small V}}) ({\rm cm}^{-2}) \approx 15.62$$ or $$\log N(\rm{C\,{\small IV}}) ({\rm cm}^{-2}) \approx 15.44$$, respectively. The uncertainties in these estimates depend mainly on the continuum placement used for the BAL profile fit. Experiments with different continua suggest that the maximum uncertainties are roughly 0.1 dex. It is important to note that the line optical depth and column densities derived above are only lower limits because BAL outflows often exhibit saturation with partial line-of-sight covering of the background light source. This can lead to weak/shallow absorption-line troughs even if the optical depths are large (e.g. Hamann 1998; Hamann & Sabra 2004; Arav et al. 2005; Capellupo et al. 2017; Moravec et al. 2017, and references therein). In one recent study, Herbst et al. (in preparation) used median composite spectra of BAL quasars from the Baryon Oscillation Spectroscopic Survey (BOSS; Ross et al. 2012; Dawson et al. 2013, part of the Sloan Digital Sky Survey-III, SDSS-III; Eisenstein et al. 2011) to show that the low-abundance doublet P v λ1118,1128 is often present and that widely separated doublets like P v, Si iv λ1393,1402, and O vi λ1032,1038 have ∼1:1 doublet ratios across a wide range of BAL strengths. These results indicate that all of the commonly measured BALs of abundant ions in quasar spectra, e.g. C iv λ1549, N v λ1240, and O vi λ1034, are typically saturated and the observed depths of the BAL troughs are controlled mainly by the line-of-sight covering fractions. 3.2 Other UV outflow features In addition to the UV BAL at 1346 Å, there are two other features in the 2000 HST-STIS spectrum that point to extreme outflows in PDS 456. First, as noted previously by O'Brien et al. (2005), the broad C iv emission line is extremely blueshifted. Our simple Gaussian fit to this feature shown in Fig. 3 indicates that the line centroid is blueshifted by 5200 ± 450 km s−1 [consistent with the O'Brien et al. (2005) measurement]. This fit also yields a rest equivalent width of REW =14.7 ± 1.8 Å and a large velocity width of FWHM =11 770 ± 1060 km s−1 (not corrected for the ∼500 km s−1 doublet separation in C iv). The low-ionization emission lines O i λ1304 and C ii λ1335 are poorly measured in this spectrum compared to C iv, but they are clearly less blueshifted. We crudely estimate their centroid blueshifts by visual inspection to be 700 ± 200 km s−1. The large C iv emission-line blueshift identifies an outflow-dominated broad emission-line region with an unusually high outflow speed. At v ≈ 5200 ± 450 km s−1, the C iv blueshift is outside the range of values reported by Coatman et al. (2016) for a sample of ∼31 157 luminous SDSS quasars (see also Richards et al. 2011). However, we can crudely estimate by extrapolation from their fig. 1 that the C iv blueshift in PDS 456 is in the upper ∼0.1 per cent of luminous quasars. To our knowledge, it is matched in the literature only by a few luminous extremely red quasars (ERQs) discovered in SDSS-III/BOSS to have prodigious outflows that include BALs in the UV and highly blueshifted [O iii] λ5007 emission lines (Zakamska et al. 2016; Hamann et al. 2017; Perrotta et al., in preparation). We will present a more thorough analysis of the C iv blueshift in a future paper. Here we note simply that large blueshifts are known to correlate with small emission-line REWs and intrinsically weak X-ray emissions (compared to other quasars/AGN at similar luminosities; e.g. Leighly 2004; Leighly et al. 2007; Richards et al. 2011; Wu et al. 2011, 2012; Luo et al. 2015). In addition, the work by Coatman et al. (2016) supports speculation in the ERQ studies that large blueshifts and other prominent outflow features are related to high accretion rates (high Eddington ratios) in the quasars. Another tentative outflow feature in the 2000 HST-STIS spectrum is a weak BAL marginally detected at ∼1513 Å (observed, see Figs 2 and 3). The reality of this feature depends on the continuum placement, but it is clearly below our best guess at the continuum and might be stronger in the HST-COS 2014 spectrum (Fig. 2). There is also tentative absorption near ∼1560 Å, but this feature is blended with telluric C iv absorption at 1549 Å and it appears to be much narrower than the BAL candidate at ∼1513 Å. Fits to the BAL candidate at 1513 Å are poorly constrained, but multiple trials indicate a central wavelength of λo ≈ 1513 ± 2 Å, FWHM ≈ 3800 ± 1000 km s−1, and REW ≈ 1.1 ± 0.5 Å in the quasar frame. If this BAL is real, the only plausible identification is C iv at v ∼ 0.19c. [The leading alternative S iv λ1398 at v ∼ 0.09c is ruled out by the absence of C iv absorption at the same velocity shift (see Rodríguez Hidalgo et al. 2011; Hamann et al. 2013).] 3.3 cloudy simulations We use the photoionization and spectral synthesis code cloudy version 17.00 (Ferland et al. 2013, 2017) to predict absorption-line strengths for different physical conditions that might produce the observed BAL at ∼1346 Å. The calculations assume twice solar metallicity and a fixed hydrogen density of nH = 107 cm−3. The specific density has no bearing on our results (e.g. Hamann 1997). The metallicity can affect mainly the metal line strengths relative to the Lyman lines. We choose twice solar metallicity to be crudely consistent with other studies of the outflows and broad emission-line environments of luminous quasars (e.g. Hamann & Ferland 1999; Dietrich et al. 2002; Warner, Hamann & Dietrich 2004; Gabel et al. 2005a; Nagao, Marconi & Maiolino 2006). The clouds are irradiated by a standard quasar ionizing spectrum consistent with measurements of PDS 456 (e.g. Nardini et al. 2015). This spectrum is defined by power laws across optical–UV and X-ray wavelengths with slopes of αuv = −0.5 and αx = −1.3, respectively (for fν ∝ να). The two power laws are joined smoothly in the far-UV by an exponential Wien function with temperature T = 250 000 K. The relative strengths of the UV and X-ray spectral segments are set by a two-point power-law index between 2500 Å and 2 keV equal to αox = −1.8 (Strateva et al. 2005; Steffen et al. 2006). This spectrum is defined in cloudy by the command: AGN T=250000K, a(ox)=-1.8, a(uv)=-0.5, a(x)=-1.3. The flux of hydrogen ionizing radiation incident on the model clouds is set by the ionization parameter,   \begin{eqnarray} U \ \equiv \ {{Q_{\rm H}}\over {4\pi c R^2\, n_{\rm H}}}, \end{eqnarray} (3)where R is the distance from the quasar light source and QH is the total emitted luminosity of hydrogen ionizing photons (#/s). Other aspects of the calculations are described below. 3.4 A comparison outflow sample Here we construct a sample of outflow quasars from SDSS-III/BOSS that can be useful analogues to test the BAL identification in PDS 456. We specifically use data from the BOSS quasar catalogue for data release 12 (DR12; Pâris et al. 2017) to select quasars with the following properties. (1) We require that the BOSS spectra have signal-to-noise ratios SNR > 5 at 1700 Å (quasar frame). (2) The emission-line redshifts must be ze > 2.4 to place the outflow lines of C iv, N v, and Ly α within the BOSS wavelength coverage at speeds up to at least ∼0.1c. Nearly 2/3 of the selected quasars have ze > 2.6 such that O vi λ1034 is also covered. (3) The quasars must have broad C iv outflow lines (BALs or mini-BALs) characterized by balnicity index BI < 3000 km s−1 and absorption index AI > 1000 km s−1, both measured at >4σ confidence (see Pâris et al. 2017, and references therein for definitions of BI and AI). These parameter limits are designed to exclude strong/deep C iv BALs (with large BI) as well as very narrow C iv lines with FWHMs < several hundred km s−1 (with small AI) that do not resemble the observed feature in PDS 456. Note that broad outflow lines are recorded in the BOSS quasar catalogue only at velocity shifts v ≲ 0.1c. (4) The minimum velocity recorded on the red side of the BAL/mini-BAL troughs is v > 6000 km s−1 (as specified by the AI integration limit vmin_civ_450 > 6000). This avoids broad lines near v ∼ 0 that can have different profiles and other properties compared to higher velocity BALs and mini-BALs. The specific value of 6000 km s−1 is a compromise that requires a substantial velocity shift while still maintaining a large sample size. Finally, (5) we reject a small fraction of the quasars (∼10 per cent) that have damped Ly α absorption (DLAs) or strong sub-DLAs at the wavelengths of interest (based on our own visual inspections of the spectra). The final sample includes 641 quasars at median redshift 〈ze〉 = 2.84 and median absolute i-band magnitude 〈Mi〉 = −26.8. We visually inspect all of the quasar spectra to assess the strengths of prominent outflow lines for comparison to PDS 456. At these redshifts and velocity shifts, the N v, Ly α, and O vi outflow lines are inevitably blended with unrelated absorption features in the Ly α forest. However, in most cases, the outflow lines are clearly discernible because they are much broader than the forest features. Fig. 4 shows spectra of four quasars we find to have the strongest well-measured N v outflow lines relative to C iv in our BOSS sample. These extreme cases with large N v/C iv are useful for our discussions of the BAL identification in PDS 456 (e.g. Section 3.6). Figure 4. View largeDownload slide Normalized spectra in the quasar frame of four outflow quasars from our BOSS sample with the strongest well-measured N v absorption lines relative to C iv. Blue arrows mark the broad outflow absorption lines of O vi, N v, and C iv from left to right, respectively. N v and O vi suffer from varying amounts of contamination in the Ly α forest (e.g. in the bottom panel there is Lyman limit absorption at wavelengths ≲912 Å and the deepest part of the broad N v trough is enhanced and offset from the absorption minima in C iv and O vi due to Ly α forest contamination). Prominent broad emission lines are labelled across the top. Also listed are the quasar names, emission-line redshifts, ze, and absorption indices, BI and AI, from the BOSS DR12 quasar catalogue. See Section 3.4. Figure 4. View largeDownload slide Normalized spectra in the quasar frame of four outflow quasars from our BOSS sample with the strongest well-measured N v absorption lines relative to C iv. Blue arrows mark the broad outflow absorption lines of O vi, N v, and C iv from left to right, respectively. N v and O vi suffer from varying amounts of contamination in the Ly α forest (e.g. in the bottom panel there is Lyman limit absorption at wavelengths ≲912 Å and the deepest part of the broad N v trough is enhanced and offset from the absorption minima in C iv and O vi due to Ly α forest contamination). Prominent broad emission lines are labelled across the top. Also listed are the quasar names, emission-line redshifts, ze, and absorption indices, BI and AI, from the BOSS DR12 quasar catalogue. See Section 3.4. It is also useful to consider the median outflow line properties in the BOSS sample. Thus, we construct a median composite spectrum in the outflow absorber frame for all 641 quasars. This composite ‘averages out’ the Ly α forest contamination to reveal the typical outflow line strengths and profiles across the sample. The result is shown in Fig. 5. We derive an absorber redshift for each quasar from the wavelength of minimum flux between the limits of the C iv AI integration (e.g. between vmin_civ_450 and vmax_civ_450) in smoothed versions of the BOSS spectra. The composite is then constructed by normalizing each spectrum to a continuum flux near 1700 Å and then shifting to the absorber frame before calculating the median. We remove broad emission lines from the final outflow quasar composite spectrum by dividing by another composite spectrum of ∼7700 BOSS quasars with similar redshifts and absolute magnitudes (Mi) but without broad outflow lines in their spectra. The redshifts of these non-outflow quasars are randomly displaced to match the distribution of shifts used to place the outflow quasars in the absorber frame. A simple division then yields the final composite spectrum shown in Fig. 5. See Baskin et al. (2013), Baskin, Laor & Hamann (2015), and Herbst et al. (in preparation) for more discussion of this procedure. Figure 5. View largeDownload slide Top panel: normalized composite spectrum in the absorber frame for 641 BOSS quasars selected to have C iv BALs/mini-BALs with moderate strengths and minimum trough velocities > 6000 km s−1 (Section 3.4). The dashed blue vertical lines with labels across the top mark absorption lines that are or might be present. The dashed red horizontal line marks the unabsorbed continuum. Bottom panel: number of quasars contributing to the composite at each wavelength. Figure 5. View largeDownload slide Top panel: normalized composite spectrum in the absorber frame for 641 BOSS quasars selected to have C iv BALs/mini-BALs with moderate strengths and minimum trough velocities > 6000 km s−1 (Section 3.4). The dashed blue vertical lines with labels across the top mark absorption lines that are or might be present. The dashed red horizontal line marks the unabsorbed continuum. Bottom panel: number of quasars contributing to the composite at each wavelength. The main result from Fig. 5 is that the O vi λ1034, N v λ1240, and C iv λ1549 outflow lines have roughly similar strengths while Ly α (in the blue wing of N v) is substantially weaker. This represents the typical situation for luminous quasars with BALs/mini-BALs crudely similar to the BAL in PDS 456. Another interesting result in Fig. 5 is the significant presence of P v λ1121 absorption and the resolved doublets with ∼1:1 ratios in O vi, P v, and Si iv. This supports the claim by Herbst et al. (in preparation) that BAL/mini-BAL outflows often have large total column densities and saturated absorption in all of the prominent lines. 3.5 Problems with the Ly α identification Fig. 6 shows theoretical line-centre optical depths from our cloudy simulations (Section 3.2) for lines that should accompany a Ly α BAL at 1346 Å in PDS 456. In particular, the solid curves show the optical depths for lines with b = 5135 km s−1 formed in clouds with a range of ionization parameters, U, but with neutral hydrogen column densities held fixed at the derived value of log NHI(cm−2) = 15.35 (Section 3.1). These predictions should be considered lower limits because NHI derived from the data is a lower limit. Holding NHI constant in the calculations leads to model clouds that have larger total hydrogen column densities, NH, at larger U. The log NH values for each log U are shown across the top of the figure. Figure 6. View largeDownload slide Line-centre optical depths versus ionization parameter, log U (bottom axis), and total hydrogen column density, log NH (top axis), predicted by cloudy simulations to accompany a Ly α BAL like the observed feature in PDS 456 (with τo = 0.35 and b = 5135 km s−1, Section 3.1). The dashed and dark solid curves are the same in both panels, representing model clouds with log NHI(cm−2) = 15.35 and vturb = 1000 km s−1. Shaded regions around the C iv, N v, and O vi curves show the ranges of optical depths that result from varying log NHI by ±0.1 (top panel) or changing the turbulence from vturb from 100 km s−1 to its maximum value, 5135 km s−1, set by the observed BAL width (bottom panel). These predictions show that the Ly α BAL identification would require log U < −2 and uniquely low column densities, log NH(cm−2) < 18.7, for a BAL outflow (see Section 3.5). Figure 6. View largeDownload slide Line-centre optical depths versus ionization parameter, log U (bottom axis), and total hydrogen column density, log NH (top axis), predicted by cloudy simulations to accompany a Ly α BAL like the observed feature in PDS 456 (with τo = 0.35 and b = 5135 km s−1, Section 3.1). The dashed and dark solid curves are the same in both panels, representing model clouds with log NHI(cm−2) = 15.35 and vturb = 1000 km s−1. Shaded regions around the C iv, N v, and O vi curves show the ranges of optical depths that result from varying log NHI by ±0.1 (top panel) or changing the turbulence from vturb from 100 km s−1 to its maximum value, 5135 km s−1, set by the observed BAL width (bottom panel). These predictions show that the Ly α BAL identification would require log U < −2 and uniquely low column densities, log NH(cm−2) < 18.7, for a BAL outflow (see Section 3.5). The shaded regions around the curves for C iv, N v, and O vi in Fig. 6 illustrate their dependence on the specific values of NHI and the turbulent velocity, vturb, used in the calculations. The solid curves for these lines (and the solid and dashed curves for C iii λ977, Si iii λ1206, and Si iv λ1398) correspond to fiducial parameters log NHI(cm−2) = 15.35 and vturb = 1000 km s−1. In the top panel, the shaded regions show the range of optical depths corresponding to the maximum uncertainty of ±0.1 dex in our log NHI measurement (see Section 3.1, where the smaller/larger NHI value sets the lower/upper envelope to the shaded curves). The optical depth changes in response to these changes in NHI are non-linear at large U (and large NH) because of radiative shielding in the far-UV that affects each ion differently. These shielding effects maintain significant optical depths in the moderate ions C iv, N v, and O vi at a fixed NHI and large U. In other calculations (not shown), we determine that the threshold for important shielding effects at large U is roughly log NHI(cm−2) ≳ 14.8. This is 0.55 dex lower than our lower limit to the H i column density inferred from the observed BAL (Section 3.1), and therefore the predictions in Fig. 6 that result from shielding at large U should be applicable. Another parameter that can affect radiative shielding is the velocity dispersion inside the cloud, which we characterize by the turbulence velocity vturb. Large internal velocity dispersions can enhance the shielding by blending together numerous absorption lines in the far-UV, which in turn affects the ionization structure deep in the cloud. The shaded regions around the C iv, N v, and O vi curves in the bottom panel of Fig. 6 depict a range of results corresponding to vturb = 100 km s−1, which yields negligible line shielding (the bottom envelope of the shaded regions), up to vturb = 5135 km s−1, which yields the maximum line shielding allowed by the observed BAL width (the upper envelope of the shaded regions). We adopt an intermediate value vturb = 1000 km s−1 as our fiducial case (here and in all of our calculations below) because it yields conservatively small amounts of line shielding and, therefore, conservatively small predicted line optical depths in C iv, N v, and O vi at large U. The main result from Fig. 6 is that a Ly α BAL should be accompanied by significant broad C iv and O vi absorption lines over a wide range of normal BAL outflow physical conditions. It is important to keep in mind that the predicted strengths of the C iv and O vi lines are lower limits based on a lower limit on NHI that follows from the assumption of no partial covering. Smaller values of NHI are not allowed because they cannot produce the observed BAL attributed to Ly α. Larger NH requiring larger NH (at a given U) would produce stronger C iv and O vi absorption. The absence of these lines in the PDS 456 spectrum (Figs 2 and 3) means that the only conditions consistent with the Ly α BAL identification are low degrees of ionization, log U < −2, and very low total column densities, log NH(cm−2) < 18.7 that are ≳3.5 orders of magnitude lower than expected for BAL outflows. For comparison, analysis of composite BAL quasar spectra by Herbst et al. (in preparation) suggests that typical BAL outflows have log NH(cm−2) ≳ 22.2 (see also Sections 1 and 3.4). Another problem for the Ly α identification is that a Ly α-only BAL system would be, to our knowledge, unprecedented among observed BAL outflows. Fig. 3 shows that the C iv and O vi BALs expected to accompany Ly α in PDS 456 would need to be <33 per cent the strength of Ly α to avoid detection. This is very different from the typical situation illustrated by Fig. 5 and by other BAL composite spectra (Baskin et al. 2013; Herbst et al., in preparation) where the C iv and O vi lines are ∼2 to ∼3 times stronger than Ly α. BAL surveys based on C iv BAL detections (e.g. Trump et al. 2006; Gibson et al. 2009; Pâris et al. 2017) are naturally biased towards stronger C iv lines and they cannot detect Ly α-only BAL systems if they exist. However, our visual inspections of the 641 outflow quasar spectra in our BOSS sample (Section 3.4) do not find any instances of C iv BALs weaker than, or even comparable to, the corresponding Ly α absorption. A third argument against the Ly α BAL identification stems from the weakening of the observed BAL between the HST observations in 2000 and 2014 (Fig. 2). Variability in BALs and other outflow lines can be attributed to outflow clouds moving across our lines of sight or to changes in the ionization caused by changes in the incident ionizing flux (cf. Arav et al. 2012, 2015; Capellupo et al. 2013, 2014; Filiz Ak et al. 2013; Misawa, Charlton & Eracleous 2014a; Grier et al. 2015; Rogerson et al. 2016; Moravec et al. 2017, and references therein). The evidence for BAL saturation discussed in Sections 3.1 and 3.4 might favour crossing clouds in many situations. However, if the BAL changes in PDS 456 were caused by ionization changes and the ionization was initially low to produce Ly α absorption without accompanying C iv and O vi lines, then the only way to make Ly α weaker at a fixed NH is with larger ionization parameters that should produce stronger C iv and O vi absorption. The continued absence of these accompanying lines in 2014 (Fig. 2) thus makes the Ly α identification more difficult because the values of NH and U in 2000 would need to be several times lower than the already-low upper limits deduced from Fig. 6. 3.6 Problems with the N v λ1240 identification The top panel in Fig. 7 shows cloudy predictions for the line-centre optical depths that should accompany the observed BAL if it is N v λ1240. As in Fig. 6, the calculations assume a cloud velocity dispersion vturb = 1000 km s−1 (to include moderate amounts of line shielding) but line optical depths derived for b = 5135 km s−1. The total column densities in the model clouds are adjusted to yield an N v column density, $$\log N(\rm{N\,{\small V}}) ({\rm cm}^{-2}) = 15.62$$, fixed to the value inferred from the observed BAL (Section 3.1). The resulting optical depths are again only lower limits because $$N(\rm{N\,{\small V}})$$ is a lower limit. Figure 7. View largeDownload slide Line-centre optical depths (black curves) and total hydrogen column densities, log NH (red curve), versus ionization parameter, log U, predicted by cloudy to accompany the observed BAL (with τo = 0.35 and b = 5135 km s−1; Section 3.1) if it is attributed to N v (top panel) or C iv (bottom panel). The ionic column densities held fixed in the model clouds are shown in the upper left of each panel. These predictions and existing spectra (Figs 1–3) rule out low degrees of ionization for the N v BAL system, and they can guide future observations to test both BAL identifications further (see Sections 3.6 and 3.7). Figure 7. View largeDownload slide Line-centre optical depths (black curves) and total hydrogen column densities, log NH (red curve), versus ionization parameter, log U, predicted by cloudy to accompany the observed BAL (with τo = 0.35 and b = 5135 km s−1; Section 3.1) if it is attributed to N v (top panel) or C iv (bottom panel). The ionic column densities held fixed in the model clouds are shown in the upper left of each panel. These predictions and existing spectra (Figs 1–3) rule out low degrees of ionization for the N v BAL system, and they can guide future observations to test both BAL identifications further (see Sections 3.6 and 3.7). The observed BAL at 1346 Å is compatible with the N v identification only if the accompanying C iv line is undetectable. This requires a ratio of N v/C iv absorption-line strengths ≳3 (Section 3.1, Fig. 3). Fig. 7 shows that moderate-to-low outflow ionizations with log U ≲ 0, corresponding to total columns log NH(cm−2) ≲ 20.2, are firmly ruled out for the N v BAL identification by the predictions for strong C iv absorption. At higher ionizations, the predicted optical depth ratio hovers around $$\tau (\rm{N\,{\small V}})/\tau (\rm{C\,{\small IV}}) \sim 3$$–5 (Fig. 7), which is marginally consistent with the observed spectrum. However, any significant partial covering effects would push the observed N v/C iv line depth ratio towards unity, such that C iv should be detectable and the N v BAL identification is ruled out. Figs 4 and 5 show more directly that absorption-line ratios N v/C iv ≳3 needed for the N v BAL identification are, at best, extremely rare in observed quasar outflows. In particular, the composite spectrum (Fig. 5) shows that typical weak-to-moderate BAL/mini-BAL systems have a median line depth ratio of N v/C iv ∼ 1 ± 0.1. Individual outflows can have larger N v/C iv ratios. Measurements in the literature of BALs/mini-BALs at speeds up to nearly 0.2c reveal N v/C iv depth ratios in the range ∼0.5–1.5 (Steidel 1990; Jannuzi et al. 1996; Barlow & Sargent 1997; Hamann et al. 1997a; Hamann, Barlow & Junkkarinen 1997c; Telfer et al. 1998; Sabra et al. 2003; Rodríguez Hidalgo 2008; Rodríguez Hidalgo et al. 2011, 2013; Rogerson et al. 2016; Moravec et al. 2017). Our visual inspections of the 641 outflow quasars in Section 3.4 find crudely ∼2 per cent of cases with N v/C iv ≳ 2. Fig. 4 shows the most extreme well-measured examples. It is difficult to assess the N v/C iv line ratios quantitatively in these spectra due to blending problems in the Ly α forest. However, the most extreme cases3 shown in Fig. 4 also appear to be inconsistent with the N v BAL identification in PDS 456. Another constraint on the N v BAL identification is that it should be accompanied by strong O vi absorption at observed wavelength ∼1122 Å. This is based on both theoretical predictions (Fig. 7) and observations of other outflow quasars (Section 3.4 and references immediately above). The wavelength 1122 Å is just outside the wavelength coverage of the 2000 STIS spectrum; however, there is no evidence for a broad O vi line wing that could be measurable if this line was present (Fig. 3). The 2014 COS spectrum does cover these O vi wavelengths and clearly shows no signs of absorption there (Fig. 2), but the BAL at 1346 Å was also much weaker in this spectrum. Thus, the constraints provided by the absence of O vi in existing spectra are weak, but they also do not support the N v BAL identification. 3.7 The case for C iv λ1549 at 0.30c One argument favouring the C iv BAL identification in PDS 456 is that it is readily compatible with existing spectra (Figs 1–3). It has none of the problems described above for Ly α and N v because no other lines are expected within the wavelength coverage. Observations of high-velocity BALs/mini-BALs in other quasars indicate that the outflow ionizations are generally high and that the most prominent lines accompanying C iv should be N v and O vi (Sections 3.4 and 3.6; Hamann et al. 2013). The bottom panel in Fig. 7 shows specific theoretical predictions for the optical depths in these and other lines that should accompany the observed BAL if it is C iv. The most observationally accessible lines in this plot are Ly α at low ionizations and N v at high ionizations, both at predicted observer-frame wavelengths near ∼1068 Å. Future observations might test for this absorption, but that will be difficult due to the short wavelength and severe reddening (Section 2). The situation is worse for other corroborating lines. In particular, the corresponding Si iv λ1398 line would be at ∼1216 Å observed, which is directly on top of the Galactic damped Ly α line, and the O vi and Ne viii λ774 lines, which should be strong at high ionizations, are at inaccessible observed wavelengths (∼899 and ∼673 Å, respectively) due to Galactic Lyman limit absorption. A second circumstantial argument favouring the C iv identification is that its velocity v ≈ 0.30c is similar to the speeds measured for X-ray outflow of PDS 456. The X-ray outflow is highly variable with multiple velocity components, but the main component identified by Fe K-shell absorption has measured velocities in the range v ∼ 0.25–0.34c with a typical value near ∼0.3c (Reeves et al. 2014, 2016; Nardini et al. 2015; Matzeu et al. 2017). The C iv BAL might provide evidence for a physical relationship between the UV and X-ray outflows in PDS 456 (see Section 4 below). A third argument is that relativistic C iv BALs/mini-BALs at speeds approaching v ∼ 0.2c have already been measured in a growing number of luminous quasars (Jannuzi et al. 1996; Hamann et al. 1997b, 2013; Rodríguez Hidalgo 2008; Rodríguez Hidalgo et al. 2011; Rogerson et al. 2016; Rodriguez Hidalgo et al., in preparation). The C iv BAL at v ≈ 0.30c in PDS 456 would set a new speed record for UV outflows, but it is not so dramatic to be a paradigm shift for our understanding of these outflows. The width and depth of the BAL in PDS 456 are roughly similar to these other high-velocity outflow features. We also note that our tentative detection of another weak C iv BAL in PDS 456 at v ∼ 0.19c (Fig. 2, Section 3.2) provides additional, albeit tentative, evidence that a relativistic UV outflow is present in PDS 456. 4 SUMMARY AND DISCUSSION The UV spectrum of PDS 456 obtained with HST-STIS in 2000 has a distinct BAL at ∼1346 Å that we identify as C iv λ1549 at velocity shift v ≈ 0.30c, FWHM ≈8550 km s−1, and minimum optical depth τo = 0.35 (Section 3.1). The C iv identification rests on its compatibility with existing spectra and the absence of lines that should accompany the alternative identifications, Ly α λ1216 and N v λ1240. Ly α is compatible with the observed BAL only if the outflow has a low degree of ionization, log U < −2, and a very low total column density, log NH(cm−2) < 18.7, that would be unprecedented in BAL outflow studies, e.g. several orders of magnitude below recent estimates (Section 3.5). The N v identification might be consistent with the absence of accompanying C iv absorption if the gas is highly ionized and it pushes the boundary of observed N v/C iv line strengths beyond what we find in our comparison of 641 outflow quasars in BOSS (Section 3.6). However, this situation should produce strong O vi absorption, which is not well constrained in existing spectra of PDS 456 but it appears to be absent (Fig. 3). Thus, the N v identification is also strongly disfavoured. The C iv BAL identification has none of these problems (Section 3.7). It would mark the fastest UV outflow line ever reported, but its velocity is consistent with the X-ray outflow in PDS 456 (see Section 1 and below) and not dramatically different from the relativistic C iv BALs/mini-BALs already measured at speeds approaching ∼0.2c in other quasars (Jannuzi et al. 1996; Hamann et al. 1997b, 2013; Rodríguez Hidalgo 2008; Rodríguez Hidalgo et al. 2011; Rogerson et al. 2016; Rodriguez Hidalgo et al., in preparation). The C iv identification is also weakly supported by our tentative identification of an additional C iv BAL feature at v ∼ 0.19c (Fig. 2, Section 3.2). Broad UV outflow lines at speeds near 0.3c are surely very rare based on the rarity of such lines at ∼0.1–0.2c (see references above). However, the incidence of C iv BALs/mini-BALs at v > 0.2c is not known because they have not been searched for in large quasar surveys like SDSS/BOSS and, in any case, they could easily be missed due to blends with unrelated lines in the Ly α forest (e.g. see the search for P v λ1121 BALs by Capellupo et al. 2017). The location of the C iv BAL outflow in PDS 456 is a critical unknown. The range of ionizations consistent with C iv absorption might favour lower ionizations and larger distances from the black hole than the X-ray outflow, perhaps at radii of the order of ∼1 pc as inferred from some variability studies of C iv BALs/mini-BALs in other quasars (Hamann et al. 2013; Capellupo et al. 2014; Moravec et al. 2017; McGraw et al. 2018). However, larger distances do not necessarily produce lower ionizations, e.g. if acceleration causes the outflow densities to drop faster than the 1/r2 behaviour expected from free expansion at a constant speed. Moreover, the relativistic speed of the C iv BAL at v ∼ 0.30c indicates that the UV outflow originated with the X-ray UFO very close to the black hole. If the measured flow speeds are similar to the gravitational escape speed at the launch radius, then the launch point is roughly at r ∼ 20–30 rg (Nardini et al. 2015; Matzeu et al. 2017). Also note that a highly ionized X-ray outflow launched from this radius and expanding freely outwards into a fixed solid angle will not necessarily become less ionized at larger distances (for C iv absorption) because the 1/r2 dilution of the radiation field is balanced by a 1/r2 decline in the densities to yield a constant ionization parameter. Lower ionizations will occur if the inner regions of the outflow radiatively shield the material downstream and/or if there are clumps with enhanced densities relative to the ambient flow. Clumping and shielding can occur to the same effect at almost any radius, and there is already evidence for dense clumps in the X-ray outflow of PDS 456 (based on lower ionizations and partial covering in the soft X-ray absorber; Matzeu et al. 2016; Reeves et al. 2016; see also Gofford et al. 2014; Hagino et al. 2015; Nardini et al. 2015). It is therefore an intriguing possibility that the C iv BAL forms directly within, or in close proximity to, the relativistic X-ray outflow. Our cloudy simulations (Fig. 7) show that the C iv BAL in PDS 456 could form over a wide range of physical conditions, including very high ionizations where C iv is just a trace constituent. This situation is illustrated in Fig. 8, which plots the ionization structure in a single model cloud with log U = 1.7. This cloud reaches the observed minimum BAL optical depth $$\tau _{\rm o} (\rm{C\,{\small IV}}) = 0.35$$, along with $$\tau _{\rm o} (\rm{N\,{\small V}}) = 0.90$$ and $$\tau _{\rm o} (\rm{O\,\,{\small VI}}) = 43$$, at total column density log NH(cm−2) = 22.9 (represented by the unshaded left-hand portion of Fig. 8). This front portion of the cloud matches our calculations in the bottom panel of Fig. 7. In this environment, the C iv and N v ion fractions are everywhere ≲4 × 10−4 and ≲3 × 10−3, respectively, and the dominant form of oxygen is O vii. This highly ionized environment capable of C iv λ1549 absorption will also produce saturated absorption in higher ion lines such as Ne viii λ774 and Mg x λ615 (see also log U = 1.7 in Fig. 7) and at the bound–free edges of O vii and O viii in soft X-rays. Figure 8. View largeDownload slide Ionization fractions versus total hydrogen column density, log NH, in a single model cloud with ionization parameter, log U = 1.7. This cloud produces the minimum observed BAL optical depth in C iv $$\tau _{\rm o} (\rm{C\,{\small IV}}) = 0.35$$, at log NH(cm−2) ≈ 22.9, such that the unshaded left-hand region represents the model cloud with log U = 1.7 in Fig. 7. This environment (at log NH(cm−2) ≤ 22.9) is highly ionized and optically thick at the bound–free edges of O vii, O viii, and Ne viii, with only trace amounts of the lower ions C iv, N v, and O vi. The lower ions dominate at larger column densities (grey shaded region) where there is more radiative shielding, but these lower ionizations are not needed for the C iv BAL in PDS 456. Figure 8. View largeDownload slide Ionization fractions versus total hydrogen column density, log NH, in a single model cloud with ionization parameter, log U = 1.7. This cloud produces the minimum observed BAL optical depth in C iv $$\tau _{\rm o} (\rm{C\,{\small IV}}) = 0.35$$, at log NH(cm−2) ≈ 22.9, such that the unshaded left-hand region represents the model cloud with log U = 1.7 in Fig. 7. This environment (at log NH(cm−2) ≤ 22.9) is highly ionized and optically thick at the bound–free edges of O vii, O viii, and Ne viii, with only trace amounts of the lower ions C iv, N v, and O vi. The lower ions dominate at larger column densities (grey shaded region) where there is more radiative shielding, but these lower ionizations are not needed for the C iv BAL in PDS 456. Detailed comparisons between the UV and X-ray outflows in PDS 456 are beyond the scope of the present study. They are subject to uncertainties caused by the outflow variabilities and by unknowns in the shape of the ionizing spectrum and the spatial locations of different outflow components relative to the ionizing far-UV and X-ray emission sources. Here we note simply that recent estimates of the X-ray outflow ionization parameters4 are roughly in the range log U ∼ 4.3–4.8 for the K-shell outflow (Gofford et al. 2014; Nardini et al. 2015) and log U ∼ 2.8 for the soft X-ray absorber (Reeves et al. 2016). The ionization parameters needed for the C iv BAL, with upper limit log U ≲ 2.4 (Fig. 7), might occur in dense clumps embedded in the X-ray outflow if the density enhancements are ≳0.4 dex relative to the soft X-ray absorber or ≳2 dex relative to the Fe K-shell outflow. The ionization structure of the UV outflow might resemble the quasar SBS 1542+541 (redshift z ∼ 2.36), which has measured UV BALs ranging in ionization from C iv up to Mg x and Si xii λ510 (Telfer et al. 1998). The lines in this quasar appear saturated with shallow troughs that reveal ion-dependent line-of-sight covering fractions from ∼15 per cent in C iv to ∼50 per cent in the higher ions. Telfer et al. (1998) infer from this a two-zone outflow that could span a decade or more in ionization parameter, with the C iv lines forming in small clumps embedded in a more highly ionized outflow medium. Telfer et al. (1998) also note that this outflow should produce substantial bound–free absorption by O vii and O viii in soft X-rays. The general picture of clumpy outflows has become commonplace in quasar outflow studies. High-quality observations of UV outflow lines often provide evidence for clumpy multi-phase outflow structures with a range of covering fractions (Ganguly et al. 1999; de Kool, Korista & Arav 2002; Hamann & Sabra 2004; Arav et al. 2005, 2008; Gabel et al. 2005a; Misawa et al. 2007, 2014b; Leighly et al. 2009, 2011, 2015; Hamann et al. 2011; Moravec et al. 2017). Detailed studies of some bright Seyfert 1 galaxies (e.g. Kaspi et al. 2002; Netzer et al. 2003; Gabel et al. 2005b) clearly demonstrate that the UV and X-ray absorption features can form together in complex outflows, with indications that the lower ionization UV lines identify clumps or filaments embedded in the X-ray outflow. Clumpy outflow structures are also predicted by recent numerical simulations (Sim et al. 2010; Takeuchi, Ohsuga & Mineshige 2013; Waters et al. 2017) and by considerations of the radiative forces that can compress the outflows into small substructures (Stern et al. 2014). There are also theoretical arguments requiring dense clumps to moderate the outflow ionizations in the absence of significant radiative shielding (De Kool 1997; Hamann et al. 2013). If the C iv-absorbing gas in PDS 456 is indeed embedded in the X-ray outflow, it could have important implications for the outflow energetics. It is a well-known problem for X-ray UFOs (e.g. Tombesi et al. 2011, 2013; Gofford et al. 2015) that their high ionizations lead to low opacities and inefficient radiative acceleration (Gofford et al. 2013, 2014; Higginbottom et al. 2014). Harnessing the full radiative power of the quasar to drive these outflows might require opacities beyond electron scattering at UV/far-UV wavelengths, near the peak of the quasar spectral energy distributions. Matzeu et al. (2017) showed recently that the outflow speeds in PDS 456 correlate with the variable X-ray luminosity, consistent with radiative acceleration (see also Saez & Chartas 2011). If radiative forces are important, then dense clumps with lower ionizations embedded in the X-ray outflows might be important to boost the opacities for radiative driving (Laor & Davis 2014; Hagino et al. 2015). The C iv BAL at v ≈ 0.30c in PDS 456 could be the first direct observational evidence for this idea. PDS 456 is a remarkable object with powerful accretion-disc outflows launched from a range of at least two decades in disc radii, from ∼0.001 pc for the X-ray outflow to ∼0.3 pc for the blueshifted CIV broad emission line (Section 1; O'Brien et al. 2005; Gofford et al. 2014). More work is needed to understand how the UV BAL fits into this outflow environment and, specifically, to test the C iv BAL identification. A search for corroborating N v absorption at ∼1077 Å will be difficult due to Galactic reddening (Section 2) and the poor sensitivity of current instruments at these wavelengths (Section 3.7). However, the alternative BAL identifications, N v and Ly α, might be ruled out or confirmed more easily by searching for O vi absorption near 1122 Å (Section 3.6, although the BAL at 1346 Å needs to be present for a meaningful test). Our team has ongoing programmes to extend the UV wavelength coverage and monitor PDS 456 in the UV and X-rays that will be described in future papers. ACKNOWLEDGEMENTS We are grateful to Gary Ferland and the cloudy development team for their continued support and public dispersement of the spectral synthesis code cloudy. FH also thanks Nahum Arav, Chris Done, Gerald Kriss, and Paola Rodriguez Hidalgo for helpful discussions. We are grateful to an anonymous referee for helpful comments on the manuscript. JNR acknowledges the support via HST grant HST-GO-14477.001-A. EN acknowledges funding from the European Union's Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement no. 664931. Footnotes 1 Throughout this paper, we treat unresolved doublets as single lines, such that, for example, C iv λ1548,1551 becomes C ivλ1549 with a summed oscillator strength (for column density estimates). 2 We note, however, that the observed spectrum is poorly fitted at wavelengths ≳2400 Å (observed). This might be caused by reddening in the quasar environment, a break in the intrinsic/emitted spectral slope, and/or blended Fe ii emission lines that are known to be strong in PDS 456 (Simpson et al. 1999). We do not investigate this further because our goal is simply to define a continuum for analysis of the outflow lines at shorter wavelengths (Section 3). 3 We need to acknowledge here that hypothetical quasars with strong N v BALs but negligible C iv absorption cannot appear in our BOSS sample because it relies on C iv line detections (via AI and BI). 4 The ionization parameter quoted in the X-ray studies is ξ = Lion/nHr2 from Tarter, Tucker & Salpeter (1969), where Lion is the quasar luminosity from 1 to 1000 Ryd. For ξ in units of erg cm s−1 and the continuum shape used in our cloudy simulations (Section 3.3), the conversion is log U ≈ log ξ − 1.2. REFERENCES Arav N. et al.  , 2001, ApJ , 561, 118 https://doi.org/10.1086/323369 CrossRef Search ADS   Arav N., Kaastra J., Kriss G. A., Korista K. T., Gabel J., Proga D., 2005, ApJ , 620, 665 https://doi.org/10.1086/425560 CrossRef Search ADS   Arav N., Moe M., Costantini E., Korista K. T., Benn C., Ellison S., 2008, ApJ , 681, 954 https://doi.org/10.1086/588651 CrossRef Search ADS   Arav N. et al.  , 2012, A&A , 544, A33 CrossRef Search ADS   Arav N. et al.  , 2015, A&A , 577, A37 CrossRef Search ADS   Barlow T. A., Sargent W. L. W., 1997, AJ , 113, 136 https://doi.org/10.1086/118239 CrossRef Search ADS   Baskin A., Laor A., Hamann F., 2013, MNRAS , 432, 1525 https://doi.org/10.1093/mnras/stt582 CrossRef Search ADS   Baskin A., Laor A., Hamann F., 2015, MNRAS , 449, 1593 https://doi.org/10.1093/mnras/stv406 CrossRef Search ADS   Borguet B. C. J., Edmonds D., Arav N., Benn C., Chamberlain C., 2012, ApJ , 758, 69 https://doi.org/10.1088/0004-637X/758/1/69 CrossRef Search ADS   Capellupo D. M., Hamann F., Shields J. C., Halpern J. P., Barlow T. A., 2013, MNRAS , 429, 1872 https://doi.org/10.1093/mnras/sts427 CrossRef Search ADS   Capellupo D. M., Hamann F., Barlow T. A., 2014, MNRAS , 444, 1893 https://doi.org/10.1093/mnras/stu1502 CrossRef Search ADS   Capellupo D. M. et al.  , 2017, MNRAS , 469, 323 https://doi.org/10.1093/mnras/stx870 CrossRef Search ADS   Cardelli J. A., Clayton G. C., Mathis J. S., 1989, ApJ , 345, 245 https://doi.org/10.1086/167900 CrossRef Search ADS   Chartas G., Brandt W. N., Gallagher S. C., Garmire G. P., 2002, ApJ , 579, 169 https://doi.org/10.1086/342744 CrossRef Search ADS   Chartas G. et al.  , 2009, New Astron. Rev. , 53, 128 https://doi.org/10.1016/j.newar.2009.07.010 CrossRef Search ADS   Coatman L., Hewett P. C., Banerji M., Richards G. T., 2016, MNRAS , 461, 647 https://doi.org/10.1093/mnras/stw1360 CrossRef Search ADS   Crenshaw D. M., Kraemer S. B., George I. M., 2003, ARA&A , 41, 117 CrossRef Search ADS   Dawson K. S. et al.  , 2013, AJ , 145, 10 https://doi.org/10.1088/0004-6256/145/1/10 CrossRef Search ADS   De Kool M., 1997, in Arav N., Shlosman I., Weymann R. J., eds, ASP Conf. Ser. Vol. 128, Mass Ejection from Active Galactic Nuclei . Astron. Soc. Pac., San Francisco, p. 233 de Kool M., Korista K. T., Arav N., 2002, ApJ , 580, 54 https://doi.org/10.1086/343107 CrossRef Search ADS   Debuhr J., Quataert E., Ma C.-P., 2012, MNRAS , 420, 2221 https://doi.org/10.1111/j.1365-2966.2011.20187.x CrossRef Search ADS   Di Matteo T., Springel V., Hernquist L., 2005, Nature , 433, 604 https://doi.org/10.1038/nature03335 CrossRef Search ADS PubMed  Dietrich M., Hamann F., Shields J. C., Constantin A., Vestergaard M., Chaffee F., Foltz C. B., Junkkarinen V. T., 2002, ApJ , 581, 912 https://doi.org/10.1086/344410 CrossRef Search ADS   Eisenstein D. J. et al.  , 2011, AJ , 142, 72 https://doi.org/10.1088/0004-6256/142/3/72 CrossRef Search ADS   Ferland G. J. et al.  , 2013, Rev. Mex. Astron. Astrofis. , 49, 137 Ferland G. J. et al.  , 2017, Rev. Mex. Astron. Astrofis. , 53, 385 Filiz Ak N. et al.  , 2013, ApJ , 777, 168 https://doi.org/10.1088/0004-637X/777/2/168 CrossRef Search ADS   Gabel J. R. et al.  , 2005a, ApJ , 623, 85 https://doi.org/10.1086/428477 CrossRef Search ADS   Gabel J. R. et al.  , 2005b, ApJ , 631, 741 https://doi.org/10.1086/432682 CrossRef Search ADS   Ganguly R., Eracleous M., Charlton J. C., Churchill C. W., 1999, AJ , 117, 2594 https://doi.org/10.1086/300882 CrossRef Search ADS   Gibson R. R. et al.  , 2009, ApJ , 692, 758 https://doi.org/10.1088/0004-637X/692/1/758 CrossRef Search ADS   Gofford J., Reeves J. N., Tombesi F., Braito V., Turner T. J., Miller L., Cappi M., 2013, MNRAS , 430, 60 https://doi.org/10.1093/mnras/sts481 CrossRef Search ADS   Gofford J. et al.  , 2014, ApJ , 784, 77 https://doi.org/10.1088/0004-637X/784/1/77 CrossRef Search ADS   Gofford J., Reeves J. N., McLaughlin D. E., Braito V., Turner T. J., Tombesi F., Cappi M., 2015, MNRAS , 451, 4169 https://doi.org/10.1093/mnras/stv1207 CrossRef Search ADS   Grier C. J. et al.  , 2015, ApJ , 806, 111 https://doi.org/10.1088/0004-637X/806/1/111 CrossRef Search ADS   Hagino K., Odaka H., Done C., Gandhi P., Watanabe S., Sako M., Takahashi T., 2015, MNRAS , 446, 663 https://doi.org/10.1093/mnras/stu2095 CrossRef Search ADS   Hamann F., 1997, ApJS , 109, 279 https://doi.org/10.1086/312980 CrossRef Search ADS   Hamann F., 1998, ApJ , 500, 798 https://doi.org/10.1086/305776 CrossRef Search ADS   Hamann F., Ferland G., 1999, ARA&A , 37, 487 CrossRef Search ADS   Hamann F., Sabra B., 2004, in Richards G. T., Hall P. B., eds, ASP Conf. Ser. Vol. 311, AGN Physics with the Sloan Digital Sky Survey . Astron. Soc. Pac., San Francisco, p. 203 Hamann F., Barlow T., Cohen R. D., Junkkarinen V., Burbidge E. M., 1997a, in Arav N., Shlosman I., Weymann R. J., eds, ASP Conf. Ser. Vol. 128, Mass Ejection from Active Galactic Nuclei . Astron. Soc. Pac., San Francisco, p. 19 Hamann F., Barlow T. A., Junkkarinen V., Burbidge E. M., 1997b, ApJ , 478, 80 https://doi.org/10.1086/303781 CrossRef Search ADS   Hamann F., Barlow T. A., Junkkarinen V., 1997c, ApJ , 478, 87 https://doi.org/10.1086/303782 CrossRef Search ADS   Hamann F., Kanekar N., Prochaska J. X., Murphy M. T., Ellison S., Malec A. L., Milutinovic N., Ubachs W., 2011, MNRAS , 410, 1957 Hamann F., Chartas G., McGraw S., Rodriguez Hidalgo P., Shields J., Capellupo D., Charlton J., Eracleous M., 2013, MNRAS , 435, 133 https://doi.org/10.1093/mnras/stt1231 CrossRef Search ADS   Hamann F. et al.  , 2017, MNRAS , 464, 3431 https://doi.org/10.1093/mnras/stw2387 CrossRef Search ADS   Higginbottom N., Proga D., Knigge C., Long K. S., Matthews J. H., Sim S. A., 2014, ApJ , 789, 19 https://doi.org/10.1088/0004-637X/789/1/19 CrossRef Search ADS   Hopkins P. F., Elvis M., 2010, MNRAS , 401, 7 https://doi.org/10.1111/j.1365-2966.2009.15643.x CrossRef Search ADS   Hopkins P. F., Hernquist L., Cox T. J., Kereš D., 2008, ApJS , 175, 356 https://doi.org/10.1086/524362 CrossRef Search ADS   Jannuzi B. T. et al.  , 1996, ApJ , 470, L11 https://doi.org/10.1086/310301 CrossRef Search ADS   Kaspi S. et al.  , 2002, ApJ , 574, 643 https://doi.org/10.1086/341113 CrossRef Search ADS   Kaspi S., Maoz D., Netzer H., Peterson B. M., Vestergaard M., Jannuzi B. T., 2005, ApJ , 629, 61 https://doi.org/10.1086/431275 CrossRef Search ADS   Knigge C., Scaringi S., Goad M. R., Cottis C. E., 2008, MNRAS , 386, 1426 https://doi.org/10.1111/j.1365-2966.2008.13081.x CrossRef Search ADS   Korista K. T., Voit G. M., Morris S. L., Weymann R. J., 1993, ApJS , 88, 357 https://doi.org/10.1086/191825 CrossRef Search ADS   Laor A., Davis S. W., 2014, MNRAS , 438, 3024 https://doi.org/10.1093/mnras/stt2408 CrossRef Search ADS   Leighly K. M., 2004, ApJ , 611, 125 https://doi.org/10.1086/422089 CrossRef Search ADS   Leighly K. M., Halpern J. P., Jenkins E. B., Casebeer D., 2007, ApJS , 173, 1 https://doi.org/10.1086/519768 CrossRef Search ADS   Leighly K. M., Hamann F., Casebeer D. A., Grupe D., 2009, ApJ , 701, 176 https://doi.org/10.1088/0004-637X/701/1/176 CrossRef Search ADS   Leighly K. M., Dietrich M., Barber S., 2011, ApJ , 728, 94 https://doi.org/10.1088/0004-637X/728/2/94 CrossRef Search ADS   Leighly K. M., Cooper E., Grupe D., Terndrup D. M., Komossa S., 2015, ApJ , 809, L13 https://doi.org/10.1088/2041-8205/809/1/L13 CrossRef Search ADS   Luo B. et al.  , 2015, ApJ , 805, 122 https://doi.org/10.1088/0004-637X/805/2/122 CrossRef Search ADS   Matzeu G. A., Reeves J. N., Nardini E., Braito V., Costa M. T., Tombesi F., Gofford J., 2016, MNRAS , 458, 1311 https://doi.org/10.1093/mnras/stw354 CrossRef Search ADS   Matzeu G. A., Reeves J. N., Braito V., Nardini E., McLaughlin D. E., Lobban A. P., Tombesi F., Costa M. T., 2017, MNRAS , 472, L15 CrossRef Search ADS   McGraw S. M., Shields J. C., Hamann F. W., Capellupo D. M., Herbst H., 2018, MNRAS , 475, 585 CrossRef Search ADS   Misawa T., Eracleous M., Charlton J. C., Kashikawa N., 2007, ApJ , 660, 152 https://doi.org/10.1086/513097 CrossRef Search ADS   Misawa T., Charlton J. C., Eracleous M., 2014a, ApJ , 792, 77 https://doi.org/10.1088/0004-637X/792/1/77 CrossRef Search ADS   Misawa T., Inada N., Oguri M., Gandhi P., Horiuchi T., Koyamada S., Okamoto R., 2014b, ApJ , 794, L20 https://doi.org/10.1088/2041-8205/794/2/L20 CrossRef Search ADS   Moravec E. A., Hamann F., Capellupo D. M., McGraw S. M., Shields J. C., Rodríguez Hidalgo P., 2017, MNRAS , 468, 4539 https://doi.org/10.1093/mnras/stx775 CrossRef Search ADS   Nagao T., Marconi A., Maiolino R., 2006, A&A , 447, 157 CrossRef Search ADS   Nardini E. et al.  , 2015, Science , 347, 860 https://doi.org/10.1126/science.1259202 CrossRef Search ADS PubMed  Netzer H. et al.  , 2003, ApJ , 599, 933 https://doi.org/10.1086/379508 CrossRef Search ADS   O'Brien P. T., Reeves J. N., Simpson C., Ward M. J., 2005, MNRAS , 360, L25 https://doi.org/10.1111/j.1745-3933.2005.00038.x CrossRef Search ADS   Pâris I. et al.  , 2017, A&A , 597, A79 CrossRef Search ADS   Reeves J. N., O'Brien P. T., Ward M. J., 2003, ApJ , 593, L65 https://doi.org/10.1086/378218 CrossRef Search ADS   Reeves J. N. et al.  , 2009, ApJ , 701, 493 https://doi.org/10.1088/0004-637X/701/1/493 CrossRef Search ADS   Reeves J. N. et al.  , 2014, ApJ , 780, 45 https://doi.org/10.1088/0004-637X/780/1/45 CrossRef Search ADS   Reeves J. N., Braito V., Nardini E., Behar E., O'Brien P. T., Tombesi F., Turner T. J., Costa M. T., 2016, ApJ , 824, 20 https://doi.org/10.3847/0004-637X/824/1/20 CrossRef Search ADS   Richards G. T. et al.  , 2011, AJ , 141, 167 https://doi.org/10.1088/0004-6256/141/5/167 CrossRef Search ADS   Rodríguez Hidalgo P., 2008, PhD thesis , Univ. Florida, Gainesville. Available at: http://etd.fcla.edu/UF/UFE0022674/rodriguez_p.pdf Rodríguez Hidalgo P., Hamann F., Hall P., 2011, MNRAS , 411, 247 https://doi.org/10.1111/j.1365-2966.2010.17677.x CrossRef Search ADS   Rodríguez Hidalgo P., Eracleous M., Charlton J., Hamann F., Murphy M. T., Nestor D., 2013, ApJ , 775, 14 https://doi.org/10.1088/0004-637X/775/1/14 CrossRef Search ADS   Rogerson J. A., Hall P. B., Rodríguez Hidalgo P., Pirkola P., Brandt W. N., Filiz Ak N., 2016, MNRAS , 457, 405 https://doi.org/10.1093/mnras/stv3010 CrossRef Search ADS   Ross N. P. et al.  , 2012, ApJS , 199, 3 https://doi.org/10.1088/0067-0049/199/1/3 CrossRef Search ADS   Rupke D. S. N., Veilleux S., 2013, ApJ , 768, 75 https://doi.org/10.1088/0004-637X/768/1/75 CrossRef Search ADS   Sabra B. M., Hamann F., Jannuzi B. T., George I. M., Shields J. C., 2003, ApJ , 590, 66 https://doi.org/10.1086/375006 CrossRef Search ADS   Saez C., Chartas G., 2011, ApJ , 737, 91 https://doi.org/10.1088/0004-637X/737/2/91 CrossRef Search ADS   Sim S. A., Proga D., Miller L., Long K. S., Turner T. J., 2010, MNRAS , 408, 1396 https://doi.org/10.1111/j.1365-2966.2010.17215.x CrossRef Search ADS   Simpson C., Ward M., O'Brien P., Reeves J., 1999, MNRAS , 303, L23 https://doi.org/10.1046/j.1365-8711.1999.02344.x CrossRef Search ADS   Steffen A. T., Strateva I., Brandt W. N., Alexander D. M., Koekemoer A. M., Lehmer B. D., Schneider D. P., Vignali C., 2006, AJ , 131, 2826 https://doi.org/10.1086/503627 CrossRef Search ADS   Steidel C. C., 1990, ApJS , 74, 37 https://doi.org/10.1086/191493 CrossRef Search ADS   Stern J., Behar E., Laor A., Baskin A., Holczer T., 2014, MNRAS , 445, 3011 https://doi.org/10.1093/mnras/stu1960 CrossRef Search ADS   Strateva I. V., Brandt W. N., Schneider D. P., Vanden Berk D. G., Vignali C., 2005, AJ , 130, 387 https://doi.org/10.1086/431247 CrossRef Search ADS   Takeuchi S., Ohsuga K., Mineshige S., 2013, PASJ , 65, 88 https://doi.org/10.1093/pasj/65.4.88 CrossRef Search ADS   Tarter C. B., Tucker W. H., Salpeter E. E., 1969, ApJ , 156, 943 https://doi.org/10.1086/150026 CrossRef Search ADS   Telfer R. C., Kriss G. A., Zheng W., Davidsen A. F., Green R. F., 1998, ApJ , 509, 132 https://doi.org/10.1086/306491 CrossRef Search ADS   Tombesi F., Cappi M., Reeves J. N., Palumbo G. G. C., Yaqoob T., Braito V., Dadina M., 2010, A&A , 521, A57 CrossRef Search ADS   Tombesi F., Cappi M., Reeves J. N., Palumbo G. G. C., Braito V., Dadina M., 2011, ApJ , 742, 44 https://doi.org/10.1088/0004-637X/742/1/44 CrossRef Search ADS   Tombesi F., Cappi M., Reeves J. N., Braito V., 2012, MNRAS , 422, L1 https://doi.org/10.1111/j.1745-3933.2012.01221.x CrossRef Search ADS   Tombesi F., Cappi M., Reeves J. N., Nemmen R. S., Braito V., Gaspari M., Reynolds C. S., 2013, MNRAS , 430, 1102 https://doi.org/10.1093/mnras/sts692 CrossRef Search ADS   Torres C. A. O., Quast G. R., Coziol R., Jablonski F., de la Reza R., Lépine J. R. D., Gregório-Hetem J., 1997, ApJ , 488, L19 https://doi.org/10.1086/310913 CrossRef Search ADS   Trump J. R. et al.  , 2006, ApJS , 165, 1 https://doi.org/10.1086/503834 CrossRef Search ADS   Warner C., Hamann F., Dietrich M., 2004, ApJ , 608, 136 https://doi.org/10.1086/386325 CrossRef Search ADS   Waters T., Proga D., Dannen R., Kallman T. R., 2017, MNRAS , 467, 3160 Weymann R. J., Morris S. L., Foltz C. B., Hewett P. C., 1991, ApJ , 373, 23 https://doi.org/10.1086/170020 CrossRef Search ADS   Wu J. et al.  , 2011, ApJ , 736, 28 https://doi.org/10.1088/0004-637X/736/1/28 CrossRef Search ADS   Wu J., Brandt W. N., Anderson S. F., Diamond-Stanic A. M., Hall P. B., Plotkin R. M., Schneider D. P., Shemmer O., 2012, ApJ , 747, 10 https://doi.org/10.1088/0004-637X/747/1/10 CrossRef Search ADS   Zakamska N. L. et al.  , 2016, MNRAS , 459, 3144 https://doi.org/10.1093/mnras/stw718 CrossRef Search ADS   © 2018 The Author(s) Published by Oxford University Press on behalf of the Royal Astronomical Society http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Monthly Notices of the Royal Astronomical Society Oxford University Press

Does the X-ray outflow quasar PDS 456 have a UV outflow at 0.3c?

, Volume 476 (1) – May 1, 2018
11 pages

/lp/ou_press/does-the-x-ray-outflow-quasar-pds-456-have-a-uv-outflow-at-0-3c-mAR0oUmek0
Publisher
Oxford University Press
ISSN
0035-8711
eISSN
1365-2966
D.O.I.
10.1093/mnras/sty043
Publisher site
See Article on Publisher Site

Abstract

Abstract The quasar PDS 456 (at redshift ∼0.184) has a prototype ultra-fast outflow (UFO) measured in X-rays. This outflow is highly ionized with relativistic speeds, large total column densities log NH(cm−2) > 23, and large kinetic energies that could be important for feedback to the host galaxy. A UV spectrum of PDS 456 obtained with the Hubble Space Telescope in 2000 contains one well-measured broad absorption line (BAL) at ∼1346 Å (observed) that might be Ly α at v ≈ 0.06c or N v λ1240 at v ≈ 0.08c. However, we use photoionization models and comparisons to other outflow quasars to show that these BAL identifications are problematic because other lines that should accompany them are not detected. We argue that the UV BAL is probably C iv at v ≈ 0.30c. This would be the fastest UV outflow ever reported, but its speed is similar to the X-ray outflow and its appearance overall is similar to relativistic UV BALs observed in other quasars. The C iv BAL identification is also supported indirectly by the tentative detection of another broad C iv line at v ≈ 0.19c. The high speeds suggest that the UV outflow originates with the X-ray UFO crudely 20–30 rg from the central black hole. We speculate that the C iv BAL might form in dense clumps embedded in the X-ray UFO, requiring density enhancements of only ≳0.4 dex compared to clumpy structures already inferred for the soft X-ray absorber in PDS 456. The C iv BAL might therefore be the first detection of low-ionization clumps proposed previously to boost the opacities in UFOs for radiative driving. line: formation, quasars: absorption lines, quasars: general, quasars: individual: PDS 456 1 INTRODUCTION Accretion-disc outflows are an important part of the quasar phenomenon that might drive ‘feedback’ to regulate black hole growth and host galaxy evolution (e.g. Di Matteo, Springel & Hernquist 2005; Hopkins et al. 2008; Hopkins & Elvis 2010; Debuhr, Quataert & Ma 2012; Rupke & Veilleux 2013). The outflows are often studied in the rest-frame UV via blueshifted broad absorption lines (BALs) or their narrower cousins called ‘mini-BALs’ (with a nominal boundary near full width at half-minimum FWHM ∼2000 km s−1; Weymann et al. 1991; Korista et al. 1993; Crenshaw, Kraemer & George 2003; Hamann & Sabra 2004; Trump et al. 2006; Knigge et al. 2008; Gibson et al. 2009, and references therein). These features appear most often at moderate velocity shifts v < 0.1c, but relativistic BALs and mini-BALs at v ∼ 0.1–0.2c have been measured in a small but growing number of quasars (Hamann et al. 1997a, 2013; Rodríguez Hidalgo 2008; Rodríguez Hidalgo, Hamann & Hall 2011; Rogerson et al. 2016). The basic physical properties for these outflows can be difficult to determine due to limited wavelength coverage and line saturation that is masked by partial line-of-sight covering of the background light source(s). The strongest measured UV lines1 are typically C iv λ1549, N v λ1240, and O vi 1034, indicating moderate to high degrees of ionization. The maximum ionizations are not known, but there might be a wide range. O vi absorption tends to be as strong as or stronger than C iv (Section 3.4 below, also Baskin, Laor & Hamann 2013; Moravec et al. 2017; Herbst et al., in preparation) and higher ionization lines such as Ne viii λ774 and Mg x λ615 have been observed in a few cases with suitable spectral coverage (e.g. Hamann et al. 1997b; Telfer et al. 1998; Arav et al. 2001). Measurements of the low-abundance line P v λ1121 help to overcome the saturation issues to reveal generally large total hydrogen column densities, log NH(cm−2) ≳ 22.3, across a wide range observed BAL strengths (Hamann 1998; Leighly et al. 2009; Leighly, Dietrich & Barber 2011; Borguet et al. 2012; Capellupo, Hamann & Barlow 2014; Capellupo et al. 2017; Moravec et al. 2017, but see also Arav et al. 2001). X-ray observations have revealed another variety of ultra-fast outflows (UFOs) that reach relativistic speeds in luminous quasars (Chartas et al. 2002, 2009; Reeves et al. 2009; Tombesi et al. 2010, 2013; Gofford et al. 2013, 2015, and references therein). UFOs are also challenging to study because they are highly variable and highly ionized to the point where the only strong absorption features appear at X-ray wavelengths. They are often characterized by Fe K-shell absorption features with derived total column densities in the range log NH(cm−2) ∼ 22–24 (see references above). UFOs also appear to have generally very large kinetic energies, sufficient to drive feedback effects in the quasar host galaxies (Tombesi et al. 2012, 2013; Gofford et al. 2014, 2015; Reeves et al. 2014). The quasar PDS 456 (at redshift ze ≈ 0.184) has the best-studied example of a powerful, relativistic X-ray UFO (e.g. Reeves, O'Brien & Ward 2003; Reeves et al. 2009, 2014, 2016; Gofford et al. 2014; Nardini et al. 2015; Matzeu et al. 2017). It is the most luminous quasar in the local Universe, with bolometric luminosity L ∼ 1047 erg s−1 and estimated black hole mass ∼(1–2) × 109 M⊙ that together indicate an accretion rate relative to Eddington that is L/LEdd ≳ 0.3 and perhaps near unity (Nardini et al. 2015). The X-ray absorber is complex and highly variable, with two main components. The main component measured via Fe K-shell absorption has speeds in the range v ∼ 0.25–0.34c, very high degrees of ionization featuring Fe xxv and Fe xxvi, and large total column densities log NH(cm−2) ≳ 23 (Gofford et al. 2014, 2015; Reeves et al. 2014; Nardini et al. 2015). Its radial distance from the black hole is estimated at a few hundred gravitational radii (based on absorber variability; e.g. Nardini et al. 2015). The kinetic power of this outflow is also remarkably large, crudely ∼20 per cent of the bolometric luminosity, which is well above the threshold needed for feedback to the host galaxy (Gofford et al. 2014; Nardini et al. 2015, and references therein). The second outflow component measured in soft X-rays has somewhat lower speeds, v ∼ 0.17–0.27c, lower ionizations, and column densities in the range log NH(cm−2) ∼ 22–23 (Reeves et al. 2016). There is evidence for time-variable covering fractions in this absorber that might be indicative of small dense clumps or substructures embedded in the overall X-ray outflow (Matzeu et al. 2016). PDS 456 also has signatures of outflow in the UV. A UV spectrum obtained in 2000 by O'Brien et al. (2005) using the Space Telescope Imaging Spectrometer (STIS) on board the Hubble Space Telescope (HST) revealed a highly blueshifted C iv broad emission line at speeds near v ∼ 5000 km s−1 plus a single BAL that might plausibly be Ly α or N v at v ∼ 0.06c or ∼0.08c, respectively. These UV features provide further evidence for exotic mass-loss from PDS 456 across a wide range of spatial scales, from the X-ray UFO that is believed to originate near the black hole at radii of ∼20–30 rg (gravitational radii, corresponding to ∼0.001–0.0015 pc; Nardini et al. 2015; Matzeu et al. 2017) to an outflow-dominated C iv broad emission-line region that we place crudely at ∼0.3 pc [based on scaling relations with luminosity (Kaspi et al. 2005) for a 109 M⊙ black hole]. The UV BAL is an important component to this outflow picture, but its measured speed and physical nature depend critically on the line identification. In this paper, we re-examine the UV BAL in PDS 456 with the main result that it is likely to be C iv λ1549 at v ≈ 0.30c. This would be the fastest UV outflow line ever reported but similar in speed to the X-ray UFO in this quasar. Throughout this paper, we adopt a redshift for PDS 456 of ze = 0.183 75 based on the emission line [Fe ii] 1.6435 μm measured by Simpson et al. (1999). We describe two archival HST spectra of PDS 456 in Section 2 below. Section 3 presents our analysis and measurements of the observed BAL, comparisons to photoionization models, comparisons to UV BALs in other quasars, and a discussion of the plausible BAL identifications. Section 4 presents a summary and discussion of the results. 2 HST SPECTRA Fig. 1 shows the spectrum of PDS 456 obtained by O'Brien et al. (2005) in 2000 using the Hubble Space Telescopea Imaging Spectrometer (STIS) with the G140L and G230L gratings. We obtained this spectrum and another HST spectrum measured in 2014 (described below) from the Mikulski Archive for Space Telescopes. We present them here without further processing. The STIS G140L grating provided wavelength coverage from 1137 to 1715 Å at resolutions ranging from ∼310 to ∼200 km s−1, while STIS G230L covered 1580–3148 Å at resolutions from ∼640 to ∼300 km s−1 (see O'Brien et al. 2005 for more details). The spectrum plotted in Fig. 1 is combined from the two gratings by masking out the extreme ends of the wavelength coverage to avoid excessive noise and then calculating variance-weighted average fluxes at the remaining wavelengths of overlap. Figure 1. View largeDownload slide HST-STIS spectrum of PDS 456 from 2000 plotted as observed (bottom panel) and reddening-corrected using a Galactic extinction curve with E(B − V) = 0.45 (top panel). The grey dotted curves are the corresponding error spectra. The red dashed curves are an approximate fit to the continuum using a power law with index αλ = −1.68 shown with (bottom panel) and without (top) Galactic reddening. The red dotted curves in the bottom panel show the same power law reddened alternatively by E(B − V) = 0.42 and 0.48. The BAL we attribute to C iv at v ≈ 0.30c is labelled above the spectrum at 1346 Å. Some broad emission-line wavelengths are marked across the top. Galactic absorption lines (including strong damped Ly α at 1216 Å) are marked by green vertical lines near the bottom. See Section 2. Figure 1. View largeDownload slide HST-STIS spectrum of PDS 456 from 2000 plotted as observed (bottom panel) and reddening-corrected using a Galactic extinction curve with E(B − V) = 0.45 (top panel). The grey dotted curves are the corresponding error spectra. The red dashed curves are an approximate fit to the continuum using a power law with index αλ = −1.68 shown with (bottom panel) and without (top) Galactic reddening. The red dotted curves in the bottom panel show the same power law reddened alternatively by E(B − V) = 0.42 and 0.48. The BAL we attribute to C iv at v ≈ 0.30c is labelled above the spectrum at 1346 Å. Some broad emission-line wavelengths are marked across the top. Galactic absorption lines (including strong damped Ly α at 1216 Å) are marked by green vertical lines near the bottom. See Section 2. The UV spectrum of PDS 456 is substantially reddened by dust in our Galaxy (due to the quasar's sky position near the Galactic plane). Previous studies based on visible spectra and photometry indicate that the visual extinction towards PDS 456 is AV ∼ 1.5 mag, corresponding to selective extinction E(B − V) ∼ 0.48 for a standard Galactic reddening curve (Torres et al. 1997; Simpson et al. 1999). We obtain an independent estimate of the reddening by fitting the UV continuum with a single power law modified by the Galactic extinction curve from Cardelli, Clayton & Mathis (1989) with RV = 3.1. The fit is constrained by the median flux in wavelength windows between ∼1170 and ∼2400 Å that avoid strong emission and absorption lines. The dashed red curve in Fig. 1 shows our preferred fit using E(B − V) = 0.45 and power-law index αλ = −1.68 (for $$F_{\lambda }\propto \lambda ^{\alpha _{\lambda }}$$). An important feature of Galactic extinction at these wavelengths is the ‘bump’ at ∼2175 Å. We experimented with different fit parameters (e.g. the dotted red curves in Fig. 1), but the values of E(B − V) and the power-law slope are well constrained2 by the data because Galactic reddening curves with a strong 2175 Å bump suppress the flux at both ends of the spectral coverage shown in Fig. 1. The top panel in this figure shows the reddening-corrected spectrum of PDS 456 together with the best-fitting power law. Fig. 2 shows a previously unpublished spectrum of PDS 456 obtained in 2014 with HST Cosmic Origins Spectrograph (COS) using the G140L grating (PI: O'Brien). This spectrum (shown by the red curve) is plotted on top of the HST-STIS spectrum (black curve) from Fig. 1 with no reddening corrections. It is combined from separate exposures totalling 5186 s. The spectral resolution ranges from roughly 200 to 100 km s−1 from blue to red across the wavelengths shown in Fig. 2. A gap in the COS wavelength coverage from ∼1170 to ∼1265 Å avoids Galactic/geocoronal Ly α absorption/emission. The blue arrows in Figs 1 and 2 mark the positions of observed or expected absorption lines discussed in Section 3 below. Figure 2. View largeDownload slide HST spectra of PDS 456 obtained with STIS in 2000 (black curve) and COS in 2014 (red) plotted at observed wavelengths, uncorrected for reddening. The red and grey dotted curves are the corresponding error spectra (1σ uncertainties per pixel). The STIS spectrum is scaled vertically by a factor of ∼0.60 to match the COS spectrum approximately in the continuum. The dashed blue curve shows our fit to the STIS spectrum from Fig. 1. The BAL we attribute to C iv λ1549 at 0.30c is labelled below the spectrum at ∼1346 Å. Also marked are the expected positions of O vi λ1034 and C iv BALs if the observed BAL is attributed, instead, to N v λ1240 at 0.08c or Ly α at 0.06c (e.g. ‘if N v, Ly α’). We tentatively identify another weak BAL at ∼1513 Å as C iv at 0.19c. Dashed blue vertical lines mark the positions of Galactic absorption features. The narrow emission spike at 1304 Å is geocoronal O i. See Fig. 1 and Section 2 for additional notes. Figure 2. View largeDownload slide HST spectra of PDS 456 obtained with STIS in 2000 (black curve) and COS in 2014 (red) plotted at observed wavelengths, uncorrected for reddening. The red and grey dotted curves are the corresponding error spectra (1σ uncertainties per pixel). The STIS spectrum is scaled vertically by a factor of ∼0.60 to match the COS spectrum approximately in the continuum. The dashed blue curve shows our fit to the STIS spectrum from Fig. 1. The BAL we attribute to C iv λ1549 at 0.30c is labelled below the spectrum at ∼1346 Å. Also marked are the expected positions of O vi λ1034 and C iv BALs if the observed BAL is attributed, instead, to N v λ1240 at 0.08c or Ly α at 0.06c (e.g. ‘if N v, Ly α’). We tentatively identify another weak BAL at ∼1513 Å as C iv at 0.19c. Dashed blue vertical lines mark the positions of Galactic absorption features. The narrow emission spike at 1304 Å is geocoronal O i. See Fig. 1 and Section 2 for additional notes. 3 ANALYSIS 3.1 The UV BAL at 1346 Å The HST-STIS spectrum in Fig. 1 clearly shows a BAL at ∼1346 Å (observed), as reported by O'Brien et al. (2005). Fig. 3 plots this spectrum again after normalizing by the reddened power law in Figs 1 and 2. We fit the BAL in this normalized spectrum using a simple Gaussian optical depth profile of the form   \begin{eqnarray} \tau _{\rm v} \ = \ \tau _{\rm o}\, {\rm e}^{-{v}^2/b^2}, \end{eqnarray} (1)where τo is the line-centre optical depth, v is the velocity shift from the line centre, and b is the Doppler parameter that sets the line width. The BAL fit shown by the magenta curve in Fig. 3 yields τo = 0.35 ± 0.01, b = 5135 ± 203 km s−1, and observed line-centre wavelength λo = 1345.7 ± 0.6 Å (where the errors quoted are 1σ uncertainties returned by the line-fitting routine). The full width at half-minimum of this fitted profile is FWHM ≈ 8550 ± 340 km s−1. In the quasar frame, the fit has rest equivalent width REW = 10.8 ± 0.6 Å. Figure 3. View largeDownload slide HST-STIS spectrum of PDS 456 in the observed frame, normalized by the continuum fit shown in Fig. 1. The smooth magenta and orange curves show our fit to the observed BAL at 1346 Å and then that fit transposed to the wavelengths of O vi and C iv assuming that the measured BAL is Ly α (magenta) or N v (orange). The weaker profiles drawn at the C iv positions show the transposed fit at 1/3 its measured strength to illustrate approximate upper limits to the actual absorption there. The absence of absorption at the transposed fit positions argues against the Ly α and N v identifications for the observed BAL (Sections 3.5 and 3.6). Other labels are the same as Figs 1 and 2. Narrow Galactic absorption lines of O i λ1304 and C ii λ1335 have been removed by interpolation for clarity. Figure 3. View largeDownload slide HST-STIS spectrum of PDS 456 in the observed frame, normalized by the continuum fit shown in Fig. 1. The smooth magenta and orange curves show our fit to the observed BAL at 1346 Å and then that fit transposed to the wavelengths of O vi and C iv assuming that the measured BAL is Ly α (magenta) or N v (orange). The weaker profiles drawn at the C iv positions show the transposed fit at 1/3 its measured strength to illustrate approximate upper limits to the actual absorption there. The absence of absorption at the transposed fit positions argues against the Ly α and N v identifications for the observed BAL (Sections 3.5 and 3.6). Other labels are the same as Figs 1 and 2. Narrow Galactic absorption lines of O i λ1304 and C ii λ1335 have been removed by interpolation for clarity. Our study concerns identification of the ∼1346 Å BAL trough. One possibility favoured by O'Brien et al. (2005) is Ly α 1216 at velocity shift v ≈ 0.06c (18 000 km s−1). Another is N v λ1240 at v ≈ 0.08c. These identifications seem reasonable because the observed BAL is on the blue side of the Ly α–N v emission-line blend and the inferred velocities are in a normal range for BAL outflows in other quasars. However, we will argue below (Sections 3.5 and 3.6) that the Ly α or N v identifications are problematic because they are not accompanied by C iv and O vi λ1034 BALs at the same velocity shifts. The absence of these other lines is illustrated by the other orange and magenta curves in Fig. 3, which show our fit to the observed BAL transposed to the positions expected for accompanying C iv and O vi BALs. At the C iv positions, the transposed BAL fit is shown twice – once at full strength and once at 1/3 of its measured strength at 1346 Å. We estimate that the 1/3 scalings represent approximate upper limits to C iv BALs that might accompany an N v or Ly α BAL at 1346 Å. We will argue below that the absence of O vi and C iv absorption at these wavelengths suggests that the BAL at ∼1346 Å is actually C iv λ1549 at v ≈ 0.30c. We test these line identifications using photoionization models (Section 3.3) and comparisons to BALs/mini-BALs observed in other quasars (Section 3.4). First, we estimate ionic column densities for each possible identification. These estimates follow from the fitted Gaussian optical depth profile by   \begin{eqnarray} N_{\rm i} \ = \ {{m_{\rm e}\, c}\over {\sqrt{\pi }\, e^2}} {{b\,\tau _{\rm o} }\over {f\lambda _{\rm r}}}, \end{eqnarray} (2)where Ni is the ionic column density, f and λr are the oscillator strength and rest-frame wavelength of the transition, and we assume implicitly that the ground-state column density equals the ionic column density. Thus, we find that, if the observed BAL is Ly α, it corresponds to a neutral hydrogen column density of log NHI(cm−2) ≈ 15.35. If the BAL is N v or C iv, then the column density is $$\log N(\rm{N\,{\small V}}) ({\rm cm}^{-2}) \approx 15.62$$ or $$\log N(\rm{C\,{\small IV}}) ({\rm cm}^{-2}) \approx 15.44$$, respectively. The uncertainties in these estimates depend mainly on the continuum placement used for the BAL profile fit. Experiments with different continua suggest that the maximum uncertainties are roughly 0.1 dex. It is important to note that the line optical depth and column densities derived above are only lower limits because BAL outflows often exhibit saturation with partial line-of-sight covering of the background light source. This can lead to weak/shallow absorption-line troughs even if the optical depths are large (e.g. Hamann 1998; Hamann & Sabra 2004; Arav et al. 2005; Capellupo et al. 2017; Moravec et al. 2017, and references therein). In one recent study, Herbst et al. (in preparation) used median composite spectra of BAL quasars from the Baryon Oscillation Spectroscopic Survey (BOSS; Ross et al. 2012; Dawson et al. 2013, part of the Sloan Digital Sky Survey-III, SDSS-III; Eisenstein et al. 2011) to show that the low-abundance doublet P v λ1118,1128 is often present and that widely separated doublets like P v, Si iv λ1393,1402, and O vi λ1032,1038 have ∼1:1 doublet ratios across a wide range of BAL strengths. These results indicate that all of the commonly measured BALs of abundant ions in quasar spectra, e.g. C iv λ1549, N v λ1240, and O vi λ1034, are typically saturated and the observed depths of the BAL troughs are controlled mainly by the line-of-sight covering fractions. 3.2 Other UV outflow features In addition to the UV BAL at 1346 Å, there are two other features in the 2000 HST-STIS spectrum that point to extreme outflows in PDS 456. First, as noted previously by O'Brien et al. (2005), the broad C iv emission line is extremely blueshifted. Our simple Gaussian fit to this feature shown in Fig. 3 indicates that the line centroid is blueshifted by 5200 ± 450 km s−1 [consistent with the O'Brien et al. (2005) measurement]. This fit also yields a rest equivalent width of REW =14.7 ± 1.8 Å and a large velocity width of FWHM =11 770 ± 1060 km s−1 (not corrected for the ∼500 km s−1 doublet separation in C iv). The low-ionization emission lines O i λ1304 and C ii λ1335 are poorly measured in this spectrum compared to C iv, but they are clearly less blueshifted. We crudely estimate their centroid blueshifts by visual inspection to be 700 ± 200 km s−1. The large C iv emission-line blueshift identifies an outflow-dominated broad emission-line region with an unusually high outflow speed. At v ≈ 5200 ± 450 km s−1, the C iv blueshift is outside the range of values reported by Coatman et al. (2016) for a sample of ∼31 157 luminous SDSS quasars (see also Richards et al. 2011). However, we can crudely estimate by extrapolation from their fig. 1 that the C iv blueshift in PDS 456 is in the upper ∼0.1 per cent of luminous quasars. To our knowledge, it is matched in the literature only by a few luminous extremely red quasars (ERQs) discovered in SDSS-III/BOSS to have prodigious outflows that include BALs in the UV and highly blueshifted [O iii] λ5007 emission lines (Zakamska et al. 2016; Hamann et al. 2017; Perrotta et al., in preparation). We will present a more thorough analysis of the C iv blueshift in a future paper. Here we note simply that large blueshifts are known to correlate with small emission-line REWs and intrinsically weak X-ray emissions (compared to other quasars/AGN at similar luminosities; e.g. Leighly 2004; Leighly et al. 2007; Richards et al. 2011; Wu et al. 2011, 2012; Luo et al. 2015). In addition, the work by Coatman et al. (2016) supports speculation in the ERQ studies that large blueshifts and other prominent outflow features are related to high accretion rates (high Eddington ratios) in the quasars. Another tentative outflow feature in the 2000 HST-STIS spectrum is a weak BAL marginally detected at ∼1513 Å (observed, see Figs 2 and 3). The reality of this feature depends on the continuum placement, but it is clearly below our best guess at the continuum and might be stronger in the HST-COS 2014 spectrum (Fig. 2). There is also tentative absorption near ∼1560 Å, but this feature is blended with telluric C iv absorption at 1549 Å and it appears to be much narrower than the BAL candidate at ∼1513 Å. Fits to the BAL candidate at 1513 Å are poorly constrained, but multiple trials indicate a central wavelength of λo ≈ 1513 ± 2 Å, FWHM ≈ 3800 ± 1000 km s−1, and REW ≈ 1.1 ± 0.5 Å in the quasar frame. If this BAL is real, the only plausible identification is C iv at v ∼ 0.19c. [The leading alternative S iv λ1398 at v ∼ 0.09c is ruled out by the absence of C iv absorption at the same velocity shift (see Rodríguez Hidalgo et al. 2011; Hamann et al. 2013).] 3.3 cloudy simulations We use the photoionization and spectral synthesis code cloudy version 17.00 (Ferland et al. 2013, 2017) to predict absorption-line strengths for different physical conditions that might produce the observed BAL at ∼1346 Å. The calculations assume twice solar metallicity and a fixed hydrogen density of nH = 107 cm−3. The specific density has no bearing on our results (e.g. Hamann 1997). The metallicity can affect mainly the metal line strengths relative to the Lyman lines. We choose twice solar metallicity to be crudely consistent with other studies of the outflows and broad emission-line environments of luminous quasars (e.g. Hamann & Ferland 1999; Dietrich et al. 2002; Warner, Hamann & Dietrich 2004; Gabel et al. 2005a; Nagao, Marconi & Maiolino 2006). The clouds are irradiated by a standard quasar ionizing spectrum consistent with measurements of PDS 456 (e.g. Nardini et al. 2015). This spectrum is defined by power laws across optical–UV and X-ray wavelengths with slopes of αuv = −0.5 and αx = −1.3, respectively (for fν ∝ να). The two power laws are joined smoothly in the far-UV by an exponential Wien function with temperature T = 250 000 K. The relative strengths of the UV and X-ray spectral segments are set by a two-point power-law index between 2500 Å and 2 keV equal to αox = −1.8 (Strateva et al. 2005; Steffen et al. 2006). This spectrum is defined in cloudy by the command: AGN T=250000K, a(ox)=-1.8, a(uv)=-0.5, a(x)=-1.3. The flux of hydrogen ionizing radiation incident on the model clouds is set by the ionization parameter,   \begin{eqnarray} U \ \equiv \ {{Q_{\rm H}}\over {4\pi c R^2\, n_{\rm H}}}, \end{eqnarray} (3)where R is the distance from the quasar light source and QH is the total emitted luminosity of hydrogen ionizing photons (#/s). Other aspects of the calculations are described below. 3.4 A comparison outflow sample Here we construct a sample of outflow quasars from SDSS-III/BOSS that can be useful analogues to test the BAL identification in PDS 456. We specifically use data from the BOSS quasar catalogue for data release 12 (DR12; Pâris et al. 2017) to select quasars with the following properties. (1) We require that the BOSS spectra have signal-to-noise ratios SNR > 5 at 1700 Å (quasar frame). (2) The emission-line redshifts must be ze > 2.4 to place the outflow lines of C iv, N v, and Ly α within the BOSS wavelength coverage at speeds up to at least ∼0.1c. Nearly 2/3 of the selected quasars have ze > 2.6 such that O vi λ1034 is also covered. (3) The quasars must have broad C iv outflow lines (BALs or mini-BALs) characterized by balnicity index BI < 3000 km s−1 and absorption index AI > 1000 km s−1, both measured at >4σ confidence (see Pâris et al. 2017, and references therein for definitions of BI and AI). These parameter limits are designed to exclude strong/deep C iv BALs (with large BI) as well as very narrow C iv lines with FWHMs < several hundred km s−1 (with small AI) that do not resemble the observed feature in PDS 456. Note that broad outflow lines are recorded in the BOSS quasar catalogue only at velocity shifts v ≲ 0.1c. (4) The minimum velocity recorded on the red side of the BAL/mini-BAL troughs is v > 6000 km s−1 (as specified by the AI integration limit vmin_civ_450 > 6000). This avoids broad lines near v ∼ 0 that can have different profiles and other properties compared to higher velocity BALs and mini-BALs. The specific value of 6000 km s−1 is a compromise that requires a substantial velocity shift while still maintaining a large sample size. Finally, (5) we reject a small fraction of the quasars (∼10 per cent) that have damped Ly α absorption (DLAs) or strong sub-DLAs at the wavelengths of interest (based on our own visual inspections of the spectra). The final sample includes 641 quasars at median redshift 〈ze〉 = 2.84 and median absolute i-band magnitude 〈Mi〉 = −26.8. We visually inspect all of the quasar spectra to assess the strengths of prominent outflow lines for comparison to PDS 456. At these redshifts and velocity shifts, the N v, Ly α, and O vi outflow lines are inevitably blended with unrelated absorption features in the Ly α forest. However, in most cases, the outflow lines are clearly discernible because they are much broader than the forest features. Fig. 4 shows spectra of four quasars we find to have the strongest well-measured N v outflow lines relative to C iv in our BOSS sample. These extreme cases with large N v/C iv are useful for our discussions of the BAL identification in PDS 456 (e.g. Section 3.6). Figure 4. View largeDownload slide Normalized spectra in the quasar frame of four outflow quasars from our BOSS sample with the strongest well-measured N v absorption lines relative to C iv. Blue arrows mark the broad outflow absorption lines of O vi, N v, and C iv from left to right, respectively. N v and O vi suffer from varying amounts of contamination in the Ly α forest (e.g. in the bottom panel there is Lyman limit absorption at wavelengths ≲912 Å and the deepest part of the broad N v trough is enhanced and offset from the absorption minima in C iv and O vi due to Ly α forest contamination). Prominent broad emission lines are labelled across the top. Also listed are the quasar names, emission-line redshifts, ze, and absorption indices, BI and AI, from the BOSS DR12 quasar catalogue. See Section 3.4. Figure 4. View largeDownload slide Normalized spectra in the quasar frame of four outflow quasars from our BOSS sample with the strongest well-measured N v absorption lines relative to C iv. Blue arrows mark the broad outflow absorption lines of O vi, N v, and C iv from left to right, respectively. N v and O vi suffer from varying amounts of contamination in the Ly α forest (e.g. in the bottom panel there is Lyman limit absorption at wavelengths ≲912 Å and the deepest part of the broad N v trough is enhanced and offset from the absorption minima in C iv and O vi due to Ly α forest contamination). Prominent broad emission lines are labelled across the top. Also listed are the quasar names, emission-line redshifts, ze, and absorption indices, BI and AI, from the BOSS DR12 quasar catalogue. See Section 3.4. It is also useful to consider the median outflow line properties in the BOSS sample. Thus, we construct a median composite spectrum in the outflow absorber frame for all 641 quasars. This composite ‘averages out’ the Ly α forest contamination to reveal the typical outflow line strengths and profiles across the sample. The result is shown in Fig. 5. We derive an absorber redshift for each quasar from the wavelength of minimum flux between the limits of the C iv AI integration (e.g. between vmin_civ_450 and vmax_civ_450) in smoothed versions of the BOSS spectra. The composite is then constructed by normalizing each spectrum to a continuum flux near 1700 Å and then shifting to the absorber frame before calculating the median. We remove broad emission lines from the final outflow quasar composite spectrum by dividing by another composite spectrum of ∼7700 BOSS quasars with similar redshifts and absolute magnitudes (Mi) but without broad outflow lines in their spectra. The redshifts of these non-outflow quasars are randomly displaced to match the distribution of shifts used to place the outflow quasars in the absorber frame. A simple division then yields the final composite spectrum shown in Fig. 5. See Baskin et al. (2013), Baskin, Laor & Hamann (2015), and Herbst et al. (in preparation) for more discussion of this procedure. Figure 5. View largeDownload slide Top panel: normalized composite spectrum in the absorber frame for 641 BOSS quasars selected to have C iv BALs/mini-BALs with moderate strengths and minimum trough velocities > 6000 km s−1 (Section 3.4). The dashed blue vertical lines with labels across the top mark absorption lines that are or might be present. The dashed red horizontal line marks the unabsorbed continuum. Bottom panel: number of quasars contributing to the composite at each wavelength. Figure 5. View largeDownload slide Top panel: normalized composite spectrum in the absorber frame for 641 BOSS quasars selected to have C iv BALs/mini-BALs with moderate strengths and minimum trough velocities > 6000 km s−1 (Section 3.4). The dashed blue vertical lines with labels across the top mark absorption lines that are or might be present. The dashed red horizontal line marks the unabsorbed continuum. Bottom panel: number of quasars contributing to the composite at each wavelength. The main result from Fig. 5 is that the O vi λ1034, N v λ1240, and C iv λ1549 outflow lines have roughly similar strengths while Ly α (in the blue wing of N v) is substantially weaker. This represents the typical situation for luminous quasars with BALs/mini-BALs crudely similar to the BAL in PDS 456. Another interesting result in Fig. 5 is the significant presence of P v λ1121 absorption and the resolved doublets with ∼1:1 ratios in O vi, P v, and Si iv. This supports the claim by Herbst et al. (in preparation) that BAL/mini-BAL outflows often have large total column densities and saturated absorption in all of the prominent lines. 3.5 Problems with the Ly α identification Fig. 6 shows theoretical line-centre optical depths from our cloudy simulations (Section 3.2) for lines that should accompany a Ly α BAL at 1346 Å in PDS 456. In particular, the solid curves show the optical depths for lines with b = 5135 km s−1 formed in clouds with a range of ionization parameters, U, but with neutral hydrogen column densities held fixed at the derived value of log NHI(cm−2) = 15.35 (Section 3.1). These predictions should be considered lower limits because NHI derived from the data is a lower limit. Holding NHI constant in the calculations leads to model clouds that have larger total hydrogen column densities, NH, at larger U. The log NH values for each log U are shown across the top of the figure. Figure 6. View largeDownload slide Line-centre optical depths versus ionization parameter, log U (bottom axis), and total hydrogen column density, log NH (top axis), predicted by cloudy simulations to accompany a Ly α BAL like the observed feature in PDS 456 (with τo = 0.35 and b = 5135 km s−1, Section 3.1). The dashed and dark solid curves are the same in both panels, representing model clouds with log NHI(cm−2) = 15.35 and vturb = 1000 km s−1. Shaded regions around the C iv, N v, and O vi curves show the ranges of optical depths that result from varying log NHI by ±0.1 (top panel) or changing the turbulence from vturb from 100 km s−1 to its maximum value, 5135 km s−1, set by the observed BAL width (bottom panel). These predictions show that the Ly α BAL identification would require log U < −2 and uniquely low column densities, log NH(cm−2) < 18.7, for a BAL outflow (see Section 3.5). Figure 6. View largeDownload slide Line-centre optical depths versus ionization parameter, log U (bottom axis), and total hydrogen column density, log NH (top axis), predicted by cloudy simulations to accompany a Ly α BAL like the observed feature in PDS 456 (with τo = 0.35 and b = 5135 km s−1, Section 3.1). The dashed and dark solid curves are the same in both panels, representing model clouds with log NHI(cm−2) = 15.35 and vturb = 1000 km s−1. Shaded regions around the C iv, N v, and O vi curves show the ranges of optical depths that result from varying log NHI by ±0.1 (top panel) or changing the turbulence from vturb from 100 km s−1 to its maximum value, 5135 km s−1, set by the observed BAL width (bottom panel). These predictions show that the Ly α BAL identification would require log U < −2 and uniquely low column densities, log NH(cm−2) < 18.7, for a BAL outflow (see Section 3.5). The shaded regions around the curves for C iv, N v, and O vi in Fig. 6 illustrate their dependence on the specific values of NHI and the turbulent velocity, vturb, used in the calculations. The solid curves for these lines (and the solid and dashed curves for C iii λ977, Si iii λ1206, and Si iv λ1398) correspond to fiducial parameters log NHI(cm−2) = 15.35 and vturb = 1000 km s−1. In the top panel, the shaded regions show the range of optical depths corresponding to the maximum uncertainty of ±0.1 dex in our log NHI measurement (see Section 3.1, where the smaller/larger NHI value sets the lower/upper envelope to the shaded curves). The optical depth changes in response to these changes in NHI are non-linear at large U (and large NH) because of radiative shielding in the far-UV that affects each ion differently. These shielding effects maintain significant optical depths in the moderate ions C iv, N v, and O vi at a fixed NHI and large U. In other calculations (not shown), we determine that the threshold for important shielding effects at large U is roughly log NHI(cm−2) ≳ 14.8. This is 0.55 dex lower than our lower limit to the H i column density inferred from the observed BAL (Section 3.1), and therefore the predictions in Fig. 6 that result from shielding at large U should be applicable. Another parameter that can affect radiative shielding is the velocity dispersion inside the cloud, which we characterize by the turbulence velocity vturb. Large internal velocity dispersions can enhance the shielding by blending together numerous absorption lines in the far-UV, which in turn affects the ionization structure deep in the cloud. The shaded regions around the C iv, N v, and O vi curves in the bottom panel of Fig. 6 depict a range of results corresponding to vturb = 100 km s−1, which yields negligible line shielding (the bottom envelope of the shaded regions), up to vturb = 5135 km s−1, which yields the maximum line shielding allowed by the observed BAL width (the upper envelope of the shaded regions). We adopt an intermediate value vturb = 1000 km s−1 as our fiducial case (here and in all of our calculations below) because it yields conservatively small amounts of line shielding and, therefore, conservatively small predicted line optical depths in C iv, N v, and O vi at large U. The main result from Fig. 6 is that a Ly α BAL should be accompanied by significant broad C iv and O vi absorption lines over a wide range of normal BAL outflow physical conditions. It is important to keep in mind that the predicted strengths of the C iv and O vi lines are lower limits based on a lower limit on NHI that follows from the assumption of no partial covering. Smaller values of NHI are not allowed because they cannot produce the observed BAL attributed to Ly α. Larger NH requiring larger NH (at a given U) would produce stronger C iv and O vi absorption. The absence of these lines in the PDS 456 spectrum (Figs 2 and 3) means that the only conditions consistent with the Ly α BAL identification are low degrees of ionization, log U < −2, and very low total column densities, log NH(cm−2) < 18.7 that are ≳3.5 orders of magnitude lower than expected for BAL outflows. For comparison, analysis of composite BAL quasar spectra by Herbst et al. (in preparation) suggests that typical BAL outflows have log NH(cm−2) ≳ 22.2 (see also Sections 1 and 3.4). Another problem for the Ly α identification is that a Ly α-only BAL system would be, to our knowledge, unprecedented among observed BAL outflows. Fig. 3 shows that the C iv and O vi BALs expected to accompany Ly α in PDS 456 would need to be <33 per cent the strength of Ly α to avoid detection. This is very different from the typical situation illustrated by Fig. 5 and by other BAL composite spectra (Baskin et al. 2013; Herbst et al., in preparation) where the C iv and O vi lines are ∼2 to ∼3 times stronger than Ly α. BAL surveys based on C iv BAL detections (e.g. Trump et al. 2006; Gibson et al. 2009; Pâris et al. 2017) are naturally biased towards stronger C iv lines and they cannot detect Ly α-only BAL systems if they exist. However, our visual inspections of the 641 outflow quasar spectra in our BOSS sample (Section 3.4) do not find any instances of C iv BALs weaker than, or even comparable to, the corresponding Ly α absorption. A third argument against the Ly α BAL identification stems from the weakening of the observed BAL between the HST observations in 2000 and 2014 (Fig. 2). Variability in BALs and other outflow lines can be attributed to outflow clouds moving across our lines of sight or to changes in the ionization caused by changes in the incident ionizing flux (cf. Arav et al. 2012, 2015; Capellupo et al. 2013, 2014; Filiz Ak et al. 2013; Misawa, Charlton & Eracleous 2014a; Grier et al. 2015; Rogerson et al. 2016; Moravec et al. 2017, and references therein). The evidence for BAL saturation discussed in Sections 3.1 and 3.4 might favour crossing clouds in many situations. However, if the BAL changes in PDS 456 were caused by ionization changes and the ionization was initially low to produce Ly α absorption without accompanying C iv and O vi lines, then the only way to make Ly α weaker at a fixed NH is with larger ionization parameters that should produce stronger C iv and O vi absorption. The continued absence of these accompanying lines in 2014 (Fig. 2) thus makes the Ly α identification more difficult because the values of NH and U in 2000 would need to be several times lower than the already-low upper limits deduced from Fig. 6. 3.6 Problems with the N v λ1240 identification The top panel in Fig. 7 shows cloudy predictions for the line-centre optical depths that should accompany the observed BAL if it is N v λ1240. As in Fig. 6, the calculations assume a cloud velocity dispersion vturb = 1000 km s−1 (to include moderate amounts of line shielding) but line optical depths derived for b = 5135 km s−1. The total column densities in the model clouds are adjusted to yield an N v column density, $$\log N(\rm{N\,{\small V}}) ({\rm cm}^{-2}) = 15.62$$, fixed to the value inferred from the observed BAL (Section 3.1). The resulting optical depths are again only lower limits because $$N(\rm{N\,{\small V}})$$ is a lower limit. Figure 7. View largeDownload slide Line-centre optical depths (black curves) and total hydrogen column densities, log NH (red curve), versus ionization parameter, log U, predicted by cloudy to accompany the observed BAL (with τo = 0.35 and b = 5135 km s−1; Section 3.1) if it is attributed to N v (top panel) or C iv (bottom panel). The ionic column densities held fixed in the model clouds are shown in the upper left of each panel. These predictions and existing spectra (Figs 1–3) rule out low degrees of ionization for the N v BAL system, and they can guide future observations to test both BAL identifications further (see Sections 3.6 and 3.7). Figure 7. View largeDownload slide Line-centre optical depths (black curves) and total hydrogen column densities, log NH (red curve), versus ionization parameter, log U, predicted by cloudy to accompany the observed BAL (with τo = 0.35 and b = 5135 km s−1; Section 3.1) if it is attributed to N v (top panel) or C iv (bottom panel). The ionic column densities held fixed in the model clouds are shown in the upper left of each panel. These predictions and existing spectra (Figs 1–3) rule out low degrees of ionization for the N v BAL system, and they can guide future observations to test both BAL identifications further (see Sections 3.6 and 3.7). The observed BAL at 1346 Å is compatible with the N v identification only if the accompanying C iv line is undetectable. This requires a ratio of N v/C iv absorption-line strengths ≳3 (Section 3.1, Fig. 3). Fig. 7 shows that moderate-to-low outflow ionizations with log U ≲ 0, corresponding to total columns log NH(cm−2) ≲ 20.2, are firmly ruled out for the N v BAL identification by the predictions for strong C iv absorption. At higher ionizations, the predicted optical depth ratio hovers around $$\tau (\rm{N\,{\small V}})/\tau (\rm{C\,{\small IV}}) \sim 3$$–5 (Fig. 7), which is marginally consistent with the observed spectrum. However, any significant partial covering effects would push the observed N v/C iv line depth ratio towards unity, such that C iv should be detectable and the N v BAL identification is ruled out. Figs 4 and 5 show more directly that absorption-line ratios N v/C iv ≳3 needed for the N v BAL identification are, at best, extremely rare in observed quasar outflows. In particular, the composite spectrum (Fig. 5) shows that typical weak-to-moderate BAL/mini-BAL systems have a median line depth ratio of N v/C iv ∼ 1 ± 0.1. Individual outflows can have larger N v/C iv ratios. Measurements in the literature of BALs/mini-BALs at speeds up to nearly 0.2c reveal N v/C iv depth ratios in the range ∼0.5–1.5 (Steidel 1990; Jannuzi et al. 1996; Barlow & Sargent 1997; Hamann et al. 1997a; Hamann, Barlow & Junkkarinen 1997c; Telfer et al. 1998; Sabra et al. 2003; Rodríguez Hidalgo 2008; Rodríguez Hidalgo et al. 2011, 2013; Rogerson et al. 2016; Moravec et al. 2017). Our visual inspections of the 641 outflow quasars in Section 3.4 find crudely ∼2 per cent of cases with N v/C iv ≳ 2. Fig. 4 shows the most extreme well-measured examples. It is difficult to assess the N v/C iv line ratios quantitatively in these spectra due to blending problems in the Ly α forest. However, the most extreme cases3 shown in Fig. 4 also appear to be inconsistent with the N v BAL identification in PDS 456. Another constraint on the N v BAL identification is that it should be accompanied by strong O vi absorption at observed wavelength ∼1122 Å. This is based on both theoretical predictions (Fig. 7) and observations of other outflow quasars (Section 3.4 and references immediately above). The wavelength 1122 Å is just outside the wavelength coverage of the 2000 STIS spectrum; however, there is no evidence for a broad O vi line wing that could be measurable if this line was present (Fig. 3). The 2014 COS spectrum does cover these O vi wavelengths and clearly shows no signs of absorption there (Fig. 2), but the BAL at 1346 Å was also much weaker in this spectrum. Thus, the constraints provided by the absence of O vi in existing spectra are weak, but they also do not support the N v BAL identification. 3.7 The case for C iv λ1549 at 0.30c One argument favouring the C iv BAL identification in PDS 456 is that it is readily compatible with existing spectra (Figs 1–3). It has none of the problems described above for Ly α and N v because no other lines are expected within the wavelength coverage. Observations of high-velocity BALs/mini-BALs in other quasars indicate that the outflow ionizations are generally high and that the most prominent lines accompanying C iv should be N v and O vi (Sections 3.4 and 3.6; Hamann et al. 2013). The bottom panel in Fig. 7 shows specific theoretical predictions for the optical depths in these and other lines that should accompany the observed BAL if it is C iv. The most observationally accessible lines in this plot are Ly α at low ionizations and N v at high ionizations, both at predicted observer-frame wavelengths near ∼1068 Å. Future observations might test for this absorption, but that will be difficult due to the short wavelength and severe reddening (Section 2). The situation is worse for other corroborating lines. In particular, the corresponding Si iv λ1398 line would be at ∼1216 Å observed, which is directly on top of the Galactic damped Ly α line, and the O vi and Ne viii λ774 lines, which should be strong at high ionizations, are at inaccessible observed wavelengths (∼899 and ∼673 Å, respectively) due to Galactic Lyman limit absorption. A second circumstantial argument favouring the C iv identification is that its velocity v ≈ 0.30c is similar to the speeds measured for X-ray outflow of PDS 456. The X-ray outflow is highly variable with multiple velocity components, but the main component identified by Fe K-shell absorption has measured velocities in the range v ∼ 0.25–0.34c with a typical value near ∼0.3c (Reeves et al. 2014, 2016; Nardini et al. 2015; Matzeu et al. 2017). The C iv BAL might provide evidence for a physical relationship between the UV and X-ray outflows in PDS 456 (see Section 4 below). A third argument is that relativistic C iv BALs/mini-BALs at speeds approaching v ∼ 0.2c have already been measured in a growing number of luminous quasars (Jannuzi et al. 1996; Hamann et al. 1997b, 2013; Rodríguez Hidalgo 2008; Rodríguez Hidalgo et al. 2011; Rogerson et al. 2016; Rodriguez Hidalgo et al., in preparation). The C iv BAL at v ≈ 0.30c in PDS 456 would set a new speed record for UV outflows, but it is not so dramatic to be a paradigm shift for our understanding of these outflows. The width and depth of the BAL in PDS 456 are roughly similar to these other high-velocity outflow features. We also note that our tentative detection of another weak C iv BAL in PDS 456 at v ∼ 0.19c (Fig. 2, Section 3.2) provides additional, albeit tentative, evidence that a relativistic UV outflow is present in PDS 456. 4 SUMMARY AND DISCUSSION The UV spectrum of PDS 456 obtained with HST-STIS in 2000 has a distinct BAL at ∼1346 Å that we identify as C iv λ1549 at velocity shift v ≈ 0.30c, FWHM ≈8550 km s−1, and minimum optical depth τo = 0.35 (Section 3.1). The C iv identification rests on its compatibility with existing spectra and the absence of lines that should accompany the alternative identifications, Ly α λ1216 and N v λ1240. Ly α is compatible with the observed BAL only if the outflow has a low degree of ionization, log U < −2, and a very low total column density, log NH(cm−2) < 18.7, that would be unprecedented in BAL outflow studies, e.g. several orders of magnitude below recent estimates (Section 3.5). The N v identification might be consistent with the absence of accompanying C iv absorption if the gas is highly ionized and it pushes the boundary of observed N v/C iv line strengths beyond what we find in our comparison of 641 outflow quasars in BOSS (Section 3.6). However, this situation should produce strong O vi absorption, which is not well constrained in existing spectra of PDS 456 but it appears to be absent (Fig. 3). Thus, the N v identification is also strongly disfavoured. The C iv BAL identification has none of these problems (Section 3.7). It would mark the fastest UV outflow line ever reported, but its velocity is consistent with the X-ray outflow in PDS 456 (see Section 1 and below) and not dramatically different from the relativistic C iv BALs/mini-BALs already measured at speeds approaching ∼0.2c in other quasars (Jannuzi et al. 1996; Hamann et al. 1997b, 2013; Rodríguez Hidalgo 2008; Rodríguez Hidalgo et al. 2011; Rogerson et al. 2016; Rodriguez Hidalgo et al., in preparation). The C iv identification is also weakly supported by our tentative identification of an additional C iv BAL feature at v ∼ 0.19c (Fig. 2, Section 3.2). Broad UV outflow lines at speeds near 0.3c are surely very rare based on the rarity of such lines at ∼0.1–0.2c (see references above). However, the incidence of C iv BALs/mini-BALs at v > 0.2c is not known because they have not been searched for in large quasar surveys like SDSS/BOSS and, in any case, they could easily be missed due to blends with unrelated lines in the Ly α forest (e.g. see the search for P v λ1121 BALs by Capellupo et al. 2017). The location of the C iv BAL outflow in PDS 456 is a critical unknown. The range of ionizations consistent with C iv absorption might favour lower ionizations and larger distances from the black hole than the X-ray outflow, perhaps at radii of the order of ∼1 pc as inferred from some variability studies of C iv BALs/mini-BALs in other quasars (Hamann et al. 2013; Capellupo et al. 2014; Moravec et al. 2017; McGraw et al. 2018). However, larger distances do not necessarily produce lower ionizations, e.g. if acceleration causes the outflow densities to drop faster than the 1/r2 behaviour expected from free expansion at a constant speed. Moreover, the relativistic speed of the C iv BAL at v ∼ 0.30c indicates that the UV outflow originated with the X-ray UFO very close to the black hole. If the measured flow speeds are similar to the gravitational escape speed at the launch radius, then the launch point is roughly at r ∼ 20–30 rg (Nardini et al. 2015; Matzeu et al. 2017). Also note that a highly ionized X-ray outflow launched from this radius and expanding freely outwards into a fixed solid angle will not necessarily become less ionized at larger distances (for C iv absorption) because the 1/r2 dilution of the radiation field is balanced by a 1/r2 decline in the densities to yield a constant ionization parameter. Lower ionizations will occur if the inner regions of the outflow radiatively shield the material downstream and/or if there are clumps with enhanced densities relative to the ambient flow. Clumping and shielding can occur to the same effect at almost any radius, and there is already evidence for dense clumps in the X-ray outflow of PDS 456 (based on lower ionizations and partial covering in the soft X-ray absorber; Matzeu et al. 2016; Reeves et al. 2016; see also Gofford et al. 2014; Hagino et al. 2015; Nardini et al. 2015). It is therefore an intriguing possibility that the C iv BAL forms directly within, or in close proximity to, the relativistic X-ray outflow. Our cloudy simulations (Fig. 7) show that the C iv BAL in PDS 456 could form over a wide range of physical conditions, including very high ionizations where C iv is just a trace constituent. This situation is illustrated in Fig. 8, which plots the ionization structure in a single model cloud with log U = 1.7. This cloud reaches the observed minimum BAL optical depth $$\tau _{\rm o} (\rm{C\,{\small IV}}) = 0.35$$, along with $$\tau _{\rm o} (\rm{N\,{\small V}}) = 0.90$$ and $$\tau _{\rm o} (\rm{O\,\,{\small VI}}) = 43$$, at total column density log NH(cm−2) = 22.9 (represented by the unshaded left-hand portion of Fig. 8). This front portion of the cloud matches our calculations in the bottom panel of Fig. 7. In this environment, the C iv and N v ion fractions are everywhere ≲4 × 10−4 and ≲3 × 10−3, respectively, and the dominant form of oxygen is O vii. This highly ionized environment capable of C iv λ1549 absorption will also produce saturated absorption in higher ion lines such as Ne viii λ774 and Mg x λ615 (see also log U = 1.7 in Fig. 7) and at the bound–free edges of O vii and O viii in soft X-rays. Figure 8. View largeDownload slide Ionization fractions versus total hydrogen column density, log NH, in a single model cloud with ionization parameter, log U = 1.7. This cloud produces the minimum observed BAL optical depth in C iv $$\tau _{\rm o} (\rm{C\,{\small IV}}) = 0.35$$, at log NH(cm−2) ≈ 22.9, such that the unshaded left-hand region represents the model cloud with log U = 1.7 in Fig. 7. This environment (at log NH(cm−2) ≤ 22.9) is highly ionized and optically thick at the bound–free edges of O vii, O viii, and Ne viii, with only trace amounts of the lower ions C iv, N v, and O vi. The lower ions dominate at larger column densities (grey shaded region) where there is more radiative shielding, but these lower ionizations are not needed for the C iv BAL in PDS 456. Figure 8. View largeDownload slide Ionization fractions versus total hydrogen column density, log NH, in a single model cloud with ionization parameter, log U = 1.7. This cloud produces the minimum observed BAL optical depth in C iv $$\tau _{\rm o} (\rm{C\,{\small IV}}) = 0.35$$, at log NH(cm−2) ≈ 22.9, such that the unshaded left-hand region represents the model cloud with log U = 1.7 in Fig. 7. This environment (at log NH(cm−2) ≤ 22.9) is highly ionized and optically thick at the bound–free edges of O vii, O viii, and Ne viii, with only trace amounts of the lower ions C iv, N v, and O vi. The lower ions dominate at larger column densities (grey shaded region) where there is more radiative shielding, but these lower ionizations are not needed for the C iv BAL in PDS 456. Detailed comparisons between the UV and X-ray outflows in PDS 456 are beyond the scope of the present study. They are subject to uncertainties caused by the outflow variabilities and by unknowns in the shape of the ionizing spectrum and the spatial locations of different outflow components relative to the ionizing far-UV and X-ray emission sources. Here we note simply that recent estimates of the X-ray outflow ionization parameters4 are roughly in the range log U ∼ 4.3–4.8 for the K-shell outflow (Gofford et al. 2014; Nardini et al. 2015) and log U ∼ 2.8 for the soft X-ray absorber (Reeves et al. 2016). The ionization parameters needed for the C iv BAL, with upper limit log U ≲ 2.4 (Fig. 7), might occur in dense clumps embedded in the X-ray outflow if the density enhancements are ≳0.4 dex relative to the soft X-ray absorber or ≳2 dex relative to the Fe K-shell outflow. The ionization structure of the UV outflow might resemble the quasar SBS 1542+541 (redshift z ∼ 2.36), which has measured UV BALs ranging in ionization from C iv up to Mg x and Si xii λ510 (Telfer et al. 1998). The lines in this quasar appear saturated with shallow troughs that reveal ion-dependent line-of-sight covering fractions from ∼15 per cent in C iv to ∼50 per cent in the higher ions. Telfer et al. (1998) infer from this a two-zone outflow that could span a decade or more in ionization parameter, with the C iv lines forming in small clumps embedded in a more highly ionized outflow medium. Telfer et al. (1998) also note that this outflow should produce substantial bound–free absorption by O vii and O viii in soft X-rays. The general picture of clumpy outflows has become commonplace in quasar outflow studies. High-quality observations of UV outflow lines often provide evidence for clumpy multi-phase outflow structures with a range of covering fractions (Ganguly et al. 1999; de Kool, Korista & Arav 2002; Hamann & Sabra 2004; Arav et al. 2005, 2008; Gabel et al. 2005a; Misawa et al. 2007, 2014b; Leighly et al. 2009, 2011, 2015; Hamann et al. 2011; Moravec et al. 2017). Detailed studies of some bright Seyfert 1 galaxies (e.g. Kaspi et al. 2002; Netzer et al. 2003; Gabel et al. 2005b) clearly demonstrate that the UV and X-ray absorption features can form together in complex outflows, with indications that the lower ionization UV lines identify clumps or filaments embedded in the X-ray outflow. Clumpy outflow structures are also predicted by recent numerical simulations (Sim et al. 2010; Takeuchi, Ohsuga & Mineshige 2013; Waters et al. 2017) and by considerations of the radiative forces that can compress the outflows into small substructures (Stern et al. 2014). There are also theoretical arguments requiring dense clumps to moderate the outflow ionizations in the absence of significant radiative shielding (De Kool 1997; Hamann et al. 2013). If the C iv-absorbing gas in PDS 456 is indeed embedded in the X-ray outflow, it could have important implications for the outflow energetics. It is a well-known problem for X-ray UFOs (e.g. Tombesi et al. 2011, 2013; Gofford et al. 2015) that their high ionizations lead to low opacities and inefficient radiative acceleration (Gofford et al. 2013, 2014; Higginbottom et al. 2014). Harnessing the full radiative power of the quasar to drive these outflows might require opacities beyond electron scattering at UV/far-UV wavelengths, near the peak of the quasar spectral energy distributions. Matzeu et al. (2017) showed recently that the outflow speeds in PDS 456 correlate with the variable X-ray luminosity, consistent with radiative acceleration (see also Saez & Chartas 2011). If radiative forces are important, then dense clumps with lower ionizations embedded in the X-ray outflows might be important to boost the opacities for radiative driving (Laor & Davis 2014; Hagino et al. 2015). The C iv BAL at v ≈ 0.30c in PDS 456 could be the first direct observational evidence for this idea. PDS 456 is a remarkable object with powerful accretion-disc outflows launched from a range of at least two decades in disc radii, from ∼0.001 pc for the X-ray outflow to ∼0.3 pc for the blueshifted CIV broad emission line (Section 1; O'Brien et al. 2005; Gofford et al. 2014). More work is needed to understand how the UV BAL fits into this outflow environment and, specifically, to test the C iv BAL identification. A search for corroborating N v absorption at ∼1077 Å will be difficult due to Galactic reddening (Section 2) and the poor sensitivity of current instruments at these wavelengths (Section 3.7). However, the alternative BAL identifications, N v and Ly α, might be ruled out or confirmed more easily by searching for O vi absorption near 1122 Å (Section 3.6, although the BAL at 1346 Å needs to be present for a meaningful test). Our team has ongoing programmes to extend the UV wavelength coverage and monitor PDS 456 in the UV and X-rays that will be described in future papers. ACKNOWLEDGEMENTS We are grateful to Gary Ferland and the cloudy development team for their continued support and public dispersement of the spectral synthesis code cloudy. FH also thanks Nahum Arav, Chris Done, Gerald Kriss, and Paola Rodriguez Hidalgo for helpful discussions. We are grateful to an anonymous referee for helpful comments on the manuscript. JNR acknowledges the support via HST grant HST-GO-14477.001-A. EN acknowledges funding from the European Union's Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement no. 664931. Footnotes 1 Throughout this paper, we treat unresolved doublets as single lines, such that, for example, C iv λ1548,1551 becomes C ivλ1549 with a summed oscillator strength (for column density estimates). 2 We note, however, that the observed spectrum is poorly fitted at wavelengths ≳2400 Å (observed). This might be caused by reddening in the quasar environment, a break in the intrinsic/emitted spectral slope, and/or blended Fe ii emission lines that are known to be strong in PDS 456 (Simpson et al. 1999). We do not investigate this further because our goal is simply to define a continuum for analysis of the outflow lines at shorter wavelengths (Section 3). 3 We need to acknowledge here that hypothetical quasars with strong N v BALs but negligible C iv absorption cannot appear in our BOSS sample because it relies on C iv line detections (via AI and BI). 4 The ionization parameter quoted in the X-ray studies is ξ = Lion/nHr2 from Tarter, Tucker & Salpeter (1969), where Lion is the quasar luminosity from 1 to 1000 Ryd. For ξ in units of erg cm s−1 and the continuum shape used in our cloudy simulations (Section 3.3), the conversion is log U ≈ log ξ − 1.2. REFERENCES Arav N. et al.  , 2001, ApJ , 561, 118 https://doi.org/10.1086/323369 CrossRef Search ADS   Arav N., Kaastra J., Kriss G. A., Korista K. T., Gabel J., Proga D., 2005, ApJ , 620, 665 https://doi.org/10.1086/425560 CrossRef Search ADS   Arav N., Moe M., Costantini E., Korista K. T., Benn C., Ellison S., 2008, ApJ , 681, 954 https://doi.org/10.1086/588651 CrossRef Search ADS   Arav N. et al.  , 2012, A&A , 544, A33 CrossRef Search ADS   Arav N. et al.  , 2015, A&A , 577, A37 CrossRef Search ADS   Barlow T. A., Sargent W. L. W., 1997, AJ , 113, 136 https://doi.org/10.1086/118239 CrossRef Search ADS   Baskin A., Laor A., Hamann F., 2013, MNRAS , 432, 1525 https://doi.org/10.1093/mnras/stt582 CrossRef Search ADS   Baskin A., Laor A., Hamann F., 2015, MNRAS , 449, 1593 https://doi.org/10.1093/mnras/stv406 CrossRef Search ADS   Borguet B. C. J., Edmonds D., Arav N., Benn C., Chamberlain C., 2012, ApJ , 758, 69 https://doi.org/10.1088/0004-637X/758/1/69 CrossRef Search ADS   Capellupo D. M., Hamann F., Shields J. C., Halpern J. P., Barlow T. A., 2013, MNRAS , 429, 1872 https://doi.org/10.1093/mnras/sts427 CrossRef Search ADS   Capellupo D. M., Hamann F., Barlow T. A., 2014, MNRAS , 444, 1893 https://doi.org/10.1093/mnras/stu1502 CrossRef Search ADS   Capellupo D. M. et al.  , 2017, MNRAS , 469, 323 https://doi.org/10.1093/mnras/stx870 CrossRef Search ADS   Cardelli J. A., Clayton G. C., Mathis J. S., 1989, ApJ , 345, 245 https://doi.org/10.1086/167900 CrossRef Search ADS   Chartas G., Brandt W. N., Gallagher S. C., Garmire G. P., 2002, ApJ , 579, 169 https://doi.org/10.1086/342744 CrossRef Search ADS   Chartas G. et al.  , 2009, New Astron. Rev. , 53, 128 https://doi.org/10.1016/j.newar.2009.07.010 CrossRef Search ADS   Coatman L., Hewett P. C., Banerji M., Richards G. T., 2016, MNRAS , 461, 647 https://doi.org/10.1093/mnras/stw1360 CrossRef Search ADS   Crenshaw D. M., Kraemer S. B., George I. M., 2003, ARA&A , 41, 117 CrossRef Search ADS   Dawson K. S. et al.  , 2013, AJ , 145, 10 https://doi.org/10.1088/0004-6256/145/1/10 CrossRef Search ADS   De Kool M., 1997, in Arav N., Shlosman I., Weymann R. J., eds, ASP Conf. Ser. Vol. 128, Mass Ejection from Active Galactic Nuclei . Astron. Soc. Pac., San Francisco, p. 233 de Kool M., Korista K. T., Arav N., 2002, ApJ , 580, 54 https://doi.org/10.1086/343107 CrossRef Search ADS   Debuhr J., Quataert E., Ma C.-P., 2012, MNRAS , 420, 2221 https://doi.org/10.1111/j.1365-2966.2011.20187.x CrossRef Search ADS   Di Matteo T., Springel V., Hernquist L., 2005, Nature , 433, 604 https://doi.org/10.1038/nature03335 CrossRef Search ADS PubMed  Dietrich M., Hamann F., Shields J. C., Constantin A., Vestergaard M., Chaffee F., Foltz C. B., Junkkarinen V. T., 2002, ApJ , 581, 912 https://doi.org/10.1086/344410 CrossRef Search ADS   Eisenstein D. J. et al.  , 2011, AJ , 142, 72 https://doi.org/10.1088/0004-6256/142/3/72 CrossRef Search ADS   Ferland G. J. et al.  , 2013, Rev. Mex. Astron. Astrofis. , 49, 137 Ferland G. J. et al.  , 2017, Rev. Mex. Astron. Astrofis. , 53, 385 Filiz Ak N. et al.  , 2013, ApJ , 777, 168 https://doi.org/10.1088/0004-637X/777/2/168 CrossRef Search ADS   Gabel J. R. et al.  , 2005a, ApJ , 623, 85 https://doi.org/10.1086/428477 CrossRef Search ADS   Gabel J. R. et al.  , 2005b, ApJ , 631, 741 https://doi.org/10.1086/432682 CrossRef Search ADS   Ganguly R., Eracleous M., Charlton J. C., Churchill C. W., 1999, AJ , 117, 2594 https://doi.org/10.1086/300882 CrossRef Search ADS   Gibson R. R. et al.  , 2009, ApJ , 692, 758 https://doi.org/10.1088/0004-637X/692/1/758 CrossRef Search ADS   Gofford J., Reeves J. N., Tombesi F., Braito V., Turner T. J., Miller L., Cappi M., 2013, MNRAS , 430, 60 https://doi.org/10.1093/mnras/sts481 CrossRef Search ADS   Gofford J. et al.  , 2014, ApJ , 784, 77 https://doi.org/10.1088/0004-637X/784/1/77 CrossRef Search ADS   Gofford J., Reeves J. N., McLaughlin D. E., Braito V., Turner T. J., Tombesi F., Cappi M., 2015, MNRAS , 451, 4169 https://doi.org/10.1093/mnras/stv1207 CrossRef Search ADS   Grier C. J. et al.  , 2015, ApJ , 806, 111 https://doi.org/10.1088/0004-637X/806/1/111 CrossRef Search ADS   Hagino K., Odaka H., Done C., Gandhi P., Watanabe S., Sako M., Takahashi T., 2015, MNRAS , 446, 663 https://doi.org/10.1093/mnras/stu2095 CrossRef Search ADS   Hamann F., 1997, ApJS , 109, 279 https://doi.org/10.1086/312980 CrossRef Search ADS   Hamann F., 1998, ApJ , 500, 798 https://doi.org/10.1086/305776 CrossRef Search ADS   Hamann F., Ferland G., 1999, ARA&A , 37, 487 CrossRef Search ADS   Hamann F., Sabra B., 2004, in Richards G. T., Hall P. B., eds, ASP Conf. Ser. Vol. 311, AGN Physics with the Sloan Digital Sky Survey . Astron. Soc. Pac., San Francisco, p. 203 Hamann F., Barlow T., Cohen R. D., Junkkarinen V., Burbidge E. M., 1997a, in Arav N., Shlosman I., Weymann R. J., eds, ASP Conf. Ser. Vol. 128, Mass Ejection from Active Galactic Nuclei . Astron. Soc. Pac., San Francisco, p. 19 Hamann F., Barlow T. A., Junkkarinen V., Burbidge E. M., 1997b, ApJ , 478, 80 https://doi.org/10.1086/303781 CrossRef Search ADS   Hamann F., Barlow T. A., Junkkarinen V., 1997c, ApJ , 478, 87 https://doi.org/10.1086/303782 CrossRef Search ADS   Hamann F., Kanekar N., Prochaska J. X., Murphy M. T., Ellison S., Malec A. L., Milutinovic N., Ubachs W., 2011, MNRAS , 410, 1957 Hamann F., Chartas G., McGraw S., Rodriguez Hidalgo P., Shields J., Capellupo D., Charlton J., Eracleous M., 2013, MNRAS , 435, 133 https://doi.org/10.1093/mnras/stt1231 CrossRef Search ADS   Hamann F. et al.  , 2017, MNRAS , 464, 3431 https://doi.org/10.1093/mnras/stw2387 CrossRef Search ADS   Higginbottom N., Proga D., Knigge C., Long K. S., Matthews J. H., Sim S. A., 2014, ApJ , 789, 19 https://doi.org/10.1088/0004-637X/789/1/19 CrossRef Search ADS   Hopkins P. F., Elvis M., 2010, MNRAS , 401, 7 https://doi.org/10.1111/j.1365-2966.2009.15643.x CrossRef Search ADS   Hopkins P. F., Hernquist L., Cox T. J., Kereš D., 2008, ApJS , 175, 356 https://doi.org/10.1086/524362 CrossRef Search ADS   Jannuzi B. T. et al.  , 1996, ApJ , 470, L11 https://doi.org/10.1086/310301 CrossRef Search ADS   Kaspi S. et al.  , 2002, ApJ , 574, 643 https://doi.org/10.1086/341113 CrossRef Search ADS   Kaspi S., Maoz D., Netzer H., Peterson B. M., Vestergaard M., Jannuzi B. T., 2005, ApJ , 629, 61 https://doi.org/10.1086/431275 CrossRef Search ADS   Knigge C., Scaringi S., Goad M. R., Cottis C. E., 2008, MNRAS , 386, 1426 https://doi.org/10.1111/j.1365-2966.2008.13081.x CrossRef Search ADS   Korista K. T., Voit G. M., Morris S. L., Weymann R. J., 1993, ApJS , 88, 357 https://doi.org/10.1086/191825 CrossRef Search ADS   Laor A., Davis S. W., 2014, MNRAS , 438, 3024 https://doi.org/10.1093/mnras/stt2408 CrossRef Search ADS   Leighly K. M., 2004, ApJ , 611, 125 https://doi.org/10.1086/422089 CrossRef Search ADS   Leighly K. M., Halpern J. P., Jenkins E. B., Casebeer D., 2007, ApJS , 173, 1 https://doi.org/10.1086/519768 CrossRef Search ADS   Leighly K. M., Hamann F., Casebeer D. A., Grupe D., 2009, ApJ , 701, 176 https://doi.org/10.1088/0004-637X/701/1/176 CrossRef Search ADS   Leighly K. M., Dietrich M., Barber S., 2011, ApJ , 728, 94 https://doi.org/10.1088/0004-637X/728/2/94 CrossRef Search ADS   Leighly K. M., Cooper E., Grupe D., Terndrup D. M., Komossa S., 2015, ApJ , 809, L13 https://doi.org/10.1088/2041-8205/809/1/L13 CrossRef Search ADS   Luo B. et al.  , 2015, ApJ , 805, 122 https://doi.org/10.1088/0004-637X/805/2/122 CrossRef Search ADS   Matzeu G. A., Reeves J. N., Nardini E., Braito V., Costa M. T., Tombesi F., Gofford J., 2016, MNRAS , 458, 1311 https://doi.org/10.1093/mnras/stw354 CrossRef Search ADS   Matzeu G. A., Reeves J. N., Braito V., Nardini E., McLaughlin D. E., Lobban A. P., Tombesi F., Costa M. T., 2017, MNRAS , 472, L15 CrossRef Search ADS   McGraw S. M., Shields J. C., Hamann F. W., Capellupo D. M., Herbst H., 2018, MNRAS , 475, 585 CrossRef Search ADS   Misawa T., Eracleous M., Charlton J. C., Kashikawa N., 2007, ApJ , 660, 152 https://doi.org/10.1086/513097 CrossRef Search ADS   Misawa T., Charlton J. C., Eracleous M., 2014a, ApJ , 792, 77 https://doi.org/10.1088/0004-637X/792/1/77 CrossRef Search ADS   Misawa T., Inada N., Oguri M., Gandhi P., Horiuchi T., Koyamada S., Okamoto R., 2014b, ApJ , 794, L20 https://doi.org/10.1088/2041-8205/794/2/L20 CrossRef Search ADS   Moravec E. A., Hamann F., Capellupo D. M., McGraw S. M., Shields J. C., Rodríguez Hidalgo P., 2017, MNRAS , 468, 4539 https://doi.org/10.1093/mnras/stx775 CrossRef Search ADS   Nagao T., Marconi A., Maiolino R., 2006, A&A , 447, 157 CrossRef Search ADS   Nardini E. et al.  , 2015, Science , 347, 860 https://doi.org/10.1126/science.1259202 CrossRef Search ADS PubMed  Netzer H. et al.  , 2003, ApJ , 599, 933 https://doi.org/10.1086/379508 CrossRef Search ADS   O'Brien P. T., Reeves J. N., Simpson C., Ward M. J., 2005, MNRAS , 360, L25 https://doi.org/10.1111/j.1745-3933.2005.00038.x CrossRef Search ADS   Pâris I. et al.  , 2017, A&A , 597, A79 CrossRef Search ADS   Reeves J. N., O'Brien P. T., Ward M. J., 2003, ApJ , 593, L65 https://doi.org/10.1086/378218 CrossRef Search ADS   Reeves J. N. et al.  , 2009, ApJ , 701, 493 https://doi.org/10.1088/0004-637X/701/1/493 CrossRef Search ADS   Reeves J. N. et al.  , 2014, ApJ , 780, 45 https://doi.org/10.1088/0004-637X/780/1/45 CrossRef Search ADS   Reeves J. N., Braito V., Nardini E., Behar E., O'Brien P. T., Tombesi F., Turner T. J., Costa M. T., 2016, ApJ , 824, 20 https://doi.org/10.3847/0004-637X/824/1/20 CrossRef Search ADS   Richards G. T. et al.  , 2011, AJ , 141, 167 https://doi.org/10.1088/0004-6256/141/5/167 CrossRef Search ADS   Rodríguez Hidalgo P., 2008, PhD thesis , Univ. Florida, Gainesville. Available at: http://etd.fcla.edu/UF/UFE0022674/rodriguez_p.pdf Rodríguez Hidalgo P., Hamann F., Hall P., 2011, MNRAS , 411, 247 https://doi.org/10.1111/j.1365-2966.2010.17677.x CrossRef Search ADS   Rodríguez Hidalgo P., Eracleous M., Charlton J., Hamann F., Murphy M. T., Nestor D., 2013, ApJ , 775, 14 https://doi.org/10.1088/0004-637X/775/1/14 CrossRef Search ADS   Rogerson J. A., Hall P. B., Rodríguez Hidalgo P., Pirkola P., Brandt W. N., Filiz Ak N., 2016, MNRAS , 457, 405 https://doi.org/10.1093/mnras/stv3010 CrossRef Search ADS   Ross N. P. et al.  , 2012, ApJS , 199, 3 https://doi.org/10.1088/0067-0049/199/1/3 CrossRef Search ADS   Rupke D. S. N., Veilleux S., 2013, ApJ , 768, 75 https://doi.org/10.1088/0004-637X/768/1/75 CrossRef Search ADS   Sabra B. M., Hamann F., Jannuzi B. T., George I. M., Shields J. C., 2003, ApJ , 590, 66 https://doi.org/10.1086/375006 CrossRef Search ADS   Saez C., Chartas G., 2011, ApJ , 737, 91 https://doi.org/10.1088/0004-637X/737/2/91 CrossRef Search ADS   Sim S. A., Proga D., Miller L., Long K. S., Turner T. J., 2010, MNRAS , 408, 1396 https://doi.org/10.1111/j.1365-2966.2010.17215.x CrossRef Search ADS   Simpson C., Ward M., O'Brien P., Reeves J., 1999, MNRAS , 303, L23 https://doi.org/10.1046/j.1365-8711.1999.02344.x CrossRef Search ADS   Steffen A. T., Strateva I., Brandt W. N., Alexander D. M., Koekemoer A. M., Lehmer B. D., Schneider D. P., Vignali C., 2006, AJ , 131, 2826 https://doi.org/10.1086/503627 CrossRef Search ADS   Steidel C. C., 1990, ApJS , 74, 37 https://doi.org/10.1086/191493 CrossRef Search ADS   Stern J., Behar E., Laor A., Baskin A., Holczer T., 2014, MNRAS , 445, 3011 https://doi.org/10.1093/mnras/stu1960 CrossRef Search ADS   Strateva I. V., Brandt W. N., Schneider D. P., Vanden Berk D. G., Vignali C., 2005, AJ , 130, 387 https://doi.org/10.1086/431247 CrossRef Search ADS   Takeuchi S., Ohsuga K., Mineshige S., 2013, PASJ , 65, 88 https://doi.org/10.1093/pasj/65.4.88 CrossRef Search ADS   Tarter C. B., Tucker W. H., Salpeter E. E., 1969, ApJ , 156, 943 https://doi.org/10.1086/150026 CrossRef Search ADS   Telfer R. C., Kriss G. A., Zheng W., Davidsen A. F., Green R. F., 1998, ApJ , 509, 132 https://doi.org/10.1086/306491 CrossRef Search ADS   Tombesi F., Cappi M., Reeves J. N., Palumbo G. G. C., Yaqoob T., Braito V., Dadina M., 2010, A&A , 521, A57 CrossRef Search ADS   Tombesi F., Cappi M., Reeves J. N., Palumbo G. G. C., Braito V., Dadina M., 2011, ApJ , 742, 44 https://doi.org/10.1088/0004-637X/742/1/44 CrossRef Search ADS   Tombesi F., Cappi M., Reeves J. N., Braito V., 2012, MNRAS , 422, L1 https://doi.org/10.1111/j.1745-3933.2012.01221.x CrossRef Search ADS   Tombesi F., Cappi M., Reeves J. N., Nemmen R. S., Braito V., Gaspari M., Reynolds C. S., 2013, MNRAS , 430, 1102 https://doi.org/10.1093/mnras/sts692 CrossRef Search ADS   Torres C. A. O., Quast G. R., Coziol R., Jablonski F., de la Reza R., Lépine J. R. D., Gregório-Hetem J., 1997, ApJ , 488, L19 https://doi.org/10.1086/310913 CrossRef Search ADS   Trump J. R. et al.  , 2006, ApJS , 165, 1 https://doi.org/10.1086/503834 CrossRef Search ADS   Warner C., Hamann F., Dietrich M., 2004, ApJ , 608, 136 https://doi.org/10.1086/386325 CrossRef Search ADS   Waters T., Proga D., Dannen R., Kallman T. R., 2017, MNRAS , 467, 3160 Weymann R. J., Morris S. L., Foltz C. B., Hewett P. C., 1991, ApJ , 373, 23 https://doi.org/10.1086/170020 CrossRef Search ADS   Wu J. et al.  , 2011, ApJ , 736, 28 https://doi.org/10.1088/0004-637X/736/1/28 CrossRef Search ADS   Wu J., Brandt W. N., Anderson S. F., Diamond-Stanic A. M., Hall P. B., Plotkin R. M., Schneider D. P., Shemmer O., 2012, ApJ , 747, 10 https://doi.org/10.1088/0004-637X/747/1/10 CrossRef Search ADS   Zakamska N. L. et al.  , 2016, MNRAS , 459, 3144 https://doi.org/10.1093/mnras/stw718 CrossRef Search ADS   © 2018 The Author(s) Published by Oxford University Press on behalf of the Royal Astronomical Society

Journal

Monthly Notices of the Royal Astronomical SocietyOxford University Press

Published: May 1, 2018

DeepDyve is your personal research library

It’s your single place to instantly
that matters to you.

over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month Explore the DeepDyve Library Search Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly Organize Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place. Access Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals. Your journals are on DeepDyve Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more. All the latest content is available, no embargo periods. DeepDyve Freelancer DeepDyve Pro Price FREE$49/month
\$360/year

Save searches from
PubMed

Create lists to

Export lists, citations