A New Relativistic Component of the Accretion Disk Wind in PDS 456

Outflows are an important phenomenon in active galactic nuclei (AGNs) and can play a key role in the co-evolution of the massive black hole and the host galaxy (Di Matteo et al. 2005; King 2010). Black holes grow by accretion and strong nuclear outflows can quench this process by shutting off their supply of matter, thereby setting the M−σ relation that is seen today (Ferrarese & Merritt 2000; Gebhardt 2000). A number of high column density (N_H ∼ 10²³ cm⁻²), fast (∼0.1c) outflows have now been found in luminous AGNs (Tombesi et al. 2010; Gofford et al. 2013), through detections of blueshifted Fe K absorption, at rest-frame energies greater than 7 keV. These ultra-fast outflows may be the missing link in the galactic feedback process, by driving massive molecular outflows out to large (∼kpc) scales in galaxies (Feruglio et al. 2015; Tombesi et al. 2015).

A prototype ultra-fast outflow occurs in the nearby (z = 0.184) quasar, PDS 456. With a bolometric luminosity of ∼10⁴⁷ erg s⁻¹, PDS 456 is the most luminous QSO in the local universe (Torres et al. 1997; Simpson et al. 1999; Reeves et al. 2000) and the radio-quiet analog of 3C 273. However, PDS 456 is most notable for its powerful and fast (∼0.25c) X-ray wind. Indeed, since its initial detection in 2001 with XMM-Newton (Reeves et al. 2003), the presence of the ultra-fast outflow in PDS 456 has now been established through over a decade's worth of X-ray observations (Reeves et al. 2009; Behar et al. 2010; Gofford et al. 2014; Hagino et al. 2015; Nardini et al. 2015; Matzeu et al. 2016, 2017; Parker et al. 2018). Intriguingly, Hamann et al. (2018) recently claimed a fast UV counterpart to the X-ray wind.

In a series of five XMM-Newton and NuSTAR observations of PDS 456 in 2013–2014, Nardini et al. (2015) detected a persistent P-Cygni-like profile from highly ionized (He or H-like) iron, blueshifted to 9 keV in the quasar rest frame. The broad P-Cygni profile established the wide-angle character of the outflow, while the wind variability provided a robust estimate of the wind radial distance, on the scale of the AGN accretion disk. From this, the large mass outflow rate inferred, of ∼10M_⊙ yr⁻¹, implied that the wind power is at least 15% of the Eddington luminosity. This is more than sufficient to provide the feedback required by models of black hole and host galaxy co-evolution (Hopkins & Elvis 2010), which likely plays a critical role in black hole growth and feedback in the early universe.

PDS 456 was subsequently observed with NuSTAR from 2017 March 23 to 26, with a total duration of 305 ks. This coincided with two simultaneous XMM-Newton observations, hereafter OBS 1 and OBS 2, taken over two consecutive satellite orbits (see Table 1) and both in Large Window mode for EPIC-pn. All data were processed using the nustardas v1.7.1, XMM-Newton sas v16.0, and heasoft v6.20 software. NuSTAR source spectra were extracted using a 50'' circular region centered on the source and background from a 65'' circular region clear from stray light. XMM-Newton EPIC-pn spectra were extracted from single and double events, using a 30'' source region and 2 × 34'' background regions on the same chip. All spectra are binned to at least 50 counts per bin, while the background rates are <10% of the net source rates. The 3–40 keV NuSTAR light curve is shown in Figure 1, where OBS 1 coincided with a pronounced dip in the source count rate ∼50 ks into the NuSTAR observation.

Zoom In Zoom Out Reset image size

Table 1.  The 2017 PDS 456 Observations and Outflow Parameters

	Mean	OBS 1	OBS 2
Observations
Telescope	NuSTAR	XMM-Newton	XMM-Newton
Detector	FPMA+FPMB	EPIC-pn	EPIC-pn
OBSID	60201020002	0780690201	0780690301
Start date	2017/03/23	2017/03/23	2017/03/25
Start time	05:31:09	19:25:01	06:27:09
Exposurea	157.0	39.6	64.9
Net rate (s−1)b	0.162 ± 0.002	0.774 ± 0.004	1.189 ± 0.004
Slower Zone
NHc	${4.2}_{-1.1}^{+1.3}$	${3.7}_{-1.1}^{+1.3}$	${3.9}_{-0.9}^{+1.0}$
$\mathrm{log}\xi$ d	5.5f	${5.5}_{-0.2}^{+0.3}$	5.5f
v/c	−0.25 ± 0.02	−0.21 ± 0.02	−0.27 ± 0.02
Δχ2	80.3	24.0	42.2
Faster Zone
NHc	${2.7}_{-1.2}^{+1.0}$	${5.0}_{-1.7}^{+1.4}$	${2.9}_{-1.1}^{+1.4}$
$\mathrm{log}\xi$ d	5.5f	5.5f	5.5f
v/c	−0.46 ± 0.02	−0.43 ± 0.02	−0.43f
Δχ2	29.1	23.6	10.4
Continuum
Γ	${2.18}_{-0.05}^{+0.08}$	2.41 ± 0.09	2.41f
F2–10 keVe	2.58	1.84	2.54
${\chi }_{\nu }^{2}$	177.9/161	133.5/148	246.5/255

Notes.

^aNet exposure after background screening and deadtime correction, in ks. ^bNet count rates, over 3–40 keV for NuSTAR and 0.4–10 keV for XMM-Newton. ^cUnits of column density ×10²³ cm⁻². ^dIonization parameter (where ξ = L/nR²) in units of erg cm s⁻¹. ^eObserved 2–10 keV flux in units of ×10⁻¹² erg cm⁻² s⁻¹. ^fDenotes parameter is tied.

Download table as:  ASCIITypeset image

Note that outflow velocities are given in the rest frame of PDS 456 at z = 0.184, after correcting for relativistic Doppler shifts. Errors are quoted at 90% confidence for one interesting parameter (or Δχ² = 2.7).

Initially we analyzed the time-averaged NuSTAR spectrum. The count-rate spectra from the FPMA+FPMB modules are shown in Figure 1, where the cross normalizations of FPMA and FPMB agree to within ±5% of each other. This is compared to a power law of a photon index of Γ = 2.3 and corrected for Galactic absorption, where N_H = 2.4 ×10²¹ cm⁻² (Kalberla et al. 2005). PDS 456 has a low flux throughout the 2017 observations, where the mean 2–10 keV flux of 2.6 × 10⁻¹² erg cm⁻² s⁻¹ is at the low end of the range measured in previous observations by Nardini et al. (2015) and Matzeu et al. (2017).

While the hard X-ray continuum is well described by a power law, the overall fit is very poor, with ${\chi }_{\nu }^{2}=327.1/169$, rejected with a null hypothesis probability of P_N = 3.6 ×10⁻¹². Two strong absorption profiles are present in the residuals of both detectors between 8 and 12 keV. Modeling these with Gaussian profiles gave rest-frame centroid energies of E = 8.93 ± 0.15 keV and E = 11.4 ± 0.3 keV, with equivalent widths of EW = −430 ± 80 eV and EW =−380 ± 100 eV, respectively. The addition of both lines significantly improved the fit by Δχ² = −61.7 and Δχ² = −39.3. The lower-energy line is consistent with the energy of the Fe K absorption profile measured previously in PDS 456 (Nardini et al. 2015; Matzeu et al. 2017); however, the second line appears at a substantially higher energy. The lines are broadened and were fitted with a common velocity width, of ${\sigma }_{v}={{\rm{20,000}}}_{-4000}^{+8000}\,\,\mathrm{km}\,{{\rm{s}}}^{-1}$, corresponding to $\sigma \,={600}_{-130}^{+250}\,\mathrm{eV}$ at 8.93 keV, consistent with the width measured by Nardini et al. (2015) previously.

If we associate the two lines with the Lyα and Lyβ transitions from H-like Fe at 6.97 keV and 8.27 keV, respectively, then the inferred outflow velocities are inconsistent, with v_out = −0.24 ± 0.02c and v_out = −0.31 ± 0.02c. Note it would also be unusual for a higher-order line to have such a high equivalent width as is observed here. Alternatively, the higher-energy feature may be associated with an Fe K absorption edge. This gave a threshold energy of E = 10.7 ± 0.2 keV, resulting in an acceptable fit. However, the velocity inferred from the edge is also inconsistent, e.g., for H-like Fe, a K-shell edge (at 9.27 keV) blueshifted to 10.7 keV gives v_out = −0.14 ± 0.02c, while the Fe xxvi Lyα line (at 6.97 keV) blueshifted to 8.9 keV gives v_out = −0.24 ± 0.02c. The velocities are also inconsistent if the two features instead arise from He-like iron.

Given the lack of a plausible identification at a self-consistent velocity, the two absorption lines are likely to arise from two outflowing systems with different velocities. If they are both associated with Fe xxvi Lyα, then the outflow velocities are v_out = −0.24 ± 0.02c and v_out = −0.46 ± 0.03c (with slightly higher velocities inferred for He-like Fe). Thus, while the lower-energy line is consistent with the outflow velocities usually measured at iron K in PDS 456 (Nardini et al. 2015; Matzeu et al. 2017), which are typically 0.25–0.3c, the higher-energy feature may originate from a new, very fast component of the wind.

3.1. Photoionization Modeling

To test this, we fitted the NuSTAR spectrum with a self-consistent xstar photoionization model, which accounts for any weak higher-order lines and edges in addition to the strong $1s\to 2p$ resonance absorption. We adopt the same absorption model grids used in Nardini et al. (2015), where the optical to X-ray SED of PDS 456 was used as the input continuum, which has an ionizing (1–1000 Ryd) luminosity of L_ion = 5 × 10⁴⁶ erg s⁻¹. A turbulence velocity width of 15,000 km s⁻¹ was used, consistent with the Gaussian line width. Solar abundances of Grevesse & Sauval (1998) were used throughout. As the FPMA and FPMB spectra were consistent within errors, these were combined using mathpha into a single mean NuSTAR spectrum to maximize S/N. The response files were combined using equal weightings for both modules, while the source to background area (backscal) scaling factors were propagated through to the combined spectrum.

The best-fit xstar model fitted to the mean PDS 456 spectrum is shown in Figure 2(a). Two outflowing absorption zones are required, one slower zone with v_out = −0.25 ± 0.02c and the faster zone with v_out = −0.46 ± 0.02c; see Table 1 for details. These velocities are consistent with the Gaussian analysis and the high ionization parameter of the absorbers ($\mathrm{log}\xi =5.5$) is consistent with the absorption lines arising from H-like iron (Fe xxvi Lyα). Figure 2(b) shows the residuals against a model including only the slower −0.25c zone, which leaves significant residuals around 11 keV and can then only be modeled by the fast 0.46c zone (see Figure 2(c)). Indeed, both absorption zones are required at >99.99% confidence (with Δχ² = −80.3, slow zone and Δχ² = −29.1, fast zone), resulting in an acceptable fit statistic of ${\chi }_{\nu }^{2}=177.9/161$. As a final consistency check, we attempted to model the high-energy feature with a lower ionization partial coverer, having the same velocity as the −0.25c zone, but where the K-shell edge is blueshifted to higher energies. This single velocity absorber is rejected as the fit statistic is worse by Δχ² = 28.2 (for Δν = 2) compared to the two-velocity model.

Zoom In Zoom Out Reset image size

The spectrum shows significant excess emission, observed redward of the absorption lines between 6 and 8 keV (see Figure 2). This was first measured by Nardini et al. (2015), who resolved the broad P-Cygni profile at Fe K from PDS 456. As per Nardini et al. (2015), the emission was modeled with an additive xstar emission grid and convolved with a Gaussian profile of width $\sigma ={0.6}_{-0.2}^{+0.3}\,\mathrm{keV}$ (or σ_v ∼ 24,000 km s⁻¹). The high equivalent width of the emission, with EW ∼ 350 eV, is consistent with the wide angle wind characterized by Nardini et al. (2015).

3.2. The XMM-Newton Observations

The spectra obtained from the two XMM-Newton observations, OBS 1 and OBS 2, were also analyzed. Simultaneous NuSTAR spectra were extracted using identical good time intervals and then fitted jointly with their corresponding XMM-Newton spectra. The combined FPMA and FPMB spectra were used after first checking that the two modules were consistent. During the OBS 1 spectrum, PDS 456 was caught at an even lower flux (F_{2–10 keV} = 1.8 × 10⁻¹² erg cm⁻² s⁻¹) coincident with the dip in the light curve. Note this may correspond to a short-lived absorption event, which will be discussed in a forthcoming paper on the broadband spectra.

The best-fit spectrum from OBS 1 is shown in Figure 3 and is well fitted by the dual velocity absorber. Both absorption lines are detected at >99.99% confidence, while the equivalent width of the high-energy line is slightly stronger than the mean value, with EW = −530 ± 150 eV. Note that the high-energy feature is also independently detected in both the NuSTAR and XMM-Newton spectra, as the bandpass of the latter extends to 12 keV in the QSO rest frame. For ease of comparison, we assumed a constant ionization across the mean, OBS 1, and OBS 2 spectra; see Table 1 for parameters. The column density of the fast zone is slightly higher in OBS 1 (${N}_{{\rm{H}}}={5.0}_{-1.7}^{+1.4}\times {10}^{23}\,{\mathrm{cm}}^{-2}$), when compared to the mean (${N}_{{\rm{H}}}={2.7}_{-1.2}^{+1.0}\times {10}^{23}\,{\mathrm{cm}}^{-2}$), to account for the increased depth of the high-energy feature. The OBS 2 spectrum is consistent with the mean, and the faster zone appears somewhat weaker compared to the lowest flux OBS 1 spectrum.

Zoom In Zoom Out Reset image size

Low-flux observations of PDS 456, obtained with NuSTAR and XMM-Newton in 2017 March, have revealed a new relativistic component of the fast wind in this quasar, with v = −0.46 ± 0.02c. Compared to studies of fast winds in other local, predominantly radio-quiet AGNs (Tombesi et al. 2010; Gofford et al. 2013), this is the fastest wind detected to date in a nearby AGN. Perhaps the only other AGNs with such an extreme fast wind is the high redshift (z = 3.9) BAL QSO, APM 08279+5255, where the outflow velocity may reach ∼−0.7c (Chartas et al. 2009; Saez et al. 2009; Saez & Chartas 2011). Note that Hagino et al. (2017) subsequently claimed the outflow velocity in APM 08279+5255 may be somewhat lower, with v ∼ 0.1–0.2c.

Here, the high-velocity absorber appears considerably faster than all of the previous Fe K wind measurements in PDS 456. Matzeu et al. (2017) studied all 12 previous X-ray observations of PDS 456, observed with either XMM-Newton, NuSTAR, or Suzaku. They showed the wind velocity in PDS 456 varies within the range 0.24c–0.34c, while the strong positive correlation observed between the outflow velocity and X-ray luminosity was interpreted as possible evidence for a radiatively driven wind. The mean X-ray luminosity in 2017, of L_{2–10 keV} = 2.6 × 10⁴⁴ erg s⁻¹, lies at the low-luminosity end of the range observed by Matzeu et al. (2017), with L_{2–10 keV} = 2.8–10.5 × 10⁴⁴ erg s⁻¹ and the velocity of the slower zone, where v ∼ −0.25c is consistent with the trend between velocity and luminosity.

The fast, ∼−0.45c component of the wind is not apparent in the past observations of PDS 456 (e.g., Nardini et al. 2015; and see Matzeu et al. 2017; Figure 1). To confirm this, we checked all the previous NuSTAR observations of PDS 456 for any significant higher-energy absorption above 10 keV, in addition to the persistent, but slower, −0.25c wind component. However, aside from the 2017 observation, none was found. The most stringent constraint comes from the first NuSTAR observation of PDS 456 in 2013 (hereafter OBS A; Nardini et al. 2015), where an upper limit of EW < 80 eV is placed on any Gaussian absorption line profile above 10 keV. Here, the 2–10 keV X-ray luminosity is 8 × 10⁴⁴ erg s⁻¹, more than 3× higher than in 2017. The upper limit on the column density of the fast zone during this high-luminosity observation is N_H < 1.1 × 10²³ cm⁻² for a given ionization of $\mathrm{log}\xi =5.5$, significantly lower than for the low-flux 2017 observations (see Table 1).

The fast component may arise from an inner streamline of an accretion disk wind (Proga & Kallman 2004; Sim et al. 2010), launched from very close to the black hole. Magnetohydrodynamical mechanisms are also capable of accelerating winds up to these velocities, especially if the illuminating X-ray continuum is steep (Fukumura et al. 2010), as is the case here with Γ > 2. For a radiatively accelerated wind, the launch radius is

$$\begin{eqnarray}&&{R}_{{\rm{w}}}\approx 2\left(\alpha \displaystyle \frac{L}{{L}_{\mathrm{Edd}}}-1\right){\left(\displaystyle \frac{{v}_{\infty }}{c}\right)}^{-2},\end{eqnarray} \tag{ 1 }$$

where R_w is the wind launch radius in gravitational units (R_G), ${v}_{\infty }$ is the terminal velocity, and α is the force multiplier factor. In PDS 456, with L/L_Edd = 1, ${v}_{\infty }=-0.45c$ and for a modest multiplier of α = 2, then R_w ∼ 10 R_G (or ∼10¹⁵ cm for a black hole mass of 10⁹ M_⊙). The maximum likely radial distance of the absorber can be derived under the assumption that ΔR/R < 1, for a given wind streamline. For an ionizing luminosity of L_ion = 5 × 10⁴⁶ erg s⁻¹, N_H = 5 × 10²³ cm⁻² and $\mathrm{log}\xi =5.5$, then R < L_ion/N_Hξ < 10¹⁷ cm. This is consistent with the radial distance of R ∼ 10¹⁶ cm estimated by Nardini et al. (2015), from the wind variability timescale in PDS 456.

If the fast wind component is launched from close to the black hole and is fully exposed to the central X-ray source, then its ionization state will likely be very high. At these distances and for iron not to become fully ionized (where $\mathrm{log}\xi \lt 6$), this requires densities of n ∼ 10⁷–10¹¹ cm⁻³ and which could be associated to matter lifted off the surface of the inner accretion disk. Given that this very fast wind component is detected during a low-flux observation of PDS 456, while it appears absent at higher fluxes (such as during OBS A in 2013), this may suggest that an inner wind zone would only be detected when its ionization state is low enough for the gas to not be fully ionized. Alternatively, the innermost wind may be shielded by denser gas near the launch point, which could arise from the high column partial covering gas often seen in the lower-flux X-ray spectra of PDS 456 (Matzeu et al. 2016).

Nonetheless, the fast zone may have important implications for the overall wind energetics. For the −0.25c wind in PDS 456, Nardini et al. (2015) estimated the kinetic power and mass outflow rate to be ∼15% and ∼50% of Eddington, respectively. As the power goes as ${\dot{E}}_{\mathrm{kin}}\propto \dot{m}{v}^{2}\propto {v}^{3}$, then the −0.45c zone could require the wind power to be a factor of 6× higher. Then the total kinetic power could reach Eddington, if the overall mass outflow rate of the fastest zone is similar to the slower one. In reality, it may be likely that we are observing a structured wind, with multiple velocity components, which are launched at different disk radii. Future high-resolution X-ray calorimeter observations, with XARM and Athena, will be able to further reveal the velocity structure in high-velocity winds such as PDS 456.

J.R. acknowledges financial support through grants NNX17AC38G, NNX17AD56G, and HST-GO-14477.001-A. A.L. acknowledges support via the STFC consolidated grant ST/K001000/1. E.N. is funded by the EU Horizon 2020 Marie Sklodowska-Curie grant No. 664931. Based on observations obtained with XMM-Newton, an ESA science mission with instruments and contributions directly funded by ESA Member States and NASA.