Black hole feedback in the luminous quasar PDS 456

See allHide authors and affiliations

Science  20 Feb 2015:
Vol. 347, Issue 6224, pp. 860-863
DOI: 10.1126/science.1259202

Finding the necessary negative feedback

The evolution of galaxies seems to be tied to the growth of the supermassive black holes at their centers, but it's not entirely clear why. Models have suggested a mechanism in which the growth of the black hole results in an outflow of gas that interrupts star formation. However, evidence for enough of this negative feedback has been lacking. Nardini et al. now see a signature in x-ray spectra of a strong persistent outflow in the quasar PDS 456. They estimate a broad solid angle spanned by the wind that enables a far greater impact on the host galaxy than narrower jet outflows.

Science, this issue p. 860


The evolution of galaxies is connected to the growth of supermassive black holes in their centers. During the quasar phase, a huge luminosity is released as matter falls onto the black hole, and radiation-driven winds can transfer most of this energy back to the host galaxy. Over five different epochs, we detected the signatures of a nearly spherical stream of highly ionized gas in the broadband x-ray spectra of the luminous quasar PDS 456. This persistent wind is expelled at relativistic speeds from the inner accretion disk, and its wide aperture suggests an effective coupling with the ambient gas. The outflow’s kinetic power larger than 1046 ergs per second is enough to provide the feedback required by models of black hole and host galaxy coevolution.

Disk winds are theoretically expected as a natural consequence of highly efficient accretion onto supermassive black holes (1), as the energy radiated in this process might easily exceed the local binding energy of the gas. In the past few years, black hole winds with column densities of ~1023 cm2 and velocities of ~0.1 times the speed of light (c) have been revealed in a growing number of nearby active galactic nuclei (AGN) through blueshifted x-ray absorption lines (2, 3). Outflows of this kind are commonly believed to affect the dynamical and physical properties of the gas in the host galaxy, and, hence, its star formation history (4). However, a complete observational characterization of how this feedback works is still missing. On its own, the detection of narrow, blueshifted features does not convey any information about the opening angle or the ejection site of the wind. This knowledge is critical for measuring the total power carried by the outflow, whose actual influence on galactic scales remains unclear (5).

The nearby (z = 0.184) radio-quiet quasar PDS 456 is an established Rosetta stone for studying disk winds (68). With a bolometric luminosity Lbol ~ 1047 erg/s and a mass of the central black hole on the order of 109 solar masses (MSun) (9), it is an exceptionally luminous AGN in the local universe and might be regarded as a counterpart of the accreting supermassive black holes during the peak of quasar activity at high redshift. Since the earliest x-ray observations, PDS 456 has regularly exhibited a deep absorption trough at rest-frame energies above 7 keV (6), which was occasionally resolved with high statistical significance into a pair of absorption lines at ~9.09 and 9.64 keV (7). Because no strong atomic transitions from cosmically abundant elements correspond to these energies, such lines are most likely associated with resonant K-shell absorption from Fe XXV Heα (6.7 keV) and Fe XXVI Lyα (6.97 keV) in a wind with an outflow velocity of ~0.3 c.

The X-ray Multi-Mirror Mission (XMM)–Newton and Nuclear Spectroscopic Telescope Array (NuSTAR) satellites simultaneously observed PDS 456 on four occasions in 2013, between 27 August and 21 September. A fifth observation was performed several months later, on 26 February 2014 (table S1). The entire campaign caught the quasar in widely different spectral states (Fig. 1). The broad Fe-K absorption trough is conspicuous at all times against a simple baseline continuum, with an equivalent width increasing in absolute value by a factor of ~4, from 75 ± 45 eV to 350 ± 60 eV (table S2; uncertainties are given at the 90% confidence level). The large blueshift to ~9 keV invariably pushes this broad feature close to the edge of the XMM-Newton energy band, preventing any evaluation of the intrinsic flux blueward of the absorption complex. The addition of the NuSTAR broadband spectra enabled the accurate determination of the intensity and slope of the high-energy continuum that was lacking before. This unveils a broadened Fe Kα emission component, which, combined with the adjacent absorption trough, gives rise to an unmistakable P-Cygni–like profile, as produced by an expanding gaseous envelope.

Fig. 1 Broadband x-ray spectra of PDS 456.

The spectra of the five XMM-Newton and NuSTAR observations (±1 SD error bars) are plotted in flux density units after factoring out the effects of the instrumental response on the raw photon count rate. All the spectra were rebinned for display purposes only, with the data from the two NuSTAR modules merged and shown in a lighter tone. (A) Observation (Obs.). 1 (black) and Obs. 2 (red). (B) Obs. 2 (shaded) and Obs. 3 (blue). (C) Obs. 3 (shaded) and Obs. 4 (green). (D) Obs. 4 (shaded) and Obs. 5 (purple). Each panel also contains the spectrum of the previous observation to highlight the extent of the variability.

Despite the overall variability, this characteristic outflow signature is clearly detected in each of the five epochs during the course of the campaign (Fig. 2), demonstrating the persistence of a wide-angle accretion disk wind in PDS 456. For a preliminary evaluation of its basic parameters, we first deconvolved the P-Cygni profile with a pair of Gaussian line shapes. For Doppler broadening relative to a uniform velocity field within a radial outflow, the profile width would suggest an aperture of θ ~ 100 degrees (fig. S1). In reality, this is unlikely to be a simple projection effect, as we might be observing the wind’s acceleration region. We therefore applied a custom P-Cygni model to the 2 to 30 keV energy range, only allowing for an absorption-induced continuum curvature (9, 10). The success of this attempt implies that the line’s profile is indeed compatible with a quasispherical (fully covering) outflow with terminal velocity of 0.35 ± 0.02 c (Fig. 3).

Fig. 2 Persistence of the P-Cygni–like feature.

The ratio of the observed emission over the continuum, which was modeled as a partially absorbed power law to reproduce the overall spectral curvature, is shown for XMM-Newton data (in black; ±1 SD error bars) and both NuSTAR modules (superimposed as green and turquoise dots). The P-Cygni–like profile is evident in each snapshot of the campaign, irrespective of the different flux and spectral states of the source. The peak of Fe Kα emission from the wind lies above 7 keV in each observation, and the absorption trough is centered around 9 keV. The line’s profile can be resolved independently at any epoch, with a full width at half-maximum for both components of ~900 eV (or 30,000 km/s at 9 keV). The vertical dotted line marks the rest-frame energy (6.97 keV) of the Fe XXVI Kα transition. (A) Obs. 1. (B) Obs. 2. (C) Obs. 3. (D) Obs. 4. (E) Obs. 5.

Fig. 3 Fit with a P-Cygni line model.

Adopting the same baseline continuum of Fig. 2 (red curve), we fitted the emission and absorption residuals characterizing the Fe-K band by means of a self-consistent P-Cygni profile from a spherically symmetric outflow (green curve). The results are shown for the merged Obs. 3 and Obs. 4, which are separated by only 3 days and are virtually indistinguishable at 2 to 30 keV (Fig. 1). The two NuSTAR modules were combined into a single spectrum (plotted in blue; ±1 SD error bars) for display purposes only. The inset contains a graphical explanation of the key parameters of this model: the characteristic energy Ec, corresponding to the onset of the absorption component, and the wind terminal velocity v = 0.35 ± 0.02 c, which can be regarded as a measure of the actual outflowing speed of the gas. The bottom panel shows the ratio between the data and the best-fit model. The residual structures above 10 keV are due to the Kβ and K edge absorption features from Fe XXVI. These are not included in the P-Cygni model but are detected with high significance (table S2) and remove any ambiguity in the identification of the ionic species.

We subsequently performed a comprehensive spectral and timing analysis, fitting the ~0.5 to 50 keV spectra from all five observations with a model that includes the standard power-law continuum and self-consistent absorption and emission components from photo-ionized gas (fig. S2). This model provides a very good description of the data, with no obvious residuals over the whole energy range (fig. S3). The main physical parameters obtained from the best fit include the hydrogen column density of the highly ionized absorber, Embedded Image, and its bulk outflow velocity, vout = 0.25 ± 0.01 c, constant over the five epochs (table S3). Partial covering by less-ionized gas accounts for most of the continuum spectral variability, which indicates clumpy obscuration within the same stream as observed in lower-luminosity active galaxies (11).

However, the general validity of the disk wind picture is still disputed. It has been proposed that blueshifted absorption might also arise from corotating optically thick plasma blanketing the accretion flow, which would be seen in x-rays reflected off the disk surface (12). Depending on the exact geometry, the extreme velocities inherent to the inner disk could produce a Fe K-shell feature anywhere between 4 and 10 keV through relativistic Doppler shifts. Previously applied to PG 1211+143, another bright quasar where a similar line complex was revealed (13), this scenario calls for a reflection-dominated x-ray spectrum. In PDS 456, this model clearly underpredicts the depth of the 9-keV absorption trough, whose energy and profile allow us to rule out this alternative interpretation (figs. S4 and S5).

Even when blueshifted lines can be safely associated with an outflow, the measure of its mass-loss rate and total energetics is subject to large uncertainties. Remarkably, these observations supply a robust estimate of the solid angle Ω filled by the wind, which is computed from the amount of absorbed ionizing radiation that is re-emitted across the spectrum. The average value is Ω = 3.2 ± 0.6π sr, and each individual observation is consistent with Ω exceeding 2π sr (9). No explicit information on Ω for individual sources existed previously, because the absorption components detected in both ultraviolet and x-ray spectra only probe the line of sight. Consequently, this quantity has thus far been assumed rather than directly determined from the data—for example, following the occurrence frequency of outflows amongst local AGN (14) or through an a priori selection of a reasonable wind geometry (15).

Besides the degree of collimation, critical information is required on the starting point of the wind (Rin), for which only indirect arguments are usually available (16). In PDS 456, the gas location is directly constrained by changes of the wind emission intensity in response to the hard x-ray continuum. The most significant variation, when the Fe K line follows a decrease by a factor of ~3 of the 7- to 30-keV flux, takes place between the first two observations, separated by ~10 days. The corresponding light-crossing time translates into a maximum distance of a few hundreds of gravitational radii (rg = GM/c2, where G is the gravitational constant and M is the black hole’s mass) from the illuminating source. This spatial extent is consistent with the degree of ionization of the gas and with predictions of hydrodynamical models of radiation-driven accretion disk winds (17). An independent corroboration of the above estimate comes from the variability time scale of ~1 week of the Fe-K absorption feature, probed during the monitoring of PDS 456 by the Suzaku x-ray satellite in early 2013 (18). Indeed, the historical behavior of the K-shell absorption line(s) appears to be in keeping with a persistent wind where the gas is in photo-ionization equilibrium with the local radiation field.

We therefore adopt Rin = 100 rg ~ 1000 astronomical units and take conservative values for the other physical and geometrical quantities involved (supplementary text). With all the relevant pieces of information now available, we determine a mass outflow rate at the base of the wind of Embedded Image~ 10 MSun/year, corresponding (for a mass-to-radiation conversion efficiency η ~ 0.1) to about half of the Eddington accretion rate, the limit at which gravitational infall is balanced by outward radiation pressure. Consequently, the kinetic power of the wind is Pkin ~ 2 × 1046 erg/s, or 20% of the bolometric luminosity of the quasar. According to models and simulations (19, 20), the deposition of a few percent of the total radiated energy is sufficient to prompt an appreciable feedback on the host galaxy. Because most of the kinetic output will be ultimately transferred to the surrounding gas (21), these conditions are met even if Rin is set to the escape radius (~30 rg for vout = 0.25 c), the minimum plausible starting point of the wind.

The mechanical energy released over a period of 107 years (i.e., one-tenth of a typical quasar lifetime) is close to 1061 erg, comparable to or exceeding the expected binding energy of the galactic bulge in a system like PDS 456. The large opening angle evinced here suggests that the coupling of the outflow with the gas in the host galaxy will be effective, as required for strong negative feedback. We are then possibly witnessing the initial stage of the sweeping process that leads to molecular mass losses of hundreds to thousands MSun/year even in sources with no powerful radio jets (22). For a commensurate kinetic luminosity, in fact, the observed galaxy-wide outflows entail considerable mass loading and momentum boost.

A fundamental correlation (the so-called Embedded Image scaling relation) exists between the mass of the central black holes and the stellar velocity dispersion of galactic bulges on kpc scales (23, 24)—that is, hundreds of times beyond the gravitational sphere of influence of the black holes themselves. It is still debated whether this is the product of a feedback-driven black hole/host galaxy coevolution (25) or of a hierarchical assembly through galaxy mergers (26). Identifying wide-angle accretion disk winds in the Eddington-limited regime lends weight to the idea that AGN have a substantial impact on the surrounding environment. As a rare, nearby analog of the luminous AGN population at high redshift, PDS 456 shows a flavor of cosmic feedback, believed to have operated at the peak of the quasar epoch about 10 billion years ago. In distant galaxies in a similar activity phase, such powerful winds would have provided the energy and momentum to self-regulate the black hole growth and control the star formation in stellar bulges, leaving the present-day scaling relations as a record of this process (27).

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S5

Tables S1 to S3

References (2863)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: This research was supported under the U.K. Science and Technology Facilities Council grant ST/J001384/1 and is based on x-ray observations obtained with the XMM-Newton and NuSTAR satellites. XMM-Newton is a European Space Agency (ESA) science mission with instruments and contributions directly funded by ESA member states and the National Aeronautics and Space Administration. The NuSTAR mission is a project led by the California Institute of Technology, managed by the Jet Propulsion Laboratory, and funded by NASA. We thank the NuSTAR Operations, Software, and Calibration teams for support with execution and analysis of these observations. We also acknowledge financial support from the Italian Space Agency under grant ASI-INAF I/037/12/0 (G.R. and G.M.); the Italian National Institute for Astrophysics under grant PRIN-INAF 2012 (G.R.); the I-CORE program of the Planning and Budgeting Committee, the Israel Science Foundation under grants 1937/12 and 1163/10, Israel’s Ministry of Science and Technology (E.B.); and NASA under grants NNX11AJ57G and NNG08FD60C (T.J.T.). The data are stored in the science archives of the two x-ray observatories involved and will become publicly available on 25 March 2015 (XMM-Newton) and with the upcoming DR6 data release (NuSTAR).
View Abstract

Stay Connected to Science

Navigate This Article