SDSS 0956+5128: A BROAD-LINE QUASAR WITH EXTREME VELOCITY OFFSETS

The offset between the broad Hβ and Mg ii emission lines is sufficiently surprising and difficult to explain that we must consider whether measurement error, fitting error, or absorption from the host galaxy might be responsible for mislocating one of the line centroids. SDSS 0956+5128 was observed a second time using the SDSS spectrographs as part of SEGUE-2 (Yanny et al. 2009; Figures 1 and 2), seven years after the initial spectrum (it was erroneously targeted as a candidate blue horizontal-branch star). The SEGUE-2 spectrum has higher signal to noise and is consistent with the previous SDSS measurement of Mg ii at z = 0.707, Hβ at z = 0.690, and the narrow lines at z = 0.714.

Zoom In Zoom Out Reset image size

Because the Hβ profile is asymmetric and has such a large offset, we performed two additional follow-up observations (Table 2). Using Keck/DEIMOS (Faber et al. 2003), a higher-resolution and higher signal-to-noise spectrum using the 830 G grating (R ∼ 3200, compared to ∼2000 for SDSS) was taken on 2011 November 23–24, under photometric conditions and 06–08 seeing. On November 23 two 600 s exposures were taken at a slit position angle (P.A.) of 90 and on November 24 900 s and 300 s exposures were taken at a slit P.A. of 0. The 1200 G grating tilted to 7800 Å was used with a OG550 blocking filter and 05 slit, resulting in a spectral resolution of 0.43 Å FWHM measured from sky lines. The data were reduced with a modified version of the DEIMOS reduction package which accounts for guider drift, variable seeing, wavelength calibration errors, and non-photometric conditions.

The spectro-photometric standard star HZ-44 was used for relative flux calibration. Regions of telluric absorption and strong spectral lines in HZ-44 were masked, and then a 10th order polynomial was fit to the ratio of measured to expected flux. After scaling out the baseline throughput, Gaussian lines were fit to the residual flux ratio at the location of known telluric absorption lines to account for atmospheric absorption. Both spectra show Hβ in the same location as the SDSS spectra (Figure 2). The Hβ profile shows similar asymmetry in both SDSS and Keck spectra.

In addition, an infrared spectrum was taken using TripleSpec (Wilson et al. 2004; Rayner et al. 2003; Cushing et al. 2004; Vacca et al. 2003; R ∼2800) for eight 240 s exposures in 2011 December on the Astrophysical Research Consortium 3.5 m at Apache Point Observatory. This spectrum reveals a broad Hα emission line at z = 0.690 with a profile comparable to Hβ (Figure 3), with no evidence of strong absorption lines in either line. We note that although the strong asymmetry in Hα and Hβ is one characteristic of disk or double-peaked emitters (Eracleous et al. 1995; Strateva et al. 2003), as above, the large offset between Hβ and Mg ii is not characteristic of the few examples with well-measured Hβ and Mg ii. We conclude that the measured centroids for the Balmer lines indeed lie at a different redshift than the narrow lines. In Table 2, we summarize the observations, both spectroscopy and imaging, of SDSS 0956+5128 that we use in this study.

Zoom In Zoom Out Reset image size

We consider the broadband SED of the source combining SDSS data with GALEX (Martin et al. 2005), WISE (Wright et al. 2010), and 2MASS (Skrutskie et al. 2006). We find that the optical data require A_V = 0.3 reddening of the SDSS composite quasar spectrum (Vanden Berk et al. 2001; Richards et al. 2006). Even this is not enough to reproduce the observed level of IR emission. Therefore, we have added a galaxy spectrum with primarily an old stellar population from Maraston (2005). We make use of the K-band (Section 3.2) image decomposition to constrain the AGN template and the galaxy template in our SED fit. This is important to reduce the degeneracies between the different models. This decomposition is shown in Figure 4. We note that some photometric data points are likely to be affected by strong emission lines (i.e., Lyα, Hα). Both the quasar and inferred host galaxy appear to be dusty.

Zoom In Zoom Out Reset image size

From our decomposition, we roughly estimate the stellar mass of the galaxy to be ∼8 × 10¹⁰ M_☉. We also use our Mg ii and the strength of the nearby continuum fit to infer a central black hole mass of log (M/M_☉) = 8.65 using the Mg ii-based virial mass estimator calibrated by McLure & Dunlop (2004). This estimate may be incorrect if there is substantial non-virial motion, although the Mg ii line appears symmetric despite its offset. The estimated black hole mass and galaxy luminosity are consistent with a central black hole–galaxy pair lying near the z = 0 M_BH − L_gal relation (Bentz et al. 2009). However, our inferred galaxy is likely a factor of ∼4 lower mass than necessary to lie on the M_BH − M_bulge relation (Häring & Rix 2004). Because the inferred host lies either on or below these relations, any different counterpart corresponding to the quasar should be nearly as bright as our inferred host.

Stronger evidence that the quasar lies in the inferred host comes from follow-up J-, H-, and K-band images using the Subaru IRCS with adaptive optics (AO; Hayano et al. 2008, 2010). We performed this imaging using the AO188 Adaptive Optics system in laser guide star (LGS) mode, with the LGS equidistant from the quasar and point-spread function (PSF) star in order to improve PSF stability. The K-band image with 52 mas pixels shown in Figure 5 was taken with a net exposure time of 66 minutes but varying conditions. The PSF was then modeled using a star of brightness similar to the main quasar point source, as a combination of a Gaussian core and Moffat wing, and rotated to correct for mismatched alignment between the source, LGS, and PSF star.

Zoom In Zoom Out Reset image size

Only the K-band image shown in Figure 5, with a PSF of 0.15 arcsec, was suitable for our purposes. The second K-band image was taken with a 20 mas pixel scale. Therefore, the flux per pixel is much lower and our integration time was too short. We did not reach the necessary depth to determine properties of the host, although extended host emission can be detected. The J, H, Kp images were taken in service mode with a P.A. = 90. This leads to different distances from the guide star to the PSF and the QSO. Anisoplanatism degrades the seeing and cannot be controlled with effectively one PSF star. So, we cannot predict the shape of the PSF at the quasar position. Therefore, we decided to reject those images in the analysis.

We then decompose the K-band image into a point source and host using GALFIT v. 3.0 (Peng et al. 2010). First a PSF-only model is fit to the data, then a Sersic profile is fit for the host galaxy contribution. We find that the AGN–host flux ratio is 1.0 ± 0.4, and the best-fit parameters for the host are those of a compact, early-type galaxy (n_Sersic ∼ 3.7–4.5, a/b = 0.50 ± 0.15, R_1/2 ∼ 1.3 kpc). The resulting picture is thus consistent with a point source lying in a single host galaxy (Figure 5).

There appear to be asymmetric structures in the central pixels, but due to the difficulties in controlling the AO PSF, we cannot be certain that they are real. Surprisingly, the quasar appears not to be located at the highest concentration of starlight, likely the galactic bulge. If SDSS 0956+5128 does not include a recoiling black hole, this is difficult to understand even in the case of a merger, particularly coupled with the high velocity. However, the apparent offset may be an artifact caused by imperfections in the point-source subtraction.

The Balmer lines in this object are strongly skewed with a strong peak blueward of the systemic velocity. Such objects cannot be fit with a typical, circular disk, even with Doppler boosting. However, they can be fit by models such as an eccentric disk (Eracleous et al. 1995) in which an additional parameter is added.

Sequences of various line profiles with different parameters can be found in Eracleous et al. (1995) and Strateva (2004). A line profile with a strong blue peak 4000 km s⁻¹ bluer than the central value and a low-amplitude red peak is predicted for models with high values of the disk orientation φ₀, high eccentricity, and a small inner pericenter distance ξ₁. We note that our ability to fit such a line shape using an eccentric disk may not be all that surprising due to the large number of parameters. The seven parameters (in addition to the central black hole mass) allowed by an eccentric disk, combined with noise and the difficulties of decomposing the spectrum into Hβ broad and narrow components and [O iii]λ4959, 5007, might mean such a model will fit almost any skewed line shape. Further, other sorts of non-circular disk may also be consistent with this line profile, i.e., such a fit is not unique.

Regardless, an eccentric disk model does not produce different line shapes for Mg ii and Hβ, since the ionization potentials of the two species are similar. The parameters producing a strongly skewed Hα and Hβ will produce a similarly shaped Mg ii line, also with a peak at z ∼ 0.690 instead of z ∼ 0.707. So, although the Balmer lines alone could be consistent with SDSS 0956+5128 as a disk emitter, Mg ii is inconsistent with this interpretation.

Merritt et al. (2006) proposed that a post-merger, recoiling black hole could produce a symmetric BLR, offset by ≳ 1000 km s⁻¹ from the narrow-line region velocity. Could SDSS 0956+5128 be an example of such a recoil? We must show that the SMBH is associated with the galaxy emitting the narrow lines at z ∼ 0.714. Because of the high redshift, we cannot rule out that the black hole lies in another galaxy along the line of sight. Rather, we must rely on indirect evidence, considering whether the observed galaxy is consistent with being the host of the SMBH responsible for Balmer and Mg ii emission.

The Subaru IRCS imaging reveals only one host, indicating that if there are multiple galaxies along the same line of sight, most of the galactic flux is coming from one, largest galaxy. The observed galaxy is a promising recoil candidate in that a decomposition into a galactic and quasar SED shows extinction that could be characteristic of a dusty galaxy. Mergers would likely be associated with strong extinction, although galaxies not undergoing mergers can also be dusty at many stages of their evolution.

Moreover, the inferred host galaxy mass is 8 × 10¹⁰ M_☉, with a luminosity close to that expected for a 10^8.65 M_☉ black hole lying on the M − L_gal relation. A 10^8.65 M_☉ central black hole is unlikely to lie in a substantially less luminous galaxy than the one detected by SDSS, and therefore the BLR is unlikely to belong to a quasar with a different galactic counterpart.

The presence of broad emission lines is evidence that if this is a recoiling black hole, the recoil took place in the recent past. Using an α-disk (Shakura & Sunyaev 1973) model, the accretion disk can temporarily survive the recoil and the BLR will continue to emit over a lifetime of (Loeb 2007)

$$\begin{equation} t_{{\rm disk}} = 8.4 \times 10^6 \alpha ^{-0.8}_{-1}\eta ^{0.4}M_7^{1.2}{v_8}^{-2.8} \hbox{ yr}. \end{equation} \tag{ 1 }$$

For a coalesced black hole mass of 10^8.65 M_☉ (M₇ = 45) and speed of v₈ = 1.21, the dimensionless combination of radiative efficiency and Eddington ratio η = 0.20 according to the bolometric luminosity given in Shen et al. (2008). Assuming a typical viscosity parameter of α = 0.1, this yields a lifetime of t_disk ∼ 1.4 × 10⁸ yr for the BLR. The BLR lifetime would be shorter if the Balmer line value of v₈ = 4.1 were used. Thus, we should expect coalescence to have taken place within the past 140 million years. This is likely comparable to the dynamical timescale for the host galaxy, so if this is indeed a recoiling black hole, follow-up observations might find evidence of a recent merger. The quasar-subtracted host galaxy fit would be consistent with a recent merger, but is insufficient to prove one has occurred. Alternatively, there might have been a long time delay between the merger and coalescence.

During these 140 million years, ignoring gravitational effects from the host galaxy, the ejected black hole could travel a distance of d ∼ v_ejt_disk ∼ 175 kpc for v_ej ∼ 1200 km s⁻¹, or well outside of the host galaxy. Depending upon the mass of the host and the current separation, it is possible that SDSS 0956+5128 contains a recoiling black hole that will escape its host. Follow-up observations capable of resolving multiple point sources and determining both the distance to the center of the galaxy and mass of the host would be needed in order to determine its eventual fate.

Thus, the Mg ii (z ∼ 0.707) line by itself, combined with the quasar narrow lines at z ∼ 0.714 and AO observations of the host, presents a picture consistent with a post-recoil SMBH. However, the Balmer lines are entirely inconsistent with this explanation; they are better explained as a disk emitter. Perhaps the explanation is that SDSS 0956+5128 is in some intermediate state in which a recoil has led to an eccentric disk, yet these two scenarios seem incompatible given the similar ionization potentials of Hβ and Mg ii.

Might these scenarios be compatible? Because of the number of degrees of freedom involved in disk emitter models, it is possible to fit the Balmer lines of SDSS 0956+5128 as centered at the z ∼ 0.707 redshift of Mg ii rather than of the narrow lines, again with large φ₀, high eccentricity, and a large inner pericenter distance ξ₁. If the impulse delivered to the central black hole at coalescence is non-adiabatic, it may drive a circular accretion disk to become highly eccentric. There might then be some brief window in which Balmer emission is coming from an eccentric disk but Mg ii is not yet. One test of this picture would be the C iv line profile, which cannot be investigated from the ground because it requires UV spectroscopy. Because C iv has an ionization potential placing it closer to the central black hole than Hβ, it must also be double peaked. If such a model is plausible, though, it requires a combination of rare events: (1) a merger; (2) the correct spin and geometry for a large recoil; and (3) timing such that the quasar is observed in the brief time when the Balmer lines have become strongly asymmetric but the slightly higher-radius Mg ii emission is not yet asymmetric. Since there are no known examples merely where (1) and (2) both occur, this scenario seems unlikely. However, this object appears to be unique among a catalog of >10⁴ quasars containing both Hβ and Mg ii, so it could be a rare glimpse of an object in transition between a circular and eccentric disk.

What is clear, however, is that SDSS 0956+5128 is a challenge for current models of quasar accretion. If related to a post-coalescence black hole, it would be particularly compelling because it was identified from the spectrum alone, without the ability to resolve the host galaxy, since most mergers occur at too high a redshift for us to resolve the host. An ability to detect such sources from a large survey such as SDSS gives us the opportunity to better understand merger rates and galaxy evolution and to better calibrate our expectations for planned gravitational wave detectors. If SDSS 0956+5128 is an exotic disk emitter, it will teach us something about the nature of quasar accretion, and in particular that Hβ and Mg ii, despite similar ionization potentials, can arise from quite distinct regions. And, if SDSS 0956+5128 is in a rare, post-merger transition state, a proper model may allow a better understanding of the quasar life cycle or duty cycle.