Blueshifted Fe ii and Balmer Broad Emission lines in a Reddened Quasar

J1540+4923 has been observed in the radio spanning more than 2 orders of magnitude in frequencies, from 0.15 GHz (7C survey; Hales et al. 2007), through, e.g., 1.4 GHz (FIRST, Helfand et al. 2015; and NVSS, White et al. 1997) to 22 GHz (Georgakakis et al. 2012). The radio source shows a compact morphology (unresolved at ∼1'' or 7 kpc) and a steep spectrum (α_{ν(1.4−22 GHz)} ≈ −0.8; S_ν ∝ ν^α ). It was imaged by 2MASS (Skrutskie et al. 2006) in the NIR in 1998, and in the optical in 2002 by SDSS (York et al. 2000). The object was then target for spectroscopic observations as a candidate radio-loud quasar and confirmed as such during BOSS (in 2013; Pâris et al. 2017) and FIRST-2MASS (F2M; in 2005 using the Echellete Spectrograph and Imager (ESI) on the Keck II 10 m telescope; Urrutia et al. 2009) surveys.

The BOSS spectrum of J1540+4923 is well consistent with the SDSS photometry, implying (1) the flux calibration of the BOSS data is reliable, and (2) the quasar might not vary between the two observations. This is confirmed by the light curve from 2005 to 2013 obtained by the Catalina Sky Survey (Drake et al. 2009). We obtain the raw ESI data from the Keck archive, ⁶ and extract 1D spectrum following a routine based on the IDL program package XIDL. ⁷ The ESI spectrum is refluxed to the BOSS spectrum, which agrees well with the SDSS photometry, using a two-order polynomial. The result is displayed in Figure 1 after corrected for the foreground Galactic reddening of E(B − V) = 0.014 using the extinction curve of Schlafly & Finkbeiner (2011), ⁸ and brought to the quasar rest frame with a redshift of z = 0.69673 ± 0.00034 measured by low-ionization [O ii] emission lines.

Zoom In Zoom Out Reset image size

We observed the quasar twice using TripleSpec spectrograph on the Palomar 200 inch Hale telescope (P200/TPSP; Wilson et al. 2004) on 2020 October 6 and 2021 May 23. Integral time of each observation is 32 minutes equally split into eight exposures in an A-B-B-A dithering mode. A nearby A0V star was observed before and after the target observation for flux calibration and telluric absorption correction. We extract the 1D spectra using Spextool software (Cushing et al. 2004), and combine the two epoch spectra (see Figure 1), which agree well with each other and are well consistent with the 2MASS photometry.

We notice with interest that the optical-to-NIR color r − K ≈ 5.12 of another quasar, J23393.84−091220.99 (J2339−0912) with a similar redshift of z = 0.6600, almost equals to that of J1540+4923 (r − K ≈ 5.31; Wang et al. 2005; Urrutia et al. 2009; Kim et al. 2015). The optical Keck/ESI data of J2339−0912 are publicly available (Urrutia et al. 2009). We also obtained NIR TPSP observations of J2339−0912 for a comparison study between the two quasars. Following the same data reduction routine as J1540+4923, we extracted the 1D ESI and TPSP spectra and show the results in Figure 1.

The close resemblance of the spectral energy distributions (SEDs) between J1540+4923 and J2339-0912 greatly simplifies the investigation of the peculiar emission-line spectrum of the former. ⁹ The [O ii] forbidden narrow emission line (NEL) of the two quasars is almost coincident. Interestingly, the profiles of Fe ii and Balmer lines of these two objects differ dramatically, though their strengths are similar. It is evident in Figure 1 that the peaks of three Balmer lines in J1540+4923, Hα, Hβ, and Hγ, and possibly also Hδ, show an obvious blueshift relative to [O ii], which is commonly used as a reliable systemic redshift indicator of quasars (e.g., Komossa et al. 2008; Shen et al. 2016). The blueshift velocity is as large as ∼1500 km s⁻¹. A similar blueshift can also be detected in relatively isolated Fe ii lines, such as λ λ 4178, 4258, 4296 in multiplet Fe ii 28 of transition b4p−z4Fo and λ λ 4303 in multiplet Fe ii 27 of transition b4P−z4Do (e.g., Véron-Cetty et al. 2004; Osterbrock & Ferland 2006). Such blueshifts might be present in other Fe ii lines as well, but are difficult to identify either due to heavy blending, or due to severe contamination by sky lines at the red end of the ESI spectrum. An unshifted component may also be detected in both of Fe ii and Balmer lines, but is much weaker than the blueshifted component.

We decompose the ESI optical spectrum of J1540+4923 by fitting the data as

$$\begin{eqnarray}&&\begin{array}{c}S(\lambda )={bB}(\lambda ;v)+{uU}(\lambda )=b[{C}_{B}(\lambda )+{\rm{Fe}}\,{{\rm{II}}}_{B}(\lambda ;v)\\ \,+{L}_{B}(\lambda ;v)]+u[{C}_{U}(\lambda )+{\rm{Fe}}\,{{\rm{II}}}_{U}(\lambda )+{L}_{U}(\lambda )],\end{array}\end{eqnarray} \tag{ 1 }$$

where B(λ; v) and U(λ) represent the blueshifted and unshifted component, respectively. We use the ESI spectrum of J2339−0912 as proxy of the unshifted component, U(λ). The blueshifted component, B(λ; v), is generated from the optical spectrum of PHL 1092, a well-studied NLS1 with superstrong Fe ii emission. We extracted the two BOSS spectra of PHL 1092 from the SDSS archive ¹⁰ and combined the data. The combined spectrum was reddened assuming the same extinction as J2339−0912 derived by Wang et al. (2005), and then blueshifted by a velocity of v. The ESI optical spectrum of J1540+4923 was fitted by a linear combination of B(λ; v) and U(λ) with b, v, and u as free parameters. The wavelength range of ∼100 Å around [O iii] was masked during the fit. The best fit was obtained by minimizing χ² and the results are displayed in Figure 2. As can be seen there, this simple model reproduces the observed data of J1540+4923 reasonably well, except that the [O iii] doublet is badly underestimated. The velocity shift of Fe ii and Balmer lines (assumed to be the same by this model) is constrained to be v = 1560 ± 70 km s⁻¹.

Zoom In Zoom Out Reset image size

In the second step of Equation (1), the blueshifted and unshifted components B(λ) and U(λ), respectively, were further decomposed into a featureless continuum C_B (λ; v) ≈ C_B (λ) and C_U (λ), a pseudocontinuum of Fe ii emission Fe II _B (λ; v) and Fe II _U (λ), and emission lines other than Fe ii, L_B (λ; v) and L_U (λ). We adopted the Fe ii template recently extracted from Mrk 493 (Park et al. 2022) ¹¹ to model Fe II _B (λ; v) and Fe II _U (λ). This Fe ii template was broadened by convolving Gaussian functions and properly scaled to find the best fit to B(λ; v) and U(λ). The results are displayed in Figure 3.

Zoom In Zoom Out Reset image size

After removal of the best-fit models of the featureless continuum and Fe ii lines, we obtained the residual spectrum contributed by other emission lines, including Hβ, Hγ, Hδ, [O ii], [O iii], and [Ne iii]. We subtracted only the featureless continuum in the Hα and Paβ regimes, without considering possible contributions from Fe ii. Besides the lines listed above, the Mg ii BEL is also evident at the blue end in both the BOSS and ESI spectra, although the signal-to-noise ratios (S/Ns) are much lower than that of the rest-frame optical (see Figure 1). We rebin the ESI data and degrade the resolution to that of the BOSS spectrum and combine the two spectra to improve the spectral quality. The continuum-subtracted spectrum in the Mg ii regime is shown in Figure 4, in which both Mg ii and UV Fe ii multiplets are clearly present. A double-peaked profile of Mg ii can be evidently identified, whose red peak remains at the systemic redshift while the blue peak is shifted with a velocity similar to those of the Balmer and the optical Fe ii BELs. We tried to decompose UV Fe ii into two components, but the result is inconclusive. Considering the fact that, for the optical Fe ii, the unshifted component is much weaker than the blueshifted one (≲1/10 in flux), it is reasonable to ignore the contribution from the unshifted component to the UV Fe ii BELs. We fit the continuum-subtracted spectrum using the UV Fe ii templates of Tsuzuki et al. (2006), following the same procedure as modeling the optical Fe ii BELs. We assume that the UV Fe ii BELs have the same blueshift velocity as that of the optical Fe ii BELs. The best-fit results are also displayed in Figure 4, and the best-fit Fe ii parameters are summarized in Table 1.

Zoom In Zoom Out Reset image size

These emission lines are plotted in their common velocity space in Figure 5, overlaid with the best-fit phenomenological models (see Table 1 for the best-fit parameters). Specifically, [O ii] can be well described by a single Gaussian. Two Gaussians are needed to fit each member line of the [O iii] doublet. We used two Lorentzians to fit Hγ, Hβ, Hα, and Mg ii, considering that Lorentzian rather than Gaussian is often used to model Balmer lines in strong Fe ii emitters (e.g., Véron-Cetty et al. 2006; Zhou et al. 2006). The Paβ regime is rather noisy. We simply scaled the best-fit Hα model to find an acceptable match.

Zoom In Zoom Out Reset image size

J1540+4923 was picked out from F2M quasar sample (Glikman et al. 2004). It was noted that the high-ionization broad-absorption-line (HiBAL) quasars of this sample show an unusually high incidence (11/12) of low-ionization broad absorption lines (LoBALs, with Al iii and Mg ii as the most commonly observed ones; Urrutia et al. 2009), about 1 order of magnitude higher than that in optically selected broad-absorption-line (BAL) quasars (LoBALQ/HiBALQ ≈ 15% and HiBALQ/Q ≈ 15%; e.g., Trump et al. 2006; Gibson et al. 2009). Even rarer are Fe ii and Balmer LoBALs (e.g., Zhou et al. 2019; Leighly et al. 2022). The densities and column densities of the BAL gases tend to increase along the sequence from HiBAL through LoBAL to Fe/BalmerLoBAL, so as to lower the ionization levels and to enhance populating the excited states. Any shielding gas, which may be present in between the UV BAL outflows and the continuum source, can further help to dilute the "redundant" high-energy photons (e.g., Murray et al. 1995). Similar trends seem to be also applicable to BEL outflows. Although much rarer than commonly seen blueshifted C iv HiBELs, significant blueshifts have been found in Mg ii LoBELs. This is interpreted as an outflow signature (Marziani et al. 2011), which is possibly associated with LoBAL outflows (Liu et al. 2019). It is reasonable to speculate that there might exit outflows possibly in a minority of quasars, where the physical conditions are optimal for emitting Fe ii and Balmer BELs, as observed in J1540+4923.

With R₄₅₇₀ ≈ 1.6, the Fe ii BELs are very strong in J1540+4923, and both the Fe ii and Balmer BELs are dominated by the blueshifted component. As pointed out by Joly (1981), standard photoionization models severely underestimate R₄₅₇₀ for the parameters of a typical BELR (e.g., a density n_H ∼ 10¹⁰ cm⁻³, a column density N_H ∼ 10²³ cm⁻², and a dimensionless photoionization parameter U ∼ 10⁻²). Increasing the density and column density may enhance the Fe ii strength. The enhancement would gradually saturate, however, if only photoionization heating were considered. A viable solution is to invoke an additional energy source, and a natural choice is mechanical heating. It is possible that the outflows in J1540+4923 are heated by shocks generated by encountering with massive medium, such as the dusty torus. The essentials of this scenario was actually proposed by Joly as early as 1981, who claimed a particular region, with a much higher density and column density than that of normal BELR, is needed to generate the large R₄₅₇₀ values (similar to that observed in strong Fe ii emitters as J1540+4923). The main energy source of this region is likely mechanical heating rather than the quasar ionization continuum. If this is the case, the outflow mass of J1540+4923 would be much larger than that of a typical BELR, which is estimated to be M_BELR ∼ 10³–10⁴ M_⊙ for high-luminosity quasars (Baldwin et al. 2003; but see also Peterson 1997). Then the mass outflowing rate could be ${\dot{M}}_{\mathrm{out}}\gtrsim 10\,{M}_{\odot }\,{\mathrm{yr}}^{-1}$. This mass-loss rate, even for the most conservative estimate, is at least 2 times larger than the mass accretion rate of ${\dot{M}}_{\bullet }=5\pm 2\,{M}_{\odot }\,{\mathrm{yr}}^{-1}$ using the bolometric luminosity of L_bol = 2.73 ± 1.11 × 10⁴⁶ erg s⁻¹ after corrected for the estimated reddening of E(B − V) ≈ 0.97 and assuming an accretion efficiency of 0.1 (Kim et al. 2015).

Binary of Black Holes? The fact that the observed optical spectrum of J1540+4923 can be well reproduced by the combination of that of PHL 1092 and J2339−0912 calls to remembrance of a binary system of black holes (BBH; e.g., Zhou et al. 2004), with SDSS J153636.22+044127.0 (J1536+0441; e.g., Boroson & Lauer 2009; but see also Gaskell 2010) as an extensively studied candidate of this phenomenon. The emission-line spectrum of J1540+4923 looks similar to that of J1536+0441: both possibly have two sets of BEL systems and one NEL system. Following Boroson & Lauer (2009), we estimate the mass of each member of the candidate BBH. Such BBH are presumed to be gravitationally bound in a single galaxy, each of which has its own BELR of different redshifts and line widths. Adopting the empirical formula in Greene & Ho (2005; their Equation (6)) and using the Hα line parameters (see Table 1), we estimate the BH masses to be M_BH,B ≈ 5 × 10⁸ M_⊙ and M_BH,U ≈ 6 × 10⁶ M_⊙ for the blueshifted and unshifted BEL systems, respectively. These estimates, though probably with rather large uncertainties, are contrary to the expectation of BBH scenario (M_BH,B ≪ M_BH,U ).

Recoiling Black Hole? It might also be noticeable of the resemblance of the emission-line spectrum between J1540+4923 and SDSS J092712.65+294344.0 (J0927+2943; Komossa et al. 2008), to some extent. J0927+2943 shows two sets of NELs separated by Δv ∼ 2600 km s⁻¹. The recoiling black hole (RBH; e.g., Damour & Gopakumar 2006) interpretation of the peculiar emission-line spectrum of J0927+2943 could also be applicable to that of J1540+4923; if the weak "unshifted" Balmer (possibly also Fe ii) lines, with an intermediate width of ∼1200 km s⁻¹, were originated from an intermediate-emission-line region (IELR) between the normal B/NELR. The difficulty of this scenario lies in that it is hard to explain the large equivalent width of the unshifted Balmer lines, which are seldom observed even in normal quasars (Li et al. 2015). Furthermore, the large equivalent width of the unshifted component of the Mg ii BEL (about 50 Å; see 33 Å in the SDSS quasar composite spectrum; Vanden Berk et al. 2001), is too large to reconcile the RBH scenario. Therefore this interpretation is strongly disfavored.

Follow up observations are essential to further distinguish between the possible interpretations. Spectral monitoring is helpful to confirm/reject the BBH scenario, by detecting the possible variability of the velocity offset between the two peaks of Balmer BELs. We can also obtain high S/N spectrum by stacking these monitoring data, which will be useful to detect weak spectral features, such as the tentatively identified absorption lines in He i*λ 3889 and Ca ii K&H in Figure 1. Newly identified emission/absorption lines would motivate high-resolution spectroscopy, which, incorporating high-spatial-resolution imaging, might be able to scrutinize the RBH scenario and provide new clues for understanding the nature of the peculiar emission-line spectrum of J1540+4923.

The emission-line spectrum of J1540+4923 should be extremely rare in "normal" blue quasars, if not absent. In an early study, Boroson & Green (1992) measured the velocity shifts of the peaks of Hβ BELs in 79 PG quasars with respect to [O iii]λ5007. It is found that the shifts are all less than 20 km s⁻¹. More recently, Shen et al. (2016) found that the peaks of Hβ BELs have velocity shifts of −109 ± 400 km s⁻¹ (negative velocity indicates blueshift), using stellar Ca ii absorption-line triplets as a robust indicator of the systemic redshifts of 151 SDSS quasars with high data quality. No J1540+4923 analog is found in either of the above two samples, both of which are selected by their blue optical colors. The F2M red quasars are suggested to be in the process of dispelling the dust cocoon by strong outflows, to subsequently evolving into normal quasars (Urrutia et al. 2009; Glikman et al. 2012). In this context, as J1540+4923 is extremely windy, it is highly likely to be a very young quasar. Trails of such outflows might in fact have been hinted in previous case studies. By carefully decomposing the BELs of a LoBAL quasar (SDSS J163345.22+512748.4), Liu et al. (2019) tentatively identified a blueshifted component in both Hα and Hβ BELs, with their velocities possibly consistent with that of Mg ii. A similar Hα profile was also identified in a reddened quasar (SDSS J160558.86+474300.1; Zhou et al. 2022). J1540+4923 might be an extreme case of these rarities and deserve further studies. Systematic exploration of the analogs of J1540+4923 is expected to shed new light on SMBH growth, galaxy buildup, and the connection between the two.