EVIDENCE FOR ACTIVE GALACTIC NUCLEUS DRIVEN OUTFLOWS IN YOUNG RADIO QUASARS^*

Supermassive black holes (BHs) and their host galaxies appear to be closely linked in their formation and evolution (e.g., Magorrian et al. 1998; Gebhardt et al. 2000; Ferrarese & Merritt 2000). While the exact origin of their relationship is still under debate (Kormendy & Ho 2013), theoretical models often invoke feedback from active galactic nuclei (AGNs) as a crucial mechanism for establishing the correlation. Outflows are believed to be associated with AGN feedback, which regulate the growth of BHs and quench star formation by expelling gas (e.g., Di Matteo et al. 2005). AGN feedback also appears to be required for cosmological models to explain the upper end of the luminosity function of galaxies (e.g., Bower et al. 2006). Despite the potential importance of AGN feedback, direct observational evidence of this process is limited. For example, extended [O iii] emission in high-z radio galaxies and ultraluminous infrared galaxies (ULIRGs) associated with AGNs are thought to be signs of AGN-driven outflows (e.g., Nesvadba et al. 2007, 2011; Harrison et al. 2012).

Numerical simulations indicate that feedback is triggered during an intense and rapidly accelerated BH fueling phase during a gas-rich galaxy merger (Hopkins et al. 2008; Narayanan et al. 2010). Objects in the early feedback phase are likely to be both highly luminous and highly obscured by dust (Haas et al. 2003). Thus, young quasars can be effectively searched using their mid-infrared (MIR) spectral energy distribution (SED). However, to avoid confusion with dusty starbursts, whose MIR SED can be similarly red, an additional diagnostic besides the MIR color is needed to identify AGN-powered sources. Here we use the presence of a strong, compact radio source to confirm the AGN nature of our MIR-selected objects, in a similar fashion as Martínez-Sansigre et al. (2005). Our goal is to see whether young radio quasars show signatures of outflows that might be indicative of AGN feedback during the early stages of their evolution.

We use the Wide-Field Infrared Survey Explorer (WISE; Wright et al. 2010) to find luminous, obscured quasars (C. J. Lonsdale et al., in preparation). The sample is selected to be bright in 22 μm (S_22 μm > 4 mJy) and to have extremely red colors in the MIR, 1.25[m(3.4 μm) − m(4.6 μm)] + [m(4.6 μm) − m(12 μm)] > 7 in the Vega magnitude system. Then we match the red objects from WISE to the NRAO VLA Sky Survey (NVSS; Condon et al. 1998) to select sources with S_1.4 GHz/S_22 μm > 1 to ensure that the sample is powered by AGNs (Ibar et al. 2008). Finally, we only include sources with no extended radio structure, in order to avoid evolved radio galaxies. This results in 156 sources. We are in the process of performing follow-up observations to explore the origin of these sources and to search for observational signatures of AGN feedback. Here we present near-infrared (NIR) spectroscopic data of a subset of these young luminous quasars. Throughout the Letter we adopt H₀ = 71 km s⁻¹ Mpc⁻¹, Ω_m = 0.27, and Ω_Λ = 0.73.

As our sources are selected from WISE and are chosen to be very dusty, they are expected to be very faint at shorter wavelengths. No reliable photometry is available in the optical or NIR. From the observed W1 (3.6 μm) flux density and an SED template for type 2 AGNs (Polletta et al. 2007), we estimate that our sources are fainter than 19–21 AB mag in the K band. We obtained NIR (simultaneous J, H, and K) spectra for 24 objects using the Folded-port InfraRed Echellette (FIRE) on the Baade 6.5 m telescope at Las Campanas Observatory on UT 2012 July 27–29. The data were taken in low-resolution prism mode using a 06 slit. The sky conditions were clear and the seeing was 08–16. The observations were taken at low airmasses (⩽1.4), with the slit aligned along the parallactic angle. Each object was dithered along the slit, with a position offset of 9'' and an exposure time of 158 s in each position. The total integration times range from 1902 s to 2536 s, split into several repeats of ABBA sequences. Following each science target, we observed a nearby A0V star (from a catalog provided by the observatories) for telluric and relative flux calibration. Since a majority of the selected A0V stars are fairly bright, we slightly offset the slit from the center of the stars to avoid saturation. Absolute flux calibration was performed using two fainter, unsaturated stars that were properly centered on the slit. Since all of the science objects are too faint to be seen on the slit image, we had to apply a blind offset from the nearest Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006) source in order to place the object on the slit. For the typical brightnesses of our sample, the relative astrometric uncertainty⁶ between WISE and 2MASS is ∼04. Given the seeing conditions during the observing run, our absolute flux scale may have been underestimated by up to 70%. Moreover, it is possible that our source blind offset acquisition procedure may have entirely missed some of the objects; this may partly account for the low detection rate reported below. The FIRE spectra cover 0.8–2.6 μm with an instrumental resolution of λ/Δλ ≈ 250 (1200 km s⁻¹), as estimated from the FWHM of the arc lines.

We used the IDL package FIREHOSE for data reduction. Flat-field correction was done using images taken with an external quartz lamp. We used He+Ne+Ar arc lamp spectra for wavelength calibration. After extracting the one-dimensional spectra, we corrected for telluric absorption and applied flux calibration using the xtellcor package within the Spextool pipeline (Cushing et al. 2004; see Vacca et al. 2003 for a detailed description).

Table 1 lists the basic properties of the sample, including the total exposure times and the quality of the observed spectra. The signal-to-noise ratio of the FIRE spectra is too low to see the underlying continuum. We positively detected identifiable emission lines in 7 out of the 24 objects observed (Table 2); their spectra are shown in Figures 1 and 2. The redshifts lie in the range z ≈ 1.6–2.5. Six show [O iii] λλ4959, 5007 and Hα, but in J1412−2020 the lines lie in spectral regions of strong telluric absorption and thus are somewhat uncertain. In J1634−1721 only Hα is detected. It is difficult to know why the majority of the sample was undetected. Possible reasons include (1) the narrow-line region is heavily reddened by dust; (2) the emission lines coincide with strong OH lines or sky absorption; or (3) the slit was misplaced due to the astrometric uncertainty between WISE and 2MASS. In particular, since the accuracy of the telescope offsets is not precisely known throughout the night, it is difficult to know which factors are dominantly responsible for the non-detection of emission lines. The detected sources are essentially indistinguishable from the undetected ones in terms of MIR flux, MIR color, or radio brightness.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 2. Spectral Measurements and Derived Physical Parameters

Name	z	FWHM([O iii])	F(Hβ)	F([O iii])	F(Hα+[N ii])	log L[O iii]	log LHα+[N ii]	log P1.4 GHz	log AV	log (MBH/M☉)	log (Mhost/M☉)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
J1308−3447	1.6457 ± 0.0007	1640 ± 394	<19.9	11.3 ± 2.3	5.0 ± 3.7	43.45 ± 0.09	43.13 ± 0.29	26.88	...	8.85–9.61	11.3–12.1
J1343−1136	2.4860 ± 0.0009	1620 ± 427	<10.9	18.2 ± 3.8	9.8 ± 3.9	43.85 ± 0.09	43.71 ± 0.13	26.34	...	9.26–9.34	11.5–11.6
J1400−2919	1.6564 ± 0.0006	2120 ± 348	5.3 ± 0.8:	18.2 ± 2.7	47.7 ± 9.8	43.66 ± 0.06	44.08 ± 0.09	26.77	$1.3^{+0.8}_{-0.6}$	9.06–9.72	11.5–12.2
J1412−2020	1.8082 ± 0.0056	1900 ± 385	<8.8	11.7 ± 9.7	9.2 ± 7.3	43.51 ± 0.36	43.45 ± 0.31	26.09	...	8.92–9.73	11.3–12.1
J1634−1721	2.0702 ± 0.0030a	...	<9.2	<12.4	25.9 ± 7.9	<43.60	43.96 ± 0.26	26.24	...	<9.53	<11.9
J1951−0420	1.5833 ± 0.0003	1320 ± 211	2.8 ± 0.4	19.7 ± 2.4	14.5 ± 4.7	43.67 ± 0.05	43.56 ± 0.14	26.33	$0.0^{+0.5}_{-0.0}$	9.07–9.10	11.5–11.6
J2000−2802	2.2772 ± 0.0028	1870 ± 962	<2.4	12.8 ± 5.5	14.9 ± 10.8	43.66 ± 0.19	43.75 ± 0.30	26.54	>0.2	9.07–9.46	11.3–11.7

Notes. Column (1): source name. Column (2): redshift based on [O iii]. Column (3): FWHM of [O iii] in units of km s⁻¹, corrected for the instrumental resolution. Column (4): flux of Hβ in units of 10⁻¹⁵ erg s⁻¹ cm⁻². A measurement possibly affected by night-sky lines is flagged with a colon. Column (5): flux of [O iii] in units of 10⁻¹⁵ erg s⁻¹ cm⁻². Column (6): flux of Hα+[N ii] in units of 10⁻¹⁵ erg s⁻¹ cm⁻². Column (7): luminosity of [O iii] in units of erg s⁻¹. Column (8): luminosity of Hα+[N ii] in units of erg s⁻¹. Column (9): radio luminosity at 1.4 GHz in units of W Hz⁻¹. Column (10): reddening (mag) in the V band estimated from the Balmer decrement. Column (11): BH mass in units of M_☉, see text for details. Column (12): stellar mass of host galaxies in units of M_☉, see text for details. ^aRedshift is derived using Hα.

Download table as:  ASCIITypeset image

We do not detect broad Hβ or Hα in any of the objects. Assuming FWHM = 5000 km s⁻¹ for a hypothetical broad component and M_BH =10⁹ M_☉, we can use the Hα virial mass formalism of Greene & Ho (2005b) to estimate the flux of broad Hα that should be present if our objects are genuinely unobscured type 1 AGNs. Comparing these expected values with the observed upper limits of broad Hα flux, we conclude that the broad-line region, if present, must be reddened by E(B − V) ⩾ 1.2 mag. This suggests that our objects are true type 2 quasars rather than the lightly reddened quasars usually selected using NIR colors (e.g., Glikman et al. 2012).

We follow the procedures outlined in Ho & Kim (2009) to fit the spectra to obtain basic parameters for the emission lines. We use a single Gaussian to fit Hβ and each component of the doublet of [O iii] λλ4959, 5007. The width is fixed to be the same for all three lines. We constrain the peak ratio of [O iii] λ4959 to [O iii] λ5007 to 2.88 and fix the separation between the two lines to the known value. For objects with no detection of Hβ, we estimate a 3σ upper limit of its flux by using the width of [O iii]. For J1634−1721, where [O iii] is also undetected, we assume a Gaussian profile with σ = 500 km s⁻¹. For Hα and [N ii] λλ6548, 6583, we simply measure the integrated flux of the lines because they are severely blended at our spectral resolution.

Although Hβ is reliably detected in only two sources (J1400−2919 and J1951−0420), the relatively high flux ratios of [O iii] λ5007 to Hβ (3.4 and 7.0, respectively) suggest that the primary ionizing source of our sample is associated with AGNs rather than young stars. This finding is consistent with the strong radio emission relative to 22 μm flux. However, because we do not resolve Hα from [N ii], we cannot use the [N ii]/Hα ratio and traditional optical line ratio diagnostic diagrams (e.g., Ho 2008) to completely rule out the possibility that star formation also contributes to the ionizing continuum. For the same reason, it is also difficult to properly constrain the extinction of the host galaxy from the Balmer decrement. To estimate the Hα flux from the total flux of Hα+[N ii], we adopt an average value of [N ii]/Hα (0.83 ± 0.42) derived from quasars selected from the Sloan Digital Sky Survey (SDSS; Shen et al. 2011). We estimate A_V = 1.3 and 0.0 mag for J1400−2919 and J1951−0420, respectively; the upper limit of Hβ for J2000−2802 yields a lower limit of A_V = 0.2 mag. These values of A_V are surprisingly low for MIR-selected sources, but they might be biased in view of the large number of nondetections, which may be heavily extinct.

Interestingly and quite surprisingly, the [O iii] lines are spectrally resolved for all detected sources. The FWHM of the line, corrected for instrumental resolution, ranges from 1320 to 2120 km s⁻¹.

4.1. Black Hole Mass and Host Galaxy Mass

4.2. AGN Feedback

There is mounting evidence that broad [O iii] emission is related to AGN-driven outflows (e.g., Nesvadba et al. 2007; Harrison et al. 2012). Figure 3 (left) shows the correlation between the FWHM and luminosity of [O iii] for various types of quasars and radio galaxies. Where the line was fit with two Gaussians in the literature, we plot the broad and narrow components open and filled symbols, respectively; single-Gaussian fits are plotted with solid symbols. Clearly, our sample has extremely broad [O iii] line widths compared to other sources with similar redshift and luminosity. Excluding the SDSS quasars,⁷ we find a weak correlation between FWHM and L_[O iii] with a Spearman's rank correlation coefficient of 0.54 (P = 0.0002). This correlation has been known for various types of AGNs (e.g., Ho et al. 2003, and references therein), although our trend is driven by sources with extreme line widths (≳1000 km s⁻¹) compared to those in previous studies (a few hundred km s⁻¹). While the origin of the correlation between FWHM and L_[O iii] is still unclear (e.g., Whittle 1992; Ho et al. 2003), the very large line widths seen in our sample almost certainly do not track gravitational motions (see Section 4.1). Instead, the most natural explanation for the supervirial velocities is that they reflect the kinematics of AGN-driven outflows, which seem to be prevalent in our sample of young radio quasars. The flux ratio of radio to [O iii] for our sources is slightly lower than in high-z radio galaxies and compact steep-spectrum (CSS) sources, indicating that our sample is moderately radio-loud.

Zoom In Zoom Out Reset image size

A jet-induced origin for the large velocities is supported by the distribution of [O iii] FWHM and radio power (Figure 3, right), which qualitatively resembles trends noted by Heckman et al. (1981) and Nesvadba et al. (2011). Integral-field spectroscopy of high-z AGNs with broad [O iii] lines finds that the emission is often associated with extended outflows of ionized gas (e.g., Nesvadba et al. 2011; Cano-Díaz et al. 2012). The energy deposition of the outflows cannot be fully explained by star formation, suggesting that an additional source of energy must come from the AGN (e.g., Harrison et al. 2012). Spatially resolved spectroscopy is needed to estimate the energy contribution from AGNs in our sample.

Our WISE-selected, radio-emitting AGNs bears an interesting similarity to CSS sources, whose small physical extent (∼ a few kpc) and steep radio spectrum suggest that they are young radio galaxies (Fanti et al. 1990). The broad and complex velocity profiles of [O iii] observed in CSS sources are thought to arise from interaction between the radio jet and the interstellar medium of the host (Gelderman & Whittle 1994; Holt et al. 2008). Figure 3 shows that bona fide CSS sources tend to have larger [O iii] widths compared to ordinary SDSS quasars; the line width enhancement is largest for more distant systems. Our sample, selected to be both dusty and compact in radio morphology, qualitatively resembles CSS sources, although we presently do not yet have sufficient radio spectral information to firmly confirm this association. Follow-up interferometric radio observations are under way.

Theoretical models of galaxy formation that invoke AGN feedback often link ULIRGs and quasars through an evolutionary sequence mediated by galaxy merging (e.g., Di Matteo et al. 2005). In this model, a massive, gas-rich merger begins with a ULIRG phase dominated by star formation, which then evolves into ULIRGs that host AGNs, young quasars, and then finally to optically revealed quasars with suppressed star formation. The strength of AGN feedback is expected to dramatically increase in the later stages. In ULIRGs with AGNs, Harrison et al. (2012) found striking evidence for AGN feedback from the detection of spatially extended [O iii] emission and broad line widths ranging from 900 to 1400 km s⁻¹. Their sample is also detected in the submillimeter, indicating substantial ongoing star formation. Interestingly, our sample has even larger line widths than theirs, although the stellar masses of the host galaxies we deduced for our sources are comparable (e.g., Hickox et al. 2012).

We thank Robert Simcoe and Rik Williams for helpful advice on the Magellan observations. An anonymous referee offered valuable input. This work is supported by a KASI-Carnegie Fellowship (M.K.) and the Carnegie Institution for Science (L.C.H.).