CO-EVOLUTION OF GALAXIES AND CENTRAL BLACK HOLES: OBSERVATIONAL EVIDENCE ON THE TRIGGER OF AGN FEEDBACK

The observed quasars were selected from the SDSS Data Release 7 (DR7) quasar catalog (Schneider et al. 2010; Shen et al. 2011). The simple selection criteria were set based on redshift (z ≃ 0.3 in order for [O iii] λ5007 to fall in the narrowband filter transmission described below) and nuclear [O iii] luminosity ($L_{\rm N[{\rm O\,\mathsc{iii}}]} > 10^{42}$ erg s⁻¹). Among the objects meeting the criteria, those with various SMBH mass and radio loudness and with the highest [O iii] luminosity were picked up. The properties of the observed quasars are summarized in Table 1. They effectively complement the existing measurements on the AGN principal-component (PC) plane explored below.

Table 1. Targets of Subaru/Suprime-Cam Observation

Quasar	r-band	Redshift	$\log L_{{\rm N[{\rm O\,\mathsc{iii}}]}}$	log MBH	Rb	$\log L_{\rm E[{\rm O\,\mathsc{iii}}]}$
	Mag.a		(erg s−1)	(M☉)		(erg s−1)
SDSS 091401.75+050750.6	17.32	0.301	42.85 ± 0.01	9.36 ± 0.10	L	42.17 ± 0.08
SDSS 095456.89+092955.7	17.92	0.298	42.93 ± 0.01	8.56 ± 0.08	L	<41.47
SDSS 113949.47+402048.5	17.52	0.314	42.97 ± 0.01	8.21 ± 0.08	Q	<41.48
SDSS 150740.92+445331.5	18.16	0.314	42.69 ± 0.01	7.34 ± 0.14	L	<41.53
SDSS 150752.66+133844.5	17.79	0.322	42.53 ± 0.01	9.67 ± 0.02	Q	41.83 ± 0.08

Notes. ^aSDSS PSF magnitude. The measurement error is <0.02 mag. ^bRadio loudness. Q: radio-quiet, L: radio-loud.

Download table as:  ASCIITypeset image

The targets were observed with the prime-focus imaging facility Suprime-Cam (Miyazaki et al. 2002) mounted on the Subaru telescope.¹ The instrument covers a wide field of view of 34 × 27 arcmin² with 10 2k × 4k CCDs, with a pixel scale of 0.20 arcsec. The narrowband NA₆₅₆ (effective wavelength λ_eff = 6570 Å, effective bandpass Δλ_eff = 140 Å) and the broadband R_c (λ_eff = 6530 Å, Δλ_eff = 1110 Å) filters were used to trace the redshifted [O iii] line and underlying continuum, respectively. The observation was carried out on 2011 May 3–4 under the open-use program S11A-028. The sky condition was sometimes non-photometric, but the measured flux of celestial objects was mostly stable within 10% variation. The seeing was 0.8–1.1 arcsec. For each target, nine 300 s or three 900 s exposures in NA₆₅₆ and nine 100 s or three 300 s exposures in R_c were obtained, depending on the source brightness. In addition, a spectrophotometric standard star SA 104-335 was observed with the same instrument configuration as the target observation.

Data reduction was performed with the Suprime-Cam Deep Field Reduction (SDFRED; Ouchi et al. 2004) software in a standard manner including bias subtraction, flat fielding, distortion correction, sky subtraction, masking bad pixels, and stacking. Photometric calibration in NA₆₅₆ was achieved by referring to the observed flux of SA 104-335, whose intrinsic magnitude was derived by convolving its spectrophotometry data (Stone 1996) with the filter transmission function. Then, the R_c images were calibrated relative to NA₆₅₆ by requiring the mean (R_c − NA₆₅₆) color to be zero for galaxies detected with high signal-to-noise ratios (point sources were not used because stellar Hα absorption could affect this calibration). The scatter of ∼0.05 mag was found around the mean between R_c and NA₆₅₆ magnitudes in the calibration.

The reduction process then proceeded to the continuum subtraction from the NA₆₅₆ images. For this purpose, the nuclear spectra of the quasars were retrieved from the SDSS DR7 archive² (an example is shown in Figure 1). They were measured within SDSS spectroscopic fibers of 3 arcsec aperture. For each spectrum, the continuum levels beneath [O iii] λ4959 and λ5007 were estimated by interpolation between continuum windows at both (shorter and longer wavelength) sides of the lines, and then the spectrum was decomposed into [O iii] and continuum. Convolving them with the filter transmissions gave the relative fluxes of [O iii] and continuum in NA₆₅₆ ($f_{{\it NA}_{656}}^{\rm [{\rm O\,\mathsc{iii}}]}$ and $f_{{\it NA}_{656}}^{\rm cont}$) and in R_c ($f_{R_{\rm c}}^{\rm [{\rm O\,\mathsc{iii}}]}$ and $f_{R_{\rm c}}^{\rm cont}$) for each object. Using these measures, accurate continuum subtraction was achieved as follows: The radial profiles of all the quasar nuclei on the obtained images are consistent with the point-spread functions (PSFs) derived from nearby stars; hence, they are dominated by the radiation from central unresolved sources. The NA₆₅₆ and R_c images were decomposed into nuclear (NA^nuc₆₅₆, R^nuc_c) and extended (NA^ext₆₅₆, R^ext_c) components by the best-fit Moffat profiles, fitted within a 10 pixel (2 arcsec) aperture on top of the quasar peak positions. Then, the nuclear [O iii] distributions are given by

$$\begin{equation} {\it NA}_{656}^{\rm nuc} \times \frac{f_{{\it NA}_{656}}^{\rm [{\rm O\,\mathsc{iii}}]}}{f_{{\it NA}_{656}}^{\rm [{\rm O\,\mathsc{iii}}]} + f_{{\it NA}_{656}}^{\rm cont}}, \end{equation} \tag{ 1 }$$

while the extended [O iii] distributions are

$$\begin{equation} \big({\it NA}_{656}^{\rm ext} - {R}_{\rm c}^{\rm ext}\big) \times \frac{f_{{\it NA}_{656}}^{\rm [{\rm O\,\mathsc{iii}}]}}{f_{{\it NA}_{656}}^{\rm [{\rm O\,\mathsc{iii}}]} - f_{R_{\rm c}}^{\rm [{\rm O\,\mathsc{iii}}]}}. \end{equation} \tag{ 2 }$$

The second term in expression (2), which is equivalent in principle to the ratio $\Delta \lambda _{{\rm eff}, R_{\rm c}}$/($\Delta \lambda _{{\rm eff}, R_{\rm c}} - \Delta \lambda _{\rm eff, {\it NA}_{656}}$), corrects for the continuum oversubtraction of the first term due to the [O iii] themselves contained in R^ext_c. Note that expression (1), although it is most accurate, is not applicable to the extended components since the relative fluxes of [O iii] and the continuum are different from $f_{\it NA_{656}}^{\rm [{\rm O\,\mathsc{iii}}]}$ and $f_{\it NA_{656}}^{\rm cont}$ measured in the SDSS spectroscopic fibers. It is worth emphasizing that the spectral information is quite useful in the above process as an accurate cross-calibrator of the images taken in the different filters.

Zoom In Zoom Out Reset image size

Thanks to the massive spectroscopic observations conducted by the SDSS, nuclear properties are now known for many quasars with EELR measurements in the literature. We compile the EELR measurements of type-1 quasars from Boroson et al. (1985), Stockton & MacKenty (1987), and Husemann et al. (2008, 2010). The continuum and line luminosities given in Stockton & MacKenty (1987) are converted to the values in the cosmology adopted here. The measurements of Husemann et al. (2008) are corrected in accordance with their re-analysis (B. Husemann et al. 2012, in preparation). Then, we cross-match the quasars with the SDSS DR7 quasar catalog (Shen et al. 2011) and obtain nuclear properties such as continuum and line luminosity, SMBH mass, and radio characteristics based on the Faint Images of the Radio Sky at Twenty cm survey (FIRST; Becker et al. 1995). For some objects without counterparts in the SDSS catalog, the nuclear properties are taken from the above papers as well as from Fu & Stockton (2007). Combined with the new observation presented in this paper, we end up with 61 type-1 quasars with EELR measurements (27 detections and 34 non-detections) and with a variety of nuclear properties. The compiled sample is listed in Table 2. We assign radio flag 0 to RQQs, flag 1 to flat-spectrum (radio spectral index α_ν < 0.5) or core-dominant (based on the FIRST morphology; Jiang et al. 2007) radio quasars, and flag 2 to steep-spectrum (α_ν > 0.5) or lobe-dominant radio quasars. The SDSS quasars are separated into radio-quiet and radio-loud sources at the radio loudness parameter R = 5 (Shen et al. 2011). For simplicity, the objects with radio flags 0, 1, and 2 are hereafter referred to as RQQs, FSRQs, and SSRQs, respectively, since there is a well-known, strong correlation between the radio morphology and spectral index.

Table 2. Nuclear Properties of Type-1 Quasars with EELR Measurements

Quasar	Redshift	log λLλ5100	log MBH	Ra	$\log L_{\rm N[{\rm O\,\mathsc{iii}}]}$	$\log L_{\rm E[{\rm O\,\mathsc{iii}}]}$	Ref.
		(erg s−1)	(M☉)		(erg s−1)	(erg s−1)
4C 15.01	0.450	45.61	9.24	2	43.58	<42.56	2
PG 0026+129	0.142	44.96 ± 0.11	8.98	0	42.53 ± 0.09	41.80 ± 0.14	3
PG 0050+124	0.061	44.31 ± 0.11	8.00	0	42.26 ± 0.09	<43.0	3
PG 0052+251	0.155	44.98 ± 0.10	8.55	0	42.70 ± 0.09	41.75 ± 0.13	1, 2, 3
SDSS 0057+1446	0.172	44.93 ± 0.01	9.38 ± 0.02	0	42.23 ± 0.03	<43.0	3
HE 0132−0441	0.154	44.69 ± 0.10	7.73	0	42.15 ± 0.16	<43.0	3
NAB 0137−01	0.334	45.17 ± 0.01	9.01 ± 0.02	0	42.62 ± 0.02	<41.53	2
SDSS 0155−0857	0.165	44.43 ± 0.02	8.82 ± 0.04	0	42.15 ± 0.02	41.43 ± 0.19	3
HE 0157−0406	0.218	44.72 ± 0.09	8.52	0	42.09 ± 0.10	<43.0	3
Mrk 1014	0.163	44.78 ± 0.01	8.06 ± 0.05	1	42.74 ± 0.02	42.22 ± 0.14	2, 3
OI 287	0.444	45.08 ± 0.01	9.09 ± 0.21	2	43.51 ± 0.01	<42.47	2
SDSS 0836+4426	0.254	45.29 ± 0.01	8.73 ± 0.06	0	43.28 ± 0.01	43.08 ± 0.09	3
SDSS 0914+0507	0.301	44.84 ± 0.01	9.36 ± 0.10	2	42.85 ± 0.01	42.30 ± 0.08	5
SDSS 0948+4335	0.226	44.76 ± 0.01	8.43 ± 0.03	0	42.44 ± 0.01	<43.0	3
SDSS 0954+0929	0.298	44.54 ± 0.01	8.56 ± 0.08	2	42.93 ± 0.01	<41.60	5
Ton 28	0.329	45.44 ± 0.01	8.03 ± 0.43	0	42.42 ± 0.02	<42.34	2
4C 13.41	0.241	45.41 ± 0.01	9.54 ± 0.05	2	42.40 ± 0.02	<41.41	2
HE 1029−1401	0.086	44.97 ± 0.05	8.70	0	⋅⋅⋅	41.78 ± 0.12	4
PG 1049−005	0.359	45.59 ± 0.01	9.17 ± 0.03	0	43.47 ± 0.01	<42.31	2
4C 61.20	0.421	45.26 ± 0.01	9.10 ± 0.09	1	43.34 ± 0.01	<42.55	2
4C 10.30	0.422	44.78 ± 0.01	9.33 ± 0.05	2	42.75 ± 0.01	<41.62	2
3C 249.1	0.313	45.42	8.96	2	43.46	42.85 ± 0.02	1, 2
SDSS 1131+2632	0.244	44.56 ± 0.02	9.08 ± 0.04	0	42.81 ± 0.01	42.11 ± 0.14	3
SDSS 1139+4020	0.314	44.63 ± 0.01	8.21 ± 0.08	0	42.97 ± 0.01	<41.61	5
4C 49.22	0.334	44.70 ± 0.01	8.50 ± 0.01	1	42.45 ± 0.01	<41.81	2
GQ Comae	0.165	44.15 ± 0.01	8.36 ± 0.03	0	42.58 ± 0.01	40.19 ± 0.30	1, 2
PKS 1217+023	0.240	45.13 ± 0.01	8.87 ± 0.08	2	42.78 ± 0.01	41.82 ± 0.16	2, 3
4C 25.40	0.268	44.73 ± 0.01	8.86 ± 0.03	2	42.66 ± 0.01	42.61 ± 0.02	2
3C 273	0.158	46.04 ± 0.09	9.06	1	42.87 ± 0.10	<41.73	1, 2, 3
SDSS 1230+6621	0.184	44.53 ± 0.01	8.26 ± 0.03	0	42.66 ± 0.01	41.90 ± 0.15	3
HE 1228+0131	0.117	44.95 ± 0.09	8.29	0	42.23 ± 0.11	<43.0	3
SDSS 1230+1100	0.236	44.73 ± 0.01	8.60 ± 0.05	0	42.64 ± 0.01	<43.0	3
PG 1307+085	0.154	44.80 ± 0.01	8.59 ± 0.05	0	42.69 ± 0.01	<41.06	2
B2 1425+267	0.364	45.14 ± 0.02	9.66 ± 0.03	2	42.99 ± 0.01	42.37 ± 0.04	1, 2
PG 1427+480	0.221	44.62 ± 0.01	7.95 ± 0.03	0	42.52 ± 0.01	<43.0	3
SDSS 1444+0633	0.208	44.48 ± 0.01	8.33 ± 0.03	0	42.44 ± 0.01	<43.0	3
HE 1453−0303	0.206	45.31 ± 0.12	8.59	0	42.54 ± 0.12	<43.0	3
SDSS 1507+4453	0.314	44.39 ± 0.01	7.34 ± 0.14	1	42.69 ± 0.01	<41.66	5
SDSS 1507+1338	0.322	44.79 ± 0.01	9.67 ± 0.02	0	42.53 ± 0.01	41.96 ± 0.08	5
4C 37.43	0.371	45.21 ± 0.01	9.37 ± 0.11	2	43.18 ± 0.01	43.02 ± 0.01	1,2
PG 1519+226	0.136	44.40 ± 0.01	7.90 ± 0.07	0	41.34 ± 0.09	<43.0	1
OR 241	0.254	44.86 ± 0.01	8.44 ± 0.05	1	42.39 ± 0.01	<41.85	2
3CR 323.1	0.265	45.20 ± 0.01	9.07 ± 0.03	2	42.99 ± 0.01	42.35 ± 0.13	1, 2, 3
4C 11.50	0.436	44.99 ± 0.01	8.87 ± 0.08	2	43.23 ± 0.01	42.68 ± 0.03	2
Ton 256	0.131	44.51 ± 0.01	7.99 ± 0.02	1	43.00 ± 0.01	42.80 ± 0.12	1, 3
SDSS 1655+2146	0.154	44.67 ± 0.01	8.28 ± 0.04	0	43.03 ± 0.01	43.03 ± 0.09	3
PG 1700+518	0.292	45.43 ± 0.09	8.98	0	42.63 ± 0.10	41.81 ± 0.18	2, 3
3CR 351	0.372	45.79 ± 0.01	9.81 ± 0.04	2	43.50 ± 0.01	<42.37	2
II Zw 136	0.061	44.40 ± 0.10	8.27	0	41.95 ± 0.09	40.90 ± 0.32	1, 3
PKS 2135−147	0.200	44.97	9.15	2	43.16	41.85 ± 0.04	2
HE 2152−0936	0.192	45.59 ± 0.10	8.76	0	42.17 ± 0.11	<43.0	3
HE 2158−0107	0.213	44.89 ± 0.11	8.60	0	42.59 ± 0.09	42.32 ± 0.13	3
HE 2158+0115	0.160	44.21 ± 0.10	7.94	0	42.01 ± 0.09	<43.0	3
4C 31.63	0.298	45.99	8.54	1	42.92	<42.08	1, 2
PG 2214+139	0.067	44.44 ± 0.11	8.64	0	41.81 ± 0.10	<43.0	3
PG 2233+134	0.326	45.19 ± 0.01	8.57 ± 0.31	0	42.50 ± 0.07	<40.58	2
PKS 2251+113	0.325	45.07	9.15	2	42.98	42.51 ± 0.04	1, 2
HE 2307−0254	0.221	44.89 ± 0.10	8.39	0	42.00 ± 0.11	<43.0	3
4C 09.72	0.433	45.17	9.30	2	42.80	<42.71	2
PKS 2349−014	0.174	44.77 ± 0.02	8.81 ± 0.03	2	42.45 ± 0.01	42.11 ± 0.11	3
HE 2353−0420	0.229	44.42 ± 0.10	8.12	0	42.58 ± 0.09	42.47 ± 0.12	3

Note. ^aRadio flag. 0: RQQ, 1: FSRQ (flat-spectrum or core-dominant radio quasar), 2: SSRQ (steep-spectrum or lobe-dominant radio quasar). References. (1) Boroson et al. 1985; (2) Stockton & MacKenty 1987; (3) Husemann et al. 2008 (including the re-analyzed data given in B. Husemann et al. 2012, in preparation); (4) Husemann et al. 2010; (5) This work.

Download table as:  ASCIITypeset image

We also collect the EELR measurements of 20 type-2 quasars (10 detections and 10 non-detections) presented by Humphrey et al. (2010) and Villar-Martín et al. (2010, 2011). Since they were selected from the SDSS type-2 quasars identified by Zakamska et al. (2003), luminosity and EW of emission lines arising from the narrow-line region (NLR) as well as optical ugriz magnitudes of the host galaxies are available in the archive. We derive the rest-frame g − i color and stellar mass of the host galaxies from the SDSS model magnitudes and spectroscopic redshifts using a k-correction code developed by Blanton & Roweis (2007). The contributions of relatively strong [O ii] λ3727 and [O iii] λ4959, λ5007 lines from quasars are removed before the calculation by subtracting Δm_X = −2.5log (1 + f_{X, Y}[EW_Y/w_X]) from the X (u, g, r, i, or z) band magnitude if line Y is included in the band, where f_{X, Y} is the filter transmission at line Y relative to the band average and w_X is the bandwidth. We summarize the derived properties of the type-2 quasars in Table 3. All but two sources, SDSS 1228+0050 and SDSS 1625+3106, are radio-quiet (SDSS 1625+3106 is in the transition range of radio-quiet and radio-loud objects; Villar-Martín et al. 2011). In addition, we use all the type-2 quasars in Zakamska et al. (2003) and ∼60,000 galaxies extracted from the SDSS galaxy sample at z = 0.3–0.6, the redshift range of the main sample (the type-2 quasars in Table 3), as control objects. Their rest-frame color and stellar mass are calculated in the same way the objects in the main sample are calculated.

Table 3. Nuclear and Host Galaxy Properties of Type-2 Quasars with EELR Measurements

Quasar	Redshift	$\log L_{\rm N[{\rm O\,\mathsc{iii}}]}$	Rest-frame	Stellar Mass	EELRc	Ref.
		(erg s−1)a	g − i (mag)b	log M⋆(M☉)
SDSS 0025−1040	0.303	42.32	0.75	11.0	+1	3
SDSS 0123+0044	0.399	42.72	1.14	10.9	+1	2
SDSS 0217−0013	0.344	42.34	0.93	11.1	+1	3
SDSS 0234−0745	0.310	42.36	0.91	10.5	−1	3
SDSS 0840+3838	0.313	42.21	1.13	11.4	+1	1
SDSS 0849+0150	0.376	41.65	1.01	10.9	−1	3
SDSS 0920+4531	0.402	42.63	0.91	11.1	−1	1
SDSS 0955+0346	0.421	42.19	1.24	11.2	+1	3
SDSS 1153+0326	0.575	43.20	0.62	11.2	+1	3
SDSS 1228+0050	0.575	42.87	0.85	11.0	−1	3
SDSS 1307−0214	0.425	42.51	1.09	11.4	+1	3
SDSS 1337−0128	0.329	42.31	0.90	11.3	+1	3
SDSS 1407+0217	0.309	42.49	0.86	10.5	−1	3
SDSS 1413−0142	0.380	42.84	1.18	11.0	−1	3
SDSS 1546−0005	0.383	41.77	1.18	11.0	−1	3
SDSS 1550+3950	0.347	43.05	0.60	11.3	+1	1
SDSS 1625+3106	0.379	42.60	1.35	11.1	−1	1
SDSS 1726+6021	0.333	42.16	0.96	10.5	−1	1
SDSS 1739+5442	0.384	42.01	1.24	11.1	−1	1
SDSS 2358−0009	0.402	42.91	0.88	11.1	+1	3

Notes. ^aThe measurement error is <0.01. ^bTypical error of a few tens of magnitude is expected, including the uncertainties in the photometry and k-correction. ^cEELR detection flag (+1: detected, −1: non-detected). References. (1) Humphrey et al. 2010; (2) Villar-Martín et al. 2010; (3) Villar-Martín et al. 2011.

Download table as:  ASCIITypeset image

While the above samples of type-1 and type-2 objects are used below to probe the different aspects of quasars, we should keep in mind that the two populations can be intrinsically different. Elitzur (2012) clarified this point by demonstrating that the two types of AGNs are drawn preferentially from the distribution of dust-torus covering factors. Since type-2 AGNs have dustier environments around them than do type-1 counterparts, their host galaxies might tend to have higher star formation rates since (1) the presence of ample dust indicates active star formation producing them and (2) the negative feedback (suppression of star formation) by the AGN radiation might be weaker due to the dust obscuration.

The Suprime-Cam [O iii] line images of the newly observed quasars are presented in Figure 2. To aid visual inspection, the gray scale consistently represents linear flux levels between zero and quasar peak values, while the white crosses indicate the PSF sizes (twice the full widths at half-maximum (FWHMs)). The most spectacular feature is found around SDSS 1507+1338. The emission feature, extending across ∼50 kpc along the NE–SW direction, is reminiscent of the ionization cones created by AGN radiation. SDSS 0914+0507 also has apparently extended emission, with a knot ∼35 kpc away from the center in the NW direction. On the other hand, the radial profiles of SDSS 0954+0929, SDSS 1139+4020, and SDSS 1507+4453 are consistent with the PSFs and show no signs of EELRs. We measure the EELR luminosity in an annulus of inner radius 10 kpc and outer radius 30 kpc. The wing components of the nuclear radiations are estimated from the PSFs and subtracted. The incomplete transmission of the [O iii] lines through the NA₆₅₆ filter is also corrected using the filter transmission function. We include in the error budget of the derived luminosity the uncertainties in the continuum subtraction and in the correction for the wing contribution of the nuclear radiation, background sky noise, and the possible flux variation assumed to be 10% due to the non-photometric sky condition. The results are listed in the last column of Table 1. The above measurements support our visual inspection finding detectable EELRs only in SDSS 0914+0507 and SDSS 1507+1338. The nuclear [O iii] luminosity ($L_{{\rm N[{\rm O\,\mathsc{iii}}]}}$) is also measured within a 3 arcsec aperture and is found to agree with the SDSS spectroscopic measurements within $\Delta \log L_{{\rm N[{\rm O\,\mathsc{iii}}]}} = \pm 0.1$.

Zoom In Zoom Out Reset image size

As described in Section 1, the presence of EELRs correlates with some AGN properties. Quasars with larger [O iii] EW, broader and bumpier Balmer lines, fainter Fe ii, and SSRQ-like radio emission have a higher possibility of accompanying EELRs around them. The correlation is fairly strong, as Fu & Stockton (2009) showed that if certain quasars are pre-selected based on their radio and nuclear [O iii] emissions, ∼50% of them would turn out to have EELRs as opposed to ∼20% with no a priori assumption. Note that the above emissions arise from different parts of AGNs: broad Balmer lines and the Fe ii lines from parsec-scale BLR, the [O iii] lines from kiloparsec-scale NLR, and radio and EELR emissions from more extended, galaxy-wide scales. The correlation among them, except for EELRs, has been well known since the seminal work by Boroson & Green (1992). With the PC analysis technique, they showed that most of the variance in AGN spectral properties is contained in two sets of correlations. A strong anti-correlation between the measures of [O iii] and Fe ii is PC 1, or eigenvector 1, with which the Balmer line profile is also associated. The general consensus is that PC 1 is driven mainly by Eddington ratio (e.g., Boroson 2002; Sameshima et al. 2011). In Figure 3, we plot Hβ line width versus the relative Fe ii strength $R_{{\rm Fe\,\mathsc{ii}}}$ ≡ EW(Fe ii)/EW(Hβ) of the SDSS quasars in Shen et al. (2011). Both measures constitute the four-dimensional eigenvector 1 (4DE1; other measures are the soft X-ray photon index and C iv λ1549 line profile) space proposed by Sulentic et al. (2000). The anti-correlation between them is evident in Figure 3, which forms a sequence of Eddington ratios. The data scatter reflects the orientation effects as well as the intrinsic variety of quasars, since one would observe larger line FWHM and $R_{{\rm Fe\,\mathsc{ii}}}$ (Fe ii is emitted from the outermost part of BLR; e.g., Matsuoka et al. 2008) as the viewing angle relative to the dust-torus plane decreases. On the other hand, PC 2 links optical luminosity and He ii λ4686 line emission. Its physical origin is thought to be mass accretion rate to SMBHs because of its strong correlation with optical luminosity. Therefore, one would expect that the ratio of PC 1 and PC 2 is proportional to SMBH mass, which was actually proved by Boroson (2002). Furthermore, Laor (2000) found that radio loudness is strongly related to SMBH mass as radio-loud/radio-quiet quasars are associated with most/least massive SMBHs. The above arguments suggest that most of the AGN properties can be understood on the PC 1–PC 2 plane.

Zoom In Zoom Out Reset image size

We plot the Eddington ratio and continuum luminosity of the compiled type-1 sample, as well as all of the SDSS quasars in the Shen et al. (2011) catalog, in Figure 4. First of all, we remark that the objects with EELR measurements cover the distribution of all of the SDSS quasars fairly uniformly on this plane. This was not anticipated since the existing EELR observations are subject to strong selection biases individually. We see the clear separation of radio-quiet and radio-loud sources as pointed out by Laor (2000), the latter being associated with most massive SMBHs. In addition, SSRQs are observed to have higher values of Eddington ratio or SMBH mass than FSRQs. We also find some correlations containing EELRs. They are shown more clearly in Figure 5 where the fraction of EELR detection is plotted as a function of the nuclear properties. The EELR fraction is strongly anti-correlated with the Eddington ratio, and its weak correlation with SMBH mass may also be present. The former relation is most likely the underlying basis of the correlation between EELR and PC 1. While no clear trend is found with respect to radio emission, we do find that EELR luminosity is systematically higher in SSRQs than in the rest of the sample; the mean luminosity is $L_{\rm E[{\rm O\,\mathsc{iii}}]}$ = 10^42.4 and 10^41.9 erg s⁻¹ for the SSRQs and RQQs, respectively. The difference will be even larger if we consider the fact that the EELR luminosity of most SSRQs measured by Stockton & MacKenty (1987) is underestimated since the emission in a central 10 kpc aperture is excluded. This explains why the EELR fraction around SSRQs is much higher than those around the other populations in the pioneering work of Stockton & MacKenty (1987).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

One of the highest dependencies of EELRs is found on nuclear [O iii] luminosity, $L_{\rm N[{\rm O\,\mathsc{iii}}]}$. In Figure 6, we show Eddington ratio versus $L_{\rm N[{\rm O\,\mathsc{iii}}]}$ of the same samples. It is clearly seen that the quasars with EELR measurements are biased to high-$L_{\rm N[{\rm O\,\mathsc{iii}}]}$ compared to the SDSS quasars. Here, we find that low-$L_{\rm N[{\rm O\,\mathsc{iii}}]}$ quasars tend to lack EELRs: only 3 out of 20 objects with $L_{\rm N[{\rm O\,\mathsc{iii}}]} < 10^{42.5}$ erg s⁻¹ have detectable, faint EELRs, while the EELR fraction is >50% at the higher $L_{\rm N[{\rm O\,\mathsc{iii}}]}$. This trend was reported previously for smaller samples (e.g., Stockton & MacKenty 1987). Figure 6 also tells us that $L_{\rm N[{\rm O\,\mathsc{iii}}]}$ does not strongly correlate with the Eddington ratio, although it correlates with PC 1 (Boroson & Green 1992). No clear correlation between the two measures indicates that PC 1 is not solely driven by the Eddington ratio—there should be another physical mechanism(s) that links nuclear [O iii] luminosity with other PC 1 constituents.

Zoom In Zoom Out Reset image size

What can we learn from these correlations? While the clear anti-correlation between EELRs and the Eddington ratio is observed, it does not seem to reflect their direct relationship through the AGN radiation since (1) the EELR fraction is highest for the AGNs with the most inefficient gas accretion to SMBHs, (2) nuclear [O iii] luminosity produced similarly by AGN radiation shows no clear correlation with Eddington ratio, and (3) a positive correlation between the EELR fraction and continuum luminosity is not observed. Therefore, it is more reasonable to consider the presence of an independent mechanism that leads to the high/low EELR fraction and low/high Eddington ratio at the same time. The dependence of EELRs on radio emission and (possibly) SMBH mass may also be related to this hidden mechanism. We will return to this issue below. The lack of a simple relation between EELRs and continuum luminosity (accretion rate) was also reported by Fu & Stockton (2009), who suggested that the correlation between nuclear and extended [O iii] emissions, which is also evident in our larger sample, may be the sequence of the available gas amount rather than AGN emission characteristics. The hypothesis that the gas availability is the primary determinant of EELR creation is supported by the probe of host galaxies in the following section.

If EELRs represent a certain stage of galaxy evolution, as the AGN feedback scenario implies, their presence may be related to stellar population properties of host galaxies. Here, an investigation of such a relation is presented for the first time, thanks to the recent measurements of EELRs around SDSS type-2 quasars at z = 0.3–0.6 (Humphrey et al. 2010; Villar-Martín et al. 2010, 2011). In Figure 7, we plot stellar mass versus rest-frame g − i color of the type-2 quasars as well as the control samples defined in Section 2.2. Most SDSS galaxies at these relatively high redshifts are luminous red galaxies (LRGs), forming a clump at the massive red end on this color–stellar mass diagram (we could see the red sequence rather than a clump if the SDSS observations were deep enough to detect fainter galaxies). The type-2 quasars are bluer than the LRGs (see also Zakamska et al. 2003) but they are redder than the star-forming galaxies that form the "blue cloud" around g − i ∼ 0.6 in the local universe (e.g., Gavazzi et al. 2010). This is consistent with the result of Schawinski et al. (2010), who found that local AGNs reside in the "green valley" between the blue cloud and the red sequence on the color–stellar mass diagram. Interestingly, we find that the EELR quasars are associated with bluer and more massive galaxies compared to their non-EELR counterparts. In order to show this trend more clearly, we define a "massive-blueness" indicator, $\mathcal {I}_{\rm mb} \equiv \log M_{\star }$ − (g − i), which increases diagonally toward the bottom right corner of Figure 7 as shown by the arrow. The number distribution of the type-2 quasars with EELR measurements as a function of $\mathcal {I}_{\rm mb}$ is plotted in Figure 8. It shows a clear, monotonic increase in the fraction of EELR quasars as $\mathcal {I}_{\rm mb}$ increases: only 1/8 of the lowest-$\mathcal {I}_{\rm mb}$ quasars have EELRs, while the fraction rises to 7/8 in the highest-$\mathcal {I}_{\rm mb}$ counterparts.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

What is the origin of this dichotomy? Blue colors point to current or recent star formation activity, which in turn indicates the presence of a certain amount of gas in the galaxies. Based on the velocity structure and mass of EELR gas around SSRQs, as well as the correlation between EELRs and radio emission, Fu & Stockton (2009) suggested that EELRs are formed in galaxy minor mergers accompanied by the production of radio jets associated with AGNs. A similar conclusion was reached by Husemann et al. (2010) for an RQQ. Villar-Martín et al. (2011) found that the majority of the type-2, radio-quiet EELR quasars show the signs of galaxy merger or interaction, while none of the non-EELR counterparts show such signs. These observations suggest that the gas may be brought in by the merging companions. The merger/interaction possibly activates the observed star formation as well as the AGN that ionizes the gas and creates the EELR, since the interstellar gas of merging galaxies can be effectively brought into the nuclear regions with rapid loss of the angular momentum (e.g., Barnes & Hernquist 1996). The fact that EELRs are preferentially found in gas-rich, massive blue galaxies is consistent with both the above scenario and the previous arguments that the presence of an EELR is regulated by the amount of available gas. Furthermore, massive galaxies are known to be the dominant hosts of radio-loud AGNs (e.g., Floyd et al. 2004; Best et al. 2005). It naturally explains the EELR–radio emission connection described above, and can also provide the "hidden" mechanism behind the anti-correlation between EELRs and the Eddington ratio since the majority of radio-loud AGNs are powered by inefficiently accreting, most massive SMBHs (see Figure 4).

The recent kinematic observations have revealed the presence of galaxy-wide, disturbed gas in quasar host galaxies (e.g., Fu & Stockton 2009; Greene et al. 2011). While only a small fraction of the gas is generally found to have high enough velocity to escape the galaxies, Greene et al. (2011) showed that these estimates are in fact lower limits and may be consistent with expectations of recent AGN feedback models. If the feedback scenario is correct, the gas responsible for EELR and star formation (and hence the blue colors of galaxies) would soon be swept away by the AGN-driven, galactic winds. In this regard, it may be noteworthy that none of the most luminous (log λL_λ5100 ⩾ 45.5) type-1 quasars in our sample have EELRs, as shown in Figure 4. It is tempting to raise the possibility of the gas stripping, but we have to be aware that (1) the relatively high Eddington ratio for these objects may be responsible for the deficit of EELRs, and (2) the bright central radiation can prevent us from detecting EELRs in the most luminous sources. More data are needed for further inspection of this issue. While the process discussed above corresponds to the so-called "quasar-mode" feedback, it is interesting to point out that a clear interaction between radio jets and EELR clouds with high-velocity dispersion is observed in SSRQs (Fu & Stockton 2009), which invokes another "radio mode" of the feedback.

In the nearby universe, Keel et al. (2012) investigated the properties of Seyfert galaxies with EELRs dominated by type-2 objects. While their morphology supports the merger origin of EELRs, they have relatively quiescent kinematics without a sign of high-velocity gas leaving the galaxies seen in higher-luminosity AGNs. This may simply indicate that the creation of the powerful galactic winds requires quasar-class energy input into the interstellar medium; hence, the relation of EELR to AGNs may be different at high and low luminosities.