OUTFLOW AND METALLICITY IN THE BROAD-LINE REGION OF LOW-REDSHIFT ACTIVE GALACTIC NUCLEI

BLR outflow is recognized through the velocity profiles of broad emission lines in AGN spectra, characterized by asymmetry or blueshift with respect to the systemic velocity of the object (e.g., Sulentic et al. 2000). Asymmetry and relative blueshift can be quantified straightforwardly based on observed emission-line spectra (e.g., Sulentic et al. 2000, 2007; Baskin & Laor 2005; Wang et al. 2011). Wang et al. (2011) proposed a new indicator, the blueshift and asymmetry index (BAI), which takes into account both the asymmetry and the relative blueshift of emission lines (see also Wang et al. 2012). In our analysis, we investigate three parameters to quantify the strength of the AGN outflow at the BLR, the asymmetry index (AI), the velocity shift index (VSI), and the BAI.

The AI is defined as the flux ratio between the blue part from the C iv profile peak and the total (Equation (1)). The VSI is defined as the velocity difference between the C iv profile peak and the laboratory center of the C iv emission, 1549.06 Å, which is the oscillator strength weighted average wavelength of the two C iv doublet lines at 1548.20 Å and 1550.77 Å (Equation (2); see, e.g., Vanden Berk et al. 2001). The C iv BAI, which combines the blueshift and asymmetric effects in the C iv velocity profile, is the flux ratio between the blue part from the C iv laboratory center and the total (Equation (3)). These indices are expressed by the following equations:

$$\begin{eqnarray}&&\mathrm{AI}=\displaystyle \frac{F(\mathrm{Blue}\ \mathrm{region}\ \mathrm{from}\ \mathrm{the}\ \mathrm{CIV}\ \mathrm{peak})}{F(\mathrm{Total}\ \mathrm{CIV})}\end{eqnarray} \tag{ 1 }$$

$$\begin{eqnarray}&&\mathrm{VSI}=v(\mathrm{CIV}\ \mathrm{peak})-v(\mathrm{CIV}\ \mathrm{lab})\end{eqnarray} \tag{ 2 }$$

$$\begin{eqnarray}&&\mathrm{BAI}=\displaystyle \frac{F(\mathrm{Blue}\ \mathrm{region}\ \mathrm{for}\ \mathrm{the}\ \mathrm{CIV}\ \mathrm{laboratory}\ \mathrm{wavelength})}{F(\mathrm{Total}\ \mathrm{CIV})}.\end{eqnarray} \tag{ 3 }$$

The systemic redshift is crucial for determining the VSI and BAI; without it, the laboratory location of C iv is not known. In principle, the best way to measure the systemic redshift of galaxies is to use stellar absorption-line features (e.g., Bae & Woo 2014; Woo et al. 2016). Although measuring systemic redshift from the stellar absorption line is very challenging in type 1 AGNs, due to strong AGN continuum emission (but see, e.g., Woo et al. 2008; Harris et al. 2012; Park et al. 2012b, 2015; Woo et al. 2015), we try to measure stellar absorption lines (Mgb, Fe ii]λ5270, Ca II triplet, and Ca H&K lines) in the SDSS spectra of our sample by using the penalized pixel fitting method (the PPXF code; Cappellari & Emsellem 2004). This does not provide a good constraint, mainly due to insufficient signal-to-noise ratios of the SDSS spectra. Therefore, we have to focus on some emission-line features, instead of stellar features, to determine the systemic redshift of each object.

Sometimes the UV Mg ii emission line from the BLR is used to determine the systemic redshift of type 1 AGNs (e.g., Richards et al. 2002) because the Mg ii line is one of the strong low-ionization emission lines from the BLR and is thus less affected by nuclear radial motions of gas outflows, as already mentioned in Section 1. However, we do not use the Mg ii line for determining the systemic redshift because some of the UV spectra of our sample do not cover the Mg ii line. Another reason is that the total mass of BLR gas is merely tiny with respect to that of the host galaxy (e.g., Baldwin et al. 2003) and thus could be easily affected by possible radial motions. Some observations actually report that the Mg ii-based redshift shows systematic differences from the systemic redshift (e.g., based on the [O iii] line; McIntosh et al. 1999; Vanden Berk et al. 2001).

More appropriate features than the Mg ii line for determining the systemic redshift are narrow optical emission lines from NLRs, since the NLR resides at much larger spatial scales (e.g., Bennert et al. 2006a, 2006b) with a larger total gas mass than the BLR (e.g., Fraquelli et al. 2003; Crenshaw et al. 2015). In previous studies, the [O iii] line is used to determine the systemic redshift (e.g., Hewett & Wild 2010). However, high-ionization NLR lines, such as [O iii], sometimes show asymmetric velocity profiles, due to radial gas motions (e.g., Komossa et al. 2008; Bae & Woo 2014; Woo et al. 2015, 2016), because such relatively high-ionization NLR lines arise from the innermost part of NLRs in AGNs (e.g., Nagao et al. 2000, 2001) and are thus easily affected by outward pressure due to AGN radiation. Therefore, in this work, we use low-ionization lines for determining the systemic redshift. Specifically, we use the [S ii] [O ii], [O i], and Hβ emission lines in this study. We fit those lines in the SDSS spectra by adopting a simple Gaussian profile. We fix the wavelength separations and widths of the [S ii] (at 6718.29 Å and 6732.67 Å in the laboratory) and [O i] (at 6302.05 Å and 6365.5 Å) doublet emissions. We treat the [O ii] emission as a single Gaussian line at 3728.48 Å (the average of the two doublet line wavelengths) because the [O ii] doublet lines are heavily blended in most cases. As for the Hβ and [O iii] wavelength regions, we subtract the blended Fe II multiplet and stellar emission before the narrow-line measurement. Note that the subtraction of Fe II and stellar emission is important not only for the NLR fitting but also for the measurement of the velocity dispersion of the broad Hβ emission for the estimation of ${M}_{\mathrm{BH}}$ (see Section 3.2). We use the Fe II template given by Tsuzuki et al. (2006), and the host galaxy template given by Bruzual & Charlot (2003), following procedures described in Park et al. (2012b, 2015). After decomposing the broad component of Hβ with the sixth order of the Gauss–Hermitian series, we fit the narrow component of Hβ and [O iii] with the single Gaussian. We do not use [N ii] emission lines since they are buried by the broad Hα line in most objects in our sample.

Using the measured central wavelength of each NLR line, we then calculate the systemic velocity (reference redshift) and estimate the BAI and VSI using the peak of the C iv line from the best-fit model. To obtain the outflow and metallicity indicators, we fit the emission lines of interest, namely, C iv, N v, Si iv, O iv], and He ii, by adopting a multicomponent fitting method (see also Shin et al. 2013). First, we divide the BLR lines into two groups according to the ionization potential and assume that all emission lines in the same group have the same emission-line profile. Second, we mask out narrow absorption lines present in some targets. Third, we adopt a double-Gaussian profile, which well reproduces the observed BLR emission lines. Here we assume that both Gaussian components originate from the BLR since the velocity width of the narrower Gaussian component is greater than 1500 km s⁻¹ in most cases, while the contribution of the NLR emission to the CIV profile does not seem to be significant as we do not detect any in our fitting analysis. Although it is possible that there is a weak narrow C iv component, its effect on our results is presumably insignificant. As expected, the high-ionization emission lines (N v, C iv, O iv], He ii) are well fitted with the same line profile, suggesting that other high-ionization emission lines also show strong outflow and C iv. However, N v, O iv], and He ii are blended with other emission lines. Thus, we use C iv to avoid any systemic uncertainty due to deblending.

In Figure 1, we compare the BAIs obtained by adopting different reference lines for a consistency check. All VSIs measured based on each redshift reference line are correlated with one another with only small scatters, showing that there is no significant systematic difference between the VSIs derived from other emission lines. The scatter is mostly within a few tens of km ${{\rm{s}}}^{-1}$ except for cases where the redshift given in the NED is adopted. In the following analysis, we exclude the reference redshift adopted from the NED. Instead, we decide to adopt the redshift reference in the order of [S ii], [O i], [O ii], and Hβ if not all lines are available. This order is determined based on line strength and uncertainties. Note that the Hβ-based redshift potentially has a large uncertainty since the Hβ narrow component in type 1 AGNs is heavily blended with the Hβ broad component, and thus we adopt the Hβ-based redshift only when the redshift based on forbidden emission lines is unavailable. The measured BAI, VSI, and AI are given in Table 2, with the flag showing which narrow line is used as a reference for the systemic redshift. To estimate errors in the BAI and VSI, we consider one resolution element of the SDSS spectrum (i.e., ∼70 km s⁻¹) as the 1σ uncertainty of the redshift by assuming that the peak of the reference lines (e.g., [S ii]) would not move more than one pixel. Regarding the uncertainty of the AI, we simply calculate it by adding the blue-part flux error and red-part flux error, which are estimated based on the signal-to-noise ratio of each pixel. The estimated errors in the BAI, VSI, and AI are also given in Table 2.

Zoom In Zoom Out Reset image size

Table 2.  Outflow Indices, AGN Properties, and UV Emission-line Fluxes

Object	AI	VSI	BAI	Flag	${\sigma }_{{\rm{H}}\beta }$	log[$\lambda {L}_{5100}$]	log[${M}_{\mathrm{BH}}/{M}_{\odot }$]	log[${L}_{\mathrm{Bol}}/{L}_{\mathrm{Edd}}$]	N v	Si iv+O iv]	C iv	He ii
		(km s−1)			(km s−1)	($\mathrm{erg}\,{{\rm{s}}}^{-1}$)			(10−14 ${{\rm{erg\; s}}}^{-1}$ ${\mathrm{cm}}^{-2}$)	(10−14 ${{\rm{erg\; s}}}^{-1}$ ${\mathrm{cm}}^{-2}$)	(10−14 ${{\rm{erg\; s}}}^{-1}$ ${\mathrm{cm}}^{-2}$)	(10−14 ${{\rm{erg\; s}}}^{-1}$ ${\mathrm{cm}}^{-2}$)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)
Mrk 106	0.59 ± 0.05	−76 ± 60	0.60 ± 0.02	1	2439 ± 26	44.45 ± 0.01	8.46 ± 0.01	−1.15 ± 0.01	26.8 ± 2.1	8.7 ± 0.9	54.2 ± 9.0	⋯
Mrk 110	0.54 ± 0.04	472 ± 65	0.43 ± 0.02	1	1992 ± 48	42.87 ± 0.01	7.45 ± 0.02	−1.72 ± 0.02	54.5 ± 5.4	47.2 ± 5.2	227.2 ± 20.3	22.2 ± 1.4
Mrk 290	0.55 ± 0.03	119 ± 66	0.46 ± 0.01	1	2683 ± 32	43.42 ± 0.01	8.02 ± 0.01	−1.75 ± 0.01	45.0 ± 4.6	22.1 ± 1.4	172.0 ± 20.7	15.1 ± 2.4
Mrk 493	0.49 ± 0.07	−133 ± 66	0.55 ± 0.02	1	887 ± 175	43.19 ± 0.01	6.83 ± 0.17	−0.79 ± 0.17	28.0 ± 6.5	18.3 ± 2.8	56.4 ± 13.2	7.1 ± 1.1
Mrk 506	0.48 ± 0.05	227 ± 65	0.45 ± 0.01	1	3078 ± 150	42.85 ± 0.01	7.85 ± 0.05	−2.14 ± 0.05	62.7 ± 7.3	34.5 ± 4.1	218.2 ± 23.7	17.5 ± 1.8
Mrk 1392	0.57 ± 0.04	9 ± 66	0.58 ± 0.01	1	2859 ± 32	43.32 ± 0.01	8.03 ± 0.01	−1.85 ± 0.01	11.4 ± 0.6	9.7 ± 0.5	87.4 ± 8.4	6.6 ± 1.0
PG 0157+001	0.55 ± 0.04	−1483 ± 63	0.82 ± 0.01	2	1515 ± 95	44.71 ± 0.01	8.14 ± 0.06	−0.57 ± 0.06	29.2 ± 2.1	⋯	64.0 ± 2.8	3.8 ± 0.2
PG 0921+525	0.52 ± 0.03	440 ± 65	0.36 ± 0.02	1	1934 ± 49	42.87 ± 0.01	7.42 ± 0.03	−1.69 ± 0.03	39.7 ± 1.9	46.9 ± 2.8	292.0 ± 12.9	13.7 ± 0.5
PG 0923+129	0.53 ± 0.07	482 ± 66	0.45 ± 0.01	1	1679 ± 53	43.11 ± 0.01	7.41 ± 0.03	−1.44 ± 0.03	66.3 ± 5.0	⋯	183.3 ± 15.8	18.4 ± 1.3
PG 0947+396	0.51 ± 0.03	−143 ± 56	0.55 ± 0.01	1	2804 ± 27	44.65 ± 0.01	8.70 ± 0.01	−1.19 ± 0.02	32.9 ± 2.0	11.5 ± 0.7	59.5 ± 2.6	7.3 ± 0.3
PG 1022+519	0.49 ± 0.07	31 ± 65	0.51 ± 0.02	1	965 ± 144	43.35 ± 0.01	7.00 ± 0.13	−0.79 ± 0.13	12.2 ± 1.1	⋯	45.7 ± 4.5	⋯
PG 1048+342	0.51 ± 0.13	−147 ± 58	0.54 ± 0.01	1	2328 ± 43	44.42 ± 0.01	8.40 ± 0.02	−1.13 ± 0.02	12.4 ± 2.9	⋯	19.1 ± 3.4	⋯
PG 1049-005	0.52 ± 0.12	−681 ± 50	0.68 ± 0.01	1	2925 ± 21	45.51 ± 0.01	9.19 ± 0.01	−0.82 ± 0.01	29.7 ± 7.3	11.3 ± 1.9	41.6 ± 6.4	4.7 ± 0.7
PG 1115+407	0.50 ± 0.04	−690 ± 59	0.63 ± 0.01	1	1830 ± 73	44.59 ± 0.01	8.26 ± 0.04	−0.81 ± 0.04	18.5 ± 1.2	⋯	34.4 ± 2.0	⋯
PG 1121+422	0.47 ± 0.05	213 ± 76	0.40 ± 0.02	4	1834 ± 45	44.94 ± 0.01	8.44 ± 0.02	−0.64 ± 0.02	14.5 ± 1.9	⋯	47.8 ± 3.5	⋯
PG 1151+117	0.51 ± 0.05	−229 ± 80	0.55 ± 0.01	4	2734 ± 31	44.67 ± 0.01	8.68 ± 0.01	−1.16 ± 0.01	35.9 ± 3.9	⋯	51.1 ± 3.5	6.6 ± 0.5
PG 1202+281	0.56 ± 0.05	−860 ± 58	0.76 ± 0.01	1	3183 ± 34	44.12 ± 0.01	8.54 ± 0.01	−1.57 ± 0.02	8.3 ± 0.8	⋯	72.9 ± 5.1	⋯
PG 1229+204	0.50 ± 0.02	−172 ± 64	0.50 ± 0.01	1	2384 ± 29	43.70 ± 0.01	8.05 ± 0.01	−1.49 ± 0.01	15.0 ± 0.6	47.5 ± 2.0	156.4 ± 4.8	11.4 ± 0.3
PG 1244+026	0.50 ± 0.08	−309 ± 65	0.67 ± 0.02	1	853 ± 267	43.41 ± 0.01	6.91 ± 0.28	−0.64 ± 0.28	3.7 ± 0.3	⋯	11.4 ± 1.2	⋯
PG 1307+085	0.54 ± 0.03	−55 ± 59	0.55 ± 0.01	1	2459 ± 25	44.74 ± 0.01	8.62 ± 0.01	−1.02 ± 0.01	64.3 ± 4.2	⋯	114.6 ± 4.6	⋯
PG 1341+258	0.50 ± 0.12	379 ± 62	0.42 ± 0.01	1	1824 ± 58	43.75 ± 0.01	7.82 ± 0.03	−1.21 ± 0.03	22.4 ± 2.6	⋯	34.7 ± 5.3	⋯
PG 1404+226	0.51 ± 0.09	−2240 ± 67	0.91 ± 0.01	2	1205 ± 110	44.13 ± 0.01	7.61 ± 0.09	−0.63 ± 0.09	9.5 ± 1.3	⋯	13.1 ± 1.6	⋯
PG 1415+451	0.57 ± 0.04	−425 ± 61	0.65 ± 0.01	1	1713 ± 77	43.95 ± 0.01	7.86 ± 0.04	−1.05 ± 0.04	44.0 ± 2.7	20.4 ± 1.1	55.0 ± 2.8	6.6 ± 0.3
PG 1425+267	0.51 ± 0.02	−442 ± 53	0.55 ± 0.01	2	4068 ± 19	45.14 ± 0.01	9.31 ± 0.00	−1.32 ± 0.01	8.6 ± 0.4	⋯	49.4 ± 1.5	⋯
PG 1427+480	0.48 ± 0.03	136 ± 55	0.44 ± 0.01	1	2019 ± 39	44.56 ± 0.01	8.34 ± 0.02	−0.92 ± 0.02	13.8 ± 0.7	9.3 ± 0.4	43.1 ± 2.0	6.1 ± 0.3
PG 1444+407	0.50 ± 0.06	−780 ± 74	0.65 ± 0.01	4	2447 ± 31	45.14 ± 0.01	8.82 ± 0.01	−0.82 ± 0.01	36.0 ± 2.1	⋯	32.6 ± 2.8	⋯
PG 1448+273	0.56 ± 0.10	−421 ± 64	0.72 ± 0.02	1	1209 ± 171	44.18 ± 0.01	7.65 ± 0.13	−0.61 ± 0.13	10.0 ± 0.8	3.7 ± 0.2	13.2 ± 1.5	4.1 ± 0.5
PG 1512+370	0.57 ± 0.08	41 ± 53	0.56 ± 0.01	2	3447 ± 28	45.18 ± 0.01	9.17 ± 0.01	−1.13 ± 0.02	32.9 ± 5.8	8.3 ± 1.0	70.2 ± 7.3	4.5 ± 0.8
PG 1519+226	0.44 ± 0.08	−411 ± 108	0.53 ± 0.02	3	1552 ± 45	44.34 ± 0.01	7.97 ± 0.03	−0.77 ± 0.03	22.4 ± 3.4	29.0 ± 6.0	49.0 ± 5.9	⋯
PG1534+580	0.54 ± 0.02	124 ± 66	0.47 ± 0.01	1	2701 ± 30	43.42 ± 0.01	8.03 ± 0.01	−1.75 ± 0.01	40.8 ± 1.5	25.1 ± 0.5	187.5 ± 6.0	14.3 ± 0.6
PG 1545+210	0.49 ± 0.04	34 ± 57	0.49 ± 0.01	2	3248 ± 20	45.15 ± 0.01	9.10 ± 0.01	−1.09 ± 0.01	36.9 ± 2.0	⋯	101.6 ± 5.4	6.9 ± 0.4
PG 1552+085	0.48 ± 0.10	−545 ± 84	0.60 ± 0.02	4	1245 ± 95	44.30 ± 0.01	7.73 ± 0.07	−0.57 ± 0.07	12.7 ± 2.2	⋯	30.7 ± 4.4	⋯
PG 1612+261	0.50 ± 0.06	−203 ± 60	0.54 ± 0.01	1	2108 ± 45	44.41 ± 0.01	8.30 ± 0.02	−1.04 ± 0.02	11.0 ± 1.3	⋯	67.8 ± 5.1	2.5 ± 0.1
PG 2233+134	0.53 ± 0.02	−1052 ± 93	0.74 ± 0.02	3	2224 ± 64	45.12 ± 0.01	8.72 ± 0.03	−0.74 ± 0.03	10.4 ± 0.8	4.4 ± 0.2	8.0 ± 0.3	⋯

Note. Col. (5): Flag for reference lines used to determine the redshift: 1 = [S ii], 2 = [O i], 3 = [O ii], and 4 = Hβ narrow component.

Download table as:  ASCIITypeset images: 1 2

Here we describe how we derive the AGN bolometric luminosity (${L}_{\mathrm{Bol}}$), ${M}_{\mathrm{BH}}$, and Eddington ratio (${L}_{\mathrm{Bol}}/{L}_{\mathrm{Edd}}$), which are compared with the AGN outflow strength in Section 4. The AGN bolometric luminosity is derived from the monochromatic luminosity of the continuum emission at ${\lambda }_{\mathrm{rest}}=5100\,\mathring{\rm A}$, where strong emission lines do not exist. We calculate the 5100A luminosity after subtracting the stellar component by adopting the template from Bruzual & Charlot (2003) since in lower-luminosity AGNs, the contribution of host galaxy emission is significant. We adopt a bolometric correction factor of 9.0 to determine the AGN bolometric luminosity (e.g., Kaspi et al. 2000). The broad component of the Hβ line seen in the SDSS spectrum is used to derive ${M}_{\mathrm{BH}}$. Though sometimes Mg ii and C iv have also been used for estimating ${M}_{\mathrm{BH}}$, we do not use them because our UV spectra are heterogeneously obtained with various instruments. Note that the C iv emission line is widely used to infer the AGN outflow, implying that the C iv velocity profile is largely affected by the AGN outflow (i.e., the C iv-emitting region may not be virialized; Sulentic et al. 2007; Wang et al. 2011; Trakhtenbrot & Netzer 2012). More specifically, we measure ${M}_{\mathrm{BH}}$ from the velocity dispersion of the Hβ broad component and the 5100 Å continuum luminosity by adopting the calibration given by Park et al. (2012a) and Woo et al. (2015). The Eddington ratio is calculated simply from the derived AGN bolometric luminosity divided by the Eddington luminosity, which is calculated using the measured ${M}_{\mathrm{BH}}$. The derived ${M}_{\mathrm{BH}}$ and ${L}_{\mathrm{Bol}}/{L}_{\mathrm{Edd}}$ are given in Table 2 with the velocity dispersion of the Hβ broad component and the 5100 Å continuum luminosity, which are used for calculating ${M}_{\mathrm{BH}}$ and ${L}_{\mathrm{Bol}}/{L}_{\mathrm{Edd}}$.

Uncertainties in the Hβ velocity dispersion and ${M}_{\mathrm{BH}}$ are estimated by performing Monte Carlo simulations. Specifically, we simulate 1000 mock spectra by randomizing the flux with the flux error (see, e.g., Bae & Woo 2014; Woo et al. 2016) and measure the velocity dispersion and ${M}_{\mathrm{BH}}$ from each mock spectrum. Then we take the standard deviations of the measurements as the uncertainties. Finally, the uncertainty in the Eddington ratio is estimated simply by combining the errors in the 5100 Å luminosity and ${M}_{\mathrm{BH}}$. Table 2 shows the derived quantities with the estimated uncertainties, which do not include systematic uncertainties.

The metallicity of gas clouds in the BLR has often been inferred from the flux ratios of N v/C iv and N v/He ii, since the nitrogen relative abundance scales to the square of the metallicity (i.e., N/H $\propto$ ${Z}_{\mathrm{BLR}}^{2}$, or equivalently, N/O $\propto$ O/H $\propto$ ${Z}_{\mathrm{BLR}}$), due to the nature of the nitrogen as a secondary element (see Hamann & Ferland 1999 and references therein). Through intensive calculations of photoionization models, Nagao et al. (2006) proposed an alternative flux ratio as an indicator of ${Z}_{\mathrm{BLR}}$, (Si iv+O iv])/C iv, which has also been used to infer ${Z}_{\mathrm{BLR}}$, especially when the deblending of N v emission from the neighboring Lyα emission is difficult (e.g., Juarez et al. 2009; Matsuoka et al. 2011; Wang et al. 2012). In this study, we examine three indicators of ${Z}_{\mathrm{BLR}}$: N v/C iv, N v/He ii, and (Si iv+O iv])/C iv. These flux ratios of PG QSOs have already been measured by Shin et al. (2013), who carefully decomposed blended lines such as N v. We adopt the same measurement method for the Markarian sample in this study. We present the measured flux—and its uncertainty—of the emission lines used for inferring ${Z}_{\mathrm{BLR}}$ in Table 2. In this table, we do not include the flux of He ii and Si iv+O iv] for some objects because it cannot be measured accurately for those cases, due to their low signal-to-noise ratio.

As described in Section 3.1, the outflow indicators investigated in this work are the AI, VSI, and BAI. The mean values and standard deviations for the AI, VSI, and BAI are (0.51, 0.03), (–258 km s⁻¹, 566 km s⁻¹), and (0.57, 0.12), respectively. For comparison, the median values of those parameters given in Wang et al. (2011) for high-redshift QSOs are 0.5 (AI), −898 km s⁻¹ (VSI), and 0.63 (BAI). While the measured AI and BAI of our sample are similar to those of the high-redshift QSO sample of Wang et al. (2011), the VSI values of our sample are systematically smaller than those of Wang et al. (2011). The reason for this systematic difference may be due to the difference in sample selection (see Section 5.1).

To examine the reliability of the measured BAI and the relation between the BAI and the other indicators (AI and VSI), we compare the three outflow indicators in Figure 2. The different colors of the symbols in this figure denote different narrow lines used for determining the systemic redshift. We do not find any systematically different behaviors for the different-color symbols, suggesting that the difference in narrow lines used for redshift determination does not produce significant systematic effects on the outflow indicators. Therefore, we combine all of the data obtained from different reference lines in the following analysis and discussion. Figure 2 shows that there is a clear correlation between the BAI and the VSI, while there is no apparent correlation between the BAI and the AI. To investigate whether or not there are statistically significant correlations between the outflow indices, we conduct Spearman's rank-order correlation test with the null hypothesis that there are no correlations between the outflow indices. This null hypothesis is not rejected for the relation between the BAI and the AI ($p=2.4\times {10}^{-1}$), while it is rejected for the relation between the BAI and the VSI with a high statistical significance ($p=2.9\times {10}^{-12}$). This result is partly due to the fact that the typical measurement uncertainty in the AI is much larger than the standard deviation of the AI distribution (see the left panel in Figure 2). It is thus inferred that the AI parameter is not an adequate indicator of AGN outflow, at least for our sample. Note that the BAI and VSI are expected to show a correlation since both parameters are defined with velocity shift from the systemic velocity. Thus, it is not surprising to observe a correlation in Figure 2 (left).

Zoom In Zoom Out Reset image size

Here we investigate the relation between outflow indicators (BAI and VSI) and AGN properties. The mean values and standard deviations of ${M}_{\mathrm{BH}}/{M}_{\odot }$, $\lambda {L}_{5100}\ (\mathrm{erg}\,{{\rm{s}}}^{-1})$, and ${L}_{\mathrm{Bol}}/{L}_{\mathrm{Edd}}$ derived in Section 3.2 in logscale are (8.20, 0.61), (45.23, 0.70), and (–1.07, 0.41), respectively. The bolometric luminosity of our sample is located at the bright part of the optical AGN luminosity function in the local universe (see, e.g., Boyle et al. 2000; Schulze et al. 2009). As for the Eddington ratio, our sample shows a wide range (∼1.5 dex), $-2.0\lesssim {L}_{\mathrm{Bol}}/{L}_{\mathrm{Edd}}\lesssim -0.5$. Thus, our sample allows us to investigate the AGN outflow in a broad dynamic range of the Eddington ratio. Note that the Eddington ratio of the sample of Wang et al. (2011) is distributed in the range of $-1.0\lesssim {L}_{\mathrm{Bol}}/{L}_{\mathrm{Edd}}\lesssim +0.5$, i.e., ∼1 dex higher than the Eddington ratio distribution of our sample. This difference is a natural consequence of the sample selection in the sense that high-redshift SDSS quasars are luminous enough to be selected in the magnitude-limited SDSS spectroscopy. The systematically smaller VSI in our sample than in the sample of Wang et al. (2011) mentioned in Section 4.1 is probably caused by this ∼1 dex difference in Eddington ratio distribution between the two samples.

In Figure 3, the possible dependence of the two outflow indicators (BAI and VSI) on the AGN properties (${M}_{\mathrm{BH}}$, ${L}_{\mathrm{Bol}}$, and ${L}_{\mathrm{Bol}}/{L}_{\mathrm{Edd}}$) is investigated. Note that investigation of the relation between the AI and those AGN properties is not useful since the dynamic range of the AI in our sample is too narrow (see Section 4.1 and Figure 2). Figure 3 shows that there is no apparent correlation between ${M}_{\mathrm{BH}}$ and the outflow indicators. However, in the center and right panels, the two outflow indicators appear to show positive correlations with the AGN bolometric luminosity and Eddington ratio, in that AGNs with a higher luminosity or a higher Eddington ratio tend to show stronger outflows. To examine the statistical significance of these possible correlations, we conduct Spearman's rank-order correlation test with the null hypothesis that there are no correlations in Figure 3. The rank-order test suggests that there is no correlation between the two outflow indicators (BAI and VSI) and ${M}_{\mathrm{BH}}$ ($p=1.6\times {10}^{-1}$ and $p=1.3\times {10}^{-1}$). It is also suggested that the inferred dependence of the BAI and VSI on the AGN bolometric luminosity and Eddington ratio is marginal, but the VSI shows more significant correlation with those two AGN parameters ($p=4.7\times {10}^{-3}$ and $p=1.1\times {10}^{-3}$) than the BAI ($p=1.9\times {10}^{-2}$ and $p=1.1\times {10}^{-2}$). These trends are qualitatively consistent with a previous study of high-redshift QSOs by Wang et al. (2011).

Zoom In Zoom Out Reset image size

The mean values and standard deviations of N v/C iv, (Si iv+O iv])/C iv, and N v/He ii derived in Section 3.3 in logscale are (–0.41, 0.29), (–0.65, 0.23), and (0.57, 0.22), respectively. As we mentioned in Section 3.3, the values are from Shin et al. (2013), and they are similar to those of previous works (e.g., Shemmer et al. 2004; Matsuoka et al. 2011).

Now we examine how the two outflow indicators (BAI and VSI) depend on the BLR metallicity. Figure 4 shows a comparison between the outflow indicators and BLR metallicity indicators. There are loose positive correlations in that AGNs with higher BLR metallicity tend to show stronger outflow, but the significance looks marginal. To examine the statistical significance of those possible correlations, we conduct Spearman's rank-order correlation test with the null hypothesis that there are no correlations in Figure 4. The rank-order test suggests that the correlations between the BAI and the metallicity indicators are very marginal with low statistical significance ($p\sim 0.01\mbox{--}0.17$). The statistical significance of correlations between the VSI and the metallicity indicators is slightly higher ($p\sim 0.003\mbox{--}0.07$) than that for the BAI, but the correlations are not very close. These results seem contradictory to the earlier results for high-redshift QSOs reported by Wang et al. (2012), who showed a close correlation between outflow strength and metallicity in BLRs. For this comparison, it should be noted that the analysis by Wang et al. (2012) is based on stacked SDSS spectra of 12844 high-redshift QSOs, while our analysis is based on individual SDSS spectra of 34 low-redshift AGNs. The significantly larger number of objects in the study of Wang et al. (2012) makes the statistical error very small.

Zoom In Zoom Out Reset image size

To investigate whether our results are consistent with those of Wang et al. (2012), we present the relation between the BAI and the (Si iv+O iv])/C iv high-redshift QSOs in Figure 4 (dashed line in the upper middle panel). Although the dispersion in our sample is large, the results of Wang et al. (2012) and ours show similar trends. This is interesting since the difference in cosmic age between our sample at $z\sim 0.1$ and the sample of Wang et al. (2012) at $1.7\lt z\lt 4$ is more than half of Hubble time, but still these two samples show similar trends in the relation between the outflow indicator and metallicity indicator.

As described in Section 4.2, the AI parameter does not work well for quantifying the outflow strength for our sample. This is attributed to the narrow range of the AI distribution. It is interesting to notice that highly asymmetric C iv lines are selectively seen in radio-loud QSOs (Sulentic et al. 2000, 2007). The small range of the AI distribution in our sample may be attributed to the fact that there are not so many radio-loud AGNs in the PG sample (Kellermann et al. 1989). Wang et al. (2011) reported a significant correlation between the AI and BAI, based on their high-redshift SDSS QSO sample, whose AI is distributed in the range of $0.3\lesssim \mathrm{AI}\lesssim 0.8$. However, the AI distribution of their sample concentrates mostly in the range of $0.4\lesssim \mathrm{AI}\lesssim 0.6$, and the wide AI range is in fact achieved owing to the large sample size. Therefore, our small sample size (which lacks radio-loud AGNs in particular) is not adequate for investigating the AI distribution and various AI-related correlations. As for the BAI and VSI parameters, the VSI shows closer correlations with the AGN properties, including the BLR metallicity, than the BAI, as shown in Sections 4.2 and 4.3. This is possibly caused by the large uncertainty in the AI, which introduces a corresponding uncertainty in the BAI.

Another potential concern regarding our sample selection is the removal of BAL AGNs. BAL features are caused by powerful outflows of ionized gas, and thus AGNs with a strong outflow may be selectively removed from our sample. This concern may not be valid if fast ionized outflow is ubiquitous in AGNs and the BAL features are observed just due to the orientation effect (i.e., we see BAL features only in ∼10% of type 1 AGNs because the covering factor of the ionized outflow is ∼10%; see, e.g., Weymann et al. 1991). It is interesting to compare the outflow indices (AI, VSI, and BAI) between BAL AGNs and non-BAL AGNs to understand the nature of the BAL phenomenon in AGNs; however, we do not further discuss this issue because it is beyond the scope of this work.

The correlations between the outflow, metallicity, and Eddington ratio seen in our sample suggest their physical connection in low-redshift AGNs. They suggest that gas radial motion (i.e., gas accretion onto the SMBH and outflow from the nucleus) is related to the metallicity in the nuclear region. The relation between metallicity and gas accretion has been reported for QSOs (e.g., Shemmer et al. 2004) and for nearby Seyfert galaxies, especially with regard to the metallicity difference between narrow-line Seyfert 1 galaxies and broad-line Seyfert 1 galaxies (e.g., Nagao et al. 2002; Shemmer & Netzer 2002; Fields et al. 2005). This can be interpreted as the result of the starburst-AGN connection; i.e., nuclear starburst activity enriches the gas in the nuclear region and triggers gas accretion onto the SMBH. On the other hand, the relation between AGN outflow and the Eddington ratio has also been discussed: gas accretion results in the increase of AGN luminosity, which exerts outward radiative pressure on the surrounding gas (e.g., Komossa et al. 2008; Wang et al. 2011; Marziani et al. 2013). Therefore, correlations between the outflow, metallicity, and Eddington ratio are reasonable characteristics of AGNs. In addition, a more direct reason for the positive correlation between the gas metallicity and the outflow is that gas with high metallicity (and consequently with high dust abundance) has greater optical depth. Therefore, such metal-rich clouds are more easily affected by the outward radiative pressure. Also, resonant scattering could contribute to the outflow. If resonant scattering exerts radiation pressure, higher-metallicity gas has greater optical depth in the resonance scattering, which leads to more powerful outflow. Interestingly, our results for low-redshift (z < 0.4) QSOs and those for high-redshift (1.7 < z < 4) QSOs show similar trends, even though these two studies examine completely different cosmic epochs. This suggests that the physics of gas accretion onto an SMBH and its relation to nuclear star formation are universal through the cosmic timescale.

On the other hand, the SMBH mass does not show a significant correlation with the outflow. The correlation coefficients are 0.25 for low-redshift QSOs and −0.39 for high-redshift QSOs (Wang et al. 2011). This implies that the SMBH mass itself is not an important factor for nuclear activity, such as nuclear star formation and outflow. This interpretation is consistent with the recent work by Shin et al. (2013), who report that there is no statistically meaningful correlation between the SMBH mass and the BLR metallicity.

Powerful AGN outflow is sometimes also seen in the NLR. Komossa et al. (2008) showed that [O iii] emission (a relatively high-ionization NLR line) in some AGNs shows a blueshift with respect to other, low-ionization narrow lines, such as [S ii]. Since it is known that high-ionization NLR clouds (emitting [O iii]) locate more inner parts in the NLR than low-ionization NLR clouds emitting [S ii], [O i], and [O ii], for example (Ferguson et al. 1997; Nagao et al. 2001), it is naturally expected that high-ionization NLR lines more frequently show outflow features than low-ionization NLR lines (e.g., Bae & Woo 2014). Then a question naturally arises—is there any physical connection between BLR outflows and NLR outflows? In terms of the gas metallicity, some earlier works reported a physical connection between gas clouds in NLRs and BLRs (e.g., Fu & Stockton 2007; Du et al. 2014). Those works inferred that the outflowing BLR gas affects the chemical property of the NLR. Therefore, it is highly interesting to examine the possible physical link between BLR outflow and NLR outflow. This can be studied by investigating the outflow indices for C iv and those for [O iii].

To investigate the possible link between BLR outflow and NLR outflow, we focus on the outflow indices for [O iii] and compare them with the outflow indices for C iv in our low-redshift AGN sample. Figure 5 shows the results of the comparison between the VSI for [O iii] and that for C iv. Here we derive the VSI by adopting different references for the systemic redshift—Hβ, [O i], [O ii], and [S ii]—to check the robustness of the inferred results. There is no apparent correlation between the VSIs of [O iii] and C iv, regardless of the adopted redshift reference. Spearman's rank-order correlation test also suggests that there is no statistically significant correlation between them (the exact results of the rank-order correlation test are given in Figure 5). However, the obtained results do not necessarily mean that the BLR outflow and NLR outflow are independent of each other, because the sample size of our low-redshift AGNs is so small that the covered range of the [O iii] VSI is very limited (from –150 to +250 km s⁻¹ roughly). Studies of [O iii] in the literature typically focus on the wing component, which represents the non-gravitational kinematic signature, finding evidence of strong outflows (Bae & Woo 2016; Woo et al. 2016, 2017). On the other hand, the velocity shift of [O iii] is reported to be relatively small (Bae & Woo 2014; Woo et al. 2016). Thus, our comparison of BLR and NLR outflows based on the velocity shift may not reveal a strong trend. A comparison of the outflow kinematics between [O iii] and C iv for much larger samples with detailed line profile analysis will be an interesting test to understand the possible link between BLR and NLR outflows.

Zoom In Zoom Out Reset image size