KILOPARSEC-SCALE RADIO STRUCTURES IN NARROW-LINE SEYFERT 1 GALAXIES

We searched for extended radio emissions by using the VLA FIRST images (version 2008 July 16) at 1.4 GHz with a ∼5'' resolution around NLS1s from three previously reported samples, including those of (i) Zhou et al. (2006), who presented 2011 optical-selected NLS1s at z ≲ 0.8 from the Sloan Digital Sky Survey (SDSS) 3rd data release (DR3; Abazajian et al. 2005); (ii) Yuan et al. (2008), who presented 23 optically selected, very radio-loud (R > 100) NLS1s from the updated version of Zhou et al. (2006) by using SDSS 5th data release (DR5; Adelman-McCarthy et al. 2007) and FIRST images; and (iii) Whalen et al. (2006), who presented 62 radio-selected NLS1s at z = 0.065–0.715 from the FIRST Bright Quasar Survey (FBQS; White et al. 2000). Some fractions of objects are listed redundantly. Because the conventional NLS1 definition of FWHM(Hβ) < 2000 km s⁻¹ was not strictly applied in any of the three NLS1 catalogs (see original papers for details), we applied the definition in this paper to obtain 1784, 17, and 47 objects from samples (i), (ii), and (iii), respectively.

We started searching for resolved radio structures by selecting FIRST-cataloged radio sources on the basis of two criteria: (1) position, <10'' in radius from the optical positions; and (2) intensity, >3 mJy beam⁻¹ in a typical image noise of 0.15 mJy beam⁻¹. To the FIRST images for the selected sources, we imposed two additional criteria: (3) the presence of extended structures associated with FIRST-cataloged radio emission determined through visual inspection and (4) the absence of relation to any possible optical or infrared counterpart for extended emission, according to the NASA/IPAC Extragalactic Database (NED). As a result, we detected three sources: PMN J0948+0022, SDSS J145041.93+591936.9, and FBQS J1644+2619. These radio sources appear to be unresolved central components that are accompanied by extended emissions.

To prevent neglecting coreless or weak-core radio galaxies and large radio galaxies with isolated radio lobes at sites distant from host galaxies, we also inspected FIRST cutout images centered at the optical positions of all NLS1 samples with the following parameters: (1) field of view, 5 × 5 arcsec²; (2) intensity, >3 mJy beam⁻¹; (3) structure, emissions symmetrically located with respect to the central position or an isolated jet/lobe-like feature pointing back to the central position as determined through visual inspection; and (4) no relation to any possible optical or infrared counterpart, according to NED. As a result, we detected one source, SDSS J120014.08−004638.7, which appears to have symmetrically located emissions without a central component.

Here, we mention cases of suspected emissions that have been subsequently exonerated. We found isolated emissions separated by ∼38'' and ∼66'' from the unresolved cores of FBQS J075800.0+392029 and FBQS J1448+3559, respectively; no optical/infrared counterparts were detected. However, we determined that these emissions are unrelated because deeper images at higher spatial resolutions by our analyses of VLA archival data at 1.4 and 5 GHz (project code: AL450, AG151, and AB982) showed no suggestive radio structure. A radio source FIRST J094901.5+002258 without an optical/infrared counterpart was located at a separation of 71'' at a position angle of 62° from PMN J0948+0022, which is one of the detected NLS1s. This source was previously reported by Komossa et al. (2006) and Foschini et al. (2010), who recognized it as an unrelated source; we determined the same result on the basis of our analyses of archival VLA data (AD489 and AK360).

For FBQS J1644+2619, one of our detected sources, archival VLA data of deeper imaging at a spatial resolution higher than that of the FIRST image were available (AP0501), which allowed us to perform more detailed investigations. We retrieved and reduced all the available archival VLA data⁷ obtained at relatively high spatial resolutions for FBQS J1644+2619. Data reduction and imaging procedures were performed by using the Astronomical Image Processing System (AIPS) according to standard operating procedures. The AP0501 data obtained by using the A-array configuration at 1.4 GHz provided the deepest image of rms noise of 1σ = 90 μJy beam⁻¹ at a resolution of 15. Flux densities of each component were measured by Gaussian-profile fitting with the AIPS task JMFIT. Errors were determined by rms of the fitting error and an amplitude calibration error of 5%.

3.1.1. New Detections of kpc-scale Radio Structures in NLS1s

3.1.2. Previously Known Source with kpc-scale Radio Structures

In this section, we use two control samples to examine whether the detection rate of the extended radio sources in NLS1s differs from that in broad-line AGNs (hereafter referred to as BLS1s) in a statistical sense. The control sample of BLS1s was created on the basis of a sample presented by Rafter et al. (2011) that found 63 sources with extended radio emissions by matching 8434 SDSS-DR4-based low-redshift type-1 AGNs (Greene & Ho 2007) to FIRST images in a manner similar to our method based on visual inspection (Section 2.1). Its criteria were (a_B) z < 0.35, (b_B) >1 mJy at radio, and (c_B) radio extended >4'' from optical position. The NLS1 control sample was created on the basis of our results that used the 1784 SDSS-DR3-based NLS1s from Zhou et al. (2006; sample (i); Section 2.1). Its criteria were (a_N) z ≲ 0.8, (b_N) FWHM(Hβ) < 2000 km s⁻¹, and (c_N) >3 mJy beam⁻¹ at radio.

From these two parent samples, we obtained two subsamples to meet the following common criteria: (1) z < 0.35, (2) >3 mJy beam⁻¹ at radio, (3) radio extended >4'' from optical position, and (4) FWHM(Hα) > 1835.4 km s⁻¹ to the former parent sample for selecting BLS1s and FWHM(Hβ) < 2000 km s⁻¹ to the latter parent sample for selecting NLS1s. In criterion (4), the threshold of FWHM(Hα) = 1835.4 km s⁻¹ is equivalent to FWHM(Hβ) = 2000 km s⁻¹, according to an empirical relationship of FWHM(Hβ) = 1.07 × 1000 (FWHM(Hα)/1000 km s⁻¹)^1.03 km s⁻¹ based on type-1 AGNs in SDSS DR3 at z < 0.35 (Greene & Ho 2005). Because the optical catalog used in Rafter et al. (2011) includes only the values of FWHM(Hα) (Greene & Ho 2007), we used this equivalent threshold on line widths for the BLS1 control sample.

When these criteria are strictly applied, the numbers of extended radio sources in the two control samples are 58 in 6868 BLS1s and 1 in 1000 NLS1s. The one NLS1 with an extended radio structure is SDSS J120014.08−004638.7. PMN J0948+0022 is definitely excluded by (1). SDSS J145041.93+591936.9 is excluded, although marginally, by (3) because the radio extension is only ∼38 from the optical position (Section 3.1.1). FBQS J1644+2619 is excluded because it is absent in the SDSS-DR3-based NLS1 sample of Zhou et al. (2006; sample (i)) and has appeared in SDSS DR4. Therefore, this object is listed in the SDSS-DR5-based very radio-loud NLS1 sample of Yuan et al. (2008; sample (ii)). The previously reported NLS1s with kpc-scale radio structures 1H 0323+342 (Antón et al. 2008) and PKS 0558−504 (Gliozzi et al. 2010) are outside the FIRST and SDSS sky coverages, respectively. Thus, considering the two marginal cases (SDSS J145041.93+591936.9 and FBQS J1644+2619), the number of extended radio sources could be 1–3 in 1000 NLS1s.

As a result, the detection rates of extended radio emissions are 58/6868 (0.84%) and 1–3/1000 (0.10%–0.30%) in BLS1s and NLS1s, respectively. Fisher's exact test rejects a null hypothesis that the detection rate in NLS1s is not lower than that in BLS1s with a significance level of 0.05 in any of the cases, because of probabilities of p = 0.0030–0.0395. Therefore, the detection rate of extended radio emissions in NLS1s is lower than that in BLS1s in a statistical sense on the basis of our two SDSS–FIRST control samples at z < 0.35.

Five of the six NLS1s exhibit two-sided structures at kpc scales (Column 2 in Table 2). On the other hand, from VLBI observations, one-sidedness in pc scales is evident in four of the five NLS1s: PMN J0948+0022 (Doi et al. 2006; Giroletti et al. 2011), 1H 0323+342 (Lister & Homan 2005 in the MOJAVE⁸ project), PKS 0558−504 (Gliozzi et al. 2010), and FBQS J1644+2619 (Figure 2(c)). The one-sided structure of FBQS J1644+2619, which can be interpreted as a consequence of the Doppler-beaming effect if jets are not intrinsically asymmetric, provided mild constraints of β_pc > 0.74 and Φ < 42°, where β is the jet speed in the unit of the speed of light and Φ is the viewing angle (Doi et al. 2011). In addition, a flat spectrum and highly variable core (Figure 2(d)), as well as a significant polarized radio flux (3.45 ± 0.46 mJy) reported in the National Radio Astronomy Observatory VLA Sky Survey (NVSS; Condon et al. 1998), indicate the presence of a blazar-like core for FBQS J1644+2619. The combination of one-sided morphology in pc scales and a two-sided structure in kpc scales is suggested by the concepts of beamed core and radio lobes as decelerated components. Different viewing angles at different geometric scales without deceleration is an alternative explanation; however, it is more possible that the viewing angle remains nearly constant because the position angles of radio morphology from pc to kpc scales are constant. Assuming intrinsically symmetric two-sided ejecta, steady jet activity, and beaming considerations (Ghisellini et al. 1993), we can estimate the speed of a lobe of β_kpc = 0.27–0.36 at Φ < 42° by using the apparent flux ratio of approaching and receding hot spots given as R_F = [(1 + βcos Φ)/(1 − βcos Φ)]^{3 − α} and by applying observed flux densities of 16.9 mJy and 1.8 mJy and a spectral index of α = −1.1 (Figure 2(D)). It should be noted that this method assumes no variation in the luminosities of the two components during a differential time delay. As an additional approach, on the basis of a geometric consideration of the differential time delay, we estimated β_kpc = 0.09–0.12 at Φ < 42° by using the apparent extent ratio of approaching to receding lobes given as R_D = (1 + βcos Φ)/(1 − βcos Φ) and by applying observed separations of 118 and 100. Therefore, a deceleration in β is required from the fast (β_pc > 0.74) jets in pc scales to the significantly slower (β_kpc = 0.09–0.36) radio lobes in kpc scales; a stronger constraint on Φ is obtained in Section 4.2.

Similarly, for PMN J0948+0022, β_pc > 0.76 and Φ < 22° were determined from the very high brightness temperature of core-dominated one-sided jets in the VLBI image (Doi et al. 2006). On the contrary, the apparent flux ratio R_F of the northern (2.2 mJy) and southern (0.8 mJy) components in kpc scales (Section 3.1.1) requires a deceleration down to β_kpc = 0.13–0.15 assuming α = −0.7 and the same range of Φ at the pc scales; a stronger constraint on Φ is obtained in Section 4.2. Significant radio flux variability, polarized radio emissions, flat/inverted radio spectra (Abdo et al. 2009b), and γ-ray detections (Abdo et al. 2009a) have also been reported for PMN J0948+0022. 1H 0323+342 is also a γ-ray-emitting NLS1 (Abdo et al. 2009c) in the NLS1s with kpc-scale radio structures. These observed properties indicate the presence of blazar-like beamed jets at pc scales in these NLS1s, whereas kpc-scale ones are rarely beamed.

Except the lobe-dominated case of SDSS J120014.08−004638.7, all NLS1s with kpc-scale radio structures exhibit a core with significantly higher luminosity than that of extended emissions (Column 8 in Table 2). These core dominances are unusually high compared to typical radio galaxies, which are lobe-dominated with respect to cores (the median values are 0.022 for FR-I and 0.003 for FR-II radio galaxies; Morganti et al. 1997). Similarly, blazars frequently show a prominent core that accompanies the extended radio structures of low brightness in deep VLA images (Murphy et al. 1993; Cassaro et al. 1999; Landt et al. 2006; Kharb et al. 2010). The possible physical origin of the high core dominances is Doppler boosting only in cores (Section 4.1), unless the jet activity has undergone extreme variations. In this subsection, we determine the Doppler factors required from the observed core dominances of these NLS1s. In addition, we constrain jet speeds and viewing angles at pc scales (Sections 4.2) and estimate deprojected sizes and kinematic ages of kpc-scale structures (Sections 4.3 and 4.5).

Two methods can be used to derive an intrinsic core power. The first uses an empirical correlation in radio galaxies given by log P_c = 0.62log P_t + 7.6, where P_c is the core radio power at 5 GHz and P_t is the total radio power at 408 MHz in W Hz⁻¹, with a scatter of 1.1 dex in P_c (Giovannini et al. 2001). Because the total radio power is measured at a low frequency and is therefore not affected by Doppler boosting, an expected intrinsic core power can be determined from the total radio power. The second method uses the fundamental plane of black hole activity given by log L_R = 0.60log L_X + 0.78log M_BH + 7.33, where L_R, L_X, and M_BH are the 5 GHz unbeamed core luminosity, 2–10-keV X-ray luminosity in erg s⁻¹, and black hole mass in M_☉, respectively, with a scatter of 0.88 dex in L_R (Merloni et al. 2003). This relationship was determined by using the samples of X-ray binaries, nearby low-luminosity AGNs, Seyfert galaxies (including seven NLS1s), and quasars. The key is that intrinsic core contributions expected from these two individual methods (r^core/lobe_int. ∼ 0.01–0.2) are much smaller than the observed values (r^core/lobe_obs. = 2.1–36.1), with the exception of SDSS J120014.08−004638.7 (Columns 8–10 in Table 2). These crucial differences can be interpreted as a Doppler-beaming effect only in cores. Some apparent discrepancies in r^core/lobe_int. between the two methods could be due to such factors as scatters on the empirical relations, luminosity variabilities, or the uncertainties of black hole masses.

We estimated a core Doppler factor of δ^core by comparing intrinsic and observed core luminosities, assuming that the core flux density is boosted by (δ^core)^{3 − α}, where $\delta = \sqrt{1-\beta ^2} (1-\beta \cos {\Phi })^{-1}$ (Rybicki & Lightman 1979), and a flat (α = 0) spectrum if α is not available.⁹ Subsequently, we can obtain the constraints of jet speed β^core and viewing angle Φ^core for each source (Table 3). Thus, all cases showing unusual core dominances require nearly pole-on-viewed jets with at least mildly or highly relativistic speeds. It is noteworthy that these jet parameters are consistent with those obtained through SED modeling for the two γ-ray-emitting NLS1s: PMN J0948+0022 and 1H 0323+342 (Abdo et al. 2009a, 2009c). Therefore, the existence of radio lobes in these NLS1s is essential as places for dissipation of the huge amount of kinetic powers of the strong jets into thermal reservoirs and radiation (Scheuer 1974; Begelman & Cioffi 1989). Conversely, the lobe-dominated structure of SDSS J120014.08−004638.7 indicates the existence of a radio galaxy in the NLS1 population as well. Thus, these NLS1s with kpc-scale radio structures can be understood in the framework of the unified scheme of radio-loud AGNs that considers radio galaxies as non-beamed parent populations of blazars (Antonucci & Ulvestad 1984, 1985; Urry & Padovani 1995), although only a small number of objects have been studied in detail thus far.

While the detected NLS1s are very radio loud (log R_obs. ≳ 1.5), the intrinsic radio loudness log R_int. based on the extended radio luminosity and the unbeamed core luminosity allows us to infer the property of their parent populations as inclined NLS1 radio sources (Table 2). PMN J0948+0022 and PKS 0558−504 should no longer be very radio loud at the rest frame (log R_int ∼ 1). In contrast, FBQS J1644+2619, SDSS J120014.08−004638.7, and 1H 0323+342 remain significantly radio loud at the rest frame (log R_int. ≳ 1.5) because of their luminous radio lobes. Thus, NLS1s with kpc-scale radio structures are probably provided from the populations of both intrinsically radio-intermediate and radio-loud objects rather than from only a radio-quiet population.

The projected radio sizes on one side of the NLS1s with extended emissions are ∼10–50 kpc (Column 4 in Table 2), which are at the smaller end of the size distribution of known radio galaxies and blazars (∼10–10³ kpc; Landt et al. 2006, and references therein) but are very large in the unit of Schwarzschild radius (Rs; Column 5 in Table 2). These sizes are equivalent to the largest known radio galaxies of several Mpc (e.g., Machalski et al. 2008), which correspond to the several × 10¹⁰ Rs for assumed black hole masses of ∼10⁹ M_☉ typically found in radio galaxies. Deprojected sizes are possibly larger than 100 kpc for high core dominances, i.e., nearly pole-on-viewed cases (Column 5 in Table 3). Thus, their radio structures have developed into intergalactic environments outside the host galaxies. In this subsection, we discuss the successful escape of radio structures of the detected NLS1s to kpc scales and the cause of the apparently compact structures of most typical NLS1s.

The majority of the NLS1s with kpc-scale structures found so far show FR-II-like radio lobes (FBQS J1644+2619, SDSS J120014.08−004638.7, 1H 0323+342, and PKS 0558−504) and appear to include edge-brightened lobes with r_s > 0.8 if their high core dominances are neglected. The edge-darkened (FR-I)/edge-brightened (FR-II) morphology reflects the subsonic/supersonic speed of their lobes. Because of strong deceleration in the nuclear dense region, an initial jet speed much faster than 0.1c is expected for progression up to kpc scales without disrupting hot spots (Kawakatu et al. 2008). Considering deceleration due to the growth of the effective cross-sectional area of lobes, the FR-I/FR-II dichotomy is determined by the ratio of jet kinetic power to ambient density $L_\mathrm{j}/\bar{n}_\mathrm{a}$ (Kaiser & Best 2007; Kawakatu et al. 2009). A threshold exceeding $L_\mathrm{j} /\bar{n}_\mathrm{a}=10^{44}$–10⁴⁵ erg s⁻¹ cm⁻³ is required to extend supersonic lobes beyond a core radius, where L_j is the jet kinetic power and $\bar{n}_\mathrm{a}$ is the ambient number density at the core radius (∼1 kpc; Kawakatu et al. 2009). As a basic topic of discussion, considering the case of $\bar{n}_\mathrm{a} \approx 0.1$–1 cm⁻³ (cf. n ∼ 0.1 cm⁻³ at the center of ellipticals; e.g., Mathews & Brighenti 2003; n ≈ 1 cm⁻³ for local interstellar medium in our Galaxy; Cox & Reynolds 1987), L_j ≳ 10⁴⁴ erg s⁻¹ is required. We estimate L_j for the NLS1s with kpc-scale radio structures using their 1.4 GHz lobe luminosities according to the relationship between jet kinetic and radio powers (Cavagnolo et al. 2010; see also, e.g., Willott et al. 1999; O'Sullivan et al. 2011; Merloni & Heinz 2007). As a result, L_j ≳ 10⁴⁴ erg s⁻¹ is estimated for all sources except SDSS J145041.93+591936.9, which shows a marginal radio structure (Column 13 in Table 2). Thus, the observed extended radio emissions actually suggest sufficient jet kinetic powers for escaping to kpc scales in the form of supersonic lobes.

The jet kinetic powers could be limited by the low-mass black holes of most typical NLS1s if the kinetic powers are supplied with accretion that is limited by the Eddington luminosity. As a basic point of discussion, we offer the case of L_j ∼ 0.1 L_Edd (Ito et al. 2008), where L_Edd = 1.3 × 10³⁸ (M_BH/M_☉) erg s⁻¹ is the Eddington luminosity. To maintain supersonic radio lobes at kpc scales, black hole masses must be M_BH ≳ 10⁷ M_☉ in order to generate L_j ≳ 10⁴⁴ erg s⁻¹. Figure 3 shows histograms of black hole masses for the control samples of NLS1s, BLS1s, and other reference samples that were defined in Sections 2.1 and 3.2. The masses have been computed by using the virial relationship of the broad-line component (e.g., Kaspi et al. 2000); we employed a standard single-epoch procedure that uses the relationship between FWHM(Hβ) and optical continuum luminosity (L_5100 Å) or between FWHM(Hα) and broad-line luminosity (L_Hα) given by Equation (5) or (6) in Greene & Ho (2005). In the distributions, all of the NLS1s with kpc-scale radio structures are driven by relatively high mass black holes of M_BH ≳ 10⁷ M_☉, except for SDSS J145041.93+591936.9, which shows a marginal radio structure. These objects appear to constitute a peculiar group at the high-mass end of the distributions among the NLS1 populations and can exceed the threshold of $L_\mathrm{j}/\bar{n}_\mathrm{a}$. On the contrary, the black hole masses of most typical NLS1s (≲ 10⁷ M_☉) suggest below the threshold. In these systems, even if jets are emanated, hot spots would be disrupted into edge-darkened (FR-I) radio lobes, which may tend to go undetected in the limited image sensitivity of FIRST (Punsly & Zhang 2011), as is the case with the discovery of FR-I morphology in the radio-quiet quasar E1821+643 exclusively in a deep VLA image (Blundell & Rawlings 2001). The lower detection rate is not due to an observational limitation specific to NLS1s because the control sample of BLS1s was also investigated using FIRST images (Section 3.2) and shares the same difficulty in detection of FR-I morphology. Therefore, lower maximal jet kinetic powers due to lower mass black holes may be related to the lower detection rate of the extended radio structures in the NLS1 population compared with those in the broad-line AGNs. The radio emissions in most typical (radio-quiet) NLS1s are compact (Ulvestad et al. 1995) and are dominated by diffuse components confined within central regions in VLBI images (Middelberg et al. 2004; Orienti & Prieto 2010; Giroletti & Panessa 2009; Doi et al. 2012), which may be relevant to the link between apparent radio size and black hole mass. Furthermore, the link between the radio size and Eddington ratio may also be a contributor, since only a small fraction of accretion luminosity is expected to be transferred to jet power in high Eddington ratio systems such as NLS1s, according to the nonlinearity in an empirical relation that is observed in radio galaxies: log L_j/L_Edd = 0.49log L_bol/L_Edd − 0.78 (Merloni & Heinz 2007). This may be another reason for the lower detection rate of the extended radio structures in NLS1s.

Zoom In Zoom Out Reset image size

It should be noted that the previous discussion considers the following three caveats. First, the adopted threshold of $L_\mathrm{j}/\bar{n}_\mathrm{a}$ was based on tentative values of jet kinetic power and ambient density. To clarify the FR-I/II dichotomy in kpc-scale radio structures of NLS1s, a study of dynamical interaction between the radio lobe and cocoon is crucial (Kaiser & Alexander 1997; Kino & Kawakatu 2005; Ito et al. 2008) and will be explored in future research.

Second, an alternative possibility is that the apparent compact structures in most cases, even for radio-loud NLS1s, are attributed to short-lived radio sources of <10⁴ pc (≲ 10⁶ years). VLBI images have been obtained for several compact radio-loud NLS1s, including RX J0806.6+7248, FBQS J1629+4007, RX J1633.3+4718, B3 1702+457 (Doi et al. 2007, 2011; Gu & Chen 2010), PKS 1502+036 (Fey & Charlot 2000; Dallacasa et al. 1998), and SBS 0846+513 (Taylor et al. 2005; Kovalev et al. 2007). However, no clear evidence of young radio lobes has yet been detected in pc scales (cf. Gallo et al. 2006, who proposed the NLS1–compact-steep-spectrum (CSS) connection on the radio-loud NLS1 PKS 2004−447; see also Orienti et al. 2012).

Third, the estimation of black hole masses through the virial relationship is controversial. The impact of this factor on our study is discussed in the following subsection.

The virial mass estimation is subject to the assumption of an isotropic geometry of a broad-line region (BLR) with random orbital inclination of clouds that are gravitationally bound to a black hole (Netzer 1990). The virial assumption could be violated in sources with high Eddington ratios because the outward force due to radiation pressure overcomes or compensates for gravitational attraction (Marconi et al. 2008; but see also Netzer 2009; Marconi et al. 2009). The mass deficit due to this effect indicates that AGNs with high Eddington ratios would be preferentially classified into NLS1s. However, such a situation may be inconsistent with the observed lower detection rate of extended radio emissions in NLS1s (Section 3.2) because large jet kinetic powers can generally be expected from such highly mass-accreting systems with larger black hole masses. If the geometry of BLR is flat, a narrow line would be the result of a small inclination and the mass would be underestimated (e.g., Decarli et al. 2008; La Mura et al. 2009). In fact, planar BLRs in blazars have been suggested (Decarli et al. 2011, and references therein; but see Punsly 2007). The mass deficit due to this effect indicates that AGNs viewed face-on would be preferentially classified into NLS1s. However, such a situation may be inconsistent with the observed lower fraction of radio-loud objects in NLS1s compared with BLS1s (Komossa et al. 2006; Zhou et al. 2006) in the framework of the Doppler-beaming effect. That is, these two possible mechanisms for inducing mass deficit do not manifest in the actual radio views of NLS1s as a population. Small masses have been presumed for NLS1s as a class on the basis of several other methods, e.g., an accretion-disk model fit (by fitting the SEDs with the disk and corona model; Mathur et al. 2001) and the analysis of X-ray variability power density spectra (Nikolajuk et al. 2004).

From the perspective of the present study, the planar BLR may influence individual objects of NLS1s with unusually high core dominances in particular, which indicate very small viewing angles (Section 4.2). The black hole masses of these blazar-like NLS1s (≳ 10⁷ M_☉ in a virial method) might be underestimated: their true masses may be even larger (≳ 10⁸ M_☉). If this is the case, our conclusion that NLS1s with kpc-scale radio structures have relatively large black hole masses remains. We note the existence of non-beamed radio-loud NLS1s that may be inclined, which include the lobe-dominated NLS1 SDSS J120014.08−004638.7 with M_BH = 10^7.4 M_☉ (Section 3.1.1) and several steep-spectrum radio sources in radio-loud NLS1s (Komossa et al. 2006; Gallo et al. 2006; Doi et al. 2011); intrinsic radio and optical natures are observed in these sources.

The ages of NLS1s with kpc-scale radio structures provide important clues of integrated activity of a rapidly growing black hole at the lower end of the AGN mass function. We define the kinematic age as t^kpc = L^deproj./β^kpcc, where L^deproj. is the deprojected size at kpc scales assuming the same viewing angle as that at the core region (Φ^core). Assuming β_kpc ∼ 0.01–0.3 at kpc scales (Column 7 in Table 3) as constant speeds of advance, jet activity should be maintained for ≳ 10⁷ years to build the entire radio structures (Column 8 in Table 3). For SDSS J120014.08−004638.7 and SDSS J145041.93+591936.9, strong constraints on t^kpc cannot be obtained because the upper limits of β^kpc are not available from their coreless and one-sided jet structures, respectively.

In comparison with Galactic black holes (Done et al. 2007, for a review), the accretion disks of NLS1s with kpc-scale radio structures could be in a very high state, which is a transitional state at very high accretion rates between low/hard and high/soft states, characterized by powerful transient relativistic ejections (Gliozzi et al. 2010). The derived kinematic ages of ≳ 10⁷ years for these NLS1s may imply the sustenance ability of the very high state on the internal secular evolution of black holes in NLS1 hosts (Kormendy & Kennicutt 2004; Orban de Xivry et al. 2011). NLS1s are considered to be in a super-Eddington accretion state and are believed to have slim disks as optically thick advection-dominated accretion flows (Abramowicz et al. 1988; Mineshige et al. 2000; Wang & Netzer 2003). Black hole masses are expected to increase exponentially by factors of ∼10 and ∼1000 in case of super-Eddington mass accretion¹⁰ at ${\sim} 10\,\dot{M}_\mathrm{Edd}$ for durations of 1 × 10⁷ years and 3 × 10⁷ years, respectively (Kawaguchi et al. 2004). These arguments suggest that the lifetime of an NLS1 with relativistic ejections is limited to ∼10⁷ years. From the perspective of the fraction of NLS1s (∼10%) in the AGN population as well, ∼10⁷ years in NLS1 phase is suggested from the AGN lifetime of ∼10⁸ years (Collin & Kawaguchi 2004; Kawaguchi et al. 2004). However, large-scale radio morphology would be detectable in radio sources with ages of ≳ 10⁷ years after their black holes have grown to ≳ 10⁷ M_☉ (Section 4.3), in which the population of broad-line AGNs is dominant (Figure 3). Hence, kpc-scale radio structures may originate in a small window of opportunity during the final stage of the NLS1 phase just before altering into broad-line AGNs. Thus, the lower detection rate of kpc-scale radio structures can be attributed to the short lifetime in addition to lower maximal jet kinetic powers due to lower mass black holes (Section 4.3).

The existence of large-scale radio structures in NLS1s apparently conflicts with the paradigm that radio galaxies are associated exclusively with elliptical host galaxies (Véron-Cetty & Véron 2001; Hota et al. 2011). Many optical and infrared investigations of host galaxies have been conducted for nearby (radio-quiet) NLS1s, revealing higher fractions of strongly barred spirals (Ohta et al. 2007) and stronger star-forming activity (Sani et al. 2010) than those of normal broad-line Seyfert galaxies. However, the host morphologies of the detected NLS1s are not clearly distinguished on the basis of their SDSS images (cf. Huertas-Company et al. 2011; Lintott et al. 2011) because they are relatively distant. 1H 0323+342 is the only NLS1 with kpc-scale radio structures whose host morphology has been investigated because of its proximity (Zhou et al. 2007; Antón et al. 2008; Foschini 2011), and it exhibits a peculiar ring morphology with a circumnuclear starburst similar to that observed in collisional ring galaxies (Antón et al. 2008). The study of hosts for radio-loud NLS1s is crucial for testing the paradigm of radio galaxies in ellipticals, even at rapidly growing phases of supermassive black holes.