THE DIFFERENT NATURE OF SEYFERT 2 GALAXIES WITH AND WITHOUT HIDDEN BROAD-LINE REGIONS

In this section, we report the distributions of several parameters for HBLR and non-HBLR Sy2s, for example, the SMBH mass, redshift, N_H, $F_{\rm 2\hbox{--}10\,keV}/F_{[\rm O\,{\mathsc{iii}}]} (F_{\rm HX}/F_{[\rm O\,{\mathsc{iii}}]})$ ratio, Kα iron-line equivalent width (EW), and mid-infrared line ratios, and both $F_{\rm [Ne\, V]}/F_{[\rm Ne\,{\mathsc{ii}}]}$ and $F_{[\rm O\,{\mathsc{iv}}]}/F_{[\rm Ne\,{\mathsc{ii}}]}$.

In Figure 1(a), we show the distribution of SMBH masses for HBLR and non-HBLR Sy2s. Since there are censored data points (upper limits; densely shaded areas) among non-HBLR Sy2s, we use the astronomical survival analysis package (ASURV; Feigelson & Nelson 1985) for statistical analysis. The distributions of the SMBH masses are different between HBLR and non-HBLR Sy2s. For the entire sample, the mean SMBH mass of HBLR Sy2s is larger than that of non-HBLR Sy2s by the amount of 0.31, with a confidence level of 96.41% (see Table 3).

Zoom In Zoom Out Reset image size

Table 3. Summary of HBLR and Non-HBLR Sy2s

Parameters	Non-HBLR Sy2s		pnull	HBLR Sy2s
	Mean	N	(%)	Mean	N
(1)	(2)	(3)	(4)	(5)	(6)
log NH	22.96 ± 0.25	52	53.66	23.18 ± 0.19	40
EW(Fe)a	1325 ± αb	32	1.83	544 ± 99	38
$F_{\rm HX}/ F_{[\rm O\, \mathsc {iii}]}$	24.34 ± 17.49	56	28.15	6.02 ± 1.55	40
logMBH	7.10 ± 0.11	52	3.59	7.41 ± 0.11	34
log z	−1.98 ± 0.05	71	2.23	−1.71 ± 0.06	49
logL1.49 GHz	29.01 ± 0.13	56	0.068	29.74 ± 0.14	38
LFIR(1010L☉)	6.21 ± 1.60	64	47.70	9.22 ± 3.05	44
LIR(1011L☉)c	1.53 ± 0.41	57	17.19	3.01 ± 1.07	43
f60/f25d	5.81 ± 0.45	60	10-4	2.99 ± 0.30	45
$\dot{M}$	0.76 ± 0.22	66	0.01	3.11 ± 1.14	49
$(F_{\rm [Ne \,{\mathsc{v}}]}/F_{[\rm Ne\,{\mathsc{ii}}]})$	0.40 ± 0.12	21	0.17	1.21 ± 0.17	12
$(F_{\rm [O \,{\mathsc{iv}}]}/F_{[\rm Ne\,{\mathsc{ii}}]})$	1.03 ± 0.27	23	0.72	2.33 ± 0.44	13
log $L_{\rm [O \,{\mathsc{iii}}]}$	41.05 ± 0.17	66	0.01	42.05 ± 0.11	49
log (Lbol/LEdd)e	−0.98 ± 0.25	49	5.75	−0.12 ± 0.15	34

Notes. Column 1: parameters; Columns 2–3 and 5–6: for each sample of non-HBLR Sy2 and HBLR Sy2 galaxies, "Mean" is the mean value of the various parameters and N is the number of data points; Column 4: the probability p_null (in percent) for the null hypothesis that the two distributions are drawn at random from the same parent population. When there are censored data, we use Gehan's generalized Wilcoxon test (hypergeometric variance) in ASURV. ^aEW(Fe) is the Fe Kα equivalent width in eV. ^bAn ASURV test does not give the value of α. ^cWe have removed three sources (NGC 1241, NGC 3362, and NGC 7682) because they have no detections in their 12 μm band. ^dDetections only. ^eHere, the Eddington ratios, L_bol/L_Edd, are given by $L_{{\rm bol}}=3500L_{[\rm O\,{\mathsc{iii}}]}$ and $L_{\rm Edd}=1.4\times 10^{38} (M_{\rm BH}/M_{\odot })\,\rm {\rm erg} \,s^{-1}$, respectively.

Download table as:  ASCIITypeset image

Figure 1(b) shows the distribution of redshifts for HBLR and non-HBLR Sy2s. The distributions of redshifts are different between HBLR and non-HBLR Sy2s. The mean value of log z of HBLR Sy2s is larger than that of non-HBLR Sy2s by the amount of 0.27. A Kolmogorov–Smirnov (K-S) test shows that the probability for the two samples to be extracted from the same parent population is 2.23%.

In Figure 1(c), $F_{\rm HX}/F_{[\rm O\,{\mathsc{iii}}]}$, which is the ratio "T," is a good indicator of obscuration (Bassani et al. 1999; Tran 2003; here, [O iii] fluxes have been corrected for extinction, and X-ray fluxes have not been corrected for absorption). Table 3 shows little difference in $F_{\rm HX}/F_{[\rm O\,{\mathsc{iii}}]}$ between the two groups. An ASURV test shows that the probability for the two samples to be extracted from the same parent population is about 28.15%. The mean value of T is 24.34 ±  17.49 for non-HBLR Sy2s and 6.02 ±  1.55 for HBLR Sy2s.

In Figure 1(d), the N_H distributions between HBLR and non-HBLR Sy2s are not significantly different (with a confidence level of 53.66%; see Table 3; since an ASURV test could not deal with a case that contained both upper and lower limits, we adopt the N_H upper limits as the measured values). Our N_H distribution is consistent with the results of Gu & Huang (2002), Tran (2003), and Shu et al. (2007). This may be explained by the following reason: since the mean value of N_H is 10^21.85±0.33 cm⁻² for the less luminous non-HBLR Sy2s (see Table 4), they greatly weaken the difference in N_H between non-HBLR Sy2s and HBLR Sy2s (Shu et al. 2007). In Section 3.3, we will find that N_H has significant differences among the luminous and less luminous non-HBLR Sy2s and HBLR Sy2s (see Table 4 and Figure 4).

Table 4. Summary of Luminous and Less Luminous Non-HBLR Sy2s and HBLR Sy2s

Parameters	Non-HBLR Sy2sA(S1)a		HBLR Sy2s(S2)		Non-HBLR Sy2sB(S3)a		pnull(%)
	Mean	N	Mean	N	Mean	N	S1–S2	S2–S3	S1–S3
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
log NH	23.82 ± 0.26	29	23.18 ± 0.19	40	21.85 ± 0.33	23	3.02	0.01	0.01
EW(Fe)b	1539 ± αc	21	544 ± 99	38	955 ± αc	11	0.05	97.97	2.91
$F_{\rm HX}/ F_{[\rm O \,{\mathsc{iii}}]}$	0.96 ± 0.31	32	6.02 ± 1.55	40	55.28 ± 39.94	24	0.04	4.92	0.01
log $L_{\rm [O\,{\mathsc{iii}}]}$	41.90 ± 0.09	42	42.05 ± 0.11	49	39.54 ± 0.19	24	20.55	10−11	10−12
log (Lbol/LEdd)d	0.27 ± 0.16	28	−0.12 ± 0.15	34	−2.57 ± 0.27	21	9.82	0.01	0.01

Notes. Column 1: parameters. Columns 2–3, 4–5, and 6–7: for each sample of the non-HBLR Sy2sA, HBLR Sy2s, and non-HBLR Sy2sB, "Mean" is the mean value of various parameters and "N" is the number of data points. Column 8: from the K-S or ASURV test of luminous non-HBLR Sy2s (S1) vs. HBLR Sy2s (S2), the probability p_null for the null hypothesis that the two distributions are drawn at random from the same parent population. Column 9: as in Column (8), but for HBLR Sy2s (S2) vs. less luminous non-HBLR Sy2s (S3). Column 10: as in Column (8), but for luminous non-HBLR Sy2s (S1) vs. less luminous non-HBLR Sy2s (S3). When there are censored data, we use Gehan's generalized Wilcoxon test (hypergeometric variance) in ASURV. ^aNon-HBLR Sy2sA and Sy2sB indicate the luminous ($L_{[\rm O\,{\mathsc{iii}}]}>10^{41} \,\rm {\rm erg} \,s^{-1}$) and less luminous ($L_{[\rm O\,{\mathsc{iii}}]}<10^{41} \,\rm {\rm erg} \,s^{-1}$) non-HBLR Sy2s, respectively. ^bEW is the Fe Kα equivalent width in eV. ^cAn ASURV test does not give the value of α. ^dHere the Eddington ratio is the same as Table 3.

Download table as:  ASCIITypeset image

In Table 3, non-HBLR Sy2s are obviously larger in terms of the mean value of EW(Fe) than HBLR Sy2s. An ASURV test shows that the difference between the two samples is present (at a level of 98.17%). Our result is not consistent with that of Tran (2003) or Shu et al. (2007). This may be explained by the following reason: due to the small sample size (only 11 objects) of the less luminous non-HBLR Sy2s with EW(Fe) measurements, adding them to the non-HBLR Sy2 sample cannot weaken the difference (99.95%) in EW(Fe) found in the luminous Sy2 sample (see Table 4; Shu et al. 2007).

The two mid-infrared line ratios, $F_{\rm [Ne \,\mathsc {v}]}/F_{[\rm Ne\,{\mathsc{ii}}]}$ and $F_{[\rm O\,{\mathsc{iv}}]}/F_{[\rm Ne\,{\mathsc{ii}}]}$, can better distinguish HBLR from non-HBLR Sy2s (see Table 3). Table 3 shows the significant differences in the two ratios between the two groups. A K-S test shows that the probabilities that the two samples are extracted from the same parent population are 0.17% and 0.72%, respectively.

In Table 3, we also provide other statistical results related to the two types of sources. We find that most of the results of the various parameters show the significant differences between the two groups. These indicate that HBLR and non-HBLR Sy2s are clearly different subsamples.

In this section, we test if HBLR Sy2s are dominated by AGNs and if non-HBLR Sy2s are dominated by starbursts. Next, we use two methods to demonstrate the different dominant mechanisms between non-HBLR and HBLR Sy2s.

As mentioned in Section 1, the AGN activity is the key to understanding the differences between the two kinds of Sy2s. AGN luminosity comes from the disk accretion onto the central SMBHs. [O iii] λ5007 luminosity represents only an indirect (i.e., reprocessed) measurement of the nuclear activity, but extinction-corrected $L_{[\rm O \,{\mathsc{iii}}]}$ is a good indicator of the AGN activity for type II AGNs (Kauffmann et al. 2003).

The IRAS f₆₀/f₂₅ flux ratio is not a good indicator of the inclination, but of the relative strength of the host galaxy and nuclear emission (Alexander 2001; Shu et al. 2007). It has been shown that HBLR Sy2s have smaller values of the f₆₀/f₂₅ ratio compared with non-HBLR Sy2s (Heisler et al. 1997). In Table 3, the mean value of the f₆₀/f₂₅ ratio is 5.81 ±  0.45 for non-HBLR Sy2s and 2.99 ± 0.30 for HBLR Sy2s. The difference (at the 99.999% level) in color between them may be due to the relative strength of the host galaxies and nuclear emissions (Alexander 2001), and this is consistent with Baum et al. (2010) who noted that the HBLR Sy2s have a higher ratio of AGN-to-starburst contribution to the spectral energy distribution (SED) than non-HBLR Sy2s, based on their distributions of several starburst and AGN tracers. So, we can show that the f₆₀/f₂₅ ratio denotes the relative strength of starburst and AGN emissions.

In Figure 2, we show the f₆₀/f₂₅ ratio versus the accretion rate (defined as $\dot{M}=L_{{\rm bol}}/\eta c^2$), with a Pearson's correlation coefficient of −0.54 and a probability of <0.0001 (here, we exclude NGC 676, NGC 1143, NGC 1358, NGC 2685, NGC 4472, and NGC 4698, because they are either not detected or they are the upper limit of fluxes and below 1 Jy at $25 \,\rm \mu m$ or $60 \,\rm \mu m$). These results show that they have a significant anticorrelation. As the f₆₀/f₂₅ ratio drops, the $\dot{M}$ value increases. HBLR Sy2s show smaller f₆₀/f₂₅ ratios and larger $\dot{M}$ values, which may indicate a higher ratio of AGN-to-starburst activity in the SED; non-HBLR Sy2s show larger f₆₀/f₂₅ ratios and smaller $\dot{M}$ values, which may indicate a lower ratio of AGN-to-starburst activity in the SED. Therefore, we suggest that non-HBLR Sy2s are dominated by starbursts, while HBLR Sy2s are dominated by AGNs.

Zoom In Zoom Out Reset image size

We also use another diagnostic for examining starburst and AGN activities. Due to the intense star formation in the nuclear region of many active galaxies, some fraction of the measured fluxes of low lying fine structure lines (excitation potential ⩽50 eV) will be produced by photoionization from stars rather than AGNs, while the high excitation lines ([O iv], [Ne v]) show little or no contamination from possible starburst components (Sturm et al. 2002). Genzel et al. (1998) found that [O iv]/[Ne ii] and [Ne v]/[Ne ii] are much higher in AGNs than in starbursts, which can now be confirmed on a broader statistical basis. [O iv] originates purely from the narrow-line region (NLR) in AGNs. In a unified scheme, the NLR line luminosity should be independently oriented and be a good tracer of AGN power, in particular when using an extinction-insensitive and modest excitation line like [O iv] (Sturm et al. 2002). Since the ionization potential of [Ne v] λ14.32 is E_ion = 97.1 eV, [Ne v] is unlikely to be strong in galaxies without an AGN (Voit 1992; Farrah et al. 2007). While [Ne ii] is a fairly good tracer of hot star emission in starburst activity, in AGNs, the [Ne ii] from the NLR is more easily contaminated by starburst emission than the higher excitation [O iv] line (Sturm et al. 2002).

In Figure 3, the relations of the infrared luminosity versus the flux ratios of [Ne v] λ14.32/[Ne ii] λ12.81 and [O iv] λ25.89/[Ne ii] λ12.81 are shown. We can see in the two regions of each plot that the upper region is primarily HBLR Sy2s and the lower one is primarily non-HBLR Sy2s. The [Ne v] λ14.32 and [O iv] λ25.89 lines are the most useful single line diagnostics (Farrah et al. 2007). As a result, we suggest that the non-HBLR Sy2s are starburst dominated, while the HBLR Sy2s are AGN dominated.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In Figures 2 and 3, we find that HBLR and non-HBLR Sy2s clearly show the differences in the power sources. In Table 3, the differences in the accretion rate ($\dot{M}=L_{\rm bol}/\eta c^2$), f₆₀/f₂₅ ratio, and the two mid-infrared line ratios, $F_{\rm [Ne\,{\mathsc{v}}]}/F_{[\rm Ne\,{\mathsc{ii}}]}$ and $F_{[\rm O\,{\mathsc{iv}}]}/F_{[\rm Ne\,{\mathsc{ii}}]}$, are significant. So, we maintain that non-HBLR Sy2s are starburst dominated, while HBLR Sy2s are AGN dominated.

In this section, we will investigate differences in the obscuration and reasons for the absence of PBLs in luminous and less luminous non-HBLR Sy2s and compare them with those of HBLR Sy2s.

With regard to the obscuration between HBLR and non-HBLR Sy2s, our result (see Table 3) and previous results (Tran 2003; Gu & Huang 2002; Shu et al. 2007) all show little difference. The reason is that adding the less luminous Sy2s to the Sy2 sample weakens the difference in obscuration found in the luminous Sy2 sample (Shu et al. 2007). Next, we discuss the possible physical explanations for the absence of PBLs in the luminous and less luminous non-HBLR Sy2s, respectively.

Table 4 shows that all the differences in N_H, EW(Fe), and $F_{\rm HX}/F_{[\rm O\,{\mathsc{iii}}]}$ between the luminous non-HBLR Sy2s and HBLR Sy2s are significant. An ASURV test shows that the probabilities for the two samples to be extracted from the same parent population are 3.02%, 0.05%, and 0.04%. These results suggest that luminous non-HBLR Sy2s show larger obscuration than HBLR Sy2s. In addition, we do not find a significant difference (with a confidence level of 20.55%) in $L_{[\rm O\,{\mathsc{iii}}]}$ between luminous non-HBLR Sy2s and HBLR Sy2s, and the mean values of log($L_{[\rm O\,{\mathsc{iii}}]}/\rm {\rm erg} \,s^{-1})$ are 41.90 ±  0.09 and 42.05 ± 0.11, respectively. So, we suggest that the obscuration is the key cause that makes PBLs weaker or nondetectable for the luminous non-HBLR Sy2s. Our explanation supports Shu et al.'s (2007) suggestion that the absence of PBLs in the luminous Sy2s arises from obscuration. However, PBLs are not detected in most (24/28) of the less luminous Sy2s in our sample. The reason is still unclear.

To explore the natural reason for the absence of PBLs in less luminous non-HBLR Sy2s, we analyze their obscuration: N_H = 10^21.85±0.33 cm⁻² and $F_{\rm HX}/F_{[\rm O\,{\mathsc{iii}}]}=55.28\pm 39.94$ (see Table 4), suggesting that the less luminous non-HBLR Sy2s have smaller obscuration than the luminous non-HBLR Sy2s or HBLR Sy2s.⁴ Their obscuration seems to be close to that of face-on Sy1s. If the scale height of the scattering zone varies with the central source luminosity (Lumsden & Alexander 2001), the absence of their PBLs may be due to either the small scale height in the scattering region or the inexistence of BLRs. Since the less luminous non-HBLR Sy2s have very small $L_{[\rm O\,{\mathsc{iii}}]}$, their scattering screens may have smaller scales than those of luminous non-HBLR Sy2s or HBLR Sy2s. However, the obscuration of this type of objects is very small and seems to be the same as that of the host galaxy. So, we suggest that the invisibility of PBLs for less luminous non-HBLR Sy2s does not arise from the obscuration.

The Eddington ratios of the less luminous non-HBLR Sy2s are generally very small and their mean value is 10^{−2.57±0.27} (see Table 4). This is consistent with what Nicastro et al. (2003) argued, that at very low accretion rates, the clouds of BLRs would cease to exist. Since the obscuration of this type of objects is very small, a key factor in the absence of PBLs is the very low Eddington ratio rather than the obscuration. When the accretion rate drops to extremely sub-Eddington values, their central engines undergo fundamental changes and the BLR disappears (Ho 2008). Recently, Tran et al. (2010) suggested that the low-luminosity AGNs are probably powered by radiatively inefficient, or advection-dominated, accretion flow, which intrinsically lack BLRs, as suggested observationally by, e.g., Tran (2001, 2003), Bianchi et al. (2008), Panessa et al. (2009), and Shi et al. (2010), and inspired theoretically by Nicastro (2000), Laor (2003), Elitzur & Shlosman (2006), Elitzur & Ho (2009), and Cao (2010).

In Table 4, we find that non-HBLR Sy2s can be classified into the luminous ($L_{[\rm O\,{\mathsc{iii}}]}>10^{41} \,\rm {\rm erg} \,s^{-1}$) and less luminous samples, when considering only their obscuration. In light of the above discussion, we hold that the invisibility of PBLs in the luminous non-HBLR Sy2s depends on the obscuration; the invisibility of PBLs in less luminous non-HBLR Sy2s depends on the very low Eddington ratio rather than on the obscuration.

Except for the 18 objects in Table 5 of Wang & Zhang (2007), all other objects of our sample have their spectropolarimetric observations, which are described in Table A1 in detail.