4 Discussion

  The new radio observations presented in this paper cover more than 50% of our southern ($\delta<0^\circ$) volume limited (cz<3600 km s^-1) sample of Seyfert galaxies; adding data from the literature increases the coverage to almost 85%. There remain 9 sources in the southern sample for which radio information at arcsec resolution is not available yet.

Our survey largely overlaps with the samples studied by Ulvestad & Wilson (1984b) and Ulvestad & Wilson (1989). We do not, therefore, expect results completely independent from theirs. We note, however, that the sample of Ulvestad & Wilson contained all Seyferts with cz < 3600 km s^-1 known at the time of their observations ($\sim 15$ years ago), and was declination-limited by the VLA horizon restrictions. Our sample, on the other hand, covers all Seyferts with reliable classification to date at southern declinations ($\delta<0^\circ$) and increases the distance-limited sample of Ulvestad & Wilson (1989) by 22%.

However, in order to further improve the statistics, we have collected additional sources from the literature, consisting of known Seyferts with cz < 4600 km s^-1, derived mainly from Ulvestad & Wilson (1989) and references therein. The selected sources are listed in Table 4. For this compilation, we selected only radio data of similar observing frequency and resolution to our observations ($\sim 0.3-1$$^{\prime\prime}$). Most of the collected data were obtained at 6 cm. The combined sample includes 71 objects of which 17 are Seyfert 1's and 54 Seyfert 2's.

Here we concentrate on the discussion of the radio characteristics of the studied sample. A discussion on the comparison between the radio and optical properties will be done in a forthcoming paper.

  

Table 4: Radio parameters from the literature

References: (1) Wilson & Ulvestad (1983); (2) Ulvestad & Wilson (1984a); (3) Ulvestad & Wilson (1984b); (4) Ulvestad & Wilson (1989); (5) Pedlar et al. (1993); (6) Kukula et al. (1993); (7) Neff & de Bruyn (1983); (8) Gallimore et al. (1996); (9) Shommer et al. (1988); (10) Su et al. (1996); (11) Wehrle & Morris (1988).

4.1 Radio spectra

We derived the spectral indices for most of our sources by combining our data with previous observations (see Sect. 3.1). The exception is NGC 3393, for which we use the 3 and 6 cm data presented here. In three cases (NGC 1097, NGC 1365 and IC 5063), we are unable to estimate the spectral index owing to complex source structure and poor matches in (u,v) coverage. For NGC 1097 and NGC 1365 detailed studies of the spectral indices are already available (Hummel et al. 1987; Sandqvist et al. 1995, respectively).

In three cases (NGC 7582, NGC 3281 and Tol 0109-383) we find that the spectral index at high frequencies ($\alpha_{\rm high}$, between 3 and 6 cm) is similar to the value at lower frequencies ($\alpha_{\rm low}$); that is, the spectra are steep in the first two cases, and flat for Tol 0109-383. In two cases (MGC-6-30-15 and NGC 1386) the spectra appear to flatten with increasing frequency ($\alpha_{\rm high} \gt \alpha_{\rm low}$), although the radio source in MGC-6-30-15 is very weak, and the derived spectral index is uncertain. New spectral indices have been derived for NGC 3393, NGC 3783 and Tol 1238-364: in all these cases the spectral index is steep ($\alpha <-0.5$). For comparison, most of the Seyfert galaxies have steep radio spectra, but flat-spectrum cores are found in a few Seyferts (Wilson 1991).

Accepting uncertainties in our spectral index measurements due to mismatched (u,v) coverage in the 3 and 6 cm maps, we measure a median spectral index steep ($\alpha \sim -0.67$) for our Seyfert nuclei. This result is in agreement, with the spectral indices of the inner $0\hbox{$.\!\!^{\prime\prime}$}1$ nuclear regions of spirals (and Seyferts) as measured by Sadler et al. (1995) using PTI observations. Their investigation of the compact radio cores in spiral and elliptical galaxies found a median spectral index for spirals (and Seyferts) of $\alpha=-1.0$, and that spirals usually have steeper core spectra than do elliptical galaxies (median spectral index $\alpha = +0.27$). The flat (or inverted) spectral index of the cores is a typical characteristic of elliptical galaxies found on all scales in which the nuclear regions have been observed (from arcsec and sub-arcsec scale, see e.g. Morganti et al. 1997 and Slee et al. 1994 to VLBI scale). Thus, our study confirms the result that the spectral indices of Seyferts are much steeper than in the cores of ellipticals, and that the spectral index remains close to the same value over $0\hbox{$.\!\!^{\prime\prime}$}1$ to 1$^{\prime\prime}$ scales.

In connection with this difference in spectral index, it is worth remembering that the nuclear regions of Seyfert galaxies appear to have a more complicated situation than in radio galaxies: following the detailed studies of few well known objects (e.g. NGC 1068, Gallimore et al. 1996; Roy et al. 1998; NGC 4151, Ulvestad et al. 1998) free-free absorption appears to be relevant in Seyfert to dim the "real" radio core. Thus the nuclear emission (and its spectral index) can be dominated not by the core itself but by bright blobs. This would be in agreement with the finding (Sadler et al. 1995) that in ellipticals most of the radio emission in the central kpc comes from the parsec-scale core, while in Seyferts this is only a small fraction (10-25%).

4.2 Radio structure

Table 3 summarizes the classification of the radio morphology for the newly observed objects, following the scheme of Ulvestad & Wilson (1984b). Table 4 provides source classifications for data taken from the literature. Including Seyferts from the literature, we find 38% of the sample are unresolved sources, 23% have slightly resolved structures and 34% have linear structure at arcsecond resolution; the remaining 5% have diffuse, amorphous, or ringed structures.

  

Figure 16: Plot of the size of the radio emission versus the radio power (at 5 GHz): open squares represent Seyfert 1 and filled squares represent Seyfert 2

We can compare these numbers with the results from Ulvestad & Wilson (1989) and references therein (see also the summary in Wilson 1991). Compared to Wilson (1991), we measure a higher fraction of unresolved sources and a lower fraction of galaxies with diffuse radio emission. This result may be due to an observational bias; our observations are not as sensitive to diffuse, extended, steep spectrum emission as are the 20 cm VLA observations of, e.g., Ulvestad & Wilson (1989). Otherwise, we measure a similar detection fraction of linear or slightly resolved structures.

Linear radio sources in Seyfert galaxies generally trace radio outflows, but it is worth checking that linear radio emission is not associated with star formation in an edge-on galaxy. Among the linear radio sources detected in this survey, only the host galaxy NGC 7172 is near edge-on. The radio structure aligns with the plane of the host galaxy, casting doubt on a jet origin; rather, it appears that the linear radio source of NGC 7172 may be associated plausibly with a nuclear starburst. Otherwise, the remaining linear radio sources are more likely associated with an AGN-driven outflow.

To compare the distributions of radio power, we adjusted all of the radio luminosities to their 6 cm values. For sources not observed at 6 cm, we adjusted the luminosities using measured spectral indices where possible. For those sources having neither 6 cm measurements or measured spectral indices, we assumed a spectral index of $\alpha^3_6 = -0.5$. Many of the low power sources are unresolved, resulting in an upper limit to the source size. Since we are interested in the compact radio emission from jets rather than extended, low surface brightness emission (which may come from star-forming regions and starburst-driven superwinds), we also assumed that the three undetected sources were unresolved (size upper limit) and radio weak (flux upper limit). To account appropriately for these limits, we employed survival analysis techniques from the ASURV package (Feigelson & Nelson 1985; Isobe et al. 1986) as it is implemented in IRAF.

We confirm a correlation between radio power and size (Fig. 16), originally discussed by Ulvestad & Wilson (1984b) and Giuricin et al. (1990). Both the power-size and flux-angular size correlations are significant at better than 1% (probability of no correlation) according to the traditional survival analysis tests (Kendall's $\tau$, Spearman's $\rho$, and the Cox Proportional History model). A summary of log-linear models for the power-size correlation is provided in Table 5. Comparing with the results of Ulvestad & Wilson (1984b), we find a steeper slope in the correlation owing to a proper treatment of limits at low radio-powers. Mrk 348 is an outlier, falling at relatively small size for its radio power. The present analysis ignores the $\sim 5$ kiloparsec radio lobes in this source, which may arise from either a starburst-driven superwind, or old nuclear-driven ejecta. Accounting for the extended radio lobes places Mrk 348 closer to the best-fit line, except that Mrk 348 now falls somewhat oversized for its luminosity according to the correlation.

  

Table 5: Log-linear fit to Radio Power vs. Size

  

Figure 17: Histograms of the spectral index distribution for Seyfert 1's and Seyfert 2's

  

Figure 18: Histograms of the radio size distribution for Seyfert 1's and Seyfert 2's. Shaded regions represent upper limits

  

Figure 19: Histograms of the radio power distribution for Seyfert 1's and Seyfert 2's. Shaded regions represent upper limits

4.3 A Comparison of Seyfert 1s and 2s

Previous radio studies of volume-limited samples of Seyfert galaxies find: a) marginal or no statistically significant difference between the distributions of radio luminosities of Seyfert 1s and 2s; b) Seyfert 2's galaxies tend to have a larger radio sources than Seyfert 1s, although at <90% significance level; c) the two types of Seyferts have essentially the same distribution of spectral indices (Giuricin et al. 1990; Rush et al. 1996) although there seems to be some evidence that flat spectrum cores are more common in Seyfert 1s than in Seyfert 2s (Ulvestad & Wilson 1989); and d) compact radio cores (on the sub-arcsec scale) are more common in Seyfert 2s than in Seyfert 1s (R94).

Examining first the distribution of spectral index (see histograms in Fig. 17), we find a median value for the spectral index ($\alpha_{6}^{20}$ or $\alpha_{3}^{6}$) of $\alpha = -0.44$ for Seyfert 1s and -0.72 for Seyfert 2s. However, the two distributions are not significantly different under a Kolmogorov-Smirnov test (13% probability that the two distributions are not different). Unfortunately, we have too few Seyfert 1's with measured spectral indices to compare the frequency of flat-spectrum cores.

We also compared the distribution of radio sizes and powers (see histograms in Figs. 18 and 19 respectively) of Seyfert 1 and Seyfert 2 galaxies. To improve the fairness of the comparison, we only included sources out to a redshift of cz = 4000 km s^-1, at which the redshift distributions of Seyferts 1 and 2 matched according to a K-S test. At marginal significance, we find that Seyfert 2 radio sources tend to be more luminous than Seyfert 1 radio sources; the probability that Seyferts 1 and 2 arise from the same parent distribution is $\sim$ 11% for most two-sample tests, $\sim$ 25% for the logrank test. Ulvestad & Wilson 1984a found a similar result for Markarian Seyferts, but they showed in a follow-up paper (Ulvestad & Wilson 1989) that the significance is reduced owing to the paucity of low-luminosity Seyfert 2s in the Markarian sample. Inspection of the power-size diagram (Fig. 12) suggests that any significance in the difference of radio powers owes to a handful of luminous Seyfert 2s, but most Seyferts 1 and 2 have comparable radio powers. Formally, we measure mean log radio powers of $\log{P\ {\rm (W\ Hz^{-1})}} = 21.03\pm 0.01$ for Seyfert 2s, and $\log{P\ {\rm (W\ Hz^{-1})}} = 20.66\pm 0.18$ for Seyfert 1s.

In contrast, we measure a significant difference in radio sizes. All of the two-sample tests report a difference in parent populations to a significance of $\ge 95\%$, with the exception of the logrank test, which is significant only to 87%. Seyfert 2s tend to be larger, with a mean size of $0.53\pm 0.12$ kpc, compared to Seyfert 1s, mean size $0.16\pm 0.04$ kpc. The errors on the mean sizes represent the dispersion in the intrinsic value. Given the statistical agreement in the distribution of radio powers, it is natural to interpret the size difference in terms of an orientation unifying scheme. The prediction of this model is the nuclear axes of Seyfert 1s are viewed more nearly pole-on, and so radio jets are foreshortened by projection.