4 Discussion

We find a mean global $L_{\rm HCN}$/ $L_{\rm CO}$ ratio of 1/6 for our sample of Seyfert galaxies, which is similar to what SDR92 find for a set of five ULIRGs. From the HCN/CO luminosity plot it is somewhat surprising that the luminosity ratio of 1/6 holds for all the detections (Fig. 5), since we expected a larger contribution from the disk in the distant sources (  km s^-1), resulting in a lower $L_{\rm HCN}$/ $L_{\rm CO}$ compared to the near-by galaxies. Typically the HCN/CO ratio in the disks of galaxies is greater than 40 (e.g. Helfer & Blitz 1993; Kuno et al. 1995), however, here both near-by and distant sources obey the same $L_{\rm CO}$ to $L_{\rm HCN}$ correlation, and so this may (partly) be an effect of a highly centralised molecular emission. From interferometric studies of the CO distribution we know this to be true for FIR luminous objects such as Mrk 231, Mrk 273, Arp 220 and NGC 7469, but also less FIR bright galaxies such as NGC 0034, NGC 1667 and NGC 7130 have very bright HCN with respect to their CO emission. It is possible that we are missing CO emission at radii greater than 6 kpc, thus getting an incorrect global $L_{\rm HCN}$/ $L_{\rm CO}$ ratio, but the more likely explanation is that the CO emission is also highly concentrated in these galaxies. This confinement of the CO is another similarity between our sample and ULIRGs (Scoville et al. 1991), as opposed to normal spirals where the HCN is much more centralised than the CO (Helfer & Blitz 1993).

Comparing the global with our measured CO luminosities of the near-by sample (Table 3), it appears that the CO extends well beyond $\approx3$ kpc in the near-by sample (based on NGC 2273), while the HCN is expected to be confined to within 1 kpc (Sect. 3), leading us to believe that we have sampled the majority of the HCN in the near-by galaxies. Why should we have different CO distributions for the distant and near-by galaxies? Examining the FIR luminosities (Table 3), the mean values of the luminosities are $L_{\rm FIR}\approx30\ 10^{10}~L_{\odot}$ for v>4000 km s^-1and $L_{\rm FIR}\approx2\ 10^{10}~L_{\odot}$ for v<4000 km s^-1. In the case of ULIRGs, the high FIR luminosity is an indicator of a high CO concentration (Bryant 1997), and so perhaps there exists a selection effect at play, in which our distant sources comprise mainly of galaxies suffering from little CO contamination in the galactic disk.

The FIR/HCN luminosity plot, Fig. 6, shows a similar correlation to that of SDR92, which holds over a large range of $L_{\rm FIR}$ and $L_{\rm HCN}$. It appears that, in our Seyfert sample, as in the ULIRGs, the FIR to HCN luminosity is similar to that of normal spiral galaxies, i.e.

$$\frac{L_{\rm FIR}}{L_{\rm HCN}}~({\rm Seyferts})\approx\frac{L_{\rm FIR}}{L_{\rm HCN}}~({\rm normal~ spirals})$$ (1)

and since normal spirals have $L_{\rm HCN}/L_{\rm CO}\approx1/80$(SDR92), from previously

$$\frac{L_{\rm HCN}}{L_{\rm CO}}~({\rm distant~Seyferts})\approx10\frac{L_{\rm HCN}}{L_{\rm CO}}~(\rm normal~ spirals),$$ (2)

and combining these equation gives

$$\frac{L_{\rm FIR}}{L_{\rm CO}}~({\rm distant~Seyferts})\approx10\frac{\rm L_{FIR}}{L_{\rm CO}}~(\rm normal ~spirals).$$ (3)

As a further check, in order to account for the distances and the extents in our sample, as SDR92, we plotted $L_{\rm FIR}/L_{\rm CO}$ against $L_{\rm HCN}/L_{\rm CO}$, Fig. 7.

Figure 7: $\log L_{{\rm FIR}}$ [ $L_{\odot}({\rm K ~km~s}^{-1}~{\rm pc}^2)^{-1}$] versus $\log L_{{\rm HCN}}$ normalised by the CO luminosity for all the detections [no units]

From this we can see that the normalisation does not significantly alter the linear fit, although three of the sample deviate further from the line. In any case, from the fit we also find $L_{\rm FIR}\approx600 L_{\rm HCN}~L_{\odot}({\rm K ~km~s}^{-1}~{\rm pc}^2)^{-1}$ (normalised by the CO luminosity) which again agrees well with the result of SDR92 ( $L_{\rm FIR}\approx750 L_{\rm HCN}~ L_{\odot}({\rm K ~km~s}^{-1}~{\rm pc}^2)^{-1}$).

With regard to what these results entail, we interpret the high $L_{\rm HCN}$/ $L_{\rm CO}$ ratio in our sample of Seyferts as largely an effect of a high degree of central concentration of the gas. The steep central potential and high gas surface densities result in large gas pressures which could force the bulk of the molecular mass to reside in high density gas. This should be true regardless of what activity dominates the FIR emission from the galaxy. SDR92 suggest that the $L_{\rm HCN}$/ $L_{\rm FIR}$ correlation means that the HCN emission traces star forming cores, which are responsible for the IR emission, i.e. the ULIRGs are powered by star-bursts. For several of the Seyferts in our sample, however, at least 50% of the FIR emission may come from the AGN activity, even if a star-burst also contributes to the total luminosity: Kohno et al. (1999) suggest that Seyferts with jets have particularly bright HCN emission and that the dense gas, which is a component of a large-scale obscuration confines the jet (Antonucci & Miller 1985; Wilson et al. 1988; Tadhunter & Tsvetanov 1989; Wilson & Tsvetanov 1994; Baker & Scoville 1998). Also, Bryant (1997); Kohno et al. (1999) discuss the connection between very high HCN/CO ratios and these nuclei and point out that Seyfert nuclei are associated with the lowest published HCN/CO ratios so far. The only similarity between these objects is that they host broad-line AGN, and similarly, in our sample, NGC 1667 and Mrk 273 both have extremely bright HCN with respect to CO, yet they differ by over an order of magnitude in their FIR luminosities.

An $L_{\rm HCN}$/ $L_{\rm CO}$ ratio of 1/6 is at the lower end of what is typically found even for galactic nuclei. For near-by galaxies, standard HCN/CO ratios appear to range between 1/15 and 1/5 on scales smaller than a kiloparsec (e.g. Helfer & Blitz 1995; Aalto et al. 1995), and so a high degree of gas concentration may not be a sufficient explanation for the high ratio we observe. The "extra'' bright HCN emission may be an additional effect of extreme gas excitation and/or unusual abundance effects.

Heckman et al. (1989) state that the CO and far infrared properties differ between the two Seyfert classes, although Seyferts as a class exhibit similar $L_{\rm CO}$ to $L_{\rm FIR}$ ratios to non-Seyferts. In order to test this, we plot the log of the $L_{\rm CO}/L_{\rm FIR}$ ratio versus the log of $L_{\rm FIR}$ (Table 3), and find that, not unexpectedly, all of our detections lie in the same range as Fig. 6 of Heckman et al. (1989), Fig. 8. From these results we calculate the mean ratio, at a 90% confidence level, to be

$$\log\left(\frac{L_{\rm CO}}{L_{\rm FIR}}\right)=-8.0\pm0.2 {\... ...tyle ...$$ (4)

$$\log\left(\frac{L_{\rm CO}}{L_{\rm FIR}}\right)=-8.0\pm0.1 {\rm ~for ~all ~the ~detections}.$$ (5)

cf. -7.76 for Seyferts and -7.84 for non-Seyferts (Heckman et al. 1989).

Figure 8: $\log L_{\rm CO}/L_{{\rm FIR}}$ [ ${\rm K ~km~s}^{-1}~{\rm kpc}^2~L_{\odot}^{-1}$] versus $\log L_{{\rm FIR}}$ [$L_{\odot }$]

This is the same range over which normal spiral galaxies are distributed (Young et al. 1984; Sanders & Mirabel 1985; Stark et al. 1986; Young et al. 1986). If we refer to Fig. 6c of Sanders & Mirabel (1996), our sample is located at around , cf. $L_{\rm FIR}<10^{11}~L_{\odot}\approx10M({\rm H}_2)~M_{\odot}$ for normal spirals and for star-burst galaxies. So we find that for similar values of $L_{\rm FIR}$, that our sample has about double the $L_{\rm FIR}/L_{\rm CO}$ ratio of normal spiral galaxies and a similar ratio to the more moderate star-burst galaxies ( $L_{\rm FIR}\sim10^{11}-10^{12}~L_{\odot}$).

As seen from Fig. 8 and Fig. 6c of Sanders & Mirabel (1996), the Seyfert sample follows the same trend for $L_{\rm CO}/L_{\rm FIR}$ to decrease over an order of magnitude as defined by normal spirals to ULIRGs, with the Seyferts being located in between these two extremes. This result could be caused by either:

1. The FIR luminosity arising from young stars, in which case the HCN would be tracing dense star-forming cloud cores.
2. The FIR luminosity arising from something other, i.e. an AGN.

In the first case we would expect the $L_{\rm FIR}/L_{\rm HCN}$ ratio to be fairly independent of the FIR luminosity of the galaxy and in the latter case we would expect $L_{\rm FIR}/L_{\rm HCN}$ to increase, i.e. a further FIR contribution in addition to that from star formation coming into play. Plotting the $L_{\rm HCN}/L_{\rm FIR}$ratio versus $L_{\rm FIR}$ (in a similar manner to Fig. 8), Fig. 9, we see that the HCN/FIR ratio may well decrease with the FIR luminosity thus indicating that the latter scenario may be the case, although with these statistics this is far from conclusive. In support of the non-AGN argument, a constant (on average) value of $L_{\rm HCN}/L_{\rm FIR}$ may be feasible (although the least squares linear fit does give a non-zero gradient, Fig. 9).

Figure 9: $\log L_{\rm HCN}/L_{{\rm FIR}}$ [ ${\rm K ~km~s}^{-1}~{\rm kpc}^2~L_{\odot}^{-1}$] versus $\log L_{{\rm FIR}}$ [$L_{\odot }$]. The least squares linear fit is shown

This is supported by Figs. 6 and 7 where a simple line does provide a good fit; a decreasing $L_{\rm HCN}/L_{\rm FIR}$ratio would demand a (slight on a log-log plot) curve in these figures. Also from Figs. 6 and 7, there appears to be no FIR to HCN excess cf. normal gas rich galaxies and ULIRGs (SDR92), although their fit may be significantly affected by the high values of $L_{\rm FIR}$ for Mrk 231 and Arp 220 which are also considered to be Seyferts, and there does remain the possibility that the HCN may be associated with the gas obscuring the AGN rather than dense star-forming cores. The low CS $3\rightarrow2$/HCN $1\rightarrow 0$luminosity ratios (<0.5) in the (Southern) sample may support this result (Curran 2000b).

A similar correlation for both Seyfert and star-burst galaxies is also found between $L_{\rm FIR}$ and $L_{{\rm H}\alpha}$ (Gu et al. 1997), and this as well as other Seyfert samples, which utilise FIR luminosities (e.g. in comparison with $L_{\rm blue}$; Whittle 1992; Gu et al. 1999 and the radio continuum; Roy et al. 1998), lead to the conclusion that the FIR flux is thermal in origin for most Seyfert galaxies.