We initially expected that the relative HNC luminosity would be lower in our sample of warm, luminous galaxies compared to the H95 sample. Instead (see Sect. 3.1.1) the very luminous galaxies even seemed to be somewhat more luminous, on average, in HNC even though this change is not statistically significant. A possibility could be that the telescope beam is picking up more extended, cooler gas in the distant, more luminous, galaxies. However, most of them are known to have very compact molecular cloud distributions and it is unlikely that significant HNC emission emerges from the outskirts of the galaxies. Our results seem to challenge the notion of HNC as a reliable tracer of cold gas.

Furthermore, the variation in ${{\rm HCN} \over {\rm HNC}}$ line ratio is large among the galaxies that otherwise have similar properties. For example, despite having similar ${{\rm CO} \over {\rm HCN}}$ 1-0 intensity ratios, Mrk 273 and Mrk 231 have quite different ${{\rm HCN} \over {\rm HNC}}$ and ${{\rm HCN} \over {\rm CN}}$ line intensity ratios. The ${{\rm HCN} \over {\rm HNC}}$ ratio of Mrk 231 is close to unity, while HNC is not detected in Mrk 273 resulting in a ${{\rm HCN} \over {\rm HNC}}$ ratio $$\mathrel{\mathchoice {\vcenter{\offinterlineskip\halign{\hfil $\displaystyle ...$ 5. It is unlikely that the dense gas is cold in Mrk 231 but warm in Mrk 273 (both are hot, ultraluminous AGN/Starburst mergers) and therefore the interpretation of HNC needs to be reevaluated.

4.1 Why is HNC often bright in centers of galaxies?

There are a number of possible explanations. Let us examine them one by one and see how they can be tested:

Is it possible that a significant fraction of the dense gas is cold in centers of starburst galaxies? Possibly. Aalto et al. (1994) discuss the presence of cold dense gas in the "mild'' starburst galaxy NGC 1808. The high dust temperatures observed towards NGC 1808 could be explained by clouds with hot surfaces and cold interiors. In this scenario, the HNC 1-0 emission would emerge from the cold cloud cores, while HCN 1-0 emission also would come from the outer, warmer parts. Gas at densities >10 $^5~\rm {cm^{-3}}$ should become thermalized with the dust. Thus, if the HNC emission is coming from dense, cold (10-20 K) gas there should be submm dust continuum emission associated with it. Therefore, a study of the submm and mm excess in conjunction with the strength of the HNC emission should be quite interesting. However, the most HNC luminous object in our sample, Mrk 213, shows very weak mm thermal dust emission and Braine & Dumke (1998) find a dust temperature of 70 K - the kinetic temperature of the associated dense gas should be at least as high. This is not consistent with the idea that the HNC emission arises from a cold component. Based on ISOPHOT data Klaas et al. (1997) find that the bulk of the dust mass in Arp 220 is at a temperature of about 50 K. Thus, the bright HNC 1-0 emission is unlikely to originate in cold cloud cores. We conclude that for objects like Mrk 231 and Arp 220, where there is no mm excess emission, and where the overall dust temperature is high, HNC 1-0 does not trace cold dense gas. Instead one or several of the scenarios below must apply. However, for less extreme objects such as NGC 1808 there could be enough mass in cold cores that at least a fraction of the HNC emission could emerge from them.

Steady state chemistry models by S92 show both a temperature and a density dependence in the ${[{\rm HCN}] \over [{\rm HNC}]}$ abundance ratio. For example, there is a very significant difference between $n({\rm H}_2)=10^4~\rm {cm^{-3}}$ and $10^7~\rm {cm^{-3}}$ . For $T_{\rm kin}= 50$ K, ${[{\rm HCN}] \over [{\rm HNC}]}=0.8$ for 10⁴ and ${[{\rm HCN}] \over [{\rm HNC}]}=67$ for 10⁷. If the bulk of the HCN and HNC emission is emerging from gas of densities 10⁴-10⁵ then the relative HNC abundance there may be substantial, despite the high temperature. The reason for this is that at lower densities reactions with HCNH⁺ (HCN and HNC reacts with H₃ ⁺ to form HCNH⁺) become more important. The ion abundance is higher and once HCN and HNC become protonised, HCNH⁺will recombine to produce either HCN or HNC with 50% probability. At higher densities, the ion abundance is likely lower and reactions like HNC + O $\rightarrow$ CO + NH become more important at high temperatures. This scenario is interesting since the electron and ion abundance is likely higher in PDRs. Therefore, in a PDR chemistry, the connection between HNC abundance and kinetic temperature may also be weak since we there expect the HCNH⁺ reactions to be important. Since the CN ${2-1 \over 1-0}$ ratios we measure indicate subthermal excitation it is reasonable to assume that most of the HCN, HNC and CN emission is indeed emerging from gas where the density is below $10^5~\rm {cm^{-3}}$ .

In Orion, S92 find peak-to-peak intensity ratios between ${{\rm HCN} \over {\rm HNC}}$ of 3 to 4 towards the hot core and ridge. However, the abundance ratio is much higher, $\approx$ 80. Thus, it is possible that the fairly bright HNC emission in some galaxy centers is caused by optical depth effects. In objects where the HCN emission is subthermally excited (like in Mrk 231) the optical depth of the 1-0 line could be quite high and perhaps explain part of the apparently too-bright HNC.

Both HCN and HNC may be pumped by an intense mid-IR radiation field boosting the emission from low density regions. There has been no direct evidence IR pumping is dominating the HCN excitation in external galaxies. However, Barvainis et al. (1997) suggest IR pumping as a possible mechanism behind the HCN emission of the Cloverleaf quasar. Ultraluminous galaxies, such as Mrk 231 and Arp 220, have central mid-IR sources with optically thick radiation temperatures well in excess of those necessary to pump the HCN molecule (Soifer et al. 1999). For HNC the coupling to the field is even stronger than for HCN, thus increasing the probability for IR pumping in extreme galaxies, such as Mrk 231. Even if the HNC abundance is lower compared to HCN the HNC emission may have a higher filling factor due to the IR pumping (if it allows emission from gas clouds otherwise at too low density to excite the HNC molecule). A comparative excitation study of HCN and HNC would help cast light on this issue. Of course, the HNC abundance must be high enough so that the pumping can be effective and it must thus happen in regions where the chemistry is dominated by ion-neutral reactions.

4.2 Interpreting the CN emission

4.2.1 CN chemistry

Studies of Galactic molecular clouds have shown that the ${[{\rm CN}] \over [{\rm HCN}]}$ abundance ratio is increasing in the outer regions of UV irradiated clouds (e.g., Greaves & Church 1996; Rodriguez-Franco et al. 1998). The abundance of the CN radical becomes enhanced at the inner edge of a PDR (at an $A_{\rm V}$ of about 2 mag) via the reaction ${\rm N} + {\rm C}_2 \to {\rm CN} + {\rm C}$ or via ${\rm N} + {\rm CH} \to {\rm CN} + {\rm H}$ . At larger depths into the cloud the CN abundance radically declines and the ${[{\rm HCN}] \over [{\rm CN}]}$ abundance ratio increases (Jansen et al. 1995). Most of the CN is present in a part of the cloud where the abundance of free electrons is rather large, $X(e) \approx 3 \times 10^{-5}$ . CN is also a photodissociation product of HCN. Thus the CN abundance should be favoured in a molecular cloud ensemble dominated by PDRs. Furthermore, chemical models (e.g., Krolik & Kallman 1983; Lepp & Dalgarno 1996) show that the CN abundance should also be enhanced when the X-ray ionization rates are high - as might be the case near an AGN.

It is of course difficult to translate a measured ${{\rm HCN} \over {\rm CN}}$ 1-0 line intensity ratio to an abundance ratio between the two species. The spingroup line ratios (see Sect. 3.2.1.) show that the CN 2-1 emission is optically thin for most galaxies we have measured. The emission of the CN molecule is distributed in a greater number of transitions than HCN, thus the optical depth per transition is often lower for CN, reducing the intensity per line. The CN luminosity then becomes a measure of the total number of CN molecules (at least in a comparative sense, given a constant excitation situation from galaxy to galaxy). For the CN 1-0 line we have only information for Arp 220 where the relative faintness of the second spingroup suggests that the optical depth of the 1-0 line is also low. So, the measured (total) ${{\rm HCN} \over {\rm CN}}$ line intensity ratio will give a reasonable idea of the abundance ratio if also the HCN line is close to being optically thin (and a lower limit to the abundance ratio if it is not) and if the same excitation temperature can be assumed. The critical density of the CN line is lower by a factor of a few, so its $T_{\rm ex}$ is likely somewhat higher.

4.2.2 AGNs or starbursts?

We were surprised to find that CN was difficult to detect in several of the brightest galaxies like Mrk 273, NGC 2623, NGC 6240, IC 694 and NGC 34. Four of these galaxies are AGNs and it is tempting to speculate that the CN deficiency is related to the nuclear activity. This seems contrary to models (see above) which predict an increase in the CN abundance in an X-ray chemistry. High resolution studies of nearby systems which contain both starburst and AGN activity (like NGC 1068) will reveal whether CN emission is associated with one or both of the activities.

$\begin{figure} \par\resizebox{14.5cm}{!}{\includegraphics{H2895f3.eps}} \end{figure}$

Figure 3: CN 1-0 spectra for the galaxies with detections. The scale is in $T_{\rm A}^*$ . The spectrum for Arp 220 shows the Gaussian fit for both spingroups, since the line is broad enough for them to be blended. The velocity resolution ranges from 10 to 50 ${\rm km\,s}^{-1}$ . It was selected to be 10% of the FWHM of the line itself, but for a few cases the S/N of the spectrum required further smoothing.

$\begin{figure} \par\resizebox{14.5cm}{!}{\includegraphics{H2895f4.eps}} \end{figure}$

Figure 4: CN 2-1 spectra for the galaxies with detections. The scale is in $T_{\rm A}^*$ . The band is centered in between the two spingroups, apart from Arp 220, where the spectrometer was centered on the J=5/2-3/2 line. The two spingroups are marked with arrows and are blended for all galaxies, apart from NGC 7130. FOR Arp 220 emission is not detected at the line center (marked by the arrow) instead, there is a tentative detection at V=5170 ${\rm km\,s}^{-1}$ (blueshifted from $V_{\rm c}$ by approximately 250 ${\rm km\,s}^{-1}$ ). The velocity resolution ranges from 10 to 50 ${\rm km\,s}^{-1}$ . It was selected to be 10% of the FWHM of the line itself, but for a few cases the S/N of the spectrum required further smoothing.

4.2.3 The CN and the [C II] 158 $\mathsf{\mu}$ m line

In ULIRGs such as Arp 220 and Mrk 231 the [CII] 158 $\mu$ m fine structure line is found to be abnormally faint compared to other, less FIR luminous, starburst galaxies like NGC 3690 (e.g., Luhman et al. 1998). This is interesting, since one would expect the emission from a standard PDR tracer, like the [C II] line, to be bright in a galaxy that is believed to be powered largely by mighty starbursts. Malhotra et al. (1997) report a decreasing trend in ${{F_{\rm [C II]}} \over {F_{\rm FIR}}}$ with increasing ${f(60) \over f(100)}~\mu$ m flux ratio.

Several possible explanations for the [C II] faintness are brought forward (e.g., Malhotra et al. 1997; Luhman et al. 1998; van der Werf 2001). The PDRs may be quenched in the high pressure, high density environment in the deep potentials of the ULIRGs and the HII regions exist in forms of small-volume, ultracompact HII regions that are dust-bounded. The [C II] line may become saturated either in low density ( $n \propto 10^2~\rm {cm^{-3}}$ ) regions of very high UV fields ( $G_0 \propto 10^3$ ) or in dense ( $n \propto 10^5~\rm {cm^{-3}}$ ) regions of more moderate UV fields (G₀ = 5-10). A soft UV field from an aging starburst is another possibility. A higher dust-to-gas ratio would also decrease the expected ${{F_{\rm [C II]}} \over {F_{\rm FIR}}}$ ratio.

The molecular ISM of ULIRGs seems to be characterized by subthermally excited CO and very bright emission from HCN (e.g., Downes & Solomon 1998; Aalto et al. 1995; Solomon et al. 1992). Crudely, this can be modelled as dense clouds ( $n = 10^4{-}10^5~\rm {cm^{-3}}$ ) embedded in a low density ( $n= 100{-}500~\rm {cm^{-3}}$ ) continous medium. This simple scenario may fit well with the two scenarios resulting in saturated [CII] emission.

We have three ULIRGs in our sample: Arp 220, Mrk 231 and Mrk 273 their ${{\rm HCN} \over {\rm CN}}$ line intensity ratio changes from 1 to $$\mathrel{\mathchoice {\vcenter{\offinterlineskip\halign{\hfil $\displaystyle ...$ 6. The deficiency of CN in Mrk 273 is consistent with the lack of [CII] emission and can be an indication that the PDRs are not forming in the dense gas. In Arp 220 the CN emission from the western nucleus is strong and an indicator that a fair fraction of the dense gas is in fact in a PDR state. The UV radiation is strongly affecting the dense molecular clouds here. Clearly we need more information on the properties of the dense gas to find an explanation for the lack of [CII] emission.

4.3 Line ratios and starburst evolution

We speculate whether line ratios of dense gas tracers can be used to explore the evolutionary stage of a starburst. The observed galaxies can be sorted into rough categories based on their line ratios.

4.3.1 Warm dense gas - a young starburst or shocks?

Objects that show bright HCN emission, but little or no HNC or CN, may be dominated by warm dense gas early in their starburst development. The Orion KL region is an example of a Galactic warm ( $$T \mathrel{\mathchoice {\vcenter{\offinterlineskip\halign{\hfil $\displaystyle ...$ K), dense core where I(HCN) is significantly greater than both I(CN) and I(HNC) (e.g., Ungerechts et al. 1997). Also the emission from the 10-9 transition of HC₃N is brighter than the CN and HNC emission which is a typical signature of warm, dense gas. However, shocked gas may be an important part of the molecular ISM in the center of a starburst galaxy - in particular if there is a bar in the center where clouds on intersecting orbits collide. The interaction between supernova remnants at the surrounding ISM may also lead to the presence of shocked gas. The effect of the shock is to compress and heat the gas, which in some respects will make it look like an ISM dominated by warm dense cores. The major difference is that the shock, partially or fully, destroys the dust grains and thereby releasing molecules, such as SiO, into the ISM. Therefore, SiO emission is often used as a tracer of shocked gas (e.g., Martin-Pintado et al. 1992). If the shocked gas is allowed to cool after the shock (which may happen quickly since it has been compressed) the HNC abundance will increase rapidly (S92). An example of a galaxy that could have a shock dominated ISM is NGC 6240. Very strong IR emission is emerging from shock-excited H₂ (van der Werf 2001) in between the two merger nuclei and both HNC and CN line emission is faint relative to that of HCN. In a high pressure environment the gas may be dense and warm - but perhaps heated by dissipation of turbulence rather than very young embedded stars. This may not occur during the very early stages of star formation, but rather be a form of aftermath.

4.3.2 PDRs

For those galaxies where I(CN) is $$\mathrel{\mathchoice {\vcenter{\offinterlineskip\halign{\hfil $\displaystyle ...$ I(HCN) a significant part of the ISM should be in PDRs. In the Galaxy they are often found in interface regions between HII regions and molecular clouds (e.g., Jansen et al. 1995) and near planetary nebulae. CN luminous galaxies are very clearly in the phase where the ISM is being strongly affected by an intense UV field. The filling factor of UV illuminated gas must be high, and the clouds probably not too large since the ${[{\rm CN}] \over [{\rm HCN}]}$ abundance ratio drops dramatically with increasing A_V. The mechanical impact of the starburst (superwinds - supernovae) may help in fragmenting the molecular clouds. Emission from complex molecules, such as HC₃N, is faint because of photodestruction. Since the chemistry now, to a large degree, involves ion-neutral reactions (see Sect. 4.1) the HNC abundance becomes less dependent on temperature and I(HNC) may be significant even from a warm PDR. For example, the temperature around a planetary nebula may become very high ( $\approx$ 100 K) but the HNC abundance is substantial enough to result in ${{\rm HCN} \over {\rm HNC}}$ line intensity ratios close to unity (e.g., Herpin & Cernicharo 2000).

For Arp 220, it appears that HC₃N is mainly emerging from the eastern nucleus (see Sect 3.2.1) which would support the notion of an evolutionary difference between the two nuclei. Rodriguez-Franco et al. (1998) show that the emission from HC₃N is bright toward hot, dense cores, while the HC₃N/CN abundance ratio is only 10^-3 in PDRs. Thus the eastern nucleus seems to be in an earlier evolutionary phase where star formation has just begun. In the mid-IR the western nucleus is more prominent than the eastern one (Soifer et al. 1999).

As discussed in Sect. 4.2.1. the absence of PDR tracers, such as CN emission and the 158 $\mu$ m [CII] fine structure line, does not necessarily mean that the burst is young (or shock dominated). In high-pressure, dusty ULIRGs it is possible that the formation of PDRs is suppressed - or that the PDRs are associated with the diffuse lower density molecular material where the classic PDR lines will not be excited.

4.3.3 An evolved burst

Two galaxies (NGC 2623 and IC 694) show fairly bright HNC emission but with undetected CN emission. Very cold (10 K) and dense gas would result in $I({\rm HCN})= 0.3{-}0.5 \times I$ (HNC) and $I({\rm CN}) < I({\rm HCN})$ . In the Galaxy such conditions dominate clouds like TMC-1 and TMC-2 (e.g., Churchwell et al. 1984). We know, however, from other studies of IC 694 that the dense gas is warm ( $$T_{\rm kin} \mathrel{\mathchoice {\vcenter{\offinterlineskip\halign{\hfil $\dis... ...{\offinterlineskip\halign{\hfil$\scriptscriptstyle ...$ K) (Aalto et al. 1998) and the compact CO nucleus of NGC 2623 likely harbours an ISM similar to that of IC 694. The starburst may have evolved beyond a strong radiative impact from the stars while the chemisty is still dominated by ion-neutral reactions at moderate gas density and relatively high electron abundance.