4. Analysis and discussion
4.1. Continuum colours
As noted by HIGW, the stellar absorption spectrum of the nucleus at m and the overall spectral shape of the near-IR continuum suggest that out to m the radiation from the centre of the galaxy is dominated by emission from late-type (super)giants in the bulge. Near-infrared photometry was published by Lawrence et al. (1985), Willner et al. (1985) and Forbes et al. (1992). The colours derived by Forbes et al. (1992) suggest relatively low extinction, but this must be interpreted as a lower limit because these authors used fairly large apertures, and, more importantly, first aligned the peaks of their J, H and K-band images, while we have noted in Sect. 3.1 that these are actually displaced from one another. The colours measured by Willner et al. (1985) suggest that the central 6" of NGC 3079 suffers a high extinction (their Figs. 2 and 3). HIGW analysed the available data and concluded to a "best" value for the central 6".
We have used our accurately aligned J, H and K-band images to produce the two-colour diagram shown in Fig. 5, and we analyse the colours of the various regions in NGC 3079 in the following sections. The curves marked "screen" and "mixed" in this figure indicate the effects upon the observed colours of extinction by respectively foreground dust and dust uniformly mixed with the stellar population. In the former case, , while in the latter case, , where and are respectively the observed and intrinsic intensities. A Galactic extinction curve is assumed.
4.1.1. The outer bulge
The colour observed from the outer bulge (i.e., positions 4" or more east of the disk) agrees well with that of typical bulges, but the colour is about bluer. These results are similar to those found by HIGW in a aperture. From the observed K-band absorption features, HIGW determined that the mean spectral type of the bulge population dominating the K-band light is M0III, reasonably consistent with the "bulges" zeropoint in Fig. 5 and with the observed colour of the eastern bulge in NGC 3079. This leaves the blueness of the eastern bulge in to be explained. HIGW note that blue colours also prevail at wavelengths shorter than m and that these colours point to the contribution of young stars at m and shorter wavelengths. There are several possibilities.
4.1.2. Disk and western dust lane
In the disk away from the central part of NGC 3079, the colours of the stellar population must correspond to a mean spectral type earlier than those indicated by the "bulges" zeropoint, i.e. close to those of the "main-sequence line" in Fig. 5. If, for instance, a mean spectral type A0 is assumed for the intrinsic "disk" colours, a "screen" reddening corresponding to would be required; the dust lane would require a mean spectral type of early B and . Since recent star formation is associated with dust, this result is reasonable. Alternatively, the "mixed" extinction model may be a more adequate representation of the relative distributions of stars and dust in the disk of the galaxy. This model suggests a mean spectral type of late F or early G, but also rather high values for the extinction optical depth. In reality, a combination of "screen" and "mixed" extinction is probably appropriate, and intermediate values for both spectral type and visual light loss are obtained by varying the relative importance of "mixed" and "screen" extinction. The present data do not allow us to draw a firm conclusion as to which combination is preferred.
4.1.3. Stellar emission and K-band excess in the central kiloparsec
In Fig. 2a, we have shown the and colours in 1" diameter apertures along the major axis of NGC 3079. A number of these positions are also marked in Fig. 5,. Towards the nucleus, the colours redden rapidly, reaching peak values , at the nuclear arcsec2 (8787 pc). As Fig. 5 shows, the () colour within from the nucleus is far too red to be explained by either the "screen" or the "mixed" extinction curves in Fig. 5, so that excess emission must be present in the K-band. The amount of excess K-band emission implied by the near-infrared colours depends, however, on the choice of extinction model. We argue here that the "screen" model is appropriate, for the following reasons.
All of these arguments indicate that the extinction towards the nuclear region is dominated by foreground material. We will therefore in the following assume foreground extinction exclusively.
4.2. The central region of NGC 3079
4.2.1. Hot dust
The JHK colours alone are insufficient to uniquely separate the effects of extinction and dust emission, as the observed points in Fig. 5 may be reached by different tracks. They do constrain, however, the range of admissible parameters.
On the one hand, Fig. 5 shows that the screen extinction limit for the central position is at arbitrarily low . For the other positions, the limit is even lower. On the other hand, for a dust emissivity , dust-mixing curves with temperatures K yield colours too blue in () to explain the observed colours of the central position for any extinction, i.e. they pass to the left of the central point in Fig. 5. As that figure shows, somewhat higher dust temperatures are in fact allowed for the other positions, but in the absence of starburst activity, we consider it unlikely that the projected central 250 pc is cooler than its surroundings. We thus find that K for the center of NGC 3079. The photometry by Lawrence et al. (1985) in a 6" aperture provides an additional constraint. Their observed () and () colours are very close to reddening lines originating in the "bulge" point. A non-negligible dust contribution, required by Fig. 5, is therefore possible only for relatively high dust temperatures that likewise bring the dust-mixing line close to the reddening line. Taking into account the fact that Lawrence et al. (1985) were pointing off the true nucleus, their Fig. 1a suggests that, were they properly centered, the L and M magnitudes may be lower by at most and respectively. The resulting colours, () = and () = , imply that the temperature of dust contributing to the near-infrared emission of the center must be between K. Lower temperatures leave no room for a significant dust contribution at K, and higher temperatures produce colours too blue in () and (). We thus conclude that K. A similar result was obtained by Armus et al. (1994) who concluded from their and m measurements to the presence of significant amounts of hot dust with K; in this respect we also note the absence of cold dust emitting at far-infrared wavelengths throughout the inner 1.5 kpc (Braine et al. 1997). Finally, with radiating hot dust contributing to the near-infrared emission, especially at K, NGC 3079 exhibits characteristics similar to those of the Seyfert 1 galaxies in which Kotilainen & Ward (1994) found significant contributions from dust radiating at temperatures of 600-1000 K.
Limits to the contribution of radiating dust to the emission at m may be estimated from the deep CO-band absorption evident in Fig. 2 of HIGW. Comparison with unreddened late-type stellar spectra by Arnaud et al. (1989) and Lancon & Rocca-Volmerange (1992) suggests that up to 25-30% of the emission from the central may be caused by dust. By assuming i. the intrinsic colours of the stellar population to be those of the typical bulge in Fig. 5, ii. a constant dust temperature K across the nuclear region, iii. identical (foreground) extinction for both radiating dust and stars and iv. no effect of scattering, we estimate from the observed J, H and K-band fluxes a peak extinction in the central pixel of , with a dust contribution in K. The inner molecular disk region has somewhat lower mean values and a dust contribution to observed K. Both values are similar to the nuclear extinction 4 estimated from the Balmer decrement by Veilleux et al. (1994), which provides further support for our conclusion that the extinction occurs mostly or entirely in the foreground. If the radiating dust suffers less extinction than the stars, these conclusions remain unchanged, but the dust contribution to the emitted (deredenned) radiation will be proportionally lower.
The dereddened peak stellar surface brightness is of the order of W mm-1 sr-1. Further analysis shows that the stellar light distribution has a halfwidth of 2" perpendicular to the major axis, and 3" along the major axis. Within the errors, the extent of the hot, radiating dust emission is the same, but its centroid appears displaced from the stellar centroid by to the west. The integrated hot dust emission, corrected for extinction, is W mm-1, corresponding to a luminosity . This is of the far-infrared luminosity of the nucleus and of the mechanical luminosity of the modelled nuclear wind (HIGW). If the dust extinction is lower, these values decrease proportionally.
4.2.2. H2 kinematics
In Fig. 6 we show three position-velocity diagrams along lines parallel to the major axis of the total H2 distribution at , and offset from one another by . This diagram indicates, again, the presence of two H2 emission components: a bright central component with a large velocity width superposed on weaker emission which clearly shows rotation with approaching velocities north-northwest of the nucleus. The bright H2 component can be identified with component C1 in the and OH absorption maps published by Baan & Irwin (1995), while the weaker emission corresponds to their component C2. Their component C3 has no counterpart in our images.
The bright central component has a FWHM diameter along the disk of (225 pc) and a broad, flat-topped velocity distribution which agrees well with the H2 S(1) spectrum obtained by HIGW, who found a trapezoid shape for the line with . The H2 distribution in the lower two panels of Fig. 6 does not show evidence for the velocity gradient of pc-1 ( arcsec-1) assigned to component C1 by Baan & Irwin. If anything, the H2 data suggest a much faster rotation of order pc-1 in the opposite sense, i.e. velocity higher in the north, lower in the south. We suspect that Baan & Irwin may have been misled by the blending of the bright component C1 with the weaker extended component C2 (see their Fig. 3 and see also Fig. 2 of Veilleux et al. (1994).
A qualitative estimate of the velocity width of the S(1) line as a function of position can be obtained by dividing the total H2 emission by the central H2 "channel" only. This shows that the H2 velocity range is smallest in the midplane of the disk, and increases away from the disk towards the east (i.e. in the direction of the conical outflow), and also somewhat to the west. The situation in NGC 3079 appears to be similar to that in the central region of NGC 4945, where Moorwood et al. (1996) found that the H2 emission covers the surface of a hollow outflow cone coated on the inside with H emission, that presumably plays a role in the collimation of this outflow.
The more extended H2 component, which is seen at upper left and lower right in the central panel of Fig. 6, appears to rotate in the regular sense, with a velocity gradient of pc-1 ( arcsec-1). This is somewhat steeper than the gradient of the rigidly rotating CO component (Fig. 5b in Sofue & Irwin 1992) which extends to about 10" from the nucleus, but it is identical to the gradient determined by Baan & Irwin (1995) for component C2. The H2 position-velocity diagram in the top panel of Fig. 6 (east of the midplane) repeats this pattern for the now less bright extended emission. In contrast, only very weak extended emission is present in the position-velocity diagram offset to the west (Fig. 6, lower panel), where extinction must be considerably higher, especially if part of the emitting H2 is outside the midplane of the galaxy. The rotation of the extended component implies a dynamical mass of within pc from the nucleus, or pc-3.
4.2.3. Structure of the central region
We are now in a position to combine the structural information from various observations. In Table 2 we have collected the available size information. In the data discussed so far, three significant scale sizes can be identified.
Table 2. Sizes of emitting components in the central region of NGC 3079
4.3. Molecular gas in NGC 3079
4.3.1. Relation of CO emission to H2 column density
We may connect the total hydrogen column density to reddening and CO intensity by the following relations:
where and are the factors by which respectively the gas-to-dust and the CO intensity to H2 column density ratios in the center of NGC 3079 differs from the canonical values; is in mag, is in K km s-1 and N is in cm-2. Our choice of both and is such that they will be less than unity in environments with metallicities higher than those in the solar neighbourhood.
Baan & Irwin (1995) derive for their extended component C1 an absorption column density , where is the unknown spin temperature. For the same extended region, we find a mean extinction corresponding to . Considering that extinction and HI absorption sample only half of the line of sight sampled by CO emission, we obtain:
From Young et al. (1988) we find that the central 8" (695 pc) yields a J=1-0 CO emission signal K km s-1, so that:
First, we obtain from this equation an upper limit to the spin temperature associated with the extended component C1 by assuming zero H2 column density towards the center: K. If the gas-to-dust ratio in the centre of NGC 3079 is less than that in the solar neighbourhood (), the limit on becomes more stringent. Second, since , we find , i.e., even for a `normal' gas-to-dust ratio ( = 1) CO luminosities in the centre of NGC 3079 correspond to at most a twentieth of the H2 column density we would obtain by applying the `standard' Galactic conversion factor. The conversion factor appropriate to NGC 3079 is thus . This value is also an upper limit because the value of used here applies to a larger area than used for the extinction. For low gas-to-dust ratios (i.e., ) and reasonable spin temperatures, and consequently rapidly become small. Conversely, CO-to-H2 conversion factors similar to that of the Galactic disk () are obtained only for very large gas-to-dust ratios (). Such large gas-to-dust ratios are more characteristic for extremely metal-poor dwarf galaxies than for the centres of spiral galaxies. These results are rather constrained and point to low values of both the spin temperature and the CO-to-H2 conversion factor X for a large range of acceptable gas-to-dust ratios.
A low value for the CO-to-H2 conversion factor is consistent with the `discrepancies' between H2 masses derived from CO and from submillimetre observations, noted in Sect. 5 of HIGW. Moreover, a rather similar result has been derived by Braine et al. (1997). On the basis of their observations, they arrive at a conservative estimate . There is some evidence that within the cavity, the nucleus itself is surrounded by a high-density, parsec-sized accretion disk. Baan & Irwin's (1995) component A exhibits a high value cm-2 K-1. Trotter et al. (1998) argue that this component is part of an inner jet emanating from a heavily absorbed nuclear engine, and that this nucleus is surrounded by a turbulent and presumably dense disk, 2 pc in diameter and traced by H2O maser emission. The nucleus may thus suffer a much higher extinction than the extended region C1. This does not change our results, because both our extinction and the HI absorption value used do not refer to the nucleus, but to the material in front of the extended cavity. In addition, the proposed circumnuclear disk has a very small filling factor with respect to the 8" region sampled in CO emission.
Even if the actual extinction were to be higher than assumed by us, we would still require the conversion factor Xto be much lower than the one applicable to the solar neighbourhood, although the constraints on spin temperature would be much relaxed. Finally, we test our result by calculating the extinction towards the extended central region required by Eq. (3) if and are both set to unity (i.e., Galactic values for the CO-to-H2 conversion factor and the gas-to-dust ratio). In this case Eq. (3) changes to the expression , so that and . This value of the extinction, although possibly appropriate to the very nucleus, is clearly ruled out for the extended circumnuclear region sampled by our data. This underlines the robustness of our conclusion that at least the CO-to-H2 conversion factor in the centre of NGC 3079 is substantially lower than the Galactic value.
The nuclear activity in NGC 3079 may be responsible for the apparent, extremely low [H2]/[CO] abundance. Theoretical models by Neufeld & Dalgarno (1989) predict that this may occur in regions exposed to dissociative shocks. Behind the shock, most of the carbon will be incorporated into CO by gas-phase reactions, but the catalytic formation of H2 is severely inhibited if the essential dust grains are heated to temperatures of the level we propose for the inner parts of NGC 3079. As a result, [H2]/[CO] abundances may be depressed by one to two orders of magnitude. These theoretical predictions at least provide a consistent framework for the interpretation of the phenomena observed in the central region of NGC 3079: the presence of fast, energetic nuclear winds, shocked molecular hydrogen, hot dust and an underabundance of molecular hydrogen with respect to carbon monoxide.
4.3.2. Gaseous content of NGC 3079
Comparison of the interferometric and single dish CO data suggest that of the order of 35 to 50% of the CO emission from NGC 3079 finds its origin in the inner molecular disk within 500 pc from the nucleus (Young et al. 1988). However, in the preceding we have presented evidence for a relatively small amount of molecular hydrogen associated with the bright central CO emission. As the main body of NGC 3079 may well be characterized by more "normal" CO-to-H2 conversion factors (e.g. Braine et al. 1997) we cannot conclude that most of the molecular mass is concentrated in the central region.
Scaling the molecular mass estimate by Young et al. (1988) to our adopted distance and our estimated CO-to-H2 conversion factor, we find for the molecular disk a mass , or of the dynamical mass. For K and , we find a mean ratio , characteristic for an ISM dominated by molecular clouds. For the rest of the galaxy, Young et al. found a mass of which reduces to after scaling to our adopted distance and a "normal" conversion factor of . This is 25 times the central molecular mass, and about a third of the mass of NGC 3079 (Irwin & Seaquist 1991). Including helium, the total mass of gas in NGC 3079 is , or about 10 of its total mass (cf. Irwin & Seaquist 1991). Such a fraction of the total mass is characteristic for late-type galaxies, as is the mean ratio . We note that our molecular gas estimates again are very similar to those derived by Braine et al. (1997) under different assumptions.
It thus appears that the molecular hydrogen content of NGC 3079 is not exceptional compared to that of other late-type galaxies. Rather, the emissivity of CO in the central molecular disk is unusually high.
© European Southern Observatory (ESO) 1998
Online publication: July 20, 1998