3.1. AGN not seen at mid-infrared wavelengths
NGC 4945 is a peculiar and interesting target for studying the relation of AGN and star formation in galaxies. Clear evidence for hidden AGN activity comes from hard X-ray observations. NGC 4945 is amongst the brighest hard X-ray emitting galaxies and exhibits variability of its 13-200 keV flux on timescales of 10hrs, which clearly establishes its AGN origin (Iwasawa et al. 1993; Guainazzi et al. 2000; Marconi et al. 2000). The AGN X-ray emission is heavily absorbed by a column density of 1024.7 cm-2 (corresponding to A), a high value, but within the range observed for Seyfert 2 galaxies (e.g. Risaliti et al. 1999). In unified schemes, the X-ray obscuration measures a line of sight towards the very center. Obscuration towards the NLR probes a different line of sight and is usually significantly lower, making the NLR visible in Seyfert 2 galaxies.
Mid-infrared high excitation lines are able to penetrate a far larger dust obscuration than their optical and UV counterparts. They are therefore ideally suited as tracers of embedded AGN activity. Mid-infrared emission lines like 14.32 µm & 24.32 µm, 7.65 µm and 25.9 µm are prominently present in the spectrum of all Seyferts observed with ISO (Moorwood et al. 1996b; Genzel et al. 1998). On the other hand, the same emission lines are also weakly visible towards, for instance, supernova remnant RCW 103 (Oliva et al. 1999). emission (at the few percent level compared to ) has also been detected in a sample of starburst galaxies (Lutz et al. 1998), again at a level much weaker than seen in typical AGNs. The origin of the weak level emission in these sources is believed to be shocks. A detection of any of the high excitation lines discussed above does therefore not automatically imply the detection of an AGN in NGC 4945.
We do not detect the lines of and . No trace of is seen in the wing of the nearby PAH emission feature (Fig. 1). From Fig. 1 it might appear that the two lines were indeed detected. However, at the level at which the features appear, instrumental effects play a significant role. In the 14.35 µm line scan a strong fringe in the relative spectral response function coincides exactly with the expected position for the line. Depending on the size of the emitting area, the feature may be entirely attributed to this instrumental effect. We therefore chose to state an upper limit for the 14.32 µm line. The feature seen in the other scan, centered at 24.4 µm, was registered by only two detectors, although 12 detectors scanned over the central wavelength. As visible in Fig. 1, the feature is redshifted with respect to the NGC 4945 systemic velocity. This shift is not observed for any other line we observed. We therefore derive an upper limit for this line too.
The only detected high ionization line in NGC 4945 is the 25.9 µm line. An AGN contribution to this line is possible - to match the limits on higher excitation lines, only part of the emission would be related to an AGN. The detection of possibly shock-related in many starbursts (Lutz et al. 1998) cautions, however, that this may be a more plausible origin of in NGC 4945. The ratio of 0.033 with respect to +0.44 is above average, but well within the range observed for the Lutz et al. (1998) starbursts, also considering that the high extinction (Sect. 3.2) will increase the observed ratio relative to the intrinsic one. A population of Wolf-Rayet stars as origin of the emission seems unlikely. Lutz et al. (1998) have shown that the emission would have to originate in widely dispersed small HII regions and would have to be relatively strong. emission at this level has not been observed in local star forming regions. A conservative analysis will hence not attribute the emission in NGC 4945 to the AGN nor to a population of Wolf-Rayet stars.
Table 2.In X-ray properties NGC 4945 is very similar to Circinus. However, IR indicators for a narrow line region are missing in NGC 4945
The limits on high excitation AGN tracers are consistent with several scenarios, or perhaps more likely a combination of them:
3.2. Starburst properties
Near-infrared broad-band and emission-line imaging has revealed the nucleus of NGC 4945 to be the site of a sizeable starburst, the presence of which is illustrated by the conically shaped starburst superwind-blown cavity traced at many near-infrared wavelengths (Moorwood et al. 1996a; Marconi et al. 2000). Hampered by the large extinction even in the near-infrared, age estimates for the nuclear starburst are sparse and intrinsically uncertain. ISO-SWS offers the possibility for the first time to observe the mid-infrared line ratio 15.56 µm/ 12.81 µm. This ratio, which is much less affected by extinction than visible and UV lines, is sensitive to the hardness of the stellar radiation field and hence is a good indicator for the age of the nuclear starburst. We observed the two lines in the same ISO-SWS aperture, which was centered on the nucleus (see Table 1).
To estimate the extinction to the NGC 4945 nuclear starburst we use the ratio of the 18.71 µm and 33.48 µm [SIII ] lines. This ratio is commonly used as a density diagnostic for the density range 102.5-104.5 cm-3 and is only mildly dependent on the temperature of the emitting gas. Assuming a typical starburst gas density of 300 cm-3 (Kunze et al. 1996; Rigopoulou et al. 1996), the intrinsic ratio should be 0.43 (i.e. the value in the low density limit, computed using the collision strengths of Tayal & Gupta 1999). The observed ratio is far lower: 0.140.06. We hence deduce a screen extinction of , which, using the Galactic center extinction law of Draine (1989; with ), amounts to AV=36. This value is to be considered an upper limit in case the [SIII ] emitting area is larger than 1427", in which case an aperture effect causes the [SIII ] ratio to be a lower limit.
Another independent estimate of the extinction is usually obtained from hydrogen recombination line strengths, assuming `case-B' conditions. For NGC 4945 we therefore observed the 4.05 µm Br and the 7.46 µm Pf line. Both were measured in the same aperture. The ratio Pf/Br is 0.250.10, whereas `case-B' recombination theory predicts a ratio of 0.32. The extinction at 7.46 µm must therefore be similar or slightly larger than at 4.05 µm. This indicates that the grain composition is unusual and probably more similar to the composition found in the line of sight towards the Galactic center (Lutz et al. 1996; Lutz 1999) than found towards other parts of our Galaxy. An extinction towards the NGC 4945 nuclear starburst can therefore at present not be derived from lines in the 4-7 µm range.
The extinction derived for the nuclear starburst is somewhat larger than the value we derive for the warm molecular hydrogen (see Sect. 3.5). This indicates that the warm molecular hydrogen and nuclear starburst emission are coming from different nuclear components, the latter possibly more enshrouded than the former. With the unusual grain composition in mind, it is striking how well the Galactic Center extinction law fits our molecular hydrogen data, resulting in a smooth excitation diagram, even for the H2 (0-0) S(3) line in the center of the 9.7 µm silicate feature (see Sect. 3.5). We are therefore confident that the extinction correction for the starburst, derived using the [SIII ] ratio, is also useful.
In order to determine the excitation of the nuclear starburst we apply the extinction correction derived from the ratio to the observed / ratio. The extinction corrections amount to A(12.8 µm)=1.51 and A(15.6 µm)=1.19. The extinction corrected / ratio is 0.064. Thornley et al. (2000) list observed / ratios for 26 starburst galaxies, all measured in the same ISO-SWS configuration. Clearly, NGC 4945 is among the lowest excitation targets in their sample. Note that the ISO-SWS aperture used is large compared to the typical size scales in starbursts. The ratios listed by Thornley et al. (2000) should therefore be regarded as aperture averaged.
For starburst galaxies is another measure of the excitation of star clusters. Depending on the upper mass cut-off, the star formation decay time scale and the age of the clusters, Thornley et al. (2000) modeled to lie between 3 and 200. The measured values for starburst galaxies range between 3 and 50. Below we will determine for the NGC 4945 nuclear starburst. We assume (i.e. no AGN contribution to ) and estimate from the dereddened 4.05 µm Br flux. For A(4.05 µm)=1.2 (applying the Galactic center law of Draine (1989) for AV=36) and we find and . Using the 12.81 µm line and the empirical scaling (Genzel et al. 1998) a similar result is obtained.
Given the variety of possible star forming histories, it is hard to constrain the age of the nuclear starburst (assuming no AGN contribution to ). However, both excitation diagnostics agree on a low excitation which suggests an evolved burst with an age in excess of 5106 years, but would also be consistent with a low IMF upper mass cut-off.
Marconi et al. (2000) show that it is possible to construct starburst models which are consistent with their near-infrared observations of NGC 4945, but differ by the total luminosity generated (their Fig. 4). An instantaneous burst would not leave space in the energy budget for a sizable contribution from the hidden AGN, whereas a combination of instantaneous burst and constant star formation would. We would like to point out here that the latter model would be inconsistent with the low / ratio observed by us. Only their instantaneous burst is in agreement with both the near-infrared and mid-infrared observations.
3.3. What powers the nucleus of NGC 4945?
The large extinction towards the nuclear starburst and the AGN buried within, makes it very difficult to assess the contributions of either component to the nuclear bolometric luminosity.
The optical, near-infrared, mid-infrared and far-infrared spectra of NGC 4945 are entirely consistent with a starburst-like nature: BLR or NLR high-excitation lines are absent; the starburst excitation indicator / has a starburst-like value; the ratios of 6 µm (ISO-PHT-S), S12 or S25 to S60 or S100 fluxes are all very low and consistent with emission from cold dust only. Furthermore, the ratio , is perfectly consistent with the low excitation of the starburst as deduced from /. And last, NGC 4945 lies on the radio-far-infrared correlation for starburst galaxies (Forbes & Norris 1998). Hence, the starburst might well account for the the entire observed bolometric luminosity.
On the other hand, Guainazzi et al. (2000), who have observed the AGN in NGC 4945 in hard X-rays, compute the AGN to be able to account for all the bolometric luminosity observed, if it has a typical quasar ratio. Since there is no such thing as a template AGN spectrum, the conversion factor applied, (Elvis et al. 1994), may have an uncertainty which could easily allow for the NGC 4945 starburst to dominate the bolometric luminosity instead.
The same uncertanties surround the accretion rate of the black hole inferred from H2O maser observations (Greenhill et al. 1997). A high but not implausible rate of 50% of the Eddington rate would suffice to power the observed bolometric luminosity. Given the wide range of efficiencies inferred for Seyferts, this information does not add anything to identify the dominant power source.
In this complex situation with two potentially dominant power sources, the most significant constraint on their relative weight is the total ratio and its implications on the of the starburst component. is directly constrained by observations, but changes for different assumptions on the starburst and AGN contributions to the total bolometric luminosity. If there is a significant AGN contribution, will be less than the global value of 28. Values as low as 3 which are possible for a zero age massive star population with Salpeter IMF (e.g. Leitherer & Heckman 1995) are strongly inconsistent with the low excitation observed in NGC 4945. Thornley et al. (2000) model [NeIII ]/[NeII ] and ratios of starbursts, taking into account clusters of different ages within the ISO-SWS aperture. An evolving starburst with [NeIII ]/[NeII ]=0.064 as in NGC 4945 must have a (their Fig. 8). This limit simply reflects the higher of later type O stars and persists if the low excitation is due to an upper mass cut-off rather than evolution. With a lower limit of 15 on , the starburst contribution to the bolometric luminosity must be at least 50%.
We hence conclude that the AGN in NGC 4945 plays a secondary although most likely not insignificant role in the energetics of this nearby galaxy. Extremely small values for the AGN contribution to the bolometric luminosity would imply an unrealistically high ratio of for the AGN. The very low inferred black hole mass, the very cold mid-infrared to far-infrared colors, and the absence of any clear line of sight towards the AGN, support our view that starburst activity dominates AGN activity in NGC 4945.
3.4. Emission and absorption features
The infrared spectrum of the central region of NGC 4945 obtained with ISO-PHT-S (see Fig. 2) presents a new view of the ISM in starburst galaxies. Even at the low spectral resolving power of R90 the spectrum is dominated by a wealth of emission and absorption features.
Especially prominent is the family of PAH emission features at 3.3, 6.2, 7.7, 8.6 and 11.3 µm, which ISO confirmed to be common-place in most galactic and extragalactic ISM spectra (e.g. Acosta-Pulido et al. 1996; Rigopoulou et al. 1999; Mattila et al. 1999; Clavel et al. 2000). Nevertheless, the weakness of the 8.6 and 11.3 µm PAH bands in NGC 4945 is unusual. Consistent with and with the strength of the absorption features discussed below, we explain this weakness by heavy extinction, which will suppress these two features placed in the wings of the silicate absorption feature.
Perhaps the most important result, however, is the rich absorption spectrum, indicating that we are observing the infrared sources in the central region of NGC 4945 through a medium containing molecular ices. Interstellar absorptions of 4.27 µm (2343 cm-1) solid CO2 and 4.68-4.67 µm `XCN'+CO are detected, the first time in an extragalactic source to our knowledge. At our resolving power and signal-to-noise we cannot determine the contribution of 4.62 µm (2165 cm-1) XCN, 4.67 µm (2140 cm-1) CO ice and of gaseous CO absorptions to the 4.58-4.67 µm absorption complex. The strength of the XCN absorption appears to be remarkable, suggestive of ice grain processing in an energetic environment (Lacy et al. 1984). We defer a more detailed analysis of the XCN/CO feature to a future paper, which will also include the results of follow-up observations with ISAAC at the VLT. A strong silicate feature is observed around 9.7 µm (see also Moorwood & Glass 1984). A deep minimum is also detected around 3.0 µm, which is suggestive of water ice (or more precise, the O-H stretch) absorption. Table 3 gives column densities for some of the absorption features discussed above. The presence and strength of these absorption features is consistent with the high starburst obscuration derived from the emission lines (but see also Chiar et al. (2000) for variations in the strength of features along lines of sight of similar AV).
Table 3. Column densities of solid state features towards the nucleus of NGC 4945 and towards the Galactic center (SgrA*). In order to derive the column densities we integrated d over the width of the band and divided the result by the bandstrength A. was determined from AV, where AV=30 for SgrA* and AV=36 for NGC 4945
At the resolution of ISO-PHT-S the molecular absorption features in NGC 4945 show striking similarities with the features seen in the ISO-SWS spectrum of the line of sight towards the Galactic center (Lutz et al. 1996; see Fig. 2). Observations at our spectral resolution do however not permit a detailed comparison. Regarding the 4.26 µm CO2 feature it is likely that the feature can be attributed to solid state CO2, since high spectral resolution ISO-SWS observations of other sources indicate that the contribution of gaseous CO2 to the observed feature is very small (see in particular van Dishoeck et al. 1996). In the 4.6-4.8 µm region, the spectra of NGC 4945 and the Galactic center differ more strongly, and a more careful inspection is required to assess the contributions of gaseous and solid CO and XCN. ISO-SWS spectroscopy of the Galactic center (Lutz et al. 1996) clearly shows that what we see at low resolution as a relatively shallow and broad feature is in fact dominated by individual lines of gaseous CO. Contributions of a potential underlying solid CO/XCN component are possible but difficult to separate until our high resolution follow-up observations have been executed.
3.5. Molecular hydrogen: physical conditions, excitation and mass
Near infrared observations of molecular hydrogen emission in NGC 4945 have been reported by several authors over the last 15 years. The most complete set of observations is published by Koornneef & Israel (1996), who observed 8 ro-vibrational transitions with IRSPEC at the ESO NTT. With ISO-SWS and ISO-PHT-S we have extended the number of observed lines from 8 to 14 by observing the pure rotational transitions S(0), S(1), S(2), S(3), S(5) and S(7) as well as the (1-0) Q(3) line. The latter was also observed with IRSPEC and can therefore be used to determine the proper aperture correction factor for the IRSPEC data set. An overview of the observed lines is presented in Table 4.
Table 4. NGC 4945 molecular hydrogen data. A() is the extinction correction in magnitudes; Aul is the Einstein coefficient for the transition from level u to level l. is the upper level energy of level u; is the statistical weight of level u; (,J) is the number of H2 molecules in upper level u
Information on the spatial extent of the H2 emitting region is only available for the 2.12 µm (1-0) S(1) line (Koornneef & Israel 1996; Moorwood et al. 1996a; Quillen et al. 1999; Marconi et al. 2000). Based on Fig. 3a of Moorwood et al. (1996a) we estimate that more than 90% of the (1-0) S(1) emission fits within the smallest ISO-SWS aperture (1420"). It is not unreasonable to expect the H2 emitting area to increase with decreasing H2 temperature. The aperture sizes used to observe the respective H2 transitions increase with increasing sensitivity to lower temperature H2. Based on this, we will assume in what follows that ISO-SWS and ISO-PHT-S have observed all available warm H2. Further to this, all three instruments were centered on the same nuclear position and viewed the nuclear region under more or less similar position angles (see Sect. 2). We will use the ratio of the 1-0 Q(3) line fluxes measured by ISO-SWS and IRSPEC to scale the other IRSPEC lines to the ISO-SWS aperture size. This ratio is 2.33.
From the 14 transitions observed it is possible to compute the upper level populations for 12 levels. We assumed the H2 levels to be optically thin. The excitation diagram in Fig. 3 shows a plot of the natural logarithm of the total number of H2 molecules (N), divided by the statistical weight (g), in the upper level of each transition detected, versus the energy of that level (E/k). The plot shows the situation after extinction correction (see below).
The excitation temperature of the gas is the reciprocal of the slope of the excitation diagram. If the warm H2 is in LTE, the excitation temperature directly corresponds to the kinetic temperature. As is clearly visible from Fig. 3, (extinction corrected; see below) the excitation temperature increases monotonically with upper level energy, from 160 K for the combination of (0-0) S(0) & S(1) to 2200 K for the ro-vibrational lines.
In a highly obscured galaxy like NGC 4945, extinction corrections to the H2 emission will be important. The extinction can be estimated from the H2 data themselves taking into account that any known excitation mechanism should produce a "smooth" excitation diagram for the pure rotational lines, and that transitions originating in a common upper level should give consistent results. More specifically, we use three criteria:
We have varied the extinction and tried several extinction laws. We present the most applicable extinction laws here:
From Fig. 3 and the criteria defined above, moderate extinctions of AV=17-23 are clearly preferred. Extinction law A provides a somewhat better fit than extinction law B. None of the 3 solutions gives a good fit to the (1-0) Q(4) data point. In the following analysis, we use the preferrred extinction correction of and extinction law A. We note that the extinction to the H2 emitting region is slightly less than that to the starburst regions (Sect. 3.2). This plausibly matches the morphological results of Moorwood et al. (1996a), who find the starburst in an obscured disk, but the H2 emission extending into a less obscured wind blown cavity.
A rough estimate of the amount of warm molecular hydrogen in the nucleus of NGC 4945 can be derived from the level populations of the pure rotational S(0) and S(1) transitions. The excitation temperature for the upper levels of these transitions (J=2 and J=3) is 160 K. Assuming the same excitation conditions for the J=0 and J=1 levels, we compute a warm molecular hydrogen mass of . This is 9% of the total H2 gas mass estimated from CO and 0.7% of the dynamical mass interior to the molecular ring (Bergman et al. 1992; see below).
As already noted, the excitation temperature changes significantly with level energy. This is the consequence of the natural fact that the emitting gas will consist of a mixture of temperatures. The rich NGC 4945 dataset allows us to address this in a more quantitative way. Experiments with fits assuming a number of discrete temperature components lead us to suggest that a power law might give a good representation of the mass distribution as a function of temperature. We obtain a good fit for the following power law: dM/dT=4.431015 T-4.793 . The quality of the fit is shown in Fig. 4.
Table 5 gives warm molecular masses for several low temperature cut-offs. Since H2 gas at temperatures below 70 K does not contribute to the (0-0) S(0) flux, nor to any of the other pure rotational lines, we cannot verify whether our power law mass distribution continues down to the lowest temperatures. Nevertheless, we have included mass estimates down to a low temperature cut-off of 50 K. This number is reasonable, since we don't expect the giant molecular clouds (GMCs) to be as cold as in the Galactic disk (10-20 K). Rather we expect conditions as found near the Galactic center, where the GMCs are believed to have temperatures exceeding 50 K (Armstrong & Barrett 1985).
Table 5. Warm molecular hydrogen mass estimates for the nucleus of NGC 4945 using the best fit power law dM/dT=4.431015 T-4.793 M/K. The total H2 gas mass estimated from CO amounts to 2.7108 M
It is interesting to compare our warm molecular hydrogen gas mass estimate with values found in the literature (see Moorwood & Oliva 1994for a review). Bergman et al. (1992), using the inner molecular rotation curve of Whiteoak et al. (1990), compute a dynamical mass interior to the molecular ring (R280 pc=15.6") of . The same authors use CO to derive a total H2 gas mass of for the ring, assuming the rather high kinetic gas temperature of 100 K. Note that a low temperature cut-off of the order 50-60 K in our H2 temperature distribution would bring our estimate of the total H2 gas mass in agreement with that derived from the low level CO observations. The total H2 gas mass of agrees well with a starburst-like position of NGC 4945 in the -M(H2) diagram (Young & Scoville 1991).
In Table 6 we list for a number of external galaxies and Galactic template sources temperatures and masses of the warm molecular hydrogen gas. With a value of 9%, NGC 4945 has a warm H2 gas fraction similar to that found for most of the other galaxies listed. Note however that the results for NGC 3256, NGC 4038/39 and Arp 220 are less well constrained than for NGC 6946 and NGC 4945: only for the latter two can the temperature of the warm H2 gas be determined from the (0-0) S(0) and S(1) transitions directly. For the same reason a comparison of the H2 gas temperatures is of limited value unless they are derived from identical line combinations. Fairly low temperatures can be derived from the (0-0) S(0) and S(1) lines, 160 K and 179 K for NGC 4945 and NGC 6946, respectively. Limits for other galaxies listed in Table 6 are consistent with a similarly low temperature. This temperature is well below that observed for an Orion type shock (500 K). It is closer to what is observed for the same line combination in PDRs (e.g. Orion Bar: 155 K, D. Rosenthal priv. comm.; S140: 159 K, Draine & Bertoldi, 1999). While a variety of regions may contribute to the galaxy-integrated temperature distribution, this comparison clearly shows Orion-like shocks to be not representative for the entire emission, and fairly normal PDRs (or less energetic shocks) to be perhaps more typical. If excited by shocks (as suggested by the morphology, Moorwood et al. 1996a), then the near-infrared H2 emission in NGC 4945 must trace a small subcomponent of faster shocks.
Table 6. Warm molecular hydrogen in external galaxies and Galactic template sources. In column 4 detected lines are printed bold, upper limits normal. T01 refers to the excitation temperature computed from the (0-0) S(0) & S(1) fluxes. Likewise, T12 refers to the excitation temperature computed from the (0-0) S(1) & S(2) fluxes. is the best fit excitation temperature to several of the lowest rotational transitions. The warm H2 gas mass is computed using the gas temperature listed in any of the preceeding three columns. The last column gives the fraction of H2 gas in the warm component
© European Southern Observatory (ESO) 2000
Online publication: June 5, 2000