5 CO line and IR flux relations

CO radio line emission and far-infrared dust emission are both good probes of the stellar mass-loss on the AGB. The gas and the dust are also connected to each other in the "dust-driven wind''-scenario (Sedlmayr & Dominik 1995). Therefore, an observational test of the correlation of the two emissions can be useful. In Olofsson et al. (1993) integrated CO line intensities, as we present in this paper, were plotted against the IRAS 60$\mu$m flux densities. The latter is known to provide a reasonable measure of the dust mass loss rate (Jura 1987). Since we have data from different telescopes and up to four transitions we selected only data from the most frequent observatory/transition combinations for a similar comparison. As argued in Olofsson et al. (1993), as a first approximation, intensities measured at different telescopes should scale in proportion to the beam-areas. In such a way CO (1-0) data from SEST where scaled to the OSO scale by multiplying them with 1.8 (the observational results of Olofsson et al. 1993, suggest a value of 1.5 for visually bright C-stars). For the other transitions only data from a single telescope were used (2-1 from SEST, and 3-2 from JCMT). Plots of $\log I$ versus $\log S_{60}$ for the three lines show positive correlations, but with some substantial scatter. The slopes of linear fits are very similar for the three transitions, $\sim 0.65$, but considerably lower than the results for the C-stars where the CO line flux varies essentially linearly with the far-infrared flux (slope$\sim$1.0, Olofsson et al. 1993). In order to study this in more detail we have made a first order correction for the distance using K-magnitudes from the literature (mainly SR_IIa, SR_IIb, Lb_I). The resulting diagrams are shown in Fig. 8.

  

Figure 8: Plots of distance corrected integrated CO line intensities versus 60$\mu$m flux densities

There is very likely a positive correlation in the expected way (e.g., the open circles in the CO(3-2) plot are approximately fitted with a line of slope 1), but the scatter is large in all diagrams. Parts of the scatter in I can be attributed to calibration uncertainties, but these are expected to be significantly less than 50% for the large majority of the stars. Parts of the scatter in S₆₀ could be a stellar contribution to the 60$\mu$m flux density. We have estimated this by fitting combinations of two blackbodies to the spectral energy distributions (see SR_III). It turns out that the stellar contribution is typically only 10% with a relatively small scatter around this value, and consequently this can only lead to minor shifts in the figures (preferentially shifting the low $S_{60}/S_{\rm K}$ objects to the right). The crude distance correction will also lead to a scatter, but it will affect $I/S_{\rm K}$ and $S_{60}/S_{\rm K}$ in the same way. However, none of these effects can explain the fact that the scatter is substantially larger in $I/S_{\rm K}$ than in $S_{60}/S_{\rm K}$. A more detailed analysis of both the CO and the dust emission is required to understand this scatter in detail, but we suggest here that it points to a difference in the mass loss properties of these stars.

We note first that in a given $S_{60}/{\rm K}$-interval the bluer objects (in $[12~\mu{\rm m}]-[25~\mu{\rm m}]$) lie significantly lower in $I/S_{\rm K}$. We find also that objects with very low $I/S_{\rm K}$-values also have low gas expansion velocities ( km s^-1). The extreme case is that of L² Pup with an expansion velocity of about 2.5 km s^-1, one of the smallest ever measured for an AGB-star. Some indication of an IR-colour dependence of the gas and dust mass-loss relation is found in Nyman et al. (1992). They found a difference in slope (of linear fits) in $\log I$ versus $\log S_{60}$-diagrammes between objects in regions II and IIIa of the IRAS two-colour diagram (although there are relatively few objects in region II, and both groups may contain C-stars).

We tentatively conclude from this that it is the intergrated CO intensity that is anomalously low for the blue, low $v_{\rm e}$ objects (as judged from the IRAS two-colour diagramme the 60$\mu$m flux densities are normal). This could be due to a higher, in a relative sense, dust content in these objects suggesting that dust plays less of a rôle for the gas mass-loss of optically bright, O-rich IRVs and SRVs than it does for higher mass-loss rate O-rich objects and C-stars. Possibly, shock waves or radiation pressure on molecules dominate here (e.g., Höfner et al. 1998). Alternatively, these envelopes are so thin that photodissociation of CO makes CO radio line emission a less reliable mass loss estimator. We cannot exclude that also temporal mass loss rate variations have an effect on this ratio since the CO line and the dust continuum emission do not probe the same regions (epochs).

Recently, Josselin et al. (1996, 1998), argued that the ratio $S_{60}/T_{\rm mb}(1-0)$ can be used to discriminate between AGB-stars and supergiants. For the former the ratio falls in the range 20-220, while the latter only show ratios (using the IRAM intensity scale for the CO data). In Fig. 9 we present the distribution of this ratio for our sample (OSO scale for the CO data; OSO and SEST data were combined in the way desribed above).

  

Figure 9: Distribution of $S_{60}/T_{\rm mb}(1-0)$ values (OSO scale)

In order to convert to the IRAM scale used in Josselin et al. our values should be divided by about 2.3 (corresponding to the beam area ratios of the OSO 20 m and the IRAM 30 m telescopes; the observed intensity ratio for visually bright C-stars is 1.9, Olofsson et al. 1993), or equivalently, on the OSO scale the range for the AGB-stars is 40-500, while for supergiants we have 450. We find that all our (bona fide) AGB stars fall well below the ratios expected for supergiants. We find also that as many as $\sim$40% of our stars fall below a ratio of 50, i.e, in a range where Josselin et al. found no stars at all. This is propably due to the fact that our observed sample differs significantly from that of Josselin et al., which is dominated by higher mass loss rate stars as judged from their considerably redder IRAS-colours. In fact, there is essentially no overlap in the IRAS two-colour diagram between the two samples. The most reasonable explanation for the low ratios found in our sample is the significantly lower gas expansion velocities of our stars, which, for a given mass loss rate, decrease the $S_{60}/T_{\rm mb}$-ratio. Possibly, also a lower, in a relative sense, dust content or a different dust composition could play a rôle here. The latter is also indicated by the fact that quite a number of our low mass loss rate objects show a strong 13$\mu$m feature, but no, or a very weak, silicate dust emission feature both in IRAS and ISO spectra (see Sect. 3.4).