6 Discussion

The first main result emerging from this work to date is the discovery of two suitable optical targets located behind and seen through the CSE of IRC +10$\degr$ 216. Star 6 is of special interest, with its offset of 37$\arcsec$, because the CS column density is relatively high, meaning that a new method is now established to study the cold, external, but still molecular layers of this carbon-rich envelope.

The large equivalent widths of the K  I lines are important ($\sim$500 mÅ) and are even larger than those found for very reddened OB stars, like Cyg OB2 #9 with E_B-V=2.25 (Chaffee & White 1982). We note that the lines appear to be composed of at least three velocity components attributable to sheets of gas, probably incomplete shells, consistent with the results of optical imaging (Mauron & Huggins 1999). A better spectral resolution than used here, ($\sim$6 km s^-1), would possibly help to resolve these components and study them separately. Here, we shall adopt a simple approach and consider the observed K  I lines as due to a single broad component. One derives from the doublet data a K  I column density $N_{\rm KI}\sim 3.0\times10^{12}$ cm^-2, and a $\lambda 7699$ line opacity of order unity if a velocity width of the order of the wind expansion speed, i.e. about 10 km s^-1, is assumed. This $N_{\rm KI}$ can be readily compared to the expected column density of potassium (i.e. all K nuclei) by using the model envelope parameters of Sect. 2 and a cosmic abundance of K of 1.3$\times$10^-7 (K is not nucleosynthetized in AGB stars). One finds $N_{\rm K}=2.4\times 10^{14}$ cm^-2, so that, on average along the line of sight in the circumstellar envelope, only about 1% of K nuclei are in the K  I form. For comparison, in the diffuse interstellar medium and for a similar H column density, i.e. for $A_{\rm v}=0.9$, one usually finds $N_{\rm KI}$ around $4.5\times 10^{11}$ cm^-2, within a factor of 2. This is obtained by scaling the data of the ten stars of the catalogue of Chaffee & White (1982) having E_B-V between 0.25 and 0.35, excluding the case of $\zeta$ Oph for which one finds $2\times 10^{12}$ cm^-2. Therefore, our data suggest that on average towards Star 6, the CSE has a $N_{\rm KI}\sim 7$ times larger than that in the common diffuse ISM, for the same $N_{\rm H}$.

A possible explanation for the low K I/K ratio of 1% in the CSE is that the large majority ($\sim$99%) of K atoms are in the ionized K  II form, due to photoionization by galactic ultraviolet ( $\lambda \leq 2860$ Å) photons penetrating the probed outer layers. It is also possible that some K could reside in molecules (e.g. KCN, KCl), or in dust grains. Similar to KCN, the NaCN molecule has indeed been detected in IRC +10$\degr$ 216 by Turner et al. (1994). However, there is no clear indication as to the radial extent of the alkali metal-bearing molecules, i.e. whether they still exist (against photodissociation) at the impact parameter of Star 6. In the case of NaCN, there is a large uncertainty, by a factor of 200, on its abundance, but if it forms in the outer envelope, up to 7% of Na can be in NaCN, according to Turner et al. (1994). Little information exists on depletion of K and Na in carbonaceous grains. At present, it is not possible to perform a detailed quantitative budget of these alkali metals and study their evolution along the gas flow. However, this situation may improve in the future with more sensitive molecular maps to be obtained with ALMA, and supplementary optical spectra of Na, K and other atoms towards other background targets.

The second main result concerns DIBs, and especially the question of whether DIB carriers can be found in circumstellar envelopes of evolved stars. Because it is currently favoured that DIBs could arise in large organic molecules, such as carbon chains, PAHs, C₆₀ and related compounds (e.g. cations), one might speculate that diffuse bands would be readily detected throughout the carbon rich envelope of IRC +10$\degr$ 216. This is not the case. Our spectra show no detectable diffuse band in the envelope of IRC +10$\degr$ 216. This is in agreement with the findings of Le Bertre & Lequeux (1993), at least for some of the carbon-rich objects they examined. For example, according to these authors, the dusty absorbing circumstellar material of NGC 7027, BD+30$\degr$ 3639, Hen 1044, CPD-56$\degr$ 8032, IRAS 21282+5050, HR 4049 and HD 213985 (which are either planetary or proto-planetary nebulae) are depleted in DIB carriers relatively to the ISM. However, as noted in Sect. 1, there are two carbon-rich objects, AC Her and especially CS 776, that might have diffuse circumstellar bands, according to the same authors.

The case of CS 776 is interesting in our context because it is the only genuine AGB outflow with which we can compare our findings. Le Bertre & Lequeux (1993) found that in the line of sight to the A-type companion of the carbon star, which they argue suffers a circumstellar absorption E_B-V=0.53, the $\lambda$5797 DIB is abnormally weak, compared to $\lambda$5780, giving support to the principle of dividing DIBs in families. The strengths of the other DIBs ( $\lambda\lambda 4430$, 5780, 6284) are in reasonable agreement with ISM-like expectations, with $\lambda$5780 being, however, twice as intense. In the context of the present work on IRC +10$\degr$ 216, we would note that the critical question for CS 776 is whether the reddening is indeed circumstellar and not interstellar (see also the comments by Herbig 1995). According to Le Bertre (1990), CS 776 has the following properties: distance 1.3 kpc, mass loss rate of the carbon star $\dot{M}_{\rm H}=5\times10^{-7}~M_{\odot}$ yr^-1, and $v_{\rm outflow}=25$ km s^-1. For its companion at an offset of 1.8$\arcsec$, these characteristics imply a line of sight (tangential) $N_{\rm H}=5.4\times10^{19}$ cm^-2, which would correspond in the ISM to E_B-V=0.0083. This is much lower than the above E_B-V=0.53 attributed to the circumstellar matter by Le Bertre (1990). This $N_{\rm H}$ column density is also much lower than that toward Star 6, for which no DIBs are observed. So, in order to maintain the conclusion of the circumstellar origin (for DIBs and reddening), one has to assume that CS 776 had a much larger mass loss rate in the past (450 yr ago), of the order of $\sim$ $3\times10^{-5}~M_{\odot}$ yr^-1. Although mass loss from AGB stars is known to undergo time variability, its seems to us that this assumption is less probable than simply envisaging an interstellar origin for the DIBs and the observed reddening. We also note that CS 776 has no mid-infrared excess (with flux densities peaking at 25 or 60 $\mu$m), as seen for many post-AGB objects and also a few carbon stars with similar ancient detached envelopes: its IRAS fluxes are 90, 29, 5 and <15 Jy at 12, 25, 60 and 100 $\mu$m, respectively, and are typical of a normal carbon star without a detached shell. Certainly, it would be useful to re-investigate the interstellar reddening in the field of CS 776, and examine again the presence or absence of atomic lines, and DIBs, attributable uniquely to the CS 776 envelope.

It is of interest also to examine our results in the context of other compact objects toward which diffuse circumstellar band carriers have been sought, or indeed observed. For example, it was suggested by our referee to examine a possible H-deficiency in IRC + 10$\degr$ 216, because another very clear case of DIB absence is that of the circumstellar disk of HR 4049, which might be H-deficient (Waters et al. 1989). Although it is true that there is no direct measurement of the abundance of H₂ in IRC +10$\degr$ 216 (i.e. a measurement of the H₂ loss rate), there are several reasons supporting the case that IRC +10$\degr$ 216 is not particularly H-poor. The models fitting many molecular observations generally adopt a ratio C/H₂ of $\sim$10^-3, and collisions with H₂ are the main source of excitation of the molecules (e.g. Groenewegen et al. 1998). A large abundance of hydrogen is also indicated by: a) abundant H-bearing molecules such as C₂H₂ or HCN; b) the presence of HCO⁺, formed from H₃⁺, itself a consequence of cosmic ray induced ionization of H₂ (Glassgold 1996), and c) detection of cold H  I at 21 cm (Le Bertre & Gérard 2001).

The absence of diffuse band carriers seen in absorption in the cool carbon-rich layers of IRC +10$\degr$ 216, and in a number of C-rich planetary nebulae (even those showing UIR features, such as NGC 7027) also suggests comparison with the Red Rectangle. For this object, Schmidt & Witt (1991) have shown that the strong optical emission features attributed to a subset of diffuse band carriers appear only at the bicone interfaces, where carbon-rich material is presumably being eroded by a bipolar plasma flow, and/or by the ultraviolet radiation from the central star. From a study of the spatial distribution and spectral structure of the 3.3 $\mu$m UIR feature, Kerr et al. (1999) suggested that through this erosion the grains might produce even completely dehydrogenated DIB-emitting molecules, perhaps monocyclic carbon ring molecules (see also Kerr et al. 1996). If this scenario were correct, the absence of DIBs in the outer layers of IRC + 10$\degr$ 216 could be understood as arising from the fact that its circumstellar material has not been processed at all in a similar way during its formation and ejection history. The effect of the external UV interstellar radiation field on the probed layers of IRC + 10$\degr$ 216 is probably insufficient to dehydrogenate the dust and fabricate DIB carriers, given the low field intensity, compared to the strong UV irradiation expected from the hot central star of the Red Rectangle. Moreover, the circumstellar dust in IRC + 10$\degr$ 216 is exposed to UV for only a short time, compared to the ISM. Finally, the average circumstellar density in the line of sight to Star 6, of order 2000 H₂/cm³, may also be too large to permit these carriers (if any) to exist freely, as they do in the diffuse low-density ISM and the bipolar cavity of the Red Rectangle.

In summary, these results concerning the prototypical mass-losing AGB star IRC + 10$\degr$ 216 reinforce the evidence that the DIB carriers are absent, or of very low abundance, in the cool winds of such carbon stars. Furthermore, it is quite plausible that some important processing, perhaps strong UV irradiation, not present in the observed layers of IRC +10$\degr$ 216, is needed to fabricate DIB carriers from carbon-rich grain or molecule precursors. The recent observations of possible circumstellar diffuse band carriers in relatively UV-poor F and G-type post-AGB supergiants, as outlined in Sect. 1, will be, if confirmed, a key route for resolving the DIB mystery.