4 A model for the carrier of the GC 3.4-$\mu$m feature

Let us now compare these bands with those which were observed towards the GC (see, e.g., Pendleton et al. 1994; Tielens et al. 1996; Chiar et al. 2000).

Depending on the IRS number (background star observed), the peak of the OH band near 3 $\mu$m shifts from $\sim$3500 to $\sim$3300 cm^-1 and has a very long red tail. Water-ice coatings are certainly deposited on carbon dust in cold environments, and are responsible for most of the observed 3.1-$\mu$m peak. However they can hardly account for the lingering red wing, unless they are much thicker than 0.1 $\mu$m (see Leger et al. 1983; Gibb et al. 2000). This has led Tielens et al. (1996) to wonder if the red tail was not due to an independent dust component, and Wada et al. (1991) to suggest H₂O trapped in SiO. Now, kerogen, too, contains such water of hydration which is responsible for its 3-$\mu$m band (see Sect. 3.1) and the latter is quite similar to the GC feature (see Figs. 2 and 3). Thus kerogen may partially contribute in this range, too. Among the other bands, the 3.4-$\mu$m band is prominent like in young kerogens (high O/C and H/C ratios), and its location and profile are pretty much the same in both cases (see Figs. 5a and 6a below). Also, the residual aromatic stretch near 3.3 $\mu$m is very weak in both.

With their highest resolution instrument pointed towards Sgr A (IRS 3), Tielens et al. (1996) detected and assigned bands at 5.5, 5.8, 6.1, 6.8 and predicted one near 7.3 $\mu$m. The 5.5-$\mu$m band is weak and was not confirmed by later observations (Schutte et al. 1998). The latter four bands are clearly present in young kerogens and have been given the same assignments. The aliphatic band at 6.8 $\mu$m has $\tau=0.16$ and $\Delta \nu=20$ cm^-1, whose product is 3.2 cm^-1 (Tielens et al. 1996). From Pendleton et al. (1994, Table 2 and Fig. 4), the 3.4-$\mu$m band has $\tau=0.23$ and $\Delta \nu=120$ cm^-1, the product of which is 28 cm^-1. The ratio of integrated intensities is $\sim$9, nearly the same as for kerogens (see above).

In young kerogens of a given type, the H content is relatively constant, so the shape and strength of the 3.4-$\mu$m blend do not depend very much on the evolutionary stage. By contrast, the relative intensities of the mid-IR are quite sensitive to evolution because of early loss of O atoms. This is evident, for instance, from Durand (1980, Chap. 6) and Espitalie et al. (1973). This indicator can therefore be used to choose a kerogen model carrier for a given astronomical spectrum. A visual inspection of the spectra in Espitalie et al., reproduced in our Fig. 3, suggests that the spectrum of sample c (type II, depth 2000 m) comes closest to Tielens et al.'s Fig. 2 (1996). Note also in sample c the presence of an underlying plateau extending from 5 to 10 $\mu$m, which is welcome in view of the GC extinction curve obtained by Lutz (1999, Fig. 5), from the ISO satellite observations: the kerogen plateau is apt to fill the well-known dip exhibited by the silicate dust absorption in the same spectral range.

Sample c has O/C=0.067 and H/C=1.24; it is, therefore, closest to model IIb in the array of models provided by Behar & Vandenbroucke (1986; see Sect. 2 above). Our Fig. 4 reproduces the corresponding chemical structure given by these authors. From their Tables 2 and 3, one can also deduce quantitative structural data such as:
- the relative concentration of aliphatic, naphtenic (in linear chains of benzenic rings) and aromatic carbon sites are, respectively, $\sim$45, 14 and 41$\%$;
- the average size of aromatic clusters is $\sim$3 benzenic rings;
- the largest fractions of O atoms are engaged in ester (-O-C=O) and ether (C-O-C) groups; about 10 percent are in hydroxyl (-OH) groups and 10 percent in 6-membered rings and quinones;
- aliphatic chains (-CH₂-)_n) are no longer than $n\sim10$;
- according to Fig. 8 of Robin et al. (1975), H(methyl)/H(aliph) $\sim 0.2{-}0.4$.

Small molecules, such as H₂O are trapped in the array by steric constraints; through H-bonding, they contribute to the inhomogeneous broadening of IR bands.

Finally, the integrated absorbance at 3.4 $\mu$m for the model chosen above is K=60 cm/mg (Robin et al. 1975) and, for a bandwidth of 120 cm^-1, the corresponding cross-section is $2.9\times 10^{-20}$ cm² per C atom.