1997A&A...322..280H

We have analysed the IRAS-LRS spectra of representative samples of O-rich Semiregular (SR) variables of types SRa and SRb and of Mira variables. The silicate features were extracted by fitting the energy distribution with two blackbodies, approximating the continuous emission from the photosphere and the circumstellar dust. The shape and strength of the silicate features in the LRS range were then studied by computing the residual fluxes in 5 selected wavelength regions covering the whole 10µm and 18µm features and parts of the 10µm feature assigned to emission from olivine and possibly corundum. We compare our approach with previous investigations and argue that a quantitative study of details in the feature shape requires subtraction of the stellar and the dust continuum and the use of flux ratios rather than a discrete classification system. The Miras form an extension of the SRb's towards lower stellar temperatures and higher dust shell opacities and they have slightly higher average dust temperatures. The SRa's seem to be more similar to the Miras in their dust shell properties. The average 10µm feature shapes of the three groups of variables agree, but taking into account the photospheric and dust shell parameters, systematic differences show up. For stars hotter than about 2900K, the 10µm feature width shows a wide range of values but no clear trend with the stellar temperature or the optical depth of the dust shell. These stars are generally SRb variables and have the thinnest dust envelopes. At cooler stellar temperatures, where mostly Miras are found, the optical depth of the dust shell determines the feature width in the sense that thicker shells have narrower features. It appears that the 13µm feature is obvious only in a narrow range of effective temperature and optical depth of the dust shell. We discuss our results in terms of radiative transfer effects, differences in the average grain size, annealing and hydration of amorphous silicates and contributions from other dust components. Of these possibilities the last one seems to be most plausible with regard to the behavior of the 10µm feature width. The observations can be interpreted in terms of changing contributions from olivine and corundum possibly caused by an increasing amount of dust processing (Miras) and the influence of the atmospheric structure on the formation of these dust components (SRb's).