next up previous
Next: Observations Up: Chapter 8: Oxygen-rich Stars Previous: Chapter 8: Oxygen-rich Stars

Introduction to O-rich stars

As discussed in chapter 2, most stars whose main sequence masses are in the range ~1-8Mtex2html_wrap_inline518 evolve into a phase known as the Asymptotic Giant Branch (AGB). These AGB stars fit into three broad categories: M, S and C types, where M-type AGB stars have oxygen-rich atmospheres, C-type stars have carbon rich atmospheres and S types stars are somewhat transitional between the two. It is believed that this is an evolutionary series where the M-type stars evolve into S- and then C-type stars. The chemistry around AGB stars is controlled by the C/O ratio (e.g. Woolf 1973). If the C/O ratio is less than unity, all the carbon will be bound into carbon monoxide (CO), which forms very easily and is very stable. The chemistry will be dominated by the remaining oxygen, leading to the formation of oxygen-rich molecules and particles, e.g. silicates and oxides. This is the case for M-type AGB stars. In this chapter, the 8-13µm spectra of M-type AGB stars will be studied with a view to identifying the oxygen-rich dust species forming around such stars.


In the late sixties Gillett, Low & Stein (1968) investigated deviations of stellar spectra from blackbodies. Their observations led to the discovery of an emission peak near 10µm in four late-type, evolved, variable stars. They proposed two explanations for these observations: 1) the effect was due to a combination of stellar opacity and temperature profile; or, 2) the effect was due to emission by circumstellar matter. This work was followed up by Woolf & Ney (1969), who attributed the emission peak near 10µm (9.7µm) to circumstellar silicate grains around such stars. Since then there has been much interest in the exact nature of the dust around cool evolved stars, how this dust forms and the structure of the dust shells. Hackwell (1972) suggested that the spectra of many M-stars were not consistent with the view that the circumstellar dust was comprised solely of silicate dust. Treffers & Cohen (1974), on the other hand, made high-resolution observations of oxygen rich stars and concurred with Woolf & Ney (1969) on the attribution of the circumstellar dust features to silicates, however they did not preclude the inclusion of other grain types.

The formation of dust grains in the circumstellar shells around oxygen-rich stars was investigated by Salpeter (1974a), who concluded that grain formation proceeded by the nucleation of small refractory seed grains (i.e. oxides) onto which an ``onion-layer'' mantle of the more abundant silicates could form. Despite this investigation into grain formation Salpeter (1974b) was unable to estimate the size of grains forming in these regions.

The emission from circumstellar grains was investigated further by Forrest, Gillett & Stein (1975), who found the ``9.7µm'' feature in many evolved M-type stars. However, the feature is not identical in each case (see also Hackwell 1971;1972 and Treffers & Cohen 1974). The variations in feature shape from star to star could not be explained in terms of optical depth or grain temperature effects, which led to the suggestion that grain size is of importance. They also suggested that the features observed in the spectra of AGB stars could be fully explained in terms of blackbody grains, silicate grains and SiC grains. The modelling of circumstellar dust shells includes the introduction of various arbitrary parameters, and so the influence of various model parameters was investigated by Jones & Merrill (1976). They found that using so-called ``clean'' (i.e. pure magnesium) silicate grains to fit the observed 9.7µm features did not yield a good fit due to the lack of absorption by these grains in the optical and near-IR. They also found that just mixing in more absorbing grains did not solve the problem. This led to the suggestion that the grains responsible for the 9.7µm feature are ``dirty'' silicates, i.e. Mg-silicates with impurities introduced into the matrix giving more opacity in the optical and near-IR. Following the interest in amorphous silicates as the cause of the 9.7µm feature, Day (1979) produced samples of highly disordered magnesium silicates and produced transmission spectra from them. The conclusion drawn from this work was that these silicates (an amorphous forsterite and an amorphous enstatite) were very good candidates for the source of the observed 9.7µm feature, however, distinguishing between the different silicates would be difficult. Forrest, McCarthy & Houck (1979) investigated the physical and chemical composition of the dust grains around cool evolved stars and found that the spectral features were smooth indicating that the grains responsible for the features are unlikely to be well ordered like terrestrial silicates and were more likely to be amorphous.

Following the suggestion of Treffers & Cohen (1974), Papoular & Pégourié (1983) explored the effects of grain size on the shape of the 9.7µm feature. They found that grain radii up to 4µm are needed to explain the observed variations in the shapes of the silicate feature. They then went on to discuss the grain-types expected in circumstellar shells (Pégourié & Papoular 1985). Their condensation model shows that the precise nature of the grains formed is determined by the elemental composition and oxidation properties of the parent atmosphere, along with the density structure of the dust shell. They found that: 1) the concentration of iron in silicates is always expected to be low (mole % Fe2SiO4 (fayalite) ~20%; and FeSiO3 ~10%); 2) Mg2SiO4 (forsterite) forms before MgSiO3(enstatite) in a cooling atmosphere, but the forsterite is converted into enstatite by gas-solid reaction. They did not predict the dust to be pure forsterite or even forsterite with some (<20%) fayalite. The disequilibrium calculations showed that the dust shells of M-type stars should also contain SiO2, solid (metal) Fe, Ca2SiO4 and Al2O3. Papoular & Pégourié (1985) concurred with Forrest, McCarthy & Houck (1979) on the physical nature of the grains, agreeing that they must be amorphous.

Vardya, de Jong & Willems (1986; hereafter VdJW) examined the possibility that, for a given stellar source, the strength of the 9.7µm feature is related to the mass-loss rate or period of variation of the star. In fact there were no obvious correlations, although they did make an interesting discovery regarding the asymmetry factor (see also Vardya 1989). The asymmetry factor f is defined as the ratio of the number of days between minimum light and the next maximum to the period, so that a symmetric light curve has f = 0.5. VdJW found that the silicate emission feature seems to occur only in the spectra of those M-type stars for which f < 0.43. The stars with f > 0.5 show a broader feature centred at about 12µm. They suggested that the 12µm feature is due to more refractory species (e.g calcium or aluminium silicates) that condense at higher temperatures. As the f-value drops the 9.7µm feature appears, due to lower temperature condensates, implying that the change in the f-value is an evolutionary trend and important in the nature of circumstellar dust formation.

Gal et al. (1987) continued the work on the effects of grains size on circumstellar features. They, again, suggested that the variations in the shapes of the silicate features could be best explained in terms of variations in the size, density and temperature of the circumstellar envelope, the size and physical state of the dust grains and the temperature and distance from the central star, rather than in terms of variations in the dust composition. However, as evidence for this scenario, they stated that the 9.7µm feature is practically always located at the same wavelength. This is not true as can be seen from the work of VdJW and our own work (see section 8.3)

Othman et al. (1988) investigated whether there was a correlation between the optical spectral type of stars and the nature of the infrared features. They found that early M-type AGB stars have predominantly featureless spectra, while many late M-type AGB stars do have spectra which exhibit the silicate feature although the correlation is only slight. They also found that simple correlations between silicate emission strength and optical spectral sub-type for both early and late M-type stars are not present. They interpreted this as evidence that the photospheric temperature is not a dominant factor in influencing the emission from oxygen-rich optically thin circumstellar dust shells around M-type AGB stars.

Schutte & Tielens (1989) examined the differences in the shape of the 9.7µm spectral feature from star to star in terms of a model of the circumstellar envelope. Their dust shell model comprised three distinct regions: 1) the regularly pulsating stellar photosphere. Oscillations of the stellar interior propagate shocks into the stellar atmosphere. These shocks transport material to large distances from the stellar surface, where it becomes part of the second distinct region: 2) the (quasi-)stationary layer. The gas temperature is fairly low (~800K) and the dust particles condense in this region; and, 3) an extended, outwardly expanding circumstellar shell or outflow, formed when radiation pressure on the dust particles in the stationary layer accelerates them outwards and drags the gas along with them. Again the smoothness of the silicate features was taken as evidence of amorphous, rather than crystalline, circumstellar grains.

Onaka, de Jong & Willems (1989; hereafter OdJW) attributed the broad 12µm band (see Hackwell 1972 and VdJW) to corundum (Al2O3). They also found that fits to nearly all their M-type star spectra could be improved by the inclusion of Al2O3 grains. They suggested that the only way to form silicate grains is for them to grow on pre-existing grains of Al2O3, which act as seed nuclei.

Little-Marenin & Little (1990; hereafter LML90) have classified the variation in the spectral features from M-type AGB stars into six categories: featureless, broad, 3 component, sil++ (a ``9.7µm'' feature with a strong feature on its long wavelength side centred at about 11.3µm), sil+ (a stronger ``9.7µm'' feature with a weaker long wavelength feature) and sil (a strong ``9.7µm'' silicate feature). They suggest that there is an evolutionary sequence in the spectral features, starting with a featureless continuum and developing a broad feature, followed by a three component feature, a two component feature and then increasingly strong silicate features. They also tried to find correlations between the emission features and the period, mass-loss rate, maser activity and other physical parameters (c.f. VdJW and Othman et al. 1988). The stars with featureless spectra are those with a slightly earlier spectral class than the rest, possibly being less evolved. The most interesting correlation is with the asymmetry factor f. Concurring with the finding of VdJW, LML90 found correlations between the feature variations and the asymmetry of the period of the stars. The mean f varied from 0.47±0.04 for the broad feature to 0.39±0.03 for the strong silicate feature. If the change in spectral feature shape is an evolutionary process, this implies that the visual light curve becomes slightly more asymmetric with age. The work of LML90 was continued by Stencel et al. (1990), who hypothesized some sort of ``chaotic silicate'' condensation. The chaotic silicate forms from a supersaturated vapour containing metal atoms, SiO, AlO and OH in a hydrogen atmosphere. Inside the chaotic silicate, where both silicon and aluminium are less than fully oxidised, the higher reduction potential of Al would initially act to produce AlO at the expense of SiO. Thus the stretching modes of solid, amorphous Al-O would grow at the expense of the 9.7µm Si-O stretch. However, Al is approximately one tenth as abundant as silicon and therefore once the aluminium is completely oxidised, the Si and SiO components of the grain should begin to oxidise and thus increase the strength of the 9.7µm Si-O stretching band. Given the overabundance of silicon relative to aluminium, the silicate feature will eventually overwhelm the 12µm aluminium oxide (corundum) feature. The three component, sil++, and sil+ features identified by LML90 are interpreted as intermediate stages between the AlO-dominated broad feature and the strong 9.7µm silicate feature.

The relationship between the 9.7µm feature and the mass-loss rate was re-examined by Hashimoto et al. (1990), using spherical dust envelope radiative transfer models and the IRAS LRS spectra. They drew several major conclusion from this work: 1) the strength of the 9.7µm silicate feature is an indicator of the mass-loss rate (c.f. Skinner & Whitmore 1988a); 2) the relationship between the 9.7µm silicate feature and the mass-loss is independent of the outer radius of the dust envelope and, therefore, independent of the duration time of the mass-loss; 3) the mass-loss rate has to be greater than about 7 × 10-8Mtex2html_wrap_inline518yr-1 for dust to form; and 4) the characteristic time of steady mass-loss for M-type AGB stars is <~104 years.

The problems of reconciling theoretical dust formation processes with observations of circumstellar dust are described by Tielens (1990). There are two basic factors which determine the species of dust formed: the thermodynamics and kinetics. According to condensation thermodynamics, the silicate condensation sequence starts with the nucleation of corundum (Al2O3) from the circumstellar gas at about 1760K. The first silicate is expected to form by a gas-solid reaction with corundum, to form Ca2Al2SiO7. As the temperature drops, further gas-dust reactions occur so that Mg substitutes for Al to form CaMgSi2O6. The aluminium released and the remaining corundum are converted to spinel (MgAl2O4 ). As further cooling occurs the CaMgSi2O6 and the spinel form a solid-solid reaction, producing anorthite (CaAl2Si2O8). At even lower temperatures (~1440K) forsterite (Mg2SiO4 ) starts to condense out. Forsterite continues to form until the temperature has dropped to ~1350K when it reacts with gaseous SiO to produce enstatite (MgSiO3). Finally, at ~1100K reactions with gaseous iron will convert some enstatite into fayalite (Fe2SiO4) and forsterite. Kinetics also plays an important role in determining which silicates form in the outflows of AGB stars. Depending on the density structure of the region circumjacent to the star the condensation sequence will be brought to a halt at different points. Thus, if the density drops rapidly with distance from the star, the only dust expected to form will be various high temperature oxides (e.g. Al2O3, CaTiO3, ZrO2), which will form very close to the photosphere. If the densities are a little higher further out in the circumstellar shell gas-grain reactions can take place, allowing the formation of calcium-aluminium silicates. If the density is high enough a little further out magnesium silicates may form as rims on the Ca-Al silicates. For magnesium silicates to nucleate, there need to be very high densities a long way out, which is highly unlikely. Feldspars are not expected to form, as the solid-solid reaction requires unrealistically high densities. Finally, Fe can only be incorporated into Mg-silicates if, initially, most of the iron is in gaseous form (rather than solid, metal form) and if the density is high enough at large distances from the star where fayalite can survive. Most laboratory studies of condensation sequences concentrate on very small sub-systems, such as silicate formation in magnesium-iron rich gases. These are obviously not realistic.

The most striking assertion from Tielens (1990) was that observations indicate the presence of crystalline grains. This conflicted with the consensus that the smooth 10µm features are indicative of amorphous, rather than crystalline, grains. The 11.3µm feature attributed to crystalline olivine was taken as evidence of crystallinity. He suggested that the magnesium silicates forming close to the stellar photosphere are crystalline, but the iron silicates are quite amorphous, which may go some way to explaining the variations in the feature.

Waters et al. (1996) found features in the ISO-SWS spectra evolved stellar objects (AGB stars, red super giants, post-AGB objects and planetary nebulae) which they have attributed to crystalline silicates. While they acknowledged that some of the features may be attributable to other dust species (e.g. water ice, oxides), they asserted that some of the features can only be explained by using the optical properties of crystalline silicates. These features are mostly longwards of 20µm, where amorphous silicates do not show prominent features. They suggested that a combination of different dust species, both crystalline and amorphous, are needed to explain all the features in the infrared spectra of evolved stars. Unfortunately most of the interesting crystalline silicate features they referred to are found outside the range of our observational spectra.

As discussed earlier, the work of LML90 divided M-type AGB star spectra into sixagroups, the sequence of which was postulated to be evolutionary. However, LML90 seem to have misclassified some of the IRAS LRS spectra (e.g. by seeing broad features where none exist) and their choice of specific classifications seems questionable. The initial and final groups of LML90 are very similar to those set out here, where the spectra exhibit either no feature or a very strong silicate feature, however the intermediate groups are much more ambiguous when it comes to separating them since they tend to merge into one another (as would be expected for an evolutionary sequence) and precise interpretation of these spectra is difficult. We have, therefore, classified 80 CGS3 10µm spectra of oxygen-rich dust shell stars into six groups. These are: featureless, broad, transition, broad+sil, sil+broad, and sil. The basic features of these groups are shown in Table 8.1. Like LML90, we see there is a possible evolutionary progression in the spectra, starting with a featureless continuum, building up a broad low feature (broad) which develops a slight 9.7µm feature (broad+sil). This 9.7µm feature becomes stronger (sil+broad) and eventually dominates the mid-IR spectrum (sil). The transition group is the stage between a dominant broad feature and a dominant silicate feature. We have then compared our classified spectra to the relevant laboratory spectra discussed in chapter 5 and the results are discussed below.

    Table 8.1: A classification system for M-type AGB star spectra
Classification Description of the spectrum
featureless no features above a blackbody energy distribution
broad broad, low, fairly smooth feature over the 8.5-12.5µm region
transition the broad with a slight 9.7µm silicate bump
broad+sil stronger silicate feature, but still a strong broad component
sil+broad silicate feature is stronger still; the broad component
starts the resemble a wing of silicate feature
sil The 9.7µm silicate feature dominates;
the broad component is no longer visible

a In fact LML90 used seven categories, however their S-feature category is viewed as separate from the suggested evolutionary track of the dust and is therefore ignored here.

Next: Observations Up: Chapter8: Oxygen-rich stars Previous: Introduction to O-rich stars

Wicked Witch's Webpage
Angela's Home Page
email me
Angela's Research page
Angela's Research Page