From what has been discussed so far, it seems that the outflows from luminous OBA stars are well understood. There is, however, accumulating evidence that currently accepted mass-loss rates may need to be revised downwards by as much as a factor often, as a consequence of previously neglected wind-'clumping', which affects most mass-loss diagnostics.
Such revisions, of course, would have dramatic consequences, not only for the stellar evolution (Section 1) but also regarding the feedback from massive stars. In the following, we will summarise the status quo, whereas a more detailed discussion can be found in Puls et al. (2006, 2007 and references therein).
The present hypothesis states that clumping (if present) is a matter of small-scale density inhomogeneities in the wind, which redistribute the matter into clumps of enhanced density embedded in a rarefied, almost-void medium. The amount of clumping is conveniently quantified by the so-called clumping factor, fcl > 1, which is a measure of the overdensity inside the clumps (relative to a smooth flow of identical average mass-loss rate). Diagnostics that are linearly dependent on the density (e.g. UV resonance lines) are insensitive to clumping, whilst those sensitive to p2 (such as Ha and free-free radio emission) will tend to overestimate the mass-loss rate of a clumped wind, by a factor +Jf\. For further details, see Abbott et al. (1981), Lamers & Waters (1984), Schmutz (1995) and Puls et al. (2006).
Until now, the most plausible physical process responsible for small-scale structure formation in massive-star winds is the so-called line-driven instability, which was found in the first time-dependent hydrodynamical simulations of such winds (Owocki et al. 1988). Nevertheless, it took some while to incorporate clumping into the atmospheric models of massive stars, firstly for Wolf-Rayet-star atmospheres, in particular to explain the strength of the observed electron-scattering wings of emission lines (Hillier 1991) and the presence and variability of substructures in these lines (e.g. Moffat & Robert 1994).
The diagnostics of OB-star winds, on the other hand, did not require (significant) clumping until recently, particularly because of the very good agreement between the theoretically predicted and observed WLRs (see above). Purely coincidental agreement seemed rather unlikely.
Though there is still hardly any direct observational evidence (but see Eversberg et al. (1998)), to date a number of indirect indications favour the presence of wind-clumping in OB-star winds, so we would like to stress here three important aspects.
(i) From detailed investigations of large samples of Galactic O stars, Puls et al. (2003), Markova et al. (2004) and Repolust et al. (2004) found that supergiants with Ha emission lie above the theoretical WLR (see Section 3), whereas the rest fits almost perfectly. Since the WLR should be independent of luminosity class (e.g. Puls et al. 1996), this discrepancy was interpreted in terms of clumpy winds, with fcl ~ 5, and mass-loss rates reduced by factors between 2 and 3. Indeed, an analogous correction has been applied in the investigation by Mokiem et al. (2007b); see Figure 31.3).
(ii) A compelling, independent indication of clumping comes from analyses of the UV P-Cygni P v 1118/28 resonance line doublet (Massa etal. 2003; Fullerton et al. 2006) observed by FUSE. Because phosphorus has alow cosmic abundance, this doublet never saturates in normal OB stars, providing useful estimates of M when P4+ is the dominant ion - as is implied to be the case at least for mid-O-star winds (Puls et al. 2007). These mass-loss rates turned out to lie considerably below those inferred from other (clumping-sensitive) diagnostics such as Ha and radio emission. The most reasonable way to reconcile these two results is to invoke extreme clumping in the wind (fci ~ 100), with actual mass-loss rates being much lower than previously thought, by a factor >10.
(iii) If clumping were indeed present, there is, of course, the additional question regarding the radial stratification of f cl. To this end, Puls et al. (2006) performed a self-consistent analysis of Ha, IR, millimetre-wave and radio fluxes, thus sampling the lower, intermediate and outer wind in parallel, on the basis of a sample of 19 well-known Galactic O-type supergiants and giants. A major result of this investigation is that in weaker winds the clumping factor is the same in the inner (r < 2R+) and outermost regions. However, for stronger winds, the clumping factor in the inner wind is larger than that in the outer one, by factors of 3-6. This finding indicates that there is a physical difference between the clumping properties of weaker and stronger winds, and is consistent with the arguments outlined in (i) above and earlier findings by Drew (1990).
Unfortunately, the latter analysis is hampered by one severe restriction. Since all employed diagnostics have a p2 dependence, only relative clumping factors could be derived, normalised with respect to the values in the outermost, radio-emitting region. In other words, M(REAL) < M(radio), since the clumping in the radio-emitting region still remains unknown. Only if fcl(radio) were unity would we have M(REAL) = Mi (radio). Thus, the issue of absolute values for M remains unresolved.
Was this article helpful?