We have observed circumstellar K I and Na I emission from three N-type carbon stars and interprete this as resonance scattering of the photospheric light. We have found the dynamics of the envelopes, as seen in the potassium lines, to be reasonably consistent with what is deduced from the CO lines which sample regions further out in the envelopes. The intensities show a rough dependence which is expected for a constant mass loss rate, gas expansion velocity and K I abundance. An interesting and astonishing fact is that the intensities are large enough to suggest that a substantial fraction of the potassium in the envelope is in neutral form. This puts limits on the chromospheric UV emission (shortwards of the K I photoionization threshold at 285.5 nm) for these stars. Since mass loss is a regular phenomenon for N-type stars, with typical mass-loss rates of 10 (Olofsson et al. 1993), an important aspect of this is that the existence or non-existence of strong K I emission may be used as an indicator of the suppression or existence of chromospheric UV emission. Systematic deep surveys of K I emission for a sample of bright N-type stars, and observations of UV fluxes of those with the Hubble Space Telescope, would be rewarding to verify this connection which could then establish a new chromospheric indicator. Also time variations (in particular the "turning off") of K I emission may be worth searching for. Another interesting fact is that the dependence of the envelope emission seems to continue inside the point in the envelope where the column optical depth should be . It seems possible, though not very probable, that this is due to inhomogeneities, clumps, in the envelope, similar in origin to those previously traced in detached CO shells. A more detailed modelling of radiative transfer and hydrodynamics in these media may be worthwhile, and seems necessary before more safe conclusions may be drawn.
One may ask why the objects which show K I emission in our survey all happen to have column optical depths around unity within the angularly resolved envelopes. This may well be a selection effect. Denser envelopes, corresponding to larger , should have smaller surface brightness, and so would considerably thinner envelopes. A deeper search to lower surface brightness for other N-type stars may be rewarding.
We finally note that the mass loss estimates from the K I emission scale as the distance d or to the star, while the CO mass loss estimate is proportional to . Thus, in principle distance estimates - indirect expansion parallaxes - are possible, though rather model dependent. Also, errors in adopted abundances may be significant and, in fact, comparisons of CO emission with K I or Na I emission might be used to estimate abundances of K and Na relative to O. However, improved models are needed before any of these possibilities may be exploited.
© European Southern Observatory (ESO) 1997
Online publication: July 8, 1998