A STUDY OF CIRCUMSTELLAR ENVELOPES AROUND BRIGHT CARBON STARS .1. STRUCTURE, KINEMATICS, AND MASS-LOSS RATE

We have performed a survey of circumstellar CO emission on a sample of bright carbon stars, which is relatively complete out to about 900 pc from the Sun. In total, 68 detections were made. All objects within 6 00 pc of the Sun were detected. The result suggests that the large maj ority of all carbon stars have circumstellar envelopes. The CO-emittin g parts of these envelopes appear to have angular sizes less than abou t 15'. The median gas expansion velocity is 12.5 km s-1, and the expan sion velocities for the majority of the objects fall in the range 9-15 km s-1. We find no significant differences between the expansion velo cities estimated from the CO (1-0) and CO (2-1) lines. The expansion v elocities show a tendency to be higher for stars that lie close to the Galactic plane. Relatively accurate stellar radial velocities have al so been obtained. The mass-loss rate estimated from the CO data correl ates well with an infrared excess estimate. This has been used to obta in mass-loss rates for the entire sample. The median mass-loss rate is 1.5 x 10(-7) M. yr-1, and the mass-loss rate for the majority of the stars lies within the narrow range (0.8-2.5 ) X 10(-7) M. yr-1. The ma ss-loss rates of the Mira variables are on the average a factor of 10 higher than those of the Lb, SRa, and SRb variables, which are, on the other hand, very similar. A corollary to this is that the mass-loss r ate increases with period and variability amplitude. We find a depende nce of the mass-loss rate on the stellar effective temperature, but no apparent dependence on the photospheric carbon excess (with respect t o oxygen) and C-12/C-13 ratio in spite of the fact that the stars span quite a range in these properties. The gas expansion velocity shows a dependence on the stellar effective temperature, and possibly also on the carbon excess. The mass-loss rate and the gas expansion velocity are weakly correlated with each other. This trend could be a luminosit y effect due to radiation pressure on grains. The substantial scatter in the data suggests a mass-loss mechanism that can produce widely dif ferent mass-loss rates with only marginal effects on the expansion vel ocity. We present results that imply that the simple estimate of the m ass-loss rate from the strength of the CO emission used in this paper systematically underestimates the mass-loss rate. The underestimate be comes larger the lower the mass-loss rate, and it amounts to about a f actor of 2-4 for the objects with the lowest mass-loss rates (< 10(-7) M. yr-1). Accurate estimates of the mass-loss rate using the circumst ellar CO emission may in fact be quite difficult to obtain before a si gnificantly better physical understanding of the envelopes is at hand. The sample contains three stars-U Ant, S Sct, and TT Cyg-where there i s convincing evidence for highly episodic mass loss, possibly caused b y a thermal pulse. The scarcity of such objects in our sample, and the assumption that all carbon stars go through this phase at least once, suggest carbon star lifetimes of about 2 X 10(5) yr or more and a car bon star birthrate consistent with other data.