2025A&A...704A.186G

We present an analysis of the results of 21-cm H I line observations of the circumstellar envelopes (CSEs) of a sample of 151 evolved stars, consisting predominantly (85%) of asymptotic giant branch (AGB) stars. This is the first time an analysis could be carried out for the neutral hydrogen constituent of a substantial sample of CSEs of AGB stars. We obtained our observations mainly with the Nançay Radio Telescope (NRT), resulting in 34 clear detections and 21 possible detections. Among the 106 AGB type stars with non-confused H I spectra, 75% are O-rich and 22% are C-rich, while 41% are SRb type semi-regular variables and 38% are Miras. We found no significant biases in the selection or observations of different types of AGB stars. The total H I masses of the detected AGB stars range from 0.002 to 0.1 M_⊙, with a mean value of 0.02 M_⊙. The mean total H I masses are not significantly different for stars of different types of variability (Miras and semi-regulars). However, there is a difference between O- and C-rich AGB stars, which is due to only three C-rich stars with exceptionally high H I masses (>0.1 M_⊙). If we disregard them, there is no significant difference among these types. We compared the total masses of atomic and molecular hydrogen in 34 AGB star CSEs, with the latter estimated from far-infrared imaging of dust, which extends out to about the same radii as the H I. We found that, on average, the H₂ masses are ∼20 times larger than the H I masses. However, in eight objects, the hydrogen in the CSE is essentially completely atomic. We examined the possible dependence of our results, in particular the H₂:H I total mass ratio, on the effective temperature (T_eff) of the central star. We find that the H I detection rate of CSEs tends to increase steadily with T_eff, but we find no obvious correlation between the H₂:H I mass ratio and T_eff over the range ∼2100-3300 K. Here, we discuss this result in the context of the theoretical prediction that the hydrogen in their CSEs should be mainly atomic for AGB stars warmer than about 2500 K, and mainly molecular for cooler stars. However, the limited fraction in our sample of stars with well-determined temperatures lying below 2500 K prevented us from definitively confirming or refuting the predictions of this model. We discuss a number of effects that might explain the predominantly molecular nature of CSEs, irrespective of stellar temperature. Advancing their interpretation would require further development of mass outflow models for AGB stars of different effective temperatures, as well as comprehensive sets of T_eff measurements of this highly time-variable class of stars. We also compared the H I and CO(1-0) line emission of AGB CSEs. The latter emission originates from much smaller radii (<0.01 pc) than the H I (0.75 pc for the resolved sources), and no H₂ masses can be determined from it. There is a large spread in the CO:H I integrated line flux ratio (by more than a factor of 100). We found that CO:H I flux ratios generally increase with the H₂:H I mass ratio.