Evolution of thermally pulsing asymptotic giant branch stars - II. Dust production at varying metallicity

We present the dust ejecta of the new stellar models for the thermally pulsing asymptotic giant branch (TP-AGB) phase computed with the COLIBRI code. We use a formalism of dust growth coupled with a stationary wind for both M-and C-stars. In the original version of this formalism, the most efficient destruction process of silicate dust in M-giants is chemisputtering by H-2 molecules. For these stars, we find that dust grains can only form at relatively large radial distances (r similar to 5R(*)), where they cannot be efficiently accelerated, in agreement with other investigations. In the light of recent laboratory results, we also consider the alternative case that the condensation temperature of silicates is determined only by the competition between growth and free evaporation processes (i.e. no chemisputtering). With this latter approach we obtain dust condensation temperatures that are significantly higher (up to T-cond similar to 1400 K) than those found when chemisputtering is included (T-cond similar to 900 K), and in better agreement with condensation experiments. As a consequence, silicate grains can remain stable in inner regions of the circumstellar envelopes (r similar to 2 R-*), where they can rapidly grow and can be efficiently accelerated. With this modification, our models nicely reproduce the observed trend between terminal velocities and mass-loss rates of Galactic M-giants. For C-stars the formalism is based on the homogeneous growth scheme where the key role is played by the carbon over oxygen excess. The models reproduce fairly well the terminal velocities of Galactic stars and there is no need to invoke changes in the standard assumptions. At decreasing metallicity the carbon excess becomes more pronounced and the efficiency of dust formation increases. This trend could be in tension with recent observational evidence in favour of a decreasing efficiency, at decreasing metallicity. If confirmed by more observational data, it would indicate that either the amount of the carbon excess, determined by the complex interplay between mass-loss, third dredge-up and hot-bottom burning, or the homogeneous growth scheme should be revised. Finally, we analyse the differences in the total dust production of M-stars that arise from the use of the two approaches (i.e. with or without chemisputtering). We find that, in spite of the differences in the expected dust stratification, for a given set of TP-AGB models, the ejecta are only weakly sensitive to the specific assumption. This work also shows that the properties of TP-AGB circumstellar envelopes are important diagnostic tools that may be profitably added to the traditional calibrators for setting further constraints on this complex phase of stellar evolution.