Chemical composition and origin of nebulae around luminous blue variables

We use the analysis of the heavy element abundances (C, N, O, S) in circumstellar nebulae around luminous blue variables to infer the evolutionary phase in which the material has been ejected. We concentrate on four aspects. (1) We discuss the different effects that may have changed the gas composition of the nebula since it was ejected : mixing with the swept up gas from the wind-blown bubble, mixing with the gas from the faster wind of the central star, and depletion by CO and dust. (2) We calculate the expected abundance changes at the stellar surface due to envelope convection in the red supergiant phase. We show that this depends strongly on the total amount of mass that was lost prior to the onset of the envelope convection. If the observed LBV nebulae are ejected during the red supergiant phase, the abundances of the LBV nebulae require a significantly smaller amount of mass to be lost than assumed in the evolutionary calculations of Meynet et al. (3) We calculate the changes in the surface composition during the main-sequence phase by rotation-induced mixing. If the nebulae are ejected at the end of the main-sequence phase, the abundances in LBV nebulae are compatible with mixing times between 5 x 10(6) and 1 x 10(7) yr. These values are reasonable, considering the high rotational velocities of main-sequence O-stars. The existence of ON stars supports this scenario. (4) The predicted He/H ratio in the nebulae, derived from the observed N/O ratios, are significantly smaller than the current observed photospheric values of their central stars. This indicates that either (1) the nebula was ejected from a star that had an abundance gradient in its envelope or (2) that fast mixing on a timescale of 10(4) yr must have occurred in the stars immediately after the nebula was ejected. Combining various arguments, we show that the LBV nebulae are ejected during the blue supergiants phase and that the stars have not gone through a red supergiant phase. The chemical enhancements are due to rotation-induced mixing, and the ejection is possibly triggered by near-critical rotation. During the ejection, the outflow was optically thick, which resulted in a large effective radius and a low effective temperature. This explains why the observed properties of the dust around LBVs closely resemble the properties of dust formed around red supergiants.