The star is surrounded by a nebula of ejected material at 0.4–1.2pc from the star. The nebula contains around, all lost from the star around 10,000 years ago. There is an 8.8pc wide empty cavity in the interstellar medium around the star, presumably cleared by fast winds earlier in the star's life. AG Carinae is apparently in a transitional phase between a massive class Oblue supergiant and a Wolf–Rayet star, where it is highly unstable and suffers from erratic pulsations, occasional larger outbursts, and rare massive eruptions. The spectral type varies between WN11 at visual minimum and an early A hypergiant at maximum. At visual minimum the star is about and 20,000–24,000K, while at maximum it is and 8,000K. The temperature varies at different minima. One study calculated that the bolometric luminosity of AG Carinae decreases during the S Doradus outbursts, unlike most LBVs which remain at approximately constant luminosity. The luminosity drops from around at visual minimum to around at visual maximum, possibly due to the energy required to expand a considerable fraction of the star. Evolutionary models of the star suggest that it had a low rotation rate for much of its life, but current observations show fairly rapid rotation. Models of LBV progenitors of type IIb supernovae list AG Carinae as matching the final stellar spectrum prior to core collapse, although the models are for stars with 20 to 25 times the mass of the Sun while AG Carinae is thought to be considerably larger. The initial mass of the star would have been around and is now thought to be.
Distance controversy
es from data release 1 of the Gaia mission suggest a much closer distance to AG Carinae and its neighbour Hen 3-519 than previously accepted, around 2 kpc. Then both stars would be less luminous than LBVs and it is argued that they would be former red supergiants whose unusual characteristics are the result of binary evolution. The earlier Hipparcos parallax for AG Carinae had a margin of error larger than the parallax itself and so gave little information about its distance. The distance of 6,000 pc is based assumptions about the properties of LBVs, models of interstellar extinction, and kinematical measurements. The Gaia DR1 parallax, derived from the combination of the first year of Gaia measurements with Tycho astrometry, is 0.40 ± 0.22 mas. The Gaia team recommend that a further 0.3 mas systematic error is allowed for. Smith et al. argue that the 0.3 mas systematic margin of error can be ignored and that the implied distance to AG Carinae is 2.50 ± 1.41 kpc. In Gaia Data Release 2, the parallax is, suggesting a distance around.
