Simulations of massive star atmospheres and winds during giant eruptive and quiescent luminous blue variable phases

Abstract

Mass loss from massive stars located in the part of the Hertzsprung-Russell diagram (HRD) where we find luminous blue variables (LBVs) is profoundly important for stellar evolution yet poorly understood. We use time-dependent radiation-hydrodynamic (RHD) simulations to examine the atmosphere and wind properties of such massive stars, computing 2D and 1D RHD models of the coupled envelopes, atmospheres, and wind outflows, tuned to this region in the HRD. Our unified simulations start deep in the stellar envelope (well below T ~ 200 kK) and include the outflowing wind, accounting for line-driving, radiative enthalpy, and photon tiring. Mass-loss rates, wind speeds, and the radiative luminosity at the photosphere are emergent properties in the simulations. A grid of models is created by slightly increasing the stellar energy at the lower boundary. This results in a natural transition from very turbulent atmospheres with line-driven winds to effectively stationary super-Eddington massive outflows. Our sub-Eddington models are essentially blue hypergiant stars with very variable surfaces, effective mass-loss rates $\dot{M} \sim 2 - 5 \times 10^{-5}$ $M_{\odot}$/year, and wind speeds $v_{\infty} \sim 200 - 300$ km/s, resembling quiescent LBVs like P Cygni. The super-Eddington models have optically thick wind envelopes and extremely inflated yellow surfaces (Teff ~ 5000 K), $\dot{M} \sim 0.1 - 1$ $M_{\odot}$/year, and $v_{\infty} \sim 400 - 500$ km/s, resembling a massive star during a great eruption like eta Carinae's. Our models naturally reproduce the overall characteristic stellar and wind parameters inferred for massive stars in their quiescent LBV and yellow giant eruptive phases. It remains an open question whether the energy increase needed to trigger a giant eruption can be obtained solely by the internal evolution of the star itself or if it requires an external energy source.

Figures (8)

• Figure 1: HRD showing the track of the MESA-model (black) up to the initial conditions and the resulting positions of the seven models discussed in this paper, with the two discussed in greater detail marked separately. Also shown are the approximate locations of S Doradus-type LBVs (grey), P Cygni, and the $\eta$ Carinae giant eruption, to guide the eye.
• Figure 2: Comparison between the structure up to the photosphere of the 1D atmosphere + wind model and MESA.
• Figure 3: Time series for the coolplate simulation. The top panel shows relative density, the bottom panel shows the radial velocity. The radial direction is displayed using the $x \equiv 1 - R_c/r$ coordinate, which enhances the visibility of the inner regions. This is a clear example of the highly turbulent nature of the simulated atmospheres. The black dashed line on each panel denotes the location of the photosphere.
• Figure 4: Space-time plots of the average density (top panel) and radial velocity (bottom panel) of the coolplate simulation, zoomed in to the lower part of the domain to increase visibility. The location of the photosphere is indicated by the black dashed line.
• Figure 5: Same as Fig. \ref{['Fig:2D-slices_380']}, but for the hotplate simulation, and using density instead of relative density in the top panel. No turbulence is present after the initial relaxation.