Multi-dimensional, time-dependent approximate NLTE unified model atmospheres with winds for hot, massive stars

1. Problem Statement

Previous multi-dimensional (multi-D), time-dependent radiation-hydrodynamics (RHD) simulations of massive star atmospheres and winds relied on a significant simplification: they assumed the flux-mean ($\kappa_F$), energy-mean ($\kappa_E$), and Planck-mean ($\kappa_B$) opacities were equal.

• The Flaw: This assumption effectively forces Local Thermodynamic Equilibrium (LTE) regarding the thermal coupling between gas and radiation. While valid in deep, dense atmospheric layers, it breaks down in the extended, outflowing wind where densities are low, collisions are inefficient, and the radiation field is diluted.
• The Consequence: In these outer regions, scattering becomes dominant. By equating $\kappa_E$ and $\kappa_B$ to $\kappa_F$, previous models failed to account for the fact that scattering does not contribute to heating or cooling. This forced the gas temperature ($T_{gas}$) to remain coupled to the radiation temperature ($T_{rad}$), preventing the formation of realistic shock-heated regions where $T_{gas} \gg T_{rad}$.

2. Methodology

The authors developed and implemented an approximate Non-Local Thermodynamic Equilibrium (aNLTE) formalism to compute $\kappa_E$ and $\kappa_B$ separately from $\kappa_F$, explicitly accounting for scattering in an accelerating medium.

• Atomic Database: The calculations utilize the "Munich Atomic Database," containing $\sim$4 million spectral lines for elements from Hydrogen to Zinc (up to the 8th ionization stage).
• Formalism:
  • Level Populations: Ionization and excitation balances are calculated using modified Saha and Boltzmann equations that account for the diluted radiation field (using a modified dilution factor $\tilde{W}$).
  • Sobolev Approximation: Line heating and cooling rates are derived using Sobolev escape probabilities ($\beta$) and effective thermalization parameters ($\chi_l$).
  • Separation of Opacities: The net heating/cooling rate ($\dot{q}$) is expressed as $c\rho\kappa_E E - 4\pi\rho\kappa_B B$. Crucially, the scattering component is removed from $\kappa_E$ and $\kappa_B$, ensuring that only true absorption/emission processes drive thermal exchange.
• Implementation Strategy:
  • Direct calculation of opacities for every grid point in a multi-D simulation is computationally prohibitive.
  • The authors pre-computed opacity multipliers ($M_E, M_B$) on a 4D grid (density, gas temperature, radiation temperature, and dilution factor).
  • These multipliers are fitted using an analytical function based on Gayley (1995) and Poniatowski et al. (2022), allowing for fast interpolation during RHD simulations.
• Simulation Setup: The method was implemented in the MPI-AMRVAC code for 2D RHD simulations of O-type stars, resolving the sub-surface iron-opacity bump, the photosphere, and the supersonic wind.

3. Key Contributions

• First Multi-D aNLTE Wind Simulations: This is the first study to apply a physically motivated aNLTE treatment of heating and cooling (separating $\kappa_E, \kappa_B$ from $\kappa_F$) in time-dependent, multi-D unified atmosphere and wind models.
• Decoupling of Temperatures: The formalism successfully decouples $T_{gas}$ and $T_{rad}$ in the wind, allowing for the natural emergence of shock-heated gas without artificial constraints.
• Quantification of Shock Heating: The study quantifies the volume filling factor of gas heated above the local radiation temperature and the stellar effective temperature.

4. Key Results

• Thermal Decoupling: In the outer wind (beyond $3R_*$), the gas temperature is no longer identical to the radiation temperature.
  • For a typical O-type supergiant, ~35% of the gas volume in the outer wind has temperatures exceeding the local $T_{rad}$.
  • ~4% of the gas is heated to temperatures exceeding the stellar effective temperature ($T_{eff}$).
• Shock Structure:
  • Structures originate from the iron-opacity bump ($\sim$150 kK) in sub-surface layers and propagate outward.
  • High-density clumps move slower than the surrounding rarefied medium due to the inverse density dependence of line-driving forces.
  • This velocity dispersion creates localized shock fronts. At these fronts, compressional heating raises $T_{gas}$ to $\sim 10^5$ K.
• Radiative Cooling vs. Advection:
  • The radiative cooling timescale ($t_{rad}$) is significantly shorter than the advection timescale ($t_{adv}$) in the dense post-shock layers.
  • Consequently, the heated gas cools rapidly and remains confined to thin, localized layers behind the shock front rather than forming an extended hot component.
• Comparison with LDI Models:
  • Unlike Line-Deshadowing Instability (LDI) models which often predict very high shock temperatures (leading to X-rays), the shocks in this study (driven by sub-surface turbulence) are strong enough to heat gas but not necessarily to the $\sim 1$ MK required for X-ray emission.
  • A comparison with a lower-luminosity O-type dwarf showed significantly less structure (only ~0.8% of gas above $T_{eff}$), highlighting the role of proximity to the Eddington limit in driving sub-surface turbulence.

5. Significance and Implications

• Physical Realism: The study demonstrates that multi-component wind structures exist not only in density and velocity but also in temperature. This is a critical advancement over previous models that assumed isothermal winds in the outflow.
• Spectral Interpretation: The presence of shock-heated gas provides a physically motivated origin for high-ionization UV lines (e.g., N V, O VI) and optical features (e.g., He II 4686 Å) observed in O-type stars. It suggests that these features arise naturally from wind shocks without needing ad hoc "hot gas" components in spectral fitting.
• Future Directions:
  • The current models do not yet reach the temperatures required for X-ray emission ($\sim 10^7$ K), likely due to the absence of LDI (which operates on smaller scales than the current resolution) or numerical diffusion.
  • Future work will focus on incorporating frequency-dependent radiation fields (bound-free processes) and improving resolution to capture LDI-driven shocks.
  • The authors are developing machine-learning approaches to replace current fitting functions for opacity multipliers to improve accuracy across varying optical depths.

In conclusion, this work establishes a new standard for modeling massive star winds by rigorously treating the non-LTE thermal balance, revealing that shock heating creates a complex, multi-temperature wind structure that fundamentally alters our understanding of stellar mass loss and spectral formation.