New self-consistent theoretical descriptions for mass-loss rates of O-type stars

TL;DR

This work presents self-consistent, NLTE-aware wind models for O-type stars by coupling hydrodynamics with detailed line-driving through the force multiplier. Using three atomic configurations (H, HHe, CNO) and NLTE TLUSTY fluxes fed into LOCUS, the authors iteratively solve for wind structure with HYDWIND to obtain robust $\dot{M}$ and wind velocities. Bayesian linear regressions provide analytic $\dot{M}$ prescriptions as functions of $T_{\rm eff}$, $\log g$, and $R_*$ for each grid, while wind momentum–luminosity relationships are calibrated and show good agreement with observations, particularly for the more complex CNO grid. The results reveal a systematic decrease in $\dot{M}$ as the radiation field becomes increasingly line-rich, highlighting the importance of accurately modeling the UV flux and line opacity for reliable wind predictions. Limitations include the solar metallicity assumption and exclusion of iron-group elements; future work will extend the grids to varying metallicities and iron-group opacities to further refine mass-loss estimates and their impact on massive-star evolution.

Abstract

Massive O-type stars lose a significant fraction of their mass through radiation-driven winds, a process that critically shapes their evolution and feedback into the interstellar medium. Accurate predictions of mass-loss rates are essential for models of stellar structure and population synthesis. We computed wind parameters for O-type stars using a self-consistent approach that couples the hydrodynamics of the wind with detailed calculations of the line acceleration. This approach follows the theory of radiation-driven stellar winds and allows us to derive mass-loss rate distributions for different atomic configurations of the stellar flux. We used the TLUSTY code for stellar atmosphere models to compute non-local thermodynamic equilibrium models; these models served as input radiation fields for the calculation of the line-force parameters, for which we used the LOCUS code. These line-force parameters were then iteratively coupled with the HYDWIND code to solve the wind hydrodynamics. The procedure was applied across a grid of stellar parameters for three chemical configurations. We obtain self-consistent wind parameters for a broad set of O-type stellar models. The results show a systematic decrease in mass-loss rates with the inclusion of more elements in the radiation field, which is attributed to a strong effect on the UV region of the spectral energy distribution. As more elements are included, resulting in a larger number of spectral lines, the contribution from the UV diminishes, leading to lower mass-loss rates. We fitted three theoretical prescriptions for $\dot{M}$ using a Bayesian approach; this yielded Pearson correlation values greater than 0.92 for all three model grids. It also allowed for the estimation of the wind momentum-luminosity relationships for each of the grids, yielding results similar to those based on observations of O-type stars.

Figures (8)

• Figure 1: Comparison of the wind temperature profiles, $T(r)$, commonly used in the atomic calculations of the force multiplier factor (Eq. \ref{['eq:forcemultiplier']}). As can be seen, it is important to approximate the temperature profile throughout the wind, given its variation along the radial coordinate, rather than relying on the two fixed values typically adopted in the isothermal treatment.
• Figure 2: $\dot{M}$ kernel density estimation distribution for each grid. The median values of each distribution are indicated by vertical dashed lines. A clear trend emerges, showing lower mass-loss rates for grids with more complex radiation fields. This indicates that simpler radiative configurations yield systematically higher mass-loss rates, while the inclusion of more atomic transitions results in predictions consistent with the low mass-loss regime.
• Figure 3: Emergent fluxes computed with Tlusty for a representative stellar model ($T_\mathrm{eff} = 38000 \ \mathrm{K}$ - $\log g = 3.4$). Each curve corresponds to a different radiative configuration, illustrating the effect of increasing atomic complexity. As can be seen, the effect, especially in the UV region, is considerable and will translate into an effect on the value of the mass-loss rate.
• Figure 4: Comparison of mass-loss rates obtained from the adjusted theoretical description of the most complex grid in terms of atomic structure (CNO; solid green line) with results from other studies. As shown, the models derived in this work exhibit a better description of low mass-loss rates, particularly in regions where $\log \dot{M}$ values fall below predictions of previous studies, reaching values as low as $10^{-8}$ [$M_\odot$ / yr].
• Figure :