Running a model

Invocation

The executable takes exactly one argument, the model folder:

$ ./agb_wind <model_folder>

For example, with the reference IRC+10216 setup shipped in the repository:

$ ./agb_wind IRC10216/

The model folder must contain a config.toml (see Configuration (config.toml)). All other input and output paths in that file are interpreted relative to the model folder.

Parallelism is controlled through OpenMP; set OMP_NUM_THREADS to choose the number of threads. The radiative transfer dominates the runtime (~90 %), and is parallelised over spectral points and grid points.

Note

Results are bit-reproducible run-to-run for a fixed thread count. They may differ at the last few digits when the thread count changes, because the parallel floating-point reductions sum in a different order; the converged physical solution is the same.

Model folder layout

A typical model folder (the IRC+10216 example) looks like:

IRC10216/
├── config.toml                 # the model configuration (required)
├── fastchem_parameters.dat     # FastChem element-abundance/parameter file
├── atmosphere_converged.dat    # starting structure (or use "grey")
├── atmosphere.dat              # output: converged structure
├── spectrum.dat                # output: emergent spectrum
├── dust.dat                    # output: dust distribution
└── hydro.dat                   # output: wind structure

External data that is not inside the model folder:

  • the opacity database (HELIOS-K cross sections), at the absolute path given by [opacity] path;

  • the dust refractive indices file, at [dust] refractive_index_file.

Starting model

The [model] starting_model key selects how the initial structure is built:

  • A path (e.g. "atmosphere_converged.dat") — read an existing structure file. This is the fastest route when restarting from a near-converged model.

  • The literal string "grey" — build a logarithmic radial grid and a hydrostatic Lucy grey starting structure on the fly, with no input file. The grid spans [grid] inner_radius … outer_radius (in units of \(R_\star\)) with [grid] n_points cells; see Wind hydrodynamics.

Structure file format

The atmospheric structure file (both the input starting model and the atmosphere.dat output) is a whitespace-separated table with two header lines followed by one row per radial grid point. The columns are:

#

Column

Meaning

1

r/R*

Radius in units of the stellar radius (the normalised radial grid).

2

r(cm)

Radius in cm.

3

rho(g/cm3)

Gas mass density.

4

p(dyn/cm2)

Gas pressure (cgs).

5

T_gas(K)

Gas temperature.

6

T_dust(K)

Dust temperature.

7

v(cm/s)

Wind velocity.

The grid is ordered from the inner boundary (index 0) outward. The first header line of the output additionally records the stellar parameters and the mass-loss rate.

Output files

Each output is written only if the corresponding [output] key is set (a missing key, an empty string, or "none" disables it). Paths are relative to the model folder.

  • [output] atmosphere — the converged structure, same format as the input (Structure file format).

  • [output] spectrum — the emergent spectrum (wavelength vs. flux); see File formats.

  • [output] dust — the dust distribution: degree of condensation, the carbon species number densities, grain number density and radius, nucleation rate \(J_\star\), growth timescale, and the moments \(K_0\dots K_3\).

  • [output] hydro — the wind structure (velocity, density, sound speed, radiative acceleration \(\alpha\), flux-mean extinction).

The outputs are written once, after the global iteration has converged (or after the maximum number of iterations is reached).

Console diagnostics

During a run the code prints, per global iteration:

  • the chemistry–hydrodynamics sub-iteration progress and the maximum change of \(\alpha\);

  • the corrector in use (Unsöld–Lucy or the latched linearisation);

  • the maximum relative temperature change, the flux-convergence deviation from the target luminosity, and the gas/dust energy-balance (radiative-equilibrium) residuals.

These diagnostics are the primary way to monitor convergence; see Numerical methods and robustness for how to read them.