From an Existing `.atm` File¶

This is the synthesis-only workflow: you supply a model atmosphere (from any code that can write a Kurucz-format .atm file — ATLAS12, ATLAS9, MARCS, PHOENIX, …), and synthe_py produces the emergent spectrum. The atmosphere is taken as-given; no emulator, no atlas_py iteration.

When this is the right workflow — and what it gives up

With a pre-supplied .atm file the abundance pattern is frozen: whatever the input atmosphere was computed for is what you get. Tweaking abundances after the fact changes only the line opacities; the temperature structure, electron density, and continuum opacity do not respond. That is fine when you trust the input atmosphere for your science case (e.g. you generated it with a dedicated 3-D or NLTE code, or you are reproducing a published grid), but it is not the right tool for asking "what does adding 1 dex of carbon do to this star?". For that, drive the full pipeline through Stellar Parameters, where any per-element abundance offset propagates into both the atmosphere and the spectrum.

When to use this workflow¶

You have a library of pre-computed .atm files and want spectra from them (grid extensions, observational mock catalogues, etc.).
Your target sits outside the emulator's training range (\(T_{\rm eff} < 2500\) K, \(\log g > 5.5\), etc.) and you want to ingest an atmosphere from a code that does cover those regimes.
You need 3-D, NLTE, or otherwise non-LTE-1-D atmospheres that ATLAS12 cannot compute — bring them from the dedicated code, ingest here.
You're benchmarking pykurucz against an external atmosphere code on shared inputs.

Dependency-light

Existing-atmosphere mode needs only the core scientific Python stack (NumPy, SciPy, Numba). PyTorch and the 3.9 GB gfpred29dec2014.bin binary are not required. If you only ever use this mode, run python scripts/download_data.py --synthe-only and skip the rest.

Step 1 — preprocess the atmosphere¶

The .atm text file is converted to a .npz cache that holds the populations and continuous opacities at every depth. This costs a few seconds and can be reused for many different wavelength windows.

python synthe_py/tools/convert_atm_to_npz.py your_model.atm results/your_model.npz

What the converter does:

parses the Kurucz-format .atm (TEFF, GRAVITY, abundance card, READ DECK6 block);
solves Saha–Boltzmann populations for every element and ionisation stage at every depth;
solves molecular equilibrium for ~50 species using the dissociation energies in data/lines/molecules.dat (see Molecular Equilibrium);
computes continuous opacity coefficients (H⁻, H I, He, metal b–f, Rayleigh, Thomson) on the synthesis grid;
writes the lot to a single .npz.

Run once per atmosphere

The .npz is keyed on the atmosphere content. If you edit the .atm (abundances, layer structure, …) you must regenerate the .npz; synthe_py checks an internal version stamp and will refuse to use a stale cache.

Molecular equilibrium needs molecules.dat

The preprocessor reads data/lines/molecules.dat (or lines/molecules.dat in the repo root). It ships in the synthesis bundle of download_data.py, so you usually have it; if you don't, re-run the data downloader.

Step 2 — synthesise¶

With the .npz ready:

python -m synthe_py.cli your_model.atm lines/gfallvac.latest \
    --npz results/your_model.npz \
    --spec results/your_model.spec \
    --wl-start 300 --wl-end 1800 \
    --resolution 300000

Most-used flags¶

Flag	Default	Purpose
`model`	required	path to the `.atm` (provides abundances, structure, header)
`atomic`	required	atomic line catalogue, normally `lines/gfallvac.latest`
`--npz`	auto	the cache from Step 1 (auto-located by sibling path if omitted)
`--spec`	`spectrum.spec`	output `.spec` path
`--wl-start`, `--wl-end`	300, 1800	wavelength range in nm
`--resolution`	300 000	resolving power \(\lambda/\Delta\lambda\)
`--no-molecular-lines`	off	drop molecular line opacity
`--no-tio`, `--no-h2o`	off	drop those specific binary catalogues
`--n-workers`	auto	parallel workers (defaults to all logical CPUs)
`--cutoff`	1e-3	opacity cutoff (fraction of continuum); matches Fortran SYNTHE default

For the full enumeration (NLTE, scattering tolerance, microturbulence, log level, environment variables), see the CLI reference.

Molecules are on by default

If data/molecules/ is populated, molecular catalogues are discovered automatically. Use --no-molecular-lines to disable the lot, or --no-tio / --no-h2o to drop just one binary while keeping the ASCII molecular catalogues.

Convenience wrapper¶

If you want the two steps in one command:

python synthesize_from_atm.py your_model.atm --wl-start 300 --wl-end 1800

This script calls convert_atm_to_npz.py and synthe_py.cli for you with sensible defaults and writes outputs under results/.

Worked example: ATLAS12 → pykurucz¶

Suppose you have a solar atmosphere sun.atm from a legacy ATLAS12 run:

# 1. preprocess once
python synthe_py/tools/convert_atm_to_npz.py sun.atm results/sun.npz

# 2. full optical + NIR at default R
python -m synthe_py.cli sun.atm lines/gfallvac.latest \
    --npz results/sun.npz \
    --spec results/sun_300_1800.spec \
    --wl-start 300 --wl-end 1800

# 3. narrow optical chunk (re-uses the same .npz, no recompute)
python -m synthe_py.cli sun.atm lines/gfallvac.latest \
    --npz results/sun.npz \
    --spec results/sun_500_510.spec \
    --wl-start 500 --wl-end 510

The third call starts synthesising immediately — the .npz cache shortcuts step 1 entirely.

Understanding the log output¶

synthe_py.cli prints timing and progress to stdout. Typical lines:

INFO: Loading atmosphere from NPZ file: results/your_model.npz
INFO: Wavelength grid: 300.00–1800.00 nm, 45012 points, R=300000
INFO: Found 1234567 atomic lines in range
INFO: Synthesis complete: 45012 points in 45.23 s

If you see warnings about missing molecules or zero populations, verify that data/molecules/ is populated and that the .atm abundance card is physically reasonable.

Next steps¶

Compare with Stellar Parameters if your question is about how a non-solar abundance pattern shapes the atmosphere itself — that workflow rebuilds the atmosphere with the abundances you specify, this one does not.
Read Output Files to load .spec, .atm, and .npz from Python.
Skim the CLI reference for advanced flags (--nlte, --scat-iterations, --microturb, …).

From an Existing .atm File¶