Bulk metallicity + alpha pattern
Use a bulk metallicity and shared alpha-element pattern when they adequately describe the star.
metallicity=-0.5, alpha_enhancement=0.2Payne Zero
Stellar spectra from
physical models,
in seconds.
Payne Zero implements one-dimensional, plane-parallel LTE Kurucz atmosphere and spectrum calculations. Atmospheres are iterated on multicore CPUs; spectra are synthesized with PyTorch on CUDA, Apple Metal, or CPU. Both model-generation workflows calculate the emergent spectrum from opacity and radiative transfer.
Atmospheres and spectra
Begin by choosing how to describe the chemical mixture. Payne Zero uses that same description to initialize the atmosphere in either workflow. You can then synthesize while holding the initialized structure fixed, or iterate the atmosphere to convergence before synthesizing. The abundance choice and the atmosphere treatment are separate decisions.
One abundance description for either workflow
Choose the mixture needed by the star
The three descriptions below are accepted by both the fixed-atmosphere and converged-atmosphere calculations.
Independent C, N, and O
Let carbon, nitrogen, and oxygen vary independently for evolved giants and other CNO-sensitive spectra.
c_over_m=-0.25, n_over_m=0.35, o_over_m=0.15Element-by-element
Start from [Fe/H] and specify the elements that differ from it when a grouped abundance pattern is not sufficient.
fe_over_h=-0.4, x_over_h={"C": -0.65, "N": -0.05}Physically converged atmosphere
Iterate the chosen mixture, then calculate its spectrum
Use this workflow when the atmospheric structure is a retained result or when the final spectrum must use an atmosphere that has passed the convergence test. The first command solves the atmosphere; the second synthesizes its structured product. The example uses a bulk metallicity and alpha pattern; the independent CNO and element-by-element inputs above can be supplied in the same place.
Command line
payne-zero-atmosphere \
--effective-temperature 5777 \
--log-surface-gravity 4.44 \
--metallicity 0.0 \
--alpha-enhancement 0.0 \
--microturbulence-km-s 1.0 \
--out runs/sun_atmosphere
payne-zero-synthesis \
runs/sun_atmosphere/payne_zero_structured_atmosphere.npz \
--wl-start-nm 500 \
--wl-end-nm 510 \
--r-grid 20000 \
--out runs/sun_converged_spectrum.npzPython API
from payne_zero_atmosphere import solve_structured_atmosphere
from payne_zero_synthesis import synthesize
solar_labels = dict(
effective_temperature=5777,
log_surface_gravity=4.44,
metallicity=0.0,
alpha_enhancement=0.0,
microturbulence_km_s=1.0,
)
atmosphere_path = solve_structured_atmosphere(
**solar_labels,
out_dir="runs/sun_atmosphere",
)
spectrum = synthesize(
atmosphere_path,
wavelength_start_nm=500,
wavelength_end_nm=510,
resolution=20_000,
device="auto",
)
spectrum.save_npz("runs/sun_converged_spectrum.npz")Fixed initialized atmosphere
Synthesize without the atmosphere iterations
Use this workflow for exploration and repeated evaluations when a separately converged atmosphere is not required. Any of the three abundance descriptions predicts a starting structure; synthesis then holds it fixed while rebuilding populations, opacity, and radiative transfer. The saved product records that the atmosphere was initialized rather than converged.
Command line
payne-zero-synthesis \
--effective-temperature 5777 \
--log-surface-gravity 4.44 \
--metallicity 0.0 \
--alpha-enhancement 0.0 \
--microturbulence-km-s 1.0 \
--wl-start-nm 500 \
--wl-end-nm 510 \
--r-grid 20000 \
--out runs/sun_initialized_spectrum.npzPython API
from payne_zero_synthesis import synthesize_from_labels
spectrum = synthesize_from_labels(
effective_temperature=5777,
log_surface_gravity=4.44,
metallicity=0.0,
alpha_enhancement=0.0,
microturbulence_km_s=1.0,
wavelength_start_nm=500,
wavelength_end_nm=510,
r_grid=20_000,
device="auto",
)
spectrum.save_npz("runs/sun_initialized_spectrum.npz")What the code implements
Opacity and radiative transfer
Accelerator-parallel synthesis makes the physical forward model practical to call directly. Every spectrum is calculated from atomic and molecular opacity, line profiles, and the LTE transfer equation rather than predicted by a label-to-flux emulator.
CPU atmosphere; parallel synthesis
Ordered atmosphere passes use compiled multicore CPU kernels. Synthesis batches independent wavelengths and line profiles on CUDA, Apple Metal, or CPU.
From model wavelengths to observed pixels
Total and continuum flux can be shifted, broadened, passed through an instrument line-spread function, and sampled onto observed pixels before normalization.
Differentiable line parameters
PyTorch differentiates selected oscillator strengths and damping terms through opacity, transfer, broadening, sampling, and the comparison with standard-star spectra.
Kurucz parity and runtime
For four matched stellar controls over 300–1000 nm at R = 300,000, the pooled RMS difference from the Kurucz calculation is below one part in a thousand in normalized flux. This remains true when Payne Zero and the original programs first solve their atmospheres independently. This is a numerical implementation comparison, not a claim of agreement with every observed stellar spectrum.
In the paper's warm repeated-call measurement, the 300–1000 nm solar spectrum took 14 seconds on one NVIDIA H100. Over the narrower APOGEE 1500–1700 nm interval, synthesizing a solar spectrum from the same atmosphere took 1.4 seconds on the H100 at R = 300,000, before instrument convolution or resampling. One physical atmosphere pass took 2.1–5.3 seconds on 16 AMD EPYC threads across the four controls. These timings exclude first-use compilation and depend on wavelength range, R, device, numerical type, and cache state.
Documentation
Installation
Install the complete package, choose a device, and calculate a first spectrum.
Generate spectra
Generate models with fixed or converged atmospheres, choose a mixture, and inspect the resulting products.
Physics
Follow atmospheric initialization, convergence, populations, opacity, transfer, and parallel execution.
Fitting
Compare generated models with observed pixels and retain a physical solution.
Line calibration
Infer selected atomic-data corrections from standard-star spectra.
About
Find the authors, citation, license, source code, and scientific lineage.