Editing Openai/6897769e-4ee4-800f-aba5-69cca34f701c (section)

=== - In my earlier runs with approximate dispersion and spectral models, the spectrum-weighted means tended to lie around ~139–140° (i.e., a couple degrees above 137.5078°). That’s why you saw 139–140° earlier. Those offsets were larger than the small ~0.47° difference you highlighted. ===
* Small differences like 0.47° are plausible if you use a particular distribution of wavelengths (very selective lines or different refractive-index model), but with realistic Solar spectra + water dispersion the weighted average sits a bit higher than the golden-angle.
* The effect is sensitive to: the precise refractive-index dispersion model (use Hale & Querry or refractiveindex.info to reduce error) and to details of the irradiance weighting (measured solar vs Planck vs line-dominated plasma). Narrow, strong emission lines can “pull” the mean angle but need to be strongly weighted to shift the mean by <1°.