Four separate pieces of work, sharing one fitting code and one line list. The first two ask whether the new grids agree with what came before, on two independent samples analysed by two independent groups. The third asks what the metallicity axis is worth. The fourth puts the grid offset the first two found under a sample where the answer already matters: stars whose seismology depends on it. Every table and figure below is generated from the fits themselves; nothing is transcribed.
428 hot subdwarfs from SDSS-V BOSS, fitted twice: on the older uli H/He grids and on Bamberg v25, same spectra, same lines, same continuum treatment. 74 stars are held by both grids, so the difference between them is the difference between the models and nothing else.
| Bamberg minus uli | median | rms | n |
|---|---|---|---|
| ΔTeff [K] | -49 | 947 | 74 |
| Δlog g [dex] | +0.092 | 0.095 | 74 |
| Δlog n(He)/n(H) [dex] | -0.010 | 0.033 | 57 |
A clean +0.092 dex gravity offset with no temperature offset, and helium unchanged — so a published uli gravity can be carried onto the new scale, while the scatter, 0.095 dex, cannot be. Every one of the 74 stars the uli grids reach is also on Bamberg, which holds 182 in all: the rest are hotter than the uli axis goes.
716 LAMOST hot subdwarfs. The published parameters come from Lei et al. (2018, 2019, 2020) — SYNSPEC 49 spectra from NLTE TLUSTY 204 model atmospheres, fitted through XTgrid — so this is not a grid comparison but the whole difference between two analyses of the same spectra. 647 pairs, 17 excluded as outliers.
| relation | Bamberg as a function of published |
|---|---|
| Teff | new = 1.218 × published − 6945 ±0.014 on the slope, 606 stars |
| log g | new = published +0.0395 offset only — 92% of the axis is real, a slope would be noise |
| log n(He)/n(H) | new = 1.024 × published + 0.0600 ±0.009 on the slope, 547 stars |
The temperatures differ by a scale, not an offset: −3.9% at 26 kK, through zero near 31–32 kK, to +7.7% at 41 kK. The slope, 1.218 ± 0.013, is close to the 1.26 Heber et al. (2026) find against Edelmann et al. (2003) — two different older analyses rescaled almost identically by the same new grid.
No slope is quoted for log g. Binned against our own gravity it looks like a clean gradient; binned against the mean of the pair it is flat at +0.04 dex. Only 92% of that axis is signal, so the apparent gradient is the shared-noise tilt — which is also why Heber et al. report no trend there. Helium is flat to ±0.03 dex except below log n(He)/n(H) −3.5, where it reaches −0.55.
The v25 delivery carries four metal bands, Z = −2, −1, 0, +1 dex about the standard sdB pattern, and everything above uses Z0. The highest signal-to-noise sdOB and He-sdOB spectra in the BOSS archive — 12 of them — were fitted on each band, and then again with metallicity as the third free parameter in place of helium.
| over two dex of metals | sdOB | He-sdOB |
|---|---|---|
| ΔTeff [K] | +1241 | +781 |
| Δlog g [dex] | +0.050 | +0.360 |
| Δlog n(He)/n(H) [dex] | +0.008 | +0.292 |
| disagreement on He II 4686 [% of continuum] | 9.6 | 1.9 |
About ±600 K per dex of Z on the temperature — larger than the formal error on these spectra, and it sits under every fit on this site. In the sdOB stars He II 4686 carries three times the band-to-band disagreement of any other line, which is the mechanism; in the helium-rich stars it does not, and helium and metals trade against each other instead. No star prefers a metal-poor model. Fitting Z in place of helium improves 10 / 12 (total Δχ² -888), with the free Z landing near +0.45.
That last comparison is not quite like for like: three parameters float either way, but in the Z run helium is held, at a value measured at Z = 0 — so it has had four quantities determined from the data where the helium run had three. Settling it means floating all four together. The interpolation for that reproduces the present three-axis code bit for bit, and that fit has now been run on eighteen stars spanning He II 4686 strength. It converges: seventeen determine Z to about 0.2 dex. What it also shows is that the parameter Z is degenerate with is not helium but temperature — see all four parameters free.
The pulsating hot subdwarfs Kepler watched in its original field, refitted uniformly from ground-based spectra so the seismology has one consistent set of temperatures, gravities and helium abundances to sit on rather than one per paper. 24 / 24 targets fitted, from 79 spectra across eight telescopes — nineteen pulsators and five non-pulsating comparison stars, at signal-to-noise up to 1912.
| fitted here minus published | median | range | n |
|---|---|---|---|
| ΔTeff [K] | -237 | -1038 to +1153 | 20 |
| Δlog g [dex] | +0.10 | -0.12 to +0.34 | 20 |
| Δlog n(He)/n(H) [dex] | +0.02 | -0.53 to +0.28 | 20 |
A grid offset, not a disagreement about the stars: on Will and Pippin, refitting the exact spectra Baran et al. (2016) and Kern et al. (2017) used gives +0.10 to +0.12 dex in gravity with no shift in temperature, and across the sample against whatever grid each paper used the offset is median +0.10 dex, positive in 18 of 20 stars — the same one-tenth-dex Bamberg-reads-higher effect as the BOSS and LAMOST samples above. Formal errors understate the truth by about seven times: nine stars observed on three or more telescopes show a median 780 K and 0.102 dex spread between instruments, against formal errors a median 7.2× and 6.7× smaller.
Every target is fitted; what remains is provenance — five stars whose published value rests on a spectrum not yet on disk here, flagged on their own page.
| Every Jname | opens that star's own page: the reduced spectrum in an interactive viewer, the line-profile fit, the fitted parameters with the grids and lines used, and what Gaia and Culpan et al. (2022) say about the star. |
| Figures | drag on either panel of a spectrum to zoom, double-click or Esc to reset. Points in the Kiel diagrams carry the star's name on hover and open its page on click. |
| Flagged fits | keep their row in the tables, with the reason beside them, and are kept out of the figures and every comparison: a fit that stopped against a grid edge has measured the edge. |
| Raw files | each star's directory holds runfit20.dat, the fit output .out, the profiles .plot and the figure .png. |