Radial abundance gradients of 18 elements in Galactic open clusters from infrared MWM spectra A detailed analysis of 655 giants in 133 clusters

H. J\"onsson; J. M. Otto; K. Cunha; P. Frinchaboy; S. Bijavara Seshashayana; V. D'Orazi

REVIEW 2 major objections 4 minor 2 references

Radial abundance gradients of 18 elements in Galactic open clusters from infrared MWM spectra A detailed analysis of 655 giants in 133 clusters

T0 review · 2 major / 4 minor · reviewed 2026-08-04 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A reanalysis of infrared spectra of 655 giant stars in 133 open clusters establishes that the Milky Way's radial iron gradient is broken—steep inside roughly 10.6 kpc and flatter beyond—and revises the open-cluster abundance scale.

desk verdict Solid, honest reanalysis of DR19 open-cluster spectra; the global gradient matches Gaia-ESO, but the claimed 10.6 kpc break rests on formal errors and distance-systematics the paper never quantifies. read the letter →

arxiv 2608.02563 v1 pith:F3YXQ6RC submitted 2026-08-03 astro-ph.GA astro-ph.SR

S. Bijavara Seshashayana , H. Jönsson , V. D'Orazi , K. Cunha , P. Frinchaboy , J. M. Otto This is my paper · ORCID

classification astro-ph.GAastro-ph.SR

keywords openclustersMilkyWaydiskradialmetallicitygradientstellarabundancesinfraredspectroscopygiantstarsGalacticchemicalevolutionNLTEanalysis

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish a more reliable map of how the Milky Way's chemistry changes with galactocentric radius, by re-analyzing infrared spectra of 655 giant stars in 133 open clusters with stricter membership control, photometrically anchored log g values, non-local thermodynamic equilibrium corrections, and visual inspection of every fitted spectrum. Its central claim is that the disk's iron gradient is not a single straight line: [Fe/H] falls at -0.078 dex/kpc inside roughly 10.6 kpc and then flattens to -0.041 dex/kpc beyond, with a global linear slope of -0.057 dex/kpc. If correct, the infrared open-cluster abundance scale is shallower and broken, in close agreement with an independent optical survey and notably different from the steeper single-slope picture from a recent automated analysis of the same spectra. The paper also reports that all 18 measured elements show negative [X/H] gradients while most [X/Fe] gradients are near flat or mildly positive, adding new open-cluster gradients for V, Cu, Zn, and Yb.

What carries the argument

The load-bearing mechanism is the photometrically anchored surface gravity (log g) inside the spectral-fitting loop: log g is not a free spectroscopic parameter but is recomputed from a fixed cluster age, distance, extinction, and apparent magnitude using an adopted isochrone, as effective temperature and metallicity are optimized. Because cluster mean abundances and hence the radial gradients are read off this abundance scale, the derived global slope, the knee at 10.6 kpc, and the inner/outer slopes all flow directly from those fixed cluster properties. Supporting machinery includes a curated atomic/molecular line list with non-local thermodynamic equilibrium corrections for key species, s

What would settle it

Redo the abundance fits on the same spectra with log g left free (spectroscopically determined) or with cluster distances/ages taken from an independent astrometric or asteroseismic source; if the knee at ~10.6 kpc and the ~-0.057 dex/kpc global slope disappear or move by more than the quoted uncertainties, the central claim is an artifact of the photometric scale rather than a property of the disk.

Watch

Extended reading notes

Core claim

The analysis claims that after re-fitting the same survey spectra with a conservative sample (655 stars, 133 clusters), cluster mean metallicities decline with galactocentric radius as -0.057 +/- 0.004 dex/kpc in a linear fit, but a broken profile is preferred: knee at 10.6 +/- 0.8 kpc, inner slope -0.078 +/- 0.009 dex/kpc, outer slope -0.041 +/- 0.009 dex/kpc. It further claims that [X/H] gradients are negative for all 18 elements, typically -0.03 to -0.08 dex/kpc, while [X/Fe] gradients are weak and near zero for alpha and iron-peak groups, with mild outward increases for Al, K, Ce, Nd, and Yb consistent with metallicity-dependent or delayed nucleosynthetic channels. The paper interprets t

Load-bearing premise

Everything downstream of the fit depends on the adopted cluster ages, distances, extinctions, and magnitudes that are held fixed while log g is computed; if those cluster properties are wrong in a way that varies with galactocentric radius, the slopes and the break location shift coherently.

Editorial extensions

If this is right

Galactic chemical evolution models must reproduce a broken radial iron gradient with a transition near 10-11 kpc and substantially flatter outer-disk slope, not a single steep linear decline.
The global open-cluster iron gradient becomes consistent between infrared and optical analyses, strengthening the case that the Milky Way's metallicity gradient is genuinely shallow.
New radial gradients for V, Cu, Zn, and Yb give the first open-cluster constraints on these odd-Z and neutron-capture species from infrared spectra, useful for nucleosynthesis and disk-enrichment models.
Weak or mildly positive [X/Fe] gradients with element-specific behavior imply that the dominant radial signal is metallicity-driven and that odd-Z/neutron-capture enrichment scales differently with environment.
Age-binned gradients, while not strictly monotonic, show a tendency for the youngest clusters to have steeper and the oldest to have flatter gradients, which the paper ties to radial migration and age-radius coupling.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Inference: The claimed break at 10.6 kpc is a concrete prediction that can be checked against other tracers (field giants, Cepheids) using the same abundance scale; existing Cepheid data already hint at a similar flattening.
Inference: Because log g is the pivot, an independent check using asteroseismic surface gravities for a subset of these giants would isolate whether the shoulder of the gradient is astrophysical or an artifact of the adopted cluster distances and ages.
Inference: The tighter scatter after visual vetting suggests that automated-pipeline gradients may be inflated by a minority of problematic spectra; quantifying how much of the slope change comes from removing those stars versus from method differences would separate membership effects from abundance-scale effects.
Inference: The non-detection of the proposed young-vs-intermediate metallicity offset at fixed guiding radius is consistent with a sample-size limitation; a targeted expansion in the outer disk could decide whether dilution signatures are absent or just unresolved.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: this is a careful, transparent reanalysis of APOGEE/MWM DR19 spectra for 655 giants in 133 open clusters. The global result — d[Fe/H]/dRgc = -0.057 +/- 0.004 dex/kpc, agreeing almost exactly with the independent optical Gaia-ESO value of -0.054 — looks solid and is the paper's main contribution. The genuinely new bits are the first IR open-cluster gradients for V, Cu, Zn, and Yb, and a cleaner 18-element NLTE-corrected gradient set than OCCAM-DR19.

The paper earns credit on several fronts. Sample construction is careful: Gaia membership with RUWE cuts, cross-checks against Hunt & Reffert and Guerco et al., S/N > 100, and visual inspection of every fit. NLTE corrections are applied for 12 elements. Machine-readable tables and the full line list go to CDS. And the authors are unusually honest: the abstract itself flags the age-radius confound, Section 3.1 says model-atmosphere and atomic-data systematics are hard to quantify, and the Palla et al. dilution test is done properly with bootstrap and permutation tests, yielding a clean null. That is good behavior. No circularity problem — the central result is an empirical measurement.

Soft spots, in proportion.

The broken radial profile — knee at 10.6 +/- 0.8 kpc, inner slope -0.078, outer -0.041 — is the weakest load-bearing claim. The break radius is fitted to the same data, and the paper never reports a formal model comparison showing the bilinear fit actually beats the single line. Otto et al. found a break too but concluded a single line was statistically preferable. I would want that test before believing the break.

The stress-test concern is real. log g is not a free spectroscopic parameter; it is computed from cluster age, distance, AV, and G magnitude adopted from Cantat-Gaudin et al. (2020) and held fixed. Distance-scale systematics therefore enter twice — once on the x-axis (Rgc) and once through log g into the abundance scale. A radius-dependent error could shift or even erase the break, and the paper does not quantify that. The quoted slope errors are purely formal. The paper admits this in Section 3.1 but does not fix it.

The Otto et al. comparison shares authors and data, so 'cleaner than OCCAM' is not an independent check. The paper is appropriately measured here, but the 0.022 dex/kpc slope difference is only meaningful if the two systematic budgets were comparable.

The Gaia-ESO agreement is genuine support for the global gradient; it does not independently establish the break, since Gaia-ESO also uses Gaia-based distances.

Who this is for: galactic archaeologists and GCE modelers who want a benchmark 18-element gradient set. The global slopes and the new element gradients carry the value; treat the break with caution.

Recommendation: send it to peer review. A good referee should push on break significance and distance-systematic propagation, but this is a solid, useful paper.

Referee Report

2 major / 4 minor

Summary. The paper presents a homogeneous re-analysis of APOGEE/MWM DR19 infrared spectra for 655 giants in 133 open clusters, using PySME with MARCS atmospheres and NLTE corrections for several species. Cluster membership is based on Gaia catalogs with strict quality cuts and visual inspection. Stellar parameters and 18 elemental abundances are derived; cluster mean abundances are used to measure radial gradients. The principal results are a global d[Fe/H]/dRgc = -0.057 ± 0.004 dex/kpc, a broken radial profile with a knee at Rgc = 10.6 ± 0.8 kpc and inner/outer slopes of -0.078 ± 0.009 and -0.041 ± 0.009 dex/kpc, and element-dependent [X/H] and [X/Fe] gradients. These are compared with the OCCAM-DR19 results (Otto et al. 2026) and the optical Gaia-ESO results (Magrini et al. 2023). The paper also examines age-dependent gradients and tests the recent-dilution signature of Palla et al. (2024), finding no significant metallicity displacement. The central claim is that the infrared open-cluster abundance scale is shallower and broken, consistent with Gaia-ESO, and that new IR gradients for V, Cu, Zn, and Yb are provided.

Significance. If the inferred gradients are robust, the paper provides a valuable independent, carefully vetted open-cluster abundance scale from infrared APOGEE spectra. The agreement of the global metallicity slope with the optical Gaia-ESO measurement supports the reliability of the overall measurement, while the proposed break at ~10.6 kpc and the new element gradients for V, Cu, Zn, and Yb constitute useful observational constraints for Galactic chemical-evolution models. The strengths of the paper include the homogeneous analysis with explicit NLTE treatment, strict membership selection, visual inspection of all fits, and the public availability of machine-readable abundance tables. These features make the dataset a significant resource even if some of the secondary conclusions (e.g., age dependence) remain tentative.

major comments (2)

[§3.1, §4.2, Table 1] The quoted slope and break-radius uncertainties are purely formal; they do not include the effect of cluster-catalog systematics. Cluster ages, distances, A_V, and Rgc are adopted from Cantat-Gaudin et al. (2020) and held fixed when deriving log g photometrically (§3.1). A radius-dependent error in those distances enters twice: directly in the x-coordinate Rgc and indirectly through log g, which shifts [Fe/H] and all element abundances. The internal ASPCAP comparison (§4.1) shows a median log g offset of -0.13 dex with 0.28 dex rms; the paper does not assess whether this offset is radius-dependent, nor does it propagate the acknowledged model-atmosphere/atomic-data systematics. This is load-bearing for the central claim of a shallow, broken gradient. A sensitivity test—e.g., recomputing log g and abundances using an alternative distance/age catalog such as Hunt & Reffert (2023), or resam
[§4.2, Fig. 4] The claim that the radial metallicity profile is 'broken' is not supported by a statistical model comparison. The text describes a grid search for the break position but never reports the improvement in fit (e.g., Δχ², ΔAIC/BIC, F-test, or bootstrap significance) of the bilinear model over a single straight line. This matters because Otto et al. (2026) found a single linear relation statistically preferable for an overlapping sample. The quoted break-radius uncertainty (±0.8 kpc) is conditional on the broken model and does not by itself establish that the break is significant. A formal significance test should be added; if the test is inconclusive, the conclusions should be softened accordingly.

minor comments (4)

[§4.1, Fig. 5] The text states that the present analysis yields 'smaller reported per-cluster uncertainties' compared to Otto et al. (2026), but the caption to Fig. 5 correctly notes that the uncertainty definitions may not be strictly identical. The text should carry the same caveat, since the comparison may conflate formal fit errors with broader uncertainty estimates.
[§4.2] There are apparent typographical remnants in the text: 'd[Fe/H]/dR Guide = -0.071 ± 0.005' should likely read 'd[Fe/H]/dRgc', and 'd[Fe/H]/dRRgc' appears in the sentence citing Magrini et al. These should be corrected.
[References] In Section 3 the atomic-data papers are cited as 'Nandakumar et al. 2023a,b, 2024a' and in Appendix B as '2023a,b, 2024b'. The reference list contains both 2024a and 2024b with the same journal and page (A&A, 684, A15); please check whether these are two distinct papers and cite them consistently.
[Fig. 6] The caption describes the teal horizontal line as the slope from the complete sample and the shaded band as its uncertainty, but it is not stated whether the band is centered on the line or represents a symmetric error region. A brief clarification would improve readability.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: gradients are empirical fits anchored externally by Gaia-ESO; only non-load-bearing self-citations and a shared-distance systematics caveat.

full rationale

The central claims are empirical measurements, not theory-derived predictions: cluster-mean [X/H] values are obtained by fitting PySME synthetic spectra (MARCS atmospheres, explicit NLTE corrections) to MWM DR19 spectra, and the radial gradients are ordinary linear and piecewise-linear regressions of those measured abundances against Rgc adopted from Cantat-Gaudin et al. (2020) (§2, §3.1, §4.2). No equation makes any reported gradient equal by construction to an input: log g is iteratively recomputed from fixed cluster age, distance, Av, and Gaia G via MIST isochrones (§3.1), so the adopted distance enters both the x-axis (Rgc) and, indirectly, the abundance scale; but [Fe/H] and [X/H] are primarily set by spectral-line fitting, not by the distance. This is a correlated-systematics risk, which the paper itself acknowledges ('systematic uncertainties arising from factors such as model atmospheres and atomic data remain more difficult to quantify,' §3.1), not a definitional reduction. The broken profile is an on-data fit whose break radius is found by grid-search RSS minimization (§4.2), cross-checked against external studies (Joshi et al. 2024; Nunnari et al. 2026), while the global slope is independently anchored by the optical Gaia-ESO value of Magrini et al. (2023), a different team, wavelength regime, and pipeline. Methodological self-citations (Bijavara Seshashayana et al. 2025 for the parameter procedure; Montelius et al. 2022, Nandakumar et al. 2023a,b, 2024a for the line list) are described in sufficient inline detail that no load-bearing result is imported solely by citation; the Otto et al. (2026) benchmark shares two co-authors, but its values do not enter the present derivation. No fitted parameter is relabeled as a prediction, no uniqueness theorem is invoked, and no ansatz is smuggled via citation. The acknowledged limitations (no external validation for V, Cu, Zn, and Yb; unquantified model-atmosphere/atomic-data systematics; distance entering both log g and Rgc) are robustness caveats about the measurement, not circularity in the derivation chain.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The ledger is dominated by domain assumptions about cluster properties, membership, model atmospheres, NLTE corrections, and atomic data. The only explicit fitted extra parameter is the break radius. No new physical entities are introduced.

free parameters (1)

break radius R_knee = 10.59 ± 0.82 kpc
Chosen by minimizing residual sum of squares over a grid of trial break positions, with continuity and requiring at least 10 clusters on each side; this is the extra parameter in the broken-gradient model.

assumptions (5)

domain assumption Cluster ages, distances, AV, and Rgc from Cantat-Gaudin et al. (2020) are accurate enough to anchor the photometric log g and define the radial coordinate.
§2 and §3.1: cluster properties are adopted wholesale and kept fixed during fitting; log g is iteratively recomputed from them. Systematic errors would shift abundances coherently with radius and alter the gradients.
domain assumption Membership probabilities > 0.7 in Cantat-Gaudin et al. (2020) or Hunt & Reffert (2023), with Gaia RUWE < 1.2, identify true cluster members.
§2: membership is the first filter on the sample; contamination would blur cluster means and gradients.
domain assumption MARCS 1D model atmospheres and the cited NLTE corrections (Amarsi et al. 2016, 2020, 2025; Mallinson et al. 2024; Caliskan et al. 2025) are adequate for cool giants in the APOGEE wavelength range.
§3: all abundances are derived through these models; errors in NLTE corrections directly enter [X/Fe] and [X/H].
domain assumption The line list with astrophysical log(gf) adjustments calibrated on solar spectra (Montelius et al. 2022; Nandakumar et al. 2023a,b, 2024) is accurate for all 18 elements.
§3 and Appendix B: atomic data drive the abundance scale; calibrated gf values could carry systematic offsets.
domain assumption MIST isochrones and the photometric relation used to derive log g are valid for the cluster giants.
§3.1: log g is not a free parameter but derived iteratively from MIST isochrones using age, distance, AV, and Gaia G; if the isochrone/photometric scale is wrong, log g and element abundances shift.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Radial abundance gradients of 18 elements in Galactic open clusters from infrared MWM spectra A detailed analysis of 655 giants in 133 clusters." pith.science (2026). https://pith.science/paper/F3YXQ6RC

@misc{pith2026260802563,
  author       = {Pith},
  title        = {Pith review of: Radial abundance gradients of 18 elements in Galactic open clusters from infrared MWM spectra A detailed analysis of 655 giants in 133 clusters},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F3YXQ6RC}},
  note         = {Machine review of arXiv:2608.02563}
}

read the original abstract

Open clusters are powerful tools for studying the Milky Way. While large spectroscopic surveys now provide spectra for many cluster members, automated pipelines and heterogeneous membership selections can introduce systematics and inflate apparent cluster scatter. Therefore, a homogeneous re-analysis with careful membership control and an explicit treatment of departures from Local Thermodynamic Equilibrium is valuable for establishing robust abundance gradients. The aim is to derive precise Galactic radial abundance gradients for multiple elements using open cluster giants, and to investigate how these gradients depend on cluster age. We re-analysed high-resolution infrared APOGEE Milky Way Mapper spectra from DR19 of the Sloan Digital Sky Survey for 655 open cluster members selected from Gaia data that satisfied strict quality cuts on signal-to-noise ratio. Stellar parameters and 18 elemental abundances were obtained using spectrum fitting with the Python version of Spectroscopy Made Easy, applying Non-Local Thermodynamic Equilibrium corrections for several key atomic species. Further quality control of the results was made by visual inspection of all fitted synthetic spectra. The metallicity of the clusters decreases with Galactocentric radius, following a global slope close to -0.06 dex/kpc. Beyond 10-11 kpc, there is modest flattening. In addition to the elements analysed in the \cite{otto2026} study, we derive open-cluster gradients for V, Cu, Zn and Yb using APOGEE spectra.

Figures

Figures reproduced from arXiv: 2608.02563 by the authors.

**Figure 2.** Same as Figure 1 but here [X/H] values are plotted against age. [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗

**Figure 3.** Same as Figure 1 but here [X/H] values are plotted against R [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: [Fe/H] vs. Rgc for our clusters. Each marker represents one cluster. Marker colour indicates the adopted cluster age according to the colour bar, while marker size scales with the number of analysed member stars (1-76 stars). The blue line shows the best-fit linear r…

**Figure 5.** Figure 5: Violin distributions of the reported per-cluster abundance uncertainties for the individual elements in the present [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]

**Figure 6.** Figure 6: The age dependence of the radial abundance gradients, [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]

**Figure 7.** Figure 7: Mean cluster [Fe/H] as a function of orbital guiding [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

2 extracted references

[1]

M., Li, W., Grevesse, N., & Jurewicz, A

Amarsi, A. M., Li, W., Grevesse, N., & Jurewicz, A. J. G. 2025, A&A, 703, A35 Amarsi, A. M., Lind, K., Asplund, M., Barklem, P. S., & Collet, R. 2016, MNRAS, 463, 1518 Amarsi, A. M., Lind, K., Osorio, Y., et al. 2020, A&A, 642, A62 Asplund, M., Amarsi, A. M., & Grevesse, N. 2021, A&A, 653, A141 Beaton, R. L., Oelkers, R. J., Hayes, C. R., et al. 2021, AJ,...

arXiv 2025
[2015]

2022; Nandakumar et al

but updated with astrophysical calibrations based on solar spectra (Montelius et al. 2022; Nandakumar et al. 2023a,b, 2024b).Article number, page 16 of 18 Bijavara Seshashayana et al.: 18 abundances in 655 Giant Stars in 133 Open Clusters Appendix C: Kiel diagram and parameter comparison with OCCAM DR19 Fig. C.1: The main panel shows the full stellar samp...

2022

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.

[1] [1]

M., Li, W., Grevesse, N., & Jurewicz, A

Amarsi, A. M., Li, W., Grevesse, N., & Jurewicz, A. J. G. 2025, A&A, 703, A35 Amarsi, A. M., Lind, K., Asplund, M., Barklem, P. S., & Collet, R. 2016, MNRAS, 463, 1518 Amarsi, A. M., Lind, K., Osorio, Y., et al. 2020, A&A, 642, A62 Asplund, M., Amarsi, A. M., & Grevesse, N. 2021, A&A, 653, A141 Beaton, R. L., Oelkers, R. J., Hayes, C. R., et al. 2021, AJ,...

arXiv 2025

[2] [2015]

2022; Nandakumar et al

but updated with astrophysical calibrations based on solar spectra (Montelius et al. 2022; Nandakumar et al. 2023a,b, 2024b).Article number, page 16 of 18 Bijavara Seshashayana et al.: 18 abundances in 655 Giant Stars in 133 Open Clusters Appendix C: Kiel diagram and parameter comparison with OCCAM DR19 Fig. C.1: The main panel shows the full stellar samp...

2022