How homogeneous was the chemical enrichment of the Milky Way 13 gigayears ago?

Anish Amarsi; Davide Massari; Donatela Romano; Laura Magrini; Martina Rossi; Riano E. Giribaldi

REVIEW 4 major objections 5 minor 21 references

How homogeneous was the chemical enrichment of the Milky Way 13 gigayears ago?

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read With high-precision atmospheric parameters and 3D non-LTE models, this paper finds that the Milky Way's most metal-poor stars formed from gas with a nearly constant magnesium-to-iron ratio — a scatter of only about 0.06 dex — overturning…

desk verdict The 0.06 dex scatter claim is not statistically supported, but the NLTE-vs-LTE comparison is a useful qualitative result that deserves referee attention. read the letter →

arxiv 2507.09743 v1 pith:IVPZENUV submitted 2025-07-13 astro-ph.GA

Riano E. Giribaldi , Laura Magrini , Martina Rossi , Anish Amarsi , Davide Massari , Donatela Romano This is my paper · ORCID

classification astro-ph.GA

keywords Galactichalometal-poorstarsmagnesiumabundance[Mg/Fe]3DNLTEspectroscopychemicalenrichmentstellarpopulationsarchaeology

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 re-examines the chemical make-up of the Milky Way's oldest, most metal-poor stars to ask whether the gas they formed from was chemically well mixed. Using precise atmospheric parameters and 3D non-LTE models, it finds that the magnesium-to-iron ratio among these stars is nearly constant, with a scatter of only about 0.06 dex in the metallicity range -3 < [Fe/H] < -2. That uniformity indicates the early halo was enriched homogeneously, and that the much larger scatter reported in previous studies was largely an artifact of simpler modeling assumptions. The analysis also uncovers a knee-like upturn in [Mg/Fe] at the lowest metallicities and suggests these stars may all have come from a single disrupted population.

What carries the argument

The central machinery is an abundance pipeline anchored to highly accurate atmospheric parameters: effective temperatures from 3D NLTE H-alpha profiles, surface gravities from the ionization equilibrium of Mg triplet lines, and NLTE iron abundances. Magnesium abundances are then synthesized under four assumptions — 1D LTE, 1D NLTE, 3D LTE, and 3D NLTE — so that the contribution of each modeling choice to the observed [Mg/Fe] scatter is isolated. The final 3D NLTE scale shifts the sequence upward by about 0.2 dex and yields the narrow scatter that drives the paper's conclusions.

What would settle it

Re-deriving [Mg/Fe] for the same spectra using an independent set of surface gravities (for example from asteroseismology or parallax) and a different NLTE magnesium model atom, and finding a scatter greater than about 0.1 dex in the range -3 < [Fe/H] < -2, would show the narrow sequence is an artifact of the adopted modeling.

Watch

Extended reading notes

Core claim

The paper shows that the placement of metal-poor stars in the [Mg/Fe] versus [Fe/H] plane depends strongly on the modeling assumptions used to convert spectra into abundances. Under 1D LTE, red giants with log g below about 1.5 show a wide dispersion; under 1D NLTE the dispersion narrows, and under 3D NLTE the whole sequence shifts up by about 0.2 dex in [Mg/Fe]. With this final abundance scale, the intrinsic scatter at -3 < [Fe/H] < -2 is roughly 0.06 dex, with a plateau near [Mg/Fe] ≈ 0.45 for [Fe/H] > -2.8 and a knee rising to about 0.65 dex near [Fe/H] ≈ -3.2. Because the sample stars are kinematically dispersed rather than clumped, the paper proposes they may be the remnants of a single disrupted population, possibly a low-mass galaxy merged into the early Milky Way.

Load-bearing premise

The analysis assumes that the atmospheric parameters and 3D NLTE abundance corrections adopted for these stars are accurate; if those corrections carry systematic errors that depend on metallicity or on the magnesium lines themselves, the narrow scatter and the knee feature would be artifacts of the models rather than real properties of the stellar population.

Editorial extensions

If this is right

If the ~0.06 dex scatter is real, models of early Galactic chemical evolution that predict large stochastic scatter at [Fe/H] < -2 must be revisited, and the first Gyr of enrichment in the Milky Way was more homogeneous than often assumed.
The +0.2 dex shift from 3D NLTE means that [Mg/Fe] ratios for metal-poor giants derived with 1D LTE or 1D NLTE analyses are systematically low, affecting comparisons of halo populations with in-situ stars.
The knee-like rise at [Fe/H] < -2.8 offers a clean chemical marker for the most primitive Milky Way stars, potentially tracing a single low-mass building-block galaxy or a family of similar mini-halos.
If all these stars belong to one disrupted population, the present-day halo preserves a chemical fossil of a specific early accretion event, which can be tested by searching for more members with the same abundance pattern.

Reading between the lines

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

The same 3D NLTE methodology applied to other alpha elements (Si, Ca, Ti) could reveal whether the early halo's homogeneity extends beyond magnesium, or whether magnesium is special.
An extremely low scatter in [Mg/Fe] at these metallicities would favor enrichment by a small number of massive early star-forming halos with efficient mixing, which is a testable prediction for cosmological simulations of the first galaxies.
If the [Mg/Fe] knee is real, its sharpness can be used to date the transition between two enrichment regimes and to constrain the masses of the first supernovae responsible for the upturn.

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

Dear [colleague],

The paper's central claim—that the [Mg/Fe] scatter among metal-poor halo stars is intrinsically as narrow as ~0.06 dex—is not actually demonstrated. The quoted scatter is essentially the same size as the paper's own measurement uncertainty of 0.05–0.10 dex, and the paper shows no error bars, no per-star uncertainties, and no statistical test comparing zero intrinsic scatter with a finite value. That makes the headline number underdetermined. But the qualitative direction deserves credit: the comparison across 1D LTE, 1D NLTE, 3D LTE, and 3D NLTE is instructive, and the fact that NLTE modeling tightens the sequence confirms a hypothesis from Arnone et al. (2005) with much better data and models.

What's actually new here is the use of high-S/N UVES spectra of 15 dwarfs and 13 giants plus 22 Gaia-ESO stars in a consistent 3D NLTE framework, and a reported +0.2 dex offset in [Mg/Fe] between 1D NLTE and 3D NLTE. The paper is honest that its parameter scale comes from previous publications, and the kinematics discussion is appropriately cautious about the lack of dynamical coherence. The knee-like rise below [Fe/H] = -2.8 is an interesting hint, though it rests on very few stars.

The main soft spot is the gap between the data and the central inference. Stating a 0.05–0.10 dex precision and then treating an observed RMS of ~0.06 dex as very narrow is not a valid subtraction of errors. A reader cannot tell whether the intrinsic scatter is 0.00 or 0.10 dex from the plots alone. This is an internal presentational flaw, not just a literature disagreement. A secondary ambiguity is that 3D NLTE corrections are applied only to dwarfs, so it's unclear which combination of models defines the sequence behind the 0.06 dex number. The paper is also thin on methodological detail and ships no data or code, which makes independent checking harder.

This is a conference proceedings paper, so the brevity is understandable. Galactic archaeology folks and stellar abundance modelers will find the qualitative model comparison useful, and it could serve as a reference for the NLTE offset. But the homogeneity claim needs more support before it can be used as evidence. For peer review: I would send it out, because the question is significant and the issue is addressable—an error budget or a likelihood fit of intrinsic scatter would settle it. I would not accept it as is.

Best, [Your name]

Referee Report

4 major / 5 minor

Summary. The paper reanalyzes magnesium abundances in a sample of 50 metal-poor halo stars (15 dwarfs from TITANS I, 13 giants from TITANS II, and 22 from Gaia-ESO) using 1D LTE, 1D NLTE, 3D LTE, and 3D NLTE spectral synthesis of the Mg I 5528 Å line. The authors report that the [Mg/Fe] versus [Fe/H] distribution in the range -3 < [Fe/H] < -2 dex is narrow (~0.06 dex) compared with previous studies, that a knee-like rise to [Mg/Fe] ≈ 0.65 dex appears below [Fe/H] ≈ -2.8, and that the stars' kinematics are widely dispersed. They interpret these results as evidence that early magnesium enrichment was more homogeneous than previously thought and that the stars may originate from a single disrupted population.

Significance. The qualitative direction of the study—comparing 1D LTE and 3D NLTE abundance corrections—is valuable, and the use of high-S/N UVES spectra with R > 40000 is a strength. If the ~0.06 dex scatter claim were supported by a proper error analysis, it would provide an important constraint on early Galactic enrichment and strengthen the case that part of the dispersion in the literature is a modeling artifact. However, as presented, the central quantitative claim is not statistically supported: the quoted scatter is of the same order as the stated measurement precision, and no intrinsic-scatter test is provided. The paper also leaves unspecified which model combination yields the 0.06 dex value, and the 3D NLTE corrections are applied only to dwarfs. These issues must be resolved before the main conclusion can be accepted.

major comments (4)

[Sections 2.1 and 4] The stated precision of the [Mg/Fe] measurements is 0.05–0.10 dex (Section 2.1), while Section 4 concludes that the scatter at −3 < [Fe/H] < −2 dex is 'remarkably narrow' at about 0.06 dex. An observed RMS of 0.06 dex is exactly what one would obtain from measurement noise alone for an intrinsically perfectly homogeneous population; the paper provides no per-star uncertainties, no error bars in Figure 3, and no statistical test (e.g., a chi-square or likelihood ratio) that distinguishes zero intrinsic dispersion from a finite value. The central claim of homogeneous early enrichment is therefore underdetermined by the data as presented, independent of the correctness of the 3D NLTE models.
[Sections 3 and 4] The paper does not specify which combination of models and sample was used to derive the 0.06 dex scatter. Section 3 states that 3D NLTE corrections were computed only for dwarfs, and Section 4 first describes the 1D NLTE sequence as 'narrow' and then introduces a +0.2 dex shift from 3D NLTE. Because giants (Precision sample II and Gaia-ESO stars) are analyzed only up to 3D LTE, the reader cannot determine whether the 0.06 dex value refers to the 1D NLTE dwarf+giant sequence, the 3D NLTE dwarf-only sequence, or some mixture. The authors should state the exact model choice and sample for the scatter estimate and, ideally, quote the scatter separately for dwarfs and giants.
[Section 4] The 'knee-like feature' below [Fe/H] < −2.8 dex, with [Mg/Fe] rising to ~0.65 dex at [Fe/H] ~ −3.2 dex, is presented without the number of stars in that range, their individual uncertainties, or a significance estimate. In a sample of 50 stars, this feature could be driven by a small number of points. Please provide the sample size, the measured values with errors, and a quantitative assessment of whether the change is statistically significant.
[Sections 2.1 and 3] The entire analysis rests on the authors' own 3D NLTE parameter scale (Teff from Hα profiles, log g from Mg triplet lines, and NLTE Fe abundances, established in Giribaldi et al. 2021, 2023). The paper presents no independent validation of this scale (e.g., asteroseismic gravities or parallax-based distances) and no discussion of how systematic errors in Teff, log g, or the Mg line formation would propagate into the [Mg/Fe] scatter. If such systematics correlate with [Fe/H], the narrow sequence and the knee could be artifacts of the model rather than stellar-population properties. A quantitative propagation of systematic uncertainties is needed to support the main claim.

minor comments (5)

[Figure 3 caption] The caption contains a typo: 'codded' should be 'coded' (in 'colour-codded according to log g').
[Introduction] The phrase 'over timein' should be 'over time', and the reference formatting is inconsistent (e.g., 'et. al' vs 'et al.').
[Sections 2.1–2.3] The term 'Precision samples I and II' is used without definition; the reader must consult the cited papers to learn what distinguishes them. A sentence describing the samples (e.g., dwarf vs giant, selection criteria) would improve the proceedings article.
[Section 2.3] The sentence 'The square box indicate the area with highest probability of enclosing GES stars defined by Massari et al. (2019)' is grammatically incorrect and does not specify whether this box is the same 3 kpc box used earlier.
[Abstract and Section 4] The abstract uses 'relatively narrow' while Section 4 uses 'remarkably narrow (~0.06 dex)'; the quantitative value and the model combination used should be stated consistently in both places.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the narrow-scatter result is a model-dependent observational reduction, not an input by construction.

full rationale

The paper's central claim is that the [Mg/Fe] scatter is ~0.06 dex after applying 1D/3D NLTE corrections, based on re-deriving abundances for archival and Gaia-ESO spectra. The atmospheric parameters come from the authors' earlier works (Giribaldi et al. 2021, 2023) and the 3D NLTE models from Amarsi et al. 2018, but these are external inputs, not outputs of the present analysis. Nothing in the reduction forces a narrow sequence: the same framework could in principle produce a wide [Mg/Fe] distribution. The comparison with the Erebus/Enceladus reference populations is a scientific interpretation, not a derivation step. The only notable weakness is that the reported 0.06 dex scatter is comparable to the stated 0.05-0.10 dex precision, and the paper does not perform an intrinsic-scatter test; this is an evidentiary/statistical concern, not circularity. No circular step is exhibited.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new free parameters in the sense of fitting data; the derived quantities are model outputs. The main assumptions are inherited from the 3D NLTE framework and the authors' previous parameter determinations, and the interpretation relies on the representativeness of a kinematically mixed sample.

free parameters (1)

No new fitted parameters
The paper does not fit free parameters to data; the 3D NLTE corrections and the 0.2 dex offset are model outputs, not fitted constants.

assumptions (4)

domain assumption The 3D NLTE H-alpha based Teff scale (Giribaldi et al. 2021, 2023) is accurate for metal-poor dwarfs and giants.
Section 2.1 states that the [Mg/Fe] scale is anchored to accurate effective temperatures from 3D NLTE H-alpha profiles.
domain assumption The Mg model atom of Bergemann et al. (2017) and the Balder/Scate codes correctly compute NLTE and 3D effects for Mg I 5528 Angstrom.
Section 3 describes Mg abundance derivation using these codes and model atoms without a validation against independent constraints.
domain assumption Ionization equilibrium using log g from Mg triplet lines is satisfied for the sample.
Section 2.1 asserts that ionization equilibrium is satisfied, which is a condition for the derived [Fe/H] and [Mg/Fe] values.
domain assumption The sample of 50 stars is sufficiently representative of the ancient halo despite its kinematic incoherence.
Section 2.3 shows the stars are widely dispersed, yet the paper uses them to infer a universal early enrichment pattern; selection effects are not quantified.

how reviews work

0 comments

Cite this review

Pith. "Pith review of How homogeneous was the chemical enrichment of the Milky Way 13 gigayears ago?." pith.science (2026). https://pith.science/paper/IVPZENUV

@misc{pith2026250709743,
  author       = {Pith},
  title        = {Pith review of: How homogeneous was the chemical enrichment of the Milky Way 13 gigayears ago?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IVPZENUV}},
  note         = {Machine review of arXiv:2507.09743}
}

abstract

We reanalyze the chemical composition of the metal-poorest tail of the Galactic halo using highly accurate atmospheric parameters Giribaldi et al. (2021, 2023) and cutting-edge 3D NLTE models Amarsi et al (2018). Most [Mg/Fe] versus [Fe/H] diagrams in the literature exhibit significant scatter at [Fe/H] $\lesssim -2$ dex, often interpreted as evidence of inhomogeneous enrichment during the early phases of galaxy evolution Rossi et al. (2021). However, our analysis of observational data reveals that in the range $-3.5 <$ [Fe/H] $< -2$ dex, the [Mg/Fe] versus [Fe/H] distribution is relatively narrow. This finding suggests a low degree of stochastic enrichment in magnesium during these epochs in the Milky Way halo.

Figures

Figures reproduced from arXiv: 2507.09743 by the authors.

**Figure 1.** Left panel: Kiel diagram. Precision samples I and II, and the Gaia-ESO sample are displayed as indicated in the legends. Symbols are colour-coded according to the metallicity. Yonsey-Yale isochrones with ages and metallicity values given the legends are over-plotted as reference. Right panel: Spatial distribution of our sample stars with respect to the Galaxy centre. Stars in the GALAH survey are displayed as backg… view at source ↗

**Figure 2.** Lindblad diagram. Each panel colour-map emphasises the Gaia-Enceladus and Milky Way (Erebus and Splash) populations with the sample stars separated by Giribaldi & Smiljanic (2023). Left panel: The symbols indicate to which sample the stars belong. Also, stars with Gaia-Enceladus-like dynamics are indicated by blue contours. The square box indicate the area with highest probability of enclosing GES stars defined by … view at source ↗

**Figure 3.** Symbols of dwarfs are size-coded according to their ages. As a reference, stars classified as insitu Milky Way (blue) and Gaia-Enceladus (gray) in Giribaldi & Smiljanic (2023) are plotted. Red, blue, and black lines display LOWESS regressions. The shades display 20 to 80% quantiles. Dwarfs (Precision sample I) have no colour coding. Red giants (Precision sample II and Gaia-ESO sample) are colour-codded according to… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

21 extracted references · 20 canonical work pages

[1]

write newline

" write newline "" before.all 'output.state := FUNCTION fin.entry add.period write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence 'output.state := if if FUNCTION not #0 #1 if FUNCTION and 'skip pop #0 if FUNCTIO...
[2]

M., Spite M., Korotin S

Andrievsky S. M., Spite M., Korotin S. A., et al. 2010, A&A , 509, A88

work page 2010
[3]

Amarsi A., Nordlander T., Barklem P. S., et. al \ 2018, A&A , 615A, 139

work page 2018
[4]

G., Argast D., Norris J

Arnone, E., Ryan S. G., Argast D., Norris J. E., & Beers T. C. 2005, A&A , 430, 507

work page 2005
[5]

Belokurov V., Erkal D., Evans N. , et. al \ 2018, MNRAS , 478, 611

work page 2018
[6]

2020, A&A , 494, 3880

Belokurov V., Sanders J., Azadeh F., et al. 2020, A&A , 494, 3880

work page 2020
[7]

M., et al

Bergemann M., Collet R., Amarsi A. M., et al. 2017, ApJ, 847, 15

work page 2017
[8]

K., Sahlholdt C

Feuillet D. K., Sahlholdt C. L., Feltzing S., & Casagrande L. 2021, MNRAS , 508, 1489

work page 2021

Show all 21 references

[9]

M., Magg E., Plez B., et al

Gerber J. M., Magg E., Plez B., et al. 2023, A&A, 669, A43

2023
[10]

E., Da Silva A., Smiljanic R., et

Giribaldi R. E., Da Silva A., Smiljanic R., et. al \ 2021, A&A , 650A, 194

2021
[11]

Giribaldi R. E. & Smiljanic R. 2023, A&A , 673, A18

2023
[12]

E., Van Eck S., Merle T., et

Giribaldi R. E., Van Eck S., Merle T., et. al \ 2023, A&A , 679A, 110

2023
[13]

C., et al

Gilmore G., Randich S., Worley C. C., et al. 2022, A&A , 666, A120

2022
[14]

2011, A&A, 529, A158

Hayek W., Asplund M., Collet R., & Nordlund A. 2011, A&A, 529, A158

2011
[15]

2021, A&A, 645, A106

Heiter U., Lind K., Bergemann M., et al. 2021, A&A, 645, A106

2021
[16]

Helmi A., Babusiaux C., Koppelman H. H., et. al \ 2018, Nature , 563, 85

2018
[17]

P., Mackereth J

Horta D., Schiavon R. P., Mackereth J. T., et al. 2021, MNRAS , 500, 1385

2021
[18]

H., Helmi A., et al

Massari D., Koppelman H. H., Helmi A., et al. 2019, A&A, 630, L4

2019
[19]

T., Miglio A., et al

Montalbán J., Mackereth J. T., Miglio A., et al. 2021, Nature Astronomy , 5, 640

2021
[20]

2022, A&A, 666A, 121

Randich S., Gilmore G., Magrini L., et al. 2022, A&A, 666A, 121

2022
[21]

2021, MNRAS , 503, 6026

Rossi M., Salvadori S., Skúladóttir A. 2021, MNRAS , 503, 6026

2021

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

write newline

" write newline "" before.all 'output.state := FUNCTION fin.entry add.period write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence 'output.state := if if FUNCTION not #0 #1 if FUNCTION and 'skip pop #0 if FUNCTIO...

[2] [2]

M., Spite M., Korotin S

Andrievsky S. M., Spite M., Korotin S. A., et al. 2010, A&A , 509, A88

work page 2010

[3] [3]

Amarsi A., Nordlander T., Barklem P. S., et. al \ 2018, A&A , 615A, 139

work page 2018

[4] [4]

G., Argast D., Norris J

Arnone, E., Ryan S. G., Argast D., Norris J. E., & Beers T. C. 2005, A&A , 430, 507

work page 2005

[5] [5]

Belokurov V., Erkal D., Evans N. , et. al \ 2018, MNRAS , 478, 611

work page 2018

[6] [6]

2020, A&A , 494, 3880

Belokurov V., Sanders J., Azadeh F., et al. 2020, A&A , 494, 3880

work page 2020

[7] [7]

M., et al

Bergemann M., Collet R., Amarsi A. M., et al. 2017, ApJ, 847, 15

work page 2017

[8] [8]

K., Sahlholdt C

Feuillet D. K., Sahlholdt C. L., Feltzing S., & Casagrande L. 2021, MNRAS , 508, 1489

work page 2021