{"id":"bf4d3231-19f9-4e4b-8304-b8406f8f6dc5","arxiv_id":"2507.09542","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Analysis of 12,000 very metal-poor stars finds a large disk-like component at low orbital heights, suggesting an ancient 'primordial' Milky Way disk.","lead":"Using a new catalog of 12,000 very metal-poor stars, the authors split orbital speeds into three components and find a surprisingly large number of ancient stars moving on flat, disk-like orbits. They argue these stars are leftovers of a 'primordial' disk that formed before the Milky Way's current disk.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Sample-purity check is likely biased: the LAMOST DR10 overlap is not representative, and the dominant residual contaminants ([Fe/H]_LAMOST between -1.8 and -1.5) have v_phi near 150 km/s at low Zmax, so they can produce the NMF 'disk' component without a true VMP disk.","rationale":"The reader identifies sample purity as the weakest assumption; I agree, but the specific threat is slightly different. The paper emphasizes the 2.6%/1.8% tail above [Fe/H]_LAMOST=-1.5, while its own success-rate definition implies a much larger 18%/7% residual at the -1.8 threshold. Most of this residual lies in the -1.8 to -1.5 interval, which is precisely the metal-weak thick disk regime whose v_phi peak (~150 km/s) matches the NMF 'disk' component. If the low-Zmax bins are substantially contaminated by such stars, the central kinematic evidence for a primordial VMP disk disappears. This is a correctness risk, not a consensus dispute: the existence of some VMP stars on disk-like orbits is supported by prior work, but the quantitative dominance claimed here depends on purity. The LAMOST overlap is the only validation presented, and LAMOST DR10's selection function is not accounted for. The proposed test is executable with existing data and would settle the issue. Because the paper should be accepted only if that test passes, the reader's conditional verdict stands; no change is needed.","tokens_in":15413,"tokens_out":12514,"duration_ms":143328,"concrete_test":"Compute selection-function-corrected purity in metallicity bins using an independent spectroscopic survey (GALAH DR4, APOGEE DR17, or DESI EDR) matched to the same parent photometric catalog, then rerun the Section 3 NMF after removing or down-weighting every star with spectroscopic [Fe/H] above -2.0. As a direct check, take the confirmed non-VMP interlopers, measure their v_phi/Zmax distribution, subtract it from the NMF input matrix, and see whether Component 2 in Figure 6 retains its amplitude at Zmax<=3 kpc. If it does, contamination is not responsible for the disk claim; if it drops toward the contamination fraction, the primordial-disk conclusion would no longer be supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"To support the claim that most low-Zmax VMP stars belong to a primordial disk, the final 12,000-star sample must be genuinely very metal-poor. Section 2's only purity check (Figure 1) uses 533 dwarfs and 1679 giants in common with LAMOST DR10. LAMOST DR10 is not an unbiased survey: its target selection enriches low-metallicity and halo stars, so the overlap success rate is likely an overestimate of true purity. The quoted 2.6%/1.8% residual above [Fe/H]_LAMOST=-1.5 is only the bright tail. At the paper's own success-rate threshold ([Fe/H]_LAMOST<=-1.8), 18% of dwarfs and 7% of giants remain spectroscopically inconsistent with VMP, mostly in the -1.8 to -1.5 interval. These are metal-weak thick disk stars, with v_phi near 150 km/s and low Zmax, exactly the phase space of NMF Component 2 ('disk') in Figure 6. In low-Zmax bins, especially for the 213 dwarfs with Zmax<1 kpc, a 10-18% contamination concentrated in those bins can create or strongly enhance the disk component. The statement that only 2.6%/1.8% have [Fe/H]>-1.5 therefore understates the threat to the central claim. Without weighting by the LAMOST selection function or an independent spectroscopic purity measurement, the 'primordial disk' interpretation is not yet established.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs a sample of about 12,000 very metal-poor stars (1604 dwarfs and 10,396 giants) from Gaia XP photometric metallicities, astrometry, and geometrical distances, applies quality and isochrone cuts, computes orbital parameters with AGAMA, and decomposes the prograde v_phi distribution across Z_max bins using non-negative matrix factorization. Three components are identified as halo, disk, and Gaia-Sausage/Enceladus; the disk-like component peaks near v_phi ~ 150 km/s and its fraction decreases with Z_max. The authors argue that these disk-like VMP stars form an independent primordial disk component rather than being solely merger debris or a rotating halo.","tokens_in":15711,"tokens_out":5219,"duration_ms":61433,"significance":"If the central claim is correct, this would be an important contribution to Galactic archaeology: it would provide a large-sample, photometric-metallicity-based argument for a primordial, low-metallicity disk component in the Milky Way, complementing earlier spectroscopic studies. The paper also delivers a publicly available catalog of VMP stars with orbital parameters, uses a transparent NMF framework, and carefully documents its astrometric and photometric quality cuts. The main significance is contingent on sample purity and on the decomposition being a valid test of the primordial-disk hypothesis rather than an assumed label assignment.","major_comments":[{"comment":"The sample-purity check is not sufficient to support the claim that the final VMP sample is clean enough for the disk-component analysis. The check is based on 533 dwarfs and 1679 giants in common with LAMOST DR10, but LAMOST DR10 target selection enriches low-metallicity and halo stars, so the overlap success rate is likely an overestimate of the purity of the XP-selected full sample. More importantly, the paper itself states that the residual contamination is primarily in the [Fe/H] interval -1.8 to -1.5, and that interval is excluded from the quoted '2.6%/1.8% above -1.5' figure. These metal-weak thick-disk stars have v_phi near 150 km/s and low Z_max, exactly the phase space assigned to NMF Component 2. For the 213 dwarfs with Z_max < 1 kpc shown in Figure 5, a contamination fraction of order 10-18% concentrated in that bin can materially enhance or even create the apparent disk component. A selection-function-weighted comparison, or an independent spectroscopic purity measurement, is required before the primordial-disk interpretation can be considered established.","section":"Section 4, Figure 7"},{"comment":"The consistency between the reconstructed and original v_phi distributions is guaranteed by the construction of an NMF fit and does not constitute independent validation of the physical decomposition. The statement that the reconstruction 'demonstrates the effectiveness and accuracy' of the method is circular in this context. The authors should validate the decomposition on synthetic data: for example, draw a pure halo + GSE population, inject a known metal-weak thick-disk or disk-like population at low Z_max, and test whether NMF recovers the injected fractions. Without such a test, the three-component decomposition is an assumption about the data rather than a measurement.","section":"Section 4, Figure 7"},{"comment":"The identification of Component 2 as an independent 'primordial disk' relies on the 150 km/s threshold and the MWTD identification from Carollo et al. (2019) and Hong et al. (2024), which are the same authors and hypothesis under test. The alternative interpretation that low-Z_max VMP stars are simply the prograde tail of a continuously rotating halo with a vertical gradient is not explicitly fitted. Please test a two-component model (one halo component whose mean v_phi varies smoothly with Z_max, plus GSE) against the three-component model and report a model comparison such as AIC or BIC. This would directly address whether a distinct disk component is required by the data and would reduce the concern that the labels are imported rather than derived.","section":"Sections 3.2 and 5"}],"minor_comments":[{"comment":"The label 'Compontent' in the left panel is a typo and should read 'Component'.","section":"Figure 6"},{"comment":"The reference Carollo et al. (2019) appears twice in the reference list; one occurrence should be removed.","section":"References"},{"comment":"The purity numbers are presented inconsistently: Section 2 reports post-cut success rates of 82% (dwarfs) and 93% (giants) for [Fe/H] <= -1.8, while Section 5 quotes a 93% success rate and a '94% to 82%' comparison. The Section 5 completeness values (89% overall, 93% versus 59%) are not defined. These numbers should be reconciled and the completeness statistic defined.","section":"Sections 2 and 5"},{"comment":"There is a duplicated letter in the appendix: two items are labeled 'E.', and the text contains 'of of' in item C. These should be corrected.","section":"Appendix A"},{"comment":"The data availability line gives the placeholder 'http://to be determined.com'. The catalog should be assigned a permanent URL before publication, as the catalog is one of the paper's main assets.","section":"Section 2"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope for a Galactic-astronomy journal, and the catalog plus the orbital-parameter set are potentially useful. My main concern is that the central claim rests on sample purity and on a decomposition whose physical labels are imported from the authors' prior work; both need to be strengthened with quantitative tests. The paper should not be rejected outright, but the revision must include a selection-function-aware purity estimate and a synthetic-data or alternative-model test of the NMF decomposition."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a serious paper that assembles a 12,000-star VMP sample from Gaia XP and finds a prograde, low-Zmax component peaking at v_phi ~150 km/s via NMF. The quantitative result is new and the sample is much larger than previous kinematic studies of VMP stars. It deserves a proper referee. But the leap from \"there is a kinematic component\" to \"this is a primordial disk\" is only as strong as the sample purity, and the paper's own purity check may not support it.\n\nWhat is actually new: the NMF decomposition of the prograde v_phi distribution in Zmax bins, the ~20% disk-like fraction at Zmax ~3 kpc, and the Hayden-criterion cross-check. These are useful numbers. The paper is honest that prior work (Hong et al. 2024, Bellazzini et al. 2024) already reported disk-like VMP stars; this is a larger, cleaner confirmation rather than a first discovery.\n\nSoft spots: the sample purity check in Fig. 1 uses only the LAMOST DR10 overlap, which is not a random sample of the photometric selection; LAMOST's target selection enriches metal-poor stars. More importantly, even on that overlap, at the paper's own success threshold ([Fe/H]_LAMOST <= -1.8), 18% of dwarfs and 7% of giants remain above it, mostly in the -1.8 to -1.5 interval. Those stars are metal-weak thick-disk stars with v_phi near 150 km/s and low Zmax, exactly the phase space of NMF Component 2. So the claim that only 2.6%/1.8% of the sample is contaminated above -1.5 misses the relevant population. The NMF reconstruction in Fig. 7 matching the input is guaranteed by the method, not independent validation. Systematic uncertainties from selection and contamination are not propagated into the reported fractions. The data table URL is a placeholder and no code is shipped.\n\nThat said, the kinematic evidence for a disk-like component among VMP stars is robust and agrees with earlier studies from this group and Bellazzini et al. The interpretation as a primordial in-situ disk is underdetermined, but the paper does consider alternatives (minor mergers, bar heating) and argues why they are unlikely to explain the full fraction. That is a reasonable scientific position, not a fatal flaw.\n\nBottom line: for a reader working on Galactic archaeology or early disk formation, this is worth reading and citing. It should go to a serious referee, but the referee should push for a contamination analysis that accounts for the LAMOST selection function, or an independent spectroscopic purity check, plus public data and code. The conclusion should be conditional, not definitive.","headline":"Larger VMP sample and clean NMF decomposition confirm a prograde low-Zmax component, but the primordial-disk conclusion leans on a purity check that likely overstates sample purity.","tokens_in":16331,"tokens_out":3619,"would_cite":true,"duration_ms":34221,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Many of the Milky Way's most metal-poor stars orbit in a flat disk, evidence that a primordial disk formed during the Galaxy's earliest phase.","keywords":["very metal-poor stars","primordial disk","Milky Way formation","Galactic archaeology","non-negative matrix factorization","Gaia XP spectra","orbital dynamics","metal-weak thick disk"],"falsifier":"Obtain high-resolution spectra for the stars on disk-like orbits ($Z_{\\rm max}\\le3$ kpc and $v_\\phi>150$ km/s); if a substantial majority turn out to have true [Fe/H] above -1.5, the disk component is contamination rather than a primordial disk. Alternatively, redo the NMF analysis excluding all stars with photometric [Fe/H] between -2.0 and -1.5 and check whether the disk component persists.","tokens_in":15147,"feed_emoji":"🌌","tokens_out":17575,"duration_ms":157522,"temperature":0.7,"pith_summary":"Very metal-poor stars (with [Fe/H] ≤ −2.0) have long been treated as halo fossils, but this paper argues that a substantial fraction of them—especially those whose orbits stay close to the Galactic plane—form an independent disk system dating back to the Milky Way's primordial formation. The authors build a sample of 12,000 VMP stars (1,604 dwarfs and 10,396 giants) from Gaia XP photometric metallicities, compute their orbits with the AGAMA package, and decompose the rotation-velocity distribution across maximum orbital height ($Z_{\\rm max}$) using non-negative matrix factorization. Three components emerge: the halo, the Gaia Sausage/Enceladus debris, and a disk component peaking at $v_\\phi \\sim 150$ km/s whose fraction decreases with $Z_{\\rm max}$ but remains about 20% at $Z_{\\rm max}\\sim 3$ kpc. The authors conclude that such a high disk fraction cannot be explained by merger debris or bar heating alone, and that most of these stars are members of a hypothesized primordial disk. If correct, the Milky Way developed a rotating disk very early in its history, and the metal-weak thick disk is a surviving relic of that phase.","feed_headline":"12,000 metal-poor stars point to a primordial Milky Way disk","feed_subtitle":"The finding suggests the Milky Way grew a rotating disk very early in its history.","key_machinery":"The load-bearing analysis is a non-negative matrix factorization (NMF) of an $8\\times10$ matrix whose entries count prograde VMP stars in ten $v_\\phi$ bins (0–350 km/s in 35 km/s steps) for eight $Z_{\\rm max}$ bins (0–8 kpc in 1 kpc steps). NMF represents the data as a sum of non-negative components, each a $v_\\phi$ profile with a per-$Z_{\\rm max}$ coefficient; the first three components account for 99.54% of the covariance and are identified as the halo (peak at 50–150 km/s), the disk system (peak near 150 km/s with $\\sigma\\sim80$ km/s), and the Gaia-Sausage/Enceladus debris (0–100 km/s, falling steeply). The disk component's coefficient is large at low $Z_{\\rm max}$ and declines with height, which is the signature the authors read as a primordial disk. Cross-checks include the energy–$J_\\phi$ distribution and the Haywood inclination-angle statistic.","core_discovery":"On the paper's own terms, the central discovery is that the rotation-velocity distribution of very metal-poor stars contains a distinct, high-rotation component peaking near $v_\\phi\\sim150$ km/s that dominates the low-$Z_{\\rm max}$ population and fades with height above the plane. This component cannot be explained as the stationary or slowly rotating halo, as debris of the Gaia-Sausage/Enceladus merger, or as disk-like orbits produced by bar and spiral-arm heating, so the authors conclude it is an independent early structure: the hypothesized primordial disk. They further argue that the metal-weak thick disk corresponds to this primordial disk, and that most VMP stars with $Z_{\\rm max}\\lesssim3$ kpc belong to it. The argument is carried by the NMF decomposition of $v_\\phi$ across $Z_{\\rm max}$ bins, corroborated by the energy–$J_\\phi$ distribution and by the Haywood inclination-angle criterion, which shows a prograde-to-retrograde ratio near 2.1:1 among low-inclination stars.","pith_inferences":["If the disk component is a distinct primordial population, its stars should carry a characteristic chemical signature, such as high [alpha/Fe] with a particular carbon or neutron-capture pattern, that differs from both the canonical thick disk and GSE debris; this can be searched for in the large samples from DESI, 4MOST, and LAMOST III.","Applying the same NMF decomposition to the full Gaia XP metallicity catalog across all [Fe/H] ranges could reveal whether the 'primordial disk' component is separate from, or continuous with, the canonical thin and thick disks, clarifying whether it is a true separate population or the low-metallicity tail of the thick disk.","A sharper contamination test would be to model, from the canonical disk's stellar population, how many metal-rich interlopers are expected inside the isochrone selection box; if that predicted number alone accounts for the full disk component, the primordial interpretation would be falsified.","The claim predicts that the disk-like VMP stars have ages near 13 Gyr and were born before the GSE merger; asteroseismic or isochrone ages for these stars, once available, could confirm or refute the primordial origin."],"forward_implications":["If the primordial disk is real, the Milky Way must have developed a rotating disk during its earliest star-forming phase, not only after mergers built up the halo.","The metal-weak thick disk should be regarded as a surviving relic of that primordial disk, distinct from the canonical thick disk, the halo, and the Gaia Sausage/Enceladus debris, rotating near 150 km/s rather than the canonical thin disk's ~200 km/s.","The observed disk fraction, about 20% at $Z_{\\rm max}\\sim3$ kpc and dominant for $Z_{\\rm max}\\lesssim3$ kpc, provides a quantitative constraint that galaxy-formation simulations must reproduce.","Accretion-only explanations would require an implausibly large number of low-energy, co-planar minor mergers to deposit so many VMP stars on disk-like orbits, making the in-situ primordial disk the more economical interpretation.","Upcoming surveys with precise ages and chemical abundances should find that the primordial-disk stars are uniformly very old and exhibit distinctive abundance patterns, offering a direct test of this interpretation."],"supporting_citations":[{"why":"Supplies the Gaia XP photometric metallicities from which the 12,000-star VMP sample is selected.","marker":"Huang et al. (2024b)"},{"why":"Provides the geometric distances used to build 6-D phase-space coordinates for the orbital calculations.","marker":"Bailer-Jones et al. (2021)"},{"why":"Sets the Z_max binning, the v_phi > 150 km/s disk threshold, and the prior evidence for disk components among VMP stars that this work extends.","marker":"Hong et al. (2024)"},{"why":"Characterizes the metal-weak thick disk rotation near 150 km/s that the NMF disk component is identified with.","marker":"Carollo et al. (2019)"},{"why":"Provides the AGAMA package used to compute all dynamical parameters, including v_phi, eccentricity, and Z_max.","marker":"Vasiliev (2019)"},{"why":"Supplies the Galactic gravitational potential assumed in the orbit calculations.","marker":"McMillan (2017)"},{"why":"Defines the Gaia Sausage/Enceladus debris signature used to label the third NMF component.","marker":"Naidu et al. (2020)"},{"why":"Presents the competing interpretation of a prograde halo with no disk below [Fe/H] < -1.6 that the paper's disk detection must argue against.","marker":"Zhang et al. (2024a)"},{"why":"Provides the LAMOST spectroscopic data used to validate the photometric metallicities and estimate the residual contamination rate.","marker":"Cui et al. (2012)"}],"fun_headline_variants":["Ancient star orbits reveal a primordial rotating disk","Metal-poor stars' rotation hints at an early Milky Way disk","New disk-like population found among the Milky Way's oldest stars","12,000 ancient stars expose a previously hidden rotating component"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the photometric metal-poor sample is genuinely metal-poor, meaning that after the quality and isochrone cuts, contamination from stars with [Fe/H] above -1.5 is only 2.6% for dwarfs and 1.8% for giants, so the disk-like component peaking near $v_\\phi\\sim150$ km/s is not mostly ordinary thick-disk interlopers.","fun_headline_variants_meta":{"raw":{"variants":["Ancient star orbits reveal a primordial rotating disk","Metal-poor stars' rotation hints at an early Milky Way disk","New disk-like population found among the Milky Way's oldest stars","12,000 ancient stars expose a previously hidden rotating component"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000201,"raw_usage":{"total_tokens":1455,"prompt_tokens":1095,"completion_tokens":360,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":711,"completion_tokens_details":{"reasoning_tokens":293}},"tokens_in":711,"tokens_out":360,"duration_ms":5150,"temperature":1.0,"reasoning_tokens":293,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T17:54:05.977340+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Obtain high-resolution spectra for the stars on disk-like orbits ($Z_{\\rm max}\\le3$ kpc and $v_\\phi>150$ km/s); if a substantial majority turn out to have true [Fe/H] above -1.5, the disk component is contamination rather than a primordial disk. Alternatively, redo the NMF analysis excluding all stars with photometric [Fe/H] between -2.0 and -1.5 and check whether the disk component persists.","supporting_citations":[],"review_version":1}