{"id":"cccaa80e-7e12-4424-8d01-f2cfe43a9841","arxiv_id":"2502.08164","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":17,"one_line_summary":"A best-fit action-based model of the Milky Way from K giant kinematics gives a virial mass of about 1.3 x 10^12 solar masses and a thicker, hotter thick disk than the 2017 model.","lead":"This paper builds new dynamical models of the Milky Way by fitting the motions of 86,109 K giant stars from the LAMOST DR8 and Gaia EDR3 surveys, and finds a best-fit model with a total virial mass near 1.3 x 10^12 solar masses and a thicker, hotter thick disk than earlier models. The models fit well near the Sun but fail in the outer Galaxy, suggesting the Milky Way has structure not captured by simple axisymmetric models.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reported virial mass is not constrained by the fit: models with M200 from 0.39 to 1.35e12 fit comparably, and the fiducial choice is made on local chi2 alone, so the headline mass is model-dependent.","rationale":"The reader's weakest_assumption identifies axisymmetry and missing non-equilibrium structures as the key risk. That is a real concern, and the paper itself admits (Section 4.6) that all models fail in the outer regions. However, the most load-bearing issue for the paper's headline numbers is more direct and visible inside the paper: the virial mass quoted in the abstract (1.35e12 Msun) is neither the fiducial model's value (1.31e12 Msun in Table 2) nor a robustly constrained quantity. Table 3 shows that models with M200 differing by a factor of more than three (Ca20d: 0.39e12; Mc17a: 1.31e12) have total reduced chi2 values within about 2 of each other, with Ca20e actually having the lowest total reduced chi2 and an intermediate mass. The fiducial model is chosen on the basis of chi2_in only, which is appropriate for local kinematics but not for a global mass claim. Section 4.5 explicitly acknowledges the mass range and states that the dark matter distribution is the critical factor. Since the local kinematics only pin down the enclosed mass near the Sun, the outer halo profile and the presence or absence of a CGM drive the virial mass, and the data do not discriminate. The paper also does not provide MCMC error bars on M200, instead giving only a qualitative statement that parameter uncertainties are below 10%, which does not cover the systematic model scatter. The central mass claim therefore rests on a model prior, not on the data. This does not invalidate the more robust local conclusions, such as the thicker, more extended thick disk and cooler thin disk relative to Wang et al. (2017), which are supported by the DF parameters across models. It does mean the headline virial mass should be presented as strongly prior-dependent, or the abstract should be revised to quote a range. The reader's conditional verdict already captures the need for revision, so I keep the verdict unchanged but emphasize that the mass claim is the specific weak point that a revision must address.","tokens_in":25201,"tokens_out":9148,"duration_ms":102352,"concrete_test":"Using the published Table 3, compute a formal model comparison (e.g., BIC or AIC) between Mc17a, Ca20d, and Ca20e over the full 30 regions, with dof from Table 3; if Ca20e or Ca20d is preferred or the delta_chi2 is within the expected scatter, then the reported M200 = 1.35e12 Msun is not supported. A stronger check: rerun the fitting pipeline with the CGM added to the Mc17 initialization and with the outer regions (21-24) excluded; if the recovered M200 moves by more than 20% under either change, the headline virial mass is systematics-dominated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 4.4 selects Mc17a (fid) solely because it has the smallest chi2_in, i.e., the fit to the 16 local regions. Yet Table 3 shows that the full-dataset reduced chi2 values are nearly indistinguishable: chi2_tot/dof = 10.36 for Mc17a, 12.54 for Ca20d, and 9.97 for Ca20e, the last being the best overall. Ca20d has M200 = 0.39e12 Msun and Ca20e has 0.84e12 Msun, while Mc17a has 1.31e12 Msun; the abstract reports 1.35e12, which does not match the fiducial model. Section 4.5 itself states that masses range from 0.39 to 1.31e12 Msun and that the dark matter distribution is the critical factor. The local kinematics only constrain the enclosed mass near the Sun; the virial mass is an extrapolation that depends on the adopted halo profile (uncontracted NFW vs. contracted Cautun et al. 2020; presence/absence of CGM). Because delta_chi2 between these models is roughly 2 over thousands of degrees of freedom, the data do not discriminate among them, and no MCMC error bars are provided. Thus the headline M200 is not a stable output of the data; it is a prior-dependent quantity that the paper nonetheless presents as a central result.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs axisymmetric, self-consistent action-based distribution-function models of the Milky Way using the AGAMA code, fitted to the three-dimensional velocity distributions of 86,109 K giants from LAMOST DR8 and Gaia EDR3 in 30 sky regions. The authors vary disk and halo density and distribution-function parameters seeded by several published models (McMillan 2017, Cautun et al. 2020, Wang et al. 2017), iterate the potential to self-consistency, and compare the models by reduced chi-square. They report a best-fit local model with a Milky Way virial mass of 1.35e12 Msun, a local stellar density of 0.0696 Msun/pc3, and a local dark matter density of 0.0115 Msun/pc3, and conclude that the thick disk is hotter and more extended while the thin disk is cooler than in Wang et al. (2017). The paper acknowledges that the fits to the outer regions and the North Galactic Pole are poor, and attributes the discrepancies to missing structures such as spiral arms, the bar, and non-equilibrium features.","tokens_in":25745,"tokens_out":7069,"duration_ms":105101,"significance":"If the local results were robust, the paper would provide a useful update to Wang et al. (2017) by adding Gaia proper motions and a larger LAMOST sample. The use of a public, well-tested modeling code (AGAMA), explicit self-consistency iteration, and comparison against several published potentials are genuine strengths, as is the honest reporting of the poor outer-region fits. However, the headline virial mass is not a stable output of the data: the paper itself shows that models with M200 from 0.39 to 1.31e12 Msun fit comparably, and the fiducial selection is based on local chi-square only. The analysis is therefore more credible as a local dynamical calibration than as a global Milky Way mass measurement, and the abstract overstates the robustness of M200.","major_comments":[{"comment":"The reported virial mass is not robustly constrained and is internally inconsistent. The abstract states 1.35e12 Msun, but Table 2 gives M200=1.31e12 for the fiducial 'Mc17a'; the value 1.35e12 corresponds to 'Mc17b'. More importantly, Section 4.4 selects 'Mc17a' solely because it has the smallest chi2_in, while Table 3 shows chi2_tot/dof = 9.97 for Ca20e (best overall), 10.36 for Mc17a, and 12.54 for Ca20d, with M200 values of 0.84, 1.31, and 0.39e12 Msun, respectively. The differences in reduced chi-square among these models are small (0.4-3) and are dwarfed by the overall poor fit (reduced chi-square ~10), so the data do not discriminate among halo models. The virial mass is an extrapolation from local kinematics that depends strongly on the assumed halo profile, baryonic contraction, and CGM treatment (Sections 3.4-3.5 and 4.5). Section 4.2 states that MCMC sampling was performed and that uncertainties are generally below 10%, but no MCMC results, posteriors, or convergence diagnostics are shown. The paper should either provide the MCMC confidence intervals or remove the virial mass from the abstract and present it explicitly as a model-dependent extrapolation.","section":"Abstract; Section 4.4; Section 4.5; Table 2; Table 3"},{"comment":"The global fit quality is poor, so the model cannot support global conclusions. The fiducial model has chi2_tot/dof = 10.36, chi2_out/dof = 27.10 for regions 21-24, and chi2_pole/dof = 21.91 for regions 25-26; even the best global model, Ca20e, has chi2_tot/dof = 9.97. Section 4.6 states that all models fail in the outer regions and that the predicted velocities are narrower than observed, and Section 4.4 notes that proper-motion fits are 2-5 times worse than line-of-sight fits. The statement that 'our model aligns well with observations near the Sun' is defensible only for the 16 local regions (chi2_in = 5.84), but even this value is substantially larger than 1 and indicates unmodeled variance in the local volume. The conclusions about a hotter, more extended thick disk and a cooler thin disk are based on fits that fail to describe a large fraction of the fitted volume; the paper should either restrict these conclusions to the local region or extend the model (e.g., with non-axisymmetric components, flaring, or a warp) and refit before presenting them as main results.","section":"Table 3; Section 4.4; Section 4.6"},{"comment":"The axisymmetric, steady-state assumption is load-bearing and is explicitly acknowledged to be violated. The model assumes a steady-state axisymmetric system with three conserved actions (Section 3.2), yet Section 4.6 invokes the bar, spiral arms, phase spirals, and the flared/warped outer disk to explain the poor fits. These structures are inside the fitted volume: regions 21-24 reach the Galactic anticenter at 3-12 kpc and cross the Perseus arm, and Section 4.4 notes that regions 02, 03, and 14 include the Local Arm. If non-axisymmetric perturbations bias the inferred distribution-function parameters, then the quoted local densities and masses are also biased. The paper should quantify the impact of this assumption, for example by repeating the fit with regions containing known spiral-arm or bar-affected stars removed, or by explicitly stating that the quoted parameters are conditional on strict axisymmetry and are not intended as unbiased estimates.","section":"Section 3.2; Section 4.6; Section 5"}],"minor_comments":[{"comment":"Equation (5) defines pdata_n as the 'observed velocity dispersion', but the text describes fitting binned histograms of vLOS, mu_alpha, and mu_delta; please clarify the notation and define how the degrees of freedom are counted for each region and how the region-averaged reduced chi-square is justified as a global comparison statistic.","section":"Section 3.1, Eq. (5)"},{"comment":"The distance corrections D = DCarlin/0.86 in the disk direction and D = DCarlin/0.97 toward the pole are adopted from Ding et al. (2021) with a small change; please state explicitly whether these factors are fixed or fitted, and whether their uncertainties are propagated into the model parameter uncertainties.","section":"Section 2.1"},{"comment":"There is a typographical issue in Table 2: the row labeled 'Mv17i' appears to be a typo for 'Mc17i', and there are two rows with similar 'Mc17i' labels but different parameter values; please disambiguate the model names.","section":"Table 2"},{"comment":"The caption for Table 3 states that chi2_pole is 'estimated from sky regions 25-16', which should presumably read '25 and 26'; please correct this typo.","section":"Table 3"},{"comment":"The sentence 'we employed MCMC sampling to estimate the uncertainties in the fitted parameters' followed by 'specific MCMC results are not included in this paper' is not verifiable; either provide the chains, a corner plot, or a clear statement that these are preliminary and model-dependent, or remove the claim of small (<10%) uncertainties.","section":"Section 4.2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript has real strengths: it uses a large, carefully cleaned sample, a public modeling code, and it openly discusses the failure of axisymmetric models in the outer disk. The main problem is that the abstract and conclusion present the virial mass as a headline measurement when the paper's own comparison shows it is highly model-dependent and not selected by the global fit. I recommend requesting a revision that either supplies the missing uncertainty quantification or substantially reframes the claims as a local calibration, with the virial mass presented as a prior-dependent extrapolation. The internal inconsistency between the abstract's 1.35e12 and Table 2's 1.31e12 for the fiducial model should also be fixed before acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Solid, careful application of AGAMA action-based distribution-function modeling to 86k K giants from LAMOST DR8 + Gaia EDR3. The new content is the dataset and the full three-velocity fit: proper motions are now included, a wider grid of starting models is tried, and the comparison is transparent. The authors deserve credit for using a well-tested public code, for presenting the bad fits alongside the good ones, and for stating plainly that all models fail in the outer disk. The local results — cooler thin disk, hotter/more extended thick disk, local stellar and dark matter densities around 0.07 and 0.0115 solar masses per cubic parsec — are plausible and consistent with prior work.\n\nThe soft spot is the headline virial mass. The fiducial model is chosen because it has the lowest chi2_in for the 16 local regions, but the full-dataset reduced chi2 values in Table 3 are almost flat: Mc17a gives 10.36, Ca20d 12.54, Ca20e 9.97 (best overall). Those models have M200 = 1.31, 0.39, and 0.84e12 solar masses respectively. The abstract's 1.35e12 doesn't even match the fiducial model's 1.31e12 in Table 2. The paper itself notes the mass range 0.39–1.31e12 and that the dark matter distribution is the critical factor. With no MCMC error bars and delta_chi2 around 2 over thousands of degrees of freedom, the data simply do not pin down M200. The virial mass is an extrapolation depending on halo profile, contraction, and CGM assumptions. That is not a fraud; it is a prior-dependent output, and the paper should say so in the abstract rather than leading with it.\n\nAlso soft, in decreasing order: the overall fit quality is poor in absolute terms (reduced chi2 around 10; outer regions 19–35), which the authors attribute to non-axisymmetric and non-equilibrium structures — reasonable, but it means the model is incomplete for the outer regions it claims to cover. The distance calibration applies region-specific scale factors (0.86 disk, 0.97 pole) after the fact, a systematic that could shift parameter values. The axisymmetry and steady-state assumption is load-bearing and only partially validated. These are not fatal for the local kinematic constraints, but they need to be quantified.\n\nBottom line: this is a useful application paper, not a measurement of the Milky Way's virial mass. Send it to review — the data and pipeline are real and the comparison is valuable — but the revision needs to demote M200 from headline to caveated extrapolation, add error bars, and fix the abstract/table inconsistency. I'd bring it to reading group as an example of model selection degeneracy in action-space fits.","headline":"Solid application of AGAMA action-based modeling to a valuable K-giant sample, but the headline virial mass is model-dependent rather than measured, and the paper's own tables show why.","tokens_in":26263,"tokens_out":2759,"would_cite":true,"duration_ms":24175,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper constructs self-consistent action-based Milky Way models and shows that the three-dimensional kinematics of 86,109 K giant stars are fit by a galaxy with a virial mass of about 1.3×10^12 M⊙, a local stellar density of 0.0696 M⊙…","keywords":["Galaxy dynamics","Milky Way Galaxy","Milky Way mass","Galaxy kinematics","Milky Way Galaxy physics","Galaxy physics","action-based distribution functions","K giant stars"],"falsifier":"Measure the local dark matter density independently with the vertical Jeans equation using Gaia RVS stars; if it differs from the paper's value of 0.0115 M⊙ $pc^{-3}$ by more than the quoted uncertainties, the fiducial axisymmetric model is falsified.","tokens_in":25025,"feed_emoji":"🌌","tokens_out":10296,"duration_ms":77382,"temperature":0.7,"pith_summary":"The paper sets out to build a self-consistent dynamical model of the Milky Way that reproduces the full three-dimensional velocity distribution of K giant stars, using proper motions from Gaia EDR3 and spectroscopy from LAMOST DR8. Its central claim is that the near-Sun kinematics of 86,109 K giants are consistent with an axisymmetric action-based model whose virial mass is about 1.3×$10^{12}$ M⊙, local stellar density 0.0696 M⊙ $pc^{-3}$, and local dark matter density 0.0115 M⊙ $pc^{-3}$. The model replaces the earlier torus-based model of Wang et al. (2017), which did not use Gaia proper motions and now fails the new proper-motion distributions. The best-fit disks are different from that earlier model: the thick disk is hotter and more extended, while the thin disk is cooler. Near the Sun the fit is good, but all models fail at 15–20 kpc, which the authors attribute to unmodeled structures such as spiral arms, the bar, or non-equilibrium features.","feed_headline":"K giants set Milky Way mass near 1.3 trillion suns","feed_subtitle":"Fits to 86,109 K giants put the virial mass near 1.3 trillion solar masses and pin down local dark matter density.","key_machinery":"The central object is a quasi-isothermal action-based distribution function f(J), written in terms of the three action integrals (J_R, J_z, J_φ) that describe radial, vertical, and azimuthal orbital motion in an axisymmetric potential. The argument is carried by the self-consistency loop: start from an initial density-potential pair, compute actions for a sample of orbits, integrate the distribution function to get a new density, solve Poisson's equation for the updated potential, and repeat until convergence. The fitted parameters are the disk surface-density scale lengths, scale heights, and the velocity-dispersion scales σ_r0, σ_z0, and R_σ for the thin and thick disks, plus the dark-matter halo parameters. Model predictions for line-of-sight velocity and both proper-motion components in 30 sky regions are compared with observations through a χ² statistic, and two optional ingredients—a contracted dark-matter halo and a circumgalactic medium—are built into some models.","core_discovery":"On its own terms, this paper claims that a self-consistent, axisymmetric Milky Way model written as an action-based distribution function can simultaneously match the line-of-sight velocities and Gaia proper motions of K giants in the solar neighborhood, and that this requires a Milky Way with a virial mass of M200 = 1.31×$10^{12}$ M⊙ for the fiducial model (the abstract quotes 1.35×$10^{12}$ M⊙), a local stellar density of 0.0696 M⊙ $pc^{-3}$, and a local dark matter density of 0.0115 M⊙ $pc^{-3}$. The fiducial best-fit model, Mc17a, starts from the McMillan (2017) potential and is selected by the lowest reduced χ² in the inner sky regions. Compared with the earlier Wang et al. (2017) torus model, the fitted thick disk has a larger vertical velocity dispersion and scale length, and the thin disk has a smaller vertical velocity dispersion. The paper also reports that proper-motion data are what exclude the older model, and that all attempted models predict velocity distributions that are narrower than observed in the outer disk, signaling missing ingredients in the axisymmetric equilibrium picture.","pith_inferences":["Beyond the paper: The systematic failure in the outer disk means the fiducial mass and densities are effectively calibrated to the inner few kiloparsecs; a model that included the bar and spiral arms could redistribute velocity dispersion and shift the inferred local dark matter density.","Beyond the paper: The distance-calibration ratios D = D_Carlin/0.86 and D = D_Carlin/0.97 are load-bearing; a change in these scale factors would rescale the stellar densities and velocities and alter the fitted distribution-function parameters.","Beyond the paper: The spread of virial masses across the paper's models, from about 0.39 to 1.31×10^12 M⊙, shows that K-giant kinematics near the Sun do not by themselves determine the total halo mass; the outer halo mass is fixed by the assumed profile rather than by the data.","Beyond the paper: A testable extension is to fit the same action-based distribution function to Gaia RVS stars or red-clump stars with independent distances; if the derived local dark matter density changes by more than the quoted uncertainties, the axisymmetric steady-state assumption is the likely cause."],"forward_implications":["If the fiducial model is right, the Milky Way's virial mass of about 1.3×10^12 M⊙ and local dark matter density of 0.0115 M⊙ pc^-3 are the values consistent with K-giant kinematics.","Proper motions are now a required constraint: any viable disk model must reproduce Gaia proper-motion distributions, not just line-of-sight velocities.","The thick disk is hotter and more extended while the thin disk is cooler, changing expectations for how the two disks formed and evolved.","The outer disk at 15–20 kpc is hotter than axisymmetric equilibrium models predict, implying missing non-axisymmetric or non-equilibrium components.","With contracted-halo models that include a circumgalactic medium, the virial mass can be as low as about 0.4×10^12 M⊙, so the outer halo mass remains uncertain even with good local kinematics."],"supporting_citations":[{"why":"Supplies the earlier torus model and K-giant sample analysis that the new proper-motion fits exclude.","marker":"Wang et al. (2017)"},{"why":"Provides the initial axisymmetric potential from which the fiducial best-fit model starts.","marker":"McMillan (2017)"},{"why":"Supplies the contracted dark-matter halo model and circumgalactic medium profile used in the lower-mass model variants.","marker":"Cautun et al. (2020)"},{"why":"Constructs the K-giant catalog from LAMOST DR8 cross-matched with Gaia EDR3 that all fits use.","marker":"Ding et al. (2021)"},{"why":"Provides the photometric distance calibration applied to the LAMOST spectra.","marker":"Carlin et al. (2015)"},{"why":"Serves as the main comparison self-consistent model whose reported fit quality motivates the new approach.","marker":"Binney & Vasiliev (2023)"},{"why":"Updated comparison model with a revised bulge, used to show the residual gap in outer-disk fits.","marker":"Binney & Vasiliev (2024)"},{"why":"Formulates the quasi-isothermal action-based distribution function adopted for the stellar disks.","marker":"Piffl et al. (2014)"},{"why":"Supplies the action-finding and self-consistent iteration machinery that turns a distribution function into a density-potential pair.","marker":"Vasiliev (2019a)"},{"why":"Supplies the Gaia EDR3 parallax zero-point correction used to derive distances.","marker":"Lindegren et al. (2021)"}],"fun_headline_variants":["86,109 K giants map Milky Way's mass and dark matter","New Milky Way model from 86k stars ups thick disk","Gaia and LAMOST data reshape Milky Way disk model","Action-space models pin Milky Way mass at 1.35 trillion suns","Thicker thick disk and cooler thin disk from 86k giants"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the Milky Way can be treated as an axisymmetric galaxy in a steady state, so that every star's orbit is fully described by three conserved actions in a fixed potential; if the survey volume contains the bar, spiral arms, or phase spirals, the fitted distribution-function parameters and the inferred mass and densities are biased.","fun_headline_variants_meta":{"raw":{"variants":["86,109 K giants map Milky Way's mass and dark matter","New Milky Way model from 86k stars ups thick disk","Gaia and LAMOST data reshape Milky Way disk model","Action-space models pin Milky Way mass at 1.35 trillion suns","Thicker thick disk and cooler thin disk from 86k giants"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000909,"raw_usage":{"total_tokens":3964,"prompt_tokens":1062,"completion_tokens":2902,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":678,"completion_tokens_details":{"reasoning_tokens":2811}},"tokens_in":678,"tokens_out":2902,"duration_ms":19770,"temperature":1.0,"reasoning_tokens":2811,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T10:12:06.868007+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the local dark matter density independently with the vertical Jeans equation using Gaia RVS stars; if it differs from the paper's value of 0.0115 M⊙ $pc^{-3}$ by more than the quoted uncertainties, the fiducial axisymmetric model is falsified.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the earlier torus model and K-giant sample analysis that the new proper-motion fits exclude."},{"cited_title":"L., Liu, C., Newberg, H","cited_arxiv_id":null,"evidence_quote":"Provides the photometric distance calibration applied to the LAMOST spectra."}],"review_version":1}