{"id":"a86c9345-f500-4c8d-8979-640c55845894","arxiv_id":"2506.12717","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"A decelerating Milky Way bar appears able to transfer angular momentum to bulge and halo stars, matching the rotation of a neural-network-selected sample of 1.18 million Gaia stars.","lead":"Using Gaia data and a neural network, the authors isolate about 1.18 million Milky Way bulge and inner halo stars that rotate together. A test-particle simulation with a slowing galactic bar reproduces their rotation profile, pointing to the bar as the engine that spun them up.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The steady-bar control may not isolate deceleration from bar mass/radial growth, so the causal attribution to deceleration is not yet established.","rationale":"The reader's weakest assumption pinpoints precisely the ambiguity in the control run, and my reading of Sections 3.1 and 4.2 confirms that the paper does not specify whether the steady-bar control includes the bar mass and radial growth used in the main run. This is the single most load-bearing gap because the entire causal argument--that deceleration, not growth or initial conditions, transfers angular momentum to the bulge and halo--rests on the clean isolation of the pattern-speed time dependence. A growing bar, even with constant pattern speed, can pump angular momentum into resonant orbits, so the current comparison cannot rule out growth as the primary driver. I do not find a more fundamental flaw: the NN methodology is applied consistently to observation and simulation, the DF initial conditions are plausible, and the agreement in spatial distribution and composition fractions is supportive, but not decisive. The proposed test--re-running the steady control with identical growth--would settle whether deceleration is indeed necessary. Since this is an addressable gap rather than a demonstrated error, the conditional verdict remains appropriate; no change to the reader's recommendation is needed.","tokens_in":14610,"tokens_out":6231,"duration_ms":73020,"concrete_test":"Rerun the steady-bar control simulation with the same bar mass growth (from M0 to 2.0 M0) and radial growth (to 1.26 times initial) as the main decelerating run, holding Omega_b constant at -35 km/s/kpc. Then compare the v_phi(R) profile of pseudo-stars with NN predictions in [0.24, 0.4] to (a) the original steady control, (b) the decelerating run, and (c) the observed Gaia profile. If the steady+growth profile rises to match the observed/decelerating profile, the deceleration-specific claim is not established; if it remains close to the original steady control, the claim is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that bar deceleration, rather than bar growth or initial conditions, produces the observed v_phi profile of the NN-selected rotating component. Section 4.2 compares the decelerating-bar run to a 'steadily rotating bar' control, but the text only states that the control has constant pattern speed Omega_b = -35 km/s/kpc, following Chiba & Schonrich (2022). It does not state whether the control also includes the same bar mass growth (x2.0) and radial growth (x1.26) that the main run has (Section 3.1: 'the mass and radial extent of the bar evolve continuously, reaching 2.0 and 1.26 times their initial values'). If the control lacks this growth, then the two runs differ in two independent physical drivers: pattern-speed evolution and bar mass/size evolution. Bar growth alone can trap stars at resonances and transfer angular momentum even at constant pattern speed, so the discrepancy visualized in Figure 4 (blue dashed vs. black/blue curves) could be caused by growth, not deceleration. Because the paper's title and conclusions explicitly attribute the rotation to the decelerating bar, this confound is load-bearing. The issue is addressable by re-running the control with identical growth, and until then the causal interpretation remains provisional.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper identifies a subset of 1,175,737 stars in Gaia DR3 with neural network predictions between 0.24 and 0.4 as an 'atypical rotating component' with net rotation of roughly 80 km/s. To explain its origin, the authors build a test-particle simulation with an axisymmetric distribution-function model of the Milky Way (following Binney & Vasiliev 2023, 2024) plus a central bar that decelerates from 56 to 35 km/s/kpc over 4 Gyr while also growing in mass and radial extent. They resample the simulation to mimic Gaia selection effects, apply the same neural network, and compare the simulated v_phi(R) profile of the selected particles with the observed one, reporting strong agreement. A comparison run with a steadily rotating bar shows a discrepancy, which the authors interpret as evidence that the decelerating bar transfers angular momentum to bulge and halo stars. The paper concludes that the rotating component is predominantly bulge, halo, and thick disk stars and that bar deceleration is pivotal in shaping the inner Galaxy's kinematics.","tokens_in":14884,"tokens_out":5380,"duration_ms":67373,"significance":"If the causal attribution holds, the paper would supply a large-sample observational check on a long-standing theoretical expectation: that a decelerating bar can secularly pump angular momentum into the bulge and inner stellar halo. The computational setup is well chosen in several respects: the simulation uses a physically motivated distribution-function model, includes synthetic Gaia-like observational errors and selection functions, and makes an explicit comparison to a steady-bar control. The authors also candidly list caveats about neglecting self-gravity and spiral arms. However, the central claim currently rests on a single prescribed bar model and a hand-selected neural network window, and the comparison in Figure 4 is visual rather than statistical. The paper is a useful contribution, but the deceleration hypothesis is not yet demonstrated at the strength claimed in the title and conclusions.","major_comments":[{"comment":"The comparison that isolates bar deceleration is not fully specified. The main run has a bar that simultaneously decelerates (Omega_b from -56 to -35 km/s/kpc) and grows in mass and radial extent (Section 3.1: reaching 2.0 and 1.26 times initial values), but the control run is described only as maintaining a constant pattern speed Omega_b = -35 km/s/kpc. If the control omits the mass and radial growth, the difference between the blue dashed and solid curves in Figure 4 could be caused by bar growth rather than by deceleration. Since the title and Section 5 attribute the net rotation to the decelerating bar, the manuscript must specify the control run's mass and size evolution or rerun it with identical growth, and show that the discrepancy persists.","section":"Section 4.2, steady-bar control"},{"comment":"The agreement between simulated and observed v_phi(R) profiles is asserted visually, with no quantitative goodness-of-fit, uncertainty estimate, or test of the difference between the decelerating and steady runs. Figure 4 shows median and 16th/84th percentile bins, but the model has many free parameters (DF component masses and action scales, bar parameters, selection thresholds), and no posterior or parameter variation is presented. A bootstrap or chi-squared comparison, and ideally a small exploration of the neural network prediction window, would be needed to support the claim of 'strong agreement' and the attribution of the rotation to bar deceleration.","section":"Section 4.2, Figure 4"},{"comment":"The definition of the rotating component depends on the hand-selected NN prediction interval [0.24, 0.4], with the right boundary chosen to exclude GSE stars. The subsequent composition analysis, Figure 3, and Figure 4 all use this specific window. The manuscript does not test the sensitivity of the v_phi(R) profile or the simulated-versus-observed agreement to the boundaries. If the profile is robust to these choices, that should be demonstrated; if not, the selection is a potential source of the claimed signal.","section":"Section 2.2, NN selection window"},{"comment":"The initial distribution function is constrained by the same APOGEE/Gaia-based data that define the observed sample (the BV24 fits listed in Tables B1 and B2), so the simulation is not fully independent of the observations. The authors should discuss the extent to which the initial conditions already encode rotating bulge or halo kinematics, and clarify whether the bulge and halo DFs permit net rotation at t=0. The internal decelerating-versus-steady comparison is the strongest guard against circularity, but the absolute agreement in Figure 4 is not an independent confirmation.","section":"Section 3 and Appendix A"}],"minor_comments":[{"comment":"The pattern speeds are given as negative numbers (Omega_b = -56 and -35 km/s/kpc), but the direction convention is not defined; the resonance variable epsilon = (Omega_phi - Omega_b)/Omega_r in Figure 5 assumes a sign convention that should be stated explicitly.","section":"Section 3.1, sign convention"},{"comment":"The sentence 'For the first time, we confirmed the net rotation of both the bulge and inner halo on a million-level sample' is stronger than the cited literature warrants, given earlier reports of bulge and inner-halo rotation cited in the introduction; suggest softening to 'on a million-level sample' without the 'first time' phrasing.","section":"Section 5, 'for the first time'"},{"comment":"The right boundary of the selected NN interval is said to exclude GSE member stars, but no quantitative criterion is given; providing the actual boundary test or a reference would reduce the impression of arbitrariness in the sample definition.","section":"Section 2.2, GSE exclusion"},{"comment":"The caption text 'The solid lines illustrate a comparison of all simulated samples with NN prediction values ranging between 0.24 and 0.4' is a little unclear; the legend label 'SimSteady' is not described in the caption, and the labels for the orange and red dashed curves would be easier to follow if the figure legend and caption were aligned.","section":"Figure 4 caption"}],"recommendation":"major_revision","confidential_remarks":"The paper is potentially suitable for publication after the load-bearing control issue is addressed. The main risk is that the causal attribution to bar deceleration is currently over-stated relative to the evidence in the manuscript; I would ask the authors to run the steady-bar control with the same mass and radial growth as the fiducial run, and to provide a quantitative comparison between the simulated and observed v_phi profiles. The manuscript also would benefit from a sensitivity test of the NN prediction window, since the entire sample definition depends on that choice."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"I read the paper. The new thing is a million-star rotating sample selected from Gaia by a neural network, plus a test-particle simulation with a decelerating bar constrained to the present-day phase angle. The authors show that the decelerating-bar model reproduces the observed v_phi profile of the selected sample, while a steadily rotating bar does not. That is a genuinely interesting comparison, and the paper is honest about the main caveats: no self-gravity, no spiral arms.\n\nWhat the paper does well: the sample selection is carefully described, the DF model is standard and well-sourced (Binney & Vasiliev, Li & Binney), the simulation uses AGAMA with eight million particles, and the comparison in Figure 4 is visually striking. The paper does not oversell the significance. If the deceleration attribution holds, it strengthens the secular-evolution picture for the bulge and inner halo rotation.\n\nThe soft spot is the control run. The 'steadily rotating bar' is described only in terms of pattern speed. The main run has the bar mass growing by 2.0x and radial extent by 1.26x. If the control lacks that growth, then the two runs differ in two ways — pattern-speed evolution and bar growth — and the discrepancy in Figure 4 could be driven by growth alone. The paper does not say one way or the other. That is load-bearing because the title and conclusions attribute the rotation specifically to deceleration. The authors should either confirm that the control includes the same growth or re-run the control with identical mass and radial evolution.\n\nA second, smaller concern: the NN selection window (0.24–0.4) is hand-chosen and no sensitivity tests are shown. The composition claim (bulge/halo/thick disk) is inferred from the simulation rather than directly measured in the data, and the agreement in Figure 4 is visual, with no model-uncertainty quantification. These are addressable in revision.\n\nThe citation pattern looks fine. The paper builds on Li et al. (2024b) and the Chiba & Schönrich work, and the relevant literature is cited. I do not see a circularity problem beyond what is normal for this kind of forward-modeling exercise: the DF is constrained by the same data that define the sample, but the simulation itself is not fitted to the v_phi curve.\n\nWho is this for? Anyone working on the Milky Way bar, bulge kinematics, or secular evolution. It deserves a serious referee — the central idea is plausible and the sample is large. But the referee should push on the control run before the causality claim is accepted. I would not cite it as evidence for bar deceleration until that control is clarified.","headline":"A plausible, readable case that a decelerating bar spins up the bulge and inner halo, but the control run does not cleanly isolate deceleration from bar growth, so the causal claim is provisional.","tokens_in":15410,"tokens_out":2310,"would_cite":false,"duration_ms":23693,"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":"The paper claims that the Milky Way's decelerating bar, not accretion or initial conditions, spun up the stars that form the observed rotating bulge and inner halo.","keywords":["Galactic bar","Galactic bulge","Milky Way stellar halo","angular momentum transfer","bar deceleration","test-particle simulation","neural network classification","Gaia DR3"],"falsifier":"Run the control simulation with a bar whose pattern speed is constant at $-35\\,\\mathrm{km\\,s^{-1}}\\,\\mathrm{kpc}^{-1}$ but whose mass and radial extent grow exactly as in the decelerating run (mass $\\times2.0$, radius $\\times1.26$ over 4 Gyr). If that control reproduces the observed $v_\\phi(R)$ profile, the claim that deceleration is the cause fails; if it does not, deceleration is supported. An independent check would measure the present-day bar pattern speed from stellar kinematics and verify that it has fallen by roughly 37.5% over the last 4 Gyr, matching the assumed braking history.","tokens_in":14413,"feed_emoji":"🌌","tokens_out":8987,"duration_ms":93028,"temperature":0.7,"pith_summary":"The paper identifies a population of 1,175,737 stars in Gaia DR3 whose neural-network scores mark them as neither ordinary disk nor obvious accreted debris, and shows they rotate at roughly $v_\\phi\\simeq80\\,\\mathrm{km\\,s^{-1}}$. It then argues that this rotation is not primordial: in a test-particle simulation, a central bar whose pattern speed $\\Omega_b$ falls from $-56$ to $-35\\,\\mathrm{km\\,s^{-1}}\\,\\mathrm{kpc}^{-1}$ over 4 Gyr transfers angular momentum to bulge and halo stars, reproducing the observed $v_\\phi(R)$ profile, while a steadily rotating bar does not. If correct, the Milky Way's bulge and inner halo acquired their net prograde motion secularly, through dynamical friction braking the bar, rather than through initial conditions or accretion alone. The result would make the decelerating bar a central engine for the inner Galaxy's kinematics.","feed_headline":"A decelerating bar spins up the Milky Way's bulge and halo","feed_subtitle":"Simulation with a braking bar reproduces the observed rotation of 1.18 million selected stars.","key_machinery":"The central object is the decelerating bar potential: a simple analytic bar whose pattern speed $\\Omega_b$ drops from $-56$ to $-35\\,\\mathrm{km\\,s^{-1}}\\,\\mathrm{kpc}^{-1}$ (about 37.5% deceleration) while its mass grows by a factor of 2.0 and its radial extent by a factor of 1.26 over 4 Gyr. The test particles are pseudo-stars sampled from an equilibrium distribution-function model of the Milky Way, evolved in the axisymmetric background plus the bar. The transfer is tracked through changes in angular momentum as a function of the resonance ratio $\\epsilon=(\\Omega_\\phi-\\Omega_b)/\\Omega_r$; particles trapped near corotation ($\\epsilon\\simeq0$) gain the most angular momentum, and the neural-network-selected bulge and halo subset shows net gains that a steadily rotating bar cannot reproduce.","core_discovery":"The paper's central claim is that the observed rotating component is not a distinct stellar population but a mixture of bulge, halo, and thick-disk stars that have been given angular momentum by the Milky Way's decelerating bar. The evidence is a test-particle simulation that initializes pseudo-stars from an equilibrium distribution-function model of the Galaxy and evolves them for 4 Gyr in an axisymmetric background plus a bar whose pattern speed drops by about 37.5% while its mass and radial extent grow. After applying observational errors, selection effects, and the same neural network used on the data, the simulated $v_\\phi(R)$ profile for the selected stars agrees with the observed profile, and the bulge and halo components individually show the rotation seen in the data. A comparison run with a steadily rotating bar produces a profile that deviates significantly from observation. The paper concludes that dynamical friction decelerating the bar is the pivotal process shaping the inner Galaxy's kinematics.","pith_inferences":["If deceleration is the active ingredient, a control run in which the bar grows in mass and length while keeping a constant pattern speed should still fail to match the data; the paper does not state that its steady control includes that growth, so this is the natural next test.","Because the simulation neglects self-gravity, the real bar may transfer angular momentum even more efficiently than modeled, making the simulated rotation a possible lower bound on the bulge and halo spin.","The same torquing mechanism could explain why even metal-poor inner-halo stars rotate: the bar acts on pre-existing old stars, so no separate accretion origin is required."],"forward_implications":["If the decelerating bar is the source, the inner Galaxy's prograde rotation is still being built today, and the bulge and inner halo should gain angular momentum as long as the bar keeps braking.","The neural-network-selected stars should be predominantly bulge, halo, and thick-disk stars, a prediction that can be checked directly with elemental abundances and stellar ages.","The characteristic $v_\\phi(R)$ shape, rising to about 3 kpc, falling to the solar circle, and rising again in the outer halo, is a fingerprint that future astrometric and spectroscopic surveys can look for.","Bar deceleration of roughly 37.5% over 4 Gyr places the Milky Way's bar in the slow-bar regime, so resonance trapping should be visible as clustered angular-momentum gains for halo stars beyond the solar radius."],"supporting_citations":[{"why":"Supplies the neural network architecture, training procedure, and threshold convention used to select the rotating sample and to score the simulated particles.","marker":"Li et al. (2024b)"},{"why":"Provides the decelerating-bar scenario and the pattern-speed range that set the bar's initial and final pattern speeds.","marker":"Chiba & Schönrich (2021)"},{"why":"Provides updated parameters for the Made-to-Measure bar model that the decelerating bar is built from.","marker":"Sormani et al. (2022)"},{"why":"Supplies the equilibrium distribution-function model of the Milky Way used to generate the initial pseudo-stars and the axisymmetric potential.","marker":"Binney & Vasiliev (2023)"},{"why":"Supplies the truncated exponential disk and truncated bulge models that define the Galactic components and the bulge extent used in the simulation.","marker":"Binney & Vasiliev (2024)"},{"why":"Supplies the original Made-to-Measure bar model whose structural parameters the decelerating bar adopts and evolves.","marker":"Portail et al. (2017)"},{"why":"Constrains the present-day bar phase angle of 28 degrees used to align the bar model with observations.","marker":"Wegg et al. (2015)"},{"why":"Provides an earlier decelerating-bar simulation and resonance analysis that the angular-momentum-transfer interpretation builds on.","marker":"Li et al. (2024a)"}],"fun_headline_variants":["Bar slowdown drives rotation in Milky Way's bulge and halo","Decelerating bar transfers spin to inner Galaxy stars","Simulation: braking bar explains rotation of 1.18M stars","Milky Way's decelerating bar spins up bulge and halo","Neural network sample and simulation link bar braking to rotation"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole case for deceleration rests on the steadily rotating bar comparison run: if that control did not include the same factor-2.0 mass growth and factor-1.26 radial growth as the main run, then the difference between the two curves could be caused by bar growth rather than by the change in pattern speed, and the paper does not state that the control includes that growth.","fun_headline_variants_meta":{"raw":{"variants":["Bar slowdown drives rotation in Milky Way's bulge and halo","Decelerating bar transfers spin to inner Galaxy stars","Simulation: braking bar explains rotation of 1.18M stars","Milky Way's decelerating bar spins up bulge and halo","Neural network sample and simulation link bar braking to rotation"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00021,"raw_usage":{"total_tokens":1401,"prompt_tokens":926,"completion_tokens":475,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":542,"completion_tokens_details":{"reasoning_tokens":390}},"tokens_in":542,"tokens_out":475,"duration_ms":5479,"temperature":1.0,"reasoning_tokens":390,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T00:44:45.020857+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the control simulation with a bar whose pattern speed is constant at $-35\\,\\mathrm{km\\,s^{-1}}\\,\\mathrm{kpc}^{-1}$ but whose mass and radial extent grow exactly as in the decelerating run (mass $\\times2.0$, radius $\\times1.26$ over 4 Gyr). If that control reproduces the observed $v_\\phi(R)$ profile, the claim that deceleration is the cause fails; if it does not, deceleration is supported. An independent check would measure the present-day bar pattern speed from stellar kinematics and verify that it has fallen by roughly 37.5% over the last 4 Gyr, matching the assumed braking history.","supporting_citations":[],"review_version":1}