{"id":"d13daec2-2650-45d3-a35c-14f7b84ea88a","arxiv_id":"2608.07209","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A dipole in the locally measured Hubble constant appears at 2-3 sigma in the lowest-redshift Pantheon+ supernova bins, points near the Shapley supercluster and CMB dipole, and disappears for higher redshift thresholds.","lead":"This paper measures how the local expansion rate of the universe, the Hubble constant H0, varies across the sky using type Ia supernovae, and finds a dipole pattern that is strongest for the nearest ones. The result suggests the local universe is not expanding uniformly in all directions, which matters for precision measurements of the Hubble constant and for the Hubble tension.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"MC-lcdm null mocks may under-represent cosmic-variance dipole from correlated peculiar velocities, biasing both p-values and the GLS covariance in the same direction; a ΛCDM velocity-field mock test is needed before the 2–3σ claim can be trusted.","rationale":"The reader's weakest assumption already flagged the faithfulness of the MC-lcdm mocks; I sharpen this to a concrete mechanism: the mocks likely omit the correlated peculiar-velocity field that dominates low-z Hubble-diagram scatter. The paper is otherwise careful: it truncates the Pantheon+ covariance correctly in Eq. (7), propagates errors through the full MCMC pipeline, uses the same mocks for both covariance and null, and includes a useful β estimator for dipole dominance. These strengths make the analysis reproducible and the central claim testable. However, the significance is entirely defined by the null distribution, and if the null is too narrow the 2–3σ headline could be spurious. The proposed test is decisive: including ΛCDM-consistent velocity correlations will either confirm the significance or reveal that the dipole is within cosmic variance, consistent with recent bulk-flow analyses. Because this is a resolvable empirical question and not a proven internal inconsistency, the appropriate verdict remains conditional on the mock test rather than an outright rejection. I partially agree with the reader because I identify the same component of the pipeline but with a sharper physical mechanism and a concrete test.","tokens_in":23496,"tokens_out":15060,"duration_ms":146517,"concrete_test":"Generate 1000 new null realizations that, in addition to Pantheon+ measurement noise, assign each SN a peculiar velocity drawn from a correlated Gaussian field with the ΛCDM linear power spectrum (or use the full Pantheon+ C_pecvel matrix if publicly available), recompute z_CMB accordingly, and rerun the full pipeline for z_min = 0.015 and 0.032. If the width of the mock A_dip distribution increases so that the observed A_dip falls below the 95th percentile, the reported 2–3σ significance is not robust. Also compare diagonal versus full peculiar-velocity mocks to isolate the effect.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim rests on the null distribution of A_dip from 1000 MC-lcdm realizations (Sec. III C, Eqs. 16–18, 21). These mocks add Gaussian magnitude noise with the Pantheon+ covariance to the same sky positions and CMB-frame redshifts, preserving survey selection and measurement correlations. However, they do not generate new large-scale-structure realizations. In the low-redshift regime (z<0.2), the dominant scatter in reconstructed H0 maps comes from peculiar velocities, which are spatially correlated over tens to hundreds of Mpc. If the Pantheon+ covariance's peculiar-velocity term is diagonal (a redshift-dependent error added in quadrature), as in the standard public release, the mocks will have a much narrower dipole-amplitude distribution than true ΛCDM realizations, which include the coherent velocity field's cosmic variance. The observed A_dip = 1.16 ± 0.28 km/s/Mpc at z_min = 0.015 would then appear artificially significant. Because the same mocks define both the null histogram (Fig. 5) and the covariance matrix Σ used in the GLS fit (Eq. 18), any such bias enters the p-value and the dipole uncertainty coherently. The paper itself attributes the dipole to a bulk flow (Sec. V), but the null mocks do not include the cosmic-variance dipole that ΛCDM permits. This is the weakest link between the mock construction and the headline significance.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reconstructs directional maps of the locally inferred Hubble constant from Pantheon+ Type Ia supernovae in the redshift interval z_min < z < 0.2, using HEALPix Nside=2 pixels with 75-degree spherical caps. A dipole is fitted to the 48-pixel H0 maps with generalized least squares, using a 48x48 covariance matrix estimated from 1000 Monte Carlo ΛCDM mocks. The authors report A_dip = 1.16 ± 0.28 km/s/Mpc at z_min = 0.015, a downward trend with increasing z_min, a dipole direction near the Shapley supercluster and CMB dipole, and claim a 2-3σ significance for z_min ≲ 0.032 that weakens at higher thresholds.","tokens_in":23858,"tokens_out":6310,"duration_ms":58536,"significance":"The paper addresses a timely and contested question: whether the local expansion rate is anisotropic. Its main strengths are the explicit treatment of the positive-definite nature of the dipole amplitude by comparing against an isotropic mock ensemble rather than A_dip = 0, and the careful propagation of the Pantheon+ covariance through the directional reconstruction into the patch covariance matrix. The tomographic z_min scan is a useful extension of earlier fixed-window analyses. If the mock-based significance is robust, the result would strengthen the case for a local bulk-flow-induced H0 dipole and would be relevant to the H0 tension debate. The analytical approximation for the patch covariance in Appendix A is a useful addition. However, the headline significance depends critically on the fidelity of the mock ensemble, which is not established.","major_comments":[{"comment":"The MC-lcdm mocks add Gaussian magnitude noise with the Pantheon+ covariance to the same sky positions and redshifts, but they do not generate new realizations of the large-scale velocity field. At z < 0.2, the dominant ΛCDM contribution to a dipole in the locally inferred H0 comes from coherent peculiar velocities, whose correlations extend over tens to hundreds of Mpc. If, as in the public Pantheon+ release, the peculiar-velocity term in the covariance is effectively diagonal, these mocks will under-represent the cosmic-variance dipole width. Because the same mocks define both the null distribution in Fig. 5 and the GLS covariance matrix Σ in Eq. (18), the p-values and the quoted dipole uncertainties are biased in the same direction, and the 2-3σ significance is not yet supported. I recommend adding mocks that include a ΛCDM velocity-field realization (e.g., linear-theory realizations or N-body mocks) or an analytic covariance that includes the velocity power spectrum, and showing how the significance changes.","section":"III C, Eqs. (16)-(18)"},{"comment":"The abstract states that A_dip decreases monotonically with z_min, but the values in Table I are not monotonic: A_dip increases from 0.87 at z_min = 0.022 to 0.91 at z_min = 0.025, and from 0.93 at z_min = 0.028 to 1.05 at z_min = 0.032. The overall trend is downward, but the non-monotonic features should be acknowledged and discussed, and the abstract should be rephrased accordingly.","section":"Abstract and IV A, Table I"},{"comment":"The claim of a \"2-3σ dipole pattern for z_min ≲ 0.032\" overstates the results. The conservative S estimator in Eq. (22) exceeds 2σ only at z_min = 0.015 (2.5σ); at z_min = 0.018 it is 1.9σ, and at z_min = 0.022 it is 1.4σ. Even with the p-value estimator, z_min = 0.028 gives 1.7σ and z_min = 0.035 gives 1.2σ. The significance statement should be restricted to the lowest redshift bin or presented bin-by-bin without the aggregate \"z_min ≲ 0.032\" wording.","section":"Abstract, IV A, Table I, Fig. 6"},{"comment":"The definition of the p-value is ambiguous: N_cross is described as \"the number of mocks crossing the red line\" in Fig. 5. It must be stated explicitly whether this is the number of mocks with A_dip ≥ A_obs, the number whose 1σ band crosses the observed line, or some other tail-count rule. A precise definition is needed for reproducibility, and the same sentence should clarify that the p-value and the S estimator are not independent because both derive from the same mock ensemble.","section":"III C, Eq. (21)"}],"minor_comments":[{"comment":"The text says the analysis uses \"ten equally spaced values\" of z_min in [0.015, 0.045], but the actual values in Table I are 0.015, 0.018, 0.022, 0.025, 0.028, 0.032, 0.035, 0.038, 0.042, 0.045, which are not equally spaced (differences alternate between 0.003 and 0.004). Please either use a regular grid or describe the chosen values as a set of ten specific thresholds.","section":"II, Table I"},{"comment":"Equation (22) defines S as a signal-to-noise-like ratio, but it is not a standard Gaussian significance because σ_obs and σ_MC are not fully independent and the null distribution is non-Gaussian. The text labels S as conservative, but a sentence explicitly stating that S is a heuristic estimator rather than a formal significance would avoid misinterpretation.","section":"III B, Eq. (22)"},{"comment":"The denominator in Eq. (25) is written as σ_{ΔH0/2}, which is ambiguous because the subscript is a ratio of two quantities. Please define it as the propagated uncertainty on ΔH0^{max}/2, for example σ_{Δ/2}, and state explicitly that the covariance between ΔH0^{max}/2 and A_dip is neglected, as the footnote already indicates.","section":"IV D, Eq. (25)"},{"comment":"The caption of Fig. 5 states that the red line and band are \"read from the fourth column in Table I,\" but it would be clearer to state that the band represents the 1σ uncertainty on A_dip and that the histograms are normalized probability densities.","section":"Fig. 5"}],"recommendation":"major_revision","confidential_remarks":"The central result is interesting and potentially important, but the significance estimate is not yet robust because the mock ensemble does not demonstrably include the cosmic-variance dipole from correlated peculiar velocities. The paper alludes to a companion paper with alternative mock strategies; if that material is available, it should be incorporated here or the present claims should be downscaled. The abstract also overstates both the monotonicity and the significance. I would not recommend rejection if the mock issue is addressed, but the current version needs substantial revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does something genuinely useful: it scans ten lower redshift thresholds and shows the H0 dipole amplitude and significance drop as low-z SNe are removed. That tomographic behavior is the real contribution, along with a mock-based covariance matrix and a beta test confirming the signal is dipole-dominated. The authors also handle the positive-definite dipole correctly, comparing against isotropic mocks rather than naively testing A_dip=0, and the appendix on covariance propagation is a nice touch.\n\nThe soft spot is the mock construction. The MC-lcdm realizations add Gaussian magnitude noise with the Pantheon+ covariance to the same sky positions and redshifts, but they do not generate new large-scale structure velocity fields. In a real ΛCDM universe each realization has a different bulk flow, and that bulk flow produces a dipole in low-redshift H0 maps. If the Pantheon+ covariance only contains diagonal peculiar-velocity errors, these mocks will have a narrower dipole-amplitude distribution than true ΛCDM, overstating the p-value and also underestimating the GLS covariance matrix. The paper itself attributes the signal to a bulk flow, but the null mocks do not include the bulk-flow cosmic variance. That is a load-bearing issue, not a technicality. I am not convinced the 2–3σ claim survives a null that properly marginalizes over coherent velocity realizations.\n\nTwo smaller issues. The abstract says the dipole amplitude decreases monotonically, but Table I shows a bump: A_dip goes 0.87, 0.91, 0.93, 1.05 between z_min=0.022 and 0.032. That is not monotonic. Also, the significance of the lowest z_min bin is quoted without a trials correction for the ten thresholds; a fluke somewhere is more likely than the raw p-value suggests.\n\nWhere the paper is solid: the analysis is careful, the handling of correlated patches is deliberate, and the literature comparison is thorough. The authors are honest about the limitations, including the inability to trace the signal to Shapley with current data.\n\nMy recommendation: send this to peer review, but make the mock construction the central referee question. The authors should either generate mocks with a realized velocity field or argue convincingly that the Pantheon+ covariance already captures the relevant cosmic variance. The abstract and trials issue also need fixing. This is a conditional accept paper, not a desk reject.","headline":"A careful Pantheon+ dipole analysis with a novel z_min scan, but the significance hinges on mocks that may miss cosmic-variance bulk flows; worth refereeing with mandatory revisions.","tokens_in":24377,"tokens_out":4333,"would_cite":false,"duration_ms":49886,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["98.80.Es"],"model":"deepseek-v4-flash","headline":"This paper claims the locally measured Hubble constant carries a directional dipole of 1.16 ± 0.28 km/s/Mpc in the lowest-redshift Pantheon+ supernovae, a signal that disappears once supernovae below z ≈ 0.038 are excluded.","keywords":["Hubble constant","H0 dipole","Pantheon+ supernovae","cosmological principle","anisotropic expansion","bulk flows","low-redshift universe","type Ia supernovae"],"falsifier":"Recompute the $z_{\\rm min} = 0.015$ dipole using a differently built null — shuffling supernova magnitudes among fixed sky positions instead of adding Gaussian noise, or using peculiar-velocity-corrected redshifts — and check whether the p-value stays below 0.001; if it does not, the signal is an artifact of the mock prescription or the redshift frame. A second check: the fitted direction near $(l, b) \\approx (307°, 61°)$ predicts a coherent bulk flow of order a few hundred km/s over $z < 0.05$, so its absence in galaxy peculiar-velocity catalogues over that volume would contradict the local-structure interpretation.","tokens_in":23337,"feed_emoji":"🌌","tokens_out":22624,"duration_ms":173857,"temperature":0.7,"pith_summary":"The paper asks whether the cosmic expansion rate measured close to us, the Hubble constant $H_0$, is the same in every direction of the sky, and argues that at the lowest redshifts it is not. Using the Pantheon+ catalogue of Type Ia supernovae, the authors divide the sky into 48 equal-area patches (a HEALPix grid), fit an expansion rate in each patch from the supernovae within a 75° circle around it, and then fit a dipole — a pattern that is higher along one sky axis and lower along the opposite one — to the resulting map. They find a dipole amplitude $A_{\\rm dip} = 1.16 \\pm 0.28$ km/s/Mpc when the sample starts at $z_{\\rm min} = 0.015$, falling monotonically to $0.35 \\pm 0.52$ km/s/Mpc, consistent with zero, at $z_{\\rm min} = 0.045$. Because even an isotropic universe yields a nonzero positive amplitude by construction, significance is judged against 1,000 mock skies built from a ΛCDM model with the same sky positions, redshifts, and noise correlations; that comparison gives 2–3σ for $z_{\\rm min} \\lesssim 0.032$, and the dipole points within about 30° of both the Shapley supercluster and the CMB dipole. The authors conclude that the $H_0$ dipole is a low-redshift feature, plausibly tied to local bulk flows rather than to cosmology on large scales.","feed_headline":"2–3σ dipole found in local Hubble rate from Pantheon+ supernovae","feed_subtitle":"The signal peaks near 1.2 km/s/Mpc for the closest supernovae, points toward Shapley, and vanishes once they are cut","key_machinery":"The load-bearing device is the reconstructed sky map of $H_0$ combined with a dipole fit on top of it. The sky is discretized into 48 equal-area HEALPix pixels, and for each pixel a spherical cap of 75° radius collects the supernovae whose distance moduli are fit, through Monte Carlo sampling, to a flat ΛCDM model while the 77 Cepheid-calibrated host supernovae fix the absolute magnitude; this yields a value $H_0(\\hat n)$ with an uncertainty for every pixel. The 48 values are then fit by generalized least squares to the dipole ansatz $H_0(\\hat n) = H_0^{\\rm mono} + \\mathbf{D}\\cdot\\hat n$, where $\\mathbf{D}$ is the dipole vector. The essential supporting object is the 48×48 covariance matrix $\\Sigma$, not taken from the Pantheon+ catalogue directly but estimated from 1,000 MC-lcdm realizations: mock skies that keep the real supernovae's sky positions and redshifts, add Gaussian magnitude noise with the full Pantheon+ covariance matrix, and run the entire map-building pipeline. From this ensemble the authors get both $\\Sigma$, which the dipole fit needs, and the null distribution of $A_{\\rm dip}$ against which significance is read. Significance is quoted two ways — a p-value counting how many mocks exceed the observed amplitude, and a more conservative estimator $S$ that also propagates the quoted error — while a third estimator, $\\Delta H_0^{\\max}$, records the largest antipodal contrast between opposite patches, with $\\beta$ quantifying whether that contrast is consistent with the fitted dipole alone.","core_discovery":"On its own terms, the paper establishes that the local expansion rate reconstructed from Pantheon+ supernovae is anisotropic, and that the anisotropy is genuinely dipolar at low redshift. With maps built from overlapping 75° caps on a 48-pixel HEALPix grid and with redshifts in the CMB rest frame, the best-fit monopole-plus-dipole model reads $A_{\\rm dip} = 1.16 \\pm 0.28$ km/s/Mpc at $z_{\\rm min} = 0.015$ and decreases monotonically as the lower redshift cut is raised, reaching $0.35 \\pm 0.52$ km/s/Mpc at $z_{\\rm min} = 0.045$. The paper does not read significance from the error bar alone, since a positive-definite amplitude is nonzero even under isotropy; instead the null distribution comes from 1,000 MC-lcdm realizations — mocks that keep the real sky positions and redshifts, add Gaussian magnitude noise with the full Pantheon+ covariance, and run the whole reconstruction pipeline. That comparison gives $p < 0.001$ (more than 3.1σ) at the lowest bin with the p-value estimator and 2.5σ with the more conservative $S$ estimator, and 2–3σ overall for $z_{\\rm min} \\lesssim 0.032$. For every threshold where the signal is significant, the direction clusters near the Shapley supercluster and the CMB dipole, and the amplitude becomes statistically indistinguishable from isotropy for $z_{\\rm min} \\geq 0.038$ — the redshift where Shapley itself begins. Residual antipodal contrast $\\Delta H_0^{\\max}$ remains nonzero at all thresholds, and because the $\\beta$ estimator stays below 1, the authors read this as a dipole the data no longer constrains rather than as evidence for higher multipoles.","pith_inferences":["A direct test the authors mention but do not run: repeat the pipeline with peculiar-velocity-corrected redshifts ($z_{\\rm HD}$) instead of CMB-frame redshifts; a surviving dipole would confirm it is a real velocity-field feature, while its disappearance would indict the redshift frame.","The quoted 2–3σ significance is conditional on the Gaussian-noise mock prescription; a null built by shuffling magnitudes among fixed sky positions, or one injecting simulated peculiar velocities, could shift the p-values either way — a comparison the authors defer to a companion paper.","Masking the Shapley region in the $z_{\\rm min} = 0.015$ sample is a cheap, decisive experiment: a surviving dipole would implicate a broader bulk flow, while its vanishing would pin the anisotropy on the supercluster itself.","Future, larger low-redshift supernova samples should sharpen the dipole if it is physical; if its significance instead shrinks as more nearby supernovae are added, the present signal was a small-sample fluctuation."],"forward_implications":["At $z_{\\rm min} = 0.015$ the expansion rate inferred from opposite patches of sky differs by up to about 3.8 km/s/Mpc, so the local $H_0$ is direction-dependent at a level current data can see.","Raising the lower redshift cut to $z_{\\rm min} = 0.038$ erases the dipole, placing whatever produces it inside the very nearby universe, below redshift $\\approx 0.038$.","Where the dipole is significant, its direction is stable across $z_{\\rm min}$ cuts and points within about 30° of both the Shapley supercluster and the CMB dipole, tying the signal to known large-scale structure and bulk flows.","Because the anisotropy is dipole-dominated at low redshift ($\\beta < 1$) while $\\Delta H_0^{\\max}$ stays nonzero where the dipole has faded, deeper data are needed to decide between an unconstrained dipole and higher multipoles.","If the local $H_0$ genuinely varies with direction, sub-percent $H_0$ values derived from low-redshift supernovae carry an angular selection effect that bears on the Hubble tension."],"supporting_citations":[{"why":"Supplies the Pantheon+ supernova catalogue and its full covariance matrix, the input data for every sky map and every mock realization.","marker":"[3, 4]"},{"why":"The prior overlapping-cap dipole detection toward Shapley whose method this paper adapts and extends to redshift tomography.","marker":"[16]"},{"why":"The hemisphere and antipodal $\\Delta H_0^{\\max}$ estimators and Cepheid-calibration treatments that the paper's third estimator compares against.","marker":"[11, 13, 34]"},{"why":"Defines the HEALPix equal-area pixelization used to lay out the 48 sky patches.","marker":"[62]"},{"why":"Gives the redshift extent of the Shapley supercluster, the scale where the dipole signal disappears.","marker":"[65]"},{"why":"Provides the supercluster's central coordinates against which the fitted dipole directions are measured.","marker":"[66]"},{"why":"The earlier dipole analysis of the Pantheon+ sample whose decreasing-significance-with-distance trend the paper's tomography mirrors.","marker":"[24]"},{"why":"Galaxy peculiar-velocity catalogue measurements cited to argue the dipole is a local velocity-field effect rather than a cosmological one.","marker":"[50, 67]"},{"why":"Underpins the claim that FLRW consistency fixes $H_0$ as an integration constant, so the null hypothesis for $\\Delta H_0^{\\max}$ is zero.","marker":"[68]"}],"fun_headline_variants":["H0 dipole in Pantheon+ supernovae weakens with redshift","2-3σ H0 dipole from Pantheon+ points toward Shapley supercluster","Low-redshift H0 anisotropy vanishes at z≈0.032","Pantheon+ supernovae reveal dipole in local Hubble rate","H0 dipole amplitude drops from 1.16 to 0.35 as redshift rises"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the 1,000 mock skies used as the isotropic comparison — realizations that keep the real supernovae's sky positions and redshifts and add Gaussian magnitude noise with the Pantheon+ covariance matrix — faithfully represent what an isotropic universe with the same selection and correlations would look like; if those mocks misrepresent the noise, both the p-values and the dipole covariance matrix are biased in the same direction.","fun_headline_variants_meta":{"raw":{"variants":["H0 dipole in Pantheon+ supernovae weakens with redshift","2-3σ H0 dipole from Pantheon+ points toward Shapley supercluster","Low-redshift H0 anisotropy vanishes at z≈0.032","Pantheon+ supernovae reveal dipole in local Hubble rate","H0 dipole amplitude drops from 1.16 to 0.35 as redshift rises"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000884,"raw_usage":{"total_tokens":4014,"prompt_tokens":1334,"completion_tokens":2680,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":950,"completion_tokens_details":{"reasoning_tokens":2579}},"tokens_in":950,"tokens_out":2680,"duration_ms":18878,"temperature":1.0,"reasoning_tokens":2579,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T12:47:50.305423+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the $z_{\\rm min} = 0.015$ dipole using a differently built null — shuffling supernova magnitudes among fixed sky positions instead of adding Gaussian noise, or using peculiar-velocity-corrected redshifts — and check whether the p-value stays below 0.001; if it does not, the signal is an artifact of the mock prescription or the redshift frame. A second check: the fitted direction near $(l, b) \\approx (307°, 61°)$ predicts a coherent bulk flow of order a few hundred km/s over $z < 0.05$, so its absence in galaxy peculiar-velocity catalogues over that volume would contradict the local-structure interpretation.","supporting_citations":[{"cited_title":"The Shapley Supercluster. II. Spectroscopic observations in a wide area and general morphology","cited_arxiv_id":"astro-ph/0007210","evidence_quote":"Gives the redshift extent of the Shapley supercluster, the scale where the dipole signal disappears."},{"cited_title":"Structure and dynamics of the Shapley Supercluster","cited_arxiv_id":"astro-ph/0509903","evidence_quote":"Provides the supercluster's central coordinates against which the fitted dipole directions are measured."}],"review_version":1}