{"id":"2a556900-c40b-4acf-9630-58f1a6296068","arxiv_id":"2411.15307","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Large late-type spiral galaxies in dense nearby filaments sit in patches where the Planck CMB is colder than in any of 10,000 simulations, at 5.7 sigma.","lead":"This paper reports that the cosmic microwave background is significantly colder in discs around large nearby spiral galaxies than in any of 10,000 standard simulations, at 5.7 sigma. It argues that a new, unknown foreground, possibly linked to dark matter, cannot be ignored.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline p-value omits the by-eye selection of filament areas and the 0.2 Mpc profile radius from the look-elsewhere correction; a modest trial factor would erase p<0.0001.","rationale":"The paper has real strengths: 10,000 simulations, multiple component-separation methods, SEVEM PR4, WMAP frequencies, and explicit attempts at look-elsewhere correction. If the effect were purely a component-separation artifact, the consistency across methods and frequencies would be surprising. The concern is not that the effect is absent but that the quantitative significance is not yet established. Because the areas and radius are chosen from the same data and not included in the Monte Carlo null, p<0.0001 is an underestimate of the true false-positive probability. This is exactly the reader's weakest assumption, and I agree with it. A single rerun with automated region selection applied identically to data and simulations would settle it. The verdict should remain conditional on that check, so I leave the reader's CONDITIONAL verdict unchanged.","tokens_in":17385,"tokens_out":11726,"duration_ms":120227,"concrete_test":"Apply the identical region-selection algorithm to the data and to each of the 10,000 SMICA simulations: define dense-filament regions by thresholding the DisPerSE filament spine density (e.g., above the 90th percentile of linear K-band luminosity density), select the three disjoint circular apertures with maximal enclosed filament density, and scan the profile radius over [0.1, 0.5] Mpc and the ℓ-cut over {1,2,5,10}; for each map record the maximum |profile depth| over the same galaxy-sample grid as in §4.2. Recompute the LEE-corrected p-value from the resulting null histogram. If the Planck value remains more extreme than all 10,000 simulations, the by-eye selection is not the driver; if not, the reported 5.7σ and p<0.0001 are inflated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing gap is that the look-elsewhere correction does not randomize the data-informed definition of the test itself. Section 3.2 and Figures 5-7 correctly maximize over galaxy size, type, redshift, inside/outside the three areas, and ℓ-cut for each simulation, but the three dense-filament areas A, B, C are fixed by eye from the observed galaxy density field (Figure 4), and the profile bin is fixed at 0.2 Mpc (Section 3.1) because it is the innermost bin with the strongest signal. The paper itself notes in Section 5 that the circle positions/radii were chosen by eye; the stated check that shifting radii by ±5° keeps >5σ is a local stability test, not a correction for the number of alternative region configurations that could have been selected. Similarly, the 0.2 Mpc radius, the DisPerSE persistence threshold (4σ) and the 'above mean' linear K-band luminosity density cut used to define dense filaments are selection variables not varied in the null. Since the null distribution is built from simulations using the same fixed regions, it is too narrow. If the complete selection procedure were applied per simulation, the maximum-deviation distribution would widen; an effective trial factor of even ~10 would move the LEE-corrected p from 1e-4 to ~1e-3, eliminating the title's p<0.0001.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports a significant detection of a CMB temperature decrement in the inner 0.2 Mpc discs around large late-type spiral galaxies in three dense filament regions (A, B, C) selected from the 2MRS galaxy distribution. Using Planck SMICA and other foreground-cleaned maps against 10,000 simulated CMB maps, the authors find that the mean profile depth for the optimal galaxy sample is lower than in any simulation (empirical p<1e-4, quoted as 5.7 after Gaussian conversion), with reduced significance (3-4) after a look-elsewhere correction over galaxy sample parameters (size, type, redshift, inside/outside the regions, multipole cut). They also report a 99.99% correlation between the largest angular scales (l<16) of the Planck CMB and a model galaxy map. The authors interpret the signal as evidence for a new CMB foreground, possibly due to photon-dark-matter interactions in magnetized halos, and argue against known foregrounds via frequency consistency tests across Planck and WMAP.","tokens_in":17699,"tokens_out":13806,"duration_ms":120287,"significance":"If the detection is correct, it would be a major discovery: a kiloparsec-to-megaparsec-scale CMB temperature decrement around galaxies that is absent in Lambda-CDM simulations, implying new physics or an unidentified astrophysical foreground. The paper's strengths are its extensive simulation-based calibration (10,000 SMICA simulations; 600-1000 for other methods), an explicit look-elsewhere correction over many sample choices, multipole filtering, consistency across four foreground-cleaned maps and two experiments, and the frequency-flatness test. The central weakness is that the look-elsewhere correction does not include the data-informed selection of the three filament regions or the profile-bin radius, so the reported p-values are conditional on these choices. A modest trial factor of order 10 would shift the headline p from <1e-4 to ~1e-3, which is why the region-selection issue is the load-bearing point for the paper's main claim.","major_comments":[{"comment":"The look-elsewhere correction is incomplete because the three filament areas A, B, C and the 0.2 Mpc profile-bin radius were chosen by eye from the observed galaxy density field, and these choices are not randomized in the null distribution. The stability check in Section 5 (shifting the radii by +/-5 degrees keeps the significance above 5 sigma) is a local test, not a correction for the number of alternative region configurations that could have been selected; the null maximization in Figures 5-7 varies galaxy sample parameters but keeps the regions fixed. Since a modest trial factor of order 10 would shift the empirical p from <1e-4 to ~1e-3, the title's p<0.0001 and the claimed 3-4 sigma LEE-corrected detection are not established. Please either (i) perform a simulation-based LEE in which the filament-region selection procedure is applied independently to each simulation, or (ii) provide a quantitative upper bound on the number of independent region configurations and apply it to the reported p-values. The strong dependence on the Cold Spot region (Section 4.1), where excluding it reduces the total significance from 5.7 sigma to 4.4 sigma, further illustrates that the region definitions are data-informed and cannot be treated as fixed in the null.","section":"Section 3.2 and Section 5; Figures 5-7"},{"comment":"The low-multipole correlation claim ('more correlated with the distribution of nearby galaxies than 99.99% of simulated CMB maps') is computed with a fixed galaxy model map built from a pre-selected population of large late-type spirals (Section 3.3), and the null simulations use the same model map. The paper does not apply a look-elsewhere correction for the choice of galaxy sample, model profile radius, or tracer population, nor does it maximize over l_max within each simulation. As a result, the reported 1/10,000 p-value is conditional on these fixed choices and is not directly comparable to the profile-depth LEE-corrected significances. Please quantify the trial factor for the model-map parameter choices or demonstrate that the correlation is insensitive to them.","section":"Section 4.5 and Figure 10"},{"comment":"The total significance depends strongly on one of the three regions: excluding galaxies within 10 degrees of the CMB Cold Spot reduces the total significance from 5.7 sigma to 4.4 sigma. Because the Cold Spot has been previously associated with this same foreground by the authors (GL2024) and the regions are chosen by eye, the contribution of this region to the headline significance should be treated as a data-informed selection in the look-elsewhere calculation, not as an independent confirmation. The paper discloses this dependence, but it does not include the Cold Spot prior in the LEE, so the reported 5.7 sigma is not a blind detection.","section":"Section 4.1 and Table 1"}],"minor_comments":[{"comment":"Typos and formatting: 'Our aim her is' should be 'Our aim here is'; '10.000' should be '10,000'; 'look-elsewhere-e ffect' should be 'look-elsewhere effect' in Section 3.2 and elsewhere; and the footnote marker in Section 2 for the 2MRS catalogue is misplaced after '1'.","section":"Section 1 and throughout"},{"comment":"The notation '8/104' is ambiguous; please write '8/10^4' or '8/10,000' to avoid confusion with the integer 104. The same notation appears in Figures 3, 5, 6, 7, 8, and 10.","section":"Table 1"},{"comment":"The sigma values in parentheses are standard-deviation units from the simulation distribution, not empirical tail probabilities; the paper should state explicitly that the empirical p-value is the quoted fraction (e.g., 0/10,000 gives p<1e-4 one-sided) and that the Gaussian conversion (e.g., 5.7 sigma) is an extrapolation beyond the directly resolvable tail.","section":"Section 4.1 and Table 1"},{"comment":"Grammar and clarity: 'an averaged y-parameter' should be 'an average y-parameter'; also fix any spacing issues such as 'theSEVEM' in the same section.","section":"Section 4.3"},{"comment":"The chi-square minimization for estimating a_lm is described with an unnumbered equation and dense notation; a brief derivation or a reference to a standard pseudo-a_lm estimator would improve reproducibility.","section":"Section 3.3"},{"comment":"In the sentence 'We cannot exclude that the signal is also present in less dense areas, a weaker signal could be confused by the CMB fluctuations themselves.', insert 'as' or use a semicolon to fix the comma splice.","section":"Section 5"}],"recommendation":"major_revision","confidential_remarks":"The paper is a follow-up to the authors' own earlier works and makes a strong, potentially paradigm-changing claim. The statistical calibration is far more careful than in the earlier papers, and the frequency consistency checks are impressive. However, the incomplete look-elsewhere correction for the region selection is a genuine gap that affects the title claim; it is fixable with additional simulation-based LEE or a quantified trial factor. I recommend major revision rather than rejection, and the editor may wish to consider whether the title should be toned down until the LEE-corrected p-value is robust to the region-definition degrees of freedom."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know up front. First, this is not a fresh discovery; it is the same group's earlier negative CMB profile around nearby spirals (L2023, H2023, GL2024, C2024), re-analyzed with more careful statistics and a tighter focus. Second, the statistical care is real, but the headline significance has a hole: the three filament areas A, B, C and the 0.2 Mpc profile radius are fixed by eye from the data, and the look-elsewhere correction does not include those choices. The reported p<0.0001 depends on that correction, so the title overstates what is established.\n\nWhat the paper does well: it answers Addison's critique in a fair way. The χ2 over a 0.1–20° annulus does dilute a central signal, and the correlated bins make χ2 hard to interpret. Using the innermost 0.2 Mpc bin as a single statistic is a reasonable move, and the authors check that the result survives removing ℓ≤2, 5, and 10. The look-elsewhere correction over galaxy size, type, redshift, and inside/outside the three areas is a genuine improvement over the earlier papers. They also test with power-spectrum-matched simulations, multiple component-separation methods, and Planck/WMAP frequencies; the frequency stability is a meaningful point.\n\nThe soft spots are in proportion. The main one is the missing trial factor for the region definitions. Figure 4 selects the three dense filament regions by eye from the observed galaxy density, and the 0.2 Mpc radius is the innermost bin, where the signal is strongest by construction. The LEE histograms in Figures 5–7 randomize the sample parameters but not the regions or the bin radius. The ±5° stability test in Section 5 is local, not a global trial factor. The DisPerSE persistence threshold and the luminosity-density cut are also fixed. So the null distribution is narrower than the actual search. I don't think this kills the paper, but it means the honest statement is 'a few sigma, with systematics budget not fully accounted for,' not 'p<0.0001.'\n\nThe low-ℓ correlation has the same issue in milder form: they show lmax=5,7,10,16 and quote the 1/10000 for lmax=16 without correcting for the four choices. There is also mild circularity in using the H2023 model map, built from the same observed signal, to define the areas and as the galaxy template for the low-ℓ correlation.\n\nWho is this for? People working on CMB anomalies, the Cold Spot, or exotic photon–dark matter interactions. It deserves a serious referee, because the claim is important and the analysis is detailed enough to engage with. The referee should ask for an automated region/parameter search or a proper injection of those degrees of freedom into the null, plus code/data for the 10,000-simulation pipeline. Engage with it, but not on the title's terms.","headline":"A statistically careful re-analysis of the group's earlier CMB-cold-spot claim, but the headline significance omits the by-eye choice of filament areas and the 0.2 Mpc bin from the look-elsewhere correction, so the title's p<0.0001 is not established.","tokens_in":18266,"tokens_out":3505,"would_cite":false,"duration_ms":34343,"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":"Planck CMB temperatures dip around nearby spiral galaxies at 5.7σ, the paper argues.","keywords":["cosmic microwave background","galactic halos","CMB foreground","dark matter","spiral galaxies","cosmic filaments","profile depth","look-elsewhere effect"],"falsifier":"Repeat the analysis as a blind search: for each of the 10,000 simulated maps, scan over all three-region definitions (centers and radii), profile-bin radii from about 0.1 to 0.5 Mpc, galaxy size and redshift cuts, morphological types, and multipole cuts, and keep only the single most extreme profile depth per simulation. If the Planck data's maximum over this full search space is not more extreme than 99.99% of the simulated maxima, the claimed p<0.0001 detection is not established.","tokens_in":17206,"feed_emoji":"🌌","tokens_out":7943,"duration_ms":71444,"temperature":0.7,"pith_summary":"Using satellite CMB maps, this paper argues that the average cosmic microwave background temperature in discs just 0.2 megaparsecs around large, late-type spiral galaxies in dense cosmic filaments is lower than in any of 10,000 simulated Planck-like CMB skies. With the largest angular scales removed, the deficit reaches 5.7σ for galaxies in three dense-filament regions; even after penalizing the many galaxy-sample choices through a look-elsewhere correction, the authors report a 3-4σ effect. The largest angular scales (ℓ < 16) of the Planck CMB are also more correlated with the nearby galaxy distribution than 99.99% of simulated maps. If correct, the finding implies an unidentified foreground process that removes or cools CMB photons inside galactic halos, a phenomenon absent from the standard ΛCDM model.","feed_headline":"CMB dips around nearby spirals at 5.7σ in Planck data","feed_subtitle":"A 3-4σ signal survives all statistical corrections; if real, a new halo process hides in Planck data.","key_machinery":"The central statistic is the 'profile depth': the mean CMB temperature in the innermost disc of radius 0.2 Mpc around a galaxy, averaged over a chosen galaxy sample. This single number replaces the full radial temperature profile, avoiding correlated bins, and is compared between the Planck data and sets of simulated Planck-like CMB maps (up to 10,000 for the SMICA-cleaned maps). Significance is calibrated by the fraction of simulations with a more negative profile depth. To control the look-elsewhere effect, the authors repeat the sample optimization inside each simulation—varying galaxy size, morphological type, maximum redshift, environment, and multipole cut—and compare the data's best deviation with the distribution of best deviations in simulations. For the low-multipole analysis, the machinery is a correlation coefficient between normalized multipole maps of the CMB and a simple foreground model map of the nearby galaxy distribution, computed up to ℓmax=16.","core_discovery":"The paper's central claim is that the CMB temperature within the inner parts of galactic halos is not what the standard model predicts. Defining 'profile depth' as the mean CMB temperature in a 0.2 Mpc disc around each galaxy, the authors find that late-type spirals larger than 8.5 kpc and closer than z=0.02, particularly those in the three densest nearby filament regions, sit in a consistently negative temperature depression of about -27 µK once low multipoles are removed. None of 10,000 ΛCDM simulations reproduces this, and the significance stays at 3-4σ when the choice of galaxy type, size, redshift, environment, and multipole cut is optimized separately for every simulation. The same profile depth is nearly identical across frequencies from 41 to 217 GHz, consistent with a blackbody-shaped decrement, while elliptical galaxies show no significant inner signal. The authors conclude that a new CMB foreground component tied to spiral galaxies and their surroundings exists and that a physical interaction mechanism, possibly involving dark matter or intergalactic magnetic fields, should be sought.","pith_inferences":["A natural next test is to measure the same 0.2 Mpc profile depth in polarization maps; if the decrement comes from scattering or conversion of photons in magnetized halos, a polarization signature should accompany it.","The by-eye selection of regions A, B, and C is an unaccounted trial factor; a fully blind scan over region geometry could reduce the global significance below the reported 3-4σ post-correction level.","If rotation or magnetic fields are the trigger, the decrement should be stronger for high-spin or high-magnetization spirals matched in mass and environment; this is not tested in the paper.","Correlating the decrement with Faraday rotation measures or dispersion measures of the halos would discriminate between plasma and particle-conversion explanations."],"forward_implications":["A real halo-scale CMB foreground would change how the lowest CMB multipoles are interpreted, since part of the decrement survives down to ℓ>1024 while another part lives on scales ℓ<16.","The temperature decrement must be included in future measurements of cosmological parameters from low multipoles and CMB anomalies.","The lack of a significant inner decrement around elliptical galaxies points to a mechanism tied to spiral properties such as rotation or magnetic fields.","Frequency independence from 41 to 217 GHz constrains any photon–dark-matter interaction to preserve the blackbody spectrum in the microwave band.","The same statistic can be applied to the CMB Cold Spot and other reported anomalies, potentially explaining them by nearby large spiral groups."],"supporting_citations":[{"why":"Earlier paper that first reported significantly lower CMB temperatures in discs around nearby spiral galaxies; this paper reproduces and extends that profile result.","marker":"L2023"},{"why":"Disputed the earlier significance by including areas far beyond galactic halos; this paper argues that the dispute missed the inner-bin signal and corrects the methodology.","marker":"A2024"},{"why":"Provided evidence of correlation between galaxies and CMB cold spots and introduced the foreground model map used for the low-multipole correlation test.","marker":"H2023"},{"why":"Found the nearby Eridanus group of large spirals at the position of the CMB Cold Spot, linking that anomaly to the proposed foreground.","marker":"GL2024"},{"why":"Reported a 99.3% correlation between the local dark matter distribution and the CMB and a frequency-consistent signal, which the paper compares with its own profiles.","marker":"C2024"},{"why":"The 2MRS redshift catalogue providing galaxy positions, redshifts, K-band magnitudes, and morphological types that define the galaxy samples.","marker":"Huchra et al. 2012"},{"why":"Describes the four foreground-cleaning methods and the common mask used to build the Planck maps and to calibrate simulations.","marker":"Planck Collaboration, IV 2020"},{"why":"Supplies the filament-finding algorithm used to identify the dense cosmic filaments that define the regions A, B, and C.","marker":"Sousbie 2011"},{"why":"WMAP nine-year maps used to show the profile depth is frequency independent across 41-94 GHz.","marker":"Bennett et al. 2013"}],"fun_headline_variants":["CMB cools near spirals: 5.7σ anomaly in Planck data","Planck data show 27 µK dip in CMB around spiral halos","Spiral halos chill CMB at 5.7σ, defying ΛCDM","CMB colder near spirals: signal survives corrections at 3-4σ"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The assumed statistical model treats the three dense-filament regions A, B, and C—their centers and radii—and the 0.2 Mpc profile-bin radius as fixed rather than as degrees of freedom searched over; the look-elsewhere correction accounts for galaxy sample parameters but not for these by-eye choices.","fun_headline_variants_meta":{"raw":{"variants":["CMB cools near spirals: 5.7σ anomaly in Planck data","Planck data show 27 µK dip in CMB around spiral halos","Spiral halos chill CMB at 5.7σ, defying ΛCDM","CMB colder near spirals: signal survives corrections at 3-4σ"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000226,"raw_usage":{"total_tokens":1500,"prompt_tokens":1011,"completion_tokens":489,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":627,"completion_tokens_details":{"reasoning_tokens":399}},"tokens_in":627,"tokens_out":489,"duration_ms":5000,"temperature":1.0,"reasoning_tokens":399,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T14:26:16.091274+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the analysis as a blind search: for each of the 10,000 simulated maps, scan over all three-region definitions (centers and radii), profile-bin radii from about 0.1 to 0.5 Mpc, galaxy size and redshift cuts, morphological types, and multipole cuts, and keep only the single most extreme profile depth per simulation. If the Planck data's maximum over this full search space is not more extreme than 99.99% of the simulated maxima, the claimed p<0.0001 detection is not established.","supporting_citations":[{"cited_title":"2012, ApJS, 199, 2","cited_arxiv_id":null,"evidence_quote":"The 2MRS redshift catalogue providing galaxy positions, redshifts, K-band magnitudes, and morphological types that define the galaxy samples."},{"cited_title":"2013, ApJ, 208, 20","cited_arxiv_id":null,"evidence_quote":"WMAP nine-year maps used to show the profile depth is frequency independent across 41-94 GHz."}],"review_version":1}