{"id":"5565a237-498b-41ea-99de-b4ba6c78efc3","arxiv_id":"2608.06480","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A joint analysis of Planck and ACT CMB polarization finds a nonzero cosmic birefringence angle of 0.277° ± 0.057°, at 4.8σ, with systematics still to be understood.","lead":"This paper combines microwave maps from two telescopes, ACT and Planck, to measure a subtle rotation in the polarization of the cosmic microwave background. The joint analysis finds a rotation angle of 0.277 degrees, 4.8 standard deviations away from zero, but the authors caution that unresolved systematics remain.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 4.8σ and 3.5σ results both depend on the zero-mean, 90%-correlated ACT miscalibration priors; a global offset in α_i is nearly degenerate with β and is not tested by the paper's robustness checks.","rationale":"The reader's weakest_assumption identifies exactly the assumption that carries the most weight in this analysis. The joint analysis is the first to combine ACT and Planck for cosmic birefringence, and the cross-experiment block adds genuinely new constraining power. However, the ACT leg has no internal handle on the α+β degeneracy: it measures only the sum of the miscalibration angle and the cosmic rotation, and the prior is the sole mechanism that separates them. The paper's own robustness test masks the Planck sky and removes foreground-dominated Planck channels, but it does not alter the ACT miscalibration priors, so the 3.5σ robustness result still depends on the same prior. The manuscript is honest about this dependence, including the caveat in the abstract and the acknowledgment that confirming the measurement requires better understanding of both datasets. This does not invalidate the paper, but it does mean the central claim should be read as conditional on the accuracy of the ACT optics-model calibration priors. The proposed global-offset test would directly quantify whether the prior mean is load-bearing. Since the reader already assigned a CONDITIONAL verdict and correctly identified this assumption, no verdict change is needed.","tokens_in":11618,"tokens_out":5900,"duration_ms":56284,"concrete_test":"Re-run the baseline joint likelihood with an additional free global offset δ added to all ACT α_i (equivalently, replace the zero mean of the ACT prior with an unknown common mean), assigning δ a conservative prior such as N(0, 0.2°) or a uniform prior of ±0.3°, while preserving the quoted 90% intra-array correlations. If the marginalized posterior for β shifts by more than ~0.05° or its significance drops below ~3σ, the reported detection is not independent of the zero-mean prior assumption; if β remains ~4σ away from zero with a similar central value, the concern is bounded and the face-value result is less vulnerable.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that β = 0.277° ± 0.057° (4.8σ) is a robust detection rests on the accuracy of the Gaussian priors on the ACT miscalibration angles α_i, with zero means, σ = 0.09°–0.11°, and 90% correlation within each polarization array. Because ACT observes only ℓ > 600 over f_sky = 0.25 with the Galactic plane masked, the ACT×ACT and ACT×Planck spectra measure ψ_i = α_i + β. A common-mode offset δ applied to all ACT α_i is exactly degenerate with β: the data can only constrain the combination α_i + β. The Planck-leg foreground constraints do not break this degeneracy for the ACT leg, since the cross-spectra at ℓ > 600 are CMB-dominated and the ACT maps are foreground-cleaned. The paper's robustness test (30% Galactic cut, b_min = 600.5 for Planck, and removal of foreground-dominated Planck bands) leaves the ACT leg and its α priors unchanged, so the quoted 3.5σ significance inherits the same prior dependence. The only test that removes the ACT priors also removes the Planck priors, and the resulting β = 0.442° ± 0.098° is dominated by the Planck-only foreground-based constraint, so it does not isolate the ACT leg. The manuscript itself flags that the ACT team reported an unexplained polarization-angle discrepancy for PA5 f090 and PA5 f150, and the abstract explicitly conditions the headline result on 'taking the reported instrumental priors at face value.' A biased prior mean or underestimated width would therefore translate almost one-to-one into a biased β, directly affecting both the 4.8σ and 3.5σ claims.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents a joint analysis of Planck DR4 and ACT DR6 polarization maps to measure the isotropic cosmic birefringence angle β. The author constructs ACT×Planck cross-power spectra, internal ACT and Planck spectra, and a joint likelihood that includes Gaussian priors on the 17 instrument miscalibration angles α_i and a filamentary dust EB model with three fitted amplitudes. The baseline result is β = 0.277° ± 0.057° (4.8σ), and a robustness test that masks the Galactic plane for Planck and removes foreground-dominated bands yields β = 0.236° ± 0.067° (3.5σ). The analysis is validated against published ACT DR6 spectra and earlier Planck-only birefringence analyses. The paper explicitly conditions the headline significance on the reported instrumental priors and acknowledges unresolved systematics in both datasets.","tokens_in":11940,"tokens_out":5648,"duration_ms":53692,"significance":"If the result is correct, this would be the first joint, multi-experiment detection of isotropic cosmic birefringence, and it would substantially strengthen the case for a parity-violating extension beyond ΛCDM. The paper's strengths are its reproducible pipeline (public code), its validation against official released spectra, the use of full covariance information including cross-experiment blocks, and the explicit identification of the assumptions on which the detection rests. The significance is real but conditional: the central number and its error are heavily dependent on the accuracy of the ACT miscalibration priors and on the dust EB model, so the detection is not yet at the level of a stand-alone, assumption-free measurement.","major_comments":[{"comment":"The ACT data probe only the combination ψ_i = α_i + β in the ACT×ACT and ACT×Planck spectra. A common-mode shift δ applied to all ACT α_i is therefore exactly degenerate with β on the ACT leg, and the Planck foreground constraints calibrate the Planck-leg angles, not the ACT angles. The robustness tests in Sec. III change the Galactic mask, b_min, and the foreground-dominated Planck bands, but they keep the ACT prior means fixed at zero; the test that removes all α priors returns β = 0.442° ± 0.098°, which is dominated by the Planck-only foreground-based constraint and does not isolate the ACT leg. Consequently, the abstract's statement that recovering β = 0 would require both the dust EB model and the instrumental priors to fail simultaneously is stronger than the analysis supports: a single common-mode error in the ACT α_i prior means would shift β without any dust-model failure. Please add a sensitivity test that either fits a common offset δ_ACT with a wide prior or shifts all ACT α_i means by ± their stated σ values, and report the resulting β and significance.","section":"Sec. II (Data and Method), Gaussian priors on ACT α_i"},{"comment":"The ACT×ACT block has PTE = 0.1%, which the paper attributes to the known poor fit around ℓ ≈ 1800–1900 reported by the ACT team. Because the baseline joint result includes this block, and because the ACT×ACT block is the part of the data most dependent on the α_i priors, this poor fit is not purely cosmetic: it signals unmodeled systematics in the very spectra that carry the prior-based constraint. Please quantify the sensitivity of β to this feature, for example by repeating the fit with the ℓ ≈ 1800–1900 bins removed or with the ACT×ACT block excluded altogether, and report the resulting β and significance. Without such a test, the 4.8σ detection remains vulnerable to the unresolved ACT systematics that the manuscript itself acknowledges.","section":"Sec. III (Results), Fig. 3 ACT×ACT panel"},{"comment":"The three dust EB amplitudes A_l are sampled with flat positive priors from the same data that constrain β, and the Planck low-ℓ foreground-based calibration is obtained in the presence of this freedom. The high-ℓ Galactic-cut robustness test reduces the foreground dependence of the Planck leg, but in that configuration the result is mostly determined by the ACT priors (Sec. III). Thus the two legs of the degeneracy breaking are each validated only in a regime that leans on the other. Please report the posterior correlation between β and the A_l parameters, and show how β changes when the A_l prior amplitudes are scaled (e.g., by 0, 0.5, and 2), so the reader can see that the detection does not trade off against the dust EB amplitude.","section":"Sec. II (Data and Method), dust EB parameters A_l"}],"minor_comments":[{"comment":"The y-axis labels read \"DEB [ K2]\"; the units are presumably μK^2 and should be typeset consistently with the text.","section":"Fig. 3"},{"comment":"The sentence \"Ref. [36] reported pre-flight an upper limit\" is grammatically awkward and should read \"reported a pre-flight upper limit\".","section":"Sec. III"},{"comment":"Reference [17] lacks a year in the citation string, and reference [40] lists the author as \"A. Collaboration\" rather than \"Astropy Collaboration\".","section":"References"},{"comment":"The notation \"E PA6 f150B100A\" and similar in Sec. III would benefit from a brief explanation or a notational table, since the reader must infer that this denotes an E-mode of PA6 f150 crossed with a B-mode of the Planck 100 GHz split.","section":"Sec. II"},{"comment":"The sentence \"Refs. [15, 19] analyzed instrumental systematic effects on β in Planck but could not explain the signal\" would be clearer if it distinguished between the two references' specific results, since Ref. [19] is a Cosmoglobe analysis and Ref. [15] is a Planck NPIPE analysis.","section":"Sec. IV"}],"recommendation":"major_revision","confidential_remarks":"The headline number is conditioned on priors taken from Ref. [25], which is currently an arXiv preprint, and on an ACT calibration model for which the ACT team itself reports an unexplained polarization-angle discrepancy for PA5 f090 and PA5 f150. The editor may wish to confirm that Ref. [25] has been peer-reviewed and that the ACT calibration values have not changed between the preprint and the final ACT DR6 release before accepting the paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First joint ACT+Planck isotropic birefringence analysis. β = 0.277° ± 0.057°, 4.8σ, with a 3.5σ dust-mitigated variant. The paper is careful, reproducible, and unusually honest about systematics; it deserves a serious referee.\n\nWhat's actually new: the ACT×Planck cross-power spectra, the combined likelihood over 17 miscalibration angles, and a public pipeline that validates well against released ACT spectra and prior Planck analyses. The 7% covariance underestimate in the ACT block is quantified and shown to shift β negligibly. The consistency of positive β across ACT×ACT, ACT×Planck, and Planck×Planck blocks is a real point in favor.\n\nThe soft spot is the one the stress-test note pins: ACT observes ℓ > 600 with f_sky = 0.25 and a Galactic cut, so its data only constrain ψ_i = α_i + β. The zero-mean, 90%-correlated Gaussian priors on α_i are what separate β from α. A common-mode offset in those priors is nearly exactly degenerate with β. The 3.5σ robustness test leaves the ACT leg and its priors unchanged, so it tests the dust model, not the prior assumption. The paper actually says this: the headline result is conditional on 'taking the reported instrumental priors at face value.' And the team's own reported PA5 f090/f150 angle discrepancy is still unexplained.\n\nOther issues are minor by comparison: the ACT×ACT block has PTE=0.1%, and the Planck-only and ACT-only β values are in 1.5σ tension. Neither is fatal, but both deserve a paragraph in a revision.\n\nMy take: this is a solid measurement paper, not a settled cosmological detection. The 4.8σ is real given the priors, but the priors do a lot of work. The fix is straightforward—quantify how much the ACT prior means would have to shift to erase the signal, and show the cross-spectra alone without the ACT×ACT block. I would not desk-reject this. Send it to a good referee; the authors can address the prior question in revision.","headline":"A credible, well-executed joint ACT+Planck birefringence analysis whose headline 4.8σ rests on ACT miscalibration priors that are degenerate with β—worth refereeing, with the prior question front and center.","tokens_in":12548,"tokens_out":2431,"would_cite":true,"duration_ms":20505,"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":"Cosmic birefringence seen at 4.8 sigma in CMB maps","keywords":["cosmic birefringence","CMB polarization","E-B cross-correlation","axion-like fields","Chern-Simons coupling","ACT DR6","Planck DR4","parity violation"],"falsifier":"Measure the ACT polarization-angle miscalibration directly from observations of a bright polarized astrophysical calibrator with known polarization angle and re-run the joint fit; if the independently calibrated α_i shift the best-fit β by more than about 0.2°, the reported 4.8σ detection is a product of the priors rather than cosmology, and if they confirm the priors the detection survives this test.","tokens_in":11352,"feed_emoji":"🌌","tokens_out":7843,"duration_ms":62969,"temperature":0.7,"pith_summary":"This paper claims that the polarization of the cosmic microwave background is rotated by a nonzero angle β = 0.277° ± 0.057° as it travels to us, a parity-violating effect known as cosmic birefringence that ΛCDM does not predict. The claim is based on a joint analysis of the Planck DR4 and ACT DR6 polarized maps, combining their auto- and cross-power spectra in a single likelihood. Excluding β = 0 at 4.8σ in the baseline fit, the paper shows the signal survives at 3.5σ when the Galactic-plane-dominated Planck channels are masked and removed. If the result holds, it would be the first statistically robust detection of isotropic cosmic birefringence from CMB data and would point to an ultra-light axion-like field coupling to electromagnetism. The paper itself stresses that unresolved systematics must be ruled out before drawing strong cosmological conclusions.","feed_headline":"Cosmic birefringence seen at 4.8 sigma in CMB maps","feed_subtitle":"A joint ACT+Planck analysis finds a parity-breaking rotation of CMB polarization that standard cosmology cannot explain.","key_machinery":"The central mechanism is the rotation of the polarization angle: cosmic birefringence with angle β mixes the intrinsic E and B modes of the CMB, producing a nonzero observed EB power spectrum whose shape follows $C^{EB,o}_\\ell = \\frac{\\tan(4\\beta)}{2}\\left(C^{EE,o}_\\ell - C^{BB,o}_\\ell\\right)$ when the intrinsic CMB EB correlation is zero. Because instrumental miscalibration angles α enter the same combination α+β for the CMB but only α for foregrounds, the analysis breaks the degeneracy using two independent handles: polarized Galactic foreground emission for Planck and Gaussian priors on α from an optics model for ACT, whose α+β is otherwise almost unconstrained since ACT masks the Galaxy and probes ℓ>600. The likelihood jointly fits β, 17 α's, and three dust-EB amplitudes, with a cross-telescope covariance built from analytic mode-coupling covariance and corrections for ground pickup and temperature-to-polarization leakage.","core_discovery":"On its own terms, the paper establishes that the observed E-B cross-correlation of the CMB, when fit jointly over ACT×ACT, ACT×Planck, and Planck×Planck spectra with 17 miscalibration angles and a filamentary dust EB model, prefers β = 0.277° ± 0.057° over β = 0 at 4.8σ. Each of the three spectrum blocks separately shows a positive EB signal, with the ACT×Planck block contributing β = 0.178° ± 0.078° (2.3σ), the ACT-only block β = 0.207° ± 0.073°, and Planck-only β = 0.387° ± 0.094°, in mild tension with the ACT legs. Masking Planck with a 30% Galactic cut and removing foreground-dominated bands yields β = 0.236° ± 0.067° (3.5σ), which the paper interprets as evidence that β = 0 can only be recovered if both the dust EB model and the instrument calibration priors fail simultaneously. The paper presents the significance of the detection as growing over the past decade of measurements and frames the result as consistent with an axion-like field coupled through a Chern-Simons term, while cautioning that unresolved systematics in both experiments cannot yet be excluded.","pith_inferences":["If the detection holds up, comparing β measured at low multipoles (foreground-anchored Planck) with β at high multipoles (ACT leg) offers a direct test of whether the rotation is truly frequency-independent and scale-independent, since the two legs break the α+β degeneracy differently.","A decisive check would come from an in-flight measurement of the ACT polarization-angle miscalibration using a polarized calibrator; the paper's ACT leg leans almost entirely on priors derived from optics modeling, so a direct calibration could either remove or confirm the main systematic vulnerability.","The mild 1.5σ tension between Planck's higher β and ACT's lower β, if not a statistical fluctuation, could indicate residual polarized dust in Planck or a small instrumental rotation in ACT; future higher-sensitivity CMB maps with more sky overlap could discriminate between these by measuring the EB spectrum in narrow multipole bins."],"forward_implications":["If the detection is real, this is the strongest evidence to date for isotropic cosmic birefringence, more significant than the previous Planck+WMAP measurement of 3.6σ.","The positive EB signal in ACT×Planck cross-spectra (2.3σ) shows the effect is not an internal artifact of either telescope's maps, since the two instruments have uncorrelated noise.","β=0 would require both the dust EB foreground model and the instrument calibration priors to be wrong at the same time, as argued from the robustness test.","A cosmological source of the rotation implies a parity-violating coupling of light to an ultra-light pseudoscalar field, opening a window onto the dark sector beyond ΛCDM.","The paper notes the signal is consistent with being frequency independent, the expected signature of a Chern-Simons coupling."],"supporting_citations":[{"why":"Developed the method of using polarized Galactic foreground emission to calibrate the instrument angle, anchoring the Planck leg.","marker":"[11]"},{"why":"Generalized that method into the likelihood framework the paper applies.","marker":"[13]"},{"why":"Provided the joint Planck-WMAP analysis with β=0.342° and the likelihood matrices A_b, B_b and dust model that the paper extends.","marker":"[18]"},{"why":"Produced the Planck DR4 NPIPE maps and estimated the bias from HFI systematics at about 0.009°.","marker":"[15]"},{"why":"Reports the Planck-only measurement β=0.387°±0.094° used as the Planck baseline and source of the 30% Galactic-cut mask.","marker":"[16]"},{"why":"Released ACT DR6 maps and reported the averaged rotation angle ⟨ψ⟩=0.20°±0.08°, which motivates the ACT comparison.","marker":"[21]"},{"why":"Released ACT DR6 spectra, masks, time-split maps and binning; the paper adopts the ACT team's baseline multipole ranges.","marker":"[22]"},{"why":"Established the Bayesian framework for ACT with Gaussian α priors, yielding β=0.215°±0.074°, whose prior settings the paper adopts.","marker":"[25]"},{"why":"Derived from an optics model the miscalibration uncertainties σ_α = 0.09°, 0.09°, 0.11° used as ACT priors.","marker":"[26]"},{"why":"Supplied the filamentary model of dust EB correlations used to model foreground contamination.","marker":"[20]"}],"fun_headline_variants":["Joint ACT+Planck analysis finds 4.8σ cosmic birefringence","CMB polarization rotation at 4.8σ from ACT plus Planck","Cosmic birefringence hinted at 4.8σ in joint CMB data","Axion-like field? CMB rotation seen in ACT+Planck joint fit","Parity-breaking CMB rotation: 4.8σ but systematics remain"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The result stands on the assumption that the Gaussian priors on the ACT miscalibration angles α_i (σ = 0.09°, 0.09°, 0.11°, correlated at 90% within each array) are as accurate as quoted, because the ACT maps probe only ℓ>600 and mask the Galaxy, leaving the α+β degeneracy on that leg broken almost entirely by those priors.","fun_headline_variants_meta":{"raw":{"variants":["Joint ACT+Planck analysis finds 4.8σ cosmic birefringence","CMB polarization rotation at 4.8σ from ACT plus Planck","Cosmic birefringence hinted at 4.8σ in joint CMB data","Axion-like field? CMB rotation seen in ACT+Planck joint fit","Parity-breaking CMB rotation: 4.8σ but systematics remain"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00034,"raw_usage":{"total_tokens":1921,"prompt_tokens":1035,"completion_tokens":886,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":651,"completion_tokens_details":{"reasoning_tokens":777}},"tokens_in":651,"tokens_out":886,"duration_ms":6969,"temperature":1.0,"reasoning_tokens":777,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T14:32:33.444877+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the ACT polarization-angle miscalibration directly from observations of a bright polarized astrophysical calibrator with known polarization angle and re-run the joint fit; if the independently calibrated α_i shift the best-fit β by more than about 0.2°, the reported 4.8σ detection is a product of the priors rather than cosmology, and if they confirm the priors the detection survives this test.","supporting_citations":[],"review_version":2}