{"id":"85b524b6-de33-4e27-8f0c-6139f55cfc6d","arxiv_id":"2501.11904","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"CLASS 90 GHz data cross-correlated with Planck give tau = 0.053 (+0.018, -0.019), the first ground-based reionization optical depth measurement, with reionization detected at 99.4% confidence.","lead":"The CLASS telescope in Chile used 90 GHz light to measure the largest-scale polarization pattern of the cosmic microwave background, then combined it with Planck satellite data to estimate the optical depth of cosmic reionization as about 0.053 plus or minus 0.019. This is the first time a ground-based experiment has contributed to such a reionization measurement.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reionization claim rests on a single multipole: the paper reports that dropping ell=4 removes the detection, so the 99.4% significance is not robust to a systematic error at that one mode.","rationale":"I read the paper as a careful, simulation-heavy analysis whose methodological steps are internally consistent and transparently validated. The map-making non-linearity is small (<3% at ell=3) and its effect on tau is expected to be roughly half as large, so I do not regard it as fatal by itself. The reader's weakest assumption is reasonable but not the most load-bearing: the real fragility is that the headline detection significance is a single-multipole statement. The paper itself discloses this in Figure 16 and the surrounding text. That disclosure is a credit to the authors, but it means the claim 'detected at 99.4%' should be interpreted with the caveat that removing one multipole erases the detection. This does not change the advisable editorial action: CONDITIONAL remains appropriate, with the condition being a more explicit robustness characterization at ell=4 and a demonstration that the quoted error covers the single-mode sensitivity. I therefore keep the reader's verdict unchanged.","tokens_in":33463,"tokens_out":9545,"duration_ms":110466,"concrete_test":"Use the existing 500 end-to-end simulations to (a) report the mean bias and scatter in the recovered tau and in C_4^EE for the full pipeline with input tau=0.054; (b) recompute the tau posterior and the tau=0 rejection significance after perturbing the data's C_4^EE by +1 sigma and -1 sigma of its total error; and (c) recompute after applying the measured map-making bias at ell=4 (from Fig. 7) as a correction. If the tau=0 rejection falls below ~3 sigma for a 1-sigma perturbation of ell=4, or if the pipeline recovery bias in tau exceeds ~0.002, the paper should present the result as a constraint consistent with reionization rather than as a detection.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—'detected cosmic reionization at 99.4% significance' with tau=0.053(+0.018,-0.019)—is carried by the EE cross-spectrum at ell=4. Section 5.4 and Figure 16 state that omitting ell=4 reduces the detection significance below threshold, and the discussion around Equation 15 and Figure 17 shows that the ell=4 likelihood gains from a tau-dependent sample-variance term as well as the shift in the theory mean. Consequently, the headline significance depends on the fidelity of one band power and on the accuracy of its total variance. The paper's end-to-end estimate of map-making non-linearity is quoted as <3% at ell=3 (Section 3.4.4) and is not corrected, but no number is quoted at ell=4, the mode that matters. Similarly, the transfer-matrix and foreground corrections are validated in aggregate (Figure 19, Table 1), not specifically at ell=4. A systematic at ell=4 at the level of a fraction of the quoted error, or a misspecified variance at that multipole, would change both tau and the 'detection' claim. This is a burden-of-proof concern about the strength of the central claim, not an accusation of error; the paper is transparent about the ell=4 sensitivity.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents the CLASS 90 GHz polarization analysis with the goal of recovering the largest-angular-scale CMB E-mode signal. The authors introduce a pixel-space transfer-matrix formalism to correct for time-domain filtering in maximum-likelihood map-making, implement it in a quadratic cross-spectrum estimator (xQML), and validate the pipeline with 500 end-to-end simulations, internal null tests, and cross-checks against Planck. The central scientific claim is a measurement of the reionization optical depth from a CLASS×Planck cross-correlation, tau = 0.053 (+0.018, -0.019), corresponding to a 99.4% rejection of the no-reionization hypothesis, which the paper labels as the first ground-based attempt at such a measurement. At intermediate multipoles (ell>30) the results are consistent with Planck.","tokens_in":33741,"tokens_out":4642,"duration_ms":51903,"significance":"If the central claim holds, this is a substantial methodological and observational milestone: it demonstrates that ground-based data, after careful filtering correction, can contribute to low-ell E-mode polarization measurements and to constraints on the reionization optical depth. The paper is unusually transparent: it provides extensive validation with 500 simulations, a full PTE table for null tests, a public implementation of the modified xQML estimator, and explicit statements of residual unmodeled biases. These strengths are real and should be credited. However, the headline detection significance and the tau value are carried by a single multipole, ell=4, and the per-multipole accuracy of the transfer-matrix correction, foreground cleaning, and noise covariance at that exact multipole is not separately demonstrated. The significance of the result is therefore conditional on the fidelity of one band power and on the accuracy of its total variance, including sample-variance contributions.","major_comments":[{"comment":"The headline claim of a 99.4% detection of cosmic reionization is not robust to the omission of a single multipole: the paper states that dropping ell=4 reduces the detection significance below threshold, and Figure 16 shows that no detection is claimed when multipoles below ell=5 are excluded. Since the entire reionization detection and the tau estimate rest on the ell=4 band power, the manuscript should provide per-multipole validation of the transfer-matrix-corrected xQML estimator at ell=4, including foreground-cleaning residuals and noise covariance accuracy at that multipole. The current validation (e.g., Figure 19 and Table 1) is aggregate over ell ranges and does not establish that the ell=4 band power and its variance are unbiased at the level needed for the 99.4% significance claim.","section":"§5.4, Figure 16"},{"comment":"The map-making non-linearity bias is characterized as less than 3% at ell=3 and is left uncorrected, but no estimate is given at ell=4, which is the multipole that drives the tau result. Because the likelihood at ell=4 depends both on the mean shift in C_ell and on the sample-variance term in Equation (15), a systematic error in the ell=4 band power or in its variance at the level of even a moderate fraction of the quoted error would alter both tau and the detection significance. The authors should report the non-linearity bias at ell=4 (and at each ell used in the likelihood) and, if it is not negligible, include it as a systematic uncertainty or correct it.","section":"§3.4.4, Figure 7"},{"comment":"The low-ell EE null test for the VPM sync split fails with PTE below 2e-4, driven by ell=8, and the paper attributes this to inadequate noise modeling when detector-pair cancellation is compromised. The authors argue that this failure does not affect the final cross-correlation because the ell=8 outlier is not present in the CLASS×Planck spectrum. However, the failure demonstrates that the noise model can be substantially inaccurate for some detector configurations, and the final analysis uses the same noise model. The manuscript should provide a direct test that the final (unsplit) CLASS noise covariance and the ell=4 band-power variance are unaffected, for example by comparing a null test constructed with the final noise model or by showing that the VPM sync split null spectrum is consistent with simulations after the proposed noise-model improvement.","section":"§4.1, Appendix B"},{"comment":"The tau measurement is not an independent ground-based determination: the CMB channel is the Planck HFI coadded map, the absolute polarization calibration is fitted to Planck high-ell spectra (Section 4.2), and the likelihood in Section 5.4 imposes a Gaussian prior on A_s e^{-2tau} derived from Planck. The paper acknowledges that calibration does not compromise tau independence only to the extent that A_s e^{-2tau} is well constrained, but the abstract and conclusion should state more explicitly that the result is a cross-correlation measurement whose parameter interpretation is tied to Planck-based calibration and priors. The 99.4% significance is the significance of the cross-spectrum being nonzero, not of CLASS independently detecting reionization; this distinction should be made explicit in the title or abstract claims.","section":"§4.2, §5.4"}],"minor_comments":[{"comment":"The abstract quotes a polarization sensitivity of 78 muK arcmin, while the body text (Section 3.3) reports 82 muK arcmin; these values should be reconciled.","section":"Abstract vs. §3.3"},{"comment":"The labels 'az-2ε' and 'az-4ε' appear to be typographical errors for 'az-2π' and 'az-4π' used in Figure 8 and the text.","section":"Figure 9"},{"comment":"The sentence 'The maximum of this bias is found to be less than 3% at ell=3' is ambiguous: it should clarify whether 3% is the maximum over all ell and occurs at ell=3, or whether ell=3 is the only multipole at which the bias is estimated.","section":"§3.4.4"},{"comment":"The use of the square root of the auto-spectrum transfer function for cross-spectra is an approximation that the paper states is negligible; a sentence quantifying the difference or citing a validation test would improve clarity.","section":"§4.2, Eq. (7)"},{"comment":"The description of the likelihood step 6 should explicitly state that the uniform prior on tau is over 0 to 0.3 and that the Gaussian prior on A_s e^{-2tau} is a choice imported from Planck; these choices are central to the reported error bars and should be restated in the results paragraph.","section":"§5.4"}],"recommendation":"major_revision","confidential_remarks":"The paper is methodologically rich and unusually transparent, and the central idea—using a pixel-space transfer matrix and xQML to recover large-scale E-modes from a ground-based experiment—is valuable. My main concern is that the headline detection claim is fragile because it rests on the ell=4 band power, while the per-multipole validation at exactly that multipole is not provided. The authors have disclosed this fragility honestly, but the claim 'detected cosmic reionization at 99.4% significance' needs either dedicated per-multipole validation, including a direct estimate of the non-linearity bias at ell=4, or a more cautious wording that emphasizes the cross-correlation nature and the dependence on Planck calibration and priors. The paper is a good fit for ApJ; I would support publication after the requested revisions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read the CLASS 90 GHz paper. Bottom line: the pixel-space transfer-matrix extension of xQML is the genuinely new and useful piece, and the reionization measurement is an honest but fragile first step from the ground. I'd send it to peer review, but I'd push the authors to make the ell=4 dependence impossible to miss in the abstract; right now the '99.4% detection' reads stronger than the actual robustness.\n\nWhat the paper does well: the transfer-matrix implementation lets the quadratic estimator handle anisotropic time-domain filtering down to ell=2, with validation against reobserved maps showing unbiased recovery to at least ell=40. That is a methodological contribution other experiments will want to copy. The end-to-end characterization of map-making non-linearity is also a real step up from the 40 GHz work, and the null-test battery is unusually thorough for a ground-based experiment. The tau value, tau = 0.053 (+0.018, -0.019), agrees with Planck-only results, which is reassuring rather than surprising.\n\nThe soft spots are exactly the ones the authors themselves flag, and that transparency is a point in their favor. The reionization significance is carried by ell=4: drop that multipole and the detection goes below threshold (Figure 16). The likelihood at ell=4 also gains from a tau-dependent sample-variance term, so part of the signal is the variance broadening, not just the mean shift. That means a systematic at that one multipole, or a misspecified variance, could move tau by more than the quoted error. The VPM sync null test failure at low-ell EE (PTE < 2e-4, driven by ell=8) is a separate unresolved issue; the authors argue it comes from noise modeling with unpaired detectors and it does not affect the final cross-spectrum, but it is exactly the kind of thing I would want revisited before anyone calls this a robust ground-based low-ell detection.\n\nOn independence: this is not a standalone ground-based tau measurement. The low-ell CMB channel is Planck HFI, the absolute calibration is pinned to Planck high-ell EE, and the likelihood uses a Planck-derived Gaussian prior on A_s e^{-2tau}. The authors are upfront about this. The result is best read as a demonstration that CLASS 90 GHz maps contain recoverable large-scale E-mode information after filtering correction. That demonstration succeeds.\n\nWho is this for? CMB experimentalists planning low-ell polarization from the ground or balloons, and cosmologists working on tau estimation. The transfer-matrix method deserves to be adopted. I would accept this for peer review; the analysis is careful, the simulations are extensive, and the central methodological claim holds up. My referee comments would center on reframing the abstract and possibly adding a per-multipole likelihood table so readers can see the ell=4 leverage directly.","headline":"The pixel-space transfer-matrix extension of xQML is the real contribution here, and the tau measurement is an honest but fragile first step from the ground that deserves a serious referee.","tokens_in":34429,"tokens_out":3217,"would_cite":true,"duration_ms":37229,"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":"A ground-based telescope has measured the CMB's largest-scale E-mode polarization and detected cosmic reionization at 99.4% significance.","keywords":["cosmic microwave background","E-mode polarization","reionization optical depth","ground-based CMB observation","pixel-space transfer matrix","quadratic maximum-likelihood estimator","CLASS","large angular scale polarization"],"falsifier":"An end-to-end simulation that applies the measured map-making nonlinearity correction at $\\ell=4$ specifically, recomputing the likelihood, would settle whether the $\\tau$ central value shifts by more than the quoted uncertainty; alternatively, an independent ground-based or balloon-borne low-$\\ell$ E-mode measurement with comparable sky coverage returning $\\tau$ outside $0.053 \\pm 0.019$ would contradict the detection.","tokens_in":33226,"feed_emoji":"🔭","tokens_out":6532,"duration_ms":59091,"temperature":0.7,"pith_summary":"This paper reports the first ground-based measurement of the reionization optical depth, $\\tau$, from the largest-angular-scale E-mode polarization of the cosmic microwave background. Using 115 detector-years of CLASS 90 GHz observations cross-correlated with Planck maps, the authors detect the reionization signal at 99.4% significance and measure $\\tau = 0.053^{+0.018}_{-0.019}$, consistent with Planck-only results. The paper's central methodological claim is that time-domain filtering in maximum-likelihood map-making can be precisely characterized and corrected with a pixel-space transfer matrix, allowing recovery of the largest angular scale polarization modes that were previously thought to be lost. If this approach is correct, ground-based experiments can participate in low-multipole reionization measurements that were previously the exclusive domain of space missions, and the path to a cosmic-variance-limited $\\tau$ measurement from CLASS becomes clear.","feed_headline":"Ground-based survey detects reionization at 99.4%","feed_subtitle":"CLASS 90 GHz data plus Planck yield tau = 0.053, opening low-ell cosmology to the ground.","key_machinery":"The paper's key object is the pixel-space transfer matrix $\\mathbf{F}_{16,N_F}$, a linear operator constructed by reobserving and downgrading single-pixel input maps with the map-making pipeline's reobservation operator $\\mathbf{R}$, which fixes the noise model to a realization-independent form. This matrix forward-models the effect of time-domain filtering on sky signals at the map level, and is paired with a modified quadratic maximum-likelihood cross-spectrum estimator (xQML) that incorporates the transfer matrix into the signal covariance, allowing unbiased power spectrum recovery down to $\\ell=2$. The transfer matrix is evaluated at $N_\\mathrm{side}=32$ for signal maps and $N_\\mathrm{side}=16$ for noise covariance matrices, and it is also used to filter foreground templates so that cleaning is performed consistently with the filtered data.","core_discovery":"The central discovery is that the large-scale E-mode signal, which encodes the reionization history and is enhanced roughly as $\\tau^2$ on the largest angular scales, is recoverable from a ground-based instrument after the time-domain filtering bias is forward-modeled and corrected. The authors demonstrate this by cross-correlating the CLASS 90 GHz map with foreground-reduced Planck 100/143 GHz maps, obtaining a detection of reionization at 99.4% confidence and $\\tau = 0.053^{+0.018}_{-0.019}$. They further show that the only uncorrected bias, arising from the nonlinearity of the maximum-likelihood map-maker's noise weighting, is less than 3% at $\\ell=3$ and is small enough not to be applied in the spectra. The cross-spectrum is validated by null tests, consistency with Planck at intermediate angular scales ($\\ell > 30$), and robustness to masks, multipole ranges, and external data sets, with the caveat that the significance weakens when multipoles below $\\ell=5$ are dropped.","pith_inferences":["Editorial extension: if the transfer-matrix correction remains valid as newer modulators (reflective half-wave plate) and a second 90 GHz telescope come online, the CLASS dataset could provide a $\\tau$ measurement with precision comparable to Planck's, and possibly an independent check on the lensing-anomaly inflating $A_s$ values.","Editorial extension: the same pixel-space transfer matrix machinery is directly transferable to other ground-based and balloon-borne CMB experiments (e.g., those with heavier time-domain filtering), including for B-mode analyses on large scales where filtering is even more aggressive.","Editorial extension: the paper's finding that the $\\tau$ constraint is driven mainly by the single multipole $\\ell=4$ implies that closer attention to the map-making nonlinearity at exactly that scale should be a priority for future low-$\\ell$ analyses; a targeted end-to-end simulation at $\\ell=4$ with more realizations could sharpen the result.","Editorial extension: the discarded closeout-period data, which represents nearly half the survey, might be recoverable if the low-$\\ell$ noise modeling improves, potentially more than doubling the effective sensitivity without new observations."],"forward_implications":["Ground-based CMB polarization experiments can recover the largest-scale E-modes after a transfer-matrix filtering correction, opening low-$\\ell$ reionization measurements to the ground.","The measured $\\tau = 0.053^{+0.018}_{-0.019}$ is consistent with Planck HFI-based values, supporting the lower range of $\\tau$ estimates that reduce the $\\sigma_8$ tension via $A_s e^{-2\\tau}$.","The xQML estimator with transfer-matrix correction is unbiased up to $\\ell=40$ in validation tests, providing a reusable tool for other experiments facing filtering-induced signal loss.","Forecasts in the paper indicate that modest filtering optimization (a 20% improvement) could bring CLASS's $\\tau$ precision close to the cosmic variance limit of about 0.003 for its sky patch."],"supporting_citations":[{"why":"Establishes that reionization boosts E-mode polarization on large angular scales, motivating the low-multipole measurement.","marker":"Zaldarriaga 1997"},{"why":"Supplies the simulation-based likelihood, Galactic masks, SRoll2 100/143 GHz maps, and the reference Planck low-ell analysis whose framework this paper adapts.","marker":"P20"},{"why":"Provides the xQML quadratic cross-spectrum estimator implementation that the paper modifies to include pixel-space transfer matrix correction.","marker":"Vanneste et al. 2018"},{"why":"Establishes the CLASS 40 GHz analysis pipeline, reobservation operator, and template-subtraction map-making that this work extends to 90 GHz.","marker":"L23"},{"why":"Provides the Planck PR4 SEVEM CMB maps and 30 GHz synchrotron template used for cross-correlation and foreground cleaning.","marker":"Planck Collaboration Int. LVII 2020"},{"why":"Introduces the pixel-space transfer function approach for filtering bias correction, which this paper refines into the transfer matrix.","marker":"BICEP2 Collaboration et al. 2016"}],"fun_headline_variants":["CLASS detects reionization at 99.4% confidence","Ground-based CMB E-mode yields tau = 0.053","First ground-based reionization optical depth from CLASS","CLASS 90GHz recovers largest-scale E-mode polarization","Reionization signal measured from ground at 99.4%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The measurement assumes the map-making noise model is linear and independent of the sky signal; the residual nonlinearity, uncorrected and measured at under 3% at $\\ell=3$, could be larger at $\\ell=4$, the multipole that drives the $\\tau$ result.","fun_headline_variants_meta":{"raw":{"variants":["CLASS detects reionization at 99.4% confidence","Ground-based CMB E-mode yields tau = 0.053","First ground-based reionization optical depth from CLASS","CLASS 90GHz recovers largest-scale E-mode polarization","Reionization signal measured from ground at 99.4%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001008,"raw_usage":{"total_tokens":4296,"prompt_tokens":1019,"completion_tokens":3277,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":635,"completion_tokens_details":{"reasoning_tokens":3192}},"tokens_in":635,"tokens_out":3277,"duration_ms":24980,"temperature":1.0,"reasoning_tokens":3192,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T17:44:59.362670+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"An end-to-end simulation that applies the measured map-making nonlinearity correction at $\\ell=4$ specifically, recomputing the likelihood, would settle whether the $\\tau$ central value shifts by more than the quoted uncertainty; alternatively, an independent ground-based or balloon-borne low-$\\ell$ E-mode measurement with comparable sky coverage returning $\\tau$ outside $0.053 \\pm 0.019$ would contradict the detection.","supporting_citations":[],"review_version":1}