{"id":"20b4c9e1-0f6e-428b-afd3-f9920881f07a","arxiv_id":"2507.04964","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Using 17 minutes of MWA data, the authors place the first comprehensive upper limits on the z=8.2 21-cm bispectrum and characterize its foreground wedge.","lead":"Astronomers analyzed 17 minutes of Murchison Widefield Array data to measure the 21-cm bispectrum from the Epoch of Reionization, finding the signal is swamped by foregrounds and only weak upper limits are possible. They also mapped the 'foreground wedge' for the bispectrum and showed that periodic missing frequency channels leak contamination into the cleanest region.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Foreground bispectrum sign/cancellation makes the quoted 2σ upper limits on the EoR signal unjustified; spherical averaging can cancel positive and negative cylindrical BS contributions.","rationale":"The reader's weakest_assumption correctly flags the unmodeled sign of the foreground bispectrum as the key issue for the upper-limit claim. I sharpen this to a concrete internal problem: the spherical and bin averaging can cancel positive and negative cylindrical BS values, so the measured Δ3 may not even represent the foreground amplitude, let alone bound the EoR signal. The paper's own figures show oscillatory sign structure, and Table 3 confirms both signs of Δ3 across bins. The paper does not report the sign distribution within the bins used for the headline limits, so a reader cannot verify that the 'foreground-dominated' measurement has a sign-definite foreground contribution. The proposed test (|B| averaging vs. standard averaging) would directly expose whether cancellation is distorting the quoted limits. Since the reader's verdict is already CONDITIONAL and this concern is the reason for the condition, the verdict remains UNCHANGED. The wedge and spike findings are well supported and not affected by this concern; the upper-limit interpretation is the load-bearing weakness.","tokens_in":32601,"tokens_out":14058,"duration_ms":153067,"concrete_test":"For the two Table 2 bins (equilateral k1=0.008 Mpc⁻¹ and squeezed k1=0.012 Mpc⁻¹), recompute the spherical BS from the cylindrical BS using (i) the standard orientation average and (ii) an average of the absolute value |B| before orientation averaging. Also compute the fraction of negative cylindrical BS contributions in each bin. If |⟨B⟩| is much smaller than ⟨|B|⟩, cancellation is significant and the quoted Δ3_UL is not a robust upper limit; a conservative limit would instead use ⟨|B|⟩ or the maximum |B| in the bin.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline 2σ upper limits (Table 2) are the measured Δ3 values after spherical averaging over triangle orientations and binning in (k1, μ, t). The cylindrical BS B(k1⊥, k2⊥, k3⊥, k1∥, k2∥) is not sign-definite: Figure 4 and Table 3 show positive and negative values across the (k1∥, k2∥) plane and across different triangle shapes. The spherical monopole (Eq. 10) averages over all orientations, so foreground contributions of opposite signs can partially cancel, making the binned Δ3 smaller than the typical foreground amplitude in the bin. The paper states 'We use these foreground-dominated estimates to place upper limits on Δ3 for the EoR 21-cm signal' (Section 4) without modeling the sign of the foreground bispectrum or demonstrating that such cancellation is negligible for the bins used in Table 2. If the measured Δ3 is the residual of a cancellation, it neither bounds the foreground amplitude nor the EoR signal, especially since the EoR bispectrum is predicted to be negative at small k1 (Section 4). Therefore the upper-limit interpretation is not internally justified.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes 17 minutes of MWA Phase II drift-scan data at 154.2 MHz to estimate the z=8.2 21-cm bispectrum using a visibility-based estimator developed in companion papers. It defines a cylindrical bispectrum B(k1⊥,k2⊥,k3⊥,k1∥,k2∥), shows that foregrounds produce a wedge analogous to the power-spectrum wedge, and partitions the (k1∥,k2∥) plane into regions A0-A3 according to how many triangle sides avoid the wedge. It identifies the A3 region as a candidate EoR window, but shows that periodic missing frequency channels introduce a periodic pattern of spikes that leaks into this window. The paper then computes the spherically averaged mean cube brightness temperature fluctuation Δ3(k1,μ,t) and quotes 2σ upper limits of Δ3_UL=(1.81×10^3)^3 mK^3 at k1=0.008 Mpc^-1 (equilateral) and (2.04×10^3)^3 mK^3 at k1=0.012 Mpc^-1 (squeezed), noting that these are foreground-dominated and far above predicted EoR values.","tokens_in":32838,"tokens_out":7548,"duration_ms":82931,"significance":"If the wedge characterization holds, the paper provides a useful methodological advance: it is one of the first attempts to map a foreground wedge in the cylindrical 21-cm bispectrum and to quantify the contamination from periodic flagging. The estimator is validated in prior companion papers on simulated data with matching baseline distribution, bandwidth, and flagging, and the thermal noise is estimated from 50 dedicated noise-only simulations. The identification of the periodic spike pattern with the 1.28 MHz flagging period is concrete and falsifiable. However, the headline upper limits are preliminary and not competitive with predicted signal levels; the main value of the paper is the foreground-wedge morphology and the demonstration that the A3 region is not clean due to flagging-induced leakage.","major_comments":[{"comment":"The interpretation of the measured foreground-dominated Δ3 values as 2σ upper limits on the EoR 21-cm bispectrum is not justified. The cylindrical bispectrum is not sign-definite, as shown by the positive and negative values in the third row of Figure 4, and the spherical monopole in Eq. (10) averages over all triangle orientations. Positive and negative foreground contributions can therefore partially cancel in the spherical averaging, so the binned Δ3 values in Table 2 can be much smaller than the typical foreground bispectrum amplitude in those bins. The paper does not model the sign of the foreground bispectrum or demonstrate that cancellation is negligible for the bins used in Table 2. The problem is compounded by the statement in Section 4 that the EoR Δ3 is predicted to be negative at small k1, whereas the quoted upper limits are positive measured values. As presented, the measured Δ3 does not bound the EoR signal, and the abstract's headline upper limits rest on an unsupported assumption.","section":"Section 4, Table 2, Eq. (10)"},{"comment":"The quantity labeled '2σ upper limit Δ3_UL' is set equal to the measured Δ3 rather than to the measured value plus 2σ (or to |Δ3| plus 2σ). Since the same table lists nonzero 1σ values, this is formally incorrect; the equality is numerically reasonable only because the 1σ values are much smaller than the measured values. In addition, for negative entries in Table 3 the quoted Δ3_UL is the absolute value of the measured Δ3 without any stated recipe, and the sign information is discarded. Because the predicted EoR bispectrum at small k1 is negative, the direction of the limit matters, and the paper should define explicitly whether the limits apply to Δ3 or to |Δ3| and how the 2σ noise contribution is included.","section":"Section 4, Table 2"}],"minor_comments":[{"comment":"The claim that the predicted EoR signal is 13–15 orders of magnitude smaller than the upper limits is inconsistent with the numbers in Table 3; for the same k1 range, the ratio is about 11–12 orders at k1=0.38 Mpc^-1 and about 6–7 orders at the smallest k1. Please recalculate the comparison.","section":"Section 4, paragraph after Figure 7"},{"comment":"There is a typo in the sentence 'the values of B(k1⊥,k2⊥,k3⊥,k1∥,k2∥), shown in Figure 4, peak around k1∥=k1∥=0'; this should read k1∥=k2∥=0.","section":"Section 5, summary paragraph"},{"comment":"The table should state explicitly that Δ3_UL is defined as |Δ3| (or as |Δ3| plus the appropriate multiple of the noise), since the table lists positive upper limits for rows where Δ3 is negative.","section":"Table 3 caption"},{"comment":"The estimator is referred to as 'Paper II' with the citation 'Gill & Bharadwaj 2025', but the reference list does not provide a full bibliographic entry; please supply a complete reference or indicate the manuscript status.","section":"Section 3.2 and References"},{"comment":"The top panels mix two different physical quantities (|B| in units of mK^3 Mpc^6 and |P| in units of mK^2 Mpc^3) with different color scales; a short sentence in the caption clarifying how the color scales are chosen for each panel would improve readability.","section":"Figure 5"}],"recommendation":"major_revision","confidential_remarks":"The paper's most defensible contribution is the foreground-wedge morphology for the cylindrical bispectrum and the characterization of the periodic flagging artifact. The upper-limit interpretation is the weakest point and should be either removed or substantially qualified; the sign-cancellation issue is central and cannot be fixed by a small edit. If the authors reframe the results as foreground-dominated measurements rather than EoR upper limits, or add a conservative treatment that avoids orientation-averaging cancellation, the paper would be suitable for publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The useful part of this paper is the wedge map. They show that the cylindrical 21-cm bispectrum has a foreground wedge analogous to the power-spectrum wedge, with regions A0-A3 depending on how many triangle sides are outside the wedge, and that the A3 “EoR window” is contaminated by periodic spikes from the missing frequency channels. That is new and will help design future bispectrum analyses. The wedge identification is supported by Figures 4 and 5, and the noise-only simulations with 50 realisations are described clearly enough to reproduce.\n\nThe main problem is the upper-limit interpretation. Table 2 quotes the measured Δ³ values as 2σ upper limits on the EoR signal. But the cylindrical bispectrum is not sign-definite: Table 3 itself contains both positive and negative Δ³ values at the same k1 for different triangle shapes. Spherical averaging over orientations can therefore cancel positive and negative foreground contributions, making the binned Δ³ smaller than the typical foreground amplitude in that bin. The paper states that the foreground-dominated estimates are used as upper limits, but it never models the sign of the foreground bispectrum or demonstrates that cancellation is negligible for the bins quoted.\n\nThis is not a minor subtlety because the predicted EoR bispectrum is negative at small k1, which the paper itself notes. If the foreground bispectrum is positive and large while the signal is negative, the measured positive residual can be the difference of two large numbers, and the true signal amplitude could be far larger than the quoted limit. In that case the measured Δ³ is not a conservative upper bound; it could even be a lower bound on the foreground amplitude. The fix is either to apply foreground subtraction or avoidance before quoting limits, or to explicitly model the foreground bispectrum sign and show that the residual is signal-dominated.\n\nThe all-triangle-shape upper-limit table (Table 3) is still a useful reference for survey design, and the flagging-spike analysis is valuable. The paper deserves a serious referee because the wedge and spike characterisation is genuinely new and helpful for the 21-cm community, even though the headline upper-limit claim needs substantial revision before publication. I would engage with it, and I would ask for the upper-limit section to be rewritten with the sign/cancellation issue addressed.","headline":"First useful wedge characterisation for the 21-cm bispectrum, but the quoted upper limits are not actually justified because the foreground bispectrum can change sign and cancel the signal.","tokens_in":33385,"tokens_out":2693,"would_cite":true,"duration_ms":30498,"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":"The 21-cm bispectrum from Epoch of Reionization data exhibits a foreground wedge analogous to the power-spectrum wedge, and the best 2σ upper limits are about (1.8×10³)³ mK³.","keywords":["Epoch of Reionization","21-cm bispectrum","foreground wedge","MWA","interferometric correlation","diffuse radiation","radio interferometry","upper limits"],"falsifier":"Inject a synthetic EoR 21-cm bispectrum of known amplitude (e.g., ~$10^3$ mK$^3$) into the flagged visibility data alongside the real foregrounds and rerun the full pipeline; if the recovered $\\Delta^3$ does not track the injected signal in the A3 region, the upper-limit interpretation is falsified. Alternatively, measure the foreground-only bispectrum from a smooth-spectrum sky model with the same flagging pattern and check its sign: a negative foreground bispectrum comparable in amplitude to the measured $\\Delta^3$ would invalidate the limits.","tokens_in":32416,"feed_emoji":"📡","tokens_out":8474,"duration_ms":68146,"temperature":0.7,"pith_summary":"The paper attempts to measure the z = 8.2 Epoch of Reionization 21-cm bispectrum using about 17 minutes of Murchison Widefield Array observations at 154.2 MHz. It establishes that the cylindrical bispectrum, like the cylindrical power spectrum, is confined to a foreground wedge whose boundary is set by the horizon condition on each side of the triangle, and it defines the region where all three sides avoid the wedge (A3) as the bispectrum's EoR window. The authors show that this window is contaminated by a periodic grid of spikes caused by the periodic pattern of missing frequency channels in the MWA data. The best foreground-avoided 2σ upper limits on the mean-cube brightness temperature fluctuations are $\\Delta^3_{\\rm UL} = (1.81\\times 10^3)^3$ mK$^3$ at $k_1 = 0.008$ Mpc$^{-1}$ (equilateral) and $(2.04\\times 10^3)^3$ mK$^3$ at $k_1 = 0.012$ Mpc$^{-1}$ (squeezed), which are foreground-dominated and roughly 13–15 orders of magnitude above the predicted signal. The significance is that it provides the first detailed characterization of the foreground structure that any 21-cm bispectrum detection must overcome.","feed_headline":"21-cm bispectrum shows its own foreground wedge","feed_subtitle":"MWA data at z=8.2 give 2σ upper limits ~10³ mK³ and map the wedge any detection must beat.","key_machinery":"The central object is the cylindrical bispectrum $B(k_{1\\perp}, k_{2\\perp}, k_{3\\perp}, k_{1\\parallel}, k_{2\\parallel})$, obtained by a two-dimensional Fourier transform of the multi-frequency angular bispectrum $B_A(\\ell_1, \\ell_2, \\ell_3, \\Delta\\nu_1, \\Delta\\nu_2)$ derived from the three-visibility correlation. The foreground wedge is defined by three horizon conditions $|k_{a\\parallel}| \\le [r/(r' \\Delta\\nu_c)] k_{a\\perp}$ for $a = 1,2,3$, which partition the $(k_{1\\parallel}, k_{2\\parallel})$ plane into regions A0, A1, A2, and A3 depending on how many triangle sides avoid the wedge; A3 is the EoR window. The estimator is a binned version that counts triangles $(k_1,k_2,k_3)$ and weights by $N_{\\rm tri}$, and the paper restricts analysis to the spherical bispectrum monopole expressed as $\\Delta^3(k_1, \\mu, t) = (2\\pi^2)^{-2} k_1^6 \\bar{B}(k_1, \\mu, t)$.","core_discovery":"Using the three-visibility correlation, the authors estimate the multi-frequency angular bispectrum (MABS) and, after a two-dimensional Fourier transform, the cylindrical bispectrum $B(k_{1\\perp}, k_{2\\perp}, k_{3\\perp}, k_{1\\parallel}, k_{2\\parallel})$ for all triangle configurations supported by the data. They find that the cylindrical bispectrum shows three bands of foreground contamination aligned with $k_{1\\parallel} = 0$, $k_{2\\parallel} = 0$, and $k_{3\\parallel} = 0$, which they identify as the footprint of a foreground wedge: each side of the triangle satisfies $|k_{a\\parallel}| \\le [r/(r' \\Delta\\nu_c)] k_{a\\perp}$, with a boundary value 3.51 for this data. Partitioning the $(k_{1\\parallel}, k_{2\\parallel})$ plane into regions A0–A3 by how many sides avoid the wedge, they identify A3, containing 83.6% of the $1.66\\times 10^{12}$ triangles, as the EoR window. They further show that the periodic flagging of one channel per 1.28 MHz produces a periodic spike pattern at $\\delta k_{\\parallel} = 0.29$ Mpc$^{-1}$ that leaks into the EoR window, and that this leakage persists after foreground avoidance. The measured values of the spherical bispectrum, expressed as mean-cube brightness temperature fluctuations $\\Delta^3$, are foreground-dominated in all triangle bins, and the tightest 2σ upper limits are $(1.81\\times 10^3)^3$ mK$^3$ at $k_1 = 0.008$ Mpc$^{-1}$ (equilateral) and $(2.04\\times 10^3)^3$ mK$^3$ at $k_1 = 0.012$ Mpc$^{-1}$ (squeezed).","pith_inferences":["The half-period (0.145 Mpc$^{-1}$) spike spacing seen for one squeezed configuration suggests that different $k_{\\parallel}$ combinations alias the same missing-frequency pattern differently, so flagging mitigation may need to be triangle-shape-specific.","The same wedge-and-spike structure should appear in any EoR bispectrum analysis with periodic channel flagging, including future HERA or SKA-LOW observations, implying that flagging-aware estimators will be necessary for a detection.","Subtracting a smooth foreground model before the bispectrum estimate, as the paper suggests for future work, could push the limits down by the two to three orders of magnitude that foreground avoidance alone achieves only at the smallest $k_1$.","A natural next step would be to combine the multiple drift-scan pointings already available in the MWA G0031 project, increasing integration time and possibly turning the current upper limits into a measurement."],"forward_implications":["The cylindrical 21-cm bispectrum has a foreground wedge whose boundary on each side of the triangle is set by the horizon condition $|k_{a\\parallel}| = [r/(r'\\Delta\\nu_c)] k_{a\\perp}$, and the A3 region where all three sides avoid the wedge contains 83.6% of the available triangles.","A periodic pattern of missing frequency channels (period 1.28 MHz) produces spikes at $\\delta k_{\\parallel} = 0.29$ Mpc$^{-1}$ that leak into the EoR window, so flagging patterns must be modeled in any bispectrum analysis.","The best 2σ upper limits are $\\Delta^3_{\\rm UL} = (1.81\\times 10^3)^3$ mK$^3$ at $k_1 = 0.008$ Mpc$^{-1}$ (equilateral) and $(2.04\\times 10^3)^3$ mK$^3$ at $k_1 = 0.012$ Mpc$^{-1}$ (squeezed); these are foreground-dominated and 13–15 orders of magnitude above the predicted signal.","Restricting to the A3 region reduces foreground contamination by a factor of about 620 at $k_1 \\approx 0.3$ Mpc$^{-1}$ for squeezed triangles, but does not remove the leakage at larger $k_1$."],"supporting_citations":[{"why":"Proposed the three-visibility correlation used here to estimate the 21-cm bispectrum and predicted its sign from ionized bubbles.","marker":"Bharadwaj & Pandey 2005"},{"why":"Paper I: validated the angular bispectrum estimator on simulated visibilities that match the MWA baseline and flagging configuration.","marker":"Gill et al. 2025"},{"why":"Paper II: generalized the estimator to multi-frequency data, producing the MABS and cylindrical bispectrum estimator validated on the exact flagging pattern.","marker":"Gill & Bharadwaj 2025"},{"why":"Provides the cylindrical power spectrum from the same MWA data, used to demonstrate the wedge analogy and the periodic spikes.","marker":"Elahi et al. 2025"},{"why":"Established the foreground wedge in the cylindrical power spectrum that this paper extends to the bispectrum.","marker":"Datta et al. 2010"},{"why":"Showed how baseline migration confines foregrounds to a wedge, the same geometry used here.","marker":"Parsons et al. 2012"},{"why":"Earlier attempt to measure the EoR 21-cm bispectrum from MWA data, providing the context and comparison for the present upper limits.","marker":"Trott et al. 2019"},{"why":"Describes the MWA drift-scan observations (project G0031) that supply the data analyzed here.","marker":"Patwa et al. 2021"}],"fun_headline_variants":["Bispectrum wedge emerges in MWA 21-cm data","First 21-cm bispectrum limits from MWA at z=8.2","EoR window for bispectrum, but spikes lurk","MWA bispectrum puts 2σ squeeze on EoR signal"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The foreground-dominated measured values of $\\Delta^3$ are treated as 2σ upper limits on the EoR 21-cm bispectrum, which assumes the foreground bispectrum does not partially cancel the EoR signal; if the foregrounds and the signal had comparable amplitudes with opposite signs, the quoted limits would not bound the true signal.","fun_headline_variants_meta":{"raw":{"variants":["Bispectrum wedge emerges in MWA 21-cm data","First 21-cm bispectrum limits from MWA at z=8.2","EoR window for bispectrum, but spikes lurk","MWA bispectrum puts 2σ squeeze on EoR signal"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000687,"raw_usage":{"total_tokens":3345,"prompt_tokens":1406,"completion_tokens":1939,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":1022,"completion_tokens_details":{"reasoning_tokens":1856}},"tokens_in":1022,"tokens_out":1939,"duration_ms":15385,"temperature":1.0,"reasoning_tokens":1856,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T19:36:10.232350+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Inject a synthetic EoR 21-cm bispectrum of known amplitude (e.g., ~$10^3$ mK$^3$) into the flagged visibility data alongside the real foregrounds and rerun the full pipeline; if the recovered $\\Delta^3$ does not track the injected signal in the A3 region, the upper-limit interpretation is falsified. Alternatively, measure the foreground-only bispectrum from a smooth-spectrum sky model with the same flagging pattern and check its sign: a negative foreground bispectrum comparable in amplitude to the measured $\\Delta^3$ would invalidate the limits.","supporting_citations":[],"review_version":1}