{"id":"7897a0d8-b001-463e-b367-5e3c90dcdf1d","arxiv_id":"astro-ph/0005036","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"low","formal_verification":"none","parameter_count":3,"one_line_summary":"The primary CMB bispectrum generated by a single non-linear coupling parameter has sign-changing acoustic peaks, and detecting it needs f_NL above roughly 600 for COBE, 20 for MAP, and 5 for Planck.","lead":"Komatsu and Spergel compute how a small primordial non-Gaussianity would appear in the cosmic microwave background when the full radiation transfer function is used, not just the large-scale Sachs-Wolfe limit. The predicted bispectrum oscillates and changes sign, which makes it a sharper probe of early-universe physics than the simple skewness and lets MAP and Planck separate it from foregrounds by shape.","discovery_kind":"new_method","skeptic_critique":{"model":"selfhosted/deepseek-v4.1-flash","headline":"The abstract's headline [s7] rests on f_NL ~ 0.01 for slow-roll, but Sec III.A and Sec VI also state Pyne-Carroll second-order GR terms give f_NL ~ O(1) and dominate for generic slow-roll; the conclusion survives, but the abstract's 'simple slow-roll' is ambiguous about which value is operative.","rationale":"The reader's weakest_assumption (the local, factorizable ansatz with a single scale-independent f_NL) is correct and is explicitly acknowledged in Sec III.A: 'It should be stressed again that this is the special case in the slow-roll limit. If the slow-roll condition is not satisfied, then f_NL = f_NL(k1,k2,k3) at equation (15).' So the reader's concern is real but acknowledged, and the paper is transparent about it. The more load-bearing unresolved issue is the f_NL value ambiguity: the abstract's 'simple slow-roll' could mean 'simplest' (0.01) or 'generic' (1), and the paper states both values without reconciling which one the 'problematic' claim uses. The conclusion survives either way, but the margin is thinner than the abstract implies. The other flagged issues (SCDM cosmology, foreground amplitudes, dust skewness 51) are acknowledged or side remarks and do not touch the central claim. The paper's main contribution, the full transfer-function bispectrum template plus the matched-filter argument, is genuinely new and well-executed. I recommend UNCHANGED: the conclusion holds, but the abstract should be revised to state the operative f_NL value explicitly.","tokens_in":41562,"tokens_out":52092,"duration_ms":955755,"concrete_test":"Recompute the cumulative S/N in Fig. 4 and the thresholds in Table V with f_NL = 1 (the Pyne-Carroll value stated in Sec III.A and Sec VI) instead of 0.01. If S/N remains below 1 for COBE, MAP, Planck, and the ideal experiment, the 'problematic' conclusion holds but the abstract should quote f_NL ~ 1 as the operative generic slow-roll value and state the margin explicitly. If any experiment reaches S/N > 1, the headline claim weakens.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract [s7] claims detection 'should be problematic for the simple slow-roll inflationary scenarios,' implicitly using f_NL ~ 0.01 from Gangui et al. (cited). But Sec III.A, last paragraph, states: 'Even if the inflation produced Gaussian fluctuations, Pyne and Carroll pointed out that the general relativistic second-order perturbation theory would produce terms of f_NL ~ O(1) [26]. For generic slow-roll models, these terms dominate the primary non-Gaussianity.' Sec VI repeats both values: 'the simplest inflationary scenario usually predicts small f_NL (~10^-2) [11,12], and the second order perturbation theory yields f_NL ~ 1 [26].' So the paper is transparent about both values, but the abstract's 'simple slow-roll' does not say which one the 'problematic' claim uses. With f_NL ~ 1, the ideal experiment (threshold 3) and Planck (threshold 5) are only factors of 3-5 away, not 100-500. The conclusion 'problematic' survives, but the margin is far thinner than the abstract's framing implies, and the abstract should state the operative value explicitly. This is not an acknowledged limitation in the sense of 'we assume X but it might be wrong'; it is an ambiguity in the headline claim's premise.","agreement_with_reader":"partial"},"referee_report":{"model":"selfhosted/deepseek-v4.1-flash","summary":"The manuscript computes the primary cosmic microwave background angular bispectrum induced by a local-type primordial non-Gaussianity parametrized by a single coupling constant f_NL (Eqs. 11, 15), using the full radiation transfer function from CMBFAST rather than the Sachs–Wolfe approximation. The authors introduce the 'reduced' bispectrum b_{l1l2l3} (Eq. 5), which factors out the Gaunt integral and carries all physical information, and derive the primary bispectrum as a line-of-sight integral over three transfer functions (Eqs. 16, 19). Numerically they find that the reduced bispectrum has a series of acoustic peaks that alternate in sign, with an oscillation period twice that of the angular power spectrum, and an amplitude scaling as l^{-4} f_NL. They then forecast signal-to-noise for COBE, MAP, Planck, and an ideal noise-free experiment via a Fisher matrix (Eqs. 48–54), finding detection requires f_NL > 600, 20, 5, and 3 respectively, and compare this to the sensitivity of the one-point skewness (Eqs. 25, 59, 60), which requires f_NL > 800, 80, 70, and 60. Finally they estimate two secondary bispectra — the Sunyaev–Zel'dovich/weak-lensing coupling (Eqs. 33–35) and Poisson-distributed extragalactic point sources (Eqs. 37–41) — and show that MAP and Planck can separate these from the primary signal on the basis of shape, because only the primary signal oscillates and changes sign.","tokens_in":21162,"tokens_out":7644,"duration_ms":135860,"significance":"This is a substantial and, if the results hold, high-impact contribution. The reduced-bispectrum formalism of Sec. II is a genuine conceptual contribution that has become the standard language for CMB bispectrum work, and the observation that the primary bispectrum oscillates and changes sign while secondaries do not is a physically well-motivated route to foreground separation. The paper is commendably explicit about its machinery: it uses the publicly available CMBFAST transfer function, states its cosmological parameters, and quotes all forecasts per unit f_NL, so the numbers in Table V are falsifiable predictions rather than fits to data. The matched-filter argument of Sec. III.C — that a computable bispectrum template is strictly more sensitive than the coarse-grained skewness — is supported quantitatively by Fig. 5 and Table V, with the ratio of required f_NL growing from 1.3 (COBE) to 20 (ideal) as resolution improves, which is exactly what the cancellation argument predicts. The internal arithmetic of the order-of-magnitude scalings in Eqs. (55)–(58) checks out against the numerical results of Tables I–III, which is a good sign for the robustness of the forecast pipeline.","major_comments":[{"comment":"The headline claim of the abstract — that detection 'should be problematic for the simple slow-roll inflationary scenarios' — is stated without specifying which value of f_NL it rests on, and the body supplies two values that differ by two orders of magnitude. Sec. III.A states that Gangui et al. find f_NL ~ 10^{-2} for quadratic and quartic potentials, but then states that Pyne and Carroll's second-order general-relativistic terms give f_NL ~ O(1) and that 'for generic slow-roll models, these terms dominate the primary non-Gaussianity'; Sec. VI repeats both numbers. The conclusion survives either way, but the margin does not: against the Planck threshold of f_NL > 5 and the ideal-experiment threshold of f_NL > 3 (Table V), an operative value of f_NL ~ 1 leaves only a factor of a few, not the factor of 10^2–10^3 that the abstract's framing suggests. The abstract should state explicitly which value is being used for the 'problematic' claim, and should note that with the O(1) second-order contribution the Planck and ideal-experiment forecasts are marginal rather than hopeless. The same sentence in the abstract [s2] that MAP and Planck 'should be able to detect weak non-Gaussianity' is in mild tension with the paper's own conclusion and would read better if it were qualified in the same way.","section":"Abstract [s7]; Sec. III.A, final paragraph; Sec. VI, first paragraph","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"All numerical thresholds quoted in Table V and in the abstract's discussion are computed for a scale-invariant SCDM model with Omega_m = 1, Omega_Lambda = 0, Omega_b = 0.05, h = 0.5, which the authors themselves describe as 'almost excluded by current observations.' The sign-alternating acoustic peak structure of Fig. 2 is robust to this choice, but the specific f_NL thresholds are not obviously so: the peak positions and the damping-tail behaviour that set the l at which d(S/N)/d ln l saturates both shift with the acoustic scale and with the epoch of matter-radiation equality. Since the numbers in Table V are the paper's most quotable output, the manuscript should either repeat the forecast for a currently viable flat-LambdaCDM model or state explicitly, with an argument, why the thresholds are expected to be insensitive to the background cosmology at the factor-of-two level. A single sentence quantifying the expected shift would be sufficient; silence on this point leaves the headline numbers less transferable than they appear.","section":"Sec. III.B and Table V","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The secondary-bispectrum budget of Sec. IV includes only the SZ-times-lensing coupling and Poisson point sources. The lensing of the primary CMB by the same large-scale structure also generates a bispectrum through the correlation between the lensing potential and the integrated Sachs–Wolfe effect, which is not estimated anywhere in the paper. This matters specifically for the separability claim in Table IV: the Planck forecast is quoted for the 217 GHz channel, where the authors themselves note that j_nu ~ 0 and the SZ-lensing term therefore vanishes, so at that channel the ISW-lensing bispectrum may be the dominant residual secondary rather than a negligible one. Because the ISW-lensing shape is smooth and non-oscillatory, it would be partially degenerate with the SZ-lensing template and could raise the degradation parameter d_primary above the quoted 1.00. The authors should either add an order-of-magnitude estimate of the ISW-lensing bispectrum in the same style as Eq. (35), or state explicitly why it is expected to be smaller than the SZ-lensing term at the relevant scales and frequencies.","section":"Sec. IV.A, Eqs. (33)–(35); Sec. V.C and Table IV","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"The statement in Sec. V.B that 'MAP and Planck experiments will separate the primary bispectrum from others at 1% or better accuracies' conflates two different quantities. The 1% figure is the degradation parameter d_primary = 1.01 from Table IV, which measures the loss in signal-to-noise, whereas the quantity that actually measures separation is the correlation coefficient r_primary,sz-lens = -0.12 for MAP and -0.059 for Planck. The qualitative claim of the abstract [s9] is supported by the r values, but the '1% accuracy' phrasing overstates the degree of separation and should be reworded to refer to the correlation coefficients.","section":"Abstract and Sec. V.B","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The dust result — a dimensionless skewness of 51 measured from the 100 micron HEALPix map with 12,582,912 pixels and mean intensity 14.8 MJy/sr — is introduced without a beam or pixel-scale specification, and the skewness of a smoothed map is resolution-dependent. Since this number is offered as motivation for a future dust-bispectrum study, the authors should state the angular resolution at which it was computed and note explicitly that it is an upper bound that will fall substantially once the Galactic plane is masked.","section":"Sec. VI, final paragraph","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The variance in Eq. (49) carries the multiplicity factor Delta_{l1l2l3} = 1, 2, 6, while the Fisher sum in Eq. (51) runs over the ordered set 2 <= l1 <= l2 <= l3. The two conventions are consistent, but the reader has to reconstruct that the Delta factor is what converts between them. A short sentence after Eq. (51) stating this explicitly would remove the ambiguity.","section":"Eqs. (49) and (51)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The Fisher matrix entries in Tables I and II are quoted as pure numbers with no indication of their dimensions or of the normalization convention used for the reduced bispectrum b_{l1l2l3}. Since b has dimensions that depend on the convention adopted in Eq. (5), the tables should state the units, or state that the entries are dimensionless under the stated normalization.","section":"Tables I and II","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The caption does not state the integration limits or step size used for the line-of-sight integral, although the text does (c(tau_0 - 2 tau_*) <= r <= c(tau_0 - 0.1 tau_*) with step 0.1 c tau_*). Repeating these in the caption would make the figure self-contained, and would also make clear that the plotted quantity is b^primary_{l1l2l3} f_NL^{-1} rather than the full bispectrum.","section":"Fig. 2 and its caption","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Eq. (34) uses the linear power spectrum P_Phi(k), while the discussion in Sec. V.C notes that Cooray and Hu find the non-linear power spectrum enhances the effect by a factor of several at l ~ 3000. A forward reference from Eq. (34) to that caveat would prevent a reader from taking the SZ-lensing estimate as a firm amplitude prediction.","section":"Sec. IV.A, Eq. (34)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Several typographical and grammatical points should be corrected: the title reads 'Bi spectrum'; the abstract has 'On smaller scales; however,' where a comma is required; 'change a sign' should read 'change sign'; 'separatability' should read 'separability'; 'signal-to-noises' should read 'signal-to-noise ratios'; and 'the ideal noise-free and infinitesimal thin-beam experiment' should read 'infinitesimally thin beam'.","section":"Throughout","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"minor_revision","confidential_remarks":"The physics is sound and the reduced-bispectrum formalism plus the shape-based foreground-separation argument are the parts most likely to be cited. My recommendation turns on two presentational points rather than on the derivation: the abstract's f_NL premise is ambiguous in a way that materially changes how impressive the forecast looks, and the numerical thresholds are tied to a background cosmology the authors themselves call excluded. Both are fixable in a revision without new computation, though the ISW-lensing omission in Sec. IV would ideally be addressed with a short estimate rather than a sentence of hand-waving."},"author_rebuttal":null,"desk_editor":{"model":"selfhosted/deepseek-v4.1-flash","letter":"This is the paper that turned the CMB bispectrum from a Sachs-Wolfe curiosity into a working template. The core of it is Eqs. (16)-(19): the reduced bispectrum computed with the full CMBFAST transfer function rather than the large-scale limit. The result in Figure 2, that the primary bispectrum oscillates and changes sign with twice the period of the power spectrum, is the thing everyone since has built on. The matched-filter argument in Sec. III.C is the other keeper: once you know the shape, the bispectrum beats the skewness by roughly an order of magnitude, and Table V makes that concrete. The Fisher-matrix machinery in Sec. V, including the degradation parameters and correlation coefficients in Table IV, is clean and reusable. The derivation is straightforward and I don't see a step that's wrong. The numerical work uses the public transfer code with stated parameters, so it's reproducible in principle. Credit where due: this is careful, honest formalism work, and the authors flag their own weak spots in the body even though the abstract doesn't carry them. The classifier record says no limitations are stated; that's wrong. Sec. III.B says SCDM is nearly excluded, and Sec. V.C says the SZ amplitudes are order-of-magnitude only. Soft spots, in order of how much they matter. The background is SCDM with Omega_m = 1, which was already in trouble in 2000; the shape result is robust to that but the S/N numbers aren't. The foreground amplitudes are estimates, and the Fisher separation neglects sky coverage, so the 1% separation claim for MAP and Planck is optimistic. The dust skewness of 51 in Sec. VI is a single number with no sky cut and no error bar; I'd drop it or do it properly. On the stress-test note: it lands, but it's minor. The abstract's 'problematic for simple slow-roll' leans on f_NL ~ 0.01, while Sec. III.A and Sec. VI both say Pyne-Carroll second-order terms give f_NL ~ 1 and dominate for generic slow-roll. With f_NL ~ 1 the conclusion still holds, but Planck's threshold of 5 is only a factor of a few away, not hundreds. One clause in the abstract would fix it. Who should read it: anyone doing CMB non-Gaussianity, and anyone who wants the reference template. Send it to referees; the revision is light.","headline":"The paper that gave the field its local f_NL bispectrum template with full transfer functions, plus the matched-filter argument; the physics holds up, the soft spots are the toy cosmology and the order-of-magnitude foreground numbers.","tokens_in":21758,"tokens_out":1022,"would_cite":true,"duration_ms":20391,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"pith_extraction":{"msc":[],"pacs":["98.70.Vc","98.80.Cq","98.65.Dx"],"model":"selfhosted/deepseek-v4.1-flash","headline":"The paper computes the three-point statistics of the microwave background with the full radiation transfer function, shows that it oscillates through a series of acoustic peaks that alternate in sign, and uses that shape to forecast how far","keywords":["cosmic microwave background","angular bispectrum","primordial non-Gaussianity","inflationary perturbations","acoustic oscillations","radiation transfer function","secondary anisotropies","Sunyaev-Zel'dovich effect"],"falsifier":"A Planck measurement of the sign-alternating acoustic peak pattern in the foreground-cleaned bispectrum, at an amplitude implying $f_{NL}$ of order ten or larger, would confirm the template and overturn the slow-roll expectation; conversely, a primordial three-point function shown to have a momentum dependence other than the local form would invalidate the template and with it every threshold in the paper's tables.","tokens_in":22134,"feed_emoji":"🌌","tokens_out":3573,"duration_ms":65615,"temperature":0.7,"pith_summary":"The cosmic microwave background is expected to be almost perfectly Gaussian, so almost all of its information sits in the two-point function, the angular power spectrum. This paper asks what the *next* statistic would look like if the primordial fluctuations carried a small non-Gaussian component, and answers it by pushing the calculation beyond the large-angle Sachs–Wolfe regime where earlier work stopped: it evaluates the angular bispectrum $b_{l_1 l_2 l_3}$ with the full radiation transfer function, so the prediction holds down to arcminute scales where MAP and Planck operate. The result is a template with a distinctive signature — a sequence of acoustic peaks that change sign, oscillating with twice the period of the power spectrum — whose overall amplitude is set by a single coupling constant $f_{NL}$. That template then does double duty: it yields concrete detection thresholds for COBE, MAP, Planck and an ideal experiment, and it acts as a matched filter that separates the primordial signal from the bispectra produced by lensing, the Sunyaev–Zel'dovich effect, and unresolved point sources. The payoff is a sharp statement about inflation: for the simplest slow-roll models, which predict $f_{NL} = O(10^{-2})$, no experiment can see this signal, so a positive detection would be evidence that the simplest picture is wrong.","feed_headline":"Microwave sky's 3-point pattern flips sign at every peak","feed_subtitle":"The alternating acoustic peaks make the bispectrum a sharper non-Gaussianity detector than skewness, and a test of inflation.","key_machinery":"The load-bearing object is the *reduced* bispectrum $b_{l_1 l_2 l_3}$, defined by factoring the angle-averaged bispectrum into a Gaunt integral times a symmetric function of the three multipoles, $B^{m_1 m_2 m_3}_{l_1 l_2 l_3} = G^{m_1 m_2 m_3}_{l_1 l_2 l_3} b_{l_1 l_2 l_3}$. The Gaunt integral carries only the triangle conditions and selection rules, so $b_{l_1 l_2 l_3}$ holds all the physics and is the quantity plotted, fitted, and compared across models. For the local coupling ansatz $b^{primary}_{l_1 l_2 l_3} = 2f_{NL} b^L_{l_1} b^L_{l_2} b^L_{l_3} + (2f_{NL} b^L_{l_1} b^L_{l_2} b^L_{l_3}) + (2f_{NL} b^L_{l_1} b^L_{l_2} b^L_{l_3})$ — written in the paper as a radial integral over products of $b^L_l(r)$ and $b^{NL}_l(r)$, the line-of-sight projections of the linear power spectrum and of the coupling term against the transfer function. The second piece of machinery is a Fisher matrix over the reduced bispectrum, $F_{ij} = 2f_{NL} b^L_{l_1} b^L_{l_2} b^L_{l_3}$, which simultaneously gives the signal-to-noise for each component, the correlation coefficients between components, and a degradation parameter measuring how much the presence of the others weakens the measurement. That matrix is what turns a template into a separability claim.","core_discovery":"The central claim is that the primary CMB bispectrum, computed with the full transfer function rather than the large-scale Sachs–Wolfe approximation, is not a slowly varying function but an oscillating one: it has a series of acoustic peaks that alternate in sign, and the period of those oscillations is twice as long as the period of the angular power spectrum. Because the sign changes are a direct imprint of the transfer function, they are absent from the secondary bispectra — the lensing–Sunyaev–Zel'dovich coupling falls off as $l^{-3}$ and the Poisson bispectrum from extragalactic sources is flat — which is what makes the primary signal separable on shape alone. Parameterising the amplitude by a single non-linear coupling $f_{NL}$, the paper finds that a signal-to-noise of one requires $f_{NL} > 600$ for COBE, $> 20$ for MAP, $> 5$ for Planck, and $> 3$ for an ideal noise-free, infinitesimally thin-beam experiment. Since slow-roll inflation gives $f_{NL} = O(10^{-2})$, and second-order general relativity gives $f_{NL} = O(1)$, the primary bispectrum should be out of reach — which is itself the useful result, because it converts a non-detection into a constraint on departures from the simplest inflationary scenario.","pith_inferences":["Because the peak pattern depends on how the three-point function depends on momentum, measuring the alternation itself is a shape measurement, not merely an amplitude measurement — a non-local or equilateral-type primordial shape would produce a different peak structure and would not be captured by this single-parameter template.","The ideal-experiment threshold of $f_{NL} > 3$ sits just above the $f_{NL} \\sim 1$ floor that second-order general relativity guarantees, so the minimum non-Gaussianity that must exist is only a factor of a few below what a perfect all-sky experiment could reach; the guaranteed signal is marginally inaccessible by design, not merely by current technology.","The Fisher machinery is written to fit several templates at once, so the same construction could be run in reverse as a null test: projecting foreground-cleaned maps at several frequencies onto the sign-flipping template while marginalising over the smooth secondary shapes.","If the skewness genuinely saturates at $\\ell \\sim 500$, then skewness-based non-Gaussianity claims from higher-resolution data add no independent information beyond that scale, which is a testable statement about published analyses."],"forward_implications":["Planck would need $f_{NL} > 5$, MAP $> 20$, COBE $> 600$, and an ideal experiment $> 3$ to reach signal-to-noise one on the primary bispectrum, so the simplest slow-roll prediction of $f_{NL} \\sim 10^{-2}$ is undetectable by any of them.","The bispectrum is a better non-Gaussianity detector than the skewness: the skewness requires $f_{NL} > 800$, $80$, $70$, and $60$ for COBE, MAP, Planck, and the ideal experiment, because integrating the sign-changing oscillations cancels signal, and the skewness saturates at MAP resolution around $\\ell \\sim 500$.","MAP and Planck can separate the primary bispectrum from the lensing–Sunyaev–Zel'dovich and point-source bispectra at the 1% level or better, with degradation parameters of $1.01$ and $1.00$, precisely because only the primary template oscillates.","Planck should detect the lensing–Sunyaev–Zel'dovich bispectrum and the point-source bispectrum, while MAP should detect only the point-source one; COBE cannot discriminate among the three.","A significant detection of the primary bispectrum or the skewness by any experiment would mean the simplest inflationary scenario needs modification, for example through features in the inflaton potential."],"supporting_citations":[{"why":"Gangui et al. computed the primordial bispectrum in the Sachs–Wolfe regime and estimated $f_{NL} = O(10^{-2})$ for quadratic and quartic inflaton potentials; the paper's full-transfer calculation is checked against and contrasted with this result.","marker":"[12]"},{"why":"Salopek and Bond supply the inflationary origin of the non-linear coupling constant $f_{NL}$ as a combination of the slope and curvature of the inflaton potential, which is the model the paper's ansatz is built on.","marker":"[11]"},{"why":"Seljak and Zaldarriaga's CMBFAST code supplies the full radiation transfer function $g_{Tl}(k)$ used in every numerical evaluation here, and its integration algorithm is reused for the $k$-integrals.","marker":"[27]"},{"why":"Spergel and Goldberg provide the cosmic-variance estimate for the bispectrum and the Fisher-matrix approach to forecasting detectability, which the paper extends from a single component to a multi-component fit.","marker":"[20]"},{"why":"Goldberg and Spergel derive the lensing–Sunyaev–Zel'dovich cross-correlation spectrum that generates the leading secondary bispectrum the primary template must be separated from.","marker":"[21]"},{"why":"Pyne and Carroll show that second-order general relativistic perturbation theory produces $f_{NL} = O(1)$ even for Gaussian inflation, which sets the floor against which the quoted detection thresholds are judged.","marker":"[26]"},{"why":"Toffolatti et al. supply the extragalactic source counts used to estimate the Poisson bispectrum amplitude from unresolved radio and infrared point sources.","marker":"[35]"},{"why":"Knox's analytic detector-noise formula provides the $C_N^l$ entering the total power spectrum in the bispectrum variance, and hence every signal-to-noise number in the tables.","marker":"[40]"}],"fun_headline_variants":["CMB bispectrum peaks alternate sign and double the power spectrum's period","Alternating acoustic peaks make the CMB bispectrum a matched filter","Planck needs f_NL above 5 to detect the primary CMB bispectrum","Primary bispectrum's sign flips separate it from secondary signals","Slow-roll inflation predicts a CMB bispectrum too faint to detect"],"cache_read_input_tokens":0,"weakest_assumption_plain":"Everything numerical rests on the assumption that the primordial three-point function has the local, factorisable form with one scale-independent constant $f_{NL}$; the paper itself notes that if slow roll fails then $f_{NL}$ becomes a function of $k_1$, $k_2$, $k_3$, and the sign-changing peaks and all the quoted thresholds change with it.","fun_headline_variants_meta":{"raw":{"variants":["CMB bispectrum peaks alternate sign and double the power spectrum's period","Alternating acoustic peaks make the CMB bispectrum a matched filter","Planck needs f_NL above 5 to detect the primary CMB bispectrum","Primary bispectrum's sign flips separate it from secondary signals","Slow-roll inflation predicts a CMB bispectrum too faint to detect"]},"model":"selfhosted/deepseek-v4.1-flash","effort":"low","attempts":1,"cost_usd":0.0,"raw_usage":{"total_tokens":3271,"prompt_tokens":1131,"completion_tokens":2140,"prompt_tokens_details":null,"completion_tokens_details":{"reasoning_tokens":2052}},"tokens_in":1131,"tokens_out":2140,"duration_ms":37403,"image_pages":[],"temperature":1.0,"reasoning_budget":50,"reasoning_tokens":2052,"transport_provider":"selfhosted","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-10-29T18:12:15.793900+00:00","model_set":{"reader":"selfhosted/deepseek-v4.1-flash"},"falsifier":"A Planck measurement of the sign-alternating acoustic peak pattern in the foreground-cleaned bispectrum, at an amplitude implying $f_{NL}$ of order ten or larger, would confirm the template and overturn the slow-roll expectation; conversely, a primordial three-point function shown to have a momentum dependence other than the local form would invalidate the template and with it every threshold in the paper's tables.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gangui et al. computed the primordial bispectrum in the Sachs–Wolfe regime and estimated $f_{NL} = O(10^{-2})$ for quadratic and quartic inflaton potentials; the paper's full-transfer calculation is checked against and contrasted with this result."},{"cited_title":"Phys. Rev. D 42, 3936 (1990); ibid. 43, 1005","cited_arxiv_id":null,"evidence_quote":"Salopek and Bond supply the inflationary origin of the non-linear coupling constant $f_{NL}$ as a combination of the slope and curvature of the inflaton potential, which is the model the paper's ansatz is built on."},{"cited_title":"Astrophys. J. 469, 437","cited_arxiv_id":null,"evidence_quote":"Seljak and Zaldarriaga's CMBFAST code supplies the full radiation transfer function $g_{Tl}(k)$ used in every numerical evaluation here, and its integration algorithm is reused for the $k$-integrals."},{"cited_title":"Phys. Rev. D 59, 103001","cited_arxiv_id":null,"evidence_quote":"Spergel and Goldberg provide the cosmic-variance estimate for the bispectrum and the Fisher-matrix approach to forecasting detectability, which the paper extends from a single component to a multi-component fit."},{"cited_title":"Phys. Rev. D 59, 103002","cited_arxiv_id":null,"evidence_quote":"Goldberg and Spergel derive the lensing–Sunyaev–Zel'dovich cross-correlation spectrum that generates the leading secondary bispectrum the primary template must be separated from."},{"cited_title":"Phys. Rev. D 53, 2920","cited_arxiv_id":null,"evidence_quote":"Pyne and Carroll show that second-order general relativistic perturbation theory produces $f_{NL} = O(1)$ even for Gaussian inflation, which sets the floor against which the quoted detection thresholds are judged."},{"cited_title":"F. Arg¨ ueso G´ omez, G. De Zotti, P. Mazzei , A. Franceschini, L. Danese and C. Burigana, Mon. Not. R. Astron. Soc., 297, 117","cited_arxiv_id":null,"evidence_quote":"Toffolatti et al. supply the extragalactic source counts used to estimate the Poisson bispectrum amplitude from unresolved radio and infrared point sources."},{"cited_title":"Phys. Rev. D 48, 3502","cited_arxiv_id":null,"evidence_quote":"Knox's analytic detector-noise formula provides the $C_N^l$ entering the total power spectrum in the bispectrum variance, and hence every signal-to-noise number in the tables."}],"review_version":1}