{"id":"7a303bbc-7794-4cac-a159-83101db369b7","arxiv_id":"2608.11119","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"For a two-detector Cosmic Explorer network, the cross-correlation estimator of the binary black hole background has skewness 0.31 and excess kurtosis 1.3 at 20 Hz, a non-Gaussianity that will matter for next-generation stochastic searches.","lead":"This paper calculates the full statistical distribution of the standard gravitational-wave background estimator, not just its average. It finds the estimator stays Gaussian for LIGO's next run but becomes skewed and fat-tailed for black-hole mergers in the planned Cosmic Explorer, which future data analysis will need to handle.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The headline BBH non-Gaussianity values at CE rest on an arbitrary z_min=0.2 cutoff; the paper cites robustness without reporting the sensitivity scan that would establish it.","rationale":"The compound Poisson cumulant formalism is standard, and the paper's decomposition into intensity, shot noise, and polarization leakage is well motivated. The numerical integration with vegas is appropriate for the heavy-tailed integrand. However, the headline quantitative prediction—non-Gaussianity at CE for BBH—depends directly on the arbitrary z_min=0.2 cutoff because the fourth moment of X_f diverges at z=0 and is dominated by nearby luminous mergers. The paper explicitly calls the cutoff 'somewhat arbitrary' and defers optimization, citing an external reference instead of demonstrating robustness in the present analysis. This matches the reader's weakest_assumption, so agreement is 'agree'. A dedicated sensitivity scan over z_min, or better, a realistic notching model, would settle whether gamma2=1.3 is a robust prediction or an artifact of the cutoff. Until then, the conditional verdict is appropriate.","tokens_in":19713,"tokens_out":6178,"duration_ms":55023,"concrete_test":"Recompute gamma1 and gamma2 for the BBH background under the CE network at f=20 Hz (and optionally across the band) for z_min = 0.05, 0.1, 0.2, 0.35, and 0.5, keeping all population and rate choices fixed. If the fractional change in gamma2 between z_min=0.1 and 0.35 exceeds ~50% (or gamma1 changes comparably), the quoted values are not robust to the cutoff and the central claim needs qualification. A stronger version would replace the redshift cutoff with an SNR-based notching selection function for CE and recompute the cumulants.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that gamma1=0.31 and gamma2=1.3 for the BBH background at CE and f=20 Hz is computed from Eq. (15), kappa_n(Cbar_f)=<N_tot>/N_seg^n <(X_f)^n> + kappa_n(nbar_f), using moments evaluated with a hard lower redshift cutoff z_min=0.2 (Sec. III C). These moments diverge as z->0, and the n=4 moment is dominated by the nearest, loudest events, so gamma2 is particularly sensitive to where the cutoff is placed. The paper asserts that results are robust for z_min below O(0.5) and cites Ref. [34], but presents no scan over z_min in this work. Moreover, z_min is only a proxy for the actual notching/resolvability threshold at CE; real resolvability is SNR- and mass-dependent, so a single redshift cutoff does not define the residual background that would remain after event subtraction. If the effective residual retains more near events than z_min=0.2 implies, the non-Gaussianity is larger than quoted; if it removes more, the headline kurtosis overstates the residual non-Gaussianity. This is a modeling choice, not a flaw in the compound Poisson derivation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper computes the cumulants up to fourth order of the frequency-domain cross-correlation estimator \\bar{C}_f for a stochastic gravitational-wave background produced by compact binary coalescences. The estimator is decomposed into mean intensity, geometrical shot noise, and polarization leakage, and the cumulants are obtained from a compound-Poisson master equation, Eq. (15), with population inputs taken from external LVK fits. Numerical results are presented for BBH, BNS, and NSBH populations under LIGO A+ (O5) and two-detector Cosmic Explorer configurations. The central claim is that \\bar{C}_f remains effectively Gaussian for all three populations at O5 and for BNS/NSBH even at CE, while for the BBH background at CE the estimator becomes non-Gaussian, with skewness \\gamma_1=0.31^{+0.04}_{-0.03} and excess kurtosis \\gamma_2=1.3^{+0.4}_{-0.3} at f=20 Hz. The paper argues that an Edgeworth expansion is adequate as a leading-order description and that notching resolved BBH events is a viable alternative in the next-generation era.","tokens_in":19974,"tokens_out":2777,"duration_ms":26385,"significance":"If the quantitative claims hold, this is a timely and useful result for the stochastic-background community: it validates Gaussian likelihood frameworks for the upcoming O5 runs, quantifies when those frameworks break down for next-generation detectors, and identifies the physical origin of the non-Gaussianity in terms of shot noise and polarization leakage. The central derivation is parameter-free in the sense that the cumulant relation, Eq. (15), is an exact property of the compound-Poisson model and is not fitted to the target non-Gaussianity; the inputs are external LVK population fits. The paper also performs useful checks of the neglected off-diagonal and self-interaction terms through the R\\tau criterion in Sec. III B and uses adaptive Vegas integration to handle heavy-tailed moments. The main weakness is the treatment of the lower-redshift cutoff z_min=0.2, which is not subjected to a dedicated sensitivity scan in this work and directly affects the headline BBH kurtosis value.","major_comments":[{"comment":"The headline result \\gamma_2=1.3^{+0.4}_{-0.3} for the BBH background at CE depends directly on the imposed lower-redshift cutoff z_min=0.2. As the paper itself notes, the moments \\langle (X_f^{(i)})^n\\rangle diverge as z\\to 0, and the fourth moment is dominated by the nearest and loudest events. The manuscript asserts robustness for z_min below O(0.5) and cites Ref. [34], but it does not present a z_min sensitivity scan, nor does it quantify how the 90% credible intervals on \\gamma_1 and \\gamma_2 would shift for, say, z_min=0.1 or z_min=0.35. Because z_min is also only a proxy for the actual SNR- and mass-dependent resolvability/notching threshold, the quoted non-Gaussianity is a modeling choice rather than a prediction for a uniquely defined residual background. I request a dedicated scan over z_min (including the value z_min=0.35 studied in Ref. [34]) and a discussion of how the headline values and their credible intervals change.","section":"Sec. III C, Eqs. (15), (27)-(28); Fig. 5"},{"comment":"The fractional-contribution plots for \\kappa_3 and \\kappa_4 show that the diagonal terms (mean intensity, shot noise, polarization leakage) do not account for the full cumulant: below 25 Hz the sum of the three diagonal contributions is only about 45% of \\kappa_3, with the remaining 55% attributed to cross-terms. The text describes this briefly, but it would strengthen the paper to state explicitly how the cross-terms are computed and to verify that the numerical evaluation of the full \\kappa_n via Eq. (15) is stable in the frequency region where the cross-terms dominate. Without this detail, the reader cannot assess whether the reported \\gamma_1(f) and \\gamma_2(f) values, especially their oscillatory high-frequency behavior, are robust.","section":"Sec. III E, Figs. 4-5"},{"comment":"For the BNS background, the paper acknowledges that the neglect of off-diagonal (overlap) and self-interaction terms may break down below about 8-12 Hz, where R\\tau reaches the marginal and dangerous zones. The conclusions for BNS non-Gaussianity are nevertheless stated as robust, relying on the argument that the induced error would need to reach O(10^4) to affect the quoted \\gamma_2 bound. This is a reasonable order-of-magnitude argument, but it is not a calculation; since the BNS background is the one most affected by the approximation, the caveat should be stated in the abstract or conclusions with the same prominence as the BBH result, rather than only in the body.","section":"Sec. III F and Sec. IV"}],"minor_comments":[{"comment":"The acknowledgment text contains a typo: the Minnesota Supercomputing Institute URL is followed by 'eduledgments' instead of 'Acknowledgments'.","section":"Acknowledgments"},{"comment":"References [43] and [44] appear to cite the same arXiv preprint (2209.01310 and 2209.01221) with different identifiers; please verify and consolidate to the published version if available.","section":"References [43]-[44]"},{"comment":"The text says 'Planck18 parameters' but the word appears as 'planet18' in the PDF; please fix the capitalization.","section":"Sec. III A"},{"comment":"The right panels use a broken vertical axis, but the captions do not explain the break or the scale change; adding a note would improve readability.","section":"Fig. 4 and Fig. 5 captions"},{"comment":"The notation P_{a;f}\\equiv P_{gw;f}+P_{na;f} is introduced but P_{gw;f} is not defined in Eq. (8) until later; please define it at first use for clarity.","section":"Sec. II A, Eq. (8)"}],"recommendation":"major_revision","confidential_remarks":"The paper is within scope for the journal and the central compound-Poisson derivation appears sound. The main reservation is the z_min sensitivity: the headline kurtosis is tail-dominated and the paper defers the robustness check to a prior work. This is fixable with a modest numerical scan and should not require a change to the overall framework. I would also encourage the editor to ask the authors to clarify whether the 90% credible intervals on \\gamma_1 and \\gamma_2 include population-model uncertainties beyond the local merger rate, since the abstract's wording could be read as a full propagation of all astrophysical uncertainties."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a real contribution, not a repackaging. The paper gives the first systematic computation of third and fourth cumulants of the frequency-resolved cross-correlation estimator for all three CBC populations, across O5 and Cosmic Explorer, and the headline result—Cbar_f stays effectively Gaussian for O5 but the BBH background at CE has skewness 0.31 and excess kurtosis 1.3 at 20 Hz—is something the field needs to know. The derivation is parameter-free: compound Poisson moments fed by external LVK population fits, no fitting to the target non-Gaussianity. The three-component decomposition (mean intensity, geometrical shot noise, polarization leakage) is clean, and they check their approximations quantitatively via the Rtau calculation, showing where off-diagonal and self-interaction terms can be neglected. That's good work.\n\nSoft spots: the z_min=0.2 cutoff is the real issue. The moments diverge as z->0, and the fourth-order moment is dominated by the nearest, loudest events, so the quoted gamma2=1.3 depends directly on where you put this cutoff. The authors assert robustness for z_min below ~0.5 and cite their earlier paper [34], but they do not show a sensitivity scan here. The 90% credible intervals they quote only cover the local merger rate, not the cutoff choice or population-model uncertainty. That's a genuine gap, though I don't think it overturns the qualitative conclusion. Also, z_min is only a proxy for the actual notching threshold; real resolvability is SNR- and mass-dependent, so the residual background after subtraction could be different. But the paper is transparent about this.\n\nOne more caveat: the neglect of frequency correlations is flagged but not quantified beyond the statement that posteriors will be overly wide. That's fine for this paper, but it's a warning for anyone using the results.\n\nWho benefits: anyone building stochastic background likelihoods for next-generation detectors, and the popcorn/non-Gaussian background community. The paper deserves a serious referee. My recommendation: send it to review, and ask for a z_min scan (or at least a figure showing the dependence) before acceptance. That's a minor-to-moderate revision.","headline":"A solid, useful forecast: the cross-correlation estimator stays Gaussian at O5 but becomes non-Gaussian for BBH at Cosmic Explorer, with the main caveat being an arbitrary low-redshift cutoff that deserves a sensitivity scan.","tokens_in":20530,"tokens_out":3213,"would_cite":true,"duration_ms":27667,"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":"The paper claims that for Cosmic Explorer the cross-correlation estimator from binary black holes becomes non-Gaussian, breaking the Gaussian likelihood assumption used in stochastic background searches.","keywords":["gravitational-wave background","cross-correlation estimator","non-Gaussianity","cumulants","compound Poisson process","Cosmic Explorer","compact binary coalescences","Edgeworth expansion"],"falsifier":"Recompute the BBH cumulants at Cosmic Explorer with $z_{\\min}$ varied from 0.05 to 0.5 under the same population model: if the excess kurtosis at 20 Hz moves from the quoted 1.3 to below 0.3 or above 3, or if a direct histogram of simulated $\\bar C_f$ values at 20 Hz shows the Edgeworth density becoming negative in the bulk, the central claim about non-Gaussianity needs revision.","tokens_in":19494,"feed_emoji":"🔭","tokens_out":5499,"duration_ms":44102,"temperature":0.7,"pith_summary":"This paper asks whether the cross-correlation estimator $\\bar C_f$ used to search for the stochastic gravitational-wave background can still be treated as Gaussian when the background is produced by compact binary coalescences. The authors derive the estimator's cumulants up to fourth order from a compound-Poisson description of individual merger events, splitting each event's contribution into mean intensity, geometrical shot noise, and polarization leakage. They find that for LIGO's upcoming O5 run the Gaussian assumption holds for black-hole, neutron-star, and mixed populations, but for a two-detector Cosmic Explorer network the binary-black-hole background makes $\\bar C_f$ measurably non-Gaussian: skewness $0.31^{+0.04}_{-0.03}$ and excess kurtosis $1.3^{+0.4}_{-0.3}$ near 20 Hz over one year. That matters because inference frameworks built on Gaussian likelihoods would misstate uncertainties, and the result shows the Edgeworth expansion is only a leading-order fix in the next-generation era.","feed_headline":"At Cosmic Explorer, black-hole background breaks Gaussian assumption","feed_subtitle":"At Cosmic Explorer, black-hole mergers push the stochastic-search statistic's skewness to 0.31 and kurtosis to 1.3.","key_machinery":"The engine is the compound-Poisson cumulant relation $\\kappa_n(\\bar C_f)=\\langle N_{\\rm tot}\\rangle N_{\\rm seg}^{-n}\\langle (X_f^{(i)})^n\\rangle_\\theta + \\kappa_n(\\bar n_f)$, which reduces the difficult many-event statistics to single-event averages. The single-event response is decomposed via Stokes parameters into three physically distinct terms: the sky-averaged mean intensity (which reproduces the standard $\\Omega_{\\rm gw}$ spectrum), the geometrical shot noise from the deviation of each event's detector response from its sky average, and the polarization leakage from finite catalog size. Adaptive Monte Carlo integration over the population models evaluates the moments, and a lower-redshift cutoff at $z_{\\min}=0.2$ makes the fourth moment converge.","core_discovery":"The central claim is that the statistical distribution of $\\bar C_f$ is not fully captured by its mean $\\Omega_{\\rm gw}$, and for the next generation of ground-based detectors the higher-order cumulants of the binary-black-hole background become too large to ignore. The key identity is $\\kappa_n(\\bar C_f)\\simeq \\langle N_{\\rm tot}\\rangle / N_{\\rm seg}^n \\langle (X_f^{(i)})^n\\rangle_\\theta + \\kappa_n(\\bar n_f)$, which ties each cumulant to the Poisson rate of mergers and the moments of a single event's detector response. Using this, the paper computes 90% credible intervals for the cumulants for the three CBC populations and finds effectively Gaussian behavior at O5 for all, but at CE a BBH-driven skewness of order 0.3 and excess kurtosis of order 1.3 in the most sensitive band, with the signal variance itself overtaking detector noise there.","pith_inferences":["The paper's compound-Poisson machinery also implies that $\\bar C_f$ values at different frequencies are positively correlated through the chirp evolution, so a full likelihood should include off-diagonal frequency covariances; the paper flags this as future work.","Because the non-Gaussianity is driven by the loudest nearby mergers, the measured skewness and kurtosis could in principle be used as an independent handle on the local merger rate or on how imperfectly resolvable events are removed.","The z_min cutoff models a residual background left after notching; a realistic notching pipeline will produce a population-dependent residual, so the CE numbers should be interpreted as a full-background upper bound on non-Gaussianity rather than a prediction for a specific analysis."],"forward_implications":["For the LIGO O5 network, current Gaussian-likelihood stochastic searches remain valid for all three compact-binary populations.","For a two-detector Cosmic Explorer network, analyses of the BBH background that keep all data must go beyond a Gaussian likelihood; neglecting the signal variance would understate uncertainties.","The Edgeworth expansion can serve as a leading-order description for the moderate non-Gaussianity, but its truncated series has limited validity in the tails.","If one notches out individually resolvable BBH signals, the remaining background can be treated as effectively Gaussian again.","BNS and NSBH backgrounds stay effectively Gaussian even at CE, so no new formalism is needed for them."],"supporting_citations":[{"why":"Sets up the cross-correlation method and the meaning of the overlap reduction function that the decomposition builds on.","marker":"[9]"},{"why":"Derives the cross-correlation estimator and its variance, the baseline the paper extends.","marker":"[12]"},{"why":"Supplies the z_min convergence argument and the residual-background analysis that justifies the redshift cutoff.","marker":"[34]"},{"why":"Provides the local merger rates and mass-distribution parameters that drive all numerical cumulant estimates.","marker":"[40]"},{"why":"Gives the general expression for the CBC energy density and shows interference terms do not affect the mean.","marker":"[42]"},{"why":"Supplies the IMRPhenomXAS waveform used to simulate each merger's strain contribution.","marker":"[63]"},{"why":"Documents the positivity and validity limits of the truncated Edgeworth expansion used to interpret the non-Gaussianity.","marker":"[73]"}],"fun_headline_variants":["Black-hole background goes non-Gaussian at Cosmic Explorer","Beyond the mean: GW background higher moments at CE","Cosmic Explorer sees skewness 0.31 in black-hole background","Next-gen GW search: kurtosis 1.3 for binary black holes"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quoted non-Gaussianity numbers rest on the $z_{\\min}=0.2$ lower-redshift cutoff imposed to make the moment integrals converge; the loudest nearby mergers dominate the fourth-order cumulant, and the paper does not include a dedicated scan over this cutoff.","fun_headline_variants_meta":{"raw":{"variants":["Black-hole background goes non-Gaussian at Cosmic Explorer","Beyond the mean: GW background higher moments at CE","Cosmic Explorer sees skewness 0.31 in black-hole background","Next-gen GW search: kurtosis 1.3 for binary black holes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000154,"raw_usage":{"total_tokens":1276,"prompt_tokens":1074,"completion_tokens":202,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":690,"completion_tokens_details":{"reasoning_tokens":129}},"tokens_in":690,"tokens_out":202,"duration_ms":2353,"temperature":1.0,"reasoning_tokens":129,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T05:50:42.062404+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the BBH cumulants at Cosmic Explorer with $z_{\\min}$ varied from 0.05 to 0.5 under the same population model: if the excess kurtosis at 20 Hz moves from the quoted 1.3 to below 0.3 or above 3, or if a direct histogram of simulated $\\bar C_f$ values at 20 Hz shows the Edgeworth density becoming negative in the bulk, the central claim about non-Gaussianity needs revision.","supporting_citations":[{"cited_title":"Expansions for nearly Gaussian distributions","cited_arxiv_id":"astro-ph/9711239","evidence_quote":"Documents the positivity and validity limits of the truncated Edgeworth expansion used to interpret the non-Gaussianity."}],"review_version":1}