{"id":"c0788206-705f-4e22-b1d9-348546d91235","arxiv_id":"1908.07706","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"No significant IceCube neutrinos were found in time coincidence with the first 33 gravitational wave events, and 90% upper limits were placed on their neutrino flux and emitted energy.","lead":"IceCube found no neutrino signal from the first 33 gravitational wave events detected by LIGO and Virgo, and set upper limits on how much neutrino energy each event could have emitted. The result narrows the possible role of compact binary mergers as high-energy neutrino sources and supports coordinated multi-messenger follow-up.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Missing catalog-level trials correction weakens the 'no correlation for any of 33' claim, though a Bonferroni check likely restores it.","rationale":"I read the paper as a catalog null result: for each of the 33 LVC events, an unbinned likelihood search within +/-500 s finds no neutrino excess above background, and 90% C.L. upper limits are placed. The internal likelihood construction is standard and the empirical p-values are based on 30,000 randomized trials, which is adequate for the reported values. The most significant single-event p-value is 0.014; with 33 events this does not constitute evidence after any standard multiple-testing correction, so the central conclusion is likely correct. However, the paper never performs or reports such a correction, and the phrase 'no significant correlation ... for any' is global. The reader's CONDITIONAL verdict is appropriate: the conclusion is sound but the statistical statement should be completed. The time-window assumption is a real astrophysical scope limitation acknowledged by the authors, but it does not undermine the claim as scoped; it only limits generalization to longer timescales. I therefore do not change the reader's verdict.","tokens_in":5946,"tokens_out":16130,"duration_ms":166203,"concrete_test":"Apply a Holm-Bonferroni correction to the 33 per-event p-values in Table 1, or alternatively compute 1-(1-p_min)^33 for the smallest p-value. If the corrected p-value remains >0.05, the paper's no-signal conclusion is unchanged; if it falls below 0.05, the 'no significant correlation for any' claim must be revised and S190728q would need to be reported as a candidate. This is a one-line computation from the existing table.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing weakness is the absence of a trials correction for the 33 simultaneous searches when the paper states in Sections 4 and 5 that 'No significant neutrino correlation was found' for 'any' of the first 33 GW events. Section 2 defines a per-event p-value from 30,000 background trials, and Table 1 lists 33 such p-values. No combined p-value or Bonferroni/Holm correction is reported. The smallest p-value, S190728q with p=0.014, would not be significant after a crude 33-trial Bonferroni correction (p about 0.46), so the conclusion probably survives; but the claim 'no significant correlation for any' is a global claim and, as written, is not backed by a corresponding global test. The +/-500 s time window (Section 2, Eq. 2.4; future work in Section 5) is an explicit and correctly scoped limitation rather than a flaw in the statistical claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a systematic search for IceCube high-energy neutrinos coincident with the first 33 gravitational-wave events detected by LIGO/Virgo through 28 July 2019. The analysis uses an unbinned maximum-likelihood ratio with LVC skymaps as spatial priors, a ±500 s time window around each trigger, and 30,000 background scrambles per event to assign p-values. For each event it reports 90% C.L. upper limits on the time-integrated flux E^2F (GeV cm^-2) and on the isotropic equivalent neutrino energy E_iso, assuming an E^-2 spectrum. No significant neutrino correlation is reported; the smallest p-value is 0.014 for S190728q. The paper also notes that the results have been circulated via GCN circulars and that longer-timescale searches are planned for future work.","tokens_in":6072,"tokens_out":8023,"duration_ms":82486,"significance":"If the null result holds, the paper provides a useful constraint on high-energy neutrino emission from compact binary mergers during the prompt phase, covering the largest sample of gravitational-wave events available at the time. The main strengths are the explicit Monte Carlo background calibration with 30,000 trials per event, the use of publicly available LVC skymaps as spatial priors, and the simultaneous reporting of flux and E_iso upper limits. The overall null conclusion appears robust: even the most significant event, S190728q with p=0.014, would not survive a simple 33-trial Bonferroni correction. The principal weaknesses are an algebraic inconsistency in the quoted test statistic, the absence of a catalog-level trials correction for the global 'no correlation' claim, and the lack of a systematic-uncertainty assessment for the quoted upper limits.","major_comments":[{"comment":"The simplification of the likelihood ratio is incomplete. Starting from the likelihood in Eq. (2.1), the ratio L(ns,hat gamma)/L(ns=0) also contains a Poisson normalization factor e^{-ns}(1+ns/nb)^N, which contributes 2N ln(1+ns/nb) to the log-likelihood-ratio test statistic. The expression shown in Eq. (2.4), Lambda = 2(-ns + sum ln[1+ns Si/(nb Bi)]) + 2 ln w, is therefore not the log-likelihood ratio of Eq. (2.1). Since the subsequent p-values and sensitivity limits are computed from this test statistic, the equation should be corrected, or the statistic should be explicitly defined as an unbinned likelihood ratio with the number of events N fixed. It should also be confirmed that the Monte Carlo trials use the same definition as the equation.","section":"Section 2, Eq. (2.4)"},{"comment":"The paper states in Sections 4 and 5 that 'No significant neutrino correlation was found' for the 33 events, but it reports only per-event p-values and does not provide a combined p-value or any multiple-trials correction. S190728q has p=0.014, which is significant at the 5% level for a single search; the quoted global conclusion therefore relies implicitly on a trials correction that is never stated. A simple Bonferroni correction for 33 independent searches would give corrected p approx 0.46, so the null conclusion is probably robust, but the global claim should be backed by either a combined test statistic or an explicit statement of the trials-corrected threshold.","section":"Table 1 and Sections 4-5"},{"comment":"The upper limits on E^2F and E_iso are presented without any assessment of systematic uncertainties. These quantities depend on detector angular resolution, energy scale, effective area, and the assumed E^-2 spectrum, and at least the dominant systematic contributions should be quantified or referenced, or the limits should be clearly labeled as preliminary. The figures carry the 'IceCube Preliminary' label, but the text does not identify which uncertainties are included in the quoted numbers.","section":"Section 4"}],"minor_comments":[{"comment":"The description of background trials says neutrinos are randomized by 'scrambling their arrival time and recomputing their direction based on their new randomly assigned time'; this should clarify that the reconstructed direction is re-projected to the detector coordinates appropriate to the scrambled time, rather than being itself randomly redrawn.","section":"Section 2, paragraph after Eq. (2.4)"},{"comment":"The FAR column entries such as '<1.00 x 10^-7' should include the unit yr^-1 for consistency with the column header and with the other rows.","section":"Table 1"},{"comment":"The statement that Eiso is tuned so that the mean expected number of neutrino events at IceCube is 2.3 should state explicitly whether the expected background in the search window is included in this calculation; if the 2.3 threshold is used regardless of background, the resulting limit is conservative but the assumption should be stated.","section":"Section 4, Eiso paragraph"},{"comment":"The captions describe GW170104 as an example but do not explain the axes or units; adding axis labels in the figures or a sentence in the caption would make the background distribution and sensitivity curves easier to interpret.","section":"Figures 1 and 2"}],"recommendation":"major_revision","confidential_remarks":"This is an ICRC2019 proceedings paper, and some of the issues are typical of short conference contributions. The main concerns are the algebraic inconsistency in Eq. (2.4) and the unstated trials correction behind the global null claim; both are fixable and do not appear to invalidate the qualitative conclusion. The systematic-uncertainty gap should also be addressed if the limits are to be cited as archival values."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis is a clean, standard null result from IceCube: no neutrinos associated with any of the first 33 GW events, with 90% upper limits on E^2F and E_iso per event. The genuinely new piece is the uniform treatment of all O3 events through July 2019; prior searches stopped at O2 or focused on individual events like GW170817. The method is textbook IceCube: an unbinned maximum likelihood with the LVC skymap as spatial prior, 30k scrambled-time background trials, and signal injection for sensitivity. The authors compare with an independent Bayesian analysis and find consistent results. They also flag the ±500 s window as a limitation and point to future longer-timescale searches, especially for BNS kilonova phase. That is honest, good practice.\n\nThe soft spot is the trials correction. Section 2 defines a per-event p-value, and Section 4 announces 'No significant neutrino correlation was found' as though it were a single global statement across all 33 searches. Table 1 lists 33 per-event p-values with no combined p-value and no Bonferroni or Holm correction. The smallest p, S190728q at 0.014, becomes about 0.46 after a crude 33-trial Bonferroni, so the conclusion almost certainly survives. But the global claim, as written, is not backed by a global test. This is a presentation gap, not a statistical failure. A short revision adding a combined p-value or a trials-corrected threshold would fix it.\n\nMinor issues: there are no systematic uncertainties quoted on the limits (e.g., from angular resolution or skymap systematics), and a few p-values are exactly 1.0, which is probably rounding but worth a check. The E_iso derivation via a mean of 2.3 expected events corresponds to the standard Poisson 90% upper limit, so that step is fine.\n\nOverall, this is a useful incremental contribution. It gives the community a single reference for IceCube limits on the first 33 GW events, and it is methodologically sound. I would send it to peer review; it should be accepted after a minor revision that adds the multiple-comparison correction and, ideally, a sentence on systematics. I would cite it as the standard IceCube search for that sample.","headline":"Clean null search from IceCube covering all first 33 GW events; conclusions hold up, but the global claim needs a trials correction that is missing.","tokens_in":6644,"tokens_out":3692,"would_cite":true,"duration_ms":100984,"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":"This paper searches for neutrinos from the first 33 gravitational-wave events and finds no significant correlation, placing 90% confidence upper limits on their time-integrated flux and isotropic equivalent energy.","keywords":["gravitational waves","neutrino astronomy","IceCube","multi-messenger astronomy","compact binary mergers","upper limits","transient searches"],"falsifier":"Take one of these 33 events and find a neutrino within its ±500 s window with a trial-corrected p-value small enough to pass the detection threshold; alternatively, rerun the same likelihood analysis on the same events with windows of hours to days and find a significant excess that the short window missed. Either result would undercut the paper's null conclusion and its reliance on the ±500 s assumption.","tokens_in":5712,"feed_emoji":"🔭","tokens_out":5509,"duration_ms":270365,"temperature":0.7,"pith_summary":"The paper searches for high-energy neutrinos arriving within ±500 seconds of each of the first 33 confirmed gravitational-wave events, using each event's sky localization as a spatial prior. It finds no significant neutrino correlation for any event, and derives 90% confidence upper limits on the time-integrated neutrino flux and on the isotropic-equivalent neutrino energy, assuming a power-law spectrum with index $E^{-2}$. If true, this means the prompt merger phase of these systems emitted less neutrino energy than the quoted limits, and that detecting neutrinos from compact mergers will require either more sensitive detectors or longer search windows. The smallest reported p-value in the table is 0.014, which the analysis does not treat as a detection.","feed_headline":"No neutrinos matched the first 33 gravitational-wave events","feed_subtitle":"IceCube sets 90% limits on neutrino energy from all 33 mergers, tightening the multi-messenger hunt.","key_machinery":"The load-bearing mechanism is an unbinned maximum-likelihood ratio test. Each neutrino event contributes a signal probability density that combines a spatial term and an energy term for a trial spectral index $\\gamma$, while the gravitational-wave sky map enters as a per-pixel prior weight $w = P_{\\mathrm{GW}}(x_s)/\\Omega_{\\mathrm{pixel}}$. The modified test statistic $\\Lambda = \\mathrm{TS} + 2\\ln w$ is maximized over all pixels, and the resulting significance is calibrated against 30,000 background trials with scrambled neutrino arrival times; sensitivity and upper limits are set by injecting Monte Carlo signal neutrinos with an $E^{-2}$ power-law spectrum.","core_discovery":"The paper reports a null result across the first 33 compact binary mergers detected by gravitational-wave observatories through July 28, 2019. In a ±500 s window centered on each gravitational-wave trigger, no neutrino event shows a statistically significant correlation with the source localization, and the smallest reported p-value (0.014) is still consistent with background. For every event the analysis sets 90% confidence upper limits on the time-integrated neutrino flux $E^{2}F$ under an $E^{-2}$ spectrum assumption and on the isotropic-equivalent neutrino energy $E_{\\mathrm{iso}}$, with values derived using the distance and sky localization information provided with each event. The paper's claim is that prompt neutrino emission from these mergers, if any, is below the quoted limits.","pith_inferences":["A stacked search that combines all 33 events in one likelihood would be more sensitive to a population of faint emitters than the per-event limits reported here; the paper leaves this as a natural extension.","If kilonova-phase neutrinos are delayed by hours to days, the null result says nothing about the dominant neutrino output of binary neutron star mergers, and the same machinery could be rerun with widened windows.","Models predicting prompt isotropic-equivalent neutrino energies above the individual 90% limits can be ruled out with the quoted numbers; comparing model predictions event-by-event would convert the table into astrophysical constraints.","As gravitational-wave localizations improve with more detectors and smaller sky areas, the same method should yield correspondingly tighter per-event neutrino flux limits, so future observing runs are a direct sensitivity upgrade."],"forward_implications":["The quoted 90% upper limits apply only to the 1000 s window centered on each trigger, so they do not constrain neutrino emission that occurs later.","A coincident neutrino with a sufficiently small trial-corrected p-value could improve source localization compared to gravitational waves alone; none of the 33 events provides such a localizing event.","With gravitational-wave detection rates near one per week for binary black holes, the analysis can grow to a population sample of dozens of mergers, enabling per-class constraints on binary black hole and binary neutron star neutrino emission.","Extending the same likelihood search to longer time windows, particularly the kilonova phase of binary neutron star mergers, may test emission scenarios the present ±500 s window cannot."],"supporting_citations":[{"why":"provides the catalog of 33 gravitational-wave events being searched","marker":"[1]"},{"why":"supplies the event classifications (binary black hole, binary neutron star, terrestrial) from the online database","marker":"[2]"},{"why":"defines the IceCube Gamma-ray Follow-Up track sample used as the neutrino dataset","marker":"[11]"},{"why":"provides the HEALPix pixelization used to grid the sky in the likelihood scan","marker":"[12]"},{"why":"is the basis for the conservative ±500 s coincidence window","marker":"[13]"},{"why":"presents the independent Bayesian joint-search whose similar results support the null conclusion","marker":"[15]"}],"fun_headline_variants":["IceCube sees no neutrinos from 33 gravitational-wave events","33 gravitational-wave events, zero neutrinos for IceCube","No neutrinos from 33 mergers: IceCube's null result","IceCube: no neutrino emission from first 33 gravitational-wave events"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The search assumes that any neutrino emission from these mergers arrives within ±500 seconds of the gravitational-wave signal, so if the neutrinos are emitted later, the reported limits do not constrain them.","fun_headline_variants_meta":{"raw":{"variants":["IceCube sees no neutrinos from 33 gravitational-wave events","33 gravitational-wave events, zero neutrinos for IceCube","No neutrinos from 33 mergers: IceCube's null result","IceCube: no neutrino emission from first 33 gravitational-wave events"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000567,"raw_usage":{"total_tokens":2621,"prompt_tokens":819,"completion_tokens":1802,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":435,"completion_tokens_details":{"reasoning_tokens":1730}},"tokens_in":435,"tokens_out":1802,"duration_ms":12977,"temperature":1.0,"reasoning_tokens":1730,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:58:19.004010+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take one of these 33 events and find a neutrino within its ±500 s window with a trial-corrected p-value small enough to pass the detection threshold; alternatively, rerun the same likelihood analysis on the same events with windows of hours to days and find a significant excess that the short window missed. Either result would undercut the paper's null conclusion and its reliance on the ±500 s assumption.","supporting_citations":[{"cited_title":"Grace DB","cited_arxiv_id":null,"evidence_quote":"supplies the event classifications (binary black hole, binary neutron star, terrestrial) from the online database"},{"cited_title":"Kintscher, J","cited_arxiv_id":null,"evidence_quote":"defines the IceCube Gamma-ray Follow-Up track sample used as the neutrino dataset"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the HEALPix pixelization used to grid the sky in the likelihood scan"},{"cited_title":"Baret et al., Astropart","cited_arxiv_id":null,"evidence_quote":"is the basis for the conservative ±500 s coincidence window"},{"cited_title":"Keivani et al., PoS(ICRC2019)930 (these proceedings)","cited_arxiv_id":null,"evidence_quote":"presents the independent Bayesian joint-search whose similar results support the null conclusion"}],"review_version":1}