{"id":"fa868900-3960-4390-9b23-5b5d97f3c08f","arxiv_id":"1908.07008","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A data-driven IceTop veto applied to 5 years of IceCube data reduces the cosmic-ray muon background to a passing fraction of about 2e-5 to 5e-6 and identifies a few candidate neutrino events that need simulation follow-up.","lead":"For five years of IceCube data, this study uses the surface IceTop array to reject cosmic-ray muons and pick out a handful of tracks that look like neutrinos. It reports a veto that leaves only about two to five events per million, which could open a new way to find southern-sky neutrinos and shape future detector designs.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The neutrino-like PDF's forced-trigger background replacement is untested; if forced-trigger hits misrepresent accidental IceTop noise, the quoted background reduction is not established.","rationale":"The central claim is a measured veto passing fraction; every number in Fig. 8 is produced by applying Lambda_cut to data, and Lambda_cut is calibrated to retain 80% of Hnu. Hnu is therefore the fulcrum of the analysis. If the forced-trigger replacement is biased, the cut is miscalibrated and the quoted reduction is not a measurement of anything well-defined. I do not see a decisive flaw: the method is reasonable, the self-subtraction in Eq. (3.2) addresses overfitting, and year-by-year consistency is a useful cross-check. But the equivalence of forced-trigger and in-ice-readout backgrounds is asserted, not shown, and the text's own caveats about needing thorough simulations underscore that the template has not been validated. The reader flagged the same assumption; my check would settle it with data already in hand. The lack of statistical uncertainties on the 5 and 2 passing events is also real, but it affects the precision of the claim, not its validity; the template question is more load-bearing. Therefore I recommend the verdict remain conditional (UNCHANGED from the reader's CONDITIONAL): the concern is concrete and testable, not a demonstrated failure.","tokens_in":5349,"tokens_out":8361,"duration_ms":93701,"concrete_test":"In the same in-ice muon data, isolate a control sample of accidental IceTop hits that should be uncorrelated with the shower, e.g., hits with |tres| > 500 ns relative to the expected shower front or with perpendicular distance d > 300 m from the reconstructed track axis. Compare the (rho, tau, delta) distribution of these hits to the forced-trigger hits used for Hnu using a 3D KS test or a binned likelihood ratio. If the distributions are incompatible at, say, p < 0.01, rebuild Hnu from the control sample and recompute Lambda_cut and the passing fraction. If they agree within uncertainties, the central claim is supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 2 constructs the neutrino-like PDF Hnu by replacing the observed IceTop hits of each muon track with hits from unbiased periodic forced triggers, and Section 3 uses Hnu to set Lambda_cut at 80% retention (99.9% for log10(MuEx) >= 5.2). The headline reduction (Section 4: 2e-5 to 5e-6) is the fraction of data passing that Lambda_cut. The load-bearing assumption is that forced-trigger hits are statistically identical to the accidental IceTop noise that would accompany a true neutrino-induced muon. This is not demonstrated. In particular, tau in Eq. (3.1) is the residual time relative to the expected shower front of the reconstructed in-ice track; forced triggers are separate readout windows with their own timing, so their tau distribution could be shifted or stretched relative to hits recorded in an in-ice trigger window. Pulse processing and deadtime conditions may also differ. A biased Hnu biases the chosen Lambda_cut and therefore the measured passing fraction; the 80% retention no longer corresponds to true signal efficiency, and the quoted reduction is not a valid estimate. The paper's own call for thorough simulations (Sections 4 and 5) acknowledges that candidate event significance is unverifiable, but the background-reduction claim itself rests on the same unvalidated template.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This proceedings paper reports a data-driven veto of cosmic-ray muons in IceCube using IceTop surface hits. The authors construct cosmic-ray and neutrino-like probability density functions from data, define an IceTop log-likelihood ratio (Eq. 3.2), and set a per-bin cut that retains 80% of the neutrino-like sample (99.9% for log10(MuEx) >= 5.2). Applying the cut to five years of data split into two samples, they find 5 and 2 events passing at log10(MuEx) >= 4.8, and quote a background reduction of ~2e-5 to 5e-6 for a minimum neutrino energy of ~100 TeV. They explicitly note that thorough simulations are necessary to establish whether the passing events are astrophysical neutrinos or rare cosmic rays.","tokens_in":5660,"tokens_out":7337,"duration_ms":73479,"significance":"If established, the result would demonstrate that IceTop can veto down-going cosmic-ray muon background at the 10^-5 level, which is valuable for southern-sky neutrino searches and for designing future surface veto arrays. The analysis is commendably data-driven, uses leave-one-out subtraction to avoid overfitting, and checks consistency across two independent data samples. However, the headline reduction is based on only five and two passing events, and its validity rests on an unvalidated assumption about how the neutrino-like PDF is constructed. The paper's own call for simulations limits the interpretation of the candidate events, but the background-reduction claim itself also depends on that same unvalidated template. The quantitative result therefore needs further support before being taken at face value.","major_comments":[{"comment":"The neutrino-like PDF H_nu is generated by replacing observed IceTop hits on muon tracks with background hits from unbiased periodic forced triggers, but the manuscript does not describe how these forced-trigger hits are assigned residual times tau relative to the expected shower front of the reconstructed in-ice track. If the forced-trigger hits have a different tau distribution than the accidental IceTop noise accompanying a true neutrino-induced muon in an in-ice trigger window, H_nu is biased. Because Lambda_cut is chosen to retain a fixed fraction of H_nu, a biased H_nu changes the actual neutrino retention efficiency and therefore the measured passing fraction. The authors should either specify the overlay procedure in detail or validate the representativeness of forced-trigger hits, for example by comparing accidental-hit rates and timing distributions in in-ice triggered events with those in forced-trigger data.","section":"Section 2 (final paragraph) and Section 3, Eq. (3.1)"},{"comment":"The passing fraction is reported as '2e-5 to 5e-6' without statistical uncertainties. The underlying counts are 5 events out of 249,694 and 2 events out of 387,576; the Poisson uncertainties are on the order of ±50-100% of the central values at 68% confidence, and the two measurements are statistically consistent with a common value. The quoted range may therefore reflect Poisson fluctuation rather than a true energy-dependent variation. The paper should present the passing fraction with confidence intervals (e.g., Clopper-Pearson or Feldman-Cousins intervals) and should not claim a reduction range without these uncertainties.","section":"Section 4, Fig. 8 and last paragraph"},{"comment":"The calculation of the passing fraction assumes 'all the passing events are cosmic rays that sneak through the veto,' while the same passing events are described elsewhere as 'candidate astrophysical neutrinos.' These statements are in tension: if the passing sample contains neutrinos, the cosmic-ray passing fraction is lower than the quoted value, and the quoted reduction is an upper limit rather than a measured value. The manuscript should explicitly distinguish between an upper limit on the cosmic-ray background and a measured reduction, and should state the assumption clearly in the abstract and conclusions.","section":"Section 4, paragraph beginning 'Assuming that all the passing events...'"}],"minor_comments":[{"comment":"The phrase 'the PDFs are shown in the Figs. 3a and 3b' should be written as 'Fig. 3(a) and 3(b)' for consistency with journal style.","section":"Section 2, near Fig. 3"},{"comment":"There is a missing space in 'eventx j'; it should read 'event x_j'.","section":"Section 3, below Eq. (3.2)"},{"comment":"The sentence '5 events out of 249694 pass the cuts for log10(MuEx)>=4.8 out of which 3 passing events have log10(MuEx)>=5.0' is a run-on; it should be split into two sentences.","section":"Section 4, first paragraph"},{"comment":"The phrase 'in absence of' should be 'in the absence of'.","section":"Section 4, second paragraph"},{"comment":"The spelling 'coloured' should be made consistent with the rest of the text (either American or British spelling throughout).","section":"Section 4, figure descriptions"},{"comment":"The labels '2012-2013' and '2014-2015-2016' are slightly misleading given that the first sample extends to May 2014 and the second begins in April 2014; consider renaming them '2012-2014' and '2014-2017'.","section":"Section 4, sample labels"}],"recommendation":"major_revision","confidential_remarks":"This is an ICRC proceedings paper, so the expected level of technical detail is limited. The main substantive concern is the unvalidated forced-trigger prescription for H_nu; if the full procedure is documented in the cited earlier work (refs [6] and [11]), the authors should make that explicit and provide a brief summary here. The missing statistical uncertainties on the passing fraction are a more straightforward fix. The paper's central claim is defensible in principle, but it is not yet quantitatively established."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis short ICRC proceeding takes the IceTop veto method from refs [6,11] and applies it to five years of IceCube data. The new thing is the measured passing fraction: assuming every event that survives the veto is a cosmic ray, they get a background reduction between 2e-5 and 5e-6 for down-going tracks with a minimum neutrino energy around 100 TeV. They also list a few candidate events above log10(MuEx)=5.0, which they correctly do not call astrophysical.\n\nThe paper has real strengths. The analysis is data-driven, using only observed events and forced triggers, so it avoids simulation-dependent efficiencies. The leave-one-out subtraction in building the PDFs is careful, and they show consistency between two merged data samples. For a conference report, the method description is clear.\n\nThe soft spot is the neutrino-like template Hnu. To create it, they replace the observed IceTop hits on each muon track with hits from periodic forced triggers. The assumption is that these forced-trigger hits have the same charge, time, and distance properties as the accidental noise that would accompany a neutrino-induced muon in an in-ice readout window. That is not demonstrated. In particular, the residual time tau is measured relative to the reconstructed shower front of the track; forced triggers are recorded in a separate readout window and could have a different timing offset or pulse-processing behavior. If Hnu is biased, the Lambda_cut chosen to retain 80% of it is biased, and the quoted passing fraction is not a valid estimate of the true background reduction. The paper does not provide any cross-check of Hnu (e.g., using time-shuffled data or neutrino simulations).\n\nThere are also more mundane issues: the passing fraction is based on only five and two events, and no statistical uncertainties are quoted. The claim of 2e-5 to 5e-6 is not supported with error bars.\n\nNone of this is disqualifying for an ICRC paper, but it means the headline number should be read as a preliminary estimate, not a measurement. The candidate events are appropriately flagged as needing thorough simulations.\n\nWho should read it: anyone working on IceCube-Gen2 surface array design or on Southern Hemisphere neutrino searches. It deserves a serious referee, but a referee should ask for an independent validation of the forced-trigger template or at least an explicit discussion of its limits, plus uncertainties on the passing fraction.\n\nMy take: accept as a conference report, with revision.","headline":"The paper reports a plausible but not yet validated veto passing fraction of 2e-5 to 5e-6, resting on an untested forced-trigger template.","tokens_in":6152,"tokens_out":5212,"would_cite":true,"duration_ms":56194,"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":"IceTop, the surface array above a deep neutrino detector, can cut the atmospheric cosmic-ray background for down-going muon neutrinos to a passing fraction between $2\\times 10^{-5}$ and $5\\times 10^{-6}$ at a minimum neutrino energy near…","keywords":["IceTop veto","muon neutrino","atmospheric background","cosmic-ray muons","log-likelihood ratio","data-driven PDF","neutrino candidate","surface veto efficiency"],"falsifier":"Generate a simulated sample of neutrino-induced muon tracks with a complete detector response, build the neutrino-like PDF from forced-trigger background hits exactly as in the paper, apply the same log-likelihood cut, and measure the passing fraction; if that fraction disagrees with the data-derived $2\\times 10^{-5}$ to $5\\times 10^{-6}$ beyond systematic uncertainty, the forced-trigger replacement is not representative and the measured veto efficiency is biased.","tokens_in":5165,"feed_emoji":"🧊","tokens_out":9781,"duration_ms":91639,"temperature":0.7,"pith_summary":"This paper claims that the surface array IceTop can act as a veto for the deep in-ice detector, identifying down-going muon neutrinos by the absence of shower-generated hits in the surface tanks. Using five years of data, it reports a reduction of the atmospheric cosmic-ray background to a passing fraction between $2\\times 10^{-5}$ and $5\\times 10^{-6}$ for events with a minimum neutrino energy around 100 TeV. The veto calibration is data-driven: neutrino-like events are made by taking real muon tracks and replacing their IceTop hits with forced-trigger background hits, so the result does not depend on a surface-detector simulation. A small number of high-energy tracks survive the veto and are presented as candidate astrophysical neutrinos whose nature must be settled by thorough simulation.","feed_headline":"IceTop veto cuts cosmic-ray background to 5 parts per million","feed_subtitle":"Five years of data show down-going neutrino searches can reach a passing fraction as low as 5×10⁻⁶.","key_machinery":"The load-bearing object is the IceTop log-likelihood ratio $\\Lambda_{\\nu,\\mathrm{CR}}$. For each of the 162 surface tanks, the charge $Q$ in vertical-equivalent muons, the residual time $t_{\\mathrm{res}}$ relative to the expected shower front, and the perpendicular distance $d$ from the shower axis are transformed into the coordinates $\\rho = \\log_{10}(Q/[\\mathrm{VEM}])$, $\\tau = \\mathrm{sign}(t_{\\mathrm{res}}/[\\mathrm{ns}])\\,\\log_{10}(|t_{\\mathrm{res}}/[\\mathrm{ns}]|+1)$, and $\\delta = \\log_{10}(d/[\\mathrm{m}]+1)$. Three-dimensional PDFs $H_{\\mathrm{CR}}$ and $H_{\\nu}$ are built from data in bins of muon energy proxy and zenith angle, with $H_{\\nu}$ constructed by replacing observed IceTop hits with forced-trigger background hits. The ratio of per-tank probabilities multiplied over all tanks gives the per-event score; the event's own contribution is removed from the PDF before scoring to avoid overfitting in low-statistics bins.","core_discovery":"The central claim is that IceTop achieves a reduction of $2\\times 10^{-5}$ to $5\\times 10^{-6}$ in the atmospheric background for a down-going muon neutrino sample at a minimum neutrino energy of about 100 TeV. The number is obtained from five years of data by comparing, tank by tank, the observed IceTop hit pattern with two three-dimensional probability distributions: one for cosmic-ray events and one for neutrino-like events in which recorded surface hits have been replaced by forced-trigger background hits. A log-likelihood ratio between the two hypotheses, with a cut chosen to retain 80% of neutrino-like events (99.9% in the highest-energy bins), defines which tracks are vetoed. Counting the surviving events, and conservatively assuming that all of them are cosmic rays that slipped through, gives the passing fraction. Seven events with $\\log_{10}(\\mathrm{MuEx}) \\geq 5.0$ survive across the two data samples and are highlighted as candidate astrophysical neutrinos.","pith_inferences":["Inference: If forced-trigger hits faithfully represent accidental surface noise, the same passing-fraction curve can be extended below 100 TeV by adding exposure, giving an empirical veto calibration in an energy range where simulations are least trusted.","Inference: The paper's veto score and its stochasticity parameter could be combined into a two-dimensional classifier; the seven surviving events are too few to validate it, but the distribution shapes suggest it could separate neutrino-induced single muons from surviving muon bundles.","Inference: The measured passing fraction at the current partial surface coverage is a benchmark for scaling to larger arrays: a denser surface detector should lower the passing fraction roughly in proportion to added solid-angle coverage, a scaling that simulation can test before construction."],"forward_implications":["At roughly 100 TeV the atmospheric cosmic-ray background for down-going muon neutrino tracks is suppressed to a passing fraction between $2\\times 10^{-5}$ and $5\\times 10^{-6}$, opening a data-driven route to Southern-Hemisphere neutrino searches.","The high-energy tracks that survive the veto are, under the conservative assumption that everything passing is a cosmic ray, candidate astrophysical neutrinos; detailed simulation is required before any one event can be claimed as astrophysical.","The cut can be tuned to retain a target number of events per year, so the method is positioned to feed a real-time neutrino alert stream.","Atmospheric muon neutrinos whose accompanying shower reaches the surface are also vetoed by the same footprint comparison; quantifying that rejection efficiency is identified as a follow-up requiring detailed simulations.","Because the veto logic is data-driven, it can be applied to simulated layouts of a larger surface array, giving future detector designs a realistic veto-efficiency benchmark."],"supporting_citations":[{"why":"previous iteration of this analysis that the present work completes","marker":"[5]"},{"why":"original presentation of the IceTop log-likelihood ratio method used here","marker":"[6]"},{"why":"spline-fit reconstruction algorithm that supplies the muon track directions correlated with IceTop hits","marker":"[7]"},{"why":"neutrino event generator used to validate the selection and produce simulated neutrino events","marker":"[8]"},{"why":"MuEx muon energy proxy on which all energy bins and cuts are defined","marker":"[9]"},{"why":"astrophysical neutrino flux used to weight simulated neutrinos when comparing expected rates","marker":"[10]"},{"why":"further description of the log-likelihood ratio approach this analysis adapts","marker":"[11]"},{"why":"cosmic-ray spectrum model used to weight simulated air showers in the background comparison","marker":"[12]"}],"fun_headline_variants":["IceTop veto slashes cosmic-ray background to 5e-6","Down-going neutrino search reaches 5e-6 passing fraction","IceTop veto: 5e-6 background in five years of data","IceTop veto enables neutrino search with 5e-6 background"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The neutrino-like surface pattern is built by replacing the IceTop hits on real muon tracks with background hits from forced triggers, so the whole veto calibration rests on those forced-trigger hits being a faithful sample of the accidental surface noise that would accompany a neutrino-induced muon.","fun_headline_variants_meta":{"raw":{"variants":["IceTop veto slashes cosmic-ray background to 5e-6","Down-going neutrino search reaches 5e-6 passing fraction","IceTop veto: 5e-6 background in five years of data","IceTop veto enables neutrino search with 5e-6 background"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000521,"raw_usage":{"total_tokens":2502,"prompt_tokens":907,"completion_tokens":1595,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":523,"completion_tokens_details":{"reasoning_tokens":1518}},"tokens_in":523,"tokens_out":1595,"duration_ms":15291,"temperature":1.0,"reasoning_tokens":1518,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:28:39.844988+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Generate a simulated sample of neutrino-induced muon tracks with a complete detector response, build the neutrino-like PDF from forced-trigger background hits exactly as in the paper, apply the same log-likelihood cut, and measure the passing fraction; if that fraction disagrees with the data-derived $2\\times 10^{-5}$ to $5\\times 10^{-6}$ beyond systematic uncertainty, the forced-trigger replacement is not representative and the measured veto efficiency is biased.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"previous iteration of this analysis that the present work completes"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"original presentation of the IceTop log-likelihood ratio method used here"},{"cited_title":"Ahrens et al., Nucl","cited_arxiv_id":null,"evidence_quote":"spline-fit reconstruction algorithm that supplies the muon track directions correlated with IceTop hits"},{"cited_title":"Gazizov and M","cited_arxiv_id":null,"evidence_quote":"neutrino event generator used to validate the selection and produce simulated neutrino events"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"MuEx muon energy proxy on which all energy bins and cuts are defined"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"astrophysical neutrino flux used to weight simulated neutrinos when comparing expected rates"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"further description of the log-likelihood ratio approach this analysis adapts"},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"cosmic-ray spectrum model used to weight simulated air showers in the background comparison"}],"review_version":1}