{"id":"e7a9d39a-0fb0-4988-8f0b-80c1568c396b","arxiv_id":"1908.07300","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A combined analysis of ANTARES and IceCube data improves upper limits on the dark matter self-annihilation cross section for WIMP masses from 50 GeV to 1 TeV, but finds no signal.","lead":"Physicists from the ANTARES and IceCube neutrino telescopes combined their data to search for neutrinos from dark matter annihilating at the center of the Milky Way. They found no excess, and the combined search sets stronger upper limits on the dark matter annihilation rate for WIMP masses between 50 GeV and 1 TeV.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Claimed improvement over ANTARES limits may stem from asymmetric under-fluctuation treatment; a fair comparison requires the same convention for combined and individual limits.","rationale":"The reader's weakest assumption correctly identifies the asymmetric treatment of under-fluctuations as the principal soft spot, and the Figure 5 caption confirms it explicitly. My stress-test found no other concern of comparable weight: the likelihood combination in Section 5 is a standard product of Poisson likelihoods, the use of PPPC4 spectra is a reasonable unification choice, and the absence of a significant excess is consistent with the presented limits. The load-bearing issue is purely comparative: the headline improvement is demonstrated by overlaying curves that obey different conventions in the under-fluctuation regime. Since an under-fluctuation in the combined fit can depress the limit below the capped ANTARES curve without representing genuine constraining power, the quantitative claim needs to be re-evaluated under a symmetric convention. I do not recommend rejection, because the combined analysis is methodologically plausible and might well still show improvement; but the paper should be accepted only if the authors either recompute the ANTARES comparison limits without the sensitivity floor, or explicitly restrict the claim to masses where the comparison is insensitive to this convention. This is exactly the condition the reader proposed, so the verdict remains CONDITIONAL.","tokens_in":5501,"tokens_out":3259,"duration_ms":38206,"concrete_test":"Recompute the ANTARES-only 90% CL observed limits from the same 9-year dataset without applying the sensitivity floor when the fitted limit under-fluctuates (i.e., keep the raw Feldman-Cousins limit), and recompute the combined limits under the identical no-floor convention. Then for each annihilation channel (especially tau+tau- and b-bbar) and halo profile in Figure 5, test whether the combined curve lies below the ANTARES-only curve at every WIMP mass from 50 GeV to 1 TeV. If the combined limit equals or exceeds the ANTARES-only limit at any mass, or if the median ratio over the mass range is not less than 1, the paper's 'improved limits' claim is not supported. An alternative cross-check is to apply the sensitivity floor to both curves; under that symmetric convention the claimed improvement should persist if it is genuine.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's central quantitative claim is that the combined 90% CL upper limits on the annihilation cross section are lower than the individual ANTARES and IceCube limits across 50 GeV to 1 TeV. The comparison is undermined by an explicitly admitted asymmetry in how under-fluctuations are handled. The Figure 5 caption states: 'this combined analysis treated under-fluctuations differently than what is presented in the ANTARES paper. When obtaining limits with lower values than sensitivities, this joint analysis kept the limit as obtained from the likelihood method while the limits were moved to sensitivities for ANTARES.' In other words, when the ANTARES-only observed limit dips below its expected sensitivity, the plotted ANTARES curve is raised to the sensitivity value, while the combined curve is allowed to retain the raw, occasionally very low, likelihood-based limit. Under-fluctuations are statistical noise; a downward fluctuation in the combined fit can produce an arbitrarily low limit that has no physical meaning as an improvement in sensitivity. Because the headline claim is a direct curve-to-curve comparison, this asymmetry alone can manufacture an apparent improvement even if the combined data are no more constraining than ANTARES alone. The concern is load-bearing: it targets the exact metric used to support the abstract's statement of 'improved limits,' and it is acknowledged in the text rather than merely hypothesized. A corrected comparison using the same under-fluctuation treatment for both curves could leave little or no improvement for some channels and masses, especially where ANTARES previously experienced downward fluctuations. This is not a challenge to the statistical method itself, which is standard, but to the fairness and interpretability of the central comparative result.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript (arXiv:1908.07300, ICRC2019 proceedings) presents the first combined search for neutrinos from dark matter annihilation in the Galactic Centre using ANTARES (2101.6 days, 2007-2015) and IceCube (1007 days, IC86). The analysis uses a binned Poisson likelihood with a shared signal fraction mu, common NFW and Burkert halo profiles, and common PPPC4 neutrino spectra, and sets 90% CL upper limits on <sigma_A v> for WIMP masses from 50 GeV to 1 TeV in the W+W-, tau+tau-, mu+mu-, and b bbar channels. No significant excess is observed. The central claim, stated in the abstract and in Section 6, is that the combined limits improve on the previously published ANTARES and IceCube limits over most of the mass range and for most channels and profiles. The paper explicitly notes in the Figure 5 caption that under-fluctuations were treated differently for the combined curve than for the ANTARES curve; this convention asymmetry directly affects the comparison that supports the headline claim.","tokens_in":5732,"tokens_out":6765,"duration_ms":72085,"significance":"The paper is significant as a benchmark: it is the first joint ANTARES+IceCube dark matter search toward the Galactic Centre, and it unifies choices (spectra, halo parameters, likelihood) that previously differed between the collaborations. The combined-likelihood construction in Section 5.3 is coherent for a fixed signal model, and the use of already published event samples makes the result checkable. The four annihilation channels and two halo profiles give a useful survey. However, the headline quantitative claim is not yet demonstrated because the comparison to ANTARES uses an asymmetric convention for handling downward fluctuations. The paper should therefore be revised to present a like-for-like comparison; once that is done, the result would serve as a useful reference point for future combined indirect dark matter searches.","major_comments":[{"comment":"Figure 5 caption explicitly states: 'this combined analysis treated under-fluctuations differently than what is presented in the ANTARES paper. When obtaining limits with lower values than sensitivities, this joint analysis kept the limit as obtained from the likelihood method while the limits were moved to sensitivities for ANTARES.' Because a downward fluctuation in observed counts can produce a likelihood limit that is stronger than the experiment's actual sensitivity, comparing a raw combined limit with a sensitivity-floored ANTARES limit can manufacture an apparent improvement. This asymmetry bears directly on the abstract's claim of 'improved limits' and on the Section 6 statement that 'the combined limits show improvements.' Please recompute both the combined and the ANTARES curves under a single convention (either both raw likelihood limits or both floored to sensitivity) and reassess the improvement claim; if the asymmetric convention is retained, the claim should be rephrased to avoid implying an equivalent comparison.","section":"Section 6 / Figure 5 caption"},{"comment":"The text claims improvements 'for almost all annihilation channels and the two DM halo profiles', but Figure 5 shows only the tau+tau- and b bbar channels for the NFW profile. The evidence base for the claim is therefore not fully presented. Please add comparison panels or a table for W+W-, mu+mu-, and the Burkert profile, or restrict the claim to the cases actually shown.","section":"Section 6"},{"comment":"The manuscript does not discuss systematic uncertainties. The published ANTARES and IceCube limits to which the comparison is made typically include detector-related and J-factor effects, and a combined limit should state whether these are propagated. If the limits shown are statistical only, the text should say so explicitly; otherwise the comparison to published limits is not apples-to-apples. At minimum, the dominant systematic (for example the J-factor uncertainty for the Galactic Centre) should be listed and, if feasible, included or shown in the figures.","section":"Sections 5 and 6"}],"minor_comments":[{"comment":"The product limits in Eq. (5.1) are written as 'max' and 'min' subscripts; this should be rendered as i = min to max (or simply as a product over bins) for readability.","section":"Equation (5.1)"},{"comment":"The IceCube PDF binning in right ascension is stated as '-2pi to 2pi' with 10 bins; since right ascension is cyclic, please clarify whether this is a wrapped interval and how the background PDF normalization handles periodicity.","section":"Section 5.2 / Figure 2"},{"comment":"In the description of the ANTARES QFit histograms, bins are said to range over Delta cos(theta) '-1 to 0.14 rad'; Delta cos(theta) is dimensionless, so the unit 'rad' appears to be a typo.","section":"Section 5.2"},{"comment":"Equation (2.1) uses a factor 1/2 for self-conjugate dark matter; please confirm that this convention is consistently applied in the ANTARES and IceCube signal normalizations, since the relative normalization affects the combined limit.","section":"Section 2"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the headline: the paper's central claim of improved combined limits is not cleanly supported because the comparison to the ANTARES-only limit treats under-fluctuations differently for the two curves. That is not a minor technicality; it directly affects the main result.\n\nThe genuinely new part is the first joint ANTARES+IceCube likelihood search for dark matter annihilation from the Galactic Centre. The authors unified the halo parameters, the neutrino energy spectra (choosing PPPC4), and the Feldman-Cousins limit procedure, and combined the binned likelihoods with a single signal fraction. This is a sensible, standard construction, and the combination is a natural benchmark for future analyses. The data sets are the published ones, so there is no new data, but the joint treatment is new.\n\nThe soft spot is load-bearing. The Figure 5 caption admits the issue: when limits fall below sensitivity, this analysis keeps the raw likelihood limit while the ANTARES values are moved up to sensitivity. A downward fluctuation in the combined fit then appears as an artificially strong limit. If the ANTARES limits were re-derived with the same convention, the improvement could be much smaller, or vanish, in some channels. The paper does not provide a corrected comparison, so the abstract's 'improved limits' sentence is premature. The absence of systematic uncertainties (halo profile, detector response) is also present; for a proceedings it is understandable, but it should at least be stated.\n\nThe b-bar/Burkert case already shows the combined limit dominated by IceCube, which is honest and consistent. The paper is readable and transparent about the main caveat, but the caveat is not resolved.\n\nThis deserves a serious referee because the combined likelihood template is useful and the issue is fixable. But the version I read should be revised before the comparative claim is repeated: use one under-fluctuation convention for all curves, or show explicitly how the comparison moves under both conventions.","headline":"The first combined ANTARES+IceCube Galactic Centre dark-matter search is a sensible technical template, but the claimed improvement over ANTARES rests on an asymmetric under-fluctuation treatment that makes the headline comparison unreliable.","tokens_in":6332,"tokens_out":2775,"would_cite":false,"duration_ms":29017,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["95.35.+d","95.85.Ry"],"model":"deepseek-v4-flash","headline":"A first combined ANTARES–IceCube search for neutrinos from dark matter annihilation in the Galactic Centre finds no signal, and sets 90% upper limits lower than either experiment's published limits for most channels in the 50 GeV–1 TeV…","keywords":["dark matter","WIMP annihilation","neutrino telescopes","Galactic Centre","combined likelihood","ANTARES","IceCube","upper limits"],"falsifier":"Recompute the 90% limits twice: once with under-fluctuating limits kept at their raw likelihood values for both the individual and combined analyses, and once with all such limits raised to the expected sensitivity; if the combined curve no longer lies below the individual curves for most channels in the 50 GeV–1 TeV range under both treatments, the paper's central claim of improved limits fails. A supporting check is to compare median expected limits rather than observed limits, which removes under-fluctuation effects entirely.","tokens_in":5332,"feed_emoji":"🔭","tokens_out":9870,"duration_ms":93359,"temperature":0.7,"pith_summary":"The paper asks whether two neutrino telescopes, one in the Mediterranean Sea and one at the South Pole, constrain dark matter more tightly together than separately. It reports the first joint search for neutrinos from dark-matter annihilation at the Galactic Centre, combining nine years of ANTARES data with three years of IceCube data inside a single binned likelihood. The data show no excess over background, so the result is a set of 90% upper limits on the annihilation cross section for WIMP masses from 50 GeV to 1 TeV. For most annihilation channels and for both the cusped NFW halo profile and the cored Burkert profile, the combined limits lie below the limits the two experiments had published individually; the $b\\bar{b}$ channel with the Burkert profile is the stated exception. The contribution matters because it shows a path to more sensitive dark-matter searches by pooling existing neutrino data and by unifying the analysis choices that had differed between the two experiments.","feed_headline":"Two neutrino telescopes together tighten dark matter limits","feed_subtitle":"ANTARES and IceCube data, combined in one likelihood, beat the single-detector limits at 50 GeV–1 TeV.","key_machinery":"The load-bearing object is the combined binned Poisson likelihood. Each telescope contributes its own likelihood built from signal and background probability density functions, and the combined likelihood is the product of the two. A single parameter, the joint signal fraction $\\mu = N_{\\rm sig}/N_{\\rm tot}$, is maximized, and each experiment's individual signal fraction is expressed as $\\mu_k = w_k \\mu$, where the weight $w_k$ encodes that sample's relative signal efficiency and background efficiency; the weights determine how much each detector pulls the shared limit. Around this core, the analysis unifies the physical inputs: the J-factor, the line-of-sight integral of the squared dark-matter density, is computed for the NFW and Burkert profiles with common parameters, and the neutrino energy spectra for all four annihilation channels come from the same published tables. The final step converts the fitted signal fraction into a 90% upper limit on $\\langle \\sigma_A \\upsilon \\rangle$ using the Feldman-Cousins prescription.","core_discovery":"The discovery the paper argues for is that a likelihood-level combination of the ANTARES and IceCube data sets improves the 90% upper limits on the thermally-averaged dark-matter self-annihilation cross section $\\langle \\sigma_A \\upsilon \\rangle$ over the WIMP (weakly interacting massive particle) mass range 50 GeV to 1 TeV, compared with the limits each experiment had previously published on the same data. The improvement is obtained by eliminating the main differences between the two analyses: the same NFW and Burkert halo profiles with common parameters, the same neutrino energy spectra for annihilation into $W^+W^-$, $\\tau^+\\tau^-$, $\\mu^+\\mu^-$, and $b\\bar{b}$, and a single combined Poisson likelihood with one shared signal-fraction parameter. Since the observed events are consistent with background, the output is upper limits rather than a detection. The improvement holds for nearly all channels and both halo profiles; the $b\\bar{b}$ channel with the Burkert profile is the exception because IceCube alone already dominates the limit across the whole mass range. The paper notes a caveat in the comparison: under-fluctuations of the background were treated differently from the ANTARES publication, with this combined analysis keeping limits that fall below sensitivity at their raw likelihood values while the ANTARES limits were raised to the expected sensitivity.","pith_inferences":["Editorial inference: the most direct test of the claimed improvement is to recompute both the combined and the individual limits with identical under-fluctuation handling; the paper's caption suggests that the size of the improvement, and possibly its existence for some channels, depends on this choice.","Editorial inference: the shared-likelihood recipe transfers to other targets visible to both telescopes, such as dwarf spheroidal galaxies or the extended Milky Way halo, and to next-generation detectors, since the unified spectra and halo parameters would carry over unchanged.","Editorial inference: one could predict where combination pays most by computing the per-experiment weight ratio $w_A/w_I$ as a function of WIMP mass and annihilation channel; channels with comparable weights should show the largest gain over the stronger single detector, which could guide future exposure allocation."],"forward_implications":["For WIMP masses from 50 GeV to 1 TeV, the combined limits are lower than the individual ANTARES and IceCube limits for almost all annihilation channels and for both halo profiles, so pooling the two data sets yields a direct sensitivity gain.","The unified likelihood, halo parameters, and spectra provide a common benchmark: future dark-matter analyses by the two experiments can be compared on equal footing, and the same combination recipe can be applied to additional years of data without re-deriving the model inputs.","Because no neutrino excess is observed, the combined limits strengthen the exclusion region for WIMP annihilation in the Galactic Centre, sitting alongside limits from gamma-ray instruments in the same mass range.","The $b\\bar{b}$/Burkert case shows where the method adds least: when one detector already dominates the sensitivity over the full mass range, the combined result inherits that detector's limit and the combination contributes little."],"supporting_citations":[{"why":"Supplies the nine-year ANTARES data sample and the individual ANTARES limits that the combined result is compared against.","marker":"[1]"},{"why":"Supplies the three-year IceCube Galactic Centre sample and the individual IceCube limits used as the comparison baseline.","marker":"[2]"},{"why":"Provides the unified dark-matter halo parameters used to compute J-factors for both experiments.","marker":"[5]"},{"why":"Computes the J-factors for the NFW and Burkert profiles that set the expected signal normalization.","marker":"[6]"},{"why":"Provides the unified neutrino energy spectra for the four annihilation channels, replacing the differing spectra used previously.","marker":"[7]"},{"why":"Gives the Feldman-Cousins prescription used to convert the fitted signal fraction into a 90% upper limit.","marker":"[11]"}],"fun_headline_variants":["Combined neutrino data sharpen dark matter limits","ANTARES + IceCube: unified search tightens dark matter bounds","Joint neutrino search tightens WIMP annihilation limits","One likelihood, two telescopes: stronger dark matter limits"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim of improved limits rests on comparing the combined limits with ANTARES limits that were raised to the expected sensitivity whenever the data fluctuated low, while the combined analysis kept its own low-fluctuation limits at the raw likelihood values; if the two sides were treated identically, the improvement could shrink or disappear.","fun_headline_variants_meta":{"raw":{"variants":["Combined neutrino data sharpen dark matter limits","ANTARES + IceCube: unified search tightens dark matter bounds","Joint neutrino search tightens WIMP annihilation limits","One likelihood, two telescopes: stronger dark matter limits"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000787,"raw_usage":{"total_tokens":3504,"prompt_tokens":1013,"completion_tokens":2491,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":629,"completion_tokens_details":{"reasoning_tokens":2427}},"tokens_in":629,"tokens_out":2491,"duration_ms":19286,"temperature":1.0,"reasoning_tokens":2427,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:20:01.920254+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the 90% limits twice: once with under-fluctuating limits kept at their raw likelihood values for both the individual and combined analyses, and once with all such limits raised to the expected sensitivity; if the combined curve no longer lies below the individual curves for most channels in the 50 GeV–1 TeV range under both treatments, the paper's central claim of improved limits fails. A supporting check is to compare median expected limits rather than observed limits, which removes under-fluctuation effects entirely.","supporting_citations":[{"cited_title":"Nesti and P","cited_arxiv_id":null,"evidence_quote":"Provides the unified dark-matter halo parameters used to compute J-factors for both experiments."}],"review_version":1}