{"id":"2bfbf592-5903-4d34-aa89-5575f0edccc2","arxiv_id":"1909.01246","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"low","formal_verification":"none","parameter_count":2,"one_line_summary":"No evidence for displaced dark-photon jets is found, and 95% confidence limits exclude cross sections above about 4 pb for H to two dark photons with decay lengths from 1.5 mm to 307 mm.","lead":"The ATLAS experiment searched for long-lived dark photons produced in Higgs boson decays, analyzing 36.1 inverse femtobarns of 13 TeV proton collisions. It found no excess over background and set new exclusion limits on dark photon production for decay lengths between about one and three hundred millimeters.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lifetime reweighting is the most load-bearing step: the formula in Sec. 9 is garbled as printed and no closure test against direct MC at other cτ values is shown.","rationale":"The reader correctly identified the ABCD independence assumption as a soft spot, and that concern is real: the SR correlation is only indirectly validated, and the mock tests are statistics-limited. However, the most load-bearing step for the headline claim is the lifetime extrapolation. The excluded cτ intervals are the paper's main quantitative result, and they depend entirely on reweighting a single generated sample. The paper does not show a closure test against direct generation at other lifetimes, and the printed weighting formula is internally inconsistent as written. This is a concrete, checkable concern rather than a speculative statistical worry. The rest of the analysis—trigger selection, BDT training, systematics, background closure—is standard and carefully presented, and the signal-region yields agree with the ABCD predictions. I therefore do not reject the paper; I recommend acceptance conditional on a successful lifetime-reweighting closure check.","tokens_in":48862,"tokens_out":18967,"duration_ms":181194,"concrete_test":"Generate independent FRVZ H→2γd+X, mH=125 GeV signal MC directly at cτ=1.5 mm and 307 mm, process through the same trigger, reconstruction, and selection chain, and compare the signal efficiency and the resulting 95% CL σ×B limit as a function of cτ against the weighted extrapolation from the nominal cτ=49.23 mm sample. Require agreement within the statistical uncertainty of the weighted sample; a difference larger than the combined error bars would indicate the extrapolation is not closure-tested.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central quantitative claim—95% CL excluded cτ intervals such as [1.5, 307] mm for H→2γd+X with mH=125 GeV—depends on extrapolating the signal acceptance from a sample generated at cτ=49.23 mm to all other lifetimes via per-DPJ weights. Section 9 states the weight as 'w_i(ti) = τref e−ti/τref · e−ti/τnew τnew', which as printed is dimensionally inconsistent and does not correspond to the standard ratio of decay densities, w_i = (τ_ref/τ_new) exp[t_i(1/τ_ref − 1/τ_new)]. Even if this is a typesetting artifact, the paper provides no closure test demonstrating that reweighting reproduces the efficiency of an independently generated sample at a different lifetime. The acceptance is strongly nonlinear in decay position (pixel layers, hadronic calorimeter, muon trigger chambers), and large extrapolation factors (about 33× down to 1.5 mm and 6× up to 307 mm) can produce large per-event weights and statistically fragile efficiency estimates. If the reweighting is biased, the cτ intervals in the abstract and Table 6 shift independently of the ABCD background method.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports a search by the ATLAS Collaboration for long-lived dark photons produced in 13 TeV pp collisions, using 36.1 fb^-1 of 2015-2016 data. The search targets the FRVZ benchmark model in which a 125 GeV Higgs boson (or an 800 GeV heavy scalar) decays to two dark fermions, each yielding one or two dark photons that decay displaced into collimated muons or light hadrons, reconstructed as dark-photon jets (DPJs). Three final states are considered: muonic-muonic, muonic-hadronic, and hadronic-hadronic DPJ pairs. Dedicated triggers and BDT-based selections are used, and the dominant multi-jet background is estimated with an ABCD method using two nearly uncorrelated variables (max track isolation and |Delta phi|). Observed event yields agree with the predicted backgrounds in all validation and signal regions (e.g., 113 observed vs 128 +/- 26 expected in the muonic-muonic signal region). No excess is found, and 95% CL upper limits on sigma x B as a function of the dark-photon proper decay length c tau are derived. The headline result is that sigma x B above 4 pb is excluded for H to 2 gamma_d + X with m_H = 125 GeV for c tau between 1.5 mm and 307 mm; further exclusions are given for four-dark-photon decays and for the 800 GeV scalar. The results are also interpreted as 90% CL limits on the kinetic mixing parameter versus dark-photon mass.","tokens_in":49073,"tokens_out":6775,"duration_ms":66730,"significance":"If the c-tau-dependent limits are correct, the search provides the first 13 TeV ATLAS constraints in this channel using the fully hadronic signature, extends Run-1 displaced-lepton-jet results, and tests the FRVZ model in a previously allowed region of low kinetic mixing. The analysis is internally consistent: the ABCD background method is validated in dedicated validation regions and in a mock signal region, the linear correlation of the ABCD variables is checked (<6%) and found to have negligible impact, cosmic-ray and beam-induced backgrounds are estimated from dedicated datasets, and the dominant systematic uncertainties are evaluated with data-driven tag-and-probe and data/MC closure methods. The manuscript is careful and the presentation is generally clear. However, the central c-tau-dependent limits rely on a lifetime-reweighting extrapolation that is both misprinted and not validated by a closure test; this is the main reason the paper does not merit immediate acceptance.","major_comments":[{"comment":"As printed, the per-dark-photon weight w_i(t_i) = tau_ref e^{-t_i/tau_ref} * e^{-t_i/tau_new} tau_new is dimensionally inconsistent and does not equal the standard decay-time reweighting ratio (tau_ref/tau_new) exp[t_i(1/tau_ref - 1/tau_new)]; if taken literally, the extrapolated efficiencies in Figure 4 and the excluded c-tau intervals in Table 6 would be incorrect. Even if this is a typesetting artifact, the paper provides no closure test demonstrating that reweighting reproduces the signal efficiency of an independently generated sample at a different c-tau. Given that the extrapolation factors are large (about 33x toward 1.5 mm and about 6x toward 307 mm) and that the acceptance is strongly nonlinear in the decay position (pixel layers, hadronic calorimeter, muon trigger chambers), the c-tau-dependent limits rest on an unvalidated extrapolation. The authors should correct the formula, add a closure test comparing reweighted efficiencies at two or three test lifetimes against dedicated MC samples, and either assign a systematic uncertainty to the reweighting procedure or demonstrate that it is negligible.","section":"Section 9 (Results and interpretation), weight formula and Figure 4"}],"minor_comments":[{"comment":"The text refers to 'J/phi -> mu mu' but should read 'J/psi -> mu mu'; the same acronym is written correctly as J/psi elsewhere in the paper.","section":"Section 8 (Systematic uncertainties), muon reconstruction paragraph"},{"comment":"The symbol for the track isolation scalar sum appears as 'Í pT' (an apparent OCR artifact); it should be written consistently as Sigma(pT) throughout.","section":"Section 7 (Multi-jet background estimation)"},{"comment":"The sentence 'The search is limited to |eta| < 2.5, corresponding to the ID coverage, to ensure that selected muons are isolated from ID tracks' is unclear; the muons are required to be unmatched to ID tracks, not 'isolated from ID tracks' in the usual isolation sense.","section":"Section 5.1 (Dark-photon jet classification)"},{"comment":"The notation 'tau ref' and 'tau new' is typeset inconsistently without subscripts; using tau_ref and tau_new as subscripts would improve readability and avoid confusion with the product structure of the weight.","section":"Section 9 (Results and interpretation), weight notation"},{"comment":"The validation-region BDT windows (e.g., '-0.75 < muBDT < 0.35' and 'muBDT > -0.7') are asymmetric but the reasons for these specific choices are not explained; a brief note on how the windows were chosen to avoid signal leakage would help the reader.","section":"Table 2 and Section 7"}],"recommendation":"major_revision","confidential_remarks":"The central technical concern is the lifetime reweighting in Section 9: as printed the formula is wrong and there is no closure test for the extrapolation. If the authors can provide a correct formula and a closure test (or otherwise justify the extrapolation), the paper should be publishable; I see no other blockable issue. The ABCD background estimation, by contrast, is well validated and is not the source of my concern."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a solid, honest ATLAS LLP search — 36.1 fb^-1 of 13 TeV data, a new fully hadronic dark-photon-jet channel, and limits that improve on Run-1 in most of the cτ range. The analysis is careful: the ABCD background estimate is validated in control regions and a mock signal region, and the yields agree (113 vs 128±26 in muonic-muonic, 179 vs 177±86 in mixed). I believe the central null result.\n\nWhat's genuinely new: the hadronic DPJ channel with the CalRatio trigger and the hBDT; the muonic channel is an update of the Run-1 search. The paper is honest about the hadronic channel's low efficiency for mH=125 GeV and about Run-1 being slightly more sensitive in some high-mass/low-epsilon regions.\n\nThe soft spot that matters: the lifetime reweighting. The formula in Sec. 9 is garbled as printed — it reads as τ_ref e^{−t_i/τ_ref} · e^{−t_i/τ_new} τ_new, which is dimensionally inconsistent. I suspect a typo, but the paper shows no closure test demonstrating that reweighting reproduces the acceptance from direct MC at other lifetimes. With per-DPJ weights up to ~33× for cτ=1.5 mm, this could shift the excluded intervals in Table 6. That's a real weakness, though likely fixable.\n\nMinor point: the ABCD uncorrelated assumption is validated with linear correlation below 6% and a mock signal-region test that agrees within statistics — but the test is limited (231±58 expected vs 184 observed in muonic). A statement on the impact of a residual ~6% correlated shift would be cleaner.\n\nThe citation pattern is normal; nothing circular here. This paper is for LLP hunters and dark-sector phenomenologists who need bounds on the FRVZ parameter space.\n\nVerdict: worth refereeing. I'd ask for a corrected formula and a closure test before acceptance; those are easy to provide. Then it's a competent, publishable measurement.","headline":"A competent, honest ATLAS LLP search whose main weakness is an under-documented and misprinted lifetime-reweighting step; deserves peer review but needs a closure test.","tokens_in":49629,"tokens_out":3220,"would_cite":true,"duration_ms":33176,"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":"No excess of displaced dark-photon jets is observed in 36.1 fb⁻¹ of 13 TeV pp collisions, and Higgs-boson decays to two dark photons are excluded above 4 pb for dark-photon decay lengths between 1.5 mm and 307 mm.","keywords":["dark photon","long-lived neutral particle","displaced vertex","exotic Higgs decay","hidden sector","lepton jets","ATLAS","LHC"],"falsifier":"A direct measurement would subdivide the signal-region control sample into smaller bins and compute the linear correlation between $\\max(\\Sigma p_T)$ and $|\\Delta\\phi|$ under the final selection; if that correlation is significantly above the few-percent level assumed, the ABCD prediction shifts and the reported exclusion windows would move.","tokens_in":48632,"feed_emoji":"⚛️","tokens_out":6795,"duration_ms":67197,"temperature":0.7,"pith_summary":"In 36.1 fb⁻¹ of proton–proton collisions at $\\sqrt{s} = 13$ TeV, the paper searches for long-lived dark photons produced in decays of a 125 GeV Higgs boson or an 800 GeV heavy scalar, where each dark photon decays into a narrow jet-like cluster of leptons or light hadrons away from the interaction point. The observed events agree with the expected background, so no new particle is claimed. Instead, the paper sets 95% confidence-level upper limits on the production cross section times branching fraction as a function of the dark-photon proper decay length. For a Higgs boson decaying into two dark photons, production above 4 pb is excluded for decay lengths between 1.5 mm and 307 mm. A reader should care because this directly constrains a well-motivated hidden-sector scenario that is otherwise hard to probe with prompt searches.","feed_headline":"No dark photons found: Higgs decay rate above 4 pb ruled out","feed_subtitle":"ATLAS's 13 TeV search sees only background and closes the 1.5–307 mm decay-length window.","key_machinery":"The load-bearing object is the displaced dark-photon jet (DPJ): a cone of collimated muons in the muon spectrometer, or a narrow energy cluster in the hadronic calorimeter, with no matching inner-detector track. Two boosted decision trees separate signal DPJs from cosmic-ray muons and from multi-jets, dedicated displaced-object triggers collect the events, and the event-level background is predicted by an ABCD method in the plane of $\\max(\\Sigma p_T)$ (inner-detector track isolation) versus $|\\Delta\\phi|$ (azimuthal opening angle between the two DPJs). Signal efficiencies are extrapolated to arbitrary proper lifetimes by reweighting each dark-photon decay time, which turns the measured event counts into limits on $\\sigma \\times B$ as a function of $c\\tau$.","core_discovery":"The paper's central claim, stated on its own terms, is that the FRVZ hidden-sector benchmark, where a Higgs boson decays to dark fermions that each emit a dark photon and the dark photon decays to Standard Model fermions, produces no observable excess above the multi-jet and cosmic-ray background in the ATLAS detector. Using the full 2015–2016 13 TeV dataset, the analysis excludes at 95% CL a cross section times branching fraction above 4 pb for $H \\to 2\\gamma_d + X$ when the dark-photon proper decay length lies between 1.5 mm and 307 mm. For two dark photons per Higgs decay, the excluded window is 3.7–178 mm, and for an 800 GeV scalar with $\\sigma \\times B = 5$ pb, the excluded windows extend to about 1.4 m and beyond. The result is an exclusion, not a discovery, and it is also interpreted as a 90% CL upper limit on the kinetic mixing parameter $\\epsilon$ as a function of dark-photon mass.","pith_inferences":["A straightforward extension would apply the same displaced-DPJ reconstruction to events with only one DPJ; the present requirement of two back-to-back jets, inherited from the two-body heavy-scalar topology, leaves single-jet or non-collinear hidden-sector signatures unconstrained.","The ABCD correlation check is the least protected part of the chain: with more luminosity, the signal-region sidebands could be subdivided further, and a residual correlation at the few-percent level could be measured directly rather than inferred.","The efficiency-versus-lifetime tables could be recast by other experiments into limits on any vector-portal model with similar kinematics, not only FRVZ, making the public result a reusable constraint.","Scaling the excluded cross section roughly as the inverse of integrated luminosity suggests that the full Run 2 dataset could push the 4 pb threshold down by about a factor of three, if the background remains smooth."],"forward_implications":["If correct, any FRVZ-model dark photon with mass near 0.4 GeV, a Higgs-decay branching fraction around 10%, and proper decay length in the excluded windows cannot be produced through Standard Model Higgs decays at the assumed rate; that parameter region is closed.","The purely hadronic channel, exploited here for the first time, means dark photons that decay to pions rather than leptons are now constrained at 13 TeV, not just muonic final states.","Because the limits are set as a function of $c\\tau$, the same result can be applied to other lifetimes without running a new search, provided the acceptance model holds.","For the 800 GeV scalar, the excluded lifetime window reaches up to about 1.4 m in the muonic channel, closing a long-lifetime region that fixed-target and beam-dump experiments do not cover."],"supporting_citations":[{"why":"Defines the FRVZ benchmark with a hidden Higgs decaying into dark fermions and dark photons; supplies the signal model.","marker":"[8]"},{"why":"Companion paper showing how Higgs decays to lepton jets could be discovered at hadron colliders; motivates the DPJ signature.","marker":"[9]"},{"why":"The previous ATLAS 8 TeV displaced lepton-jet search whose detector and trigger strategy this analysis extends, and from which the tri-muon trigger systematic is taken.","marker":"[11]"},{"why":"Describes the ATLAS detector whose inner tracker, calorimeters, and muon spectrometer define the displaced-jet acceptance.","marker":"[50]"},{"why":"Provides the full relation between dark-photon lifetime, kinetic mixing, and mass, plus the branching fractions used in simulation.","marker":"[56]"},{"why":"Supplies the multivariate analysis toolkit used to train the boosted decision trees for the muon and hadron DPJ classifiers.","marker":"[89]"},{"why":"Documents the displaced-decay trigger algorithms that select the three signal channels.","marker":"[94]"},{"why":"Defines the CLs procedure used to derive the 95% and 90% confidence-level limits.","marker":"[99]"}],"fun_headline_variants":["ATLAS dark-photon hunt: no signal, 4 pb ceiling for Higgs decays","Long-lived dark photons: ATLAS sees none, sets new limits","Higgs to dark photons? ATLAS data says no above 4 pb","Dark photon search: null result, decay-length window closed","ATLAS excludes dark photons from Higgs, 1.5-307 mm"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The ABCD background estimate assumes that the two event-level variables, inner-detector track isolation $\\max(\\Sigma p_T)$ and opening angle $|\\Delta\\phi|$, are nearly uncorrelated in the signal region, so that the background there can be extrapolated from three sidebands; the paper measures a linear correlation below 6% in validation-region data and in multi-jet simulation, but the test is limited by the size of the control sample.","fun_headline_variants_meta":{"raw":{"variants":["ATLAS dark-photon hunt: no signal, 4 pb ceiling for Higgs decays","Long-lived dark photons: ATLAS sees none, sets new limits","Higgs to dark photons? ATLAS data says no above 4 pb","Dark photon search: null result, decay-length window closed","ATLAS excludes dark photons from Higgs, 1.5-307 mm"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00112,"raw_usage":{"total_tokens":4672,"prompt_tokens":967,"completion_tokens":3705,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":583,"completion_tokens_details":{"reasoning_tokens":3608}},"tokens_in":583,"tokens_out":3705,"duration_ms":25370,"temperature":1.0,"reasoning_tokens":3608,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T05:23:46.526524+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct measurement would subdivide the signal-region control sample into smaller bins and compute the linear correlation between $\\max(\\Sigma p_T)$ and $|\\Delta\\phi|$ under the final selection; if that correlation is significantly above the few-percent level assumed, the ABCD prediction shifts and the reported exclusion windows would move.","supporting_citations":[{"cited_title":"Triggers for displaced decays of long-lived neutral particles in the ATLAS detector","cited_arxiv_id":"1305.2284","evidence_quote":"Documents the displaced-decay trigger algorithms that select the three signal channels."}],"review_version":1}