{"id":"0bb63fa3-aef8-47bc-be0d-5413eea4bb35","arxiv_id":"1908.02699","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":3,"one_line_summary":"At 95% confidence, fewer than 4.6% of 125 GeV Higgs bosons decay to a photon plus an invisible massless dark photon, based on 137 fb^-1 of CMS data.","lead":"The CMS experiment searched for Higgs bosons that decay into a photon plus an invisible particle, a possible sign of dark matter. No such events were seen, and the result limits this decay to at most 4.6% of 125 GeV Higgs bosons at 95% confidence.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 4.6% limit assumes one normalization factor for both electron-misID and genuine-photon WZ backgrounds; the 231-event control region constrains only the first, leaving the second component's shape and normalization unvalidated.","rationale":"The reader's weakest assumption points to simulated mT shapes in low-statistics control regions. I sharpen this to a more concrete mechanism: the WZ background's two components are tied to a single normalization factor, but the 231-event control region only constrains the electron-misID component. The genuine-hard-photon component is unvalidated and could bias the dominant background estimate directly at the mT values where the signal lives. The proposed control-region or two-parameter test would settle whether this matters; if it shows the ratio is correct, the original ACCEPT verdict stands, but conditional acceptance is appropriate until the test is performed. A secondary concern is the abstract's claim of 'first limits' given the similar 8 TeV search in Ref. [20]; that claim should be checked against the earlier paper's interpretation, though it does not affect the numerical limit.","tokens_in":32894,"tokens_out":17831,"duration_ms":204779,"concrete_test":"Redo the maximum-likelihood fit in Sec. 6 introducing two separate normalization parameters, muWZ_misID and muWZ_photon, with the second constrained by a new control region requiring two same-flavor leptons, a real isolated photon, and large MET with the Z->ll mass constraint (the W lepton not identified), or equivalently by a theory prior reflecting the uncertainty on the ratio. If the best-fit values of the two parameters differ by more than ~1 sigma, the shared-muWZ assumption biases the signal-region WZ yield. Alternatively, use the observed mT distribution in the existing 231-event control region to reweight the predicted WZ shape and propagate the reweighting through the fit; if the resulting 95% CL upper limit on B(H->invisible+gamma) shifts by more than ~0.5 percentage points from 4.6%, the central limit is not robust.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"At mH=125 GeV, the leading systematic risk is not the statistical size of the control regions but the structure of the WZ background model in Sec. 5.2. The WZ prediction entering the signal region has two components: (a) WZ->lnu ll where the W-decay lepton is misidentified as a photon, and (b) WZ->lnu ll where the W lepton is lost and a genuine hard photon is radiated. The fit uses a single normalization parameter muWZ for the sum. The control region with 231 events is constructed by using the W-decay lepton in place of the photon, so it constrains only component (a). Component (b) is estimated purely from simulation and is constrained only by its predicted ratio to (a) via the shared muWZ. If the simulated ratio of (b) to (a) is wrong—for example because the electron mis-ID rate measured in Z->ee does not apply in the high-pT phase space, or because the hard-photon radiation is mis-modeled—the signal-region WZ yield is biased by an amount not captured by the quoted systematic uncertainties. Since the 95% CL limit near 4.6% corresponds to roughly eight signal events against 13.3 background events, a one-event shift in the WZ estimate changes the limit by about 0.5 percentage points, which is a large fraction of the quoted result. The mT shape for component (b) is not validated by any data control region.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a search by the CMS Collaboration for a Higgs boson produced in association with a Z boson and decaying to an undetected particle together with an isolated photon, focusing on the massless dark photon interpretation. The analysis uses the full 13 TeV data set of 137 fb^-1 and selects events with a Z->ll candidate, a high-pT photon, large missing transverse momentum, and low jet activity. A binned maximum-likelihood fit is performed on the transverse mass mT spectrum in the signal region together with three dedicated control regions (e-mu for nonresonant backgrounds, three-lepton for WZ, and four-lepton for ZZ), with separate bins in photon pseudorapidity. Background normalizations for the dominant WZ and ZZ processes are fitted to data, nonresonant backgrounds are estimated from an e-mu control region, and a set of experimental and theoretical systematic uncertainties is included. No significant excess over the standard model background is found: 14 events are observed against a total background estimate of 13.3 +/- 3.8. Upper limits at 95% CL are set on sigma_ZH times B(H->invisible+gamma) as a function of mH, and for mH=125 GeV, assuming the SM ZH production cross section, the observed (expected) limit on the branching fraction is 4.6% (3.6%).","tokens_in":33200,"tokens_out":11734,"duration_ms":113799,"significance":"If the result holds, this is the first experimental limit on a Higgs boson decay to a massless dark photon plus invisible particles in the ZH production channel. The analysis follows a well-established profile-likelihood strategy, uses data-driven control regions for the dominant backgrounds, and documents the systematic uncertainties. The result is of direct interest to dark-sector and BSM Higgs searches, and it provides a template for future searches of this signature. The paper is clearly written and the statistical methodology is sound. The main caveat is the validation of a subdominant WZ background component, discussed in the major comments.","major_comments":[{"comment":"The WZ background in the signal region contains two components: (a) events in which the electron from the W boson decay is misidentified as a photon, and (b) events in which the W lepton is not identified and a genuine hard photon is radiated. In the likelihood of Eq. (1), a single normalization parameter muWZ multiplies both components, and the three-lepton control region directly constrains only component (a), because it requires the W-decay lepton to be identified and used as the photon proxy. Component (b) is thus constrained only through its simulated ratio to component (a), and its mT shape is not validated by any data control region. Since WZ is the dominant background (8.1 +/- 2.0 out of a total of 13.3 +/- 3.8) and the observed limit corresponds to roughly eight signal events, an O(1) event shift in this background would change the mH=125 GeV limit by about 0.5 percentage points. The authors should report the relative contribution of component (b) and either validate it with a dedicated control region or assign an explicit systematic uncertainty to the ratio of the two components.","section":"Section 5.2, Eq. (1)"}],"minor_comments":[{"comment":"Signal samples are generated only at mH = 125, 200, and 300 GeV, while limits are shown as a continuous function of mH between 125 and 300 GeV. The paper should state how the signal acceptance and mT shapes are interpolated between the generated masses.","section":"Section 6, Section 8"},{"comment":"The systematic uncertainties are described in the text, but a summary table listing each nuisance parameter, its prior, and its typical impact on the signal and background yields would improve the reproducibility and readability of the analysis.","section":"Section 7"},{"comment":"The sentence describing the electron-to-photon misidentification rate would be clearer if it specified whether the quoted 1-5% range is per-candidate and how it is applied as a function of |eta_gamma| and pT.","section":"Section 5.2"},{"comment":"The signal predictions in Table 2 are quoted for a signal size of 0.1 sigma_ZH; for clarity, the corresponding branching fraction assumed for the signal should be stated explicitly in the table caption or text.","section":"Table 2 and Figure 3"}],"recommendation":"major_revision","confidential_remarks":"The paper is a competent CMS search and the central result (no excess, first limits) is likely robust. My main concern is the unvalidated genuine-photon component of the WZ background, which is tied to the dominant background through a single normalization parameter. The impact may be modest compared to the large statistical uncertainties, but the issue is concrete and should be addressed with a quantitative systematic or a dedicated control region. I recommend major revision to allow the authors to resolve this point; I do not see a need for rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, worth your time: this is a tidy search with a genuinely new result. Using the full 137 fb^-1 of Run 2, CMS looks for ZH with Z→ll and H→γ+invisible (massless dark photon), and finds no excess. The observed (expected) 95% CL limit on B(H→invisible+γ) is 4.6 (3.6)% at mH=125, the first dedicated limit on that final state. The analysis follows well-trodden CMS practice: a binned mT likelihood, e-mu, 3-lepton, and 4-lepton control regions for the nonresonant, WZ, and ZZ backgrounds, CLs limits, and documented systematics. The data (14 events) agree with the 13.3±3.8 background estimate. It's a solid extension of the 8 TeV search, with much more data and a new dark-photon interpretation, and the paper is clearly written.\n\nThe one soft spot worth pushing is the WZ background model in Sec. 5.2. The fit uses a single normalization μ_WZ for the sum of two physically different contributions: WZ where the W electron is misidentified as the photon, and WZ where the W lepton is lost and a genuine hard photon is radiated. The 231-event control region replaces the photon by the W lepton, so it validates the topology where the W lepton is reconstructed—not the lost-lepton-plus-hard-photon component. That second piece comes purely from simulation and is tied to the first only through the shared μ_WZ. If the simulated ratio of those two components is wrong, the signal-region WZ yield could shift by a small fraction of an event. At the sensitivity of this analysis (~8 signal events at the 4.6% upper limit), that could move the limit by several tenths of a percentage point. That is within the stated uncertainties but larger than one would like; I would ask the authors to quantify the two-component split and the simulation uncertainty on the hard-photon piece. The sparse e-mu control region (3 events) is a minor worry, but the nonresonant background is only 2.4 events, so it's not a big deal.\n\nNone of this undermines the core result. The analysis is honest, the control-region agreement is reasonable, and the limit is properly labelled as assuming SM ZH production for the branching-fraction interpretation. If I worked on exotic Higgs decays or dark sectors, I'd cite it. It deserves a serious referee; I'd accept it with a request for a clarifying note on the WZ decomposition.","headline":"First dedicated H→invisible+massless dark photon limit from CMS, clean and standard; the single WZ normalization for two subcomponents is the one question I'd push.","tokens_in":33738,"tokens_out":6005,"would_cite":true,"duration_ms":63603,"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":"No Higgs-to-dark-photon decays found; rate capped at 4.6 percent","keywords":["dark photon","Higgs boson exotic decay","invisible Higgs decay","missing transverse momentum","transverse mass","CLs upper limit","binned likelihood fit"],"falsifier":"Recompute the 95% confidence-level limit twice, once with the WZ background $m_T$ shape taken directly from the trilepton control-region data and once from simulation; if the two limits differ by more than the quoted systematic uncertainty, the background-shape assumption breaks. Alternatively, a dedicated data sample of WZ events with a fully identified third lepton, selected with the same missing-transverse-momentum and photon requirements as the signal region, would reveal any $m_T$ shape mismatch between data and simulation in the relevant phase space.","tokens_in":32689,"feed_emoji":"⚛️","tokens_out":5818,"duration_ms":55472,"temperature":0.7,"pith_summary":"This paper asks whether the 125 GeV Higgs boson can decay to a single visible photon plus an invisible massless dark photon, a signature that would point to a charged dark sector. The search, using the full 2016–2018 proton-proton collision data recorded at 13 TeV (an integrated luminosity of 137 fb$^{-1}$), finds 14 candidate events against 13.3 expected from standard model background processes; there is no excess. Interpreting the null result in the dark-photon model gives the first upper limits on such Higgs decays: at 125 GeV the branching fraction is capped at 4.6% at 95% confidence, with 3.6% expected. Because some models allow this branching fraction up to about 5%, the search is sensitive to the most plausible signal rates and rules them out.","feed_headline":"No Higgs-to-dark-photon decays found; rate capped at 4.6 percent","feed_subtitle":"First search for Higgs decays to a photon plus an invisible massless dark photon sees only background in 137 fb-1 of 13 TeV data.","key_machinery":"The discriminating variable is the transverse mass $m_T=\\sqrt{2 p_T^{\\text{miss}} p_T^{\\gamma}[1-\\cos\\Delta\\phi(\\vec p_T^{\\text{miss}},\\vec p_T^{\\gamma})]}$ of the missing-transverse-momentum and photon system. Signal events from $H\\to\\gamma+\\text{invisible}$ populate a Jacobian peak with an endpoint near $m_T\\sim m_H$, while the dominant WZ, ZZ, and nonresonant backgrounds are flat or rise toward low $m_T$. The extraction is a binned maximum-likelihood fit to the $m_T$ distribution, split into low-$|\\eta_\\gamma|$ and high-$|\\eta_\\gamma|$ regions, with three data control regions (same-flavor lepton pairs, trilepton WZ, and four-lepton ZZ) fixing the background normalizations; a modified frequentist CLs procedure produces the limits.","core_discovery":"The analysis targets the process $pp\\to Z(\\ell\\ell)H$ with $H\\to\\gamma\\gamma_D$, where $\\gamma_D$ is a massless dark photon that escapes detection, leaving a final state of a same-flavor lepton pair, one isolated photon, and large missing transverse momentum. After event selection, 14 events are observed against 13.3 $\\pm$ 3.8 expected background events, and a binned maximum-likelihood fit to the transverse-mass spectrum of the photon plus missing momentum finds no significant signal. At $m_H=125$ GeV, assuming the standard model ZH production cross section, the observed (expected) 95% confidence-level upper limit on the branching fraction $\\mathcal{B}(H\\to\\text{invisible}+\\gamma)$ is 4.6 (3.6)%. The same fit excludes the product $\\sigma_{ZH}\\times\\mathcal{B}(H\\to\\text{invisible}+\\gamma)$ from about 40 fb at 125 GeV to about 4 fb at 300 GeV, and the paper states these are the first limits on Higgs boson decays to final states that include an undetected massless dark photon.","pith_inferences":["I infer that the same analysis, applied to the full High-Luminosity LHC data set of roughly 3000 fb$^{-1}$, would push the branching-fraction limit below 1%, directly testing the ~5% benchmark that motivates the search.","The result can likely be reinterpreted for a massive dark photon: a nonzero $\\gamma_D$ mass would shift the $m_T$ endpoint below $m_H$, so the existing limit curve could be recast as a constraint on the dark photon mass within the model.","The same-flavor control region used for nonresonant backgrounds contains only 3 events, so that background shape is validated on very sparse data; a future measurement with higher statistics in that region would be the most direct check on the quoted limit.","Combining this ZH-associated search with gluon-fusion production, as the earlier 8 TeV analysis did, could roughly double the signal acceptance and tighten the limit further."],"forward_implications":["Any 125 GeV Higgs-like scalar with a branching fraction above 4.6% to a photon plus an invisible massless particle is excluded, independent of the specific dark-photon model details.","The exclusion extends to heavy neutral Higgs bosons, ruling out $\\sigma_{ZH}\\times\\mathcal{B}(H\\to\\text{invisible}+\\gamma)$ from about 40 fb at 125 GeV down to about 4 fb at 300 GeV.","Fitting the $m_T$ shape rather than counting events improves the expected sensitivity by 30–50%, showing that shape-based extraction is the effective route in this final state.","The irreducible standard model background from $H\\to Z\\gamma\\to\\nu\\nu\\gamma$ is below 0.1 events, so the search is limited by instrumental and reducible backgrounds, not by the standard model Higgs itself."],"supporting_citations":[{"why":"Supplies the dark-photon model with a charged dark sector and shows the branching fraction can be as large as 5%, setting the signal hypothesis this search tests.","marker":"[15]"},{"why":"The earlier 8 TeV CMS search for Higgs decays to undetectable particles plus photons; this analysis extends and improves it with a shape-based fit.","marker":"[20]"},{"why":"Provides the same-flavor lepton-pair control-region method used to estimate the nonresonant WW and top-quark backgrounds.","marker":"[52]"},{"why":"Defines the LHC Higgs combination likelihood procedure that the binned maximum-likelihood fit follows.","marker":"[53]"},{"why":"Supplies the standard model ZH production cross sections used to convert the cross-section limit into a branching-fraction limit at mH = 125 GeV.","marker":"[29]"},{"why":"The CLs technique used to compute the 95% confidence-level upper limits.","marker":"[62]"},{"why":"Provides the asymptotic formulae for the test statistic used in the limit calculation.","marker":"[64]"},{"why":"The POWHEG NLO generator used to simulate the ZH signal and the dominant diboson background processes.","marker":"[23]"}],"fun_headline_variants":["No sign of Higgs decaying to photon plus dark photon; limit at 4.6%","First limits on Higgs decays to photon plus invisible dark photon","Dark photon in Higgs decays? CMS finds nothing, sets first limits","Higgs to photon+invisible? No excess, limit 4.6% from CMS","First search for Higgs to photon plus dark photon: no excess"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The limit assumes the simulated shapes of the $m_T$ distributions for the WZ and ZZ backgrounds are correct in the signal region, because the analysis normalizes those simulations in control regions containing only a handful of events and then relies on the same shapes to model the signal region.","fun_headline_variants_meta":{"raw":{"variants":["No sign of Higgs decaying to photon plus dark photon; limit at 4.6%","First limits on Higgs decays to photon plus invisible dark photon","Dark photon in Higgs decays? CMS finds nothing, sets first limits","Higgs to photon+invisible? No excess, limit 4.6% from CMS","First search for Higgs to photon plus dark photon: no excess"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001663,"raw_usage":{"total_tokens":6636,"prompt_tokens":1018,"completion_tokens":5618,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":634,"completion_tokens_details":{"reasoning_tokens":5519}},"tokens_in":634,"tokens_out":5618,"duration_ms":37589,"temperature":1.0,"reasoning_tokens":5519,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:37:04.487393+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the 95% confidence-level limit twice, once with the WZ background $m_T$ shape taken directly from the trilepton control-region data and once from simulation; if the two limits differ by more than the quoted systematic uncertainty, the background-shape assumption breaks. Alternatively, a dedicated data sample of WZ events with a fully identified third lepton, selected with the same missing-transverse-momentum and photon requirements as the signal region, would reveal any $m_T$ shape mismatch between data and simulation in the relevant phase space.","supporting_citations":[{"cited_title":"Search for exotic decays of a Higgs boson into undetectable particles and one or more photons","cited_arxiv_id":"1507.00359","evidence_quote":"The earlier 8 TeV CMS search for Higgs decays to undetectable particles plus photons; this analysis extends and improves it with a shape-based fit."},{"cited_title":"Procedure for the LHC Higgs boson search combination in Summer 2011","cited_arxiv_id":null,"evidence_quote":"Defines the LHC Higgs combination likelihood procedure that the binned maximum-likelihood fit follows."}],"review_version":1}