{"id":"f8c97a74-995e-4344-a682-7738e589a910","arxiv_id":"2608.12065","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"CONUS+ presents a complete Monte Carlo-based background decomposition of its run 1 data, showing cosmic muons and neutrons dominate below 1 keV while reactor-correlated backgrounds remain an order of magnitude below the neutrino signal.","lead":"This paper breaks down every source of background noise in the CONUS+ neutrino detector and shows that cosmic rays, not the nuclear reactor, dominate the low-energy signal region. The full model matches measured rates below 1 keV, strengthening the earlier claim of coherent neutrino scattering at a reactor.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Cosmic-neutron input is the load-bearing external assumption: the largest ROI background component rests on a 2004 Yorktown Heights spectrum scaled to KKL with no local measurement and no spectral-shape validation.","rationale":"Good-faith reading: the paper is a careful, largely transparent background decomposition built on measured reactor-neutron input, muon simulations validated against no-veto data, radon line normalizations, and known activation lines. The central claim that reactor-correlated backgrounds are subdominant is well supported by the direct room-neutron measurement and by the small simulated reactor-neutron rate. My stress-test does not find an internal contradiction that invalidates the decomposition; the weakest load-bearing point is external: cosmic neutrons dominate the ROI, and their input is a 2004 surface spectrum scaled by altitude and latitude with no local spectral constraint. The reader identified exactly this as the weakest assumption, and I agree. The reader's other points (the 'first measurement at a reactor' overclaim and the partially empirical ROI components) are correct but secondary; they do not threaten the cosmic-dominance and reactor-subdominance conclusion as directly as the cosmic-neutron input. Because the concern is a serious caveat rather than a demonstrated error, the conditional verdict stands; the natural path to acceptance is an independent spectral check or a local neutron-flux measurement. I therefore leave the reader's CONDITIONAL verdict unchanged.","tokens_in":23990,"tokens_out":13497,"duration_ms":121377,"concrete_test":"Re-run the Section 5.2 building-shield simulation with an independent high-energy cosmic-neutron spectrum measured at a similar altitude and latitude (e.g., a PSI, CERN-EU, or CRN spectrum) instead of the Gordon et al. Yorktown spectrum, keeping the measured (7.3±0.1) m w.e. overburden fixed. Compute the resulting [0.4,1.0] keVee cosmic-neutron rate for all three detectors. If the rate changes by more than the quoted ±15% component uncertainty, or if the model-data residual in [0.4,1.0] keVee shifts beyond 1σ, the external input is under-constraining the dominant background and the model requires a local cosmic-neutron normalization before it can anchor the CEνNS likelihood fit.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 5.2 adopts the Gordon et al. 2004 surface cosmic-neutron spectrum from Yorktown Heights (altitude 170 m) and scales only its integral flux to the KKL location (altitude 350 m), explicitly assuming no change in spectral shape. Section 9.3 states that no reliable cosmic-neutron spectrum could be derived from the KKL characterization measurement, and a dedicated measurement is only planned. Table 8 makes this component the largest single term in the CEνNS ROI: 50.3% (C5), 42.8% (C2), and 44.2% (C3) of the reactor-on [0.4,1.0] keVee model. Because the 7.3 m.w.e. overburden and 20 cm lead strongly attenuate the spectrum, the low-energy recoil rate is controlled by the high-energy (>10 MeV) tail of the input spectrum, whose normalization and slope are exactly the parts most sensitive to altitude, geomagnetic latitude, and local overburden. The quoted ~15% uncertainty on this component covers only the integrated surface flux and concrete density, not a shape uncertainty. If the fast-neutron tail were e.g. 30% higher or lower than the Yorktown representation, the ROI model would shift by roughly 10-15%, moving the [0.4,1.0] keVee model-data agreement from sub-1σ to several σ and changing the background normalization used in the likelihood fit of [1]. The strong internal checks (muon simulation versus no-veto data, radon line intensity ratios, gamma-peak scaling) do not constrain the cosmic-neutron spectrum, since they use other normalizations or do not isolate this component. This is not an internal contradiction, but the central claim that the background model is safe for the CEνNS fit is only as secure as this unvalidated external spectrum.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents a complete Monte Carlo based decomposition of the CONUS+ run 1 background for the three germanium detectors used in the CEνNS analysis. Using the MaGe/Geant4 framework, the authors simulate reactor neutrons, inert gases, cosmic muons, cosmic neutrons, muon-induced neutrons in the overburden, radon, cosmogenic activation products, 210Pb, and two detector-specific empirical components. They report that cosmic muons and cosmic neutrons together dominate the sub-keV region, that reactor-correlated backgrounds are subdominant and about an order of magnitude below the expected CEνNS signal below 350 eVee, and that the final model reproduces the measured spectra with an average rate of (47.5 ± 3.2) d^-1 kg^-1 in the [0.4, 1.0] keVee reactor-on region versus (48.0 ± 0.6) d^-1 kg^-1 measured. The model is presented as the input to the likelihood fit used in the companion CEνNS analysis [1].","tokens_in":24254,"tokens_out":8835,"duration_ms":78564,"significance":"If the model is correct, this paper provides the background basis for the first observation of reactor CEνNS by CONUS+ and is a useful template for shallow-depth, low-threshold germanium experiments. The work has clear strengths: the muon component is anchored to a measured local muon flux and validated against no-veto data; the reactor-neutron input comes from a dedicated Bonner-sphere measurement; radon and cosmogenic components are normalized to observed gamma lines; and the sub-400 eVee region was kept blinded during model construction. The manuscript is also transparent about its main weakness, namely the absence of a local cosmic-neutron spectrum measurement. Because the cosmic-neutron component is the largest single term in the CEνNS ROI, and because the high-energy agreement of the model is not as good as the abstract suggests, the central claim needs additional systematic support before the model can be considered validated for the likelihood fit.","major_comments":[{"comment":"The cosmic-neutron component, which is the largest single term in the [0.4,1] keVee ROI (50.3% for C5, 42.8% for C2, and 44.2% for C3 in reactor-on data), is normalized using the Gordon et al. (2004) surface spectrum from Yorktown Heights scaled in integral flux only, with no change to the spectral shape. Since the 7.3 m.w.e. overburden preferentially transmits the high-energy tail, the ROI recoil rate is controlled by exactly the part of the input spectrum that is most sensitive to altitude, latitude, and local overburden. The ~15% uncertainty quoted in Section 9.3 covers the integrated surface flux and concrete density but not the spectral shape. The authors should quantify a shape systematic by reweighting the cosmic-neutron spectrum within plausible variations (for example, changing the fast-neutron spectral index or comparing with an alternative surface spectrum model) and report the resulting change in the [0.4,1] keVee prediction and in the likelihood fit of [1]. Without this, the claim that the ROI model is safe for the CEνNS extraction is not fully supported.","section":"Section 5.2 and Table 8"},{"comment":"The abstract and Section 9.2 state that the model agrees with the data in all energy regions, with differences mostly within 1-2σ, but Table 8 contains several larger discrepancies. For example, the C5 reactor-off [100,250] keVee bin has model 283.7 ± 4.8 d^-1 kg^-1 versus data 242.6 ± 3.6 d^-1 kg^-1, a difference of about 6.9σ; the C3 reactor-on [100,250] keVee bin has model 658.9 ± 11.3 d^-1 kg^-1 versus data 606.6 ± 2.3 d^-1 kg^-1, about 4.5σ; and the C5 reactor-on [30,100] keVee bin has model 204.5 ± 4.6 d^-1 kg^-1 versus data 220.9 ± 1.4 d^-1 kg^-1, about 3.4σ. The text should qualify the all-energy agreement claim and discuss whether these high-energy deficits indicate normalization issues in components such as radon, 210Pb, or metastable germanium states that could affect the low-energy model through shared systematics.","section":"Table 8 and Section 9.2"},{"comment":"The leakage-test background is an empirical term fitted to data in the same sub-keV region used to validate the model. For C2 it contributes 14.3% of the model below 0.4 keVee and 6.5% in [0.4,1] keVee; for C3 the contributions are 9.8% and 1.9% (Table 8). The agreement below 1 keVee for these two detectors is therefore not a fully independent check of the background model. The authors should state the fit range and data set used to determine θ1 and θ2, report the values with uncertainties, and provide a cross-check such as fitting the leakage parameters to reactor-off data alone or demonstrating that the [0.4,1] keVee model-data agreement is insensitive to the parameter values within their uncertainties.","section":"Section 8.3 and Eq. (2)"},{"comment":"The muon-veto efficiency below 15 keVee is described by a polynomial that is normalized at 15 keVee and applied to the dominant muon-induced component in the ROI. The polynomial is derived from the ratio of simulated untagged to tagged muon events, and the text does not give the fitted polynomial parameters, their uncertainties, or the covariance between them. Since the muon component contributes roughly 33-46% of the [thr,0.4] and [0.4,1] keVee model rates, the authors should either report the fit results and propagate their uncertainty, or show via a variation test that the ROI prediction is insensitive to the polynomial shape.","section":"Section 5.1 and Eq. (1)"}],"minor_comments":[{"comment":"The phrase 'Similar agreement is found in all energy regions' should be reworded to acknowledge the high-energy discrepancies listed in Major Comment 2, for example by stating that the agreement is best below 1 keVee and that some high-energy bins show larger deviations.","section":"Abstract and Section 9.2"},{"comment":"The sentence containing Eq. (1) states that the fitted values are 'taken to be the inverse muon veto efficiency' and then writes epsilon = 1 - FV_norm; the terminology is confusing because epsilon is the efficiency, not the inverse efficiency, and the normalized polynomial FV_norm is not defined explicitly.","section":"Section 5.1"},{"comment":"The text refers to spectra drawn in black, blue, red, and green, but the figure caption does not define the colors; a legend entry matching the colors to the surface spectrum, the room total, cosmogenic neutrons, and muon-induced neutrons should be added.","section":"Section 5.2 and Figure 12"},{"comment":"Equation (2) would be easier to use for reproducibility if the fitted numerical values of θ1 and θ2 for C2 and C3 were reported, together with the energy range over which the fit was performed.","section":"Section 8.3"},{"comment":"Figure 3 compares the reactor-neutron simulation to 'C5 run 1 data' without specifying whether the shown data are from the reactor-on or reactor-off period; the text implies reactor-on, but the caption should state this explicitly.","section":"Section 4.1"},{"comment":"The sentence 'The line was recently measured for the first by the CONUS+ experiment' should read 'for the first time', and the reference [33] appears to lack venue information.","section":"Section 7"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is well within the scope of physics.ins-det and is important as the background model behind a published reactor CEνNS result. The main risk is the unvalidated spectral shape of the cosmic-neutron input, which is the dominant term in the CEνNS ROI; this is a standard difficulty for shallow-depth experiments, and I would not reject on that basis if the shape systematic is quantified in the revision. The high-energy discrepancies in Table 8 also need to be addressed in the text, since the abstract currently overstates the all-energy agreement."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThis is the background model behind the CONUS+ run 1 CEνNS claim, and it deserves a serious referee. The genuinely new pieces are the full decomposition of the three detectors' run 1 spectra, the observation that cosmic neutrons dominate the sub-keV region at 7.3 m w.e., and the muon-veto inefficiency mechanism from muons that never cross the scintillator plates. The reactor-neutron component is computed from the measured Bonner-sphere flux and comes out an order of magnitude below the expected CEνNS signal; that is the useful result.\n\nThe paper is mostly careful. The MaGe framework, Lindhard quenching, radon line scaling, and cosmogenic activation are carried over from the CONUS background paper, but that is the right toolkit and the authors are transparent about which components are simulated, scaled, or empirical. The ROI agreement (47.5 vs 48.0 d^-1 kg^-1) is good, and the muon simulation is validated against no-veto data.\n\nThe soft spots are real but not fatal. First, the abstract and introduction claim CONUS+ achieved the first measurement of CEνNS at a nuclear reactor. That is wrong on the record: Dresden-II reported it in 2022. The paper cites Dresden-II, so the claim is internally inconsistent. Second, the abstract says similar agreement in all energy regions, but Table 8 shows a clear discrepancy in the [100,250] keVee reactor-off interval, roughly 17% high for C5 and several sigma. That should be reworded at minimum. Third, part of the low-energy agreement is data-driven: the leakage background for C2/C3 is an exponential fitted to the same sub-keV region, and the 60Co contamination in the cryostat is a guess with no line. Those terms are small fractions of the ROI, but they do mean the statement \"prediction\" is softer than it sounds.\n\nThe largest substantive concern is the cosmic-neutron input. The surface spectrum comes from a 2004 Yorktown Heights measurement, scaled in integral flux to KKL with no spectral-shape change; Section 9.3 admits no reliable local cosmic-neutron spectrum could be extracted and a dedicated measurement is only planned. Since cosmic neutrons are the single largest ROI component (40-50%), the model's low-energy normalization rests on an external spectrum whose fast-neutron tail is unvalidated at this site. The quoted 15% uncertainty covers flux and concrete density, not shape. I would not call this a fatal flaw--the method is standard and the authors disclose it--but the paper should deal with it more explicitly, ideally by showing how the ROI result shifts under a conservative shape variation.\n\nWho this is for: anyone working on shallow-depth germanium backgrounds or reactor CEνNS. It deserves peer review, with the wording fixes and a sharper treatment of the cosmic-neutron shape uncertainty before it is used as the official background model reference.","headline":"Solid background decomposition for CONUS+ run 1, worth refereeing despite an overbroad first-measurement claim and a load-bearing cosmic-neutron spectrum taken from a 2004 Yorktown measurement.","tokens_in":25040,"tokens_out":2540,"would_cite":true,"duration_ms":22486,"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 establishes a complete decomposition of the CONUS+ run 1 background: cosmic-ray muons and neutrons dominate below 1 keVee, and reactor-correlated backgrounds are one order of magnitude below the CEνNS signal in the search window.","keywords":["coherent elastic neutrino-nucleus scattering","reactor antineutrinos","background decomposition","cosmic-ray neutrons","muon-induced background","high-purity germanium detectors","low-background shielding","Monte Carlo simulation"],"falsifier":"A direct measurement of the cosmic neutron flux and energy spectrum inside the experiment's room at KKL, for example with Bonner spheres or a fast-neutron spectrometer, would settle the question: the simulated in-room flux is $(0.9 \\pm 0.2)\\,\\mathrm{cm}^{-2}\\,\\mathrm{d}^{-1}$, and the cosmic-neutron component contributes $(21.6 \\pm 3.1)\\,\\mathrm{d}^{-1}\\,\\mathrm{kg}^{-1}$ in the $[0.4, 1.0]$ keVee window. A measured flux outside those ranges, or a measured spectrum with a different fast-neutron peak, would falsify the dominant term and require rebuilding the model's normalization.","tokens_in":23689,"feed_emoji":"☢️","tokens_out":10851,"duration_ms":94886,"temperature":0.7,"pith_summary":"The paper tries to establish a complete, component-by-component accounting of every background event seen by the three germanium detectors in the first physics run of a reactor neutrino experiment. Its central assertion is that the backgrounds tied to reactor operation—fast neutrons and noble-gas fission products—are subdominant everywhere, and in the search window below 350 eVee they sit one order of magnitude below the expected coherent-scattering signal (CEνNS). The dominant low-energy backgrounds are instead cosmic-ray muons, cosmic neutrons, and muon-induced neutrons, together about 75–90% of the rate below 1 keVee. If this is right, the reactor on/off difference is not reactor neutrons masquerading as neutrinos, and the likelihood fit that extracted the experiment's first reactor CEνNS signal uses a sound background model. The paper reports agreement between the model and the measured spectra across all energy ranges, including $(47.5 \\pm 3.2)$ versus $(48.0 \\pm 0.6)\\,\\mathrm{d}^{-1}\\,\\mathrm{kg}^{-1}$ in the $[0.4, 1.0]$ keVee window.","feed_headline":"Cosmic rays, not reactor neutrons, dominate CONUS+ backgrounds","feed_subtitle":"Full decomposition places reactor-correlated noise an order of magnitude below the coherent-scattering signal","key_machinery":"The load-bearing object is a Monte Carlo background model assembled source by source. Each candidate component—reactor neutrons started from a measured in-room spectrum, cosmic neutrons propagated from a measured surface spectrum through a model of the reactor building and its overburden, muons propagated from a literature flux through the shield, radon scaled to fitted gamma-line intensities, cosmogenic isotopes from screened material activities, and two detector-specific empirical residuals—is simulated, converted to electron-equivalent energy with the Lindhard quenching model, folded with detector dead-layer, transition-layer, and bulk efficiencies, broadened by the measured resolution, and summed into one predicted spectrum. The argument is carried by the model's blind construction: the sub-400 eVee region is never used to normalize any component, so agreement there, including the one-order-of-magnitude separation between reactor-correlated background and the CEνNS signal, is an extrapolated prediction rather than a fit.","core_discovery":"The core discovery is a full spectral decomposition of the CONUS+ run 1 background that leaves no significant unexplained component below 300 keVee. Reactor neutrons contribute only $(0.2 \\pm 0.1)\\,\\mathrm{d}^{-1}\\,\\mathrm{kg}^{-1}$ between 0.4 and 1 keVee, about two orders of magnitude below cosmic neutrons and below 1% of the total rate in that window; inert gases add roughly 1–3% of the rate between 400 and 1000 eVee. Cosmic-ray muons, cosmic neutrons, and muon-induced neutrons from the overburden together provide about 75–90% of the rate below 1 keVee, with the direct muon component alone around 30–40% in the 400–1000 eVee band. The model was constructed with the CEνNS search window kept blind: all normalizations come from data above 400 eVee and from the reactor-off period, and the resulting prediction agrees with the measured spectra in every energy range for both reactor states.","pith_inferences":["The same practice of scaling a foreign surface neutron spectrum by altitude and latitude is used by other shallow-depth CEνNS projects; a direct measurement at KKL would therefore test not just this model but a widespread technique.","The identified low-energy veto inefficiency comes mostly from muons that miss both scintillator layers, so adding an inner veto layer or exploiting pulse-shape rejection of near-surface events could cut the largest low-energy background component further.","Once the cause of the high-energy efficiency loss is identified, the muon-simulation-based correction could be replaced by a physical correction, extending trustworthy model predictions above 300 keVee.","The success of the blinded construction suggests applying the same discipline—normalize above the ROI and on source-off data—to future low-threshold reactor experiments, including the larger detectors already planned for run 2."],"forward_implications":["Reactor on/off differences are explained by the drywell-lid overburden change, the lower radon level during outage, the disappearance of reactor neutrons, and transient inert-gas release—not by reactor neutrons mimicking neutrinos.","The background model is safe to use as the input to the likelihood fit that extracted the first reactor CEνNS signal, with reactor-correlated backgrounds below 350 eVee one order of magnitude below the expected signal.","A dedicated cosmic-neutron measurement at the experiment's location is the clear next step, because the cosmic-neutron component carries the largest uncertainty while being the dominant low-energy term.","Above tens of keVee, radon in the detector chamber and $^{210}$Pb in the inner lead layer dominate, accounting for up to roughly 65% of the count rate between 100 and 250 keVee in reactor-on data.","The predicted model rate in the $[0.4, 1.0]$ keVee window agrees with the measured rate within uncertainties, supporting the extrapolation of the model into the previously blinded search region."],"supporting_citations":[{"why":"The CEνNS measurement whose likelihood fit this background model feeds; validating the model protects that first reactor CEνNS result.","marker":"[1]"},{"why":"Supplies the simulation chain, energy-response treatment, cosmogenic activation rates, and leakage-test parametrization reused in this work.","marker":"[6]"},{"why":"Supplies the measured reactor-neutron spectrum, muon flux, room survey, and building-propagation result used as simulation inputs.","marker":"[7]"},{"why":"Documents the CONUS+ shield, detector upgrades, and muon-veto geometry implemented in the simulation.","marker":"[9]"},{"why":"Provides the surface muon momentum and angular parameterization propagated through the overburden for the muon background.","marker":"[24]"},{"why":"Provides the atmospheric muon flux model at depth used together with the surface parameterization to normalize the muon component.","marker":"[25]"},{"why":"Supplies the cosmic neutron surface flux and spectrum, scaled to KKL, that normalizes the dominant low-energy component.","marker":"[29]"},{"why":"Provides the Lindhard quenching model that converts neutron energy depositions into electron-equivalent energy in the post-processing chain.","marker":"[19]"}],"fun_headline_variants":["Muons, not reactor, dominate CONUS+ background spectrum","CONUS+ background: cosmic rays rule, reactor negligible","Reactor background an order below signal in CONUS+","CONUS+ background fully modeled: cosmic rays lead","Cosmic muons and neutrons set CONUS+ background"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the cosmic neutron surface flux and spectrum at the experiment's location follow a 2004 measurement made at another site, scaled by altitude and latitude without changing the spectral shape, because no direct cosmic-neutron measurement at KKL exists; since cosmic neutrons are the dominant low-energy component, an incorrect input spectrum directly shifts the model's largest term.","fun_headline_variants_meta":{"raw":{"variants":["Muons, not reactor, dominate CONUS+ background spectrum","CONUS+ background: cosmic rays rule, reactor negligible","Reactor background an order below signal in CONUS+","CONUS+ background fully modeled: cosmic rays lead","Cosmic muons and neutrons set CONUS+ background"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000852,"raw_usage":{"total_tokens":3760,"prompt_tokens":1057,"completion_tokens":2703,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":673,"completion_tokens_details":{"reasoning_tokens":2622}},"tokens_in":673,"tokens_out":2703,"duration_ms":18615,"temperature":1.0,"reasoning_tokens":2622,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:17:41.576954+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct measurement of the cosmic neutron flux and energy spectrum inside the experiment's room at KKL, for example with Bonner spheres or a fast-neutron spectrometer, would settle the question: the simulated in-room flux is $(0.9 \\pm 0.2)\\,\\mathrm{cm}^{-2}\\,\\mathrm{d}^{-1}$, and the cosmic-neutron component contributes $(21.6 \\pm 3.1)\\,\\mathrm{d}^{-1}\\,\\mathrm{kg}^{-1}$ in the $[0.4, 1.0]$ keVee window. A measured flux outside those ranges, or a measured spectrum with a different fast-neutron peak, would falsify the dominant term and require rebuilding the model's normalization.","supporting_citations":[{"cited_title":"Ackermann et al.First observation of reactor antineu- trinos by coherent scattering, 2025","cited_arxiv_id":null,"evidence_quote":"The CEνNS measurement whose likelihood fit this background model feeds; validating the model protects that first reactor CEνNS result."},{"cited_title":"Reyna.A Simple Parameterization of the Cosmic-Ray Muon Momentum Spectra at the Surface as a Function of Zenith Angle, 2006","cited_arxiv_id":null,"evidence_quote":"Provides the surface muon momentum and angular parameterization propagated through the overburden for the muon background."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the atmospheric muon flux model at depth used together with the surface parameterization to normalize the muon component."},{"cited_title":"Gordon et al.Measurement of the flux and energy spectrum of cosmic-ray induced neutrons on the ground","cited_arxiv_id":null,"evidence_quote":"Supplies the cosmic neutron surface flux and spectrum, scaled to KKL, that normalizes the dominant low-energy component."},{"cited_title":"Lindhard, M","cited_arxiv_id":null,"evidence_quote":"Provides the Lindhard quenching model that converts neutron energy depositions into electron-equivalent energy in the post-processing chain."}],"review_version":1}