{"id":"d0201f2e-3a96-4d20-9cbd-5e00f539727b","arxiv_id":"2608.12616","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A beam-dump experiment using a 10-ton hybrid Cherenkov-scintillation detector found no axion-like particles, but demonstrated background rejection about six times stronger than its predecessor, yielding improved exclusion limits with less exposure.","lead":"This paper reports a search for axion-like particles using a 10-ton liquid argon detector that separates fast Cherenkov light from slower scintillation light. No signal was found, but the new hybrid detector technique rejected background about six times better than the previous version of the experiment, pointing to a promising approach for future rare-particle searches.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Prompt beam-neutrino background in the ROI is not modeled; the uniform prebeam extrapolation may overstate the hybrid-detector sensitivity gain.","rationale":"The reader's weakest-assumption assessment identifies the same load-bearing point: the background model extrapolates a prebeam steady-state rate into the prompt ROI without quantifying beam-induced, prompt backgrounds. I agree with that assessment and with the conditional verdict. The central null result is supported by the data and the fit: the best-fit signal has only 0.86 sigma significance, and the observed ROI count is consistent with the prebeam rate after profiling the background normalization. The concern therefore does not threaten the no-excess claim; it threatens the strength of the sensitivity-improvement claim, because the expected background and the exclusion contour depend on the assumption that no prompt neutrino component survives the LLR selection. The paper itself acknowledges that prompt neutrinos reach the detector in the same window but neither estimates their rate nor validates the LLR against them. The neutron-dominated control sample is not a substitute, since those events arrive later and have higher energy and different topology. The proposed simulation check is decisive: if the predicted beam-neutrino background is negligible relative to the 0.17 events/ns prior width, the uniform background model is safe; if it is not, the limits should be recomputed and the comparison to CCM120 revisited. Until that check is done, CONDITIONAL is the appropriate verdict, and no change to the reader's recommendation is needed.","tokens_in":18112,"tokens_out":11416,"duration_ms":146446,"concrete_test":"Generate the full Lujan pion/muon decay-at-rest neutrino flux with the same GEANT4 target model used for the ALP flux, propagate nu_mu, anti-nu_mu, and nu_e through the CCM200 detector simulation and the exact reconstruction and LLR selection, and count predicted events in the -600 to -424 ns ROI for 0.2-10 MeV. If the predicted prompt neutrino background is below about 0.17 events/ns, the width of the background-normalization prior, the uniform prebeam model is adequate. If it exceeds a few percent of the 11.8 events/ns background, refit the limits with the neutrino-induced time and energy template included and compare the resulting 90% CL contour to the CCM120 contour in Fig. 13.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's sensitivity claim depends on the assumption, introduced in Sec. IV.A and Sec. IV.B.1, that all background in the physics ROI (-600 to -424 ns) is the same steady-state population measured before the beam spill. This is where the central claim is least secure. The ROI is deliberately chosen to begin with the arrival of prompt relativistic particles from the target, and the text in Sec. I notes that prompt neutrinos and near-light-speed BSM particles reach the detector with minimal delay. Neutrino interactions from pion/muon decay-at-rest are therefore expected in this window, yet no beam-neutrino background template is built or validated. The LLR discriminant in Sec. III.E is constructed from prebeam data and ALP Monte Carlo; the only data-driven validation sample is the neutron-dominated region, which is explicitly excluded from the ROI and has a different time/energy structure. A prompt neutrino component passing the LLR>1 cut would produce observed counts and an energy/time shape that the uniform, prebeam-derived background model cannot describe. Because the background normalization is constrained by a Gaussian prior of width 0.17 events/ns in Sec. IV.B.1, even a modest unmodeled beam-induced rate would force the nuisance parameter or, worse, the ALP signal template to absorb the excess, biasing the fitted signal and the resulting 90% confidence-level contour. Since the headline improvement over CCM120 is expressed as a factor of roughly six in prebeam rejection in Table II, not as a neutrino-background-subtracted sensitivity, this omission is load-bearing for the 'surpass CCM120' claim. The null result itself would survive a larger background, so the concern is about the strength of the sensitivity gain, not about the existence of a signal.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports a search for axion-like particles (ALPs) in the CCM200 liquid argon detector at the Lujan spallation source, using the hybrid Cherenkov-scintillation readout. The analysis defines four discriminating observables (prompt uncoated-PMT hit multiplicity, directionality, pulse-shape ratio, and spatial RMS), combines them into a log-likelihood-ratio discriminant, and applies it to 1.23×10^21 protons on target. A two-dimensional template fit in reconstructed energy and event start time, with a uniform prebeam background and three profiled nuisance parameters, finds no significant excess; the most signal-like fit has a local significance of 0.86σ. The authors present 90% CL exclusion limits and argue that, despite using only ~70% of the CCM120 exposure, the ~6-fold improvement in prebeam background rejection gives comparable or better sensitivity across the 10^-3–10 MeV mass range.","tokens_in":18339,"tokens_out":10315,"duration_ms":103038,"significance":"If the background model is correct, this is a valuable proof-of-principle: it is the first ALP search with a hybrid Cherenkov-scintillation detector at a beam dump, and the Cherenkov-based rejection is physically well motivated. The analysis is honest in reporting a null result, and it rests on several strengths: a data-driven prebeam background estimate, direct measurement of the six-fold steady-state rejection (0.49% vs 3.2%), a neutron-dominated control sample, and a calibrated simulation chain (position reconstruction validated with 22Na, energy response validated at the ~10% level). The main risk is not the internal consistency of the prebeam model but its extrapolation into the physics ROI, where prompt neutrinos are expected; this threatens the sensitivity claim as stated.","major_comments":[{"comment":"The background model is a uniform PDF normalized to the prebeam rate of 11.82 ± 0.17 events/ns, but the physics ROI (-600 to -424 ns) begins with the arrival of prompt relativistic particles, and Sec. I explicitly notes that \"prompt neutrinos and near-speed-of-light BSM particles reach the detector with minimal delay.\" The paper does not construct a beam-neutrino background template, a beam-on/beam-off subtraction, or a data-driven estimate of the neutrino rate after the LLR>1 selection. Because the Gaussian prior on the background normalization is only 0.17 events/ns, a prompt-neutrino rate at or above that level could bias the fitted background and signal, and thereby the 90% CL limits in Fig. 13 and the comparison with CCM120 in Table II. Please add a quantitative estimate of this background or a control measurement that validates the uniform extrapolation into the ROI.","section":"Sec. IV.B.1 and Sec. IV.A"},{"comment":"The only data-driven validation of the LLR discriminant is the neutron-dominated sample, which the paper excludes from the ROI because of its later arrival and different event structure. This control cannot validate the LLR behavior for a prompt neutrino component, whose electromagnetic final states resemble the ALP signal. An additional in-beam control, for example a comparison of the ROI events with a scaled beam-off sample or a dedicated neutrino simulation, is needed to support the claim that the background is fully described by the prebeam model.","section":"Sec. III.F and Sec. I"}],"minor_comments":[{"comment":"The choice of the LLR threshold (LLR>1) is described qualitatively; if it was selected after inspecting the data, the possible selection bias should be discussed, and if it was fixed a priori, that should be stated.","section":"Sec. III.E"},{"comment":"The prebeam time distribution is shown only down to -1000 ns; adding a marker or inset for the ROI (-600 to -424 ns) would make the uniform extrapolation easier to assess.","section":"Fig. 7"},{"comment":"The table reports individual cut efficiencies, but it is not clear whether the 18.29% and 80.93% entries are sequential efficiencies for events passing the preceding cuts or inclusive fractions; this should be clarified.","section":"Table I"},{"comment":"The ~5 cm per-coordinate position resolution is quoted for simulated electron events, while the 22Na validation shows widths of order 10-13 cm because the source emits gammas that convert at a distance; the main text should state this distinction where the resolution is quoted.","section":"Appendix A.1"},{"comment":"Minor wording issues include the subject-verb agreement in \"allows ... demonstrate\" (should be \"demonstrates\") and the repeated use of \"proof-of-concept\" where \"proof-of-principle\" is used elsewhere; these should be harmonized.","section":"Abstract and Section V"}],"recommendation":"major_revision","confidential_remarks":"To the editor: The main risk is the unquantified prompt-neutrino background in the ROI; this is a correctable issue, not a fundamental flaw. The collaboration has the tools to estimate it from previous CCM neutrino flux papers and simulations. I recommend major revision rather than rejection. The paper is within scope; the novelty is the hybrid detector technique rather than the physics reach."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a legitimate null result and a worthwhile technical demonstration. The claim that the upgraded detector and new selection improve steady-state background rejection about sixfold over CCM120 is directly supported by the measured prebeam efficiencies (0.49% vs 3.2%). The new exclusion limits are modest, but real, and were obtained with 70% of the prior exposure. I think the paper deserves peer review.\n\nWhat is actually new: this is the first ALP search using hybrid Cherenkov-scintillation discrimination, with four observables (uncoated-PMT prompt hits, directionality, pulse-shape ratio, spatial RMS) combined into a likelihood-ratio classifier. The detector work is also nicely validated with 22Na calibration and the GraphNeT-based position reconstruction is a clear improvement. The paper is honest about the null result: the best-fit signal is 0.86 sigma and the nuisance parameters barely move.\n\nThe soft spots are mostly documentation gaps. The largest is the background model in the physics ROI. The ROI is intentionally chosen to start when prompt relativistic particles from the target arrive, but the background template is the prebeam steady-state sample, assumed time-independent. The paper does not give an explicit estimate of beam-neutrino induced events in that window. That is a real omission, but the stress-test note overstates the risk. The observed time distribution in the ROI (Fig. 12) is consistent with uniform, and the selection (0.2-10 MeV, LLR>1) is unlikely to pass many neutrino-related events. Still, a single paragraph with an order-of-magnitude estimate would close the gap and should be requested by referees.\n\nTwo smaller points. The LLR>1 threshold is stated without documenting how it was chosen, and the systematic uncertainties are limited to three nuisance parameters with no variation of the signal PDFs. For a proof-of-concept with limited statistics, these are acceptable, but they are exactly what a referee should probe.\n\nThe citation pattern is fine: the signal model and optical calibrations come from earlier CCM papers that are already published, and the same collaboration validated the detector response. No circularity concern.\n\nWho is this for? Experimentalists working on beam-dump BSM searches, liquid-argon detector R&D, and hybrid optical detection. It deserves a serious referee, with the request to quantify the beam-neutrino background and clarify the LLR optimization.","headline":"A solid null ALP search and useful proof-of-concept for hybrid Cherenkov-scintillation detection; the main soft spot is an unquantified prompt neutrino background in the ROI, but it does not undermine the central result.","tokens_in":715,"tokens_out":2341,"would_cite":true,"duration_ms":56569,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["14.80.Va","29.40.Mc"],"model":"deepseek-v4-flash","headline":"No axion excess found; hybrid detector beats limits with less data","keywords":["axion-like particles","hybrid Cherenkov-scintillation detector","liquid argon detector","beam dump experiment","Cherenkov radiation","log-likelihood ratio","background rejection","ALP exclusion limits"],"falsifier":"Measure the event rate in the identical prompt window with the proton beam diverted or the target removed: if the rate exceeds the prebeam extrapolation in a way that scales with beam intensity, the uniform-background assumption fails and the derived limits would need revision.","tokens_in":17862,"feed_emoji":"⚛️","tokens_out":11442,"duration_ms":87390,"temperature":0.7,"pith_summary":"This paper reports the first axion-like particle (ALP) search carried out with a hybrid Cherenkov-scintillation optical detector in a beam-dump environment. Using the CCM200 liquid-argon detector, in which 80% of the photomultiplier tubes are wavelength-shifted to see scintillation light while 20% remain bare to catch prompt Cherenkov photons, the analysis builds four timing and topology observables and combines them into a log-likelihood ratio. No statistically significant excess over the expected steady-state background is found for ALP masses between $10^{-3}$ MeV and $10$ MeV. The paper's central positive claim is that the Cherenkov-based rejection reduces the surviving background to 0.49%, roughly six times lower than the predecessor CCM120 analysis, so the search excludes new ALP parameter space even though it uses only about 70% of the previous exposure. If this claim holds, hybrid optical readout can substitute optical information for raw exposure in rare-event searches.","feed_headline":"No axion excess found; hybrid detector beats limits with less data","feed_subtitle":"Sixfold better background rejection extends axion limits using 30% less beam exposure.","key_machinery":"The load-bearing mechanism is the hybrid readout: 80% of the photomultiplier tubes are coated with tetraphenyl butadiene (TPB), which shifts the 128 nm liquid-argon scintillation light into the visible, while 20% are left uncoated and therefore preferentially see the prompt visible component of Cherenkov radiation. Because Cherenkov photons arrive promptly and with a characteristic angular pattern, while scintillation light is delayed and isotropic, the difference in PMT type and in nanosecond timing gives the detector four discriminating handles. Those handles feed a log-likelihood ratio whose probability densities are built from simulated ALP events (signal) and prebeam steady-state data (background); the LLR $> 1$ cut is what reduces the background by a factor of about six relative to CCM120. The same time structure defines the physics region of interest: relativistic ALPs arrive in the prompt window from $-600$ to $-424$ ns, before the slower neutron flux.","core_discovery":"The paper's central claim is that separating Cherenkov light from scintillation light in a single liquid-argon detector is a working background-rejection tool for accelerator beam-dump searches, and that it makes a smaller-exposure search more sensitive than a larger predecessor. The argument runs through four observables, each capturing a different fingerprint of the directional, prompt Cherenkov light that ALP-induced electromagnetic showers produce: prompt hits on uncoated PMTs, a charge-weighted directionality metric, a pulse-shape ratio, and a charge-weighted spatial RMS. Combined into a log-likelihood ratio with a cut at LLR $> 1$, these observables reduce the prebeam steady-state sample to 0.49% of its original size, versus 3.2% for CCM120. Fitting the two-dimensional energy-time distribution in the prompt window gives a best-fit ALP signal at $m_a = 0.18$ MeV and $g_{a\\gamma} = 1.63 \\times 10^{-4}\\,\\mathrm{GeV}^{-1}$ with local significance $0.86\\sigma$, which the paper reads as no evidence for ALPs; the resulting 90% confidence exclusion contour still pushes beyond the CCM120 bound over much of the $10^{-3}$ to $10$ MeV range.","pith_inferences":["The same hybrid separation could be applied to other beam-dump signatures with electromagnetic final states, such as dark photons, millicharged particles, or coherent neutrino scattering, where neutron-induced backgrounds are the main obstacle.","The uniform-background assumption could be tested with a target-out or beam-off run: if the prompt-window rate depends on beam intensity, part of the 'steady-state' background is actually beam-induced and the limits would shift.","If future detectors push the uncoated-PMT fraction or timing resolution further, the directionality metric could evolve into a true Cherenkov ring-imaging observable, converting the current proof-of-concept into a precision background tagger.","A 5 MeV feature in the selected background, possibly from neutron capture on argon, suggests that previous-spill neutrons are captured in the detector; quantifying this component could improve background modeling in the next analysis."],"forward_implications":["Within the probed mass range, the observed 90% confidence exclusion contour is the first ALP constraint produced by a hybrid Cherenkov-scintillation detector at a beam-dump facility.","The sixfold improvement in steady-state background rejection means hybrid optical detection can compensate for reduced exposure in rare-particle searches, not just add modest discrimination.","The successful use of the neutron-dominated sample as a data-driven cross-check indicates the same four-observable selection can be transferred to other electromagnetic final-state searches in the same detector.","Extending the calibrated energy range to roughly 50 MeV with Michel electrons would let the CCM200 program reach higher-mass ALPs, where the current 10 MeV cutoff suppresses efficiency.","With the full dataset, roughly $3\\times10^{21}$ protons on target, the same hybrid selection is expected to strengthen these limits further, since the signal scales linearly with beam exposure."],"supporting_citations":[{"why":"Establishes the calibration-level demonstration that uncoated PMTs can tag prompt Cherenkov light, the technique this selection extends to the full detector volume.","marker":"[1]"},{"why":"Supplies the optical model, pulse unfolding, and 22Na calibration that underpin the energy and timing reconstruction used in the fit.","marker":"[2]"},{"why":"Defines the previous CCM120 ALP search whose sensitivity and 3.2% background efficiency this analysis is compared against.","marker":"[3]"},{"why":"Establishes the beam-spill timing, neutron arrival delay, and the prompt ROI boundaries used to isolate ALP events.","marker":"[32]"},{"why":"Computes Primakoff production and inverse-Primakoff and diphoton detection cross sections for the ALP signal grid.","marker":"[38]"},{"why":"Simulates the target gamma-ray flux and the detector optical response used to build signal templates.","marker":"[40]"},{"why":"Propagates ALPs from target to detector and generates final-state photons for the signal simulation chain.","marker":"[41]"},{"why":"Provides the binned likelihood formalism that accounts for finite Monte Carlo statistics in the template fit.","marker":"[46]"}],"fun_headline_variants":["Hybrid detector sets axion limits with 30% less beam data","Axion search uses Cherenkov trick to beat bigger detector","No axions but hybrid light split beats previous limits","CCM200 hybrid method tightens axion bounds on less data","First hybrid Cherenkov-scintillation axion search sharpens limits"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the background measured during the quiet time before each beam pulse also holds inside the brief prompt window where ALPs would appear, so any beam-related neutrino events arriving in that window would be silently counted as uniform background.","fun_headline_variants_meta":{"raw":{"variants":["Hybrid detector sets axion limits with 30% less beam data","Axion search uses Cherenkov trick to beat bigger detector","No axions but hybrid light split beats previous limits","CCM200 hybrid method tightens axion bounds on less data","First hybrid Cherenkov-scintillation axion search sharpens limits"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000292,"raw_usage":{"total_tokens":1772,"prompt_tokens":1081,"completion_tokens":691,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":697,"completion_tokens_details":{"reasoning_tokens":602}},"tokens_in":697,"tokens_out":691,"duration_ms":6215,"temperature":1.0,"reasoning_tokens":602,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:03:35.296345+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the event rate in the identical prompt window with the proton beam diverted or the target removed: if the rate exceeds the prebeam extrapolation in a way that scales with beam intensity, the uniform-background assumption fails and the derived limits would need revision.","supporting_citations":[{"cited_title":"7), with a best-fit value of11.82±0.17events/ns","cited_arxiv_id":null,"evidence_quote":"Establishes the calibration-level demonstration that uncoated PMTs can tag prompt Cherenkov light, the technique this selection extends to the full detector volume."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the optical model, pulse unfolding, and 22Na calibration that underpin the energy and timing reconstruction used in the fit."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the previous CCM120 ALP search whose sensitivity and 3.2% background efficiency this analysis is compared against."},{"cited_title":"First Demonstration of a Hybrid Cherenkov and Scintillation Detector in a Proof-of-Principle Axion Search at a Beam Dump","cited_arxiv_id":"2607.07476","evidence_quote":"Propagates ALPs from target to detector and generates final-state photons for the signal simulation chain."}],"review_version":1}