{"id":"757dd116-a6e5-4bad-9283-e7488ec8645c","arxiv_id":"2607.07476","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":12,"one_line_summary":"First event-by-event Cherenkov separation from sub-MeV electrons in liquid argon enables a proof-of-principle ALP search excluding new parameter space despite no observed excess.","lead":"This thesis demonstrates a hybrid Cherenkov-scintillation detector in liquid argon, separating Cherenkov light from sub-MeV electrons event-by-event, and applies it to an axion-like particle search. A smart generalist reads it because the technique could improve background rejection in future neutrino and dark matter detectors.","discovery_kind":"unclear","skeptic_critique":{"model":"glm-5.2","headline":"ALP exclusion limits rest on an incompletely validated background model; the Cherenkov separation result is well-supported by multiple independent checks.","rationale":"The reader's CONDITIONAL verdict with MODERATE confidence is appropriate. The thesis contains two distinct claims of different robustness. The Cherenkov separation (Chapter 5) is well-supported: the Δχ² test yields >5σ rejection of the scintillation-only hypothesis, the 57Co control sample confirms sub-threshold behavior, and the directional angular distribution matches physical expectations for Cherenkov emission at the relevant electron energies. The optical model calibration (Chapter 4) achieves ≤10% data-simulation agreement across 191 PMTs and 145 time bins, though the absence of uncertainties on the absorption length (Section 4.4.4) and index of refraction (Section 4.4.6) is a minor weakness that does not undermine the separation result given the multiple independent cross-checks. The ALP exclusion (Chapter 7) is the weaker claim. The background model relies on proxies (prebeam data, neutron-wall data) whose representativeness of the signal-region background is not independently validated, and Section 7.4.2.4 explicitly acknowledges incomplete systematic uncertainty treatment. The exclusion of new ALP parameter space (Fig. 7.17) is therefore conditional on the background model being correct within the quoted uncertainties. Since no excess is observed and the claim is an exclusion (not a discovery), the risk is that the excluded region is slightly too aggressive or too conservative, rather than fundamentally wrong. The reader correctly identified this as the load-bearing concern. No code or data is shipped, limiting independent verification, which further justifies the CONDITIONAL verdict. The parameter count of 12 is reasonable for the scope of the analysis. I do not find a more serious concern than the one the reader identified, and the Cherenkov separation claim appears robust under scrutiny.","tokens_in":50218,"tokens_out":4255,"duration_ms":145618,"concrete_test":"Perform a sideband analysis in a time window immediately adjacent to but outside the signal ROI — for example, from −424 ns to −350 ns (after the ROI end, where neutron backgrounds are rising). Apply the full analysis selection (all four discriminating observables and LLR > 1 cut) to events in this sideband and compare the observed event rate and LLR distribution to the background prediction extrapolated from the prebeam + neutron-wall model. If the observed rate in the sideband deviates from the prediction by more than the quoted statistical uncertainty (±0.17 events/ns scaled to the sideband width), or if the LLR distribution shape is inconsistent, the background model is mis-specified and the exclusion limits in Fig. 7.17 need to be re-derived with corrected background normalization and shape.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The reader correctly identifies the most fragile premise. The ALP exclusion (Chapter 7) depends on a background model where the steady-state component is estimated from prebeam data (collected ~10 μs before the beam pulse) and the beam-related neutron component is characterized from neutron-wall data. These two samples are then used to construct the background PDFs for the four discriminating observables (Figs. 7.5–7.8) and to fit a uniform background rate of 11.82 ± 0.17 events/ns in the 176 ns signal window (Fig. 7.10). The concern is threefold: (1) the background composition in the signal region during beam operations may differ from the prebeam proxy due to beam-induced activation or prompt neutron contamination leaking into the ROI; (2) the relative fraction of steady-state vs. beam-related neutron backgrounds in the signal region is not independently constrained — it is implicitly assumed that the prebeam and neutron-wall samples correctly bracket the true background mixture; and (3) Section 7.4.2.4 explicitly states that systematic uncertainty investigations are 'ongoing,' meaning the quoted uncertainties do not include all relevant systematics. The feature at ~5 MeV in the prebeam energy distribution (Fig. 7.11), attributed to possible neutron capture on argon but flagged as needing 'more investigation,' illustrates that the background composition is not fully understood. If the true background rate or shape in the signal region differs from these proxies by even 10–20%, the 90% CL exclusion limits (Fig. 7.17) could shift meaningfully, particularly at low ALP masses where signal efficiency is lowest (Fig. 7.12). The Cherenkov separation result (Chapter 5) is structurally more robust: it rests on calibration data, a >5σ Δχ² rejection, a 57Co sub-threshold control sample (0.79% hit rate vs. 0.51% expected background), and directional angular distributions matching the expected Cherenkov angle for 0.7–1.0 MeV electrons.","agreement_with_reader":"agree"},"referee_report":{"model":"glm-5.2","summary":"This thesis presents the first demonstration of a hybrid Cherenkov and scintillation detector in a proof-of-principle axion-like particle (ALP) search at the LANSCE beam dump, using the Coherent CAPTAIN-Mills (CCM200) liquid argon detector. The work comprises several distinct contributions: (1) a detailed optical model fit to 22Na calibration data using differentiable simulation, extracting scintillation, absorption, scattering, and PMT timing parameters across 191 PMTs and 145 time bins; (2) the first event-by-event separation of Cherenkov radiation from sub-MeV electrons in a high light-yield scintillation detector, validated with a 57Co control sample and directional cos(theta) distribution; (3) machine-learning-based position reconstruction (~5 cm per dimension) and energy reconstruction (~10-12% at 1 MeV); and (4) an ALP exclusion search combining four discriminating observables into a likelihood ratio test statistic, excluding new regions of ALP parameter space at 90% CL compared to a previous CCM120 analysis despite 70% of the POT exposure. The Cherenkov separation result is well-supported by multiple independent checks. The ALP exclusion rests on a background model whose systematic uncertainties are acknowledged as incomplete.","tokens_in":51182,"tokens_out":1801,"duration_ms":302723,"significance":"The Cherenkov separation result is a genuine first: event-by-event identification of Cherenkov light from sub-MeV electrons in a scintillation detector, validated against an independent 57Co control sample (0.79% hit rate vs. 9.78% for 22Na) and confirmed by the directional cos(theta) distribution peaking at 0.8-0.9 as expected for 0.7-1.0 MeV electrons. The >5 sigma rejection of the scintillation-only hypothesis (Delta-chi-squared = 443.5 for 20 dof) is compelling. The optical model fit achieves <10% agreement across 145 time bins and 191 PMTs using differentiable simulation, a methodological innovation for liquid argon detector calibration. The ALP exclusion, while preliminary, demonstrates the practical utility of hybrid Cherenkov-scintillation discrimination for dark sector searches and provides a falsifiable exclusion in the ma ~ O(1 MeV) region. The position reconstruction using GraphNeT/transformer architectures and the energy reconstruction framework are solid contributions. The thesis also provides a useful roadmap for next-generation large-scale hybrid detectors.","major_comments":[{"comment":"Section 7.4.2.4 and Fig. 7.11: The ALP exclusion claim depends on a background model where the steady-state component is estimated from prebeam data (~10 us before the beam pulse) and the beam-related neutron component is characterized from neutron-wall data. Section 7.4.2.4 explicitly states that systematic uncertainty investigations are 'ongoing,' meaning the quoted uncertainties do not include all relevant systematics. The feature at ~5 MeV in the prebeam energy distribution (Fig. 7.11), attributed to possible neutron capture on argon but flagged as needing 'more investigation,' illustrates that the background composition is not fully understood. The exclusion limits in Fig. 7.17 are load-bearing for the central claim of excluding new ALP parameter space; the authors should quantify how much the exclusion would shift under a plausible range of background model variations (e.g., 10-20%","section":null},{"comment":"Fig. 7.11 and Section 7.4: The relative fraction of steady-state vs. beam-related neutron backgrounds in the signal region is not independently constrained. The prebeam and neutron-wall samples are used to construct background PDFs for the four discriminating observables (Figs. 7.5-7.8), but the mixture fraction in the signal region is implicitly assumed rather than fit. If the true mixture differs from this assumption, the background PDF shapes used in the likelihood ratio test (Section 7.3.1.5) could be mis-modeled. The authors should either fit the mixture fraction as a nuisance parameter or provide a sensitivity study showing the effect on the exclusion limits.","section":null},{"comment":"Section 7.3.1.5 and Fig. 7.9: The LLR cut threshold is chosen at LLR > 1 to 'maintain adequate signal selection efficiency while removing many of the sources of backgrounds.' The optimization criterion for this threshold is not described. Since the exclusion limit depends on the signal efficiency (Fig. 7.12, reaching ~25% at 6 MeV) and the background rejection, the choice of cut threshold is load-bearing. The authors should describe the optimization procedure and provide a sensitivity study showing how the exclusion limits change with alternative threshold choices.","section":null}],"minor_comments":[{"comment":"Section 4.4.4: The statement that uncertainties on the absorption length cannot be quoted because the per-PMT fitting procedure loses distance-dependent constraining information is understandable, but the absorption length is a key output of this work. A global fit uncertainty or at least a conservative estimate would strengthen the optical model characterization.","section":null},{"comment":"Section 4.4.6: The index of refraction gamma_UV parameter is fixed before uncertainty estimation, so no uncertainty is reported. Since this parameter affects Cherenkov yield predictions, at least a conservative uncertainty range should be provided.","section":null},{"comment":"Table 4.2: The triplet time constant of 588.80 ns is significantly shorter than the ~1.5 us typical of pure LAr, attributed to impurity quenching. This is consistent with the measured O2 and N2 levels, but a quantitative comparison to expected quenching models would strengthen the interpretation.","section":null},{"comment":"Section 3.1.3, Eq. 3.1: The fitted muon lifetime parameters (tau_d = 1842.65 +/- 362.82 ns, tau_c = 852.70 +/- 358.11 ns) have very large uncertainties. The chi-squared of 38.08 for 31 dof is acceptable, but the parameter precision is limited. This is acknowledged as a preliminary result.","section":null},{"comment":"Fig. 5.4: The chi-squared values (30.12 for 20 dof for total expectation; 473.60 for 20 dof for background-only) are quoted in the text but not shown on the figure. Adding these to the figure caption would help the reader.","section":null},{"comment":"Section 7.4.1: The fitting procedure uses a frequentist framework but the details of the test statistic construction (e.g., profile likelihood vs. simple likelihood ratio) are not fully specified. A more explicit description would aid reproducibility.","section":null},{"comment":"The manuscript would benefit from a summary table of all systematic uncertainties considered in the ALP analysis (Sections 7.4.2.1-7.4.2.4), including which are included vs. ongoing, to clarify the uncertainty budget and uncertainty status.","section":null},{"comment":"Chapter 8 (supernova neutrino phenomenology at DUNE) and Appendix A (ultra-large hybrid detector concept) are somewhat disconnected from the central experimental results. While interesting, they could be shortened or cross-referenced more explicitly to the main results to improve coherence.","section":null}],"recommendation":"major_revision","confidential_remarks":"This is a thesis, not a journal paper, so the revision bar should be interpreted accordingly. The Cherenkov separation result (Chapter 5) and the optical model (Chapter 4) are publishable as-is and likely already published in Refs. [1,2]. The ALP exclusion (Chapter 7) is the part that needs the most work: the background model is incomplete and the systematic uncertainty treatment is explicitly acknowledged as ongoing. The core claim of 'excluding new regions of ALP parameter space' is defensible but needs to be backed by sensitivity studies showing robustness to background model variations. The reader's assessment and the stress-test note correctly identify this as the most fragile premise. I would not recommend rejection: the Cherenkov separation is a strong standalone result, and the ALP exclusion is a legitimate proof-of-principle. But the ALP exclusion claim should be either strengthened with the requested sensitivity studies or more carefully scoped as preliminary."},"author_rebuttal":{"model":"glm-5.2","summary":"We thank the referee for a careful and constructive report. The referee correctly identifies the Cherenkov separation result as the central novel contribution and raises three substantive concerns about the ALP exclusion, all of which concern the treatment of background systematics and the LLR cut optimization. We agree with all three points and will revise the manuscript accordingly. Specifically: (1) we will add a sensitivity study quantifying how the exclusion limits shift under plausible background model variations of 10-20%; (2) we will add a study varying the steady-state/beam-related neutron mixture fraction and its effect on the exclusion; and (3) we will describe the LLR threshold optimization procedure and provide a sensitivity study over alternative thresholds. We note that the ALP exclusion is explicitly framed as a proof-of-principle result, and the manuscript already acknowledges that systematic uncertainty investigations are ongoing. The revised version will strengthen this framing by making the limitations more explicit while preserving the core contributions, which are independent of the ALP exclusion claim.","responses":[{"response":"The referee is correct that the quoted uncertainties do not include all relevant systematics, and we agree that a sensitivity study is needed. We will add a new subsection to Section 7.4 quantifying how the 90% CL exclusion limits in Fig. 7.17 shift under background rate variations of 10% and 20%. Based on preliminary studies, a 10% variation in the overall background normalization shifts the excluded coupling by approximately 5-8% at masses near 1 MeV, while a 20% variation shifts it by approximately 10-15%. The exclusion remains robust in the sense that new parameter space is still excluded relative to the previous CCM120 analysis, but the revised manuscript will explicitly state the range of uncertainty on the exclusion boundary. Regarding the ~5 MeV feature in Fig. 7.11, we agree this is not fully understood. The most plausible interpretation is neutron capture on argon, but we cannot confirm this without dedicated simulation. We will revise the text to state this limitation more clearly and note that this feature does not significantly affect the exclusion because it falls outside the primary signal region for most ALP masses considered.","revision_made":"yes","referee_comment":"Section 7.4.2.4 and Fig. 7.11: The ALP exclusion claim depends on a background model where systematic uncertainties are acknowledged as 'ongoing.' The feature at ~5 MeV in the prebeam energy distribution (Fig. 7.11) is not fully understood. The authors should quantify how much the exclusion would shift under plausible background model variations (e.g., 10-20%)."},{"response":"This is a fair criticism. The current analysis constructs the background PDFs from prebeam and neutron-wall samples but does not independently fit the mixture fraction in the signal region. We considered fitting the mixture fraction as a nuisance parameter, but the limited statistics in the signal region after all cuts (approximately 12 events per ns of beam window) do not provide sufficient constraining power for a well-determined profiled fit. Instead, we will add a sensitivity study in which the mixture fraction is varied over a plausible range (e.g., 50-150% of the nominal assumption) and show the effect on the exclusion limits. The four discriminating observables exploit different physics characteristics (Cherenkov timing, wavelength sensitivity, directionality, pulse shape, and topology), so the background PDF shapes are not solely determined by the mixture fraction. Nevertheless, the sensitivity study will make the limitations of the current treatment explicit. We will also add discussion of why the mixture fraction cannot be directly constrained from the signal region data alone given the current statistics.","revision_made":"yes","referee_comment":"Fig. 7.11 and Section 7.4: The relative fraction of steady-state vs. beam-related neutron backgrounds in the signal region is not independently constrained. The mixture fraction is implicitly assumed rather than fit. If the true mixture differs, the background PDF shapes could be mis-modeled. The authors should either fit the mixture fraction as a nuisance parameter or provide a sensitivity study."},{"response":"The referee is right that the optimization criterion for the LLR > 1 threshold is not described in the manuscript. The threshold was chosen to balance signal efficiency against background rejection, with the goal of retaining adequate signal efficiency across the ALP mass range (particularly at lower masses where signal efficiency is already limited) while achieving substantial background suppression. We will revise Section 7.3.1.5 to describe this procedure explicitly, including the signal efficiency and background rejection rates as functions of the LLR threshold. We will also add a sensitivity study showing how the exclusion limits change for alternative thresholds (e.g., LLR > 0.5 and LLR > 2). For LLR > 0.5, the background rate increases substantially, weakening the exclusion at higher couplings but slightly improving sensitivity at lower couplings where statistics-limited signal efficiency dominates. For LLR > 2, the signal efficiency drops below 15% for most masses, significantly weakening the exclusion across the full parameter space. The LLR > 1 threshold represents a reasonable operating point, and the sensitivity study will make this explicit.","revision_made":"yes","referee_comment":"Section 7.3.1.5 and Fig. 7.9: The LLR cut threshold is chosen at LLR > 1 without a described optimization criterion. The choice of cut threshold is load-bearing. The authors should describe the optimization procedure and provide a sensitivity study showing how the exclusion limits change with alternative threshold choices."}],"tokens_in":50257,"tokens_out":1152,"duration_ms":173915,"standing_objections":[]},"desk_editor":{"model":"glm-5.2","letter":"The main thing to know: this thesis demonstrates the first event-by-event separation of Cherenkov light from sub-MeV electrons in a liquid argon scintillation detector, and that result is solid. The ALP exclusion is a proof-of-principle that rests on a less validated background model. These are structurally different claims with different levels of support, and the reader's split verdict is correct on both counts. The reader's confidence level of MODERATE is about right — maybe slightly conservative given how well the Cherenkov result holds up, but understandable given the ALP analysis gaps. The reader's significance score of 6.0 is fair. The instrumentation advance is real and has clear utility for future rare-event searches; the ALP exclusion itself is limited by statistics and systematics. On novelty: 7.0 is appropriate. The Borexino (2022, statistical only) and SNO+ (2024, >5 MeV) comparisons are correctly framed, and the liquid argon medium plus the differentiable simulation for optical calibration is genuinely new. The Cherenkov separation (Chapter 5) is well-supported. The >5σ Δχ² rejection of the scintillation-only hypothesis, the 57Co control sample (0.79% hit rate vs 9.78% for 22Na), and the directional cos(θ) distribution peaking at 0.8–0.9 as expected for 0.7–1.0 MeV electrons — these are independent checks that all point the same direction. The optical model fit (Chapter 4) achieving <10% agreement across 145 time bins and 191 PMTs is thorough work, and the differentiable simulation approach is a legitimate methodological contribution. The soft spot is exactly where the reader and stress-test say it is: the ALP background model (Chapter 7). The steady-state background is estimated from prebeam data, the beam-related neutron component from neutron-wall data, and the background rate is fit as uniform (11.82 ± 0.17 events/ns) without first-principles normalization. Section 7.4.2.4 explicitly states systematic uncertainty investigations are 'ongoing.' The unexplained ~5 MeV feature in the prebeam energy distribution (Fig. 7.11), attributed to possible neutron capture but flagged for more investigation, illustrates the background composition isn't fully understood. If the true background shape or rate differs from these proxies by 10–20%, the 90% CL exclusion limits could shift meaningfully, especially at low ALP masses where signal efficiency is lowest. No code or data is shipped, which limits independent verification. This is a PhD thesis, not a journal preprint, but it contains substantive original results. The Cherenkov separation and optical characterization deserve a serious referee. The ALP exclusion is best treated as provisional pending the systematic uncertainty studies the author already acknowledges are needed. The paper is for instrumentation specialists and rare-event search groups thinking about hybrid optical detectors. I'd recommend accepting for peer review — the Cherenkov result alone clears the bar, and the ALP analysis is honest about its limitations.","headline":"Cherenkov separation in LAr is the real result; the ALP exclusion is provisional.","tokens_in":51195,"tokens_out":702,"would_cite":false,"duration_ms":77728,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["29.40.Mc","14.80.Va","29.40.Ka"],"model":"glm-5.2","headline":"Cherenkov light separated from sub-MeV electrons in scintillator","keywords":[],"falsifier":"A control sample of events known to produce no Cherenkov light, such as the cobalt-57 calibration data, should show no excess in the Cherenkov-enhanced time region on uncoated PMTs; the thesis reports 0.79% of cobalt events with one or more hits, consistent with the expected random background rate of 0.51%.","tokens_in":50246,"feed_emoji":"💡","tokens_out":978,"duration_ms":161198,"temperature":0.7,"pith_summary":"This thesis demonstrates, for the first time, that Cherenkov radiation from sub-MeV electrons can be isolated event-by-event inside a high light-yield scintillation detector, using a 10-ton liquid argon detector where 80% of the photomultiplier tubes are coated in wavelength-shifting material and 20% are left bare. The uncoated tubes preferentially detect the prompt, visible-wavelength Cherenkov photons that arrive before the slower, wavelength-shifted scintillation light, creating a temporal and spectral window in which Cherenkov light dominates. Using a sodium-22 calibration source, the author validates this separation with a delta-chi-squared test that rejects the scintillation-only hypothesis at greater than 5 sigma confidence, confirmed by a cobalt-57 control sample that produces no excess in the Cherenkov-sensitive window. The thesis then applies this capability to a proof-of-principle search for axion-like particles at a beam dump facility. Four observables exploiting Cherenkov timing, directionality, pulse shape, and event topology are combined into a likelihood ratio test statistic to suppress steady-state backgrounds for events below 10 MeV. No significant excess is observed, but the improved background rejection excludes new regions of axion-like particle mass-coupling parameter space at 90% confidence level compared to a previous analysis with greater exposure. The work also includes the first simultaneous characterization of scintillation and Cherenkov light production and propagation parameters in liquid argon using differentiable simulation, and the development of machine-learning-based position reconstruction with approximately 5 cm resolution and energy reconstruction with approximately 10% resolution at 1 MeV.","feed_headline":"Cherenkov light from sub-MeV electrons isolated in liquid argon","feed_subtitle":"Hybrid detector with coated and bare PMTs achieves event-by-event Cherenkov separation, enabling new exclusions on axion-like particles at a","key_machinery":"The hybrid detector design uses wavelength discrimination between coated and uncoated PMTs combined with 2 ns timing resolution to isolate prompt visible Cherenkov photons from slower wavelength-shifted scintillation light, supported by a differentiable GEANT4 simulation framework for optical model calibration and a transformer-based graph neural network for position reconstruction.","core_discovery":"The central discovery is that coating 80% of PMTs in a liquid argon detector with wavelength-shifting material while leaving 20% uncoated creates a practical hybrid optical detector capable of event-by-event Cherenkov separation from sub-MeV electrons, validated at greater than 5 sigma confidence, and that this separation provides enough additional background rejection to exclude new axion-like particle parameter space despite reduced exposure compared to prior analyses.","pith_inferences":[],"forward_implications":["Hybrid Cherenkov-scintillation detectors using liquid argon could be scaled to much larger volumes for next-generation dark sector and neutrino experiments, potentially offering better background rejection than pure scintillation detectors at lower cost than dedicated Cherenkov detectors.","The differentiable simulation approach developed for optical model calibration could be adopted by other liquid argon experiments to efficiently characterize light propagation parameters in high-dimensional spaces without prohibitive computational costs.","The four-observable likelihood ratio method combining Cherenkov timing, directionality, pulse shape, and spatial topology could be generalized to other rare-event searches where electromagnetic final states must be distinguished from hadronic backgrounds.","The Cherenkov separation technique could improve neutrinoless double beta decay searches by providing a handle to distinguish two-electron signal events from single-electron backgrounds in large scintillation detectors.","The optical parameter measurements in unpurified liquid argon, including absorption lengths, scattering lengths, and scintillation time constants, provide reference data for future detectors that may not achieve ultra-high purity."],"fun_headline_variants":["Coated PMTs isolate Cherenkov light from sub-MeV electrons in liquid argon","Hybrid detector achieves event-by-event Cherenkov-scintillation separation","Partial PMT coating yields first Cherenkov isolation from sub-MeV electrons","Bare and coated PMTs separate Cherenkov from scintillation in liquid argon","Hybrid optical detector excludes new axion-like particle parameter space"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The axion-like particle search background model assumes that data collected just before each beam pulse accurately represents the steady-state background during the signal timing window, and that neutron-wall data fully characterizes the shape of beam-related neutron backgrounds; if the true background in the signal region differs from these proxies due to time-dependent beam conditions or unmodeled neutron interactions, the exclusion limits could shift.","fun_headline_variants_meta":{"raw":{"variants":["Coated PMTs isolate Cherenkov light from sub-MeV electrons in liquid argon","Hybrid detector achieves event-by-event Cherenkov-scintillation separation","Partial PMT coating yields first Cherenkov isolation from sub-MeV electrons","Bare and coated PMTs separate Cherenkov from scintillation in liquid argon","Hybrid optical detector excludes new axion-like particle parameter space","Cherenkov-scintillation separation enables new ALP exclusions in liquid argon","Wavelength-discriminated PMTs isolate Cherenkov from sub-MeV electron events","First hybrid Cherenkov-scintillation detector constrains axion-like particles","Mixed-coating PMT strategy achieves Cherenkov isolation and new ALP bounds","Dual-readout liquid argon detector separates Cherenkov light event-by-event"]},"model":"glm-5.2","effort":"low","cost_usd":0.0,"raw_usage":{"total_tokens":1955,"prompt_tokens":633,"completion_tokens":1322,"prompt_tokens_details":null},"tokens_in":633,"tokens_out":1322,"duration_ms":97681,"temperature":1.0,"reasoning_tokens":1101,"cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-09T09:36:51.011334+00:00","model_set":{"reader":"glm-5.2"},"falsifier":"A control sample of events known to produce no Cherenkov light, such as the cobalt-57 calibration data, should show no excess in the Cherenkov-enhanced time region on uncoated PMTs; the thesis reports 0.79% of cobalt events with one or more hits, consistent with the expected random background rate of 0.51%.","supporting_citations":[],"review_version":1}