{"id":"3ec5cabf-0c36-495a-96f2-fa9f2cb5d2a8","arxiv_id":"2505.10353","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"The 20 R5912 PMTs for the SABRE South veto meet performance requirements in bench tests, and prototype pulse-shape discrimination separates gammas from neutrons at the hundreds-of-keV scale.","lead":"Pre-calibration measurements of the R5912 photomultiplier tubes for the SABRE South liquid scintillator veto indicate the detector can tag backgrounds down to tens of keV. The results matter because a well-understood veto is central to SABRE South's test of the DAMA/LIBRA dark matter claim.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Low-threshold veto claim hinges on an unmeasured 12 photons/keV light-yield assumption; a factor-of-two reduction would break the 'most thresholds >50 keV' guarantee.","rationale":"The paper's central quantitative claim converts simulated PE statistics into low-threshold veto performance. Equation 2.3 and the 0.75 PE/keV figure are the multiplicative bridge between physical inputs and every efficiency in Table 1. The measured PMT DE in Sec. 4.5 is higher than assumed, so the PMT side is conservative; the weak point is the scintillator light yield. It is cited from other LAB-based experiments but never measured for this detector or the prototype, and the prototype's charge windows are not converted to an absolute energy scale, so no in-situ constraint is provided. Since the 20-50 keV and 50-80 keV bins have mean PE counts of order 10-40 total, they are the first to degrade as light yield drops. A simple binomial estimate at LY=6 gives a large efficiency loss for the PMTs>=3, PEs>=3 threshold at 50 keV, which directly contradicts the 'most thresholds' wording in the conclusion. This is a validation gap rather than a demonstrated error: the PMT characterization itself appears sound, and the design conclusions are qualitatively unchanged. The reader already conditioned acceptance on in-situ light-yield validation, so the verdict should remain conditional.","tokens_in":21076,"tokens_out":15623,"duration_ms":154080,"concrete_test":"Re-run the Geant4 simulation used for Table 1 with the light yield set to 8 and then 6 photons/keV, leaving geometry, reflectivity, QE and thresholds unchanged, and recompute the 20-50 and 50-80 keV efficiencies for at least the PMTs>=3, PEs>=3 and PMTs>=4, PEs>=1 rows. If either row drops below ~99% in the 50-80 keV bin at LY=6, the conclusion's 'most thresholds guarantee >50 keV' is not robust; if both remain ~100%, the concern does not land.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing input is the assumed scintillator light yield of 12 photons/keV (Sec. 2.1). Together with a ~25% photon collection probability and 25% PMT QE, it produces the 0.75 PE/keV figure used to scale every threshold in Table 1 and the conclusion. The PMT detection-efficiency side is actually measured and conservative (DE_R5912 = 0.286 at 405 nm vs 0.25 assumed, Sec. 4.5), but the light yield is only referenced to other LAB scintillators [16,17]; the paper reports no absolute light-yield measurement for the SABRE fill or for the Sec. 5 prototype. The thresholds sit near the Poisson/coincidence edge: at 50 keV the simulated mean is ~2.1 PE per PMT, and the PMTs>=3, PEs>=3 row of Table 1 reports 100% in the 50-80 keV bin. If the true light yield is 6 photons/keV instead of 12, the mean drops to ~1.0 PE per PMT, the probability that a PMT sees >=3 PE falls from ~0.35 to ~0.09, and the same threshold would fail on a large fraction of 50-80 keV deposits. The 'most thresholds guarantee >50 keV' claim therefore rests on an unvalidated factor of 2 in light yield.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports the pre-calibration campaign for the Hamamatsu R5912 photomultiplier tubes that will instrument the 12 kL linear-alkylbenzene liquid scintillator veto of SABRE South. The measurements cover single-photoelectron charge and gain, temperature-dependent dark rate, timing and transit time spread, relative detection efficiency, linearity and saturation, afterpulsing, and spontaneous light emission from the oil-proof base. The paper also presents a Geant4-based simulation that translates PMT coincidence and photoelectron thresholds into veto efficiencies for 40K backgrounds, and a small prototype study of gamma/neutron pulse-shape discrimination using the same PMT model and liquid scintillator, including boosted decision trees and frequency-domain variables. The central claim is that, with the characterized PMTs, the veto can operate at low thresholds and efficiently veto background deposits down to tens of keV, with an expected 0.75 PE/keV and near-100% efficiency above 50 keV for several threshold configurations.","tokens_in":21479,"tokens_out":10552,"duration_ms":104959,"significance":"If the quantitative claims hold, the paper is a valuable reference for the SABRE South veto system and for low-threshold liquid scintillator veto operation generally. The PMT characterization is careful and useful: it covers twenty tubes, checks against Hamamatsu specifications, and includes effects such as afterpulsing, base light emission, and temperature-dependent dark rate that are often omitted from similar studies. The measured detection efficiency of 0.286 at 405 nm is reassuringly close to the assumed 25% quantum efficiency. The prototype PSD study is a sensible exploratory step toward particle identification in the final detector. However, the headline veto-efficiency numbers and the keV-scale conclusions rest on an unvalidated light-yield assumption and on in-sample machine-learning evaluation, so the quantitative significance of those specific claims is currently conditional.","major_comments":[{"comment":"The quoted 0.75 PE/keV figure and every veto efficiency in Table 1 scale linearly with the assumed light yield LY = 12 PE/keV, cited to other LAB scintillators [16,17], together with an assumed average photon-collection probability of about 25%. No absolute light-yield measurement is reported for the SABRE fill or for the Sec. 5 prototype. This is load-bearing because the low-energy rows sit near the Poisson detection edge: with 0.75 PE/keV and 18 PMTs, a 50 keV deposit gives about 2.1 PE per PMT, and the \"PMTs>=3, PEs>=3\" row reports 100% efficiency in the 50-80 keV bin. If the true LY were a factor of two lower, the per-PMT mean would drop to about 1.0 PE, the probability that a PMT sees at least 3 PE falls from about 0.35 to about 0.09, and the same threshold would miss a substantial fraction of 50-80 keV deposits. Please provide a measured LY for the veto scintillator, or at least a sensitivity scan of Table 1 as a function of LY, and temper the conclusion accordingly.","section":"Secs. 2.1–2.2, Table 1, Sec. 6"},{"comment":"The machine-learning performance quoted in Sec. 5.3.1 and Sec. 5.5 (about 80% efficiency for the best single variable and about 90% for the BDT at 10% false-positive rate) is reported without any described train/test split or cross-validation. Evaluating a classifier on the same data used to train it inflates the reported discrimination. The same concern applies to the frequency-domain variable F of Eq. (5.4), which is normalized by dataset averages computed over the same events later scored by the BDT. Please add an out-of-sample evaluation, such as k-fold cross-validation or a held-out dataset, report the resulting ROC values, and reassess the claim in Sec. 5.5 that pulse-shape variables will remain effective down to the 100s-of-keV scale.","section":"Secs. 5.3.1, 5.4, 5.5"},{"comment":"The convolution of the Poisson photon-number distribution with the binomial detection process is written as an integral over d n, but n is an integer photon count and Pois(n; ...) is defined only for integer n. As written, the expression is not well defined. The compound distribution has the closed form Poisson(LY*E * sum_i QE_i P_Di) if the detection probabilities are independent and exclusive; please either write the sum over n or state the closed form and confirm that the simulation uses an equivalent implementation. This matters because these equations are the formal basis for Table 1.","section":"Eqs. (2.3)–(2.4)"}],"minor_comments":[{"comment":"The parameter N1PE is used as a mixing fraction between 0 and 1 but is described as a number of SPEs; renaming it to something like f_1PE would avoid confusion.","section":"Sec. 4.2, Eq. (4.1)"},{"comment":"The quoted DE_R5912 = 0.2862 +/- 0.0009 includes only the statistical error; the systematic uncertainty from the assumed R11065 QE of 0.3325, the DE=QE approximation, beam-splitter nonuniformity, and the applied charge and peak-height cuts should be estimated and reported.","section":"Sec. 4.5"},{"comment":"The measured transit time spread is compared with the Hamamatsu nominal value, but the laser trigger itself has a quoted spread of 5 +/- 3 ns; the text should state explicitly that this contribution is folded into the measured TTS.","section":"Sec. 4.4"},{"comment":"The fit to the spontaneous light emission rate versus bias voltage is described only as \"loosely fitted\"; if the curve is to be used quantitatively, give the fit function and parameters, otherwise label it as illustrative.","section":"Sec. 4.8"},{"comment":"The conclusion that \"for most thresholds, energy deposits above 50 keV are guaranteed to be detected\" is stronger than Table 1 supports: the \"PMTs=18, PEs>=1\" row shows 25.2% efficiency in the 50-80 keV bin and 58.0% in the 80-100 keV bin, and the \"PMTs>=1, PEs>=6\" row shows 72.6% in 50-80 keV. Please qualify the wording.","section":"Sec. 6"},{"comment":"The source is spelled both \"Am-Be\" and \"Am-BE\"; please use a single spelling throughout.","section":"Sec. 5.2"}],"recommendation":"major_revision","confidential_remarks":"The PMT characterization work is solid and genuinely useful, and the simulation/PSD study is a reasonable first look at veto performance. The main barrier to acceptance is the unvalidated 12 PE/keV light-yield assumption that drives the headline low-threshold claim; this can be addressed with a measured light yield or with a sensitivity scan of Table 1, and the BDT results need out-of-sample evaluation. I do not see a need for rejection if those points are addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a competent detector-characterization paper with a real batch-specific dataset; the headline low-threshold veto efficiencies are conditional on a borrowed light yield, and the ML discrimination numbers are in-sample, so treat those parts as indicative rather than final.\n\nWhat's new: measured SPE/gain/dark-rate/timing/DE/afterpulse parameters for 20 oil-proof Hamamatsu R5912 PMTs, including the relative DE against a reference R11065 at 405 nm (0.286 ± 0.001, conservative relative to the 0.25 assumed in simulation), a voltage-dependent spontaneous light emission study for the potted bases, and a prototype LAB scintillator study with the same PMT model showing neutron/gamma PSD with BDTs and a frequency-domain noise variable. The measurement methods are standard but applied carefully; the numbers look consistent with manufacturer specs and previous experiments. That is the solid core.\n\nSoft spots: the simulated veto efficiencies and the 0.75 PE/keV figure depend on assuming 12 photons/keV light yield for the liquid scintillator, taken from other LAB scintillators. The paper presents no measurement of the light yield for the SABRE fill or for the prototype. Everything in Table 1 and the conclusion scales linearly with this number. If the real yield were 6 photons/keV, the statement that energy deposits above 50 keV are guaranteed would not survive for the stricter coincidences. The PMT-side detection efficiency is actually measured and is not the weak link; the light-yield assumption is. The conclusion should flag this explicitly, and the collaboration should measure or bound the light yield in situ before quoting these thresholds as requirements.\n\nAlso, the PSD/BDT performance is evaluated without a train/test split, and the frequency-domain variable F is normalized by the dataset average and scored on the same dataset. That can inflate the quoted separation. For a prototype methods study this is forgivable, but the 80–90% efficiency numbers should be presented as proof-of-principle, not as expected veto performance.\n\nMinor: afterpulsing and spontaneous-light studies use one or two PMTs; fine for a first look.\n\nVerdict: worth a serious referee. The pre-calibration measurements are reproducible and the paper is honest about what is measured versus assumed, except the conclusion overreaches. I would send to JINST with requested revisions: add an explicit light-yield uncertainty or a sensitivity scan, move the PSD numbers to cross-validated form, and soften the 'guaranteed' language. The paper is useful for anyone instrumenting a liquid scintillator veto with R5912s or planning PSD at low PE.","headline":"Solid PMT pre-calibration for SABRE South's veto, but the low-threshold efficiency claim rests on an unmeasured 12 PE/keV light-yield assumption and in-sample ML scores.","tokens_in":22150,"tokens_out":3023,"would_cite":false,"duration_ms":31508,"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":"Pre-calibrated R5912 photomultiplier tubes allow the SABRE South liquid scintillator veto to reject 40K background deposits above 20–50 keV, supporting a low-threshold test of the DAMA/LIBRA dark matter annual modulation signal.","keywords":["photomultiplier tube characterisation","liquid scintillator veto","SABRE South","dark matter annual modulation","low energy threshold","pulse shape discrimination","40K background","single photoelectron response"],"falsifier":"Fill the 12 kL veto with its final scintillator mixture and measure the position-averaged photoelectron yield using a calibrated gamma-ray source at known energy, or use the installed optical calibration system; if the yield falls below 0.75 PE/keV by more than the simulation's spatial spread, then a 20 keV deposit no longer produces the roughly 15 photoelectrons needed for the claimed veto thresholds.","tokens_in":20939,"feed_emoji":"🔬","tokens_out":7000,"duration_ms":65735,"temperature":0.7,"pith_summary":"This paper argues that the 12 kL liquid scintillator veto of SABRE South, once its R5912 photomultiplier tubes are pre-calibrated, can veto 40K background decays in the sodium-iodide crystals down to tens of keV, supporting a low-threshold test of the DAMA/LIBRA annual modulation claim. Using a Monte Carlo model with a light yield of 12 photoelectrons per keV and 25% average photon-to-photoelectron conversion, the authors derive a position-averaged detection probability of 0.75 PE/keV and show that for most threshold requirements every energy deposit above 50 keV is detected, with the least strict requirements catching everything above 20 keV. The supporting measurements—single-photoelectron charge, gain, dark rate, timing, relative detection efficiency, afterpulsing, and spontaneous base-light emission—are presented as a pre-calibration campaign on up to 20 tubes. A small liquid scintillator prototype with the same PMT model shows that pulse-shape discrimination can separate neutrons from gammas even at about 100–400 pC, and a frequency-domain variable can flag electronic noise. If these results hold, the veto is the key component that lets SABRE South confront the DAMA/LIBRA signal with a background-rejection capability no previous NaI(Tl) experiment had.","feed_headline":"Veto catches 40K deposits down to 20 keV","feed_subtitle":"Pre-calibrated R5912 photomultipliers give 0.75 PE/keV, enough to background-tag a DAMA-style dark matter search.","key_machinery":"The load-bearing object is the photoelectron-yield model: a Poisson distribution for the number of scintillation photons generated by an energy deposit, a binomial distribution for how many of those photons convert to photoelectrons at each PMT (quantum efficiency times a per-PMT, position-dependent hit probability map), and a sum over PMTs that yields the total detected-photoelectron distribution. This model produces the 0.75 PE/keV figure and the veto-efficiency table. The second key object is the single-photoelectron charge-response model, a two-Gaussian SPE fit for bulk gain calibration and a three-Gaussian model including pedestal and under-amplified pulses for simulation-grade digitisation; it gives the gain curve $\\mu_{\\text{gain}} = A V^{kN}$ and defines the charge scale used in thresholds and afterpulse counting. The pulse-shape work uses a machine-learning classifier over variables such as CAP25, CAP50, and charge ratio to establish that discrimination survives at low PE.","core_discovery":"The paper's central claim is that a liquid scintillator veto instrumented with the oil-proof R5912 PMTs can be operated at low photoelectron thresholds and thereby veto the dominant 40K background in NaI(Tl) crystals down to tens of keV. In simulation the veto achieves about 99.9% efficiency for all 40K-related energy deposits under a requirement of at least one triggered PMT with at least one photoelectron; the same requirement vetoes 100% of deposits in the 20–50 keV band and 92.5% in 0–20 keV. With a coincidence of at least four PMTs at one PE each, the 0–20 keV efficiency drops to 71% but the 20–50 keV band remains at 100%. The authors infer a 0.75 PE/keV position-averaged detection probability from a Poisson-binomial model with 12 PE/keV light yield and 25% detection efficiency, and their measured relative detection efficiency of 0.286 for the R5912 at 405 nm is consistent with that assumed 25% at 390 nm. The paper also claims that afterpulsing is rare enough (1.7e-4 per PE) and dark rate low enough that single-PE triggering is viable, and that prototype pulse-shape discrimination keeps neutron/gamma separation at keV-scale energies.","pith_inferences":["Beyond the paper: if the real 12 kL detector's light yield or effective PMT efficiency falls below the assumed 12 PE/keV and 25% conversion, the quoted 20–50 keV thresholds scale linearly; a direct in-situ calibration of photoelectrons per keV would renormalize the thresholds.","Beyond the paper: the frequency-domain noise-rejection variable could be deployed not only in offline analysis but also as a data-compression or trigger-level filter in the full detector, reducing the low-amplitude electronic-noise rate that dominates low-PE samples.","Beyond the paper: if the additional PMTs bring the instrumented total to 32, the resulting denser photocathode coverage could push the practical veto threshold to 20 keV and open the door to supernova-neutrino detection and early-warning participation.","Beyond the paper: the prototype-trained machine-learning pulse-shape classifier may transfer to the full veto only after per-PMT retraining, and the observed loss of discrimination power with decreasing PE statistics suggests a per-detector PE threshold below which particle identification should not be trusted."],"forward_implications":["With the 0.75 PE/keV position-averaged detection probability, a 20 keV deposit yields roughly 15 photoelectrons, enough for multi-PMT coincidence thresholds that reject 40K efficiently.","For most studied threshold requirements, 40K energy deposits above 50 keV are vetoed at 100% efficiency, and the least strict requirements veto all deposits above 20 keV.","The measured SPE charge, gain stability, timing, and relative detection efficiency allow single-PE thresholds to be set across the PMT array, which is necessary for the low-energy veto claims.","Prototype pulse-shape studies show neutron/gamma discrimination remains viable at about 100–400 pC (roughly 112–449 PE), implying effective veto particle identification down to hundreds of keV; the machine-learning classifier reaches about 90% efficiency at a 10% false-positive rate.","The afterpulsing probability of 1.7e-4 per PE and the temperature-dependent dark-rate measurements set the noise budget for low-threshold operation and for monitoring PMT degradation over time."],"supporting_citations":[{"why":"Defines the annual-modulation claim that the experiment is designed to test, setting the scientific stakes for the veto.","marker":"[1]"},{"why":"Supplies the detector configuration, DAQ context, and the initial threshold targets of a few hundred keV with the possibility of 50 keV.","marker":"[2]"},{"why":"Provides the 40K background model and veto-efficiency benchmark that the PMT requirements are built on.","marker":"[9]"},{"why":"Characterises spontaneous light emission from oil-proof PMT bases in a similar liquid scintillator experiment, the effect this paper tests and mitigates.","marker":"[14]"},{"why":"Supplies the optical-photon transport simulation used to build per-PMT hit probability maps.","marker":"[15]"},{"why":"Provides a liquid-scintillator light-yield reference supporting the 12 PE/keV assumption.","marker":"[16]"},{"why":"Provides another scintillator characterisation supporting the same light-yield assumption used in the PE statistics.","marker":"[17]"},{"why":"Gives the manufacturer's nominal specifications for the R5912 PMT, including the 25% quantum efficiency and transit-time spread used in the simulations.","marker":"[18]"},{"why":"Supplies a comparable PMT characterisation method for linearity and saturation used as a cross-check.","marker":"[30]"}],"fun_headline_variants":["SABRE South veto hits 99.9% on 40K down to 20 keV","Oil-proof R5912 PMTs give 0.75 PE/keV for SABRE veto","Single-photoelectron veto catches NaI 40K backgrounds","SABRE South veto: 100% rejection in 20-50 keV band","Low threshold veto for DAMA test achieved with R5912s"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire low-threshold veto claim rests on the assumed scintillator light yield of 12 photoelectrons per keV and an average photon-to-photoelectron conversion of 25%; if the real 12 kL detector delivers less, the energy threshold at which background is reliably vetoed rises.","fun_headline_variants_meta":{"raw":{"variants":["SABRE South veto hits 99.9% on 40K down to 20 keV","Oil-proof R5912 PMTs give 0.75 PE/keV for SABRE veto","Single-photoelectron veto catches NaI 40K backgrounds","SABRE South veto: 100% rejection in 20-50 keV band","Low threshold veto for DAMA test achieved with R5912s"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000315,"raw_usage":{"total_tokens":1814,"prompt_tokens":1001,"completion_tokens":813,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":617,"completion_tokens_details":{"reasoning_tokens":704}},"tokens_in":617,"tokens_out":813,"duration_ms":7499,"temperature":1.0,"reasoning_tokens":704,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T21:10:31.078750+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Fill the 12 kL veto with its final scintillator mixture and measure the position-averaged photoelectron yield using a calibrated gamma-ray source at known energy, or use the installed optical calibration system; if the yield falls below 0.75 PE/keV by more than the simulation's spatial spread, then a 20 keV deposit no longer produces the roughly 15 photoelectrons needed for the claimed veto thresholds.","supporting_citations":[{"cited_title":"Bernabei et al.,First results from DAMA/LIBRA–phase2,Nucl","cited_arxiv_id":null,"evidence_quote":"Defines the annual-modulation claim that the experiment is designed to test, setting the scientific stakes for the veto."},{"cited_title":"The SABRE South Technical Design Report Executive Summary","cited_arxiv_id":"2411.13889","evidence_quote":"Supplies the detector configuration, DAQ context, and the initial threshold targets of a few hundred keV with the possibility of 50 keV."},{"cited_title":"Barberio et al.,Simulation and background characterisation of the SABRE south experiment,The European Physical Journal C83(Sept., 2023)","cited_arxiv_id":null,"evidence_quote":"Provides the 40K background model and veto-efficiency benchmark that the PMT requirements are built on."},{"cited_title":"Characterization of water-based liquid scintillator for Cherenkov and scintillation separation","cited_arxiv_id":"2006.00173","evidence_quote":"Provides another scintillator characterisation supporting the same light-yield assumption used in the PE statistics."},{"cited_title":"Hamamatsu photonics: Large cathode area photomultiplier tubes","cited_arxiv_id":null,"evidence_quote":"Gives the manufacturer's nominal specifications for the R5912 PMT, including the 25% quantum efficiency and transit-time spread used in the simulations."},{"cited_title":"Cryogenic R5912-20Mod Photomultiplier Tube Characterization for the ProtoDUNE Dual Phase Detector","cited_arxiv_id":"1806.04571","evidence_quote":"Supplies a comparable PMT characterisation method for linearity and saturation used as a cross-check."}],"review_version":1}