{"id":"e94d4592-c1b3-4a1c-b194-401bc820b875","arxiv_id":"2506.15164","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":12,"one_line_summary":"JUNO's two PMT types differ strongly in dark count rate stability: NNVT MCP-PMTs need about 50 hours to cool and shift 4-12% per degree Celsius with temperature, while HPK dynode PMTs stabilize faster and shift only about 2% per degree.","lead":"This paper measures how the dark count rate of JUNO's 20-inch photomultiplier tubes changes with cooling time and temperature, and it tracks sudden rate spikes. The results help JUNO set operating conditions and identify 'flasher' PMTs that could add noise to neutrino measurements.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The bare-NNVT 12%/°C temperature coefficient rests on the unverified assumption that drawer temperature sensors track PMT temperature after 2-3 h; the text itself admits incomplete equilibrium during cooling.","rationale":"The reader's weakest assumption is exactly the point I would flag: the temperature coefficients in Tables 2-3 depend on the drawer sensor tracking the PMT temperature, and the paper contains no direct evidence for that tracking. The paper itself provides internal admissions of incomplete thermal equilibrium, so this is not an abstract worry. I also considered the cooling-time fit in Section 3.1, where the 51.7 h slow time constant for NNVT is extracted from only about 50 h of monitoring, making the 56% asymptotic value somewhat extrapolated. However, the temperature coefficient is more directly load-bearing for the paper's operational recommendation that temperature be controlled to ±1°C, and it is not protected by any independent cross-check. The concern is addressable by a dedicated measurement, so a conditional acceptance remains the right disposition. No change to the reader's verdict is needed.","tokens_in":14115,"tokens_out":5042,"duration_ms":49234,"concrete_test":"Repeat the bare-NNVT heating/cooling cycle (Figs 6, 14-28°C) with a calibrated thermocouple attached to the PMT glass envelope or, better, to a spare PMT with an internal sensor, and hold each set point until both the drawer sensor and the PMT sensor are stable to <0.2°C over at least 1 h. Recompute the temperature coefficient using the PMT-sensor temperature rather than the drawer-sensor temperature. If the resulting slope differs from 12%/°C by more than ~30% or the heating and cooling curves do not overlap, the thermal-lag bias is confirmed and the reported coefficient should be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3.2 derives the temperature coefficients from drawer-mounted sensors ('aside the PMT', Figs 6-8) and asserts that holding each set point for 2-3 h allows thermal equilibrium. No independent measurement of PMT bulb, photocathode, or MCP temperature is presented. The concern is not merely formal: §3.2.1 states that NNVT PMTs 'require more time to achieve thermal equilibrium while cooling, particularly at the highest temperature', and §3.2.2 concedes that in the 14-21°C range 'the waiting time ... was shorter'. If the PMT temperature lags the drawer sensor, the normalized DCR versus sensor temperature is not the true thermodynamic response. For the steepest and most consequential result—bare NNVT at ~12%/°C (about 6 kHz/°C)—a lag of 1-2°C would shift the reported slope by roughly 10-20% and could also explain part of the heating/cooling asymmetry in Table 2. The ±1°C temperature-control recommendation in §3.4 is based on this coefficient, so the assumption bears the weight of a headline operational conclusion. The alternative—that the coefficient is real and the lag is negligible—is plausible but currently unsupported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports an empirical characterization of the dark count rate (DCR) of the two types of 20-inch PMTs used in JUNO (HPK dynode and NNVT MCP-PMTs), based on data from the Pan-Asia mass-testing facility. It quantifies DCR evolution during post-loading cooling (Section 3.1), the dependence of DCR on temperature (Section 3.2), and long-term DCR stability including transient spikes (Section 3.3). A two-exponential fit to the normalized cooling curve yields asymptotic DCR fractions (56% for NNVT and 83% for HPK after 12 h normalization) and time constants (about 52 h and 25 h, respectively). Temperature coefficients are reported for bare and potted PMTs, with bare NNVT showing about 12%/°C, potted NNVT about 4%/°C, and HPK about 2%/°C. The paper also describes a preliminary investigation of DCR spikes, including flasher identification using coincidence counting and a single-photon camera.","tokens_in":14425,"tokens_out":6466,"duration_ms":55680,"significance":"If the reported coefficients hold, the paper provides valuable input for JUNO operations: the cooling-time curves justify different PMT-type-specific waiting periods, and the temperature coefficients inform the required temperature stability of the detector. The large-sample, real-world mass-testing context is a strength, as is the direct comparison of two PMT technologies under the same HVAC-controlled container system. The long-term monitoring of 117 PMTs and the short-term spike statistics give a useful baseline for DCR-related false triggers. However, the quantitative headline values (12%/°C, 4%/°C, 2%/°C and the 56%/83% asymptotic fractions) currently lack published uncertainties and rest on a thermal-equilibrium assumption that is not independently validated. The flasher identification is explicitly preliminary and appropriately framed. Overall, the paper is a useful experimental characterization, but the load-bearing temperature coefficients need additional support before the numbers can be taken at face value.","major_comments":[{"comment":"The temperature coefficients are derived from drawer-mounted sensors ('inside the drawers aside the PMT'), not from measurements of the PMT bulb, photocathode, or MCP temperature. The text itself admits incomplete equilibrium: §3.2.1 states that NNVT PMTs 'require more time to achieve thermal equilibrium while cooling, particularly at the highest temperature', and §3.2.2 concedes that in the 14–21°C range 'the waiting time ... was shorter'. If the PMT temperature lags the sensor by 1–2°C, the steepest coefficient (bare NNVT, ~12%/°C or ~6 kHz/°C) would be biased by roughly 10–20%, and part of the heating/cooling asymmetry in Table 2 could be an artefact of the lag. Since §3.4's ±1°C recommendation is derived from this coefficient, the assumption is load-bearing. Please add a thermal-equilibration check (e.g., a dedicated measurement with a temperature sensor attached to a PMT, or a quantitative thermal model with estimated lag) or explicitly report the coefficients as effective sensitivities with the associated systematic uncertainty.","section":"3.2.1, Figs. 6–8, Tables 2–3"},{"comment":"The headline numbers—the asymptotic DCR fractions (c = 0.56 and 0.83), the stabilization times (25 h and 52 h), and the temperature coefficients (1.4–12.6%/°C)—are quoted without uncertainties, sample sizes, or goodness-of-fit statistics. Table 1 parameters are fit outputs, but no error bars or fit quality (e.g., χ²/ndf or residual plots) are given, and the extrapolation to an 'ideal DCR' is not validated against an independent data set. Please report the number of PMTs contributing to each average, the statistical uncertainty on each coefficient, and a measure of fit quality for Eq. (3.1).","section":"Tables 1–3 and Eq. (3.1)"},{"comment":"The recommendation in §3.4 that 'a minimum of 7 hours is sufficient for HPK PMTs' appears inconsistent with the 25-hour stabilization time reported in §3.1 and Table 1 (τ2 ≈ 25 h). Please clarify the criterion used for 'sufficient' (e.g., time to reach a specified DCR drift threshold) and reconcile the two statements, or the reader cannot tell which operational guidance to follow.","section":"3.1 and 3.4"}],"minor_comments":[{"comment":"The coverage figure is given as 'approximately 75%' in the abstract and 'exceed 78% (75% with LPMTs alone)' in §1; please make the numbers consistent.","section":"Abstract and §1"},{"comment":"'heeting' should be 'heating' in the caption.","section":"Figure 7 caption"},{"comment":"'In the additional to the 12 hours DCR measurement...' is ungrammatical; please rephrase.","section":"§3.1"},{"comment":"The text says 117 PMTs were monitored, while §5 says 'around 110 PMTs'; please use the exact number consistently.","section":"§3.3.1"},{"comment":"Please define how the 'variation ratio' (max. and min.) is computed and how the %/°C coefficient is extracted (e.g., linear regression over the full range, or average of point-to-point slopes).","section":"Tables 2 and 3"},{"comment":"The statement that holding each set point for 2–3 h 'allowing the PMTs to achieve thermal equilibrium' is an assertion; consider softening to 'assumed to allow' or add a reference to a thermal-equilibration study.","section":"§3.2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript fits JINST's scope. The main technical concern is the thermal-equilibrium assumption behind the temperature coefficients; this is fixable with additional measurements or a careful systematic-error discussion. I would not recommend rejection based on the current draft, but the quantitative claims need to be either supported or downgraded."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"If you work with large PMTs, this is worth a look: it gives the first quantitative picture of how JUNO's 20-inch MCP and dynode PMTs cool down and respond to temperature in mass testing. The two-exponential cooling fit and the bare-vs-potted temperature coefficients are new, and the spike statistics give a concrete sense of how many tubes misbehave. The camera image of a candidate flasher at +500V is a nice piece of evidence, even if it's a single event.\n\nWhat the paper does well: the qualitative message is consistent across short- and long-term data. NNVT MCP-PMTs cool more slowly (about 52 h to stabilize vs 25 h for HPK), respond more strongly to temperature (roughly 12%/C bare, 4%/C potted, vs about 2%/C for HPK), and are more likely to show DCR spikes. The monitoring system they built is sensible, and the operational recommendations (at least 50 h cooling for NNVT, ±1 C temperature control) follow logically from the data if the measurements are right.\n\nThe soft spots are real but not fatal. The biggest one is that the temperature coefficients in Tables 2 and 3 are derived from sensors mounted inside the drawers, not on the PMTs themselves. The text says the PMTs were held 2-3 h at each set point for equilibrium, but it also admits that the NNVT tubes needed more time at the highest temperature and that the waiting time was shorter in the 14-21 C range. If the PMT lags the drawer sensor, the bare-NNVT 12%/C slope—the steepest and most consequential number—could be biased by 10-20%. That's worth fixing or at least quantifying.\n\nSecond, the fit parameters in Tables 1-3 have no uncertainties. The two-exponential form is plausible, but without errors on a1, tau1, etc., it's hard to know whether the NNVT/HPK differences are significant. Third, the bare vs potted comparison in Fig. 5 may involve different PMT samples and different readout thresholds (0.3 p.e. vs 0.25 p.e.); the text doesn't explicitly clarify. Fourth, the 20% and 7% spike rates need sample sizes and confidence intervals.\n\nNone of this undercuts the central story. The qualitative picture is coherent, and the paper is honest about its limitations. For a JUNO audience, these numbers are directly useful. I'd send it to a serious referee, with a request for uncertainties and a clearer statement about the temperature measurement premise.","headline":"A useful engineering characterization of JUNO PMT dark rates; the qualitative NNVT/HPK difference holds up, but the headline temperature coefficients rest on an unverified thermal-equilibrium assumption.","tokens_in":15083,"tokens_out":2144,"would_cite":true,"duration_ms":20668,"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":"The paper establishes that JUNO's NNVT PMTs keep cooling for about 52 h, with a settled dark count 44% below the 12-hour value, while HPK tubes settle near 83% in about 25 h, and that the two types have different dark-count temperature…","keywords":["20-inch photomultiplier tubes","dark count rate","JUNO","cooling time","temperature dependence","PMT stability","flasher events","mass testing"],"falsifier":"Repeat the 14–28°C step cycle on a spare bare NNVT PMT with a thermocouple attached directly to the photocathode glass and to the microchannel-plate package; if the slope of DCR versus measured tube temperature differs materially from the slope computed against the drawer sensor, the reported 12%/°C bare-NNVT coefficient is an artifact of thermal lag.","tokens_in":2113,"feed_emoji":"🔬","tokens_out":2164,"duration_ms":123133,"temperature":0.7,"pith_summary":"This paper establishes how the dark count rate (random pulses with no light) of two types of 20-inch photomultiplier tubes used in JUNO depends on cooldown time, temperature, and long-term operation. Using data from the JUNO PMT mass-testing containers, it finds that NNVT tubes are slower to settle and much more temperature-sensitive than HPK tubes: after the standard 12-hour cooling period, the NNVT dark count keeps falling to about 56% of the measured value over about 52 hours, while HPK reaches about 83% in about 25 hours. It also reports that roughly 20% of tested tubes show at least one dark-count spike during cooling, about 7% of 117 long-term monitored tubes are unstable, and some spikes are consistent with flasher events, including one MCP flasher imaged at elevated voltage. If these numbers hold, JUNO's acceptance procedure and dark-count noise budget should treat the 12-hour reading as provisional rather than the settled value.","feed_headline":"NNVT PMT dark count settles 44% lower after 52 h","feed_subtitle":"Mass-test data quantify cooling and temperature drift in the PMT noise that feeds JUNO's energy measurements.","key_machinery":"The central object is the normalized DCR-versus-time curve and the two-exponential cooling fit $y = (a_1/\\tau_1)e^{-x/\\tau_1} + (a_2/\\tau_2)e^{-x/\\tau_2} + c$, where the constant $c$ gives the asymptotic fraction of the 12-hour DCR (0.56 for NNVT, 0.83 for HPK), the fast component $\\tau_1\\approx 3$–4 h captures the light-exposure and thermal transient, and the slow component $\\tau_2$ gives the stabilization time (about 25 h for HPK, 52 h for NNVT). The temperature coefficients are carried by normalized ratios $\\mathrm{DCR}(T)/\\mathrm{DCR}(21^\\circ\\mathrm{C})$ measured during HVAC step cycles, with the temperature read by sensors mounted inside the drawer beside each PMT and a 2–3 hour wait at each set point to approach thermal equilibrium.","core_discovery":"At an operating gain of $1\\times10^7$, the central claim is that the dark count rate of JUNO's 20-inch PMTs is not fully characterized by the standard 12-hour acceptance measurement. For potted NNVT tubes the DCR continues to decline after 12 hours, following a two-exponential decay that asymptotes at $c=0.56$ times the 12-hour value with a slow time constant near 52 hours; potted HPK tubes asymptote at $c=0.83$ with a slow time constant near 25 hours. The temperature dependence is also type-specific: bare NNVT tubes change by about 12%/°C (about 6 kHz/°C), potted NNVT by about 4%/°C (about 0.8–0.9 kHz/°C), and HPK by about 2%/°C (about 0.2–0.4 kHz/°C) over the 14–28°C range, with long-term room-temperature monitoring consistent with the short-term step tests. DCR monitoring further shows that about 20% of tubes have at least one spike above 50 kHz or 50% during cooling, about 7% of 117 long-term monitored tubes are unstable, and at least one NNVT flasher originating from the microchannel plate was imaged when the tube was operated 500 V above its nominal voltage.","pith_inferences":["If the asymptotic DCR fractions hold in the final detector environment, JUNO's in-situ NNVT noise floor after months of operation could be well below the acceptance-test value; early detector data would then show a slowly declining background as tubes settle.","The drop in NNVT temperature coefficient after potting (from about 12%/°C to about 4%/°C) suggests the bare-tube sensitivity may be dominated by the exposed glass envelope or divider rather than the MCP or photocathode itself; a test that heats only the envelope or only the photocathode would separate these contributions.","Because the spike search used 30-minute sampling with a >50 kHz or >50% threshold, continuous fast sampling would likely reveal shorter flashers that are averaged out, and correlating spikes with HVAC, HV, and drawer-swap timing would test whether many reported spikes are operational transients rather than intrinsic PMT behavior.","The +500 V flasher image suggests an accelerated screening strategy: briefly run NNVT tubes above nominal voltage in a dark box with coincidence or camera monitoring, and reject tubes showing MCP flashers before potting."],"forward_implications":["For NNVT tubes, a DCR quoted after the standard 12-hour cooldown overstates the settled dark count: the asymptotic rate is 44% lower than the 12-hour reading, so acceptance thresholds based on 12-hour values are conservative relative to steady-state operation.","DCR measurements with uncertainty below 1 kHz require a longer cooling period for NNVT tubes (about 50 hours) than for HPK tubes (about 7 hours).","With the measured temperature coefficients, controlling the detector temperature to about ±1°C keeps the DCR-induced contribution below 1 kHz across the studied temperature range.","The observed spike rate (about 20% during cooling) and long-term instability rate (about 7%) imply that acceptance testing should include a stability-monitoring window and a drawer or HV-divider swap step to separate electronics issues from intrinsic PMT problems.","Coincidence-rate analysis among neighboring PMTs and single-photon camera imaging can identify flasher candidates, as demonstrated for the NNVT tube imaged 500 V above nominal high voltage."],"supporting_citations":[{"why":"Supplies the accepted mean DCR values (15.3 kHz HPK, 49.3 kHz NNVT) and the standard 12-hour cooling protocol that this paper quantifies and extends.","marker":"[4]"},{"why":"Describes the container-based mass-testing facility whose HVAC-controlled drawers generated the cooling, temperature, and stability data.","marker":"[15]"},{"why":"Gives the 1F3 electronics threshold (2±0.1 mV, about 0.25 p.e.) used for the potted-PMT DCR measurements.","marker":"[19]"},{"why":"Provides the HPK 20-inch dynode tube performance baseline, including its dark count below 20 kHz, for comparison.","marker":"[12]"},{"why":"Characterizes a 20-inch microchannel-plate PMT and supplies the NNVT-type baseline for dark count and gain behavior.","marker":"[13]"},{"why":"Documents the R&D program for the 20-inch MCP-PMTs used as NNVT tubes in JUNO.","marker":"[14]"},{"why":"Studies flasher signals from the HV divider of 20-inch PMTs and is used to interpret the observed DCR spikes as possible flashers.","marker":"[21]"},{"why":"Studies the 20-inch PMT flasher signal and coincidence signatures that inform the flasher identification method.","marker":"[22]"},{"why":"Provides prior stability results for 20-inch microchannel-plate PMTs, serving as the baseline for the long-term DCR stability monitoring.","marker":"[33]"}],"fun_headline_variants":["Dark count keeps falling: JUNO PMT test reveals 52-h drift","20% of JUNO PMTs show dark count spikes","PMT dark count sensitivity: JUNO finds up to 12%/°C","Long-term drift in JUNO PMTs: 7% unstable","JUNO PMT flasher imaged at 500 V overdrive"],"cache_read_input_tokens":17024,"weakest_assumption_plain":"The temperature coefficients assume that the air-temperature sensor mounted beside each PMT in the drawer tracks the temperature that actually controls the dark count, and that waiting two to three hours at each set point brings the PMT itself into thermal equilibrium.","fun_headline_variants_meta":{"raw":{"variants":["Dark count keeps falling: JUNO PMT test reveals 52-h drift","20% of JUNO PMTs show dark count spikes","PMT dark count sensitivity: JUNO finds up to 12%/°C","Long-term drift in JUNO PMTs: 7% unstable","JUNO PMT flasher imaged at 500 V overdrive"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000464,"raw_usage":{"total_tokens":2436,"prompt_tokens":1183,"completion_tokens":1253,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":799,"completion_tokens_details":{"reasoning_tokens":1152}},"tokens_in":799,"tokens_out":1253,"duration_ms":12090,"temperature":1.0,"reasoning_tokens":1152,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T19:42:21.069554+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeat the 14–28°C step cycle on a spare bare NNVT PMT with a thermocouple attached directly to the photocathode glass and to the microchannel-plate package; if the slope of DCR versus measured tube temperature differs materially from the slope computed against the drawer sensor, the reported 12%/°C bare-NNVT coefficient is an artifact of thermal lag.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the accepted mean DCR values (15.3 kHz HPK, 49.3 kHz NNVT) and the standard 12-hour cooling protocol that this paper quantifies and extends."},{"cited_title":"Wonsak, A","cited_arxiv_id":null,"evidence_quote":"Describes the container-based mass-testing facility whose HVAC-controlled drawers generated the cooling, temperature, and stability data."},{"cited_title":"Check on the features of potted 20-inch PMTs with 1F3 electronics prototype at Pan-Asia","cited_arxiv_id":"2208.08264","evidence_quote":"Gives the 1F3 electronics threshold (2±0.1 mV, about 0.25 p.e.) used for the potted-PMT DCR measurements."},{"cited_title":"Maekawa, C","cited_arxiv_id":null,"evidence_quote":"Provides the HPK 20-inch dynode tube performance baseline, including its dark count below 20 kHz, for comparison."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Characterizes a 20-inch microchannel-plate PMT and supplies the NNVT-type baseline for dark count and gain behavior."},{"cited_title":"Chang et al.,The r&d of the 20in","cited_arxiv_id":null,"evidence_quote":"Documents the R&D program for the 20-inch MCP-PMTs used as NNVT tubes in JUNO."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Studies the 20-inch PMT flasher signal and coincidence signatures that inform the flasher identification method."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides prior stability results for 20-inch microchannel-plate PMTs, serving as the baseline for the long-term DCR stability monitoring."}],"review_version":2}