{"id":"af5355ab-b684-4daa-a7fb-d60fec4a72aa","arxiv_id":"2501.09988","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":19,"one_line_summary":"The JUNO liquid scintillator's emission spectrum, absorption length, alpha/beta fluorescence time profiles, and resulting pulse-shape discrimination figures are measured and reported.","lead":"This paper reports laboratory measurements of the light emission, absorption, and fluorescence timing of the liquid scintillator mixture built for the JUNO neutrino detector. These numbers feed the detector simulations JUNO needs for energy and position reconstruction and for separating alpha and beta particle signals.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Finite 1600 ns window leaves ~9% of the slow component unmeasured; the Eq. 5.2 tail correction assumes no slower component, so Table 2's tau4/q4 and the PSD claim may be window-dependent.","rationale":"The paper is careful and internally consistent: the four-exponential model is checked against three- and five-exponential alternatives, the IRF is fixed from a dedicated measurement, the Cherenkov crosscheck is a real independent check, and the PSD caveats are stated. I do not question the data or the fit procedure. My concern targets the one place where the analysis depends on an unverifiable assumption about time scales beyond the measured window: the 1600 ns truncation. Equation 5.2 explicitly relies on the fourth component being a pure exponential beyond the window, and the five-exponential test cannot see a component whose lifetime is much longer than the window. This is not an external-consensus disagreement but an internal completeness issue. It is more specific than the reader's general 'missing systematics' point and, for the slow-tail parameters, more load-bearing than the IRF assumption; the IRF issue would also need systematic treatment, but the proposed finite-window test directly probes whether Table 2 is stable. The reader's CONDITIONAL verdict is appropriate, and my concern sharpens the condition without moving the verdict.","tokens_in":11549,"tokens_out":9635,"duration_ms":102794,"concrete_test":"Take a second dataset with the same 244Cm and 60Co sources and setup but extend the DAQ window to at least 5000 ns; refit with the 4-exponential model and with a 5-exponential model starting tau5 near 2 microseconds. If tau_i/q_i shift by more than the Table 2 statistical errors, or if the 5th component has nonzero weight, Table 2 is window-dependent and the claim needs revision. Also run a toy-MC injection of a 0.5-1% component with tau5 = 2 microseconds through the 1600 ns analysis to check whether the 4-exponential fit recovers the injected tau_i/q_i; if the fit is biased by more than the quoted errors, the 'railed to zero' observation cannot exclude a slower component.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central deliverable is Table 2, but the parameters are extracted from a 1600 ns acquisition window while the slowest fitted component has tau4 = 598 ns, so about 9.1% of that component lies beyond the window. Equation 5.2 corrects the normalization by assuming the fourth exponential continues unchanged to infinity and that no additional slower component exists. The paper's check that a fifth exponential 'railed to zero' is not a discriminating test: a component with tau of order 2 microseconds and sub-percent weight would be strongly suppressed inside the window and would be absorbed into tau4 and q4 (and, through the fit, into tau3 and q3) without being visibly preferred as a fifth exponential. The quoted errors in Table 2 are statistical only, so this missing-tail systematic is unquantified. It matters for the headline decay parameters and directly for the Sec. 5.4 PSD tail-to-total claim, since the alpha/beta separation depends on the slow-tail fractions. The IRF concern in the reader's report is real, but the Cherenkov crosscheck mainly validates the main-peak width, not the detailed afterpulse structure; the finite-window tail assumption is the more load-bearing unresolved systematic.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports optical characterization of the JUNO liquid scintillator (LAB + 2.5 g/L PPO + 3 mg/L Bis-MSB): emission spectrum from a spectrofluorimeter, absorption length from a spectrophotometer, and fluorescence time profiles measured with a custom TCSPC setup for alpha (244Cm) and beta (60Co) excitation. The timing analysis models the decay as a sum of four exponentials with a common rise time (Eq. 5.1), convolved with an instrument response function measured with a 405 nm pulsed laser and represented by seven Gaussians. The best-fit parameters are given in Table 2, with tau1 about 4 ns and tau4 about 570-600 ns; alpha excitation produces a slower tail than beta. The authors use the measured profiles in a toy Monte Carlo to estimate alpha/beta discrimination via a tail-to-total ratio, claiming 99.5% alpha rejection at 99% beta acceptance for 1 MeV events.","tokens_in":11915,"tokens_out":10735,"duration_ms":96965,"significance":"If correct, Table 2 provides the actual scintillation timing parameters for the JUNO liquid scintillator, which are needed for position/energy reconstruction and alpha/beta discrimination in JUNO. The paper has notable strengths: the IRF is measured and fixed before the fluorescence fit; the four-exponential model is compared with three- and five-exponential alternatives; the Cherenkov emission is used as a crosscheck of the IRF main-peak width; and the PSD estimate is a clearly described toy-MC realization of the measured profiles rather than a circular claim. The main deliverable is a parameter table with sub-nanosecond precision on the fast component, which is valuable for the JUNO Monte Carlo.","major_comments":[{"comment":"The tail correction assumes that the fourth exponential is the slowest component and that it continues unchanged beyond t_w = 1600 ns. Since tau4 is about 600 ns, roughly 9% of that component lies outside the window; any additional slower component with tau of order microseconds and sub-percent weight would be almost linear over the window and would be absorbed into tau4 and q4 (and, through the fit, into tau3 and q3). The five-exponential test reported in the same section ('railed to zero') is not a discriminating test for such a component because its in-window contribution is small. The errors in Table 2 are statistical only, so this bias is unquantified, and it propagates directly to the Sec. 5.4 PSD tail-to-total prediction. Please quantify this systematic, e.g., by fitting with a fifth exponential with a prior, by generating pseudo-data with a slow component and refitting with the four-exponential model, or by extending the acquisition window.","section":"Sec. 5.3, Eq. (5.2), Table 2"},{"comment":"The IRF is measured at 405 nm with pure LAB and the IRF parameters are fixed during the fluorescence fits. The Cherenkov crosscheck validates only the FWHM of the main peak; it does not validate the afterpulse structure (e.g., the bump at about 103 ns), which lies in a region where the scintillation profile changes slope and therefore has a strong influence on the fitted tau_i and q_i. The statement that the IRF does not change with wavelength is supported only by a Gaussian FWHM comparison at >530 nm, not by a full-shape comparison. Please either measure the full IRF shape at representative wavelengths or perform a sensitivity study in which the IRF parameters are varied within plausible ranges and the resulting changes in Table 2 are reported as a systematic uncertainty.","section":"Sec. 5.1.1, Figs. 7-8, Table 2"},{"comment":"The notation in Eq. (5.2) is ambiguous about whether the q_i entering the right-hand side are the fitted (in-window) weights or the final corrected weights. If the q_i (i=1..3) are the final fractions that sum to 1 with q4, then the formula gives tilde q4 = q4/[1-exp(-(t_w-t0)/tau4)], which is larger than q4 and is inconsistent with Table 2. If instead they are the pre-correction fitted values, the text should say so explicitly and give the derivation. Please clarify what quantity is reported in Table 2 and how Eq. (5.2) is used in the fit.","section":"Sec. 5.3, Eq. (5.2)"}],"minor_comments":[{"comment":"The sentence 'By using a toy Monte-Carlo we also verified that the seven Gaussian model ensures a description of the fluorescence time distributions with an accuracy on the order of ~ %.' contains an incomplete number; it should read '~1%' or give the actual value.","section":"Sec. 5.1.1"},{"comment":"The statement that the four-exponential model is 'the best choice' would be easier to evaluate if the chi2 or goodness-of-fit values for the three-, four-, and five-exponential fits were reported.","section":"Sec. 5.3"},{"comment":"The convolution is written with N_i and N_j, but these quantities are not defined; please define the normalization factors or use a clearer notation.","section":"Eq. (5.4)"},{"comment":"Please state explicitly in the caption that the quoted uncertainties are statistical only, and consider adding a systematic error row or a systematic budget in the text.","section":"Table 2"},{"comment":"The phrase 'Figure 4 shows the error-weighted mean trend with 5σ error bar' is unusual; please clarify whether the error bars are 5σ or whether the text means something else.","section":"Sec. 4"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of JINST and the measurement is useful for JUNO. The main concern is that the headline precision (sub-1% on tau1) is not accompanied by a systematic error budget; the finite-window tail and IRF-transfer issues need to be addressed before acceptance. The authors should also clarify the normalization in Eq. (5.2)."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Plain take: this is a careful, useful measurement paper for JUNO's MC, and the main numbers are credible. The missing piece is quantified systematic error, especially around the finite-window tail.\n\nThe genuinely new content: fluorescence time constants and weights for alpha and beta excitation on the actual JUNO purification-plant scintillator (LAB + 2.5 g/L PPO + 3 mg/L Bis-MSB), with a common rise time, plus the merit-factor curve for a simple tail-to-total PSD. The emission spectrum and absorption length for this sample are also measured. The analysis is methodologically solid: the IRF is measured with a 405 nm laser and fixed before the fluorescence fit, the four-exponential model is checked against three and five exponentials, and the Cherenkov crosscheck supports the IRF width estimate. The reproducibility runs give compatible results.\n\nWhere it's soft: Table 2 quotes only statistical errors. The systematic uncertainty from the IRF model, the filter choices, and the fit range is not propagated, so the claimed ~1% accuracy on tau_1 and q_1 is probably optimistic. More importantly, the 1600 ns window leaves about 9% of the slowest component unmeasured, and Eq. 5.2's correction assumes that this single exponential continues to infinity. The check that a fifth exponential 'railed to zero' does not rule out a sub-percent slower component; a ~2 microsecond component would be absorbed into tau4/q4. That could bias the PSD tail-to-total claim in Sec. 5.4. This doesn't sink the paper, but a referee should ask for a quantified estimate or an extended-window measurement. The IRF concern about wavelength dependence is real, but the Cherenkov measurement at several wavelengths partially covers it; it is the less severe issue.\n\nWho this is for: anyone building the JUNO detector simulation, and the broader liquid-scintillator timing community. It deserves a serious referee. My recommendation: send it to review with a request for systematics and a discussion of the tail assumption. I would cite the table values in my own MC work.","headline":"Credible new timing parameters for JUNO's scintillator, but the unquantified finite-window tail correction is the main thing a referee should push on.","tokens_in":12797,"tokens_out":2390,"would_cite":true,"duration_ms":22788,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The fluorescence of the JUNO liquid scintillator is described by four exponential components with a common rise time, and the measured alpha/beta timing difference supports 99.5% alpha rejection at 99% beta acceptance.","keywords":["liquid scintillator","JUNO","fluorescence decay time","pulse shape discrimination","linear alkyl benzene","PPO","Bis-MSB","time-correlated single photon counting"],"falsifier":"Measure the instrument response function at several wavelengths inside the emission band (for example 430 nm and 480 nm) and with the full JUNO scintillator in the cuvette instead of pure LAB, then refit the four-exponential model with each IRF; if the fitted $\\tau_i$ move by more than the quoted statistical errors (about 1% for the fast components) or the weights $q_3+q_4$ change by more than a few percent, the fixed seven-Gaussian IRF is not representative and the reported parameters are biased.","tokens_in":11365,"feed_emoji":"⚛️","tokens_out":12848,"duration_ms":112599,"temperature":0.7,"pith_summary":"JUNO, a 20-kiloton liquid-scintillator neutrino detector, needs precise knowledge of how its scintillator emits light to reconstruct event energies and positions and to reject backgrounds. This paper measures the emission spectrum, absorption length, and fluorescence time profiles of the actual JUNO scintillator mixture (LAB with 2.5 g/L PPO and 3 mg/L Bis-MSB) using a dedicated single-photon counting setup on a sample from the JUNO purification plants. The central result is a four-exponential description of the fluorescence time profile with a common rise time, giving distinct parameters for alpha and beta excitations, with statistical precision near 1% for the fastest components. If these numbers are right, they are the timing inputs JUNO's simulation and pulse-shape discrimination analysis will rely on, and they imply that a simple tail-to-total cut can reject 99.5% of alpha backgrounds while retaining more than 99% of beta signals at about 1 MeV.","feed_headline":"JUNO scintillator timing enables 99.5% alpha rejection","feed_subtitle":"Four-exponential fit separates alpha and beta signals, anchoring JUNO's background rejection.","key_machinery":"The central object is the four-exponential fluorescence model $F_{\\mathrm{fit}}(t) = N \\sum_{i=1}^{4} \\frac{q_i}{\\tau_i - \\tau_r}\\left(e^{-(t-t_0)/\\tau_i} - e^{-(t-t_0)/\\tau_r}\\right)\\Theta(t-t_0)$, where $\\tau_r$ is a common rise time, $\\tau_i$ and $q_i$ are the decay times and relative weights of the four components, and $\\Theta$ is the step function marking the start of the pulse. The model is convolved analytically with the instrument response function (the system's time response to an instantaneous flash), which is itself represented as a sum of seven Gaussians fitted to a 405-nm laser pulse scattered in pure LAB; a finite-window correction accounts for the 1600-ns acquisition truncating the slowest component. This machinery separates the fast nanosecond component from the slow delayed-fluorescence tail and attributes the $\\alpha$/$\\beta$ difference to the increased weight of the slower components for $\\alpha$ excitation.","core_discovery":"On its own terms, the paper establishes that the JUNO liquid scintillator's fluorescence time distribution is well fitted by a sum of four exponentials sharing one rise time, after convolution with an instrument response function modeled as seven Gaussians. For $\\alpha$ excitation the decay times are $\\tau = (4.06, 17.29, 91.3, 598)\\,\\mathrm{ns}$ with weights $(54.0, 23.7, 14.1, 8.9)\\%$; for $\\beta$ excitation they are $\\tau = (3.86, 14.52, 81.3, 570)\\,\\mathrm{ns}$ with weights $(67.8, 20.0, 7.5, 5.0)\\%$. The slower components carry a larger weight for alphas, which is the physical handle for pulse-shape discrimination. Using these profiles in a toy Monte Carlo at 1500 photoelectrons (about 1 MeV), a tail-to-total cut achieves a separation merit factor $D = 3.7$, corresponding to 99.5% $\\alpha$ rejection at more than 99% $\\beta$ acceptance.","pith_inferences":["Because the rise time is a single effective parameter common to all four components, a natural testable extension is to measure the time profile through narrow-band emission filters; if the energy transfer from LAB to PPO to Bis-MSB is wavelength-dependent, the fitted rise time and fast decay time should shift with wavelength, which the paper's Cherenkov crosscheck does not probe.","Interpolating between the measured alpha and beta parameter sets could yield a continuous $\\mathrm{d}E/\\mathrm{d}x$-dependent timing parameterization for other particles (protons, heavier ions), which JUNO background simulations would need but the paper does not provide.","The small-sample PSD estimate omits detector-scale light propagation, absorption, re-emission, and scattering; if those effects smear the time profile, the full-detector alpha rejection at fixed beta acceptance could be lower than 99.5%, with JUNO's roughly threefold higher photoelectron statistics serving as the compensating factor — a trade-off the paper leaves for the full Monte Carlo."],"forward_implications":["JUNO's full Monte Carlo can adopt these $\\tau_i$ and $q_i$ as fixed scintillation timing parameters, making simulated events match the measured single-photon time structure.","An alpha/beta separation merit factor of $D=3.7$ at 1500 photoelectrons means a simple tail-to-total cut already reaches 99.5% alpha rejection at more than 99% beta acceptance for roughly 1 MeV events in the small-sample geometry.","Because a pseudocumene-based scintillator measured in the same setup separates alpha and beta better, while JUNO collects about three times more photoelectrons per unit energy, the two detectors should end up with comparable alpha/beta discrimination at similar energies.","The measured emission spectrum peaking at 400 and 420 nm matches the quantum efficiency of JUNO's large PMTs, supporting the light-collection assumptions behind energy-resolution predictions.","The absorption length is inferred to be at least 18 m at wavelengths above 430 nm, so attenuation at those wavelengths will not dominate light loss in the 20-kton detector."],"supporting_citations":[{"why":"Supplies the time-correlated single-photon counting method and the four-exponential decay model for LAB-based scintillators, plus the Borexino comparison data.","marker":"[20]"},{"why":"Documents the purification plants that produced the liquid scintillator sample, grounding the measurement in the actual JUNO batch.","marker":"[11]"},{"why":"Defines the optimized pulse-shape discrimination techniques and the Borexino alpha/beta context against which the tail-to-total estimate is assessed.","marker":"[10]"},{"why":"Provides the quantum-efficiency curve of the JUNO large PMTs used to verify the emission-spectrum match.","marker":"[21]"},{"why":"Reports the precise attenuation length of the JUNO scintillator used to corroborate the at-least-18 m extrapolation at wavelengths above 430 nm.","marker":"[23]"}],"fun_headline_variants":["Four-exponential fit separates alpha and beta in JUNO scintillator","JUNO scintillator fluorescence timing enables 99.5% alpha rejection","Alpha-beta pulse-shape discrimination in JUNO via four-exponential decay","JUNO scintillator four-exponential decay yields D=3.7 alpha separation","JUNO scintillator timing separates alpha from beta"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The instrument response function, measured with a 405-nanometer laser scattering off a diffuser in pure LAB and held fixed during the fluorescence fit, is assumed to be identical for the real scintillation wavelengths and for the entire 1600-nanosecond window; if the response varies with wavelength or with the full scintillator mixture, the reported decay times and weights shift beyond the quoted errors.","fun_headline_variants_meta":{"raw":{"variants":["Four-exponential fit separates alpha and beta in JUNO scintillator","JUNO scintillator fluorescence timing enables 99.5% alpha rejection","Alpha-beta pulse-shape discrimination in JUNO via four-exponential decay","JUNO scintillator four-exponential decay yields D=3.7 alpha separation","JUNO scintillator timing separates alpha from beta"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00042,"raw_usage":{"total_tokens":2151,"prompt_tokens":927,"completion_tokens":1224,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":543,"completion_tokens_details":{"reasoning_tokens":1127}},"tokens_in":543,"tokens_out":1224,"duration_ms":8760,"temperature":1.0,"reasoning_tokens":1127,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T19:27:37.349332+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the instrument response function at several wavelengths inside the emission band (for example 430 nm and 480 nm) and with the full JUNO scintillator in the cuvette instead of pure LAB, then refit the four-exponential model with each IRF; if the fitted $\\tau_i$ move by more than the quoted statistical errors (about 1% for the fast components) or the weights $q_3+q_4$ change by more than a few percent, the fixed seven-Gaussian IRF is not representative and the reported parameters are biased.","supporting_citations":[{"cited_title":"Lombardi, F","cited_arxiv_id":null,"evidence_quote":"Supplies the time-correlated single-photon counting method and the four-exponential decay model for LAB-based scintillators, plus the Borexino comparison data."},{"cited_title":"Landini, et al., Distillation and gas stripping purification plants for the JUNO liquid scintillator, NIM A, 1069 (2024)","cited_arxiv_id":null,"evidence_quote":"Documents the purification plants that produced the liquid scintillator sample, grounding the measurement in the actual JUNO batch."},{"cited_title":"Basilico et al., Optimized𝛼/𝛽 pulse shape discrimination in Borexino, Phys","cited_arxiv_id":null,"evidence_quote":"Defines the optimized pulse-shape discrimination techniques and the Borexino alpha/beta context against which the tail-to-total estimate is assessed."},{"cited_title":"Abusleme et al., Prediction of Energy Resolution in the JUNO Experiment, Chinese Physics C 49, (1) (2025)","cited_arxiv_id":null,"evidence_quote":"Provides the quantum-efficiency curve of the JUNO large PMTs used to verify the emission-spectrum match."},{"cited_title":"Yin et al., Precise measurement of attenuation length of the juno liquid scintillator, Radiation Detection Technol- ogy and Methods, 4 (3) (2020)","cited_arxiv_id":null,"evidence_quote":"Reports the precise attenuation length of the JUNO scintillator used to corroborate the at-least-18 m extrapolation at wavelengths above 430 nm."}],"review_version":1}