{"id":"5075c178-b859-46c3-877c-ebb87593228b","arxiv_id":"2501.16434","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"After heavy-ion irradiation, CsI shows afterglow pulses lasting 5-9 times longer than NaI, so the BTO team selects NaI for its low-Earth-orbit transient detector.","lead":"NaI and CsI scintillator crystals keep emitting small light pulses after being hit by high-energy particles, and the effect is several times longer in CsI. This measurement, done for the BTO gamma-ray detector on NASA's COSI mission, led the team to choose NaI to avoid wasted observing time.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Afterglow rate estimate uses incident kinetic energy thresholds rather than deposited energy, which can misstate the 0.02 s vs 9 s lost-time comparison.","rationale":"After a good-faith reading, the measured afterglow durations (Table 2) are the paper's strongest evidence: the analysis is careful (gain correction with Ba-133, more than 50 waveforms per setup, explicit return-to-baseline criterion), and the qualitative result that CsI afterglow is longer and stronger than NaI is well supported. The vulnerability is in the translation from these beam measurements to orbit losses. The paper's own deposited-energy check for CsI (74.5 MeV) shows the authors are aware that deposited, not incident, energy matters, but the rate calculation in §3.1 uses incident kinetic energy thresholds. This is a real methodological gap: in the space radiation environment, the hadron spectrum is dominated by protons, whose energy deposition per centimeter is far lower than that of the He and C ions used at HIMAC; a proton above 920 MeV can deposit far less than 275 MeV in the small BTO crystal, so the NaI afterglow event rate of 0.014/s is likely overestimated. Conversely, the CsI threshold at 164 MeV incident energy corresponds to a deposited energy near the Rau et al. (2005) threshold, so that rate may be roughly right. The net effect is that the absolute lost-time numbers and the ratio could shift, though the qualitative preference for NaI is probably robust because CsI's longer afterglow and lower threshold make it very unlikely to lose less time. The reader flagged the threshold premise; the present concern sharpens it to the incident-versus-deposited energy mismatch. A single MEGAlib re-analysis using deposited-energy thresholds would settle whether the 0.02 s vs 9 s figures stand. Because the central qualitative conclusion is secure, the conditional verdict (with a request to redo the rate estimate on a deposited-energy basis) remains appropriate.","tokens_in":13944,"tokens_out":12750,"duration_ms":103041,"concrete_test":"Re-run the MEGAlib background simulation for the BTO orbit, tallying the energy deposited in the BTO crystal by each simulated hadron. Count events whose deposited energy exceeds the afterglow thresholds expressed in deposited energy: for CsI use the paper's 74.5 MeV value; for NaI derive a threshold from the He/NaI non-detection (e.g., bracket 100–300 MeV, since 275 MeV deposited by He produced negligible afterglow). Compare the resulting per-orbit lost time for NaI and CsI against the paper's 0.02 s and 9 s. If the CsI/NaI lost-time ratio changes by more than a factor of ~2, or if either absolute rate changes by more than an order of magnitude, the quantitative trade-study claim is not robust and the conclusion would rest only on the directly measured afterglow durations.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The quantitative case for NaI rests on the afterglow event rates in §3.1, where thresholds are set in terms of incident particle energy (920 MeV for NaI, 164 MeV for CsI) and the MEGAlib hadronic flux is integrated above these values. But the HIMAC measurements show afterglow scales with energy deposited in the crystal: the 920 MeV He beam deposits only 275 MeV in NaI and 325 MeV in CsI, and the C beam deposits 1.65/2 GeV; the afterglow durations in Table 2 are tied to these deposited energies. A 920 MeV proton traversing a 3.8 cm crystal deposits an order of magnitude less energy than a 920 MeV He, so counting all hadrons above 920 MeV likely overestimates the NaI afterglow rate; conversely, a 164 MeV proton may or may not deposit enough energy to exceed the CsI afterglow threshold. The authors themselves translate the CsI threshold to a deposited energy of 74.5 MeV (validated against Rau et al. 2005), yet do not use deposited-energy thresholds in the rate calculation. Thus the claimed ~70 s vs ~1.4 s event spacings and the 0.02 s vs 9 s lost-time ratio are not securely grounded. An internal inconsistency underlines the fragility: §3.1 quotes 0.7 particles/s (1.4 s spacing), while §3.2 quotes 0.7 s spacing for CsI.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports measurements of afterglow in NaI(Tl) and CsI(Tl) scintillator modules with SiPM readout, irradiated with 230 MeV/u He and 350 MeV/u C beams at HIMAC. The authors measure afterglow durations of 2.40 ± 0.5 ms (CsI, C) and 0.28 ± 0.07 ms (NaI, C), with corresponding He values of 0.9 ± 0.2 ms and 0.16 ± 0.04 ms, and use these to argue that CsI afterglow is 8.6× and 5.6× longer than NaI for C and He beams, respectively. They then use MEGAlib simulations of a BTO-like low-Earth orbit to estimate afterglow-inducing event rates, assign the measured C-beam durations as dead times, and conclude that NaI loses ~0.02 s per orbit to afterglow versus ~9 s for CsI, motivating the choice of NaI for BTO. The direct duration measurements are repeated over 50+ pulses per configuration with gain correction and error bars, but the orbital impact calculations rely on several assumptions about afterglow thresholds and dead times that are not robustly justified.","tokens_in":14277,"tokens_out":4377,"duration_ms":38551,"significance":"If the measured afterglow contrast is correct, the paper provides a useful, quantitatively documented input for scintillator selection in space missions using SiPM readout, and the HIMAC dataset with per-pulse statistics is a strength. The central qualitative conclusion—that CsI exhibits substantially stronger and longer afterglow than NaI under heavy-ion irradiation—is well supported by the waveform analysis. However, the quantitative claims that drive the trade-study conclusion (event spacings of ~70 s vs ~1.4 s and lost time of 0.02 s vs 9 s per orbit) depend on threshold and dead-time choices that are not securely grounded, so the paper's main quantitative predictions need revision before the conclusion can be considered robust.","major_comments":[{"comment":"The afterglow-inducing thresholds are defined as incident kinetic energies (920 MeV for NaI, 164 MeV for CsI) and used to integrate the MEGAlib hadronic flux, but the HIMAC data establish afterglow as a function of energy deposited in the crystal: the 920 MeV He beam deposits only 275 MeV in NaI and 325 MeV in CsI, while the C beam deposits 1.65/2 GeV. A 920 MeV proton traversing a 3.8-cm crystal deposits far less energy than the 920 MeV He used in the experiment, so integrating all hadrons above 920 MeV likely misstates the event rate. The authors should apply the energy cut on deposited energy in the scintillator (as they do for the 74.5 MeV CsI validation) or explicitly justify why incident kinetic energy is the relevant variable; as written, the claimed event spacings and the 0.02 s vs 9 s lost-time comparison are not securely grounded.","section":"§3.1"},{"comment":"The CsI threshold of 164 MeV is obtained by extrapolating the afterglow-duration-versus-excitation-energy relation from the same HIMAC data (including the 662 keV calibration point) and is then used to compute the CsI afterglow event rate of 0.7 particles/s. This makes the 'every 1.4 s' rate partly an encoding of the measured durations rather than an independent prediction, and the uncertainty in the threshold from the polynomial fit is not propagated. The paper should present this as an explicit calibration step and propagate the threshold uncertainty into the event-rate and lost-time estimates, or use an independently justified threshold.","section":"§3.1"},{"comment":"The dead-time-per-event assignment uses the C-beam afterglow durations (2.4 ms for CsI, 0.28 ms for NaI) for all events above threshold, even though the HIMAC data show a strong dependence on deposited energy (the He beam gives 0.9 ms in CsI). If many events just above threshold deposit less energy, the assigned dead time overestimates the lost time; if the threshold is set too low, the event rate may be overestimated. A sensitivity study over reasonable threshold and dead-time choices is needed to support the 0.02 s vs 9 s conclusion, which is the central quantitative basis for the trade-study decision.","section":"§3.1 and §3.2"},{"comment":"The rate calculation explicitly excludes the South Atlantic Anomaly ('outside of the SAA'), yet the text states that afterglow in the SAA 'will be strong' and that BTO will remain powered during SAA crossings. Since trapped hadrons in the SAA may dominate the afterglow-inducing event rate, the quoted lost-time estimates are incomplete as stated. The authors should either quantify the SAA contribution or explicitly restrict the conclusion to non-SAA portions of the orbit.","section":"§3.1"}],"minor_comments":[{"comment":"There is an internal inconsistency in the CsI event spacing: §3.1 reports 0.7 particles/s (one event every 1.4 s), while §3.2 states 'every ~0.7 seconds in the CsI detector'; the latter should read ~1.4 s.","section":"§3.2"},{"comment":"The NaI threshold is justified by saying the He afterglow signal 'can be approximated as zero,' but Table 2 lists a nonzero NaI He afterglow duration of 0.16 ± 0.04 ms; this approximation should be stated more carefully, and its effect on the rate estimate should be addressed.","section":"§3.1"},{"comment":"The claim that the lost observing time in NaI is '~0.01×' that in CsI is numerically inconsistent with the quoted values 0.02 s vs 9 s, which give a ratio of ~0.002.","section":"§3.1"},{"comment":"The caption refers to a 'post-post residual' in the third panel; this appears to be a typo for 'post-pulse residual.'","section":"Figure 4"}],"recommendation":"major_revision","confidential_remarks":"The direct afterglow measurements seem sound and valuable, but the orbital extrapolation in §3.1 is the load-bearing part of the trade-study conclusion and it currently has a deposited-energy/incident-energy mismatch, a partially circular threshold, and an unquantified SAA contribution. These are fixable within the manuscript's scope, so I recommend major revision rather than rejection. The inconsistency between '1.4 s' and '0.7 s' for CsI event spacing should also be corrected."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What you should know: this is a genuinely useful measurement paper. The authors took Scionix NaI(Tl) and CsI(Tl) 51B51 modules with SiPM readout to HIMAC, hit them with 230 MeV/u He and 350 MeV/u C beams, and measured afterglow durations, amplitudes, and decay constants over 50+ pulses per setup. The CsI afterglow is 8.6x (C) and 5.6x (He) longer than NaI, and even the weakest CsI afterglow is stronger than the strongest NaI case. That contrast is robust and is the paper's real contribution. Prior afterglow studies mostly used PMTs or photodiodes; this is the first quantitative SiPM-module data I know of for these commercial packages, and it lands directly on a mission trade study for BTO on COSI. The analysis is careful: gain correction with Ba-133, removal of the scintillator/electronics baseline with a series of exponentials, and explicit uncertainties from the fit chain.\n\nThe soft spot is the rate extrapolation in Section 3.1. The authors set afterglow thresholds in terms of incident particle energy (920 MeV for NaI, 164 MeV for CsI), but the HIMAC data show afterglow scales with deposited energy: the 920 MeV He beam deposits only 275 MeV in NaI and 325 MeV in CsI. Counting all hadrons above 920 MeV incident almost certainly overestimates the NaI afterglow rate, since a 920 MeV proton deposits far less than 275 MeV in a 3.8 cm crystal. The CsI threshold is a little better—they translate it to 74.5 MeV deposited and validate it against Rau et al.—but then they don't use a deposited-energy cut in the MEGAlib rate integration. So the 0.02 s vs 9 s per-orbit loss numbers are order-of-magnitude estimates at best. There's also a text inconsistency: abstract and §3.1 give a 1.4 s event spacing for CsI, while §3.2 says every ~0.7 s.\n\nNone of this overturns their conclusion. The CsI threshold is still far lower than NaI's even in deposited energy, and the afterglow dead time per event is ~8x longer, so NaI remains the sensible choice. But I'd ask the authors to redo the integration with deposited-energy thresholds and propagate uncertainties into the lost-time numbers before publication.\n\nThis paper deserves a serious referee. It's aimed at detector engineers and small gamma-ray missions, and the measured durations are directly reusable. I'd send it to NIM A or Astroparticle Physics with the threshold issue flagged.","headline":"Careful new afterglow measurements for SiPM-read Scionix modules, but the orbit-loss numbers rest on a threshold treatment that mixes incident and deposited energy.","tokens_in":14896,"tokens_out":3510,"would_cite":true,"duration_ms":30226,"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":"This paper shows CsI(Tl) afterglow lasts 5.6–8.6 times as long as NaI(Tl) after heavy-ion hits, and concludes NaI is the better detector for BTO because it loses only 0.02 s per orbit to afterglow versus ~9 s for CsI.","keywords":["gamma-ray scintillators","scintillator afterglow","all-sky gamma-ray survey","time-domain astrophysics","NaI(Tl)","CsI(Tl)","silicon photomultipliers","detector dead time"],"falsifier":"Measure afterglow in NaI(Tl) for deposited energies between roughly 100 and 900 MeV (for example by placing absorbers in front of a heavy-ion beam or using a tuneable beam energy). If NaI shows afterglow durations well above zero at energies below 920 MeV, the assumed 920 MeV threshold is wrong and the NaI dead-time loss of 0.02 s per orbit is an underestimate; the trade-study conclusion would need revisiting. Alternatively, a flight measurement of afterglow trigger rates inside the South Atlantic Anomaly could test the rate prediction directly.","tokens_in":13742,"feed_emoji":"🔭","tokens_out":7051,"duration_ms":53800,"temperature":0.7,"pith_summary":"The paper aims to settle a detector choice for the Background and Transient Observer (BTO), a gamma-ray transient monitor flying on the COSI mission. By irradiating NaI(Tl) and CsI(Tl) scintillators with helium and carbon beams, it measures the afterglow pulses that follow large energy deposits. It finds CsI's afterglow lasts 8.6 times (carbon beam) and 5.6 times (helium beam) as long as NaI's, and that CsI's stronger afterglow would force either a much longer dead time per event or a much higher trigger threshold. Combining the measured durations with simulated orbital background rates, the paper estimates that CsI would lose about 9 seconds of observing time per orbit to afterglow, while NaI would lose about 0.02 seconds. On this basis it concludes that NaI is the better scintillator for BTO.","feed_headline":"CsI afterglow would cost 9 seconds per orbit; NaI loses 0.02","feed_subtitle":"Measured afterglow durations pick NaI over CsI for the gamma-ray transient monitor flying on COSI.","key_machinery":"The central object is the afterglow signature itself: the train of small, delayed scintillation pulses that follows a saturating energy deposit. The paper's key measurement procedure is waveform analysis in which the post-pulse trend (scintillation decay plus electronics baseline) is fitted with a series of exponential functions, subtracted, and the residual voltage spread (sigma) is tracked in time bins until it returns to the pre-pulse baseline. The afterglow duration is defined as the time for that spread to reach baseline; this duration is then used as the dead time per afterglow-inducing event. A second load-bearing piece is the energy threshold for afterglow induction, estimated as 920 MeV for NaI (the helium-beam energy, since He produced essentially no NaI afterglow) and 164 MeV for CsI via a polynomial extrapolation constrained by a 662 keV no-afterglow calibration. These thresholds convert simulated hadronic background counts into afterglow event rates.","core_discovery":"On its own terms, this paper establishes that both NaI(Tl) and CsI(Tl) detector modules produce measurable afterglow after heavy-ion irradiation, but with dramatically different severities. For a 350 MeV/u carbon beam, afterglow pulses persist for a median 2.40 ms in CsI versus 0.28 ms in NaI; for a 230 MeV/u helium beam, 0.9 ms versus 0.16 ms. The CsI afterglow is also more intense, with a residual voltage spread about twice that of NaI even though NaI has a higher gain. Because afterglow events look like small pulses that can falsely trigger a gamma-ray detector, the paper translates these durations into dead-time costs: after estimating that a BTO-like instrument in low Earth orbit would see afterglow-inducing particles every ~74 s in NaI and every ~1.4 s in CsI, it finds the observing-time loss per orbit is ~0.02 s for NaI and ~9 s for CsI. The paper's conclusion is that NaI is the better choice for BTO despite CsI's higher light yield and radiation hardness.","pith_inferences":["The measured energy thresholds and durations suggest a testable scaling law: afterglow duration in these crystals likely grows with total deposited energy, and the polynomial fit used for CsI could be extended to NaI to predict its afterglow at energies between a few hundred MeV and 1 GeV, a regime more typical of trapped protons than the heavy-ion beams used here.","Because trapped protons in the South Atlantic Anomaly have lower energies than the carbon beam but much higher flux, the orbital loss estimate may be dominated by how often a 100–500 MeV proton deposit crosses the threshold; the paper's threshold method could be sharpened with proton-beam measurements.","A similar trade-study logic applies to other scintillator-based transient monitors: the choice between high light yield and long afterglow can be framed as an observing-time budget, where the product of afterglow duration and event rate determines which material wins."],"forward_implications":["If NaI is used on BTO, afterglow-inducing events are expected every ~74 seconds per detector, meaning afterglow dead time costs about 0.02 s per orbit.","If CsI were used instead, afterglow-inducing events would arrive every ~1.4 seconds, and 2.4 ms dead time per event would cost about 9 s per orbit, roughly 400 times more lost time.","CsI afterglow would force either a trigger threshold about 3 times higher or a dead time about 7 times longer than NaI to avoid false triggers, both of which degrade the 30 keV–2 MeV science band.","These results imply that any space-based gamma-ray instrument using CsI(Tl) with SiPM readout should plan for afterglow mitigation such as post-saturation dead time or coincidence requirements."],"supporting_citations":[{"why":"Provides the MEGAlib background simulation toolkit used to compute the BTO orbital background rates and afterglow-inducing particle rates.","marker":"Zoglauer et al., 2006"},{"why":"Supplies the 8–80 MeV deposited-energy range for afterglow in CsI, used to validate the paper's 164 MeV energy-threshold estimate.","marker":"Rau et al., 2005"},{"why":"Defines the afterglow mechanism and the practice of fitting pulse shapes with a series of exponentials, which the paper's analysis method follows.","marker":"Lecoq, 2020"},{"why":"Reports typical afterglow fractions in NaI(Tl) on millisecond timescales, providing the baseline expectation the measurements are compared against.","marker":"Koppert et al., 2018"},{"why":"Documents that CsI(Tl) afterglow under high-energy irradiation is strong enough to prevent its use in many applications, motivating the trade-study focus.","marker":"Alfassi et al., 2009"}],"fun_headline_variants":["NaI beats CsI for gamma-ray monitor: 0.02 s vs 9 s dead time per orbit","Afterglow tests: CsI dead time 450x worse than NaI for BTO","CsI afterglow pulses last 2.4 ms vs NaI's 0.28 ms under carbon beam","For gamma-ray transients, NaI afterglow beats CsI's by 450x in dead time","BTO chooses NaI: CsI afterglow would eat 9 s per orbit vs 0.02 s"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The comparison rests on assuming that only particles above a particular energy threshold (920 MeV in NaI, 164 MeV in CsI) produce afterglow and that each such event costs exactly the measured carbon-beam afterglow duration in dead time; if significant afterglow occurs at lower energies, or if the South Atlantic Anomaly contributes more events than the simulation assumes, the 0.02 s versus 9 s loss comparison and the choice of NaI could change.","fun_headline_variants_meta":{"raw":{"variants":["NaI beats CsI for gamma-ray monitor: 0.02 s vs 9 s dead time per orbit","Afterglow tests: CsI dead time 450x worse than NaI for BTO","CsI afterglow pulses last 2.4 ms vs NaI's 0.28 ms under carbon beam","For gamma-ray transients, NaI afterglow beats CsI's by 450x in dead time","BTO chooses NaI: CsI afterglow would eat 9 s per orbit vs 0.02 s"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000906,"raw_usage":{"total_tokens":4040,"prompt_tokens":1231,"completion_tokens":2809,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":847,"completion_tokens_details":{"reasoning_tokens":2671}},"tokens_in":847,"tokens_out":2809,"duration_ms":20948,"temperature":1.0,"reasoning_tokens":2671,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T13:15:50.697768+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure afterglow in NaI(Tl) for deposited energies between roughly 100 and 900 MeV (for example by placing absorbers in front of a heavy-ion beam or using a tuneable beam energy). If NaI shows afterglow durations well above zero at energies below 920 MeV, the assumed 920 MeV threshold is wrong and the NaI dead-time loss of 0.02 s per orbit is an underestimate; the trade-study conclusion would need revisiting. Alternatively, a flight measurement of afterglow trigger rates inside the South Atlantic Anomaly could test the rate prediction directly.","supporting_citations":[{"cited_title":"2006, New Astronomy Reviews, 50, 629 11","cited_arxiv_id":null,"evidence_quote":"Provides the MEGAlib background simulation toolkit used to compute the BTO orbital background rates and afterglow-inducing particle rates."},{"cited_title":"V ., Hurley, K., & Lichti, G","cited_arxiv_id":null,"evidence_quote":"Supplies the 8–80 MeV deposited-energy range for afterglow in CsI, used to validate the paper's 164 MeV energy-threshold estimate."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the afterglow mechanism and the practice of fitting pulse shapes with a series of exponentials, which the paper's analysis method follows."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reports typical afterglow fractions in NaI(Tl) on millisecond timescales, providing the baseline expectation the measurements are compared against."},{"cited_title":"2009, Nuclear In- struments and Methods in Physics Research Section A: Accelerators, Spec- trometers, Detectors and Associated Equipment, 606, 585","cited_arxiv_id":null,"evidence_quote":"Documents that CsI(Tl) afterglow under high-energy irradiation is strong enough to prevent its use in many applications, motivating the trade-study focus."}],"review_version":1}