{"id":"4f6ccfbc-aed1-41fe-a971-efaa411f0b2e","arxiv_id":"2607.29435","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"Persistence in H4RG-15 detectors varies widely between detectors and with temperature; only 2 of 15 show the previously reported 65 K persistence peak.","lead":"ESO measured persistence—residual signal from bright exposures—in 15 infrared detectors for its MOONS, MICADO, and HARMONI instruments and found each detector behaves differently with temperature. The study maps this behavior and recommends that instruments be built with flexible cooling so the operating temperature can be optimized after each detector is tested.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Temporal stability of lab-measured trap maps is an acknowledged but unresolved assumption; the on-sky ERIS/NIX counterexample directly challenges the persistence-correction pipeline's validity over instrument lifetime.","rationale":"The paper is a transparent instrument-characterization study. It honestly acknowledges the temporal-stability assumption and the unexplained ERIS/NIX decrease, which is a strength. However, that transparency does not remove the load-bearing nature of the assumption: the correction pipeline and the operating-temperature recommendation both rely on lab-measured trap maps being valid for years of on-sky use. The reader's CONDITIONAL verdict already captures this, so no change is needed. I considered whether the lack of repeated measurements per temperature or the internal '5 values' vs. 6 exponentials inconsistency (Section 2.2) might be more central, but those are secondary; the stability issue is the one the paper itself identifies as unresolved and the one that directly undermines the applied value. A controlled continuous-cold experiment or on-sky commissioning data would settle whether the concern lands.","tokens_in":15040,"tokens_out":11893,"duration_ms":134990,"concrete_test":"Perform the §2.1 quick persistence verification (≈50 ke- flat followed by 300 s dark) on a fully characterized H4RG-15 detector kept continuously at its operating temperature (e.g., 82 K) for 6–12 months without thermal cycling, repeating monthly; also run the same test on MOONS/MICADO detectors during commissioning. If the median persistence amplitude or spatial trap pattern drifts by more than the lab repeatability (established through repeated measurements), the stability assumption fails and the correction pipeline requires periodic recalibration; if stable, the ERIS/NIX decrease is likely detector-specific.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's applied claims—automatic pipeline correction and the recommendation to set operating temperature after lab characterization—depend on the assumption that persistence trap maps and time constant vectors measured in the lab remain representative after detector installation and operation in the final instrument. Section 5 states this directly: \"One important factor in this approach is that the persistence trap maps and time constant vectors must remain stable over time for this strategy to work.\" The same section then reports the opposite: the ERIS/NIX detector's persistence amplitude decreased with time at the telescope until persistence is \"almost never seen,\" with no explanation beyond speculation about temperature or continuous cold. The paper provides no quantitative stability bound, no on-sky H4RG validation (MOONS commissioning is upcoming), and the one long-term lab stability datum (EG20370, 11 cool-downs) is not equivalent to continuous cold operation. This is a genuine external-validity gap, not a disagreement with consensus; the pipeline's usefulness hinges on it.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper reports a systematic characterization of image persistence in 15 H4RG-15 HgCdTe detectors procured for ESO's MOONS, MICADO, and HARMONI instruments. The authors describe a standardized test protocol (LED illumination, 10,000 s soak, 14 h non-destructive readout) and fit detrapping curves with a sum of six exponentials with fixed time constants to produce per-detector time-constant spectra and pixel-level trap-density maps. They classify detector behavior into five categories (increasing, flat, decreasing, 65 K peak, high contact resistance) based on the temperature dependence of median trap density between 40 and 85 K, and they demonstrate the use of these maps in a persistence correction pipeline implemented in the ESO HDRL and in the Pyxel simulator. The paper concludes that each detector is unique and recommends instrument designs with flexible thermal interfaces so operating temperature can be chosen after full detector characterization.","tokens_in":15259,"tokens_out":6301,"duration_ms":67229,"significance":"The main value of this work is the empirical dataset: 15 science-grade H4RG-15 detectors characterized under a consistent protocol, with public access to the modeling framework and an example notebook. The categorization of persistence behavior is practically useful for instrument design, and the implementation of the correction algorithm in operational software is a concrete deliverable. If the stability assumption is validated, the approach could enable automatic persistence correction for ESO NIR instruments. However, the paper's quantitative claims are limited by the absence of uncertainty estimates and by the unresolved temporal-stability counterexample, which temper the strength of the conclusions.","major_comments":[{"comment":"The central taxonomy and model parameters are presented without uncertainties. The text states that each temperature is typically measured only once and that per-pixel S/N is insufficient for individual spectral fits. Nevertheless, Table 1 lists median trap densities to 0.1% and assigns detectors to categories based on their temperature dependence. Without error bars, repeat measurements, or goodness-of-fit statistics, categories such as 'flat' versus 'increasing' cannot be robustly separated (e.g., EG20373: 0.6%/0.7% vs SG20369: 0.6%/1.5%). Please report uncertainties on N_i, on the median trap densities, and on the classification, e.g., via bootstrap or repeat measurements.","section":"§2.1–2.2, Eq. (1), Table 1"},{"comment":"The paper states that 'the persistence trap maps and time constant vectors must remain stable over time for this strategy to work,' then immediately reports that the ERIS/NIX detector's persistence amplitude decreased on sky until it is 'almost never seen,' with no explanation. This is a direct counterexample to the stability assumption underlying the correction pipeline and the recommendation to set operating temperature after lab characterization. The manuscript provides no quantitative stability bound, no monitoring protocol, and no on-sky H4RG validation (MOONS commissioning is listed as future work). This external-validity gap must be addressed before the pipeline can be considered operationally validated.","section":"§5"},{"comment":"The example correction shown in Figure 6 appears to be an in-sample demonstration: the input dark frame and the persistence trap maps were collected in the same test campaign (the mask shadow is mentioned). An in-sample fit does not demonstrate predictive performance on new data. Please clarify whether the trap maps were derived from independent exposures, and if not, provide an out-of-sample test, such as prediction on a different cool-down, a different illumination level, or on-sky data.","section":"§5, Figure 6"},{"comment":"The two-stage fitting procedure is underspecified: after co-adding regions to obtain a normalized time-constant spectrum, 'every pixel' is fit for total trap number. Given the stated per-pixel S/N limitation, what is the uncertainty on the pixel-level trap maps? The paper does not provide residual maps or validation of the fixed-shape assumption. Please clarify the fitting method and quantify the resulting uncertainties on the trap maps.","section":"§2.2"}],"minor_comments":[{"comment":"The text says 'a single vector with 5 values' but Eq. (1) has six time constants (1, 10, 100, 1000, 10000, 100000 s); should be 6.","section":"§2.2"},{"comment":"Typo in caption: 'tempeartures' should be 'temperatures'.","section":"Figure 1"},{"comment":"For SG19907 and SG19910, the category 'increasing or 65K peak' makes the phrase 'only 2 of 15 definitively' somewhat misleading; consider stating '2 definitive, 2 ambiguous' explicitly.","section":"Table 1, §3.4"},{"comment":"The phrase 'the rare case where a science exposure is contaminated' seems to understate the earlier discussion that persistence is expected in several observing modes; consider rewording.","section":"§5"},{"comment":"Several references are to 2026 SPIE proceedings; please ensure DOIs or arXiv identifiers are included where available.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The manuscript reads as an SPIE proceedings paper; for a full research journal, the lack of uncertainty quantification and external validation is more significant. The paper is nevertheless a useful data paper, and the gaps are addressable with additional analysis and a more cautious framing. I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a solid, clearly written instrument characterization paper. The new piece is the dataset—persistence trap maps and time-constant spectra for 15 science-grade H4RG-15 detectors from the MOONS and MICADO batches, measured at multiple temperatures between 40 and 85 K over 2019–2025. The five-category classification (increasing, flat, decreasing, 65 K peak, high contact resistance) is useful shorthand, and the finding that only 2 of 15 detectors definitively show the 65 K peak justifies revising ESO's earlier advice to avoid 45–75 K. That revision is the most actionable result.\n\nThe paper does several things well. The test protocol is concrete and reproducible: soak, reset, 14-hour non-destructive readout, fixed time-constant bins, co-added regions for signal-to-noise. The analysis is transparent, and the limitations are stated rather than hidden. The Pyxel example notebook with real trap-map data is a genuine community resource. The heavy reliance on the authors' own prior work (Tulloch & George, Ives et al.) is natural—this is an application and extension of that model, not a claim to a new physical mechanism.\n\nSoft spots, in order of importance. First, the persistence-correction pipeline's central assumption—that lab-measured trap maps and time-constant vectors remain stable after installation and over instrument lifetime—is not just unverified; the paper's Section 5 reports that ERIS/NIX persistence almost vanished at the telescope, with no explanation. That doesn't kill the paper, but it means the correction algorithm should be treated as a prototype pending MOONS on-sky validation. Second, there are no uncertainty estimates on fitted trap densities or time-constant spectra, and each temperature point is usually measured once. The classification is qualitative enough that this is a minor issue for the main conclusions, but it limits quantitative comparisons across detectors. Third, the model is an empirical sum of exponentials; the pipeline 'prediction' is an application of fitted parameters, not an independent test. The paper mostly avoids overstating this.\n\nThe central recommendation—flexible thermal interfaces so operating temperature can be optimized after detector characterization—is well supported by the data. I didn't find load-bearing errors. This is an honest, useful engineering paper for anyone building or operating H4RG-15 systems. It deserves peer review; the stability question is the one to put to the authors.","headline":"Useful empirical survey of persistence in 15 H4RG-15s with a new classification and revised temperature recommendation; pipeline stability is acknowledged but unresolved.","tokens_in":15749,"tokens_out":2671,"would_cite":true,"duration_ms":28127,"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":"A six-year, 15-detector lab survey shows that H4RG-15 image persistence is a per-device property: one time-constant spectrum plus a trap map describes it, and only 2 of 15 detectors show the previously feared 65 K peak.","keywords":["image persistence","H4RG-15","HgCdTe detectors","infrared detector characterization","trap density map","detrapping time constants","operating temperature optimization","detector persistence model"],"falsifier":"Take a fully characterized detector, measure its persistence with the standard quick test after installation at the same operating temperature, and compare the observed amplitude and decay with the lab-built trap map; a deviation beyond lab repeatability would falsify the claim that lab characterization predicts in-situ persistence. The paper's own ERIS/NIX observation — a large initial trap amplitude that essentially vanished over time — is already an indicator to check.","tokens_in":14949,"feed_emoji":"🔭","tokens_out":5294,"duration_ms":55722,"temperature":0.7,"pith_summary":"The paper reports a six-year laboratory campaign that characterized image persistence in fifteen H4RG-15 near-infrared detectors built for the MOONS and MICADO instruments, testing each between 40 and 85 kelvin. It claims that for most detectors the persistence behaviour across the entire array can be captured by a single time-constant spectrum plus a per-pixel trap-density map, and that the detectors fall into five behavioural types: persistence rising with temperature, flat, falling with temperature, a distinct peak near 65 kelvin, and artefacts from high contact resistance. Crucially, only two of the fifteen detectors definitively show the 65 kelvin peak, so the earlier recommendation to avoid operating temperatures between 45 and 75 kelvin should be reconsidered for the majority. Because each detector has a unique optimal temperature, the paper recommends flexible thermal interfaces in instrument design so the operating temperature can be chosen after full characterization.","feed_headline":"Only 2 of 15 detectors show the 65 K persistence peak","feed_subtitle":"Lab-derived trap maps let ESO predict and correct infrared image memory; most arrays can run at any temperature from 40 to 85 K.","key_machinery":"The persistence model: detrapping curves from long dark reads are fit to a sum of six exponentials with fixed time constants tau = 1, 10, 100, 1000, 10000, 100000 s; the fitted coefficients form a time-constant spectrum N, normalized to a trap-fraction vector that is fixed across the array, while a second fit per pixel yields a 4096x4096 maximum-trap map (and a trap-density map relative to illumination). This decomposition lets persistence be predicted and subtracted without knowing the underlying defect physics, and the same data can drive simulations.","core_discovery":"The central claim is that a single persistence model — a normalized vector of detrapping time constants (1, 10, 100, 1000, 10000, 100000 seconds) plus a full-frame trap-density map — is sufficient to describe and correct image persistence for most H4RG-15 detectors at a given temperature, and that the device-to-device variation in persistence is large enough that operating temperature must be chosen per detector. The paper's key empirical finding is that only 2 of 15 detectors exhibit the previously identified 65 K persistence peak, meaning the blanket recommendation to avoid 45–75 K is not generally valid. The paper also shows that some detectors contain distinct pixel populations (hot-pixe","pith_inferences":["The documented decline of persistence in one on-telescope detector suggests trap maps may drift over time; a periodic quick-verification test after installation would tell whether recalibration is needed, something the paper leaves open.","If the 0.13 eV trap is a process-related defect, a short screening test at 60–65 K on future deliveries could identify 65 K-peak devices early, before cryostat design is frozen.","The same characterization protocol could be applied to other HxRG generations (e.g., H2RG) to build a common ESO-wide persistence archive, extending the results beyond H4RG-15.","A testable extension: use the trap-density maps to predict persistence for arbitrary illumination patterns (e.g., moving OH lines in MOONS/HARMONI) and validate against on-sky darks; agreement would confirm the model's extrapolation to non-flat illumination."],"forward_implications":["If the model holds, persistence correction frames can be generated automatically from the exposure history ahead of each science frame, as already deployed for at least one instrument.","Operating temperatures for MOONS, MICADO, and HARMONI can be tuned per detector, trading persistence against noise and quantum efficiency.","Only detectors with hot-pixel or high-contact-resistance regions need more than a single time-constant spectrum; for the majority, one spectrum and one trap map suffice.","The rarity of the 65 K peak means future instruments need not avoid the 45–75 K range as a blanket rule, simplifying cryostat design.","The characterization data can be fed into the Pyxel simulator to test observing strategies before the detector is in the instrument."],"fun_headline_variants":["2 of 15 H4RG-15s trigger the 65 K persistence peak","65 K persistence peak rare: only 2 of 15 H4RG-15s","Most H4RG-15s avoid the 65 K persistence peak","For H4RG-15s, 65 K persistence peak is the exception, not the rule","Persistence peak at 65 K is rare: tune each H4RG-15 individually"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The lab-measured persistence trap maps and time-constant vectors remain stable and representative after the detector is installed and operated in the final instrument over its lifetime; the paper states this requirement and reports an unexplained decline of persistence in an on-telescope detector.","fun_headline_variants_meta":{"raw":{"variants":["2 of 15 H4RG-15s trigger the 65 K persistence peak","65 K persistence peak rare: only 2 of 15 H4RG-15s","Most H4RG-15s avoid the 65 K persistence peak","For H4RG-15s, 65 K persistence peak is the exception, not the rule","Persistence peak at 65 K is rare: tune each H4RG-15 individually"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001338,"raw_usage":{"total_tokens":5287,"prompt_tokens":763,"completion_tokens":4524,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":507,"completion_tokens_details":{"reasoning_tokens":4425}},"tokens_in":507,"tokens_out":4524,"duration_ms":31474,"temperature":1.0,"reasoning_tokens":4425,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T07:08:07.004883+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a fully characterized detector, measure its persistence with the standard quick test after installation at the same operating temperature, and compare the observed amplitude and decay with the lab-built trap map; a deviation beyond lab repeatability would falsify the claim that lab characterization predicts in-situ persistence. The paper's own ERIS/NIX observation — a large initial trap amplitude that essentially vanished over time — is already an indicator to check.","supporting_citations":[],"review_version":1}