{"id":"e27716af-e5e8-44ac-ac1a-34d2dde44e8e","arxiv_id":"2412.14850","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A 610-day aging test of a scientific CMOS X-ray detector shows no significant degradation in key performance parameters, with a projected gain loss of 2.4% over ten years.","lead":"A scientific CMOS X-ray detector was aged for 610 days on the ground, and its bias, dark current, noise, gain, and energy resolution stayed essentially unchanged. The paper projects only about 2.4% gain drift over ten years, which matters for space missions that rely on CMOS sensors.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 10-year gain projection is not statistically supported: the fitted rate constant k=2.4e-3/yr is indistinguishable from zero given the ±0.02 eV/DN errors on the gain values in Table 2.","rationale":"The reader's CONDITIONAL verdict already identifies the extrapolation as a weakness; this pass sharpens the concern. The weakest link is not primarily the 23% pixel mask, which is standard practice and mostly affects generalization, but the statistical content of the gain fit itself. Table 2 reports gain uncertainties comparable to the fitted decline, yet no uncertainty is propagated to k or to the 0.73% and 2.41% projections. The proposed test would settle this: if the confidence interval for k excludes zero and is tight, the projection stands; if it includes zero, the abstract's numerical lifetime predictions should be withdrawn or reframed as an upper limit. The direct 610-day stability evidence for the analyzed pixels is plausible and consistent with the presented plots, so rejection of the paper would be too harsh. I would retain the reader's CONDITIONAL verdict, adding the explicit requirement of a confidence interval on k and, ideally, release of the underlying gain time series. I partially agree with the reader because the same broad extrapolation concern was flagged, but the decisive issue is that even the in-sample fit is statistically underdetermined.","tokens_in":9444,"tokens_out":6080,"duration_ms":49721,"concrete_test":"Re-fit the eight -30C gain entries in Table 2 (days 0, 32, 94, 134, 198, 260, 323, 427) to G=G0 exp(-kt) by weighted least squares, using the quoted ±0.02 eV/DN errors, and report the 95% confidence interval for k via profile likelihood or bootstrap. If the interval contains zero, recompute the abstract's 3-year and 10-year projected degradations from the upper confidence bound; if the upper-bound projection is materially larger than 2.41%, the paper must present the projection as an upper limit with uncertainty rather than a point prediction. Repeat the fit omitting the day-427 point, the epoch with the reported setup issue visible in Figure 4, to check whether k changes sign; a robust conclusion should not depend on a single endpoint.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing concern is the lifetime projection, not the direct 610-day stability claim. The projection G=G0 exp(-kt), with k=2.4e-3/yr, is obtained from the -30C gain column of Table 2: 6.62, 6.62, 6.62, 6.62, 6.62, 6.61, 6.61, 6.60 eV/DN, each with a quoted uncertainty of ±0.02 eV/DN. The total drop over the 427-day fit window is 0.02 eV/DN (about 0.3%), equal to the stated uncertainty, and the series is non-monotonic: the -30C gains on days 584 and 610 return to 6.61 eV/DN. No uncertainty, confidence interval, or goodness-of-fit is reported for k. A simple endpoint propagation gives k approximately 0.0026 ± 0.004 yr^-1, so the 95% range includes k=0 and even negative k. The abstract's 0.73% (3-year) and 2.41% (10-year) figures are therefore consistent with no aging and cannot be regarded as predictions. The 10-year number is also an extrapolation roughly six times beyond the 427-day fit interval, assuming an unchanged first-order mechanism. The direct 610-day measurements do support 'no observable degradation' for the analyzed pixels, so this concern targets the quantitative lifetime claim in the abstract rather than the core stability assertion.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a 610-day aging test of a scientific CMOS X-ray detector (EP4K), combining 16 months at -30 °C with 5 months at 20 °C and periodic measurements of bias, dark current, readout noise, X-ray conversion gain, and energy resolution. The authors find no observable degradation in these quantities over the direct measurement period, observe a small number of pixels with bias or noise variations, and then use a first-order kinetic fit to the gain time series to project a gain degeneration of 0.73% over 3 years and 2.41% over 10 years. The central stability claim is based on direct measurements, while the quantitative lifetime projection is an extrapolation from a fitted model.","tokens_in":9797,"tokens_out":3823,"duration_ms":29391,"significance":"If the stability claim holds, the paper provides valuable empirical support for the use of scientific CMOS detectors in long-lived space missions, complementing prior radiation-tolerance studies. The 610-day direct data set, including the degraded-pixel census and the explicit statement of excluded regions, is a useful contribution. However, the paper's headline quantitative lifetime prediction is not statistically supported: the fitted rate constant is consistent with zero, no uncertainty is reported, and the projection is based on an extrapolation well beyond the measured interval. The direct stability measurements remain the paper's main strength; the projection should be substantially reframed or removed.","major_comments":[{"comment":"The first-order kinetic fit G = G0 exp(-kt) is applied to the -30°C gain values of Table 2, which are 6.62, 6.62, 6.62, 6.62, 6.62, 6.61, 6.61, and 6.60 eV/DN, each with a quoted uncertainty of ±0.02 eV/DN. The total change over 427 days is 0.02 eV/DN, equal to the stated uncertainty, and the series is not strongly monotonic. The paper reports k = 2.4e-3 yr^-1 with no confidence interval, goodness-of-fit, or residual analysis. A simple endpoint propagation from the quoted errors yields k approximately 0.0026 ± 0.004 yr^-1, so the 95% range includes k = 0 and negative values. The abstract's statements that aging predicts 0.73% degeneration over 3 years and 2.41% over 10 years are therefore not statistically supported by the data as presented. In addition, the 10-year projection corresponds to an extrapolation about a factor of eight beyond the 427-day fitting window, and the model is imported from organic solar cell lifetime studies (reference [34]) without evidence that it applies to CMOS gain aging. The authors should either provide a proper uncertainty estimate for k, demonstrate that the model is identifiable, and state the extrapolation caveat explicitly, or remove the quantitative lifetime claim from the abstract and conclusions.","section":"Section 4, Table 2"},{"comment":"The analysis excludes roughly 3.9 million pixels, or 23% of the 4k x 4k array, as having inherent defects or light leaks. Consequently, the statement that 'the bias map, dark current, readout noise, gain, and energy resolution exhibited no observable degradation' applies only to the remaining 77% of pixels. The manuscript does not check whether the excluded regions themselves are stable during aging, nor does it demonstrate that the exclusion criteria are time-independent. The degraded-pixel statistics in Section 3.5 are also computed on the reduced pixel set, so the abstract's claim of 'less than 50 pixels' is potentially misleading if additional aging-related changes occur in the masked areas. The authors should either analyze the excluded regions for stability or explicitly state that the stability conclusion is restricted to the non-excluded pixels.","section":"Section 2, Figure 1"},{"comment":"The intercept values a0 in Table 2 vary between 30 and 50 eV with uncertainties of 14-25 eV, and the FWHM values vary between 194 and 200 eV with uncertainties of about 1 eV. The text states that the intercepts 'remain unchanged' and that energy resolution 'exhibits no significant degeneration,' but no statistical test or trend analysis is provided. Given the scatter and the small number of epochs, a statement of consistency within uncertainties would be more precise. This is not central to the stability claim, but it affects the quantification of 'no observable degradation.'","section":"Section 3.4, Table 2"}],"minor_comments":[{"comment":"The sentence 'The 14 µs signal can be seen as the the true bias level' contains a duplicated 'the'.","section":"Section 3.1"},{"comment":"The phrase 'every a few months' should be 'every few months' or 'every couple of months'.","section":"Section 2"},{"comment":"The phrase 'a aging test' should be 'an aging test'.","section":"Section 4"},{"comment":"The sentence 'There are less than 50 pixels within the 4k x 4k array which show a decrease of the bias under 50 ms integration time by over 10 digital numbers (DNs)' should specify that this count refers to pixels not excluded from analysis, given the 23% exclusion described in Section 2.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The direct 610-day stability data are credible and useful, and the paper is transparent about excluded regions. The main issue is the lifetime projection, which is presented in the abstract and conclusions as a quantitative prediction but is statistically indistinguishable from no aging. This point is fixable by removing or heavily qualifying the projection and by adding uncertainty estimates. I recommend major revision rather than rejection because the core stability claim appears sound; the paper's scope can accommodate the needed changes."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Roman, here's my take.\n\nThe genuinely new thing is the dataset: 610 days of ground aging on an EP4K-class scientific CMOS X-ray sensor, with bias, dark current, readout noise, gain, and energy resolution tracked under controlled conditions. That fills a real gap — annual-scale CMOS aging measurements are rare, and the paper's direct measurements support the no-observable-degradation claim for the pixels that were analyzed. The degraded-pixel section is a nice touch, especially the suppression of random telegraph signal in some pixels after aging.\n\nThe weak spot is the lifetime projection, not the stability claim. The -30°C gain series in Table 2 is flat within the quoted ±0.02 eV/DN errors, non-monotonic (6.61 on days 584 and 610), and the fitted k = 2.4e-3/yr has no reported uncertainty or goodness-of-fit. Propagating the errors, k is consistent with zero or even negative values, so the abstract's 0.73% (3-year) and 2.41% (10-year) numbers are not statistically distinguishable from no aging. The reader's stress-test note is right. This is a moderate flaw: the projection is presented as a quantitative prediction when it's really an upper-limit-style extrapolation.\n\nTwo smaller issues. First, the analysis excludes 23% of pixels (3.9e6) as defective or light-leak regions before computing stability statistics. The paper is transparent about this, but the conclusion about sensor stability technically applies to the unmasked 77%, and the assumption that those exclusions are static is untested. Second, the single-sensor design limits generalization, though for an engineering reliability study that's fairly typical.\n\nOn balance, I'd trust the direct 610-day stability result; it's an honest, useful engineering measurement. The lifetime numbers need to be walked back or re-derived with proper error propagation before publication. The paper deserves peer review — a serious referee should engage with the dataset, not desk-reject it. My recommendation: send it out, but flag the projection issue as the key revision.\n\nIf I were still doing detector work, I'd cite the stability dataset.","headline":"A valuable new aging dataset for scientific CMOS X-ray sensors, but the 10-year gain projection is statistically indistinguishable from zero aging.","tokens_in":10297,"tokens_out":1874,"would_cite":true,"duration_ms":15440,"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":"After 610 days of aging, an X-ray CMOS sensor shows no measurable performance degradation, according to this study.","keywords":["X-ray detector","CMOS sensor","long-term aging","dark current","readout noise","conversion gain","energy resolution","space telescope"],"falsifier":"Re-run a 610-day aging test on the same sensor batch without masking any pixels, tracking every pixel's bias map and noise; if noticeably more than the reported ~50 pixels develop bias shifts above 10 DN or noise shifts above 15 $e^-$, or if the measured gain after several years deviates from the fitted exponential beyond the quoted $0.73\\%$–$2.41\\%$ band, the stability claim would need revision.","tokens_in":1788,"feed_emoji":"🛰️","tokens_out":2338,"duration_ms":50473,"temperature":0.7,"pith_summary":"This paper tries to establish that a scientific CMOS X-ray detector can operate for years in space without losing performance. The authors aged one EP4K sensor for 610 days, first at $-30^\\circ$C for 16 months and then at $20^\\circ$C, measuring bias, dark current, readout noise, conversion gain, and X-ray energy resolution at regular intervals. All key metrics stayed within measurement scatter, with fewer than 50 of the 16 million pixels showing a bias drop larger than 10 DN. From the gain evolution they project a gain degeneration of $0.73\\%$ over 3 years and $2.41\\%$ over 10 years, small enough to be corrected by routine calibration. If correct, this supports the use of CMOS sensors in long-lived X-ray astronomy missions.","feed_headline":"610-day aging test leaves X-ray CMOS sensor's performance intact","feed_subtitle":"Bias, dark current, noise, and X-ray spectrum all stay flat; projected gain drift is under 0.75% in 3 years.","key_machinery":"The central object is the EP4K, a 4k $\\times$ 4k back-illuminated scientific CMOS sensor with 15 $\\mu$m pixels and a 10 $\\mu$m epitaxial layer, operated at 20 Hz in high-gain mode with the PGA register set to 7.5. The aging protocol repeatedly collects dark frames at integration times from 14 $\\mu$s to 1000 s and $^{55}$Fe X-ray spectra; the bias map is the per-pixel median DN, dark current is the slope of bias versus integration time, readout noise is the per-pixel standard deviation at 14 $\\mu$s, and conversion gain comes from linear fits to X-ray line centroids. The lifetime projection is carried by the first-order kinetic model $G = G_0 \\exp(-kt)$, fitted to the gain measurements from the $-30^\\circ$C period, which yields the decay constant used for the 3- and 10-year extrapolations.","core_discovery":"The central claim is that a customized large-format scientific CMOS sensor, the EP4K, shows no observable degradation in bias map, dark current, readout noise, conversion gain, or energy resolution after a 610-day aging period that combines 16 months at $-30^\\circ$C and 5 months at $20^\\circ$C in vacuum. The gain at $-30^\\circ$C remains at $6.62\\pm0.02$ eV/DN throughout the test, and the FWHM of the Mn K$\\alpha$ line stays near 197 eV with only a few eV of scatter. A first-order kinetic fit to the gain decay gives a rate constant $k = 2.4\\times10^{-3}\\,\\mathrm{yr}^{-1}$, translating to a projected gain loss of $0.73\\%$ over 3 years and $2.41\\%$ over 10 years, which the authors describe as low enough for most current and upcoming astronomical missions and easily correctable by annual calibration.","pith_inferences":["The 10-year projection of $2.41\\%$ gain loss is an extrapolation from a fit to only 427 days of $-30^\\circ$C data; including the $20^\\circ$C accelerated-aging points in the fit might change the decay constant, so the true decade-scale uncertainty is likely larger than the quoted figure.","The analysis masks 23% of pixels as inherent defects or light leaks; those pixels are excluded from every statistic, so the 'no degradation' conclusion applies only to the selected 77% subarray. A follow-up test that tracks the excluded pixels would reveal whether latent defects grow with age.","The observed suppression of random telegraph noise hints that aging could partially heal radiation-induced defects, suggesting a combined radiation-plus-aging experiment (irradiate, then age, then compare defect densities) as a natural next test.","The protocol developed here—repeated dark and $^{55}$Fe exposures at two temperatures over two years—could serve as a standard pre-flight qualification for CMOS X-ray detectors on future missions."],"forward_implications":["The sensor can support space missions requiring continuous operation for several hundred days without performance loss, as exemplified by the currently flying LEIA and EP-WXT instruments that use this sensor type.","The number of degraded pixels remains tiny—roughly 40 bias-varied and 10 noise-varied pixels after 610 days—and these are non-clustering, so they can be flagged and excluded without affecting the array's overall performance.","Aging appears to suppress dark current and random telegraph signal in some pixels, suggesting that extended operation may anneal certain defect sites rather than create new ones.","Because inherent aging is negligible, any mild performance degradation observed in orbit should be attributed to radiation damage or other environmental factors, not to the sensor's own decay.","The predicted gain drift of $0.73\\%$ over 3 years can be compensated by routine annual gain calibration, removing a potential concern for long-duration X-ray surveys."],"supporting_citations":[{"why":"Provides the standard data extraction and event grading procedures used to build X-ray spectra and derive conversion gain and energy resolution.","marker":"[10]"},{"why":"Supplies the first-order kinetic model $G = G_0 \\exp(-kt)$ used to fit the gain evolution and project 3- and 10-year degeneration.","marker":"[34]"},{"why":"Describes the EP4K sensor, the same device type tested here, including its aluminum coating and X-ray performance.","marker":"[13]"},{"why":"Proton irradiation study whose observed degradation is compared against the aging result, establishing that aging cannot account for radiation damage.","marker":"[14]"},{"why":"Total ionizing dose irradiation study used in the same comparative way to separate aging effects from radiation effects.","marker":"[15]"},{"why":"Correlogram method for examining crosstalk, part of the standard data-processing pipeline referenced for the gain measurements.","marker":"[6]"},{"why":"Describes the camera used to operate the CMOS sensor at 20 Hz with real-time temperature control and data readout.","marker":"[9]"}],"fun_headline_variants":["610-day aging test leaves X-ray CMOS sensor intact","Space-grade X-ray CMOS sensor remains stable after 610 days","X-ray CMOS detector: no degradation after 610-day test","Projected gain loss of only 0.73% in 3 years for X-ray CMOS","610-day test: X-ray CMOS sensor performance stays flat"],"cache_read_input_tokens":12416,"weakest_assumption_plain":"The load-bearing premise is that the 23% of pixels excluded as inherently defective or light-leaking are static and do not mask aging damage, and that the gain decay fitted from 427 days at $-30^\\circ$C follows the same exponential for ten years.","fun_headline_variants_meta":{"raw":{"variants":["610-day aging test leaves X-ray CMOS sensor intact","Space-grade X-ray CMOS sensor remains stable after 610 days","X-ray CMOS detector: no degradation after 610-day test","Projected gain loss of only 0.73% in 3 years for X-ray CMOS","610-day test: X-ray CMOS sensor performance stays flat"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000735,"raw_usage":{"total_tokens":3287,"prompt_tokens":947,"completion_tokens":2340,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":563,"completion_tokens_details":{"reasoning_tokens":2250}},"tokens_in":563,"tokens_out":2340,"duration_ms":12666,"temperature":1.0,"reasoning_tokens":2250,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T11:50:50.543127+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run a 610-day aging test on the same sensor batch without masking any pixels, tracking every pixel's bias map and noise; if noticeably more than the reported ~50 pixels develop bias shifts above 10 DN or noise shifts above 15 $e^-$, or if the measured gain after several years deviates from the fitted exponential beyond the quoted $0.73\\%$–$2.41\\%$ band, the stability claim would need revision.","supporting_citations":[{"cited_title":"X-ray performance of a customized large-format scientifc CMOS detector","cited_arxiv_id":"2209.15295","evidence_quote":"Provides the standard data extraction and event grading procedures used to build X-ray spectra and derive conversion gain and energy resolution."},{"cited_title":"De Bettignies, J","cited_arxiv_id":null,"evidence_quote":"Supplies the first-order kinetic model $G = G_0 \\exp(-kt)$ used to fit the gain evolution and project 3- and 10-year degeneration."},{"cited_title":"An Aluminum-coated sCMOS sensor for X-Ray Astronomy","cited_arxiv_id":"2310.14887","evidence_quote":"Describes the EP4K sensor, the same device type tested here, including its aluminum coating and X-ray performance."},{"cited_title":"Radiation effects on scientific CMOS sensors for X-ray astronomy: I. proton irradiation","cited_arxiv_id":"2312.01851","evidence_quote":"Proton irradiation study whose observed degradation is compared against the aging result, establishing that aging cannot account for radiation damage."},{"cited_title":"Ling, et al., A correlogram method to examine the crosstalk of sCMOS sensors, Journal of Instrumentation 16 (3) (2021) P03018","cited_arxiv_id":null,"evidence_quote":"Correlogram method for examining crosstalk, part of the standard data-processing pipeline referenced for the gain measurements."}],"review_version":1}