{"id":"1cd11772-8800-4f5e-a5bf-50c8c4736b37","arxiv_id":"2607.19615","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"SiSeRO RNDR detectors are predicted to hold their 1/√N noise reduction to at least 1000 cycles at 173 K, with thermal-leakage and impact-ionization noise constrained below the current detection threshold.","lead":"A Monte Carlo model of SiSeRO X-ray detector readout predicts that two dark-current-like noise mechanisms — trap-assisted thermal leakage and impact ionization — stay below the measurable floor at 173 K for up to 1000 repetitive readout (RNDR) cycles. The paper also sets statistical upper limits on this excess charge from first-generation devices. Read it if you want to know whether SiSeRO detectors can meet the sub-electron-noise needs of future X-ray observatories.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Impact-ionization 'rule out' in §2.3 depends on an unquantified correction factor K and on 200-cycle data that live in an unpublished companion paper; the central exclusion is not reproducible.","rationale":"The reader's CONDITIONAL verdict is appropriate. My independent reading identifies the same load-bearing weakness: the impact-ionization exclusion is not tied to a measurable physical electric field because K is unquantified, and the only empirical constraint used to calibrate or exclude the model is deferred to an in-preparation companion paper. This is not a reason to reject the paper—the model is coherent, the Monte Carlo approach is reasonable, and the thermal-leakage prediction at 173 K is robust to modest trap-density assumptions—but it prevents an independent reader from verifying the central 'rule out' statement. The concrete test above would settle whether the exclusion survives once actual device voltages and the 200-cycle dataset are made available. Until then, conditional acceptance with the requirement that K and the companion analysis be provided is the right posture. I agree with the reader's flagged concerns and see no additional internal inconsistency or evidence of bad faith.","tokens_in":9193,"tokens_out":6647,"duration_ms":70352,"concrete_test":"Recompute §2.3 impact-ionization noise from the device's actual operating point: report measured source, drain, and gate voltages, channel length, and threshold voltage; set K=1 (or use a TCAD-derived field profile); and generate the predicted cumulative-noise curve at N=100–200. Overlay the Pan et al. [24] 200-cycle dataset and its 2σ upper limit (9.61 e−/pixel/s). If the predicted excess stays below the upper limit, the exclusion is confirmed with a physical field; if it crosses it, the 'rule out' claim fails and K must be constrained.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that impact ionization is ruled out rests on two unpinned links. First, Eq. 4 introduces K as a fudge multiplier on the drain field because 'it is not possible to calculate this value accurately using a simple linear model,' but K is never assigned a value or range. The Fig. 7 curves are therefore labeled only by effective field (e.g., 4.74×10^4 V/cm), and the paper never reports the source/drain/gate voltages, channel length, or threshold voltage that would let a reader connect those effective fields to the actual CCID-93 device. 'Known operating voltages' are asserted, not given. Second, the experimental anchor that 'rules out' the high-field curves is the 100–200 cycle RNDR dataset whose F-test, bootstrap, and 2σ upper limit on excess slope live entirely in Pan et al. [24], a preprint listed 'in preparation.' The logic itself is sound—if a high effective field produced a detectable excess by N=200, the data would show it—but the sensitivity of that 200-cycle test is not documented here. Without the companion analysis one cannot compute the minimum excluded excess rate, so the conclusion 'we can rule out such high electric field' is not checkable. The thermal-leakage claim is less exposed: at 173 K SRH rates are exponentially small even for the assumed near-maximal trap density, so that part of the claim is robust. The impact-ionization exclusion, by contrast, depends on an unquantified parameter and an unavailable dataset.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops a Monte Carlo noise model for first-generation SiSeRO devices operated in repetitive non-destructive readout (RNDR) mode. The model includes thermal read noise, 1/f noise, trap-assisted thermal leakage via Shockley-Read-Hall generation, and impact ionization in the output MOSFET. The authors claim that for the first-generation CCID-93 device at 173 K, neither thermal leakage nor impact ionization produces measurable excess dark-current-like noise up to at least 1000 RNDR cycles, and that a high electric-field impact-ionization contribution can be ruled out on the basis of no observed deviation from the 1/sqrt(N) trend in 100–200 cycle laboratory data. The empirical anchor for that exclusion is deferred to an in-preparation companion paper (Pan et al. 2026, ref. [24]).","tokens_in":9511,"tokens_out":4167,"duration_ms":40417,"significance":"If the claims hold, the paper provides useful evidence that RNDR readout of SiSeRO devices can be extended well beyond the currently demonstrated 200 cycles without encountering a fundamental noise floor from thermal generation or impact ionization in the output stage. The modeling is largely built on standard, externally parameterized physical models: the SRH generation rate and the Van Overstraeten-De Man impact-ionization coefficients. The authors are also transparent about the main caveats, explicitly noting that the correction factor K in Eq. (4) cannot be accurately calculated with their simple linear model and that the 100–200 cycle empirical constraint is deferred to ref. [24]. These strengths make the overall approach credible. However, the central exclusion for impact ionization is not independently checkable from this manuscript, because it depends on an unquantified parameter and on an unpublished dataset. The significance of the paper is therefore conditional on the companion analysis and on a quantitative treatment of K.","major_comments":[{"comment":"The impact-ionization exclusion rests on the correction factor K, introduced because the electric field near the drain cannot be calculated accurately with a simple linear model. K is never assigned a value or a plausible range, and the Fig. 7 curves are labeled only by effective field. The manuscript states that known operating voltages are used, but the source/drain/gate voltages, channel length, and threshold voltage are not given. Consequently the reader cannot connect the effective fields such as 4.74×10^4 V/cm to the actual device or assess how sensitive the exclusion is to K. Please provide the K value or range, the device parameters, and either a sensitivity analysis or a calibration of K against measured voltages. Without this, the statement “we can rule out such high electric field” is not reproducible.","section":"§2.3, Eq. (4), Fig. 7"},{"comment":"The empirical constraint that anchors the impact-ionization exclusion and the upper limits of Fig. 8 is not contained in this paper. The F-test, bootstrap analysis, and 2σ upper limit are all deferred to Pan et al. 2026 (in prep.). The logical structure is reasonable: if a high effective field produced a detectable excess by 200 cycles, the data would show it. But the sensitivity of the 200-cycle test is not documented here, so the minimum excluded excess-slope cannot be computed by the reader. If the companion analysis is incorrect or contains different systematics, the constraints could collapse. Please include the observed noise versus cycle data, the F-test and bootstrap distributions, and a statement of the minimum detectable excess rate, or clearly mark the conclusions as conditional on ref. [24].","section":"§2.4, Fig. 8, ref. [24]"},{"comment":"The thermal-leakage prediction uses N_trap = 1×10^12 cm^-3 as a “hypothetical near-maximal” value, but no justification or calibration is given for this choice. This is not fatal for the main thermal-leakage claim, because the SRH rate at 173 K is exponentially small and the conclusion is robust to large variations in trap density. However, if this near-maximal value is meant to represent an upper constraint rather than merely a plausible scenario, the manuscript should cite a source or derive the value from the measured dark-current floor. Otherwise the wording “near-maximal” is unsupported.","section":"§2.2, Fig. 4"}],"minor_comments":[{"comment":"The heading “T rap-Assisted Thermal Leakage” contains a formatting artifact (“T rap”).","section":"§2.2"},{"comment":"The text defines E_c as the conduction-band edge, but Eq. (1) uses E_C. Please use consistent notation for E_C, E_c, and E_t, and define E_t explicitly as the trap energy level.","section":"Eq. (1)"},{"comment":"The sentence mentioning “the known length of the MOSFET channel, the operating voltages of the source, drain, and gate, and a nominal threshold voltage” would benefit from an explicit table of these parameters. Without them, the model is not reproducible even aside from K.","section":"§2.3"},{"comment":"The caption states that the probability increases with the electric field, but it does not mention that the curves depend directly on the unquantified K. Adding a sentence noting the dependence for a representative K value would help the reader interpret the figure.","section":"Fig. 6 caption"},{"comment":"Ref. [24] has the same title as [23] except for lowercasing “sisero.” Please clarify the relationship between the two papers and update the status if the companion paper has become available.","section":"References [23] and [24]"},{"comment":"The Monte Carlo simulation setup would benefit from reporting the number of realizations or pixels used for the standard-deviation estimates in Figs. 2, 4, and 7. This affects the apparent scatter of the simulated curves.","section":"§2.1"}],"recommendation":"major_revision","confidential_remarks":"This is a promising modeling paper with a clear physical framework, but the central exclusion claim is currently not checkable: the empirical anchor is in an unpublished companion paper and the impact-ionization model contains an unquantified K. I would not reject outright because both issues are fixable within the manuscript's scope: supply the missing empirical analysis or explicitly mark the claim as conditional, quantify K, and give the device operating parameters. If the companion paper is submitted concurrently, the editor should require that it be posted or included so that the referee and readers can verify the 100–200 cycle constraint."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: useful, coherent modeling paper for a detector that deserves scrutiny, but the main exclusion it advertises — no impact-ionization excess noise at 173 K out to 1000 RNDR cycles — is not checkable from the text. The thermal-leakage half of the claim is robust; the impact-ionization half depends on an unquantified correction factor K and on 100–200 cycle laboratory data deferred to an in-preparation companion paper.\n\nWhat is actually new: first model-based look at RNDR noise beyond the measured 200 cycles for SiSeRO, combining SRH thermal leakage and Van Overstraeten–De Man impact ionization inside a Monte Carlo of the RNDR waveform. Equations (1)–(8) are standard and correctly assembled. At 173 K the thermal-leakage prediction is not fragile: SRH emission is exponentially small even for the deliberately high trap density used, so “negligible contribution” holds. The paper is also appropriately modest about the current data, pointing to Pan et al. for the F-test/bootstrap analysis and calling the situation “encouraging” rather than proven.\n\nSoft spots. Biggest: the 200-cycle anchor is in another manuscript. Fig. 8's upper limits (2.26–9.61 e−/pixel/s) are reported, but the datasets, fits, and bootstrap distributions are not, so a reader cannot check whether the 200-cycle test would have detected a real excess. That is the difference between a claim and a checkable claim. Second, K in Eq. 4 is introduced because the drain field is nonlinear, but no value or range is given, and the actual source/drain/gate voltages are never listed. The effective-field labels in Fig. 7 cannot be tied to the CCID-93 device; if K is a free dial, the “we can rule out” statement needs a sensitivity scan. Third, the “near-maximal” trap density is not defined; the conclusion is fine as a stress test, but the basis for calling it near-maximal should be stated. The unspecified 1/f pulse amplitude is minor because 1/f alone does not alter the cumulative noise trend.\n\nBottom line: read this if you are evaluating SiSeRO or building RNDR noise models. The physics is standard, the logic is coherent, and the thermal-leakage part holds up. But I would not rely on the impact-ionization exclusion until the companion analysis is public and K is pinned down. I would send it to peer review with the explicit condition that the underlying data be made available.","headline":"Useful, coherent RNDR noise model for SiSeRO, but the impact-ionization exclusion depends on an unpublished dataset and an unquantified K factor; the thermal-leakage claim holds up.","tokens_in":10183,"tokens_out":5494,"would_cite":false,"duration_ms":50306,"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":"At 173 K, SiSeRO repetitive readout gains no measurable noise from thermal leakage or impact ionization, preserving the 1/sqrt(N) averaging trend through 1000 modeled cycles.","keywords":["SiSeRO","repetitive non-destructive readout","noise modeling","thermal leakage","impact ionization","single-photon detection","low-noise CCDs","X-ray detectors"],"falsifier":"Check the companion paper's F-test and bootstrap: if the slope of excess charge vs. cycle is not consistent with zero, the impact-ionization exclusion is void. Alternatively, run the same RNDR sequence at a higher temperature, say 220 K, where the model predicts thermal-leakage noise; a measured deviation from 1/sqrt(N) would confirm the SRH parameters, and its absence would strengthen the bound.","tokens_in":9010,"feed_emoji":"🔭","tokens_out":8032,"duration_ms":63375,"temperature":0.7,"pith_summary":"This paper argues that two mechanisms that could add dark-current-like noise to SiSeRO detectors — thermal generation from crystal traps and impact ionization in the output transistor — do not measurably perturb repetitive non-destructive readout (RNDR) at the operating temperature of 173 K. The authors build a Monte Carlo model of the RNDR waveform, injecting electron emission from Shockley-Read-Hall trap physics and Van Overstraeten-De Man impact ionization as Poisson processes, and simulate cumulative read noise out to 1000 cycles. The model predicts that thermal leakage is negligible at 173 K and that impact ionization would require drain electric fields high enough to be excluded by existing laboratory measurements up to 200 cycles. If right, this clears the low-noise path for single-photon X-ray and UV detectors that rely on averaging many readout cycles.","feed_headline":"At 173 K, SiSeRO readout shows no excess noise over 1000 cycles","feed_subtitle":"Model rules out thermal leakage and impact ionization, clearing the path to ultra-low-noise single-photon detectors.","key_machinery":"The argument rides on two probabilistic electron-generation mechanisms and the Monte Carlo waveform that ties them to the observable. Trap-assisted thermal leakage is described by Shockley-Read-Hall emission: Remit = sigma v_th n_c exp[-(Ec-Et)/kT], with the probability of n electrons following a Poisson distribution whose mean scales with trap density, depletion volume, and integration time. Impact ionization is described by the Van Overstraeten-De Man ionization coefficient alpha = A exp(-B/(K E)), where K is a correction factor for the poorly known drain electric field. The carrying observable is the cumulative RNDR noise, which in an ideal device falls as 1/sqrt(N); any time-dependent ex","core_discovery":"The central claim is that at 173 K, first-generation SiSeRO devices exhibit no excess dark-current-like noise from trap-assisted thermal leakage or impact ionization over the full range of RNDR cycles examined. The evidence is a Monte Carlo simulation of the RNDR process that models transistor thermal noise and 1/f noise, then adds Poisson-distributed electron emission from two physical mechanisms: Shockley-Read-Hall generation at mid-gap traps in the depleted internal gate, and impact ionization in the high-field region near the MOSFET drain using Overstraeten impact coefficients. The thermal-leakage model predicts negligible contribution at 173 K even for a near-maximal trap density of 1e1","pith_inferences":["The model's temperature dependence implies that, for a fixed noise budget, reducing trap density in future fabrication runs could allow the same RNDR performance at higher operating temperatures, easing cryostat requirements — a trade-off the paper does not explicitly quantify.","The impact-ionization exclusion is only as strong as the assumed channel geometry; a direct measurement or 3D field simulation of the drain region would either tighten the bound or force a revised correction factor K, making the exclusion testable before second-generation devices ship.","The reported upper limit of 9.61 e-/pixel/s is orders of magnitude above the dark-current goals of planned Earth-like exoplanet coronagraphs, so meeting those goals depends on suppressing image-area dark current, which this paper deliberately does not model.","A natural next experiment is to repeat the RNDR measurement at higher temperatures (for example 200–220 K), where the model predicts thermal-leakage noise should appear; the cycle count at which the 1/sqrt(N) trend breaks would calibrate the actual trap density and validate the Shockley-Read-Hall parameters."],"forward_implications":["If correct, the 1/sqrt(N) noise-averaging benefit of RNDR extends beyond 200 cycles at 173 K, making it feasible to push SiSeRO devices to sub-electron noise by running hundreds of readout cycles.","The 2-sigma bound of 9.61 e-/pixel/s on excess charge gives a quantitative floor that second-generation devices can be tested against.","Since the model is generic, it can be applied to other RNDR architectures, such as Skipper CCDs, DEPFETs, and CMOS sensors, to predict where their dark-current-like limits appear.","The same model warns that higher operating temperatures will bring thermal-leakage noise into the readout, so temperature control is a key lever for long RNDR chains."],"fun_headline_variants":["SiSeRO at 173 K: zero excess noise over 1000 RNDR cycles","No trap leakage or impact ionization noise in SiSeRO at 173 K","Model: SiSeRO has zero excess noise from traps or ionization at 173 K","At 173 K, SiSeRO rules out thermal leakage and impact ionization noise","SiSeRO at 173 K: no excess noise from thermal leakage or impact ionization"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The constraints collapse if the still-unpublished 200-cycle laboratory dataset does not actually show a clean 1/sqrt(N) trend, or if the real trap density exceeds the assumed 'near-maximal' level used in the thermal-leakage model.","fun_headline_variants_meta":{"raw":{"variants":["SiSeRO at 173 K: zero excess noise over 1000 RNDR cycles","No trap leakage or impact ionization noise in SiSeRO at 173 K","Model: SiSeRO has zero excess noise from traps or ionization at 173 K","At 173 K, SiSeRO rules out thermal leakage and impact ionization noise","SiSeRO at 173 K: no excess noise from thermal leakage or impact ionization"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000609,"raw_usage":{"total_tokens":2649,"prompt_tokens":696,"completion_tokens":1953,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":440,"completion_tokens_details":{"reasoning_tokens":1843}},"tokens_in":440,"tokens_out":1953,"duration_ms":11665,"temperature":1.0,"reasoning_tokens":1843,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T12:13:04.011792+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Check the companion paper's F-test and bootstrap: if the slope of excess charge vs. cycle is not consistent with zero, the impact-ionization exclusion is void. Alternatively, run the same RNDR sequence at a higher temperature, say 220 K, where the model predicts thermal-leakage noise; a measured deviation from 1/sqrt(N) would confirm the SRH parameters, and its absence would strengthen the bound.","supporting_citations":[],"review_version":1}