{"id":"7621e10b-5861-489a-bc94-b60a0ed33a30","arxiv_id":"2506.10823","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"A TCAD plus Allpix2 simulation chain, averaging the CMS and Mandic trapping models, reproduces irradiated pixel charge collection and predicts ITk planar modules remain fully efficient at HL-LHC fluences.","lead":"This paper tests four radiation damage trapping models in TCAD and Allpix2 simulations against CMS testbeam data for irradiated silicon pixel sensors, then uses the best models to predict charge collection for ATLAS's future ITk pixel modules at HL-LHC fluences. The predictions suggest the modules should collect enough charge to stay fully efficient even at end-of-life fluences, if the simulation approach is trusted.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Single-point validation (150 µm, 2.1e15 neq/cm²) cannot support extrapolation to 100 µm sensors and fluences up to 3.5e15; a second fluence check is needed.","rationale":"The reader's weakest_assumption identifies exactly the concern I find most load-bearing: the validation at one fluence and one thickness is insufficient to guarantee the TCAD radiation damage model and the averaged CMS/Mandić trapping model for the full ITk range. My analysis confirms this by noting that ref [16] already contains data at Φ = 1×10^16 neq/cm², which the paper does not use for validation despite explicitly mentioning this fluence in Section 2.2.1. A concrete, low-cost check is to simulate that second fluence and compare to the published data. This would directly test the fluence-scaling assumption and would either support or undermine the extrapolation to ITk. I do not find a separate, more fundamental flaw in the simulation chain: the unirradiated normalization agrees within 1%, the voltage-dependent behavior at the single validation point is reproduced, and the full-efficiency claim, while based on an effective-threshold heuristic rather than a full efficiency simulation, is a secondary concern because the predicted MPVs are several times the effective threshold. The recommended verdict remains CONDITIONAL: the paper should either perform the second-fluence validation or explicitly reframe the ITk predictions as an unvalidated extrapolation. Since the reader already reached CONDITIONAL, my read does not change the verdict.","tokens_in":9861,"tokens_out":5072,"duration_ms":52469,"concrete_test":"Re-run the exact simulation setup (LHCb TCAD radiation damage model, Allpix2 with the average of CMS and Mandić trapping models, same digitization settings) for the 150 µm passive CMOS sensor at Φ = 1×10^16 neq/cm², and compare the simulated MPV versus bias voltage curve to the corresponding data reported in ref [16] (Glessgen et al.), which the paper already cites as containing measurements at this fluence. If the simulated curve deviates from the data by more than the semi-difference uncertainty used in Figures 5–9, the transferability of the model to higher fluences is not established, and the ITk predictions at 3.5×10^15 must be reframed as unvalidated projections.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The entire extrapolation to ATLAS ITk planar sensors rests on one validation point: a 150 µm passive CMOS sensor at Φ = 2.1×10^15 neq/cm² (Section 3.2). From this single point, the paper concludes in Section 4 that 100 µm and 150 µm ITk sensors will collect enough charge for full efficiency at all fluences up to 3.5×10^15 neq/cm² and at voltages well below 600 V. This is a two-directional extrapolation: sensor thickness (150→100 µm) and fluence (2.1→3.5×10^15). Neither direction is independently validated. The LHCb TCAD radiation damage model is chosen because it reproduces the electric-field-driven voltage dependence at this one fluence, but its defect introduction rates and charge states have not been shown to scale correctly to other fluences or to the different processing technology of ITk planar sensors. The trapping model is the post-hoc average of CMS and Mandić models, justified only by bracketing the data at 2.1×10^15; whether this bracket remains valid at 3.5×10^15 is untested. Section 2.2.1 notes that ref [16] reports measurements at 1×10^16 neq/cm², yet no validation against those data is presented. If the radiation damage model or the trapping-model average mispredicts the fluence dependence, the predicted MPV values—and hence the 'full efficiency' conclusion—could shift significantly at the highest ITk fluences. The paper's confidence in moderate-voltage operation depends directly on this unvalidated extrapolation.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper validates a combined TCAD (Silvaco) and Allpix2 Monte Carlo simulation chain for irradiated hybrid planar pixel sensors. Validation is performed against CMS passive CMOS n-on-p sensors (150 µm thick, 50×50 µm² pitch) irradiated to Φ = 2.1×10^15 neq/cm², comparing cluster-charge MPV versus bias voltage for four trapping models (CMS, Ljubljana, Atlas, Mandić). The authors find that the CMS and Mandić models bracket the data and adopt their average, with the semi-difference as uncertainty. They then apply this validated setup to predict the collected charge of ATLAS ITk planar sensors (100 µm and 150 µm thick) at fluences from 1.5×10^15 to 3.5×10^15 neq/cm², concluding that enough charge is collected for full efficiency at all considered fluences and voltages, and that operation at moderate bias voltages (well below 600 V) should be possible.","tokens_in":10178,"tokens_out":2782,"duration_ms":32951,"significance":"If the predictions are correct, the paper would provide valuable guidance for operating the ATLAS ITk pixel detector at HL-LHC fluences and for preparing radiation-damage-aware Monte Carlo simulations. The manuscript has genuine strengths: the unirradiated simulation reproduces the data MPV within 1% (Section 3.1); the voltage dependence of the collected charge in the irradiated case is reproduced by all four trapping models, which independently supports the TCAD electric-field modelling; and the trapping-model parameters are taken from external, published fits rather than fitted to the validation data. However, the central extrapolation to ITk sensors rests on a single validation point, and the adopted uncertainty band is constructed post hoc. These issues currently limit the strength of the conclusions.","major_comments":[{"comment":"The validation is performed on a single sensor (150 µm thick, n-on-p passive CMOS, Φ = 2.1×10^15 neq/cm²), and this single point is used to justify extrapolation in Section 4 to 100 µm thick \"thin\" ITk sensors and to fluences up to 3.5×10^15 neq/cm². This is a two-directional extrapolation in both thickness and fluence, and no independent data are used to test either direction. Reference [16] reports measurements at other fluences, including up to 1×10^16 neq/cm², and the paper itself notes these data in Section 2.2.1, but they are not used for validation. The claim in Section 4 that \"at all fluences and voltages and for both thicknesses enough charge is collected to have full efficiency\" is therefore not supported by the evidence presented.","section":"Section 3.2 and Section 4"},{"comment":"The averaging of the CMS and Mandić trapping models, with the uncertainty taken as their semi-difference, is decided after inspecting the data at one fluence and one thickness. This makes the uncertainty band a post hoc envelope rather than a predictive uncertainty. The two models have different functional forms (linear in Φ for CMS, power-law for Mandić), so their difference at Φ = 2.1×10^15 neq/cm² does not guarantee that they will bracket the true collected charge at Φ = 3.5×10^15 neq/cm² or for 100 µm sensors. The fluence-dependence of the uncertainty band in Figures 6–9 is therefore not justified and should be either validated against additional data or presented as an assumption.","section":"Section 4"},{"comment":"The statement \"At all fluences and voltages and for both thicknesses enough charge is collected to have full efficiency\" is based on comparing the simulated MPV of the cluster charge to an \"effective threshold\" of about 1.3 ke, not on a simulated hit efficiency. The simulation uses 120 GeV pions at normal incidence with essentially no charge sharing, which is an optimistic scenario. For inclined tracks, clusters with more charge sharing, or with the actual readout threshold and cluster algorithm, the hit efficiency could be lower even if the MPV exceeds the threshold by a wide margin. The paper should either compute a realistic hit efficiency or qualify the \"full efficiency\" claim to say that the collected charge is expected to be well above threshold under the idealized conditions simulated.","section":"Section 4"}],"minor_comments":[{"comment":"Table 1 lists a digitization threshold of 600 e, whereas Section 3.2 states that \"a hit threshold of 1.24 ke was used in both data and simulations.\" Please clarify which threshold applies to the validation and which to the ITk predictions, and whether the difference is intentional.","section":"Section 2.2 and Section 3.2"},{"comment":"The sentence \"It is interesting to notice that the Mandić and CMS trapping models essentially bracket the data in the 'depleted' region\" is followed by a statement that Mandić is closer to data, but then the average is adopted. The rationale for preferring the average over the single better-performing Mandić model is not fully articulated; please spell out the reasoning.","section":"Section 3.2"},{"comment":"The definition of the \"effective threshold\" as the quadratic sum of the threshold with 5 times the quadratic sum of noise and threshold dispersion should be justified; the factor of 5 appears arbitrary and is not motivated in the text.","section":"Section 4"},{"comment":"The vertical axis label in Figure 8 reads \"MIP [ke-]\", which is inconsistent with all other figures that use \"MPV [ke]\"; please correct.","section":"Figure 8"},{"comment":"There is a typographical spacing error in \"CMStrackergroupreported\" near the end of Section 3.1; the text should read \"CMS tracker group reported\".","section":"Section 3.1"}],"recommendation":"major_revision","confidential_remarks":"The paper is essentially a proceedings-style report and is reasonable in scope, but the single-point validation is the key weakness. The authors should be asked either to add validation at a second fluence (or at least a second thickness) or to substantially weaken the operational claims. The post hoc averaging of two trapping models is also problematic for the uncertainty quantification. I see no grounds for rejection, provided these concerns are addressed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"What you should know: this is a genuinely useful comparison of four trapping models against the CMS passive-CMOS RD53A testbeam data at 2.1e15 neq/cm2, and it makes a credible case that a TCAD field model from LHCb Velo can reproduce the voltage dependence of collected charge. The unirradiated normalization agrees with data at the percent level, and all four models get the turn-on shape right, which is a real success for the electric-field modeling. The paper is honest enough to admit that no single model wins: it uses the average of the CMS and Mandić models, with their semi-difference as a band, which is a reasonable engineering choice.\n\nThe soft spots are real, though not fatal. The validation is at one fluence and one thickness. Everything about the ITk predictions for 100 um sensors and fluences up to 3.5e15 is extrapolation. The paper cites 1e16 data already, so it would have been easy to add a second check; without it, the 'at all fluences and voltages... full efficiency' sentence should be downgraded to a projection. Also, the data points in Figure 4 appear without error bars; the chi2 comparison is hard to interpret without them. The 'effective threshold' argument is a reasonable proxy, but it is not a real hit-efficiency simulation; the MPVs are far enough above threshold that the conclusion is plausible, but it is not demonstrated. And the text has a confusing pronoun: it says 'the latter' underestimates, which is the opposite of what the plots show; that should be corrected.\n\nWho gets value: detector simulation people, ITk operations groups, and anyone building radiation-damage digitizers. It is a competent engineering paper with a useful dataset comparison, not a fundamental advance. I would send it to peer review with a request for a second validation point or a careful softening of the extrapolation claims. The underlying simulation chain looks sound; the paper just needs to match its conclusions to its evidence.","headline":"A probing but single-point validation of a TCAD/Allpix2 chain; the ITk 'full efficiency' claim outruns the evidence.","tokens_in":10754,"tokens_out":4581,"would_cite":true,"duration_ms":50123,"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":"Combined TCAD and Monte Carlo simulations predict ATLAS ITk planar pixel modules will collect enough charge at HL-LHC fluences to maintain full hit efficiency.","keywords":["silicon pixel detectors","radiation damage","charge trapping","TCAD","Allpix2","HL-LHC","ATLAS ITk","Monte Carlo simulation"],"falsifier":"Measure the cluster charge MPV of an ATLAS ITkPixV2 planar module (100 µm or 150 µm thick) irradiated to 3.5e15 neq/cm2 and biased at 400 V. If the measured MPV is below the effective threshold of about 1.3 ke, or deviates by more than the predicted uncertainty band from the simulated 5 ke (thin) or 6 ke (thick), the transferability of the validated models to the full ITk fluence range is ruled out.","tokens_in":9604,"feed_emoji":"⚛️","tokens_out":12918,"duration_ms":80151,"temperature":0.7,"pith_summary":"The paper addresses the main threat to tracking performance at the HL-LHC: loss of signal charge in silicon pixel detectors from radiation-induced carrier trapping. It validates a two-stage simulation chain—TCAD device simulation for the electric field and weighting potential, followed by the Allpix2 Monte Carlo for charge collection—against test-beam data from a 150 µm thick n-on-p passive CMOS sensor irradiated to a fluence of 2.1e15 neq/cm2. Comparing four trapping models, the paper finds that the CMS and Mandić parameterizations bracket the data, and their average reproduces both the voltage dependence and the saturated value of the most probable collected charge. Applying this validated chain to the ATLAS ITk planar pixel sensors (100 µm and 150 µm thick) at fluences up to 3.5e15 neq/cm2, the paper concludes that collected charge remains far above threshold at all simulated voltages and thicknesses, so full hit efficiency should be achieved without running at the maximum allowed voltage.","feed_headline":"Pixel modules keep full hit efficiency at HL-LHC, simulations show","feed_subtitle":"Validated TCAD+Monte Carlo chain predicts enough charge for ATLAS ITk planar sensors without running at maximum voltage.","key_machinery":"The load-bearing object is a two-stage simulation chain. Silvaco TCAD produces 3D maps of the electric field and weighting potential for a 50x50 µm2 n-on-p pixel cell, using the LHCb VELO radiation damage model to describe defect states created by irradiation. Allpix2 v3.2.0 then transports charge from 120 GeV pions impinging at normal incidence, applying a mobility model, digitization thresholds, and a carrier-trapping model. The chosen trapping model is the average of the CMS and Mandić parameterizations, whose predictions bracket the measured data; the semi-difference of the two is assigned as the systematic uncertainty.","core_discovery":"The central claim is that, at all fluences and voltages simulated and for both sensor thicknesses, enough charge is collected to have full efficiency for normally incident particles: for thin sensors the predicted most probable charge is above 5 ke at 400 V even at the largest fluence of 3.5e15 neq/cm2, and for thick sensors it is always above 6 ke. This follows from the validation step, where the combination of the LHCb radiation damage model (two deep acceptors, one donor) in TCAD and the average of the CMS and Mandić trapping models matches the measured cluster charge as a function of bias voltage for an irradiated 150 µm sensor, correctly predicting the voltage at which saturation begins. Consequently, the paper states that it will not be necessary to operate the ITk pixel modules at the maximum 600 V bias.","pith_inferences":["The validation rests on a single fluence–thickness point; measuring a second point at 3.5e15 neq/cm2 on a 100 µm sensor would directly test whether the LHCb damage model and the averaged trapping model transfer across the full ITk fluence range.","The simulations use particles at normal incidence, which suppresses charge sharing; real ATLAS tracks arrive at a range of angles, so per-pixel collected charge will be lower than simulated, and the 'full efficiency' claim is safest for cluster-level charge.","The success of averaging two bracketing trapping models suggests a single unified trapping-rate parameterization covering 1e15 to 1e17 neq/cm2 could simplify future simulations.","The same validated chain could be applied to 3D sensors, though the paper notes their different geometry and fluence range (up to 1.6e16 neq/cm2) require a dedicated study."],"forward_implications":["ATLAS ITk planar pixel modules can be operated at 300–400 V bias instead of the maximum 600 V, reducing power draw and easing the cooling budget.","The validated chain can produce radiation-damage look-up tables for the ATLAS Monte Carlo generator, allowing tracking and vertexing algorithms to be tested under realistic HL-LHC charge-collection conditions.","Even after 40% or more signal loss, the collected charge is predicted to remain several times above the effective threshold of about 1.3 ke, implying high hit efficiency throughout the detector lifetime.","The same simulation setup can be extended to future sensor designs or readout chips once new irradiated test-beam data become available."],"supporting_citations":[{"why":"Supplies the irradiated-device test-beam data (cluster charge MPV vs bias at 2.1e15 neq/cm2) used to validate the simulations.","marker":"[16]"},{"why":"Provides the unirradiated sensor characterization and normalization point used to calibrate the simulation charge scale.","marker":"[15]"},{"why":"Supplies the LHCb silicon bulk radiation damage model implemented in the TCAD device simulations.","marker":"[27]"},{"why":"Defines the Mandić trapping model, one of the two bracketing models whose average is used for predictions.","marker":"[30]"},{"why":"Corrects the Mandić model; the paper notes the Allpix2 implementation had to be reimplemented to match the published version.","marker":"[32]"},{"why":"Provides the underlying trapping-rate measurements on which the CMS trapping model is based.","marker":"[31]"},{"why":"Describes the Allpix2 Monte Carlo framework that carries the charge-transport and digitization simulation.","marker":"[11]"}],"fun_headline_variants":["Pixel sensors stay efficient at HL-LHC radiation, simulations confirm","Simulations show full efficiency for ITk pixels at all fluences","Planar pixel modules keep hit efficiency without max voltage","Validated simulation predicts full pixel efficiency at HL-LHC","No need for 600V: pixel modules efficient at HL-LHC fluences"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The validation is performed at one fluence (2.1e15 neq/cm2) on one sensor thickness (150 µm), and this single point is taken as sufficient to guarantee the LHCb radiation damage model and the averaged CMS–Mandić trapping model are accurate for 100 µm and 150 µm ATLAS ITk planar sensors at fluences from 1.5 to 3.5e15 neq/cm2.","fun_headline_variants_meta":{"raw":{"variants":["Pixel sensors stay efficient at HL-LHC radiation, simulations confirm","Simulations show full efficiency for ITk pixels at all fluences","Planar pixel modules keep hit efficiency without max voltage","Validated simulation predicts full pixel efficiency at HL-LHC","No need for 600V: pixel modules efficient at HL-LHC fluences"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000193,"raw_usage":{"total_tokens":1286,"prompt_tokens":818,"completion_tokens":468,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":434,"completion_tokens_details":{"reasoning_tokens":380}},"tokens_in":434,"tokens_out":468,"duration_ms":5203,"temperature":1.0,"reasoning_tokens":380,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T04:16:35.270404+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the cluster charge MPV of an ATLAS ITkPixV2 planar module (100 µm or 150 µm thick) irradiated to 3.5e15 neq/cm2 and biased at 400 V. If the measured MPV is below the effective threshold of about 1.3 ke, or deviates by more than the predicted uncertainty band from the simulated 5 ke (thin) or 6 ke (thick), the transferability of the validated models to the full ITk fluence range is ruled out.","supporting_citations":[{"cited_title":"Glessgen, Characterization of irradiated passive cmos sensors for tracking in hep experiments, Nucl","cited_arxiv_id":null,"evidence_quote":"Supplies the irradiated-device test-beam data (cluster charge MPV vs bias at 2.1e15 neq/cm2) used to validate the simulations."},{"cited_title":"Glessgen, M","cited_arxiv_id":null,"evidence_quote":"Provides the unirradiated sensor characterization and normalization point used to calibrate the simulation charge scale."},{"cited_title":"Folkestad and others, Development of a silicon bulk radiation damage model for sentaurus tcad, Nucl","cited_arxiv_id":null,"evidence_quote":"Supplies the LHCb silicon bulk radiation damage model implemented in the TCAD device simulations."},{"cited_title":"Mandic, V","cited_arxiv_id":null,"evidence_quote":"Defines the Mandić trapping model, one of the two bracketing models whose average is used for predictions."},{"cited_title":"Mandic, V","cited_arxiv_id":null,"evidence_quote":"Corrects the Mandić model; the paper notes the Allpix2 implementation had to be reimplemented to match the published version."},{"cited_title":"Adam, et al., Trapping in proton irradiated p+-n-n+ silicon sensors at fluences anticipated at the hl-lhc outer tracker, Journal of Instrumen- tation 11 (04) (2016) P04023","cited_arxiv_id":null,"evidence_quote":"Provides the underlying trapping-rate measurements on which the CMS trapping model is based."}],"review_version":1}