{"id":"45cf52fd-eb96-436a-a632-677ab3947f79","arxiv_id":"2607.16151","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Background Monte Carlo plus material assays predict under one millihertz per silicon qubit chip in SNOLAB's CUTE cryostat, with ~10 eV deposits able to cause correlated multi-qubit errors.","lead":"A SNOLAB collaboration measured the radioactive contamination of materials around a planned superconducting qubit setup and simulated how much background radiation will hit the qubit chips inside an underground cryostat. The projected rate is under one millihertz per silicon chip, and the same simulations show how radiation bursts can disturb several qubits at once.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The <1 mHz/Si-chip headline depends on excluding the direct 210Pb HPGe assays, which push the rate to 5.57 mHz (Table 3); the replacement (secular equilibrium + radon plate-out model) is ~2,000–5,000× below the measured values, and only a dedicated 210Pb/210Po assay on the actual chips can settle wh","rationale":"The paper's stated goal is to estimate the radiogenic background rate for a superconducting-qubit payload in CUTE and to quantify the benefit of a 133Ba calibration source. The conclusions — <1 mHz per Si chip and a ~50× controlled excess from 133Ba — are the central claims. For them to hold, the 210Pb content of the deployed materials must be as modeled: negligible in the bulk (secular equilibrium with the 226Ra break) and at the radon-exposure level of the medium scenario on surfaces. The least secure point is the exclusion of the direct 210Pb HPGe assays. The paper itself flags the issue (Sec. 2.3: 'appear unrealistically high', 'likely imprecise') and the omission is stated twice (Table 3 note, Sec. 6), so there is no hidden assumption. But the justification is qualitative, the data cannot distinguish the two interpretations (the assay is consistent with both zero and ~65 Bq/kg at 90% C.L.), and the difference between interpretations is a factor of ten in the headline rate. The radon plate-out model, which replaces the excluded measurements, predicts ~160× less 210Pb on the wafers than the direct assays would imply if read as surface contamination — so the model is not validated by the assay data; it is simply a different assumption. The reader's weakest_assumption identifies exactly this point, and I agree. The G4CMP multiplicity claim (O(10 eV) → correlated multi-qubit errors, whole substrate above ~100 eV) is secondary: it does not feed the rate projection (only the Δt=15 µs split time comes from G4CMP, and Appendix C shows a 10× larger split time changes rates by ≤10%). The downsampling-to-200-eV caveat the reader mentions is real but mostly affects the M4≈100% claim for β/α events, not the O(10–100 eV) threshold statement, which comes from the 2–192 eV ER/NR/phonon runs. So the 210Pb exclusion remains the single most load-bearing concern. Because the paper is fully transparent about this choice and presents the 'with 210Pb' numbers, a conditional verdict is appropriate: the quantitative claims should be adopted only together with the caveat that they depend on excluding the direct 210Pb assays. My read does not change the reader's CONDITIONAL verdict, so I recommend UNCHANGED. The concrete test is an independent high-sensitivity 210Pb/210Po assay with surface-vs-bulk discrimination on the actual chips — this settles whether the true rate is ~0.7 mHz or ~5.6 mHz.","tokens_in":37345,"tokens_out":18473,"duration_ms":135598,"concrete_test":"Perform a dedicated 210Pb/210Po assay on the actual QUTEbits Si wafers/chips with surface-vs-bulk discrimination: 210Po alpha spectrometry after sequential surface etch (≈1 µm) and bulk dissolution, plus low-background ICP-MS on the etchate and digestate. Feed the measured bulk and surface specific activities into the Geant4 model, replacing the secular-equilibrium and radon-scenario inputs for the Si chip (and, if measured, for the SMA connectors and microwave switch). For the headline <1 mHz to survive, the measured bulk 210Pb in the chip must be below roughly 1.7 Bq/kg (≈1,740 mBq/kg, the contribution that would add ≈0.30 mHz on top of the Table 7 total); at the HPGe central value (29 Bq/kg) the projected rate exceeds 5 mHz. The required sensitivity of ~0.1 mBq per chip (57 mg) is routinely achievable with low-background alpha counting of 210Po.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim — <1 mHz per Si chip in CUTE (Table 7) — is a factor of ten away from what the paper's own data imply if the direct 210Pb HPGe assays are taken at face value: Table 3 shows the bulk-only Si-chip rate rising from 0.551 mHz to 5.57 mHz, and the claimed 'fifty-fold' 133Ba excess (Table 8: 38.8/0.698 ≈ 56) would shrink to 38.8/5.57 ≈ 7. The load-bearing step is the exclusion of those assays (Secs. 2.3, 3.3, 6). The stated justification is qualitative: HPGe 210Pb sensitivity is poor (Table 1) and the 46.5-keV line is attenuated, so the measurements are 'imprecise'. But the actual Si-wafer assay (Table 10: 29,123±22,150 mBq/kg) is consistent with zero at only ~1.3σ, while its 90% C.L. upper limit (~65 Bq/kg) is still ~5,000× above the secular-equilibrium bound (<12.6 mBq/kg from the 226Ra break). The two interpretations are not statistically distinguishable, yet choosing between them moves the headline by an order of magnitude. No quantitative bias argument (control measurement, re-analysis of the 46.5-keV region, or surface-vs-bulk separation) is given. The treatment is also asymmetric: assay upper limits elsewhere (Si 238U, 232Th) are included at face value (Appendix C), while these low-confidence measurements are excluded. The radon plate-out model does not independently rescue the exclusion: for the medium scenario, Eq. (3.1) predicts ~0.012 mBq/cm² on the wafer, whereas the direct wafer assay, if read as surface contamination, implies ~1.9 mBq/cm² (~160× more). Thus the headline rests on an unvalidated judgment that the direct assays are biased high by roughly three orders of magnitude. The paper is transparent about the choice, so this is a correctness risk rather than an internal inconsistency, but it is the weakest point of the argument.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a material assay campaign for superconducting qubit components and uses Geant4 to project radiogenic background rates in the CUTE facility at SNOLAB. The central claim is a total projected background rate below 1 mHz per Si chip (Table 7), about fifty times lower than the rate expected when a 133Ba calibration source is inserted (Table 8). The paper further uses G4CMP simulations to estimate phonon collection times (~15 μs) and qubit hit multiplicities, concluding that energy deposits of O(10 eV) have a high probability of causing correlated multi-qubit errors and that the whole 7×7 mm substrate is affected above ~100 eV. The headline rate is obtained by excluding direct 210Pb HPGe bulk assays and instead assuming secular equilibrium with the bottom 238U chain, supplemented by a radon plate-out surface model.","tokens_in":37966,"tokens_out":5219,"duration_ms":42992,"significance":"If the central projection were robust, the work would be a valuable contribution to the design of low-background quantum experiments and to the understanding of radiation-induced correlated errors. The paper has clear strengths: the assay data are publicly released on radiopurity.org; the simulation campaign is detailed (482 component-isotope combinations, 3.63×10^12 primary decays); the use of G4CMP to ground the event-splitting parameter Δt=15 μs in a phonon-collection-time calculation is a good practice; and the 133Ba excess and multiplicity predictions are falsifiable. However, the principal '<1 mHz' claim is conditional on a data-selection choice that is not quantitatively justified, and the paper's own data contain a factor-of-ten higher rate if the direct 210Pb assays are included. The significance of the headline as stated is therefore limited.","major_comments":[{"comment":"The headline '<1 mHz per Si chip' is built on the 'without 210Pb' column of Table 3 (0.551 mHz bulk), carried into Table 7 (0.698 mHz total). The 'with 210Pb' column gives 5.57 mHz bulk — a factor of ten higher. The exclusion of the direct 210Pb assays (Sec. 2.3) is justified only qualitatively (poor HPGe sensitivity, 46.5-keV attenuation). No control measurement, re-analysis of the 46.5-keV region, or surface-vs-bulk separation is presented to demonstrate that the assay is biased. The Si-wafer assay (Table 10: 29,123±22,150 mBq/kg) is a measurement, not an upper limit; even its 90% C.L. upper limit (~65 Bq/kg) is ~5,000× above the secular-equilibrium bound (<12.6 mBq/kg). This is a load-bearing data-selection choice, and the central claim as stated is not supported by the paper's own data.","section":"Table 3; Secs. 3.3, 6; Table 7 note"},{"comment":"The radon plate-out model does not independently rescue the exclusion of the direct assays. Eq. (3.1) with the medium scenario (Table 5) predicts a total surface 210Pb emission rate of ~256 mBq summed over all components; for the Si wafer this corresponds to ~0.012 mBq/cm² (the manuscript does not quote the per-area value). In contrast, if the direct wafer assay (29,123 mBq/kg on 57 mg) is interpreted as surface contamination, it implies ~1.9 mBq/cm² on the two 7×7 mm faces. The two interpretations differ by more than two orders of magnitude, yet the manuscript does not quantitatively compare the model prediction to the wafer assay or explain why the model should be preferred for this specific component.","section":"Sec. 3.5, Eq. (3.1); Table 5"},{"comment":"The second central quantitative claim — that a 133Ba source provides a fifty-fold excess over background — is equally affected by the 210Pb data-selection choice. Table 8 gives 38.8 mHz for the best 133Ba deployment, and Table 7 gives 0.698 mHz total background, yielding 38.8/0.698 ≈ 56. If the direct 210Pb assays are included (5.57 mHz bulk; total >5.6 mHz), the excess ratio falls to ~7. The paper should present both scenarios in the abstract/conclusions or provide the missing quantitative justification for excluding the direct assays before claiming either the <1 mHz background or the fifty-fold controlled-exposure capability.","section":"Sec. 4.2, Table 8"}],"minor_comments":[{"comment":"The text reads 'GCPMP version V09-09-02'; this is presumably a typo for 'G4CMP'.","section":"Sec. 5 (paragraph 2)"},{"comment":"The 210Pb entry for the Si chip is formatted as '(29±22)·10 3' in the source; this is ambiguous. Use standard notation, e.g., (29±22)×10³ mBq/kg, or better, 29,000±22,000 mBq/kg.","section":"Table 2 and Table 10"},{"comment":"The split time is quoted as Δt=15 μs with no uncertainty. Since it is informed by the G4CMP τmax value of 14.7±1.1 μs, report the rounded value with a clear statement that it is a rounded quantity and discuss the sensitivity (appendix C does this for 1 ms but not for the ±1 μs variation).","section":"Sec. 3.2 and Sec. 5.1"},{"comment":"The abstract states a 'fifty times higher rate' from the 133Ba source without qualification. Given that this factor holds only for the no-210Pb projection, the abstract should flag the conditional nature of the statement or the revised factor for the alternative projection.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a competent simulation study with valuable public data and a transparent exposition of its assumptions. However, the central numerical claim is currently driven by an exclusion of direct measurements, and the paper's own Table 3 shows a factor-of-ten alternative. I recommend major revision: the authors should either (a) provide a quantitative validation of the 210Pb exclusion using control measurements or a rigorous surface/bulk decomposition, or (b) reframe the abstract and conclusions around the with-210Pb scenario (5.6 mHz, 133Ba excess ~7). The paper is publishable in either case, but not with the current '<1 mHz' headline."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this is a solid, transparent engineering study that will be useful to anyone planning radiation-sensitive quantum experiments in an underground facility. The new content is real: the component assays, the Geant4 background budget for the QUTEbits payload in CUTE, and the G4CMP phonon collection-time and multiplicity curves for a 4-transmon chip. The G4CMP numbers (mean ~1.2 μs, max ~15 μs collection time) are plausible and directly usable for hit-window choices. The paper does careful work: full system geometry in Geant4, 482 component-isotope combinations, cross-checks on split time, and it publishes assay results on radiopurity.org.\n\nThe soft spot is the one you already see: the headline <1 mHz per Si chip is only reached by excluding the directly measured 210Pb bulk assays. Table 3 is honest about this—including those assays gives 5.57 mHz instead of 0.55 mHz. The stated justification is that HPGe is insensitive to the 46.5 keV line, which is fair, but the paper doesn't provide a quantitative argument for why the measured values (e.g., 29±22 Bq/kg on the wafer) are biased high. The radon plate-out model doesn't independently settle it: the medium scenario predicts ~0.012 mBq/cm², while the direct assay, read as surface contamination, implies ~1.9 mBq/cm²—two orders of magnitude higher. So the choice is not neutral. The treatment is also asymmetric: upper limits for 238U and 232Th are used at face value, while the 210Pb assays are replaced. This is a correctness risk, not an inconsistency—the paper is upfront—but it matters because the abstract and conclusions lean on the <1 mHz number.\n\nThe G4CMP multiplicity conclusions are based on downsampled 200 eV deposits for high-energy particles. That's a reasonable practical choice, and the qualitative conclusions—O(10 eV) deposits can hit multiple qubits, >100 eV affects the whole substrate—are probably robust. Not a major concern.\n\nWho's this for? Experimentalists planning underground qubit runs, and anyone building low-background cryogenic setups. It deserves a serious referee. I'd send it to review, with a note asking for a quantitative treatment of the 210Pb exclusion (or a clearly caveated headline). If that issue is resolved, it's a solid contribution.","headline":"Useful engineering study with a transparent but load-bearing choice: the <1 mHz/Si-chip background excludes the direct 210Pb assays, which would give ~5.6 mHz.","tokens_in":38544,"tokens_out":2813,"would_cite":true,"duration_ms":23574,"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":"The paper projects that a superconducting qubit chip operated in the SNOLAB cryostat CUTE will see a radiogenic background near 0.7 mHz — about one ionizing event every 24 minutes — and that a 133Ba calibration source can raise that rate ro","keywords":["superconducting qubits","radiopurity material assays","radiation background projection","SNOLAB CUTE facility","210Pb contamination","correlated qubit errors","phonon transport simulation","underground quantum device testing"],"falsifier":"Measure bulk 210Pb directly in the same silicon wafers and neighbouring components — by 46.5 keV gamma counting on a detector with good 210Pb sensitivity, or by alpha-counting 210Po after ingrowth — and compare with the secular-equilibrium value the paper substitutes. Or simply take the first CUTE run's measured hit rate: a rate near 5.6 mHz rather than ~0.7 mHz would indicate the discarded bulk 210Pb assays were the correct input, with the 133Ba source providing an in-situ rate calibration for the comparison.","tokens_in":37303,"feed_emoji":"⚛️","tokens_out":17692,"duration_ms":139603,"temperature":0.7,"pith_summary":"This is a preparation study for the first deep-underground operation of superconducting qubits, inside the CUTE cryostat at SNOLAB, 2 km below the surface. By measuring the radioactivity of every component that will sit near the qubit chip and feeding those assays into particle-transport simulations, the paper projects a radiogenic background of about 0.7 mHz per 7×7 mm silicon chip — roughly one energy-depositing event every 24 minutes — with the chip's own substrate and the components in direct contact with it dominating, while the rock, muons, and external shielding contribute almost nothing. It further shows that a 133Ba gamma source already available in the facility can raise the hit rate about fiftyfold, turning CUTE into a controllable radiation laboratory for coherence studies. A third result, from crystal-dynamics simulations, is that energy deposits of order 10 eV already have a high probability of disturbing several qubits on the same chip, and deposits above about 100 eV affect the entire substrate, implying that correlated error bursts can only be mitigated by phonon engineering rather than by material purity alone. The load-bearing input — acknowledged in the paper — is the replacement of imprecise direct 210Pb assays with a secular-equilibrium assumption plus a modeled surface 210Pb contribution; taking the direct assays at face value raises the projected rate tenfold.","feed_headline":"Projected: one radiation hit per qubit chip every 24 minutes","feed_subtitle":"Material assays and simulations put the chip under 1 mHz of background; a gamma calibration source can raise exposure fiftyfold.","key_machinery":"The load-bearing machinery is a material-assay pipeline feeding two simulation layers. First, high-purity germanium detectors measure the 238U (split at 226Ra), 235U, 232Th, 210Pb, 40K, and cosmogenic content of every component; then a Monte Carlo particle-transport code (Geant4) models radioactive decay and tracks emitted particles through a full geometry of the CUTE cryostat and payload, recording energy deposits in the silicon chip with events coalesced in a 15 μs window. The decisive step is the 210Pb treatment: the paper sets aside the directly measured bulk 210Pb values as too insensitive and instead assumes bulk 210Pb is in secular equilibrium with the bottom of the 238U chain, adding","core_discovery":"On the paper's own terms: a screened superconducting-qubit payload, operated 2 km underground in the shielded CUTE cryostat, is projected to see about 0.7 mHz of radiogenic background per 7×7 mm silicon chip — under one energy-depositing event every 24 minutes — dominated by contaminants in the chip itself and in the circuit board and holder touching it. The projection deliberately uses secular-equilibrium values for bulk 210Pb instead of the imprecise direct 210Pb assays, plus a modeled surface 210Pb component from radon exposure. The same simulations show that a 133Ba source raises the rate about fiftyfold, enabling controlled exposure studies. Crystal-dynamics simulations add that deposit","pith_inferences":["The paper's own comparison suggests an arbiter for its central assumption: the first run's measured hit rate will fall near 0.7 mHz if the secular-equilibrium substitution is right, or near 5.6 mHz if the discarded bulk 210Pb assays were the better input; the two projections differ almost entirely in that one isotope.","Because the phonon response is nearly independent of particle type above ~100 eV, the 133Ba gamma source may serve as a faithful surrogate for rare alpha and neutron events when testing correlated-error mitigation — a substitution the paper establishes in simulation but does not yet claim experimentally.","The reported collection times (mean ~1.2 μs, maximum ~15 μs, across all particle types) imply that any multi-qubit error-mitigation logic must operate on tens-of-microsecond coincidence windows; the observed phonon caustics pattern also suggests qubit hit patterns might be used to reconstruct where on the chip an ionizing event struck.","The 15 μs versus 1 ms comparison in the appendix hints that timing structure alone can separate decay chains in the data: longer coincidence windows suppress the 238U-chain contribution by ~10% while leaving 232Th nearly unchanged, offering a data-driven way to de-mix spectral components after the first run."],"forward_implications":["A direct radiation test becomes possible: identical devices measured at CUTE and in a surface lab should differ in coherence only through the radiation environment, since the projected underground background is reduced to about one hit every 24 minutes per chip.","The background budget is local: external sources (cavern rock, muons, interstitial air) contribute below the percent level, so further rate reduction means cleaning the payload's own materials — the silicon substrate, the PCB, and the holder dominate the budget.","The 133Ba source is a built-in dose lever: at its best rotation it delivers ~39 mHz in a chip, roughly fifty times the projected background, letting the collaboration measure coherence versus radiation exposure without hardware changes; the source's 15% activity uncertainty dominates the systematic error.","Correlated errors are a substrate-wide phenomenon: above ~100 eV any single deposit affects the whole 7×7 mm chip, so layout changes cannot prevent them; the paper argues for phonon mitigation — absorbers, backside coatings, mechanical isolation of qubit islands.","Fault-tolerant-scale extrapolation: scaling the rate linearly to a 20×20×0.5 mm processor chip gives ~10 mHz of background, limiting uninterrupted algorithm execution to ~100 s in a CUTE-like environment, versus a few seconds in an unshielded surface laboratory."],"fun_headline_variants":["Underground qubit chip: a radiation hit every 24 minutes","At 2 km deep, qubit chip sees ~0.7 mHz background","SNOLAB qubits: one energy deposit per chip per 24 min","Shielded qubits: 0.7 mHz background, one hit per 24 min","Deep-underground qubits: chip background under 1 mHz"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The sub-millihertz projection rests on treating the directly measured bulk 210Pb activities as too imprecise to use and instead assuming bulk 210Pb sits in secular equilibrium with the uranium decay chain while adding a modeled surface 210Pb term from a specific radon-exposure scenario (2 years on the surface, 14 days underground); if the direct 210Pb measurements are used at face value, the projected per-chip rate rises from 0.55 mHz to 5.57 mHz.","fun_headline_variants_meta":{"raw":{"variants":["Underground qubit chip: a radiation hit every 24 minutes","At 2 km deep, qubit chip sees ~0.7 mHz background","SNOLAB qubits: one energy deposit per chip per 24 min","Shielded qubits: 0.7 mHz background, one hit per 24 min","Deep-underground qubits: chip background under 1 mHz"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000254,"raw_usage":{"total_tokens":1412,"prompt_tokens":756,"completion_tokens":656,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":500,"completion_tokens_details":{"reasoning_tokens":564}},"tokens_in":500,"tokens_out":656,"duration_ms":5999,"temperature":1.0,"reasoning_tokens":564,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T21:10:49.987452+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure bulk 210Pb directly in the same silicon wafers and neighbouring components — by 46.5 keV gamma counting on a detector with good 210Pb sensitivity, or by alpha-counting 210Po after ingrowth — and compare with the secular-equilibrium value the paper substitutes. Or simply take the first CUTE run's measured hit rate: a rate near 5.6 mHz rather than ~0.7 mHz would indicate the discarded bulk 210Pb assays were the correct input, with the 133Ba source providing an in-situ rate calibration for the comparison.","supporting_citations":[],"review_version":1}