{"id":"54bba2d4-acb6-481f-a5eb-9ae36b6ead12","arxiv_id":"2607.26732","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Geant4 simulations of the newly installed CROSS 100Mo bolometer array project BI ≈ 3.2(5)×10⁻³ cnts/keV/kg/yr, enabling a world-leading T1/2 limit ~4×10²⁴ yr in one live year.","lead":"CROSS built a 4.9 kg 100Mo cryogenic calorimeter array at Canfranc and projects a background index of 3.2×10⁻³ counts/keV/kg/yr near 3034 keV. That level would let it set the strongest 100Mo neutrinoless double-beta half-life limit in about one year of running.","discovery_kind":"extension","skeptic_critique":{"model":"grok-4.5","headline":"Omitted surface radioactivity plus CUPID-Mo-transferred bulk activities and uniform best-case resolution remain the soft spot under the BI=3.2e-3 that drives the one-year world-leading claim.","rationale":"The reader correctly isolates the transferred radiopurity numbers, the complete omission of surfaces, and the uniform best-case detector response as the weakest assumptions under the headline BI and sensitivity. Those choices are load-bearing: the paper’s own Fig. 13 shows that non-muon terms are already comparable to the statistical uncertainty on the total, and surface events on the nearest copper are known from prior LMO arrays to populate the ROI once bulk contamination is at the µBq/kg level. The authors hedge with factor-3/10 degradation curves, so the manuscript remains a credible construction-plus-projection paper; the appropriate stance is still CONDITIONAL pending confrontation with commissioning data. No stronger internal inconsistency or calculation error was found; the muon-veto dead-time and 2ν pile-up treatments are standard and cross-checked. Hence the reader’s verdict and weakest-assumption diagnosis stand without adjustment.","tokens_in":19345,"tokens_out":720,"duration_ms":28653,"concrete_test":"Re-run the Geant4 model of Sect. 3 with surface 226Ra/228Th activities on the Cu frames and 10 mK screen set to the surface specific activities reported in the CUPID-Mo background model (or to the 90 % C.L. upper limits from the CROSS HPGe screening if tighter); recompute the 100 keV BI at 3034 keV. If the new total BI exceeds ∼1×10^{-2} ckky, the one-year “world-leading” claim in Sect. 4 no longer holds under the paper’s own efficiency and live-time assumptions.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central projection (BI = 3.2(5)×10^{-3} ckky in a 100 keV window at 3034 keV, Sect. 3.5; lim T_{1/2}∼4×10^{24} yr in 1 yr, Sect. 4) rests on three linked modeling choices stated in Sect. 3.2–3.3 and Table 1: (i) bulk 226Ra/228Th of LMO crystals and Cu frames are set equal to CUPID-Mo values rather than measured on the as-built CROSS parts; (ii) surface contamination is omitted entirely (“this simulation does not consider any special contribution to the radioactive components in the surface of the materials”); (iii) every crystal and LD is assigned the single best prototype resolution and LHR of Ref. [48], with no NTL gain and no performance spread. Because the 10 mK Cu screen and frames face the crystals, degraded surface α/β events land directly in the β/γ band near Qββ (explicitly called “the most harmful background” in the text and visible in Fig. 11). Muons dominate the quoted BI, yet any realistic surface term on the nearest Cu can raise the non-muon floor by the same factor 3–10 the authors already flag. The sensitivity curve therefore inherits an unquantified systematic that is only bounded, not measured, by the conservative scenarios of Fig. 14.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The manuscript describes the construction and underground installation of the CROSS array of 42 dual-readout cryogenic calorimeters (32 100Mo-enriched Li2MoO4 crystals totaling 4.9 kg of 100Mo) at LSC, together with a Geant4-based background model and a projected sensitivity to 100Mo 0ν2β decay. Using assayed or sister-experiment activities, a detector-response model taken from CROSS prototypes, muon veto anti-coincidences, and standard selection cuts, the authors predict a background index of 3.2(5)×10−3 cnts/keV/kg/yr in a 100 keV window at Qββ=3034 keV. With 18% muon-veto dead time and 90% duty cycle this yields a projected 90% C.L. half-life limit ∼4×10^24 yr in one year of elapsed time, with conservative factor-3 and factor-10 worse BI scenarios still competitive after two years.","tokens_in":19683,"tokens_out":1538,"duration_ms":40844,"significance":"CROSS is a timely intermediate-scale 100Mo bolometric experiment that bridges CUPID-Mo/AMoRE-I and the forthcoming tonne-scale CUPID/AMoRE-II programs, while also serving as a technology demonstrator for NTL light detectors and a low-mass Cu/PLA holder. A documented as-built geometry, a full facility MC, and an explicit sensitivity projection with conservative BI envelopes are valuable to the community even before physics results appear. The construction narrative, cleaning protocols, wiring redesign, and muon-veto dead-time quantification are concrete contributions. The projection itself is falsifiable once commissioning data are released, which strengthens rather than weakens the paper.","major_comments":[{"comment":"Sect. 3.3 (and the explicit statement that “this simulation does not consider any special contribution to the radioactive components in the surface of the materials”): surface 226Ra/228Th on the copper frames and 10 mK screen is omitted entirely, yet the text and Fig. 11 identify precisely these facing surfaces as “the most harmful background” because degraded α/β events populate the β/γ band near Qββ. A bounding estimate (e.g., a surface activity scaled from CUPID-Mo or from the CROSS holder R&D) should be added so that the non-muon floor is not left unquantified; the factor-3/10 envelopes in Fig. 14 are not a substitute for a surface term in the baseline model.","section":"Sect. 3.3, Fig. 11"},{"comment":"Table 1 and Sect. 3.3.1: bulk 226Ra/228Th activities of the LMO crystals and copper frames are taken equal to CUPID-Mo values rather than measured on the as-built CROSS parts. While the materials and producers are the same, the crystals underwent a different assembly/gluing campaign and the frames are a new low-mass design. The paper should either report (or cite) CROSS-specific HPGe/bolometric assays or propagate an explicit systematic on these two entries into the total BI uncertainty, instead of absorbing the risk only into the ad-hoc factor-3/10 scenarios.","section":"Table 1, Sect. 3.3.1"},{"comment":"Sect. 3.2: every crystal and LD is assigned the single best prototype resolution and LHR of Ref. [48], with no performance spread and with NTL gain deliberately set to zero “worst case.” The sensitivity arithmetic in Sect. 4 then adopts a 17.1 keV ROI width taken from CUPID-Mo. Because energy resolution and light-yield uniformity directly set both the analysis ROI and the PSD/pile-up rejection efficiency, the baseline BI and the 70.2% total efficiency should be shown under at least one degraded-resolution scenario (e.g., FWHM scaled by the observed prototype spread), not only under a global BI rescaling.","section":"Sect. 3.2, Sect. 4"}],"minor_comments":[{"comment":"Abstract and Introduction state that physics data taking has been ongoing since mid-November 2025 / April 2026; the arXiv datestamp is 29 Jul 2026. Clarify the exact timeline (commissioning vs. physics) so that the reader knows whether any in-situ validation of the BI model is already available or still forthcoming.","section":"Abstract, Sect. 1"},{"comment":"Eq. (2): the numerical prefactors 30, 20, 5, 3 that convert crystal energy into LD scintillation are given without units or a clear derivation from the quoted 0.3 keV/MeV LHR; a one-sentence justification would help reproducibility.","section":"Sect. 3.2, Eq. (2)"},{"comment":"Fig. 13 error bars combine activity and MC statistical uncertainties, but the text does not state how the two are combined or whether activity upper limits (lead shielding) are treated as one-sided. A brief note would suffice.","section":"Sect. 3.5, Fig. 13"},{"comment":"The muon energy spectrum (Eqs. 3–5) is an above-ground approximation folded with a simple exponential overburden; a short comparison to the measured LSC muon spectrum (or a citation to the validation in Ref. [38]) would strengthen confidence that the dominant BI term is not biased.","section":"Sect. 3.3.4"},{"comment":"Typographical / consistency items: “molibdate” → “molybdate” (Conclusions); “undeground” → “underground” (Introduction); Q-value is written both Q2β and Qββ; the abstract’s “mid-November 2025” vs. body “April 2026” start of physics running should be aligned.","section":null},{"comment":"Sect. 3.3 mentions that neutron-induced background is “not expected to represent a dominant contribution” on the basis of a ∼2 counts/day thermal-neutron rate, yet no MC estimate is given. A one-paragraph upper bound (or a clear deferral to a future note) would close the source list.","section":"Sect. 3.3"}],"recommendation":"minor_revision","confidential_remarks":"The paper is a solid instrumentation-plus-projection contribution appropriate for EPJC. The three modeling choices flagged above are real but already partially mitigated by the conservative BI curves the authors themselves plot; I therefore recommend minor rather than major revision. No novelty or citation-pattern concerns. The timeline language (data taking “ongoing since” dates that post-date or straddle the arXiv stamp) is slightly awkward and should be cleaned for the journal version so the status of the experiment is unambiguous."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"Punchline: this is the construction record and background/sensitivity forecast for the 42-module CROSS array that is already taking physics data at LSC. It is not a measured BI. What is new is the as-built inventory (32 enriched LMOs = 4.9 kg 100Mo, TeO and depleted/natural companions, NTL Ge/Si LDs, Cu/PLA holder at 6% Cu-to-crystal mass), the facility-specific muon-veto model with 18% dead time, and the end-to-end Geant4 number BI = 3.2(5)×10^{-3} ckky in a 100 keV window at 3034 keV, which under 90% duty cycle gives lim T1/2 ~ 4×10^{24} yr in one elapsed year.\n\nThey do the standard things carefully. Geometry includes crystals, LDs, frames, Kapton, cryostat screens, internal/external Pb, and a simplified veto. Response model uses 1 ms integration, resolution smearing from prototype fits (~6 keV FWHM at 2615 keV), LHR light summing, single-hit + LHR band + veto anti-coincidence + pile-up cuts. Activities are HPGe assays or CUPID-Mo numbers; muons use LSC flux/angular data; 2ν2β and pile-up use Decay0. Feldman–Cousins sensitivity arithmetic is transparent, and they plot conservative BI = 1×10^{-2} and 5×10^{-2} cases so a 2-year run still looks competitive or best-in-class for 100Mo.\n\nSoft spot, in proportion: surface radioactivity is omitted by design, crystal/frame bulk 226Ra/228Th are transferred from CUPID-Mo, and every channel is given the best prototype resolution/LHR with no NTL gain and no spread. The text itself flags the 10 mK screen and frames as the most harmful near-ROI source (Fig. 11). Muons dominate the quoted total, so a realistic surface floor on nearest Cu could move the non-muon piece by the same factor 3–10 they already show. That is a modeling systematic, not a hidden contradiction; commissioning data (promised) will settle it.\n\nMath, citations, and methods look solid for an instrumentation-and-projection paper. Self-cites are to their own prototypes, muon-veto paper, and CUPID-Mo background model—appropriate here. Who it is for: anyone running or designing cryogenic 100Mo arrays (CUPID, AMoRE, next CROSS stages). Bring it to reading group if the group does bolometers or underground background models; skip if you only care about final half-life limits. I would send it to peer review without hesitation and would cite the construction details and the BI breakdown when comparing projections. Engage.","headline":"Solid as-built construction + Geant4 forecast for the running CROSS array; the BI=3.2e-3 and one-year world-leading claim are projections that lean on transferred radiopurity and no surface term, but the authors already bound that with factor-3/10 scenarios.","tokens_in":20638,"tokens_out":729,"would_cite":true,"duration_ms":21002,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["29.40.Vj","23.40.-s","07.20.Mc","95.55.Vj"],"model":"grok-4.5","headline":"CROSS’s Monte Carlo background model predicts 3.2×10⁻³ counts/keV/kg/yr at the 100Mo Q-value, enough for a world-leading half-life limit near 4×10²⁴ yr in one year of running.","keywords":["neutrinoless double-beta decay","100Mo","scintillating bolometers","lithium molybdate","background index","Geant4","cryogenic detectors","Canfranc"],"falsifier":"After one year of physics data, measure the actual event rate in the 100Mo ROI after the same single-crystal, light-band and muon-veto cuts; if it substantially exceeds ~3×10⁻³ cnts/keV/kg/yr the projected sensitivity fails.","tokens_in":20187,"feed_emoji":"❄️","tokens_out":1081,"duration_ms":22173,"temperature":0.7,"pith_summary":"CROSS has built and installed a 42-module array of scintillating cryogenic calorimeters holding 4.9 kg of 100Mo at Canfranc, and this paper reports the detector construction plus a full Geant4 projection of background in the neutrinoless double-beta region of interest. The simulations give a background index of 3.2(5)×10⁻³ counts/keV/kg/yr in a 100 keV window around 3034 keV. With the muon-veto dead time and ordinary facility duty cycle folded in, that level would let the experiment set a half-life limit around 4×10²⁴ yr after one year—competitive with or better than existing 100Mo bounds. Even if materials or detector performance push the background three- to ten-fold higher, two years of data would still remain competitive. The paper therefore establishes both that the hardware is in place and that the projected background supports a leading 100Mo search.","feed_headline":"CROSS projects 4×10²⁴ yr 100Mo limit in one year","feed_subtitle":"Monte Carlo gives 3.2×10⁻³ counts/keV/kg/yr at 3034 keV for the new Canfranc array","key_machinery":"A Geant4 geometry of crystals, light detectors, copper structure, cryostat screens, lead shielding and muon veto, with detector response (energy resolution, light-to-heat ratio, 1 ms integration, single-crystal and veto cuts) applied event-by-event to simulated U/Th chains, 100Mo 2ν2β pile-up and cosmic muons; the resulting spectra are integrated in the ROI to obtain the background index.","core_discovery":"Detailed Geant4 modeling of the installed CROSS array, cryostat, shielding and muon veto predicts a total background index of 3.2(5)×10⁻³ cnts/keV/kg/yr in a 100 keV interval centered on the 100Mo Q-value. Combined with 18 % muon-veto dead time, 90 % duty cycle, ~70 % total efficiency and 4.9 kg of 100Mo, this background yields a projected 90 % C.L. half-life sensitivity of order 4×10²⁴ yr in one year of elapsed time, still world-leading or competitive after two years even if the true background is several times worse.","pith_inferences":["If surface α contamination proves higher than bulk assumptions, light-detector pulse-shape and NTL gain become the main remaining handles for recovering the projected BI.","The large muon contribution at Canfranc depth implies that any future ton-scale follow-on at the same site will need either deeper overburden or a higher-efficiency veto to stay background-free.","Successful operation of the uncoated LMO + NTL-LD configuration validates the fallback design chosen after surface-coating R&D underperformed on large crystals."],"forward_implications":["One year of CROSS data at the predicted background would set the strongest published limit on 100Mo 0ν2β decay.","A factor-of-three worse background still allows a world-best limit within roughly 1.5–2 years.","The same low-mass copper/PLA structure and NTL light detectors can be reused as a technology pathfinder for larger 100Mo arrays.","Muon-induced secondaries dominate the ROI, so further veto or shielding gains would directly improve sensitivity."],"fun_headline_variants":["CROSS projects 4×10²⁴ yr 100Mo 0ν2β limit in one year","Geant4 pegs CROSS BI at 3.2×10⁻³ for 4×10²⁴ yr 100Mo reach","CROSS array eyes 4×10²⁴ yr 100Mo sensitivity after 1 yr live","Canfranc CROSS sim: 3.2×10⁻³ BI enables leading 100Mo limit","One-year CROSS data projects 4×10²⁴ yr 100Mo half-life bound"],"cache_read_input_tokens":16512,"weakest_assumption_plain":"Crystal and copper-frame radioactivities are taken from an earlier experiment, surface contamination is left out of the geometry, and every detector is assigned the best prototype resolution and light yield.","fun_headline_variants_meta":{"raw":{"variants":["CROSS projects 4×10²⁴ yr 100Mo 0ν2β limit in one year","Geant4 pegs CROSS BI at 3.2×10⁻³ for 4×10²⁴ yr 100Mo reach","CROSS array eyes 4×10²⁴ yr 100Mo sensitivity after 1 yr live","Canfranc CROSS sim: 3.2×10⁻³ BI enables leading 100Mo limit","One-year CROSS data projects 4×10²⁴ yr 100Mo half-life bound"]},"model":"grok-4.5","effort":"low","cost_usd":0.006092,"raw_usage":{"total_tokens":1677,"prompt_tokens":938,"num_sources_used":0,"completion_tokens":121,"cost_in_usd_ticks":60924000,"prompt_tokens_details":{"text_tokens":938,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":618,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":938,"tokens_out":121,"duration_ms":10997,"temperature":1.0,"reasoning_tokens":618,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-30T22:40:34.284000+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"After one year of physics data, measure the actual event rate in the 100Mo ROI after the same single-crystal, light-band and muon-veto cuts; if it substantially exceeds ~3×10⁻³ cnts/keV/kg/yr the projected sensitivity fails.","supporting_citations":[],"review_version":1}