{"id":"03e6d0ec-6b1c-4de3-89a8-72dc92f34a39","arxiv_id":"1908.07734","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A prototype camera and LED system for the IceCube Upgrade can, based on calibrated simulations, measure refrozen ice bubble column radii to about 1 cm and bulk ice scattering lengths to about 10 m.","lead":"A camera and LED calibration system for the IceCube Upgrade was prototyped and tested in a swimming pool, with simulations showing it can map refrozen drill-hole ice and bulk ice properties. This matters because better ice calibration reduces uncertainties in neutrino direction and energy measurements across the whole IceCube detector.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The O(10 m) bulk-ice scattering-length precision is only asserted under an LED-orientation assumption that the paper never shows is identifiable; tilt and scattering length may be degenerate.","rationale":"The reader's weakest assumption is broader, citing both the air-derived calibration and LED orientation; I agree partially but single out the orientation–scattering-length identifiability because Fig. 5's caption and text condition the headline precision on it, and the paper provides no evidence that the condition is met. The pool test (two LEDs at 25 m separated with 10 cm uncertainty) is genuine support for imaging hardware, but it does not constrain the in-ice orientation–scattering-length degeneracy. The bubble-column claim is less affected because it is a geometric measurement along a single string and does not require cross-string orientation knowledge. If the proposed two-parameter check were run and the marginalized error remained O(10 m), the paper's claim would stand; if not, the CONDITIONAL verdict would need to be kept or tightened to require an orientation-calibration demonstration. Since the paper already states the orientation caveat and the concern is addressable, I do not move the verdict from CONDITIONAL, hence UNCHANGED.","tokens_in":6438,"tokens_out":5710,"duration_ms":62302,"concrete_test":"Generate a continuous grid of simulated images with effective scattering lengths 25–55 m and LED tilts drawn from a Gaussian with σ=3° (or the actual installation tolerance), then perform a joint two-parameter fit with scattering length and tilt as free parameters and marginalize over tilt. Report the marginalized 1σ uncertainty on scattering length. If it exceeds ~10 m, the quoted precision requires an independent orientation calibration, not just the stated assumption. Alternatively, test identifiability by fitting with the wrong tilt: if a 3° tilt error shifts the best-fit scattering length by more than 10 m, the degeneracy is confirmed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central bulk-ice claim in Sec. 4 is that the camera system can measure the effective scattering length to O(10 m) from a Δχ2 comparison to a high-statistics template, assuming the LED orientation is known to a few degrees. This assumption is load-bearing and unsupported. Table 1 specifies a 5° camera-alignment tolerance but says nothing about how the illumination module on a neighboring string is oriented relative to the camera at deployment; DOM azimuthal orientation is not controlled, and the system is supposed to determine DOM orientation from the same light-cone images. The paper never demonstrates that the orientation can be recovered to a few degrees without absorbing the scattering-length information. Moreover, the simulations use only five discrete tilts and three scattering lengths, and Fig. 5 right subtracts the per-tilt χ² minimum; this does not establish that tilt and scattering length are separately identifiable in a continuous joint fit. The claim that multiple depth measurements reduce the orientation uncertainty is an assertion without a combined-fit simulation. If the two parameters are partially degenerate, a few degrees of unknown tilt could push the scattering-length uncertainty well beyond 10 m.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper describes the design, prototype tests, and simulation-based performance projections of a camera system for the IceCube Upgrade. The system pairs CMOS cameras installed inside optical modules with LED illumination modules to image refrozen drill-hole ice and scattered light in the surrounding bulk ice, with the additional goal of recovering module position and orientation. After presenting engineering requirements and a quality-control workflow, the authors show swimming-pool images documenting basic functionality, then use a photon Monte Carlo with the SPICE ice model to simulate expected in-ice images. From these simulations they report O(1 cm) precision on the bubble-column radius and O(10 m) precision on the bulk-ice effective scattering length, the latter contingent on the LED orientation being known to within a few degrees.","tokens_in":6640,"tokens_out":4256,"duration_ms":48420,"significance":"If the headline precision claims hold, the camera system would supply a new in-situ calibration of refrozen-hole-ice and bulk-ice optical properties, directly serving the IceCube Upgrade's detector-calibration goals. The paper's strengths include a concrete prototype with characterized components, a plausible acceptance-test workflow, demonstration images from water that show basic imaging out to 25 m, and use of the standard IceCube SPICE photon-propagation model. The quoted precisions, however, come from a simulation-only self-consistency study: the photon-to-pixel scaling is calibrated in air, the expected images are generated with the same simulation chain, and the analysis recovers input parameters that were put into the simulation. That is legitimate as design validation, but it is not an end-to-end validation of the measurement, and the central scattering-length claim rests on an orientation-assumption that the paper does not demonstrate is identifiable.","major_comments":[{"comment":"The claimed O(10 m) precision for the bulk-ice scattering length is shown only for five discrete LED-tilt values, with the per-tilt chi-square minimum subtracted. This demonstrates sensitivity to the scattering length when the tilt is fixed, but it does not establish that the scattering length and LED orientation can be separately determined in a joint fit. Since the same camera system is intended to determine DOM orientation from the light-cone images, and Table 1 only specifies camera alignment within 5 degrees, the assumption that the LED orientation is 'known within a few degrees' is load-bearing and unsupported. A continuous two-parameter fit, or an explicit demonstration that the orientation can be recovered from the same simulated images without degrading the scattering-length sensitivity, is needed before the precision claim is credible. The statement that multiple depth measurements will reduce the orientation uncertainty is an assertion that is not backed by a combined-fit simulation.","section":"Section 4, Fig. 5 (right)"},{"comment":"The precision of the delta-chi-squared analysis depends on absolute image intensities, yet the photon-to-pixel scaling factor is derived from a test setup in air, with no uncertainty quoted for the factor and no validation that the calibration transfers to ice. The in-ice environment differs in glass refraction, scattering, and the angular response of the illumination module, so the transferability of an air-derived absolute calibration is not self-evident. The authors should either provide a systematic uncertainty for the scaling factor, validate it with an in-ice or water measurement, or explicitly restrict the precision claims to the idealized simulation model.","section":"Section 4, photon-to-pixel calibration"},{"comment":"The bubble-column-radius precision is reported as 'a few centimetres', but the measurement procedure is described only as 'measuring size of the visible features in the image', and the figure does not show fit lines, per-configuration scatter, or systematic error estimates. With six configurations and ten pseudo-experiments each, the claimed O(1 cm) precision is not yet robustly demonstrated against the main identified systematics, in particular the photon-to-pixel calibration and the choice of effective scattering length in the column. The authors should define the radius estimator, report the measured versus true radius for each configuration, and state which uncertainties enter the quoted precision.","section":"Section 4, Fig. 5 (left)"}],"minor_comments":[{"comment":"The red box mentioned in the text is not clearly marked in the figure, and the axis label appears garbled as 'values χ2Δ' with '3 10×'; the axes should be labeled as 'Δχ2' with correct scientific notation.","section":"Section 4, Fig. 5 (right)"},{"comment":"The paper uses 'LED orientation', 'camera alignment', and 'relative orientations of light' without a consistent definition of the relevant angles; defining the reference frame and the tilt convention in a short paragraph or diagram would substantially improve reproducibility.","section":"Section 2"},{"comment":"The swimming-pool images demonstrate the basic functionality well, but the text says the captured intensity 'varied with distance' without quantitative comparison to expectation; one sentence with the measured-versus-expected trend would make this subsection more informative.","section":"Section 3"},{"comment":"The sentence 'The camera system is to be installed in all three types of DOMs' is clear, but the subsequent description of camera count per DOM is easy to misread; a small table summarizing the number and pointing direction of cameras and illumination modules per DOM type would help the reader.","section":"Section 1"}],"recommendation":"major_revision","confidential_remarks":"This is a short conference proceedings, so the bar for full derivations is lower than for a full journal article. However, the two headline precision claims are load-bearing and both require additional work: the scattering-length claim needs a joint fit with LED orientation, and both claims need an explicit treatment of the air-derived calibration uncertainty. These are achievable within the scope of the existing simulation study. If the authors prefer to keep the paper as a pure design-validation note, they should soften the precision statements accordingly. The prototype and pool-test sections are solid and should be preserved."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a modest, honest instrumentation paper. The genuinely new parts are the Upgrade-specific camera design (three cameras plus illumination modules per DOM), the swimming pool verification, and the first Monte-Carlo-based precision estimates for this configuration. The hardware story looks solid: the prototype images in Fig. 3 show real capability, and the pool test resolving two LEDs 40 cm apart at 25 m with ~10 cm uncertainty is a concrete, useful result.\n\nWhere it gets softer is the simulation section. The headline precision numbers—centimeter-level bubble column radius and O(10 m) bulk scattering length—come from a self-consistency exercise: the simulation is calibrated to a prototype air measurement, then used to generate images with known input ice parameters, then fit back to those parameters. That is a reasonable feasibility study, but it is not an external validation.\n\nThe specific concern from the stress-test is on target. The O(10 m) bulk-ice claim is conditioned on 'the LED orientation is known within a few degrees,' and the paper never shows that the orientation is identifiable in the same fit. The simulations use five discrete tilts and subtract the per-tilt chi-square minimum; that does not establish that tilt and scattering length are separately measurable in a continuous joint fit. Table 1 specifies a 5-degree camera alignment tolerance but nothing about how the illumination module orientation is controlled or determined at deployment. The claim that multiple depth measurements reduce the orientation uncertainty is an assertion without a supporting combined-fit simulation. So the few-meter final claim in the last line of Sec. 4 is not yet supported.\n\nNone of this kills the design. The system is still worth having for the hole-ice images and geometry information. But the bulk-ice scattering-length precision should be read as 'expected under ideal orientation knowledge,' not 'demonstrated.' The paper's own language is mostly careful—'suggest', 'indicate'—so this is a gap in analysis depth rather than an overreach in tone. Minor additional gripe: Fig. 5 right shows chi-square differences without statistical uncertainties, so even the relative sensitivity is not fully quantified.\n\nBottom line: good proceedings material, deserves a referee. I'd accept after a minor revision that either adds a joint fit simulation or explicitly labels the precision claims as preliminary. This is a paper for IceCube collaborators and instrumentation people; I wouldn't cite it in my own work but I'd bring it to a detector reading group.","headline":"A solid, honest instrumentation proceedings paper: the hardware and pool tests are real, but the simulated bulk-ice precision claims lean on an untested LED-orientation assumption.","tokens_in":7154,"tokens_out":2323,"would_cite":false,"duration_ms":23592,"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":"IceCube Upgrade cameras can map refrozen ice to within a few centimetres, simulations show.","keywords":["IceCube Upgrade","camera system","refrozen hole ice","bubble column","scattering length","ice optical properties","detector calibration","neutrino detector"],"falsifier":"Deploy the camera system in a borehole with an independently known bubble-column radius and a measured LED tilt, then compare the image-derived radius and chi-squared scattering-length estimate against those known values; a discrepancy larger than the quoted few-centimetre and 10-metre precisions would disprove the claim.","tokens_in":6251,"feed_emoji":"📷","tokens_out":3613,"duration_ms":97293,"temperature":0.7,"pith_summary":"The paper establishes that a compact CMOS camera and LED illumination system, built for the IceCube Upgrade optical modules, can photograph the refrozen ice in each drill hole and the glacial ice between strings. Simulation studies calibrated against lab measurements show that images of a bubble column in the refrozen ice yield its radius to a precision of a few centimetres, and that images of scattered light from an LED on a neighbouring string can determine the bulk-ice effective scattering length to roughly 10 metres when the LED orientation is known to within a few degrees. If these precisions hold in the deployed detector, the system gives IceCube an in-situ calibration of the two ice regions that most affect Cherenkov light propagation, and provides geometry information on module positions and cable locations. The authors present the prototype hardware, the quality-control tests, and a swimming-pool demonstration that two LEDs 40 cm apart can be separated at 25 m distance.","feed_headline":"Upgrade cameras can map refrozen ice to within a few centimetres","feed_subtitle":"The same system measures bulk-ice scattering to about 10 metres, sharpening neutrino reconstructions.","key_machinery":"The system pairs a CMOS image sensor with a 170-degree-field-of-view lens (about 120 degrees effective in ice) and a 465 nm illumination module whose 40-degree air cone narrows to 22.5 degrees in ice. The predictive machinery is a photon-propagation Monte Carlo using the SPICE ice model, the same code used for Cherenkov light in IceCube analyses; a photon-to-pixel scaling factor derived from an air test setup converts simulated arrival directions into expected images. The bubble-column analysis measures the visible column footprint, while the bulk-ice analysis builds delta-chi-squared maps comparing simulated images against a 40-metre scattering-length, zero-tilt template.","core_discovery":"The central claim is that a camera and LED pair inside every upgraded optical module can turn IceCube's ice into an imageable calibration target. A downward-facing camera viewing an upward-pointing LED in the DOM below maps the refrozen hole ice, including the central bubble column; a sideways camera viewing an LED on an adjacent string maps scattering in the bulk ice. In Monte Carlo, the bubble-column radius is recovered to a few centimetres, and a chi-squared comparison of measured images to high-statistics templates recovers effective scattering lengths of 30 to 50 metres to about 10 metres precision, provided LED tilt is known within a few degrees. The same light-cone positions give the orientation and position of DOMs and cables, which can calibrate detector geometry in Monte Carlo simulations.","pith_inferences":["If the air-derived pixel calibration does not transfer to ice, the centimetre and 10-metre numbers would degrade; an in-ice cross-check against a known LED brightness and geometry would settle this before scientific deployment.","The same image archive could double as a dust log for the refrozen column, since bubble and impurity distributions are visible in the reflected-light images.","Because the illumination modules are monochromatic at 465 nm, extending the design to a second wavelength could separate scattering from absorption in the bulk-ice measurement.","The demonstrated centimetre-level separation of two LEDs at 25 metres suggests the camera could also track slow DOM motion in the hole during the refreezing phase, not just static geometry."],"forward_implications":["If deployed as planned, every pDOM, D-Egg, and mDOM will carry three cameras, giving overlapping views of both the hole ice and the surrounding glacial ice.","Measuring the refrozen-ice bubble column and bulk-ice scattering length in situ should reduce the dominant optical uncertainties in IceCube event reconstruction.","Comparing images taken with LEDs pointing in different directions allows studies of anisotropy in the ice optical properties.","Light-cone positions in the images provide DOM orientation and cable locations, feeding directly into Monte Carlo geometry calibration.","Repeated measurements at each depth should shrink the LED-orientation uncertainty, pushing scattering-length precision toward a few metres."],"supporting_citations":[{"why":"Supplies the SPICE ice model used as the optical medium in the photon-propagation simulations.","marker":"[2]"},{"why":"Previous camera studies on which the present simulation tool and concept are based.","marker":"[11]"},{"why":"Documents the earlier detection of the refrozen-hole-ice bubble column that this system aims to measure.","marker":"[12]"},{"why":"Provides the photon-propagation Monte Carlo code that simulates LED light in the ice.","marker":"[14]"},{"why":"A similar camera system successfully deployed in Antarctic ice, demonstrating the operating principles.","marker":"[15]"}],"fun_headline_variants":["IceCube upgrade cameras map refrozen ice to centimetres","New camera system sharpens IceCube reconstructions","Cameras in IceCube DOMs image hole ice for calibration","Borehole cameras calibrate IceCube geometry and ice optics"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The quoted precisions rest on a simulation chain whose photon-to-pixel calibration was measured in air and whose ice model is SPICE; if that calibration does not hold in the refrozen or bulk ice, or if the deployed LED orientation departs from the assumed few degrees, the centimetre and 10-metre numbers would not be reached.","fun_headline_variants_meta":{"raw":{"variants":["IceCube upgrade cameras map refrozen ice to centimetres","New camera system sharpens IceCube reconstructions","Cameras in IceCube DOMs image hole ice for calibration","Borehole cameras calibrate IceCube geometry and ice optics"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000178,"raw_usage":{"total_tokens":1314,"prompt_tokens":983,"completion_tokens":331,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":599,"completion_tokens_details":{"reasoning_tokens":264}},"tokens_in":599,"tokens_out":331,"duration_ms":6728,"temperature":1.0,"reasoning_tokens":264,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:57:50.185639+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Deploy the camera system in a borehole with an independently known bubble-column radius and a measured LED tilt, then compare the image-derived radius and chi-squared scattering-length estimate against those known values; a discrepancy larger than the quoted few-centimetre and 10-metre precisions would disprove the claim.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the SPICE ice model used as the optical medium in the photon-propagation simulations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Previous camera studies on which the present simulation tool and concept are based."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents the earlier detection of the refrozen-hole-ice bubble column that this system aims to measure."},{"cited_title":"Chirkin for the IceCube Collaboration, Nucl","cited_arxiv_id":null,"evidence_quote":"Provides the photon-propagation Monte Carlo code that simulates LED light in the ice."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"A similar camera system successfully deployed in Antarctic ice, demonstrating the operating principles."}],"review_version":1}