{"id":"784e316a-82c1-4a59-afd1-3cd8cf1bfb35","arxiv_id":"1908.11697","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A compact 4,617-source celestial reference frame was built by stacking normal equations from 8, 24, and 32 GHz VLBI catalogs with full covariance transfer.","lead":"The authors combined 40 years of radio telescope data at three frequencies into a single celestial reference frame, preserving the full uncertainty relationships between all 4,617 source positions. The result gives astrometry, geodesy, and deep space navigation a rigorous, multifrequency reference frame prototype for future ICRF realizations.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"XKa weighting is tuned to suppress rotations that are then reported as absent; alignment/deformation claims are circular.","rationale":"The reader identified the core-shift assumption as the weakest point, and that is a genuine concern: the Sec. 3 angular-separation test has low power because individual source positions are too inaccurate, and the two K-band outliers are dismissed as artifacts. However, I consider the more load-bearing problem to be the circularity in the XKa weighting. The central claim of a deformation-free, rigorously combined frame is validated by comparisons to ICRF3 after the XKa weight has been tuned specifically to make the XKa-induced R2 rotation disappear. This is not an attack on the normal-equation stacking itself; the stacking method and full covariance transfer are plausible, and the catalog is a useful product. Yet until the weighting is derived from data, for example by variance-component estimation, or the sensitivity to the tuning is shown to be negligible, the reported alignment and absence of deformation should not be treated as independent evidence. The core-shift omission remains under-tested and may affect individual positions, but it does not by itself invalidate the frame-level claim if the offsets are random. The proposed sensitivity experiment would settle whether the 3 µas alignment is a genuine property of the combination or a consequence of the chosen down-weighting. Because both issues warrant further analysis but the method remains viable, the reader's CONDITIONAL verdict is appropriate; no change is needed.","tokens_in":24426,"tokens_out":6470,"duration_ms":62167,"concrete_test":"Re-run the ComboKXKa combination without re-tuning: set the XKa variance factor to 1 and to 2, and add noise floors w in {0, 0.01, 0.02, 0.05, 0.1, 0.2} mas^2; compute R1, R2, R3, |R|, and |D| with respect to ICRF3 for each setting. If R2 varies from about -30 µas at w=0 to near zero only at the chosen w=0.05, the alignment claim is an artifact of the tuning. As a complementary check, run a formal Helmert variance-component estimation for XKa from the post-fit residuals of the combination and compare the estimated variance factor and implied noise floor with the ad hoc values 2 and 0.05 mas^2 used in the paper.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing flaw is the empirically tuned XKa down-weighting in Sec. 4, step 3. The authors state that no a posteriori variance factor was available for XKa, so they set it to 2; then, because \"initial results have shown that the XKa solution introduces significant rotations around R2,\" they chose a weighting parameter of 0.05 mas^2 \"in such a manner that the originally dominant rotations are just no longer discernible in the residuals.\" Section 5.2 then reports R2 = -2 ± 1 µas with respect to ICRF3 and notes that \"without the scaling this parameter amounts to -30 µas and clearly dominates the rotations.\" This is a circular validation: a free parameter is adjusted until the frame defect disappears, and the absence of that defect is later cited as evidence that the frame is deformation-free. The reported 3 µas axis stability and \"no significant deformations\" are therefore not independent outcomes of the combination; they are partly enforced by construction. This also weakens the \"full variance-covariance\" claim, because a hand-set variance floor, not an estimate from the data, controls the XKa contribution and the final formal errors. This concern is more direct than the core-shift issue: even if core shifts are truly random, the deformation claim is still not independently tested.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a method for combining multifrequency VLBI celestial reference frame catalogs (SX, K, XKa) via stacking of normal equation systems, thereby carrying full variance-covariance information into the combined solution. The resulting catalog, ComboKXKa, contains 4617 sources and is compared with ICRF2, ICRF3, and Gaia-CRF2 using vector spherical harmonic decomposition. The central claims are that the combination is rigorous, that the frame is aligned with ICRF3 within 3 microarcseconds, that no significant deformations are present, and that core-shift effects are negligible at current precision.","tokens_in":24792,"tokens_out":4830,"duration_ms":39612,"significance":"If the claims hold, the paper provides a useful methodological advance for future ICRF realizations: it is the first application of full covariance transfer through normal equation stacking to a multifrequency VLBI frame, and the resulting catalog with cross-frequency covariance information would be a valuable community product. The authors also ship the catalog and covariance matrix publicly, which is a concrete strength. However, the central validation is weakened by the empirical tuning of the XKa weight, the self-referential comparison with ICRF3, and the admitted low power of the core-shift test, so the 'deformation-free' claim is not established as an independent outcome.","major_comments":[{"comment":"The XKa down-weighting is circular with respect to the deformation claim. Section 4 step 3 states that the weighting parameter of 0.05 mas^2 'was determined empirically in such a manner that the originally dominant rotations are just no longer discernible in the residuals,' and Section 5.2 reports that for ComboKXKa 'the rotation in R2 is very small' with the explicit note that without the scaling it would be -30 microas. The absence of a significant R2 rotation is therefore enforced by construction rather than demonstrated, yet the abstract and conclusions cite the small rotations as evidence that 'No significant deformations can be identified.' This circularity also affects the stated axis stability of 3 microas, since the stability is estimated from the same weighted solution. The authors should present the weighting as a deliberate regularization choice, quantify the resulting systematic uncertainty (e.g., by repeating the analysis with a range of weights), and avoid citing the suppressed R2 as an independent validation.","section":"Sec. 4, step 3 and Sec. 5.2"},{"comment":"The alignment claim in the abstract ('aligned with ICRF3 within 3 microas') is not supported by the reported transformation parameters. Table 3 gives |R| = 13 +/- 1 microas for ComboKXKa with respect to ICRF3 and |R| = 31 +/- 7 microas with respect to ICRF2. The 3 microas value that appears in Sec. 5.2 refers to the scatter of rotation parameters across different source subsets (axis stability), not to the absolute alignment. The abstract and Sec. 6 should be reworded to distinguish 'axis stability of 3 microas' from 'rotation magnitude of about 13 microas with respect to ICRF3,' and the wording in Sec. 6 ('aligned with ICRF2 within +/-3 microas') should be corrected; as written it is inconsistent with Table 3.","section":"Abstract and Sec. 6"},{"comment":"The comparison of the combined product with ICRF3 is partly self-referential. As the paper notes, the GSF SX solution 'is identical in content to that used for the determination of the SX catalog of ICRF3,' so the small deformation parameters between ComboKXKa and ICRF3 largely reflect the propagation of the same SX input. Because the K and XKa catalogs are down-weighted and the XKa weight is empirically tuned, the agreement with ICRF3 does not provide an independent external check of the frame. The authors should state this limitation explicitly when interpreting the ICRF3 comparison, and should place more weight on the Gaia-CRF2 comparison (which they admit is inconclusive) or on comparisons against realizations not built from the same input.","section":"Sec. 2.1.1 and Sec. 5.2"},{"comment":"The core-shift analysis does not substantiate the claim that core shifts 'only add white noise.' The angular-separation test in Fig. 3 has low power, as the authors themselves conclude: 'the failure of any statistical testing of core shifts of a large number of sources, as the individual source position is too inaccurate at the current state and the standard deviations are too optimistic.' Absence of detection is not evidence of random orientation, and the two K-band outliers (3C119, 2018+295) are dismissed as analysis artifacts without a quantitative argument beyond their large separations. The conclusion in Sec. 1 that 'we demonstrate that the effect is of a random nature for catalog combinations' therefore overstates what the data show. The paper should present the randomness of core shifts as an assumption required by current precision, not as a demonstrated property, and should discuss the potential impact of a systematic frequency-dependent component on the combined positions.","section":"Sec. 3"}],"minor_comments":[{"comment":"The column header 'CommboKXKa' contains a typo; it should be 'ComboKXKa'.","section":"Table 3"},{"comment":"The abstract uses 'Ghz' but the standard unit symbol is 'GHz'; please correct for consistency.","section":"Abstract and throughout"},{"comment":"In the text, 'aM2,0 which describes a sharing of the two hemispheres' should read 'shearing,' not 'sharing,' to match the earlier definition in Sec. 2.4.","section":"Sec. 5.4"},{"comment":"The column labels 'wmean' and 'σwmean' are ambiguous; consider relabeling as 'weighted mean' and 'weighted standard deviation' or adding a footnote explaining the notation.","section":"Tables 1 and 4"}],"recommendation":"major_revision","confidential_remarks":"The manuscript makes a useful methodological contribution and provides a public data product, but the central claim of a deformation-free frame is not independently validated due to the empirically tuned XKa weight and the partial self-reference to ICRF3. The authors can likely address this in a revision by reframing the weighting choice as a robustness analysis and by clearly separating the axis-stability statement from the absolute alignment numbers. The paper is within the scope of A&A astrometry; I recommend major revision rather than rejection because the core combination machinery is sound and the limitations are fixable in the text and analysis."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper builds a combined CRF from SX, K, and XKa inputs and delivers a 4617-source catalog with full covariance across all frequencies. That product is new: ICRF3 offers separate aligned catalogs, not a single stacked solution with complete covariance transfer. The NEQ reconstruction from the XKa covariance matrix, the Helmert blocking, and the datum handling are standard and carefully laid out. I like that they verify their NEQ pipeline by reproducing the near-zero GSF-vs-ICRF3 rotations, and the VSH residual analysis is thorough. They are also candid about the Gaia comparison being inconclusive and about the core-shift caveat. That honesty earns real credit.\n\nThe soft spot is exactly the one the stress test flags. In Sec. 4, step 3, they set the XKa variance factor to 2 without justification, then choose the 0.05 mas^2 weight empirically 'in such a manner that the originally dominant rotations are just no longer discernible.' The abstract and Sec. 5.2 then report R2 = -2 ± 1 µas and 'no significant deformations' as if these were independent outcomes. They are partly enforced by construction. This does not break the combination machinery, but it means the rotation-free claim is not an independent result, and the reported formal errors inherit a hand-set variance floor rather than a data-driven estimate. I also note the XKa solution predates the ICRF3 XKa catalog by six months, and the GSF-vs-ICRF3 comparison is near self-reference since GSF is the ICRF3 SX input.\n\nThe core-shift treatment is reasonable but limited. The random-orientation assumption is plausible and supported by Plavin et al., yet the angular-separation test has low power because individual positions are too inaccurate, and the two K-band outliers are dismissed rather than explained. The authors actually acknowledge this in the text, which is good, but it means the multifrequency combination still rests on an assumption rather than a demonstrated null effect.\n\nOverall the central method holds up. The product is a useful prototype for future ICRF realizations and for frame ties, and the full covariance transfer is a real step forward. The flaws are in the validation narrative, not in the stacking approach. A serious referee should ask them to present the XKa weighting as a tuned parameter, to soften the independent-validation language, and to show how the final formal errors depend on that choice.\n\nBottom line: this deserves peer review. It is a solid, specialized contribution to VLBI astrometry and reference frame work, and the main claim needs honest qualification rather than rejection. I would bring it to a reference-frame reading group and would cite it if I were working on CRF combination methodology.","headline":"Useful and genuinely novel as a first full-covariance multifrequency VLBI frame combination, but the headline deformation-free claim is partly enforced by a tuned XKa weight, so the validation is weaker than the abstract suggests.","tokens_in":25194,"tokens_out":1468,"would_cite":true,"duration_ms":15971,"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":"Astronomers stack three radio frequency bands into one celestial reference frame that stays aligned with ICRF3 to within 3 microarcseconds and shows no detectable deformation.","keywords":["celestial reference frame","ICRF3","VLBI","normal equation systems","Helmert blocking","core shift","multifrequency astrometry","covariance transfer"],"falsifier":"Measure the angular separation between radio cores of the same compact sources at 8, 24, and 32 GHz with a VLBI campaign reaching per-source precision well below 0.1 mas; if the offsets are not isotropically distributed but systematically aligned with jet directions or grow with frequency separation for many sources, the white-noise assumption fails.","tokens_in":24213,"feed_emoji":"📡","tokens_out":3829,"duration_ms":36313,"temperature":0.7,"pith_summary":"This paper tries to establish that a celestial reference frame built by stacking independent VLBI catalogs at 8, 24, and 32 GHz, rather than by aligning finished position lists, carries the full covariance information across all sources and frequencies, and that the resulting ComboKXKa frame is free of detectable deformation. A sympathetic reader would care because the current ICRF3 treats its three frequency catalogs as standalone products aligned through common sources, whereas this method jointly re-determines common sources from all observations. The claimed product contains 4,617 compact radio sources, aligns with ICRF3 within 3 microarcseconds at the defining sources, and shows an average positional uncertainty of 0.1 mas in right ascension and declination. The paper also argues that core shifts are not an obstacle at current VLBI precision because their offsets have random orientation and add only white noise.","feed_headline":"One stacked catalog merges 4,617 radio sources across three bands","feed_subtitle":"Full covariance transfer through normal-equation stacking keeps the combined frame aligned with ICRF3 to 3 microarcseconds.","key_machinery":"The load-bearing object is the normal equation system (NEQ) from each VLBI global solution, with all non-source parameters already eliminated, carrying the variance-covariance relationships of the estimated source positions. The combination itself is Helmert stacking: the NEQ matrices and right-hand side vectors of common parameters are summed, a no-net-rotation datum is imposed with the ICRF2 defining sources, and the stacked system is solved. Because the XKa input arrived as a covariance matrix rather than NEQs, it is first converted back to datum-free normal equations using the Grafarend-Sanso identity; the XKa system is then scaled down empirically so its network-geometry rotations no longer dominate. This machinery makes the full covariance matrix of all 4,617 sources, across frequency bands, a direct output of the combination rather than an a posteriori construct.","core_discovery":"The paper's central claim is that ComboKXKa is a rigorous, multifrequency realization of the ICRS: 4,617 compact radio sources positioned jointly from SX (8 GHz), K (24 GHz), and XKa (32 GHz) observations, with all variance-covariance information carried through the combination. The frame is aligned with ICRF3 to within 3 microarcseconds and has an average positional uncertainty of 0.1 mas in both coordinates; rotation and deformation parameters from vector-spherical-harmonic analysis show no significant deformations once the XKa solution is down-weighted. Adding the higher-frequency catalogs also extends the frame southward, since 16 of the 31 XKa-only sources lie below -30 degrees declination. Comparisons with Gaia-CRF2 are called inconclusive because the transformation parameters depend strongly on source selection, yet the paper states that significant differences between all frames are attested.","pith_inferences":["A dedicated VLBI campaign measuring the same compact sources in all three bands with matched networks could convert the random core-shift assumption into a measured correction; if core-shift vectors align with jet position angles, the stacked frame would need per-source frequency offsets.","The empirical down-weighting of XKa (a variance factor of 2 plus an additional 0.05 mas-squared inflation) indicates the combination is only as good as the weakest geometry; future frames could formalize such weights from VSH residuals instead of tuning them.","If Gaia DR3 confirms the radio-optical differences seen here, the combined multifrequency radio frame could serve as a clean intermediate frame for disentangling optical structure effects from true source position offsets."],"forward_implications":["The same pipeline can be applied to future ICRF realizations directly from normal equations of all analysis centers, replacing the monolithic-solution-plus-alignment scheme.","Common sources present in more than one catalog are effectively re-determined from the union of observations, which improves formal errors for some sources and mitigates network deficiencies.","The southern extension from XKa-only sources improves sky coverage in the deep south, a region where SX catalogs remain sparsely populated.","The full covariance matrix enables proper statistical interpretation of the frame, including frame ties and comparisons with Gaia, without requiring ad hoc error inflation.","The absence of detectable deformation suggests that the weak network geometry of the XKa solution can be prevented from propagating into a stacked product by appropriate weighting."],"supporting_citations":[{"why":"Supplies the three input catalogs (SX, K, XKa) and the ICRF3 alignment baseline that the combined frame is compared against.","marker":"Charlot et al. 2019"},{"why":"Provides ICRF2 and its defining sources, which are used for the celestial datum, quick-look residual checks, and the ICRF2 comparisons.","marker":"Fey et al. 2015"},{"why":"Supplies the no-net-rotation condition used to impose the celestial datum on the stacked normal equation systems.","marker":"Jacobs et al. 2010"},{"why":"Provides the classical Helmert blocking/stacking formalism by which the individual NEQ matrices and right-hand side vectors are summed.","marker":"Helmert 1872"},{"why":"Establishes the feasibility of rigorous VLBI solution combination with transfer of full variance-covariance information, the foundation of the method.","marker":"Iddink et al. 2015"},{"why":"Supplies the angular-separation and normalized-separation test used to assess core-shift evidence between the frequency catalogs.","marker":"Mignard, F. et al. 2016"},{"why":"Provides the measured core-shift magnitudes between frequencies that motivate the random-orientation white-noise assumption.","marker":"Plavin et al. 2019b"},{"why":"Supplies the identity used to reconstruct datum-free normal equations from the XKa covariance matrix.","marker":"Grafarend and Sanso 1985"}],"fun_headline_variants":["A three-band celestial frame from 4617 sources, aligned to ICRF3","Multifrequency frame: 4617 sources via normal-equation stacking","Stacked VLBI data produce 4617-source frame with full covariance","New CRF: 4617 sources at SX, K, XKa, ICRF3-aligned to 3 µas","Combining 40 years of VLBI into a 4617-source multifrequency frame"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that frequency-dependent source position offsets, or core shifts, are random in orientation and too small to be detected at current VLBI precision, so combining 8, 24, and 32 GHz positions adds only white noise.","fun_headline_variants_meta":{"raw":{"variants":["A three-band celestial frame from 4617 sources, aligned to ICRF3","Multifrequency frame: 4617 sources via normal-equation stacking","Stacked VLBI data produce 4617-source frame with full covariance","New CRF: 4617 sources at SX, K, XKa, ICRF3-aligned to 3 µas","Combining 40 years of VLBI into a 4617-source multifrequency frame"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00093,"raw_usage":{"total_tokens":4018,"prompt_tokens":1015,"completion_tokens":3003,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":631,"completion_tokens_details":{"reasoning_tokens":2887}},"tokens_in":631,"tokens_out":3003,"duration_ms":19227,"temperature":1.0,"reasoning_tokens":2887,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T10:08:03.745468+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the angular separation between radio cores of the same compact sources at 8, 24, and 32 GHz with a VLBI campaign reaching per-source precision well below 0.1 mas; if the offsets are not isotropically distributed but systematically aligned with jet directions or grow with frequency separation for many sources, the white-noise assumption fails.","supporting_citations":[{"cited_title":"The Third Realization of the International Celestial Reference Frame by Very Long Baseline Interferometry","cited_arxiv_id":null,"evidence_quote":"Supplies the three input catalogs (SX, K, XKa) and the ICRF3 alignment baseline that the combined frame is compared against."}],"review_version":1}