{"id":"0c627e5e-0e80-4844-861a-2344ca00d3f6","arxiv_id":"2506.19933","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A new infrared astrometric reference frame for the Galactic center, built from 14 epochs of Hubble data tied to Gaia, reaches 0.03 mas/yr precision and is consistent with radio maser frames.","lead":"Astronomers used 14 years of Hubble images of the Milky Way's center, tied to the Gaia satellite's coordinate grid, to make the most precise star-motion catalog yet within one parsec of the central black hole. The new reference frame is accurate enough to weigh the invisible mass around Sagittarius A* and to cross-check radio measurements of maser stars.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Frame-consistency claim is calibrated on the same 40 stars used to build the frame; a leave-one-out cross-check against Gaia-DR3 is needed to show the 0.025 mas/yr tie is not self-referential.","rationale":"The reader's weakest assumption identifies the same risk: the 40 primary reference stars are assumed to be an unbiased realization of the ICRS after iterative outlier rejection. My stress-test sharpens this into a concrete circularity: the Section 3.3 consistency numbers are measured on the very stars used to define the transformation, so they measure internal agreement, not external alignment with Gaia-CRF3. The paper is otherwise technically careful: the Gaussian-process modeling is a genuine methodological contribution, the PA-dependent systematic is documented and modeled, and the comparison to radio masers is an honest external check even though it currently shows a possible ~2.5-sigma offset. The conditional verdict is therefore appropriate. I do not see an internal inconsistency or a reason to move to reject; the missing piece is an independent validation of the 0.025 mas/yr error floor, which a leave-one-out test on the existing 40 reference stars could provide immediately, and which Gaia-DR4 should eventually provide externally.","tokens_in":31171,"tokens_out":3256,"duration_ms":36468,"concrete_test":"Run leave-one-out cross-validation on the 40 primary reference stars: for each star, recompute the second-order polynomial transformation (including the same 5-sigma outlier rejection on the remaining 39 stars), predict that star's position and proper motion at the common epoch 2016.0605, and compare with Gaia-DR3. Report the RMS and error-weighted scatter of the 40 held-out residuals in alpha* and delta. If the held-out scatter exceeds the quoted 0.025 mas/yr in proper motion or 0.044 mas in position by more than ~1.5x, the frame-consistency claim is not supported by the internal data. As a secondary check, re-run the transformation with all 50 original Gaia candidates included to quantify how strongly the quoted bias depends on the 10-star outlier cut.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim (abstract and Section 3.3), that the HST-Gaia frame agrees with Gaia-CRF3 to 0.025 mas/yr in proper motion and 0.044 mas in position, is computed from the error-weighted mean residual of the 40 primary reference stars that also define the transformation. This is not an independent external consistency test: the second-order polynomial transformation is fit by linear least squares to these same stars, so residual scatter on them is minimized by construction. Moreover, the 40 stars are the survivors of iterative 5-sigma outlier rejection that removed 10 of the original 50 Gaia candidates (Section 3.1). If the rejected stars carry real astrometric offsets, for example from crowding, saturation, or the PA-dependent systematic in Appendix A, then the retained sample is biased toward zero residual and the quoted bias uncertainties are underestimated. The bootstrap-derived sigma_trans in Section 3.1 measures internal transformation precision, not accuracy relative to the ICRS. No quasars are observable in this field, so there is no direct external ICRS anchor. The only independent check is the 13-maser radio comparison (Section 4.2.3), which shows a possible ~2.5-sigma offset in alpha* for both position and proper motion and is dominated by SgrA* position uncertainties. The headline claim that this is the first ICRS frame precise enough to probe extended mass rests on an error floor that has not yet been validated against an independent ICRS realization.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a new near-infrared astrometric reference frame near the Galactic center, constructed by transforming 14 epochs of HST WFC3-IR astrometry into the Gaia-CRF3 frame using 40 Gaia-DR3 stars that overlap the HST field. The authors introduce a Gaussian-process framework for modeling stellar proper motions that simultaneously handles time-correlated systematics (e.g., a position-angle-dependent shift of up to ~2 mas for faint stars) and spatially correlated confusion. The resulting catalog contains 2,876 sources within ~25\" of SgrA*, with the brightest stars achieving median proper-motion and position errors of ~0.03 mas/yr and ~0.11 mas, respectively. The central claim is that the HST-Gaia frame is consistent with Gaia-CRF3 to 0.025 mas/yr in proper motion and 0.044 mas in position, and that it is the first ICRS-based frame precise enough to probe the extended mass distribution near SgrA*. The authors also compare the frame to the radio maser-based frame of Darling et al. (2023), finding consistency within 0.041 mas/yr and 0.54 mas at 99.7% confidence, while noting a possible ~2.5-sigma offset in the alpha* direction.","tokens_in":31505,"tokens_out":8665,"duration_ms":91835,"significance":"If the frame-accuracy claim is correct, this is a valuable contribution: it is the first ICRS-tied NIR reference frame in the central parsec that is independent of radio masers, and the catalog (with ~2,900 sources, including 40 Gaia primary stars and 13 masers) will be a useful resource for GC dynamics. The Gaussian-process methodology for modeling systematic errors is a genuine methodological advance, and the paper provides a detailed pipeline description with machine-readable tables. The validation is reasonably thorough: the GP models are tested against leave-one-out cross-validation and the g-chi-squared_LOO distributions, and the radio maser comparison is an independent external check. However, the headline consistency with Gaia-CRF3 is not an independent validation, since it is measured on the same 40 stars that define the transformation; the only fully external test, the radio comparison, shows a possible ~2.5-sigma offset in alpha*. The significance of the extended-mass claim therefore rests on a precision estimate that is partly self-referential, and this needs to be addressed before the central claim can be accepted.","major_comments":[{"comment":"The quoted 0.025 mas/yr proper-motion and 0.044 mas position consistency with Gaia-CRF3 is computed as the error-weighted mean residual of the 40 primary reference stars that are the same stars used to fit the second-order polynomial transformation. For a least-squares fit, the residuals on the fitting sample are minimized by construction, so these numbers quantify the internal scatter of the transformation residuals (i.e., its precision relative to the same data), not an independent measure of the accuracy of the frame tie to the ICRS. The authors should either add a leave-one-out or split-sample cross-validation (fit the transformation on a subset of primary stars and test on the remaining stars) or explicitly relabel the §3.3 result as an internal consistency check. This is load-bearing because the abstract and §5.3 use the 0.025 mas/yr value to argue that the frame is precise enough to probe the extended mass distribution near SgrA*.","section":"§3.3, Eq. (10)"},{"comment":"The only fully independent external validation, the 13-maser comparison, shows a possible tension in the alpha* direction: the average proper-motion offset is -0.076 ± 0.030 mas/yr and the position offset is -0.536 ± 0.230 mas (after adding systematics). The paper acknowledges a ~2.5-sigma tension, but the abstract and conclusions state consistency 'to within 0.041 mas/yr and 0.54 mas at 99.7% confidence,' which is misleading because the observed offset is larger than the quoted 1-sigma uncertainty and the 99.7% confidence interval includes non-zero values. The authors should report the offset and its significance explicitly in the abstract and conclusions, and discuss what a real offset of this magnitude would imply for the frame accuracy claim. If the offset is real, it would be comparable to the claimed 0.025 mas/yr precision and would weaken the extended-mass claim.","section":"§4.2.3"},{"comment":"The iterative 5-sigma outlier rejection removed 10 of the 50 Gaia reference stars (20%) from the sample that defines the transformation. The rejection is based on the differences between HST-Gaia proper motions and Gaia-DR3 proper motions, but the HST-Gaia proper motions are computed using the transformation that is fit to the same stars, introducing a circularity into the outlier selection. The paper notes that many removed stars had truncated PSFs or close neighbors, but the possibility that the retained 40-star sample is biased toward zero residual (and thus that the transformation errors are underestimated) should be addressed quantitatively. The authors should report the astrometric properties of the rejected stars and/or demonstrate that the outlier rejection does not alter the conclusions by repeating the analysis with a leave-one-out or a more robust fitting procedure.","section":"§3.1"}],"minor_comments":[{"comment":"The phrase 'consistent with zero to within 0.041 mas/yr at 99.7% confidence' is imprecise; the observed offset is -0.076 mas/yr with a 1-sigma uncertainty of 0.041 mas/yr. Please state the offset and its uncertainty, and phrase the consistency statement in terms of the confidence interval for the true offset.","section":"Abstract / §4.2.3"},{"comment":"The abbreviation 'D23' is used in §4.1 and §4.2 without being defined at first use in the text; please spell out Darling et al. (2023) when the radio maser frame is first mentioned.","section":"§3.1"},{"comment":"The table note lists 'σ_α*, σ_δ, σ_μδ, σ_μδ'; the last two entries should be σ_{μα*} and σ_{μδ}. Also, several table captions contain the typo 'T able'.","section":"Table 3 note"},{"comment":"The ELPD model-selection threshold of ≥6 is stated as corresponding to ~3-sigma preference, but the mapping between the ELPD difference and a significance level is not derived; please provide a reference or a brief justification for this threshold.","section":"§3.2"},{"comment":"The predicted Gaia-DR4 improvement relies on the empirical constant α_obs ≈ 2.4 in Eq. (19), but the derivation of this constant is not described; please state how it is obtained from the current transformations and discuss its uncertainty.","section":"§5.2"}],"recommendation":"major_revision","confidential_remarks":"The paper is well-written and the pipeline is described in unusual detail, with a clear presentation of the GP methodology and validation. The main concern is that the headline consistency with Gaia-CRF3 is measured on the same stars used to build the frame, so it is not an independent accuracy test. The only independent anchor (the radio maser comparison) shows a possible ~2.5-sigma offset in alpha*, which the authors acknowledge but do not adequately foreground in the abstract. I would urge the editor to require either a split-sample validation or a clear reframing of the 0.025 mas/yr claim as an internal precision estimate, and to ensure that the alpha* tension is reported prominently. The catalog and method are valuable, so I do not recommend rejection, but the central claim needs to be made more defensible."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here’s my take. The substantive new thing is a ~2,900-star HST WFC3-IR proper-motion catalog in the central parsec, tied to ICRS via Gaia DR3, plus a Gaussian-process framework for modeling epoch-correlated systematics in HST astrometry. That is a genuinely useful product for Galactic-center dynamics. The paper is careful about measurement errors, uses bootstrap transformation uncertainties, checks its kinematic models with leave-one-out diagnostics, and is unusually honest about the PA-dependent systematic it found. The citation pattern looks appropriate, with proper benchmarks against Plewa, Sakai, and Griggio.\n\nThe main soft spot is the headline 0.025 mas/yr consistency with Gaia-CRF3. That number is computed from the same 40 primary stars used to define the second-order polynomial transformation. A least-squares fit will absorb smooth distortions and reduce residuals on those stars, so the residual scatter mostly measures internal fit consistency, not how well the frame independently realizes the ICRS. Calling it “consistent with Gaia-CRF3 to 0.025 mas/yr” overstates the external validation. This is not fatal: the 13-maser radio comparison is an independent external test and it broadly agrees, but it is limited in power, dominated by SgrA* position uncertainties, and shows a 2.5-sigma offset in alpha* that the authors acknowledge. Also, 10 of 50 Gaia candidates were iteratively rejected as outliers; if those stars carry real astrometric signal, the tie inherits that bias. The paper does not quantify sensitivity to that cut.\n\nSecondary issues: the DR4 projection rests on an empirical scaling constant, alpha_obs = 2.4, so treat that factor-of-two forecast as an estimate rather than a measurement. No analysis code is released, which makes the GP model harder to audit, although the appendix is fairly detailed. None of these are load-bearing flaws for the catalog itself.\n\nOverall, this is a serious, honest paper. It is not a paradigm shift, but it is the first ICRS-tied infrared reference frame in this crowded field and it will likely be widely used. It deserves peer review rather than desk rejection. I would ask for a leave-one-out cross-validation of the Gaia tie, fitting on 39 stars and predicting the held-out star, to show the 0.025 mas/yr figure is not just the fit absorbing the data; a sensitivity analysis to the outlier-removal step; and release of the analysis code and machine-readable tables. With those revisions I would be comfortable citing the catalog and the method.","headline":"A solid, carefully-built ICRS-tied reference frame for the Galactic center; the headline Gaia-consistency number is partly self-referential, but the independent maser check and the GP proper-motion method make this a serious referee-worthy paper.","tokens_in":32072,"tokens_out":2620,"would_cite":true,"duration_ms":30524,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Fourteen epochs of HST imaging, anchored by 40 Gaia-DR3 stars, define the first near-infrared ICRS-tied reference frame at the Galactic center, with a systematic tie to Gaia-CRF3 of $0.025$ mas yr$^{-1}$ in proper motion and $0.044$ mas…","keywords":["Galactic center","astrometric reference frame","proper motions","ICRS","Gaia-CRF3","Gaussian process","Sgr A*","Hubble Space Telescope"],"falsifier":"Recompute the HST-to-Gaia-CRF3 transformation after re-adding the ten rejected Gaia-DR3 stars with conservative weights and compare the resulting proper-motion bias; a shift larger than $0.025$ mas yr$^{-1}$ would show the outlier rejection biases the frame. Independently, fit the 40 primary stars from the two position-angle epochs separately: if the PA-dependent systematic is real, the two half-frame proper-motion ties should differ by more than their formal errors.","tokens_in":30931,"feed_emoji":"🌌","tokens_out":12759,"duration_ms":117327,"temperature":0.7,"pith_summary":"This paper sets out to establish the first high-precision near-infrared astrometric reference frame at the Milky Way's Galactic center that is tied to the International Celestial Reference System (ICRS). Because the central parsec is too dusty and crowded for Gaia to see, the authors bridge the gap by transforming 14 epochs of Hubble Space Telescope imaging into the Gaia-CRF3 frame using 40 stars in common with Gaia-DR3, and they publish a catalog of 2,876 stars around Sgr A*. They report that the frame agrees with Gaia-CRF3 to $0.025$ mas yr$^{-1}$ in proper motion and $0.044$ mas in position, making it, they argue, the first ICRS-based frame precise enough to probe the extended mass distribution inside the orbit of the star S0-2. The catalog is also the first near-infrared reference frame at the Galactic center that is independent of radio maser astrometry, which lets it test the maser-based frames used so far.","feed_headline":"ICRS star frame reaches 0.025 mas/yr at the Galactic center","feed_subtitle":"HST proper motions match Gaia-CRF3 to 0.025 mas/yr, enough to weigh the mass around Sgr A*.","key_machinery":"Three components carry the argument. First, a second-order polynomial transformation fits HST pixel coordinates to Gaia-CRF3 sky coordinates using 40 carefully selected primary reference stars, with bootstrap estimates of the transformation error at each epoch. Second, each star's motion is modeled as a first-order polynomial plus one of several Gaussian-process correlation kernels (squared-exponential, step, confusion, or additive error), selected by leave-one-out cross-validation and an expected-log-probability criterion; this is what absorbs a magnitude- and position-angle-dependent systematic shift of up to $\\sim2$ mas that would otherwise distort proper motions. Third, the frame is converted into SgrA*-at-Rest coordinates by subtracting the adopted ICRS position and proper motion of Sgr A* itself, so the catalog can be used directly for orbital dynamics. The paper identifies the position-angle systematic but leaves its physical origin open.","core_discovery":"The central claim is that the HST-Gaia frame inherits the ICRS orientation through Gaia-CRF3 with no measurable bias: the error-weighted differences between HST-Gaia and Gaia-DR3 for the 40 primary stars are $-0.015\\pm0.020$ mas yr$^{-1}$ in $\\mu_{\\alpha*}$ and $0.002\\pm0.015$ mas yr$^{-1}$ in $\\mu_\\delta$, with a combined proper-motion precision of $0.025$ mas yr$^{-1}$, and position consistency of $0.044$ mas. The brightest catalog stars reach median errors of $0.03$ mas yr$^{-1}$ in proper motion and $0.11$ mas in position, roughly a factor of $20$ better than previous ICRS proper motions in this field. The paper further claims that a comparison of 13 stellar masers with the radio-based reference frame (D23) shows consistency within $0.041$ mas yr$^{-1}$ and $0.54$ mas at $99.7\\%$ confidence, while flagging a possible $\\sim2.5\\sigma$ offset in the $\\alpha_*$ direction that requires further observations to attribute.","pith_inferences":["If the frame tie is as stable as reported, the same Gaussian-process modeling should create ICRS-tied reference frames in other crowded, reddened fields where Gaia stars are sparse but multi-epoch HST imaging exists.","A direct stress test of the central assumption is to repeat the transformation after re-adding the ten rejected Gaia candidates with conservative weights; a shift in the proper-motion bias above $0.025$ mas yr$^{-1}$ would implicate the outlier-rejection step.","The paper's own Table 6 suggests that the position part of the SgrA*-at-Rest comparison is limited by the adopted ICRS position of Sgr A* rather than by HST-Gaia; improving that VLBI anchor would tighten the maser comparison more than extra HST epochs would.","If the position-angle-dependent shift is really intra-exposure pointing drift, it should also appear in other long-exposure WFC3-IR programs with two roll angles, and confirming that would permit a physical correction instead of absorbing the effect into the GP kernel."],"forward_implications":["Galactic-center stellar orbits can now be tied to an inertial ICRS frame rather than only to a radio-maser frame, so dynamical fits gain an independent constraint on the reference frame itself.","With the reported $0.025$ mas yr$^{-1}$ proper-motion consistency, the frame should detect an extended mass of roughly $3000\\,M_\\odot$ inside S0-2's orbit, dropping to roughly $1000\\,M_\\odot$ once Gaia-DR4 improves the tie.","The 2,823 secondary reference stars inside the ground-based AO field allow existing AO datasets to be re-registered to the ICRS without new observations.","The 13-maser cross-check gives the first independent near-infrared confirmation of the radio-maser frame at the $0.041$ mas yr$^{-1}$ level, with the paper leaving open whether the $\\sim2.5\\sigma$ $\\alpha_*$ offset is real."],"supporting_citations":[{"why":"Defines Gaia-CRF3, the frame into which the HST astrometry is transformed; supplies the quasar-based orientation used for the consistency claim.","marker":"Gaia Collaboration et al. 2022a"},{"why":"Provides the Gaia-DR3 astrometric error model and the quasar proper-motion systematic maps used for the HST-Gaia-to-ICRS uncertainty budget.","marker":"Lindegren et al. 2021"},{"why":"Supplies the KS2 software used to combine per-epoch HST images and produce the starlists on which all measurements rest.","marker":"Anderson et al. 2008"},{"why":"Defines an earlier maser-based Galactic-center near-infrared reference frame used as the precision baseline for comparison.","marker":"Plewa et al. 2015"},{"why":"Provides the AO maser-based reference frame and the catalog of 542 common stars used to compare individual star precisions.","marker":"Sakai et al. 2019"},{"why":"The VVV-based ICRS near-infrared proper-motion catalog in the region; sets the factor-of-20 precision improvement claim.","marker":"Griggio et al. 2024"},{"why":"The radio maser reference frame (D23) used for the independent cross-check of 13 masers.","marker":"Darling et al. 2023"},{"why":"Supplies the ICRS position and proper motion of SgrA* used to convert HST-Gaia measurements into SgrA*-at-Rest coordinates.","marker":"Xu et al. 2022"},{"why":"Provides the Gaussian-process formalism and kernel machinery at the heart of the new proper-motion modeling.","marker":"Rasmussen & Williams 2006"}],"fun_headline_variants":["HST-Gaia frame at GC matches ICRS to 0.025 mas/yr","Galactic center proper motions now tied to ICRS at 0.025 mas/yr","HST-Gaia: first ICRS frame at GC, 20x sharper","New HST-Gaia frame: 0.025 mas/yr precision at Sgr A*"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The frame tie rests on the assumption that the 40 Gaia-DR3 stars retained as primary references are an unbiased realization of the ICRS in a crowded, high-extinction field; if the ten rejected candidates carry real astrometric signal, or if the retained 40 have correlated errors, the quoted $0.025$ mas yr$^{-1}$ consistency with Gaia-CRF3 would not hold.","fun_headline_variants_meta":{"raw":{"variants":["HST-Gaia frame at GC matches ICRS to 0.025 mas/yr","Galactic center proper motions now tied to ICRS at 0.025 mas/yr","HST-Gaia: first ICRS frame at GC, 20x sharper","New HST-Gaia frame: 0.025 mas/yr precision at Sgr A*"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00091,"raw_usage":{"total_tokens":4016,"prompt_tokens":1156,"completion_tokens":2860,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":772,"completion_tokens_details":{"reasoning_tokens":2762}},"tokens_in":772,"tokens_out":2860,"duration_ms":23430,"temperature":1.0,"reasoning_tokens":2762,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:22:50.157238+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the HST-to-Gaia-CRF3 transformation after re-adding the ten rejected Gaia-DR3 stars with conservative weights and compare the resulting proper-motion bias; a shift larger than $0.025$ mas yr$^{-1}$ would show the outlier rejection biases the frame. Independently, fit the 40 primary stars from the two position-angle epochs separately: if the PA-dependent systematic is real, the two half-frame proper-motion ties should differ by more than their formal errors.","supporting_citations":[{"cited_title":"2021, , 649, A2","cited_arxiv_id":null,"evidence_quote":"Provides the Gaia-DR3 astrometric error model and the quasar proper-motion systematic maps used for the HST-Gaia-to-ICRS uncertainty budget."},{"cited_title":"2008, , 135, 2055","cited_arxiv_id":null,"evidence_quote":"Supplies the KS2 software used to combine per-epoch HST images and produce the starlists on which all measurements rest."},{"cited_title":"M., et al","cited_arxiv_id":null,"evidence_quote":"Defines an earlier maser-based Galactic-center near-infrared reference frame used as the precision baseline for comparison."},{"cited_title":"2019, , 873, 65","cited_arxiv_id":null,"evidence_quote":"Provides the AO maser-based reference frame and the catalog of 542 common stars used to compare individual star precisions."},{"cited_title":"2024, arXiv e-prints, arXiv:2403.12219","cited_arxiv_id":null,"evidence_quote":"The VVV-based ICRS near-infrared proper-motion catalog in the region; sets the factor-of-20 precision improvement claim."},{"cited_title":"J., et al","cited_arxiv_id":null,"evidence_quote":"The radio maser reference frame (D23) used for the independent cross-check of 13 masers."}],"review_version":2}