{"id":"5031fc8f-68fd-4772-a39e-d71d116394e7","arxiv_id":"2501.18276","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Quad-band VGOS source positions show source-structure-driven variations and offsets from S/X positions, so a dedicated quad-band catalog is needed.","lead":"This paper uses 177 VGOS sessions to track how AGN positions vary in quad-band (3.3, 5.5, 6.6, 10.5 GHz) VLBI observations. It finds that source structure drives position changes and that many quad-band positions differ significantly from legacy S/X positions, arguing for a dedicated quad-band catalog.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 15%/6% offset fractions rest on a hand-drawn declination error floor with no algorithm or uncertainty; if the floor is mis-estimated, the headline quantitative conclusions weaken.","rationale":"The reader identified the error floor in Table 1 as the weakest assumption, and I agree. The paper's central claim—that source structure drives quad-band position variations and that a dedicated quad-band catalog is mandatory—has two components: a qualitative component supported by imaging and detailed case studies, and a quantitative component expressed as 15% and 6% significant-offset fractions. The quantitative component depends directly on the uncertainty inflation procedure described in Section 5, whose input is the declination-dependent error floor. The floor is defined as the lower edge of the residual distribution after a 7-point smoothing, but no algorithm, binning rule, or uncertainty is specified for drawing the grey curve in Fig. 7. Because it is a lower envelope, it is vulnerable to small-sample effects: with ~96 sources, the minimum wrms in a declination bin can be dominated by one or two exceptionally stable sources or by the particular choice of smoothing window. The scaling with sqrt(N_sessions) is also approximate, and the paper itself notes the 500-observation session threshold 'might need further justification.' The Rayleigh comparison is suggestive but not an external check. I do not see a stronger, more load-bearing concern: the paper is transparent about limitations, the data and software are standard, and the case studies provide real evidence for the structure-driven variability. The error-floor issue is precisely the place where the quantitative headline could change under reasonable alternative choices, so it warrants a concrete reproducibility check. If the recomputation yields similar fractions, the conditional verdict should stand as an accept-level result; if the fractions move substantially, the quantitative conclusions need revision, though the qualitative need for a quad-band catalog would remain.","tokens_in":22952,"tokens_out":3421,"duration_ms":40534,"concrete_test":"Recompute Table 1 without the hand-drawn envelope: for the 96 sources with ≥30 sessions, take the 5th percentile (or a bootstrap median) of the smoothed wrms residuals in each 15-degree declination bin, keeping the 7-point filter unchanged, and re-derive the inflated arc lengths and the 15%/6% fractions in Section 5. If using a 5-point or 11-point smoothing window changes the >3σ fraction by more than 3 percentage points, or changes the count of offsets >0.8 mas, the error floor construction is not stable enough to support the quantitative claims.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 4 defines the error floor (Table 1, Fig. 7, grey curve) as the lower envelope of per-source wrms residuals after a 7-point smoothing filter, but gives no algorithm or uncertainty for this envelope. Section 5 then uses this floor, scaled by sqrt(N_sessions) and added in quadrature to formal errors, to compute normalized arc lengths via Eq. (2) and to claim that 15% of sources have >3σ offsets from S/X positions and 6% have offsets >0.8 mas. Because the envelope is the minimum of only ~96 residual values per coordinate, it is sensitive to the arbitrary smoothing window length, to sparse-session sources for which 7 points spans much more than 3 months, and to the few most stable sources. If the true noise floor is higher than the drawn envelope, the inflated uncertainties grow and the significant-offset fractions shrink; if it is lower, the fractions are overestimated. The Rayleigh comparison in Fig. 9 is a consistency check, not an independent validation, since the floor and the arc-length distribution derive from the same residuals. The central claim that a quad-band catalog is necessary is supported by the case studies (2229+695, 3C418, 0723−008), but the quantitative 15%/6% statistics are a headline result and are not robustly anchored.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper analyzes 177 quad-band (3.3/5.5/6.6/10.5 GHz) VLBI sessions from the IVS VGOS program and derives source position time series from dedicated global solutions. The authors identify three types of position variability (stable, continuous motion, back-and-forth motion), attribute the variability to source structure, and introduce a declination-dependent 'error floor' obtained as the lower envelope of smoothed wrms residuals. They inflate quad-band position uncertainties with this floor and compare the resulting positions with ICRF3 S/X positions, reporting that 15% of sources show >3σ offsets and 6% show offsets larger than 0.8 mas. The paper also presents a 377-source quad-band position catalog and argues that a dedicated quad-band catalog is mandatory for processing VGOS observations.","tokens_in":23227,"tokens_out":3560,"duration_ms":39807,"significance":"If the results hold, the paper is a valuable step toward establishing a VGOS celestial reference frame. The analysis is unusually thorough in its handling of session-wise editing, global solutions, and datum definition, and the combination of astrometric time series with closure-only imaging for sources such as 2229+695 and 3C418 is a clear strength. The authors are also careful to distinguish 'visible' and 'invisible' source structure and to connect these concepts to astrometric position offsets. The release of the quad-band catalog and the machine-readable Table 3 is a useful community resource. However, the headline quantitative claims (15% and 6% offsets) rest on an error floor that is specified only as a hand-drawn lower envelope, with no algorithm, uncertainty, or robustness test; this is the main weakness. The qualitative conclusion that source structure affects quad-band positions is well supported by the case studies, but the statistical fractions are not yet anchored to a reproducible error-floor definition.","major_comments":[{"comment":"The declination-dependent error floor is defined as the lower envelope of the smoothed wrms residuals shown as the grey curve in Fig. 7, but no algorithm is given for constructing this envelope and no uncertainty is attached to it. Because Table 1 is used in Section 5 to inflate the uncertainties that enter Eq. (2), the reported 15% and 6% offset fractions depend directly on this envelope. Please specify the envelope construction (e.g., binned quantile, curve fit, or explicit interpolation rule) and quantify its sensitivity to the 7-point smoothing window, the declination bin size, and the subset of most stable sources.","section":"Section 4, Table 1"},{"comment":"The Rayleigh comparison in Fig. 9 is presented as a consistency check, but it is not an independent validation: the error floor and the normalized arc-length distribution are derived from the same quad-band position residuals. An error floor chosen as the lower envelope will, by construction, make the inflated uncertainties smaller for the most stable sources and therefore influence the tail of the X_rho distribution. Please provide an external or hold-out validation, for example by estimating the floor from a subset of sessions and testing on a disjoint subset, or by comparing the inflated uncertainties against independent astrometric results such as those of Petrov (2024).","section":"Section 5, Fig. 9"},{"comment":"The largest offset in the sample, 4.73 +/- 0.29 mas for 0723-008, is derived from a single quad-band session (the text states this is the only VGOS session in which the source was observed). This single-epoch position is then included in the 6% statistic for offsets larger than 0.8 mas. A single-session measurement is more vulnerable to unmodeled session-specific errors than the multi-epoch averaged positions used for most other sources. Please either repeat the offset statistics excluding single-session sources or provide a robustness test showing that the 6% result is unchanged.","section":"Section 5, Table 2, source 0723-008"},{"comment":"The inflation procedure scales the error floor by the square root of the number of sessions, with sessions counted only when the source has more than 500 good observations and with partial credit for sparser sessions. The authors state that the 500 threshold is 'not critical' between 100 and 1000, but no test is shown. Since Eq. (2) depends on sigma_rho, which includes this scaling, a sensitivity analysis over the session-count threshold and over the treatment of fractional sessions is needed to establish that the 15% and 6% percentages are robust rather than artifacts of the weighting choice.","section":"Section 5, N_sessions scaling"}],"minor_comments":[{"comment":"In the paragraph discussing 1803+784, the text refers to '1804+784' when discussing extended structure; this appears to be a typo for 1803+784.","section":"Section 3.1"},{"comment":"The text says the error floor is interpolated from Table 1, but does not specify the interpolation scheme (linear, nearest-bin, or spline) between the 15-degree declination bins; please state the scheme.","section":"Section 4, Table 1"},{"comment":"The grey curve is described only in the caption as the lower edge of the red-dot distribution; please add a legend or textual description of how the grey curve was traced, since it is central to the error-floor definition.","section":"Figure 7"},{"comment":"The terms 'invisible structure' and 'in-beam structure' are used interchangeably in the abstract and in Section 7; define them at first use in both places and keep the terminology consistent throughout.","section":"Abstract and Section 7"},{"comment":"The sentence comparing the present results with 'Figs. 8 and 9 in Petrov (2024)' would be clearer if it stated that the comparison refers to the distribution of position offsets at different frequencies, not to the full contents of those figures.","section":"Section 5"}],"recommendation":"major_revision","confidential_remarks":"The paper's qualitative conclusions and the case-study evidence are solid, and the catalog is a useful product. The main issue is that the quantitative 15%/6% claims rest on an error floor whose construction is not reproducible and whose sensitivity has not been tested. I see no reason for rejection if the authors can supply an algorithmic error-floor definition and robustness checks; if they cannot, the statistical percentages should be presented with appropriate caveats or removed from the abstract."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First pass on Xu & Charlot, arXiv:2501.18276. Bottom line: this is the first dedicated quad-band VGOS position time-series analysis with a catalog, and it deserves a serious look. The strongest parts are the case studies and the care taken in the global-solution design. The weakest part is the error floor that backs the headline 15%/6% offset fractions.\n\nWhat is actually new: 177 IVS quad-band sessions processed with one global solution per source to get session-wise position time series, a 377-source catalog, and a declination-dependent error floor. The visible/invisible source-structure split follows Porcas (2010) and is credited properly. The case studies — 1803+784 stable at tens of µas, 2229+695 with a jet component dragging the measured position 1.7 mas, 3C418 wandering back and forth, 0723−008 where the quad-band position coincides with the bright jet component rather than the core — are coherent and genuinely informative. They alone support the qualitative claim that a quad-band-specific catalog is needed.\n\nThe soft spot is real but contained. The error floor (Table 1, grey curve in Fig. 7) is a hand-drawn lower envelope of per-source wrms residuals after a 7-point smoothing filter, with no algorithm or uncertainty given, and it depends on the smoothing window and on which sources happen to sit at the bottom. The 15% and 6% fractions scale directly with that floor, so those numbers are less robust than the qualitative conclusions. The Rayleigh comparison in Fig. 9 is a consistency check, not an independent validation, since floor and arc lengths come from the same residuals. That said, the floor is not fitted to the arc-length distribution, so the circularity is only partial. The paper is also transparent about its own limits: it admits the inflated uncertainties may still be underestimated and that frequency is not the only driver of the offsets.\n\nWho it is for: anyone processing VGOS geodetic or astrometric sessions, and anyone building the future VGOS celestial frame. The catalog is a deliverable in itself. It deserves a serious referee; the referee should push on the floor methodology and ask for a sensitivity test, but the paper is solid enough to warrant that time.\n\nRecommendation: send to peer review, expect moderate revision on the error floor. I would cite it if I worked in this area.","headline":"First dedicated quad-band VGOS position catalog and time-series analysis with convincing source-structure case studies; the headline 15%/6% offset statistics rest on a hand-drawn declination-dependent error floor that needs to be made reproducible before those numbers are quoted.","tokens_in":23758,"tokens_out":4007,"would_cite":true,"duration_ms":32719,"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":"The apparent position of a quasar in quad-band VLBI is set by unresolved jet structure inside the beam, which moves 15% of sources significantly away from their legacy S/X positions and makes a dedicated quad-band catalog mandatory.","keywords":["celestial reference frame","VLBI","VGOS","AGN source structure","quad-band astrometry","ICRF3","source position stability","radio-optical frame tie"],"falsifier":"Recompute the quad-band catalog with the 177 sessions split into two disjoint halves (or with the 13 stations split into two sub-networks) and apply the same Table 1 error floor to each half. If the same sources appear with >3 sigma offsets to S/X in both halves, the offsets are stable, structure-driven signals; if the set of 'significant' sources changes between halves, the error floor understates the noise and the quoted fractions are artifacts. A second, time-domain check: monitor 0723-008 and 2229+695, since the framework predicts that when the currently brightest jet component fades below the core, the measured quad-band position will jump back toward the core position.","tokens_in":22717,"feed_emoji":"📡","tokens_out":15525,"duration_ms":125584,"temperature":0.7,"pith_summary":"Quasars observed by very long baseline interferometry (VLBI) anchor the celestial reference frame used for astronomy and geodesy, but this study shows that a quasar's measured position is not a stable point: it is the brightness centroid of the radio emission inside the interferometer beam, and that centroid moves as jet components brighten and fade. Analyzing 177 sessions of the new quad-band VGOS system (simultaneous 3.3, 5.5, 6.6, and 10.5 GHz) from 2017 to 2024, the authors find that for sources above $20^\\circ$ declination this unresolved 'invisible' structure is the dominant cause of position instability, while for southern sources the sparse network's poor sensitivity to declination dominates. Comparing their 377-source quad-band catalog with the legacy S/X (2.2/8.6 GHz) positions of the ICRF3, they find 15% of sources offset by more than $3\\sigma$ and 6% offset by more than 0.8 milli-arcseconds; the extreme case, 0723$-$008, is offset by 4.73 mas, with the quad-band position tracking a bright jet knot rather than the core. The operational conclusion is that quad-band observations must be processed against a quad-band catalog, because treating the offsets as noise to be absorbed by inflated uncertainties would hide real physics.","feed_headline":"15% of quasar positions are shifted by unresolved jet structure","feed_subtitle":"In-beam AGN jet knots, not noise, move VGOS positions up to 4.7 mas and demand a dedicated quad-band catalog.","key_machinery":"The argument is carried by per-source position time series obtained from 190 dedicated global solutions of 177 quad-band sessions, in which a stable datum of 46 ICRF3 sources holds the frame fixed while each source's coordinates are estimated session by session. On top of these time series, the concept of source structure from Porcas (2010) does the explanatory work: emission inside the beam ('invisible' structure) displaces the measured position to its brightness centroid, while emission outside the beam ('visible' structure) barely moves the position but creates closure delays. A 7-point first-order polynomial smoothing filter separates the systematic structure-driven wander from white noise, and the lower envelope of the smoothed residuals defines a declination-dependent error floor — about 40 $\\mu$as in the north, growing to 0.32 mas in RA$^*$ and 0.47 mas in Dec below $-45^\\circ$ — which is used to inflate the catalog uncertainties. The significance statistic is the normalized arc length $X_\\rho = \\rho/\\sigma_\\rho$ between quad-band and S/X positions, with $\\sigma_\\rho$ obtained by projecting the two-coordinate uncertainties onto the offset direction; the authors test the inflation by checking that the $X_\\rho$ distribution approaches the Rayleigh distribution expected for pure noise.","core_discovery":"On the paper's own terms, the discovery is that the apparent position of an AGN in quad-band VLBI is governed by the structure of the source on scales smaller than the beam — the 'invisible' structure — whose brightness centroid shifts as jet components are ejected, travel outward, and fade, dragging the measured position along by $0.2$–$1.0$ mas even though the structure never appears in images. The 'visible' structure on larger scales, by contrast, changes positions by only tens of micro-arcseconds but produces closure delays that inflate the delay residuals of geodetic solutions. This single mechanism organizes the three observed behaviors — stable positions near the $0.02$–$0.04$ mas noise level (1803+784), continuous milli-arcsecond drift (2229+695, moving 1.7 mas eastward as its brightest component is a jet knot, not the core), and back-and-forth wander along the jet as components cross the beam boundary (3C418) — and it explains why quad-band, S/X, and Gaia optical positions can point at different physical features of the same object. The authors show this explicitly for 0723$-$008, where the quad-band position lies on a jet component 4.7 mas from the optical core, and conclude that a dedicated quad-band catalog is mandatory and that inflating uncertainties to force agreement with S/X would be the wrong response.","pith_inferences":["Editorial extension: because the 'invisible' scale is set by the beam ($0.4$–$2.0$ mas across the four VGOS bands), the framework predicts a frequency dependence — the fraction and size of structure-induced offsets should shrink at higher observing frequencies; comparing K/Ka-band or single-band VGOS positions against the same S/X catalog would test this scaling.","Editorial extension: phase-referencing astrometry that uses AGN calibrators inherits the calibrator's in-beam structure offset, so the optical–radio frame tie via VGOS-era phase referencing will remain limited by this effect unless calibrators are screened for position stability and jet geometry.","Editorial extension: the back-and-forth model makes a concrete prediction — when a bright jet component crosses from inside to outside the beam, the position shifts back toward the core; a re-analysis of the existing time series could detect these crossing events directly and measure their amplitude against the jet's position angle.","Editorial extension: the paper attributes the three-month-timescale systematic wander to source structure by ruling out tropospheric effects, but a cleaner test would compare the wander direction with the jet position angle measured from the closure-only images for the full sample; a correlation between wander direction and jet angle would confirm the mechanism across all 96 sources rather than th"],"forward_implications":["A quad-band source catalog must be used to analyze quad-band VLBI observations; continuing to use ICRF3 S/X positions as a priori injects systematic errors for the roughly 15% of sources whose positions differ between the bands.","For northern ($\\delta > 20^\\circ$) core-dominated sources, the accuracy of VLBI positions is limited by in-beam source structure at the $0.1$–$0.2$ mas level, so no amount of additional observing time reduces this floor.","Radio–optical position offsets measured against Gaia are substantially caused by in-beam structure pulling the radio position along the jet, so the offsets cannot be removed by improving random errors alone.","For jet-dominated sources such as 2229+695, 0642+449, and 0723$-$008, a catalog position averaged over years does not represent the source's position at any single epoch, since positions can drift by more than a milli-arcsecond.","Southern sources cannot yet reach the quad-band noise floor because the VGOS network lacks long north–south baselines; the declination-dependent error floor quantifies how much the network must be extended."],"supporting_citations":[{"why":"Supplies the ICRF3 S/X positions used as a priori and as the comparison catalog for the offset statistics.","marker":"Charlot et al. 2020"},{"why":"Defines the visible/invisible structure split and the brightness-centroid response that organizes the whole interpretation.","marker":"Porcas 2010"},{"why":"Provides the method for projecting RA/Dec uncertainties onto the offset direction when computing the normalized arc length X_rho.","marker":"Petrov et al. 2019"},{"why":"Simulation study establishing that visible structure shifts source positions by less than roughly 0.1 mas, used to attribute larger wander to in-beam structure.","marker":"Plank et al. 2016"},{"why":"Structure-correction analysis confirming the small position impact of visible structure in real geodetic data.","marker":"Petrov & Kovalev 2017b"},{"why":"Prior result relating inter-band position offsets to quad-band position estimates, used to argue in-beam structure shifts the group-delay position.","marker":"Xu et al. 2022"},{"why":"MOJAVE images and jet kinematics underpinning the 0723-008 core/jet identification and the back-and-forth motion interpretation.","marker":"Lister et al. 2018"},{"why":"The earlier quad-band astrometry study whose session-wise strategy found no such large offsets; the contrast motivates the present analysis.","marker":"Petrov 2024"}],"fun_headline_variants":["AGN jet knots shift quad-band positions by up to 4.7 mas","Unresolved jet structure drives quasar position offsets in VLBI","Quad-band VLBI reveals jet-induced position shifts in AGNs","Invisible jet structure moves quasar positions, demanding new catalog","Jet components, not noise, explain quad-band quasar position drift"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the declination-dependent error floor of Table 1 — read by eye as the lower envelope of the smoothed residual scatter, with no algorithm or uncertainty stated — truly represents the noise floor of quad-band positions; if that floor is drawn too low, the inflated uncertainties are too small and the 15% and 6% offset statistics lose their significance.","fun_headline_variants_meta":{"raw":{"variants":["AGN jet knots shift quad-band positions by up to 4.7 mas","Unresolved jet structure drives quasar position offsets in VLBI","Quad-band VLBI reveals jet-induced position shifts in AGNs","Invisible jet structure moves quasar positions, demanding new catalog","Jet components, not noise, explain quad-band quasar position drift"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000257,"raw_usage":{"total_tokens":1679,"prompt_tokens":1146,"completion_tokens":533,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":762,"completion_tokens_details":{"reasoning_tokens":441}},"tokens_in":762,"tokens_out":533,"duration_ms":4945,"temperature":1.0,"reasoning_tokens":441,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T00:06:45.304844+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the quad-band catalog with the 177 sessions split into two disjoint halves (or with the 13 stations split into two sub-networks) and apply the same Table 1 error floor to each half. If the same sources appear with >3 sigma offsets to S/X in both halves, the offsets are stable, structure-driven signals; if the set of 'significant' sources changes between halves, the error floor understates the noise and the quoted fractions are artifacts. A second, time-domain check: monitor 0723-008 and 2229+695, since the framework predicts that when the currently brightest jet component fades below the core, the measured quad-band position will jump back toward the core position.","supporting_citations":[{"cited_title":"2010, in Sixth International VLBI Service for Geodesy and Astronomy","cited_arxiv_id":null,"evidence_quote":"Defines the visible/invisible structure split and the brightness-centroid response that organizes the whole interpretation."},{"cited_title":"S., McCallum, J","cited_arxiv_id":null,"evidence_quote":"Simulation study establishing that visible structure shifts source positions by less than roughly 0.1 mas, used to attribute larger wander to in-beam structure."}],"review_version":1}