{"id":"817559a4-dea9-4bbc-8d75-a2f595aa9064","arxiv_id":"2608.06453","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Spectroscopic follow-up confirms about 60% of Gaia DR3 astrometric compact-object candidates, identifies 27 neutron-star candidates, and measures the DR3 parallax zeropoint for orbital solutions as -0.0362 +/- 0.0053 mas.","lead":"A four-year spectroscopic campaign with 1,292 high-precision radial velocities checked the reality of 227 candidate black hole, neutron star, and white dwarf binary systems found in Gaia data. About 60% of the astrometric candidates held up, and the data yielded a new measurement of Gaia's distance-scale error for binary orbits.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The parallax-zeropoint fit fixes fG=0 for all 40 sources rather than fitting flux ratio and zeropoint jointly, so faint luminous companions below the detection threshold could bias Z by several times the quoted 0.0053 mas uncertainty.","rationale":"I read the paper in good faith. The observational program is impressive and the public data are a real contribution. The joint astrometry+RV modeling is careful and the consistency checks with Lindegren are strong. The central claim, however, is a precision measurement of Z, and the quoted uncertainty is 0.0053 mas. The weakest link is not the existence of a nonzero zeropoint, which is independently supported, but the assumption that the 40-source sample is exactly dark. The paper's own flux-ratio analysis shows that fG is only constrained to about ±0.02, and the fG and Z fits are not done jointly. Because Equation 24 couples fG to the same physical ratio that defines Z, undetected low-level companion light can masquerade as a parallax shift. The proposed rerun with fG free and a hierarchical Z population is a small additional computation using existing code and would settle whether the central number and its error bar are robust. The reader's weakest_assumption points to the same area, so I partially agree. I do not think the paper should be rejected; the main astrophysical conclusions (purity, NS/WD population, applicability of the single-star zeropoint) are well supported. But the headline precision should be conditional on this check.","tokens_in":57306,"tokens_out":12427,"duration_ms":120201,"concrete_test":"Rerun the Section 5.1.3 joint zeropoint fit with fG as a free parameter for each of the 40 sources, imposing a prior fG≥0, and fit a hierarchical model allowing per-source zeropoints Z_i to scatter around a common mean. Report the resulting common Z and its uncertainty, and compare the scatter of the per-source Z_i to their formal errors. If Z shifts by more than 0.01 mas or the hierarchical uncertainty exceeds about 0.01 mas, the headline precision is not robust; if Z stays within 0.005 mas and the fG posteriors all peak below 0.01, the dark-companion assumption is confirmed.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central value Z=-0.0362±0.0053 mas in Section 5.1.3 is obtained from joint astrometry+RV fits that fix the G-band flux ratio fG=0. The flux-ratio test in Section 5.1.2 is run separately, with raw and Lindegren-corrected parallaxes, and is not coupled to the shared-Z fit. This matters because fG and Z are partially degenerate in Equation 24: a positive fG reduces the photocenter semimajor axis exactly as a more distant (or less negative Z) solution would. The reported fG uncertainties are about 0.02 per source, and the distribution after the L21 correction is centered on zero, but a population with mean fG≈0.01 cannot be excluded. For a typical source with ϖ≈1 mas and q≈0.5, fG=0.01 changes the inferred true parallax by roughly 3% (≈0.03 mas), which is about six times the quoted uncertainty. If such a population exists, the measured Z would be biased in the positive direction (less negative), and the true zeropoint for orbital solutions could differ from both the reported value and the single-star value. Additionally, the per-source Z_i scatter in Figure 9 appears larger than expected from the Fisher-matrix uncertainty; a bootstrap or hierarchical analysis would likely yield an uncertainty closer to 0.01 mas than 0.0053 mas. The paper's external consistency with Lindegren et al. (2021) is reassuring, but it does not by itself settle the fG-Z degeneracy because the same L21 correction is used to define the 'dark' sample.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a systematic spectroscopic follow-up of 227 Gaia DR3 compact-object binary candidates selected from astrometric and single-lined spectroscopic orbits. The authors present 1292 new RVs from TRES and FEROS, joint astrometry+RV fits with per-source jitter and instrumental offsets, SED-based stellar parameter and mass estimates, and a classification of the candidate samples into reliable binaries, spurious solutions, hierarchical triples, and compact-object candidates. A central result is the inference of a Gaia DR3 parallax zeropoint for astrometric orbital solutions, Z = -0.0362 ± 0.0053 mas under the convention varpi_true = varpi - Z, obtained from 40 systems with apparently dark companions and consistent with the single-star Lindegren et al. (2021) correction. The paper also reports that about 60% of the astrometric candidates have reliable orbits and that roughly 27 systems are neutron-star candidates, while cautioning that tight WD+WD binaries can masquerade as neutron stars.","tokens_in":57652,"tokens_out":11595,"duration_ms":99078,"significance":"If the zeropoint result holds, it is important because it validates applying the single-star DR3 parallax zeropoint to 12- and 15-parameter astrometric binary solutions, with direct consequences for Gaia DR4 sample selection and for the inferred physical parameters of compact-object binaries. The paper's strengths include a large homogeneous RV dataset at typical 50 m/s precision, explicit joint modeling of astrometry and RVs with jitter and instrumental offsets, a flux-ratio consistency test, a direct comparison with the external Lindegren et al. (2021) prescription, and public release of spectra, RVs, and machine-readable orbital solutions. The conclusion that the binary zeropoint is consistent with the single-star zeropoint is additionally supported by an ensemble of published astrophysical-standard-candle measurements.","major_comments":[{"comment":"The shared-zeropoint fit fixes the G-band flux ratio to fG=0 for all 40 sources, but Section 5.1.2 reports that one included source, Gaia DR3 220012968211559296, has a best-fit fG=0.055±0.019. Because Eq. (24) shows that a positive fG reduces the photocenter semimajor axis in a way that can be partially absorbed by a less negative Z, this source can bias the inferred Z by roughly (1−factor) times its parallax, which is about 0.3 mas for its ϖ≈3.2 mas and q≈1.4. Even a population with mean fG=0.01 would bias Z by roughly 0.03 mas, several times the quoted 0.0053 mas uncertainty. Please report a joint fit in which fG and a shared Z are fit simultaneously, and/or repeat the zeropoint measurement after excluding sources with fG>0.03, stating how Z changes.","section":"Section 5.1.3, Eq. (24)"},{"comment":"The quoted uncertainty of Z=-0.0362±0.0053 mas is computed from the Fisher matrix, but the per-source best-fit zeropoints in the upper-right panel of Figure 9 show visibly larger scatter, and the bootstrap median, Z=-0.028 (+0.008/-0.012), implies an effective uncertainty of order 0.01-0.02 mas rather than 0.0053 mas. This discrepancy suggests that the formal Fisher uncertainty underestimates the actual spread, possibly because of astrometric systematics or a small number of outliers. A hierarchical or bootstrap analysis of the shared-Z fit should be reported; if the robust uncertainty is substantially larger, the headline precision should be revised even though the consistency with Lindegren et al. (2021) would remain.","section":"Section 5.1.3, Figure 9"},{"comment":"The paper uses the Lindegren et al. (2021) zeropoint correction in the flux-ratio test that supports treating the 40-source sample as dark, and then compares the independently fitted Z with the same Lindegren et al. (2021) prescription. This is not circular, but it means the two consistency checks are not fully independent. The manuscript should state explicitly that the 40-source dark-companion selection does not depend on the L21 zeropoint, or should demonstrate that the zeropoint result is unchanged when the selection is made without any zeropoint correction.","section":"Section 5.1.2-5.1.3"}],"minor_comments":[{"comment":"The abstract says the inferred zeropoint is 'perfectly consistent' with the single-star zeropoint; given the systematic concerns discussed in Section 5.1.3, 'consistent within uncertainties' would be more precise.","section":"Abstract"},{"comment":"The text moves from 41 reliable joint fits to 40 sources in the zeropoint sample without explicitly stating that Gaia BH1 is excluded; please state the exclusion criterion at the start of the subsection.","section":"Section 5.1.3"},{"comment":"The lower-right panel would benefit from a caption clarifying that the red dashed line is the median Lindegren et al. (2021) zeropoint for the same 40 sources and that the abscissa is Z in mas, since several panels share similar axis labels.","section":"Figure 9"},{"comment":"Define q immediately before Eq. (24) and explicitly state that negative fG values are unphysical but allowed in the fits, as this convention is used later in the discussion of Gaia BH1.","section":"Section 5.1.2, Eq. (24)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is well within the journal's scope and the data products are valuable. The main technical issue is the fG-Z degeneracy in the zeropoint inference; it is fixable with a robustness analysis such as a joint fG and Z fit or an outlier-excluded rerun. The independent agreement with Lindegren et al. (2021) suggests the qualitative conclusion is stable, so I recommend major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is a big, careful piece of work. The team obtained 1292 RVs at ~50 m/s precision for 227 systems, characterized more than 90% of them, and released everything. That alone is valuable for anyone using Gaia DR3 astrometric or SB1 orbits. The genuinely new result is the direct measurement of the parallax zeropoint for astrometric orbital solutions: Z = -0.0362 ± 0.0053 mas, consistent with the Lindegren et al. single-star prescription. The purity statistics (~60% for astrometric candidates, ~50% for SB1) and the WD+WD masquerade discussion are also new and will shape DR4 target selection.\n\nThe joint astrometry+RV fitting is done well: they model jitter, instrument offsets, and M1 priors from SEDs, and they explicitly test the dark-companion assumption by fitting flux ratios. The stress-test worry about fG-Z degeneracy is real but not fatal. The shared-Z fit fixes fG=0, and a population with mean fG around 0.01 could bias Z by several times the quoted Fisher uncertainty. But the paper's Section 5.1.2 shows the fG distribution centers on zero after applying the L21 correction, and the independent per-source Z_i fits give a median of -0.028 (+0.008/-0.012), consistent with the joint value. The external agreement with L21 is reassuring. My main quibble is that the 0.0053 mas error likely understates the systematics; the per-source scatter suggests something closer to 0.01 mas. That doesn't change the takeaway, but it would be honest to report a bootstrap-based uncertainty as the headline.\n\nThe only other soft spot is the acknowledged selection bias in the purity fraction: they vetted the most promising candidates more thoroughly, so 60% is an upper limit. They say this clearly.\n\nWho should read it: anyone working with Gaia binary orbits, compact-object search groups, and anyone calibrating parallaxes. It deserves a serious referee; I'd recommend accept with minor revisions. I'd also bring it to the next reading group.\n\nCitation-wise, it leans on the authors' previous catalogs, but that is appropriate here since this is the systematic follow-up of those samples.\n\nShort version: solid paper, real new results, one honest caveat about the error bar. Send it out.","headline":"A systematic RV campaign that vets essentially all Gaia DR3 compact-object candidates and gives the first direct parallax-zeropoint measurement for orbital solutions; the zeropoint is convincing, though the quoted uncertainty may be a bit optimistic.","tokens_in":58278,"tokens_out":1984,"would_cite":true,"duration_ms":20846,"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":"A four-year spectroscopic campaign shows that about 60% of Gaia DR3 high-mass astrometric binary candidates host compact objects, and measures the orbital-solution parallax zeropoint as $Z=-0.0362\\pm0.0053$ mas.","keywords":["Gaia DR3 astrometric binaries","compact object binaries","radial velocity follow-up","parallax zeropoint","black hole binaries","neutron star candidates","white dwarf binaries","spectroscopic binary purity"],"falsifier":"Observe the 40 calibration binaries with high-contrast imaging or wait for Gaia DR4 astrometry to directly measure their G-band flux ratios; if more than a few show flux ratios above roughly 2-5%, the inferred $Z=-0.0362$ mas is biased. Alternatively, measure the same systems' distances from wide, resolved companions with single-star parallaxes and compare them with the joint-fit distances; a systematic difference would falsify the shared zeropoint.","tokens_in":57098,"feed_emoji":"🔭","tokens_out":9436,"duration_ms":73681,"temperature":0.7,"pith_summary":"This paper reports a four-year spectroscopic follow-up of 227 candidate compact-object binaries drawn from Gaia DR3 astrometric and single-lined spectroscopic orbital solutions. Its central claim is that about 60% of the high-mass astrometric candidates genuinely host dark companions -- black holes, neutron stars, or massive white dwarfs -- whereas the spectroscopic candidate sample is dominated by impostors such as post-mass-transfer binaries, hierarchical triples, and spurious orbital solutions. The paper's main quantitative discovery comes from jointly fitting Gaia astrometry with 1,292 new radial velocities for binaries whose companions are dark: each such fit yields a parallax-independent distance. Combining 40 such systems gives a Gaia DR3 parallax zeropoint for astrometric orbital solutions of $Z=-0.0362\\pm0.0053$ mas under the convention $\\varpi_{\\mathrm{true}}=\\varpi-Z$, matching the single-star zeropoint. If the claim holds, binary solutions should be corrected with the same zeropoint as single stars, and the same technique can calibrate distances for the larger DR4 sample.","feed_headline":"Gaia's dark companions fix the parallax zero-point at -0.036 mas","feed_subtitle":"Four years and 1,292 radial velocities show most Gaia compact-object candidates are real and calibrate binary distances.","key_machinery":"The central object is the joint astrometry+RV orbital model for a binary whose companion is assumed dark (Section 5.1.1). Astrometry fixes the angular photocenter orbit through Thiele-Innes elements, while the RV data fix the physical orbit through the velocity semi-amplitude; the likelihood is the product of a multivariate Gaussian on the Gaia parameters and a Gaussian on the radial velocities. Equation 24 extends the model to a luminous companion by replacing the photocenter semimajor axis with a flux-ratio-dependent expression, allowing the same fits to measure the G-band flux ratio and thereby test the dark-companion assumption. For the zeropoint, the paper fits 40 systems simultaneously with a shared additive parallax offset $Z$, exploiting the fact that RV plus inclination gives a parallax-independent distance against which the Gaia parallax can be calibrated. The astrometric mass-ratio function is the selection tool that identifies candidate binaries with massive dark companions before follow-up begins.","core_discovery":"The core discovery is that Gaia DR3 astrometric orbital solutions carry the same parallax zeropoint as single-star solutions, measured directly rather than assumed. Using 40 binaries whose companions contribute negligible G-band light, the paper fits each system's Gaia astrometry jointly with high-precision follow-up radial velocities and obtains a shared zeropoint $Z=-0.0362\\pm0.0053$ mas; this agrees with the median value predicted by the Lindegren et al. (2021) prescription for these sources, and the joint fit prefers it by $\\Delta\\ln P=14.3$. The same fits verify the dark-companion assumption by constraining G-band flux ratios, show that about 60% of astrometric candidates host compact objects (two black holes, 27 neutron-star candidates, and roughly a dozen massive white dwarfs), and reveal that tight WD+WD binaries can masquerade as neutron stars. They also show the high-mass SB1 sample is impure: roughly half of the spectroscopically testable sources have spurious solutions, and most of the rest are post-mass-transfer binaries or hierarchical triples. Applying the zeropoint lowers the inferred companion masses by a median of $0.018\\,M_\\odot$.","pith_inferences":["If the zeropoint is truly shared, then every published DR3 astrometric binary distance should shrink by roughly $0.036$ mas divided by its parallax squared, which matters most for distant systems.","A natural next step the paper does not take is to split the 40 systems by color, magnitude, and sky position to test whether the binary zeropoint varies with these quantities the way the single-star zeropoint does.","The WD+WD masquerade suggests that some confirmed neutron-star candidates could be resolved by ultraviolet spectroscopy or by DR4 astrometric detection of the inner binary's motion.","For future surveys, the low purity of SB1-selected candidates argues for requiring astrometric orbital solutions or combined SED and light-curve vetting before committing multi-year radial-velocity resources."],"forward_implications":["Because the measured zeropoint matches the single-star prescription, astrometric binary solutions from DR3 should be corrected with $\\varpi_{\\mathrm{true}}=\\varpi-Z$ just like single stars, and the same correction should be applied to mass measurements derived from these orbits.","Uncorrected astrometric binary masses are systematically overestimated, by a median of $0.018\\,M_\\odot$ in this sample, and the bias grows for the more distant binaries expected in Gaia DR4.","Reliable astrometric orbits can be separated from spurious ones only statistically: cuts on the Gaia significance and goodness-of-fit remove most impostors, but no single cut is clean.","Some neutron-star candidates are probably tight WD+WD binaries rather than single neutron stars; near-circular orbits or UV excess can flag them.","The 40 dark-companion binaries provide a set of parallax-independent distance anchors that can be revisited with DR4 data to check whether the zeropoint depends on magnitude, color, or ecliptic latitude."],"supporting_citations":[{"why":"Supplies the DR3 non-single-star orbital-solution catalog from which the astrometric and spectroscopic candidate samples are drawn.","marker":"Gaia Collaboration et al. 2023c"},{"why":"Provides the DR3 astrometric binary orbit catalog and the nsstools package used to compute photocenter semimajor axes and the astrometric mass-ratio function.","marker":"Halbwachs et al. 2023"},{"why":"Supplies the DR3 SB1 spectroscopic orbit catalog and mass functions used to select the spectroscopic candidate sample.","marker":"Gosset et al. 2025"},{"why":"Defines the astrometric mass-ratio function that converts photocenter orbits into companion-mass estimates.","marker":"Shahaf et al. 2019"},{"why":"Provides the single-star DR3 parallax zeropoint prescription that the paper tests against and applies to binary solutions.","marker":"Lindegren et al. 2021"},{"why":"Earlier identification of the black-hole candidate sample and Gaia BH1 that anchors the highest-mass astrometric candidates.","marker":"El-Badry et al. 2023a"},{"why":"Establishes the companion-mass threshold used to separate neutron-star candidates from white dwarfs and supplies the prior 21-source neutron-star sample.","marker":"El-Badry et al. 2024a"},{"why":"Independent hierarchical-triple analysis that reached the same conclusion that the single-star zeropoint applies to astrometric binaries.","marker":"Nagarajan & El-Badry 2024"}],"fun_headline_variants":["Gaia parallax zero-point from 40 dark binaries: -0.036 mas","Dark companions fix Gaia's astrometric zero-point at -0.036 mas","Most Gaia compact-object candidates pass follow-up scrutiny","Tight WD+WD binaries masquerade as neutron stars in Gaia","Direct calibration of Gaia orbital parallaxes: -0.036 mas offset"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The distance and zeropoint results assume that the light we see from each binary comes almost entirely from the visible star, so the photocenter tracks that star; if a substantial share of the 40 calibration systems actually contain faint but luminous companions, the inferred distances and zeropoint would be biased.","fun_headline_variants_meta":{"raw":{"variants":["Gaia parallax zero-point from 40 dark binaries: -0.036 mas","Dark companions fix Gaia's astrometric zero-point at -0.036 mas","Most Gaia compact-object candidates pass follow-up scrutiny","Tight WD+WD binaries masquerade as neutron stars in Gaia","Direct calibration of Gaia orbital parallaxes: -0.036 mas offset"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000686,"raw_usage":{"total_tokens":3236,"prompt_tokens":1194,"completion_tokens":2042,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":810,"completion_tokens_details":{"reasoning_tokens":1946}},"tokens_in":810,"tokens_out":2042,"duration_ms":12524,"temperature":1.0,"reasoning_tokens":1946,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T14:33:28.450017+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Observe the 40 calibration binaries with high-contrast imaging or wait for Gaia DR4 astrometry to directly measure their G-band flux ratios; if more than a few show flux ratios above roughly 2-5%, the inferred $Z=-0.0362$ mas is biased. Alternatively, measure the same systems' distances from wide, resolved companions with single-star parallaxes and compare them with the joint-fit distances; a systematic difference would falsify the shared zeropoint.","supporting_citations":[],"review_version":1}