The Gaia-CRF2 paper (Gaia Collaboration, Mignard et al. 2018) closed with a set of systematics-limited open questions for the Gaia optical celestial reference frame: residual quasar proper-motion systematics of order 10 microarcsecond per year blocked use of the quasar proper-motion field to measure the Galactic-acceleration (secular aberration drift) dipole and to estimate the energy flux of the low-frequency stochastic gravitational-wave background; several tens of ICRF3-prototype sources showed unexplained multi-milliarcsecond radio–optical position offsets requiring case-by-case investigation; the distribution of normalized separations between Gaia and VLBI positions had a non-Rayleigh tail; and the frame still depended on external catalogue cross-matches for quasar identification. Later literature resolved the acceleration strand — Gaia EDR3 measured the solar-system acceleration dipole (5.05 +/- 0.35 microarcsecond per year) from 1.2 million quasars. The open problem is threefold: (i) reduce and characterize the remaining large-scale proper-motion systematics (about 12 microarcsecond per year in Gaia DR3 on scales above 15 degrees) to below about 1 microarcsecond per year, confirmed by a stated null test on the residual quasar proper-motion field; (ii) determine, at the individual-source level, the physical origin of the statistically significant Gaia–ICRF3 radio–optical offsets and of the fat-tailed normalized-separation distribution, and their consequences for the mutual alignment of the radio and optical frames; (iii) replace external cross-matching by a self-consistent, high-purity Gaia-internal quasar selection with quantified non-point-source systematics.

Background

Positions on the sky are referred to the International Celestial Reference System (ICRS), a conventionally non-rotating coordinate system whose materialization is the International Celestial Reference Frame (ICRF), built from the positions of distant extragalactic radio sources observed with Very Long Baseline Interferometry (VLBI, an array technique combining radio telescopes on different continents into a single interferometer). The current radio realization, ICRF3 (Charlot et al. 2020), contains 4536 compact sources, mostly quasars — the extremely distant, compact active nuclei of galaxies, whose large distances make their true angular motions negligible so that they define a quasi-inertial frame. ESA's Gaia satellite performs microarcsecond-level global astrometry — positions, parallaxes, and annual proper motions for about 2 billion sources — and since Gaia Data Release 2 (DR2) also provides an independent optical realization of the ICRS, the Gaia celestial reference frame (Gaia-CRF2: 556,869 quasars; Gaia-CRF3 in DR3: about 1.6 million), defined by the positions and proper motions of quasars selected by cross-matching Gaia with external quasar catalogues (Gaia Collaboration, Mignard et al. 2018; Gaia Collaboration, Klioner et al. 2022). Because a reference frame must be non-rotating, the residual quasar proper motions are expanded in vector spherical harmonics (VSH), where the antisymmetric first-order term is a global solid rotation ('spin') and the symmetric term a dipolar 'glide'; all coefficients should vanish up to physical signals and systematics.

Two fundamental-physics signals live in the quasar proper-motion field. First, the acceleration of the solar-system barycentre — dominated by the Galactic gravitational field — produces secular aberration drift: a steadily growing aberrational displacement whose apparent proper motion is a dipole of amplitude about 5 microarcsecond per year directed toward the Galactic centre. Second, a stochastic background of low-frequency gravitational waves (of primordial/cosmological origin plus unresolved compact binaries) induces a characteristic quadrupolar, correlated pattern in the apparent positions and proper motions of distant sources (Gwinn et al. 1997) — the astrometric analogue of the Hellings–Downs angular correlation exploited by pulsar timing arrays. Gaia-CRF2 found that large-scale systematic errors of order 10 microarcsecond per year in the DR2 quasar proper motions — comparable to or larger than both signals — prevented the Galactic-acceleration measurement and a fortiori any estimate of the gravitational-wave background energy flux, concluding that an order-of-magnitude reduction in systematics (expected from better calibrations and longer time baselines in future releases) was required.

A third issue concerns consistency between the radio and optical frames. A VLBI position refers to the compact radio core (centroid of the radio emission), while Gaia measures the optical photocentre (the flux-weighted centre of optical emission); in active galactic nuclei (AGN) these points need not coincide because of parsec-scale jets, extended host-galaxy light, or dual/lensed nuclei. When the Gaia and VLBI positions of a common source are compared, the normalized separation — the angular offset scaled by the combined, covariance-aware positional uncertainties — should follow a standard Rayleigh distribution if all errors are Gaussian and correctly reported; values above 3 are expected for only about 1% of sources, so a pronounced non-Rayleigh tail signals either underestimated formal errors in one or both catalogues, or real physical offsets. Gaia-CRF2 found several tens of sources with multi-milliarcsecond radio–optical offsets against a prototype of ICRF3 whose origin was unknown, a distinctly non-Rayleigh tail (148 sources with normalized separation above 10), and stated that the frame's reliance on external catalogue cross-matches would eventually be superseded by a self-consistent quasar identification from Gaia's own photometry and astrometry. The official ICRF3 was adopted in 2018 and published by Charlot et al. (2020), allowing exactly the revisit that Gaia-CRF2 called for; Gaia EDR3/DR3 (Klioner et al. 2022) provides the current state of that comparison.

Problem Statement

With the Galactic-acceleration (secular aberration drift) strand of the original Gaia-CRF2 question already achieved by Gaia EDR3 (Klioner et al. 2021), the problem — completing the physical validation of the Gaia optical celestial reference frame — has three parts:

(1) Systematics and the gravitational-wave background. Determine and control the residual large-scale quasar proper-motion systematics of the Gaia frame (about 10 microarcsecond per year in DR2 and about 12 microarcsecond per year on angular scales above 15 degrees in EDR3/DR3), reducing them by the order of magnitude that Gaia-CRF2 identified as necessary for estimating the background's energy flux — to below about 1 microarcsecond per year on scales above 15 degrees — and confirm the reduction with a stated null test on the residual quasar proper-motion field.

(2) Radio-optical offsets and the normalized-separation tail. This part comprises three sub-claims, each mandatory and independently assessable. (2a) Determine, at the individual-source level, the physical origin of the statistically significant radio-optical position offsets between Gaia and the official ICRF3 (433 of 3142 common S/X sources at normalized separation squared above 25, with 127 sources above 4 mas and extremes up to tens of mas), discriminating extended and time-variable radio structure, optical photocentre shifts from jets or host galaxies, dual or lensed quasars, and catalogue mismatches. (2b) On that basis, establish whether the fat-tailed distribution of normalized separations is dominated by real physical offsets rather than underestimated formal errors. (2c) Determine what these offsets imply for the attainable accuracy of the mutual alignment of the radio and optical reference frames.

(3) Self-consistent quasar selection. Replace external-catalogue cross-matching by a self-consistent identification of quasars from Gaia's own astrometry, spectrophotometry, and variability, with quantified purity and with the systematics induced by non-point-like sources (host-galaxy morphology, dual/lensed quasars) measured and controlled, as the CRF2 paper anticipated for future releases.

The common objective of the three parts is a Gaia celestial reference frame whose quasar positions and proper motions are physically validated at the microarcsecond level: which sources can be trusted, what their residual systematics are, and how the optical frame relates physically to the radio ICRF3 — the prerequisite for using quasar astrometry for fundamental physics (gravitational-wave background) and for a rigorous radio-optical frame link.

References

  1. Gaia Collaboration: F. Mignard, S. A. Klioner, L. Lindegren, J. Hernández, U. Bastian, A. Bombrun, D. Hobbs, U. Lammers, D. Michalik, M. Ramos-Lerate, M. Biermann, et al. 2018, 'Gaia Data Release 2: The Celestial reference frame (Gaia-CRF2)', Astronomy & Astrophysics 616, A14. DOI 10.1051/0004-6361/201832916, https://doi.org/10.1051/0004-6361/201832916; arXiv:1804.09377, https://arxiv.org/abs/1804.09377
  2. Gaia Collaboration: S. A. Klioner, F. Mignard, L. Lindegren, U. Bastian, P. J. McMillan, J. Hernández, D. Hobbs, M. Ramos-Lerate, M. Biermann, A. Bombrun, A. de Torres, et al. 2021, 'Gaia Early Data Release 3: Acceleration of the solar system from Gaia astrometry', Astronomy & Astrophysics 649, A9. DOI 10.1051/0004-6361/202039734, https://doi.org/10.1051/0004-6361/202039734; arXiv:2012.02036, https://arxiv.org/abs/2012.02036
  3. Gaia Collaboration: S. A. Klioner, L. Lindegren, F. Mignard, J. Hernández, M. Ramos-Lerate, U. Bastian, M. Biermann, A. Bombrun, A. de Torres, E. Gerlach, R. Geyer, et al. 2022, 'Gaia Early Data Release 3: The celestial reference frame (Gaia-CRF3)', Astronomy & Astrophysics 667, A148. DOI 10.1051/0004-6361/202243483, https://doi.org/10.1051/0004-6361/202243483; arXiv:2204.12574, https://arxiv.org/abs/2204.12574
  4. P. Charlot, C. S. Jacobs, D. Gordon, S. Lambert, A. de Witt, J. Böhm, A. L. Fey, R. Heinkelmann, E. Skurikhina, O. Titov, E. F. Arias, S. Bolotin, G. Bourda, C. Ma, Z. Malkin, A. Nothnagel, D. Mayer, D. S. MacMillan, T. Nilsson, R. Gaume 2020, 'The third realization of the International Celestial Reference Frame by very long baseline interferometry', Astronomy & Astrophysics 644, A159. DOI 10.1051/0004-6361/202038368, https://doi.org/10.1051/0004-6361/202038368
  5. C. R. Gwinn, T. M. Eubanks, T. Pyne, M. Birkinshaw, D. N. Matsakis 1997, 'Quasar Proper Motions and Low-Frequency Gravitational Waves', The Astrophysical Journal 485, 87. DOI 10.1086/304424, https://doi.org/10.1086/304424
  6. Y. Y. Kovalev, L. Petrov, A. V. Plavin 2017, 'VLBI-Gaia offsets favor parsec-scale jet direction in active galactic nuclei', Astronomy & Astrophysics 598, L1. DOI 10.1051/0004-6361/201630031, https://doi.org/10.1051/0004-6361/201630031
  7. L. Petrov, Y. Y. Kovalev, A. V. Plavin 2019, 'A quantitative analysis of systematic differences in the positions and proper motions of Gaia DR2 with respect to VLBI', Monthly Notices of the Royal Astronomical Society 482, 3023. DOI 10.1093/mnras/sty2807, https://doi.org/10.1093/mnras/sty2807
  8. A. V. Plavin, Y. Y. Kovalev, L. Y. Petrov 2019, 'Dissecting the AGN Disk-Jet System with Joint VLBI-Gaia Analysis', The Astrophysical Journal 871, 143. DOI 10.3847/1538-4357/aaf650, https://doi.org/10.3847/1538-4357/aaf650
  9. S. Lambert, N. Liu, E. F. Arias, C. Barache, J. Souchay, F. Taris, J. C. Liu, Z. Zhu 2021, 'Parsec-scale alignments of radio-optical offsets with jets in AGNs from multifrequency geodetic VLBI, Gaia EDR3, and the MOJAVE program', Astronomy & Astrophysics 651, A64. DOI 10.1051/0004-6361/202140652, https://doi.org/10.1051/0004-6361/202140652
  10. N. Liu, S. B. Lambert, P. Charlot, Z. Zhu, J.-C. Liu, N. Jiang, X.-S. Wan, C.-Y. Ding 2021, 'Comparison of multifrequency positions of extragalactic sources from ICRF3 and Gaia EDR3', Astronomy & Astrophysics 652, A87. DOI 10.1051/0004-6361/202038179, https://doi.org/10.1051/0004-6361/202038179
  11. J. Darling 2025, 'A New Approach to the Low-frequency Stochastic Gravitational-wave Background: Constraints from Quasars and the Astrometric Hellings-Downs Curve', The Astrophysical Journal Letters 982, L46. DOI 10.3847/2041-8213/adbf0d, https://doi.org/10.3847/2041-8213/adbf0d
  12. S. Frey, O. Titov, A. Melnikov, S. Lambert 2025, 'New look at old friends: EVN imaging of prominent radio-loud active galactic nuclei with extremely large radio-optical positional offsets', Proceedings of the 16th European VLBI Network Symposium (Bonn), pp. 141-144. arXiv:2501.06513, https://arxiv.org/abs/2501.06513
  13. A. G. A. Brown, U. Bastian, S. A. Klioner 2025, 'The Gaia Astrometric Catalogue and Secular Aberration Drift in Proper Motions', Astronomy & Astrophysics 695, A272. DOI 10.1051/0004-6361/202452459, https://doi.org/10.1051/0004-6361/202452459; arXiv:2503.03389, https://arxiv.org/abs/2503.03389
  14. M. Vaglio, M. Falxa, G. Mentasti, A. I. Renzini, A. Kuntz, E. Barausse, C. Contaldi, A. Sesana 2025, 'Searching for Gravitational Waves with Gaia and its Cross-Correlation with PTA: Absolute vs Relative Astrometry', arXiv:2507.18593, https://arxiv.org/abs/2507.18593
  15. S. Liao, Q. Wu, Y. Ding, Q. Xu, Z. Qi 2025, 'Characterizing the astrometric quality of AGNs in Gaia-CRF3', Proceedings of IAU Symposium 401 'Advancing Reference Systems, Ephemeris, and Standards' (La Plata). arXiv:2511.02204, https://arxiv.org/abs/2511.02204
  16. Q. Wu, S. Liao, Z. Qi, Q. Xu, Y. Ding, K. Zhu 2026, 'Optical-morphology-based assessment of astrometric quality in Gaia-CRF3 quasars', arXiv:2606.10655 (submitted to Astronomy & Astrophysics), https://arxiv.org/abs/2606.10655
  17. G. Fichet de Clairfontaine, M. Perucho, J. M. Martí, Y. Y. Kovalev 2026, 'Stellar mass loading drives dissipation and reacceleration in AGN jets: Explaining VLBI-Gaia offsets and constraining jet power', Astronomy & Astrophysics 711, A227. DOI 10.1051/0004-6361/202660436, https://doi.org/10.1051/0004-6361/202660436
  18. Z.-Y. Zhang, N. Liu, X.-X. Zhang, I. Nurul Huda, S.-F. Guo, Z. Zhu, J.-C. Liu, J. Yao, Z.-W. Wang, H.-F. Yu, D.-D. Zhang 2026, 'Astrometric properties of reference frame sources as a function of redshift', Astronomy & Astrophysics 708, A83. DOI 10.1051/0004-6361/202558731, https://doi.org/10.1051/0004-6361/202558731