A Planetary Illusion’s Funeral: Nondetection of a Gaia DR3 Exoplanet Candidate, and the Role of Intermediate-precision Radial Velocities in Gaia Exoplanet Follow-up

Many of the methods used for the discovery of exoplanets involve the detection of the subtle influence they have on their parent stars. Perhaps the earliest of these methods to be applied is astrometry, which may be used to indirectly detect the existence of planets through the detection of the minute reflex orbital motion of their host stars as they move across the sky (A. Sozzetti 2005; A. Quirrenbach 2010).

Though the discovery of exoplanets through astrometry has been explored for a long time, the method has unfortunately enjoyed rather few successes. Ground-based detections have been attempted since the middle of the twentieth century (e.g., S. L. Lippincott 1960; P. van de Kamp 1963, 1969; G. Gatewood & H. Eichhorn 1973; J. L. Hershey 1973; G. Gatewood 1974, but the level of precision required means that for most stars planet detection is only possible with space-based observations (see A. Sozzetti 2005, section 4.1 for a review). Even in the case of the Hipparcos mission (M. A. C. Perryman et al. 1997), only a small number of planets then known could be detected (S. Reffert & A. Quirrenbach 2011).

However, this state of affairs is changing thanks to the recently completed Gaia mission (Gaia Collaboration et al. 2016). Gaia has performed a deep all-sky astrometric survey with unprecedented precision and duration. Early-mission yield estimates set the expectation that Gaia will astrometrically detect in excess of >10⁴ exoplanets (M. Perryman et al. 2014), a considerably larger count than the entire sample of exoplanets currently known. There is therefore significant anticipation for the exoplanet science results from the 5 yr Gaia nominal mission, included in Gaia Data Release 4 (DR4), which is expected in 2026 December.⁴

Gaia DR3, released in 2022 June based on the first three years of Gaia observations (Gaia Collaboration et al. 2023a), included the first batch of non-single-star (NSS) orbital solutions from the mission. The detection pipeline and the properties of the orbital solutions are primarily discussed in Gaia Collaboration et al. (2023b). The Gaia DR3 NSS solutions have already proved fruitful for exoplanet research. J. N. Winn (2022) explored joint constraints on exoplanet parameters by combining the Gaia astrometric orbital solutions with archival RV observations, demonstrating consistency between solutions in the best cases. Follow-up RV observations of the star HIP 66074 led to the first confirmation of an exoplanet reported in the Gaia astrometry, Gaia-3 b (A. Sozzetti et al. 2023). Most recently, Gaia-4 b and Gaia-5 b have been confirmed as massive companions (10–20 M_J) orbiting low-mass stars (G. Stefánsson et al. 2025), while Gaia-6 B is a confirmed long-period and high-eccentricity companion, though with significantly revised parameters (M. Pinamonti et al. 2026).

However, much care must be taken to identify false positives among the Gaia astrometric exoplanet candidates. The main astrophysical false positive scenario for these systems involve binary systems with near-equal luminosities. As Gaia observes the motion of the center-of-light much smaller than the angular resolution of the instrument, for a binary where both components are luminous in the GaiaG-band the orbital motion will be attenuated according to the flux ratio. The first system to be recognized as such a false positive is HD 68683, identified in B. Holl et al. (2023) on the basis of published spectroscopic observations that show it is a double-lined spectroscopic binary (SB 2). Three additional Gaia DR3 candidates were discovered to be SB 2s in M. L. Marcussen & S. H. Albrecht (2023), demonstrating that these false positives are not uncommon but can easily be detected with spectroscopic follow-up. It is also possible to detect these false positives through imaging; HD 3221, highlighted as a planet candidate orbiting a young star in Gaia Collaboration et al. (2023b), was independently resolved as a near-equal luminosity binary by M. Bonavita et al. (2022).

A more pathological false positive scenario involves fully spurious orbital solutions. At the time of writing, a total of four Gaia DR3 NSS orbits have been identified as spurious and have been retracted, being caused by software issues.⁵ These manifested as systems with NSS astrometric solutions that failed to demonstrate the expected signals of either planetary or stellar companions in follow-up observations. The first of these to be reported was HD 113283, with a stellar companion reported in Gaia DR3 that went undetected in follow-up RV observations (D. Spaeth et al. 2023). Another casualty was the candidate companion of WD 0141-675, the only planet candidate orbiting a white dwarf detected in Gaia DR3 (Gaia Collaboration et al. 2023b, their Section 8.8; see further L. K. Rogers et al. 2024). The NSS solution for HIP 66074 (Gaia-3) was also retracted, which appears unexpected given the earlier confirmation of a planet with a similar orbital period in A. Sozzetti et al. (2023) and complicates its claim to being the first exoplanet to be discovered from Gaia astrometry.

The last of the four now-retracted exoplanet candidates pertains to the bright (V = 6.6) solar-type star HD 12800. The NSS solution was highlighted as “the only candidate companion around a main-sequence solar-type star with a mass well in the planetary regime” newly reported from the Gaia DR3 targeted search (Gaia Collaboration et al. 2023b; their Section 8.7), with orbital period P = 401 ± 12 days, orbital semimajor axis a₀ = 0.25 ± 0.05 milli-arcseconds (mas), and estimated mass 5.6 ± 1.4 M_J. However, it was ultimately discovered that the two-body solution was spurious, and the Gaia DR3 NSS solution was retracted on 2024 May 27.

Soon after the release of Gaia DR3, we identified the candidate companion of HD 12800 as an interesting target for follow-up and immediately began collecting RV data. However, our observations failed to detect the expected orbital signal, which demonstrated that the candidate companion did not exist prior to its formal retraction. In this work, we report on our observations of HD 12800 and their implications for follow-up of Gaia astrometric exoplanet candidates in the context of the future release of Gaia DR4.

HD 12800 (54 Cas, Gaia DR3 522135261462534528) is a V = 6.6 star at a distance of 27.018  ±  0.012 parsecs (Gaia Collaboration et al. 2023a) with a spectral type conventionally given as F8, originating from the Henry Draper catalog (see A. J. Cannon & E. C. Pickering 1918). Despite its status as a bright and nearby Sun-like star, HD 12800 has little presence in the astronomical literature and has largely avoided study.

In Gaia DR3 the star had a OrbitalTargetedSearch solution in the nss_two_body_orbit table. Though the two-body solution has now been retracted, we go through the process of extracting the orbital information from the Gaia orbital solution in order to make full comparisons with our RV observations. The Gaia DR3 solution has an orbital period of P = 401 ± 12 days, eccentricity e = 0.22 ± 0.16, time of periastron T_p = 118 ± 69 days (i.e., BJD 2457507  ±  69 relative to epoch 2016.0), and the following Thiele–Innes orbital coefficients: A = 0.157 ± 0.096, B = 0.020 ± 0.079, F = −0.152 ± 0.087, and G = −0.140 ± 0.051. To convert the reported Thiele–Innes coefficients to the corresponding Campbell terms, we used the nsstools package (see J.-L. Halbwachs et al. 2023, appendices A, B).⁶ This provides a semimajor axis of a₀ = 0.25 ± 0.05 mas, argument of periastron ω = 54° ± 32°, orbital inclination i = 1078 ± 83, and longitude of node Ω = 304 ± 57 for the Gaia DR3 two-body solution. These values align with those reproduced in Gaia Collaboration et al. (2023b).

In the case of a two-body orbit where the secondary contributes no flux, as appropriate for a star-planet system, the observable amplitude of the astrometric orbit a₀ is related to the system properties through the following equation:

where a₀ is in milli-arcseconds, M_* and M_p are the mass of the star and companion respectively, a is the orbital semimajor axis in AU, and D is the distance in parsecs (A. Sozzetti 2005, Equation 6). Assuming for the moment M_* = 1 M_⊙, the a₀ = 0.25 ± 0.05 mas originally reported in Gaia DR3 entails a companion mass of M_p ≈ 6.3 M_J.

We may then estimate the expected radial velocity semiamplitude K using the classical expression:

where G is the gravitational constant (C. Lovis & D. Fischer 2010, Equation 12). We provisionally estimated a reflex semiamplitude of K ≈ 170 m s⁻¹, which notionally could be easily detected in RV observations. This motivated our attempt to detect the RV signal from the planet candidate.

Though HD 12800 is a bright Sun-like star, it happens to have been omitted from the major long-term exoplanet RV surveys, perhaps on account of its relatively blue colors (B – V = 0.59, E. Høg et al. 2000). The most recent RV data published in the literature are low-precision observations (σ = 0.33 km s⁻¹) from D. W. Latham et al. (2002), though the star has occasionally been used as a spectral standard in subsequent literature (ex. B. Tingley et al. 2011). We report the relevant observational parameters in Table 1.

Table 1. Properties of HD 12800

Parameter	Value	References
R.A. αJ2000 …	02:09:08.26	Gaia Collaboration et al. (2023a)
decl. δJ2000 …	+71:33:07.22	Gaia Collaboration et al. (2023a)
Parallax ϖ (mas) …	37.012  ±  0.017	Gaia Collaboration et al. (2023a)
Distance d (pc) …	27.018  ±  0.012	Gaia Collaboration et al. (2023a)
V (mag) …	6.58  ±  0.01	E. Høg et al. (2000)
BT (mag) …	7.225  ±  0.015	E. Høg et al. (2000)
VT (mag) …	6.633  ±  0.010	E. Høg et al. (2000)
G (mag) …	6.448  ±  0.02	Gaia Collaboration et al. (2023a)
GBP (mag) …	6.717  ±  0.02	Gaia Collaboration et al. (2023a)
GRP (mag) …	6.009  ±  0.02	Gaia Collaboration et al. (2023a)
J (mag) …	5.519  ±  0.027	M. F. Skrutskie et al. (2006)
H (mag) …	5.313  ±  0.038	M. F. Skrutskie et al. (2006)
KS (mag) …	5.241  ±  0.017	M. F. Skrutskie et al. (2006)
Teff (K) …	6000  ±  50	This work
( cm s−1) …	4.26  ±  0.10	This work
[M/H] (dex) …	−0.24 ± 0.08	This work
vbroad (km s−1) a …	4.0  ±  0.5	This work
M* (M⊙) …		This work
R* (R⊙) …	1.082  ±  0.024	This work
L* (L⊙) …	1.36  ±  0.07	This work
Age (Gyr) …		This work

Note. ^aIncludes both and macroturbulence.

Download table as:  ASCIITypeset image

We therefore began collecting new observations of HD 12800 soon after the 2022 June 13 release of Gaia DR3. We observed the star using the TRES spectrograph on the 1.5 m telescope at the Fred Lawrence Whipple Observatory, Mt. Hopkins, AZ, USA. We acquired a total of 28 TRES observations of HD 12800 with exposure times varying between 45 and 540 s, achieving a median signal-to-noise per resolution element of 160. Our first observation was made on 2022 July 8; further observations were then delayed until October due to seasonal weather at Mt. Hopkins. We then regularly observed the star 17 times between 2022 October 4 and 2023 February 5 at an approximately weekly cadence. However, it became apparent that the expected large RV variability was not present, and the target was temporarily placed on hold. We subsequently acquired 10 further observations between 2023 October 4 and 2024 February 13, which confirmed the constant RV.

To determine the RVs, we perform spectral extraction and multiorder cross correlations following the process outlined in L. A. Buchhave et al. (2010). We account for variations in the instrument zero-point through nightly observations of standard stars as in Q. Zhou et al. (2023). Our RV observations achieve a median precision of 11 m s⁻¹, and have a standard deviation of only 9 m s⁻¹. We present the TRES RVs collected for this work in the Appendix.

In order to fully compare our RV results to the Gaia DR3 NSS astrometric solution, we require an accurate estimate for the stellar mass M_*. We therefore use the Stellar Parameter Classification tool (SPC; L. A. Buchhave et al. 2012) to estimate the observable stellar parameters from the TRES spectra as in A. Bieryla et al. (2024). We estimate T_eff = 6000 ± 50 K, [M/H] = −0.24 ±0.08 dex, (cm s⁻¹), and a line broadening of 4.0 ± 0.5 km s⁻¹. We then model the physical parameters of HD 12800 using the MIST isochrones (J. Choi et al. 2016; A. Dotter 2016) with a model applied in our previous work (A. Venner et al. 2024). We use space-based photometry from Tycho-2, Gaia, and 2MASS, and adopt the spectroscopic priors on effective temperature and metallicity from the SPC results. We estimate a stellar mass of , radius R_* = 1.082 ± 0.024 R_⊙, luminosity L_* = 1.36 ± 0.07 L_⊙, and age  Gyr. These properties suggest that HD 12800 is similar to the Sun, mainly differing in being less metal-enriched and more evolved. We report our stellar parameters in Table 1.

With this updated value for the stellar mass, we may now recalculate the companion mass implied by the Gaia DR3 two-body solution using Equation (1), resulting in M_p = 6.3 ± 1.3 M_J. Then, following Equation (2), the expected reflex RV semiamplitude is K = 170 ± 35 m s⁻¹.

We plot the zero-point normalized TRES radial velocities of HD 12800 in Figure 1. We compare it to the RV signal expected from the retracted Gaia DR3 two-body solution assuming the median orbital parameters. It is evident that our RV observations can unequivocally reject the existence of the candidate companion with high confidence, and could do so even prior to the official retraction of the orbital solution on 2024 May 27.

Zoom In Zoom Out Reset image size

The immediate result of this work is the nondetection of the (now-retracted) astrometric companion candidate of HD 12800 originally reported in Gaia DR3. This is far from the first case of an astrometric exoplanet candidate that has gone undetected in follow-up RV observations; a notable pre-Gaia example is VB 10 (G. Anglada-Escudé et al. 2010; J. L. Bean et al. 2010). However, in the vein of L. K. Rogers et al. (2024), we seek here to leverage our nondetection to extract some lessons in the process of following up Gaia astrometric exoplanet candidates for the forthcoming release of Gaia DR4.

A major lesson drawn from past and present research is that radial velocities provide a powerful complement to astrometry in the area of exoplanet detection. For solar-type FGK dwarfs, astrometry is mainly sensitive to long-period giant planets (P ≳ 1 yr, M_p ≳ 1 M_J) at the limits of current precision, which are normally well within the range of detectability of RVs. Furthermore, since astrometry achieves higher sensitivity for more proximate stars (Equation (1)), it performs best on nearby bright stars which are often the best RV targets.

Perhaps one of the main lessons to be taken from previous studies combining Hipparcos-Gaia astrometry with radial velocities is that most bright solar-type stars have decades of RV survey observations fit to be combined with astrometry (e.g., Y. Li et al. 2021; A. Venner et al. 2021; G.-Y. Xiao et al. 2023; Q. An et al. 2025). This is in significant part a result of the fact that most of the classical RV surveys assembled their target lists from bright solar-type stars in the Hipparcos catalog (e.g., S. Udry et al. 2000; H. R. A. Jones et al. 2002), which is >90% complete for stars brighter than V ≤ 8 (C. Turon et al. 1992). It is therefore unlikely that the Gaia DR4 astrometry will discover many new planets orbiting solar-type stars brighter than V ≤ 8, simply due to the fact that most of these stars have already been observed in RV surveys. Here HD 12800 is an exception, being a Sun-like star omitted from the main northern hemisphere RV surveys. We speculate that this occurred due to its relatively early spectral type (F8). As a result, for the subsection of stars brighter than V ≲ 8, we argue that Gaia may mainly discover new planets around earlier-type stars (≲F8) typically unsuitable for precise RV observations due to their propensity for rapid rotation.

Beyond V  >  8, the likelihood that a given solar-type star has existing RV observations declines due to a combination of the increasing likelihood of absence from Hipparcos and idiosyncratic choices in target brightness limits among different RV surveys. Gaia is therefore more likely to discover new planets around stars fainter than V  >  8, which is neatly demonstrated by Gaia-3 b, Gaia-4 b, Gaia-5 b, and Gaia-6 B (all V  >  8; e.g., A. Sozzetti et al. 2023; M. Pinamonti et al. 2026; G. Stefánsson et al. 2025), and their confirmation will entail new RV observations. Follow-up of the aforementioned targets have been undertaken by high-precision instruments (i.e., ≈1 m s⁻¹ precision; HIRES, HARPS-N, NEID, HPF; A. Sozzetti et al. 2023; G. Stefánsson et al. 2025), which have proven adept for the confirmation of these Gaia DR3 planet candidates. However, access to observing time on high-precision spectrographs is restricted or highly competitive which may limit the accessibility of follow-up observations, especially considering the quantity of planet candidates expected in Gaia DR4 (M. Perryman et al. 2014).

We reason that there is a good analogy to be made between follow-up of Gaia DR4 planet candidates and the follow-up of exoplanets discovered by the TESS mission (G. R. Ricker et al. 2014). TESS is performing an all-sky survey for transiting planets, predominantly those orbiting Sun-like stars between magnitudes 7 ≲ T ≲ 13 (where T reflects the red-optical TESS bandpass), which has detected thousands of planet candidates to date (N. M. Guerrero et al. 2021). The number, stellar properties, and magnitude distribution of TESS targets compare favorably to the properties of planet hosts expected to be discovered by Gaia (M. Perryman et al. 2014). In the context of TESS science, intermediate-precision spectrographs (≈10 m s⁻¹ precision) have played a vital role in the follow-up and confirmation of TESS planet candidates through both outright confirmation of reflex RV signals as well as by excluding eclipsing binary false positive scenarios. In particular, we may highlight TRES, CHIRON (A. Tokovinin et al. 2013), MINERVA-Australis (B. Addison et al. 2019), and NRES (R. J. Siverd et al. 2018), along with the older FEROS (A. Kaufer et al. 1999) and CORALIE (D. Queloz et al. 2000), as “workhorse” spectrographs which have cumulatively contributed to the confirmation of hundreds of TESS exoplanets. These spectrographs typically achieve 10–100 m s⁻¹ RV precision for the relevant stars, down to V ≈ 12.

We argue that RV observations with comparable precision and breadth will play a key role in the follow-up of Gaia DR4 astrometric planet candidates. The hosts of Gaia planets are expected to mainly lie between 6 ≲ G ≲ 16, with a peak at G ≈ 12 (where G reflects the Gaia bandpass; M. Perryman et al. 2014, figure 1a). We infer from M. Perryman et al. (2014) that approximately ≈50% of planets detected from the Gaia astrometry will orbit stars between 8 < G < 12, in the range amenable to observations by intermediate-precision spectrographs. While the conversion between the Gaia-observed astrometric semimajor axis a₀ and the RV semiamplitude K depends on several parameters, (Equations (1), (2)), for a representative solar-mass star at 50 pc we suggest that typical values for K may lie within an order of magnitude of ≈100 m s⁻¹ for the planets expected to be detected by Gaia. If such large values for K are broadly accurate, a substantial fraction of Gaia astrometric exoplanet discoveries could be directly confirmed using intermediate-precision spectrographs; for the remaining systems where K is too small for detection, intermediate-precision spectrographs may still provide important information by providing vetting for false positives, especially binaries undetected in the Gaia data.

Beyond the similarities in host astrophysical properties between TESS and Gaia planet candidates, we further argue that TESS follow-up provides a reasonable model for the logistics of exoplanet follow-up for Gaia DR4. The TESS follow-up program is primarily collaborative, which has helped greatly to maximize the scientific output of TESS. Given the even larger scale of exoplanet science expected from Gaia, it can be argued that a similarly collaborative approach to Gaia astrometric exoplanet follow-up would serve to enhance Gaia DR4 exoplanet science.

We acknowledge and pay respect to Australia’s Aboriginal and Torres Strait Islander peoples, who are the traditional custodians of the lands, waterways and skies all across Australia. AVe would like to thank Jason Wang and Sarah Blunt for productive discussions on astrometric exoplanet detection that have helped to shape this work. AVe and CXH are supported by ARC DECRA project DE200101840.

This work has made use of data from the European Space Agency (ESA) mission Gaia (https://www.cosmos.esa.int/gaia), processed by the Gaia Data Processing and Analysis Consortium (DPAC; https://www.cosmos.esa.int/web/gaia/dpac/consortium). Funding for the DPAC has been provided by national institutions, in particular the institutions participating in the Gaia Multilateral Agreement.

Facilities: Gaia - , FLWO:1.5m (TRES) - .

Software: nsstools (J.-L. Halbwachs et al. 2023).

In Table 2 we present the multiorder, zero-point corrected TRES RVs collected for this work.