On the Exoplanet Yield of Gaia Astrometry

TL;DR

This work readdresses Gaia's exoplanet yield using updated giant-planet occurrence statistics, a refined local stellar population model, and Gaia’s astrometric performance. It blends a compact semi-analytic framework with more realistic time-series simulations to quantify detectable planets for DR4 and DR5, including how false positives from unresolved binaries may bias the catalog. The authors predict about $7{,}500 \pm 2{,}100$ detections in DR4 and $120{,}000 \pm 22{,}000$ in DR5, with roughly $1{,}900 \pm 540$ and $38{,}000 \pm 7{,}300$ planets having masses and orbital parameters measured to within 20%, respectively; most detections are super-Jupiters on multi-AU orbits around GKM-type stars within a few hundred parsecs. The study also provides mock Gaia exoplanet catalogs and demonstrates that genuine planets will outnumber impostors by substantial factors, though the false-positive rate can be significant for M-dwarfs at small separations. Overall, the results offer concrete expectations and a resource suite to aid community preparation, follow-up planning, and future comparisons with Gaia data.

Abstract

We re-examine the expected yield of Gaia astrometric planet detections using updated models for giant-planet occurrence, the local stellar population, and Gaia's demonstrated astrometric precision. Our analysis combines a semi-analytic model that clarifies key scaling relations with more realistic Monte Carlo simulations. We predict $7{,}500 \pm 2{,}100$ planet discoveries in the 5-year dataset (DR4) and $120{,}000 \pm 22{,}000$ over the full 10-year mission (DR5), with the dominant error arising from uncertainties in giant-planet occurrence. We evaluate the sensitivity of these forecasts to the detection threshold and the desired precision for measurements of planet masses and orbital parameters. Roughly $1{,}900 \pm 540$ planets in DR4 and $38{,}000 \pm 7{,}300$ planets in DR5 should have masses and orbital periods determined to better than $20$%. Most detections will be super-Jupiters ($3$ - $13 M_{\rm J}$) on $2$ - $5$AU orbits around GKM-type stars ($0.4$ - $1.3 M_\odot$) within $500$ pc. Unresolved binary stars will lead to spurious planet detections, but we estimate that genuine planets will outnumber them by a factor of $5$ or more. An exception is planets around M-dwarfs with $a < 1$AU, for which the false-positive rate is expected to be about $50$%. To support community preparation for upcoming data releases, we provide mock catalogs of Gaia exoplanets and planet-impostor binaries.

Figures (15)

• Figure 1: Masses and semi-major axes of known planets and simulated Gaia planets. In both cases, we have restricted the sample to planets with $P_\mathrm{orb}$ and $m_p$ constrained to within $20$% (i.e., $P_\mathrm{84th}/P_\mathrm{16th}\,{<}\,1.2$ and $m_\mathrm{84th}/m_\mathrm{16th}\,{<}\,1.2$). Blue and green points are real planets discovered with the Doppler and transit methods. Red and black points are from our mock Gaia planet catalogs for DR4 and DR5 (see Section \ref{['sec:mock_catalog']}). For comparison, the values of Jupiter and Saturn are highlighted with gray points labeled "J" and "S." The Gaia planets are primarily super-Jupiters on several-AU orbits. About $100$ such planets are currently known; Gaia promises to expand this sample to thousands in DR4 and tens of thousands in DR5.
• Figure 2: The top panel shows the absolute $G$-band magnitude versus stellar mass for main-sequence stars, from PecautMamajek2013 (and online updates). The best-fit fourth-order polynomial is plotted in red. The blue curve is the best-fit single-parameter function $L_G \propto M_\star^4$ or, equivalently, $M_G\,{=}\,-10\log_{10}(M_\star/M_\odot)\, {+}\,C$. The bottom panel shows the derivative $d M_G/dM_\star$, in units of $M_\odot^{-1}$, which is needed to convert the luminosity function into a mass function.
• Figure 3: The top panel shows the $G$-band volumetric luminosity function for main-sequence stars within $100$ pc GaiaCollaboration2021. A cubic spline interpolation is shown in red. The bottom panel shows the corresponding volumetric mass function, based on the two choices for the mass-magnitude relations in Figure \ref{['fig:mass-mag']}. For reference, a simple exponential approximation is plotted in gray. In both panels, the units of volume and mass are pc$^3$ and $M_\odot$, respectively.
• Figure 4: Model for the astrometric precision of individual Gaia measurements (i.e., field-of-view crossings) as a function of apparent $G$-band magnitude. DR3 astrometric uncertainties from Holl2023a are shown in black, along with the interpolation used by the Gaiamock code (ElBadry2024; see Section \ref{['sec:astrom_data']}). Our simple piecewise model (Equation \ref{['eqn:astrom_unc']}; red) was fitted to the Holl2023a data. For bright stars ($G\,{<}\,14$), the astrometric uncertainty is approximately constant, whereas for fainter stars ($G\,{>}\,14$), the astrometric uncertainty rises with $G$ in the manner expected due to photon-counting noise.
• Figure 5: Predictions for the maximum distance (left column) and maximum apparent $G$-band magnitude (right column) of stars for which Gaia can detect planets in DR5. Within each panel, curves are shown for three different planet masses. Each of the three panels is for a different orbital distance. The kinks are due to the break in our model for Gaia's astrometric precision as a function of apparent magnitude (Figure \ref{['fig:astrom_unc']}). In the middle panel of each column, the gray dashed curve shows the scaling expected from a simple analytic model. In the top panels, the arrows indicate how the vertical spacing between curves scales with planet mass. In the bottom panels, the gray bands highlight the cases for which the orbital period exceeds $9.5$ years, which will make astrometric detection challenging. The most distant stars for which Gaia can detect planets are several hundred parsecs away. The faintest $G$ magnitude depends sensitively on stellar mass.