Astrometry Method

The wobble in two dimensions

The standard intuition for exoplanet detection by stellar wobble is the radial-velocity method: the star moves toward and away from the observer along the line of sight, and Doppler shifts reveal the motion. But the same dynamics imply a complementary signal — the star also moves perpendicular to the line of sight, tracing an ellipse on the sky. Measuring that 2D motion is the astrometry method.

If a planet has mass m_p, orbits at semi-major axis a, and the system is at distance d from Earth, the star moves in its own orbit around the common center of mass. That orbit has semi-major axis a_star = a × m_p / m_star. The angular size of the star's orbit on the sky is

α = a_star / d
  = (m_p / m_star) × (a / d)

For a Jupiter (m_p / m_Sun = 9.5 × 10⁻⁴) at 5 AU (a = 5) around a Sun-mass star at 10 parsecs (d = 10 pc = 2 × 10⁶ AU), α = 9.5 × 10⁻⁴ × 5 / 2 × 10⁶ = 2.4 × 10⁻⁹ radians = 500 microarcseconds.

Five hundred microarcseconds is the angular size of a one-euro coin held 8,000 kilometres away. It is small. It is also vastly larger than what current astrometric missions can resolve. Gaia's per-epoch precision is ~7 μas at G=15 — about 70× smaller than the Jupiter-at-5-AU-at-10-pc signal. Hundreds of Sun-like stars within ~50 pc are in range for Jupiter detection by Gaia.

Three advantages of astrometry

Astrometry's signal scales differently from radial velocity. The RV semi-amplitude K and the astrometric semi-amplitude α scale as:

K_RV   = (2π G / P)^(1/3) × (m_p sin i) / m_star^(2/3)   ∝  a^(-1/2)
α_astr = (m_p / m_star) × (a / d)                         ∝  a

where P is orbital period. Three consequences follow:

1. Astrometry favours wide orbits. α grows linearly with a; K_RV shrinks as 1/√a. RV is sensitive to hot Jupiters (close, fast) and Earth-like in habitable zone (close, hard); astrometry is sensitive to Jupiter analogs (5 AU, 12-year orbit) and gas giants further out.
2. Astrometry returns true mass. The shape of the ellipse on the sky fixes the orbital inclination directly — you can measure how tilted the orbit is by the projection. RV gives only m sin i because line-of-sight velocity has no inclination information. Combined RV + astrometry breaks the sin i degeneracy.
3. Astrometry is immune to stellar activity in a different way. Stellar oscillations and spots produce spurious RV signals at the m/s level but produce negligible astrometric shifts because the photocenter of the star is set by integrated brightness, not by a few active spots. (There are subtler astrometric jitter sources, but the dominant RV systematics do not apply.)

These advantages have made astrometry the long-promised future of exoplanet detection for decades. The wait has been for the technology to reach the precision required.

From Bessel to Gaia

Year	Mission / observer	Precision	Notes
1844	F. W. Bessel	~100 mas	Detected Sirius B by wobble of Sirius A; first invisible companion
1963	P. van de Kamp	~10 mas	Claimed (incorrectly) Barnard's Star planets
1989–93	Hipparcos (ESA)	~1 mas	First space-based astrometry; brown dwarf detections
2002–05	HST FGS	~0.5 mas	GJ 876b mass measurement (first astrometric exoplanet mass)
2013–present	Gaia (ESA)	~7 μas at G=15	Billion-star catalog; mission yield ~70,000 expected
2024–25	VLTI GRAVITY+	~10 μas (interferometric)	Direct astrometry of nearby imaged planets
2030+ (planned)	Theia / GaiaNIR concepts	~0.3 μas	Targeting Earth-class planets nearby

The 19th-century discovery that gave astrometry its credentials: Friedrich Bessel in 1844 noticed Sirius wobbled with an 50-year period and concluded an invisible massive companion existed. The companion, Sirius B, was finally seen visually in 1862 — and turned out to be a white dwarf. Astrometry had detected an invisible mass before its nature was known. The technique is older than spectroscopy, older than photography of stars.

Modern space astrometry began with Hipparcos (ESA, 1989–93), which achieved ~1 mas single-measurement precision and produced the first global catalog of precise stellar positions and parallaxes. Hipparcos did not have the precision to find planets but it cleaned up calibration of the parsec scale and discovered brown-dwarf companions. Gaia (launched 2013) is the successor, with three orders of magnitude better precision and a billion-star catalog. As of Data Release 3 (2022) Gaia has begun publishing astrometric exoplanet candidates; the full mission catalog (DR5, ~2030) will deliver the bulk of the exoplanet harvest.

Worked example: detecting Jupiter from 10 parsecs

How precisely must we measure Sun-like stars to detect Jupiter-mass planets at 5 AU?

Sun's mass = 2 × 10³⁰ kg. Jupiter's mass = 1.9 × 10²⁷ kg. Mass ratio m_p / m_star = 9.5 × 10⁻⁴. Jupiter's orbital semi-major axis a = 5.2 AU = 7.8 × 10¹¹ m. Distance d = 10 pc = 3.1 × 10¹⁷ m. Star's orbital semi-major axis (in physical units):

a_star = (m_p / m_star) × a
       = 9.5 × 10⁻⁴ × 7.8 × 10¹¹
       = 7.4 × 10⁸ m
       ≈ 0.005 AU

Angular size on the sky:

α = a_star / d
  = 7.4 × 10⁸ / 3.1 × 10¹⁷
  = 2.4 × 10⁻⁹ rad
  = 500 μas

The orbital period is Jupiter's: P = 11.86 years. Over the 10-year Gaia mission, the star completes ~0.85 orbits — almost a full cycle, more than enough to characterise the orbit. With Gaia precision σ_epoch = 7 μas and ~70 epochs per source over the mission, the expected per-source mean-position precision is σ_mean ≈ σ_epoch / √70 ≈ 0.8 μas — orders of magnitude smaller than the 500 μas signal. Detection is straightforward.

Now the same analysis for an Earth at 1 AU around the same star: m_p / m_star = 3 × 10⁻⁶, a = 1 AU. Astrometric signal:

α_Earth = 3 × 10⁻⁶ × 1 / 2 × 10⁶
        = 1.5 × 10⁻¹² rad
        = 0.3 μas

Even with full Gaia mission averaging, this is right at the noise floor — and that floor includes systematic errors that may not average down. Detecting Earth analogs at 10 pc requires the next generation of missions (Theia, GAIA-NIR), with sub-microarcsecond per-epoch precision.

Variants and extensions

• Combined astrometry + RV. A planet's orbit has seven Keplerian parameters. RV measures K, P, eccentricity, longitude of perihelion (4 parameters; constraint on m sin i). Astrometry adds inclination i and longitude of node Ω (2 more parameters; constraint on m). Together: complete 3D orbit and true mass. The combination is the gold standard.
• Reflex motion of multiple-planet systems. A star with two or more planets traces a superposition of ellipses. Astrometric fitting recovers each planet's contribution if periods are well separated and the time baseline covers many cycles. Gaia DR4 onward will deliver multi-planet systems.
• Differential astrometry. Measure target star position relative to nearby reference stars, not absolute. Cancels common-mode error from spacecraft attitude. Gaia uses both absolute (parallax baseline) and differential (relative shift) astrometry; ground-based interferometry can only do differential.
• Interferometric astrometry. The VLT Interferometer (VLTI) combines four 8-m UTs at Cerro Paranal. With GRAVITY+ (2024 upgrade) it achieves ~10 μas differential astrometry at K-band, enabling pointed follow-up of nearby imaged planets like β Pictoris b. Different sensitivity profile from all-sky Gaia.
• Microlensing astrometry. The Roman Space Telescope (launching 2027) will combine its microlensing exoplanet survey with bonus astrometry of source stars during events — a hybrid that breaks the lens-mass degeneracy in microlensing.

Where astrometry matters

• Population statistics of Jupiter analogs. The 5-AU wide-orbit population is hard to find with transit (long periods, low transit probability) or RV (long monitoring required). Astrometry is the natural tool. Gaia's expected ~70,000 detections will define the giant-planet occurrence rate at intermediate separations.
• Mass measurement for transiting planets. Transit method gives radius. Combined with astrometric mass: planet density and bulk composition. Especially useful for low-mass planets where RV reflex velocity is below precision floor.
• Brown-dwarf/planet boundary. Astrometry can probe down to ~13 Jupiter masses (the deuterium-burning boundary). The mass function in this regime is otherwise poorly determined.
• Stellar physics. Gaia's parallaxes (which are astrometric byproducts) calibrate the distance scale to thousands of nearby stars and clusters; the same data product underpins much of modern stellar physics.
• Future habitable-zone Earth detection. Theia, GAIA-NIR, and similar mission concepts target the ~0.3 μas regime needed for Earth at 10 pc. If launched in the 2030s, these will turn astrometric Earth detection from theoretical aspiration to reality.

Common pitfalls

• Confusing astrometric wobble with parallax. Parallax (~2 × 10⁵ μas at 10 pc) is much larger than typical exoplanet wobbles. Astrometric data analysis must first subtract parallax and proper motion before searching for orbital signals — the orbital ellipse is a tiny perturbation on a much larger annual parallax ellipse.
• Treating astrometry as a small RV-equivalent. Astrometry returns true mass and 3D orbit — both extras over RV. The data analysis is correspondingly different (epoch-by-epoch 2D positions versus epoch-by-epoch line-of-sight velocity).
• Ignoring photocentric jitter. Starspots, irradiation, and faint binary companions can shift the apparent center of brightness of a star on the sky, mimicking small astrometric wobble. Gaia DR2-onward analysis explicitly models these.
• Overestimating precision over short baselines. Astrometric precision improves with mission length. Gaia's 10-year baseline is what enables 0.8 μas mean-position precision; halving the baseline doubles the noise.
• Forgetting the m_p signal scales with planet mass, not period. Doubling the orbital period at fixed mass keeps the astrometric signal the same. Doubling the planet mass doubles the signal regardless of period.