On the Completeness of Reflex Astrometry on Extrasolar Planets near the Sensitivity Limit

We provide a preliminary estimate of the performance of reflex astrometry on Earth-like planets in the habitable zones of nearby stars. In Monte Carlo experiments, we analyze large samples of astrometric data sets with low to moderate signal-to-noise ratios. We treat the idealized case of a single planet orbiting a single star, and assume there are no non-Keplerian complications or uncertainties. The real case can only be more difficult. We use periodograms for discovery and least-squares fits for estimating the Keplerian parameters. We find a completeness for detection compatible with estimates in the literature. We find mass estimation by least squares to be biased, as has been found for noisy radial-velocity data sets; this bias degrades the completeness of accurate mass estimation. When we compare the true planetary position with the position predicted from the fitted orbital parameters, at future times, we find low completeness for an accuracy goal of 0.3 times the semimajor axis of the planet, even with no delay following the end of astrometric observations. Our findings suggest that the recommendation of the ExoPlanet Task Force (Lunine et al. 2008) for “the capability to measure convincingly wobble semi-amplitudes down to 0.2 $\mu$as integrated over the mission lifetime,” may not be satisfied by an instrument characterized by the noise floor of the Space Interferometry Mission, $\sigma_\mathrm{floor}\approx0.035\mu$as. An important, unsolved, strategic challenge for the exoplanetary science program is figuring out how to predict the future position of an Earth-like planet with accuracy sufficient to ensure the efficiency and success of the science operations for follow-on spectroscopy, which would search for biologically significant molecules in the atmosphere.

💡 Research Summary

The paper presents a quantitative assessment of the capability of reflex astrometry to detect and characterize Earth‑like planets in the habitable zones of nearby stars when the signal‑to‑noise ratio (S/N) is low to moderate. The authors adopt an idealized scenario: a single planet orbiting a single star, with no non‑Keplerian perturbations or additional sources of uncertainty. They generate large ensembles of synthetic astrometric data sets using Monte Carlo simulations, assigning random orbital elements (semi‑major axis, eccentricity, inclination, argument of periastron, mean anomaly) and a planetary mass drawn from an Earth‑like distribution. Observations are simulated over a five‑year mission, with twelve measurements per year, each corrupted by Gaussian noise with a standard deviation equal to the Space Interferometry Mission (SIM) noise floor, σ_floor ≈ 0.035 µas.

Detection is performed with Lomb‑Scargle periodograms, adopting a false‑alarm probability threshold of 1 %. The resulting detection completeness matches prior estimates: roughly 50 % of planets are recovered at S/N ≈ 2, rising to about 90 % at S/N ≈ 4. After a detection, the seven Keplerian parameters (including planetary mass) are estimated via a non‑linear least‑squares fit (Levenberg‑Marquardt algorithm). The authors find that, especially for low S/N, the mass estimates are systematically biased low. This bias mirrors the well‑known “noise‑induced bias” observed in radial‑velocity studies, where the non‑linear dependence of the observable on the parameters leads to a skewed estimator when the data are noisy. The bias reduces the completeness for the goal of “mass accurate to within 20 %,” dropping it well below the detection completeness.

A central focus of the study is the ability to predict the future position of the planet after the astrometric campaign ends. The authors propagate the fitted orbital solution forward in time and compare the predicted sky‑plane location with the true position at intervals up to ten years beyond the mission. They adopt an accuracy requirement of 0.3 × the planetary semi‑major axis (≈0.3 a). Even with no observational gap after the mission, the completeness for meeting this requirement remains low: at S/N ≈ 5 it is under 30 %, and at S/N ≈ 2 it falls below 10 %. The degradation is driven by uncertainties in orbital phase and eccentricity that accumulate over time, making long‑term predictions unreliable.

The authors conclude that the ExoPlanet Task Force recommendation—“the capability to measure convincingly wobble semi‑amplitudes down to 0.2 µas integrated over the mission lifetime”—is insufficient to guarantee the scientific outcomes needed for follow‑up spectroscopy. While a 0.2 µas wobble may be detectable, the associated uncertainties in mass and future position are too large for efficient targeting of direct‑imaging or spectroscopic missions. They argue that achieving the positional accuracy needed to schedule high‑contrast spectroscopy (e.g., within 0.3 a) would require S/N ≈ 10, far beyond the performance of an instrument limited by σ_floor ≈ 0.035 µas.

Strategically, the paper suggests several avenues to mitigate these limitations: (1) replace simple least‑squares fitting with Bayesian inference or Markov‑Chain Monte‑Carlo sampling to properly account for parameter covariances and reduce bias; (2) combine astrometry with complementary techniques such as radial velocity, transit timing, or direct imaging to break degeneracies; (3) schedule intermittent astrometric follow‑up after the primary mission to refine orbital phase; and (4) incorporate realistic stellar jitter and non‑Keplerian effects into the model to better reflect the true error budget.

In summary, the study validates that reflex astrometry can detect Earth‑like planets near the sensitivity limit, but it also highlights a critical shortfall: the precision required to predict future planetary positions for subsequent atmospheric characterization is not met by the nominal noise floor of current mission concepts. Overcoming this shortfall will demand improved data‑analysis methodologies, multi‑technique synergy, and possibly more stringent instrument performance than presently envisioned.