Which Radial Velocity Exoplanets Have Undetected Outer Companions?

(Abridged) The observed radial velocity (RV) eccentricity distribution for extrasolar planets in single-planet systems shows that a significant fraction of planets are eccentric ($e > 0.1$). Here we investigate the effects on an RV planet’s eccentricity produced by undetected outer companions. We have carried out Monte Carlo simulations of mock RV data to understand this effect and predict its impact on the observed distribution. We first quantify the statistical effect of undetected outer companions and show that this alone cannot explain the observed distribution. We then modify the simulations to consist of two populations, one of zero-eccentricity planets in double-planet systems and the other of single planets drawn from an eccentric distribution. Our simulations show that a good fit to the observed distribution is obtained with 45% zero-eccentricity double-planets and 55% single eccentric planets. Matching the observed distribution allows us to determine the probability that a known RV planet’s orbital eccentricity has been biased by an undetected wide-separation companion. Our simulations show that moderately-eccentric planets, with $0.1 < e < 0.3$ and $0.1 < e < 0.2$, have a $\sim 13%$ and $\sim 19%$ probability, respectively, of having an undetected outer companion. We encourage both high-contrast direct imaging and RV follow-up surveys of known RV planets with moderate eccentricities to test our predictions and look for previously undetected outer companions.

💡 Research Summary

The paper tackles a long‑standing puzzle in radial‑velocity (RV) exoplanet surveys: a surprisingly large fraction of apparently single‑planet systems exhibit moderate to high orbital eccentricities (e > 0.1), whereas planet‑formation models often predict near‑circular orbits for isolated planets. The authors ask whether undetected wide‑separation companions could be biasing the measured eccentricities of the inner, RV‑detected planets.

To address this, they perform extensive Monte Carlo simulations. In the first set they generate synthetic RV data for a single inner planet while adding a distant, massive companion whose orbital parameters (mass, semi‑major axis, eccentricity, phase) are drawn from the observed distribution of long‑period exoplanets. The synthetic data are then fit with a standard single‑Keplerian model, and the recovered eccentricities are compared to the input circular orbit. This exercise shows that the presence of an outer companion can indeed inflate the apparent eccentricity, but the effect alone cannot reproduce the full observed eccentricity distribution, especially the high‑e tail.

Consequently, the authors construct a two‑population model. Population A consists of double‑planet systems in which the inner planet’s true orbit is circular (e = 0) and the outer planet is a wide, massive companion. Population B consists of genuinely eccentric single‑planet systems. By varying the fraction f of Population A (0 % ≤ f ≤ 100 %) and repeating the Monte Carlo procedure for each mixture, they generate a family of synthetic eccentricity histograms. These are compared to the empirical RV eccentricity distribution (primarily from the Butler et al. 2006 and Wright et al. 2011 catalogs) using a Kolmogorov‑Smirnov test. The best match occurs for f ≈ 45 % (i.e., 45 % of the observed sample are circular inner planets with hidden outer companions) and 55 % are truly eccentric single planets. This mixed model reproduces the observed frequency of planets with 0.1 < e < 0.3 remarkably well.

From the optimal mixture the authors compute the probability that any given observed planet’s eccentricity is contaminated by an unseen companion. For planets with measured eccentricities between 0.1 and 0.3, the contamination probability is ≈13 %; for the narrower range 0.1 < e < 0.2 it rises to ≈19 %. In other words, a non‑negligible subset of “moderately eccentric” RV planets may in fact be on circular orbits, with the apparent eccentricity arising from the gravitational influence of a distant, undetected companion.

The study’s strengths lie in its quantitative treatment of observational bias, the use of realistic companion parameter distributions, and the direct statistical comparison with the actual RV catalog. However, several caveats are acknowledged. The outer‑companion population is modeled on currently known long‑period planets, potentially missing a large reservoir of low‑mass, ultra‑wide companions. Stellar activity and instrumental noise are treated in a simplified manner, whereas real RV data often contain correlated noise that could further affect eccentricity estimates. Moreover, the assumption that the inner planet in double‑planet systems is perfectly circular ignores possible secular perturbations that could induce small but non‑zero eccentricities.

Despite these limitations, the paper provides a clear, testable prediction: RV planets with moderate eccentricities are prime targets for high‑contrast direct imaging (e.g., with VLT/SPHERE, Gemini/GPI, or upcoming ELT instruments) and for extended RV monitoring to search for long‑period trends. Detecting or ruling out outer companions in a statistically significant sample would directly validate or refute the proposed 45 %/55 % mixture, thereby refining our understanding of the true eccentricity distribution of exoplanets and informing models of planetary system architecture and dynamical evolution.