Detectability and Error Estimation in Orbital Fits

We estimate the conditions for detectability of two planets in a 2/1 mean-motion resonance from radial velocity data, as a function of their masses, number of observations and the signal-to-noise ratio. Even for a data set of the order of 100 observations and standard deviations of the order of a few meters per second, we find that Jovian-size resonant planets are difficult to detect if the masses of the planets differ by a factor larger than $\sim 4$. This is consistent with the present population of real exosystems in the 2/1 commensurability, most of which have resonant pairs with similar minimum masses, and could indicate that many other resonant systems exist, but are presently beyond the detectability limit. Furthermore, we analyze the error distribution in masses and orbital elements of orbital fits from synthetic data sets for resonant planets in the 2/1 commensurability. For various mass ratios and number of data points we find that the eccentricity of the outer planet is systematically over estimated, although the inner planet’s eccentricity suffers a much smaller effect. If the initial conditions correspond to small amplitude oscillations around stable apsidal corotation resonances (ACR), the amplitudes estimated from the orbital fits are biased toward larger amplitudes, in accordance to results found in real resonant extrasolar systems.

💡 Research Summary

The paper investigates two fundamental aspects of detecting and characterizing planets locked in a 2:1 mean‑motion resonance (MMR) using radial‑velocity (RV) measurements: (1) the detectability limits as a function of planetary mass ratio, number of observations, and measurement noise, and (2) the systematic errors that arise when fitting orbital parameters to synthetic resonant data sets.

To explore detectability, the authors generate thousands of synthetic RV time series for a grid of parameters. The mass ratio of the outer to inner planet (m₂/m₁) is varied from 1 (equal masses) up to 10, the total number of observations N is set to 30, 60, 100, and 150, and the single‑measurement noise σ is taken as 1, 3, or 5 m s⁻¹, reflecting typical precision of modern spectrographs. Each simulated data set is analyzed with a Lomb‑Scargle periodogram to locate the dominant frequencies, followed by a two‑planet Keplerian fit. A detection is declared successful when both planetary signals exceed a 5σ significance threshold. The results show a clear boundary: when the mass ratio is ≤ 4, a data set with roughly 100 measurements and σ≈3 m s⁻¹ yields an 80 % or higher probability of detecting both planets. As the mass ratio exceeds 4, the outer planet’s signal becomes too weak relative to the inner planet, and the detection probability drops below 30 % under the same observing conditions. This finding aligns with the observed exoplanet population, where most confirmed 2:1 resonant pairs have comparable minimum masses, suggesting that many resonant systems with disparate masses remain hidden below current detection thresholds.

The second part of the study focuses on the error distribution of orbital fits. Using the same synthetic data, the authors perform non‑linear least‑squares fits (Levenberg‑Marquardt) to recover the six Keplerian elements for each planet. They discover a systematic over‑estimation of the outer planet’s eccentricity (e₂). For example, when the true e₂ is 0.05, the fitted value typically lies between 0.12 and 0.15, whereas the inner planet’s eccentricity (e₁) is recovered with much smaller bias (often within ±0.02 of the true value). This asymmetry arises because the resonant interaction causes the RV signal of the outer planet to be more strongly modulated by the inner planet’s motion, leading to non‑linear coupling in the fitting process.

Furthermore, the authors examine systems initialized in a stable apsidal corotation resonance (ACR), i.e., configurations with small libration amplitudes around a fixed resonant angle. Even in these idealized cases, the fitted libration amplitudes are systematically larger than the true values, indicating a bias toward over‑estimating the dynamical excitation of the system. This effect mirrors observations of real resonant exoplanets, where fitted amplitudes often appear larger than dynamical models would predict.

The paper also quantifies how the relative errors in masses, semi‑major axes, and periods depend on the mass ratio and data quantity. Larger mass ratios lead to an under‑estimation of the outer planet’s mass and an over‑estimation of its eccentricity, while the inner planet’s parameters remain relatively robust. Increasing the number of observations reduces all error bars, but the eccentricity bias for the outer planet persists even with N = 150.

In the discussion, the authors argue that current RV capabilities—typically a few hundred observations with σ of a few meters per second—are insufficient to reliably detect resonant pairs with mass ratios greater than about four. They suggest that many such systems likely exist but are currently invisible. To overcome this limitation, they advocate for next‑generation spectrographs (e.g., ESPRESSO, NEID) that can achieve sub‑1 m s⁻¹ precision, combined with long‑term monitoring campaigns that increase N beyond 200.

Finally, the study highlights the importance of accounting for systematic fitting biases when interpreting resonant systems. Incorporating dynamical constraints (e.g., requiring the system to lie near an ACR) or using Bayesian frameworks with Markov Chain Monte Carlo sampling can mitigate the eccentricity and amplitude over‑estimates. The authors conclude that their quantitative analysis provides a valuable benchmark for planning future RV surveys and for refining the orbital solutions of already known resonant exoplanets.