The SNR of idealised radial velocity signals

TL;DR

This paper derives analytical expressions for the signal-to-noise ratio (SNR) of idealised radial velocity signals, considering factors like eccentricity and orbital alignment, providing insights for exoplanet detection and characterization.

Contribution

It presents the first comprehensive derivation of RV SNR formulas accounting for eccentricity, orbital alignment bias, and the Rossiter-McLaughlin effect, filling a gap in prior theoretical work.

Findings

SNR scales as K√T, with eccentricity and argument of periastron dependence.

Bias towards orbital alignment with the observer is confirmed.

Discriminatory SNR for eccentricity detection is typically an order of magnitude less than for the planet signal.

Abstract

One of the most basic quantities relevant to planning observations and assessing detection bias is the signal-to-noise ratio (SNR). Remarkably, the SNR of an idealised radial velocity (RV) signal has not been previously derived beyond scaling behaviours and ignoring orbital eccentricity. In this work, we derive the RV SNR for three relevant cases to observers. First, we consider a single mass orbiting a star, revealing the expected result that $\mathrm{SNR}\propto K \sqrt{T}$, where $T$ is the observing window, but an additional dependency on eccentricity and argument of periastron. We show that the RV method is biased towards companions with their semi-major axes aligned to the observer, which is physically intuitive, but also less obviously that the marginalised bias to eccentricity is negligible until one reaches very high eccentricities. Second, we derive the SNR necessary to discriminate eccentric companions from 2:1 resonance circular orbits, although our result is only valid for eccentricities $e\lesssim0.3$. We find that the discriminatory SNR is $(9/8) e^2 (1-e^2)^{-1/2}$ times that of the eccentric planet solution's SNR, and is thus typically an order-of-magnitude less. Finally, we have obtained a semi-empirical expression for the SNR of the idealised Rossiter-McLaughlin effect, revealing the bias with respect to spin-orbit alignment angle. Our formula is valid to within 10% accuracy in 95.45% of the training samples used (for $b\leq0.8$), but larger deviations occur when comparing to different RM models.