Uncovering the planets and stellar activity of CoRoT-7 using only radial velocities

Abstract

Stellar activity can induce signals in the radial velocities of stars, complicating the detection of orbiting low-mass planets. We present a method to determine the number of planetary signals present in radial-velocity datasets of active stars, using only radial-velocity observations. Instead of considering separate fits with different number of planets, we use a birth-death Markov chain Monte Carlo algorithm to infer the posterior distribution for the number of planets in a single run. In a natural way, the marginal distributions for the orbital parameters of all planets are also inferred. This method is applied to HARPS data of CoRoT-7. We confidently recover the orbits of both CoRoT-7b and CoRoT-7c although the data show evidence for the presence of additional signals.

Heavy Metal Rules. I. Exoplanet Incidence and Metallicity

“…However, the detection of low-mass/small-sized planets is not an easy task not only because of the very small photoelectric and RV signals, but also because of the activity signals coming from the host stars that often have the same magnitude as the planetary signal and can strongly perturb the detection of these planets and/or mimic a planetary signal [e.g. [248][249][250][251][252][253][254]. These difficulties not only limits the number of detected low-mass planets, but also make very hard to construct a control sample of stars without low-mass planets for RV surveys and make practically impossible 30 for transit surveys [255].…”

The HARPS search for southern extra-solar planets

“…These signals are often very complex, and a physical model is therefore usually hard to produce or has the strong disadvantage of being computationally expensive. Gaussian process regression (GP) has become a widely used non-parametric method to model this variability without the necessity of specifying a model (Rasmussen & Williams 2006;Haywood et al 2014;Rajpaul et al 2015;Faria et al 2016), although it may be prone to inducing false positives if care is not taken (Dumusque et al 2017).…”

Radial-velocity fitting challenge

“…For short-term activity, which is by far the most difficult stellar signal to deal with due to the nonperiodic, stochastic, long-term signals arising from the evolution and decay of active regions, several correction techniques have been investigated: -fitting sine waves at the rotation period of the star and harmonics (Boisse et al 2011), -using red-noise models to fit the data (e.g. Feroz & Hobson 2014;Gregory 2011;Tuomi et al 2013), -using the FF method if contemporaneous photometry exists (Dumusque et al 2015;Haywood et al 2014;Aigrain et al 2012), -modeling activity-induced signals in RVs with Gaussian process regression, whose covariance properties are shared either with the star's photometric variations (Haywood et al 2014;Grunblatt et al 2015) or a combination of several spectroscopic indicators (Rajpaul et al 2015), or determined from the RVs themselves (Faria et al 2016), -using linear correlations between the different observables, i.e., RV, bisector span (BIS SPAN) and full width at half maximum (FWHM) of the cross correlation function (CCF, Baranne et al 1996;Pepe et al 2002), photometry (Robertson et al 2015(Robertson et al , 2014Boisse et al 2009;Queloz et al 2001), and magnetic field strength (Hébrard et al 2014), -checking for season per season phase incoherence of signals (Santos et al 2014;Dumusque et al 2014bDumusque et al , 2012, -avoiding the impact of activity by using wavelength dependence criteria for RV signal (e.g. in HD40307 and HD69830, Tuomi et al 2013;Anglada-Escudé & Butler 2012).…”