A Gaussian Process Regression Reveals No Evidence for Planets Orbiting Kapteyn's Star
Abstract
A significant issue with radial velocity (RV) planet searches is that they are often polluted by signals caused by activity on the star's surface. Stellar activity can mimic or mask changes in the radial velocities caused by orbiting planets, resulting in false positives. Here we test whether Gaussian process (GP) regression (e.g. Haywood et al. 2014) can distinguish star rotation from wobbles due to planets. We examine the RV and Hα data for Kapteyn's star, which has reported planets Kapteyn b and c with periods 48.6 and 121.5 days, respectively (Anglada-Escude et al. 2014). Based on the Hα observations, which trace magnetic activity in M dwarfs, Robertson et al. (2015) argued that Kapteyn b was an alias of the rotation. Here we use a quasiperiodic GP kernel to model rotation in the RV and Hα data jointly, requiring that their period and decorrelation timescale be the same. From this model, we construct residual RV datasets after subtracting off realizations of the GP rotation model. We find no evidence that Kapteyn's star hosts any planets, as there were no significant signals that remained in the residual data. We suggest that the periodic signals of the previously reported planets are both products of the star's rotation.
- Publication:
-
American Astronomical Society Meeting Abstracts #234
- Pub Date:
- June 2019
- Bibcode:
- 2019AAS...23410901J