Stellar-activity analysis of the nearby M dwarf GJ 526. Multi-dimensional Gaussian-process modelling of RV, FWHM and S-index

TL;DR

GJ 526 is analysed with a multi-instrument, multi-quantity GP framework to disentangle stellar activity from potential planets. The joint RV, FWHM, and S-index modelling yields a rotation period of $P_ ext{rot}=48.7\pm0.3$ d and a magnetic-cycle period of $P_ ext{cyc}=1680^{+50}_{-40}$ d, with no compelling planetary signals detected; ESPRESSO data drive the inference and a phase-space trajectory view highlights activity dynamics. The study shows that the activity-driven signals can mimic or obscure planets, sets stringent detection limits (e.g., around $1\,\text{m s}^{-1}$ for certain periods), and introduces phase-space trajectories as a diagnostic tool for stellar magnetism effects on RV measurements. The results have important implications for planet searches around active M dwarfs and demonstrate a robust methodology for separating activity from planetary signals using multi-dimensional GP kernels.

Abstract

M dwarfs are the most abundant stars in the Galaxy, and their low masses make them natural favourites for exoplanet radial-velocity (RV) searches. Nevertheless, these stars often demonstrate strong stellar activity that is yet to be fully understood. We use high-precision ESPRESSO, HIRES, and HARPS spectroscopy to perform a stellar-activity analysis on the nearby early M dwarf GJ 526 ($d=5.4$ pc). We carry out joint modelling of: (i) stellar rotation in RV, FWHM, and Mount Wilson S-index through Gaussian processes, (ii) long-term trends in these three physical quantities, (iii) RV planetary signals. We constrain the stellar-rotation period to $P_\text{rot}=48.7\pm 0.3$ d, and discover a weak sinusoidal signature in RV, FWHM and S-index of period $P_\text{cyc}=1680^{+50}_{-40}$ d. We propose phase-space trajectories between RV and activity indicators as a novel means to visualise and interpret stellar activity. Current evidence does not support planetary companions of GJ 526.

Figures (14)

• Figure 1: Timeseries of mean-subtracted: (a) RV, (c) FWHM, (e) S-index. Measurements are marked depending on the data source: pink downward triangles for HIRES, teal upward triangles for HARPS, blue circles for ESPRESSO18, and green squares for ESPRESSO19. (b,d,f) Associated wide-period GLSPs of RV and activity indicators. Three FAP levels: 10%, 1% and 0.1%, split GLSP ordinates in bands of different colour. We highlight the three most prominent peaks in each GLSP.
• Figure 2: Bayesian-evidence comparison between different planetary configurations and stellar activity kernels. Planetary configurations include a circular-orbit component (C$_\text{b}$; $e=0$). For stellar activity, we utilised the MEP kernel and the ESP kernel with 2, 3 and 4 harmonics. The model that we further elect for analysis assumed a planet-free model with a MEP kernel (red border; $\ln Z=-434.4$). We give the Bayesian factor $\Delta\ln Z$ of remaining models relative to this model.
• Figure 3: Raw time series (markers) against our best model (black; LTF+GP) in two selected intervals. Data points come with two error bars: one assigned by the instrument (saturated), and another that includes the model jitter (semi-transparent). We provide with a direct comparison and associated O-C diagrams in: (a,b,c,d) RV, (e,f,g,h) FWHM, (i,j,k,l) S-index.
• Figure 4: Phase-space trajectories built from our best-model predictions in Fig. \ref{['fig:bestmodel_fit']}, in the same selected intervals. We compare RV with: (a,b) FWHM, (c,d) the time derivative of FWHM, (e,f) S-index, (g,h) the time derivative of S-index. Time runs from blue to yellow, as defined by the colorbars at the right-hand side. The same colorbars contain the temporal location of ESPRESSO18 (circles) and ESPRESSO19 (squares) measurements. Every trajectory comes with a mark at the bottom right that guides the direction of evolution, as well as the median model uncertainty in each dimension.
• Figure 5: Detection limits of GJ 526. Shaded bands show the $3\sigma$ RV semi-amplitude posterior of a modelled planet in our best model, in 100 log-spaced orbital-period bins. Two dotted lines show the RV semi-amplitudes that an Earth-mass planet and an Neptune-mass planet would inject in the system.