| dc.description.abstract | In the last 20 years, the radial velocity (RV) method has successfully detected and characterised hundreds of exoplanets, from blazing hot giants to small super-Earths. With the recent sub-m s−1 precision reached by ultra-stable spectrographs, the signals of rocky extra-solar planets in long-orbits have finally become detectable. The greatest challenge for the characterisation of exoplanets is now stellar variability. The effects of activity on the surface of stars often strongly dominate the RV budget, and can easily obscure or mimic Keplerian signals. Gaussian Processes (GPs) have been proven to be a very successful tech- nique for the mitigation of stellar effects, as they are able to model the variability without making any assumption about its functional form. In this work, I introduce MAGPy-RV. MAGPy-RV is a Gaussian process regression pipeline with Markov Chain Monte Carlo parameter space searching algorithm I developed in the context of exoplanet detection and characterisation. It allows to simultaneously model stellar activity, described by a GP with the chosen covariance function, and Keplerian signals in the RVs as well as transits in photometric data.
I then use this pipeline for the analysis of two planetary systems: TOI-2134 and HD 48948. The moderately active, bright K5V star TOI-2134 is orbited by an inner mini-Neptune in a 9.2292005±0.0000063 day orbit and an outer mono-transiting sub-Saturn planet in a 95.50(+0.36, −0.25) day orbit. Based on the analysis of TESS data, I determine the radii of TOI-2134b and c to be 2.69±0.16 R⊕ for the inner planet, and 7.27±0.42 R⊕ for the outer one. The masses of both planets are derived based on HARPS-N and SOPHIE RVs via
Gaussian process regression to be 9.13 (+0.78, −0.76) M⊕ for TOI-2134b and 41.89 (+7.69, −7.83) M⊕ for TOI-2134c. The outer planet is computed to have a significant eccentricity of 0.67 (+0.05, −0.06) from a combination of photometry and RVs. The HD 48948 system comprises of three super-Earth planetary candidates with orbital periods of roughly 7.3, 38, and 151 days, and minimum masses estimated to be 4.96 ± 0.42 M⊕, 7.45 ± 0.75 M⊕, and 10.67 ± 0.90 M⊕, respectively. The outermost planet is also found to reside within the (temperate) habitable zone, positioned at a projected distance of 0.029′′ from its star.
Both these analyses highlighted the need for caution and systematic model testing when employing GPs to model stellar-induced signals.
In parallel, I also analyse solar data in order to develop a better understanding of the processes driving stellar variability. The Sun is a fairly representative star of the sample of targets that are generally selected for RV surveys, and most crucially it is the only star we can resolve. I use the SolAster pipeline to derive disc-integrated longitudinal magnetic field data in order to asses its uses in radial-velocity surveys. I show that the mean longitu- dinal magnetic field is an excellent rotation period detector and a useful tracer of the solar magnetic cycle. In order to put these results into context, I compare the mean longitudi- nal magnetic field to three common activity proxies derived from HARPS-N Sun-as-a-star data: the full-width at half-maximum, the bisector span and the S-index. I find that the mean longitudinal magnetic field cannot be used as a one-to-one proxy, but that it out- performs all other considered indicators as a solar rotational period detector, and can be used to inform our understanding of the physical processes happening on the surface of the Sun. | en_GB |
