Exoplanet Detection Around M-Dwarfs with Near Infra-Red Andvisible

Home , 2018 AH, Doppler spectroscopy, ELODIE spectrograph, Gliese 49, Gliese 784, Hot Neptune

AIX-MARSEILLE UNIVERSITÉ ECOLE DOCTORALE 352 LABORATOIRE D’ASTROPHYSIQUE DE MARSEILLE

Thèse présentée pour obtenir le grade universitaire de docteur

Discipline : Physique et Sciences de la Matière Spécialité : Astrophysique et Cosmologie

Melissa J. HOBSON

Exoplanet Detection Around M Dwarfs with Near Infrared and Visible Spectroscopy

Détection des exoplanètes autour de naines M par spectroscopie proche infra-rouge et visible

Soutenue le 08/10/2019 devant le jury composé de :

Eduardo MARTIN Centro de Astrobiología (INTA-CSIC) Rapporteur Peter WHEATLEY Department of Physics, University of Warwick Rapporteur Magali DELEUIL Laboratoire d’Astrophysique de Marseille Présidente du jury Xavier DELFOSSE Institut de Planetologie et d’Astrophysique de Grenoble Examinateur Christophe LOVIS Observatoire de Genève Examinateur Isabelle BOISSE Laboratoire d’Astrophysique de Marseille Co-directrice de thèse François BOUCHY Laboratoire d’Astrophysique de Marseille Directeur de thèse

Numéro national de thèse/suﬃxe local : 2019AIXM0317/049ED352 Cette oeuvre est mise à disposition selon les termes de la Licence Creative Com- mons Attribution - Pas d’Utilisation Commerciale - Pas de Modiﬁcation 4.0 Internatio- nal. Résumé

Récemment, la science exoplanètaire s’est focalisé sur les étoiles M comme cibles intéressantes pour la détection et caractérisation des exoplanètes. Il y a plusieurs rai- sons pour ça : Les naines M sont les étoiles les plus communes de la galaxie ; leur petit taille implique qu’on peut détecter des planètes plus petits qu’autour des étoiles G ; la zone d’habitabilité d’eau liquide est plus proche à l’étoile. La population émergente de planètes autour de naines M montre des caractéristiques intrigantes par rapport aux planètes des étoiles FGK, comme l’absence des Jupiters chauds, et l’incertitude de la corrélation planète-metalicité. L’objectif de cette thèse est d’explorer la détection d’exoplanètes autour de naines M par la méthode de vitesse radiale, dans les domaines du visible et du prochaine infrarouge. J’ai aussi réalisé un analyse statistique de la population connue de planètes autour de naines M, comme il était au début et à la fin de la thèse. Dans le visible, j’ai travaillé avec le spectrographe SOPHIE à l’Observatoire d’Haute- Provence, comme partie du consortium SOPHIE exoplanètes. Cet groupe, qui regroupe des chercheurs dans multiples institutes en France et des pays voisins, a effectué diverses programmes de recherche d’exoplanètes, dont une se focalise sur la recherche de planètes autour de naines M (avec spécial attention aux Neptunes et superTerres). J’ai appliqué une algorithme "template-matching" aux cibles de ce sous-programme, et ana- lysé les séries de vitesse radiale résultantes. À partir e cet analyise, j’ai pu confirmer l’importance des signaux périodiques que, si bien présents dans l’analyse CCF traditionnel, étaient partiellement cachés par le bruit. J’ai aussi étudié une variété d’indices d’acti- vité, trouvant ceux qui sont mieux adaptés aux spectres de SOPHIE. Les premiers quatre exoplanètes issus de ce sous-programme ont récemment été publiées ; j’ai été première auteur pour deux des trois articles. Dans le prochaine infra-rouge, j’ai travaillé avec le spectropolarimetre SPIRou au Canada France Hawaii Telescope, comme partie du consortium SPIRou. Ce nouvelle instrument a été conçu spécifiquement pour l’observation des naines M, qui sont faibles dans le visible et émettent la plupart de son radiation dans le prochaine infra-rouge. J’ai travaillé sur le développement du système de réduction des données, particulièrement sur la solution en longueur d’onde - c’est à dire, la correspondance entre la position en pixels et la longueur d’onde, qui est crucial pour mesurer des vitesses radiales précises. J’ai développé et testé des façons de combiner différents calibreurs de longueur d’onde, pour obtenir une solution en longueur d’onde précise.

Mots clés : exoplanètes, spectroscopie, naines M, vitesses radiales Abstract

In recent years, exoplanet science has begun to focus on M-dwarf stars as highly interesting targets for exoplanet detection and characterisation. The reasons for this are multiple: M dwarfs are the most common stars in the galaxy; their small size means smaller exoplanets can be detected than around G-type stars; the liquid-water habitable zone is closer to the star, hence these planets are faster and easier to detect. The emerging population of planets hosted by M dwarfs shows intriguing characteristics compared to planets hosted by FGK stars, such as a lack of hot Jupiters, and an uncertain planet- metallicity correlation. The aim of this thesis is to explore the detection of exoplanets around M dwarfs via the radial velocity method, in both the near infrared and visible domains. I also carried out a statistical analysis of the known population of planets around M dwarfs, as it stood both at the start of the thesis and at its conclusion. In the visible, I worked with the SOPHIE spectrograph at the Observatoire de Haute- Provence, as part of the SOPHIE exoplanets consortium. This group, which nucleates researchers in multiple institutes in France and neighbouring countries, has been carrying out several long-term exoplanet surveys, one of which focuses on the search for planets around M dwarfs (with special attention to Neptunes and superEarths). I applied a template-matching algorithm to the targets of this subprogramme, and analysed the resulting radial velocity time series. Through this analysis, I was able to confirm the significance of periodic signals that, while apparent in the traditional CCF analysis, were partially hidden by noise. I also studied a variety of stellar activity indicators, identifying those most suited to SOPHIE spectra. The first four exoplanets from this subprogramme have recently been published; I was lead author for two of the three papers. In the near infrared, I worked with the SPIRou spectropolarimeter at the Canada France Hawaii Telescope, as part of the SPIRou consortium. This new instrument was conceived specifically for the observation of M dwarfs, which are faint in the visible and emit most of their radiation in the infrared. I worked on the development of the data reduction pipeline, with specific focus on the wavelength solution - that is, the correspondence between pixel position and wavelength, which is crucial to the measurement of precise radial velocities. I developed and tested ways to combine different wavelength calibrators, for an accurate wavelength solution.

Keywords: exoplanets, spectroscopy, M dwarfs, radial velocities Contents

Résumé 3

Abstract 4

Contents 5

List of Figures 7

List of Tables 9

Résumé étendue en français 11 0.1 Introduction...... 11 0.1.1 La méthode de vitesse radiale...... 12 0.1.2 Recherche d’exoplanètes autour de naines M...... 13 0.2 Recherche en vitesse radiale des naines M avec le spectrographe SOPHIE 14 0.3 Développement du spectropolarimetre SPIRou ...... 16 0.4 Proprietés des planètes autours des naines M ...... 18 0.5 Conclusions...... 19

Introduction 21 1.1 A brief history of exoplanets...... 22 1.2 Characteristics of M-dwarf stars...... 24 1.3 The radial velocity method ...... 28 1.3.1 Limitations of the RV method...... 30 1.3.2 Visible and nIR spectrographs and surveys ...... 32 1.4 M-dwarf stars and their planets (as of 2016)...... 34 1.5 Thesis objectives ...... 39

2 M-dwarf RV search with the SOPHIE spectrograph 41 2.1 The SOPHIE spectrograph ...... 42 2.2 The M-dwarf sample...... 43 2.3 The template-matching method...... 45 2.4 Instrumental instabilities...... 48 2.4.1 Charge transfer inefficiency ...... 48 2.4.2 Nightly drift ...... 49 2.4.3 Long-term variation of the zero-point...... 50 2.5 Summary of SOPHIE results on M dwarfs...... 52 2.5.1 Discovery of new exoplanets...... 55 2.5.2 Confirmation or non-confirmation of published planets ...... 81 2.5.3 Stellar activity mitigation...... 83 2.6 Other contributions to SOPHIE RV programmes...... 90 2.6.1 High precision RV search for super-Earths...... 90 2.6.2 Simultaneous FP background correction...... 92

3 Development of the SPIRou nIR spectropolarimeter 95 3.1 The SPIRou spectropolarimeter...... 96 3.2 The SPIRou Data Reduction System...... 99 3.3 Validation tests and commissioning...... 102 3.4 Wavelength calibration development ...... 103 3.4.1 Hollow-cathode lamps ...... 103 3.4.2 Combination with Fabry-Pérot reference ...... 107 3.5 Validation and performances of the wavelength solution ...... 110 3.5.1 Impact of previous calibrations...... 111 3.5.2 Performances of the HC wavelength solution ...... 114 3.5.3 Combined HC-FP wavelength solution ...... 115 3.5.4 Impact on RV error ...... 117 3.5.5 Upcoming changes to the Data Reduction System ...... 119 3.6 SPIRou science programs ...... 120 3.6.1 Spirou Legacy Survey ...... 120 3.6.2 Synergy with SOPHIE...... 122

4 Properties of M-dwarf planets as of 2019 125 4.1 Overview and growth of the M dwarf planet population sample ...... 126 4.2 Stellar hosts...... 127 4.3 Mass-period diagrams ...... 129 4.4 Types of planets ...... 136 4.5 Habitable zone planets...... 140 4.6 A mass-metallicity correlation? ...... 142 4.7 Multiplanetary systems orbiting M dwarfs...... 143

Conclusions 145 5.1 Overview of results and conclusions...... 146 5.2 Future perspectives...... 147 5.2.1 SOPHIE-red ...... 148 5.2.2 Synergies between the SPIRou Legacy Survey and other M dwarf surveys ...... 148 5.2.3 Overlap with TESS (SPIRou - SOPHIE) ...... 149 5.2.4 Beyond SOPHIE and SPIRou: New and upcoming spectrographs . 150

Bibliography 153 List of Figures

1.1 Exoplanets by detection type ...... 23 1.2 Exoplanet mass-period diagram...... 24 1.3 Comparison of G, K, and M stellar spectra ...... 26 1.4 Habitable Zone for diﬀerent spectral types...... 27 1.5 Radial velocity method...... 28 1.6 Keplerian orbital elements...... 29 1.7 RV curve dependency on e and ω ...... 30 1.8 Mass-period diagram for M dwarf planets (2016)...... 37 1.9 Planet mass or sum of masses as function of stellar metallicity ...... 38

2.1 A raw SOPHIE spectra of an M3 dwarf ...... 43 2.2 Magnitude, distance, and spectral type histograms for the SOPHIE M dwarf targets ...... 45 2.3 A red SOPHIE spectral order of an M3 dwarf...... 46 2.4 Test of slope and width parameters...... 48 2.5 Radial velocity offset as a function of the signal-to-noise ratio ...... 49 2.6 Correction of the long-term variation of the zero-point...... 52 2.7 Number of observations and RV dispersion for SP3 stars with at least ten observations...... 54 2.8 Phase-folded curve for Gl 411...... 80 2.9 Phase-folded curve for GJ 687...... 81 2.10 Time series and phase-folded curve for Gl 686...... 82 2.11 Periodogram for Gl 15A...... 83 0 2.12 Median log(RHK) index values from HARPS and SOPHIE spectra . . . . 84 2.13 RV periodograms for different reductions...... 86 2.14 Activity indicator periodograms...... 87 2.15 Periodograms for Gl104, Gl617A, and Gl251...... 89 2.16 Correction of the long-term variation of the zero-point for sun-like stars . 91 2.17 Inadequacy of existing SOPHIE background routines for simultaneous FP 92 2.18 Comparison of the dark and FP behaviour with exposure time per order . 93 0 2.19 log(RHK) for different background corrections...... 94

3.1 SPIRou instrument design...... 97 3.2 SPIRou optical path...... 98 3.3 SPIRou calibration unit ...... 99 3.4 SPIRou DRS flowchart...... 100 3.5 Intermediate data products of the SPIRou calibration sequence ...... 101 3.6 Quick-look SPIRou flux tool...... 102 3.7 SPIRou HC lamp raw spectra...... 104 3.8 UNe catalogues comparison ...... 104 3.9 SPIRou FP étalon raw spectra...... 108 3.10 Variability of the FP cavity width ...... 110 3.11 Construction of a stable lines catalogue ...... 111 3.12 Impact of changing the SHAPE file on the wavelength solution...... 113 3.13 FP lines at the centre and edge of the detector ...... 114 3.14 Night-to-night variations of the FP solution ...... 116 3.15 CCF RV differences per order...... 119 3.16 H magnitude and distance of SLS-PS targets ...... 121

4.1 Evolution of the M dwarf planet population ...... 127 4.2 Spectral types of M-dwarf stars hosting planets...... 128 4.3 Distances and V magnitudes of M dwarfs with planets...... 129 4.4 Mass-period diagram for M dwarf planets (2019)...... 131 4.5 Mass-period diagram for G star planets (2019)...... 132 4.6 Mass-period diagrams colour-coded by Teﬀ , M∗, and [Fe/H]...... 133 4.7 Mass ratio-period diagram for M dwarf planets (2019)...... 135 4.8 M dwarf planets mass bar chart (2019)...... 137 4.9 Mass-radius plot for M dwarf planets...... 138 4.10 Composition models for M dwarf planets...... 139 4.11 Habitable Zone planets orbiting M dwarfs...... 141 4.12 Planet mass or sum of masses as function of stellar metallicity ...... 143 5.13 Current and new SOPHIE CCDs...... 148 5.14 TESS observing sectors ...... 150 List of Tables

1.1 Characteristics of M dwarf spectral classes...... 25

2.1 SOPHIE long-term drift offsets...... 51 2.2 Summary of the SOPHIE M dwarf RVs periodogram analysis...... 53 2.3 Summary of the SOPHIE M dwarf RV residuals periodogram analysis. . . 54 2.4 Most significant periods in RVs and activity indices...... 88 0 2.5 Median log(RHK) for different background corrections ...... 94

3.1 Summary of HC wavelength solution performances...... 114 3.2 Summary of HC-FP wavelength solution performances...... 115 3.3 Summary of CCF RV diﬀerences...... 118

4.1 Solar System planets’ densities ...... 138 4.2 Rocky habitable zone planets ...... 142

Résumé étendue

Contents

0.1 Introduction...... 11 0.1.1 La méthode de vitesse radiale...... 12 0.1.2 Recherche d’exoplanètes autour de naines M...... 13 0.2 Recherche en vitesse radiale des naines M avec le spectrographe SOPHIE 14 0.3 Développement du spectropolarimetre SPIRou ...... 16 0.4 Proprietés des planètes autours des naines M ...... 18 0.5 Conclusions...... 19

0.1 Introduction

Dès l’Antiquité, les civilisations humaines ont rêvé de découvrir d’autres mondes comme le nôtre. La première véritable découverte d’un exomonde remonte à 1992, lorsque Wolszczan & Frail(1992) démontrèrent l’existence d’au moins deux corps de la taille d’une planète orbitant autour du pulsar PSR B1257 + 12 en se basant sur des variations minimes dans la synchronisation des émissions radio. Trois ans plus tard, Mayor & Queloz(1995) entrèrent dans l’histoire pour avoir détecté, à l’aide du spectrographe ELODIE de l’Observatoire de Haute Provence (OHP), la première planète autour d’une étoile dans la séquence principale, à savoir 51 Pegasi. Cette planète fut la première d’une catégorie appelée Jupiters chauds. Il s’agit de planètes sans analogue dans le Système solaire, car elles sont à la fois de taille similaire à celle de Jupiter et situées très proches de leurs étoiles hôtes (a . 0.1 AU, P . 10 d). Au cours des vingt-quatre années qui nous séparent de la découverte de 51 Peg b, les détections d’exoplanètes se sont accrues de façon exponentielle pour atteindre un total dépassant les 4000 objets. Ces planètes ont été détectées par diverses techniques, dont les plus prolifiques sont la méthode de transit et celle des vitesses radiales. Bien que les premières exoplanètes découvertes aient été des Jupiters chauds, l’augmentation de la sensibilité des instruments et des échelles de temps sondées a permis de révéler une grande variété de populations planétaires. Après les planètes géantes à longue période, les seuils de detection arrivent aujourd’hui à identifier des planètes dont les tailles sont comprises entre celles de la Terre et de Neptune. Paradoxalement, des ré- gions dépeuplées sont également apparues dans les distributions masse-rayon, tels que 12 Résumé étendue en français le désert de Neptunes chauds (Lecavelier Des Etangs 2007, Beaugé & Nesvorný 2013, Mazeh et al. 2016) ou encore la lacune de Fulton qui est observée entre les super- Terres et les mini-Neptunes (Fulton et al., 2017). La diversification des planètes dé- couvertes s’est accompagnée d’une diversification tout aussi grande vis-à-vis des hôtes stellaires. Aujourd’hui, la science des exoplanétes est toujours en pleine croissance. Les mesures simultanées de la masse et du rayon des exoplanètes ont permis de développer des modèles de structure interne toujours plus sophistiqués (Baraffe et al. 2014, Brugger et al. 2017). Par ailleurs, les atmosphères elles-mêmes commencent à être accessibles d’un point de vue observationnel, stimulant les études par modélisation (e.g., Sing et al. 2016). Enfin, de nouveaux instruments toujours plus sensibles traquent des analogues de notre planète, c’est-à-dire des corps de la taille de la Terre situés dans les zones habitables de leurs étoiles hôtes.

0.1.1 La méthode de vitesse radiale

La vision d’une planète qui orbite autour de son étoile hôte est une simpliﬁcation. En réalité, la planète et l’étoile gravitent ensemble autour du barycentre du système étoile-planète. Cette différence est fondamentale, car elle implique le rapprochement et l’éloignement périodique d’une étoile abritant une planète pour un observateur im- mobile qui regarde le système. Par effet Doppler, ce mouvement induit un décalage des raies spectrales vers le bleu lorsque l’étoile se rapproche, ou bien vers le rouge lorsqu’elle s’éloigne. C’est sur ce principe que repose la méthode de la vitesse radiale encore ap- pelée spectroscopie Doppler. Pour une étoile se déplaçant avec la vitesse v sous un angle θ par rapport à la ligne de visée de l’observateur, le décalage en longueur d’onde ∆λ sera donné par l’équation relativiste de Doppler. Pour v c et θ π/2, la vitesse radiale se réduit à : ∆λ vr = v cos(θ) ≈ c (0.1) λem

En traduisant les décalages spectraux mesurés en vitesse radiale, il est possible de construire une courbe de vitesse radiale qui caractérise complètement la trajectoire de la planète. En effet, une orbite képlerienne est décrite par sept paramètres orbitaux (la période orbitale P , le demi-grand axe a, l’excentricité e, l’heure du périastre tp, l’argument du périastre ω, l’inclinaison i et la longitude du nœud ascendant Ω), dont les cinq premiers s’obtiennent à partir de la courbe de vitesse radiale. Pour déterminer la ou les périodicités les plus significatives dans les données, on utilise couramment le périodogramme généralisé de Lomb-Scargle (GLS, Zechmeister & Kürster 2009 après Scargle 1982). Une fois la période P estimée, les autres paramètres sont ensuite déter- minés à l’aide d’algorithmes d’optimisation, généralement la méthode de minimisation des moindres carrés. Les conditions initiales que nécessitent ces algorithmes proviennent de méthodes telles que celles données par Delisle et al.(2016). Comme toute technique d’observation, la méthode de la vitesse radiale présente des limitations. Une première limite fondamentale de la méthode, liée à sa précision, 0.1. Introduction 13 provient du bruit des photons. Ce bruit est inhérent à toutes les mesures de flux stellaire, car les photons sont discrets et leurs arrivées sur le détecteur sont indépendantes les unes des autres. Cela signifie que le nombre de photons peut être décrit par une statistique de Poisson,√ ce qui confère à un nombre N de photons détectés un écart-type intrinsèque de N. Bouchy et al.(2001) a étudié l’impact du bruit de photons sur les vitesses radiales. Le spectrographe lui-même est le second facteur naturellement limitant de la précision de vitesse radiale. Parmi l’ensemble des limitations instrumentales dont un aperçu complet peut être trouvé dans Fischer et al.(2016), nous pouvons citer : les stabilités thermique, mécanique et de pression; les effets de détecteur; les incertitudes des calibreurs de longueur d’onde; les fibres et leur brouillage. Finalement, les étoiles sont un troisième facteur limitant puisqu’elles ne sont pas des sources lumineuses sta- bles mais sont animées par de nombreux processus physiques capables d’induire des dé- calages Doppler. Les échelles de temps de ces processus varient de la minute à plusieurs années. De judicieuses stratégies d’observation permettent de gérer les oscillations stellaires de périodes courtes. Les tâches stellaires et les plages sont certainement les plus complexes à traiter, car elles induisent des signaux quasi périodiques qui ressemblent à ceux qu’émettrait une planète. Pour les identifier, il est d’usage d’avoir recours à une multitude d’indicateurs d’activité stellaire.

0.1.2 Recherche d’exoplanètes autour de naines M

Plusieurs programmes ont été partiellement ou totalement consacrés aux naines M. Leurs pétits masses inférieures comparées aux étoiles semblables au soleil permettent la détection d’exoplanètes plus petites. De même, les planètes de la taille de la Terre situées dans la zone habitable (HZ, la région entourant une étoile où une planète reçoit une irradiation suffisante pour pouvoir disposer d’eau liquide à sa surface, Kopparapu et al. 2013), parce que la HZ est plus proche de l’étoile. La recherche d’exoplanètes autour de M nains cherche à répondre à plusieurs questions scientifiques importantes. Premièrement, les statistiques des planètes en orbite autour de naines M restent plus incertains; les grandes enquêtes cherchent à mieux décrire l’occurrence et les propriétés générales de leurs planètes. Deuxièmement, les planètes potentiellement habitables sont plus facilement détectées autour de M nains. Troisième- ment, les planètes les plus proches du système solaire seront généralement celles qui se prêteront le mieux à la caractérisation en profondeur des installations futures. La grande majorité des étoiles voisines sont des nanies M, de sorte que les enquêtes en volume limité sur ces étoiles devraient fournir les meilleurs candidats pour une telle caractérisation. Plusieurs enquêtes RV dans le visible des nains cherchent à répondre à ces questions. Certaines des enquêtes principales sont les programme sur: HARPS (Bonfils et al., 2013; Delfosse et al., 2013; Astudillo-Defru et al., 2017b); SOPHIE Bouchy et al.(2009a); HARPS-N Affer et al.(2016); LCES/KECK-HIRES (Butler et al., 2017). Alors que ces enquêtes dans le visible ont produit (et continuent de produire) de bons résultats, elles ont été inévitablement limitées par la faible luminosité de ces étoiles dans le visible. C’est pour l’observation des naines M en particulier que des spectrographes prochaine 14 Résumé étendue en français infrarouge de haute précision ont été conçus; Un bref résumé des principaux instruments actuels et à venir est disponible dans Wright & Robertson(2017). Je souligne ici quelques-uns des principaux sondages actuels et à venir: CARMENES (?Quirrenbach et al., 2016); SPIRou (Artigau et al., 2014; Moutou et al., 2017; Fouqué et al., 2018); NIRPS (Bouchy et al., 2017).

0.2 Recherche en vitesse radiale des naines M avec le spectrographe SOPHIE

SOPHIE est un spectrographe d’échelle à dispersion croisée et alimenté par fibres, monté sur le télescope de 1.93 m de diametre de l’Observatoire d’Haute-Provence, opéra- tionnel depuis 2006 (Bouchy & Sophie Team 2006, Perruchot et al. 2008). En 2011, SOPHIE a été considérablement améliorée par l’installation de fibres octogonales et d’un double brouilleur (Perruchot et al., 2011), ce qui a permis d’améliorer la précision de vitesse radiale d’un facteur ∼ 6. Le spectrographe couvre la gamme de longueurs d’onde 387–694 nm sur 39 ordres spectraux. Deux modes de résolution sont disponibles: haute efficacité (HE, R ∼ 39 000) et haute résolution (HR, R ∼ 75 000). En plus de la fibre étoile (fibre A), SOPHIE possède une seconde fibre de calibration (fibre B) qui peut être utilisée soit pour la surveillance du ciel, soit pour la calibration simultanée. Pour la calibration simultanée, il y a deux calibreurs: une lampe à cathode creuse de ThAr (jusqu’au semestre 2017A), et un étalon Fabry-Pérot (à partir du semestre 2017B). Les spectres SOPHIE sont traités par un pipeline entièrement automatisé, connu sous le nom de système de réduction de données (DRS), adapté du pipeline HARPS (Bouchy & Sophie Team 2006, Bouchy et al. 2009a). La mesure des vitesses radiales s’effectue par la méthode CCF (fonction de corrélation croisée). Cette méthode est basée sur la corrélation croisée des spectres stellaires avec un masque binaire pondéré (voir, par exemple, Queloz 1995, Pepe et al. 2002a). Pour mon travail avec SOPHIE dans le visible, j’ai travaillé sur les données issues d’un programme de recherche des exoplanètes autour des naines M mené par le consortium exoplanètes SOPHIE depuis 2006, décrit en (Bouchy et al., 2009a). Ce programme a comme objectifs détecter les superTerres et Neptunes habitables; contraindre les statistiques des planètes autour de naines M; et trouver des compagnons potentiellement en transit. Traditionnellement, les vitesses radiales étaient généralement calculées par la méth- ode CCF; ceci est vrai pour SOPHIE. La méthode CCF est bien adaptée aux étoiles de type G, avec un continuum bien défini et des raies spectrales claires. Les naines M, en revanche, ont des spectres complexes avec une multitude de petites lignes et de bandes moléculaires. Cela signifie que la méthode CCF, qui ne sélectionne que des lignes spé- cifiques dans son masque binaire, sous-utilise les informations Doppler présentes. La correspondance de modèles ("template-matching") avec un spectre stellaire réel permet une détermination plus précise de la vitesse radiale (Anglada-Escudé & Butler 2012, 0.2. Recherche en vitesse radiale des naines M avec le spectrographe SOPHIE 15

Astudillo-Defru et al. 2015, etc.). De plus, il permettre de mieux gérer toute contamination tellurique. J’ai adapté un algorithme développé par N. Astudillo-Defru ( citealt Astudillo15, citealt Astudillo17b) aux spectres SOPHIE des naines M, et l’ai appliqué à toutes les cibles du programme avec au moins dix observations HR depuis 2011 (56 cibles). J’ai aussi étudié les effets instrumentaux à différents échelles de temps (inefficacité de trans- fert de charges sur chaque pose, dérive nocturne du spectrographe, et dérive à long terme). J’ai utilisé la plate-forme web DACE (Data Analysis Center for Exoplanets) a pour analyser les vitesses radiales calculées. Grâce à cette ré-analyse, nous avons pu détecter et publier quatre planètes autour de naines M: la détection de Gl 96 b et la confirmation indépendant de Gl 617A b, deux planètes de la taille de Neptune proche à leurs zones habitables (Hobson et al., 2018a); la détection du Gl 411 b, une superTerre tempéré orbitant une des étoiles les plus proches au Système Solaire (Díaz et al., 2019); et la détection du Gl 378 b, une Neptune tiède proche au limite du désert de Neptunes chauds (Hobson et al., 2019). Il y a une autre planète à publier prochainement (Delfosse en prep), et deux candidats qui sont suivies en priorité. On a aussi confirmé deux planètes publiés par d’autres groupes, et réfuté une troisième. J’ai également étudié les effets de l’activité stellaire sur les spectres SOPHIE. J’ai 0 obtenu des différentes indices d’activité: Hα, log(RHK), Na i et He i, ainsi que l’étendue de la bissectrice du CCF. Les trois premiers indices retracent l’activité chromosphérique dans les régions supérieure, inférieure et moyenne à inférieure de la chromosphère, re- spectivement (Gomes da Silva et al., 2011), tandis que la bissectrice CCF suit l’activité photosphérique. La DRS SOPHIE fournit automatiquement la bissectrice CCF et des 0 codes sont disponibles pour calculer les indices Hα et log(RHK). J’ai modifié le code qui 0 calcule le log(RHK) pour utiliser le nouvel étalonnage de Astudillo-Defru et al.(2017a), et implémenté le calcul de Na I et He i. Lors de l’analyse des séries chronologiques de RV pour chaque cible, j’ai également inspecté les indicateurs d’activité pour déterminer si les périodicités observées pouvaient être causées par une activité stellaire. 0 Avec cette analyse, j’ai trouvé que les indices Hα et log(RHK) sont les indicateurs les plus efficaces de l’activité stellaire dans les spectres des naines M obtenues avec SOPHIE. Il semble claire aussi que les vitesses radiales obtenues avec le "template- matching" sont plus sensibles à l’activité stellaire que celles issues de l’algorithme traditionnel du CCF. Cela souligne l’importance d’avoir indicateurs d’activité stellaire ro- bustes et fiables, par une coté; et par l’autre, qu’on doit avoir précaution avec les algorithmes de "template-matching", et vérifier soigneusement l’impacte de l’activité stellaire. En dehors de mon travail sur les naines M, j’ai aussi fait d’autres contributions au travail du consortium exoplanètes. Pour le programme de recherche des superTerres autour d’étoiles comme le soleil, j’ai fait des mises à jour périodiques de la base de données. Cela impliquait de générer la correction des dérives á long terme de l’instrument adapté pour les étoiles de type solaire, l’appliquer aux données de vitesse radiale, et générer une

a. Disponible sur https://dace.unige.ch 16 Résumé étendue en français document avec une sommaire de tous les observations. Pour ce document, j’analysais le periodogramme pour chaque étoile pour chercher des signaux signiﬁcatifs. Par autre côté, j’ai trouvé un défaut en la manière où la correction du fond était réalisé pour des observations avec du Fabry-Pérot simultanée sur la ﬁbre B. J’ai crée et implémenté une correction améliorée. Ce travail est présenté dans le chapitre2.

0.3 Développement du spectropolarimetre SPIRou

SPIRou (SpectroPolarimetre InfraRouge) est un spectroplarimètre proche infrarouge, monté sur le télescope Canada France Hawaii (CFHT) de 3.6 m, en opération depuis février 2019. SPIRou a été développé par un grand consortium international, dirigé par la France et impliquant plusieurs instituts de sept pays (France, États-Unis, Canada, Suisse, Brésil, Taïwan et Portugal). Il couvre la plage spectrale 0.98–2.35 (bandes Y, J, H et K) à une résolution de R ∼ 75 000, fournissant ainsi le spectre complet de l’étoile dans le proche infrarouge, et peut être utilisé dans polarisation circulaire ou linéaire. La conception de l’instrument est décrite dans Artigau et al.(2014). Je souligne ici l’unité de calibration, décrite dans Boisse et al.(2016). L’unité de calibration a été construite en collaboration par l’Observatoire d’Haute-Provence et l’Observatoire de Genève, ce dernier fournissant l’étalon de Fabry-Pérot constituant le module de référence pour les vitesses radiales. Mon travail avec SPIRou étant axé sur les calibrations, j’ai donc pu collaborer avec les ingénieurs et techniciens de l’Observatoire d’Haute-Provence, notamment dans le cadre des tests de validation et de la mise en service. Le premier élément de l’unité est le module de sources de lumière, avec des emplacements pour les deux lampes à cathodes creuses, des injections de fibres pour la source de lumière blanche et le module FP, et un emplacement réservé pour des futures modifications. Il y a aussi une position de stationnement (pas d’éclairage). L’unité d’étalonnage possède deux liaisons de fibres optiques avec le spectrographe. L’un est injecté dans l’unité Cassegrain qui lie le télescope au spectrographe, de sorte que la lu- mière suive le même chemin que les fibres scientifiques. L’autre va directement au spec- tropgraphe (après avoir traversé le module de balance de flux) et sert pour la mesure simultanée de la dérive. L’éclairage de chacune de ces fibres est sélectionné par des chariots devant les lampes. Ma principale implication avec SPIRou a eu lieu dans le cadre du développement du système de réduction des données (DRS). La DRS du SPIRou est basé à l’origine sur les pipelines développés pour HARPS et SOPHIE (Pepe et al. 2004, Bouchy et al. 2009a); le développement a été dirigé par Marseille (LAM, I.Boisse et moi-même), Montréal (iREx, E. Artigau et N. Cook) et Genève (Observatoire de Genève, F. Bouchy). Le DRS est écrit en Python 3 (rétrocompatible avec Python 2.7), mis à jour et contrôlé par version sur Github. Le but de la DRS est d’effectuer la réduction complète des données spectrales, des images brutes aux vitesses radiales et aux produits polarimétriques. La première étape de la réduction est le pré-traitement; Il s’agit d’une étape entièrement nouvelle 0.3. Développement du spectropolarimetre SPIRou 17 qui n’existait pas dans les pipelines HARPS / SOPHIE, mais qui est nécessaire pour gérer correctement les images du nouvelle détecteur H4RG. Il s’applique à tous les fichiers et est utilisé pour supprimer certains effets de détecteur tels que la diaphonie de l’amplificateur (routine cal_preprocess_spirou.py). Ensuite, les calibrations sont réduits en générant: une carte du courant d’obscurité (cal_DARK_spirou.py); une carte des pixels défectueux, comprenant deux petits trous et une égratignure sur le détecteur (cal_BADPIX_spirou.py); la localisation des ordres pour les fibres A, B et C (cal_loc_RAW_spirou.py); une carte du profil de la fente à travers le détecteur; les profils de flux pour toutes les fibres extraites (cal_FF_RAW_spirou.py); et une carte de longueur d’onde pour toutes les fibres (extraites) (cal_HC_E2DS_EA_spirou.py, cal_WAVE_E2DS_EA_spirou.py). Enfin, les spectres stellaires sont extraits et corrigés de la contamination tellurique, puis les vitesses radiales (via la méthode CCF) et / ou les produits polarimétriques sont générés. Avant d’être expédié à Hawaï et monté sur le CFHT, SPIRou a d’abord été assemblé et testé à l’IRAP de Toulouse. J’ai assisté aux tests de validation des performances de l’unité de calibration, décrits dans Perruchot et al.(2018). En particulier, j’ai développé des routines de calibration de longueur d’onde, en utilisant soit la lampe à cathode creuse seule, soit une combinaison de la lampe à cathode creuse et de l’étalon Fabry- Pérot. J’ai également crée des outils de visualisation et de mesure rapides permettant d’afficher une image brute et un profil de section transversale le long d’une ligne ou d’une colonne, notamment pour l’estimation du niveau de flux dans un ordre particulier (pour une lampe par exemple). Après validation à Toulouse, SPIRou a été expédié au CFHT et des tests de mise en service ont été entrepris. Je me suis principalement concentré sur la correction et l’amélioration des routines de longueur d’onde, mais j’ai également aidé à résoudre divers problèmes et bugs signalés à l’équipe DRS. L’unité d’étalonnage SPIRou possède deux logements pour lampes à cathode creuse (HC); ceux actuellement installés sont une lampe UNe et une lampe ThAr. Ces lampes fournissent un étalonnage absolu de la longueur d’onde, offrant des raies spectrales (dont les longueurs d’onde ont été cataloguées et sont donc connues en principe) sur l’ensemble du détecteur. Lors des tests de validation et de mise en service de SPIRou, il est apparu que les lampes HC ne fournissaient pas à elles seules une solution de longueur d’onde suffisamment précise et stable. La prochaine étape consistait donc à combiner les lampes HC avec le spectre d’étalon de Fabry-Pérot (FP). Le FP fournit une multitude de lignes sur l’ensemble du détecteur, dont l’espacement est connu a priori puisqu’il est donné par l’équation de FP (il y a aussi une dépendance sur la longueur d’onde, mais il est mesurable). Cependant, leurs longueurs d’onde absolues ne sont pas connues mais doivent être déterminées à partir d’une autre source. Par conséquent, en ancrant les longueurs d’onde de la ligne FP aux lignes HC, nous pouvons obtenir une solution plus précise de la longueur d’onde. J’ai mis en œuvre et testé différentes méthodes permettant de générer une correspondance longueur d’onde-pixel, en utilisant soit des lampes à cathode creuse seules, soit des lampes à cathode creuse combinées et un étalon de Fabry-Pérot. Les lampes HC ne fournissaient pas à elles seules une précision suffisante, car elles étaient au niveau du ∼2 m s−1. Les solutions combinées HC-FP, en revanche, ont un excellent RMS interne de ∼0.15 m s−1. La stabilité nuit à nuit est compliquée par la dépendance de la solution 18 Résumé étendue en français de longueur d’onde sur les calibrations précédents, en particulier sur la détermination de la fente. La fixation de tous les calibrations produit un décalage RV visible mais constant entre les solutions; lorsque ce décalage est supprimé, les variations de nuit à nuit s’améliorent. La description complète de ce travail se trouve dans le chapitre3.

0.4 Proprietés des planètes autours des naines M

Enfin, j’ai étudié la population actuelle de planètes trouvées autour des naines M dans une perspective globale. Le nombre d’exoplanètes connues autour de naines M a considérablement augmenté pendant le développement de cette thèse, passant de 118 planètes détectées en 2016 à 183 connues aujourd’hui - une augmentation de 55% sur trois ans. J’ai pris le catalogue de l’Explanet Encyclopedia, avec les critères de sélection suivants: Pour l’étoile hôte, masse stellaire dans la gamme 0, 06M ≤ M∗ ≤ 0, 6M et type spectral sur Simbad ou Vizier correspondant à une naine M de séquence principale; pour la planète, masse planétaire Mp < 13MJ . La plupart des planètes orbitent autour d’étoiles M précoces à moyens. Ceci est co- hérent avec les stratégies de sélection des cibles adoptées par les enquêtes de vitesse radiale (VR); même si l’échantillon n’exclut pas spécifiquement les naines M tardives, elles sont principalement supprimés par des coupures de magnitude car elles sont intrinsèque- ment très faibles dans le visible. Néanmoins, récemment des exoplanètes autour d’étoiles M tardives ont été découvertes, et on peut s’attendre à ce que leur nombre augmente avec l’augmentation du nombre d’enquêtes dans le prochaine infra-rouge. Les planètes en orbite autour de naines M brillantes devraient être les meilleurs candidats pour une caractérisation en profondeur dans un proche avenir. Alors que des planètes ont été détectées sur de grandes distances (notamment par la méthode de la microlentille), la majorité tourne autour d’étoiles relativement proches. Les planètes des zones habitables, en particulier, sont toutes a moins de 11 pc et la plupart sont hébergées par des étoiles brillantes, ce qui en fait d’excellentes cibles pour les efforts de caractérisation atmosphérique et par imagerie directe. Quant aux types de planètes trouvés, la plupart ont une masse relativement faible et une période courte (1 à 10 M⊕, 1 à 200 jours). On constate clairement un dé- ficit de Neptunes et Jupiters chauds. Vu que ces planètes sont les plus simples à dé- tecter, grâce à son grande taille et courte période orbitale, cette manque n’est pas un effet de sélection observationnel sinon une vrai caractéristique de la population. Les planètes proches de la limite inférieure du chaud désert de Neptune ont tendance à être hébergées par des étoiles plus riches en métaux, et a être seules dans ses systèmes planétaires. Actuellement, seul une petite partie des exoplanètes découvertes ont leur rayon et leur masse mesurés. La poignée de planètes dont les masses et les rayons sont suffisamment précis pour permettre des études de composition s’étalent sur une vaste gamme de compositions, allant d’une composition proche à celle de Mercure (grand noyau ferreux), 0.5. Conclusions 19 passant par des compositions comme celle de la Terre (petit noyau, grand manteau rocheuse), ou 50% manteau et 50% océan, jusqu’à 100% d’eau. J’ai estimé les limites de zones d’habitabilité conservateurs et optimistes, comme définis par Kopparapu et al.(2013). J’ai trouvé quatorze planètes dans leurs respectives zones d’habitabilité conservatives, plus huit dans les zones optimistes. Ces planètes sont distribués dans quinze systèmes. L’habitabilité réelle des planètes en orbite autour des naines M a été vivement débattue (voir e.g. Tarter et al. 2007, Shields et al. 2016). En raison de leurs courtes périodes orbitales, beaucoup de ces planètes sont attendues être en rotation synchrone avec leur étoile. Cela conduirait à des côtés permanents de jour et de nuit, dont les températures extrême pourraient nuire à l’habitabilité. Les naines M émettent aussi beaucoup de lumière dans les EUV et les rayons X, en particulier pendant leur jeunesse. Ce rayonnement pourrait dépouiller une planète de beaucoup de son atmosphère (e.g., King et al. 2018), bien qu’une magnétosphère planétaire peut fournir une protection. La réponse à savoir si une planète spécifique est vraiment habitable nécessite donc une étude minutieuse et une modélisation de l’étoile hôte et de la planète lui-même; les limites de la HZ ne constituent qu’une première approximation. Les métallicités stellaires des naines M sont difficiles à mesurer du à leur spectre complexe. Celles du catalogue proviennent de sources variées; toute tendance doit donc être considérée avec prudence. Cependant, il semble exister une corrélation moyen- nement significative entre les hôtes riches en métaux et une masse planétaire supérieure. Cela correspond bien aux découvertes précédentes (par exemple, Hobson et al. 2018b, Courcol et al. 2016, Sousa et al. 2019). 87 des 183 planètes connues à ce jour font partie de systèmes multiplanétaires, et des statistiques préliminaires indiquent que la plupart des systèmes devraient être multiples. Avec une surveillance continue et des instruments plus sensibles, nous pouvons nous attendre à que ce nombre augmente à mesure que des planètes auparavant inac- cessibles deviennent détectables. L’analyse détaillée est dans le chapitre4.

0.5 Conclusions

Aa cours de cette thèse, j’ai étudié la détection des exoplanètes autour de naines M par vitesse radiale. Mes contributions sont centrés sur l’amélioration de la précision des vitesses radiales. Cette précision est indispensable surtout pour la détection des planètes de petites masses, mais aussi pour la meilleure caractérisation de tout l’ensemble de la population planétaire. En particulier, je me suis focalisée sur deux aspects: la calibration en longueur d’onde, et le calcul des vitesses radiales. Pour la calibration en longueur d’onde, j’ai développé des nouvelles méthodes pour combiner les lampes à cathodes creuses avec les étalons Fabry-Pérot, ce qui permet d’avoir une calibration à très haute précision. Cet 20 Résumé étendue en français calibration précise est indispensable pour la mesure des vitesses radiales. Pour le calcul des vitesses radiales, j’ai adapté et testé un nouvel algorithme de comparaison des modèles pour un grand échantillon de naines M, montrant par un coté qu’il permet de detecter des signaux planétaires à plus haute puissance; et par l’autre qu’il est plus sen- sible à l’activité stellaire. Il est donc important de vériﬁer s’il y a une possible impact de l’activité stellaire pour tout planète issu des vitesses radiales obtenus avec cette méth- ode. Néanmoins, il reste une méthode très prometteuse pour la recherche d’exoplanètes autour des naines M. Introduction

Contents

1.1 A brief history of exoplanets...... 22 1.2 Characteristics of M-dwarf stars...... 24 1.3 The radial velocity method ...... 28 1.3.1 Limitations of the RV method...... 30 1.3.2 Visible and nIR spectrographs and surveys ...... 32 1.4 M-dwarf stars and their planets (as of 2016)...... 34 1.5 Thesis objectives ...... 39 22 Introduction

1.1 A brief history of exoplanets

Since antiquity, humans have dreamed of finding other worlds like our own. Exo- planet detections have been reported since the 1800s (Jacob, 1855; See, 1896). How- ever, most of these early findings turned out to be mistaken (though a few were later confirmed, e.g. Campbell et al. 1988 by Walker 2012). The first true discovery was in 1992: Wolszczan & Frail(1992), based on minute variations in the timing of the radio emissions of pulsar PSR B1257+12, demonstrated the existence of two planet-sized bodies orbiting it, and indications of a third. Pulsars, however, are very different objects to the Sun, being rotating neutron stars with strong magnetic fields. This makes the planetary-sized bodies found around them hard to compare with the planets of our own Solar System. Three years later, Mayor & Queloz(1995) made history with the detection of the first planet around a main-sequence star, 51 Pegasi, using the ELODIE spectrograph at the Observatoire de Haute-Provence (OHP). The planet, named 51 Peg b, is a 0.47 MJ minimum-mass planet at 4.23 d orbital period, which proved to be the first of a class of planets that would become known as hot Jupiters. These are Jupiter-size planets orbiting at short distances from their host stars (a . 0.1 AU, P . 10 d), which have no analogues in our Solar System. In the twenty-four years since the discovery of 51 Peg b, exoplanet detections have mounted exponentially, to a total of over 4000 planets known today. These planets have been detected by a variety of techniques, of which the most productive by far have been the transit and radial velocity methods (Fig. 1.1). The first consists in the detection, via photometric monitoring of a star, of periodic dips in its light curve due to an orbiting planet that passes between the star and the observer. While productive from the ground, its true potential is best exemplified by space missions such as Kepler (Borucki et al., 2010) which have produced thousands of planet candidates. The second exploits the Doppler shift induced in the spectra of a planet-hosting star by its motion around the star-planet centre of gravity. Large dedicated surveys have provided hundreds of planets, while follow-up spectroscopic observations of transiting candidates allow us to confirm or refute their planetary nature. This thesis is based on the radial velocity method, which is therefore described in more detail in Section 1.3. The first exoplanets found tended to be large, close-in planets around bright, sun- like stars (e.g. Butler et al. 1997, Fischer et al. 1999). There are several reasons for this. First, the strength of a planet’s signal (e.g. the brightness diminution or Doppler shift it induces) is proportional to its size, and for radial velocities inversely proportional to its period. Next, faint stars are costly to observe since they require longer exposure times, and tiny planetary signals become harder to distinguish from the increased photon noise. Lastly, sun-like stars have naturally been prioritised in the search for Solar System analogues. As instrumentation became more sensitive and the temporal baseline of surveys increased, however, the detection possibilities widened, and a wide variety of planetary populations began to emerge (Fig. 1.2). In addition to hot and warm Jupiters, a population of long-period giant planets has also been found (primarily by radial velocities, microlensing, and direct imaging). Lower mass planets, Earth- to Neptune-sized, are likewise detected nowadays (mostly by radial velocities and transits). True gaps in the distribution also became apparent, such as the hot Neptune desert (e.g. Lecavelier Des 1.1. A brief history of exoplanets 23

Etangs 2007, Beaugé & Nesvorný 2013, Mazeh et al. 2016), and the Fulton gap between super-Earths and mini-Neptunes (Fulton et al., 2017). A broader range of stellar hosts likewise came into evidence. The speciﬁc case of M-dwarf stars, which this thesis focuses on, is detailed in Section 1.4. Today, exoplanetary science continues to be a growing ﬁeld. Simultaneous mass and radius measurements have led to the development of internal structure models (e.g. Baraffe et al. 2014, Brugger et al. 2017). Exoplanet atmospheres begin to be accessible to observation and modelling (e.g. Sing et al. 2016, Pino et al. 2018). New instruments, using novel methods to achieve ever-higher sensitivity, hunt for Earth analogues - that is, Earth-sized planets in the habitable zones of their host stars.

Figure 1.1– Histogram of exoplanets by detection type (according to "The Exoplanet Encyclopaedia", data obtained 19th April 2019). Transit and radial velocity detections account for over 90% of the total. 24 Introduction

Figure 1.2– mass-period diagram for all known exoplanets with these parameters reported in "The Exoplanet Encyclopaedia" (data obtained 19th April 2019). Msin(i) lower limits are used as a mass proxy when the true mass is not known for RV detections. Masses derived from mass-radius relationships for transiting planets are also included. The dashed line at 13 MJ marks the IAU’s proposed planet-brown dwarf boundary. The points are color-coded by stellar mass; planets without a reported stellar mass are black. Some outliers stand out: the handful of ultra-short-period giants at ∼ 10 MJ, ∼0.05 d are mostly hosted by pulsars, and may actually be eroded (sub)stellar companions; the −5 ultra-low-mass planets at ∼ 7 × 10 MJ are KIC 12557548 b, a transiting disintegrating super-Mercury at 0.5 d period (Rappaport et al., 2012), and PSR 1257 12 b, the ﬁrst planet around a pulsar, at 25.3 d period (Wolszczan & Frail, 1992).

1.2 Characteristics of M-dwarf stars

As this thesis is focused on planets orbiting M dwarfs, I describe here the main characteristics of these stars. M dwarfs constitute the smallest and coolest stars on the stellar main sequence. They take their name from their spectral classiﬁcation under the Harvard system (Cannon, 1915), where the M class corresponds to spectra with strong metal lines and molecular bands. The class is further divided into ten subclasses, whose main properties are summarised in Table 1.1. M dwarfs span a very wide range of masses and temperatures, with a difference of 0.54 M and 1500 K between M0 and M9 stars. For comparison, G-type stars range from 0.84 − 1.15 M , and K dwarfs range from 0.5 − 0.8 M . This wide range means that early and late M dwarfs have widely different characteristics. In particular, while early M dwarfs have a similar internal structure to the Sun, with a radiative core and a convective outer layer, M dwarfs with masses below 0.32−0.37 M (i.e., of spectral types M4 and later) are fully convective (Mann et al., 2015). This means 1.2. Characteristics of M-dwarf stars 25

Table 1.1– Characteristics of M dwarf spectral classes, adapted from Kaltenegger & Traub(2009). The columns a(HZ) and P(HZ) indicate the orbital semi-major axis and period of a planet in the middle of the habitable zone.

Spectral Teﬀ [K] Radius [R ] Mass [M ] L/100 [L ] a(HZ) [AU] P(HZ) [d] type M0 3800 0.62 0.60 7.2 0.268 66 M1 3600 0.49 0.49 3.5 0.190 43 M2 3400 0.44 0.44 2.3 0.152 33 M3 3250 0.39 0.36 1.5 0.123 26 M4 3100 0.26 0.20 0.55 0.075 17 M5 2800 0.20 0.14 0.22 0.047 10 M6 2600 0.15 0.10 0.09 0.030 6 M7 2500 0.12 ∼0.09 0.05 0.022 4 M8 2400 0.11 ∼0.08 0.03 0.019 3 M9 2300 0.08 ∼0.075 0.015 0.013 2

that He does not gather in the core but is mixed throughout the star, allowing them to consume far more of their nuclear fuel than G-type stars, and thus leading to extended lifetimes of the order of ∼ 1013 years (longer than the age of the universe, Laughlin et al. 1997). The convective boundary has been a particular target of magnetic field studies, as it was expected that partially and fully convective stars would have different dynamo mechanisms at play. Observational evidence, however, has shown a wide variety of magnetic topologies, with no clear difference across the convection boundary (e.g. Morin et al., 2010; Shulyak et al., 2014; Moutou et al., 2017). Stellar activity in M dwarfs is interesting both in itself, as a tracer of, for example, the stellar rotation or magnetic dynamo (e.g. Newton et al., 2017; Astudillo-Defru et al., 2017a), and also due to its strong impact on the detection of exoplanets (e.g. Moutou et al. 2017 and references therein). In particular, starspots, which are generated by magnetic activity, can mimic an exoplanet. Estimates of the surface covered by spots vary widely, from . 2% (Hébrard et al., 2016) to ∼ 40% (Jackson & Jeffries, 2013). Many methods have been developed to quantify the stellar activity of these stars, which are discussed in Sect. 2.5.3. Another notable characteristic of M dwarfs is the richness and complexity of their spectra. Figure 1.3 shows spectra for G, K, and M stars. While the G spectrum has clear, sharply defined lines, the M dwarfs’ spectra show a multitude of bands and a poorly defined continuum. This means that spectroscopic stellar parameters, whose determination relies on precise measurements of spectral lines, are hard to measure (e.g. Delfosse et al. 2000; Neves et al. 2014; Mann et al. 2015, 2019). In particular, metallicities are extremely complex to determine. Nevertheless, progress has recently been made using moderate resolution nIR spectra, where sharp metallicity-dependent lines have been identified. Their dependence on metallicity is calibrated via G-M binaries, for which the metallicities of both components are assumed to be identical, and the metallicity of the G star is well determined (Rojas-Ayala et al., 2010, 2012; Terrien et al., 2012; Newton et al., 2015; Hobson et al., 2018b). An alternative method based on high-resolution spec- 26 Introduction tra in the visible was developed by Neves et al.(2014); it uses most lines and features in the spectra, and was calibrated through FGK-M binaries.

Figure 1.3– Comparison of G, K and M stellar spectra from the atlas of Pickles(1998). The spectra have been arbitrarily shifted in intensity for clarity.

M-dwarf stars emit far less radiation than stars like the Sun, with luminosities ranging from 0.015% to 7.2% L (Table 1.1). This means that the Habitable Zone (HZ), the region around a star where a planet receives sufficient irradiation to be able to have liquid water on its surface, as defined by e.g. Kopparapu et al. 2013, will be much closer to the star than for Sun-like stars. The evolution of the HZ distance across spectral types is shown in Fig. 1.4. However, the actual habitability of planets orbiting M dwarfs has been greatly debated (see e.g. Tarter et al. 2007, Shields et al. 2016 for reviews). Due to their short orbital periods, many of these planets (particularly those orbiting the lowest-mass stars) are expected to be tidal-locked. This would lead to permanent day and night sides, whose extreme temperature might impede habitability (although atmospheric circulation may mitigate this). M dwarfs also emit strongly in the EUV and X-Ray, particularly during their youth, which is long compared to G-type stars. This radiation could strip a planet of significant proportions of its atmosphere (e.g. King et al. 2018), though a planetary magnetosphere may provide protection. 1.2. Characteristics of M-dwarf stars 27

(a) Plot of the Habitable Zone range for diﬀerent stellar spectral types, from F through M. Credit: PHL@UPR Arecibo.

(b) Plot of the Habitable Zone range for M dwarfs, using the limits of Kopparapu et al.(2013) and the stellar parameters from Table 1.1.

Figure 1.4– Plots of the Habitable Zone range for diﬀerent stellar spectral types. 28 Introduction

1.3 The radial velocity method

While we generally speak of a planet as orbiting its host star, this is a simpliﬁcation; both planet and star actually orbit the barycentre of the star-planet system. Therefore, when we observe a star with a planet, it will periodically move towards and away from us (provided the orbit is not in the plane of the sky). This movement will induce a Doppler shift of the spectra towards the blue when the star approaches, and towards the red when it recedes (Fig. 1.5). These spectral shifts can be exploited to detect planets, in what is known as the radial velocity or Doppler spectroscopy method. For a star moving with velocity v at an angle θ to the observer’s line of sight, the shift in wavelength ∆λ will be given by the relativistic Doppler equation (adapted from Perryman 2018):

! 1 + β cos(θ) ∆λ = λob − λem = λem p − 1 (1.2) 1 − β2 with β = v/c.

Figure 1.5– Illustration of the radial velocity method. When the star recedes from the observer, the radial velocity is positive and the spectral lines are red-shifted (left panel); when the star and planet are aligned, the line-of-sight velocity is zero (center panel); when the star approaches the observer, the radial velocity is negative and the spectral lines are blue-shifted (right panel). 1.3. The radial velocity method 29

For v c and θ π/2, the radial velocity (i.e. line-of-sight velocity) reduces to: ∆λ vr = v cos(θ) ≈ c (1.3) λem

By measuring the spectral shifts and translating them to radial velocity through equation 1.3, we can construct a radial velocity curve. From this curve, we seek to characterise the planet. A Keplerian orbit is described by seven orbital parameters: the orbital period P , orbital distance a, eccentricity e, time of pericentre tp, argument of pericentre ω, inclination i, and longitude of ascending node Ω (Fig. 1.6). The ﬁrst ﬁve of these can be obtained from the radial velocity curve. The radial velocity of the star is given by (formulation from Perryman 2018):

vr = K[cos (ω + ν) + e cos ω] (1.4)

with ν the true anomaly (angle between the direction of pericentre and the position of the body) and K the semi-amplitude, given in turn by

2π a? sin i K = √ (1.5) P 1 − e2

3 Using Kepler’s third law, a /P 2 = G(M? + Mp)/4π2, K can be re-written as

2πG1/3 M sin i 1 K p √ = 2/3 (1.6) P (M? + Mp) 1 − e2

Figure 1.6– Keplerian orbital elements. Credit: Lasunncty at the English Wikipedia, licensed under CC BY-SA 3.0. 30 Introduction

From equations 1.4 to 1.6, it is clear that the shape of the curve will be deﬁned by e and ω, as shown in Fig. 1.7, and its amplitude by e, P , and Mp sin i (with this last linked to a and P through Kepler’s third law). As the inclination cannot be determined from the radial velocity curve, only a minimum mass Mp sin i for the planet can be established.

Figure 1.7– RV curve dependency on e and ω: simulated radial velocity curves with, from left to right, e = 0 and ω = 0; e = 0.5 and ω = 0; e = 0.5 and ω = 90◦. Created with the NAAP radial velocity simulator b.

To determine the most significant periodicity(ies) in the data, the generalised Lomb- Scargle periodogram (GLS, Zechmeister & Kürster 2009 following Scargle 1982) is com- monly used. With the period P to be fitted estimated, the best-fit parameters are generally determined through numerical least-squares minimisation algorithms, which provide estimates of both the parameters and their errors. First-guess parameters to ini- tialise the minimisation algorithms can be obtained from analytical formulations, such as that given by e.g. Delisle et al.(2016).

1.3.1 Limitations of the RV method

As with any observational technique, the radial velocity method for detecting exoplanets naturally possesses limitations. On the one hand, the precision of the individual data points will be determined by a combination of multiple factors such the photon noise and the instrumental characteristics. On the other, not all radial velocity variations - nor even all apparently periodic radial velocity variations - are induced by orbiting planets; the intrinsic variability of the star also needs to be considered. A ﬁrst, and fundamental, limit to radial velocity precision is set by the photon noise. Photon noise is inherent to all measurements of stellar ﬂux, since the photons are dis- crete and their arrival at the detector is independent of each other. This means that the photon count can be described by Poisson√ statistics, and therefore a count of N photons has an inherent standard deviation of N. Bouchy et al.(2001) studied the impact of photon noise on radial velocities. They found that the uncertainty in velocity change is given by:

c δVRMS = √ (1.7) Q Ne−

b. Available at https://astro.unl.edu/classaction/animations/extrasolarplanets/ radialvelocitysimulator.html 1.3. The radial velocity method 31

where c is the speed of light, Q is a quality factor given by the stellar spectral profile within the wavelength range (dependent on spectral type, stellar v sin i, and spectral resolution), and Ne− is the total number of counted photoelectrons over the spectral range (dependent on the telescope area, the total efficiency, the exposure time, and the stellar magnitude). As an example, they show that CORALIE can reach δVRMS = −1 −1 1 m s on a 6 magnitude star with Teff = 4500 K, v sin i = 0 km s in an 8 minute −1 exposure. For the same star, they estimate HARPS would reach δVRMS <0.1 m s in only 4 minutes. The spectrograph itself will also limit the RV precision. A comprehensive overview of the instrumental limitations can be found in Fischer et al.(2016). Some of the main points are: • Thermal, mechanical, and pressure stability: stabilised spectrographs are crucial for high-precision RVs. For example, Pepe et al.(2002b) note that a 1 mbar change in pressure would induce drifts of the order of 100 m s−1 on HARPS. • Detector effects: CCD imperfections have been found to create noticeable RV effects. To cite two notable examples, the HARPS CCD stitching was found to in- troduce spurious signals of a few m s−1 (Dumusque et al., 2015); and the SOPHIE CCD’s charge transfer inefficiency was shown to induce S/N-dependent RV changes (Bouchy et al., 2009b). These effects are generally measurable and can be modelled out. • Wavelength calibrators: Both thorium-argon lamps and iodine cells, the most common calibrators, are limited to a precision of ∼1 m s−1, according to Fischer et al. (2016). New calibrators such as laser-frequency combs and Fabry-Pérot étalons should improve on this precision significantly. • Fibres and scrambling: optical fibres are used to couple light from the telescope to the (environmentally stabilised) spectrograph in a stable way. Multi-mode fibres are generally used, for which the modes must be scrambled, as otherwise changes in the input illumination affect the output density distribution. Circular fibres have good azimuthal but poor radial scrambling; the incorporation of non-circular fibre sections has been shown to efficiently scramble the modes (e.g. Chazelas et al., 2010). Interference between the modes also generates modal noise, which is higher for few-mode fibres; it can be handled by fibre agitation (e.g. Petersburg et al., 2018) or AO tip-tilt strategies (Blind et al., 2017).

Finally, stars are not quiescent light sources, but display a wide range of physical processes that can induce Doppler shifts. The time scales of these processes vary greatly: from oscillations over a few minutes; granulation changes over minutes to days; surface phenomena such as spots and plages which move with the stellar rotation (typically days to months); through to year-to-decade long magnetic cycles (see, e.g., Saar & Donahue 1997, Gomes da Silva et al. 2011, Dumusque et al. 2015, and references therein). The shortest-period variations can be handled through observing strategies, such as adopting sufficiently long exposure times to average them out (e.g. Mayor et al. 2003). Granula- tion features are also expected to have much lower contrast and shorter timescales for M dwarfs than for sun-like stars (e.g. Ludwig et al., 2002), making them easier to average out. Perhaps the most complex to deal with, in terms of distinguishing stellar-induced 32 Introduction from planetary signals, are magnetic surface features such as spots and plages. These induce quasi-periodic signals at around the stellar rotation period, which generally falls within the period range in which most planets are detected. If the features are persistent over time, they will generate significant peaks in the GLS periodogram that can be hard to separate from those due to true planets. A variety of activity indicators have been developed to help weed out these false positives. As part of my thesis work, I studied these indicators in visible spectra (see Sect. 2.5.3). Non-parametric models using Gaussian processes to describe and fit out the stellar activity have also recently been developed (e.g. Haywood et al. 2014).

1.3.2 Visible and nIR spectrographs and surveys

It was through spectroscopy in the visible that the ﬁrst exoplanet around a sun-like star was detected, with the ELODIE spectrograph at the OHP (described in Baranne et al., 1996). Since then, more than seven hundred exoplanets have been found or conﬁrmed via this technique, by a multitude of instruments. An overview of the principal high-precision Doppler instruments in the visible (as they stood at the start of this thesis), and their associated main surveys, is given in Fischer et al.(2016). These surveys use high-resolution spectroscopy, ranging from 45 000 (UCLES) to 115 000 (HARPS, HARPS-N); are wavelength calibrated using either ThAr hollow-cathode lamps or iodine cells; and reach measurement precisions of 0.8 m s−1 (HARPS, HARPS-N) to 5 m s−1 (Tull). Within these surveys, several programmes have been focused partly or wholly on M dwarfs, which are particularly interesting targets for several reasons. Firstly, their lower masses compared to sun-like stars enable the detection of smaller exoplanets (since the RV semi-amplitude K is inversely proportional to the stellar mass, as seen in Eq. 1.6). Likewise, earth-size planets in the habitable zone are easier to detect, because the HZ is closer to the star than for sun-like stars (Fig. 1.4). As an example, a 5M⊕ planet in the HZ of a G2 star will orbit at 1 AU with a 1 yr period and produce only a K ≈ 0.5 m s−1 signal. The same planet in the HZ of an M2 dwarf will orbit at 0.25 AU with a 77 d period, producing a K ≈ 1.5 m s−1 signal. Finally, M dwarfs comprise around ∼ 70 − 75% of the stars in the Galaxy, and are by far the most common stars in the solar neighbourhood (Henry et al., 2006). The search for exoplanets around M dwarfs looks to answer several important science questions. First, the statistics of planets orbiting solar-type stars have been fairly well constrained by RV and transit surveys (e.g. Mayor et al. 2011, Silburt et al. 2015). For M dwarfs, they remain more uncertain; large surveys seek to better describe the occurrence and general properties of their planets. Second, as noted above, potentially habitable planets are more easily detected around M dwarfs; these planets are close to their stars, and therefore easier to detect and characterise. Third, M dwarfs are interesting targets in the context of planetary formation studies, because their protoplanetary disks are much less massive than those of G stars, enabling the study of the impact of disk mass on planetary formation (e.g. Mordasini et al., 2012; Gaidos, 2017). Finally, the closest planets to the Solar System will generally be the most amenable to deep characterisation with future facilities (e.g. atmospheric studies with JWST, direct imaging with ELTs). 1.3. The radial velocity method 33

The vast majority of nearby stars are M dwarfs, so volume-limited surveys on these stars should provide the best candidates for such characterisation. Several M dwarf RV surveys in the visible seek to answer these questions. Some of the main surveys are: • The HARPS M dwarf survey: The target list comprises 102 southern M dwarfs nearer than 11 pc, brighter than V=14, with rotational velocities smaller than 6.5 km s−1. The survey uses the HAPRS spectrograph, which is mounted on ESO’s 3.6 m telescope at La Silla (Chile), covers the 380–690 nm wavelength range at resolution 115 000, is calibrated with a ThAr lamp, and reaches a precision of 0.8 m s−1 (for S/N of 200 at 550 nm). Observations began in 2003. The sample, and results on observations from 2003-2009, are described in Bonfils et al.(2013); updates re- porting new planets are in Delfosse et al.(2013) and Astudillo-Defru et al.(2017b). The stellar activity of the M dwarfs was studied in Gomes da Silva et al.(2011) and Astudillo-Defru et al.(2017a). • The SOPHIE M dwarf subprogramme: Targeting a volume-limited sample of 181 M dwarfs nearer than 12 pc, the programme - which began in 2006 - was described in Bouchy et al.(2009a). It employs the SOPHIE spectrograph, which is mounted on the 1.93 m telescope at the Observatoire de Haute-Provence (France), covers the 387–694 nm wavelength range at resolution 75 000, is calibrated with a ThAr lamp, and reaches a precision of 1.1 m s−1 (for S/N of 200 at 550 nm). Four planets have been published, in Hobson et al.(2018a), Díaz et al.(2019), and Hobson et al. (2019). Part of my thesis work was dedicated to this programme; this work is presented in Chapter2. • The HArps-n red Dwarf Exoplanet Survey (HADES): This survey uses HARPS-N, a HARPS twin mounted on the 2.58 m Telescopio Nazionale Galileo at La Palma (Canaries, Spain). The original sample was comprised of 106 stars with spectral types between dM0 and dM3, listed in the APACHE catalogue, with V<12 and a high number of GAIA scans. After the first semester of observations, in 2013, 27 stars were removed, leaving a filtered sample of 79 targets. The survey was described in Affer et al.(2016), and has detected seven exoplanets so far, with the latest (and an overview to date) presented by Pinamonti et al.(2019). • The LCES HIRES/Keck survey: Not specifically an M dwarf program, but a massive survey of 1624 stars observed over more than 20 years (since 1994), including M-dwarf stars. The survey is conducted with the HIRES spectrograph, which is mounted on the 10 m Keck telescope at Mauna Kea (Hawaii, USA), covers the 364–800 nm wavelength range at resolution 55 000, is calibrated with an Iodine cell, and reaches a 1.5 m s−1 precision (for S/N of 200 at 550 nm). The survey overview paper, Butler et al.(2017), does not quantify the targets per spectral type, but cross-matching the full list with Simbad c spectral types identifies 168 of the observed stars as M dwarfs. For 28 of these, Butler et al.(2017) reported periodic RV signals; 17 are published planets, 6 are new planetary candidates, and 5 are marginal signals requiring confirmation.

c. Simbad Astronomical database, CDS, Strasbourg, France, at http://simbad.u-strasbg.fr/ simbad/ (Wenger et al., 2000). 34 Introduction

While these M dwarf surveys in the visible have produced (and continue to produce) good results, they have been unavoidably limited by the low luminosity of these stars in the visible. It is for the observation of M dwarfs in particular that high-precision nIR spectrographs have been conceived; a brief summary of the main current and near- future instruments can be found in Wright & Robertson(2017). I highlight here some of the principal current and upcoming surveys:

• CARMENES: a combination of two stabilised spectrographs (Quirrenbach et al., 2014, 2016), one in the visible(R = 94 600, 520–960 nm), one in the near infrared (R = 80 400, 960–1710 nm), mounted on the 3.5 m telescope at Calar Alto (Spain). A GTO survey with a 750-night allocation is observing 324 M dwarfs, since 1 Jan 2016. The sample is described in Reiners et al.(2018b), and was selected taking the brightest targets of every spectral subtype, visible from Calar Alto, that are not members of close multiple systems. The team has published several planets, but to date have used only the visible arm RVs, due to high uncertainties on the nIR RVs. • SPIRou: a nIR spectropolarimeter covering the YJHK bands (980–2350 nm) at 70 000 resolution (Artigau et al., 2014), mounted on the 3.6 m Canada-France- Hawaii Telescope at Mauna Kea (Hawaii, USA) and in operation since 2019. The Spirou Legacy Survey (input catalogue creation described in Moutou et al. 2017, Fouqué et al. 2018) will observe 100 M dwarfs over 300 nights in four years, and began observations in semester 2019A. It is described in more detail in Sect. 3.6.1. • NIRPS, an upcoming nIR (YJH bands) companion to HARPS (Bouchy et al., 2017) based on the SPIRou design, to be mounted on the ESO’s 3.6 m telescope at La Silla (Chile). It aims for ≈ 70 000 resolution, and will be operated simultaneously with HARPS.

1.4 M-dwarf stars and their planets (as of 2016)

In this section, I aim to explore the general properties of M-dwarf planet hosts and of the planets themselves, as they stood at the beginning of the thesis in November 2016. In order to carry out this study, the first step is the selection of an exoplanet catalogue from which to construct a catalogue of M dwarf planets. Two recent publications investigated the main catalogues in the field: Christiansen(2018) presented an overview, focused primarily on the Exoplanets Encyclopaedia (http://exoplanet.eu/), the NASA Exoplanet Archive (https://exoplanetarchive.ipac.caltech.edu/), and the Open Exoplanet Catalogue (http://www.openexoplanetcatalogue.com/), while Bashi et al.(2018) carried out detailed statistical comparisons in which they also included the Exoplanet Data Explorer (http://www.exoplanets.org/) of the Exoplanet Orbit Database. The Exoplanet Orbit Database will no longer be regularly updated as of June 2018 (and had not been updated for two years prior to the final update), so it was excluded from consideration. The Open Exoplanet Catalogue is an open-source, community- maintained catalogue based on github; as such, the update schedule is inconsistent and planet incorporation can be slow (at the time of writing, no planets discovered in 2019 1.4. M-dwarf stars and their planets (as of 2016) 35

are included). Additionally, the sources of planetary and stellar parameters are available only in github commit messages, making them less accessible compared to the Exoplan- ets Encyclopaedia and the NASA Exoplanet Archive (both of which list them directly on the respective websites). Therefore, it was also removed from consideration. The final choice between the last two is the most complex. The Exoplanets Encyclopaedia provides all published parameters for each planet, which may come from different sources, while the NASA Exoplanet Archive designates a single source of default parameters per planet. This means that the Exoplanet Encyclopaedia’s parameters can be more complete but may not be self-consistent, while the NASA Exoplanet Archive’s parameters are self-consistent for each system but may have missing values. Bashi et al.(2018) found an overall relatively good agreement between these two catalogues, with some differences in the unique subsets (i.e. the planets found in one catalogue but not the other). In particular, with regards to stellar mass, their CDFs suggest that the Exoplanet Ency- clopaedia’s unique subset includes more M-dwarf hosts. The result of filtering both tables by stellar mass (in the range 0.06 M ≤ M∗ ≤ 0.6 M , following Cox 2000, Kaltenegger & Traub 2009) is in agreement with this, with the Exoplanet Encyclopaedia providing 359 records and the NASA Exoplanet Archive only 298. As we are interested in these low-mass stars and their planets in particular, I chose to work with the database provided by the Exoplanet Encyclopaedia. I manually inspected the 359 planets listed by the Exoplanet Encyclopaedia (as of th 19 April 2019) as having a stellar mass between 0.06 M and 0.6 M (following the mass limits given by Cox 2000, Kaltenegger & Traub 2009) to verify their host’s spectral type, by consulting Simbad and Vizier d (when the spectral type was not listed on Sim- bad). I removed any planets hosted by stars other than main-sequence M dwarfs (consisting mainly of evolved stars such as subdwarfs and white dwarfs, plus some brown dwarfs and K stars), leaving a filtered list of 316 planets. Of these, 108 have no reported planetary mass and were removed. These are Kepler detections in multiplanetary systems, that were validated in Rowe et al.(2014) by statistical analyses, and are also included in the NASA Exoplanet Archive’s catalogue. Next, I set a planetary mass limit of 13 MJ (in line with the definition of the IAU’s Working Group on Extrasolar Planets, Boss et al. 2007) to filter out 25 brown-dwarf companions, for a final set of 183 planets. It should be noted here that there is a further caveat to be taken into account for the Exoplanet Encyclopaedia: they compile planets not only from published papers, but also from submitted pre-prints (e.g. ArXiv), conference announcements, and selected pro- fessional websites. However, inspecting the sources of the M dwarf planets shows that the vast majority (168) come from refereed papers. I manually reviewed the remainder (2 announced in conferences, 13 submitted to journals), finding that in all cases subsequent published papers are available, and are the source of the Encyclopaedia’s reported parameters. Of the 183 total planets, 118 were detected in or before 2016: 54 by radial velocities, 31 by the transit technique, 23 by microlensing, 8 by imaging, 1 by transit timing variations, and 1 by the pulsar method. Comparison of the planets found by these different techniques is naturally somewhat limited, as each method has its own subset of

d. VizieR catalogue access tool, CDS, Strasbourg, France, athttp://vizier.u-strasbg.fr/ viz-bin/VizieR (Ochsenbein et al., 2000). 36 Introduction the parameter space it covers (e.g. Fischer et al., 2014). For example, radial velocities provide a lower limit to the planetary mass but cannot fix the orbital inclination; transits offer planetary radii and orbital inclinations but (save for cases of transit timing variations in multiplanetary systems) no masses; microlensing and imaging are sensitive to planets at a few AU, contrasting with RVs and transits that favour close-in planets; the spectral classes of stars hosting planets detected by microlensing are uncertain as the stars are distant. Some overlaps, in which a planet can be detected by two techniques, are possible and provide additional information, such as RV confirmations of transiting planets. Nevertheless, despite these limitations, the combined analysis of the planets detected by all methods can provide interesting information. Figure 1.8 shows an analogous mass-period diagram to Fig. 1.2 but for M-dwarf planets only, this time colour-coded by eccentricity, as it stood in 2016. Empirically determined masses (obtained by radial velocities, transit timing variations, and microlensing) are differentiated from theoretical mass estimates (from upper limits, mass-radius relationships, or evolutionary models). Most of the planets are relatively small, clustering in the 1 − 10 M⊕ range (median 7.98 M⊕), over periods ranging from 1–200 d (median 16.6 d); Saturn or Jupiter-sized and long-period planets are rare. Likewise, the majority have relatively low eccentricities (median 0.08). This is consistent with results for FGK stars that low-mass planets tend to be less eccentric (Adibekyan et al., 2013), and is in accordance with pebble-accretion and migration formation models, which predict generally small eccentricities for lower-mass planets (e.g. Raymond et al., 2018; Izidoro et al., 2019). Using the mass ranges established by Stevens & Gaudi(2013), we can divide the planets into 40 Earths (Mp < 2 M⊕), 21 superEarths (2 M⊕ ≤ Mp < 10 M⊕), 28 Nep- tunes (10 M⊕ ≤ Mp < 100 M⊕), and 16 Jupiters (100 M⊕ ≤ Mp < 13 MJ). The lack of hot Jupiters is particularly interesting, as these planets are (as discussed in Sect. 1.1) the easiest to detect, due to their large sizes and short periods; therefore, their absence is presumably real and not due to detection effects (see e.g. Bonfils et al. 2013, who found frequencies of f . 1% for giant planets with periods shorter than 1–10 d, compared to +0.25 f = 0.36−0.10 for superEarths in the same period range). Similarly to the overall planet distribution for all stellar hosts, hot Neptunes are rare around M dwarfs. The M dwarf planets follow the lower boundary of the hot Neptune desert (as defined by Mazeh et al. 2016, see Fig. 1.8) well; the consistency with the upper boundary is less defined due to the overall lack of giant planets. 1.4. M-dwarf stars and their planets (as of 2016) 37

Figure 1.8– mass-period diagram for all known exoplanets (as of 2016) hosted by M dwarfs, with these parameters reported in "The Exoplanet Encyclopaedia". Minimum mass M sin i values are used as mass when the inclination is not known. The points are colour-coded by eccentricity; planets without reported eccentricities are plotted in black. Diamond symbols represent measured masses (RV, TTV, microlensing); circles represent mass estimates (upper limits, mass-radius relations, evolutionary models). The dashed black lines show the limits of the hot Neptune desert as deﬁned by Mazeh et al.(2016).

The question of possible links between stellar metallicity and planetary presence has a long history in exoplanetary science. While solar-type stars hosting hot Jupiters tend to be metal-rich (e.g. Fischer & Valenti 2005, Ghezzi et al. 2010, Adibekyan 2019 and references therein), the answer is not so clear for lower-mass planets (e.g. Buchhave et al. 2012, Courcol et al. 2016, Petigura et al. 2018, Sousa et al. 2018), nor for M dwarfs (Hobson et al. 2018b and references therein). Figure 1.9 shows the planet mass (or minimum mass) as a function of stellar metallicity, for planets with measured masses. The different mass-determination methods are distinguished; all but four planets have masses determined by RVs. For stars hosting planets with RV-determined masses, there is a suggestion of correlation between increasing metallicity and higher planetary mass (Pearson correlation coefﬁcient ρ = 0.518, p-value = 1.642×10−4). Among these stars are two multi-planet hosts with very low metallicities, Kapteyn’s star with two planets and [Fe/H] = −0.86 (the least metallic star in the sample), and GJ 667 C with six planets and 38 Introduction

[Fe/H] = −0.55, which may bias the sample. Removing them, the correlation coefficients become ρ = 0.570, p-value = 1.222 × 10−4. Likewise, if we consider the sum of planetary masses, the correlation coefficients are ρ = 0.449, p-value = 1.879 × 10−2. However, it must be noted that the reported metallicities in the Exoplanet Encyclopaedia come from a variety of sources, and may not be mutually consistent, especially given the difficulties of deriving M dwarf metallicities. Several works have attempted to establish homogeneous stellar parameters for large samples of exoplanet hosts (e.g. SWEET-CAT, Santos et al. 2013, Sousa et al. 2018 focused on FGK stars; CONCH-SHELL, Gaidos et al. 2014, focused on M dwarfs). The CONCH-SHELL catalogue reports metallicities for 23 of the M dwarf hosts from the Exoplanet Encyclopaedia. There is a median difference of [Fe/H]diff − 0.03 between the two catalogues, which is well below the typical uncertainty of 0.11 for CONCH-SHELL.

(b) Sum of RV-determined planetary (a) Planet mass (or minimum mass M sin i) masses (or minimum masses M sin i) per as a function of stellar metallicity. system as a function of stellar metallicity.

Figure 1.9– Planet mass or minimum mass M sin i (left, 1.9a), or sum of masses/minimum masses per system (right, 1.9b), as a function of stellar metallicity for all known exoplanets (as of 2016) hosted by M dwarfs, with these parameters reported in "The Exoplanet Encyclopaedia". Only planets with measured masses are shown in 1.9a, and only RV-determined masses in 1.9b.

Statistical studies suggest that most M-dwarf planetary systems should be multiple - see e.g. Dressing & Charbonneau 2015, who estimate a cumulative planet occurrence rate of 2.5 ± 0.2 planets per M dwarf from Kepler data, though arguments have also been made for a more complex two-population model divided into one-to-two-planets systems and ﬁve-or-more-planets systems (e.g. Ballard 2019 and references therein). In the sample analysed here, the 118 planets form a total of 80 systems, of which 23 are multiple. The most populated multiplanetary system is the one hosted by GJ 667 C, with up to six planets; it is followed by GJ 876 with four planets, nine three-planet systems, and twelve two-planet systems. 1.5. Thesis objectives 39

1.5 Thesis objectives

The main goal of this thesis is the study of exoplanets around M dwarfs through radial velocity. More specifically, I sought to improve the data reduction and analysis, from calibration through to radial velocity extraction, in order to enable the detection of the smallest planets. Within this broad goal, I focused on two main topics, each corresponding to a different wavelength range and instrument. In the optical domain, I improved the detection and characterisation of M dwarf planets for the SOPHIE M dwarf survey. To do this, I applied template-matching methods for the radial velocity determination. I also identified and corrected several instrumental effects, and implemented an improved computation of activity indices. This work is described in Chapter2. In the nIR domain, I worked on the data reduction system for the SPIRou spectropolarimeter, a new instrument that will dedicate most of its observing time to the search for exoplanets orbiting M-dwarf stars. The core of my work was on the wavelength calibration, for which I developed and tested a variety of methods. This is presented in Chapter3. Section 1.4 of this chapter described the known exoplanets orbiting M-dwarf stars at the beginning of my thesis. During the three years of this PhD, we have seen a veritable explosion of M dwarf exoplanet detections. With this increased sample, we can begin to attempt firmer answers to some of the science questions raised in Sect. 1.3.2. I carried out a general study of the known M dwarf exoplanets, updating and expanding on the analysis of Sect. 1.4, in Chapter4. Finally, the conclusions and future perspectives are presented in Conclusions.

2 M-dwarf RV search with the SOPHIE spectrograph

Contents

2.1 The SOPHIE spectrograph ...... 42 2.2 The M-dwarf sample...... 43 2.3 The template-matching method...... 45 2.4 Instrumental instabilities...... 48 2.4.1 Charge transfer inefficiency ...... 48 2.4.2 Nightly drift ...... 49 2.4.3 Long-term variation of the zero-point...... 50 2.5 Summary of SOPHIE results on M dwarfs...... 52 2.5.1 Discovery of new exoplanets...... 55 2.5.1.1 Two planets around M-dwarfs Gl 617A and Gl 96 . . . . 55 2.5.1.2 A warm Neptune around the M dwarf Gl 378 ...... 71 2.5.1.3 Other planets...... 80 2.5.2 Confirmation or non-confirmation of published planets ...... 81 2.5.3 Stellar activity mitigation...... 83 2.6 Other contributions to SOPHIE RV programmes...... 90 2.6.1 High precision RV search for super-Earths...... 90 2.6.2 Simultaneous FP background correction...... 92 42 Chapter 2. M-dwarf RV search with the SOPHIE spectrograph

2.1 The SOPHIE spectrograph

SOPHIE a is a fibre-fed, cross-dispersed echelle spectrograph mounted on the 1.93 m telescope at the Observatoire d’Haute-Provence, which has been operational since 2006 (Bouchy & Sophie Team 2006, Perruchot et al. 2008). In 2011, SOPHIE was significantly upgraded by the installation of octagonal fibres and a double scrambler (Perruchot et al., 2011), improving its RV precision by a factor of ∼ 6. The upgraded instrument was re- named SOPHIE+ (Bouchy et al., 2013), but is generally referred to simply as SOPHIE un- less the distinction is relevant. I will follow this convention here. The spectrograph covers the 387–694 nm wavelength range over 39 spectral orders. It has a Zerodur R2 echelle grating (blaze angle 65°) and a double-pass prism cross- disperser (PBL25Y glass, angle 31°). The detector is a 2k × 4k CCD with 15 µm pixels, with an overall quantum efficiency of ∼ 80%, cooled to −100 ◦C. There are two resolution modes available: high efficiency (HE, R ≈ 39 000) and high resolution (HR, R ≈ 75 000). For SOPHIE+, Bouchy et al.(2013) estimate typical precisions of 1–2 m s−1 for HR and 3–4 m s−1 for HE. In addition to the star fibre (fibre A), SOPHIE possesses a second calibration fibre (fibre B) which can be used either for sky monitoring, or for simultaneous calibration. Until semester 2017A, the only calibration lamp available was a ThAr hollow-cathode lamp. Since semester 2017B, a Fabry-Pérot (FP) étalon is also available, and is the recommended calibrator. Part of a raw SOPHIE spectrum, with an M dwarf on fibre A and the FP on fibre B, is shown in Fig. 2.1. The SOPHIE spectra are processed by a fully automated pipeline, known as the Data Reduction System or DRS, adapted from the HARPS pipeline (Bouchy & Sophie Team 2006, Bouchy et al. 2009a). For each observing night, the DRS first deals with the afternoon calibrations, generating dark current and flat field maps, creating a wavelength solution from a ThAr spectrum, and setting a zero-point for the nightly drift from ThAr and FP spectra. Next, for stellar observations, it takes the raw spectra, performs dark and flat field corrections, extracts wavelength-calibrated two-dimensional and one-dimensional spectra, and measures the radial velocities through the CCF (cross-correlation function) method. This method is based on cross-correlating the stellar spectra with a weighted binary mask (see e.g. Queloz 1995, Pepe et al. 2002a).

a. Spectrographe pour l’Observation des Phénomènes des Intérieurs stellaires et des Exoplanètes; instrument website: http://www.obs-hp.fr/guide/sophie/sophie-eng.shtml 2.2. The M-dwarf sample 43

Figure 2.1– Part of a raw SOPHIE spectra for an M3 dwarf, with the star on the science ﬁbre and the FP illuminating the calibration ﬁbre. A bad column on the CCD can be seen on the left of the image as a black vertical line crossing the second spectral order.

2.2 The M-dwarf sample

Since 2006, the SOPHIE exoplanet consortium b has been carrying out several exoplanet- hunting programmes (Bouchy et al., 2009a). Subprogramme 3, generally known as SP3 (PI X. Delfosse), is the subprogramme dedicated to the search for exoplanets around M-dwarf stars. Its objectives are: • to detect habitable superEarths and Neptunes; • to constrain the statistics of planets around M dwarfs; • to ﬁnd potentially transiting companions.

The initial target list for the SP3 consisted of a volume-limited sample of 180 M- dwarf stars, at a distance of less than 12 pc, with δ ≥ 0° and V ≤ 14.0 (Bouchy et al., 2009a). RV measurements from ELODIE (the precursor spectrograph to SOPHIE, see

b. The consortium nucleates SOPHIE exoplanet observers (primarily in France, Switzerland, and Portugal), with the aim of optimising observing time and sharing expertise. 44 Chapter 2. M-dwarf RV search with the SOPHIE spectrograph

Baranne et al. 1996) and adaptive optics imaging allowed the removal of fast rotators and of binaries with a separation smaller than 700 from the volume-limited sample, leaving 120 targets. The brightest stars (V<9) were to be observed with simultaneous calibration, as they are bright enough that moonlight contamination should not be a problem, and contaminating flux from the calibration spectrum on fibre B is negligible. The fainter stars (V>9) would be observed with simultaneous sky, in order to detect potential contamination from the moon. In this case, a calibration exposure (ThAr) was to be taken just before the stellar observation, in order to monitor and correct potential instrumental drift. A second magnitude limit was established for selecting the resolution mode: bright targets with V<12 would be observed in HR mode, faint stars with V>12 in HE mode. As a general rule, the programme aimed for a minimum of thirty measurements for each target before deciding whether to stop or continue observations. Over the duration of the programme, the target list evolved somewhat. A set of more distant stars in the 12–25 pc range, with δ ≥ 0°, V ≤ 12.0, and known parallaxes from the NSTARS database, were added, incorporating a total of 61 new targets. Figures 2.2a, 2.2b, and 2.2c show V magnitude, distance, and spectral type histograms for this new 181-target list. Most of the stars are early-to-mid M dwarfs, with a few late K stars also included. Further modifications were subsequently made regarding the observing modes. Fol- lowing the installation of the octagonal fibres, the HE mode was finally judged to be insufficiently accurate for the programme goals. 37 of the HE targets were therefore no longer observed, while the remaining 41 were changed to HR mode with increased exposure times. At the start of the program, simultaneous and prior calibrations were performed with the ThAr lamp. After the installation and validation of the FP étalon, in semester 2017B, these calibrations were switched to the FP in semester 2018A. Addi- tionally, observations were stopped for any stars that showed strong variability in the Ca II H and K lines (at 396.9 nm and 393.4 nm respectively) once it came into evidence, on a case-by-case basis. 2.3. The template-matching method 45

(a) Histogram of the V magnitudes of the (b) Histogram of the distances of the cur- current SP3 catalogue. The limits for rent SP3 catalogue. The limit of the simultaneous calibration and for HR ob- volume-complete sample is indicated. servation are indicated.

Figure 2.2– Magnitude (top left), distance (top right), and spectral type (bottom left) histograms for the ﬁnal 181-target catalogue of the SOPHIE M dwarf subprogramme.

Part of my thesis was dedicated to a re-analysis of the SP3 measurements. I applied a template-matching algorithm to re-determine the radial velocities more precisely, and analysed the resulting data-sets for periodic signals induced by planets. I also calculated several activity indices from the literature, with the aim of determining which are the most suited to detecting stellar-induced periodicities in SOPHIE data.

2.3 The template-matching method

Traditionally, radial velocities have been generally been calculated by the CCF method; this is true for the SOPHIE DRS, as described in Section 2.1. The CCF method is well adapted to G-type stars, with a well-deﬁned continuum and sharp spectral lines. M dwarfs, on the other hand, have complex spectra with a multitude of small lines and 46 Chapter 2. M-dwarf RV search with the SOPHIE spectrograph molecular bands (Fig. 2.3). This means that the CCF method, which selects only spe- ciﬁc lines in its binary mask, underutilises the Doppler information present. Template- matching with a true stellar spectrum allows for a more precise determination of the radial velocity (e.g. Anglada-Escudé & Butler 2012, Astudillo-Defru et al. 2015). Addi- tionally, template-matching can also enable a better handling of any telluric contamination: a template constructed from observations over a wide span of barycentric radial velocities will remove any telluric lines present in the individual spectra, preserving only the stellar lines.

Figure 2.3– One red order of a SOPHIE spectra for an M3 dwarf (blue, the most common spectral type in the SP3 sample), and a G5 star (orange, presented for comparison).

I adapted an algorithm developed by N. Astudillo-Defru (Astudillo-Defru et al. 2015, Astudillo-Defru et al. 2017b) to SOPHIE M dwarf spectra, and applied it to all SP3 targets with at least ten HR observations since 2011. The date cut-off corresponds to the instrument upgrade from SOPHIE to SOPHIE+; data before the upgrade have much lower accuracy, so the template-matching algorithm is not expected to provide signiﬁcant improvements. As the stellar template is built from the set of input spectra, a minimum amount of ten observations are needed in order to build a high signal-to-noise template. A minimum time span is also advisable, to allow for a good correction of the telluric contamination. Using the CCF RV from the SOPHIE DRS as a ﬁrst guess for each spectra, the al- 2.3. The template-matching method 47 gorithm follows these steps to derive radial velocities (as described in Astudillo-Defru 2015):

• Construction of the stellar and telluric templates: The spectra are shifted to the stellar frame (for the stellar template, using first-guess stellar RVs) or the rest frame (for the telluric template, using the BERV), and resampled. The template is constructed as the median of the spectra. The process iterates once. • Determination of the precise RV for each epoch by χ2 minimisation: the stellar template is shifted over a grid of RVs around the first-guess RV. For each RV, the following process takes place: · The derivatives of the spectrum with respect to the (scaled) telluric and stellar templates are computed, and the ratio between them is calculated over regions of fixed width. · Regions where the ratio is below a set value (slope) are rejected as unin- formative, and a 5-σ clipping is performed. · This process iterates once, then the χ2 is computed. A Voigtian is fitted to the overall χ2 profile, and a Gaussian to its centre; the centroid of the Gaussian is taken as the final RV. Errors are calculated following Bouchy et al.(2001).

Two free parameters, the width of the region where the ratio of the derivatives is calculated (henceforward "width", measured in pixels), and the cut-off threshold for the ratio (henceforward "slope", dimensionless), had been set by Astudillo-Defru(2015) using HARPS spectra to defaults of width = 7 and slope = 10. These needed to be tested for SOPHIE spectra. I used spectra from Gl411 and Gl514 (two of the program standard stars) to run tests, varying the width between 3 and 15 for a ﬁxed slope of 10, and the slope between 0 and 25 for a ﬁxed width of 7. Figure 2.4 shows the RV scatter as a function of the width and slope respectively, colour-coded by percentage of the spectra that was rejected. The aim was to balance a low RV scatter with a low percentage of suppressed regions. The RV scatter varies very little with changing width (standard deviation ∼0.07 m s−1 for both targets), but is lowest for values of 5-7. I therefore decided to keep the initial value of width = 7, since it minimises the rejection percentage for both. For changing slope, the RV scatter descends fairly smoothly (save for slope = 0, which is equivalent to no rejection) until values of 10-15, then rises again. Once again, I chose to keep the initial parameter of slope = 10, since it provides some of the lowest RV scatters while not rejecting too large a percentage of the spectra. 48 Chapter 2. M-dwarf RV search with the SOPHIE spectrograph

(a) Variation of the width parameter for (b) Variation of the slope parameter for slope=7. width=10.

Figure 2.4– RV scatter and percentage of spectra rejected when varying the slope and width parameters.

2.4 Instrumental instabilities

Several instrumental effects need to be corrected for on SOPHIE data: • The charge transfer inefﬁciency effect (CTI); • The nightly spectrograph drift; • The long-term variation of the zero-point.

2.4.1 Charge transfer ineﬃciency

The CTI effect is a well-known phenomenon on CCD detectors. In order to read out a CCD, multiple charge transfers are performed: for each row, first, the charges are transferred through successive rows until they reach the final row, the serial register; then, they are transferred along the register, one pixel at a time, to the output amplifier. These charge transfers are imperfect, and electrons are lost along the way; the loss is quantified by the Charge Transfer Efficiency (CTE) or its inverse, the Charge Transfer Inefficiency (CTI). In terms of radial velocities, Bouchy et al.(2009b) showed that for SOPHIE, where the echelle orders are essentially perpendicular to the serial register, the CTI effect translates into an SNR-dependent drop in RV. For the SP3 targets, an inspection of the output RVs from the template-matching showed a clear drop towards low SNR (Figure 2.5). 2.4. Instrumental instabilities 49

Figure 2.5– Radial velocity offset as a function of the signal-to-noise ratio in order 26, for a selection of SP3 targets. The offset is computed as RV − median(RV) for each star, where the median was performed only over observations with SNR > 75 to minimise any bias from CTI-affected RVs. The selected targets have at least 30 observations, −1 and σRV <10 m s .

The CTI effect for SOPHIE was ﬁrst characterised in Bouchy et al.(2009b), where a correction of the extracted two-dimensional spectra was proposed following Goudfrooij et al.(2006). An alternative empirical correction of the radial velocities as a power law of the SNR was proposed by Santerne et al.(2012). However, this correction, which was derived from blue sky spectra, is not a good ﬁt for the SP3 sample. The extracted two-dimensional spectra correction, on the other hand, provided an average RV RMS improvement of ∼ 15%. Therefore, I implemented the correction given in Bouchy et al. (2009b) in the template-matching algorithm.

2.4.2 Nightly drift

The nightly drift of SOPHIE is regularly monitored through calibration exposures, using a ThAr lamp (up to semester 2017A) or a FP étalon (since semester 2017B). While 50 Chapter 2. M-dwarf RV search with the SOPHIE spectrograph both calibrators allow the drift to be measured, there is a conceptual difference. When a ThAr spectrum is taken during the observing night, a new wavelength solution is generated, the drift is obtained by comparing with the previous wavelength calibration, and the zero-point is reset to the new wavelength solution. The FP, on the other hand, provides only a relative drift from the zero-point set by the most recent wavelength solution generated (for normal operation, this will be from the afternoon calibrations). Observers are requested to take a calibration exposure (i.e. a spectrum of the calibrator on both ﬁbres) every two hours approximately. Generally, the spectrograph drifts around 3 m s−1 during a night. When using the FP to monitor the drift, observers are further requested to reset the zero-point by performing an additional ThAr exposure if the drift measured by the FP exceeds 5 m s−1. For SP3 targets, I corrected the drift using either the simultaneous calibration (for bright stars), or the interpolation between previous and following calibrations (for fainter targets).

2.4.3 Long-term variation of the zero-point

In addition to the intra-night drift, SOPHIE displays a long-term variation of the zero-point, an effect first characterised by Courcol et al.(2015). The variation takes the form of a long-term drift, with occasional sharp jumps, which are identified as related to different instrumental changes (e.g. lamp changes, thermal regulation). This characterisation was constructed on a sample of FGK stars, and it was suspected that at least one of these jumps (at BJD = 56775) had a chromatic dependency, as it was not seen in the radial velocities of the SP3 standard stars. Therefore, I followed the procedure described by the authors to construct an analogous correction using M dwarf spectra. I tested two different corrections: one constructed using M dwarf spectra alone, and one incorporating the four G-type "super-constants" defined by Courcol et al. (2015), HD185144, HD9407, HD22154, and HD89269A, reprocessed with template- matching. I found that re-processing with template-matching eliminated the offset that was suspected of chromatic dependency, including for the G-type stars. This jump, which had been attributed by Courcol et al.(2015) as due to the re-coating of the secondary mirror, I found to be more specifically caused by an instability of the wavelength solution on the blue spectral orders (whose weight was modified by the mirror re-coating). The effect of this instability was less pronounced in the CCF RVs for M-dwarf stars because they are faint in the blue, and is not apparent in the template-matching RVs for either G or M stars. Finally, I analysed the dispersion of the corrected RVs for each constant correction. Those calculated with the correction using M dwarfs alone showed a higher dispersion (by 1.5 m s−1 on average), so I chose to employ the correction built from the M dwarfs plus the four super-constants. The constant correction, computed with data up to 30th Mar. 2019, is shown in Fig. 2.6. Ten stars were used to create this correction: The four super-constants HD185144, HD9407, HD22154, and HD89269A, which were defined by the SP1 subprogramme on 2.4. Instrumental instabilities 51 sun-like stars (Bouchy et al., 2009a) and at least one of which is observed each night with consortium observations; the SP3 standards Gl411, Gl514, and Gl686, defined at the start of the SP3 programme, at least one of which is observed each night with SP3 observations; and the additional SP3 stars Gl521, Gl15A, and Gl694, incorporated automatically by the script thanks to their low RV variations. Several offsets due to instrumental changes are highlighted; the dates and causes are summarised in Table 2.1.

Table 2.1– SOPHIE long-term drift offsets and their cause. BJD 2 400 000 Calendar Date Cause 55872 6th Nov. 2011 replacement of the ThAr calibration lamp 56274 12th Dec. 2012 installation of octagonal fibres after double scrambler 56690 1st Feb. 2014 replacement of the ThAr calibration lamp 56730 13th Mar. 2014 installation of a new calibration unit 56941 10th Oct. 2014 modification of the set current for the ThAr lamp 57435 16th Feb. 2016 installation of new thermal regulation 58419-58431 28th Oct. - 9th Nov. 2018 pressure leak in the vacuum enclosure 52 Chapter 2. M-dwarf RV search with the SOPHIE spectrograph

Figure 2.6– Correction of the long-term variation of the zero-point (red line), together with the radial velocities used to construct it (black dots). Jumps due to instrumental changes are indicated by vertical lines. The correction has a dispersion of 2.96 m s−1 and a peak-to-peak variation of 16 m s−1.

2.5 Summary of SOPHIE results on M dwarfs

Of the initial sample of 181 M dwarfs, 56 currently have at least ten measurements performed in HR mode with SOPHIE+ (i.e., after the implementation of the double scrambler and octagonal ﬁbres). The rest of the sample was either observed in HE mode, completed prior to the instrument upgrade, or stopped early after proving to be active. I applied the template-matching algorithm and instrumental corrections discussed in Sects. 2.3 and 2.4 to the SOPHIE spectra of these 56 targets. Additionally, for the fainter targets with simultaneous sky monitoring, I checked for moon contamination, using a merit function developed by X. Delfosse. It is used for observations where the RV difference between the moon and the star is less than 20 km s−1, and is calculated using this RV difference and the S/N and CCF contrast in ﬁbre B. 2.5. Summary of SOPHIE results on M dwarfs 53

I used the Data Analysis Center for Exoplanets (DACE) web platform c to analyse the radial velocities calculated with template-matching for each target. Given a series of radial velocities, DACE automatically generates a plot of the time series, and a generalised Lomb-Scargle (GLS) periodogram with false alarm probability (FAP) levels indicated. By default the FAP levels are calculated analytically following Baluev(2008), but they can also be calculated numerically by bootstrap resampling. I inspected each periodogram for significant peaks, which could indicate the presence of planets. Table 2.2 summarises the number of targets whose highest peak is above the 10%, 1%, or 0.1% FAP line respectively - i.e., the number of targets whose most significant signal’s FAP is below 10%, 1%, or 0.1% respectively (signals below 1% or 0.1% are not included in the 10% count, and signals below 0.1% are not included in the 1% count). 23 stars of the 56-target sample had no signals below 10% FAP. They have a median of 34 observations, a median V magnitude of 10.57, and a median RV dispersion of 4.1 m s−1. That they show no signals does not necessarily mean that these stars do not host planets, only that any planets they may host are not detectable from the SOPHIE spectra gathered to date. Conversely, a significant signal does not necessarily indicate the existence of a planet; it may also be originated by stellar activity.

Table 2.2– Summary of the SOPHIE M dwarf RVs periodogram analysis. Minimum FAP level Number of targets 10% 13 1% 8 0.1% 12

The number of observations and RV dispersion for each star are shown in Fig. 2.7, colour-coded by V magnitude and distinguished by FAP category. Three stars have much higher dispersion than the rest; the variability does not show any periodicity, but is presumably due to stellar activity. Observations were ceased for these stars once this high non-periodic variability was conﬁrmed. Setting aside these stars, the dispersion and V magnitudes are similar throughout the different FAP categories. While stars with signals below 0.1% FAP are overall those most observed, this is generally a natural result of intensive follow-up in priority once an emerging signal was found.

c. Available at https://dace.unige.ch 54 Chapter 2. M-dwarf RV search with the SOPHIE spectrograph

Figure 2.7– Number of observations and RV dispersion for each star of the SP3 sample with at least ten observations. The points are colour-coded by V magnitude; the symbols distinguish the FAP category of the strongest signal present in the corresponding periodogram.

Regarding the twelve targets with signals below 0.1% FAP, nine are held to be true planets, and are discussed in Sections 2.5.1 (new planets) and 2.5.2 (confirmation of published planets) respectively. For these, I also inspected the residuals of the keplerian fits to the planets. The results are summarised in Table 2.3, which is analogous to Table 2.2. Two of the stars show no significant signals whatsoever in their residuals, and none have signals below 0.1% FAP.

Table 2.3– Summary of the SOPHIE M dwarf RV residuals periodogram analysis, for stars hosting planets, with the planet removed by a keplerian ﬁt. Minimum FAP level Number of targets 10% 4 1% 3 0.1% 0

Of the three remaining stars with signals below 0.1% FAP, one is clearly attributable 0 to stellar activity, with significant signals in both Hα and log(RHK) at the same periods, 2.5. Summary of SOPHIE results on M dwarfs 55 and is discussed in Section 2.5.3. The final two are less clear, but show season-to-season variations that make them suspect. In both cases, the results from the DRS reduction and template-matching were inconsistent. For the first target, Gl AAA, the DRS reduction shows a significant signal at 155 d which is not visible in the template-matching reduction. However, in the template-matching periodogram a significant signal can be found at 0.5 d, but has only become apparent with the latest season of data. For the second, Gl BBB, a significant signal was found in the DRS reduction at 2 d, but was never visible in the template-matching reduction, and its importance in the DRS reduction decreased after 100 observations. Meanwhile, for the template-matching a significant signal was found at 16.5 d in the latest reduction, but is likely due to activity as it is driven by the latest season only. However, there is also a moderately significant signal at 9 d present in both periodograms; analysis is currently ongoing to attempt to determine whether it is another activity signal, or a planet being masked by stellar activity.

2.5.1 Discovery of new exoplanets

2.5.1.1 Two planets around M-dwarfs Gl 617A and Gl 96

The ﬁrst two planets from the SP3 survey were presented in Hobson et al.(2018a), for which I led the publication: the discovery of Gl 96 b and the independent detection of Gl 617A b. I summarise here the planets’ main properties, and present the full paper in the subsequent pages.

Gl 96 b A Neptune-mass (19.66 M⊕) planet with a 73.9 d orbital period, and the third highest eccentricity (e = 0.44) of known planets around M dwarfs. The host is an M2 star. Gl 96 b is close to the inner edge of the habitable zone, but its high eccentricity takes it too close to the star at periastron.

Gl 617A b A Neptune-mass (31.29 M⊕) planet, orbiting an M1 star at 86.7 d orbital period and potentially within the habitable zone, ﬁrst published by Reiners et al.(2018a). We present an independent detection, and a re-analysis of the combined data to compare and reﬁne orbital parameters. A&A 618, A103 (2018) Astronomy https://doi.org/10.1051/0004-6361/201832732 & © ESO 2018 Astrophysics

The SOPHIE search for northern extrasolar planets?,?? XIII. Two planets around M-dwarfs Gl617A and Gl96 M. J. Hobson1, R. F. Díaz2,3, X. Delfosse4, N. Astudillo-Defru5,6, I. Boisse1, F. Bouchy6, X. Bonﬁls4, T. Forveille4, N. Hara6,7, L. Arnold8, S. Borgniet4, V. Bourrier6, B. Brugger1, N. Cabrera4, B. Courcol1, S. Dalal9, M. Deleuil1, O. Demangeon10, X. Dumusque6, D. Ehrenreich6, G. Hébrard9,8, F. Kiefer9, T. Lopez1, L. Mignon4, G. Montagnier9,8, O. Mousis1, C. Moutou1,12, F. Pepe6, J. Rey6, A. Santerne1, N. Santos10,11, M. Stalport6, D. Ségransan6, S. Udry6, and P. A. Wilson9,13

1 Laboratoire d’Astrophysique de Marseille, UMR 7326, CNRS, Aix-Marseille Université, 13388 Marseille Cedex 13, France e-mail: [email protected] 2 Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Buenos Aires, Argentina 3 Instituto de Astronomía y Física del Espacio (IAFE), CONICET – Universidad de Buenos Aires, Buenos Aires, Argentina 4 CNRS, IPAG, Université Grenoble Alpes, 38000 Grenoble, France 5 Departamento de Astronomía, Universidad de Concepción, Casilla 160-C, Concepción, Chile 6 Observatoire Astronomique de l’Université de Genève, 51 Chemin des Maillettes, 1290 Versoix, Switzerland 7 ASD/IMCCE, CNRS-UMR8028, Observatoire de Paris, PSL, UPMC, 77 Avenue Denfert-Rochereau, 75014 Paris, France 8 Observatoire de Haute-Provence, CNRS, Institut Pythéas UMS 3470, Aix-Marseille Université, 04870 Saint-Michel-l’Observatoire, France 9 Institut d’Astrophysique de Paris, UMR7095 CNRS, Université Pierre & Marie Curie, 98bis Boulevard Arago, 75014 Paris, France 10 Instituto de Astrofísica e Ciências do Espaço, CAUP, Universidade do Porto, Rua das Estrelas, 4150-762 Porto, Portugal 11 Departamento de Física e Astronomia, Faculdade de Ciências, Universidade do Porto, Rua do Campo Alegre, 4169-007 Porto, Portugal 12 Canada-France-Hawaii Telescope Corporation, 65-1238 Mamalahoa Hwy, Kamuela, HI 96743, USA 13 Leiden Observatory, Leiden University, Postbus 9513, 2300 RA Leiden, The Netherlands Received 30 January 2018 / Accepted 27 June 2018

ABSTRACT

We report the detection of two exoplanets and a further tentative candidate around the M-dwarf stars Gl96 and Gl617A, based on radial velocity measurements obtained with the SOPHIE spectrograph at the Observatoire de Haute-Provence. Both stars were observed in the context of the SOPHIE exoplanet consortium’s dedicated M-dwarf subprogramme, which aims to detect exoplanets around nearby M-dwarf stars through a systematic survey. For Gl96 we present the discovery of a new exoplanet at 73.9 d with a minimum mass of 19.66 earth masses. Gl96 b has an eccentricity of 0.44, placing it among the most eccentric planets orbiting M stars. For Gl617A we independently conﬁrm a recently reported exoplanet at 86.7 d with a minimum mass of 31.29 earth masses. Both Gl96 b and Gl617A b are potentially within the habitable zone, although the high eccentricity of Gl96 b may take it too close to the star at periapsis. Key words. techniques: radial velocities – planetary systems – stars: late-type – stars: individual: Gl617A – stars: individual: Gl96

1. Introduction habitable zone is located closer to the stars. Hence, low-mass short-period habitable planets are easier to detect for M-dwarfs M-dwarf stars are both interesting and promising targets for exo- than for sun-like stars. Some examples are Gl 667C c, the ﬁrst planet hunts. They are the most common stars in the Galaxy, habitable-zone Earth-size planet around an M-dwarf (Delfosse and studies suggest their planet occurrence rates are high (e.g. et al. 2013); LHS 1140 b, one of the most recently detected plan- Bonﬁls et al. 2013; Dressing & Charbonneau 2015). Moreover, ets, which orbits the brightest M-dwarf with a transiting planet they are interesting candidates for habitable planet searches. in the habitable zone (Dittmann et al. 2017); TRAPPIST-1 e, f, g, Their relatively small masses (0.07 0.6 M ; Reid & Hawley − three potentially habitable telluric planets in a seven-planet sys- 2005) mean that small planets will still induce detectable sig- tem (Gillon et al. 2017); GJ 273, with two super-Earths one of nals, while their faintness compared to G-type stars means the which is in the habitable zone (Astudillo-Defru et al. 2017c); ? Based on observations collected with the SOPHIE spectrograph on and K2-18b, a transiting habitable-zone planet whose mass was the 1.93 m telescope at the Observatoire de Haute-Provence (CNRS), characterised by RV follow-up, revealing a density that may cor- France, by the SOPHIE Consortium. respond to a rocky planet with extended atmosphere or to a water ?? Full Tables A.1 and A.2 (RV tables) are only available at the CDS world (Foreman-Mackey et al. 2015; Cloutier et al. 2017). via anonymous ftp to cdsarc.u-strasbg.fr (130.79.128.5) or via Currently, 146 exoplanets around main sequence M-dwarf http://cdsarc.u-strasbg.fr/viz-bin/qcat?J/A+A/618/A103 stars are known, compared to 997 planets around FGK

A103, page 1 of 15 Open Access article, published by EDP Sciences, under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. A&A 618, A103 (2018) stars1. However, this number is expected to grow as several cur- median exposure time of 900 s and a median S/N at 550 nm of rent or near-future projects have M-dwarf stars as part of their 84.1. (or their sole) primary targets, e.g. SPIRou (Artigau et al. 2014), The SOPHIE pipeline (Bouchy et al. 2009a) was used to TESS (NASA mission, launch 2018; Ricker 2016), TRAPPIST reduce and extract the spectra. The spectra were then cross- (e.g. Gillon et al. 2017), CARMENES (e.g. Quirrenbach et al. correlated with an M3 stellar spectral mask in order to obtain 2014, 2016), HADES (e.g. Affer et al. 2016), NIRPS (Bouchy the cross-correlation functions (CCFs), from which radial veloc- et al. 2017), and ExTrA (Bonfils et al. 2015). Given the relatively ities (RV), FWHM, contrast, and bisectors were measured. The low number of detected exoplanets around M-dwarfs, each new mask was built from the median of a large number of spectra of detection provides valuable information on the population that Gl581 obtained with HARPS (La Silla, ESO) and degraded to can be used to refine observing strategies. the SOPHIE spectral resolution. Since 2006, the SOPHIE exoplanet consortium has been car- For Gl617A, we removed nine spectra with S/N < 35. For rying out several planet-hunting programmes using the SOPHIE Gl96, we removed a total of eight observations: two spectra fol- spectrograph at the Observatoire de Haute-Provence (Bouchy lowing the same S/N criteria as for Gl617A, and six spectra due et al. 2009a). Subprogramme 3, or SP3, consists of a systematic to moon contamination. survey of nearby M-dwarfs whose aims are to detect habitable SuperEarths and Neptunes, to constrain the statistics of planets 3. Data analysis around M-dwarfs, and to find potentially transiting companions. A complete description of the SP3 is beyond the scope Radial velocities for stabilised spectrographs with ThAr-derived of this paper; the programme will be presented in detail in a wavelength calibration are usually obtained by the CCF method, forthcoming paper. in which the spectra are correlated with a weighted binary mask In this work we report the results of the SP3 study of two (see Queloz 1995 and Pepe et al. 2002 for full descriptions M-dwarfs, Gl96 and Gl617A. Section2 presents the observa- of the method). Although the CCF method is very effective tions. In Sect.3 we describe the analysis of the data. Our for FGK stars, which have strong spectral lines and a well- results are presented in Sect.4, together with an analysis of defined continuum, this is not the case for M-dwarfs, where the HIPPARCOS photometry of these stars in Sect.5, and are numerous overlapping molecular bands complicate the contin- discussed in Sect.6. uum determination. For these stars, the use of these binary masks which target only clearly defined lines under-utilises the Doppler information present in the spectrum. Therefore, other methods 2. Observations have been developed to better exploit this information, such as template-matching using a true stellar template (as done by The M-dwarf stars Gl96 and Gl617A were observed with the HARPS-TERRA code of Anglada-Escudé & Butler 2012). the SOPHIE spectrograph as part of the SOPHIE consortium Template-matching also allows a more precise removal of the search for exoplanets around M-dwarfs. SOPHIE is a fibre- telluric lines, and of any parts of the spectra that are not com- fed, environmentally stabilised, cross-dispersed echelle spectro- pliant with the template or that have no spectral information. In graph mounted on the 193 cm telescope at the Observatoire de this work, we make use of an algorithm developed by Astudillo- Haute-Provence (Perruchot et al. 2008). In 2011, SOPHIE was Defru for this purpose (Astudillo-Defru et al. 2015, 2017c); it upgraded by inserting an octagonal-section fibre in the fibre constructs stellar and telluric templates from the observed spec- link (Perruchot et al. 2011; Bouchy et al. 2013). The upgraded tra, discards the telluric-contaminated zones, and derives the spectrograph is known as SOPHIE+. radial velocity by χ2 minimisation, using the RV determined by The observations presented here were performed after the the CCF method as a first guess. We applied this method to all 2011 upgrade, using the high-resolution (HR) mode of the spec- the SOPHIE+ spectra. trograph for a resolving power of λ/δλ 75 000. The SOPHIE Before employing the template-matching procedure, we per- spectrograph provides two modes for fibre≈ B: thosimult mode, formed a correction for the charge transfer inefficiency (CTI) in which a simultaneous ThAr calibration is performed and effect, following the characterisation of Bouchy et al.(2009b). is used to trace the spectrograph drift during the night, and Once the RVs were determined, we added a further correction objAB mode, in which fibre B is used to monitor the sky bright- for the instrumental drift using the simultaneous ThAr calibra- ness. The choice of modes for the SP3 targets depends on the tion (for Gl617A) or an interpolation between ThAr calibrations brightness of the stars (with a limit at V = 9). Gl96, which is performed before and after the exposure (for Gl96). above this limit, was observed in the objAB mode to control The SOPHIE+ spectrograph also presents long-term vari- for possible moonlight contamination and to avoid any potential ations of the zero-point, an effect described in Courcol et al. ThAr contamination. A calibration lamp spectrum was obtained (2015). In order to correct for these variations, the authors itera- immediately before each observation to monitor potential drifts. tively construct a master RV time series from RV constant stars. Gl617A, which is brighter, was observed in the thosimult mode This master series is used to correct the measurements of a given since it is bright enough that the contamination by moonlight target by subtracting the velocity in the master series, interpo- can be neglected and contamination from the ThAr spectrum on lated at the dates of observation. We followed the procedure fibre B is negligible. described in Courcol et al.(2015) to construct an analogous mas- Observations were gathered between 2011 and 2017. In total, ter for the SP3 programme, using the ensemble of SP3 targets 79 spectra were obtained for Gl96 with SOPHIE+, with a median with at least ten SOPHIE+ observations plus the four “super- exposure time of 1800 s and a median S/N at 550 nm of constants” from the SP1 programme (defined by Courcol et al. 82.7. A total of 163 spectra were obtained for Gl617A, with a 2015) as a starting point. We chose to use primarily SP3 targets 1 Retrieved on 11 Oct 2017, from The Exoplanet Encyclopaedia, when constructing our master correction in order to mitigate any 1995–2017, considering only the stars for which spectral type is reported potential bias or offset due to differing spectral types. The final in the catalogue, and filtering out those not on the main sequence. master employs 14 stars: the SP1 (G-type) constants HD185144,

A103, page 2 of 15 M. J. Hobson et al.: The SOPHIE search for northern extrasolar planets. XIII.

Table 1. Stellar parameters.

Parameter Gl617A Gl96 Spectral type M1a M2a V 8.896b 9.345b B-V 1.010b 1.519b V-K 3.943c 3.791c Mass [M ] 0.60 0.07d 0.60 0.07d ± ± Π [mas] 93.15 0.23e 83.75 1.14 f log(R ) 4.75± 0.14 4.77± 0.06 HK0 − ± − ± Prot [d] 28.8 6.1 29.6 2.8 ± g ± h Teff [k] 4156 73 3785 62 ± g ± g L? [L ] 0.1069 0.0153 0.0888 0.0135 Fe/H [dex] 0.19 ± 0.08g 0.14 ± 0.08g ± ± Notes. (a)Gaidos et al.(2014). (b)Zacharias et al.(2012). (c)Zacharias et al. (2012); Cutri et al.(2003). (d)Delfosse et al.(2000). (e)Gaia Collabora- Fig. 1. Zero-point drift correction (red line) and radial velocities of the tion(2016). ( f )van Leeuwen(2007). (g)Gaidos & Mann(2014). (h)Mann 14 stars used to construct it (black dots). The vertical lines indicate et al.(2015). identified jumps and their causes (see Sect.3 for details). The zero- 1 point correction spans 6 years; it has a dispersion of 3.15 m s− , and a 1 indicators of chromospheric activity are the Hα index, which peak-to-peak variation of 16.9 m s− . measures the flux in the Hα line, and the log(RHK0 ) index, which is based on the flux in the Ca II H and K emission lines. We HD9407, HD22154, and HD89269A; three M-dwarfs which are computed the Hα index following the definitions of Boisse et al. systematically observed for all observation seasons and consid- (2009). For the Ca lines, we followed Boisse et al.(2010) to ered as our SP3 constants, Gl411, Gl514, and Gl686; and the obtain the S -index scaled to Mount Wilson values. The log(RHK0 ) additional SP3 stars G239-25, Gl133, Gl15A, Gl436, Gl521, index was originally defined by Noyes et al.(1984), using a pho- Gl694, and Gl728 (all additional stars used for the constant cor- tometric correction based on the B-V index. However, in that rection have a corrected rms after the first iteration lower than work the conversion from the S -index to log(RHK0 ) was not cal- 1 3 m s− , as defined by Courcol et al. 2015). We also tested a mas- ibrated for redder M-dwarfs, and the B-V index is not ideal for ter constructed using only the SP3 targets, but found the resulting M-dwarfs which are too faint in the B band. Therefore, we used 1 RVs had higher dispersion (around 1.5 m s− higher on average). the calibrations of Astudillo-Defru et al.(2017a) for M-dwarfs to Figure1 shows the master used for the zero-point drift correction calculate the log(RHK0 ) index employing V-K colours. and the RVs from which it was derived. Gomes da Silva et al.(2011) carried out a study of activ- The zero-point drift correction clearly reflects several instru- ity indices for M-dwarfs. They found that in addition to the Hα ment modifications: a jump at 55872 (6 Nov. 2011) corre- and log(RHK0 ) indices, the Na I D1 and D2 lines correlate well sponding to a ThAr calibration lamp change; a jump at 56274 with stellar activity in these stars. Therefore, we also calculated (12 Dec. 2012) following the installation of octagonal fibres after the NaI index as defined by the authors. The CCF bisector is the double scrambler; a second ThAr lamp-change related jump also known to correlate with stellar activity for short rotational at 56690 (01 Feb. 2014); a jump at 56730 (13 Mar. 2014) after the periods. We obtained the bisector for our observations from the installation of a new calibration unit; a jump at 56941 (10 Oct. SOPHIE pipeline. 2014) due to a change in the current of the ThAr calibration lamp; and a jump at 57435 (16 Feb. 2016) corresponding to the instal- 3.2. Stellar parameters lation of a new thermal regulation. These events are indicated in Fig.1. An additional long-term drift is also seen. The stellar parameters are listed in Table1. Spectral types were obtained from Gaidos et al.(2014); masses from Delfosse et al. The first two events described were also noted by Courcol (2000); metallicities and luminosities from Gaidos & Mann et al.(2015). We do not, however, find the jump at 56775 (2014); and temperatures from Mann et al.(2015) when available which the authors correlated at the time to the recoating of the and from Gaidos & Mann(2014) when they were not. Magni- secondary mirror (including when we regarded a test master con- tudes and colour indices were taken from Zacharias et al.(2012), structed using only the template-matching derived RVs of the except for the K magnitude which is from Cutri et al.(2003). SP1 super-constants employed by Courcol et al. 2015). We have For Gl617A, a Gaia DR1 parallax is available, while for Gl96 identified the actual cause of the jump at 56775 as an instability we take the HIPPARCOS parallax. The mean and standard devia- of the wavelength solution on blue spectral orders, whose weight tion of log(R ) were calculated from the SOPHIE spectra. We changed after mirror recoating. As this effect is not present in the HK0 used the log(R ) log(P ) relation from Astudillo-Defru et al. template-matching procedure, the jump no longer appears. HK0 rot (2017a) to estimate− the rotation period from the mean log(R ), The full corrected radial velocity sets for Gl96 and Gl617A HK0 with error bars calculated by propagation. are given in Tables A.1 and A.2, respectively (available at the CDS). 3.3. Radial velocity analysis 3.1. Activity indicators To analyse the radial velocities, we employed the Data and Anal- ysis Center for Exoplanets (DACE) web platform2, which is In order to study the stellar activity of our targets, we calculated several indicators. Two of the most widely used spectral 2 Available at https://dace.unige.ch.

A103, page 3 of 15 A&A 618, A103 (2018)

Fig. 2. Time series of the radial velocities calculated with template- matching for Gl96 from the SOPHIE+ measurements. based at the University of Geneva. This platform allows the generation of generalised Lomb–Scargle periodograms and the calculation of false alarm probabilities (FAP). Signals can be ﬁt by Keplerian models (DACE employs the formalism of Delisle et al. 2016 for this purpose), and polynomial drifts and stellar jitter can be added. DACE also provides a Markov chain Monte Carlo (MCMC) analysis facility, described in Díaz et al. (2014, 2016b).

4. Results 4.1. Gl96 Fig. 3. Periodogram of (top panel) the radial velocities calculated with template-matching for Gl96 from the SOPHIE+ measurements, cor- The RVs calculated with template-matching from the SOPHIE+ rected from the zero-point drift; (middle panel) the time series of the observations of Gl96 were analysed using DACE. The time series master correction for the zero-point drift applied; (bottom panel) the and periodogram of the Gl96 RVs are shown in Figs.2 and3, uncorrected radial velocities for Gl96 (prior to the application of respectively. Figure3 also shows the periodogram of the zero- the master correction). The horizontal lines correspond to 50%, 10%, point correction applied, and of the data prior to this correction. and 1% FAP, respectively. In Fig.4 we show periodograms of the activity indicators H α, log(RHK0 ), NaI, and the CCF bisector. We used DACE to carry out an MCMC analysis of the single- The RV periodogram shows a peak at 75d below 1% FAP, planet model for Gl96 in order to better constrain the parameters. which bootstrap resampling places below 0.05% FAP. There The results are summarised in Table2. Figure8 shows the phase- is no corresponding peak for any of the activity indicators, folded data points. The best-fit solution results in a rather highly nor does the zero-point drift correction applied show any sig- eccentric orbit. In order to analyse whether contamination from nal at this period. Additionally, the signal also appears in the stellar-activity driven RV variations influences the results, we uncorrected time series periodogram. We also applied an l1 tested two approaches: the addition of a Keplerian fit to the periodogram, as defined by Hara et al.(2017); this technique 29d peak, and a red-noise model. The two-Keplerian model searches for a number of signals simultaneously using com- did not modify the planetary parameters greatly; in particular, +0.22 pressed sensing techniques. As such, it is much less prone the resulting eccentricity is of 0.46 0.14, which is indistinguish- to aliases and other problems arising in the traditional peri- able from the one-Keplerian result− within the error bars. The odogram. The resulting periodogram is shown in Fig.6, where red noise was modelled using a Gaussian process with a quasi- the signal at 74 days can clearly be seen to dominate the data; periodic kernel (for details of the model, see e.g. Astudillo-Defru the FAP of this signal is conservatively estimated (using an ana- et al. 2017b). We included the rotational period of the star in lytical formula from Baluev 2008) as log10(FAP) = 2.6072. the model by using an informative prior on the correspond- The remaining signals are consistent with the stellar− rotation ing hyper-parameter. The posterior distribution of this parameter period and half this period, suggesting they originate in stellar is narrower, and we find Prot = 28.4 1.4 days, where the activity. uncertainties correspond to the 1σ credible± interval. For the We fit this signal by a Keplerian model with DACE. The planet eccentricity, the maximum a posteriori estimate is 0.50, highest peak in the residuals, at 29d, is only below 50% FAP, in agreement with the model without red noise. The inclusion of while a second peak at 14d at similar FAP corresponds to half correlated noise, on the other hand, seems to allow for lower val- this period (Fig.5). Additionally, the highest peak is close to ues of the eccentricity: the 95% highest density interval extends the peaks seen at 28–29d, well below 1% FAP in the peri- between 0.0 and 0.68 (Fig.7), showing that in the presence of odograms of the Hα and log(RHK0 ) indices. Furthermore, these red noise the eccentricity is effectively unconstrained. periods are consistent with the estimated stellar rotation period To quantify the significance of the detection, we estimated of 29.6 2.8d (see Sect. 3.2). Therefore, we cannot justify the posterior odds ratio (POR) for several competing models. treating it± as a potential second planet. If we assume equal prior probability for all models, the POR

A103, page 4 of 15 M. J. Hobson et al.: The SOPHIE search for northern extrasolar planets. XIII.

Hα index Gl96 l1-periodogram 1.6 Candidates tested at: 73.8511 13.2861 28.3776 1.1994 days log10(FAP) : -2.6072 0 0 0 1.4

1.2

0.8

0.6 Amplitude (m/s) 0.4 log (R'HK) index 0.2

0 101 102 103 Period (days)

Fig. 6. l1 periodogram of the SOPHIE RVs for Gl96. The signal at 74 days is clearly predominant, while the other two peaks are probably related to activity. Na index

1.6

1.4

1.2 CCF bisector 1.0

0.8

0.6 Probability density

0.4

0.2

Fig. 4. Periodograms of activity indicators for Gl96 – from top to bottom 0.0 0.0 0.2 0.4 0.6 0.8 1.0 panels:Hα index, log(RHK0 ) index, NaI index, and CCF bisector. The horizontal lines correspond to 50%, 10%, and 1% FAP, respectively. For Orbital eccentricity the NaI index and CCF bisector, no horizontal lines are visible because the entire periodogram is beneath the 50% FAP line. Fig. 7. Posterior probability of the orbital eccentricity of Gl96 b under the model including correlated noise.

Table 2. Best-ﬁt solution for the planetary system orbiting Gl96.

Parameter Units Gl96 b +0.33 P [d] 73.94 0.38 1 +−0.72 K [m s− ] 4.69 0.62 −+0.09 e 0.44 0.11 −+12.45 ω [deg] 339.58 14.52 − Fig. 5. Periodogram of the residuals of a Keplerian fit with P = 75d to T [d] 55556.39+10.57 the radial velocities calculated with template-matching for Gl96 from P 8.98 +−11.56 the SOPHIE+ measurements. The horizontal line corresponds to 50% TC [d] 55568.90 9.82 − FAP level. +0.005 Ar [AU] 0.291 0.005 −+2.42 reduces to the ratio of marginal likelihoods, i.e. the Bayes fac- M sin i [MEarth] 19.66 2.30 · − tor (BF). We used the estimation introduced by Perrakis et al. γ [m s 1] 37874.84+0.31 (2014), based on importance sampling, to compute the marginal SOPHIE − − 0.32 σ 1 +0.93− likelihoods for a model without a Keplerian (k0), a model with JIT [m s− ] 3.45 0.91 1 − a quadratic long-term trend (k0d2), and a model with a sin- σ(O C) [m s− ] 3.37 − gle Keplerian (k1). All three models included an additional +1.68 log (Post) 196.77 2.46 white noise component whose amplitude was an additional nui- − − sance parameter. The results are shown in Fig.9 as a function of Notes. For each parameter the median of the posterior is reported, with the sample size used for the estimation. As expected, the Perrakis error bars computed from the MCMC chains using a 68.3% confidence estimator is biased, but for sample sizes larger than around interval. σO C corresponds to the weighted standard deviation of the 3000, the bias is negligible. We find logarithmic Bayes factors residuals around− the best solutions. log (Post) is the posterior likelihood. log(BF ) = 5.21 0.06 and log(BF ) = 7.91 0.06 for All the parameters probed by the MCMC can be found in Table B.1. k1;k0 ± k1;k0d2 ±

A103, page 5 of 15 A&A 618, A103 (2018)

Fig. 8. Phase-folded radial velocities of Gl96 for a one-planet model with P = 74.7d.

N(5.21,0.06) N(7.91,0.06) BF(k1,k0) BF(k1,k0d2) 200 − Fig. 10. Calculated orbit for Gl96 b with respect to the optimistic and conservative habitable zones, as deﬁned by Kopparapu et al.(2013b). 205 −

210 k1 −200 k0 − 200 200 k0d2

Log Marginal likelihood −215 −− 205 − 205 1 2 3 4 205− 10 10 10 10 5.25 5.50 7.75 8.00 8.25 − 210 Sample size k1 Log Bayes Factor − 210 k1 210 k0k1 − k0 − k0d2k0

Log Marginal likelihood 215 Fig.− 9. Left panel: marginal likelihoodk0d2 estimated by the Perrakis method Log Marginal likelihood 215 k0d2

Log Marginal likelihood 215− vs. sample− 101 size for102 three10 models:3 10 a4 single Keplerian plus white noise 101 102 103 104 (k1; blue),101 a quadratic10Sample2 drift size103 plus10 white4 noise (k0d2; green), and pure Sample size white noise (k0; orange).Sample size The one-Keplerian model is clearly favoured by the data. The error bars correspond to the 95% confidence interval. Right panel: histogram of 1000 Monte Carlo realisations of the Bayes factor between model k1 and k0 (orange) and k0d2 (green); the solid curves are normal distributions with the mean and variance equal to those of the Monte Carlo sample. the comparison between k1 and k0, and k1 and k0d2, respec- Fig. 11. Time series of the radial velocities calculated with template- tively, where the reported values are the empirical means and matching for Gl617A from the SOPHIE+ measurements. standard deviations obtained by repeating the calculation for each model 5000 times, and drawing 1000 random pairs. The present an independent detection of this planet from our SOPHIE resulting distributions are very approximately normal. The anal- data, and describe a further potential candidate at 500 d. ysis shows therefore that the posterior probability of a model We analysed the RVs calculated by template-matching from with a single Keplerian is much higher than any of the com- the SOPHIE+ observations of Gl617A with the DACE plat- peting models. This is so in spite of the strong penalisation the form. The time series and periodogram of the Gl617A RVs are Bayes factor gives to more complex models. shown in Figs. 11 and 12, respectively. Figure 12 also shows the One of the goals of the SP3 programme is to detect habitable periodograms of the zero-point correction applied and of the planets around M-dwarf stars. The habitable zone calculator3 uncorrected data (i.e. the Gl617A RVs prior to the application based on the work of Kopparapu et al.(2013b) places Gl96 b of this correction). inward of the conservative HZ, though within the optimistic one; In Fig. 13 we show periodograms of the activity indica- however, due to its high eccentricity it would probably move tors Hα, log(RHK0 ), NaI, and the CCF bisector. The Hα index inward of the optimistic HZ at periapsis (Fig. 10). To better exhibits a periodicity at around 21.8 days, probably related to the quantify the habitability of this eccentric planet, we follow the rotational period of the star, as well as long-period peaks. The method employed by Díaz et al.(2016a), who calculated the log(RHK0 ) and NaI periodograms present only very long-period mean incident flux over an orbit as defined by Williams & Pollard peaks, in the 500–1000 day range. The CCF bisector shows a (2002), and compared it with the limits given by Kopparapu et al. small peak at 10.9 days, which is close to half the rotation period (2013a). For Gl96 b, the mean incident flux is F /F = 1.168, and therefore is probably due to stellar activity. h i ⊕ placing it between the recent Venus and runaway greenhouse The RV periodogram shows a very strong signal at 86d well limits. below 1% FAP, which bootstrap resampling places below 0.01% FAP, and which is not present in any of the activity indicators 4.2. Gl617A or in the periodogram of the zero-point correction applied. We employed DACE to fit it with a Keplerian model. Figure 14 shows A planet at 86.54 d around this star was recently announced by the periodogram of the residuals. There are two signals below the CARMENES team (Reiners et al. 2018). In this section, we 10% FAP: one at 21d and one at around 500d, and one further 3 Available at https://depts.washington.edu/naivpl/ signal below 50% FAP at 29d. The 21d signal is probably due content/hz-calculator to stellar activity, as it coincides with a signal below 1% FAP

A103, page 6 of 15 M. J. Hobson et al.: The SOPHIE search for northern extrasolar planets. XIII.

Hα index

log(R'HK) index

Na index

CCF bisector

Fig. 12. Periodograms: radial velocities calculated with template- matching for Gl617A from the SOPHIE+ measurements, corrected for the zero-point drift (top panel); middle panel: master correction for the zero-point drift applied; bottom panel: uncorrected radial veloci- Fig. 13. Periodograms of activity indicators for Gl617A. From top to ties for Gl617A (prior to the application of the master correction). The bottom panels:Hα index, log(RHK0 ) index, NaI index, and CCF bisector. horizontal lines correspond to 50%, 10%, and 1% FAP, respectively. The horizontal lines correspond to 50%, 10%, and 1% FAP, respectively. in the Hα periodogram, and is close to the estimated rotation period (Table1), although there is no signal at 21d in the other activity indicators. The 29d signal is removed by the addition of a quadratic drift, suggesting it may be an artefact of the window function (although the period may also point to moon contamination; as this star is observed with simultaneous wavelength calibration, we do not have a measure of the sky). The long-period signal is intriguing; the addition of a linear or quadratic drift does not affect it, whereas a test Keplerian model removes it completely and results in a low-eccentricity Fig. 14. Periodogram of the residuals of a Keplerian fit with P = 86d to Keplerian fit. While the periodograms of the Hα, log(RHK0 ), and NaI indices show long-period peaks, they are not well fitted the radial velocities calculated with template-matching for Gl617A from by the model obtained from the RVs. More concerning is the the SOPHIE+ measurements. The horizontal lines correspond to 50% and 10% FAP, respectively. periodogram of the master correction, which shows a peak at 500 days; nevertheless, we note that the long-period signal is also present in the uncorrected RVs (see Fig. 12). However, we stress The planet at 86d is consistent in both models, with prac- that this signal is of only moderate significance. We therefore tically all parameters indistinguishable within the error bars present it as a tentative detection to be confirmed, not a definite between the two best-fit solutions. The two-Keplerian model planet. actually provides a better fit to the data, with a lower σ(O C) As was done for Gl96, we also applied an l1 periodogram, and BIC. Figures 16 and 17 show the phase-folded RVs− for which is shown in Fig. 15; the signal at 86 days dominates the Gl617A b and Gl617A c, respectively, using the parameters data, with log10(FAP) = 17.1869. The remaining signals at 21d derived from the two-Keplerian model. In the one-Keplerian and 500d are consistent− with the discussion above. model, the drift is significant; removing it slightly decreases In∼ order to explore the parameter space more thoroughly, we the amplitude and increases the eccentricity of the Keplerian employed the MCMC sampler implemented in DACE. We tested fit. In the two-Keplerian model, however, the drift is only two models: a one-Keplerian model at 86d, and a two-Keplerian marginally significant. This is coherent with the fact that in model with initial periods at 86d and 500d, both including a the one-Keplerian model, the drift may attempt to absorb the quadratic drift. The resulting best-fit solutions are summarised long-period signal that is fitted by a Keplerian in the second in Table3. model.

A103, page 7 of 15 A&A 618, A103 (2018)

Table 3. Best-ﬁt solutions for the planetary system orbiting Gl617A: one-Keplerian and two-Keplerian models plus drift.

Parameter Units Gl617A ba Gl617A bb Gl617A c +0.18 +0.20 +35.45 P [d] 86.93 0.19 86.72 0.18 496.90 21.82 1 +−0.41 +−0.36 +−0.43 K [m s− ] 6.56 0.39 6.57 0.38 3.16 0.42 −+0.08 −+0.07 −+0.15 e 0.32 0.09 0.23 0.08 0.15 0.10 −+10.04 −+13.55 −+980.42 ω [deg] 102.20 11.48 97.25 13.46 311.97 63.26 − +4.75 − +4.97 − −+146.26 TP [d] 55466.60 4.32 55470.17 4.85 55214.75 182.87 −+4.47 −+4.30 −+109.85 TC [d] 55465.26 4.08 55469.18 4.41 55275.54 121.34 − − − +0.006 +0.006 +0.051 Ar [AU] 0.324 0.006 0.323 0.005 1.036 0.036 −+2.10 −+2.20 −+3.84 M sin i [MEarth] 30.56 2.12 31.29 2.15 27.26 3.72 · − − − 1 +5.38 +5.58 γSOPHIE [m s− ] 18737.35 5.74 18715.26 5.63 1 1 − +2.42− − +2.−65 lin [m s− yr− ] 7.71 2.3 3.13 2.34 1 2 −+0.23 − +−0.27 quad [m s− yr− ] 0.86 0.25 0.30 0.26 1 − +−0.56 +−0.49 σJIT [m s− ] 4.86 0.62 3.71 0.46 1 − − σ(O C) [m s− ] 4.55 3.90 − log (Post) 466.65+1.82 440.08+2.28 − 2.56 − 3.07 BIC 105.93− 118.52−

Notes. (a)Parameters obtained from the one-Keplerian model for Gl617A b. (b)Parameters obtained from the two-Keplerian model for Gl617A b. For each parameter, the median of the posterior is reported, with error bars computed from the MCMC chains using a 68.3% conﬁdence interval. σO C corresponds to the weighted standard deviation of the residuals around this best solution. log (Post) is the posterior likelihood. All the parameters− probed by the MCMC can be found in Table B.3 (one-Keplerian model) and Table B.5 (two-Keplerian model).

Gl617A l1-periodogram 5 Candidates tested at: 86.57387 21.09149 561.1842 49.80399 days log10(FAP) : -17.1869 -3.72901 -0.754693 0 4

2 Amplitude (m/s)

0 101 102 103 Period (days)

Fig. 15. l1 periodogram of the SOPHIE RVs for Gl617A. The signal at 86 days is clearly predominant, while that at 21 days is probably related Fig. 16. Phase-folded radial radial velocities velocities of of Gl617A Gl617A for for a aP P==86d planet, to activity. The 500d signal is discussed in the text. ∼ using the parameters derived from the two-Keplerian model.

We compared the parameters obtained by our analysis with those recently presented by Reiners et al.(2018). The orbital period and semi-major axis are compatible within error bars; however, we ﬁnd a somewhat larger mass and distinctly larger eccentricity from our data. The mass reported by Reiners et al. +2.4 (2018) of M sin i = 24.7 1.8 MEarth is compatible with our calculated mass· at 2σ. This− is true for both the one-Keplerian and the two-Keplerian models, with the eccentricity higher for the one-Keplerian model (see Table3). As was done for Gl96, we employed the habitable zone calculator to estimate the location of the HZ for Gl617A. The orbits calculated by the MCMC analysis place Gl617A b at 0.32 AU, closer to the star than the conservative inner limit, but within the optimistic HZ (Fig. 18). This differs from the results of Reiners et al.(2018). As Gl617A b is moderately eccentric, we Fig. 17. Phase-folded radial velocities of Gl617A for a P = 497d planet, also calculate the mean incident ﬂux (as was done for Gl96 b) using the parameters derived from the two-Keplerian model.

A103, page 8 of 15 M. J. Hobson et al.: The SOPHIE search for northern extrasolar planets. XIII.

Table 4. Best-ﬁt solution for the planetary system orbiting Gl617A, from SOPHIE, CARMENES, and KECK combined data.

Parameter Units Gl617A b +0.16 P [d] 86.78 0.15 1 +−0.22 K [m s− ] 5.83 0.24 +−0.04 e 0.07 0.04 −+30.39 ω [deg] 97.22 41.83 − +8.52 TP [d] 55468.26 10.67 −+4.12 TC [d] 55466.91 4.11 − +0.005 Ar [AU] 0.323 0.006 −+1.49 M sin i [MEarth] 28.55 1.45 · − 1 +1.64 γCARMENES [m s− ] 0.40 1.60 1 −+2.63 γKECK PUB [m s− ] 1.76 Fig. 18. Calculated orbit for Gl617A b with respect to the optimistic and − 1 2.68 1 − − +1.54 conservative habitable zones, as deﬁned by Kopparapu et al.(2013b). γSOPHIE [m s− ] 18721.59 1.50 1 − +4.14− σJIT [m s− ] 15.89 1.82 1 − σ(O C) [m s− ] 3.87 − log (Post) 939.95+2.11 − 2.76 BIC 1668.19−

Notes. Same notes as Table3. All the parameters probed by the MCMC can be found in Table B.7.

resulting best-fit solution is summarised in Table4, and the phase-folded combined data is shown in Fig. 20. We compared the parameters resulting from the analysis of the combined data with those we obtained from SOPHIE data alone, and with those presented by Reiners et al.(2018). In all three cases, the orbital periods are compatible at the 1σ level (though slightly larger for the SOPHIE and combined data). M sin i is also compatible at the 1σ level between SOPHIE data alone· and the combined data, but Reiners et al.(2018) present +2.4 a slightly smaller value of M sin i = 24.7 1.8 MEarth, only compatible with the others at the· 2σ level. The− eccentricity is also Fig. 19. Periodograms: combined SOPHIE, CARMENES, and KECK different; the value derived from the SOPHIE data alone is radial velocities of Gl617A (top panel); bottom panel: residuals of a larger, while those from the combined data and from Reiners Keplerian fit with P = 86d to the data. The horizontal lines correspond et al.(2018) are compatible at 1 σ. The amplitude of the sig- to 50%, 10%, and (top panel only) 1% FAP, respectively. nal thus differs, being largest for the SOPHIE data alone and smallest in the Reiners et al.(2018) analysis; this is consistent in order to better quantify its habitability. The mean incident with the fact that for fixed e, K grows with M sin i, and for fixed flux is F /F = 1.053, placing it between the recent Venus M sin i K grows with e (see e.g. Seager 2010).· h i ⊕ and runaway greenhouse limits as defined by Kopparapu et al. · Recently, Feng et al.(2018) published an analysis of this (2013a). system in a Research Note of the AAS, also employing the CARMENES, HIRES, and SOPHIE data. Their use of our Combination with CARMENES and KECK data SOPHIE data, however, is biased; they do not take into account the CTI effect and zero-point drift described in Sect.3 that we We combined our observations of Gl617A with the CARMENES correct for in this work. data presented by Reiners et al.(2018) and the KECK data of Butler et al.(2017). Figure 19 shows the resulting periodogram, 5. Photometry with a strong signal at 86d, to which we fit a Keplerian model using DACE; and the periodogram of the residuals of this fit. Stellar activity can be reflected not only in the radial velocity The only strong signal is at 21d, and presumably corresponds to but also in the photometric observations of a star, where we also the stellar activity as discussed previously. The 500d signal that hope to find signals linked to the rotation period. In order to is present in the SOPHIE data is not in evidence in the combined analyse whether this is the case for our targets, we obtained the observations, suggesting that it may be spurious. HIPPARCOS photometry for both from ESA(1997). The posterior distribution of the model parameters was sam- For Gl96, we retrieved 121 measurements over a time span pled using an MCMC algorithm implemented in DACE. A of three years. Figure 21 shows the photometric data points and one-Keplerian model with a quadratic drift was used. The the corresponding periodogram. No signals below 50% FAP can

A103, page 9 of 15 A&A 618, A103 (2018)

8.74 8.748.74 8.72 8.728.72 8.70 8.708.70 8.68 8.688.68 8.66 8.668.66 8.64 8.648.64 Hipparcos magnitude 8.62 Hipparcos magnitude Hipparcos magnitude 8.628.62 8.60 8.608.60 8.58 8000 8200 8400 8600 8800 8.588.58 80008000 82008200Date (BJD-2,440,000.0)84008400 [d] 86008600 88008800 DateDate (BJD-2,440,000.0) (BJD-2,440,000.0) [d] [d] Fig. 20. Phase-folded radial velocities of of Gl617A Gl617A for for a aP P==86d planet, using the combined SOPHIE, CARMENES, and KECK data.

9.60 9.609.60 9.55 9.559.55 9.50 9.509.50 9.45 9.459.45 Fig. 22. HIPPARCOS photometry for Gl617A (top panel) and its corre- 9.40

Hipparcos magnitude sponding periodogram (bottom panel). 9.40 Hipparcos magnitude 9.40 Hipparcos magnitude 9.35 9.359.35 6. Discussion and conclusions 8000 8200 8400 8600 8800 9000 80008000 82008200 Date 8400(BJD-2,440,000.0)8400 86008600 [d] 88008800 90009000 We have presented the detection of a new Neptune-like exoplanet DateDate (BJD-2,440,000.0) (BJD-2,440,000.0) [d] [d] orbiting the M-dwarf Gl96 and the independent detection of a second Neptune-like exoplanet orbiting the M-dwarf Gl617A for which we refine the planetary parameters, using the SOPHIE+ spectrograph on a 1.93m telescope. The planets have minimum masses of 29 and 31 Earth masses and orbital periods of 74 and 87 days, respectively, and are located close to the inner limit of the HZ. For Gl96 we find no evidence of further planetary companions. Gl617A shows an intriguing signal of moderate significance at 500d in the periodogram that is best fit by a Keplerian model.∼ For Gl617A, we also analysed the combination Fig. 21. HIPPARCOS photometry for Gl96 (top panel) and its corre- of our data with that from CARMENES (Reiners et al. 2018) sponding periodogram (bottom panel). and KECK (Butler et al. 2017). The 87d signal is clear in the combined data, though the resulting planetary parameters dif- be found in the periodogram. There is a small peak at 29d, the fer slightly from those obtained by the SOPHIE data alone. The 500d signal, however, is no longer significant, which suggests period found in the RV, Hα, and log(RHK0 ) periodograms that is presumed to be related to activity, but it is not particularly that it is probably spurious. We may also suspect the influ- relevant in the periodogram of photometric measurements which ence of a magnetic cycle here; complementary observations in is dominated by a forest of peaks at around 2d. This set of peaks polarimetry with SPIRou should help to resolve the question. is probably linked to the sampling; as can be seen in the upper As mentioned in Sect.1, 146 exoplanets around M-dwarf panel of Fig. 21, this star appears to have been observed in groups stars are presently known, of which 75 were detected by radial of several measurements within two days, with the groups set velocities and 33 by transits. The two planets presented here fall weeks or months apart. in the intermediate- to long-period range, and the intermediate- For Gl617A, we retrieved 103 measurements over a time to high-mass range of this sample, as shown in Fig. 23 (median span of 2.5 yr. The photometric data points and the correspond- period = 13.5 d, median mass = 14.3 MEarth). Gl96 b is one of ing periodogram are shown in Fig. 22. No signals below 50% the most eccentric known planets around M-dwarf stars (mean FAP can be found in the periodogram. Nevertheless, it is worth e = 0.12), surpassed only by Wolf 1061 d (Wright et al. 2016) remarking that the two highest peaks are at 10.4d and 20d; these and GJ 317 c (Johnson et al. 2007), with eccentricities of 0.55 values are very close to the 21d period seen in the RV and Hα and 0.81, respectively. periodograms and half this period. Both host stars are metal-rich, as noted in Table1. This For both stars, we note that the HIPPARCOS photometry has a is consistent with a trend found for M-dwarfs hosting plan- relatively high RMS (2.62% for Gl96, 1.97% for Gl617A). Activ- ets to be preferentially metal-rich (e.g. Courcol et al. 2016, ity signals at the level attained by our RV measurements would Hobson et al. 2018). Additionally, the masses determined fall well below this, so any photometric activity tracers may be for the planets are compatible with the upper mass bound- absorbed in the uncertainty of the data. ary determined for Neptune-like planets around M-dwarfs by A103, page 10 of 15 M. J. Hobson et al.: The SOPHIE search for northern extrasolar planets. XIII.

Bonfils, X., Almenara, J. M., Jocou, L., et al. 2015, in Techniques and Instrumen- tation for Detection of Exoplanets VII, Proc. SPIE, 9605, 96051L Bouchy, F., Hébrard, G., Udry, S., et al. 2009a, A&A, 505, 853 Bouchy, F., Isambert, J., Lovis, C., et al. 2009b, EAS Pub. Ser., ed. P. Kern, 37, 247 Bouchy, F., Díaz, R. F., Hébrard, G., et al. 2013, A&A, 549, A49 Bouchy, F., Doyon, R., Artigau, É., et al. 2017, The Messenger, 169, 21 Butler, R. P., Vogt, S. S., Laughlin, G., et al. 2017, AJ, 153, 208 Cloutier, R., Astudillo-Defru, N., Doyon, R., et al. 2017, A&A, 608, A35 Courcol, B., Bouchy, F., Pepe, F., et al. 2015, A&A, 581, A38 Courcol, B., Bouchy, F., & Deleuil, M. 2016, MNRAS, 461, 1841 Cutri, R. M., Skrutskie, M. F., van Dyk, S., et al. 2003, VizieR Online Data Catalog: II/246 Delfosse, X., Forveille, T., Ségransan, D., et al. 2000, A&A, 364, 217 Delfosse, X., Bonfils, X., Forveille, T., et al. 2013, A&A, 553, A8 Delisle, J.-B., Ségransan, D., Buchschacher, N., & Alesina, F. 2016, A&A, 590, A134 Díaz, R. F., Almenara, J. M., Santerne, A., et al. 2014, MNRAS, 441, 983 Díaz, R. F., Rey, J., Demangeon, O., et al. 2016a, A&A, 591, A146 Díaz, R. F., Ségransan, D., Udry, S., et al. 2016b, A&A, 585, A134 Fig. 23. Orbital period vs. minimum mass for planets around M-dwarfs Dittmann, J. A., Irwin, J. M., Charbonneau, D., et al. 2017, Nature, 544, 333 detected by radial velocities or transits. Each point represents a planet, Dressing, C. D., & Charbonneau, D. 2015, ApJ, 807, 45 with Gl96 b indicated by a pentagon and Gl617A b by a diamond. The ESA 1997, The HIPPARCOS and Tycho Catalogues. Astrometric and Photomet- colours indicate the orbital eccentricity. ric Star Catalogues Derived from the ESA HIPPARCOS Space Astrometry Mission, ESA SP, 1200 Feng, F., Jones, H. R. A., & Tuomi, M. 2018, Res. Notes Am. Astron. Soc., 2, 23 Courcol et al.(2016), which corresponds to around 35 Earth Foreman-Mackey, D., Montet, B. T., Hogg, D. W., et al. 2015, ApJ, 806, 215 masses for the metallicities of these stars. Gaia Collaboration (Brown, A. G. A., et al.) 2016, A&A, 595, A2 Gaidos, E., & Mann, A. W. 2014, ApJ, 791, 54 Low-mass planets orbiting M-dwarfs are mainly found in Gaidos, E., Mann, A. W., Lépine, S., et al. 2014, MNRAS, 443, 2561 multi-planet systems. Follow-up observations with SPIRou in Gillon, M., Triaud, A. H. M. J., Demory, B.-O., et al. 2017, Nature, 542, 456 spectropolarimetry will permit us to disentangle the stellar activ- Gomes da Silva, J., Santos, N. C., Bonfils, X., et al. 2011, A&A, 534, A30 ity and planetary signals, refine the mass, and identify possible Hara, N. C., Boué, G., Laskar, J., & Correia, A. C. M. 2017, MNRAS, 464, 1220 Hobson, M. J., Jofré, E., García, L., Petrucci, R., & Gómez, M. 2018, Rev. Mex. additional rocky planets. Astron. Astrofis., 54, 65 Johnson, J. A., Butler, R. P., Marcy, G. W., et al. 2007, ApJ, 670, 833 Acknowledgements. We warmly thank the OHP staff for their support on Kopparapu, R. K., Ramirez, R., Kasting, J. F., et al. 2013a, ApJ, 770, 82 the observations. X.D., X.B., I.B., and T.F. received funding from the Kopparapu, R. K., Ramirez, R., Kasting, J. F., et al. 2013b, ApJ, 765, 131 French Programme National de Physique Stellaire (PNPS) and the Pro- Mann, A. W., Feiden, G. A., Gaidos, E., Boyajian, T., & von Braun K. 2015, ApJ, gramme National de Planétologie (PNP) of CNRS (INSU). N.S. and O.D. 804, 64 were supported by Fundação para a Ciência e a Tecnologia (FCT, Portugal) Noyes, R. W., Hartmann, L. W., Baliunas, S. L., Duncan, D. K., & Vaughan, through national funds and by FEDER through COMPETE2020 in the context A. H. 1984, ApJ, 279, 763 of the projects and grants reference UID/FIS/04434/2013 & POCI-01-0145- Pepe, F., Mayor, M., Galland, F., et al. 2002, A&A, 388, 632 FEDER-007672, PTDC/FIS-AST/1526/2014 & POCI-01-0145-FEDER-016886, Perrakis, K., Ntzoufras, I., & Tsionas, E. G. 2014, Comput. Stat. Data Anal., 77, and IF/00169/2012/CP0150/CT0002. This publication makes use of the Data & 54 Analysis Center for Exoplanets (DACE), which is a facility based at the Univer- Perruchot, S., Kohler, D., Bouchy, F., et al. 2008, in Ground-based and Airborne sity of Geneva (CH) dedicated to extrasolar planets data visualisation, exchange, Instrumentation for Astronomy II, Proc. SPIE, 7014, 70140J and analysis. DACE is a platform of the Swiss National Centre of Competence in Perruchot, S., Bouchy, F., Chazelas, B., et al. 2011, in Techniques and Instrumen- Research (NCCR) PlanetS, federating the Swiss expertise in Exoplanet research. tation for Detection of Exoplanets V, Proc. SPIE, 8151, 815115 The DACE platform is available at https://dace.unige.ch. We thank the Queloz, D. 1995, in New Developments in Array Technology and Applications, anonymous referees for their comments which helped to improve this article. eds. A. G. D. Philip, K. Janes, & A. R. Upgren, IAU Symp., 167, 221 Quirrenbach, A., Amado, P. J., Caballero, J. A., et al. 2014, in Ground-based and Airborne Instrumentation for Astronomy V, 9147, 91471F References Quirrenbach, A., Amado, P. J., Caballero, J. A., et al. 2016, in Ground-based and Airborne Instrumentation for Astronomy VI, 9908, 990812 Affer, L., Micela, G., Damasso, M., et al. 2016, A&A, 593, A117 Reid, I. N., & Hawley, S. L. 2005, New Light on Dark Stars Red Dwarfs, Anglada-Escudé, G., & Butler, R. P. 2012, ApJS, 200, 15 Low-Mass Stars, Brown Stars (Springer-Praxis books in astrophysics and Artigau, É., Kouach, D., Donati, J.-F., et al. 2014, in Ground-based and Airborne astronomy) Instrumentation for Astronomy V, Proc. SPIE, 9147, 914715 Reiners, A., Ribas, I., Zechmeister, M., et al. 2018, A&A, 609, L5 Astudillo-Defru, N., Bonfils, X., Delfosse, X., et al. 2015, A&A, 575, A119 Ricker, G. R. 2016, AGU Fall Meeting Abstracts, P13C-01 Astudillo-Defru, N., Delfosse, X., Bonfils, X., et al. 2017a, A&A, 600, A13 Seager, S. 2010, The Exoplanet Encyclopaedia, http://exoplanet.eu/ Astudillo-Defru, N., Díaz, R. F., Bonfils, X., et al. 2017b, A&A, 605, L11 van Leeuwen, F. 2007, A&A, 474, 653 Astudillo-Defru, N., Forveille, T., Bonfils, X., et al. 2017c, A&A, 602, A88 Williams, D. M., & Pollard, D. 2002, Int. J. Astrobiol., 1, 61 Baluev, R. V. 2008, MNRAS, 385, 1279 Wright, D. J., Wittenmyer, R. A., Tinney, C. G., Bentley, J. S., & Zhao, J. 2016, Boisse, I., Moutou, C., Vidal-Madjar, A., et al. 2009, A&A, 495, 959 ApJ, 817, L20 Boisse, I., Eggenberger, A., Santos, N. C., et al. 2010, A&A, 523, A88 Zacharias, N., Finch, C. T., Girard, T. M., et al. 2012, VizieR Online Data Bonfils, X., Delfosse, X., Udry, S., et al. 2013, A&A, 549, A109 Catalog: I/322

A103, page 11 of 15 A&A 618, A103 (2018)

Appendix A: Radial velocities Appendix B: MCMC analysis – full probed In this appendix, we present for each star the full set of radial parameters velocities derived by template-matching, with the short-term We present the full set of parameters probed by the DACE instrumental drift and long-term zero-point corrections applied. MCMC analysis for Gl96 and Gl617A, and the derived physical parameters. Table A.1. Radial velocities for Gl96 derived by template-matching, with the short-term instrumental drift and long-term zero-point corrections applied.

BJD 2400000 [d] RV [km s 1] sigma RV [km s 1] − − − 55813.6543 37.8703 0.0023 55916.385 −37.8778 0.003 56177.6458 −37.8799 0.002 ...... − ...

Notes. The complete version of this table is available at the CDS.

Table A.2. Radial velocities for Gl617A derived by template-matching, with the short-term instrumental drift and long-term zero-point corrections applied.

BJD 2400000 [d] RV [km s 1] sigma RV [km s 1] − − − 55827.3208 18.7247 0.0019 55982.7022 −18.714 0.0018 56328.6907 −18.7186 0.0019 ...... − ...

Notes. The complete version of this table is available at the CDS.

Table B.1. Parameters probed by the MCMC used to ﬁt the RV measurements of Gl96.

Parameter Units Max(Like) Med Mod Std CI(15.85) CI(84.15) CI(2.275) CI(97.725) Prior Likelihood log (Post) 193.285494 196.771970 195.921884 1.870757 199.232726 195.096198 202.542797 194.075043 − − − − − − − log (Like) 191.922894 195.623787 195.175969 1.973270 198.178506 193.846439 201.754278 192.754170 − − − − − − − log (Prior) 1.362600 1.078578 1.136538 0.427678 1.579763 0.619581 2.223648 0.226630 − − − − − − − M? [M ] 0.621188 0.600424 0.596743 0.026436 0.570211 0.630397 0.540296 0.660160 U σ [m s 1] 3.00 3.45 3.21 0.93 2.54 2.30 1.78 0.77 JIT − U γ [m s 1] 37874.88 37874.84 37874.90 0.28 37875.15 37874.53 37875.50 37874.22 SOPHIE − − − − − − − − U P [d] 73.939773 73.937730 73.993314 0.325983 73.553342 74.264366 73.092750 74.719936 U K [m s 1] 5.20 4.69 4.42 0.60 4.07 5.41 3.50 6.26 − U e 0.495245 0.439034 0.430347 0.088190 0.332797 0.529659 0.200769 0.609855 U ω [deg] 341.181282 339.576769 338.551057 11.790040 325.056172 352.022293 307.050528 358.379033 U λ [deg] 64.471037 63.383026 67.891413 45.293155 9.598258 108.428683 55.974137 170.848142 0 − U Notes. The maximum likelihood solution (Max(Like)), the median (Med), mode (Mod) and standard deviation (Std) of the posterior distribution for each parameter is shown, as well as the 68.3% (CI(15.85), CI(84.15)) and 95.45% (CI(2.275), CI(97.725)) conﬁdence intervals. The prior for each parameter can be of type: : uniform, : normal, : split normal, : truncated normal. U N SN TN

A103, page 12 of 15 M. J. Hobson et al.: The SOPHIE search for northern extrasolar planets. XIII.

Table B.2. Physical parameters derived from the MCMC chains used to ﬁt the RV measurements of Gl96.

Parameter Units Max(Like) Med Mod Std CI(15.85) CI(84.15) CI(2.275) CI(97.725) Prior Likelihood

M? [M ] 0.621188 0.600424 0.596743 0.026436 0.570211 0.630397 0.540296 0.660160 U P [d] 73.939773 73.937730 73.993314 0.325983 73.553342 74.264366 73.092750 74.719936 U K [m s 1] 5.20 4.69 4.42 0.60 4.07 5.41 3.50 6.26 − U e 0.495245 0.439034 0.430347 0.088190 0.332797 0.529659 0.200769 0.609855 U ω [deg] 341.181282 339.576769 338.551057 11.790040 325.056172 352.022293 307.050528 358.379033 U TP [d] 55556.833035 55556.392386 55553.872826 9.006008 55547.408500 55566.959102 55534.517800 55579.539962

TC [d] 55567.178726 55568.899883 55564.573169 9.663639 55559.078770 55580.463887 55546.269103 55593.191399 Ar [AU] 0.294178 0.290799 0.290868 0.004383 0.285732 0.295728 0.280513 0.300435 M sin i [M ] 0.067975 0.061855 0.058986 0.006479 0.054626 0.069477 0.047715 0.077017 · Jup M sin i [M ] 21.602453 19.657445 18.745664 2.058966 17.359998 22.079752 15.163668 24.475879 · Earth Notes. The maximum likelihood solution (Max(Like)), the median (Med), mode (Mod) and standard deviation (Std) for the posterior distribution of each parameter is shown, as well as the 68.3% (CI(15.85), CI(84.15)) and 95.45% (CI(2.275), CI(97.725)) conﬁdence intervals. The prior for each parameter can be of type: : uniform, : normal, : split normal, : truncated normal. U N SN TN

Table B.3. Parameters probed by the MCMC used to ﬁt the RV measurements of Gl617A – 1-Keplerian model plus quadratic drift.

Parameter Units Max(Like) Med Mod Std CI(15.85) CI(84.15) CI(2.275) CI(97.725) Prior Likelihood log (Post) 462.619011 466.651637 466.520190 1.964561 469.209353 464.831362 472.561763 463.627502 − − − − − − − log (Like) 461.941145 466.049364 465.293525 2.015104 468.693270 464.197613 472.107721 462.923059 − − − − − − − log (Prior) 0.677866 0.555253 0.500711 0.254783 0.874032 0.292873 1.220718 0.097352 − − − − − − − M? [M ] 0.611414 0.600358 0.595467 0.026063 0.570942 0.630052 0.540715 0.658600 U σ [m s 1] 4.44 4.86 4.73 0.52 4.30 4.24 3.78 3.55 JIT − U γ [m s 1] 18739.60 18737.35 18736.94 4.89 18743.09 18731.97 18748.80 18726.19 SOPHIE − − − − − − − − U lin [m s 1 yr 1] 8.53 7.71 7.09 2.09 5.41 10.13 3.01 12.66 − − U quad [m s 1 yr 2] 0.93 0.86 0.84 0.21 1.11 0.63 1.36 0.38 − − − − − − − − − U P [d] 86.916721 86.925044 86.904924 0.164238 86.730726 87.102317 86.523585 87.278783 U K [m s 1] 6.70 6.56 6.47 0.35 6.16 6.97 5.78 7.37 − U e 0.349308 0.316142 0.319096 0.073835 0.229602 0.396643 0.132376 0.468753 U ω [deg] 101.004334 102.200583 103.915448 9.807382 90.720509 112.245164 75.691327 123.345529 U λ [deg] 238.987027 239.918335 240.420912 16.482165 220.387042 257.567720 199.451375 275.038557 0 U Notes. The maximum likelihood solution (Max(Like)), the median (Med), mode (Mod) and standard deviation (Std) of the posterior distribution for each parameter is shown, as well as the 68.3% (CI(15.85), CI(84.15)) and 95.45% (CI(2.275), CI(97.725)) conﬁdence intervals. The prior for each parameter can be of type: : uniform, : normal, : split normal, : truncated normal. U N SN TN

Table B.4. Physical parameters derived from the MCMC chains used to ﬁt the RV measurements of Gl617A – 1-Keplerian model plus quadratic drift.

Parameter Units Max(Like) Med Mod Std CI(15.85) CI(84.15) CI(2.275) CI(97.725) Prior Likelihood

M? [M ] 0.611414 0.600358 0.595467 0.026063 0.570942 0.630052 0.540715 0.658600 U P [d] 86.916721 86.925044 86.904924 0.164238 86.730726 87.102317 86.523585 87.278783 U K [m s 1] 6.70 6.56 6.47 0.35 6.16 6.97 5.78 7.37 − U e 0.349308 0.316142 0.319096 0.073835 0.229602 0.396643 0.132376 0.468753 U ω [deg] 101.004334 102.200583 103.915448 9.807382 90.720509 112.245164 75.691327 123.345529 U TP [d] 55466.686102 55466.601086 55465.672814 4.040069 55462.278510 55471.350467 55457.295504 55476.587804

TC [d] 55465.481743 55465.258184 55464.278744 3.805920 55461.181843 55469.727111 55456.842749 55474.584025 Ar [AU] 0.325940 0.323965 0.324425 0.004713 0.318459 0.329242 0.312843 0.334318 M sin i [M ] 0.098576 0.096167 0.094936 0.005835 0.089490 0.102775 0.083150 0.110012 · Jup M sin i [M ] 31.327394 30.561882 30.170671 1.854279 28.440031 32.661895 26.425117 34.961725 · Earth Notes. The maximum likelihood solution (Max(Like)), the median (Med), mode (Mod) and standard deviation (Std) for the posterior distribution of each parameter is shown, as well as the 68.3% (CI(15.85), CI(84.15)) and 95.45% (CI(2.275), CI(97.725)) conﬁdence intervals. The prior for each parameter can be of type: : uniform, : normal, : split normal, : truncated normal. U N SN TN

A103, page 13 of 15 A&A 618, A103 (2018)

Table B.5. Parameters probed by the MCMC used to ﬁt the RV measurements of Gl617A – 2-Keplerian model plus quadratic drift.

Parameter Units Max(Like) Med Mod Std CI(15.85) CI(84.15) CI(2.275) CI(97.725) Prior Likelihood log (Post) 434.164980 440.075963 440.047078 2.381702 443.144186 437.799233 446.780069 436.227907 − − − − − − − log (Like) 433.765721 439.525314 439.289290 2.390269 442.596771 437.243499 446.189031 435.601729 − − − − − − − log (Prior) 0.399259 0.486225 0.368662 0.298923 0.882288 0.243982 1.520647 0.083642 − − − − − − − M? [M ] 0.581188 0.598699 0.589123 0.026717 0.569082 0.629830 0.538640 0.660030 U σ [m s 1] 3.47 3.71 3.83 0.42 3.25 3.22 2.86 2.68 JIT − U γ [m s 1] 18714.90 18715.26 18716.78 4.92 18720.89 18709.68 18726.32 18703.97 SOPHIE − − − − − − − − U lin [m s 1 yr 1] 3.13 2.98 2.46 2.19 5.47 0.48 8.10 1.92 − − − − − − − − U quad [m s 1 yr 2] 0.32 0.30 0.24 0.23 0.04 0.57 0.22 0.85 − − − U P [d] 86.685971 86.716536 86.644378 0.163173 86.533594 86.911495 86.364597 87.082150 U K [m s 1] 6.51 6.57 6.62 0.33 6.19 6.93 5.83 7.32 − U e 0.232050 0.230947 0.204544 0.066766 0.152388 0.303804 0.070509 0.372129 U ω [deg] 91.237907 97.248455 96.102520 12.239918 83.790120 110.796187 66.715021 127.067212 U λ [deg] 215.291405 220.718337 216.160185 16.196583 202.266759 239.462980 185.195600 257.350058 0 U P [d] 485.429776 496.902333 484.032769 26.820241 475.087228 532.349347 456.095460 579.472904 U K [m s 1] 3.38 3.16 3.03 0.37 2.74 3.59 2.33 4.03 − U e 0.134236 0.146812 0.021121 0.110116 0.043096 0.298966 0.005614 0.458703 U ω [deg] 358.202334 311.969148 380.751216 423.482663 375.233350 668.453999 478.847476 714.163530 − − − − − U λ [deg] 240.692904 274.875882 256.683399 77.426560 207.978823 379.031065 146.361972 491.545801 0 U Notes. The maximum likelihood solution (Max(Like)), the median (Med), mode (Mod) and standard deviation (Std) of the posterior distribution for each parameter is shown, as well as the 68.3% (CI(15.85), CI(84.15)) and 95.45% (CI(2.275), CI(97.725)) conﬁdence intervals. The prior for each parameter can be of type: : uniform, : normal, : split normal, : truncated normal. U N SN TN

Table B.6. Physical parameters derived from the MCMC chains used to ﬁt the RV measurements of Gl617-A – 2-Keplerian model plus quadratic drift.

Parameter Units Max(Like) Med Mod Std CI(15.85) CI(84.15) CI(2.275) CI(97.725) Prior Likelihood

M? [M ] 0.581188 0.598699 0.589123 0.026717 0.569082 0.629830 0.538640 0.660030 U P [d] 86.685971 86.716536 86.644378 0.163173 86.533594 86.911495 86.364597 87.082150 U K [m s 1] 6.51 6.57 6.62 0.33 6.19 6.93 5.83 7.32 − U e 0.232050 0.230947 0.204544 0.066766 0.152388 0.303804 0.070509 0.372129 U ω [deg] 91.237907 97.248455 96.102520 12.239918 83.790120 110.796187 66.715021 127.067212 U TP [d] 55470.128617 55470.170612 55469.066335 4.360219 55465.320275 55475.139385 55459.955671 55480.110606

TC [d] 55469.947886 55469.184485 55469.744721 3.803915 55464.775258 55473.489374 55460.507190 55477.410804 Ar [AU] 0.319911 0.323187 0.321331 0.004833 0.317754 0.328762 0.311972 0.333917 M sin i [M ] 0.096018 0.098450 0.097920 0.005996 0.091680 0.105364 0.085691 0.113023 · Jup M sin i [M ] 30.514385 31.287511 31.118932 1.905555 29.135990 33.484633 27.232456 35.918729 · Earth P [d] 485.429776 496.902333 484.032769 26.820241 475.087228 532.349347 456.095460 579.472904 U K [m s 1] 3.38 3.16 3.03 0.37 2.74 3.59 2.33 4.03 − U e 0.134236 0.146812 0.021121 0.110116 0.043096 0.298966 0.005614 0.458703 U ω [deg] 358.202334 311.969148 380.751216 423.482663 375.233350 668.453999 478.847476 714.163530 − − − − − U TP [d] 55177.869828 55214.752874 55283.022634 135.078647 55031.879715 55361.011891 54952.197623 55438.122406

TC [d] 55276.199082 55275.541145 55277.491526 107.612883 55154.204758 55385.387258 55055.822719 55543.038427 Ar [AU] 1.008817 1.035866 1.026245 0.039843 0.999836 1.087041 0.965852 1.152976 M sin i [M ] 0.090084 0.085769 0.082725 0.010332 0.074076 0.097851 0.062734 0.109579 · Jup M sin i [M ] 28.628775 27.257285 26.290082 3.283456 23.541287 31.097041 19.936952 34.824225 · Earth Notes. The maximum likelihood solution (Max(Like)), the median (Med), mode (Mod) and standard deviation (Std) for the posterior distribution of each parameter is shown, as well as the 68.3% (CI(15.85), CI(84.15)) and 95.45% (CI(2.275), CI(97.725)) conﬁdence intervals. The prior for each parameter can be of type: : uniform, : normal, : split normal, : truncated normal. U N SN TN

A103, page 14 of 15 M. J. Hobson et al.: The SOPHIE search for northern extrasolar planets. XIII.

Table B.7. Parameters probed by the MCMC used to ﬁt the combined SOPHIE, CARMENES, and KECK RV measurements of Gl617A.

Parameter Units Max(Like) Med Mod Std CI(15.85) CI(84.15) CI(2.275) CI(97.725) Prior Likelihood log (Post) 934.737509 939.951019 940.359405 2.181801 942.707691 937.845539 946.160275 936.299385 − − − − − − − log (Like) 931.716450 937.193692 936.740517 2.633389 940.751319 934.848899 944.875654 933.276083 − − − − − − − log (Prior) 3.021059 3.015167 3.129632 1.055714 3.056399 0.160778 3.122142 0.006795 − − − − − − − M? [M ] 0.617017 0.599486 0.595975 0.025964 0.569661 0.628938 0.539536 0.658046 U σ [m s 1] 14.15 15.89 15.31 2.25 14.07 20.03 12.57 21.49 JIT − U γ [m s 1] 0.51 0.40 0.14 1.44 1.20 2.03 2.85 3.80 CARMENES − − − U 1 γKECK PUB [m s− ] 2.55 1.76 1.60 2.37 4.44 0.87 7.25 3.72 − 1 − − − − − U γ [m s 1] 18721.42 18721.59 18721.85 1.34 18723.09 18720.05 18724.63 18718.48 SOPHIE − − − − − − − − U lin [m s 1 yr 1] 0.20 0.12 0.21 0.40 0.57 0.32 1.06 0.79 − − − − − − − U quad [m s 1 yr 2] 0.01 0.00 0.02 0.04 0.05 0.04 0.09 0.09 − − − − − − U P [d] 86.717889 86.776472 86.776265 0.135383 86.630677 86.933133 86.478695 87.110679 U K [m s 1] 5.77 5.83 5.76 0.20 5.59 6.05 5.38 6.28 − U e 0.067519 0.071525 0.072428 0.035663 0.030243 0.112694 0.004864 0.159102 U ω [deg] 74.887523 97.219922 98.412412 34.567998 55.385453 127.614449 16.938101 176.384852 − U λ [deg] 220.983105 227.841194 223.343831 14.690130 211.859018 244.846508 195.405489 263.911937 0 U Notes. The maximum likelihood solution (Max(Like)), the median (Med), mode (Mod) and standard deviation (Std) for the posterior distribution of each parameter is shown, as well as the 68.3% (CI(15.85), CI(84.15)) and 95.45% (CI(2.275), CI(97.725)) conﬁdence intervals. The prior for each parameter can be of type: : uniform, : normal, : split normal, : truncated normal. U N SN TN

Table B.8. Physical parameters derived from the MCMC chains used to ﬁt the combined SOPHIE, CARMENES, and KECK RV measurements of Gl617A.

Parameter Units Max(Like) Med Mod Std CI(15.85) CI(84.15) CI(2.275) CI(97.725) Prior Likelihood

M? [M ] 0.617017 0.599486 0.595975 0.025964 0.569661 0.628938 0.539536 0.658046 U P [d] 86.717889 86.776472 86.776265 0.135383 86.630677 86.933133 86.478695 87.110679 U K [m s 1] 5.77 5.83 5.76 0.20 5.59 6.05 5.38 6.28 − U e 0.067519 0.071525 0.072428 0.035663 0.030243 0.112694 0.004864 0.159102 U ω [deg] 74.887523 97.219922 98.412412 34.567998 55.385453 127.614449 -16.938101 176.384852 U TP [d] 55464.808054 55468.264643 55469.553885 9.043231 55457.599658 55476.783117 55440.927623 55489.460498 TC [d] 55467.985292 55466.914469 55465.796807 3.657568 55462.799889 55471.033793 55458.170838 55475.281484 Ar [AU] 0.326432 0.323420 0.322034 0.004684 0.317906 0.328658 0.312283 0.333674 M sin i [M ] 0.090837 0.089847 0.090072 0.004054 0.085275 0.094534 0.080783 0.099046 · Jup M sin i [M ] 28.867969 28.553526 28.624925 1.288239 27.100269 30.042914 25.672842 31.476762 · Earth Notes. The maximum likelihood solution (Max(Like)), the median (Med), mode (Mod) and standard deviation (Std) for the posterior distribution of each parameter is shown, as well as the 68.3% (CI(15.85), CI(84.15)) and 95.45% (CI(2.275), CI(97.725)) conﬁdence intervals. The prior for each parameter can be of type: : uniform, : normal, : split normal, : truncated normal. U N SN TN

A103, page 15 of 15 2.5. Summary of SOPHIE results on M dwarfs 71

2.5.1.2 A warm Neptune around the M dwarf Gl 378

My second publication from the SP3 survey, Hobson et al.(2019), presented the discovery of the warm Neptune Gl 378 b. The planet is brieﬂy described in the following paragraph, and the full paper follows.

Gl 378 b A warm Neptune (13.02 M⊕) around an M1 dwarf, with an orbital period of 3.82 d, at the lower boundary of the hot Neptune desert. Its eccentricity of 0.1, non- negligible and similar to fellow warm Neptunes GJ 436 b and GJ 3470 b, may point to a history of high-eccentricity migration. It is an interesting candidate for transit searches, and (since it is probably strongly irradiated and surrounded by a giant exosphere) for UV/Ly-α observations. A&A 625, A18 (2019) Astronomy https://doi.org/10.1051/0004-6361/201834890 & © M. J. Hobson et al. 2019 Astrophysics

The SOPHIE search for northern extrasolar planets XV. A warm Neptune around the M dwarf Gl 378? M. J. Hobson1, X. Delfosse2, N. Astudillo-Defru3, I. Boisse1, R. F. Díaz4,5, F. Bouchy6, X. Bonﬁls2, T. Forveille2, L. Arnold7, S. Borgniet2, V. Bourrier6, B. Brugger1, N. Cabrera Salazar2, B. Courcol1, S. Dalal8, M. Deleuil1, O. Demangeon9, X. Dumusque6, N. Hara6,11, G. Hébrard8,7, F. Kiefer8, T. Lopez1, L. Mignon2, G. Montagnier8,7, O. Mousis1, C. Moutou1,12, F. Pepe6, J. Rey6, A. Santerne1, N. C. Santos9,10, M. Stalport6, D. Ségransan6, S. Udry6, and P. A. Wilson13,14,8

1 CNRS, CNES, LAM, Aix-Marseille Université, Marseille, France e-mail: [email protected] 2 CNRS, IPAG, Université Grenoble Alpes, 38000 Grenoble, France 3 Departamento de Astronomía, Universidad de Concepción, Casilla 160-C, Concepción, Chile 4 Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Buenos Aires, Argentina 5 Instituto de Astronomía y Física del Espacio (IAFE), CONICET – Universidad de Buenos Aires, Buenos Aires, Argentina 6 Observatoire Astronomique de l’Université de Genève, 51 Chemin des Maillettes, 1290 Versoix, Switzerland 7 Observatoire de Haute-Provence, CNRS, Aix-Marseille Université, Institut Pythéas UMS 3470, 04870 Saint-Michel-l’Observatoire, France 8 Institut d’Astrophysique de Paris, UMR7095 CNRS, Université Pierre & Marie Curie, 98bis boulevard Arago, 75014 Paris, France 9 Instituto de Astrofísica e Ciências do Espaço, Universidade do Porto, CAUP, Rua das Estrelas, 4150-762 Porto, Portugal 10 Departamento de Física e Astronomia, Faculdade de Ciências, Universidade do Porto, Rua do Campo Alegre, 4169-007 Porto, Portugal 11 ASD/IMCCE, CNRS-UMR8028, Observatoire de Paris, PSL, UPMC, 77 Avenue Denfert-Rochereau, 75014 Paris, France 12 Canada-France-Hawaii Telescope Corporation, 65-1238 Mamalahoa Hwy, Kamuela, HI 96743, USA 13 Department of Physics, University of Warwick, Coventry CV4 7AL, UK 14 Centre for Exoplanets and Habitability, University of Warwick, Coventry CV4 7AL, UK

Received 15 December 2018 / Accepted 12 February 2019

ABSTRACT

We present the detection of a warm Neptune orbiting the M dwarf Gl 378, using radial velocity measurements obtained with the SOPHIE spectrograph at the Observatoire de Haute-Provence. The star was observed in the context of the SOPHIE exoplanet consortium’s sub-programme dedicated to ﬁnding planets around M dwarfs. Gl 378 is an M1 star, of solar metallicity, at a distance of 14.96 pc. The single planet detected, Gl 378 b, has a minimum mass of 13.02 MEarth and an orbital period of 3.82 days, which place it at the lower boundary of the hot Neptune desert. As one of only a few such planets around M dwarfs, Gl 378 b provides important clues to the evolutionary history of these close-in planets. In particular, the eccentricity of 0.1 may point to a high-eccentricity migration. The planet may also have lost part of its envelope due to irradiation. Key words. techniques: radial velocities – planetary systems – stars: late-type – stars: individual: G1 378

1. Introduction Statistics on M-dwarf planets remain less certain than those on planets around Sun-type stars, due to the comparatively small The mass-period diagram is an important diagnostic of the for- number of detections, though these detections are expected to mation and evolution of planetary systems. There is a known increase thanks to several current or upcoming projects such as dearth of Neptune-size exoplanets at short orbital periods com- SPIRou (Artigau et al. 2014), CARMENES (e.g. Quirrenbach pared to both Jupiter-size and Earth-size planets, which is gener- et al. 2014, 2016), HADES (e.g. Affer et al. 2016), and NIRPS ally referred to as the Neptune or sub-Jovian desert (Lecavelier (Bouchy et al. 2017) in radial velocity; or TESS (NASA mission, Des Etangs 2007; Davis & Wheatley 2009; Szabó & Kiss 2011; launched in April 2018, Ricker 2016), TRAPPIST (e.g. Gillon Beaugé & Nesvorný 2013; Helled et al. 2016; Mazeh et al. 2016). et al. 2017), SPECULOOS (Delrez et al. 2018), and ExTrA It is unlikely to be an observational bias, but is more proba- (Bonﬁls et al. 2015) in transits. Nevertheless, it is clear that bly due to photoevaporation and/or high-eccentricity migration while hot Jupiters are rare around M dwarfs, short-period Earths (Owen & Lai 2018; Ionov et al. 2018). and superEarths are numerous; however, hot Neptunes remain ? Based on observations collected with the SOPHIE spectrograph on unusual, making up only about 3% of the sample of known exo- the 1.93 m telescope at the Observatoire de Haute-Provence (CNRS), planets around M dwarfs (e.g. Bonﬁls et al. 2013; Dressing & France, by the SOPHIE Consortium. Charbonneau 2015; Hirano et al. 2018).

A18, page 1 of8 Open Access article, published by EDP Sciences, under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. A&A 625, A18 (2019)

The SOPHIE exoplanet consortium has led several ongoing exoplanet-hunting programmes on the SOPHIE spectrograph at the Observatoire de Haute Provence since 2006 (Bouchy et al. 2009). Sub-programme 3, also known as SP3, is dedicated to the hunt for exoplanets around M-dwarf stars. Via a systematic survey of a volume-limited sample of M dwarfs within 12 parsecs of the Sun, it seeks to detect superEarths and Neptunes, constrain the statistics of planets around M dwarfs, and find potentially transiting companions. With a general radial velocity precision 1 of 1–2 m s− on solar-type stars, SOPHIE has proved to be a successful planet hunter. For the SP3 in particular, we recently published the first two exoplanets from this sub-programme: the detection of Gl 96 b and the independent confirmation of Gl 617A b (Hobson et al. 2018). In this work, we report the detection of a warm Neptune on the lower boundary of the hot Neptune desert, orbiting the M dwarf Gl 378, which was observed as part of this survey. We describe the data and its analysis in Sects.2 and3, respectively. The results are presented in Sect.4 and discussed in Sect.5. Fig. 1. Correction for the long-term variations in the zero-point (red Finally, we conclude in Sect.6. line) and the data points used to construct it (black dots). The points correspond to the ten stars detailed in Sect.3. The correction spans 1 7 yr, with a dispersion of 2.87 m s− and a peak-to-peak variation of 1 2. Observations 16.3 m s− .

Observations for Gl 378 were gathered between 2015 and 2018 uses ten stars: the four super-constants HD185144, HD 221354, with the SOPHIE+ spectrograph (Perruchot et al. 2011; Bouchy HD 89269A, and HD 9407; the three SP3 constants Gl 411, et al. 2013). A total of 62 spectra were obtained. All the Gl 514, and Gl 686; and the additional SP3 stars Gl 521, Gl 15A, observations were performed with simultaneous sky measure- and Gl 694, selected because they have a corrected rms after the ment in order to check for potential moonlight contamination. 1 first iteration lower than 3 m s− (as defined by Courcol et al. Additionally, a ThAr or FP calibration spectrum was obtained 2015). Figure1 shows the correction and the data used to derive immediately prior to each observation in order to measure the 1 1 it . instrumental drift (average value: 1 m s− ). For the observations The final radial velocities, which have a mean error bar of where the velocity difference between the moon and the star was 1 1 3 m s− (including photon noise and instrumental error, follow- less than 20 km s− , a merit function was applied, computed from ing Astudillo-Defru et al. 2015), are reported in AppendixA, and the velocity difference and the S/N and CCF contrast in fibre B, were analysed with the Data and Analysis Center for Exoplanets in order to identify possible contamination. In this way, 18 obser- (DACE) web platform2. We employed the Keplerian fitting tools vations were found to be contaminated by the moon and were (which follow Delisle et al. 2016) and the Markov chain Monte discarded, leaving a total of 44 spectra. We note that retaining Carlo (MCMC) analysis facilities (described in Díaz et al. 2014, these observations does not change the final planetary parame- 2016). ters within the uncertainties, but it does increase the noise. With an exposure time of 1800 s, the spectra have a median S/N of 83 (at 650 nm), resulting in a photon noise of 3 m s 1. For this 3.1. Activity indicators ≈ − star the photon noise is a little higher than that of instrument The main spectral activity indicators are the Hα index, the 1 systematics (1–2 m s− ). log(RHK0 ) index, and the CCF bisector. We followed Boisse et al. (2009) to compute the Hα index; Astudillo-Defru et al.(2017a) to determine the log(RHK0 ) index from the Ca II H and K lines 3. Data analysis measured in the spectrum; and obtained the CCF bisector from The data were reduced using the SOPHIE Data Reduction the SOPHIE DRS. Since the Na I D lines have been shown to be Software (DRS, Bouchy et al. 2009), which computes the good activity indicators for M dwarfs (Díaz et al. 2007; Gomes radial velocity (RV) by cross-correlation functions (CCF). For da Silva et al. 2011), we also calculated the Na index as defined M dwarfs this approach does not use all the Doppler content, by Gomes da Silva et al.(2011) from our SOPHIE spectra. The so we extracted RVs through a template-matching algorithm. values obtained are given in AppendixA. We shifted all the spectra to a common reference frame using the DRS RVs, and co-added them to build a high S/N stellar 3.2. Stellar parameters template. This template was Doppler shifted over a series of The stellar parameters are listed in Table1. Gl 378 was char- guess RVs, producing a Chi-square profile whose minimum cor- acterised in Gaidos et al.(2014), where we obtained spectral responds to the maximum likelihood RV (Astudillo-Defru et al. type, magnitudes and colour indices (except for the K magnitude, 2015, 2017b). which was taken from Cutri et al. 2003), effective temperature, SOPHIE shows long-term variations in the zero-point, an and luminosity. The coordinates, parallax, and distance were effect first described in Courcol et al.(2015). We constructed taken from the Gaia DR2 (Gaia Collaboration 2016, 2018). We an up-to-date correction from the SP3 stars, plus the four solar-type “super-constant” stars of the SOPHIE high-precision 1 This constant correction for SOPHIE RVs (applicable to M-dwarf programmes as defined by Courcol et al.(2015), in the same stars) is available upon request. way as in Hobson et al.(2018). Our updated constant correction 2 Available at https://dace.unige.ch

A18, page 2 of8 M. J. Hobson et al.: The SOPHIE search for northern extrasolar planets. XV.

Table 1. Stellar parameters. 20 Parameter Gl 378 10 Spectral type M1 (a) (a) V 10.19 0 B–V 1.26 (a)

(b) s] RV[m/ # V–K 2.03 -10 Mass (M ) 0.56 0.01 (c) ± (d) Radius (R ) 0.56 0.02 -20 α (h m s) 10 02 21.7516441689± (e) (e) δ (d m s) +48 05 19.687165248 7400 7500 7600 7700 7800 7900 8000 8100 8200 8300 (e) Π (mas) 66.8407 0.0322 Date (BJD - 2,450,000.0) [d] Distance (pc) 14.9609 ± 0.0072 (e) ± ( f ) Fig. 2. SOPHIE RVs for Gl 378, obtained using template-matching, log(RHK0 ) 4.98 0.06 − ± (a) and corrected for the nightly drift and the long-term variations of the Teﬀ (K) 3879 67 zero-point. ± (a) L? (L ) 0.06 0.01 Fe/H (dex) 0.06 ± 0.09 (g) ± References. (a)Gaidos et al.(2014). (b)V: Gaidos et al.(2014), K: Cutri 0. 4 et al.(2003). (c)This work, following Mann et al.(2019). (d)This work, (e) ( f ) following Mann et al.(2015). Gaia Collaboration(2018). This work, 0. 2 following Astudillo-Defru et al.(2017a). (g)This work, following Neves

et al.(2014). r Powe d Normalize

1 10 100 1000 obtained the mean and standard deviation of log(RHK0 ) from the Pe riod [d] SOPHIE spectra, and used the log(RHK0 ) log(Prot) relation from Astudillo-Defru et al.(2017a) to estimate− a rotation period of Fig. 3. Periodogram of the SOPHIE RVs for Gl 378, obtained using 40.5 4 days, with error bars calculated by propagation. We template-matching, and corrected for the nightly drift and the long-term employed± the MCAL code of Neves et al.(2014) to estimate the variations of the zero-point. The horizontal lines indicate the 50, 10, 1, metallicity from our SOPHIE spectra. Finally, we used the dis- and 0.1% FAP levels (from bottom to top). tance measurement from Gaia Collaboration(2018) to estimate Gl378 l1-periodogram a more precise radius (following Mann et al. 2015, with errors 3 Candidates tested at : 3.8218 6.2396 1.3495 1.0439 days calculated by propagation) and stellar mass (using the MCMC log10(FAP) : -3.3386 -0.031596 0 0 routine provided by Mann et al. 2019, based on masses from 2.5 log10(Bayes F) Laplace approx.: 1.8531 -1.2598 -4.4837 -3.3139 Delfosse et al. 2000) than were available in the literature. 2 1.5

1 4. Results Amplitude (m/s) 0.5

The time series of the radial velocities is shown in Fig.2, and its 0 periodogram in Fig.3. The periodogram shows a clear peak at 100 101 102 103 3.82 d, which by bootstrap resampling we place below a 0.01% Period (days) false alarm probability (FAP). The other notable peaks, at 0.79 d Fig. 4. l1-periodogram following Hara et al.(2017). The FAP is com- and 1.35d, are 1-day aliases of the 3.82 d period; they are system- puted according to the Baluev(2008) formula, and the Bayes factor is atically weaker than the 3.82 d peak, and attempted Keplerian fits computed via a Laplace approximation with the same methodology as have higher σ(O C). We also computed the l1-periodogram as in in Nelson et al.(2018, Appendix A.4). Hara et al.(2017−), which is shown in Fig.4, and confirms the 3.82 d signal as the most significant. The periodograms of the four activity indicators described We employed the DACE platform to fit a Keplerian signal in Sect. 3.1 are shown in Fig.5. None of them shows any to the 3.82 d period. The highest peak in the periodogram of peaks below 10% FAP, or any peak whatsoever at the 3.82 d the residuals has a FAP of 17%, and is therefore not significant period found in the RVs. Likewise, there is no anti-correlation (Fig.6). In order to sample the joint posterior distribution of the in evidence between the RVs and the CCF bisector. Addition- model parameters, we carried out an MCMC analysis. We used ally, for an M dwarf, a rotation period of 3.82 d would lead to a model with a single Keplerian and an additive stellar jitter. The an extremely high activity level, with saturated chromospheric resulting parameters are summarised in Table2, with the full out- emission of Hα (Delfosse et al. 1998) and Ca (Astudillo-Defru puts available in AppendixB. The phase-folded data and fitted et al. 2017a), which is clearly not the case for Gl 378. Finally, Keplerian are shown in Fig.7. Houdebine(2010) found Gl 378 to be a slow rotator, with 1 v sin i = 2.25 km s− . Therefore, we conclude that the 3.82 d 5. Discussion peak cannot be of stellar origin, and that Gl 378 shows no evidence of clear stellar activity (at the estimated rotation period of With a minimum mass of 13.63 MEarth and an orbital period of PRot = 40.5 4 d, indicated by the shaded grey regions in Fig.5, 3.82 d, Gl 378 b is a warm Neptune-like exoplanet. Depending or any other± period) in the SOPHIE spectra. on the heat redistribution factor and albedo assumed, we obtain

A18, page 3 of8 A&A 625, A18 (2019)

Hα index Table 2. Best-ﬁt solution for the planetary system orbiting Gl 378. 0.3 Param. Units Gl 378 b 0.2 +0.001 P (d) 3.822 0.001 0.1 1 +−1.24 K (m s− ) 7.96 1.23 Normalized Power −+0.131 e 0.109 0.077 1 Log(R' )10 index 100 1000 − HK ω (deg) 210.6+79.8 0.4 Period [d] 116.9 − +1.27 TP (d) 2455500.02 1.3 −+0.72 TC (d) 2455502.564 1.05 0.2 − +0.00023 a (AU) 0.039435 0.00023 Normalized Power +−2.03 0.4 M sin i (MEarth) 13.02 2.01 1 Na index10 100 1000 − Period [d] γ (m s 1) 9697.333+0.85 SOPHIE − − 0.83 1 +0.83− σJIT (m s− ) 4.610 0.2 0.74 1 − σ(O C) (m s− ) 4.86 − +1.41 log (Likelihood) 138.07 2.09 Normalized Power − − 1 10 100 CCF bisector1000 Notes. For each parameter the median of the posterior is given, with Period [d] 0.2 error bars computed from the MCMC analysis using a 68.3% conﬁdence interval. σO C corresponds to the weighted standard deviation of the residuals around− the best solutions. All the parameters probed by the 0.1 MCMC can be found in Table B.1. Normalized Power

1 10 100 1000 20 Period [d] 10 Fig. 5. Periodogram of the activity indices for Gl 378. From top to bottom panels:Hα index, log(RHK0 ) index, Na index, and CCF bisector. The horizontal lines indicate the 50 and 10% FAP levels. The red 0 vertical dashed line marks the orbital period of Gl 378 b. The shaded V[/s] RV[m/ # grey region indicates the probable rotation period, as estimated from -10 the log(RHK0 ). -20

0. 3 0 50 100 150 200 250 300 350 Mean Anomaly [deg] 0. 2 Fig. 7. SOPHIE RVs for Gl 378, phase-folded to the one-planet model 0. 1 with P = 3.82 d. The curve indicates the fitted model. omlz oer Powe d Normalize 1 10 100 1000 The lower boundary is believed to have its origin in photoevapo- Pe riod [d] ration (Owen & Lai 2018). Therefore, Gl 378 b may have lost at least part of its gaseous envelope due to the X-ray and EUV irra- Fig. 6. Periodogram of the residuals of the SOPHIE RVs for Gl 378 diation from its host star. We note that the young active phase is following the fit of a planet at 3.82 d. The horizontal line indicates the 50% FAP level. long for M dwarfs compared to Sun-type stars, giving more time for evaporation to work. If we assume that the mass of Gl 378 b is close to its M sin i, equilibrium temperatures in the range of Teq 630 K (for a heat it is likely similar to GJ 436 b (Ehrenreich et al. 2015; Lavie redistribution factor of 1 and an albedo of 0.3,≈ close to that of et al. 2017) and GJ 3470 b (Bourrier et al. 2018a). These plan- Neptune) to Teq 830 K (for a redistribution factor of 2 and an ets are warm Neptunes, in the same region of the mass-period albedo of 0, as a≈ lower limit). Its orbital parameters place it on diagram as Gl 378 b (Fig.8), and orbit M-dwarf stars. Both the lower boundary of the Neptune desert as defined by Mazeh are surrounded by giant hydrogen exospheres; GJ 436b possibly et al.(2016), as shown in Fig.8, although they note that the lower became a warm Neptune recently due to a late high-eccentricity boundary is somewhat blurry, and other authors (e.g. Owen & migration and is thus currently not losing much mass (Bourrier Lai 2018) have placed the external limit in period at P 3 d et al. 2015, 2016, 2018b), while GJ 3470b is much more irra- rather than P 5 d. This location on the lower boundary rather≈ diated by its younger and earlier-type star and could have lost than within the≈ desert, and its range of probable equilibrium tem- up to 35% of its mass already (Bourrier et al. 2018a). This perature well below the methane condensation temperature of suggests that the warm Neptune population at the border of 1200 K, lead us to classify it as a warm rather than a hot Neptune. the desert is particularly sensitive to atmospheric escape, and

A18, page 4 of8 M. J. Hobson et al.: The SOPHIE search for northern extrasolar planets. XV.

increasing metallicity (Petigura et al. 2018, although it is a study on the Kepler survey where the planets are thus categorised by radius rather than mass). The activity indicators analysed show no evidence for quasi- periodic stellar activity signals. Likewise, the residuals of the Keplerian fit to the 3.82 d planet show no significant periodicity, although statistics on M dwarfs indicate that most of their planets are found in multi-planet systems (e.g. Bonfils et al. 2013; Dressing & Charbonneau 2015). However, the σO C (weighted standard deviation of the residuals around the best− solution) is high compared to the mean error bar of the observations, suggesting that there are further effects in the data, i.e. additional planets, stellar activity signals, and/or systematics. As an M dwarf, Gl 378 is faint in the visible, emitting most of its radiation in the infrared; consequently, high-precision infrared spectroscopy could help to detect further planets or place limits on their existence. Fig. 8. Hot Neptune desert. Gl 378 b is indicated by the red star. All planets with known masses and periods <10 d are shown (taken from exoplanets.eu on 18 Oct. 2018). Planets orbiting M dwarfs are high- 6. Conclusions lighted (orange hexagons); GJ 436 b and GJ 3470 are indicated by cross ( ) and plus signs (+), respectively. The location of the hot Neptune We have presented the detection of a Neptune-mass planet × desert is indicated (grey region), using the boundaries of Mazeh et al. orbiting the M dwarf Gl 378 at a period of 3.82 d. Its orbital (2016; black lines). The boundaries are dotted beyond 5 d, which reflects parameters place it at the edge of the hot Neptune desert, though the Mazeh et al.(2016) warning that beyond this orbital period the with a sin(i) degeneracy on the mass. Transit observations could existence of the desert is uncertain. help to break this degeneracy: if the planet transits the inclination can be measured, while a non-detection can place limits on it. Transit measurements would also provide the radius and supports this mechanism as the reason why hot Neptunes are therefore the density, permitting a characterisation of its prob- missing. The three planets also have similar non-zero eccen- able composition. Finally, if the planet transits, we should be tricities (Deming et al. 2007; Kosiarek et al. 2019), which may able to characterise its atmosphere in the UV/Ly-α line, given the point to high-eccentricity rather than disk-driven migrations. brightness and close distance of its host star to us, and because Therefore, objects like Gl 378 b are crucial to understanding the planet is a warm Neptune and therefore likely surrounded by the evolution of close-in planets, providing we can characterise a giant hydrogen exosphere. This exosphere can extend beyond them. the Roche lobe, resulting in an effective planet radius in UV an A fuller characterisation of the planet would require knowl- order of magnitude larger than the optical radius, and hence a edge of its density, and therefore of its radius. The transit slightly increased transit probability. Therefore, UV/Ly-α obser- probability of a planet detected by radial velocity can be vations at the expected transit times would be interesting even if approximated by P(transit) R /a, with R the stellar radius ≈ ? ? the planet does not transit in the optical. and a the semi-major axis of the planetary orbit (Borucki & Although we do not detect any other periodic signals in our Summers 1984). For Gl 378 b, we obtain a transit probability data, it is statistically likely that more planets are present. Moni- of P(transit) = 6.5 0.5%. As Gl 378 b is most likely a Neptune- ± toring this star with infrared spectroscopy should help to resolve mass exoplanet, the analysis of Stevens & Gaudi(2013) suggests this question, and to refine the ephemeris of Gl 378b for transit this transit probability is likely underestimated: they derive the searches. We hope to observe Gl 378 with SPIRou in the near posterior transit probability from the prior distributions of plan- future. etary masses and inclinations, finding that physically motivated distributions from planet formation models yield increased pos- Acknowledgements. We warmly thank the OHP staff for their support on the terior transit probabilities for Neptune-mass planets. The transit observations. We thank the anonymous referee for the careful reading and depth is given by ∆F = (R /R )2. Using the mass-radius relation valuable comments which helped to improve the manuscript. This work was p ? supported by the Programme National de Planétologie (PNP) of CNRS/INSU, of Chen & Kipping(2017), we estimate the radius of Gl 378 b as co-funded by CNE. X.De., X.B., I.B., and T.F. received funding from the French Rp 4.67 R ; when combined with the stellar radius, this value Programme National de Physique Stellaire (PNPS) and the Programme National gives≈ a transit⊕ depth of ∆F 0.58%. This could be observed by de Planétologie (PNP) of CNRS (INSU). X.B. acknowledges funding from the ground-based surveys, and should≈ be easily detectable by space- European Research Council under the ERC Grant Agreement n. 337591-ExTrA. This work has been supported by a grant from Labex OSUG@2020 (Investisse- based missions such as TESS (Barclay et al. 2018) or CHEOPS ments d’avenir – ANR10 LABX56). This work is also supported by the French (Rando et al. 2018). National Research Agency in the framework of the Investissements d’Avenir pro- The metallicity estimated for the host star, Gl 378, from gram (ANR-15-IDEX-02), through the funding of the “Origin of Life” project of our SOPHIE spectra is [Fe/H] = 0.06 0.09. This solar metal- the Univ. Grenoble-Alpes. V.B. acknowledges support from the Swiss National licity is in line with the tendency for± M dwarfs hosting plan- Science Foundation (SNSF) in the frame of the National Centre for Compe- tence in Research PlanetS, and has received funding from the European Research ets to be comparatively more metal rich (e.g. Courcol et al. Council (ERC) under the European Unions Horizon 2020 research and innova- 2016, Hirano et al. 2018, who find [Fe/H] & 0 for planet hosts) tion programme (project Four Aces; grant agreement No 724427). This work with respect to the sub-solar average metallicity of nearby was supported by FCT – Fundação para a Ciência e a Tecnologia through M dwarfs (e.g. Schlaufman & Laughlin 2010, who report national funds and by FEDER through COMPETE2020 – Programa Opera- [Fe/H] = 0.17 0.07, and Passegger et al. 2018, who find cional Competitividade e Internacionalização by grants UID/FIS/04434/2013 & − ± POCI-01-0145-FEDER-007672; PTDC/FIS-AST/28953/2017 & POCI-01-0145- mainly subsolar values for the CARMENES target sample) and FEDER-028953; and PTDC/FIS-AST/32113/2017 & POCI-01-0145-FEDER- also with findings that hot Neptunes are more common with 032113. N.A-D. acknowledges support from FONDECYT #3180063. N.H.

A18, page 5 of8 A&A 625, A18 (2019) acknowledges the financial support of the National Centre for Competence in Delrez, L., Gillon, M., Queloz, D., et al. 2018, SPIE Conf. Ser., 10700, 107001I Research PlanetS of the Swiss National Science Foundation (SNSF). X.Du. is Deming, D., Harrington, J., Laughlin, G., et al. 2007, ApJ, 667, L199 grateful to the Branco Weiss Fellowship–Society in Science for continuous sup- Díaz, R. F., Cincunegui, C., & Mauas, P. J. D. 2007, MNRAS, 378, 1007 port. This publication makes use of the Data & Analysis Center for Exoplanets Díaz, R. F., Almenara, J. M., Santerne, A., et al. 2014, MNRAS, 441, 983 (DACE), which is a facility based at the University of Geneva (CH) dedicated to Díaz, R. F., Ségransan, D., Udry, S., et al. 2016, A&A, 585, A134 extrasolar planets data visualisation, exchange, and analysis. DACE is a platform Dressing, C. D., & Charbonneau, D. 2015, ApJ, 807, 45 of the Swiss National Centre of Competence in Research (NCCR) PlanetS, feder- Ehrenreich, D., Bourrier, V., Wheatley, P. J., et al. 2015, Nature, 522, 459 ating the Swiss expertise in Exoplanet research. The DACE platform is available Gaia Collaboration (Prusti, T., et al.) 2016, A&A, 595, A1 at https://dace.unige.ch. Gaia Collaboration (Brown, A. G. A., et al.) 2018, A&A, 616, A1 Gaidos, E., Mann, A. W., Lépine, S., et al. 2014, MNRAS, 443, 2561 Gillon, M., Triaud, A. H. M. J., Demory, B.-O., et al. 2017, Nature, 542, 456 References Gomes da Silva, J., Santos, N. C., Bonfils, X., et al. 2011, A&A, 534, A30 Hara, N. C., Boué, G., Laskar, J., & Correia, A. C. M. 2017, MNRAS, 464, Affer, L., Micela, G., Damasso, M., et al. 2016, A&A, 593, A117 1220 Artigau, É., Kouach, D., Donati, J.-F., et al. 2014, in Ground-based and Airborne Helled, R., Lozovsky, M., & Zucker, S. 2016, MNRAS, 455, L96 Instrumentation for Astronomy V, Proc. SPIE, 9147, 914715 Hirano, T., Dai, F., Gandolfi, D., et al. 2018, AJ, 155, 127 Astudillo-Defru, N., Bonfils, X., Delfosse, X., et al. 2015, A&A, 575, A119 Hobson, M. J., Díaz, R. F., Delfosse, X., et al. 2018, A&A, 618, A103 Astudillo-Defru, N., Delfosse, X., Bonfils, X., et al. 2017a, A&A, 600, A13 Houdebine, E. R. 2010, MNRAS, 407, 1657 Astudillo-Defru, N., Díaz, R. F., Bonfils, X., et al. 2017b, A&A, 605, L11 Ionov, D. E., Pavlyuchenkov, Y. N., & Shematovich, V. I. 2018, MNRAS, 476, Baluev, R. V. 2008, MNRAS, 385, 1279 5639 Barclay, T., Pepper, J., & Quintana, E. V. 2018, ApJS, 239, 2 Kosiarek, M. R., Crossfield, I. J. M., Hardegree-Ullman, K. K., et al. 2019, AJ, Beaugé, C., & Nesvorný, D. 2013, ApJ, 763, 12 157, 97 Boisse, I., Moutou, C., Vidal-Madjar, A., et al. 2009, A&A, 495, 959 Lavie, B., Ehrenreich, D., Bourrier, V., et al. 2017, A&A, 605, L7 Bonfils, X., Delfosse, X., Udry, S., et al. 2013, A&A, 549, A109 Lecavelier Des Etangs, A. 2007, A&A, 461, 1185 Bonfils, X., Almenara, J. M., Jocou, L., et al. 2015, in Techniques and Instrumen- Mann, A. W., Feiden, G. A., Gaidos, E., Boyajian, T., & von Braun, K. 2015, tation for Detection of Exoplanets VII, Proc. SPIE, 9605, 96051L ApJ, 804, 64 Borucki, W. J., & Summers, A. L. 1984, Icarus, 58, 121 Mann, A. W., Dupuy, T., Kraus, A. L., et al. 2019, ApJ, 871, 63 Bouchy, F., Hébrard, G., Udry, S., et al. 2009, A&A, 505, 853 Mazeh, T., Holczer, T., & Faigler, S. 2016, A&A, 589, A75 Bouchy, F., Díaz, R. F., Hébrard, G., et al. 2013, A&A, 549, A49 Nelson, B. E., Ford, E. B., Buchner, J., et al. 2018, ArXiv e-prints Bouchy, F., Doyon, R., Artigau, É., et al. 2017, The Messenger, 169, 21 [arXiv:1806.04683] Bourrier, V., Ehrenreich, D., & Lecavelier des Etangs, A. 2015, A&A, 582, A65 Neves, V., Bonfils, X., Santos, N. C., et al. 2014, A&A, 568, A121 Bourrier, V., Lecavelier des Etangs, A., Ehrenreich, D., Tanaka, Y. A., & Vidotto, Owen, J. E., & Lai, D. 2018, MNRAS, 479, 5012 A. A. 2016, A&A, 591, A121 Passegger, V. M., Reiners, A., Jeffers, S. V., et al. 2018, A&A, 615, A6 Bourrier, V., Lecavelier des Etangs, A., Ehrenreich, D., et al. 2018a, A&A, 620, Perruchot, S., Bouchy, F., Chazelas, B., et al. 2011, in Techniques and Instrumen- A147 tation for Detection of Exoplanets V, Proc. SPIE, 8151, 815115 Bourrier, V., Lovis, C., Beust, H., et al. 2018b, Nature, 553, 477 Petigura, E. A., Marcy, G. W., Winn, J. N., et al. 2018, AJ, 155, 89 Chen, J., & Kipping, D. 2017, ApJ, 834, 17 Quirrenbach, A., Amado, P. J., Caballero, J. A., et al. 2014, in Ground-based and Courcol, B., Bouchy, F., Pepe, F., et al. 2015, A&A, 581, A38 Airborne Instrumentation for Astronomy V, Proc SPIE, 9147, 91471F Courcol, B., Bouchy, F., & Deleuil, M. 2016, MNRAS, 461, 1841 Quirrenbach, A., Amado, P. J., Caballero, J. A., et al. 2016, in Ground-based and Cutri, R. M., Skrutskie, M. F., van Dyk, S., et al. 2003, VizieR Online Data Airborne Instrumentation for Astronomy VI, Proc. SPIE, 9908, 990812 Catalog: II/246 Rando, N., Asquier, J., Corral Van Damme, C., et al. 2018, SPIE Conf. Ser., Davis, T. A., & Wheatley, P. J. 2009, MNRAS, 396, 1012 10698, 106980K Delfosse, X., Forveille, T., Perrier, C., & Mayor, M. 1998, A&A, 331, 581 Ricker, G. R. 2016, AGU Fall Meeting Abstracts, P13C-01 Delfosse, X., Forveille, T., Ségransan, D., et al. 2000, A&A, 364, 217 Schlaufman, K. C., & Laughlin, G. 2010, A&A, 519, A105 Delisle, J.-B., Ségransan, D., Buchschacher, N., & Alesina, F. 2016, A&A, 590, Stevens, D. J., & Gaudi, B. S. 2013, PASP, 125, 933 A134 Szabó, G. M., & Kiss, L. L. 2011, ApJ, 727, L44

A18, page 6 of8 M. J. Hobson et al.: The SOPHIE search for northern extrasolar planets. XV.

Appendix A: Radial velocities In this appendix, we present the radial velocities obtained for Gl 378 from SOPHIE+ using template-matching, corrected for the nightly drift and the long-term variations of the zero-point.

Table A.1. Radial velocities for Gl 378.

1 1 1 BJD ( 2400000 d) RV (km s− ) σRV (km s− ) Bisector (m s− ) Hα index σHα log(R0 ) σlog(R ) Na index σNa − HK HK0 57 383.6037 9.7084 0.003 15.167 0.2353 0.0012 4.9599 0.0001 0.2107 0.0103 57 393.6361 −9.6932 0.003 11.333 0.2242 0.0012 −4.9868 0.0001 0.2116 0.0101 57 419.5259 −9.7026 0.0028 2 0.2255 0.0011 −4.9107 0.0001 0.2085 0.0094 57 474.4968 −9.691 0.003 20.667− 0.2236 0.0012 −5.0253 0.0001 0.2079 0.0100 57 476.428 −9.7083 0.0027 21.333 0.2248 0.0010 −5.002 0.0001 0.2045 0.0089 57 496.4268 −9.6977 0.0034 16.333 0.2343 0.0015 −4.9589 0.0001 0.2163 0.0110 57 502.3933 −9.7169 0.0046 8.667 0.2386 0.0021 −4.9313 0.0001 0.2212 0.0145 57 512.3882 −9.6888 0.0044 13− 0.2267 0.0019 −4.9842 0.0001 0.2115 0.0139 57 524.3675 −9.6926 0.0047 1 0.2389 0.0022 −5.0567 0.0001 0.2199 0.0156 57 525.3946 −9.698 0.0047 −1.333 0.2416 0.0022 −5.0498 0.0001 0.214 0.0150 57 760.5835 −9.6862 0.0031 15.167− 0.2227 0.0013 −4.9891 0.0001 0.217 0.0104 57 764.615 −9.6899 0.0029 7.167 0.2223 0.0012 −4.9965 0.0001 0.2174 0.0101 57 766.6183 −9.7047 0.0028 5.333 0.2258 0.0011 −5.0322 0.0001 0.2127 0.0010 57 768.5861 −9.6921 0.003 8.167 0.2228 0.0012 −4.9964 0.0001 0.2254 0.0105 57 770.6187 −9.7065 0.0032 3 0.2263 0.0013 −5.0235 0.0001 0.2221 0.0111 57 771.6751 −9.6978 0.003 18 0.2226 0.0012 −5.0034 0.0001 0.2163 0.0101 57 815.5618 −9.7035 0.0027 11.833 0.222 0.0011 −5.0024 0.0001 0.2122 0.0093 57 858.4545 −9.7078 0.0027 10.333 0.232 0.0010 −4.996 0.0001 0.2051 0.0092 57 861.4145 −9.7003 0.0039 19.5 0.2341 0.0017 −5.0434 0.0001 0.2229 0.0129 57 890.4053 −9.697 0.0026 13.5 0.2314 0.0009 −4.9607 0.0001 0.1997 0.0084 57 891.3779 −9.6905 0.0026 10 0.2285 0.0009 −4.9447 0.0001 0.2038 0.0086 58 071.6969 −9.6988 0.0031 15 0.2358 0.0012 −4.9848 0.0001 0.2123 0.0104 58 073.6999 −9.6926 0.0026 9.833 0.2326 0.0010 −4.9674 0.0001 0.2074 0.0092 58 074.6453 −9.6911 0.0027 21 0.2421 0.0010 −4.9242 0.0001 0.2085 0.0094 58 076.6457 −9.7075 0.004 3.167 0.233 0.0018 −4.9704 0.0001 0.2162 0.0130 58 078.6855 −9.6967 0.0034 2 0.2365 0.0014 −4.9643 0.0001 0.2133 0.0110 58 090.7094 −9.6895 0.0032 3.333 0.2317 0.0014 −4.941 0.0001 0.2131 0.0108 58 091.6594 −9.6944 0.003 4.333 0.2339 0.0012 −4.9547 0.0001 0.2121 0.0102 58 092.6222 −9.677 0.0031 7.833 0.2373 0.0013 −4.9128 0.0001 0.2173 0.0107 58 111.6548 −9.6941 0.0025 15 0.254 0.0010 −4.8508 0.0001 0.2127 0.0091 58 129.7054 −9.7063 0.0043 0.333 0.2243 0.0019 −4.9616 0.0001 0.2457 0.0146 58 130.6467 −9.6992 0.0029 14.667 0.2311 0.0011 −4.9647 0.0001 0.2151 0.0010 58 131.6165 −9.6911 0.0028 3.333 0.2299 0.0011 −4.9829 0.0001 0.211 0.0097 58 136.5933 −9.6916 0.0032 8.833 0.2296 0.0014 −4.9599 0.0001 0.2241 0.0111 58 142.6058 −9.6874 0.0026 11.167 0.2363 0.0010 −4.9272 0.0001 0.2125 0.0094 58 185.4633 −9.6842 0.0027 4.833 0.2277 0.0010− 999.99 0.0001 0.215 0.0093 58 186.4198 −9.6977 0.0026 8 0.2313 0.0010 999.99 0.0001 0.2103 0.0089 58 213.5248 −9.7063 0.003 15.833 0.2405 0.0011 999.99 0.0001 0.2094 0.0097 58 214.3723 −9.7034 0.0027 11.833 0.2334 0.0010 999.99 0.0001 0.2095 0.0092 58 215.3927 −9.6859 0.0032 14.833 0.2378 0.0013 999.99 0.0001 0.2147 0.0103 58 257.4077 −9.6952 0.0032 10.667 0.2226 0.0012 999.99 0.0001 0.2029 0.0099 58 258.3895 −9.6947 0.0027 5.5 0.2254 0.0010 999.99 0.0001 0.2029 0.0090 58 262.4213 −9.7012 0.003 11.5 0.2231 0.0011 999.99 0.0001 0.2035 0.0097 58 263.3966 −9.7051 0.0029 7.667 0.2248 0.0011 999.99 0.0001 0.2006 0.0093 −

A18, page 7 of8 A&A 625, A18 (2019) 20) - , − U U U U U U U U − (0 U : truncated TN : normal, or N 147.57 to 135.37] − 9700.136 to 9694.614] − : uniform, U 143.08 to 135.84] [ − 9699.075 to 9695.588] [ − 140.16 to 136.66] [ − 9698.165 to 9696.485] [ − Star Noise Offset Gl 378b 138.07 [ 9697.333 [ − − is given at reference epoch: 2 455 500.0 BJD. 0 λ 138.41 1.85 9697.329 0.864 − − 137.44 9697.317 − − 135.12 0.098 0.000 0.132 0.103 0.109 [0.032–0.234] [0.004–0.393] [0.000–0.595] 9697.488 − − ) 8.16 7.91 7.96 1.24 7.96 [6.73–9.20] [5.51–10.47] [4.19–11.95] ) ) 4.073 4.521 4.664 0.806 4.610 [3.874–5.441] [3.199–6.417] [2.658–7.737] ) 0.00003982 0.00003913 0.00003914 0.00000607 0.00003911 [0.00003306–0.00004521] [0.00002699–0.00005133] [0.00002020–0.00005759] – ) 13.26 13.03 13.03 2.02 13.02 [11.01–15.05] [8.99–17.09] [6.73–19.17] – ) 0.5458 0.5591 0.5600 0.0100 0.5599 [0.5500–0.5699] [0.5399–0.5801] [0.5297–0.5900] 1 1 1 ) 0.04172 0.04100 0.04100 0.00636 0.04098 [0.03464–0.04736] [0.02828–0.05378] [0.02116–0.06033] – J ) 218.4 244.6 197.0 90.7 210.6 [93.7–290.4] [13.1–346.8] [0.8–359.4] ) 203.8 197.3 190.8 76.3 192.3 [113.5–270.3] [26.3–337.0] [1.1–358.6] − − −

⊕

◦ ◦ ( ( M (d) 3.82264 3.82248 3.82249 0.00135 3.82248 [3.82119–3.82381] [3.81973–3.82522] [3.81806–3.82698] M M M ( (AU)(AU) 0.000002853 0.000002750 0.039102 0.000002756 0.000000426 0.000002755 0.039416 [0.000002329–0.000003180] [0.000001904–0.000003612] [0.000001437–0.000004076] 0.039436 0.000235 0.039435 [0.039200–0.039668] [0.038960–0.039906] [0.038709–0.040128] – ( ( ( (mas) 66.8313 66.8364 66.8405 0.0322 66.8402 [66.8083–66.8721] [66.7762–66.9052] [66.7465–66.9380] (BJD)(BJD) 2455501.200 2455500.16 2455502.749 2455502.399 2455500.42 0.918 2455500.00 2455502.564 1.09 [2455501.511–2455503.286] [2455500.131–2455503.723] [2455500.011–2455503.815] 2455500.02 – [2455498.72–2455501.29] [2455498.18–2455501.83] [2455498.09–2455501.91] – (m s (m s errinc3) Parameters probed by the MCMC used to fit the RV measurements of Gl 378. − S S P 0 C S e a P ω K m m m λ a The maximum likelihood solution, median, mode, and standard-deviation of the posterior distribution for each parameter are shown, as well as the 68.27, 95.45, and 99.73% confidence T T Π M Avg. act. (m s (Likelihood) Parameter Units Max(Likelihood) Mode Mean Std Median 68.27% 95.45% 99.73% Prior log NAIRA(DRS γ Appendix B: MCMC parameters In this appendix we present the parameters probed by theTable B.1. MCMC analysis that was applied to the radial velocities of Gl 378. Notes. intervals. Parameters with priors listed are fitting parameters, while the rest are derived. The prior for each parameter can be of the following types: normal. Priors without given ranges are improper. The mean longitude

A18, page 8 of8 80 Chapter 2. M-dwarf RV search with the SOPHIE spectrograph

2.5.1.3 Other planets

Gl 411 b A temperate super-Earth (2.99 M⊕) with a 12.95 d orbital period. This M1.9 star was originally chosen as a standard star for the SP3 programme due to its brightness and low variability; the subsequent 157 measurements over seven years revealed the low-amplitude (K=1.49 m s−1) signal. The phase-folded curve is shown in Fig. 2.8. Gl 411 b is the third closest conﬁrmed exoplanet at 2.55 pc, making it an excellent candidate for atmospheric characterisation with next-generation 30 m telescopes. It was published in Díaz, Delfosse, Hobson et al. (2019).

Figure 2.8– Phase-folded curve and residuals to the ﬁt for Gl 411 b. The black and red curves represent the MAP-estimate and mean Keplerian model over the posterior samples obtained with the MCMC algorithm, respectively. Extracted from Díaz et al. (2019).

Three unpublished planets A further publication on a new planet is in preparation (Delfosse et al. in prep). Planet XXX b is an 11 M⊕ minimum mass planet with a 20 d orbital period. The remaining two planets still require additional observations in order to properly constrain their parameters, and are currently observed in priority. Planet YYY b is a 13 M⊕ minimum mass planet with a 25 d orbital period, planet ZZZ b is a 7 M⊕ minimum mass planet with a 16 d orbital period. 2.5. Summary of SOPHIE results on M dwarfs 81

2.5.2 Conﬁrmation or non-conﬁrmation of published planets

With the increase of surveys focused on M dwarfs (see Sect. 1.3.2), detecting exoplanets around M dwarfs is becoming a highly competitive domain. Most if not all the surveys will have many targets in common, and compete for first detection with increasing frequency. The positive side of these overlaps is that they allow the independent confirmation (or, in some cases, refutation) of planets, and a better constraint of the orbital parameters. For three of the SP3 targets, planets have been published by other teams. We are able to confirm two of these, GJ 687 b and Gl 686 b, using SOPHIE data. On the other hand, our measurements are not in agreement with the published planet around Gl 15A.

GJ 687 b Burt et al.(2014) published a 18.394 ± 2.167 M⊕ planet with a 38.140 ± 0.015 d orbital period around the M3 star GJ 687, using APF and Keck/HIRES data. We have 72 SOPHIE RVs for this star, obtained between 2011 and 2016, whose periodogram shows a strong signal below 0.1% FAP at 38 d. A keplerian fit to this signal on the SOPHIE data alone results in a planet with a 38.18 ± 0.03 d period and 19.3 M⊕ mass, consistent with the published parameters. Combining the SOPHIE and HIRES data, the best-fit keplerian is for a planet with a 38.155 ± 0.005 d period and 16.7 M⊕ minimum mass. No significant signals are seen in the residuals. The combined phase-folded curve is shown in Fig. 2.9.

Figure 2.9– Phase-folded curve for the combined GJ 687 data. SOPHIE RVs are in purple, HIRES RVs from Burt et al.(2014) in orange. The ﬁtted keplerian curve is overplotted in grey.

Gl 686 b Affer et al.(2019) published a 7.1 ± 0.9 M⊕ planet with a 15.53209 ± 0.00167 d orbital period around the M1 dwarf Gl 686, using HIRES, HARPS, and HARPS-N data. Like Gl 411, Gl 686 was monitored as a potential standard star for the SP3, so that we have 177 measurements between 2011 and 2019. The periodogram shows a highly significant signal at 15.5 d; a keplerian fit results in a 15.54 ± 0.01 d period and a 4.95 M⊕ 82 Chapter 2. M-dwarf RV search with the SOPHIE spectrograph mass. The period is therefore consistent; the mass is low, but still within 3σ of the published mass. It is also worth noting that a recent confirmation by CARMENES (Lalitha et al., 2019) also found a somewhat lower mass of 6.64 ± 0.54 M⊕, and a non-zero eccentricity of 0.077. Combining the data from HIRES, HARPS, HARPS-N, CARMENES, and SOPHIE, the best-fit keplerian corresponds to a 5.91 M⊕ planet at 15.5294 ± 0.0008 d period (time series and phase-folded data in Fig. 2.10). Like Lalitha et al.(2019), we find a non-zero eccentricity, in this case of 0.18 ± 0.03. The SOPHIE data bridge a 2- year gap between the older HIRES/HARPS points and the newer CARMENES/HARPS-N RVs.

(a) Timeseries for the combined Gl 686 data.

(b) Phase-folded curve for the combined Gl 686 data.

Figure 2.10– Time series and phase-folded curve for the combined Gl 686 data. Except for the SOPHIE RVs, the data was taken from Lalitha et al.(2019). The ﬁtted keplerian curve is overplotted in grey.

Within the residuals of the combined dataset after ﬁtting out Gl 686 b, I ﬁnd multiple signals below 0.1% FAP: three aliased peaks at 36.8 d, 41.4 d, and 45.9 d, and several 2.5. Summary of SOPHIE results on M dwarfs 83 longer-period peaks (268 d, 545 d, 1170 d, 1954 d). The three aliased peaks are at similar periods to signals noted by Lalitha et al.(2019) in the different RV and stellar activity datasets, and are probably caused by stellar activity. Of the long-period peaks, the most important is the one at 1170 d, which is compatible with a signal noted by Lalitha et al. +53 (2019) at 1163−81 d.

Gl 15A b A planet with a minimum mass of 5.35 ± 0.75 M⊕ and an orbital period of 11.4433 ± 0.0016 d was announced in Howard et al.(2014) around the M2 dwarf Gl 15A, based on HIRES data. We have 53 SOPHIE measurements, taken between 2012 and 2016, whose periodogram shows no signal below 10% FAP at any period (Fig. 2.11). Attempts to force a keplerian ﬁt to the parameters of Howard et al.(2014) did not converge. Recently, Trifonov et al.(2018) analysed both CARMENES and HIRES data for this star, ﬁnding that the 11.4 d signal appears only in early HIRES data (2009-2011), but not in later HIRES data (2012-2014) or in the 2016-2017 CARMENES data. They therefore suggest the 11.4 d signal may have been due to intense stellar activity in the 2009-2011 period, noting peaks in the S-index and H-index. The non-detection of the signal in SOPHIE data supports the idea that the 11.4 d signal is likely not a planet, though we note that we see no particular evidence of stellar activity from SOPHIE spectra in either the full time series or the 2012-2014 data.

Figure 2.11– Periodogram for the SOPHIE RVs of Gl 15A computed through template- matching. No signiﬁcant signals are in evidence.

2.5.3 Stellar activity mitigation

In order to detect planets through radial velocities, we seek to measure periodic variations, as described in Sect. 1.3. However, these periodic variations can be mimicked by stellar activity, inducing quasi-periodic variations at the rotation period or harmonics thereof (see, e.g., Saar & Donahue 1997, Santos et al. 2000, Boisse et al. 2011, Reiners et al. 2013, Oshagh et al. 2017). Therefore, it becomes necessary to quantify the stellar activity. Several spectroscopic activity indicators are available. The majority measure the ﬂux in certain spectral lines that are known to vary with stellar activity in the chromosphere. Periodic variations of these indicators, at the same period as a signal in the RV 84 Chapter 2. M-dwarf RV search with the SOPHIE spectrograph time series, point to the signal being induced by the star’s activity instead of by an orbiting planet. The principal such indicators in the visible are listed below.

The link between the Ca II H and K emission lines (at 396.9 nm and 393.4 nm) and chromospheric activity is well documented (see e.g Noyes et al. 1984 and references 0 therein). The log(RHK) index, deﬁned by Noyes et al.(1984), quantiﬁes this link using the Mount Wilson S index (Vaughan et al., 1978) and the B − V colour. However, their calibration is valid only for stars with B − V > 1.2, and therefore not suitable for M 0 dwarfs. A new calibration of the log(RHK) index based on V − K colours, and applicable to M dwarfs, was proposed by Astudillo-Defru et al.(2017a). This calibration was built from HARPS M dwarf spectra; for ten of the stars used, we have SOPHIE spectra, enabling a direct comparison. Figure 2.12 shows the HARPS and SOPHIE median 0 log(RHK) values for these stars. There is a very good correspondence between the values, with most falling on the identity line within error bars. Therefore, I concluded that the calibration is apt for SOPHIE spectra. Finally, the S index for SOPHIE was calibrated in Boisse et al.(2010).

0 Figure 2.12– Median log(RHK) index values from HARPS and SOPHIE spectra, for ten M dwarfs that are in common between the sample used by Astudillo-Defru et al. (2017a) and the SP3 sample. There is an overall good correspondence between the values from the two instruments.

The Hα absorption line at 656.281 nm is known to be affected by stellar activity, 2.5. Summary of SOPHIE results on M dwarfs 85 with chromospheric emission producing a brightening or filling-in of the core of the line (e.g. Zarro & Rodgers, 1983; Pasquini & Pallavicini, 1991). The Hα index quantifies this filling-in as a relative flux quotient between the line and nearby continuum regions. It was first defined by Kürster et al.(2003), and modified by Bonfils et al.(2007). For SOPHIE, suitable spectral regions to compute this index were defined by Boisse et al. (2009). Studies have shown that for low-mass stars, the behaviour of the Hα line varies with level of activity and spectral type. For increasing magnetic activity, the lines will first become deeper in absorption, then develop emission peaks on their wings, and finally become emission lines (Cram & Mullan, 1979). Likewise, the emission in Hα correlates with that in Ca II H and K only above a certain threshold (Rauscher & Marcy, 2006). Regarding spectral types, West et al.(2004) found (using SDSS spectra for classification, L and the Hα/Lbol luminosity ratio measured from these spectra as an activity indicator) that the activity strength is approximately constant for M0-M5 stars, and declines from M6 onward. Since the vast majority of the SP3 sample are early to mid M stars, with no stars later than M6, this should not have an impact here. 0 While the Hα and log(RHK) indices are the most-used spectral activity indicators, others have also been proposed. The Na I index (D1 and D2 lines at 589.592 nm and 588.995 nm respectively), first defined by Díaz et al.(2007) and modified by Gomes da Silva et al.(2011), and the He I index (587.562 nm), defined by Boisse et al.(2009), were both tested by Gomes da Silva et al.(2011) for a sample of M dwarfs using HARPS spectra. They found the Na I index correlated well with stellar activity, while the He I index showed small variability and a poorer correlation. In addition to changes in the flux of specific spectral lines caused by chromospheric activity, the overall shape of the line profiles can also be affected by photospheric activity. This change in line profile can be quantified by the bisector of the CCF, taking into account that the CCF is effectively a mean line profile (Queloz et al., 2001; Dall et al., 2006; Boisse et al., 2009). In this case, an anticorrelation between the CCF bisector span and the radial velocities indicates a potential stellar origin. However, this correlation is weaker for slow rotators (Santos et al., 2003; Desort et al., 2007), and also depends on the spectral SNR and the resolution of the instrument (Boisse et al., 2011). 0 For the SP3 sample, I computed the Hα, log(RHK), Na I, and He I indices, and the CCF bisector span. The first three indices trace chromospheric activity in the upper, lower, and middle-to-lower regions of the chromosphere respectively (Gomes da Silva et al., 2011), while the CCF bisector traces photospheric activity. The SOPHIE DRS provides the CCF bisector automatically, and codes are available to calculate the Hα 0 0 and log(RHK) indices. I modified the code that calculates the log(RHK) index to employ the new calibration of Astudillo-Defru et al.(2017a), and implemented the calculation of the Na I and He I indices. When analysing the RV time series for each target, I also inspected the activity indicators to determine whether the observed periodicities could be caused by stellar activity. Examples of this analysis can be seen in Hobson et al.(2018a) (Sect. 2.5.1.1) for Gl96 and Gl617A, both of which show signals in the residuals periodograms after removing the planets, whose periods match signals in the activity indicators’ periodograms. 86 Chapter 2. M-dwarf RV search with the SOPHIE spectrograph

A crucial fact to note about the quasi-periodicities induced by stellar activity is that they are expected to be wavelength-dependent, in contrast to true planetary signals that are the same at all wavelengths. Therefore, we can expect that the importance of their signals in the periodogram could vary from the DRS RVs to the template-matching RVs, as the two reductions weight the orders and spectral lines differently. We first detected this effect for Gl270. Fig. 2.13 shows the periodograms for both reductions; for the DRS reduction, the 31.5 d signal is far more important than the 15.8 d signal, while for the template-matching RVs they are at similar levels. I tested re-running the template-matching separately on the redder and bluer orders, finding that there was effectively a colour dependency: the twenty bluer orders favoured the 15.8 d peak, while the nineteen redder orders favoured the 31.5 d peak. Looking at the activity indicators 0 (Fig. 2.14), we can also see very strong signals at 31.5 d in Hα and log(RHK). The rest of the indicators are not so clear, with only a hint of a signal at around 16 d for the Na I index, no significant signals whatsoever in the He I index, and only a weak anticorrelation between the RVs and the CCF bisector span (R2 = −0.23).

Figure 2.13– Periodograms for the template-matching (top) and DRS (bottom) reductions for Gl270. Two regions of interest, around 15.8 d and 31.5 d, are highlighted. The template-matching reduction shows similarly strong signals at 15.8 d and 31.5 d, while for the DRS reduction only the 31.5 d peak is signiﬁcant. 2.5. Summary of SOPHIE results on M dwarfs 87

Figure 2.14– Periodograms for the template-matching RVs and activity indicators 0 for Gl270. Top to bottom: RV, Hα, log(RHK), Na i, and He i. The 31.5 d region, which corresponds to a signiﬁcant signal in both template-matching and DRS RVs, is 0 highlighted. Both Hα and log(RHK) show clear signals at 31.5 d. 88 Chapter 2. M-dwarf RV search with the SOPHIE spectrograph

0 Given the strength of the signals in Hα and log(RHK), it seemed clear that the signals in the RV periodogram of Gl270 were caused by stellar activity. The question was then raised of whether this effect was also observed for other stars. To answer to this question, I searched all the SP3 periodograms for signals that differed between the DRS and template-matching periodograms, and analysed the activity indicators for these targets. 0 Likewise, I calculated an estimate of the rotation period PRot, using the log(RHK) − Prot relationship determined by Astudillo-Defru et al.(2017a). I found interesting signals in ﬁve targets of the sample: Gl270 as previously described, Gl104, the RV residuals of Gl96 and of Gl617A after ﬁtting out the respective planets presented in Hobson et al.(2018a), and Gl251. The signals are summarised in Table 2.4. In all cases, the activity signals are strongest or only present in Hα and 0 log(RHK), so only these are listed.

Table 2.4– Periods of the most signiﬁcant peaks in the periodograms of the DRS RVs, template-matching (T-M) RVs, and activity indices. 0 Star DRS RVs T-M RVs Hα log(RHK)PRot Spec. type Gl270 (16d)† 31d† 16d† 31d† 31d 31d 23.9 ± 2.5 M1 Gl104 31d∗ (22d†) 31d, 34d (31d, 34d) 39.9 ± 7.7 M3 (22d†, 31d∗, 34d∗†) 34d∗† Gl96 (r) 29d∗, 32d∗ 29d 29d 29d 29.6 ± 2.8 M2 Gl617A (10d)† 21d† 21d†, 30d† 21d 21d 28.8 ± 6.1 M1 (r) Gl251 (71d), 71d, (71d), >100d 100 ± 20 M3 >100d >100d >100d Brackets denote a period at lower signiﬁcance; a * symbol indicates 1-year aliases, a † symbol indicates harmonics.

For four of these stars (Gl270, Gl104, Gl96 residuals, and Gl617A residuals), the two RV periodograms favour signals at different aliases or harmonics of a specific period. In all cases, the period favoured by template-matching coincides with highly significant signals in Hα, and for three (Gl270, Gl96 residuals, Gl617A residuals) also in 0 log(RHK). The signals in Hα are all consistent with the estimated rotation period at the 3-σ level. Figures 2.15a and 2.15b show the CCF RVs, template-matching RVs, and Hα periodograms for Gl104 and Gl617A, as examples of periods at different aliases and in the residuals of a keplerian fit respectively. For the final star, Gl251, a significant signal in the template-matching RVs (and of moderate significance in CCF RVs) coincides with a moderate signal in Hα, but at 71d is beyond the typical range of periods in which Hα is considered a good activity tracer (Figure 2.15c). Likewise, both RV sets and Hα show a multitude of peaks at P>100d, which is the range where the estimated PRot also falls. The exact origin of these signals is therefore somewhat unclear, but it seems more likely to be stellar activity than a planet. There does not appear to be a spectral type dependency, with these stars going from M1 to M3 (which is the centre of the spectral type distribution of SP targets, as shown in Fig. 2.2c). 2.5. Summary of SOPHIE results on M dwarfs 89

(a) Periodograms for Gl104. (b) Periodograms for Gl617A.

Figure 2.15– Periodograms for Gl104, Gl617A, and Gl251: CCF RVs (top, blue), template- matching RVs (middle, green), and Hα index (bottom, purple) for each star. Regions of interest, corresponding to the periods listed in Table 2.4, are highlighted.

This analysis suggests that template-matching may be more sensitive to stellar activity signals than CCF, with more signiﬁcant signals appearing in the template-matching RV periodograms at the same periods as signals in activity indices. With surveys focusing on M dwarfs increasing, template-matching will undoubtedly see an increase in use; it is therefore crucial to understand any biases and limitations it may present. The importance of this is highlighted by disputes in the literature over potential stellar activity origins for several proposed planets found with template-matching pipelines, such as around GJ 667C (Anglada-Escudé et al., 2012; Robertson & Mahadevan, 2014; Anglada- Escudé et al., 2013) or around Kapteyn’s star (Anglada-Escude et al., 2014; Robertson 90 Chapter 2. M-dwarf RV search with the SOPHIE spectrograph et al., 2015; Anglada-Escudé et al., 2016). The heightened sensitivity of template-matching to stellar activity is likely due to the fact that the template is constructed from the entire spectra (discarding only telluric- affected zones), and is therefore likely to include more lines sensitive to stellar activity than the more selective CCF masks. Therefore, both periodograms should be generated whenever possible, and discrepancies between CCF and template-matching RV periodograms should be carefully analysed. This is especially important in the case of signals at different aliases or harmonics of the same period. 0 Finally, for SOPHIE M-dwarf spectra, the Hα and log(RHK) indices proved to be better activity indicators than the Na I and He I indices or the CCF bisector. With regards to the Na I and He I indices, this means that our data agrees with Gomes da Silva et al. (2011) in that He I is a poor activity indicator for M dwarfs, but does not agree with their support for Na I as an activity indicator. Regarding the CCF bisector, I searched Simbad for measured rotational velocities for the SP3 targets. I found measurements for 18 of the stars in the sample; all are relatively slow rotators, ranging from 0.46–3.21 km s−1 with a median value of 2.17 km s−1. Assuming these values are representative of the full sample, the targets probably rotate too slowly for the CCF bisector anticorrelation to be a strong indicator of stellar activity.

2.6 Other contributions to SOPHIE RV programmes

While the core of my work with SOPHIE was in the context of the SP3 subprogramme, I was also able to make some other contributions to the work of the exoplanet consortium. This was facilitated by my supervisors, who have lead the SOPHIE high- precision proposals as PIs.

2.6.1 High precision RV search for super-Earths

One of the exoplanet programmes led by the SOPHIE exoplanet consortium is the high precision search for super-Earths around sun-like stars, known as subprogramme 1 or SP1 (Bouchy et al., 2009a). This programme aims to detect planets with masses lower than 0.1 MJ, and targets a sample of 190 inactive bright solar-type stars. The stars were pre-selected using information from the ELODIE survey and from another SOPHIE subprogramme consisting in a volume-limited search for giant planets. Like the SP3 subprogramme, the SP1 aims for 30 observations before deciding whether to stop or continue observing a target, according to whether or not there is any emerging evidence of periodic signals. Observations are also stopped if the star shows strong stellar activity. It was shown by Courcol et al.(2015) that in order to reach the precision required over long timescales, the long-term drift of the spectrograph must be measured and corrected. Using the code developed by B. Courcol (Courcol et al., 2015), I performed periodic updates of the SP1 database with the long-term drift applied. The long-term 2.6. Other contributions to SOPHIE RV programmes 91 drift correction is shown in Fig. 2.16. It is analogous to the correction generated for M dwarfs in Sect. 2.4.3, but is created using DRS CCF radial velocities for 44 G and K dwarfs (out of a 200-target sample). In particular, the use of the CCF-generated RVs means that the offset at BJD = 56775 due to wavelength solution instabilities is clearly in evidence. Later offsets at BJD = 57665, BJD ∼ 58050, not seen in the M dwarf correction, are probably of similar origin.

Figure 2.16– Correction of the long-term variation of the zero-point for sun-like stars (red line), together with the CCF-generated radial velocities used to construct it (black dots).

With the corrected RVs, I next generated a summary document of the targets, con- 0 taining the time series of the RVs, CCF bisectors, and log(RHK) values; the RV peri- 0 odograms; and RV-bisector and RV-log(RHK) correlation plots, as well as brief comments on targets of interest. This involved a manual inspection of the updated periodograms, to verify the persistence of known planetary signals and identify emerging ones. 92 Chapter 2. M-dwarf RV search with the SOPHIE spectrograph

2.6.2 Simultaneous FP background correction

The SOPHIE DRS possesses routines to correct the extracted science fibre from contamination due to the calibration fibre, depending on if the calibration fibre is on the sky or on the ThAr lamp. These routines are employed in the activity index calculations. I found that both routines proved inadequate to correct the contamination in the case of a simultaneous FP exposure, significantly overestimating the true background (Fig. 2.17). This is due to the fact the FP proved to have no true "dark" level; even the floor level between two peaks is illuminated.

Figure 2.17– Existing SOPHIE background correction routines applied to a simultaneous FP exposure (order 0): In red, the FP spectrum; in green, the stellar spectrum; in blue, the background routine used to correct for simultaneous ThAr; in pink, the background routine used to correct for simultaneous sky. The X axis indicates pixel values, the Y axis ﬂux level. Both background corrections are clearly inadequate, as the stellar lines drop well beneath them.

Consequently, the activity indices for simultaneous FP exposures were incorrectly calculated. It was therefore necessary to create a new calibration based on direct measurements: I requested DARK-FP frames (exposures with no illumination in the science fibre, and FP spectra in the calibration fibre) to measure the contamination on the science fibre from the FP spectrum. For SOPHIE simultaneous FP observations, a neutral density wheel is used to preserve the flux level of the FP approximately constant for varying observation times. A set of DARK-FP frames for representative exposure times (300 s, 600 s, 800 s, 900 s, 1800 s) and corresponding neutral density levels showed the FP 2.6. Other contributions to SOPHIE RV programmes 93

ﬂux is not exactly constant, but increases slightly with exposure time; the DARK level also changes with exposure time (in a non-linear but consistent fashion, as shown in Fig. 2.18). Therefore, I constructed a "master" correction that is weighted by the FP ﬂux of the simultaneous calibration, and by a scaling factor per order related to the observed difference in behaviour between the DARK and the FP with exposure time (plotted in Fig. 2.18b).

(a) median ﬂux quotient per order for darks (b) Dark-to-FP scale factor per order for (dashed, points) and FPs (dotted, carets). each exposure time (dotted lines), and Each colour corresponds to an exposure the median (solid black line). time, as indicated in the legend. The 1800s frame is the reference point for the quotients.

Figure 2.18– Comparison of the dark and FP behaviour with exposure time per order.

The resulting activity indices for simultaneous FP exposures show much better consistency with the ones for simultaneous ThAr exposures. An example is shown in Fig. 0 2.19), displaying the log(RHK) values obtained using different background corrections for Gl411 spectra from the last three observing seasons with S/N>10 in order 1 (where 0 the Ca II H and K emission lines are located). The median log(RHK) values for the different series are summarised in Table 2.5. The ﬁrst of these seasons was observed with simultaneous ThAr (labelled THsimult) and is shown as a reference point. For the simultaneous FP exposures (labelled FPsimult), the results for a ThAr-like background correction (Th back), no background correction (no back), and the newly developed FP background correction (new back) are shown. The simultaneous sky background correc- 0 tion was also tested, but removed so much of the Ca lines that the log(RHK) could not be calculated. It is clear from both the time series plot and the median values that the indices using the new background correction are closer to the values for the simultaneous ThAr exposures than the rest. 94 Chapter 2. M-dwarf RV search with the SOPHIE spectrograph

0 Table 2.5– Median log(RHK) for diﬀerent background corrections, for the last three observing seasons on Gl411. THsimult indicates the observations with simultaneous ThAr, FPsimult with simultaneous FP. 0 Correction Median log(RHK) THsimult -5.6156 FPsimult - Th back -5.8731 FPsimult - no back -5.4541 FP simult - new back -5.6624

0 Figure 2.19– Example of log(RHK) for diﬀerent background corrections, for Gl 411. The inset shows a zoom of the values for the 2018A observing season, when the change from simultaneous ThAr to simultaneous FP took place. The values computed with the new FP background correction are far more consistent with the ones for simultaneous ThAr exposures than the rest. 3 Development of the SPIRou nIR spectropolarimeter

Contents

3.1 The SPIRou spectropolarimeter...... 96 3.2 The SPIRou Data Reduction System...... 99 3.3 Validation tests and commissioning...... 102 3.4 Wavelength calibration development ...... 103 3.4.1 Hollow-cathode lamps ...... 103 3.4.2 Combination with Fabry-Pérot reference ...... 107 3.5 Validation and performances of the wavelength solution ...... 110 3.5.1 Impact of previous calibrations...... 111 3.5.2 Performances of the HC wavelength solution ...... 114 3.5.3 Combined HC-FP wavelength solution ...... 115 3.5.4 Impact on RV error ...... 117 3.5.5 Upcoming changes to the Data Reduction System ...... 119 3.6 SPIRou science programs ...... 120 3.6.1 Spirou Legacy Survey ...... 120 3.6.2 Synergy with SOPHIE...... 122 3.6.2.1 Open-time proposal ...... 123 96 Chapter 3. Development of the SPIRou nIR spectropolarimeter

3.1 The SPIRou spectropolarimeter

SPIRou a (SpectroPolarimetre InfraRouge) is a near-infrared spectroplarimeter, mounted on the 3.6 m Canada France Hawaii Telescope (CFHT), and in operation since February 2019. SPIRou was developed by a large international consortium, led by France and in- volving multiple institutes in seven countries (France, USA, Canada, Switzerland, Brazil, Taiwan and Portugal). It covers the 0.98–2.35 µm spectral range (Y, J, H, and K bands) at a resolution R ≈ 75 000, thus providing a star’s entire nIR spectrum, and can be used in either circular or linear polarization. The instrument design (Fig. 3.1) is described in detail in Artigau et al.(2014); I highlight here some of the main characteristics: • 16-million-pixel Hawaii 4RG (H4RG) detector. • Double-pass ZnSe prism train cross disperser. • R2 diffraction grating. • In a cryogenic dewar cooled to 77 K (−200 ◦C) and thermally stable at a few mK. The cooling is required to minimise thermal contamination from the instrument itself on the spectra, while the thermal stability is necessary to avoid RV drifts (Sect. 1.3). • Special fluoride fibres, made of ZrF4 which enables good transmission into the K band (unlike off-the-shelf Si fibres). There are two fibres for the polarimetric science channels and one for reference calibrations. • Linked to the telescope (via the fibres) by a Cassegrain unit, containing an atmospheric dispersion corrector, a tip-tilt module, a calibration wheel, and an achromatic polarimeter. • A pupil slicer transfers the output beams from the fibres to the spectrograph. It is composed of a collimating parabolic mirror and four flat slicing mirrors. • A calibration module allows the illumination of science, calibration, or all channels simultaneously, by: A cold source (Black Acktar surface at −25 ◦C) for observation of faint stars without simultaneous calibration; A white lamp (LDLS) for blaze measurement; One of two hollow-cathode (HC) lamps for wavelength calibration; A Fabry-Pérot (FP) étalon for simultaneous calibration and drift monitoring; A future upgrade or visitor instrument (e.g. a laser frequency comb).

a. Instrument website: http://spirou.irap.omp.eu/ 3.1. The SPIRou spectropolarimeter 97

Figure 3.1– SPIRou instrument design. The spectrograph (including a parabolic mirror, prism train cross disperser, échelle grating, camera, and detector) is within a cryostat. The Cassegrain and calibration units are also shown. Adapted from Artigau et al. (2014).

A schematic view of the optical path, including the different fibres, is shown in Fig. 3.2. Light from the telescope will pass through the Cassegrain unit, including the injection calibration and polarimetric modules, then be conveyed by the science fibres to the telescope. Light from the calibration module may be injected via the Cassegrain unit to the science fibres, or directly to the spectrograph through the reference fibre. Within the spectrograph unit, the light from the fibres first passes through a pupil slicer, then through the slit. The beam from the slit is collimated by the parabola; makes a first pass through the prism train cross disperser, is diffracted by the grating, and passes through the cross-dispersing prisms a second time; is focused by the parabola onto the flat mirror, which folds it back onto the parabola, where it is collimated once more; and finally is sent through the refractive camera to the H4RG detector. 98 Chapter 3. Development of the SPIRou nIR spectropolarimeter

Figure 3.2– Schematic view of the SPIRou optical path, from the telescope/calibration unit to the detector. Taken from Boisse et al.(2016).

Two elements of the design in particular merit further description in the context of this thesis. The first is the detector. Unlike the HARPS and SOPHIE detectors, which are CCD (charge coupled device) sensors, the SPIRou detector is a CMOS (complementary metal oxide semiconductor) sensor. The main operational difference between CCD and CMOS devices lies in the readout. In CCDs, there is a singular output amplifier and converter, to which the charge of each pixel is conveyed by a series of transfers. In CMOS sensors, each pixel has its own converter and amplifier. This change in detector was mandated by the the move from visible to nIR spectroscopy, as CCDs are not sensitive in the nIR. The detector chosen is an H4RG-15 from Teledyne (Blank et al., 2012), with 4096 × 4096 15 µm pixels and readout noise < 5e−. The second element I wish to highlight is the calibration unit, which is described in Boisse et al.(2016). The calibration unit was built in collaboration by the Observatoire d’Haute-Provence and the Observatoire de Genève, with the latter providing the Fabry- Pérot etalon that constitutes the RV reference module (Cersullo et al., 2017). As my work with SPIRou focused on the calibrations, I was therefore able to collaborate with the engineers and technicians from the Observatoire d’Haute-Provence, especially in the context of the validation tests and commissioning (Sect. 3.3). A schematic of the calibration unit is shown in Fig. 3.3. The first element is the light sources module, with slots for the two HC lamps, fibre injections for the white light source and the FP, and a reserved slot for future upgrades. The parking position (no illumination) is also 3.2. The SPIRou Data Reduction System 99 indicated. The calibration unit possesses two fibre links to the spectrograph, as indicated in Fig. 3.2. One is injected into the Cassegrain unit, so that the light follows the same path as the science fibres. The other goes directly to the pupil slicer (after passing through the flux balance module of the calibration unit, which contains a continuous variable density wheel and is used to keep the same flux level for varying exposure times) and is used for the simultaneous drift measurement. The illumination for each of these fibres is selected by trolleys in front of the lamp slots, as illustrated in Fig. 3.3.

Figure 3.3– Schematic view of the SPIRou calibration unit, showing the diﬀerent modules. Taken from Boisse et al.(2016).

3.2 The SPIRou Data Reduction System

My main involvement with SPIRou was in the context of the development of the data reduction system (DRS). The SPIRou DRS is originally based on the pipelines developed for HARPS and SOPHIE (Pepe et al. 2004, Bouchy et al. 2009a); the development has been led by Marseille (LAM, I.Boisse and myself), Montreal (iREx, E. Artigau and N. Cook), and Geneva (Observatoire de Genève, F. Bouchy). The DRS is written in Python 3 (backwards-compatible with Python 2.7), maintained and version-controlled on Github b. The purpose of the DRS is to perform the complete reduction of spectral data, from raw frames to radial velocities and polarimetric products. Figure 3.4 shows a ﬂowchart

b. https://github.com/njcuk9999/spirou_py3/; private repository, access provided as needed. 100 Chapter 3. Development of the SPIRou nIR spectropolarimeter summarization of the structure of the code (as of version 0.5.000 c).

Figure 3.4– Flowchart illustrating the inputs and outputs of the SPIRou DRS routines, from pre-processing of the raw frames, through calibrations, to radial velocities and polarimetric products.

The first step in the reduction is the pre-processing; this is an entirely new step that did not exist in the HARPS/SOPHIE pipelines, but is required to correctly deal with the H4RG images. It applies to all files and is used to remove certain detector effects such as the amplificator crosstalk (routine cal_preprocess_spirou.py). Next, the calibrations are reduced, generating: • a dark map (cal_DARK_spirou.py); • a map of the bad pixels, including two small holes and a scratch on the detector (cal_BADPIX_spirou.py); • the localization of the orders for fibres A, B, and C (cal_loc_RAW_spirou.py); • a map of the slit profile across the detector (cal_SHAPE_spirou.py); • the blaze profiles for all extracted fibres (cal_FF_RAW_spirou.py);

c. DRS updates are tracked and numbered generally following the conventions described in https: //semver.org/. The first number indicates the overall version, and is currently set to zero to reflect that the DRS is still in a beta development phase. The second number indicates large milestone changes. The last number marks minor modifications and bug fixes. The version numbers are set by N. Cook, who maintains the DRS Github repository and master branch. 3.2. The SPIRou Data Reduction System 101

• the wavelength map for all extracted ﬁbres (cal_HC_E2DS_EA_spirou.py, cal_WAVE_E2DS_EA_spirou.py).

Two of the data products, the bad pixel map and the order localization, are shown in Fig 3.5. Finally, the stellar spectra are extracted and telluric-corrected, and the radial velocities (via the CCF method) and/or polarimetric products are generated.

(a) The bad pixel map produced by (b) The polynomial fits of the sci- cal_BADPIX_spirou.py. It is a binary ence orders (red curves) produced mask with bad pixels set to 1. The bad by cal_loc_RAW_spirou.py, superim- pixel fraction is 0.75%. The two holes posed on the spectrum used to produce at x ∼ 3500, y ∼ 2500 and the scratch at it (white lamp on the science fibres, dark x ∼ 2000 − 3500, y ∼ 200 can be distin- on the calibration fibre). guished.

Figure 3.5– Intermediate data products of the SPIRou calibration sequence: The bad pixel map (left) and the order localisation for the science ﬁbres (right). Note that the order localisation has been ﬂipped in x and y with respect to the bad pixel map.

I worked primarily on the wavelength calibration, which will be the focus of Section 3.4. I also assisted with the rest of the calibration sequence, participating to weekly DRS meetings, and fixing bugs that were reported (primarily by CFHT, in the course of routine reductions). While the SPIRou team assembles many experts in the construction of spectrographs and the coding of associated pipelines, this does not mean that the development has been a simple process. SPIRou is not simply a new instrument, it is a new concept with many challenges that are not seen in visible instruments such as HARPS. The detector is not a CCD but a CMOS; the orders are significantly curved (e.g. Fig. 3.5b), which also means the slit inclination varies strongly across the detector; thermal emission impacts the redder orders; etc. All these factors need to be first understood and then correctly handled by the pipeline. This is a process that is still ongoing, as improvements in one direction can (and often do) reveal suboptimal handling elsewhere. 102 Chapter 3. Development of the SPIRou nIR spectropolarimeter

3.3 Validation tests and commissioning

Prior to being shipped to Hawaii and mounted on the CFHT, SPIRou was first assembled and tested at IRAP, Toulouse. I assisted in the performance validation tests of the calibration unit, that are described in Perruchot, Hobson et al. (2018). The tests were carried out with an engineering-grade H2RG detector of 2048 × 2048 18 µm pixels; the final science-grade H4RG detector with 4096 × 4096 15 µm pixels was installed directly in situ at CFHT. Not only is the H2RG smaller than the H4RG, leading to the edges of the orders being cut off, it had ∼ 20% of bad pixels. These bad pixels posed significant difficulties to the DRS development; in particular, they introduced spurious edges that caused issues for the order localisation and the wavelength solution. Nevertheless, the H2RG provided a first look at true SPIRou data, and enabled the DRS team to proceed with the adaptation of the pipeline to the data. For the validation tests, I developed wavelength calibration routines, using either the hollow-cathode lamp alone or a combination of the hollow-cathode lamp and the Fabry-Pérot etalon. I also developed quick-look visualisation and measurement tools allowing the display of a raw image and a cross-section profile along a row or column, notably for use in estimating the level of flux in a particular order (e.g. for a lamp or due to thermal background). An example is shown in Fig 3.6.

Figure 3.6– Output of the quick-look visualisation for a white lamp (Tungsten) spectrum. The left-hand plot shows a zoom of the raw image; the red line marks the measurement row. The right-hand plot shows the ﬂux, calculated as the mean value on ±10 rows around the measurement row, excluding dead pixels.

After being validated at Toulouse, SPIRou was shipped to the CFHT. It arrived at Mauna Kea during late January 2018, and was integrated and tested during the first months of 2018. The first tests were done with the H2RG detector, in order to match the Toulouse configuration and verify the instrumental performances had not suffered through the disassembly, transport, and reassembly. Subsequently, the final H4RG detector was installed, and intensive testing began. First light on the H4RG was on the 24th of April 2018. 3.4. Wavelength calibration development 103

Together with most of the DRS team, I attended the ﬁrst commissioning with the H4RG at Hawaii. Having the team assembled on-site was crucial to providing a rapid response, which time zone offsets (6 hours between Europe and Montreal, 12 hours between Europe and Hawaii) otherwise make practically impossible. This rapid response was particularly important for these ﬁrst tests with the new detector. The difference between the two detectors meant that many DRS parameters had to be re-calibrated, and scripts re-adjusted. I focused mainly on the correction and improvement of the wavelength routines, but also assisted with diverse problems and bugs reported to the DRS team, as needed.

3.4 Wavelength calibration development

This section presents the different wavelength calibration routines I developed for SPIRou. The calibration unit possesses two types of lamps that can be used as wavelength calibrators: The hollow-cathode lamps, which provide absolute wavelength references, and the Fabry-Pérot étalon, which provides relative wavelengths to an absolute anchor. The routines using the hollow-cathode lamps alone are described in Sect. 3.4.1, and the incorporation of the Fabry-Pérot in Sect. 3.4.2.

3.4.1 Hollow-cathode lamps

The SPIRou calibration unit possesses two slots for hollow-cathode (HC) lamps; the ones currently installed are a UNe lamp and a ThAr lamp. The pre-processed H4RG spectra are shown in Fig. 3.7. These lamps provide an absolute wavelength calibration, offering spectral lines (whose wavelengths have been catalogued and are therefore in principle known) across the entire detector. During the validation tests, I examined extracted spectra for both lamps, manually identifying the main catalogued peaks by using the estimated order coverage in order to create a ﬁrst-guess wavelength solution. This required a selection of wavelength catalogues. At the start of this thesis, the only catalogue of UNe lines in the nIR was that of Redman et al.(2011) (hereafter R11), with 9767 lines in the SPIRou wavelength range. This catalogue was therefore used in the initial development of the HC wavelength solution during validation tests. Last year, however, Sarmiento et al.(2018) (hereafter S18) published an updated catalogue of U lines in the 500-1700 nm wavelength range. While this is far from covering the full wavelength range of SPIRou, it cross-matches well with the catalogue of R11 (Fig. 3.8) and provides a substantial increase in the range it does cover, incorporating 3787 new lines. For the ThAr lamp, the only catalogue that covered the SPIRou domain at the time is that of Redman et al.(2014). The UNe catalogue has far more lines than the ThAr catalogue, even before the incorporation of the lines from S18 (9767 vs 1587 in the SPIRou wavelength range). 104 Chapter 3. Development of the SPIRou nIR spectropolarimeter

Figure 3.7– Pre-processed SPIRou spectra of the two HC lamps, ThAr (left) and UNe (right), taken with the H4RG detector. The exposure time is 600s for both, and the colour scales have been matched.

Figure 3.8– Comparison of wavelengths for common lines between the catalogues of Redman et al.(2011) (R11) and Sarmiento et al.(2018) (S18). The blue points represent the diﬀerence in wavelengths, the black lines the S18 error bars. Points without error bars indicate that S18 reported no uncertainty for that line. The diﬀerences in wavelength are generally within the error bars when known, and always small. 3.4. Wavelength calibration development 105

During the line identification process, I found that both lamps presented numerous lines that were not identified in the R11 or R14 catalogues. Many more lines were identified in the UNe spectra than in the ThAr spectra (as is logical given the difference in the catalogues), and preliminary wavelength solutions were more accurate for UNe than for ThAr. Additionally, the ThAr spectra show many more, and more strongly, saturated lines than the UNe spectra. Saturated lines are useless for fitting a wavelength solution, since their centres cannot be precisely determined; worse, if they saturate in a single detector readout their cores will be set to NaNs, introducing spurious edges. They will also ’bleed’ into neighbouring orders and contaminate them. Therefore, the UNe lamp was adopted as the primary absolute wavelength calibrator. Two main methods were developed and tested for the wavelength solution based on the HC lamp. The principal conceptual difference between them lies in the way in which the HC lines are identified. The first, method HC1, is analogous to the SOPHIE/HARPS wavelength solution: for each line in the catalogue, the region where it should be located is selected and a gaussian fit is attempted, with poor fits being discarded. In the second, method HC2 (developed in collaboration with E. Artigau), gaussians are fitted to every peak in the HC spectra, and the best match to the catalogue is identified. The detailed structure of the full routine is as follows. Data identification and reading: The input files are verified via FITS header keys to be extracted two-dimensional spectra (E2DS) corresponding to a HC lamp. The data and header are read, and the lamp and fibre are identified. If more than one HC file is given, the routine verifies they all correspond to the same lamp and fibre, then they are added together. Calibration set-up: The calibration files that are closest in time to the input file(s) are copied from the calibration database (including the previous wavelength solution). The previous wavelength solution is read and checked for compatibility with the current parameter set-up (order of the polynomial fits employed). The correct HC line catalogue is read in. Identification of the HC lines: The lines are identified following one of the procedures outlined above: • Method HC1: For each line in the catalogue, a 4σ region in wavelength around its expected position is selected, based on the previous wavelength solution. A gaussian fit to the region is attempted, and kept if the σ of the fitted gaussian is less than 4 (to discard very broad, flat lines whose centres will be imprecise), and the amplitude is less than 1e8 (to avoid fitting on saturated lines). • Method HC2: For each spectral order, a 13-pixel-wide window is moved along the order in four-pixel shifts. If the maximum flux is at least at a 2σ level above the local RMS and is within four pixels of the centre of the segment, a gaussian fit is attempted. The gaussian is kept if the residual of the fit normalised by the peak value is between 0 and 0.2, and the equivalent width in pixels is between 0.7 and 1.1. Once all the peaks in the order are identified, the 20 brightest (which are generally the least likely to be spurious, and most likely to be catalogued) are selected and the closest catalogue line to each one is identified. Then, all possible three-line combinations are used to fit a second-order polynomial to the 106 Chapter 3. Development of the SPIRou nIR spectropolarimeter

order, test wavelengths are calculated for all lines using the polynomial fit, and the velocity offset for each from its identified catalogue line is calculated. The polynomial with the most lines within 1 km s−1 of the catalogue is held to be the correct identification, and all peaks within 1 km s−1 of the catalogue are kept. Fitting the solution: With the HC lines identified, a fourth-order polynomial fit is performed for each order between the pixel positions given by the centres of the gaussians, and the catalogued wavelengths. In method HC2, continuity of the polynomial coefficients across the orders is also imposed at this step (attempts to impose similar cross-order continuity for method HC1 resulted in less stable solutions, so they were discarded). Littrow solution check: A verification of the cross-order continuity of the wavelength solution. It is evaluated for several pixel positions (currently every 500 pixels). For each position x, a fourth-order polynomial fit is performed between the inverse echelle orders and the fractional wavelength contribution at x for each order (normalised by the wavelength for the first order), with an iterative step to remove the largest outlier. For an ideal spectrograph, the residuals to these fits would be zero. The polynomial fits, the residuals, and the minimum, maximum, and rms values of the residuals are stored. Extrapolation of the reddest orders: For the last two orders, very few HC lines are catalogued (13 in the last order and 68 in the penultimate, compared to an average of 300 for the rest) and even fewer are identified (around ten in the last order and 25 in the second-to-last for method HC2). This means that the fitted solution often fails or is highly unstable. Therefore, for these orders it is not fitted from the HC lines, but extrap- olated from the Littrow solution: the cross-order polynomial fits are used to generate pixel-wavelength pairs at the Littrow evaluation positions, and these values are used in turn to fit a fourth-order polynomial for each spectral order. Quality controls: The structure of an E2DS file (one order per row, wavelength increasing along the order) means that along each column, the wavelength must be increasing - that is, each pixel of order N must have a smaller wavelength value than the corresponding pixel of order N+1. A first rough quality check, therefore, verifies that this is in fact the case (when this fails, it generally points to a problem with the slit shape or order identification calibrations). A second quality check is applied to the minimum, maximum, and rms values of the Littrow check residuals. Logging statistics: The mean and rms of the deviation from the catalogued lines in m s−1 is logged, as is the total number of lines used to fit the solution. Most importantly, the internal precision of the solution is determined, as the rms of the residuals of the fit to the catalogued lines, divided by the number of lines used to generate the fit. Saving the solution: The wavelengths per pixel generated from the polynomial fits for each order are saved as an E2DS file. The coefficients of the polynomial fits are stored in the header. If the quality controls were all passed, the new wavelength solution is copied to the calibration database, and to the header of the input HC E2DS spectra. 3.4. Wavelength calibration development 107

Performance tests (discussed in more detail in Sect. 3.5) showed that while the first method has slightly better internal accuracy than the second, it is much less stable night-to-night. In any case, neither method reaches the expected accuracy budgeted for the SPIRou wavelength solution. The first method is also very sensitive to drifts from the initial wavelength solution (this was particularly highlighted by the earthquake in May 2018, which caused a detector shift of several pixels), to which the second method is more robust. The second method was therefore adopted for the HC wavelength solution, and is the only one available in version 0.5.000 of the DRS. During the SPIRou validation and commissioning tests, it became clear that the HC lamps alone did not provide a sufficiently accurate and stable wavelength solution, with internal accuracy measurements of ∼2–4 m s−1 compared to the 0.45 m s−1 accuracy demanded by the SPIRou error budget for an overall 1 m s−1 RV precision. There are likely several contributing factors to this lack of accuracy: the low flux levels in the edges of the bluest orders, the low number of lines found for the reddest orders, imprecision in the catalogue wavelengths, among others. The next step chosen, therefore, was to combine the HC lamps with the Fabry-Pérot (FP) etalon spectra.

3.4.2 Combination with Fabry-Pérot reference

The design of the SPIRou Fabry-Pérot etalon is described in Cersullo et al.(2017). Its spectrum provides a wealth of lines across the entire detector, as shown in Fig. 3.9, whose spacing is a priori known since it is given by the FP equation (though we will see that for a physical as opposed to an ideal etalon, this is not entirely accurate as the cavity width is wavelength-dependent). The absolute wavelengths of the FP lines, however, are not known but must be determined from some other source. Therefore, by anchoring the FP line wavelengths to the HC lines, we aim to derive a more precise wavelength solution than could be obtained with the HC alone. In principle, the wavelength of a FP line is given by the FP equation:

2d λ = (3.1) m m where λm is the wavelength of the line of (integer) line number m and d is the effective cavity width of the interferometer. In an ideal FP interferometer, d is constant (and known, since it is set by the manufacturer); therefore, once the wavelength λm,r is known for a specific reference line, its line number mr can be determined. Then, for any other line, the line number can be determined simply by counting from the reference line, and its wavelength calculated with Eq. 3.1. However, in real FP étalons, the cavity width d is not in fact constant. Bauer et al.(2015) showed it to be wavelength- dependent, due to a varying penetration depth; photons of different energy (i.e., different wavelength) will penetrate the soft coating to different depths. This wavelength dependence needs to be calibrated in order to allow the use of the FP lines in a wavelength solution. Here, again, two methods were developed. The first, method FP1, follows the approach described by Bauer et al.(2015); we first use the HC lines to generate a rough 108 Chapter 3. Development of the SPIRou nIR spectropolarimeter

Figure 3.9– Pre-processed SPIRou spectra of the Fabry-Pérot etalon (left) and zoom to the central part of the image (right). wavelength solution, from which we obtain first-guess FP wavelengths; these first-guess wavelengths are in turn used to fit the cavity width. The second, method FP2, is based on the one developed by C. Lovis for ESPRESSO (private communication). Fractional FP line numbers are assigned to the HC lines, which are then used to fit the cavity width directly. The overall structure of the algorithm is as follows. Data identification and reading: The input files are verified via FITS header keys to be E2DS files corresponding to a HC lamp and the FP etalon respectively. The data and header are read, and the lamp and fibre are identified. Fibre correspondence between the HC and FP files is checked. If more than one HC file is given, they are added together. Currently, providing more than one FP file is not supported. Calibration set-up: Previous calibration files are copied from the calibration database (including the blaze file and the previous wavelength solution). The previous wavelength solution is read and checked for compatibility with the current parameter set- up (order of the polynomial fits employed). The correct HC line catalogue is read in. Generation of a first-guess wavelength solution: The HC spectra is used to generate a wavelength solution, using Method HC2 from Sect. 3.4.1. Quality controls are applied to the first-guess solution. In version 0.5.000 they only print a warning if failed; in the next version, if failed, the FP incorporation will not be entered, and the code will exit with a warning to the user. Incorporation of the FP lines: The FP lines are identified and gaussians fitted to them: for each order, the highest value is identified, and a gaussian fit attempted on a 7-pixel box around it. Fits that do not fail and are centred within ±1 pixel of the highest value are stored, the rest are rejected. Finally the gaussian fit is subtracted (or the region set to zero if the fit fails), and the process iterates until no more lines are found. With the 3.4. Wavelength calibration development 109

FP lines’ pixel positions known, the initial wavelength solution generated from the HC spectra alone is used to fit first-guess FP wavelengths from the FP line pixel positions. Using the FP equation and an input cavity width value, the FP line number is obtained for the last peak of the reddest order. The rest of the lines are then numbered by counting along each order, using wavelength matching across orders (with gaps due to missed peaks accounted for). Wavelength dependence of the cavity width: With the FP line numbers identified, the wavelength dependence of the cavity width is dealt with using one of the two methods named above: • Method FP1: Using the line numbers and the first-guess wavelengths of the FP lines (derived from the first-guess HC solution previously generated), a cavity width is calculated for each peak. A ninth-order polynomial is fitted to the cavity width (Fig. 3.10a), and corrected FP line wavelengths are calculated from this polynomial fit. • Method FP2: First, a fourth-order polynomial is fitted to the FP line numbers as a function of pixel positions per order. This fit is used to generate fractional line numbers for the "best" HC lines (selected as the lines at blaze values of more than 30%, and with velocity offset from the catalogue less than 0.25 km s−1). Using these fractional line numbers and the catalogue wavelengths, the cavity width is calculated for each HC line using the FP equation. Ninth-order polynomials are then fitted to these cavity width values (Fig. 3.10b) as a function of both line number and wavelength, and corrected FP line wavelengths are calculated from the fit. Optionally, a previous cavity width fit can be read in; in this case we assume only an achromatic change (i.e. a shift) may have taken place, and correct the fit for this shift by subtracting the median of the residuals between the newly calculated cavity widths for the HC lines and the previous fit. Fitting the solution: The HC and FP lines are combined. A fourth-order polynomial fit is performed between the pixel positions given by the centres of the gaussians, and the line wavelengths (catalogued for the HC lines, generated from the cavity width fit for the FP lines). Calculating the FP RV: The radial velocity of the FP spectrum is calculated using the CCF method, through cross-correlation with a FP mask. It is stored in order to give a FP RV zero-point, from which the drift of the spectrograph for a later observation of a star with simultaneous FP can be measured (by subtracting the RV of the wavelength solution’s FP from the RV of the simultaneous FP). Littrow solution check: Similar to the HC solution. Quality controls: Similar to the HC solution. Logging statistics: Similar to the HC solution. Saving the solution: Similar to the HC solution. If the quality controls were all passed, the header of the input FP E2DS spectra is also updated. The HC and HC-FP solutions are stored under different file names, so that the HC-FP solution does not overwrite the HC solution. 110 Chapter 3. Development of the SPIRou nIR spectropolarimeter

Saving results tables: Two tables are stored. The first logs the statistics of the Littrow solution quality check. The second is a list of all lines used for the solution, containing the order, wavelength, difference in velocity of the final fit from the input line value, weight, and pixel position for each line.

(a) FP cavity width versus FP line number, (b) FP Cavity width versus line number, as calculated for all the FP lines (Method as calculated for the HC lines using frac- FP1). tional line numbers (Method FP2).

Figure 3.10– Variability of the FP cavity width for method FP1 (left) and method FP2 (right). The results from the two methods are very similar. For method FP1 some outliers (poorly ﬁtted FP lines) can be seen, while for method FP2 the number of lines drops oﬀ towards low line numbers (i.e. high wavelengths) and no lines are selected for the reddest order.

As will be discussed in Sect. 3.5, the two methods are comparable, though method FP2 has slightly better accuracy and stability. In version 0.5.000 of the DRS, only method FP1 is available for general use, as method FP2 was still in development. In forthcoming versions both will be offered as options in the ﬁnal wavelength solution algorithm.

3.5 Validation and performances of the wavelength solution

To test the performance of the different wavelength solution generation methods, I ran all scripts on the calibrations of a two-week SPIRou run in February 2019. I used three different wavelength catalogues: the R11 catalogue, a combination of the R11 and S18 catalogues (with the S18 wavelengths kept for matching lines), and a selection of the most stable lines (i.e., those consistently identiﬁed for different HC frames) with updated, more accurate wavelength values. This selection of lines was derived using method FP2. First, the cavity width was ﬁtted from the best HC lines (at blaze values of more than 30%, and with velocity offset from the catalogue less than 0.25 km s−1) for each of 50 HC exposures taken during commissioning (between May and November 3.5. Validation and performances 111

2018). For each of these exposures, the wavelengths, fractional line numbers and calculated cavity widths of the lines used for the fit were saved. Then, all the lines were combined, and their wavelengths and cavity widths were fitted together to generate a very accurate cavity width fit (Fig 3.11a). Using this accurate cavity width fit, it can be seen that while for each catalogue line the different cavity widths measured for each exposure cluster together, these clusters can be significantly offset from the overall fit (Fig 3.11b). This would mean that for these offset values, the catalogue wavelengths are inaccurate. Therefore, each measured line’s wavelength was recalculated from its fractional line number and the FP equation. Finally, each line that was selected in at least two exposures was assigned an updated wavelength, as the median of its recalculated wavelengths. The stable lines catalogue is therefore not just a selection of the best lines from the others, but a new catalogue with updated wavelength values for each lines.

(a) Overall cavity width fit from the HC (b) Zoom of the cavity width fit, showing lines for fifty exposures. lines with offset values.

Figure 3.11– Construction of a stable lines catalogue. Left: Overall cavity width fit (black line) from the HC lines (dots) for fifty exposures. Right: Zoom showing how the values for each line cluster together, but can be offset from the main fit.

3.5.1 Impact of previous calibrations

As shown in Fig. 3.4, the wavelength solution is the last step of a long calibration sequence. This made its development particularly challenging, as changes upstream frequently meant large modifications in the data that proved destabilising for the wavelength solution, especially in the earliest versions of the DRS. As the DRS has evolved, this has somewhat ameliorated. Nevertheless, it is interesting to investigate whether the wavelength solution shows any important dependence on other calibrations. In particular, it was suspected that the precise determination of the slit profile (generated by cal_SHAPE_spirou.py) could impact the wavelength solution. Therefore, I generated wavelength solutions for two re-extractions of the HC and FP files used for these tests: one using a single set of calibrations, one fixing all calibrations save the slit profile file (SHAPE file hereafter), which is calculated for each night. Figure 112 Chapter 3. Development of the SPIRou nIR spectropolarimeter

3.12 shows the RV offsets between consecutive wavelength solutions for both reductions, for method FP2. Fixing all calibrations save the SHAPE file, most pairs of solutions show similar structures in their RV differences. In particular, there are clear effects at the ends of the orders. This is consistent with the fact that towards the edges of the detector, the flux is lower as the blaze function falls off, and the orders are more strongly curved (Fig. 3.13); so we can expect the determination of the slit profile to be less precise. A constant SHAPE file introduces large jumps between nights, of the order of ∼10–50 m s−1. However, removing these drifts (by subtracting the median RV difference), the pairs of solutions are much more consistent than those obtained with varying SHAPE files, and the edge effects diminish. Since we do not expect the spectrograph slit profile to actually change significantly from one night to the next, this also suggests that there are small instabilities in its determination that still need addressing. The next version of the DRS is expected to improve on this, which should in turn improve the stability of the wavelength solution. In order to explore the performance of the wavelength solutions alone, therefore, the tests presented in the rest of this section are all carried out with the HC and FP files extracted using a single set of calibrations, including a single SHAPE file. 3.5. Validation and performances 113

(a) Night-to-night RV differences between (b) Night-to-night RV differences be- wavelength solutions with a changing tween wavelength solutions with a fixed SHAPE file. SHAPE file.

Figure 3.12– Impact of changing the SHAPE file on the wavelength solution. Top row: night-to-night differences in RV, for wavelength solutions calculated with a changing (left, 3.12a) or fixed (right, 3.12b) SHAPE file. Bottom row: night-to-night differences in RV for wavelengths solutions with a fixed SHAPE file, removing the median RV. 114 Chapter 3. Development of the SPIRou nIR spectropolarimeter

Figure 3.13– FP lines, used to build the SHAPE ﬁle, at the centre (left) and edge (right) of the detector. The lines at the centre are much brighter, and the spectral order is practically straight; at the edges, the lines are fainter and the order curvature is pronounced. The four slices of the pupil slicer can also be clearly distinguished.

3.5.2 Performances of the HC wavelength solution

To test the performance of the HC wavelength solutions, I ran both methods on the UNe spectra taken as part of the afternoon calibrations for the two-week SPIRou run in February 2019, for all three wavelength catalogues. The spectra were reduced with a single set of calibrations. Table 3.1 summarises the results. The "internal RMS" row is a median of the internal accuracies reported for each solution. The "night-to- night variation" represents the median difference between consecutive nights’ solutions (computed as the median of the absolute drift-corrected RV differences, with the drift corrected by subtracting the overall median, which removes the large offsets created by the use of a single SHAPE file). The "lines used" row is the median of the number of HC lines that were identified and used to fit each wavelength solution.

Table 3.1– Summary of HC wavelength solution performances. Method HC1 Method HC2 Internal RMS 1.88 m s−1 3.87 m s−1 R11 catalogue Night-to night variation 16.4 m s−1 6.3 m s−1 Lines used 5607 4770 Internal RMS 1.88 m s−1 3.95 m s−1 R11+S18 catalogue Night-to night variation 15.1 m s−1 7.6 m s−1 Lines used 6186 5310 Internal RMS 1.46 m s−1 2.21 m s−1 Stable lines Night-to night variation 13.9 m s−1 5.7 m s−1 Lines used 2363 2124

Method HC1 has slightly better internal accuracy than method HC2, but is less stable from one night to the next. As noted in Sect. 3.4.1, its sensitivity to the input wavelength solution was particularly highlighted by the earthquakes suffered by the 3.5. Validation and performances 115

CFHT during SPIRou validation, on 3rd and 4th May 2018. The multi-pixel displace- ment meant all lines were shifted completely out of the search windows defined from the previous (pre-earthquake) solutions. This required me to create new algorithms to identify the pixel shifts and generate shifted first-guess solutions, in order for method HC1 to be able to run. Concern over this sensitivity was in fact one of the driving motivations for the development of method HC2, which (since it identifies all peaks in the spectrum and then generates a best match to the catalogue) is more robust to such shifts. Regarding the catalogues, adding the lines from S18 does not seem to create a substantial change in accuracy or stability. Most of the lines added have fairly low relative intensities reported by S18, so they are likely small and their gaussian fits may be less precise. The "stable lines" catalogue provides somewhat more accurate and stable wavelength solutions for both methods. In any case, neither method reaches the required internal precision of 0.45 m s−1 with any catalogue.

3.5.3 Combined HC-FP wavelength solution

I used the same two-week SPIRou run in February 2019, processed with a single set of calibrations, to test the performances of both the combined HC-FP wavelength solution methods. To generate the ﬁrst-guess HC solution, I applied method HC2 in both cases. Once again, I tested the three different wavelength catalogues. Table 3.2 summarises the results, with the same rows as for Table 3.1.

Table 3.2– Summary of HC-FP wavelength solution performances. Method FP1 Method FP2 Internal RMS 0.18 m s−1 0.12 m s−1 R11 catalogue Night-to night variation 3.1 m s−1 1.5 m s−1 Lines used 23612 21662 Internal RMS 0.18 m s−1 0.12 m s−1 R11+S18 catalogue Night-to night variation 3.0 m s−1 1.9 m s−1 Lines used 23691 21874 Internal RMS 0.18 m s−1 0.13 m s−1 Stable lines Night-to night variation 4.3 m s−1 3.3 m s−1 Lines used 22990 20466

In this case, the two methods are very comparable, though method FP2 has somewhat better internal accuracy and stability. Rather perplexingly, for the combined HC-FP solutions the "stable lines" catalogue provides the least night-to-night stability! This is particularly evident for method FP2, where the HC lines are used to fit the cavity width directly. Since this catalogue was generated from a cavity width fit using multiple HC exposures, each reduced with the corresponding nightly calibrations, this may perhaps be a derived effect of the previous calibrations’ instability. Several important changes are planned for the next version of the DRS (Sect. 3.5.5); once it is in place, I will generate a new "stable lines" catalogue and test it against the current one. Nevertheless, 116 Chapter 3. Development of the SPIRou nIR spectropolarimeter in all cases the internal accuracy is excellent, and the night-to-night stability is much improved compared to the HC solutions. As was described in Sect. 3.4.2, for method FP2 there is an option to read in a previous cavity width fit and correct it from any achromatic shift, instead of generating it anew. The reasoning behind this is that the chromatic dependence is an intrinsic prop- erty of the soft coating; while it may evolve slowly over the lifetime of the instrument, it is not expected to change from one night to the next. An achromatic shift, meanwhile, would correspond to a change in the physical separation of the FP, which could be caused by pressure changes. I tested the implementation of this option, redoing the analysis with an initial cavity width read in. I found a median internal RMS of 0.14 m s−1, and a median night-to-night variation of 0.8 m s−1, regardless of the catalogue used. This implies that a substantial part of the night-to-night variability for the combined HC-FP solutions is in fact coming from the cavity width fit. An example of a night-to-night comparison is shown in Fig. 3.14, for the solutions computed with method FP2 using a fixed cavity width, for the nights of 21st and 22nd February 2019, respectively. It is clear that the night-to-night variations are significantly reduced compared to e.g. Fig. 3.12a, and that the redder orders are driving most of the remaining variability.

Figure 3.14– Night-to-night variations of the FP wavelength solution: difference (in RV space) between the solutions generated with method FP2 for the 21st and 22nd February 2019. The constant shift induced by the fixed SHAPE file has been subtracted. 3.5. Validation and performances 117

3.5.4 Impact on RV error

Ultimately, the interest of an accurate and stable wavelength solution for a spectrograph lies in being able to measure precise radial velocities. In the SPIRou DRS, radial velocities are measured by the CCF method - that is, by cross-correlating a binary stellar mask with the observed spectrum. This cross-correlation is first performed order by order; these are then summed together, and a gaussian fit is performed, the final RV being measured as the centre of the gaussian. The RVs per order and the combined RV are all stored. If the star is observed with simultaneous FP on the calibration fibre, the instrumental drift is computed by the same recipe, cross-correlating the FP spectrum to a binary FP mask. To evaluate the impact of the wavelength solution on the radial velocities, I selected two stellar observations, one of Gl411 and one of Gl338B. I chose these as the brightest targets observed on the 17th of February 2019, which is the closest night to the centre of the run for which stars were observed (due to significant bad weather throughout the run). For each star, I computed the radial velocities changing the input wavelength solution. To generate the wavelength solutions, I adopted method FP2 and a fixed cavity width fit, as this was shown to provide the most accurate and stable set of wavelength solutions. I used the CCF computation from the SPIRou DRS to obtain the RVs. Both stars were observed with simultaneous FP calibration, so the drift was also computed from the FP CCFs. The results are summarised in Table 3.3. The standard deviation of the differences from the RV obtained for the 13th of February (taken as the reference point) is 0.67 m s−1 for Gl411, and 0.40 m s−1 for Gl338B. The large drifts are due to the fixing of a single SHAPE file, which induces offsets between the wavelength solutions, as described in Sect. 3.5.1. Although the overall variations are small, it is worthwhile to inspect the CCFs in more detail. Fig. 3.15 shows the difference in CCF RVs per order for each star, without and with drift correction. Gaps correspond to orders for which no CCF could be calculated (generally due to a very low or null atmospheric transmission). It is clear that some orders are far more variable than others, and may be driving the RV differences. In particular, for Gl411 the RVs for orders 9 and 1 have standard deviations of ∼5 m s−1, while the rest are below ∼3 m s−1. Likewise, for Gl338B, the RVs for orders 22 and 9 have standard deviations of ∼12 m s−1 and ∼8 m s−1 respectively, while the rest are below ∼3 m s−1. Recalculating the RVs excluding these two orders, the differences are generally slightly reduced (last column of Table 3.3), with a standard deviation of 0.50 m s−1 for Gl411 and 0.39 m s−1 for Gl338B, though the impact is not large. As to the cause of these variations, order 22 is adjacent to a low-transmission region, and may have suffered from atmospheric effects. Ongoing analyses of the CCF RVs by other members of the SPIRou team have found that orders 0-1 and 9-12 are affected by telluric contamination; future versions of the DRS will be adapted to better handle these regions. 118 Chapter 3. Development of the SPIRou nIR spectropolarimeter

Table 3.3– Summary of CCF RV differences using different wavelength solutions. Star Night RV [km s−1] drift [m s−1] RV diff (all RV diff (sel. orders) orders) [m s−1] [m s−1] 13 Feb -75.92096 82.42 — — 14 Feb -75.96668 35.81 0.89 0.89 15 Feb -76.01213 -8.95 0.20 0.20 16 Feb -76.01567 -12.51 0.22 0.22 17 Feb -76.01211 -9.41 0.68 0.68 18 Feb -76.01640 -13.27 0.26 0.25 Gl411 19 Feb -76.01066 -7.63 0.36 0.35 21 Feb -75.99916 4.62 -0.39 -0.40 22 Feb -75.99447 8.84 0.07 0.07 23 Feb -75.99277 10.52 0.09 0.09 24 Feb -75.97533 28.60 -0.54 -0.55 25 Feb -75.95954 44.62 1.93 -0.78 13 Feb 21.01181 82.21 — — 14 Feb 20.96621 35.60 1.01 1.01 15 Feb 20.92063 -9.17 0.20 0.23 16 Feb 20.91696 -12.71 0.07 0.10 17 Feb 20.91939 -9.60 -0.61 -0.58 18 Feb 20.91627 -13.47 0.14 0.17 Gl338B 19 Feb 20.92215 -7.84 0.39 0.42 21 Feb 20.93418 4.40 0.18 0.19 22 Feb 20.93831 8.63 0.08 0.10 23 Feb 20.94013 10.31 0.22 0.23 24 Feb 20.95789 28.38 -0.09 -0.08 25 Feb 20.97379 44.39 -0.20 -0.20 3.5. Validation and performances 119

Figure 3.15– CCF RV differences per order for Gl411(top row) and Gl338B (bottom row), using wavelength solutions from different nights. Left: absolute RV differences; the offset generated by fixing the single SHAPE file for the wavelength solutions is evident. Right: drift-corrected RV differences. The drifts computed from the FP are effective to shift all the CCFs into a common frame.

3.5.5 Upcoming changes to the Data Reduction System

The previous sections are based primarily on the current stable version of the DRS, 0.5.000. Signiﬁcant changes have been planned for the next version, which the DRS team is currently working on. Several of these changes are either directly on the wavelength solution, or are expected to impact it:

• A new unified wavelength solution code, which will handle inputs of both HC alone and HC+FP. If only HC files are given, it will create a wavelength solution using method HC2. If FP files are also included, it will use this wavelength solution (provided it passes quality controls) as a first guess for either method FP1 or FP2. The choice between the FP methods will be set through configuration parameters. Cross-order continuity of the solutions will also be tested, though it may not be implemented directly in the next version. • The implementation of a set of "master" calibrations that are not expected to change on a nightly basis, but only per run or even per thermal cycle. I highlight 120 Chapter 3. Development of the SPIRou nIR spectropolarimeter

here the SHAPE file which traces the slit profile. A master file will be generated by combining a great quantity of FP files, and establishing the slit profile from this super-FP. Each night will then be mapped onto this master SHAPE by comparing with the FPs of the corresponding calibration sequence. This should improve the stability of extraction, and therefore of the wavelength solution. • An improved telluric line correction, and an enhanced CCF computation taking into account the atmospheric transmission. Both these changes will improve the CCF RV precision and stability.

3.6 SPIRou science programs

3.6.1 Spirou Legacy Survey

The SPIRou Legacy Survey (SLS) d consists of two main science topics: 1. Detection and characterisation of exoplanets around low mass stars, with two components: a) (SLS-PS) Systematic RV monitoring of nearby M dwarfs; b) (SLS-TF) RV follow-up of transiting planet candidates. 2. (SLS-MP) Study of the impact of magnetic ﬁelds on star and planet formation, through spectropolarimetry of protostars. The SLS was allocated 300 nights over 4 years (semesters 2019A to 2022B) by CFHT. Observations started in February 2019, and are performed in queue mode. For all three SLS components, the exposures are taken in polarimetric mode, producing Stokes I (unpolarised) and Stokes V (circular polarisation) spectra. For the SLS-PS, the detection and conﬁrmation of habitable Earths and superEarths is of particular interest. Small planets in the habitable zone are much easier to detect around M dwarfs compared to sun-like stars, as described in Sect. 1.3.2. A large RV survey will enable us not only to detect such planets, but also to constrain their statistics. Additionally, searching for planets around nearby M dwarfs in particular means that we will be able to identify the best targets for future atmospheric characterisation (e.g. with the James Webb Space Telescope or the upcoming ELTs). The SLS-PS will observe 100 M dwarfs selected from the SPIRou input catalogue of closest M dwarfs, using merit functions based on low stellar activity and expected RV uncertainty (Moutou et al., 2017; Fouqué et al., 2018). Each stars will be observed on average once per night while it is visible from Mauna Kea. The H magnitudes and distances are shown in Fig. 3.16. The expected yield of the SLS-PS was estimated in Cloutier et al.(2018), using a simulated target list and an assumed time allocation of 300 nights over 3 years, to +29.3 +16.8 +7.6 be of 85.3−12.4 planets, among which are 20.0−7.2 habitable-zone planets and 8.1−3.2

d. Main website at https://spirou-legacy.irap.omp.eu/doku.php, full description at http:// spirou.irap.omp.eu/Observations/The-SPIRou-Legacy-Survey 3.6. SPIRou science programs 121

(a) Histogram of the H magnitudes of the (b) Histogram of the distances of the SLS- SLS-PS targets. PS targets.

Figure 3.16– Histograms of the H magnitudes (left) and distances (right) of the 100 SLS-PS targets, as reported in Simbad. The targets are all bright in the nIR (median H magnitude 6.967) and close (median distance 7.88 pc).

Earth-like planets. They also studied the effect of several modifications to the simula- tion; of particular relevance are the short and degraded versions. The first assumes only half the measurements will be performed (so 150 nights over 3 years, compared to the 225 nights expected over the first three years for the actual allocated time). In this +27.8 +13.7 +4.9 case, the yield is of 65.7−11.1 planets, with 16.4−5.9 habitable-zone planets and 5.2−2.1 Earth-like planets. The second studies the impact of a degraded RV precision of 2 m s−1 −1 +27.1 +13.3 instead of 1 m s ; the expected yield becomes 65.2−11.0 planets, including 15.9−5.7 +4.8 habitable-zone planets and 5.0−2.0 Earth-like planets. In either case, SPIRou is still expected to provide dozens of new discoveries, including habitable-zone and Earth-like planets. The follow-up of transiting planet candidates is another important part of SPIRou exoplanet science, and is the focus of the SLS-TF component. On the one hand, simultaneous mass and radius measurements for a planet allow the determination of its density and thus the modelling of its probable internal structure. On the other, transiting planets are prime candidates for atmospheric characterisation, for example with JWST. With several current and upcoming photometric surveys focusing on M dwarfs, nIR radial- velocity follow-up will be crucial to validate the planets and determine their masses. The SLS-TF will observe 65 stars, and will mostly concentrate on the follow-up of interesting candidates from TESS (Ricker, 2016) and ExTra (Bonfils et al., 2015). Each target for which a planet is confirmed will be monitored until a 5σ mass detection has been secured. The second science topic is somewhat outside the scope of this thesis. Briefly, the SLS-MP will monitor 120 class-I, class-II, and class-III protostars in the three closest star- forming regions with spectropolarimetry. It will study the magnetic topologies of these objects, aiming to understand how they depend on mass, age and rotation. It could also potentially detect hot Jupiters around these objects, providing an observational test to current formation and migration theories. 122 Chapter 3. Development of the SPIRou nIR spectropolarimeter

To organise the work on the obtained data, the SLS has defined five Work Packages (WPs). The first three are each dedicated to a science topic, while the last two are transversal. They are structured as follows: • WP1: work on the SLS-PS science component (coordinators: X. Delfosse, E. Arti- gau, C. Moutou). • WP2: work on the SLS-TF science component (coordinators: R. Doyon, G. Hébrard). • WP3: work on the SLS-MP science component (coordinators: S. Alencar, J. Bouvier, J.-F. Donati). • WP4: optimizing the results of WP1-3, through RV optimization, jitter filtering, and studies of system dynamics, star-planet interactions, and habitable zones (coordinator: I. Boisse). • WP5: complementary science, through spectral analysis, mapping of large-scale fields of low-mass stars, and study of Earth’s atmosphere (coordinator: J. Morin).

3.6.2 Synergy with SOPHIE

While the SLS-PS and the SOPHIE SP3 programmes do not have the same selection criteria, with the SLS-PS taking the least active, closest M dwarfs and the SP3 a mostly volume- and magnitude-limited sample, there is a natural overlap between these sets of stars. Therefore, the SLS-PS target list includes 47 of the SOPHIE SP3 targets, among which is Gl411, for which the SP3 has published a superEarth at a 13 d period (Díaz et al., 2019). There is a strong potential for synergy between SPIRou and SO- PHIE: • While planetary signals must be present at all wavelengths, stellar activity signals should be wavelength-dependent, since the flux ratio between starspots and the stellar photosphere decreases with wavelength, meaning the RV signal is generally expected to be smaller in the nIR (Reiners et al., 2010; Barnes et al., 2011; Marchwinski et al., 2015). However, this dependence can be complex (see e.g. Barnes et al. 2011, who find the shape and equivalent width of the spot profile to be important for high contrast ratios; Reiners et al. 2013, who describe how line broadening due to to the Zeeman effect creates an RV signal that increases at longer wavelength). The combination of nIR and visible data will therefore allow the verification of suspected stellar signals. It will also enable the testing and identification of robust activity indicators in the SPIRou nIR range. The currently most-used nIR activity indicator is the CaII triplet (CaII IRT, at 849.8 nm, 854.2 nm 0 and 866.2 nm), which correlates well with the log(RHK) index (e.g. Andretta et al. 2005, Martin et al. 2017), but may be less sensitive to the rotation period (Mittag et al., 2017). However, it falls outside the SPIRou wavelength range, as do other indicators explored by Schöfer et al.(2019). Therefore, new indicators will have to be found for SPIRou; comparison with well-tested visible activity indices from SOPHIE should help to identify them. • The SOPHIE precision on M dwarfs is primarily limited by photon noise. Thanks to its wavelength range and the larger telescope on which it is mounted, SPIRou can achieve much higher S/N ratios for these stars, enabling the detection of additional 3.6. SPIRou science programs 123

low-mass planets. To illustrate the importance of this effect, on SOPHIE 30-minute exposures are required to reach average S/N levels of 85 at 650 nm. For these same targets, the SPIRou exposure time calculator estimates a S/N level of 200 can be reached in less than four minutes on average. While the RV content in the nIR is lower than that in the visible (e.g. Artigau et al. 2018, who also note that current models underpredict the RV content in the K band), this is at least partially offset by the ability to reach much higher S/N levels. • SOPHIE has a much more ﬂexible on-sky schedule compared to SPIRou, which competes for time with multiple other instruments at CFHT, and is not mounted in dark time. In semester 2019A, it was mounted on the telescope in 10-15 day runs with gaps of at least 15-20 days. Coordinated SOPHIE observations should therefore allow us to cover gaps in the SPIRou coverage. This will be particularly important for planets whose orbital periods are longer than the typical SPIRou run. • There is a signiﬁcant overlap between the SOPHIE-SP3 and SPIRou-SLS teams, which should enable an optimal exploitation of the joint data.

The SP3 sub-programme is close to nominal completion (that is, to having at least 30 measurements for all targets in its volume-deﬁned sample of M dwarfs); for the near future, the aim is to continue only a handful of most interesting targets, and to reorient the programme towards a synergy with SPIRou. This reorientation will begin in the upcoming semester, through the incorporation of selected SPIRou targets to the SP3 sample.

3.6.2.1 Open-time proposal

While several of the SOPHIE SP3 targets are part of the SLS, there are many interesting targets from the SP3 that are not included in the SLS target list. I submitted an open-time proposal to observe three of these, in order to study: • The potential of SPIRou to detect additional planets that are obscured by photon noise in the visible, through the observation of a SOPHIE target, Gl 378, with a known short-period planet and high O-C residuals compared to the SOPHIE error bar. These additional planets are statistically likely to exist, as most M dwarf planetary systems are expected to be multiple (e.g. Dressing & Charbonneau 2015). • The effect of the SPIRou sampling (conditioned by when SPIRou is on sky) on the detectability of a long-period planet, Gl 96 b, seen in the SOPHIE spectra. • The impact of stellar activity in the near infrared, through the study of a target, Gl 270, with clear and important quasi-periodic activity signals in the visible.

The programme has been accepted for semester 2019B, in priority B, for the first and third targets. For Gl 378, I have requested 30 observations, over consecutive nights whenever possible. This star has a warm Neptune, published in Hobson et al.(2019), and the high residuals of the fit to the SOPHIE data suggest additional effects (more planets, stellar activity, or instrumental effects). As well as searching for additional planets, we aim to cover the known planet’s 3.82 d period several times, and to refine its orbital parameters by analysis of the combined SOPHIE and SPIRou data. Since the 124 Chapter 3. Development of the SPIRou nIR spectropolarimeter planet has a non-negligible transit probability, refined orbital parameters will be valuable for transit searches. Gl 270 shows strong RV signals at 15.5 d and 31.5 d, and strong signals at 31.5 d in 0 Hα and log(RHK), as discussed in Sect. 2.5.3. This makes it an excellent target to study quasi-periodic stellar activity. For this target, I have requested 35 observations, to be spread over 3 runs in order to sample the quasi-period from the visible RVs and activity indices well. This will allow us to verify whether the same periodicity is seen in the nIR RVs, and - if it is - how its strength compares to the signal strength in the visible. Additionally, we will also be able to search for nIR lines/bands in the SPIRou domain that are sensitive to stellar activity. 4 Properties of M-dwarf planets as of 2019

Contents

4.1 Overview and growth of the M dwarf planet population sample ...... 126 4.2 Stellar hosts...... 127 4.3 Mass-period diagrams ...... 129 4.4 Types of planets ...... 136 4.5 Habitable zone planets...... 140 4.6 A mass-metallicity correlation? ...... 142 4.7 Multiplanetary systems orbiting M dwarfs...... 143 126 Chapter 4. Properties of M-dwarf planets as of 2019

4.1 Overview and growth of the M dwarf planet population sample

The number of exoplanets known orbiting M dwarfs greatly increased during the development of this thesis, from 118 planets detected up to 2016, to 196 known today - a 66% increase over three years. The full choice of catalogue and verification of the sample is described in Sect. 1.4; I briefly reiterate the main selection criteria that were applied to the Exoplanet Encyclopaedia’s catalogue (originally downloaded 19th April 2019; for this chapter, the catalogue was re-downloaded in August 2019, and new M-dwarf planets fitting the criteria below were manually incorporated): • For the host star:

Stellar mass in the range 0.06 M ≤ M∗ ≤ 0.6 M (Cox 2000, Kaltenegger & Traub 2009). Spectral type on Simbad or Vizier corresponds to a main-sequence M dwarf (to remove evolved stars, brown dwarfs at the lower mass limit, and K stars at the upper mass limit). • For the planet:

Planetary mass Mp < 13 MJ.

Figure 4.1a shows the cumulative distribution of these M dwarf planets from the year 1998 (the ﬁrst planet around an M dwarf, GJ 876 b, independently detected by Delfosse et al.(1998) and Marcy et al.(1998)) to 2019. It is clear that 2017 and 2018, in particular, were extremely productive years, with 31 and 28 detected planets respectively; with 21 planets already detected halfway through the year, 2019 seems likely to match or exceed them. With regards to detection methods, 83 were detected by radial velocities, 51 by the transit technique, 52 by microlensing, eight by imaging, one by transit timing variations, and one by the pulsar method. Compared to the same statistics for 2016 reported in Sect. 1.4, it is clear that the increase in planets of the past three years has been driven by the radial velocity, transit, and microlensing methods, as shown in Fig. 4.1b. 4.2. Stellar hosts 127

(a) Cumulative distribution of M dwarf (b) Productivity of detection methods for planets by year of discovery. M dwarf planets as of 2016 (blue) and 2019 (orange).

Figure 4.1– Evolution of the M dwarf planet population, as of 19th April 2019: detections over time (cumulative distribution, left), and productivity of diﬀerent methods (bar plots, right).

The aim of this chapter is to provide an overview of the main properties of the planets orbiting M dwarfs and of the stars that host them. In it, I will discuss the mass- period diagram, the mass-metallicity relation, the types of planets detected and their potential compositions, and the multiplicity of the systems.

4.2 Stellar hosts

While most of this chapter will deal primarily with the properties of the exoplanets orbiting M dwarfs, it is also relevant to brieﬂy detail the main properties of their host stars. Their spectral types are shown in Fig. 4.2; most of the known planets orbit early-to-mid M dwarfs. This is consistent with the target selection strategies adopted by RV surveys such as those detailed in Sect. 1.3.2: even when the sample does not speciﬁcally exclude late M dwarfs, they will mostly be removed by magnitude cuts as they are intrinsically very faint in the visible. Nevertheless, some planets around late M stars are known, and we may expect their numbers to increase with the rise of nIR surveys. 128 Chapter 4. Properties of M-dwarf planets as of 2019

Figure 4.2– Histogram of spectral types of M-dwarf stars known to host planets.

As noted in Sect. 1.4, planets around bright, nearby M dwarfs are expected to be the best candidates for near-future in-depth characterisation. Fig. 4.3 shows the distances and V magnitudes of M dwarfs currently known to host planets. Stars hosting planets in the habitable zone are highlighted. While planets have been detected out to large distances (notably through the microlensing method), the majority orbit relatively nearby stars. Habitable zone planets, in particular, are all within 11 pc, and most are hosted by bright stars, making them excellent targets for atmospheric and direct imaging characterisation efforts. With new nIR instruments enabling the observation of fainter stars, and surveys focusing primarily on nearby stars, as discussed in Sect. 1.3.2, we can expect that most new systems will ﬁll in the left side of the diagram. 4.3. Mass-period diagrams 129

Figure 4.3– Distances and V magnitudes of M dwarfs with planets. Stars hosting habitable zone planets are highlighted in orange. The circle sizes are proportional to the stellar masses.

4.3 Mass-period diagrams

The mass-period diagram is extremely useful as a tool to study an overall planetary population, enabling us to distinguish features such as subgroups, gaps, deserts, which in turn provide information about the formation and evolution of the planetary systems considered (e.g. Mordasini et al., 2012; Beaugé & Nesvorný, 2013; Mazeh et al., 2016; Owen & Lai, 2018; Jin & Mordasini, 2018). Figure 4.4 shows the mass-period diagram for the exoplanets hosted by M dwarfs, colour-coded by eccentricity. The masses are separated into two categories, indicated by diamonds and circles respectively: empirically measured masses or mass lower limits (determined by radial velocities, transit timing variations, or microlensing) and estimated masses (upper limits from RV or TTV non-detections, mass-radius relations applied to transiting planets, or evolutionary models applied to directly imaged planets). 124 planets have measured masses, 46 only estimated masses. The overall structure of the diagram has not signiﬁcantly changed since 2016, with the vast majority of the new planets joining the cluster of low-mass, short-period planets 130 Chapter 4. Properties of M-dwarf planets as of 2019

in the 1 − 10 M⊕ and 1–200 d ranges. The new median mass of 6.63 M⊕ and median period of 11.4 d are slightly lower than the corresponding 2016 values (7.98 M⊕, 16.6 d), reflecting that the increase has primarily been in lower-mass, shorter-period planets. The median eccentricity has likewise decreased, with a new value of 0.04 (compared to 0.08 in 2016), despite the addition of some high-eccentricity planets such as Gl 96 b (e = 0.44, Hobson et al. 2018a) or Gliese 49 b (e = 0.393, Perger et al. 2019) to the sample. Comparing this mass-period diagram to an analogous diagram for planets hosted by G-type stars (Fig. 4.5), several differences are immediately evident. The first is the lower number of planets overall (there are 646 planets in the G stars diagram). As noted in Sect. 4.1, the sample of planets around M dwarfs is rapidly growing, and many upcoming surveys will target these stars specifically (Sect. 1.3.2), so this may even out in the future. Beyond that, there is an overall lack of Jupiter-sized planets around M dwarfs at all periods, compared to the large populations of hot and cool Jupiters orbiting G-type stars. The range of orbital periods covered is similar, which is logical as it will be dictated primarily by observational limits; however, there are more long-period planets around G stars. These long-period planets are generally massive (as small planets at long periods are not detectable with current instrumentation); their relative under-representation around M dwarfs is likely due to the general paucity of large planets orbiting these stars. As expected due to the inverse relation between RV amplitude and stellar mass (Sect. 1.4), lower-mass planets have been detected around M dwarfs than around G stars. The Hot Neptune desert remains essentially unpopulated for planets hosted by M dwarfs; of the planets potentially within it, only GJ 436 b has a well-characterised mass of 23.4 M⊕ from RVs (Southworth, 2010). The other two, K2-137b at 0.18 d with an upper mass limit of 158.9 M⊕ (Smith et al., 2018), and K2-22b at 0.38 d with an upper mass limit of 445.0 M⊕ (Sanchis-Ojeda et al., 2015), have only upper limits placed on their masses through RVs. Likewise, Hot Jupiters remain extremely rare, with only two −6 discovered since 2016 (HATS-17A b, mass 0.45 ± 0.24 MJ, period 3.7955213 ± 1.1 × 10 +0.066 d, Bakos et al. 2018, and NGTS-1b, mass 0.8120.075 MJ, period 2.647298 ± 0.000020 d). 4.3. Mass-period diagrams 131

Figure 4.4– mass-period diagram for all known exoplanets (as of 2019) hosted by M dwarfs, with these parameters reported in "The Exoplanet Encyclopaedia". Minimum mass M sin i values are used as mass when the inclination is not known. The points are colour-coded by eccentricity; planets without reported eccentricities are plotted in black. Diamond symbols represent measured masses or minimum masses (RV, TTV, microlensing); circles represent mass estimates (upper limits, mass-radius relations, evolutionary models), and are semitransparent. The dashed black lines show the limits of the Hot Neptune desert as deﬁned by Mazeh et al.(2016); they are dotted beyond 5d to reﬂect the authors’ warning that the desert is uncertain beyond that orbital period. 132 Chapter 4. Properties of M-dwarf planets as of 2019

Figure 4.5– Analogous to Fig. 4.4, but for exoplanets hosted by G-type stars (0.84 M ≤ M∗ ≤ 1.15 M ).

The dependence of the (radius-period) boundaries on different stellar parameters was recently analysed by Szabó & Kálmán(2019). They found a strong dependence on stellar effective temperature Teff , with more close-in Neptunes for cooler stars (Teff <5600 K). However, for these cooler stars, they also find an inverse dependence to this one on stellar mass: planets around more massive (warmer) stars can have shorter periods. They also confirm the tendency reported by Dong et al.(2018) and Petigura et al.(2018) for planets close to the lower boundary to be hosted by metal-rich stars. Meanwhile, Dong et al.(2018) also find that Kepler-detected hot Neptunes tend to be in single-planet systems. I analysed the M dwarf planets with measured masses (or minimum masses) determined by RVs and TTVs to see if these tendencies are also found for this sample (microlensing-detected planets are excluded since they are large and distant, and therefore not relevant to this analysis). Mass-period diagrams color-coded by Teff , M∗, [Fe/H], and multiplicity are shown in Fig. 4.6. 4.3. Mass-period diagrams 133

(a) (b)

Figure 4.6– Same as Fig. 4.4, but showing masses/minimum masses measured by RVs (diamonds) and TTVs (circles) only, and colour-coded by Teﬀ (4.6a), M∗ (4.6b), [Fe/H] (4.6c), and multiplicity (4.6d) respectively. The red dot-dashed line indicates the limit between planets that will be considered close to or far from the lower boundary of the hot Neptune desert, following a visible gap in the distribution.

To test whether the reported dependencies on planet parameters can be seen in this sample, I selected all planets with P<5 d beneath the lower boundary, and established a dividing line parallel to the boundary following a visible gap in the distribution (marked in red in Figs. 4.6). The two sets of planets thus defined seem to differ in metallicity, with median values of [Fe/H]median = 0.05 for the close planets and [Fe/H]median = −0.13 for the more distant planets. However, it must be noted that these metallicities are from different sources and may be inconsistent, especially given the difficulties of estimating metallicities for M dwarfs (as discussed in Sect. 1.2). The tendency for planets close to the boundary to be hosted by metal-rich stars seems therefore to hold for M dwarf planets in the mass-period plane. Regarding the stellar masses and temperatures, I found slightly more massive, warmer stellar hosts for planets close to the boundary, though the differences are not particularly significant as the dispersions are quite large (median values of Teff = 3324 ± 260 K, 134 Chapter 4. Properties of M-dwarf planets as of 2019

M∗ = 0.25 ± 0.15 M for planets close to the boundary, Teff = 3306 ± 356 K, M∗ = 0.18 ± 0.15 M for further planets). This is consistent with the tendency noted by Szabó & Kálmán(2019) for their sample of cooler stars, in which shorter-period planets orbited the more massive members of the sample. It appears that for M dwarfs, this tendency may dominate over the inverse tendency of the full stellar sample analysed by Sz- abó & Kálmán(2019) for hot Neptunes to be hosted by cooler stars. Like Dong et al.(2018), I find an overall tendency for warm Neptunes to be single; there are 7 single planets and 2 multiple-system planets among those close to the desert boundary (78% singles), and 14 multiple-system planets among those further from the boundary (with no singles). Additionally, the three planets with measured masses that fall within the desert are all singles. Dong et al.(2018) suggest this tendency could point to migration rather than in-situ origins for hot Neptunes (and hot Jupiters). It has recently been observed by Eigmüller et al.(2019) that the traditional mass- period diagram may not be the most adequate representation of a planet population, in that it disregards the potential impact of differing stellar masses. They suggest that a mass ratio-period diagram might instead be more appropriate. Their motivation for adopting this diagram is mainly due to inconsistencies on the upper edge of the Hot Neptune desert between the empirical boundaries of Mazeh et al.(2016), the high- eccentricity migration boundary determined by Owen & Lai(2018), and the actual up- to-date period-mass diagram for the full exoplanet population. This motivation may not be entirely relevant to M dwarf planets, as Hot Jupiters are also absent and the upper boundary of the desert is not well determined. Nevertheless, it is interesting to test this representation for the M dwarf planets, which is shown in Fig. 4.7. As was done in Eigmüller et al.(2019), the boundaries of the hot Neptune desert from Mazeh et al. (2016) are scaled assuming a solar mass for the host star (a reasonable assumption, since the sample used by Mazeh et al.(2016) was likely of primarily G-type hosts given its time frame; note also that the boundaries fit the mass-period diagram in Fig. 4.5 well). 4.3. Mass-period diagrams 135

Figure 4.7– mass ratio - period diagram for all known exoplanets (as of 2019) hosted by M dwarfs, with masses measured by RVs, TTVs, or microlensing. Minimum mass M sin i values are used as mass when the inclination is not known. The points are colour-coded by eccentricity; planets without reported eccentricities are plotted in black. The dashed black lines show the limits of the Hot Neptune desert as deﬁned by Mazeh et al.(2016), scaled assuming a solar-mass host; they are dotted beyond 5 d to reﬂect the authors’ warning that the desert is uncertain beyond that orbital period.

Interestingly, in this representation the clump of ﬁve hot Jupiters at 0.5 MJ mass and 3.5 d period are all above the upper boundary, instead of lying on it, while the lower boundary cuts through the low-mass planet population rather than tracing its edge. It should be noted here that in Eigmüller et al.(2019)’s own mass ratio - period diagram (their Fig. 8b), the same effect of multiple planets "moving up" into the desert can be seen, though it is not discussed by the authors. It appears that the mass ratio - period diagram as deﬁned by Eigmüller et al.(2019) may be a good tool for the analysis of the upper boundary of the desert, but does not adequately represent its lower boundary. Since their reasoning for adopting this representation was based on the high-eccentricity limits of Owen & Lai(2018) for the upper boundary, this would support different origins for the two boundaries - as proposed for example by Owen & Lai(2018), who suggest a photo-evaporation origin for the lower boundary. This is not necessarily contradictory with the migration hypothesis raised by Dong et al.(2018) for hot Neptunes, as they also propose - based on similarities in stellar metallicity, frequency, and tendency to 136 Chapter 4. Properties of M-dwarf planets as of 2019 be single - that these planets share common formation and migration processes with hot Jupiters, and that hot Neptunes could originate from photo-evaporated hot Jupiters (Baraffe et al., 2005). Therefore, warm Neptunes could have migrated inwards as hot Jupiters, scattering any other planets, then been evaporated down to the lower boundary of the desert. To summarise, the mass-period diagram of M-dwarf planets shows an abundance of low-mass, short-period, low eccentricity planets, and a dearth of hot Neptunes and hot Jupiters. The distribution follows the lower boundary of the hot Neptune desert as proposed by Mazeh et al.(2016). Accordance with the upper boundary is less clear as hot Jupiters are also mostly absent, but the few hot Jupiters known with measured masses lie along it. The stellar hosts of the planets close to the lower boundary show a slight tendency to be metal-rich, warmer, and more massive compared to the hosts of low-mass planets further from it. Most of the planets close to this boundary are in single-planet systems.

4.4 Types of planets

Using the mass ranges established by Stevens & Gaudi(2013) (Earths: Mp < 2 M⊕, superEarths: 2 M⊕ ≤ Mp < 10 M⊕, Neptunes: 10 M⊕ ≤ Mp < 100 M⊕, Jupiters: 100 M⊕ ≤ Mp < 13 MJ), the planet population can be divided into 63 Earth-mass planets, 35 superEarths, 39 Neptunes, and 19 Jupiters (taking the minimum mass M sin i as a mass proxy when the inclination is not known). Considering only those planets with masses determined by RVs, TTVs, or microlensing, the divisions become 49 Earth-mass planets, 34 superEarths, 35 Neptunes, and 8 Jupiters (Fig. 4.8). There is a clear abundance of low-mass planets and relative lack of giant planets; since more massive planets are easier to detect, this is unlikely to be a detection bias, but probably reﬂects the true distribution of planet masses around M dwarfs. 4.4. Types of planets 137

Figure 4.8– M dwarf planets categorised by mass (Earths: Mp < 2 M⊕, su- perEarths: 2 M⊕ ≤ Mp < 10 M⊕, Neptunes: 10 M⊕ ≤ Mp < 100 M⊕, Jupiters: 100 M⊕ ≤ Mp < 13 MJ). The paler columns in each colour correspond to the total number of planets with measured masses (through RVs, TTVs, or microlensing) in the category.

For a subset of 22 planets, both the mass and radius are known, allowing for a precise characterisation. Fig. 4.9 shows these planets, colour-coded by density. In this plot, RV-measured masses are true masses and not M sin i lower limits, because the planets transit and therefore the inclination is known. As a reference point, the densities of the Solar System planets are listed in Table 4.1. The highest-mass, highest- −3 radius planets at ∼ 10 R⊕, ∼ 100 M⊕ all have densities below 1 g cm , and are therefore presumably gas giants with massive atmospheres. Most of the low-mass, low-radius planets have densities in the 2–10 g cm−3 range, similar to the rocky Solar System planets. However, some are closer to 1 g cm−3, and may be more analogous to the Solar System ice giants. 138 Chapter 4. Properties of M-dwarf planets as of 2019

Table 4.1– Solar System planets’ densities Planet Density [g cm−3] Mercury 5.427 Venus 5.243 Earth 5.514 Mars 3.933 Jupiter 1.326 Saturn 0.687 Uranus 1.271 Neptune 1.638

Figure 4.9– Mass-radius plot for M dwarf planets, colour-coded by density. RV-measured masses are indicated by diamonds, TTV-measured masses by hexagons.

For a more precise idea of planetary composition than given by the densities alone, we can turn to internal structure models. Taking the planets with masses in the 0.1 − 20 M⊕ range, I placed them on a grid of composition models developed by Brugger et al.(2017) (Fig. 4.10). For the TRAPPIST-1 system (using the more accurate parameters derived by Grimm et al. 2018), most of the planets have similar positions close to the 100% mantle com- 4.4. Types of planets 139 position, with TRAPPIST-1 e in particular lying on the Earth-like composition line. More detailed modelling by Grimm et al.(2018) suggests that planets c and e (the two closest to the Earth-like model here) should have rocky interiors, while the rest require envelopes of volatiles. In contrast, the Kepler-138 planets (using the updated parameters from Almenara et al. 2018 have more varied compositions; planet b is between the 100% mantle and 50% mantle - 50% water models, planet c on the 100% mantle line, and planet d clearly above the 100% water line. This is reasonably consistent with the photodynamical modelling by Almenara et al.(2018), which results in a rocky composition for Kepler-138 c, and signiﬁcant envelopes for the other two, with higher volatile content for Kepler-138 d. The planets with RV-measured masses are fairly evenly distributed, with three (GJ 1132 b, GJ 357 b, and LHS 1140 c) between the 100% mantle and Earth-like compositions, one (LHS 1140 b) with a Mercury-like composition, one (K2-18 b) between the 50% mantle - 50% water and 100% water models, and two (GJ 1214 b and GJ 3470 b) above the 100% water line.

Figure 4.10– Composition models from Brugger et al.(2017) compared to M dwarf planets, colour-coded by mass-measurement method. The planets’ names are indicated. GJ 357 b and GJ 1132 b have practically identical masses and radii and appear superposed. TR-1 refers to the TRAPPIST-1 planets, abbreviated for ease of viewing. 140 Chapter 4. Properties of M-dwarf planets as of 2019

4.5 Habitable zone planets

As discussed in Sect. 1.3.2, M-dwarf stars are particularly interesting targets in the search for habitable zone planets. Since they are comparatively faint, planets orbiting them receive less stellar flux than they would at the same distance from a G star. Therefore, the liquid water habitable zone moves inward, closer to the star. Kop- parapu et al.(2013) developed a model for estimating the limits of the HZ around main-sequence stars, based on their effective temperature Teff and luminosity L∗. They define both a conservative HZ, with inner and outer limits given by water-loss and maximum greenhouse respectively; and an optimistic HZ, based on recent Venus (inner limit) and early Mars (outer limit) models. I computed these limits for all the M dwarf stars in the sample with catalogued Teff . As the L∗ does not appear in the catalogue, I estimated values from the mass-luminosity relationship for main-sequence stars, L∗ M∗ 3.5. I found 15 planets within the conservative HZ of their respective hosts, L = ( M ) and a further 7 within the optimistic HZ. Fig. 4.11 illustrates these planets within their systems. The liquid water habitable zone, as defined in Sect. 1.2, is the range of distance from the host star in which a planet could have liquid water on its surface. While the HZ limits provide this range of distances, it is also necessary for the planet to have a solid surface on which water can be found. This means a habitable planet must have a rocky composition, which implies a small size. While it has been speculated that giant planets could host potentially habitable exomoons (e.g. Kaltenegger, 2010), these are not detectable with current technology. The Habitable Exoplanets Catalog a proposes two size criteria: conservative mass and radius ranges of 0.5 R⊕ < Rp ≤ 1.5 R⊕ or 0.1 M⊕ < Mp ≤ 5 M⊕, and optimistic ranges of 1.5 R⊕ < Rp ≤ 2.5 R⊕ or 5 M⊕ < Mp ≤ 10 M⊕. Planets within the HZ falling in either of these ranges are listed in Table 4.2. There are twelve planets in the conservative mass range, seven within the conservative HZ limits and five in the optimistic HZ. A further four are in the optimistic mass range, three of which are within the conservative HZ limits. It should be noted here that the existence of GJ 667 C e and f is disputed, and they may be stellar noise artefacts (Feroz & Hobson, 2014). Nevertheless, this is an interesting sample of planets for potential in depth-characterisation.

a. Located at http://phl.upr.edu/projects/habitable-exoplanets-catalog, maintained by the University of Puerto Rico at Arecibo 4.5. Habitable zone planets 141

Figure 4.11– Systems with planets in the Habitable Zone orbiting M dwarfs. The optimistic and conservative HZs deﬁned by Kopparapu et al.(2013) are plotted in light and dark green respectively. Each system is shown in a diﬀerent colour, and stellar host names are indicated. The planet sizes are proportional to their masses.

As described in Sect. 1.2, whether or not M dwarf planets can in fact be habitable is a question that is still hotly debated. Many factors such as tidal-locking, atmospheric circulation, stellar irradiation, among others, will influence the habitability of these planets. Therefore, the answer to if any specific one of the planets highlighted in Table 4.2 is truly habitable will require careful study and modelling of the host star and the planet itself; the HZ limits provide merely a first approximation. Future atmospheric studies of M dwarf planets, with JWST and/or ELTs, will be able to characterise the chemical composition and thermal profiles, and hunt for biomarkers (e.g. Birkby et al. 2013, Serindag & Snellen 2019, Lustig-Yaeger et al. 2019). 142 Chapter 4. Properties of M-dwarf planets as of 2019

Table 4.2– Likely rocky habitable zone planets (top half of the table) and potentially rocky habitable zone planets (bottom half).

Name Mass [M⊕] Radius [R⊕] Semimajor axis [AU] HZ range +0.033 +0.00066 TRAPPIST-1 d 0.297−0.034 0.784 ± 0.023 0.02144−0.00063 cons. +0.142 +0.00047 TRAPPIST-1 c 1.156−0.131 1.095 ± 0.031 0.01521−0.00047 cons. +0.19 +0.0041 Proxima Centauri b 1.27−0.19 0.0485−0.0041 cons. +0.21 +0.0017 Ross 128 b 1.40−0.21 0.0496−0.0017 cons. +1.6 +0.02 GJ 667 C e 2.7−1.4 0.213−0.02 cons. +1.4 +0.014 GJ 667 C f 2.7−1.2 0.156−0.017 cons. +0.89 +0.008 Kapteyn’s b 4.80−0.95 0.168−0.008 cons. +0.154 +0.00034 TRAPPIST-1 b 1.016−0.143 1.121 ± 0.032 0.01111−0.00034 opt. +0.17 +0.00093 YZ Cet d 1.14−0.17 0.02764−0.00093 opt. +0.44 +0.0017 GJ 1132 c 2.64−0.44 0.0476−0.0017 opt. +0.45 +0.0029 Wolf 1061 c 3.40−0.41 0.089−0.0031 opt. +1.5 +0.012 GJ 667 C c 3.8−1.2 0.125−0.013 opt. +5.8 +0.03 GJ 682 c 8.7−4.6 0.176−0.009 cons. +1.8 +0.024 GJ 667 C d 5.1−1.7 0.276−0.03 cons. +3.9 +0.018 GJ 357 d 7.7−3.5 0.204−0.022 cons. +0.99 +0.00018 GJ 3293 d 7.63−0.99 0.194−0.00018 opt.

4.6 A mass-metallicity correlation?

As discussed in Sect. 1.4, the idea of a link between stellar metallicity and planetary presence has long been hypothesised: stars hosting planets would tend to be more metal- rich than those without planets. Likewise, it has been proposed that the mass of the initial circumstellar disk from which the planets are formed (e.g. Gáspár et al. 2016), the masses of the planets hosted (e.g. Courcol et al. 2016), and the total sum of planetary masses (e.g Bonﬁls et al. 2007) all increase with stellar metallicity. Figures 4.12a and 4.12b show the planetary masses vs metallicity and sum of planetary masses vs metallicity respectively, for M dwarf planets with known stellar metallicities and precisely determined masses. The vast majority of these correspond to RV detections, which is only to be expected as precise metallicities are generally obtained from stellar spectra. The only exceptions are OGLE-2012-BLG-0563L b, a microlensing exoplanet with stellar metallicity determined from J − Ks and H − Ks colours, and the TRAPPIST-1 system, with masses determined by TTVs and stellar metallicity from nIR spectroscopy. 4.7. Multiplanetary systems orbiting M dwarfs 143

(a) Planet mass (or minimum mass M sin i) (b) Sum of RV-determined planetary as a function of stellar metallicity. The masses (or minimum masses M sin i) per boundary from Courcol et al.(2016) is system as a function of stellar metallicity. indicated by the dashed black line.

Figure 4.12– Planet mass or minimum mass M sin i (left, 1.9a), or sum of masses/minimum masses per system (right, 1.9b), as a function of stellar metallicity for all known exoplanets (as of 2019) hosted by M dwarfs, with these parameters reported in "The Exoplanet Encyclopaedia". Only planets with masses measured by RVs, TTVs, or microlensing are shown in 4.12a, and only RV-determined masses in 4.12b.

The stellar metallicities in Fig. 4.12 come from varied sources, so any trends must be viewed with caution. However, there does seem to be a moderately significant correlation between metal-rich hosts and higher planetary mass for RV-dermined masses (Pear- son correlation coefficient ρ = 0.46, p-value = 5.02 × 10−5). In particular, the planets with masses below 40 M⊕ follow the metallicity-dependent boundary defined by Courcol et al.(2016) for this mass range well. Likewise, the total sum of planetary masses also shows a moderate correlation with metallicity (Pearson correlation coefficient ρ = 0.42, p-value = 6.05 × 10−3). This aligns well with previous findings, allowing us to conclude that it seems the tendency for planetary masses to be larger for more metallic hosts holds for M dwarf planets. This tendency in turn supports the core accretion model of planetary formation (see e.g. Hobson et al. 2018b, Courcol et al. 2016, Sousa et al. 2019).

4.7 Multiplanetary systems orbiting M dwarfs

Twelve multiplanetary systems have been discovered since 2016, bringing the total up to 32 systems containing 87 planets. The best-known of these is the TRAPPIST- 1 system (Gillon et al., 2017), composed of seven earth-sized planets orbiting an M8 star. The system is very compact, with all the orbits within 0.1 AU (for comparison, Mercury orbits at 0.4 AU from the Sun). Three of the planets orbit within the liquid water habitable zone. 144 Chapter 4. Properties of M-dwarf planets as of 2019

For the rest, there are sixteen two-planet systems, twelve three-planet systems, three with four planets, and one with up to six planets. The six-planet system, GJ 667C, is particularly interesting in that (unlike, for instance, TRAPPIST-1), the planets were not all discovered simultaneously. Rather, they were successively unfolded through follow-up observations and detailed analysis (Anglada-Escudé et al. 2012, Anglada-Escudé et al. 2013), and the existence of some of the planets is debated (Feroz & Hobson 2014, Robertson & Mahadevan 2014, Cuartas-Restrepo et al. 2016). Three other systems also have planets discovered at different moments, at least one of which emerged during the last three years: GJ 1132 (Planet b: Berta-Thompson et al. 2015, using MEarth photometry and HARPS RVs; planet c: Bonﬁls et al. 2018 using HARPS RVs), HIP 57050 (Planet b: Haghighipour et al. 2010, using Keck/HIRES RVs; planet c: Trifonov et al. 2018, using CARMENES RVs) and LHS 1140 (Planet b: Dittmann et al. 2017; planet c: Ment et al. 2019, both using MEarth photometry and HARPS RVs). This highlights the importance of continued observations of known planet- hosting stars. Conclusions

Contents

5.1 Overview of results and conclusions...... 146 5.2 Future perspectives...... 147 5.2.1 SOPHIE-red ...... 148 5.2.2 Synergies between the SPIRou Legacy Survey and other M dwarf surveys ...... 148 5.2.3 Overlap with TESS (SPIRou - SOPHIE) ...... 149 5.2.4 Beyond SOPHIE and SPIRou: New and upcoming spectrographs . 150 146 Conclusions

5.1 Overview of results and conclusions

In the preceding chapters, I have presented the results of my work on the radial velocity planet hunt around M dwarfs in the visible and nIR domains. I have also carried out a general analysis of the entire M dwarf exoplanet sample. For my work with SOPHIE in the visible, I re-analysed the entire M dwarf sample, applying a template-matching algorithm to SOPHIE spectra for the first time. Through this re-analysis, including the calibration of nightly and long-term instrumental effects, we have been able to conclusively detect and publish four planets around M dwarfs: Gl 96 b, Gl 617A b, Gl 411 b, and Gl 378 b. Template-matching was essential to these detections, as with the CCF analysis the signals were partly hidden by noise. I also studied the effects of stellar activity on the SOPHIE spectra, finding that the Hα and 0 log(RHK) indices are the most effective indicators of stellar activity in SOPHIE M dwarf spectra. Stellar activity appears to impact the template-matching radial velocities more strongly than the usual CCF radial velocities, generating more significant signals at the rotation period (and/or aliases/harmonics thereof). Therefore, it is even more important when using template-matching to evaluate the stellar activity and take it into account. This work is presented in Ch.2. For my work with SPIRou in the nIR, I led the development of the wavelength calibration. I implemented and tested different methods of generating a wavelength-pixel correspondence, using either hollow-cathode lamps alone or combined hollow-cathode lamps and a Fabry-Pérot étalon. The HC lamps alone did not provide sufficient accuracy, being at the level of ∼2–3 m s−1 internal RMS, while the error budget prevision was of <0.45 m s−1. The combined HC-FP solutions, on the other hand, have an excellent internal RMS of ∼0.15 m s−1. The night-to-night stability is complicated by the dependence of the wavelength solution on the previous calibrations, especially on the slit determination. Fixing all the calibrations produces a noticeable ∼10–50 m s−1 but constant RV offset between solutions; when this offset is removed, the night-to-night variations are greatly diminished. I analysed the impact of changing the wavelength solution on the radial velocity calculations, finding that the calculated RVs remain fairly consistent with ∼0.5–0.7 m s−1 dispersions, and that the drift computation is efficient at removing the RV offset between wavelength solutions computed with a fixed set of previous calibrations. The full description of this work is found in Ch.3. Finally, I studied the entire population of planets found around M dwarfs in a global perspective. There is a clear lack of both hot Neptunes and hot Jupiters. While hot Neptunes are lacking across the entire planet population regardless of stellar host, in what is known as the hot Neptune desert, the lack of hot Jupiters is a feature of the M dwarf planets specifically. Planets close to the lower boundary of the hot Neptune desert tend to be hosted by more metal-rich, larger, warmer stars. Most of the planets found are relatively low-mass and short-period (1 − 10 M⊕, 1–200 d). Simultaneous masses and radii are available for only a few planets, which are spread over a wide range of compositions, from Mercury-like to 100% water. For sixteen systems, there is at least one planet within the optimistic HZ limits. There appears to be a moderately significant correlation between stellar metallicity and planetary mass, with larger planets tending 5.2. Future perspectives 147 to be hosted by more metallic stars. 87 of the 196 planets currently known are part of multiplanetary systems; with continued monitoring and more sensitive instruments, we can expect this number to grow as previously inaccessible planets become detectable. The detailed analysis is in Ch.4. To conclude, through the development of this thesis, I have been able to study the detection of exoplanets around M dwarfs via radial velocities, from raw spectra through to planet characterisation, in both the nIR and visible domains. The radial velocity method has already shown itself extremely productive in the detection of new planets and the confirmation of transiting candidates in the visible, and we expect this productivity to greatly increase with the advent of high-precision nIR spectroscopy. The major contributions of my work have been oriented to enhancing this productivity by improving the accuracy of the radial velocities measured. This need for more accurate radial velocities is particularly driven by the desire to detect less massive planets, which are expected to be very frequent around M dwarfs. However, it also allows a better characterisation of all planets, which in turn enables more accurate population studies and the testing of formation and evolution theories. One major factor in the accuracy of radial velocity measurements is the wavelength calibration; the traditional hollow-cathode lamps are not accurate below the m s−1 level in the visible, and I have shown that they are even less so in the near infrared. I have developed methods to combine these lamps with a Fabry-Pérot étalon for a very precise wavelength solution; these are currently implemented in the SPIRou data reduction, and there are plans to adapt them to other instruments (such as SOPHIE). A second factor that needs to be considered is the calculation of the radial velocities themselves. By adapting and applying a template-matching algorithm to a large sample of M dwarfs, I was able to firmly detect several planetary signals, showcasing the great potential of template-matching for M dwarf RV surveys. Conversely, I also found a greater sensitivity of this method to stellar activity. Any survey using template-matching for radial velocity determination should therefore make every effort possible to study stellar activity (e.g., through activity indicators) in order to verify that it is not the originator of the RV periodicities. Nevertheless, template-matching remains a highly promising technique, and will probably be the RV calculation method of choice for most (if not all) M dwarf surveys.

5.2 Future perspectives

The hunt for exoplanets around M dwarfs with radial velocities is a rapidly growing field. Since my thesis work was carried out with SPIRou and SOPHIE, I first highlight some avenues in which they are expected to make important contributions: an upcoming modification of SOPHIE that will enhance its capabilities for M dwarf science; the synergies of the SPIRou Legacy Survey with other M dwarf surveys; and the overlap with the NASA TESS mission, for which both SPIRou and SOPHIE are expected to contribute to the validation of transiting planet candidates. Next, I give a brief overview of some new and near-future spectrographs and how they will contribute to the field. 148 Conclusions

5.2.1 SOPHIE-red

SOPHIE-red is the project name given to a planned extension of the SOPHIE domain towards the red though CCD replacement. The current CCD is a 2k × 4k detector, covering the 387–694 nm wavelength range. The overall quantum efficiency is of Q ∼ 80%, but drops off steeply in the red. It also shows an important CTI effect (as described in Sect. 2.4), and ageing. The goal is therefore to replace it with a new 4k × 4k CCD. The chosen detector has an efficiency of Q ∼ 90%, and will allow the SOPHIE domain to be extended around 5 orders further into the red - a ∼ 12% increase of the current 38 orders. Fig. 5.13 shows the quantum efficiency curves of the two detectors. The detector change is planned to take place before summer 2020, so that SOPHIE-red should be available for semester 2020B.

Figure 5.13– Quantum eﬃciency curves for the current (red) and planned (yellow) SOPHIE CCDs. The future CCD is signifcantly more responsive in the red.

With this upgrade, the RV content available for M dwarfs will be substantially aug- mented, thereby allowing SOPHIE to remain a competitive instrument for M dwarf exoplanet science. An enhancement of the SOPHIE DRS wavelength solution is also planned, by adapting the FP routines developed for SPIRou to SOPHIE data.

5.2.2 Synergies between the SPIRou Legacy Survey and other M dwarf surveys

As described in Sect. 3.6.1, the SPIRou Legacy Survey has among its targets 47 stars that have been observed with SOPHIE. Likewise, similar numbers have been previously observed by the KECK/HIRES survey (Butler et al., 2017) and/or the HARPS M dwarf program (Bonfils et al., 2013). This leads to several avenues for synergy. Previously published planets can be independently confirmed (or refuted), and additional planets 5.2. Future perspectives 149 can be searched for. Examples of such independent confirmations and refutations (using SOPHIE) were discussed in Sect. 2.5.2. All the KECK/HIRES data up to 2017, and a large part of the HARPS data, is publicly available. Therefore, combined analysis can be performed for common targets. Such analyses can provide: better constraints on planetary parameters (through a higher number of data points, at different wavelengths); coverage of gaps in the sampling of each instrument; distinguishing between aliased signals; a long temporal baseline, on which to search for longer-period planets; among others. Stellar activity can also be investigated in both wavelength domains, an important task as most M dwarfs are at least moderately active and the stellar activity can mimic planets.

5.2.3 Overlap with TESS (SPIRou - SOPHIE)

TESS (Transiting Exoplanet Survey Satellite) is a NASA mission designed to search for transiting planets around bright, nearby stars (Ricker et al., 2015). It is planned to cover around 85% of the sky, observing successive 24◦ × 96◦ sectors from near the ecliptic to beyond the ecliptic pole. It will sweep through 13 sectors in each hemisphere. Observations were started in July 2018, on the southern hemisphere. Currently (June 2019) TESS is observing the 13th and final sector of this hemisphere; it is scheduled to start sector 14, the first sector of the northern hemisphere, on the 18th of July 2019. The TESS sectors are shown in Fig. 5.14. The expected yield of TESS is described in Barclay et al.(2018). Their simulations predict some 500 planet candidates found around M dwarfs, around 70 of which are in the optimistic habitable zone. Ballard (2019) explored the expected yield for M dwarfs in more detail, predicting that TESS should find 1274 ± 241 planets around 1026 ± 182 stars, with a 20% multiplicity rate. Importantly, their work also suggests that TESS will miss around 93 planets per 200 host stars, due to the sampling rate and mission sensitivity limits. If they transit, they could potentially be detected by photometric follow-up from the ground or space (e.g. by TRAPPIST, ExTrA, CHEOPS). 150 Conclusions

Figure 5.14– TESS observing sectors, in celestial coordinates.

These transiting planet candidates will require RV follow-up in order to confirm their planetary natures and measure their masses. Both SPIRou and SOPHIE are well placed to perform this follow-up. For SPIRou, the SPIRou Legacy Survey has reserved time within its transit follow-up component specifically to observe TESS candidate planets over the next four years. Meanwhile, thanks to its geographical location as the most northern precision radial velocity instrument, SOPHIE is ideally placed to follow up the northern TESS targets, including the continuous viewing zone around the ecliptic pole that TESS will observe for a year. As well as confirming transiting planets, spectroscopic follow-up could also potentially identify additional planets that were not detected by TESS, e.g. because their orbital periods are too long or because they do not transit.

5.2.4 Beyond SOPHIE and SPIRou: New and upcoming spectrographs

While it is not specifically aimed at M dwarf exoplanetary science, in the quest for extremely precise radial velocities to detect earth-like planets, it is not possible to omit mention of ESPRESSO. Described in González Hernández et al.(2018), ESPRESSO is a visible spectrograph covering the 380–780 range, on the Very Large Telescope. It can use light from one or up to all four of the 8.2 m Unit Telescopes (UT), and has three resolution modes: HR (one UT) at R = 134 000, UHR (one UT) at R = 200 000, and MR (one to four UTs) at R = 59 000. The main factor that makes ESPRESSO unique is its RV precision of 10 s−1, which will enable the detection of Earth-like planets at distances of 1 AU. In addition to the detection of exoplanets, spectroscopy can also be used to characterise planetary atmospheres of transiting planets. During transit, part of the stellar light will pass through the atmosphere, imprinting atmospheric spectral features onto 5.2. Future perspectives 151 the stellar spectrum. Thus, by comparing out-of-transit and in-transit spectra, the atmosphere can be characterised. This technique is most useful for close-in planets with large, inflated atmospheres. The first such detection was achieved by Charbonneau et al. (2002), who found sodium in the atmosphere of HD 209458 with the HST STIS spectrograph. While the first studies relied on medium resolution space-borne spectroscopy, results are also being achieved from the ground with high-resolution spectrographs (see e.g. Pino et al. 2018 and references therein). Some detections have been achieved for M dwarf planets (e.g. GJ 436 b, Knutson et al. 2014, Morley et al. 2017; GJ 1214 b, Kreidberg et al. 2014), and are expected to increase in the future as more good targets are discovered and instrument precision improves. For M dwarfs specifically, many efforts are moving toward the nIR spectral domain, as described in Sect. 1.3.2. Many new nIR spectrographs are starting operations or will do so in the near future. In addition to CARMENES and NIRPS, which were highlighted in Sect. 1.3.2, there are also:

• IRD: Described in Kotani et al.(2018), IRD is a spectrograph covering the Y, J, and H bands at R = 70 000, on the 8.2 m Subaru telescope. It uses adaptive-optics-fed fibre injection, with two multi-mode and two single-mode fibres any two of which can be used simultaneously. It has 2 H2RG detectors, and uses a laser frequency comb at 12.5 GHz for wavelength calibration, for an internal error close to 1.3– 1.9 m s−1. A survey comprising 170 nights of GTO over 5 years started in Feb 2019, dedicated to observing 150 quiet nearby middle-to-late M dwarfs. • HPF: Described in Mahadevan et al.(2014), HPF covers the 0.8–1.27 µm wavelength range at R = 55 000, at the 10 m Hobby-Eberly Telescope. The instrument possesses an H2RG detector, a double scrambler and fibre agitator, and is wavelength-calibrated with a laser frequency comb at 30 GHz. A GTO programme of 900 hours over 5 years, started mid 2018, is observing mid-to-late M dwarfs. • iSHELL: Described in Cale et al.(2018), iSHELL is a spectrograph covering the 1.1– 5.3 µm range at R = 70 000, on the 3 m NASA Infrared Telescope Facility. The wavelength in the K band is calibrated through a methane gas cell. iSHELL has been on sky for 1.5 yr, and currently reaches 2.5 m s−1 precision. Current programmes aim to do TESS follow-up, and to study stellar activity through RV colour. • GIANO-B and GIARPS: Described in Claudi et al.(2018), GIANO-B is a modification of the GIANO spectrograph at the 3.58 m Telescopio Nazionale Galileo. GIANO covers the 0.95–2.45 µm range at R = 50 000, and is wavelength-calibrated using telluric lines, for a precision of ∼8 m s−1. The GIANO-B modifications consisted in moving the instrument to the Nasmyth-B focus, and changing the light input from a fibre-fed system to pre-slit optics train. These changes enable it to be used jointly with HARPS-N; the combined observing mode is called GIARPS. Since 2018, the GAPS2.0 programme has been searching for very young Hot Jupiters. • iLocater: Described in Crepp et al.(2016), iLocater is an upcoming spectrograph for the Large Binocular Telescope. It will cover the 960–1270 nm range at R = 200 000. The instrument uses single-mode fibres with AO, and a custom grating. Commissioning will take place in summer 2019, with a nominal delivery date of fall 2020. 152 Conclusions

First results for many of these instruments were presented at the EPRV IV conference b. All showed similar RV precision on Barnard’s star, with values in the 1.5–2.7 m s−1 range. Likewise, several challenges were seen by many of the teams: fibre modal noise, CMOS detector effects, complex wavelength calibration, difficulties of telluric correction, among others. These difficulties will need to be solved for nIR spectroscopy to reach its full potential. Given the strong community investment and the proliferation of these instruments, we can confidently expect significant advances to be made, further pushing the limits on exoplanet detection around M dwarfs via radial velocities. Some important questions that we can hope to answer in the upcoming years are: the statistics and properties of planets around mid-to-late M dwarfs, most of which are too faint for visible spectroscopy, which will also enable the study of planet formation around very low-mass stars; a more complete census of low-mass planets orbiting nearby stars; the identification of Earth-mass planets in the habitable zone, including the best candidates for atmospheric characterisation and biomarker searches.

b. Event website, including presentation materials: https://indico.cern.ch/event/743355/ Bibliography

Adibekyan, V. 2019, Geosciences, 9, 105 Adibekyan, V. Z., Figueira, P., Santos, N. C., et al. 2013, A&A, 560, A51 Affer, L., Damasso, M., Micela, G., et al. 2019, A&A, 622, A193 Affer, L., Micela, G., Damasso, M., et al. 2016, A&A, 593, A117 Almenara, J. M., Díaz, R. F., Dorn, C., Bonfils, X., & Udry, S. 2018, MNRAS, 478, 460 Andretta, V., Busà, I., Gomez, M. T., & Terranegra, L. 2005, A&A, 430, 669 Anglada-Escude, G., Arriagada, P., Tuomi, M., et al. 2014, MNRAS, 443, L89 Anglada-Escudé, G., Arriagada, P., Vogt, S. S., et al. 2012, ApJ, 751, L16 Anglada-Escudé, G. & Butler, R. P. 2012, ApJS, 200, 15 Anglada-Escudé, G., Tuomi, M., Arriagada, P., et al. 2016, ApJ, 830, 74 Anglada-Escudé, G., Tuomi, M., Gerlach, E., et al. 2013, A&A, 556, A126 Artigau, É., Kouach, D., Donati, J.-F., et al. 2014, in Proc. SPIE, Vol. 9147, Ground-based and Airborne Instrumentation for Astronomy V, 914715 Artigau, É., Malo, L., Doyon, R., et al. 2018, AJ, 155, 198 Astudillo-Defru, N. 2015, Theses, Université Grenoble Alpes Astudillo-Defru, N., Bonfils, X., Delfosse, X., et al. 2015, A&A, 575, A119 Astudillo-Defru, N., Delfosse, X., Bonfils, X., et al. 2017a, A&A, 600, A13 Astudillo-Defru, N., Forveille, T., Bonfils, X., et al. 2017b, A&A, 602, A88 Bakos, G. Á., Bayliss, D., Bento, J., et al. 2018, arXiv e-prints, arXiv:1812.09406 Ballard, S. 2019, AJ, 157, 113 Baluev, R. V. 2008, MNRAS, 385, 1279 Baraffe, I., Chabrier, G., Barman, T. S., et al. 2005, A&A, 436, L47 Baraffe, I., Chabrier, G., Fortney, J., & Sotin, C. 2014, in Protostars and Planets VI, ed. H. Beuther, R. S. Klessen, C. P. Dullemond, & T. Henning, 763 Baranne, A., Queloz, D., Mayor, M., et al. 1996, A&AS, 119, 373 Barclay, T., Pepper, J., & Quintana, E. V. 2018, ApJS, 239, 2 Barnes, J. R., Jeffers, S. V., & Jones, H. R. A. 2011, MNRAS, 412, 1599 154 Bibliography

Bashi, D., Helled, R., & Zucker, S. 2018, Geosciences, 8, 325 Bauer, F. F., Zechmeister, M., & Reiners, A. 2015, A&A, 581, A117 Beaugé, C. & Nesvorný, D. 2013, ApJ, 763, 12 Berta-Thompson, Z. K., Irwin, J., Charbonneau, D., et al. 2015, Nature, 527, 204 Birkby, J., de Kok, R., Brogi, M., et al. 2013, The Messenger, 154, 57 Blank, R., Beletic, J. W., Cooper, D., et al. 2012, in Society of Photo-Optical Instrumen- tation Engineers (SPIE) Conference Series, Vol. 8453, Proc. SPIE, 84530V Blind, N., Conod, U., & Wildi, F. 2017, arXiv e-prints, arXiv:1711.00835 Boisse, I., Bouchy, F., Hébrard, G., et al. 2011, A&A, 528, A4 Boisse, I., Eggenberger, A., Santos, N. C., et al. 2010, A&A, 523, A88 Boisse, I., Moutou, C., Vidal-Madjar, A., et al. 2009, A&A, 495, 959 Boisse, I., Perruchot, S., Bouchy, F., et al. 2016, in Society of Photo-Optical Instrumenta- tion Engineers (SPIE) Conference Series, Vol. 9908, Proc. SPIE, 990868 Bonfils, X., Almenara, J. M., Cloutier, R., et al. 2018, A&A, 618, A142 Bonfils, X., Almenara, J. M., Jocou, L., et al. 2015, in Proc. SPIE, Vol. 9605, Techniques and Instrumentation for Detection of Exoplanets VII, 96051L Bonfils, X., Delfosse, X., Udry, S., et al. 2013, A&A, 549, A109 Bonfils, X., Mayor, M., Delfosse, X., et al. 2007, A&A, 474, 293 Borucki, W. J., Koch, D., Basri, G., et al. 2010, Science, 327, 977 Boss, A. P., Butler, R. P., Hubbard, W. B., et al. 2007, Transactions of the International Astronomical Union, Series A, 26, 183 Bouchy, F., Díaz, R. F., Hébrard, G., et al. 2013, A&A, 549, A49 Bouchy, F., Doyon, R., Artigau, É., et al. 2017, The Messenger, 169, 21 Bouchy, F., Hébrard, G., Udry, S., et al. 2009a, A&A, 505, 853 Bouchy, F., Isambert, J., Lovis, C., et al. 2009b, in EAS Publications Series, ed. P. Kern, Vol. 37, 247–253 Bouchy, F., Pepe, F., & Queloz, D. 2001, A&A, 374, 733 Bouchy, F. & Sophie Team. 2006, in Tenth Anniversary of 51 Peg-b: Status of and prospects for hot Jupiter studies, ed. L. Arnold, F. Bouchy, & C. Moutou, 319–325 Brugger, B., Mousis, O., Deleuil, M., & Deschamps, F. 2017, ApJ, 850, 93 Buchhave, L. A., Latham, D. W., Johansen, A., et al. 2012, Nature, 486, 375 Burt, J., Vogt, S. S., Butler, R. P., et al. 2014, ApJ, 789, 114 Butler, R. P., Marcy, G. W., Williams, E., Hauser, H., & Shirts, P. 1997, ApJ, 474, L115 Bibliography 155

Butler, R. P., Vogt, S. S., Laughlin, G., et al. 2017, AJ, 153, 208 Cale, B., Plavchan, P., Gagné, J., et al. 2018, arXiv e-prints, arXiv:1803.04003 Campbell, B., Walker, G. A. H., & Yang, S. 1988, ApJ, 331, 902 Cannon, A. J. 1915, JRASC, 9, 203 Cersullo, F., Wildi, F., Chazelas, B., & Pepe, F. 2017, A&A, 601, A102 Charbonneau, D., Brown, T. M., Noyes, R. W., & Gilliland, R. L. 2002, ApJ, 568, 377 Chazelas, B., Pepe, F., Wildi, F., et al. 2010, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 7739, Proc. SPIE, 773947 Christiansen, J. 2018, Exoplanet Catalogs, 150 Claudi, R., Benatti, S., Carleo, I., et al. 2018, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 10702, Proc. SPIE, 107020Z Cloutier, R., Artigau, É., Delfosse, X., et al. 2018, AJ, 155, 93 Courcol, B., Bouchy, F., & Deleuil, M. 2016, MNRAS, 461, 1841 Courcol, B., Bouchy, F., Pepe, F., et al. 2015, A&A, 581, A38 Cox, A. N. 2000, Allen’s astrophysical quantities Cram, L. E. & Mullan, D. J. 1979, ApJ, 234, 579 Crepp, J. R., Crass, J., King, D., et al. 2016, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 9908, Proc. SPIE, 990819 Cuartas-Restrepo, P. A., Melita, M., Zuluaga, J. I., et al. 2016, MNRAS, 463, 1592 Dall, T. H., Santos, N. C., Arentoft, T., Bedding, T. R., & Kjeldsen, H. 2006, A&A, 454, 341 Delfosse, X., Bonﬁls, X., Forveille, T., et al. 2013, A&A, 553, A8 Delfosse, X., Forveille, T., Mayor, M., et al. 1998, A&A, 338, L67 Delfosse, X., Forveille, T., Ségransan, D., et al. 2000, A&A, 364, 217 Delisle, J.-B., Ségransan, D., Buchschacher, N., & Alesina, F. 2016, A&A, 590, A134 Desort, M., Lagrange, A. M., Galland, F., Udry, S., & Mayor, M. 2007, A&A, 473, 983 Díaz, R. F., Cincunegui, C., & Mauas, P. J. D. 2007, MNRAS, 378, 1007 Díaz, R. F., Delfosse, X., Hobson, M. J., et al. 2019, A&A, 625, A17 Dittmann, J. A., Irwin, J. M., Charbonneau, D., et al. 2017, Nature, 544, 333 Dong, S., Xie, J.-W., Zhou, J.-L., Zheng, Z., & Luo, A. 2018, Proceedings of the National Academy of Science, 115, 266 Dressing, C. D. & Charbonneau, D. 2015, ApJ, 807, 45 Dumusque, X., Pepe, F., Lovis, C., & Latham, D. W. 2015, ApJ, 808, 171 156 Bibliography

Eigmüller, P., Chaushev, A., Gillen, E., et al. 2019, A&A, 625, A142 Feroz, F. & Hobson, M. P. 2014, MNRAS, 437, 3540 Fischer, D. A., Anglada-Escude, G., Arriagada, P., et al. 2016, PASP, 128, 066001 Fischer, D. A., Howard, A. W., Laughlin, G. P., et al. 2014, in Protostars and Planets VI, ed. H. Beuther, R. S. Klessen, C. P. Dullemond, & T. Henning, 715 Fischer, D. A., Marcy, G. W., Butler, R. P., Vogt, S. S., & Apps, K. 1999, PASP, 111, 50 Fischer, D. A. & Valenti, J. 2005, ApJ, 622, 1102 Fouqué, P., Moutou, C., Malo, L., et al. 2018, MNRAS, 475, 1960 Fulton, B. J., Petigura, E. A., Howard, A. W., et al. 2017, AJ, 154, 109 Gaidos, E. 2017, MNRAS, 470, L1 Gaidos, E., Mann, A. W., Lépine, S., et al. 2014, MNRAS, 443, 2561 Gáspár, A., Rieke, G. H., & Ballering, N. 2016, ApJ, 826, 171 Ghezzi, L., Cunha, K., Smith, V. V., et al. 2010, ApJ, 720, 1290 Gillon, M., Triaud, A. H. M. J., Demory, B.-O., et al. 2017, Nature, 542, 456 Gomes da Silva, J., Santos, N. C., Bonﬁls, X., et al. 2011, A&A, 534, A30 González Hernández, J. I., Pepe, F., Molaro, P., & Santos, N. C. 2018, ESPRESSO on VLT: An Instrument for Exoplanet Research, 157 Goudfrooij, P., Bohlin, R. C., Maíz-Apellániz, J., & Kimble, R. A. 2006, PASP, 118, 1455 Grimm, S. L., Demory, B.-O., Gillon, M., et al. 2018, A&A, 613, A68 Haghighipour, N., Vogt, S. S., Butler, R. P., et al. 2010, ApJ, 715, 271 Haywood, R. D., Collier Cameron, A., Queloz, D., et al. 2014, MNRAS, 443, 2517 Hébrard, É. M., Donati, J. F., Delfosse, X., et al. 2016, MNRAS, 461, 1465 Henry, T. J., Jao, W.-C., Subasavage, J. P., et al. 2006, AJ, 132, 2360 Hobson, M. J., Delfosse, X., Astudillo-Defru, N., et al. 2019, A&A, 625, A18 Hobson, M. J., Díaz, R. F., Delfosse, X., et al. 2018a, A&A, 618, A103 Hobson, M. J., Jofré, E., García, L., Petrucci, R., & Gómez, M. 2018b, Rev. Mexicana Astron. Astroﬁs., 54, 65 Howard, A. W., Marcy, G. W., Fischer, D. A., et al. 2014, ApJ, 794, 51 Izidoro, A., Bitsch, B., Raymond, S. N., et al. 2019, arXiv e-prints, arXiv:1902.08772 Jackson, R. J. & Jeffries, R. D. 2013, MNRAS, 431, 1883 Jacob, W. S. 1855, MNRAS, 15, 228 Jin, S. & Mordasini, C. 2018, ApJ, 853, 163 Bibliography 157

Kaltenegger, L. 2010, ApJ, 712, L125 Kaltenegger, L. & Traub, W. A. 2009, ApJ, 698, 519 King, G. W., Wheatley, P. J., Salz, M., et al. 2018, MNRAS, 478, 1193 Knutson, H. A., Benneke, B., Deming, D., & Homeier, D. 2014, Nature, 505, 66 Kopparapu, R. K., Ramirez, R., Kasting, J. F., et al. 2013, ApJ, 765, 131 Kotani, T., Tamura, M., Nishikawa, J., et al. 2018, in Society of Photo-Optical Instrumen- tation Engineers (SPIE) Conference Series, Vol. 10702, Proc. SPIE, 1070211 Kreidberg, L., Bean, J. L., Désert, J.-M., et al. 2014, Nature, 505, 69 Kürster, M., Endl, M., Rouesnel, F., et al. 2003, A&A, 403, 1077 Lalitha, S., Baroch, D., Morales, J. C., et al. 2019, A&A, 627, A116 Laughlin, G., Bodenheimer, P., & Adams, F. C. 1997, ApJ, 482, 420 Lecavelier Des Etangs, A. 2007, A&A, 461, 1185 Ludwig, H. G., Allard, F., & Hauschildt, P. H. 2002, A&A, 395, 99 Lustig-Yaeger, J., Meadows, V. S., & Lincowski, A. P. 2019, AJ, 158, 27 Mahadevan, S., Ramsey, L. W., Terrien, R., et al. 2014, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 9147, Proc. SPIE, 91471G Mann, A. W., Dupuy, T., Kraus, A. L., et al. 2019, ApJ, 871, 63 Mann, A. W., Feiden, G. A., Gaidos, E., Boyajian, T., & von Braun, K. 2015, ApJ, 804, 64 Marchwinski, R. C., Mahadevan, S., Robertson, P., Ramsey, L., & Harder, J. 2015, ApJ, 798, 63 Marcy, G. W., Butler, R. P., Vogt, S. S., Fischer, D., & Lissauer, J. J. 1998, ApJ, 505, L147 Martin, J., Fuhrmeister, B., Mittag, M., et al. 2017, A&A, 605, A113 Mayor, M., Marmier, M., Lovis, C., et al. 2011, arXiv e-prints, arXiv:1109.2497 Mayor, M., Pepe, F., Queloz, D., et al. 2003, The Messenger, 114, 20 Mayor, M. & Queloz, D. 1995, Nature, 378, 355 Mazeh, T., Holczer, T., & Faigler, S. 2016, A&A, 589, A75 Ment, K., Dittmann, J. A., Astudillo-Defru, N., et al. 2019, AJ, 157, 32 Mittag, M., Hempelmann, A., Schmitt, J. H. M. M., et al. 2017, A&A, 607, A87 Mordasini, C., Alibert, Y., Benz, W., Klahr, H., & Henning, T. 2012, A&A, 541, A97 Morin, J., Donati, J. F., Petit, P., et al. 2010, MNRAS, 407, 2269 Morley, C. V., Knutson, H., Line, M., et al. 2017, AJ, 153, 86 Moutou, C., Hébrard, E. M., Morin, J., et al. 2017, MNRAS, 472, 4563 158 Bibliography

Neves, V., Bonﬁls, X., Santos, N. C., et al. 2014, A&A, 568, A121 Newton, E. R., Charbonneau, D., Irwin, J., & Mann, A. W. 2015, ApJ, 800, 85 Newton, E. R., Irwin, J., Charbonneau, D., et al. 2017, ApJ, 834, 85 Noyes, R. W., Hartmann, L. W., Baliunas, S. L., Duncan, D. K., & Vaughan, A. H. 1984, ApJ, 279, 763 Ochsenbein, F., Bauer, P., & Marcout, J. 2000, A&AS, 143, 23 Oshagh, M., Santos, N. C., Figueira, P., et al. 2017, A&A, 606, A107 Owen, J. E. & Lai, D. 2018, MNRAS, 479, 5012 Pasquini, L. & Pallavicini, R. 1991, A&A, 251, 199 Pepe, F., Mayor, M., Galland, F., et al. 2002a, A&A, 388, 632 Pepe, F., Mayor, M., Queloz, D., et al. 2004, A&A, 423, 385 Pepe, F., Mayor, M., Rupprecht, G., et al. 2002b, The Messenger, 110, 9 Perger, M., Scandariato, G., Ribas, I., et al. 2019, A&A, 624, A123 Perruchot, S., Bouchy, F., Chazelas, B., et al. 2011, in Proc. SPIE, Vol. 8151, Techniques and Instrumentation for Detection of Exoplanets V, 815115 Perruchot, S., Hobson, M., Bouchy, F., et al. 2018, in Society of Photo-Optical Instrumen- tation Engineers (SPIE) Conference Series, Vol. 10702, Ground-based and Airborne Instrumentation for Astronomy VII, 1070265 Perruchot, S., Kohler, D., Bouchy, F., et al. 2008, in Proc. SPIE, Vol. 7014, Ground-based and Airborne Instrumentation for Astronomy II, 70140J Perryman, M. 2018, The Exoplanet Handbook Petersburg, R. R., McCracken, T. M., Eggerman, D., et al. 2018, ApJ, 853, 181 Petigura, E. A., Marcy, G. W., Winn, J. N., et al. 2018, AJ, 155, 89 Pickles, A. J. 1998, PASP, 110, 863 Pinamonti, M., Sozzetti, A., Giacobbe, P., et al. 2019, arXiv e-prints, arXiv:1903.11853 Pino, L., Ehrenreich, D., Allart, R., et al. 2018, A&A, 619, A3 Queloz, D. 1995, in IAU Symposium, Vol. 167, New Developments in Array Technology and Applications, ed. A. G. D. Philip, K. Janes, & A. R. Upgren, 221 Queloz, D., Henry, G. W., Sivan, J. P., et al. 2001, A&A, 379, 279 Quirrenbach, A., Amado, P. J., Caballero, J. A., et al. 2014, in Ground-based and Air- borne Instrumentation for Astronomy V, Vol. 9147, 91471F Quirrenbach, A., Amado, P. J., Caballero, J. A., et al. 2016, in Ground-based and Air- borne Instrumentation for Astronomy VI, Vol. 9908, 990812 Rappaport, S., Levine, A., Chiang, E., et al. 2012, ApJ, 752, 1 Bibliography 159

Rauscher, E. & Marcy, G. W. 2006, PASP, 118, 617 Raymond, S. N., Boulet, T., Izidoro, A., Esteves, L., & Bitsch, B. 2018, MNRAS, 479, L81 Redman, S. L., Lawler, J. E., Nave, G., Ramsey, L. W., & Mahadevan, S. 2011, ApJS, 195, 24 Redman, S. L., Nave, G., & Sansonetti, C. J. 2014, ApJS, 211, 4 Reiners, A., Bean, J. L., Huber, K. F., et al. 2010, ApJ, 710, 432 Reiners, A., Ribas, I., Zechmeister, M., et al. 2018a, A&A, 609, L5 Reiners, A., Shulyak, D., Anglada-Escudé, G., et al. 2013, A&A, 552, A103 Reiners, A., Zechmeister, M., Caballero, J. A., et al. 2018b, A&A, 612, A49 Ricker, G. R. 2016, AGU Fall Meeting Abstracts Ricker, G. R., Winn, J. N., Vanderspek, R., et al. 2015, Journal of Astronomical Tele- scopes, Instruments, and Systems, 1, 014003 Robertson, P. & Mahadevan, S. 2014, ApJ, 793, L24 Robertson, P., Roy, A., & Mahadevan, S. 2015, ApJ, 805, L22 Rojas-Ayala, B., Covey, K. R., Muirhead, P. S., & Lloyd, J. P. 2010, ApJ, 720, L113 Rojas-Ayala, B., Covey, K. R., Muirhead, P. S., & Lloyd, J. P. 2012, ApJ, 748, 93 Rowe, J. F., Bryson, S. T., Marcy, G. W., et al. 2014, ApJ, 784, 45 Saar, S. H. & Donahue, R. A. 1997, ApJ, 485, 319 Sanchis-Ojeda, R., Rappaport, S., Pallè, E., et al. 2015, ApJ, 812, 112 Santerne, A., Díaz, R. F., Moutou, C., et al. 2012, A&A, 545, A76 Santos, N. C., Mayor, M., Naef, D., et al. 2000, A&A, 361, 265 Santos, N. C., Sousa, S. G., Mortier, A., et al. 2013, A&A, 556, A150 Santos, N. C., Udry, S., Mayor, M., et al. 2003, A&A, 406, 373 Sarmiento, L. F., Reiners, A., Huke, P., et al. 2018, A&A, 618, A118 Scargle, J. D. 1982, ApJ, 263, 835 Schöfer, P., Jeffers, S. V., Reiners, A., et al. 2019, A&A, 623, A44 See, T. J. J. 1896, AJ, 16, 17 Serindag, D. B. & Snellen, I. A. G. 2019, ApJ, 871, L7 Shields, A. L., Ballard, S., & Johnson, J. A. 2016, Phys. Rep., 663, 1 Shulyak, D., Reiners, A., Seemann, U., Kochukhov, O., & Piskunov, N. 2014, A&A, 563, A35 Silburt, A., Gaidos, E., & Wu, Y. 2015, ApJ, 799, 180 160 Bibliography

Sing, D. K., Fortney, J. J., Nikolov, N., et al. 2016, Nature, 529, 59 Smith, A. M. S., Cabrera, J., Csizmadia, S., et al. 2018, MNRAS, 474, 5523 Sousa, S. G., Adibekyan, V., Delgado-Mena, E., et al. 2018, A&A, 620, A58 Sousa, S. G., Adibekyan, V., Santos, N. C., et al. 2019, MNRAS, 485, 3981 Southworth, J. 2010, MNRAS, 408, 1689 Stevens, D. J. & Gaudi, B. S. 2013, Publications of the Astronomical Society of the Paciﬁc, 125, 933 Szabó, G. M. & Kálmán, S. 2019, MNRAS, 485, L116 Tarter, J. C., Backus, P. R., Mancinelli, R. L., et al. 2007, Astrobiology, 7, 30 Terrien, R. C., Mahadevan, S., Bender, C. F., et al. 2012, ApJ, 747, L38 "The Exoplanet Encyclopaedia". 1995-2017, The Exoplanet Encyclopaedia, http:// exoplanet.eu/ Trifonov, T., Kürster, M., Zechmeister, M., et al. 2018, A&A, 609, A117 Vaughan, A. H., Preston, G. W., & Wilson, O. C. 1978, PASP, 90, 267 Walker, G. A. H. 2012, New A Rev., 56, 9 Wenger, M., Ochsenbein, F., Egret, D., et al. 2000, A&AS, 143, 9 West, A. A., Hawley, S. L., Walkowicz, L. M., et al. 2004, AJ, 128, 426 Wolszczan, A. & Frail, D. A. 1992, Nature, 355, 145 Wright, J. T. & Robertson, P. 2017, Research Notes of the American Astronomical Society, 1, 51 Zarro, D. M. & Rodgers, A. W. 1983, ApJS, 53, 815 Zechmeister, M. & Kürster, M. 2009, A&A, 496, 577