PONTIFICIA UNIVERSIDAD CATÓLICA DE CHILE FACULTAD DE FÍSICA DEPARTAMENTO DE ASTRONOMÍA Y ASTROFÍSICA
TOWARDS RüCKY PLANETS USING THE PLANET FINDER SPECTROGRAPH
BY
PAMELA V. ARRIAGADA P.
Thesis presented to the Department of Astronomy and Astrophysics of
the Pontificia Universidad Católica de Chile to obtain the degree of PhD in Astrophysics.
Advisors: Dante Minniti (PUC) and R. Paul Butler (CIW, DTM)
Readers: Andrés Jordán (PUC), Patricia Arévalo (UNAB), Julio Chanamé (PUC) and
Gaspar Galaz (PUC)
-- _, ...... __ ...... ,._ 1 RECIBIDO CONICYi PROGRAMA CAPITAL HUMANO AVANZAO'J July, 2012 2 o SEP 2012 Santiago, Chile @2012, Pamela Arriagada HORA: ____~-- FIRMA: ~tr ZDlZ @2012, Pamela Arriagada.
Se autoriza la rer,..ToducciÓJJ total o pctrcia.l, co11 fi11es académicos, por C:lmlquier medio o procedimiento, incluyendo la cita bil1liográfica del documento. Acknowledgments
First, I would like to thank my supervisors, R. Paul Butler and Dante Minniti. Paul has been the most fun and caring mentor. He has always trusted in my work, thank you for guiding me during this process. Dante has been the most understand ing supervisor, thank you for making this journey a non-stressful one. I would also like to thank the Magellan Planet Search Team: Stephen Shectman, Jeff Crane and Ian Thompson, for allowing me to be part of the Planet Finder Spectrograph team. I also grateful to all the other collaborators who have put time and effort in the project : Guillem Anglada-Escudé, Steve Vogt, Eugenio Rivera, James J enkins and Erad Carter. I wouldn't like to forget the staff at Las Campanas: Don Tito and the kitchen staff, who have prepared wonderful meals and beautiful desserts for us on the moun tain, always with a smile on their faces; the telescope operators, who have prevented me from falling asleep during our long observing runs and pretended to like the music I listened to, I couldn't have observed anything without them; the technical and engineering staff, who kecp thc tclescope working and did not let me break PFS when somehow, I was involved in the task of dismounting it. I am obliged to my fellow postgraduate students and friends Paula and Pía, who have had to put up with my craziness for the past 3 years. To my friends at the Theoretical Physics department Tío :Miguel, Bastin and Diego for our enlightening conversations during lunch hours. I'd like to acknowledge CONICYT (Comisión Nacional de Investigación de la Científica y Tecnológica) for having support me financially during my studies. I am also grateful to the Department of Astronomy at Católica for the MECESUP grant which allowed me to spend two months working with Paul at the Carnegie Institution of Washington in Washington DC. This work was also partially funded by the Fondap Center for Astrophysics Nr. 15010003 and by the BASAL Center for Astrophysics and Applied Technologies Nr. 0609. Finally, I want to thank my family: Ivonne, Gustavo, Nacho, Blibli, Claudia, and Matias, for pushing me in every step of the way. Contents
Resumen ix
Abstract XI
1 Introduction 1 1.1 Doppler veloci ty technique 2 1.2 Planets arouncl M clwarfs . 5 1.3 M clwarfs ancl Planet formation 5
1.4 Habitability ancl 'f}Earth 6 1.5 Thesis outline . . . . . 8
2 Chromospheric Activity of Southern Stars 9 2.1 Introcluction ...... 9 2.2 Observations ancl elata recluction . 10 2.3 Analysis ...... 12 2.3.1 Derivation of S índices from MIKE/lVIagellan observations 12
2.3.2 Converting from s~HKE to Mount \Vilson s~d\N 13 2.3.3 Uncertainties ancl ranclom errors . 13 2.3.4 log R;rK ...... · · · · 14 2.3.5 Rotation periocls ancl Ages 15 2.3.6 Radial velocity jitter 16 2.4 Conclusions ...... 17
3 A new deconvolution routine for M-dwarfs 24 3.1 Introcluction ...... 24 3.2 Keck/HIRES observa.tions 25 3.3 The role of Deconvolution in Doppler Analysis 26 3.4 Deconvolution algorithms 27 3.4.1 The problem of deconvolution 27 3.4.2 J ansson technique ...... 28 3.4.3 The Bayesian approach 29 3.4.4 The least squares method using B-splines 31 3.4.5 Comparison of various deconvolution routines on Keck/HIRES template spectra 32 3.5 Results . . . 35 3.6 Conclusions 36
4 The Magellan Planet Search 44 4.1 Introduction ...... 44 4.2 The Old Magellan Planet Search 45 4.2.1 New exoplanets ...... 45 4.3 The New Magellan Planet Search 47 4.3.1 Targets ...... 47 4.4 Observations and data reduction . 48 4.5 S values ...... 49 4.5.1 PFS Caii H index. 49 4.5.2 From SPFS to SMw 49 4.6 Velocity Precision ..... 51 4.7 Preliminary results . . . . 51 4.7.1 A planetary system around the nearby M dwarf GJ 667C with one super-Earth in its habitable zone ...... 51 4.7.2 Two planetary companions to the Nearby M dwarf GJ 221 57 4.8 Summary ...... 65
5 Conclusions and Future Work 71 5.1 Summary and conclusions 71 5.2 Ongoing and future work . 73
Bibliography 75
ll List of Tables
3.1 Resulting uncertainties using each of the deconvolution algorithms for each star: Jansson, Boosted 1VIaximum Likelihood (BJ\!IL), Damped Richardson-Lucy (DRL), lVIaximum Entropy (ME) ancl B-Splines de convolution (Decosp). The three rows for each star correspond to the RMS of all velocities (top), RMS of velocities averaged in 2-hour bins 1 (middle) ancl interna! errors (bottom). These are given in m s- . 38
1 1 3.2 Velocity dispersion in m s- of stable (RiviS< 6 m s- ) J\!I-dwarfs from the California-Carnegie Planet search. The four left columns show the RJ\!IS obtained using all observations (including interna!). The four columns on the right, show the RlVIS obtained using only post-fix (taken with improved CCD mosaic) observations (including interna!)...... 41
4.1 Stellar Properties 65 4.2 Best Keplerian solution to the planetary system around GJ 667C. The numbers in parenthesis indicate the uncertainty in the last two siguificant digits of thc paramctcr valucs. Uuccrtaiutics havc bccn obtained using a Bayesian lVIGMC analysis (Ford 2005) and represent the 68% confidence levels around the preferred solution. All orbital
elements are referred to JD0 = 2453158.7643. The assumed mass of GJ 667C is 0.31 lVLz...... 67 4.3 Best Keplerian solution to the planetary system around GJ 221. The numbers in parenthesis indicate the uncertainty in the last two sig nificant digits of the parameter values. Uncertainties correspond to the standard deviation of the ma.rginalized MCMC samplings. The assumed mass of GJ 221 is 0.77 l\(...... 68
iii 1 Derived S-values from MIKE spectra converted to MW system, SvuKE; chromospheric activity índices, log R~K ; rotation periods, Pro,; ages, log(Agejyears); and estimated jitter, o-~v (Isaacson & Fischer (2010) and Wright (2005)) labeled as 1 and 2 respective! y...... 84
lV List of Figures
1.1 Orbital elements 3
1.2 The orbital region that remaim; continuously habitable during at least 5 Gyr as a function of the stellar mass. Dark gray and light gray areas show the Habitable Zone defined using different criteria (see Selsis et al. 2007 for a more detailecl clescription). The dotted bound aries corresponcl to the extreme theoreticallimits, found with a 100% cloucl cover. The clashed line inclicates the clistance at which a 1 IVIEart.h planet on a circular or·bit becomes tidally locked in less than 1 Gyr. Figure from Selsis et al. 2007 ...... 7
2.1 V, K, H ancl R channels in a representative l\IIIKE spectrum. The relative flux is in arbitrary units. vVavelengths have been shiftecl to zero velocity in orcler to make the measurements. 18
2.2 Comparison between sl\HKE measnrements t.aken in clifierent. epochs. In the left panel, fillecl circles correspond to measured S values using data taken between May 2004 and Sept. 2005 vs. S measured using data taken after Sept. 2005. The solid line corresponcls to a linear regression, yielcling a slope of 0.995 and an intercept of -0.002. In t.he right. panel, filled circles correspond Lo mea:mred S valnes nsing data taken before May 2004 vs. measured using data taken after Sept. 2005. The solid line corresponcls to a linear regression, yielcling a slope of 0.987 andan intercept of -0.011...... 19
V 2.3 Comparison between SMIKE measurements with previous measure
ments that are already calibrated to the S,,1w system. Solid lines correspond to linear regressions for each of the datasets, while the dashed lines show relations of slope unity. For clarity, the Jenkins et al (2008), Henry et al. (1996) and Gray et al. (2006) datasets have an offset of 0.4, 0.8 and 1.2 respedively ...... 20
2.4 Upper panel : Comparison between log R~K indices obtained from MIKE spectra in this work and those obtained from FEROS spectra by Jenkins et al. (2008). The line denotes a 1:1 relationship. Lower
panel : the level of difference between both samples, or rY values vs.
our log R~K indicies. The line denotes O level of difference between the samples...... 21
2.5 Top: T Ceti S-values from Magellan/MIKE to illustrate my measure ment errors, in contrast with the 1% scatter obtained at Mount Wil son. MIKE Observations have a standard deviation of 4.9%, consistent with what has been previously obtained at Keck and Lick (Wright et al. 2005). Bottom: Chromospherically active star, HD 219764, S values show a 0.07 dex standard deviation (7%) in contrast to Tau Ceti, 0.008 dex...... 22
2.6 Distribution of the chromospheric activity parameter log R~K for the target stars of the Magellan Planet Search Program. The bulk of stars are inactive, with a peak at values below -4.9. The inactive tail includes a bulk of stars that could evolving off the main sequence as demonstrated by Wright et al. (2004b) and Jenkins et al. (2008). 23
3.1 Observed template spectra of HD 1326 (black dots) superimposed to the deconvolved solutions using the old Jansson routine (in blue) and the alternative deconvolution methods (in red). From top to bottom: Boosted Maximum Likelihood, Damped Richardson Lucy, Maximum Entropy, and B-Spline deconvolution...... 37 3.2 RMS for stable star HD 1326 obtained using the Jansson deconvolu- tion (left) and-Boosted -Maximum-Likelihood (right-)...... 39 3.3 RMS for stable star GJ 699 obtained using the Jansson deconvolution (left) and Boosted Maximum Likelihood (right)...... 39
Vl 3.4 RJ\!IS for stable star GJ 908 obtainecl using the Jansson cleconvolution (left) ancl Boostecl :Maximum Likelihoocl (right)...... 40
3.5 RJ\!IS for stable star GJ 412A obtainecl using the Jansson cleconvolu- tion (left) ancl Boostecl J\!Iaximum Likelihoocl (right)...... 40
3.6 RMS for stable star 95735 obtainecl using the Jansson cleconvolution (left) ancl Boostecl Maximum Likelihoocl ( right)...... 42
3.7 Observecl template spectra (black clots) superimposecl to the clecon volvecl solutions using the olcl Jansson routine (in blue) ancl the acloptecl Boostecl Maximurn likelihoocl (in red) of the five lllost stable M clwarfs of the planet search program. From top to bottom: GJ 908, GJ 699, HD 1326, HD 95735 ancl GJ 412A...... 43
4.1 J\!Iean SPFS values comparecl with Sl\nKE values for a set of 67 calibra tion stars. The solid line represents the hest lea,st-squa,re linear fit, while the clashecl line represents a slope of unity...... 50
4.2 Four PFS stable stars with spectra types ranging from miel G to early Tvl...... 52
4.3 Detection perioclograms of the 3 cancliclate planets plus long periocl trencl cletectecl in the RV measurements of GJ 667C. The signals are listecl from top to bottom in orcler of cletection. Each perioclogram is compnt.ed on t.he resiclua,ls t.o t.he previons fully Kepleri 4.4 Phase-folclecl RV measurements of the 4 signals cliscussecl in the text. The red clots represent the new 143 HARPS measurements. The 21 PFS measurements are shmvn in clark blue ancl the green clots corre sponcl to the 20 HIRES observations. Each preferrecl Keplerian moclel is shown as a solicl black line. 61 4.5 Top view of the orbits for the three inner cancliclates. The classical habitable zone of GJ 667C is shown in blue...... 62 vii 4.6 Left. Periodograms of the three activity indicators discussed in the text. Both the S-index and the FWHM show a significant signal around 105 days. On the right, we show each activity indicator folded to the period with greatest significance : 105 days for the S-index and the FWHM, and 1.2008 days for the BIS. The shape of the"' 105 days signal is apparent to the eye in both the S-index and FWHM. . . . . 63 4. 7 Periodograms of the two candidate planets detected in the RV mea surements of GJ 221. The signals are listed from top to bottom in order of detection. 64 4.8 Keplerian solution for GJ221. Top panel: phased Keplerian fit for the 125 day component. Bottom panel: phased Keplerian fit for the 3.87 day component. Red dots represent HARPS measurements, while blue dots represent PFS measurements...... 66 4.9 Plot of orbital elements of all known exoplanets (black dots) and the new discovered planets presented in this work (red dots)...... 69 Vlll Resumen Gracias a los lllÚltiples ava11ces en el área de búsqueda de exoplanetas, y búsquedas utili:1.m1do el método Doppler en particular, los límites estdísticos del censo planetario galád ico para planetas gigantes con semi-amplitudes de velocidades radiales de J( > 10 m s-1 y períodos P < 10 años han sido aclarados. A pesar de esto, las preguntas más importantes, y la razón por la cual seguimos buscando planetas extrasolares siguen aún si11 ser contestadas: ¿Cómo se fonnan los planetas? ¿Existen 1nás planetas rocosos habitables, como la Tierra? ¿Cuán comunes son? Con las esperanza ele arrojar luz dentro del problema, presento los resultados ele 3 años ele esfuerzo para encontrar planetas rocosos alrededor ele enanas i\II, el tipo ele estrella más abundante dentro ele 10 pe. Además, tienen Zonas Habitables cerradas (0.1 - 0.2 UA), lo que conespomle al espacio de pará1netros donde las búsquedas con Doppler son más sensibles, haciéndolas blancos fáciles para espedrgrafos de alta 1 precisión de última generación ( <2 m s- ). El primer paso hacia la detección de planetas del tamaño ele la Tierra fue medir el valor esperado ele la inestabilidad ele las velocidades radiales causadas por la variabil idad intrínseca de las estrellas que imitaría la seíía.l planetaria. Se midieron valores S y log R;m a partir de '"'"'8000 espectros acumulados desde el 2004. Usando estos valores se derivaron edades, periodos de rotación e inestabilidades de velocidades esperadas para'"'"' 670 estrellas F,G,K y IVI de la secuencia principal. En segundo lugar, y particularmente con el objetivo de alcanzar la precisión requerida para detectar exoplanetas de baja ~nasa, se desanolló una mejora en el algoritmo de deconvolución usado para derivar velocidades radiales usando espectros estelares de Keck. Se probaron cuatro nuevas ruLinas de deconvolución en cinco de las estrellas más estables de la muestra. El algoritmo de I'vláxima Verosimilitud Impulsado delllostró ser la mejor opción, produciendo velocidades radiales que eran casi iguales a las producidas pm la técnica de J aussou eu los pemes casos, y rnejmó la precisión en 1 m s-1 en los mejores casos. I'vlás aún, este algoritmo mejoró las incertidumbres internas en todos los casos. :Mejor precisión es un paso más adelante hacia encontrar planetas rocosos. Fimdlllelite, el Prognuna de Búsqueda de Planetas de iVlagallanes es descrito desde sus cmnieuzos, cuaudo operaba cou el el:lpectrógrafo J\!IIKE, y ahora, usaudo el nuevo Planet Fincler Spectrograph (PFS). Se presentan resultados de ambas: 12 lX gigantes gaseosos de largo periodo, así como 2 super-Tierras alrededor de GJ667C, una en el medio de la zona habitable, y una super-Tierra cercana con un Neptuno de largo periodo alrededor de GJ221. Todos estos planetas descubiertos recientemente están ayudando a explicar las propiedades estadísticas de los cxoplanetas y sistemas multi-planetarios, lo cual a su vez, nos ayudará a contestar las preguntas presentadas en el primer párrafo. X ABSTRACT Thanks to the many advances in the field of exoplanet surveys, and in Doppler velocity surveys in particular, the statistical outlines of the galactic planetary census 1 for giant planets with radial velocity half-amplitudes J( > 10 m s- and periods P < 10 yr lmvc bccn darificd. In spitc of this, thc most important qncstions, ancl the the reason we are looking for extrasolar planets are still unanswered: how are planets formed? are there more rocky habitable planets, like the Earth? ancl how common are they? Vlith hope of shedding some light into this problem, I present the results of 3 ycars of cff"ort to fiud rocky plaucts arouud M dwarfs, thc most abuudaut typc of stars within 10 pe. In addition, they have close-in Habitable Zones (0.1 - 0.2 AU), the parameter space where Doppler velocity surveys are most sensitive, making them 1 attainable targets to state-of-the-art high precision (<2m s- ) spectrometers. The first step towards the detection of earth-sized planets was to estimate the ra dial velocity jitter from stellar activity expected on our target stars, which could mimic a planet's signature. S -values and log R;,K were measured from rv8000 archiva! MIKE spectra. Using the these values, ages, rotation periods ancl expected velocity jitter were derived for"' 670 main sequence F,G,K and M stars. Second, ancl with the particular goal of achieving the required precision to detect low-mass exoplanets, an upgrade on the deconvolution algorithm used to derive radial velocities was performed using Keck spectra. Four new deconvolution routines wcrc tcstcd on fivc of thc most stablc stars in thc sarnplc. Thc Boostcd lVIaximmn Likelihood algorithm proved to be the best option, yielding RVs that were nearly equal to the J ansson technique at worst and improved the precision by 1 m s- 1 m the best cases. Moreover, this algorithm improved the interna! uncertainties in all cases. Better precisiou moves oue step closer towanls fiudiug rocky planets. Finally, the J\!Iagellan Planet Search Progra.m is described from its early stage, when it operated on the 1J!IKE spectrograph, and the current stage, using the new Planet Finder Spectrograph (PFS). Results from both are presented: 12 long-period gas giants detected with MIKE, as well as two super-Earths around GJ 667C, one of the in middle of the habitable zone, ancl a close-in super-Earth ancl long-period Neptune around GJ 221. All of these newly discovered planets are helping to clarify the statistical prop- xi erties of exoplanets and multiple-planet systems, which in turn will help us answer the questions presented in the first paragraph. ----~~ ~------ Xll Chapter 1 Introduction Before the discovery of the first extrasolar planet in 1995, most theoretical models were based on our own Solar System, and thus, predicted planetary systems with gas giants orbiting beyond 5 AU and rocky planets in closer-in orbits, all in co-planar, and almost circular orbits due to eccentricity damping in protoplanetary disk During the past fifteen years, more than 700 extra solar planets have been un covered around late F, G, K, and lVI stars within 100 parsecs, so we know that these low mass companions are a common consequence of star formation. The current planetary sample covers a wide variety of masses, orbital periods and eccentricites (Butler et al. 2006, U dry & Santos 2007). Many ongoing efforts to improve Doppler precision push this census clown into the regime involving radial ve locity half-amplitudes K < < 10m s-1 and periods P < 1 yr. In this regime, a spate of recent detections from various groups reveal a quite unexpected rich new category of exoplanets that are midway between Earth-mass rocky planets and Jupiter-mass or Saturn-mass gas giants. These include the so-called "super-Earths" (2 MEarth :::; lVI :::; 10 lVIEarth ), potentially rocky planets with likely water-rich deep-ocean and/or deep-atmosphere mantles, and the somewhat larger water-dominated "ice-giants" with masses between that of Uranus all the way up to gas giants like Saturn. These have orbital periods of days to weeks (Rivera et al. 2005; Udry et al. 2006; Mayor et al. 2009; Vogt et al. 2009; Rivera et al. 2010). ------~ln-the-regime-Gf-per-iGds-12-~l-0-yeai's,----these-efferts-ha-ve--a-lso-leacl-to-t-he--fir-st detections of solar system analogs (Jorres et al. 2010; Marcy et al. 2002). E ven though Doppler searches have clarified the statistical outlines of the galactic 1 planetary census for giant planets with radial velocity semi-amplitudes J( > 10 m s-1 ancl periods P < 10 yr (see e.g. Ma.rcy et al. 2005; Cumming et al. 2008), the ultimate questiolls ill tlte fielcl are yet to be mtswerecl; ltow are pl;wets fonnecl? are there more rocky habitable planets, like the Earth? ancl how common are they? 1.1 Doppler velocity technique J\!Iany methocls are being usecl to cletect planets outside our Solar System. Most of these take aclvantage on the fact that an exoplanet can cause a noticeable effect on its host star such as blocking part of its light, which is callee! the transit methocl (Henry et al. 2000; Charbonneau et al. 2000), a.ltering its position or the astrometry meLhod (BenedicL eL al. 200:2; McArLlmr et al. :2004: Bean et. al. 2007; RefTert. & Quirrenbach 2011), altering the periocl of cyclic pulses CWolszczan & Frail, 1992) or of other transits (Holman et al. 2010; Cochran et al. 2011) callee! pulsar timing ancl transit timiug respectively. There is a.lso a. method that relies 011 the effect a plauet has on the radial velocity of its parent star because of its gra.vitational pull, which makes the star-planet system orbit arouncl their center of mass. If a time series of high resolution spectra of the star is observecl, ancl one looks at the absorptic t line Doppler shifts of the series, a perioclic variation shoulcl be founcl when a planet is present. This is callee! the Doppler velocity technique (Mayor & Queloz 1995; Dutler et al. 2006; Uclry et al. 2007). The orbit of a binary system can be clefinecl by seven orbital elements: the orbital periocl P, the inclination of the orbital plane with respect to the plane of the sky (plane of reference) i, tlte positiüll a.llgle n of tlte lille of llüdes (tlte illterseccioll between the orbital plane ancl the plane of the sky) which is measurecl from North to East from a reference point, w the angle between the ascencling nocle (the nocle where the star crosses the tangent plane while rececling from the observer), the semi-major axis of the orbit a ancl its eccentricity e, a.ncl finally the time of periastron passage T. These are showu iu figure 1.1. Now, the radial velocity curve of the primary star in a spectroscopic binary can be expressecl in terms of the velocity amplitucle K 1 ancl the position angle from the periastron, or true anomaly v (shmvn in figure 1.1) 11 =; + K 1 [cos(u + w) +e cos(w)], (1.1) 2 - Figure 1.1: Orbital elements 3 where the time clepenclency líes in v. Kepler's Equation states E - e sm. E = -27r (L - T) (1.2) p where E corresponcls to the eccentric anomaly. Using the relation (see Murray & Dermott 1999) 1 tan -v = J + e tan E (1.3) 2 1- e 2 it is possible to determine P, T, e ancl w. i, on the other hancl, cannot be cleterminecl using spectroscopic observations alone, but can be clerivecl using photometric elata from the eclipse, in case of a binary system, or transit, in case of a planet. The semimajor axis of the primary is relatecl to the amplitucle K 1 by p a sini = -~K1 (1.4) 1 27r Accorcling to Kepler's Thircl Law, where a = a 1 + a 2 as the semi-major axis of the relative orbit of the system (1.5) Using equations lA ancl 1.5 together with the clefinition of the center of mass m 1a 1 = 1n2a 2 results in the mass function of the system: (1.6) Assuming that the seconclary is a planetary mass companion, then m 2 < < m 1 a.ll(l thc cquatiou can be simplificcl to 1/3 2 3 2 1 2 1n sin i = (___!__. ) J( 1n l (1 - e ) 1 (1.7) )J 21rC ' * then, basecl on RV elata, the mínimum mass 7711' sin i of a planet can be known assuming a mass for the host star. 4 1.2 Planets around M dwarfs Over the past 3 decades, Doppler velocity surveys have been progressing in precision from 300m s-1 in 1980 to 1 m s-1 in 2010, resulting in major breakthroughs, such as the first confirmed extrasolar planet (Mayor & Queloz 1995), the first multiple planet system (Butler et al. 1999), the first transiting planet (Henry et al. 2000; Charbonneau et al. 2000), the first neptune-mass planet (Butler et al. 2004), the first terrestrial mass planets (Rivera et al. 2005; :VIayor et al. 2009), thc first system with two super-Earths ( Queloz et al 2009), and thc first rocky planets in the middle of the Habitable Zone of their host star (Anglada-Escudé et al. 2012). The most recent discoveries have made it clear that with the current precision available, it is possible to detect Earth-mass planets around the least massive stars: M-dwarfs, given that these stars have the greatest reflex motion due to an orbit ing planet since the mass contrast between the star and the planet is significantly reduced. Moreover, these stars have the desirable feature that their habitable zones, the region where an Earth analog planet can maintain liquid water on its surface, are typically at 0.1 to 0.2 AU, corresponding to orbital periods of ""20-50 days, making them targets to which current Doppler surveys are most sensitive. In addition to these favorable characteristics, M dwarfs are the most abundant types of stars in the solar neighborhood, and so there are at least 239 M dwarfs (Henry et al. 2006) within 10 pe with masses in the range of 0.08 to 0.5 M8 , making them a natural choice for extrasolar planet studies, astrobiology, and next generation interferometry and direct imaging missions. It is specially interesting for astrometric searches, where closeness is the primary virtue. In spite of these facts spectroscopic extrasolar planet surveys have concentrated on G dwarfs, in hopes of finding solar system analogs, until recently when they have started to focus more on M dwarfs. 1.3 M dwarfs and Planet formation -~-~From~th-e-theoretical pointof view, recent studíes-ar-eínvestig-attngthe- connectíon between planet formation processes and the host stellar mass, as this parameter would affect the physical conditions that dominate the development of its planets, 5 such as the stellar gravity, which changes disk rotation speecl; the position of the ice line controllecl by the stellar temperature, ancl fina1ly, the clisk's mass which scales linearly with the stellar mass. According to the "core accretion" scenario (Laughlin et al. 2004; Ida & Lin 2005; and Kennecly & Kenyon 2008), in which planetesimals are assemblecl into planetary cores that then accr·ete gas, giant planets could not be as easily formed arouncl J\!I dwarfs as they do arouncl G-dwarfs, due to the lmv surface dust density of in the disk whid1 woulcl prcvcut thc cmergcncc of massivc corcs befare suifering gas dcpletiou . In turn, close-in Neptune-mass ice giants shoulcl be more common around these low-mass stars. vVithin the same scenario, but assuming that the distribution of initial disk pa rameters do not depend on stellar mass, Kornet et al. (2006) preclicts that the number of stars with giant planets increases \vith decreasing stellar mass. In the other hand, Boss (2006) cliscusses the formation of giant plants aromK low mass stars accorcling to the disk instability paracligm, -vvhich enables the formation of gas giant protoplanets befare gas clissipation of the disk. This paraclignt oft"crs a way to form giant planets around rvr dwarfs. The cliscovery of a statistically significan!, nnmber of planets arouncl M dwarfs will help to unveil where, when ancl how these proposecl mechanisms are most important in the formation process of exoplanets. 1.4 Habitability and 7JEarth The last two questions presented at the beginning of this chapter regarding the frequency of habitable planets, ancl in particular, of Earth-like planets has been the main goals of current exoplanetary science. It is known that earth-mass planets exist, but it is still not clear what fraction of stars form such planets, ancl how many of these planets have survived possible clynamical interactions after they were formed. J\!Ioreover, what is the fraction of these terrestrial mass planets that can be consiclerecl as habitable, also, known as 1]Eart.h · The com:ept of lmbitability can be clcfinecl i11 countless ways. T11 exoplanetary scicucc, thc most general cldiuitiou of a habitable pla.11ct is a plauct whosc orbit lies in the Habitable Zone of its host star. The Habitable Zone of a star has been 6 defined as the ranges of distances from the stars where a lie planet could lie and have liquid water on its surface. The inner and outer boundaries of this range of circumstellar distances depend both on the star's properties and any assumptions rnade with respect to the planet's atrnosphere and its response to the stellar flux. Kasting et al. (1993) derived these boundaries for the Sun as well as other stars in the Zero Age Main Sequence, based on the assumption that a planet would have an atmospheric composition similar to the Earth's. This study puts the M-dwarfs' habitable orbits very close to the star, as seen in Figure 1.2. This characteristic is the most desirable when searching for habitable planets, as it falls right within the reach of today's most precise instruments. 1.0 Sun 1 .. ·· 1 .· ¡.-' ¡ .. .-1, ,: J 1 1 1 1 ¡O) J.S IZJ /..Q ¡¡:;/:tJ 1 1 1 1 1 1 0.1 1.0 10.0 orbital distance (AU) Figure 1.2: The orbital region that remains continuously habitable during at least 5 Gyr as a function of the stellar mass. Dark gray and light gray areas show the Habitable Zone defined using different criteria (see Selsis et al. 2007 for a more detailed description). The dotted boundaries correspond to the extreme theoretical limits, found with a 100% cloud cover. The dashed line indicates the distance at which a 1 MEarth planet on a circular orbit becomes tidally locked in less than 1 Gyr. ·---- Figm:e-from.Selsis-et-al.~200'l~-~.------ 7 1.5 Thesis outline Thc: main focns of this dissc:rtation is to show that it is possiblc: to find c:xtrasohr rocky planets arouncl nearby early JVI clwarfs. In orcler to attain the necessary preci sion, first a chromospheric activity stucly hacl to be performecl in orcler to select the most chromospherically stable stars that show the lowest intrinsic jitter. Seconcl, an upgracle on the existing radial velocity recluction cocle hacl to be performecl with the aim of improving the existing precision using archival elata. Chapter 2 is a transcript of the peer reviewecl material on chromospheric activities entitlecl "Chromospheric activity of Southern Stars from the Magellan Planet Search Program (Arriagacla, 2011). This stucly consists of new measurements of two activity K ) inclicators (S-values ancl log R; 1 of a sample of ~670 stars from ~8000 archival high-resolution echelle spectra from the Magellan Planet Search Program. Ages, rotation periocls ancl expectecl intrinsic stellar jitters clerivecl from the measurecl S values are also incluclecl in this chapter. Chapter 3 presents the work clone on Keck M dwarf spectra in orcler to improve the precision already obtainecl by updating the template cleconvolution technique. The importance of deconvolution in the radial velocity recluction cocle is shown. Then, five alternative cleconvolution techniques are clescribecl, along the implemen tation of each algorithm, testing proceclures ancl the results. Results from the Niagellan Planet Search Program are clescribed in Chapter 4. Thc first part is bascd in thc publicatious madc by thc tcmn uutil 2010 (Lópcz• :Morales et al. 2009, Niinni ti et al. 201 O ancl Arriagada et al. 201 O), ancl prc;;;ents the planets discoverecl using the NIIKE spectrometer, which were crucial cluring the clevelopment of the N ew Planet Search, presentecl also in this chapter. Then, the Planet Finder Spectrograph is introclucecl, together with the observations and elata recluction technique, activity measurements ancl the current precision acquirecl with the new spectrometer. The last part of this chapter is based on the publication of the first terrestrial-mass extrasolar planet in the miclclle of the habitable zone of an M-clwarf (Anglacla.-Escuclé et al. 2012) a.nd a two-planet system ·with a super-earth ancla long-periocl giant planet around an J\!I-clwarf (Arriagacla et al. in prep.). 8 Chapter 2 Chromospheric Activity of Southern Stars 2.1 Introduction The Magellan Planet Search Program monitored radial velocities of 690 G, F and K main sequence stars from December 2002 to July 2009 at high-spectral resolu tion with the main goal of finding extra solar planets. "J'vleasurements that are done ------with a precision of a small fraction of the line width ( of at least 1/1,000) m "Y be affected by any physical process happening in the stellar surface that would change the lines profiles. Precision Dopplcr surveys rcly on finding a repeating pattern on the radial velocities, and thus, one must take into account major sources of error that may mimic a planet's signature, or example, starspots rotating into view and periodically changing the line centro id, and thus, the measured velocity ( Queloz et al. 2001; Henry, Donahue, & Baliunas 2002; Santos et al. 2003; Paulson et al. 2004). Photospheric features such as spots and other and rotationally modulated inhomo geneities on the stellar surface, as well as magnetic cycles can be associated with the presence of active regions, inducing radial velocity variabilities otherwise called "photospheric jitter". Activity measurements are, therefore, essential for selecting the most inactive, stable stars for a planet search survey. Indirectly, · these magnetic activity levels and photospheric phenomena can be studied from chromospheric activity indicators, determined from the strength of emission in the Ca II H & K line cores of a star's spectrum. This indicator gives 9 an estímate of the stellar jitter and the rotation period, both critica] values for understanding ancl interpreting the noise present in radial velocity measure 1ents (Noyes et al. 1984; Saar & Fischer 2000; Santos et al. 2000). Since 1966, the Niount \i\Tilson program has been monitoring Ca II H & K emission of more than 1200 Nor!.hern-Hemisphere clwarfs Fmcl gian!.s, clefining Lhe j'vloun(. vVilson S-valne (S,.¡w ) (Duncan et al. 1991). Dueto the long term monitoring of their stars, the Smv has been used by other programs as the standard metric of photospheric activity. Since the S,nv index includes both, the chromospheric ancl photospheric information, the K MW I-IK team clefinecl thelog R; 1 index (N oyes et al. 1984) where photospheric eiiecLs are removed by normalizing Lhe chromospheric emission Lo Lhe bolomeLric K luminosity of the star. The log R; 1 is the value used to estimate the expected RV jitter asociated with the stellar activity (Saar & Donahue 1997; Santos et al. 2003; Wright 2005; Isaacson & Fischer 2010; Lovis et al. 2011; Comes da Silva et al. 2012). In t he Son( hern-I-Iemisphere, Lwo large programs have macle Lhe eiiorLs (.o provide measurements ancl analysis of photospheric activity of solar-type stars. Henry et al. (1996) observed more than 800 stars at CTIO using the Cassegrain Spectrograph at the 1.5m telescope. J\;Iore recently, Gray et al (2006) has observed more than 1600 dwarf and giants as part of the Nstars project. Although they provide us with a useful testbed to the conclusions drmvn in the north, ancl also an independent samrle for statistical analysis, both were done using low-resolution spectra ancl none provicle the long-tenn coverage as Mmmt vVilson's. Other efForts il1 tl1e Southem Hemisphere were made by Tinney et al. (2002) ancl Jenkins et al. (2006), who performed a long term analysis of rv200 stars as part of the Anglo-Australian Planet Search (AAPS) using high-resolution spectra, as well as chromospheric measurements of 353 bright stars by J enkins et al. ( 2008) . 2.2 Observations and data reduction The Magellan Planet Search Program has been monitoring ""'690 stars since late 2002. The observations have been done using the MIKE echelle spectrograph (Dernstein et al. 2003) mounted on the 6.5-m Clay Telescope (Magellan II) at Las Campanas Observatory. Using a 0.35 arc-sec slit, MIKE obtains spectra with a resolution of R ""' 70000 in the blue ancl rv50000 in the red, covering the wavelength range from 3900-6200 A divided into a red anda blue CCD. The Iodine spectrum (5000- 6200 10 Á) falls on the red CCD while the blue CCD captures the Ca II H and K lines to monitor stellar activity. The detectors on MIKE are Lincoln Labs 2048x4096 15,um CCDs , the blue one has a C'a TT HK slit-to-detector efficiency of '""30 per cent at 3800 A. The targets presented here are part of the Magellan Planet Search Program stars, which are 690 F7-M5 dwarfs and subgiants, that were observed in seeing that ranged 1 from 0.5 to 1.5 arcsec. The dispersion at the Ca II H & K lines is 0.02 A pixel- . Exposure times ranged from 90 seconds on brightest objects to 600 on the fainter ones, giving a SNR per pixel that would fall between 20 and 100 at the Ca II H & K lines. Extraction from raw CCD images was carried out using a modified version of the pipeline used to extract spectra from the red CCD images used then to calculate velocities. Besides the standard reduction, this pipeline also measures scattered light from inter order pixels and subtracts it. No sky subtraction is done as the sky brightness around the Ca II H & K lines (and throughout the whole echellogram) is negligible compared to the brightness of our sources. A cosmic ray removal is also performed on the two dimensional echellogram, as with all MIKE data. Wavelength calibration was done using ThAr spectra acquired each night to focus the instrument. Since the blaze function removal is not part of the standard pipeline, and no fiux standards are observed during the Planet Search prograrn, a different approach had to be applied. In order to correct for the blaze function we fit a 7th order polynomial to a quartz lamp spectrurn, and divided it out of all spectra in the region of interest. During the first nine observing runs quartz larnp spectra were not properly taken and a slightly modified reduction code was used in order to extract the spectra. Wavelength calibration was done using ThAr spectra that was acquired each ·· · night-to-focus-the-instrument-and-to-set--the grating-in-place:--All-stars-were-then cross-correlated with a binned NSO solar spectrum, to correct for the the barycentric velocity of each star. 11 2. 3 Analysis 2.3.1 Derivation of S indices from MIKE/Magellan observa tions Duncan et al. (1991) defines the S-index as the ratio between the total counts H and J( bandpasses centered on the H and K lines, ancl the total counts in the R and V bandpasses centered in the continuum. The MvV project calculate<-: this index individually each night using the HKP-2 spectrometer, which is a specialized multichannel spectrometer. The H ancl J( hanclpasses, have triangular profiles wiLh full wiclth at half maximums of 1.09 A and are centered at the very cores of the H ancl K lines (3933.667 ancl 3968.470 A respectively). The other two, R ancl V, have rcctaugular profilcs with wiclths of 20 A and are centerecl in the continuum at 3901 ancl4001 A. NH + NJ( Sri'!W = et ~ ~ (2.1) j R + j V o: is a constant that was calculated to be 2.4, and that would make the mean S corresponcl to the previously measurecl mean F-value of standard stars. The F- value was the original chromospheric measurements determined at M\iV in the same way as S but using the old HKP-lspectrometer which had difFerent bandpass widths. Following the prescription of Duncan et al (1991), I have simulated the mea surement of the Mount vVilson spectrometers by clefining two triangular bandpasses with 1.09 A F\iVHM centered on the H and K lines, and two rectangular channels 20 A wicle centerecl at 3901 and 4001 A. Figure 1 shows the position of these channels in onr NIIKE spcctra. I havc snmmcd connts ·within thcsc four dfcctivc hanclprrsscs ancl taken the ratio between the sum of H ancl K ancl the sum of R ancl V. The relation used to determine MIKE S values is : NH + NJ( Sl\IIKE = (2.2) j~ R + j~ V In order to correct for systematic effects due in part to clifferences in removing the blaze function as noted in Section 2.2, and to upgrades of the MIKE blue CCD chip, which presented linearity problems between lVIay 2004 ancl September 2005, we performed linear calibrations between the two earlier epochs ancl the latter one, 12 shown in Figure 2.2, in the same way that MIKE S values are then calibrated into the Mount Wilson system as detailed in the next section. Thus, using equation 2.2 we produced a set of SwnKE that were then calibrated into the Mount Wilson System. Individual S values will be available online. 2.3.2 Converting from SMIKE to Mount Wilson SMw A conversion had to be applied in order to calibrate MIKE S values into the Mount Wilson standard frame. Tinney et al. ( 2002), Jenkins et al. ( 2006) and J enkins et al. (2008) have demonstrated that to bring S values derived from high resolution spectra using the Duncan et al. (1991) prescription into the MW system only requires a linear calibration. To find this rclation, a statistically significant number of stars with s~1W values must be used. Only two stars previously observed at MW were observed during the program (HD 10700 and HD 119217), so a different approach had to be taken in order to perform the calibration. Stars which have already been part of low and high resolution chromospheric studies and that were calibrated into the MW system were selected. 195 stars had already been observed by Gray et al. (2006)(G06), 117 by Henry (1996) (H96), 115 by Jenkins et al. (2008)(J08) and 7 by Wright et al. (2004)(W04). Figure 2.3 shows the median SMrKE value against the reported values from G06, H96, JOS and W04. A tight correlation is seen between SMrKE and all four studies, although the scatter increases when comparing against values obtained from low resolution spectra (G06, H96) and where the number of points is low (W04). Based on this result, we decided to use only the stars from J08, giving a linear least squares fit that has a sk·Je of 1.049±0.05 and a zero point offset of 0.012±0.004. The derived calibrated S values are presented in Table 1 of the Appendix, including both stars from Duncan et al. (1991). The values derived in this study agree within uncertainties. 2.3.3 Uncertainties and random errors Since chromospherically-based quantities depend on intrinsic variability, which fiuc -- - --tuates-on-alHimescales-such-as-that-of"-activity-cycles-or-rotational-periods,-our-final log R~K estimations represent median values of the points we obtained during the period they were observed, but not true averages for the stars. 13 j\!Ieasurement errors are causecl by the quality of the spectra as well as the quality of the calibration into the M\"l values. The 0.017 scatter shown in Figure 2.3 bet\veen the JOS elata set is in part clue to the stella.r va.riability, since MIKE elata are not contemporaneous with the .JOS data and many stars are in different parts of their activity cycles. Measurements for sta.rs observecl more frequently ancl for the full cluration of the Planet Search program will have correspondingly lower uncertainties. Previous works (Vhight et al. 2004, Jenkins et al. 2006) have usecl measureu1ents of the stable star T Ceti (HD 10700) as a proxy for the ranclom errors introducecl in the recluction proceclure (scatterecl light removal, blaze function correction, cosmic ray removal ancl barycentric velocity correction). Consiclering that, first, T Ceti has been shown to be extremely stable (Baliunas et al. 1995a, \"lright et al. 2004, Lovis et al. 2011), ancl seconcl, because it is an excellent somce with which to search for any systematic errors in om precision velocities, this star has been observecl continuously throughout the program; T Ceti serves as an excellent cliagnostic star. 127 useful observations of T Ceti have been acquirecl, as shown in Figure 5, with a standard cleviation of 4.9%, which is similar to that of the Keck ancl Lid: errors from \"lright et al. (2004). 2.3.4 log R~K The S inclex inclucles both photospheric ancl chromospheric information. However, for an activity analysis, we are only interestecl in the chromospheric component. The photospheric contribution, which clepencls on the stellar temperatme (measmecl as (B- 11)), is removed following the methoclology by Noyes et al. (19S4) in or._.er to generate the log R;,K which appea.rs in Table 1 of the Appenclix. This transformation has been well calibratecl for 0.44 < (B- 11) < 0.9, which means that for stars reclcler than 0.9, the calibration has greater uncertainties. K Figure 2.4 shmvs log R; 1 values from this stucly plottecl against values from JOS. The line denotes a 1:1 relationship, ancl visual inspection reveals a goocl match between the two sets. Since calibration sta.rs were only observecl once or twice by JOS, we expectecl a high clispersion in the elata clue to propagatecl errors, such as the intrinsic variability of the stars, scatter in JOS values ancl scatter in my own values, which were cliscussecl in the previous section. In orcler to test if there is goocl agreement l1d.ween the samplcs, first, I compare my valnes with .TOS nsing 14 (2.3) to look for any systematic differences. These values are plotted in Figure 2.4, where we see no clear trend. The mean value for O" is -0.00013. A two-sample Kolmogorov Smirnov test was also performed to check for any difference between both data set distributions, obtaining a statistical estimator D=0.124 and a corresponding probability estimator of P =0.3, rneaning that my sarnple is not significantly different frorn J08. 2.3.5 Rotation periods and Ages Using the results from the activity analysis, one can also compute rotation periods and ages using empirically derived log R~K relationships. From Mamajek & Hille brand (2008) Ro= { (0.233 ± 0.015)- (0.689 ± 0.063)(1og R~K + 4.23) if log R~K > -4.3 (0.808 ± 0.014) - (2.966 ± 0.098) (log R~K + 4.52) if -5.0 < log R~K < -4.3 (2.4) where Ro=log(Prot/T) and T is the convective turnover time as adopted frorn Noyes et al. (1984), 2 3 1.362- 0.166x + 0.025x - 5.323x if X> 0 logr ~ { (2.5) 1.362- 0.14x if X< 0 where x = 1 - (E- V) and the ratio of mixing length to scale height is 1.9. Noyes et al. (1984) note that this relationship has a rms about the mean curve of"' 0.08, which rneans that predicted rotational periods using equations 2.4 and 2.5 rnay be fairly accurate. To estimate ages for individual stars, we use the age-chromospheric activity re lation of Mamajek & Hillebrand (2008): logt = -38.053 -17.912logR~K -1.6675(1ogR~K? (2.6) where t is the stellar age in years. Marnajek & Hillebrand (2008) report an rms of rv0.07 dex in logtjyr for stars with -5.1< logR~K < -4.3. I have extrapolated this 15 relation to lower log R; 1K in arder to get an estímate of the ages, therefore for less active stars the estimated age will not be as accurate. I should also note at this point that R; 1K corresponds to the activity level averaged over many activity cycles. For this reason, one should obtain as many observations as possible in orcler to derive a value closer to the actual age of the star using this relation, which is not always the case for the program stars. I present a histogram of the distribution of the median activity levels for the lVIag ellan target stars in Figure 2.6. The bimoclal distribution of activity in log R;,K notecl previously by Gray et al. (2003), Jenkins et al. (2006, 2008) is eviclent, although less pronounced. This slight discrepancy might be explained by the clepende11-::y of bimodality with metallicity ([rdjH]), as shown by Gray et al. (2006). Another work, by Hall et al. (2007), shows a similar distribution to our sample. 2.3.6 Radial velocity jitter Since intrinsic stellar variations can lead to false planet cletections, when it shows periodic signals, or to non detections, when jitter is larger than planetary signature; it is essential to quantify this source in orcler to take it into account when searching for planets. Saar, Butler & Marcy (1998), Santos et al. (2000) derived the first empirical relations cletected RV dispersion, or jitter ( vVright (2005) usecl ~ 450 stars from the California and Carnegie Planet Search to derive empiricalmodels to preclict a star's radial velocity jitter basecl on its (B- V) color, activity level ancl absolute magnitucle. Since log R; 1K is not well calibratecl in the B- V < 0.4 or B- V > 0.9, an alternative metric of stellar activity, F(c. ri) is used instead. Upclated relations to estímate the baseline expectecljitter as a function of cxccss activity 6.S ancl B-V colors were recently published by Isaacson & Fischer (2010) using more than 2600 stars from the California Planet Search. Using Hipparcos (Penyman et al. 1997) colors ancl parallaxes, I have applied both vVright's empirical model and Isaacson's relations to my calculated activity S values to predict the expectecl radial velocity jitter of the lVIagellan program stars. These results are given in the last two columns in Table 1 of the Appenclix. 16 2.4 Conclusions Chromospheric activity log R~K and S-values from over 9,000 archival spectra from the Magellan Planet Search Program, taken over a baseline of 6 years, have been measured. The spectra were taken using the l\IIIKE spectrograph at Las Campanas Observatory. Analysis of the measured level of activity of T Ceti yields a random error of ""'5% and the S-values and log R~K indexes presented here correspond to the median activity levels of each star during the time span of the observations. As stellar activity has proven to mimic an extrasolar planet signal, an expected value of the RV jitter due to the star's intrinsic variability is an essential tool when it comes to selecting targets, as it can constrain the mínimum RV amplitude variation detectable due to a true planetary companion. I have computed the expected stellar RV jitter for all the stars in the sample, with the main goal of selecting the most stable targets for the new Magellan Planet Search, which is making use of the new Planet Finder Spectrograph (Crane et al. 2006; Crane et al. 2008). Advantages of this new spectrograph include higher throughput, higher resolution, active and pas sive tempcrature st.abilization, fixed format, and all optics optimized for the Iodinc region ( 5000 to 6200 Angstroms). Data collected in the first five months of scientific operation indicate that velocity precision better than 1m s-1 RMS is being achieved (Crane et al. 2010). 17 3890 3900 3910 3920 3930 3940 Wavelenglh Á 3930 3940 3950 3960 3970 3980 Wavelenglh Á 3980 3990 4000 4010 4020 4030 Wavel.en.!blb_ Á_ Figure 2.1: 11, K, H and R channels in a representative MIKE spectrum. The rdative flnx is in arhitrary nnits. \iVavelengths ha.ve heen shifted to 7,ero vdocity in arder to make the measurements. 18 2.0 1.0 1 .¡ 11 1 0.8 1.5 1 1 Lñ" iO o o o o 0.6 /· N N c. c. al al (/) 1.0 (/) (¡j (¡j 1 0.0 0.0 ·/ 1 l 0.0 0.5 1.0 1.5 2.0 0.0 0.2 0.4 0.6 0.8 1.0 S(between May 2004 and Sept 2005) S(before M ay 2004) Figure 2.2: Comparison between SMIKE measuremcnts taken in differcnt epochs. In the left panel, filled circles correspond to measured S values using dat.a taken between May 2004 and Sept. 2005 vs. S measured using data taken after Sept. 2005. The solid line corresponds toa linear regression, yielding a slope of 0.995 andan intercept of -0.002. In the right panel, filled circles correspond to measured S values using data taken befare May 2004 vs. measured using data taken after Sept. 2005. The solid line corresponds to a linear regression, yielding a slope of 0.987 and an intercept of -0.011. 19 1 .5 1 .O Jenkins et o1.(2008) rms=0.017 0.5 Wright et ol (20041 rms=0.019 0.0 ~~~~~~~~~~~~~~~~ 0.0 0.1 0.2 0.3 0.4 0.5 Figure 2.3: Comparison between SllnKE measurements with previous measurements that are alreacly calibratecl to the s),!W system. Solicl lines corresponcl to linear regressions for each of the clatasets, while the clashecl lines show relations of slope unity. For clarity, the Jenkins et al (2008), Henry et al. (1996) ancl Gray et al. (2006) daLaseLs have an o1Ise(, of 0.4, 0.8 and 1.2 respecLively 20 Figure 2.4: Upper panel : Comparison between log R~K índices obtained from MIKE spectra in this work and those obtained from FEROS spectra by Jenkins et al. (2008). The line denotes a l:l relationship. Lower panel : the level of difference between both samples, or cr values vs. our log R~K indicies. The line denotes O level of difference between the samples. 21 0.18 • • • • • • .. • • • 1 • • 0.17' • • • • • • • • • • • • 1 • • • C/) • • • • • 1 0.16 r- •• • 1: • 1 a: • •• • L • • • • • 0.15 r- • 1 1• • 0.14 2003 2004 2005 2006 2007 2008 Year 1.3 r- - - 1.2 • • • • C/) 1.1 • - • 1.0 • - 0.9 2005 2006 2007 2008 2009 Year Figure 2.5: Top: T Ceti S-values from IVIagellanjj\JIIKE to illustrate my measnrement errors, in contrast with the 1% scatter obtained at JVIount vVilson. MIKE Observa tions have a standard deviation of 4.9%, consistent with what has been previously obtained at Keck ancl Lick (Wright et al. 2005). Bottom: Chromospherically active star, HD 219764, S-values shmv a 0.07 clex standard cleviation (7%) in contrast to Tau Ceti, 0.008 dex. 22 80 60 z 40 20 -5.5 -4.5 -4.0 Figure 2.6: Distribution of the chromospheric activity parameter log R~K for the target stars of the Magellan Planet Search Program. The bulk of stars are inactive, with a peak at values below -4.9. The inactive tail includes a bulk of stars that could evolving off the main sequence as demonstrated by Wright et al. (2004b) and Jenkins et al. (2008). 23 Chapter 3 A new deconvolution routine for M-dwarfs 3.1 Introduction Doppler spedroscopy is uniquely able Lo find l.errest.rial nutss planees and Solar Sys tem analogs around the nearest 2,000 stars within 50 pe. Only Doppler programs 1 that can achieve rv 1 m s- precision will continue to fin el terrestrial mass planets in this dcca.dc, a.ud thc first Sa.tum aua.logs iu thc 2020s. Thcsc will be thc prima.ry targets of interferometric astrometry (VLTI, Keck-I, SIIVI), clirect imaging, and exo planet spectroscopy over the next century. The goal of this work was to improve the precision cunently obtainecl with the Keck/HIRES system of > 3m s-1 on nearby late K- early IVI clwarfs. In orcler to reach 3m s-1 precision on :M-clwarfs, I polishecl the current velocity recluction code (Butler et al. 1996) used to derive radial velocities for the California-Carne'Sie Exoplanet Survey, where the majar improvements that one could make lie in the deconvolution of the template spectra. The next section describes thE: Radial Velocity observations taken with Keck/HIRES as part of the Lick-Carnegie Exoplanet Survey. Section 3.3 describes the role of cle convolution in the radial velocity analysis. Later, in section 3.4 I describe the various tested deconvolution algorithms ancl the method usecl to test each one. In section 24 3.5 results and shown. Finally, section 3.6 is a summary of the main results and conclusions. 3.2 Keck/HIRES observations The Lick-Carnegie Planet Search has been obtaining radial velocity observations of a large sample of about 1000 of the nearest stars for the past 15 years. These were acquired using the HIRES (Vogt et al. 1999) spectrometer installed in the Keck I telescope. In order to measure Doppler shifts with this telescope/instrument configuration, a sealed and temperature controlled (50 ± 0.1 a C at Keck) iodine absorption cell is placed just ahead of the spectrometer slit in the converging f/15 beam from the telescope. Inside the sealed cell the column density of the iodine inside remains constant, hence the absorption cell superimposes a rich forest of iodine lines on the stellar spectrum, providing a wavelength scale against which doppler changes are measured and proxy for the point-spread function (PSF) of the spectrometer (Butler et al. 1996). For the Keck planet search program, the HIRES spectrometer operates at a spectral resolving power R "' 70,000 and wavelength range of 3700 - 8000 A, though only the region 5000-6200 A (with iodine lines) is used in the current Doppler analysis. Doppler shifts from the spectra are determined with the spectral synthesis technique described by Butler et al. (1996) and seen in more detail in the next section. Sin ce August of 2004, the focal plane of HIRES has been upgraded from a Tek2048 CCD, which had various issues such as a non-fiat focal plane, nonlinearity of charge transfer efficiency (CTE), charge diffusion in the silicon substrate, overly large pixels, etc., to a three-chip CCD mosaic of fiatter and more modern MIT-Lincoln Labs CCDs. As a consequence, larger uncertainties on the 2004 pre-August velocities ( "pre-fix") are evident. 25 3.3 The role of Deconvolution in Doppler Analysis In order to carry out a Doppler analysis using iodine as a wavelength scale and instru mental PSF proxy a complete modeling of the spectroscopic observations should be made. This modcl should consider various instrumental efl'ects such as t.he spectro graph wavelength zero point, changes in dispersion and changes in the instrumental PSF. As explained in Butler et al. (1996), the model of the observed spectrum is the result of the convolution of the product of the intrinsic stellar spectrum, ! 5 , and the iodine transmission spectrum T12 with the instrumental PSF. The result is binned to the wavelength range of the used CCD pixels. In the case of Keck/HIRES spectra, this is the 5000 - 6200 A region, which consists of 15 orclers, each spread over 4021 pixels. Observations are modelecl as (3.1) where ,6,,\ is the Doppler shift ancl k is the normalization factor. * represents the convolution. The iodine region is dividecl into '""700 chunks of 2 A. Each chunk produces an independent measure of the wavelength scale, PSF, ancl Doppler shift. Thc finalmeasnred velocity is the weighted mean of the velocities of thc individual chunks. The two input functions, T12 ancl Is are obtained in different wa.ys. The iodine transmission spectra, T12 is measurecl clirectly at the National Institute of Stanrlards and Technology (NIST) using the Fourier Transform Spectrograph (FTS). The reso 6 lution of this spectrum is ,\f,6,,\ =10 , which makes the Iocline spectrum fully resolved and oversamplecl comparecl to astronomica.l spectrographs. A cliffcrent. procedure nmst he ta.ken in order to ol1tain the intrinsic stcllar spec trum, ! 5 , since a FTS can only reach 1st magnitucle stars ancl what is needecl is a high resolution, high S/N spectrum for the moclel. Observations taken through the echelle without the iocline cell still have the instrumental smearing from the PSF, so in this case, the answer to this problem is to cleconvolve the observecl stellar spectrum with the known instrumental PSF. The instrumental PSF is estima.tecl by observing a ra.piclly rotating B star through the iocline cell. Since B stars are virtually featureless, they serve as incanclescent lamps illuminating the telescope ancl instrument in the same wa.y of any other stellar 26 observation, and providing the absorption spectrum of the iodine. Using this obser vation, the PSF is extracted by comparing the B-star Iodine observation versus the reference FTS !2 atlas. This process is described in more detail by Valenti et al. (1995). After successfully deriving the PSF, a deconvolution of the stellar spectrum is ready to be performed. 3.4 Deconvolution algorithms Butler et al. (1996) uses the modified Jansson technique (Gilliland et al. 1992) to perform the template spectra deconvolution. This is the routine used in the Keck/HIRES velocity reduction code as well as various other instruments (MIKE, UCLES, UVES and PFS). The advantage of this routine is that it's a simple iterative method. Although the routine performs well in F, G and early K dwarfs, it fails to do so in late K and early M dwarfs. The reason for this is that the routine makes use of the National Solar Observatory (NSO) solar spectrum to define the continuum which is a primary constant. This approach succeeds for F, G and early K dwarfs because the spectra can be compared with the sun's without any problems. In the case of late K and early M dwarfs, this comparison is no longer suited since the spectra does not have a continuum due to the TiO molecular bands which domínate the spectrum. In this section I will discuss the problem of deconvolution and all the alternative algorithms to the Jansson technique that were studied. I will discuss the testing method used as well as the current results in precision using the chosen algorithm. 3.4.1 The problern of deconvolution Deconvolution is the process by which a convolution is reversed, in the same sense that division reverses multiplication. From the practical point of view, it is an algorithm-based way to reverse a physical convolution such as the signal distortion effect of an electrical filter or of the finite resolution of a spectrometer. ·· Let 's consider a signal, in this case, an intrinsic stellar spectrum, f (x) that has been distorted by the instrumental profile of thc spectrograph (PSF), r(x) it has been measured with. What is observed, y(x), can be described mathematically by a 27 linear integral equation : y(x) = ./ r(u)f(x- u)cht (3.2) in the presence of noise n(x), which is always the case in the experimental worlcl, it becomes y(x) I: r(u)f(.T- u)du + n(x) (3.3) (r * f)(x) + n(x) (3.4) Looking at the problem in fourier spa.ce, this equation can be re-written as fj(v) = P(v)/(v) + i?.(v) (3.5) Here is where one of the main problems m·ises: any noise aclclecl to the signal after the broadening or low-pass filter operator has been appliecl \Vill be greatly amplifiecl when the Fourier transform of the signal is cliviclecl by the Fourier transform of the broaclening operator, because the high frequency components of the broaclening operator (the clenominator in the clivision of the Fourier transforms) are typically very small, resulting in a great amplification of high frequency noise in the resulting cleconvolved signal. l\iloreover, trying to determine f(x) knowing y(x) ancl r(x) is recognizecl to be an ill-posed prohlem. Dne Lo Lhe smoo!.hing effec!.s of in!.egnüion, small pert nrhaLions seen in y could corresponcl to large perturbations in f, hence there is no unique or stable solution. Our only hope is to recover the "best" estimate of the original so urce. 3.4.2 J ansson technique Thc J a.ussou mcthocl is a moclificcl vcrsim1 of thc Van Cittcr rcstora.tiou which 1s given by .rn+l = ./"' + a:(y _ 1. * .f') (3.6) where cv is a convergence parameter, generally taken as 1, ancl fn is the n-th estímate 28 of the object. This algorithm converges quickly but generally diverges in the presence of noise. Janson introduced a modification in order to deal with this problem by considering constraints on the solution. If we wish that A :S .fk :S B, the iteration becomes 1 r+ (x) = r(x) + s(x)[y- r * rJ(x) (3.7) with s(x) = C[l- 2(B- A)-1 J_r(x)- T 1 (A + B)J] (3.8) where e is a constant. 3.4.3 The Bayesian approach Consider an observed spectrum y and true intrinsic stellar signal .f. The goal is to reconstruct f from y. According to Bayes' Theorem, p(.fJy) = p(.f)p;~~;)' (3.9) where p(y) and p(.f) are the probabilities of the experimental data and of the real data over all other possible image outcomes, respectively. p(yJ.f) and p(.fJy) are the probabilities of observing y given an intrinsic spectrum f and off being an intrinsic spectrum given that y has been measured. The solution given by the Bayes' method is the one that maximizes the right part of equation 3.9. Maximum likelihood The Bayesian solution is the one that maximizes the p(.fJy) probability distrit. 1tion over the real data f. Assuming gaussian noise, the probability p(yJ.f) then becomes, (3.10) The maximum likelihood solution assumes that p(.f) is a constant. Then, maxi mizing p(.f Jy) is equivalent to minimizing 29 .. IIY-1'*./'W .7(./)= 22 (3.11) a N Now some restriction must be appliecl in orcler to obtain a regularizecl solu tion. The Lanclweber methocl (Lanclweber 1951), otherwise callecl the Jacobi methocl (Bertero & Boccacci 1998) achieve this by minimizing equation 3.11 using an iterative algorithm: r+l = r + (y- In * r) * 1· (3.12) An aclvantage of this methocl is that it can be easily regularizecl by constraining the number of iterations, by demancling positivity of the cleconvolvecl spectrum, by constraining a spatial clomain, etc. Maximum Entropy This theory formulates a solution from our knowleclge of nothing but the positivity of f by cleriving the probability basecl on its entropy. Consiclering the entropy function H of the solution f, the probability can be written as p(f) = exp[-o:H(f)] (3.13) leacling to the iteration In= exp(m.n * r) ( L"' y ) (3.14) La: exp(m." * r) where "'(l'n*1') )] 1nn+ 1 = 1T/.n + [(Y ~"' . 1 - In * T (3.15) ¿"'Y Richardson-L ucy Now, assuming Poisson noise insteacl of Gaussian noise, the probability p(IIY) be comes . [(7' * I)(x)]y(a:)e-(nf(;¡:)) p(yl.f) = (3.16) II;¡· y (X·)' . , 30 where the constraint is given by the iteration r+l = r ( ;n * r) (3.17) which is the Richardson-Lucy algorithm (Richardson 1972; Lucy 1974; Shepp & Vardi 1982), also sometimes called the expectation maximization (EM) method (Dempster, Laird, & Rubin 1977). If the unknown noise contribution is characterized by a deviation, (}", then the difference between the signal being approximated, f, and the approximation r cannot be improved further than (}". Any reduction in this difference woulc:i turn into an exaggeration of the noise and we would be entering the domain of noise enhancement. A noise damping mechanism has been proposed (R.L. White 1994) in order to control these noise enhancement effecLs. This damping can be achieved by replacing the y j ¡n factor with a simple corrected version, for instan ce (3.18) where r is sorne positive number. The idea here is to suppress those iterative cor rections that have a frequency higher than that of the frequency at which the power spectrum of the true signal is lost below the power spectrum of the noise. Since this is where more of the power of the noise is, almost no enhancement of noise occurs. 3.4.4 The least squares method using E-splines The method of performing deconvolution of signals using B-splines has been explored by Dierckx (1984). T will briefty summarize the basic idea in this section. Let's consider equation 3.2. As a first objective, the routine tries to determine a spline function of degree l, s(x) which when convolved with a spline approximation uf the instrumental PSF, R(x), it will fit Lhe measured data, y(x), closely enough. Second, as regularization, it will be required for s(x) to be smooth in the sense that the discontinuities on its lth derivative should be as small as possible. Mathematically speaking, Dierckx (1984) introduces measures of both, closeness of the fit and--smootlmess of s(x). As the first criterion he uses the sum of the squares residuals. Considering a set of data values (xq, yq), with weights Wq, q = 1,2, ... ,m(xq < Xq+l), 31 m 2 6(c) = I:wq(Yq- g(xq)) (3.19) q=l where e are the spline coefficieuts aud g(x) is the convolution of s(x) ancl R(x). As the latter criterion, the lack of smoothness can be measured by (3.20) with if j > q - l - 1 or j > q, (3.21) if q - l - 1 :::; .i :::; q and (3.22) where t.i, j = 1, 2, ... , n are the knots of the spline function s(x) of degree l. The prime denotes the derivative with respect to t. In summary, the deconvolution criterion is formulated by minimizing r¡(c) con stra.ined by 6(c) :::; S, where S is a non-negative user-supplied constant to control the degree of smoothing, therefore called the "smoothing factor". 3.405 Comparison ofvarious deconvolution routines on Keck/HIRES template spectra Deconvolution tests of template Keck spectra. using the previously described decon 1 volution algorithms \Vere carried out on 5 of the most stable (RiviS< 5 m s- ) M-dwarfs (B - V :::; 1.2) of the Lick-Carnegie Extrasolar Planet Search. All of the algorithms were implemented as IDL functions and the itera.tions were turned into IDL routines to be used by the precision velocity code. The Maximum Entropy (J\;IE) and Maximum Likelihood (ML) deconvolution rou- tines were based on the deconvolution functionw max_entropy o pro and max_likelihood o pro written by Fra.nk Varosi (NASA/GSFC, 1992). In my implementation of the decon- 32 volution routine I rnodified the max_entropy. pro function so that the convolution in equation 3.13 was perforrned by the routine num_convol. pro, which utilizes a nurnerical convolution instead of a convolution with fast fourier transforrns. The Darnped Richardson Lucy (DRL) algorithrn was irnplernented as described in the section 3.4.3. I started frorn Varosi's Maximum Likelihood function, but changed the iteration to the one given by equation 3.17 using the factor given by the relation 3.18. The E-splines routine made use of Paul Dierckx's fortran code DECDSP. This routine needs a spline approximation of the PSF, the data points with their weights, the degree of the spline approxirnation of the solution, s( x), and the srnoothing factor S as inputs. I cornputed the spline approxirnation of the PSF with the IDL routine IMSLBSLSQ. pro. The degree of spline approximation was set to 3 for all of the test stars frorn trial and error. The tricky part was setting the srnoothing factor, as it was not the sarne for all stars. Moreover, it looked like the sarne number was not going to work for all chunks, as sorne values of S gave beautiful results in sorne parts of the spectrurn, but eventually rnade the code crash. I tried setting different S values depending on the chunk's SNR and the surn of the derivatives at each data point ( w hich should account for the number and deepness of lines), but the code kept crashing. Finally, I cornprornised and chose a single S per star that would (a) not rnake the code crash and (b) showed a reasonable solution in rnost of the spectrurn. Deconvolution with boosting M. Morhác et al. (2009) found that although positive definite deconvolution rneth ods irnprove substantially the resolution of the spectra, these solutions reach a stable state where it is necessary to stop the nurnber of iterations, and this is not efficient enough to decornpose closely positioned peaks. They propose a non-linear "boost ing" operation for these deconvolution algorithrns to irnprove their effectiveness. The boosted algorithrn consists of perforrning an iterative deconvolution with a de termined number of iterations, let's say, L. After cornpleting the Lth iteration, the boosting operation (3.23) rnust be applied to the deconvolved spectrurn jL, where p > 1 is the boosting co- 33 efficient. Then, the deconvolution and posterior boosting operations are repeated a requirecl number of times R. This number is set by trial ancl error. I appliecl boosting to two of the cleconvolution routines, IVIL ancl l\IIE, at the very early stages of this work. \IVhile Boostecl IVIL showecl noticeable improvement, Boostecl IVIE clicl not improve the results as clearly. Later, when testing the DRL rou tine, I also triecl boosting, but improvements were not eviclent either, thus I cleciclecl to keep the boosüng operation only on the IVIL routine, hereafter Boostecl Maximum Likelihoocl (BlVIL). A visual inspection of ME ancl DRL spectra shows more ringing than the BfliiL. vVorst results using these two methocls coulcl be due to a noise en hancement effect of these ringing features proclucecl by the boosting algorithm. As describecl in the previous section, each chunk is cleconvolvecl inclepenclently. Befare performing the deconvolution, the spectrum is oversampled by a factor of 4, using a spline interpolation. Changing the orcler between interpolating and decon volving did not change the resulting velocities, so I decidecl to keep it as the original code. In order to set a maximum number of iterations for the three iterative meth ocls, I first staxtcd by dcfiuiug; a x2 which would accouut for thc cliffcrcncc bctwccn the observed spectrum ancl convolution of the deconvolved solution with the PSF. This method dicl not prove to be effective as the threshold value of the :x:2 varied greatly and arbitrarily between stars ancl dicl not give better results than the Jansson technique. After ( hü;; firf>L aLLempL, I tried seLLing a defined nnmber of itera( ions ancl Les( ed it in one of the test stars ( GJ 908). I started with 50 i terations ancl increased the number by 10 until I obtained a precision close, or better if possible, to that given by the Jansson deconvolution method. Then, I tried the routine in the remaining 4 test stars using the same number of iterations. I noticecl that some of them required more iterations, while other, less, in orcler to obtain equal or better precision than the old routine. As one would expect, this was relatecl to the signa! to noise ratio (SNR) of the observed spectrum, since higher SNR spectra required more iterations than lower SNR spectra, although the number of iterations dicl not increase linearly with SNR. 34 The final number of iterations was set by trial and error and was different for each of the deconvolution algorithms. I defined different SNR intervals and assigned each an increasing number of iterations. When no more improvement was seen,, then the number of iterations for cach interval was fixed. In order to decide on which algorithm to use in all of the program stars, radial velocities (RVs) for each of the test stars were derived using the best deconvolved template spectrum obtained from every deconvolution routine described above. The outcoming scatter of all RVs as well as the internal uncertainties, measured as mean of all of the observations' standard deviation, defined as the scatter of RVs resulting from each chunk, were compared to the Jansson scatter and internal uncertainties. Both of these uncertainties refl.cct the quality of the performance of the deconvolution routines, as the internal uncertainties demonstrate how stable the routine is within the same spectrum (for example between chunks with less or more continuum), while the standard deviation of all RVs demonstrate long term stability. Once a better result was found, I applied this routine to a.ll the M dwarfs in the Program. 3.5 Results To exemplify the results from all the deconvolution routines Figure 3.1 shows de convolved spectra of two "chunks" from different echelle orders of the star HD 1326 using each of the deconvolution algorithms. A visual inspection is necessary to detect noise enhancing behavior such as ringing (left panel, blue lines and red lines in the second and third from top to bottom), underfitting (left panel, bottom), or peculiar shapes in the spectral lines. The uncertainties from the resulting velocities obtained using every deconvolution routine in our 5 test stars are summarized in Table 3.1. For each star the numbers in the top row correspond to the uncertainties obtained considering all data points as individual points. The second row shows the uncertainties obtained by binning all - ---observations on 2 hr timescales:- Finally, in the third row, the internal·uncertainties, refl.ecting systematic errors from characterizing and detcrmining thc PSF, detector imperfections, optical aberrations, effects of undersampling the iodine lines, etc. are 35 presentecl. These results show that from all testecl cleconvolution routines, the smallest un certa.inties are obtainecl by both BML ancl DRL, which show similar results. Looking a.t the binnecl velocities, Bl\IIL gives a better result as Jansson in 3 out the 5 stars, and a similar result in the other two. DRL gives a better result in 3 out of the five, but worse in the other two. Between BML ancl DRL, better internal uncertainties are obtainecl using BJVIL. A visual inspection of wicle lines in the cleconvolvecl spec tra (Figure 3.1, for instance) reveals more ringing in DRL templates than in .JML templates. Finally, the same spectra.l line from a.ll 5 test J\II-clwarfs are plottecl in Figure 3. 7 along with the Bl\IIL ancl Jansson templa tes. Deeper lines, less ringing ancl the conservation of stellar flux are achieved with RML cleconvolution. In Figures 3.2 to 3.5 are plottecl the velocities acquired using B:tviL (on the right) ancl Jansson (on the left). 3.6 Conclusions Accorcling to the results shown in the previous section, 1 cleciclecl to test the Bl\IIL cleconvolution routine on all the JVI-clwarfs of the Keck-Carnegie Planet Search Pro 1 gram. The most stable (RiVIS< 6 m s- ) are presentecl in table 3.2. Using BIVIL cleconvolution, velocity clispersions have clecreasecl by a median of 15%. Internal un certaillties, llave i11tproved by 20%. H COllsideri11g o11ly post-fix observatio11s, velocity clispersions have improvecl by 10%, ancl internally by 20%. The main goal of this project was to improve the precision obtained with the HIRES/Keck system, which is currently attaining 2 m s-1 on F,G ancl K clwarfs, but <3 m s-1 on l\II-clwarfs. One of the main sources of high velocity clispersion comes from the quality of the stellar templa.te spectrnm, which is the result of a cleconvolution of a high-resolution spectrum of the target star with the instrumental point spreacl function. Template spectra of 5 stable stars from the Program were clerivecl using four clif- 36 o 20 40 60 80 o 20 40 60 80 pixel #-3730 pixel #-690 Echelle Order= 1O Echelle Order=3 o 20 40 60 80 o 20 40 60 80 pixel #-3730 pixel #-690 Echelle Order= 1 O Echelle Order=3 2.0x 105 r-~~~-;:-~~__:;_::_::_..~_:;._,~~~~ o 20 40 60 80 o 20 40 60 80 pixel #-3730 pixel #-690 o 20 40 60 80 o 20 40 60 80 pixel #-3730 pixel #-690 Figure 3 ]· Observed template spectra of HD 1326 (black dot.s) snperimposed to the deconvolved solutions using the old Jansson routine (in blue) and the alternative deconvolution methods (in red). Ero m top to bottom: Boosted Maximum Likelihood, Damped Richardson Lucy, Maximum Entropy, and B-Spline deconvolution. 37 Table 3.1. Resulting uncertainties using each of the cleconvolution algorithms for each star: Jansson, Boostecl Iviaximum Likelihoocl (BJ\!IL), Damped Richardson-Lucy (DRL), :iVIaximum Entropy (ME) andE-Splines deconvolution (Decosp). The three rows for each star corresponcl to the RJ\!IS of all velocities ( top), RMS of velocities averagecl in 2-hom bins (miel elle) ancl internal errors (bottom). These are given in m s-1 Star Jansson BlVIL DRL rdE Decosp GJ 699 3.32 3.20 3.20 4.17 4.80 3.19 3.17 3.13 4.18 4.84 1.75 1.50 1.56 1.76 2.31 GJ 908 2.87 2.87 3.13 3.79 4.65 2.79 2.80 3.20 3.87 4.70 1.17 1.18 1.27 1.37 4.20 HD 1326 3.76 3.33 3.29 3.29 3.70 3.65 3.28 3.19 3.20 3.70 1.11 1.01 1.04 1.04 1.16 HD 95735¡, 3.15 3.26 3.25 3.38 2.94 2.94 2.87 2.99 1.38 1.25 1.28 1.35 GJ 412A" 6.25 4.08 5.23 4.06 6.16 6.40 4.03 5.23 3.97 6.47 2.25 1.62 1.63 1.72 3.02 aThis st.ar has a "pre-fix" Lemplat.e that. clelivers 1 RJ\!J S ~ 3.6m s- . The goal of the new deconvolution is obtain close-to or better precision using "post-fix" templates. hThe Decosp deconvolution routine clelivered veloci ties tha.t were rejectecl by the raclial-velocity cocle. 38 30~~~~~~~~~~~~~~~~ 30 1 1 u ALL =3.72 ms- a ALL =3.33 ms- 1 1 a BIN =3.65 ms- a EIN =3.28 ms- 1 1 20 a INT =1.11 ms- 20 a !NT =1.01 ms- 10 10 "' m "' 5 ilJ ilJ $ " ilJ $ 5 ~ ill$' 1 ilJ 1 ilJ ~ o $ ilJ ilJ ,4¡¡,$%~'" mnr o $ ~ ill$ ~i. $ rn ~ Jll lilJ ...,,., q¡p & m 'lbrn ~,~, ~ m ·a ill~ $ ~ rn ill ilJ ·1!!'·¡~¡ rn m ·¡¡ ilJ o ilJ o w -10 w -10 :> :> -20 -20 -30 -30 1000 2000 3000 4000 5000 6000 1000 2000 3000 4000 5000 6000 Julian Date - 2450000 Julian Date - 2450000 Figure 3.2: RMS for stable star HD 1326 obtained using the Jansson deconvolution (left) and Boosted Maximum Likelihood (right). 30 1 30 1 a ALL =3.30 ms- a AU. =3.20 ms- 1 1 a BIN =3.19 ms- a BIN =3.17 ms- 20 criNT =1.75 ms-1 20 aiNT =1.51 ms-1 10 [fJ 10 "' .. ~ 5 ~ . 5 o $ $ $~ , ,., o $ f 1~, ~ ..., ·a f ~~~ ~ ~ $ $ ·a o q ~ $ H :> -20 -20 -30 -30 1000 2000 3000 4000 5000 1000 2000 3000 4000 5000 Julian Date - 2450000 Julian Date - 2450000 E-igllre-3-3._RMS-for-st EJble-foltar-GJ-699-ohtained-using-t-he-Jansson-deconvolut-ion-- (left) and Boosted Maximum Likelihood ( right). 39 30~~~~~~~~~~~~~~~~ 1 30 1 u ALL =2.87 ms- a .w. =2.87 ms- 1 1 a BJN =2. 79 ms- a B!N =2.80 ms- 1 1 20 aiNT =1.17 ms- 20 10 10 "' ljl "' ljJ ljJ s l~iJ ljJ IP ~ s o 1/1 IIP Ql ~ lil~ 1 ill ilJ !¡fJ m IP o iiJ ~"' .:;- (jJ 1p qm , , Ul ' '&~ .:;- ~ 11 11 q r11 1• rj1 ·¡; 1 ilJ ·¡; Ir o o Gl -10 Gl -10 :> :> -20 -20 -30 -30 1000 2000 3000 4000 5000 6000 1000 2000 3000 4000 5000 6000 Julian Dale - 2450000 Julian Dale - 2450000 Figme 3.4: RMS for stable star GJ 908 obtainecl using the Jansson cleconvolütion (left) ancl Boostecl Maximum Likelihoocl ( right). 30 1 30 1 cr AIL =6.25 ms- a ALL =4.08 ms- 1 1 u 8 m =6.40 ms- a l!IN =4.03 ms- 1 a mr 1.62 ms- 1 20 a 1':?.' =2.25 ms- 20 = 111 111 ~ "' 10 ~~ 10 . ljl ljJ 1~ 1 "' J¡ 1~ 11 ¡\¡ rL ih s ~ 111 ~] ~ ' ' ' s ~ 1 1 1' qll o 1~ Ir i' i\1 o d+ r!J 1 '~w~~~~·~ ~P.~w~P .:;- Ir 'ill 'P .,., ~ rp qi 1" 'Pd P '"' ¡(¡ '~~ ~¡¡¡¡~~~ ·¡; H 111 i' ""() 'Ir •. ~"q~ ~ ~ C¡~h~~¡~1] t;~ 'P o o ljl 'P 1 i ~ ql .¡- Gl -10 t 11 \1 '• Gl -10 :> ~ ~ 1 :> ~r ljl 111 -20 -20 -30 -30 1000 2000 3000 4000 5000 6000 1000 2000 3000 4000 5000 6000 Julian Dale - 2450000 Julian Dale - 2450000 Figme 3.5: RMS for stable star GJ 412A obtainecl using the Jansson cleconvolution (left) ancl Boostecl Maximum Likelihoocl ( right). 40 1 1 Table 3.2. Velocity dispersion in m s- of stable (RMS< 6 m s- ) M-dwarfs from the California-Carnegie Planet search. The four left columns show the RMS obtained using all observations (including interna!). The four columns on the right, show the RMS obtained using only post-fix (taken with improved CCD mosaic) observations (including interna!). Star BML Int Jansson Int BML Int Jansson Int all all post-fix post-fix HD 202560 2.66 0.62 2.77 0.66 2.66 0.62 2.77 0.66 HD 42581 3.82 1.24 4.49 1.40 2.40 1.28 2.68 1.40 HD 95735 2.95 1.26 2.96 1.38 2.90 1.15 2.68 1.26 HD 97101B 2.77 1.78 3.86 2.30 2.87 1.76 3.94 2.26 HIP 114411 5.97 3.05 12.46 4.41 5.94 2.81 10.70 4.36 HIP 1532 5.86 1.50 5.89 2.07 5.86 1.50 5.89 2.07 HIP 19165 5.59 1.60 5.10 1.93 4.50 1.57 4.06 1.88 HIP 22762 5.93 2.88 7.41 3.60 5.24 2.52 6.73 3.25 HIP 24284 5.45 2.44 5.58 3.55 5.82 2.38 5.19 3.39 HIP 33241 4.84 1.86 6.54 2.10 5.29 1.73 5.54 1.99 HIP 36338 5.16 2.52 7.48 3.19 5.68 2.24 7.34 3.03 HIP 36834 5.53 2.69 7.62 3.16 5.12 2.68 6.95 2.98 HIP 37217 5.40 3.20 5.83 4.12 3.56 3.05 6.07 3.95 HIP 41689 3.95 2.05 4.87 2.36 4.31 1.68 4.15 2.33 HIP 46769 5.25 1.91 6.32 2.32 4.73 1.87 4.12 2.16 HIP 5663 3.60 1.24 3.44 1.79 3.60 1.24 3.44 1.79 41 30~~~~~~~~~~~~~~~ 30~~~~~~~~~~"~~~~3~.2~.6~m~,-~,~~ 1 a ALL =3.15 ms- 1 a mN =2.94 ms- U BIN =2.94 J. ,-l 1 U INT =1.25 ffiS-l 20 a INT ::::1.38 ms- 20 "' 10 "' 10 $ E- . [¡) E- ( •l·~·r ~ ¡¡¡~¡tJ ~~1;ifN' ,., o $ ~ '1' ·~t •1• ~if ,... o ,... ~ !11~ [J !jJ ~~t$~m .,.., ~ 'Pt +' m% m ~ ·¡¡ •ft' ~r $ 'P ~ ', ~ ~"~' •1• ¡pw ~ ·¡¡ ~$ ~IP¡~ m o o al -10 Q) -10 :> :> -20 -20 -30 -30 1000 2000 3000 4000 5000 6000 1000 2000 3000 4000 5000 6000 Julian Date - 2450000 Julian Date - 2450000 Figure 3.6: RlVIS for stable star 95735 obtainecl using the Jansson cleconvolution (left) ancl Boostecl :Maximum Likelihoocl (right). ferent cleconvolution algorithms, of which three of them were Bayesian-basecl, ancl one of them is basecl on E-splines. Radial velocities were clerivecl using each of the cle convolvecl template spectra ancl later velocity clispersions were comparecl with what it was obta.inecl using the clefault routine ba.secl on Jansson cleconvolution. Apart from obtaining better cleconvolvecl spectra, which from the visual point o[ view, had less ringing ami l1eLter mat.ching of Lhe observed st.ellar flnx, Lhe B:\11 cleconvolution algorithm resultecl in more precise velocities as well as smaller intemal uncertainties, as it was clemonstratecl by the computed velocities of the most stable 1vi-clwarfs in the program. This cleconvolution routine for :t\11-clwarfs has been a.cloptecl by the Lick-Carnegie Exopla.net team since October, 2011. 42 Echelle Order=5 Sx10s,-----~------~~~~~~------~------~------, o 20 40 60 so pixel #-61 O Echelle Order=5 8.0x1 05 20 40 60 so pixel #-610 Echelle Order=5 Sx10 5 r------~~~~~------~------, 20 40 60 so pixel #-690 Echelle Order=5 20 40 60 so pixel #-61 O Echelle Order=5 Figure 3.7: Observed template spectra (black dots) superimposed to the deconvolved solutions using the old Jansson routine (in blue) and the adopted Boosted Maximum likelihood (in red) of the five most stable M dwarfs of the planet search pro¡;ram. ~- - -~-~~From~topto~bottom:o G-J~9Q8; GJ-699; HD -1326~ HB 95735 and GJ 41-2A-. 43 Chapter 4 The Magellan Planet Search 4.1 Introduction Aftcr more than 20 ycars of cxoplanct discovcrics, starting with thc first founcl cxo planet, 51 Peg (lviayor & Queloz, 1995), this perspective of planet formation based on our own solar system has been turnecl arouncl. In particular, Doppler velocity surveys have uncoverecl more than 500 extrasolar planets arouncl late F, G, K, ancl .:vi stars withiu 100 parsccs, ma.kiug it possiblc to charactcrizc sta.tistically siguificaut trends that shecllight on planet formation processes. Although each discoverecl exoplanet is equally important for the statistics usecl to probe planet formation theories, the exciting parameter space for most D(J 1)pler surveys now includes both long-period Jupiter analogs as well as low mass planets in close-in orbits, vvhich is why Doppler surveys continue to push towarcls better precision. Currently, the most precise Doppler planet search facilities can achieve an internal long-term precision of 1-3 m s-1 (Mayor et al. 2011, Vogt et al. 2010), which is enough to cletect short-periocl Earth-mass planets around the habitable zone of J\!I clwarfs. Section 4.2 of this chapter presents the long periocl extrasolar planets cliscoverecl cluring the Olcl lVIagellan Planet Search, as well as a clescription of the survey. Section 4.3 presents a new ancl more stable instrument usecl to carry out the New Magellan Pla.net Search, the Planet Fincler spectrograph, along with the targets that are being obscrvcd, ancl thc obscrvations ancl data rcdnction proccdurc bcing nscd to find 44 planets. A brief description of the chromospheric activity indicators is shown in Section 4.5. Stability and the first results, two exoplanetary systems including a potentially habitable super-earth around GJ 221 and GJ 667C, are presentend in Sections 4.6 and 4.7. Finally, a summary is given in Section 4.8. 4.2 The Old Magellan Planet Search The Magellan Planet Search Program started in December 2002. Observations were carried out at Magellan Clay telescope, using the MIKE spectrograph (Bernstein et al. 2003), as decribed in Chapter 2. To determine the stability of the Magellan/MIKE system, a number of stable main sequence stars were monitored which were of spectral types ranging from late F to mid K. as shown in Figures 1 and 2 of Minniti et al. 2009. These figures 1 demonstrate the precision achieved by Magellan/MIKE is of 5 m s- . The in: 2rnal 1 measurement uncertainty of these observations is typically 2 to 4 m s- , suggesting the Magellan/MIKE system suffers from systematic errors at the 3 to 4 m s-1 level. The Old Magellan Planet Search monitored stars ranging from late F to the brightest early M dwarfs. Stars earlier than F7 do not contain enough Doppler 1 information to achieve precision of 5 m s- , while stars later than M5 are too faint even for a 6.5-m telescope. Stars with known stellar companions within 2 arcsec were also removed from the observing list as it is operationally difficult to get an uncontaminated spectrum of a star with a nearby companion. Otherwise there was no bias against observing multiple stars. The Old Magellan target stars also contain no bias against brown dwarf companions or host stars' metallicity. 4.2.1 New exoplanets Until2009, the Magellan Planet Search Team had reported 12 exoplanets in total. All of these planets are gas giants, in short and long period objects. A brief description of each system follows. The first twoplanets (López=J't.fmales et al. 2008) thaL we1e detected lrave masses of Msin i = 4.96MJUP and Msin i = 1.37MJUP . They first one orbits the G3 IV star HD154672 with an orbital period of 163.9 days. The second planet is orbiting the F7 45 V star HD205739 with an orbital periocl of 279.8 clays. Both planets are in eccentric orbits, with e = 0.61 ancl 0.27, respectively. Next, ·we reportecllow 111a.ss companions orbiting five more Solar-type stars (:Vfin niti et al. 2009) emerging from the velocity survey, with mínimum masses (Nisin i ) ranging from 1.2 to 25 IVI.JuP . The companions to the brightest two of these stars hacl previously been reportecl by the CORALIE survey team. Four of these companions (HD 48265-b, HD 143361-b, HD 28185-b, HD 111232-b) are Jupiter-like planets in eccentric intermecliate ancllong-periocl orbits. On the other hancl, the companion to HD 43848 appears to be a long periocl brown clwarf in a very eccentric orbit. Arriagada eL al. (2010) reporLecl five new planeLs orbi!.ing G and K dwarfs emerged from the i\!Iagellan velocity survey. These companions are jovian-mass plan ets in eccentric (e 2 0.24) intermecliate ancllong-periocl orbits. HD 86226b orbits a solar metallicity G2 clwarf. The Nisin i mass of the planet is 1.5 M.JuP , the semi majar a.,-xis is 2.6 AU, ancl the eccentricity 0.73. HD 129445b orbits a metal rich G6 clwarf. The mínimum mass of the planet is lVIsin i =1.6 NI.1up , the semi-majar axis is 2.9 AU, ancl the eccentricity 0.70. HD 164604b orbits a K2 clwarf. The Msin i is 2.7 M.JuP , semi-majar axis is 1.3 AU, ancl the eccentricity is 0.24. HD 175167b orbits a metal rich G5 star. The 1/fsin i is 7.8 M.JuP , the semi-major axis is 2.4 AU, ancl the eccentricity 0.54. HD 152079b orbits a G6 clwarf. The Msin i of the planet is 3 M.mP , the semi-major axis is 3.2 AU, ancl the eccentricity is 0.60. With time, planet searches have been pushing the planetary census into the lower mass regime. Recent cletections from RV surveys as well as the Kepler space mission have revealecl an unexpectecl rich new category of exoplanets miclway between Earth mass rocky planets ancl Neptune-mass ice giants. These so-callecl super-Earths are potentially rocky planets with water-rich oceans ancl cleep atmospheres. At least 30% of G ancl K clwarfs host Neptune or lower mass planets orbiting with periocls of 50 clays or less (Mayor et al. 2009). The fact that cloppler velocity searches are more sensitive to close-in orbits has encouragecl the research on terrestrial-mass planets arouncl M clwarfs, in particular those in the HZ, (Joshi et al. 1997; Segura et al. 2005; Boss 2006; Scalo et al. 2007; G renfell et al. 2007; Tarter et al. 2007). Discoveries of the lowest mass, lowest amplitudes planets have requirecl high 1 caclence observations with systems that achieve precision ~ 1 m s- . This was the 46 main motivation to turn from the old MIKE/Magellan system toan instrument that could achieve world-class precision. 4.3 The New Magellan Planet Search The New Magellan Planet Search is looking for planets around a sample of 500 of the nearest stars ( < 100 pe), with sorne overlap with Doppler surveys from col laborators such as the Lick-Carnegie Planet Search being carried out at Keck and the Anglo Australian Planet Search (AAPS) being carried out with the Australian Astronomical Telescope. Following the success of the Old Magellan Planet Search Program, the New Mag ellan Planet Search makes use of the Carnegie Planet Finder Spectrograph (PFS), a temperature controlled high resolution spectrograph (Crane et al. 2004). PFS was commissioned at the 6.5 meter Magellan Clay telescope at Las Campanas Obser vatory (LCO) in Chile since September 2009. The spectrograph is maintained at a constant ternperature of 25°C±O.Ol. The internal optical focus remains cm: 3tant without need for adjustment and the refractive index of the air will be stable. An Iodine absorption cell (Marcy & Butler 1992; Butler 1993) is mounted in front of _____uhe_instrurnent's_entrance_slit,irnprinting_the_reference__lQdine_spectnJ m directly___illL____ the incident starlight, providing a wavelength scale anda proxy for the spectrorneter point-spread-function (Butler et al. 1996). The Iodine cell is a ternperature con- trolled sealed pyrex tube, such that the colurnn density of Iodine remains constant indefinitely. PFS was custorn to find extrasolar planets around nearby stars using the Radial Velocity technique. 4.3.1 Targets Our target list consists of 505 objects and includes stars F-K dwarfs from the Old Magellan Planet Search. Known active stars for which our previously derived activity indicator log R~K had a value greater than -4.6 were removed frorn the search. We added 103 close (d <50 pe) M dwarfs (E- V> 1.2) frorn the Henry-Draper, Gliese and Hipparcos cattdogs . Their declina:tions mnge from 88º to =t=l9º, thurla:vin~ sorne overlapping with the KECK/HIRES systern, which is also surveying the closest M dwarfs in the northern hemisphere. Sorne of the stars are also being followed by 47 our australian collaborators at the AAT. 4.4 Observations and data reduction The observations are made using PFS with a 0.5 arc-sec slit, yielding a resolution of R ~ 80000 at the iodine region ( 5000-6300 A), covering a wavelength range of 3880 to 6680 A. Only the iodine region is used in the Doppler analysis, while the Ca II H and K lines region is used to monitor stellar activity. Total exposure times ranged from 300 seconds on brightest objects to 720 seconds on the fainter ones. This scheme ensures S N R > 300 in the iodine region per spectral resolution element. Cosmic rays are removed in the spectral extraction process of the raw reduction. A mínimum exposure time of 300 seconds per star allows proper time-averaging over the stars unresolved low-degree surface p-modes. Calibrations were aquired at the beginning and at the end of each night. These include thirty Quartz-flat fi.elcl iumges, t\vo iocliue shots usiug the O.T' aucl 0.5" slits, and two rapidily rotating B star shots near the meridian. vVe also take ThAr shots. Extraction of the spectra throughout the whole echellogram from raw CCD im ages was carried out using an IDL-based reduction pipeline which performs flat fieldillg, rellloves coslllic rays a11cl llleasures a11cl substracts scatterecl light from i11ter orcler pixels ancl subtracts it. No sky subtraction is clone as the sky brightness around the iodine region is negligible comparecl to the brightness of the sources. Doppler shifts from the spectra are cletermined with the spectral synthesis tech nique clescribecl in chapter 3 ancl cliscussecl in more cletail in Butler et al. (1996). In the case of PFS, the iodine regim1 is cliviclecl into ~ 800 chunks of 2 A each. Each chunk produces an independent measure of the wavelength scale, PSF, ancl Doppler shift. The final measurecl velocity is the weightecl mean of the velocities of the individual chunks. The interna] uncertainties presentecl for all of the PFS RVs show the intrinsic systematic errors arising from cletermining the instrumental PSF, detector imperfec lions, Lime-variable opLical aberralions, efiecLs of unclersampling Lhe iocline lines, eLe, ancl corresponcl to the standard deviation of the mean velocity ca.lculated from all the chunks. The other two ma.in sources of error are photon noise ancl stella.r jitter. In aclclition to selecting the most chromospherically quiet stars, we have chosen the observing strategy clescribecl above to mitigate stellar jitter. 48 4.5 S values 4.5.1 PFS Call H index In order to monitor the chromospheric activity of our program stars, we derived S values from the same spectra used in the velocity reduction. The method is similar to what was done with MIKE spectra, but it diflers in that only the ftux in the center of the Ca II H line was measured, because the K component in PFS spectra is in a spectral order with very low typical ftuxes, which introduces undesirable noise. To correct for the blaze function I make use of the Iodine shot closest in time to the program observation, súce iodine shots have no 12 lines in the Ca II H & K region. I smoothed it within 100 pixels, and divided it out of the spectra. The wavelength calibration was done using the wavelength solution given by the iodine lines during the velocity reduction, which was extrapolated to the Ca II H line order. All stars were then cross-correlated with a binned spectrum of the sun taken from the NSO, to correct for the barycentric velocity of each star. The S-index was then directly measured on the blaze-corrected spectra using the definitions given by Baliunas et al. (1995), but applied only to the Ca II H line. 4.5.2 Frorn SPFS to SMw As explained in chapter 2, although the PFS S index will already be very close to the Mt. Wilson scale, a simple linear calibration between the PFS and Mt. Wilson systems should be made when comparing such indexes. As with MIKE spectra, the main problem of calibrating the S scale is that a number of standard stars obseived with both instruments are needed. To solve this problem, I turned to our previously measured values for sorne of these stars and performed a least-squares linear fit (equation 4.1), as seen in Figure 4.1. SMw = 1.33Spps- 0.06 ( 4.1) 49 1 .O • • 0.8 • • /./ 0.6 w • y: ¿ • (J) • • / 0.4 / . . / • • • • 0.2 • • • 0.0 0.0 0.2 0.4 0.6 0.8 1 .O SPFS Figme 4.1: :Mean Spps values comparecl with sl\HKE values for a set of 67 calibration stars. Thc: solid linc: rc:prc:sc:nts thc: bc:st lc:ast-sqnarc: linear fit, whilc: thc: dashc:d linc: represents a slope of unity. 50 4.6 Velocity Precision In order to illustrate the precision performance of a spectrograph, repeated obser vations of a known "stable" star over a prolonged period of time must be done. Fortunately, stable stars have emerged as a byproduct of many years of execution of radial velocity searches. Since the start of the project, we have monitored a number of stable main se quence stars from late F to early M, including the well-known stable star Tau Ceti (HD 10700). Figure 4.2 shows the RVs of 3 of these stars during a period spanning more than 2 years of observations. The Magellan/PFS system achieves measure 1 ment precision of 1.5 m s- , as demonstrated by this figure. This number includes contributions from stellar jitter, unknown planets, photon statistics and systematic errors. 4. 7 Preliminary results Since the beginning of the new Planet Search Program, two exoplanetary systems have been discovered (Anglada-Escudé el al. :2012 , Arriagada et al. in prep.) using 2.5 years of observations with PFS. Below, these results are described in more detail. 4.7.1 A planetary system around the nearby M dwarf GJ 667C with one super-Earth in its habitable zone. Magellan/PFS observations of the nearby M1.5 dwarf GJ 667C were combined with a re-analysis of 4 years of HARPS sped.ra (Anglada-Escudé & Butler 2012) as well as observations made with the Keck/HIRES system, yielding 184 RV measurements, to reveal3 additional clear signals in the RVs of GJ 667C beyond the previously reported 7.2 days planet, with periods of 28 days, 75 days, and a secular trend consiBtent with the presence of a gas giant (Prv 10 years). The 28days signa!, which appears with an empirical false alarm probability of '"" 0.03%, implies a super-Earth with a mínimum mass of 4.5 MEarth orbiting solidly in the host star's classical habitable zone. The 75 days signa! is less certain, being significantly affected by aliasing interactions among a potential 91 days signa!, and the likely rotation period of the star at 105 days detected in two activity índices. If confirmcd by fnrther observations, this 51 15 r-- HIP 48336- cr=1.47 m s·' MO ~-~• o ' ' ' ~ ' - -15 1- - --.. - HD 53706- "'C/) 15 cr=1.51 m s' K0.5V E --->. ___, -- ...... o \ • tlt . () • o ' ' Q) -15 - - > : 15 - HD 10700- cr=1.68 m s·' G5V o - • , • . - t ' -15 ' - - 200 400 600 800 1000 JD-2455000 Figure 4.2: Four PFS stable stars with spectra types ranging from miel G to early M. 52 75 days period implies a third super-Earth in this system, with a mínimum mass of 5.6 MEarth , orbiting just outside, but quite near, the cold edge of the classical Habitable Zone. GJ 667C is the common proper motion companion to the GJ 667 AB binary, which is known to be metal poor compared to the Sun. Observations 21 new measurements were obtained with the Planet Finder Spectrograph (PFS) between August 2011 and October 2011. A detailed description of observations using PFS has already been described in Section 4.4. GJ 667C has 143 HARPS observations obtained between June 2004 and October 2008 which available through the ESO Public Archive. In 2009 at a conference, a planet candidate with a'"" 7 days period was announced to orbit this star . Recently, Bonfils et al. 2011 reported the detection of a plausible signa! with a period of 28 days similar to one of the candidates reported here. However, these authors present no details on their analysis of the data. Typical exposure times of these observations ranges between 900 and 1500 sec onds and the average SNR is 64 at 6100 A. A re-analysis of these spectra has been performed using HARPS-TERRA (Anglada-Escudé & Butler 2012), which obtains precision RV measurements using least-squares matching of each observed spectrum toa high signal-to-noise ratio template derived from the same observations instead of using the Cross Correlation Function (CCF) technique typically used by the HARPS team. The re-analyzed RV measurements use all the spectra between 4500 A and 6800 A, and a cubic polynomial is used to correct for the blaze variability of each order. Although a typical HARPS spectrum extends down to 3800 A, the signal-to noise ratio below 4500 A is too low to contribute significantly to the RV precision 1 on M dwarfs. The root-mean-square (RMS) of these RVs is 3.89 m s- , which is 1 significantly larger than the median photon noise ('"" 1.1 m s- ) and the typical RMS found on other stable M dwarfs (Bonfils et al. 2011). The public CCF RV measurements provided by the HARPS data reduction software are noisier ('"" 4.3 1 m s- ). As a consequence, the signals reported here would appear much less' gnif- ~~~icant~when~using4hese~G~::w~RVs·and~might~explain~why~the~new··candidates~were not announced before. GJ 667C was also observed with HIRES/Keck (Vogt et al. 1994) using the Iodine 53 cell methocl for just over a clecacle, potentially clelivering tighter constraints on long periocl signals. As clescribecl in the previous chapter, in August 2004, the HIRES CCD array was replacecl ancl elata obtainecl prior to this upgracle are of markeclly inferior qnality. Post-fix HIRES mcasnrcmcnts show similar scattcr to thc HARPS ancl PFS RVs, so only post.-fix HIH.ES elata (:20 meansurement.s) were usecl in Chis analysis. It is notecl that that the signals cliscussecl here were first cletectecl using only HARPS-TERRA measurements ancl that the contribution of the new PFS ancl HIRES elata is to irnprove the ::;ampliug cacleuce aml iucrea::;e the siguificauce of the cletection::;. Properties of GJ 667 G.J 667C (M1.5V Skiii 2010) is a common proper motion companion Lo G.J 667 AB (HD 156384). GJ 667AB is a K3V+K5V binary with an orbital periocl of 42.1 years, a semi-majar axis of"' 12.9 AU, ancl estimatecl masses of 0.73 ancl 0.69 IVI8 , respectively ( Cvetkovic & Ninkovic 2011; Tokovinin 2008). GJ 667C is currently at an angular separation of 32" which corresponcls to a. mínimum orbital clista.nce of 228 AU from the AB pa.ir. The metallicity of GJ 667 AB has been cleterminecl in prclious spectroscopic stuclies ( e.g. Perrin et al. 1988). The reportecl metallicity is [Fe/H] "' -0.60±0.1, which means that the system is metal poor relative to the Sun. Tha.nks to the association with GJ 667 AB, GJ 667C is one of the few M clwarfs with a well known metallicity. The same spectroscopic stuclies inclica.te tha.t the central pair is well within the main sequence, implying an age larger than 2 Gyr. GJ 667 has a 1 moclerate heliocentric velocity (36.9 km s- ), inclicating that it is likely a member of the Galactic disk. Using the most recent parallax measurement from HIPPARCOS on the GJ 667 AB pair (146.29±9 mas, van Leeuwen 2007), the clistance to the system is 6.8 ± 0.42 pe. Assuming this clistance ancl K bancl photometry from 2MASS (K = 6.04±0.02), we obtain a ma.ss of 0.31± 0.05 M8 for GJ 667C using the empirical relations given by Delfosse et al. (2000). Orbital analysis for GL 667C Keplerian orbital fits to the combi11ed RV data were obtained with the system·ic interface (Meschiari et al. 2009), which allows the interactive least-squares acljust ment of complex multiplanetary systems to several elata sets. To determine whether 54 there is a significant periodicity remaining in the data, a custom-made version of a least-squares periodogram (e.g., Cumming 2004) that is able to adjust for a flc .üing zero-point to each instrument (HARPS,PFS, and HIRES) has been used. To quantify the significance of a new signal, its False Alarm Probability (FAP) has been estimated empirically. 105 synthetic sets were created by randomly permutating the RV measurements over the same observing epochs (while retaining membership within each instrument). Then, the periodogram of each synthetic set is computed and the highest power found in a file is saved. A false alarm is obtained when a synthetic data set generates a peak higher than the power of the signal under inspection. The number of false alarms is then divided by the number of simulations to derive the FAP, which is used as a measure of the probability that a spurious detection arises due to an unfortunate arrangement of noise. This method is widely used to assess the likelihood of periodic signals in time series and detailed description can be found elsewhere(e.g., Cumming 2004). A new planet will be added to our solution if its FAP is lower than 1%. The 1% FAP thresholds are illustrated as dotted lines in Figure 4.3. The first detected signal is an extremely significant periodicity at 7.2 days (see Figure 4.3). N o false alarms are found in any of the 105 synthetic data sets indicating a FAP < 0.001%. The signal corresponds toa 5.2 MEarth planet (GJ 667Cb) ar- i the data favors a slightly eccentric orbit. After subtracting a Keplerian solution for GJ 667Cb, a secular trend is the next most significant signal. The magnitude of the trend ('"'"' 1.8 m s-1 yc 1) is compatible with the gravitational pull from the GJ 667 AB pair (maximum value is '"'"' 3.6 m s-1 yc1) but could also be caused by an additional unseen long-period companion. Its FAP is 0.055%, a very significant detection. A tcntative solution with a period of 7000 days provides a slight improvement to the fit due to sorne curvature detected when combining HIRES+PFS measurements (see Figure 4.4). The next signal is a planet candidate at 28.15 days, a minimum mass of '"'"' 4.5 MEarth (GJ 677Cc), and also a very low a FAP (0.034%). Even though the period is el ose to the lunar aliasing frequency ( rv 27.3 d)' the orbital phase of this candi date is still well sampled thanks to the multi-year time-span of the observations. Since tire arnplitade for this signal is stnall-totnpar-ed Lo the noise, ettelltric solations ar-e still allowed and cannot be ruled out (Shen & Turner 2008; OToole et al. 2009; Anglada-Escudét al. 2010). A Monte Carlo Markov Chain analysis (Ford 2005) 55 indicates that this eccentricity (with 98% confidence) must be less than 0.27. The classical habitable zone (HZ) of this star (Kasting et al. 1993; Kane & Gelino ?011) lies between 0.11 to 0.21 AU (see Figure 4.5). The semi-major axis of the orbit for GJ 667Cc is ~ 0.123 AU so, as long as the actual mbit is not very eccentric (i.e., e<0.2), this planet vvoulcl spencl all its time within the HZ of GJ 667C. After fitting for the 28.1 dr,ys candidate, a group of canclidate periodicities be tween 75 ancl 105 clays is founcl. vVhile the preferrecl periocl seems to be 75 clays, strong aliases are affecti11g tliis time dmnai11. For example, tlie HA RPS data alone favors a periocl of 91 clays, which is uncomfortably close to a very clear signal we also cletect on two activity inclicators (see Section 4.7.2). Even with the new PFS ancl HIRES observations, all the cancliclate mbits are still sparsely samplecl. Therefore, a precise cletermination of its orbital parameters ancl/ m its relation to the activity cycles of the star cannot yet be cleciclecl. Its FAP (0.021 %) is still very low. Therefore the signal cannot be ignorecl even in the presence of aliases m poor phase sampling. Even though a 75/91 clays mbit (0.235/0.268 AU) woulcllie slightly beyoncl the outer edge of the HZ, such a planet woulcl have similar characteristics to the super-Earth GJ 581cl (Mayor et al. 2009). Recent stuclies (Wmclsworth et al. 2011) indica te that ( clepending on its atmospheric composition) GJ 581cl could support life, so the same argnment wonld apply if this candidate is finally confirmed. For reference, we provide our best orbital fit to the 75 clays signal in Table 4.7.2, but reiterate that this candiclate be considered with clue caution. Periodic signals in the activity indicators Below, the analysis of the time series of three activity inclicators clerived from the HARPS spectra is clescribed. J\!Ieasurements of the Bisectm Span (BIS), ancl the full-wiclth-half-maximum (FvVHlVI) of the Cross Correlation Function (CCF) are ob tainecl by the HARPS-ESO pipeline for each spectrum available through the archive. The S-inclex in the Mount vVilson system(Baliunas et al. 1995) is directly measurecl on the blaze correctecl spectrum (also providecl by the ESO-HARPS pipeline) us ing the clefinitions given by Lovis et al. (2011). Briefly, the BIS is a measure of the asymmetry of the average spectral line ancl shoulcl correlate with the RV if the obscrvccl offscts are causccl by spots or pla.gcs mtatiHg with thc sta.r( Qucloz ct al. 2001). The FvVHIVI variability is thought to be a direct consequence of changes in 56 Observations This star has been observed wit PFS since January 2010 and17 new measurements have been obtained. A detailed description of observations using PFS has already been described in Section 4.4. GJ 221 also has 61 HARPS observations obtained between October 2004 and August 2008 which were available through the ESO Public Archive. As with GJ 667C, these were also re-analyzed using HARPS-TERRA. Stellar properties of GJ 221 GJ 221 has been classified as both a K7 dwarf (Reid et al. 1995) and MOV (Upgren et al. 1972) v,rith with an apparent visual magnitude of V = 9.69 and a color B - V = 1.35. The Hipparcos paralla.x (Perryman et al. 1997) gives a distance of 20.31 pe, yielding an absolute magnitude of 1\1v = 8.34 . Its metallicity has been measured to be [M/H] = -0.26 (Casagrande et al. 2008), meaning that it is of subsolar metallicity. GJ 221 has a mass of 0.767 rd 0 , which was derived following the mass-luminosity relation by Delfosse et al. (2000), using JHK-band 21viASS photometry ancl the Hipparcos parallax. According to the UV\i\1 Galactic velocities (Hawley et al. 1997), this star s.l1ould belong to the young disk, giving a lower threshold for its age of 0.1 Gyr, which is in accordance with -vve obtain using the (V-Ic)-age relation given by Gizis et al. (2002). On the other hancl, in terms of magnetic activity, the chromospheric activity indicator log(R'111<)=-4.8 suggests an age closer to 4 Gyr (i\!Iamajek et al. 2008). All stellar properties have been summarizecl in Table 4. 7.2. Orbital Analysis for GJ 221 ·we usecl the SYSTEMIC Console (lvieschiari et al. 2009) interface to derive the orbital Kcplcria.u fits to thc RV data. For cach plauct, wc list bcst- fit pcriod (P), eccentricity (e), semi-amplitude (K), mean anomaly (1110 ), longitude of pericenter ( w), mínimum mass (i\!Isill'i ) , and semi-major axis (a). The top panel of Figure 4. 7 shows the Lomb-Scargle periodogram of the combined elata set (black line) with the empirically estimated false alarm probability (FAP) for thc first ca.ucliclatc (Cuunuiug 2004; Augla.da.-Escuclé 2012). Thc bottom paucl shows the Lomb-Scargle periodogram (black line) of the residuals long with the empirically 58 estimated FAP for the second candidate. The 1% FAP threshold are illustra~ od as solid lines in these figures. The empirical FAPs are computed using a custom-made version of a least-squares periodogram (Cumming, 2004) which consists of generating 105 sets of randomly scrambled realizations of the RV data set and determining the maximum peri odogram power for each. In order to minimize the chances of detecting a false positive, only signals with an FAP < 1% were added to the solution. The first signal found has a 125-days periodicity and has an estimated FAP < 0.001%. Using the stellar parameters listed in table 4.7.2 for the mass, the best keplerian fit yields a planet with a minimum mass of 0.16 MJUP in an orbit with a small eccentricity, e= 0.17. The periodogram of the residuals (Fig. 4.7, bottom panel), shows a dominant peak at 3.87 days with a FAP of 0.007%. Using this period, the best-fit suggests an Msin i of 6.34 MEarth in a slightly eccentric orbit (e = 0.17). A Monte Carlo Markov Chain analysis (Ford, 2005) suggests that its eccentricity must be less than 0.3. Activity indicators In order to rule out the possibility that any of the found periodicities in thc RVs are caused by chromospheric activity, and since the relationship between RV jitter and acitivity is not well understood, we analyzed the S-index time series. S-indexes were derived from HARPS data using the prescription given by Lovis et al. 2011 and from PFS spec!.ra following the definition of Baliunas et al. (1995) and described in Section 4.5. We found no significant periodicity in the S-indexes around the periods given by the RV signals. 59 400 .._ 300 ~ Q_ 200 100 20 (¡j 15 3: !f 10 20 m 1s 3: !f 10 20 (¡j 15 3: !f 10 10 100 1000 Period [days] Figure 4.3: Detection periodograms of the 3 candidate planets plus long period trend detected in the RV measurements of GJ 667C. The signals are listed from top to bottom in order of detection. Each periodogram is computed on the residuals !.o Lhe previous fully Keplerian fit to the elata. Given Lhe presence of a few peaks el ose to the 1% thresholcl in the bottom panel, additional planets might be detected whenever additional RV measurements become available. 60 16.00 12.00 b e P=7.2006d 8.00 'iii ! 4.00 > 0.00 "' -4.00 -8.00 -12.00 -3.53 -2.11 -0.69 0.73 2.15 -8.30 -2.71 2.88 8.48 14.07 16.00 Trend 12.00 8.00 !"' 4.00 ~ 0.00 -4.00 -8.00 -12.00 -36.99 -22.22 -7.44 7.33 22.11 538.95 1077.90 1616.85 2155.79 2694.74 Orbital phase [days] Orbital phase [days] Figure 4.4: Phase-folded RV measurements of the 4 signals discussed in the text. The red dots represent the new 143 HARPS measurements. The 21 PFS measurements are shown in dark blue and the green dots correspond to the 20 HIRES observations. Each preferred Keplerian model is shown as a solid black line. 61 0.11 AU 0.21 AU Figure 4.5: Top view of the orbits for the three inner candidates. The classical habitable zone of GJ 667C is shown in blue. 62 20 0.7 15 S-index • • .... X Cl) 1%FAP 0.6 ([) "O :: 10 e o 0.5 J; D.. 5 0.4 20 .... 15 -5 Cl) -10 l ::o 10 (/) D.. co 5 -15 20 20 ííi" .... 15 10 l Cl) 1'-"' :: 10 o o o "'~ D.. -10 I 5 slL -20 o 10 100 1000 o 0.25 0.5 0.75 Period [days] Phase Figure 4.6: Left. Periodograms of the three activity indicators discussed in the text. Both the S-index and the FWHM show a significant signal around 105 days. On the right, we show each activity indicator folded to the period with greatest significance : 105 days for the S-index and the FWHM, and 1.2008 days for the BIS. The shape of the ,...., 105 days signal is apparent to the eye in both the S-index and FWHM. 63 • 126 days 60 o :¡:¡ ~ 40 u.1 o 10 100 1000 Period [days] 25 • • 3.87 days 10 5 o 10 100 1000 Period [days] Figure 4.7: Periodograms of the two candidate planets detected in the RV measure ments of GJ 221. The signals are listed from top to bottom in order of detection. 64 Table 4.1. Stellar Properties Parameter GJ 667C GJ 221 Sp M1.5V K7/MOV V [mag] 10.22 9.69 K [mag] 6.036 6.305 7r [mas] 146.29 49.23 [M/H] [dex] -0.59 -0.26 1 f.tR.A. [mas yr- ] 1129.7 -1.08 1 f.tDec [mas yr- ] -77.02 -346.17 1 Hel. RV [km s- ] 6.5 40.1 Age (log(R' HK)) [Gyr] >2 <4 M* [M o] 0.310 0.767 1 UVWLsR [km s- ] (19.5,29.4,-27.2) (-16,-40,-23) Teff [K] 3700 4020 L*/Lc.1 0.0137 0.0253513 4.8 Summary Both stages of the the Magellan Planet Search program has been presented in this chapter. First, the 12 exoplanets discovered using the MIKE spectrograph, and later, the first results obtained using the new Planet Finder spectrograph, which has demonstrated to achieve a long term stability of 2m s-1 or better on all late F through early M dwarfs. Also, the program's target list is described along with the observing technique and the chromospheric activity measurements. Finally, two exoplanetary systems emerging from combined PFS RVs with addi tional observations are introduced. In the case of the M dwarf GJ 667C, precision RV measurements from HARPS public spectra were obtained using a least-squares matching approach. Additional observations with PFS and HIRES confirm the de tection of these signals and further constrain the orbital parameters. Even though the public CCF Doppler measurements are not as precise, the shape of the CCF still provides useful information on stellar activity that can be used to investigate the origin of signals detected in the RV. 65 20 Pb = 125.0 days K = 8.4 -1 + e m s 0.15 ...... 10 e e = t/) -_,E >0::: o -1 o o 50 100 Orbital phase [days] 15 = P e 3.87 days 10 -1 Ke = 3.06 m s ...... ee = 0.15 t/) 5 -_,E > 0::: o t+ tt -5 -1 o o 1 2 3 Orbital phase [days] Figure 4.8: Keplerian solut.ion for G.J221. Top panel: phased Keplerian fit for the 125 day component. Bottom panel: phased Keplerian fit for the 3.87 day compo nent. Red dots represent HARPS measurements, while blue dots represent PFS measurements. 66 Table 4.2. Best Keplerian solution to the planetary system around GJ 667C. The numbers in parenthesis indicate the uncertainty in the last two significant digiLs of the parameter values. Uncertainties have been obtained using a Bayesian MCMC analysis (Ford 2005) and represent the 68% confidence levels around the preferred solution. All orbital elements are referred to JD0 = 2453158.7643. The assumed mass of GJ 667C is 0.31 M0 . Parameter b e (d?) Trend P [days] 7.20066(67) 28.155(17) 74.79(12) 7100(3000) Msin i [MJUP ] 0.01789(75) 0.0143(12) 0.0178(17) 0.25(12) Msin i [MEarth ] 5.68(23) 4.54(38) 5.65(54) 79(40) Mo [deg] 106.6(3.5) 144(25) 211(11) 231(10) e 0.172( 43) <0.27 O(fixed) O(fixed) w[deg] 344(12) 238(20) O(fixed) O(fixed) Detection FAP < 0.001% 0.034% 0.021% 0.055% 1 K [m s- ] 3.90 2.02 1.84 8.41 a [AU] 0.049 0.123 0.235 2.577 Statistics NHARPS 143 Total Nobs 184 1 1 RMSHARPS [m s- ] 1.89 RMS [m s- ] 2.0.; Npps 21 x2 310.99 1 RMSpps [m s- ] 2.37 xe 1.88 NHIRES 20 1 RMSHIRES [m s- ] 2.85 67 Table 4.3. Best Keplerian solution to the planetary system arouncl GJ 221. The numbcrs in parcnthcsis indicatc thc: unccrtainiy in thc: last two significant digits of the parameter values. Uneertainties eorresponcl to the standard cleviation of the marginalizecl IVICl\IIC samplings. The assumecl mass of GJ 221 is 0.77 IVI8 . Planet b e P [clays] 124.97 (0.16) 3.8727(0.0003) IVI* [Tvl.mp ] 0.16(0.01) 0.025(0.006) J\!Jo [cleg] 339(65) 170(51) e 0.15 (0.06) 0.15 (0.19) w [cleg] 155(63) 267(54) Deteetion F AP < 0.001% 0.007% 1 K[m s- ] 8.4 3.06 a[ A U] 0.448 0.044 vVith an orbital periocl of 28.15 clays ancla minimum mass of 4.5 J\!IEarth , GJ 667Ce is the most seeurely cleteetecl planet eancliclate with a goocl ehanee of supporting life. Basecl on statistieal extrapolations ancl transit surveys, sueh low-mass planets shoulcl be abunclant arouncl JVI clwarfs (l\!Iayor et al. 2009; Borueki et al. 2011). For the J\!I clwarf GJ 221, preeision RV measurements from HARPS speetra were also obtainecl using the HARPS-TERRA software. Combinecl with 17 aclclitiona.l PFS ohsc:rvations, wc wcrc ahlc to rctric:vc t.wo significant signals of low-mass companions. Figure 4.9 shmvs our planets in the semimajor axis-mass ancl semimajor axis eeeentrieity parameter spaees of all known extrasolar planets. All of our new cleteetecl planets lie well within the parameter spaee envelope. One of them, GJ221b lies in the miclclle of the so-eallecl "clesert" preclietecl by lela & Lin (2004). Finally, the star is slightly metal clepletecl eompareel to the Sun ([M/H]~-0.26) giving further inclieation that low metallieity (e.g. Anglacla-Eseuclé et al. 2012) cloes not inhibit the formation of planets in the super-Earth/Neptune mass regime. Using the relation given by Charbonneau et al. (2007), the reportecl eancliclates have a non-negligible proba.bility of transiting in front of the star (~ 2.7%, 1.1%, 0.6% ancl 8% for planets GJ 667C b, e, el, ancl GJ 221 b respeetively). \iVith the aclvent of the new generation optieal anel infrarecl speetrographs, many snch l\II dwarfs will l1c efficicntly snrvcycd for low mass plancts. If thc: dctcction rat.c 68 • .• .. • . • •. •• •..... • •. .• • . . . . .· ...... • : ' . . . :· .. ; \:.-s:t-~·.·- . 1000 ~-- : ·).--'.:.: •• _ ····~·.M.., . , ... : - . ·~...... '· ~·!.' • • ..... ~ ...... ,... 1.- -• ~.r:.. .,. ·: ...... - • ,..:~. ...a :·...... ~ • ,. • "' • J'• • ' • 100- ·". ••• 1 • ': ..... • •• - •\;.• ··-~· :·., ... . • • • • .•• ...... , .. .. . •• .~ .. :, ,~~- •• •,. •••• --_,¡;_ • ., '. • 10- • ..,...... - • • • . .."'...••..... • • • •. •. . . • • • 1 1- • - 0.01 0.10 1.00 10.00 a (AU) 1.0~------~------~.------~------~ • • .. "' • . •, • • • 1 ·:• • • •• fl' • >. • 'ü • •• •• • 1 • ·;::- • • ...... • • : , e • • .• . •. .• .. (]) • • • •• t .... (.) 0.4 • ...... -...... (]) . . , ...... ~ ...... ,_. ~·- .... ··! ... , ·~,~···· • • ... =-. . . t!!" .... ~. • •• .. . . :.r·~ .. ... • • • l. • .. • J ""'···. • .. ~= -"'·"' ..·: "'.,..., ..... • ' --=-· ...... t- • • • • ·--~. ·~ •• • ~.: , •• ...... 0.10 1.00 10.00 a (AU) Figure 4.9: Plot of orbital elements of all known exoplanets (black dots) and the new discovered planets presented in this work (red dots). 69 holds, very soon we may have a real chance of searching for spectral signatures of life in one of these worlcls. 70 Chapter 5 Conclusions and Future Work In this thesis I have presented the reoults of the work that have kept me focuoed and busy the last three years, motivated by the possibility of finding rocky planets around late K, early M dwarfs, further verifying their recently discovered existence. From the chromospheric study of our target stars and the update of the deconvolu tion routine used to derive radial velocities, to the first rocky planets discovered by the Magellan Planet Search team, the results discussed here contribute not only to the understanding of how exoplanets form and evolve, or how they are distributed and how many of these could potentially harbor life, but also contributes to nurture our relationship with the only planet that is, for now, known to support life, and the understanding of what our role is in the universe. 5.1 Summary and conclusions Chapter 2, presents the chromospheric activity measurements of "' 670 F, G, K and M main sequence stars in the Southern Hemisphere, from cv8000 archival high resolution echelle spectra taken at Las Campanas Observatory since 2004. These stars were targets from the Old Magellan Planet Search, and are now potential tar gets for the New Magellan Planet Search that is looking for rocky and habitable planets. Activity indexes (S-values) were derived from Ca II H & K line cores and then converted to the Mt. Wilson system. From these measurements, chromospheric (log R~K ) indexes are derived, which are then used as indicators of the level of in- 71 trinsic radial-velocity jitter expected for each of these stars. The measurements presented here not only benefit our Exoplanet Search Team, but also anyone who is interested in the stars themselves, as they provide valuable stellar age estimates and roti1tion p~riods for lmndrcds mor~ south~rn fidd stms nsing ~xp~nsiv~ Mi1g~llm1 archival data. The current goal of most radial velocity programs is to achieve 1m s-1 prec1s1on to make possible the discovery of low-mass planets and long period ( P > 25 yrs.) gas giants. Chapter 3 is clevotecl to the work macle to improve the precision obtainecl with Keck/HIRES on M clwarfs. The radial velocity cocle makes use of the Jansson cleconvolution algorithm, which clelivers velocities with precisions of > 2 m s-1 on 1 F, G ancl K clwarfs, but only > 3m s- on M chmrfs. This is th~ first tim~ tlmt i1 major upgracle on the cleconvolution algorithm of the RV cocle has been mack and it consistecl on testing four other cleconvolution algorithms found in the literature: Boosted JVIaximum Likelihoocl, Niaximum Entropy and Damped Richardson-Lucy, which are basecl on Bayesian theory, ancl a E-splines based deconvolution technique. The one that delivered the best results (i.e. lowest radial velocity uncertainties) on 5 of the program's most stable stars was appliecl to the rest of the NI clwarfs of the program. These RVs were comparecl with the ones obtainecl using the default Jans son algorithm. Besicles the RV uncertainties test, I carried out a visual inspection of the cleconvolvecl templates of the 5 test stars to compare noise enhancement ef fects (ringing) present on the resulting template from each algorithm. The Boosted Maximum Likelihoocl algorithm proved to be the best option, yielcling RVs that were nearly equal to the J ansson technique at worst and improvecl the precision by 1 m s-1 or more in the best cases. JVIoreover, this algorithm improved the internal uncertainties in all cases. In view of how successful this upgrade was on Keck elata, I am currently working on cleveloping an upgracle for PFS. Finally, I present a description of the Niagellan Planet Search Program, the new PlaneL Finder SpecLrograph ancl Lhe firsL cliscoverecl exoplanets in ChapLer 4. ¡\ L Lhe beginning, the Niagellan Planet Search was carried out using the MIKE spectrograph, 1 which clelivers a precision of"' 4 m s- . Tltese first efforts were fruiLful, aml tlte team was able to harvest 12 new long-periocl giant exoplanets. The same spectra used in the Doppler search was then usecl to carry out the chromospheric activity stucly 72 presented in Chapter 2. Once the new Planet Finder Spectrograph was installed, we began taking data of ~ 500 F-M dwarfs, of which ~100 are late-K, early-M dwarfs. The observing technique and data reduction are described in this Chapter, along with the chromospheric activity measurements. I also present 3 of our stable M-dwarfs that demonstrate the precision achieved by this instrument. In the last part of the chapter, I show the first resuHs obtained with PFS data: i.wo planetary systems orbiting around the stars GJ667C and GJ221. The first one consists of two and possibly three super Earths, one of them in the middle of the star's Habitable Zone. The second, a super Earth and a Neptune in short and long periods respectively. These results confirm the fact that it is possihle to dctect low-mass planets around these stars, especially around their Habitable Zones. In the context of the general planetary population and of M dwarfs in particu lar, these four new planets contribute to clarify the various statistical properties of extrasolar planets. With less than 100 planets found around low-mass (<::: 0.7M8 ) stars, every detection counts. Furthermore, as these planets are also part of multiple planet systems, they also help in the analysis of stars that harbar multiple planetary companions and any correlations emerging in the distribution of orbital elements as suggested by observational clues (Wright et al. 2009). 5.2 Ongoing and future work Obtaining S -values, log R~K measurements, age and rotation period estimates, as well as velocity jitter for our target stars was the most straightforward result from the chromospheric study, but there is still more work to be done with this data. I looked for periodicities in the S-values but the number of data points per star was too low and too sparse to obtain conclusive periods. By combining this data with new PFS S-values and log R~K measurements a thorough study of the effect stellar activity has on radial velocity measurements (see for example, Lovis el al. 0 011) should be made. Many more improvements can be made to the deconvolution routine that was adopted for Keck/HIRES. Although I decided to settle on the Boosted Maximum Likelihood algorithm for now, I would like to explore with the Damped Richardson Lucy algorithm which delivered only slightly worse results. Since there is still a fair 73 leve! of ringing of the DRL templates, I woulcllike to see how a parabolic smoothing algorithm would help by comparing these results with the ones I obtain using BJVIL. I woulel also like to explore the possibility of performing JVIonte Cario simulations in oreler to optimize the eleconvolution parameters usecl in the B splines cleconvolution routine. As it was previously mentioned, I am currently working on implementing the BlVIL algorithm on PFS elata. I have chosen our 5 most stable M elwarfs from the j\Jiagellan Planet Search ancl began performing the same tests clone to the Keck spec tra. lVIy goal is to achieve < lm s-1 precision for these stars. After 2 years of successful anel exciting high caclence observing runs, the Mag ellan Planet Search Program promises to yielel more cliscoveries in the near future, <1lthough mor~ ohs~rvations ::cr~ n~~dcd to confirm onr curr~nt ~m~rging candidat~s. In aelclition, these neecl to be followecl up by photometric observations, as m"st of these planets are close-in, short-period planets, where the probability of finding a transit is non-negligible. I woulel be eager to perfonn such photometric observations and comhin~ th~m with radial vclociti~s, as th~y off~r th~ bcst sc~nario to lat.~r d~rivc the basic planet parameters (mass, raclius, perioel, inclination anel semimajor axis) ancl truly reveal any potential habitability of these planets. Current planet searches suggest that at least 30% of G ancl K dwarfs host N eptune or lower mass planets orbiting with perioels of 50 clays or less (Mayor et al. 2009) with progressively lower-mass planets preclicteel to be more plentiful (eg. ·wittenmyer et al. 2011). vVith these numbers in mind, anel consiclering that with the current precision achieveel with PFS, we expect to cletect at least 30 of these short-perioel low-mass planets in the next couple of years. At this point, and as clemonstrateel by the two eliscoveries presenteel in the previous chapter, the rate of planet detection is limitecl only by telescope time. For now, the low-mass planet sample is incomplete in the planet's periocl-mass range that is out of our present reach given our current precision and time-caclence coverage. In arder to estímate this incompleteness, the cletectability of planets anel the number of missecl planets using present elata shoulel to be calculatecl, as clone by \iVittenmyer et al (2011) for AAT "One :tvieter Per Seconel" stars. 74 Bibliography Anglada-Escudé, G., López-Morales, M., & Chambers, J. E. 2010, ApJ, 709, 168 Anglada-Escudé, G., Arriagada, P., Butler, R.P. et al. 2012, ApJ Letters, 751, 16 Anglada-Escudé, G. & Butler, R.P. 2012, ApJS, 200, loA Arriagada, P., Butler, R. P. , :Vlinniti, D., Lópcz-Moralcs, M., et al., 2010, ApJ, 711, 1229. Arriagada, P., 2011, ApJ, 734, 70. Baliunas, S. L., Donahue, R. A. ,Soon, W. H., et al. 1995, ApJ, 438, 269 Baliunas, S. L., Donahue R. A., Soon W., Gilliland R., Soderblom D. R. 1995b, BAAS, 27, 839 Baranne, A., et al. 1996, A&AS, 119, 373 Bean, J. L., McArthur, B. E., Benedict, G. F., Harrison, T. E., Bizyaev, D., Nelan, E., & Smith, V. V. 2007, AJ, 134, 749 Benedict, G. F., McArthur, B. E., Forveille, T., Delfosse, X., Nelan, E., Butler, R. P., Spiesman, W., Marcy, G., Goldrnan, B., Perrier, C., Jefferys, W. I-I., & Mayor, M. 2002, ApJ, 581, L115 Bennet, D.P., 2009, e-prints arXiv:0902.1761 75 Bernstein, R., Shectman, S. A., Gunnels, S. NI., Niochnacki, S., & Athey, A. E. ..2003, SPIE, 4841, 1694 Bertero, :M., & Boccacci, P. 1998, Introduction to Inverse Problems in Ima.ging (Lon don: Inst. Phys.) Bo11fils, X., Delfosse, X., Udry, & et al. 2011, e-pri11ts arXiv:llll ..J019 Borucki, Vv. J., Koch, D. G., Ba.sri, G., & et al. 2011, ApJ, 736, 19 Boss, A.P. 2006, ApJ, 643, 501 Butler, R.P. 1993, ApJ, 415, 323 Butler, R. P., Nia.rcy, G. \N., Vlilliams, E., lVIcCarthy, C., Dosanjh, P., & Vogt, S. S. 1996, PASP, 108, 500 Butler, R.P., lVIarcy, G.vV., Fischer, D.A., & Brown, T.rd. et al., 1999, ApJ, 52t, 890 Butler, R.P. , Vogt, S.S., Ma.rcy, G.vV. , Fischer, D.A. et al., 2004, ApJ, 617, 580 Butler, R. P., Johnson, J. A., Niarcy, G. vV., vVright, Jason T., Vogt, S. S., & Fischer, D.A. 2006, PASP, 118, 1685 Casagrande, L., Flynn, C. & Bessell, M. 2008, NINRAS, 389, 585 Chambers, J. E. 1999, lVINRAS, 304, 793 Cha.rbonneau, D., Brown, T.rd., Latham, D.vV., & Tvia.yor, IVI. 2000 ApJ Letters 529, L45 Charbonneau, D., Brown, T. 1VI., Burrows, A., & La.ughlin, G. 2007, Protostars and Pla.nets V, 701 76 Charbonneau, D., Berta, Z. K., Irwin, J., Burke, C. J., Nutzman, P., Buchhave, L. A., Lovis, C., Boufils, X., Latham, D. W., Udry, S., Murray-Clay, R. A., Holman, M. J., Falco, E. E., Winn, J. N., Queloz, D.; Pepe, F., Mayor, M., Delfosse, X., Forveille, T. 2009, Nature, 462, 891 Crane, J. D., Shectman, S. A., Butler, R. P., & et al. 2010, in SPIE Conference Series, Vol. 6269 Crane, J. D., Shectman, S. A., Butler, R. P., & et al. 2008, in SPIE Conference Series, Vol. 7014 Crane, J., Shectman, S. A., Butler, R. P., Thompson, I.B., Birk, C., Jones, P., & Burley, G.S. 2010, in SPIE Conference Series, Vol. 7735 Cochran, W.D., et al. 2011, ApJS, 197, 7 Cumming, A. 2004, MNRAS, 354, 1165 Cumming, A., Butler, R. P., Marcy, G. W., Vogt, S. S., Wright, Jason T. & Fischer, Debra A. 2008, PASP, 120, 531 Cvetkovic, Z., & Ninkovic, S. 2011, VizieR On-line Data Catalog: J / other /Ser /180.71, 4201, 18001 Delfosse, X., Forveille, T., Ségransan, D., Beuzit, J.-L., Udry, S., Perrier, C., & Mayor, M. 2000, A&A, 364, 217 Dempster, A., Laird, N., & Rubin, D. 1977, J. Royal Stat. Soc. B, 39 Dierckx,P. 1984, Journal of Computational Physics, 52, 163 Duncan, D. K., Vaughan, A. H., Wilson, O. C., et al. 1991, ApJS, 76, 389. Ford, E. B. 2005, AJ, 129, 1706 77 Gilliland, R. L., fi!Iorris, S. L., vVeymann, R. J., Ebbets, D. C., & Lincller, D. J. 1992, PASP, 104, 367 Gizis, J. E., Reicl, I. N. & Hawley, S. L. 2002, AJ, 123, 33560 Comes da Silva, .T., Santos, N.C., Bonfils, X., Delfosse, X., Forveille, T., Udry, Dn musque, S & C. Lovis 2012, A&A, 541, A9 Gray, R. 0., Corbally, C. J., Garrison, R. F., iVIcFaclden, M. T., Bubar, E. J., Mc Gahee, C. E., O'Donoghue, A. A., & Knox, E. R. 2006, AJ, 132, 161 Hawley, S. L., Gizis, J. E. & Reicl, l. N. 1996, AJ, 112, 2799H Henry, T. J., Soclerblom, D. R., Donahue, R. A., & Baliunas, S. L. 1996, AJ, 111, 439 Henry, G.\iV., rviarcy., G.W., Butler, R.P., Vogt, S.S. 2000 ApJ Letters 529, L41 Henry, G. vV., Donahue, R. A., & Baliunas, S. L. 2002, ApJ, 577, 111 Henry, T. J., Jao, vV., Subasavage, J. P., Beaulieu, T. D., Ianna, P. A., Costa, E., ::VIénde7., R. A., 2006, A.T, 132, 2360 Holman, M.J., et al. 2010, Science, 330, 51 Ida, S. & Lin, D. N. C., 2005, ApJ, 626, 1045 Irwin, J., Berta, Z. K., Burke, C. J., Charbonneau, D., Nutzman, P., Vlest, A. A., & Falco, E. E. 2011, ApJ, 727, 56 Isaacson & Fischer 2010, AJ, 725, 875. Jcnkins, J. S., Joncs, II. R. J\ .. Pavlcnko, Y., Pillficld. D. J., Ba.mcs, & J. R., Lyubchik, Y. 2008, ApJ, 485, 571. Jenkins, J. S., Jones, H. R. A., Tinney, C. G., et al. 2006, iVINRAS, 372, 163. 78 Jorres, H. R. A., Butler, R. P., Tirrrrey, C. G., O'Toole, S., & Witterrmyer, R. 2010, MNRAS, 403, 1703 Karre, S. R., & Gelirro, D. M. 2011, ApJ, 741, 52 Kastirrg, J. F., Whitmire, D. P., & Reyrrolds, R. T. 1993, Icarus, 101, 108 Kerrrredy, G. M., & Kerryorr, S. J. 2008, ApJ, 682, 1264 Kornet, K., Wolf, S., & Rózyczka, M. 2006, A&A, 458, 661, S. J. 2008, ApJ, 682, 1264 Larrdweber, L. 1951, Am. J. Math., 73, 615 Laughlirr G., Boderrheimer, P., Adams F.C. Lissauer, J. J., & Rivera, E. J. 2001, ApJ, 554, 1141 López-f./Iorales, M., Butler, R. P., Fischer, D. A., :Mirrrriti, D., Shectmarr, S. A., Takeda, G., Adams, F. C., Wright, J. T., & Arriagada, P. 2008, AJ, 136, 1901 Lovis, C., et al. 2011, ArXiv e-prirrts Lucy, L. 1974, AJ, 79, 745 Mamajek, EricE. & Hillerrbrarrd, Lyrrrre A. 2008, ApJ, 687, 1264 Marcy, G. W., Butler, R. P., Fischer, D. A., Laughlirr, G., Vogt, Steverr S.;,Henry, G. W., Pourbaix, D., 2002, ApJ, 581, 1375 Marcy, G. W., Butler, R. P., Fischer, D. A., Vogt, Steverr S., Wright, J.T., Tirrrrey, C.G. & Jorres, H.R.A. 2005, Progress of Theoretical Physics Supplemerrt No. 158, ~~~2005 Mayor, M. & Queloz, D. 1995, Nature, 378, 355 79 Nlayor, M., et al. 2009, A&A, 507, 487 Mayor, rd. et al. 2011, ArXiv e-prints ·McArthur, B.E., Encll, M., Cochran, 'N.D., G., Beneclict, G.F., Fischer, D.A., Nlarcy, G.vV., Butler, R.P., Naef, D., IVlayor, NI., Queloz, D., Uchy, S. & Harrison, T.E. 2004, ApJ, 614, L81 .Meschiari, S., \i'lolf, A. S., Rivera, E., Laughlin, G., Vogt, S., & Butler, P. 2009, PASP, 121, 1016 Minniti, D., Butler, R. P., López-lVIorales, M., Shectman, S. A., Aclams, F. C., Ar riagacla, P., Boss, A. P., &Chambers, J. E. 2009, ApJ, 693, 1424 Morhác, :M., & :viatousek, V. Digital Signal Processing, 2009, 19, 372 Murray, C.D., & Dermott, S.F. 1999, Solar Syslem Dynamics, Cambridge: Cam bridge University Press, 592 pp. Noyes, R. vV., Hartmann, L. \"Al., Baliunas, S. L., Duncan, D. K., Vaughan, A. H. 1984a, ApJ, 279, 763. O'Toole, S. J., Tinney, C. G., Jones, H. R. A., Butler, R. P., lVIarcy, G. Vl., Carter, B., & Bailey, J. 2009, ri/INRAS, 392, 641 Pepe, F., J\;Iayor, M., Gallancl, & et al. 2002, A&A, 388, 632 Pepe, F., Rupprecht, G., Avila, G., & el al. 2003, in SPIE Conference Series, Vol. 4841, 1045-1056 Pepe, F., et al. 2011, A&A, 534, A58+ Penin, M.-N., Cayrel ele Strobel, G., & Dennefelcl, M. 1988, A&A, 191, 237 80 Perryman, M. A. C. et al. 1997, A&A, 323, L49. The Hipparcos Catalog Queloz, D., Henry, G. W., Sivan, J. P., Baliunas, S. L., Beuzit, J. L., Donahue, R. A., Mayor, M., Naef, D., Perrier, C., & Udry, S. 2001, A&A, 379, 279. Queloz, D., Bouchy, F., Mout0u, C. et al., 2009, A&A, 506, 303 Reffert, S., & Quirrenbach, A. 2011, A&A, 527, A140+ Reid, I. N., Hawley, S. L. & Gizis, J. E. 1995, AJ, 110, 1838R Richardson, W. 1972, J. Opt. Soc. Am., 62, 55 Rivera, E.J., Lissauer, J.J., Butler, R.P., Marcy, G.\iV., Vogt, S.S. et al. 2005, ApJ, 634, 625 Rivera, E.J., Butler, R.P., Vogt, S.S. et al. 2010, ApJ, 708, 1492 Saar, S. H.; & Fischer, D. 2000 ApJ, 534, 105. Saar S. H., Butler R. P. & Marcy G. W., 1998, ApJ, 498, L153. Saar S. H., & Donahue R. A. 1997, ApJ, 485, 319 Santos, N. C., Mayor, M., Naef, D., et al. 2003, A&A, 406, 373. Santos, N. C., Udry, S., Mayor, M., Naef, D., Pepe, F., Queloz, D., Burki, G., Cramer, N., & Nicolet, B. 2000, A&A, 361, 265. Shen, Y., & Turner, E. L. 2008, ApJ, 685, 553 Shepp, L., & Vardi, Y. 1982, IEEE Trans. Medical Imaging, 2, 113 81 Skifi, B. A. 2010, Vi7,ieR Online DaLa CaLalog, 10, 2023 Soderblom D. R., Duncan D. K., Johnson D. R. H., 1991, ApJ, 375, 722. Tinney, C. G., l\IcCarthy, C., Jones, H. R. A., et al. 2002, i\IINRAS, 332, 759 Tokovinin, A. 2008, i\IINRAS, 389, 925 Uclry, S., :Mayor, f.il., Benz, vV. et al., 2006, A&A, 447, 361 Uclry, S. & Santos, N. 2007, ARA&A 45, 397 Upgren, A. R., Grossenbacher, R., Penhallow, vV. S., MacConnell, D. J. & Frye, R. L. 1972, AJ, 77, 486U Valenti, J. A., Butler, R. P., & :Marcy, G. W. 1995, PASP, 107, 966 van Leeuwen, F. 2007, A&A, 474, 653 Vogt, S. S., Allen, S. L., Bigelow, B. C., & et al. 1994, in SPIE Conference Series, ecl. D. L. Crawforcl & E. R. Craine, Vol. 2198, 362 Vogt, S. S., Butler, R. P., Rivera, E. J., Haghighipour, N., Henry, G. vV., & \iVilliamson, i\II. H. 2010, ApJ, 723, 954 Vogt, S.S., vVittenmyer, R.A., Butler,R.P. et al. 2010, ApJ, 708, 1366 \iVhite, R.L., 1994, The Restoral'ion of HST Images and Spectra, Hanish, R.J. ancl vVhite, R.L., ecls. Space Telescope Science Institute, 104. \iVittenmyer, R. A., Tinney, C. G., Butler, R. P., O'Toole, S. J., Jones, H. R. A., Carter, B. D., Bailey, J. & Horner, J., ApJ, 738, 81 Wolszczan, A. & D. A. Frail, D.A. 1992 Nature, 355, 145 82 Wordsworth, R. D., Forget, F., Selsis, F., Millour, E., Charnay, B., & Madeleine, J.-B. 2011, ApJL, 733, L48 Wright, J. T. 2005, PASP, 17, 657. Wright, J. T., Marcy, G. W., Butler, R. P., Vogt, S. S. 2004, ApJS, 152, 261. _1 83 Appendix : Individual S-values, log R~.K , Age, Rotation Periods and Jitter estimates. Tablc 1: Derivccl S-values from 1\IIIKE spcctra convcrtccl to MvV system, SwKE; chromosphcric activity inclices, log R;,K , rotation pcriocls, P,",; agcs, log(Agcjycars); ancl cstimatccljitter, O";,v (Isaacson & Fischcr (2010) ancl Wright (2005)) labclccl as 1 ancl 2 rcspcctivcly. K ame B -V N""' SwKE log R;,K P,", log(Agcjycars) (HD) (clays) 224789 0.863 7 0.458 -4.47 14. 8.71 3.5 8.8 225155 0.7Lll 4 0.150 -5.08 9.91 2.1 2.2 225299 0.710 6 0.193 -4.86 28. 9.61 2.3 2.1 23 0.577 8 0.173 -4.89 15. 9.67 2.8 3 ·'i 55 1.076 4 0.454 1.6 2.1 361 0.624 6 0.195 -4.81 17. 9.52 3.1 4.6 798 0.448 6 0.157 -4.97 5. 9.78 2.4 3.0 1002 0.640 4 0.143 -5.12 9.92 2.2 2.4 1237 0.750 1 0.458 -4.36 6. 8.33 3.5 18.'1 hip1532 1.318 3 0.839 2.4 2.1 1690 1.354 2 0.166 0.5 2.2 1893 0.769 11 0.244 -Ll. 73 26. 9.36 2.5 4.'1 1910 1.075 6 0.952 1.6 7.1 2222 0.675 7 0.139 -5.16 9.97 2.1 2.2 3222 0.854 7 0.189 -4.95 44. 9.75 2.2 2.1 3359 0.780 5 0.144 -5.12 9.93 2.1 2.1 4113 0.716 9 0.149 -5.08 9.91 2.1 2.1 L1333 0.414 4 0.175 -4.86 3. 9.62 2.4 4.0 4631 0.774 7 0.159 -5.04 9.86 2.1 2.1 5L199 0.987 3 0.114 1.6 4.3 5349 0.991 2 0.115 1.6 4.3 6107 0.650 8 0.143 -5.12 9.93 2.2 2.4 Conl:inucrl on ne:cí pagc 84 Table 1 - Continued frorn previous page N ame E-V Nobs s~liKE log B!¡,K P,·ot log(Age/years) a~v 1 (}~V 2 (HD) ( days) (m s-1 ) (m s-1 ) hip4845 1.365 3 1.270 ... 3.5 2.2 6156 0.800 5 0.355 -4.54 17. 8.90 3.0 8.0 6236 0.591 4 0.146 -5.08 9.91 2.3 2.6 6434 0.613 8 0.144 -5.10 9.90 2.2 2.5 6790 0.561 5 0.154 -5.00 9.82 2.5 2.7 6880 0.764 4 0.234 -4.75 27. 9.41 2.5 4.1 6910 0.651 6 0.126 -5.29 10.14 1.9 4.3 7134 0.589 6 0.151 -5.04 9.87 2.4 ~.6 7399 0.476 8 0.134 -5.17 9.99 2.1 3.0 7449 0.575 8 0.168 -4.92 15. 9.71 2.7 2.6 7661 0.753 4 0.397 -4.43 10. 8.59 3.3 13.4 7786 0.494 6 0.171 -4.88 7. 9.64 2.8 5.0 8049 0.876 8 0.678 -4.30 3. 8.14 4.5 16.2 8076 0.622 14 0.220 -4.71 14. 9.32 3.6 6.9 8129 0.702 4 0.180 -4.91 29. 9.70 2.2 2.1 8326 0.970 7 0.340 ... 2.7 2.1 8406 0.656 4 0.162 -4.98 27. 9.80 2.5 2.3 8581 0.569 7 0.133 -5.20 10.04 2.1 2.7 9175 0.656 7 0.145 -5.10 9.90 2.2 2.3 9782 0.593 7 0.141 -5.12 9.93 2.2 2.6 9847 0.712 5 0.140 -5.15 9.97 2.0 2.5 9905 0.761 8 0.172 -4.97 38. 9.78 2.2 2.1 hip7554 1.416 2 1.782 ... 4.8 2.3 10008 0.797 3 0.421 -4.45 11. 8.62 3.4 11.3 10226 0.607 3 0.176 -4.89 18. 9.66 2.8 2.5 10576 0.590 9 0.142 -5.11 9.91 2.2 2.6 10370 0.714 5 0.164 -5.00 35. 9.81 2.2 :....1 10519 0.617 4 0.138 -5.16 9.98 2.1 2.5 10678 0.714 7 0.213 -4.79 25. 9.48 2.4 4.1 10611 0.853 7 0.477 -4.44 12. 8.61 3.6 10.0 Continued on next page 85 Table 1 - Contin:ued from prev'io·us page N ame B-V JVol" shl!Kió logR;m pro! log(Agejyears) CT~v 1 (}~V 2 (HD) ( clays) (rn s-1 ) (m s-1 ) 10700 0.727 127 0.160 -5.02 9.84 2.1 2.1 1126Ll 0.66Ll 3 0.140 -5.15 9.97 2.1 2.2 11231 0.463 7 0.127 -5.26 10.10 1.9 3.0 117M 0.682 6 0.137 -5.17 9.99 2.1 2.6 12058 l.H3 L1 0.944 1.6 Ll,5 13060 0.797 5 0.217 -4.82 33. 9.55 2.4 2.1 12951 0.576 9 0.130 -5.23 10.08 2.0 2.6 12585 0.662 11 0.157 -5.02 9.85 2.'1 2.3 12617 1.020 5 0.510 1.6 4.1 13350 0.766 4 0.137 -5.17 9.99 2.0 2.6 13386 0.880 5 0.174 -5.02 9.8L1 2.1 2.1 13724 0.667 8 0.205 -4.79 21. 9A8 3.3 4.5 13808 0.870 4 0.235 -Ll.83 37. 9.56 2A ?..1 hip10337 1.346 2 1.689 Ll,6 2.2 13789 1.055 4 0.711 1.6 5.3 14629 1.031 3 0.569 1.6 Ll,5 14635 1.079 L1 0.671 ... 1.6 4.2 H758 0.6L14 5 0.158 -5.01 9.83 2.5 2.3 1523L1 OA17 7 0.170 -4.89 3. 9.66 2A 3.0 15507 0.670 5 0.161 -5.00 29. 9.82 2.5 2.2 15590 0.652 4 0.131 -5.23 10.07 2.0 2.6 15767 0.935 3 OA4L1 3.3 5.'1 16348 1.045 5 0.52L1 1.6 3.7 16950 0.616 6 0.128 -5.26 10.11 2.0 2.6 17155 1.0'10 3 0.294 ... 1.6 2.1 17289 0.588 3 0.139 -5.1¿1 9.96 2.2 2.6 1713L1 0.634 5 0.143 -5.12 9.92 2.2 2.'1 hip12961 1.389 2 1.336 3.6 2.3 hip13389 1.549 1 0.255 ... 1.3 2.5 18168 0.933 3 0.516 ... 3.7 7.0 Cont'irmcd on ncxl pagc 86 Table 1 - Continued fmm previous page N ame E-V Nohs SMIKE logR~K prot log(Age/years) a~v 1 CT~v 2 (HD) (days) (rn s-1 ) (m s-1 ) 18754 0.736 6 0.134 -5.19 10.02 2.0 4.3 18708 0.651 5 0.134 -5.20 10.03 2.0 2.3 18819 0.597 5 0.120 -5.37 10.23 1.8 3.0 18894 0.598 5 0.179 -4.87 16. 9.62 2.9 3.9 19493 0.695 2 0.135 -5.19 10.02 2.0 2.6 19330 0.566 8 0.1i)1 -4.84 13. 9.59 2.9 4.7 19880 0.715 4 0.134 -5.20 10.03 2.0 2.6 20003 0.771 6 0.187 -4.91 36. 9.70 2.3 2.1 19603 0.603 6 0.142 -5.12 9.92 2.2 2.5 hip14593 1.340 2 0.778 2.2 2.2 19916 0.585 7 0.137 -5.15 9.97 2.2 2.6 20155 0.622 6 0.139 -5.15 9.97 2.1 2.6 20299 0.428 4 0.161 -4.94 4. 9.74 2.3 3.0 20407 0.586 12 0.151 -5.04 9.87 2.4 2.6 20339 0.614 6 0.143 -5.11 9.91 2.2 2.5 20584 0.584 9 0.123 -5.33 10.19 1.9 3.8 20657 0.800 5 0.136 -5.18 10.00 2.0 3.8 21036 0.730 4 0.290 -4.60 17. 9.05 2.8 7.8 21411 0.716 7 0.232 -4.73 22. 9.36 2.5 5.0 21749 1.130 6 0.589 ... 1.6 2.1 21841 0.666 8 0.235 -4.69 17. 9.27 3.8 ti.6 21938 0.591 6 0.143 -5.11 9.91 2.2 2.6 22177 0.717 6 0.150 -5.08 9.90 2.1 2.1 22323 0.401 5 0.222 -4.66 2. 9.21 3.1 16.0 22582 0.730 5 0.148 -5.09 9.92 2.1 2.1 23295 1.089 3 0.376 ... 1.6 2.1 23576 0.587 6 0.141 -5.12 9.92 2.2 2.6 24085 0.595 6 0.148 -5.07 9.89 2.3 2.6 24331 0.913 7 0.273 ... 2.6 2.1 24650 0.569 6 0.140 -5.13 9.94 2.2 2.7 Continued on next page 87 Table 1 - Contúz.-ued .frmn JFrev'io·ns page 0 N ame B-V J\Tob';; sl\fiKB )ogR;rK P,·ot log(Agejyears) 0 ~v 1 ~v 2 (HD) ( clays) (m s-1 ) (m s-1 ) hip18280 1.366 5 0.992 2.7 2.2 2505'1 0.536 5 0.154 -5.00 9.82 2.5 2.7 2500Ll 1.214 5 1.221 1.6 4.0 25797 0.422 6 0.207 -4.71 2. 9.32 3.1 11.9 26071 0.742 16 0.152 -5.07 9.90 2.1 '>.6 26040 0.576 4 0.152 -5.03 9.85 2A 2.6 lüp19976 1.602 3 0.867 2A 2.5 27426 1.052 13 0.671 1.6 5.0 hip20142 1.654 2 1.040 1.6 2.5 27,166 0.673 7 0.243 -'1.67 17. 9.23 3.9 7.0 27905 0.627 3 0.150 -5.06 9.89 2.3 2.4 28471 0.650 4 0.144 -5.11 9.91 2.2 2.3 28701 0.650 Ll 0.145 -5.10 9.90 2.2 2.3 28185 0.750 12 0.154 -5.06 9.88 2.1 2.1 28821 0.683 5 0.158 -5.02 9.8Ll 2.4 2.2 29086 0.990 11 0.691 ... 4.3 7.7 29220 1.106 5 0.787 ... 1.6 4A 29161 0.632 8 0.222 -4.71 15. 9.32 3.6 6.7 30306 0.747 Ll 0.150 -5.08 9.91 2.1 2.1 29813 0.621 4 0.153 -5.03 9.86 2.4 2.4 30278 0.746 10 0.155 -5.05 9.88 2.1 2.1 30669 0.795 12 0.159 -5.04 9.87 2.1 2.1 3321Ll 0.890 4 0.326 -4.67 27. 9.24 2.8 L1.2 hip22772 1.271 3 0.721 1.6 ?.1 31392 0.792 10 0.270 -4.69 25. 9.27 2.7 4.9 32564 0.741 5 0.150 -5.08 9.91 2.1 2.1 32724 0.618 11 0.139 -5.14 9.96 2.2 2.5 32842 0.492 7 0.179 -4.83 7. 9.57 2.9 6.3 hip23708 1.386 1 1.737 4.7 2.3 35877 0.573 2 0.216 -4.70 10. 9.30 3.5 8.5 Con[irmed on ncü pagc 88 Table 1 - Continued from previous page N ame B-V Nobs s~l!KE logR~K prot log(Agejyears) a~v 1 a~v2 1 1 (HD) (days) (m s- ) (m s- ) 33093 0.606 10 0.130 -5.23 10.07 2.0 2.6 35267 0.513 5 0.129 -5.24 10.08 2.1 2.9 33873 0.704 3 0.242 -4.69 20. 9.28 2.5 !'i.8 33822 0.708 8 0.141 -5.14 9.95 2.1 2.1 33693 0.472 15 0.137 -5.14 9.95 2.1 3.0 hip24392 1.645 3 1.021 ... 2.1 2.5 34195 0.567 8 0.128 -5.26 10.11 2.0 2.6 35041 0.636 9 0.304 -4.50 8. 8.78 5.0 15.3 36379 0.561 5 0.134 -5.19 10.02 2.1 2.7 37351 0.594 5 0.191 -4.81 15. 9.52 3.1 5.0 37761 0.764 4 0.125 -5.27 10.12 2.0 4.3 37706 0.769 8 0.186 -4.92 36. 9.70 2.3 2.1 37986 0.801 8 0.151 -5.09 9.91 2.1 2.1 38459 0.861 1 0.471 -4.46 13. 8.66 3.6 9.3 38683 0.571 3 0.180 -4.85 13. 9.60 2.9 3.8 38277 0.637 7 0.139 -5.15 9.97 2.1 2.4 38467 0.672 6 0.145 -5.10 9.90 2.2 2.2 38554 0.704 8 0.175 -4.93 31. 9.73 2.2 2.1 38677 0.581 6 0.137 -5.16 9.98 2.1 2.6 hip27323 1.377 3 1.828 ... 4.9 2.2 39503 0.732 5 0.160 -5.02 9.85 2.1 2.1 hip27803 1.321 3 0.859 ... 2.5 ~.1 40105 0.891 8 0.137 -5.17 10.00 2.0 4.3 39855 0.700 9 0.173 -4.94 31. 9.75 2.7 2.1 hip28153 1.404 3 1.361 ... 3.6 2.3 41158 0.519 2 0.142 -5.09 9.92 2.3 2.8 41155 0.559 5 0.121 -5.36 10.22 1.9 2.6 42044 0.543 5 0.117 -5.43 10.28 1.8 2.6 42538 0.593 5 0.126 -5.29 10.14 1.9 2.6 42719 0.657 5 0.128 -5.27 10.11 1.9 2.6 Continued on next page 89 Table 1 - Conün:ued jf'O'm pre·u'iO'Its puge N ame B-\1 JVoh> sl'diKE log R~ 11 , prot log(Agejyears) (}~V 1 ()RV 2 1 1 (HD) (days) (m s- ) (m s- ) 43197 0.817 7 0.146 -5.11 9.91 2.1 2.1 43,170 0.539 2 0.134 -5.18 10.01 2.1 2.8 43848 0.927 6 0.173 ... 2.1 2.1 L14310 0.837 2 0.156 -5.07 9.90 2.1 2.1 H573 0.915 4 0.409 3.2 5A H569 0.619 10 0.157 -5.01 9.83 2.5 2.5 ,15987 0.655 10 0.146 -5.10 9.92 2.2 2.3 45738 0.460 11 0.163 -'1.93 6. 9.72 2.5 3.6 '16435 0.702 2 0.224 -'1.7'1 22. 9.39 2.4 4.9 4689,1 0.778 5 0.136 -5.17 10.00 2.0 Ll,3 47017 0.609 6 0.129 -5.25 10.09 2.0 2.5 46872 0.536 2 0.205 - 11.73 8. 9.36 3.4 8.8 47855 0.465 7 0.189 -4.79 5. 9.48 3.0 8.2 47186 0.71'1 8 0.150 -5.08 9.91 2.1 2.1 hip31862 1.377 4 1.285 3.5 2.2 hip31878 1.257 9 2.339 ... 1.6 7.0 48265 0.747 16 0.132 -5.21 10.04 2.0 '1.3 48056 0.576 2 0.160 -4.97 17. 9.78 2.6 ~.6 48286 0.563 7 0.153 -5.01 9.83 2.5 2.7 49035 0.755 2 0.228 -4.76 27. 9.44 2.5 4.0 hip32530 1.306 6 0.757 2.3 2.1 49866 0.663 2 0.1S7 -5.02 9.85 2.'1 4.3 51608 0.771 1 0.159 -5.03 9.86 2.1 2.1 52217 0.823 8 0.229 -11.81 33. 9.52 2.4 2.1 53,166 0.551 1 0.133 -5.20 10.03 2.1 2.8 53680 1.180 7 0.514 ... 1.6 2.1 hip3L1785 1.330 L1 0.811 2.3 2.1 56Ll13 0.754 2 0.184 -4.92 35. 9.70 2.3 2.1 56662 0.6011 18 0.148 -5.07 9.90 2.3 2.5 hip35173 1.010 5 0.320 ... 1.6 2.1 Contirmed on next page 90 Table 1 - Continued from previous page N ame B-V Nobs SM!hE logR~K prot log(Agc/ycars) J~v 1 a~v2 1 1 (HD) (days) (m s- ) (m s- ) 57553 0.506 1 0.172 -4.88 8. 9.64 2.8 2.8 hip35937 1.274 2 0.815 ... 1.6 2.1 hip35943 1.365 3 1.741 ... 4.7 2.2 58556 0.568 14 0.204 -4.74 11. 9.39 3.3 7.2 59711 0.637 6 0.156 -5.02 9.84 2.4 2.4 61475 0.853 7 0.340 -4.61 23. 9.10 2.9 5.4 61214 1.049 8 0.815 ... 1.6 6.7 61051 0.766 3 0.158 -5.04 9.86 2.1 2.1 60779 0.564 10 0.149 -5.05 9.87 2.4 2.7 hip37727 0.700 15 0.296 -4.56 14. 8.95 4.8 9.7 62549 0.612 10 0.150 -5.05 9.88 2.3 2.5 64114 0.721 5 0.218 -4.77 25. 9.45 2.4 4.2 hip38594 1.385 7 1.087 ... 2.9 2.3 hip38910 1.136 8 0.621 ... 1.6 : 1 66039 0.582 2 0.199 -4.77 13. 9.45 3.2 6.1 67200 0.595 1 0.131 -5.22 10.06 2.0 2.6 68475 0.887 6 0.305 -4.71 29. 9.31 2.7 3.8 68607 0.862 1 0.277 -4.73 30. 9.37 2.6 3.6 69611 0.584 8 0.147 -5.06 9.89 2.3 2.6 70081 0.660 2 0.140 -5.14 9.96 2.1 2.3 72579 0.790 8 0.157 -5.05 9.88 2.1 2.1 73744 0.607 5 0.157 -5.00 21. 9.82 2.5 2.5 73267 0.806 3 0.154 -5.07 9.90 2.1 2.1 73322 0.910 22 0.458 ... 3.4 6.6 hip42507 1.361 2 1.654 ... 4.5 2.2 hip42601 1.360 6 0.994 ... 2.7 2.2 hip42881 1.432 5 1.153 3.1 2.3 75351 0.625 3 0.142 -5.13 9.93 2.2 2.6 75288 0.673 2 0.149 -5.08 9.90 2.3 2.2 77417 0.770 16 0.177 -4.95 38. 9.75 2.2 2.1 Contin·ued on next page 91 Tablc 1 Continued from previous page Name B-V J\Tob;; sl'diKE ]ogR;rK pn>t log(Agc/ycars) (}~V 1 (}~V 2 (HD) (clays) (m s-1 ) (m s-1 ) 77825 0.960 21 0.755 ... L1.7 10.8 hip44722 1.418 7 1.937 5.2 2.3 78747 0.575 22 0.146 -5.07 9.90 2.3 2.6 78558 0.617 17 0.151 -5.05 9.88 2.4 2.5 78612 0.610 22 0.142 -5.11 9.92 2.2 2.5 hip44899 1.316 10 1.007 2.9 2.1 79985 0.659 3 0.126 -5.28 10.13 1.9 2.6 81238 0.509 1 0.157 -4.97 10. 9.78 2.5 2.9 81110 0.72L1 15 0.193 -4.87 29. 9.62 2.3 2.1 81659 0.670 18 0.251 -4.64 15. 9.17 4.1 7.7 hip46549 1.305 8 1.021 3.0 :¿_1 82342 0.985 9 0.272 2.4 2.1 82516 0.927 9 0.228 ... 2.3 2.1 85249 0.492 1 0.132 -5.19 10.03 2.1 3.0 85390 0.855 10 0.210 -4.89 40. 9.66 2.3 2.1 hi¡JL18336 1.4,16 7 0.764 2.0 2.4 85380 0.577 10 0.141 -5.12 9.92 2.2 2.6 hip48502 1.375 3 0.835 2.3 2.2 86006 0.708 2 0.140 -5.15 9.97 2.0 2.6 86249 0.943 4 0.341 ... 2.8 2.1 86226 0.647 15 0.168 -4.95 25. 9.75 2.6 2.3 86397 0.725 7 0.175 -4.94 33. 9.7'1 2.2 2.1 86652 0.630 2 0.153 -5.04 9.86 2.4 2.4 87966 0.415 1 0.168 -4.90 3. 9.68 2.3 3.0 hip,19577 1.331 4 0.936 2.7 2.1 87931 0.760 3 0.161 -5.02 9.85 2.1 2.1 87998 0.622 21 0.147 -5.08 9.91 2.3 2.4 89591 0.642 5 0.165 -4.96 25. 9.77 2.6 2.6 89418 0.569 20 0.156 -4.99 16. 9.81 2.5 2.7 89839 0.523 3 0.130 -5.22 10.06 2.1 ~: 9 Continucd on ncx/, pagc 92 Table 1 - Continued from previous page N ame B-V Nobs s~l!KE !ogR~K prot log(Agejyears) o-~v 1 a~v 2 1 (HD) (days) (m s-1 ) (m s- ) 89942 0.700 1 0.144 -5.11 9.92 2.2 2.1 89988 0.681 2 0.141 -5.14 9.95 2.1 2.6 90028 0.656 12 0.154 -5.04 9.86 2.4 2.3 90926 0.748 2 0.158 -5.03 9.86 2.1 2.1 90520 0.643 1 0.141 -5.14 9.95 2.2 ;c.6 90884 1.045 14 0.830 ... 1.6 7.0 91682 0.696 1 0.145 -5.11 9.91 2.2 2.3 91320 0.588 1 0.137 -5.16 9.98 2.1 2.6 91901 0.920 11 0.4:20 3.2 5.5 94009 0.770 1 0.134 -5.19 10.02 2.0 4.3 hip52296 1.464 6 1.167 3.2 2.4 hip52341 1.460 2 1.062 ... 2.9 2.4 92719 0.622 13 0.179 -4.88 19. 9.64 2.8 2.4 93083 0.945 10 0.185 ... 2.1 2.1 93489 0.620 12 0.155 -5.02 9.85 2.4 2.4 94151 0.718 14 0.172 -4.95 33. 9.76 2.2 2.1 94270 0.579 11 0.149 -5.05 9.87 2.4 2.6 94482 0.562 15 0.138 -5.14 9.96 2.2 2.7 94690 0.710 2 0.169 -4.96 33. 9.77 2.2 2.1 94838 0.635 10 0.214 -4.74 16. 9.38 3.4 5.9 95090 0.639 1 0.141 -5.13 9.94 2.2 2.4 95136 0.660 6 0.151 -5.06 9.89 2.3 2.3 95338 0.878 8 0.195 -4.94 44. 9.75 2.2 2.1 95521 0.637 16 0.163 -4.97 24. 9.79 2.5 ;.4 96417 0.563 1 0.135 -5.17 10.00 2.1 2.7 hip54373 1.430 8 1.762 ... 4.7 2.3 hip54569 1.400 5 1.260 3.4 2.3 97782 1.108 12 0.452 ... 1.6 2.1 hip54966 1.340 7 1.003 2.8 2.2 98179 0.500 1 0.140 -5.11 9.91 2.2 3.0 Continued on next page 93 Table l - Contúmed from ¡neu'io·us puge N ame B-11 J\Tob';; S,!IKE JogR;,K prot log(Age/years) O"~v 1 0 ~v 2 1 (HD) ( clays) (m s-1 ) (m s- ) hip55119 L383 7 1.5L10 Ll.l 2.3 98459 0.623 1 0.130 -5.2L1 10.09 2.0 2.4 98640 0.765 1 0.338 -4.53 15. 8.88 3.0 2.6 98553 0.59'1 17 0.1M -5.02 9.8'1 2.4 2.6 98649 0.658 5 0.154 -5.04 9.86 2.4 2.3 98727 0.521 1 0.166 -4.91 10. 9.70 2.7 Ll,O 100508 0.828 9 0.158 -5.06 9.88 2.1 2.1 100850 0.661 3 0.135 -5.20 10.03 2.0 2.3 101117 0.638 1 0.251 -4.63 13. 9.13 Ll.l 9.1 101093 0.561 7 0.161 -4.96 15. 9.77 2.6 2.7 101171 0.771 4 0.193 -'1.89 35. 9.66 2.3 2.1 101197 0.659 1 0.152 -5.05 9.88 2.3 2.3 101930 0.908 13 0.188 2.2 2.1 102574 0.585 13 0.130 -5.23 10.07 2.0 2.6 hip57688 1.078 8 0.700 1.6 Ll.5 10280¿1 0.413 1 0.202 -4.73 2. 9.37 2.9 10.2 1028L13 0.787 7 0.186 -4.92 38. 9.71 2.3 2.1 103220 0.7'10 2 0.151 -5.07 9.90 2.1 2.6 103743 0.640 1 0.351 -4.42 6. 8.53 5.8 21.3 103760 0.650 19 0.1'18 -5.08 9.91 2.3 2.3 103836 1.024 11 0.474 1.6 3.6 103949 0.985 8 0.350 2.7 2.1 hip58688 1.419 6 1.182 3.1 2.3 104760 0.645 4 0.157 -5.01 9.83 2.'1 2.3 1QL1982 0.651 14 0.166 -'1.96 26. 9.77 2.6 2.3 105671 1.137 10 0.869 1.6 4.2 105837 0.570 13 0.159 -4.97 16. 9.79 2.5 2.7 10590¿1 0.742 6 0.143 -5.13 9.94 2.1 2.1 106275 0.892 7 0.202 -'1. 94 4Ll. 9.73 2.3 2.1 106290 0.623 1 0.170 -Ll. 93 21. 9.72 2.7 2.4 Contirmcd on nc.Tt pagc 94 Table 1 ·· Continued from previous page N ame B-V Nob' SMIKE logR~K prot log(Agejyears) J~v 1 (}~V 2 (HD) (days) (m s-1 ) (m s-1 ) 106515A 0.815 7 0.156 -5.06 9.89 2.1 2.6 106515B 0.815 8 0.161 -5.04 9.87 2.1 2.1 106869 0.574 8 0.140 -5.12 9.93 2.2 2.7 106937 0.762 8 0.141 -5.14 9.95 2.1 2.6 107008 0.841 7 0.257 -4.76 31. 9.42 2.6 2.1 107077 0.435 7 0.165 -4.92 4. 9.70 2.4 3.0 107352 0.571 1 0.212 -4.71 10. 9.33 3.5 8.1 107388 1.009 14 0.549 1.6 4.8 107576 1.064 12 0.440 ... 1.6 2.1 107820 0.863 1 0.140 -5.16 9.97 2.0 4.3 107956 0.475 1 0.202 -4.73 5. 9.36 3.3 1'1.9 108446 0.872 11 0.302 -4.70 28. 9.29 2.7 4.0 108953 0.800 11 0.179 -4.96 41. 9.76 2.2 2.1 109067 0.405 1 0.164 -4.93 3. 9.72 2.2 3.0 109591 0.666 12 0.148 -5.09 9.91 2.3 2.2 hip61629 1.470 11 1.705 ... 4.6 2.4 bd-043319b 0.980 2 0.308 ... 2.5 2.1 109930 0.894 7 0.202 -4.94 45. 9.74 2.3 2.1 109908 0.583 11 0.268 -4.55 7. 8.93 4.4 15.3 109952 1.154 5 0.623 ... 1.6 2.1 110273 0.538 2 0.142 -5.10 9.90 2.3 2.8 110605 0.730 8 0.171 -4.96 35. 9.77 2.2 2.1 110619 0.664 14 0.169 -4.95 27. 9.75 2.6 2.2 111261B 1.380 2 0.881 ... 2.4 2.2 111261A 1.380 6 0.779 ... 2.1 2.2 111232 0.701 18 0.160 -5.02 9.84 2.1 2.1 111431 0.627 11 0.131 -5.23 10.07 2.0 2.6 112121 0.728 4 0.149 -5.09 9.91 2.1 2.1 112456 0.569 4 0.160 -4.97 16. 9.78 2.6 2.7 113027 0.569 13 0.187 -4.82 12. 9.53 3.0 t;.2 Continued on next page 95 Tablc 1 Contimted from pTevio·us paqe Name B-V JVol" SwKco ]ogR;IK prol log(Agc/ycars) (}~V 1 cr~v ~ 1 (HD) (clays) (m s-1 ) (m s- ) 1134,19 0.847 4 0.581 -4.34 6. 8.27 4.1 1... 9 113538 1.362 11 1.160 3.2 2.2 113693 1.035 5 0.273 1.6 2.1 114386 0.982 7 0.24t1 2.2 2.1 114432 0.756 9 0.3E:7 -4.50 13. 8.77 3.1 10.5 114747 0.923 5 0.283 2.6 2.1 115080 0.683 5 0.156 -5.03 9.86 2.4 2.2 115106 0.660 1 0.134 -5.20 10.04 2.0 2.6 116220 0.667 1 0.138 -5.16 9.99 2.1 2.2 116259 0.721 1 0.141 -5.14 9.95 2.1 2.1 115674 0.667 13 0.186 -4.86 24. 9.62 2.9 2.2 116920 0.911 7 0.251 2.5 2.1 117860 0.624 12 0.281 -4.54 9. 8.90 4.6 13.4 118926 1.376 3 0.928 2.5 2.2 hip66678 1.337 16 0.803 2.3 2.2 119217 1.288 3 0.910 1.6 2.1 119119 0.589 2 0.126 -5.29 10.14 1.9 2.6 119329 0.735 1 0.147 -5.10 9.90 2.1 2.5 119782 0.856 9 0.305 -4.67 27. 9.2'1 2.8 4.4 12003613 1.322 4 1.367 3.9 :) 1 120036A 1.322 8 1.525 4.3 2.1 120056 0.724 3 0.135 -5.19 10.02 2.0 4.3 120250 0.461 5 0.126 -5.27 10.12 1.9 3.0 120329 0.739 5 0.143 -5.13 9.94 2.1 2.6 120744 0.913 7 0.205 2.2 2.1 121504 0.593 20 0.175 -4.89 17. 9.66 2.8 3.6 122245 0.485 1 0.16'1 -4.92 7. 9.71 2.6 4.1 124227 0.529 1 0.172 -4.88 10. 9.64 2.8 4.6 hip69454 1.500 5 0.992 2.9 2.5 12,1553 0.593 8 0.159 -4.98 19. 9.80 2.5 2.6 Contimtcd on ncxt. pagc 96 Table 1 - Continued fmrn previous page N ame E-V Nobs SMIKE log~K P,·ot log(Agejyears) a~v 1 O"~v 2 1 1 (HD) (days) (m s- ) (m s- ) 124595 0.610 5 0.149 -5.06 9.89 2.3 2.5 124925 0.920 2 0.155 ... 2.0 4.3 125595 1.107 5 0.537 ... 1.6 2.1 125906B 0.584 8 0.155 -5.01 9.83 2.5 2.6 125906A 0.584 14 0.156 -5.00 18. 9.82 2.5 2.6 126525 0.682 9 0.157 -5.03 9.85 2.4 2.2 126653 0.574 2 0.164 -4.94 16. 9.74 2.6 2.6 127339 1.403 5 1.144 ... 3.1 2.3 127321 0.615 1 0.130 -5.24 10.09 2.0 2.5 127423 0.569 13 0.225 -4.67 9. 9.23 3.7 :).8 hip71135 1.270 8 0.783 1.6 2.1 128214 0.720 2 0.173 -4.95 33. 9.76 2.2 2.6 128760 0.568 4 0.158 -4.98 16. 9.79 2.5 2.7 128987 0.710 13 0.360 -4.45 9. 8.64 3.1 14.2 129679 0.600 1 0.315 -4.46 6. 8.66 5.2 21.1 129642 0.936 9 0.213 2.2 2.1 129903 0.576 1 0.386 -4.32 2. 8.22 6.5 43.3 129445 0.756 12 0.161 -5.02 9.84 2.1 2.1 129946 0.710 1 0.280 -4.60 16. 9.06 2.7 8.0 130102 0.544 1 0.178 -4.86 11. 9.61 2.9 4.8 130930 0.913 6 0.190 2.2 2.1 131719 1.005 19 0.607 1.6 5.7 131900 0.747 1 0.170 -4.97 37. 9.79 2.2 2.1 hip73362 1.318 1 1.242 ... 3.5 2.1 131664 0.667 8 0.179 -4.90 25. 9.67 2.8 2.2 hip73427 1.253 4 1.202 1.6 2.1 132648 0.721 9 0.191 -4.87 29. 9.63 2.3 2.1 132899 1.145 6 1.032 1.6 5.0 132996 0.613 15 0.154 -5.02 9.84 2.4 :.5 133131A 0.622 13 0.160 -4.99 23. 9.80 2.5 2.4 Continued on next page 97 Tablc 1 - Conl:inued jr·om prcvio·ns page Name B-V iV.,¡,, SWKIO ]ogR;tK P"', log(Agc/ycars) IT~v 1 (J~v 2 (HD) ( clays) (m s-1 ) (m s-1 ) 133131B 0.622 2 0.164 -4.96 22. 9.77 2.6 2.4 134048 0.727 3 0.139 -5.16 9.98 2.0 2.6 133639 0.678 5 0.163 -L1.99 30. 9.81 2.5 2.2 13,166¿1 0.662 9 0.172 -4.93 26. 9.73 2.7 2.3 134928 0.759 2 0.212 -4.82 30. 9.54 2.4 2.1 134929 0.810 1 0.193 -4.91 39. 9.69 2.3 2.1 135446 0.649 7 0.124 -5.31 10.16 1.9 2.6 135309 0.640 3 0.170 -4.94 23. 9.73 2.7 2.4 135725 0.748 6 0.166 -4.99 38. 9.81 2.2 2.1 135562 0.652 1 0.129 -5.25 10.09 1.9 ?6 135625 0.617 3 0.146 -5.09 9.91 2.3 2.5 136762 0.62L1 3 0.236 -4.66 13. 9.21 3.8 2.5 13689Ll 0.721 8 0.173 -4.95 33. 9.75 2.2 2.1 137214 0.593 2 0.149 -5.06 9.88 2.3 2.6 137628 1.010 5 0.348 1.6 2.1 137388 0.891 6 0.265 -4.78 35. 9.47 2.6 2.1 139061 0.412 1 0.148 -5.04 9.87 2.0 3.0 139763 1.296 6 1.405 1.6 2.1 1406,13 0.898 2 0.293 -4.74 32. 9.38 2.7 2.1 141514 0.588 10 0.152 -5.03 9.86 2.4 2.6 141366 0.683 2 0.142 -5.13 9.94 2.2 4.3 1¿11599 0.757 5 0.209 -4.82 30. 9.55 2.4 2.1 141815 0.542 2 0.129 -5.24 10.08 2.1 2.8 1L12072 0.670 12 0.310 -4.51 10. 8.81 5.1 13.1 141885 0.652 2 0.122 -5.34 10.19 1.8 Ll,3 142137 0.629 3 0.142 -5.12 9.92 2.2 4.2 142709 1.118 8 0.419 1.6 2.1 hip78353 1.497 1 1.353 3.8 2.5 143120 0.780 3 0.131 -5.21 10.04 2.0 4.3 1L13137 0.518 2 0.129 -5.23 10.07 2.1 2.9 Coniinuerl on ne:Di page 98 Table 1 - Continued jTom pTevio·us page Name E-V Nobo SMIKE log~K prot log(Agejyears) O"~v 1 a~v2 1 1 (HD) (days) (m s ) (m s- ) 143361 0.773 13 0.167 -5.00 41. 9.82 2.2 L.1 143673 0.802 4 0.163 -5.02 9.85 2.1 2.2 143846 0.600 3 0.165 -4.94 19. 9.75 2.6 2.5 144087 0.750 6 0.313 -4.57 16. 8.97 2.9 8.3 144088 0.850 4 0.3G4 -4.59 21. 9.03 3.0 6.0 144550 0.682 4 0.159 -5.01 9.84 2.4 2.2 144766 0.562 12 0.157 -4.98 15. 9.80 2.5 2.7 144848 0.648 2 0.152 -5.05 9.87 2.4 2.3 142022 0.790 6 0.192 -4.90 37. 9.68 2.3 2.1 144899 0.660 4 0.150 -5.07 9.89 2.3 2.3 145518 0.616 6 0.209 -4.75 15. 9.40 3.4 6.0 145666 0.603 11 0.193 -4.80 15. 9.51 3.1 4.9 146070 0.614 2 0.212 -4.73 14. 9.37 3.4 6.3 146124 0.760 6 0.293 -4.61 19. 9.09 2.8 6.8 146835 0.585 3 0.199 -4.77 13. 9.45 3.2 6.0 146817 0.666 10 0.149 -5.07 9.90 2.3 2.2 hip80018 1.550 5 0.680 ... 2.4 2.5 147018 0.763 9 0.219 -4.80 29. 9.50 2.4 3.6 144167 0.651 4 0.191 -4.84 21. 9.57 3.0 2.3 147873 0.575 4 0.128 -5.26 10.10 2.0 J.6 147619 0.604 1 0.125 -5.31 10.16 1.9 2.5 148156 0.560 2 0.151 -5.03 9.85 2.4 2.7 148303 0.980 1 0.465 3.3 4.5 148290 0.502 1 0.1~8 -5.03 9.86 2.4 3.0 148628 0.538 3 0.139 -5.13 9.94 2.2 2.8 148587 0.573 10 0.139 -5.13 9.95 2.2 2.7 149189 0.653 1 0.136 -5.18 10.01 2.1 2.3 149192 1.096 3 0.799 1.6 4.8 149396 0.704 6 0.257 -4.65 18. 9.19 2.6 6.8 149079 0.576 10 0.142 -5.11 9.90 2.3 2.6 Continued on next page 99 Tablc 1 - Contirmed from previous page N ame B-V JVoh-; sl\liKE logR;m prot log(Agc/ycars) O"~v 1 CT~v 2 1 1 (HD) (clays) (m s- ) (m s- ) 14919,1 0.559 6 0.151 -5.03 9.86 2.4 2.7 149606 0.965 3 0.227 ... 2.2 2.1 151450 0.566 3 0.154 -5.01 9.83 2.5 '2.7 hip82357 1.330 Ll 1.509 4.2 2.1 150761 0.564 1 0.1,15 -5.08 9.90 2.3 2.7 152079 0.711 11 0.164 -4.99 34. 9.81 2.2 2.1 152138 0.490 10 0.2C2 -4.73 5. 9.36 3.3 10.5 152388 0.904 4 0.364 3.0 4.7 153026 1.181 6 0.308 1.6 2.1 154518 1.015 5 0.442 1.6 2.1 154697 0.735 2 0.222 -4.77 26. 9.45 2.4 4.1 154682 0.617 7 0.143 -5.11 9.91 2.2 2.5 154672 0.713 14 0.156 -5.04 9.87 2.1 2.1 hip84452 1.273 2 1.498 ... 1.6 3.6 156152 0.644 6 0.143 -5.12 9.92 2.2 2.3 156239 0.942 2 0.118 1.8 L1.3 156643 0.627 5 0.161 -4.98 23. 9.79 2.5 2.4 157830 0.685 7 0.195 -4.84 24. 9.57 3.1 3.6 hip85523 1.553 4 1.530 4.8 2.5 158469 0.556 6 0.133 -5.19 10.02 2.1 2.7 158630 0.600 9 0.157 -5.00 9.82 2.5 2.5 159902 0.682 1 0.149 -5.08 9.90 2.3 22 160L111 0.655 5 0.122 -5.34 10.19 1.8 3.9 161098 0.676 7 0.163 -4.99 30. 9.81 2.5 2.2 160859 0.617 11 0.170 -4.92 20. 9.72 2.7 2.5 162907 0.758 12 0.159 -5.03 9.86 2.1 2.1 163568 0.556 4 0.152 -5.02 9.85 2.4 2.7 164604 1.396 15 0.547 1.4 2.3 165011 0.576 3 0.141 -5.12 9.93 2.2 2.6 165204 0.770 5 0.160 -5.03 9.86 2.1 2.1 Contirmed on ne:rt page lOO Table 1 - Continued from previo'us page N ame B-V Nobs s~fiKE log R~K pwt log(Agejyears) 0~v1 (7~V2 (HD) (days) (m s-1 ) (rn s-1 ) 165385 0.581 5 0.130 -5.24 10.08 2.0 2.6 165271 0.655 5 0.137 -5.17 10.00 2.1 2.6 165920 0.900 3 0.152 -5.11 9.91 2.0 2.1 166184 1.020 2 0.578 ... 1.6 4.9 165499 0.592 4 0.156 -5.00 19. 9.82 2.5 2.6 166348 1.297 4 1.180 ... 1.6 2.1 166745 0.763 4 0.155 -5.06 9.88 2.1 2.1 168788 0.845 4 0.263 -4.75 31~ . 9~40 T6 . 3.6 169303 0.564 4 0.130 -5.23 10.07 2.0 2.6 170641 0.438 13 0.148 -5.03 9.86 2.2 3.0 169506 0.608 3 0.130 -5.24 10.09 2.0 3.0 172063 0.756 6 0.179 -4.93 36. 9.73 2.2 2.1 172582 0.685 9 0.160 -5.01 9.83 2.5 2.2 173206 0.619 4 0.155 -5.02 9.84 2.4 2.5 171990 0.593 13 0.139 -5.14 9.95 2.2 2.6 173872 0.889 3 0.198 -4.94 45. 9.74 2.2 2.1 174494 0.694 4 0.144 -5.12 9.92 2.2 2.6 174541 0.631 4 0.143 -5.12 9.92 2.2 2.4 174545 0.884 5 0.194 -4.95 45. 9.76 2.2 2.1 175073 0.857 2 0.323 -4.64 25. 9.17 2.9 4.9 175224 1.433 11 1.858 ... 5.0 2.4 175167 0.751 9 0.152 -5.07 9.90 2.1 2.6 176151 0.698 4 0.140 -5.15 9.96 2.1 2.6 176110 0.698 16 0.144 -5.12 9.92 2.2 3.2 176986 0.939 5 0.275 ... 2.5 2.1 176354 0.888 7 0.178 -5.01 9.83 2.1 4.3 177122 0.591 3 0.151 -5.04 9.86 2.4 2.6 175169 0.721 4 0.159 -5.02 9.85 2.1 2.1 178076 0.787 9 0.398 -4.47 12. 8.68 3.3 10.8 178340 0.757 4 0.163 -5.01 9.83 2.2 2.1 Continued on next page 101 Tablc 1 - Contúmed from previmts page Name B-11 J\Tobs sl\IIKE iogR;rK P,·ot log(Agcjycars) (J~v 1 O"~v 2 1 1 (HD) ( clays) (m s- ) (m s- ) 177409 0.600 7 0.166 -4.94 19. 9.74 2.6 2.5 179205 0.579 4 0.266 -4.55 7. 8.94 4.'1 15.4 1796L10 0.738 7 0.300 -4.58 17. 9.01 2.8 8.1 179699 0.596 4 0.131 -5.22 10.06 2.0 2.6 hip94674 1.420 4 0.733 1.9 2.3 180409 0.570 11 0.150 -5.04 9.86 2.4 2.7 180204 0.607 3 0.145 -5.09 9.92 2.3 2.5 180257 0.644 1 0.130 -5.24 10.08 2.0 2.5 181010 0.842 3 0.207 -4.88 39. 9.65 2.3 2.1 182228 0.630 3 O.HS -5.08 9.90 2.3 2.4 181433 1.006 6 0.180 1.6 2.1 182498 0.601 4 0.165 -4.95 19. 9.75 2.6 2.5 183783 1.088 4 0.398 1.6 2.1 183804 0.564 4 0.273 -4.53 6. 8.87 4.5 18.2 184081 0.757 2 0.165 -5.00 9.82 2.2 3.4 184317 0.614 10 0.153 -5.03 9.86 2.4 2.5 185615 0.734 3 0.156 -5.0L1 9.87 2.1 2.1 185283 1.040 6 0.230 1.6 2.1 hip96998 1.347 12 0.903 2.5 2.2 185928 0.947 4 0.128 1.8 4.3 186265 0.790 5 0.146 -5.11 9.91 2.1 2.1 186061 0.990 7 0.328 2.6 2.1 186235 1.509 3 0.410 1.3 2.5 186194 0.697 2 0.150 -5.07 9.90 2.3 2.1 186268 0.468 2 0.174 -4.86 5. 9.61 2.8 2.7 186803 0.689 9 0.254 -4.65 17. 9.18 4.1 7.2 186853 0.669 6 0.200 -4.81 22. 9.52 3.2 4.1 187154 0.624 S 0.288 -4.53 9. 8.86 4.7 14.3 187101 0.584 5 0.141 -5.12 9.93 2.2 2.6 hip98105 1.340 1 0.826 ... 2.3 2.2 Continued on next page 102 Table 1 - Continued fmm previous page Narne E-V Nobs s~JIKE log l?!¡,K pn>t log(Agejyears) O"~v 1 (j~V2 (HD) (days) (rn s-1 ) (rn s-1 ) 187456 1.058 6 0.361 ... 1.6 2.1 188474 1.025 5 0.319 ... 1.6 2.1 188559 1.050 6 0.449 ... 1.6 2.1 190613 0.629 5 0.145 -5.10 9.90 2.2 2.4 190647 0.743 4 0.147 -5.10 9.90 2.1 2.6 190125 0.708 10 0.278 -4.60 16. 9.07 2.7 8.0 189310 0.896 3 0.381 -4.60 23. 9.06 3.1 5.3 ~~ - 191760 0.668 4 0.145 -5.11 9.91 2.2 2.6 hip99764 1.351 1 1.402 ... 3.9 2.2 192961 1.167 5 0.528 ... 1.6 2.1 hip100356 1.594 1 1.446 5.1 2.5 193567 0.788 3 0.158 -5.05 9.87 2.1 2.1 193995 0.716 3 0.140 -5.15 9.96 2.0 2.6 195284 0.709 2 0.301 -4.56 14. 8.95 2.8 9.5 195145 0.738 6 0.178 -4.94 34. 9.73 2.2 2.1 196877 1.324 7 0.767 ... 2.2 2.1 197210 0.711 6 0.183 -4.90 30. 9.68 2.2 2.1 197069 0.579 5 0.168 -4.92 16. 9.72 2.7 2.6 197499 0.576 5 0.148 -5.05 9.88 2.4 2.6 197818 0.620 8 0.159 -4.99 23. 9.81 2.5 2.4 197823 0.759 4 0.222 -4.79 28. 9.48 2.4 3.7 hip103019 1.332 5 1.244 ... 3.5 2.1 199065 0.658 4 0.246 -4.65 15. 9.18 4.0 7.8 199704 0.955 5 0.315 ... 2.7 2.1 200733 0.519 3 0.135 -5.16 9.99 2.2 2.9 200869 0.851 6 0.157 -5.07 9.90 2.1 2.1 201796A 0.654 7 0.313 -4.49 9. 8.77 5.1 14.7 201796B 0.654 7 0.317 -4.49 9. 8.74 5.2 15.2 201757 0.714 6 0.140 -5.14 9.96 2.1 2.6 202041 0.578 3 0.151 -5.03 9.86 2.4 2.6 Continued on next page 103 Tablc 1 - Continued fmm pTevious page 1 Name E-V 1\~¡\¡:; sI'<.IIKE logR;,K prot log(Agc/ycars) (J~v 1 ()~V 2 1 (HD) ( clays) (m s-1 ) (m s- ) 202206 0.714 10 0.214 -'1.78 25. 9.47 2.4 4.1 202457 0.689 6 0.146 -5.10 9.90 2.2 2.1 202746 1.010 4 0.998 ... 1.6 11.9 198477 0.900 6 0.227 -4.88 41. 9.64 2.4 2.1 203432 0.741 6 0.186 -4.90 33. 9.68 2.3 2.1 203,113 1.063 5 0.545 ... 1.6 2.1 203850 0.924 Ll 0.243 ... 2.4 2.1 204313 0.697 8 0.161 -5.00 9.82 2.5 2.1 204807 0.717 2 0.197 -4.85 28. 9.59 2.3 2.1 204941 0.878 4 0.218 -4.88 40. 9.65 2.3 2.1 205067 0.656 4 0.158 -5.01 9.83 2.4 2.3 205045 0.591 6 0.137 -5.16 9.99 2.1 2.6 205158 0.596 13 0.130 -5.24 10.09 2.0 2.6 205310 1.379 3 0.856 ... 2.3 2.2 205739 0.546 15 0.139 -5.12 9.93 2.2 2.8 205860 0.500 4 0.142 -5.09 9.91 2.3 3.0 206025 0.555 14 0.134 -5.18 10.01 2.1 2.7 206255 0.727 6 0.131 -5.22 10.06 2.0 L1.3 206683 0.657 7 0.150 -~.07 9.89 2.3 2.3 207315 0.401 2 0.170 -4.89 3. 9.66 2.2 3.0 hip107772 1.328 1 0.848 ... 2.4 2.1 207970 0.733 8 0.155 -5.05 9.88 2.1 2.1 207496 1.000 5 0.603 ... 3.9 5.9 208573 1.035 3 0.475 1.6 2.1 208704 0.640 3 0.152 -5.05 9.88 2.3 2.4 209566 0.758 5 0.156 -5.05 9.88 2.1 2.1 209659 0.609 6 0.142 -5.12 9.93 2.2 2.6 209913 0.763 6 0.167 -4.99 40. 9.81 2.2 2.1 210193 0.660 8 0.172 -4.93 25. 9.73 2.7 2.3 211366 0.659 10 0.172 -4.93 25. 9.72 2.7 2.3 Continv.ed on next page 10¿1 Table 1 - Continued from previous page N ame prot log(Age/years) B-V N""' s~!IKE logR~K a~v 1 J~V2 1 (HD) (days) (m s-1 ) (m s- ) 211369 0.961 5 0.463 ... 3.3 5.0 211970 1.329 8 1.184 ... 3.3 2.1 hip110655 1.312 3 1.282 3.7 2.1 213241 0.463 5 0.149 -5.03 9.85 2.3 3.0 213717 0.817 9 0.189 -4.93 40. 9.72 2.3 2.1 213401 0.669 2 0.151 -5.06 9.89 2.3 2.5 213941 0.670 4 0.178 -4.91 26. 9.69 2.8 2.2 214100 1.411 3 0.831 2.2 2.3 214385 0.640 5 0.155 -5.03 9.85 2.4 2.4 214691 0.735 5 0.149 -5.08 9.91 2.1 2.1 215456 0.636 7 0.133 -5.20 10.04 2.0 2.4 216054 0.741 8 0.176 -4.94 35. 9.74 2.2 2.1 216133 1.452 5 1.350 ... 3.6 2.4 hipll3850 1.643 2 1.326 3.6 2.5 218511 1.201 7 1.208 ... 1.6 4.3 218572 1.004 5 0.461 1.6 3.9 218760 0.966 2 0.321 ... 2.6 2.1 219011 0.727 6 0.149 -5.09 9.91 2.1 2.1 hip114719 1.449 4 1.325 3.6 :¿_4 219533 0.684 3 0.138 -5.16 9.98 2.1 2.4 219556 0.759 3 0.158 -5.04 9.86 2.1 2.1 219495 1.130 2 0.863 ... 1.6 4.3 219509 1.049 2 0.311 1.6 2.1 219764 1.154 12 1.089 ... 1.6 5.1 220256 0.853 4 0.184 -4.96 45. 9.77 2.2 2.1 220426 0.581 4 0.146 -5.07 9.90 2.3 2.6 220476 0.682 5 0.367 -4.42 7. 8.54 6.1 17.9 220689 0.603 5 0.159 -4.99 20. 9.81 2.5 2.5 hip115752 1.272 2 0.809 ... 1.6 2.1 220718 0.651 5 0.132 -5.22 10.05 2.0 2.6 Continued on next page 105 Tablc 1- Cont'irmed from previo·as page Name B-11 JV.,¡,, S,liKE logR~K P, ..,, log(Agcjycars) (J~v 1 O"~y 2 1 1 (HD) (clays) (m s- ) (m s- ) 220829 0.906 3 0.166 2.1 2.1 220945 0.934 7 0.303 2.7 2.1 220981 0.740 5 0.142 -5.1L1 9.95 2.1 2.1 hip115955 1.316 2 1.633 4.6 2.1 221257 0.636 3 0.166 -4.96 23. 9.76 2.6 2.4 221275 0.799 2 0.236 -4.77 30. 9.45 2.5 2.1 hip116491 1.370 10 1.490 4.0 2.2 221954 0.750 4 0.145 -5.11 9.92 2.1 2.6 222422 0.731 4 0.237 -4.72 23. 9.35 2.5 4.9 222743 0.559 4 0.153 -5.02 9.84 2.5 2.7 222669 0.608 3 0.166 -4.94 20. 9.75 2.6 2.5 223957 0.607 4 0.205 -4.76 14. 9.42 3.3 4.3 224010 0.628 6 0.146 -5.09 9.91 2.3 2.4 224022 0.572 9 0.147 -5.06 9.89 2.3 2.7 224143 0.640 5 0.278 -4.56 11. 8.95 4.6 11.9 224228 0.973 5 0.640 4.1 7.6 224376 0.486 3 0.186 -4.80 6. 9.50 3.0 7.5 224433 0.748 5 0.167 -4.99 38. 9.80 2.2 2.1 224464 0.465 2 0.158 -4.96 6. 9.77 2.5 3.0 224538 0.581 4 0.153 -5.02 9.84 2.4 2.6 224607 1.042 8 0.640 1.6 4.9 106