188 Host stars Table 8.1: An incomplete list of some of the larger compilations 3.8 of [Fe/H] relevant to exoplanet investigations, mostly for FGK H 74 dwarfs. The compilations often provide other parameters such GCC 96 as logg and Teff. N is the total sample size, Np is the number g BL 86 4.0 of planet stars hosts when quoted. References are for the latest SS 73 descriptions in the case of progressively enlarged samples. E 93 4.2 N Np Reference B 90 3356 – Cayrel de Strobel et al. (2001) G 97 14 000 – Nordström et al. (2004b) Surface gravity, log Surface gravity, 4.4 1040 99 Valenti & Fischer (2005) 160 27 Takeda & Honda (2005) 4.6 TF 83 υ And 100 000 – Ammons et al. (2006) 216 55 Luck & Heiter (2006) 6400 6200 6000 5800 1907 – Robinson et al. (2007) Effective temperature,T eff 118 28 Bond et al. (2008) 451 – Sousa et al. (2008) υ Figure 8.6: logg versus Teff for And. The black circle, with error bars, marks the spectroscopically derived values. The hosting planets have significantly higher metal con- inclined greyscale bar indicates the most probable parameter tent than the average solar-type star in the solar neigh- space allowed by the accurate Hipparcos distance. Grey cir- cles represent results from earlier work, with diameters propor- bourhood (Gonzalez, 1997). While the Sun and other tional to derived metallicity. A systematic shift of the bar as a nearby solar-type dwarfs have [Fe/H] ∼ 0 (Reid, 2002), result of different metallicity scales is indicated by the vertical typical exoplanet host stars have [Fe/H] ∼> 0.15. Values arrow for a decrease in [Fe/H] by 0.1 dex. From Fuhrmann et al. of [Fe/H] = +0.45 for two early discoveries, 55 Cnc and (1998, Figure 2), reproduced with permission © ESO. 14 Her, placed them amongst the most metal-rich stars in the solar neighbourhood (Gonzalez & Laws, 2000). ing rapidly evaporated due to a high ultraviolet flux of Although planets around even very low-metallicity their (presumably young and hot) host stars. From their stars have since been found, an overall correlation be- chromospheric activity indices, none were found to be tween metallicity and planet occurrence has been con- younger than 0.5 Gyr. firmed by subsequent work, using different samples and different analysis procedures. 8.4 Element abundances Comparison stars Consistent agreement in determin- Chemical abundance analysis, using high-resolution ing effective temperatures and metallicities has proven high signal-to-noise spectroscopy, provides a funda- notoriously difficult. Published results for a given star mental diagnostic of host star properties, and an impor- frequently formally disagree, as a combined result of dif- tant if indirect probe of planetary formation and subse- fering spectral resolution, the choice of spectral lines, quent evolution. analysis procedures, and the adopted scales of metallic- υ ity and Teff. For the case of And shown in Figure 8.6, for example, nine publications pre-2000 give values span- 8.4.1 Metallicity =− = ning the range [Fe/H] 0.23, Teff 6000 K (Hearnshaw, =+ = An important aspact of a star’s chemical composition is 1974) to [Fe/H] 0.17, Teff 6250 K (Gonzalez, 1997). the fraction of metals (i.e. elements heavier than He in To establish statistical differences between stars with astronomy usage). Iron abundance, expressed as [Fe/H], and without planets at the level of 0.1–0.2 dex, a secure is frequently used as the reference element for exoplanet sample of comparison stars is required. The comparison host star studies (e.g. Fuhrmann et al., 1997, 1998; Gon- sample should be demonstrably companion-free, and zalez, 1997, 1998; Gonzalez et al., 1999; Gonzalez & Laws, analyses for both samples should preferably be based on 2000; Giménez, 2000; Murray et al., 2001; Santos et al., the same sets of spectroscopic lines, observed and anal- 2001; Murray & Chaboyer, 2002; Laws et al., 2003; Santos ysed in the same way. A number of such spectroscopic et al., 2003, 2004c, 2005, and others). host star samples and comparison sets have now been © Perryman, Michael, May 31, 2011, The Exoplanet Handbook Cambridge University Press, Cambridge, ISBN: 9781139082549 The abundances of other elements are providing constructed and investigated. increasingly valuable diagnostics (e.g. Gonzalez, 1998; Santos et al. (2001) first presented a spectroscopic Gonzalez & Vanture, 1998; Gonzalez & Laws, 2000; San- study of a volume-defined set of 43 F8–M1 stars within tos et al., 2000; Gonzalez et al., 2001b; Smith et al., 2001; 50 pc included in the CORALIE programme, and for Sadakane et al., 2002; Zhao et al., 2002; Bodaghee et al., which constant radial velocities over a long time interval 2003; Ecuvillon et al., 2004a,b, and others). provided evidence that the comparison stars are planet- Already from the earliest studies of just four systems free. A further 54 comparison stars were added by Santos (51 Peg, 55 Cnc, υ And, and τ Boo) it appeared that stars et al. (2005), yielding two large and uniform samples of 8.4 Element abundances 189 28 (2005), who gave the incidence of Doppler-detected 1.0 planet giant planets as < 3% for [Fe/H] <−0.5, and 25% for 0.8 hosts 24 [Fe/H] >+0.5. Over the range −0.5 < [Fe/H] < 0.5, and 0.6 for FGK-type main-sequence stars, they expressed the 0.4 probability of formation of a gas giant planet, with or- 20 0.2 Cumulative fraction Cumulative − bital period shorter than 4 yr and K > 30ms 1,as 0.0 16 –1 –0.5 0 0.5 2.0[Fe/H] Number [Fe/H] P(planet) = 0.03 × 10 N /N 2 12 = 0.03 Fe H , (8.16) comparison (N /N ) stars Fe H 8 the second expression following from the definition of [Fe/H] (box on page 186). 4 As discussed further below, the correlation between occurrence and metallicity may not extend to giant stars, 0 to stars of intermediate metallicity, to M dwarfs, or to the –1 –0.5 0 0.5 Metallicity, [Fe/H] occurrence of low-mass planets. Transiting planets The correlation between metallic- Figure 8.7: Metallicity distribution for 119 planet-host stars (shown as the dashed line, shaded), and for a volume-limited ity and occurrence appears to extend to the close-in comparison sample of 94 stars with no known planets (contin- giant planets discovered by transit photometry. uous line, unshaded). The average metallicity difference of the To explain the observed radius anomalies for tran- two samples is 0.24 dex. Inset: cumulative distribution func- siting planets known at the time (including HD 209458 tions. A statistical Kolmogorov–Smirnov test shows that the and OGLE–TR–10 considered to be anomalously large, probability that both distributions belong to the same popula- − and HD 149026 considered to be anomalously small), tion is ∼ 10 12. From Santos et al. (2005, Figure 1), reproduced Guillot et al. (2006) suggested an exoplanet composi- with permission © ESO. tion/evolution model which included an additional in- ternal energy source equal to 0.5% of the incoming stel- 119 planet-host stars, and 94 stars without known plan- lar luminosity. This additional heat source acts to slow ets, all of which have accurate stellar parameters and the cooling of the planet. [Fe/H] estimates. These samples have been the basis of With this adjustment to bring the radii into bet- various subsequent abundance analyses (Santos et al., ter consistency with theoretical models, they showed 2003; Bodaghee et al., 2003; Santos et al., 2004c; Israelian that for the nine transiting planets known at the time, et al., 2004; Gilli et al., 2006). A further 64 comparison the amount of heavy elements that had to be added stars were added by Sousa et al. (2006). to match their observed radii was a steep function of Other large uniform spectroscopic surveys of exo- the host star metallicity: from less than 20M⊕ of heavy planet host stars and comparison stars include (see also elements around stars of solar composition, to up to Table 8.1): the 99 planet host stars from the 1040 FGK 100M⊕ for stars with three times the solar metallicity dwarfs of the Keck, Lick, and AAT programme, selected (Figure 8.8). These results add to the picture that heavy according to magnitude, colour, and luminosity (Valenti elements play a key role in the formation of close-in & Fischer, 2005); the 27 planet host star and 133 com- giant planets. parison stars observed at Okayama (Takeda et al., 2005; A uniform determination of spectroscopic parame- Takeda & Honda, 2005); the 28 planet host stars and ters for 13 host stars of transiting planets was made by 90 comparison stars from the Anglo–Australian planet Ammler-von Eiff et al. (2009), and supplemented by a search programme (Bond et al., 2006, 2008); and the 216 compilation of results for a total of 50 transit host stars. star sample of the ‘nearby stars project’ (Heiter & Luck, A systematic offset in the abundance scale was found for 2003; Luck & Heiter, 2005, 2006). the TrES and HAT objects. Luck & Heiter (2006) detail the overlap between these and other samples, including the extensive Giant stars Pasquini et al. (2007) found that planet © Perryman, Michael, May 31, 2011, The Exoplanet Handbook Cambridge University Press, Cambridge, ISBN: 9781139082549 Geneva–Copenhagen spectroscopic and kinematic sur- occurrence around a sample of 14 giant stars does not vey (Nordström et al., 2004a,b). correlate with increasing metallicity, in contrast with main sequence stars. While they favoured an expla- Occurrence versus metallicity A comparison based nation based on the accretion of metal-rich material on the host star and control samples of Santos et al.