References: Sozzetti, A., et al. 2007, ApJ, 664, 1190 Mandushev, G., et al. 2007, ApJ, 667, L195 Sozzetti, A., et al. 2007a, ApJ, in preparation Sozzetti, A., et al. 2007b, ApJ, in preparation Kovacs, G., et al. 2007, ApJL, in press (arXiv ) Observational Tests of Planet Formation Models A. Sozzetti 1,2, D. W. Latham 1, G. Torres 1 B. W. Carney 3, J. B. Laird 4, R. P. Stefanik 1, A. Sozzetti 1,2, D. W. Latham 1, G. Torres 1, B. W. Carney 3, J. B. Laird 4, R. P. Stefanik 1, A. P. Boss 5, D. Charbonneau 1, F. T. O’Donovan 6, M. J. Holman 1, J. N. Winn 7 1) CfA; 2) INAF-OATO; 3) UNC; 4) BGSU, 5) CIW 6) Caltech, 7) CIW The Planet-Metallicity Connection I The Planet-Metallicity Connection II The planet-metallicity connection is one of the most important aspects of the close relationship between characteristics and frequencies of planetary systems and the physical properties of the host stars which have been unveiled by the present sample of over 200 extrasolar planets. In particular, the likelihood of finding a planet around a given star rises sharply with stellar metallicity. Furthermore, a correlation may alsoexist between estimated inner core masses of transiting giant planets and the hosts' metal content. In both cases, the evidence collected so far appears to strongly support the orthodox mechanism of giant planet formation by core accretion, as opposed to the heretic formation mode by disk instability. However, the relatively small numbers of metal-poor stars screened for planets so far, and the large uncertainties often present in the determination of both planet and stellar properties in transiting systems prevent one from drawing conclusions. We will describe two experiments designed to put the observed trends on firmer observational grounds, thus ultimately helping to discriminate between proposed planet formation models. First, we will present results from a Doppler survey for giant planets orbiting within 2 AU of a well-defined sample of 200 field metal-poor dwarfs. Our data will crucially help to gauge the behavior of planet frequency in the metal-poor regime. Then, I will describe a novel method for improving on the knowledge of stellar and planetary parameters of transiting systems through a careful analysis of spectro-photometric measurements. With this approach, structural and evolutionary models of irradiated planets can be better informed, allowing for refined estimates of the heavy-element content of transiting planets and for improved understanding ot the core mass - stellar metallicity correlation. Abstract The M c – [Fe/H] Connection ? Burrows et al. (ApJ, 2007): “The core mass of transiting planets scales linearly (or more) with [Fe/H]” Guillot et al. (A&A, 2006): “The heavy element content of transiting extrasolar planets should be a steep function of stellar metallicity” ? Do inferred exoplanets core masses depend on metallicity? The M p -R p Relation Roughly OK Very large core? Coreless?? Default models have trouble! Transiting planets come in many flavors What are their actual interiors? How did they form? Improving Stellar and Planetary Parameters Must determine R *, M *, to derive M p, R pMust determine R *, M *, to derive M p, R p R *, M * are inferred by comparison with stellar evolution modelsR *, M * are inferred by comparison with stellar evolution models Use T eff and a proxy for L *, such as log(g)Use T eff and a proxy for L *, such as log(g) But…But… 1)T eff must be reliable: check relative agreement of multiple methods 2)Log(g) is usually not known precisely enough: use a new method based on observables from the light-curve A Better Proxy: a/R * Main adjustable light-curve parameters: ρ*ρ*ρ*ρ* X X Theoretical values of a/R * are compared with the results from the light-curve fit Results We have determined precisely the stellar and planetary properties of a number of transiting systems (TrES-2, TrES-3, TrES-4), using a combination of high-quality spectroscopic and photometric dataWe have determined precisely the stellar and planetary properties of a number of transiting systems (TrES-2, TrES-3, TrES-4), using a combination of high-quality spectroscopic and photometric data New approach: compare theoretical isochrones with a reliable spectroscopic T eff and the photometric a/R *, rather than the spectroscopic log(g).New approach: compare theoretical isochrones with a reliable spectroscopic T eff and the photometric a/R *, rather than the spectroscopic log(g). As a result, uncertainties in stellar and planetary parameters (R *, M *, R p, M p, log(g p )) are improved by factors of 2-5As a result, uncertainties in stellar and planetary parameters (R *, M *, R p, M p, log(g p )) are improved by factors of 2-5 The comparison between the measured planet and host’s parameters for these systems complicate the picture for planet structure theories. In particular, metal-rich stars appear to be orbited by hot Jupiters whose radii can be both small (requiring large core masses) as well as quite inflated (consistent with coreless models). At present, at least three systems (HAT-P-4, WASP-1, and TrES-4) do not seem to support the existence of a simple relation between host star metallicity and planet’s core mass.The comparison between the measured planet and host’s parameters for these systems complicate the picture for planet structure theories. In particular, metal-rich stars appear to be orbited by hot Jupiters whose radii can be both small (requiring large core masses) as well as quite inflated (consistent with coreless models). At present, at least three systems (HAT-P-4, WASP-1, and TrES-4) do not seem to support the existence of a simple relation between host star metallicity and planet’s core mass. This method is being applied to the other transiting systems, to best inform structural and evolutionary (and ultimately formation) modelsThis method is being applied to the other transiting systems, to best inform structural and evolutionary (and ultimately formation) models How do You Go About it? The star-planet interplay is complexThe star-planet interplay is complex The parameter space is largeThe parameter space is large Uncertainties are no lessUncertainties are no less Determine stellar and planetary properties as best as you can! From top to bottom, metallicities for the parent stars are: 0.23, 0.24, 0.02, 0.14, and The f p – [Fe/H] Relation I Ida & Lin (ApJ, 2004), Kornet et al. (A&A, 2006): “The probability of forming gas giant planets by core accretion is roughly a linear function of Z” Boss (ApJL, 2002): “The probability of forming gas giant planets by disk instability is remarkably insensitive to Z” N/A Do giant planets form by Core Accretion, Disk Instability, or both? F p ~ Z, for Z > 0.02) F p ~ const, for Z < 0.02 Linear Dependence? Flat tail for [Fe/H] < 0.0? Low statistics for [Fe/H] < -0.5 Santos et al. (A&A, 2004): No P, K, [Fe/H] thresholds: Fischer & Valenti (ApJ, 2005): K > 30 m/s, P 30 m/s, P < 4 yr, -0.5<[Fe/H]<0.5: Quadratic dependence? Flat tail for [Fe/H]<0.0? Low statistics for [Fe/H] < -0.5 The f p – [Fe/H] Relation II Keck/HIRES Metal-Poor Planet Search 200 stars from the Carney-Latham and Ryan samples200 stars from the Carney-Latham and Ryan samples No close stellar companionsNo close stellar companions Cut-offs: -2.0 < [Fe/H] < -0.6, T eff < 6000 K, V < 12Cut-offs: -2.0 < [Fe/H] < -0.6, T eff < 6000 K, V < 12 Reconnaissance for gas giant planets within 2 AUReconnaissance for gas giant planets within 2 AU Campaign duration: 3 yearsCampaign duration: 3 years Typical radial-velocity rms: ~ 9 m/sTypical radial-velocity rms: ~ 9 m/s No clear RV trends are seen as a function of V, T eff, [Fe/H], and ΔT Results COMPLETENESS: - 6 observations, 3-yr baseline; - s RV = 9 m/s % confidence level - Sensitivity to companions: 1M J 100 m/s), with orbital periods between a few days and 3 years - Strong dependence of detection thresholds on eccentricity WE FIND NONE… Sozzetti et al. 2007a (in prep.): K > 100 m/s, P 100 m/s, P < 3 yr, -1.0<[Fe/H]<0.5: We observe a dearth of gas giant planets (K > 100 m/s) within 2 AU of metal-poor stars (-2.0 < [Fe/H] < -0.6), confirming and extending previous findings The resulting average planet frequency is F p < 0.67% (1σ) F p ( ), but it’s indistinguishable from F p (-0.5<[Fe/H]<0.0) Is F p ([Fe/H]) bimodal or not? It is consistent with being so. However, need larger and better statistics to really discriminate… 1) Expand the sample size; 2) lower the mass sensitivity threshold; 3) search at longer periods. Next generation RV surveys and future high-precision space-borne astrometric observatories (SIM, Gaia) can help!