arXiv:0705.3228v1 [astro-ph] 22 May 2007 G M NVRIY LUDWIG-MAXIMILIANS-UNIVERSIT JOINT STANFORD A AUSTIN, AT UNIVERSITY, TEXAS IS STATE UNIVERSITY, OF PENNSYLVANIA UNIVERSITY WHICH THE THE OF specifically TELESCOPE, PROJECT to first HOBBY-EBERLY the was sur- (2006) ∗ radial-velocity al. The et Sozzetti form. around of to vey able lowest are the planets find which forma- to planet need of We around impor- physics the tion. extremely planets of is understanding our for metallicity for low tant searching relatively Thus, with stars ob- likely 2006). most the is (Gonzalez explanation for primordial a stellar responsible correlation; served post-formation is there that self-enrichment far, evidence So photospheric compelling formation. little planet of is H- detritus for (or He-poor) metal-rich searching the and from by stellar result might planet-host addressed that in been enhancement elemental has of issue patterns This environ- of metal-rich physics ment? a favor the pol- significantly does planets formation in- or planet metallicity, apparent of photospheric an presence on causing in observed region the crease concentrated convective the does stellar area in the is, lute this causality That in of 2007). correlation. work direction Lineweaver early the & 2005; determining the Grether al. et of Santos 2005; Much 2003; Luck Valenti & & samples Fischer Heiter control 2002; selected Reid sev- carefully investigations, (e.g. included of which subsequently number of has large eral finding a dif- by This without significant supported a and samples. been was these with there in that stars ference conclude comparison of to detailed and abundances found a planets, chemical do are to the first (1997, of stars the Gonzalez was sequence 1999) stars. 1998a,b, main metal-rich around to preferentially companions etary lcrncades [email protected] address: Electronic AE NOSRAIN BANDWT THE WITH OBTAINED OBSERVATIONS ON BASED OTTINGEN. ¨ NHN N GEORG-AUGUST-UNIVERSIT AND UNCHEN, ¨ hr snwsrnl opligeiec htplan- that evidence compelling strongly now is There oApa nTeAtohsclJunl etme 07issue 2007 September 1 L Journal, using Astrophysical typeset The Preprint in Appear To w lnt aeobtlproso 9. n 3. as iheccen M with 0.50 days, and 530.3 0.89 and are 195.0 planets these of for periods orbital minimum have planets two 70yas hl ag netite eani h ria eccentric orbital stab the dynamically in be would remain orbits derived uncertainties periast our large plane of that arguments the indicate While and calculations that eccentricities massive, . their sufficiently with 2700 are other, planets each the with and other, each otn forSnwssilal ofr ytmo tlattoJov two least at of AU. system 0.6–1.2 a of wit headings: form axes Subject to a semi-major able Thus, to still plane evolve found. was giant orbits yet K1III star the our host with of planet tied content any is of 155358 HD metallicity -0.68, lowest of [Fe/H] metallicity a erpr h eeto ftopaeayms opnost h s the to companions planetary two of detection the report We ila .Ccrn ihe nl oetA itnyradJ and Wittenmyer A. Robert Endl, Michael Cochran, D. William 1. A INTRODUCTION T E tl mltajv 08/22/09 v. emulateapj style X coadOsraoy h nvriyo ea tAsi,Au Austin, at Texas of University The Observatory, McDonald lntr ytm tr:idvda H 538 ehius r techniques: — 155358) (HD individual stars: — systems planetary H OETMTLIIYPAE OTSTAR HOST PLANET METALLICITY LOWEST THE oApa nTeAtohsclJunl etme 07iss 2007 September 1 Journal, Astrophysical The in Appear To LNTR YTMAON D155358: HD AROUND SYSTEM PLANETARY A ABSTRACT AT AT ¨ ¨ ewe h w Csflsa 5936 at falls CCDs two the between eouineeet h rs-ipre sstt obtain to 4076 set between is spectrum cross-disperser the The 2048 two element. of resolution mosaic a on recorded is Cs hc ape h pcrmat spectrum the samples which CCDs, 0.Tednenro I of narrow temperature dense molec- a at a The stabilized through 70C. cell passes absorption then iodine Starlight ular telescope. the of tr(ul19)o h . ob-bryTelescope Hobby-Eberly 400 m A 9.2 2 tends 1998). Spectrom- the al. Resolution of et (Ramsey High 1998) (Tull the eter using mode scheduled this of implications 5. given host the are Section theories the and in formation 4, of planet for Section parameters system in fascinating Detailed ex- given is are 3.4. solution star dy- Section orbital The preferred in com- our plored a solutions. achieve of orbital to de- stability the possible algorithm 3.3 namical of for Section genetic exploration a solutions 3.2. of plete and Orbital use 3.1 observa- our Section 2. the scribes in Section given of are in Details planets given planets. are dis- to with tions element begin stars heavy to the of order of HET tribution in tail our low-metallicity 2004) in the al. explore sample et included the (Cochran were of survey one that planet is stars This low-metallicity 155358. HD of star the to panions metallicity. on formation and eovn oe of power resolving t esrmn Bte ta.19;Ed ta.20) A 2000). al. et Endl veloc- 1996; radial al. 250 precise the et for (Butler model 1995) measurement metric ity al. to velocity et the us (Valenti provides enables function and spread spectrum point stellar instrumental the on posed agtlwmtliiysasi re oudrtn nde- in understand to low- order the in tail stars metallicity low target l bevtoso D155 eemd nqueue in made were 155358 HD of observations All com- planetary two of discovery the here present We J µ epciey h risaecoeeog to enough close are orbits The respectively. pcrgahetac ltte ie spectral a gives then slit entrance spectrograph m ′′ . ntesytak h trars h oa plane focal the across star the tracks sky the on 0 Z o ayn ihproso 2300– of periods with varying ron n ftedpnec fpaeaysystem planetary of dependence the of end saegaiainlyinteracting gravitationally are ts ny2%o h heavy-element the of 21% only h a-aspaesadhv their have and planets ian-mass rcte f01 n .8 The 0.18. and 0.11 of tricities lrtp trH 538 The 155358. HD star olar-type efra es 10 least at for le tnT 78712 TX stin 2. te,orobtlintegration orbital our ities, R otsa D456frthe for 47536 HD star host t OBSERVATIONS = ue λ/δλ cbL Bean L. acob 2 ∗ pcrmwihi superim- is which spectrum n 7838 and A ˚ µ pia brwihsub- which fiber optical m 60 = da velocities adial 8 , er.With years. ,alwn otof most allowing A, ˚ 0.Tespectrum The 000. × .Asalgap small A A. ˚ 10pxlE2V pixel 4100 ≈ iesper pixels 4 2 Cochran et al. the I2 absorption spectrum to be recorded on the “blue” CCD chip. TABLE 1 Relative Velocities for We obtained a total of 71 high signal/noise radial ve- HD 155358 locity observations of HD155358 between June 2001 and March 2007. All spectra were processed using standard BJD Velocity σ IRAF routines, and velocities were computed using our -2 400 000 m s−1 ms−1 high-precision code. Table 1 gives the rel- 52071.904837 19.49 3.28 ative radial velocities for HD 155358. The observation 52075.886827 24.07 3.17 times and velocities have been corrected to the solar sys- 52076.889856 25.29 2.93 tem barycenter. The uncertainty σ for each velocity in 52091.846435 31.10 3.08 52422.937143 3.70 3.43 the table is an internal error computed from the vari- 53189.839130 -30.67 3.28 ance about the mean of the velocities from each of the 53205.795629 -20.54 3.08 480 small chunks into which the spectrum is divided for 53219.753215 -16.15 3.35 the velocity computation. Thus, it represents the relative 53498.776977 24.06 3.72 53507.957634 41.99 4.30 uncertainty of one velocity measurement with respect to 53507.962870 41.12 3.81 the others, based on the quality and observing conditions 53511.953195 41.06 2.88 of the spectrum. This uncertainty does not include other 53512.939314 34.47 3.29 53590.733229 -13.27 2.86 intrinsic stellar sources of uncertainty, nor any unidenti- 53601.700903 -3.25 2.77 fied sources of systematic errors. 53604.701932 -3.05 2.97 53606.702997 -5.41 3.20 3. ORBITAL SOLUTION 53612.684038 3.44 2.61 3.1. 53625.645464 29.05 2.65 The First Planet 53628.624309 28.41 3.61 A periodogram of the HET velocities of HD 155358 53629.619464 23.41 2.85 53633.616275 20.48 2.83 show a very strong peak at ∼195 days, with substantial 53755.042134 -35.55 4.38 additional power in the 300-500 range. The Scargle 53758.042744 -41.84 3.70 (1982) false-alarm probability of this 195-day signal is 53765.035558 -30.94 4.02 −10 53774.021743 -45.17 4.33 4.5 × 10 . We first fit a Keplerian orbit to the data 53779.979871 -26.31 4.27 using the GaussFit generalized least squares software 53805.930694 -5.23 3.16 of Jefferys et al. (1988). This planet, with a period of 53808.892822 -4.14 4.28 −1 53769.026048 -42.02 3.56 193.8 d, eccentricity of 0.16, and K velocity of 31.0 m s 53866.964277 30.26 3.87 has a minimum mass of M sin i = 0.79 MJ and a semi- 53869.954988 29.05 3.55 major axis of 0.63AU. The RMS dispersion of the data 53881.945116 43.83 2.66 around the orbital solution is 10.2 m s−1, and the reduced 53889.682359 47.51 2.83 2 53894.672403 42.03 2.80 chi-squared χν =9.67. This goodness-of-fit parameter is 53897.895406 34.74 2.39 significantly higher than would be expected based, as it 53898.685344 27.26 2.75 is based purely on the internal errors quoted in Table 1. 53899.892253 36.21 2.91 Wright (2005) gives a median “jitter” of 4.4 m s−1 for 53902.657932 17.67 2.53 53903.906452 35.85 2.87 main-sequence stars similar to HD 155358. Wright also 53904.662338 26.89 3.03 mentions that subgiants typically have jitters of about 53905.860234 27.26 3.10 5ms−1, as do blue stars (with B-V < 0.6). HD 155358 53907.629415 28.25 2.75 appears to fall into both categories (cf. section 4, with 53908.876910 23.17 3.05 53910.639802 26.13 3.61 our derived age of 10 Gyr and Hipparcos B-V=0.545. 53911.845252 17.72 2.70 While some of the additional scatter might be explained 53912.636138 15.77 3.03 as intrinsic stellar velocity variability expected in this 53917.641443 9.13 3.12 53924.818858 14.17 3.21 star, there is probably some additional source for much 53925.807994 10.44 2.60 of the scatter. Thus, we have searched carefully for ad- 53926.824657 -0.29 2.83 ditional planets in the system. 53927.822039 9.12 2.98 53936.790961 -7.51 2.98 3.2. The Second Planet 53937.790352 -12.14 6.01 53937.804652 -0.17 2.83 A Lomb-Scargle periodogram of the residuals around 53941.762397 -14.31 2.78 the single-planet orbital fit shows very significant power 53943.766914 -11.76 2.83 53954.756895 -10.92 2.49 over a broad range of periods from 200 to 800 days, with 53956.734767 -15.38 2.47 the strongest peak around 530days. The false-alarm 53958.747552 -11.09 2.50 probability of this power at 530 days is less than 10−6. 53960.714843 -9.99 2.52 Smaller peaks around 330 days show FAP slightly less 53966.708140 -11.85 2.34 −3 53971.697987 -12.96 2.51 than 10 . Thus, we investigated the possibility of a 53985.651466 12.67 3.81 second long-period planet HD155358c. We used Gauss- 53988.637456 6.17 2.83 Fit to compute simultaneous least-squares solutions of 53990.639454 19.80 3.29 53993.644925 20.57 3.27 double-Keplerian orbits to the observed velocities. We 54136.010987 -17.45 4.47 investigated periods around both 330 days and 530 days. 54137.005715 -9.57 3.98 While we were able to find formal Keplerian solutions for 54165.933759 -30.96 3.73 54167.924067 -24.81 3.85 both periods, the solutions with Pc ∼ 530days gave sig- nificantly better fits than did shorter period second plan- Planetary System Around HD155358 3

to survive and to multiply. Solutions that are above this TABLE 2 2 2 threshold (e.g. χ < χbest + 0.1%), and are not able HD155358 Double Planet 2 Keplerian Solution to move back into this χ range, are terminated after a few generations (typically 6 generations). Repeating this Parameter Value process many times (each time with new randomly se-

Pb 195.0±1.1 days lected starting values) allowed us to probe a much larger T0b 2453950.0±10.4 BJD parameter space than is usually possible with standard −1 Kb 34.6±3.0 ms orbital fit programs. eb 0.112±0.037 ± For HD155358 we performed > 10, 000 trial runs, al- ωb 162 20 degrees lowing the period of HD155358c to vary between 250 Pc 530.3±27.2 days T0c 2454420.3±79.3 BJD and 2100 days (the time span of observations). We con- −1 Kc 14.1±1.6 ms strained the period of HD155358b to the range of 185 ec 0.176±0.174 to 200 days and its eccentricity to e < 0.5. ± b ωc 279 38 degrees Three χ2 minima were found by the genetic algorithm: Mb sin i 0.89±0.12 MJ ab 0.628±0.020 AU the first with periods for HD155358c around 326 days, Mc sin i 0.504±0.075 MJ the second around with Pc 530 days and a third one with ac 1.224±0.081 AU −1 Pc around 1540 days. The parameters of the solutions RMS 6.0ms found for the first two minima coincide with the solutions found by GaussFit. However, the third minimum at 1540 days was intriguing, as it represents another possible pe- ets. When we couple this with the results of our genetic riod for HD 155358c. A detailed examination of the so- algorithm orbital fits (section 3.3) and dynamical calcu- lutions in this minimum revealed that their eccentricities lations (section 3.4), we have concluded that the 530 day were all in the range of 0.835 to 0.86. With such a high period is correct for HD 155358 c. This solution is given eccentricity the orbits of the two planets would cross and in Table 2 and is shown in Figure 1. The uncertainties the system would quickly disrupt itself. It appears that reported in Table 2 were generated by GaussFit from this 1540day solution represents a good formal fit for a a maximum likelihood estimation that is an approxima- physically impossible orbital configuration. In order to tion to a Bayesian maximum a posteriori estimator with further rule out the 1540 days as the true period of the a flat prior (Jefferys 1990). The observed HET velocity outer planet we performed 1000 additional trials in the residuals to the HD 155358 c fit, phased to the period of period range of 1500 to 1600 days for HD155358c and HD155358b are shown in the upper panel of Figure 2, using an upper limit on e of 0.6. This time no solutions and vice-versa in the lower panel. c was found with comparable low χ2 values. We thus con- In fitting a Keplerian orbit to observed data, the ec- clude that the 530 day period is the most likely period centricity is often the least well-determined of the orbital for the second planet. elements, and is the one element that is most often able to absorb any additional signals or periodicities that may be 3.4. Dynamical Stability Calculations in the data. For example, in comparing the HD 155358 b single-planet orbit fit with the elements of HD 155358 b While the orbits given in Table 2 represent the least- in the two-planet fits, the most significant changes in squares Keplerian orbit solution to the observations, they the orbital elements of HD155358b are that the eccen- may not necessarily represent physically realistic orbits tricity dropped somewhat and the K value increased. for two planets in this system. The inclusion of the sec- The uncertainty in the eccentricity of HD 155358 c is ond planet HD 155358 c drops the eccentricity of the first rather large. This is probably due to some regions of planet HD 155358 b slightly, but the new outer planet so- sparse phase coverage, particularly in the phase intervals lution has a somewhat non-zero eccentricity itself. While 0.30-0.45 and 0.8-1.0. The rms scatter of the observa- the orbits for this solution do not cross each other, they tions around this two-planet fit is 6.0 m s−1. This rms do have sufficiently close approaches that the planets may is quite consistent with a stellar internal “jitter” of 4.5 well be interacting dynamically. to 5.0 m s−1 and our typical internal observed velocity To investigate the orbital stability of the two-planet uncertainty of 2-4ms−1. If we add a stellar “jitter” of Keplerian orbital solution for the HD 155358 system, we −1 conducted dynamical simulations using the N-body inte- 5ms in quadrature with the internal errors given in 1 Table 1, we then get a reduced chi-squared for our two- grator SWIFT . A full description of SWIFT’s capabil- planet fit of 1.15. ities is given by Levison & Duncan (1994). We adopted the M∗=0.87 M⊙ as discussed below in Sec- 3.3. Genetic Algorithm Investigation tion 4, and the parameters of the two planets were those given in Table 2. The planets were assumed to be copla- In order to understand fully the nature of the orbit nar, and the planet masses were taken to be the mini- of the second planet in this system, we used a standard mum values (i.e. sin i = 1). All simulations which used genetic algorithm to explore further the χ2-landscape of 2 shorter period in the range of 300 days discussed above a 2 planet solution and to find possible additional χ in Section 3.3 for the outer HD 155358 c planet resulted minima. The initial orbital parameters were distributed in ejection of one or both planets within 105 yr. Thus, randomly over a certain range of start values and during we reject all solutions with Pc in the 300day range. The each iteration these parameters were randomly mutated. longer period (P = 530.3 days) for HD155358c is fa- The solutions “evolve” by using χ2 as the environmental c 2 selection mechanism. Solutions that improve χ or at 1 SWIFT is publicly available at least do not degrade it by a certain threshold are allowed http://www.boulder.swri.edu/∼hal/swift.html. 4 Cochran et al.

2002 2004 2006 2008

50

0

-50

52000 52500 53000 53500 54000 54500 JD - 2400000

Fig. 1.— The HET data velocities of HD 155358 with the best double-planet fit (solid line). The RMS scatter of the data around the fit is 6.0ms−1. vored dynamically, and was used in all subsequent simu- approximation. lations. Throughout the dynamical simulations, we have as- The three-body system (star and two planets) was inte- sumed that the two planets are in coplanar orbits. It grated for a total of 108 yr. Initial orbits for the planets is, however, conceivable that the planetary orbits are in- were taken from Table 2. The system remained stable clined with respect to each other. We have conducted for the duration of the simulation. As shown in Figure 3, further tests with mutually inclined planetary orbits to the two planets exchanged eccentricities on a relatively determine the maximum mutual inclination for which the short timescale, with a period of approximately 2700 yr. system remained stable. For the purposes of these exper- The eccentricity of planet b varied between 0.02 and 0.18, iments, the initial inclination of the inner planet was set while the eccentricity of planet c varied between 0.08 and to 0◦ and that of the outer planet was assigned a grid of 0.22. The argument of periastron ω for both planets cir- starting values ranging from 5◦ up to 50◦. The planetary culated with a period of about 2300 yr. The two planets masses were again assumed to be the minimum values, ◦ 7 spend most of their time with |ωb − ωc| between 120 and the systems were integrated for 10 yr. Systems with and 240◦. mutual inclinations less than 45◦ remained stable for the Noting the significant interaction between the two duration of the simulations. For higher values of the in- planets in the minimum-mass case, we tested the effect clination between the two planets, the eccentricities of of inclination of the entire system by adjusting the plan- both planets experienced chaotic variations resulting in etary masses and repeating the simulations. More mas- the ejection of the outer planet within 106 yr (45◦) and sive planets (sin i < 1) should interact more strongly 2 × 105 yr (50◦). with each other, and thus may be less stable. Systems 4. STELLAR PROPERTIES with sin i & 0.33 (i &20◦) remained stable for the du- ration of the tests (107 yr). These results indicate that We measured the photospheric abundance, [Fe/H], the dynamical stability evidenced in Figure 3 does not of HD155358 by analyzing the “template” spectrum used require the special circumstance of a nearly edge-on in- in the radial velocity analysis. We assumed that [Fe/H] clination and thus minimum-mass planets. Even though was an effective proxy for the general photospheric metal- the two planets are definitely interacting dynamically, licity, [M/H]. Our analysis method is the same as de- the timescale of this interaction is long compared with scribed by Bean et al. (2006) for solar-type stars. We the span of our observations. Thus, our two-planet least- briefly describe the technique here, and refer the reader squares Keplerian solution given in Section 3.2 is a valid to that paper for a complete description. We fit synthetic spectra to the profiles of 30 Fe I lines Planetary System Around HD155358 5

Fig. 2.— The best fit double-planet Keplerian orbit solution to the Hobby-Eberly Telescope data for HD 155358. The upper panel shows the phased inner planet (HD 155358 b) orbit, along with the observed data which have had the outer planet (HD 155358 c) orbit removed. The lower panel shows the phased HD 155358 c orbital solution, along with the observed data which have had the the HD 155358 b orbit removed. Two cycles are shown, with the second cycle in grey for clarity. The rms scatter of the data around this double-planet fit is 6.0ms−1. in the observed spectrum. We generated the synthetic (B−V ) color, and parallax from the Hipparcos catalog spectra with an updated version of the plane-parallel, (Perryman et al. 1997). local thermodynamic equilibrium (LTE), stellar analysis We determined [Fe/H], microturbulence ξ, and macro- computer code MOOG (Sneden 1973). We assumed as- turbulence η for HD 155358 by fitting the line profiles trophysical log gf values for the Fe I lines, which were in the observed spectrum simultaneously. As the con- determined by fitting a solar spectrum and assuming the straints on the stellar Teff and log g described above are solar iron abundance log ǫ(Fe)⊙ = 7.45. Our analysis both dependent on [Fe/H], these parameters were also was therefore differential to the sun. varied simultaneously. We fit the observed spectrum by We also adopted model atmospheres computed with using an adaptation of the “Marquardt” χ2 minimization the general-purpose stellar atmosphere code PHOENIX algorithm (Marquardt 1963; Press et al. 1986). The un- (version 13, Hauschildt et al. 1999) for our analysis. certainty in the determined [Fe/H] was calculated from The model atmosphere effective temperature, Teff , was the scatter in the abundances determined for each line constrained using the {(B−V ), [Fe/H]} – Teff rela- individually and deviations due to the uncertainties in tionship of Ram´ırez & Mel´endez (2005). The surface the adopted Teff and log g. gravity, log g, was constrained by interpolating the In addition to the standard spectroscopic parameters, Bertelli et al. (1994) evolutionary isochrones for a given we have also estimated the mass and age of HD155358 MV , Teff , and [Fe/H]. We took the needed V magnitude, by interpolating the Bertelli et al. (1994) evolutionary 6 Cochran et al.

Fig. 3.— Results of dynamical simulations of the HD 155358 system. The evolution of semimajor axis (a), eccentricity, and the argument of periastron (ω) are shown for a typical 104 interval. The simulation was run for a total of 108 years. The two planets exchange eccentricities on a timescale of about 2700 yr, and ω circulates with a period of about 2300 yr. The semimajor axes of both planets are stable throughout the simulation. These computations were for a star of mass 0.87 M⊙, and planet masses fixed at the minimum mass (sin i = 1). However, similar results are obtained for sin i & 0.4. isochrones as we did to constrain the log g. The pa- rameters that we determined for HD155358 are given in TABLE 3 Derived Stellar Parameters Table 3. Table 4 lists the literature values of the stel- for HD155358 lar parameters for comparison. Most notably, we found [Fe/H] = -0.68 ± 0.07. This is in very good agreement Parameter Value with the literature values, which have a range -0.72 ≤ Teff 5760±101 K [Fe/H] ≤ -0.60, and median value [Fe/H] = -0.67. The log g 4.09±0.10 (cgs) derived value of log g, coupled with the mass, age, and [Fe/H] -0.68±0.07 effective temperature of HD 155358 indicate that the star ξ 1.55±0.15 km s−1 1 has evolved off the main sequence. η 1.89±0.20 km s− M⋆ 0.87±0.07 M⊙ Age ∼ 10 Gyr 5. DISCUSSION The planetary system around HD155358 is quite re- markable in that it contains two Jovian-mass planets system formation and early dynamical evolution. in relatively short-period orbits around one of the most The final configuration of any planetary system results metal-poor planet-host star yet found. Thus, this system from a competition of several different physical processes, is crucial to our understanding of the physics of planetary each occurring on its own independent timescale. The Planetary System Around HD155358 7

sub-critical through gas disk dissipation, then the plane- TABLE 4 tary core will remain a Neptune or a super-Earth. Since Stellar Parameters for HD155358 from Other Sources the disk dissipation has a much weaker (if any) depen- dence on stellar metallicity, gas-giant planet formation

Teff log g [Fe/H] Mass Age Source around low Z stars is significantly more difficult than (K) (cgs) (M⊙) (Gyr) around stars of solar or higher metallicity. Unless the 5760 4.09 -0.68 0.87 10.1 1 circumstellar disk is rather massive to begin with, there 5870 4.19 -0.67 0.75 ··· 2 simply isn’t time to grow critical-mass cores before the 5868 4.19 -0.67 ······ 3 gas disk dissipates. Thus, one might expect a critical 5914 4.09 -0.61 ······ 4 lower metallicity threshold for the formation of gas giant ········· 0.84 10.2 5 5860 4.10 -0.60 ······ 6 planets. The value of this threshold would depend on the 5818 4.09 -0.67 0.89 12.7 7 physics governing the overall mass of the disk, which are 5888 ··· -0.69 0.83 13.8 8 poorly understood. ············ 15.7 9 The third relevant timescale for determining the final 5808 ··· -0.72 0.89 11.9 10 configuration of a planetary system is that of the plan- etary system dynamical evolution, which transforms the References. — (1) This Paper, (2) Lambert et al. (1991), (3) Edvardsson et al. system from the one which formed the planets into the (1993), (4) Gratton et al. (1996), (5) Ng & Bertelli system which we detect today. There are two basic phys- (1998), (6) Th´evenin & Idiart (1999), (7) ical processes governing post-formation dynamical evolu- Chen et al. (2001), (8) Feltzing et al. (2001), (9) tion. These are tidal migration prior to disk dissipation, Lambert & Reddy (2004), (10) Nordstr¨om et al. (2004) and planet-planet or planet-planetessimal scattering fol- lowing disk dissipation. Only tidal migration can have planets themselves must form from the remnant disk of any significant dependence on stellar metallicity. The material left over from star formation. This must happen disk evolution is driven by the disk viscosity. In type before the circumstellar disk is dissipated. After planet II migration, in which the planet has opened a gap in formation, the system must evolve into the final state the disk as a result of depleting the local gas in the disk in which we detect it. The dependence of each of these from runaway gas accretion, the planet is locked into processes on the metallicity of the material from which the viscous evolution of the disk. Ida & Lin (2005) give the entire system forms then determines the range of a disk migration timescale that depends linearly on the allowed planetary systems as a function of metallicity. disk viscosity, α. The dependence of α on metallicity Matsuo et al. (2007) investigated the circumstances is not known, and most models assume a constant α of under which stars of various metallicity and mass could order 10−4 for all disks. Thus, it appears that once a form planets within the framework of both the core- system of several planets is able to form in a disk, the accretion model and the disk-instability model. They subsequent dynamical evolution of that system is prob- derive a low metallicity threshold for planet formation ably reasonably independent of Z. Detailed modeling of within the core accretion model of [Fe/H] = -0.85 for a the dependence of disk evolution and type II migration disk of 5 times the mass of the minimum mass solar neb- on Z is needed. If the dynamical evolution of the plan- ula (MMSN) and [Fe/H] = -1.17 for a disk of 10 times etary system is due to tidal migration in the disk, this the MMSN. They also showed that the disk instability migration must still occur before the disk is dissipated. model has essentially no dependence on stellar metallic- If the final configuration of the HD155358 system is due ity, but would require a disk mass of at least 10 times the to post-formation planet-planet scattering, or dynamical MMSN to form these planets. Thus, the two giant plan- interactions of the planets with remnant planetessimals, ets we have discovered in the HD 155358 system could then we would not necessarily expect any significant de- have formed from either formation mechanism. In either pendence of the process on the stellar metallicity. For model they require a disk which was substantially more low metallicity systems, the theoretical challenge is to massive than the minimum-mass solar . form the planets of the required mass before the disk Ida & Lin (2004, 2005) investigated the dependence of dissipates. the formation of planetary systems on stellar metallic- ity within the core-accretion model. The motivation for their model was to explain the observed correlation be- tween high stellar metallicity and planetary systems eas- This material is based on work supported by the ily detected by RV techniques. The basic idea is that National Aeronautics and Space Administration under in systems of high metallicity, the dust surface density Grants NNG04G141G and NNG05G107G issued through in the protoplanetary disk will be enhanced. This will the Terrestrial Planet Finder Foundation Science pro- lead to planetary cores being formed on a much shorter gram. The dynamical simulations used in this work were timescale. Ida & Lin (2005) derived a planetary core carried out at the Texas Advanced Computing Center mass at time t (prior to depletion of the feeding zone) (TACC). The Hobby-Eberly Telescope (HET) is a joint that is proportional to 103Z , where Z is the stellar metal- project of the University of Texas at Austin, the Penn- [Fe/H]⋆ licity (10 Z⊙). For a giant planet to grow, a core sylvania State University, Stanford University, Ludwig- must achieve a mass above a critical value of about 10- Maximilians-Universit¨at M¨unchen, and Georg-August- 20M⊕ before the gas disk is dissipated. If the core does Universit¨at G¨ottingen. The HET is named in honor of reach a super-critical mass while there is still significant its principal benefactors, William P. Hobby and Robert gas content in the disk, it will rapidly accrete gas and E. Eberly. grow to become a gas-giant planet. If the core remains Facilities: HET 8 Cochran et al.

REFERENCES Bean, J. L., Sneden, C., Hauschildt, P. H., Johns-Krull, C. M., & Matsuo, T., Shibai, H., & Ootsubo, T. 2007, Ap.J. in press, Benedict, G. F. 2006, ApJ, 652, 1604 astro-ph/0703237 Bertelli, G., Bressan, A., Chiosi, C., Fagotto, F., & Nasi, E. 1994, Ng, Y. K. & Bertelli, G. 1998, A&A, 329, 943 A&AS, 106, 275 Nordstr¨om, B., Mayor, M., Andersen, J., Holmberg, J., Pont, E., Butler, R. P., Marcy, G. W., Williams, E., McCarthy, C., Jørgensen, B. R., Olsen, E. H., Udry, S., & Mowlavi, N. 2004, Dosanjh, P., & Vogt, S. S. 1996, PASP, 108, 500 A&A, 418, 989 Chen, Y. Q., Nissen, P. E., Benoni, T., & Zhao, G. 2001, A&A, Perryman, M. A. C., Lindegren, L., Kovalevsky, J., Hoeg, E., 371, 943 Bastian, U., Bernacca, P. L., Cr´ez´e, M., Donati, F., Grenon, Cochran, W. D., Endl, M., McArthur, B., Paulson, D. B., Smith, M., van Leeuwen, F., van der Marel, H., Mignard, F., Murray, V. V., MacQueen, P. J., Tull, R. G., Good, J., Booth, J., C. A., Le Poole, R. S., Schrijver, H., Turon, C., Arenou, F., Shetrone, M., Roman, B., Odewahn, S., Deglman, F., Graver, Froeschl´e, M., & Petersen, C. S. 1997, A&A, 323, L49 M., Soukup, M., & Villarreal, M. L. 2004, ApJ, 611, L133 Press, W. H., Flannery, B. P., & Teukolsky, S. A. 1986, Numerical Edvardsson, B., Andersen, J., Gustafsson, B., Lambert, D. L., recipes. The art of scientific computing (Cambridge: University Nissen, P. E., & Tomkin, J. 1993, A&A, 275, 101 Press, 1986) Endl, M., K¨urster, M., & Els, S. 2000, A&A, 362, 585 Ram´ırez, I. & Mel´endez, J. 2005, ApJ, 626, 465 Feltzing, S., Holmberg, J., & Hurley, J. R. 2001, A&A, 377, 911 Ramsey, L. W., Adams, M. T., Barnes, T. G., Booth, J. A., Fischer, D. A. & Valenti, J. 2005, ApJ, 622, 1102 Cornell, M. E., Fowler, J. R., Gaffney, N. I., Glaspey, J. W., Gonzalez, G. 1997, MNRAS, 285, 403 Good, J. M., Hill, G. J., Kelton, P. W., Krabbendam, V. L., —. 1998a, A&A, 334, 221 Long, L., MacQueen, P. J., Ray, R. B., Ricklefs, R. L., Sage, J., Gonzalez, G. 1998b, in ASP Conf. Ser. 134: Brown Dwarfs and Sebring, T. A., Spiesman, W. J., & Steiner, M. 1998, Proc. Soc. Extrasolar Planets, ed. R. Rebolo, E. L. Martin, & M. R. Z. Photo-opt. Inst. Eng., 3352, 34 Osorio, 431 Reid, I. N. 2002, PASP, 114, 306 —. 1999, MNRAS, 308, 447 Santos, N. C., Israelian, G., Mayor, M., Bento, J. P., Almeida, —. 2006, PASP, 118, 1494 P. C., Sousa, S. G., & Ecuvillon, A. 2005, A&A, 437, 1127 Gratton, R. G., Carretta, E., & Castelli, F. 1996, A&A, 314, 191 Scargle, J. D. 1982, ApJ, 263, 835 Grether, D. & Lineweaver, C. H. 2007, preprint, astro-ph/0612172 Sneden, C. A. 1973, PhD Thesis, The University of Texas at Hauschildt, P. H., Allard, F., & Baron, E. 1999, ApJ, 512, 377 Austin Heiter, U. & Luck, R. E. 2003, AJ, 126, 2015 Sozzetti, A., Torres, G., Latham, D. W., Carney, B. W., Stefanik, Ida, S. & Lin, D. N. C. 2004, ApJ, 616, 567 R. P., Boss, A. P., Laird, J. B., & Korzennik, S. G. 2006, ApJ, —. 2005, Progress of Theoretical Physics Supplement, 158, 68 649, 428 Jefferys, W. H. 1990, Biometrika, 77, 597 Th´evenin, F. & Idiart, T. P. 1999, ApJ, 521, 753 Jefferys, W. H., Fitzpatrick, M. J., & McArthur, B. E. 1988, Tull, R. G. 1998, Proc. Soc. Photo-opt. Inst. Eng., 3355, 387 Celestial Mechanics, 41, 39 Valenti, J. A., Butler, R. P., & Marcy, G. W. 1995, PASP, 107, Lambert, D. L., Heath, J. E., & Edvardsson, B. 1991, MNRAS, 966 253, 610 Wright, J. T. 2005, PASP, 117, 657 Lambert, D. L. & Reddy, B. E. 2004, MNRAS, 349, 757 Levison, H. F. & Duncan, M. J. 1994, Icarus, 108, 18 Marquardt, D. W. 1963, J. Soc. Ind. Appl. Math., 11, 431