The HARPS search for southern extra-solar planets. XX. Planets around the active star BD-08:2823
G. Hebrard, S. Udry, G. Lo Curto, N. Robichon, D. Naef, D. Ehrenreich, W. Benz, F. Bouchy, A. Lecavelier des Etangs, C. Lovis, M. Mayor, C. Moutou, F. Pepe, D. Queloz, N. C. Santos, D. Segransan
aa r X i v : . [ a s t r o - ph . E P ] D ec Astronomy&Astrophysicsmanuscript no. article c (cid:13)
ESO 2018October 30, 2018
The HARPS search for southern extra-solar planets ⋆ XX. Planets around the active star BD − ◦ G. H´ebrard , S. Udry , G. Lo Curto , N. Robichon , D. Naef , , D. Ehrenreich , , W. Benz , F. Bouchy , ,A. Lecavelier des Etangs , C. Lovis , M. Mayor , C. Moutou , F. Pepe , D. Queloz , N. C. Santos , D. S´egransan Institut d’Astrophysique de Paris, UMR7095 CNRS, Universit´e Pierre & Marie Curie, 98bis boulevard Arago, 75014 Paris, France Observatoire de Gen`eve, Universit´e de Gen`eve, 51 Chemin des Maillettes, 1290 Sauverny, Switzerland ESO, Karl-Schwarzschild-Strasse 2, D-85748 Garching bei M¨unchen, Germany Observatoire de Paris, GEPI, 5 Place Jules Janssen, F-92195 Meudon, France ESO, Alonso de Cordoba 3107, Vitacura Casilla 19001, Santiago, Chile Laboratoire d’Astrophysique de Grenoble, CNRS (UMR 5571), Universit´e J. Fourier, BP53, 38041 Grenoble cedex 9, France Physikalisches Institut Universit¨at Bern, Sidlerstrasse 5, 3012 Bern, Switzerland Observatoire de Haute-Provence, CNRS / OAMP, 04870 Saint-Michel-l’Observatoire, France Laboratoire d’Astrophysique de Marseille, Universit´e de Provence, CNRS (UMR 6110), BP 8, 13376 Marseille Cedex 12, France Centro de Astrof´ısica, Universidade do Porto, Rua das Estrelas, 4150-762 Porto, PortugalReceived TBC; accepted TBC
ABSTRACT
We report the detection of a planetary system around BD − ◦ Hipparcos photometry database.This program preferentially selected active stars and did not allow the detection of new transiting planets. It allowed however theidentification of the K3V star BD − ◦ − ◦ − ◦ . ± . ⊕ and an orbital period of 5.60 days, whereas BD − ◦ . ± .
03 M
Jup and an orbitalperiod of 237.6 days. This new system strengthens the fact that low-mass planets are preferentially found in multiplanetary systems,but not around high-metallicity stars as this is the case for massive planets. It also supports the belief that active stars should not beneglected in exoplanet searches, even when searching for low-mass planets.
Key words. planetary systems – techniques: radial velocities – stars: individual: BD − ◦
1. Introduction
The HARPS spectrograph (Mayor et al. 2003) is operating since2003 at the 3.6-m ESO telescope in La Silla, Chile. This isa fiber-fed, environmentally stabilized high-resolution echellespectrograph dedicated to high-precision radial velocity mea-surements. Thanks to the exoplanetology programs that are con-ducted with it, it allowed numerous extra-solar planets studiesand discoveries, in the ranges of Jupiter-mass planets (e.g. Pepeet al. 2004, Moutou et al. 2009a), low-mass planets (e.g. Lovis etal. 2006, Mayor et al. 2009), planets around early- (e.g. Desortet al. 2008, Lagrange et al. 2009) or late-type stars (e.g. Bonfilset al. 2007, Forveille et al. 2009), or transiting planets (e.g.Bouchy et al.. 2008, Queloz et al. 2009). The essential qualityof HARPS (the High Accuracy Radial velocity Planet Searcher)is its high stability, which results in a sub-m s − accuracy in theradial velocity measurements, on time-scales of several years.This allows the detection of planets in the Neptune or Super-Earth mass ranges, on progressively increasing orbital periods asthe time-span of the monitoring is growing up. This instrument ⋆ Based on observations made with HARPS spectrograph on the 3.6-m ESO telescope at La Silla Observatory, Chile, under the programs ID072.C-0488, 074.C-0364 and 078.C-0044. hitherto plays a major role in the improvement of the knowledgesof exoplanets.We report here the detection of a new multiplanetary system.It serendipitously originates from the radial velocity follow-upaccompanying a search for planetary transits in the
Hipparcos epoch photometry annex (Perryman et al. 1997). Despite its sig-nificant stellar activity, the star BD − ◦ Hipparcos photometry in §
2. The HARPS observations of BD − ◦ § §
4, then the re-sults are discussed in §
2. Search for transiting planets in the
Hipparcos database
The goal of our initial search in the
Hipparcos database was totry to find new transiting planets, especially around bright stars.Transiting planets could allow numerous important studies tobe performed. These studies include planetary radii, masses and
H´ebrard et al.: Planets around the active star BD − ◦ densities measurements, accurate determination of the inclina-tions and the eccentricities of the orbits, possible detections oftransit timing variations due to additional companions, measure-ments of the spin-orbit (mis)alignment angles, detections of theemission or absorption planetary spectra, or even satellites orrings detections. The famous planets transiting in front of thebright stars HD 209458 and HD 189733 are those that allow themost accurate measurements and the deepest investigations.The Hipparcos epoch photometry annex contains between ∼
40 and ∼
300 photometric measurements performed duringthe 1226-day duration of the mission for each of the 118 204stars of the catalog, in the magnitude limit V ≤
10. With about0.1% chance that a given star harbors a transiting extra-solarplanet, the
Hipparcos all-sky survey must contain photomet-ric measurements for tens of transiting hot Jupiters. This be-lief is reinforced by the two cases of transiting planets a pos-teriori re-discovered in
Hipparcos data: HD 209458b (Robichon& Arenou 2000, Castellano et al. 2000, S¨oderhjelm 1999) andHD 189733b (Bouchy et al. 2005, H´ebrard & Lecavelier desEtangs 2006). Thanks to the long time baseline, these data alsoallow the orbital period to be measured with an accuracy of theorder of the second.Up to now, there are the only two transiting planets that havebeen found in the
Hipparcos database. Indeed, with photometricvariations of about 1% or less, such transits are di ffi cult to iden-tify in the Hipparcos data, that present a mean photometric accu-racy of the same order of magnitude. The poor time coverage, bycomparison with dedicated surveys as SuperWASP, XO, CoRoTor Kepler, is another di ffi culty. Jenkins et al. (2002) concludedthat because of its poor photometric quality, the Hipparcos cata-log does not represent a likely place to detect planets in the ab-sence of other informations. It might however provide planetarytransit candidates for follow-up observations. Laughlin (2000)has searched in the
Hipparcos epoch photometry for transit-ing planets within 206 metal-rich stars. None have been con-firmed thereafter.We present in Appendix A the systematic search we man-aged in the
Hipparcos epoch photometry annex for periods com-patible with planetary transits. We constructed a ranked list ofcandidates for follow-up radial velocity measurements, based onthe depth and significance of the planetary transits that couldbe detected in the
Hipparcos photometry. The simulations weperformed indicated that our detection rate for transiting planetsis low, of the order of 2 %; so the chances to find new transit-ing planets in the
Hipparcos data are small (see Appendix A).However, the particularly high interest for the potential discov-ery of a planet transiting a bright star pushed us to perform aradial velocity follow-up of our candidates.We observed 194 of these selected, ranked targets withHARPS, in order to search for radial velocity variations in agree-ment with the transiting candidates found in the
Hipparcos pho-tometry. The HARPS setup, the spectra extraction and the ra-dial velocity measurements were identical to those describedbelow in § − were measured for 37 stars,i.e. 19% of our observed sample. Those variations are mainlycaused by stellar activity. Indeed, most of the targets with largeradial velocity variations show emissions in the Ca ii H & K lines(3934.8 Å and 3969.6 Å). The level of these emissions is quanti-fied with the activity S-index (Mount Wilson System), which isconverted to the log R ′ HK index (Santos et al. 2000). The majorityof the variable targets have log R ′ HK indexes larger than − .
7, in- dicating prominent chromospheric activity; this can explain suchlevel of radial velocity variations (Santos et al. 2000).Some targets first seemed to exhibit periodic radial velocityvariations, but subsequent monitoring with HARPS showed thatthese variations were not persistent with time, in agreement withtransient activity processes on the stellar surface due to activity(flares, spots, plages...) modulated by the stellar rotation. In ad-dition, in most of the cases, analysis of the line profiles usingbisectors of the cross-correlation functions shown variations inthe shape of the lines with time. This indicates that the observedradial velocity variations are not due to Doppler shifts of thelines, but rather to deformations of the shape of the spectral lines.Some cases revealed a clear anti-correlation between radial ve-locity and line-bisector orientation, which could be understoodas the signature of cool spots on the stellar photospheres (see,e.g., Queloz et al. 2001, Melo et al. 2007, Desort et al. 2007,Boisse et al. 2009).Thus, our procedure seems to preferentially select activestars. This should be contrasted with transiting candidates ob-tained from the photometric surveys dedicated to transitingplanet searches, which are mainly eclipsing binaries or transit-ing planets (see, e.g., Pont et al. 2005). None of the targets thatwe observed through this program have shown radial velocityvariations in agreement with a reflex motion due to a hot Jupiter.Instead, most of the radial velocity variations seem to be causedby stellar activity.However, these observations allowed the serendipitously dis-covery of a new planetary system, without transit detection inthe
Hipparcos data. This system orbits BD − ◦ R ′ HK in-dexes within our monitored targets. This activity level compli-cates planet detection but does not prohibit it in this case. Weconcentrate below on this target and the detection of its harboredplanetary system.
3. Observations and properties of BD − ◦ − ◦ − ◦ R =
115 000, with a fiber diameter of 1 arcsec. The spectra wereextracted from the detector images with the HARPS pipeline,that includes localization of the orders on the 2D-images, opti-mal order extraction, cosmic-ray rejection, wavelength calibra-tion and corrections of flat-field. The pipeline then performs across-correlation of the extracted spectra with a numerical mask(K5-type in this case), and finally measures the radial velocitiesfrom Gaussian fits of the cross-correlation functions (CCFs), fol-lowing the method described by Baranne et al. (1996) and Pepeet al. (2002). The full dataset we use for BD − ◦ − ◦ − ◦ § / early-2007 as part ofthe program 078.C-0044, but most of the data were secured as ´ebrard et al.: Planets around the active star BD − ◦ part of the Guaranteed Time Observations (GTO) survey pro-gram 072.C-0488 (Mayor et al. 2003). This is the case in par-ticular for the intensive series in 2007-2008 (BDJ − Hipparcos original program.
Fig. 1.
Top:
Radial velocity HARPS measurements ofBD − ◦ § § Bottom:
Residuals of the fitwith 1- σ error bars.The derived radial velocities are reported in Table 1. Theaccuracies are between 1.0 and 2.4 m s − , typically around1.6 m s − . This includes photon noise but not the jitter due tostellar oscillations or activity ( § − were not included in the final dataset of83 spectra.The radial velocities show a significant 11-m s − disper-sion (43 m s − peak-to-peak), well over the expected accuracy(Fig. 1, upper panel). The CCF from which those radial veloci-ties were measured show parameters that also significantly varywith time. Their full widths at half maximum fluctuates between6 . ± .
01 and 6 . ± .
01 km s − , and their contrasts from46 . ± .
05 and 47 . ± .
05 % of the continuum. The bisectorspan of the CCF also show a dispersion, at the level of 7 m s − (see Fig. 2, upper panel). Thus, Doppler shifts of the spectrallines of BD − ◦ Table 2 summarizes the stellar parameters of BD − ◦ V = .
86. Its
Hipparcos parallax
Table 1.
HARPS measurements of BD − ◦ † . BJD RV σ (RV) FWHM Bis. log R ′ HK -2 400 000 km s − km s − km s − km s − † : the six columns are respectively the Barycentric Julian Day of theobservations, the radial velocity and its uncertainty, the full widthat half maximum of the CCF, the span of its bisector, and the Ca ii activity index. ( π = . ± .
61 mas) implies a distance of 42 . ± . Hipparcos color is B − V = . ± . T e ff = ±
63 K, thegravity log g = . ± .
26, the metallicity [Fe / H] = − . ± .
03, and an uncertain age of 4 . ± . M ∗ = .
74 M ⊙ , with a formal error bar of ± . ⊙ .Following Fernandes & Santos (2004), we rather adopt a conser-vative ±
10 % uncertainty, corresponding to ± .
07 M ⊙ . We derivea projected rotational velocity v sin i ⋆ = . − from the pa-rameters of the CCF using a calibration similar to that presentedby Santos et al. (2002). Acoustic oscillations are not averagedout in the data, as there are on timescales similar or shorter thanthe exposure times we used. However, for a K-type star, oscil-lation amplitudes are expected to be negligible by comparisonwith stellar activity e ff ects on the measured radial velocities. Table 2.
Stellar parameters for BD − ◦ Parameters Values m v . B − V . ± . . ± . . ± . v sin i ⋆ [km s − ] 1 . P rot [days] 26 . ± . R ′ HK − . ± . / H] − . ± . T e ff [K] 4746 ± g [cgs] 4 . ± . ⊙ ] 0 . ± . . ± . The core of the large H & K Ca ii absorption lines ofBD − ◦ R ′ HK = − . ± .
1, butit significantly varies with time, between extrema log R ′ HK = − . ± .
01 and − . ± .
01. Such stellar activity would im-ply a significant jitter on the observed stellar radial velocities.For a K-type star with this level of activity, Santos et al. (2000)predict a dispersion up to 10 m s − for the stellar jitter. This isthe order of magnitude of the dispersion of our measurementsof BD − ◦ § H´ebrard et al.: Planets around the active star BD − ◦ tivity as the main cause of radial velocity variations is to lookfor anti-correlation between radial velocity and the bisector span(see, e.g., Queloz et al. 2001, Boisse et al. 2009). The upperpanel of Fig. 2 shows the bisector spans as a function of theradial velocities for BD − ◦ − ◦ − ◦ Hipparcos data of BD − ◦ ff erent estimation of the rotation period could be ob-tained from the HARPS data through the stellar activity indi-cators, as shown in Fig. 3. In this figure are plotted the Lomb-Scargle periodograms (Press et al. 1992) of five HARPS signals:the radial velocities, full widths at half maximum (FWHMs),contrasts and bisectors of the CCFs, and of the log R ′ HK activityindexes (Table 1). The fourth last parameters could show signa-tures of the stellar activity, and in particular their modulationswith the stellar rotation; the radial velocity variations could aswell show signatures of the stellar activity and rotation, but alsothe Doppler reflex motion due to companion(s). The five peri-odograms in Fig. 3 all clearly show a peak around 26 days. Weinterpret it as the signature of modulations of the stellar surfacedue to activity (flares, spots, plages...). This kind of phenomenaare expected for such an active star. As there are sporadic eventsaltering the surface of the star and having limited life times, theywould imply quasi-periodic variations of the shapes of the spec-tral lines, with periods near the stellar rotation period, and un-conserved phases. The rotation period of BD − ◦ − ◦ − peak-to-peak, see upper panel of Fig. 1) andthe two orbital periods (near 5.6 and 238 days) imply projectedmasses M p sin i well below the mass of Jupiter. The radial ve-locity variations detected by HARPS would thus originate bothfrom stellar activity jitter and planetary companions.
4. A planetary system around BD − ◦ We first fitted the radial velocities using a two-planet Keplerianmodel without mutual interactions. The results are plotted inFigs. 1 and 4. The inner planet, BD − ◦ K = . − ,corresponding to a planet with a minimum mass M p sin i = . ⊕ , thus similar to the mass of Uranus. Its orbit has a periodof 5 .
60 days and is circular or with low eccentricity. The outerplanet, BD − ◦ K = . − ; this corresponds to a planet with a minimum Fig. 2.
Bisector span as a function of the radial velocity (upperpanel) and of the residuals of the fit with two planets as shownin Figs. 1 and 4 (lower panel). On the lower panel, the filled cir-cles indicate the measurements secured on the 101-day interval,where we fitted the stellar jitter ( § x - and y -axes on both panels.mass M p sin i = .
33 M
Jup , slightly above the Saturn mass.The orbital period is 237 . e = . χ of the Keplerian fit is 3.2, and the standarddeviation of the residuals is σ O − C = . − . This is reducedwhen compared to the 11-m s − dispersion of the original radialvelocities, but this is clearly higher than the 1.6-m s − typical er-ror bars on the individual measurements. The extra dispersion, ofthe order of 4 m s − , is mainly due to stellar activity jitter as seenabove. By fitting the two planets without including the stellar ac- ´ebrard et al.: Planets around the active star BD − ◦ Fig. 3.
Lomb-Scargle periodograms for BD − ◦ R ′ HK activity indexes.Each of these five signals shows a peak in the power around26 days, which could be interpreted as the stellar rotation period.On the upper panel, the radial velocities periodogram shows twoextra peaks at 5.6 and 238 days, which are not seen in the fourother panels. The horizontal dashed lines correspond to the false-alarm probability of 1 × − .tivity, we assume that the stellar jitter would be averaged out overthe five years of observations, as activity induces quasi-periodice ff ects that are not exactly duplicated in time with the stellar ro-tation. This results in a quite large dispersion around the fit withrespect to the error bars, whereas the periodic signal of the twoplanets remains coherent over the five years of observations.Fig. 5 strengthens the interpretation in term of planets ofthe two signals at 5.6 and 238 days. The upper panel showsthe periodiogram of the radial velocities, exactly as the upperpanel of Fig. 3. On the second panel of Fig. 5 is shown the pe-riodiogram of the radial velocity residuals, after a fit includingBD − ◦ σ O − C = . − . On this periodogram, the peak at238 days of course is no longer visible. The main peak is thisat 5.6 days. Its false-alarm probability is 5 . × − . It is clearlystronger than the peaks due to activity. We note also the pres-ence of two other peaks, at 0.8 and 1.2 day, that are the one-dayaliases of the 5.6-day signal in the frequency space (1 ± / . Fig. 4.
Phase-folded radial velocity curves for BD − ◦ P = .
60 d, top) and BD − ◦ P = . ff ect of the other planet. The HARPS radial ve-locity measurements are presented with 1- σ error bars, and theKeplerian fits are the solid lines. Orbital parameters correspond-ing to the fits are reported in Table 3.We adopted 5.6 days as the actual period of this signal insteadof one of the two aliases, because it is more likely according tothe sampling. In addition, its peak is higher than those of thetwo aliases, and Keplerian fits performed with 0.8 or 1.2 day oforbital periods produce higher residuals dispersion, and eccen-tric orbits.Similarly, the third panel of Fig. 5 shows the periodiogram ofthe radial velocity residuals, after a fit including BD − ◦ × − . Wenote that in addition to the clear peak at one day (which is dueto the aliases of all the detected signals), all the four panels inFig. 5 show a possible peak near 700 days; this could be the sig-nature of a third, outer planet. Fits with a curvature in addition tothe two planets also suggest a possible long-period signal (with-out significant e ff ects on the parameters of BD − ◦ − ◦ H´ebrard et al.: Planets around the active star BD − ◦ Table 3.
Fitted orbits and planetary parameters for theBD − ◦ σ error bars. Parameters BD − ◦ − ◦ P [days] 5 . ± .
02 237 . ± . e . ± .
15 0 . ± . ω [ ◦ ] 30 ± − ± K [m s − ] 6 . ± . . ± . T (periastron) [BJD] 2 454 637 . ± . ± M p sin i [M Jup ] 0 . ± . ‡ . ± . ‡ M p sin i [M ⊕ ] 14 . ± . ‡ ± ‡ a [AU] 0 . ± . ‡ . ± . ‡ V r [km s − ] 53 . ± . σ O − C [m s − ] 4.3reduced χ N ‡ : using M ⋆ = . ± .
07 M ⊙ get on a longer time baseline are mandatory to establish or notthe presence of the hypothetic planet BD − ◦ − ◦ In the previous section we have shown a Keplerian fit of the twoplanets without attempting any fit of the stellar activity jitter sig-nal in the radial velocities. Our knowledge of the activity is poor;for example the number, locations and sizes of potential stellarspots are far to be controlled. We attempt here a naive, phe-nomenological approach, based on the fact that activity showssignals in the periodograms, so stellar jitter has a periodical na-ture that is linked to the rotation of the star. Unfortunately thesesignals are only quasi-periodical : they do not reproduce them-selves periodically in an identical form for a long time, as allthese phenomena have limited life times. Fitting the stellar jitterwith sinusoids over the 5-year time span do not provide satisfac-tory fits.We thus tried such sinusoid fits on a shorter time span. Wechose a 101-day interval, between BJD-2 400 000 = − ). Thisshould improve the coherence of the stellar jitter in term of pe-riodicity, as well as allowing the 5.60-day-period planet to bewell sampled. Using sinusoids of periods near 26 days, we fit-ted on this 101-day interval the activity indicators studied in Fig. 5.
Lomb-Scargle periodograms of the HARPS radial veloc-ities. The upper panel shows the periodogram computed on theinitial radial velocities, without any fit removed. The second andthird panels show the periodograms computed on the residualsof the fits including BD − ◦ − ◦ × − .Fig. 3, namely the FWHMs, contrasts and bisectors of the CCFs,and the log R ′ HK indexes. The best solution we obtained wasfor a 26.6-day period. Together with the periodograms shownin Fig. 3, this allowed us to determine the rotation period:26 . ± . − ◦ § − ◦ ´ebrard et al.: Planets around the active star BD − ◦ Fig. 6.
Fit of the two planets and the stellar activity on a 101-dayinterval. The upper plot shows the HARPS radial velocities ofBD − ◦ middle plot show the residuals of this fits. The lower plot showson five panels the measurements of the FWHMs, contrasts andbisectors of the CCFs, the log R ′ HK and the stellar jitter, respec-tively from top to bottom. Sinusoid fits with 26.6-day period areoverplotted. The stellar jitter in the lower panel is the sinusoidthat is added to the two Keplerian in the upper plot. The verticaldashed-line helps visualize the correlations. Error bars at 1 σ areplotted for all measurements in this figure.Keplerian in the upper panel. It has an amplitude of ± − .The residuals of the radial velocities fit with two Keplerian anda sinusoid is plotted in the middle panel of Fig. 6. Its dispersionis reduced down to 2.5 m s − , which is an improvement by com-parison to the 4.3 m s − dispersion obtained on the full datasetin § − dispersion if computed only on the 101-day interval, again with-out stellar jitter fit). It remains larger than the error bars, because of the imperfection of the sinusoid model we used for the stellarjitter (and perhaps also because of possible extra planets). Werestrained this fit to a single sinusoid for the stellar jitter, with-out including extra sinusoids at the periods of the stellar rotationharmonics. Indeed, these harmonics are not detected in the radialvelocities (see § ff erent parameters on this 101-day timespan (Fig. 6, lower panels). The FWHMs and contrasts of theCCFs are anti-correlated, implying spectral line deformationsdue to activity that let nearly constant their equivalent widths.This could be understood as the surface of the CCF is an indi-cator of the stellar metallicity (Santos et al. 2002). The veloc-ity jitter is barely correlated with log R ′ HK , as well as with theFWHM and the bisector of the CCF. The correlation betweenthe FWHM and log R ′ HK is di ff erent from what was seen for ex-ample in the case of the spotted star CoRoT-7, for which thosetwo values are rather anti-correlated (Queloz et al. 2009). Also,the apparent correlation between log R ′ HK and the radial veloc-ity jitter is di ff erent from the picture seen in the cases of theactive stars HD 166435 (Queloz et al. 2001), GJ 674 (Bonfils etal. 2007) or HD 189733 (Boisse et al. 2009). And finally, thereis an apparent correlation between the bisector and the radialvelocity jitter on this 101-day time span, possibly with a smallphase o ff set (see Fig. 6). This possible correlation is also shownon the lower panel of Fig. 2. All these relations drawn a picturefor BD − ◦ ff erent from a simplescenario where the radial velocity jitter is mainly due to darkspots on the stellar surface that modulate the shape of the linesas the star is rotating. Other phenomena are likely to occur onthis star; they may include pulsations, convections, flares, plages,hot spots...The e ff ect of the stellar activity on the observed radial ve-locities of BD − ◦ − . This is lowerthan the semi-amplitudes measured for the two detected plan-ets, but not negligible. The error bars reported in Table 3 wherederived from χ variations and Monte Carlo experiments withand without stellar activity modeling, as well as from trials anderrors with di ff erent kinds of sinusoids for the stellar jitter. Ourpoor understanding of the stellar activity of BD − ◦
5. Discussion
Finding transiting planets in the
Hipparcos epoch photometryannex does not look promising. Our attempt for a-priori de-tections did not succeed, and up to now, only two a-posteriori detections were performed within the
Hipparcos data, in thecases of HD 209458b and HD 189733b that were first revealedfrom ground observations. Three main limitations make it dif-ficult. First, the error bars on individual photometric measure-ments are of the same order of magnitude than the expected sig-nal for transits of giant planets, or even slightly larger. Second,the sparse time-coverage allows only a few points to be ob-tained in a potential transit. These two limitations make any tran-sit identification at the limit of detection. Third, stellar activitycould produce false positives, which are di ffi cult to identify withthe sparse time-coverage. The ESA mission Gaia (Perryman etal. 2001) is awaited as the successor of Hipparcos . Its poten-tial for planetary transit discoveries in front of bright stars couldbe considered, as its sensitivity and accuracy would be better
H´ebrard et al.: Planets around the active star BD − ◦ than Hipparcos . However, the time coverage won’t be betterthan that of
Hipparcos , which will prohibit well-resolved lightcurve studies. Stellar activity should thus also be a limitation fortransit detections with Gaia. Dedicated surveys have proven tobe more e ffi cient for transit detections. Ground-based programsallow Jupiter-size planets to be detected, with improving accu-racies that now point toward planets with smaller radii (Bakoset al. 2009), whereas space-based programs allow planets witheven smaller radii to be detected (L´eger et al. 2009, Quelozet al. 2009), as well as planets on longer periods (Moutou etal. 2009b). If accepted, the ESA-proposed space mission PLATO(Catala et al. 2009) could permit the detection of such kind oftransiting planets in front of brighter stars.If this search in the Hipparcos data did not provide any de-tection of new transiting planets, it allowed the serendipitous dis-covery of a new planetary system around BD − ◦ − ◦ ⊕ . Its 5.60-day orbit could be circular orslightly eccentric. The outer planet, BD − ◦ Jup .Its orbit is moderately but significantly eccentric, and has a pe-riod of 237.6 days. As the masses are low and the orbits aredistant and nearly circular, the mutual interactions between thetwo planets are negligible.The reflex motions that these two planets induce to their hoststar have semi-amplitudes of 6.5 and 13.4 m s − , which can bedistinguished from the 4-m s − jitter due to stellar activity. Wesummarize here the arguments that allow us to conclude thatthese two signals are due to planets and not to stellar activity: – the signals with periods of 5.60 and 237.6 days are seen onlyin the radial velocities, and are not seen in the shapes of thelines nor in the activity indexes; – these two periods do not correspond to the rotation period ofthe star, nor to its harmonics or its aliases; – with the available data, the signal is coherent over the 5-yeartime span of the observations.The radial velocities can be fitted on a short time span bytwo Keplerian and a sinusoid that models the stellar activity.Acknowledging that we do not have a good understanding ofthe stellar activity processes, we can alternatively fit the radialvelocities using two Keplerian only. By doing that we assumethat the stellar jitter is damped and averaged out over the 5-yeartime-span, and would only lead to a dispersion of the residualsof the fit that are larger than expected from the accuracy of theradial velocity measurements. Additional planetary companionsare possible in this system, so the monitoring of BD − ◦ − ◦ − ◦ Fig. 7.
Stellar metallicities as a function of the planetary masses,for 304 known extra-solar planets ( top ), and for the 96 onesamong them with orbital periods shorter than 20 days ( bot-tom ). BD − ◦ − ◦ − ◦ . While planets with masseslarger than 1 M Jup are more numerous around overmetallic stars,there are clearly more numerous around undermetallic stars forplanetary masses below 20 M ⊕ . The same e ff ect is seen if weselect only the planets with orbital periods of 20 days or shorter,i.e. in the period regime where most of the low-mass planets aredetected up to now (lower panel of Fig. 7). Indeed, among the ∼
30 planets with a projected mass lower than 0.1 M
Jup , onlyfour have orbital periods longer than 20 days (namely Gl 581d,HD 40307d, HD 69830c and d). http: // exoplanet.eu´ebrard et al.: Planets around the active star BD − ◦ Planet formation models based on the core accretion scenario(Mordasini et al. 2009a, 2009b) have shown that the sharply ris-ing probability of detecting giant planets with stellar metallic-ity can be at least partially accounted for by the fact that metalrich systems favor the formation of massive planets and that theradial velocity technique is most sensitive to massive bodies.Observations also seem to indicate that this correlation vanishesfor Neptune-mass planets, a trend also found in population syn-thesis calculations (Mordasini et al, in preparation). These cal-culations even show that the correlation reverses for very smallplanetary masses ( < − ⊕ ), a prediction that will have tobe confirmed by observations. In fact, a careful analysis of thetheoretical models indicate that the critical parameter is the over-all mass of heavy elements rather than the metallicity. Since thismass is determined by the metallicity and the mass of the initialproto-planetary disk (which is unknown), it makes a straight in-terpretation more di ffi cult. For example, models predict that lowmass objects orbiting metal rich stars or relatively massive plan-ets orbiting metal poor stars are also possible albeit they shouldbe rare.The BD − ◦ ∼
30 planets with projected masses lowerthan 0.1 M
Jup are found in multiplanet systems, whereas this ra-tio is only ∼
25 % for all the ∼
350 known planets. Consideringagain only the known planets with orbital periods shorter than20 days, less than ∼
20 % of them are found to be in multiplanetsystems: those ∼
20 % are mainly low-mass planets.No photometric search for transits have been managed forBD − ◦ i of the orbit, the transitprobability for BD − ◦ Hipparcos photometric accuracy.
Acknowledgements.
We would like to thank F. Pont, F. Arenou, I. Boisse,X. Bonfils, J.-M. D´esert, R. Ferlet, J. Laskar, D. Sosnowska and A. Vidal-Madjar for helps and discussions, as well as the di ff erent observers from otherHARPS programs who have also measured BD − ◦ ff for their support on the HARPS instrument. GH, FB andALDE acknowledge support from the French National Research Agency (ANR-08-JCJC-0102-01). DE acknowledges financial support from the CNES. NCSwould like to thank the support by the European Research Council / EuropeanCommunity under the FP7 through a Starting Grant, as well from Fundac¸˜ao paraa Ciˆencia e a Tecnologia (FCT), Portugal, through programme Ciˆencia 2007, andin the form of grants reference PTDC / CTE-AST / / / CTE-AST / / References
Bakos, G. A., Torres, G., P´al, A., et al. 2009, ApJ, submitted [arXiv:0901.0282]Baranne, A., Queloz, D., Mayor, M., et al. 1994, A&AS, 119, 373Boisse, I., Moutou, C., Vidal-Madjar, A., et al. 2008, A&A, 495, 959Bonfils, X., Mayor, M., Delfosse, X., et al., A&A, 474, 293Bouchy, F., Udry, S., Mayor, M., et al. 2005, A&A, 444, L15Bouchy, F., Queloz, D., Deleuil, M., et al. 2008, A&A, 482, L25Bouchy, F., Mayor, M., Lovis, C., et al. 2009, A&A, 496, 527Castellano, T., Jenkins, J., Trilling, D. E., Doyle, L., & Koch, D. 2000, ApJ,532, L51Catala, C., 2009, ExA, 23, 329Desort, M., Lagrange, A.-M., Galland, F., Udry, S., Mayor, M. 2007, A&A,473, 983Desort, M., Lagrange, A.-M., Galland, F., Beust, H., Udry, S., Mayor, M., LoCurto, G. 2008, A&A, 491, 883Fernandes, J., Santos, N. C. 2004, A&A, 427, 607Forveille, T., Bonfils, X., Delfosse, X., et al. 2009, A&A, 493, 645H´ebrard, G., & Lecavelier des Etangs, A. 2006, A&A, 445, 341Jenkins, J. M., Caldwell, D. A., & Borucki, W. J. 2002, ApJ, 564, 495Lagrange, A.-M., Desort, M., Galland, F., Udry, S., Mayor, M. 2009 A&A,495, 335Laughlin, G. 2000, ApJ, 545, 1064 L´eger, A., Rouan, D., Schneider, J., et al., 2009, A&A, 506, 287Lovis, C., Mayor, M., Pepe, F., et al. 2006, Nature, 441, 305Mamajek, E. E., & Hillenbrand, L. A. 2008, ApJ, 687, 1264Mayor, M., Pepe, F., Queloz, D., et al. 2003, Msngr., 114, 20Mayor, M., Udry, S., Lovis, C., et al. 2009, A&A, 493, 639Melo, C., Santos, N. C., Gieren, W., et al. 2007, A&A, 467, 721Mordasini, C., Alibert, Y., Benz, W. 2009a, A&A, 501, 1139Mordasini, C., Alibert, Y., Benz, W., Naef, D. 2009b, A&A, 501, 1161Moutou, C., Mayor, M., Lo Curto, G., et al. 2009a, A&A, 496, 513Moutou, C., Bruntt, H., Guillot, T., et al. 2009b, A&A, 488, L47Noyes, R. W., Hartmann, L. W., Baliunas, S. L., Duncan, D. K., Vaughan, A. H.1984, ApJ, 279, 763Pepe, F., Mayor, M., Galland, F., et al. 2002, A&A, 388, 632Pepe, F., Mayor, M., Queloz, D., et al. 2004, A&A, 423, 385Perryman, M. A. C., Lindegren, L., Kovalevsky, J., et al. 1997, A&A, 323, L49Perryman, M. A. C., de Boer, K. S., Gilmore, G., et al. 2001, A&A, 369, 339Pont, F., Bouchy, F., Melo, C., Santos, N. C., Mayor, M., Queloz, D., & Udry, S.2005, A&A, 438, 1123Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. 1992,Numerical recipes in C. The art of scientic computing, 2nd (Cambridge:University Press)Queloz, D., Henry, G. W., Sivan, J. P., et al. 2001, A&A, 379, 279Queloz, D., Bouchy, F., Moutou, C., et al. 2009, A&A, 506, 303Robichon, N., & Arenou, F. 2000, A&A, 355, 295Santos, N. C., Mayor, M., Naef, D., et al. 2000, A&A, 361, 265Santos, N. C., Mayor, M., Naef, D., et al. 2002, A&A, 392, 215Santos, N. C., Israelian, G., Mayor, M. 2004a, A&A, 415, 1153Santos, N. C., Bouchy, F., Mayor, M., et al. 2004b, A&A, 426, L19S¨oderhjelm, S. 1999, Informational Bulletin on Variable Stars, 4816, 1Udry, S., Mayor, M., Benz, W., et al. 2006, A&A, 447, 361
Appendix A: Selection of targets for transitingplanets search in the
Hipparcos database
We present in this Appendix the systematic search we managedin the
Hipparcos epoch photometry for planetary transit candi-dates ( § Hipparcos epoch photometry catalog, according to the follow-ing criteria: – B − V > . – π > σ π (defined parallax); – empty H48 field (reference flag for photometry); – H52 field di ff erent from D, P, or R (variability types); – R ⋆ < ⊙ (small stellar radius).Then we kept only targets with at most one epoch brighterthan 3 σ from the average magnitude (in order to remove re-maining variables) and at least 40 di ff erent available epochs. Weperformed the periods’ research on the 17 800 remaining stars.For each target, we took the two faintest epochs and we scanall the possible periods in order to have those two points in atransit. Then we adopted the period producing the lowest flux inthe drop, and we quantified this solution with the parameter α ,defined as: α = < V in > − < V out > q σ < V in > + σ < V out > where < V in > and < V out > are the averaged magnitudes ofthe epochs respectively in and out the drop, and σ the standarddeviations. The higher α is, the deeper and more significant thedrop is. We also computed the χ for the fit of the epochs inthe drop with a transit curve for a planet with a radius R p = (0 . ± .
04) R ⊙ , implying a ∆ V . drop in magnitude: χ = X transit ( V i − < V out > − ∆ V . ) σ V i + σ < V out > + σ ∆ V . . − ◦ We finally computed the goodness of fit (gof) for this χ . Theselection of candidates is based on those two parameters: α andthe goodness of fit. go f go f Simu of constant starsSimu of stars with planets go f Simu of constant starsSimu of stars with planetsHipparcos data
Fig. A.1.
Threshold tuning simulations. The parameters α and α ′ characterize the depth and the significance of the transit, and gof the goodness of fit. These parameters are defined in the text. Top : Constant stars.
Middle : Constant stars (black) and stars withtransits (red).
Bottom : Actual data (blue) and simulations forconstant stars (black) and stars with transits (red). We performed two simulations in order to tune the thresh-old: for the 17 800 stars selected above, we assumed in the firstsimulation that all the stars are constant, whereas in the secondsimulation, we assumed that all the stars host a transiting planetwith a radius chosen randomly within 0 . − .
15 R ⊙ . We keptthe individual error on the photometry on each epoch.The upper panel of Fig. A.1 shows the goodness of fit as afunction of α for the first simulation. The distribution is skewas α and the goodness of fit are correlated. In order to makeeasier the selection, we fitted the distribution with a line, andcomputed α ′ , which is corrected (with a first-order polynomial)from this skew. The middle panel of Fig. A.1 shows the good-ness of fit as a function of α ′ for the two simulations. α ′ > . − / =
2% the detection rate. With 0.06% chancethat a given star harbors a transiting extra-solar planet, the 17800targets of our latest sample should include ∼
11 transiting plan-ets. The expected transiting planet detection number with ourmethod should thus be slightly less than unity (11 × = . α ′ and highgoodness of fit are eclipsing binaries that were not identified inthe Hipparcos catalog. In order to remove those binaries, we re-moved targets with goodness of fit larger than 2.5. We also re-moved targets with R ∗ > . ⊙ , and obtained a list of candi-dates, sorted by decreasing α ′ .Targets referenced in SIMBAD as binaries, active, or vari-able stars were removed from the obtained list, as well as tar-gets with already known planets at the time of the observ-ing run (HD 70642, HD 39091, HD 10647, HD 4208, HD 17051,HD 13445, HD 75289, and HD 83443 – we checked that the Hipparcos photometric data did not include periodic variationsat their periods).Using HARPS, we performed follow-up observations of 194of these selected targets in December 2004, as part of the pro-gram 074.C-0364. The apparent magnitudes of the observedstars range from 4.9 to 11.5. Exposures of typically a few min-utes duration were obtained, allowing 70 to 90 targets to beobserved each night. Errors on the measured radial velocitiesare typically of the order of 2 m / s. In order to identify hotJupiters, targets were removed from the candidate list after one,two, or three observations, according to the following observa-tional strategy: – observations were stopped after one HARPS measurementif: 1) there are two peacks in the CCF; only one SB2 wasfound during our observations (HD 23919), binaries hav-ing been previously removed from the candidate list from Hipparcos flags and SIMBAD checks; 2) the full width athalf maximum of the CCF is larger than 15 km / s, which pro-hibit accurate radial velocity measurements (17 stars); – observations were stopped after two HARPS measurementsif: 1) the radial velocity di ff erence ∆ RV between the twomeasurements is larger than 2.5 km / s (no SB1 were found);2) ∆ RV < / s (113 stars); – observations were stopped after three HARPS measurementsif ∆ RV <
20 m / s (26 stars).Radial velocity variations larger than 20 m s − were mea-sured for 37 stars, i.e. 19% of our observed sample. However,those variations are mainly caused by stellar activity (see §§