s r

RANSITING EXTRA SOLAR PLANETS e TTRANSITING EXTRA--SOLAR PLANETS,, v r

FOLLOWFOLLOW THETHE FLAMESFLAMES …… e s b THE PAST YEAR WITNESSED A BLOSSOMING IN THE FIELD OF TRANSITING EXOPLANETS. UNTIL 2003, ONLY ONE O TRANSITING EXOPLANET WAS KNOWN, THE FAMOUS HD 209568, FIRST DETECTED BY RADIAL-VELOCITY SUR- VEYS. THEN IN QUICK SUCCESSION, SIX TRANSITING PLANETS WERE DISCOVERED AMONG THE CANDIDATES IDEN- m TIFIED BY PHOTOMETRIC TRANSIT SURVEYS. FIVE OF THESE TRANSITING EXOPLANETS WERE CHARACTERIZED BY o

OUR TEAM WITH FLAMES ON THE VLT. r f s t

FRRÉDÉRICÉDÉRIC POONTNT1, FRRANÇOISANÇOIS BOOUCHYUCHY2, DIIDIERDIER QUUELOZELOZ1, CLLAUDIOAUDIO MEELOLO3, r

NUNO SANTOS4, STEPHANE UDRY1, MICHELICHEL MAYOR1, CLAIRE MOUTOU2 o p

1OBSERVATOIRE DE GENÈVE, SWITZERLAND; e 2 3 LABORATOIRE D’ASTROPHYSIQUE DE MARSEILLE, FRANCE; ESO R 4CENTRO DE ASTRONOMIA E ASTROFÍSICA DA UNIVERSIDADE DE LISBOA, OBSERVATÓRIO ASTRONÓMICO DE LISBOA, PORTUGAL

HE ORBITS OF SOME extra-solar ing their mass. FLAMES is a multi-fibre link monstrated that the apparent disadvantage of planets intersect the line of to the UVES and GIRAFFE spectrographs on targeting faint planetary transit candidates sight between the star and the the VLT. can be compensated by their compact loca- observer, blocking part of the In a previous Messenger article (June tion on the sky that allow to benefit from the star’s light and producing a 2004) we presented the detection of the plan- full multiplexing capability of FLAMES. TdetectableT photometric transit. The probabil- ets OGLE-TR-113b and OGLE-TR-132b that At present, seven transiting planets are ity to have such a geometry is about 10% for we made with FLAMES. In this article we known (see Table 1). If we look carefully, short-period companions. After the first dis- summarize our whole campaign including the these measurements are puzzling in various coveries of planets on short orbits (often detection of two more planets (OGLE-TR- respects. HD 209458 is almost three times called “hot Jupiters”) many ground-based 10b and OGLE-TR-111b). Complete details less dense than Jupiter, a characteristic that is surveys have been initiated to search for tran- on this campaign may be found in a series of not yet understood. Three of the OGLE plan- sit signatures from these hot planets. articles (Bouchy et al. 2004, 2005; Moutou et ets (OGLE-TR-56b, 113b and 132b) have Transiting planets are especially precious al. 2004; Pont et al. 2004, 2005, 2005b). periods around 1.5 days, much shorter than to study the structure and the evolution of The main outcome of our FLAMES cam- the lower limit detected by Doppler surveys extra-solar planets because the transit signal paign is the identification and the measure- of around 3 days, revealing a new class of offers a unique way to measure directly the ment of four true planetary transits, but also “very hot Jupiter” (i.e. gas giants even clos- radius of the planet. If the transit signal can the measurement of the mass of many tran- er to their star than the “hot Jupiters” with be combined with radial-velocity measure- siting low-mass stellar companions initially periods above 3 days). All three of these ob- ments, a direct measurement of the mass of tagged as possible planetary transit. Consid- jects are also heavier than normal hot Jupiters. the planet is obtained as well, the usual sin(i) ering the stringent requirements needed to OGLE-TR-10b, OGLE-TR-111b and TrES-1 uncertainty being removed by the edge-on carry out this programme on both the accura- are more typical hot Jupiters, resembling inclination of the orbit. With both the mass cy of the radial velocity and the need for high those found in abundance by radial-velocity and radius measured, the planet mean densi- efficiency to quickly survey a large number searches. They have radii ranging from ty becomes known. This allows us to “flesh of faint objects, we enthusiastically report the Jupiter size to the anomalous HD 209458, in” exoplanets in much more detail than for unrivalled efficiency of FLAMES/UVES to showing that the density of hot Jupiters can the planets without transit, for which only an carry out such a programme. This facility de- vary even under similar orbital conditions. estimate of the minimum mass (M sini) is available. Table 1: The seven presently known transiting exoplanets and the instrumentation used to Since 2002 many transiting extra-solar measure their radial-velocity orbit – HIRES on the Keck telescope in Hawaii, ELODIE on the planet candidates were identified as a result 1.93-m at Haute-Provence Observatory, and FLAMES on the VLT. of the OGLE survey. To follow up on those Period Mass Radius Instrument for spectroscopy announcements, we decided to start a system- [days] [MJ] [RJ] atic survey of all OGLE candidates. Using OGLE-TR-56 1.21 1.45 ± 0.23 1.23 ± 0.16 HIRES ESO facilities and in particular FLAMES on OGLE-TR-113 1.43 1.35 ± 0.22 1.08 ± 0.06 FLAMES the VLT, we conducted intensive spectro- OGLE-TR-132 1.69 1.19 ± 0.13 1.13 ± 0.08 FLAMES scopic follow-up observations to detect the TrES-1 3.03 0.76 ± 0.05 1.04 ± 0.07 HIRES radial-velocity signatures of announced plan- OGLE-TR-10 3.10 0.57 ± 0.12 1.24 ± 0.09 FLAMES/HIRES et transit candidates, with the objective of HD 209458 3.52 0.69 ± 0.02 1.42 ± 0.12 HIRES/ELODIE assessing their planetary nature by measur- OGLE-TR-111 4.02 0.53 ± 0.11 1.00 ± 0.09 FLAMES © ESO – June 2005 19 The mass-radius diagram of exoplanets con- stitutes the main outcome of follow-up cam- paigns on transiting planet candidates (Fig-

ure 1). It has already spurred many studies HD 209458 attempting to relate the observed sizes and 10 56 TrES-1 132 densities to the incoming stellar flux, the tidal 111 effects and the presence or absence of a rocky 113 core.

THE OGLE TRANSIT SURVEY OGLE (Optical Gravitational Lensing Exper- iment) was originally conceived as a micro- lensing survey in the Galaxy and towards the Magellanic Clouds. Beginning in 2001, the Figure 1: The mass-radius relation for the seven Figure 2: Complete light curve, zoom on the tran- OGLE team started a survey for transiting known transiting extra-solar planets. sit, and finding chart for one OGLE candidate, OGLE-TR-132. Planets produce shallow transits planets, using the wide-field camera on the that require a high photometric accuracy and 1.3-m Warsaw telescope at Las Campanas in long-term stability. This can be difficult to achieve Chile. Fields in the Galactic disc were ob- in crowded fields. OGLE-TR-132 is a candidate very near the detection threshold of ground-based served every season, providing 64 transit- surveys that was found to harbour a planet. ing candidates towards the Galactic bulge from the 2001–2002 season, 73 from the tons can be collected to compute accurate ra- case of transiting candidates, the orbital peri- 2002–2003 season in Carina, 40 from the dial velocities, but we were faced with the od and phase are known from the photomet- 2003–2004 season in Centaurus and Musca problem of uncertainties related to the cen- ric transit signals, so that such measurements (Udalski et al. 2002ab, 2005). Photometric tring of the star in the slit, that prevented us are sufficient to constrain the orbital motion accuracies below 1% are obtained for up to from obtaining a radial velocity with an accu- of a planetary companion and measure its 100 000 objects in each season, allowing the racy better than 100 m/s. When FLAMES be- mass accurately. detection of transit signals of depth from a came available, the fibre scrambling effect on few per cent down to slightly below one the incoming light from the target solved SUPPORTING ACTORS IN THE OGLE PLANET per cent. the centring problems and made our follow- FOLLOW-UP: UVES AND FORS Dozens of ground-based planet transit up programme feasible, with the additional Although the main actor in our spectroscop- searches are currently under way, but the benefit of multiplexing (the ability to ob- ic follow-up is doubtlessly FLAMES/UVES, OGLE survey is dominating the scene. This serve seven targets simultaneously with the there are other supporting actors in this is probably the consequence of an adequate FLAMES link on UVES). movie. The first of them is FORS. combination of continuous observation with In 2003 and 2004, we resolved the nature The shape and depth of a transit light a 1-m-class telescope equipped with a large of 60 OGLE transiting candidates in just eight curve are related to the size of the transit- detector (8 k × 8 k) and long experience of nights with FLAMES and we found four plan- ing companion, but in some cases, such as the team in stable photometric reductions. ets. A mean rate of two nights per planet is OGLE-TR-132, the transit depth is so near to Most other surveys either rely on very small about ten times more efficient than radial- the detection limit that little information can telescopes covering large fields of several velocity survey programmes! Regular radial- be obtained on its actual shape (see Figure 2). square degrees, or larger telescopes available velocity survey programmes use typically In such cases, FORS can be used to obtain a only during a very limited time. one night per target through the duration of much better light curve, leading to im- the survey, and one planet is detected out of proved mass and radius determination. That OUR FLAMES FOLLOW-UP 20 targets on average, meaning that about is what we have done for OGLE-TR-132 (see OF OGLE TRANSITING CANDIDATES 20 nights are spent to get one planet. This Figure 3). With the benefits of the UT aper- The OGLE survey provided more than one large difference of efficiency reflects the ben- ture, the quality of the FORS camera and hundred planetary transit candidates, but we efit of the multiplex advantage provided by the good seeing of Paranal, we measured the found that most of these transits are in fact the FLAMES fibre link, and of course the transit shape at the milli-magnitude level, an eclipsing binaries. The measurements of the benefit of the previous photometric transit orbital motion are therefore essential to deter- survey. mine whether the transiting object is a star To optimize telescope use, we have also or a planet. The OGLE targets are between opted to react in real-time during the obser- 15 and 18 in V magnitude, and typical plan- vations. During the eight nights of the cam- etary orbits cause velocity variations of the paign we analyzed our spectra on the fly, so order of 100 m/s or lower. The observational that we could re-allocate one of the seven challenge posed by the OGLE follow-up is FLAMES fibre links to another candidate as therefore to obtain radial velocities more soon as we had enough information to iden- precise than 100 m/s for objects down to tify the transiting body as a stellar compan- 18th magnitude in V. ion. In 2002, before FLAMES was available, Thanks to the fibre link, we could reduce we attempted to measure a few of the most the systematics in the radial-velocity meas- promising OGLE candidates both with UVES urements to less than 30 m/s. When this sys- in regular slit mode and with HARPS. Due to tematic error is combined with the photon the faintness of the targets, the signal-to- noise error, we are reaching typical accura- noise per pixel in HARPS spectra was often cies of 40–60 m/s on each individual radial- rather low, causing a radial-velocity uncer- velocity measurement. Measurements with Figure 3: Light curve of the transit of OGLE-TR-132 with FORS2. To be compared with the OGLE tainty higher than 100 m/s due to photon such errors would be too large to search for a detection light curve on Figure 2. Individual errors noise. With UVES in slit mode, enough pho- planet without prior information, but in the and systematics are near the level of 1 millimag. 20 The Messenger 120 achievement comparable to space perform- Figure 4 (left): The mass-radius dia- ance for such a faint object. gram from stars to planets, showing the objects from our follow-up of the In order to measure the mass and radius of OGLE survey, solar-system gas the transiting body, one needs to have an indi- giants, bright transiting planets, bright cation of the mass and radius of the primary eclipsing binaries, and stars with interferometric radius determinations. star. In order to do so, high signal-to-noise stellar spectra (S/N about 100) are need- ed. Most of our spectra collected with FLAMES/UVES have a S/N that is too low to carry out a precise spectroscopic analysis (although they are good to have precise radi- Figure 5 (below): Two stars and three planets... The transiting stellar com- al velocities!). UVES in slit mode can be used panion OGLE-TR-122b has a size to collect higher S/N spectra, refining the comparable to Jupiter, and is smaller previous determination of masses and radii than the planet HD 209458b. It pro- for the primary and its companion. duces a transit light curve indistin- guishable from a planetary transit.

THE MASS-RADIUS DIAGRAM FROM STARS TO PLANETS The main product of our spectroscopic fol- low-up with FLAMES is of course the char- acterization of the five planets described above. But another important result was the characterization of 55 candidates that were not planets. This yielded two important by- products: – Mass and radius for several very low-mass stars, down to the Hydrogen-burning limit. – A complete image of the range of config- urations that can be confused with plane- tary transits. Our follow-up of the OGLE candidates has Figure 6 (left): Radial-velocity accu- yielded a very rich wealth of empirical infor- racy versus target magnitudes for different facilities used in the follow- mation in the mass-radius plots for low-mass up of faint transiting candidates: stars and planets (Figure 4), comparable to all FLAMES/UVES on the VLT, UVES in other previous studies combined. slit mode, HARPS on the ESO 3.6-m, HIRES on the Keck Telescope in Two very small stellar objects are par- Hawaii with a Thorium calibration (Th) ticularly interesting: OGLE-TR-122 and or an Iodine cell (I). The stars indi- OGLE-TR-123. With masses near the Hydro- cate the velocity dispersion due to gen-burning limit and radii comparable to orbital motion for detected transiting planets, normal hot Jupiters in black that of the hot Jupiters, they provide an obser- and “very hot Jupiters” in white. The vational confirmation of the theoretical pre- yellow band shows the zone most diction that stellar radii reach down to about relevant for the detection of normal hot Jupiters in the OGLE survey. 0.1 Rᓃ before entering the regime of a de- generate equation-of-state. These objects also confirm that in some cases it is not possible to distinguish a hot Jupiter transit from a stel- are too faint and the photon noise on the the most performing facility in the relevant lar transit from the photometry alone, and that velocity exceeds the amplitude of a typical range – even before taking into account the small stellar companions occur at frequencies planetary orbit. Slit spectrographs on 8-m or multiplex facility that allows several targets comparable to hot Jupiters (see Figure 5). 10-m class telescopes (UVES, HIRES) can to be followed at once! With FLAMES in con- This is a major source of concern for very collect the necessary S/N ratio to push the junction with the results from transit surveys, deep transit surveys (such as with the HST), photon noise low enough, but only at the cost ESO instrumentation has demonstrated its were spectroscopic confirmation is not pos- of large systematic uncertainties due to the capability to make a major contribution to the sible. centring of the object in the slit. The use of empirical knowledge of the mass-radius re- an Iodine cell can solve this problem, but only lation for planets, brown dwarfs and low- FLAMES AHEAD at the cost of losing transmission efficiency mass stars. The combination of 8-m-class telescope, and most of the spectral range, so that again fibre-feed, high resolution, simultaneous the photon noise becomes too large. As a re- ACKNOWLEDGEMENTS Thorium calibration, and multiplex capacity, sult, even if it had a single fibre, FLAMES is We wish to express our gratitude to all persons puts FLAMES/UVES in a league of its own the only facility able to derive a planetary or- that participate in the development of the UVES FLAMES link, making available to the ESO com- for the follow-up of deep transit surveys. No bit in five measurements for an object such munity such a unique and unrivalled facility. other facility in the world combines both the as OGLE-TR-132. Adding to that, the 7-fiber faint multi-object capability and the capaci- multiplex capacity makes it the king of the REFERENCES ty to measure accurate radial-velocity. jungle for transit follow-up of faint targets. Bouchy et al. 2004, A&A Letters 421, L13 Specialized radial-velocity spectrographs Figure 6 presents a graphic comparison of Moutou et al. 2004, A&A Letters 424, L31 Pont et al. 2004, A&A Letters 426, L15 on smaller telescopes, such as HARPS on the the capacity of some facilities for radial- Bouchy et al. 2005, A&A 431, 1105 ESO 3.6-m at La Silla, can reach a higher velocity follow-up of faint planetary candi- Pont et al. 2005, A&A Letters 433, 21 absolute precision, but most OGLE targets dates. This diagram shows that FLAMES is Pont et al. 2005, A&A, in press, astro-ph/0501615 Pont F. et al., Transiting extra-solar planets © ESO – June 2005 21