Astronomical Science

Behind the Scenes of the Discovery of Two Extrasolar Planets: ESO Large Programme 666

Dante Minniti 1, 2 ets and brown dwarfs), and to measure The telescopes and instruments Claudio Melo3 their masses, radii, and mean densities. Dominique Naef 3 We hunt selected OGLE transit candi- We use the Very Large Telescope UT2 Andrzej Udalski 4,5 dates using spectroscopy and photom- in order to measure radial velocities, and Frédéric Pont 6 etry in the ‘twilight zone’, stretching UT1 and 2 for photometry in order to Claire Moutou 7 the limits of what is nowadays possible measure the transits. Most of the spec- Nuno Santos 8 with the VLT. troscopic observations required real-time Didier Queloz 6 decisions to be taken and therefore were Tsevi Mazeh 9 made in visitor mode, whereas photom- Michel Gillon 6 The programme etry collected at precise transit times was Michel Mayor 6 carried out in service mode by expert Stephane Udry 6 Transiting extrasolar planets are essential mountain personnel. Rodrigo Diaz10 to our understanding of planetary struc- Sergio Hoyer11 ture, formation and evolution outside the The spectroscopic runs were done with Sebastian Ramirez1 Solar System. The observation of transits FLAMES in GIRAFFE mode which allows Grzegorz Pietrzynski 4,12 and secondary eclipses gives access to simultaneously more than 100 to Wolfgang Gieren12 such quantities as a planet’s true mass, be observed at resolutions ranging from Maria Teresa Ruiz11 radius, density, surface temperature and about 5 000 to 20 000. At the same time, Manuela Zoccali1 atmospheric spectrum. the other seven to eight fibres feed UVES Omer Tamuz 9 at the other Nasmyth focus of the tele- Abi Shporer 9 The OGLE search for transiting planets scope, collecting spectra with a resolu- Marcin Kubiak 4,5 and low-mass stellar companions has tion of 50 000. Both GIRAFFE and UVES Igor Soszynski 4,5 been the first photometric transit survey allow us to acquire spectra of a compari- Olaf Szewczyk 4,5 to yield results. Follow-up of existing son lamp simultaneously with the target Michal Szymanski 4,5 OGLE low-amplitude transit candidates observations. The lamp spectrum is used Krzysztof Ulaczyk 4,5 prior to our Large Programme has un - afterwards to correct spectrograph shifts. Lukasz Wyrzykowski 5,13 covered five extrasolar planets and has This is essential in order to be able to yielded the measurement of their radii measure with a precision of and therefore their densities. Despite a few m/s. Accurate radial velocity meas- 1 Departamento de Astronomía y these successes, many important points urements require high-resolution spectra, Astrofísica, Pontificia Universidad remain to be understood: how hot Jupi- for this reason we placed the best can- Católica de Chile, Santiago, Chile ters form, how they evolve, what is the didates in the UVES fibres in order to 2 Specola Vaticana, Città del Vaticano, frequency of hot Jupiters, why the density measure more accurate velocities. The Italy range of hot Jupiters is so large, etc. The photometric runs were acquired with 3 ESO main difficulty in answering these ques- FORS1 or FORS2, yielding milli-magni- 4 Warsaw University Observatory, Poland tions is the limited number of transiting tude photometry and a high-quality light 5 The OGLE Team planets detected so far. curve, essential to derive accurate physi- 6 Observatoire de Genève, Sauverny, cal parameters for the transiting planets. Switzerland Our Large Programme 177.C-0666 7 Laboratoire d’Astrophysique de (LP666 for short) proposed to enlarge Marseille, France this sample of confirmed OGLE extra- The team preparations 8 Centro de Astrofísica, Universidade do solar planets, and also to populate the Porto, Portugal mass-radius diagram for low-mass ob- Three teams were competing for the 9 School of Physics and Astronomy, jects, including planets, brown dwarfs same resources: the OGLE team, the R. and B. Sackler Faculty of Exact and late M-type stars. 177 transiting can- Geneva team, and the Chilean team. The Sciences, Tel Aviv University, Israel didates from the OGLE survey had been idea arose of working together and the 10 Instituto de Astronomía y Física del published when we started this LP666, final details were discussed at the work- Espacio, Buenos Aires, Argentina and as part of this programme three new shop in Haute-Provence “10 years of 11 Departmento de Astronomía, OGLE seasons produced 62 new candi- 51 Peg”. In preparation for this Large Universidad de Chile, Santiago, Chile dates. Programme, we had to re-examine the 12 Departamento de Astronomía, OGLE database to check old candidates Universidad de Concepción, Chile We used VLT+FLAMES to obtain radial and improve the ephemeris, and to run 13 Institute of Astronomy, University of velocity orbits in order to measure the OGLE pipelines to select new can- Cambridge, United Kingdom the mass of all interesting OGLE transit- didates. Our teams have put together ing candidates, and also VLT+UVES their respective spectroscopic and pho- and VLT+FORS to measure their precise tometric databases in order to select the This is the story of the Large Pro- radius from high-resolution spectroscopy most promising candidates, but it was gramme 666, dedicated to discover of the primary and high-definition transit not easy to decide which were these best sub-stellar objects (extrasolar plan - light curves. candidates, and this was extensively dis-

The Messenger 133 – September 2008 21 Astronomical Science Minniti D. et al., ESO Large Programme 666

cussed by many team members. The (2005) for , and Bouchy et al. can didates from season VI (up to OGLE- spectroscopic run of February 2006 by (2005) for the Galactic Bulge. In addition, TR-238) are still being analysed. the Swiss team was used to cull some the flexibility of the OGLE telescope candidates. The photometric run of the has allowed us to obtain the candidates The spectroscopic runs had two goals: last period of March 2006 of the Chilean needed, producing a new set of candi- (1) to sort out the non-planetary can- team was used to observe some of dates for this LP666, which we have fol- didates; and (2) to measure an orbital these most promising candidates. We lowed-up during periods P78–P82. These motion for the planetary candidates. The finally started the planet hunting for were selected carefully among periodic need to discard as quickly as possible this Large Programme in April 2006 with low-amplitude transits observed with the impostors required that data reduc- many promising candidates. the Warsaw telescope during the last tion and radial velocity measurement had seasons: season IV from candidates to be done in real time. This was achieved The first step is to acquire radial-velocity OGLE-TR-178 to TR-200; and season V by a combination of the FLAMES/UVES information for the most promising OGLE from candidates OGLE-TR-201 to pipeline and our own code. The previous planetary transit candidates, and to iden- TR-219. These are listed in Table 1. New nights results were analysed by our team tify real transiting planets among them. This requires five to eight radial velocity points with UVES in good observing con- Object Period I Depth Status ditions. As the OGLE candidates are dis- OGLE-TR-178 2.97115 16.56 0.016 faint target, not observed posed in a few square degrees of the sky, OGLE-TR-179 12.67106 15.13 0.034 flat CCF a few of them (typically two to five targets) OGLE-TR-180 1.99601 16.74 0.012 faint target, not observed can be observed simultaneously using OGLE-TR-181 2.3896 16.29 0.01 fast rotator (synch.?) the FLAMES configuration. The weather OGLE-TR-182 3.98105 15.86 0.01 transiting planet conditions of this first run were rather OGLE-TR-183 4.78217 15.32 0.015 fast rotator (synch.?) poor, with five clouded nights and three OGLE-TR-184 4.92005 15.57 0.015 fast rotator (synch.?) clear nights. As our targets are faint and OGLE-TR-185 2.78427 16.72 0.035 fast rotator (synch.?) we need < 100 m/s radial velocity accu- OGLE-TR-186 14.81481 16.54 0.054 faint target, not observed racy, the clouded nights were of very lim- OGLE-TR-187 3.45686 14.07 0.008 double-lined spectroscopic binary (SB2) ited use, despite the occasional gaps in OGLE-TR-188 6.87663 16.38 0.031 blend of two line systems the clouds. OGLE-TR-189 1.73937 15.03 0.006 not observed OGLE-TR-190 9.38262 16.06 0.043 not observed This Large Programme had another com- OGLE-TR-191 2.51946 15.57 0.007 fast rotator (synch.?) ponent, many hours of service observa- OGLE-TR-192 5.42388 14.41 0.008 flat CCF tions on FORS to obtain high-accuracy OGLE-TR-193 2.95081 14.99 0.008 not observed measurements of the transits of the plan- OGLE-TR-194 1.59492 14.69 0.006 flat CCF ets that we expected to discover. In fact, OGLE-TR-195 3.62174 14.19 0.006 not obseved the results of these initial runs suggested OGLE-TR-196 2.1554 15.57 0.012 fast rotator (synch.?) three possible planets, but because of OGLE-TR-197 2.40587 14.59 0.019 flat CCF bad weather we could not reach solid OGLE-TR-198 13.63141 15.44 0.018 not observed conclusions on these objects. Due to the OGLE-TR-199 8.8347 14.88 0.017 single-lined spectroscopic binary (SB1) bad weather, it was more important OGLE-TR-200 6.48845 15.63 0.023 not observed at this initial stage to recover some of the OGLE-TR-201 2.368 15.6 0.016 fast rotator spectroscopic time that was lost. We OGLE-TR-202 1.6545 13.6 0.017 not observed therefore requested to ESO to swap some OGLE-TR-203 3.3456 15.6 0.014 not observed of our photometric time for spectroscopic OGLE-TR-204 3.1097 14.8 0.026 SB2 time in service mode, a request that was OGLE-TR-205 1.7501 16 0.015 not observed kindly (and quickly) approved. OGLE-TR-206 3.2658 13.8 0.006 no variation OGLE-TR-207 4.817 14.3 0.021 SB2 OGLE-TR-208 4.5025 15.3 0.022 SB2 The candidates OGLE-TR-209 2.2056 15 0.022 no variation Table 1. List of targets OGLE-TR-210 2.2427 15.2 0.032 fast rotator selected for follow-up We measured and analysed the various OGLE-TR-211 3.6772 14.3 0.008 planet from OGLE seasons IV candidates found during the OGLE OGLE-TR-212 2.2234 16.3 0.016 blend? and V. The photometric season III, from OGLE-TR-138 to OGLE- OGLE-TR-213 6.5746 15.3 0.036 SB2 period, I-band magni- tude and transit depth TR-177. This was done using new data OGLE-TR-214 3.601 16.5 0.023 SB1 are based on the OGLE from the present LP666 in combination OGLE-TR-215 4.9237 14.8 0.016 no variation data. The last column with data already in hand. A detailed pub- OGLE-TR-216 1.9763 14.6 0.011 blend? is our final assessment lication for these candidates is in prepa- OGLE-TR-217 5.7208 16.1 0.037 no CCF of the status of the OGLE transit candidates ration. This paper would be a large effort, OGLE-TR-218 2.2488 14.5 0.02 fast rotator after the spectroscopic like the previous papers by Pont et al. OGLE-TR-219 9.7466 15.1 0.032 SB2 follow-up with FLAMES.

22 The Messenger 133 – September 2008 members in Europe. Based on their feed- 177.C-0666 FLAMES/P78 Figure 1. The cross-cor- back, FLAMES configuration files were 1 relation functions (CCF) from the spectra of the updated or simply discarded. We will 0.98 different candidates here discuss the first two confirmed plan- shown as an example of 0.96 ets, but along this work we have analysed all the typical bad CCF about 100 OGLE candidates (from OGLE- 0.94 cases. 0.92 TR-138 to OGLE-TR-238), including sev- Broad Bad #1 Asymmetric Bad #2 SB2 Bad #3 No Dip eral new ones that were selected for this 0.9 n o programme from OGLE seasons IV, V i

t 1 c and VI. n u

F 0.98

n o i t 0.96 a l e r r

The non-planets r 0.94 o C -

s 0.92 s o A vast number of candidates are pro- r Bad #4 No Dip Bad #5 No Dip Bad #6 Broad duced by OGLE, which needed to C 0.9 be confirmed spectroscopically and pho- 1 tometrically. Therefore, a large part of 0.98 the work we did was dedicated to elimi- 0.96 nate non-planetary candidates from the 0.94 OGLE selection. 0.92 Broad Broad Bad #7 SB2? Bad #8 Dip? Bad #9 Faint The shape of the cross correlation func- 0.9 tion (CCF) can be efficiently used for bad –100 –5050–010 0 10 01–5050000 –100 –5050010 0 Radial Velocity (km/s) candidate detection and rejection. Fig- ure 1 represents the CCFs obtained for some new P78–P80 targets for which 177.C-0666 FLAMES/P78 Figure 2. The CCF of a we were able to conclude, after a single good candidate (OGLE- 1 TR-235). FLAMES observation, that a follow- up was useless. All these data were ob- tained in the February 2007 run, when

five new fields were started. Only one con- n o i tained a good candidate out of 10 new t c n objects; that is OGLE-TR-235, shown in u F

0.8

Figure 2. From the photometric point of n o i t view, the light curve produced by OGLE a l e r was always compatible with a planet. r r o

That is one of the reasons why a follow- C - s

up is mandatory (the other important s o reason being the need to determine an r C accurate planetary mass, of course). 0.6

Following the procedure described in the papers, we have discarded several candi- dates. For example, we analysed 42 new OGLE candidates, finding two planets –100 –50 0 50 10 0 plus the following: Radial Velocity (km/s) – obvious or suspected single- or double- lined spectroscopic binaries (blend, FLAMES has been very efficient in identi- planetary candidates and measuring their binary with high-mass companion): fying bad candidates. It is not an exag- parameters. In the near future we would 11 candidates; geration to say that FLAMES-UVES is in a go back and sift through the whole data- – broad CCF (synchronised, fast rotator): league of its own for follow-up of faint set, looking for low-mass companions 9 candidates; transit candidates, such as the ones dis- (M-dwarfs and brown-dwarfs), fitting the – no CCF dip detected, no radial velocity covered by OGLE. We have observed best parameters. We expect some more (false detections, faint candidates): the complete sample of all accessible interesting results, other than the few new 8 candidates; new OGLE candidates. We note that all planets that we have found, which are – too faint candidates, outside the field the data acquired with FLAMES-UVES still being analysed, along with hundreds (not observed): 11 candidates; are fully reduced already. We have so far of eclipsing binary stars, for which we – others: 3 candidates. concentrated on looking for the best have radial velocities.

The Messenger 133 – September 2008 23 Astronomical Science Minniti D. et al., ESO Large Programme 666

The first planet OGLE-TR-182-b Figure 3. Transit light curves for OGLE- 1.01 TR-182 from OGLE (top panel), and FORS1 for the nights of 6 June (middle 1 After the first year, we had discarded panel), and 9 June 2007 (bottom several candidates, but we also had ten- 0.99 panel). tative orbits for three planet candidates. These good ones, however, did not have 0.98 enough velocities to exclude random radial velocity fluctuations as the cause of the orbital signal. In addition, the ephem- x u l eris was not confirmed by the photome- F try: there was a slight discrepancy, which required collection of more data to re- solve. Then we started the year 2007, full 1.01 of hopes about these three objects. Below we tell the story of the first two. 1

0.99 The masses and radii of the stars were not accurately measured, and therefore 0.98 we could not estimate precisely the –0.05 0 0.05 masses and radii of our good planet can- Phase didates. We again asked ESO to allow a change in strategy, swapping a few hours Planet ID OGLE-TR-182-b OGLE-TR-211-b Table 2. Measured of FORS time into UVES time in order Period (days) 3.97910 +/– 0.00001 3.67724 +/– 0.00003 parameters for the new planets. to get high S/N echelle spectra of a cou- Transit (JD) 2454270.572 +/– 0.002 2453428.334 +/– 0.003 ple of stars to confirm that they are main- RV semi-amplitude (m/s) 120 +/– 17 82 +/– 16 sequence stars and to measure their Semi-major axis (AU) 0.051 +/– 0.001 0.051 +/– 0.001 parameters (temperature, gravity, lumi- Radius ratio with primary 0.102 +/– 0.004 0.085 +/– 0.004 nosity, and chemical abundances). ESO Orbital inclination angle 85.7 +/– 0.3 > 82.7 kindly and quickly approved our change, Planet Radius (RJ) 1.13 (+ 0.4 – 0.06) 1.36 (+ 0.18 – 0.09) which allowed us to measure the stellar Planet Mass (MJ) 1.01 +/– 0.15 1.03 +/– 0.20 mass, M = 1.14 +/– 0.05 M, and radius, R = 1.14 + 0.23 – 0.06 R. Figure 4. Radial velocity curve for The main effort has been devoted to 22.6 OGLE-TR-182 measured with FLAMES, phased with the photometric transit obtaining the ephemeris, which for the signal. two most promising candidates has proven to be difficult. In one case we sus- pect that we missed the transit because 22.4 s) /

it was observed during the flat bottom m k (

r portion before the observations ended, V and in another case we just caught the ingress, but we observed a full transit on 19 June 2007 (Figure 3). 22.2

Finally, we succeeded in confirming can- didate OGLE-TR-182-b as a planet, in 0 0.2 0.4 0.6 0.8 1 a paper that also contains analysis Phase of OGLE-TR-178 to TR-200 (Pont et al., 2008). In Figure 4 we show the fit to The second planet OGLE-TR-211-b the FLAMES orbital fit for OGLE-TR-211 the OGLE-TR-182 radial velocities ob- obtained with a radial velocity amplitude tained with P = 3.9789 days. This object The second planet was not easier to con- of 82 m/s and period P = 3.67718 days. was very difficult to confirm because firm, but we managed to conclude that It can be seen that the velocity curve is the period is almost a multiple of one day. OGLE-TR-211 is also a planet, in a paper very well sampled, from which we meas-

Table 2 summarises the derived parame- that contains the analysis of candidates ure a planetary mass of M = 1.0 MJupiter. ters. OGLE-TR-201 to OGLE-TR-219 (Udalski et al., 2008). Again, we used UVES to The photometric transit measured with determine the stellar parameters: mass FORS (Figure 6) gives a radius of R =

M = 1.33 +/– 0.05 M, and radius R = 1.36 RJ (see Table 2). This radius is about 1.64 + 0.21 – 0.07 M. Figure 5 shows 20 % larger than the typical radius of

24 The Messenger 133 – September 2008 Figure 5. Radial velocity of impostors in an efficient way, and that curve for OGLE-TR-211 V = 18.827 +/– 0.11 km/s K = 81.5 +/– 16 m/s the candidates that cannot be observed 0 measured with FLAMES. 19 spectroscopically will remain candidates. In particular, we pushed to the limit the FLAMES radial velocity capabilities, and after careful characterisation of these 18.9

) radial velocities, found a precision around s / 30–50 m/s. More recently, we under- m k ( stood why these two (or three) planets re-

r 18.8 V quired so much time to be confirmed. They lay in a place that we called the Twi- light Zone (Pont et al., 2008). When both 18.7 the photometric and spectroscopic signals are marginal, many more obser- vations are necessary until reasonable 18.6 certainty can be achieved about the 0 0.5 1 1.5 2 presence of a planetary companion. The Phase uncertainties on the light curve make it difficult to phase the radial velocity data. hot Jupiters with similar masses. Indeed, ets around metal rich stars (Torres et The high radial velocity uncertainties this new planet seems to be one of the al., 2008). The issue is clearly not settled hinder the identification of an orbital rare examples of an inflated hot Jupiter, and, in this sense, every new planet motion with the correct period, and the with an unusually low mean density, like counts. elimination of eclipsing binary blend sce- HD 209458b. In this case there is some narios. The OGLE survey is the first to evidence that the velocity of the centre of The question can be asked if it isn’t more explore this ‘twilight zone’ in real condi- the mass of the -planet orbit is chang- efficient to have a space mission like tions, since other ground-based surveys ing. Also, slight variations of transit times COROT that focuses on brighter targets? target brighter stars, for which very pre- lead us to suspect that there may be Aside from the obvious answer that cise radial velocities can be obtained, so another massive planet in the system that a space mission is much more risky and that the significance of the radial velocity may be perturbing the orbit, but only expensive, the LP666 programme, signal can be established relatively easily. future observations can confirm if this is and the observations carried out prior to All this savoir-faire is already being used real (e.g. due to a companion in the sys- it, helped the whole team to gain enor- in the COROT mission, and would again tem), or to an instrumental artefact. mously in experience. The team learned be useful when the KEPLER mission flies. how to obtain milli-magnitude photom- etry, and more importantly, to understand Lessons learned the systematics involved in these kinds References of observations (e.g. the red-noise – Bouchy, F., et al. 2005, A&A, 431, 1105 We discovered two new planets! We can Pont, 2006). This understanding led to a Guillot, T., et al. 2006, A&A, 453, L21 now tell with certainty that they are gas- better (more realistic) prediction of the Pont, F. 2006, MNRAS, 373, 231 eous planets like Jupiter and Saturn, and space mission yields. Pont, F., et al. 2005, A&A, 438, 1123 not rocky like Earth or Venus, nor most- Pont, F., et al. 2008, A&A, in press Torres, G., et al. 2008, ApJ, 677, 1324 ly liquid like Uranus or Neptune. Transit- In terms of spectroscopic follow-up we Udalski, A., et al. 2008, A&A, 482, 299 ing planets are the only way to get infor- learned how to identify the signatures mation about their internal structure (and therefore mechanism of formation). De- spite the scarce number of such plan- OGLE-TR-211 ets, the number starts to be large enough to reveal interesting features. For exam- 14.51 ple, it seems that in order to explain the observed radii, one needs to add a cer- 14.52 tain quantity of metals in the form of a ) g solid core. It turns out that the mass of a 14.53 m (

this solid core is proportional to the met- I allicity of the parent star. This trend sup- 14.54 ports the mode of formation via core accretion (Guillot et al., 2006). However 14.55 it seems that metallicity is not the only Figure 6. Light curve of the transit of OGLE- parameter controlling the size of these 0.92 0.94 0.96 0.98 1 1.02 1.04 1.06 1.08 TR-211 measured with giant planets, since there are inflated plan- Phase FORS2.

The Messenger 133 – September 2008 25