SSETTINGETTING NNEWEW SSTTANDANDARDSARDS WITHWITH HARPSHARPS

BY OCTOBER 1ST, 2003, ESO'S NEW AND UNIQUE PLANET-HUNTING MACHINE HARPS (HIGH-ACCURACY PLANETARY SEARCHER) HAS BECOME OPERATIONAL. THE MEASUREMENTS MADE DURING THE COMMISSIONING PHASE AND THE FIRST WEEKS OF OPERATION ARE OF OUTSTANDING QUALITY. IN THIS ARTICLE WE REPORT AMONG OTHER EXAMPLES ON THE FIRST EXTRA-SOLAR PLANET DISCOVERED WITH HARPS AND ON THE DETECTION OF TINY STELLAR OSCILLATIONS. THE RESULTS PRESENTED DEMONSTRATE THAT HARPS IS CURRENT- LY THE MOST PRECISE DOPPLER-MEASUREMENTS MACHINE IN THE WORLD. WITH THIS ACQUISITION ESO PLACES ITSELF AT THE HEAD OF A SCIENTIFIC DOMAIN, WHOSE INTEREST HAS CONTINUED TO GROW DURING THE PAST .

1 1 HE HARPS PROJECT was AN UNEQUALLED FACILITY FOR M. MAYOR1, F. PEPE1, born in May 1998 when ESO RADIAL-VELOCITY MEASUREMENTS 1 2 D. QUELOZ1, F. BOUCHY2, issued an Announcement of HARPS is a fibre-fed, cross-dispersed 3 Opportunity asking for the echelle spectrograph. Two fibres, an object G. RUPPRECHT3, G. LO design, construction, and pro- and a reference fibre, feed the spectro- 4 3 CURTO , G. AVILA , TcurementT of an instrument dedicated to graph with the light from the telescope and the search for extrasolar planets and aim- W. BENZ5, J.-L. BERTAUX6, the calibration lamps. The fibres are re- ing at an unequalled precision of 1 m/s. In imaged by the spectrograph optics onto a 1 4 X. BONFILS , TH. DALL , response to ESO’s call the Observatoire de mosaic of two 2k4 CCDs, where two H. DEKKER3, B. DELABRE3, Gen `eve together with the Physikalisches echelle spectra of 72 orders are formed for Institut der Universität Bern, the Observa- 4 1 each fibre. The covered spectral domain W. ECKERT , M. FLEURY , toire de Haute-Provence, and the Service ranges from 380 nm to 690 nm. The resolu- A. GILLIOILLIOTTE4, D. GOJAK3, d’Aéronomie du CNRS have formed a tion of the spectrograph is given by the Consortium which realized, in collabora- fibre diameter and attains a value of about UZMAN4 OHLER7 J.C. GUZMAN , D. KOHLER , tion with ESO, this ambitious project in R=115,000.At this resolution each spectral J.-L. LIZONIZON3, A. LONGINOTTI3, only three and a half years.The instrument element is still sampled by 3.2 CCD pixels. was installed by the Consortium on ESO’s A summary of the main HARPS features C. LOVIS1, D. MÉGEVAND1, 3.6-m Telescope at La Silla in January 2003 is given in Table 1. For a detailed descrip- 3 3 L. PASQUINI3, J. REYES3, and first light took place on February 11th. tion of the instrument we refer to the arti- 7 1 HARPS was commissioned during the cle published in Pepe et al. (2002) and to J.-P. SIVIVAN7, D. SOSNOWSKA1, periods 70 and 71. During period 71 a first the HARPS Users Manual. 4 1 R. SOTO , S. UDRY , GTO run of seven nights had been allocat- HARPS is an ordinary spectrograph ed for the Consortium as well. These A. VAN KESTEREN3, with outstanding efficiency and spectral observations have already produced many resolution. The main characteristics of 1 4 L. WEBER , U. WEILENMANN exciting results. HARPS is however its extraordinary sta- The instrument was handed over to bility. The ThAr-reference technique is ESO La Silla by the end of September able to measure and correct the tiniest 2003 and made available to the Communi- instrumental drifts; nevertheless a lot of 1OBSERVATOIRE DE GENÈVE, ty for period 72. Together with the 3.6-m effort was put into making the spectro- WITZERLAND S ; telescope HARPS is now delivering fan- graph intrinsically stable, in order to avoid 2LABORATOIRE D'ASTROPHYSIQUE tastic scientific frames to DE MARSEILLE, FRANCE; the observers. But this is not the only data product Table 1: HARPS spectrograph characteristics 3ESO GARCHING; of HARPS. The powerful 4ESO LA SILLA; Optical design fibre-fed, cross-dispersed HARPS pipeline provides echelle spectrograph 5 any observer with extract- PHYSIKALISCHES INSTITUT DER Technique simultaneous ThAr Reference UNIVERSITÄT BERN, SWITZERLAND; ed and wavelength cali- brated high-resolution Number of fibres 2 6SERVICE D'AÉRONOMIE, spectra, as well as the pre- Fibre aperture on sky 1 arcsec VERRIÈRES LE BUISSON, FRANCE; cise radial velocity of the Collimated beam diameter 208 mm 7 OBSERVATOIRE DE HAUTE- observed .The observ- Covered spectral range 380 nm to 690 nm PROVENCE, FRANCE er will thus leave the Spectral resolution R=115,000 Observatory with fully Spectral format 72 echelle orders reduced spectra. Addi- 61.44 x 62.74 mm tional information on how CCD chip mosaic, 2xEEV2k4 to observe with HARPS pixel size=15µm can be found at Sampling 3.2 pixels/SE http://www.ls.eso.org/lasilla/ Min. inter-order 33 pixels sciops/harps/. 20 The Messenger 114

The extraordinary stability translates directly into the stability of the radial velocity measurements. This is most impressively shown by the measurement series presented in Figure 2b. Both fibres have been exposed repetitively with the ThAr calibration light and the drift – expressed in m/s – was computed as a func- tion of time. During several hours the total drift remains well below 1 m/s and the measurement is dominated by “noise”.This noise is introduced by the CCD whose temperature varies by ±0.02 K and pro- duces microscopic dilatation of the chip. In fact, both fibres show the same behaviour and when subtracting one fibre from the other this “noise” disappears and we obtain the dispersion values expected from the photon-noise level of the simultaneous ThAr reference. These measurements prove that the simultaneous ThAr refer- ence technique is perfectly able to track instrumental drift at a level of 0.1 m/s rms. Figure 1:The HARPS spectrograph inside the air-conditioned room at La Silla just before being HARPS has been optimised for radial- closed and evacuated. velocity efficiency, i.e. to obtain in the shortest possible exposure time the most as far as possible any kind of second-order uum vessel controls the air temperature to precise radial-velocity measurement. How- instrumental errors. As a consequence, we 17gC with a long-term stability of the order ever, this requirement is not necessarily have decided to operate the spectrograph of 0.01gC. Because of the huge thermal synonymous with optical efficiency. For in vacuum (see Figure 1), since ambient inertia of the vacuum vessel and the excel- example, spectral resolution is an impor- pressure variations would have produced lent thermal insulation between spectro- tant factor in reducing the photon noise of huge drifts (typically 100 m/s per mbar). graph and vessel provided by the vacuum, the radial-velocity measurement but is The operating pressure is always kept the short-term temperature stability obtai- often in competition with the optical effi- below 0.01 mbar such that the drifts never ned is even better. Over one we have ciency of the spectrograph or with the size will exceed the equivalent of 1 m/s per day. measured variations of the order of 0.001K of the instrument and the telescope. On Not only the pressure but also the tem- rms.An example of the excellent control is the other hand the instrumental errors perature is strictly controlled. In fact, a given by the behaviour of the echelle grat- must also be low, at least as low as the goal two-stage air-conditioning around the vac- ing temperature plotted in Figure 2a. for the photon-noise equivalent. The

Figure 2: a) Tempera- ture of the echelle grating over 24 hours recorded at two differ- ent days. b) Drift of the spectrograph dur- ing a ThAr series of several hours proving that the spectrograph is extremely stable and that the simulta- neous ThAr reference is able to track and correct tiniest instru- mental drifts.

Figure 3: a) Total effi- ciency of HARPS (right-hand ordinate) and efficiency of single components (left-hand ordinate). b) Compari- son of the measured SNR with the values calculated by means of the Exposure Time Calculator. For com- parison we performed a 71s exposure on

HD10700 (mv=3.5) at an airmass of 1.1. The DIMM monitor was indicating a seeing of 0.9 arcsec. © ESO - December 2003 21

1.7 m/s and partly due to photon noise. The residual noise is a combination of cal- ibration errors (< 0.5 m/s) and effects due to the star itself, which can be due to pul- sations, activity and jitter. It is also worth noting that the HARPS vaccum has been broken between the two commissioning runs. We conclude that the short-term pre- cision (1 night) of HARPS is well below 1 m/s. Even on the time scale of a commis- Figure 4: a) Series of 7 hours and 420 exposures on α Centauri B proving the extraordinary short-term precision of HARPS. b) Zoom of figure a) to illustrate the presence of a periodic sig- sioning period (2 weeks) we have been nal produced by the stellar pulsation. able to measure standard which showed radial-velocity dispersions of 1 m/s including photon noise. The same level of precision is obtained on HD 83443 over a time scale of 4 months. Apart from photon noise a major contri- bution to the residuals probably comes from the star itself (pulsations, activity,jit- ter). Only an in-depth, long-term study will allow us to identify the relative importance of the various error sources.

Figure 5: a) Power spectrum of α Cen B. The acoustic modes corresponding to the 4-minutes CATCHING THE TINY MELODY oscillation are clearly identified and emerge well above the noise. b) Autocorrelation of the OF A SOLAR-LIKE STAR power spectrum of α Centauri B. Stars, which are spheres of hot gas, prop- agate very well in their interiors acoustic design of HARPS is consequently the that the average contribution of modes to waves which are generated by turbulent result of an accurate trade-off between all the total dispersion reaches 0.44 m/s. By convection near their surfaces. Frequen- the relevant factors. Despite various com- quadratic subtraction of this value from cies and amplitudes of these acoustic promises regarding for example the fibre the measured dispersion we obtain a waves, also called oscillation modes or p- diameter (1 arcsec on the sky) and the noise level of 0.26 m/s, which is in good modes, depend on the physical conditions fibre-feed itself, the optical efficiency agreement with the value extrapolated prevailing in the layers crossed by the obtained is remarkable. Figure 3a shows from the mean white noise level meas- waves and provide a powerful seismolog- the calculated total optical efficiency of ured in the high frequency range between ical tool. Helioseismology, which moni- the instrument and some of its subsys- 6 and 8 mHz in the power spectrum. The tors the oscillation modes of our Sun, has tems. These values have been adopted in photon noise on a single measurement is been used since the 1970’s and led to the current version of the HARPS-Expo- of 0.17 m/s, which leaves less than 0.2 m/s major revisions in the “standard model” sure Time Calculator (ETC) which serves for all other possible error sources (ThAr of the Sun and provided, for instance, to estimate the SNR obtained during an noise, guiding errors, influence of the measures of the Sun’s inner rotation, the exposure. In Figure 3b we compare the atmosphere, instrumental errors). size of the convective zone and the struc- results obtained using the ETC with real The long-term precision of the instru- ture and composition of the external lay- measurements on the star HD 10700. ment cannot be checked easily because it ers. Solar-like oscillation modes generate requires a long time base on one hand, periodic motions of the stellar surface HARPS DELIVERING PRECISION and the knowledge of stable stellar with periods in the range 3 − 30 min., but NEVER REACHED BEFORE sources on the other hand. Especially the with extremely small amplitudes.The cor- During the HARPS commissioning we latter point represents a new challenge responding amplitudes of the stellar sur- monitored the star α Centauri B, which is since the intrinsic stability of the stars has face velocity modulations are in the range a K-type star substantially smaller than never been studied at this level of preci- 10 − 100 cm/s. our Sun. During 7 hours we collected a sion before. Nevertheless we have total of 420 spectra with typical SNR of been able to gain some indication of 500 each at λ=550 nm. The radial-velocity the long-term precision of HARPS measurement sequence plotted in Figure by observing the star HD 83443 4a indicates a dispersion of 51 cm/s but which has a planetary companion the zoom shown in Figure 4b shows that with well known orbital parameters. this dispersion is completely dominated Figure 6 shows a total of nine radial- by 4-minutes stellar oscillations. In fact, velocity data points, some of them the power spectrum of this sequence collected during the first commis- shown in Figure 5 clearly exhibits a series sioning period in February 2003, and of peaks around 4 mHz, corresponding to some others about 4 months later individual acoustic modes of the star, with during the second commissioning amplitude in the range 10-20 cm/s. The held in June 2003. All the data fit positive interference of several oscillation well the calculated orbit whose modes may lead to amplitudes much larg- parameters are fully consistent with Figure 6: Radial-velocity data of HD 83443, which harbours a known extra-solar planet. The er than the amplitude of single modes. the previously known solution. The data fit well the calculated orbit with a weighted From the power spectrum we estimate obtained dispersion (O−C) is only dispersion of 1.7 m/s.

22 The Messenger 114

Two years ago, we were able to detect Table 2: Stellar parameters and oscillation modes properties with the CORALIE spectrograph an Spectral Peak to Peak Star Period RMS Dispersion unambiguous oscillation signal of 31 cm/s type modulation amplitude of the star α Centauri A, a nearby solar twin (Bouchy & Carrier, α Cen B K1V 4 min. 1.5 m/s 0.5 m/s 2001 & 2002). With these measurements several oscillation modes were separated HD20794 G8V 5 min. 2.5 m/s 0.8 m/s and clearly identified. The measurements made on α Cen- HD160691 G5IV-V 9 min. 5 m/s 1.7 m/s tauri B with HARPS have lower frequen- cy resolution because of the limited dura- tion of the tests. To obtain the same reso- β Hydri G2IV 16 min. 5 m/s 1.7 m/s lution we would have had to observe the star for 13 consecutive nights. On the other hand the measurement obtained sequences (several nights are needed to interference leads to modulation of the with HARPS are impressive mostly reach a sufficient frequency resolution to Doppler time series of up to 10 times because of efficiency, cycle time and pre- characterize individual p-modes) of high their value. This is clearly a limitation for cision. Figure 4 illustrates clearly the sampling radial-velocity measurements high precision surveys at the detectivity obtained with HARPS: will permit to measure acoustic waves level of 1 m/s and requires us to define a despite the small amplitude the 4-minutes with amplitudes as tiny as a few cm/s. precise strategy of target selection and stellar oscillation can be detected “by observation. Short-period oscillations (4− eye” directly in the time series, and they DOWN TO THE STELLAR LIMIT 8 min.) of small amplitudes, which occur become even more evident in the Fourier After the amazing results obtained on α in G and K dwarfs, can be easily averaged Transform space shown in Figure 5a. Fig- Centauri B we have decided to monitor a out with typical exposures times of 15 ure 5b shows the autocorrelation of the small set of solar type stars. Figure 7 minutes. Sub-giants and giants, which can power spectrum. The plot clearly indi- shows short sequences of radial velocity show larger amplitudes and longer peri- cates that the peaks have a comb-like measurements obtained on these objects. ods, should however be carefully structure, with a main peak separation of On each of these sequences the stellar removed from the sample. This kind of about 160 µHz. This is the typical signa- oscillations are clearly visible. Stellar precaution should permit us to detect ture of oscillation modes of identical parameters and observed oscillation at the level of 1 m/s. angular degree expected for solar-like modes are listed in the Table 2.As expect- stars. The intermediate peaks appearing ed, these measurements clearly show that OBSERVING FAINT OBJECTS in the figure correspond to the correlation period and amplitude of the oscillation The OGLE-III programme, involved in between modes with angular degree l=1 modes are directly related to the stellar an extensive ground-based photometric and l=0 or 2. properties. They demonstrate the full survey for planetary and low Such a level of accuracy in the detec- capability of HARPS for asteroseismolo- object transits, has recently announced tion of solar-like oscillation modes has gy, not only on very bright stars. the detection of transiting candidates never been reached before. This result These results show however also that around stars located in the direction of obtained with HARPS on a very short the main limitation for HARPS will come the Galactic centre. The radii of these sequence (7 hours), clearly shows the from the stellar “noise” itself. The individ- objects were estimated to be in the range unique potential of this instrument in the ual mode amplitudes of these stars are in between 0.5 and 5 Jupiter radii, meaning domain of asteroseismology. Long the range 10−50 cm/s but their additive that some of these objects are probably in the planetary domain. In order to meas- ure the of these objects, and then determine their real nature, a radial velocity follow-up is necessary. In order to test the limiting magnitude of HARPS, we observed the faint candi-

date OGLE-TR-56 (mv=16.6). One hour exposure allowed us to reach a signal-to- noise ratio between 3 and 4 correspon- ding to a photon noise uncertainty on the radial velocity of about 30 m/s. The five RV measurements presented in Figure 8 demonstrate the full capability of HARPS to find and characterize hot Jupiters around stars as faint as OGLE- Figure 7: TR-56.We have fitted to these data points Summary plot a sinusoidal curve with fixed period as of the radial- given by the photometric study.The meas- velocity of ured amplitude is however much different stars for which we have made and in contrast to the results published by long observa- Konacki et al. (2003).A complete analysis tions series. of these measurements will be required in All of them order to determine whether this RV mod- show clear indications of ulation is due to a planetary companion star pulsation. or to a binary system blended and

Mayor M. et al., Setting New Standards with HARPS © ESO - December 2003 23

velocity measurements of several hun- ed to the mass of the convective zone and dred stars. Many of them show a varying probably finds its origin in the chemical radial velocity, sometimes with a clear composition of the primordial molecular periodicity. In the case of HD 330075 the cloud. To add new constraints to the link signal is most probably due to an exo- between star chemical composition and planet orbiting this star in 3.37 days. The frequency (or properties) of exoplanets, radial-velocity curve of HD 330075 is we will carry out two additional pro- shown in Figure 9. Precise photometric grammes: The first programme is a search measurements carried out on the SAT for exoplanets orbiting solar-type stars Danish Telescope and by the Swiss-Euler with notable deficiency (for most of them Telescope have shown the absence of a [Fe/H] between −0.5 and −1.0). Among photometric transit, which would have the existing detections of exoplanets only allowed us to determine the radius and two or three have been found with metal- the of the planet. Nev- licity in that range. We aim at being able ertheless, these measurements have also of estimating the frequency of exoplanets Figure 8: Radial-velocity of OGLE-TR-56 demonstrated that the photometric vari- in that domain of and, if possi- measured with HARPS. Despite the faint mag- nitude of the object, the dispersion of the data ability is low, and that we can exclude ble, to compare their characteristics points around the fitted curve is only 30 m/s. therefore the radial-velocity variation to (, orbits) to planets orbiting metal be produced by stellar activity.This makes rich stars. The second programme aims at HD 330075 the first HARPS extra-solar exploring the link between stellar metal- masked by the main star. This kind of planet candidate. licity and properties of exoplanets. Visual study has a huge interest for the future Only a few of the hundred detected binaries with solar-type stars of almost radial-velocity follow-up of exoplanets planets have masses less than the mass of identical magnitudes have been selected. detected by the satellite COROT. , and due to the present precision We will search for exoplanets orbiting of radial velocity surveys the distribution one of the components of these systems. THE HARPS of planetary masses is heavily biased (or For those including giant planets a EXOPLANET PROGRAMME completely unknown) for masses less detailed chemical analysis will be done The HARPS Consortium GTO pro- than half the mass of Jupiter.We will take for both stellar components to search for gramme will be devoted exclusively to the advantage of the very high precision of possible differences in their chemical study and characterization of exoplanets, HARPS to search for very low mass plan- compositions. in continuation of a planet-search pro- ets. For a sample of pre-selected non- Follow-up radial velocity measure- gramme initiated 10 years ago (Queloz & active solar-type stars we will be able to ments for stars with planetary transits Mayor, 2001). For a large, volume-limited explore the domain of the mass-function detected by the COROT space mission sample we will do a first screening in for planetary masses less than the mass of will be made with HARPS. Photometric order to identify new “Hot Jupiters” and Saturn down to a few Earth masses for transits provide an estimate of the radius other Jovian-type planets. Increasing the short periods. of the transiting planet as well as the list of “Hot Jupiters” will offer a chance to A systematic search for planets will be and phase. Complementary find a second star with a planetary transit made for a volume-limited sample of M- ground-based spectroscopic measure- among relatively bright stars. Better sta- dwarfs closer than 11 . Such a sur- ments with HARPS will constrain the tistics are needed to search for new prop- vey of very low mass stars will give us a planetary mass and then the planet mean erties of the distribution of exoplanet chance to derive the frequency of planets density.The main scientific return for the parameters. as a function of the . These planetary programme of the COROT This part of the programme already objects are of prime importance for mission will come from the combination started in July 2003 within the first GTO future astrometric studies to be carried of the photometric and radial velocity period assigned to the Consortium. In this out with the VLTI or space-mission like data. short run of only 9 nights HARPS SIM. ACKNOWLEDGEMENTS unveiled all its characteristics, its optical Stars with detected giant planets We would like to acknowledge the many peo- efficiency, the unique precision and, in exhibit an impressive excess of metallicity ple whose names do not appear among the particular, the outstanding efficiency of in contrast to stellar samples without authors but have contributed to the great suc- the reduction pipeline. In fact, we have giant planets (Santos et al., 2001). The cess of the HARPS project through their valu- collected about 500 spectra and radial excess of metallicity does not seem relat- able and dedicated work. This project has been financed by the Swiss National Science Foun- dation, the French region “Provence, Alpes et Côte d’Azur”,the Institut National des Sciences de l’Univers INSU (F), the Université de Figure 9: First HARPS Geneve` (CH), the Observatoire de Haute- extra-solar planet candidate Provence (F), the Physikalisches Institut der HD 330075. The planet has a Universität Bern (CH), and the Service

minimum mass of 0.8 MJ and d’Aéronomie du CNRS (F). The HARPS Pro- orbits the star in 3.37 days at ject is a collaboration between the HARPS a distance of 0.044 AU from Consortium and the European Southern its parent star, and belongs Observatory (ESO). therefore to the category of Hot Jupiters. The induced REFERENCES radial-velocity variation has a Bouchy F. & Carrier F., 2001, ESO Messenger, semi-amplitude of 107 m/s. 106, 32 The total dispersion with Bouchy F.& Carrier F.,2002, A&A, 390, 205 regard to the orbital solution Konacki M. et al., 2003, Nature, 421, 507 is (O-C)=2.5 m/s, and includes Pepe F.et al., 2002, ESO Messenger, 110, 9 photon noise, stellar pulsa- Queloz D. & Mayor M., 2001, ESO Messenger, tion and possible jitter. 105, 1 Santos N. et al., 2001, A&A, 373, 1019 24 The Messenger 114