1

No. 105 – September 2001

From CORALIE to HARPSThe Way Towards 1m s–1 Precision Doppler MeasurementsD. QUELOZ1, M. MAYOR1

With the collaboration of the CORALIE Team: S. UDRY1, M. BURNET1, F. CARRIER1, A. EGGENBERGER1, D. NAEF1, N. SANTOS1

And for the HARPS Project: F. PEPE 1, G. RUPPRECHT3, G. AVILA3, S. BAEZA2, W. BENZ 5, J.-L. BERTAUX 6, F. BOUCHY1, C. CAVADORE 3, B. DELABRE 3, W. ECKERT2, J. FISCHER5, M. FLEURY1, A. GILLIOTTE 2, D. GOYAK 2, J.C. GUZMAN 2, D. KOHLER 4, D. LACROIX4, J.-L. LIZON 3, D. MEGEVAND1, J.-P. SIVAN4, D. SOSNOWSKA1, U. WEILENMANN 2

1Observatoire de Genève; 2ESO, La Silla; 3ESO, Garching; 4Observatoire de Haute-Provence;5Physikalisches Institut, Bern; 6Service d’Aeronomie

1. Search for Extrasolar Planets by PreciseDoppler Measurements

Precise Doppler measurements of stars is a very efficient wayto search for extrasolar planets orbiting nearby stars similar tothe Sun. The gravitational interactions between a planet and itshost star produces a change of the radial velocity of the star thatcan be detected by Doppler measurements with precision of fewm s–1. Thanks to the effort of observations done by preciseDoppler surveys conducted world-wide, 67 companions withmass less than 15 MJ have been discovered. Today ongoingsurveys regularly monitor the radial velocity of a total sample of3000 G, K and M stars (Queloz 2001).

The detection of a planet by the mea-surement of the orbit of its host star bringsinformation on the mass of the planet, theorbital eccentricity and the orbital period.Actually the mass is only known withinthe uncertainty of the projection factorrepresented by the sin i of the orbit.However, the sin i statistics is so sharpthat one has 87% probability to be with-in a factor of 2 (sin i between 1 and 0.5).

Figure 1: Semi-majoraxis and eccentricitydistribution of the extra-solar planets(blue dots) dis-covered as of July2001. In red squaresare shown the giantplanets of the SolarSystem.

2

The first planet discovered orbitingthe star 51 Peg (Mayor & Queloz 1995)is a giant planet on a very close orbit(4.2 days). Other similar systems havebeen later detected. The direct obser-vation of a transit of one of these shortorbit planets (Charbonneau et al. 2000,Mazeh et al. 2000) brought the finalconfirmation of their existence. Theseplanetary systems, called “hot Jupi-ters”, do not fit in the paradigm of theplanet formation based on the observa-tion of our Solar System (Boss 1995;Lissauer 1995). Extra mechanismssuch that planet migration or multi-plan-et gravitational interactions have beensuggested to explain their formation(see references in the review by Marcyet al. 2000).

The orbital characteristics of the ex-trasolar planets detected so far contrastwith the orbital characteristics of giantplanets in our solar system (see Fig. 1).The orbits of giant planets in our SolarSystem are almost circular, a naturalconsequence of their formation in theprotoplanetary disk. The extrasolarplanets show a wide range of eccen-tricity surprisingly similar to the eccen-tricity distribution of binary systems withstellar companions (Mayor & Udry2000). This is not understood in theframe of a global planetary formationtheory. However the interpretation ofdata should be made carefully andspecifically the comparison with our so-lar system. Actually ongoing planet sur-veys have not been able yet to detect aJupiter analogue (1 MJ object at orabove 5 AU). The orbit of Jupiter pro-duces on the Sun a complete radial ve-locity variation of 13 m s–1 amplitude in11 years. Intensive high-precisionDoppler surveys while reaching enoughprecision (about 3 m s–1) do not yet

have enough timecoverage. Extrayears of measure-ments are needed totackle Jupiter massplanets with orbitalperiods longer than10 years. Moreover,it is important to re-call that Doppler sur-vey are sensitive toa detection system-atic that ties togetherthe mass of the plan-et with the semi-ma-jor axis of its orbit,making easier thedetection of a planeton a short orbit thanof a planet like Ju-piter.

Doppler surveysare providing thefirst results on plan-et mass distribution.While we can stilldebate on the maxi-mum mass of plan-

ets (See Jorissen et al. 2001) we knowfor sure that the planet mass function isdramatically rising towards low masses(Fig. 2). If we restrict the analysis toplanets with masses larger than 1 MJand with orbits having semi-major axesless than 3 AU we find a rise dn/dm ~m–1. Below 1 MJ the detection thresholdlimits the detection to systems onshorter orbits, therefore the observedmass function artificially decreases.With the reasonable assumption thatthe planet mass function should contin-ue to rise at least in the giant planetmass domain, we can expect that anyimprovements in the detection thresh-old should convert into a significant in-crease of the planet detection rate.

The studies of the content of the atmo-sphere of stars having a planet showspectroscopic features that distinguishthem from field stars with no planet de-tection. Statistically the stars with plan-ets are metal richer compared to starswith no planet detection (Santos et al.2001). A recent study on the Li6 contentof the atmosphere of the star HD82943, hosting a planet, suggests thatextra Lithium hasbeen brought to theatmosphere of thatstar, possibly fed bya planet (Israelian etal. 2001). These ob-servations rise theissue of a possibletrace of the planet

formation visible in the atmosphere ofthe star.

2. The CORALIE Spectrograph

CORALIE is the fibre-fed echellespectrograph installed in April 1998 atthe Swiss 1.2-m Leonard Euler tele-scope at the ESO La Silla Observatory(Fig. 3). It has been built as a joint proj-ect with the Observatoire de Haute-Provence which owns the first speci-men, named ELODIE (Baranne et al.1997; Queloz et al. 2000). TheCORALIE spectrograph is fed by two fi-bres each including a double scramblerdevice to improve the stability of the in-put illumination of the spectrograph.The spectrograph itself is located in anisolated and stable environment with anaccurate temperature regulation. Thefibres are connected to the Euler tele-scope Nasmyth focus by a front-endadapter. The calibration lamps and theentrance fibre viewer device are part ofthe front-end adapter.

Figure 2: Extra-solar planet mass distribution. Below 1 MJ themass distribution is biased by the detection threshold that makesplanets on far-off orbits harder to detect.

Figure 3: 1.2-m Swiss Leonard Euler tele-scope at La Silla. The CORALIE front-endadapter is visible at the Nasmyth focus onthe top of the right fork of the telescope.

Figure 4: Perspectivedrawing of CORALIE.The fibre entrance ison the left. The grat-ing is coloured ingreen.

3

The optical design of CORALIE is thesame as that of ELODIE. The echellegrating is used with tan φ = 4 blaze an-gle and the cross-disperser is made ofa prism and a grism in order to obtainequal spacing between orders throughthe whole useful wavelength range.The optical system is made of numer-ous surfaces but allows a compact im-age with a maximal resolution of100,000. (See Fig. 4 and Table 2.)

CORALIE has been designed toachieve precise radial velocity meas-urements and to deliver the measure-ments shortly after the end of the expo-sure. The telescope and the spectro-graph are operated by a single observ-er in a semiautomatic mode. The se-quence of observations can be pre-pared before the beginning of the nightand run automatically during the night.

The measurements of stellar radialvelocities at a few meter per secondprecision rely on two key elements.First, one needs enough photons andspectral information to compute the ra-dial velocity with a high precision.Second, one needs a stable referenceto measure and to correct the system-atic errors of the instrument. The3000 Å wavelength range ensures arich spectral information for the compu-tation of the radial velocity balancingthe modest size of the telescope. Forexample, for a K0 dwarf with v sin i = 2km s–1 we reach in 10 minutes a 3 m s–1

precision on the radial velocity meas-urement for a 7.5 magnitude star. Thestable reference is provided by the in-trinsic very high stability of the instru-ment itself and regular wavelength cal-ibrations with the thorium spectrum.Moreover, the simultaneous use of thethorium during science exposures cor-rects from any short-term drifts of theinstrument. The technique is known asthe simultaneous thorium referencingtechnique (Queloz et al. 1999). OnFigure 5 a CORALIE CCD frame with asimultaneous thorium reference is dis-played.

CORALIE has an automatic reduc-tion software that provides fully cali-brated spectra and a measurement ofthe radial velocity by cross-correlation.The general description of this softwarecan be found in (Baranne et al. 1996).Since the CORALIE commissioning,significant modifications in reduction al-gorithms have been made, leading toan improvement of the long-term preci-sion from 7 m s–1 to 2 m s–1 (rms).

The long-term reference is providedby the thorium spectrum. The proce-dure setting the global wavelength so-lution has been modified in 2001. Thewavelength solution fit includes aweighting scheme that takes into ac-count positioning uncertainties of thephoto centre of each pixel (about 50 ms–1). The fit of the global wavelength so-lution is done with 1200 thorium spec-tral lines. The fit on the solution has 80

m s–1 rms residual, corresponding to 2m s–1 error on the zero point of the cal-ibration.

The short-term tracking of the instru-ment by simultaneous thorium showsno instrument error apart the photonnoise (Fig. 6). The measured disper-sion of a sequence of radial velocitymeasurements on the Sun is 1.2 m s–1.Part of it comes from the solar oscilla-

tion signal, the rest from the photonnoise on the measurement of the in-stantaneous drift of the instrument.When we average the measurements,the sun oscillation signal averages aswell. The convergence of the observeddispersion with the expected disper-sion due to the photon noise demon-strates that no extra instrumental noiseis seen down to 30 cm s–1. This result

Figure 5: CCD frame of a CORALIE stellar exposure with its simultaneous thorium reference.The emission lines (black dots) of the thorium spectrum are clearly visible between the or-ders of the stellar spectrum

Figure 6: Observedradial velocity dis-persion on a 4-hourobservation se-quence of the Sun.We display the dis-persion for differentnumbers ofaveraged measure-ments in order to average the Sun oscillation signal aswell. The hatchedcurve indicates theexpected uncertain-ties on the radial ve-locity from the pho-ton-noise error onthe measurement ofthe simultaneoustracking. With an av-erage of 10 radial-velocity measure-ments we averageout the solar oscilla-tion signal.

4

makes possible asteroseismology pro-grammes to detect Solar-type oscilla-tions by radial velocity measurementson dwarf stars.

The long-term precision of radial ve-locity measurements is tied to the qual-ity and the reliability of the daily zeropoints. Actually no systematics have beenobserved apart the 2 m s–1 error on thewavelength calibration. The bright star82 Eri is one of the star used since theinstallation of CORALIE as a proxy forthe long-term precision tracking. Wemeasure a 3.6 m s–1 dispersion over a2.5-year duration with no indications oflong-term systematics (Fig. 7). More-over, the yearly average has a dispersionless than 1 m s–1. It demonstrates that infew years, with a longer time-base, thedetection of Jupiter-mass objects on far-off orbits similar to our giant planets Ju-piter or even Saturn is to be expected.

3. The CORALIE Planet SearchProgramme

In mid-1998, right after the suc-cessful commissioning of CORALIE,started the CORALIE planet searchprogramme. The survey sample ismade of 1650 dwarf stars, brighter than10th V-mag located in the southernhemisphere, selected according to theirdistance from the Sun. Actually, differ-ent distance criteria have been used forG and K dwarfs in order to compensatefor the magnitude difference betweenspectral types (see in Udry et al. 2000).In the sample 80% of the stars arebrighter than 9th V-mag. Active stars

and fast rotators are not excluded fromthe sample but once they are identifiedthey are observed with a lower priority.The exposure time of each observationis set to reach a photon-noise error ofabout 5 m s–1 per radial-velocity meas-urement. It corresponds to a signal-to-noise ratio per pixel between 30 and100, depending on the spectral type.

After three years of activity the planetsearch programme totals 14,000 pre-cise radial velocity measurements.CORALIE has discovered 23 extrasolarplanets (Table 1) and contributed to theprecise measurement of orbital charac-teristics of 5 otherplanets. Amongstthe many discov-eries made byCORALIE, a veryinteresting multi-ple system is dis-played in Figure 8.This multiple sys-

tem is made of two Saturn-mass ob-jects trapped in a 1:10 resonance.

4. Catching the Sound of Starswith CORALIE

Acoustic waves or solar oscillationsare observed on the Sun. They arethought to be excited by turbulent con-vection near the surface. Observationsof solar oscillations place importantconstraints on the internal structure ofthe Sun and provides a strong test ofevolutionary theory as well. The radialvelocity effect from the 5-min solar os-cillations has a radial velocity amplitudeof 23 cm s–1. Many attempts have beenmade, to detect similar oscillations onother stars. With CORALIE we havebeen able to detect for the first time anunambiguous oscillation signal of 31cm s–1 amplitude on the star α CentauriA, a nearby solar twin. This corre-sponds to a wave with an amplitude of40 metres at the surface of that star.

The star α Centauri A was observedwith CORALIE during 13 nights in May2001. In total, 1850 spectra were col-lected with typical signal-to-noise ratiosin the range 300–420 at 550 nm. Theradial-velocity measurement sequenceshows a dispersion of 1.53 m s–1. Thepower spectrum shown in Figure 9 ex-hibits a series of peaks between 1.8and 2.9 mHz modulated by a broad en-velope. This is the typical signature ofsolar-like oscillations (Bouchy & Carrier2001).

In the low frequency range of thepower spectrum (ν, < 0.6 mHz), thepower of the signal scales inversely tothe square of the frequency as expect-ed for a white noise contribution. Themean white noise level in the powerspectrum, computed in the range0.6–1.5 mHz, is 4.3 cm s–1 correspon-ding to a velocity precision of 1.0 m s–1

per measurement.

Planet name m2 sin i a O–C DiscoveryMJ [AU] m s–1

Gl86b 3.4 0.11 9 1998HD192263b 0.68 0.15 11 1999HD130322b 0.95 0.32 15 1999HD83443c 0.15 0.17 6 2000HD168746b 0.23 0.07 9 2000HD108147b 0.31 0.10 11 2000HD83443b 0.34 0.04 6 2000HD75289b 0.40 0.04 8 2000HD6434b 0.44 0.15 14 2000HD121504b 0.81 0.32 9 2000HD52265b 0.98 0.50 11 2000HD19994b 1.7 1.23 10 2000HD169830b 2.8 0.82 10 2000HD1237b 3.2 0.49 11 2000HD92788b 3.4 0.97 12 2000HD162020b 13 0.07 13 2000HD202206b 13.6 0.76 11 2000HD168443c 13.7 2.67 6 2000HD82943c 0.80 0.73 7 2001HD82943b 1.48 1.16 7 2001HD213240b 3.3 1.6 12 2001HD28185b 5.3 1.01 10 2001HD141937b 8.8 1.49 11 2001

Table 1: List of Extra-solar planets discovered by CORALIE. In the O–C column are the resid-uals to the planet model fit. Residuals higher than 10 m s–1 usually origin from stellar intrin-sic activity at the surface of some young star or they suggest another companion yet unde-tected. References can be found on http://obswww.unige.ch/~udry/planet/plan-et.html

Figure 7: Radial ve-locity measure-ments of the brightG8 star 82 Eri. Thedispersion of thedata is 3.6 m s–1. Inred we indicate theyearly average. Thedispersion of theyearly average isless than 1 m s–1,suggesting noinstrumentalsystematics downto this level.

5

The strongest modes are identified inFigure 10. From the measurement ofthe large splitting (∆ν = νn,l – νn – 1,l ) andthe small splitting (δν0 = νn,0 − νn-1,2) weconstrain the mass and the age of thestar. Preliminary results (Carrier et al.2001) suggest that α Cen A is slightlymore evolved than the Sun, with amass in the range 1.10–1.16 M�.

5. HARPS: the 1 m s–1 PrecisionInstrument

HARPS is a fibre-fed, cross-dis-persed echelle spectrograph design tomeasure radial velocities of stars with aprecision better than 1 m s–1. It will beinstalled on the 3.6-m ESO telescope atLa Silla, Chile. HARPS is the result ofan Announcement of Opportunity madeby ESO in 1998 for the design, the con-struction, and procurement of a High-Accuracy Radial velocity PlanetarySearcher (HARPS) instrument.

In response to ESO’s announcementthe Observatoire de Genève hasformed a Consortium that has been re-inforced considerably by the active par-ticipation of the ESO La Silla Obser-vatory and the ESO Garching Cryo-genic Group and Optical DetectorTeam. At present, all design reviewshave been passed and the project is inits manufacturing phase. The instru-ment commissioning is scheduled forthe end of 2002. Besides the guaranteetime for the Consortium, a large amountof HARPS time will be available to theastronomical community for a broad va-riety of observational programmes indifferent domains including for examplethe search for extrasolar planets andasteroseismology.

The strategical choices of theHARPS project are based on the expe-rience gathered with the ELODIE andthe CORALIE instruments. Moreover,to cope with the short track develop-ment of the project we have tried toavoid as much as possible any devel-opment risk that could jeopardise theproject. In general, we preferred toadopt conservative solutions every timethe consequences of a proposed newsolution on the final result were notknown precisely. HARPS design isbased on three fundamental technicalchoices. First, we decided to adopt a fi-bre-fed illumination with two fibres forsimultaneous thorium referencing.Apart the fact that this technique has al-ready proven its efficiency with ELODIEand CORALIE, it is about 4–6 times

Figure 8: Radial-ve-locity measurementsof the star HD 83443.The curve is the bestfit to the data with atwo-planet model.The period of theshortest orbit is2.9853 d and the longone is 29.85 d. Bothplanets have aboutthe same mass asSaturn. Interestingly,the system may betrapped in a 1:10 res-onance.

Figure 9: Power spectrum of 13 nights of radial-velocity measurements on the star α Cen A.The series of peaks between 1.8 and 2.9 mHz is the signature of solar-like oscillations.

Figure 10: Identified p-mode oscillations inthe power spectrum of radial-velocity meas-urements of α Cen A. l corresponds to thenumber of knots of the various pulsation or-ders (n-number), where l = 0 is the radial pul-sation. Typical identified pulsation moden-numbers range from 15 to 25.

6

more efficient in terms of photon needthan using an alternative technique likethe iodine cell for example (Bouchy etal. 2001). A fundamental aspect toreach 1 m s–1 accuracy on a large sam-

ple of stars. Second, we decided tobuild an instrument using the largestmonolithic echelle grating available(837 × 208 mm grating developedfor UVES) in order to achieve a very

high spectral reso-lution. For starswith unresolvedabsorption lines,the precision ofthe measurementof the radial veloc-ity scales with the1.5 power of thespectral resolution(Hatzes & Cochran

1992). A good compromise between slitlosses and best resolution was finallyfound to be R = 90,000. More complexsolutions for increasing the efficiencyand the spectral resolution, like for ex-ample using adaptive optics or an im-age slicer, have been considered butwere found not suitable for HARPS.Finally, while the simultaneous thoriumreferencing technique monitors the in-strumental drifts in order to removethem, we made additional efforts to in-crease the intrinsic opto-mechanicalstability of the spectrograph. In order toeliminate the atmospheric pressurevariation, which could produce wave-length drifts (100 m s–1/mbar) and to ex-clude any convective cell circulation inthe spectrograph, the entire spectro-graph is operated in vacuum. More-over, the vacuum vessel protects thespectrograph from rapid temperaturevariations. The vacuum vessel itself isinstalled inside a temperature-con-trolled environment which ensures along-term stability better than 0.1 K. Toimprove the stability of the spectro-graph input illumination as well, each fi-bre includes a double scrambler. Moredetails on HARPS design can be foundin Pepe et al. 2000.

The optical design, proposed by B.Delabre and adapted by D. Kohler, isvery similar to that of UVES. A ray trac-ing of the optical design is shown inFigure 11. Two fibres, an object and areference fibre feed the spectrographwith the light from the telescope. The fi-bres are re-imaged by the spectrographoptics onto a mosaic of two 2 × 4 kCCDs (EEV, 15 µm), where two echellespectra of 68 orders are formed. Thespectral domain ranges from 380 nm to690 nm with no order lost for the objectfibre. A summary of the spectrograph’sparameters is given in Table 2.

Realisation of the spectrograph’sopto-mechanics is under the re-sponsibility of the Observatoire de

Figure 11: Ray tracing of the HARPS optical design. The optical design is very similar to that of UVES. The main difference is the use of agrism for the cross disperser instead of a reflection grating. This solution is more stable and allows a compact mechanical mount.

HARPS CORALIE

Optical design fibre-fed, cross-dispersed echelle spectrograph# of fibres 2 (object and reference)Accepted field on sky 1 arcsec 2 arcsecCollimated beam diameter 208 mm 100 mmCovered spectral range 380 nm to 690 nmSpectral format 68 echelle orders

61.44 × 62.74 mm 26 × 26 mmSpectral resolution 90,000 50,000CCD chip mosaic, 2 × EEV 2k4 EEV 2k2

pixel size = 15 µm pixel size = 15 µmSampling/Spectral element (FWHM) 4 pixels 3.3 pixelsImage quality < 1.5 pixels < 1.5 pixelsMinimum inter-order spacing 30 pixels 10 pixelsSpectrograph peak efficiency at 550 nm 28 % 7%Total peak efficiency at 550 nm 4.5% 1.5%

Figure 12: TheHARPS dewar con-sisting of the detec-tor head, the cryo-stat, and the inter-face below. Therigid central part re-places the vacuumvessel and allows tosimulate the work-ing condition of thedewar.

Table 2: Spectrograph characteristics.

References

Baranne, A., Queloz D., Mayor M. 1996, etal., AASS 119, 373.

Boss, A., 1995, Science 267, 360.Bouchy, F., Carrier, F., 2001, A&A, 374, L5.Bouchy, F., Pepe, F., Queloz, D., 2001, AA

374, 733.Butler, R.P., Marcy, G.W., Williams, E., et al.

1996, PASP 108, 500.

7

Haute-Provenceand made in col-laboration with thePhysikalisches In-stitut of the BernUniversity. Thespectrograph op-tics is mounted ona 2.5-metre op-tical bench madeof plated steel.The orientation ofthe optical planeis vertical, theechelle grating be-ing mounted onthe top side, andthe grism and thecamera on thebottom side of the bench (Figure 11).

HARPS uses a standard VLT detec-tor head and the ESO controller FIERA.ESO’s Optical Detector Team will pro-vide the Consortium with the DetectorUnit including detector-head electron-ics, the LCU, and the Continuous-FlowCryostat adapted by the ESO Cryogenicgroup to the HARPS-specific vacuumvessel solution (See Fig. 12).

The vacuum vessel containing thespectrograph will be installed inside theair-conditioned coudé room. It is manu-factured under the responsibility of theGeneva Observatory. It consists of apolished stainless steel vessel of 1 mdiameter and about 3 m long, evacuat-ed at about p = 10–2 mbar.

The HARPS Cassegrain FibreAdapter is the interface to the tele-scope. It is entirely made by the La SillaObservatory. It incorporates several in-strumental functions and an Atmo-spheric Dispersion Corrector (ADC). Itis presently in an advanced realisationphase (see Fig.13).

HARPS should be an unrivalled facil-ity for conducting planet search pro-grammes and asteroseismology meas-urements. The improvement made onHARPS compared to CORALIE will re-duce the instrumental errors well belowthe 1 m s–1 threshold. The expectedperformances of HARPS are shown inFigure14. For a G8 dwarf star a radial-velocity measurement at 1 m s–1 accu-racy is reached in 1 minute exposurefor a star of magnitude 7.5. More detailson the photon-noise errors of radial-ve-locity measurements for different stellarspectral types and different v sin i canbe found in Bouchy et al. (2001).

Carrier, F., Bouchy, F., Provost, J., et al.,2001, IAU Colloquium 185, in press.

Charbonneau D., Brown, T., Latham, D.,Mayor, M., ApJ 529, L45.

Hatzes, A.P., Cochran, W.D., 1992, in“Workshop on High-Resolution Spec-troscopy with the VLT”, M. Ulrich, ed.,275.

Israelian, G., Santos, N. C., Mayor, M.,Rebolo, R. 2001, Nature 411, 163.

Jorissen, A., Mayor, M., Udry, S., 2001, A&Asubmitted, astro-ph/0105301.

Lissauer, J. 1995, Icarus 114, 217.Marcy, G., Cochran, W.D., Mayor, M., 2000,

PPIV, V. Mannings, A.P. Boss, S.S.Russell ed., 1285.

Mazeh, T., Naef, D., Torres, G., et al., 2000,ApJ 532, L55.

Mayor M. Queloz, D. 1995, Nature 378, 355.Mayor, M., Udry, S., 2000, in “Disks, plan-

etesimals and Planets”, F. Garzon, C.Eiroa, D. de Winter and T.J. MahoneyEds., ASP Conf. Ser. 219, 441.

Pepe, F., Mayor M., Delabre B., et al., 2000,in “Optical and IR Telescope Instru-mentation and Detectors” SPIE 4008,582.

Queloz, D., Casse, M., Mayor, M., 1999,ASP Conf Ser. 185, 13.

Queloz D., Mayor M., Naef D., et al., 2000,In “VLT Opening symposium opening:From Extrasolar Planets to Browndwarfs”, J. Bergeron & A. Renzini (eds.),ESO Astrophysics Symposia Ser., 548.

Queloz, D., 2001, in “11th Cool stars, stellarsystems and the Sun”, ASP Conf Ser.223, R.J. Garcia Lopez, R. Rebolo, M. R.Zapatero Osorio (eds), 59.

Santos, N.C., Israelian, G., Mayor, M., 2001,A&A 373, 1019.

Udry, S., Mayor, M., Naef, D., et al., 2000,A&A, 356, 590.

Figure 13: TheCassegrain Fibre

Adapter body duringmanufacturing at La

Silla, ESO.

Figure 14: Signal-to-noise ratio per spectral bin at λ = 550 nm. The dynamic range of the CCDand the estimated 1 m s–1 limit for a G8 star are shown.