Introduction

FLAMES (Fibre Large Array MultiElement Spectrograph) is the VLT FibreFacility, installed and commissioned atthe Nasmyth A focus of UT2 (KueyenTelescope). FLAMES was built and as-sembled in about four years through aninternational collaboration between teninstitutes in six countries and three con-tinents. It had first light with the fibre linkto the red arm of UVES on April 1, andwith the GIRAFFE spectrograph on

1

No. 110 – December 2002

Installation and Commissioning of FLAMES,the VLT Multifibre FacilityL. PASQUINI1, G. AVILA1, A. BLECHA2, C. CACCIARI3, V. CAYATTE4, M. COLLESS5, F. DAMIANI6, R. DE PROPRIS5, H. DEKKER1, P. DI MARCANTONIO7, T. FARRELL8, P. GILLINGHAM8, I. GUINOUARD4, F. HAMMER4, A. KAUFER9, V. HILL4, M. MARTEAUD4,A. MODIGLIANI1, G. MULAS10, P. NORTH2, D. POPOVIC8, E. ROSSETTI3, F. ROYER2, P. SANTIN7, R. SCHMUTZER9, G. SIMOND2, P. VOLA4, L. WALLER8, M. ZOCCALI1

1European Southern Observatory, Garching, Germany; 2Observatoire de Genève, Sauverny, Suisse; 3INAF-OABo, Bologna, Italy; 4Observatoire de Paris-Meudon, Meudon, France;5Research School of Astronomy and Astrophysics, Australian National University, Canberra, Australia;6INAF-OAPa, Palermo, Italy; 7INAF-OATs, Trieste, Italy; 8Anglo Australian Observatory, Sydney, Australia;9European Southern Observatory, Chile; 10INAF-OACa, Capoterra, Cagliari, Italy

Figure 1: The fibre positioner and GIRAFFEas seen during the GIRAFFE integration onthe Nasmyth platform. The positioner is look-ing towards the Nasmyth Focus, where the cor-rector is placed (only the corrector support isvisible). GIRAFFE is open, and all the opto-mechanics components are visible. The pic-ture was taken when the OzPoz enclosure hadnot yet been installed.

(www.eso.org/instruments/FLAMES/manuals).

The characteristics of the differentobserving modes are given in Table 1,which summarizes the observationalcapabilities of FLAMES.

UVES-Fibre

UVES is the high-resolution spectro-graph of the VLT UT2 (D’Odorico et al.2000). Each positioner plate has eightfibres connected to the UVES Red Arm.With an aperture on the sky of 1 arcsec,the fibres project to five UVES pixelsgiving a resolving power of ~ 47,000.Only the three standard UVES Red set-ups are offered, with central wave-lengths of 520, 580 and 860 nm, re-spectively.

GIRAFFE

GIRAFFE is a medium-high resolu-tion spectrograph (R = 6000–33,000)for the entire visible range (370–950nm). It is equipped with two echelles forlow and high resolution and uses inter-ference order sorting filters to select therequired spectral range within an order.The typical spectral coverage in oneexposure is 60–100 nm in low resolu-tion and 20–40 nm in high resolution.

The fibre system feeding GIRAFFEconsists of the following components(cf. Avila et al. 2002 for details):

• MEDUSA fibre slits, one per posi-tioner plate. Up to 132 separate objects(including sky fibres) are accessible inMEDUSA single fibre mode, each withan aperture of 1.2 arcsec on the sky.

• IFU slits. 15 deployable IntegralField Units (IFU) per plate, each con-sisting of an array of 20 square micro-lenses, for a total (almost rectangular)aperture of ~ 3 × 2 arcsec. For eachplate there are also 15 IFU dedicated tosky measurements.

• 1 ARGUS slit. A large integral unitconsisting of a rectangular array of 22by 14 micro-lenses, fixed at the centreof one positioner plate. Two scales areavailable: one with a sampling of 0.52arcsec/micro-lens and a total apertureof 12 by 7 arcsec, and one with a sam-

pling of 0.3 arcsec/micro-lens and a to-tal coverage of 6.6 by 4.2 arcsec. 15deployable ARGUS single sky fibresare also available.

GIRAFFE is operated with 30 fixedset-ups (22 high-resolution + 8 low-res-olution modes). For performance esti-mates (based on expected transmis-sion curves and performances) theuser is referred to the Exposure TimeCalculator (ETC), available at the ESOETC web page: www.eso.org/observ-ing/etc

The FLAMES observing software(OS) is coordinating the operations ofthe Telescope, the Positioner and theUVES and GIRAFFE spectrographs. Inaddition, it allows COMBINED observa-tions: that is the simultaneous acquisi-tion of UVES and GIRAFFE spectrawith the specific observing modes listedin Table 1.

LAYOUT

Figure 1 shows a view of the maincomponents of the FLAMES facility, thefibre Positioner and GIRAFFE, as seenon the telescope platform. The Cor-rector support is also visible, attachedat the Nasmyth rotator.

1. Corrector

The optical corrector is a big doubletof 880 mm free aperture. The functionof the corrector is to give an excellentimage quality over the whole 25 arcminFLAMES field of view and to provide apupil located at the centre of curvatureof the focal plate, to avoid vignetting foroff-axis fibres. When the whole opticaltrain is taken into account (including tel-escope optics and vignetting), the ef-fective transmission of the corrector de-pends on the observing wavelengthand on the distance of the object to thefield centre, varying from about 78% at370 nm to 90% in the visible, and 82%at 1 micron.

A system of lamps has been installedto illuminate the Nasmyth screen,which is located in the telescope cen-trepiece, about 2 m in front of the focalplane.

2

July 3. We report here on the complexprocess of integration and commission-ing, and we compare the expected andobserved astronomical requirements.

Overview

FLAMES is the multi-object, interme-diate and high-resolution fibre facility ofthe VLT. Mounted at the Nasmyth Aplatform of UT2 it offers a large correct-ed field of view (25 arcmin diameter)and it consists of several components,developed by several consortia (Aus-tralis (AAO/ANU/UNSW), Paris-Meu-don (OPM), Geneva and Lausanne(OGL), Bologna, Cagliari, Palermo andTrieste (ITAL Consortium)) and ESO:

• An optical Corrector, providing ex-cellent image quality and telecentricityover the full field of view of 25 arcmindiameter (developed at ESO).

• A Fibre Positioner hosting twoplates. While one plate is observing,the other one is positioning the fibresfor the subsequent observations, there-fore limiting the dead time between ob-servations (developed at AAO).

• A link to the UVES spectrograph(Red Arm) via eight single object fibres /plate (developed at OPM in collabora-tion with ESO).

• A high and intermediate resolutionoptical spectrograph, GIRAFFE (devel-oped at OPM and ESO).

• Three types of fibre systems: MEDU-SA, IFU, ARGUS (developed at OPM).

• A coordinating observing software,that allows simultaneous UVES and GI-RAFFE observations (developed at ITALConsortium in collaboration with ESO).

• As for all VLT instruments, a fullData Reduction Software (DRS) pack-age is provided, integrated in the ESOData Flow Software (developed at OGLwith contributions from OPM (GI-RAFFE) and ITAL Consortium (UVES-Fibre)).

The FLAMES components are de-scribed in detail elsewhere (cf. e.g. Avi-la et al. 2002, Blecha et al. 2000, Gilling-ham et al. 2000, 2002, Jocou et al. 2000,Mulas et al. 2002, Pasquini et al. 2000,Royer et al. 2002), on the ESO web siteand in the FLAMES User Manual

Spectrograph Mode Nr. Objects Aperture R Coverage (nm)

UVES8 Red Arm (with sky) 8 1 47000 200, 400UVES7 Red Arm (with sky) Simult. Calibration 7 1 47000 200GIRAFFE MEDUSA 132 (with sky) 1.2 20500* λ/22*GIRAFFE MEDUSA 132 (with sky) 1.2 7000* λ/7*GIRAFFE IFU 15 (+15 sky) 2 × 3 33000* λ/22*GIRAFFE IFU 15 (+15 sky) 2 × 3 11000* λ/7*GIRAFFE ARGUS 1 11.5 × 7.3 33000* λ/22*

Or 6.6 × 4.2 GIRAFFE ARGUS 1 11.5 × 7.3 11000* λ/7*

Or 6.6 × 4.2

Table 1. Observational capabilities of FLAMES.

*The GIRAFFE resolving power (R) and wavelength coverage given here are only average values. They will vary for the different set-ups by ± ~ 20%. Apertureis in arcseconds.

2. Fibre Positioner (OzPoz)

The Fibre Positioner (OzPoz) is atthe core of the FLAMES facility. OzPozis a rather large and complex systemequipped with four plates, two of whichare currently in use (cf. Gillingham et al.2002). The Positioner can be subdivid-ed into the following subsystems:

• Plates: Two metallic disks, on whichthe magnetic buttons holding the fibresare attached. Each of the plates has ahole in the centre. In one plate this holehosts ARGUS. Each plate has a curva-ture of 3950 mm to match the curvatureof the corrector focal surface.

• Retractors: Mechanical systemsmaintaining the fibres in constant ten-sion. Each fibre is equipped with oneretractor. The retractors are externallythe same for all fibres.

• Exchanger: Main structure holdingthe plates. The exchanger can performtwo main movements: it can retract (orapproach) the Nasmyth adapter to en-gage the plate (or disengage it).Furthermore, it can rotate the upperpart in order to perform the plate ex-change.

• r-θ system and Gripper: This unitgrips and releases the magnetic but-tons at the positions reached via the r-θ (polar) robot. The gripper requires aback-illumination system, sending abrief light flash along each fibre fromthe spectrograph towards the plate. Acamera records this back-illuminationlight and performs an image analysis.The back-illumination light is used forseveral purposes: to reach the requiredpositioning accuracy and to detect if themagnetic button was properly liftedfrom the plate.

• OzPoz is equipped with a calibra-tion box which can direct the light eitherfrom a tungsten, or from a Th-Ar, orfrom a Ne lamp. In this way FF, Th-Arand Ne calibrations can be obtained forGIRAFFE and for UVES.

• Field Acquisition Bundles (FACBs):Four magnetic buttons on each plateare equipped with a system of 19 co-herent fibres each. This bundle is usedto obtain images of fiducial stars, oneper button. The four images are viewedby an ESO technical CCD; the imagecentroids are computed and the properoffsets are calculated to centre the fidu-cial stars into the bundles. Each FACBbundle has an effective diameter of 2.4arcseconds.

• Positioning Software: It is writtenaround the so-called delta-task pro-gram, developed initially for the 2dFsystem at AAO. This program allowscrossing between the fibres and it de-termines the button movements se-quence.

3. Buttons and Fibre Systems

FLAMES is equipped with differenttypes of fibres for UVES and for the dif-

ferent modes forGIRAFFE. It isworth mentioningthat, in addition tothe object or skyfibres coming fromthe Positioner,each GIRAFFE fi-bre slit has five fi-bres devoted to si-multaneous wave-length calibration,which provide si-multaneous cali-bration spectra foreach observationacquired with GI-RAFFE (Hammeret al. 1999).

The histogram distribution of thetransmission is given in Figure 2 for thedifferent fibre types. The FLAMES fi-bres have been proven very robust, sothat out of more than 1000 fibres, onlya few were broken (and replaced) at theend of the whole process.

3.1 UVES fibres

Each of the positioner plates hostseight 55-metre fibres to guide the lightto the UVES spectrograph located onthe opposite Nasmyth Platform. Fromthe UVES simultaneous calibrationbox, one additional 5-metre fibre reach-es the UVES-Fibre slit. The fibre cen-tres are separated by 1.7 fibre core,implying that there is some degree ofcontamination between adjacent fibres;the UVES-Fibre Data Reduction Soft-ware (Mulas et al. 2002) has been de-veloped to obtain a complete de-blend-ing of the spectra.

3.2 MEDUSA fibres

Each plate also hosts 132 MEDUSAfibres. The separation between each ofthe MEDUSA fibres in a sub-slit is 2.26× the fibre core; this ensures a fibre-to-fibre contamination below 0.5 per cent.

During our tests we have experi-enced some (about 2%) increasedstray light due to reflections from the fil-ter and the (polished) exit slit. We willattempt to reduce this with a stray lightmask that reduces the width of the re-flecting exit slit viewed from the side ofthe filter.

3.3 IFU fibres

Each Integral Field Unit (IFU) buttonis composed of twenty micro-lensesarranged in a rectangular shape. Themicro-lenses are 0.52″ squares. Themovable IFUs are a unique characteris-tic of GIRAFFE and can be placed allover the FLAMES field of view, with theexception of a small fraction of the cen-tre annulus. Each plate hosts fifteenIFUs plus fifteen Sky IFUs. The sepa-ration between the fibres’ centre is only

1.47 times the fibre diameter core,which implies that the contaminationbetween adjacent fibres is about 10 percent. With the chosen arrangement ofthe fibres this is acceptable since innormal seeing conditions a higher levelof contamination will be present at thefibre entrance level.

3.4 ARGUS fibres

The ARGUS system is a fixed arrayof 14 × 22 micro-lenses, similar to theIFUs, located in the middle of Plate 2.ARGUS is also equipped with a lens toswitch between a scale of 0.52″/micro-lens to a finer scale of 0.3″/pixel. AnADC ensures that the object images atdifferent wavelengths are maintained inthe same locations up to a Zenithal dis-tance of 60 degrees. In addition to theobject fibres, fifteen ARGUS sky fibresare present on the plate.

4. GIRAFFE

GIRAFFE is a fully dioptric spectro-graph. The fibres are arranged in 5 longslits (IFU 1/2, MEDUSA 1/2, ARGUS) atthe curved focal plane of the collimator.One of these fibre slits is placed in theworking position by a translation stage.The other slits are masked. After leav-ing the fibres at F/5, the light first pass-es through one order sorting filter be-

3

Figure 2: Transmission distribution of all GIRAFFE fibres, as meas-ured in the laboratory.

Figure 3: Transmission of order sorting filterHR 17.

fore it is directed to the collimator by a45-degree mirror. The collimated beamis dispersed by one of the two echellegratings that are mounted back to backon the grating turret. After passingthrough the collimator again an inter-mediate spectrum is produced. Finally,a re-imaging system (“camera”) pro-duces a 2.5 times demagnified imageon the 2K × 4K CCD at F/2. A view ofGIRAFFE is given in Figure 1. GI-RAFFE has been conceived to mini-mize maintenance and night calibra-tions; special requirements have beenintroduced to reduce set-up shifts andto obtain accurate re-positioning, to beable to use wavelength calibrations andflats obtained in the afternoon. To beable to monitor instrument drifts, five fi-bres are always used to obtain simulta-neous calibration spectra.

A summary of the most relevantGIRAFFE characteristics is given inTable 2

GIRAFFE is operated with 30 fixedset-ups; 22 are used to cover the wholespectral range with the high resolutiongrating, 8 with the low resolution grat-ing. The different GIRAFFE sub-unitsare described in more detail in the fol-lowing sections.

4.1 Slit Unit

The slit unit is the most complex GI-RAFFE mechanical subsystem, be-cause it needs a very high stability andreproducibility. It has two linear move-ments: one to exchange the fibre slit,the other one is used for focusing andto press the non-used slits against abaffle that is also used for back-illumi-

nation. The slit unit is equipped with anumber of back-illumination LEDs,which are switched on before pick-upand after placing a fibre, to allow thegripper camera to view the fibre output.

4.2 Filters

After the fibre slit, an order-sorting in-terference filter is placed in the beam.These filters must have excellent trans-mission and steep bandpass edgesand out-of-band blocking, in order toavoid pollution from adjacent spectralorders. One example of filter transmis-sion is given in Figure 3. The transmis-sions of all filters, averaged over eachset-up, are given in Figure 4. This figureshows also the efficiencies of other GI-RAFFE components, as well as theoverall GIRAFFE efficiency.

4.3 Optics and gratings

The gratings are placed back to backon a turntable. The high-dispersiongrating is a MgF2/Ag coated, 204 × 408mm 63.6 degree echelle with a largegroove density (316 lines/mm). The370–950 nm spectral range is coveredby twenty-two set-ups in grating orders15 to 6. Silver was selected because ofits good VIS/NIR efficiency and the lowpolarization it produces on this grating.While the efficiency on one edge of thegrating is good through the used spec-trum, the efficiency degrades below500 nm as one approaches the otheredge. This is due to inhomogeneity inthe MgF2 layer. We are working to finda replacement up to specifications.

The low-resolution grating has 600

lines/mm coated with MgF2/Al. Its first-order blaze lies at 1.96 µm, so in therange 370–950 nm it works in orders5–2; eight set-ups are needed to coverthe full range. After being dispersed,the light passes again through the colli-mator, forms a real image at an inter-mediate focal plane and is finally im-aged by an F/2 reimaging camera with7 elements.

4.4 Spectral Format and Efficiency

In GIRAFFE the spectra are parallelin dispersion along the long side of thedetector (readout direction), while theshort side is along the slit direction.Spectra are slightly curved due to opti-cal distortion and lateral chromatism,with the central part closer to each oth-er than the edges. As is the case in anylong-slit spectrograph, the lines of con-stant wavelength are arranged on low-curvature arcs. An example of a Th-Aspectrum taken with MEDUSA is givenin Figure 5.

The resolving power and coverageare both a function of the grating angleand grating order: two or three differentgrating angles are required to fully cov-er a grating order, which causes the dif-ferences in resolution and spectral cov-erage between the different set-ups.The higher resolving power of the IFUand ARGUS modes (compared to theMEDUSA mode) is due to the smallersize of the fibres, which projects to~ 2.2 pixels instead of ~ 4.3 pixels ofMEDUSA , but the spectral coveragefor a given set-up is the same in theMEDUSA, IFU and ARGUS modes.

5. Integration andCommissioning Results

FLAMES components were installedbetween October 2001 and April 2002on UT2. The integration sequence hasroughly followed the light path: Correc-tor, UVES fibres, Positioner, GIRAFFE.GIRAFFE had first light in Europe inDecember 2001 and was re-assembledin Paranal in April 2002 without majorinconveniences, after the re-integrationof the positioner and in perfect timingwith the rest of the project.

4

Fibre slit height 76.8 mmType Echelle + Order selection FiltersCollimator aperture F/5Collimator beam 18 cmH-R Echelle 204 × 408 mm, 316 lines/mm, 63.4º BlazeL-R Echelle 156 × 204 mm, 600 lines/mm, 33.7º BlazeReimager demagnification ratio 2.5Effective Camera Focal Length 360 mmDetector 2048 × 4096, 15 µm EEV CCDScale 0.20 arcsec/pixel

Table 2. Summary of GIRAFFE construction characteristics.

Figure 4: Component efficiencies per set-up and the resulting overall DQE (slit to detected photoelectrons) of GIRAFFE. All reported valuesare averages; note that grating efficiencies can vary by up to a factor of two within a set-up. The optics efficiency includes vignetting at thegrating, this is why the reported LR values are slightly lower than HR values at the same wavelengths.

During the first run at the end ofMarch 2002 the Positioner was inte-grated and tested and we could gatherthe first light with UVES-Fibres. Three10-minute spectra were obtained for 7stars (one fibre was dedicated to sky)belonging to the Globular clusterOmega Cen. Three of them were previ-ously observed with UVES in long slitmode (Pancino et al. 2002). A smallportion of the spectra around the lineHα line is shown in Figure 6. The starshad magnitude varying between V =11.5 and 12.9. After data reduction theequivalent widths of a few hundredlines were compared for the UVES –Fibre and UVES-long slit spectra of thesame star, finding an excellent agree-ment: EQW(Fibre) = EQW(Slit) + A withA in the range of 4–6 mÅ, with an rmsof 4–7 mÅ, which is largely due toequivalent widths measurement uncer-tainties.

During this March run, malfunctionswere experienced in the positioner ro-bot, due to the lower temperature at thetelescope compared to those prevailingduring the tests in Australia and in theAssembly Hall in Paranal. We tem-porarily fixed this using three heaters,

clearly not a suitable permanent solu-tion. This was corrected prior to a sec-ond commissioning period in June butnew mechanical problems, related tothe interface with the telescope rotator-

adapter were encountered. Theseproblems were cured in advance of theAugust commissioning run. On July 3,first GIRAFFE spectra could be ob-tained. Two sets of spectra of stars inan astrometric field were acquired,each of 10 minutes. The set-up usedwas LR4, Low Resolution centred at awavelength close to the V filter. 66MEDUSA fibres were positioned (somein sky positions), and the reduced spec-tra of some of them are shown in Figure7. The stars have R magnitudes be-tween 13 and 14. In all spectra theMagnesium triplet at 518 nm is clearlyvisible, as well as the 557 sky emissionline.

The stability and repeatability of GI-RAFFE on the telescope, measured bytaking a ThAr spectrum every 4 hoursover a period of 2 weeks is shown inFigures 8 and 9.

The tests have been very satisfac-tory, showing that the instrument ex-ceeds specifications.

A further commissioning took place inAugust 2002, enabling us to advancethe commissioning tests. ManyMEDUSA ‘rasters’ have been per-formed, to determine the telescope-plate geometry.

In these rasters an astrometric field isobserved with MEDUSA and the tele-scope is offset along a grid around thepointing position. For every star, thedistance between its nominal position(the centre of the grid) and the positionsof its maximum flux is then computedand used to correct the astrometricmodel. The results have been so farvery encouraging, and the found resid-uals are shown in Figure 10 , with anRMS of 0.15 arcseconds, which is verygood when considering that this num-ber includes all sources of errors, e.g.,the accuracy of the stellar astrometry.In Figure 11 we show the image of 4

5

Figure 5: Th-A spectrum taken with a MEDUSA slit. The fibres in the centre of the slit havelines moved towards smaller pixels; e.g. fibres corresponding to the central fibres have slight-ly redder wavelength coverage than the ones at the CCD edges.

Figure 6: Reduced spectra of 7 giants in the globular cluster Omega Cen taken with UVES-fibre around the Hα line.

Figure 7: Reducedspectra of GIRAFFEfrom the first com-missioning run: thestars belong to anastrometric field.The Mg triplet andthe 557 nm skyemission line areclearly visible in allspectra.

stars from the Tycho catalogue wellcentred in the FACBs after the tele-scope field acquisition has been com-pleted.

Two recurrent questions when usingfibres are the capability of obtaininggood flat field corrections, and to sub-tract the sky. Sky subtraction tests areunder way, while flat field correction ca-pabilities have been evaluated. In Fig-ure 12 is shown the comparison be-tween the measured S/N ratio in a por-

tion of the spectraof several stars inthe SMC, acquiredwith the L8 set-up,a spectral rangeplagued by strongfringing. We search-ed for a ‘line free’region in the spec-tra and the S/N ra-tio has been com-puted in this small(typically less than2 nm) spectral re-gion. In ordinatethe Poisson noise

based on the number of detected elec-trons is given. The agreement betweenthe measured and Poisson S/N ratio isquite good.

In the past, several groups have re-ported problems in obtaining accurate

6

Figure 8: the average position of ThAr lines in a time series that wastaken at 4-h intervals over a period of 2 weeks. Shifts are given inpixels relative to the first exposure. These exposures were taken inMedusa HR mode, with configuration changes between exposures.The jump on May 7 cannot be correlated to any known event like anearthquake.

▲ Figure 9: Histogram of relative shifts after4, 12 and 24 hours, using the time series ofFigure 12. In HR mode, 0.1 pixel corre-sponds to about 300 m/sec. These plotsshow the intrinsic stability of Giraffe and willbe helpful to plan the frequency of calibra-tions. Note that we expect to obtain higherradial velocity precision by using the simul-taneous calibration fibres.

Figure 10: Residuals vectors after the subtraction of the position cen-troid from the object fibre, after a successful modelling of the plategeometry.

Figure 11: After a successful acquisition iscompleted, the four fiducial stars should bewell centred in the FACBs, as in the case ofthese 4 stars from the Tycho Catalogue. Theobservations were made at airmass 1.8which shows the good handling of atmos-pheric effects by the positioner SW. The fig-ure shows the Graphic User Interface at theUT2 Console.

▼

flat fielding while using fibres, because,if spatial filters are present in the opticalpath, variable fringing pattern can beproduced in the spectra.

By dividing 3 FF taken with the fibrepositioner by the mean of 36 FF takenwith the Nasmyth screen, a S/N ratio of400 was obtained. This S/N ratio ob-tained was less than what expected byPoisson noise statistics, but it showsthat this problem is limited in theFLAMES system to very high S/N ob-servations. Note that the 3 positionerFF represent the standard calibrationsthat the observer will receive. Similarlywe have tested the stability of photo-metric response of the fibres to torsionand movements: by taking Nasmyth FFand after rotating by a full circle, thephotometric stability has been shown tobe better than 0.5 %. Finally, in order toperform an accurate sky subtraction,the relative fibre transmission needs tobe known accurately. By comparing fi-bre to fibre transmission by using 3 po-sitioner flats (standard calibrations)with that obtained by the average of the36 Nasmyth flats, we find that the twomeasurements agree to better than 3%rms. This is a good result, although notyet within the specifications (2%).

We have performed several compar-isons with the ESO Exposure TimeCalculator (ETC, www.eso.org/observ-ing/etc). In Figure 13 we show the re-sults of the comparison between theflux observed vs. the flux predicted asobtained by observing an astrometricfield (the astrometric solution in use inthis test had a residual rms of 0.24 arc-seconds). The Figure shows the num-ber of counts recorded vs. R magni-tude, while the curve shows the predic-tion by the ETC in the L8 set-up for aG0 star and the parameters (seeing,exposure time) appropriate for the ob-servation. As shown in the bottom pan-el, at any given magnitude, the numberof collected e– per object is represent-ed by a relatively broad distribution.This is due to several factors, such asthe variation in the fibre transmission,

the poor astrometry of some targets,and the difference in spectral typeamong the stars in the catalogue. On

the other hand, it is important that thepotential users become familiar withthe concept that for FLAMES the S/N

7

Figure 12: S/N ratio as measured in cleanregions of spectra taken in the L8 set-up,compared with the Poisson S/N ratio ex-pected on the number of counts recorded.

Figure 13: Comparison between observed flux and predicted ETC flux for an astrometric fieldwith the L8 set-up. The histogram shows the ratio measured/expected. The ETC calculationsare based on a G0 spectral type.

Figure 14: Distribution of the difference between RV single measurements and RV mean for3 exposures taken in 1 night of 60 stars of the globular cluster NGC 6809, shown vs. the stel-lar magnitude. Points from the different observations are shown with different colours. In thelower panel, the left distribution refers to all the measurements, irrespective of stellar magni-tude. The right distribution includes only the bright (V < 14.3) subsample.

ratio predicted by the ETC will alwayshave to be intended in a statisticalsense.

A number of tests have been per-formed to characterize the GIRAFFEradial velocity (RV) capabilities, observ-ing the same objects several time in anight, as well as in different nights.Three not consecutive observations of15 minutes in the GIRAFFE H9 set-upwere obtained in the same night for 96giants at the centre of the GlobularCluster NGC6809. The stars were se-lected between magnitude 13 and 15and the observations made with a nonoptimal astrometric model. Out of thefull sample (96 stars) we analysed only60 objects giving a well determinedcross correlation peak.

Out of these 11 were further exclud-ed because their RV were either dis-cordant from the bulk of the cluster, orbecause they showed a very large RVdifference in the 3 measurements (4stars). Based on the 3 measurementsof the 49 stars left, the cluster velocityis found at 178.81 km/sec (note thatthis RV is computed with respect to ourdigital mask, no attempt has beenmade to transform it to an absoluteRV), with a dispersion of 4.057 km/sec.In Figure 14 the distribution of the dif-ferences of all the measurements withrespect to their mean values areshown. This figure clearly shows howthe Radial Velocity accuracy degradeswith increasing stellar magnitude (andtherefore decreasing fluxes); the threeobservation points are indicated withdifferent colours. When taking all thepoints, irrespective of the stellar magni-tudes, the σ of the distribution is of 113m/sec, while if only the bright end isconsidered, V < 14.3, the σ decreases

Red Giant Branch of the SmallMagellanic Cloud; the magnitudes varybetween V = 18.5 and 18.8, the expo-sure time was of 1 hour with an 0.9 arc-second seeing.

Spectra were acquired for a field ofgalaxies with the Giraffe IFUs, and theirreduction is not yet completed. We aimat making a release of FLAMES com-missioning data in the first quarter of2003.

6. Acknowledgements

The development of FLAMES, whichwas recommended by the ESO STC in1998, presented some remarkable as-pects, showing the feasibility of a com-plex, multi-institute facility in a shorttime. This could not have been possiblewithout appropriate structures butmostly without the dedicated work of avery large number of people, whosenames do not appear in the author list.We thank all the persons at ESO

8

to 70 m/sec. These numbers are in-dicative of the accuracy reached in thetime scale of a few hours and normaloperations. Royer et al. (2002) havereached a short-term accuracy of 13m/sec with GIRAFFE, in a more stableenvironment, using much higher S/Nobservations of the solar light in theGarching Integration Hall.

We had the opportunity to observethe same NGC 6809 field also with GI-RAFFE in low resolution mode (L4) andin Figure 15 a fraction of the spectra ofthe same stars (4 giants belonging tothe cluster) observed in L4 and H9 at aresolving power of R = 6000 and R =26,000 respectively are shown. The in-tegration time in the HR9 was 2 timeslonger than in the LR4 .

We are acquiring spectra to verify thebehaviour of the instruments at the faintend of the magnitude range; Figure 16shows a collection of Hα spectra takenwith the high resolution grating of GI-RAFFE (set-up H14) of giants on the

Figure15: GIRAFFE Low (L4, black lines) and High (H9, red lines)resolution spectra of 4 giants belonging to the Globular Cluster NGC6809

Figure 16: Hα Spectraof 8 red giants in theSMC, no sky subtractionwas performed, and skyHα in emission is visibleat 6562.8 nm. The starshave V magnitude be-tween 18.5 and 18.8.The exposure time was1 hour, in the GIRAFFEH14 setup. Fluxes are inLog scale.

(Garching and Paranal) and at the dif-ferent institutions who have contributedto the realization of this ambitious proj-ect: J. Alonso, P. Ballester, J.-L. Be-ckers, P. Biereichel, B. Buzzoni, C. Ca-vadore, N. Cretton, C. Cumani, B. Dela-bre, A.V. Kesteren, H. Kotzlowski, J.-L.Lizon, R.S. Moreau , W. Nees, R. Pal-sa, E, Pozna, (ESO), P. Barriga, R.Castillo, J. Spyromilio AND MANYOTHERS at Paranal, C. Evans, R.Haynes, U. Klauser, B. Hingley, D.Mayfield, S. Miziarski, R. Muller, W.Saunders, G. Schafer, K. Shortridge,(AAO), L. Jocou and the fibre team, T.Melse and the mechanics workshopteam, S. Baratchart, L. Chemin, H. Flo-res, D. Horville, J.M., Huet, F. Rigaud,F. Sayède (GEPI, OPM), F.R. Ferraro,

P. Molaro, R. Pallavicini, I. Porceddu(ITAL Consortium).

ReferencesAvila, G. Guinouard, I., Jocou, L., Guillon, F.,

Balsamo, F. 2002, in Instrument Designand Performance for Optical/InfraredGround-Based Telescopes Proc. SPIE4841 (in press).

Blecha, A., North, P., Cayatte, V., Royer, F., Si-mond, G. 2000, Proc. SPIE Vol. 4008, p. 467.

D’Odorico, S. et al. 2000, Proc. SPIE Vol.4005, p. 121.

Gillingham, P., Miziarski, S., Klauser, U.2000, Proc. SPIE Vol. 4008, p. 914.

Gillingham P.R., Popovic D., Waller L.G.,Farrell T.J. 2002, in Instrument Designand Performance for Optical / InfraredGround-Based Telescopes, Proc. SPIE4841 (in press).

Hammer, F., Hill, V., Cayatte, V. 1999, Jour-nal des Astronomes Français, 60, 19.

Jocou, l. et al. 2000, Proc. SPIE Vol. 4008,p. 475.

Mulas, G., Modigliani, A., Porceddu, I.,Damiani, F. 2002, in Instrument Designand Performances for Optical/InfraredGround-Based Telescopes, Proc. SPIE4841 (in press).

Pancino, E. et al. 2002: ApJ 568, L101. Pasquini, L. et al. 2000, Proc. SPIE Vol.

4008, p. 129. Pasquini , L. et al. 2002, in Instrument

Design and Performances for Optical/Infrared Ground-Based Telescopes, Proc.SPIE 4841 (in press).

Royer, F., Blecha, A., North, P., Simond, G.,Baratchart, S., Cayatte, V., Chemin, L.,Palsa, R. 2002, in Astronomical DataAnalysis II, Proc SPIE 4847 (in press).

9

HARPS: ESO’s Coming Planet SearcherChasing Exoplanets with the La Silla 3.6-m Telescope

F. PEPE1, M. MAYOR1, G. RUPPRECHT3

email: Francesco.Pepe@obs.unige.ch; Michel.Mayor@obs.unige.ch; grupprec@eso.org

with the collaboration of the HARPS Team: G. AVILA3, P. BALLESTER3, J.-L. BECKERS3, W. BENZ5,J.-L. BERTAUX6, F. BOUCHY1, B. BUZZONI3, C. CAVADORE3, S. DEIRIES3, H. DEKKER3, B. DELABRE3,S. D’ODORICO 3, W. ECKERT 2, J. FISCHER5, M. FLEURY1, M. GEORGE 1, A. GILLIOTTE 2, D. GOJAK 2,3,J.-C. GUZMAN2, F. KOCH3, D. KOHLER4, H. KOTZLOWSKI, D. LACROIX3, J. LE MERRER10, J.-L. LIZON3,G. LO CURTO2, A. LONGINOTTI3, D. MEGEVAND1, L. PASQUINI3, P. PETITPAS4, M. PICHARD1, D. QUELOZ1,J. REYES3, P. RICHAUD4, J.-P. SIVAN 4, D. SOSNOWSKA1, R. SOTO 2, S. UDRY1, E. URETA2, A. VANKESTEREN 3, L. WEBER1, U. WEILENMANN 2, A. WICENEC 3, G. WIELAND 3 and advice by the InstrumentScience Team: J. CHRISTENSEN-DALSGAARD7, D. DRAVINS8, A. HATZES9, M. KÜRSTER9, F. PARESCE3,A. PENNY11

1Observatoire de Genève; 2ESO, La Silla; 3ESO, Garching; 4Observatoire de Haute Provence; 5Physikalisches Institut, Bern; 6Service d’Aéronomie du CNRS; 7Aarhus University, 8Lund Observatory,9Tautenburger Landessternwarte; 10Laboratoire d’Astrophysique de Marseille; 11Rutherford Appleton Laboratory

Introduction

An extensive review of past, presentand future research on extrasolar plan-ets is given in the article “ExtrasolarPlanets” by N. Santos et al. in the pres-ent issue of The Messenger. Here wewant to mention only that the search forextrasolar planets and the interpreta-tion of the scientific results haveevolved in recent years into one of themost exciting and dynamic researchtopics in modern astronomy.

As a consequence, ESO decided tohave a dedicated spectrograph built tobe installed at the La Silla 3.6-m tele-scope. An Announcement of Oppor-tunity was issued in 1998 and a con-sortium chosen for the realization of theproject which consists of Observatoirede Haute Provence (F), Physikalisches

Institut der Universität Bern (CH) andService d’Aéronomie du CNRS (F) un-der the leadership of the Observatoirede Genève (CH) where the PrincipalInvestigator and the Project Office arelocated. Further contributions comefrom ESO-La Silla and ESO Head-quarters (Garching). The agreementwas finally signed in August 2000.

According to this agreement, theHARPS Consortium bears the cost forthe spectrograph and all its compo-nents whereas ESO provides theCassegrain Fibre Adapter, the fibre link,the dedicated HARPS room in the tele-scope building and the complete detec-tor system. In return, the HARPSConsortium will be granted 100 HARPSobserving nights per year for a period of5 years after successful ProvisionalAcceptance in Chile. HARPS will of

course also be offered to the astronom-ical community like any other ESO in-strument.

The Radial Velocity Method:Error Sources

A description of the radial velocitymethod for the detection of extrasolarplanets in general and of the CORALIEspectrograph installed at the SwissLeonhard Euler Telescope at La Silla inparticular was given by Queloz & Mayor(2001); it describes the technique ap-plied also with HARPS and the spec-trograph which is its direct ancestor. Aninstrument like CORALIE achieves anaccuracy in radial velocity determina-tion of about 3 m/s, and with this accu-racy planets down to a minimum massof about one third the mass of Saturn