1

No. 98 – December 1999

Js versus Js–Ks colour-magnitude diagrams of NGC 3603. The left-hand panel contains all stars detected in all three wave-bands in the entire field of view (3.4′ × 3.4′, or 6 pc × 6 pc); the centre panel shows the field stars at r > 75″ (2.25pc) around the cluster, and the right-hand panel shows the cluster population within r < 33″ (1pc) with the field stars statisticallysubtracted. The dashed horizontal line (left-hand panel) indicates the detection limit of the previous most sensitive NIR study(Eisenhauer et al. 1998). The right-hand panel also shows the theoretical isochrones of pre-main-sequence stars of differentages from Palla & Stahler (1999) and the main sequence for dwarfs. For comparison, some corresponding stellar masses havebeen plotted next to the isochrones.

First Scientific Results with the VLT in Visitor and Service Modes

Starting with this issue, The Messenger will regularly include scientific results obtained with the VLT. One of theresults reported in this issue is a study of NGC 3603, the most massive visible H II region in the Galaxy, with VLT/ISAACin the near-infrared Js, H, and Ks-bands and HST/WFPC2 at Hα and [N II] wavelengths. These VLT observationsare the most sensitive near-infrared observations made to date of a dense starburst region, allowing one to in-vestigate with unprecedented quality its low-mass stellar population. The sensitivity limit to stars detected in allthree bands corresponds to 0.1 M0 for a pre-main-sequence star of age 0.7 Myr. The observations clearly showthat sub-solar-mass stars down to at least 0.1 M0 do form in massive starbursts (from B. Brandl, W. Brandner,E.K. Grebel and H. Zinnecker, page 46).

1. Introduction

After several years on a bumpy road,the support for the VLT Interferometerproject has increased dramatically overthe last year with the appointment ofeight new staff members, with the foun-dation of the NOVA1-ESO VLTI ExpertiseCentre (NEVEC) at the Leiden Ob-

servatory and with institutes and institu-tions from other European countries (Italy,Switzerland and Belgium) on the vergeof joining the project in one way or an-other.

Major contracts for the Delay Line Sys-tems and for the Auxiliary Telescope Sys-tems (ATs) were signed in 1997 and 1998respectively, the work on the science in-struments MIDI and AMBER has pro-gressed very well in the last two yearsand the commissioning instrument VIN-CI is coming closer to reality. Several

smaller contracts for the Test Siderostats,for the Coudé Optical Trains of the fourUnit Telescopes (UTs), and for a Feasi-bility Study of the dual-feed facility PRI-MA were also placed.

We are now planning for first fringeswith the commissioning instrument andthe Siderostats before the beginning ofthe new Millenium.

In this article we will give a report onthe status of the project and we will de-scribe the strategy for making the VLTIthe observatory of the 21st century.

2

TELESCOPES AND INSTRUMENTATION

The VLTI – The Observatory of the 21st CenturyA. GLINDEMANN, R. ABUTER, F. CARBOGNANI, F. DELPLANCKE, F. DERIE,A. GENNAI, P. GITTON, P. KERVELLA, B. KOEHLER, S. LÉVÊQUE, G. DE MARCHI,S. MENARDI, A. MICHEL, F. PARESCE, T. PHAN DUC, M. SCHÖLLER, M. TARENGHI andR. WILHELM

European Southern Observatory

Figure 1: The optical layout of the VLTI with two telescopes. The telescopes represent both UTs and ATs that have the same optical design.The Coudé Optical Trains are the mirrors after the tertiary mirror up to the Coudé Focus. Two Delay Lines are shown to demonstrate the prin-ciple of operation. The VLTI laboratory is represented by the beam-combining lens forming fringes.

1NOVA is the Netherlands Research School forAstronomy.

2. The Sub -Systems of the VL TI

The layout of the VLTI is displayed inFigure 1, for the sake of simplicity withonly two telescopes. A star at infinity il-luminates the apertures in the two tele-scopes with a plane wave that is guidedthrough the Coudé Optical Trains into theDelay Line Tunnel. The delay in arrivalof the light at telescope 1 with respect totelescope 2 is compensated by the de-lay lines so that the beams have zero Op-tical Path Difference (OPD) when they in-terfere on the detector in the VLTI labo-ratory. The field of view of the VLTI is 2arcsec. However, the dual-feed facilityPRIMA will allow picking two stars at theCoudé focus of ATs or UTs each in a 2arcsec field of view and separated by upto 1 arcmin.

The measurements of the contrastof the fringe pattern and of its phase,i.e. the position of the white light fringewith respect to the nominal zero OPDposition are the tasks of the VLT Inter-ferometer. Doing these measurementsfor many different baselines (differentin length and orientation) allows re-constructing an image with the angularresolution of a single telescope with a di-ameter equal to the longest baseline.Since the longest baselines are 130 m forthe Unit Telescopes and 200 m for theAuxiliary Telescopes and since the spec-ified precision for the measurements ofthe fringe contrast is very high, the re-quirements e.g. for the Delay Line sys-tems are very tough. Here, the beam tiltaccuracy has to be better than 1.5 arc-seconds (15 milliarcsec on the sky) andthe absolute position accuracy is 1µmover 65 m of travel in the tunnel. Similaraccuracy values are required for all mir-rors and mirror mounts of the VLTI, in-cluding the 8-m primary mirrors. Also, theenvironmental conditions have to be verystable in order to minimise internal tur-bulence.

It is the advantage of the VLTI thatthese considerations have been driv-ing the design of the Unit Telescopesand of the Auxiliary Telescopes, aswell as the construction of the infra-structure. Several measurement cam-paigns have confirmed that both opto-mechanical and environmental specifi-cations are met.

The status of the VLTI sub-systems isas follows:

Test Siderostats

For tests with the VLTI and the in-struments, two Test Siderostats were de-signed and built (for details see [3]). Thefree aperture of 400 mm allows for about20 stars to be observed in the N-band at10 µm. Since for the VLTI Control Sys-tem the Siderostats will ‘look’ the sameas the UTs and ATs, it is planned to com-mission the VLTI with the Siderostats.They will be located on the AT stationsand will thus be able to use baselines be-tween 8 and 200 m. The systems are

readily manufactured and they will be de-livered to Paranal in January 2000. Theimplementation of the Siderostat controlsoftware will be finished by May 2000. Itis planned to have first fringes with theSiderostats and VINCI in December2000.

Delay Line Systems

Three Delay Lines have been ordered;the first two will be installed in the De-lay Line Tunnel in July 2000, the third onein December 2000. The systems consistof the cat’s-eye mirror system retro-re-flecting the incoming parallel beam, andof the carriage transporting the cat’s-eyealong the delay lines. Figure 2 shows thecarriage on its three wheels in the clean-room at the manufacturer. The three-mir-ror optical design of the cat’s-eye has avariable curvature mirror (VCM, see [4])in the focus of the cat’s eye in order tore-image the telescope pupil into a fixedposition in the VLTI laboratory while theDelay Line System is tracking. Thecat’s-eye can handle two input beams asrequired for a dual feed system. A totalof eight Delay Line Systems can be host-ed in the Delay Line Tunnel.

Auxiliary Telescope Systems

In June 1998, two 1.8-m Auxiliary Tele-scopes were ordered, and in June 1999,the option for the third AT was exerted.A model of the telescope is displayed inFigure 3. The first two telescopes will beready for the VLTI in June 2002, the thirdin October 2002. The telescopes are re-locatable on 30 stations of the VISA (VLTInterferometer Sub-Array) providingbaselines between 8 and 200 m. Using

three telescopes and thus three baselinesat the same time will allow the applica-tion of closure phase techniques elimi-nating the influence of atmospheric tur-bulence on fringe position. Each AT willbe equipped with a tip-tilt system cor-recting for the fast image motion inducedby atmospheric turbulence. Under theseeing conditions at Paranal, tip-tilt cor-rection on a 1.8-m telescope in the nearinfrared means almost diffraction limitedimage quality. One should note that theATs are available exclusively for the VLTI,forming an observatory that is operatedindependently of the Unit Telescopes.

Coudé Optical Trains and TransferOptics

The five mirrors after the tertiary mir-ror of the Unit Telescopes form the CoudéOptical Train with a field of view of2 arcmin (see Fig. 1). The last mir-ror before the Coudé Focus is very closeto the telescope exit pupil and can be re-placed by the deformable mirror of theadaptive optics system. The image qual-ity at 2.2 µm is diffraction limited onthe optical axis and better than 97% overthe full field of view. The Coudé OpticalTrains for all four UTs have been or-dered and will be installed at Paranal inOctober 2000.

The transfer optics between the CoudéFocus and the Delay Lines are current-ly being designed. The concept is to avoidthe need of human intervention in the de-lay line tunnel when switching betweenAT stations or UTs. Thus, the mirrors inthe Delay Line Tunnel reflecting the lightinto the Delay Lines and from the DelayLines into the VLTI Laboratory will be re-motely controlled.

3

Figure 2: The carriage for the cat’s-eye of the Delay Line System in the Clean Room at Fokkerin Leiden. The carriage has three wheels running on the 65-m steel rails. The diameter of thewheels is about 40 cm and the length of the carriage 2.25 m.

VLTI Laboratory

The concept driver for the optical lay-out of the VLTI Laboratory (see Fig. 4)was to provide the same beam diameter

for the interferometric instruments bothwhen observing with UTs or ATs. There-fore, a beam compressor reduces the 80mm beam diameter from the UTs to18 mm matching the diameter of 18 mm

provided by the ATs. The nominal posi-tion of equal optical path length is set forall beams at the same distance after theswitchyard (indicated by the red lineZOPD in Fig. 4) to simplify the opticalalignment of the interferometric instru-ments.

3. Phase A – Two TelescopeInterferometry

In accordance with the recommen-dation by the Interferometry ScienceAdvisory Committee (ISAC) laid outin the VLTI New Plan [12], the VLTIin its first phase will combine two tele-scopes, and near- and mid-infrared in-strumentation will be available. An adap-tive optics system for the UTs and afringe sensor unit (FSU) will be imple-mented as well. In the near infrared,adaptive optics on the UTs is mandato-ry if the effective telescope aperture is notto be reduced to the size of the Fried Pa-rameter, r0, which is about 1 m in the nearinfrared.

4

Figure 3: A model of an Auxiliary Telescope during Observations. The 1.8-m telescope (withan Alt-Az mount like the Unit Telescopes) is rigidly anchored to the ground by means of a spe-cial interface. The light is directed via the Coudé Optical Train to the bottom of the Telescope.From the Coudé Relay Optics it is sent on to the underground Delay Line Tunnel

Figure 4: The layout of the VLTI Laboratory.The switchyard can direct the beam into fourdirections: (1) to the interferometric instrumentson the left without beam compression, (2) tothe interferometric instruments after beamcompression, (3) to the Differential Delay Lines(DDLs) after beam compression, and (4) to theDifferential Delay Lines without beam com-pression. The beam compressor reduces thebeam diameter from 80 mm to 18 mm and itre-images the telescope pupil into the PRIMAFSU, MIDI and VINCI (the position of the pupilis indicated by a diagonal line). A possible lo-cation for a visitor instrument is shown on theright. D

VINCI – The CommissioningInstrument

The commissioning instrument of theVLTI is a conceptual copy of FLUOR, thenear-infrared interferometric instrumentof the IOTA interferometer on Mount Hop-kins in Arizona (US) [2]. The idea is to lim-it the technical risk and to facilitate thecommissioning of the VLTI by using thedesign of an existing well functioning in-strument (see e.g. [6]). The main com-ponent of VINCI (see Fig. 5) is a fibrebeam combiner using the light from twotelescopes as input and producing fouroutputs, two photometric and two inter-ferometric signals. By varying the OPDbetween the beams with an internal mod-ulator, a temporally modulated fringe pat-tern is produced on the detector. In ad-dition to serving as the interferometric in-strument, VINCI provides alignmenttools and reference sources for the VLTIand the scientific instruments. VINCI willbe delivered to Paranal in October2000.

MIDI

An interferometric instrument in themid-infrared at 10 µm in combination with8-m telescopes is a novelty, promis-ing exciting results on new types of ob-jects.

MIDI is being designed and built by aEuropean consortium led by the Max-Planck-Institute for Astronomy in Heidel-berg. The design philosophy is to havea simple instrument concept combiningtwo beams and providing a moderatespectral resolution (P 200). The chal-lenge lies in controlling the high thermalbackground. It is estimated that the in-dividual beams will have an emissivity ofabout 50%, which corresponds to anequivalent photon noise on the sky of53 mJy per Airy disk (with λ/D = 0.26″ foran aperture of 8 m) for a broad band 10µm filter (∆λ = 4 µm). Since the signal-to-noise-ratio scales as S/N∝ D2 the avail-ability of the Unit Telescopes is a hugeadvantage over smaller apertures. Thedetails of the instrument are described in[7]. MIDI will be delivered to Paranal inJune 2001; first light with the Siderostatsis planned for September 2001.

AMBER

The near infrared instrument of theVLTI, AMBER, will operate between 1 and2.5 µm, at first with two telescopes witha spectral resolution up to 10,000. As not-ed above, for UT observations in thenear-infrared adaptive optics is manda-tory. The magnitude limit of AMBER onthe UTs is expected to reach K = 20 whena bright reference star is available (i.e.with a dual feed facility) and K = 14 oth-erwise. The European consortium incharge of designing and manufacturingthis instrument is led by the Universitiesof Nice and Grenoble. AMBER has beendesigned for three beams to enable im-

aging through phase closure techniques[10]. It is planned to start commissioningAMBER with the Siderostats in February2002.

MACAO

The adaptive optics system MACAOwill have a 60-actuator bimorph mirrorand a curvature wavefront sensor. Thedeformable mirror will replace one ofthe mirrors of the Coudé optical train,thus requiring no additional optical ele-ments. As noted above, MACAO is es-sential for all near-infrared instrumenta-tion to be used with the Unit Telescopesincluding the Fringe Sensor Unit operat-ing in the H-band. This means that alsoa mid-infrared instrument like MIDIneeds adaptive optics in order to improvethe limiting magnitude by using a FringeTracker.

The first MACAO system will be in-stalled on one of the Unit Telescopes inDecember 2001. It is planned to haveMACAO ready for interferometric obser-vations with two UTs in June 2002.Clones of the MACAO system will beused for future VLT instrumentation likeSINFONI. MACAO is an in-house de-velopment [1].

4. Phase B – Imaging

The main drivers for the secondphase of the VLTI are increasing the sen-sitivity of the interferometer and addingimaging modes. In addition, an astro-metric mode will open the door for manythrilling scientific programmes.

Increasing the sensitivity of the VLTIcalls for a dual feed facility. Then, the

sensitivity is improved by using a brightguide star for fringe tracking – simi-lar to the guide star in adaptive optics forwavefront sensing – in one of the two‘feeds’ allowing increasing the expo-sure time on the science object in the oth-er feed up to 10–30 minutes dependingon the position in the sky. With ahigh-precision laser metrology systemused to determine the internal opticalpath length, a dual feed system also pro-vides an imaging mode. The VLTI dualfeed facility PRIMA is described in the fol-lowing.

Imaging with an interferometer relieson filling efficiently the UV plane withmany different baseline vectors. Al-though this can be done by observing withtwo telescopes in many different positionsusing a dual feed system, it is more effi-cient to use three or more telescopes atthe same time and eliminate the influenceof atmospheric turbulence by applyingphase closure techniques. Therefore, thethird and fourth delay lines as well as thethird Auxiliary Telescope are included inthis phase.

PRIMA

The Phase Referenced Imaging andMicro-arcsec Astrometry (PRIMA) facil-ity is a dual-feed system adding a faint-object imaging and an astrometry modeto the VLTI [5, 11]. PRIMA enables si-multaneous interferometric observationsof two objects – each with a maximumsize of 2 arcsec – that are separated byup to 1 arcmin, without requiring a largecontinuous field of view. One object willthen be used as a reference star for fringetracking whilst the other object will be the

5

Figure 5: The concept of the VLTI commissioning instrument VINCI. The beam combining unit,VINCI, and the reference source unit, LEONARDO, are placed on individual tables. Their re-spective positions in the VLTI Laboratory are shown in Figure 4. The main elements on theVINCI table are the fibre beam combiner and the dewar of the infrared detector. The insertedphotographs show details of the FLUOR instrument at the IOTA interferometer. VINCI is a con-ceptual copy of FLUOR.

science target. As a detector for PRIMAeither the two scientific instrumentsMIDI and AMBER can be used makinguse of the fringe stabilisation provided byPRIMA, or a dedicated PRIMA detectorfor high-precision astrometry that will bedesigned in the next years.

PRIMA is the key to access: (1) High-er sensitivity, the limiting magnitude willbe about K = 20, (2) imaging of faintobjects with high angular resolution(<10 milliarcsec), (3) high-precision as-trometry (P 10 µarcsec over a 10-arcsecfield).

The principle of operation relies onfinding within the isoplanatic angle (P 1arcmin) of the science target a sufficientlybright star (H P 12) that can be used asa reference star for the stabilisation of thefringe motion induced by atmosphericturbulence (see Fig. 6). Controlling all op-tical path lengths of the reference star andof the science star inside the interfer-ometer (OPDint) with a laser metrologysystem introduces the capability of im-aging faint objects and of determining theprecise angular separation between thetwo stars. The measurement has to berepeated for up to 30 min in order to av-erage out the variations of the differen-tial OPD caused by atmospheric turbu-lence (OPDturb).

PRIMA can be subdivided into thefour sub-systems (see Fig. 7): Star Sep-arator, Laser Metrology System, Differ-ential Delay Lines and Fringe SensorUnit. These sub-systems contain a num-ber of technological challenges, in par-ticular in the area of laser metrology,that require careful analysis of thepossible technical solutions. Therefore,a feasibility study for the four subsystemswas performed by Dornier Satellite Sys-tems, Friedrichshafen, and by ONERA,Paris, to find the best technical so-

lutions and to ob-tain a thorough fi-nancial estimatefor the manufac-turing of the sub-systems. The re-sults of the studybecame availablein July 1999 and aCall for Tender forthe manufacturingof the system isplanned early in2000.

Dual feed ob-servations withPRIMA can startas soon as the starseparator, thefringe sensor unitand the differentialdelay lines areready. Then, a ref-erence star can beused for fringetracking while inte-grating on thefainter science ob-ject as describedabove. If the fringepattern can be sta-bilised over 10–100 sec the ex-pected limitingmagnitudes areabout K P 16 andN P 8.

Phase informa-tion required forimaging and as-trometry becomes available if the lasermetrology system is installed. An OPDmeasurement accuracy of 500 nm rmsover 10 min sets the limiting magnitudesto about K P 20 and N P 11. The Strehlratio in the reconstructed image can beas good as 30% in the K-band and80% in the N-band depending on theUV coverage. Reaching the final goal of5 nm rms over 30 min allows 10 µarcsec

astrometry. PRIMA shall be operationalby 2003.

5. First Fringes and BeyondThe major challenge in making the

VLTI work is the complexity of the sys-tem. Not only are the individual sub-sys-tems high-tech instruments but they arealso scattered over an area up to 200 min diameter. The quality of the control sys-

6

Figure 6: Principle of phase referenced im-aging and astrometry with an interferometer.The difference in the positions of the white lightfringes of object and reference star are de-termined by the OPD given by the product of∆S, the angular separation vector of the stars,and B, the baseline vector, by the phase φofthe visibility function of the science object, bythe OPD caused by the turbulence, and by theinternal OPD.

Figure 7: Functional block diagram of the VLTI and the PRIMA com-ponents. The VLTI has the following subsystems: Telescope 1 and 2– two UTs or two ATs, AO – Adaptive Optics, Delay Lines, AMBER/MIDI– Science Instruments. The PRIMA subsystems are: Star Separator,Differential Delay Line, Laser Metrology, Fringe Sensor Unit.

Figure 8: The UV coverage on the left and the point spread function (PSF) on the right with afull width at half maximum of 4 mas respectively 8 mas in the narrow respectively wide direc-tion of the PSF at 2.2 µm. The UV coverage and the PSF are calculated for –15° declinationand 8 hours of observing when combining all four UTs. Producing images with this quality isthe ultimate goal for the VLTI.

tem is a key issue for the smooth oper-ation of the complete system. From thebeginning of 2000 onwards, every 6months major new sub-systems and in-struments arrive at Paranal and have tobe integrated into the VLTI. Thus, thestrategy for commissioning has to be de-fined very carefully.

For the first milestone of the VLTI pro-gramme, First Fringes with the Sidero-stats and VINCI, we have adopted the fol-lowing approach: after installation andcommissioning of the Siderostats, of theDelay Lines and of the Transfer Optics,the optical alignment of the full system willbe done with a Technical CCD. Then, alldegrees of freedom but one can be ad-justed and the dynamic behaviour of thedelay lines can be tested with a star ob-served with the siderostats. With thecommissioning instrument VINCI the lastremaining degree of freedom, the opticalpath difference, will be adjusted to pro-duce interferometric fringes before theend of 2000. After first fringes, in the firsthalf of 2001, the VLTI will be optimisedby using different AT stations for theSiderostats and by integrating the thirdDelay Line into the system. While this istaking place the Coudé Optical Trains inthe UTs and the remaining transfer op-tics are being installed.

The next instruments to arrive atParanal are MIDI in the middle and AM-

BER at the end of2001. Both scienceinstruments will becommissioned withthe Siderostats.MIDI can be testedshortly after firstfringes on theSiderostats withthe UTs even be-fore the adap-tive optics systemMACAO arrives.After installationand commission-ing of MACAO withthe VLTI and withVINCI in the sec-ond half of 2002,AMBER will betested on the UTs.

Also in 2002, atotal of three Aux-iliary Telescopeswill be commis-sioned, first indi-vidually then withthe VLTI and VIN-CI. The science in-struments will fol-low suit. The firstsub-systems ofPRIMA arrive aswell in 2002. Dur-ing the course of2003 both dualfeed and closurephase operationcan be tested withthe VLTI.

In order to operate the VLTI as an ob-servatory, support astronomers, instru-ment operators and engineers are re-quired for regular operations and formaintenance. Although the number ofnights with two (or more) UTs is proba-bly restricted to a small fraction of the to-tal number of available nights in the firstfew years, first the Siderostats and thenthe Auxiliary Telescopes are availableevery night for VLTI observations. Thus,the planning for the VLTI has to foreseea full-time coverage with technical andscientific personnel. One has to keep inmind that the VLTI is running in parallelto the VLT observatory with Unit Tele-scopes and requires the same level oftechnical support for telescopes and in-struments.

The final goal of the VLTI is to pro-duce images with a few milliarcsecondsresolution. Figure 8 shows a simulatedpoint spread function when observingfor eight hours with all four UTs. The re-sult shows a very impressive albeit elon-gated PSF with a full width at half max-imum (FWHM) of about 4 × 8 milli- arc-sec for a wavelength of 2.2 µm. Havingthis goal in mind, one must not forget thatfor several years the results will be in-dividual measurement points in the UVplane represented by curves like thosein Figure 9. However, this does not di-minish the scientific content of these

results. The full scientific potential of theVLTI in its different phases is describedin [9] summarised by F. Paresce [8].

6. The Future

The scenario of VLTI observations de-scribed in the article will without any doubtprovide a wealth of scientific discoveriesbut it still does not exploit the full capac-ity of the infrastructure at Paranal. Thelayout of the interferometric tunnel andof the VISA array allows combining morethan three telescopes providing either aneven better coverage of the UV plane orthe operation of different instruments atthe same time.

Therefore, the four final Delay LineSystems and all remaining Auxiliary Tele-scopes for a total of 8 ATs shall be in-stalled for VLTI operations. The comple-tion of the final phase will allow VLTI ob-servations with 28 simultaneous base-lines. The planning also has to includenew, the second-generation, instrumen-tation allowing the combination of six toeight telescopes and at the same timeproviding a dual feed facility for each pairof telescopes.

Beyond the fully equipped VLTI therehave to be kilometric arrays with 8+ m tel-escopes in order to drive the angular res-olution into the sub-milliarcsec range andto improve the sensitivity well beyond K= 20. The VLTI will be the stepping stonetowards these observatories.

References

[1] Bonaccini, D., Rigaut, F., Glindemann, A.,Dudziak, G., Mariotti, J.-M. and Paresce,F. 1998, Proc. SPIE 3353, 224–232.

[2] Coudé du Foresto, V., Perrin, G., Mari-otti, J.-M., Lacasse, M. and Traub, W.A.1996, Integrated Optics For Astronomi-cal lnterferometry, Grenoble, P. Kern andF. Malbet eds., 110–125.

[3] Derie, F., Brunetto, E. and Ferrari, M.1999, The Messenger 97, 12–13.

[4] Ferrari, M. and Derie, F. 1999, The Mes-senger 97, 11.

[5] Glindemann, A. and Lévêque, S. 1999VLT Opening Symposium, in press.

[6] Kervella, P., Coudé du Foresto, V.,Traub, W. A. and Lacasse, M.G.1999,ASP Conf. Series XX, 51.

[7] Leinert, Ch. and Graser, U. 1998, Proc.SPIE 3350, 389–393.

[8] Paresce, F., et al. 1996, The Messenger83, 14–21.

[9] Paresce, F., Ed. 1997 Science with theVLT lnterferometer. Springer-Verlag.

[10] Petrov, R., Malbet, F., Richichi, A. and Hof-mann, K.-H. 1998, The Messenger 92,11–14.

[11] Quirrenbach, A., Coudé du Foresto, V.,Daigne, G., Hofmann, K.-H., Hofmann,R., Lattanzi, M., Osterbart, R., Le Poole,R., Queloz, D., Vakili, F. 1998, Proc. SPIE3350, 807–817.

[12] von der Lühe, O., Bonaccini, D., Derie,F., Koehler, B., Lévêque, S., Manil, E.,Michel, A. Verola, M. 1997, The Mes-senger 87, 8-14.

7

Figure 9: The type of results obtained with VINCI. The two top rowsdisplay the photometric signals from each telescope, the third row dis-plays one of the interferometric outputs and the fourth row displays thesame signal as in the third row calibrated with the photometric signalsto obtain a clean fringe pattern.

E-mail: aglindem@eso.org

and very low read noise are required.Thus photon noise is the only unavoid-able contributor to the WFS noise. Evenso, with Natural Guide Stars (NGS) theperformance in the K-band of a medi-um-sized AO system drops significantlywhen mV ≥15. Figure 2 shows the ex-pected K-band Strehl performance for a60-element curvature system with theVLT as a function of NGS magnitudeand its angular distance from the sci-ence target. Note that Strehl curves de-pend by the seeing at 6th power. Ouranalytical computations are for 0.66"

seeing at 0.5 µm, 6 msec τo, and 0.4Strehl contribution from field anisopla-natism at 20" elongation, in K-band.

The analytical simulations allow aparametric exploration of the AO per-formance. Although this is less accuratethan a full numerical simulation, the re-sults are according to our estimates ac-curate to within 10-15%, and serve wellthe purpose of this discussion. The ef-fect of anisoplanatism on AO systems isvery evident in Figure 2. For example, a10th magnitude NGS which is 25 arc-seconds away from the science target

8

1. Introduction

ESO has been studying the possibil-ity of introducing a Laser Guide StarFacility (LGSF) for the VLT to serve theadaptive optics (AO) instruments fore-seen on the UT3 telescope of theParanal Observatory: NAOS-CONICAand SINFONI.

Although full project funding, accord-ing to the ESO Long Range plan, willonly start in 2001, we have performed acomprehensive feasibility study, withsome innovative developments, espe-cially on the possible injection of thelaser light through a single-mode fibre(see the following article). The LGSFconcept is the basis of this article.

The LGSF is foreseen as an add-onto the VLT telescope, able to create anartificial guide star in the mesosphericsodium layer, at 90 km average altitude.The LGSF would be used as a slavesystem to the AO systems, to producean artificial reference star for the wave-front sensor, close to the observed ob-ject, whenever necessary. Figure 1 il-lustrates how the baseline LGSF con-cept would be implemented on the VLT.

The activities reported here havebeen carried out at ESO, with the en-thusiastic scientific collaboration of theMax-Planck-Institut für Extraterrestri-sche Physik (MPE) in Garching,. MPEhas built the Calar Alto Laser for theALFA system (Davies et al, 1999).

The definition of the LGSF is at theconceptual design stage, with a solidbaseline solution explored in detail andtwo open options: the final choice forthe laser, and the use of a fibre relay forthe laser beam instead of a mirror sys-tem. The final choices will be made dur-ing the LGSF Conceptual DesignReview.

This article gives an overview of theconceptual design we are currentlyworking on.

2. Project Rationale

The most important limitation of cur-rent NGS-AO systems, is the scarcity ofa good reference star for the wavefrontsensor (WFS) to exploit the full poten-tial gain of the adaptive optics loop. TheNGS should be bright enough to over-come both sensor read-out and photonnoise. If it lies outside of a small field,optimum correction cannot beachieved, due to anisoplanetism ef-fects.

To minimise the former, wavefrontsensors with a high quantum efficiency

Laser Guide Star Facility for the ESO VL TD. BONACCINI, W. HACKENBERG, M. CULLUM, M. QUATTRI, E. BRUNETTO, J. QUENTIN, F. KOCH, E. ALLAERT, A. VAN KESTEREN

European Southern Observatory

Figure 1: Layout of the LGSF. The laser and its controls are mounted under the Nasmyth plat-form, in a well-insulated laser room. The laser beam is relayed to the 50-cm-diameter launchtelescope located behind the UT secondary mirror.

gives the same Strehl ratio as a 17thmagnitude NGS on-axis.

The scientific justification of theLGSF comes from sky-coverage con-siderations. Put in simple terms, weneed the answer to the question: ‘whatfraction of the sky can be observed withAO, and how many of my objects canbe observed?’ (Le Louarn et al., 1998,Di Serego Alighieri et al., 1994) The re-quirement of a relatively bright NGS, atshort distance from the science object,limits the use of AO systems. The high-er the Strehl Ratio required the greaterthe limitation. Figure 3 shows the medi-an sky coverage iso-probability curvesfor a NGS plotted on a logarithmicscale.

Using a LGS, the reference signalfor the AO wavefront sensor can be

made equivalent to that of a NGS of mV≅ 9.5 with seeing-limited diameter. It israther like providing the telescope witha headlight, enabling the AO system tofind a reference star wherever the tele-scope points. This opens the possibilityof observing many more extragalactictargets than with a NGS-AO system.Two problems still prevent us reachingcomplete (100%) sky coverage: (a) the‘focus anisoplanatism’ or cone effect(Tyler, 1994), and (b) the intrinsic inde-termination of the image motion (tip-tilt) signal. The cone effect, for an 8-mtelescope in the Near Infra-Red (NIR),and one single LGS, gives an unavoid-able but tolerable reduction of theStrehl performance.

The tilt problem can be overcome byusing a natural star in the field in addi-

tion to the LGS. The image blur is cor-rected using the LGS, the image motion(tip-tilt) using a NGS. The NGS for tip-tilt can, however, be much fainter andtypically 2.5 times further away than ahigh-order NGS reference.

The ultimate sky coverage and AOperformance depends largely on thetip-tilt sensor and is one of the reasonswhy ESO has put considerable effortto develop a high-performance tip-tiltcorrection subsystem, the so-calledSTRAP system. We will report on theSTRAP system in a next issue of TheMessenger.

The K-band iso-Strehl contours for aLGS-AO 60-element curvature sensingsystem are shown in Figure 4. TheNGS magnitude and elongation axis re-fer to the tip-tilt reference star. The LGSis assumed to be pointed close to thescience target. Compared to Figure 2,the gain in sky coverage is very evidenteven though the cone effect limits themaximum achievable Strehl. There isan even larger gain if one looks at theencircled energy improvement for aspectrograph (Le Louarn et al., 1998,Bonaccini, 1996).

Figure 5 combines the informationcontained in Figures 2 and 4, andshows the overall sky coverage (com-bining NGS magnitude and elongation)achieved with a given Strehl ratio, forNGS-AO and LGS-AO systems. Thesimulated system corrects 80 Karhu-nen-Loewe modes1. Using a LGS-AO,

9

Figure 2: K-band iso-Strehl curves, computed for a 60-element cur-vature system, vs NGS magnitude (x-axis) and elongation betweenscience object and NGS (y-axis, arcsec).

Figure 3: Logarithmic iso-probability curves for NGS. V-band NGS mag-nitude on the horizontal axis, NGS elongation from the science objectin arcsec, on the vertical axis.

Figure 4: K-bandiso-Strehl curves,for a 60-elementcurvature sensingAO system with aLGS, using a tip-tiltnatural guide star.The NGS V-magni-tude and elongation(arcsec) from thescience target arethe horizontal andvertical axes re-spectively. This plotshould be comparedwith Figure 2.

1Karhunen-Loewe modes: they are an orto-nor-mal basis for the expansion of the wavefront. Theyare often used in AO systems, as the modes are alsostatistically independent (see e.g. Adaptive Opticsin Astronomy, p. 31, F. Roddier ed., Cambridge PressUniversity 1999).

the maximum K-band Strehl obtainableis 0.52 in our example, and this is seenin the saturation of the LGS-AO skycoverage curve. The sky coverage for aStrehl Ratio < 0.45 in K-band is alwayslarger than NGS-AO in this configura-tion.

The LGS-AO system foreseen for theVLT will not reach Strehl ratios muchgreater than 40% in K-band, due to thecone effect. Note that the PSF coreachieves the resolution of the diffractionlimit (0.057″ in K-band), therefore thespatial information can be recovered.The images may be further boostedusing post-processing techniques suchas deconvolution if necessary (Conanet al. 1998, Christou et al., 1999). Fig-ure 6 shows the long-exposure Modu-lation Transfer Function (MTF) for theuncorrected seeing, for the LGS-AO

and for the ‘perfect’ telescope. TheLGS-AO system allows the attenuationof all the spatial frequencies transmittedby the telescope to be restored by post-facto deconvolution.

Techniques to reduce or remove theLGS cone-effect are being studied atvarious institutes, for example by usingmultiple guide stars (Fried and Belsher,1994), or with a hybrid system wherethe low orders are corrected with theNGS, the remaining high orders withthe LGS. These will be necessary forvisible AO compensation on future ex-tremely large aperture telescopes.

3. Project Status

ESO’s Scientific Technical Commit-tee has confirmed the scientific rele-vance of the LGSF, and asked to pre-

pare a proposal for equipping with it theVLT Unit Telescope hosting AO sys-tems. A baseline solution has beenstudied in detail during 1998, for thesubsystems of the laser, the beam relayand the launch telescope. The prelimi-nary design activities were frozen inFebruary 1999, except for the studieson a single-mode fibre relay. This wasto allow the feasibility of the fibre relaysystem to be thoroughly investigatedbefore proceeding further.

The past months have been spent onthe development of a prototype single-mode fibre relay system, from the laseroutput to the launch telescope. The fi-bre-optics industry has been involved inorder to have access to state-of-the-artfibre production technologies. We haveused a 532 nm, 5W Continuous Wave(CW), Verdi* laser from CoherentGmbH, to do the laboratory experi-ments and studies. The results of ourstudies and tests so far indicate that thefibre relay solution, although not aseasy as it seems, is feasible and alsovery attractive. We plan to conclude thisdevelopment with working fibre relaysin June 2000, and resume LGSF de-tailed design activities in the summer2000. Assuming that funding is forth-coming as planned, one could envisagecommissioning of the LGSF on the VLTby the end of 2003.

A brief explanation of the baselineLGSF design is given below and repre-sents the present conceptual designaround which we are working. It is like-ly to be modified somewhat during thepreliminary and final design phases.

4. High Level Operational andFunctional Requirements

The LGSF is seen by the AO instru-ments as a telescope server facility. It istherefore compliant with the VLT re-quirements/standards and integratedhardware and software with the tele-scope environment. The AO systems(NAOS and MACAO) on UT3 will beable to send basic commands and re-ceive the diagnostic status.

The LGSF will never take independ-ent initiatives except for safety reasons,such as air-traffic alerts from the dedi-cated monitoring camera. In this case,it notifies to the clients that it is going toshut down the beam propagation be-fore doing it.

Concerning light pollution from theLGS, work is in progress to determine itquantitatively (Delplancke et al., 1998).The monochromatic laser beam scat-tering does not disturb the IR instru-ments. The operational scheme pro-posed is to create an area-of-avoidancearound the laser beam: the LGSF mon-itors the pointing of the other Paranaltelescopes, and creates an alert status

10

Figure 5: Sky coverage of an AO system correcting 80 Karhunen-Loewe modes, as a functionof the K-band Strehl Ratio. The model atmosphere is as described in the article. NGS and LGScurves are plotted. The assumed single sodium LGS has a mV = 10 and 1 arcsec FHWM. Thesaturation of the LGS curve at SR = 0.52 is due to the cone effect: Irrespective of how goodthe tip-tilt reference signal is, higher Strehl ratios are not possible.

*The Verdi is an all-solid-state diode pumpedNd:YVO4 laser, recently produced by Coherent.

Figure 6: MTF in K-band, for the VLT with uncorrected seeing, LGS-AO correction, and theperfect telescope

when another telescope is pointingclose to the cone-of-avoidance. The ac-tions to take in this case are still to bedecided.

The LGSF is required to produce amV = 9.5 equivalent artificial referencestar with a FWHM of 1.5″ (seeing limit-ed as goal). For median Paranal see-ing, down to 60 degrees zenith dis-tance, a 6.5W CW Sodium laser willserve the purpose. For worse seeingconditions, or in the presence of thin cir-rus clouds, a laser power in the rangeof 10W would be preferable.

The LGS has to be independentlypointed within 20″ from the defined op-tical axis of the VLT, with a relativepointing accuracy of 0.3″. Automatic fo-cussing of the LGS on the sodium layerwill be performed, up to zenith anglesof 60 degrees.

The safety measures of the LGSFwill comply with the Paranal observato-ry strict safety regulations, Chilean law,and the FAA regulations as applicablein Chile.

5. LGSF Baseline DesignSolution

The baseline solution study has di-vided the LGSF into three different sub-systems: (a) the laser room, includingthe laser and its servo-controls, (b) thebeam relay and (c) the launch tele-scope with its diagnostics. We reporthere briefly on (a) and (c). A companionarticle in this issue describes the proto-type fibre beam relay subsystem in de-tail. We will report in future issues ofThe Messenger design details on sub-systems (a) and (c).

We propose to use a relatively low-consumption laser (4KW), and to builda thermally stabilised laser room un-der the Nasmyth platform of UT3. Thelaser choice takes into account com-mercial availability, support, stabilityand servicing issues, so as to minimisethe LGSF operation and overheadcosts.

The baseline solution for the relayof the 20-mm-diameter laser beam from

the laser room to the launch telescopecould in principle use flat mirrors in ashielded light-path. The laser beamwould be hidden behind one of thesecondary mirror (M2) support spiderswhen crossing above the primary mir-ror. The mirror relay system was in-vestigated in detail. It is cumbersomeand difficult to implement in practice.Turbulence effects in the relay pathas well as dust contamination andthermal effects on the optics, can de-grade the LGS quality considerably.Air tightness, sev-eral servo-con-trols and alsofrequent mainte-nance would benecessary.

The alternativesolution for thelaser relay wepropose uses a30 m long, single-mode optical fi-bre. This is an in-novative solutionand not easy,given the powerdensities involved:one Watt of inject-ed power in a 5.5micron core fibrecorresponds to apower density of4 MW/cm2. Wehave found a vi-able solution tothis problem. Itwill deliver a dif-fraction-limitedbeam at thelaunch telescopeinput, with >70%throughput. Theuse of a single-mode fibre wouldresult in savingsin laser power,money and man-power.

A 50-cm diam-eter laser launchtelescope is locat-

ed behind M2. The aperture is dictatedby the optimisation of the LGS beam di-ameter for the best Paranal seeingconditions. The projected beam diame-ter will be adjusted to the actual seeingconditions. The telescope is a polychro-matic reflective design, which allowsthe LGS to be positioned within a 1-ar-cminute field on the sky, with a visiblemonitoring camera. Figure 1 shows aschematic diagram of the foreseen set-up.

The level of automatism to be builtinto the LGSF is large, in order to allowroutine and reliable operation at theVLT. This increases the initial cost andcomplexity of the facility, but we believeit will prove highly beneficial later, dur-ing operation. A number of servo-sys-tems and automatic computations areplanned, based on dedicated hard-ware/software.

6. Baseline Laser and LaserOptions

Only a complex micro-macro pulsemodulation scheme would allow slightlygreater flux returns from the mesos-pheric sodium than a CW laser (Milonniand Fugate 1997). Considering thatCW lasers are intrinsically more stableand simpler, that coatings for CW lasers

11

Figure 7: Baselinelaser conceptualscheme. Two solid-state pump lasersare combined in asingle beam, 20Wat 532 nm, pump-ing a ring-dyeresonator laser,using Rodhamine6G as dye andservo-tuned to589 nm.

Figure 8: Layout of the laser room designed, with the laser optical tableand the necessary accessories. The room is thermally isolated and con-ditioned.

are less demanding, and that CWlasers safety issues are easier to han-dle than for pulsed lasers, we are con-centrating on CW lasers solutions.

The baseline laser is a modified ver-sion of a commercial product from Co-herent, made of two 10W Nd-YV04 CWVerdi diode laser pumps at 532 nm, com-bined into the Coherent jet-streamed ringdye resonator model 899-21, using Rho-damine 6G as dye. The general advan-tages of solid state lasers as comparedto ion-lasers for pumping, are the muchhigher conversion efficiency and greatercompactness.

In addition the wavelength of fre-quency-doubled light from Nd-dopedlaser crystals is around the maximumof the absorption spectrum of Rh6G,whereas the light of argon-ion lasers isclearly off the absorption peak.

This laser configuration delivers >6.5W of CW power on a 20 MHzFWHM TEM00* line, or a LGS with mV≅ 9 after beam conditioning. A similarring-dye laser has proven to be robustand stable during the past years atCalar Alto.

The laser line is broadened to 500MHz, to avoid both the saturation ofmesospheric sodium atoms and nonlinear effects in the fibre.

We have the capability of doublingthis power in the future, by adding twosuch laser systems on the same op-tical table. This scenario would pro-duce 13W CW of sodium laser power,and would use two independent single-mode fibres to relay the beams to thelaunch telescope. The orthogonallypolarised beams at the fibre outputwould be combined in the launch tele-scope.

For safety and operational stabilityreasons, the laser bench would be in-stalled in an isolated room underneaththe Nasmyth platform. This room alsocontains the dye solution chiller, thepump laser heat-exchanger, two elec-tronics cabinets and workspace formaintenance work (Fig. 8).

The dimensions of the laser room al-low convenient access to the opticaltable. The room is temperature stabilisedand supplied with filtered air to avoidcontamination of the optics. In order notto impact on the telescope seeing thelaser room is thermally insulated, withthe outer walls at the same temperatureas the air in the telescope enclosure.

Vibration, air turbulence, temperaturevariations and dust contamination direct-ly influence the line width as well as thepower and beam-pointing stability of ajet-streamed dye laser. The dye-laser os-cillator therefore is installed in a fixedspatial position on a rigid optical table thatde-couples the laser head from vibra-tions, e.g. when the telescope is moving.

Figure 8 shows the layout of the laserroom together with the accessory de-

vices. The laseroptical table sizeis chosen so as toallow a later addi-tion of a second6.5W laser. Theroom is thermallyisolated and con-ditioned.

The total massof the laser benchequipment is es-timated to be 1.3metric tons. Weestimate the en-tire laser room toweight not morethan 4 tons. Ananalysis of thestatic and dynamic impact of the laserroom mounted under the Nasmyth plat-form on UT3 has been done and foundto be negligible. The power, communi-cation and cooling budgets have indi-cated that one VLT Service ConnectionPoint (SCP) will be sufficient for thelaser room.

7. Alternative Laser Choices

Two alternative laser choices arecurrently being monitored.

One solution, in which the fibre itselfis the laser, is being studied by Lite-cycles Inc. (Murray et al., 1998). This isstill in the feasibility phase and it has

12

Figure 9: Fibre laser layout. The pump laser is a diode-pumped dou-ble-clad Yb-doped fibre that outputs 1.1137 µm. The Raman is a high-delta germanosilicate single-mode fibre, which is directly coupled tothe pump fibre. Feedback for both the pump and Raman fibres are pro-vided by FBGs. The second harmonic of the 1.178 µm first Stokes isgenerated within the fibre Raman laser cavity, and is ejected by a dichroicmirror. (Courtesy J. Murray, LiteCycles Inc.)

Figure 10: The launch telescope is a 25X beam expander. It is mounted behind the M2 unit,together with its accessory beam monitors: power, alignment, polarization meters, and beampointing/focussing servos. Two cameras are present, one to monitor the LGS in the field, andone for air-traffic surveillance.

Yb: Double-Clad FibreGermanosilicate

Fibre

Diode

*TEM00 is the fundamental Transversal Electro-magnetic Mode of the resonator.

13

recently attracted the interest of theGemini AO project as well. The con-cept is shown in Figure 9.

Two 100W CW laser diode-bar arraysare used, efficiently coupled in a double-clad Yb-doped fibre which outputs >50%of the input power at 1.1137 µm from thesingle-mode fibre core. The Yb-doped fi-bre is fusion-spliced with a single-modegermanosilicate fibre 1 km long. TheStimulated Raman Scattering (SRS) ef-fect in this long fibre is used to stretch thewavelength and tune it to 1.178 µm, i.e.twice the 0.589 µm of the D2 sodium line,by the use of fibre Bragg gratings (FBG)written on the fibre and servo-tuned bypiezos. The fibre output beam is focusedon a second harmonic generation crys-tal (SHG) for intracavity doubling, whichwill convert the 1.178 µm photons to0.589 µm. The longitudinal mode spac-ing of such a long fibre Raman laseris 100 kHz, giving a truly broad-band ex-citation of the mesospheric sodium. Thegoal for the line FWHM is ≤ 2 GHz,and the CW output power should be inexcess of 10W CW, with a goal of 30W.However, an unpolarised sodium 2 GHzlaser line would return about half the fluxof a 20 MHz FWHM line. A 10W broad-band laser would therefore be equiva-lent to 5W of ‘narrow-band’ power. Per-ceived risks of this technology are thelarger than expected linewidth, excessivenon-linearity (Stimulated Brillouin Scat-tering) losses and a lower than expect-ed efficiency of the intracavity SHG crys-tal. We are taking advantage of our mod-els of the non-linear effects, developedfor the fiber relay studies, to further in-vestigate this laser design.

The great advantages of this solutionare that the single-mode fibre outputplus frequency doubling crystal give ahigh beam quality, the long fibre length(1 km!) allows the laser source to beplaced far away from the telescope, ina VME-rack sized box. If this laserwould be commercially available, itwould be the ideal solution.

The second laser choice under in-vestigation is a recent development atMPE in Garching. It is a two-stage laserwith a seeded dye-laser amplifier whichwould deliver at least 10 W CW with 500MHz line width. This laser would give anequivalent LGS magnitude mV = 8.

8. Launch T elescope

The input interface to the launch tel-escope (LT) is a collimated beam 20-mm in diameter. The output is a diffrac-tion limited spot at 90 km altitude withthe possibility to focus it between 70and 200 km. The LT optical qualityspecification is to have a Strehl ratio>0.95 within 10 arcsec of field radius,and with 80% encircled energy diame-ter of 0.7" within 1 arcminute field ra-dius. This is required to ease the align-ment tolerances and to have a CCDmonitoring camera looking at the sur-rounding sky field (notching out the

LGS). The launchtelescope is thus a25X beam ex-pander. We haveexplored a reflec-tive solution basedon a two-parabolic-mirror stigmatic de-sign and verifiedthe feasibility of itsopto-mechanicaltolerances. Light-weight mirrors canbe used; the totalweight of the LTand its structure isestimated to be80 kg.

A wedged, anti-reflection coatedexit window pro-vides shielding fromturbulence and dust,and a 10–3 laserflux return, to mon-itor the exit beamquality and power.We will also moni-tor the beam polar-isation state and itsfrequency spec-trum. The volumefrom the fibre endto the exit windowis airtight.

The space avail-able behind the UT3 secondary mirroris limited and we are forced to expandthe beam by a factor 25 within 600 mmof design space. The current design isshown in Figure 10. It delivers a verygood performance with a central obscu-ration of 0.7% in area for a 2-ar-cminute-diameter field (see Table 1).However it has tight tolerances and re-quires an athermalised design: its fea-sibility has been checked. We will alsohave to introduce alignment diagnos-tics. The pointing has to be remotelycontrolled within a minimum range of 1arcmin radius. Fast steering is also nec-essary to compensate the LGS jitter in-duced by the atmosphere on the up-ward-going 50 cm beam which differsfrom the downward path toward the col-lecting 8-m telescope. The expectedrms jitter is 0.34 arcsec for the medianParanal atmosphere and therefore afast steering range of ± 2-arcsec isspecified. The steering signal is provid-ed by the AO high order WFS that islooking at the LGS. The bandwidth ofthe steering device has to be betterthan 200Hz at –3dB.

The pointing and steering would bedone at the level of the fibre output. Therange of fibre motion is ± 0.79 µm forslow pointing with 0.1 µm precision, and± 12 µm for fast steering. Commercialdevices are available which meet thesespecifications together with their controlelectronics.

We plan to enclose the launch tele-scope and its liquid-cooled electronicsin an aerodynamic wind-shield, behindM2 unit. The static and dynamic impacton M2 structural stability have beenstudied and shown to be negligible.

Conclusion

The conceptual design of the LGSFhas been completed and the main in-terfaces with UT3 as well as the envi-ronmental and structural impacts havebeen assessed. Further optimisationwill be carried out in the next phases.Laser beam relay system designs usingboth conventional optics and mono-mode fibre have been explored. Thislatter solution has many advantages

Table 1: Launch T elescope Design – Optical Performance

Design Requirements / Value Achieved

Field 10 arcsec radius 1 arcmin radiusStrehl Ratio at 589 nm >0.95 / 0.99 -- / 0.97580% Encircled energy diam. (“) -- / 0.45 0.7 / 0.5

Figure 11: View of the laboratory setup where the fibre experimentshave been carried out. The setup allows the measurement of the spec-trum, power and propagation properties of the laser beam. The in-jection, fibre output and SBS return beams can also be characterised.

14

and we now feel confident that this so-lution can be adopted.

We have however to finish thetests of the final fibre-relay modules inthe laboratory. A burn-in test is beingdiscussed with MPE, to be performedwith the 4.5W CW sodium laser atCalar Alto. After these tests have beensatisfactorily completed, the detaileddesign activities for the LGSF can beresumed by the second half of nextyear, pending approval by the ESOmanagement.

References

Ageorges N., Delplancke F., Hubin N., Red-fern M. and Davies R.: “Monitoring of laserguide star and light pollution”, SPIE Pro-ceedings 3763, in press, 1999.

Avicola et al.: ‘Sodium Layer guide-star Ex-perimental Results’, JOSA A, Vol.11,825–831, 1994.

Bonaccini, D: ‘Laser Guide Star Adaptive Op-tics Performance Analysis’, ESO TechnicalReport VLT-TRE-ESO-11630-1202, 1996.

Christou, J.C., Bonaccini, D., Ageorges, N. andMarchis, F: ‘Myopic Deconvolution of Adap-tive Optics Images’, The Messenger No. 97,14, Sept. 1999.

Conan, J.M., Fusco, T., Mugnier, L.M., Kersale,E. and Michau, V: ‘Deconvolution of Adap-

tive Optics Images with imprecise knowl-edge of the point spread function: resultson astronomical objects’, in ESO/OSA Top-ical meeting on Astronomy with AdaptiveOptics: Present Results and Future Pro-grams, p. 121, D.Bonaccini ed., 1998.

Davies,R, Hackenberg, W., Eckart, A., Ott, T.,Butler, D., and Casper, M: ‘The ALFA LaserGuide Star operation and results, acceptedfor publication in Experimental Astronomy.

Delplancke F., Ageorges N., Hubin N., O’Sul-livan C.: “LGS light pollution investigationin Calar ALto” Proceedings of the ESO/OSAtopical meeting on Astronomy with AdaptiveOptics, Present Results and Future Pro-grams – Sonthofen, September 1998.

Di Serego Alighieri, S., Bonaccini, D., Oliva,E., Piotto, G., Ragazzoni, R. and Richichi,A: ‘Adaptive Optics for the TelescopioNazionale Galileo’, Technical Report No. 41,Astronomical Observatory of Padua-Asia-go, 1995.

Fried, D.L. and Belsher, J.F.: ‘Analysis of fun-damental limits to artificial guide star adap-tive optics system performance forastronomical imaging’, JOSA A, Vol. 11, p.277, 1994.

Le Louarn, M., Foy, R., Hubin, N. and Tallon,M.: ‘Laser Guide Star for 3.6 and 8mtelescopes: performance and astrophysi-cal implications’, MNRAS 295, 756,1998.

Le Louarn, M, Hubin, N., Foy, R. and Tallon,M: ‘Sky coverage and PSF shape with LGS-

AO on 8m telescopes’, SPIE ProceedingsVol. 3353, 364, 1998.

Milonni, P. W. and Fugate, R.Q.: ‘Analysis ofmeasured photon returns from SodiumGuide Stars’, in Proceedings of the ESOWorkshop on Laser Technology for LaserGuide Star Adaptive Optics Astronomy, N.Hubin ed., p. 77, 1997.

Milonni, P.W., Fugate R.Q. and Telle, J.M.:‘Analysis of Measured Photon Returns fromSodium Beacons’, JOSA A, Vol. 15, p.217–233, 1998.

Murray, J.T., Roberts, W.T., Austin, L. andBonaccini, D.: ‘Fiber Raman laser forsodium guide star’, in SPIE Proc. Vol. 3353,p. 330, 1998.

Tyler, G.A.: ‘Rapid Evaluation of d0: the ef-fective diameter of a laser guide star adap-tive optics system’, JOSAA, 11, 325, 1994.

Viard E., Delplancke F., Hubin N. andAgeorges N.: “LGS Na spot elongationand Rayleigh scattering effects on Shack-Hartmann wavefront sensor perform-ances”, SPIE Proceedings 3762, in press,1999.

Viard E., Delplancke F., Hubin N., AgeorgesN. and Davies R.: “Rayleigh Scattering andlaser spot elongation problems at ALFA”, ac-cepted for publication in Experimental As-tronomy.

VLT Laser Guide Star Facility Subsystems DesignPart I: Fibre Relay Module

W. HACKENBERG, D. BONACCINI, G. AVILA, ESO

The advantages of the LGSF fibre re-lay approach are twofold: (1) we avoid thecumbersome optomechanical relay, and(2) transfer a diffraction limited beam. Theopto-mechanical relay system would re-quire a sealed tube for the laser beam op-tical path to avoid turbulence and dustcontamination, with air-tight sliding jointsat the elevation axis, two servo-controlledsteering mirrors to keep the optical align-ment and several safety interlocks if thebeam has lost alignment. Point (2) is veryimportant, as much effort needs to bespent to ensure a good beam quality. As-suming the LGS is an extended object,the adaptive optics wavefront sensor er-ror is proportional to the area of the LGSdivided by its flux. As it is both costly anddifficult to increase the laser power, allpossible efforts have to be taken to min-imise the LGS spot size, i.e. the laserbeam quality. Some laser relay systemsin other LGS projects are even consid-ering incorporating active or low orderadaptive optics in the laser beam relayoptical path, to ensure a good outgoingbeam quality.

Incidentally, injection of high laserpower in a single-mode fibre is veryuseful in long distance fibre communi-

cation systems to save repeaters. Infact, much work has been done and lit-erature is available on the subject(Tsubokawa et al., 1986; Cotter, 1982).However the successful injection ofseveral watts of continuous wave (CW)visible laser power has not beendemonstrated for long fibres. As we areconcerned with a fibre length of 30 mfor the VLT LGSF system, we havebeen able to achieve this in the labora-tory and have also learned a few tricksin the process. A fibre relay (Figure 1)provides a very flexible solution andsignificantly simplifies the integration. Itwould also be readily applicable to mul-tiple LGS systems needed for futurehigh-order AO systems on large tele-scopes.

The potential causes of power lossesin the fibre are:

1. Coupling losses from the laser beamto the fibre core. These include Fresnelreflections and laser to waveguide modemismatches..

2. Extinction losses in the bulk mate-rial. In a 30 m long low-loss, pure fusedsilica fibre, the bulk losses from scatter-ing and absorption are expected to beabout 8%.

3. Scattering at the fibre core-claddinginterfaces.

4. Mode conversion from low-losstrapped modes to high-loss claddingmodes.

5. Stimulated Raman Scattering(SRS).

6. Stimulated Brillouin Scattering(SBS).

7. Four-wave mixing between forwardand backward Stokes modes.

Points 1 and 6 are the most importantcauses of losses in our case.

The basic requirements of our fibre re-lay system (Fig. 1) are:

(i) stable throughput above 70 % over30 m for 589-nm CW laser light in the mul-ti-Watt regime

(ii) diffraction limited output for thesmallest possible LGS spot size

(iii) polarisation preserving. The latter allows the output of sever-

al fibres to be combined, and to circu-larly polarise the light for the opticalpumping of the mesospheric sodiumatoms. In order to achieve these re-quirements, three fundamental prob-lems have to be solved:

1. Efficient coupling of a free-spacelaser beam into a small-frame optical fi-

E-mail: dbonacci@eso.org

bre. This demands an accuracy and longterm stability of the order of a fraction ofa micron.

2. Handling of high power densities atair-glass interfaces. The power densityat the fibre core interface is in the orderof several tens of MW/cm2. This is abovethe damage threshold of high power anti-reflection coatings (0.5 MW/cm2). Acareful thermal management in the fibreconnectors is therefore crucial.

3. Nonlinear effects. At high power lev-els the scattering processes becomestimulated and influence both effectivethroughput efficiency and spectral puri-ty of the transmitted light.

Three simple measures have been tak-en to overcome these problems: We have(i) maximised the area of the illuminatedair-glass surface, (ii) maximised the ef-fective cross-sectional fibre mode areaand (iii) maximised the effective laser linewidth.

The limits are set by: (i) the numeri-cal aperture of the fibre, (ii) the require-ment of having single-mode output, (iii)the finite bandwidth of the mesosphericsodium excitation process.

Design of the Single-Mode Fibreand Input Coupling Unit

The input laser beam at the fibre-waveguide is assumed to be in TEM00mode, i. e. the lowest-order resonatormode. Gaussian beam propagation hasto be considered in the design of the cou-pling optics and their tolerances, so thatthe mode fields of the input beam and ofthe optical waveguide optimally match.If we consider a fibre, which is single-mode over its entire length for a givenwavelength, the field distribution of theguided fundamental mode can be ap-proximated by a Gaussian. That modefield diameter increases not only with in-creasing core radius but also with de-creasing numerical aperture. The latteris a measure of the light acceptance an-gle of a fibre. Restricting the waveguidelosses due to bending and micro-bend-ing to no more than 1% (minimum bendradius 10 cm) results in a minimum nu-

merical aperture NA = 0.08. For opera-tion as close as possible at the single-mode cut-off wavelength, the maximumfibre core diameter is 5.5 µm for light at589 nm.

The linear loss coefficient of fibre witha core of pure synthetic fused silica is0.003 m–1 at 589 nm. With antireflectioncoatings on the fibre ends the theo-retical maximum throughput of a 30-m fi-bre becomes 90%. Since the damagethreshold of high-power antireflectioncoatings is around 500 kWcm–2 in the vis-ible, the air-glass surface has to be en-larged at the fibre ends. For this we usea fused silica window, either in opticalcontact with the fibre or fusion spliced(Fig. 2). To avoid Fabry-Perot effects be-tween the fibre ends, the output surfacehas to be wedged.

The state of polarisation at the exit ofa polarisation-preserving fibre is stableand linear if the input state of polarisa-

tion is stable, linear and aligned to thebirefringence axis.

For focusing the free-space laser beamonto the fibre we use an aspheric lens,mounted and pre-aligned to the fibre in-put inside an athermalised housing. Wefirst carried out an analytical optical de-sign, and then verified numerically thecoupling efficiency using CODE V. Theexperiments reported in the last sectionof this article confirm the design model.Figure 3 shows the dependence of cou-pling efficiency on decentre between theoptical axis and the fibre axis, for variousangles of incidence. For optimum cou-pling efficiency, the incoming laser beamhas to be shaped so that the waist dia-meter of the focal spot produced by theexternal lens exactly matches the fibremode field diameter. In our case this is15 % larger than the nominal core di-ameter.

Aligning the coupling unit with respectto the incoming beam requires three de-grees of freedom. A power and polari-sation sensor analysing the fibre outputwill provide the error signal for a piezo-based servo-loop stabilisation of the op-timum injection.

In order to prevent the single-modecontaminating cladding modes (i. e. high-er order modes that are not guided by thecore), as well as for an optimal heat re-moval, the fibre connectors have to befabricated out of glass that matches thefibre core index. A protective and flexiblestainless steel sheath is required over theentire length of the fibre.

Critical Fibre Input PowerDensity

The most important nonlinear lossprocess in the fibre is stimulated Bril-

15

Figure 1: LGSF fibre relay subsystem layout. The fibre input is on the laser optical table, theoutput is at the launch telescope. The power and polarisation sensors at the fibre end are usedto drive the alignment actuators at the fibre input.

Figure 2: The glass-air interface area of the fibre is enlarged using a window of the same ma-terial of the fibre core. The window is either spliced to the fibre, or optically contacted. Bothfibre ends have a fibre window.

louin scattering (SBS). This is the scat-tering of light by sound waves that arecreated in the fibre by the incidentlight due to electrostiction1 in a phonon-

photon interaction process. The stand-ing acoustic waves create a regulardensity grating along the fibre. As aconsequence of the selection rules inthe waveguide, only scattering in thebackward direction occurs in the fi-bre due to SBS.

Energy conservation demands that theamplified backward light (the Stokeswave) is shifted downwards in frequen-cy with respect to the incoming (pump)light, by an amount equal to the acousticfrequency of the material excited. TheStokes line frequency shift is in our caseabout 30 GHz, and the spectral (or Bril-louin) width of the backscatterd light is ∆νB= 110 MHz FWHM.

The effect of SBS is that a fraction ofthe light injected in the fibre is sentbackwards, shifted in frequency andwith a characteristic linewidth of ∆νB.SBS is a non-linear effect and its onsetis abrupt, once the pump laser powerexceeds a certain threshold. Thisthreshold depends on the fibre materialand the frequency distribution of thepump. Solving the differential equationsfor the Stokes and pump intensities inthe steady regime, we obtain a criticalpump power2 of 720 mW. This is forsingle-frequency, narrow-band sodiumlight injected into a 30-m single-modefused silica fibre whose parameters areoptimised for minimum power density inthe core. In an equivalent polarisation-maintaining fibre the SBS thresholdpower is reduced by a factor of two dueto polarisation scrambling.

As long as the coherence length of thepump light is much smaller than the fibrelength, the frequency-dependent SBSgain coefficient is given by the convolu-tion of the pump spectral profile and theSBS gain profile. This is a key point. Asit means the critical pump power can beincreased, if the effective laser linewidthis broadened, for example by means ofphase or frequency modulation of thelaser beam.

If we assume a narrow-band single-frequency CW laser source whose ef-fective linewidth is broadened external-ly by a simple phase modulationscheme (see below), then the pumpspectrum will consist of several longitu-dinal modes equally separated by a fre-quency ∆νm corresponding to the mod-ulation bandwidth. The modes of theStokes wave will be also equidistant,separated by the same distance ∆νm. Ifagain the pump coherence length ismuch smaller than the characteristicSBS interaction length, i.e. for fibresmore than a few metres long, each ofthe Stokes modes interact with all of thepump modes, and vice versa. Since theinteraction strength is determined bythe de-tuning from resonance, the inter-action bandwidth is limited to theBrillouin linewidth ∆νB. In other words, if∆νm >> ∆νB then the SBS source polar-isations generated by different pumpmodes are not phase matched and donot interfere coherently, as each pumpmode generates its own Stoke wave. Inthis situation and with equal powerpump modes, the resulting SBS gain is

16

Figure 3: Coupling efficiency for the designed optics. The coupling efficiency is plotted againstdecentring error between the laser beam and the waveguide axis. Several curves for differentbeam incident angles are shown.

Figure 4: Increase of the SBS threshold pump power, ρeff–1

, under multiline excitation, normalisedto the single-mode threshold, as a function of the frequency modulation bandwidth to Brillouinlinewidth ratio ∆νm / ∆νB, and the achievable peak phase shift φm. It has been assumed thatthe pump coherence length is much smaller than the fibre length, Lf, and that the frequencymodulation is achieved with a single transverse electro-optical modulator.

1The type of piezoelectric effect where the inducedmaterial deformation depends on the square of theelectrical field amplitude is called electrostiction.

2Here defined as that input power for which thetotal SBS power equals 1% of the pump power atthe fibre entrance.

reduced by a factor equal to the num-ber of pump modes. This reduction ofthe SBS gain in case of multimodeexcitation can be summarised in afactor ρeff, averaged over all modesand normalised to the SBS gain for asingle pump mode with the same totalintensity and intrinsic linewidth. In gen-eral, ρeff will also depend on the rela-tive power distribution within the pumpmodes. For a frequency modulationscheme with a transverse electro-opticlight modulator (see below), the rel-ative strength of pump mode j is givenby the square of the Bessel function oforder j, evaluated at the peak phaseshift φm which can be achieved withthe modulator. The inverse of ρeff (orthe effective increase in the critical in-put power) is shown in Figure 4 as afunction of the externally applied peakphase shift and the frequency modula-tion bandwidth to Brillouin linewidthratio. It can be seen that at φm = 1.5 rad,2.7 rad, 3.8 rad (etc.) local maxima inρeff

–1 appear. Those extremes corre-spond to the most equal power dis-tribution within the generated pumpmodes.

Mesospheric Sodium ExcitationEfficiency

The effective backscatter cross sec-tion of the mesospheric sodium atomsis given by the convolution of the spec-tral illumination profile and the sodiumabsorption profile. We want to max-imise it.

With a fixed LGS spot size and in-creasing layer illumination, saturationof the atoms can be avoided by broad-ening the effective laser linewidth.Since the width of the Doppler-broad-ened sodium D2 absorption profile isabout 2 GHz, there is an optimum ef-fective laser linewidth, which optimisesthe LGS photon return. In Figure 5 thepredicted LGS brightness is shown as afunction of the phase modulation pa-rameters, in case of a beam relay withan optimised polarisation preserving

30-m-fibre that is single-mode over itsentire length.

The throughput of the laser launch-ing optics has been assumed to be90%. The atmospheric parameters(seeing, transparency, sodium columndensity, layer height and temperature)assumed for Figure 5 represent medianParanal values, with the telescopepointing the LGS to zenith. As can beseen, the maximum brightness occursat modulation bandwidth to Brillouinlinewidth ratios in the range between 1and 2. For smaller ratios the lower criti-

cal input power limits the brightness, in-dependently of the peak phase shift.For larger bandwidth ratios, the de-crease in the overlap integral betweenthe excitation spectral profile and theDoppler-broadened absorption profilelimits the LGS magnitude. The bright-ness gain starts to saturate for optimumpeak phase shift values above 2.7 rad.This, together with a minimisation of theelectrical power requirements for thelaser light modulation, implies an over-all optimum peak phase shift of φm = 3.8rad. The corresponding optimum mod-ulation to SBS bandwidth ratio is ∆νm /∆νB = 1.0. With these parameters themaximum LGS brightness will be limit-ed to 9.5 equivalent V-band magnitudein case of unpolarised excitation.

Experimental Results

Fibre-transmission experiments havebeen carried out with the ALFA sodiumlaser on Calar Alto in 1998. The resultsof the SBS threshold achieved agree wellwith theory. In Garching, a prototype in-jection setup has been built with a 532nm, 5W CW all solid-state laser. Thescheme is shown in Figure 6. HPL is thehigh power laser. For the laboratory ex-periment this is a diode-pumped Nd:YV04laser (Coherent Verdi) capable of emit-ting 5.5 W of single-frequency CW pow-er at 532 nm with high beam quality. Tosuppress any back-reflections into the

17

Figure 5: Predicted LGS brightness mV in case of laser light relay with an optimised single mode,30-m-long, polarisation-preserving, pure fused silica fibre. The LGS brightness is computedas a function of the frequency modulation bandwidth to Brillouin linewidth ratio, ∆νm / ∆νB, andpeak phase shift φm, achieved with a single electro-optic modulator. The parameters for theatmosphere are chosen to represent median values. For the transmission of the launching op-tics 90 % was assumed with the LGS at zenith.

Figure 6: Layout of the laboratory set-up for the results reported in this article. See text for ex-planation of the acronyms.

laser head a Faraday Isolator (FI) is in-stalled near the laser output coupler. M1and M2 are folding mirrors for controllingthe four degrees of freedom of the laserbeam. PH1-PH4 are pinholes for align-ment purposes (see below). For the samereason an attenuator (A) is installed in thebeam. Its working principle is based onFresnel reflection at counter-movingglass plates, which ensures minimumbeam position deviation during operation.In combination with the position an-gle of the exit polariser of FI, any at-tenuation between 0 dB (without in-sertion losses) and 27 dB can be se-lected. BC and BE form the beam shap-ing optics. These are 2-element 4xbeam compressor and expander, re-spectively. The purpose of these two tel-escopes is twofold: First, BC reduces thebeam size so that the beam can passthrough the electro-optical light modula-tor (EOM) without clipping. The apertureof the electro-optical crystal is minimisedin order to relax the electrical powerrequirements to drive it. Second, by slight-ly decollimating BC and BE, the waist lo-cation and diameter of the beam enter-ing the focusing lens L1 is controlled foroptimum launching, i. e. mode matchingwith the fibre waveguide.

Several diagnostics tools are inte-grated for on-line analysis of the launched

beam, the back-reflected light from thefibre entrance and output. These allow themeasurement of the light spectrum (SA),polarisation (POL), power (photodiodesPD 1–4, bolometer B), beam quality(BPA) and jitter as well as waist locationand size (beam propagation analyserBPA, position sensitive device PSD). Thediagnostics tools are fed with the help ofwindows (W1–3), beamsplitters (BS) andflipper mirrors (F1–8). Metallic coatedneutral density filters (NDF) are used toadjust light levels. All high power lensesand windows including the phase mod-ulator crystal are antireflection coated.The high power mirrors are dielectriccoated. To reduce scattering and beamdegradation, all high-power surfacesand the elements in the arm of the beam-propagation analyser have a surfacequality of better than 60 nm PTV and arepolished down to 10–5 scratch-dig. Thethroughput of the high power beam pathis 80 %. HeNe and DPSSL are low pow-er lasers for alignment purposes.

To allow the optics in the high-powerarm to thermalise and also for safety rea-sons the coarse-alignment of the highpower laser beam is started first with anattenuated beam. Figure 7 summarisesthe results on the high power transmis-sion of a 5-m-long single-mode test fibre.The conclusion from this first experiment

has been that it is possible to achievean optimum launching and efficient SBSsuppression for the test fibre. More-over, our theoretical calculation andexperimental results are in good agree-ment, which gives us confidence in thereliability of the model of the system. Withthe current HF amplifier driving theelectro-optic light modulator, a maximumpeak phase shift of φm = 2.1 rad at ∆νm= 110 MHz can be achieved. At 532 nmthis corresponds to a SBS gain reductionfactor ρeff = 0.41. In Figure 7 the fibreoutput and backscattered power areshown as a function of the input pow-er on a 5-m-long, non-polarisation-pre-serving single-mode fibre. Below 2.4 Winput power only Fresnel reflection isobserved in the backscattered light. Thisagrees with the calculated critical SBSpump power of 2.5 W. The slight dis-crepancy between the former values canbe explained by the tolerances in the fi-bre parameters. For input powers high-er than 2.5 W and with the laser fre-quency un-modulated, the SBS back-re-flected power increases to 29% of thepump power above the Fresnel reflection.In the spectrum of the back-reflected lightthe Brillouin-shifted input started to ap-pear at the critical input power. With themodulator switched on, no SBS was ob-served up to the maximum achievable in-put power level. This agrees with the cal-culated decrease of the SBS gain. Themodulated throughput at the maximuminput level of 4.2 W is 85 %. For com-parison the throughput of a low powerHeNe-laser is 88 %, measured with anoptical setup optimised for the He-Nelaser, which had a mode quality compa-rable to the Verdi laser. The difference canbe explained by the larger beam jitter ofthe Verdi laser, which should account fornearly 3% coupling loss due to concen-tricity errors. The maximum theoreticalthroughput is about 90% including Fres-nel reflection at the uncoated fibre endsurfaces (8% loss) and extinction (linearloss coefficient 0.005 m–1 at 532 nm) butneglecting all other loss sources. Themaximum fibre output power was stableduring the experiment. A burn-in test withlonger fibres is scheduled.

Conclusions

We have demonstrated that a fibre re-lay module is feasible. The experimen-tal results agree well with the theoreticalmodel. Further work on an optimised,end-face-protected, 30-m-long single-mode fibre is in progress and the first ex-periments will be carried out in Garching.A burn-in trial on Calar Alto sodium laseris foreseen in the first half of next year.

There appear to be at least two pos-sibilities for further increasing the LGSbrightness within a fibre-fed relay system:

1. To combine the polarised output oftwo fibres at the launch telescope. Thebrightness limit would increase by 0.7 mag.

2. To separate the two purposes of thefibre relay, namely flexible transportation

18

Figure 7: Measured fibre output power Po and backscattered power Pb as a function of the in-put power Pi for a 5-m-long pure fused silica non-polarisation preserving single-mode fibre at532 nm. The cases of unmodulated and frequency modulated input are shown (light modula-tor EOM on/off). For comparison, the extrapolated low power HeNe-laser transmission and themaximum theoretical throughput is displayed. See text for further details on the setup.

of polarised high power light and dif-fraction-limited output, into two physicalfibre sections: a longer multimode sec-tion and a shorter mode cleaning sec-tion.

The disadvantage of the first solutionis the increase in complexity, however weknow it is feasible. The advantage of atwo-sectioned fibre is that its designwould be no longer limited by the criticalinput power of a single-mode fibre ofequivalent length. Instead higher powerlevels could be considered. For a sodi-um laser with 10 W output power, whosespectrum is optimised for maximum LGS

return flux, a hybrid fibre is under study.Assuming the same launching and at-mospheric parameters as for Figure 5,without optical pumping of the sodiumatoms via circular polarisation, this lasersetup should be able to provide a see-ing-limited LGS of brightness V=8 mag-nitude. This would be sufficient for adap-tive optics correction even under poorseeing conditions, pointing at 60 degZenith angle.

ReferencesCotter, D.: ‘Suppression of Stimulated Brillouin

Scattering during Transmission of High-

power narrow-band laser light in monomodefibre’, Electronics Letters, Vol. 18, No. 15,p. 638–640, 1982.

Smith, R.G.: ‘Optical Power Handling Capacityof Low Loss Optical Fibres Determined byStimulated Raman and Brillouin Scattering’,Applied Optics, Vol. 11, p. 2489, 1992.

Tsubokawa, M, Seikai, S., Nakashima, T., Shi-bata, N.: ‘Suppression of Stimulated Bril-louin Scattering in a Single Mode fibreby an Acousto-Optic Modulation’, Elec-tronics Letters, Vo. 22, No. 9, p. 473–475,1986.

19

New Pictures from the VL T

Spiral Galaxy Messier 83: This photo shows the central region of a beautiful spiral galaxy, Messier 83, as observed with the FORS1 instrumentat VLT ANTU. It is based on a composite of three images, all of which are now available from the ESO Science Data Archive. The three frameswere taken in March 1999 through three different filters: B (wavelength 429 nm; Full-Width-Half-Maximum (FWHM) 88 nm; exposure time 10min; here rendered as blue), R (657 nm; 150 nm; 3 min; green) and I (768 nm; 138 nm; 3 min; red) during a period of 0.8 arcsec average see-ing. The field shown measures about 6.8 x 6.8 arcmin and the images were recorded in frames of 2048 x 2048 pixels, each measuring 0.2 arc-sec. North is up; East is left.

E-mail: dbonacci@eso.org

First Images from FORS2 at VLT KUEYEN on Paranal

Top: Three-colour composite of the well-knownCrab Nebula as observed on November 10,1999. It is the remnant of a supernova explosion at adistance of about 6000 light-years, observed forthe first time almost 1000 years ago, in the year1054.

Right: Three-colour composite of the young ob-ject Herbig-Haro 34 (HH-34), now in a protostarstage of evolution. It is based on CCD framesobtained with FORS2 in imaging mode, onNovember 2 and 6, 1999. The object has a remarkable, very complicatedappearance that includes two opposite jets (redstraight line, down centre) that ram into the sur-rounding interstellar matter. This structure is pro-duced by a machine-gun-like blast of “bullets” ofdense gas ejected from the star at high velocities(approaching 250 km/sec). This seems toindicate that the star experiences episodic “out-bursts” when large chunks of material fall onto itfrom a surrounding disk.Note also the enigmatic yellow arc to the left, afeature that is still unexplained.

21

In the June 1998 issue of The Mes-senger (No. 92) there was an article de-scribing the hectic days and months be-fore the first light for UT1. First light wasa great public event and of great signif-icance for the project and the organisa-tion. From a technical point of view how-ever, it was a non-event. The telescopehad met all specifications for first lightahead of time and the weeks before theofficial announcement were more a sittight, don’t tell and don’t break it time. Fol-lowing the big event we were free againto work on the machine. For those withkeen eyes looking back at the beautifulimages taken for the first light, flaws couldeasily be found. In fact a number of mes-sages and private comments arrived atParanal describing in great detail what wealready knew. A lot of work lay ahead ofus.

The working schedule was set out.During the day Peter Gray, Toomas Erm,Gerd Hudepohl, Marc Sbaihi and JuanOsorio would fix all the things we brokeduring the night and continuously im-proved the opto-mechanical and elec-tronic robustness of the system. Duringthe night Jason would sleep on his sofawhile the others tested out the latest and

newest ideas that had been implement-ed the day before. Pretty pictures, whichbefore first light were obtained when theseeing was better than 0.6 arcseconds,were banished to be taken at times whenthe seeing was better than 0.5 arcsec-onds. Then the limit was changed to 0.4arcseconds as seeing was good too of-ten and test time was being limited. It wasoften said that the seeing was best whenJason was asleep in the control room.Ivan Muñoz and Anders Wallander set outto prove this by obtaining a 0.27 arc-second image in June under those op-erating conditions. A press release on thesubject – politely avoiding the sleepingissue – can be found on the web.

Antu, or UT1 as it was still known inthose early days, suffered from windshake. The telescope was not stiffenough to remove the effects of the windon the focal plane. This was not a sur-prise since from the design phase alreadyit was known that the UTs would be sus-ceptible to such problems. Before firstlight Gianlucca Chiozzi and Robert Kar-ban had done a splendid job in gettingfield stabilisation to work in a matter of acouple of weeks. Now was the time totune it up, remove the lag as much as

possible and get every last millisecondout of the loop. Working with a 33 MHz68040 processor as the central enginewas a challenge. However, together withBirger Gustafsson, Antonio Longinotti andPhilippe Duhoux (M2 and CCD experts)the field stabilisation was brought to aneffective update frequency of 20 Hz. Nowwas the chance to have some real fun.Out of the drawer Birger produced the lat-est and best toy. A Power PC processorfor our guiding LCU. It was tested on thecontrol model in Garching, deemed to begood and hey presto! 50 Hz frequencyon the M2. The secondary mirror of theVLT is 1.2 metres across. Moving it at 50Hz is not for the weak of heart. If you gottime on the VLT and you have got somedata, the 'little' mirror at the top of the tel-escope was moving about all the timemaking the focal plane stay put while thebig thing (telescope) shook in the wind.The weakness of the telescope in wind-shake rejection was thought to be with usfor life. Toomas Erm would not accept thatand continued working on this problem.

Before first light we only needed thetelescope to point to 10 arcseconds or so.Even for a big machine like UT1 this wasnot very difficult to achieve thanks to thefull integration of the pointing model intothe tracking software. Now we had thereal specifications to meet. Pointing anytelescope to 1 arcsecond rms all over thesky is hard enough. Doing it with the UTsof the VLT is an interesting challenge.First we had to stabilise the pointing mod-el which was varying from night to night.It does not take a rocket scientist to workout that if the primary mirror moves in-side the cell then the pointing errors ofthe telescope change. Keeping the mir-ror in position was provided for with themirror cell hardware, which permits us toposition the mirror in x,y,z as well as thethree angles about these axes. By re-po-sitioning the mirror after each pre-set westabilised the pointing model. A new set-up keyword for the telescope was need-ed and the food theme on Paranal began.When TEL.PSOUP (pronounced peasoup) is set to true the m1 cell PassiveSuppOrt moves the mirror UP. Now wehad a stable pointing solution and wecould start worrying about the actual per-formance of the telescope. Trying as hardas we might, neither the residual of thefit nor the actual pointing performancecould be brought below 1.5 or so arc-seconds rms, and often the performancewas worse. But then again we asked our-selves why are we bothering with this.The telescope guide probe moves auto-matically to find a guide star. Why not off-set the telescope position to make theguide star land where the guide probe ex-pected it to be. The pre-setting accura-cy would then be equivalent to the as-

Commissioning of the Unit T elescopes of the VL TJ. SPYROMILIO, A. WALLANDER AND M. TARENGHI, ESO

Figure 1: The plot shows the rms tracking error as measured on the telescope encoders forAzimuth on Kueyen. The specification is 0.1 arcsecond and the achieved performance is clear-ly much better than that. The direction of the arrow indicates the direction the telescope wasmoving during the measurements.

22

trometric accuracy of the guide star. Twobirds killed with one stone. One was thepointing of the telescope. The other wasthe reproducibility of the pointing. The lat-ter was to show its true worth later whenFORS1 was being commissioned.

Now that some basic things were outof the way we could start having fun.Again Gianlucca Chiozzi and Robert Kar-ban were set to work. We want to chopand field stabilise at the same time. Surethey said. It is part of the specs. Onebeam is always under control while theother is the sky beam that no one caresabout. Somewhere in the specificationsthis had slipped through. Maybe it camefrom the days of single-element IR de-tectors. In any case, the new more com-plex specification was given out. As-suming the guide star is going to land onchip in both beams of the chop then wewish to field stabilise on both beams.Back to the drawing board. PhilippeDuhoux to the rescue with a port of theDIMM software to the guider and againour wish was fulfilled. ISAAC folks wouldbe happy. But what about the comet andmoving-target community? Of course theVLT can track on a moving target. Butcould it guide on it as well? So the chal-lenge was set. How would this be done?Back to the original design to see whatwas thought to be the way. Move the guid-ing box on the CCD with the appropriatespeed and everything will be OK was thetheory. But why do that? Why not tell theguide probe to simply track its star. Thenif the centre moved or not was not a prob-lem since the guide probe would happi-ly hunt down its guide star. It actuallyturned out that Birger Gustafsson had al-ready coded almost all of this simply tobe able to pre-set the probe to the guidestar. All he had to do was to keep theprobe alive at all times. Easier said thandone but, as usual, nothing is more funthan the impossible. So having this work-ing, we could now chop and field stabilisewhile the telescope tracked a moving tar-get. Unfortunately, in doing this webroke the combined offset mechanism.Traditionally the probe is offset in the oth-er direction to the telescope. By makingthe probe live when the telescope offsetthe probe would automatically follow itsown star. After a bit of soul searching werealised that this was actually the betterway of doing things. The autoguider waspromoted above the tracking software asthe master and we were back in business.

While all of this was going on, fromnowhere came a couple of Nasmythadapters and were miraculously mount-ed on the telescope without interferingwith our work (we suspect a phantomFrancis Franza and his twin Paul Gior-dano had something to do with this butMartin Cullum denies all knowledge asto how this miracle occurred). The engi-neering crew was growing by the day.George Harding, the two Pablos (Gutiér-rez and Barriga) and the formidable Patri-cio Ibanez all started putting their stampon things. Toomas Erm kept worrying

about this wind shake thing. Gerd Hude-pohl and Pierre Sangaset would fretabout the cell while Gustavo Rahmer wasfinally about to get some serious work todo. Thanh Phan Duc would come and goand let us all know exactly how many celloperations we had made and would fixwhatever needed fixing. It was like havinga policeman checking up all the time. Ri-cardo Schmutzer moved between Garch-ing and UVES and Paranal and the testcamera without missing a beat on eitherof them. FORS1 arrived on the mountainand immediately aroused suspicions. Didthey really think that big yellow thingwould fit under our delicate telescope?

Meanwhile the final phases of the com-missioning of another telescope onParanal were taking place. The ASM (As-tronomical Site Monitor) was getting to thephase where it could be left alone. Ste-fan Sandrock and Rodrigo Amesticawould try to make our robotic telescopework. Anders expressed wishes that theUTs be so simple to operate. Switch it onand the ASM will know when to open andwhen to close, where to find a star andwhat to measure. Not too hard a re-quirement really. Only the ASM was hav-ing problems finding alpha Cen let aloneany other star. Abit of investigating on thepointing problem and a few ingenious so-

lutions by Stefan and Rodrigo and theASM was up and running. Now the ra-dio need not sound all night long with thecall: “Julio! What is the seeing?”

Before putting FORS1 on to the tele-scope we had to satisfy another client.Bruno Leibundgut together with Mark Fer-rari and Eline Tolstoy arrived for scienceverification with the test camera. Brilliantplans were laid out, astronomical chal-lenges and problems to be solved. Thefirst couple of nights went fine. In fact, anew image quality record was set at 0.26arcseconds, the image taken with theADC mounted. Then the weatherchanged, the seeing became poor, thecontrol room was not a happy place andbefore you could say “bad luck” the twoweeks of SV were over. Bruno took acopy of the data with him and the videoconference to Garching fell silent.

Soon afterwards The FORS1 team ledby Professor Appenzeller together withGero Rupprecht as the ESO responsibleput their pride and joy on the Cassegrainfocus of UT1 and then had to take it offagain due to a broken cooling pipe insidethe adapter which dripped water insidetheir instrument. We apologised, helpedto clean up the mess and on went FORS1again. The moment of excitement arrived.The shutter was opened and the first im-

Figure 2: The top left plot shows the measured centroid for a star during successive 30 sec-ond exposures taken during a 3.5 hour track. The difference in the centroid between the firstimage and all successive images is the measured error in tracking. The three other plots showthe error in tracking as a function of altitude, adapter angle and hour angle. The telescopecrossed the meridian during this test and the peak error was below 0.15 arcsecond duringthe entire experiment.

23

age obtained. The image quality wasquickly derived to be 0.6 arcseconds. Aquiet note of satisfaction was felt acrossthe control room and now the real workcommissioning FORS1 could start inearnest. A separate article on the FORS1commissioning has already been pub-lished, so not much more will be said here.

In the mean time ISAAC was sitting inthe integration lab waiting to be installedon the telescope. Jean-Luis Lizon hadbeen working furiously under difficult con-ditions to complete the re-integration ofthe instrument. In the couple of weeks be-tween FORS1 and ISAAC we checkedout the Nasmyth A focus and verified allwas well. ISAAC went on to the telescopeon schedule and also produced beauti-ful images straight away. Finally we hadan instrument that could exercise the tel-escope in all the weird and wonderfulways that only infrared astronomers canfind. Alan Moorwood, Jean-Gabriel Cubyand Chris Lidman all worked furiously atcharacterising the instrument. Our first fel-low on Paranal, Monica Petr, arrivedaround this time and immediately start-ed learning and driving both instrumentsand telescope. Joar Brynnel finally couldsee all his electronics work used for as-tronomy rather than just tests. ManfredMeyer and Gert Finger got the last ounceof efficiency out their IRACE system toread the ISAAC detectors. Peter Bier-eichel was always at hand to make theinstrument software that little bit better.Nicolas Devillard demonstrated the lat-est pipeline capabilities.

It was Christmas time by now and theFORS1 crew were back. To welcome

them we put water inside their instrumentagain, but by now we were all experts atdismounting and cleaning FORS1 up.Our apologies were gracefully acceptedagain and we started to work together to-wards the final integration of the instru-ment into its new home. Unfortunately theweather did not co-operate and most ofthe second commissioning run was lost.We scheduled another run for Januaryand with the help of Thomas Szeifert,Bernard Muschielok and WolfgangGaessler the instrument was Paranalised.Following this it was handed to Fernan-do Comerón to system test. Of course,Fernando kept on finding things we hadmissed. Roberto Gilmozzi would agreewith him and then we had to changeeverything again. It was fun.

Time had come to welcome ISAACback. Remember how well the telescopeand instrument had behaved in the firstcommissioning run. Well, this time thingswere not to be so easy going. The UTs’direct drive motors (big fancy magnetswith a protective metal cover) move themachine silently and very elegantlyaround the sky. On side A (not the ISAACside) one of the covers came off and gotcrunched inside the motor. PatricioIbanez, Pablo Gutiérrez, Gustavo Rah-mer and Hans Gemperlein worked non-stop for a week to disassemble one wholeside of the telescope and put it back to-gether again. The badge of honour at thetime was a piece of yellow duct tape,which was worn with great pride for theduration of the work. You might havethought that the control room would havebeen a quiet place with a broken tele-

scope. Taking advantage of the absenceof the head of commissioning, MichaelLoose, our architect on Paranal, trans-formed the control room into Beirut at theheight of the Lebanon troubles. The floorand the ceiling disappeared. If a smallthermonuclear weapon had gone off itwould have made a smaller mess. In spiteof this, Jean-Gabriel, Jens Knudstrup anda few other brave souls continued work-ing, wearing surgical masks to limit theamount of dust entering their lungs andthe ear-plugs to reduce the noise level tothat of a firing range.

All was brought together just in time forthe end of the commissioning of ISAACand we were ready for the operations dryruns which were scheduled for March. Inthe mean time two big events were on thehorizon. UT2 (soon to be renamed toKueyen) was getting ready for first lightand the inauguration of Paranal was alsorapidly approaching. Without commis-sioning really noticing, Peter Gray and hiscrew, with some help from the usual sus-pects in Garching – Bruno Gilli, MaxKraus and Fabio Biancat Marchet comeimmediately to mind – had another 8-me-tre telescope ready to go. Cells movingup and down the mountain, mirrors coat-ed, servo loops tuned and all the thingsnecessary to make a telescope readysomehow were all done without slowingUT1 work. Two telescopes running be-fore the inauguration and a nice imageto show from Antu would make a goodshowing. Massimo in a typically chal-lenging order stated that nothing worsethan a third of an arcsecond for the im-age from Antu would satisfy him. On thenight before the inauguration, UT1 hadalready been handed to Science Oper-ations for their dry runs. The telescopewas theirs. What could they do? Just toset the standard for the future, HermannBoehnhardt and Roberto Gilmozzi wereat the controls with Massimo directing thatFORS1 be switched to high-resolutionmode since the seeing was good. It is awell-known fact that the presence of thedirector on Paranal improves the seeing.The combination of all of the effects de-scribed resulted in a 600-second I bandimage with quality better than 0.25 arc-seconds. Now that the first goal was metwe could have some extra fun. LotharNoethe and Stephane Guisard had beenlabouring for some time to make the tel-escope spell its own name. Anyone theysaid could do it moving the telescopearound until the star trails spelt out VLT.Even the NTT managed that. So theytook charge of the shape of the primarymirror and they made stars whoseshapes spelt out VLT! True control of theshape of the primary was demonstratedbeyond a doubt.

On Kueyen as soon as we had a pri-mary and a secondary mirror installed wehad a look. It was not going to be as easyas Antu but it would work. Lothar Noe-the, Stephane Guisard and RobertoAbuter set about their merry way to makethe thing make stars that looked round.

Figure 3: Measured pointing errors on 140 astrometric stars covering the whole sky. The meas-ured rms pointing on this night was 0.85 arcsecond.

24

Krister Wirenstrand sorted out the point-ing with a bit of help from a big piece ofpaper being held at the Nasmyth Focusand using a real bright star. Two tele-scopes working and the best ever image.We now felt that the observatory couldbe inaugurated.

Before the telescope was released intofull operations we threw away the old soft-ware. Giorgio Filippi came down with apresent from Garching. AY2K compliantsystem had been fully tested on theGarching control model and was installedon all systems on Paranal. Carlos Guiraobrought some new data flow computersand Michèle Peron, Nick Kornweibel,Michele Zamparelli, Gerhard Mekiffer,Klaus Banse, Miguel Albrecht, Paola Am-ico, Reinhard Hanuschik, Palle Møller,Michael Rauch, Almudena Prieto, DaveSilva and Maurizio Chavan all spent quitesome sleepless nights on Paranal en-suring the data-flow system was fully op-erational before and immediately afterAntu was released.

Somewhere earlier in this storyToomas Erm was still worrying about thiswind-shake problem. So, after a longthink about it all, he asked if Stefan San-drock could give him a helping hand withan idea he had. A couple of days later anew version of our tachometer was im-plemented in software and the enchila-da was born (a soft taco being an en-chilada). Wind shake is not a thing of thepast but Kueyen is now stiffer than wehad ever dreamed of. Enchilada gaveway to Enchilada turbo and by now theazimuth axis of Kueyen was exceedingtracking specifications by over an orderof magnitude and altitude was in greatshape also exceeding specifications bya factor of 2 under most operating con-ditions. The windscreens of the enclosurehave been improved and tested andretrofitted thanks to the work of MartinaGarcía (who somehow also managed tohave a baby while all of this was hap-pening) and Erich Bugueno. Juan Oso-rio fixed things on the electronics of theenclosure and German Ehrenfeld neverseemed to stop making things and mov-ing big pieces around. Nelson Montanois now our hydrostatic pad man and rou-tine preventative maintenance started inearnest. The benefits of having more thanone telescope to work on were coming

to the fore. In any normal telescope ob-servatory Antu would have been releasedand the improvements coming fromKueyen would have been implementedin some distant future. At Paranal the cy-cle still has some way to run.

On Antu, for each activity undertakenby commissioning, a test procedure wassupposed to be followed documenting ex-actly how each test was to be performed.This was to be a self-documenting sys-tem with the procedure also providing thefinal documentation for the test. In somecases, such as tracking and field-stabil-isation maps, this was done while for oth-ers the procedures were developed dur-ing the commissioning, as we better un-derstood what needed to be done. OnKueyen this documentation effort wasmuch improved and expanded. Also au-tomation of a number of tests was un-dertaken. The use of the engineering datastream was expanded dramatically to pro-vide input for preventive maintenance butalso to better understand the system. TheVLT control-system engineering-datastream includes for example every aber-ration measured by the system. In anygiven month, as many as 20 thousand ac-tive-optics corrections are made. Look-ing at the logs and using parallel meas-urements with the test camera wavefrontsensor, we identified some cross-talk be-tween defocus and spherical aberration.The effect was small resulting in addi-tional noise in the system of order 0.03arcseconds in the average image quali-ty delivered. Lothar was looking at thedata in Garching and promised a solutionby the birth of his son. On schedule onthe 3rd of August the problem was iden-tified and corrected for. A baby boy wasalso born on the same day. The cross-talk was identified as being due to an in-correct sign in the force setting on the pri-mary. The improvement in the stability ofthe defocus was obvious (the rms valueof the aberration was reduced by 500nanometres). The value of the engi-neering data stream cannot be over-stated.

Continuing on the food theme (peasoup, enchilada) we started worryingabout the sausages the telescope wasproducing. Close to the zenith and themeridian, when 430 tons of glass andmetal are spinning around at an impres-

sive pace, the stars were not coming outround but rather long like sausages. Herewas a problem that had to have a solu-tion. So we all worked on it. Kristerchanged the tracking software. Stefanchanged it back and then back again andonce more. Rodrigo and Anders meas-ured in ever more imaginative ways. Inparallel, continuous improvements in thesystem were made, all contributing in theirown ways in attacking this problem andother annoyances. Mario Kiekebusch au-tomated the adapter calibration proce-dures and gave us smarter secondaryguiding. Ivan Muñoz made the pointingmodelling trivial by automating it and in-troduced us to Pinky and the Brain. Fran-cisco Delgado kept improving the activeoptics software to allow us to better con-trol the mirror position. Sausage meas-urements were made, analysed, modelsand simulators written. The sausageproblem was reduced by a factor of 10and is now an operational annoyance.We keep working on it. Same with thepointing that has been measured as goodas 0.85 arcseconds all over the sky.Roberto Rojas has inherited the thermalcontrol software, from Roberto Abuterwho escaped to Enteferometry [sic], andput it into full operation. This promises inthe medium to long term to provide a sig-nificant improvement in how we managethe telescope environment.

This article is being written as UVESended its first commissioning run onKueyen. UVES is looking good. Verygood! FORS2 is already mounted on thetelescope and has seen stars. Kueyen iswell on its way to be released on sched-ule for Period 65. Norma Hurtado andJulio Navarrete are driving the telescopeand providing valuable operator feedbackto the commissioning team. Baring un-foreseen events in 2000 both Melipal andYepun will have their first lights and wehave new ideas for further improvementsto the system.

To know who has made all of this pos-sible, access the web and get a list of allESO staff irrespective of location, posi-tion or function. We would like to thankKrister Wirenstrand personally for his tire-less efforts towards building and com-missioning the telescopes.

The Science V erification Plan for FORS2 and UVESat UT2/KueyenA. RENZINI and P. ROSATI, ESO

1. Introduction

The Science Verification (SV) obser-vations for FORS2 and UVES at Kueyen(UT2) will follow the same approach andpolicy as already described for the SV

• FORS2 SV: February 28 – March10 (New Moon is March 6). Raw, cali-bration and calibrated data will be madepublicly available to the ESO communi-ty as soon as the reductions are com-pleted and the data prepared for re-

of Antu (UT1) and its instruments(www.eso.org/science/utlsv/). SV ob-servations are now planned as fol-lows:

• UVES SV: February 6–17 (NewMoon is February 5)

E-mail: jspyromi@eso.org

25

lease. The SV Teams plan to releaseUVES data by March 31, 2000, and theFORS2 data by April 30, 2000.Observations will be made in ServiceMode with SV Astronomers on Paranalfor part of the time. SV observationsmay actually be taken from February6 through March 31 in a flexible sched-uling, depending on the needs of oth-er commissioning and training activitiesthat may arise.

SV observations will be conducted incoordination with the PIs of the two in-struments, the Instrument Scientists, andthe Kueyen (UT2) and instrument Com-missioning Teams.

The scientific programmes have beenselected in such a way as to produce first-quality data for cutting-edge science inareas of interest for a wide section of thepotential users in the ESO community.Moreover, programmes that require verygood seeing are complemented by oth-ers that would still be well feasible in lessdemanding conditions.

The SV Teams includes the followingscientists: Joao Alves, Stephane Arnauts.Jacqueline Bergeron, Tom Broadhurst,Stefano Cristiani, Chris Gorski, VanessaHill, Richard Hook, Rodrigo Ibata, Tae-Sun Kim, Markus Kissler-Patig, BenoitPirenne, Mario Nonino, Francesca Pri-mas, Michael Rauch, Alvio Renzini,Piero Rosati, and Eline Tolstoy.

The following list of SV programmeshas to be taken as indicative. In par-ticular, some additional relatively smallprogramme could be accommodated, de-pending on the actual observing condi-tions on Paranal. Any modification will bepromptly reported in the SV WEB pages(http://http.hq.eso.org/science/ut2sv/),where more detailed information can be

found, including target lists, modes andplanned exposure times.

2. UVES SV Plan Three main SV Programmes have

been planned:• A Spectroscopic Survey of Inter -

mediate Redshift QSOs . Five out of alist of 8 QSOs with 2 <~ z <~ 3 will be ob-served. These observations are meant tobe the first step towards a public surveyof QSO absorbers which aims at creat-ing a large and homogeneous dataset toallow the study of the intergalactic medi-um in the redshift interval z = 1.6–3. Ateam co-ordinated by ESO will submit aproposal to carry out such a Public Sur-vey in Period 66.

• Abundance Ratios of ExtremelyMetal Poor Stars from the HamburgESO Survey . At least 5 stars with [Fe/H]<~ –3 will be observed. The high S/N spec-tra will allow to determine the abundancesof a number of interesting elements, suchas Li, α-elements, and s- and r-processelements.

• Chemical Abundances in LMCClusters in a W ide Age Range . A setof Red Giants (16.5 < V < 17) will beobserved in seven LMC Globular Clus-ters with ages from 0.1 to ~ 14 Gyr. Thisshould allow a detailed reconstruction ofthe chemical enrichment history of LMC.

3. FORS-2 SV Plan The following programmes have been

planned for the SV observations ofFORS2:

• Deep Imaging of the Cluster ofGalaxies MS1008-1224 . This z = 0.3cluster was extensively observed forSV of FORS1. With FORS2 SV one in-tends to complete the database by

adding deep U and Gunn z images of thecluster.

• Hα Rotation Curves of DwarfGalaxies . This is a “bad seeing” pro-gramme. Long slit spectra should allowto constrain the mass distribution at largeradii in a set of three dwarf, dark matterdominated galaxies.

• Kinematics of Previously Mi -crolensed Stars in LMC . Over twodozens microlensing events have beendetected so far towards LMC. The planforesees to take spectra of at least 10 ofthe stars that were microlensed, in orderto obtain their radial velocity with an ac-curacy of ~ 10 km/s. Constraints on thenature and location of the lensing objectscould then be derived.

• Lyα emission at high redshift . TheNTT Deep Field will be imaged with a cus-tom narrow-band filter (FW = 7Å) search-ing for Lyα emission from large-scalestructure centred on a damped Lyα sys-tem at z = 3.2.

• Near-IR Imaging of the NTT DF .Imaging in IF915 and Gunn z filters will en-sure the completion of the deep multi-colour database of this deep field, thusallowing a photometric redshift search forvery distant galaxies.

• Spectroscopy of High RedshiftGalaxies . Distant, photometric redshiftselected galaxies to IAB Q 25 from pub-lic NTT data will be observed in MOSmode, possibly using the mask option.

• Abundance Measurements in Gi -ant Ellipticals . MOS and long-slit spec-troscopy for the abundance determina-tion of globular clusters and the dwarfgalaxies surrounding NGC4472 and ofthe galaxy itself out to ~ 3Re.

The First Six Months of VL T Science OperationsR. GILMOZZI, ESO

Science Operations started at UT1Antu on April 1, 1999. This article is abrief account of the first six months(Period 63). As the statistics below willshow, operations have been fairly suc-cessful. Visitor and Service modeswere intermixed during P63, in a 50/50ratio for the available time (i.e. dis-counting Guaranteed time, which is bydefinition in Visitor mode): we had 62nights in Visitor, and 60 in Service. Tothis one should add 10 Service nightsfor the Calibration Plans, and any timeallocated but not used for technicaltime, target of opportunity time and di-rector discretionary time (so that in theend, more than 85 nights were devotedto Service Mode observations and cali-brations).

Period 63 started under less thanpromising omens: the mask wheel of

ISAAC had just had to be taken out ofthe instrument to be repaired, so ISAACcould only work in imaging mode; thedecontamination of the FORS1 CCDhad not resolved the infamous “quad-rangle” problem, and the CCD had tobe taken out to be baked more thor-oughly, so FORS could not be used.And the weather was the worst seen inthe last year (at least by the ScienceOperations staff): foggy (!), cold andvery windy. April 1 and 2 were spentwith the Paranal Engineering depart-ment, the Garching Optical DetectorGroup and other Instrumentation Divi-sion staff working overtime on one sideto bring ISAAC up in imaging mode sothat the first Service Mode episode ofthe Period could start, on the other todecontaminate the CCD in time for thefirst dark moon and the first FORS1 run.

It is a testimony of the excellent workof all the people involved that on April 3,the wind and humidity having abated tothe more normal Paranal values, thefirst OBs could be executed on the sky.Paranal was finally a full-fledged astro-nomical Observatory!

From then on, the month spent in‘dry runs’ to acquaint ourselves withthe telescope and the instrumentsshowed its results: it was immediatelyclear that the efficiency of the opera-tions was quite high, even if some pro-grammes were not as efficient as theycould have been (the fact that ESO haddecided to absorb the overheads beingpossibly a reason for this), the seeingwas co-operating very nicely (thoughnot all through the Period, alas), the in-struments were behaving very well.Total downtime, both the weather and

E-mail: arenzini@eso.org

26

the technical flavours, was less than20%.

Ten days later the next challenge: thefirst Visiting Astronomers started com-ing to the mountain. I should probablyexplain that Service Mode and VisitorMode, as seen from the Operationspoint of view, differ only in who makesthe scientific decisions about ‘what todo next’. A Paranal astronomer is al-ways present to run the instrument, toadvise the Visitor and to help preparingthe OBs beforehand. Therefore theVisiting Astronomers can devote theirtime to analysing the data and decideon the observing strategy.

The experience with Visitor Modehas been very positive, the interactionwith the Operations Astronomers on thewhole resulting in satisfactory observa-tions (or at least this is what we gatherfrom the end of run reports!).

We have learned a lot during the firstsix months, and we have also made afew mistakes.

One thing that we have to clarify bet-ter to our “proposers” in the future isthat a successful Service ObservingMode (SM) needs to include pro-grammes designed for various types ofconditions: if everybody wants 0.4 arc-second seeing, photometric sky and beone day from new moon, it’s very diffi-cult that we can satisfy all requests...

In Period 63 there were two main is-sues that caused some confusion withour users: the oversubscription (espe-cially of category A) SM programmes,and the constraints set.

The first is not a problem per se:oversubscribing the available time addsto the scheduling flexibility and makesfor an efficient operation. Of course, italso creates unhappy users, who haveto submit their OBs without guaranteeof execution. The OPC ranking, whichis our primary decision input, helped inselecting among programmes compet-ing for the same conditions.

In P63 we were not helped by thetime being assigned in terms of shutteropen time: of the 600 available hours inService Mode, 720 were allocated inP63. With the measured values of over-head (~30%) and downtime (~20%), wewere oversubscribed by (1.3×720)/(0.8×600) ~ 2. Add to this that weathervariability sometime caused the condi-tions to change during an exposure, sothat it had to be repeated, adding to theoversubscription. We are actively study-ing ways to decrease the impact of thisproblem; in P63 we were lucky enoughthat part of the technical time (andsome of the discretionary time) werenot used, so that we put it into the SMpool, and were able to carry out moreobservations (see the statistics below).

The other problem we had is the con-straints set: the number of programmesthat could be observed when theseeing was > 1″ or the moon was up(for FORS), or for nights with cirrus wasvery small. In part this was compensat-

ed by the oversubscription, but in somecases it has been very difficult to decidewhat to observe. This problem is moreacute at the end of a period (again, weare thinking of ways to improve this).Particularly demanding was the Augustfull moon episode when ISAAC was notavailable for 7 nights, and we had touse FORS1 with very few programmesthat made sense to execute.

Sometimes there is no alternative butto violate some constraint: when wehad to do that, we always tried to com-pensate in some way (for example:higher than requested moon back-ground usually meant that we would ob-serve in better than requested seeing,to improve the contrast, or in thelongest possible wavelength, where theeffect is smallest, or both). Not alwaysthis has been possible; ranking hasplayed a role (if a choice had to bemade, we tried to avoid affecting veryhighly ranked programmes). But wehave found ourselves in situations (for-tunately few) where it was either ob-serve something outside the con-straints or keep the telescope idle.

Discussions are ongoing within ESO,and with the community, to minimisethese problems in the future. Users canhelp too, by not requesting indiscrimi-nately conditions that happen only asmall fraction of the time.

In the following I would like to offersome statistics for Period 63. Some ofthis statistics comes from the engineer-ing data stream that is recorded everynight, much in the same way that satel-lite engineering data are recorded (in-deed, this has proven to be among thebest tools produced by the Commis-sioning team to understand the per-formance of the telescope and instru-ments).

Performance of Antu

During period 63 there have been3907 presets (21 per night), 22,491 off-sets (6 per preset) and 121,537 imageanalyses (98% of which resulted in an

active optics correction). For those ofyou who have not yet had a chance tosee the active optics in action, it issomething really impressive: once youacquire a new field, the guide probegoes automatically to a guide star, ac-quires it, starts the image analysis and,usually about 30 seconds later, sendsthe first correction to the mirror actua-tors: at this point you see the guide starshrink to the outside seeing value, andyou can start your science exposure (see http://www.eso.org/outreach/press-rel/pr-1999/phot-34-99.html).

This is the most visible differencewith the way we used to do astronomy:from this moment on, the telescope iskept at the best focus, with the best op-tical figure for the pointing, all in paral-lel with the science observations andwithout requiring any additional time.The typical overhead to move to a newfield, acquire a guide star, and start theactive optics is about 5 minutes.

Downtime Statistics

One way to “measure” how well wedid in Period 63 is to compute the totaldowntime. The technical downtime was13,561 minutes (11.4% of the totalavailable time). This statistics includesany observing time affected by thetechnical problem (e.g. a problemwhich can be fixed in 5 minutes butwhich has caused the loss of a, say, 30min exposure would count as 35 mindowntime).

The weather downtime was 12,700minutes (10.7% of time). This includesclouds (5.6% of time), high winds(>18m/s) and high humidity (or combi-nations). Considering that P63 was dur-ing the southern winter, this percentloss does not seem surprising.

In total, therefore, we lost 22.1% ofthe time.

Weather

Of the 163 nights that were not lostto bad weather, 118 (72%) were ap-

Figure 1.

27

parently photometric, 18 (11%) possiblyphotometric, 17 (11%) had thin cirrus and10 (6%) had thick cirrus. We are currentlyfinalising the analysis of the standardstars observations to confirm the numberof “true” photometric nights.

The statistics on the seeing is basedon 116,761 ASM data points, each av-eraging one minute of measurements.The median seeing in P63 was 0.76”. Theseeing was better than 1” 72% of the time,and better than 0.5” 13% of the time.

You may have heard that sometimesthe seeing measured by the instru-ments on Antu is better than the outsideseeing. Discounting wavelengths effect(i.e. yes, it is true that ISAAC with its0.15″ pixels sometimes undersamplesthe atmospheric PSF), we have indeedseen this effect. It is however fairly rare,and needs special circumstances (inparticular, the appropriate wind speed,a pointing such that the wind can flushthe primary at the right angle, and anexcellent seeing to start with). On someoccasions, in service mode, we havehad enough programmes in the queuethat we could choose to observe in theright direction, and so took advantageof this. In general, however, and espe-

cially in visitor mode, it is an unlikely oc-currence.

Operations Statistics

ISAAC was used 73 nights, or 40%of the time. FORS1 was used 110nights. This reflects in part the instru-mental problems described above, inpart the actual time allocation (or theuser preferences).

Once one discounts the downtime,there were 1484 available “observable”hours. Of these, 651 were used forFORS1 and 403 for ISAAC. This resultsin a total “shutter open” efficiency of71% (see Figure 1). In period 63, the ar-chive “ingested” 19,794 ISAAC framesand 24,039 FORS1 frames.

In Service Mode, we had 75 accept-ed programmes (40 in category A, 19 inB and 16 in C, or 40/19/16 for simplici-ty). The completion statistics is 34/16/11 (of which 13/10/8 are partial comple-tions, usually meaning that most of theprogramme was carried out, the re-mainder being impossible to observebecause of target RA or non realisedconstraints etc). The programmes notinitiated were 0/3/5. The remainder

(6/0/0) are category A open pro-grammes that the Director General hasallowed to carry over into Period 64 sothat they can be finished.

The success of the first six months ofoperations is the result of the commitmentand professionality of a very large num-ber of people, both in Europe and in Chile.Naming all of them would look very muchlike the ESO internal organigram! How-ever, I would like to mention at least theScience Operations Team at Paranal(now under the leadership of GautierMathys), the Paranal Engineering De-partment (Peter Gray), the User SupportGroup (Dave Silva) and the Data FlowOperations Group (Bruno Leibundgut) fortheir dedication and enthusiasm, and fortheir crucial contributions. Jason Spy-romilio and his team delivered fully com-missioned telescope and instruments.Massimo Tarenghi as director of Paranalprovided oversight and direction to thewhole process.

Special thanks to Jason, Gautier,Dave, Chris Lidman and Jose Parra forproviding the statistics in this article.

Configuration Management of the V ery LargeTelescope Control SoftwareF. CARBOGNANI, G. FILIPPI, P. SIVERA, ESO, Garching

1. Introduction

One of the elements of the successof the development, integration andcommissioning of NTT, VLT and the at-tached Instruments has been the Con-figuration Management of the relatedControl Software. It has been basedon the Code Archive and the VLTSoftware Problem Report (VLTSPR)procedure.

2. Code Archive

The code archive supports code con-figuration during development and inte-gration on geographically distributedsites. The design keys are the following:

• The configuration item is the soft-ware module. This allows a reasonablebut still simple flexibility in system con-figuration (a system is made up of 15 to100 items, identified by their name andversion, corresponding to 2000 to20,000 files). Each module is a set offiles organised in a fixed directory struc-ture and has a unique name. Figure 1gives the modules divided by VLT com-mon software and applications.

• There is only one central archive.Users get local copies using a simple

client server mechanism. In our proj-ects, there is typically only one personat a time dealing with a specific part. A

modification must be started, imple-mented and tested, then archived.During this period the module is locked.

Figure 1.

E-mail: rgilmozz@eso.org

28

• To allow experiments or patches onolder versions, branches are also sup-ported.

The code archive is implemented us-ing RCS and a set of ad hoc pro-grammes and scripts (cmm) imple-menting the client-server interactionand the user interface. The archive isphysically located at the headquartersand used also by teams in Chile andseveral institutes.

Figure 2 gives the accesses permonth for modification and for read-only mode. Currently approximately60% of the accesses are from outsidethe headquarters and 25% from non-ESO sites. It can be remarked that thesystem was able to deal with more than500 archive and 2500 read-only month-ly accesses.

Figure 3 shows in detail the accessmade by non-ESO sites.

The central archive is not only the soft-ware repository, but it is also a manage-ment tool. In archiving a module, the de-veloper tells all the other people “Hey,there is a new consistent and tested setof files that you can use!”. This is veryuseful when several teams operating ondifferent time zones do development andintegration. Comparing the current con-figuration against the status of the

archive, the integration responsible cansee that new items have been produced.A glance to the comments stored by thedeveloper to qualify the newly archived

version is normally enough to decidewhether to take the new version or not.In this way NTT, VLT and instrumentshave been developed, integrated and

Figure 2.

Figure 3.

29

commissioned on top of mountains at2500 m in the middle of a desert, but in-volving people in several countries onboth sides of the ocean. The CentralArchive is hosted on a RAID system lev-el 5, with hot-standby disk, and backedup in two different ways every day. Thisyear the downtime of the server has beenless than 8 days, corresponding to 2%,mainly during weekends. We are devel-oping a better client-server mechanismand we plan to make the cmm softwareavailable on the public domain.

3. VLT Software Problem Report(VLTSPR)

The VLTSPR System is meant to beused by both internal and externalusers of VLT software to report errors incode or documentation or to propose achange. VLTSPR is built using the com-mercial tool Action Remedy (c) and hasa Web Browser interface. This is the ba-sic workflow of the system:

• Problem submitted, depending onthe subject, some people are notifiedimmediately

• The SPR is discussed in the SoftwareConfiguration Control Board meeting and

a Responsible Person is appointed for theproblem

• Responsible works on problem• Responsible can close the SPR

(must add a final remark on it)• People can add comments any

time.At present there are about 120 names

in the user database, 75% ESO users(both Europe and Chile), the remaining25% from 13 external projects, some proj-ect have more sites. Figure 4 shows thepresent distribution of the VLTSPR users.

In Figure 5 the trend of the SPRarchive over the years is shown. Thenumber of open SPRs has always beenkept under a physiological limit (about500) that corresponds to what we are ableto treat between two releases.

Figure 6 reports the distribution of theSPRs per area in 1997 and in 1998. Asthe project evolved, SPRs concerning thecommon software are decreasing, whilethe ones for the application part are go-ing up, which is a sign of the integrationactivity that is now taking place.

4. Conclusion

Tools and Methodologies developedwithin the VLT Software EngineeringGroup have proved to be effective insupporting the VLT Software develop-ment. They can be seen as a concreteguideline and reference for all ESO in-ternal projects and are proposed asbaseline for all external collaborations.

This article is based on reportspresented at ICALEPCS 99:[1] G. Filippi, “Software Engineering for

ESO’s VLT project”, ICALEPCS 93.[2] G. Filippi, F. Carbognani, “Software prac-

tices used in the ESO Very Large Tele-scope Control Software”, ICALEPCS 99.

Figure 4.

Figure 5.

Figure 6.

E-mail: fcarbogn@eso.org

30

Overview

The first part of this article discussesthe Java computing language and a num-ber of concepts based on it. After learn-ing why Java is best suited for many ap-plications in astronomy including ESO’sJSky initiative and certain Web servicesthere is a second part containing a col-lection of application briefs from softwareprojects at ESO and ST-ECF. Appendedis a list of contact points in house (Table1) and an applet gallery (Table 2).

1. Why Java?

Here are the key features [1] that theauthors felt were the most important ar-guments why Java was the first choicefor their projects:

• Portability: There is only one versionof source code and one version of bina-ries for all operating system-architecturesand hardware platforms. As a result of thearchitecture-neutral capabilities, Javaprogrammes execute on any platform for

which a Java Virtual Machine (JVM) ex-ists (e.g. Windows, Solaris, Mac OS, andLinux). Some web browsers come witha built-in JVM.

• JavaBeans Component Model = built-in support for code re-use: Java incor-porates JavaBeans which is a softwarecomponent model. The component mod-el describes how to build an applicationfrom software components. In contrast tosoftware libraries Beans provide an ad-ditional interface in order to support theirre-use and customisation. The Java-Beans concept is widely supported by de-velopment tools.

• Common look and feel: There is built-in support of powerful GUI widgets(Swing classes) and common look andfeel on different platforms.

• Applets: Applets are programmesrunning via web browsers. Therefore, noinstallation is necessary on the side of theweb client and also no maintenance.Computations are performed on the clientcomputer thereby off-loading our servers.

The next two bullet lists summarise

more concepts and advantages as wellas the weak points of Java technology.

More advantages and supported concepts

• Remote Method Invocation (RMI):RMI allows the execution of tasks on oth-er machines.

• Jini: Jini defines a protocol for com-munication with devices. A device is, forinstance, a disk drive, a TV, VCR or cell-phone. With Jini there is no need for sep-arate driver software. It's a very generalimplementation of plug and play capa-bilities.

• Enterprise JavaBeans (EJB): EJB isa specification for server side compo-nents based on JavaBeans. EJB com-ponents neither deal with server-side sys-tem-level programming nor with client-side interfaces and therefore form a mid-dle tier containing business logic only.This gives much more flexibility whencoping with changes of business rules.

• Lots of free software and a lively com-munity with newsgroups

• Object-oriented: This implies en-capsulation of data in objects and sup-port of code-reuse.

• Security: Various security mecha-nisms prevent malicious code from be-ing executed.

• Multi-threaded: Java supports threadsynchronisation and enables an appli-cation to perform several tasks in paral-lel even if the underlying operating sys-tem is not capable of multitasking.

• Late binding: Java is capable of dy-namically linking in new class libraries orinstance variables at runtime.

• Robustness: Strict type checking, nopointers, no explicit memory (de)alloca-tion, enforcing boundary checking whenaccessing arrays

Applet specific advantages• No installation: Just requires a Java

enabled web browser.• Automatic software update: The lat-

est software version is loaded wheneveran applet is invoked. There exists only onecopy of the software on the web server.

Weaknesses• Rapid development led to a steady

stream of new Java versions and frame-work specifications rendering it a movingtarget. It is only slowly settling down.

• Write once but install (in case of ap-plications) and test everywhere: Platformspecific bugs such as incompatibilities be-tween JVMs are hard to detect and cir-cumvent.

Java for Astronomy: Software Development at ESO/ST-ECFM. DOLENSKY1, M. ALBRECHT2, R. ALBRECHT1, P. BALLESTER2, C. BOAROTTO2, A. BRIGHTON2, T. CANAVAN2, M. CHAVAN2, G. CHIOZZI3, A. DISARÒ2, B. KEMP2

1ESA/ST-ECF; 2ESO/DMD; 3ESO/VLT

Applet Name Description URL

ESO ETC ESO Exposure Time Calculator http://www.eso.org/observing/etc/

Isql JDBC interface to databases http://archive.eso.org:8009/sample/gateway.html

JIPA Download Page of Java http://archive.eso.org/java/jipa1.35/Image Preview Application

JPlot Applet for Jitter Analysis http://archive.eso.org/archive/jplot.html

Spectral HST Preview Applet for Spectra http://archive.eso.org/preview/preview/preview_hst/Y29E0305T/fits/spectral/4

TimeGraph [10] Applet for System Load Monitoring http://www.hq.eso.org/bin/tg?schema=sysload

Table 2: Applet Gallery.

Service Contact Point

Java Interest Group (JIG) Mailing List jig-list@web3.hq.eso.org Send mail to Tim Canavan tcanavan@eso.org for registration.

Java tools in archive and dmd Bob Kemp bkemp@eso.org UNIX domains

Java tools for archive operations and Markus Dolensky mdolensk@eso.org ecf domain

JSky Home Page http://archive.eso.org/JSky/

JSky Mailing List jsky@egroups.com Send mail to jsky-subscribe@egroups.com for registration.

JSky Repository ftp://ftp.archive.eso.org/pub/jsky

P2PP Project Page http://www.eso.org/~amchavan/projects/jp2pp/development-docs.html

Table 1: Contact Points and Java Links in House.

31

• Imposes strict adherence to objectoriented design in contrast to C++. In fact,this could be seen as an advantage aswell.

• Programmer has no direct controlover system resources. Therefore flawsof built-in services like the garbage col-lection algorithm are hard to circumvent.

Applet specific weaknesses• Security restrictions make it difficult

to implement basic features like savingthe programme status and printing.

• Lacking browser support, latencyproblem due to limited bandwidth and thelack of adequate caching mechanismscan lead to frustration of users.

2. Application Briefs

DBB – The Database BrowserProject

The Database Browser Project isaimed at developing a library of Javaclasses which can be used to create in-teractive database browsing and editingtools. Many data-centred applicationsshare a common structure: A user wantsto select a set of items (rows) from a data-base according to some searching andsorting criteria, reviews a summary of thereturned items, edits one or more itemsand saves the edited items back to thedatabase. The DBB classes provide theapplication programmer with a set of toolsto quickly develop interactive tools of thatkind. By specialising the library, a pro-

grammer can fine tune the look and feelof the application to the desired level.DBB is used for the new Phase II Pro-posal Preparation System (P2PP) V.2.

DocumentView: A way of linkingGUIs and Objects

DocumentView is a set of classes thatlink object properties with GUI components.For instance, one can link a text fieldwidget with the title of an ObservationBlock (OB). Changes in each affect theother, and thus multiple GUI componentsdisplaying the property are kept synchro-nised. DocumentView is a componentused for the P2PP tool described below.

A New Generation Phase IIProposal Preparation Tool

The successor to the current versionof P2PP (Phase II Proposal PreparationSystem) is an application being devel-oped using the Java programming lan-guage. The new version is largely back-wards compatible with the current sys-tem, and its basic purpose is to providehelp for the preparation of ObservationBlocks (OBs) both at the observer's homeinstitution and at the telescope.

The decision to rewrite version 2 of thesystem in Java was based on two fun-damental requirements:

• To provide easier support for multi-ple platforms, since Java libraries areavailable for most operating systems andare extremely consistent in their behav-iour across platforms.

• To provide a cleaner, more intuitiveuser interface. One of the main aims ofthe Swing graphics library for Java is toprovide platform-independent high-levelgraphical components.

Design Patterns with names like Fac-tory and Visitor [2] were used whereverreasonable in order to reduce the learn-ing curve for future maintainers. P2PP V.2can function as a 2- or 3-Tier client/serv-er system with a relational database backend and an optional business object mid-dle tier. However, users can work com-pletely off-line, saving unfinished work ontheir local machine. Connection to theESO OB repository is only needed in or-der to submit their OBs for execution.

Java Interface to VLT ControlSoftware

This class library allows Java appli-cations to interact with the non-Java VLTControl System. It provides interfaces tothe the VLT Central Common Software.These interfaces are implemented asJava Native Methods, relying on the lega-cy VLT Common Software libraries im-plemented in C/C++.

The new library allows the develop-ment of high-level interfaces and VLTcontrol applications in Java. It wasmade available for the first time as partof the OCT99 VLT Software Release.

P2PP V.2 uses this interface to trans-mit OBs to the VLT Unit Telescopes inParanal and to get status informationabout their execution.

Figure 1: Alpha Release of STScI's archive browser StarView II. It uses ECF's utility JIPA to display a sky region from ESO's copy of the DSScatalogue together with an HST aperture overlay.

32

Java Calibration Database Manager

The Calibration Database includes acollection of observation data and re-duction procedures representing ascompletely as possible the observingmodes and configurations of the differ-ent VLT instruments. The CalibrationDatabase Manager is the standard ap-plication for browsing, editing, populatingand aligning the local calibration databasecontents. This Java application is the in-terface between the pipeline local cali-bration database and the ESO archivesystem. It allows the ESO Quality Con-trol Scientists of the Data Flow Operationsgroup to request and submit files to thearchive, and to structure calibrationinformation to the formats supported bythe pipeline and needed for long-term ar-chival and trend analysis. The CalibrationDatabase Manager and other quality con-trol applications are described in more de-tail in a previous issue of The Messen-ger [3].

ESO Exposure Time CalculatorThe user interface of the on-line Ex-

posure Time Calculators provided for var-ious VLT and NTT instrument as well asthe Wide-Field Imager was implement-ed as a combination of HTML and Java.An observer specifies the relevant pa-rameters in a web form, the necessarycomputation takes place on an ESO serv-er and the results are returned in a pagecontaining HTML text and a plot applet.

The applet is based on Leigh Brook-shaw’s GNU plot library [4].

ST-ECF Applets for the ArchiveInterface

ECF started its efforts in this field aboutfour years ago. Early on applets were re-leased on the web. They are used for pre-viewing HST spectra and images [5].There is JPlot – a plot utility for jitter analy-sis of HST observations [6]. JPlot wasused for quality control purposes in firstplace and later on put on-line as a pub-lic archive service. There is also Skymap,a simple applet that draws locations ofHST parallel observations on a map.

As a logical next step, more ambitiousprojects followed which are currently ac-complished in collaboration with other in-stitutions like STScI and ISO IDC [7], [8].The aim is a pure Java interface to bothHST and ESO archives (Fig. 1) that willserve as a more powerful alternative tothe current web interface.

JSky – Reusable JavaComponents for Astronomy

The JSky development effort aims toprovide a library of reusable Java com-ponents for astronomy. Based on expe-riences with the Skycat application, thefirst components being developed are forimage and catalogue display and ma-nipulation. The goal is to eventually havea collection of general-purpose and as-tronomy-related JavaBeans that may beeasily assembled into applications using

a Bean Builder, ideally with little or nohand-written code necessary.

The prototype can display FITS im-ages and supports zooming, panning,colour maps, and automatic cut level set-ting. Initial tests indicate that the per-formance is satisfactory for small images,as long as the JAI native library (mlib) isavailable. At the time of writing, this is onlythe case under Windows and Solaris, al-though a Linux version is due out soon.The current version does not yet handlelarge images well, but can be optimisedto do so. The FITS reader is based onThomas McGlynn’s Java implementation,which currently always reads in the en-tire image. A future version will providesupport for splitting the image into tiles,which is the only way to efficiently han-dle large images in JAI.

The image display application (Fig. 2)currently has a hard-coded layout. In theend, the goal is to build the applicationfrom JavaBeans and allow the users tochange the layout, add new panel items,remove unwanted ones, etc.

The Java implementation of the cata-logue browser looks pretty much like theSkycat version, so far, although this willlikely change. The panel layout can begenerated automatically, based on an eX-tensible Markup Language (XML) cata-logue description and the catalogue dataitself may also be in XML [9]. The newcatalogue interface should be more flex-ible than the Skycat version and allow var-ious catalogue types to be supported bymeans of a catalogue registry.

JSky is an open development platformand contributions of general-purpose orastronomy-related software and compo-nents are welcome. There is a JSky homepage [Table 1] as well as a JSky ftp repos-itory [Table 1].

References

[1] A. Walsh, J. Fronchkowiak, 1998, “JavaBible”, IDG Books Worldwide Inc, p. 7–18.

[2] Gamma et al., 1995, “Design Patterns: El-ements of Reusable Object-OrientedSoftware”, Addison-Wesley.

[3] P. Ballester, A. Disaro, D. Dorigo, A. Modi-gliani, J.A.Pizarro de la Iglesia, “The VLTData Quality Control System”, The Mes-senger 96, p. 19.

[4] GNU Graph Class Library http://www.sci.usq.edu.au/staff/leighb/graph/

[5] Dolensky et al., 1998, “HSTArchive Ser-vices Implemented in Java”, The Mes-senger 93, 26.

[6] Dolensky et al., 1998, “How the Analysisof HST Engineering Telemetry Supportsthe WFPC2 Association Project and En-hances FOS Calibration Accuracy”, TheMessenger 93, 23.

[7] Binegar et al., 1999, “Browsing astronom-ical archives with StarView II and inter-acting with analysis tools such as JIPA”,in ASP Conf. Ser., ADASS 99, in press.

[8] http://archive.eso.org/~mdolensk/publ/jskybof/[9] http://vizier.u-strasbg.fr/doc/astroxml.htx

[10] TimeGraph Download Page http://www.ece.utexas.edu/~tbieleck/

Figure 2: A first prototype of an image viewer has been developed, based on the JAI (Java Ad-vanced Imaging) package from Sun.

E-mail: mdolensk@eso.org

33

1. Background

In the era of 4-m-class telescopes itwas common to select targets and/orprepare observations using deep Schmidtplates for which full sky coverage wasavailable down to m ~ 23. Now, in theblossoming era of 8-m-class telescopesone needs much deeper survey datato feed high-throughput, high-multiplexspectrographs able to reach at least twomagnitudes deeper, or to find relativelyrare objects such as clusters of galaxies,brown dwarfs, or trans-neptunian objects(for an extensive list of such potential tar-gets see Renzini 1998, The Messenger,91, 4). To a large extent, the scien-tific outcome of an 8-m-class telescopecritically depends on the availability ofpreparatory imaging surveys, and the de-mand for deep and wide surveys will in-tensify as other large-aperture tele-scopes become operational world-wideand the competition becomes more andmore fierce.

CCD and computer technology haveevolved at a rapid pace over the last twodecades, and multi-colour, digital imag-ing surveys of appropriate area and depthare within reach. Even though there area number of ongoing optical/infrared dig-ital all-sky surveys (DENIS, SDSS,2MASS), none of them are particularlysuitable for feeding targets for large tel-escopes (for most possible targets theyare not deep enough) nor will they bepublicly available in the near future. Sur-veys more suitable to large telescopesare being conducted at several obser-vatories (e.g., CFHT, CTIO, La Palma, KittPeak) with the products becoming pub-lic within their respective communities.Groups with specific science goals arealso conducting ambitious surveys, butagain these data are unlikely to becomeaccessible to general ESO users in ashort-time scale.

To meet this challenge, in 1996, ESOtook the initiative of proposing to car-ry out a public survey to fulfill at leastsome of the short-term needs of theESO community in view of the first sci-entific runs of the VLT. Without hamper-ing the possibility of groups to conducttheir own surveys, a public survey offersa number of advantages. In particular,data can be used for a broad rangeof scientific applications, thus result-ing in a substantial saving of telescopetime (that can be as high as a factor of10!), compared to the case in whicheach individual science group has to se-cure its own survey data. Moreover, pub-lic survey data can be easily comple-

mented by “privately” acquired data forspecific applications, such as in the caseof second-epoch or narrow-band filter ob-servations.

A Working Group (WG) was thereforeappointed in 1996 to design the surveyand to supervise the work of the SurveyTeam charged to carry out the observa-tions, data reduction and distribution ofsurvey products. Both the WG and theSurvey Team were assembled from thecommunity itself, with ESO providing thecoordination and the required facilitiesand resources.

The main objectives of the resultingESO Imaging Survey (EIS) were toconduct a public optical-IR imagingsurvey at the NTT, to reduce the data,construct object catalogues, and selectfrom them special classes of objectsof potential interest for VLT pro-grammes. All this within a short timescale, before the start of operation of thefirst Unit Telescope of the VLT. In addi-tion, EIS was seen as a first step towardsdeveloping an appropriate environmentfor carrying out future surveys, pub-lic as well as private. Indeed, the softwaretools developed by the programme wereregarded as an integral part of thesurvey products, to be disseminatedthroughout the ESO community to facil-itate the reduction and analysis of wide-area imaging data. The dissemination ofsuch tools was seen as essential for thecommunity to take full advantage of ded-icated imaging telescopes such as theMPG/ESO 2.2-m telescope, now alreadyin operation, and the VLT Survey Tele-scope (VST, Arnaboldi et al. 1998, TheMessenger, 93, 30) to be installed onParanal.

2. The EIS Project and the WFIPilot Survey

In practice, to combine as far as pos-sible wide-area coverage and depth, EISconsisted of two complementary projects:EIS-WIDE and EIS-DEEP. The goal ofEIS-WIDE was to cover with EMMI@NTTfour patches in the sky, 6 square degreeseach, providing V and I data to a limitingmagnitude of VAB ~ 24.5 and IAB ~ 24.This part of the survey was optimised todetect candidate clusters of galaxies atmoderate to high redshift (z <~ 1). A ~ 2-square-degree fraction of one of thepatches was also to be observed in U andB to allow for the identification of QSOs,metal-poor stars, etc.

The goals of EIS-DEEP were to cov-er with SUSI2 and SOFI three adjacentpointings in the Hubble Deep Field

South including the fields covered bythe various HST cameras, and fouradjacent pointings in the AXAF deep field.In both cases the multicolour UBVRI-JHK data should have been deep enoughto provide accurate photometric red-shifts and to select Lyman-break galax-ies. In practice, this required limitingmagnitudes in the range 26–27 in theoptical passbands and 23–24 in theinfrared.

The one-year period (July 1997–July1998) available to achieve the primarygoals of EIS-WIDE represented a formi-dable task, as no consolidated experi-ence existed at ESO or in the communi-ty. Apart from the limitations due to theweather, the goals were only met thanksto the enthusiastic response of many sci-entists from the community who broughttheir specific experience to ESO and tothe resources made available at ESO tosupport the temporary relocation ofthese scientists. The contribution ofESO and ECF staff was also critical forseveral aspects of data processing,archiving, and distribution.

With the advent of the Wide Field Im-ager (WFI) at the 2.2-m telescope(Baade et al 1999, The Messenger, 95,15) a “WFI Pilot Survey” was designedby the WG and recommended by theOPC to be conducted by the EIS Team.The scientific goal of the Pilot Survey wasessentially to complete that part of theEIS-WIDE project that could not be donedue to bad weather during the allocatednights (it is worth recalling the impact ofEl Niño in 1997–98). The “pilot” desig-nation was meant to reflect the devel-opment period required for the major up-grade of the survey software to enable itto pass from one-CCD (EMMI) or two-CCD camera data to the eight 2k × 4kCCDs of WFI.

The companion article by da Costa etal. in this issue of The Messenger de-scribes in some detail what was achievedby these surveys, with emphasis on thesurvey products. All the data from EIS-WIDE, EIS-DEEP and the WFI Pilot Sur-vey were delivered to the community indue time. Over the last three Calls for Pro-posals (P63–P65), over 100 proposals forfour different Panels, were based on theEIS data, or referred to EIS. Follow-up ob-servations were conducted using ESOtelescopes as well as other facilities toprune samples in preparation of VLT pro-posals. Targets selected from the EISdata were also used during Science Ver-ification of UT1/Antu as well as ofFORS1 Commissioning and Science Ver-ification.

E IS – THE ESO IMAGING SURVEY

The ESO Public Imaging SurveyA. RENZINI and L. DA COSTA, ESO

34

3. The Public Survey , Period 64to 67

The usefulness of Public Surveys insupport of VLT programmes beinggenerally recognised, the WG at itsmeeting in September 1998 recom-mended ESO to issue a “Call for Ideas”for future public surveys to be con-ducted with the 2.2-m and NTT tele-scopes (The Messenger 94, 31). The re-sulting “ideas” were discussed in depthby the WG in March 1999, and two opti-mised proposals emerged. A small short-term proposal to complete the EIS-WIDEsurvey reaching the goals as originallystated, and a two-year Public Surveyprogramme. The latter consists of threeparts: (1) a deep multi-band (UBVRI)optical survey with the WFI over threefields of one square degree each; (2) adeep near IR survey with SOFI, cover-ing two 450 square arcmin fields select-ed within the area observed in the opti-cal; and (3) a shallower set of WFI ob-servations in V and I of selected stellarfields in preparation for FLAMES atUT2/Kueyen (hereafter the “Pre-FLAMESSurvey”).

The selected fields for the deep op-tical/IR survey (hereafter Deep PublicSurvey, DPS) are given in Table 1. Themain scientific drivers for this part ofthe Public Survey are the search forhigh-redshift objects: galaxies (e.g. Ly-man-break galaxies), clusters and QSOs,as well as the possibility to de-rive photometric redshifts from such alarge, homogeneous, multicolour data-base. The goals and requirements interms of limiting magnitudes are sim-ilar to those of EIS-DEEP describedabove except that DPS is meant to cov-er a much larger area. The same data-base will also provide a unique and ur-gently needed tool for studies of Galac-tic structure and stellar populations. TheDPS is therefore designed to provide tar-gets for several of the first-generation VLTinstruments, with emphasis on FORS1and FORS2, ISAAC, UVES, VIMOS, andNIRMOS.

On the other hand, the Pre-FLAMESSurvey is of crucial importance for stel-lar astronomy studies with the FLAMESSystem on UT2/Kueyen, where theOzPoz fibre positioner can pick up~ 130 targets over a field of 25 diameter(well matched by the field of view of WFI)and feed the MEDUSA mode of GI-RAFFE. Alternatively, 8 fibres fromOzPoz could send the light of as manytargets to UVES, over to the other Nas-

myth platform of the same telescope (seeRenzini 1999, The Messenger, 96, 13).The accuracy of fibre positioning, whichrequires accurate target coordinates tostart with, is critical. Existing stellar cat-alogues fail to give the information thatis necessary to operate FLAMES, i.e., co-ordinates with the required astrometricaccuracy <~ 0.″1 rms) combined with mul-ticolour information. The tentatively se-lected fields for the Pre-FLAMES Surveyare listed on the EIS web pages. Inter-ested scientists are welcome to give sug-gestions to complement and/or improvethe list.

The Public Survey proposal was sub-mitted as a Large Programme in re-sponse to the Call for Proposals for Pe-riod 64, applying for a total of 81 nights:for the DPS 54 nights at the 2.2-m tele-scope and 15 nights at NTT for itsIR part, plus 12 nights at the 2.2-mtelescope for the Pre-FLAMES Survey.This Large Programme was proposedfor a two-year period, hence being dis-tributed over ESO Periods 64 through67. Details of the Public Survey, includ-ing the original proposal, can be found at http://http.hq.eso.org/science/eis/public-surveys/

This Public Survey “large proposal”was approved and recommended by theOPC, and 18 nights with WFI@2.2 and8 nights with SOFI@NTT have beenscheduled in Period 64. To increase theparticipation of the community in the Pub-lic Survey, an Announcement of Oppor-tunity was also issued for groups in thecommunity to carry out the observationsand data reduction of the infrared part ofthe DPS, following the same guidelinesof the EIS Team. Following an examina-tion of the proposals, the WG has alreadyselected the Team that will be in chargeof this part of the survey (PI H. Jørgensen,Copenhagen).

The Public Survey in periods 64–67 willprovide the ESO community with a com-petitive edge in the realm of the high-red-shift universe as well as in stellar stud-ies. Furthermore, it will allow the contin-uation of the development effort of EISthat will guarantee a smooth transition be-tween the WFI@2.2 and the VST that willbe equipped with ΩCam, four times larg-er than WFI in both field of view and num-ber of pixels.

4. The Public Survey Productsand Distribution

Among the many requirements of aPublic Survey, the most important is per-haps that the data products be easily re-trievable by external users. While thismay seem trivial, practice has shown thatit is not always simple to reconcile theneeds of different users – with the impliedlarge variety and size of the products –with other practical constraints such asefficiency and resources. In addition, datafrom ground-based optical observationsare subject to atmospheric conditions and

thus require more information to be prop-erly characterised, relative to radio orspace-based data.

The EIS/Pilot Survey project madeavailable four kinds of data products: as-trometrically and photometrically cali-brated images, filtered and verified ob-ject catalogues drawn from single-colourcoadded frames, derived cataloguessuch as colour catalogues, and targetlists. Each release was also accom-panied by papers describing the ob-servations, data reduction, the differenttypes of catalogues available and the tar-get lists produced. The object catalogueswere also essential to validate the databy computing simple statistics such as thenumber counts of stars and galaxies,colour-colour diagrams for point-sourcesand the two-point angular correlationfunction for galaxies. This allowed thephotometric zero-points, star/galaxy clas-sification, and uniformity of the data to betested.

The primary reason for adopting thisdata release model was the strong de-sire to have verified data publicly avail-able in the shortest possible time to meetthe stringent deadlines imposed by theimminent start of the VLT regular opera-tions, which began with the submissionof the proposals for Period 63 (deadlineOctober 1, 1998). However, this modelis not the most suitable for long-term proj-ects such as the Public Survey describedin the previous section. The time base-line for the completion of the project ismuch longer, and intermediate productsmay be valuable for different applications.While it is obviously highly desirable tocut to the minimum the time lag betweenthe collection of the survey data andthe VLT follow-up, the sheer volume ofthe data expected from WFI@2.2m re-quires more time for detailed verifi-cation of the data products. In addition,users may have their own ideas on howto best extract catalogues for their spe-cific scientific purposes, while final re-sults for a given sky area can only be pro-duced after all the observations are com-pleted.

One solution to accommodate thesevarious constraints is to provide botha prompt release, soon after an ob-servation run is completed, and amore complete release on a time tablesynchronised with the ESO observingperiods. In this framework one pro-vides rapid access to reduced – butstill unverified – data, thus allowingusers to extract objects directly fromthe images. Meanwhile, the SurveyTeam will work to prepare completeand fully verified data releases. The aimis to strike a balance both between theneeds of the community and the opera-tional requirements, and between speedand quality. Following this approach,the distribution of data accumulatedfrom the Public Survey will consist ofprompt and final releases, which will oc-cur on different time scales, as describedbelow.

Field α(J2000.0) δ(J2000.0)

Deep 1 22 43 00 –39 58 00Deep 2 03 32 28 –27 48 00Deep 3 11 22 00 –21 35 00

Table 1: Deep Public Survey SelectedFields.

35

4.1. Prompt releases

The prompt releases will consist ofproducts that are generated directly bythe EIS-WFI pipeline. These will providethe opportunity for users to have a “quick-look” at the data soon after they havebeen acquired. These releases will con-sist of:

• Astrometrically and photometricallycalibrated coadded images of dithered,single-pointing exposures and associat-ed weight maps. The astrometric andphotometric accuracy shall be better than0.″05 and 0.1 mag,

• Catalogues of objects extracted fromthese images using SExtractor, in theform of FITS binary tables.

Such releases are expected to startas soon as the ongoing work of up-grading the EIS-WFI pipeline is com-pleted and the new version properly test-ed, which should occur by the end of1999.

These data together with SExtractor,LDAC and Drizzle softwares, which arepublicly available and can be retrievedfrom the EIS home page, or directly fromtheir respective authors, will enableusers to work in parallel to the SurveyTeam, starting from the same basic dataand at the same time. Users will, there-fore, be able to evaluate the data and pre-pare, if they so desire, their own cus-tomised catalogues without having to relysolely on those derived by the SurveyTeam. Moreover, new tools are in theprocess of being installed which will fur-ther enhance the scope of work that canbe done outside ESO without the needfor distributing large amounts of raw orprocessed data.

4.2. Complete releases

Complementing these prompt re-leases, complete releases of verifieddata including all the data accumulatedduring all previous observing periodswill occur every 6 months, namely onFebruary 15 for odd periods and August15 for even periods. These will includenot only images but also a host of de-rived products which require more timefor their proper preparation and verifi-cation. These releases will include:

• Tools to allow the extraction of imagesections (cut-outs) from image mosaicsand associated weight maps

• Co-added, mosaiced images of in-dividual patches and associated weightmaps

• Colour catalogues for multi-passbandobservations

with all image products being photo-metrically and astrometrically calibratedas specified above.

Each release will be followed by anupdate of the web and will be ac-companied by papers prepared by theSurvey Team. These papers are an in-tegral part of the data distribution asthey describe in detail the observations,

the quality of the data, the data reductionand verification procedures, and the de-scription of the content of the derived cat-alogues and of the methodology usedin their preparation. These papers arealso important as a reference, therebyrecognising the commitment of ESO inproducing the data and in making themavailable to the ESO community. Whileno direct scientific exploitation of the datawill be done in these papers, they wouldstill represent a reasonable reward for themembers of the Survey Team for their ef-fort in conducting such a survey of pub-lic utility.

5. Data Distribution Logistics

A daunting task for public wide-fieldimaging surveys is finding the waysand means of exporting the data to out-side users. To give an indication of thescale of the problem it suffices to men-tion that the Pilot Survey alone has pro-duced over 0.5 Terrabytes of data inroughly 11 nights of WFI observing, while66 nights have been assigned for thePublic Survey at the WFI@2.2. Even af-ter co-addition of all dithered images, asingle-passband 6-square-degree mo-saic consists of 6.5 Gigabytes, and acomplete data set for the recently ob-served EIS patches (combining data fromEMMI and WFI) comprises 64 Gb of im-ages and weight maps, and 32 Gb of con-text maps. The stellar fields comprise an-other 63 Gb of images. Therefore, withthe presently available technology it is un-realistic to consider the distribution of pix-el maps for an undetermined number ofrequests. To make the distribution pos-sible, with the resources currently avail-able at ESO for this task, a certain num-ber of constraints are unavoidable. Thedistribution of bulk data must be limitedto well-defined packages of reduced data,produced semi-automatically; distributionof large volumes of data must be limitedto certain media, e.g. DLT model 7000which store 35 Gb and are produced inabout 2 hours (5 Mb/s). Large volumesof data cannot be requested by a simpleclick on the web page, but will have to berequested by submitting a statement ofpurpose which should include a scientif-ic justification and demonstrate that re-sources are available at the home insti-tute to handle the requested amount ofdata.

Note that even a single WFI frame (267Mb) cannot be transferred by FTP in areasonable time, under normal condi-tions. Major improvements will have toawait the development of new storagemedia and of broadband technology. Thisalso raises the issue of how rapidly andwidely new storage media become avail-able throughout European institutes. Bynow it should be clear that the rate atwhich images can currently be dissemi-nated lags far behind the rate at whichthey are acquired.

In order to overcome this problem, thecommunity is urged to consider using

compressed images, which can lead toa significant reduction (as much as a fac-tor of 500) in the size of the images. How-ever, such compression schemes in-evitably involve some loss of informationand the scientific value of the resultingproducts depends on the specific appli-cation. Experiments conducted by theSurvey Team indicate that for astromet-ric applications very little is lost, even forvery large compression factors. In fact,the loss in accuracy is <~ 0.5 of pixel size,comparable to the estimated internal er-ror of the astrometric solution. Examplesof compressed images have been madeavailable in the most recent data release.With the advent of the WFI it seems thatthe combination of compressed imagesfor large fields and full resolution cut-outsfor small regions should be seriously con-sidered.

In summary, to a large extent the scopeof the Survey Team work and the rangeof data products have been dictated bythe recognition of the difficulties in ex-porting full-resolution images, as result-ing from the direct experience with theEIS project, and the expected trend withthe WFI@2.2 and the ΩCam at the VLTSurvey Telescope. One needs to effi-ciently translate terabytes of imaging datainto megabytes of useful information inthe form of the various data productsmade available by the Survey Team (sin-gle and multi-band object catalogues, tar-get lists, cut-outs, postage stamps). In thiscontext it is important to emphasise oncemore that the primary goal of the ESOpublic surveys has been to support VLTobservations and not to be an end in it-self.

6. Future Plans

The Survey Team and the ESOArchive Group continue to explore oth-er ways to provide external users withmore flexibility in their use of the sur-vey data. In the short term it will bepossible to use the SKYCAT interfaceto run SExtractor on image cut-outsextracted from WFI mosaics, of suf-ficient size (12′ × 12′) to include thefield-of-view of the FORSes and of in-dividual VIMOS and NIRMOS cameras,regardless of their respective orienta-tions on the sky. In the long term, effortswill be made to implement an object-oriented database which will allow usersto fully explore the multi-dimensional po-sition, magnitude and colour space de-fined by the objects extracted from the im-aging surveys. The development ofsuch a database, where astronomical ob-jects extracted from survey data can bestored, updated and associated with allthe information available from internal andexternal sources, and of tools for search-ing special categories of objects is an es-sential element for the efficient use ofpublic survey data, as well as of all oth-er imaging data that will also becomepublicly available after the proprietaryperiod.

36

7. Conclusion

Besides the data, arguably one of themost important contributions of the EISproject has been to foster a collaborationbetween different European groups to de-velop a comprehensive imaging dataanalysis pipeline for the timely and fullexploitation of the data that will becomeavailable from wide-field imagers. EIS hasboth benefited from and contributed tothis ongoing effort, by capitalising on theavailable expertise and by providing dataand a realistic framework and time-tableto test concepts, designs and imple-mentations of pipelines, database archi-tecture and data archiving and distribu-tion schemes. Since the first feasibilitystudy carried out in 1995, over 30 peo-ple have contributed to the EIS effortamong EIS visitors, WG members andESO staff and fellows.

EIS has also provided a stimulating en-vironment for the introduction of studentsto modern techniques of data reductionand analysis and the means to dissem-inate their experience throughout thecommunity. This has been achieved bythe EIS Visitor Programme which hassponsored long-term visits, ranging from6 months to over 2 years, of 15 people

over the past two years as well as short-term visits of leading experts acting asconsultants. While about half of the long-term visitors have already returned totheir home institutes, a large fraction con-tinues to contribute to EIS in differentways.

As stated over two years ago, EIS wasconceived as an experimental project notonly to provide targets for the first yearof VLT (see da Costa et al. in this issueof The Messenger) but also to explore thefeasibility of conducting public surveys us-ing dedicated wide-field imaging tele-scopes. Indeed, with its 9 × 9 arcmin fieldof view, EMMI@NTT was far from beingthe ideal instrument for wide area sur-veys. Yet, without the experience gainedwith EIS it would now be much more dif-ficult to efficiently deal with the WFI@2.2data, and in the medium term perspec-tive with the almost 10 times larger dataflow from the ΩCam@VST. EIS hasserved to prototype and build a frame-work for this type of long-range pro-gramme, essential for the whole ESOcommunity to benefit from such tele-scopes, and to do so to the largest pos-sible extent independently of the effec-tive resources available at the individualinstitutes.

The new policy outlined here is meantto bring the data promptly to the users,while allowing the Survey Team moretime to deliver the final products. Budgetconstraints obviously limit the size of theSurvey Team, which cannot grow at thesame pace as the data flow. Hopefully,this new model will stimulate new usesand checks of the data. Experience hasshown that, regardless of the care tak-en, handling large volumes of data is adifficult task and only by the exhaustiveuse of the data in different applicationscan problems be detected. Therefore, theparticipation and the feedback from theusers is essential to ensure the quality ofthe data. Users are also encouraged toperiodically visit the EIS home page tomonitor any new version of the variousreleases, and report any possible error.

8. Acknowledgements

We would like to thank the EIS Teamand the ESO Archive Group for their con-tinuing efforts to explore ways of im-proving the accessibility of the data by ex-ternal users, as well as the NTT and 2.2-m Telescope Teams for their cooperation.

ESO IMAGING SURVEY: Past Activities and Future ProspectsL. DA COSTA1, S. ARNOUTS1, C. BENOIST1, E. DEUL1,2, R. HOOK3, Y. -S. KIM1, M. NONINO1,4, E. PANCINO1,5, R. RENGELINK1,2, R. SLIJKHUIS1, A. WICENEC1, S. ZAGGIA1

1European Southern Observatory, Garching b. München, Germany2Leiden Observatory, Leiden, The Netherlands3Space Telescope – European Coordinating Facility, Garching b. München, Germany4Osservatorio Astronomico di Trieste, Italy5Dipartimento di Astronomia, Universtà di Padova, Italy

1. Introduction

The ESO Imaging Survey (EIS) proj-ect is an ongoing effort to carry out pub-lic imaging surveys in support of VLT pro-grammes. Background information on theoriginal and future goals of the pro-gramme can be found in Renzini & daCosta (1997) and in a companion articleby Renzini & da Costa in this issue of TheMessenger.

The first phase of the project, whichstarted in July 1997, consisted of a mod-erately deep, large-area survey (EIS-WIDE) and a deep optical/infrared sur-vey (EIS-DEEP), with the observationsbeing conducted at the NTT. EIS hasrecently reached another milestone withthe completion of a Pilot Survey us-ing the Wide-Field Imager (WFI) mount-

ed on the MPG/ESO 2.2-m telescope atLa Silla.

The purpose of this contribution is tobriefly review the results of the originalEIS and to give an update of the resultsobtained from the observations carriedout as part of the Pilot Survey. The on-going work to develop an advancedpipeline for handling data from large CCDmosaics and of facilities to make accessto the data products easier to externalusers are also discussed.

2. EIS-WIDE

This part of the survey consistedof a mosaic of overlapping EMMI-NTTframes (9 × 9 arcmin) with each po-sition on the sky being sampled twice

for a total integration time of 300 sec.The observations were carried out i nthe period July 97–March 98 and cov-ered four patches of the sky southof δ = –20° and in the right ascensionrange 22h < α < 10h, thus producing tar-gets suitable for the VLT almost year-round. The locations of these fields areshown in Figure 1 and their approxi-mate centres are given in Table 1, whichalso lists the area covered in eachpassband. The typical depth of the sur-vey, as measured from the co-added mo-saics and with the magnitudes ex-pressed in the AB system, is given inTable 2 which lists: in column (1) the pass-band; in column (2) the median seeing;in column (3) the 1σ limiting isophote;in column (4) the 80% completenessmagnitude (e.g., Nonino et al. 1999); in

E-mail: arenzini@eso.org

37

column (5) the 3σ limiting magnitudewithin an aperture twice the medianFWHM; and in column (6) the originallyproposed magnitude limits (Renzini &da Costa 1997) expressed as in column(5). Note that these numbers are rep-resentative values since, as pointed outin the original papers (Nonino et al.1999, Prandoni et al. 1999, Benoist et al.1999), the observing conditions variedsignificantly during the course of the ob-servations. Comparison of Tables 1 and2 to the original goals of the survey(Renzini & da Costa 1997, and revisedin da Costa 1997) shows that, whilethe desired limiting magnitudes were

reached, not all the objectives weremet. Among the most important omis-sions were the observations of a com-parable area in V and of patch B in theU-band. The incompleteness was due toa combination of poor weather conditionsand unexpected overheads in the ob-servations. A total of about 42 nights wereassigned to this part of the project out ofwhich about 30% were lost due to theweather alone, compromising especial-ly the coverage of EIS patches A and B.At the time, the Working Group (WG) incharge of overseeing the survey decid-ed to postpone the U-band observationsof patch B and the V-band observations

of the other patches, giving priority tothe I-band survey.

Despite the smaller total area and thevarying quality of the data, it was possi-ble to meet most of the primary sciencegoals of the survey. Particularly suc-cessful has been the I-band survey cov-ering ~ 17 square degrees to a limitingmagnitude IAB ~ 23.5, currently the largestsurvey of its kind in the southern hemi-sphere. In addition the survey provided~ 3 square degrees in V I and ~ 1.5square degrees in B V I. From these data,samples of distant clusters of galaxiesand quasar candidates were compiled(Olsen et al. l999a,b, Scodeggio et al.1999, Zaggia et al. 1999) and used in sev-eral follow-up observations by differentteams. The data, comprising astromet-rically and photometrically calibratedpixel maps, derived catalogues and tar-get lists, were made publicly available inMarch and July 1998 (Nonino et al. 1999,Prandoni et al. 1999, Benoist et al. 1999),meeting the planned deadlines. To easethe access to the data, a web interfacewas created to enable external users torequest survey products, to extract imagecutouts and to examine the selected tar-gets on-line.

The accuracy of the relative astrome-try is ~ 0.03 arcsec, making the co-ordinates sufficiently accurate for thepreparation of multi-object spectroscop-ic observations. The internal accuracy ofthe photometric zero-point has been es-timated to be better than 0.1 mag, con-sistent with the results obtained fromcomparisons with other data (e.g., DE-NIS). However, as more data in differentpassbands became available, gradientsof the photometric zero-point, on a degreescale, have been noticed and are cur-rently being investigated. The results ofthese investigations will be reported indue time.

3. EIS-DEEP

This part of the survey consisted ofdeep optical/infrared observations usingSUSI2 and SOFI at the NTT of the HDF-S and AXAF fields shown in Figure 1. Theoriginal observations were carried outin the period August–November 1998,with all the data being publicly releasedDecember 10–12, 1998 (da Costa et al.1999b, Rengelink et al. 1999). The re-lease included fully processed and cali-brated images, single-passband andcolour catalogues, and a tentative list ofU -and B dropouts. The centres of thedifferent fields observed with SUSI2and SOFI, chosen to provide someoverlap, are given in Table 3. The dataavailable for these fields are sum-marised in Tables 4 and 5 which pro-vide, for each field, the following infor-mation: in column (1) the passbandsavailable; in column (2) the total inte-gration time; in column (3) the number offrames; in column (4) the final FWHM asmeasured on the co-added image; in col-umn (5) the 1σ limiting isophote; in col-

Figure 1: Sky projection showing the location of the original EIS-WIDE patches (red), the EIS-DEEP AXAF and HDF-S fields (blue), and the stellar fields of the Pilot Survey (orange). Alsoin blue are the new fields selected for the Deep Public Survey to be carried out in periods 64–67.Note that one of the regions selected coincides with patch A and the other with the EIS-DEEPAXAF field (see Fig. 2). The third field (Deep-3) has been selected in a region with lower ab-sorption and less crowding than patch D. The right ascension range was chosen to distributeas much as possible the observations throughout the year.

Table 1: EIS-WIDE: Sky Coverage (square degrees).

Patch α2000 δ2000 B V I

A 22 43 03 –39 58 30 – 1.2 3.2B 00 48 22 –29 31 48 1.5 1.5 1.6C 05 38 24 –23 51 54 – – 6.0D 09 51 40 –21 00 00 – – 6.0

Total 1.5 2.7 16.8

Table 2: EIS-WIDE: Limiting Magnitudes (AB system).

Filter FWHM µlim 80% 3σ 3σ (proposed)(arcsec) (mag arcsec–2) (2 × FWHM) (2 × FWHM)

B 1.0 26.5 24.6 24.2 24.2V 0.8 26.2 24.4 24.0 24.5I 0.8 25.5 23.7 23.7 24.0

38

umn (6) the limiting magnitude of a 3σ de-tection within an aperture twice theFWHM. In Table 4 the last column liststhe originally proposed limiting magni-tudes for EIS-DEEP, expressed as thosein column (7), assuming a 1 arcsec see-ing (da Costa & Renzini 1998). To helpvisualise the fields and the different setsof data available, Figure 2 shows aschematic view of the EIS-DEEP obser-vations of the HDF-S and AXAF fields,with the colours indicating the com-pleteness of the multi-colour data. In thecase of HDF-S the location of theWFPC2, STIS and NIC3 HST fields andtheir relation to the location of the EIS-DEEP fields are also shown.

The highest priority was given to theobservations of the HDF2 (WFPC2) fieldfor which data in eight passbands areavailable, by and large reaching the re-quired magnitude limits, with the notableexception of the I-band. Note that in thisfield the J and Ks observations reach half-magnitude fainter limits than those in theother fields, as suggested by the WG (daCosta & Renzini 1998). Even though itwas not possible to complete the obser-vations in all the passbands for all thefields, the data gathered provide ~ 150square arcmin in J Ks and ~ 80 squarearcmin in UBVRI. When combined, theoptical/infrared deep observations pro-vide a coverage of 15 and 25 square ar-cmin in eight and seven passbands, re-spectively. These numbers include the re-cently released data for the HDF-S STISand NICMOS fields from observationsconducted in July 1999 (see EIS homepage). Unfortunately, due to bad weath-er, no new optical data were obtained dur-ing the four half nights scheduled in July1999 at the NTT, and the R I images ofHDF1, taken in 1998, proved to be un-suitable due to poor seeing and stray

Figure 2: Schematic view of the various data sets available for the HDF-S and AXAF fields from optical/infrared observations conducted withSUSI2 and SOFI at NTT. The different colours denote the completeness of the multi-colour data.

Field αSUSI2 δSUSI2 αSOFI δSOFI

HDF1 22 33 29 –60 33 50 22 33 32 –60 33 30HDF2 22 32 42 –60 33 50 22 33 00 –60 33 30HDF3 – – 22 33 00 –60 37 59

AXAF1 03 32 16 –27 46 00 03 32 17 –27 46 10AXAF2 03 32 38 –27 46 00 03 32 37 –27 46 10AXAF3 – – 03 32 17 –27 50 35AXAF4 – – 03 32 37 –27 50 35

Table 3: EIS-DEEP: SUSI2 and SOFI Pointings (J2000.0).

Table 4: EIS-DEEP: HDF-S Observations (magnitudes in the AB system).

Filter ttotal Nƒ FWHM µlim 3σ 3σ σ (proposed)(sec) (arcsec) (mag arcsec–2) (2 × FWHM) (2 × FWHM)

HDF1

J 10800 180 1.37 25.7 23.6 24.2H 3600 60 0.91 25.1 23.4 23.7Ks 10800 180 0.90 24.9 23.1 23.1

HDF2

U 17800 22 1.00 28.2 26.3 26.6B 6600 22 0.84 28.2 26.5 25.7V 12250 49 1.27 28.3 26.5 25.8R 5500 22 1.05 27.3 25.4 26.0I 8800 44 1.11 27.0 25.0 26.2J 10800 180 0.90 26.2 24.4 24.7H 7200 120 0.85 24.9 23.2 24.2Ks 18000 300 0.96 25.5 23.7 23.6

HDF3

J 5820 97 1.18 25.5 23.5 24.2Ks 7440 124 0.86 24.9 23.2 23.1

39

light. The incompleteness of the EIS-DEEP data relative to the original goalsis primarily due to the optical observationswhich were affected by both poor weath-er and a variety of technical problems in1998. An attempt to complete the opticalcoverage of the HDF1 (STIS) field will bemade in June–July 2000, while the AXAFfield has been chosen to be one of thethree fields selected for the new DeepPublic Survey as shown in Figure 1 (seeRenzini & da Costa 1999).

4. WFI@2.2 Pilot Survey

With the announcement of the com-missioning of the WFI@2.2m, the WGrecommended the undertaking of a pilotmulti-colour survey over the area alreadycovered in I – band, including the miss-ing U-band observations of patch B. Themotivation was to complete the originalgoals of EIS-WIDE and to steer the up-grade of the pipeline to cope with CCD-mosaics and the corresponding order ofmagnitude increase in the volume of data.

Even though originally proposed to becarried out in dark time over the observ-ing Periods 62 and 63, some adjustmentshad to be made to accommodate the WFIcommissioning period. In the end, a to-tal of 30 hours was used for the PilotSurvey observations during the com-missioning period in January 1999, and

a total of 8.5 nights in the months of April,May and September. The observationsconsisted of five 4-minute dithered ex-posures, which allowed the coverage of0.5 square degrees/hour in a single pass-band including overheads. A total of 18square degrees were observed, cover-ing patches C and D in B and BV, re-spectively. The B-filter was used first be-cause it was the only broad-band filteravailable during commissioning. In mid-September new observations of Patch-es A and B were conducted in V and Ubut the data have notyet been reduced atthe time of writing.Table 6 shows the cur-rent status of the EIS-WIDE patches aftercombining the dataavailable from EMMIat the NTT with thosefrom the Pilot Survey,including the Septem-ber 1999 run. These data together withthe 17 hours scheduled in service modefor Period 64 should allow the com-pletion of about 1.5 square degrees inUBVI and 15 square degrees in BVI.These data should greatly broaden therange of applications that can bene-fit from the EIS public survey to producesamples for follow-up observations withthe VLT. Preliminary estimates of themedian seeing and limiting magnitudes

reached in B and V from the data gath-ered so far using the WFI are given inTable 7.

Some of the scheduled time was un-suitable for the observation of the patch-es, either because the EIS-WIDE fields

Filter ttotal Nƒ FWHM µlim 3σ(sec) (arcsec) (mag arcsec–2) (2 × FWHM)

AXAF1

U 17000 21 0.90 28.0 26.1B 6600 22 1.10 28.0 26.4V 5500 22 0.88 27.7 25.7R 5500 22 0.89 27.5 25.9I 12600 21 1.31 27.0 24.9J 10800 180 0.99 25.7 23.8Ks 10800 180 0.95 24.9 23.2

AXAF2

U 13000 16 0.90 27.8 26.0B 5400 18 1.10 27.9 25.8V 5500 22 0.88 27.7 26.1R 5500 22 0.89 27.5 25.7J 10800 180 0.90 25.8 24.1Ks 10800 180 0.90 24.8 23.1

AXAF3

J 10800 180 0.90 25.4 23.7Ks 10800 180 1.00 24.7 22.9

AXAF4

J 10800 180 0.70 25.6 24.1Ks 7200 120 1.00 24.4 22.6

Table 5: EIS:DEEP: AXAF Observations (magnitudes in the AB system).

Table 6: WFI Pilot Survey: Sky Coverage(square degrees).

Patch U B V I

A – 1.0 1.2 3.2B 1.25 1.5 1.5 1.6C – 6.0 6.0 6.0D – 6.0 6.0 6.0

Total 1.25 14.5 14.7 16.8

Table 7: WFI Pilot Survey: Limiting Magnitudes (AB system).

Band FWHM µlim 80% 3σ(arcsec) (mag arcsec–2) (2 × FWHM)

B 0.9 26.8 24.5 25.0V 1.0 26.4 24.1 24.6

Figure 3: Vector representation of the patternof the PSF anisotropy and the amplitudeof its components for each chip of the CCDmosaic, as measured on a V-band image ob-tained from a 20-min exposure.

Figure 4: Colour image of a section of EISpatch D combining BV images of WFI withI-images of EMMI. E

42

were not visible or because the nightswere too bright. In these periods, a setof selected stellar fields were observedas a test case for future preparatory pro-grammes for GIRAFFE and UVES usingthe fibre-positioner FLAMES, thus an-ticipating part of the Public Survey ob-servations now planned for P64–67, theso-called Pre-Flames Survey recom-mended by the WG in their meeting inMarch 1999 (see Renzini & da Costa1999). The selected fields (Figure 1) in-

upgraded EIS-WFI pipeline, and publiclyreleased on September 9, 1999.

Preliminary results show that the per-formance of the WFI@2.2 exceeds ex-pectations and demonstrate its compet-itiveness in the area of wide-field imag-ing surveys. Analysis of the imagesshows that the point-spread function isvery uniform across the entire field ofview and that PSF distortions are small~ 2%. This is illustrated in Figure 3where the spatial distribution of the po-

cluded Baade’s window, globular (GC)and open clusters (OC). A total of 22 fields(5.5 square degrees) given in Table 8were observed in different passbands. Afull description of the available data canbe found on the web.

Altogether the observations from Jan-uary to April 1999 yielded 400 Gby of rawdata at a rate of about 40 Gby/night (0.5Mb/sec), including calibration frames(standard stars, dome and sky flats). Allthe data have been processed, using the

Figure 5: A 16.1 × 9.4 arcmin section of the co-added I-band image, rebinned to a scale of 0.5 arcsec/pixel, of the cluster NGC 6558 locatedin Baade's window. The regions devoid of objects are associated with the interchip gaps.

43

larisation vector, representing the am-plitude and direction of the PSF distor-tions, and the distribution of the ampli-tude of its two components are shownfor each CCD chip of the WFI camera.

To illustrate the results obtained bythe pipeline, Figure 4 shows a colourcomposite of a section of patch D (23 ×19 arcmin) constructed from the combi-nation of WFI B and V mosaics with theEMMI I-band mosaic. Since one of theprimary goals of Pilot Survey is to com-plement the I-band observations, theco-addition of the WFI images was car-ried out on the EMMI pixel scale, about10% larger than that of the WFI. Bydithering the WFI exposures and by co-adding the available frames using theweight maps, the final coadded mosaicshows no noticeable evidence of the in-ter-chip gaps, with the least sampledregions having an exposure time about60% of that obtained in the best sam-pled regions. Another example can befound in Figure 5, where the co-addedB-band image of the globular clusterNGC 6558, observed with 0.6 arcsecseeing, is shown. Note that in this caseonly two images have been co-addedand the regions devoid of stars are theremains of the interchip gaps.

The co-addition is carried out afterthe astrometric calibration of each inputframe. Figure 6 shows the distributionof the differences in the coordinates of700,000 pairs of stars in patch D fromindependent astrometric solutions ob-tained for different WFI V-band frames.The rms of the distribution is ~ 0.04 arc-sec, an estimate of the internal accura-cy of the astrometric solution. This re-

sult shows that the position of the tar-gets extracted from the survey data canbe directly used to position the slits andfibers of VLT spectrographs such asFORS1, FORS2, VIMOS, NIRMOS and

FLAMES. However, even though a sat-isfactory astrometric solution was de-termined for the stellar field shown inFigure 5, problems are still found in cal-ibrating crowded fields. Procedures toastrometrically calibrate these fields, in-cluding proper modelling of the opticaldistortions, are still being investigated.These issues as well as a full descrip-tion of the observations, data reductionand derived products will be presentedin forthcoming papers.

5. Data Handling

One of the most demanding tasks ofthe EIS project has been the develop-ment of a data reduction pipeline inparallel to the observations which haveused four different instruments (EMMI,SUSI2, SOFI and WFI@2.2m) in a pe-riod of two years. Adaptation of the soft-ware to this rapidly changing environ-ment has been by far the most chal-lenging aspect of the project, which willhopefully subside in the next couple ofyears. This not only impacts the devel-opment effort but also the data reduc-tion and verification which depend onthe stability of the pipeline.

The original EIS pipeline was devel-oped over a period of one year, startingin April 1997, by assembling existingsoftware (IRAF, SExtractor, LDAC andDrizzle) adapted to the specific needs(detector and observing strategy) of theEIS-WIDE survey. The task was facili-tated because most tools were devel-

Figure 6: Distribution of the differences in the coordinates of 700,000 pairs of stars in patch Dobtained from the independent astrometric calibration of different WFI frames. The solid con-tour corresponds to the half-maximum of the distribution, encompassing 62% of the points.

Object Name Object Type α1950 δ1950 Filters

Baade Window (1) Bulge 17 55 28 –28 45 20 B, V, I, zBaade Window (2) Bulge 17 56 04 –29 13 14 B, V, I, zBaade Window (3) Bulge 18 00 00 –29 52 13 B, V, IBaade Window (4) Bulge 18 05 33 –31 27 04 B, V, IBaade Window (5) Bulge 18 07 02 –31 46 27 V, IBaade Window (6) Bulge 18 15 10 –33 58 38 V, I

NGC 6218 GC 16 47 14 –01 56 52 B, V, INGC 6397 GC 17 40 41 –53 40 25 B, V, INGC 6544 GC 18 07 20 –24 59 51 B. V, INGC 6809 GC 19 39 59 –30 57 44 B, V, INGC 7089 GC 21 33 29 –00 49 23 B, V, INGC 6752 GC 19 10 51 –59 58 55 B, V, INGC 6541 GC 18 08 02 –43 42 20 B, V, INGC 6681 GC 18 43 12 –32 17 31 B, V, INGC 6656 GC 18 36 24 –23 54 12 B, V, INGC 6717 GC 18 55 06 –22 42 03 B, V, INGC 6723 GC 18 59 33 –36 37 54 B, V, I

IC 4651 OC 17 24 42 –49 57 00 B, V, INGC 6405 OC 17 36 48 –32 11 00 B, V, INGC 6208 OC 16 45 30 –53 4400 B, V, INGC 5822 OC 15 01 30 –54 09 00 B, V, INGC 6705 OC 18 51 06 –06 16 00 B, V, I

Table 8: WFI Pilot Survey: Stellar Fields.

44

oped by astronomers in the MemberStates who participated in this develop-ment. At the time, the data volume wasrelatively small, which greatly facilitatedthe handling and processing of thedata. Most of the work focused on theinterfacing of the different processingblocks and in producing scripts to con-trol the processing essentially in batchmode. A lot of effort was also spent inthe design and implementation of adatabase which was primarily used forbookkeeping.

After the first EIS data release inMarch 1998, an upgrade was carriedout to incorporate ECLIPSE (Devillard,Jung & Cuby 1999) for the reduction ofSOFI data and to handle the two-CCDmosaic of SUSI2. Even though time-consuming, implementing these add-ons did not require major changes inthe design of the pipeline.

The advent of the WFI, on the otherhand, required a major overhaul drivenby the sheer volume of the data. Newhardware was required as well as a ma-jor re-design of the pipeline architec-ture. Over the past year new machineshave been added, the number ofprocessors available in each server in-creased, disk and memory capacitiesexpanded and DLT tape units installed.In parallel, during the past six monthssignificant modifications have beenmade to the pipeline software, eventhough the basic algorithms remain es-sentially the same. A more sophisticat-ed control system has been developedto administer the resources of the sys-tem, to retrieve/store images in the tapelibrary, to supervise the reduction proc-ess and to communicate with an ex-panded implementation of the databasein a variety of ways. The new pipelineallows the reduction of both single CCD(SOFI and other single-chip instru-ments) and mosaics (e.g., SUSI2, WFI)and parts of the code have been opti-mised and parallelised to increase theoverall throughput to match the expect-ed input data rate. Finally, the pipelinehas been made more user-friendly anduniform in order to facilitate its mainte-nance and integrity, to prevent disconti-nuities caused by the constant changesof Survey Team members and to allowit to be run by other interested users.

Broadly speaking the pipeline con-sists of four logical blocks:

1. Image Processing – this moduleperforms the usual image processingsteps removing instrument signaturesand astrometrically and photometricallycalibrating each individual WFI frame. Itthen performs the co-addition of dith-ered WFI frames, producing a singleFITS image from which object cata-logues are extracted. These are the ba-sic products produced automatically bythe pipeline which are then transferredto the archive for quick-look releases.

2. Image Mosaicing – this moduledeals with the production of image mo-saics, which implies adjustments of the

photometric zero-point, of final objectcatalogues, extracted from the co-added images using the most recentand extensively tested version ofSExtractor, and of colour cataloguesconstructed either by the association ofobjects detected in each passband orby the use of a reference image (e.g.,χ2-image).

3. Quality control – this module car-ries out the quality control of the data. Itmonitors image distortions, computesthe rms distribution of the residuals ofthe astrometric solution, compares theastrometry of the objects extracted indifferent passbands and checks thespatial uniformity of the photometriczero-point. It also performs a prelimi-nary scientific evaluation of the data. Itculls object catalogues, computes sim-ple statistics such as number counts,colour-colour diagrams, colour distribu-tions and the two-point angular correla-tion function for objects classified asgalaxies. The results are comparedwith other available data to check thephotometric zero-points, the star-galaxy classification and uniformity ofthe object catalogues.

4. Target Selection – this module isused to produce a variety of targetlists. It also produces image postagestamps (typically 2 × 2 arcmin) in differ-ent passbands of the selected objectsfor visual inspection. In time, other fa-cilities are expected to be added, in par-ticular for searches for variable andproper-motion objects. Among manypossible applications, a major goal is todefine secondary-standards in the sur-roundings of the traditional Landoltfields, essential to monitor the zero-point of the different chips in the CCD-mosaic.

Even though the backbone of thepipeline is in place, a lot of work re-mains to be done, especially on thecontrol system, in order to allow theprocess to be fully automated. Timebenchmarks of different parts of thepipeline demonstrate that most of themodules are operating at a rate ~ 0.2Mpixels/sec. For instance, it currentlytakes about 7 hours to process 20 WFIframes and produce the results re-quired for a quick-look release of asquare degree worth of data. This cor-responds to 2 hours of observations inthe wide-angle-survey mode adoptedby the Pilot Survey. Even though pre-liminary, these numbers show that thecurrent pipeline is capable of copingwith the volume of data being producedby the survey. Furthermore, the pipelineis not yet fully implemented and testedand there still is considerable room forimproving the system throughput. Thefinal consolidation of the softwareshould last until the end of 1999, with afinal target date of April 1, 2000. A de-tailed description of the available soft-ware will be presented in the “EIS-WFIPipeline Primer” currently in prepara-tion by the Survey Team.

6. Data Access

Unlike radio surveys and other publicsurveys carried out in space, ground-based optical imaging surveys requireconsiderably more information to char-acterise the data available. An arduousand time-consuming task, given thetime pressure and limited personnel,has been to establish adequate chan-nels of communication between theSurvey Team and the community atlarge to describe the progress of theproject and to provide the necessary in-formation about the observations andthe various data products available. Inthis first phase this was accomplishedby posting as much information as pos-sible on the web, by periodically re-viewing the status of the project in TheMessenger (da Costa 1997, da Costaet al. 1998, da Costa & Renzini 1998)and by submitting and circulating pa-pers with each data release.

During the first phase, requests forlarge volumes of data from differentgroups have been processed by the ESOArchive Group and about 100 Gby of datahave been delivered in various forms. Therequests originated from 11 countries, in-cluding all of the ESO member states. Ac-cording to current guidelines, cataloguesare available world-wide, but images arerestricted to ESO member states, withthe exception of those of the HDF-Sfields. Table 9 shows a summary of datarequests listing the product and thenumber of requests up to September1999. The access to the EIS home pagehas increased by a factor of 5 overthe period of the survey. It currentlyaverages about 30,000 hits/month,reaching peaks twice as high after adata release. About one-third of thesehits are from research institutes in Eu-rope and abroad, constantly monitor-ing the progress of the survey, while therest are from the general public. In thisrespect, it is worth noting the popularityof the EIS image gallery, with about35,000 images being retrieved from theweb. A total of about 4500 data requestshave been processed so far, about halfof which asked for derived products suchas target lists and catalogues. By ex-panding the scope of the observations toother areas of application and by im-

Product Number of Requests

Bulk Data 50Cutouts 1128Mini-images 319Target Lists 737EIS-Deep catalogues 1440HDF-S images 821TOTAL 4495Gallery Images 34538

Table 9: Summary of Data Requests (Sep-tember 1999).

45

proving the interface with the public, thesenumbers are likely to increase signifi-cantly.

7. Future Developments

With the recent approval of a two-yearLarge Programme to conduct a PublicSurvey (see Renzini & da Costa 1999)with the WFI and SOFI, it is now possi-ble to envision a more stable operationleading to the production of more stan-dard data sets. In this framework, newways of making the data accessible to thecommunity will be explored. Quick-lookreleases of the data, distribution of com-pressed data and the implementation ofnew on-line features are being consid-ered as described by Renzini & da Cos-ta (1999). These, together with the oth-er products already available should pro-vide users with a large range of optionsand greater flexibility to explore the avail-able data.

A considerable effort is also underwayto implement a flexible design to store theobjects extracted from the public surveyimages in a database which will allowusers to directly select and display objectsaccording to their positions in the sky,their magnitudes and their colours with-out any reference to a specific catalogue.The database will be a dynamic entity thatwill grow in time as information from newobservations of different sky regionsor repeated observations of the same ob-jects are continuously added to it. The im-plementation of such a database anddatabase tools where astronomical ob-jects extracted from survey data can bestored, updated and associated with allthe information available from internal andexternal sources is an essential elementfor the successful use of the Public Sur-

vey data. The need for such databasestranscends EIS and several groups in Eu-rope involved in wide-field imaging sur-veys are tackling the same problem (e.g.Terapix, ΩCam). These various groupsare exchanging ideas, and a preliminaryimplementation has recently been test-ed by the EIS Survey Team, in collabo-ration with the ESO Archive Group, us-ing the Objectivity database. As the workon the pipeline is consolidated, the at-tention of the software developers in theSurvey Team will focus on this criticalarea which is a key element for the effi-cient scientific exploitation of the surveydata.

8. Conclusions

After two years of observations, soft-ware development, data reduction anddistribution, nearly all the data from theongoing Public Surveys (EIS-WIDE,EIS-DEEP and Pilot Survey) are publiclyavailable. The only exceptions are thedata gathered from observations con-ducted with the WFI in September 1999and some data taken with the narrow-band z filter for which no proper calibra-tion is currently available. The EIS ex-perience has been extremely valuable forthe new Public Survey to be conductedin Periods 64–67, as important lessonshave been learned in handling large vol-umes of data, in setting up rules for au-tomatically processing images andrecipes for data quality control, and inways and means of exporting the datato the users in an adequate form and ina timely fashion. With the accumulatedexperience from four previous releasesand with the major development phasedrawing to a close, new ways of makingthe data available are being considered

which will hopefully provide interestedusers with a broader range of productsas well as more flexibility and thuscontribute to the scientific exploitationof the VLT.

9. Acknowledgements

We would like to thank A. Renzini forhis effort in steering the imaging survey,the WG members for their scientific in-put and former EIS team members fortheir contribution over the years. Specialthanks to E. Bertin for addressing the in-numerable questions and requests posedby the EIS team and for introducing newfeatures to SExtractor as needed and toT. Erben for his help in the analysis of thePSF.

ReferencesBenoist et al. 1999, A&A, 346, 58.da Costa, L., 1997, The Messenger, 88, 34.da Costa. L. et al., 1998, The Messenger, 91,

49.da Costa, L. & Renzini, A., 1998, The Mes-

senger, 92, 40.da Costa, L. et al., l999b, A&A, in press (as-

tro-ph/9812105).Devillard, N., Jung, Y. & Cuby, J.G., 1999, The

Messenger, 95, 5.Nonino, M. et al. 1999, A&A Supp., 137, 51.Olsen et al. 1999a, A&A, 345, 681.Olsen et al. 1999b, A&A, 345, 363.Prandoni et al. 1999, A&A, 345, 448.Renzini, A. & da Costa, L., 1997, The Mes-

senger, 87, 23.Renzini, A. & da Costa, L., 1999, in this issue

of The Messenger.Rengelink et al., 1999, A&A, submitted (astro-

ph/9812190).Scodeggio et al. 1999, A&A Supp., 137, 83.Zaggia, S. et al. 1999, A&A Supp., 137, 75.

E-mail: ldacosta

46

1. Abstract

We have studied NGC 3603, themost massive visible HII region in theGalaxy, with VLT/ISAAC in the near-in-frared (NIR) Js, H, and Ks-bands andHST/WFPC2 at Hα and [N II] wave-lengths. In this Messenger article wedescribe the data analysis and somefirst results from both our complemen-tary observations.

Our HST/WFPC2 gave us an un-precedented high-resolution view ofthe interstellar medium in the giant HIIregion and its ionisation structure.Among the findings are two giganticgas columns (similar to the famous“elephant trunks” in M16) and three pro-plyd-like structures. The emission neb-ulae are clearly resolved; all three neb-ulae are tadpole shaped, with the brightionisation front at the head facing thecentral cluster and a fainter ionisationfront around the tail pointing away fromthe cluster. Typical sizes are 6000 A.U.× 20,000 A.U. The nebulae share theoverall morphology of the proplyds(“PROto PLanetarY DiskS”) in Orion,but are 20 to 30 times larger in size.

The VLT observations are the mostsensitive near-infrared observationsmade to date of a dense starburstregion, allowing us to investigate with un-precedented quality its low-mass stellarpopulation. Our sensitivity limit to starsdetected in all three bands correspondsto 0.1 M0 for a pre-main sequence starof age 0.7 Myr. Our observations clear-ly show that sub-solar-mass stars downto at least 0.1 M0 do form in massive star-bursts.

2. Introduction

NGC 3603 is located in the Carina spi-ral arm (RA = 11h, DEC = –61°) at a dis-tance of 6–7 kpc. It is the only massive,Galactic HII region whose ionising cen-tral cluster can be studied at optical wave-lengths, due to only moderate (mainlyforeground) extinction of AV = 4–5 mag(Moffat 1983; Melnick et al. 1989). TheOB stars (> 10 M0) and stars with the

spectral signatures of Wolf-Rayet (W-R)stars contribute more than 2000 M0 tothe cluster mass. Normally, W-R stars areevolved supergiants that have long leftthe main sequence and have ages of 3–5Myr. In NGC 3603, however, the W-Rstars also show hydrogen absorptionlines in addition to the typical W-R fea-tures. It is believed that these stars arestill main-sequence, core hydrogen-burning stars that are so massive and soclose to the Eddington limit that they areloosing their outer envelopes through fastwinds, and thus resemble evolved W-Rstars. In comparison to the Orion Trapez-ium system, NGC 3603 with its more than50 O and W-R stars producing a Lymancontinuum flux of 1051 s–1 (Kennicutt1984; Drissen et al. 1995) has about 100times the ionising power of the Trapezi-um cluster.

With a bolometric luminosity Lbol > 107

L0, NGC 3603 has about 10% of the lu-minosity of 30 Doradus and looks in manyrespects very similar to its stellar coreR136 (Brandl et al. 1996). In fact, it hasbeen called a Galactic clone of R 136 butwithout the massive surrounding clusterhalo (Moffat, Drissen & Shara 1994). Inmany ways NGC 3603 and R136 can beregarded as representative buildingblocks of more distant and luminous star-burst galaxies (Brandl, Brandner & Zin-necker 1999, and references therein). Sothe best way to study collective star for-mation in a violent environment is by ex-amining these regions on a star-by-starbasis. Until recently, observations of theentire stellar populations were limited tothe massive stars by sensitivity or wave-length restrictions. The formation of low-mass stars in starburst regions is of par-ticular interest. Fundamental questionsthat arise in this context are: Does theslope of the IMF vary on small scales?Do low-mass stars in a starburst eventform together with the most massive starsor do they form at different times or ondifferent timescales? And finally, onemight even ask if low-mass stars form atall in such environments?

In addition, NGC 3603 is an ideal lab-oratory to study individual (!) star forma-tion. Some young stars in the vicinity ofmassive stars are surrounded by par-tially ionised circumstellar clouds withcometary shape, so-called protoplanetarydisks (“Proplyds”), ionised from the out-

side (Churchwell et al. 1987; O’Dell et al.1993). FUV photons (13.6 eV > hν > 6eV) heat up the inside of the proplyd en-velope and lead to the dissociation of mol-ecules in the outer layers of the circum-stellar disk (Johnstone et al. 1998). Theresulting evaporation flow provides asteady supply of neutral atoms to the ion-isation front and leads to the developmentof a cometary tail (McCullough et al. 1995;Störzer & Hollenbach 1999). Until re-cently, only one other proplyd had beenfound outside the Orion nebula. It is lo-cated in the vicinity of the O7V star Her-schel 36 in the Lagoon Nebula (M8,Stecklum et al. 1998).

3. VLT/ISAAC Observations

We observed NGC 3603 through theJs = 1.16–1.32 µm, H = 1.50–1.80 µm,and Ks = 2.03–2.30 µm broadband filtersusing the NIR camera ISAAC on ANTU,the first VLT unit telescope. The obser-vations were made during the 4 nights ofApril 4–6 and 9, 1999, in service modewhen the optical seeing was equal to,or better than, 0.4″ in 1-minute expo-sures. Such seeing was essential foraccurate photometry in the crowdedcluster and increased our sensitivityto the faintest stars. The majority ofour data were taken under photometricconditions.

Our observing strategy was to usethe shortest possible frame times of1.77 seconds to keep the number ofsaturated stars to a minimum. However,due to the system’s excellent sensi-tivity, about two dozen of the brighteststars ended up being saturated. Never-theless, this does not impose a prob-lem to our study of the low-mass stars.Thirty-four short exposures were co-added to an effective one-minute expo-sure, the minimum time per pointingrequired to stabilise the telescope’s ac-tive optics control system. Between the1-minute pointings we moved the tele-scope by up to 20″ offsets in a randompattern. This approach has several ad-vantages:

• Enlargement of the observed field ofview (FOV) with maximum signal-to-noise(S/N) in the cluster centre.

• Reduction of residual images andother array artefacts, using the medianfiltering technique.

OBSERVERS WITH THE VLT

VLT/ISAAC and HST/WFPC2 Observations of NGC 3603B. BRANDL1, W. BRANDNER2, E.K. GREBEL3 AND H. ZINNECKER4*

1Cornell University, Ithaca; 2University of Hawaii, Honolulu; 3University of Washington at Seattle; 4Astrophysikalisches Institut Potsdam

*Our collaborators on the project described in thisarticle are Y.-H. Chu, H. Dottori, F. Eisenhauer, A.F.J.Moffat, F. Palla, S. Richling, H.W. Yorke and S. D.Points.

47

• Derivation ofthe “sky” from thetarget exposuresusing the medianfiltering technique.No additional timefor “blank” skyframes outside thecluster was re-quired.

The sky frameshave been comput-ed using between15 and 37 subse-quent exposuresper waveband andnight, and carefule y e - i n s p e c t i o nshowed that allsources have beenefficiently removedusing our modifiedmedian filteringtechnique which re-turns the lower 1/3instead of the mean(1/2) value.

We subtractedthe sky backgroundand flat-fielded eachexposure using thetwilight flat-fieldsprovided by ESO.The relative positionoffsets were derivedfrom cross-correlat-ing the images; theexposures were co-aligned on a 0.5 ×0.5-pixel sub-gridfor better spatialresolution, and thenadded together us-ing the median fil-tering technique.The resulting im-ages are 3.′4 × 3.′4 insize with pixels of 0.″074. The effectiveexposure times of the final broadband im-ages in the central 2.′5 × 2.′5 are 37, 45,and 48 minutes in Js, H, and Ks, re-spectively.

Figure 1 shows the impressive 3-colourcomposite image that has recently beenthe subject of an ESO press release.(http://www.eso.org/outreach/press-rel/pr-1999/pr-16-99.html). The brightest star inthe FOV (80″ northeast of the core) is thered supergiant IRS 4 (Frogel, Persson,& Aaronson 1977). To the south of thecluster is a giant molecular cloud(GMC). Ionising radiation and fast stel-lar winds from the starburst cluster areexcavating large gaseous pillars. Lo-cated about 20″ to the north of thecluster centre is the blue supergiantSher 25. This supergiant is unique be-cause its circumstellar ring and bipo-lar outflows form an hourglass struc-ture similar to that of SN 1987A (Brand-ner et al. 1997a, 1997b). The image alsoshows the three proplyd-like objectsthat have recently been discovered byBrandner et al. (2000) (see Fig. 2 for

details). About 1′ south of the centralcluster, we detect the brightest mem-bers of the deeply embedded pro-tocluster IRS 9.

In order to derive the photometricfluxes of the stars we used the IRAFimplementation of DAOPHOT (Stetson1987). We first ran DAOFIND to detectthe individual sources, leading toP 20,000 peaks in each waveband.Many of these may be noise orpeaks in the nebular background andappear only in one waveband. In orderto reject spurious sources, we requiredthat sources be detected independ-ently in all three wavebands, and thatthe maximal deviation of the sourceposition centroid between differentwavebands be less than 0.075″. Theresulting source list contains 6967 ob-jects in the entire FOV. We then flux-calibrated the images using the faintNIR standard stars from the lists byHunt et al. (1998) and Persson et al.(1998). Because of the stringent re-quirements on the seeing, the PSFdid not noticeably change during our

observations and the systematic pho-tometric errors are dominated by un-certainties in the aperture offsets. (Adetailed error analysis will be partof a subsequent paper). Comparingour photometric fluxes of numeroussources with the fluxes derived byEisenhauer et al. (1998) yields asystematic offset of 0.1m in Js and0.05m in Ks.

4. HST/WFPC2 Observations

On March 5, 1999 we obtaineddeep narrow-band Hα (F656N, 2 ×500s) and [NII] (F658N, 2 × 600s)observations. The Planetary Camera(PC2) chip was centred on the bipo-lar outflow structure around the bluesupergiant Sher 25. The three WideField Camera (WF) chips coveredthe central cluster and the HII regionto the South of Sher 25. In addition,we retrieved and analysed archivalHST data, which had originally beenobtained in July 1997 (PI Drissen). The

Figure 1: Three-colour image of NGC 3603 composed from Js (blue), H (green), and Ks-band (red) images. Intensi-ties are scaled in logarithmic units. The field of view is 3.′4 × 3.′4. North is up, East to the left. The insert to the lowerright is a blow-up of the central parsec2.

48

PC2 was centred on the cluster, andthe three WF chips covered the areanorth-west of the cluster. We combinedindividual short exposures in F547M(8 × 30s), F675W (8 × 20s), and F814W(8 × 20s) to produce images with effec-tive exposure times of 240s, 160s, and160s, respectively.

The HST/WFPC2 observations arepresented in Figure 2. The figureshows an overlay of two compositecolour images. The upper part of theimage consists of the archive data withthe following colour coding: F547M(blue), F675W (green), F814W (red).Overlaid are our new WFPC2 data with

the F656N data in the red channel,the average of F656N and F658N inthe green channel, and F658N in theblue channel. The locations of theproplyds are marked by small boxes,and enlargements of the boxes areshown in the upper part of Figure 2.Proplyd 3 has only been observed inintermediate and broad-band filtersand thus stands out less clearly againstthe underlying background when com-pared to Proplyds 1 and 2. The insertat the lower right shows a colour com-posite of HST/WFPC2 F656N (blue)and F658N (green) data and VLT/ISAAC Ks data (red).

5. Results and Interpretation

All three proplyds are tadpole shapedand rim brightened, with the extendedtails facing away from the starburstcluster. The portion of the ionised rimspointing towards the cluster arebrighter than the rims on the oppositeside. The central parts of the proplydsare fainter than the rims, with a notice-able drop in surface brightness be-tween the head and the tail. Proplyds 2and 3 exhibit a largely axisymmetricmorphology, whereas Proplyd 1, whichis also the one closest to the clus-ter, has a more complex structure. Un-

Figure 2: WFPC2 observations of NGC 3603. North is up and East is to the left. The upper part of the image consists of the archive data withthe following colour coding: F547M (blue), F675W (green), F814W (red). Overlaid are our new WFPC2 data with the F656N data in the redchannel, the average of F656N and F658N in the green channel, and F658N in the blue channel. The location of the three proplyd-like emis-sion nebulae is indicated. The insert at the lower right is a combination of WFPC2 F656N (blue) and F658N (green) and VLT/ISAAC Ks (red)observations.

49

like the convex shape of the heads of theother proplyds, Proplyd 1 has a heart-shaped head with a collimated, outflow-like structure in between. One possibleexplanation for the more complex mor-phology of Proplyd 1 might be that it isactually a superposition of two (or maybeeven three) individual proplyds or thatthe photoevaporative flows of severaldisks in a multiple system interact toproduce this complex single structure.

At distances of 7.4″ and 2.9″ from Pro-plyd 1 and 2, respectively, faint arc-likeHα emission features are seen on theWFPC2 frames. The arcs are located inthe direction of the cluster, and may bethe signatures of bow shocks created bythe interaction of proplyd winds with thewinds from the massive stars in the cen-tral cluster. Additional faint filaments lo-cated between the nebulae and the cen-tral ionising cluster can be interpreted asbow shocks resulting from the interactionof the fast winds from the high-mass starsin the cluster with the evaporation flowfrom the proplyds.

Low-resolution spectra obtained withEFOSC2 at the ESO/MPI 2.2-m tele-scope of the brightest nebula, which isat a projected separation of 1.3 pc fromthe cluster, reveal that it has the spec-tral excitation characteristics of an ul-tra compact H II region with electrondensities well in excess of 104 cm–3.The near-infrared data reveal a pointsource superimposed on the ionisationfront.

The striking similarity of the tadpole-shaped emission nebulae in NGC 3603to the proplyds in Orion suggests thatthe physical structure of both types of ob-jects might be the same. Our hydro-dynamical simulations (Brandner et al.2000) reproduce the overall morpholo-gy of the proplyds in NGC 3603 very well,but also indicate that mass-loss rates ofup to 10–5 Mo yr–1 are required in orderto explain the size of the proplyds. Dueto these high mass-loss rates, the pro-plyds in NGC 3603 should only survive105 yrs. Despite this short survival time,we detect three proplyds. This indicatesthat circumstellar disks must be commonaround young stars in NGC 3603 and thatthese particular proplyds have only re-cently been exposed to their presentharsh UV environment.

The point source close to the head ofProplyd 3 is already detected on thebroadband HST/WFPC2 observations(see Figure 2). The WFPC2 images showthat the point source is actually locatedin front of Proplyd 3 and thus very likelynot physically associated with it. The in-frared source in Proplyd 1 is also detectedas an underlying, heavily reddened con-tinuum source in the spectrum. A moredetailed analysis of ground based opti-cal images and spectra of the proplydsand other compact nebulae in NGC 3603will be presented in Dottori et al. (2000).

The figure on page 1 of this issue ofThe Messenger shows the resultingcolour-magnitude diagrams (CMD) de-

rived from our VLT data. The left plot con-tains all stars detected in all 3 wavebandswithin the entire FOV of 3.′4 × 3.′4 (6 pc× 6 pc). Since NGC 3603 is located in theGalactic Plane we expect a significantcontamination from field stars. To reducethis contribution we followed a statisticalapproach by subtracting the averagenumber of field stars found in the regionsaround the cluster at r > 75″ (central plot)per magnitude and per colour bin (0.5mag each).

The accuracy of our statistical sub-traction is mainly limited by three factors:first, we cannot rule out that low-masspre-main-sequence stars are also pres-ent in the outskirts of the cluster. Sec-ond, because of crowding on one handand dithering which leads to shortereffective integration times outside thecentral 2.′5 × 2.′5 on the other hand, thecompleteness limit varies across theFOV. Third, local nebulosities may hidebackground field stars. However, noneof these potential errors affects ourconclusions drawn from the CMD. Theresulting net CMD for cluster stars with-in r < 33″ of NGC 3603 is shown at theright in the figure on page 1. We over-laid the theoretical isochrones of pre-main sequence stars from Palla &Stahler (1999) down to 0.1 M0. We as-sumed a distance modulus of (m–M)0

= 13.9 based on the distance of 6 kpc(De Pree et al. 1999) and an averageforeground extinction of AV = 4.5m fol-lowing the reddening law by Rieke &Lebofski (1985).

Applying the isochrones to measuredmagnitudes could be misleading sincethe theoretical calculations include onlystellar photospheres while the stars stillmay be surrounded by dust envelopesand accretion disks. This would lead toexcess emission in the NIR and makethe stars appear younger than they ac-tually are. Typical excess emission ofclassical T-Tauri stars in the Taurus-Auriga complex have been determinedas DH = 0.2m, and DK = 0.5m (Meyer,Calvet & Hillenbrand 1997). The upperpart of the cluster-minus-field CMDclearly shows a main sequence with amarked knee indicating the transition topre-main-sequence stars. The turn-onoccurs at Js P 15.5 mag (m P 2.9M0). Below the turn-on the main-se-quence basically disappears. We notethat the width of the pre-main se-quence in the right part of the figure onpage 1 does not significantly broadentoward fainter magnitudes, indicatingthat our photometry is not limited byphotometric errors. The scatter may infact be real and due to varying fore-ground extinction, infrared excess andevolutionary stage. In that case the leftrim of the distribution would be repre-sentative of the “true” colour of themost evolved stars. Fitting isochronesto the left rim in the CMD yields anage of only 0.3–1.0 Myr. Our result isin good agreement with the study byEisenhauer et al. (1998) but extends

the investigated mass range by aboutone order of magnitude toward smallermasses and covers also a much largerarea.

Because of P 10 magnitudes rangein luminosities in the crowded core re-gion, our sensitivity limits of Js P 21m,H P 20m and Ks P 19m don’t appearto be exceedingly faint (and are about3 magnitudes above ISAAC’s detectionlimit for isolated sources). However, onlyVLT/ISAAC’s high angular resolution,PSF stability, and overall sensitivity en-abled us to study the sub-solar stellarpopulation in a starburst region on a star-by-star basis.

Acknowledgements: We would liketo thank the ESO staff, in particular thoseinvolved in the service observations, fortheir excellent work.

References

Brandl, B., et al. 1996, ApJ, 466, 254.Brandl, B., Brandner, W. & Zinnecker, H. 1999,

Star Formation 1999, eds. T. Nakamoto etal., in press.

Brandl, B., et al. 1999, A&A Letters, 352,L69.

Brandner, W., et al. 1997a, ApJ, 475, L45.Brandner, W., et al. 1997b, ApJ, 489,

L153.Brandner, W., et al. 2000, AJ, in press (Jan-

uary 2000 issue).Churchwell, E., Felliw, M., Wood, D.O.S. &

Massi, M. 1987, ApJ, 321, 516.De Pree, C.G., Nysewander, M.C. & Goss,

W.M. 1999, AJ, 117, 2902.Dottori, H., et al. 2000, subm. to A&A.Drissen, L., Moffat, A.F.J., Walborn, N. &

Shara, M.M. 1995, AJ, 110, 2235.Eisenhauer, F., Quirrenbach, A., Zinnecker, H.

& Genzel, R. 1998, ApJ, 498, 278.Frogel, J.A., Persson, S.E., Aaronson, M.

1977, ApJ, 213, 723. Hunt, L.K., et al. 1998, AJ, 115, 2594.Johnstone, D., Hollenbach, D. & Bally, J. 1998,

ApJ, 499, 758.Kennicutt, R.C. Jr. 1984, ApJ, 287, 116.McCullough, P.R., Fugate, R.Q., Christou, J.C.

Et al. 1995, ApJ, 438, 394.Melnick, J., Tapia, M. & Terlevich, . 1989, A&A,

213, 89.Meyer, M.R., Calvet, N., & Hillenbrand, L.A.

1997, AJ, 114, 288.Moffat, A.F.J. 1983, A&A, 124, 273.Moffat, A.F.J., Drissen, L., & Shara, M.M. 1994,

ApJ, 436, 183.O'Dell, C.R.O., Wen, Z. & Hu, X. 1993, ApJ,

410, 686.Palla, F. & Stahler, S. W. 1993, ApJ, 418,

414.Palla, F. & Stahler, S.W. 1999, ApJ, in press

(20 Nov).Persson, S.E. et al. 1998, AJ, 116, 247.Stecklum, B., Henning, T., Feldt, M., et al.

1998, ApJ, 115, 767.Stetson, P.B. 1987, PASP, 99, 191.Störzer, H. & Hollenbach, D. 1999, ApJ, 515,

669.Zinnecker, McCaughrean & Wilking 1993,

Protostars and Planets III, eds. E.H. Levy& J.I. Lunine (University of Arizona Press),429.

E-mail: brandl@astrosun.tn.cornell.edu

1. Introduction

The fact that the spectra of planetarynebulae (PNs) are dominated by strongemission lines, with a very weak contin-uum, has two interesting consequencesfor extragalactic work. First, PNs in oth-er galaxies are easily detectable: weneed two images of a galaxy, one takenthrough a narrow-band filter transmit-ting a strong nebular emission (e.g.[O III] λ5007) and another takenthrough an off-band filter which doesnot transmit any strong nebular line. Inthe on-band image the PN appears asa point source. Since the nebular con-tinuum is so weak, the point sourceshould be invisible in the off-band im-age. By blinking the two images, andprovided that the off-band image isdeep enough, it is easy to identify thePNs. Second, for each detected PN wecan measure an accurate radial veloci-ty, because all the flux we have detect-ed is concentrated at the redshiftedwavelength of the emission line. Thuswe can take a good spectrogram in anexposure time not much longer thanwhat we need for the imaging. SincePNs are preferentially discovered inthe outskirts of galaxies, where thesurface brightness is lower, this meansthat PNs are ideal test particles forstudies of rotation and mass distri-bution in the halos of the correspondinggalaxies. Typical examples of suchstudies are Hui et al. (1995) on NGC5128, and Arnaboldi et al. (1996) on theVirgo cluster galaxy NGC 4406. Herebegins an unusual chain of unexpectedresults.

2. Intracluster PNs

The giant galaxy NGC 4406 is char-acterised by a negative redshift (–230km s–1), and indeed many of its PNs,discovered by Jacoby et al. (1990) andmeasured by Arnaboldi et al. (1996),have the expected redshifts. However,3 of these PNs turn out to have red-shifts typical of the Virgo cluster(around 1300 km s–1). The conclusionwas that these 3 PNs do not belong toNGC 4406; they must be members ofan intergalactic or intracluster diffusestellar population in the Virgo cluster,

and only by chance they appear pro-jected upon NGC 4406. Direct evi-dence of the existence of red giantsbelonging to this Virgo intracluster stel-lar population was subsequently re-ported by Ferguson et al. (1998), whoworked with the Hubble Space Tel-escope. These red giants are too faintfor spectroscopic studies. In contrast,the intracluster PNs offer an ideal op-portunity to learn about the kinemat-ic behaviour and perhaps even theabundances of the diffuse stellar popu-lation. A search for intracluster PNs indifferent positions across the Virgocluster started immediately and pro-duced dozens of PN candidates(Méndez et al. 1997, Feldmeier et al.1998).

The “on-band/off-band” narrow-bandfilter technique used to discover the PNcandidates allows the detection of asingle emission line. This is not neces-sarily the desired [O III] λ5007; it mightbe another emission line at higher z,redshifted into the on-band filter, like[O II] λ3727 at z = 0.35, or Lyα at z =3.14. In previous work (e.g. Méndez etal. 1997) we argued that most of ourdetections had to be real PNs because(a) the surface density of emission-linegalaxies derived from previous studies(for example Thompson et al. 1995)was not high enough to explain all thedetections; (b) the luminosity functionof the detected sources, if we assumethem to be at the distance of the Virgocluster, is in good agreement with thePN luminosity functions derived inseveral Virgo galaxies (Jacoby et al.1990).

Then we decided to take deepVLT+FORS spectrograms of several in-tracluster PN candidates, located inField 1 of Feldmeier et al. (1998) and inthe smaller field observed by Méndez etal. (1997; in what follows the “La PalmaField” or LPF, because the discoveryimages were taken with the 4.2-mWilliam Herschel Telescope at LaPalma). Our purpose was first of all tomeasure radial velocities and providespectroscopic confirmation of the PNnature of the candidates, since a few ofthem could be high-redshift sources.Second, we were hoping to detect thevery faint diagnostic emission lines nec-

essary for reliable abundance determi-nations, at least for a few of the bright-est intracluster PN candidates. Thiswas not to be; in what follows we reportthe unexpected outcome of our firstVLT run. Here we will restrict our de-scription to the LPF, where the most in-teresting objects are located. A paperwith a complete report has been sub-mitted to ApJ.

3. Observations and FirstResults

In April 1999 we used the VLT UT1with FORS in multi-object spectroscop-ic mode (MOS) with grism 300V. Wedefined 1 arcsec wide slitlets for sever-al PN candidates in the LPF, and addedslitlets for 2 objects suspected to beQSOs or starbursts because they werevisible (although much weaker) in theoff-band discovery image (Méndez etal. 1997). Some other slitlets wereplaced on stars or galaxies in the field,in order to check the slitlet positioning(done by taking a short exposure with-out grism through the slitlets) and tohelp locate the dispersion lines as afunction of position across the field,which is important for the correct ex-traction of spectra consisting of isolatedemission lines.

After completing 5 exposures of theinitial MOS configuration, each 40 min-utes long, the surprising result was thatour targets in the LPF were confirmedas emission-line objects, but none ofthem is a PN. If we expose longenough, a PN must show not only [O III]λ5007, but also [O III] λ4959, which isthree times weaker. In none of oursources could we find the second [OIII]line. The spectra show only a strong,isolated and narrow emission line atwavelengths from 5007 to 5042 Å (en-compassing the full width of the on-band discovery filter), and nothingelse.

Of the 2 QSO or starburst candi-dates, one was confirmed as a QSO atz = 3.13, showing a typical broad-linedLyα and several other emissions, in-cluding C IV λ1550. The other one ap-pears to be a starburst region (onestrong, isolated and narrow emissionline, with faint continuum).

50

From Intracluster Planetary Nebulae to High-Redshift L yα EmittersR.P. KUDRITZKI and R.H. MÉNDEZ, Munich University ObservatoryJ.J. FELDMEIER and R. CIARDULLO, Dept. of Astron. and Astrophys., Penn State UniversityG.H. JACOBY, Kitt Peak National Observatory, TucsonK.C. FREEMAN, Mt. Stromlo and Siding Spring ObservatoriesM. ARNABOLDI and M. CAPACCIOLI, Osservatorio Astronomico di Capodimonte, NapoliO. GERHARD, Astronomisches Institut, Universität BaselH.C. FORD, Physics and Astronomy Dept., Johns Hopkins University, Baltimore

In this way we found ourselves un-able to work on the expected intra-cluster PNs, and facing a differentproblem, namely, we had to identifythe isolated emission. Was it [O II]λ3727 at z = 0.35, or could it be Lyαat z = 3.13? We defined a new MOSconfiguration, placing the slitlets so asto cover as much wavelength range to-wards the infrared as possible. Wewere able to take three exposures withthis new MOS configuration. Based onthe absence of any other emission linein our spectra, we could rule out allthe alternatives, e.g. if the detectedemission were [O II] λ3727 redshiftedinto the on-band filter, then we shouldsee other features, like Hβ, the [O III]lines and Hα, at the correspondingredshift. A similar argument can beused to reject Mg II λ2798 at z = 0.79.There is only one possible identifica-tion: Lyα at z = 3.13. We had unwilling-ly found a nice collection of high-red-shift Lyα emitters, which was an ade-quate compensation for the loss of theexpected intracluster PNs.

4. How Could This Happen?

Not just a few, but all the sources wetested with FORS are high-redshiftgalaxies. However it would be wrong toconclude that intracluster PNs do notexist. In fact, multi-object spectroscopywith the AAT and the 2-degree-field(2df) fibre spectrograph has confirmedthe PN nature of many Virgo intraclus-ter candidates, by detecting both [O III]emissions at 4959 and 5007 Å (Free-man et al. 1999). What happened tous is that the surface density of intra-cluster PNs in the LPF appears to bemuch lower than in other places acrossthe Virgo cluster, probably indicatingsome degree of clumpiness in the dis-tribution of the diffuse intracluster pop-ulation. This will be verified by wide fieldsurveys currently in progress.

There is another aspect to consider:the rather small area of the LPF (50arcmin2). Given a small area and a lowsurface density, the total number of in-tracluster PNs is small, and it becomesvery improbable to find bright PNs. Itseems that the fraction of high-redshiftLyα sources in the on-band/off-bandsamples is higher at faint magnitudes,thus explaining the detections in theLPF. We conclude that it is more effi-cient to search for intracluster PNs us-ing wide-field survey cameras, like theWFI at the ESO 2.2-m telescope. Witha larger area the total number of PNsincreases and it is more probable to findbright ones; at brighter magnitudes thefraction of high-redshift Lyα emitters issmaller.

5. Properties of the L yα Emitters

The weak (in most cases undetect-ed) continuum implies a large Lyα

equivalent width, between 20 and 200Å in the rest frame. For these typicalequivalent widths the most probableexplanation is H II regions ionised byrecently formed massive stars (e.g.Chariot and Fall 1993). The observedLyα fluxes are between 2 × 10–17 and 2× 10–16 erg cm–2 s–1. Adopting z = 3.13,H0 = 70 km s–1 Mpc–1 and q0 = 0.5 weget a luminosity distance of 1.8 × 104

Mpc, which implies Lyα luminosities forour sources between 2 × 108 and 2 ×109 L0. These numbers depend on ourassumptions about the cosmologicalparameters: for a flat universe withΩ0 = 0.2, ΩΛ = 0.8, and the same zand H0 as above, the luminosity dis-tance becomes 3 × 104 Mpc, and theLyα luminosities become larger by afactor 2.8. Our Lyα emitters are quitesimilar to those recently found by Hu(1998), Cowie & Hu (1998) and Hu etal. (1998) at redshifts from 3 to 5.

Using simple population synthesismodels, on the assumption that thesesources are produced by star for-mation with a Salpeter initial massfunction and a lower mass cut-off of0.5 M0, we have concluded (Kudritzkiet al. 1999 ApJ submitted) that thenebulae are nearly optically thick andmust have a very low dust content, inorder to explain the high observedLyα equivalent widths. These galaxiesmust differ substantially from the typ-ical Lyman break galaxies detectedthrough their U-dropout; those arecharacterised by a strong continuum(needed for their very detection) andweak or absent Lyα emission. We at-tribute this difference to a lower dustcontent of our objects.

For the cosmological and star forma-tion parameters we adopted, the totalstellar mass produced would seem to

correspond to the formation of rathersmall galaxies, some of which areperhaps destined to merge. However,one of our sources might become aserious candidate for a proto-giantspheroidal galaxy if we assumed con-tinuous star formation, a low mass cut-off of 0.1 M0 in the IMF, and a flat ac-celerating universe with Ω0 = 0.2 andΩΛ = O.8.

The implied star formation densityin our sampled comoving volume isprobably somewhat smaller than, butof the same order of magnitude as,the star formation density at z ~ 3 de-rived by other authors from Lyman-break galaxy surveys (see e.g. Steidelet al. 1999). This result agrees with theexpectation that the Lyα emitters area low-metallicity (or low-dust) tail in adistribution of star-forming regions athigh redshifts. Finally, the Lyα emittersmay contribute as many H-ionising pho-tons as QSOs and Ly-break galaxies atz ~ 3 (see e.g. Madau et al. 1998). Theyare therefore potentially significant forthe ionisation budget of the early uni-verse.

6. Future W ork

Now we would like to do someadditional work on these Lyα emitters.In particular we would like to useISAAC; we expect the strongest [O II]and [O III] forbidden lines as well asHβ to be detectable with the ShortWavelength arm of ISAAC in MediumResolution spectroscopic mode. Thiswould allow us to either determineor (in the case of non-detection) putan upper limit to the metallicity. It mightalso be possible to clarify to whatextent the production of Lyα photonscan be attributed to AGN activity in-

51

Figure 1. A section of an image produced by coadding 3 MOS exposures of the La Palma Field(LPF), taken with FORS. Here 4 of the 19 FORS slitlets are shown. These spectra, taken dur-ing new moon, are dominated by strong sky emissions. From top to bottom: (1) one of our in-tracluster PN candidates, now re-identified as a high-redshift Lyα emitter; (2) a foreground star;(3) a low-redshift galaxy showing a bright H II region at the edge, where Hβ and the [OIII] emis-sions λλ4959, 5007 are visible; (4) one of the QSO or starburst candidates, confirmed as aQSO. The strong and broad emission line is Lyα at z = 3.13. A weak continuum is visible. CIV λ1550 is visible to the right, flanked by strong sky emissions.

stead of recent massive star formation.The firm detection of an infrared contin-uum would help to constrain the char-acteristics and total mass of the stellarpopulations through comparison withpopulation synthesis models.

References

[1] Arnaboldi M., Freeman K.C., MéndezR.H. et al. 1996, ApJ 472, 145.

[2] Charlot S., Fall S.M. 1993, ApJ 415, 580.[3] Cowie L.L., Hu E.M. 1998, AJ 115,1319.

[4] Feldmeier J.J., Ciardullo R., Jacoby G.H.1998, ApJ 503, 109.

[5] Ferguson H.C., Tanvir N.R., von HippelT. 1998, Nature 391, 461.

[6] Freeman K.C. et al., 1999, in GalaxyDynamics: from the Early Universe tothe Present, ASP Conf. Ser., eds. F.Combes, G.A. Mamon, V. Charmandaris,in press, astro-ph/9910057.

[7] Hu E.M. 1998, ASP Conf Ser. 146, p. 148.[8] Hu E.M., Cowie L.L., McMahon R.G.

1998, ApJ 502, L99.[9] Hui X., Ford H.C., Freeman K.C., Dopi-

ta M.A. 1995, ApJ 449, 592.

[10] Jacoby G.H., Ciardullo R., Ford H.C.1990, ApJ 356, 332.

[11] Madau P., Haardt F., Rees M.J. 1998,preprint astro-ph/9809058 (4 Sep 1998).

[12] Méndez R.H., Guerrero M.A., FreemanK.C. et al. 1997, ApJ 491, L23.

[13] Steidel C.C., Adelberger K.L., Giavalis-co M. et al. 1999, preprint astroph/9811399v2 (23 Jan 1999).

[14] Thompson D., Djorgovski S., Trauger J.1995, AJ 110, 96.

52

VLT Spectroscopy of the z = 4.11 Radio Galaxy TN J1338–1942C. DE BREUCK1,2, W. VAN BREUGEL2, D. MINNITI 2,3,G. MILEY1, H. RÖTTGERING1, S.A. STANFORD2 and C. CARILLI4

1Sterrewacht Leiden, The Netherlands (debreuck,miley,rottgeri@strw.leidenuniv.nl)2Institute of Geophysics and Planetary Physics, Lawrence Livermore

National Laboratory, Livermore, U.S.A. (wil,adam@igpp.llnl.gov)3P. Universidad Católica, Santiago, Chile (dante@astro.puc.cl)4National Radio Astronomy Observatory, Socorro, USA (ccarilli@nrao.edu)

High-redshift radio galaxies (HzRGs)play an important role in cosmology. Theyare likely to be some of the oldest and

most massive galaxies at high red-shifts, and can therefore constrain theepoch at which the first generation of

stars were formed. The near-IR Hub-ble K – z diagram of powerful radiogalaxies shows a remarkably tight cor-relation from the present time out toz = 5.19, despite significant K-correc-tions (van Breugel et al. 1998, 1999).This indicates that we can follow theevolution of the hosts of HzRGs fromnear their formation epoch out to lowredshift (z <~ 1), where powerful radiosources inhabit massive elliptical gal-axies (e.g. Lilly & Longair 1984; Best,Longair & Röttgering 1998). For exam-ple, at z ~ 3, we observe a changein the observed K-band morphologiesof HzRGs from large-scale low-sur-face brightness emission with brightradio-aligned clumps at z <~ 3 to smooth,compact structures, sometimes show-ing elliptical shapes, like their ocalUniverse counterparts (van Breugel etal. 1998). These surrounding clumpshave properties similar to the UV drop-out galaxies at similar redshifts (Pen-tericci et al. 1999), and indicate thatHzRGs often reside in (proto-)clusterenvironments. This evolution pictureseems consistent with the hierarchicalclustering formation models (e.g. Kauff-mann et al. 1999), where massive ob-jects form by accretion of smaller sys-tems located in over-dense regions.

Figure 1: VLT spectrum of TN J1338–1942.The lower panel has been boxcar smoothedby a factor of 9 to better show the shape ofthe Lyα forest and the Lyman limit.

E-mail: kudritzki@usm.uni-muenchen.de

Detailed studies of HzRG can then becompared with the predictions of thehierarchical clustering scenarios.

Using newly available radio surveys,we have begun a systematic search forz > 3 radio galaxies (De Breuck et al.2000), to be followed by more detailedstudies of selected objects.

During the second week of VLTAntu operations in April 1999, we usedFORS1 to obtain a spectrum of thesecond highest redshift radio galaxyfrom our sample, TN J1338–1942 at z =4.11. We discovered this object inMarch 1997 using the ESO 3.6-m tele-scope (De Breuck et al. 1999a). Thepurpose of these VLT observations wasto study the Lyα emission and the UV-continuum.

We used the 600R grism with a 1.3″wide slit, and integrated for 47 min. Thespectrum, shown in Figures 1 and 2, isdominated by the bright Lyα line (FLyα =2 × 10–15 erg s–1 cm–2) with a rest-frameequivalent width of 210±50Å, a typicalvalue for HzRGs.

Most notable from Figure 2 is thelarge asymmetry in the emission pro-file, consistent with a very wide (~ 1400km/s) blueward depression over theentire spatial extent of the emission(~10 kpc). Similar asymmetries havebeen detected in other HzRGs, andhave been interpreted as due to ab-sorption by cold HI gas in a halo sur-rounding the radio galaxy (e.g. van Ojiket al. 1997; Dey 1999). The Lyα profileis well fit by a simple model consistingof a Gaussian emission profile and asingle Voigt absorption function (seeFigure 2). This model constrains theHI column density in the range 3.5 ×1019 – 1.3 × 1020 cm–2. From this, wederive the total mass of the absorberat 2 – 10 × 107 M0, which is compara-ble or slightly less than the total HIImass, as derived from the Lyα emis-sion (De Breuck et al. 1999b). Theseresults show that TN J1338–1942 issurrounded by a large reservoir ofgas.

We also detect continuum emissionin TN J1338–1942, only the secondtime this is reported in the spectrum ofa z > 4 radio galaxy. The continuum,shown in the bottom panel of Figure 1,shows a discontinuity across the Lyαline, and a tentative detection of theLyman limit. The flux deficit bluewardsof Lyα is interpreted as HI absorp-tion along the cosmological line of sight,and is described by the parameter

DA =

(Oke & Korycanski 1982). For TNJ1338–1942, we measure DA = 0.37 ±0.1, similar to the DA = 0.45 ± 0.1measured for the z = 4.25 radio galaxy8C 1435+64 (Spinrad et al. 1995). Boththese DA values for z > 4 radiogalaxies are significantly lower thanthe DA values found for quasars atthese redshifts (Schneider et al. 1991).Although based on only two measure-

ments, our results suggest that the fore-ground HI absorption (≡ DA) in quasarscould have been overestimated be-cause the highest redshift (z > 4) qua-sars have been found from sampleswith an optical colour selection, therebyexcluding quasars with relatively smallbreaks across Lyα.

It is therefore particularly importantto expand the number of objects atz > 4 selected by techniques otherthan optical colours. Radio galaxiesare the brightest objects at these red-shifts which are not out-shined bytheir central AGN like in quasars. Usingthe high sensitivity of the VLT instru-ments, we will be able to study at thesame time the origin and evolution oftheir giant luminous halos or ionisedgas, and the neutral HI gas surroundingthem.

Acknowledgements. The work byC.D.B., W.v.B., D.M. and S.A.S. atIGPP/LLNL was performed under theauspices of the US Department ofEnergy under contract W-7405-ENG-48. D.M. is also supported by Fondecytgrant No. 01990440 and DIPUC. Thiswork is supported in part by theFormation and Evolution of Galaxiesnetwork set up by the EuropeanCommission under contract ERBFMRX-CT96-086 of its TMR pro-gramme.

Bibliography

[1] Best, P. N., Longair, M. N., & Röttgering,H.J.A. 1998, MNRAS, 295, 549.

[2] De Breuck, C., van Breugel, W., Röttger-ing, H., Miley, G., & Carilli, C. 1999a, in“Looking Deep in the Southern Sky”, ed.R. Morganti & W. Couch (Berlin Heidel-berg: Springer), p. 246.

[3] De Breuck, C. van Breugel, W., Minniti,D., Miley, G., Röttgering, H., Stanford, S.A., & Carilli, C. 1999 A&A, 1999b, 352,L51.

[4] De Breuck, C., van Breugel, W., Röttger-ing, H., & Miley, G.2000, A&AS, submit-ted.

[5] Dey, A. 1999, in 'The Most Distant RadioGalaxies', ed. H. Röttgering, P. Best M.Lehnert (Amsterdam: KNAW), p. 19.

[6] Kauffmann, G., Colberg, J.M., Diaferio, A.,& White, S. D. M. 1999, MNRAS, 303, 188.

[7] Lilly, S. J. & Longair, M., 1984, MNRAS,211, 833.

[8] Oke, J.B., & Korycanski, D.G., 1982, ApJ,255, 11.

[9] Pentericci, L. et al. 1999, A&A, 341, 329.[10] Schneider, D. P., Schmidt, M, & Gunn,

J.E. 1991, AJ, 101, 2004.[11] Spinrad, H., Dey, A., & Graham, J.R.

1995, ApJ, 438, 51.[12] van Breugel, W., Stanford, S. A., Spinrad,

H., Stern, D., & Graham, J.R. 1998, ApJ,502, 614.

[13] van Breugel, W., De Breuck, C., Stanford,S. A., Stern, D., Röttgering, H., & Miley,G. 1999, ApJ, 518, 61.

[14] van Ojik, R., Röttgering, H., Miley, G., &Hunstead, R. W. 1997, A&A, 317, 358.

53

Figure 2: Part of the spectrum around the Lyα line. The solid line is the model consisting of aGaussian emission profile (dashed line) and a Voigt absorption profile with the indicated pa-rameters.

54

“The VLT is Europe’s great leap for-ward, heralded as a new window on thedistant universe. Surely we can think ofsome joint projects that will turn theAmericans green!” Trying to rally thetroops at the annual meeting of our EC-funded “Galaxy Formation” networkforced me to think about where the VLTpay-off might really come. A talk fromAlan Moorwood provided some valu-able ideas. The first efficient near-IRspectrograph on an 8-metre telescopecould detect Hα past redshift 2 and [OII]3727 to redshift 5. How about gettingkinematics for distant galaxies likethose in the Hubble Deep Fields or theinfamous “Steidel” objects? This wouldhave to clarify their relation to nearer,dearer, but (perhaps) more boringgalaxies. Of course, I’d never taken aninfrared spectrum of anything, but whynot start now?

In a new field it’s always important toget good help. My chance came with thenews that Nicole Vogt was moving to IoAin Cambridge, one of the partners in our

network. Californian graduate studentsnow regularly present theses full of 10-metre data, and Nicole’s contained Keckrotation curves for 15 galaxies out to red-shift 0.8. Data for fifty more were ru-moured to be in the pipeline. Some dis-cussion showed that ISAAC could get Hαrotation curves for her most distant Keckgalaxies (observed in [OII] 3727 onHawaii) and, more interestingly, could ex-tend the accessible redshift range by atleast a factor of two. All we had to do wasto find suitable candidates and to bangaway for a few hours under the famousParanal observing conditions. A propos-al was duly dispatched suggesting a pi-lot study in the Hubble Deep Field South.We were elated when the Observing Pro-grammes Committee gave us (most of)the time.

Travelling to Paranal from Garchingmakes one appreciate the virtues of re-mote observing. The shops at Paris CDGand in Buenos Aires were interesting butthe attractions of neither place compen-sated for the long hours spent on planes

and in airport lounges. The ESO guesthouse in Santiago, seemingly stuck in agenteel time warp between encroachinghigh-rise apartments, provided a wel-come rest, but was spoiled somewhat bythe need to get up long before breakfastin order to catch a taxi to catch a planeto catch a taxi to catch the once-a-dayESO bus to Paranal. The bus itself, com-plete with green baize curtains, recliningseats and movie videos, was a surreal butsurprisingly comfortable way to end theodyssey. It dumped us under a clear bluesky in front of a small city of temporarybuildings which appeared to have beenbuilt on the Moon. Not Ithaca, perhaps,but it turned out to be too deceptively lux-urious to be considered Spartan.

In the two days before our run startedwe learned the novel joys of P2PP, cre-ating enough OBs (Observing Blocks) forthe more seasoned observers to decidethat we were quite definitely crazy. Ourproblem was that it isn't easy to decidehow bright a distant object is likely to bein the Hα line when all you have is broad-band photometry.

Furthermore, we had to make do withphotometric redshifts for most of ourcandidates, so we weren’t sure whereHα would fall relative to the night sky.Finally, there just aren’t any “big” spiralsbeyond redshift 1 in the Hubble DeepField South; our best candidates wereall dwarf or irregular systems. As a re-sult we needed plenty of flexibility to ad-just our programme in light of experi-ence at the telescope – and so plenty ofOBs. The inevitable drawbacks of a pi-lot project...

Our first night at the telescope welearned a number of things.

(1) ESO looks after its VLT observersextremely well. The night assistant at-tended to all telescope-related issues,including guide-star acquisition. Wenever had to think about such things.The support scientist knew the instru-ment and its performance inside outand had written all the available quick-look software. As a result he could helpus assess our data as soon as theywere taken. Finally, the shift manageralways showed up whenever things gotcomplicated.

(2) The telescope points and tracks likea dream. Our reference stars always ar-rived within a few seconds of the centreof the field, our offsets all worked to asmall fraction of an arcsecond, even overseveral arcminutes, and the telescopedrifted by only a couple of tenths of anarcsecond even after tracking for twohours.

(3) The instrument and the conditionsreally can be “as advertised”. The see-

Weighing Young Galaxies – an Occasional Observer Goes to ParanalS. WHITE, Max-Planck Institute for Astrophysics, Garching

H-band spectrum of the Hα emission line for disk galaxy HDFS–0620, observed with VLT+ISAACin high-resolution mode. The galaxy has a redshift of z = 1.29, placing the Hα emission at λ =1.5 µm, in a clear region of the H-band IR window of the atmosphere between night-sky emis-sion lines. The gas can be traced across almost 2 arcseconds (15 kiloparsecs) along the ma-jor axis of the galaxy, yielding a rotation curve with a total observed amplitude of order 180 km/s.

55

ing on the detector rarely got worse than0.5 arcsec and there were extended pe-riods when it was 0.3 arcsec. Furthermorethe long-slit spectra taken by ISAAC sky-subtracted well and had a cosmetic qual-ity quite comparable to that of good op-tical spectra.

(4) Even with the best equipment un-der the best conditions, a hard project isstill hard. Our first couple of candidatesshowed no clear detection after a cou-ple of hours of integration, leaving usuncertain whether their photometric red-

shifts were in error, their emission layunder a sky line, or was just very weak.We were greatly relieved when our thirdtarget showed strong Hα at z ~ 1.3 witha clear detection of galactic rotation.

Our second night was equally goodand brought several more detectionsas well as some further “blanks”. Wefinished with a feeling of achievement;galaxy kinematics really can be studiedbeyond redshift one. This night alsobrought an unexpected (and probablyillegal) glass of champagne as the

UVES commissioning team, working20 metres away on the UT2 console,obtained their first bona fide quasarspectrum. It was certainly impressive;the Lyα forest could be seen boomingthrough right down to the atmosphericcut-off. As we boarded the bus thefollowing afternoon at the start of thelong trek home, we felt sure we wouldbe returning soon despite the dis-tance.

Sunset on Paranal (Photographer: Herbert Zodet).

(swhite@mpa-garching.mpg.de)

“Science with the Ata-cama Large MillimetreArray” was the title of aninternational conferenceheld at the Carnegie In-stitution in Washington,DC, in early October. Inorder to increase publicattention on the ALMAproject, NRAO had or-ganised a press brief-ing on October 7, en-abling journalists tomeet with scientists andkey representatives ofthe ALMA partnership.The press briefing fol-lowed a reception inthe US Congress. Inaddition, ESO, PPARCand NRAO mountedsmall exhibitions at theCarnegie Institution it-self.

C. Madsen

56

OTHER ASTRONOMICAL NEWS

Science with the Atacama Large Millimetre Array

57

ANNOUNCEMENTS

The La Silla 2000+ Report Now Available

The NEON Observing School

The Network of European Observatories in the North (NEON)organises an observational Euro summer school sponsored by theEuropean Community. This school is taking over from the former,discontinued, ESO/OHP summer school and intends to run on ayearly basis.

The participating observatories are: Asiago Observatory (Italy),Calar Alto Observatory (Germany-Spain) and Haute-Provence Ob-servatory (France), with additional tutorial assistance from ESO.

The purpose of the school is to provide an opportunity to gainpractical observational experience at the telescope, in observa-tories with state-of-the-art instrumentation. To this effect, the schoolproposes tutorial observations in small groups of 3 students, un-der the guidance of an experienced observer, centred around asmall research project and going through all steps of a standardobserving programme. Some complementary lectures will be giv-en by experts in the field.

The school is open to students working on a PhD thesis in As-tronomy and which are nationals of a Member State or an Asso-ciated State of the European Union. The working language is Eng-lish. Up to eighteen participants will be selected by the organising

committee and will have their travel and living expenses paid, ifthey satisfy the EC rules (age limit of 35 years at the time of theEuro Summer School).

The first NEON Euro summer school will take place at the CalarAlto Observatory in Spain, from

July 10 to 22, 2000

Instructions and full practical details can be found on the schoolWeb site when this announcement appears in print.

Applicants are expected to fill in an application form that will beavailable on the Web site, with a curriculum vitae and a descrip-tion of previous observational experience, and to provide a letterof recommendation from a senior scientist.

The provisional application deadline is March 31st, 2000. The school Web site is hosted by the EAS Web site at: http://www.iap.fr/eas/schools.htmlSecretary of the school: Mrs Brigitte RABAN at IAP, 98bis, Bd

Arago, F-75014 PARIS.

Michel Dennefeld, Co-ordinator of the NEON school.

The Working Group “La Silla 2000+” has completed its work,based on the replies to the questionnaire survey of the Users Com-mittee (see The Messenger No. 95, p. 38). A total of 256 colleaguesfrom the ESO countries participated in this survey, a very gratify-ing turnout.

The report of the Working Group, which contains a set of rec-ommendations for the future of the La Silla facilities (1999–2006)has been submitted to the Director General of ESO (June 1999).

The final “La Silla 2000+ Report” (with colour figures) is now available on the Web at the following address:

http://www.eso.org/gen-fac/commit/ls2000pl1.html

The Working Group wishes to thank all of those who answeredthe questionnaire. Special thanks go to the ESO Remedy Team,and in particular to Rein Warmels who helped to design the ques-tionnaire and prepare the figures in the Report.

The report will now be discussed in the relevant ESO committees.

Birgitta NordströmChair, ESO Users Committee and the WG “La Silla 2000+”

PERSONNEL MOVEMENTS(1st October 1999–30th November 1999)

International Staff

ARRIVALS

EUROPE

BRISTOW, Paul (GB), Science Data Analyst/ProgrammerBROWN, Anthony (NL), FellowDIMMLER, Martin (D), Electronics/Control EngineerESKDALE, Jane (AUS), Secretary/Group AssistantGAILLARD, Boris (F), AssociateGONÇALVES C., Anabela (P), FellowRITTER, Heiderose (D), AssociateRIVINIUS, Thomas (D), FellowSARTORETTI, Paola (I), Scientific Application DeveloperTUELLMANN, Ralph (D), StudentVAN ECK, Sophie (B), FellowZAGGIA, Simone (I), Fellow

CHILE

ATHREYA, Ramana (IND), FellowGALLIANO, Emmanuel (F), StudentMARCONI, Gianni (I), Operations Staff AstronomerTAMAI, Roberto (I), Mechanical Engineer

DEPARTURES

EUROPE

BALON, Serge (F), Technical AssistantBRESOLIN, Fabio (I), FellowFERRARI, Marc (F), AssociateFUCH, Rainer (D), Legal AdvisorHILL, Susan (GB), Archive OperatorKORNWEIBEL, Nicholas (AUS), Software EngineerLATSCH, Hedwig (D), Administrative ClerkSCODEGGIO, Marco (I), Fellow

CHILE

SCHUMANN, Erich (D), Catering La Silla

Local StaffDEPARTURES

DIAZ,Jorge, RAMIREZ, Luis, LECAROS, Fernando, FERNANDEZ DE C., Rodrigo,

ARRIVAL

TAPIA, Mario, Mechanical Engineer

58

INAUGURATION CEREMONYOF PARANAL OBSERVATORY5 MARCH 1999

Introduction to the ESO Annual Report 1998by Prof. Riccardo Giacconi, Director Gen-eral of ESO 96, 2

Discourse of Prof. Riccardo Giacconi, Direc-tor General of ESO 96, 2

Discourse of Mr. Grage, President of the ESOCouncil 96, 4

Discourse of His Excellency the President ofthe Republic of Chile, Don Eduardo FreiRuiz-Tagle 96, 5

The VLT Opening Symposium 96, 6

TELESCOPES ANDINSTRUMENTATION

A. Moorwood, J.-G. Cuby, P. Ballester, P. Bier-eichel, J. Brynnel, R. Conzelmann, B. De-labre, N. Devillard, A. Van Dijsseldonk, G.Finger, H. Gemperlein, C. Lidman, T. Her-lin, G. Huster, J. Knudstrup, J.-L. Lizon, H.Mehrgan, M. Meyer, G. Nicolini, A. Silber,J. Spyromilio, J. Stegmeier: ISAAC at theVLT 95, 1

N. Devillard, Y. Jung, J.-G. Cuby: ISAACPipeline Data Reduction 95, 5

A. Kaufer, O. Stahl, S. Tubbesing, P. Nørre-gaard, G. Avila, P. François, L. Pasquini, A.Pizzella: Commissioning FEROS, the NewHigh-resolution Spectrograph at La Silla95, 8

N. Ageorges and N. Hubin: Monitoring ofthe Atmospheric Sodium above La Silla95, 12

D. Baade, K. Meisenheimer, O. Iwert, J. Alon-so, Th. Augusteijn, J. Beletic, H. Bellemann,W. Benesch, A. Böhm, H. Böhnhardt, J.Brewer, S. Deiries, B. Delabre, R. Donald-son, Ch. Dupuy, P. Franke, R. Gerdes, A.Gilliotte, B. Grimm, N. Haddad, G. Hess,G. Ihle, R. Klein, R. Lenzen, J.-L. Lizon, D.Mancini, N. Munch, A. Pizarro, P. Prado, G.Rahmer, J. Reyes, F. Richardson, E. Ro-bledo, F. Sanchez, A. Silber, P. Sinclaire,R. Wackermann, S. Zaggia: The Wide FieldImager at the 2.2-m MPG/ESO Telescope:First Views with a 67-Million-Facette Eye95, 15

R. Kurz and P. Shaver: The ALMA Project96, 7

A First for the VLT: Observations of theGamma-Ray Burst GRB990510, and Dis-covery of Linear Polarisation 96, 11

A. Renzini: FORS1 and ISAAC Science Ver-ification at Antu/UT1 96, 12

B. Renzini: UT2/Kueyen: a Stellar Astron-omer’s Dream Comes True 96, 13

G. Monnet: VLT Instrumentation Renewal 96,15

Centerfold: Satellite image showing the pro-posed location of ALMA 96, 15

P. Ballester, A. Disaro, D. Dorigo, A. Modigliani,J.A. Pizarro de la Iglesia: The VLT DataQuality Control System 96, 19

E. Allaert: The VLT Software Workshop 96,22

M. Tarenghi: News from the VLT 97, 1E. Ettlinger, P. Giordano, and M. Schneer-

mann: Performance of the VLT Mirror Coat-ing Unit 97, 4

M. Sarazin and J. Navarrete: Climate Varia-bility and Ground-Based Astronomy: TheVLT Site Fights Against La Niña 97, 8

C. Castillo F. and L. Serrano G.: Analysis ofthe Anomalous Atmospheric Circulation inNorthern Chile During 1998 97, 10

M. Ferrari and F. Derie: Variable CurvatureMirrors 97, 11

F. Derie, E. Brunetto, and M. Ferrari: The VLTITest Siderostats Are Ready for First Light97, 12

J.C. Christou, D. Bonaccini, N. Ageorges, andF. Marchis: Myopic Deconvolution of Adap-tive Optics Images 97, 14

Latest News: “First Light” for VLT High-Res-olution Spectrograph UVES 97, 22

A. Glindemann, R. Abuter, F. Carbognani, F.Delplancke, F. Derie, A. Gennai,P. Gitton,P. Kervella, B. Koehler, S. Lévêque, G. DeMarchi, S. Menardi, A. Michel, F. Paresce,T. Phan Duc, M. Schöller, M. Tarenghi, R.Wilhelm: The VLT – The Observatory of the21st Century 98, 2

D. Bonaccini, W. Hackenberg, M. Cullum, M.Quattri, E. Brunetto, J. Quentin, F. Koch, E.Allaert, A. Van Kesteren: Laser Guide StarFacility for the ESO VLT 98, 8

W. Hackenberg, D. Bonaccini, G. Avila: VLTLaser Guide Star Facility Sub-system De-sign. Part I: Fibre Relay Module 98, 14

New Pictures from the VLT 98, 19J. Spyromilio, A. Wallander, M. Tarenghi:

Commissioning of the Unit Telescopes ofthe VLT 98, 21

A. Renzini, P. Rosati: The Science VerificationPlan for FORS2 and UVES at UT2/Kuyen98, 24

R. Gilmozzi: The First Six Months of VLT Sci-ence Operations 98, 25

F. Carbognani, G. Filippi, P. Sivera: Configu-ration Management of the Very Large Tele-scope Control Software 98, 27

M. Dolensky, M. Albrecht, R. Albrecht, P.Ballester, C. Boarotto, A. Brighton, T.Chavan, G. Chiozzi, A. Disaro, B. Kemp98, 30

THE LA SILLA NEWSPAGE

O.R. Hainaut: News from the NTT 95, 17C. Lidman: New SOFI Grisms – NTT and IR

Teams 95, 17M. Sterzik: A New Control Room for the 3.6-

m Telescope 95, 18O. Hainaut and the NTT Team: News from the

NTT 97, 24M. Sterzik, U. Weilenmann and the 3.6-m Up-

grade Team: 3.6-m Telescope Control Sys-tem Upgrade Completed 97, 25

EIS – THE ESO IMAGINGSURVEY

A. Renzini, L. Da Costa: The ESO ImagingSurvey 98, 33

L. Da Costa, S. Arnouts, C . Benoist, E.Deul, R. Hook, Y.-S. Kim, M. Nonino, E.

Pancino, R. Rengelink, R. Slijkhuis, A.Wicenec, S. Zaggia: ESO Imaging Sur-vey: Past Activities and Future Prospects98, 36

VLT DATA FLOW OPERATIONSNEWSP.A. Woudt and D. Silva: The NTT Service Ob-

serving Programme: On the Efficiency ofService Observing 95, 18

D. Baade: NGC 4945 in 3 Colours 95, 19“First Light” of UT2! 95, 24

OBSERVERS WITH THE VLT

First Scientific Results with the VLT in Visitorand Service Mode 98, 1

B. Brandl, W. Brandner, E.K. Grebel, H. Zin-necker: VLT/ISAAC and HST/WFPC2 Ob-servations of NGC 3603 98, 46

R.P. Kudritzki, R.H. Mendez, J.J. Feldmeier,R. Ciardullo, G.H. Jacoby, K.C. Freeman,M. Arnaboldi,M.Capaccioli,O.Gerhard, H.C.Ford: From Intracluster Planetary Nebulaeto High-Redshift Lyα Emitters 98, 50

C. De Breuck, W. Van Breugel, D. Minniti, G.Miley, H. Röttgering, S.A. Stanford, C. Ca-rilli: VLT Spectroscopy of the z = 4.11 Ra-dio Galaxy TN J1338–1942 98, 52

S. White: Weighing Young Galaxies – anOccasional Observer Goes to Paranal98, 54

REPORTS FROM OBSERVERS

B. Stecklum, H.-U. Käufl, A. Richichi: The Lu-nar Occultation of CW Leo – a Great Finalefor TIMMI 95, 25

L. Guzzo, H. Böhringer, P. Schuecker, C.A.Collins, S. Schindler, D.M. Neumann, S. DeGrandi, R. Cruddace, G. Chincarini, A.C.Edge, P.A. Shaver, W. Voges: The ReflexCluster Survey: Observing Strategy andFirst Results on Large-Scale Structure95, 27

J.U. Fynbo, B. Thomsen, P. Møller: NTTService Mode Observations of the Ly-man-Limit Absorber towards Q1205–3095, 32

C. Aerts, P. de Cat and L. Eyer: Long-TermSpectroscopic Monitoring of Pulsating BStars: a Tribute to the CAT 96, 23

H. Lamy and D. Hutsemékers: AProcedure forDeriving Accurate Linear Polarimetric Mea-surements 96, 25

F. Courbin, P. Magain, S. Sohy, C. Lidman, G.Meylan: Deconvolving Spectra of LensingGalaxies, QSO Hosts, and More… 97, 26

P. Leisy, P. Francois, and P. Fouqué: Emission-Line Object Survey in the LMC with the WFI:New Faint Planetary Nebulae 97, 29

OTHER ASTRONOMICAL NEWS

H.W. Duerbeck, D.E. Osterbrock, L.H. BarreraS., R. Leiva G.: Halfway from La Silla toParanal – in 1909 95, 34

B. Nordström: The Questionnaire Survey for“La Silla 2000+” 95, 38

MESSENGER INDEX 1999 (Nos . 95–98)

S U B J E C T I N D E X

AC. Aerts, P. de Cat and L. Eyer: Long-Term

Spectroscopic Monitoring of Pulsating BStars: a Tribute to the CAT 96, 23

N. Ageorges and N. Hubin: Monitoring of theAtmospheric Sodium above La Silla 95, 12

E. Allaert: The VLT Software Workshop96, 22

BD. Baade, K. Meisenheimer, O. Iwert, J. Alon-

so, Th. Augusteijn, J. Beletic, H. Bellemann,W. Benesch, A. Böhm, H. Böhnhardt, J.Brewer, S. Deiries, B. Delabre, R. Donald-son, Ch. Dupuy, P. Franke, R. Gerdes, A.Gilliotte, B. Grimm, N. Haddad, G. Hess,G. Ihle, R. Klein, R. Lenzen, J.-L. Lizon, D.Mancini, N. Munch, A. Pizarro, P. Prado, G.Rahmer, J. Reyes, F. Richardson, E. Ro-bledo, F. Sanchez, A. Silber, P. Sinclaire,R. Wacker-mann, S. Zaggia: The WideField Imager at the 2.2-m MPG/ESO Tele-scope: First Views with a 67-Million-Facette Eye 95, 15

D. Baade: NGC 4945 in 3 Colours 95, 19P. Ballester, A. Disaro, D. Dorigo, A. Modigliani,

J.A. Pizarro de la Iglesia: The VLT DataQuality Control System 96, 19

J. Bergeron: The VLT Opening Symposium96, 6

J. Bergeron and U. Grothkopf: Publications inRefereed Journals Based on TelescopeObservations 96, 28

D. Bonaccini, W. Hackenberg, M. Cullum, M.Quattri, E. Brunetto, J. Quentin, F. Koch, E.Allaert, A. Van Kesteren: Laser Guide StarFacility for the ESO VLT 98, 8

B. Brandl, W. Brandner, E.K. Grebel, H. Zin-necker: VLT/ISAAC and HST/WFPC2 Ob-servations of NGC 3603 98, 46

CF. Carbognani, G. Filippi, P. Sivera: Configu-

ration Management of the Very Large Tele-scope Control Software 98, 27

C. Castillo F. and L. Serrano G.: Analysis ofthe Anomalous Atmospheric Circulation inNorthern Chile During 1998 97, 10

J.C. Christou, D. Bonaccini, N. Ageorges, andF. Marchis: Myopic Deconvolution of Adap-tive Optics Images 97, 14

F. Courbin, P. Magain, S. Sohy, C. Lidman, G.Meylan: Deconvolving Spectra of LensingGalaxies, QSO Hosts, and More... 97, 26

DL. Da Costa, S. Arnouts, C. Benoist, E. Deul,

R. Hook, Y.-S. Kim, M. Nonino, E. Panci-no, R. Rengelink, R. Slijkhuis, A. Wicenec,S. Zaggia: ESO Imaging Survey: Past Ac-tivities and Future Prospects 98, 36

C. De Breuck, W. Van Breugel, D. Minniti, G.Miley, H. Röttgering, S.A. Stanford, C. Ca-rilli: VLT Spectroscopy of the z = 4.11 Ra-dio Galaxy TN J1338–1942 98, 52

M. Dennefeld: The NEON Observing School98, 57

F. Derie, E. Brunetto, and M. Ferrari: The VLTITest Siderostats Are Ready for First Light97, 12

N. Devillard, Y. Jung, J.-G. Cuby: ISAAC Pipe-line Data Reduction 95, 5

M. Dolensky, M. Albrecht, R. Albrecht, P. Bal-lester, C. Boarotto, A. Brighton, T. Chavan,G. Chiozzi, A. Disaro, B. Kemp 98, 30

H.W. Duerbeck, D.E. Osterbrock, L.H. BarreraS., R. Leiva G.: Halfway from La Silla toParanal – in 1909 95, 34

EE. Ettlinger, P. Giordano, and M. Schneer-

mann: Performance of the VLT Mirror Coat-ing Unit 97, 4

FM. Ferrari and F. Derie: Variable Curvature Mir-

rors 97, 11Discourse of His Excellency the President of

the Republic of Chile, Don Eduardo FreiRuiz-Tagle 96, 5

J.U. Fynbo, B. Thomsen, P. Møller: NTT Ser-vice Mode Observations of the Lyman-Lim-it Absorber towards Q1205–30 95, 32

GIntroduction to the ESO Annual Report 1998

by Prof. Riccardo Giacconi, Director Gen-eral of ESO 96, 2

Discourse of Prof. Riccardo Giacconi, Direc-tor General of ESO 96, 2

R. Gilmozzi: The First Six Months of VLT Sci-ence Operations 98, 25

A. Glindemann, R. Abuter, F. Carbognani, F.Delplancke, F. Derie, A. Gennai, P. Gitton,P. Kervella, B. Koehler, S. Lévêque, G. DeMarchi, S. Menardi, A. Michel, F. Paresce,T. Phan Duc, M. Schöller, M. Tarenghi, R.Wilhelm: The VLT – The Observatory of the21st Century 98, 2

Discourse of Mr. Grage, President of the ESOCouncil 96, 4

L. Guzzo, H. Böhringer, P. Schuecker, C.A.Collins, S. Schindler, D.M. Neumann, S. DeGrandi, R. Cruddace, G. Chincarini, A.C.Edge, P.A. Shaver, W. Voges: The ReflexCluster Survey: Observing Strategy and FirstResults on Large-Scale Structure 95, 27

HW. Hackenberg, D. Bonaccini, G. Avila: VLT

Laser Guide Star Facility Sub-system De-sign. Part I: Fibre Relay Module 98, 14

O.R. Hainaut: News from the NTT 95, 17O. Hainaut and the NTT Team: News from the

NTT 97, 24K

A. Kaufer, O. Stahl, S. Tubbesing, P. Nørre-gaard, G. Avila, P. François, L. Pasquini, A.Pizzella: Commissioning FEROS, the NewHigh-resolution Spectrograph at La Silla95, 8

R.P. Kudritzki, R.H. Méndez, J.J. Feldmeier,R. Ciardullo, G.H. Jacoby, K.C. Freeman,M. Arnaboldi, M. Capaccioli, O. Gerhard,H.C. Ford: From Intracluster Planetary Neb-ulae to High-Redshift Lyα Emitters 98, 50

R. Kurz and P. Shaver: The ALMA Project96, 7

LH. Lamy and D. Hutsemékers: AProcedure for

Deriving Accurate Linear Polarimetric Mea-surements 96, 25

P. Leisy, P. François, and P. Fouqué: Emis-sion-Line Object Survey in the LMC withthe WFI: New Faint Planetary Nebulae97, 29

C. Lidman: New SOFI Grisms – NTT and IRTeams 95, 17

MC. Madsen: ESO at the Hannover Fair

96, 29C. Madsen: Science with the Atacama Mil-

limetre Array 98, 56G. Monnet: VLT Instrumentation Renewal

96, 15A. Moorwood, J.-G. Cuby, P. Ballester, P. Bier-

eichel, J. Brynnel, R. Conzelmann, B. De-labre, N. Devillard, A. Van Dijsseldonk, G.Finger, H. Gemperlein, C. Lidman, T. Her-lin, G. Huster, J. Knudstrup, J.-L. Lizon, H.Mehrgan, M. Meyer, G. Nicolini, A. Silber,J. Spyromilio, J. Stegmeier: ISAAC at theVLT 95, 1

NB. Nordström: The Questionnaire Survey for

“La Silla 2000+” 95, 38B. Nordström: The La Silla 2000+ Report Now

Available 98, 57

RA.Renzini: FORS1 and ISAAC Science Veri-

fication at Antu/UT1 96, 12A. Renzini: UT2/Kueyen: a Stellar Astron-

omer’s Dream Comes True 96, 13A. Renzini, P. Rosati: The Science Verification

Plan for FORS2 and UVES at UT2/Kuyen98, 24

A. Renzini, L. Da Costa: The ESO ImagingSurvey 98, 33

59

View of Northern Chile 96, 1Translation of the Speech by H.E. the Presi-

dent of the Republic of Chile, Don Eduar-do Frei Ruiz-Tagle, at the Inauguration ofthe Paranal Observatoy, 5th March 199996, 28

J. Bergeron and U. Grothkopf: Publications inRefereed Journals Based on Telescope Ob-servations 96, 28

C. Madsen: ESO at the Hannover Fair 96, 29C. Madsen: Science with the Atacama Mil-

limetre Array 98, 56

ANNOUNCEMENTS

Catherine Cesarsky – ESO’s Next DirectorGeneral 95, 38

ESO Studentship Programme 95, 39Personnel Movements 95, 39List of Scientific Preprints 95, 39Minor Planet Mariotti 95, 40Personnel Movements 96, 30ESO Fellowship Programme 2000 96, 31Contributed Software by Observers in the ESO

Community 96, 31List of Scientific Preprints (April–June 1999)

96, 31New ESO Proceedings Available 96, 32Personnel Movements 97, 31List of Scientific Preprints 97, 32B. Nordström: The La Silla 2000+ Report Now

Available 98, 57M. Dennefeld: The NEON Observing School

98, 57Personnel Movements (1st October – 30th No-

vember 1999) 98, 57

A U T H O R I N D E X

60

SM. Sarazin and J. Navarrete: Climate Variability

and Ground-Based Astronomy: The VLTSite Fights Against La Niña 97, 8

J. Spyromilio, A. Wallander, M. Tarenghi: Com-missioning of the Unit Telescopes of the VLT98, 21

B. Stecklum, H.-U. Käufl, A. Richichi: The Lu-nar Occultation of CW Leo – a Great Finalefor TIMMI 95, 25

M. Sterzik, U. Weilenmann and the 3.6-mUpgrade Team: 3.6-m Telescope ControlSystem Upgrade Completed 97, 25

M. Sterzik: A New Control Room for the3.6-m Telescope 95, 18

TM. Tarenghi: News from the VLT 97, 1

WS. White: Weighing Young Galaxies – an

Occasional Observer Goes to Paranal98, 54

P.A. Woudt and D. Silva: The NTT Service Ob-serving Programme: On the Efficiency ofService Observing 95, 18

ESO, the European Southern Observatory,was created in 1962 to “… establish and op-erate an astronomical observatory in thesouthern hemisphere, equipped with pow-erful instruments, with the aim of furtheringand organising collaboration in astronomy…” It is supported by eight countries: Bel-gium, Denmark, France, Germany, Italy, theNetherlands, Sweden and Switzerland. ESOoperates at two sites. It operates the La Sil-la observatory in the Atacama desert, 600km north of Santiago de Chile, at 2,400 maltitude, where several optical telescopeswith diameters up to 3.6 m and a 15-m sub-millimetre radio telescope (SEST) are nowin operation. In addition, ESO is in the pro-cess of building the Very Large Telescope(VLT) on Paranal, a 2,600 m high mountainapproximately 130 km south of Anto-fagasta, in the driest part of the Atacamadesert. The VLT consists of four 8.2-metreand three 1.8-metre telescopes. These tele-scopes can also be used in combination asa giant interferometer (VLTI). The first 8.2-metre telescope (called ANTU) is since April1999 in regular operation, and also the sec-ond one (KUEYEN) has already deliveredpictures of excellent quality. Over 1200 pro-posals are made each year for the use ofthe ESO telescopes. The ESO Headquar-ters are located in Garching, near Munich,Germany. This is the scientific, technical andadministrative centre of ESO where tech-nical development programmes are carriedout to provide the La Silla and Paranal ob-servatories with the most advanced instru-ments. There are also extensive astro-nomical data facilities. In Europe ESO em-ploys about 200 international staff members,Fellows and Associates; in Chile about 70and, in addition, about 130 local staff mem-bers.

The ESO MESSENGER is published fourtimes a year: normally in March, June, Sep-tember and December. ESO also publish-es Conference Proceedings, Preprints,Technical Notes and other material con-nected to its activities. Press Releases in-form the media about particular events. Forfurther information, contact the ESO Edu-cation and Public Relations Department atthe following address:

EUROPEAN SOUTHERN OBSERVATORYKarl-Schwarzschild-Str. 2 D-85748 Garching bei München Germany Tel. (089) 320 06-0 Telefax (089) 3202362ips@eso.org (internet) URL: http://www.eso.org

The ESO Messenger:Editor: Marie-Hélène DemoulinTechnical editor: Kurt Kjär

Printed by J. Gotteswinter GmbHBuch- und OffsetdruckJoseph-Dollinger-Bogen 22D-80807 MünchenGermany

ISSN 0722-6691

ContentsFirst Scientific Results with the VLT in Visitor and Service Modes . . . . . . . . . 1

TELESCOPES AND INSTRUMENTATION

A. Glindemann, R. Abuter, F. Carbognani, F. Delplancke, F. Derie, A. Gennai,P. Gitton, P. Kervella, B. Koehler, S. Lévêque, G. De Marchi, S. Menardi,A. Michel, F. Paresce, T. Phan Duc, M. Schöller, M. Tarenghi, R. Wilhelm:The VLTI – The Observatory of the 21st Century . . . . . . . . . . . . . . . . . . . 2

D. Bonaccini, W. Hackenberg, M. Cullum, M. Quattri, E. Brunetto, J. Quentin,F. Koch, E. Allaert, A. Van Kesteren: Laser Guide Star Facility for the ESOVLT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

W. Hackenberg, D. Bonaccini, G. Avila: VLT Laser Guide Star Facility Sub-system Design. Part I: Fibre Relay Module . . . . . . . . . . . . . . . . . . . . . . . 14

New Pictures from the VLT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19J. Spyromilio, A. Wallander, M. Tarenghi: Commissioning of the Unit

Telescopes of the VLT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21A. Renzini, P. Rosati: The Science Verification Plan for FORS2 and UVES at

UT2/Kuyen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24R. Gilmozzi: The First Six Months of VLT Science Operations . . . . . . . . . . . . 25F. Carbognani, G. Filippi, P. Sivera: Configuration Management of the Very

Large Telescope Control Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27M. Dolensky, M. Albrecht, R. Albrecht, P. Ballester, C. Boarotto, A. Brighton,

T. Chavan, G. Chiozzi, A. Disarò, B. Kemp: Java for Astronomy: SoftwareDevelopment at ESO/ST-ECF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

EIS – THE ESO IMAGING SURVEY

A. Renzini, L. Da Costa: The ESO Public Imaging Survey . . . . . . . . . . . . . . . 33L. Da Costa, S. Arnouts, C. Benoist, E. Deul, R. Hook, Y.-S. Kim, M. Nonino,

E. Pancino, R. Rengelink, R. Slijkhuis, A. Wicenec, S. Zaggia:ESO Imaging Survey: Past Activities and Future Prospects . . . . . . . . . . . 36

OBSERVERS WITH THE VLT

B. Brandl, W. Brandner, E.K. Grebel, H. Zinnecker: VLT/ISAAC andHST/WFPC2 Observations of NGC 3603 . . . . . . . . . . . . . . . . . . . . . . . . . 46

R.P. Kudritzki, R.H. Mendez, J.J. Feldmeier, R. Ciardullo, G.H. Jacoby,K.C. Freeman, M. Arnaboldi, M. Capaccioli, O. Gerhard, H.C. Ford: FromIntracluster Planetary Nebulae to High-Redshift Lyα Emitters . . . . . . . . . 50

C. De Breuck, W. Van Breugel, D. Minniti, G. Miley, H. Röttgering,S.A. Stanford, C. Carilli: VLT Spectroscopy of the z = 4.11 Radio GalaxyTN J1338–1942 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

S. White: Weighing Young Galaxies – an Occasional Observer Goes toParanal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

OTHER ASTRONOMICAL NEWS

C. Madsen: Science with the Atacama Millimetre Array . . . . . . . . . . . . . . . . . 56

ANNOUNCEMENTS

B. Nordström: The La Silla 2000+ Report Now Available . . . . . . . . . . . . . . . . 57M. Dennefeld: The NEON Observing School . . . . . . . . . . . . . . . . . . . . . . . . . 57Personnel Movements (1st October – 30th November 1999) . . . . . . . . . . . . . 57

MESSENGER INDEX 1999 (Nos. 95–98) . . . . . . . . . . . . . . . . . . . . . . 58