1

No. 99 – March 2000

Successful Commissioning of UVES at KueyenFinally, we are there: as of April 1,

2000, UVES, the UV-Visual EchelleSpectrograph built by ESO, will start op-eration at the Nasmyth focus of the VLTtelescope Kueyen. The instrument com-missioning has been completed inDecember 1999, eight years after thefirst proposal to build a high-resolutionspectrograph for the VLT was circulatedand in perfect schedule with the first de-tailed VLT planning dated March 1994.UVES will be the third instrument afterFORS1 and ISAAC to enter into regularuse at the VLT. The second version ofFORS, FORS2, will start operating at thesame date at the Cassegrain focus ofthe telescope.

The figure to the right is an un-processed section of a 1-hour integra-tion in the blue arm showing the centralportion of a few echelle orders centred at380 nm. It clearly illustrates some ofUVES prime capabilities: the UV-blue ef-ficiency and the image quality of the at-mosphere/telescope/spectrograph sys-tem (see page 2).

The two parallel tracings correspondto the two images of a gravitationallylensed QSO (HE1104-1805) separatedon the sky by 3.2 arcsec, blue magni-tudes ~ 16.7 and 18.6 respectively. TheCCD was read-out in binned 2 × 2 mode.The vertical line at the bottom is a sky-emission line and it visualises well thespectral resolution of ~ 55,000. The red-shifts, widths and equivalent widths ofthe absorption lines along the 2 lines ofsight provide a mini-“tomography” of theintergalactic/interstellar medium up tothe distance corresponding to z = 2.1.Some of the broad absorption lines(Lyα) and the narrow metal absorption(e.g. in the second order from the bot-tom) reveal column density and velocityvariations over a scale of a few kpc.

Introduction

The ESO scientific community has arich tradition of research based on thedetailed analysis of high-resolutionspectra of stars. One of the very firstspectrographs to enter into operation atESO was in 1968 the coudé high-reso-lution spectrograph at the 1.52-m at LaSilla, followed in 1973 by the Echelec atthe same telescope.

The first ESO-built echelle spectro-graph for the 3.6-m, CASPEC, was in-stalled in 1983 and opened the possi-bility to study faint stellar and extra-galactic sources, thanks to its new (atthat time almost exotic) CCD detector(a high efficiency 320 × 512 RCA CCD,with a read-out-noise of 40 e–/pixel anda dark current of 15 e–/hr/pixel). Forthe first time, the ESO community hadaccess to an instrument that was fullycompetitive with, and in some areassuperior to, other high-resolution spec-

trographs at large telescopes world-wide. CASPEC was followed by theCES spectrograph coupled to the 1.5-mCAT giving resolving powers larger than100,000 on stars down to magnitudes~ 10 and in 1991 by the echelle modeof the EMMI spectrograph at the NTT,which again offered optimal perform-ance in R = 25,000 spectroscopy offaint sources in the visual-red spectralregion.

In 1993, the HIRES echelle spectro-graph came into operation at the first10-m Keck telescope. The excellentquality of the instrument and the col-lecting power of the largest telescopeever built successfully combined toform a unique tool for all programmes,which require spectra of faint targets atresolution up to 50,000 in the 400–800nm range. ESO astronomers workingon research topics, which rely on high-resolution spectra, for both stellar andextragalactic targets, often had to face

a powerful, almost unbeatable competi-tion or to concentrate on objects at dec-linations lower than –40 degrees. Aftersix years of justified frustration, theESO community has now access to aninstrument, UVES, which is more effi-cient, provides larger spectral coveragein a single exposure, and can reachhigher resolving power with propersampling than its main competitor in itspresent configuration.

Not surprisingly, in the first semesterof Kueyen observations (Period 65,starting April 1, 2000), about 70% of thetime has been assigned to UVES obser-vations. The Observing ProgrammesCommittee has selected programmesfor 78 nights in UVES visitor mode andfor 312 hours of service observations.Scanning through the titles of theseapproved programmes, we find mostof the research topics identified asscientific drivers of the instrument whenit was first proposed. Going from thevery close to the distant universe,UVES observations will aim at highlyaccurate radial velocity measurementsof nearby stars to search for associatedplanets, at the determination of theabundance of various critical elementsin the atmospheres of stars in theGalaxy and in nearby systems and atthe study of absorption systems downto the atmospheric cut-off in QSO’sspectra.

The Instrument Layout

There are a few basic choices takenearly in the project which have beencrucial in determining its present goodperformance: the configuration fixedwith respect to gravity (advantages:less weight and space restrictions,more stability, shortcomings: need ofderotator and more relay optics), thesplitting of the optical path in a UV-blueand in a visual-red arm (giving the pos-sibility to optimise the efficiency overthe entire spectral range from 300 to1000 nm) and the early selection of de-tectors of a format (2k × 4k, 15 µm pix-els) for which we had to find a supplierwho would deliver to specifications. Theoptical design of both arms is of the

2

TELESCOPES AND INSTRUMENTATION

UVES at Kueyen: Bright Prospects for High-Resolution Spectroscopy at the VL TS. D’ODORICO, ESO

After the successful completion of the commissioning period, the UV-Visual Echelle Spectrograph startsregular operation at the Paranal Observatory.

Figure 1: This areal view of the Nasmyth platform of Kueyen at the end of the integrationshows UVES with the top of its enclosure partially lifted to have access to the table where thevarious components are mounted. The bar connecting the enclosure to the telescope fork isfor safety in case of earthquake.

white pupil type. To maximise the reso-lution while keeping the beam size rea-sonable (200 mm) we went for two 214× 840 mm mosaics (each made by tworeplicas on a single blank) R4 echellegratings, the first ones of this type andsize ever produced. The design ofUVES is fully ESO made, the variouscomponents were produced in Europeand the USA: the optics in France, themechanics in Germany and Switzer-land, the gratings in the USA and Rus-sia, the detectors in the UK and USAand most of the high-level software inItaly at the Observatory of Trieste.

Two main parts compose UVES (seeFig. 1). The preslit area is attached tothe Nasmyth rotator and includes thecalibration unit with the arc lamp anddifferent FF lamps for each spectralrange, an insertable iodine cell, a slidemounting three image slicers for obser-vations at the highest resolution inmediocre seeing conditions and thederotator. Along the optical path, nowon the steel table bolted to the Nasmythplatform, the beam encounters the at-mospheric dispersion corrector, a depo-larise slide, a variable pupil stop andthe arm selector unit, which can feedthe two arms individually or in parallelwith dichroics. The blue and red slitunits are adjustable in width and height,they reflect the light over a field of 30arcsec diameter to two CCDs whichare used for target acquisition andcentring, for monitoring the telescopeguiding and for recording the slit po-sition on the sky. After the slits, eachof the two parallel arms (which in-tersect each other to minimise the over-all volume) includes an order sorter fil-ter wheel, the collimator mirrors, theechelle grating, the exposimeter andthe cross-disperser unit with two grat-ings mounted back to back. The cam-eras are dioptric with an external focusto facilitate detector exchange; thelargest lenses are CaF2 (220 mm di-

ameter) and SFPL51 (246 mm) in theblue and red respectively. The blue-armCCD is an EEV-44 device with en-hanced UV efficiency (55% at 340nm).The red detector is a mosaic of oneEEV-44 device and one MIT-LL CCID-20 device, to optimise the spectral re-sponse with wavelength. The CCDs areoperated by the ESO-built FIERA con-troller. In the configurations they areoffered in UVES, both detectors areread out in ~ 40 s with a rather goodr.o.n. of 2 and 4 e– r.m.s. The operatingtemperature of the CCDs (~ 150° K) ismaintained by liquid nitrogen fed from atank which secures an autonomy of atleast 10 days. The table is protectedfrom dust and light by a motorised en-closure that can be lifted to give accessto the functions. It provides a passivethermal insulation which, combinedwith the air conditioning of the tele-scope enclosure during the day,smoothes out the temperature varia-tions inside the instrument.

A more detailed description of thespectrograph and its main componentscan be found in the UVES User Manualand in Dekker et al. (Proceedings of theSPIE Conference 4008, Munich, 2000).

Who is Who in the UVESProject?

The table includes the names of theengineers, technicians and astron-omers who have contributed to the de-sign, building and testing of the varioussubsystems and of the instrument in thelast seven years. This serves as recog-nition of a job well done and also as areminder of the different expertises thatare needed to complete and put suc-cessfully into operation a complex in-strument at the VLT. Starting from theoptical designer to the astronomerswho verify the quality of the first astro-nomical data, from the skilled techni-cians who integrate and test the opto-mechanical functions to the softwarespecialists who wrote more than140,000 lines of code for instrumentcontrol, all had to complete their taskproperly and in schedule for the instru-ment to be successful. A total of ap-proximately 40 person-years and 6.7MDM went into project.

UVES through Commissioning

Hardware and software were first puttogether, tested and optimised in theESO Integration Laboratory in Gar-ching. The results in the laboratoryconfirmed the quality of the optics, thecapability to reach the specified re-solving power, and the robust, reliablebehaviour of the electro-mechanics andof the software. The tests were com-pleted in May 1999, the instrument wasthen fully dismounted, its hundreds ofcomponents properly packed and sentpart by plane, part by ship to Chile. At

3

Figure 2: During the frantic days of the instrument integration at the telescope, J.L. Lizontakes advantage of the robust design of the UVES functions for a short rest. Most of the sub-systems are already mounted on the table fixed to the Nasmyth platform: from the left thepreslit units, the shiny back of the blue echelle mount, the blue CD unit, camera and CCD.

TABLE 1. THE UVES TEAM

Project Manager, Optical Engineering: H. Dekker

Instrument Scientist: S. D’Odorico

Optical Design: B. Delabre

Mechanical Engineering and Design: H. Kotzlowski, G. Hess

Control Electronics: S. Moureau

Control Software: A. Longinotti, P. Santin and P. Dimarcantonio (Obs.Trieste), R. Schmutzer

CCD Detector Integration and Testing: R. Dorn, C. Cumani

Opto-mechanical Integration and Testing, Cryogenics: J.L. Lizon à l’Allemand, C. Dupuy, A. Silber

Data Flow System (Pipeline, Instrument P. Ballester, O. Boitquin, M. Chavan,Model and ETC, P2PP): A. Modigliani, S. Wolf

Testing in Europe, Commissioning,Calibration and Operation at Paranal: A. Kaufer

Astronomical Support, Documentation, S. Cristiani, V. Hill, L. Kaper, T. Kim, Data Reduction, Testing of Pipeline: F. Primas

(All from ESO, except where indicated differently)

Figure 4: The over-all efficiency ofUVES derived fromobservations of thestandard stars EG21and Fei 110 takenwith a wide openslit. The values referto the top of theblaze function ineach order and havebeen obtained fromobservations withtwo dichroic stan-dard settings. Theyare corrected forlosses in the atmos-phere and in thethree-mirror reflec-tions of the tele-scope. The de-crease of efficiencytoward the UV andthe far-infrared are mostly due to the lower efficiency of the CCDs and of the cross-dispers-er gratings at these wavelengths. Both components could be easily substituted with new onesof higher performance, when they become available.

the Observatory, first the table was in-stalled on the Nasmyth platform of UT2,the Kueyen Telescope, and then thecomplex layout of optics, mechanics,cables, detectors was reassembled andthe optical path re-aligned (Fig. 2). Wewent through this usually critical phase(it is the time one discovers whetherproper care has been taken of the in-terfaces with the telescope) with aminimum of bad surprises: a few holeswere not in the right places, the tableplane a few millimetres below the ex-pected height. The optics, and espe-cially the large lenses of the camerashad survived the loading from the planeand the bumpy road from Antofagastato Paranal but the blue camera didshow a degraded optical quality andhad to be dismounted and re-aligned. Inthe last week of September, we were fi-nally ready and eager to verify whetherthe operation of UVES at the telescopeand its overall efficiency on the skywere in line with the model prediction. Acrucial point to check was the acquisi-tion and guiding of the targets on the slitplane and the parallel operation of thetwo arms in the dichroic modes.

The first stellar photons entered thespectrograph on September 26. JasonSpyromilio and Anders Wallander hadjust concluded the commissioning onthe Kueyen telescope with the TestCamera reporting record performancein image quality and tracking perform-ance. We had already gone through ex-tensive testing and optimisation withthe calibration lamps and knew that theinstrument’s optical quality was basical-ly all right. The first target was a fluxstandard star. From the very beginningwe were ready to carry out target ac-

quisition, instrument setting, observa-tions and archiving of the data startingfrom Observation Blocks prepared inadvance with the instrument templates.We launched the first OB and the starlanded within one arcsecond from bothslits and was centred without problems(except the mandatory wrong sign inone of the formula to convert pixels tocoordinates) using the slit viewers. Onthe same night, a quick analysis of thespectra produced an overall efficiencythat is very close to the predicted val-ues at all wavelengths but in the far-redregion. The following night, on Sep-tember 27 (the official first light) wasblessed by a seeing between 0.6 and0.4 for 90% of the time and we could

confirm the capability of UVES on thefirst scientific targets (see the ESO15/99 press release on the ESO Webpages). At that point, we knew that withUVES+Kueyen, we had on line themost powerful combination for high-res-olution spectroscopy available to theastronomical community worldwide. Itwas, and still is, a nice feeling whichhelped us to go through the subse-quent less exciting but necessary threeweeks of testing and calibration of allthe instrument modes, of the acquisi-tion procedures and of the software in-terfaces.

The smoothness of the commission-ing of UVES is best testified by the verylow number of hours lost due to techni-cal problems of the telescope or instru-ment: around 7 in total over threeweeks of continuous operation, ofwhich three were due to the suddendeath of the power supplier of the in-strument WS, three to a failure of thetelescope M2 unit, due to a gust of windwell above the safe operation condi-tions and one to a rebooting of the in-strument control software.

A substantial chunk of data of scien-tific value taken for Commissioning hasbeen released with their calibration forpublic use. They correspond to morethan 94 hours of integration on the sky.

A second, shorter commissioning pe-riod in December was mainly dedicatedto the optimisation of the blue arm opti-cal alignment, to the test of the iodinecell and the final editing of the instru-ment observing templates.

The performance of the instrument atthe time of the handing-over to the Ob-servatory in December 1999 is sum-marised in Table 2. An additional im-portant parameter is the stability of thewavelength calibration. When allowanceis made for the variation of the refrac-tive index of air with temperature andpressure (both recorded in the fileheaders), the velocity stability of the

4

Figure 3: After the successful first light, part of the UVES team proudly poses close to the in-strument with its partially lifted enclosure. From the left, back: A. Kaufer, C. Dupuy; front: A.Longinotti, P. Santin, P. Dimarcantonio, H. Dekker, S. Moureau and R. Schmutzer. The almosttotal compliance with the Observatory safety regulations, even at a time of undisputed eu-phoria, is worth praise.

instrument over days was found tobe of the order of 50 m/s. The useof the iodine cell, not yet tested in full,should further lower this limit to a fewmeters/sec but it requires an effort by a

specialised team on the data-reductionsoftware.

While the overall status of the instru-ment at the end of Commissioning isvery satisfactory, there are a few pend-

ing problems on which we need towork. The tracking of targets observedwith the image slicers was found to re-quire an unplanned modification of theTCS. The Cross-Dispersers #1 and #4

5

Figure 5: This plot, a by-product of the automatic data-reduction pipeline running for the instrument standard wavelength settings, maps theinstrument resolution from measurements of the FWHM widths of the lines of the ThA lamp distributed over the whole CCD (see subplot bot-tom-right). Resolution and FWHM in pixels for each line are given as a function of position on the chip and wavelength. This particular dia-gram refers to a spectrum obtained with a 0.5-arcsec-wide slit and to the region covered by the EEV CCD-44 chip in the red arm.

TABLE 2. UVES OBSERVING CAPABILITIES AND MEASURED PERFORMANCE

Blue Arm Red Arm

Wavelength range 300 –500 nm 420 –1100 nm

Echelle 41.59 g/mm, R4 31.6 g/mm, R42 mosaicked replicas on a Zerodur block 2 mosaicked replicas on a Zerodur block

Cross-dispersers CD1: 1000 g/mm, λb 430 nm CD3: 600 g/mm, λb 560 nmCD2: 660 g/mm, λb 460 nm CD4: 312 g/mm, λb 770 nm

CCD format and pixel scale 2048 × 4096, 4096 × 4096,⊥ disp (1 pixel = 15 µm) windowed to 2k × 3k (.25“/pix) 2 x 1 mosaic (.18”/pix)

Resolution-slit product/wavelength bin 41400 38700 0.0019 nm at 450 nm 0.0025 nm at 600 nm

Max. resolution 80,000 (0.4” slit or IS) 115,000 (0.3” slit or IS)

Throughput (TEL+UVES, no slit, 10 % at 400 nm 12% at 600 nmno atmosphere)

Limiting magnitude (90m. exp., 18 (R = 58,000) at 360 nm 19.2 (R = 62,000) at 600 nmS/N =10, 0.7” slit & seeing)

λλ /frame,CD1, CD3 85 nm in 33 orders 200 nm in 37 orders CD2, CD4 126 nm in 31 orders 403nm in 33 orders

Order separation (minimum) 10“ ↔ 40 pixels 12“ ↔ 70 pixels

are not the ones ordered for the instru-ment but prototypes of inferior quality.This results in a lower efficiency espe-cially in the far red and in stronger opti-cal ghosts in the UV and far red (in themost unfavourable configurations up toa few per cent of the primary spectrum).The final gratings should be installed bythe end of this year.

In February 2000, UVES was usedin service mode for nine nights ofScience Verification on a variety ofscientific programmes (see for detailshttp://www.eso.org/science/ut2sv).

Everything went smoothly and theweather co-operated: a total of morethan 70 hours of integration mostly inexcellent seeing conditions were suc-

cessfully completed. The data will bereleased to the community in latespring.

Acknowledgements

The UVES project team gratefully ac-knowledges the steady support of manyESO staff, in particular C. Nieuwen-kamp and G. Wieland (Contracts andProcurement) and E. Zuffanelli (INS) inGarching and of the Paranal staff P.Gray, G. Gillet, G. Rahmer and P.Sansgasset during the instrument in-tegration. Special thanks are due to J.Spyromilio who was there to help, ad-vise and support when requested,throughout the entire commissioning.

In the framework of a long-standingcollaboration with ESO, the Observa-tory of Trieste provided a substantial,highly professional contribution to theproject with the work of P. Santin andP. Dimarcantonio on the instrumentcontrol, observing and maintenancesoftware.

We are particularly grateful to themembers of the UVES Science TeamB. Gustafsson (Uppsala Observatory),H. Hensberge (R.O.B., Brussels), P.Molaro (Osservatorio di Trieste) andP.E. Nissen (Aarhus University) for theiradvice and the support to the projectthroughout its development.

6

E-mail: sdodoric@eso.org

VLT Pipeline Operation and Quality Control: FORS1 and ISAACR. HANUSCHIK and P. AMICO

Quality Control, ESO

1. Introduction

It is well known in the community thatApril 1, 1999, marked the beginning ofoperations for ANTU, the first VLT tele-scope. It is less well known, however,that this date also marked the begin-ning of real life for VLT data-flow oper-ations (DFO) at ESO Headquarters inGarching. Forming the back end of theData-Flow life cycle, DFO has to act asdata-production, data-distribution anddata-storage machine. All these func-tions form what is called Quality Control(QC). For achieving data products ofthe highest possible quality, all compo-nents have to perform well and collabo-rate closely.

At the moment of writing this, twoVLT instruments are operational:FORS1 and ISAAC, with the next twowaiting at the front door (FORS2 andUVES). We will briefly describe in thefollowing the different QC tasks for thetwo first VLT instruments.

2. General Functions

The most important tasks for QCGarching are:

• commission the instrument data-re-duction pipeline,

• create master calibration data andcalibration solutions,

• reduce science data,• sort and distribute all kinds of data

(raw, reduced, calibration, logs, list-ings) for the Service Mode package,

• check the quality of processed data,• provide instrument health checks,• perform trend analysis of quality

parameters.

Present ESO strategy for processingand distributing data is as follows: cali-bration data are processed irrespectiveof the observing mode (both Visitorand Service Mode), science data areprocessed for Service Mode (SM) ob-serving only. SM programmes (of sup-ported instrument modes) receive a fullset of raw, reduced and calibration data.Processed calibration data will becomegenerally available as soon as theArchive storage project has been re-alised.

Hence a Visitor Mode (VM) night,from the QC point of view, requires onlyprocessing of calibration data, while aSM night needs the full machinery pro-ducing master calibration and reducedscience files. For a typical 50:50 mix ofSM/VM nights and an average QC fish1

with 4-days-per-week duty, there ispresently about one QC working dayper ANTU operational night available.Any time more than that will produce abacklog.

3. FORS1

Data. Being a complex instrumentwith many different modes, FORS1 pro-duces data from the very beginning ofoperations in a huge amount and vari-ety. Period 63 produced about 24,000raw FORS1 files – about 200 GB – ,about half of them in Service Mode.68.0% of all raw files were calibrationdata, 17.9% science data, the rest

(12.7%) TEST data (acquisition, slitview, etc.). The vast majority of allFORS1 files (70.9%) was obtained inimaging mode (IMG), the secondlargest fraction (18.4%) in multi-objectspectroscopy mode (MOS), 5.2% inlong-slit spectroscopy (LSS), 3.9% inpolarisation imaging (IPOL) or polarisa-tion MOS (PMOS). Typically 100–200files are produced per SM night whichcorrespond to 1–1.5 GB of raw data re-sulting in another 1–1.5 GB of reduceddata.

Pipeline operations . Due to thecomplexity of the task, we decided tostart pipeline operations with the sim-plest modes, IMG and LSS. These to-gether cover already 76% of all FORS1data. Master calibration files routinelycreated are:

• Master BIAS files for all 4 CCDmodes (high and low gain; 1-portand 4-port readout).

• Master flats for IMG mode: masterSCREEN_FLAT_IMG, master SKY_FLAT, master NIGHT_FLAT. Theycome in two CCD modes (high andlow gain, 4-port readout). MasterSKY_FLATs are used for sciencereduction (see below) and henceare measured in all filters availablefor imaging, the most commonlyused are the Bessell UBVRI filters.They are frequently measured indusk and dawn.

• Tables with photometric zeropoints(ALIGNED_PHOTOMETRY_TABLE)from standard star frames for IMGmode: these are exposed in the fivestandard Bessell filters (UBVRI),usually in the high-gain, 4-portCCD mode. Sets of such five stan-

1Since this process bears some resemblance totrout held in purification plant basins to indi-cate water quality, we have dubbed ourselves the‘QC fishes’.

Master files aremedian averagesfrom input sets oftypically 3–5 rawfiles. The mastercreation recipeuses a kappa-sig-ma clipping rou-tine to reducerandom noise,suppress cos-mics and stellarsources.

Science dataare reduced us-ing these cali-bration products.S C I E N C E _IMG files are de-biassed and flat-tened. The pipe-line uses twilightSKY_FLATs tak-en in dusk ordawn. These flatsremove all small-scale CCD struc-ture (‘fixed-pat-tern noise’), thefour-port patternand all largescales except forthe largest onesof order 1000 pix-els. This is due toillumination gra-dients differingbetween nightand twilight, andamounts to 1–2%.A NIGHT_FLATwould removeeven this gradi-ent perfectly, butis not routinelyavailable. If pos-sible, the pipe-line extracts suchflats from jitteredscience images.Success dependson the offsetchosen for jitter-ing, the nature ofthe sources andtheir density. Ifavailable, theseNIGHT_FLATsare delivered aspart of the SMdata package,but they are notused for pipe-line science re-ductions. MasterSCREEN_FLATsare available, buttheir illumination

pattern is very different from sky con-ditions. They are primarily used for mon-itoring the CCD performance.

Photometric standard stars are re-duced the same way as science data,with the added step of source identifi-

dard exposures are taken at leastonce, usually several times pernight.

• Master flats SCREEN_FLAT_LSSand dispersion solutions WAVE_DISPERSION_LSS for LSS mode.

7

Figure 1: FORS1 trend plot of the BIAS QC parameters median value (diagram 1) and read noise (diagram 2) for thefour CCD modes low and high gain / 1-port and 4-port readout. The period covered is 1999-10-01 to 2000-01-01.These plots are used to assess the stability of the CCD system and identify outliers.

Figure 2: Trend plot for QC parameters of SCR_FLAT_IMG calibration files. Diagram 1 shows mean values, diagram2 photon noise (in raw frames) and fixed-pattern noise, diagram 3 large-scale structure (both in master files). Data arefor the five Bessell UBVRI filters. The last plot clearly shows the CCD contamination slowly increasing with time, main-ly affecting the U filter.

These come in 6 grisms whichcould be combined with any of 9longslits.

Hence for a typical IMG mode nightabout 30–40 master calibration fileshave to be created by the pipeline.

cation and extraction. At the moment,they produce photometric solutions(zero points) for the night. Currently thisinformation is used to assess the quali-ty of the night and trace telescope effi-ciency. SM data packages receive thezeropoint tables and the reduced stan-dard star files.

LSS data are de-biassed and flat-tened using master SCREEN_FLAT_LSS files which contain high spatial fre-quencies only. The data are rebinned towavelength space but not extracted.Hence fixed-pattern noise, slit noiseand slit function are removed, as is slitcurvature.

Planned next steps are the photo-metric calibration of the IMG files, andremoving of the instrumental efficiencycurve for LSS data. Finally the MOSpipeline will become operational inPeriod 65.

A total of about 3000 master calibra-tion files, and about the same numberof reduced science files, has been cre-ated in Period 63. 93% of these files areIMG files.

Distribution . The proper distributionof files in the Service Mode packages isnot trivial, if calibration files are con-cerned. As the simplest example, look

at science files taken in IMG mode andassume that the OB was taken with justone filter and in one CCD mode. Thiswould make a minimum set of one mas-ter BIAS and one master SKY_FLATtaken with these parameters (plus, ofcourse, the corresponding raw and re-duced science files). Since it is notknown, however, whether the pro-gramme requires photometry data, wealways add all available STD_IMG files(raw, reduced, photometry table). Tofurther enable the PI to reprocess all re-duction steps, this requires all applica-ble SKY_FLATs and BIASses as well.Hence we actually blow up the amountof data delivered by a factor which cango up to 5 or more, but only this ap-proach guarantees completeness. Toslim down the CD-ROM package a littlebit, we usually do not include those rawcalibration files that successfully pro-duced master files. These raw calibra-tion files can be retrieved by the userfrom the ESO Archive (see article byLeibundgut et al. in this issue).

For the LSS mode, things becomemore complicated since spectrophoto-metric standard stars are taken in MOSmode. Hence LSS programmes receivefull sets of LSS and MOS calibration

files, and the blow-up factor is evenlarger than for IMG files.

Quality Control . Post-pipeline oper-ations involve quality checks of the rawand the produced data. As a simple buttime-consuming check, scanning thenight logs is fundamental. This ispresently done in the old-fashionedway, i.e. reading and, if needed, editingby hand. In the near future there will betools to have night-log information ac-cessible for automatic processing, dis-tribution and storage.

On raw and produced calibrationdata, several checks are done. Fromthe BIAS frames, median values for thebias level (both across the whole CCDand per port), for the value of large-scale structure, and for the read noisefor raw and master files are determined(Fig. 1). On the SCREEN_FLATs andthe SKY_FLATs, the mean values(across the whole CCD and per port),the random photon noise, the fixed-pat-tern noise, and the large-scale structureare measured (Fig. 2). SCREEN_FLATs are also used to measure actualgain values. It is checked how randomthe ‘random’ noise is.

Photometric zeropoints are deter-mined. The quality of the LSS disper-

8

Figure 3: Snapshot of the Gasgano GUI in use with ISAAC data. The tool allows interactive selection of classified frames and their input tothe corresponding pipeline recipes.

9

Supported imaging Period Descriptiontemplates

Jitter 63 It reduces images taken in jitter and jitter + offset modes. The jitter data reduction process is divided into flat-fielding/dark subtraction/bad pixel correction, sky estimation

Jitter+Offset 64and subtraction, frame offset detection, frame re-centring, frame stacking to a single frame, and optional post-processing tasks.

Darks (including 63 Creation of master dark frame. The process sorts out frames with an identical DIT and spectroscopy) produces the averaged frames. It computes also the read-out noise.

Zero-points 64 Calculation of zero points. It computes the number of counts, and relates that measure-ment to a standard star database. The identified infrared star database at present contains about 800 star positions with magnitudes in bands J, H, K, Ks, L and M.

Twilight flats 63 Creation of master flat frames. It takes as input a list of files taken at twilight and produces the flat-field of the detector by observing this rapidly increasing ordecreasing signal. Since it computes a characteristic curve per pixel, it also creates a bad pixels map.

Illumination frame 64 Creation of master illumination frames. It subtracts dark, divides by flat-field and correctsbad pixels if the adequate calibration files are available. The final product is a 2dpolynomial surface normalised to a value of 1.

Supported spectroscopy Period Descriptiontemplates

NodOnSlit 64 It reduces images taken in jitter spectroscopic mode. The process is divided into the (partly) classification of the input files, correction of the distortion, shifting the frames and aver-

aging, wavelength calibration, creation of a combined image, detection and extraction of a spectrum. The wavelength-calibrated spectrum is provided for all calibration standard star observed in this mode. It is not provided for science frames.

Spectroscopic flatfield 64 Creation of master spectroscopic flat frames. This algorithm is applied to each pair of frames (lamp on and off). The difference ‘on’-’off’ is computed and the result frame is divided by its mean.

Arcs 65 It detects vertical or horizontal arcs in a spectral image, models the corresponding deformation (in the x or y direction only) and corrects the found deformation along the x or y direction. Finally, it computes a wavelength calibration using a lampspectrum catalogue.

Star trace 65 It performs star-trace analysis. It takes as input an image and produces two tables of output: a line position table (containing the fitted coordinates of the curved lines) and a polynomial coefficient table (describing the found deformation).

Slit position 65 It finds the exact position of a slit.

Response function 65 Determination of the spectroscopic response function, by means of the extraction and wavelength calibration of a standard star spectrum.

Table 1: List of templates supported by the ISAAC pipeline recipe set. The recipe's products are made available to the user community dur-ing the indicated observing period. The description briefly explains what the recipe does. For more detailed explanations, please refer tothe ISAAC web page and references therein (http://www.eso.org/instruments/isaac/).

sion solutions and the effective resolu-tion are measured.

All these parameters are stored in ta-bles and their trends monitored. Thereare also checks whether random noiseand fixed noise scale with signal as ex-pected.

In reduced science IMG files, thequality of the flattening process is con-trolled.

Feedback . Since part of the resultsof the QC process is a direct healthcheck for the CCD and the instru-ment, a natural task for QC Garching isproviding feedback to the CCD groupand to Paranal Science Operations.This is mostly channelled through theInstrument Operation Teams which

combine expertise about the instru-ment.

Generally, it is important to store andprovide QC results in a centralised wayopen to anyone interested. Options areputting results onto the web, have qual-ity-control parameters stored in a QCdatabase, and ingest QC informationinto the Archive. As a first step, checkthe newly created QC page to be foundunder http://www.eso.org/observing/dfo/quality/index_fors1.htm.

Software for operations . ForFORS1 and ISAAC, software for thelower-level functions existed when op-erations started, provided by ESO orby the instrument consortia: the datareduction pipeline and tools for or-

ganising the data and processing them.All higher-level tasks, such as distrib-

uting raw and product files to the finalSM data packages, pre-selecting datafor processing, assessment of dataquality, storage of QC parameters, etc.started during Period 63 without soft-ware support. Tools had to be devel-oped during operations. Such ‘hot de-velopment’ offers the advantage of be-ing extremely efficient since any newscript could be tested and improved un-der real life conditions. Evolutionary cy-cles were short. However, the price topay was a very tough schedule sincecertain elementary tasks had to be pro-vided, no matter whether tools existedor not.

It soon became clear that there isonly one option for keeping one’s headabove water: create (UNIX shell) scriptsand (MIDAS) procedures for automaticprocessing. The strategy to survive is:clearly identify the jobs which can beroutinely done and those which can’t.Then leave the routine stuff to the ma-chine preferably for overnight process-ing, and do the non-routine work duringdaytime. This primarily involves deci-sion making, i.e. quality assessment ofmaster calibration and reduced files,commissioning of pipeline recipes, andkeeping control of the whole process.

The backbone of the FORS1 QC jobis formed by about 30 shell scriptswhich translate the basic steps of data-flow operations into well-defined func-tionalities. This package is called‘SMORS’ (Service Mode OptimisedReduction Scheme). It produces resultswhich are repeatable and predictable,and its operation is safe. With this pack-age, we do the full data processing fromthe very beginning (provide listings fornewly arrived raw data) up to the end ofthe life cycle (delete all data for an SMprogramme once the CD-ROMs havebeen distributed).

SMORS being the backbone, a sec-ond package is the ‘brain’ of FORS1quality control: ‘qc_dec’ (QC decision),a number of MIDAS procedures devel-oped for post-pipeline assessment ofdata quality, measuring QC parametersand trending. These tools enable deci-sion making, e.g. accept or reject amaster calibration file, measure thefixed-pattern noise in a master_screen_flat, check the removal of stellarsources in a master_night_flat, checkthe degree to which a SCIENCE_IMGfile has been flattened, create nightlyaverages for photometric zeropoints.

As a by-product of the tools devel-oped for pipeline operations, a scriptpackage ‘Pipe’ has recently been in-stalled on Paranal to facilitate the oper-ation of the FORS1 quick-look pipeline.This tool can be used by staff as-tronomers to create their own mastercalibration files and obtain photometriczeropoints during the night, so that real-time assessment of the quality of thenight becomes possible.

4. ISAAC

Quality-control operations for ISAACresemble in broad sense those de-scribed in the previous section forFORS1. The differences between thetwo instruments, especially in terms ofoperations, lead to a different approachfrom a QC point of view.

P63 statistics . During Period 63there have been 92 nights with ISAACdata (science and calibration frames)and a total of 59 service mode nights(with science data), which have pro-duced a total of 24087 files (includingcommissioning and science verificationdata), divided into 13349 science

frames, 9985 calibration frames and753 test frames. The total number ofSM programmes was 35, 4 of whichrequired quick releases (that is, releaseof the data soon, typically 1 day, afterobservations). At the end of the Period,29 programmes had been shipped tothe users, while the remaining 6 wereput on hold by User Support Group(USG) and Science Operations inParanal (PSO) to allow for follow up ob-servations during Period 64. A total of69 CDs were prepared and cut, thebiggest programme received 12 CDs,the smallest only 1. The longest pro-gramme spanned 5 months and thedensest was observed in 12 differentnights.

CD packing . Each PI of a SM pro-gramme receives a set of CDs contain-ing data subdivided by night of obser-vation. Each “night” contains the follow-ing data:

• Science raw frames and pro-gramme specific calibration files (ifrequired by the observer).

• Calibration raw frames taken fol-lowing the calibration plan (onlythose pertaining the specific set-tings required for the science rawframes). For example, a pro-gramme that required observationsin short wavelength mode (SWI1)in Ks band with a single value of thedetector integration time (DIT) willreceive the twilight flats in Ks, thestandard star frames for that nightin all bands and the dark frames forall the DITs used (science plus cal-ibration frames).

• Master calibration data – the mas-ter files as created by the ISAACpipeline recipes set. For the aboveexample, the user will receive themaster dark frames for all DITs andthe master twilight frame in Ks.

• Reduced science frames – all sci-ence frames sets observed usingthe autojitter and autojitter + offsettemplates are sky-subtracted andcoadded.

During Period 63 the pipeline pro-duced a total of 268 master calibrationframes (SWI1 mode only, master flatsand darks) and 55 coadded images (injitter mode). The average number offrames used to produce a single masterframe was 18 for calibration and 17 forscience frames.

In addition to the data, for each nightof observation and each set of data(raw science and calibration, reducedscience and calibration) a table with thelist of files and their most relevant key-words (e.g. RA, DEC, DIT, filters, cen-tral wavelength, etc.) is included. EachCD contains the night logs for the rele-vant night of observations with all perti-nent log entries written during observa-tion by the operation staff astronomersin Paranal, and a data reduction log,which contains information on the datapackage, the reduced frames, the OBlist, etc.

The packing process for a specificprogramme starts upon receipt of a“completion” signal from USG and fin-ishes when the package is sent to thePI. In the worst case in Period 63, thedelivery time had been 40 days afterthe signal while in the best case lessthan 1 day. It has to be noted that thedelivery time decreased steadily duringthe Period, and now, for Period 64, theaverage is 1 day after receiving thecompletion signal.

When operations for ISAAC startedin April 1999, it was decided to con-centrate on SWI1 mode only and toprogressively increase the number ofproducts delivered to the users. Thischoice was driven mainly by the specif-ic need of further testing for the recipeset in short wavelength spectroscopicmodes (SWI1 and SWI2) and by themore general fact that, running opera-tions for the first time ever, was a taskfull of unknowns. Now that the wholedata-flow process is better understood,we progressively add tasks to qualitycontrol and services for the user com-munity.

ISAAC Pipeline Recipes . Each VLTinstrument has its own pipeline set ofrecipes, which support all or part of theinstrument modes. In the case ofISAAC, presently the pipeline supportsthe short wavelength modes, imagingand spectroscopy and will probably beextended in the future to include thelong wavelength modes. In Period 63,the ISAAC pipeline set of recipes wasinto SWI recipes, all developed aspart of the Eclipse software (N.Devillard, “The eclipse software”, TheMessenger No. 87 – March 1997 andhttp://www.eso.org/projects/aot/eclipse/),and SWS recipes, developed in MIDASby Y. Jung. As for Period 64, and start-ing with operations in Period 65, allSWS recipes have been also includedin the Eclipse software, thanks to thework of Y. Jung and T. Rogon. Table 1lists the entire set of templates support-ed by the pipeline.

Among the responsibilities of QC,testing of the pipeline recipes is one ofthe most important, since the work ofQC relies entirely upon this set ofrecipes for the great part of the work.The quality control scientist producesmaster calibration frames and certifiestheir quality before shipment to theusers. In the near future all frames willbe inserted in the calibration databaseand made available to the user com-munity. As for Period 64/65 the follow-ing by-products of the pipeline are cal-culated and their trending monitored:read-noise of the detector, zero pointsfor each night. Zero point values arealso made available to the users withSM programmes. Shortly, they will bepublished on the Web for the entireuser community. In addition, the good-ness of sky-subtraction and coaddition,as well as image quality is monitored inall coadded images produced for SM

10

programmmes. Further checks will beintroduced for all products produced bythe pipeline (e.g. spectroscopic jittermode images).

For additional information on thesupported templates and operatingmodes see the ISAAC home page on-line and the ISAAC manual (http://www.eso.org/instruments/isaac/ andreferences therein) and the PSO webpages for ISAAC (http://www.eso.org/paranal/sciops/ISAAC_Info.html).

Software for Operations . When op-eration started in April 1999, we hadclearly understood the general pictureof the data-flow, but we missed first-hand experience on the actual amountand the type of work, on the most effi-cient way to do it and of course on allthose unknowns, which are to be ex-pected every time a new enterprise isstarted.

To perform the first 4 tasks listed insection 2, QC could from the beginningmake use of the data-flow system,which includes, among the others, theData Organiser (DO), a software whichclassifies the raw data and creates re-duction blocks, which are in turn usedby the Reduction Block Scheduler(RBS) software to fire the proper recipeand run it for the list of raw frames pre-viously classified. Both DO and RBSwork in a completely automated way.For the particular case of ISAAC, thedifferent science operations needs,which change according to the particu-lar observing programme to be execut-ed, and the wish to keep them as flexi-ble and efficient as possible, require agreater level of “human” interactionthan what allowed by automated soft-ware. In the majority of cases, it is nec-essary to classify the files, to select theframe-set as input to a data reductionrecipe and to tailor the configuration pa-rameters of the recipe itself manually.Given the great amount of data thatreaches Quality Control and that has tobe processed and distributed, a newsoftware tool had to be foreseen. Themain requirements for it were: flexibilityand interactivity of operations, compat-ibility with the data-flow model and withthe needs of QC work, speed (in a typ-ical ISAAC night a minimum of 300 filescan be produced and in a typical QCworking session many nights of datamust be loaded, classified and reducedat the same time) and configurability.The Software Engineering Group(SEG), namely N. Kornweibel and M.Zamparelli, developed a softwarenamed Gasgano (see Fig. 3), whichprovides these and many other func-tionalities, which make it the tool rou-tinely used by QC for ISAAC. The toolhas been recently officially released toPSO, but has been tested and used byQC since its very first “unofficial” re-lease.

Since the creation of a CD packagefor a SM programme is not yet feasibleby means of Gasgano, a set of UNIX

shell scripts, similar to those created forFORS1, was developed for ISAAC.These scripts allow to quickly and(semi) automatically assemble all sci-ence and calibration raw data, their cor-responding reduced frames and the re-duction logs; in addition the scripts pro-duce a reduction and packing log, filelistings, useful statistics (number offrames divided by type – science or cal-ibration –, Programme ID and OB IDand list of rejected frames, typically fileswith incorrect or missing keywords) andcheck logs. The latter are comparedwith information retrieved by the scriptsthemselves from various database ta-bles of the ESO archive (OB repository,Observations, etc.). The difficulty of thepacking process lies almost entirely inthe non-uniform distribution in time ofthe calibration files and in the vastnessof the science data parameter space:the script must be able to “intelligently”choose a proper set of calibrationframes, observed as a rule under a dif-ferent program id and in general in dif-ferent days than those of scienceframes. Those may also vary, for differ-ent SM programmes and within a singleprogramme, in all possible modes al-lowed by the instrument.

We have already started to developan extra set of scripts/tools to check thequality of processed data, up to nowperformed manually on each frame, toprovide instrument performanceschecks and to perform trend analysis ofquality parameters. It is foreseen thatthey will be fully ready to support QCoperations within Period 65.

5. Instrument Operations T eams

From the user point of view, QualityControl activities represent the last ele-ment of the data flow life-cycle chain, inthe sense that the final delivery of thedata, which ends the cycle, relies uponit. Less evident for outside observers isthat the entire operation process reck-ons upon the work of a fairly largegroup of persons with different respon-sibilities within the chain and owes itssuccess to the interactions that occuramong them. This group forms the so-

called Instrument Operations Team(IOT) and every instrument for the VLThas a similar team assigned to it to en-sure operations.

6. Lessons Learnt

After more than one period of opera-tion, it is clear that the basic concepts ofQuality Control are routinely working.Data are processed and their qualitychecked, PIs receive their SM datapackages. These data packages addvalue to the simple traditional raw filecollections.

Some important issues could beidentified during operations. One is:keep the instruments simple, if youwant to have simple operations. Thereis a close correlation between, e.g.,the many modes offered by FORS1 andthe complexity of its data flow opera-tions.

Operationally, do as much as possi-ble in automatic mode. This is reliable,reproducible, and can be done in batchmode. Scripts are preferable over inter-active tools if you go for mass produc-tion. The evolutionary development ap-proach, though dictated by circum-stances, proved to be efficient.

Data integrity is very important. Wecannot afford to manually correct errorsintroduced upstream: either files arriv-ing in Garching are syntactically (DICBconform) and logically integral (e.g.have proper programme and OB ID), orthey are useless. DICB conformity isalso extremely important for Archive in-tegrity. Data with wrong keyword con-tents will never be properly retrieved.They are just wasting disk space.

Relevant information has to be keptcentral. Facilities like web pages, rela-tional databases and the Archive arecrucial.

All in all, the tasks of QC Garchingcombine astronomical challenges withinformation technology challenges.They close the loop for VLT data pro-duction, provide added value to thecommunity and create a sound data-base for assessment of instrument per-formance.

11

Table 2: Members and their respective roles within the team for ISAAC and FORS1.

Role ISAAC FORS1

Instrument Scientist Jean-Gabriel Cuby Gero Rupprecht

Operations Staff Chris Lidman, Gianni Marconi Hermann Boehnhardt,Astronomers Thomas Szeifert

User Support Almudena Prieto, Fernando Comerón Palle MøllerAstronomer

Pipeline Nicolas Devillard, Yves Jung, Stefan BogunDevelopment Thomas Rogon

Quality Control Paola Amico Reinhard HanuschikScientist (future: Ferdinando Patat)

rhanusch@eso.org, pamico@eso.org

12

The science archive at ESO doesn’tjust collect the data from the tele-scopes, it has a number of other func-tions. These include the distribution ofthe service mode data obtained atthe VLT, the immediate availability ofall calibration data from all telescopesfrom which data are archived (in addi-tion to the VLT data we are currentlyarchiving the data from the 3.6-m, theNTT, and the 2.2-m), the public accessto the science data after the propri-etary period has expired, and the col-lection of all master calibrations. Otherservices offered through the archiveare the seeing databases from LaSilla and Paranal, several astronom-ical catalogues (e.g. USNO, GSC I,Tycho-2), and on-line access to thedigital sky survey, among other things.The archive is a joint operation ofthe ECF and ESO and contains all theHST data as well. The archive staffis shared between these two groups.The ESO archive is accessed athttp://archive.eso.org or by clicking on‘Observing Facilities and Operations’on the ESO home page (http://www.eso.org) and then choosing the‘Science Archive Facility’. This articlegives an overview over the current sta-tus and some of the new features in theESO/HST archive.

1. Data Available in the Archive

Currently the archive collects datafrom FORS1 and ISAAC at UT1,EFOSC2 and CES at the 3.6-m, EMMI,SUSI2, and SOFI at the NTT and theWide-Field Imager (WFI) at theMPG/ESO 2.2-m. A graphical overviewover the number of data sets in thearchive at the beginning of February

2000 can be seen in Figure 1 for theESO and Figure 2 for the HST archive,respectively. The commissioning andscience verification data of UVES andFORS2 have already entered thearchive. Relevant information on eachobservation is entered into a relationaldatabase, which can be queriedthrough a Web page. The database canbe searched for specific celestial ob-jects by their regular names (resolvedto positions by querying either SIMBADor NED), position on the sky, by ESO’sprogramme identification, data type, tel-escope or instrument, or any combina-tion of these parameters. In addition,the FITS headers of the VLT data areincluded online and for FORS1 imagingdata previews can be investigated.Archive data can be requested throughthe results page of any query. Pleasenote that you have to be a registereduser of the ESO/ECF archive to retrievedata. Registration can be done directlyfrom the archive Web page.

EMMI/NTT data have systematicallybeen archived since 1991 (see TheMessenger No. 93, page 20). The othertelescopes and instruments have beenadded continuously. The one-year pro-prietary period for the first VLT data willexpire at the beginning of April and allfurther science data will successivelybecome public. The raw commissioningand science verification data of FORS1,ISAAC, and UVES are already publiclyavailable through the archive. The sci-ence verification data can also be re-trieved in processed form at http://www.eso.org/science/ut1sv and http://www.eso.org/paranal/sv. The raw ESOImaging Survey data (da Costa et al.1999, The Messenger No. 98, page 36)can also be retrieved from the archive.

Data from the wide-field imager are en-tering the public domain as well.

To ease the access to the variouspublic data sets we have created a spe-cial page(http://archive.eso.org/archive/public_datasets.html). This page con-tains links to the above data sets andsome new test data that are publiclyavailable. Recently, data of polarisationmeasurements of the bright Type IISupernova SN 1999em (IAU Circulars7296, 7296, 7305, and especially 7355)have become available. These datahave been taken during the technicalnight of 2 November 1999 shortly afterthe supernova was discovered. Theraw data with some information fromthe night logs can be found there. Wewill make data sets public in the futurethrough this page, so please keep youreyes open.

2. Data Delivery

Once you made your selection fromthe archive, you can request data setsfrom the results page. After submissionof the request you will be notified byemail at the address you entered whenyou registered. Please make sure thatthis address is up to date. You have achoice of which medium you prefer forthe data delivery. In most cases forsmall (< 200 Mb) requests the data areprovided through ftp. Larger requestsreceive the data by regular mail onCD-ROMs, like in the case of servicemode programmes. For high data vol-umes, e.g. requests for WFI data above10 Gb, the choice is limited to DLTtapes. Other media options are DATtapes (DDS-1 and DDS-3 formats) andExabyte tapes. We are now also start-ing to ship data on DVD-R disks.

Access to VL T Data in the ESO ArchiveB. LEIBUNDGUT, B. PIRENNE, M. ALBRECHT, A. WICENEC, and K. GORSKI

Figure 1. Figure 2.

Please contact catalog@eso.org forspecial cases.

Once the data set has been createdyou are again notified by e-mail aboutthe dispatch of the data from ESO.

3. Future Developments

We are currently redesigning thearchive web pages and their layout. Afirst example is the archive entry page(Fig. 3). This should result in an im-proved presentation of the availabledata and their usefulness for specificresearch projects. We also will add newparameters to the database to makesure that we cover most of the requestsand make the selection criteria as var-ied as possible to make the searchesfor specific data sets easier.

A typical example is the search forcalibration data that correspond to agiven scientific data set. Right now,these calibrations have to be selectedmanually which is a tedious process.We are planning to implement a moreautomatic procedure in the future,which would simplify the search signifi-cantly. One of the main ongoing proj-ects is the inclusion of the master cali-bration data produced by the qualitycontrol group into the archive. You thenshould be able to find the correspon-ding calibration data in a processedform in the archive.

Please let us know if you find defi-ciencies in the archive so that we canaddress them. You can contact any ofus at the e-mail addresses given at theend of this article.

bleibund@eso.org; bpirenne@eso.org – malbrech@eso.org; awicenec@eso.org –kgorski@eso.org

13

Chile Astroclimate, a Biannual UpdateM. SARAZIN, ESO

Not long ago (The Messenger 97,September 99), climate change wasidentified as the main responsible forthe degradation of observing conditions(seeing) at Paranal. It was pointed outin particular that the weakening of thetraditional westerly wind pattern wasmore frequently allowing turbulent airfrom inland to blow over the coastalcordillera.

Figure 1: Seeing Statistics at Paranal sinceUT1 first light: monthly average (red), medi-an (black) and 5th percentile (blue). Thedashed lines give the respective long-term(1989–1995) site characteristics. Seeing isreconstructed from DIMM measurementstaken at 6 m above ground, at 0.5 micronand at zenith. Because of the finite outerscale of the atmospheric turbulence, actuallarge-telescope image quality can be betterthan predicted by DIMM (see e.g.: The see-ing at the William Herschel Telescope, R.W.Wilson et al., MNRAS 309, 379–387, 1999).

Six months later, and in spite of muchwishful thinking, the site quality has onlymarginally improved and remains waybelow the standards established duringthe extensive site survey (dashed lines,Fig. 1). This means for the observatorythat Period 64 should not be better thanPeriod 63 which provided sub-half arc-second seeing only 13% of the time (R.

Gilmozzi, The Messenger 98, Decem-ber 99, to be compared 21% in the pe-riod 1989–1995). During that same pe-riod, La Silla, which is not undergoingany visible climate change but is ratheron a favourable phase of its own cycles,had been producing 8% of such good-quality observing time and promiseseven more in Period 64.

It was reported (The Messenger 90,December 1997) that cloudiness atParanal was obviously increasing withwarmer sea water, i.e., El Niño events.The dependency of Paranal seeing toEl Niño cycles had been indeed similar-ly tested over a decade in the past(1988–1997) but without unveiling anycorrelation (yellow squares in Fig. 2). Itwas thus concluded that the basic Pa-ranal observing conditions were weath-er independent. The seeing increase ofthe past 20 months (green squares inFig. 2 corresponding to the periodshown in Fig. 1) is mainly due to a par-ticular North-East wind pattern whichlasts part of the night, a few times permonth. As shown in Figure 2, all thesepoor months belong to the current LaNiña and the seeing trend even showssome correlation with the standardisedSouthern Oscillation Index (SOI) whichis commonly used to define the state ofthe Pacific Ocean surface temperature.

The El Niño and La Niña cycles arehardly predictable and many past at-tempts failed. Some success was ap-parently obtained by a model based onsolar-activity cycles which correctly pre-dicted the 1997–1998 El Niño event(http://www.microtech.com.au/daly/sun-enso/sun-enso.htm). If one can be-lieve such models, the next El Niñoevent should arrive in 2002, perhapsbringing to an end the current phase ofpoorer than average astroclimate onParanal.

Moreover, recent analyses of seasurface elevation measured by theTopex-Poseidon satellite (NASA/JPLNews release, Jan. 20, 2000) lead re-searchers to suspect the Niño-Niña os-cillations to sit on, and therefore partial-ly hide, a much wider (20–30 years pe-riod) so-called Pacific decadal oscilla-tion. If this phenomenon was confirmedand quantified, it would provide newperspectives to astroclimatological sur-veys; let us thus wait and see.

14

Figure 2: Correlation of the standardised monthly Southern Oscillation Index (SOI) withmonthly average seeing at Paranal during 1988–1997 (yellow) and since April 1998 (green).The SOI represents the sea level pressure anomaly between Darwin and Tahiti(http://www.cpc.noaa.gov/data/indices/). A negative index corresponds to warmer waters (ElNiño), a positive index to cooler ones (La Niña).

msarazin@eso.org

ESO Demonstration Project with the NRAO 12-m Antenna R. HEALD (NRAO) and R. KARBAN (ESO)

During the months of Septemberthrough November 1999, an ALMA jointdemonstration project between theEuropean Southern Observatory (ESO)and the National Radio AstronomyObservatory (NRAO) was carried out inSocorro/New Mexico. During this peri-od, Robert Karban (ESO) and RonHeald (NRAO) worked together on theESO Demonstration Project. The proj-ect integrated ESO software and exist-ing NRAO software (a prototype for thefuture ALMA control software) to controlthe motion of the Kitt Peak 12-m anten-na. ESO software from the VLT provid-ed the operator interface and coordi-

nate transformation software, while PatWallace’s TPOINT provided the point-ing-model software.

On the 26 to 28 November, the proj-ect had its highlight – the final test withthe Kitt Peak 12-m antenna at theNRAO Observatory in Tucson/Arizona.Since the test period lasted only 72hours, it was essential to prepare, planand test the software thoroughly andsystematically. To accomplish this,practices of ESO Software Engineeringwere applied. ESO configuration man-agement, systematic regression test-ing, build procedures, development en-vironment, test preparation and docu-

mentation procedures were used.Using these methods enabled us tomanage efforts among the various per-sons in the project locally, as well as toprovide remote support from ESO. Theproject was successfully completed.For the test results and more details onthe project, see http://www.alma.nrao.edu/development/computing/news/index.html

We would like to thank Bob Freundand the other members of the Tucsonoperations staff who provided us excel-lent system support during these threedays at the 12-m.

Since the last article in this series,the NTT has been through a completerejuvenation process: during a period of5 technical nights in December 1999,the main mirror has been re-aluminisedand the telescope, the walls of the tele-scope room, and the main parking lothave been painted. The NTT looksbrand new, while it has just passed its10th anniversary.

The result of the aluminisation is ex-cellent: the reflectivity is back to 91%and the micro-roughness to 10Å. Thesevalues are similar to those we obtainedafter the previous aluminisation, whichtook place 3.5 years ago. It should how-ever be noted that, thanks to the week-ly CO2 cleaning and the water cleaningevery 3–6 months, just before the alu-minisation, the reflectivity was still at87%, and the micro-roughness at 60Å.The amount of diffused light was meas-ured before and after the aluminisation(using the radial profile of bright stars):a significant improvement is noted inthe U, B, and V filters, of the order of 40,30, and 20% respectively, at 40″ fromthe star. No significant improvementwas measured in R, I, J, H and K.

The outside parking lot and platformwere repainted in white, in order to min-imise the amount of heat that is accu-mulated during the day and releasedduring the night. Be sure not to forgetyour sunglasses when you exit the tel-escope during daytime: the platform’s

albedo is now very close to 1! The innerwalls of the telescope room have beenpainted with a high-diffusion paint to cutdown the reflection of the Moon duringthe night. Indeed, the former grey wallswere quite glossy, and we often hadsome nasty reflections when observingwith the Moon up. The telescope itselfhas also been painted (we did notchange the colours) to protect the struc-ture, which was starting to oxidise inplaces.

During Period 64, the NTT scheduleincludes 14 nights of service observing.Unfortunately, the weather is not veryco-operative, but the programmes arebeing executed. The User SupportGroup maintains a web page where theprogress of these programmes can bemonitored, see URLhttp://www.hq.eso.org/observing/dfo/

Since the last article, the data-reduc-tion pipeline has significantly grownthanks to the intensive work of B.Joguet (NTT). All the imaging modesof the three instruments are now sup-ported, and data from some of thespectroscopic modes are also proc-essed: a couple of minutes after youtake a RILD spectrum, you get a wave-length-calibrated, flat-fielded version ofyour spectrum. The remaining modes(i.e. echelle spectroscopy, IR polarime-try imaging) should be implemented inthe coming months. Consequently, weare now keeping a library of standard

calibrations: you can request from yoursupport astronomer the flat-field, bias-es, darks, etc., from the previousweeks.

SOFI, the IR spectro-imager, is stillmisbehaving. Over the past months, wehad an alert with the closed-cycle cool-ing system (now under control), with theDetector Control System (under investi-gation, but now behaving properly) andwith the Grism wheel. The latter is stillnot working, and is kept in the “open”position, in order to permit imaging ob-servations. An intervention on thatwheel is foreseen in April, but until thatdate, no spectroscopic (or polarimetric)observations will be possible. The PIsof the affected programmes have beencontacted. After the April intervention,the problem should be solved in a per-manent way.

To conclude this message, I am hap-py to announce that the change from1999 to 2000 did not cause any prob-lem at the NTT: we stopped the wholesystem on December 31 at 18:00 (thecontrol room with all its 17 switched-offmonitors is a depressing sight), andrestarted everything at 21:30 (January1, 2000, 00:30 UT). By 23:30, i.e. afterthe time to start everything, we were onthe sky.

Finally, the IR staff position has final-ly been filled: Leonardo Vanzi, formerNTT fellow, is now the SOFI instrumentscientist.

15

NEWS from the NTTO. Hainaut and the NTT Team

New Pictures from Paranal Observatory

16

history of a system. Some of the physi-cal parameters that affect a CMD arestrongly correlated, such as metallicityand age, since successive generationsof stars may be progressively enrichedin the heavier elements. Thus, detailednumerical simulations of CMD morphol-ogy are necessary to disentangle thecomplex effects of different stellar pop-ulations overlying each other and makean effective quantitative analysis ofpossible SFHs (e.g., Tosi et al. 1991;Tolstoy and Saha 1996; Dohm-Palmeret al. 1997). For every galaxy for whichan accurate CMD has been derived,down to the Horizontal Branch (HB) lu-minosity (MR ~ 0. ± 0.5) or fainter wehave learnt something new and funda-mentally important about the SFH thatwas not discernable from images con-taining the red giant branch (RGB)alone (e.g., Smecker-Hane et al. 1994;Tolstoy et al. 1998; Cole et al. 1999).

Accurate CMD analysis benefitsenormously from the high spatial reso-lution and excellent image quality, ascrowded-field stellar photometry is ex-

tremely sensitive to seeing, which af-fects both the degree of crowding andthe speed with which an image be-comes sky noise limited. Previous pro-grammes on ESO telescopes havebeen carried out with the 2.2-m tele-scope (e.g., Tosi et al. 1989) and morerecently with the NTT (e.g., Minniti andZijlstra 1996). Here we show that in ide-al conditions, and with a large, high-performance telescope and closed-loopactive optics spectacular improvementscan be obtained on previous results.

Because of the significant gains inimage quality and collecting area nowavailable with the VLT on Paranal, it isworthwhile and fundamentally impor-tant to survey resolved stellar popula-tions down to the HB of all nearbygalaxies in our Local Group and beyond(see Fig. 1). This will provide a uniformpicture of the evolutionary properties ofgalaxies with a wide variety of mass,metallicity, gas content, etc. and thusguide our understanding of galaxy evo-lution in conditions of extremely lowmetallicity, presumably similar to those

Abstract

In August 1999 we used FORS1 onUT1 in excellent seeing conditions overthree nights to image several nearbygalaxies through the B and R broad-band filters. The galaxies observed,Cetus, Aquarius (DDO 210) and Phoe-nix, were selected because they arerelatively close-by, open-structureddwarf irregular or spheroidal systems.Owing to the excellent seeing condi-tions we were able to obtain very deepexposures covering the densest centralregions of these galaxies, without ourimages becoming prohibitively crowd-ed. From these images we have madevery accurate Colour-Magnitude Dia-grams of the resolved stellar populationdown below the magnitude of theHorizontal Branch region. In this waywe have made the first detection of RedClump and/or Horizontal Branch popu-lations in these galaxies, which revealthe presence of intermediate and oldstellar populations. In the case ofPhoenix, we detect a distinct and popu-lous blue Horizontal Branch, which indi-cates the presence of quite a number ofstars >10 Gyr old. These results furtherstrengthen evidence that most, if not all,galaxies, no matter how small or metalpoor, contain some old stars. Anotherstriking feature of our results is themarked difference between the Colour-Magnitude diagrams of each galaxy,despite the apparent similarity of theirglobal morphologies, luminosities andmetallicities. For the purposes of accu-rately interpreting our results we havealso made observations in the same fil-ters of a Galactic globular cluster,Ruprecht 106, which has a metallicitysimilar to the dwarf galaxies.

1. Introduction

Deep Colour-Magnitude Diagrams(CMDs) of resolved stellar populationsprovide powerful tools to follow galaxyevolution directly in terms of physicalparameters such as age (star formationhistory, SFH), chemical compositionand enrichment history, initial massfunction, environment, and dynamical

REPORTS FROM OBSERVERS

Imaging W ith UT1/FORS1: The Fossil Record ofStar -Formation in Nearby Dwarf GalaxiesE. TOLSTOY1, J. GALLAGHER2, L. GREGGIO3, M. TOSI3, G. DE MARCHI1

M. ROMANIELLO1, D. MINNITI4, A. ZIJLSTRA5

1ESO; 2University of Wisconsin, Madison, WI, USA; 3Bologna Observatory, Italy; 4Universidad Católica, Santiago, Chile; 5UMIST, Manchester, United Kingdom

Object Distance MV [Fe/H] type ref(kpc) (dex)

Aquarius 800 –10.0 –1.9 dI/dSph Mateo 1998Phoenix 445 –10.1 –1.9 dI/dSph Mateo 1998Cetus 800 –10.1 –1.7 dSph Whiting et al. 1999

Ruprecht 106 20 –6.45 –1.7 globular cluster Da Costa et al. 1992

Galaxy date filter exp. time <seeing>(secs) (arcsec)

Aquarius 17Aug99 R 3000 0.45B 3600 0.45

Phoenix 19Aug99 R 1600 0.80B 1800 0.80

Cetus 17Aug99 R 3000 0.45B 3600 0.55

Ruprecht 106 19Aug99 R 30 0.60B 80 0.75

Table 1: The Sample

Table 2: The Observations

17

expected in the early Universe. A com-plete survey of the SFH of the nearbyUniverse should also be broadly con-sistent with those determined from highredshift surveys (e.g., Steidel et al.1999).

2. The Sample

Despite the advances in ground-based image quality at the VLT, we stillhave to carefully select galaxies thatwill be relatively uncrowded, i.e. sys-tems with a stellar density of 0.01–0.5stars/arcsec2, down to the magnitude ofthe HB region. Dwarf Irregular (dI) andSpheroidal (dSph) galaxies at 400–800kpc distance fit perfectly into this cate-gory (see Table 1). These types ofgalaxies are also the most numerous inthe Local Group (see Fig. 1) and be-yond, and there is considerable evi-dence for widely varying evolutionaryhistories, with periods of active star for-mation interspersed with quiescent pe-riods (e.g., Smecker-Hane et al. 1994;Gallart et al. 1999). They have thusbeen suggested as good candidates forpresent-day counter-parts to the “faintblue galaxies” seen in redshift surveys

(e.g., Ellis 1997). Two of the objects welooked at are known as “transition” ob-jects, which means they are intermedi-ate in class between dSph (no currentstar formation, or HI gas) and dIs (cur-rent star formation and HI gas), and areparticularly interesting because theymay help us to understand why galax-ies may turn on and off their star-for-mation process galaxy-wide and thuswhy dwarf galaxies can exhibit suchwidely differing SFHs.

3. The Observations

Our observing run, from 17–19August 1999, had varying seeing con-ditions (0.3″–0.9″ on the seeing moni-tor), but there were several periods last-ing 1–2 hours with stable, excellentseeing. It was these periods which weused to image the resolved stellar pop-ulations of nearby galaxies. At othertimes we observed in narrow-band fil-ters, or went to a separate programmeof spectroscopy of individual stars innearby galaxies. We present the datafor three of the galaxies we imaged dur-ing our run (see Table 2). The sensitivi-ty characteristics of the FORS1 CCD

(Tektronix) led us to observe in B and Rfilters. The B filter is very useful forcharacterising HB morphology, espe-cially the blue HB.

We typically split our observationsinto short (500–600 sec) ditheredgroups of images to help minimise flat-fielding problems and removal of cos-mic rays and bad pixels. The readoutcharacteristics of the CCD with FIERA(4-port readout in 27 secs, with a readnoise of ~ 6 e–) make this an efficientway to observe. To make an accuratephotometric solution, we made obser-vations at a different airmass of a stan-dard field (PG1657) containing severalstars over a large range in colour (B –R = –0.21–1.64) on our first night. Wethen observed two fields (PG1657 andPG2331) on the second night and oneon the third at airmass close to those atwhich our observations were made toconfirm that the photometric solutionobtained on night 1 was stable allthrough our run, and to confirm that allthree nights were clear and photomet-ric. The photometric solution is:

R – R’ = 27.38 – 0.018 ∗ X – 0.025 ∗(B’ – R’)

B – B’ = 27.21 – 0.213 ∗ X – 0.039 ∗(B’ – R’)

where R’ and B’ are the observed mag-nitudes (in e–/s), R and B are the truemagnitudes, and X is the airmass. InFigure 2 we show the central 4 arcminof the combined 3600 sec of B filter im-aging of Cetus. The average FWHM ofthe more than 11,000 stars “photome-tered” over the whole 6.8 arcminFORS1 image is 0.45″. All the extend-ed objects visible in Figure 2 are distantgalaxies behind Cetus.

4. Results

We carried out PSF fitting photome-try using a modified version ofDoPHOT, following the precepts laidout by Saha at al. (1996), over each ofour reduced and combined images, andmatched the stars detected with suffi-cient signal-to-noise in both filters. Theresulting calibrated but not reddeningcorrected CMDs are shown in Figure 3.These are the most detailed CMDs evermade from the ground of such distantsystems.

In each CMD in Figure 3 we haveclearly detected the Red Clump (RC) /HB region (MR ~ 0 ± 0.5). This regioncontains an evolutionary sequence ofHelium Burning low-mass stars 1 Gyrand older. The relative number of starsin each phase determines the age mixof the stellar populations older than1 Gyr. There is a clear diversity of mor-phologies in Figure 3, where eachgalaxy (and the globular cluster) are alldistinctly different suggesting quite dif-ferent past SFHs.

Figure 1: The spatial distribution of the Local Group plus neighbouring galaxies in X-Z galac-tic coordinates. Our Galaxy is at the origin of this plot, and our dwarf spheroidal neighboursare all marked by red dots. M31 and its colony of neighbouring dwarf ellipticals and spher-oidals are marked in green. The more free-floating dwarf irregular/spheroidal galaxy compo-nents of the Local Group are marked in light blue dots (including Aquarius, Phoenix andCetus) and also labelled. In black are the more distant galaxies on the fringes of the LocalGroup. This figure was kindly provided by Mike Irwin, and comes from Whiting et al. (1999).

18

The Aquarius dwarf which was alsostudied in the original ESO-MPI/2.2-mstudy (Greggio et al. 1993),has recent-ly been shown to be at a distance con-sistent with Local Group membership(Lee et al. 1999). Its CMD (Fig. 3a) hasa narrow RGB (B – R >1.2), and whatlooks like a relatively young RC (MR ~–0.5, B – R ~ 1.1), and very little, if any,older HB population. It also has evi-dence of a fairly young stellar popula-tion from He burning blue loop stars(sequence going left to right in the CMDbetween MR ~ –4 and –0.5 and incolour B – R ~ 0 and 1 and intersectingwith the RC) and a main sequence (ver-tically at MR ~ 0). Aquarius thus ap-pears to resemble Leo A, a galaxywhich is completely dominated by starsyounger than 2 Gyr (Tolstoy et al.1998).

The Phoenix dwarf CMD (Figure 3b)has an extraordinarily complex HB andpossibly an overlying, young RC. Thedistinct and well populated blue HB (atMR ~ 0 and B – R ~ 0 ± 0.2 to B – R ~0.4) is an unambiguous indication thatthis galaxy contains quite a number ofstars that are older than 10 Gyr. Thereare also a sprinkling of young main-se-quence and blue loop stars, as definedfor Aquarius. This looks like a galaxythat has been forming stars, perhaps onand off, for a Hubble time.

The Cetus dwarf is a newly discov-ered member of the Local Group(Whiting et al. 1999). From this CMD(Fig. 3c) it looks like it may contain ablue HB population, although muchless prominent than the one in Phoe-nix. It has a very densely populatedred HB/RC (lying between B – R ~ 0.5and 1.2), with a lot of structure thatmight resemble that predicted in the re-cent models of Girardi (1999). Thisgalaxy is probably predominantly of in-termediate age, and there is no evi-dence for star formation in the last 1.5Gyr or so.

Ruprecht 106, a Galactic globularcluster, known to have a similar metal-licity as these galaxies (Da Costa et al.1992), was also observed as a basictest of our calibration and modellingprocedures. From the CMD (Fig. 3d) wemade an RGB and an HB fiducial line,to facilitate the comparison between itand the galaxy CMDs. These lines areoverplotted in red on each of the galaxyCMDs in Figure 3. The normalisation issomewhat arbitrary because carefulmodelling has not yet been made to as-certain the optimum reddening and dis-tances with these new accurate CMDs.Ruprecht 106 is a single old population(12–13 Gyr old), whereas all the galax-ies are composite with a spread of ageand/or metallicity broadening theirHB/RC and RGB. However, Ruprecht106 serves as a good starting point tocharacterise the properties of eachgalaxy, and we will also use it to helpdetermine how accurate the theoreticalisochrones are.

5. Interpreting Colour -MagnitudeDiagram Morphology

These data will be fully interpretedusing quantitative synthetic modellingtechniques – making use of theoreticalstellar evolution tracks (e.g., Girardi etal. 2000) and, where possible, globu-lar/open cluster observations.

The Main-Sequence and Blue-Loopstars are fairly straightforward indica-tors of what the star formation rateshave been in these galaxies for the last0.5–1 Gyr. The RGB is an indicator ofstar formation which has taken placemore than ~ 1 Gyr ago. Unfortunately itis not straight forward to disentanglethe details of the precise SFH morethan 1 Gyr ago because of difficultieslike the age-metallicity degeneracy.That the galaxy RGBs are spread incolour in comparison to Ruprecht 106RGB is clear, but the effects whichcause this can be either or both of ageand metallicity variations.

We can obtain additional informationon the older populations from theRC/HB populations. The RC evolvesstrongly with age through the 1–6 Gyrage range (Caputo et al. 1995, Tolstoy1998), and if we are lucky enough tohave a population in this age range,quite a lot can be said about the SFH of

this galaxy. After about 6–8 Gyr, westart to see a red HB extension to theRC. This feature in Cetus and Phoenixprobably means that in addition to RCpopulations a few Gyr old, there arealso populations older than ~ 6 Gyr. Inthe case of Phoenix the presence of ablue HB population is a definitive indi-cator of an ancient stellar populationolder than 10 Gyr.

However, although the HB region ofa CMD provides the brightest unam-biguous indicators of old and intermedi-ate stellar populations, it is difficult touse them to uniquely quantify ancientstar formation rates. Basically the prob-lem is that the colour spread of ob-served HBs cannot be reproduced bytheoretical isochrones without invokingvariable amounts of mass loss from theRGB progenitors. This is often calledthe “Second Parameter Effect” (e.g.,Fusi Pecci & Bellazzini 1997) which re-sults in some metal poor clusters hav-ing redder HBs than models of theirmetallicity would predict. Thus it is pos-sible for globular clusters of identicalage and metallicity to have very differ-ent HBs. The implications for a com-posite population in a galaxy of un-known age and metallicity variationsare clearly dire. However, careful mod-elling of the number of stars on a HB

Figure 2: This is a 4′ square piece of a FORS1 image in the centre of the Cetus dwarf galaxy,taken through the B filter. It is the composite of six 600sec images. Each image was regis-tered to the nearest pixel and co-added. On the resulting images the stellar sources have aPSF measuring FWHM ~ 0.45 ″. The extended sources seen in this image are galaxies be-hind Cetus.

19

(and not where they lie) in conjunctionwith the RC and the RGB can constrainthe plausible SFHs and put limits on thenumber of stars which could have beenformed more than 6–8 Gyr ago, even ifthe precise SFH at these times cannotbe extracted (e.g., Han et al. 1997;Gallagher et al. 1998; Tolstoy et al.1998; Tolstoy 1998). The ratio of red toblue HB stars can also be used as anindicator of the age and metallicityspread of the oldest stellar populationsin a galaxy (e.g., Da Costa et al. 2000).

6. Summary

The earliest epochs of star formationin the Universe must have occurred inlow-metallicity environments. The onlyplace where we can study on-going ex-tremely low metallicity star formation indetail is in nearby dwarf irregular gal-axies. They typically have metallicitiesin the range 3–10% of solar values(e.g., Skillman et al. 1989). Looking atgalaxies which appear to be in transi-tion between a period of star formationand a period of quiescence may help todetermine what can effect star forma-tion on a galaxy wide-scale (albeit asmall galaxy) and result in the tre-

mendous diversity of SFHs seen indwarf galaxies.

The SFH of the nearby universeshould also reflect that determined fromredshift surveys (e.g., Steidel et al.1999). To make this comparison an ac-curate one, we need data like these todetermine when nearby galaxies, eventhe smallest ones, formed most of theirstars and if bursts were ever commonenough or bright enough to producefaint blue galaxies at intermediate red-shifts (e.g., Tolstoy 1998). The giantgalaxies (M 31, the Galaxy, M 33) con-tain something like 90% of the stellarmass in the Local Group, so this is un-deniably where the dominant mode ofstar formation occurs. However, theconcern remains that, because theproperties of faint blue galaxies resem-ble more late-type irregular galaxies,the redshift surveys are picking updwarf systems which are very bright foronly a short time and do not constitutea significant fraction of the star forma-tion in the universe (e.g., Fukugita et al.1998, Tolstoy 1998).

To date, the limiting factors in deter-mining accurate SFHs in the LocalGroup have been crowding, sensitivityand resolution limits for accurate stellar

photometry. HST first showed us themagnificent improvements in under-standing SFHs from CMDs made pos-sible by high-resolution imaging (e.g.,Tolstoy 1999). The large collecting areaand impressive image quality and sta-bility of UT1 combined with the ex-tremely good seeing attainable atParanal makes it possible to obtain ac-curate CMDs of relatively uncrowdeddI/dSph systems. FORS1 also allowsus to extend our knowledge of faint stel-lar populations into the blue, somethingwith which HST has difficulty, and alsoto image the entire area of a dwarfgalaxy (typically 3–5’ across) in oneshot. FORS1 is thus an extremely im-portant complement to WFPC2 on HST,which is needed to push deeper downthe main sequence to detect faint main-sequence turnoffs to quantify the contri-bution of the older populations seen inthe HB. FORS imaging would inevitablyhave crowding problems at these faintmagnitudes and with the increase incrowding.

We have shown that under goodconditions the VLT can deliver high-quality images that result in exquisitelydetailed CMDs which can be used todetermine the SFHs of all the galaxiesin our Local Group and ways beyond. Itis also worth looking further afield tostart surveying the high-mass stellarpopulations of other groups.

Acknowledgements: We thank thesupport staff at Paranal Observatory(particularly Thomas Szeifert, AndreasKaufer and Chris Lidman) who helpedus to make very efficient use of the tel-escope.

References

Caputo F., Castellani V. & Degl’Innocenti S.(1995) A&A, 304, 365.

Cole A.A., Tolstoy E., Gallagher J.S.,Hoessel J.G., Mould J.R., HoltzmanJ.A., Saha A., & the WFPC2 IDT team1999 AJ, 118, 1657.

Da Costa G.S., Armandroff T.E. & NorrisJ.E. 1992 AJ, 104, 154.

Da Costa G.S., Armandroff T.E., Caldwell N.& Seitzer P. 2000 AJ, in press (as-troph/9911020).

Dohm-Palmer R.C., Skillman E.D., Saha A.,Tolstoy E., Mateo M., Gallagher J.S.,Hoessel J. & Chiosi C. 1997 AJ, 114, 2514.

Ellis R. 1997 ARAA, 35, 389.Fukugita M., Hogan C.J. & Peebles P.J.E.

(1998) ApJ, 503, 518.Fusi-Pecci F. & Bellazzini M. 1997 Third

Conference on Faint Blue Stars, in press(astro-ph/9701026).

Gallagher J.S., Tolstoy E., Saha A., HoesselJ., Dohm-Palmer R.C., Skillman E.D. &Mateo M. 1998, AJ, 115, 1869.

Gallart C., Freedman W.L., Aparicio A.,Bertelli G. & Chiosi C. 1999 AJ, 118, 2245.

Girardi L. 1999 MNRAS, 308, 818.Girardi L., Bressan A., Bertelli G. & Chiosi C.

2000 A&A, 141, 371.Greggio L., Marconi G., Tosi M. & Focardi P.

1993 AJ, 105, 894.Han M., Hoessel J.G., Gallagher J.S.,

Holtzman J. & Stetson P.B. 1997, AJ,113,1001.

Figure 3: Here are plotted the Colour-Magnitude Diagrams which resulted from the data sum-marised in Table 2 for the Cetus, Aquarius and Phoenix dwarf galaxies, and the globular clus-ter Ruprecht 106. Representative error bars are also plotted for each data set. These datahave not been corrected for any reddening effects. From the Ruprecht 106 data a fiducialmean was found for the ROB and HB. This is then overplotted on each of the galactic CMDsin red. An accurate fit has not been made, these have just been overlaid on each CMD forthe purposes of comparison only. Before accurate conclusions can be drawn a detailed analy-sis of the reddening and distance uncertainties in all cases have to be made.

20

Minniti D. & Zijlstra A. 1996 ApJ, 467, L13.Lee M.G., Aparicio A., Tikonov N., Byun Y.-I.

& Kim E. 1999 AJ, 118, 853.Saha A., Sandage A., Labhardt L., Tammann

G.A., Macchetto F.D. & Panagia N. 1996ApJ, 466, 65.

Skillman E.D., Kennicutt R.C. & Hodge P.1989 ApJ, 347, 875.

Smecker-Hane T. A., Stetson P. B., HesserJ. E . & Lehnert M. D. 1994 AJ, 108,507.

Steidel C.C., Adelberger K.L., Giavalisco M.,Dickinson M. & Pettini M. 1999, ApJ,519, 1.

Tolstoy E. & Saha A. 1996, ApJ, 462,672.

Tolstoy E., Gallagher J.S., Cole A.A.,Hoessel J., Saha A., Dohm-Palmer R.C.,Skillman E.D., Mateo M. & Hurley-KellerD. 1998, AJ, 116, 1244.

Tolstoy E. 1998, XVIIIth Moriond Astro-physics Meeting: Dwarf Galaxies and

Cosmology, eds. T.X. Thuan et al., inpress (astro-ph/9807154).

Tolstoy E. 1999, IAU Symposium 192, TheStellar Content of Local Group Galaxies,eds. Whitelock & Cannon, p. 218.

Tosi M., Focardi P., Greggio L. & Marconi G.1989, The Messenger, 57, 57.

Tosi M., Greggio L., Marconi G. & Focardi P.1991, AJ, 102, 951.

Whiting A.B., Hau G.K.T. & Irwin M. 1999,AJ, 118, 2767.

FIRES at the VL T: the Faint InfraRed Extragalactic SurveyM. FRANX1, A. MOORWOOD2, H.-W. RIX 3, K. KUIJKEN4, H. RÖTTGERING1,P. VAN DER WERF1, P. VAN DOKKUM1 , I. LABBE1, G. RUDNICK 3

1Leiden Observatory, 2ESO, 3MPIA, 4University of Groningen

One of the unique capabilities of theVLT is the near-infrared imaging modeof ISAAC. The wide field of view, detec-tor stability, and image quality thatISAAC can provide are currently unpar-alleled at other observatories.

In order to take advantage of thiswindow of opportunity, we have pro-posed a non-proprietary, deep imagingsurvey. The survey consists of verydeep exposures in the Js, H and Ksbands of the Hubble Deep Field South(Williams et al. 2000), and somewhatshallower imaging on a wider field in thecluster MS1054-03 at z = 0.83 (Gioiaand Luppino 1994, van Dokkum et al.1999, 2000). The survey is aimed at thestudy of distant galaxies in both fields,although many other applications willbe possible. The same fields are ob-served at other wavelengths by severalother groups. The full complement ofdata will range from the radio to the X-ray. The fields were selected based ontheir superb optical imaging from theHubble Space Telescope. Both fieldswill be observed for a total of 96 hours,split evenly between Js, H and Ks. In thefield of MS1054-03, this integration timewill be divided between four point-ings to cover the full Hubble SpaceTelescope mosaic of images (vanDokkum et al. 2000). For the HubbleDeep Field South, one single pointingwill cover the WFPC image. The ex-pected depth of the images is a Ks mag-nitude of 24.4 (26.3) in the HDF South,and a Ks magnitude of 23.7 (25.6) inMS1054-03; these are 3σ limits inJohnson and AB magnitudes, respec-tively. The OPC has been very gener-ous and has allocated the full request-ed time to this survey.

The first data on the Hubble DeepField South were obtained at the end of1999. Even the first quarter of the totalintegration provided Ks-band data ofunprecedented image depth and quali-ty for the HDFS. A colour image is

shown in Figure 1, which is a combina-tion of the HST I band data, and Js andKs ISAAC data. The variations amongthe infrared colours of galaxies arestriking, and several very red galaxiesare immediately apparent in the image.

Some of these galaxies are absent, orvery faint in the extremely deep I banddata from the Hubble Space Telescope.

Some examples of distant galaxiesare shown in Figure 2. The morpholo-gies of the very red galaxies span a

Figure 1. A colour image of the Hubble Deep Field South, observed with the Hubble SpaceTelescope, and ISAAC on the VLT. The colour image is constructed from the I band imagesfrom HST, and the Js and Ks images taken with ISAAC. The outline indicates the size andshape of the WFPC2 field, and the galaxy colours outside of that are yellow-green becauseof the absence of I band data in the image. The great variety of galaxy colours is striking.These differences are often caused by redshift: at higher redshifts, the Balmer break and4000 Angstrom break will move in or beyond the bandpass of the I band filter. Some very redgalaxies can be identified, these have very low I and Js fluxes.

21

wide range: some are compact at allwavelengths, others are compact in theinfrared, but much more extended in theoptical wavelengths. The very red col-ours of these galaxies are likely causedby the redshifted Balmer or 4000-Ang-strom breaks or by dust. The spectralenergy distributions of three of the gal-axies are shown in Figure 3. The redshiftof the galaxy in Figure 3a is expected tobe above 1.6, whereas the others areexpected to be above 2 based on theirspectral energy distribution.

Some of the morphologies are quiteremarkable. The very large galaxy inFigure 2d is an example. It is very ex-tended in the I band, which most likelycorresponds to the rest-frame UV. TheHST images show spiral structure, sim-ilar to that of spiral galaxies nearby. TheKs band image show a centrally con-centrated component, which is likelythe bulge of the galaxy. This galaxy istherefore possibly the progenitor of alarge-disk galaxy at low redshift. If thisis confirmed with deeper data, it wouldbe in conflict with theories of disk for-mation which predict generally smalldisks at high redshift. Another strikingexample is shown in Figure 2e. Thisgalaxy has extremely weak I bandemission, and was not included in thesource list of galaxies for the HDFSouth (Williams et al. 2000), but we

note that it is close to the edge betweenWF and PC. The galaxy in Figure 2c isred in the I-Js, but very blue in the opti-cal colours. It is a regular Ly-Breakgalaxy, with a remarkable difference inmorphology between the IR and optical.The Ly-break would predict the Balmerbreak to occur further to the red (red-ward of Js). Further study is required tosee whether dust or other factors mayplay a role.

The first results of the survey arevery promising. When the survey iscompleted, it will be unique in terms ofdepth and coverage. The data can beused for many purposes: the study of Jdrop-out objects, which are possibly thehighest redshift galaxies (e.g., Dickinsonet al. 1999); the study of the rest-frameoptical properties of Ly-break galaxies,the determination of photometric red-shifts of a large sample of galaxies, es-pecially for galaxies with redshifts be-tween 1 and 3; the weak lensing signalin the Hubble Deep Field South andMS1054-03, aided by accurate photo-metric redshifts; the faint end of the lu-minosity function at lower redshifts,both for the field and MS 1054-03, etc.

One of the interesting applications isthe selection of distant galaxies purelybased on their observed infrared fluxes.As we have shown in Figures 2 and 3,some galaxies are very red in their Js-

Ks and/or Js-H colours. These galaxiesare so faint in the optical, that they willgenerally not be in samples selected bythe Ly-break technique (Steidel et al.1996). Nevertheless, their contributionto the total (rest-frame) optical luminos-ity density at high redshift can be quitesubstantial. Given the fact that theirmass-to-light ratio is also expected tobe higher than that of Ly-break galax-ies, their contribution to the stellar massdensity at high redshift could be signifi-cant. Analysis of this aspect is inprogress.

The reduced data of the survey will bemade available to the public. We expecta first release of the preliminary HDFSouth data later this year. The data re-duction is relatively complex due to thelarge number of frames taken in jittermode, some of which were affected byscattered moonlight. For the most accu-rate astrometry and combination withother images, it is also necessary to cor-rect for a small amount of field distortion.

We have currently written our ownreduction routines based on the IRAF“dimsum” package. The observationson the cluster field MS1054-03 are ex-pected to be done soon, but they havenot yet started.

Updates on the programme can befound at our web site http://www.strw.leidenuniv.nl/~franx/fires.

Figure 2. A close-up of several high-redshift galaxies in the Hubble Deep Field South. The individual panels show the images of the galaxiesin the optical from HST (U, B, V, I), and in the infrared from the VLT (Js, H, Ks). The differences in morphology are quite striking. Thanks tothe good image quality of the VLT data, many of the galaxies are resolved in the infrared. The large galaxy in panel d is similar to spiral galax-ies in the nearby universe. It is likely at a redshift above 2. The galaxy in panel e has not been identified as a galaxy from the HST-WFPCdata, whereas it is bright in the infrared. Such very red galaxies can contribute substantially to the stellar mass density at high redshift.

22

Acknowledgement

It is a pleasure to thank the staff atESO who contributed to the construc-tion and operation of the VLT andISAAC. This project has only beenpossible because of their enormous ef-forts.

ReferencesDickinson, M., et al, 1999, preprint, as-

troph/9908083.Gioia, I., and Luppino, G. A., 1994, ApJS,

94, 583.van Dokkum, P. G., Franx, M., Fabricant, D.,

Kelson, D., Illingworth, G. D., 1999, ApJL,520, L95.

van Dokkum, P. G., Franx, M., Fabricant, D.,Kelson, D., Illingworth, G. D., 2000, sub-mitted to ApJ.

Steidel, C. C., Giavalisco, M., Pettini, M.,Dickinson, M., Adelberger, K. L., 1996,ApJL, 462, L17.

Williams, R. E., et al, 2000, in prepara-tion.

Figure 3: The normalised spectral energy distribution of 3 galaxies. From left to right we show a regular Ly-break galaxy (Fig. 2c), the “spiral”galaxy (Fig. 2d), and the very red galaxy from Figure 2e. The red continuum feature of the last two galaxies can be due to the Balmer/4000Angstrom break or due to dust. Only one of these would be selected by the regular Ly-break selection technique, as the others are too faintin the optical (rest-frame UV).

Optical Observations of Pulsars: the ESO ContributionR.P. MIGNANI1, P.A. CARAVEO2 and G.F. BIGNAMI3

1ST-ECF, rmignani@eso.org; 2IFC-CNR, pat@ifctr.mi.cnr.it; 3ASI bignami@asi.it

Introduction

Our knowledge of the optical emis-sion properties of neutron stars hasbeen remarkably improved by the re-sults obtained during the last 15 years.At the beginning of the 80s, only two ofthe about 500 isolated neutron stars atthat time detected as radio pulsars hadbeen identified also at optical wave-lengths. These were the two young(2000–10,000 years) optical pulsars inthe Crab (Cocke et al. 1969), the firstand for about 10 years the only one,and Vela (Wallace 1977) supernovaremnants. Soon after the identificationof the Crab, a model to explain theoptical emission of young pulsars wasdeveloped by Pacini (1971) in termsof synchrotron radiation emitted bycharged particles injected in the pul-sar’s magnetosphere. According to thismodel, the optical luminosity of a pulsaris predicted to scale proportionally toB4P–10, where B and P are its magnet-ic field at the light cylinder and its peri-od, respectively. This relation, known asthe “Pacini’s Law”, proved correct forboth the Crab and the Vela pulsars andwas thereon assumed as a reference.

The panorama of neutron stars’ opti-cal astronomy changed rapidly whenfew more pulsars started to be identi-fied (or discovered) in the X-ray data ofthe EINSTEIN satellite (see Sewardand Wang 1988 for a review) and apossible X-ray counterpart for the enig-

matic gamma-rays source Geminga,not yet recognised as an X/gamma-raypulsar, was proposed. This triggeredthe search for their optical counterparts,

and ESO telescopes gave to theEuropean astronomers the chance toboost a virtually new field of investiga-tion (see Table 1 for a summary of the

Table 1: Summary of (published) ESO observations of the pulsars with an optical counterpart.For each pulsar the observing epochs and the telescopes are listed. Observations performedby the authors are marked by an asterisk. The observing modes are specified by the colourcode. Arrows indicate HST follow-ups. The results are summarised in Table 2 and describedin the text.

23

published ESO observations). Also tak-ing advantage of the improved perform-ances of the NTT and its new detectors,more optical identifications of X-ray pul-sars were achieved in few years, somebrand new, some the confirmation ofprevious detections obtained with the3.6-m. In particular, middle-aged (≥100,000 years) pulsars were observedfor the first time. This opened the wayto the study of their optical emissionproperties, which turned out to be verydifferent from the ones of young Crab-like pulsars, thus changing a well-es-tablished scenario. In parallel, fast pho-tometry observations were pursued atthe 3.6-m, also experimenting the newtechnology of the MAMA detectors, tomonitor the light-curve evolution of theknown optical pulsars and to search fornew ones (a case for all: the long questfor a pulsar in SN 1987A). Last, but notleast, precise proper-motion measure-ments of pulsars, so far obtained only inthe radio band, were started at ESO us-ing classical optical astrometry tech-niques, yielding results of comparable(or even higher) accuracy.

Thereafter, the stage was taken bythe HST, which, exploiting its highersensitivity in B/UV, obtained three newlikely identifications and complementedthe explorative work done with ESO tel-escopes for the pulsars already identi-fied. Indeed, one can go as far as say-ing that almost all the HST time so farallocated for the study of isolated neu-tron stars has been a follow-up of ESOprogrammes (see Table 1).

Observations Review

The total number of pulsars with anoptical counterpart, either secured ortentative, amounts now to nine. Theavailable optical database (consistingof timing, spectroscopy, polarimetry

and photometry) is summarised inTable 2, where the objects have beensorted according to their spin-down age(P/[2dP/dt]). From the table, it can beimmediately appreciated how the ob-servations with the ESO telescopesplayed an important role in the opticalstudies of pulsars, claiming a number ofabsolute firsts. For each pulsar, the ma-jor results achieved so far by ESO ob-servations are discussed in the follow-ing sections.

The Crab Pulsar

The bright pulsar (33 ms) in the CrabNebula (PSR B0531+21) was the firstone to be identified in the optical(Cocke et al. 1969). Although the Crabis relatively bright (V = 16.6), the firstgood spectrum was taken only 20 yearsafter the pulsar discovery with EMMI atthe NTT (Nasuti et al. 1996a,b). Thepulsar spectrum appears flat (see Fig.1of Nasuti et al. 1996a), as expectedfrom a synchrotron origin of the opticalradiation, apart from an unidentifiedbroad absorption feature observedaround 5900 Å, which could be origi-nated in the pulsar magnetosphere.Photometry of the Crab pulsar, per-formed at different epochs from ESO,was used to critically investigate the re-ality of the so-called secular decreaseof the pulsar’s optical luminosity. Thiseffect, predicted by the Pacini’s Law asa consequence of the pulsar’s spin-down, was never convincingly meas-ured. By comparing the V flux meas-urements of the Crab taken over 15years, a decrease of 0.008 ± 0.004mag/yr was indeed found (Nasuti et al.1996a,b), consistent with the theoreti-cal prediction (0.005 mag/yr), but stilltoo uncertain to prove the reality of theeffect.

PSR B1509-58

With a spin-down age close to 1500years, the pulsar PSR B1509-58 is theyoungest after the Crab. Its period (P =150 ms) is relatively long with respect tothe Crab one, but it spins down muchfaster than almost any other pulsar(Kaspi et al. 2000). A candidate coun-terpart to PSR B1509-58 (V = 22.1) wasfirst detected by the NTT (Caraveo etal. 1994b), with a corresponding opticalluminosity much higher than the oneexpected from the Pacini’s Law. Theproposed identification was investigat-ed in the following years through multi-colour imaging, spectroscopy and tim-ing performed at the NTT (EMMI andSUSI1/2) and at the 3.6-m, which, how-ever, lead to inconclusive results (seeMignani et al. 1998a for a summary). Inparticular, the non-detection of opticalpulsations raised doubts on the pro-posed identification. Recently, the pul-sar field was observed in polarimetrymode with the FORS1 instrument at theVLT/UT1. Exploiting excellent seeingconditions (0.46″), the proposed coun-terpart of Caraveo et al. (1994b) wasresolved in a triplet of objects. Of these,only one (R = 25.7) showed evidence ofa significant polarisation (Wagner andSeifert 2000), as expected from puremagnetospheric optical emission froma young pulsar. Although its luminositywould still exceed the predicted one,the polarisation signature makes thisnew candidate a viable counterpart toPSR B1509-58.

PSR B0540-69

The fourth youngest pulsar (50 ms)known so far (~ 1,700 years) is PSRB0540-69 in the Large MagellanicCloud. With the discovery of optical pul-sations (Middleditch & Pennypacker

Table 2: Summary of the existing optical database for all the pulsars identified so far. The objects (first column) are sorted according to theirspin-down age in years (second column, in logarithmic units). For PSR B1509-58 both the original and the most recent candidate counter-parts are reported. The row colour identifies (in increasing intensity) young, middle-aged and old pulsars. The 3rd column gives the identifi-cation evidence either from positional coincidence (Pos), timing (Tim), proper motion (PM) or polarimetry (Pol). Proper-motion measurement,timing – either resulting in the detection (P) or non-detection (UP) – of optical pulsations, spectroscopy and polarimetry observations areflagged in columns 4–7. Null results (–) and upper limits (u.l.) are also noted. The remaining columns list the available (time integrated) UB-VRI photometry.

Pulsar Age Id. PM Tim Spec Pol I R V B U

Crab 3.10 Tim Y P Y Y 15.63 16.21 16.65 17.16 16.69B1509-58 3.19 Pos UP - u.l. 19.8 20.8 22.1 23.8

Pol Y 25.7B0540-69 3.22 Pos,Tim P Y Y 21.5 21.8 22.5 22.7 22.05Vela 4.05 Tim Y P Y 23.9 23.6 23.9 23.8B0656+14 5.05 Pos ,Tim Y P 23.8 24.5 25 24.8 24.1Geminga 5.53 PM,Tim Y P Y <26.4 25.5 25.5 25.7 24.9B1055-52 5.73 Pos 24.9B1929+10 6.49 Pos >26.2 25.7B0950+08 7.24 Pos 27.1

IDENTIFIED FROM ESO

MEASUREMENTS FIRST OBTAINED FROM ESO

24

1985), it became the third optical pulsarafter the Crab and Vela ones. Whenbeaming effects were taken into ac-count (Pacini and Salvati 1987), the op-tical luminosity of PSR B0540-69 wasfound to be consistent with the predic-tions of the Pacini’s Law. The pulsarwas monitored between 1989 and 1991through fast-photometry observationsperformed at the ESO/3.6-m (Gouiffeset al. 1992). These observations al-lowed to measure very accurately thepulsar’s timing parameters (dP/dt andd2P/dt2), to be used as input to derivethe value of its braking index, i.e. aquantity of paramount importance forneutron star models (see e.g. Shapiro &Teukolsky 1983). Since the discovery ofthe weak radio emission from PSRB0540-69 is quite recent, optical (aswell as X-ray) observations were theonly way to perform a very accuratepulsar timing.

Although optical pulsations had beendetected from the direction of PSRB0540-69 soon after the X-ray discov-ery, its counterpart remained unidenti-fied due to the lack of a very precise po-sitioning. The first deep search was car-ried on in January 1992 with SUSI atthe NTT (Caraveo et al. 1992). The cluefor identifying the optical counterpart ofthe pulsar came from an Hα image ofthe host supernova remnant (SNR0540-69), which unveiled a strange spi-ral-like structure, somehow recalling asort of jet-like emission activity. A point-like source (V = 22.5) was indeed ob-served coincident with the centre of the“spiral”. The identification proposed byCaraveo et al. (1992) was later con-firmed by time-resolved imaging of thefield performed at the NTT with theguest TRIFFID camera equipped with aMAMA detector (Shearer et al. 1994),which allowed to determine the positionof the optical pulsar. Recently, a newpiece of information was added by thefirst measurement of the pulsar’s opti-cal polarisation, obtained by Wagner &Seifert (2000) using FORS1 at theVLT/UT1.

The Vela Pulsar

The Vela pulsar (PSR B0833-45) isthe older (11,000 years) of the “youngpulsar” group. Together with the Crab, itis the only one originally detected on aphotographic plate (Lasker 1976). Itsoptical pulsations (89 ms) were exten-sively studied by Gouiffes (1998)through fast-photometry observationscarried out at the 3.6-m thus yieldingthe most accurate characterisation ofthe pulsar’s light curve.

Multi-epoch imaging with the NTTand the 3.6-m was fundamental to as-sess the actual value of the Vela pulsarproper motion, measured several timesin radio with different instruments andtechniques but with conflicting results.Indeed, optical astrometry is not affect-ed by pulsar’s timing irregularities, very

frequent in the case of Vela, whichhamper significantly radio proper-mo-tion measurements. After an upper lim-it first obtained by Bignami & Caraveo(1988) with the 3.6-m, the measure-ment of the proper motion was carriedout by Ogelman et al. (1989) at the 2.2-m. A few years later, the proper motionwas revisited with the aid of new NTTobservations by Nasuti et al. (1997a,b),who computed the angular displace-ment of the pulsar during 20 years andobtained a value of 52 ± 5 mas/yr, re-cently confirmed by DeLuca et al.(2000) using the HST.

Amongst the most recent results, im-aging observations of the Vela pulsar,performed with EMMI at the NTT, al-lowed to improve its multicolour pho-tometry and to add the first detection inR (Nasuti et al. 1997a). As expectedfrom this young pulsar, the UBVR fluxvalues are clearly indicative that the op-tical emission is of magnetospheric ori-gin. This conclusion is supported by thefirst measurement of a significant opti-cal polarisation from the pulsar(Wagner & Seifert 2000), obtained us-ing FORS1 at the VLT/UT1. However,as remarked by Nasuti et al. (1997a),the broad-band spectral behaviour ofVela is significantly different from theone observed in the other young opticalpulsars Crab and PSR B0540-59. Whilein all cases the UBV colours appear tofollow the same flattish spectral distri-bution, the relatively lower R flux of Velarepresents a clear spectral turnover inthe red region of the spectrum. The ori-gin of this turnover is unclear. We notethat a similar trend was observed forthe Crab in the NIR region and was in-terpreted as a synchrotron self-absorp-tion taking place in the pulsar magne-tosphere (see Nasuti et al. 1997a andreferences therein).

PSR B0656+14

With a spin-down age of 100,000years, PSR B0656+14 (384 ms) is oneof the two middle-aged pulsars identi-fied from ESO. A possible optical coun-terpart was detected by Caraveo et al.(1994a) in a faint (V ~ 25) point sourcedetected in images obtained in 1989(3.6-m) and 1991 (NTT) and later con-firmed by an HST/WFPC2 observation(Mignani et al. 1997b). Since propermotion is a very distinctive characteris-tic of each pulsar, the known radio dis-placement of PSR B0656+14 wassearched in the optical but, owing to thepoorer angular resolution of the 3.6-mand NTT images, only a marginal resultwas obtained.

Geminga

The optical identification of the mid-dle-aged (340,000 years) Geminga pul-sar (237 ms) with a faint star named G”(V = 25.5), proposed by Bignami et al.(1987), was initially substantiated by its

unusual colours (Halpern & Tytler1987). In particular, multicolour pho-tometry observations performed at the3.6-m (Bignami et al. 1988) showed forthe first time that its optical flux distribu-tion could not be explained by a simplespectral model, as happens e.g. for theCrab, and that different emission mech-anisms were at work. A few years later,the optical identification of Gemingawas supported by the measurement ofthe G” proper motion, obtained thanksto a new NTT/SUSI observation(Bignami et al. 1993) and later re-assessed both with the NTT/SUSI(Mignani et al. 1994) and with theHST/WFPC2 (Caraveo et al. 1996).The definitive proof of the optical identi-fication came when the gamma-raylight curve of Geminga was found to besensitive to the G” proper motion(Mattox et al. 1996), finally closing theloop.

NTT/SUSI observations of the Ge-minga field were also fundamental totie, through a multi-step astrometricprocedure, the Hipparcos referenceframe to the HST/WFPC2 one and toobtain the absolute coordinates of thepulsar with an accuracy of only 40 mas(Caraveo et al. 1998). This, togetherwith the refined measurement of theproper motion, obtained by the HST(Caraveo et al. 1996), makes it possibleto compute the absolute position ofGeminga at a given epoch, a critical in-formation to phase together its gamma-ray pulsations over 20 years of obser-vations (Mattox et al. 1998).

The last, although still weak, proof ofthe optical identification of Gemingacame recently with the tentative detec-tion of optical pulsations from G” (237ms), a result partially obtained using theTRIFFID camera at the 3.6-m (Sheareret al. 1998).

The Chase Goes On

A few more pulsars were observedthrough the years from ESO but no newoptical identification was achieved.Among the best-studied targets, we re-call the middle-aged (540,000 years)pulsar (197 ms) PSR B1055-52, ob-served for the first time from the 3.6-m(Bignami et al. 1988) and later alsofrom the NTT. With a flux comparable tothe one of Geminga (see Table 2), thispulsar would probably have been de-tected if it were not buried in the PSF’swings of a very close and bright star(see Fig. 1 of Mignani et al. 1997a),which made mandatory the use of theHST. The more recent (and deeper) op-tical investigations are those of twoyoung pulsars: the Crab-like PSRJ0537-6910 (5,000 years) and the Vela-like PSR B1706-44 (17,000 years), ob-served respectively with the NTT andthe VLT.

The X-ray pulsar PSR J0537-6910 inthe Large Magellanic Clouds superno-va remnant N157B is the fastest (16

25

ms) known isolated (non-recycled) pul-sar and the most energetic (togetherwith the Crab). Although the X-ray de-tection has been confirmed by all oper-ating X-ray satellites, PSR J0537-6910is still undetected in radio (see Mignaniet al. 2000 and references therein). Inthe optical, the field was observed in1998 in the BVI bands with the SUSI2camera at the NTT. Few objects weredetected close/inside the X-ray errorcircle but none of them could be con-vincingly associated to the pulsar,which appears undetected down to V~ 23.4 (Mignani et al. 2000). This resulthas been independently confirmed bythe upper limits on the pulsed opticalflux obtained by Gouiffes & Ogelman(2000) from the reanalysis of fast-pho-tometry observations of the field per-formed with the ESO fast photometer atthe 3.6-m.

PSR B1706-44 is a radio/gamma-raypulsar (P = 102 ms) and a weak soft X-ray source (see Mignani et al. 1999 andreferences therein), very similar to theVela pulsar both in its dynamical pa-rameters and in its multiwavelengthphenomenology. After a few (unpub-lished) trials with the NTT, deep opticalobservations of PSR B1706-44 wereobtained in August 1998 during the

Science Verification programme of theVLT/UT1 using the Test Camera (Mig-nani et al. 1999; Lundquist et al. 1999).Unfortunately, no candidate opticalcounterpart to the pulsar has been de-tected down to V ~ 25–27, dependingon the computed image astrometry.

The General Picture

Although still incomplete, the infor-mation obtained so far is sufficient todraw a first, general, picture of the opti-cal emission properties of pulsars (seeMignani 1998 for a review). To sum-marise, two different processes cancontribute to the optical emission: (i)synchrotron radiation from the magne-tosphere, fuelled by the neutron star’srotational energy loss dE/dt, and (ii)thermal (T ≈ 105–106 °K) radiation fromthe cooling neutron star’s surface,which peaks in the soft X-ray band.

The first process is dominant inyoung and bright objects but its effi-ciency seems to decay rapidly (Fig. 1),to wit, the case of the Vela pulsar, whichis only 5 times older than the Crab butchannels in the optical approximately a1,000 times smaller fraction of its glob-al energy budget. For the middle-agedand fainter pulsars, the scenario is less

clear. First, the optical luminosity fallswell below the value predicted by thePacini’s Law and its dependence on thespin-down age is smoother. Second,the available multicolour flux measure-ments are clearly inconsistent with asingle spectral model, thus implyingthat both the emission processes de-scribed above probably contribute tothe optical luminosity. In the case ofPSR B0656+14, the optical flux distri-bution can be fit by the combination ofboth a magnetospheric and a thermalcomponent (Pavlov et al. 1997). On theother hand, for the slightly olderGeminga, the optical continuum seemsto be thermal with a broad emissionfeature at 6,000 Å (Mignani et al.1998b), which, following the originalidea of Bignami et al. (1988), has beeninterpreted as a possible Hydrogen cy-clotron emission line originated in theneutron star’s atmosphere (Jacchia etal. 1999). For older pulsars (>1,000,000years), thermal emission is thought tobecome finally dominant over the mag-netospheric one (Pavlov et al. 1996).

The Future

Although the last years marked a bigstep forward in the optical astronomy ofpulsars, much work remains to be doneto understand the emission propertiesof these objects. While the above pic-ture seems substantially correct, thequestion of when and how the opticalemission of young pulsars starts to fadeaway and switches from a pure magne-tospheric to a thermal regime is still anopen issue. Optical observations ofmore pulsars, especially in the age in-terval 5000–100,000 years, are thuscrucial to find an answer. Starting fromthe updated pulsar database, nowcounting ≈ 500 additional detections(e.g. Camilo et al. 2000), future Chan-dra and XMM observations should pin-point targets worth of optical follow-ups.

The knowledge of the pulsars spec-tra in the optical domain is also ratherpoor. Up to now, it was based on spec-tral information gathered through multi-colour photometry and hidden in timingand polarimetry data. Only for the brightCrab pulsar (Nasuti et al. 1996a,b) andPSR B0540-69 (Hill et al. 1997) a ratheraccurate spectroscopy is available.Future observations of pulsars shouldcertainly be more focussed on spec-troscopy. Amongst young optical pul-sars, Vela is certainly the one that couldbenefit more from a comprehensivespectral study. An accurate determina-tion of its spectral shape would be cru-cial both to unveil emission/absorptionmechanisms in the pulsar’s magnetos-phere and to trace the transition in theemission physics between young andmiddle-aged pulsars. Spectroscopicobservations are also needed to re-solve the different components (mag-netospheric and thermal) so far onlyhinted in the multicolour flux distribu-

Figure 1: Extinction-corrected optical luminosities of identified pulsars plotted vs. their spin-down ages. In all cases but Geminga, for which an optical parallax measurement exists(Caraveo et al. 1996), the luminosities have been calculated assuming the nominal radio dis-tances. Available upper limits are reported for PSR J0537-6910 and PSR B1706-44 (in thiscase for the extreme values of the interstellar extinction). The different symbols indicatesyoung (hexagons), middle-aged (triangles) and old (squares) pulsars, with luminosities com-puted in the V (green), R (red) and U (magenta) bands.

26

tion of the middle-aged PSR B0656+14and Geminga, including possible at-mospheric cyclotron lines. A joint fit ofthe optical/X-ray thermal componentswould provide a better measurement ofthe observed neutron star surface tem-perature, important to investigate bothits internal structure and the physical/chemical properties of its atmosphere.On the other hand, fitting well-resolvedcyclotron lines would allow to obtain thefirst direct measurements of the neu-tron star magnetic field (Bignami 1998).For Geminga, a spectrum was recentlyobtained from the ground (Martin et al.1998) but with a very low S/N, while nospectroscopy observation has yet beentried for the slightly brighter PSRB0656+14.

To pursue all the above goals, deepimaging, timing and spectroscopic ob-servations, as feasible with 8-m-classtelescopes, are required. Needless tosay, in all cases the VLT could offer aninvaluable contribution. In particular,with FORS1 and FORS2, both Antu andKueyen fulfil the imaging/spectroscopyrequirements. On the other hand, tim-ing observations can be performed ei-ther by using FORS2 in trailer mode(see e.g. ESO/PR 17/99) or by usingMAMA detectors attached to the visitorfocus of Yepun. Thus, we can certainlysay that the VLT has all the potentiali-ties to open a new era in the optical as-tronomy of pulsars, giving back to ESOthe leadership achieved before the ad-vent of the HST.

References

Bignami, G.F., Caraveo, P.A., Paul, J.A.,Salotti, L. and Vigroux, L. 1987, ApJ 319,358.

Bignami, G.F., Caraveo, P.A. and Paul, J. A.1988, A&A 202, L1.

Bignami, G.F., Caraveo, P.A and Vacanti, G.1988, A&A 196, 191.

Bignami, G.F. and Caraveo, P. A 1988, ApJ325, L5.

Bignami, G.F., Caraveo, P.A. andMereghetti, S. 1993, Nature 361, 704.

Bignami, G.F. 1998, Adv.Sp.Res. Vol . 21/1,243.

Camilo, F. et al. 2000, Proc. of IAU Coll.N.177 PULSAR ASTRONOMY – 2000and Beyond.

Caraveo, P.A., Bignami, G.F. andMereghetti, S. 1994a, ApJ Lett. 422, L87.

Caraveo, P.A., Mereghetti, S. and Bignami,G.F. 1994b, ApJ 423, L125.

Caraveo, P.A., Bignami, G.F, Mignani, R.and Taff, L.G. 1996, ApJ 461, L91.

Caraveo, P.A., Lattanzi, M., Massone, G.,Mignani, R., Makarov V.V., Perryman,M.A.C. and Bignami, G.F. 1998, A&A 329,L1.

Cocke, W.J., Disney, M.J. and Taylor, D.J.1969, Nature 221, 525.

De Luca, A., Mignani, R.P. and Caraveo,P.A. 2000, A&A 354, 1001..

Gouiffes, C., Finley, J.P. and Ogelman, H.1992, ApJ 394, 58.

Gouiffes, C. and Ogelman, H. 2000, Proc. ofIAU Coll. N.177 PULSAR ASTRONOMY –2000 and Beyond.

Gouiffes, C. 1998, Proc. Neutron Stars andPulsars: 30 years after the discovery Eds.N. Shibazaki, N. Kawai, S. Shibata and T.Kifune, p.363, Universal Academic Press.

Halpern, J.P. and Tytler, D. 1987, ApJ 330,201.

Hill, R.J. et al. 1997, ApJ 486, L99.Jacchia, A., DeLuca, F., Lazzaro, E.,

Caraveo, P.A., Mignani, R. and Bignami,G.F. 1999, A&A 347, 494.

Kaspi, V.M. et al. 2000, Proc. of IAU Coll. N.177 PULSAR ASTRONOMY – 2000 andBeyond.

Lasker, B. M. 1976, ApJ 203, 193.Lundqvist, P. et al. 1999, A&A 343,L15.Martin, C, Halpern, J.P. and Schiminovich,

D. 1998, ApJ 494, L211.Mattox, J.R., Halpern, J.P. and Caraveo,

P.A. 1996, A&A Supp. 120, 77.Mattox, J.R., Halpern, J.P. and Caraveo,

P.A. 1998, ApJ 493, 891.

Middleditch, J. and Pennypacker, C. 1985,Nature 313, 659.

Mignani, R., Caraveo, P.A. and Bignami,G.F. 1994, The Messenger No.76, 32.

Mignani, R., Caraveo, P.A. and Bignami,G.F. 1997a, ApJ 474,L51.

Mignani, R., Caraveo, P.A. and Bignami,G.F. 1997b The Messenger 87, 43.

Mignani, R. et al. 1998a, The Messenger94, 25.

Mignani, R., Caraveo, P.A. and Bignami,G.F. 1998b, A&A 332, L37.

Mignani, R. 1998, Proc. Neutron Stars andPulsars: 30 years after the discoveryEds. N. Shibazaki, N. Kawai, S. Shibata &T. Kifune, p.335, Universal AcademicPress.

Mignani R., Caraveo, P.A. and Bignami, G.F.1999, A&A 343, L5.

Mignani, R., Pulone, L., Marconi, G.,Iannicola, G. and Caraveo, P.A. 2000,A&A – in press.

Nasuti, F. P., Mignani, R., Caraveo, P. A. andBignami, G.F. 1996a, A&A 1996, 314,849.

Nasuti, F. P., Mignani, R., Caraveo, P.A. andBignami. G.F. 1996b, The Messenger No.83,38.

Nasuti, F.P., Mignani, R., Caraveo, P.A. andBignami, G.F. 1997a, A&A 323,839.

Nasuti, F.P., Mignani, R., Caraveo, P.A., andBignami, G.F. 1997b, The Messenger No.87, 48.

Ogelman, H. et al. 1989, ApJ 342, L83.Pacini, F., 1971 ApJ 163, L17.Pacini, F. and Salvati, M. 1987, ApJ 321,

447.Pavlov, G.G., Stringfellow, G.S. and

Cordova, F.A. 1996, ApJ 467,370.Pavlov, G.G., Welty, A.D. and Cordova, F.A.,

1997 ApJ 489, L75.Seward, F. D. and Wang, Z. 1988, ApJ 332,

199.Shapiro, S. L. and Teukolsky, S.A. 1983, in

Black Holes, White Dwarfs and NeutronStars, Eds. Wiley and Sons.

Shearer, A. et al. 1994, ApJ 423, L51.Shearer, A. et al. 1998, A&A 335,L21.Wagner, S. and Seifert, W. 2000, Proc. of

IAU Coll. N.177 PULSAR ASTRONOMY –2000 and Beyond.

Wallace, P.T. et al. 1977, Nature 266, 692

1. Introduction

The redshift evolution of the spacedensity of galaxy clusters has longserved as a valuable tool with which totest models of structure formation andto set constraints on fundamental cos-mological parameters. Being recognis-able out to large redshifts, clusters arealso ideal laboratories to study the evo-

lutionary history of old stellar systems,such as E/S0s, back to early cosmiclook-back times. It is therefore not sur-prising that a considerable observation-al effort has been devoted over the lastdecade to the construction of homoge-neous samples of clusters over a largeredshift baseline. Until a few years ago,however, the difficulty of finding high-redshift clusters in deep optical images

The ROSAT Deep Cluster Survey: Probing theGalaxy Cluster Population out to z = 1.3P. ROSATI1, C. LIDMAN1, R. DELLA CECA2, S. BORGANI3, M. LOMBARDI1, S.A. STANFORD4, P.R. EISENHARDT5, G. SQUIRES6, R. GIACCONI7, C. NORMAN7

1European Southern Observatory; 2Osservatorio Astronomico di Brera, Milano; 3INFN, Sezione di Trieste;4Institute of Geophysics and Planetary Physics, Lawrence Livermore National Laboratory;5Jet Propulsion Laboratory; 6California Institute of Technology; 7Johns Hopkins University

and the limited sensitivity of early X-raysurveys had resulted in only a handfulof spectroscopically confirmed clustersat z > 0.5. The ROSAT satellite, with itsimproved sensitivity and spatial resolu-tion, made clusters high-contrast, ex-tended objects in the X-ray sky and hasthus allowed for a significant leap for-ward. About a thousand clusters havenow been selected from the ROSAT

27

All-Sky Survey at z Lp 0.3 and severalstatistical complete subsamples havebeen used to obtain a firm measure-ment of the local abundance of clustersand their spatial distribution (cf. Böhrin-ger et al. 1998). Soon after the ROSATarchive of pointed observations wasopened, it was realised that deep PSPCpointings could be used for seren-dipitous searches for distant clusters.

In this spirit, we embarked in 1994 onthe ROSAT Deep Cluster Survey withthe aim of constructing a large, X-rayflux-limited sample of distant galaxyclusters. In the period 1995–1998 weengaged in an intense optical identifi-cation programme, a substantial frac-tion of which was carried out with theEFOSC spectrograph at the 3.6-m at La

Silla. Over the last two years, our effortshave shifted to the near IR study of themost distant candidates and, most re-cently, to spectroscopic follow-up ofthese systems with the VLT. This ob-servational effort has produced thelargest sample of spectroscopically-confirmed distant clusters to date, withfour clusters confirmed beyond redshiftone. We summarise here some resultsand highlights from the entire survey.

2. Cluster Selection and theOptical Follow -up Work

The RDCS was designed to compilea purely X-ray selected sample ofgalaxy clusters, selected via a seren-dipitous search for extended X-ray

sources in ROSAT-PSPC deep pointedobservations (Rosati et al. 1995). Thelimiting X-ray flux and the solid angle ofthe survey were chosen to probe an ad-equate range of X-ray luminosities overa large redshift baseline. Approximately160 candidates were selected down tothe flux limit of 1 × 10–14 erg s–1 cm–2,over a total area of p 50 deg2, using awavelet-based detection technique.This technique is particularly efficient indiscriminating between point-like andextended, low surface brightnesssources.

Optical follow-up observations of thecluster candidates in both hemispheresconsisted primarily of optical imaging inthe I-band with 2- and 4-m-class tele-scopes at NOAO and ESO, followed by

Figure 1: RXJ0152.7-1357 (z = 0.83) – one of the most X-ray luminous clusters known at z > 0.5. The ROSAT PSPC X-ray contours are over-laid on a (V+R, J, K) composite image obtained with VLT/FORS1 and NTT/SOFI. The field of view is 4.7′ × 4.7′ (corresponding to 2.3 Mpch –1

50 at z = 0.83). The overall morphology of the X-ray emission, and the apparent galaxy distribution, are clearly elongated with two clumpspresent.

28

multi-object spectroscopy with 4-m-class telescopes at KPNO and ESO.Candidates were observed in order ofdecreasing X-ray flux to have a flux-lim-ited sample at any given time. EFOSCat the 3.6-m was used for the identifica-tion of the southern and equatorial clus-ters, and indeed, it proved to be an ide-al instrument for this work. Typically, a10-minute snapshot image was ob-tained and if a galaxy overdensity wasvisible around the X-ray peak, a MOSexposure was obtained the followingnight. Even with the old IHAP system,overheads were relatively small andmask alignment very reliable. This con-tributed to the overall efficiency of iden-tification, and approximately 60 south-ern cluster redshifts were secured over3 years with 20 allocated nights (80%clear). About a quarter of these clusterswere confirmed at z > 0.5. The highestredshift cluster (RXJ0152.7-1357) wasidentified with EFOSC1 at z = 0.83(Rosati et al. 1998, Della Ceca et al.2000). This cluster, shown in Figure 1,is one of the most X-ray luminousknown at z > 0.5 and possibly also oneof the most massive distant systems,perhaps akin to MS1054.4-0321 (Gioia& Luppino 1994).

To date, the entire RDCS samplecontains 115 new clusters or groupswith secure spectroscopic redshifts, 21of which lie at z > 0.6 and 10 at z > 0.8.The redshift distribution is shown inFigure 2 and compared with the EMSScluster sample, which has been the ba-sis of numerous studies of distant clus-ters in previous years. The RDCS hasclearly considerably extended the high-redshift tail.

The spectroscopic identification is90% complete to 3 × 10–14 erg s–1 cm–2;at this limiting flux, the completenesslevel and the selection function are bothwell understood. At fainter fluxes, theeffective sky coverage progressivelydecreases and the X-ray surface bright-ness limit becomes increasingly impor-tant. However, it is precisely in this low-est flux bin that the most distant clus-ters of the survey are expected to lie. Toimprove the success rate of identifyingz Lp 1 clusters, we have begun a pro-gramme of near-IR imaging of unidenti-fied faint candidates in the RDCS.Near-IR imaging is essential at theselarge redshifts to compensate the k-cor-rection, which significantly dims thedominant population of early-type clus-ter galaxies at observed optical wave-lengths. A dramatic example of this ef-fect is illustrated in Figure 3. A deep I-band image with a 4-m-class telescope(CTIO 4-m) had not shown any signifi-cant galaxy overdensity at the X-ray po-sition. On the other hand, a moderatelydeep image with SOFI at the NTT (bot-tom right) clearly revealed a red clumpof galaxies with a narrow J – K colourdistribution typical of early-type galax-ies at z > 1 (see the “yellow clump” inthe colour composite image). The same

clump is barely visible in a 1-hour R-band exposure with FORS1 (bottomleft). Cluster galaxies are fainter than K= 17, with colours as red as R – K = 6.FORS1 spectroscopy in March 1999yielded a redshift z = 1.23 for the 3brightest cluster members, and furtherspectroscopic work is scheduled withFORS1 in March 2000. It is worth em-phasising that spectroscopy of early-type galaxies at these redshifts is ex-tremely hard, even with the VLT, requir-ing no less than 5 hours integration tosecurely identify a few absorption fea-tures (primarily the H+K break) embed-ded in the OH sky forest. The cluster inFigure 3 is the highest redshift clusterconfirmed to date in the southern sky.Three other RDCS clusters in the north-ern sky were spectroscopically identi-fied with the Keck telescope at z > 1(Stanford et al. 1997, Rosati et al.1999).

3. Some RDCS Highlights

An immediate advantage of an X-rayselected cluster sample, like the RDCS,is that one can use statistics such asnumber counts and luminosity functionsto quantify the evolution of the clusterpopulation, similar to what is done forgalaxy redshift surveys. In Figure 4, weshow the X-ray luminosity function(XLF, i.e. the number of clusters per co-moving Mpc3, per unit X-ray luminosity)of a complete, spectroscopically identi-fied RDCS subsample with a limitingflux of 3.0 × 10–14 erg s–1 cm–2. Thisrepresents the best determination todate of the space density of distant

clusters out to z Q 1.2. A maximum like-lihood analysis of the sample confirmsthe visual impression that there is nosignificant evolution of the clusterspace density out to z Q 1, at luminosi-ties below the local L*X (P 4 × 1044 ergs–1 [0.5–2 keV], roughly the Coma clus-ter). Careful analysis of the bright endof the XLF instead shows some mildnegative evolution, i.e. a lack of themost luminous, possibly massive sys-tems (LX Lp 4 × 1044 erg s–1) at high red-shifts. This result is consistent with theoriginal findings of the EMSS (Gioia etal. 1990), although the strength of theeffect is still a matter of some debate(Rosati et al. 1998, 1999).

The identification of the most distantclusters in the RDCS allows an esti-mate of the cluster abundance to bemade for the first time at z P 1. This isshown in Figure 4 as a lower limit dueto the incomplete optical identificationat very low flux levels. According to thisestimate, there is at least one cluster asluminous as 1/5 of the Coma cluster in(150h–1

50)3 Mpc3 comoving, at z Q 1.

Figure 2: The cumulative redshift distribution of spectroscopically confirmed RDCS clustersto date. For comparison, the distribution of the EMSS sample (Gioia & Luppino, 1994) isshown.

EFigure 3: RJK composite image of the mostdistant RDCS cluster in the southern sky (z= 1.23) obtained combining FORS1 andSOFI images. The field of view is 3.8′ × 3.8′.Bottom: R-band and K-band image obtainedwith VLT/FORS1 and NTT/SOFI (1 hour ex-posure each), with overlaid ROSAT-PSPCcontours (scale in arcmin).

29

30

That the space density of the bulk ofthe cluster population is approximatelyconstant out to z Q 0.8 has interestingconsequences for cosmology. Despitethe fact that we are not measuring clus-ter masses but X-ray luminosities,which depend of the thermodynamicalhistory of the intra-cluster gas in a com-plicated fashion, it can be shown thatsignificant constraints can be placed oncosmological models within the uncer-tainties of the LX – M relation and itsevolution (see Fig. 5, Borgani et al.1999a, 1999b). This analysis showsthat it is difficult to reconcile an ΩM = 1universe with the RDCS data and ourcurrent knowledge of the LX – TX rela-tion for distant clusters. The fact thata large fraction of relatively massiveclusters is already in place at z p 1indicates that the dynamical evolutionof structure has proceeded at rela-tively slow pace since z Q 1, a scenariowhich fits naturally in a low-density uni-verse.

4. Conclusions

The sheer number of distant clustersspectroscopically identified in theRDCS illustrates the efficiency of themethodology used, as well as the strat-egy of the optical follow-up observa-tions which were largely carried out with4-m-class telescopes. By boosting thenumber of known clusters at z > 0.5 bya factor 5, the RDCS provides the basisfor several follow-up studies. Some of

these programmes, such as the evolu-tion of the cluster galaxy populationsand the cluster mass distribution vialensing and dynamical methods, havealready begun with the VLT.

In particular, the spectrophotometricproperties of cluster galaxies at z > 1are important diagnostics for constrain-ing the mode and epoch of formation ofE/S0 galaxies. The near-IR colours ofthe reddest members of the RDCSclusters which we have studied so far atz > 1 show that these galaxies form aremarkably homogeneous population,

a property which is often exploited toconstrain the formation scenario ofE/S0s (e.g. Stanford et al. 1997). DeepISAAC imaging would be crucial to in-vestigate the (unknown) faint end of thecluster galaxy population at severalmagnitudes below L*, at these high red-shifts. The time-consuming spectro-scopic work will benefit considerablyfrom the planned upgrade of FORS2with a red-sensitive CCD, coupled withhigh-throughput gratings.

Many exciting new observations ofthese clusters will be made in the nearfuture. Further insight on the physicalproperties (temperature, metallicity) ofthe gaseous component of the mostdistant RDCS clusters will be possiblewith scheduled Chandra and XMM ob-servations. In addition, significantprogress towards our understanding ofthe formation of the galaxy populationsat these large look-back times will comefrom combining VLT data with plannedobservations of these clusters with theAdvanced Camera on HST (scheduledin 2001).

References

Böhringer et al. 1998, The Messenger, 1998,No. 94, p. 21.

Borgani et al. 1999a, ApJ, 517, 40.Borgani et al. 1999b, in “Large Scale

Structure in the X-ray Universe”, eds. M.Plionis & I. Georgantopoulos, Santorini,Greece, September 1999 (astro-ph/9912378).

Della Ceca et al. 2000, A&A, 353, 498.Gioia et al. 1990, ApJ, 356, L35.Gioia, I.M. & Luppino, G.A. 1994, ApJS, 94,

583.Henry et al. 1992, ApJ, 386, 408.Rosati et al. 1995, ApJ, 445, L11.Rosati et al. 1998, ApJ, 492, L21.Rosati et al. 1999, AJ, 118, 76.Rosati et al. 1999, in “Large Scale Structure

in the X-ray Universe”, eds. M. Plionis & I.Georgantopoulos, Santorini, Greece,September 1999 (astro-ph/0001119).

Stanford et al. 1997, ApJ, 114, 2232.

Figure 4: The best determination to date of the cluster X-ray Luminosity Function out to z P1.2. Data points at z < 0.85 are derived from a complete RDCS sample of 103 clusters over47 deg2, with Flim = 3 × 10–14 erg s–1 cm–2. The triangles represent a lower limit (due to in-complete optical identification) to the cluster space density obtained from a sample of 4 clus-ters with ⟨z⟩ = 1.1 and with Flim = 1.5 × 10–14 erg s–1 cm–2.

Figure 5: Constraints in the plane of the cosmological parameters Ω0 (matter density param-eter)–σ8 (rms mass fluctuation on a 8h–1 Mpc scale) derived from the observed evolution ofthe cluster abundance in the RDCS sample (Borgani et al. 1999a, 1999b). Contours are 1σ,2σ and 3σ confidence level. The three parameters (A, α, β) describe the uncertainties in con-verting cluster masses into temperatures (T p M2/3 / b), and temperatures into X-ray lumi-nosities (LX p Tα (1 + z)A). The two values for each parameter bracket the range which is al-lowed from current X-ray observations of distant clusters.

31

Introduction: Why Spectral PSFSubtraction?

The first FORS follow-up run of theSTIS/NICMOS survey for Damped Lyα(DLA) galaxies was completed inAugust 1999. The purpose of the FORSspectroscopy is to search for Lyα emis-sion lines from candidate DLA galaxieswith projected distances from the un-derlying quasars as small as 0.30 arc-sec. The candidate DLA galaxies aretypically some 8–10 magnitudes fainterthan the underlying quasars.

To be able to look for the faint Lyαemission lines hidden under brightquasar spectra, we shall need a spec-tral reduction code optimised to sub-tract the two-dimensional (henceforth2D) signal of the quasars with ex-tremely high accuracy. By a strike ofgood fortune, a 1D spectral extractioncode was designed several years agowith the specific aim to provide the bestpossible determination of the SpectralPoint Spread Function along the slit.As outlined below there are minor,but important, differences between a2D imaging PSF and the 1D SpectralPSF. To avoid confusion, we shall inwhat follows write SPSF for the 1DSpectral PSF.

The example presented above, thesearch for high redshift DLA galaxies, ismerely one field in which SPSF sub-traction is essential. The technique ispotentially extremely useful in any fieldof astronomy dealing with separation ofobjects with small projected distanceson the sky, or even objects coveringeach other. In this and in a following is-sue of The Messenger we are going topresent three reports aimed at detailingthe exact properties of the algorithm tobe used for the reduction of our FORSspectra. This first report presents thedetails of the algorithm itself. The sec-ond report (Møller et al., 2000) presents,as a simple example of a research fieldin which exact SPSF subtraction can pro-vide interesting new results, the seren-dipitous discovery of extended Lyα emis-sion from the host galaxy of a radio-quiethigh-redshift QSO. The third and lastreport will present the first confirmedidentification of an intervening DLA gal-axy from the STIS/NICMOS survey.

2. Maximum S/N Extraction of1D Spectra

When 2D photon-counting detectorswere first attached to spectrographs, itwas immediately realised that by apply-ing a proper weighting scheme to allpixels of the 2D spectrum, one may op-timise the signal-to-noise (the S/N) ofthe extracted 1D spectrum. Determina-tion of the optimal weights requiresknowledge of the SPSF. The more pre-cise guess of the actual SPSF one isable to obtain, the higher the resultingS/N of the extracted 1D spectrum. It isalso true, however, that the improve-ment in S/N when going from a “fairlygood SPSF” to the “exact SPSF” will beminor in most cases. Therefore, for thepurpose of 1D extraction there is nostrong incentive to optimise the modelSPSF beyond a certain point. Extrac-tion algorithms not optimised in thissense present no problem as far as theextraction of 1D spectra is concerned,but they will not be adequate for ourpurposes. This point is of some impor-tance and will be detailed below.

In what follows we shall consider a2D spectrum which falls roughly alongthe columns of a CCD. The rows willthen (if we for now ignore the 2D dis-tortion in wavelength space) representbins in wavelength space.

A number of effects complicate thedetermination of the optimal weights:

• The SPSF is wavelength depend -ent. The seeing is typically better in thered than in the blue. For broadband im-aging this means that a blue object will

have a slightly larger PSF than a redobject. In spectroscopy, it means thatthe SPSF will be broad in the blue end,and narrow in the red end of the spec-trum.

• Focus changes along the spec -trum. Most spectrographs are not si-multaneously in focus over the entirewavelength range. One therefore has tochoose a compromise for the bestover-all focus. This may introduce aquite unpredictable change of theSPSF width along the spectrum.

• The spectrum does not lie exact -ly along the detector rows (or col -umns). Most spectrographs have a dis-torted spectrum which is “C” or “S”shaped. Hence, the SPSF is typicallyshifting from one column into severalother columns along the way.

The SPSF itself is a continuous andsmooth function. However, what we ac-tually observe is a discontinuous step-function induced by the process of ob-serving the continuous function with anarray of finite sized pixels. To empha-sise the difference (which is extremelyimportant in what follows) we shall referto the observed step-function as a “re-alisation” of the SPSF. For a givenSPSF, there will be an infinite number ofdifferent realisations, all differing byshifts on the detector smaller than a pix-el. An example of several different real-isations of the same SPSF from aFORS1 spectrum are shown in Figure 1.

Horne (1986) reported on an ex-tremely powerful and simple algorithmthat, as long as the spectrum is alignedwell with the CCD rows or columns, de-termines the SPSF realisations alongthe spectrum with very good accuracy.A generalisation of the Horne algorithmto the case of spectra not aligned wellwith the CCD was found by Marsh(1989). As pointed out above, optimalweights for maximum S/N 1D extractionroutines are derived from the SPSF re-alisations, and that was the goal of theHorne and Marsh algorithms. In the lim-it of very bright objects, the optimalweights are all identical and equal to 1,for faint objects the SPSF realisationsare needed to determine the weights.For very faint sources and for sourceswith strong absorption features, severaladditional problems come into play.First of all the SPSF is essentially im-possible to determine where there arebright skylines and where there isstrong absorption (such as in theLyman forest of quasar spectra).Hence, some sort of smoothing or fittingalgorithm must be used to interpolatebetween the poorly determined bins.The smoothing (or in the case of the

Spectral PSF Subtraction I: The SPSF Look -Up-Table MethodP. MØLLER, ESO

Figure 1: Attempt to represent the Look-Up-Table (LUT) used for the prediction of theSpectral PSF (SPSF) at any wavelength binof the 2D spectrum. In this example, the in-crease in sampling (n) is 25 and the lengthof the SPSF (l) is 13. Each LUT entry (a rowin Fig. 1) is a string of 13 numbers making upa single SPSF realisation (see Section 2).

32

Horne algorithm the polynomial fitting)introduces some imperfections in theweights. For the purpose of extractingthe 1D spectrum, those imperfectionsare mostly of no consequence. For thepurpose of searching for faint objectshidden under the bright spectrum, thereare three specific problems. First, theSPSF determination is in situ. Thismeans that the hidden object will betreated by the code as part of theSPSF, and it will be included in the datafor the polynomial fitting. A narrowemission line then becomes somewhatharder to detect, a broad emission linewill most likely be missed entirely, hav-ing been interpreted as a wavelength-dependent change in the SPSF.Second, most codes ignore the extend-ed wings of the SPSF because theycontain essentially no information (forthe purpose of extracting a 1D spec-trum). Fitting to low S/N wings in situwould in any case be a fit to mostly ran-dom noise. The wings are neverthelessimportant to remove if one wishes tosearch for faint extended emission.Third, since the main goal of our projectis to search for faint emission in theLyman forest of high redshift QSOs, wewould expect major problems if wewere to use the SPSF determined insitu on the basis of the few bits of QSOleft between the many strong absorp-tion lines in the forest.

All three problems listed above arerelated to, or made worse by, the factthat the SPSF determination is per-formed locally in each separate wave-length bin (in situ). Hence, the solutionis to use a global SPSF determinationalgorithm. A code based on this principle

was developed and installed at Copen-hagen University Observatory in 1989. Ithas been used only by myself and col-laborators, mostly for extraction of qua-sar spectra but lately also for extractionof faint galaxy spectra of early-type gal-axies at intermediate redshifts (Treu etal., 1999). A short summary of the meth-od was given in Møller & Kjærgaard(1992) but details of the algorithm havenever previously been published.

The code was developed for the pur-pose of 1D spectral extraction, but anextra option was included: The possibil-ity to subtract the 2D version of the ex-tracted spectrum. The 2D spectrum-re-moval option was intended mostly as acheck that the algorithm had workedproperly. In addition, since the code in-cluded “cosmic” rejection via sigmaclipping, it provided a possibility tocheck if cosmics had been overlookedby the code. The 2D-subtraction optionhas, however, already been successful-ly applied in the search for DLA galax-ies (Møller 1999).

3. The SPSF Look -Up-TableMethod

3.1 General Overview: The FourSteps

The algorithm is iterative and basedon four simple steps: (i) Trace the spec-trum (determine the centroid of theSPSF in each wavelength bin), (ii) de-termine the width of the SPSF in eachwavelength bin, (iii) determine the am-plitude of the SPSF in each wavelengthbin, (iv) predict the realisation of theSPSF in each wavelength bin.

Figure 2 shows an example of thefirst two steps, the “centroid trace” and“width” determinations on a FORS1/Grism-600B 2D spectrum. The secondand third steps in the above list are inreality performed simultaneously via atwo-parameter minimum χ2 fit as fol-lows. For a given guess of the width,the predicted SPSF realisation isscaled linearly in width, and the best-fit-ting amplitude is determined. The bestglobal two-parameter fit is found. It iseasy to show that minimum χ2 fitting ofan assumed SPSF realisation to the ac-tual data in each wavelength bin ismathematically equivalent to a maxi-mum S/N 1D extraction. After comple-tion of the first three steps, a new andbetter determination of the SPSF ispossible, and a new iteration of the foursteps can be performed. When the iter-ations converge towards a global mini-mum of the total χ2, the process may bestopped and the fitted 2D spectrumsubtracted from the input data frame.

The determination of the SPSF reali-sations is done on basis of the latest“trace”. The trace command determinesthe centroid of the SPSF with high ac-curacy in each wavelength bin. Thecode then subdivides all pixels into nsubpixels in the spatial direction (alongthe slit). Effectively it then builds up anempirical Look-Up-Table (LUT) of n dif-ferent realisations of the SPSF. Each ofthose realisations correspond to thesubset of rows in the input data framewhere the trace command determinedthat the centroid fell inside the samegiven sub-pixel. In practical terms, theLUT is an array of (I × n) numberswhere l is the selected length of theSPSF (in CCD pixels) and n is the res-olution into sub-pixels which is deemednecessary/possible for the given data.Typically we use n = 25, and will do soin the example given below.

3.2 The Core Algorithm: Creatingthe LUT

The final quality of the 2D spectralsubtraction is depending critically uponhow carefully the LUT was constructed.The first step of the LUT construction isthe summing of all data-rows with thesame sub-pixel trace position, whichpass a set of additional quality condi-tions (S/N, no cosmics found in thatbin). After the distribution of raw LUTdata is completed, all n realisations arenormalised. In Figure 1 we show (inblue) the normalised LUT determinedfrom the example FORS spectrum usedin Report 2. Clearly, the first entry(named LUT(1) in Figure 1) and the lastentry (LUT(25)) are neighbour realisa-tions in the same sense that e.g.LUT(3) and LUT(4) are, except thatthey are shifted by a full CCD pixelalong the slit. This is exemplified by thetwo realisations reproduced in red inFigure 1. E.g. LUT(25+) is identical toLUT(25) but shifted by one pixel. As the

Figure 2: Output of the “trace” and “width” commands. The left frame plots the centroid of theSPSF and covers three CCD columns. The right frame plots the relative width of the SPSF asa function of wavelength (blue is down, red is up). The data are from FORS1/GRISM-600B tak-en under excellent seeing conditions, but note that we have rotated the FORS data 90 degrees.

33

last step in the LUT construction algo-rithm, this process of expanding the 2DLUT by “adding on” shifted 1D LUT re-alisations is repeated (I × n) times. Byselection of the column correspondingto the leftmost pixel of LUT(1), the codeobtains a single string of the (I × n)numbers making up the complete LUT.This string of numbers hence containthe entire LUT, only now re-ordered in away so that it is a single smooth andcontinuous function. This function cor-responds to the input SPSF at n timesimproved sampling, but still containingthe observationally induced CCD pixelsmoothing.

It can be seen that if a single, or afew, of the individual entries in the 2DLUT grid are missing (by unfortunatelocations of cosmics, sky-lines and ab-sorption lines), then a whole row will bemissing in the 2D LUT grid. In the con-tinuous high-resolution representationof the LUT, however, a missing entrywill only result in a set of single missingvalues at each 25th sub-pixel. Thosecan easily be determined via interpola-tion between good values. After inter-polation to determine unknown values,possibly some smoothing if the spec-trum in general has low S/N, theprocess is reversed, and the 2D LUT isreconstructed from the interpolated/smoothed high-resolution representa-tion. This is the last step of a given iter-ation. If another iteration is started, thenew LUT will be used.

4. Degeneracy of Results andConcluding Remarks

We have presented an algorithm thatis specifically aimed at performing thebest possible removal of a 2D spectrumto reveal faint emission-line objects“hidden” under a bright spectrum. A fewnotes should be added. First: As de-tailed in Section 2, an algorithm withthis goal has to use a global approachto the determination of the SPSF. Aglobal approach is clearly incompatiblewith the fact that the SPSF varies alongthe spectrum; some additional assump-tion is required. In the code describedhere, we have assumed that the SPSFat any wavelength can be obtained viaa linear scaling in width of the samebase SPSF. Second: Clearly, if twopoint sources are super-imposed at ex-actly the same position, there is no waythe code can tell that there are in facttwo different objects. It will regard thesummed spectrum as that of a singlepoint source. Third: If the emission-lineobject is not a point source, and if it ex-tends across the bright point source,the code will try (as a default) to assignas much flux as possible to the pointsource. This means that part of theemission-line object will be assigned tothe spectrum of the bright point source,and the code will “dig a hole” in the ex-tended object at the position of thebright point source. In both of those twocases, there is hence a certain degen-

eracy in the final solution as to howmuch flux should be assigned to thetwo objects. This degeneracy can bebroken in several ways. Typically, onemay assume that either the spectrum ofthe point source is smooth and continu-ous in the given spectral region, or onemay assume that the extended objecthas a smooth and continuous surfacebrightness distribution at the position ofthe bright point source. In the case oftwo compact and well separated ob-jects no additional assumptions areneeded, the solution will be unique.

Acknowledgements

I wish to thank Michael I. Andersenwho supplied the routines that performthe optimal fitting of Chebyshev polyno-mials, and Sandra Savaglio who pro-vided extensive comments on severalother functionalities of this data-reduc-tion tool. I am grateful to Keith Horne formany useful comments on an earlierversion of this report.

References

Horne, K., 1986, PASP, 118, 609.Marsh, T.R., 1989, PASP, 101, 1032.Møller, P., Kjærgaard, P., 1992, A&A, 258, 234.Møller, P., 1999, in Astrophysics with the

NOT, Karttunen, H. and Piirola, V. (eds.),University of Turku, p. 80-88.

Møller, P., et al., 2000, The Messenger, 99, 33.Treu, T., Stiavelli, M., Casertano, S., Møller,

P., Bertin, G., 1999, MNRAS, 308, 1037.

SPSF Subtraction II: The Extended L yα Emission of a Radio Quiet QSOP. MØLLER1, S.J. WARREN2, S.M. FALL3, P. JAKOBSEN4, J.U. FYNBO5

1ESO, Karl-Schwarzschild-Str. 2, Garching, Germany2Blackett Laboratory, Imperial College of Science, Technology and Medicine, London, UK3Space Telescope Science Institute, Baltimore, USA4Astrophysics Division, Space Science Department of ESA, ESTEC, Noordwijk, The Netherlands5Institute of Physics and Astronomy, University of Århus, Århus, Denmark

1. Introduction

A common trait of high-redshift radiogalaxies is their extended Lyα emis-sion. This emission-line nebulosity isgenerally aligned with the radio axis,and similar emission has been reportedaround radio-loud QSOs (Schneider etal. 1987, Heckman et al. 1991a,b, Hu etal. 1991). For radio-quiet QSOs, it ap-pears that extended Lyα emission ismuch less common. Hu et al. (1991)found extended emission-line nebulosi-ty around three of ten radio-loud QSOs,but from none of the seven radio-quietQSOs in their sample. Steidel et al.(1991) and Bergeron et al. (1999) re-ported the detection of extended Lyαemission from two radio-weak QSOs,

and Bremer at al. (1992) reported theso far only detection of extended Lyαemission from a radio-quiet QSO.

Two methods have been used in allof the work quoted above, narrow-bandimaging and direct inspection of two-di-mensional (2D) CCD spectra to look forregions of emission extending awayfrom the spectrum of the QSO. A toolthat would allow modelling and subtrac-tion of the 2D spectrum of the QSOwould not only improve the chances ofdetecting faint emission lines in thevicinity of the QSO, it could also allowmore detailed analysis. Our first reporton spectral PSF (SPSF) subtraction(Møller, 2000) described a technique forSPSF subtraction. In this second report,as a simple example of an application

of the tool, we report the serendipitousdiscovery of extended Lyα emissionfrom the host galaxy of the radio quiet z= 2.559 QSO Q2206-199.

2. Observations

On August 12, 1999, we obtainedfour FORS1/Grism 600B long slit spec-tra, each of 2000 sec integration time,of the radio quiet QSO Q2206-199.Our observing strategy was to obtainseveral spectra with different offsetsalong the slit, to minimise the effectsof flat-field errors and other syste-matic effects. This is a crucial point tokeep in mind when one is in searchof extremely faint features close to, oron top of, a much brighter object.

34

The spectra were obtained with a PAof 236 deg. and the seeing from the au-tomatic seeing monitor (the DIMM) dur-ing all exposures was recorded as0.40–0.45 arcsec. After flat-fielding,shifting and co-addition of the four indi-vidual spectra, the measured combinedFWHM of the QSO SPSF along the slitis 0.60 arcsec, which is an acceptabledegradation of the SPSF as comparedto the imaging PSF.

Following the co-addition we usedthe code described in Report 1 to per-form the tracing of the spectrum, to de-termine the SPSF as a function ofwavelength, and to extract the maxi-mum signal-to-noise spectrum of theQSO. In Figure 1, we present the ex-tracted spectrum. The vertical scale istotal CCD counts per bin (the ADU con-version is 1.48 e–/count). The spectrumis clearly of very high S/N in the regionabove 4000 Angstrom (red part of thespectrum in Fig. 1), but below 4000Angstrom (blue part), the combined in-strument efficiency drops rapidly. Ourtarget DLA system (at 3549 Angstrom)is deep into the low efficiency part of thespectrum, but despite the low efficiency,we obtained a very clear detection ofLyα emission from the DLA galaxy (tobe reported in our third and last SPSFreport).

3. Application of SPSF Sub -traction: QSO Host Galaxies

The search for the host galaxies ofQSOs and the interpretation of theirspectral energy distributions has re-cently started to move from low (z =0.1–0.3) towards higher (z = 2–3) red-shifts, and has become a topic of cur-rent debate (Terlevich and Boyle 1993,Aretxaga et al. 1998, Fynbo, Burud andMøller 2000). A proposal as to how one

may detect extended line emissionacross QSO spectra was recently pre-sented by Courbin et al. 1999.Spectroscopic detection of emissionlines from the QSO host galaxies wouldbe an important step forward, as itwould clarify many of the current ques-tions concerning the high redshiftQSOs and their host galaxies.

1. Is the redshift of the host galaxythe same as that of the QSO broadLyα and CIV lines? It has long beenknown that the QSO high ionisationlines are significantly blue shifted withrespect to the low ionisation lines. Thelow ionisation lines are expected to rep-resent the true systemic redshift of theQSO (Espey et al. 1989, Corbin 1990,Møller, Warren and Fynbo 1998).Emission lines from the host galaxy

should be found at the systemic red-shift, hence they should be redshiftedfrom the broad Lyα and CIV of theQSO.

2. What is the velocity width of theextended line emission? If the emis-sion from the QSO host galaxy is in re-ality caused by simple dust reflection(see e.g. Fynbo, Burud and Møller2000) of the QSO spectrum, such ashas been reported for radio galaxies,then the host galaxy spectrum musthave both Lyα redshift and line velocitywidths identical to that of the QSO itself.If that is not the case then one can ruleout the dust-reflection hypothesis.

3. What is the dynamical mass ofthe QSO host galaxy? If the extendedline emission is not due to reflectedQSO light, then the velocity profile overthe spatial extent of the line emissionreflects the dynamical state of the emit-ting gas. It may even be possible to de-termine a rotation curve. This couldprovide limits on the mass of the darkmatter halos in which the high redshiftquasars form.

Data of the quality shown in Figure 1with high S/N at Lyα, excellent seeingand good sampling of the SPSF, clear-ly provide the best possible conditionsto address those questions. In Figure 2we show (left panel) a section of the 2DQSO spectrum centred on the Lyαemission line of the QSO. After maxi-mum S/N extraction of the 1D spectrumas described above, we subtracted theminimum χ2 fit of the 2D spectrum. Forillustration purposes, we have (centralpanel of Fig. 2) only subtracted the cen-tral part of the fitted QSO spectrum.The residuals clearly show evidence forextended Lyα emission. The rightmostpanel of Figure 2 shows the residualsafter the PSF subtraction and after sim-ple box-car smoothing to enhance thecontrast of the extended emission overthe background noise.

Figure 1: FORS1/600B raw extracted spectrum (8000 seconds) of the z = 2.559 radio quietQSO Q2206-199.

Figure 2: 2D spectrum covering the region around Lyα. Left: 2D spectrum after sky subtrac-tion. Centre: 2D spectrum after sky subtraction and subtraction of the minimum χ2 fit of thecentral part of the 2D QSO spectrum. Right: Same as the spectrum in the centre but aftersimple box-car smoothing to enhance the contrast of the extended emission over the back-ground noise. The horizontal scale (along the slit) in all three panels is 0.2 arcsec per pixel.

35

Clearly, we detect extended Lyαemission over a region of almost 10arcsec (roughly 50 h–1 kpc), and thethree questions asked above can nowbe addressed.

4. Extended L yα of a High-Redshift QSO Host Galaxy

The CCD columns very close to thecentral part of the 2D spectrum showlarge residuals at all wavelength bins.This is due to the large photon shot-noise. Faint objects cannot ever be de-tected close to the central part of theQSO SPSF because of this. To obtainthe spectrum of the extended host galaxywe therefore ignored the central noisy

columns, and averaged the regions leftand right of the residuals of the middleframe of Figure 2. This spectrum is plot-ted in red in Figure 3. Also in Figure 3,we have plotted the normalised spec-trum of the broad Lyα and NV lines ofthe QSO Q2206-199. The expected po-sition of the quasar emission lines (forzQSO = 2.559) are marked in Figure 3.The emission redshift of Q2206-199 istaken from the literature and is basedon the Lyα, SiIV, and CIV emission lines.

It is immediately clear that the broadlines of the QSO are blue shifted by1500 km/s with respect to the extendedLyα, and also that the extended Lyα ismuch narrower than the broad QSOLyα line. Those two observations con-

firm that the ex-tended Lyα is in-deed at the sys-temic redshift ofthe QSO andmust be the signa-ture of gas in thehost galaxy. It isnot caused bydust reflection ofquasar light. TheFWHM of the nar-row Lyα is 760km/s (at PA 236deg E of N) and510 km/s (at PA56 deg E of N). Itis interesting to

note that there is no hint of Lyα absorp-tion at the redshift of the host galaxy.

From visual inspection of Figure 2(right frame) it appears that the extend-ed Lyα cloud is rotating. We shall nowquantify this observation. For each CCDcolumn where the extended Lyα wasstrong enough, we fitted a gaussian tothe velocity profile. Also here, we ignoredthe noise-dominated columns close tothe central part of the QSO spectrum.The centroids of the velocity profiles, asa function of the projected distance fromthe QSO, are plotted in Figure 4. It isclear from Figure 4 that the host galaxygas is indeed rotating. With only a sin-gle slit orientation, we cannot know ifwe are on the major or minor axis of thehost galaxy or somewhere in between,and we also do not know what the incli-nation angle of the galaxy is. Hence,the apparent rotation velocity of about50–100 km/s at a projected radius ofabout 25h–1 kpc, is a lower limit, whichplaces a lower limit of a few times 1010

h–1 Solar masses on the dynamicalmass inside this radius. The most strik-ing result from Figure 4 is, however,that the rotation velocity appears tokeep growing outwards. Hence, there islikely a significant amount of mass lo-cated at even larger distances from thecentral QSO. This would support theview that QSOs at high redshifts areformed in deep potential wells.

Acknowledgements

Based on observations collected atthe European Southern Observatory,Paranal, Chile (ESO Programme63.0-0618).

References

Aretxaga I., Terlevich R.J., Boyle B.J., 1998,MNRAS 296, 643.

Bergeron, J., Petitjean, P., Cristiani, S.,Arnouts, S., Bresolin, F., Fasano, G.,1999, A&A 343, L40.

Bremer M.N., Fabian A.C., Sargent W.L.W.,Steidel C.C., Boksenberg A., JohnstoneR.M., 1992, MNRAS 258, L23.

Corbin, M.R., 1990, ApJ, 357, 346.Courbin, F., Magain, P., Sohy, S., Lidman, C.,

Meylan, G., 1999, The Messenger, 97, 26.Espey, B.R., Carswell, R.F., Bailey, J.A.,

Smith, M.G., Ward, M.J., 1989, ApJ, 342,666.

Fynbo, J.U., Burud, I., Møller, P., 2000, A&A,submitted.

Heckman T.M., Lehnert M.D., van BreugelW., Miley G.K., 1991a, ApJ 370, 78.

Heckman T.M., Lehnert M.D., Miley G.K.,van Breugel W., 1991b, ApJ 381, 373.

Hu E.M., Songaila A., Cowie L.L., StocktonA., 1991, ApJ 368, 28.

Møller, P., Warren, S.J., Fynbo, J.U., 1998,A&A, 330, 19.

Møller, P., 2000, The Messenger, 99, 31.Schneider D.P., Gunn J.E., Turner E.L.,

Lawrence C.R., Schmidt M., Burke B.F.,1987, AJ 94, 12.

Steidel C.C., Sargent W.L.W., Dickinson M.,1991, AJ 101, 1187.

Terlevich R.J., Boyle B.J., 1993, MNRAS262, 491.

Figure 3: Blue: Spectrum of the Lyα line region of the QSO Q2206-199 normalised to 1 inthe continuum. Red: Spectrum of the extended Lyα line of the host galaxy of Q2206-199.Note that the extended Lyα is much narrower than the QSO Lyα, and that the QSO emission-line redshift is blue shifted by 1500 km/s. The extended Lyα amplitude was scaled arbitrarilyfor easy comparison to the QSO spectrum.

Figure 4: Relativeprojected motion ofthe host galaxy gas.The data clearly indi-cate rotation.

36

The many deviations from known su-pernovae and the probable connectionwith the γ-ray burst triggered a substan-tial interest in SN 1998bw. It has beencalled the ‘Rosetta stone’ of γ-ray burstsand a detailed understanding of this ex-plosion could provide insights into thenature of relatively close-by GRBs (therecession velocity of the parent galaxyis only 2550 km s–1). The collapse ofthe core and subsequent explosion ofmassive stars had been proposed forGRBs before SN 1998bw (e.g. Woosley1993). Supernovae from stars whichhave lost their hydrogen and even heli-um envelope can come from eithermassive stars which shed their upperlayers in stellar winds (e.g. Wolf-Rayetstars) or binary stars which undergo acommon envelope phase in which onestar is stripped to the core. The con-nection of the γ-ray burst and the su-pernova explosion can be studied in de-tail for SN 1998bw.

The peculiarity of SN 1998bw wasrecognised almost immediately at ESO,and a dedicated follow-up programme

initiated. These data map the evolutionduring the first observing season exten-sively and will appear soon (Patat et al.2000). Here we present the data of asmaller programme that also monitoredSN 1998bw into the second year. Themain reason for the interest at theseepochs is in the possibility to directlystudy the nucleosynthesis and the hy-drodynamic properties of the explosion.

2. Observations

We followed the spectral and photo-metric evolution of SN 1998bw with the3.6-m/EFOSC2 and UT1/FORS1 from33 days until 504 days after the outburst(Sollerman et al. 2000). The decreasein brightness can be appreciated inFigure 1 where the images from the3.6-m from 33 days and from UT1 414days after explosion are shown. Thesupernova faded from V = 14.7 to V =21.7 over the year. The FORS1 imageshows the supernova also superposedon an HII region which makes the pho-tometry at late phases very difficult.

1. Introduction

Some supernovae are special. Su-pernova SN 1998bw was discoveredwhen the optical counterpart of the γ-ray burst GRB980425 was sought inthe BeppoSax error box. It became oneof the most luminous stellar explosionsever observed in the optical (Galama etal. 1998), the brightest radio supernovaat a time when most other supernovaeare still deeply enshrouded in anionised cocoon which blocks all radioemission (Kulkarni et al. 1998), and thetemporal and positional coincidencemake a connection to the γ-ray burst it-self very likely (Galama et al. 1998). ButSN 1998bw was peculiar in severalother aspects as well. The spectrumlooked different from any other knownsupernova and defied a clear classifica-tion for some time. Due to the lack ofobvious hydrogen, helium or siliconlines near maximum light, each of theseindicate a Type II, a Type lb, and a TypeIa supernova, respectively, it was final-ly assigned a Type Ic classification.

The Late Phase of SN 1998bwB. LEIBUNDGUT1, J. SOLLERMAN1,2, C. KOZMA 2, C. FRANSSON2, P. LUNDQVIST2, F. RYDE2, P. WOUDT1

1ESO, 2Stockholm University

SN 1998bw is of exceptional interest. The time and position coincidence between the explosion of the SNand the explosion of GRB 980435 makes it likely that the two phenomena are related. The SN evolution hasbeen extensively studied first with the 3.6-m/EFOSC2 during the first year, then was monitored with theVLT/FORS1 during the second year when the object became fainter than mV = 21.

We report here on the late phase of SN 1998bw.

Figure 1: SN 1998bw in May 1998 (3.6-m with EFOSC2; left panel) and June 1999 (UT1 with FORS1; right panel). The supernova is super-posed on a faint HII region.

37

The light curve of SN 1998bw showsan exponential decline in luminosityfrom about 50 to 400 days after explo-sion in V, R and I (Fig. 2). Only the lastfew points after 500 days start to showa deviation from this decline. A slightflattening of the light curve can be per-ceived at this epoch. We believe thatthe uncertainties in the backgroundsubtraction are not entirely responsiblefor this flattening and the supernovamay be entering a new phase of its evo-lution.

The spectroscopic evolution of SN1998bw is displayed in Figure 3. At ear-ly times, only very broad and not-welldefined features can be seen. The lackof sharp lines is an indication of the highvelocities in the ejecta and blending ofmany lines (e.g. Iwamoto et al. 1998).At late phases the regular features ofType Ib/c supernovae in their nebularphase, i.e. once the supernova ejectahave become optically thin, are ob-served. The familiar emission lines ofMg I] (λ 4571Å) or [Fe III], [Na I] (λλ5890, 5896Å), [O I] (λλ 6300, 6364Å),[OII] (λλ 7320, 7330Å), [CaII] (λλ 7292,7324Å), and the CaII IR triplet (λλ 8498,8542, 8662Å) are all present andchange little in shape, but fade awaycontinuously.

A slight narrowing of the emissionlines with time can be observed. Of in-terest is further the blue continuum be-low 5500Å which is likely to be due tomany blended Fe II and Fe III transi-tions. The narrow lines of Hα, Hβ,[O III], and [S II] are from the underlyingHII region.

The late-epoch spectroscopy henceconfirms that SN 1998bw indeedcomes from a massive star that has lostits envelope and we can observe theexposed core of the progenitor.

3. Interpretation

The large luminosity and the high ve-locities in the ejecta have been inter-preted as due to an extremely powerful

explosion (Iwamoto et al. 1998,Woosley et al. 1999) of carbon-oxygenstars of 6 to 13 M0 (note that this is themass at explosion and not the main-se-quence mass which is more than about40 M0) and the production of about 0.5to 0.7 M0 of 56Ni. This 56Ni mass isabout 5 to 10 times higher than what isobserved in other core-collapse super-novae. Other interpretations have pro-posed asymmetric explosions (Höflichet al. 1999) with less nickel synthesisedin the explosion. We have modelled thelight curve at late epochs and find thata large amount of nickel is indeed re-quired to sustain the observed luminos-ity. Since asymmetries play a minor roleat late phases, this measurement ar-gues for a Ni mass around 0.7 M0,within a factor of two.

The late-phase spectrum showsrather high velocities still, at least10,000 km s–1. Models based on themassive explosions proposed to fit theearly spectrum and light curve do notfare too well at these late phases.Strong macroscopic mixing has to beapplied to the models to obtain therather peaked line shapes observed.Without the mixing, the lines would ba-sically trace the mass shells in the ejec-ta and with a stratified composition,they would produce flat-topped lineshapes. But the models also predict ve-locities that are too high to fit the ob-servations.

Since the energy input into the ejec-ta is down-scattering of the γ-rays com-ing from the decay of 56Co, the daugh-ter nucleus of the original 56Ni, it de-pends directly on the column density ofmatter in the ejecta. The optical depthto γ-rays is proportional to Mej /v 2exp. Ahigher expansion velocity will thin outthe ejecta faster, produce a steeper

Figure 2: The V, R and I light curves of SN 1998bw. Our photometry (Sollerman et al. 2000)is plotted in the upper left panel. The other panels also include data from Galama et al. (1998;days 1–57), McKenzie & Schaefer (1999; days 63–187), and Patat et al. (2000; days324–417).

Figure 3: Spectral evolution of SN1998bw. For clarity, the spectra have been shifted by a con-stant factor. The wavelength scale has been corrected for the velocity of the parent galaxy(2550 km s–1).

38

slope of the light curves, and hence re-quire more nickel to power the late lightcurve. Decreasing the velocity booststhe late luminosity, and can reproducethe observed line shapes. However,these velocities were not predicted bythe models that fit the early spectrumand light curve.

4. Conclusions

The monitoring of SN 1998bwthrough its second year has providedimportant new clues on the nature ofthis object. The late spectrum verymuch resembles the ones of other TypeIb/c supernovae. Hence, SN 1998bwcan be tied to a class of objects weknow fairly well. We confirm the largenickel mass required to power the opti-cal radiation of this event, but find dis-crepancies by fitting the late spectrumwith the models which were used to in-terpret the early phases. The impliedenergies are still unique for any super-nova ever observed.

The connection of SN 1998bw toGRB980425 is still unclear, but we donot expect to see any signature of theburst at these late phases. The radioobservations have been linked to the γ-ray burst and the relativistic expansionof material (Kulkarni et al. 1998, Li &Chevalier 1999). The X-ray emissionfrom the GRB afterglow is coincidentwith the supernova as well (Pian et al.1999). SN 1998bw remains a fascinat-ing and puzzling object.

The combination of two instrumentsand telescopes to follow SN 1998bw tolate phases has been very useful. The3.6-m/EFOSC2 provided the early cov-erage and only at the end was the su-pernova ‘handed’ to UT1/FORS1. Wewere therefore able to secure a spec-trum about every month and could fol-low this object further than would havebeen possible with the 3.6-m.

It is unlikely that there will be a thirdobserving season for SN 1998bw.Unless the luminosity becomes con-stant, it will be too faint to be recovered.

There are known processes that couldlead to such a constant flux, e.g. inter-action with the circumstellar material orinput from the accretion on a black hole.We have been following several suchobjects already and SN 1998bw will beworth at least a check in the next fewmonths.

References

Galama, T.J., et al. 1998, Nature, 395, 670.Höflich, P., Wheeler, J. C., & Wang, L. 1999,

ApJ, 521, 179.Iwamoto, K., et al. 1998, Nature, 395, 672.Kulkarni, S.R., et al. 1998, Nature, 395,

663.McKenzie, E. H. & Schaefer, B.E. 1999,

PASP, 111, 964.Li, Z.-Y. & Chevalier, R.A. 1999, ApJ, 526,

716.Patat, F., et al. 2000, in preparation.Pian, E., et al. 1999, A&ASS, 138, 463.Sollerman, J., et al. 2000, ApJ, submitted.Woosley, S.E. 1993, ApJ, 405, 273.Woosley, S. E., Eastman, R.G., & Schmidt,

B.P. 1999, ApJ, 516, 788.

MISTRAL: Myopic Deconvolution Method Applied toADONIS and to Simulated VL T-NAOS ImagesJean-Marc Conan1, Thierry Fusco1, Laurent M. Mugnier1, Franck Marchis2

1ONERA, Département d’Optique Théorique et Appliquée, France (name@onera.fr); 2ESO, Santiago, Chile (fmarchis@eso.org)

1. Introduction

The performance of high-resolutionimaging with large astronomical tele-scopes is severely limited by the at-mospheric turbulence. Adaptive optics[1, 2, 3] (AO) offers a real-time com-pensation of the turbulence. The cor-rection is however only partial [2, 4, 5,6, 7] and the long-exposure imagesmust be deconvolved to restore the finedetails of the object.

Great care must be taken in the de-convolution process if one wants to ob-tain a reliable restoration with goodphotometric precision. Two aspects addto the difficulty: the fact that the residualpoint spread function (PSF) is usuallynot perfectly known [8, 9, 10], and thefact that astronomical objects are usu-ally a mix of sharp structures and smoothareas. “MISTRAL” (Myopic IterativeSTep Preserving ALgorithm) [11, 8] hasbeen developed to account for these twopoints. It is based on a rigorous Bay-esian approach which allows us to eas-ily account for the noise in the image,the imprecise knowledge of the PSF,and the available a priori information onthe object (spatial structure, positivi-ty…). A specific edge preserving objectprior is proposed, which is in particularwell adapted for planetary-like objects.

The notion of AO partial correction isfirst discussed in Section 2. The princi-ple of our deconvolution technique isbriefly summarised in Section 3. In Sec-tion 4, the photometric accuracy of MIS-TRAL is first demonstrated on simulat-ed AO images. The simulation parame-ters correspond to NAOS, the AO sys-tem of the VLT. MISTRAL is then ap-plied to ADONIS images of Io taken atthermal wavelengths using the COMICcamera. This allows an accurate map-ping of Io’s surface volcanic activity. Wealso used our deconvolution method onbroadband filter (J, H, K) images of Ura-nus taken with SHARPII+. The structuresof the rings and its innermost satelliteshave been successfully detected.

2. Partially Corrected AO Images

Within the isoplanatic angle, the in-tensity i(r) at the focal plane of the sys-tem consisting of the atmosphere, ofthe telescope and of the AO bench isgiven by:

i(r) = h(r)∗ o(r) + n(r), (1)

where r is the spatial coordinate, o(r) isthe observed object, h(r) is the systemPSF and n(r) is an additive zero meannoise.

We consider here the case of AOcorrected long exposure images. Suchan image is presented in Figure 1. Inthis numerical simulation, we consid-ered an 11th magnitude planetary-likeobject observed in the visible with theNAOS AO system [12] installed on theVLT. This system will provide high per-formance in the near IR (SR P 70% athigh flux). Here we consider the case ofobservations at visible wavelength (λ =0.5 µm). In such conditions, the imageblur is very severe, the expected SR isonly 2.1% for a 0.73 arcsec seeing.Neither the fine structures on the sur-face of the object, nor the stars in thebackground are apparent in the correct-ed image. A deconvolution is thereforerequired.

The deconvolution procedure needsa measurement of the PSF. The usualprocedure consists in recording the cor-rected image of a nearby unresolvedstar shortly after observing the object ofinterest. Since the correction quality de-pends on the observing conditions (tur-bulence strength, magnitude of thesource used for wavefront sensing), theunresolved star image is not a perfectmeasurement of the PSF associatedwith the image to be deconvolved [8,11, 13, 14]. Actually, the main source ofPSF variability is the seeing fluctuation.

39

The sensibility of the Optical TransferFunction (OTF) to the seeing variationsis illustrated in Figure 2. The OTFs areestimated for a seeing ranging from0.65 to 0.93 arcsec. The correspondingSRs go from 3.8% down to 0.4%. Thestructure of the OTFs is typical ofan AO corrected OTF [4, 5, 6,7]: a low-frequency lobe and a high-frequencywing going up to the telescope cut-offfrequency. The spatial frequencies be-tween ro /λ and D/λ, which would belost without correction, are now pre-served. The high-frequency level ishowever low and very dependent of theseeing conditions. Here it changes by afactor of ten for rather realistic seeingchanges.

The OTFs presented here derivefrom a careful study of the system per-formance [12, 15]. The correspondingPSF at 0.73 arcsec seeing is shown inFigure 3. Note that despite the very lowSR in the visible, a coherent peak is stillclearly seen above the broad halo. Thisis characteristic of high-order correctionsystems such as the VLT-NAOS oneworking in the visible: the residualphase variance is large due to the shortwavelength, but the phase is mainlyconstituted of high-order modes, hencethe particular PSF profile. We will showin Section 4 that MISTRAL can restorehigh-resolution maps out of these lowSR visible VLT-NAOS images. Note thatthis suggests that the SR is not a goodmeasurement of the image quality,when quality means “restorability” [7].

Of course, the deconvolution canalso be applied to IR images with morereasonable SRs as shown on ADONISdata in Section 5. But we first recall thedeconvolution approach in the followingsection.

3. Deconvolution Approach

Most deconvolution techniques boildown to the minimisation (or maximisa-

tion) of a criterion. The first issue is thedefinition of a suitable criterion for thegiven inverse problem. The criteria pre-sented here will be derived from a prob-abilistic approach. The second problemis then to find the position of the criteri-on’s global minimum which is defined asthe solution. In some cases, it is givenby an analytical expression, but most ofthe time one has to use an iterative nu-merical method to solve the problem.

In the following sections, we first con-sider the case of an assumedly knownPSF, so-called “classical” deconvolu-tion; the method is then extended to thejoint estimation of the object and thePSF, called here “myopic” deconvolution.

3.1 Deconvolution with known PSF

Following a probabilistic approach,called maximum a posteriori [MAP] [16],

the deconvolution problem can be stat-ed as follows: we look for the most like-ly object o given the observed image i.This reads:

ômap = argomax p(o|i) =

arg omax p(i|o) × p(o) =

arg omin[Jn(o) + Jo(o)]. (2)

The criterion to be minimised, J = Jn+ Jo, is composed of a first term (Jn =–ln p(i o)) accounting for the noise sta-tistics in the image, plus a second term(Jo = –ln p(o)) which allows to use the apriori knowledge we have on the object.This last term, of course, is a function ofthe type of object being observed. Thechoice of Jo for planetary-type objects isdiscussed in the Section 3.1.2.

If no prior knowledge is available,one can still use the previous equationswith p(o) = 1. One therefore only max-imises p(i o), also called likelihood ofthe data, to obtain a maximum-likeli-hood solution. In this case the criterionis only constituted of the term Jn.

3.1.1 Maximum likelihood with photon noise

If the image is corrupted solely byphoton noise, the maximum likelihood[ML] solution minimises the followingcriterion, directly derived from the ex-pression of Poisson statistics:

Jpn

oisson(o) = –ln p(i o) =

∑r

(h∗ o)(r) – i(r)ln[(h∗ o)(r)]. (3)

The Richardson-Lucy algorithm [RL][17, 18] is an iterative numericalmethod which converges towards theglobal minimum of Jp

noisson.

Figure 1: Planetary-type object (mv = 11) and its simulated VLT-NAOS image at 0.5 µm andfor a 0.73 arcsec seeing. The Strehl Ratio (SR) is 2.1%. The number of detected photons inthe image is 108 photo – e–, the background noise has a 31 e– standard deviation. The Fieldof View is 0.8 arcsec, 128 × 128 pixel image. The planetary disk is constituted of a uniformlevel plus a broad feature 10% brighter, and three small spots, 30% brighter. Four stars areadded in the field with a 2.5 magnitude difference between the brightest and the faintest. Thefaintest corresponds to 40,000 detected photons, which corresponds to a maximum contri-bution of 155 photons/pixel in the image. The true object top-right display gives a log-scalerepresentation.

Figure 2: The OTFs are estimated for the following seeing values: from top to bottom, 0.65,0.73, 0.79, 0.85 and 0.93 arcsec. The corresponding SRs are respectively: 3.8%, 2.1%, 1.3%,0.8% and 0.4%. The mean OTF is also drawn (dashed line).

40

TRAL has the ability to estimate boththe object and PSF from the image andsome imprecise PSF measurement.The Eqs. 2 and 4 can indeed be gener-alised, in the same probabilistic frame-work, to the case of a joint estimation of[o, h]. One obtains:

[ô, h] = argo,hmax p(o, h i) =

argo,hmax p(io, h ) × p(o) × p(h) =

argo,hmin [Jn(o, h) + Jo(o) + Jh(h)].

(6)

The myopic criterion is given by Eqs.4 and 5, now a function of o and of h,plus an additional term Jh = –ln p(h)which accounts for the knowledge, al-though partial, available on the PSF.Assuming stationary Gaussian statis-tics for the PSF, Jh reads:

(7)

where h~

m = E [h~] is the mean OTF, and

PSDh = E [ h~

(ƒ) – h~

m(ƒ) 2] is the asso-ciated spatial Power Spectral Density[PSD].

Such a regularisation obviously en-sures that the actual OTF is close to themean OTF with respect to error barsgiven by the PSD, which characterisesthe fluctuations around the mean. Inpractice, the mean PSF and the PSDare estimated by replacing, in their def-initions, the expected values (E[·]) byan average on the different imagesrecorded on the unresolved star. Ideallyone would want to estimate the PSFfrom the wavefront sensing data [10,25] which would avoid the errors due toseeing fluctuations. But even in thiscase, the myopic approach can be in-teresting to account for the PSF uncer-tainties due to constant aberration cali-

bration errors [9] or to the wavefrontsensing noise for faint stars.

Note that the new criterion is convexin o for a given h, convex in h for a giv-en o but it is not convex on the wholeparameter space. However, it is possi-ble to use starting points that are closeto the global minimum, and we did notencounter minimisation problems withthe conjugate gradient method. A posi-tivity constraint is also used on the PSF(h = b2).

4. Deconvolution of SimulatedVLT-NAOS Images

The application of MISTRAL to thesimulated VLT-NAOS visible image pre-sented in Section 2 is discussed here.

Figure 4 shows the results obtainedin the ideal case of a classical decon-volution using the true PSF. The decon-volution obtained with MISTRAL at con-vergence (360 iterations) of the minimi-sation of Eq. 4 is compared to the RLestimation stopped respectively at1000, 6830 and 50,000 iterations. Ineach case a log-scale version of the re-stored object is shown in order to checkthe detection of the faint stars in thebackground.

A quantitative measurement of therestoration quality can be given in termsof distance to the true object, rms valueexpressed in photons/pixel and com-puted on the full field of view. The dis-tance is 5150 photons/pixel for theMISTRAL estimate, and respectively11,800, 9900 and 13,900 photons/pixelfor the RL cases. The evolution of thedistance with the computation time isgiven in Figure 5. The estimate ob-tained with 6830 iterations and shownin Figure 4 therefore corresponds to thebest RL estimate. The starting point isalways the image thresholded to aslightly positive level for implementation

Figure 3: VLT-NAOS PSF at 0.5 µm, for 0.73 arcsec seeing and a guide star magnitudemV = 11. The PSF is normalised to its Strehl Ratio SR P 2.1%.

It is however well known that therestoration of the object using the soledata is an unstable process (see in par-ticular Refs. [19] and [16] for reviews).The noise is highly amplified in the so-lution. Of course, one can stop the RLbefore convergence to limit the noiseamplification but in this case the solu-tion is poorly defined. For sure, it is nomore the minimum of any criterion. Abetter solution is to regularise the prob-lem with an adequate object prior asproposed in Section 3.1.2.

3.1.2 Edge preserving regularised deconvolution

In the MISTRAL algorithm, we con-sider a white non-stationary Gaussiannoise in the image, which is a good ap-proximation of a mix of photon andbackground (detector or sky-back-ground) noise. Furthermore, the decon-volution is regularised by an object pri-or particularly adapted for planetary-likeobjects. This prior avoids the usual ring-ing artefacts [20] given by standard de-convolution techniques on such sharpedge objects [21, 8]. The criterion to beminimised is:

J(o) = Jn(o) + Jo(o) =(4)

– (o∗ h)(r))2 + Jo(o),

where σ 2(r) is the image thresholded tothe background noise variance. In theabsence of background noise, this ex-pression of Jn is actually a first-orderdevelopment of Eq. 3. The regularisa-tion term is defined as:

(5)

where ∆ο(r) = √∆xo(r)2 + ∆yo(r)2), ∆xoand ∆yo are the object finite differencegradients along x and y respectively.

This regularisation, called L1 – L2, isan isotropic version of the expressionsuggested by Brette [22]. The globalfactor µ and the threshold d have to beadjusted according to the noise leveland the structure of the object. This iscurrently done by hand but an automat-ic procedure is under study.

We use a fast conjugate gradientmethod [23] to minimise the global cri-terion J given in Eq. 4. This method iswell adapted since the so-defined crite-rion is convex. An additional positivityconstraint is used in MISTRAL, it is en-forced with a reparameterisationmethod (o = a2) [24] where a are thenew parameters used in the minimisa-tion.

3.2 Myopic deconvolutionAs mentioned in Section 2, the true

residual PSF is seldom available. MIS-

41

of the positivity. The RL estimate neverreaches the distance obtained withMISTRAL and diverges for a large num-ber of iterations. It goes from a low-resolution estimate with ringing arte-facts to a very noisy one. The restora-tion quality is obviously much higherwith our regularised algorithm. Theglobal photometry is very precisely re-stored. The stars in the backgroundare well detected and only slightlysmoothed. This large dynamic is per-mitted by a good model of the imagenoise. Note also that MISTRAL reachesconvergence in a quite reasonableamount of time (P 390 s and 360 it-erations), roughly the same time re-quired for RL to reach its best estima-

tion (P 330 s and 6800 iterations). Notealso that MISTRAL is able to both re-store the edge of the object and thestructures on the surface.

We then consider the case of a poorestimation of the PSF. We recall thatthe image was obtained with a PSF cor-responding to a 0.73 arcsec seeing. Weassume that 5 images of an unresolvedstar were recorded shortly before or af-ter. The seeing is supposed to be un-stable and the seeing is actually 0.65,0.73, 0.79, 0.85 and 0.93 arcsec re-spectively for each of these calibrationimages. The OTF estimates are shownin Figure 2 as well as the mean OTF.Since it can be difficult to estimate pre-cisely the seeing conditions to select

the correct OTF, one may want to use aclassical deconvolution assuming thatthe true OTF is the mean OTF. The re-sult obtained with MISTRAL with thisassumption is shown in Figure 6.Despite the fact that the mean OTF isclose to the true one (SR = 1.7% in-stead of 2.1%), the restoration is poor:artefacts on the surface, apparent di-ameter underestimated, no detection ofthe surrounding stars. The distance tothe true object is large (P 23,000 pho-tons/pixel). The other approach is touse MISTRAL in the myopic mode (min-imisation at convergence of Eq. 6) withthe same mean OTF and a PSD whichis nothing but the variance estimatedout of these 5 OTF measurements foreach spatial frequency. The restorationis very similar to that obtained with thetrue PSF. The distance to the true ob-ject is P 6800 photons/pixel. Note how-ever that part of the dynamic is lost.Only the two brightest stars are detect-ed. The computation time required inthe myopic case is P 1900 s (1600 iter-ations) which is still quite reasonable.

5. Deconvolution ofExperimental ADONIS Data

ADONIS, the AO system mounted onthe 3.6-m telescope of the La Silla ob-servatory, has been routinely used bythe ESO community since 1993. Thewavefront distortions of the visible in-coming light are measured using one ofthe two Shack-Hartmann wavefrontsensors (WFS). The 52-actuators de-formable mirror and tip-tilt mirror controlis performed by a modal optimisation.ADONIS is the only AO system provid-ing an imaging facility in the 1–5 µmrange via two NIR cameras. The plane-

Figure 4: Classical deconvolution with MISTRAL and RL estimates with 1000, 6830 and 50,000 iterations. The PSF is the true PSF. The de-convolution is performed on the 0.8 arcsec field of view but only a 0.4 arcsec subfield is displayed. In each case we show the correspondinglog-scale map (right-hand side panel in each pair of images). The distance to the true object is 5150 photons/pixel with MISTRAL and re-spectively 11,800, 9900 and 13,900 photons/pixel for the RL cases.

Figure 5: Distance to the true object, rms value expressed in photons/pixel, versus computa-tion time: MISTRAL (solid line), RL (dashed line).

Mistral Classic Lucy 1000 it

Lucy 6830 it Lucy 50,000 it

42

tary observations presented belowhave been performed without prefocaloptics using broad-band filters.

5.1 Monitoring of Io’s volcanism

Io’s volcanic activity, attributed to in-ternal heating due to tidal effects be-tween Jupiter and Io, was first discov-ered from space with Voyager 1 and 2in 1979. They have shown the pres-ence of active volcanic centres, calledhot spots, detectable by their IR emis-sion. Since then, and because of the in-crease in IR detector sensitivity, Io’svariable volcanos of the Jupiter-facinghemisphere have been studied byground-based observations when Io islocated in the shadow of Jupiter [27].Since October 1996, ADONIS AO sys-tem coupled with its COMIC thermalcamera (CEA/LIR/LETI detector, 128 ×128, 100 mas/pixel) has been used tomonitor Io’s volcanic activity [28].Observations performed with a L’broad-band filter (λc = 3.8 µm) using thesatellite itself as reference (angular sizep 1.0 arcsec, mv p 5) allow a completemapping of its surface. In this spectralrange, the AO correction is quite effi-cient and the typical SR obtained isaround 45% with a 0.8 arcsec seeing.

An AO corrected image and the cor-responding PSF are presented in Fig-ure 7. Only the bright Loki hot spot isdetected on the AO corrected image.A deconvolution is required to studyother structures. In addition to theresidual blur, the image incorporates arather uniform high-level backgroundemission produced by the sky and thebench optical elements [29] and avariable and inhomogeneous back-ground modulated by the AO correction[30]. A good background subtraction isof course important for the deconvo-lution.

We present in Figure 8 two con-secutive images of Io’s Jupiter-facinghemisphere taken in September 1998and processed with MISTRAL [31].We used the myopic mode to accountfor seeing variations. Loki, well-knownhot spot located on the Jupiter-facinghemisphere, is quite active and sur-rounded by secondary outbursts.Standard deconvolution processescannot be applied on these data.Indeed, in the absence of edge-pre-serving regularisation term, the wholeenergy of the disk tends to be con-centrated in the bright Loki feature.Our observations are well correlated to,and complement those, made by

Galileo/NIMS spectrocamera [26] (seeFig. 8). The coupling of AO system witha thermal camera and the use of a spe-cific deconvolution process for plane-tary objects, such as MISTRAL, is verypromising for our understanding of Io’svolcanism which can only be accom-plished by a frequent monitoring of itsactivity.

5.2 Study of the Uranian system

Uranus has been observed withbroad-band filters (J,H,K) using theSHARPII+ camera (Rockwell NIC-MOS3 detector, 1–2.5 µm, 100 mas/pixel) on May 2, 1999. These observa-tions have been performed thanks tothe capability of the Shack-HartmannWFS to analyse the wavefront on anextended object. Because of the rela-tively small angular size of the planet(2.6 arcsec) and the excellent seeingcondition (0.5 arcsec and very stable),the correction quality was high and wegot a SR of 48% in K band (2.2 µm).The centroid position error, in eachsub-pupil, is however higher on extend-ed objects, hence a degraded AO per-formance on such extended objects.The image of an unresolved star istherefore a biased measurement of the

MISTRAL with mean PSF Myopic MISTRAL

Figure 6: MISTRAL classical deconvolution with mean PSF considered as the true PSF and myopic deconvolution. In each case we show thecorresponding log-scale map. The deconvolution is performed on the 0.8 arcsec field of view but only a 0.4 arcsec sub-field is displayed. Thedistance to the true object is respectively 23,000 and 6800 photons/pixel (see Section 4).

Figure 7: ADONIS image of Io and a log-scale representation of the corresponding PSF. Only the bright Loki hot spot is detected on the Io’sdisk.

43

PSF and a myopic deconvolutionmethod is definitely necessary to re-store the initial sharpness of the im-ages. Figure 9 displays a set of data af-ter deconvolution with MISTRAL. In J(1.2 µm) and H (1.6 µm) bands, theplanetary disk is not uniform as ob-served in the visible and shows brightpolar haze distributed along a latitude.In K band, due to the methane atmos-pheric absorption band, the planet isdark and the brightest feature is theEpsilon ring with its longitudinal anom-aly. Looking at the low intensity levels in

J and H band (Fig. 10), one can see theEpsilon ring and also some innermostones. The exterior satellites Ariel,Miranda and Puck can be detected withstandard deconvolution such as RL.But the myopic deconvolution processalso reveals the presence of the inner-most satellites, Portia, Rosalind, Biancaand Juliet (see Fig. 10) which havebeen discovered by Voyager 2 in 1986and never re-observed since then. Afterthese first successful observations, themonitoring of Uranus and its environ-ment will continue. The AO system fa-

cility and the accurate MISTRAL de-convolution method will allow us tostudy the atmospheric activity of theplanets, the colour and composition ofthe rings. The comparison of the posi-tions of the faintest satellites withephemerides will also better constrainphysical parameters (body masses,flattening factor . . .) and to elaborate amore accurate analytical theory of thesatellite motion [32].

6. Conclusion

MISTRAL is a myopic deconvolutionalgorithm particularly adapted for AOcorrected images of astronomical ob-jects. It accounts for the noise in the im-age, for the presence of sharp struc-tures in the object and for the fact thatthe PSF is usually not perfectly known.Its ability to provide high photometricprecision estimates with a quite reason-able computation time has been illus-trated on simulated data. The simula-tion conditions correspond to a VLT-NAOS observation in the visible. Sincethe system is optimised in the near IR,the correction quality in the visible islow (SR P 2% here). Even in such se-vere conditions, a diffraction-limitedrestoration is obtained.

MISTRAL has also been applied toADONIS images of Io in the thermal do-main, and images of Uranus in the nearIR. In addition to the bright hot spotLoki, secondary outbursts were ob-served on Io’s Jupiter-facing hemi-sphere. Such observations are verypromising for our understanding of Io’svolcanism which can only be accom-plished by frequent monitoring of its ac-tivity. Concerning Uranus, the struc-tures of the rings and its innermostsatellites have been successfully de-tected. We will continue a monitoring ofUranus and its environment. Solar-sys-tem studies (atmospheric activity of theplanets, colour and composition of therings, position of the faint satellites) re-quire high-resolution and high-photo-metric-precision data. These can be ob-

Figure 8: Two consecutive images of the Jupiter-facing hemisphere of Io observed with ADO-NIS/COMIC in L’ band and processed with MISTRAL. The green boxes correspond to the pro-jection of the hot spots detected by Galileo/NIMS during the first four orbits [26]. The rightpanel indicates the geometry of Io at the date of the observation and the name of the knownhot spots (courtesy Bureau des Longitudes). See Section 5.

Figure 9: Uranus observed in May 1999 (north is down, east is left) after deconvolution by MISTRAL. In J and H bands, the hazy atmosphericregions are clearly visible around the pole. In K band, the methane atmospheric band absorbs the solar light and the bright feature observedis the Epsilon ring and its longitudinal anomaly.

44

[17] W.H. Richardson, “Bayesian-based iter-ative method of image restoration”, J.Opt. Soc. Am. 62, 55–59 (1972).

[18] L. B. Lucy, “An iterative technique forrectification of observed distributions”,Ap.J. 79, 745–754 (1974).

[19] D. M. Titterington, “General structureof regularization procedures in imagereconstruction”, A&A 144, 381–387(1985).

[20] J.-P. Véran, “Estimation de la réponseimpulsionnelle et restauration d’image enoptique adaptative – Application au sys-tème d’optique adaptative du TélescopeCanada-France-Hawaii”, Ph.D. thesis,Ecole Nationale Supérieure desTélécommunications, 1997.

[21] L. M. Mugnier, J.-M. Conan, T. Fusco,and V. Michau, “Joint Maximum aPosteriori Estimation of Object and PSFfor Turbulence Degraded Images”, inBayesian Inference for Inverse problems,3459, 50–61 (San Diego, CA (USA),1998).

[22] S. Brette and J. Idier, “Optimized SingleSite Update Algorithms for ImageDeblurring”, in Proceedings of theInternational Conference on ImageProcessing, p. 65–68 (Lausanne, Switz-erland, 1996).

[23] Groupe Problèmes Inverses, “GPAV:une grande œuvre collective”, internal re-port, Laboratoire des Signaux etSystèmes, CNRS/Supélec/UniversitéParis-Sud (1997) .

[24] E. Thiébaut and J.-M. Conan, “Strict apriori constraints for maximum-likelihoodblind deconvolution”, J. Opt. Soc. Am. A12, 485–492 (1995).

[25] S. Harder, “Reconstruction de laréponse impulsionnelle du système d’op-tique adaptative ADONIS a Dartir desmesures de son analyseur de surfaced’onde et Étude photométrique de la vari-abilité des étoiles YY Orionis”, Ph.D. the-sis, Université Joseph Fourier, Grenoble1, 1999.

[26] R. Lopes-Gautier, A. Davies, R. Carlson,W. Smythe, L. Kamp, L. Soderblom, F. E.

tained with large telescope AO obser-vations in conjunction with a high-preci-sion deconvolution technique.

7. Acknowledgements

The authors thank Gérard Rousset,Vincent Michau, Claude and FrançoisRoddier, Jérome Idier and Guy LeBesnerais for many fruitful discussions.We are indebted to ESO-3.6-m supportteam, in particular the Telescope Op-erators for their assistance during thesedifficult and unusual planetary objectobservations.

References

[1] J.W. Hardy, J. E. Lefevbre, and C.L.Koliopoulos, “Real time atmosphericcompensation”, J. Opt. Soc. Am. 67,360–369 (1977).

[2] G. Rousset, J.-C. Fontanella, P. Kern, P.Gigan, F. Rigaut, P. Léna, C. Boyer, P.Jagourel, J.-P. Gaffard, and F. Merkle,“First diffraction-limited astronomical im-ages with adaptive optics”, A&A 230,29–32 (1990).

[3] Adaptive Optics in Astronomy, F. Rod-dier, ed., (Cambridge University Press,1999).

[4] F. Rigaut, G. Rousset, P. Kern, J.-C.Fontanella, J.-P. Gaffard, F. Merkle, andP. Léna, “Adaptive optics on a 3.6-m tel-escope: results and performance”, A&A250, 280–290 (1991).

[5] M.C. Roggemann and C.L. Matson,“Power spectrum and Fourier phasespectrum estimation by using fully andpartially compensating adaptive opticsand bispectrum postprocessing”, J. Opt.Soc. Am. A 9, 1525–1535 (1992).

[6] J.M. Conan, P.Y. Madec, and G.Rousset, “Image formation in adaptiveoptics partial correction”, in Active andAdaptive Optics, (Garching, Germany,1993).

[7] J.-M. Conan, “Étude de la correctionpartielle en optique adaptative”, Ph.D.thesis, Université Paris XI Orsay. 1994.

[8] J.-M. Conan, T. Fusco, L. Mugnier, E.Kersalé, and V. Michau, “Deconvolutionof adaptive optics images with impreciseknowledge of the point spread fonction:results on astronomical objects”, in As-tronomy with adaptive optics: present re-sults and future programs, (Sonthofen,1998).

[9] T. Fusco, J.-P. Véran, J.-M. Conan, andL. Mugnier, “Myopic deconvolution meth-od for adaptive optics images of stellarfields”, A&A Suppl. Ser. 134,1–10 (1999).

[10] J.-P. Véran, F. Rigaut, H. Maître, and D.Rouan, “Estimation of the adaptive opticslong exposure point spread function us-ing control loop data”, J. Opt. Soc. Am. A14, 3057–3069 (1997).

[11] J.-M. Conan, L. M. Mugnier, T. Fusco, V.Michau, and G. Rousset, “Myopic Decon-volution of Adaptive Optics Images usingObject and Point Spread Function PowerSpectra”, Appl. Opt. 37, 4614–4622(1998).

[12] G. Rousset et al., “Design of the Nas-myth Adaptive Optics System (NAOS) ofthe VLT”, in Astronomical Telescopes &Instrumentation, D. Bonaccini and R. K.Tyson, eds., 3353 (Kona, Hawaii, 1998).

[13] J.C. Christou, “Deconvolution of adap-tive optics images”, in ESO/OSA topicalmeeting on astronomy with adaptive op-tics present results and future programs,(1998).

[14] E. Tessier, “Image quality with currentadaptive optics instruments”, A&A Suppl.Ser. 125, 581–593 (1997).

[15] J.-M. Conan, D. Mouillet, E. Gendron,and G. Rousset, “NAOS analysis reportIssue 2.0”, Rapport Technique VLT-TRE-NAO-11650-1-000000-0002, ONERA(1999), contract ESO.

[16] G. Demoment, “Image Reconstructionand Restoration: Overview of CommonEstimation Structures and Problems”,IEEE Trans. Acoust. Speech SignalProcess. 37, 2024–2036 (1989).

Figure 10: Logarithmic display of the MISTRAL images (north is up and east is left) showing the lowest intensity levels in the J and H images.Innermost rings and faintest satellites (first observed with Voyager in 1986) are also detected.

45

Leader, and R. Mehlman, “Hot spots onIo: Initial results from Galileo’s near in-frared mapping spectrometer”, Geophys.Res. Lett. 24, 2439 (1997).

[27] J. R. Spencer, J. A. Stansberry, C.Dumas, D. Vakil, R. Pregler, M. Hicks,and K. Hege, “History of high-tempera-ture Io volcanism: February 1995 to May1997”, Geophys. Res. Lett. 24, 2451(1997).

[28] F. Marchis, R. Prangé, and J.C.Christou, “Adaptive Optics mapping of

Io’s volcanism in the thermal IR”, Icarus(1999), submitted.

[29] D. Le Mignant, E. Gendron, and F.Marchis, “The ESO ADONIS adaptiveoptics system: Study of the 3-5µm back-ground emissivity”, in ESO conferenceand workshop proceedings 56, 575(Garching, Germany, 1998).

[30] H. Geoffray, “Thermal background blur-ring on ADONIS system”, in ESO confer-ence and workshop proceedings 56, 531(Garching, Germany, 1998).

[31] R. Prangé, F. Marchis, and T. Fusco,“Mapping and monitoring of Io’s hot spotsby use of the ESO Adaptive Optics sys-tem.”, in 31st Annual Meeting of the DPS,(Padova, Italy, 1999).

[32] F. Marchis, J. Berthier, P. Descamps, T.Fusco, R. Prangé, and T. Sekiguchi,“Ground-based high resolution observa-tions of the Uranian system in the nearIR”, in 31st Annual Meeting of the DPS,(Padova, Italy, 1999).

A New Look at the Sombrero Galaxy

This image of Messier 104, also known as Sombrero Galaxy, because of its particular shape, was obtained with FORS1 at VLT Antu onJanuary 30, 2000. The colour image was made by a combination of three CCD images obtained by Peter Barthel from the Kapteyn Instituteat Groningen, Netherlands. He and Mark Neeser, also from the Kapteyn Institute, produced the composite images.The “Sombrero” is notable for its dominant nuclear bulge, composed primarily of old stars, and its nearly edge-on disk composed of stars,gas and intricately structured dust. The complexity of this dust and the high resolution of this image are most apparent directly in front of thebright nucleus, but it is also evident as dark absorbing lanes throughout the disk.

46

Physics is everywhere. The laws of physics govern theUniverse, the Sun, the Earth and even our own lives. In today’srapidly developing society, we are becoming increasingly de-pendent on high technology. Computers, transport, and commu-nication are just some of the key areas that are the result of dis-coveries by scientists working in physics.

But how much do the citizens of Europe really know aboutphysics? Here is a unique opportunity to learn more about thiselusive subject!

Beginning in February 2000, three major European researchorganisations are organising a unique Europe-wide programmeto raise the public awareness of physics and related sciences.

“Physics on Stage” is launched by the European Laboratoryfor Particle Physics (CERN), the European Space Agency (ESA)and the European Southern Observatory (ESO), with supportfrom the European Union. Other partners are the EuropeanPhysical Society (EPS) and the European Association forAstronomy Education (EAAE).

This exciting programme is part of the European Week forScience and Technology and will culminate in a Science Festivalduring November 6–11, 2000, on the CERN premises at theFrench-Swiss border near Geneva.

Why “Physics on Stage”?

The primary goal of “Physics on Stage” is to counteract thecurrent decline in interest and knowledge about physics amongEurope’s citizens by means of a series of highly visible promo-tional activities. It will bring together leading scientists and edu-cators, government bodies and the media, to confront the dimin-ishing attraction of physics to young people and to developstrategies to reverse this trend.

The objective in the short term is to infuse excitement and toprovide new educational materials. In the longer term, “Physicson Stage” will generate new developments by enabling expertsthroughout Europe to meet, exchange and innovate.

“Physics on Stage” in 22 European Countries

“Physics on Stage” has been initiated in 22 European coun-tries1. In each of these, a dedicated National Steering Committeeis being formed which will be responsible for its own national pro-gramme.

“Physics on Stage” is based on a series of high-profilephysics-related activities that will inform the European public ingeneral and European high-school physics teachers and mediarepresentatives in particular about innovative ways to convey in-formation about physics. It will stress the intimate connection ofthis natural science with our daily lives. It will be accompanied bya broad media debate on these subjects.

What will happen during “Physics on Stage”?

During the first phase of “Physics on Stage”, from now untilOctober 2000, the individual National Steering Committees(NSCs) will survey the situation in their respective countries. TheNSCs will collaborate with national media to identify new and ex-citing educational approaches to physics. These may involvedemonstrations, interactive experiments, video and CD-Rompresentations, Web applications, virtual reality, theatre perform-ances, etc.

Nationally run competitions will select some of the best andmost convincing new ideas for presentations and educationalmaterials which will receive development support from “Physicson Stage”.

The project will culminate in November 2000, with approxi-

ANNOUNCEMENTS

CERN, ESA and ESO Launch “Physics On Stage” (Taken from ESO Press Release 04/00)

mately 400 delegates converging on CERN, in Geneva, for thePhysics on Stage Festival. During this event, the national com-petion winners, science teachers, science communicators, pub-lishers, top scientists and high-level representatives of the min-istries and European organisations will brainstorm future solu-tions to bolster physics’ popularity. The programme will also in-clude spectacular demonstrations of new educational tools; thebest will be disseminated over the national TV networks and oth-er media to the European public.

Why CERN, ESA and ESO?

As Europe’s principal organisations in physics research (par-ticle physics, space and astronomy), the three recognised theirmutual responsibility to address the issue through the creation ofa new initiative and the creative use of their own research to at-tract the public and teachers alike.

About the “European Science and T echnologyWeek”

The objective of theEuropean Science andTechnology Week is to im-prove the public’s knowl-edge and understanding ofscience and technology – in-cluding the associated ben-efits for society as a whole.The Week focuses on theEuropean dimension of re-search, such as pan-Eu-ropean scientific and tech-nological co-operation.

The rationale for holdingthe Week has its roots in theimportance of the role of sci-ence and technology inmodern societies and theneed, therefore, to ensure that the public recognises its signifi-cance in our lives.

The Week is a framework for special TV programmes, exhibi-tions, contests, conferences, electronic networking, and otherscience-related activities to promote the public understanding ofscience and technology.

The Week was launched in 1993, on the initiative of theEuropean Commission. Raising public awareness of science andtechnology is now the subject of a clearly defined action withinthe Human Potential Programme of the Fifth FrameworkProgramme.

1The 22 countries are the member countries of at least one of the partici-pating organisations or the European Union: Austria, Belgium, Bulgaria, theCzech Republic, Denmark, Finland, France, Germany, Greece, Hungary,Ireland, Italy, Luxembourg, the Netherlands, Norway, Poland, Portugal, theSlovak Republic, Spain, Sweden, Switzerland, United Kingdom.

ALMA Science AdvisoryCommittee

The Atacama Large Millimetre Array (ALMA) project hasformed a new committee to provide scientific advice to theproject and outreach to the wider community. The committeeholds monthly teleconferences and other meetings at regularintervals. The minutes of the teleconferences and reportsfrom the meetings are placed on the following web site.

http://www.eso.org:8082/committees/ASAC/index.html

A list of the committee members can also be found on thisweb site. Comments or questions can be addressed directlyto the individual committee members or to the committee byemail via the web site. In addition, we are all willing to givecolloquia or other presentations on the ALMA project.

47

PERSONNEL MOVEMENTSInternational Staff(1 December 1999 – 31 March 2000)

ARRIVALS

EUROPEDESSAUGES-ZAVADSKY, Miroslava (CH), StudentHAGGOUCHI, Karim (F), System Software EngineerMODIGLIANI, Andrea (I), Astronomical Data Analysis SpecialistREJKUBA, Marina (Croatian), StudentREYES, Javier (E), Electronics/Electrical EngineerRICHICHI, Andrea (I), AssociateTOKIVININE, Andrei (Russian), Associate

CHILEHUTSEMEKERS, Damien (B), Operations Staff AstronomerLAGER, Mikael (S), Microwave EngineerROBERT, Pascal (F), Electronics Engineer

DEPARTURESEUROPE

ENGELBART, Lore (D), SecretaryGIANNONE, Gino (I), Software Engineer for Scheduling ToolsNICOLINI, Gianalfredo (I), Infrared Laboratory TechnicianWOLF, Sebastian (D), Scientific Applications Developer

for UVESWOUDT, Patrick (NL), Fellow

CHILEAUGUSTEIJ N, Thomas (NL), AstronomerGEMPERLEIN, Hans (D), Infrared Instrumentation EngineerGUNNARSSON, Lars-Göran (S), Microwave EngineerJOGUET, Benoit (F), StudentPETR, Monika (D), FellowSCOBBIE, James (GB), Telescope Software ScientistTESCHNER, Klaus (D), Programmer

Local Staff(1 December 1999 – 31 March 2000)

ARRIVALS

ILLANES, Esteban, Public Relations Officer, VitacuraKAISER, Cristian, Personnel Assistant, VitacuraMARDONES, Pedro, Electronics Engineer, ParanalMIRANDA, Marcela, Administrative Secretary, VitacuraQUINTANA, Rolando, Warehouse Administrative Assistant,

Paranal RIJO, Ariela, Electronics Technician, ParanalSAGUEZ, Claudio, Warehouse Assistant, ParanalTAPIA, Francisco, Warehouse Assistant, Paranal ZARATE, Andres, Telescope Instrument Operator, Paranal

Scientific Preprints(October 1999 – February 2000)

1341. C.J. Cesarsky and M. Sauvage: A Mid and Far Infrared Viewof Galaxies. In: “Toward a New Millennium in GalaxyMorphology”, edited by D. L. Block, I. Puerari, A. Stockton andD. Ferreira (Kluwer, Dordrecht).

1342. F. Kerber et al.: ISO Observations of Dust Formation inSakurai’s Object. Monitoring the Mass Loss of a Very LateHelium Flash Star. A&A.

1343. J. U. Fynbo, B. Thomsen and P. Møller: Lyα Emission from aLyman Limit Absorber at z = 3.036. A&A.

1344. M.-H. Ulrich, A. Comastri, S. Komossa and P. Crane: TheSteep Spectrum Quasar PG1404+226 with ASCA, HST andRosat. A&A.

1345. V. Testa et al.: The Large Magellanic Cloud Globular ClusterNGC 1866: New Data, New Models, New Analysis. AJ.

1346. A. Fontana et al.: High Redshift Evolution of Optically and IR-Selected Galaxies: a Comparison with CDM Scenarios. MNRAS.

1347. D. Elbaz et al.: Source Counts from the 15 µm ISOCAM DeepSurveys. A&A.

1348. M. Chadid, K. Kolenberg, C. Aerts and D. Gillet: FirstDetection of a Frequency Multiplet in the Line-ProfileVariations of RR Lyrae: Towards an Understanding of theBlazhko Effect. A&A.

1349. T. Douvion, P. O. Lagage and C. J. Cesarsky: Element Mixingin the Cassiopeia A Supernova. A&A Letters.

1350. M. Romaniello, M. Salaris, S. Cassisi, N. Panagia: HST Ob-servations of the LMC Field Around SN 1987A: DistanceDetermination with Red Clump and Tip of the Red GiantBranch Stars. ApJ.

1351. O. Marco and D. Alloin: Adaptive Optics Images at 3.5 and4.8 µm of the Core Arcsec of NGC 1068: More Evidence fora Dusty/Molecular Torus. A&A.

1352. D. Baade: Observed Periodic Phenomena.S. Štefl and T. Rivinius: Heros Be Star Campaigns. The BePhenomenon in Early-Type Stars. ASP Conference Series,Vol. 3 × 108, 2000. M. A. Smith, H. F. Henrichs and J.Fabregat, eds.

1353. F. Paresce and G. De Marchi: On the Globular Cluster IMFBelow 1 MA. ApJ.

1354. A. Pasquali et al.: R4 and Its Circumstellar Nebula: Evidencefor a Binary Merger? AJ.

1355. F. R. Ferraro, P. Montegriffo, L. Origlia and F. Fusi Pecci: ANew IR-Array Photometric Survey of Galactic GlobularClusters: A detailed Study of the RGB Sequence as a StepTowards the Global Testing of Stellar Models. AJ.

1356. M. Kürster et al.: An Extrasolar Giant Planet in an Earth-likeOrbit. Precise Radial Velocities of the Young Star ι Horologii= HR 810. A&A.

ESO Studentship ProgrammeThe European Southern Observatory research student programme aims at providing the opportunities and the facilities to enhance the

post-graduate programmes of ESO member-state universities by bringing young scientists into close contact with the instruments, activities,and people at one of the world’s foremost observatories. For more information about ESO’s astronomical research activities please consultResearch Projects and Activities. (http://www.eso.org/projects/ or http://www.eso.org/science/). Students in the programme work on an ad-vanced research degree under the formal tutelage of their home university and department, but come to either Garching or Vitacura-San-tiago for a stay of up to two years to conduct part of their studies under the supervision of an ESO staff astronomer. Candidates and theirnational supervisors should agree on a research project together with the potential ESO local supervisor. This research programme shouldbe described in the application and the name of the ESO local supervisor should also be mentioned. It is highly recommended that the ap-plicants start their Ph.D. studies at their home institute before continuing their Ph.D. work and developing observational expertise at ESO.

The ESO studentship programme comprises about 14 positions, so that each year a total of up to 7 new studentships are available ei-ther at the ESO Headquarters in Garching or in Chile at the Vitacura Quarters. These positions are open to students enrolled in a Ph.D.programme in the ESO member states and exceptionally at a university outside the ESO member states.

The closing date for applications is June 15, 2000.Please apply by using the ESO Studentship application form now available on-line (http://www.eso.org/gen-fac/adm/pers/forms)

European Southern ObservatoryStudentship ProgrammeKarl-Schwarzschild-Str. 2, D-85748 Garching bei München, Germanyksteiner@eso.org

DEPARTURESBRIONES, Jorge, Mecánico de Cúpulas, La SillaPINO, Flavia, Personnel Administrative Assistant, VitacuraPRADO, Pablo, Scientific Instrument Operator / Night

Assistant, La Silla ZAPATA, Joel, Warehouse Administrative Assistant, Paranal

ESO, the European Southern Observa-tory, was created in 1962 to “… establishand operate an astronomical observatoryin the southern hemisphere, equippedwith powerful instruments, with the aim offurthering and organising collaboration inastronomy …” It is supported by eightcountries: Belgium, Denmark, France,Germany, Italy, the Netherlands, Swedenand Switzerland. ESO operates at twosites. It operates the La Silla observatoryin the Atacama desert, 600 km north ofSantiago de Chile, at 2,400 m altitude,where several optical telescopes with di-ameters up to 3.6 m and a 15-m submil-limetre radio telescope (SEST) are now inoperation. In addition, ESO is in the pro-cess of building the Very Large Telescope(VLT) on Paranal, a 2,600 m high moun-tain approximately 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. Thesetelescopes can also be used in combina-tion as a giant interferometer (VLTI). Thefirst 8.2-metre telescope (called ANTU) issince April 1999 in regular operation, andalso the second one (KUEYEN) has al-ready delivered pictures of excellent qual-ity. Over 1200 proposals are made eachyear for the use of the ESO telescopes.The ESO Headquarters are located inGarching, near Munich, Germany. This isthe scientific, technical and administrativecentre of ESO where technical develop-ment programmes are carried out to pro-vide the La Silla and Paranal observato-ries with the most advanced instruments.There are also extensive astronomicaldata facilities. In Europe ESO employsabout 200 international staff members,Fellows and Associates; in Chile about 70and, in addition, about 130 local staffmembers.

The ESO MESSENGER is publishedfour times a year: normally in March,June, September and December. ESOalso publishes Conference Proceedings,Preprints, Technical Notes and other ma-terial connected to its activities. PressReleases inform the media about particu-lar events. For further information, contactthe ESO Education and Public RelationsDepartment at the 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

48

ContentsSuccessful Commissioning of UVES at Kueyen . . . . . . . . . . . . . . . . . . . . . . 1

TELESCOPES AND INSTRUMENTATIONS. D’Odorico: UVES at Kueyen: Bright Prospects for High-Resolution

Spectroscopy at the VLT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2R. Hanuschik and P. Amico: VLT Pipeline Operation and Quality Control:

FORS1 and ISAAC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6B. Leibundgut, B. Pirenne, M. Albrecht, A. Wicenec and K. Gorski: Access to

VLT Data in the ESO Archive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12M. Sarazin: Chile Astroclimate, a Biannual Update . . . . . . . . . . . . . . . . . . . . 13R. Heald and R. Karban: ESO Demonstration Project with the NRAO 12-m

Antenna . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14O. Hainaut and the NTT Team: News from the NTT . . . . . . . . . . . . . . . . . . . 15New Pictures from Paranal Observatory . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

REPORTS FROM OBSERVERSE. Tolstoy, J. Gallagher, L. Greggio, M. Tosi, G. De Marchi, M. Romaniello,

D. Minniti and A. Zijlstra: Imaging With UT1/FORS1: The Fossil Recordof Star-Formation in Nearby Dwarf Galaxies . . . . . . . . . . . . . . . . . . . . . . 16

M. Franx, A. Moorwood, H.-W. Rix, K. Kuijken, H. Röttgering, P. van derWerft, P. van Dokkum, I. Labbe and G. Rudnick: FIRES at the VLT: theFaint InfraRed Extragalactic Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

R.P. Mignani, P.A. Caraveo and G.F. Bignami: Optical Observations ofPulsars: the ESO Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

P. Rosati, C. Lidman, R. Della Ceca, S. Borgani, M. Lombardi, S.A. Stanford,P.R. Eisenhardt, G. Squires, R. Giacconi and C. Norman: The ROSATDeep Cluster Survey: Probing the Galaxy Cluster Population out toz = 1.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

P. Møller: Spectral PSF Subtraction I: The SPSF Look-Up-Table Method . . . 31P. Møller, S.J. Warren, S.M. Fall, P. Jakobsen and J.U. Fynbo: SPSF

Subtraction II: The Extended Lyα Emission of a Radio Quiet QSO . . . . . 33B. Leibundgut, J. Sollerman, C. Kozma, C. Fransson, P. Lundqvist, F. Ryde

and P. Woudt: The Late Phase of SN 1998bw . . . . . . . . . . . . . . . . . . . . 36J.-M. Conan, T. Fusco, L. M. Mugnier and F. Marchis: MISTRAL: Myopic

Deconvolution Method Applied to ADONIS and to Simulated VLT-NAOSImages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

A New Look at the Sombrero Galaxy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

ANNOUNCEMENTSCERN, ESA and ESO Launch “Physics On Stage” . . . . . . . . . . . . . . . . . . . . 46ALMA Science Advisory Committee . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46ESO Studentship Programme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47Personnel Movements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47Scientific Preprints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

NEW ESO PROCEEDINGS AVAILABLE

The Procedings of the Bäckaskog Workshop on

EXTREMELY LARGE TELESCOPES(ESO Conference and W orkshop Proceedings No. 57)

have been published. The 306-page volume, edited by Torben Andersen, Arne Ardebergand Roberto Gilmozzi, is available at a price of DM 60.– (prepayment required).

Payments have to be made to the ESO bank account 2102002 with CommerzbankMünchen, or by cheque, addressed to the attention of

ESO, Financial Services, Karl-Schwarzschild-Str. 2,D-85748 Garching bei München, Germany

1357. K. Adelberger: Star Formation and Structure Formation at 1 <~ z <~ 4. Clustering at HighRedshift, ASP Conference Series, Vol 1999, A. Mazure and O. LeFevre, eds.

1358. J.U. Fynbo, W. Freudling and P. Møller: Clustering of Galaxies at Faint Magnitudes.A&A.1359. M.-H. Ulrich: The Active Galaxy NGC 4151: Archetype or Exception? A&A Review.

Variability of Active Galactic Nuclei. Contribution for the Encyclopedia of Astronomy andAstrophysics, Oxford Institute of Physics and McMillan Publ. Co. 1999.

1360. T. Broadhurst and R.J. Bouwens: Young Red Spheroidal Galaxies in the Hubble DeepFields: Evidence for a Truncated IMF at ~ 2 MA and a Constant Space Density to z ~2.

1361. G. A. Wade et al.: Magnetic Field Geometries of Two Slowly Rotating Ap/Bp Stars: HD12288 and HD 14437. A&A.S. Hubrig et al.: Rapidly Oscillating Ap Stars versus Non-Oscillating Ap Stars. A&A.M. Gelbmann et al.: Abundance Analysis of roAp Stars. V. HD 166473. A&A.

1362. F. R. Ferraro et al.: Another Faint UV Object Associated with a Globular Cluster X-RaySource: The Case of M92. ApJ.

1363. A. Grazian et al.: The Asiago-ESO/RASS QSO Survey. I. The Catalog and the LocalQSO Luminosity Function. AJ.