Table Of Content

Astronomy & Astrophysics manuscript no. 2257b February 2, 2008 (DOI: will be inserted by hand later) ⋆ First results from the ESO VLTI Calibrators Program A. Richichi and I. Percheron European Southern Observatory,Karl-Schwarzschildstr. 2, D-85748 Garching bei Mu¨nchen, Germany e-mail: [email protected] 5 Received / accepted 0 0 Abstract. The ESO Very Large Telescope Interferometer (VLTI) is one of the leading interferometric facilities. 2 It is equipped with several 8.2 and 1.8m telescopes, a large number of baselines up to 200m, and with several n subsystems designed to enable high quality measurements and to improve significantly the limits of sensitivities a currentlyavailable tolong-baseline interferometry.The fullscientificpotential oftheVLTIcanbeexploited only J if a consistent set of good quality calibrators is available. For this, a large number of observations of potential 5 calibratorshavebeenobtainedduringthecommissioningphaseoftheVLTI.Thesedataarepubliclyavailable.We 2 brieflydescribetheinterferometer,theVINCIinstrumentusedfortheobservations,thedataflowfromacquisition to processed results, and we present and comment on the volume of observations gathered and scrutinized. The 1 result is a list 191 calibrator candidates, for which a total of 12066 observations can be deemed of satisfactory v quality. We present a general statistical analysis of this sample, using as a starting point the angular diameters 2 previously available in theliterature. Wederivethegeneral characteristics of theVLTI transfer function,and its 3 5 trendwith timein theperiod 2001 through mid-2004. Asecond paperwill bedevotedtoadetailed investigation 1 of a selected sample, aimed at establishing a VLTI-basedhomogeneous system of calibrators. 0 5 Key words. Techniques: high angular resolution – Techniques: interferometric – Catalogs – Stars: fundamental 0 parameters / h p - 1. Introduction trackingandbeamfilteringthroughmonomodefibers,the o process ofcalibrationremainsa fundamentalrequisite for r Long-baseline interferometers provide at present the t the success of interferometric measurements, and this as- s highest angular resolution possible from the ground a pect has only become more pressing in view of the needs v: at optical and infrared wavelengths. In the last few in terms of accuracy demanded by the most recent inter- years, the CHARA, Keck and VLTI interferometers (ten i ferometers mentioned above. Indeed, the accuracy of cal- X Brummelaar et al. 2003, Colavita & Wizinowich 2003, ibration is at present the limiting factor in measurements r Glindemann et al. 2003) have started operation, while performedwiththeVLTI,whichadditionallyislocatedin a other facilities such as OHANA (Perrin et al. 2004) and thesouthernhemispherewhereinterferometriccalibrators MRO (Creech-Eakman et al. 2003) are being readied. have been traditionally more scarce. These new large interferometers couple for the first time The problem of calibration, although crucial for the hectometric baselines with large telescopes, offering a success of interferometric observations, has remained un- wavelength coverage from the visual to the thermal in- til now a relatively uncoordinated effort. In general, the frared and aiming also at nulling interferometry. tendencyhasbeentousedifferent,oftenindependentlists Ground-based long-baseline interferometry (LBI) is a of calibratorsat each interferometric facility. Another ap- technique that is heavily affected by atmospheric turbu- proachhasbeentoestimatetheangulardiameterforaset lence. This sets limits on the integration times, on the of suitable stars. Cohen et al. (1999) have produced a list quality ofthe wavefrontsthat are combinedfromthe var- of over 400 sources, with typical magnitudes and diame- ioustelescopes,andingeneralontheaccuracyofthemea- terswhicharewellsuitedmostly forbaselinesoflessthan surements. Although significantimprovements can be ob- 100m and relatively small telescopes. Bordé et al. (2002) tained by selecting observationalsites of excellent quality have refined this list, imposing selection criteria derived andbyadoptingtechniquessuchasadaptiveoptics,fringe from their experience at the IOTA interferometer, while Mérandetal.(2004)haveextendedittostarswithsmaller Send offprint requests to: A. Richichi ⋆ Table 4 is only available in electronic form at the CDS angular diameter and suitable for hectometric baselines. via anonymous ftp to cdsarc.u-strasbg.fr (130.79.128.5) or via The recent CHARM2 catalogue (Richichi et al. 2004) http://cdsweb.u-strasbg.fr/cgi-bin/qcat?J/A+A/ is an effort to include all known direct measurements at 2 A.Richichi and I. Percheron: First results from theESO VLTI Calibrators Program high angular resolution of angular diameters and binary Table 1. Schedule of VLTI subsystems and instruments. stars,aswellasindirectestimatessuchasthosementioned above.However,the best approachto the problem of cal- System Description Status (Oct 04) ibrators, at least those that can be resolved by interfero- VINCI Test instrument available metric facilities, is undoubtedly represented by direct ob- MIDI 10µm instrument available servations. DL 1-6 Six Delay Lines available AMBER 1−2.4µm instrument commissioning In this paper, we present the first results of an effort FINITO IR Fringe tracker commissioning to establish a network of calibrators for the VLTI. We MACAO 1-3 UT AdaptiveOptics available review the main characteristics of the interferometer and MACAO 4 UT AdaptiveOptics integration its instrumentation, and the observations carried out as AT1 Auxiliary telescope available part of the VLTI commissioning. We also briefly describe AT2 Auxiliary telescope integration theVINCIinstrumentusedforsuchobservations,andthe AT3&4 Auxiliary telescopes Europe process of automated data reduction employed. We then SID Test Siderostats to be decomm. presentthelistofobservedobjectsandthemainstatistical PRIMA Dual-feed & phase from 2006 conclusions, including an assessment of the performance referencing of the VLTI, its accuracy and stability. A detailed analysis and discussion of individual results is the subject of a second paper. with a broad K-band filter using a Hawaiiarrayas detector. For most of its operation, VINCI has been used with the fiber-basedbeamcombinerMONA(seeSect.3).Note 2. VLTI overview that at the time of writing another beam combiner is be- TheESOVeryLargeTelescopeInterferometer(VLTI),lo- ing used with VINCI, basedonintegratedoptics (IONIC, catedonCerroParanalinChile,isbasedonthesamesite Laurent et al. 2002, Le Bouquin et al. 2004). IONIC is and infrastructure as the widely known VLT observatory available in two versions, that cover the H and the K (Glindemann et al. 2000). In addition to the four 8.2m bands respectively. No extensive calibrator observations unit telescopes (UT), it includes a number of 1.8m aux- havebeencollectedandmadepublicyetwithIONIC,and iliary telescopes (AT). These are expecially designed for wewillrestrictourselvesinthispapertoVINCIoperation interferometry, with very compact structures and an op- inabroad-bandKfilter withthe MONAbeam-combiner. tical systemwhich replicatesthe light path inside the UT While VINCI was designed and built mainly to test telescopes (Koehler & Flebus 2000). Currently, one AT is and commission the VLTI, the two facility instruments available, two more are being delivered and a fourth one are MIDI and AMBER, both of which are installed and is in an advanced phase of construction. Also, two 0.4m operatingatthe VLTI.Inparticular,MIDI (Leinertetal. siderostattesttelescopes(SID,Derieetal.2000)areavail- 2000)operatesintheNband,withaforeseenextensionto able,whichhaverepresentedtheworkhorseofthecommis- the Q band. It offers various levels of spectral resolution sioning activities of the VLTI (Scho¨ller et al. 2003). The anditrepresentsoneofveryfewinstrumentsofitskindin ATs and SIDs can be moved over an array of 30 stations. the world, and the only one in the southern hemisphere. In particular, the ATs can be moved semi-automatically AMBER (Petrov et al. 2000) is based on a concept of a on a system of tracks and can be relocated in a matter fiber-based beam combination similar to VINCI. It offers of hours; this is a unique feature among large optical in- 3bandsinthenear-infraredwithvariouslevelsofspectral terferometers. Altogether, 254 independent baselines are resolution, and the possibility to combine either 2 or 3 available, leading to 3025 closure phases, with a rather beams. In order to achieve the highest possibile accuracy homogenous distribution in baseline lengths (from 8 to on differential phase measurements demanded by some of 205m) and orientations (Richichi et al. 2000). itsscientificgoals,AMBERemploysalsoaschemeofrapid The VLTI is following a dense implementation plan, beam switching. with several subsystems and instruments becoming avail- The main characteristic of the VLTI instruments are able and being commissioned in close sequence (Scho¨ller summarized in Table 2. We stress that in the case of the et al. 2003). Table 1 provides an overview of the main AMBER and MIDI instruments, which are still in a pre- steps. More details can be found in Derie (2000) for the liminary phase of operation without the full VLTI or not delay lines, Gai et al. (2003) for FINITO, Donaldson et entirely within specifications, the values listed are rela- al. (2000) for MACAO, Koehler & Flebus (2000) for the tive to the foreseenperformance only.Of particular inter- ATs, Derie et al. (2003) for PRIMA. est are the angular resolution, the sensitivity limits and The VLTI instruments and their characteristics are the accuracy of the visibility measurements. These rep- crucial in defining the science capabilities of the interfer- resent the main parameters required for the selection of ometer,andthespecificationsonthecalibrators.Thefirst VLTI calibrators. However, a number of additional steps of these instruments is VINCI, developed by the Paris are currently under study, that could change significantly Observatory in collaboration with ESO (Kervella et al. the scenario provided so far. For example, the dual-feed 2000). It achieved first fringes with the SIDs in March PRIMA facility (Derie et al. 2003) will push by several 2001,andwiththe UTs inOctober2001.VINCI operates magnitudes the sensitivity limits, in those cases in which A.Richichi and I. Percheron: First results from theESO VLTI Calibrators Program 3 Table 2. Summary of VLTI first generation instruments 6 and their characteristics related to the needs of calibration. 5 3 System VINCI MIDI AMBER Bands K N J,H,K 4 4 λλ (µm) 1.9-2.4 8-13 1.0-2.4 a Ang.Resol. (8m) 38.7 182-364 22.7-40.0 3 2 a Ang.Resol. (202m) 1.5 7.2-14.4 0.9-1.6 Spectr.Resol. (λ/∆λ) 5 30-260 35-14000 Beams 2 2 3 VINCI DPAocifflca.urrpizahacatysioe(n∆V/V) 10nn−oo3 10nn−oo2 1y0ne−os4 2 H I J K L M lae)LngiEmtxh.psMroeasfsgtenhdietiusndheo0r′.′t0e0s1t afonrd5l0o1%n0bgevsitsibbailnitd4yc,awthtehneacpe1pn3ltidcraabllwe.ave- A 1B C D E F G 0 0 0 b)effectivelyachievedduringcommissioning,inK-band,using 0 MACAO. 0 0 1 c) N-band magnitude (goal) for operation with the FINITO 0 fringe tracker.From Leinert et al. (2003). 0 0 0 d) K-band magnitude (goal) for 5σ fringe detection in 100ms 1 2 on UTs equippedwith MACAO (Strehl=0.5). From Petrov et 1 1 al. (2001). 1 23423 1 2 H I J K L M anearbybrightstarcouldbeusedforfringetracking.Itis 5 1 frtoheraeecshEeueanrbootphueatat2nt0hSmepalaigcmeaitbAiongugetns4ecnyhsoEituiSrvAsitoyafriinentKceugbrrraaentnidotnlwy.oEsutSludOdtyahinnendg A B C D E F G the possibility to build GENIE, a ground demonstrator Fig.1. Schematic layout of the VLTI interferometer, of the DARWIN nulling interferometer that will operate showingthebaselinesusedforthecalibratorsobservations later in space (Fridlund et al. 2003). Finally, a number of as listed in Table 3. North is up. concepts for a secondgenerationof VLTI instruments are also being considered (Glindemann et al. 2003). For the immediate purpose of this paper, we will not discuss the implications of such developments, that would require a ibility measurements. Thanks also to the stability of the new set of calibrators. VLTI environment,VINCI attains often uncalibratedvis- ibility accuracies at the 0.1% level for many hours during nightsofgoodseeing.Thispotentialaccuracyisoftennot 3. VINCI/MONA observations and data analysis fully exploited, due to the lack of suitable calibrators. The generallayoutoftheVLTI systemisshowninFig.1. The two interferometric and two photometric output Two afocal beams of light from either two siderostats or signals are imaged simultaneously on the detector. The twoUTsarereducedindiameterandsentthroughanun- fringesaredetectedbymodulationoftheopticalpathdif- dergroundlight duct system to the interferometric tunnel ference with an amplitude of typically 200µm to 300µm wheretheyarecompensatedforopticalpathdifferencesby using a piezo element within the VINCI instrument, and delaylines.The beams arethen directedinto the interfer- kept within the scan length by sending an offset signal ometriclaboratory.Here,theyarecompressedindiameter to the delay lines at slow rates of about 1 Hz. The two from80to18mm(notethiswillnotberequiredforATop- interferometric outputs, which are in opposite phase, are eration),andaresenttotheVINCIinstrumentwherethey subtracted to get a single interferogram which is free of arefocusedontosingle-modeopticalfibersandafterwords systematic system noise. The squared coherence factor is combinedbyafiberbeamcombinerprovidingtwointerfer- obtained for eachinterferogram.A detailed descriptionof ometricoutputs.Beforebeamcombination,thephotomet- thealgorithmsusedinprocessingtheVINCIdataisgiven ricsignalsareseparatedfromeachofthetwoinputbeams by Kervella et al. (2004). Note that two main approaches inordertomonitorthephotometricfluctuations.Thisde- can be selected, based on Fourier and wavelet transforms sign allows the higher order modes of the incoming radi- respectively.Experienceinthe VLTI groupwithrepeated ation to be filtered and achieves interference with beams measurementsonthe samesourceandwithlargevolumes which are only affected by the lower order modes, specifi- ofdatahasshownthatthetwoapproachesleadessentially callybytip-tiltinducedbyatmosphericpistonnoise.This to the same results and accuracy, although the wavelet permits optimal performance in terms of accuracy of vis- analysisismoreaccurateonsinglemeasurements.Wehave 4 A.Richichi and I. Percheron: First results from theESO VLTI Calibrators Program usedthis latter for ourcomputations of the transferfunc- Table 3. Statistics of the VINCI/MONA observations tion presented in Sect. 5. An observation consists of the executionofanobservingblock(OB),whichisasequence Baseline Length Nights Total Accepted of typically 300 to 500 scans. This sequence is also called (m) # OB Cal. # OB B3-C3 8.0 27 860 426 a batch. B3-D1 24.0 63 2675 1667 A description of the VLTI data flow, in connection B3-M0 139.7 81 1856 1118 also with VINCI, can be found in Ballester et al. (2002, D0-H0 64.0 50 825 690 2004). In summary,two data reduction systems are avail- E0-G0 16.0 294 8317 5606 able.Firstly,anautomateddatareductionthroughtheso- E0-G1 66.0 145 3147 2188 called pipeline is triggered by the execution of an OB on U1-U2 56.6 1 8 11 Paranal.A similar pipeline can be initiated manually off- U1-U3 102.4 15 422 250 line with parameter options in Garching as we did. This U1-U4 130.2 1 41 9 produces a synthetic ASCII summary called the quality U2-U3 46.6 9 149 85 control(QC)log.QClogsincludeamongothersthesource U2-U4 89.4 2 52 16 name, time of observation, baseline, telescope type, num- Total 688 18352 12066 ber of scans utilized, the coherence factor and errors of each interferogram as well as their average, and quality 4. The interferometric transfer function flags. Secondly, various data reduction packages are avail- The squared visibility of a source with angular diameter able, some of which are based on the same core of algo- φ, observed with an interferometer having a baseline B, rithms as the pipeline with an added library of graphical at the wavelength λ, is defined as interfaces and scan selection routines, and some of which 2 2 2J1(πφB/λ) start from the data produced by the pipeline and pro- V = (cid:20) πφB/λ (cid:21) vide additional selection and fitting algorithms. This in- teractivedatareductionisusedforanaccurateanalysisof Here,andthroughoutthe paper,weassume alwaysa uni- individual observations. However, for the large volume of form stellar disc. The effect of limb-darkening can be in- calibratorobservationsdiscussedherethisapproachisnot vestigatedby long-baselineinterferometry(Wittkowskiet practical.We reprocessedallthe rawdatarelevanttothis al. 2001, 2004), but it is relatively small and can be ne- paper with a single homogeneous version of the pipeline glectedforthepurposeofthispaper.Duetodegradations installed at the ESO headquarters in Garching, and we introducedbytheatmosphere,aswellasbytheopticsand used the average values present in the QC logs. mechanics of the interferometer, the actual squared visi- 2 2 2 Note that the pipeline applies flags to filter out indi- bilityobservedisVo,withVo <V .Thetransferfunction vidual scans and full OBs which do not meet a number TF is defined as the ratio of criteria. The settings that we used were such that a 2 2 TF=Vo/V scanwasrejectedifthesignal-to-noiseratio(SNR)onthe two photometric beams differed by more than a factor of In order to perform scientific measurements onstars with 5; if the fringe packet was too wide or too narrow in the an unknown angular diameter, it is necessary to estimate time domain (more than 50% away from the theoretical TF from stars for which the angular diameter is known. value); if the fringe peak was not at the correct position This is in essencethe process ofcalibrationof aninterfer- in the frequency domain (more than 30% difference from ometer. Since the TF is subject to the influence of atmo- the theoretical position); if the fringe peak was too wide spheric turbulence, it must be measured as frequently as or too narrow in the frequency domain (more than 40% possible. It is alsodesirable to estimate it from starswith differencefromthetheoreticalvalue);iftherewasanopti- a similar elevation, and ideally also in a similar direction calpathdifference(OPD)jumpofmorethan20µmbefore to the scientific target stars. the scan; if fringes were located at edge of the scan; and It can be assumed that the process of determining an if the scanvisibility was more than 3σ from the OB aver- interferometrictransferfunction andthe guidelines forits age. A full OB was rejected if less than 10 interferograms validity could be similar to those established for example couldbesuccessfullyprocessed;iflessthan5%ofthescans in studies of speckle interferometry (Aitken 1989, Leinert could be successfully processed; if the photometric signal 1994). We note however that some differences could ex- averaged SNR less than 1. More details on these criteria ist, for example linked to the different atmospheric prop- can be found in the paper by Kervella et al. (2004). erties over distances significantly longer than the outer Asummaryofthetotalobservations,orderedbybase- turbulence scale, and this effect could be site-dependent line,isgiveninTable3.Thelasttwocolumnslistforeach (Abahamid et al. 2004). There is no general study of the baseline,respectively,thetotalnumberofOBssuccesfully regime of validity of transfer function determinations for processed and the number of candidate calibrator OBs long-baseline interferometry. whichwereacceptedafterthetestsmentionedabove.Our For the purpose of this paper, we have used the whole definition of a candidate calibrator is provided in Sect. 4, database of VINCI measurements after automated pro- andmoredetailsonthedatastatisticsaregiveninSect.5. cessing by the pipeline, using only OBs which were not A.Richichi and I. Percheron: First results from theESO VLTI Calibrators Program 5 flaggedforrejectionasexplainedinSect.3.Wehaveiden- 5. Statistics of the observations tified a number of stars which satisfied the main criteria TheVLTIstartedoperationsinthemiddleofMarch2001, to be considered as candidate calibrators. These include using the VINCI instrument and the SID. Since then, ob- the stability,the absence ofnearbycompanions,andsuit- servations have continued almost uninterrupted on every able spectral types. More details are given by Percheron useful night. The main goal of this long period of obser- et al. (2003). A summary of the observations satisfying vations has been the commissioning of the VLTI, includ- these criteria is given in the last column of Table 3. ing the characterization of several baselines (see Fig. 1), Essentially,wehaveusedasthemainsourcetheVLTI of the delay lines, of the several subsystems, and of the Calibrator catalog available to users of the ESO web complex software overviewing the whole system (Scho¨ller tools. This in turn is based on entries from CHARM2 et al. 2003). More recently, scientific observations have (Richichi et al. 2004), from the list of MIDI calibrators started, including the so-called Science Demonstration (T.Verhoelst,priv.comm.)andfromthecatalogofBordé (SD), Guaranteed Time Observations (GTO) for MIDI, et al. (2002). Very few stars come from other miscella- and Open Time (OT) proposals by the community. All neous sources. Details are given in the on-line material commissioning and SD data taken on the sky with possi- (see Table 4). Our list includes stars whichcannot always ble scientific relevance are being releasedto the public on be considered good calibrators in a strict sense. In par- a regular basis. ticular,since MIDI is an instrument workingin the 10µm These public data form the basis of the material dis- range, the compilers of its list of calibrators were less in- cussed in this work. We have considered observations ob- clined to reject stars with some known variability in the tainedfromthenightoffirstfringes,16March2001,to22 near-IR,as long as it did not affect its mid-IR behaviour. July 2004, which corresponds to the latest public release. As aresult,someofthe MIDI calibratorshavequite large This makes a total of 1225 nights. The actual number of angular diameters and/orare variable in the near-IR.We nights with data useful for our reprocessing is about 60% donotconcernourselvesherewiththisissue,andwepost- of this, as detailed in Table 3. The rest of the time was pone a critical selection to the next paper. spent mostly on integration and commissioning activities On the basis of the available angular diameters, we withVINCIaswellwithotherinstruments,notonthesky, have computed the TF for each calibrator j and for i,j or for GTO and OT scientific observationswhich arepro- each night i, and the associated error ∆TF . We have i,j tected by proprietary rules. Overall, the fraction of time thencomputed foreachnightthe weightedaveragetrans- inwhichthe VLTIwasidlefromobservationsortechnical fer function and its error workwaslessthan2monthsoverthefirst3.4years,which represents in itself an impressive performance. ∆TF−2·TF TF = j i,j i,j Some general statistics are presented in Table 3. The i P ∆TF−2 total volume of data processed amounted to 20849 OBs, j i,j P or ≈150 Gb of raw FITS files. Of these, 18402 were suc- cessfullyprocessedbythepipeline.Noteaslightdifference 1 2 to the total number provided in Table 3, because for 50 ∆TFi = ∆TF−i,j2 OBs the baseline information was not reliable. After the Xj selections described in Sect. 3, 15123 OBs were accepted   as having a satisfactory quality, for a total of 287 stars. We stress that this is a rather rough approximation, A significant fraction of these stars were actually science since it is based on a pre-defined set of φ values and we targets,or otherwise consideredunsuitable as calibrators. j know thatmany of the listed calibratordiametersare not After removingthese entries,ourdatabaseincludes12066 accurate to the precision demanded by the VLTI/VINCI OBs of goodquality on 191candidate calibrators.Details measurements.Thevalueswehaveusedarethoseprovided aregiveninTable4,whichisprovidedason-linematerial. in the references mentioned above,and in many cases the In particular, we provide the following: a list of the can- diameters are either for a wavelength other than the K didate calibrators; their main parameters such as coordi- band, or several determinations are available. We have nates, magnitudes, spectral types; the baselines on which not attempted to derive a definitive value for each star, they were observed and how many times. but have taken instead what we considered a representa- Figs. 2 and 3 show the magnitude and brightness dis- tive value. Nevertheless, we estimate that the derived TF tributions of the observed candidate calibrators. The K as a function of time is a sufficient approximation for a magnitudeshavebeenobtainedfollowingthesamecriteria statistical discussion as provided in Sect. 5. It will be the and using the same information as described in Richichi subject of a second paper to go into more detail for a se- et al. (2004). Note that nominal values of the diameters lected sample of calibrators, and in particular to provide are used, as stressed in Sect. 4. It can be seen that most refinedangulardiametervalues.Notethattheerrorterms ofthesourcesarewithinthebrightnesslimitsofK=0and ∆TF includealsotheuncertaintiesontheangulardiam- K=3mag. This is a consequence of the typical brightness i,j eters ∆φ , which at this stage are necessarily as tentative limit attained with VINCI on the small SID telescopes. j as the initial values of angular diameters themselves. The detector integration times used with VINCI during 6 A.Richichi and I. Percheron: First results from theESO VLTI Calibrators Program Fig.4. Diameter accuracy distribution of the observed Fig.2. K magnitude distribution of the observed candi- candidate calibrators. date calibrators. Fig.5.Skydistributionoftheobservedcandidatecalibra- Fig.3.Angulardiameterdistributionoftheobservedcan- tors. didate calibrators. very large angular diameters: these are mainly from the MIDI list, and are unsuitable for near-IR calibration. the commissioning observations varied between 0.1 and Ideally,the bestcalibrationaccuracyisachievedwhen 20ms, howeverthe most commonly used values (account- the calibrator candidate can be assumed to be effectively ing for 80% of the data) were between 0.8 and 1.0ms. unresolvedat the given wavelengthand baseline. In prac- Theseshouldbeunderstoodastimespentonascan.When tice, as just mentioned in the context of Fig. 3, calibra- the total number of scans in an OB is considered, the tors can be marginally resolved. In this case, the relative correspondingtypical integration times were about 0.4 to accuracy with which the angular diameter is known will 0.45s. affect directly the accuracy of the final calibrated visibil- Also,itcanbenotedfromTable3thatthemajorityof ity of the science target. Fig. 4 shows that about 70% of the observations were carriedout on SID telescopes, with the observed candidate calibrators have angular diame- the UTs accounting for only 3.0% of the total number of ters known, or presumed, with better than 5% accuracy. acceptedOBs.Westressthatforthepresentworkonlyob- Inthisfigure,itcanbenotedhowsomeofthepointswith servationsonUTs withoutMACAOwereused.Typically, higheraccuracyandsmallerangulardiametersseemtobe there is a difference of about <∼4-5mag between the per- lined up along a few precise curves. These are candidate formance of VINCI at the UTs without MACAO and at calibratorsforwhichtheangulardiameterhasbeendeter- theSID.However,theVLTICalibratorcatalogwasbased minedindirectly,suchasthosepresentinthelistofCohen ontheSIDperformanceandthisexplainswhyFig.2does et al. (1999). The curves are the consequence of a precise notextendto K≈10whichwouldbe a morerealisticlimit link between the angular size and the accuracy of the de- for the UTs. termination,andareprobablyproducedbythe processof Theangulardiameterspeakat3mas,providingagood modelling and the associated parameter uncertainties. coverageof near-IRcalibrators especially for baselines up It can be appreciated from Fig. 5 how the sky distri- to about 100m. For longer baselines, this set of calibra- bution coverage is relatively uniform. For each of the 191 ◦ tors can be used but most of them will be significantly candidate calibrators, the nearest one is on average 5.8, ◦ resolvedandtheaccuracyofcalibrationcouldbeaffected. and never more than 15.8, away. More important is the For longer wavelengths, such as in the 10µm range, all distanceto the nearestcalibratorforanypointinthesky. VLTIbaselinescanbeadequatelycalibrated.Itshouldbe Restrictingthedeclinationrangebetween−85◦and+35◦, noted that a small number of candidate calibrators have equivalenttoanairmassof2asseenfromParanal,theav- A.Richichi and I. Percheron: First results from theESO VLTI Calibrators Program 7 ◦ eragedistancetothenearestcalibratoris9.0,anditnever 800 ◦ exceeds30 .Ifthe rangeofdeclinationsis restrictedsoas to limit the airmass to no more than 1.5, the averagedis- tancetothenearestcalibratoris7◦.0,anditneverexceeds hts 600 ◦ g 26 . ni # AsmentionedinSect.4,unfortunatelynoprecisestud- e 400 bl iesareavailableonthe dependence ofthe accuracyofcal- a ail ibration from the angular distance between science and v A 200 calibratorsources.Itisgenerallyassumedthatsuchadis- tance should be less than ≈ 15◦ for reliable results, and it is probably desirable to keep it even smaller. Davis et 0 al. (2004) found almost no changes in the VLTI transfer 1 2 5 10 15 20 30 function over a wide range of zenith angle, although this OBs per night studywasadmittedly limitedtoonenightonly.Fromthis Fig.6. Histogram of the number of nights according to point of view, it appears that the coverage provided by the number of OBs (observation blocks) carried out on the list of candidate calibrators observed by the VLTI is candidatecalibrators.The maximumnumber ofOBsever sufficient for most applications. observed in one night was 97. 0.6 6. The VLTI Transfer Function One main result that can be drawn from the large n database of VLTI observations of candidate calibrators is ctio0.4 n the evaluation ofthe VLTI transfer function, its accuracy u abnehdaivtsiosutra.bFiloitryt,hbios,thwoenhaavneigchotmlypubtaesdistahnedtirtasnlsofnerg-ftuenrmc- ansfer F0.2 2eh4tra6ca1ar6pa6 Tr hr6630 tion TF according to the definitions given in Sect. 4. We hr6682 lamsgr stress again that our computations are based on a set of tetcen angular diameters, and their uncertainties, which at this 0.0 -1 1 3 5 7 9 11 stage are only first approximations. Given the large vol- UT Time [hours] ume ofobservationsandthe number ofcandidatecalibra- tors,weshouldnotbeaffectedbysystematicbiasesinour Fig.7. Calibrators observations on the night of 2 June conclusions. In order to further reduce this risk, we have 2002, using the E0-G1 baseline. See the text for a discus- chosen to use only nights in which at least 23 OBs were sion on the rejection of discrepant visibility points. available. This number has been chosen as a compromise between having a good number of useful nights (about 0.3 200 according to this choice), and keeping at the same epssco time onlynightswithaveryhighnumberofobservations. gamaql Ahistogramofthe number ofnights asa function ofOBs ction0.2 etasgr observed is shown in Fig. 6. un F Additionally, we have chosen to discard from each er sf nightthoseOBs whichproduceda TFvalue departingby an0.1 Tr more than 3 standard deviations (σ) from the night average.Fig.7showsanexampleinwhichthreeOBsrecorded between 1 and 5 hours UT are obviously discrepant and 0.0 1 2 3 4 5 6 7 werediscardedaccordingtotheabovecriterium.Itcanbe UT Time [hours] noted that another OB appears to be discrepant, around 7.5 hours UT, but it wasnot discardedbecause it did not Fig.8. SameasFig.7,forthe nightof23July2001.The violate the 3σ criterium. baselineusedwasE0-G0.Nomeasurementswererejected, On the other hand, in spite of this simple rejection as explained in the text. criteria, the remaining validated OBs are not necessarily alwayssatisfactory.An example is shown in Fig. 8, where intuitively one wouldbe ledto considerthe OBsrecorded thenight,andnoobviouscorrelationisseenwithambient after 6 hours UT as discrepant.However,they do notdif- parameters such as seeing, wind, coherence time. fer by more than 3σ from the night average. Also, it is We did not push our analysis to identify specific rea- difficult to find an obvious reason for this apparent dis- sons for each rejection, or for the inclusion of apparently crepancy,sincetheyarerelatedtoastar(η Sgr)forwhich discrepant points. As stated earlier, our goal is mainly to several apparently good results were obtained earlier in deriveglobalstatisticalproperties,andwewilldevote an- 8 A.Richichi and I. Percheron: First results from theESO VLTI Calibrators Program 0.90 0.40 0.80 Transfer Function 00..3389 betcet Transfer Function 000...567000 betdor alfcma chisco delcma 0.40 delvel rscl hr4831 0.37 0.30 0 1 2 3 4 5 6 7 8 -2 -1 0 1 2 3 4 5 6 7 8 9 UT Time [hours] UT Time [hours] Fig.9. Calibrators observations on the night of 14 Fig.10. Calibratorsobservationsonthenightof20April November 2001, using the E0-G0 baseline. The weighted 2003, using the B3-M0 baseline. The weighted average average transfer function is 0.384 with a relative error of transfer function is 0.574 with a large relative error of only0.5%.Thiscanbeconsideredasanexampleofagood 15.8%. This can be considered as an example of a very night at the VLTI. badnight atthe VLTI.Note the largescatterin the measurements of each star. otherpapertotheindividualevaluationofselectedresults and to a refined list of angular diameters of selected cal- As such, it is quite evident in the plot of Fig. 11, however ibrators. In particular, as mentioned in Sect. 4, some of itdidnotaffectsignificantlyobservationscollectedduring the MIDI calibrators are known to be unsuitable for the the same night. near-IR. AsecondconsiderationisthattheVLTItransferfunc- Figs. 9-10 show examples of two nights of calibrator tion variedat eachbaseline change, but not dramatically. observations, and the corresponding TF average value. It Thisisduetothefactthatsuchchangescaninvolvemove- can be appreciated how, depending also on ambient con- mentsandswapsofoneormorerelaymirrorsinthedelay ditions and in particular on the atmospheric seeing and linetunnel.Inparticular,itcanbenotedthatthereareno coherencetime,theobservationsmightshowdifferentlev- majorsystematicchangesofthetransferfunctionbetween els of scatter and quality. On the night of November 14, SIDandUTbaselines.Thebiggestchangesinthetransfer 2001,theseeingwasgenerallyconstantatthelevelof0′.′6, functioncanbeeasilyassociatedwithtemperatureadjust- increasing towards 1′.′0 only at the end of the night. The ments in the MONA beam combiner, realignments of the coherence time was also approximately constant around generaloptics, and similar intervertions.When a baseline 4ms.Notethatthesevaluesarecomputedatvisualwave- wasinplaceforalongperiodoftime,long-termdriftsare lengths, and at a location on the mountain top separate evident,whichcanalsoberelatedtomanualinterventions from the VLTI telescopes. On the night of April 20, 2003 on the components of the VLTI. on the contrary, the seeing fluctuations were more pro- The smoothestpartof the long-termtransferfunction nounced, with values between 0′.′3 and 1′.′1 and increasing is betweenOctober 19,2001andMay 6,2003.Beforeand toover1′.′5attheendofthenight.Alsothecoherencetime after this range, the MONA problems mentioned above changedsignificantly during the night, ranging between 2 altered the long-term transfer function significantly. For and 8ms. the 115 nights inside this period, each having more than The overall TF of the VLTI, computed under the as- 23 OBs on calibrators, the average transfer function was sumptions illustrated so far, is shown in Fig. 11 for the 0.405±0.034. Of course, it should be kept in mind that totality of the available data. Firstly, one can notice a this relates to the long-termbehaviour only. On a nightly long period, in the summer of 2001, in which almost no basis the relative accuracies ranged from 0.1% to 2.9%, transfer function is available. Indeed, in this period sky with an average of 0.68% which can be considered as a observations at the VLTI were almost completely halted, satisfactory result. duetoaproblemintheMONAbeamcombiner.Theprob- lem,causedbytemperature-dependentpolarizationofthe 7. Conclusions single-modefibers,recurredseveraltimesinthelifetimeof this beam combiner. It required frequent adjustments of We have presented and discussed a large set of interfero- thelocaltemperatureofthefibers,anditbecameparticu- metricobservationsofcandidatecalibratorsobtainedwith larly worrysomein the summer of 2003,when a period of the VLTI, equipped with the VINCI/MONA beam com- verylowvaluesofthetransferfunctionscanbenoticed.It biner, over a period of more than three years. The obser- shouldbestressedthatthisinstabilityoftheinterferomet- vations have been obtainedin the frameworkof the VLTI ric efficiency of the MONA beam combiner is character- commissioning, and are publicly available from the ESO ized by relatively long time scales, of the order of weeks. web site. The data have been processed by an automated A.Richichi and I. Percheron: First results from theESO VLTI Calibrators Program 9 Fig.11. Long term plot of the night-averageVLTI transfer function, as computed from the nominal diameters of the candidate calibrators.Changes of baseline are markedby the verticalsegments at the top (short for SID, long for UT baselines). At this scale, the error bars are barely visible. pipeline, and have been subjected to a number of quality Aitken G.J.M. 1989, PASP,101, 471 criteria.Asubsetof12066observationshasbeenselected. Ballester P., Chavan A.M., Glindemann A., Guirao C., Jaffe From this, we have computed the VLTI transfer func- W. et al. 2002, SPIE Proc. Vol. 4844, Quinn P. (ed.), p. tion and discussed its main statistical properties both on 300 Ballester P., Licha T., McKay D., Percheron I., Peron M. et a nightly basis and in the long term. The typical transfer al. 2004, SPIE Proc. Vol. 5493, Oschmann J.M. Jr. (ed.), function accuracy (in the squared visibility sense) of the p.16 VLTI with VINCI/MONA of order of 0.7% on a nightly Bordé P., Coudé du Foresto V., Chagnon G., Perrin G. 2002, basis,onasampleof115nights,eachhavingatleast23ob- A&A393, 183 servations of candidate calibrators.For the long term be- Colavita M.M., Wizinowich P.L. 2003, SPIE Proc. Vol. 4838, haviour, the transfer function shows marked fluctuations Traub W.A. (ed.), p.79 whichgenerallycanbeexplainedintermsofslowdegrada- Cohen M., WalkerR.G., Carter B. et al. 1999, AJ 117, 1864 tionsoftheMONAbeamcombiner,andtoalesserextent Creech-Eakman M.J., Buscher D., Haniff C., Howell P., intermsofbaselinechanges,realignments,andsimilarin- Jorgensen A.et al. 2003 AAS,202, 2403 terventions.FromOctober2001toMay2003,the average DavisJ.,RichichiA.,BallesterP.,GittonPh.,GlindemannA. value of the transfer function was 0.405±0.034. et al. 2004, Astronomische Nachrichten,in press Derie F., Brunetto E., Duchateau M., et al. 2000, SPIE Proc. We present as on-line material a list of 191 candidate Vol. 4006, Interferometry in Optical Astronomy, P. Léna, calibrators, together with their main characteristics and A.Quirrenbach (eds.), p. 99 the statistics of their observations at the VLTI. This list DerieF.2000,SPIEProc.Vol.4006,InterferometryinOptical ispreliminary,inthe sensethatsomeofthesesources,ex- Astronomy,P. Léna, A.Quirrenbach (eds.), p. 25 tracted from a catalog of calibrators for the mid-infrared HogenhuisH.,VisserM.,DerieF.2003,SPIEProc.Vol.4838, beam combiner MIDI, are not ideally suited for calibra- Traub W.A. (ed.), p.1148 tioninthenear-infrared.Alsotheangulardiametervalues Donaldson R., Bonaccini D., Brynnel J., et al. 2000, SPIE aretentative,havingbeenextractedfromtheavailablelit- Proc.Vol.4007,InterferometryinOpticalAstronomy,P.L. erature without any effort to perform a consistency check Wizinowich (ed.), p. 82 againsttheVLTIobservations.Anotherpaperisinprepa- Fridlund M., Gondoin P. 2003, Ap&SS286, 93 Gai M., Corcione L., Lattanzi M.G., Bauvir B., Bonino D. et ration, with the goal of deriving a new set of angular di- al. 2003, Mem. Soc. Astron.It. 74, 472 ameters for a restricted sample of validated calibrators, Glindemann, A., Coudé du Foresto V., Delplancke F., et al. by means of a global solution to the VLTI observations 2000, SPIE Proc. Vol. 4006, Interferometry in Optical presented here. Astronomy,P. Léna, A.Quirrenbach (eds.), p. 2 Glindemann, A., Algomedo J., Amestica R., Ballester P., Acknowledgements. The VLT Interferometer is the result of Bauvir B. et al. 2003, SPIE Proc. Vol. 4838, Traub W.A. more than a decade of work in planning and engineering, and (ed.), p.89 of several intenseyears of efforts on thesite. Wewould liketo Kervella P., Coudé du Foresto V., Glindemann, Hofmann R. recognize the many tens of people involved in this enterprise, 2000, SPIE Proc. Vol. 4006, Interferometry in Optical even without being able to acknowledge each by name. This Astronomy,P. Léna, A.Quirrenbach (eds.), p. 31 researchhasmadeuseoftheSimbad database,operatedatthe Kervella P., Ségransan D., Coudé du Foresto V. 2004, A&A Centre de Données Astronomiques de Strasbourg (CDS), and 425, 1161 of NASA’s Astrophysics Data System Bibliographic Services Koehler B., Flebus C. 2000, SPIE Proc. Vol. 4006, (ADS). Interferometry in Optical Astronomy, P. Léna, A. Quirrenbach (eds.), p.13 Laurent E., Rousselet-Perraut K., Benech. P., Berger. J.P., References Gluck S. et al. 2002, A&A 390, 1171 Abahamid A., Vernin J., Benkhaldoun Z., Jabiri A., Azouit LeBouquin J.B., Rousselet-Perraut K., Kern P., Malbet F., M., Agabi A. 2004, A&A 422, 1123 HaguenauerP. et al. 2004, A&A 424, 719 10 A.Richichi and I. Percheron: First results from theESO VLTI Calibrators Program LeinertC.1994,inStarFormationandTechniquesinInfrared and mm-Wave Astronomy, Ray T.P., Beckwith S.V.W. (eds.), SpringerVerlag, p.215 LeinertC.,GraserU.,WatersL.B.F.M.,etal.2000,SPIEProc. Vol. 4006, Interferometry in Optical Astronomy, P. Léna, A. Quirrenbach (eds.), p.43 Leinert C., Graser U., Richichi A., Schöller M., Waters L.F.B.M. 2003, ESO Messenger, 112, 13 MérandA.,BordéP.,CoudéduForestoV.2004,A&Ainpress Percheron I., Richichi A., Wittkowski M. 2003, Astrophysics and SpaceScience , 286,105 PerrinG.,LaiO.,WoillezJ.2004,AAS,204,816AJ117,1864 PetrovR.,MalbetF.,RichichiA.,etal.2000,SPIEProc.Vol. 4006, Interferometry in Optical Astronomy, P. Léna, A. Quirrenbach (eds.), p. 68 Petrov R., Malbet F., Richichi A., Hofmann K.-H., Mourard D. et al. 2001, C. R. Acad.Sci. Paris, t,67 Richichi A., Blöcker T., Foy R., et al. 2000, SPIE Proc. Vol. 4006, Interferometry in Optical Astronomy, P. Léna, A. Quirrenbach (eds.), p. 80 RichichiA.,PercheronI.,KhristoforovaM.2004,A&Ainpress SchöllerM.,GittonP.,ArgomedoJ.,BallesterP.,BauvirB.et al. 2003, SPIEProc. Vol. 4838, Traub W.A. (ed.),p. 870 ten Brummelaar T.A., McAlister H.A., Ridgway S.T. Turner N.H.,SturmannL.2003,SPIEProc.Vol.4838,TraubW.A. (ed.), p. 69 Wittkowski M., Hummel C.A., Johnston K.J., Mozurkewich D., Hajian A.R., WhiteN.M. 2001, A&A 377, 981 Wittkowski M., Aufdenberg J.P., Kervella P. 2004, A&A 413, 711