REPO RT SFROMOBSE RV ERSA Study of the Activity of G and K

Home , ELODIE spectrograph, Geneva Observatory

R E P O RT S F R O M O B S E RV E R S A Study of the Activity of G and K Giants Through Their Precise Radial Velocity; Breaking the 10-m/sec Accuracy with FEROS J. SETIAWAN 1, L. PASQUINI 2, L. DA SILVA3, A. HATZES4, O. VON DER LÜHE1, A. KAUFER2, L. GIRARDI 5, R. DE LA REZA3, J.R. DE MEDEIROS6

1Kiepenheuer-Institut für Sonnenphysik, Freiburg (Breisgau), Germany; 2European Southern Observatory 3Observatorio Nacional, Rio de Janeiro, Brazil; 4Thüringer Landessternwarte Tautenburg, Tautenburg, Germany 5Dipartimento di Astronomia, Università di Padova, Padova, Italy 6Universidade Federal do Rio Grande do Norte, Natal, Brazil

1. Scientific Background giant stars and their progenitors. In par- ible stellar surface (as a result of stellar ticular, when combining accurate dis- rotation), they will also induce variabili- Asteroseismology is an indispensa- tances (e.g. from HIPPARCOS), the ty in the core of deep lines, as the Ca II ble tool that uses the properties of stel- spectroscopic determination of the H and K, which are formed in the chro- lar oscillations to probe the internal chemical composition and gravity along mosphere (see e.g. Pasquini et al. structure of stars. This can provide a di- with the oscillation spectrum, the stellar 1988, Pasquini 1992). Since FEROS rect test of stellar structure and evolu- evolutionary models will be required to allows the simultaneous recording of tion theory. Precise stellar radial veloc- fit all these observations. This could the most relevant chromospheric lines, ity (RV) measurements made in recent provide an unprecedented test bench it will be possible to test directly from years have not only discovered the first for the theories of the stellar evolution. our spectra the rotational modulation extra-solar planets, but have also un- Before this can be done, however, one hypothesis. In Figure 1, CaII K line covered new classes of low-amplitude must derive the full oscillation spectrum spectra of one target star taken at dif- variable stars. One such is represented (periods and amplitudes) for a signifi- ferent epochs are over-imposed: no ev- by the K giant stars which exhibit RV cant number of giant stars. idence for strong variability in the line variations with amplitudes in the range Although the short-period variations core is detected. of 50–300 m/s (Walker et al. 1989, in giants can only arise from radial and The expected rotational periods for Hatzes & Cochran 1993,1994 ab). This non-radial pulsation, the nature of the some giants are consistent with the ob- variability is multi-periodic and occur- long-period RV variations is still open to served periods and some evidence for ring on two time- scales: less than 10 debate. Possibilities include planetary equivalent-width variations of activity days and several hundreds of days. companions, rotational modulation by indicators accompanying the RV varia- In the most detailed study of a K gi- surface structure, or non-radial pulsations have been found in two K giants ant star, Merline (1996) found 10 pulsation. Each of these hypotheses has a (Larson et al. 1993; Lambert 1987). If tion modes with periods ranging from high astrophysical impact, and there the RV variations were caused by sur- 2–10 days in Arcturus. These modes are strong arguments to support each face inhomogeneities, then these were equally spaced in frequency, the of them. would be large enough to be resolved characteristic signature of p-mode os- When surface structures, such as ac- by future ground-based interferometry. cillations analogous to the solar 5- tive regions and spots, move on the vis- These stars will thus make excellent minute oscillations. The relatively large number of modes that may exist in giant stars means that these objects may be amenable to asteroseismic techniques. Asteroseismology is one of the next milestones in astrophysics, and presently there are relatively few classes of stars on which these techniques can be applied, so it is important when more such objects are discovered. Asteroseismology can be used to derive such fundamental stellar parameters as mass and radius. This is partic- ularly useful for giant stars as these oc- cupy a region of the H–R diagram where it is difficult to obtain accurate stellar parameters. Furthermore, the evolutionary tracks of main-sequence stars in the spectral range A-G, all con- verge on the giant branch, so it is im- possible to establish the nature of the progenitor star only from these theoretical tracks. Asteroseismology may play Figure 1: Ca II K line observations of one of the target stars. Observations taken in different a key role in understanding the stellar nights are overlapped. In case of rotational modulation induced by surface inhomogeneities, properties, structure, and evolution of the chromospheric core of this line will show detectable variations.

13 (2) its high effi- proaching that of other, state-of-art ciency permitted techniques. We will outline our method- the acquisition of ology in the data reduction and provide data on a large some tips for the potential users. sample of stars in one night, and fi- 3. Pushing the FEROS nally, Capabilities (3) the radial velocity accuracy Although not conceived explicitly for of 23 m/s, which very accurate radial-velocity measure- was shown during ments (original requirements were set the commission- to an accuracy of 50 m/sec), FEROS ing period (Kaufer has been equipped with a double-fibre et al. 1999) was at system, where the second fibre can be a level adequate used either for recording the sky or for for the study of gi- obtaining simultaneous calibration ant variability. We spectra to the science exposure (see also were confi- Fig. 3, which shows a part of a frame dent that the containing the stellar and simultaneous FEROS perform- calibration spectra), following the tech- ances could be nique pioneered by the Geneva group improved with a with ELODIE and CORALIE (Baranne targeted strategy et al. 1996). Figure 2: H-R diagram for the 63 stars observed so far for our pre - and by upgrading By using a simple cross-correlation cise radial velocity survey with FEROS. The sample well covers the the software and (Kaufer et al., 1999, see also FEROS Red Giant Branch and the “clump” region. the data analysis. user manual) in the commissioning As part of the time, it was shown that FEROS could targets for VLTI observations (von der ESO time on FEROS we obtained in obtain a radial-velocity accuracy of 23 Lühe et al. 1996). Periods 64 and 65 a total of 5 full and 6 m/sec by observing every night the G The non-radial g-mode oscillations half nights. An additional 12 half nights star HD10700 (Tau Cet), an object seem unlikely, since these are not ex- were allocated as part of the Brazilian which has been demonstrated to have pected to propagate through the deep time. Early in our programme, after sur- a radial velocity constant to better than outer convection zone in these stars, veying only 27 stars, we had indications 5 m/s (Butler et al. 1996). but nature is always full of surprises. that 80% of our target objects showed A “constant’’ star is extremely useful Alternatively, these long-period varia- variations on the time-scales of a few to to estimate the long-term accuracy of tions may represent more exotic pulsa- hundreds of days. This is consistent FEROS and to provide a standard for tions such as toroidal modes. The plan- with the results of earlier, more limited optimising the radial velocity analysis. et hypothesis is supported by these pe- surveys. Variability in G and K giants Demonstrating a lack of RV variability riods being long lived (for more than 12 may indeed be ubiquitous. In Period 65 in a standard star would make us more years) and coherent. In the case of the we could enlarge the sample which confident of our results on variable star Aldebaran, the long-period RV vari- now consists of 63 stars covering a stars in the programme. For this pur- ations were not accompanied by line- large fraction of the upper HR-diagram. profile variations (Hatzes & Cochran The H-R diagram of the observed stars 1998). Although these seem to exclude is shown in Figure 2. From this figure it rotations and pulsations, conclusive ev- is clear that the sample spans the idence in support of the planetary hy- whole Red Giant Branch as well as the pothesis has proved elusive. region of the “clump”. The targets were selected on the basis of accurate HIP- 2. Observations PARCOS parallaxes in order to ensure a reliable determination of the basic K giant RV variability is still a largely stellar parameters. The high resolution unexplored area. Not only is the nature and high S/N ratio of the FEROS spec- of the long-period variations unknown, tra will also allow the spectroscopic de- but a knowledge of the characteristics termination of effective temperatures, of the short-period (p-mode) oscilla- gravities and metallicities. With such an tions as a function of a giant’s position extensive data base it will be possible in the H–R diagram is also lacking. to investigate the dependence of the ra- For these reasons, in ESO Period 64 dial velocity variability on a wide variety (October 1999 – April 2000), we started of stellar parameters, including the es- an unprecedented survey of precise ra- timated stellar mass and evolutionary dial-velocity measurements as a Euro- status. This is important for calibrating pean-Brazilian collaboration aimed at theoretical evolutionary tracks. obtaining a sample of about 100 G-K gi- While the scientific results of this sur- ants and subgiants along the whole up- vey will require many more observa- per part of the HR-diagram (see Fig. 2). tions and their full analyses, in this We opted for using the FEROS spec- Messenger contribution we mostly con- trograph (Kaufer et al. 1999), at the centrate on the data-analysis process 1.52-m telescope, because: in order to demonstrate the capabilities (1) its large spectral coverage al- of FEROS. lowed the observer to record simulta- We believe that our results (8.3- neously fundamental lines like the H m/sec RV accuracy) are so encourag- and K of CaII, suitable for investigating ing that they open a new possible use Figure 3: Portion of a FEROS frame show - rotationally induced modulations (cf. of the FEROS spectrograph as a plan- ing the stellar spectrum and the simultane - Section 1); et hunter with an RV accuracy ap- ous calibration Th-A spectrum.

14 plying some addi- better than 2.5 m/sec. This indicates tional zero-point that the spectrograph can deliver excel- offset corrections. lent RV performances. The first four In order to determine the optimal steps are easily spectral range to be used for computing accomplished by the radial velocity, we checked the dis- the FEROS re- persion of each order by cross-correlat- duction in MIDAS ing the double Th-A(fibre 1 and fibre 2) without the need with the Th-A mask prepared for FER- for any major OS and the results are shown in Figure modification. Our 4. Orders 7 to 34 may be suitable for results testify the computing of accurate radial velocities excellent work (note that the order numbering used in made by the Hei- this work is the extraction order num- delberg group in ber, not the real echelle order number). the data reduction We mention that since the binary package, which is template for the stars was originally also suitable for prepared for the ELODIE spectrograph, the stringent needs it covers only the wavelength range of precise radial from 3600–6997 Angstrom correspon- velocity studies. ding to the FEROS orders 7 to 31. A From the three more extended mask covering also or- Figure 4: Determining the orders to be used for computing the pre - extraction meth- ders 31–34 will be prepared, and the cise radial velocity by cross-correlating Th-Aspectra from each fibre ods available, we addition of this signal may improve the with the numerical Thorium template. prefer the stan- final RV accuracy. dard extraction pose one or two observations of mode with cosmic-ray elimination. 4. Computing the Radial HD10700 were acquired on each night Moreover, it is important to have the Velocity of our programme. wavelength rebinning done with a step The analysis of the data is in princi- fine enough to avoid any spurious ef- To get the radial velocity variations, ple straightforward and consists of the fects. In our case we use a rebinning we compute the difference of the cross- following steps: step of 0.03 Angstrom. After spectra re- correlations of the two fibres order by (1) Wavelength calibration with 2-fi- binning, cross-correlations with the ap- order. In addition, for each order a zero- bre (“double”) Th-A exposures. Wave- propriate numerical masks (steps 5 and point offset is finally applied, consider- length solutions are constructed using 6) are performed by using the TACOS ing the velocity difference between the two 25-second exposures of a Th-A package, developed by the Geneva ob- two fibres in the double Th-A expo- hollow cathode lamp in the “object-sky” servatory. sures. The results found so far are mode. The required numerical masks have shown in the Figure 6 where the radial- (2) Spectral extraction. The extrac- been prepared for FEROS by modifying velocity measurements of Tau Cet are tion procedure has been implemented existing masks kindly provided by the shown: in 245 days spanning our ob- in the FEROS reduction pipeline inte- Geneva observatory. The stellar mask servations, the rms around the constant grated in MIDAS version 98 and 99. is based on a K giant spectrum. radial velocity is 8.3 m/sec. Several extraction options are avail- In order to evaluate the FEROS in- This puts FEROS as a full member of able. trinsic capabilities, the La Silla 2.2-m the select family of spectrographs ca- (3) Removing the blaze function. This team kindly acquired a series of 10 pable of obtaining an accuracy below is done by the FEROS package using Th-A spectra with the calibration lamp 10 m/s, a regime capable of detecting Flat Fields (i.e. observations of spec- illuminating the two fibres; these spec- extra-solar planets. trophotometric standards are not re- tra were cross-correlated with the Th-A This is quite remarkable, considering quired). template mask, to investigate the that these performances are more than (4) Rebinning of the spectra. The behaviour of the spectra are rebinned order by order instrument. We rather than using the merged spectrum need in fact to normally produced by the FEROS on- evaluate within line reduction pipeline. which limits the (5) Order-by-order cross-correlation separation be- of the object spectrum with an appro- tween the two fi- priate stellar numerical mask. bres can be con- (6) Order-by-order cross-correlation sidered constant. of the calibration spectrum, which is A total of 50 called the “simultaneous Th-A”, with a T h - A e x p o s u r e s numerical template. This template is were acquired for based on a Th-A atlas that has been this purpose. The modified for our own purpose. T h e result is shown in modification takes into consideration Figure 5, where the spectrograph resolution and line the shift: fibre 1- blending, and is therefore instrument- fibre 2 cross-cor- dependent. The whole Th-A spectrum relation is shown. needs to be subdivided in small por- The results are tions and each portion of the mask is extremely en- “cleaned” by analysing the correspon- couraging and in- dent cross-correlation function. dicate that the (7) Radial-velocity calculation by two fibres “follow” Figure 5: Determining the FEROS calibration stability by checking the subtracting the cross-correlations of the each other with shifts between fibre 1 and fibre 2. 50 Th-Aexposures have been tak - stellar and calibration spectra and ap- an accuracy of en for this purpose; the results indicate and RMS of 2.3 m/sec.

15 of FEROS a num- We are also conscious that these ber of tips on the performances could be improved; there handling of the may be effects which we did not take data. into consideration enough so far; 1. When rebin- among them we can foresee: ning the wave- (i) Being interested in giants, our length within the stellar mask spectrum is taken from FEROS pipeline a K giant template, while Tau Ceti DRS, if the com- (HD 10700) is a G dwarf. An ad-hoc mand “REBIN/ mask for this star might already im- FEROS” is used, prove the results. this will apply the (ii) With the help from the Geneva barycentric cor- observatory we will try to extend the rection also to the star template to the red-part regions. simultaneous Th- This may enable us to gain more infor- A spectrum. This mation from the spectra. should be cor- (iii) FEROS is not equipped with an rected with the exposure meter, therefore we can not next MIDAS re- accurately determine the median time lease. of our observations (in terms of ac- 2. The FEROS quired photons); our observations are Figure 6: Radial velocity of HD10700 (Tau Ceti) taken for determin - spectra are auto- rather short (a few minutes), but this ef- ing the long-term accuracy of FEROS. After monitoring the radial ve - locity of this constant star for almost one year (245 days), we obtain matically correct- fect could become relevant in long ob- an rms of 8.28 m/s of constancy. ed for barycentric servations, poor guiding and/or cloudy correction. We nights. use the MIDAS (iv) In our first observations we only 5 times better than the original specifi- commands “COMPUTE/PREC” and took two double Th-A exposures every cations of the instrument. But mostly “COMP/BARY” for this purpose, but night. Several Th-A observations in the considering that, thanks to its impres- these will be checked against more ac- middle of the observing run would be sive efficiency and large spectral cover- curate methods. This step should be required to monitor any possible shifts age, FEROS is already a quite unique done carefully in order to improve the during the night and to provide addi- high-resolution spectrograph. results. It is also relevant to note that tional calibrations for the offset deter- On the technical side, such a good the FEROS pipeline computes the mining in the radial velocity compu- performance is somewhat unexpected barycentric correction reading the stel- tation. since FEROS fibres are not equipped lar right ascension and declination from with a light scrambler, and one could the 1.52-m telescope headers and it image that the spectrograph pupil could takes the UTat the beginning of the ob- 6. Acknowledgement s u ffer from small shifts or variable servations. In our experience, at this asymmetries in the light distribution. In accuracy the 1.52-m telescope stellar We acknowledge the strong support addition, FEROS is equipped with an co-ordinates (contained in the frame and help by Didier Queloz, Dominique image slicer. While this device will help FITS headers) may not be accurate Naef and Francesco Kienzle in sharing in scrambling the light, and therefore in enough; also the UT should be com- their time, experience and the TACOS. mitigating the problem above, the light puted for the middle of the exposure. We also thank Emanuela Pompei, is divided into two parts (half moons), 3. The real geographic co-ordinates Rolando Vega and Arturo Torrejon for which will have a slightly diff e r e n t of the 1.52-m telescope should be used their assistance during the observing wavelength solution. In principle, if the when computing barycentric correction. runs. The presence of J.S. at ESO was relative illumination of the slices We have not developed a full auto- supported by the ESO DGDF2000 pro- changes slightly from one observation matic procedure yet, which is supposed gramme. to the next (remember that 8 m/sec cor- to enable comput- responds to ~ 1/300 of pixel), this will ing the radial ve- be enough to introduce additional noise locity shortly after in the precise radial velocity measure- the MIDAS pipe- ments. line reduction and So far, in addition to HD10700, an- directly linked to other star has been reduced: the K1 III the cross-correla- giant HD18322, and the result is very tion process using promising: as shown in Figure 7 the TACOS. peak-to-peak radial-velocity variations Also, we found are of ~ 200 m/sec and they are con- that in some ob- sistent with a 40-day period (which is servations, one or also shown in the same figure). We two of the se- stress that due to the small number of lected orders de- points the 40-day period we detect may part strongly on be an alias of another period. More the expected solu- sampling is needed to determine the tion and they need period accurately. to be discarded. This operation is at the moment 5. Some Tips for Potential Users performed manu- ally, and we need Figure 7: Results of the radial velocity measurements of HD18322. The real process, of course, is in to build an algo- Although there are still many data points missing (no observing runs practice a bit more complex than de- rithm to select the between December 1999 and June 2000), we are able to show the scribed in the previous section, and we good orders auto- variability of the star, which is consistent with a periodicity of about would like to give to the potential users matically. 40 days and a peak-to-peak variation of about 200 m/s.

16 References Hatzes, A.P. & Cochran, W.D. 1994b, ApJ Pasquini, L., Pallavicini, R. Pakull, M. 1988, 432, 763. A&A 191, 266. Baranne, A., et al., 1996, A&A Suppl Ser Hatzes, A.P. & Cochran, W.D. 1998, MNRAS Pasquini, L. 1992, A&A 266, 347. 119,1. 293, 469. von der Lühe, O., Solanki, S., Reinheimer, Butler, R.P., et al. 1996, PASP 108, 500. Kaufer, A., et al. 1999, The Messenger 95, 8. Th. 1996, in IAU Symp. 176, S t e l l a r Hatzes, A.P. & Cochran, W.D. 1993, ApJ Larson, A. et al. 1993, PASP 105, 825. Surface Structure 147–163. Wa l k e r, 413, 339. Lambert, D.L. 1987, ApJS 65, 255. G.A.H., Yang, S., Campbell, B., and Irwin, Hatzes, A.P. & Cochran, W.D. 1994a, ApJ Merline, W.J. 1996, ASP Conf. Ser. 135, p. A.W. 1989, ApJL 343, L21. 422, 366. 208.

Crowded Field Photometry with the VLT: the Case of the Peculiar Globular Cluster NGC 6712 F. PARESCE1, G. DE MARCHI2, G. ANDREUZZI 3, R. BUONANNO 3, F. FERRARO4, B. PALTRINIERI3, L. PULONE 3

1European Southern Observatory, Garching, Germany 2European Space Agency, STScI, Baltimore, USA 3Osservatorio Astronomico di Roma, Rome, Italy 4Osservatorio Astronomico di Bologna, Bologna, Italy

1. Introduction study that best illuminate the perform- 2. VLT Observations of ance of this instrument combination for NGC 6712 The Hubble Space Te l e s c o p e crowded field photometry that might be opened up the exciting possibility of of interest to anyone contemplating do- NGC 6712 (a = 18h 53m 04.3², d = carrying out very deep, high-precision ing this sort of work with the VLT in the –08º 42¢ 21.5²) is a relatively metal stellar photometry in very crowded future. Results from a preliminary study poor, small and sparse GC ([Fe/H] = fields such as those routinely encoun- of this object using the VLT Te s t –1.01 and concentration ratio c = 0.9; tered in the core of dense stellar clus- Camera during the UT1 Science Harris 1996) located in the midst of a ters and galaxies. This capability has Verification period were reported in De rich star field at the centre of the allowed, for example, the reliable de- Marchi et al. (1999). Scutum cloud (Sandage 1965), which tection of stars all the way down to the bottom of the main sequence (MS) at the centre of nearby globular clusters (GC) six or seven magnitudes below the turn-off (TO) and well into the brown dwarf substellar region of nearby star- forming regions and of resolving the evolved stellar populations in nearby dwarf galaxies. Enormous progress in our understanding of the age, distance, star formation rates, and mass func- tions of a large sample of stellar populations has been a direct result of the last ten years of HST. The advent of the VLT with wide-field optical and IR cameras such as FORS and ISAAC present us with a golden opportunity to push further our quest for good quality photometry of faint sources in crowded fields. The potential of very good and stable seeing coupled with the huge collecting area of an 8-m diameter telescope with wide field and broad spectral coverage detectors cer- tainly rivals and even surpasses, in principle, even the exceptional HSTca- pability in this endeavour. In order to test these capabilities in practice, our group proposed to study in depth a par- ticularly interesting example of a crowded stellar environment namely the GC NGC 6712. We report here the preliminary results of our study of this cluster using FORS1 and UT1 obtained with 12 hours of observations during Period 63. Figure 1: Locations of the four FORS1 fields on NGC 6712. The centre of the cluster is lo - We emphasise those aspects of the cated at the origin of the co-ordinate system.