NAOS-CONICA Performance in a Crowded Field – the Arches Cluster A. STOLTE1, W. BRANDNER1, E.K. GREBEL1, DON F. FIGER2, F. EISENHAUER3, R. LENZEN1, Y. HARAYAMA3

1Max-Planck-Institut für Astronomie, Heidelberg, Germany 2Space Telescope Science Institute, Baltimore, USA 3Max-Planck-Institut für Extraterrestrische Physik, Garching, Germany

1. NAOS-CONICA – Near-Infrared Way. With an age of only ∼ 2–3 Myr it field (Schödel et al. 2002 and Ott et al. Camera with Adaptive Optics contains a rich massive stellar popula- in this issue), NIR wavefront sensing tion with about 150 O stars located in with NIR bright reference targets (K ∼10 NAOS-CONICA is a near-infrared the cluster centre (Figer et al. 1999), mag) is capable of improving the per- imaging camera and spectrograph and several more found in the immedi- formance in a highly extincted region. (CONICA) attached to an adaptive optics ate vicinity. The core density has been All images were taken in auto-jitter 5 –3 (AO) system (NAOS) for wavefront cor- estimated to be 3 ⋅ 10 M pc , and the mode. The auto-jitter template allows 4 rections (Lenzen et al. 1998, Rousset et total stellar mass exceeds 10 M automatic, random offsets between in- al. 2000). The AO system is designed to (Figer et al. 1999). The massive cluster dividual exposures to facilitate back- deliver diffraction-limited observations at centre contains ∼ 12 Wolf-Rayet stars ground subtraction and avoid after- an 8-metre-class telescope. NAOS is (Cotera et al. 1996, see also Blum et al. glows at the position of bright sources. equipped with two Shack-Hartmann 2001), intertwined with an intermediate- A set of additional exposures with short wavefront sensors operating at visual mass main-sequence population. At the detector integration times were taken to and NIR wavelengths, respectively. faint end, where pre-main-sequence obtain photometry of the bright cluster Most existing AO systems feed near-in- stars would be expected, the stellar stars (8 × 1 s DIT in H, 8 × 0.5 s DIT in frared (NIR) cameras, but operate at vi- population is severely contaminated by Ks). Seeing conditions varied during the sual wavelengths. In addition to the fact bulge stars, as Arches is located in the Arches observations. In H-band, the that the spectral energy distribution of central region of the Milky Way at a pro- seeing ranged from 0.45 to 0.85, while bright stars peaks at optical wave- jected distance of only 25 pc (roughly during K-band observations the seeing lengths, visual detectors offer higher 10 arcminutes) from the Galactic degraded from 0.8 to 1.3. On April 4, sensitivity and lower read-out noise Centre (Fig. 1). the moon was only 6° from our target, than NIR detectors, thus allowing one With a distance of about 8 kpc and a and increased background fluctuations to use fainter targets for wavefront foreground extinction in the visual of 24 decreased the detectability of faint sensing. The restriction to visual refer- to 34 mag, Arches is a very challenging sources in the cluster vicinity. ence sources, however, excludes object. The high foreground extinction The data were reduced within IRAF, deeply embedded objects located in re- obscures the faint cluster population at using twilight flat fields taken during gions with active star formation or high visual wavelengths, so that NIR obser- Comm 3 and combined with the nacop foreground extinction such as the vations, where extinction is less severe, twflat algorithm in eclipse (Devillard Galactic Centre. With wavefront sen- are necessary to detect the cluster pop- 1997). Eclipse is the ESO pipeline re- sors operating both at visual and NIR ulation. Given these extreme proper- duction package, which allows special wavelengths, NAOS combines the ad- ties, and the existence of data sets ob- treatment of particular instrument char- vantages of visual, faint reference tar- tained with HST/NICMOS and the acteristics adapted to the available gets with the option to penetrate highly Gemini/ Hokupa’a AO system allowing ESO instrumentation. Sky frames were obscured regions containing bright in- a detailed comparison, the Arches clus- created from the cluster observations frared sources. NAOS-CONICA was ter was chosen as NAOS-CONICA themselves, which is not ideal, but was commissioned on Yepun (UT4) in the commissioning target to test the NACO sufficient for our technical study. Star- course of Period 69. For more detailed performance on a dense, crowding-lim- light was excluded from these skies by information on NACO performance and ited stellar field. rejecting the bulk of the bright pixels. As first light, see Brandner et al. (2002) in the CONICA detector has a significant The Messenger 107. 3. NAOS-CONICA fraction of time-varying hot pixels, a We have used the CONICA JHK im- Commissioning Data cosmic ray rejection routine was used aging mode with the S27 camera in to obtain a bad pixel mask. This mask combination with the NAOS visual Arches H and Ks images were ob- was applied during the drizzling wavefront sensor for performance tests tained during Commissioning 3 of process, in which images are shifted in the crowded stellar field of the dense, NAOS-CONICA in March and April and combined. The 7 exposures with massive star cluster Arches. The S27 2002. 20 exposures with integration the highest resolution were combined camera with a pixel scale of 27 milliarc- times of 1 min (2 × 30 s detector inte- in Ks to one 7 min integrated frame, and second (mas) and corresponding field gration time (DIT)) each were obtained the 14 best exposures in H, where see- of view (FOV) of 27″ × 27″ is optimized in H, and 15 × 1 min exposures in Ks (4 ing conditions were much more stable, for diffraction-limited observations in × 15 s DIT). As the NIR wavefront sen- could be used to obtain a 14 min H- the J to K band wavelength range. At sor was not available during Comm 3, a band image. During the K-band obser- 2.2 µm, the diffraction limit of an 8-m blue foreground giant with V = 16 mag vations, the AO performance changed telescope is 57 mas, such that the S27 (V–R ∼1 mag) served as the reference with time as the seeing worsened. As camera allows Nyquist sampling. source for the visual wavefront sensor. crowding is the limiting factor on the Although the brightest star in the Arches field, the gain in resolution by 2. Why Arches? Arches field, this star is already close to rejecting the large fraction of frames the limiting magnitude (∼17 mag) of the with lower quality supersedes the loss The Arches cluster is one of the few visual wavefront sensor. As has been in photometric depth due to the shorter starburst clusters found in the Milky demonstrated on the Galactic Centre resulting total integration time on the fi- 9 Figure 1: A JHK composite 2MASS image of the Galactic Plane, including the Arches Cluster and the Galactic Centre, is shown in the left panel. The Arches cluster is the little blob of stars pointed out in the North of the image. The CONICA field of view (FOV) is indicated. The NAOS-CONICA H and Ks two-colour composite taken during Commissioning 3 is shown in the right panel. The FOV of this mosaic is 25″ × 27″, approximately the FOV of the CONICA S27 camera. The dense popu- lation in the Arches cluster centre has been resolved with NACO for the first time. nal image (see also Tessier et al. 1994). density gradient over small spatial in a very dense stellar field, background The best K-band photometry was ob- scales. fluctuations and saturation peaks fre- tained when combining only the frames quently cause false detections. where the best AO performance was 4. Luminosity Functions As sources in the Arches cluster are achieved. The resulting images have hidden behind ∼ 30 mag of foreground Strehl ratios of 14% in H and 20 % in K, In Figure 2, luminosity functions extinction, reddening influences the de- yielding a spatial resolution of ∼ 85 mas (LFs) of the NACO commissioning data tection of stars more severely at short- in both filters. are compared to HST/NICMOS LFs of er wavelengths. Consequently, the H- Standard DAOPHOT PSF fitting pho- the same field. Arches was observed band LF shown in Fig. 2 represents the tometry was used to obtain the final with NICMOS2 for 256 s integration photometric performance more realisti- magnitudes. A constant PSF function time in the broadband filters F110W, cally, as the K-band LF is artificially has been used, as isolated PSF stars F160W and F205W corresponding to J, truncated by the requirement that are rare in the dense Arches field. A H, and K. The NICMOS detection limits sources have to be detected in H as Penny function, which has a Gaussian were approximately 21 mag in H well. core with elongated Lorentzian wings, (F160W), and 20 mag in K (F205W). While the field star LF is comparable yielded the best fit to the data. Within a factor of 2 longer integration for NACO and NICMOS down to the The NACO data were calibrated time in K and a factor of 3.3 in H, NACO NICMOS detection limit of H ∼ 21 mag, using existing HST/NICMOS photom- reaches detection limits of 22 mag in H NACO clearly gains a large number of etry of the same field (Figer et al. 1999). and 21 mag in K, one magnitude deep- sources in the cluster centre. Here, the All calibrated data shown below are er than the NICMOS observations. significantly better resolution of 85 mas in the NICMOS photometric system. Note that NICMOS does not have to with VLT/NACO vs. 210 mas with This allowed a large number of stars to fight the large NIR background from the HST/NICMOS allows us to resolve the be used for zero-pointing and for the Earth’s atmosphere, but also has the cluster centre for the first time. Note study of photometric residuals over the much smaller collecting area of a 2.4-m that the diffraction limit of the 2.4 m field. We find the NACO performance to mirror as opposed to Yepun’s 8-m mir- HST mirror of 183 mas in H and 210 be very stable over the field. No trend ror. mas in K does not allow a much higher of the amplitude of photometric residu- Only objects detected in both filters resolution with NICMOS, while better als with field position is observed. Note are included in the NACO LF, as the re- Strehl ratios may yield resolutions that the measurement of anisopla- quirement that objects have to be de- down to the VLT diffraction limit of 57 natism on the Arches field is complicat- tected in both filters is a very efficient mas in H and 69 mas in K with NAOS- ed by the high stellar density and large method to reject artefacts. In particular CONICA. The cluster centre LF dem-

10 Figure 2: H-band luminosity function of the Arches cluster. The left LF contains only objects at a distance R > 10″ from the cluster centre, dis- playing the field performance in comparison to HST/NICMOS, while the right panel shows the cluster centre regions with stars within R < 5″. While the field performance between NACO and NICMOS is comparable until the NICMOS detection limit of H ∼ 21 mag is approached, the cluster centre LF shows the large gain in the NACO data due to the higher spatial resolution. It is particularly intriguing that already bins fainter than H = 15 mag are affected. This indicates that the cluster centre population was until now never truly resolved, not even at the brighter end. onstrates not only the deeper photome- neighbours vanishing in the PSF pat- sion) from the Geneva set of models try with NACO, but also the strongly en- tern of a bright star, or close pairs where (Lejeune & Schaerer 2001) was used to hanced resolution, as in nearly each only one component has been re- transform K-band magnitudes into stel- magnitude bin above H = 15 mag more solved. Figure 4 shows the comparison lar masses.1 A distance modulus of stars are resolved. This affects statisti- of resolved sources with K < 15 mag on 14.52 mag and extinction of AV = 24 cally derived properties of the Arches a sample field in the cluster centre. mag, corresponding to the cluster cen- stellar population, such as stellar densi- tre, have been applied. Magnitudes ty, total mass and the slope of the mass 6. Mass Function function. The NACO CMD was used to derive 1Note that the current best age estimate from high- 5. Colour-Magnitude Diagrams the mass function (MF) of the Arches resolution spectra of the bright Arches main se- quence O stars is 2.5 Myr (Figer et al. 2002), but cluster population. A 2 Myr isochrone we prefer to use the 2 Myr isochrone to allow direct A comparison of colour-magnitude with twice solar metallicity (see Figer et comparison of the NACO MF with our prior results diagrams (CMDs) of the Arches cluster al. 1999, Stolte et al. 2002 for discus- given in Figer et al. 1999 and Stolte et al. 2002. centre is shown in Figure 3. Here, Gemini/Hokupa’a low Strehl data (2.5% in H, 7% in K, Stolte et al. 2002), HST/NICMOS observations (Figer et al. 1999), and the new NACO observa- tions are compared. The Gemini data cover the smallest field of view, and care was taken to select only the area in common to all data sets to allow di- rect comparison. The NACO and NIC- MOS data were selected accordingly. While the low-Strehl Hokupa’a data show a large scatter in the main se- quence due to the contamination of stellar fluxes with extended seeing ha- los, the NACO data show a well-con- fined, narrow main sequence. The NIC- MOS data are also very well confined, but in accordance with the results from luminosity functions the NACO main sequence appears more densely popu- lated. Derived physical and statistical properties of the cluster population are consequently less biased by crowding losses. Figure 3: Arches colour-magnitude diagrams. A comparison between VLT/NAOS-CONICA, The Hokupa’a data set has the high- HST/NICMOS and Gemini/Hokupa’a data is shown. While the low Strehl ratio of the Hokupa’a est incompleteness, being 90% com- data of only a few per cent severely limits resolving the cluster population and caused large plete only down to K = 15 mag. When photometric uncertainties, the moderate Strehl NAOS-CONICA observations exhibit a nar- the number of stars down to this mag- row, well-defined main sequence, reflecting the photometric quality of the data. The NICMOS data are limited by the small telescope diameter of only 2.4 m, and corresponding larger dif- nitude limit is compared, we find 131 fraction limit. In addition, the strong diffraction pattern in the NICMOS PSF hampers the de- stars with Hokupa’a, 157 stars with tection of faint sources in the cluster centre. Despite the moderate Strehl ratio, the NACO NICMOS within the same field, and 169 CMD clearly displays the advantages of AO at an 8-m-class telescope. The right axis of the stars with NACO. The 12 stars unde- NACO CMD is labelled with present-day masses from a 2 Myr main sequence isochrone used tected with NICMOS are either fainter to derive the mass function.

11 Figure 4: Sources brighter than 15 mag, the Hokupa’a 90% completeness limit, overlaid on the 7 min NACO K-band image (9 ′ × 9′FOV, left panel). For comparison, the same area on the NICMOS K-band (F205W) image is shown in the right panel. Blue circles are NACO sources and dots are NICMOS detections in both filters. Green, larger circles are Hokupa’a sources. Several close neighbours of bright sources and equal brightness pairs are exclusively detected in the high resolution NACO data.

have been transformed to the NICMOS chosen bin width of ∆log(M/M) = 0.1, The cluster centre MF is shown in system using the colour equations de- resulting in ten MFs with consecutive Figure 5 (right). Only stars within the in- termined by Brandner et al. 2001. A lin- starting points. Averaging the resultant nermost 5″ of the cluster centre are in- ear change in extinction is observed MF slopes yields an MF slope of Γ = cluded. The binning width was now when moving outwards from the cluster –0.91 ± 0.15 for the Arches main se- chosen to be ∆log(M/M) = 0.2 for sta- centre. While the innermost 5″ of the quence, where Salpeter is Γ = –1.35 tistical reasons. Again, only the bright cluster population are consistent with a (Salpeter 1955). Although the NACO end of the MF was fitted for comparison constant reddening of AV = 24.1 mag, LF is more complete, this slope is iden- purposes. The average slope derived the extinction increases with increasing tical to the slope derived from NICMOS for the cluster centre with the same radial distance at radii R > 5″. The ob- data before (Stolte et al. 2002). This in- method as described above is Γ = –0.6 served H-K colour as well as the K- dicates that at the bright end (K < 16 ± 0.2, flatter than the slope for the en- band magnitude have been corrected mag, M > 10 M), all bins contributing tire cluster. This is consistent with mas- for this trend before transformation into to the LF and MF gain the same fraction sive star cluster formation scenarios, stellar masses. A colour cut with 1.35 < of stars, such that the relative fraction where primordial or rapid dynamical (H-K)corr < 1.77 mag was applied to se- of stars in each bin remains the same. segregation causes massive stars to lect Arches main sequence stars from Note that we only consider the massive end up in the cluster centre, while a the corrected CMD. end of the MF, up to K = 16 mag, such larger fraction of low-mass stars is The present-day mass function for that the faint end of the LF, where the found in the outskirts. stars in the Arches cluster derived from largest discrepancy to the NICMOS LF Although the entire mass range in the the NACO CMD (Fig. 3) is shown in is observed, is not taken into account cluster centre appears to be consistent Figure 5. A power law can easily be fit- here, as this would require field sub- with a very flat MF (Γ ∼ 0) down to ted to the upper end of the MF for M > traction. log(M/M) ∼ 0.5, M = 3 M, where the 10 M, below which the MF appears to This slope is slightly flatter than the K-band LF is more than 75% complete become flat. When regarding the entire average slope of Γ = –1.1 ± 0.3 found in (derived from stars recovered in both H NACO field, it appears that at this point star forming regions in the Milky Way and K), it is currently not clear how the field contamination sets in, as this flat (Massey et al. 1995). A comparably flat interplay of incompleteness correction portion of the MF is very pronounced in slope is seen in the MF of the massive and field contamination will alter the MFs created with the above procedure starburst cluster in NGC 3603, which has shape of the MF for masses below 10 at large distances from the cluster cen- physical properties similar to Arches, M. Neighbouring Galactic Centre tre. Analysis of Galactic Centre fields in although forming in a much more mod- fields and thorough incompleteness the vicinity of Arches is required to esti- erate star-forming environment in the testing will be required to obtain a mate the relative fractions of field and Carina spiral arm. In NGC 3603, the MF meaningful slope in the cluster centre cluster stars in the faint regime. As the slope is found to be Γ = –0.7 in the down to the low-mass regime. combined field of view obtained during mass range 1M < M < 30 M Comm 3 does not cover sufficient area (Eisenhauer et al. 1998), close to the 7. Outlook at large enough distances from the Arches MF slope. The flat slope indi- cluster to perform statistical back- cates that the Arches cluster was either Velocity studies of the bright cluster ground subtraction, the MF discussed very efficient in forming massive stars, stars in the vicinity of Arches are nec- here focuses on the massive end with or that dynamical mass segregation essary to draw final conclusions on the M > 10 M, where field contamination has led to the ejection of a significant formation scenario. Stars dynamically is negligible. number of intermediate-mass stars. If ejected during interaction processes To eliminate the dependence of bin- the cluster is mass segregated (either should obtain large velocities directed ning on the derived MF slope, the start- due to primordial or due to dynamical away from the cluster centre. Such a ing point of the first mass bin has been processes), this should be particularly velocity study can yield new insights shifted by δlog(M/M) = 0.01 at the pronounced in the cluster centre. into the cluster dynamics as well as for- 12 Figure 5: Mass Function of the Arches cluster. Left panel: The area within the Gemini FOV has been selected, such that the MF corresponds to the CMD displayed in Figure 3. A linear change in extinction over the field has been corrected, and a colour cut of 1.35 < H-K < 1.77 mag has been applied to the corrected CMD to select Arches main-sequence stars. The Arches main sequence can clearly be identified down to 18.3 mag in K-band, or 2.5 M (logM/M = 0.4), beyond which the bulge population dominates the CMD. Right panel: The mass function of Arches cluster centre stars observed with NACO. Stars within R < 5″ are included. A weighted least-squares fit has been performed on data points above 10 M. The MF in the cluster centre appears flatter than the integrated MF, indicating mass segregation. mer cluster membership of ejected in R < 5″ are 75% complete down to H Brandner, W., Rousset, G., Lenzen, R., stars. High-resolution NIR spectro- = 19.4 mag and K = 17.7 mag or ∼ 3 Hubin, N. et al. 2002, The Messenger No. 107, 1. graphs such as CRIRES expected to M, while HST/NICMOS and Gemini/ Hokupa’a observations were limited to Cotera, A.S., Erickson, E.F., Colgan, S.W.J., go into operation in the next years have Simpson. J.P., Allen, D.A., & Burton, the capability to overcome the huge line ∼10–20 M in the immediate cluster M.G., 1996, ApJ, 461, 750. of sight obscuration towards the centre. While a lower Strehl ratio is re- Devillard, N. 1997, The Messenger No 87, Galactic Centre. The determination of sponsible for the lower performance in 19. spectral types with SINFONI/SPIFFI al- the Hokupa’a data, the NICMOS reso- Eisenhauer, F., Quirrenbach, A., Zinnecker, lows one to disentangle the cluster lution is limited by the smaller HST mir- H., Genzel, R. 1998, ApJ, 498, 278. members from the dense background ror and correspondingly larger diffrac- Figer, D.F., Najarro, F., Gilmore, D., et al. tion limit. As we have shown, even 2002, ApJ, 581, 258. population, thereby yielding accurate Figer, D.F., Kim, S.S., Morris, M., et al. 1999, estimates on the field star contamina- moderate AO performance using faint ApJ, 525, 750. tion when deriving mass functions. With visual reference sources (V = 16 mag) Lejeune, T., & Schaerer, D. 2001, A&A, 366, these prospects in mind, spectroscopic now achieves higher spatial resolution 538. observations may shed light on the un- from the ground while at the same time Lenzen, R., Hoffmann, R. et al. 1998, SPIE solved questions of massive star for- exploiting the light collecting power and 3354, 606. mation in a dense environment. sensitivity of 8-m-class telescopes. Massey, P., Johnson, K.E., Degioia- Thus, a much more complete census of Eastwood, K. 1995, ApJ, 454, 151. Rieke, G.H., & Lebofsky, M.J. 1985, ApJ, 8. Conclusions stellar populations in dense fields in our 288, 618. own Galaxy as well as in nearby re- Rousset, G., Lacombe, F. et al. 2002, SPIE NAOS-CONICA data of the dense solved galactic systems is possible with 4007, 72. stellar field in the Arches cluster centre ground-based AO instruments such as Salpeter, E.E. 1955, ApJ, 121, 161. yield significantly better spatial resolu- NAOS-CONICA. Scalo, J.M. 1998, in ASP Conf. Ser., Vol. tion than NICMOS on HST. About 50% 142, The stellar initial mass function, 201. more stars are resolved in the cluster Schoedel, R., Ott, T., Genzel, R. et al. 2002, Nature 419, 694. centre down to the same magnitude References limit. The luminosity functions reveal Stetson, P.B. 1987, PASP, 99, 191. Stolte, A., Grebel, E.K., Brandner, W., Figer, that each magnitude interval up to the Blum, R. D., Schaerer, D., Pasquali, A., et al. D.F., 2002, A&A, 394, 462. brightest stars was affected by crowd- 2001, AJ, 122, 1875. Tessier, E., Bouvier, J., Beuzit, J.-L., ing losses before. The NACO observa- Brandner, W., Grebel, E. K., Barbá, R. H., et Brandner, W. 1994, The Messenger No. tions of the densest cluster region with- al. 2001, AJ, 122, 858. 78, 35.

Early Galactic Chemical Evolution with UVES M. PETTINI, Institute of Astronomy, Cambridge, UK

The UVES instrument on the ESO and to probe the chemical composition shop served as a focus to bring togeth- VLT2 telescope has now been provid- of gas in the high-redshift universe. In er scientists interested in the general ing high-resolution spectra of faint stars November of last year, a workshop on theme of how the chemical enrichment and galaxies for nearly three years. the theme “Early Galactic Chemical of galaxies progressed over the cosmic European astronomers have taken ad- Evolution with UVES” was held at the ages, from the Big Bang to the present vantage of this new facility to conduct ESO headquarters in Garching. time. The meeting proved to be a valu- extensive and accurate studies of the Attended by more than 50 stellar and able opportunity to exchange ideas, as- first stellar generations of our Galaxy, extragalactic astronomers, the work- sess progress to date, and to chart fu- 13