OBSERVERS WITH THE VLT

VLT/ISAAC and HST/WFPC2 Observations of NGC 3603 B. BRANDL1, W. BRANDNER2, E.K. GREBEL3 AND H. ZINNECKER 4*

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

1. Abstract spectral signatures of Wolf-Rayet (W-R) side (Churchwell et al. 1987; O’Dell et al. contribute more than 2000 Mᓃ to 1993). FUV photons (13.6 eV > hν > 6 We have studied NGC 3603, the the cluster mass. Normally, W-R stars are eV) heat up the inside of the proplyd en- most massive visible HII region in the evolved supergiants that have long left velope and lead to the dissociation of mol- , with VLT/ISAAC in the near-in- the main sequence and have ages of 3–5 ecules in the outer layers of the circum- frared (NIR) Js, H, and Ks-bands and Myr. In NGC 3603, however, the W-R stellar disk (Johnstone et al. 1998). The HST/WFPC2 at Hα and [N II] wave- stars also show hydrogen absorption resulting evaporation flow provides a lengths. In this Messenger article we lines in addition to the typical W-R fea- steady supply of neutral atoms to the ion- describe the data analysis and some tures. It is believed that these stars are isation front and leads to the development first results from both our complemen- still main-sequence, core hydrogen- of a cometary tail (McCullough et al. 1995; tary observations. burning stars that are so massive and so Störzer & Hollenbach 1999). Until re- Our HST/WFPC2 gave us an un- close to the Eddington limit that they are cently, only one other proplyd had been precedented high-resolution view of loosing their outer envelopes through fast found outside the Orion . It is lo- the interstellar medium in the giant HII winds, and thus resemble evolved W-R cated in the vicinity of the O7V Her- region and its ionisation structure. stars. In comparison to the Orion Trapez- schel 36 in the Lagoon Nebula (M8, Among the findings are two gigantic ium system, NGC 3603 with its more than Stecklum et al. 1998). gas columns (similar to the famous 50 O and W-R stars producing a Lyman “elephant trunks” in M16) and three pro- continuum flux of 1051 s–1 (Kennicutt 3. VLT/ISAAC Observations plyd-like structures. The emission neb- 1984; Drissen et al. 1995) has about 100 ulae are clearly resolved; all three neb- times the ionising power of the Trapezi- We observed NGC 3603 through the ulae are tadpole shaped, with the bright um cluster. Js = 1.16–1.32 µm, H = 1.50–1.80 µm, 7 ionisation front at the head facing the With a bolometric luminosity Lbol > 10 and Ks = 2.03–2.30 µm broadband filters central cluster and a fainter ionisation Lᓃ, NGC 3603 has about 10% of the lu- using the NIR camera ISAAC on ANTU, front around the tail pointing away from minosity of 30 Doradus and looks in many the first VLT unit telescope. The obser- the cluster. Typical sizes are 6000 A.U. respects very similar to its stellar core vations were made during the 4 nights of × 20,000 A.U. The nebulae share the R136 (Brandl et al. 1996). In fact, it has April 4–6 and 9, 1999, in service mode overall morphology of the proplyds been called a Galactic clone of R 136 but when the optical seeing was equal to, (“PROto PLanetarY DiskS”) in Orion, without the massive surrounding cluster or better than, 0.4″ in 1-minute expo- but are 20 to 30 times larger in size. halo (Moffat, Drissen & Shara 1994). In sures. Such seeing was essential for The VLT observations are the most many ways NGC 3603 and R136 can be accurate photometry in the crowded sensitive near-infrared observations regarded as representative building cluster and increased our sensitivity made to date of a dense starburst blocks of more distant and luminous star- to the faintest stars. The majority of region, allowing us to investigate with un- burst (Brandl, Brandner & Zin- our data were taken under photometric precedented quality its low-mass stellar necker 1999, and references therein). So conditions. population. Our sensitivity limit to stars the best way to study collective star for- Our observing strategy was to use detected in all three bands corresponds mation in a violent environment is by ex- the shortest possible frame times of to 0.1 Mᓃ for a pre-main sequence star amining these regions on a star-by-star 1.77 seconds to keep the number of of age 0.7 Myr. Our observations clear- basis. Until recently, observations of the saturated stars to a minimum. However, ly show that sub-solar-mass stars down entire stellar populations were limited to due to the system’s excellent sensi- to at least 0.1 Mᓃ do form in massive star- the massive stars by sensitivity or wave- tivity, about two dozen of the brightest bursts. length restrictions. The formation of low- stars ended up being saturated. Never- mass stars in starburst regions is of par- theless, this does not impose a prob- 2. Introduction ticular interest. Fundamental questions lem to our study of the low-mass stars. that arise in this context are: Does the Thirty-four short exposures were co- NGC 3603 is located in the spi- slope of the IMF vary on small scales? added to an effective one-minute expo- ral arm (RA = 11h, DEC = –61°) at a dis- Do low-mass stars in a starburst event sure, the minimum time per pointing tance of 6–7 kpc. It is the only massive, form together with the most massive stars required to stabilise the telescope’s ac- Galactic HII region whose ionising cen- or do they form at different times or on tive optics control system. Between the tral cluster can be studied at optical wave- different timescales? And finally, one 1-minute pointings we moved the tele- lengths, due to only moderate (mainly might even ask if low-mass stars form at scope by up to 20″ offsets in a random foreground) extinction of AV = 4–5 mag all in such environments? pattern. This approach has several ad- (Moffat 1983; Melnick et al. 1989). The In addition, NGC 3603 is an ideal lab- vantages: OB stars (> 10 Mᓃ) and stars with the oratory to study individual (!) star forma- • Enlargement of the observed field of tion. Some young stars in the vicinity of view (FOV) with maximum signal-to-noise massive stars are surrounded by par- (S/N) in the cluster centre. *Our collaborators on the project described in this tially ionised circumstellar clouds with • Reduction of residual images and article are Y.-H. Chu, H. Dottori, F. Eisenhauer, A.F.J. Moffat, F. Palla, S. Richling, H.W. Yorke and S. D. cometary shape, so-called protoplanetary other array artefacts, using the median Points. disks (“Proplyds”), ionised from the out- filtering technique.

46 • Derivation of the “sky” from the target exposures using the median filtering technique. No additional time for “blank” sky frames outside the cluster was re- quired. The sky frames have been comput- ed using between 15 and 37 subse- quent exposures per waveband and night, and careful eye-inspection showed that all sources have been efficiently removed using our modified median filtering technique which re- turns the lower 1/3 instead of the mean (1/2) value. We subtracted the sky background and flat-fielded each exposure using the twilight flat-fields provided by ESO. The relative position offsets were derived from cross-correlat- ing the images; the exposures were co- aligned on a 0.5 × 0.5-pixel sub-grid for better spatial resolution, and then added together us- Figure 1: Three-colour image of NGC 3603 composed from Js (blue), H (green), and Ks-band (red) images. Intensi- ing the median fil- ties are scaled in logarithmic units. The field of view is 3.′4 × 3.′4. North is up, East to the left. The insert to the lower tering technique. right is a blow-up of the central parsec2. The resulting im- ages are 3.′4 × 3.′4 in size with pixels of 0.″074. The effective details). About 1′ south of the central observations and the systematic pho- exposure times of the final broadband im- cluster, we detect the brightest mem- tometric errors are dominated by un- ages in the central 2.′5 × 2.′5 are 37, 45, bers of the deeply embedded pro- certainties in the aperture offsets. (A and 48 minutes in Js, H, and Ks, re- tocluster IRS 9. detailed error analysis will be part spectively. In order to derive the photometric of a subsequent paper). Comparing Figure 1 shows the impressive 3-colour fluxes of the stars we used the IRAF our photometric fluxes of numerous composite image that has recently been implementation of DAOPHOT (Stetson sources with the fluxes derived by the subject of an ESO press release. 1987). We first ran DAOFIND to detect Eisenhauer et al. (1998) yields a m (http://www.eso.org/outreach/press-rel/pr- the individual sources, leading to systematic offset of 0.1 in Js and m 1999/pr-16-99.html). The brightest star in Ȃ 20,000 peaks in each waveband. 0.05 in Ks. the FOV (80″ northeast of the core) is the Many of these may be noise or red supergiant IRS 4 (Frogel, Persson, peaks in the nebular background and & Aaronson 1977). To the south of the appear only in one waveband. In order 4. HST/WFPC2 Observations cluster is a giant molecular cloud to reject spurious sources, we required (GMC). Ionising radiation and fast stel- that sources be detected independ- On March 5, 1999 we obtained lar winds from the starburst cluster are ently in all three wavebands, and that deep narrow-band Hα (F656N, 2 × excavating large gaseous pillars. Lo- the maximal deviation of the source 500s) and [NII] (F658N, 2 × 600s) cated about 20″ to the north of the position centroid between different observations. The Planetary Camera cluster centre is the blue supergiant wavebands be less than 0.075″. The (PC2) chip was centred on the bipo- Sher 25. This supergiant is unique be- resulting source list contains 6967 ob- lar outflow structure around the blue cause its circumstellar ring and bipo- jects in the entire FOV. We then flux- supergiant Sher 25. The three Wide lar outflows form an hourglass struc- calibrated the images using the faint Field Camera (WF) chips covered ture similar to that of SN 1987A (Brand- NIR standard stars from the lists by the central cluster and the HII region ner et al. 1997a, 1997b). The image also Hunt et al. (1998) and Persson et al. to the South of Sher 25. In addition, shows the three proplyd-like objects (1998). Because of the stringent re- we retrieved and analysed archival that have recently been discovered by quirements on the seeing, the PSF HST data, which had originally been Brandner et al. (2000) (see Fig. 2 for did not noticeably change during our obtained in July 1997 (PI Drissen). The

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

PC2 was centred on the cluster, and the F656N data in the red channel, 5. Results and Interpretation the three WF chips covered the area the average of F656N and F658N in north-west of the cluster. We combined the green channel, and F658N in the All three proplyds are tadpole shaped individual short exposures in F547M blue channel. The locations of the and rim brightened, with the extended (8 × 30s), F675W (8 × 20s), and F814W proplyds are marked by small boxes, tails facing away from the starburst (8 × 20s) to produce images with effec- and enlargements of the boxes are cluster. The portion of the ionised rims tive exposure times of 240s, 160s, and shown in the upper part of Figure 2. pointing towards the cluster are 160s, respectively. Proplyd 3 has only been observed in brighter than the rims on the opposite The HST/WFPC2 observations are intermediate and broad-band filters side. The central parts of the proplyds presented in Figure 2. The figure and thus stands out less clearly against are fainter than the rims, with a notice- shows an overlay of two composite the underlying background when com- able drop in surface brightness be- colour images. The upper part of the pared to Proplyds 1 and 2. The insert tween the head and the tail. Proplyds 2 image consists of the archive data with at the lower right shows a colour com- and 3 exhibit a largely axisymmetric the following colour coding: F547M posite of HST/WFPC2 F656N (blue) morphology, whereas Proplyd 1, which (blue), F675W (green), F814W (red). and F658N (green) data and VLT/ is also the one closest to the clus- Overlaid are our new WFPC2 data with ISAAC Ks data (red). ter, has a more complex structure. Un-

48 like the convex shape of the heads of the rived from our VLT data. The left plot con- the investigated mass range by about other proplyds, Proplyd 1 has a heart- tains all stars detected in all 3 wavebands one order of magnitude toward smaller shaped head with a collimated, outflow- within the entire FOV of 3.′4 × 3.′4 (6 pc masses and covers also a much larger like structure in between. One possible × 6 pc). Since NGC 3603 is located in the area. explanation for the more complex mor- Galactic Plane we expect a significant Because of Ȃ 10 magnitudes range phology of Proplyd 1 might be that it is contamination from field stars. To reduce in luminosities in the crowded core re- m actually a superposition of two (or maybe this contribution we followed a statistical gion, our sensitivity limits of Js Ȃ 21 , even three) individual proplyds or that approach by subtracting the average H Ȃ 20m and Ks Ȃ 19m don’t appear the photoevaporative flows of several number of field stars found in the regions to be exceedingly faint (and are about disks in a multiple system interact to around the cluster at r > 75″ (central plot) 3 magnitudes above ISAAC’s detection produce this complex single structure. per magnitude and per colour bin (0.5 limit for isolated sources). However, only At distances of 7.4″ and 2.9″ from Pro- mag each). VLT/ISAAC’s high angular resolution, plyd 1 and 2, respectively, faint arc-like The accuracy of our statistical sub- PSF stability, and overall sensitivity en- Hα emission features are seen on the traction is mainly limited by three factors: abled us to study the sub-solar stellar WFPC2 frames. The arcs are located in first, we cannot rule out that low-mass population in a starburst region on a star- the direction of the cluster, and may be pre-main-sequence stars are also pres- by-star basis. the signatures of bow shocks created by ent in the outskirts of the cluster. Sec- the interaction of proplyd winds with the ond, because of crowding on one hand Acknowledgements: We would like winds from the massive stars in the cen- and dithering which leads to shorter to thank the ESO staff, in particular those tral cluster. Additional faint filaments lo- effective integration times outside the involved in the service observations, for cated between the nebulae and the cen- central 2.′5 × 2.′5 on the other hand, the their excellent work. tral ionising cluster can be interpreted as completeness limit varies across the bow shocks resulting from the interaction FOV. Third, local nebulosities may hide of the fast winds from the high-mass stars background field stars. However, none References in the cluster with the evaporation flow of these potential errors affects our from the proplyds. conclusions drawn from the CMD. The Brandl, B., et al. 1996, ApJ, 466, 254. Low-resolution spectra obtained with resulting net CMD for cluster stars with- Brandl, B., Brandner, W. & Zinnecker, H. 1999, EFOSC2 at the ESO/MPI 2.2-m tele- in r < 33″ of NGC 3603 is shown at the Star Formation 1999, eds. T. Nakamoto et scope of the brightest nebula, which is right in the figure on page 1. We over- al., in press. at a projected separation of 1.3 pc from laid the theoretical isochrones of pre- Brandl, B., et al. 1999, A&A Letters, 352, L69. the cluster, reveal that it has the spec- main sequence stars from Palla & Brandner, W., et al. 1997a, ApJ, 475, L45. tral excitation characteristics of an ul- Stahler (1999) down to 0.1 Mᓃ. We as- Brandner, W., et al. 1997b, ApJ, 489, tra compact H II region with electron sumed a distance modulus of (m–M)ᓃ L153. densities well in excess of 104 cm–3. = 13.9 based on the distance of 6 kpc Brandner, W., et al. 2000, AJ, in press (Jan- The near-infrared data reveal a point (De Pree et al. 1999) and an average uary 2000 issue). m source superimposed on the ionisation foreground extinction of AV = 4.5 fol- Churchwell, E., Felliw, M., Wood, D.O.S. & front. lowing the reddening law by Rieke & Massi, M. 1987, ApJ, 321, 516. The striking similarity of the tadpole- Lebofski (1985). De Pree, C.G., Nysewander, M.C. & Goss, shaped emission nebulae in NGC 3603 Applying the isochrones to measured W.M. 1999, AJ, 117, 2902. Dottori, H., et al. 2000, subm. to A&A. to the proplyds in Orion suggests that magnitudes could be misleading since Drissen, L., Moffat, A.F.J., Walborn, N. & the physical structure of both types of ob- the theoretical calculations include only Shara, M.M. 1995, AJ, 110, 2235. jects might be the same. Our hydro- stellar photospheres while the stars still Eisenhauer, F., Quirrenbach, A., Zinnecker, H. dynamical simulations (Brandner et al. may be surrounded by dust envelopes & Genzel, R. 1998, ApJ, 498, 278. 2000) reproduce the overall morpholo- and accretion disks. This would lead to Frogel, J.A., Persson, S.E., Aaronson, M. gy of the proplyds in NGC 3603 very well, excess emission in the NIR and make 1977, ApJ, 213, 723. but also indicate that mass-loss rates of the stars appear younger than they ac- Hunt, L.K., et al. 1998, AJ, 115, 2594. up to 10–5 Mo yr–1 are required in order tually are. Typical excess emission of Johnstone, D., Hollenbach, D. & Bally, J. 1998, to explain the size of the proplyds. Due classical T-Tauri stars in the Taurus- ApJ, 499, 758. Kennicutt, R.C. Jr. 1984, ApJ, 287, 116. to these high mass-loss rates, the pro- Auriga complex have been determined m m McCullough, P.R., Fugate, R.Q., Christou, J.C. plyds in NGC 3603 should only survive as DH = 0.2 , and DK = 0.5 (Meyer, Et al. 1995, ApJ, 438, 394. 5 10 yrs. Despite this short survival time, Calvet & Hillenbrand 1997). The upper Melnick, J., Tapia, M. & Terlevich, . 1989, A&A, we detect three proplyds. This indicates part of the cluster-minus-field CMD 213, 89. that circumstellar disks must be common clearly shows a main sequence with a Meyer, M.R., Calvet, N., & Hillenbrand, L.A. around young stars in NGC 3603 and that marked knee indicating the transition to 1997, AJ, 114, 288. these particular proplyds have only re- pre-main-sequence stars. The turn-on Moffat, A.F.J. 1983, A&A, 124, 273. cently been exposed to their present occurs at J Ȃ 15.5 mag (m Ȃ 2.9 Moffat, A.F.J., Drissen, L., & Shara, M.M. 1994, s ApJ, 436, 183. harsh UV environment. Mᓃ). Below the turn-on the main-se- O'Dell, C.R.O., Wen, Z. & Hu, X. 1993, ApJ, The point source close to the head of quence basically disappears. We note 410, 686. Proplyd 3 is already detected on the that the width of the pre-main se- Palla, F. & Stahler, S. W. 1993, ApJ, 418, broadband HST/WFPC2 observations quence in the right part of the figure on 414. (see Figure 2). The WFPC2 images show page 1 does not significantly broaden Palla, F. & Stahler, S.W. 1999, ApJ, in press that the point source is actually located toward fainter magnitudes, indicating (20 Nov). in front of Proplyd 3 and thus very likely that our photometry is not limited by Persson, S.E. et al. 1998, AJ, 116, 247. not physically associated with it. The in- photometric errors. The scatter may in Stecklum, B., Henning, T., Feldt, M., et al. frared source in Proplyd 1 is also detected fact be real and due to varying fore- 1998, ApJ, 115, 767. as an underlying, heavily reddened con- ground extinction, infrared excess and Stetson, P.B. 1987, PASP, 99, 191. Störzer, H. & Hollenbach, D. 1999, ApJ, 515, tinuum source in the spectrum. A more evolutionary stage. In that case the left 669. detailed analysis of ground based opti- rim of the distribution would be repre- Zinnecker, McCaughrean & Wilking 1993, cal images and spectra of the proplyds sentative of the “true” colour of the Protostars and Planets III, eds. E.H. Levy and other compact nebulae in NGC 3603 most evolved stars. Fitting isochrones & J.I. Lunine (University of Arizona Press), will be presented in Dottori et al. (2000). to the left rim in the CMD yields an 429. The figure on page 1 of this issue of age of only 0.3–1.0 Myr. Our result is The Messenger shows the resulting in good agreement with the study by colour-magnitude diagrams (CMD) de- Eisenhauer et al. (1998) but extends E-mail: [email protected]

49