Age spread in W3 Main: LBT/LUCI near-infrared spectroscopy of the massive stellar content
A. Bik, Th. Henning, A. Stolte, W. Brandner, D. A. Gouliermis, M. Gennaro, A. Pasquali, B. Rochau, H. Beuther, N. Ageorges, W. Seifert, Y. Wang, N. Kudryavtseva
aa r X i v : . [ a s t r o - ph . GA ] S e p Draft version October 4, 2018
Preprint typeset using L A TEX style emulateapj v. 11/10/09
AGE SPREAD IN W3 MAIN: LBT/LUCI NEAR-INFRARED SPECTROSCOPY OF THE MASSIVE STELLARCONTENT A. Bik , Th. Henning , A. Stolte , W. Brandner , D. A. Gouliermis , M. Gennaro , A. Pasquali , B. Rochau ,H. Beuther , N. Ageorges , W. Seifert , Y. Wang , N. Kudryavtseva Draft version October 4, 2018
ABSTRACTWe present near-infrared multi-object spectroscopy and
JHK s imaging of the massive stellar contentof the Galactic star-forming region W3 Main, obtained with LUCI at the Large Binocular Telescope.We confirm 15 OB stars in W3 Main and derive spectral types between O5V and B4V from theirabsorption line spectra. Three massive Young Stellar Objects are identified by their emission linespectra and near-infrared excess. The color-color diagram of the detected sources allows a detailedinvestigation of the slope of the near-infrared extinction law towards W3 Main. Analysis of theHertzsprung Russell diagram suggests that the Nishiyama extinction law fits the stellar population ofW3 Main best ( E ( J − H ) /E ( H − K s ) = 1.76 and R K s = 1.44). From our spectrophotometric analysis ofthe massive stars and the nature of their surrounding H ii regions we derive the evolutionary sequenceof W3 Main and we find evidence of an age spread of at least 2-3 Myr. While the most massivestar (IRS2) is already evolved, indications for high-mass pre–main-sequence evolution is found foranother star (IRS N1), deeply embedded in an ultra compact H ii region, in line with the differentevolutionary phases observed in the corresponding H ii regions. We derive a stellar mass of W3 Mainof (4 ± × M ⊙ , by extrapolating from the number of OB stars using a Kroupa IMF and correctingfor our spectroscopic incompleteness. We have detected the photospheres of OB stars from the moreevolved diffuse H ii region to the much younger UCH ii regions, suggesting that these stars have finishedtheir formation and cleared away their circumstellar disks very fast. Only in the hyper-compact H ii region (IRS5), the early type stars seem to be still surrounded by circumstellar material. Subject headings:
HII regions – infrared: stars – stars: formation – stars: massive – techniques:spectroscopic INTRODUCTION
Despite the impact on their surroundings, the forma-tion and early evolution of massive stars is poorly con-strained, primarily because of their scarcity and shortlifetimes. OB stars are usually observed in very youngstar-forming regions, e.g. clusters and associations,which are still embedded. Near-infrared observations are,in most cases, the only way to study the stellar con-tent of these young regions. However, a pure photomet-ric characterization of young embedded stellar clustersis strongly hampered by highly varying extinction, un- [email protected] Based on data acquired using the Large Binocular Telescope(LBT). The LBT is an international collaboration among insti-tutions in Germany, Italy and the United States. LBT Cor-poration partners are: LBT Beteiligungsgesellschaft, Germany,representing the MaxPlanck Society, the Astrophysical InstitutePotsdam, and Heidelberg University; Istituto Nazionale di As-trosica, Italy; The University of Arizona on behalf of the Arizonauniversity system; The Ohio State University, and The ResearchCorporation, on behalf of The University of Notre Dame, Uni-versity of Minnesota and University of Virginia. Max-Planck-Institut f¨ur Astronomie, K¨onigstuhl 17, 69117Heidelberg, Germany Argelander Institut f¨ur Astronomie, Auf dem H¨ugel 71,53121 Bonn, Germany Astronomisches Rechen Institut, M¨onchhofstrasse 12 - 14,69120 Heidelberg, Germany Max-Planck-Institut f¨ur extraterrestrische Physik, Giessen-bachstrasse 1, 85748 Garching, Germany Landessternwarte K¨onigstuhl, Zentrum f¨ur Astronomie Hei-delberg, K¨onigstuhl 12, 69117 Heidelberg, Germany Purple Mountain Observatory, Chinese Academy of Sciences,210008, Nanjing, PR China known distances and infrared excess of the young clustermembers.On the other hand, spectroscopy offers an unambigu-ous identification of the massive stellar content and hasproven to be a powerful method to find and character-ize the newborn OB stars. Stellar properties, like ef-fective temperature and luminosity, are derived basedon the spectral features and extinction, distance andpossible infrared excess can be determined reliably (e.g.Watson & Hanson 1997; Blum et al. 2000; Hanson et al.2002; Bik et al. 2005; Puga et al. 2006; Bik et al. 2010;Puga et al. 2010). By comparing the effective tempera-ture and luminosity to stellar isochrones, the age as wellas the mass of the stellar population in the young clustercan be derived more reliably than by photometry alone(Bik et al. 2010), as the large extinction variations af-fects the identification of a clearly reddened cluster mainsequence. Apart from the basic properties of the starclusters, also age spread inside the cluster, the evolu-tion of circumstellar disks and the early evolution of theyoung OB stars can be characterized and studied.Analyses of increasing samples of stars in the same starcluster have produced evidence for age spreads in severalcases. Based on the massive star content, Clark et al.(2005); Negueruela et al. (2010) find very little agespread (less than 1 Myr) in starburst cluster Wester-lund 1. However, in the Orion Nebula Cluster age spreadhas been found based on pre-main-sequence (PMS) stars(Da Rio et al. 2010b). Kraus & Hillenbrand (2009) pre-sented evidence for an age spread in Taurus. Additional Bik et al.evidence for an age spread has been found in S255 wherethe most massive star is still deeply embedded and driv-ing an outflow, while the PMS population has an age of1-3 Myr (Wang et al. 2011). Kumar et al. (2006) foundevidence for stellar clusters around high-mass protostel-lar objects, suggesting a similar age spread.Additionally, young stellar clusters are the places tolook for signatures of how massive stars have formed.Remnant accretion disks are being dispersed on a shorttimescale (Hollenbach et al. 1994). However, massiveYoung Stellar Objects (YSOs) have been identified inseveral high-mass star forming regions (e.g. Blum et al.2004; Bik et al. 2006). Studying the frequency of mas-sive YSOs as function of cluster age and mass will allowus to place time scales on the disk evaporation process.Before arriving on the main sequence, low- and inter-mediate mass stars are still contracting and are coolerand bigger than they will be on the main sequence.The intermediate-mass stars are detected as G and K-type stars and will evolve to A and F stars respectively(Bik et al. 2010). Theoretical predictions show that starsmore massive than 10 M ⊙ do not have an optically visiblePMS phase (Palla & Stahler 1990). However, using deepnear-infrared observations we can probe objects deeplyembedded behind up to 50 magnitudes of visual extinc-tion and we might follow the PMS phase of the massivestars.Theoretical modeling of high-mass protostellar evolu-tion predicts that massive protostars might migrate to-wards the main sequence following a path similar to in-termediate mass PMS stars (Yorke & Bodenheimer 2008;Hosokawa & Omukai 2009; Hosokawa et al. 2010). Dueto the high accretion rates, the massive protostars willswell up to a maximum radius of 125 R ⊙ and exhibita relatively cool temperature of ∼ K -band multi-object spec-troscopy as well as JHK s imaging of the massive stellarcontent of W3 Main, obtained with LUCI at the LBT. Weperform for the first time a spectral classification of itsmassive stellar content, allowing the detailed assessmentof the evolutionary status of these stars and their H ii re-gions. The W3 (Westerhout 1958) region is part of an ex-tended star formation complex in the Perseus spiral armand is associated with W4 and W5 (see Megeath et al.2008, for an exhaustive review), spanning an area of 200 ×
70 pc. W3 is located at a distance of 1.95 ± ii re-gions; IC 1795 and NGC 896 of which IC 1795 has anage of 3-5 Myr (Oey 2005; Roccatagliata et al. 2011). Amuch younger part of W3 is still deeply embedded inthe molecular cloud surrounding the OB association andconsists of 3 embedded clusters of compact H ii regions:W3-North, W3-OH and W3 Main. The latter region isthe subject of this paper.Radio continuum observations reveal several com-pact and diffuse H ii regions in the W3 Main area(Wynn-Williams 1971; Harris & Wynn-Williams 1976) as well as ultra-compact H ii (UCH ii ) and hyper-compact H ii (HCH ii ) regions (Claussen et al. 1994;Tieftrunk et al. 1997). The location of the C O(Tieftrunk et al. 1995) and NH (Tieftrunk et al. 1998)molecular emission peaks in a radio quiet area south ofthe region W3 D. Sub-mm dust continuum emission isfound here too, as well as centered on IRS5 (Moore et al.2007).Along with the first detection of the radio contin-uum sources, the first near-infrared imaging revealed sev-eral bright point sources associated with these H ii re-gions (Wynn-Williams et al. 1972). More sensitive imag-ing surveys (Hayward et al. 1989; Tieftrunk et al. 1998;Ojha et al. 2004) have revealed the stellar richness of theW3 Main region and identified all the candidate ionizingsources of the H ii regions via photometry. In the X-rays,where the extinction is much less severe, the entire PMSpopulation is detected, and the cluster shows a sphericalgeometry with a size of 7 pc, about twice as large as seenin the near-infrared (Feigelson & Townsley 2008).The different evolutionary phases of the H ii regionsmake W3 Main an ideal target to study age spread andthe evolution of circumstellar disks around massive stars.The paper is organized as follows; in Section 2 we dis-cuss the LUCI imaging and spectroscopic observationsand data reduction. Section 3 presents the near-infraredphotometry, the spectroscopic classification and the de-rived HRD. In section 4 we discuss the evolutionary stageof the massive stars as well as the detected age spread,and summarize the conclusions in Section 5. OBSERVATIONS AND DATA REDUCTION
Near-infrared observations of W3 Main were obtainedwith LUCI 1 (LBT NIR spectroscopic Utility withCamera and Integral-Field Unit for Extragalactic Re-search), mounted at the Gregorian focus of the LargeBinocular Telescope (LBT) on Mount Graham, Ari-zona (Hill et al. 2006). LUCI 1 (Ageorges et al. 2010;Seifert et al. 2010) is a near-infrared multimode instru-ment operating in seeing-limited imaging, long-slit spec-troscopy as well as multi-object spectroscopy (MOS)mode (Buschkamp et al. 2010).
Imaging observations
W3 Main at α (2000) = 02 h m s , δ (2000) =+62 ◦ ′ ′′ , was imaged in JHK s with LUCI. The J and H data were taken on December 18 and December19, 2009, respectively, while the K s data were obtainedon Feb 5, 2010 with the N3.75 camera, providing a 4 ′ × ′ field of view at an image scale of 0.12 ′′ /pixel.The imaging data were taken with a detector integra-tion time (DIT) of 5 sec and a NDIT (number of integra-tions) of 10. A random dither pattern has been appliedusing 8 up to a maximum of 10 different positions and ajitter box of 30 ′′ . The total exposure times per filter arelisted in Table 1. Sky frames were obtained using thesame instrument settings immediately afterwards, cen-tered on a off-cluster field at α (2000) = 02 h m s , δ (2000) = +62 ◦ ′ ′′ . The final image quality of the ofthe J and H images is 0.6 ′′ and 0.9 ′′ for the K s image. Spectroscopic observations
W3 Main was observed with the LUCI MOS unit in the K -band on December 17 - 19, 2009. To ensure a properge spread in W3 Main 3 TABLE 1Observing details
Observations Date (UT) DIT(s) a Exp time (s) Objects J H K s a Detector Integration Time alignment of the masks for the MOS observations, K -band pre-imaging was acquired on November 01, 2009.Three masks were created to cover the stellar contentranging from the most massive stars to the intermediate-mass PMS stars. We designed the masks such that thedifference in magnitude between the stars in the maskwas around 1 magnitude, ensuring comparable signal-to-noise ratio for all spectra in one mask.We used the 210 zJHK grating in combination with aslit width of 1.0 ′′ , providing us with a resolution R =4000.The spectra were collected with a spatial resolution of0.25 ′′ /pixel (N1.8 camera), providing the largest wave-length coverage (∆ λ = 0.328 µ m). The MOS slits weretypically 4 ′′ long, to allow for a 2 ′′ nodding for sky sub-traction while minimizing the contamination by otherstars.In addition to the MOS spectra, we obtained a longslit spectrum of IRS4 and IRS7 on March 6, 2011 with atotal integration time of 2400 sec. The instrument setupwas the same as for the MOS spectroscopy. For the long-slit observations, the nodding offset was set to 45 ′′ . Thedetails of the observations are shown in Table 1, whichlists the different DITs and total exposure times usedfor the masks and long-slit spectra. The seeing of thespectroscopy data varied between 0.6 ′′ and 0.9 ′′ .Telluric standard stars of early B type were observedimmediately before and after the science frames. Thestandard star was observed using the same mask as thescience observations, as well as being placed in two dif-ferent slits, covering the blue and the red part of thespectral band respectively. Immediately after the sci-ence and standard star observations, arc lamp and flatfields were taken to minimize the effects of flexure. Data reduction
Imaging
The near-infrared
JHK s images were reduced withstandard IRAF routines, as described in Pasquali et al.(2011). The images where corrected for dark currentand flat fielded using sky flats taken during morningand evening twilight. A sky frame was created by com-bining the images taken at the offset position using aweighted mean and rejecting the lowest 3 and highest 8values to remove the stars from the frames. The mas-ter sky frame was subtracted from the individual scienceframes. After background subtraction, the images were IRAF is distributed by the National Optical Astronomy Ob-servatory, which is operated by the Association of Universities forResearch in Astronomy, Inc., under cooperative agreement withthe National Science Foundation. corrected for geometric distortion. An astrometric so-lution was derived using stars in common with 2MASS(Skrutskie et al. 2006). The corrected frames where com-bined and weighted by the average level of the back-ground, producing the final images (Fig. 1).Photometry on the
JHK s images was performed usingthe starfinder software (Diolaiti et al. 2000). This pack-age is designed to fit an empirical point spread function(PSF) to a crowded stellar field. The stellar PSF wasconstructed using 22 bright and isolated sources in theframe. The radial extent of the PSF was set to 40 pixels,beyond the radius at which the PSF structure was dom-inated by background noise. The final PSF was createditeratively after selection and subtraction of neighboringsources from the image. Three iterations were performedbefore the final PSF was extracted. All sources at ≥ σ above the local background noise were included in thedetected source list, and three PSF fitting iterations wereperformed to extract the faintest sources.We verified the results of our starfinder photometry byperforming an additional run with the daophot package(Stetson 1987) within the IRAF environment. daophot treats the background and the PSF model differentlythan starfinder , allowing us thus to check our photomet-ric results from different approaches. Stellar sources weredetected with the daofind task and aperture photometrywas performed with the phot task in radii ∼ × theFWHM of the PSF. For each filter a reference PSF wasconstructed by combining the PSF of at least 20 objectswith the tasks pstselect and psf . PSF-fitting photome-try was performed with the allstar task, using the PSFmodels to fit all objects identied with a 3 σ confidencelevel over the local background. Our results from bothphotometric runs are very similar, however, starfinder isable to better treat the saturated stars in the images.Several of the bright stars in the H ii region appearedsaturated in our JHK s images. The J -band image has7 stars with counts above the linearity limit, while the H and K s images have 8 stars in the non-linear regimeof the detector. To obtain reliable magnitudes for thesebright stars, starfinder was used to fit the wings of thesaturated stars with the determined PSF to reconstructthe flux of the saturated part. However, a few of thebrightest stars could not be recovered by starfinder dueto their heavily saturated cores having negative counts.For those stars, the negative core of the PSF was removedand the peak of the PSF was flattened. The same areawas also flattened in the reference PSF, and the best fitdetermined, such that the differences in the linear partof the star’s and the reference PSF were minimized.This procedure is compared to the recovery done by Bik et al. W3 JW3 KW3 A W3 DW3 HW3 EW3B
IRS N4 IRS7IRS N7IRS2IRS2c IRS2bIRS2a IRS4IRS3a IRS N8IRS N1IRS N5IRS N3 NR559NR386
W3 M,
IRS5IRS N6 IRS N2
W3F W3 CW3 Ca
Fig. 1.—
LUCI
JHK s color image of W3 Main, North is up and East is left. The observed field of view is 4.5 ′ × ′ , correspondingto 2.55 × ii region W3 A. The twodiffuse, spherical H ii regions in the south are W3 K (south-east) and W3 J (south-west). In the center, the deeply embedded massive YSOIRS5 can be found. Annotated are the H ii regions (Tieftrunk et al. 1997) as well as the candidate massive stars of which the spectra arepresented in this paper (in italic). Additionally, the location of the 2 massive stars (NR386, NR559) identified by Navarete et al. (2011)are identified. starfinder on the non-saturated stars allowing the deter-mination of the uncertainties of the used method. Themeasured uncertainties where 0.1 mag for J , 0.07 mag forthe H -band and 0.08 mag for K s . Comparison betweenthe obtained magnitudes and the 2MASS values for theisolated stars shows a match within 0.1 mag for all threebands, suggesting that the above described method de-livers reliable values.Finally, we cross-matched the obtained catalogs foreach filter to identify the sources that are detected inmore than one band. We allowed a maximum off-set of 4 pixels (0.48 ′′ ) for the center of each pointsource. We calibrated the LUCI photometry with2MASS (Skrutskie et al. 2006). The matching radius was4 pixels after transforming the 2MASS positions to theLUCI x,y coordinate system. Only stars with magni-tudes 10 < J < . < H < . < K s < | mag(LUCI) - mag(2MASS) | ≤ J , 49 stars in H and 53 starsin K s as final calibrators, yielding the zero points J zpt =29.205 ± H zpt = 29.072 ± K s , zpt = 28.373 ± starfinder photometrywas derived for the auxiliary datasets. The photometricuncertainties are derived from the difference in magni-tude between the auxiliary datasets. For faint sources,not detected in the auxiliary datasets, the largest error inthe same magnitude range, as derived for other sources,was applied as a conservative error estimate.The final photometric catalog, containing all thesources detected in the 3 bands, has a limiting magni-tude of ∼ J , ∼ H and ∼ K s . A more detailed assessment of the completenessge spread in W3 Main 5limits will be done in a future paper where the entirestellar content of W3 Main will be discussed. Multi-object spectroscopy
The MOS spectra were reduced with a modified versionof lucired , which is a collection of IRAF routines devel-oped for the reduction of LUCI MOS spectra. First theraw frames were corrected for the tilt of the slit using aspectroscopic sieve mask as well as for distortion throughan imaging pinhole mask. After correction, the slits wereoriented horizontally on the frame and were extracted tobe treated separately. The science data were corrected bya normalized flat field. The flat field was normalized inthe spatial direction for every slit to remove illuminationeffects as well as the slit response.The long slit spectra were reduced using standardIRAF routines and no distortion or tilt correction wasapplied. The flat field was normalized by dividing withthe mean of the flat field as the spatial variation in thecounts of the flat field was very low.For both observing modes, the wavelength calibrationwas performed using the Ar and Ne arc wavelength cal-ibration frames. After the wavelength calibration, thesky background was subtracted using the procedure byDavies (2007) to remove the OH line residuals. Finally,the 1D spectra were extracted using the iraf task doslit and combined.The removal of the telluric absorption lines was per-formed in two steps. First the telluric absorption lines inthe standard star spectrum were removed using a high-S/N telluric reference spectrum , providing a first rudi-mentary correction in order to reduce the absorption con-tamination of the telluric standard star. The Br γ absorp-tion line was fitted with a Lorentzian profile. The bestfitting profile of Br γ was removed from the original stan-dard star spectrum, resulting in a atmospheric transmis-sion spectrum taken at the same airmass as the scienceobservations. Finally, the IRAF task telluric was used tocorrect the science observations for the telluric absorp-tion lines. This procedure was applied to the standardstar observations before and after the science observa-tions as well as in the two different slits. The correctionshowing the least telluric residuals was chosen as the finalreduction. Finally, the spectra are normalized by fittinga spline function to the continuum. RESULTS
Photometry
The final reduced 3-color
JHK s image is displayedin Fig. 1. Apart from the stellar content, also severalpatches of extended emission can be seen. These areascoincide with the location of the H ii regions in the radiomaps (Tieftrunk et al. 1997). W3 A is the brightest H ii region in the near-infrared in the north-east of the im-age. The two diffuse H ii regions W3 K and W3 J can beidentified in scattered light in the J -band due to the lowextinction. Several dark dust lanes can be seen as well,showing that the extinction is highly variable.The K s , H − K s color-magnitude diagram (CMD) and J − H , H − K s color-color diagram (CCD) confirm thislarge extinction variation over the observed field (Fig. obtained at NSO/Kitt Peak Observatory K s = 15mag. In the CMD, no clear reddened, cluster main se-quence can be detected. Apart from a foreground mainsequence (at H − K s ≈ H − K s color of the sources in W3 Main is observed,caused by large extinction variations inside W3 Main.The CCD of the bright stars with K s ≤
15 mag confirmsthese huge extinction variations, as the location of thereddened main sequence stars (solid lines) is populatedto at least A K s = 5 mag. Several sources are locatedto the right of the reddened main sequence, due to aninfrared excess, most likely originating in a circumstellardisk (see Fig. 2). Overplotted in the CMD is the loca-tion of the upper end of the un-reddened main sequence 1Myr isochrone (Lejeune & Schaerer 2001), with the spec-tral types annotated as well as the reddening vectorsfrom the Indebetouw et al. (2005, hereafter Ind05) andNishiyama et al. (2009, hereafter NI09) extinction laws.The candidate massive stars in W3 Main are expectedto be located in between the reddening lines departingfrom a B3V and a O3V star respectively (diagonal linesin the CMD).Ojha et al. (2004) identified 13 candidate OB starswhich are bright near-infrared point sources without in-frared excess and associated with H ii regions detected inthe radio (IRS2 – IRS N4 in Table 2). The two candi-date massive stars (IRS4 and IRS7) have only H - and K s detections, but are added to the CMD, and are notplotted in the CCD. IRS5 is only detected in K s . Toidentify massive Young Stellar Objects (YSOs) we addedbright sources possessing an infrared excess and addedtwo more bright sources in the field to complete the tar-get list for the spectroscopic observations (IRS N5 - IRSN8 in Table. 2). Recently, Navarete et al. (2011) pub-lished K -band spectra of 4 sources in W3 Main. Two ofthem (NR386 and NR559) are added to our sample andincluded in our analysis. The two other stars, IRS N3and IRS N4, in common with our sample, are used as across check of the spectral type determination.A comparison between our LUCI photometry andthose of Ojha et al. (2004) and 2MASS shows signifi-cant differences. The LUCI photometry of the blue andisolated sources is fully consistent with the photometryby Ojha et al. (2004) and 2MASS. However, for the redsources (IRS4 and IRS7), as well as sources contaminatedby strong background emission (IRS2a and IRS2b), a sig-nificant difference (in H up to 0.7 mag and in K s up to0.9 mag) is found. All these sources are located insideH ii regions and the lower resolution images of Ojha et al.(2004) as well as 2MASS possibly include more nebularcontribution, altering the magnitudes. Additionally, forthe extremely red sources, a possible small color termin the conversion between the different photometric sys-tems, for either our or the photometry of Ojha et al.(2004), could also contribute to the observed differences.Due to the finer pixel scale and better seeing of our LUCIobservations we consider our LUCI photometry more re-liable.The CMD (Fig. 2) can be used to estimate the spec-tral type of the candidate massive stars by compari-son with the absolute magnitudes predicted by the 1Myr isochrone of Lejeune & Schaerer (2001) and adopt-ing a distance of 1.95 kpc (Xu et al. 2006). The 1 Myr Bik et al. Fig. 2.— L eft: H − K s vs K s color-magnitude diagram of W3 Main. Plotted as squares are the candidate massive stars identified byOjha et al. (2004), in this work and the two stars identified by Navarete et al. (2011). Overplotted as a dashed line is the main sequenceisochrone with an age of 1 Myr from Lejeune & Schaerer (2001) with the location of the spectral types between O3V and B3V are indicated.The two diagonal dotted lines show the reddening lines for an O3V and a B3V star using the extinction law of Ind05. Additionally, thereddening vectors for A K s =1 mag from Ind05, (steep vector) and NI09, (shallow vector) are plotted. R ight: J − H , H − K s color-colordiagram of W3 Main. The plotting symbols and lines are the same as in the CMD. The stars with an infrared excess, located more than3 σ away from the reddened main sequence are plotted as triangles. TABLE 2Catalogue of the massive stars in W3 Main
Object a H ii α (J2000) δ (J2000) J H K s region (h m s) ( ◦ ′ ′′ ) mag mag magIRS2 W3 A 02:25:44.3 +62:06:11.4 12.23 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± b ± ± ± b W3 A? 02:25:47.4 +62:06:55.3 15.04 ± ± ± b ± ± ± b ± ± ± c ± ± ± c ± ± ± a taken from Ojha et al. (2004) b naming convention discussed in main text c Taken from Navarete et al. (2011) isochrone has been chosen as it likely represents the ageof the region better than the ZAMS (see also Sect. 4 fora discussion on the age). Using the relation between T eff and spectral type from Martins et al. (2005) for the Ostars and Kenyon & Hartmann (1995) for the B stars wederived the matching spectral types.The so derived photometric spectral types are verysensitive to the chosen extinction law. Especially forthe high-extinction sources, a different extinction lawchanges the spectral type significantly by up to 3-4 sub-types. We have tested several extinction laws and com-pared them to the observed CCD (Fig. 2). The ex-tinction laws of Rieke & Lebofsky (1985), Ind05 andNI09 provide a reasonable fit to the observed CCD, hav-ing a very similar E ( J − H ) /E ( H − K s ) value of ∼ E ( J − H ) /E ( H − K s ) = 1.21 (Cardelli et al. 1989),and , E ( J − H ) /E ( H − K s ) = 1.52 (Rom´an-Z´u˜niga et al.2007), or too steep slope , E ( J − H ) /E ( H − K s ) = 1.84(Fitzpatrick 1999) of the reddened main sequence. In theCMD and CCD (Fig. 2) we have plotted the extinctionlaw of Ind05.The other parameter of the extinction law, the to-tal over selective extinction ( R λ = A λ /E ( H − K s )),is important when absolute values of A K s are derivede.g. to place the objects in the HRD or derive a pho-tometric spectral type. R K s is difficult to determinefrom JHK s data only. Ind05, Rieke & Lebofsky (1985)and NI09 provide different values for R λ , while hav-ing a similar slope of the near-infrared extinction law.Ind05 derive R K s = A K s /E ( H − K s ) = 1 . ± . Fig. 3.—
Normalized K -band spectra of the massive stars in W3 Main as taken with the multi-object-mode of LUCI. Annotated withdashed lines are the spectral features which are important for classification of the stellar spectra. For the details of the classification seethe main text. Rieke & Lebofsky (1985) find R K s = 1 .
78, while NI09derive R K s = 1 . ± .
01. This parameter can only beconstrained when the absolute magnitude of a reddenedsource is known. The effect of this parameter on theHRD will be discussed in Sect. 4.1.Table 3 shows the resulting photometric spectral typesfor the extinction laws of Ind05 and NI09. For the lowextinction sources, A K s < ii re-gions, other candidate OB stars can be found in theCMD. Navarete et al. (2011) identified one of the blue sources as a O7V star, not related to any H ii region.This would suggest that all the other bright stars couldwell be O stars, while the stars around K s =13 mag and H − K s ≈ σ from the reddening line.The majority of the sources are faint ( K s ≈
14 mag) andare most likely intermediate mass PMS stars showingcircumstellar disk emission. Among the brighter sources,two are spectroscopic targets: IRS3a and IRS N8, withadditionally IRS2a and IRS7 located very close to the3 σ line. The K -band spectra of these sources are thus Bik et al.crucial to establish the true nature of their IR excess. Spectral classification K -band spectroscopy of the candidate massive starsprovides more reliable spectral types, based on the pres-ence of photospheric absorption and emission lines, thanphotometry alone. Additionally, it reveals the nature ofthe infrared excess sources detected in the CCD.The K -band spectra of the massive stars are shown inFig. 3. The majority of the spectra shows Br γ absorptionwith a narrow emission component. The absorption hasa photospheric origin, while the narrow emission line isemitted by the surrounding H ii region. Due to variationsof the nebular emission on small spatial scales, we madeno attempt to correct the spectra for nebular emission.Other lines present in OB star atmospheres are He I µ m, the N III complex at 2.115 µ m (IRS2a and IRS3a)and He II at 2.1185 µ m (IRS2).The LUCI spectra are compared with high-resolutionreference spectra of optically classified O and early Bstars from Hanson et al. (2005) and Bik et al. (2005).Reference spectra of mid- and late-B stars were addedfrom Hanson et al. (1996). The reference spectra weredegraded to the resolution of the LUCI spectra and ar-tificial noise was added to make the spectra comparablein S/N. A visual comparison is performed to determinethe best matching spectral type.As an additional check, we determined the spectraltypes by comparing the measured equivalent widths(EW) of the Br γ , He I and He II absorption lines withthe observations of Hanson et al. (1996) and theoreticalpredictions of Lenorzer et al. (2004). Since the Br γ ab-sorption lines are contaminated by the nebular emission,special care has been taken to measure the EW of Br γ re-liably. We fit a Moffat profile to the absorption wings ofBr γ which are unaffected by the narrow nebular emissionline, providing a more reliable EW determination. TheMoffat fit resembles the observed line profile well and re-produces the line shapes of IRS N3 and IRS N5, whichare unaffected by nebular emission. The two methods de-liver very similar spectral types. Navarete et al. (2011)obtained K -band spectra of IRS N3 and IRS N4, andderived the same spectral types as we derive based onour LUCI spectra.As shown in Fig. 2, four stars with spectra (IRS2a,IRS3a, IRS N7 and IRS N8) show evidence for a near-infrared excess in the CCD. Two of them, IRS N7 andIRS N8, display an emission line spectrum. Additionally,the K -band spectrum of the deeply embedded sourceIRS5 also has these characteristics. These spectra showBr γ emission and IRS N7 also shows the CO bandheademission at 2.3 µ m, most likely originating from dense,hot circumstellar material around the star (Bik & Thi2004; Blum et al. 2004). These emission line spectra, aswell as rising continua, suggest that these objects aremassive YSOs, object surrounded by remnant accretiondisk, possibly as a result of their formation (Bik et al.2006).The other sources with a near-infrared excess, IRS2aand IRS3a, however, do not show a spectrum indica-tive of circumstellar matter. Their K -band spectra showphotospheric absorption and emission lines and they areclassified as O8V-O9V (IRS2a) and O5V-O7V (IRS3a).Possibly, free-free emission of the bright H ii region could be the cause of their IR excess. This shows the impor-tance of spectroscopy to identify the stellar content reli-ably and disentangle the massive YSO candidates fromthe sources contaminated by nebular emission.Veiling by free-free emission makes the absorption linesmore difficult to detect and could increase the uncer-tainty of the spectral type determination. In both IRS2aand IRS3a, the N III complex at 2.115 µ m is detected,narrowing down the spectral type determination to O5V-O9V (Hanson et al. 1996) and making the effect of veil-ing on the spectral type determination less important.Using LUCI MOS spectroscopy, we have determinedthe spectral type of 13 OB stars in W3 Main as well asidentified 3 massive YSOs. Adding the two sources fromNavarete et al. (2011) to our sample results in a total of15 OB stars. The final derived spectral types are given inTable 3 and vary from O5V - O7V for IRS3a to B3V-B4Vfor IRS N6.When we compare the spectral types with those de-rived from the photometry (Sect. 3.1), it turns out thatthe photometry usually predicts earlier spectral types.For some objects, like IRS2a, IRS3a, the spectral typesderived with both methods agree well. However, in manycases the spectral types are rather different, e.g. forIRS2, IRS N1 and IRS N4. In the case of IRS N4, thephotometric spectral type is O7V (for both extinctionlaws), for which the N III emission line at 2.115 µ m andthe He II absorption line at 2.185 µ m are expected. How-ever, in the spectrum of IRS N4 these lines are not ob-served, resulting in a much later spectral type (B0.5V -B2V). HR-diagram
After spectroscopic classification, the objects canbe placed in the HRD. The intrinsic H − K s col-ors as well as the bolometric corrections are takenfrom Martins & Plez (2006) for the O stars andKenyon & Hartmann (1995) for the B stars. First, the E ( H − K s ) is calculated and converted into A K s . For thedetermination of A K s , the two extinction laws (Ind05;NI09) provide different values. In table 3, the valuesof A K s are given for the Ind05 extinction law. To ob-tain the values using the NI09 extinction law, a factor of0.79 × E ( H − K s ) needs to be applied.After accounting for extinction, and applying the dis-tance modulus and the bolometric correction, the lumi-nosities of the massive stars in W3 Main are derived andthey are placed in the HRD (Fig 4). The left panel showsthe HRD with the extinction correction by Ind05 ( R K s = 1.82), while the right diagram shows the correction byNI09 ( R K s = 1.44). The error budget of the luminosityis dominated by the errors on the extinction coefficients,while the error in T eff is given by the uncertainty of thespectral classification, which in turn results in an extrauncertainty in the bolometric correction.Overplotted in the HRDs are the zero age main se-quence (ZAMS) and main sequence isochrones from 1to 3 Myr from Lejeune & Schaerer (2001), the PMSisochrones for intermediate mass stars ( M < M ⊙ ,Siess et al. 2000) as well as the theoretical birth-line (Palla & Stahler 1990). Overplotted also arethe high-mass protostellar evolutionary tracks fromHosokawa & Omukai (2009) for 10 − and 10 − M ⊙ yr − accretion rates.ge spread in W3 Main 9 TABLE 3Spectral types of the massive stars in W3 Main
Object a S/N ratio Photom. Spectro. A K s b Class. a Class. magIRS2 135 O4V/O4.5V O6.5V – O7.5V 2.32 ± ± ± ± ± ± ± ± ± ± ± ± ± > O3V YSO —IRS N8 70 > O3V/O3V YSO —NR559 — O7V/O7V O7V 0.33 ± ± a Spectral types derived with the Ind05 and NI09 extinction laws b Values derived using the Ind05 extinction law, when adopting theNI09 law, the values need to be increased by a factor 0.79 × E ( H − K s ) Fig. 4.—
HRD of the massive stars in W3 Main. The dashed lines represent the main-sequence isochrones from Lejeune & Schaerer(2001) for 1, 2 and 3 Myr. The dotted lines show the PMS evolutionary tracks for a 4 and 5 M ⊙ star (Siess et al. 2000) as well as high-massprotostellar evolution tracks for accretion rates of 10 − and 10 − M ⊙ yr − from Hosokawa & Omukai (2009). The solid line representsthe theoretical birth-line, above which a PMS star is not detectable in the optical (Palla & Stahler 1990). left : The stars are de-reddenedusing the extinction law of Ind05 extinction law. right: The same HRD, but using the extinction correction of NI09. See also discussion inthe text.
L/L ⊙ >
5, where IRS2 and IRS3a are located. The middle areaof the HRD (3.5 < log L/L ⊙ < L/L ⊙ < K s =5.42 ± σ - with their expected main se-quence location. For spectral types between O9V andB1V, where most of the objects are found, the change inT eff and K -band magnitude is large between e.g. O9Vand B0V. However, the the change in observable spectralfeatures is not that dramatic. This is reflected in thelarge error bars of e.g. IRS N5 and IRS7. IRS N1 andIRS N4, however, are the furthest away from the mainsequence. As their extinction is relatively low, their posi-tion does not change much adopting different extinctionlaws. Below we discuss possible scenarios explaining thisoffset.A mis-classification of the spectral type is unlikely.Similarly as described in Sect. 3.2, the observed spectraof IRS N1 and IRS N4 are not compatible with the ex-pected spectral type derived from their luminosity (O7Vand B1V).Another explanation for the offset in the HRD is thatthe sources are at different distances. Calculating thespectrophotometric distance of the stars, assuming thatthey all are on the main sequence, would result in adistance between 1 and 3.4 kpc suggesting that subre-gions of W3 Main would be 1 kpc in the fore- or back- ground. Heyer & Terebey (1998) show that the entireW3/W4/W5 molecular cloud complex forms a connect-ing structure, both in spacial projection and velocity,suggesting that the entire complex is at the same distance(see also Megeath et al. 2008). Therefore, we consider ithighly unlikely that the offsets in the HRD can be ex-plained by different distances for the individual sources.Other effects, however, can be responsible for this offsetin the HRD. Most of the massive stars are observed tobe binary stars (e.g. Bosch et al. 2001; Apai et al. 2007).An equal mass binary would result in a mismatch in theobserved luminosity by a factor of 2 (∆ log L/L ⊙ = 0.3dex), with a multiple star system increasing the offseteven more. Both sources would have to be a system of5-6 equal mass stars to explain their location above themain sequence, though.Stellar evolution can also be the reason why these ob-jects are located away from the main sequence. The loca-tion of IRS N4 would be compatible with a 10 Myr mainsequence track, while IRS N1 would require an even olderisochrone. This is highly unlikely, in particular for IRSN1, as the star is located inside an UCH ii region, whichis still very young ( ∼ yrs, Wood & Churchwell 1989).IRS N4 is the central star of a diffuse and more evolvedH ii region; as shown by Lada & Lada (2003) a timespanof 10 Myr would be enough to disperse the surroundinggas.Another possible explanation for the HRD loci of thestars is that they are still in the PMS phase. Abovethe “birth-line” no optically visible PMS stars are ex-pected (Palla & Stahler 1990). However, using deepnear-infrared observations we might start to probe thePMS phase of massive stars. Theoretical modeling of theprotostellar evolution predicts that high-mass protostarsmight have a similar evolution to intermediate PMS starstowards the main sequence (Hosokawa & Omukai 2009).The exact path of the protostars in the HRD dependson the accretion rate. Overplotted in Fig. 4 are twotracks for accretion rates of 10 − M ⊙ yr − and 10 − M ⊙ yr − . The location of IRS N1 would be consistent withan accretion rate between 10 − and 10 − M ⊙ yr − , whileIRS N4 would require a slightly higher accretion rate.The location of IRS N1 and IRS N4 are consistent withthe very last stages of the PMS evolution where the starsare contracting to their main-sequence temperature andluminosity. In this contraction phase, the accretion lumi-nosity only contributes 20% in the case of the 10 − M ⊙ yr − and 10% 10 − M ⊙ yr − track respectively. The ab-sorption lines in the K-band spectra of IRS N1 and IRSN4 as well as the lack of near-infrared excess are con-sistent with the stellar contribution being the dominantsource of luminosity.Surprising for those young objects however, is the lackof infrared emission coming from a circumstellar disk.The presence of a disk would make these objects similarto e.g. IRS N7 or IRS N8, which show an emission linespectrum and a near-infrared excess. An explanation ofthe lack of disk emission could be the fast dispersal ofthe disk by FUV photons of the central star or externalFUV photons from other cluster members in W3 Main(see Section 4.1).Summarizing, the spectral classification of the mas-sive stars in W3 Main allows the comparison with stellarevolution models in the HRD. The location of the mostge spread in W3 Main 11massive stars in the upper HRD suggests that IRS2 hasan age of 2-3 Myr and IRS3a is slightly younger, whilea speculative explanation of the displacement of IRS N1and IRS N4 in the HRD could be high-mass PMS evo-lution. However, due to the large impact a change ofextinction law has on the appearance of the HRD, it isdifficult to assess with confidence that these stars are inthe PMS phase. To prove this hypothesis, a good deter-mination of the stellar surface gravity, by means of mod-eling the absorption lines of high signal-to-noise spectrais needed to show that they have larger radii than mainsequence stars. Energy budget of the H ii regions Apart from IRS N6, NR559 and NR386, all OB starsin W3 Main are located in H ii regions detected at radiowavelengths (Tieftrunk et al. 1997). Assuming the radioemission to be optically thin, we recalculated the numberof ionizing photons responsible for the radio flux usingthe distance of 1.95 kpc instead of 2.3 kpc quoted byTieftrunk et al. (1997). With Martins et al. (2005) andSmith et al. (2002), the number of ionizing photons areconverted in a spectral type of a single star that would beresponsible for the ionization of the H ii region (Table 4).These spectral types are compared to the spectral typesof the OB stars inside the H ii regions. In the case ofW3 A (radio spectral type O6V), 5 OB stars are foundin a single H ii region. Nevertheless, IRS2, the earliesttype star, is the main source of ionization as the numberof ionizing photons drops very fast with spectral type(Martins et al. 2005). IRS2 is of spectral type O6.5V –O7.5V emitting 10 . − . Lyman continuum photons,while the 2nd most massive star, IRS2a (O8V – O9V)only emits 10 . − . ionizing photons.For the H ii regions W3 A (O6.5V), W3 B (O7.5V), W3C (O9.5V), W3 E (B1V) and W3 F (B0V), the most lu-minous stars provide sufficient Lyman continuum flux toionize these H ii regions. For the diffuse H ii regions W3 H(O9V), W3 J (O8.5V) and W3 K (O7.5V), on the otherhand, the Lyman continuum flux of the brightest sourceis not sufficient to explain the radio flux. IRS N3 (B0V– B2V) in W3 J and IRS N4 (B0.5V – B2V) in W3 Kare the brightest stars, located in the center of these H ii regions, and therefore are expected to be the main ion-izing sources. However, the expected spectral type fromthe radio flux is much earlier and inconsistent with theobserved spectrum of the ionizing sources. Interestingly,the photometric spectral type of IRS N4 (O7V) agreesbetter with that expected from the radio flux of W3 K.However, as discussed in Sect. 3.2, a O7V spectrum isnot compatible with the the observed spectrum of IRSN4, so a miss classification of the spectrum is not likely.The spectroscopic classification of IRS N3 and IRS N4 byNavarete et al. (2011) confirms our spectral classificationbased on a different data set.Leakage of FUV photons or dust attenuation would beable to explain why the ionizing star would be of earlierspectral type (e.g. Kurtz et al. 1994), but will not resultin the fact that the observed radio emission cannot beaccounted for by the ionizing star. Smith et al. (2002)show that the number of ionizing photons is much higher(log Q = 47.8 s − ) for a B1 super giant instead of log Q = 46.5 s − for a B1 dwarf. This would come closer to the value derived from the radio emission. However,their position in the HRD (Fig 4) does not comply withthis explanation, since these stars are both too faint tobe early B supergiants, which would have a luminosityof ∼ . L ⊙ (Martins et al. 2005), while the observedluminosity is ∼ . L ⊙ (IRS N3) and ∼ . L ⊙ for IRSN4. The presence of supergiants is also not consistentwith the derived age of W3 Main. Cluster mass
Determinations of the mass of embedded regions likeW3 Main are hampered by the extreme extinction vari-ations observed in the cluster (Fig. 2). No clearly red-dened main sequence is visible which can be comparedto main sequence isochrones directly (e.g. Ascenso et al.2007; Brandner et al. 2008). However, the spectroscopyof the massive stars results in a direct mass estimate forthe stars of the upper end of the mass function.The masses of the stars were calculated by compar-ing the stellar luminosities to the 1 Myr main sequenceisochrone. Only for IRS2, the 3 Myr isochrone is used asit already evolved away from the location of the 1 Myrisochrone. This results in stellar masses of 35 M ⊙ forIRS2 as the most massive star and 5 M ⊙ for IRS N6, thelowest mass star in our sample.In order to obtain an extinction-limited sample, whichis complete to the highest extincted massive star we havedetected with spectroscopy, we took the mass of our red-dest OB star, IRS 7 (11 M ⊙ ) as our lower mass limitwhere we are complete. In this way, we select 13 outof 15 OB stars studied in this paper. Inspecting theCMD shows an additional 12 sources above this masslimit without spectroscopic classification. Navarete et al.(2011) identified one of the bright stars, not related toany H ii region, as a massive star, this suggests that theother stars might well be members of W3 Main. As-suming all these 12 sources are part of W3 Main, weend up with a total of 25 stars between 11 and 35 M ⊙ .This number does not include the deeply embedded starsaround IRS5 as they have even higher extinction. Addi-tionally, our observations only cover the central 2.5 pc × ii regions tracing the massive stars are all locatedin the central 2.5 pc. We calculate the mass of the clus-ter assuming that we are complete in the detection of themassive stars with A K s ≤ M ⊙ and -0.3 for massesbetween 0.08 and 0.5 M ⊙ and the total number of starsbetween 11 and 35 M ⊙ (25 stars), we can extrapolate themass function and obtain an estimate of the total stellarmass in the cluster. For the extrapolation we carried outMonte Carlo simulations assuming a randomly populatedIMF as described in Brandner et al. (2008), resulting ina total mass of (4 ± × M ⊙ .The quoted uncertainty is only due to the effects ofrandom sampling. When only the stars where spec-troscopy has confirmed their cluster membership are used(15 stars), then the derived mass is about a factor of twolower. Additionally, binaries are not taken into accountas well as the deeply embedded population like IRS5.This would make the real mass likely higher. With 40002 Bik et al. TABLE 4Ionizing content of the Hii regions in W3 Main H ii log N Ly Radio Stars NIR log Q region s − Sp. type Sp. type s − W3 A 48.93 O6V IRS2 O6.5V – O7.5 48.44 – 48.80W3 B 48.43 O7.5V IRS3a O5V – O7V 48.63 – 49.26W3 C 47.46 O9.5V IRS4 O8V – B0.5V 47.00 – 48.29W3 E 46.50 B1V IRS N1 B2V – B3V ≤ ≤ ≤ M ⊙ , W3 Main is about twice as massive as the OrionNebula Cluster (O’Dell 2001).The total mass of the W3 GMC is estimated to be5 × M ⊙ based on CO observations (Dickel 1980), butencompasses more than just W3 Main alone. When look-ing at just the molecular gas towards W3 Main, Dickel(1980) derive a minimum mass for the high-density gastowards the W3 Main region of 5 × M ⊙ , which is con-firmed by C O observations (Thronson 1986). To-gether with our mass estimate of 4 × M ⊙ , this wouldresult in a estimate of the star-formation-efficiency (SFE)of ∼
44 %. Note that the southern part of the W3 Mainregion most likely already evacuated the gas, and there-fore this should be considered as a rough upper limit.Such high value of the SFE is consistent with the pre-diction of Bonnell et al. (2011) for clusters forming inthe gravitationally bound regions of a molecular cloud(SFE ∼
40 %). Whether the cluster will remain bound ifthe gas is removed (Portegies Zwart et al. 2010) remainsuncertain and cannot be constrained with our currentobservations. DISCUSSION massive Young Stellar Objects
Apart from the 15 OB stars, 3 massive YSOs have beendetected. Their spectra are dominated by circumstellaremission and no spectral type can be derived from their K -band spectra. However, based on their location anddetection at other wavelengths, we can derive some upperlimits to their luminosity. IRS5 is located in a HCH ii re-gion and is a small multiple system (Megeath et al. 2005;van der Tak et al. 2005; Rod´on et al. 2008) with a totalluminosity of 3 × L ⊙ ionized by a handful of earlyB stars (Campbell et al. 1995). The two other massiveYSOs in our sample, IRS N7 and IRS N8 are not as-sociated with a detectable H ii region. Also, they haveno detectable sub-mm counter part. IRS N7 is locatedat the southern edge of the sub-mm clump centered onIRS5, while IRS N8 is located east of the clump centeredon IRS4 (Moore et al. 2007).In contrast to the very embedded IRS5, both sourcesare detected in JHK s and therefore less embedded andprobably more evolved than IRS5. The lack of radioemission would suggest that they are mid- or late B stars,as for more massive stars, an H ii region would be ex-pected. Similar objects have been detected in other starforming regions (Hanson et al. 1997; Bik & Thi 2004;Blum et al. 2004; Bik et al. 2006), but no clear reasonhas been identified for why these stars are still sur-rounded by circumstellar matter, while other OB stars already cleared their surroundings. Bik et al. (2006)showed that these objects have similar spectral and pho-tometric properties to Herbig AeBe stars, which are stillsurrounded by circumstellar disks even when they be-come optically visible.We detected OB stars in the diffuse, compact andUCH ii regions. Only in the HCH ii region W3 M, cir-cumstellar emission from IRS5 has been detected in the K -band spectrum. This shows that the accretion disksaround the massive stars in W3 Main are destroyedwithin 10 yrs. When an HCH ii region evolves to theUCH ii phase, circumstellar disks are destroyed by thestrong FUV field and stellar winds of the recently formedmassive stars (Hollenbach et al. 1994). Evolutionary sequence in W3 Main
W3 Main harbors several different evolutionary stagesof H ii regions, ranging from very young HCH ii re-gions (few 10 yrs), UCH ii regions ( ∼ yrsWood & Churchwell 1989) to evolved, diffuse H ii regions(few 10 yrs). All these regions are most likely formedout of the same molecular cloud. This provides thepossibility to study the evolution of young H ii regionsand their stellar content in great detail. Tieftrunk et al.(1997) derived an evolutionary sequence for the H ii re-gions in W3 Main based on the morphology of the radiosources. The youngest are the HCH ii regions W3 M andW3 Ca, with the UCH ii regions W3 F, W3 C and W3E, slightly older, the compact H ii regions W3 B and W3A even more evolved, and the diffuse H ii regions, W3 Kand W3 J being the oldest H ii regions in W3 Main. Thisclassification can be compared to the ages of the massivestars deduced from their position in the HRD.We have detected OB stars in three diffuse H ii regions,two compact H ii regions and three UCH ii regions withHCH ii region W3 M harboring the high-mass protostarIRS5 (see Sect. 4.1). The position of IRS2 in the HRDsuggests an age of 2-3 Myr, consistent with its locationin a relatively evolved compact H ii region (W3 A) witha similar or younger age derived for IRS3a, harbored ina younger UCH ii region. For the lower-mass stars (lateO, early B) the isochrones are too close to each otherto derive any age information. For IRS N1, located in-side the UCH ii region W3 E, the offset from the mainsequence could be explained by high-mass PMS evolu-tion (Hosokawa & Omukai 2009), consistent with the ex-pected young age of the UCH ii region.A similar sequence can be seen in the extinction to-wards the different H ii regions. The extinction variesfrom A K s = 0.9 mag for the diffuse H ii regions W3 J andW3 K to very high extinction ( A K s = 5.9 mag) towardsge spread in W3 Main 13W3 F. Only IRS3a, located in the compact H ii regionW3 B does not follow the trend and shows a very highextinction (A K s = 5.4 mag). Such a sequence in extinc-tion indicates that the stars in diffuse H ii regions havealready cleared out their surroundings and destroyed themolecular cloud, while those in UCH ii regions are stillembedded in their parental molecular cloud.Based on the presence of different evolutionary stagesof H ii regions as well as the location of the most massivestars in the HRD, we can conclude that an age spread ofa few Myr is most likely present for the massive stars inW3 Main. The presence of an older population of a fewMyr is supported by the detection of several 100s classIII stars in the X-rays by Feigelson & Townsley (2008).These authors suggest that, at least some, of the OBstars have formed after the lower-mass PMS stars. Aspectroscopic analysis of a sample of intermediate-massPMS stars will provide more insights on this suggestion(Bik et al, in prep).A large number of young stellar clusters show evi-dence for an age spread, usually based on the analysisof the PMS population in the HRD. In Orion an agespread of a few Myr has been found (Palla & Stahler1999; Da Rio et al. 2010b), as well as in the LMC in theyoung stellar cluster LH95 (Da Rio et al. 2010a). In star-burst clusters, however, upper limits on the age spreadof less than 1 Myr have been found for Westerlund 1(Clark et al. 2005; Negueruela et al. 2010) based on themassive stars. Similar results are found for the PMSpopulation of NGC 3603 and Westerlund 1 (Kudryavt-seva et al, in prep). This suggests that W3 Main, similarin mass to the Orion Nebula cluster, is not formed inone star formation burst as expected for starburst clus-ter, but, more likely, through a temporal sequence of starformation events. CONCLUSIONS
We have obtained deep
JHK s near-infrared imaging aswell as K -band multi-object spectroscopy of the massivestellar content of W3 Main using LUCI at the LBT. Wehave confirmed the nature of 15 candidate OB stars bymeans of their absorption line spectra in the K-band.Three additional sources are identified as massive YoungStellar Objects based on their emission line spectra.From the analysis of the CCD and CMD, the extinc- tion laws of Ind05 and NI09 provide the best fit of theslope of the reddened main sequence. Using either of thethese two extinction laws the OB stars are placed in theHRD, with the NI09 providing a better match betweenthe derived positions of the stars and the main sequence.Our most massive star, IRS2 (35 M ⊙ ) is alreadyevolved away from the ZAMS and its location is con-sistent with an age of 2-3 Myr. Additionally, we findevidence for high-mass PMS evolution in the displace-ment of IRS N1 and IRS N4 from the main sequence.A total cluster mass of (4 ± × M ⊙ has been de-rived by extrapolating the number of massive stars usinga Kroupa IMF.The evolutionary sequence of the H ii regions seen inthe radio morphology is consistent with a increase inextinction from the older (diffuse H ii ) regions to theyounger (UCH ii ) regions. We have detected the photo-spheres of OB stars from the more evolved diffuse H ii re-gion to the much younger UCH ii regions, suggesting thatthe OB stars have finished their formation and clearedaway their possible circumstellar disks very fast. Onlyin the HCH ii phase (IRS 5), the massive star is still sur-rounded by circumstellar material.The above properties of the stellar content and H ii regions present evidence of an age spread of at least 2-3 Myr in W3 Main, similar to what has been observedin Orion and other young open clusters and in contrastto the single age population found in starburst clusters.This suggests a sequential star formation mode being atwork for W3 Main, instead of a single star formationevent.We thank the anonymous referee for useful comments,helping to improve the paper. We thank the LBTO stafffor their support during the observations. We thankJaron Kurk and Neil Crighton for obtaining the long slitspectra of IRS4 and IRS7. We thank Giovanni Cresciand Jaron Kurk for providing the lucired pipeline andTakashi Hosokawa and Rolf Kuiper for the high-massPMS tracks. A.B thanks Angela Adamo for reading themanuscript. D.A.G. kindly acknowledges the GermanAerospace Center (DLR) and the German Federal Min-istry for Economics and Technology (BMWi) for theirsupport through grant 50 OR 0908. Facilities:
LBT(LUCI I).
REFERENCESAgeorges, N., Seifert, W., J¨utte, M., Knierim, V., Lehmitz, M.,Germeroth, A., Buschkamp, P., Polsterer, K., Pasquali, A.,Naranjo, V., Gemperlein, H., Hill, J., Feiz, C., Hofmann, R.,Laun, W., Lederer, R., Lenzen, R., Mall, U., Mandel, H.,M¨uller, P., Quirrenbach, A., Sch¨affner, L., Storz, C., & Weiser,P. 2010, Proc. SPIE, 7735, 53Apai, D., Bik, A., Kaper, L., Henning, T., & Zinnecker, H. 2007,ApJ, 655, 484Ascenso, J., Alves, J., Beletsky, Y., & Lago, M. T. V. T. 2007,A&A, 466, 137Bik, A., Kaper, L., Hanson, M. M., & Smits, M. 2005, A&A, 440,121Bik, A., Kaper, L., & Waters, L. B. F. M. 2006, A&A, 455, 561Bik, A., Puga, E., Waters, L. B. F. M., Horrobin, M., Henning,T., Vasyunina, T., Beuther, H., Linz, H., Kaper, L., van denAncker, M., Lenorzer, A., Churchwell, E., Kurtz, S.,Kouwenhoven, M. B. N., Stolte, A., de Koter, A., Thi, W.-F.,Comeron, F., & Waelkens, C. 2010, ApJ, 713, 883Bik, A., & Thi, W.-F. 2004, A&A, 427, L13 Blum, R. D., Barbosa, C. L., Damineli, A., Conti, P. S., &Ridgway, S. T. 2004, ApJ, 617, 1167Blum, R. D., Conti, P. S., & Damineli, A. 2000, AJ, 119, 1860Bonnell, I. A., Smith, R. J., Clark, P. C., & Bate, M. R. 2011,MNRAS, 410, 2339Bosch, G., Selman, F., Melnick, J., & Terlevich, R. 2001, A&A,380, 137Brandner, W., Clark, J. S., Stolte, A., Waters, R., Negueruela, I.,& Goodwin, S. P. 2008, A&A, 478, 137Buschkamp, P., Hofmann, R., Gemperlein, H., Polsterer, K.,Ageorges, N., Eisenhauer, F., Lederer, R., Honsberg, M., Haug,M., Eibl, J., Seifert, W., & Genzel, R. 2010, Proc. SPIE, 7735,236Campbell, M. F., Butner, H. M., Harvey, P. M., Evans, N. J. I.,Campbell, M. B., & Sabbey, C. N. 1995, ApJ, 454, 831Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345,245Clark, J. S., Negueruela, I., Crowther, P. A., & Goodwin, S. P.2005, A&A, 434, 9494 Bik et al.