Galactic structure based on the ATLASGAL 870mum survey
H. Beuther, J. Tackenberg, H. Linz, Th. Henning, F. Schuller, F. Wyrowski, P. Schilke, K. Menten, T. P. Robitaille, C. M. Walmsley, L. Bronfman, F. Motte, Q. Nguyen-Luong, S. Bontemps
aa r X i v : . [ a s t r o - ph . S R ] D ec Accepted for ApJ, draft from November 15, 2018
Preprint typeset using L A TEX style emulateapj v. 03/07/07
GALACTIC STRUCTURE BASED ON THE ATLASGAL 870 µ M SURVEY
H. Beuther , J. Tackenberg , H. Linz , Th. Henning , F. Schuller , F. Wyrowski , P. Schilke , K. Menten ,T.P. Robitaille , C.M. Walmsley ,, L. Bronfman , F. Motte , Q. Nguyen-Luong , S. Bontemps Accepted for ApJ, draft from November 15, 2018
ABSTRACTThe ATLASGAL 870 µ m continuum survey conducted with the APEX telescope is the first surveycovering the whole inner Galactic plane (60 o > l > − o & b < ± . o ) in submm continuum emissiontracing the cold dust of dense and young star-forming regions. Here, we present the overall distributionof sources within our Galactic disk. The submm continuum emission is confined to a narrow rangearound the galactic plane, but shifted on average by ∼ Subject headings:
Stars: formation — Galaxy: structure — dust, extinction — stars: pre-main se-quence — ISM: clouds INTRODUCTION
Since the location of our solar system is within theGalactic disk, studying the Galactic structure of ourMilky Way is always a challenging problem. There-fore, we cannot derive such comprehensive and in-tuitive pictures of our disk as extragalactic stud-ies are able to do for other spiral galaxies (e.g.,Kennicutt et al. 2003; Nieten et al. 2006; Walter et al.2008). Nevertheless, based on a diverse set of stud-ies over all wavelengths, in the last few decades wehave derived a reasonably comprehensive picture of ourGalactic spiral structure (for recent work, see, e.g.,Benjamin 2008; Reid et al. 2009). The Galactic planehas been observed in the optical/near-/mid-infraredbands (e.g., Dobashi et al. 2005; Skrutskie et al. 2006;Churchwell et al. 2009; Carey et al. 2009) as well as atlonger wavelengths, e.g., in CO or cm continuum emis-sion (e.g., Dame et al. 2001; Stil et al. 2006). How-ever, until the arrival of the two (sub)mm Galacticplane surveys ATLASGAL (The APEX Telescope LargeArea Survey of the GALaxy at 870 µ m) and BGPS Max-Planck-Institute for Astronomy, K¨onigstuhl 17, 69117Heidelberg, Germany, [email protected] Max-Planck-Institute for Radiostronomy, Auf dem Hgel 71,53121 Bonn, Germany University of Cologne, Z¨ulpicher Str. 77, 50937 K¨oln, Germany Harvard-Smithsonian Center for Astrophysics, 60 GardenStreet, Cambridge, USA Osservatori Astrofisico di Arcetri, Largo E. Fermi, 5, Firenze,Italy Dublin Institute for Advanced Studies (DIAS), 31 FitzwilliamPlace, Dublin, Ireland Departamento de Astronomia, Universidad de Chile, Casilla36-D, Santiago, Chile Laboratoire AIM, CEA/IRFU - CNRS/INSU - Universit ParisDiderot, CEA-Saclay, 91191 Gif-sur-Yvette Cedex, France Universite de Bordeaux, OASU, Bordeaux, France (Bolocam Galactic Plane Survey) (Schuller et al. 2009;Aguirre et al. 2011), no survey at (sub)mm wavelengthsexisted that trace the cold dust emission stemming fromdense and young star-forming regions at adequate spatialresolution (the COBE (Cosmic Background Explorer)and WMAP (Wilkinson Microwave Anisotropy Probe)data have too coarse resolution to isolate individual star-forming regions). Here, we employ the 870 µ m submmcontinuum survey ATLASGAL to study the general dis-tribution of the dense dust and gas within our Galacticplane. OBSERVATIONS AND SOURCE EXTRACTION
The 870 µ m data are taken from the APEX Tele-scope Large Area Survey of the Galaxy (ATLAS-GAL, Schuller et al. 2009). The 1 σ rms of thedata is ∼
50 mJy beam − and the FWHM ∼ . ′′ .Using the clumpfind source identification algorithmby Williams et al. (1994) with a 6 σ threshold of300 mJy beam − , we identified 16336 clumps within theGalactic plane for longitudes 60 o > l > − o and lati-tudes b < ± . o . In the context of this paper, we are notaiming for exact fluxes, column densities or masses, butwe just want to evaluate source number counts withinthe Galactic plane. Therefore, the specific clump identi-fication algorithm or the used thresholds are not of greatimportance. To test this, we also derived correspondingsource catalogs using 4 σ or 8 σ thresholds. While theabsolute number of sources obviously varies significantlywith changed thresholds, the structural results presentedbelow are not significantly affected by that. As an addi-tional test, instead of deriving clumps, we just extractedthe total submm fluxes above the 6 σ threshold in thegiven latitude and longitude bins. Again the structuraldistributions in longitude and latitude are very similar. Beuther et al.Since other Galactic plane surveys usually also work onsource counts (e.g., GLIMPSE (Galactic Legacy InfraredMidplane Extraordinaire) or MSX (Midcourse Space Ex-periment); Churchwell et al. 2009; Robitaille et al. 2008;Egan et al. 2003), for the remainder of the paper weadopt the 6 σ source catalog. The clump masses rangebetween 100 and a few 1000 M ⊙ , and these clumps formclusters with certain star formation efficiencies. There-fore, the ATLASGAL data largely trace gas/dust clumpscapable of forming intermediate- to high-mass stars atdistances between several 100 out to more than 10000 pc(Schuller et al. 2009; Tackenberg et al. subm.).Tackenberg et al. (subm.) analyzed the ATLASGALdata in the longitude range between 10 and 20 deg indepth via correlating them with the GLIMPSE andMIPSGAL (MIPS Galactic Plane Survey) near- to mid-infrared surveys of the Galactic plane (Churchwell et al.2009; Carey et al. 2009), and at long wavelength with theNH spectral line data from Wienen et al. (subm.). Outof 210 starless clump candidates, Tackenberg et al couldextract NH spectral information – and by that kine-matic distances – for 150 sources. To resolve the kine-matic distance ambiguity, these targets were comparedto the GLIMPSE and MIPSGAL images. Clumps associ-ated with GLIMPSE/MIPSGAL shadows were assignedthe near distance, and the other clumps were assigned asfar. This way, 115 clumps are likely on the near side ofthe Galaxy and 35 on the far side. Tackenberg et al. findthat the mean distances of starless clumps in the lon-gitude range between 10 and 20 deg on the near and farside of the Galaxy are 3.1 and 13.8 kpc, respectively. Oneshould keep in mind that the rotation curve of the innerGalaxy is far from circular (e.g., Reid et al. 2009), mak-ing the absolute determination of kinematic distances adifficult task.For comparison with somewhat more evolved evolu-tionary stages, namely young stellar objects (YSOs),we resort to the Spitzer red source catalog (with acolor criterion of [4 . − [8 . ≥
1, for more detailssee Robitaille et al. 2008). While the intrinsically redsources are contaminated by approximately 30% AGBstars, Robitaille et al. (2008) extracted statistically theseAGB stars and provided a YSO catalog with reducedAGB contamination. This YSO catalog contains 11649sources again over the Galactic longitude range 60 o >l > − o . Because deriving distances and masses for in-dividual YSOs is difficult, Robitaille & Whitney (2010)conducted a population synthesis analysis of the sample.They find that their detected sources consist mainly ofintermediate- to high-mass stars (between 3 and 25 M ⊙ )at distances of several kpc. Again dividing their sourcesin near and far sources with respect to the Galactic cen-ter and Galactic bar, they as well find mean distancesin the longitude range between 10 and 20 deg of 4.9 and13.1 kpc, respectively.Figure 1 presents a histogram of the kinematic dis-tances in the Galactic longitude range between 10 and20 deg derived for the ATLASGAL sample (Tackenberget al. subm.) and population synthesis distances for theGLIMPSE sources (Robitaille & Whitney 2010), respec-tively. Clearly, both distributions show a near and fardistance peak (see also Fig. 9 in Dunham et al. 2011).For the ATLASGAL sources, the far peak has less sourcesbecause, at the given spatial resolution of 19 . ′′ , close Fig. 1.—
Histogram of distances derived for the ATLASGALclumps (black, only the starless clumps) and the GLIMPSE YSOs(red) in the Galactic longitude range from 10 to 20 deg, respectively(Tackenberg et al. subm., Robitaille & Whitney 2010). clumps that would be spatially resolved at the near dis-tance merge and appear as one clump at the far sideof the Galaxy (Tackenberg et al. subm.). To be moreprecise, Tackenberg et al. (subm.) simulated the dis-tance smoothing effect for the ATLASGAL data, andthey found that an artificial sample of 328 clumps at3 kpc distance would appear as only 20 clumps when putto 15 kpc distance (see also the corresponding discussionin Dunham et al. 2011). This effect is far less severe forthe much better resolved GLIMPSE sources. Addition-ally, GLIMPSE sources easier saturate at the near side ofthe Galaxy. Combining these effects with the larger ob-served volume at the far side of the Galaxy, more sourcesare found at the far side for that sample. Although thenear peaks of the ATLASGAL and GLIMPSE samplesare shifted with respect to each other a little bit, takinginto account the inherent uncertainties of kinematic dis-tances for the ATLASGAL sample and population syn-thesis for the YSO sample ( ∼ RESULTS
Figure 2 presents the longitude distribution of thesubmm continuum and YSO sources within our Galac-tic plane (see Schuller et al. (2009) for a first version ofsuch a plot based on a far smaller initial dataset). Whilethe YSO distribution based on the Spitzer data is rela-tively flat, the submm continuum emission shows a seriesof distinctive peaks. The most prominent one is towardthe Galactic center where the source count increases ap-proximately by a factor of 4. In addition to this, thereare a few more clear submm source count peaks at pos-itive and negative longitudes. They can mainly be at-tributed to tangential points of spiral arms (for a dis-cussion of older COBE 240 µ m data, see Drimmel 2000)as well as to a few prominent star formation complexesin the nearby Sagittarius arm. The most important ofthese are marked in Figure 2. For comparison, a sketchof our Galaxy as it would be viewed face-on is shown inFig. 3 where several lines of sight are marked correspond-ing to increased number counts in Fig. 2. For longitudes > −
10 deg, a similar distribution was found in the BGPSsurvey (Rosolowsky et al. 2010). The implications sug-gested by this result will be discussed further in section4. Another way to represent the 870 µ m source distribu-tion is in a 2-dimensional binning in Galactic longitudeand latitude (Fig. 4 top panel). In addition to an increasein source counts toward specific Galactic longitudes, wealso identify a tight confinement to the Galactic mid-plane with only a narrow spread north and south of themid-plane. To derive the approximate latitude for whichthe submm emission peaks in each longitude bin, we fit-ted Gaussians to their latitude distribution in each longi-tude bin. The Gaussian fit peak positions are marked inFig. 4 (top panel). These fits indicate that the dominantdust and gas distribution is slightly shifted to negativeGalactic latitudes with a mean offset over the whole planeof − . ± .
008 deg (the mean values are derived fromGaussian fits to 10 deg longitude bins, see below).We can produce the same plot for the YSO distribu-tion derived from the Spitzer data (Fig. 4 bottom panel).Similarly, the mean value of the peaks of the YSO dis-tribution is also shifted below the Galactic mid-plane,again at − . ± .
008 deg (the mean values are alsoderived from Gaussian fits to 10 deg longitude bins). Al-ready a visual inspection of the two distributions indi-cates that the YSOs appear to cover a broader range inGalactic latitude than the dust and gas clumps tracedby the submm continuum emission. Fitting Gaussiansto the latitude distributions in each longitude bin for thesubmm clumps as well as the YSOs allows us to bet-ter quantify this effect. Since the latitude distributionsare not as smooth on the scales of individual degrees inGalactic longitude, we average over 10 deg in longitudefor smoothing purposes. Figure 5 presents Gaussian ex-ample fits at different Galactic longitudes outlining theapplicability of the Gaussian assumption to these dis-tributions. The corresponding Gaussian full width halfmaximum (FWHM) for the two distributions are shownin Figure 6. One clearly sees that the YSO distribu-tions is broader over the whole Galactic plane than the dense gas and dust distribution. Below Galactic lon-gitudes of -30 deg, the ATLASGAL distribution showsa tendency of increased FWHM. However, this effect isconfined to only three bins, one with a particularly largeerror-bar. Therefore, in the context of this paper we re-frain from further interpretation. The mean values of theFWHM for the dust continuum and YSO distributionsare ∼ ± .
02 and ∼ ± .
02 deg, respectively. Thiscorresponds to characteristic scale heights H (distancewhere distribution has dropped to 1/e, H ≈ × FWHM)of ∼ ∼ DISCUSSION
Longitude distribution
To first order, it appears surprising that the cold dustclump distribution and the YSO distribution do not re-semble each other more closely. While the submm contin-uum emission is optically thin and should therefore tracealmost all cold dust along each line of sight, the YSOdistribution is more strongly affected by extinction. Itmay well be that a considerable fraction of YSOs are notidentified because of too high extinction. Regarding thedifferent properties of the ATLASGAL clumps and YSOs Beuther et al.
Fig. 2.—
Histogram of source number counts with Galactic longitude. The grey-scale shows the ATLASGAL submm continuum sources,and the red histogram presents the YSOs derived from the Spitzer data (Robitaille et al. 2008). The data are binned in longitude in 1 degbins. For ATLASGAL we use the data for latitudes between ± .
25 deg whereas the GLIMPSE YSO data are restricted to latitudes between ± . toward the Galactic center (Fig. 2), a similar emission in-crease toward the Galactic center was recently also foundin the dense gas emission of NH (the HOPS survey,e.g., Walsh et al. 2011, Purcell et al. subm., Longmoreet al. in prep.). The lack of a prominent peak towardthe Galactic center in the GLIMPSE YSO data may bepartly an observational bias but also partly a real phys-ical effect. Observationally, the extinction toward theGalactic center increases which increases the GLIMPSEdetection threshold. However, in contrast to that, alsosurveys of H ii regions, tracing more evolved high-massstar-forming regions, as well as H O and CH OH masersurveys exhibit no strong emission peaks toward theGalactic center (e.g., Wilson et al. 1970; Lockman 1979;Bronfman et al. 2000; Anderson et al. 2011; Walsh et al.2011; Green et al. 2011). Does that imply a relativelylow gas-to-star conversion in that specific part of ourGalaxy? For a more detailed discussion, see Longmoreet al. in prep..Simon et al. (2006) presented a similar plot to ourFig. 2 for the distribution of IRDCs in the Galacticplane. While the general features for the IRDCs ap-pear similar to the dust continuum distribution, the spi-ral arm and star-forming regions are less pronounced.Anderson et al. (2011) performed a similar study of theH ii region distribution of the Galactic plane accessibleto the northern hemisphere, whose results are noticeablydifferent to what we find. The H ii region distribution does not exhibit a peak toward the Galactic center re-gion, but a clear peak is found at approximately +30 deg,corresponding to our peak for the Scutum arm. PreviousH ii region surveys show a similar H ii number increase atnegative longitudes around -30 deg (Wilson et al. 1970;Lockman 1979; Bronfman et al. 2000).Recently, Green et al. (2011) report on the CH OHmaser distribution in the Galactic plane between longi-tude ±
28 deg. Among other source count peaks, in par-ticular they report an increased detection rates at longi-tudes around +25 and −
22 deg, very similar to what wefind. While the peaks at +25 and +31 deg are likely asso-ciated with the end of the long Galactic bar and the be-ginning of the Scutum-Centaurus spiral arm, the −
22 degpeak should be associated with a tangent point of the3 kpc arm (Fig. 3). In summary, the submm contin-uum emission is an excellent tracer of the Galactic densegas structure, even in our Milky Way where our locationwithin the plane complicates the picture so severely.
Latitude distribution
The finding that the average peak of the Gaussian lati-tude distribution is below the Galactic plane has alreadybeen inferred by other groups, e.g., for (sub)mm con-tinuum emission (Schuller et al. 2009; Rosolowsky et al.2010), in the infrared (Churchwell et al. 2006,2007; Robitaille & Whitney 2010), for CO emission(Cohen & Thaddeus 1977), clusters (Mercer et al.alactic structure based on ATLASGAL 5
Fig. 3.—
Sketch of the Galactic plane with several prominentlines of sight marked. Artist impression (by MPIA graphics de-partment) of face-on view of the Milky Way following the Galacticstructure discussed in Reid et al. (2009). ii regions (Lockman 1979; Bronfman et al.2000) or infrared bubbles (Kendrew priv. comm.,Simpson et al. in prep.). Even the Galactic centeritself is located at 0.05 deg below the Galactic plane(Reid & Brunthaler 2004). While a common inter-pretation of that effect is based on a poor definitionof the Galactic plane where neither the sun nor theGalactic center itself are located directly in the planeat 0 deg latitude (e.g., Humphreys & Larsen 1995; Joshi2007; Schuller et al. 2009), Rosolowsky et al. (2010)recently suggested that this effect may also simply becaused by individual star formation complexes and notso much reflect a global Galactic property. However,they state that the offset is mainly a feature of themolecular interstellar medium, whereas different studiesof the GLIMPSE survey indicate that the stellar com-ponent shows the same effect (e.g., Mercer et al. 2005;Churchwell et al. 2006, this study). Although we cannotconclusively differentiate these scenarios, the data hereare indicative of a real global offset of the Galacticmidplane from its conventional position where the axisbetween the sun and the Galactic center are located at b = 0 deg.Are the different Galactic latitude distributions of thesubmm clumps and the YSOs a real physical effect orcould they be caused by observational biases? As out-lined in section 2, the mass distributions and the dis-tances on the near and far side of the Galaxy of thetwo samples are similar. One may now ask whetherthe number of near sources were larger for the YSOsthan for the submm clumps. However, there are sev-eral effects that counteract this: At the relatively coarsespatial resolution of ATLASGAL (19 . ′′ ), clumps that Fig. 4.—
The color-scale shows the two-dimensional source countdistribution for ATLASGAL submm continuum (top panel) andGLIMPSE YSO (bottom panel) sources. The bin sizes in Galacticlongitude and latitude are 1 and 0.1 deg, respectively. The whitelines mark the peak positions of Gaussian fits to the latitude dis-tributions at each given longitude.
Fig. 5.—
Examples of the Gaussian fits to the latitude distribu-tions. The centers of the 10 deg longitude bins are labeled in eachpanel.
Beuther et al.
Fig. 6.—
Resulting FWHM of Gaussian fits (and associatederrors) to the latitude distribution in 10 deg Galactic longitudebins. The black histogram is from the ATLASGAL data, and thered histogram from the GLIMPSE data. would be separate entities on the near side of the Galaxymerge into single objects on the far side, and the to-tal number of sources on the far side is lower than thaton the near side (Tackenberg et al. subm.). In contrastto that, at the higher spatial resolution of Spitzer (2 ′′ ),most sources remain point sources independent of thedistance, and therefore less suffer from the “merging-problem”. Furthermore, the observed volume at the farside of our Galaxy is larger than that on the near side,and GLIMPSE sources on the near side more easily sat-urate. These combined effects even cause a YSO numberincrease on the far side compared to the near side of ourGalaxy (Fig. 1, Robitaille & Whitney 2010). Therefore,the mass and distance distributions of the submm clumpsand YSOs are unlikely to be the cause for the difference inthe latitude distribution. Furthermore, Robitaille et al.(2008) statistically excluded the AGB star populationfrom their catalogue, which implies that contaminationby post-main-sequence sources is not responsible for thedifference in latitude distribution as well. Similar to theeffect seen for the longitude distribution discussed in theprevious section ( § O masers of approximately 0.4 deg,earlier finding a scale height for the CH OH class II masers in a similar range (Walsh et al. 1997). Assum-ing a comparable distance distribution for the maser aswell as the submm continuum sources, the mean physi-cal scale-height of the masers should also be ≈
46 pc. Forultracompact H ii regions, Wood & Churchwell (1989)found a scale height of 0.6 deg, intermediate between ourvalues for the submm clumps and the YSOs. Later,Becker et al. (1994) reported a smaller mean physi-cal scale-height for ultracompact H ii regions of ∼
30 pc (about 40% of the Wood & Churchwell 1989 value),claiming that the Wood & Churchwell (1989) sample isbiased by its large fraction of B-stars. Similar meanphysical scale heights for high-mass star-forming regionsof ∼
44 pc and ∼
29 pc were reported by Bronfman et al.(2000) and Urquhart et al. (2011). Hence masers as typ-ical tracers of star-forming regions (a fraction of the H Omasers may also stem from AGB stars), dust contin-uum emission as a tracer of the dense gas, and younghigh-mass stars exhibit similar scale height distribu-tions in the Milky Way. In comparison, Bronfman et al.(1988) find an approximate scale height of 70 pc for CO(rescaled to a galactocentric solar distance of 8.5 kpc),and Dame & Thaddeus (1994) derive a values of ∼
120 pcfor the thick CO disk (which is an average of their3 cited values). The reported cold HI scale height isaround 150 pc (Kalberla & Kerp 1998; Kalberla 2003;Dedes et al. 2005). Therefore, while tracers of theyoungest evolutionary stages of star formation (submmcontinuum emission and masers) are all found closest tothe Galactic plane, more evolved evolutionary stages likeYSOs as well as the less dense atomic and molecular gasappear to be located on average slightly further from theplane.Using the estimated mean physical scale heights for thedust continuum and YSO distributions of 46 and 80 pc,respectively, we can calculate approximate velocities re-quired to move the 30 pc difference in the given YSO life-times of 1–2 Myrs. This estimate results in required YSOvelocities between 15 and 30 km s − . While velocities ofthat order are found (e.g., PV Cephei, Goodman & Arce2004), they are apparently not the rule. Therefore, wepropose that the most likely explanation for the spreadin scale height for different populations appears to bea combination of extinction effects and dissolving youngclusters from their natal birth sites. CONCLUSIONS