The initial conditions of stellar protocluster formation. II. A catalogue of starless and protostellar clumps embedded in IRDCs in the Galactic longitude range 15<l<55
A. Traficante, G. A. Fuller, N. Peretto, J. E. Pineda, S. Molinari
aa r X i v : . [ a s t r o - ph . S R ] J un Mon. Not. R. Astron. Soc. , 1–19 (2011) Printed 12 July 2018 (MN L A TEX style file v2.2)
The initial conditions of stellar protocluster formation. II. Acatalogue of starless and protostellar clumps embedded in IRDCs inthe Galactic longitude range ◦ l ◦ A. Traficante ⋆ , G.A. Fuller , N. Peretto , J.E. Pineda , and S. Molinari Jodrell Bank Centre for Astrophysics, School of Physics and Astronomy, University of Manchester, Oxford Road, Manchester M13 9PL, UK School of Physics and Astronomy, Cardi ff University, Queens Buildings, The Parade, Cardi ff CF24 3AA, UK Max-Planck-Institut fr extraterrestrische Physik (MPE), Germany IAPS - INAF, via Fosso del Cavaliere, 100, I-00133 Roma, Italy
ABSTRACT
We present a catalogue of starless and protostellar clumps associated with infrared dark clouds(IRDCs) in a 40 ◦ wide region of the inner Galactic Plane ( | b | ◦ ). We have extracted thefar-infrared (FIR) counterparts of 3493 IRDCs with known distance in the Galactic longituderange 15 ◦ l ◦ and searched for the young clumps using Hi-GAL, the survey of theGalactic Plane carried out with the Herschel satellite. Each clump is identified as a compactsource detected at 160, 250 and 350 µ m. The clumps have been classified as protostellar orstarless, based on their emission (or lack of emission) at 70 µ m. We identify 1723 clumps,1056 (61%) of which are protostellar and 667 (39%) starless. These clumps are found within764 di ff erent IRDCs, 375 (49%) of which are only associated with protostellar clumps, 178(23%) only with starless clumps, and 211 (28%) with both categories of clumps. The clumpshave a median mass of ∼ ⊙ and range up to > M ⊙ in mass and up to 10 L ⊙ inluminosity.The mass-radius distribution shows that almost 30% of the starless clumps identified inthis survey could form high-mass stars, however these massive clumps are confined in only ≃
4% of the IRDCs. Assuming a minimum mass surface density threshold for the formation ofhigh-mass stars, the comparison of the numbers of massive starless clumps and those alreadycontaining embedded sources suggests an upper limit lifetime for the starless phase of ∼ years for clumps with a mass M >
500 M ⊙ . The star formation process begins in massive clouds with a reser-voir of gas and dust su ffi cient to sustain the creation of a clusterof stars (e.g. Lada & Lada 2003), which in some clouds may even-tually form high-mass stellar objects (e.g., Rathborne et al. 2006;Peretto & Fuller 2009). Some of these dense, cold clouds whichare not yet dominated by star formation absorb the IR emission ofthe background radiation and therefore can be observed as darkstructures in the mid-IR images (infrared dark clouds, IRDCs,e.g. Simon et al. 2006). These relatively undisturbed clouds arefavoured places to study the very early stages of star formation(Peretto & Fuller 2009, 2010; Battersby et al. 2011; Ragan et al.2012).Within these clouds, the earliest stages are initially charac-terised by the formation of a starless, dense clump with a size of ≃ / sub-mm region of the spectrum, but are still IR-quiet at wave-lengths µ m (Elia et al. 2010; Motte et al. 2010; Giannini et al.2012). As a protostellar core and then protostar eventually formswithin the gravitationally bound starless clumps (the pre-stellar clumps) it heats the surrounding envelope. The now protostellarclump becomes visible at wavelengths µ m and its bolomet-ric luminosity increases, sustained by the warm inner core(s). As a consequence, the protostellar luminosity correlates well with theclump emission observed at 70 µ m (Dunham et al. 2008).Large surveys of the Galactic Plane in the FIR-sub-mm regionallow us to make a census of these clumps and the young stellar ob-jects (YSOs) within them across the Galaxy. Such surveys providethe significant samples of sources required to understand the starformation mechanisms across a wide range of mass and luminosityregimes as well as identify rare and / or short-lived classes of ob-jects. Several surveys have been carried out in the past few years atincreasing sensitivity and spatial resolution in order to build a sta-tistically significant sample of young stellar objects in the GalacticPlane. The APEX Telescope Large Area Survey of the Galaxy (AT-LASGAL, Schuller et al. 2009) and the Bolocam Galactic PlaneSurvey (BGPS, Aguirre et al. 2011) have mapped the inner Galac-tic Plane in the sub-mm range, at 870 µ m and 1.1 mm respectively,allowing a census of the sub-mm thermal emission from high-massregions. However, unprecedented opportunities to make a multi-wavelength FIR / sub mm study of the sky arrived with the Herschel satellite (Pilbratt et al. 2010), opening a new window of our un-derstanding of the cold universe and, in this context, Galactic starformation. For example, the earliest phases of star formation asso-ciated with IRDCs have been studied with the EPoS
Herschel keyprogram, for both low-mass (Launhardt et al. 2013) and high-mass c (cid:13) A. Traficante et al. objects (Ragan et al. 2012). This survey studied a sample of 12 low-mass and 45 high-mass regions respectively in order to characterisethe temperature and column densities of the cores and clumps atdi ff erent mass regimes.A comprehensive Herschel mapping of the Galactic coldand dusty regions has been recently completed with the
Herschel infrared Galactic Plane survey (Hi-GAL, Molinari et al. 2010b),which has mapped the whole Galactic Plane at latitudes | b | ◦ andfollowing the Galactic warp in the wavelength range 70 λ µ m. The aim of this work is to produce a first extensive cata-logue of young clumps embedded in IRDCs combining the Hi-GAL data with the most comprehensive catalogue of IRDCs todate (Peretto & Fuller 2009, hereafter Paper I). This initial cata-logue looks in a specific region of the inner Galactic Plane, in thelongitude range 15 ◦ l ◦ , which encompasses ≃ / sub-mm IRDCs counterparts; the source extraction and thecatalogues generation are described in Section 3. In this Section wealso briefly describe Hyper , a new algorithm used in this work anddeveloped for source extraction and photometry in complex back-grounds and crowded fields (Traficante et al. 2015). In Section 4 wedescribe the statistical distributions of the starless and protostellarclumps properties. Finally, in Section 5 we draw up our conclusionsabout this first release of clumps in IRDCs.
The IRDCs survey produced in Paper I contains ≃ µ m survey of the Galactic Plane (Benjamin et al.2003) delimited by | l | ◦ , | b | ◦ . The IRDCs have been iden-tified as connected regions with column density higher than N H > × cm − and a diameter greater than 4 ′′ (Paper I).We focused on a subsample of 3659 IRDCs observed in theregion 15 ◦ l ◦ , | b | ◦ . This region has been selectedfor its overlap with the Galactic Ring Survey (GRS, Jackson et al.2006). The GRS emission along the IRDCs line of sight (LOS) hasbeen used to identify clouds and to estimate the IRDCs kinematicdistances. The GRS survey mapped the CO J = ′′ and aspectral resolution of 0.212 km s − (Jackson et al. 2006). To obtainkinematic distances, we developed an automated procedure that ex-tracts the CO spectrum towards the centre of each IRDC, identi-fies the number of velocity components in it, integrates each com-ponent in a 2 km / s interval around the central velocity, and then cal-culates the ratio map for each of the integrated intensity maps withthe Herschel column density, smoothed at 46 ′′ resolution, centredat the IRDC position. We then select the velocity component forwhich the ratio map is the flattest, i.e. with a minimum dispersion.In some cases, no CO(1-0) components were found, and a -999flag was returned for the velocity. We identified 166 clouds witha -999 flag, therefore our starting IRDC catalogue is composed of3493 clouds. This procedure is simple, and rather easy to imple-ment for such a large number of sources. Kinematical distanceswere then calculated using the Reid et al. (2009) Galactic rotationmodel, always assuming the clouds located at the near distance incase of a distance ambiguity.
The Hi-GAL survey (Molinari et al. 2010b) has been carried out us-ing both the PACS (Poglitsch et al. 2010) and SPIRE (Gri ffi n et al.2010) photometry instruments on-board Herschel (Pilbratt et al.2010) in parallel mode, observing the sky at five wavelengths si-multaneously (70 and 160 µ m with PACS and 250, 350 and 500 µ m with SPIRE). The maps have been reduced with the ROMA-GAL pipeline (Traficante et al. 2011), an enhanced version of thestandard Herschel pipeline specifically designed for Hi-GAL. Aweighted post-processing on the maps has been applied to help withimage artefact removal (Piazzo et al. 2011). The maps have beenflux calibrated (by means of an o ff set subtraction) following theprescription of Bernard et al. (2010) and are expressed in MJy / sr.Due to the fast scan-speed and the co-addition on-board Herschel of 8 samples at 70 µ m and 4 samples at 160 µ m, the measured PACSbeams in the maps are slightly larger than the nominal ones. TheHi-GAL beams measured on the maps are 10.2 ′′ , 13.55 ′′ , 18.0 ′′ ,24.0 ′′ and 34.5 ′′ at 70, 160, 250, 350 and 500 µ m respectively. Themap pixel size is 3.2 ′′ , 4.5 ′′ , 6.0 ′′ , 8.0 ′′ , 11.5 ′′ at 70, 160, 250, 350and 500 µ m respectively (Traficante et al. 2011).For each IRDCs in our sample we selected the correspond-ing region in the Hi-GAL 70, 160, 250 and 350 µ m maps. We donot include the 500 µ m maps in the analysis due to the poor beamresolution compared to the other wavelengths.Some of the IRDCs of our sample have been already identi-fied in the Hi-GAL survey by Peretto et al. (2010). In this work theauthors showed that the IRDC FIR extension is slightly bigger thantheir appearance at 8 and 24 µ m. Therefore, for each IRDC, weisolate an Hi-GAL region which enlarged the 24 µ m IRDC size bythe equivalent of a 70 µ m beam (10.2 ′′ ) in each spatial direction.We assume that all the sources extracted between these extendedboundaries belong at the same IRDC. Due to the boundaries exten-sion and the proximity of some IRDCs, some sources can be asso-ciated with two di ff erent clouds. In these cases we associate eachsource with one IRDC randomly selected from the two clouds. Source extraction and photometry is a well known problem in as-tronomy, in particular in complex fields such as the Galactic Plane.The high resolution and sensitivity of the new IR instruments haveshown the complexity of the sites in which the newly born starsare located (such as the filamentary structures, e.g. Molinari et al.2010a). In particular the background variability can be significantin FIR Galactic Plane data, especially at longer wavelengths (e.g.,Molinari et al. 2010a; Peretto et al. 2010), and models must ac-count for its high variations across each source. Furthermore theGalactic Plane is dense and crowded and it is often the case thatsources are partially blended together. The problem is further com-plicated in case of multi-wavelengths study since each band has adi ff erent spatial resolution and often single sources are resolved inmultiple objects in the high-resolution maps.Various approaches based on di ff erent techniques (e.g. Cu-tex , Molinari et al. (2011); getsources , Men’shchikov et al. (2012))have been developed specifically for the new
Herschel
FIR data,in particular for the Galactic surveys.
Cutex in particular is thestandard source extraction algorithm used by the Hi-GAL team.It identifies the compact source in the second derivative image ofthe sky, and fits the sources with a 2d-Gaussian model. In this ap-proach the physical source diameter, FWHM dec , also needed to con-straint the source flux, is evaluated through the deconvolution of c (cid:13) , 1–19 catalogue of starless and protostellar clumps embedded in IRDCs. the HPBW with the FWHM λ of the source measured at each bandand it is therefore wavelength-dependent (e.g. Molinari et al. 2011;Elia et al. 2010; Veneziani et al. 2013).We decide to adopt a di ff erent approach and to estimate theflux consistently within the same volume of gas and dust at allwavelengths regardless of the di ff erent spatial resolutions. A simi-lar approach have been adopted by Olmi et al. (2013) to study theClMF in Hi-GAL fields.For this purpose we used a new algorithm, Hyper (HY-brid Photometry and Extraction Routine), fully described inTraficante et al. (2015). It has been designed specifically to takeinto account the complexities generated by the new datasets suchas the Galactic Plane observations made with
Herschel , in particu-lar the high background variability and the source crowding.
Hyper is based on a hybrid approach between the classical aperture pho-tometry and a 2d-Gaussian modelling of the sources. In the caseof multi-wavelength analysis, the region of integration is definedat a particular wavelength and the same region of the sky is usedto integrate the flux at all wavelengths. In the following Sections,along with the main results of this work we will describe the
Hyper parameters tuned for this source extraction and photometry.
The sources are initially identified at each wavelength separately ona modified high-pass filtered map using in sequence the find and gcnrtd
IDL routines (Traficante et al. 2015). These routines fit aGaussian model to recognize the peaks above a given threshold σ t defined as a multiple of the r.m.s. of the filtered map, σ f . A ref-erence threshold value for each Herschel wavelength, which min-imizes the false positives and maximizes the identification of realsources is provided in Traficante et al. (2015). However the sourcerecovery for a given value of σ t depends on variations of the (local)background, the source crowding and the cloud size, which di ff erfrom cloud to cloud (see also the discussion in Section 4.1.1). Wetested several values of σ t around the reference values discussedin Traficante et al. (2015) and after visual inspections of randomlyselected IRDCs we set σ t = [6.0,5.0,4.0,4.0] · σ f at [70,160,250,350] µ m respectively. These thresholds are a conservative compromisebetween source recovery and false identifications. From our visualinspections, less than 1% of the sources appear as false positives.The completeness of our catalogue based on this extraction is dis-cussed in Section 4.1.1.From this extraction we produced a catalogue of sources in-dependently identified at each wavelength. Hyper identified 8220,5393, 4967, 3413 sources at 70, 160, 250 and 350 µ m respectively.They are associated with 2070, 1640, 1621, 1246 IRDCs, respec-tively ≃ ≃ ≃
46% and ≃
36% of the whole sample. Thehigh number of 70 µ m sources compared to the other wavelengthsis likely due to the high spatial resolution of the 70 µ m data, whichallows the resolution of close sources possibly unresolved at longerwavelengths. Conversely, the low spatial resolution of the 350 µ mband blends sources resolved at shorter wavelengths. The longitudedistribution of sources at 70, 160 and 350 µ m is in Figure 1. The250 µ m source distribution is very similar to that of the 160 µ msources and is not showed.The multi-wavelength catalogues contain the photometry ofthe sources observed at all the specified wavelengths, as discussedin the next Sections, starting from a reference wavelength used toidentified the candidates. The reference wavelength is 160 µ m, thehighest resolution wavelength that is in common among protostel-lar and starless clumps. We initially generated a merged catalogue Figure 1.
Longitude distribution of sources independently identified at 70(brown), 160 (green) and 350 (blue) µ m. The distribution of 250 µ m sourcesis very similar to the 160 µ m distribution and it is therefore not shown. of 160, 250 and 350 µ m sources to produce a list of starless andprotostellar clump candidates in our selection of IRDCs. At the lo-cation of each 160 µ m source, a 250 and / or 350 µ m counterpart isassociated if a source is identified within a radius equal to half the160 µ m beam, 6.7 ′′ . The merged catalogue contains 1723 clumpsassociated with 764 di ff erent IRDCs ( ≃ Herschel wavelengths and may not be real clouds(e.g. Wilcock et al. 2012), however the small fraction of IRDCswith identified
Herschel compact sources here is due to a com-bination of factors. First, the merged catalogue does not considersources detected at only one or two wavelengths, nor objects de-tected at 160, 250 and 350 µ m but with centroids separated by morethan 6.7 ′′ from the 160 µ m centroid. Second, most IRDCs smallerthan ≃ ′′ are too small to be associated with a compact Herschel sources. Note that in particular with this choice we are not includ-ing in the catalogues the sources observed at 250 and 350 µ m withno counterparts at 160 µ m. We identified ≃
400 sources observedat 250 and 350 µ m only. Some of these sources are potentially ex-tremely young, their envelope so cold and / or di ff use that they donot emit above the background at wavelengths λ µ m. How-ever, we cannot extract physical parameters for sources identifiedat only two wavelengths, therefore we restrict the analysis to theclumps already visible at 160 µ m. A multi-wavelength analysis ofthese very cold clumps is the subject of a subsequent work.An independent extraction run used to produce a list of 70 µ msources has been combined with the previous sample to producetwo final catalogues:1. Sources identified at 160, 250 and 350 µ m but without coun-terparts at 70 µ m . These sources are classified as starless clumps.The catalogue contains 667 clumps associated with 389 IRDCs.The IRDCs and source longitude distributions are shown in Fig-ures 2 and 3 respectively.2. Sources identified at 160 , 250 and 350 µ m with (at least)one counterpart at 70 µ m . A 70 µ m source is associated with eachclump if its distance from the centroid of the 160 µ m source is lessor equal to half the 160 µ m beam, 6.7 ′′ . These sources are classi-fied as protostellar clumps. This catalogue contains 1056 sourcesassociated with 586 IRDCs. The IRDC and source longitude distri-butions are shown in 4 and 5 respectively. c (cid:13)000
400 sources observedat 250 and 350 µ m only. Some of these sources are potentially ex-tremely young, their envelope so cold and / or di ff use that they donot emit above the background at wavelengths λ µ m. How-ever, we cannot extract physical parameters for sources identifiedat only two wavelengths, therefore we restrict the analysis to theclumps already visible at 160 µ m. A multi-wavelength analysis ofthese very cold clumps is the subject of a subsequent work.An independent extraction run used to produce a list of 70 µ msources has been combined with the previous sample to producetwo final catalogues:1. Sources identified at 160, 250 and 350 µ m but without coun-terparts at 70 µ m . These sources are classified as starless clumps.The catalogue contains 667 clumps associated with 389 IRDCs.The IRDCs and source longitude distributions are shown in Fig-ures 2 and 3 respectively.2. Sources identified at 160 , 250 and 350 µ m with (at least)one counterpart at 70 µ m . A 70 µ m source is associated with eachclump if its distance from the centroid of the 160 µ m source is lessor equal to half the 160 µ m beam, 6.7 ′′ . These sources are classi-fied as protostellar clumps. This catalogue contains 1056 sourcesassociated with 586 IRDCs. The IRDC and source longitude distri-butions are shown in 4 and 5 respectively. c (cid:13)000 , 1–19 A. Traficante et al.
Figure 2.
Longitude distribution in the range 15 ◦ l ◦ of IRDCs withat least one starless clump (389, blue) together with the distribution of theIRDCs with at least one 160 µ m detection (1640, green) and the distributionof the total number of IRDCs in our catalogue with known distances (3493,brown). Figure 3.
Longitude distribution of sources. In green the 160 µ m sources(5393) and in azure the starless clumps (667). In total 211 clouds ( ≃
6% of the total) contain both starlessand protostellar clumps.Figures 1–5 show that the majority of the sources are locatedin the longitude range 25 ◦ l ◦ , and a second peak in thesource distribution occurs around l = ◦ , for both starless andprotostellar clumps. The longitude distribution of the clumps as afunction of their distance is shown in Figure 6. The clump distancesare assumed equal to the distance of their parent IRDCs. The me-dian distances are d = . µ m sources. The region around 25 ◦ l ◦ corresponds tosources mostly located at a distance d ≃ Figure 4.
IRDCs longitude distribution of IRDCs with at least one proto-stellar clump (586, red). The other bars are as shown in Figure 2.
Figure 5.
Longitude distribution of protostellar clumps in light red (1056).The other bar is as shown in Figure 3. and there are likely to be associated with the Scutum-Crux arm.The source distribution along this LOS is not dominated by starlessor protostellar clumps. The peak around l = ◦ corresponds to thetangent point of the Sagittarius-Carina arm, but passes also throughthe Perseus and the Norma-Cygnus arms. Some of the clumps inthis region come from the furthest IRDCs present in our catalogue(see Figure 6), located at ∼
10 kpc and are likely to be associatedwith the Perseus arm. The peak at l = ◦ in the source distribu-tion is more evident in the starless clumps. Although some 70 µ msources could be too faint to be observed in the furthest clouds,the mean IRDCs distance around l = ◦ is similar to the meandistance of the clouds around l = ◦ , 5 kpc and 4.6 kpc respec-tively. Therefore, the peak at l = ◦ in the starless distribution islikely to be a sign of younger star-forming region along this LOScompared to the Scutum-Crux region. The distribution in Figure 6indicates that the Hi-GAL sources are mainly located on the Galac-tic arms, although some are in interarms regions, as already notedby Russeil et al. (2011). c (cid:13) , 1–19 catalogue of starless and protostellar clumps embedded in IRDCs. Figure 6.
Starless and protostellar clump longitude distribution as functionof their parent IRDC distance. The two distributions look similar and thereare no regions with a clear predominance of starless or protostellar clumps.The majority of the sources are located around 25 ◦ l ◦ and l = ◦ .The median distances of the two distributions are d = .
21 kpc and d = . The flux of each source is estimated by integrating the emissionwithin the same elliptical region at all wavelengths. The semi-axes a and b of the elliptical aperture are equal to the FWHM G of the2d-Gaussian fit to each source estimated at a reference wavelength,which can be di ff erent from the wavelength used to initially identifythe clumps (Traficante et al. 2015). For the analysis here we set thereference wavelength to λ = µ m. The FWHM G of each fit canvary from a minimum of 1 · FWHM λ (18 ′′ at λ = µ m) up to2 · FWHM λ (36 ′′ ). This size at λ = µ m encompasses a regionof at least 1 . · FWHM λ at λ = µ m within the integration area.The range of allowed FWHM G has been chosen in order to accountfor point-like and slightly elongated compact sources and, at thesame time, to avoid highly elongated fits likely contaminated by theunderlying filaments aligned with some of the sources. The profileat λ = µ m also defines the source size as described in Section4.1. We note that within each elliptical aperture the clump couldbe resolved in multiple objects at 70 and 160 µ m. In these cases,the closest source in the cluster is identified as the counterpart andwe assign each clump the 70 and 160, µ m flux arising from all thesources within the elliptical integration area. This choice is con-sistent with our approach of evaluating the flux within the same area, since the sources are blended within the beam at longer wave-lengths and they all contribute to the observed emission. In thecase of these resolved clusters, a specific keyword in the catalogue(“cluster”) indicates the number of multiple sources observed at 70and 160 µ m within each integration area.The source flux evaluation is done at each wavelength after thebackground emission is subtracted and, in case of blended sources,the flux arising from the companions is removed. Hyper evaluatesthe local background by selecting various square regions acrosseach source and modelling the emission in each region with polyno-mial functions of di ff erent orders (from the zeroth up to the fourth).The fit which produces the lowest r.m.s. of the residuals is assumedas best fit and subtracted as background contribution.Two (or more) sources are considered blended if their cen- troids are closer than a fixed distance. This distance is automati-cally evaluated by the algorithm and it is equal to twice the maxi-mum allowed FWHM G (e.g. 72 ′′ ), the maximum distance at whichtwo elliptical apertures can be partially overlapped. Each sourceand its blended companions are fitted simultaneously with a multi-Gaussian function using the mpfit IDL routine (Markwardt 2009).The parameters of the fit are used to build 2d-Gaussian mod-els of the companions which are then subtracted from the image.The flux of the companion-subtracted source is then evaluated asin the isolated source case. A detailed description of the back-ground subtraction and source de-blending strategies can be foundin Traficante et al. (2015). However the source fluxes still requireto be corrected for two factors: the aperture and colour corrections.
Aperture corrections A c are needed in aperture photometry to com-pensate for the flux distributed outside the integration area due tothe convolution of the true sky signal with the instrumental beam,the PSF response. These corrections therefore depend on the beamat each di ff erent wavelength and, for point-like sources, can be esti-mated by measuring the source emission of reference sources withknown flux. A further correction factor has to be applied for ex-tended sources.For PACS, the aperture corrections are known for circularapertures as function of aperture radius . The equivalent apertureradius r eq of each source, described in Section 4.1, is used to esti-mate the corresponding A c .For SPIRE, aperture corrections are known for circular aper-tures as function of aperture radius and source spectral index . Thecorrections are estimated at fixed aperture radii of 22 ′′ , 30 ′′ and 40 ′′ at 250, 350 and 500 µ m respectively. Our aperture radius dependsthe 2d-Gaussian fit at λ = µ m to each source and it is fixedat each wavelength (Section 3.2). However, the range of allowedFWHM G is from 18 ′′ to 36 ′′ , therefore we decided to adopt theproposed corrections without further assumptions and assuming aspectral index of β = . Colour corrections are needed to convert the flux density measuredby PACS and SPIRE into the monochromatic flux density for eachobserved object, which depends on the intrinsic temperature andspectral energy distribution (SED) of the source. The colour correc-tions are available online for both PACS and SPIRE instruments.The colour corrections have been measured for di ff erent spectral in-dexes and blackbody (and greybody) temperatures for both PACSand SPIRE instruments. We adopted a fixed spectral index β = . T
30 K for both the PACS70 and 160 µ m filters are shown in Figure 7. In order to extrapolatethe colour corrections for our sample however we need to know the http : // herschel . esac . esa . int / twiki / pub / Public / PacsCalibrationWeb / pacs bolo fluxcal report v1 . pdf http : // herschel . esac . esa . int / twiki / pub / Public / PacsCalibrationWeb / http : // herschel . esac . esa . int / Docs / SPIRE / spire handbook . pdf http : // herschel . esac . esa . int / twiki / pub / Public / PacsCalibrationWeb / cc report v1 . pdf http : // herschel . esac . esa . int / twiki / pub / Public / PacsCalibrationWeb / http : // herschel . esac . esa . int / Docs / SPIRE / spire handbook . pdfc (cid:13) , 1–19 A. Traficante et al.
Figure 7.
PACS colour correction curves for both the 70 and 160 µ m filters,extrapolated from the colour correction values at T =
10, 15, 20 and 30 K de-scribed in the PICC-ME-TN-038 PACS report. The green and black crossesshow the colour corrections applied for the starless and protostellar clumpsflux distributions respectively. Their values are 1.307 and 0.946 for starlessand 1.223 and 0.955 for protostellar clumps at 70 and 160 µ m respectively,corresponding to a temperature of T = greybody temperature in advance and this poses a circular prob-lem (Pezzuto et al. 2012). So we estimate the colour correctionsfollowing an iterative procedure. We first fixed the colour correc-tions assuming T =
15 K for both starless and protostellar clumps,in agreement with previous estimations of their envelope tempera-ture (e.g. Veneziani et al. 2013). We obtained a first estimation ofthe greybody temperatures of our sources resulting in median tem-perature values of T = = = = = ffi ciently close to the previous estimate, that no further iterate wascarried out. The adopted colour corrections are 1.307 and 0.946 forstarless and 1.223 and 0.955 for protostellar clumps at 70 and 160 µ m respectively. The SPIRE colour corrections do not vary signif-icantly in the range 10-20 K at β = .
0. Therefore, we adopted thepublicly available values assuming T =
15 K for both starless andprotostellar clumps. The values are 1.223 and 0.955 for the 250and 350 µ m fluxes respectively. The quoted error for the flux in the PACS and SPIRE referencemanuals is, after aperture and colour corrections, < Hyper photometry. The error quoted in the catalogues is the flux error es-timated as the local sky r.m.s. per pixel multiplied for the num-ber of pixels in the area over which the source flux is integrated(Traficante et al. 2015). In addition, the precision in recovering thesource flux depends on the intensity of each source and the localbackground variability. For each source we assume a conservative error of 5% of its flux associated with the aperture and colour cor-rections plus an error of 15% from
Hyper measurements. Henceeach source flux has associated a conservative error of 20% of itsflux at every wavelength.
An example of source extraction and photometry at the fourconsidered wavelengths in the SDC23.271-0.263, catalogued asHGL23.271-0.263, is shown in Figure 8.
Hyper identified sevensources observed at 160, 250 and 350 µ m simultaneously. Sixsources have a 70 µ m counterpart and are therefore classified asprotostellar clumps, one is classified as starless clump. The 2d-Gaussian fits of the clumps evaluated at 250 µ m are shown ingreen and blue for starless and protostellar clumps respectively. Thesources identified at 160 µ m are indicated as black crosses in the160 µ m map.The catalogue parameters obtained for the sources extracted inHGL23.271-0.263 are listed in Table 1. The quantities reported foreach source are: the IRDC name and the source identification num-ber as extracted from Hyper (Columns 1–2), which corresponds tothe number in the associated region file; the wavelength (Column3); the source peak flux in MJy / sr and Jy, evaluated at the centroidpixel (Columns 4–5); the source integrated flux (not corrected) andits associated error (Columns 6–7); the sky r.m.s. before and afterthe background subtraction (Columns 8–9); the polynomial orderused to estimate the background (Column 10); the source semi-axes and the position angles as obtained from 2d-Gaussian fittingat the reference wavelength, 250 µ m (Columns 11–13); the conver-gence status of the fit (Column 14). This status is 0 if the fittingroutine converges, otherwise it assumes the value − > .
5) or − . · FWHM µ m ); the position of the sourcecentroids at each di ff erent wavelength, in both Galactic and Equato-rial (J2000) coordinates (Columns 15–18); the distance between the160 µ m centroid and the centroids of the source counterparts at theother wavelengths (Column 19); the number of sources with their2d-Gaussian profile overlapped and de-blended before the sourceflux integration (Column 20). The number of multiple sources re-solved at wavelengths other than the 250 µ m within the integrationarea (Column 21). c (cid:13) , 1–19 catalogue of starless and protostellar clumps embedded in IRDCs. Figure 8.
The FIR counterpart of the IRDC catalogued as HGL23.271-0.263. From top to the bottom: 70, 160, 250 and 350 µ m images. The protostellarclumps are shown in blue and the starless clumps are shown in green. The ellipses correspond to the elliptical region, defined at 250 µ m, used to estimate thesource fluxes at all wavelengths. The black crosses in the 160 µ m image shows the sources initially identified in this map.c (cid:13) , 1–19 A . T r afi c an t ee t a l . map sou. band peak fl. peak fl. flux err flux sky nob. sky ord. fwhm fwhm PA status glon glat ra dec dist. deb. clust( µ m) (MJy / sr) (Jy) (Jy) (Jy) (Jy) (Jy) ( ′′ ) ( ′′ ) ( ◦ ) ( ◦ ) ( ◦ ) ( ◦ ) ( ◦ ) ( ′′ )23.271-0.263 1 70 2249.96 0.54 16.60 2.21 0.213 0.142 3 21.91 36.07 149.14 0 23.242 -0.240 278.623 -8.726 3.35 2 123.271-0.263 1 160 2623.40 1.25 37.29 3.12 0.474 0.282 1 21.91 36.07 149.14 0 23.243 -0.239 278.623 -8.725 0.00 2 123.271-0.263 1 250 1102.27 0.93 24.60 4.16 0.734 0.501 1 21.91 36.07 149.14 0 23.243 -0.239 278.622 -8.724 1.39 2 123.271-0.263 1 350 435.34 0.65 14.70 5.08 0.949 0.815 1 21.91 36.07 149.14 0 23.243 -0.240 278.623 -8.725 3.11 2 123.271-0.263 2 70 12507.02 3.01 100.25 4.41 0.297 0.265 1 25.02 36.07 110.66 0 23.256 -0.240 278.630 -8.714 2.94 2 123.271-0.263 2 160 10611.79 5.05 114.53 4.45 0.520 0.377 4 25.02 36.07 110.66 0 23.257 -0.240 278.630 -8.713 0.00 2 123.271-0.263 2 250 4201.20 3.56 95.12 9.93 1.295 1.119 1 25.02 36.07 110.66 0 23.257 -0.240 278.630 -8.713 0.83 2 123.271-0.263 2 350 1572.63 2.37 64.42 5.86 1.031 0.881 1 25.02 36.07 110.66 0 23.257 -0.240 278.630 -8.712 1.75 2 123.271-0.263 3 70 3661.37 0.88 43.39 3.02 0.241 0.189 1 23.10 36.07 102.28 0 23.250 -0.225 278.613 -8.712 2.16 2 123.271-0.263 3 160 3240.27 1.54 70.44 4.08 0.345 0.359 1 23.10 36.07 102.28 0 23.250 -0.225 278.613 -8.712 0.00 2 123.271-0.263 3 250 1454.45 1.23 49.61 2.31 0.399 0.271 3 23.10 36.07 102.28 0 23.251 -0.224 278.613 -8.711 2.97 2 123.271-0.263 3 350 375.82 0.56 14.01 2.41 0.546 0.376 1 23.10 36.07 102.28 0 23.252 -0.225 278.614 -8.710 8.79 2 123.271-0.263 4 70 6063.55 1.46 89.55 5.59 0.283 0.280 2 36.07 36.07 252.40 0 23.271 -0.256 278.651 -8.707 2.39 0 423.271-0.263 4 160 13647.29 6.50 394.78 10.69 0.767 0.753 4 36.07 36.07 252.40 0 23.271 -0.257 278.651 -8.708 0.00 0 223.271-0.263 4 250 7821.71 6.62 298.13 9.59 0.945 0.900 3 36.07 36.07 252.40 0 23.271 -0.257 278.651 -8.708 1.27 0 123.271-0.263 4 350 3162.00 4.76 138.29 4.12 0.660 0.516 3 36.07 36.07 252.40 0 23.271 -0.256 278.651 -8.708 1.46 0 123.271-0.263 5 70 1848.66 0.44 132.85 7.69 0.451 0.385 1 36.07 36.07 230.89 0 23.321 -0.298 278.711 -8.682 8.72 0 223.271-0.263 5 160 4595.00 2.19 243.37 8.49 0.756 0.597 3 36.07 36.07 230.89 0 23.321 -0.295 278.709 -8.681 0.00 0 223.271-0.263 5 250 3167.15 2.68 161.36 6.69 0.782 0.628 4 36.07 36.07 230.89 0 23.321 -0.294 278.708 -8.681 3.63 0 123.271-0.263 5 350 1553.36 2.34 93.40 4.34 0.591 0.543 1 36.07 36.07 230.89 0 23.321 -0.295 278.709 -8.681 0.86 0 123.271-0.263 6 70 1610.35 0.39 21.64 1.18 0.084 0.059 4 36.07 36.07 117.66 0 23.339 -0.327 278.746 -8.680 4.49 0 123.271-0.263 6 160 2004.29 0.95 47.64 2.10 0.268 0.148 4 36.07 36.07 117.66 0 23.340 -0.327 278.747 -8.679 0.00 0 123.271-0.263 6 250 971.20 0.82 45.75 2.06 0.434 0.193 4 36.07 36.07 117.66 0 23.340 -0.327 278.747 -8.679 2.17 0 123.271-0.263 6 350 465.66 0.70 22.41 2.02 0.418 0.252 4 36.07 36.07 117.66 0 23.341 -0.328 278.748 -8.678 4.99 0 123.271-0.263 7 160 1392.66 0.66 13.40 1.02 0.243 0.116 1 21.17 23.21 232.83 0 23.300 -0.251 278.659 -8.679 0.00 0 123.271-0.263 7 250 1663.01 1.41 32.07 2.93 0.449 0.447 1 21.17 23.21 232.83 0 23.300 -0.250 278.658 -8.679 3.78 0 123.271-0.263 7 350 755.46 1.14 22.97 2.34 0.439 0.476 1 21.17 23.21 232.83 0 23.300 -0.249 278.658 -8.679 5.28 0 1 Table 1.
Output parameters for protostellar and starless clumps extracted in HGL23.271-0.263. The catalogues for protostellar (sources 1–6) and starless clumps (source 7) are separated by a blank line in this table.Column 1: IRDC name. Column 2:
Hyper source number. Column 3: wavelength. Columns 4–5: source peak flux in MJy / sr and Jy. Columns 6–7: source integrated flux and source flux error in Jy. This source fluxis not corrected for aperture or colour corrections. The flux error is estimated from the local sky r.m.s. multiplied by the number of pixels in the area over which the source flux is integrated. Column 8–9: r.m.s. ofthe sky evaluated in the rectangular region used to model the background before and after the background subtraction respectively. Column 10: polynomial order used to model the background. Columns 11–12:minor and maximum FWHM of the 2d-Gaussian fit. Column 13: Source Position Angle. Column 14: goodness of the 2d-Gaussian fit. Status is equal to 0 if the fit has regularly converged. Columns 15–18: sourcecentroids Galactic and Equatorial coordinates (J2000). Column 19: distance from the source centroids in the reference wavelength and the source counterparts at the other wavelengths. Column 20: number of sourcesidentified as companions and de-blended. Column 21: number of source counterparts at each wavelength. It normally is equal to 1 but it can be higher if the source is resolved in more than one counterpart in thehigh resolution maps. c (cid:13) R A S , M N R A S , catalogue of starless and protostellar clumps embedded in IRDCs. The source photometric properties and the source distances arecombined to build the SEDs and estimate equivalent radius, tem-perature, mass and FIR luminosity. We assume a single temper-ature greybody model with a fixed spectral index to describe theemission of the cold dust envelope of both starless and protostel-lar clumps. We model the flux of each source S ν at frequency ν as(e.g., Elia et al. 2010; Giannini et al. 2012): S ν = M κ d νν β B ν ( T ) Ω (1)where ν and κ are respectively a normalisation frequency and thedust mass absorption coe ffi cient evaluated at ν ( ν =
230 GHz and κ = with a gas to dust ratio of 100, Preibisch et al.1993). The spectral index β has been fixed to β = .
0, in agree-ment with the standard value for cold dust emission (Hildebrand1983) and with previous studies of starless and protostellar clumps(e.g. Motte et al. 2010; K¨onyves et al. 2010; Veneziani et al. 2013).The total mass (gas + dust) is M and Ω is the solid angle in whichthe flux has been evaluated. B ν ( T ) is the Planck blackbody functionat frequency ν and temperature T . Finally, d is the distance of thesource.For both starless and protostellar clumps we used the fluxesat 160, 250 and 350 µ m to model the SED and estimate mass anddust temperature. The FIR luminosity is evaluated by integratingthe flux in the range 70 λ µ m. The 500 µ m is extrapolatedfrom the SED fit in both the distributions, where the 70 µ m flux isextrapolated from the fit for starless clumps and is obtained fromthe measured flux for protostellar clumps. The 70 µ m flux (and thatat shorter wavelengths) of the protostellar clumps is partly influ-enced by the warm core(s) in the centre (e.g., Dunham et al. 2008;Motte et al. 2010). A single temperature greybody model cannotaccount for this second, warm component. Therefore, we do notinclude the 70 µ m emission to estimate the protostellar clump dusttemperature and mass.We note that the temperature gradients across the wholeIRDCs can be up to ≃
10K (Peretto et al. 2010), but it is neg-ligible across the starless clumps. Furthermore, the di ff erences inthe starless clumps mass estimation assuming a single-temperaturegreybody model or with a detailed radiative transfer model is negli-gible (Wilcock et al. 2011). At the same time, the central regions ofthe protostellar clumps partially warmed-up by the central core(s)are not included in the model, therefore the model underestimatesthe protostellar clumps central temperature. As a consequence ofEquation 1, the model could overestimate the protostellar masses(see in Section 4.1). The adopted single clump-averaged tempera-ture model is however a reasonable approximation to describe theemission arising from the cold, extended dust envelope of bothtypes of objects. A more complete analysis of protostellar prop-erties requires to model at least two di ff erent temperature compo-nents, using more sophisticated models (e.g. Robitaille et al. 2007)and using shorter wavelengths to constraint the warm component(e.g. Motte et al. 2010).The best fit for the greybody model is evaluated with achi-square minimisation analysis using the mpfit IDL routine(Markwardt 2009). In order to restrict the analysis to the sourceswith good temperature, mass and luminosity estimates, onlysources with fits with χ
10 are considered further. To furtherdelimit the sample, we excluded few sources with χ
10 but withunrealistic temperature values (T >
40 K), likely a consequence of constraining the SED with only three points. The majority of thesources have good fits however, and the final catalogue contains649 out of a total of 667 starless clumps and 1030 out of 1056 pro-tostellar clumps.Figure 9 shows the SEDs for the starless and protostellarclumps observed in HGL23.271-0.263. The plots for the protostel-lar clumps also show the emission at 70 µ m, which is in excess ofthe SED fits in all cases. Each SED plot reports the best-fit parame-ters, the distance of the source and the value of the χ . These valuesare also reported in Table 4. We adopt a source size given by the equivalent radius r eq , definedas the radius of a circle with the same area, A = π r eq , as the ellip-tical region evaluated from the 250 µ m source fit. Unlike previouswork on Herschel data we do not assume the deconvolved radiusto describe the source size, which is a description of the intrinsicshape of the source but it is wavelength-dependent (e.g. Motte et al.2010; Giannini et al. 2012; Elia et al. 2013). Instead, r eq = √ A /π describes the dimension of the region across which the flux is eval-uated at all wavelengths.The distribution of r eq for both starless and protostellar clumpsis shown in Figure 10. A Kolmogorov-Smirnov (KS) test shows thatthere is no evidence that the distributions are di ff erent. The medianvalues of the two distributions are similar, ¯ r eq ≃ . µ m beam is equivalent to ≃ . r . ≃ ≃ . R eq (see Section 4.1), to twice the FWHM µ m , 36 ′′ , which correspondsto R eq ≃ = µ m is made to isolate compact structures embeddedin the clouds and we expect the clumps to be relatively circular,with aspect ratios of less than 2 on average (e.g. Urquhart et al.2014). The few sources with radii r eq > d ≃
10 kpc incorrespondence of the Perseus arm at l ≃ ◦ (see Section 3.1).The distribution of source temperature for both the starless andprotostellar clumps are shown in Figure 11. The median tempera-tures of the distributions is T ≃ ≃ < − , although it is clear that the protostellar clumps are onlyslightly warmer than starless clumps on average. Since the emis-sion in the 160-350 µ m spectral range arises from the dust enve-lope for both starless and protostellar clumps, this result impliesthat the protostellar envelope is only partially warmed-up by thecentral core(s), which may be an indication of the youth of the pro-tostellar systems.The mass distributions of the sources are shown in Figure12. The red vertical line defines M com =
105 M ⊙ which indicatesthe mass completeness limit as explained in the next Section. Thislimit lies below the turnover point which is around M ≃ M ⊙ .The identified sources span a wide range of masses, from few solarmasses up to few ≃ M ⊙ with average clump mass of ≃
193 M ⊙ and ≃
272 M ⊙ for starless and protostellar clumps respectively. The c (cid:13) , 1–19 A. Traficante et al.
Figure 9.
Fluxes at 70, 160, 250 and 350 µ m and SEDs for the 6 protostellar (blue) clumps and the starless (bottom, red) clumps in HGL23.271-0.263. Thespectral index is fixed to β = .
0. The free parameters of the fit are the temperature and the mass, while the luminosity is obtained integrating the emission inthe range 70 λ µ m. The χ value of the fit is also shown. c (cid:13) , 1–19 catalogue of starless and protostellar clumps embedded in IRDCs. IRDC source type R M L T T err χ distance(pc) (M ⊙ ) (L ⊙ ) (K) (K) (kpc)23.271-0.263 1 protostellar 0.71 228 926 17.7 2.1 0.595 5.2123.271-0.263 2 protostellar 0.76 1240 3110 16.0 1.6 0.055 5.2123.271-0.263 3 protostellar 0.73 152 1748 22.7 3.1 1.730 5.2123.271-0.263 4 protostellar 0.91 2537 8591 17.7 1.9 0.042 5.2123.271-0.263 5 protostellar 0.91 1628 6175 17.4 1.9 0.326 5.2123.271-0.263 6 protostellar 0.91 563 1073 15.8 1.5 0.129 5.2123.271-0.263 7 starless 0.56 1148 200 11.7 0.7 0.777 5.21 Table 2.
Temperature, mass and luminosity of the 7 clumps identified in HGL23.271-0.263, six protostellar and one starless. Columns 1-2: IRDC name andsource number. Column 3: protostellar or starless type. Columns 4–6: radius, mass and luminosity of each source. Columns 7–8: temperature and the associatederror estimated with the mpfit routine. Column 9: χ value of the greybody fit. Column 10: source distance, equal to the parent IRDC distance. Figure 10.
Equivalent radius distribution of the clumps identified in the cat-alogue for both starless (red) and protostellar (blue) clumps. The mean val-ues are 0.61 pc and 0.63 pc for starless and protostellar clumps respectively.The radii have been evaluated as the radii of circles with the same area ofthe ellipses estimated from the 2d-Gaussian fits at 250 µ m, as explained indetail in the text. distributions are similar for both starless and protostellar dumps,likely due to the fact that the envelope of these clumps are not sub-stantially perturbed by the internal gravitational collapse which isoccurring in the protostellar cores (Giannini et al. 2012). Some ofthese clumps are potentially the birth site of high-mass objects, asdiscussed in Sections 4.2 and 4.3.Some of the clumps in the sample have also been observedas part of the Herschel
EPOS survey of high-mass star formingIRDCs by Ragan et al. (2012) . They found 496 protostellar coresand clumps simultaneously observed at 70, 100 and 160 µ m associ-ated with 45 IRDCs across the Galactic Plane. Longer wavelengthswere excluded from the analysis in order to preserve the high spatialresolution of the PACS data. In the Ragan et al. (2012) catalogue 64protostellar clumps fall in the longitude range 15 ◦ l ◦ andoverlap with our sample. All these 64 sources were detected in ouranalysis at 70 µ m, but only 35 of them have observable counterpartsat 160, 250 and 350 µ m and are part of the protostellar clump cat-alogue. In addition 3 sources have been identified as starless, sincethe 160 µ m centroids are located more than 6 . ′′ away from the 70 µ m source (Section 3.1). The properties of the two overlapping samples are not straight-forward to compare, since the higher spatial resolution of theRagan et al. (2012) sample allows them to model the properties ofthe inner part of the clumps, and in some cases of the single cores.Indeed in our catalogue the dust temperature are, on average, 5 Kcolder and the mass are, on average, a factor of ≃ ≃ ⊙ whereas Ragan et al.(2012) find a mean of M ≃
740 M ⊙ . Similarly for the the 35 pro-tostellar clumps, in our catalogue the mean mass is M ≃ ⊙ whereas Ragan et al. (2012) find M ≃
215 M ⊙ . The inclusion of the70 µ m flux in the SED fitting by Ragan et al. (2012) significantlyraises the temperature estimation which contributes to these massdi ff erences. The sources have a mean temperature of 14.4 and 18.3K for starless and 17.5 and 21.7 K for protostellar clumps, in ourcatalogue and in the Ragan et al. (2012) catalogue respectively.It is important to note that the mass (and mass surface den-sity) values have significant uncertainties due to the assumed dustmodel. The opacity and the spectral index of the dust can varyamong the di ff erent models by a factor of 2 or more (Ormel et al.2011). High-resolution, multi-wavelength observations at sub-mm / mm wavelengths are required to better constraint the dust prop-erties and hence the mass of each clump. The mass completeness of the catalogue is determined by the abil-ity to recover faint sources in the presence of highly variable back-grounds, which vary from cloud to cloud by orders of magnitude.Also, we are extracting sources from thousands of relatively smallclouds located at a range of distances, d , which will have masses ∝ d (see Equation 1). Therefore a single completeness limit for thecatalogue is poorly representative of the whole sample of clouds.Instead, a mass completeness needs to be evaluated locally for eachcloud which accounts for the background confusion and the sourcedistance.In order to derive the local mass completeness for each cloudwe compared the mean mass of the embedded clumps with the cor-responding mean background-equivalent source mass M bg of eachcloud. For each clump we define the background-equivalent sourceas a point source with a flux equal to the clump background flux in-tegrated in a circular region with a radius equal to the FWHM µ m at all wavelengths (see Section 3.2). The clump background fluxper pixel at each wavelength is defined as the average flux emis-sion per pixel evaluated in a region surrounding each clump, with c (cid:13)000
215 M ⊙ . The inclusion of the70 µ m flux in the SED fitting by Ragan et al. (2012) significantlyraises the temperature estimation which contributes to these massdi ff erences. The sources have a mean temperature of 14.4 and 18.3K for starless and 17.5 and 21.7 K for protostellar clumps, in ourcatalogue and in the Ragan et al. (2012) catalogue respectively.It is important to note that the mass (and mass surface den-sity) values have significant uncertainties due to the assumed dustmodel. The opacity and the spectral index of the dust can varyamong the di ff erent models by a factor of 2 or more (Ormel et al.2011). High-resolution, multi-wavelength observations at sub-mm / mm wavelengths are required to better constraint the dust prop-erties and hence the mass of each clump. The mass completeness of the catalogue is determined by the abil-ity to recover faint sources in the presence of highly variable back-grounds, which vary from cloud to cloud by orders of magnitude.Also, we are extracting sources from thousands of relatively smallclouds located at a range of distances, d , which will have masses ∝ d (see Equation 1). Therefore a single completeness limit for thecatalogue is poorly representative of the whole sample of clouds.Instead, a mass completeness needs to be evaluated locally for eachcloud which accounts for the background confusion and the sourcedistance.In order to derive the local mass completeness for each cloudwe compared the mean mass of the embedded clumps with the cor-responding mean background-equivalent source mass M bg of eachcloud. For each clump we define the background-equivalent sourceas a point source with a flux equal to the clump background flux in-tegrated in a circular region with a radius equal to the FWHM µ m at all wavelengths (see Section 3.2). The clump background fluxper pixel at each wavelength is defined as the average flux emis-sion per pixel evaluated in a region surrounding each clump, with c (cid:13)000 , 1–19 A. Traficante et al.
Figure 11.
Temperature distribution of the 1723 clumps, starless (red) andprotostellar (blue). The median values are T = = Figure 12.
Mass distribution of the 1723 clumps, starless (blue) and proto-stellar (red). The dashed red line is the mass completeness, fixed at M = ⊙ . The mass range spans several order of magnitudes, from few solarmasses up to few 10 M ⊙ . The median values are M =
194 M ⊙ and M = ⊙ for starless and protostellar clumps respectively. the clump masked and with a sigma-clipping procedure to avoidsignificant contamination from companion sources.The background-equivalent source mass is then evaluated us-ing the same SED fitting procedure described in the previous Sec-tion and assuming for the background sources the same distance astheir corresponding IRDC. The mean value of all the background-equivalent source masses in each cloud determines the local IRDCM bg . The mean mass of the background-equivalent sources com-pared with the sources in each IRDC is shown in Figure 13. Sourceswith a mass up to M bg are hidden in the background emissionand unidentifiable in that specific cloud. M bg sets a local masscompleteness limit which accounts for the local background varia-tion and the IRDC distance. There are a few tens of clouds withM bg > M ⊙ and, at the same time, several clouds for whichsources with M
10 M ⊙ are easily identifiable. On average the mean source mass is few times greater than the corresponding meanM bg , however there are identified sources with masses close to thecorresponding mean M bg .To set an estimate of the mass sensitivity limit for the wholecatalogue, we identify the mass value M com where 90% of cloudshave M bg M com (the green dashed line in Figure 13). In otherwords, in 90% of the clouds all sources with masses M > M com aredetected. This mass value is M com ≃
105 M ⊙ . We define this value asa mass completeness limit for the catalogues. The remaining 10%of the clouds with M bg above the limit are among the furthest in thecatalogue, with a mean distance of 5.4 kpc (in comparison with theaverage distance of the whole catalogue, 4.2 kpc, see Section 3.1).We stress that this mass completeness limit is strongly in-fluenced by the IRDC distance and has to be taken “cum granosalis”, in general a local mass completeness should be used foreach cloud. As showed in Figure 14, the distribution of M bg sub-stantially changes if we divide the source sample in two subsets,one including all the sources located at d . d > . com ( d . = ⊙ and M com ( d > . =
167 M ⊙ for the close and the far samplerespectively. To further demonstrate that the mass completeness isdominated by the di ff use emission and depends on the source dis-tance, we estimated the mass of a background-equivalent source lo-cated at the mean distance of d = µ m arising from residual noise only (0.6 Jy / pixel, Traficante et al.2011). Its mass is ≃
10 M ⊙ (the orange-dotted line in Figure 14).For comparison, the average M bg of all clouds located at ≃ ≃
30 M ⊙ , while ≃
10 M ⊙ is the mean M bg of clouds located at ≃ . com gives a reference value which only defines that we re-cover sources with masses > M com in 90% of the clouds and it isnot related to the underlying intrinsic clump mass spectrum, forwhich a statistically complete sample would be required. As a con-sequence, for example, it is not clear whether the turnover point inthe source mass distribution, around M ≃ M ⊙ (see Figure 12), ismeaningful beyond describing the actual distribution of clumps inthe catalogue. The local mass completeness values for each cloudis available with the source catalogues. In the past it has been suggested that IRDCs are favourableplaces for high-mass star formation (e.g. Carey et al. 1998). Thissuggestion has been corroborated by several observations ofmassive YSOs associated with IRDCs (e.g. Beuther et al. 2013;Beltr´an et al. 2013; Peretto et al. 2013). However, IRDCs are alsoforming low-to-intermediate stars in regions devoid of high-massstars (e.g. Sakai et al. 2013).One way to determine if the clumps associated with the IRDCsare ongoing high-mass star formation is to look at the mass-radiusdistribution under the assumption that, in order to form high-massstars, a large amount of material needs to be concentrated in toa relatively small volume. Recently an empirical threshold forthe formation of mass stars in IRDCs has been proposed. Thisthreshold is, in its original formulation, m ( r ) >
870 M ⊙ ( r / pc) . (Kau ff mann & Pillai 2010, hereafter KP). This model accounts forsingle fragments within each star-forming region and it is wellsuited to be compared with our clumps catalogue. In Figure 15we show the mass-radius relationship for our catalogue of starless c (cid:13) , 1–19 catalogue of starless and protostellar clumps embedded in IRDCs. Figure 13.
Distribution of the mean source mass per IRDC as function ofthe M bg . The blue line is the y = x line. The majority of the clumps have M > M bg . Figure 14. M bg distribution for sources located at d . d > . com , which definesthe mass value above which all sources are identified in 90% of the clouds,changes significantly with distance with values of M com =
37 M ⊙ for thedistribution of the closest sources and M com =
167 for the distribution ofthe more distant sources. The orange-dotted line shows for comparison themass of a background-equivalent source located at the mean distance ofd = µ m arising from residual noise only. clumps, colour-coded for their dust envelope temperature. The lightbrown shaded area delimits the KP region above which massivestars can be formed. For comparison, the figure also shows Larson’sthird law as it appeared in its original formulation, m ( r ) >
460 M ⊙ ( r / pc) . (green dashed line), which describes the universality ofthe scaling relation between mass and radius of molecular clouds(Larson 1981).We identify 171 starless clumps above the KP threshold, dis-tributed in 130 IRDCs. Interestingly by this criterion ≃
26% of thestarless clumps are likely to form high-mass stars. These are situ-ated in ≃
33% of the IRDCs with identified starless clumps (389,Section 3.1). At the same time, the clouds which encompass thesehigh-mass starless clumps are only 4% of the total clouds in oursample (3493, see Section 2). While the mass completeness limit influences the percentage of high-mass starless clumps, the num-ber of clouds which embed these high-mass clumps is much lessa ff ected since it is unlikely that we are missing the most massiveobjects associated with the clouds. We can conclude that only asmall fraction of IRDCs will potentially form high-mass stars, inagreement with the results of Kau ff mann & Pillai (2010).The dash-dotted red line in Figure 15 defines the empiricalthreshold for high-mass star formation proposed by Urquhart et al.(2014). This threshold is based on the analysis of high-mass starforming clumps identified in the ATLASGAL survey and corre-sponds to clumps with a constant mass surface density of Σ = .
05g cm − . This threshold however sets less stringent constraints thanthe KP threshold for clumps with an equivalent radius of R ff ected by mass-loss injets or outflows, or mass segregated in the central core(s) as wellas overestimating the mass as a result of using a single temperaturemodel and incompletely accounting for the heating by the centralprotostars. However, we find 456 protostellar clumps ( ≃
43% of thetotal) associated with 291 IRDCs ( ≃
50% of the IRDCs with em-bedded protostars, and ≃
8% of the total) above the KP thresholdand so are capable of forming high mass stars.The low percentage of IRDCs possibly forming high-massstars does not exclude however that the majority of the high-massstar formation activity is associated with IRDCs. This issue canbe addressed, for example, by comparing the star formation activ-ity within the IRDCs and in a large sample of massive molecularclouds not associated with IRDCs, or tracing the star formation rateinside and outside IRDCs. Both these analyses are the main topicof a forthcoming work.The temperature distribution in the diagrams show that thecolder clumps are the bigger and the more massive. For starlessclumps this may be an indication of their evolutionary phase, withthe smaller clumps having already started central accretion andlikely being just at the verge of a protostar core formation, or maybewith a very young protostar already formed but still obscured at70 µ m. For protostellar clumps this is likely a consequence ofthe adopted model, since the single clump-averaged temperature isdominated by the extended cold dust envelope emission with lesscontribution from the warmer central regions as the clump size in-creases. The environment of massive star formation can be also investigatedthrough the mass surface density, Σ , versus mass diagram. Thework of Krumholz & McKee (2008) suggested Σ = − asa mass surface density threshold required to prevent fragmentationinto low-mass cores as a result of radiative feedback, thus allowinghigh-mass star formation. However, this threshold has a large scat-ter and the calculation does not include the contribution of magneticfields, which can prevent fragmentation without imposing a mini-mum mass surface density (e.g. Butler & Tan 2012). Several high-mass clumps and cores are indeed observed with 0 . Σ . − (Butler & Tan 2012; Tan et al. 2013). One of the most mas-sive cores observed in the inner Galaxy, associated with the IRDCSDC335, with a mass of ≃
500 M ⊙ in ≃ .
05 pc (Peretto et al. c (cid:13) , 1–19 A. Traficante et al.
Figure 15.
The mass of the starless clumps as function of their equiva-lent radius, colour-coded for the temperature of the dust envelope. Themass increases with increasing equivalent radius. The unshaded area de-limits the region of the high-mass star formation in IRDCs as determinedby the Kau ff mann & Pillai (2010) empiric analysis. Almost one third of thestarless clumps (171 out of 667) lie above this threshold. The line-dottedred line is the threshold adapted from Urquhart et al. (2014) who proposean empirical lower limit value for high-mass star formation equivalent to aconstant mass surface density of Σ = .
05 g cm − . The green dashed lineis the Larson (1981) universal scaling relation between mass and radius ofmolecular clouds, shown for comparison. Figure 16.
The mass of the protostellar clumps as function of theirequivalent radius, colour-coded for the temperature of the dust envelope.There are ≃
43% of the protostellar clumps (456 out of 1056) above theKau ff mann & Pillai (2010) threshold. Σ ≃
60 g cm − . However the mass surface densityof the embedding clump, with a mass of ≃ ⊙ in ≃ . Σ ≃ . − . The dark cloud, with amass of ≃ ⊙ within ≃ . Σ ≃ .
25 g cm − . On average, high-mass star forming regionsassociated with IRDCs are usually observed within 0 . Σ
1g cm − (Tan et al. 2014). As discussed in the previous Section, aless stringent empirical threshold for high-mass star formation inclumps has been suggested by Urquhart et al. (2014) based on AT-LASGAL clump studies and corresponds to Σ > .
05 g cm − .In Figures 17 and 18 we plot the mass surface density-mass di-agrams for the starless and protostellar clumps respectively, colour- coded for the temperature of the dust envelope. The light-brownshaded area delimits the Σ > . − region. There are 144starless and 308 protostellar clumps, respectively 22% and 29% ofthe two samples, within this region where massive star formationis possible. Adopting the Urquhart et al. (2014) threshold limit in-creases these numbers to 241 starless and 542 protostellar clumps( ≃
36% and ≃
51% of the respective samples). Following Tan et al.(2014), we also show straight lines corresponding to constant ra-dius, constant escape speed and constant hydrogen number density.The apparent radius threshold at R ≃ v esc = (10 /α vir ) / σ , with α vir beingthe virial parameter and σ the velocity dispersion of the clump.Assuming a virial parameter α vir = σ = / s (e.g.Henshaw et al. 2013; Kau ff mann et al. 2013) the escape speed is v esc ≃ / s. Therefore, for a typical high-mass star form-ing clump with mass of M ≃ ⊙ (Tackenberg et al. 2012;Urquhart et al. 2014), the clump is bound if its mass surface den-sity reaches Σ ≃ . − . Low values of viral parameters havebeen observed in region of high-mass star formation (e.g. Tan et al.2013; Li et al. 2013). Assuming α vir = . v esc ≃ . / s and a M ≃ ⊙ clumpwill be bound only if reaches Σ ≃ . − or, equivalently, if itsmass is confined in a radius R ≃ Σ > . − , ≃
6% and ≃
7% of thetwo samples respectively. We find only 4 starless and 4 protostellarclumps with Σ > − .The hydrogen number density n H is analogous to constantfree-fall time, following the relation t f f = [3 π/ (32 G ρ )] / , where ρ is the volume density of the clump. From the diagram we note that,setting the high-mass star formation threshold at Σ > . − ,the minimum hydrogen number density of high-mass star-formingclumps is n H = cm − . This value corresponds to a free-fall timeof 1 . × year. Assuming as a lower limit for the mass surfacedensity Σ > . − to form high-mass stars, from the diagramwe deduce a minimum hydrogen number density of n H ≃ . × cm − , or a free-fall time of 3 . × year. With the surface densityproposed by (Urquhart et al. 2014), from our clumps distributionwe deduced a minimum hydrogen number density of n H ≃ × cm − , or a free-fall time of 5 . × year.For comparison we highlighted the regions in the plots oc-cupied by cores / clumps associated with a sample of 10 IRDCsselected by Butler & Tan (2012), the 51 massive star formingclumps identified by Mueller et al. (2002) and the CO molecularclouds identified in Roman-Duval et al. (2010). The SDC335 cen-tral clump and cloud values are also showed in the plot with darkgrey asterisks.
Figure 19 shows the FIR luminosity distributions of the clumps.The median values of the luminosities are L =
83 L ⊙ and L =
590 L ⊙ for starless and protostellar clumps respectively. This di ff erence inthe FIR luminosity is principally due to the emission arising fromthe warm core(s) in the protostars. In the starless clumps the lumi-nosity of the dust envelope is the only contribution to the bolomet-ric luminosity. On the other hand, the warm core(s) in the proto-stellar clumps contributes substantially to the integrated FIR lumi-nosity. Depending on the evolutionary state of a protostar there canalso be significant emission at 24 µ m or shorter wavelengths. The c (cid:13) , 1–19 catalogue of starless and protostellar clumps embedded in IRDCs. Figure 17.
Mass surface density as function of mass for starless clumps, colour coded for the temperature of the dust envelope. The light brown shaded areadefines the region of sources with surface density Σ > . − . The lines correspond to constant values of various parameters: red lines : constant radius; blue lines : constant H number density, analogous of constant free-fall time; green lines : constant escape velocity. Also shown are the regions corresponding toIRDC cores observed by Butler & Tan (2012, blue shaded area), the massive star formation clumps (and cores) observed by Mueller et al. (2002, red shadedarea) and the CO molecular clouds from the catalogue of Roman-Duval et al. (2010, green shaded area). The dark grey asterisks show the location of theSDC335 central clump and cloud (Peretto et al. 2013). estimated FIR luminosity is therefore a lower limit of the protostarsbolometric luminosity (e.g. Giannini et al. 2012).We identified 71 protostellar clumps ( ≃
7% of the total) withL > L ⊙ . There are 52 starless clumps with L > L ⊙ , relativelyhigh for starless clumps. The majority of them (30 out of 59) havea bolometric luminosity vs. envelope mass (L bol / M env )
1, in-dicative of objects without warm cores. These are among the mostmassive starless clumps identified in this catalogue. The remaining22 starless clumps with luminosity L > L ⊙ and L bol / M env > µ m sources not identified withour selection threshold (see Section 3.1) or associated with brightdi ff use 70 µ m emission. A potentially useful tool to infer the evolutionary properties of theyoung clumps is the L bol – M env diagram. This diagram has beensuccessfully used in the past to describe the evolutionary phasesof low-mass objects (Saraceno et al. 1996), and subsequently ex-tended to predict the evolutionary paths of high-mass objects(Molinari et al. 2008). A source evolves following specific tracks,depending on its initial mass and luminosity, and the model pre-dicts the mass and luminosity of the object when it reaches the zero age main sequence (ZAMS). For high-mass objects the paths fol-low the two-phases model of McKee & Tan (2003). According tothis model, when a cloud starts its gravitational collapse the massslightly decreases due the accretion and molecular outflows, whilethe luminosity of the object increases significantly, sustained by thecollapse. The source moves along almost vertical paths in the dia-gram. At the end of this phase the object is surrounded by an HIIregion, the luminosity remains constant while the mass is expelledthrough radiation and molecular outflows. The object then followsan almost horizontal path, which will ends when it becomes visiblein the optical and almost all the envelope has been expelled. Ac-cordingly to the evolutionary phase in the diagram, the high-massobjects are classified in analogy with the low-mass regime classifi-cation, from Class 0 to Class II prior to the ZAMS phase. This clas-sification can be misleading though, and some caveats of this modelhave to be kept in mind, in particular when applied to our sample:1) the transition from Class 0 to Class I and Class II sources in thelow-mass regime is not sharp. Sources classified as Class 0 basedon their NIR–MIR fluxes can instead be classified as Class I in theFIR (Hennemann et al. 2010; Elia et al. 2013); 2) the evolutionarytracks have been initially modelled for single cores, not for clumpsand 3) in high-mass regions the observed clumps can contain mul-tiple cores in di ff erent stages of evolution and the clump luminosityis likely dominated by the most evolved core(s). Nevertheless, theL bol – M env diagram has been used in the past to discuss the evo-lutionary tracks of YSOs with Herschel data in a wide range of c (cid:13) , 1–19 A. Traficante et al.
Figure 18.
Mass surface density as function of mass for protostellar clumps, colour coded for the temperature of the dust envelope. The shaded areas and linesare as shown in Figure 17.
Figure 19.
FIR luminosity distributions for the 1723 clumps, evaluated inthe range 70 λ µ m as detailed in the text. The median values areL =
83 L ⊙ and L =
590 L ⊙ for starless and protostellar clumps respectively.This di ff erence is a direct consequence of the internal luminosity of thecore(s) embedded in the protostellar clumps which are partly visible in the70 µ m emission. di ff erent conditions, both for low-mass (e.g. Bontemps et al. 2010;Hennemann et al. 2010) and for intermediate-to-high mass regime(e.g. Elia et al. 2010; Veneziani et al. 2013).The L bol – M env diagram for the starless clumps is shownin Figure 20. The distribution is superimposed to the models of Saraceno et al. (1996) and Molinari et al. (2008), which shows thetracks for stars with final masses of 6.5, 8, 13.5, 18 and 35 M ⊙ fromleft to right respectively. The best log-log fit for Class I and Class0 high-mass objects (Molinari et al. 2008) are also shown. The ma-jority of the sources lie below the best log-log fit of Class 0 ob-jects. Also, the distribution lies on average below the distribution ofyoung high-mass clumps identified in Urquhart et al. (2014), whichincludes clumps at di ff erent stages of evolution but does not includestarless ones. We find a mean L bol / M env ≃ .
1, compared with amean L bol – M env ≃
10 of Urquhart et al. (2014), consistent withthe majority of these starless clumps being in a very early stage oftheir evolution. Figure 21 shows the L bol – M env diagram for theprotostellar clumps. The distribution overlaps that of the starlessclumps, proving little discrimination between the relative evolu-tionary stages of the sources. However the mean L bol – M env ≃ . Comparing the number of starless to protostellar clumps providesan estimate of the lifetime of the starless phase. This is of particular c (cid:13) , 1–19 catalogue of starless and protostellar clumps embedded in IRDCs. Figure 20. L bol – M env diagram for starless clumps superimposed withthe model of Saraceno et al. (1996) and Molinari et al. (2008, green lines).The solid and dashed lines are the best log-log fit for Class I and Class 0sources respectively, extrapolated in the high-mass regime by Molinari et al.(2008). The red vertical line is our mass completeness limit, M =
105 M ⊙ .The sources are sparse in the diagram but the majority are below the Class0 fit, showing that they are in a very early stage of evolution. Figure 21.
The same diagram of Figure 20 but for protostellar clumps. interest for those regions most likely to form high mass stars, themost massive clumps with high surface densities. For clumps witha mass M >
500 M ⊙ and a surface density Σ > . − the ra-tio of the number of starless clumps to protostellar clumps is 0.46,which increases only slightly to 0.48 when considering clumps witha surface density Σ > . − . This suggests a lifetime for thestarless phase of about half that of the protostellar phase. Adopt-ing an estimate of few × yr for the lifetime of the embeddedphase of young massive stars (e.g. Davies et al. 2011), the starlessphase duration would be ∼ yr, in agreement with the lifetimeestimation of starless clumps identified in the ATLASGAL survey(Tackenberg et al. 2012). Since some starless clumps may containcurrently unidentified protostars as suggested by the high luminos-ity to mass ratio for some of the apparently starless clumps (Section4.4), this lifetime represents an upper limit. We have produced a catalogue of starless and protostellar clumpsidentified in a sample of 3493 IRDCs extracted from thePeretto & Fuller (2009) catalogue in the region of the GalacticPlane delimited by 15 ◦ l ◦ , | b | ◦ . Using the Herschel observations acquired as part of the Hi-GAL survey, we first iden-tified the FIR counterparts of the IRDCs in the Hi-GAL data at 70,160, 250 and 350 µ m. The compact sources were extracted from theHi-GAL IRDCs counterparts using Hyper , a new algorithm whichcombines the advantages of the aperture photometry with a min-imal use of source modelling to take into account the high vari-able background and the source crowding (Traficante et al. 2015).The sources were initially identified at 160 µ m and we found 5393160 µ m sources distributed in 1640 IRDCs ( ≃
47% of the initialsample). All the sources with a counterpart at 250 and 350 µ m,but no counterpart(s) at 70 µ m are identified as starless . If at leastone 70 µ m counterpart is present, they are identified as protostellar clumps. We identified 1056 protostellar in 586 IRDCs ( ≃ ≃ µ m with a single-temperature greybody model anda fixed spectral index of β = .
0. The adopted model is a goodapproximation to describe the cold envelopes with only small tem-perature gradients across the clumps, but it does not account for theemission arising from the central, warmer regions of protostars. Asa consequence it underestimates the average dust temperature (andtherefore overestimates the mass) of the protosteallr clumps. Themean equivalent radius is ≃ . ≃ . ff erent, as demonstrated bya KS test, however the mean temperature values di ff er for less than2 K, T = = ⊙ and 272 M ⊙ respectively, and extending up to ≃ and ≃ · M ⊙ for the more massive starless and pro-tostellar clumps respectively. The similarities in temperature andmass indicate that the dust envelope of the protostellar clumps hasbeen not yet warmed-up by the inner core(s) and it is relativelyuna ff ected by the gravitational collapse occurring in the core(s).Both these conclusions indicates that the population of protostellarclumps observed in this sample of IRDCs is very young.We demonstrated that, due to the heterogeneity of the cloudstructures, the mass completeness has to be assumed locally foreach cloud. We have however set a reference mass completeness ofM =
105 M ⊙ for the whole sample, which corresponds to the massabove which clumps can be detected in 90% of the clouds.The mass vs. radius distribution of the starless clump is usedto identify the clumps which may lead to high-mass stars, assumingthe Kau ff mann & Pillai (2010) empiric threshold for high-mass starformation. We show that almost one third of the starless clumps(171 out of 667), distributed in 130 IRDCs (only 4% of the wholesample), are above the high-mass stars threshold. In other wordsonly a small fraction of the clouds are likely to form massive stars.The environment of massive star formation has been furtherinvestigated through the mass vs. surface density diagram. Assum- c (cid:13) , 1–19 A. Traficante et al. ing that high-mass star formation occurs in regions with Σ > . − (Butler & Tan 2012; Tan et al. 2014) we identify 144 starlessand 308 protostellar clumps that will likely form high-mass stars.From the mass surface density diagram of the starless clumps weidentify the maximum free-fall time of our sample of t ≃ . × year. Fixing the mass surface density threshold to Σ > .
05 g cm − ,as proposed by Urquhart et al. (2014), this time rises to t ≃ . × year. The numbers of massive (M >
500 M ⊙ ), high surface densitystarless clumps and those already containing embedded sources in-dicate an upper limit lifetime for the starless massive clump phaseof ∼ years, similar to the free-fall time.The FIR luminosity distributions di ff ers with mean values ofL =
83 L ⊙ and L =
590 L ⊙ for starless and protostellar clumps re-spectively, principally because the protostellar clumps luminosityis highly influenced by the luminosity of the embedded core(s). Atotal of 71 protostellar clumps, ≃
7% of the sample, have FIR lumi-nosity > L ⊙ . The FIR luminosity is a good approximation forthe bolometric luminosity only for starless clumps, and we com-pared the starless L bol – M env diagram with the evolutionary mod-els for low- and high-mass objects developed by Saraceno et al.(1996) and Molinari et al. (2008). The starless clumps distributionis sparse but most of the sources are below the line which de-scribes the high-mass counterparts of low-mass Class 0 regime.Although a clear analogy with the low-mass classes classificationis not straightforward in the high-mass regime, the mean L bol / M env ≃ . / mm surveys of the Galactic Plane such as ATLASGAL and theBGPS.The catalogues are freely available at the webpage . ACKNOWLEDGEMENTS
AT and GAF acknowledge the support from the STFC consolidatedgrant ST / L000768 / REFERENCES
Aguirre J. E., Ginsburg A. G., Dunham M. K., et al. 2011, ApJS,192, 4Battersby C., Bally J., Ginsburg A., et al. 2011, A&A, 535, A128Beltr´an M. T., Olmi L., Cesaroni R., et al. 2013, A&A, 552, A123Benjamin R. A., Churchwell E., Babler B., et al. 2003, PASP, 115,953Bergin E. A., Tafalla M., 2007, ARAA, 45, 339Bernard J., Paradis D., Marshall D., et al. 2010, A&A, 518, L88 + Beuther H., Linz H., Tackenberg J., et al. 2013, A&A, 553, A115Bontemps S., Andr´e P., K¨onyves V., et al. 2010, A&A, 518, L85Butler M. J., Tan J. C., 2012, ApJ, 754, 5Carey S. J., Clark F. O., Egan M. P., et al. 1998, ApJ, 508, 721Davies B., Hoare M. G., Lumsden S. L., et al. 2011, MNRAS,416, 972 Dunham M. M., Crapsi A., Evans II N. J., et al. 2008, ApJS, 179,249Elia D., Molinari S., Fukui Y., et al. 2013, ApJ, 772, 45Elia D., Schisano E., Molinari S., et al. 2010, A&A, 518, L97Giannini T., Elia D., Lorenzetti D., et al. 2012, A&A, 1202, 1413Gri ffi n M. J., Abergel A., Abreu A., et al. 2010, A&A, 518, L3 + Hennemann M., Motte F., Bontemps S., et al. 2010, A&A, 518,L84Henshaw J. D., Caselli P., Fontani F., et al. 2013, MNRAS, 428,3425Hildebrand R. H., 1983, QJRAS, 24, 267Jackson J. M., Rathborne J. M., Shah R. Y., et al.. 2006, ApJS,163, 145Kau ff mann J., Pillai T., 2010, ApJ, 723, L7Kau ff mann J., Pillai T., Goldsmith P. F., 2013, ApJ, 779, 185K¨onyves V., Andr´e P., Men’shchikov A., et al. 2010, A&A, 518,L106Krumholz M. R., McKee C. F., 2008, Nature, 451, 1082Lada C. J., Lada E. A., 2003, ARA&A, 41, 57Larson R. B., 1981, MNRAS, 194, 809Launhardt R., Stutz A. M., Schmiedeke A., et al. 2013, A&A, 551,A98Li D., Kau ff mann J., Zhang Q., et al. 2013, ApJL, 768, L5Markwardt C. B., 2009, in Bohlender D. A., Durand D., Dowler P.,eds, ADASS XVIII Vol. 411 of ASPC, Non-linear Least-squaresFitting in IDL with MPFIT. p. 251McKee C. F., Tan J. C., 2003, ApJ, 585, 850Men’shchikov A., Andr´e P., Didelon P., et al. 2012, A&A, 542,A81Molinari S., Pezzuto S., Cesaroni R., et al. 2008, A&A, 481, 345Molinari S., Schisano E., Faustini F., et al. 2011, A&A, 530, A133Molinari S., Swinyard B., Bally J., et al. 2010a, A&A, 518, L100 + Molinari S., Swinyard B., Bally J., et al. 2010b, PASP, 122, 314Motte F., Zavagno A., Bontemps S., et al. 2010, A&A, 518, L77Mueller K. E., Shirley Y. L., Evans II N. J., Jacobson H. R., 2002,ApJS, 143, 469Olmi L., Angles-Alcazar D., Elia D., et al. 2013, A&A, 551, A111Ormel C. W., Min M., Tielens A. G. G. M., et al. 2011, A&A,532, A43Peretto N., Fuller G. A., 2009, A&A, 505, 405Peretto N., Fuller G. A., 2010, ApJ, 723, 555Peretto N., Fuller G. A., Duarte-Cabral A., et al. 2013, A&A, 555,A112Peretto N., Fuller G. A., Plume R., et al. 2010, A&A, 518, L98 + Pezzuto S., Elia D., Schisano E., et al. 2012, A&A, 547, A54Piazzo L., Ikhenaode D., Natoli P., et al. 2011, ITIP, 21, 3687Pilbratt G. L., Riedinger J. R., Passvogel T., et al. 2010, A&A,518, L1 + Poglitsch A., Waelkens C., Geis N., et al. 2010, A&A, 518, L2 + Preibisch T., Ossenkopf V., Yorke H. W., et al. 1993, A&A, 279,577Ragan S., Henning T., Krause O., et al. 2012, A&A, 547, A49Rathborne J. M., Jackson J. M., Simon R., 2006, ApJ, 641, 389Reid M. J., Menten K. M., Zheng X. W., et al. 2009, ApJ, 700,137Robitaille T. P., Whitney B. A., Indebetouw R., et al. 2007, ApJS,169, 328Roman-Duval J., Jackson J. M., Heyer M., et al. 2010, ApJ, 723,492Russeil D., 2003, A&A, 397, 133Russeil D., Pestalozzi M., Mottram J. C., et al. 2011, A&A, 526,A151 c (cid:13) , 1–19 catalogue of starless and protostellar clumps embedded in IRDCs. Sakai T., Sakai N., Foster J. B., et al. 2013, ApJL, 775, L31Saraceno P., Andre P., Ceccarelli C., et al. 1996, A&A, 309, 827Schuller F., Menten K. M., Contreras Y., et al. 2009, A&A, 504,415Simon R., Rathborne J. M., Shah R. Y., et al. 2006, ApJ, 653, 1325Tackenberg J., Beuther H., Henning T., et al. 2012, A&A, 540,A113Tan J. C., Beltr´an M. T., Caselli P., et al. 2014, Protostars andPlanets VI, pp 149–172Tan J. C., Kong S., Butler M. J., et al. 2013, ApJ, 779, 96Traficante A., Calzoletti L., Veneziani M., et al. 2011, MNRAS,416, 2932Traficante A., Fuller G. A., Pineda J. E., et al. 2015, A&A, 574,A119Urquhart J. S., Moore T. J. T., Csengeri T., et al. 2014, MNRAS,443, 1555Veneziani M., Elia D., Noriega-Crespo A., et al. 2013, A&A, 549,A130Wilcock L. A., Kirk J. M., Stamatellos D., et al. 2011, A&A, 526,A159Wilcock L. A., Ward-Thompson D., Kirk J. M., et al. 2012, MN-RAS, 422, 1071 c (cid:13)000