GASTON: Galactic Star Formation with NIKA2. Evidence for the mass growth of star-forming clumps
A. J. Rigby, N. Peretto, R. Adam, P. Ade, M. Anderson, P. André, A. Andrianasolo, H. Aussel, A. Bacmann, A. Beelen, A. Benoît, S. Berta, O. Bourrion, A. Bracco, M. Calvo, A. Catalano, M. De Petris, F.-X. Désert, S. Doyle, E. F. C. Driessen, P. García, A. Gomez, J. Goupy, F. Kéruzoré, C. Kramer, B. Ladjelate, G. Lagache, S. Leclercq, J.-F. Lestrade, J. F. Macías-Pérez, P. Mauskopf, F. Mayet, A. Monfardini, L. Perotto, G. Pisano, N. Ponthieu, V. Revéret, I. Ristorcelli, A. Ritacco, C. Romero, H. Roussel, F. Ruppin, K. Schuster, S. Shu, A. Sievers, C. Tucker, E. J. Watkins
MMNRAS , 1–21 (2020) Preprint 17 February 2021 Compiled using MNRAS L A TEX style file v3.0
GASTON: Galactic Star Formation with NIKA2. Evidence for the massgrowth of star-forming clumps
A. J. Rigby, ★ N. Peretto, † R. Adam, P. Ade, M. Anderson, P. André, A. Andrianasolo, H. Aussel, A. Bacmann, A. Beelen, A. Benoît, S. Berta, O. Bourrion, A. Bracco, M. Calvo, A. Catalano, M. De Petris, F.-X. Désert, S. Doyle, E. F. C. Driessen, P. García, , A. Gomez, J. Goupy, F. Kéruzoré, C. Kramer, , B. Ladjelate, G. Lagache, S. Leclercq, J.-F. Lestrade, J. F. Macías-Pérez, P. Mauskopf, , F. Mayet, A. Monfardini, L. Perotto, G. Pisano, N. Ponthieu, V. Revéret, I. Ristorcelli, A. Ritacco , C. Romero, H. Roussel, F. Ruppin, K. Schuster, S. Shu, A. Sievers, C. Tucker, and E. J. Watkins , Affiliations are listed at the end of the paper
Accepted 2021 January 20. Received 2021 January 20; in original form 2020 July 21
ABSTRACT
Determining the mechanism by which high-mass stars are formed is essential for our understanding of the energy budget andchemical evolution of galaxies. By using the New IRAM KIDs Array 2 (NIKA2) camera on the Institut de Radio AstronomieMillimétrique (IRAM) 30-m telescope, we have conducted high-sensitivity and large-scale mapping of a fraction of the Galacticplane in order to search for signatures of the transition between the high- and low-mass star-forming modes. Here, we present thefirst results from the Galactic Star Formation with NIKA2 (GASTON) project, a Large Programme at the IRAM 30-m telescopewhich is mapping ≈ of the inner Galactic plane (GP), centred on ℓ = . ◦ , 𝑏 = . ◦
05, as well as targets in Taurus andOphiuchus in 1.15 and 2.00 mm continuum wavebands. In this paper we present the first of the GASTON GP data taken, andpresent initial science results. We conduct an extraction of structures from the 1.15 mm maps using a dendrogram analysis and, bycomparison to the compact source catalogues from
Herschel survey data, we identify a population of 321 previously-undetectedclumps. Approximately 80 per cent of these new clumps are 70 µ m-quiet, and may be considered as starless candidates. Wefind that this new population of clumps are less massive and cooler, on average, than clumps that have already been identified.Further, by classifying the full sample of clumps based upon their infrared-bright fraction – an indicator of evolutionary stage –we find evidence for clump mass growth, supporting models of clump-fed high-mass star formation. Key words:
Galaxy: disc – stars: formation – stars: massive – ISM: evolution – ISM: structure – surveys
The stellar initial mass function (IMF) is an essential ingredient incosmological simulations of galaxy formation and evolution, and itsorigin remains one of the most fundamental and important questionsin the field of astronomy. In the post-
Herschel era, a link betweenfilamentary structures in the interstellar medium (ISM) and star-formation has been firmly established (Molinari et al. 2010; Andréet al. 2010, 2014). At the low-mass end of the IMF, most stars withmasses in the range 𝑚 ∗ ∼ . 𝑀 (cid:12) are found to formwithin dense filaments (Könyves et al. 2015, 2020; Marsh et al.2016). If the mass-per-unit-length exceeds a critical value set by thelocal sound speed, gravitational instabilities can develop, leadingto the fragmentation of the filament, and the formation of stellarsystems from this mass reservoir (Inutsuka & Miyama 1997; Clarke ★ E-mail: rigbya@cardiff.ac.uk † E-mail: peretton@cardiff.ac.uk et al. 2017). This core-fed scheme of fragmentation of trans- orsupercritical virialized filaments may explain the origin of the IMFfrom ∼ . ∼ 𝑀 (cid:12) (Lee et al. 2017; André et al. 2019). However,outside this range, the proposed formation mechanisms for browndwarfs (Whitworth et al. 2007; Luhman 2012), and the formation ofhigh-mass ( 𝑚 ∗ (cid:38) 𝑀 (cid:12) ) stars (Tan et al. 2014; Motte et al. 2018),remain controversial.It is well established that the Jeans-like fragmentation of molec-ular clouds is unable to produce dense cores which are sufficientlymassive to be the progenitors of high-mass stars (Bontemps et al.2010; Sanhueza et al. 2017), and so the formation pathway mustincorporate additional mechanisms. There are broadly two familiesof models for high-mass star formation that remain actively debatedwithin the field: ones in which the formation of high-mass stars re-sembles a scaled-up version of the low-mass star formation models(e.g. McKee & Tan 2003), where the high-mass protostellar objectaccretes material from a compact ( (cid:54) . © a r X i v : . [ a s t r o - ph . GA ] F e b A. J. Rigby et al. grow in mass as a result of the large-scale ( (cid:62) core-fed and clump-fed scenarios, respectively (Wang et al. 2010).A growing weight of observational evidence supports a picturein which large-scale gravitational collapse, resulting in large infallrates, plays a key role in the formation of high-mass stars. Systemswith the so-called ‘hub-filament’ configuration (Myers 2009) arealso routinely found in high-spatial resolution observations of high-mass clumps, with stellar protoclusters (Schneider et al. 2012; Liuet al. 2016), or the most massive dense cores found at the filamentintersections (e.g. Peretto et al. 2013, 2014). The filaments withinthese systems have also been seen to exhibit longitudinal velocitygradients that could indicate large-scale gas flows (Kirk et al. 2013;Peretto et al. 2014; Williams et al. 2018; Lu et al. 2018; ? ) thatsupply additional kinetic support at the hub centre, whilst funnellingsufficient mass to the high-mass cores on short timescales. However,it is neither clear whether longitudinal filamentary accretion andglobal collapse in high-mass star-forming hub-filament systems isuniversal, nor at which masses the transition from the solar-type starforming mode to the high-mass mode occurs.While the execution of such high-spatial resolution studies acrossthe entire Galaxy remain unfeasible, large and unbiased samples ofdense clumps now exist from single dish Galactic Plane surveys(Schuller et al. 2009; Aguirre et al. 2011; Moore et al. 2015; Moli-nari et al. 2016a). The Apex Telescope Large-Area Survey of theGalaxy (ATLASGAL; Schuller et al. 2009) at 870 µ m identifiedall of the high-mass clumps ( 𝑀 > 𝑀 (cid:12) ) within the Galaxy(Urquhart et al. 2018), and Jackson et al. (2019) showed that themajority of high-mass clumps are undergoing gravitational collapse.However, limitations in sensitivity or angular resolution prevent thesesurveys from being able to detect the precursors of intermediate-massstar-forming clumps, where any transition between the core-fed andclump-fed regimes is expected to occur. Studies of clump morphol-ogy with higher-angular resolution single dish facilities may allowprogress on this front, (e.g. Rigby et al. 2018), but large samplesmust still be acquired.In this paper we present the first results from GASTON, a new200-hour Large Programme (PI: Peretto) being undertaken at theIRAM 30-m telescope using the NIKA2 camera (Bourrion et al.2016; Calvo et al. 2016; Adam et al. 2018; Perotto et al. 2020).The GASTON project is comprised of three parts aimed at inves-tigating the enigmatic origin of the IMF from different angles: i) a ∼ survey of a section of the inner Galactic plane searchingfor the transition between the solar-type and high-mass star form-ing scenarios; ii) a search for pre-brown dwarf cores, iii) a studyof dust properties in nearby resolved pre-stellar cores. NIKA2 is astate-of-the-art focal plane array at the Institut de Radio AstronomieMillimétrique (IRAM) 30-m telescope at Pico Veleta in Spain, whichcompleted its commissioning campaign in April 2017. NIKA2 ob-serves in 260 GHz and 150 GHz (hereafter 1.15 mm and 2.00 mm)continuum wavebands simultaneously by means of a dichroic mir-ror, with angular resolutions of 11.1 and 17.6 arcsec, respectively.There are two 1.15 mm arrays, each made up of 1140 kinetic induc-tance detectors (KIDs), and there are 616 detectors on the 2.00 mmarray, filling the 6 . (cid:48) / at 1.15 mm and 9 mJy s / at 2.00 mm which,when combined with the large instantaneous FoV, result in mappingspeeds that are an order of magnitude larger than the previous gener-ation of instrumentation in operation at the IRAM 30-m telescope. In this paper, we describe the observations and present the firstresults from the GASTON GP project, in which we explore thepotential for the NIKA2 observations to identify signatures of thecore-fed or clump-fed scenarios of high-mass star formation. Thepaper is structured in the following way. In Sect. 2, we describe theobservations and data reduction, while in Sect. 3 we describe ouranalysis of the 1.15 mm data and present the results, in Sect. 4. Wesummarise our results in Sect. 5 The GASTON GP field, centred on ℓ = . ◦ 𝑏 = . ◦
05, was selectedfor the project due to its extremely high density of compact dustcontinuum sources, as identified by the
Herschel infrared GalacticPlane Survey (Hi-GAL; Molinari et al. 2016a), as well as a largeof number of infrared-dark clouds (IRDCs; Peretto & Fuller 2009).The GP part of GASTON has been allocated a total of 73.6 hoursof telescope time. The approximately 2 deg field was divided intothree sub-maps, centred on ℓ = . ◦
3, 23 . ◦
9, and 24 . ◦
5, and 𝑏 = . ◦ ± ◦ with respect to the normal to the Galactic plane, asused by ATLASGAL (Schuller et al. 2009), but a ± ◦ strategy wasfinally adopted for most scans, representing a compromise betweenreducing striping artefacts from data reduction, and maintaining agood mapping efficiency. For the first of the observing runs (N2R12and N2R14), scans were made with a length of 78 arcmin, with 195arcsec row spacing (i.e. one half of the 6 . (cid:48) 𝜏
225 GHz = .
08 to 0.44, and a mean value of 0.19, as measuredby the IRAM 30-m taumeter. The mean source elevation was 39 . ◦ (cid:38)
80 per cent) of the observations were obtainedbetween the hours of 08:00–14:00, indicating that the primary beamsare likely to be relatively stable around their nominal values of 11.1(1.15 mm) and 17.6 (2.00 mm) arcsec (Perotto et al. 2020, se their Image generated using the python package multicolorfits: https://github.com/pjcigan/multicolorfits
MNRAS , 1–21 (2020)
ASTON: Galactic Star Formation with NIKA2 . ◦ . ◦ . ◦ . ◦ . ◦ . ◦ . ◦ . ◦ − . ◦ Galactic Longitude G a l a c t i c L a t i t ud e NIKA2 f.o.v.
Figure 1.
Scanning strategy for the GASTON GP field. Rectangles show the area covered by the central pixel of the arrays in the various scans centred onthe white crosses. Dotted white rectangles show the initial strategy, and solid white rectangles show the final adopted strategy. The background image is afour-colour composite consisting of Herschel
Hi-GAL (Molinari et al. 2016a) 70 µ m, 160 µ m, and 350 µ m images in cyan, yellow, and magenta, and NIKA21.15 mm signal-to-noise ratio (to exclude noisy edges) in orange (this study). The circle in the lower left corner shows the extent of the NIKA2 FoV. Fig. 12). The remainder were observed between 15:00–19:00, duringwhich time the primary beams tend to degrade slightly to ∼ . The data were reduced using the IDL pipeline developed by the NIKACore Team, which converts the raw time-series data into calibratedmaps. While a detailed description of this pipeline will be presentedin Ponthieu et al. (in prep), we summarise the main steps here.First, individual KIDs that do not meet the performance criteria,are masked from the timelines, as is also done for samples associ-ated with detected cosmic ray impacts. Next, the pointing informationfrom the telescope control system is used in conjunction with a recordof the positional offsets of each KID with respect to a reference po-sition to determine the position of each KID on the sky as a functionof time. The KIDs are then inter-calibrated, using coefficients deter-mined from their relative gains, before applying a correction for theinstantaneous atmospheric opacity, and finally applying an absolutecalibration determined for the specific observing run (see Sect. 2.3).At this stage, a model of the correlated low-frequency noise com-ponents, which consist of electronic noise that is coupled to the read-out electronics of subsets of KIDs, and atmospheric noise, whichis common to all KIDs, is created; this is the so-called ‘common-mode’. We use the ‘Most Correlated Pixels’ method (Perotto et al.2020), which works by first discarding any samples in each timeline that, when combined with the telescope pointing information and asource mask, are thought to coincide with an astrophysical source.Next, cross-correlation coefficients between all KIDs are determinedand, for each KID, a model of the low-frequency noise is built byco-adding the timelines of that KID and the 15 most correlated KIDs.A linear fit of this common mode is made and subtracted from eachtimeline, leaving, in principle, the true astrophysical signal.Since the GASTON GP field is known to contain a large number ofbright and diffuse structures with complex morphologies, we adoptan iterative method to define the source mask used in the calculationof the common mode. For the initial iteration, we use a source maskcreated from a Hi-GAL map of the region at 500 µ m (Molinari et al.2016a). On subsequent iterations, the source mask is determined fromthe astrophysical signal from the previous iteration, masking out allsamples where the signal-to-noise ratio (S/N) is greater than 2, andthe process is repeated to improve the estimate of the common mode,and thereby the astrophysical signal. We used a total of 55 iterations,at which point, the change in the flux is negligible compared to themap produced by the penultimate iteration.Finally, the timelines, when combined with the telescope pointing The source mask is created by first rescaling the 500 µ m Hi-GAL mapto 1.15 mm, assuming a greybody spectral energy distribution with a dustemissivity spectral index of 𝛽 = . − . MNRAS000
Hi-GAL (Molinari et al. 2016a) 70 µ m, 160 µ m, and 350 µ m images in cyan, yellow, and magenta, and NIKA21.15 mm signal-to-noise ratio (to exclude noisy edges) in orange (this study). The circle in the lower left corner shows the extent of the NIKA2 FoV. Fig. 12). The remainder were observed between 15:00–19:00, duringwhich time the primary beams tend to degrade slightly to ∼ . The data were reduced using the IDL pipeline developed by the NIKACore Team, which converts the raw time-series data into calibratedmaps. While a detailed description of this pipeline will be presentedin Ponthieu et al. (in prep), we summarise the main steps here.First, individual KIDs that do not meet the performance criteria,are masked from the timelines, as is also done for samples associ-ated with detected cosmic ray impacts. Next, the pointing informationfrom the telescope control system is used in conjunction with a recordof the positional offsets of each KID with respect to a reference po-sition to determine the position of each KID on the sky as a functionof time. The KIDs are then inter-calibrated, using coefficients deter-mined from their relative gains, before applying a correction for theinstantaneous atmospheric opacity, and finally applying an absolutecalibration determined for the specific observing run (see Sect. 2.3).At this stage, a model of the correlated low-frequency noise com-ponents, which consist of electronic noise that is coupled to the read-out electronics of subsets of KIDs, and atmospheric noise, whichis common to all KIDs, is created; this is the so-called ‘common-mode’. We use the ‘Most Correlated Pixels’ method (Perotto et al.2020), which works by first discarding any samples in each timeline that, when combined with the telescope pointing information and asource mask, are thought to coincide with an astrophysical source.Next, cross-correlation coefficients between all KIDs are determinedand, for each KID, a model of the low-frequency noise is built byco-adding the timelines of that KID and the 15 most correlated KIDs.A linear fit of this common mode is made and subtracted from eachtimeline, leaving, in principle, the true astrophysical signal.Since the GASTON GP field is known to contain a large number ofbright and diffuse structures with complex morphologies, we adoptan iterative method to define the source mask used in the calculationof the common mode. For the initial iteration, we use a source maskcreated from a Hi-GAL map of the region at 500 µ m (Molinari et al.2016a). On subsequent iterations, the source mask is determined fromthe astrophysical signal from the previous iteration, masking out allsamples where the signal-to-noise ratio (S/N) is greater than 2, andthe process is repeated to improve the estimate of the common mode,and thereby the astrophysical signal. We used a total of 55 iterations,at which point, the change in the flux is negligible compared to themap produced by the penultimate iteration.Finally, the timelines, when combined with the telescope pointing The source mask is created by first rescaling the 500 µ m Hi-GAL mapto 1.15 mm, assuming a greybody spectral energy distribution with a dustemissivity spectral index of 𝛽 = . − . MNRAS000 , 1–21 (2020) A. J. Rigby et al. information, are projected (tangential projection) onto a pixel gridwith 2.5 arcsec pixels. Inverse-variance noise weighting is used toassign the data samples onto the pixel grid, using a nearest grid pointmethod, and to combine the maps associated with each individualscan into a mosaic of the GP field.The correlated noise removal is the origin of the spatial filteringwithin the reduced maps, which is common to all submillimetre andmillimetre-continuum imaging from ground-based telescopes. Thelevel of spatial filtering applied here is typically on the order of theinstantaneous FoV which, for NIKA2, is approximately 6 . (cid:48)
5. We donot calculate the pipeline transfer function that describes the ratio ofthe power spectrum within the reduced image to an estimate of theabsolute source power spectrum over the sky region in this paper. Anexample of the 2.00 mm NIKA2 transfer function towards a galaxycluster may be found in Ruppin et al. (2018, see their Fig. 3), whofound that around 95 per cent of emission at angular scales betweenthe 2.00 mm beam size and the FoV is recovered, falling off rapidlyat larger scales, and we expect a similar level of recovery here.
Uranus and Neptune are used as the primary flux calibrators, andso colour corrections must be applied to our photometric maps toaccount for the difference between the spectral energy distributions(SEDs) of those planets, and the approximate SEDs of the sourcesis the NIKA2 imaging. We apply colour correction factors of 1.02and 0.97 to the 1.15 and 2.00 mm bands, respectively, which wereinterpolated from Table 12 of Perotto et al. (2020), assuming an SEDof the form 𝐼 𝜈 ∝ ( 𝜈 / 𝜈 ) + 𝛽 , with the mean Galactic plane value forthe dust emissivity spectral index of 𝛽 = . We make use of two spectral line data sets in order to determinevelocities to continuum sources in Sect. 3.3. Firstly, we use CO (1–0) data (110.201 GHz) from the FOREST unbiased Galactic planeimaging survey with the Nobeyama 45 m telescope (FUGIN; Umem-oto et al. 2017), which have an angular resolution of 21 arcsec, and anrms of 𝜎 ( 𝑇 ∗ A ) = .
87 K per 0.65 km s − velocity channel. The spatialextent of the FUGIN data covers the entire GASTON GP field, andthe velocity range of − < 𝑣 LSR <
200 km s − , covers all of theGalactic spiral arms along this sight-line (Dame et al. 2001).We also make use of data from the pilot study for the RadioAmmonia Mid-Plane Survey (RAMPS; Hogge et al. 2018), a Galacticplane survey of ammonia and water maser emission carried out withthe Green Bank Telescope. The L23 and L24 regions of the pilot studyhave a considerable overlap with the GASTON GP field, spanning22 . ◦ (cid:38) ℓ (cid:46) . ◦
5, with | 𝑏 | (cid:46) . ◦
4, and we make use of the moment1 maps of NH (1,1) inversion emission at 23.694 GHz. These datahave an angular resolution of 32 arcsec, and 0.018 km s − spectralchannels.Data from Hi-GAL (Molinari et al. 2016a) at 160 and 250 µ m areused in Sect. 3.5 for colour temperature determinations. The 160 µ m PACS (Poglitsch et al. 2010) data have an angular resolution of 12arcsec, and the 250 µ m SPIRE (Griffin et al. 2010) data have anangular resolution of 18 arcsec. In both cases the uncertainties aredominated by calibration, which we take as 7 per cent for PACS and 6.5 per cent for SPIRE , including a 1 per cent contribution forextended sources due to uncertainty in the SPIRE beam.We also use 8 µ m data from the Galactic Legacy Infrared Mid-Plane Survey Extraordinaire (GLIMPSE; Benjamin et al. 2003;Churchwell et al. 2009) in Sect. 4.3. These Spitzer data have anangular resolution of 2.0 arcsec.
In Fig. 2, we present the NIKA2 photometric maps, after croppingto the largest contiguous rectangular field where the sensitivity isapproximately uniform. The total integration time obtained so far is ∼
30 hours in 108 scans, corresponding to roughly 40 per cent ofwhat will be acquired upon the completion of GASTON.Although we do not use the 2.00 mm map in the analysis presentedin this paper, we present it in Fig. 2 for the sake of completeness, andto display the data quality. We will make use of these data in futureGASTON studies and, in particular, the 1.15 and 2.00 mm data willbe used in combination to study the dust emissivity spectral index, 𝛽 , as in Rigby et al. (2018). A source extraction was carried out on the 1.15 mm maps us-ing the python astrodendro package, which uses a dendrogram(Rosolowsky et al. 2008) scheme to segment the emission whilemaintaining the ability to track its hierarchical structure. The den-drogram separates emission features when they are considered tobe significant given their size, and difference in their intensity withrespect to the background and to overlapping emission features.The 1.15 mm map was first cropped to 22 . ◦ (cid:54) ℓ (cid:54) . ◦
85 and − . ◦ (cid:54) ℓ (cid:54) . ◦
65 (as in Fig. 2) in order to exclude the noisieredge regions. The map was then smoothed to 13 arcsec resolutionin order to increase the S/N. Measuring the sensitivity in such deepmaps of the inner Galactic plane is problematic because of sourceconfusion, and so several methods were used to measure the rmsnoise level. The preferred method was to perform a Gaussian fit to thedistribution of pixel values taken from the null maps within a sampleof four octagonal sky apertures with radii of ∼ − .Based on the estimated rms noise level, we set the minimum value MNRAS , 1–21 (2020)
ASTON: Galactic Star Formation with NIKA2 . ◦ . ◦ . ◦ . ◦ . ◦ . ◦ − . ◦ Galactic Longitude G a l a c t i c L a t i t ud e F l u x d e n s i t y [ m J y / b e a m ] . ◦ . ◦ . ◦ . ◦ . ◦ . ◦ − . ◦ Galactic Longitude G a l a c t i c L a t i t ud e F l u x d e n s i t y [ m J y / b e a m ] Figure 2.
GASTON GP field at 1.15 mm (top panel) and 2.00 mm (bottom panel), which have FWHM resolutions of 11.1 and 17.6 arcsec, respectively. TheFoV and primary beam sizes are shown in the bottom left corners as opaque white, and black circles, respectively. MNRAS000
GASTON GP field at 1.15 mm (top panel) and 2.00 mm (bottom panel), which have FWHM resolutions of 11.1 and 17.6 arcsec, respectively. TheFoV and primary beam sizes are shown in the bottom left corners as opaque white, and black circles, respectively. MNRAS000 , 1–21 (2020)
A. J. Rigby et al. ( min_value ) above which emission is considered to be real, and theminimum difference between substructures ( min_delta ) to 3 timesthe global rms, while the minimum structure size ( min_npix ) to be16 pixels – equivalent to half of the 13-arcsec beam area. The latterchoice was made for the minimum size criterion because pure pointsources will be reported as smaller than the beamsize due to theclipping of the wings that extend below the minimum intensity level(as in the ‘bijection’ scheme described in Rosolowsky et al. 2008),and so would be filtered from the data if the full beam area was used.This is especially true for low S/N sources.The application of a single average rms noise level over the en-tire area results in a small number of spurious detections in thenoisier-than-average regions and, conversely, some undetected com-pact sources are visible in the residuals in the regions where thesensitivity is greater than this average value. To enable us to elimi-nate spurious sources, we calculate the local S/N of each source ineach waveband. As mentioned earlier, the null maps contain someresidual positive or negative emission, and are unsuitable for use asnoise maps for all sources. We therefore produce a noise map in eachwaveband, that is calculated from the weight maps produced by theIDL pipeline. The per-pixel weights are proportional to the inversevariance of the flux densities measured at each position on the pixelgrid from the contributing observations. The noise at each pixel po-sition, 𝜎 𝑖 , is therefore calculated as 𝜎 𝑖 = 𝜎 √︁ 𝑤 / 𝑤 𝑖 , where 𝜎 isthe rms measured from the pixels within the four sky apertures, 𝑤 is the mean value of the pixel weights within the four sky apertures,and 𝑤 𝑖 is the per-pixel weight. We apply a ‘pruning’ scheme in thedendrogram construction such that any objects that do not have alocal S/N that exceeds 6 – in keeping with the dendrogram’s inputparameters – in at least one of their constituent pixels are removed.In the dendrogram nomenclature, emission features may describedas ‘trunks’, ‘branches’, or ‘leaves’, depending on their position withinthe hierarchy. The trunks are the lowest-lying structures in the hier-archy, at the minimum detection threshold and thus describe themaximum extent of any isolated emission region, while branches areat a higher contour level within a trunk, and may contain multipleleaves. In the following analysis, we refer to the dendrogram leaves– those emission features that contain no further discernible sub-structure – as ‘clumps’, whereas we use the term ‘source’ to referto any emission feature extracted using the dendrogram, regardlessof its position within the hierarchy. That is to say that a source maycontain multiple clumps. The dendrogram extraction found a total of2446 sources of all kinds, consisting of 488 isolated emission struc-tures (trunks) with 1467 clumps (leaves). We summarise the basicproperties – flux densities and source sizes – in Appendix B. Determining the systemic velocity, and thereby a kinematic distance,for each extracted source is a necessary step in calculating manyof its physical properties. However, due to both the high sensitivityof the GASTON GP maps at 1.15 mm, and the particularly compli-cated line-of-sight for the survey region, source confusion presents asignificant challenge in their determination. We therefore adopted aprocedure with several stages to establish the most reliable possiblevelocities, which are summarised in this Section, and further detailsmay be found in Appendix C.To determine a velocity for each GASTON source, first, an inte-grated CO (1–0) spectrum is extracted from the FUGIN data of allpixels that lie within the source footprint. Many of the spectra ex-hibit multiple emission features, and so the velocity with the greatestintensity is assigned as the centroid velocity, 𝑣 cen , of each source. In cases where the line of sight contains multiple velocity components,this velocity is assumed to trace the highest-column density materialwhich is most likely to be associated with the dust continuum source.A quality flag was assigned to each velocity assignment, with a valueof 3 given to the most robust assignments, and 1 to the poorest, anda value of 2 is given for intermediate quality assignments.The main advantage of this method is its simplicity, but it mayobviously provide wrong velocity assignments in some cases, pri-marily due to the difference in densities traced by the molecular andcontinuum data. In principle, the dust continuum emission traces thefull column density of molecular hydrogen, but the spatial filteringapplied by the data reduction pipeline removes much of the emissionfrom diffuse gas, leaving the more compact, and higher column-density gas. Observations of dense gas tracers would provide moresuitable velocities for the GASTON GP sources than CO (1–0),and so, as a second stage, we use data from the RAMPS pilot study(Hogge et al. 2018) of NH (1,1), which has a critical density of 𝑛 crit ∼ × cm − to determine dense-gas velocities. We adopt ve-locities for 517 (21 per cent) of the sources from the first componentfrom the line-fitting procedure, and a further 513 (21 per cent) fromthe first-moment maps of Hogge et al. (2018), assigning a qualityflag of 3 for all ammonia-derived velocities.The hierarchical structure of the 1.15 mm emission is recordedwithin the dendrogram catalogue, and since each source now hasan individual velocity assignment, along with a quality flag, thestructure can be used to inform us about likelihood of the moredubious velocity assignments being the correct ones. For example,if a velocity assignment of 50.5 km s − is given for a source witha poor quality flag, but its parent structure has a robust velocityassignment of 50.9 km s − , then it is considered to be part of thatsame velocity group. If a poor velocity assignment is made to asource at a significantly different velocity to its neighbours withinthe hierarchy that have robust assignments, then we adopt the medianvelocity of those neighbours for that source. In this way, the velocityassignments for all sources (where possible) with a quality flag ofeither 1 or 2 were refined, and we adopt the quality flag of the mostrobust velocity within that group for all its constituent sources. Werefer to the velocities refined in this way as 𝑣 group . In Tab. 1 wepresent the distribution of quality flags assigned to the velocitiesdetermined and refined in this Section.Finally, in order to ensure that velocity-coherent structures are re-tained, we perform a friends-of-friends grouping. Clusters of sourceswere identified by recursively linking groups that lie within a toler-ance of 0 . ◦
06 in the ℓ and 𝑏 axes, and within 2.5 km s − in line-of-sightvelocities, 𝑣 group , which approximates the characteristics of small(10 pc-diameter) molecular clouds (e.g. Roman-Duval et al. 2010) ata distance of 5 kpc – the median near-kinematic distance for the re-vised source velocities along this sight-line. This allows, for example,the prominent large-scale filament that extends from [ ℓ, 𝑏, 𝑣 group ] = [ . ◦ , . ◦ ,
94 km s − ] to [ . ◦ , . ◦ ,
98 km s − ] to be identi-fied as a single cluster. 2249 (92 per cent) of all GASTON sources areconnected to one of the 211 clusters identified in this way, which con-tain up to 233 individual sources, with a median size of 25 sources.For each cluster, we also calculate the median group velocity, 𝑣 cluster ,and the associated standard deviation, 𝜎 cluster , using only the sourceswhich have the most reliable velocity flags from that cluster.The usage of 21 arcsec-resolution spectral line data with our 13arcsec resolution element means that there is an element of beam-averaging that is likely to blend together velocity components thatwould be distinct at matching resolution. We have not made useof available survey data in CO (3–2) (Dempsey et al. 2013) at16.6 arcsec resolution, due to optical depth considerations. While
MNRAS , 1–21 (2020)
ASTON: Galactic Star Formation with NIKA2 Table 1.
Distribution of quality flags for each stage in the velocity determination process. For each flag value, the total number of sources with that value ( 𝑁 )and the percentage of the sample made up by those sources (%) is given.Stage Description Flag value3 2 1 𝑁 % 𝑁 % 𝑁 %i) FUGIN CO (1–0) 1685 68.9 298 12.2 463 18.9ii) RAMPS NH (1,1) 1916 78.3 165 6.7 365 14.9iii) Dendrogram refinement 2169 88.7 75 3.1 202 8.3 new facilities (e.g. Klaassen et al. 2019; Stanke et al. 2019) andinstrumentation (e.g. Frayer et al. 2020) hold much promise, unbiasedGalactic plane surveys in dense gas tracers such as N H + , HCN, NH or even C O at sufficiently high sensitivity and angular resolutionare not currently available, and the FUGIN and RAMPS data presentthe best data for the moment. We note that CHIMPS2 (Eden et al.2020) aims to cover the GASTON GP field in CO (3–2) which, atan angular resolution of 15 arcsec and with a higher-critical densitymolecular gas tracer than that of the FUGIN, should complementGASTON upon its completion, though it is limited to | 𝑏 | < . ◦
5. Wenote that we do not use data from the Galactic Ring Survey (GRS;Jackson et al. 2006), which have velocity resolution and sensitivitysuperior to the FUGIN data, because the higher angular resolutionof the latter are more important in this case.
Distances to sources lying within the Galactic plane are typicallydetermined by measuring their radial velocity, and by assuming thatthe source is following a circular orbit described by a model of theGalactic rotation curve. For sources lying within the Solar Circle,each radial velocity corresponds to two kinematic distance solutions.In order to determine whether a source lies at the near- or far- kine-matic distance, one either requires additional information, or someassumptions must be made. We determined kinematic distance solu-tions to each source using the Galactic rotation model of Reid et al.(2019) based on their ℓ, 𝑏, 𝑣 group coordinates. As a first estimate, weuse version 2.4.1 of the Bayesian distance calculator (Reid et al.2016), adopting the recommended weightings for the priors basedupon the spiral arm model, Galactic plane scale height, and proxim-ity to sources with a measured parallax. This gives an independentdistance estimate to each source (regardless of position within thehierarchy), but we note that this preferentially concentrates sourcesinto the spiral arms. However, since the spiral structure of the MilkyWay is very much an ongoing matter of research, we do not adoptthese distance solutions directly. Rather, we use the distances calcu-lated with these various priors to distinguish between the near andfar kinematic distances.2274 (93 per cent) of the 2446 sources are assigned to the nearkinematic distance in this way, with only 120 (5 per cent) assignedto the far distance, and the remainder are located at the tangent point( 𝑑 ≈ . http://bessel.vlbi-astrometry.org/ N o . s o u r ce s S c u t u m ( n e a r ) N o r m a ( n e a r ) S ag i tt a r i u s ( n e a r ) S c u t u m ( f a r ) S ag i tt a r i u s ( f a r ) P e r s e s O u t e r T a n g e n t p o i n t Figure 3.
Distribution of kinematic distances in 0.5 kpc-wide bins, as deter-mined for all sources (blue), and the subset of clumps (orange), in Sect. 3.4.The approximate locations of the intersection of the spiral arms with the lineof sight, as in the Reid et al. (2019) model, are shown as vertical lines. with the adopted Galactic rotation model, and a similar figure of79 per cent is found for 1.1 mm sources sources from the BolocamGalactic Plane Survey (BGPS; Ellsworth-Bowers et al. 2015). Only35 per cent of clumps within the Hi-GAL catalogue of Elia et al.(2017) were assigned distances on the near-side of the tangent point.However, in the latter case, there is a bias towards far distances, asobjects for which data (such as extinction data) that may allow anear determination to be made were unavailable are assumed to belocated at the far distance. After removing clumps with this defaultassumption from the Hi-GAL catalogue, the number rises sharply to89 per cent. We show the resulting distribution of distances in Fig.3, and compare the values to their counterparts in other cataloguesin Appendix E. We find that two-thirds of our distance estimatesmatch with those of the other catalogues to within 1 kpc and can beconsidered to be in general agreement.It is not surprising that a large fraction of sources are found tobe located at the near distance as a result of the combination ofMalmquist bias, and the fact that the ℓ = ◦ sight-line covers nearbysections of the Sagittarius, Scutum, and Norma spiral arms (Reid et al.2019). However, we caution that the friends-of-friends analysis usedhere will inevitably lead to cases where sources at the far distanceappear in clusters of sources for which our methodology favours anear distance.In a manner similar to the determination of the cluster velocities,the median distance to each cluster is determined from the set ofdistances stemming from its constituent sources with the most robustvelocity measurements, and then assigned to all sources within thecluster. Finally, we account for clusters which host maser sources MNRAS000
Distribution of kinematic distances in 0.5 kpc-wide bins, as deter-mined for all sources (blue), and the subset of clumps (orange), in Sect. 3.4.The approximate locations of the intersection of the spiral arms with the lineof sight, as in the Reid et al. (2019) model, are shown as vertical lines. with the adopted Galactic rotation model, and a similar figure of79 per cent is found for 1.1 mm sources sources from the BolocamGalactic Plane Survey (BGPS; Ellsworth-Bowers et al. 2015). Only35 per cent of clumps within the Hi-GAL catalogue of Elia et al.(2017) were assigned distances on the near-side of the tangent point.However, in the latter case, there is a bias towards far distances, asobjects for which data (such as extinction data) that may allow anear determination to be made were unavailable are assumed to belocated at the far distance. After removing clumps with this defaultassumption from the Hi-GAL catalogue, the number rises sharply to89 per cent. We show the resulting distribution of distances in Fig.3, and compare the values to their counterparts in other cataloguesin Appendix E. We find that two-thirds of our distance estimatesmatch with those of the other catalogues to within 1 kpc and can beconsidered to be in general agreement.It is not surprising that a large fraction of sources are found tobe located at the near distance as a result of the combination ofMalmquist bias, and the fact that the ℓ = ◦ sight-line covers nearbysections of the Sagittarius, Scutum, and Norma spiral arms (Reid et al.2019). However, we caution that the friends-of-friends analysis usedhere will inevitably lead to cases where sources at the far distanceappear in clusters of sources for which our methodology favours anear distance.In a manner similar to the determination of the cluster velocities,the median distance to each cluster is determined from the set ofdistances stemming from its constituent sources with the most robustvelocity measurements, and then assigned to all sources within thecluster. Finally, we account for clusters which host maser sources MNRAS000 , 1–21 (2020)
A. J. Rigby et al. with a known parallax distance. A total of 11 masers with parallaxdistances are located in the field, and after identifying the clusterassociated with the 𝑙, 𝑏, 𝑣 position of each maser source, we assignthe parallax distance to all sources within that cluster. There are atotal of nine clusters that contain maser sources and are assigneddistances in this way. Two of the eleven maser sources, G023.00-0.41 ( 𝑣 = ± − , 𝑑 = . ± .
36 kpc) and G023.20-0.37( 𝑣 = ±
10 km s − , 𝑑 = . ± .
61 kpc), are associated with objectswithin the same cluster, and since the associated distances are signifi-cantly different, a choice must be made. For this cluster, we adopt theparallax distance associated with G023.00-0.41, which is in strongeragreement with both the mean of the predetermined kinematic dis-tances ( 𝑑 = .
20 kpc) and systemic velocities (77 . ± . − ).A further two of the maser sources, G023.65-0.12 and G23.6, arelocated within the same cluster, but their distances and velocitiesare identical, and so present no conflict. The median discrepancybetween the kinematic and maser parallax distances to these sevenclusters is 0.41 kpc, though the largest difference is 1.73 kpc.At this point, we have derived two distances for each object withinthe dendrogram, 𝑑 group and 𝑑 cluster , both of which are derived prin-cipally from the sources’ line-of-sight velocities, and the assumptionthat they simply follow circular motions about the Galactic centre.However, it is well known that line-of-sight streaming motions of upto ∼
10 km s − can be present for molecular clouds, and peculiar mo-tions within molecular cloud complexes further complicate this pic-ture. To mitigate these effects in the subsequent analysis, the clusterdistances were, therefore, used to determine any distance-dependentproperties, and we note that the largest internal velocity dispersionof the clusters identified using the friends-of-friends grouping is2.3 km s − . The difference between the two sets of distances is min-imal and, when considering the population of clumps, 98 per centhave a difference between 𝑑 group and 𝑑 cluster of less than 0.2 kpc. Estimating dust temperatures ( 𝑇 d ) and column densities ( 𝑁 H ) ofmolecular hydrogen from dust continuum imaging is typically doneby fitting a modified blackbody to the spectral energy distribution(SED): 𝐼 𝜈 = 𝜇 H 𝑚 H 𝑁 H 𝜅 𝜈 𝐵 𝜈 ( 𝑇 d ) , (1)where 𝐼 𝜈 is the specific intensity, 𝜇 H is the mean molecular weightper hydrogen molecular (with a value of 2.8), 𝑚 H is the mass of ahydrogen atom, 𝜅 𝜈 is the specific dust mass absorption coefficient,and 𝐵 𝜈 is the Planck function evaluated at the frequency 𝜈 for the dusttemperature. Three or more photometry points taken from imagingat submillimetre or millimetre wavelengths, convolved to the sameresolution, are usually required (assuming both 𝜇 and 𝜅 𝜈 are fixed),providing a fit at the resolution of the longest wavelength being used– typically 36 arcsec for Herschel µ m data. However, the usageof dust temperatures derived at 36 arcsec-resolution for the 13 arcsec-resolution scales of the 1.15 mm data would mean that local minimaand maxima would be beam-diluted. The different spatial frequenciesprobed by the ground- and space-based observatories also presents afurther limitation.To overcome these limitations, we adopt colour temperatures de-rived from the ratio of Hi-GAL 160 to 250 µ m flux densities (e.g.Peretto et al. 2016; Rigby et al. 2018), which have an effective reso- lution of 18 arcsec. The colour temperature is related to the ratio offlux densities in the following way: 𝑆 𝑆 = 𝐵 ( 𝑇 col ) 𝐵 ( 𝑇 col ) (cid:18) μ m160 μ m (cid:19) 𝛽 . (2)where 𝑆 𝜈 is the source-integrated flux density 𝑆 𝜈 = ∫ 𝐼 𝜈 𝑑 Ω . Byadopting a fixed value of 𝛽 = .
8, and using 160 and 250 µ mphotometry from Hi-GAL imaging, we sampled 𝑇 col from a grid ofvalues ranging from 2.7 to 50 K, with 0.1 K intervals. Uncertaintieson the Herschel photometry result in uncertainties on 𝑇 col of ∼ µ m image to that of the 250 µ m image following Aniano et al.(2011), and both images were high-pass filtered to remove spatialfrequencies larger than 6.5 arcmin using the nebuliser applicationof the Cambridge Astronomy Survey Unit Tools software package to approximate the spatial filtering present in the NIKA2 data.Many of the sources in the GP field consist of a mixture of compactand extended structure, and the flux densities of compact sources,therefore, usually contain contributions both of the compact objectitself, and the extended emission structure on which they reside.We have therefore adopted the ‘clipping’ scheme (Rosolowsky et al.2008) of flux estimation for 𝑆 𝜈 (and thereby the mass) for each source,in which the local background level of flux is subtracted from thetotal, or put another way, only flux above the contour level definingthe structure is counted in the integrated flux. Although Rosolowskyet al. (2008) prefer the ‘bijection’ scheme, in which the entirety ofthe flux density integrated across the source’s sky position is used forthe mass calculation within molecular clouds, the clipping schemeis used here, since we are most interested in tracing the mass of thesmallest structures within the hierarchy.We determine the appropriate background level for each source byfirst performing a binary dilation on each of the source masks, whichexpands the perimeter by one pixel, and allows the encompassingperimeter pixels to be isolated (i.e. pixels just below the 1.15 mmcontour level at which the source was identified). The backgroundflux density at 1.15 mm is the contour level at which the source wasseparated from its parent structure, and was determined as the meanof the minimum pixel value within the source, and the maximum pixelvalue within the perimeter pixels. A slightly different approach mustbe taken to determine the background flux densities for the 160 and250 µ m, because the morphology of emission at those wavelengthsmay differ from the morphology at 1.15 mm at which the source wasdefined. In contrast to the 1.15 mm case, some of the pixels within theboundary region may be brighter than some of those within the sourcearea. At these wavelengths, the background value is therefore takenas half of the sum of the mean value of pixels within the perimeter,and the minimum pixel value within the source, thus approximatingthe 1.15 mm contour.The masses of the sources identified in the dendrogram are calcu-lated from 𝑆 𝜈 , the monochromatic flux densities integrated over thesolid angle subtended by the source at 1.15 mm using 𝑀 = 𝑆 𝜈 𝑑 𝜅 𝜈 𝐵 𝜈 ( 𝑇 col ) , (3)where the dust absorption coefficient is given by 𝜅 𝜈 = . ( 𝜈 / ) 𝛽 (Beckwith et al. 1990). We use a value evaluated at 1.15 mm (260 http://casu.ast.cam.ac.uk/surveys-projects/software-release/background-filtering MNRAS , 1–21 (2020)
ASTON: Galactic Star Formation with NIKA2 GHz) of 𝜅 = .
009 cm g − , assuming 𝛽 = .
8, and incorporatinga gas-to-dust mass ratio of 100. This value is close to the value of0.008 cm g − used in our previous NIKA study (Rigby et al. 2018),and that adopted by Hi-GAL in Elia et al. (2017). We discuss theimpact of different values of 𝛽 upon our results in Sect. 4.4. Forall sources, we use their cluster distance, 𝑑 cluster , as determined inSect. 3.4, and we decrease the integrated flux densities by 4 percent to account for the median level of CO (2–1) contamination(determined in Appendix D), adding a 4 per cent contribution tothe corresponding uncertainties. Uncertainties are calculated usingMonte Carlo methods for the primary quantities for each source (i.e.,distance, absolute flux calibration uncertainty), and propagated to allderived quantities, such as mass and colour temperature. We notethat the contribution of the factor of ∼ 𝜅 𝜈 is notincluded in our mass uncertainties since its effects are likely to bemostly systematic across the region, and will consequently not havean impact on the analyses presented here.We note that, as a result of the 18 arcsec resolution of the colourtemperature estimates, smaller-scale temperature variations will notbe detected. This may result in the over-estimation of source masseswhere compact sources warmer than the beam-smeared average arepresent, and vise-versa for colder compact sources. At large radii, thetemperature of protostellar sources is of the form 𝑇 ( 𝑟 ) ∝ 𝑟 /( + 𝛽 ) ,where 𝛽 is the dust emissivity spectral index (Terebey et al. 1993)and, by adopting 𝛽 = .
8, we can expect to underestimate dusttemperatures of point sources at 18 arcsec by up to ∼
11 per centcompared to those at 13 arcsec. For resolved sources, this effect isunlikely to play a significant role, but we conservatively increase theuncertainties on 𝑇 col for all sources by 11 per cent to account for thiseffect.We discuss the resulting distributions of mass and temperature inSect. 4.2 and Sect. 4.3, and present the overall distribution in Fig. 6. This particular region of the Galactic plane is well covered in manyrecent surveys of (sub-)millimetre continuum emission. The mostsimilar ground-based surveys to our NIKA2 observations are the1.1 mm BGPS at 33 arcsec resolution (Aguirre et al. 2011; Ginsburget al. 2013) and, at 870 µ m and with 19.2 arcsec resolution, AT-LASGAL (Schuller et al. 2009). While the data from these surveysare very well matched to GASTON in terms of wavelength, theywere designed to cover far wider regions of the GP (covering 192and 360 deg , respectively), and so their sensitivities are lower thanthose of the data presented here.We have identified a total of 1467 GASTON clumps (dendrogramleaves), compared to 346 ATLASGAL compact sources (Urquhartet al. 2018) and 164 from the BGPS (Ellsworth-Bowers et al. 2015)within the same sky area. Of the 164 sources from the BGPS, only 2sources do not have centroid coordinates lying within three convolvedbeam radii of a GASTON clump centroid. In both cases the BGPSsource is located within a large-scale diffuse region in which theGASTON image has partially resolved and detected clumps aroundthe periphery, and have centroid coordinates located far away fromthe particular region of emission extracted from the BGPS data. All Our search parameters adopt convolved beam radii, that are the quadraturesum of the (smoothed) 13 arcsec GASTON beam and the beam size of thedata set in question. of the ATLASGAL clumps lie within three convolved beam radii ofa GASTON compact source, and they all fall within the boundariesof emission extracted by the dendrogram analysis in this work. Thecentroid coordinates of 1237 (84 per cent) and 1114 (75 per cent)out of the 1467 GASTON clumps are located further than threeconvolved beam radii of any BGPS or ATLASGAL source centroid,respectively. Although this positional matching based on centroidcoordinates will be less effective for sources that are not centrallyconcentrated, ∼
95 per cent of GASTON clumps have angular radiiof less than three beam radii (i.e. 39 arcsec), and so it is clear thatthe majority of GASTON clumps are new identifications.In Fig. 4, we compare the GASTON 1.15 mm imaging in a regioncentred on ℓ = . ◦ , 𝑏 = − . ◦
205 to observations from ATLAS-GAL and the BGPS, and 8 µ m emission from the GLIMPSE. Thegreater angular resolution of the GASTON GP data compared toBGPS and ATLASGAL clearly allows new substructures to be re-solved within previously-detected emission regions, in addition tothe detection of new faint compact sources. By comparison to the Spitzer /GLIMPSE 8 µ m image, it can also be seen that the 1.15 mmcontinuum is effective at tracing structures seen in absorption asinfrared-dark features as well as compact regions of bright emissionat 8 µ m. Not all features from the 8 µ m image have discernible1.15 mm counterparts, and although most of the faint and extended8 µ m emission (presumably arising from hot dust grains or PAHemission at ionization fronts) is not seen in the GASTON image,there are visible counterparts to some of these features, such as thecolumn extending from [24 . ◦ , − . ◦
10] to [24 . ◦ , − . ◦
20] and, sur-prisingly, the bow shock at [24 . ◦ , − . ◦ ℓ = . ◦ , 𝑏 = . ◦
0, witha radius of 0 . ◦
55 – partially overlapping with the GASTON GP field– inside which they reached a median sensitivity of 9 mJy beam − with a resolution of 8.5 arcsec. After rescaling to 1.15 mm, assuminga spectral energy distribution of the form 𝐼 𝜈 ∝ 𝜈 . , to match ourNIKA2 observations, the point-source sensitivity would correspondto ∼ . − , making them around half as sensitive as thosepresented here. There are a total of 993 AzTEC compact sourceswithin the region that overlaps with GASTON, after restricting thesource catalogue to those considered by Heyer et al. (2018) to berobust, i.e. probability of a false-positive detection of < >
50 arcsec) compared to that of NIKA2 ( ∼ . (cid:48) MNRAS000
50 arcsec) compared to that of NIKA2 ( ∼ . (cid:48) MNRAS000 , 1–21 (2020) A. J. Rigby et al. − . ◦ − . ◦ − . ◦ G a l a c t i c L a t i t ud e GLIMPSE µ m GASTON 1.15 mm . ◦ . ◦ . ◦ . ◦ − . ◦ − . ◦ − . ◦ Galactic Longitude G a l a c t i c L a t i t ud e ATLASGAL 870 µ m . ◦ . ◦ . ◦ . ◦ Galactic Longitude
BGPS 1.1 mm
Figure 4.
Comparison of the GASTON 1.15 mm imaging (upper-right panel) to imaging from
Spitzer /GLIMPSE 8 µ m (upper-left; Benjamin et al. 2003;Churchwell et al. 2009), APEX/ATLASGAL 870 µ m (lower-left; Schuller et al. 2009) and CSO/BGPS (lower-right; Ginsburg et al. 2013) imaging at 1.1 mm.FWHM beam sizes are shown in the lower-right corners of each image. All images are shown at their native resolution and pixel sizes. . ◦ . ◦ . ◦ Galactic Longitude − m J y b e a m − . ◦ . ◦ . ◦ . ◦ . ◦ . ◦ Galactic Longitude G a l a c t i c L a t i t ud e × M J y s r − Figure 5.
Herschel /Hi-GAL (Molinari et al. 2016a) 250 μ m (left panel) and NIKA2/GASTON 1.15 mm (right panel) continuum emission maps for a 0 . ◦ × . ◦ region centred on ℓ = . ◦ , 𝑏 = . ◦ . The positions and sizes of the compact 250 µ m sources from the Molinari et al. (2016a) Hi-GAL catalogue are shownas orange ellipses, and sources identified in the band-merged (BM) Hi-GAL catalogue of Elia et al. (2017) are shown as red circles. The positions of the 1.15 mmGASTON clumps are shown by markers with black borders: those that have no cospatial 250 μ m counterpart are given in white, while 1.15 mm sources withcounterparts are shown as pink ellipses. Both panels contain an inlaid zoomed region, with Herschel markers only shown on the left panel, while in the rightpane1 – for the sake of clarity – we show the boundaries of the GASTON clumps, adopting the same colour scheme as for the ellipses in the main image.MNRAS , 1–21 (2020) ASTON: Galactic Star Formation with NIKA2 Table 2.
Summary of the results of catalogue cross-matching procedure used to compare the GASTON source statistics with other surveys in Sect. 4.1 and Sect.4.2. For each data set (i), we identify: (ii) the catalogue used; (iii) the wavelength; (iv) the angular resolution; (v) 𝑁 , the total number of sources in the cataloguefalling within the cropped GASTON GP field; (vi) 𝑁 source , the number of source centroids that fall upon a pixel belonging to an identified GASTON source;(vii) 𝑁 clump , the number of sources coincident with a GASTON clump centroid; (viii) 𝑁 source , the number of GASTON clump centroids coincident with acatalogued source centroid; (ix) 𝑁 unique , the number of GASTON clump centroids that are not coincident with a catalogued source centroid. Data Reference Wavelength Resolution
𝑁 𝑁 source 𝑁 clump 𝑁 matched 𝑁 unique (i) (ii) (iii) (iv) (v) (vi) (vii) (viii) (ix)BGPS Ellsworth-Bowers et al. (2015) 1.1 mm 33 (cid:48)(cid:48)
164 162 162 230 1237ATLASGAL Urquhart et al. (2018) 870 µ m 19.2 (cid:48)(cid:48)
346 346 345 365 1102AzTEC Heyer et al. (2018) 1.1 mm 8.5 (cid:48)(cid:48)
993 698 420 336 357Hi-GAL Elia et al. (2017) BM (cid:48)(cid:48)
857 731 734 702 765Hi-GAL Molinari et al. (2016a) 70 µ m 6 (cid:48)(cid:48) µ m 12 (cid:48)(cid:48) µ m 18 (cid:48)(cid:48) µ m 24 (cid:48)(cid:48) µ m 35 (cid:48)(cid:48)
938 771 798 838 629 Sources are regarded as coincident when the separation of their centroids is less than or equal to three convolved beam radii (Sect. 4.1). GASTON sources are regarded as coincident with a catalogued source as above with the exception of the Hi-GAL sources for which the circular or ellipticalfootprints are checked for overlapping GASTON sources (Sect. 4.2). Sources in this catalogue require at least three consecutive detections in the 160 µ m, 250 µ m, 350 µ m, and 500 µ m wavebands. We adopt the most conservative beam size for the band-merged Hi-GAL catalogue.
NIKA2 is expected to be more sensitive to cold and compact dustsources within the Milky Way than
Herschel , owing to its greaterangular resolution at longer wavelengths. To determine whether wehave identified compact structures within the GASTON GP data thatwere not identified using the various extraction techniques applied tothe
Herschel data, we compared our catalogue to two sets of Hi-GALcompact source catalogues. We performed a series of cross-matchingexercises to identify any 1.15 mm GASTON clumps that have notbeen identified by either the five monochromatic source catalogues ofMolinari et al. (2016a), or the ‘high-reliability’ band-merged (BM)catalogue of Elia et al. (2017).To carry out the source matching, we adopted the elliptical sourcesizes from the monochromatic catalogues, and the circular sourcesizes for the BM catalogue – after convolving with a Gaussian kernelrepresentative of the 1.15 mm NIKA2 beam – to look for overlapswith the footprint of each dendrogram structure, and record the num-ber of matches in each case (see Tab. 2). By comparison to themonochromatic catalogues, we found the lowest fraction of GAS-TON clumps that do not match to any Hi-GAL compact source inthe 250 µ m band, with 321 GASTON clumps (22 per cent) with-out a 250 µ m counterpart, and 765 (40 per cent) with no band-merged counterpart. We compare the Herschel µ m data andthe GASTON GP 1.15 mm data for a 0 . ◦ × . ◦ ℓ = . ◦ , 𝑏 = . ◦
15 in Fig. 5, overlaying markers representing cata-logued sources from the 1.15 mm dendrogram extraction and the twoHi-GAL catalogues. The highest match fraction with sources in the250 µ m was expected, as these particular Herschel observations rep-resent a compromise between the greater angular resolution availableat the shorter wavelengths, and the greater sensitivity to the coldeststructures at the longer wavelengths. Although the 500 µ m cataloguerepresents the closest match in wavelength, the large difference inangular resolution results in a substantially different segmentation ofthe emission under the different source extraction methods. In Fig. 6, we compare the distributions of mass and colour temper-ature, 𝑇 col , for the full sample of sources. Their statistical propertiesare summarised in Tab. 3. The new clumps identified by our NIKA2observations with no counterpart in the Hi-GAL catalogues fall, onaverage, towards lower mass and temperature than those associatedwith Herschel -catalogued sources. We compare the log ( 𝑀 / 𝑀 (cid:12) ) and 𝑇 col distributions for the clumps that do have 250 µ m Hi-GALmatches and those that do not, using two-sample Anderson-Darling.The results show that the distributions of both properties in the twopopulations are inconsistent at a high confidence level, with 𝑝 -valuesof < .
02 in both cases.There are 14 clumps with masses in excess of 100 M (cid:12) that arenew detections. We note that to compare these values to othersurvey data, such as ATLASGAL, one must consider the massesmeasured in the same way, and the ATLASGAL clump masses ofUrquhart et al. (2018) adopt a method more similar to the ‘bijec-tion’ scheme (as discussed in Sect. 3.5). The approximate conver-sion between the ‘clipped’ masses reported in this study, and theequivalent bijected masses, calculated from combined
Herschel andNIKA2 column density maps (following the method of Rigby et al.2018) is 𝑀 bijection ≈ × 𝑀 . , with a simple linear least-squaresfit in log-space. In this metric, 33 of the 321 new clumps have 𝑀 bijection > 𝑀 (cid:12) , and may be considered as high-mass. 256of the 321 new clumps (80 per cent) have no compact 70 µ m coun-terpart found by matching to the Molinari et al. (2016a) catalogue,and so may be considered as candidates for starless sources, thoughwe caution that this is an upper limit, and a lack of a 70 µ m matchdoes not guarantee a lack of embedded star formation (e.g. Traficanteet al. 2017).We point out here that the various catalogues used in the cross-matching exercises in this Section and in the previous one, weregenerated using different source extraction techniques. We cautionthat we expect the total number of sources in the different data sets tovary with the methodology, but it is clear that we detect a substantial MNRAS000
Herschel andNIKA2 column density maps (following the method of Rigby et al.2018) is 𝑀 bijection ≈ × 𝑀 . , with a simple linear least-squaresfit in log-space. In this metric, 33 of the 321 new clumps have 𝑀 bijection > 𝑀 (cid:12) , and may be considered as high-mass. 256of the 321 new clumps (80 per cent) have no compact 70 µ m coun-terpart found by matching to the Molinari et al. (2016a) catalogue,and so may be considered as candidates for starless sources, thoughwe caution that this is an upper limit, and a lack of a 70 µ m matchdoes not guarantee a lack of embedded star formation (e.g. Traficanteet al. 2017).We point out here that the various catalogues used in the cross-matching exercises in this Section and in the previous one, weregenerated using different source extraction techniques. We cautionthat we expect the total number of sources in the different data sets tovary with the methodology, but it is clear that we detect a substantial MNRAS000 , 1–21 (2020) A. J. Rigby et al. T col / K012345 l og ( M / M (cid:12) ) F r e q u e n c y µ m Figure 6.
Mass and colour temperature distributions for all sources (grey), clumps (purple), clumps with no band-merged (BM) Hi-GAL counterpart (green),and clumps with no Hi-GAL 250 µ m counterpart (yellow). Table 3.
Statistical results – mean, standard deviation, samplesize, standard error on the mean – of the masses and colourtemperatures for the various samples presented in Fig. 6.Sample Mean 𝜎 𝑁 𝜎 /√ 𝑁 log ( 𝑀 / 𝑀 (cid:12) ) All sources 2.24 1.08 2443 0.02Clumps 1.60 0.63 1464 0.02Clumps no BM 1.35 0.49 763 0.02Clumps no 250 µ m 1.26 0.46 319 0.03 𝑇 col / KAll sources 16.66 3.59 2443 0.07Clumps 15.84 3.56 1464 0.09Clumps no BM 15.45 3.53 763 0.13Clumps no 250 µ m 15.35 3.42 319 0.19 Note that 3 clumps have 160/250 µ m flux ratios that correspondto colour temperatures that outside of the range 2 . (cid:62) 𝑇 col (cid:54)
50 K, and so have no colour temperatures or masses derived here. number of previously undetected sources, with the 321 sources thatdo not coincide with 250 µ m compact sources representing the mostconservative estimate. To characterise our sample, we have computed the infrared-brightfraction for each source, 𝑓 IRB , based on GLIMPSE 8 µ m imaging(Benjamin et al. 2003; Churchwell et al. 2009). The infrared-brightfraction is defined as the fraction of pixels within the source boundarythat are brighter than the median value of all pixels within a 4.8-arcmin box centred on each pixel (see Watkins et al. in prep for a fulldescription). The filter size was selected so that the most extendedsources within the 8 µ m image – typically bubbles associated withH ii regions – could be compared to their local background. Althoughthis filter scale represents a different spatial scale for the nearest andfarthest sources within the region, it offers robust measurements forsources with angular sizes less than half of the filter scale, whichis true for all GASTON clumps. We note that a very similar theclassification scheme was used by Battersby et al. (2011), who usedGLIMPSE 8 µ m images to classify a sample of Hi-GAL sources.In Sect. 4.2, each clump was also matched to the 70 µ m catalogueof Molinari et al. (2016a) and we record the sum of the positionally-matched 70 µ m source integrated flux densities. The 70 µ m luminos-ity is known to be a good tracer of the bolometric luminosity of anyembedded sources (e.g. Dunham et al. 2008; Ragan et al. 2012). Wecalculate the bolometric luminosity by using the Elia et al. (2017) MNRAS , 1–21 (2020)
ASTON: Galactic Star Formation with NIKA2 . . . . . . f IRB T c o l / K T col / K = (5 . ± . f IRB + (13 . ± .
1) 01 f L bol /M [ L (cid:12) /M (cid:12) ] Figure 7.
Colour temperature as a function of the infrared-bright fraction, 𝑓 IRB . The coloured points show the relation for all GASTON sources whichhave at least one matched Hi-GAL 70 µ m compact source from Molinari et al.(2016a), and are colour-coded by the ratio of their bolometric luminosityto 1.15 mm mass. The empty circles show the relationship for unmatchedGASTON sources. The trend determined by a linear least-squares fit is shownas the solid orange line. The red points show, on the secondary 𝑦 -axis, thefraction of GASTON sources that are associated with compact 70 µ m sourcesas a function of 𝑓 IRB in 0.1-wide bins. relationships: 𝐿 bol 𝐿 (cid:12) = . (cid:16) 𝑆 Jy (cid:17) . (cid:16) 𝑑 kpc (cid:17) if 𝑆 (cid:62)
50 Jy , . (cid:16) 𝑆 Jy (cid:17) . (cid:16) 𝑑 kpc (cid:17) if 𝑆 <
50 Jy , (4)where 𝑆 is the sum of the integrated flux densities of any compactsources from the catalogue of Molinari et al. (2016a) that lie withinone beam-radius of a GASTON clump boundary.The infrared-bright fraction, 𝑓 IRB , may serve as an indicator for theevolutionary stage of a source: sources that are completely infrared-dark ( 𝑓 IRB =
0) are likely to be at the earliest stages of evolution,containing no observable indications of star-formation in terms oftheir 8 µ m flux and, conversely, sources that are completely infrared-bright ( 𝑓 IRB =
1) are likely to be sources exhibiting signs of advancedstar-formation, most commonly in the form of emission from youngstellar objects, H ii regions, or heated polycylcic aromatic hydrocar-bon (PAH) molecules. Although a more detailed analysis will bepresented in Watkins et al. ( in prep. ), we demonstrate the basic util-ity of 𝑓 IRB as an evolutionary indicator in several ways in Fig. 7.Firstly, there is a positive correlation (Pearson correlation coefficient 𝜌 = . 𝑝 − value (cid:28) .
1) between 𝑓 IRB and the source tempera-ture as traced by 𝑇 col . Secondly, there is also a positive correlation( 𝜌 = . 𝑝 − value (cid:28) .
1) between 𝑓 IRB and log ( 𝐿 bol / 𝑀 ) . Tem-perature and 𝐿 bol / 𝑀 are both generally regarded as robust tracersof the time-evolution of clumps (e.g. Urquhart et al. 2014, 2018;Molinari et al. 2016b; Svoboda et al. 2016; Elia et al. 2017). Finally,we show that the fraction of GASTON sources that are associatedwith compact 70 µ m sources from Hi-GAL (Molinari et al. 2016a)also increases with 𝑓 IRB , and reaches a plateau of ∼
85 per cent forsources with 𝑓 IRB (cid:38) . 𝑓 IRB parameter, since darkness at 8 µ m usually requiresa bright background against which the radiation can be absorbed.We therefore reduce our sample in the following analysis to include only those clumps that lie at distances between 3.5 and 7.0 kpc.This criterion serves a dual function; firstly, it limits the effect of thedistance bias in 𝑓 IRB since infrared-dark clouds are generally locatedat the near kinematic distance (e.g. Ellsworth-Bowers et al. 2013),and the tangent for kinematic distances at this longitude range is ∼ . ∼
10 per cent of the sample.We define four categories through which we divide the sample ofclumps based upon their infrared-bright fractions that can be used tobroadly trace evolutionary phase: i) 𝑓 IRB < .
040 for predominantlyinfrared-dark clumps; ii) 0 . (cid:62) 𝑓 IRB < .
229 for clumps that aremostly infrared-dark; iii) 0 . (cid:62) 𝑓 IRB < .
663 for clumps that arepartially infrared-bright; iv) 𝑓 IRB (cid:62) .
663 for clumps that are pre-dominantly infrared-bright. The specific values of 𝑓 IRB were chosensuch that the four sub-samples are of equal size, each containing atotal of 321 clumps. The fraction of clumps in each category thatare associated with compact 70 µ m counterparts, 𝑓 also increasesmonotonically with the proposed evolutionary sequence, with 21,33, 59 and 69 per cent of the clumps having 70 µ m counterparts insamples i)–iv), respectively.Tracing the precise time-evolution of star-forming regions is nota trivial matter, but although we caution that 𝑓 IRB does not give an absolute value of ‘evolutionary stage’ or age for individual sources,it may be regarded as relative age indicator and, therefore, a suitablemeasurement when dealing with large samples of sources. Using asimilar methodology to that used in this paper, Battersby et al. (2017)estimated the fraction of a sample of dense molecular regions abovea column density threshold associated with high-mass star formationthat were starless or star-forming. By comparing the relative fractionsof these categories to those associated with methanol masers andultra-compact H ii regions – which have established lifetimes – theauthors estimated the lifetime of the starless and star-forming phasesas 0.2–1.7 and 0.1–0.7 Myr, respectively. These two timescales maybroadly map onto stages i)–ii) (predominantly starless) and iii)–iv)(predominantly star-forming), respectively, indicating a total durationof (cid:46) . 𝑓 IRB = 𝑓 IRB = µ m. These mean valuessuggest that the clumps, on average, gain a moderate amount of mass– increasing by ∼ . 𝜎 level between stages i) and ii), at the 4.3 𝜎 level betweenstages i) and iii), and at the 4.0 𝜎 level of significance between stagesiii) and iv). The increase in source counts as the GASTON GP projectprogresses will allow these values to be revisited with greater sta-tistical power. The mean values and their associated standard errorsare given in Tab. 4. A further striking feature of Fig. 8 is the increas-ing spread in temperatures along the evolutionary sequence. Thisbehaviour is expected as a consequence of the more rapid evolutionof the highest-mass sources, in addition to the mass function. Themore numerous low-mass sources are expected to evolve relativelyslowly, only increasing in temperature, and losing mass at later stagesas the clump is dispersed. The rarer high-mass clumps are expectedto rapidly gain in mass and temperature at early sages, before losingmass at later stages. MNRAS000
663 for clumps that are pre-dominantly infrared-bright. The specific values of 𝑓 IRB were chosensuch that the four sub-samples are of equal size, each containing atotal of 321 clumps. The fraction of clumps in each category thatare associated with compact 70 µ m counterparts, 𝑓 also increasesmonotonically with the proposed evolutionary sequence, with 21,33, 59 and 69 per cent of the clumps having 70 µ m counterparts insamples i)–iv), respectively.Tracing the precise time-evolution of star-forming regions is nota trivial matter, but although we caution that 𝑓 IRB does not give an absolute value of ‘evolutionary stage’ or age for individual sources,it may be regarded as relative age indicator and, therefore, a suitablemeasurement when dealing with large samples of sources. Using asimilar methodology to that used in this paper, Battersby et al. (2017)estimated the fraction of a sample of dense molecular regions abovea column density threshold associated with high-mass star formationthat were starless or star-forming. By comparing the relative fractionsof these categories to those associated with methanol masers andultra-compact H ii regions – which have established lifetimes – theauthors estimated the lifetime of the starless and star-forming phasesas 0.2–1.7 and 0.1–0.7 Myr, respectively. These two timescales maybroadly map onto stages i)–ii) (predominantly starless) and iii)–iv)(predominantly star-forming), respectively, indicating a total durationof (cid:46) . 𝑓 IRB = 𝑓 IRB = µ m. These mean valuessuggest that the clumps, on average, gain a moderate amount of mass– increasing by ∼ . 𝜎 level between stages i) and ii), at the 4.3 𝜎 level betweenstages i) and iii), and at the 4.0 𝜎 level of significance between stagesiii) and iv). The increase in source counts as the GASTON GP projectprogresses will allow these values to be revisited with greater sta-tistical power. The mean values and their associated standard errorsare given in Tab. 4. A further striking feature of Fig. 8 is the increas-ing spread in temperatures along the evolutionary sequence. Thisbehaviour is expected as a consequence of the more rapid evolutionof the highest-mass sources, in addition to the mass function. Themore numerous low-mass sources are expected to evolve relativelyslowly, only increasing in temperature, and losing mass at later stagesas the clump is dispersed. The rarer high-mass clumps are expectedto rapidly gain in mass and temperature at early sages, before losingmass at later stages. MNRAS000 , 1–21 (2020) A. J. Rigby et al. l og ( M / M (cid:12) ) i) f IRB < . n = 321 f = 21% . k p c . k p c ii) 0 . ≥ f IRB < . n = 321 f = 33%0 10 20 30 T col / K01234 l og ( M / M (cid:12) ) iii) 0 . ≥ f IRB < . n = 321 f = 59% 0 10 20 30 40 T col / K iv) f IRB ≥ . n = 321 f = 69% 1234 l og ( L b o l / L (cid:12) ) Figure 8.
Mass–colour temperature relationships for the four subsamples of GASTON clumps, classified by their infrared-bright fraction, 𝑓 IRB . Masses arecalculated using the ‘clipping’ method. Filled orange squares show the mean value of the quantities plotted in each axis, while the open squares show the meanvalues for the previous stages in the proposed evolutionary sequence. In each case, the standard error on the mean in both axes is shown in black within themarker shapes. The colour of the points shows the total bolometric luminosity from matched 70 µ m compact sources, while the grey points have no compact70 µ m counterparts from Molinari et al. (2016a). Red ellipses show the approximate 1- 𝜎 and 2- 𝜎 confidence intervals for the distribution, containing ∼
68 and95 per cent of the data points, respectively. The dashed and dotted purple curves correspond to the 1.15 mm point-like and extended source sensitivities for thecombination of the data and source extraction parameters at the near- and far-distance limits of the sample.
Table 4.
Mean values and associated standard errors of the distributions ofthe four evolutionary stages presented in Fig. 8.Stage log ( 𝑀 / 𝑀 (cid:12) ) 𝑇 col / KMean Error Mean Errori) 1.52 0.03 13.61 0.15ii) 1.66 0.03 14.60 0.14iii) 1.71 0.03 16.49 0.16iv) 1.52 0.04 18.20 0.21 This reflects the behaviour that might be expected from a clump-fed model of star formation, whereby the mass gain of the clump isthe result of accretion from its direct environment, and the clump is subsequently dispersed by feedback as star formation progresses. Bycontrast, in a core-fed scenario, such clumps do not exhibit an initialincrease in mass, since all of the mass is in situ before star-formationbegins, which seems incompatible with Fig. 8. Furthermore, themovement of the mean points in Fig. 8 bears a striking resemblanceto the mass-temperature tracks described by the clump-fed modelsof Peretto et al. (2020, see their Fig. 7), supporting the coeval massgrowth of clumps and embedded protostars within. However, it isimportant to note that the sizes of the GASTON sources presentedhere are, on average, larger than those presented in Peretto et al.(2020). This might suggest that the same accretion process fromlarge to small scales occur over a relatively wide range of scales.A qualitatively similar behaviour as that seen in Fig. 8 is observedby Elia et al. (2017, see their Fig. 22). In that study, clumps werecategorised as starless or star-forming based upon the presence of70 µ m emission, with the star-forming sample further divided ac- MNRAS , 1–21 (2020)
ASTON: Galactic Star Formation with NIKA2 cording to the detection or non-detection of 21–24 µ m counterparts,and coincidence with H ii regions.We highlight that each of the four subsamples is likely to containa mix of both core-fed and clump-fed sources, and so the signalpresent here in the progressive movement of the mean position willbe moderated by the core-fed sources toward the low-mass end ofthe mass distribution. If the clump-fed scenario is more likely to beimportant for the highest-mass clumps, then this behaviour couldbe reflected in the upper envelope of the distributions. The 95thpercentile in mass changes by + . − .
03, and + .
06 dex, whilethe 95th percentile in temperature increases by 0.7 K, 2.7 K and4.4 K between stages i)–ii), ii)–iii), and iii)–iv), respectively. Thelack of a reduction in the 95th percentile in logarithmic mass inthe final stages is more difficult to interpret, as it suggests that theclump dispersal phase of the most massive clumps is either notbeing sampled well here, or lasts a relatively long time. The formerexplanation could easily be true, as the number of sources in the 95thpercentile is only 16 in each stage, and thus the measurement suffersfrom small numbers. Urquhart et al. (2014) argued that, althoughthe accretion phase for high-mass protostars is probably very rapid,they may then take a relatively long time to begin dispersing theenveloping clump, resulting in the clustering of sources at the apexof their mass-luminosity evolution. Qualitatively, this would resultin the same behaviour seen here with these highest-mass clumpsspending much of their time in the high-mass, high-temperature partof this diagram. This is also supported by the clump-fed models ofPeretto et al. (2020), whose evolutionary tracks for the highest-massstars shows an accretion phase that is relatively short-lived comparedto the lifetime approaching their maximum mass.It is unlikely that this sequence arises as a result of observationaleffects such as sensitivity or completeness. The point-source sensi-tivity limits are displayed in each panel of Fig. 8, and do not displayany signs of correlation with the movement of the mean position ofthe distributions in stages i)–iii); lower-mass clumps are more likelyto be observable at later stages when temperatures are higher, andso this effect may in fact be suppressing mean value of the mass inthe middle stages, and the average mass gain in high-mass sources istherefore likely to be greater than the value of 60 per cent deducedfrom this study.The source extraction, which was carried out on the 1.15 mm map,is completely independent of the 𝑓 IRB parameter, and so we do notexpect any systematic bias in terms of completeness as a functionof evolutionary stage. The effects of mass completeness manifestthemselves in this parameter space by the position of the peak ofthe distribution. The real distribution of clump masses is expected toconsist of a power law 𝑁 ∝ 𝑀 𝛼 , where 𝛼 <
0, and so the turnoverclose to the mean value reflects the completeness that is a function ofthe sensitivity and source extraction methodology. Since both of thesefactors are consistent across all samples, it is difficult to conceive ascenario in which differences in completeness could be dominatingthe position of the mean.
In the previous Section we have provided evidence, acquired from thecatalogue of clumps identified at 1.15 mm, that the overall populationis most massive at intermediate evolutionary stages, as traced by the 𝑓 IRB proxy for relative evolution. In this Section, we explore variouscaveats to this analysis, and test the robustness of the observed trend.Firstly, our methods for acquiring clump distances in Sect. 3.4have resulted in a large fraction – 93 per cent – of sources that arelocated at the near kinematic distance. We expect a high fraction . . . . . . T col / K1 . . . T col / Kf) l og ( M / M (cid:12) ) Figure 9.
Mean evolutionary trend after altering the samples to: a) displacea randomly-selected third of clumps to the far kinematic distance (usingthree sets); b) adopt 𝑑 group as the distance estimate; c) adjust the distancelimits to 4 < 𝑑 cluster < 𝑓 IRB limits for the subsamples;f) use alternative values of 𝛽 . In each case, the original trend is shown asgrey squares, with mean positions of the altered samples overlaid in colouredcircles. The shaded regions show the ± 𝜎 confidence limits. along this line of sight, and it is, perhaps, not surprising that ahigher fraction is recovered than for clumps in ATLASGAL (77 percent), BGPS (79 per cent), or Hi-GAL (89 per cent of the ‘good-quality’ assignments), due simply to the greater mass sensitivityof GASTON, and the greater prevalence of low-mass compared tohigh-mass clumps. However, we have tested the robustness of theproposed evolutionary trend by repeating the analysis three timesafter first randomly assigning 1/3 of the population of clumps to theirfar kinematic distance solutions. The resulting trends are illustratedin Fig. 9, and the overall effect is, in two of the three cases, toslightly reduce the significance of the differences between stages i)–ii), i)–iii), and iii)-iv) compared to the original distance assignments,while the significance is increased in the third case. This tendencyto slightly reduce the significance can be mainly ascribed to theresulting smaller sample sizes, as the clumps now assigned to thefar-kinematic distance fall outside the distance limits of the sample,consequently increasing the error on the mean.Another choice affecting the results was the decision to adopt thecluster-averaged distances, 𝑑 cluster , as opposed to the dendrogramgroup-averaged distances, 𝑑 group . If we adopt the latter in place ofthe former, and repeat the analysis, we see almost no change in theresults, since the average difference in the two distances in on theorder of one hundred parsecs. In fact, as can be seen in panel b) ofFig. 9, the evolutionary trend is slightly strengthened in this scenariowith the largest change coming between stages i) and iii), now atthe 4.5- 𝜎 level of significance, compared to 4.3- 𝜎 in the originalanalysis. MNRAS000
Mean evolutionary trend after altering the samples to: a) displacea randomly-selected third of clumps to the far kinematic distance (usingthree sets); b) adopt 𝑑 group as the distance estimate; c) adjust the distancelimits to 4 < 𝑑 cluster < 𝑓 IRB limits for the subsamples;f) use alternative values of 𝛽 . In each case, the original trend is shown asgrey squares, with mean positions of the altered samples overlaid in colouredcircles. The shaded regions show the ± 𝜎 confidence limits. along this line of sight, and it is, perhaps, not surprising that ahigher fraction is recovered than for clumps in ATLASGAL (77 percent), BGPS (79 per cent), or Hi-GAL (89 per cent of the ‘good-quality’ assignments), due simply to the greater mass sensitivityof GASTON, and the greater prevalence of low-mass compared tohigh-mass clumps. However, we have tested the robustness of theproposed evolutionary trend by repeating the analysis three timesafter first randomly assigning 1/3 of the population of clumps to theirfar kinematic distance solutions. The resulting trends are illustratedin Fig. 9, and the overall effect is, in two of the three cases, toslightly reduce the significance of the differences between stages i)–ii), i)–iii), and iii)-iv) compared to the original distance assignments,while the significance is increased in the third case. This tendencyto slightly reduce the significance can be mainly ascribed to theresulting smaller sample sizes, as the clumps now assigned to thefar-kinematic distance fall outside the distance limits of the sample,consequently increasing the error on the mean.Another choice affecting the results was the decision to adopt thecluster-averaged distances, 𝑑 cluster , as opposed to the dendrogramgroup-averaged distances, 𝑑 group . If we adopt the latter in place ofthe former, and repeat the analysis, we see almost no change in theresults, since the average difference in the two distances in on theorder of one hundred parsecs. In fact, as can be seen in panel b) ofFig. 9, the evolutionary trend is slightly strengthened in this scenariowith the largest change coming between stages i) and iii), now atthe 4.5- 𝜎 level of significance, compared to 4.3- 𝜎 in the originalanalysis. MNRAS000 , 1–21 (2020) A. J. Rigby et al.
The choice of distance limits used could also have an effect uponthe recovered trend, and we repeated the experiment by adoptingmore conservative distance limits, such that a maximum factor of1.5 difference in the spatial resolution at the near and far cut-off wasused, i.e. 4 < 𝑑 cluster < ( 𝑀 / 𝑀 (cid:12) ) between stages i) andiii) is still significant at the 3.7- 𝜎 level (panel c) of Fig. 9).We also note the presence of potential biases in both distanceand angular size as a function of 𝑓 IRB . Relative to one another, theaverage values of both quantities shows the same behaviour as theaverage mass throughout the four proposed evolutionary stages. Abias towards nearer distances for the more infrared-dark sources canbe understood by their requirement for a bright infrared backgroundagainst which they can be seen, though it is not clear why thereshould be a bias for the brightest sources to also be located closer tothe observer. The bias towards larger sources at intermediate values of 𝑓 IRB can be understood by the nature of the median filter, by whichsources should tend towards 𝑓 IRB = . ( .
35 kpc / 𝑑 cluster ) , and we includethe modified evolutionary trend in panel d) of Fig. 9. The bias indistance is extremely mild, and has little effect upon our conclusions.However, the effect of the bias in angular size is more difficult to test,and so we caution that there may be an effect at play here though,again, the effect is probably very mild, considering the overlappingdistributions. In Fig. 10, we also show the lack of any significant biasin 𝑓 IRB as result of the latitude of the clumps, indicating that thisparticular line of sight has a sufficiently bright infrared-backgroundso as to make any latitude-dependent bias negligible.In Sect. 4.3, the limits to 𝑓 IRB were selected to equalise the sizeof the four samples, and thereby eliminate the statistical effects ofvarying sample sizes. Since 𝑓 IRB is expected to trace the relativeevolution of each clump (however non-linearly), a secondary effectof this choice of samples is that a similar fraction of the clumps’ totallifetimes may be covered in each stage. We can, however, investigatewhether these particular boundaries of 𝑓 IRB that define the evolu-tionary stages, are responsible for the apparent sequence. Therefore,we conducted a further test, by defining the subsamples with equally-spaced 𝑓 IRB limits at 𝑓 IRB = . , . , .
75. The four samples i)–iv)now contain 656, 202, 148, and 248 clumps, and, again, the samegeneral trend is present (panel e) of Fig. 9), though at a reducedsignificance, with 2.35-, 1.57- and 2.73- 𝜎 changes in the value oflog ( 𝑀 ) between stages i)–ii), i)–iii), and iii)–iv), respectively. thatequal time is spent – on average – in each stage. It is conceivablethat other 𝑓 IRB limits could be adopted to either amplify or diminishthe significance of the proposed evolutionary tend, but with a lack ofphysical or statistical justification, we explore these no further here.Our calculation of temperatures and masses in Sect. 3.5, could alsobe having an effect upon the evolutionary trend. We adopted a value of 𝛽 = .
8, based upon the Galactic plane average value, as measuredby
Planck (Planck Collaboration 2011), and thereby calculated avalue of 𝜅 . However, the value of 𝛽 has been shown to vary as afunction of frequency (e.g. Planck Collaboration: et al. 2014), and sowe test the impact of a two-valued 𝛽 upon the reported evolutionarytrend. Following Planck Collaboration: et al. (2014), we adopt a far-infrared value of 𝛽 FIR = . ± .
08 for our calculation of colour d c l u s t e r / k p c R e q /a r c s ec . . . . . . f IRB − . − . . . . b / d e g Figure 10.
Average values of distance (top), angular size (middle) and cen-troid latitude (bottom) of GASTON clumps as a function of the infrared-brightfraction. The red data points show the mean values for samples i)–iv), and theblue data points show the mean values for clumps sampled in 0.1-wide binsof 𝑓 IRB . The vertical error bars indicate the standard deviation of the popu-lations, and the horizontal error bars indicate the 𝑓 IRB range of the sample.Dashed vertical lines indicate the boundaries used for the samples i)–iv). temperatures (as per Eq. 2), and a millimetre value of 𝛽 mm = . ± .
06, for the determination of 𝜅 used in the mass calculation (Eq.3). In panel f) of Fig. 9, it can be seen that the relative evolutionarytrends are unaffected by these changes, with the exception that thevalue of log ( 𝑀 / 𝑀 (cid:12) ) is systematically shifted downwards by ∼ In this paper, we present the Galactic Star Formation with NIKA2(GASTON) project, a guaranteed-time large programme using theIRAM 30 m telescope’s new millimetre continuum camera, NIKA2.The GASTON large-programme consists of two projects aimed atprobing the origin of different regions of the stellar initial mass func-tion: i) a survey of high- to intermediate-mass star-forming clumpswithin a 2 deg region of the inner Galactic plane (GP) centred on ℓ = . ◦ , 𝑏 = . ◦
05; ii) a search for pre-brown dwarf cores; and athird project: iii) consisting of a study of dust property variations innearby well-resolved cores.We have described the observing strategy for the GASTON GPproject, and presented the first results obtained primarily throughanalysis of the 1.15 mm photometric maps after ∼
40 per cent ofthe total integration time. A dendrogram extraction technique hasallowed us to isolate resolved and unresolved compact sources aswell as more extended structures, whilst maintaining a descriptionof the hierarchy of emission complexes. By extracting CO (1–0)
MNRAS , 1–21 (2020)
ASTON: Galactic Star Formation with NIKA2 spectra for each source using FUGIN (Umemoto et al. 2017) andRAMPS (Hogge et al. 2018) data in combination with the latestmodels of the Galactic rotation curve (Reid et al. 2019), we haveassigned kinematic distances to each structure and, in concert withHi-GAL (Molinari et al. 2016a) data, have been able to calculatetheir masses.Of the 2446 dendrogram structures that we have identified, a to-tal of 1467 structures are either compact or unsubstructured (withinthe constraints of the observations) and 22 per cent of these appearto be a new population of clumps with no counterpart in the Her-schel
Hi-GAL catalogues (Molinari et al. 2016a; Elia et al. 2017).These new clumps are, on average, less massive, and colder thantheir 250 µ m-detected counterparts, and represent a population ofsources for which the Herschel resolution element at wavelengths inexcess of 250 µ m is not sensitive to. The majority of these sourcesare candidates for starless clumps, with 80 per cent having no com-pact 70 µ mcounterparts. By the end of the GASTON GP observingcampaign, we expect this number to increase significantly.We have proposed a categorisation that describes the relative evo-lutionary stage of the clumps in terms of their infrared-bright frac-tion, 𝑓 IRB . The mean temperature of the clumps, along with thefraction of 70 µ m-bright sources, and the ratio bolometric luminos-ity to mass in each stage increases monotonically with the proposedsequence, supporting the use of 𝑓 IRB as an evolutionary indicator.In mass-temperature space, the mean position of the distribution ofclumps follows a path that agrees well with clump-fed scenarios forhigh-mass star formation, in which high-mass star-forming cores ac-crete mass from their parsec-scale environment, before finally losingmass at later stages as the dense gas initially associated with thestar-forming region is dispersed and accreted.Upon completion of the GASTON GP field, we will publish thefull catalogue of sources, and revisit the measurements we have madein this study.
ACKNOWLEDGEMENTS
We wish to thank the anonymous referee for their helpful commentsand suggestions that have improved the quality of this paper. AJRwould like to thank his newborn son, Joseph, whose tardy arrivalassisted with the completion of this work. AJR and NP would liketo thank the STFC for financial support under the consolidated grantnumber ST/N000706/1. We would like to thank the IRAM stafffor their support during the campaigns. The NIKA dilution cryo-stat has been designed and built at the Institut Néel. In particular,we acknowledge the crucial contribution of the Cryogenics Group,and in particular Gregory Garde, Henri Rodenas, Jean Paul Leg-geri, Philippe Camus. This work has been partially funded by theFoundation Nanoscience Grenoble and the LabEx FOCUS ANR-11-LABX-0013. This work is supported by the French National ResearchAgency under the contracts "MKIDS", "NIKA" and ANR-15-CE31-0017 and in the framework of the "Investissements d’avenir” program(ANR-15-IDEX-02). This work has benefited from the support of theEuropean Research Council Advanced Grant ORISTARS under theEuropean Union’s Seventh Framework Programme (Grant Agree-ment no. 291294). F.R. acknowledges financial supports provided byNASA through SAO Award Number SV2-82023 issued by the Chan-dra X-Ray Observatory Center, which is operated by the SmithsonianAstrophysical Observatory for and on behalf of NASA under con-tract NAS8-03060. A.B. acknowledges the support of the EuropeanUnion’s Horizon 2020 research and innovation program under theMarie Skłodowska-Curie Grant agreement No. 843008 (MUSICA). This research made use of the python packages astropy , (The As-tropy Collaboration et al. 2013, 2018), astrodendro , ipython (Perez & Granger 2007), numpy , scipy (Virtanen et al. 2020),matplotlib (Hunter 2007) and multicolorfits . This researchhas also made use of NASA’s Astrophysics Data System Biblio-graphic Services, topcat (Taylor 2005), and saoimageds9 (Joye& Mandel 2003). DATA AVAILABILITY
In keeping with the IRAM Large Program policy , the GASTONdata – including raw and processed products – will be made publiclyavailable 18 months after the last scheduling semester in which it isobserved. REFERENCES
Adam R., et al., 2018, A&A, 609, A115Aguirre J. E., et al., 2011, ApJS, 192, 4André P., et al., 2010, A&A, 518, L102André P., Di Francesco J., Ward-Thompson D., Inutsuka S.-I., Pudritz R. E.,Pineda J. E., 2014, Protostars and Planets VI, pp 27–51André P., Arzoumanian D., Könyves V., Shimajiri Y., Palmeirim P., 2019,A&A, 629, L4Aniano G., Draine B. T., Gordon K. D., Sandstrom K., 2011, Publications ofthe Astronomical Society of the Pacific, 123, 1218Battersby C., et al., 2011, A&A, 535, A128Battersby C., Bally J., Svoboda B., 2017, The Astrophysical Journal, 835,263Beckwith S. V. W., Sargent A. I., Chini R. S., Guesten R., 1990, The Astro-nomical Journal, 99, 924Benjamin R. A., et al., 2003, PUBL ASTRON SOC PAC, 115, 953Bonnell I. A., Bate M. R., 2006, Monthly Notices of the Royal AstronomicalSociety, 370, 488Bontemps S., Motte F., Csengeri T., Schneider N., 2010, A&A, 524, A18Bourrion O., et al., 2016, Journal of Instrumentation, 11, P11001Calvo M., et al., 2016, Journal of Low Temperature Physics, 184, 816Churchwell E., et al., 2009, Publications of the Astronomical Society of thePacific, 121, 213Clarke S. D., Whitworth A. P., Duarte-Cabral A., Hubber D. A., 2017,Monthly Notices of the Royal Astronomical Society, 468, 2489Dame T. M., Hartmann D., Thaddeus P., 2001, The Astrophysical Journal,547, 792Dempsey J. T., Thomas H. S., Currie M. J., 2013, ApJS, 209, 8Drabek E., et al., 2012, Monthly Notices of the Royal Astronomical Society,426, 23Dunham M. M., Crapsi A., Evans II N. J., Bourke T. L., Huard T. L., MyersP. C., Kauffmann J., 2008, Astrophysical Journal Supplement Series, The,179, 249Eden D. J., et al., 2020, Monthly Notices of the Royal Astronomical Society,498, 5936Elia D., et al., 2017, MNRAS, 471, 100 https://astropy.org/ http://dendrograms.org/ https://ipython.org/ https://numpy.org/ https://scipy.org/ https://matplotlib.org/ https://github.com/pjcigan/multicolorfits http://ds9.si.edu/ MNRAS000
Adam R., et al., 2018, A&A, 609, A115Aguirre J. E., et al., 2011, ApJS, 192, 4André P., et al., 2010, A&A, 518, L102André P., Di Francesco J., Ward-Thompson D., Inutsuka S.-I., Pudritz R. E.,Pineda J. E., 2014, Protostars and Planets VI, pp 27–51André P., Arzoumanian D., Könyves V., Shimajiri Y., Palmeirim P., 2019,A&A, 629, L4Aniano G., Draine B. T., Gordon K. D., Sandstrom K., 2011, Publications ofthe Astronomical Society of the Pacific, 123, 1218Battersby C., et al., 2011, A&A, 535, A128Battersby C., Bally J., Svoboda B., 2017, The Astrophysical Journal, 835,263Beckwith S. V. W., Sargent A. I., Chini R. S., Guesten R., 1990, The Astro-nomical Journal, 99, 924Benjamin R. A., et al., 2003, PUBL ASTRON SOC PAC, 115, 953Bonnell I. A., Bate M. R., 2006, Monthly Notices of the Royal AstronomicalSociety, 370, 488Bontemps S., Motte F., Csengeri T., Schneider N., 2010, A&A, 524, A18Bourrion O., et al., 2016, Journal of Instrumentation, 11, P11001Calvo M., et al., 2016, Journal of Low Temperature Physics, 184, 816Churchwell E., et al., 2009, Publications of the Astronomical Society of thePacific, 121, 213Clarke S. D., Whitworth A. P., Duarte-Cabral A., Hubber D. A., 2017,Monthly Notices of the Royal Astronomical Society, 468, 2489Dame T. M., Hartmann D., Thaddeus P., 2001, The Astrophysical Journal,547, 792Dempsey J. T., Thomas H. S., Currie M. J., 2013, ApJS, 209, 8Drabek E., et al., 2012, Monthly Notices of the Royal Astronomical Society,426, 23Dunham M. M., Crapsi A., Evans II N. J., Bourke T. L., Huard T. L., MyersP. C., Kauffmann J., 2008, Astrophysical Journal Supplement Series, The,179, 249Eden D. J., et al., 2020, Monthly Notices of the Royal Astronomical Society,498, 5936Elia D., et al., 2017, MNRAS, 471, 100 https://astropy.org/ http://dendrograms.org/ https://ipython.org/ https://numpy.org/ https://scipy.org/ https://matplotlib.org/ https://github.com/pjcigan/multicolorfits http://ds9.si.edu/ MNRAS000 , 1–21 (2020) A. J. Rigby et al.
Ellsworth-Bowers T. P., et al., 2013, ApJ, 770, 39Ellsworth-Bowers T. P., Rosolowsky E., Glenn J., Ginsburg A., Evans II N. J.,Battersby C., Shirley Y. L., Svoboda B., 2015, ApJ, 799, 29Frayer D., et al., 2020, in Bulletin of the American Astronomical Society. p.272.17Ginsburg A., et al., 2013, ApJS, 208, 14Griffin M. J., et al., 2010, A&A, 518, L3Heyer M., et al., 2018, Monthly Notices of the Royal Astronomical Society,473, 2222Hogge T., et al., 2018, ApJS, 237, 27Hunter J. D., 2007, Computing in Science and Engineering, 9, 90Inutsuka S.-i., Miyama S. M., 1997, ApJ, 480, 681Jackson J. M., et al., 2006, ASTROPHYS J SUPPL S, 163, 145Jackson J. M., et al., 2019, The Astrophysical Journal, 870, 5Joye W. A., Mandel E., 2003, Astronomical Data Analysis Software andSystems XII, 295, 489Kirk H., Myers P. C., Bourke T. L., Gutermuth R. A., Hedden A., WilsonG. W., 2013, The Astrophysical Journal, p. 14Klaassen P., et al., 2019, Astro2020: Decadal Survey on Astronomy andAstrophysics, APC white papers, no. 58; Bulletin of the American As-tronomical Society, 51, 58Könyves V., et al., 2015, A&A, 584, A91Könyves V., et al., 2020, A&A, 635, A34Lee Y.-N., Hennebelle P., Chabrier G., 2017, ApJ, 847, 114Liu T., et al., 2016, ApJ, 829, 59Lu X., et al., 2018, ApJ, 855, 9Luhman K. L., 2012, Annu. Rev. Astron. Astrophys., 50, 65Marsh K. A., et al., 2016, Mon. Not. R. Astron. Soc., 459, 342McKee C. F., Tan J. C., 2003, ApJ, 585, 850Molinari S., et al., 2010, A&A, 518, L100Molinari S., et al., 2016a, A&A, 591, A149Molinari S., Merello M., Elia D., Cesaroni R., Testi L., Robitaille T., 2016b,ApJ, 826, L8Moore T. J. T., et al., 2015, Mon. Not. R. Astron. Soc., 453, 4265Motte F., Bontemps S., Louvet F., 2018, Annu. Rev. Astron. Astrophys., 56,41Myers P. C., 2009, ApJ, 700, 1609Peñaloza C. H., Clark P. C., Glover S. C. O., Shetty R., Klessen R. S., 2017,Mon. Not. R. Astron. Soc., 465, 2277Peretto N., Fuller G. A., 2009, A&A, 505, 405Peretto N., et al., 2013, A&A, 555, A112Peretto N., et al., 2014, A&A, 561, A83Peretto N., Lenfestey C., Fuller G. A., Traficante A., Molinari S., ThompsonM. A., Ward-Thompson D., 2016, A&A, 590, A72Peretto N., et al., 2020, Monthly Notices of the Royal Astronomical Society,496, 3482Perez F., Granger B. E., 2007, Computing in Science Engineering, 9, 21Perotto L., et al., 2020, A&A, 637, A71Planck Collaboration 2011, A&A, 536, A25Planck Collaboration: et al., 2014, A&A, 564, A45Poglitsch A., et al., 2010, A&A, 518, L2Ragan S., et al., 2012, A&A, 547, A49Reid M. J., Dame T. M., Menten K. M., Brunthaler A., 2016, ApJ, 823, 77Reid M. J., et al., 2019, ApJ, 885, 131Rigby A. J., et al., 2018, A&A, 615, A18Roman-Duval J., Jackson J. M., Heyer M., Rathborne J., Simon R., 2010,ApJ, 723, 492Rosolowsky E. W., Pineda J. E., Kauffmann J., Goodman A. A., 2008, ApJ,679, 1338Ruppin F., et al., 2018, A&A, 615, A112Sanhueza P., Jackson J. M., Zhang Q., Guzmán A. E., Lu X., Stephens I. W.,Wang K., Tatematsu K., 2017, ApJ, 841, 97Schneider N., et al., 2012, Astronomy and Astrophysics, 540, L11Schuller F., et al., 2009, Astronomy and Astrophysics, 504, 415Stanke T., et al., 2019, Astro2020: Decadal Survey on Astronomy and Astro-physics, science white papers, no. 542; Bulletin of the American Astro-nomical Society, 51, 542Svoboda B. E., et al., 2016, ApJ, 822, 59 . ◦ . ◦ . ◦ . ◦ . ◦ . ◦ . ◦ . ◦ − . ◦ Galactic Longitude G a l a c t i c L a t i t ud e -20-1001020 F l u x d e n s i t y [ m J y / b e a m ] Figure A1.
Null map for the GASTON Galactic Plane field at 1.15 mm. Thepositions of the four sky apertures are shown as white polygons.Tan J. C., Beltrán M. T., Caselli P., Fontani F., Fuente A., Krumholz M. R.,McKee C. F., Stolte A., 2014, Protostars and Planets VI, pp 149–172Taylor M. B., 2005, in Astronomical Data Analysis Software and SystemsXIV. p. 29Terebey S., Chandler C. J., Andre P., 1993, The Astrophysical Journal, 414,759The Astropy Collaboration et al., 2013, A&A, 558, A33The Astropy Collaboration et al., 2018, AJ, 156, 123Traficante A., Fuller G. A., Billot N., Duarte-Cabral A., Merello M., Moli-nari S., Peretto N., Schisano E., 2017, Monthly Notices of the RoyalAstronomical Society, 470, 3882Umemoto T., et al., 2017, Publications of the Astronomical Society of Japan,69Urquhart J. S., et al., 2014, A&A, 568, A41Urquhart J. S., et al., 2018, Monthly Notices of the Royal AstronomicalSociety, 473, 1059Vázquez-Semadeni E., Gómez G. C., Jappsen A.-K., Ballesteros-Paredes J.,Klessen R. S., 2009, ApJ, 707, 1023Vázquez-Semadeni E., Palau A., Ballesteros-Paredes J., Gómez G. C.,Zamora-Avilés M., 2019, Monthly Notices of the Royal AstronomicalSociety, 490, 3061Virtanen P., et al., 2020, Nature Methods, 17, 261Wang P., Li Z.-Y., Abel T., Nakamura F., 2010, ApJ, 709, 27Whitworth A., Bate M. R., Nordlund Å., Reipurth B., Zinnecker H., 2007,Protostars and Planets V, pp 459–476Williams G. M., Peretto N., Avison A., Duarte-Cabral A., Fuller G. A., 2018,A&A, 613, A11
APPENDIX A: NULL MAPS
We display the 1,15 mm null map, used in Sect. 3.2, in Fig. A1.
APPENDIX B: BASIC SOURCE PROPERTIES
In Section 3.2, we described the dendrogram-based procedure usedto identify discrete sources within the 1.15 mm data, and here wedescribe their basic properties. We have measured two types of inte-grated flux density, termed ‘bijected’ and ‘clipped’, as in the schemeof Rosolowsky et al. (2008). The bijected integrated flux density of asource is calculated from the sum of pixel values within the source.By contrast, the clipped integrated flux density is a background-subtracted measurement, in which an estimate of the local back-ground (as described in Sect. 3.5) is first subtracted from each pixelvalue before calculating their sum. We further define a contrast pa-rameter for the clipped flux densities, which is the ratio of the peak tothe background flux density value, and has a value of unity for faint
MNRAS , 1–21 (2020)
ASTON: Galactic Star Formation with NIKA2 ( S peak / mJy beam − )0123456 l og ( S b i j ec t e d / m J y ) . . . . l og ( R e q /a r c s ec ) ( S peak / mJy beam − )0123456 l og ( S c li pp e d / m J y ) C o n t r a s t ( S clipped / mJy)0123456 l og ( S b i j ec t e d / m J y ) C o n t r a s t Figure B1.
Left panel: peak versus integrated (bijected) flux density for all sources, colour-coded according to sources’ equivalent radii. The dashed 1:1 lineshows the expected relationship for isolated point sources in all panels. The dotted line shows the lower bound of the data, where the integrated flux densityis equal to half of the expected value for a point source. Middle panel: peak versus integrated (clipped) flux density for all sources, colour-coded according tothe contrast parameter (see text). The dashed black line again shows the expected relationship for isolated point sources. The dotted line shows the minimumintegrated flux density of a point source at the detection threshold. Right panel: clipped versus bijected integrated flux density, coloured by contrast parameter. ( S peak / mJy beam − )10 N o . s o u r ce s All sourcesClumpsComplexes 0 2 4 6log ( S bijected / mJy)10 ( S clipped / mJy)10 ( R eq / arcsec)10 Figure B2.
Distributions of various source properties, separated into all sources (grey), clumps (sources containing no discernible substructure), and complexes(sources containing two or more clumps). sources at the detection threshold of our dendrogram extraction. Noneof the sources in our dendrogram extraction have a contrast value ofless than unity although, in principle, point sources located in regionswith a bright background could result in values of less than one.Figure B1 illustrates the relationship between peak flux density,and the two types of integrated flux density. The relationship be-tween peak and bijected flux densities shows that the majority ofsources lie above the line of equality, which isolated point sourcesat very high S/N would be expected to follow. Sources lying abovethis line are extended sources, as can be seen by the colour scaleindicating the sources’ equivalent radii, 𝑅 eq (the radius of a circlewith the equivalent area). In the case of the clipped flux densities,sources lying in regions with significant backgrounds may lie belowthe line of equality, and indeed the most significant departures oc-cur for sources located in the region around ℓ = . ◦ 𝑏 = − . ◦ 𝑆 bijected = . 𝑆 . .In Fig. B2 we display histograms of the peak and integrated fluxdensities, and equivalent radii of all sources identified by the den-drogram. The distributions of further subsamples are shown, andwe show the population of clumps (sources containing no furtherobserved substructures), and complexes (dendrogram ’trunks’ con-taining more than one clump). Note that in the case of the peak fluxdensities, the peak flux within a complex is the same value as thepeak flux of its brightest constituent clump, and so the complex dis-tribution is a sub-set of the clump distribution. The brightest clumpshave peak flux densities of ∼
10 Jy beam − , and the largest com-plexes have integrated flux densities of ∼
100 Jy. Source sizes rangefrom the minimum detectable size of 𝑅 eq = . APPENDIX C: VELOCITY ASSIGNMENTS
In Sect. 3.3, we derived an initial velocity assignment for each sourcebased on the channel within a source area-averaged spectrum withthe maximum intensity. A level of robustness for each velocity as-signment was estimated by determining the standard deviation ofthe velocities associated with the three brightest channels withineach spectrum. In cases where this standard deviation is less than1 km s − (i.e. the three brightest channels are almost perfectly adja- MNRAS000
In Sect. 3.3, we derived an initial velocity assignment for each sourcebased on the channel within a source area-averaged spectrum withthe maximum intensity. A level of robustness for each velocity as-signment was estimated by determining the standard deviation ofthe velocities associated with the three brightest channels withineach spectrum. In cases where this standard deviation is less than1 km s − (i.e. the three brightest channels are almost perfectly adja- MNRAS000 , 1–21 (2020) A. J. Rigby et al. . ◦ . ◦ . ◦ . ◦ . ◦ . ◦ . ◦ Galactic Longitude G a l a c t i c L a t i t ud e − . ◦ . ◦ . ◦ . ◦ . ◦ . ◦ . ◦ Galactic Longitude25 50 75 100km s −
01 193801 N o r m a li s e d C O ( ) i n t e n s i t y − −
50 0 50 100 150 200Velocity [km s − ]01 1289 ID 𝑣 CO flag 𝑆 / 𝑁 𝑣 m1 𝑣 c1 (km s − ) (km s − ) (km s − )1289 39.4 ∗ ∗ ∗ ∗ – Figure C1.
Example of the velocity determinations of four sources within a sub-set of the data. In the left panel, the 1.15 mm flux density map is shown (aftersmoothing to 13 arcsec resolution), with contours overlaid corresponding to the footprints of catalogued sources with ID numbers 1289, 1620, 1705, and 1938.In the central panel, the corresponding CO (1–0) spectra are presented, which were extracted over the footprint of the sources, with the velocity associated withthe maximum intensity is assigned to that source indicated by a solid line above the spectra, and any velocity information determined from either the first-fittedvelocity component or first moment maps of the RAMPS data are indicated by dot-dashed and dotted lines below the spectra. The right-hand image shows theRAMPS NH (1,1) first moment (intensity weighted coordinate) map from Hogge et al. (2018), with the same source outlines overlaid. In the Table, details ofthe CO-derived velocity ( 𝑣 CO ), initial velocity flag, the S/N of the CO (1–0) spectrum, and the source velocities derived from the RAMPS NH (1,1) firstmoment ( 𝑣 m1 ) and first velocity component ( 𝑣 c1 ), are given for each source. In each case, an asterisk denotes the adopted velocity, 𝑣 cen . cent), the assignment is considered to be robust and assigned a flagvalue 3; where the standard deviation is more than 5 km s − , theassignment is considered to be poor and assigned a flag value of 1;cases between those limits are assigned an intermediate robustnessflag value of 2. In practice, these flags allow the most robust velocities– where there is a single dominant peak – to be isolated from caseswith either a low S/N or from spectra in which there are multiplevelocity components of similar strength. In this initial velocity as-signment, we record 1680 (69 per cent) of sources with a robustnessflag of 3, 305 (12 per cent) sources with an intermediate robustnessflag, and 461 (19 per cent) with a poor robustness flag.Following this initial velocity assignment, we make further veloc-ity estimates by using the more appropriate – but limited in coverage– data of the NH (1,1) inversion transition from the RAMPS survey.The RAMPS pilot study (Hogge et al. 2018) presented, as data prod-ucts, maps of source velocities derived from a line-fitting procedure,as well maps of the first moment, with the latter having a greater ex-tent, but the former being more accurate. For each GASTON source,and for each RAMPS velocity map, we record the mean velocity ofpixels that fall within the source area. After excluding measurementswhere either the mean error (in the case of the fitted first compo-nents) or the standard deviation on the mean velocities (in the caseof the first moments) are greater than 5 km s − , we record a totalof 523 first-component velocities and 1097 first moment velocities.Where available, we replace the CO-derived 𝑣 cen values with their NH -derived counterparts, preferring the line-fitted measurementsover the first moment-derived velocities, and adopt a robustness flagof 3. At this stage, the 𝑣 cen assignments are made up of 1407 (58per cent) CO-derived velocities, 516 (21 per cent) with NH firstmoment-derived velocities, and the remaining 523 (21 per cent) withvelocities derived from the NH fitted first velocity components.There are 1913 (78 per cent), 174 (7 per cent) and 359 (15 per cent)of 𝑣 cen assignments with quality flags of 3, 2, and 1, respectively. 888(85 per cent) of the NH -derived velocities agree with their initial CO-derived assignment to better than 5 km s − , indicating that theinitial assignments are generally reliable.In Fig. C1, we illustrate the method used in Sect. 3.3 for theinitial velocity assignments for each of the sources identified in Sect.3.2. Where possible, we also display the velocity derived from theavailable RAMPS data. At this point, we refer the reader back to themain text of Sect. 3.3 for a description of how the emission structuresare used to refine the individual velocity measurements. APPENDIX D: LINE CONTAMINATION
The 1.15 mm NIKA2 band spans a range in frequency from 210–300GHz (taking these edges as 20 per cent transmission), and as such,the continuum may contain important contributions from the CO(2–1) line at 230.538 GHz, especially for observations of the inner
MNRAS , 1–21 (2020)
ASTON: Galactic Star Formation with NIKA2 Galactic plane. To estimate the level of this line contamination in the1.15 mm maps, we made use of CO (1–0) data from the FORESTunbiased Galactic plane imaging survey with the Nobeyama 45 mtelescope (FUGIN Umemoto et al. 2017). FUGIN covers the entireGASTON GP region, with an angular resolution of 20 arcsec.The CO (1–0) data were integrated over the full velocity range of −
50 to 200 km s − , and multiplied by a line ratio of 𝑅 − / − = . − )to surface brightness units (Jy beam − ) following the procedure out-lined in Drabek et al. (2012), which accounts for the NIKA2 filterresponse. Finally, we applied a 6.5 arcmin filter to the CO data toapproximate the spatial filtering applied by the NIKA2 data reduc-tion pipeline, using the nebuliser application. Both the CO (2–1)and NIKA2 1.15 mm surface brightness maps were smoothed to acommon resolution of 22 arcsec, and we measured their ratio forpixels that had a 1.15 mm S/N of more than 10.The level of CO (2–1) line contamination in the valid pixels rangesfrom 0 to ∼
15 per cent, with a median value of 4 per cent. The upperlimit of this measurement is most likely to be a worst-case scenario,and would apply in cases where NIKA2 is fully recovering emissionup to 6.5 (cid:48) scales. Furthermore, the adopted 𝑅 − / − value is thehigher value from what is usually measured as a bimodal distribution,with a lower value of 0.3 often found in regions of sub-thermallyexcited low density CO. We do not, therefore, directly account forCO (2–1) line contamination in our analysis, but incorporate anadditional factor of 4 per cent as the associated uncertainty on theflux density.We note that these values are comparable to those estimated by(Rigby et al. 2018) for NIKA, who estimated CO contamination at the1–3 per cent level. A slightly higher value is found here for a numberof reasons: i) the FUGIN data allow us to estimate the contaminationat higher angular resolution; ii) The FUGIN data allow us to estimatethis quantity in the correct isotopologue of CO whereas the Rigbyet al. (2018) study used an approximate conversion from CO, whichis less effected by optical depth effects; iii) larger spatial frequenciesare able to recovered with NIKA2 than with NIKA, and so morelarge-scale CO emission survives the pipeline’s filtering.
APPENDIX E: DISTANCE COMPARISON
In Fig. E1, we compare the distances determined for GASTONclumps in Sect. 3.4 with the distances to their closest positionally-matched counterparts from the BGPS, ATLASGAL, and Hi-GALsurveys, whose catalogues were cross-matched in Sect. 4.1. We findthat 64 and 69 per cent of GASTON clumps have distances that differby less than 1 kpc when compared to the clump catalogues from theBGPS (Ellsworth-Bowers et al. 2015) and ATLASGAL (Urquhartet al. 2018), respectively. While only 25 per cent of distance determi-nations agree with Hi-GAL clumps from the band-merged catalogueof Elia et al. (2017), this value rises to 67 per cent once sourceswithout a ‘good’ distance quality flag are removed. School of Physics and Astronomy, Cardiff University, Queen’s Build-ings, The Parade, Cardiff, CF24 3AA, UK LLR (Laboratoire Leprince-Ringuet), CNRS, École Polytechnique,Institut Polytechnique de Paris, Palaiseau, France AIM, CEA, CNRS, Université Paris-Saclay, Université ParisDiderot, Sorbonne Paris Cité, 91191 Gif-sur-Yvette, France Univ. Grenoble Alpes, CNRS, IPAG, 38000 Grenoble, France | d catalogue − d k | / kpc0 . . . . . . C u m u l a t i v e f r a c t i o n BGPSATLASGALHi-GAL (good)Hi-GAL (all)
Figure E1.
Cumulative distribution showing the difference in distances de-termined for GASTON clumps and their counterparts from the BGPS (blue),ATLASGAL (orange), and Hi-GAL (red) catalogues. Institut d’Astrophysique Spatiale (IAS), CNRS and Université ParisSud, Orsay, France Institut Néel, CNRS and Université Grenoble Alpes, France Institut de RadioAstronomie Millimétrique (IRAM), Grenoble,France Univ. Grenoble Alpes, CNRS, Grenoble INP, LPSC-IN2P3, 53, av-enue des Martyrs, 38000 Grenoble, France Rudjer Bošković Institute, Bijenička cesta 54, 10000 Zagreb, Croa-tia Dipartimento di Fisica, Sapienza Università di Roma, PiazzaleAldo Moro 5, I-00185 Roma, Italy Chinese Academy of Sciences South America Center for Astron-omy, National Astronomical Observatories, CAS, Beijing 100101,PR China Instituto de Astronomía, Universidad Católica del Norte, Av. Ang-amos 0610, Antofagasta 1270709, Chile Centro de Astrobiología (CSIC-INTA), Torrejón de Ardoz, 28850Madrid, Spain Instituto de Radioastronomía Milimétrica (IRAM), Granada, Spain Aix Marseille Univ, CNRS, CNES, LAM (Laboratoired’Astrophysique de Marseille), Marseille, France LERMA, Observatoire de Paris, PSL Research University, CNRS,Sorbonne Universités, UPMC Univ. Paris 06, 75014 Paris, France School of Earth and Space Exploration and Department of Physics,Arizona State University, Tempe, AZ 85287, USA Univ. Toulouse, CNRS, IRAP, 9 Av. du Colonel Roche, BP 44346,31028, Toulouse, France Département de Physique, Ecole Normale Supérieure, 24, rueLhomond 75005 Paris, France Department of Physics and Astronomy, University of Pennsylvania,209 South 33rd Street, Philadelphia, PA, 19104, USA Institut d’Astrophysique de Paris, Sorbonne Université, CNRSUMR7095, 75014 Paris, France Kavli Institute for Astrophysics and Space Research, MassachusettsInstitute of Technology, Cambridge, MA 02139, USA Astronomisches Rechen-Institut, Zentrum f ur Astronomie der Uni-versit at Heidelberg, M onchhofstraße 12-14, D-69120 Heidelberg,Germany
This paper has been typeset from a TEX/L A TEX file prepared by the author.MNRAS000