Molecular line mapping of the giant molecular cloud associated with RCW 106 - III. Multi-molecular line mapping
N. Lo, M. R. Cunningham, P. A. Jones, I. Bains, M. G. Burton, T. Wong, E. Muller, C. Kramer, V. Ossenkopf, C. Henkel, G. Deragopian, S. Donnelly, E.F. Ladd
MMon. Not. R. Astron. Soc. , 1–25 () Printed 1 November 2018 (MN L A TEX style file v2.2)
Molecular line mapping of the giant molecular cloudassociated with RCW 106 -
III. Multi-molecular line mapping
N. Lo, , (cid:63) M. R. Cunningham, P. A. Jones, , I. Bains, , M. G. Burton, T. Wong, , , E. Muller, , † C. Kramer, V. Ossenkopf, , C. Henkel, G. Deragopian, S. Donnelly and E. F. Ladd School of Physics, University of New South Wales, Sydney, NSW 2052, Australia Australia Telescope National Facility, CSIRO, PO Box 76, Epping, NSW 1710, Australia Departamento de Astronom ˜Aa, Universidad de Chile, Casilla 36-D, Santiago, Chile Centre for Astrophysics and Supercomputing, Swinburne University of Technology, P.O. Box 218, Hawthorn, VIC 3122, Australia Astronomy Department, University of Illinois, 1002 W. Green St, Urbana, IL 61801, USA Department of Astrophysics, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8602, Japan Instituto de Radioastronomia Milimetrica (IRAM), Avda. Divina Pastora 7, E-18012 Granada, Spain I. Physikalisches Institut, Universit¨at zu K¨oln, Z¨ulpicher Stra β e 77, 50937 K¨oln, Germany SRON, Netherlands Institute for Space Research, PO Box 800, 9700 AV Groningen, The Netherlands Max-Planck-Institut f¨ur Radioastronomie, Auf dem H ¨ u gel 69, 53121 Bonn, Germany Centre for Astronomy, James Cook University, Townsville, Australia Department of Physics and Astronomy, Bucknell University, Lewisburg PA 17837, USA
Accepted ***. Received ***; in original form ***
ABSTRACT
We present multi-molecular line maps obtained with the Mopra Telescope towards thesouthern giant molecular cloud (GMC) complex G333, associated with the H ii regionRCW 106. We have characterised the GMC by decomposing the 3D data cubes with gaussclumps , and investigated spatial correlations among different molecules withprincipal component analysis (PCA). We find no correlation between clump size andline width, but a strong correlation between emission luminosity and line width. PCAclassifies molecules into high and low density tracers, and reveals that HCO + andN H + are anti-correlated. Key words: stars: formation - ISM: clouds - ISM: molecules - ISM: structure - radiolines: ISM
This is the third paper (paper I by Bains et al. 2006 andpaper II by Wong et al. 2008) in a series of multi-molecularlines observations of the giant molecular cloud (GMC)associated with RCW 106, G333. The G333 giant molecularcloud, a southern massive star forming region, spans1 . ◦ × . ◦ l ∼ ◦ , b ∼ − . ◦ α J = 16 h m , δ J = − d m ), at a distance of3.6 kpc (Lockman 1979). The aim of the multi-molecularline study was to investigate the relationship between thedynamics of the interstellar medium and star formation.Further investigation and analysis planned include powerspectra to study the role of turbulence in the GMC, andsearching for evidence of triggered star formation. (cid:63) E-mail: [email protected] † Bolton Fellow, ATNF
The G333 GMC has previously been studied in otherwavelengths, for example in the far-infrared by Karniket al. (2001) and the 1.2-mm dust continuum by Mookerjeaet al. (2004). A search for water masers was carried outby Breen et al. (2007). There are also numerous workson specific regions in this GMC, such as RCW 106 (e.g.Rodgers, Campbell & Whiteoak 1960, Russeil et al. 2005)and G333.6 − ii regions (e.g. Goss & Shaver 1970, Becklin et al. 1973,Storey et al. 1989, Colgan et al. 1993, Fujiyoshi et al.1998, 2001, 2005 and 2006). However, to date, there hasbeen no systematic, 3-mm multi-molecular line mappingof this GMC. This study demonstrates the full capabil-ity of the Mopra Telescope’s new digital filter bank, The Australia Telescope Mopra telescope is part of the Aus-tralia Telescope, which is funded by the Commonwealth of Aus- ' RAS a r X i v : . [ a s t r o - ph . S R ] F e b N. Lo et al.
Table 1.
List of observed molecular transitions mapped in G333. The columns are: (1) molecule; (2) transition; (3) rest frequency (Lovas,Johnson & Snyder 1979); (4) upper energy levels from the Cologne Database for Molecular Spectroscopy (CDMS, M¨uller et al. 2001,2005); (5) calculated critical density: unless specified, n crit = A ul / (cid:104) σ ( v ) v (cid:105) where A ul is the Einstein A coefficient obtained from CDMS, (cid:104) σ ( v ) (cid:105) = 10 − cm is the collision cross section and v is the velocity of collision particles assumed to be 1 km s − ; (6) observationseason; (7) backend used, where MPCOR stands for Mopra Correlator, which was decommissioned in 2005 and replaced with a newdigital filter bank, the UNSW-Mopra Spectrometer (UNSW-MOPS); (8) notes, where [hf] denotes hyperfine structures, with the restfrequency quoted being that of the strongest hyperfine component, and [ (cid:132) ] denotes molecules that have detectable emission at a fewplaces only. Their emission maps are not presented in this work but will appear in subsequent papers on specific regions in the G333cloud. Molecule Transition Rest frequency E u /k n crit Observation season Backend Notes (GHz) (K) (cm − )13 CO 1 − .
353 5.29 3 × O 1 − .
173 5.27 3 × Jul, 2005 MPCORC H 1 − .
925 4.19 2 × Jul, 2006 MOPS [hf]CH OH 2(0 , − , A + 96741 .
377 6.97 7 × Sep, 2006 MOPS [ (cid:132) ]CS 2 − .
953 7.05 2 × Sep, 2006 MOPSC S 2 − .
961 6.94 2 × Sep, 2006 MOPS [ (cid:132) ]HCN 1 − .
847 4.25 2 × Jul, 2006 MOPS [hf]H CN 1 − .
167 4.14 2 × Jul, 2006 MOPS [hf, (cid:132) ]HCCCN 10 − .
989 24.01 6 × Sep, 2006 MOPS [ (cid:132) ]HCCCN 11 −
10 100076 .
385 28.82 8 × Sep, 2006 MOPS [ (cid:132) ]HCO + − .
526 4.28 4 × Jul, 2006 MOPSH CO + − .
330 4.16 4 × Jul, 2006 MOPS [ (cid:132) ]HNC 1 − .
572 4.35 3 × Jul, 2006 MOPS [hf]N H + − .
480 4.47 4 × Sep, 2006 MOPS [hf]SiO 2 − ν = 0 86847 .
010 6.25 3 × Jul, 2006 MOPS [ (cid:132) ]SO 3(2) − .
905 9.23 1 × Sep, 2006 MOPS [ (cid:132) ]4(5) − .
565 38.58 3 × a Sep, 2006 MOPS [ (cid:132) ] a Ungerechts et al. (1997) the UNSW-Mopra Spectrometer (UNSW-MOPS). Othersimilar surveys utilising the UNSW-MOPS are the H OSouthern Galactic Plane Survey (HOPS) by Walsh et al.(2008) and the Central Molecular Zone of the Galaxy byJones et al. (2008).We present in this paper results of molecular linemapping carried out during July to November in 2006, con-sisting of molecular lines from 83 to 101 GHz. For a detailedanalysis of CO and C O data please refer to previouswork by Bains et al. (2006, hereafter BWC2006) and Wonget al. (2008, hereafter WLB2008). In Section 2 we describethe observing technique, and the new UNSW-MOPS digitalfilter bank. In Section 3 we present the results, and examinethe velocity and spatial distribution of the molecularemission from the different molecules. In Section 4 and5, we present two different approaches to characterisingemission distribution in the GMC: Clump finding with gaussclumps and principal component analysis (PCA). InSection 6 we discuss results, including specific regions ofthe GMC. Finally, we summarise our findings in Section 7.Data presented here will be made available in the nearfuture, please contact the authors for details.
The data were collected with the 22 metre Mopra Telescope,which is a centimetre- and millimetre-wavelength antennahaving a full width to half-maximum (FWHM) beam sizeof ∼ (cid:48)(cid:48) at 100-GHz (Ladd et al. 2005). The observationswere carried out with the narrow band mode of the newUNSW-Mopra Spectrometer (UNSW-MOPS) digital fil-terbank back-end, and a Monolithic Microwave IntegratedCircuit (MMIC) 77 to 116 GHz receiver. The observingparameters, including rest frequencies and dates are listedin Table 1. UNSW-MOPS has a 8-GHz bandwidth withfour overlapping 2.2-GHz subbands, each subband havingfour dual-polarisation 137.5-MHz-wide windows giving atotal of sixteen dual-polarisation windows. Each windowhas 4096 channels providing a velocity resolution of ∼ − per channel at 100 GHz. The velocities presented inthis work are with respect to the kinematic local standardof rest (LSR), with v LSR = −
50 km s − being the systemicvelocity of the G333 complex.The brightness temperature T b is related to theantenna temperature T ∗ A by T b = T ∗ A /η ν , where η ν isthe frequency dependent beam efficiency. According toLadd et al. (2005) the main beam efficiency at 86 GHz is η
86 GHz = 0 .
49, and at 110 GHz is η
110 GHz = 0 .
44. Thebeam efficiencies are η
86 GHz = 0 .
65 and η
86 GHz = 0 . T ∗ A unlessotherwise specified.We followed a similar mapping procedure to ourprevious work ( CO; BWC2006). The ∼ ×
300 square arcsec fields, ' RAS, MNRAS , 1–25 ulti-molecular lines mapping of the GMC associated with RCW 106 with pointing centres separated by 285 arcsec in rightascension and declination, allowing a 15-arcsec overlapbetween adjacent fields. Each field was mapped with twopasses, the first pass scanning in right ascension and thesecond pass scanning in declination. The observing modewas ‘on-the-fly’(OTF) raster scanning, at a scan rate of3.5 arcsec s − and averaging data over a 2-s cycle time.We used the CO map as a guide, mapping the brightest CO fields first, then extending these to cover the majorityof the emission in most detected molecules, resulting inmaps covering ∼ ∗ A during the whole observing season iswithin 10 per cent.The data were reduced with the livedata and gridzilla packages available from the ATNF, written byMark Calabretta . livedata performs a bandpass calibra-tion using the preceding reference (off-source) scan, thenfits the spectral baseline with a first degree polynomial. Dueto low sensitivity near the start and end of the window, thefirst and last 750 channels were discarded before baselinefitting. gridzilla grids the data according to user specifiedweighting and beam parameter inputs. In this work, ourdata was weighted by the relevant system temperature(T sys ) and Gaussian smoothed during data reduction stage.The data cubes were gridded with three pixels across eachbeam width, therefore each pixel is ∼ ×
15 arcsec fordata with 87-GHz frequency setting, and ∼ ×
12 arcsec for the 97-GHz setting. Note that all data cubes presentedin this work are regridded to 15 ×
15 arcsec pixel size forease of comparison, and the velocity resolution is left as ∼ . − . Different molecules may trace different physical conditionswithin the region, such as density, chemistry, temperature,evolutionary stage, and so on. Our selection of moleculescovers those which may trace differences in density, temper-ature and dynamic features such as outflow, allowing us tostudy environmental varieties within this giant molecularcloud. Since some molecules (e.g. CH OH, HCCCN) havedetectable emission at a few places only, their emissionmaps and velocity profiles are not presented here, but insubsequent works focusing on specific regions. Table 2.
A list of the emission range, peak brightness and 1 σ level of the integrated emission maps shown in Figures 1 and 2.Molecule Emission range 1 σ Peak brightness (km s − ) (K km s − ) (K km s − )13 CO −
70 to −
40 0.3 95C O −
70 to −
40 0.2 19CS −
70 to −
40 0.2 30HCO + −
70 to −
40 0.1 21HCN −
80 to −
30 0.1 21HNC −
70 to −
40 0.1 18N H + −
80 to −
30 0.2 11C H −
60 to −
40 0.2 5
Shown in Figure 1 and 2 are the integrated intensity(zeroth moment) maps of the observed molecular lines(contours) overlaid on the
Spitzer
Galactic Legacy InfraredMidplane Survey Extraordinaire (GLIMPSE; Benjaminet al. 2003) 8.0- µm image (grey scale). The integratedvelocity range is −
70 to −
40 km s − for CO (BWC2006),C O (WLB2008), CS, HCO + and HNC, and −
60 to − − for C H. A wider velocity range was chosen formolecules with hyperfine components: −
80 to −
30 km s − for HCN and N H + , so as to include emission from allcomponents. The choice of velocity range reflects emissionfrom the GMC only. Emission outside this range comesfrom spiral arms and intermittent clouds. The maps wereGaussian smoothed with an FWHM of 45 arcsec. The labelsin the CO total intensity map (Figure 1a) indicate someof the molecular regions in proximity to bright H ii regionsand IRAS sources, as marked by rectangles. The startingcontour levels are based on the lowest detectable emissionabove noise, hence outlining the emission structure of thecloud. Note that contours near the edges of the maps arenoisy; this is because the edges were scanned with onepass only and thus have lower sensitivity. Details of theintegrated emission maps are summarised in Table 2.The CS total intensity map (Figure 1c) shows a moreconfined distribution than the CO emission. This isas expected from a dense gas tracer such as CS, with acritical density of ∼ cm − . However CS has a similardistribution to another CO isotope presented here − C O.HCO + (Figure 1d), a common ionic species found inmolecular clouds also presents a similar distribution to CS,as do HCN (Figure 2a) and HNC (Figure 2b).A comparison of HCN and HNC total intensity maps(Figures 2a and b), shows that they both have similardistributions in general. This is surprising because ofthe strong chemical differences between HCN and HNC,meaning that its abundance ratio can depend heavily on,for example, temperature (e.g. Schilke et al. 1992; Hirotaet al. 1998). We do note there are local variations wheretemperature changes, such as the massive dense cold corewhere HCN and HNC have different spatial distribution(Lo et al. 2007).The total emission map of N H + (Figure 2c) showsthat this molecule has a more compact and clumpier dis-tribution than other molecules. It also shows regions withintense N H + emission which show diffuse emission in othermolecules; here we highlight two such regions on the map(indicated by arrows). The source G333.125 − ' RAS, MNRAS000
30 km s − for HCN and N H + , so as to include emission from allcomponents. The choice of velocity range reflects emissionfrom the GMC only. Emission outside this range comesfrom spiral arms and intermittent clouds. The maps wereGaussian smoothed with an FWHM of 45 arcsec. The labelsin the CO total intensity map (Figure 1a) indicate someof the molecular regions in proximity to bright H ii regionsand IRAS sources, as marked by rectangles. The startingcontour levels are based on the lowest detectable emissionabove noise, hence outlining the emission structure of thecloud. Note that contours near the edges of the maps arenoisy; this is because the edges were scanned with onepass only and thus have lower sensitivity. Details of theintegrated emission maps are summarised in Table 2.The CS total intensity map (Figure 1c) shows a moreconfined distribution than the CO emission. This isas expected from a dense gas tracer such as CS, with acritical density of ∼ cm − . However CS has a similardistribution to another CO isotope presented here − C O.HCO + (Figure 1d), a common ionic species found inmolecular clouds also presents a similar distribution to CS,as do HCN (Figure 2a) and HNC (Figure 2b).A comparison of HCN and HNC total intensity maps(Figures 2a and b), shows that they both have similardistributions in general. This is surprising because ofthe strong chemical differences between HCN and HNC,meaning that its abundance ratio can depend heavily on,for example, temperature (e.g. Schilke et al. 1992; Hirotaet al. 1998). We do note there are local variations wheretemperature changes, such as the massive dense cold corewhere HCN and HNC have different spatial distribution(Lo et al. 2007).The total emission map of N H + (Figure 2c) showsthat this molecule has a more compact and clumpier dis-tribution than other molecules. It also shows regions withintense N H + emission which show diffuse emission in othermolecules; here we highlight two such regions on the map(indicated by arrows). The source G333.125 − ' RAS, MNRAS000 , 1–25
N. Lo et al.
Figure 1.
Integrated emission maps (contours) for CO ( J = 1 − O ( J = 1 − J = 2 −
1) and HCO + ( J = 1 −
0) overlaidon a
Spitzer
IRAC 8.0- µm image (grey scale). The maps are integrated over a velocity range of −
70 to −
40 km s − and were clipped ata 3 σ level. 1 σ is 0.3 K km s − for CO, 0.2 K km s − for C O and CS, and 0.1 K km s − for HCO + . The maps were then Gaussiansmoothed with an FWHM of 45 arcsec. The contour levels start at 1.5 K km s − for CO, 2.0 K km s − for C O, 1.4 K km s − for CSand 1.0 K km s − for HCO + . Then each successive contour level is double the value of the previous one. The choice of lowest contourlevel is based on the lowest detectable emission above noise, except for CO where there is emission detected throughout the map.The temperatures are in terms of the antenna temperature, T ∗ A . Previously designated H ii regions and the associated IRAS sources aremarked with white rectangles and names are shown in the CO map. The scale bar shown in the CO map indicates 5-pc at a distanceof 3.6-kpc and the tiny circle in the lower left corner of each panel show the FWHM beam size after smoothing. The velocity range usedhere reflects emission from the GMC only; emission outside this range comes from more distant spiral arms or intervening clouds. ' RAS, MNRAS , 1–25 ulti-molecular lines mapping of the GMC associated with RCW 106 Figure 2.
Integrated emission maps (contours) for HCN ( J = 1 − J = 1 − H + ( J = 1 −
0) and C H ( J = 1 −
0) overlaidon
Spitzer
IRAC 8- µm images (grey scale). The integrated velocity ranges are of −
80 to −
30 km s − for HCN and N H + , −
70 to − − for HNC and −
60 to −
40 km s − for C H. The maps were clipped at a 3 σ level. 1 σ is 0.1 K km s − for HCN and HNC, and0.2 K km s − for N H + and C H. The maps were then Gaussian smoothed with an FWHM of 45 arcsec. The contour levels start at 2.0K km s − for HCN, 1.3 K km s − for HNC, 1.6 K km s − for N H + and 1.2 K km s − for C H, then each successive level is doublethe previous one. The choice of lowest contour levels are based on the lowest detectable emission above noise. The temperatures are interms of T ∗ A and the tiny circle in the lower left corner of each panel show the FWHM beam size after smoothing. Previously designatedH ii regions and the associated IRAS sources are marked with white rectangles (see Figure 1 for their naming). The arrows indicate twoN H + emission peaks, which are not prominent in other emission maps. Note that the differences in velocity range between moleculesare due to hyperfine splitting such that some molecules have a wider velocity distribution. The velocity range used here reflects emissionfrom the GMC only; emission outside this range is from more distant spiral arms or intervening clouds. ' RAS, MNRAS000
40 km s − for C H. The maps were clipped at a 3 σ level. 1 σ is 0.1 K km s − for HCN and HNC, and0.2 K km s − for N H + and C H. The maps were then Gaussian smoothed with an FWHM of 45 arcsec. The contour levels start at 2.0K km s − for HCN, 1.3 K km s − for HNC, 1.6 K km s − for N H + and 1.2 K km s − for C H, then each successive level is doublethe previous one. The choice of lowest contour levels are based on the lowest detectable emission above noise. The temperatures are interms of T ∗ A and the tiny circle in the lower left corner of each panel show the FWHM beam size after smoothing. Previously designatedH ii regions and the associated IRAS sources are marked with white rectangles (see Figure 1 for their naming). The arrows indicate twoN H + emission peaks, which are not prominent in other emission maps. Note that the differences in velocity range between moleculesare due to hyperfine splitting such that some molecules have a wider velocity distribution. The velocity range used here reflects emissionfrom the GMC only; emission outside this range is from more distant spiral arms or intervening clouds. ' RAS, MNRAS000 , 1–25
N. Lo et al.
13 to 19 K derived from the NH (J,K) = (1,1) and (2,2)inversion line and from the spectral energy distribution(SED). It has broad thermal SiO emission, and is believedto be a deeply embedded young stellar object in an earlystage of star formation (Lo et al. 2007). In the northernpart of the map, the N H + emission indicated by the toparrow is similarly interesting. Here, bright N H + emissionis detected where other molecules show diffuse weakemission. A comparison with the GLIMPSE 8.0- µ m imageclearly shows that N H + aligns well with the infrared darkfilaments, as discussed later in Section 6.3. While N H + seems to be correlated with infrared dark filaments, C Hdoes the opposite; it seems to correlate with GLIMPSE8.0- µ m emission.Among the eight molecules we present, C H emissionis the weakest. Its spatial distribution is more extendedthan the quiescent gas tracer N H + , but not as extendedas other molecules. Spatially averaged spectral emission profiles of the moleculespresented in Section 3.1 are shown in Figure 3. Note thatthe apparent lower signal-to-noise of the N H + and C Hspectra is due to these two molecules being less extendedspatially. To illustrate this, spectra of specific regions(IRAS16172 − − − ∼ −
50 km s − , the systemic velocity ofthe G333 cloud, and spans a velocity range of ∼ −
70 to ∼ −
40 km s − . CO, C O and CS have similar velocitystructure, such as the emission ‘shoulders’ at ∼ −
40 and ∼ −
55 km s − . The ∼ −
55 km s − emission feature alsoappears in the HCO + and HNC profiles. HCN and N H + have the widest velocity range, but this is due to hyperfinesplitting rather than actual differences in distribution. The CO velocity feature centred at −
70 km s − comes from adifferent cloud along line of sight, as discussed in BWC2006.From the CO data, BWC2006 noted a linear velocitygradient of ∼ − pc − across the GMC (1 arcminis 1 pc for this GMC at distance of 3.6 kpc), which is fivetimes larger than the velocity gradient due to Galacticrotation ( ∼ .
04 km s − pc − for an angular separationof ∼
65 arcmin at b = 333 ◦ , see BWC2006). To investigatewhether molecules other than CO are involved in thisbulk gas motion, we plotted the position-velocity (pv) mapof CS, HCO + , HNC and C H across the GMC as shownin Figure 5. The pv slice is positioned such that it cutsthrough the major emission ridge of the GMC, as shownin the top panel inset. The centre position (zero angularoffset) is roughly at α J = 16 h m , δ J = − d m near IRAS16172 − ∼ − . Similar to CO,the velocity gradient is also noticeable in CS, HCO + , HNCand C H, ∼
10 km s − over roughly 50 arcmin ( ∼ − pc − ), comparable to that of the CO. The HCO + pv map shows similar overall structure to CS in general,however the centre velocity of the strong emission clump at −
30 arcmin (G333.6 − − − for HCO + , compared to −
48 km s − for CS, withHCO + emission being more extended towards the negativevelocities ( ∼ −
60 km s − ) than CS. Where the CS emissionclump peaks at −
48 km s − , the HCO + emission clump isless intense and appears to host two components. In theHNC emission, the structure of this feature is consistentwith CS. In the 3-D data cube, the HCO + spectrumshows a deep self-absorption feature at this position, whencompared with the optically thin H CO + , explaining theabsence of HCO + emission. In general, the HNC and C Hpv maps show a similar velocity structure to CS, exceptthere is no detectable C H at ∼
28 arcmin at −
58 km s − .In the next two sections, we will present two differentapproaches in characterising the distribution of the emis-sion in the GMC, clump finding with gaussclumps andprincipal component analysis. One of the aims of this multi-molecular line mapping isto examine how different molecules correlate with eachother in the GMC. To characterise the distribution ofmolecules, we have used the Gaussian clump decompositionalgorithm gaussclumps (Stutzki & Guesten 1990; Krameret al. 1998) to decompose the observed three-dimensionaldata cubes into individual ‘clumps’ of emission. There hasbeen some discussion in recent years about the best typeof automatic clump decomposition algorithm to use ingiant molecular clouds (see Sheth et al. 2008 for a gooddiscussion on this subject). gaussclumps and clumpfind (Williams, de Geus & Blitz 1994) are algorithms whichwork by identifying peak pixels within a data cube orimage, and then identifying other nearby pixels whichmay be part of the same clump. clumpfind stops addingpixels to a clump when it reaches pixels below a userspecified contour level (see BWC2006 for a more lengthydiscussion of this algorithm). gaussclumps decomposesthe data cube by iteratively fitting the peak position as aGaussian distribution, then subtracting the fitted clumpsand fitting the residual map. Both these methods havethe disadvantage that they tend to break the emissioninto many small clumps close to the resolution element ofthe data set, particularly with a low signal-to-noise dataset. Rosolowsky & Leroy (2006) have described a differentmethod, implemented by the algorithm cprops , whichuses a combination of moments and principal componentanalysis to identify clumps in data cubes, which they findis more robust against the effects of resolution and noise.However, of the three methods, we find gaussclumps provided the best set of physically plausible clumps for themolecular transitions presented in this paper, which tracedense confined gas, with cprops finding very few clumpsin each data cube. This is consistent with the finding ofRosolowsky & Leroy (2006) that gaussclumps is the mosteffective algorithm for separating tight blends of clouds.Consequently, we have chosen to use the gaussclumps ' RAS, MNRAS , 1–25 ulti-molecular lines mapping of the GMC associated with RCW 106 Figure 3.
Spatially averaged emission profiles of the whole GMC. The temperatures are in terms of T ∗ A . Note that the lower signal-to-noise ratio of the N H + and C H spectra is due to these two molecules being less extended spatially; the 1 σ level is comparable to othermore extended molecules. algorithm, but have also undertaken a careful check ofthe clumps produced to confirm that they are physicallyplausible.The molecules selected for such clump finding areCS, HCO + , HNC and C H. The selection is based onCS and HCO + emission not having hyperfine splitting,HNC hyperfine splitting not being resolvable, and the C Hhyperfine component separation being so wide that onlyone component falls into the velocity range. The datacubes were smoothed with a hanning window width offive channels and binning window of two channels, givinga velocity resolution of ∼ . − , to improve the signal-to-noise ratio. The rms levels after binning are 0.070,0.065, 0.06 and 0.06 K (in antenna temperature, T ∗ A ) per0.3 km s − for CS, HCO + , HNC and C H respectively. Wehave set the intensity threshold at the 5 σ level, thereforeclumps with peak temperature below 5 σ are discardedfor CS, HCO + and HNC. For C H, due to its compara-tively weak emission, we only discarded clumps with peaktemperature below the 3 σ level. For all molecules thespectral line emission towards the clumps was examined toconfirm that the emission is real. We also excluded clumpsthat have angular sizes smaller than 1.5 beam-widths toreduce the number of false detections. Clumps that are ' RAS, MNRAS000
Spatially averaged emission profiles of the whole GMC. The temperatures are in terms of T ∗ A . Note that the lower signal-to-noise ratio of the N H + and C H spectra is due to these two molecules being less extended spatially; the 1 σ level is comparable to othermore extended molecules. algorithm, but have also undertaken a careful check ofthe clumps produced to confirm that they are physicallyplausible.The molecules selected for such clump finding areCS, HCO + , HNC and C H. The selection is based onCS and HCO + emission not having hyperfine splitting,HNC hyperfine splitting not being resolvable, and the C Hhyperfine component separation being so wide that onlyone component falls into the velocity range. The datacubes were smoothed with a hanning window width offive channels and binning window of two channels, givinga velocity resolution of ∼ . − , to improve the signal-to-noise ratio. The rms levels after binning are 0.070,0.065, 0.06 and 0.06 K (in antenna temperature, T ∗ A ) per0.3 km s − for CS, HCO + , HNC and C H respectively. Wehave set the intensity threshold at the 5 σ level, thereforeclumps with peak temperature below 5 σ are discardedfor CS, HCO + and HNC. For C H, due to its compara-tively weak emission, we only discarded clumps with peaktemperature below the 3 σ level. For all molecules thespectral line emission towards the clumps was examined toconfirm that the emission is real. We also excluded clumpsthat have angular sizes smaller than 1.5 beam-widths toreduce the number of false detections. Clumps that are ' RAS, MNRAS000 , 1–25
N. Lo et al.
Figure 4.
Sample spectra towards two of the IRAS sources (IRAS16172 − − − T ∗ A scale. found on the edges of the maps are also discarded, due tothe lower sensitivity as the map edges contain one pass only. gaussclumps finds 129, 186, 128 and 78 clumps in theCS, HCO + , HNC and C H data cubes, respectively. gaussclumps reports centre positions and velocities, linewidths, angular sizes, peak temperatures and internalvelocity gradients of the identified clumps. We have listed ' RAS, MNRAS , 1–25 ulti-molecular lines mapping of the GMC associated with RCW 106 Figure 5.
Position-velocity diagrams of CS, HCO + , HNC and C H centred at approximately α J = 16 h m , δ J = − d m .The inset in the top panel shows the position-velocity cut presented in the diagrams. The contour levels all start at the 3 σ level, 0.3 K,increasing to 3.0 K with increments of 0.5 K. The temperatures are in terms of T ∗ A . Positive angular offset corresponds to the southernregion of the GMC. The arrow points toward the H ii region G333.6 − ' RAS, MNRAS000
Position-velocity diagrams of CS, HCO + , HNC and C H centred at approximately α J = 16 h m , δ J = − d m .The inset in the top panel shows the position-velocity cut presented in the diagrams. The contour levels all start at the 3 σ level, 0.3 K,increasing to 3.0 K with increments of 0.5 K. The temperatures are in terms of T ∗ A . Positive angular offset corresponds to the southernregion of the GMC. The arrow points toward the H ii region G333.6 − ' RAS, MNRAS000 , 1–25 N. Lo et al.
Table 3.
The 20 brightest CS clumps selected according to decreasing peak temperature with line intensities above the 5 σ level of theHanning smoothed data cube. Also in Tables 4, 5 and 6: The listed properties are centre position (RA and Dec), centre velocity ( v inkm s − ), full width to half maximum (FWHM) line width (∆ V in km s − ), FWHM of the two principal axes ( D x and D y in arcmin),peak temperature ( T b = T ∗ A /η ν in K), internal velocity gradient across the clump ( dv/dr in km s − arcmin − ), calculated luminosity ( L in K km s − pc ), L = ( d [pc]) ( π × ) ( r x r y ) R T b δv (see text for details) and associated clumps in other molecules. See the onlineversion for a complete list of 129 clumps. v ∆ V D x D y Peak T b dv/dr L Associated clumps in other moleculesCS01 16:21:02.0 -50:35:04.9 -52.2 2.8 1.7 1.4 7.6 0.4 48.7 HCOp02, HNC02, C2H01CS02 16:21:01.3 -50:35:22.9 -49.5 2.5 1.6 1.6 5.2 0.1 33.4 HCOp02, HNC02, HNC05, C2H01CS03 16:22:07.8 -50:06:22.9 -47.8 4.1 2.0 1.3 4.9 0.4 49.1 CS14, HCOp03, HCOp09, HNC01,HNC10CS04 16:21:02.0 -50:34:58.9 -55.3 3.1 1.0 1.2 4.4 0.1 14.8 HCOp05, C2H05CS05 16:20:10.2 -50:53:16.9 -57.4 3.4 1.5 1.6 3.8 0.8 29.8 CS09, CS15, HCOp18, HNC11,HNC17, C2H06CS06 16:21:27.1 -50:24:52.9 -50.4 2.6 1.9 2.4 3.6 0.2 42.2 HCOp03, HNC03, C2H08CS07 16:21:30.3 -50:26:46.9 -52.2 3.2 1.4 1.7 3.6 0.2 26.7 CS16, HCOp16, HCOp20, HNC12,C2H02CS08 16:21:13.3 -50:39:46.9 -55.6 2.0 1.3 1.7 2.9 0.5 12.4 CS17, HCOp07, HNC07, C2H07CS09 16:20:11.5 -50:53:22.9 -54.2 3.2 1.1 0.9 2.6 0.5 8.1 CS05, HCOp18, HNC17CS10 16:20:26.0 -50:41:22.9 -56.0 2.4 1.2 1.4 2.6 0.3 9.9 HCOp14, HNC16, C2H11CS11 16:20:46.9 -50:38:46.9 -53.3 2.6 1.4 1.2 2.5 0.3 10.4 HNC18, C2H04CS12 16:21:37.8 -50:24:52.9 -50.8 3.0 1.1 2.4 2.0 0.5 16.1 C2H15CS13 16:19:37.1 -51:03:28.9 -50.6 3.7 1.7 1.2 2.0 0.7 14.6 C2H12CS14 16:22:07.8 -50:06:04.9 -45.3 2.0 1.7 1.4 2.0 0.3 8.5 CS03, HCOp04, HNC01, C2H03CS15 16:20:09.6 -50:53:16.9 -60.1 2.4 1.4 1.4 2.0 0.6 8.6 CS05, HNC11CS16 16:21:30.3 -50:26:52.9 -54.8 2.5 1.2 1.6 1.9 0.6 9.6 CS07, HCOp16, HCOp20, HNC12,C2H02CS17 16:21:16.5 -50:39:40.9 -57.6 2.0 1.2 1.6 1.9 0.1 7.3 CS08, HCOp07, HNC07, C2H07CS18 16:20:38.0 -50:41:34.9 -54.8 2.1 0.9 2.0 1.9 0.3 6.9 − CS19 16:21:18.3 -50:30:34.9 -52.2 2.0 0.9 2.2 1.9 0.4 7.1 C2H13CS20 16:21:13.3 -50:33:40.9 -51.5 2.1 1.1 2.0 1.8 0.4 8.3 HNC15 the parameters of the brightest 20 clumps of CS, HCO + ,HNC and C H emission according to their peak intensities,along with the derived luminosity in Table 3, 4, 5 and 6,respectively. A complete list of the clumps is available inthe online version. The temperatures listed are correctedfor the beam efficiency, T b = T ∗ A /η ν , where η ν is theextended beam efficiency. For the twenty brightest clumpsof each molecular species, we also listed their associatedclumps in other molecules in the last column of thetables. The association criteria are: the clumps coincidewithin one beam and the line widths overlap. We havealso plotted the clump positions on the correspondingintegrated intensity map in Figures 6 and 7, with ellipsesof size proportional to the angular sizes of the clumps. Asummary of the clump properties is listed in Table 7 forcomparison. The centroid velocities and line widths of theclumps among the four molecules are similar. The angularsizes of the clumps are also comparable among the fourmolecules, but note that the lower limit of angular size isjust above the threshold value (1.5 beam-width) we have set. Among the four molecules decomposed by gaussclumps ,CS and HCO + have lines from rare isotopes, so-called ‘iso-topologues’ (C S and H CO + ) in our data set. Therefore itis possible to estimate optical depths for these molecules. As-suming identical excitation temperatures and optically thinemission from the rare isotopologues, the opacity τ relates to Table 7.
A summary of the clump properties for the CS, HCO + ,HNC and C H data cubes, derived from gaussclumps (see Sec-tion 4). CS HCO + HNC C HNumber of clumps 129 186 128 78Centre velocity (km s − ) lowest − . − . − . − . − . − . − . − . − . − . − . − . σ (arcmin) minimum 0.9 0.8 0.8 0.9maximum 2.3 2.0 2.3 1.8mean 1.4 1.4 1.3 1.3 σ b (K) minimum 0.7 0.7 0.7 0.5maximum 7.6 5.4 4.4 1.8mean 1.3 1.6 1.8 0.7 σ (km s − ) minimum 0.4 0.5 0.4 0.3maximum 4.1 3.8 3.7 3.2mean 1.7 1.1 1.1 1.5 σ (K km s − pc ) minimum 0.05 0.04 0.02 0.03maximum 5.2 5.0 4.0 1.8mean 0.6 0.4 0.5 0.2 σ ' RAS, MNRAS , 1–25 ulti-molecular lines mapping of the GMC associated with RCW 106 Figure 6.
The integrated emission maps (contours and grey scale) of CS and HCO + (for the contour levels, see Figure 1) overlaid withthe corresponding gaussclumps clump fits of the 3-D data cube. The ellipses show the orientation and size of the clumps. Refer to Tables3 and 4 for details of the clumps identified. ' RAS, MNRAS000
The integrated emission maps (contours and grey scale) of CS and HCO + (for the contour levels, see Figure 1) overlaid withthe corresponding gaussclumps clump fits of the 3-D data cube. The ellipses show the orientation and size of the clumps. Refer to Tables3 and 4 for details of the clumps identified. ' RAS, MNRAS000 , 1–25 N. Lo et al.
Figure 7.
The integrated emission maps (contours and grey scale) of HNC and C H (for the contour levels, see Figure 2) overlaid withthe corresponding gaussclumps clump fits of the 3-D data cube. The ellipses show the orientation and size of the clumps. Refer to Tables5 and 6 for details of the clumps identified. ' RAS, MNRAS , 1–25 ulti-molecular lines mapping of the GMC associated with RCW 106 Table 4.
Same as Table 3 but for the 20 brightest HCO + clumps with peak temperatures above 5 σ level of the Hanning smoothed datacube. The clumps are listed in descending peak T b order. See the online version for a complete list of 186 clumps. v ∆ V D x D y Peak T b dv/dr L Associated clumps in other moleculesHCOp01 16:22:08.5 -50:06:28.9 -49.1 2.4 1.7 1.1 5.4 0.6 23.8 CS03, HNC01, HNC10, C2H03HCOp02 16:21:03.2 -50:35:28.9 -51.4 3.8 1.4 1.8 4.9 0.6 47.3 CS01, CS02, HNC02, C2H01HCOp03 16:21:25.8 -50:24:40.9 -50.1 2.8 1.4 1.5 4.4 0.5 26.1 CS06, HNC03, C2H08HCOp04 16:22:09.7 -50:06:22.9 -46.7 1.7 1.9 1.2 3.0 0.2 11.6 CS03, CS14, HNC01, C2H03HCOp05 16:21:02.6 -50:35:10.9 -55.5 3.2 0.9 1.2 2.9 0.1 9.8 CS04, C2H05HCOp06 16:20:05.8 -50:57:10.9 -56.0 2.0 1.1 1.4 2.9 0.6 8.5 HNC13, C2H20HCOp07 16:21:14.6 -50:39:52.9 -56.0 3.1 1.5 1.9 2.9 0.8 23.6 CS08, CS17, HNC07, C2H07HCOp08 16:21:40.3 -50:23:28.9 -50.8 1.9 1.4 2.1 2.6 0.3 13.8 HNC08HCOp09 16:22:11.0 -50:06:28.9 -51.5 2.8 1.6 1.2 2.5 0.3 13.6 CS03, HCOp01, HNC10HCOp10 16:20:38.1 -50:44:16.9 -53.1 2.4 1.2 1.4 2.3 0.4 8.4 HNC04, C2H10HCOp11 16:20:44.4 -50:43:28.9 -60.5 1.7 1.6 2.9 2.1 0.2 15.1 HNC14HCOp12 16:21:36.7 -50:41:10.9 -59.8 2.5 1.1 1.4 2.1 0.5 8.0 HNC09HCOp13 16:20:39.3 -50:41:34.9 -54.6 2.0 1.0 1.4 2.0 0.6 5.1 CS18, C2H16HCOp14 16:20:27.3 -50:41:16.9 -55.3 2.3 1.3 1.8 1.9 0.5 10.1 CS10, HNC16, C2H11HCOp15 16:21:43.5 -50:28:04.9 -50.1 2.0 1.5 2.5 1.8 0.2 12.8 HNC20, C2H17HCOp16 16:21:32.2 -50:27:04.9 -52.4 2.4 1.1 1.4 1.8 0.3 6.5 CS07, CS16, HNC12, C2H02HCOp17 16:20:53.1 -50:43:58.9 -54.0 2.2 1.2 2.4 1.8 0.6 10.6 HNC19HCOp18 16:20:11.5 -50:53:28.9 -54.4 2.6 1.2 1.1 1.8 0.4 5.7 CS05, CS09, HNC17HCOp19 16:19:48.6 -51:02:28.9 -58.9 2.0 1.2 1.5 1.8 0.1 5.6 − HCOp20 16:21:32.1 -50:26:22.9 -49.7 1.9 1.2 2.6 1.7 0.3 10.3 CS07, CS16, HNC12
Table 5.
Same as Table 3 but for the 20 brightest HNC clumps with peak temperature above 5 σ level of the Hanning smoothed datacube. The clumps are listed in descending T b order. See the online version for a complete list of 128 clumps. v ∆ V D x D y Peak T b dv/dr L Associated clumps in other moleculesHNC01 16:22:09.1 -50:06:22.9 -47.1 3.7 1.1 1.7 4.4 0.2 28.1 CS03, CS14, HCOp01, HCOp04,C2H03HNC02 16:21:02.6 -50:35:22.9 -51.7 3.0 1.6 2.0 4.2 0.4 35.4 CS01, CS02, HCOp02, HNC05,C2H01HNC03 16:21:25.8 -50:24:46.9 -50.4 2.9 1.3 1.6 3.4 0.5 19.8 CS06, HCOp03, C2H08HNC04 16:20:38.7 -50:44:22.9 -52.7 2.4 1.2 1.4 2.9 0.5 10.9 HCOp10, C2H10HNC05 16:21:01.3 -50:35:34.9 -49.2 2.1 1.4 1.7 2.8 0.2 13.2 CS02, HCOp02, HNC02, C2H01,C2H18HNC06 16:21:02.6 -50:35:10.9 -54.6 2.6 0.9 1.2 2.8 0.1 7.4 CS04, HCOp02, HCOp05, HNC02,C2H05HNC07 16:21:15.2 -50:39:46.9 -56.1 3.1 1.4 1.6 2.4 1.1 15.2 CS08, CS17, HCOp07, C2H07HNC08 16:21:39.1 -50:23:52.9 -50.8 2.3 1.4 2.8 2.3 0.1 19.1 CS12, HCOp08, C2H15HNC09 16:21:36.7 -50:41:04.9 -59.2 3.1 1.2 1.4 2.2 1.1 10.6 HCOp12HNC10 16:22:09.7 -50:06:34.9 -49.7 2.8 1.8 1.3 2.2 0.3 12.9 CS03, HCOp01, HCOp09, C2H03HNC11 16:20:08.9 -50:53:22.9 -58.1 3.3 1.9 2.1 2.1 0.6 25.7 CS05, CS15, C2H06HNC12 16:21:30.3 -50:26:52.9 -52.1 3.0 1.5 1.5 2.1 0.3 12.6 CS07, CS16, HCOp16, HCOp20,C2H02HNC13 16:20:05.8 -50:57:04.9 -56.1 2.0 1.1 1.0 2.0 0.6 4.3 HCOp06, C2H20HNC14 16:20:44.4 -50:43:10.9 -60.5 1.7 1.4 2.5 1.9 0.1 10.7 HCOp11HNC15 16:21:13.9 -50:33:58.9 -51.5 2.0 1.1 1.8 1.8 0.4 7.1 CS20HNC16 16:20:27.3 -50:41:28.9 -56.1 2.3 1.2 1.7 1.8 0.2 7.9 CS10, HCOp14, C2H11HNC17 16:20:11.5 -50:53:28.9 -54.6 2.9 1.1 1.1 1.8 0.5 5.6 CS05, CS09, HCOp18HNC18 16:20:47.5 -50:38:46.9 -53.1 2.0 1.2 1.7 1.8 0.0 6.8 CS11, C2H04HNC19 16:20:51.7 -50:43:46.9 -54.2 2.0 1.3 2.5 1.7 0.4 10.0 HCOp17HNC20 16:21:43.5 -50:28:04.9 -50.3 1.9 1.2 2.0 1.6 0.1 6.9 HCOp15, C2H17 the brightness (beam efficiency corrected) of optically thickand thin emission lines as follows: T b (thick) T b (thin) = 1 − e − τ thick − e − τ thin = 1 − e − τ thick − e − τ thick /X , (1)where X is the abundance ratio, assuming [CS/C S] = 22.5(Chin et al. 1996) and [HCO + /H CO + ] = 50, based on[CO/ CO] from Wilson & Rood (1994).After solving for τ , we next calculate the excitation tem- perature T ex from the optically thick emission lines, CS andHCO + . From the equation of radiative transfer, T ex can bederived from T b = f [ J ( T ex ) − J ( T bg )][1 − e − τ ] . (2) f is the beam filling factor which we assume to be 1, and J ( T ) = T / ( e T /T − τ and T ex are known, thecolumn density of the optically thick line can be derived ' RAS, MNRAS000
Same as Table 3 but for the 20 brightest HNC clumps with peak temperature above 5 σ level of the Hanning smoothed datacube. The clumps are listed in descending T b order. See the online version for a complete list of 128 clumps. v ∆ V D x D y Peak T b dv/dr L Associated clumps in other moleculesHNC01 16:22:09.1 -50:06:22.9 -47.1 3.7 1.1 1.7 4.4 0.2 28.1 CS03, CS14, HCOp01, HCOp04,C2H03HNC02 16:21:02.6 -50:35:22.9 -51.7 3.0 1.6 2.0 4.2 0.4 35.4 CS01, CS02, HCOp02, HNC05,C2H01HNC03 16:21:25.8 -50:24:46.9 -50.4 2.9 1.3 1.6 3.4 0.5 19.8 CS06, HCOp03, C2H08HNC04 16:20:38.7 -50:44:22.9 -52.7 2.4 1.2 1.4 2.9 0.5 10.9 HCOp10, C2H10HNC05 16:21:01.3 -50:35:34.9 -49.2 2.1 1.4 1.7 2.8 0.2 13.2 CS02, HCOp02, HNC02, C2H01,C2H18HNC06 16:21:02.6 -50:35:10.9 -54.6 2.6 0.9 1.2 2.8 0.1 7.4 CS04, HCOp02, HCOp05, HNC02,C2H05HNC07 16:21:15.2 -50:39:46.9 -56.1 3.1 1.4 1.6 2.4 1.1 15.2 CS08, CS17, HCOp07, C2H07HNC08 16:21:39.1 -50:23:52.9 -50.8 2.3 1.4 2.8 2.3 0.1 19.1 CS12, HCOp08, C2H15HNC09 16:21:36.7 -50:41:04.9 -59.2 3.1 1.2 1.4 2.2 1.1 10.6 HCOp12HNC10 16:22:09.7 -50:06:34.9 -49.7 2.8 1.8 1.3 2.2 0.3 12.9 CS03, HCOp01, HCOp09, C2H03HNC11 16:20:08.9 -50:53:22.9 -58.1 3.3 1.9 2.1 2.1 0.6 25.7 CS05, CS15, C2H06HNC12 16:21:30.3 -50:26:52.9 -52.1 3.0 1.5 1.5 2.1 0.3 12.6 CS07, CS16, HCOp16, HCOp20,C2H02HNC13 16:20:05.8 -50:57:04.9 -56.1 2.0 1.1 1.0 2.0 0.6 4.3 HCOp06, C2H20HNC14 16:20:44.4 -50:43:10.9 -60.5 1.7 1.4 2.5 1.9 0.1 10.7 HCOp11HNC15 16:21:13.9 -50:33:58.9 -51.5 2.0 1.1 1.8 1.8 0.4 7.1 CS20HNC16 16:20:27.3 -50:41:28.9 -56.1 2.3 1.2 1.7 1.8 0.2 7.9 CS10, HCOp14, C2H11HNC17 16:20:11.5 -50:53:28.9 -54.6 2.9 1.1 1.1 1.8 0.5 5.6 CS05, CS09, HCOp18HNC18 16:20:47.5 -50:38:46.9 -53.1 2.0 1.2 1.7 1.8 0.0 6.8 CS11, C2H04HNC19 16:20:51.7 -50:43:46.9 -54.2 2.0 1.3 2.5 1.7 0.4 10.0 HCOp17HNC20 16:21:43.5 -50:28:04.9 -50.3 1.9 1.2 2.0 1.6 0.1 6.9 HCOp15, C2H17 the brightness (beam efficiency corrected) of optically thickand thin emission lines as follows: T b (thick) T b (thin) = 1 − e − τ thick − e − τ thin = 1 − e − τ thick − e − τ thick /X , (1)where X is the abundance ratio, assuming [CS/C S] = 22.5(Chin et al. 1996) and [HCO + /H CO + ] = 50, based on[CO/ CO] from Wilson & Rood (1994).After solving for τ , we next calculate the excitation tem- perature T ex from the optically thick emission lines, CS andHCO + . From the equation of radiative transfer, T ex can bederived from T b = f [ J ( T ex ) − J ( T bg )][1 − e − τ ] . (2) f is the beam filling factor which we assume to be 1, and J ( T ) = T / ( e T /T − τ and T ex are known, thecolumn density of the optically thick line can be derived ' RAS, MNRAS000 , 1–25 N. Lo et al.
Table 6.
Same as Table 3 but for the 20 brightest C H clumps with peak temperature above 3 σ level of the Hanning smoothed datacube. The clumps are listed in descending peak T b order. See the online version for a complete list of 78 clumps. The C H clumpspresented here refer to the main hyperfine component only, hence the luminosity is not calculated. v ∆ V D x D y Peak T b dv/dr Associated clumps in other moleculesC2H01 16:21:02.6 -50:35:34.9 -50.8 2.8 2.2 1.6 1.8 0.5 CS01, CS02, HCOp02, HNC02,HNC05, C2H18C2H02 16:21:30.3 -50:26:52.9 -52.6 2.4 1.5 1.8 1.6 0.2 CS07, CS16, HCOp16, HNC12C2H03 16:22:07.8 -50:06:34.9 -47.8 3.2 1.1 1.8 1.6 0.2 CS03, CS14, HCOp01, HCOp04,HNC01, HNC10C2H04 16:20:49.4 -50:38:52.9 -53.7 2.1 1.2 1.2 1.4 0.3 CS11, HNC18C2H05 16:21:02.6 -50:35:10.9 -54.0 2.1 0.9 1.4 1.3 0.3 CS04, HCOp02, HCOp05, HNC02,HNC06C2H06 16:20:10.8 -50:53:04.9 -58.3 2.7 1.6 1.2 1.1 0.2 CS05, HNC11C2H07 16:21:16.5 -50:39:40.9 -56.7 2.3 1.3 1.7 1.1 0.9 CS08, CS17, HCOp07, HNC07C2H08 16:21:27.1 -50:24:40.9 -50.4 1.8 1.6 1.7 1.1 0.5 CS06, HCOp03, HNC03C2H09 16:19:53.0 -51:01:34.9 -47.6 2.6 1.0 1.5 1.0 0.4 − C2H10 16:20:38.7 -50:44:04.9 -52.6 2.0 1.1 1.3 1.0 0.7 HCOp10, HNC04C2H11 16:20:26.7 -50:41:16.9 -55.5 2.4 1.1 1.8 1.0 0.5 CS10, HCOp14, HNC16C2H12 16:19:35.8 -51:03:34.9 -50.1 2.5 1.2 1.7 1.0 0.2 CS13C2H13 16:21:17.7 -50:30:34.9 -52.4 1.8 1.0 2.2 0.9 0.3 CS19C2H14 16:22:04.1 -50:12:16.9 -49.7 2.0 1.5 2.3 0.9 0.4 − C2H15 16:21:36.5 -50:25:04.9 -50.8 2.0 1.0 2.6 0.9 0.2 CS12, HNC08C2H16 16:20:41.2 -50:42:04.9 -52.9 1.8 1.0 1.8 0.8 0.1 CS18, HCOp13C2H17 16:21:42.9 -50:28:10.9 -50.5 1.4 1.3 2.1 0.8 0.3 HCOp15, HNC20C2H18 16:21:02.6 -50:36:04.9 -48.8 1.3 0.9 1.1 0.8 0.3 CS05, CS09, HCOp02, HNC05C2H19 16:21:20.0 -50:10:04.9 -43.5 1.9 1.1 1.6 0.8 0.1 − C2H20 16:20:05.8 -50:57:16.9 -56.0 1.8 1.2 1.3 0.7 0.6 HCOp06, HNC13 with (cf Purcell et al. 2006) N = 8 kπν hc g u A ul τ − e − τ e − E u /kT ex Q ( T ) Z T b dv, (3)where A ul is the Einstein A coefficient in s − , ν is the tran-sition frequency in Hz, R T b dv is the integrated brightnesstemperature in K km s − , Q ( T ) is the partition function, g u is the upper state degeneracy and E u is the upper stateenergy in m − . At positions, where no C S or H CO + wasdetected, we have assumed also that the main isotopologuewas optically thin and the excitation temperature was setto 3 K. Eleven CS and two HCO + clumps fall into thiscategory. The choice of 3 K for the excitation temperaturewas motivated by the median value of T ex from clumps thatshow emission in the main and the rare species (Figures 10and 11).Shown in Figure 8 and 9 are histogram plots of CSand HCO + clump column density. The solid line shows thecolumn density after correction for optical depth, and thedashed line shows the derived column density distributionby assuming CS and HCO + are optically thin with anexcitation temperature of 20 K. The plots show that aftercorrection, the column density increases by an order ofmagnitude. An examination of the clump spectra confirmsthat clumps with column density below 10 cm − arethose with minimal or no optically thin line detections,while those with column density of 10 cm − or above areoptically thick clumps.An excitation temperature of 3 K is low. One possiblecause could be due to the optically thin lines (C S andH CO + ) being spatially compact, and the assumptionof having the same beam efficiency as the more commonisotopologue may not hold. We thus also calculated theexcitation temperatures of both CS and HCO + with Figure 8.
Histogram of CS clump column density with opticaldepth correction (solid). The dashed line shows the distributionobtained by assuming that CS is optically thin and has an exci-tation temperature of 20 K. extended beam efficiency ( η xb ) for CS and HCO + , andmain beam efficiency ( η mb ) for C S and H CO + . Theexcitation temperatures of CS and HCO + then increasesto ∼ ∼
36 arcsec at 100-GHz). The clump finding algorithmsees these fragmented clumps as one single clump. Thus,the assumption of the beam filling factor ( f ) being unitydoes not hold, despite the fact that all identified clumpsare larger than 1.5 beam widths. Various studies showthat CS clumps are typically fragmented, with a volumefilling factor as low as 0.2 (e.g. Stutzki & Guesten 1990;Juvela 1998). To investigate the effect of beam filling factor ' RAS, MNRAS , 1–25 ulti-molecular lines mapping of the GMC associated with RCW 106 Figure 9.
Histogram of HCO + clump column density with opti-cal depth correction (solid). The dashed line shows the distribu-tion obtained by assuming that HCO + is optically thin and hasan excitation temperature of 20 K. Figure 10.
Histogram of CS clump excitation temperature witha beam filling factor of 0.2 (solid line). The dashed line shows thedistribution assuming a unity beam filling factor.
Figure 11.
Histogram of HCO + clump excitation temperaturewith a beam filling factor of 0.2 (solid line). The dashed line showsthe distribution assuming a unity beam filling factor. Figure 12.
Plot of variances of the principal components. Theinset presents a zoomed version of the plot, showing the secondand higher order components in more detail. on the excitation temperature, we derived the excitationtemperatures of CS and HCO + clumps with various beamfilling factors. We found that with f = 0 .
2, approximately74 per cent of the CS clumps, and 56 per cent of the HCO + clumps have T ex between 6 to 20 K. Histograms of CS andHCO + excitation temperatures are shown in Figure 10 and11 to illustrate the effect of lowering beam filling factor. Themean excitation temperature increases greatly by loweringthe beam filling factor, suggesting the clumps may indeedfragmented and not resolvable with MOPRA beam. Thiswould provide an explanation for the lack of correlation be-tween clump size and line width (see Section 6.1 for details). An alternative way to study the distribution of themolecules is by performing principal component analysis(PCA) on the integrated emission maps of the region. PCAis a multivariate data analysis technique, its purpose beingto reduce the dimensionality of a data set. Mathematicallythe derivation of principal components involves finding theeigenvalues and eigenvectors of a covariance or correlationmatrix (see e.g. Jolliffe 2002; Heyer & Schloerb 1997;Ungerechts et al. 1997, and references therein).The purpose of performing PCA on the integratedemission maps is to characterise differences in moleculardistribution. The G333 cloud consists of star forming sitesat different stages of evolution, from cold dense starlesscores to H ii regions. Hence we expect to see chemicaldifferences across the cloud. The data set covers varietiesof chemical probes. Representative ones are the outflowtracer HCO + , the large scale gas tracer CO, the dense gastracer CS, and the cold dense gas tracer N H + . Interestingaspects are chemical differences and a possible temperaturedependance of HCN to HNC abundance ratio. In thissection, we will discuss the result of a PCA decompositionon the whole G333 molecular cloud; the PCA of individualregions of interest will be discussed in later sections.We have chosen eight molecule lines with high signal-to-noise ratio for this analysis: CO, C O, CS, HCO + ,HCN, HNC, N H + and C H (Figures 1 and 2). Among ' RAS, MNRAS000
2, approximately74 per cent of the CS clumps, and 56 per cent of the HCO + clumps have T ex between 6 to 20 K. Histograms of CS andHCO + excitation temperatures are shown in Figure 10 and11 to illustrate the effect of lowering beam filling factor. Themean excitation temperature increases greatly by loweringthe beam filling factor, suggesting the clumps may indeedfragmented and not resolvable with MOPRA beam. Thiswould provide an explanation for the lack of correlation be-tween clump size and line width (see Section 6.1 for details). An alternative way to study the distribution of themolecules is by performing principal component analysis(PCA) on the integrated emission maps of the region. PCAis a multivariate data analysis technique, its purpose beingto reduce the dimensionality of a data set. Mathematicallythe derivation of principal components involves finding theeigenvalues and eigenvectors of a covariance or correlationmatrix (see e.g. Jolliffe 2002; Heyer & Schloerb 1997;Ungerechts et al. 1997, and references therein).The purpose of performing PCA on the integratedemission maps is to characterise differences in moleculardistribution. The G333 cloud consists of star forming sitesat different stages of evolution, from cold dense starlesscores to H ii regions. Hence we expect to see chemicaldifferences across the cloud. The data set covers varietiesof chemical probes. Representative ones are the outflowtracer HCO + , the large scale gas tracer CO, the dense gastracer CS, and the cold dense gas tracer N H + . Interestingaspects are chemical differences and a possible temperaturedependance of HCN to HNC abundance ratio. In thissection, we will discuss the result of a PCA decompositionon the whole G333 molecular cloud; the PCA of individualregions of interest will be discussed in later sections.We have chosen eight molecule lines with high signal-to-noise ratio for this analysis: CO, C O, CS, HCO + ,HCN, HNC, N H + and C H (Figures 1 and 2). Among ' RAS, MNRAS000 , 1–25 N. Lo et al.
Figure 14.
Constructed images of the first four principal components of the GMC. The contour levels are multiples of 20 per cent of thepeak. The dashed contours represent an anti-correlation, while the solid contours represent a positive correlation in each of the principalcomponents. The grey scale in each of the images shows the first principal component, overlaid with contours of the other principalcomponents for ease of comparison. The lower left circles of each images indicate the beam size. The arrow indicates an example ofanti-correlation between HCO + and N H + as discussed in Section 6.2.2. ' RAS, MNRAS , 1–25 ulti-molecular lines mapping of the GMC associated with RCW 106 Table 8.
The correlation matrix of the input molecular data set for principal component analysis.CS HCO + HNC C O C H HCN N H + 13 COCS 1.00HCO + O 0.77 0.80 0.81 1.00C H 0.60 0.70 0.67 0.61 1.00HCN 0.81 0.94 0.90 0.75 0.70 1.00N H + CO 0.77 0.82 0.81 0.88 0.63 0.79 0.69 1.00
Table 9.
The eigenvectors and eigenvalues of the principal components (PC) derived from the correlation matrix listed in Table 8.Percentage of variance CS HCO + HNC C O C H HCN N H + 13 COPC 1 80.0 0.36 0.38 0.38 0.35 0.30 0.37 0.34 0.35PC 2 6.3 -0.25 0.05 -0.10 -0.17 0.86 0.15 -0.36 -0.06PC 3 4.7 -0.25 -0.11 -0.22 0.58 -0.07 -0.16 -0.38 0.61PC 4 3.8 -0.05 -0.33 -0.17 0.21 0.38 -0.50 0.64 -0.03PC 5 2.1 0.80 -0.26 0.07 -0.05 0.15 -0.34 -0.39 0.04PC 6 1.4 0 0.03 0.13 0.68 0.01 0.02 -0.18 -0.70PC 7 0.8 -0.25 -0.63 0.72 -0.03 0.01 0.13 -0.05 0.09PC 8 0.7 0.22 -0.52 -0.48 0.10 -0.01 0.66 0.11 -0.03 these, C H has the lowest signal-to-noise level and COhas the highest. The data sets were imported into idl (a data visualisation and analysis platform) and each oftheir standard deviations were obtained. We decided toderive the principal components of the correlation matrix,rather than the covariance matrix. This is to avoid theprincipal components being dominated by a single variable(Jolliffe 2002, Chapter 3.3). An alternative approachinvolves normalising the data set first, and then derivingthe principal components from the covariance matrix. Wein fact derived the principal components using these twodifferent approaches and obtained essentially the sameresult.Listed in Table 8 is the correlation matrix of the inputmolecules, which describes how well the molecules correlatewith each other. The correlation matrix shows there arefour pairs of molecules which have correlation coefficientsequal to or above 0.9. These are HCO + and HCN, HCO + and HNC, HCN and HNC, HNC and CS. This is consistentwith the integrated intensity maps shown in Figure 1 and2, where CS, HCO + , HCN and HNC have similar largescale spatial distribution. However CS does show smalldifferences in distribution, as discussed in Section 3.1.Among the eight molecules, C H and N H + have the lowestcorrelation coefficient. After forming the correlation matrixof the data set, we derived the eigenvalues and eigenvectorsof the correlation matrix (Table 9). The eigenvalues indicatethat the first principal component accounts for 80 per centof total variation, while the first three principal componentstogether account for over 90 per cent of total variationin the data set (see Table 9 and Figure 12). Figure 12also shows that the first four principal components containfeatures above the noise level, while the variations containedin higher components are insignificant.From the eigenvectors we constructed images ofprincipal components (PC), by projecting the data setonto each of the eigenvectors. Shown in Figure 13 areplots of eigenvectors of each molecule in the first fourprincipal components. A (negative) positive value indicates the molecule is (anti-)correlated with others. The largerthe value, the stronger the correlation. From the PC1axis of Figure 13a it becomes clear that all of the eightmolecules are positively correlated with each other. Thiscan be visualised in Figure 14a, which is the whole dataset projected onto the first principal component only. Thisalso resembles an ‘ideal’ molecular distribution of the giantmolecular cloud - the average intensity distribution of thespecies - and can be interpreted as the eight molecules beingpositively correlated on large scales. From the eigenvectorplot of the second principal component (PC 2 axis of Figure13a), C H stands out from other molecules; we suspect thisis caused by scanning stripes resulting from the on-the-flymapping procedure. This can be seen in the image of thesecond principal component (Figure 14b); note the verticaland horizontal stripes of solid contours which resemblethe scanning patterns presented in the C H integratedemission map (Figure 2d). In fact, the scanning stripesare present in all data taken simultaneously as listed inTable 1 (the July, 2006 observation season). However, dueto the lower signal-to-noise level of the C H emission, thesescanning patterns become dominant in this map. We haveexplored the possibility of removing scanning artefacts byreconstructing the images without the second principalcomponent. This indeed reduced the level of the artifacts,but it also subtracts a component of real emission from theimage. Hence it provides a qualitative improvement, butthe corrected image should not be used for quantitativestudies.From the eigenvector plot of the third principalcomponent (Figure 13b) we can see C O and CO areanti-correlated with other molecules. Spatially this anti-correlation appears as compact negative contours againsta diffuse positive region in the third principal componentimage (Figure 14c). This could be due to the differencesbetween high and low density tracers discussed in Section6.2.1, with the CO isotopologue distribution being moreextended.The fourth principal component eigenvector plot ' RAS, MNRAS , 1–25 N. Lo et al.
Figure 13.
Plots of eigenvectors of the first three principal com-ponents of the data set. Each of the eigenvectors represent thecomponent of that molecule in the relevant principal component. (Figure 13c) shows an anti-correlation between N H + andHCO + . The anti-correlation also appears clearly in thefourth principal component image (Figure 14d), notablyacross IRAS16172 − + , N H + , and C H. Implications will be discussed in the next section.
In this section we discuss the results from this multi-molecular line survey and the analysis techniques applied.We also focus on particularly interesting regions in theG333 giant molecular cloud.
To examine the correlation among the clump properties ofCS, HCO + and HNC, we have compared luminosities of themolecular line emission, line widths and radii in Figure 15.The C H clumps presented here were obtained analysingthe main hyperfine component only. Hence the entire lineluminosity has not been calculated. The luminosity was cal-culated with L = ( d [ pc ]) ( π × ( r x r y ) Z T d v , (4)where d is the distance to the GMC (3.6 kpc), r x and r y are the radii (arcsecond) of the two principal axes and R T d v (K km s − ) is the sum of emission at the maximumposition. We have discarded clumps with line widths below0.7 km s − , as this is less than twice the effective spectralresolution of 0.3 km s − after hanning smoothing andbinning the data. The slope of the ordinary least square(OLS) bisector (Feigelson & Babu 1992) line of best fit ( γ ),estimates the structural relationship between variables X and Y , without assuming whether Y depends on X or viceversa. Values inside the curly brackets are the slopes ofOLS ( X | Y ), which minimises the residuals in X , and OLS( Y | X ) which minimises the residuals in Y . The number ofclumps ( N ) are shown on the top left corners of correlationplots.Larson (1981) empirically showed, utilising data fromseveral molecular cloud surveys, that the line width ∆ V appears to scale with cloud size r , ∆ V ∝ r γ . The com-mon value quoted for γ is ∼ . ± . CO and C O line emission (WLB2008), we find nocorrelation between line width and radius for CS, HCO + and HNC clumps. The line width-size (note radius is given,not diameter) plots (Figure 15a-c) show a large scatter,hence the fitted slopes are poorly constrained. Note thelarge differences between OLS ( X | Y ) and ( Y | X ) slopes.Other line width-size relation studies (e.g. Schneider &Brooks 2004, and references therein) also found no linewidth-size relation, and suggested its possible dependanceon clump identification procedure.There is a clear correlation between luminosity andradius among CS, HCO + and HNC clumps. The fittedOLS bisector slopes γ (3.8 to 4.6) are slightly steeperthan found for CO and C O ( ∼ . ' RAS, MNRAS , 1–25 ulti-molecular lines mapping of the GMC associated with RCW 106 Figure 15.
Plots of relationship between physical parameters for CS, HCO + and HNC clumps: (a) to (c) radius and line width; (d) to(f) luminosity and radius; (g) to (i) luminosity and line width. γ indicates the slope of the ordinary least square (OLS) bisector fittedsolid line, the values in curly brackets are the slope of the OLS ( X | Y ) and ( Y | X ). N is the number of clumps. Note that clumps withline widths smaller than 0.7 kms − are discarded. implicitly depends on r , one would expect γ ≈
3, assuminga constant column density across the clumps. However, nostatistical significance can be assigned to the value of slope,due to the inherent bias in clump finding algorithms, thattend to find larger clumps in denser regions (Schneider &Brooks 2004).Similarly, luminosity and line width exhibits a strongcorrelation; this is not surprising as luminosity is calculatedfrom the clump sizes and line widths. We do note that the CO and C O clumps from previous work (WLB2008)were found using a different clump-finding algorithm, cprops . In Section 5 we presented the results of principal componentanalysis, finding correlations or anti-correlations betweenmolecular species. In the following sections, we discuss thesein terms of the physical and chemical properties of the GMC.
The eigenvector plot of the third principal component showsthat CO and C O are anti-correlated with the othermolecules (Figure 13b); this can be visualised spatially fromthe projected principal component image as shown in Figure14c. In this third PC image the compact dashed (negative)contours contrast spatially with the diffuse solid (positive)contours. This is due to C O and CO ( J = 1 → CO and C O are present in both low and high densityregions, whereas the other molecules are present in densegas only. Hence in diffuse gas CO and C O are in ‘excess’compared to the high density tracers. H + and HCO + The eigenvector plot for the fourth principal component(Figure 13c) indicates the two ionic species N H + andHCO + are anti-correlated. According to models of denseclouds the major formation pathway of N H + is via ion-molecule reactions between N and H +3 (Nejad, Williams &Charnley 1990). It is mainly destroyed by reactions with CO(Tafalla, Myers, Caselli & Walmsley 2004) and by recombi- ' RAS, MNRAS , 1–25 N. Lo et al. nation with electrons in hot regions (Sternberg & Dalgarno1995). The following chemical equations summarise the men-tioned reactions,N + H +3 −→ N H + + H , (5)N H + + CO −→ HCO + + H , (6)N H + + e −→ N + H . (7)HCO + can be formed by ion-molecule reaction betweenN H + and CO (Equation 6), or between CO and H +3 (Stern-berg & Dalgarno 1995) as shown in Equation 8 below,CO + H +3 −→ HCO + + H . (8)In cold dense cores, CO, the major destroyer of N H + , willbe depleted, so that the N H + abundance increases. Withthe lack of CO molecules, HCO + loses both of its majorpathways (CO reacting with N H + or H +3 ), so the abun-dance of HCO + decreases. On the other hand, in warm re-gions where CO is not depleted, it reacts with N H + andproduces HCO + causing a decrease in N H + and increasein HCO + . CO reactions with H +3 further increase the HCO + abundance.Prominent sites to study this phenomenon are locatedin the vincinity of H ii regions, i.e. in photodissociation orphoton dominated regions (PDRs), where at the cloud sur-face a high HCO + and a low N H + abundance is expected.In contrast, in the inner part of the cloud, where materialis collapsing to form a dense core, CO is depleted causingan increase in N H + and a drop in HCO + abundance. ThusHCO + and N H + should be anti-correlated. − The H ii region associatedwith IRAS16172 − H + and HCO + are anti-correlated. As suggestedby the compact intense N H + emission, this source has acold high density component, but is immersed in a strongPDR according to the bright 8.0- µ m emission. To excludethe effects of large-scale spatial contamination, we appliedPCA to the integrated emission map of this region. Listedin Table 10 are the eigenvalues and eigenvectors for thisanalysis; the eigenvectors of the second, third and fourthprincipal components are also plotted in Figure 16; wherethe anti-correlation of HCO + and N H + is clearly seen.In the eigenvector plot of the second and third principalcomponents (Figure 16a), C H has the strongest positivecorrelation among the molecules. This is due to its spa-tially wide-spread emission towards the south-eastern re-gion, compared to, for example, N H + . Figure 17 shows theGLIMPSE 8.0- µ m image (grey scale) overlaid with contoursof the second principal component of the H ii region associ-ated with IRAS16172 − H, HCO + (thick contours) and N H + (thin contours) contribute sig-nificantly to the variance in the second component and areanti-correlated, as expected for the hypothesis above. Notethat the 8.0- µ m emission (PAHs and warm gas) coincideswell with the thick contours (HCO + ).The second and third principal components of this re-gion also show that CO and C O are correlated with
Figure 16.
The eigenvectors of principal components two,three and four of the gas near the H ii region associated withIRAS16172 − Figure 17.
An image of the second principal component(contours) of the gas near the H ii region associated withIRAS16172 − µ m image (greyscale). The thin contours show regions with negative correlationto this component (mainly N H + ) and the thick contours areregions with positive correlation (mainly HCO + ). ' RAS, MNRAS , 1–25 ulti-molecular lines mapping of the GMC associated with RCW 106 Table 10.
The eigenvectors and eigenvalues of the principal components (PC) of the gas near the H ii region associated withIRAS16172 − + HNC C O C H HCN N H + 13 COPC 1 84.8 0.36 0.37 0.37 0.35 0.31 0.36 0.35 0.35PC 2 6.7 0.07 0.27 0.02 -0.39 0.66 0.20 -0.34 -0.41PC 3 3.4 -0.42 -0.23 -0.28 0.37 0.64 -0.29 0.13 0.20PC 4 2.3 0.64 -0.30 0.05 -0.23 0.19 -0.53 0.35 -0.11PC 5 1.4 0.35 -0.07 0.05 0.20 0.04 -0.21 -0.78 0.43PC 6 0.6 0.17 -0.03 0.03 0.70 -0.06 -0.02 -0.11 -0.68PC 7 0.4 -0.21 -0.63 0.72 -0.06 0.10 0.18 -0.08 -0.04PC 8 0.3 -0.31 0.50 0.51 0 -0.05 -0.62 -0.02 -0.06
Table 11.
The eigenvectors and eigenvalues of the principal components (PC) of the gas near the H ii region associated withIRAS16177 − + HNC C O C H HCN N H + 13 COPC 1 86.0 0.36 0.36 0.37 0.35 0.31 0.36 0.35 0.36PC 2 5.4 -0.01 -0.08 0.08 -0.46 0.79 0.18 -0.09 -0.34PC 3 3.0 0.04 -0.36 -0.19 0.45 0.40 -0.16 -0.53 0.41PC 4 2.8 0.29 -0.43 -0.04 0.15 0.17 -0.54 0.61 -0.17PC 5 1.1 -0.81 0.25 -0.14 0.21 0.27 -0.16 0.33 0.11PC 6 0.7 -0.30 -0.70 0.24 -0.02 -0.14 0.52 0.22 0.16PC 7 0.6 -0.01 0 -0.12 0.62 -0.01 0.31 -0.07 -0.71PC 8 0.4 -0.18 -0.01 0.85 0.12 -0.04 -0.36 -0.26 -0.17
Figure 18.
The eigenvectors of principal components two,three and four of the gas near the H ii region associated withIRAS16177 − N H + (Figure 16a), which contradicts the idea suggestingthat CO is the main destroyer of N H + . However, we be-lieve this can be explained by considering the different den-sity conditions that these two species are tracing. CO is alow density tracer having a critical density of ∼ cm − ,while N H + is a high density tracer with a critical den-sity of ∼ cm − . Therefore CO is tracing both low andhigh density gas along the line of sight. Since maps of inte-grated emission were used for the PCA, the CO and C Omay arise from a molecular envelope almost devoid of N H + ,while the bulk of the N H + emission is arising from the core. − Another example is themolecular emission associated with IRAS16177 − Figure 19.
An image of the third principal component (contours)of the gas near the H ii region associated with IRAS16177 − µ m image (grey scale). The thincontours showing regions with negative correlations (mainlyN H + ) and the thick contours are regions with positive corre-lations (mainly CO and C O). shown in Figures 19 and 20. The eigenvalues and eigen-vectors of the principal components are listed in Table 11,while the eigenvectors of the third and fourth principalcomponents are plotted in Figure 18. Similar to the othersource IRAS16172 − H + and HCO + in the fourth principalcomponent (Figure 20). On the other hand, shown in thethird principal component (Figure 19) is an anti-correlationof N H + and CO isotopologues.The PCA also indicates that HCN and HCO + arecorrelated, as shown in the eigenvector plots of the entire ' RAS, MNRAS000
An image of the third principal component (contours)of the gas near the H ii region associated with IRAS16177 − µ m image (grey scale). The thincontours showing regions with negative correlations (mainlyN H + ) and the thick contours are regions with positive corre-lations (mainly CO and C O). shown in Figures 19 and 20. The eigenvalues and eigen-vectors of the principal components are listed in Table 11,while the eigenvectors of the third and fourth principalcomponents are plotted in Figure 18. Similar to the othersource IRAS16172 − H + and HCO + in the fourth principalcomponent (Figure 20). On the other hand, shown in thethird principal component (Figure 19) is an anti-correlationof N H + and CO isotopologues.The PCA also indicates that HCN and HCO + arecorrelated, as shown in the eigenvector plots of the entire ' RAS, MNRAS000 , 1–25 N. Lo et al.
Figure 20.
Similar to Figure 19 but the fourth principal com-ponent (contours) overlaid on the GLIMPSE 8.0- µ m image (greyscale). The thin contours show regions with negative correlationto this component (mainly HCO + ) and the thick contours areregions with positive correlation (mainly N H + ). region (Figure 13). This correlation has also been suggestedby other studies, such as by Turner & Thaddeus (1977)for Orion-KL. IRAS16172 − − ii region, containcold molecular gas of high density (bright compact N H + emission), and have a low photon exposure. As expected bychemical models, both sources clearly show that N H + andHCO + are anti-correlated. Another note worthy feature in G333 is a ring of mid-infrared emission as shown in Figure 21, the northern por-tion of which is aligned with an arc of N H + emission (seethe top arrow on Figure 2). This ring feature is visible inthe GLIMPSE 5.8 and 8.0- µ m (Figure 21 top right andbottom left panels) images, but is marginally detectablein 3.6- (Figure 21 top left panel) and 4.5- µ m images. TheGLIMPSE 5.8 and 8.0- µ m channels of the Spitzer
IRAC in-strument are dominated by PAH emission excited by nearbyultraviolet sources (Reach et al. 2006). Lying inside theemission ring but offset towards the edge, is bright MIPS-GAL (Carey et al. 2006) 24- µ m emission (Figure 21 bottomright panel), spatially coincident with the MSX point sourceG333.5114 − IRAS source IRAS16182 − µ m furthersuggests that the emission originates from PAHs, and thatthere is a lack of thermal dust emission.Comparing the MIPSGAL 24- µ m emission with the843-MHz radio continuum (Figure 21d contours) from theMolonglo Galactic Plane Survey (MGPS, Green et al. 1999),it is clear that the radio continuum is associated with the24- µ m emission. A search of the SIMBAD astronomical database and VizieR catalogue service did not lead to theidentification of any supernova or X-ray source, suggestingthere is an H ii region inside the 8- µ m emission ring. Shownin Figure 21b is the 5.6- µ m emission ring (grey scale) over-laid with 1.2-mm dust continuum contours (Mookerjea et al.2004); the 1.2-mm dust continuum lies next to the bright 8- µ m rim and is coincident with a dark filament (appearingwhite in the grey scale image). According to Mookerjea et al.(2004), this source (MMS16) has a density of ∼ × cm − .The 1.2-mm dust continuum emission agrees well with thearc of N H + emission (Figure 21c) described above. There isno detectable dust continuum from the south-western regionof the ring. Molecular emission is detected from this partbut it is weak and diffuse, indicating a relative low densityregion compared to the dust. Given the completeness andsymmetry of the ring, one might expect the H ii region tolie near the centre. Since it does not, we suggest the possi-bility that this is due to density differences in the gas sur-rounding the infrared ring. In general H ii regions are densitybounded, and the pressurised H ii gas breaks out of the cloudinto lower-density gas, creating a champagne flow (Stahler& Palla 2005). The inhomogeneous density (from our molec-ular line data) allows the ionised gas to spread out furthertowards the south-west compared to the north-east rim, cre-ating a near circular 8- µ m emission structure with the driv-ing source lying on the edge. However the extreme symmetryalso suggests the ring could be a pre-existing structure thatis currently being illuminated by the H ii region. In eitherway this infrared emission ring must be related to the H ii region.Another notable feature is the spatial coincidence ofmm dust continuum, N H + emission and the infrared darkfilament (which appears as white in the grey scale image) tothe north of the ring (Figure 21b and c). As CO is frozenon to grains in dense cold gas (see e.g. Kramer et al. 1999;Bacmann et al. 2002), causing it to be depleted, and theion-molecule formation reaction for N H + proceeds regard-less of temperature, N H + is a good candidate for tracingcold dense gas where CO is depleted. With mm dust contin-uum tracing cold dust along with N H + tracing cold densegas, it is not surprising that they coincide with the infrareddark filament. In this paper we have presented data from a multi-molecularline survey of the southern star forming region, the G333giant molecular cloud complex. For this survey, we have ex-ploited the Mopra MMIC receiver and the 8-GHz bandwidthUNSW-Mopra Spectrometer, resulting in over twenty multi-molecular transition maps, with velocity resolution of ∼ . − . We have presented total intensity maps of moleculeswith bright emission and have also discussed the velocitystructure of the G333 molecular cloud. To further charac-terise the physical and chemical properties, we have carriedout common analysis techniques such as gaussclumps toobtain distributions of CS, HCO + , HNC and C H emission.We have also performed principal component analysis on http://cdsweb.u-strasbg.fr/ ' RAS, MNRAS , 1–25 ulti-molecular lines mapping of the GMC associated with RCW 106 Figure 21. (a) GLIMPSE 3.6- µ m image; (b) GLIMPSE 5.8- µ m image overlaid with 1.2-mm dust continuum contours (Mookerjea et al.2004); (c) GLIMPSE 8.0- µ m images overlaid with N H + integrated emission contours and (d) MIPSGAL 24- µ m image overlaid withMGPS 843-MHz radio continuum contours of the infrared ring discussed in Section 6.3. The grey scale is stretched to 0 - 20 MJy sr − for the GLIMPSE 3.6- µ m, 0 - 120 MJy sr − for the 5.8- µ m image, 0 - 300 MJy sr − for the 8- µ m image and 0 - 800 MJy sr − for theMIPSGAL image. The contours are at multiples of 20 per cent level of each of the emission peaks. The hatched circles represent the beamsize of the respective contour maps. The small circle indicates the position of the MSX point source G333.5114-00.2798, the trianglemarks the position of the
IRAS source, IRAS16182 − ∼ ∼ µ m. Note the ring feature present at 5.8 and 8.0- µ m, which isdominated by PAHs, the bright 24- µ m emission which lies inside this ring. Both the 1.2-mm dust continuum and the integrated N H + emission coincide with the infrared dark filaments, which appear white in the 5.8 and 8.0- µ m grey scale images. the data set, to visualise and parameterise the differencesbetween the spatial distribution of molecules. In this work,we have found:(i) Differences in spatial and velocity distribution amongdifferent molecules. We found that the spatial distributionof CS, HCO + , HCN and HNC are similar on large scales,while N H + seems to trace preferentially the very densestregions. C H is only detected close to bright infrared emis-sion regions. The detected molecules all have similar velocitydistributions.(ii) The velocity gradient across the GMC complex noted in CO (BWC2006) and C O (WLB2008) is also presentin CS, HCO + and HNC.(iii) CS, HCO + and HNC emission maps were decom-posed with gaussclumps in three dimensions. We foundno correlation between clump radius and line width, but aclear correlation between luminosity and radius. Account-ing for saturation effects in the CS ( J = 2 →
1) and HCO + ( J = 1 →
0) lines toward clumps, we obtain column densi-ties of ∼ to ∼ cm − .(iv) An alternative approach used to characterise thisdata set was principal component analysis (PCA). PCAseparates molecules into low ( CO and C O) and high(the rest) density tracers, identifies anti-correlations be- ' RAS, MNRAS000
0) lines toward clumps, we obtain column densi-ties of ∼ to ∼ cm − .(iv) An alternative approach used to characterise thisdata set was principal component analysis (PCA). PCAseparates molecules into low ( CO and C O) and high(the rest) density tracers, identifies anti-correlations be- ' RAS, MNRAS000 , 1–25 N. Lo et al. tween HCO + and N H + , correlations between HCN andHCO + , and helps to explore scanning patterns of the ‘on-the-fly’ mapping.(v) A noteworthy ‘ring-like’ structure of the GMC ispresent in the GLIMPSE 8- µ m image, associated with anH ii region as suggested by radio continuum emission insidethe ring. The molecular line data (especially N H + ) showsgas being swept up and compressed by the ring. ACKNOWLEDGMENTS
The Mopra Telescope is part of the Australia Telescopeand is funded by the Commonwealth of Australia for oper-ation as National Facility managed by CSIRO. The Univer-sity of New South Wales Mopra Spectrometer Digital Fil-ter Bank used for the observations with the Mopra Tele-scope was provided with support from the Australian Re-search Council, together with the University of New SouthWales, University of Sydney and Monash University. PAJ ac-knowledges partial support from Centro de Astrof´ısica FON-DAP 15010003 and the GEMINI-CONICYT FUND. Thisresearch (GLIMPSE and MIPSGAL images) has made useof the NASA/IPAC Infrared Science Archive which is oper-ated by the Jet Propulsion Laboratory, California Instituteof Technology, under contract with NASA.
REFERENCES
Bacmann A., Lefloch B., Ceccarelli C., Castets A.,Steinacker J., Loinard L., 2002, A&A, 389, L6Bains I., Wong T., Cunningham M., Sparks P., Brisbin D.,Calisse P., Dempsey J. T., Deragopian G., et al., 2006,MNRAS, 367, 1609Becklin E. E., Frogel J. A., Neugebauer G., Persson S. E.,Wynn-Williams C. G., 1973, ApJ L, 182, L125+Benjamin R. A., Churchwell E., Babler B. L., Bania T. M.,Clemens D. P., Cohen M., Dickey J. M., Indebetouw R.,et al., 2003, PASP, 115, 953Breen S. L., Ellingsen S. P., Johnston-Hollitt M., Wother-spoon S., Bains I., Burton M. G., Cunningham M., Lo N.,Senkbeil C. E., Wong T., 2007, MNRAS, 377, 491Carey S. J., Noriega-Crespo A., Mizuno D. R., Shenoy S.,Paladini R., Kraemer K. E., Kuchar T. A., Marleau F. R.,et al., 2006, 38, 1023Caselli P., Myers P. C., 1995, ApJ, 446, 665Chin Y.-N., Henkel C., Whiteoak J. B., Langer N., Church-well E. B., 1996, A&A, 305, 960Colgan S. W. J., Haas M. R., Erickson E. F., Rubin R. H.,Simpson J. P., Russell R. W., 1993, ApJ, 413, 237Egan M. P., Price S. D., Kraemer K. E., 2003, 35, 1301Feigelson E. D., Babu G. J., 1992, ApJ, 397, 55Fujiyoshi T., Smith C. H., Caswell J. L., Moore T. J. T.,Lumsden S. L., Aitken D. K., Roche P. F., 2006, MNRAS,368, 1843Fujiyoshi T., Smith C. H., Moore T. J. T., Aitken D. K.,Roche P. F., Quinn D. E., 1998, MNRAS, 296, 225Fujiyoshi T., Smith C. H., Moore T. J. T., Lumsden S. L.,Aitken D. K., Roche P. F., 2005, MNRAS, 356, 801Fujiyoshi T., Smith C. H., Wright C. M., Moore T. J. T.,Aitken D. K., Roche P. F., 2001, MNRAS, 327, 233 Goss W. M., Shaver P. A., 1970, Australian Journal ofPhysics Astrophysical Supplement, 14, 1Green A. J., Cram L. E., Large M. I., Ye T., 1999, ApJS,122, 207Heyer M. H., Schloerb F. P., 1997, ApJ, 475, 173Hirota T., Yamamoto S., Mikami H., Ohishi M., 1998, ApJ,503, 717Jolliffe I., 2002, Principal Component Analysis. SpringerJones P. A., Burton M. G., Cunningham M. R., MentenK. M., Schilke P., Belloche A., Leurini S., Ott J., WalshA. J., 2008, MNRAS, 386, 117Juvela M., 1998, A&A, 329, 659Karnik A. D., Ghosh S. K., Rengarajan T. N., Verma R. P.,2001, MNRAS, 326, 293Kramer C., Alves J., Lada C. J., Lada E. A., Sievers A.,Ungerechts H., Walmsley C. M., 1999, A&A, 342, 257Kramer C., Stutzki J., Rohrig R., Corneliussen U., 1998,A&A, 329, 249Ladd N., Purcell C., Wong T., Robertson S., 2005, Publi-cations of the Astronomical Society of Australia, 22, 62Larson R. B., 1981, MNRAS, 194, 809Lo N., Cunningham M., Bains I., Burton M. G., Garay G.,2007, MNRAS, 381, L30Lockman F. J., 1979, ApJ, 232, 761Loren R. B., 1989, ApJ, 338, 925Lovas F. J., Johnson D. R., Snyder L. E., 1979, ApJ Supp,41, 451Mac Low M.-M., Klessen R. S., 2004, Reviews of ModernPhysics, 76, 125Mookerjea B., Kramer C., Nielbock M., Nyman L.-˚A.,2004, A&A, 426, 119M¨uller H. S. P., Schl¨oder F., Stutzki J., Winnewisser G.,2005, J. Molecular Structure, 742, 215M¨uller H. S. P., Thorwirth S., Roth D. A., WinnewisserG., 2001, A&A, 370, L49Nejad L. A. M., Williams D. A., Charnley S. B., 1990,MNRAS, 246, 183Plume R., Jaffe D. T., Evans II N. J., Martin-Pintado J.,Gomez-Gonzalez J., 1997, ApJ, 476, 730Purcell C. R., Balasubramanyam R., Burton M. G., WalshA. J., Minier V., Hunt-Cunningham M. R., Kedziora-Chudczer L. L., Longmore S. N., et al., 2006, MNRAS,367, 553Reach W. T., Rho J., Tappe A., Pannuti T. G., BroganC. L., Churchwell E. B., Meade M. R., Babler B., Indebe-touw R., Whitney B. A., 2006, ApJ, 131, 1479Rodgers A. W., Campbell C. T., Whiteoak J. B., 1960,MNRAS, 121, 103Rosolowsky E., Leroy A., 2006, PASP, 118, 590Russeil D., Adami C., Amram P., Le Coarer E., GeorgelinY. M., Marcelin M., Parker Q., 2005, A&A, 429, 497Schilke P., Walmsley C. M., Pineau Des Forets G., RoueffE., Flower D. R., Guilloteau S., 1992, A&A, 256, 595Schneider N., Brooks K., 2004, Publications of the Astro-nomical Society of Australia, 21, 290Sheth K., Vogel S. N., Wilson C. D., Dame T. M., 2008,ApJ, 675, 330Stahler S. W., Palla F., 2005, The Formation of Stars.The Formation of Stars, by Steven W. Stahler, FrancescoPalla, pp. 865. ISBN 3-527-40559-3. Wiley-VCH , January2005.Sternberg A., Dalgarno A., 1995, ApJS, 99, 565 ' RAS, MNRAS , 1–25 ulti-molecular lines mapping of the GMC associated with RCW 106 Storey J. W. V., Harnett J. I., Lugten J. B., CrawfordM. K., Stacey G. J., Watson D. M., Genzel R., 1989, MN-RAS, 237, 1001Stutzki J., Guesten R., 1990, ApJ, 356, 513Tafalla M., Myers P. C., Caselli P., Walmsley C. M., 2004,A&A, 416, 191Turner B. E., Thaddeus P., 1977, ApJ, 211, 755Ungerechts H., Bergin E. A., Goldsmith P. F., IrvineW. M., Schloerb F. P., Snell R. L., 1997, ApJ, 482, 245Walsh A. J., Lo N., Burton M. G., White G. L., PurcellC. R., Longmore S. N., Phillips C. J., Brooks K. J., 2008,Publications of the Astronomical Society of Australia, 25,105Williams J. P., de Geus E. J., Blitz L., 1994, ApJ, 428, 693Wilson T. L., Rood R., 1994, ARA&A, 32, 191Wong T., Ladd E. F., Brisbin D., Burton M. G., Bains I.,Cunningham M. R., Lo N., Jones P. A., Thomas K. L.,Longmore S. N., Vigan A., Mookerjea B., Kramer C.,Fukui Y., Kawamura A., 2008, MNRAS, 386, 1069This paper has been typeset from a TEX/ L A TEX file preparedby the author. ' RAS, MNRAS000