Herschel Observations of the W3 GMC: Clues to the Formation of Clusters of High-Mass Stars

0.89 MB·English

by  A. Rivera-Ingraham

Draftversion January17,2013
PreprinttypesetusingLATEXstyleemulateapjv.5/2/11
HERSCHEL OBSERVATIONS OF THE W3 GMC: CLUES TO THE FORMATION OF CLUSTERS OF
HIGH-MASS STARS
A. Rivera-Ingraham,1,2,3 P. G. Martin,4 D. Polychroni,5 F. Motte,6 N. Schneider,6,7,8 S. Bontemps,7,8
M. Hennemann,6 A. Menshchikov,6 Q. Nguyen Luong,4 Ph. Andr´e,6 D. Arzoumanian,6 J.-Ph. Bernard,3
J. Di Francesco,9,10 D. Elia,11 C. Fallscheer,10,9 T. Hill,6 J. Z. Li,12 V. Minier,6 S. Pezzuto,11 A. Roy,4
K. L. J. Rygl,11 S. I. Sadavoy,10,9 L. Spinoglio,11 G. J. White,13,14 C. D. Wilson15
Draft version January 17, 2013
3
1 ABSTRACT
0
The W3 GMC is a prime target for the study of the early stages of high-mass star formation.
2
We have used Herschel data from the HOBYS key program to produce and analyze column density

n and temperature maps. Two preliminary catalogs were produced by extracting sources from the
a column density map and from Herschel maps convolved to the 500µm resolution. Herschel reveals
J that among the compact sources (FWHM<0.45pc), W3 East, W3 West, and W3 (OH) are the most

6 massiveandluminous andhavethe highestcolumn density. Consideringthe unique properties ofW3
1 EastandW3West,theonlyclumpswithon-goinghigh-massstarformation,wesuggesta‘convergent
constructive feedback’ scenario to account for the formation of a cluster with decreasing age and

] increasing system/source mass toward the innermost regions. This process, which relies on feedback
A
by high-mass stars to ensure the availability of material during cluster formation, could also lead to
G the creation of an environment suitable for the formation of Trapezium-like systems. In common
. with other scenarios proposed in other HOBYS studies, our results indicate that an active/dynamic
h process aiding in the accumulation, compression, and confinement of material is a critical feature of
p
the high-mass star/cluster formation, distinguishing it from classical low-mass star formation. The
-
o environmentalconditions and availabilityof triggersdetermine the form in whichthis process occurs,
r implying thathigh-massstar/clusterformationcouldarisefromarangeofscenarios: fromlargescale
t convergence of turbulent flows, to convergentconstructive feedback or mergers of filaments.
s
a Subject headings: ISM:dust,extinction—ISM:individual(Westerhout3)—Infrared: stars—Stars:
[ formation — Stars: early-type

1
v 1. INTRODUCTION clusters, and high-mass star forming sites (see e.g.,
5
0 W3 is a ∼ 4 × 105M⊙ (Moore et al. 2007; Megeath et al.2008foradetaileddescriptionofthefield
and a review of recent literature). The relatively close
8 Polychroniet al. 2012) Giant Molecular Cloud (GMC)
3 well known for its rich population of H II regions, distance of W3 (∼ 2kpc; e.g., Hachisuka et al. 2004;
Xu et al.2006;Navarete et al.2011)hasmadethiscloud
.
1 1Department of Astronomy and Astrophysics, University of a prime target for the study of cluster/high-mass star
0 Toronto, 50St. GeorgeStreet, Toronto,ONM5S3H4,Canada formation,whichcomparedtothelow-masscaseismuch
3 2Universit´edeToulouse,UPS-OMP,IRAP,F-31028Toulouse lesswellunderstood. Thequestionofwhetherhigh-mass
1 cedex4,France
: 3CNRS, IRAP, 9 Av. colonel Roche, BP 44346, F-31028 star formation is simply a scaled-up version of low-mass
v Toulousecedex4,France star formation, or if it is the result of a completely dif-
i 4Canadian Institute forTheoretical Astrophysics, University ferent process(sometimes defined as ‘bimodality’ in star
X ofToronto,60St. GeorgeStreet,Toronto,ONM5S3H8,Canada formation), remains one of the main outstanding issues
5Department of Astrophysics, Astronomy and Mechanics,
r in star formation theory (e.g., Zinnecker & Yorke 2007).
a Faculty of Physics, University of Athens, Panepistimiopolis,
15784Zografos,Athens,Greece Furthermore, while low-mass stars are able to form in
6Laboratoire AIM, CEA/DSM/Irfu - CNRS/INSU - Univer- isolation, most star formation occurs in clusters embed-
sit´e Paris Diderot, CEA-Saclay, F-91191 Gif-sur-Yvette Cedex, ded in their parent GMCs (Lada & Lada 2003). This
France
7Univ. Bordeaux,LAB,UMR5804,F-33270Floirac,France situationisparticularlytrueforhigh-massstars,making
8CNRS,LAB,UMR5804,F-33270Floirac,France clusterstudiescrucialtoinvestigatingandunderstanding
9NationalResearchCouncilCanada,HerzbergInstituteofAs- the origin of high-mass stars.
trophysics, 5071 West Saanich Road, Victoria, BC, V9E 2E7, The W3 GMC was observed with Herschel
Canada
10Department of Physics and Astronomy, University of Vic- (Pilbratt et al. 2010) as part of the Guaranteed
toria,POBox355,STNCSC,Victoria,BC,V8W3P6,Canada Time Key ProgramHOBYS16 (Herschel imagingsurvey
11INAF-Istituto di Astrofisica e Planetologia Spaziali, via of OB Young Stellar objects; Motte et al. 2010). The
FossodelCavaliere100,I-00133Rome,Italy
12National Astronomical Observatories, ChineseAcademyof program is specifically designed to address the major
Sciences, Beijing,China outstanding issues in high-mass star formation with the
13DepartmentofPhysicalsciences,TheOpenUniversity,Mil- analysis of all major regions with high-mass stars at
tonKeynes,UK distances less than d∼3kpc.
14RALspace, The RutherfordAppleton Laboratory, Chilton,
The paper presents a first look at the W3 GMC with
Didcot,UK
15Department of Physics and Astronomy, McMaster Univer-
sity,Hamilton,ON,L8S4M1,Canada 16 http://www.herschel.fr/cea/hobys/en/
2 Rivera-Ingraham,A. et al.
the recently acquired HOBYS data, focusing on those lier stage, prior to IC 1795 (Feigelson & Townsley 2008;
structurescurrentlyhostingtheyoungesthigh-massstars Rivera-Ingrahamet al. 2011).
in this field. By identifying and characterizingthe prop- KR 140 is a small H II region, 5.7pc in diame-
erties and origin of these systems, we aim to constrain ter (Kerton et al. 2001). Located west of the main
the pre-requisites for the formation of clusters of high- star-forming activity in the W3 complex and ∼ 40pc
mass stars. This study is the first of a series of Her- SW of IRS5, this H II region is powered by a central
schel-based papers on W3 currently in preparation, and and isolated O8.5 star (VES 735) about 1−2Myr old
complements our previous analysis of the YSO content (Ballantyne et al. 2000). Contraryto W3 Main, KR140
ofthis field(Rivera-Ingrahamet al.2011)andthe study has been suggested to be the result of a rare case of
of CO by Polychroniet al. (2012) . spontaneous high-mass star formation (Ballantyne et al.
This paper is organized as follows. An introduction 2000).
to the W3 GMC is given in Section 1.1. Section 2 in- North of KR 140 there is a filamentary-like struc-
troduces the Herschel datasets and images. A global ture called ‘KR 140-N’ (Rivera-Ingrahamet al. 2011) or
overviewofthepropertiesofW3anditsfieldsasseenby the ‘Trilobite’ (Polychroniet al. 2010; Polychroniet al.
Herschel based on the column density and temperature 2012). Its morphology and associated population
maps is described in Section 3. In Section 4, we focus of young stellar objects (YSOs) suggest the Trilo-
on the unique properties characteristic of the high-mass bite could be a case of ‘Radiative Driven Implosion’
star formation in W3 Main. Section 5 outlines a new (RDI; Rivera-Ingrahamet al. 2011), in which an ioniza-
high-mass formation scenario, described as ’convergent tion/shock front, such as that driven by an expanding
constructive feedback.’ Our conclusions are summarized HIIregion,causesaneighboringoverdensitytocollapse,
in Section 6 . triggering the formation of stars.
Several of these prominent features are labeled below
1.1. The W3 GMC
in Figures 2 and 3.
W3 contains high-mass stars in various evolutionary
stages (e.g., see Tieftrunk et al. 1997). The most active 1.2. Constraining the High-Mass Star Formation
star forming sites are W3 Main, W3 (OH), and AFGL Process
333. All of these regions (see annotations on Figure 2 Emerging evidence presented in recent studies of sev-
below) are located in a prominent and dense structure eral of the HOBYS fields suggests the importance of
definedintheliteratureasthe‘high-densitylayer’(HDL; dynamical processes in high-mass/cluster formation, as-
e.g., Oey et al. 2005). The HDL forms the western edge sociated preferentially with filamentary-like regions of
of the W4 ‘Heart Nebula’ bubble, powered by eleven O high column density of the order of N ∼ 1023cm−2
starsintheclusterIC1805(2h32m42s+61◦27′(J2000)). (‘ridges’; Hill et al. 2011; Nguyen LuoHn2g et al. 2011a;
W3Maincontainsthemostprominenthigh-masspop- Hennemann et al. 2012). These structures could be
ulation of the entire GMC. The two brightest and most formed by convergence of flows, as proposed for W43
central infrared sources in this region, IRS5 and IRS4 (Nguyen Luong et al. 2011b) and the DR 21 Ridge in
(Wynn-Williams et al. 1972), are the youngest high- Cygnus-X,thelatterbeingoneofthemostprominentex-
mass systems in the GMC, as indicated by the pres- amples (Schneider et al. 2010; Hennemann et al. 2012).
ence of numerous hypercompact H II (HCH II) regions Theeffectsofthisprocessinturbulentenvironmentshave
(Tieftrunk et al. 1997). IRS5 has been suggested to been investigated extensively in previous studies (e.g.,
containa proto-Trapeziumsystem (Megeath et al.2005; Klessen et al. 2004; Klessen et al. 2005; Heitsch et al.
Rodo´n et al.2008)poweredbyaclusterofOBstars(e.g., 2006). Other studies of high-mass star forming regions
Claussen et al. 1994; van der Tak et al. 2005). suggest that enhanced accretion at the mergers of fil-
The W3 (OH) region is also comprised of two main aments is necessary in order to form clusters on short
regions: W3 (OH) itself, a young ultracompact H II timescales (Dale & Bonnell 2011; Schneider et al. 2012).
(UCH II) region rich in OH masers (Dreher & Welch Such a rapid formation process is requiredto agree with
1981), and a hot core with a younger massive proto- estimates of the lifetime of the prestellar phase of mas-
binarysystem,∼6′′eastofW3(OH)(Chen et al.2006). sive YSOs, which may be as short as a single free-fall
The region associated with the binary has prominent time (e.g., Motte et al. 2007).
H2O emission and it is commonly referred in the liter- Questions remain as to whether such models can suc-
ature as W3 (H2O) (Little et al. 1977; Turner & Welch cessfully explain the origin of other known high-mass
1984). systems in other regions. If more than one scenario is
W3 (OH) and W3 Main are both located in the shell needed to explain these systems, it is essential to iden-
surrounding the cluster IC 1795. This 3 − 5Myr old tify the conditions common to all that might constitute
cluster is powered by various OB stars, the most mas- the basic requisites for the formation of OB stars and
sive of which, BD +61◦411, has a spectral type O6.5V their associatedclusters. We return to this in Sections 5
(Oey et al. 2005). The location of W3 Main and W3 and 6.
(OH) at the edges of the shell around this cluster has
been suggested as evidence for their formation having 2. DATAPROCESSINGANDHerschel IMAGES
beeninducedbyIC1795,itselfcreatedbyanearlierburst W3 was observedby Herschel in March2011. Contin-
of star formation in the W4 region (Oey et al. 2005). uum data were obtained in parallel mode at 70µm and
While a triggered origin by IC 1795 appears to be the 160µm with PACS (Poglitsch et al. 2010), and 250µm,
case for W3 (OH), the characteristics of the cluster of 350µm, and 500µm with SPIRE (Griffin et al. 2010).
low mass stars associated with W3 Main suggest that The most prominent thermal emission features in W3
this regionmight actually havebegun to form atanear- areshowninunprecedenteddetailbythe Herschel data.
Herschel Observations of High-Mass Star Formation in the W3 GMC 3
Taylor et al. 2003) were used to investigate the location
and distribution of the ionized regions.
3. COLUMNDENSITYANDDUSTTEMPERATUREMAPS
3.1. Data Analysis
To derivecolumn density anddust temperature maps,
the PACS and SPIRE maps were first convolved to the
resolution at 500µm (∼36′′; ∼0.35pc at a distance of
2kpc)andprojectedontoa common9′′pix−1 grid(that
of the 500µm map).
The intensities at each pixel were fitted with the ap-
proximation20
I =(1−exp(−τ ))B (T), (1)
ν ν ν
where B (T) is the Planck function for dust tempera-
ν
ture T. The common assumption in such studies is that
T is constant along the line of sight, obviously a great
simplification given the variety of structures and radia-
tion fields. Thus only Herschel bands ≥ 160µm were
considered in creating the maps. The optical depth is
parametrized as τ = τ (ν/ν )β with ν = 1200 GHz
ν ν0 0 0
(250µm). The optical depth is related to the total H
column density by τ =σ N through the opacity σ .
Figure 1. Three color image of the W3 GMC using Herschel ν ν H ν
We assumed a fixed dust emissivity index of β = 2
HOBYSdataat70µm(blue),160µm(green)and250µm(red). The
color tables have been manipulated to bring out the structural and adopted the same opacity standard as in other
detailinthemap. HOBYS and Herschel-based studies (e.g., Motte et al.
Datasets were reduced from level 0 to level 1 using 2010;Andr´e et al.2010),0.1cm2 gm−1 at1THz,equiv-
HIPEv8.1,andthefinalmapswereproducedwithv17.0 alently σ0 = 3.4×10−25cm2H−1 using 1.4 as the mean
ofthe Scanamorphos17 softwarepackage(Roussel 2012). atomicweightperHnucleon. Thiscanbeusedtodeduce
AstrometrywascorrectedusingtheTwoMicronAllSky the total H column density from τ. However, again to
Survey18 and Spitzer data19. The pointing accuracy is be consistent with the scale of column densities in these
estimatedtobebetterthan2′.′0. Mapsweretransformed earlier papers, we used 2.33 rather than 2.8 as the mean
toMJy/srandoffsetsweredeterminedusingPlanck and atomic weight per molecule, so that what is called NH2
IRAS data (Bernard et al. 2010). isreallythecolumndensityof‘meanmolecules’(seealso
A small saturated area was detected in W3 (OH) at Kauffmann et al. 2008). There is no scaling effect on
70, 250, and 350µm (∼2h27m4s +61◦52′24′′). This the mass surface density or the masses of clumps. More
corresponds to the source IRS8 (Wynn-Williams et al. importantly, we note that intrinsic uncertainties in the
1972). Other small (a few pixels) regions were also sat- adoptedopacitymightalterthederivedcolumndensities
urated in W3 Main at 70µm (2h25m40s.5 +62◦05′50′′) systematically by a factor of ∼ 2, and that there might
and 250µm (2h25m40s.5 +62◦05′50′′; 2h25m30s.5 be systematic differences in opacity across the field or
+62◦06′). These correspond to the sources IRS5 and with column density (Martin et al. 2012). Compared to
this, photometric calibration errors are relatively small,
IRS4, respectively (Wynn-Williams et al. 1972). These
15%and20%forSPIREandPACS160µm,respectively
regions were reobserved with Herschel using the ‘bright
(Griffin et al. 2010; Poglitschet al. 2010).
source’ mode. These images showed a good correlation
Our goal is to assess the column densities and tem-
with the originals in the unsaturated areas of overlap
perature within the W3 GMC itself. The Herschel in-
and were used to replace the intensities in the saturated
tensities I , however,contain contributions from dust in
pixels. ν
theforegroundandbackground,eachwithitsownσ ,T,
A three-color image of W3 is shown in Figure 1. The ν
and N ; the right hand side of Equation [1] is summed
five individual continuum images are presented in Ap- H
over all components. While a challenging exercise, ed-
pendix A. Our analysis focuses on the ‘common survey
area’, ∼1.5◦ in extent in RA and Dec, scanned in two ucated subtraction of the non-GMC components is ad-
vantageousfor providing a more accurate representation
roughly orthogonal directions by PACS and SPIRE ac-
of the true local conditions in the GMC. The process of
cording to the telescope focal plane orientation on the
subtractionis described in Appendix B. Subtracting the
date of observation.
foreground and background is most important for char-
Supplementary Stokes I continuum images at 1420
acterization of regions where the column density in the
MHz from the Canadian Galactic Plane Survey (CGPS;
GMC is relatively low; it has little effect on the regions
17 http://www2.iap.fr/users/roussel/herschel/index.html of highest column density.
18 The Two MicronAll Sky Survey (2MASS) is a jointproject The final corrected column density map is shown in
oftheUniversityofMassachusettsandtheInfraredProcessingand Figure 2. The accompanying temperature map is shown
AnalysisCenter/CaliforniaInstitute ofTechnology, fundedbythe
National Aeronautics andSpaceAdministrationandthe National 20 Atthecolumndensitypeaks(Section4),theopticaldepthat
ScienceFoundation. 160µmexceeds0.3. Itisofcourseevenhigherat70µm,butthose
19 http://irsa.ipac.caltech.edu/applications/Gator/ dataarenotusedinourassessment.
4 Rivera-Ingraham,A. et al.
Figure 2. Column density map of the W3 GMC after correction for dust emission associated with foreground/background atomic and
molecular material, as described in Appendix B. A variety of filaments, pillars and structures are found throughout the GMC. Labels
markprominent features inW3(Section 1.1). The‘HDL’isthe denseregioncomprisingW3Main,W3 (OH), andthe AFGL333Ridge.
ContoursareNH2 ≈[30,60,200]×1021cm−2.
in Figure 3. Both figures label the most prominent fea-
Table 1
tures of the W3 GMC. Unless mentioned otherwise, we GlobalParametersfortheW3Fields
used these corrected maps as the default images for our
analysis. Field Massunca Massb Area Σa Σb
Forpurposesofcomparisonanddiscussion,weseparate [104M⊙] [104M⊙] [deg2] [M⊙/pc2] [M⊙/pc2]
Main/(OH) 7.2 6.0 0.43 136 114
the GMC into four different‘fields,’ labeled accordingto
AFGL333 8.1 6.3 0.50 133 103
their physical location with respect to the center of the KR140 9.4 5.9 0.74 104 65
entire cloud or a major feature present in the field. The NW 6.5 4.8 0.53 101 76
fields are: W3 Main/(OH), W3 NW, AFGL 333, and a Massandsurfacedensityfromuncorrected (original)maps.
b Massandsurfacedensitycorrectedforforegr./backgr. material.
KR 140 field (see Fig. 3). The W3 Main/(OH) and
AFGL 333 fields together comprise what we define as using a simple temperature/columndensity thresholdto
the ‘eastern’ (or HDL) fields in W3, while the other two identify the youngest stages of high-mass star formation
are the ‘western’ (‘quiescent’ and more ‘diffuse’) ones. in large scale surveys.
Taking into account the material at all temperatures,
3.2. Global Overview wefindthatthefourfields,thoughofunequalareawithin
the common science region, have comparable mass. The
Figure4presentstherelationshipbetweencolumnden-
main parameters for the four fields, including masses,
sityanddusttemperaturefor the fourfields,inthe form
areas,andsurfacedensities, Σ, for eachofthe four fields
of a two-dimensional histogram. The W3 Main/(OH)
are shown in Table 1. This also includes high column
field ranges to higher temperatures not observed in the
density material above the limits displayed in Figure 4,
cooler and more diffuse western fields (W3 NW and
which belongs mainly to the W3 Main/(OH) field.
KR 140). The AFGL 333 field shows a similar high-
Thetotal(allfields)correctedmassoftheW3GMCis
temperature extension, associated with material at the
boundarywithW4,butwithpeaktemperaturesbetween foundtobe∼2.3×105M⊙ (∼3.1×105M⊙,uncorrected
for foreground-background ISM contribution). This es-
those in W3 Main/(OH) and the western fields. In this
timate is about half that derived by Polychroniet al.
field there is also low-temperature material, not present
(2012) from molecular data. Moore et al. (2007) found
in the W3 Main/(OH) field, associated with the struc-
tures of the East Loop (Fig. 2), and with properties ∼3.8×105M⊙ from13COforaconstantexcitationtem-
perature of ∼30K. Our mass estimates would be raised
consistent with those observed in the W3 NW and KR
(lowered) systematically by adopting a lower (higher)
140 fields.
opacity.
While most regions in W3 show a clear anticorrela-
tion between high column densities and low tempera-
3.3. Stellar Influence and Cloud Structure
tures, the Main/(OH) field is unique due to the associa-
tionofhighcolumndensitieswithwarmertemperatures. TheionizinganderodingactivityoftheOstarspower-
These structures, warmed by embedded OB stars, are ing W4 is clearly observed along the eastern edge of our
also associated with the youngest high-mass activity in mapped field. The ionization front can be traced in the
W3 (e.g., W3 Main). This illustrates the difficulty of radiocontinuum,asshownwithStokesIcontinuumcon-
Herschel Observations of High-Mass Star Formation in the W3 GMC 5
Figure 3. Dust temperature map of the W3 GMC after correction for dust emission associated with foreground/background atomic
and molecular material, as described in Appendix B. Contours are of Stokes I continuum at 1420 MHz: Tb =8,12,15K (magenta), and
Tb = 30,100, and 240K (black). Colors for contours have been chosen for better contrast in cold and warm regions. White circle with
crossmarkstheintersectionofthefourfieldsinW3. Fromlefttorightandtoptobottom: W3Main/(OH)field,W3-NW,AFGL333,KR
140fields. Figureincludeslabelsforthefourfieldsandthelocationofthe‘Y’-shapedhotstructure(seetext).
to be part of the coldest structures in the W3 GMC.
There are groups of south-oriented pillars in the cen-
tral regions of W3, marked P2 and P3 in Figure 5 and
seen in more detail in Figure 1 and the original single-
band Herschel images in Appendix A. Interaction with
W4 alonecannotaccountfor the levelofcomplexityand
diversityofsuchfeaturesfoundwestoftheHDL;instead
they are suggestive of more local stellar influence in the
area. Furthermorphologicaldetailsandstellarcontentof
thecentrally-located(P2,P3)andW4-facingpillarswere
examinedwithSpitzer imagesbyRivera-Ingrahamet al.
(2011).
Figure 4. Two-dimensional histogram of dust temperature and
column density in each W3 field. Color bar indicates number of
pixels (log units). Black contour marks the distribution of the
W3NWfieldasareferenceforthedistributionintheotherfields.
tours in Figure 3. Diffuse radio emission is observed to
bemostprominentintheHDL,withthestrongestpeaks
coincident with W3 Main, and decreasing progressively
into the colder western fields.
Various elongated, dense cloud structures emerging
paralleltooneanotherfromtheHDLareobservedtoex-
tend more than 10 pc in projection along the boundary
with W4, pointing toward the east and the O-star clus-
ter IC 1805 (see Fig. 5). The most prominent of these
structures (defined as ‘pillars’ in this work), is located
just northeast of AFGL 333 and has signatures of ongo-
Figure 5. Same as Fig. 3, but with intensity and color range
ing star formation (Rivera-Ingrahamet al. 2011). The modified to highlight the coldest structures. Circles mark the lo-
locationofthisstructure(P1)andotherpillar-regionsin cationofpillar-likestructures. TheIC1805clusterislocatedeast
W3 are marked below in Figure 5, which shows several oftheAFGL333Ridge,at2h32m42s +61◦27′ (J2000).
6 Rivera-Ingraham,A. et al.
Contrary to the HDL and eastern fields, stellar influ- N >6×1022cm−2)asshowninFigures7and8which
H2
ence in the western fields is very localized, as shown in concentrate on these most dense regions in the GMC.
Figure 3. Examples of these ‘hot’ spots are the KR 140
H II region and a ‘Y’ shape structure (Fig. 3) at ∼
2h22m37s+61◦49′41′′. Abow-shockshaped‘high’tem-
perature feature wraps around the eastern side of KR
140-N (Trilobite); displacement in the direction of the
YSOs associated with this structure supports a possi-
ble case of triggered RDI origin (Rivera-Ingrahamet al.
2011).
4. CLUESTOTHEELUSIVEHIGH-MASSSTAR
FORMATIONPROCESS:THECASEOFW3MAIN
Withthepresenceofongoinghigh-massstarformation
andN oftheorderof∼1023cm−2,theW3GMCoffers
H2
afavorableopportunitytocharacterizerarehigh-column
density structures and understand how their properties,
origin,andevolutionmightbelinkedtotheonsetofhigh-
mass activity.
4.1. Identification of High-Column Density Structures
Figure 7. Same as Fig. 6, but for W3 Main, W3 (OH), and
Figure 6 shows the temperature map of what we de-
fine as the ‘AFGL 333 Ridge’ (Fig. 2), in the innermost W10231Ncmor−th2.with an additional (white) contour at NH2 ≈ 200×
regions of the AFGL 333 field. It is cold (<T>∼ 15K,
T ∼14K,for N >6×1022cm−2). The AFGL 333
min H2
Ridge is also the only filamentary-shaped structure in
theGMCreachingpeakcolumndensitiesof∼1023cm−2.
Thismakesitsimilartothosefilamentsdefinedas‘ridges’
in previous HOBYS papers (e.g., Hill et al. 2011).
Figure 8. SameasFig. 7,butfocusingontheW3Mainregion.
The highest column density is in W3 (OH), N ∼
H2
[4.5±1.0]×1023cm−2, comparablewithin errorsto that
of W3 West, N ∼[3.3±0.8]×1023cm−2.
Figure 6. Temperature map of the structural details around Column denHs2ities of this order are consistent
the central regions of the AFGL 333 field. Black contours are
NH2 ≈[20,30]×1021cm−2. WhitecontoursareNH2 ≈[60,100]× with the findings reported by White et al. (1983),
1021cm−2. Richardson et al. (1989), and the presence of a strong,
highly variable water maser in the W3 (OH) region
Krumholz & McKee (2008) predicted a star formation (Little et al. 1977). W3 East is found to have a peak
thresholdforhigh-massstarsofΣ=0.7gcm−2 forastar column density 1.4 times lower than W3 West (NH2 ∼
with M ∼ 10M⊙. This surface density corresponds to [2.3 ± 0.6] × 1023cm−2), which agrees with the col-
N ∼ 1.8× 1023cm−2 in our maps. Although theirs umn density ranking of Tieftrunk et al. (1998b) based
H2
is a specific model, we will for convenience refer to this on molecular C18O data. We note, however, that these
column density as a ‘massive star formation threshold’ authors predict a column density estimate for W3 (OH)
(MSFT). Only three structures in the W3 GMC reach from NH comparable to that of W3 East.
3
columndensitiesofthisorder,theW3EastandW3West W3East,W3West,andW3(OH)aretheonlyclumps
clumps (FWHM∼ 0.45pc) in W3 Main and W3 (OH) with confirmed ongoing (clustered) high-mass star for-
(FWHM∼ 0.43pc). Compared to the AFGL 333 Ridge, mation in the W3 GMC, as shown by the presence of
these are all warm (<T>∼ 25K, T ∼ 18K, for masers, HCH II, or UCH II regions. While the MSFT
min
Herschel Observations of High-Mass Star Formation in the W3 GMC 7
value is used in this work just as a point of reference The identification of properties exclusive to W3 East,
whenrankingcolumndensities, ifsuchathresholdholds W3 West, and W3 (OH) (structural and stellar con-
in practice, then a total combined mass of ∼ 2600M⊙ tent)mightrevealcluesabouttheiroriginandaboutthe
for N ≥MSFT, about ∼ 1% of the mass of the W3 general process of high-mass star and cluster formation.
H2
GMC, is at present possibly associated with high-mass Therefore,we attempt here to find some such properties
star formation. by comparing the Herschel-derived results with charac-
teristics of the extensively studied stellar population of
4.2. Masses and Luminosities the sources in W3 Main (W3 East and W3 West), the
only sources in the W3 GMC with recent/ongoing high-
As a first reconnaissance, we ran the multi-scale,
mass star formation (containing HCHII regions).
multi-wavelength source extraction software getsources
(Men’shchikov et al. 2012; v.1.120828) on the column
density map (Fig. 2) made from the Herschel images.
Some properties of the three most massive clumps asso-
ciated with the highest column density peaks are sum-
marized in Table 2, assuming a distance of 2.0 kpc. Be-
causegetsources estimatesthelocalbackground,thecor-
responding values above the background, as well as the
errors as estimated by getsources, are presented. The
contributions to the luminosity within the footprint of
thestructureareevaluatedusingthemassandT (Fig.3)
associated with each pixel.
In fitting a modified blackbody, analogous to Equa-
tion 1, to the spectral energy distributions (SEDs) from
these data,we usedβ =2andfound the temperatureT,
the luminosity L,and massM using the aforementioned
opacity. The resulting parameters from our ‘cold’ SED
fits (λ ≥ 160µm), and their uncertainties as estimated
from Monte Carlo simulations, are included in Table 3.
Ideallyonewouldmeasurethefluxdensitiesbyrunning
getsources ontheoriginalHerschel images. However,our
choice to use the 500µm resolution for all maps avoids
the use of the flux density scaling recipe, used in other
Figure 9. DistributionofHIIregions(triangles;Tieftrunketal.
HOBYS fields to correct flux densities for the different 1997)andOBstars(bluestars;Biketal.2012)inW3Main. Con-
resolutionspriortofittingtheSED,butwhoseuseisonly tours are the same as in Fig. 8. Ellipses are the FHWM ellipses
applicableforcoresundercertainconditionsandphysical provided bygetsources duringsource extraction onthe convolved
Herschel maps(blue)andonthecolumndensitymap(red). Axis
characteristics (Motte et al. 2010; Nguyen Luong et al.
units are in pixels (9′′) offset from position RA/Dec:2h25m35s.5
2011a). +62◦05′59′′. Labels mark the location of high-mass stars men-
All of these estimates are in agreement and con- tionedinthetext.
firm that the masses of the ∼ 0.4pc W3 (OH), W3
West, and W3 East dense clumps are of the order ∼ Both clumps are unique in the W3 GMC due to
103M⊙. These are also in agreement with previous esti- their large population of young high-mass stars (e.g.,
mates(e.g.,Campbell et al.1995;Tieftrunk et al.1998b; Tieftrunk et al. 1997). These stars are present not only
Megeath et al. 2008 and references therein). within the clumps, but also surrounding their outer
Regardless of the catalog used (that derived from the boundaries. This can be observed in Figure 9, which
columndensitymaporthemultibandmaps)theHerschel shows the location of the OB population with respect
data indicate that the clumps currently hosting the on- to the Herschel detections. The spectral types of these
going high-mass star formation of W3 (W3 East, W3 stars were analyzed in detail by Bik et al. (2012).
West,andW3(OH))arethemostmassiveandluminous
of the entire GMC. Only two sources from the column 4.3.1. W3 East
density source catalog, located in the AFGL 333 Ridge,
W3 East is the most active of the two in terms of the
have a mass of this order (greater than W3 East, albeit
associated(externalandinternal)stellarpopulation. Us-
less than W3 (OH) andW3 West). They have,however,
ing the technique of Rivera-Ingrahamet al. (2010) we
a lower peak column density and a mean FWHM twice
estimate that the luminosity of the W3 East clump is
that of the three high extinction sources. Similarly, the
equivalent to a single-star ZAMS spectral type (Panagia
only other detection (in addition to W3 (OH) and W3
1973) of ∼O7. However, because we did not use the
West)fromtheconvolvedmultibandcatalogwithamass
70µm datum in fitting the SED, the temperature and
greater than that of W3 East (also associated with the
luminosityofthis hotsourcearepoorlyconstrained,and
AFGL 333 Ridge), also shows a mean FWHM500µm ∼ so the allowed range of spectral type is wide, B0.5 to
1.5 times larger. There are no reliably-measuredsources
O6. When including the 70µmmeasurementthe SEDis
with a luminosity reaching that of W3 East, W3 West,
muchmoretightlyconstrained,withthe resultT≈39K,
or W3 (OH).
M≈ 560M⊙, L≈ [1.5± 0.2]× 105L⊙, and an implied
spectral type of O6.5. Use of the luminosity Class V ta-
4.3. Stellar Content in a Herschel Context
ble in Martins et al. (2005) would make the type ‘later’
8 Rivera-Ingraham,A. et al.
Table 2
ThehighextinctionstructuresintheW3Main/(OH)field: parametersfromtheNH2 andT maps
Name R[hAms] D[◦e′c′′] P[1e0a2k3cNmH−22a] [PKe]akTb [M10a3ssM⊙] [<pFc]WHM>c
W3West 022530.1 620607 2.9±0.8 27.2±2.2 1.54±0.03 0.45
W3East 022540.8 620552 1.8±0.6 30.4±3.9 0.87±0.03 0.42
W3(OH) 022703.7 615221 4.1±1.0 25.0±1.9 1.85±0.03 0.42
a Measuredatthecoordinatecenter abovethelocalgetsources-estimatedbackground
b Measuredatthecoordinatecenter.
c GeometricmeanFWHMofellipticalfootprint.
Table 3
ThehighextinctionstructuresintheW3Main/(OH)field: parametersfromSEDfittinga
Name RA Dec Mass L T <FWHM>b offsetc
[hms] [◦′′′] [103M⊙] [104L⊙] [K] [pc] [′′]
W3West 022530.5 620613 1.7±0.5 3.5±1.8 24.9±3.2 0.46 6.8
W3East 022540.6 620553 0.8±0.3 10.3±9.1 32.6±6.2 0.42 1.9
W3(OH) 022703.8 615223 1.6±0.5 2.3±1.0 23.3±2.7 0.42 2.0
a Usinggetsources parametersfrommapsconvolved to500µmresolution.
b GeometricmeanFWHMofellipticalapertureat500µm.
c DistancefromtheNH2 peaklistedinTable2
by ∼ 0.5−1. The YSO and proto-Trapezium system The other members of the young population of W3
IRS5 is located at the column density peak. Bik et al. West lie close to or beyond the boundary of the col-
(2012) provided no spectral type estimate for IRS5, but umn density structure (Fig. 9). The location of a YSO
our value is compatible with that of IRS3a, the earliest and other high-mass stars toward the southern and SW
reported star within the FWHM of W3 East, with an boundariescouldsuggestinteractionwiththediffuseHII
estimated spectral type of O5–O7. Also located within regions at the south.
the FWHM (boundary), offset ∼ 0.1pc from the col-
umn density peak, is IRS7, a late O/early B star associ- 4.4. High Column Density Structures Lacking
ated with an UCHII region and, closer still to the peak High-Mass Star Formation Indicators
(∼0.05pc), IRSN7 (Fig., 9), alsoa YSO but olderthan
In addition to W3 (OH), Herschel identifies only two
IRS5 (Bik et al. 2012).
other structures in the W3 GMC with column densities
Despite the significant population of high-mass stars
of the order N ∼ 1023cm−2: W3 SE (Fig. 8), in the
within its boundaries, W3 East shows evidence of exter- H2
W3 Main region, and the AFGL 333 Ridge (Fig. 6).
nal heating. The highest dust temperatures of the en-
tire GMC (∼ 32K) are located outside the NE and SW W3 SE is the coolest of the three clumps in W3 Main.
It is located ∼ 1.3pc from IRS5, with the closest high-
boundary of the N clump (Fig. 8), and are coincident
H2 mass star indicator being a diffuse HII region <1pc to
with the many late O/early B stars . 0.5pc from the
the southwest.
peak (Fig. 9). These outer stars range in spectral type
AlsolocatedintheHDL,theAFGL333Ridge,despite
from early B to O6.5 (IRS2) (Bik et al. 2012).
beingattheboundarywithW4,showsevidenceofamore
locally triggered origin. For example, it has:
4.3.2. W3 West
i) anelongatedmorphologyonthe eastcurvedaround
Despite having more mass and higher column density an embedded cluster IRAS 02245+6115 (this cluster is
than W3 East, W3 West has a lower dust temperature ∼ 1pc from the strongest column density peak in the
and is relatively more quiescent. High-mass star forma- Ridge and contains a compact HII region powered by a
tionappearsnotto haveyetprogressedorbeeninitiated B0.5-type star; e.g., Hughes & Viner 1982);
intheinnermostregions;thereisnoindicationofinternal ii) a distribution of YSOs that follow the curvature of
high-mass star phenomena coincident with the Herschel the structure and is abundant in the boundary between
NH2 peak. Corroborating this, there is an NH3 peak at theRidgeandthe cluster(Rivera-Ingrahamet al.2011);
thepositionofthe columndensitypeak(Tieftrunk et al. and
1998b), and this peak is offset from any PACS (hot) in- iii) an overall much younger population compared
fraredsourceoranyHIIregion. Bycontrast,thecolumn to all the other YSO groups in the rest of the field
density peak of W3 East lacks significant NH3 emission (Rivera-Ingrahamet al. 2011). Indeed, the AFGL 333
suggesting a more advanced state of evolution in which Ridge contains∼70%ofthe Class 0/Ipopulation in the
star formationhas already influenced the parentalcloud AFGL333fieldbutonly∼5%oftheClassIIpopulation
locally (Tieftrunk et al. 1998b). (this census excludes the population in the East Loop
Only IRS4 is found within the boundaries of the W3 whoseenvironmentalconditionsaremoreconsistentwith
Westclump, ∼0.1pcfromtheNH2 peak(Fig. 9). Itisa the westernfields than the HDL; Rivera-Ingrahamet al.
B0.5–O8starpredictedto be as youngas IRS5 basedon 2013,in preparation).
the presence of HCHII regions (Bik et al. 2012). This is W3SEandtheAFGL333Ridgearebothformingstars
inexcellentagreementwiththesingle-starmain-sequence (Rivera-Ingrahamet al. 2011) and both appear to have
(ZAMS)spectraltypeB0.5(O9.5)derivedfromtheHer- the potential to form high-mass stars. They also have
schel luminosity of L∼[3.5±1.8]×104L⊙. a possible ‘trigger’, i.e., a high-mass star, in their local
Herschel Observations of High-Mass Star Formation in the W3 GMC 9
neighborhood that could aid in the process. However, effectsofsub-parsectriggeringactingwithinthehighcol-
they have notreachedthe column densities,masses,and umn density structure being formed, together with the
degree of stellar activity (internal and external to the amountofmassandlimitedrangeofthetriggering,could
clumps) characterizing W3 East and W3 West. sustainlastingperiodsofstarformationinthemostcen-
tral regions, therefore emphasizing the differential age
5. FORMATIONOFCLUSTERSWITHHIGH-MASSSTARS effect.
AND‘TRAPEZIUM-LIKE’SYSTEMSBY‘CONVERGENT
CONSTRUCTIVEFEEDBACK’ 5.1.2. Constructive
A very interesting point noted in previous studies of The ‘positive’ effects of stellar feedback by high-mass
both W3 East and W3 West (e.g., Tieftrunk et al. 1997; starshavebeenstudiedextensivelyintheGalaxy,includ-
Bik et al. 2012) is a progressive decrease in age of the ingotherHOBYSfields(e.g.,Zavagno et al.2010;Minier
stellar population from the outskirts toward the peak of et al. 2013, A&A, in press). Here we argue that high-
the density structure, from the farthest diffuse H II re- mass stars (‘triggers’) can collectively influence not only
gions, to more centrally located evolved compact, and the creation of new high-mass stars, but also the new,
HCH II regions. This is a critical clue. The physical massive structures hosting this new population of high-
properties inferred from Herschel for W3 East and W3 mass stars. Indeed, it is a requirement of our scenario
West are unique in the entire W3 GMC. In order to in- thatinthedenseenvironment,theprogressiveformation
terprettheminthecontextoftheir(alsounique)geome- of high-mass stars will in addition result in the creation
try,stellarpopulationcharacteristics,andstarformation ofevenhigher columndensities byfeedback,ratherthan
history,weproposeascenarioforformationofamassive simply disruption and dispersal by their mechanical and
clumpsuitableforhostingaclusterofhigh-massstarsas radiative output.
well as for formation of the individual high-mass stellar This is in contrast to alternative models of high-mass
members. We now describe the key features of what we cluster formation, like that presented in Peters et al.
call ‘convergent constructive feedback’. (2010). In the scenario presented by these authors a
centralstar forms first, followedby the formationof sec-
5.1. Key Features ondary stars in the accretion flow (i.e., star formation
propagates outward).
5.1.1. Feedback
5.1.3. Convergent
Low and high-mass star formation induced (‘trig-
gered’) by external OB stars is a well-studied phe- To achieve the constructive behavior, we suggest that
nomenon supported by extensive theoretical and obser- the configuration of the high-mass triggers is a key fac-
vational studies (e.g., White et al. 1999; Tothill et al. tor in the formation of the most compact and massive
2002; Minier et al. 2009). Local stellar feedback could systems (i.e., Trapezium-like systems), like those in W3
result in much faster and more efficient star formation Main. When acting on a dense region with enough
thaninthequiescentmodeattheinteractionboundaries mass, different populations of high-mass stars with the
between the ‘triggering’ stars and a dense environment right ‘confining’ configurationcan lead to a ‘convergent’
(e.g., Elmegreen & Lada 1977). process, creating/enhancing a central massive structure,
We might have observed this effect between W4 and moving/trapping the material, and ensuring the avail-
the HDL, near AFGL 333 (e.g., Rivera-Ingrahamet al. ability of mass for accretion during the early otherwise
2011; Rivera-Ingrahamet al. 2013, in preparation). En- disruptive stages of high-mass star formation. The cen-
hanced star formation in triggeredregions has also been tralcolumndensitywouldcontinuetogrowasnewhigh-
reportedinpreviousstudies(e.g.,Thompson et al.2012; mass stars form in a sequential process by sub-parsec
Koenig et al. 2012). triggeringatthe ‘boundaries’ofthe highcolumndensity
However, the progress of triggering is predicted to be region, where triggering is most effective.
dependentontheenvironmentalconditions. Ahigh-mass A particularly favorable case could arise where a mas-
star in a dense environment (e.g., a shell or ridge) can siveanddensestructureisformedbetween separatehigh-
only induce further compression only at smaller (sub- mass star populations, due to their combined effect in
parsec) scales; the disruptive effects of newly formed compressing and confining the material. W3 West ex-
high-mass stars are not efficient in relatively dense re- emplifies this possibility. This prominent quiescent col-
gions (Dale & Bonnell 2011) and are therefore of more umn density peak has infrared sources at its periphery,
limited range. Nevertheless, this further compression with a clear anticorrelation between molecular and ion-
could propagate the triggering process and the forma- ized gas (e.g., Tieftrunk et al. 1998b). A similar confin-
tion of new high-mass stars on the relevant small scales, ing arrangement is also observed for W3 East, and fur-
starting at the boundaries and progressing toward the thermore, higher resolution studies have shown an even
more dense (inner) regions of the compressed structure. smaller central core containing IRS5, within our Her-
This process would explain, for instance, the presence schel clump, surrounded by the other HII regions (e.g.,
of molecular cores and condensations at the perime- Tieftrunk et al. 1995), suggesting that W3 East con-
ter of the ionized regions (e.g., Tieftrunk et al. 1995; tainsasmallerscale(andalreadyactiveversion)ofwhat
Tieftrunk et al. 1998a) in W3 Main (W3 East). is occurring in W3 West. In addition, Tieftrunk et al.
While (low/intermediate-mass) star formation likely (1998b) suggested that the lack of ammonia associated
has taken place in the dense region prior to the present withW3EastwasduetoNH being‘thinnedout’orde-
3
triggering, accounting for the large cluster of low mass stroyedbyongoingactivitywithout dispersal,whichalso
starsintheregion(Feigelson & Townsley2008),therate supportsthe‘confinement’aspectoftheprocess. Incon-
andefficiencywilllikelybeenhancedatlaterstages. The trast,anammoniapeakis stillpresentinW3 West. The
10 Rivera-Ingraham,A. et al.
combined evidence supports a similar process for both provide a useful framework for addressing some of these
clumps, but with W3 West at an earlier stage. outstanding problems in high-mass star formation.
The concept of convergence is critical for high-mass First, given the age spreads in W3 observed by
star/cluster formation as an active process aiding in the Bik et al. (2012), a progressive formation of the cen-
continuing supply of material, beyond that required to tral column density clump and cluster members must
form a low-mass star. Convergence might arise for a have occurred over a 2 − 3Myr period. Therefore,
variety of different reasons. Other than our scenario there is no need to form such structures fast enough
above, evidence for different geometries is presented in to prevent major internal fragmentation or to invoke
recent Herschel HOBYS studies for the importance of long ‘starless’ lifetimes (e.g., several free-fall times;
convergence of flows (Hennemann et al. 2012) and of McKee & Tan 2003), which allow for the material to
junctions of filaments (Schneider et al. 2012). In ad- be gathered before the star formation process is initi-
dition to those scenarios aiming to describe the origin ated (e.g., Zinnecker & Yorke 2007). The ‘older’ halo
of the parsec-scale progenitors of massive clusters, there cluster of low-mass stars surrounding the high-mass
are various models describing the origin of the individ- star population in W3 Main (e.g., Megeath et al. 1996;
ual cluster members. These invoke processes acting at Feigelson & Townsley2008) that formed throughoutthe
sub-parsec scales, among them small-scale convergence regionmightalreadyhaveinitiatedorenhancedthe pro-
of flows (Csengeri et al. 2011) in addition to turbulent cess of compression in the center, as well as contributed
cores (McKee & Tan 2003). to the formation of the first population of high-mass
The compressionof convergentfeedback mightalso be stars.
able to explain the ‘pinched’ morphology of observed Second, the simultaneous, small-scale (sub-parsec)
magnetic fields (e.g., Roberts et al. 1993; Greaves et al. triggering by high-mass stars could provide more tur-
1994;Tieftrunk et al.1995),withanenhancementofthe bulent, as well as warmer environments. The latter in
component of the field parallel to the compressed struc- particular could inhibit fragmentation by increasing the
ture (e.g., Peretto et al. 2012). minimum Jean’s mass, leading to the formation of new
massive cores (e.g., Zinnecker et al. 1993; McKee & Tan
5.1.4. The combination in W3 Main 2003;Peters et al.2010)andanincreaseincharacteristic
stellarmasstowardthemorecentralregions,asobserved
Akeyconsequenceofthe‘convergentconstructivefeed-
fortheIRS5clumpforthehighandlow-masspopulation
back’processisthatstarswouldformprogressivelycloser
(Megeath et al. 1996; Ojha et al. 2009). The combina-
to the central regions, each generation ‘aiding’ in the
tion of high efficiency of triggering and higher temper-
formationofnew high-massstars,andresultinginasys-
atures could then be key to the formation and tell-tale
tematic decrease in age toward the innermost regions of
characteristics of rich clusters of high-mass stars. The
the clump. Returning to our motivation, this behavior
final morphology of the cluster would resemble that of
is indeed observed in the high-mass stellar population
a more evolved cluster after mass segregation. Indeed,
in W3 Main (e.g., Tieftrunk et al. 1997), as well as in
from their timescale analysis of W3 Main, Ojha et al.
theenhancedconcentrationofyoungClass0starsatthe
(2009) suggested that the apparent mass segregation
boundariesoftheHIIregionswithintheW3Eastclump
must not be dynamical in origin.
(Ojha et al. 2009) suggestive of induced star formation.
Third,whenahigh-massstarformsclosetothecentral
Thus the new scenario can account for the unique stel-
(andmostdense)regionsoftheclump,thelimitedrange
lar distribution and characteristics of W3 East and W3
ofthestellarinfluence(Dale & Bonnell2011)andtheef-
West (spatial, age, and mass distributions, and multi-
ficiencyoftriggeringcouldthenleadtothemostcompact
plicity) in conjunction with the Herschel-based proper-
andrichestsystems,byformingnewoverdensities,induc-
ties. A sequential process of high-mass star formation
ing the collapse of preexisting ones, or by direct interac-
in W3 Main was already suggested in previous studies
tionbetweentheeffectsoftheembeddedhigh-massstars
(e.g., Tieftrunk et al. 1997; Feigelson & Townsley 2008:
suchasoutflowsandshocks(Phillips et al.1988). InW3
option 4 in their analysis).
East, this could account for the high star formation effi-
ciencyandmultiplicityobservedintheinnermostregions
5.2. Implications
of the clump, local to the IRS5 system (Megeath et al.
When trying to address even the most basic processes 2005;Rodo´n et al.2008). Thesub-parsecconvergenceof
ofthehigh-massstarformationprocess,currenttheoret- flowsscenariofromCsengeri et al.(2011)wouldalsoben-
ical models face several challenges such as: efit from the confined environment created by the con-
i) the low core accretion rate m∗ (and therefore the vergent constructive feedback and/or converging flows,
long formation times) due to initially low temperatures minimizing the disruptive effects.
in the case of the standard protostellar collapse model Fourth, in a scenario with multiple and simultaneous
(e.g., m∗ ∝T3/2; for isothermalcore collapsein the case triggeringbyvarioushigh-massstars,asinW3Main,the
of spherical collapse; Shu 1977); resultingcentralstructurecouldreachrare,highcolumn
ii) the suppression of accretion due to radiation pres- densitiessuitablefortheformationofcentralTrapezium-
sure and ionization by the forming high-mass star like systems.
(Zinnecker & Yorke 2007); Fifth, the continuingconfinementandinfluence by the
iii) formation in clustered environments; and high-mass stars at the outer boundaries of clumps could
iv) primordial mass segregation with anomalous age aidtheaccretionrequiredtoproduceverymassiveproto-
distributions (e.g., young central massive systems sur- starsinthecentralregion,sustainingthefeedingprocess
rounded by a cluster of older low-mass stars). bycounteractingorminimizingmasslossduetothestel-
The ‘convergent constructive feedback’ process might lar outflows, winds, and radiative energy of the newly