Characterizing Precursors to Stellar Clusters with Herschel

1.6 MB·

by  C. Battersby

Astronomy&Astrophysicsmanuscriptno.battersby˙higalirdcs˙astroph (cid:13)c ESO2011
January26,2011
⋆
Characterizing Precursors to Stellar Clusters with Herschel
C.Battersby1,J.Bally1,A.Ginsburg1,J.-P.Bernard2,3,C.Brunt4,G.A.Fuller5,P.Martin6,S.Molinari7,J.Mottram4,
N.Peretto5,8,L.Testi9,10,andM.A.Thompson11
1 CenterforAstrophysicsandSpaceAstronomy,UniversityofColorado,Boulder,CO80309,USA
2 Universite´deToulouse(UPS-OMP),InstitutdeRechercheenAstrophysiqueetPlane´tologie
3 CNRS,UMR5277,9Av.colonelRoche,BP44346,F31028Toulousecedex4,France
4 SchoolofPhysics,UniversityofExeter,StockerRoad,Exeter,EX44QL,UK
5 JodrellBankCentreforAstrophysics,SchoolofPhysicsandAstronomy,UniversityofManchester,Manchester,M139PL,UK
1 6 CanadianInstituteforTheoreticalAstrophysics,UniversityofToronto,Toronto,ONM5S3H8,Canada
1 7 INAF-Instituto,FisicaSpazioInterplanetario,viaFossodelCavaliere100,00133Roma,Italy
0 8 LaboratoireAIM,CEA/DSM-INSU/CNRS-Universite´ParisDiderot,IRFU/SApCEA-Saclay,91191Gif-sur-Yvette,France
2 9 INAF-OsservatorioAstrofisicodiArcetri,Firenze,Italy
n 10 EuropeanSouthernObservatory,GarchingbeiMuenchen,Germany
a 11 CentreforAstrophysicsResearch,ScienceandTechnologyResearchInstitute,UniversityofHertfordshire,Hatfield,UK
J
Preprintonlineversion:January26,2011
4
2
ABSTRACT

] Context.Despite their profound effect onthe universe, the formationof massive starsand stellar clusters remainselusive. Recent
A
advances in observing facilitiesand computing power have brought us closer to understanding thisformation process. Inthe past
G decade,compellingevidencehasemergedthatsuggestsInfraredDarkClouds(IRDCs)maybeprecursorstostellarclusters.However,
. the usual method for identifying IRDCsisbiased by the requirement that they are seen in absorption against background mid-IR
h emission,whereasdustcontinuumobservationsallowcold,densepre-stellar-clusterstobeidentifiedanywhere.
p Aims.WeaimtounderstandwhatphysicalpropertiescharacterizeIRDCs,toexplorethepopulationofdustcontinuumsourcesthat
- arenotmid-IR-dark,andtoroughlycharacterizethelevelofstarformationactivityindustcontinuumsources.
o
Methods. We use Hi-GAL 70 to 500 µm data to identify dust continuum sources in the ℓ=30◦ and ℓ=59◦ Hi-GAL Science
r
t Demonstration Phase (SDP) fields, to characterize and subtract the Galactic cirrus emission, and perform pixel-by-pixel modified
s
blackbodyfitsoncirrus-subtractedHi-GALsources.WeutilizearchivalSpitzerdatatoindicatethelevelofstar-formingactivityin
a
eachpixel,frommid-IR-darktomid-IR-bright.
[
Results.WepresenttemperatureandcolumndensitymapsintheHi-GALℓ=30◦andℓ=59◦SDPfields,aswellasarobustalgorithm

1 for cirrussubtraction and source identification using Hi-GAL data. Wereport on the fraction of Hi-GAL source pixels which are
v mid-IR-dark,mid-IR-neutral,ormid-IR-brightinbothfields.Wefindsignificanttrendsincolumndensityandtemperaturebetween
4 mid-IR-darkandmid-IR-bright pixels;mid-IR-darkpixelsareabout 10Kcolder andhaveafactor of 2higher columndensity on
5 average than mid-IR-bright pixels. Wefind that Hi-GAL dust continuum sources span a range of evolutionary states frompre- to
6 star-forming,andthatwarmersourcesareassociatedwithmorestarformationtracers.Additionally,thereisatrendofincreasingtem-
4 peraturewithtracertypefrommid-IR-darkatthecoldest,tooutflow/masersourcesinthemiddle,andfinallyto8and24µmbright
. sourcesatthewarmest.Finally,weidentifyfivecandidateIRDC-likesourcesonthefar-sideoftheGalaxy.Thesearecold(∼20K),
1
highcolumndensity(N(H )> 1022cm−2)cloudsidentifiedwithHi-GALwhich,despitebrightsurroundingmid-IRemission,show
0 2
littletonoabsorptionat8µm.ThesearethefirstinnerGalaxyfar-sidecandidateIRDCsofwhichtheauthorsareaware.
1
1 Keywords.ISM:clouds–ISM:dust,extinction–ISM:evolution–stars:formation–stars:pre-main-sequence
:
v
i
X
1. Introduction proto-clusters is complicated: they are rare, and therefore fur-
r ther away on average than isolated low mass star-forming re-
a Massive stars play a dominant role in shaping the Universe
gions, proto-clusters have high column densities, meaning that
through their immense ionizing radiation, winds, and spectac- the proto-stars are highly embedded, and massive stars evolve
ular explosive death, yet their formation mechanism remains
rapidlyandquicklyheatandionizetheirsurroundings,disrupt-
poorly understood. The dominant mode of star formation, and
ingtheirnatalmolecularcloud.
perhaps the only mode for massive star formation seems to be
clustered(Lada&Lada2003;deWitetal.2005).Thedefinition Searchesforproto-clustersormassiveproto-starsareusually
ofclusteredcanbecalledintoquestion(e.g.Bressertetal.2010; targetedatlong-wavelengths,asdustcontinuumemissionfrom
Gieles&PortegiesZwart2011),butthesearchforyoung,mas- thesecold,densesourcespeaksinthefar-IRtosub-mm.Surveys
sivestarformingregionsisstilldirectedtoward‘proto-clusters:’ of molecular lines are another approach for identifying poten-
cold, dense, massive molecular clumps. Direct observation of tialproto-clusters,althoughthesesurveysmaybetime-intensive
and the strength of a molecular line is complicated by exci-
⋆ Herschel in an ESA space observatory with science instruments tation conditions and optical depth. The discovery of Infrared
providedbyEuropean-ledPrincipalInvestigatorconsortiaandwithim- Dark Clouds (IRDCs Eganetal. 1998; Peraultetal. 1996;
portantparticipationbyNASA. Omontetal.2003),openedanewwindowtoviewingthesecold,
1
C.Battersbyetal.:CharacterizingPrecursorstoStellarClusterswithHerschel
dense potential proto-clusters in silhouette against the bright and 500µm using the PACS (Poglitschetal. 2010) andSPIRE
Galacticmid-IRbackground.Inthepastdecade,compellingev- (Griffinetal.2010)imagingcamerasinparallelmode.Two2◦×
idencehasemergedthatsuggeststhatsomeIRDCsmaybepre- 2◦ regions of the Hi-GAL survey were completed during the
cursorstomassivestarsandclusters(e.g.Rathborneetal.2006; ScienceDemonstrationPhase(SDP;Molinarietal.2010a),cen-
Raganetal. 2006; Beuther&Sridharan 2007; Parsonsetal. teredatapproximately[ℓ,b]=[30◦,0◦]and[59◦,0◦].
2009; Battersbyetal. 2010). While there exist many small DatareductionwascarriedoutusingtheHerschelInteractive
IRDCs (e.g. Peretto&Fuller 2010; Kauffmann&Pillai 2010), Processing Environment(HIPE, Ott 2010) with custom reduc-
the most massive ones (M ∼ 103−4 M , n > 105cm−3, T < tion scripts that deviated considerably from the standard pro-
⊙ H
25K,Rathborneetal.2006;Eganetal.1998;Careyetal.1998) cessing for PACS (Poglitschetal. 2010), and to a lesser extent
areconsistentwithexpectationsforaproto-cluster.However,the forSPIRE(Griffinetal.2010).Adetaileddescriptionoftheen-
identificationofanIRDCrequiresthatitbeonthenear-sideof tire data reduction procedure can be found in Traficanteetal.
abrightmid-IRbackground.Thislimitsourpotentialtounder- (2010).
standtheGalacticdistributionofpotentialproto-clusters. The zero-level offsets in the Herschel maps were estab-
Far-IR and submm dust continuum surveys are a powerful lished by comparison with the IRAS and Planck data at com-
waytoidentifyproto-clustersthroughouttheGalaxy,asthecold, parable wavelengths, following the same procedure as de-
dense dust is optically thin at these wavelengths.Surveyssuch scribed in Bernardetal. (2010). We compared the Herschel-
as the Bolocam Galactic Plane Survey (BGPS, Aguirreetal. SPIRE and PACS data with the predictions of a model pro-
2010) at 1.1 mm, the APEX Telescope Large Area Survey of vided by the Planck collaboration (Planck Core-Team, private
the Galaxy (ATLASGAL, Schulleretal. 2009) at 870 µm, and communication) and constrained on the Planck and IRIS data
now Hi-GAL (The Herschel Infrared Galactic Plane Survey, (Miville-Descheˆnes&Lagache 2005). The model uses the all-
Molinarietal. 2010a) from 70 to 500 µm are promising tools skydusttemperaturemapsderivedfromtheIRAS100µm and
for understanding star cluster formation on a Galactic scale. the two highest Planck frequencies to infer the average radi-
However, the sources identified in these surveys may span a ation field intensity for each pixel at the common resolution
largerangeofevolutionarystates,frompre-star-formingtostar- of the Planck and IRAS resolution of 5’. The Dustem model
forming, and further analysis or intercomparison may be nec- (Compie`gneetal. 2010) with the above value for the radiation
essary to identifythe pre-star-formingregions.Since dusttem- field intensity was then used to predictthe expectedbrightness
peratures and column densities can be derived from the multi- intheHerschel-SPIREandPACSbands,usingthenearestavail-
wavelengthHi-GALdata,thisdatasetallowsforthedistinction able Planck or IRAS band for normalization and taking into
tobemadebetweenpre-andstar-formingregions. accountthe appropriatecolorcorrectionin the Herschelfilters.
For this study, we utilize data from the Hi-GAL survey Thepredictedbrightnesswascorrelatedwiththeobservedmaps
(Molinarietal. 2010a) from 70 to 500 µm in the ℓ=30◦ and smoothed to the 5’ resolution over the region observed with
ℓ=59◦ SDP fields to characterize dust continuum sources. We Herschel and the offsets were derived from the zero intercept
investigatedifferencesinthephysicalpropertiesofmid-IR-dark ofthecorrelation.We estimatetheaccuracyoftheoffsetdeter-
and mid-IR-brightcloudsidentified in the dust continuum,and minationtobetterthan5%.
also the physical properties of sources associated with vari-
ous star formation tracers. We use Extended Green Objects
2.2.ArchivalData
(EGOs,alsoknownas“greenfuzzies”,Cyganowskietal.2008;
Chambersetal. 2009) to trace outflows from young stars, We utilize the wealth of archivaldata in the Galactic Plane for
CH3OH masers to trace sites of massive star formation, and 8 ouranalysis.Weusethemid-IRdatatakenaspartoftheGalactic
and24µmemissiontoindicateanaccretingproto-starorUCHII Legacy Infrared Mid-Plane Survey Extraordinaire (GLIMPSE;
Region(Battersbyetal.2010). 3.6, 4.5, 5.8, and 8.0 µm, Benjaminetal. 2003) and the
This paper is organized as follows. In Section 2 we intro- MIPSGALsurvey(24µm;Careyetal.2009).Wealsomakeuse
duce the Hi-GAL observing strategy and discuss archival data of the IRDC catalog of Peretto&Fuller (2009), the catalog of
usedinouranalysis.InSection3wedescribetheGalacticcirrus Extended Green Objects (Cyganowskietal. 2008, EGOs), and
emission removal and source identification methods, the mod- the 1.1 mm data and catalog from the Bolocam Galactic Plane
ified blackbody fitting procedure, and how the star formation Survey (BGPS; Aguirreetal. 2010; Rosolowskyetal. 2010).
tracers were incorporated. Section 4 describes our results, in- The Multi-Array Galactic Plane Imaging Survey (MAGPIS;
cludingadiscussionofuncertainties,thepropertiesofthecirrus Helfandetal.2006;Whiteetal.2005),whichprovidescompre-
cloud emission, the temperature and column density maps, the hensive radio continuum maps of the first Galactic quadrantat
association of Hi-GAL sources with mid-IR-dark and mid-IR- highresolutionandsensitivity,isalsousedinouranalysis.
brightsources,andstarformationtracersinHi-GALsources.In
Section5wepresentfivecandidateIRDC-likecloudsonthefar-
side of the Galaxy, and finally, in Section 6 we summarize our 3. Methods
conclusions.
3.1.SourceDefinitionandRemovaloftheGalacticCirrus
The Hi-GAL data reveal a wealth of structure in the Galactic
2. ObservationsandArchivalData Plane, from cirrus clouds (Martinetal. 2010) to filaments and
clumps,asdiscussedinMolinarietal.(2010b).Figure1demon-
2.1.Hi-GAL
strates the complicated association of mid-IR and dust contin-
The Herschel Infrared Galactic Plane Survey, Hi-GAL uumsourcestowardtheGalacticRing.Inthispaper,weexplore
(Molinarietal. 2010a), is an Open Time Key Project of the thephysicalpropertiesofthedensestcomponentsoftheGalactic
HerschelSpaceObservatory(Pilbrattetal.2010).Hi-GALwill Plane;the potentialprecursorsto massive stars and clusters. In
performa5-bandphotometricsurveyoftheGalacticPlaneina orderto properlycharacterizethesedense objects,a carefulre-
|b| ≤ 1◦ -wide strip from -70◦ ≤ l ≤ 70◦ at 70, 160, 250, 350, movaloftheGalacticcirrusisrequired.Wehaveexploredava-
2
C.Battersbyetal.:CharacterizingPrecursorstoStellarClusterswithHerschel
Fig.1.Athree-colorGLIMPSEimagewithlogarithmicBGPS1.1mmdustcontinuumcontoursonourtestregionintheHi-GAL
SDPℓ=30◦field.Red:8µm,Green:4.5µm,andBlue:3.6µm.ThisfiguredemonstratesthedistinctionbetweentheIRDCanddust
continuumpopulation.Somedustcontinuumsourcesaremid-IR-bright,somearemid-IR-dark,andsomehavenoIRcorrelation.
The objectin the center is particularlywell-suited asa test case forthis study asit containsbothmid-IR-brightandmid-IR-dark
dustcontinuumsourcesradiatingasfilamentsoneithersideofayoungHIIregioncomplex.
riety of methodsforthe removalofthe cirrusemission and we method works well when applied to a single IRDC, as it was
brieflydiscussthe pros,cons,andsystematiceffectsof thedif- usedinPerettoetal.(2010),itisnotrobustenoughtobeusedto
ferentsubtractionmethods. createafullGalacticcirrusemissionmapfromhigh(|b|&1◦)to
Our firststep is to projectallthe data ontoa commongrid, low(|b|=0◦)Galacticlatitudes.Wethentriedathirdmethodto
with a common resolution and common units for comparison. createthedensesourcemasksusingasimplecontrastmap(con-
Wecroptheimagestotheuseablesciencefield,convertthedata trast=data-background)oftheSPIRE500µmdataandapplied
to units of MJy/sr, Gaussian convolve to a common resolution a cutoff. This method improved over the previous, however, it
(36 ′′ and 25 ′′ for the lower and higher resolution modified suffered from large negative bowls, creating source masks that
blackbodyfits), andregridtoacommongridwithareasonable weremuchsmallerthanthephysicalsourcesizes.Inthefourth
pixel size (roughly 1/4 the Θ ). These images are used in method,wedefinethebackgroundfirstbyasecond-orderpoly-
FWHM
theremainderoftheanalysis.Weincludeimagesthroughoutthe nomial plane fit (along Galactic latitude) to the smoothed 500
paperofa “testfield,” (approximately,[ℓ,b]= [29.95,-0.35]to µm image. This plane fit to the Galactic cirrus was then sub-
[30.50, 0]) chosen for it’s diversity of mid-IR-bright and mid- tracted from the original data and a cutoff was applied (deter-
IR-darksources. The analysiswas orginallyrun and optimized mined by eye) to define the dense source masks. This was the
onthisregionandthenexpandedtothefullHi-GALfields. first method to do a reasonable job of identifying sources in a
range of Galactic latitudes. We found, however, that a polyno-
The first method for determining the dense clump source
mialplanewasnotaverygoodapproximationtotheshapeofthe
masks was to use Bolocat clump masks derived from the
Galacticcirrusemission,andthatthefitwasparticularlypoorat
Bolocam Galactic Plane Survey (BGPS Aguirreetal. 2010;
highGalacticlatitudes(|b|∼0.8−1◦ ).
Rosolowskyetal. 2010) at 1.1 mm. While these masks are a
robust tracer of the cold, dense gas, they only extend to |b| ≤ Figure2depictsthefinalmethodusedtoidentifythedense
0.5◦inthemajorityofoursciencefieldandthesensitivityinthe sourcesandtoseparatethosefromthecirruscloudemission.The
ℓ=59◦fieldistoopoortotracetheclumpsseeninHi-GAL. firstpanel(a)showstheoriginalSPIRE500µmimage.Thefirst
The following methods all use the SPIRE 500 µm data step (shown in panel b) is to convolve the 500 µm image with
to determine the source masks, as the cirrus decreases to- a Gaussian thatis largeenoughto smoothoverthe sourcesbut
wardslongerwavelengths(Gautieretal.1992)andPerettoetal. small enough to capture variations in the cirrus emission. We
(2010) demonstrate that the SPIRE 500 µm data is well-suited decided through trial and error that a Gaussian with a FWHM
forcirrus/densesourcedistinction.Thesecondmethodwetried of12’wasareasonablecompromise.We thenfitaGaussianin
wasbasedonthatofPerettoetal.(2010),whofoundthattoward latitudetoeachGalacticlongitude,asshowninFigure3andin
IRDCs,theSPIRE500µmfluxdistributionshowedaclearpeak panel(c)ofFigure2.AGaussianisareasonableapproximation
at low fluxes with a long tail toward higher fluxes. While this forthevariationintheGalacticcirrusacrosstheGalacticPlane
3
C.Battersbyetal.:CharacterizingPrecursorstoStellarClusterswithHerschel
andworkedtoidentifysourcesathighandlowGalacticlatitudes
equallywell.WethensubtracttheGaussianfitapproximationof Sr] 800 l = 30.51o
the Galactic cirrus from the original 500 µm data to achieve a y/
“differenceimage.”This“differenceimage”isthefirstguessat MJ 600
thecirrus-subtractedsourcemap. ce [
Sli 400
We apply a cutoff to this first guess cirrus-subtracted g
n
sourcemap,suchthateverythingabovethecutoffisconsidered Alo 200
‘source.’ We determine this cutoff by fitting a Gaussian to the x
u
histogramof pixelvalues(asshown in Figure4). The negative Fl 0
fluxvaluesinthedistributionofthis“differenceimage”arerep- −1.0 −0.5 0.0 0.5 1.0
Galactic Latitude
resentativeoftherandomfluctuationsinthedata,so wemirror
thenegativefluxvaluesaboutzero,fitaGaussiantothatdistri-
250
bution (representative of the noise in the map) and then apply
acutoffof4.25σtothe“differenceimage.”We calculatedthis Sr] 200 l = 30.51o
y/
cutoff in quarterσ intervalsfrom3 to 6 σ, a range overwhich MJ 150
it grows smoothly. The choice of 4.25 σ was selected because x [
u 100
ℓth=a5t9c◦ufitoeffldsbewstherenpirnesspeencteteddtbhye esoyue.rcWeshiilnebthoethchthoeicℓe=o3f0c◦uatonffd dual Fl 50
isimportantindeterminingthefinalphysicalproperties,thereis esi 0
R
nothing special about this cutoff; the properties vary smoothly −50
aboveandbelowthisvalue. −1.0 −0.5 0.0 0.5 1.0
Galactic Latitude
Oncethesourcecutoffhasbeendetermined,thesourcesare
masked out in the original 500 µm image (as shown in solid Fig.3. A slice across the Galactic Plane at l = 30.51◦ of the
whiteinpanel(d)ofFigure2)andweconvolve(withaGaussian
SPIRE 500 µm image, demonstrating the cirrus emission re-
of FWHM 12’) the data outside the source masks, the cirrus
moval. The top panel shows the SPIRE 500 µm data, with a
emission,treatingdatainthesourcemasksasmissing.Thiscre-
Gaussian fit to the smoothed background on the first iteration
ates a second guess at the cirrus emission, equivalent to panel
showninasolidgrayline.Thedashedgraylineistheconvolved
(b).We thenrepeatthe processof fittinga Gaussianin latitude
backgroundonthefirstiteration,afterthesourcemaskhasbeen
tothisimage(asinpanel(c)),determiningasourcecutoffbyfit-
applied.TheblacksolidlineistheGaussianfittothesmoothed
tinga Gaussiantothemirrorednegativefluxdistributioninthe
background on the final iteration and the black dashed line is
“differenceimage,” applyingthese masks (panel(d)),and con-
the finalconvolvedbackgroundafterthe source maskhas been
volvingaroundthemasks(panel(e))toproducethenextguess
applied. Note that the final background(dashed black line) fits
atthecirrusemission(panel(b)).Weiterateonthisprocessuntil
the low-lyingdiffuse emission nicely,includingit’s asymmetry
the sourcemasks converge.Figure 5 showsthe convergenceof about b = 0◦ . The bottom panel shows the final background
thesourcecutoffin bothfields. We choseiteration16asthe fi-
subtractedscienceimagecutalongthesameGalacticlongitude,
nalsourcemaskcutoffasitwasrepresentativeoftheconverged
with the final source cutoff drawn as a dashed black line at 65
valuearoundwhichtheiterationsvaried,thoughtheexactchoice
MJy/sr. The success of the background subtraction method is
doesnotmattermuch,asthecutoffvariesverylittleafteritera-
demonstratedbythefactthatthestructureisflatacrossGalactic
tion 10in bothfields. Inpanel(d)of Figure2 the white masks
latitude and that the cutoff allows for the identification of the
arethefirstiterationsourcemasksandthewhitecontourshows significant,thoughfaint,sourcenearb=1◦.
the final (16th) iteration source masks. Once the source masks
aredetermined,theyareappliedtoeachwavelengthimage,and
that image is convolved (with a Gaussian FWHM 12’ and the
mask pixels ignored) to create the cirrus image at that wave- Galactic latitudes covered by the data and the range of source
length.Subtractingthecirrusimageateachwavelengthfromthe confusionandactivity observed.The sensitivityof thismethod
original data produces the source images used in the modified isdependentuponthefinalmaskcutofforconfusion,ratherthan
blackbodyfits, whilethe smoothcirrusimagesareusedforthe thesensitivityofHi-GAL.Fortheℓ=30◦finalmaskcutoffof65
modifiedblackbodyfitstothecirrusemission. MJy/sr, we are sensitive above the cirrus background to a col-
The top panel of figure 3 shows a slice of 500 µm data umndensityofN(H2)=2.9×1021cm−2for20KdustorN(H2)
through Galactic latitude. The final convolved background = 9.7 × 1020 cm−2 for 40 K dust. In the ℓ=59◦ field the final
(dashed black line) is representative of the low-lying emis- maskcutoffisabout24MJy/sr,whichgivesasensitivityabove
sion across Galactic latitude, including it’s asymmetry about thecirrusbackgroundtoacolumndensityofN(H2)=1.1×1021
b=0◦.Noticeinthebottompanelthatthestructureisflatacross cm−2for20KdustorN(H2)=3.6×1020cm−2for40Kdust.
Galactic latitude and allows for the identification of the source
near b=1◦ . Figure 4 shows that as the iterations converge,we
3.2.ModifiedBlackbodyFits
movefluxfromnegativebowlsbackto sourcesaspositivefea-
tures and the mirrored negative distribution then more closely Weperformedpixel-by-pixelmodifiedblackbodyfitstoboththe
resemblesa Gaussian(theblacklineisthe finalGaussian fitto cirrus cloud and the dense source clump emission. The dense
iteration16). source pixel-by-pixel fits were first performed on the images
Thismethodhasbeenfine-tunedbyeyetoreproducethesig- convolvedto36”resolution.Followingthat,theresultsfromthe
nificant structure picked out by a human, but is entirely auto- 36”resolutiontestswereusedasinputtopixel-by-pixelfitsper-
matedandwasrunintheexactsamewayforboththeℓ=30◦and formed on the images convolved to 25” resolution. While the
ℓ=59◦ fields, despite their vast differences.We have foundthat fitstothe25”resolutionimagesformallyutilizelessdatapoints,
this method faithfully identifies sources across the range of theyallowforamoredetailedcomparison.
4
C.Battersbyetal.:CharacterizingPrecursorstoStellarClusterswithHerschel
0.1 (a) (b) (c)
e
d 0.0
u
atit −0.1
L
c
cti −0.2
a
al −0.3
G
−0.4
0.1 (d) (e) (f)
e
d 0.0
u
atit −0.1
L
c
cti −0.2
a
al −0.3
G
−0.4
30.4 30.2 30.0 29.8 30.4 30.2 30.0 29.8 30.4 30.2 30.0 29.8
Galactic Longitude Galactic Longitude Galactic Longitude
Fig.2.Adepictionofthecirrussubtractionmethodforthefirstiteration.Panel(a)istheoriginalSPIRE500µmimage(allimages
onthesamelinearreversegrayscale,from-10to900MJy/sr).Panel(b)isthesmoothed(FWHMGaussianconvolutionkernelof
12’)SPIRE500µmimagethatisusedtofittheGaussianshowninpanel(c).Panel(c)isthensubtractedfrompanel(a)toproduce
a contrast image. A 4.25 σ cutoff is then applied to the contrast image to produce the source masks shown in panel (d). Panel
(d) is the original SPIRE 500 µm image with the sources masked out in white. This is the first iteration source masks; the final
sourceboundariesareshownaswhitecontours.Panel(d)isthenconvolvedwithaGaussianFWHMconvolutionkernelof12’with
the maskstreatedas missing data to producepanel(e). Panel(e)is consideredthe cirrusimagefor the nextiteration.In the next
iteration,aGaussianwillbefittothecirrusimage,asafunctionoflatitudeateachlongitude,andstepsb-ewillrepeat.Panel(f)
isthedifferenceimage(originalSPIRE500µmimage-convolvedcirrusimage)thatwilleventuallybeusedforgreybodyfittingin
thefinaliteration.
Weusethemodifiedblackbodyexpressionintheform 1022 cm−2,andβfrom1to3.WedonotusetheBGPS1.1mm
point in the fits because uncertainties in both the absolute cal-
Sν = c2(e2khhTνν−3 1)(1−e−τν) (1) ifbroramtiosingnainfidcathnetlysphaetlipailnfiglttoercfounnsctrtaioinntphreevfietnatttthheisBtiGmPe.S point
Wethenfitamodifiedblackbodyateachpositionwithinthe
where
sourcemasksofthecirrus-subtractedscienceimages(e.g.Figure
τ =µ m κ N(H ) (2) 2panel(f))tothedenseclumps.Forthesefitsofdenseclumps,
ν H2 H ν 2 we ignore the 70 µm point since the optically thin assumption
where µH2 is the mean molecularweight for which we adopta may not be valid (τ70µm = 1 at N(H2) = 1.2 × 1023 cm−2 us-
value of µH2 = 2.8 (Kauffmannetal. 2008), mH is the mass of ingEquation2).Infact,IRDCsoftenappearinabsorptionat70
hydrogen,N(H2)isthecolumndensity,andκν isthedustopac- µm indicating that in these very high column density regimes,
ity. We determine the dust opacity as a continuous function of the 70 µm data is not optically thin and that it can no longer
frequencybyfittingapower-lawoftheformκν = κ0(νν0)β tothe be modelled by a single dust temperature. Additionally, it has
tabulatedOssenkopf&Henning(1994)dustopacitiesintherel- been found (e.g. Desertetal. 1990; ?) that a large fraction of
evantrangeof frequencies.We used the Ossenkopf&Henning thefluxat70µmislikelycontaminatedbyemissionfromvery
(1994) MRN distribution with thin ice mantles that have coag- small grains whose temperature fluctuations with time are not
ulatedat 106 cm−3 for105 yearsmodelfor dustopacity,which representativeoftheequilibriumtemperatureofthelargegrains.
isareasonableguessforthesecold,denseclumps.Wehaveas- Withoutthe70µmpointwehaveonlyfourpoints(usuallyjust
sumedagastodustratioof100,whichyieldedaκ of4.0atν onthe Rayleigh-Jeansslope)to constrainthe fit, so we fix β to
0 0
of505GHzwithaβ=1.75. 1.75(TheOssenkopf&Henning1994,fitvalueforanMRNdis-
We first fit a modified blackbody to each pixel in the cir- tributionwiththinicemantlesthathavecoagulatedat106 cm−3
rusimage(e.g.Figure2panel(e))usingMPFITFUN(fromthe for 105 years; a reasonable guess for these environments),and
Markwardt IDL Libarary, Markwardt 2009). We assign a cali- leave only the temperature and the column density as free pa-
brationuncertaintyof20%totheSPIREandPACSdatapoints rameters.Wediscusshowavalueofβof1.5or2wouldalterthe
and use the covariancematrix returnedby MPFITFUN to esti- resultsinSection4.1.Weacknowledgethatperceivedchangesin
mateerrorsintheparameters.Inthemodifiedblackbodyfit,we temperaturemayactuallybechangesinβanddiscussthismore
leavethecolumndensity,β,andthetemperatureasfreeparame- alsoinSection4.1.Weare,toanextent,lookingforchangesin
ters.Thetemperaturewasrestrictedtorangebetween0and100 the dust properties (β or temperature) across different environ-
K, the column density was free to range between 0 and 100 × ments,whichwecanstillachieve.
5
C.Battersbyetal.:CharacterizingPrecursorstoStellarClusterswithHerschel
(the SPIRE 500 µm pointmust be excludedbecause of resolu-
105 Iteration 1 tion), they allow for a more detailed comparison.The physical
Iteration 2 propertiesreturnedbythefitsatdifferentresolutionsagreewell.
Iteration 3
n Iteration 4 The final temperature and column density maps (as shown in
ux bi IItteerraattiioonn 812 Figure 6) show the derived temperatures and column densities
h fl IItteerraattiioonn 1260 from the 25” map within the source masks, and the smoothed
eac 104 cirrus-background temperatures and column densities outside
n thesourcemasks.
er i We notehere thatstar-formingregionscontainstructureon
b
m many scales, and that fitting a single temperature and column
u
N density over a region of 25” (about 0.5 pc at a distance of 4
kpc)isavastoversimplification.Therecertainlyexistlargevari-
103 ations in the temperature and column density in these sources
−100 −50 0 50 100
on smaller scales. In this paper, we present the beam-diluted
Difference Image Flux [MJy/Sr]
average source properties on 25” scales. The column densities
presented here are therefore characteristic of the larger-scale
Fig.4. The flux distribution in the difference images (original
clump structure and under-estimates of the peak column den-
500µmimage-convolvedcirrusimage,e.g.Figure2panel(f))
sities,whilethetemperatureswillbeunderestimatesinveryhot
for 8 iterations from 1 to 20 in the ℓ=30◦ field. In each itera-
regions(hotcores)andoverestimatesinverycoldregions(star-
tion, the negative flux distribution was mirrored about 0 and a
lessIRDCcores).
Gaussian was fit to the distribution to determine the character-
istic fluctuations so that a source identification cutoff (of 4.25
σ) could be applied. Plotted here are the full distributions, not 3.3.StarFormationTracerLabelMaps
themirroreddistributionsusedtofittheGaussian.Asweiterate,
Inordertorobustlycomparethephysicalpropertiesdetermined
pointsin the negativeend are transferredto the positive endof
inthispaperwithvariousstarformationtracers,wehavecreated
thedistributionasfluxis restoredin thenegativebowlsaround
star formation tracer label maps. These maps are on the same
brightsources. The Gaussian fit to the distribution forthe final
grid as the science images (temperature, column density, etc.)
iteration (16) is shown as the black curve. While the final dis-
and have a binary denotion in each pixel: 1 if the star forma-
tributionisnotperfectlyGaussian,theiterationsshowgreatim-
tiontracerispresentand0ifitisnot.We createstarformation
provement,and a Gaussian that is characteristic of the random
tracer label maps for the MIPS 24 µm emission, 8 µm emis-
fluctuationsinthedistribution.Wenotethatthedistributiondoes
sion, ExtendedGreen Objects (EGOs;Cyganowskietal. 2008;
notpeakatzerobecausetheaveragemustbezeroandthereisa
Chambersetal. 2009), and 6.7 GHz methanol maser emission
largepositivewinginthisdistribution.
(Pestalozzietal.2005).TheMIPS24µmimagesrequiredstitch-
ing together using MONTAGE before the label map could be
created.Sincetheℓ=30◦ fieldismorepopulated,confused,and
120
therefore has a wider range of star formation activity, we only
100 perform the analysis in this field. The results of this compari-
Sr] l=30o soncanbefoundinSection4.5.The8µmemissionlabelmaps,
y/ 80 however, were created and analyzed for both the ℓ=30◦ and
J
M Final Mask Cutoff ℓ=59◦ fields so that we could understand the relationship be-
off [ 60 tween mid-IR and Hi-GAL sources in both fields (see Section
ut 4.4).
C
k 40 The GLIMPSE 8 µm band shows emission from warm
as l=59o
M Final Mask Cutoff dust and Polycyclic Aromatic Hydrocarbons (PAHs). Regions
20 with bright 8 µm emission may contain warm, diffuse dust or
UV-excited PAH emission from an Ultra-Compact HII Region
0
(Battersbyetal. 2010).The absorptionof8 µm emissionbyan
5 10 15 20
Iteration Infrared Dark Cloud (IRDC) is indicative of a high column of
colddustobscuringthebrightmid-IRbackground.Wenotethat
Fig.5.The 4.25σ mask cutoffover20iterations.The cutoffis mid-IR-brightregionsaregenerallyindicativeofactivestarfor-
from the Gaussian fit (see Figure 4) to the flux distribution of mation,andthatthepeakofmid-IRlightcanbeoffsetfromthe
pointsinthedifferenceimage(e.gFigure2panel(f)).Boththe youngestregionsofstarformation(e.g.Beutheretal.2007).We
ℓ=30◦ andℓ=59◦ fields convergeafter severaliterations,but to denoteregionsofbright8µmemissionasmid-IR-bright(mIRb),
verydifferentvalues.Wechoseiteration16asthefinalcutoffas darkregionsofabsorptionat8µmasmid-IR-dark(mIRd),and
itishighenoughtobeassuredofconvergenceandisrepresenta- regionswithoutstarkemissionorabsorptionat8µmasmid-IR-
tiveofthevaluetowhichtheiterationsconverge. neutral (mIRn). The 8 µm emission label map has both a neg-
ative(IRDC,mIRd),positive(mIRb),andneutral(mIRn)com-
ponent. This map is created by first making a contrast image.
These pixel-by-pixel modified blackbody fits to the cirrus- The GLIMPSE 8 µm image is convolved with a Gaussian of
subtracteddenseclumpswerefirstperformedontheimagescon- FWHM5’torepresentthesmooth,slowlyvaryingbackground.
volved to 36” resolution. Following that, the results from the 5’ was determined(see Peretto&Fuller 2009, for more details
36”resolutiontestswereusedasinputtopixel-by-pixelfitsper- on this smoothingkernelchoice) to be largeenoughto capture
formed on the images convolved to 25” resolution. While the the largest IRDCs and mid-IR-bright objects, while still small
fitstothe25”resolutionimagesformallyutilizelessdatapoints enoughtocapturethebackgroundvariations.Thisimageisthen
6
C.Battersbyetal.:CharacterizingPrecursorstoStellarClusterswithHerschel
Temperature [K] Column Density [x1022 cm−2]
10 20 30 40 50 0 1 2
0.1
0.0
e
d
u
atit −0.1
L
c
acti −0.2
al
G
−0.3
−0.4
30.4 30.2 30.0 29.8 30.4 30.2 30.0 29.8
Galactic Longitude Galactic Longitude
Fig.6. Temperature (left) and Column Density (right) maps in the ℓ=30◦ field test region. In these maps, the backgroundcirrus
emissiontemperatureandcolumndensitiesareplottedoutsideofthesourcemasks,whilethebackground-subtractedtemperature
andcolumndensitiesareplottedinsidethesourcemasks.Thesemapsareat25”resolutionandassumeβ=1.75insidethesource
masks.
mIRn.We relyonoureyestopickoutregionsofemissionand
absorption,andhaveselectedthecutoffsthatbestrepresentavi-
sualinterpretationofthatwhichisdarkandbright,asshownin
Figure7.
A portion of this label map is shown in Figure 7. This
methoddoesaverygoodjobofpickingoutthebrightanddark
regions, and agrees reasonably well (considering the different
methods, resolution used, and cutoffs) with the IRDC catalog
of Peretto&Fuller (2009). Two important biases to note with
thismethodare:1) the contrastimage techniquecreates‘nega-
tivebowls’aroundbrightobjects,andthereforethesizesofthe
mIRdandmIRbregionsareunderestimated,and2)convolving
thecontrastimageto25”resolutioncausesustomisssomesmall
IRDCs.Additionally,anydecrementsinthebackgroundwillbe
denoted as being mIRd. This issue is partly resolved when we
multiplythislabelmapwithoursourcemasks,sothatourclas-
sificationsonlyapplywhereaHi-GALsourcealsoexists.
Fig.7.GLIMPSE8µmimageoverlaidwithcontoursoftheclas-
Emission at 24 µm observed using the MIPS instrument
sification as mid-IR-bright (mIRb, dark blue) or mid-IR-dark
traces warm dust. MIPS 24 µm emission can trace warm, dif-
(mIRd,cyan).The classification is only appliedwherea far-IR
fusedustintheISMorwarmdustassociatedwithstarformation
sourceexists,asdenotedbytheblackcontours.
such as material accreting onto formingstars, the gravitational
contractionofayoungstellarobject,orwarmdustsurrounding
anewlyformedstar. TheMIPS24µm emissionlabelmapwas
subtractedfromtheoriginalimage,whichisthendividedbythe created using the same method as the 8 µm label maps, with
smoothed background. The resulting ‘contrast map’ enhances the exception that we only include a positive contrast cutoff.
thecontrastofmIRbandmIRdcloudsabovethebrightmid-IR WhilesomeIRDCsremaindarkat24µm,thenegativecontrast
background.Toobtainafaircomparisonwiththelowerresolu- isveryminimalwhenconvolvedto25”resolution.Additionally,
tion(∼25”)scienceimages,thiscontrastimageisconvolvedto the highest negative contrast in the contrast image is from the
thesame resolution.We thenapplyacutoffsuchthatanypixel negative bowls around bright sources, so identifying IRDCs in
with a positivecontrastgreaterthan orequalto 10%is consid- the24µm imagesusingthismethodisnotadequate.Theposi-
ered mIRb, and any pixel with a negativecontrast greater than tivecontrastthreshholdfortheMIPS24µmcontrastmapswas
or equal to 5% is considered mIRd, while all others are then 25%,also chosenby eye.A highercutoffthanat8 µm wasre-
7
C.Battersbyetal.:CharacterizingPrecursorstoStellarClusterswithHerschel
Table1.EGOsUsedinThisPaper
EGOName R.A.(J2000) Decl.(J2000) lxbSize(”) Ref.a
G29.23-0.05 18:44:51.4 -03:18:41.7 14x12 1
G29.25-0.69 18:47:10.7 -03:35:42.0 14x14 1
G29.84-0.47 18:47:28.8 -02:58:03.0 9x9 2
G29.91-0.81 18:48:47.6 -03:03:31.1 8x8 1
G29.96-0.79 18:48:49.8 -03:00:28.7 32x32 2
G30.40-0.30 18:47:52.6 -02:23:12.2 17x17 1
G30.42-0.23 18:47:40.8 -02:20:31.6 12x12 1
G30.60+0.18 18:46:33.7 -01:59:30.3 14x14 1
G30.79+0.20 18:46:48.2 -01:48:54.4 12x12 1
G30.90-0.01 18:47:45.9 -01:49:00.0 14x7 1
Notes. (a) 1 - This work (see Section 3.3 and Figure 8), 2 -
Cyganowskietal.(2008)
EGO,andeachpixelinthisboxwasgivenaflagof1,whilefor
pixelsoutsidetheboxtheflagis0.
Finally,weemploythepresenceorabsenceof6.7GHzClass
II CH OH masers as a star formation tracer. We use the com-
3
pletecatalogcompiledbyPestalozzietal.(2005)toidentifyun-
biasedsearchesfor6.7GHzmethanolmasersintheℓ=30◦field.
While there havebeen numeroustargetedsearchesforCH OH
3
masers, we utilize the unbiased Galactic Plane searches of
Szymczaketal. (2002) and Ellingsenetal. (1996), on the 32m
Torun Radio Telescope (FWHM of 5.5’) and the University of
Tasmania 26m Radio Telescope (FWHM of 7’), respectively.
6.7 GHz methanol masers are found to be exclusively associ-
ated with massive star formation (Minieretal. 2003), and are
often offset from the radio continuum emission indicative of
an UCHII Region (e.g. Walshetal. 1998; Minieretal. 2001)
and are thought to represent an earlier stage of massive star
formation. We chose to include only the unbiased catalogs of
Ellingsenetal. (1996) and Szymczaketal. (2002) so as not to
choosemIRbregionspreferentially.Thesecatalogstogethergive
22 methanol maser regions, which we use to create a label
map where the “size” of the methanol maser regions is given
Fig.8. GLIMPSE 3-color images (Red: 8 µm, Green: 4.5 µm,
by 2× the positional accuracy RMS of the observations (30”
Blue:3.6µm)ofEGOsintheℓ=30◦ field.TheEGOsshownin
and36”forSzymczaketal.2002;Ellingsenetal.1996,respec-
therightcolumnofthe3rdand4throwwereidentifiedas“pos-
tively). Methanol maser emission comes from very small ar-
sible” EGOs by Cyganowskietal. (2008) and the others were
easonthesky(e.g.Walshetal.1998;Minieretal.2001),how-
identifiedbytwoindependentviewersasEGOsasdescribedin
ever,methanolmasersareoftenclustered(Szymczaketal.2002;
Section3.3.ThepositionsarelistedinTable1.
Ellingsen1996, find multiplemaser spotstowardsthe majority
ofsources),sothese“sourcelabelsizes”aremeanttoberepre-
sentativeoftheextentofthestar-formingregion.
quiredtoselectthe24µmsourcesabovethebrightbackground.
Thesecutoffsgreatlyaffecttheresultingtrends.SeeSection4.1
4. Results
foradiscussionoftheeffectsofcutoffsonsourceidentification.
We include the presence of Extended Green Objects We discuss here the results of the modified blackbody fits, the
(EGOs, also called “green fuzzies” Cyganowskietal. 2008; propertiesoftheGalacticcirrusemission,acomparisonofmIRb
Chambersetal. 2009) as a star formation tracer. EGOs are and mIRd pixels, and the association with star formation trac-
regions of enhanced and extended 4.5 µm emission and are ers.WediscusswhatdefinesthepopulationsofmIRbandmIRd
thoughttobeindicativeofshocksinoutflows(Cyganowskietal. sources, and the objects which have no mid-IR association.
2009). Cyganowskietal. (2008) catalogued EGOs throughout Before the discussion of the results, we first mentionthe many
the Galaxy in the GLIMPSE fields. There are two “possible” caveatsanduncertaintiesthataccompanythesefindings.
EGOs in the ℓ=30◦ field which we include in our EGO la-
belmap.WesupplementedtheCyganowskietal.(2008)catalog
4.1.ARemarkonCaveatsandUncertainties
with an independentvisual searchof the ℓ=30◦ field forEGOs
that may have been missed. An independent, cross-checked A major uncertainty in this analysis is the modified black-
search identified eight further candidates, which we include in body fits. Firstly, star-formingregionsshow structure on many
our analysis. A list of the locations of the EGOs used in this scales, and so fitting a single column density and temperature
analysisisgiveninTable1,andtheyareshowninFigure8.For to any point does not adequately represent the whole region;
eachEGO,aboxwasdrawnaroundtheregionthatcontainsthe ratheritservesasawaytocategorizeanddifferentiatebetween
8
C.Battersbyetal.:CharacterizingPrecursorstoStellarClusterswithHerschel
GLIMPSE 8µm Sr] 10000 N(H2): 0.2 T+e/−m p0:.1 4 x6 1+P0/−o22 s1c.m4 1 −K2 N(H2): 2.2 T+e/−m p0:.7 2 x0 1+P0/−o22 s2c.m K6−2
−0.10 Flux [MJy/ 1000
de 100
Galactic Latitu −−00..2105 87654321 Flux [MJy/Sr] 101000000 N(H2): 0.2 T+e/−m p0:.1 4 x4 1+P0/−o22 s1c.m3 2 −K2 N(H2): 1.1 T+e/−m p0:.4 1 x9 1+P0/−o22 s2c.m K7−2
9 100
10
N(H): 0.4 +/− 0.2 x 1022 cm−2 N(H): 0.7 +/− 0.2 x 1022 cm−2
−0.25 Sr] 10000 2 Temp: 37 +P/−o s8. K3 2 Temp: 21 +P/−o s2. K8
30.30 30.G25alactic L3o0n.g2i0tude 30.15 30.10 Flux [MJy/ 1000
100
ure [K] 4500 1 2 3 4 TNe(mH2p) 122...5050 cm]22−2 Flux [MJy/Sr] 101001000000 N(H2): 0.6 T+e/−m p0:.3 3 x6 1+P0/−o22 s8c.m K4−2 N(H2): 0.6 T+e/−m p0:.2 2 x1 1+P0/−o22 s2c.m K9−2
Temperat 2300 5 6 7 8 9 10 01..50N(H) [x 12 Flux [MJy/Sr] 101000000 N(H2): 2.2 T+e/−m p0:.8 2 x5 1+P0/−o22 s4c.m K5−2 N(H2): 0.6 T+e/−m p0:.2 1 x9 1+P0/−o22 s2c.m K1−02
10 0.0 100
0 100 200 300 400 50 100 200 500 50 100 200 500
Wavelength [µm] Wavelength [µm]
Distance along filament ["]
Fig.9.TopLeft:GLIMPSE8µmimageinourℓ=30◦testregion.Right:Modifiedblackbodyfitstothepointsmarkedonthefilament
inthetopleftpanel.Sincethe70µmpointisnotincludedinthesefits(seeSection3.2)itisplottedasanXwhiletheotherpoints
are shown with their 20% calibration error bars. The errors quoted in temperature and column density are the formal fit errors.
As discussedin Section 4.1, these are notnecessarilyrepresentativeof the trueerrors.BottomLeft: Thetemperatureandcolumn
densitiesderivedalongthefilament,with10pointsmarked.Notetheinverserelationbetweentemperatureandcolumndensityin
thisexample,especiallyfrompoints8to10.
bulk, large-scale physical properties of regions. Additionally, dependent not only on the background that is subtracted, but
this means that the quantities derived are likely to be highly alsoonpossibledeparturesbetweenthegainscaleofthePACS
beam-diluted.Thetruepeakcolumndensityinanygivenregion and the SPIRE data. Improvementsto the general issue of the
isalmostcertainlyhigherthanthebeam-averagedcolumnden- Planck-Herschel cross calibration with respect to the accuracy
sityreportedhere.Likewise,thebeam-averagedtemperaturesre- ofthe offsetdeterminationaswellasto thegainvalueswillbe
portedforcoldobjectsareoverestimatesofthetruetemperature adressedinthefuture.
minimum, while the beam-averaged temperatures reported for
Theassumptionofafixedβcanbeconsideredtocontribute
hotobjects is an under-estimateof the peak. These caveats ex-
thebulkoftheuncertaintyinthetemperaturedetermination.The
plainthesomewhatsmalldynamicrangeobservedinbothtem-
degeneracyofβ-ThasbeendiscussedinParadisetal.(2010).If
peratureandcolumndensity.
we adopta valueof β= 1.5(insteadof 1.75),the temperatures
Inaddition,theparametersderivedfromthemodifiedblack- are about 4K higher on average and the column densities are
body fits themselves are rather uncertain. For most of the aboutthesame.Ifweadoptavalueofβ=2.0(insteadof1.75),
sources, the temperature is higher than 18 K, so all the points thetemperaturesareabout3Kloweronaverage,andagain,the
usedinthefitlie nearthepeakoftheSED orontheRayleigh- columndensitiesareaboutthesame.Withonlyfourdatapoints
Jeans slope. This equates to large uncertainties in the fits, es- toconstrainthefits,fixingβisnearlyessential,withthechoiceof
peciallyathightemperatures.Themedianerrorsinthe derived 1.75beingareasonableguessfortheseenvironments.Variations
temperatureareabout4and3K(fortheℓ=30◦andℓ=59◦fields, inβfrom1.75to1.5or2correspondtochangesintemperature
respectively), while the median errors in the derived column ofabout3or4K.Thereisthepossibilitythatsystematicchanges
density are about 0.12 and 0.04 × 1022 cm−2 (in ℓ=30◦ and inβ,forexampleasystematicdecreaseinβtowardthecenterof
ℓ=59◦ fields, respectively). The error is a strong function of IRDCs due to dust coagulation (Ossenkopf&Henning 1994),
the parameter value, with a median temperature error of 13% could systematically change our results. In the dark centers of
and 18% and median column density error of 44% and 48% IRDCs, however,a variationinβ onlychangesthe temperature
(in the ℓ=59◦ and ℓ=30◦ field, respectively). These uncertain- byabout2K(2Khotterforβ=1.5,and2Kcolderforβ=2.0),
tiesarebasedonthe20%calibrationuncertaintyassignedtothe so while thisis an importanteffectto consider,itis unlikelyto
Hi-GAL fluxes. Additionally, the derived quantities are highly negatetheobservedtrendsintemperature.
9
C.Battersbyetal.:CharacterizingPrecursorstoStellarClusterswithHerschel
Fig.10.Top:Temperature(onalinearscalefrom0to50K)andBottom:ColumnDensity(onalogscalefromN(H )=0to5×1022
2
cm−2)intheℓ=30◦(left)andℓ=59◦(right)fields.Inthesemaps,thebackgroundcirrusemissiontemperatureandcolumndensities
areplottedoutsideofthesourcemasks,while thebackground-subtractedtemperatureandcolumndensitiesareplottedinsidethe
sourcemasks.Thesemapsareat25”resolutionandassumeβ=1.75insidethesourcemasks.
The association with star formation tracers is also bridled inthecenterofFigure9highlightstheimportanceofcomparing
with uncertainty.Each of the tracers includedare not perfectly valuesona pixel-by-pixelbasis,ratherthandenotingtheentire
representedwithourmethods.The8and24µmlabelmapsare objectaseithermIRbormIRd.
entirely dependenton the cutoffs chosen. While we have done
ourbesttochoosereasonablecutoffs,a changein thesecutoffs
wouldhaveasignificanteffectonthepropertiesinferred.Theas- 4.2.PropertiesoftheCirrusCloudEmission
sociationwithEGOssuffersfromasimilarbias.Wehavemade
Whileweprimarilyconsiderthecirruscloudemissionas“back-
a careful attempt to identify probable EGO candidates, but the
ground”inthispaper,wereportheresomeusefulphysicalprop-
objectsthemselvesaresomewhatill-defined,soagain,weresort
erties derived in this analysis. These maps were smoothed to
totheby-eyeclassification.TheassociationwithCH OHmasers
3 12’ (Θ ) resolution, so we report only the average prop-
isrobustinthatitisbasedonuniformsensitivityblindsurveys, FWHM
ertiesin Table 2. Thesedata suggestthatdustin diffuse clouds
however,thesesurveyssufferfromlargebeamsizesandconfu-
in the molecular ring (ℓ=30◦ field) is warmer, has higher col-
sionwhichconsequentlydilutetheirtruephysicalproperties.All
umndensities,andasteeperspectralindex(β)thandustindif-
ofthese tracersalso sufferfromtheinherentscatterthatcomes
fuse clouds toward the mid-outer Galaxy, outside the molecu-
alongwithvaryinglevelsofextinctionanddifferentdistances.
lar ring (ℓ=59◦ field). In the cirrus background determination,
Despitethesecaveats,thetrendsobservedovermanypixels we fit a Gaussian of the form: g = B + Ae(b−b0)2/2σ2 across
remain robust. The pixel-by-pixel comparison is an important GalacticlatitudeateachGalacticlongitude.We reportherethe
technique for achieving the best possible associations with the Gaussian fit to the median value across Galactic longitudes.In
givendata, and therefore,the mostrobuststatistics. The object theℓ=30◦ field,thebestfitparametersareA=285MJy/sr,B=
10