The Luminosities of Protostars in the Spitzer c2d and Gould Belt Legacy Clouds

0.35 MB·English

by  Michael M. Dunham

Accepted for Publicationin AJ
PreprinttypesetusingLATEXstyleemulateapjv.08/22/09
THE LUMINOSITIES OF PROTOSTARS IN THE SPITZER C2D AND GOULD BELT LEGACY
CLOUDS
Michael M. Dunham1,2, H´ector G. Arce1, Lori E. Allen3, Neal J. Evans II4, Hannah Broekhoven-Fiene5,
Nicholas L. Chapman6, Lucas A. Cieza7, Robert A. Gutermuth8, Paul M. Harvey4, Jennifer Hatchell9, Tracy
L. Huard10, Jason M. Kirk11, Brenda C. Matthews5, Bruno Mer´ın12, Jennifer F. Miller10,13, Dawn
E. Peterson14, & Loredana Spezzi15
Accepted forPublication inAJ
3
1 ABSTRACT
0
Motivatedbythelong-standing“luminosityproblem”inlow-massstarformationwherebyprotostars
2
are underluminous compared to theoretical expectations, we identify 230 protostars in 18 molecular

n clouds observed by two Spitzer Space Telescope Legacy surveys of nearby star-forming regions. We
a compile complete spectral energy distributions, calculate L for each source, and study the proto-
bol
J stellar luminosity distribution. This distribution extends over three orders of magnitude, from 0.01

8 L⊙ – 69 L⊙, and has a mean and median of 4.3 L⊙ and 1.3 L⊙, respectively. The distributions are
1 very similar for Class 0 and Class I sources except for an excess of low luminosity (L . 0.5 L )
bol ⊙
Class I sources compared to Class 0. 100 out of the 230 protostars (43%) lack any available data in

] the far-infrared and submillimeter (70 µm < λ < 850 µm) and have L underestimated by factors
A bol
of 2.5 on average, and up to factors of 8−10 in extreme cases. Correcting these underestimates for
G eachsourceindividuallyonceadditionaldatabecomesavailablewilllikelyincreaseboththemeanand
. medianofthe sampleby35%–40%. We discussandcompareourresultstoseveralrecenttheoretical
h studies of protostellar luminosities and show that our new results do not invalidate the conclusions
p
of any of these studies. As these studies demonstrate that there is more than one plausible accretion
-
o scenariothatcanmatchobservations,futureattentionisclearlyneeded. Thebetterstatisticsprovided
r by our increased dataset should aid such future work.
t
s Subjectheadings: stars: formation-stars: low-mass-stars: luminosityfunction,massfunction-stars:
a protostars
[

1 1. INTRODUCTION to refer to the hydrostatic object at the center of a col-
v
lapsing core. More evolvedyoungstellar objects (YSOs)
6 Low-massstarsformfromthe gravitationalcollapseof
nolongerembeddedwithinandformingfromtheirnatal
1 dense molecular cloud cores of gas and dust (e.g., Be-
dense cores are not considered protostars by this defini-
4 ichman et al. 1986; Di Francesco et al. 2007). During
tion.
4 the collapseprocessmaterialaccretesfromthe coreonto
Despite several decades of progress, many details re-
. the protostar. In this paper the term protostar is used
1 lating to the accretionof materialfrom dense cores onto
0 1Department of Astronomy, Yale University, P.O. Box 208101, protostars remain poorly understood. As mass accretes
3 NewHaven,CT06520, USA onto protostars the gravitationalenergy is liberated and
1 [email protected] radiated away as accretion luminosity. This luminosity,
v: 34NDaeptiaorntamleOntptoifcaAlsAtrsotnroonmoym,yThOebUsenrivvaetrosriiteys,ofTTucesxoans,aAtZA,uUstSiAn, which depends on the mass accretion rate, current pro-
i 2515Speedway, StopC1400, Austin,TX78712-1205, USA tostellar mass, and current protostellar radius, can be
X 5HerzbergInstitute,NationalResearchCouncilofCanada,5071 usedtostudythe massaccretionprocessanddistinguish
r W.SaanichRoad,Victoria,BCV9E2E7,Canada between different accretion models.
a 6Center for Interdisciplinary Exploration and Research in As-
Observational studies of protostellar luminosities are
trophysics (CIERA), Department of Physics & Astronomy, 2145
SheridanRoad,Evanston,IL60208, USA hinderedbythefactthatprotostarsaredeeplyembedded
7Institute forAstronomy,UniversityofHawaiiatManoa,Hon- indensecores,withmostoftheiremittedluminositiesre-
olulu,HI96822,USA processed to mid-infrared, far-infrared, and submillime-
8Department of Astronomy, University of Massachusetts,
terwavelengthsbythedustinthecores. Thefirstsignifi-
Amherst,MA,USA
9AstrophysicsGroup,Physics,UniversityofExeter,ExeterEX4 cantstudyoftheprotostellarluminositydistributionwas
4QL,UK presented in a series of papers by Kenyon et al. (1990,
10Department of Astronomy, University of Maryland, College 1994) and Kenyon & Hartmann (1995). They identi-
Park,MD20742, USA
11School ofPhysicsandAstronomy,CardiffUniversity,Queens fied 23 protostars in the Taurus-Auriga molecular cloud
Buildings,TheParade,Cardiff,CF243AA,UK andcalculatedbolometricluminositiesbyintegratingthe
12HerschelScienceCentre,ESAC-ESA,P.O.Box78,28691Vil- observed spectral energy distributions (SEDs) using In-
lanuevadelaCan˜ada, Madrid,Spain frared Astronomical Satellite (IRAS) 12−100 µm pho-
13HarvardSmithsonian Center for Astrophysics, 60 Garden
tometryandlonger-wavelength(sub)millimeter photom-
Street,Cambridge,MA02138,USA
14SpaceScienceInstitute,4750WalnutStreet,Suite205,Boul- etry from the ground, when available. They found that
der,CO80301 the protostellar luminosity distribution extended from
15European Southern Observatory (ESO), Karl-Schwarzschild- 0.09–22L ,withameanandmedianof2.3L and0.7
Strasse2,D-85748GarchingbeiMu¨nchen,Germany ⊙ ⊙
L , respectively, and a strong peak around 0.3 L .
⊙ ⊙
2
As first noted by Kenyon et al. (1990), their observed plete SEDs including far-infrared and (sub)millimeter
protostellar luminosities are lower than expected from photometry from the literature, and integrating these
simple theoretical predictions. Their argument is as fol- SEDs to determine L . They found a total of 112
bol
lows. First,theyassumedthatallobservedluminosityis protostars in the seven c2d clouds. Enoch et al. cal-
accretion luminosity, culated mean and median values similar to those found
byKenyon&Hartmann(1995),butwiththeirimproved
GMM˙ samplestatisticstheynotedthepresenceofalargerfrac-
acc
L =f , (1)
acc acc R tion of sources at low luminosities (.1.0 L⊙). Evans et
al.(2009)includedacorrectionforforegroundextinction
wherefacc isanefficiencyfactortakentobe1, M andR andcalculatedrevisedmeanandmedianvaluesof5.3L⊙
arethemassandradiusoftheprotostar,andM˙ isthe and 1.5 L . In a separate study, Kryukova et al. (2012)
acc ⊙
accretion rate onto the protostar. By further assuming alsoderivedtheprotostellarluminositydistributionfora
that the peak of the luminosity distribution is produced number ofstar-formingclouds,including the c2dclouds.
by low-mass stars with M = 0.1 M and R = 1 R , Theyfoundanevenlargerexcessoflow-luminosityproto-
⊙ ⊙
theycalculatedanimpliedmassaccretionrateof∼10−7 starsthan foundbyEnochetal.andEvanset al. Offner
M yr−1. If some fraction of the observed luminosity & McKee (2011) argued that the higher observed lumi-
⊙
arises from the protostar itself (contraction, deuterium nosities found when extinction corrections are applied,
burning, etc.), the implied mass accretion rate is even combinedwithamorerealisticvalueoftheefficiencyfac-
lower. torinEquation1 off ∼0.5 totake intoaccountboth
acc
In the simplest model, the collapse of a singular the poweringofjets andwinds andthe effects ofunseen,
isothermal sphere initially at rest as first considered episodic accretion bursts, can essentially resolve the lu-
by Shu (1977) and later extended by Terebey, Shu, & minosityproblem,althoughexplainingthe largefraction
Cassen(1984)toincluderotation(oftencalledthe“stan- of sources at very low luminosities remains a challenge.
dard model” of star formation), collapse proceeds in an Several recent theoretical studies have explored possi-
“inside-out”fashion, beginning in the center of the core, bleresolutionstotheluminosityproblem,manyofwhich
moving outward at the sound speed, and giving rise to were originally proposed by Kenyon et al. (1990). One
a constant mass accretion rate of M˙ ∼ 2×10−6 M possibility is that accretion is variable or episodic, with
acc ⊙
yr−1 for 10 K gas. This is over ten times higher than prolonged periods of low accretion punctuated by short
inferred by Kenyon et al. (1990), and will only scale up- bursts of rapid accretion. Numerous origins for such a
wardas M˙ ∝T3/2 for higher gas temperatures. Mod- process have been proposed, including gravitational in-
acc
stabilities in protostellar disks (e.g., Vorobyov & Basu
ifications to the standard model, including non-zero ini-
2005, 2006, 2010; Machida et al. 2011; Cha & Nayak-
tial inwardmotions (Larson1969;Penston1969;Hunter
shin 2011),a combinationof gravitationalandmagneto-
1977; Fatuzzo, Adams, & Myers 2004), magnetic fields
rotational instabilities in protostellar disks (e.g., Ar-
(Galli & Shu 1993a, 1993b; Li & Shu 1997; Basu 1997),
mitageetal.2001;Zhuetal.2009a,2009b,2010),quasi-
and isothermalspheres that are not singular but feature
periodic magnetically driven outflows in the envelope
flatteneddensity profilesatsmallradii(Foster& Cheva-
(Tassis & Mouschovias 2005), decay and regrowth of
lier 1993; Henriksen, Andr´e, & Bontemps 1997) all tend
MRI turbulence (Simon et al. 2011), close interaction
to increase the accretion rate over that predicted by the
in binary systems or in dense stellar clusters (Bonnell &
standard model, making reconciliation between theory
Bastien 1992; Pfalzner et al. 2008), and disk-planet in-
and the Kenyon et al. observations difficult. This has
teractions (Lodato & Clarke 2004; Nayakshin & Lodato
become known as the “luminosity problem.”
2011). Indeed, Dunham & Vorobyov(2012)showedthat
Identification of protostars and determining their lu-
the L distribution predicted by the Vorobyov & Basu
minosities were both greatly improved by the launch bol
(2005,2006,2010)simulations,whichfeaturehighlyvari-
of the Spitzer Space Telescope (Werner et al. 2004) in
able accretion with episodic bursts, provides a reason-
2003. Manysitesofstarformationhavebeenobservedat
able match to the c2d observations presented by Evans
wavelengthsrangingfrom3.6to160µmthroughvarious
etal. (2009). Alternatively,Offner & McKee (2011)pre-
Spitzer surveys. One suchsurveywas the Legacysurvey
sented analytic derivations of the protostellar luminos-
“FromMolecularCores to PlanetForming Disks” (here-
ity function for several different accretion scenarios and
after c2d; Evans et al. 2003), which observed 7 large,
showed that accretion models that tend toward a con-
nearby molecular clouds and ∼ 100 isolated dense cores
stantaccretiontimeratherthanaconstantaccretionrate
and resulted in the discovery of very low luminosity ob-
jects (VeLLOs), protostars with internal luminosities16 provide a good match to the Evans et al. c2d observa-
tions. As a third alternative, Dalba & Stahler (2012)
≤ 0.1 L embedded in dense cores (Young et al. 2004).
⊙
recently argued that external accretion onto collapsing
Dunham et al. (2008) identified 15 VeLLOs in the c2d
coresfromthesurroundingbackgroundcloudwillreduce
dataset, and detailed studies of several have confirmed
accretion rates and luminosities.
their very low luminosities and status as embedded pro-
With 112 protostars spread over more than three or-
tostars (Dunham et al. 2006; Bourke et al. 2006; Lee et
dersofmagnitude inL , the c2dsampleofprotostellar
al. 2009; Dunham et al. 2010b; Kauffmann et al. 2011). bol
luminosities is still somewhat limited by small number
BothEnochetal.(2009)andEvansetal.(2009)stud-
statistics. As a follow-up to c2d, the Spitzer Gould Belt
ied the c2d protostellar luminosity distribution by using
Legacy Survey (hereafter GB; L. Allen et al. 2012, in
the Spitzer data to identify protostars, compiling com-
preparation) observed most of the remaining clouds in
16 Theinternalluminosity,Lint,istheluminosityofthecentral the Gould Belt. In this paper we extend the identifica-
sourceandexcludes luminosityarisingfromexternalheating. tionof protostarsandcalculations ofLbol fromEvans et
3
TABLE 1
Molecular Clouds Surveyedbythe c2dandGBSurveys
Distance Distance
Cloud Survey (pc) Referencea DataReference(s)b
Aquila GB 260 Mauryetal.(2011) Gutermuthetal.(2008); Mauryetal.(2011)
Auriga/California GB 450 Ladaetal.(2009) H.Broekhoven-Fiene etal.(2012, inpreparation)
Cepheus GB 200–325c Kirketal.(2009) Kirketal.(2009)
ChamaeleonI GB 150 Bellocheetal.(2011a)
ChamaeleonII c2d 178 Whittet etal.(1997) Y·o·u·ngetal.(2005); Porrasetal.(2007); Alcal´aetal.(2008)
ChamaeleonIII GB 150 Bellocheetal.(2011a)
CoronaAustralis GB 130 Neuha¨user&Forbrich(2008) P·e·t·ersonetal.(2011)
IC5146 GB 950 Harveyetal.(2008) Harveyetal.(2008)
LupusI c2d 150 Comer´on(2008) Chapmanetal.(2007); Mer´ınetal.(2008)
LupusIII c2d 200 Comer´on(2008) Chapmanetal.(2007); Mer´ınetal.(2008)
LupusIV c2d 150 Comer´on(2008) Chapmanetal.(2007); Mer´ınetal.(2008)
LupusV GB 150 Comer´on(2008) Spezzi etal.(2011)
LupusVI GB 150 Comer´on(2008) Spezzi etal.(2011)
Musca GB 160 Knude&Hog(1998) T.Huardetal.(2012, inpreparation)
Ophiuchus c2d 125 deGeus etal.(1989) Padgett etal.(2008)
OphiuchusNorth GB 130 Wilkingetal.(2008) Hatchell etal.(2012)
Perseus c2d 250 Enochetal.(2006) Jørgensenetal.(2006); Rebulletal.(2007)
Serpens c2d 429 Dzibetal.(2010, 2011) Harveyetal.(2006,2007a, 2007b)
a
Referenceforthedistancequotedinthiswork.
b ReferencespresentingtheSpitzer IRACandMIPSobservations.
c
DifferentregionswithinCepheusarelocatedatdifferentdistances;seeKirketal.(2009)fordetails.
al.(2009)tothecombinedc2d+GBdataset. Ourworkis obtained 3.6–8.0 µm images with the Spitzer Infrared
motivated by a desire for better underlying statistics in ArrayCamera(IRAC; Fazio etal. 2004)and24–160µm
theobservedprotostellarluminositydistributionandim- imageswiththe Multiband ImagingPhotometer(MIPS:
proving the accuracy of the L calculations by includ- Rieke et al. 2004) of all 18 clouds. A standard pipeline
bol
ing additional data not yet available when the Evans et developedbyc2dwasusedfordatareduction,sourceex-
al. study was conducted. The organizationof this paper traction, and band-merging to produce final source cat-
is as follows: We describe our method in §2, including alogs for both surveys and has been described in detail
overviews of the c2d and GB surveys in §2.1, the iden- elsewhere (Harvey et al. 2006; Evans et al. 2007).
tification of protostars in §2.2, the compilation of full Table 1 lists each cloud, the survey in which it was
source SEDs in §2.3, and the calculation of L in §2.4. imaged (c2d or GB), the assumed distance to the cloud,
bol
§3 summarizes our basic results. A discussion of these the reference for the distance, and references of individ-
results is contained in §4. In particular, in §4.1 we com- ual studies of each cloud where the observation strategy
pareourresultstotheexistingc2d(§4.1.1)andKryukova and basic results are presented. These clouds were cho-
et al. (2012) (§4.1.2) results, in §4.2 we discuss several sen to represent nearly all of the significant sites of star
recenttheoreticalinvestigationsofprotostellarluminosi- formationwithin the Gould Belt, with two major excep-
ties, and in §4.3 we evaluate the accuracy of our L tions: the Taurus and Orion molecular clouds. These
bol
measurements for sources with observed SEDs that are two clouds were each the focus of separate, dedicated
not well sampled in the far-infrared and submillimeter, Spitzer Legacy surveys led by other groups, and fold-
andtheeffectsofthisincompletesamplingonouroverall ing in their results with the c2d+GB clouds will be the
results. Finally,weoutlineimportantfutureworkneeded focus of a future paper. The clouds listed in Table 1
to further advance this topic in §5, and summarize our span very large ranges of properties. For example, the
findings in §6. total cloud masses range from a few hundred M (e.g.,
⊙
Chamaeleon II; Evans et al. 2009) to ∼ 105 M (Au-
2. METHOD ⊙
riga/California Molecular Cloud; Lada et al. 2009), the
2.1. Overview of the Surveys starformationratesandstarformationratesurfaceden-
TheSpitzer c2dsurvey(PI:N.J.Evans)conductedan sities both span approximately two orders of magnitude
imaging survey of seven large, nearby molecular clouds (Evans et al. 2009; Heiderman et al. 2010), and the ra-
andabout100isolatedmolecularcloudcores,andaspec- tio of protostars to pre-main sequence stars (indicative
troscopic survey of selected targets. The science ques- of the amount of current star formation still on-going
tionsmotivatingthissurveyandasummaryoftheobser- in the cloud) range from none (e.g., Lupus V and VI;
vationstrategyare givenby Evans etal. (2003). A sum- Spezzi et al. 2011) to values in excess of 30% (e.g., Au-
mary of the results from the survey of the large molec- riga/California Molecular Cloud, Cepheus, IC5146, and
ular clouds is given by Evans et al. (2009). The Spitzer Perseus;H.Broekhoven-Fieneetal.2012,inpreparation;
GB survey (PI: L. E. Allen) was designed as a follow-up Kirk et al. 2009; Harvey et al. 2008; Evans et al. 2009).
to the clouds portion of c2d and conducted an imaging We referthe readerto the individualcloudstudies listed
survey of 11 nearby molecular clouds, completing most in Table 1 for more details and additional references.
of the remaining clouds in the Gould Belt (L. Allen et We caution that the distances to the 18 clouds sur-
al. 2012, in preparation; see also Gutermuth et al. 2008; veyed are not all well-known, and some cloud distances
Harveyetal.2008;Kirketal.2009;Petersonetal.2011; are still under significant debate. One such example is
Spezzietal.2011;Hatchelletal.2012). Thetwosurveys thedebateoverthedistance(s)totheSerpensandAquila
4
regions. Recent VLBA parallax measurements led to a the data required to identify and remove such objects in
65%increaseinthedistancetoSerpenscomparedtothat the other clouds. Oliveira et al. (2009)found a contami-
assumed by the c2d team (429 vs. 260 pc; Straiˇzys et nation rate of 25% in their Serpens study. Serpens (and
al.1996;Harveyet al.2006;Dzib etal. 2010,2011),and Aquila)arelikelytheworstcasesduetotheircloseprox-
there remains debate whether or not Aquila is also lo- imity to the Galactic plane (spanning Galactic latitudes
catedatthisnew,fartherdistanceorevenifallofAquila ranging from 2◦ to 10◦), although Romero et al. (2012)
is itself locatedat the same distance (e.g.,Gutermuth et recently suggested the contamination rate is at least as
al. 2008; Maury et al. 2011). We do not list formal dis- high in other clouds as well, and Hatchell et al. (2012)
tance uncertainties in Table 1 as such uncertainties are found that 27% of their sample of candidate YSOs in
verypoorlycharacterizedinatleastsomeclouds. Instead OphiuchusNorthselectedviathec2dcriteriawerelikely
we refer to the references listed in Table 1 for detailed to be background giants based on proper motion argu-
discussions on the various methods used to derive dis- ments. However, 80% of the contaminants identified by
tances and the uncertainties in these methods. Future Oliveira et al. and 75% of the contaminants identified
distance revisions will require future revisions to the re- by Hatchell et al. are classified as Class III YSOs, thus
sults presented in this study. even if the overall contamination rate is as high as 25%
– 30%, our inability to remove these contaminants will
2.2. Sample Selection
not significantly affect this study since it is only focused
Our method for selecting protostars from the c2d and on the subset of YSOs that are considered to be proto-
GB observations closely follows the selection method stars. Finally, a few known YSOs missing from the list
used by Evans et al. (2009)for the c2d clouds. We sum- of candidate YSOs due to missing photometry at one or
marize the mainpoints here andrefer to Evanset al. for more Spitzer wavelengths caused by saturation or non-
more details. detections from being too deeply embedded were added
The data reduction pipeline creates band-merged by hand.
sourcecatalogsincorporating2MASS andSpitzer 1.25– Theaboveprocessresultedinafinallistof2966YSOs
70µmphotometryforeachcloud. Candidateyoungstel- (since all 2966 sources passed visual inspection, we have
lar objects (YSOs) are identified using a standard clas- followedthe terminologyused by Evanset al.[2009]and
sification method developed for the Spitzer c2d and GB dropped the word “candidate” at this point). This is
projects. ThismethodisdescribedindetailinHarveyet nearly a factor of three increase over the 1024 YSOs
al. (2007b) and Evans et al. (2007) and summarized in identified in the c2d clouds alone by Evans et al. (2009).
allofthepublicationspresentingindividualcloudstudies Many of these YSO populations have already been pre-
listed in Table 1. Briefly, this method uses the Spitzer sented and discussed in detailed studies of individual
SWIRE Legacy survey of the ELAIS N1 extragalactic clouds (see Table 1 for references) and in an analysis
field (Lonsdale et al. 2003), processed to simulate the of the star formation rates and efficiencies of the c2d
sensitivity and extinction distribution of the clouds in and GB clouds based on a preliminary version of the fi-
the c2d and GB surveys, to determine the positions of nal YSO catalog (Heiderman et al. 2010). A complete
galaxies in three different Spitzer color-magnitude dia- analysisofthefullYSOpopulation,implicationsforstar
grams. Each source extracted in the c2d and GB cloud formation rates and efficiencies in the Gould Belt, and
catalogswithinfraredcolorsindicativeofthepresenceof the evolution and lifetimes of YSOs will be presented in
dust (sources with colors that can not be explained by a forthcoming paper (L. Allen et al. 2012, in prepara-
extincted background stars) is then assigned an unnor- tion). Here we focus only on the observed luminosities
malized“probability”ofbeingagalaxyorYSObasedon of protostars.
itspositionineachcolor-magnitudediagram,itsK−[4.5] Thefinalsampleofprotostarsisidentifiedfromthelist
color,whetheritwasfoundtobeextendedineitherofthe of 2966 YSOs by examining the full SEDs compiled for
two shortest Spitzer IRAC bands (3.6 and 4.5 µm), and each source (see below) and selecting only those sources
its flux density at 24 and 70 µm. The color and magni- associatedwithat leastone (sub)millimeter detectionat
tude boundaries, along with the final boundary between λ≥350µm,resultinginafinalsampleof230protostars.
candidate YSO and candidate galaxy in unnormalized This is identical to the procedure followed by Evans et
“probability”,are set to provide a nearly complete elim- al. (2009) except they used a cutoff wavelength of 850
ination of SWIRE sources. We refer the reader to Har- µm; we modified this to 350 µm because of the large
veyetal.(2007b)forfurtherdetailsonthisclassification increase in available data at this wavelength. No intrin-
method. Similar classification methods have been pre- sic protostellar colors were assumed and no additional
sented by other Spitzer studies of galactic star-forming color criteria were imposed. This decision is motivated
regions (e.g., Gutermuth et al. 2009; Rebull et al. 2010; by numerous recent studies that have used dust radia-
Kryukova et al. 2012). tive transfer models to show that protostars observed
In total, we identified 3239 candidate YSOs in the throughoutflowcavitiescanresemblemoreevolvedClass
18 c2d and GB catalogs. All sources were visually in- II or Class III sources in the infrared (e.g., Whitney
spected to remove residual contaminants, including re- et al. 2003; Robitaille et al. 2006; Crapsi et al. 2008;
solved galaxies misclassified as candidate YSOs and im- Dunham et al. 2010a). By selecting all sources asso-
ageartifactsidentifiedaspoint-sourcesbytheautomated ciated with (sub)millimeter detections we recover such
pipelinebutlackingtruepoint-sourcedetectionsinoneor sources and identify all YSOs that are associated with
morebands(seeEvansetal.2009fordetails). Follow-up dense cores, although future follow-up observations are
optical spectroscopy of targets in Serpens presented by required to remove true Class II or III sources simply
Oliveiraetal.(2009)ledtotheidentificationandremoval seen in projection against a dense core.
of 11 background giants with infrared excesses. We lack By requiring a (sub)millimeter detection, our method
5
requirestheavailabilityof(sub)millimetersurveyscover- Ophiuchus, Perseus, and Serpens), plus a partial survey
ingthefullextentsofthecloudssurveyedbyc2dandGB. ofAquilaandpiecemealcoverageofothercloudsfromthe
This is not always the case, as described in more detail SCUBALegacyCatalog(DiFrancescoetal.2008). This
in the next section below. The effects of this limitation incomplete (sub)millimeter coverage will affect both our
will be discussed in detail in §4.1.2, where we compare luminositycalculationsandabilitytoidentifyprotostars,
to a recent study that used very different methods for and these effects are discussed in detail in §4.3 and §5.1.
selecting protostars and did not require (sub)millimeter Finally, before using the SEDs to calculate bolometric
detections. luminosities, we correct the photometry for foreground
extinction. We wish to only correct for the foreground
2.3. Constructing Full SEDs and Correcting for cloud extinction and not the local extinction from the
Extinction densecoreitself, asinthe lattercasethe extinctedemis-
sionisreprocessedtolongerwavelengthsandincludedin
Similar to Evans et al. (2009), we compiled as com-
our observed SEDs. Determining the true line-of-sight
plete SEDs as possible for each of the 2966 YSOs. In
extinction to a protostar from the foreground cloud is
addition to the 2MASS and Spitzer 1.25–70 µm pho-
not a trivial task. Following Evans et al. (2009), we as-
tometry provided by the source catalogs, we included
sign extinction values to all 2966 YSOs (a sample which
the following: (1) optical photometry, where available
includes the 230 protostars identified in this work) as
from the literature, (2) Wide-field Infrared Survey Ex-
follows:
plorer (WISE) 12 and 22 µm photometry from the all-
sky catalog17, (3) selected other ground-based optical 1. We adopt extinction values from the literature for
and infrared data as compiled by the authors of the de- Class II and III YSOs (classified via infrared spec-
tailed studies of individual clouds (see references in Ta- tral index; see Evans et al. 2009 for details) in-
ble 1), (4) Spitzer 160 µm photometry for sources de- cluded in published optical studies.
tected and not located in saturated or confused regions,
calculated using aperture photometry and aperture cor- 2. We de-reddenthe remainingClassII andIII YSOs
rections as given by the MIPS Instrument Handbook18; to the intrinsic near-infrared colors of an assumed
(5) SHARC-II19 350 µm photometry, when available, spectral type of K7, found to be fairly representa-
from a targeted survey of protostellar sources (Wu et tive of the majority of Class II and III YSOs in
al. 2007; M. M. Dunham et al. 2012, in preparation); the c2d clouds (Oliveira et al. 2009, 2010; see also
(6) SCUBA20 450 and 850 µm photometry, when avail- Evans et al. 2009 for details).
able, from the SCUBA Legacy Catalog (Di Francesco et
al.2008);and(7)other(sub)millimeterphotometryfrom 3. We de-redden all of the Class I and Flat spectrum
unbiased surveys of molecular clouds, where available. YSOs (again classified via infrared spectral index)
For the last item above, other (sub)millimeter pho- in each cloud using the mean extinction towardall
tometry from unbiased surveys of molecular clouds, we Class II YSOs in that cloud.
used photometry from the following surveys: (1) A
The extinction values adoptedfor eachof our230proto-
MAMBO221 1.2 mm survey of part of Aquila (Maury
starsfollowingthisprocedurearelistedinTable4. Most
et al. 2011); (2) A LABOCA22 870 µm survey of
ofthe protostarsinagivencloudhavethesameadopted
Chamaeleon I (Belloche et al. 2011a); (3) A SIMBA23
extinction value since most protostars are classified as
1.2mmsurveyofChamaeleonII(Youngetal.2005);(4)
ClassI orflatspectrumviatheirinfraredspectralindex,
ALABOCA870µmsurveyofChamaeleonIII (Belloche
althoughsomehavedifferentvaluessincenointrinsicpro-
et al. 2011b); (5) A Bolocam 24 1.1 mm survey of Ophi- tostellarcolorswereassumedbyourselectioncriteriaand
uchus(Youngetal.2006);(6)ABolocam1.1mmsurvey thus some Class II YSOs are classifiedas protostars(see
of Perseus (Enoch et al. 2006); and (7) A Bolocam 1.1 §2.2 above).
mm survey of Serpens (Enoch et al. 2007). Once the extinction values are assigned, we use these
Summarizingtheaboveinformation,wehaveaccessto values combined with the Weingartner & Draine (2001)
complete(sub)millimetersurveysforonly6outofthe18 extinctionlawforR =5.5tocorrectthephotometryfor
V
clouds (Chamaeleon I, Chamaeleon II, Chamaeleon III, extinction. The choice of the R = 5.5 law rather than
V
theR =3.1lawismotivatedbyseveralstudiesshowing
17Availableathttp://irsa.ipac.caltech.edu/cgi-bin/Gator/nph-scan?mtishsaiotn=tVhiresafo&rsmubemriits=mSeolercet&apprporjoshporrita=tWefIoSrEthedenseregions
18Availableathttp://irsa.ipac.caltech.edu/data/SPITZER/docs/
in which stars form (e.g., Chapman et al. 2009). While
mips/mipsinstrumenthandbook/
19 The Submillimeter High Angular Resolution Camera II wedocautionthatourapproachissomewhatcrude,itis
(SHARC-II)isa350µmbolometerarrayoperatedattheCaltech the best that can currently be done and is significantly
SubmillimeterObservatory(Dowelletal.2003). morereliablethanignoringthe effects ofextinctionalto-
20TheSubmillimeterCommon-UserBolometerArray(SCUBA) gether.
wasa450and850µmbolometerarrayoperatedattheJamesClerk
MaxwellTelescope.
21 The Max-Planck Millimeter Bolometer 2 (MAMBO2) was a 2.4. Calculation of Evolutionary Indicators
1.2mmbolometerarrayoperated attheIRAM30-mtelescope.
22TheLargeApexBolometerCamera(LABOCA)isan870µm OncewehaveconstructedfullSEDsasdescribedabove
in §2.3, we use these SEDs to calculate the bolometric
bolometer array in operation at the Atacama Pathfinder Experi-
menttelescope (Siringoetal.2009). luminosities (Lbol) and bolometric temperatures (Tbol).
23TheSESTImagingBolometerArraywasa1.2mmbolometer L is calculated by integrating over all detections,
bol
arrayinoperationattheSwedish-ESOSubmillimeterTelescope.
24Bolocamisa1.1and2.1mmbolometerarrayoperatedatthe ∞
CaltechSubmillimeterObservatory(Glennetal.1998) Lbol =4πd2Z Sνdν . (2)
0
6
Thebolometrictemperatureisdefinedtobethetemper-
ature of a blackbody with the same flux-weighted mean
frequencyasthe source(Myers&Ladd1993). Following 0.20
Myers & Ladd, T is calculated as
bol
∞
νS dν
Tbol =1.25×10−11 RR00∞Sννdν K . (3) er Bin 0.15
T can be thought of as a protostellar equivalent of P
Tebfofl for stars; Tbol starts at very low values (∼ 10 K) on 0.10
for cold, starless cores and eventually increases to Teff cti
a
once the core and disk have fully dissipated. The inte- r
F
grals defined in Equations 2 and 3 are calculated using
0.05
the trapezoid rule to integrate over the finitely sampled
SEDs. To avoid model or fitting uncertainties and focus
only on the observations themselves, we do not extrapo-
latebeyondtheshortestandlongestfrequencesatwhich 0.00
dataareavailableandwedonotinterpolateovermissing --22 --11 00 11 22
data. Instead, we explore the effects of missing data on log Lbol (LO •)
our L calculations in §4.3. We calculate L and T
bol bol bol
twice, once with the original, observed photometry and Fig.1.— Histogram showing the distribution of extinction cor-
once with the extinction-corrected photometry. rectedLbolforall230protostarsinlogspace. Thebinsare1/3dex
wide, and the error bars show the statistical (√N) uncertainties.
The solid vertical line shows the approximate completeness limit
3. RESULTS of0.2L⊙ forthec2d+GBsample.
For each of the 230 protostars identified following the
selectionmethoddescribedabove,Table4listsarunning
index, the cloud in which the protostar is located, the
0.25
Spitzer source name (which also gives the coordinates),
the assumed A for extinction corrections, the infrared
V
spectralindex25 (α), Tbol,andLbol calculatedfromboth 0.20 Class 0
the observed and extinction corrected photometry, and
a flag indicating whether or not each protostar has any n Class I
Bi
availabledataat70µm<λ<850µm(see §4.3). InTa- r 0.15
ble 4 the extinction corrected values are denoted as α′, Pe
T′ , and L′ to differentiate them from the observed n
bol bol o
values. Throughout the remainder of this paper we con- cti 0.10
sider only the extinction corrected values and drop the a
r
primes for simplicity. We do not give uncertainties for F
the L derived in this work. Statistical uncertainties
bol 0.05
calculated by propagating through the uncertainties in
theobservedfluxesareonthe orderof10%,butthetrue
uncertainties are dominated by incomplete sampling of 0.00
the SEDs andareimpossible tocalculateforeachsource --22 --11 00 11 22
individually. These uncertainties will be discussed fur- log L (L )
bol O •
ther in §4.3.
Figure 1 shows the distribution of the extinction cor- Fig.2.—HistogramshowingLboldistributionsinlogspacewith
rected values of Lbol for all 230 protostars in log space. 1/3dexbins. Theshadedhistogramshowsthedistributionderived
With a minimum and maximum of 0.01 L and 69 L , inthisstudyforthe65outof230objectsinthecombinedc2dand
⊙ ⊙
GBsamplesclassifiedasClass0protostars. Thedashedhistogram
respectively, this distribution extends over greater than
shows the same thing, except for the 120 out of 230 objects clas-
three orders of magnitude. The mean and median are sified as Class I protostars. The classification is based on Tbol
4.3 L and 1.3 L , respectively. These statistics are calculated according to Equation 3 using the extinction-corrected
⊙ ⊙
summarized in Table 2. Also listed in Table 2 are four photometryandtheClassboundariesdefinedbyChenetal.(1995).
The solid vertical line shows the approximate completeness limit
dimensionless quantities calculated from the luminosity
of0.2L⊙ forthec2d+GBsample.
distribution: thestandarddeviationoflogL ,theratio
bol
of the median to mean L , the ratio of the maximum
bol
Class 0 and Class I sources. We have used T , calcu-
to mean L , and the fraction of protostars with L bol
bol bol
latedusingEquation3,toclassifyoursources,sinceT
≤ 0.1 L . These particular quantities are motivated by bol
⊙
is one of several commonly used indicators of class and
the recent theoretical study of protostellar luminosities
evolutionary status (e.g., Dunham et al. 2008; Enoch et
by Offner & McKee (2011), to which we compare our
al. 2009; Evans et al. 2009; Maury et al. 2011). Follow-
results below in §4.2.
ing Chen et al. (1995), Class 0 sources are selected with
Figure 2 shows the L distributions separately for
bol
the criterion that T < 70 K and Class I sources are
bol
25 The infraredspectral index, α, is calculated over all 2MASS selected with the criterion that 70 ≤ Tbol ≤ 650 K. In-
andSpitzer detections from2 24µm(Evansetal.2007). spection of Figure 2 reveals that the peak and extent of
−
7
TABLE 2
Luminosity Distribution
Statistics
Parameter Value
Total Number 230
Mean 4.3a L⊙
Median 1.3a L⊙
Minimum 0.01L⊙
Maximum 69L⊙
StandardDeviationoflog 0.73
Median/Mean 0.3
Maximum/Mean 16.0
Fraction 0.1L⊙ 0.07
≤
a
Asdescribedin§4.3,oncefar-infrared
and submillimeter photometry becomes
availableforthe43%ofthesamplelack-
ingany available data at 70 µm < λ<
850µm,themeanandmedianwilllikely
increase to approximately 5.8 and 1.8
L⊙, respectively. The effects of includ-
ingsuchdataontheoveralldistribution
ofLbol,andthusontheotherquantities
listedinthisTable,canonlybeinvesti-
gatedoncesuchdataareavailable.
the L distributions are similar for Class 0 and Class I
bol
sources. Thedistributionshavemean(median)valuesof TABLE 3
4.5 L and 3.8 L (1.4 L and 1.0 L ) for the Class 0 Lbol andLint forVeLLOs
⊙ ⊙ ⊙ ⊙
andI sources,respectively. However,there is one signifi-
Source Lbol Lint Referencea
cantdifferenceinthatthereisanexcessoflowluminosity
Class I sources comparedto the Class 0 population. For L1014-IRS 0.34 0.09 1
IRAM04191-IRS 0.13 0.08 2
theClassIpopulation,36%haveL <0.5L ,whereas
bol ⊙ L1521F-IRS 0.13 0.05 3
for the Class 0 population, only 20% have such lumi-
L328-IRS 0.18 0.05 4
nosities. A K-S test on the two distributions returns a L673-7-IRS 0.18 0.04 5
value of only 0.04, demonstrating that the difference at L1148-IRS 0.13 0.08–0.13 6,7
low luminosities is statistically significant. These results a References: (1) Young et al. (2004); (2) Dunham
are similar to those obtained by Enoch et al. (2009) for et al. (2006); (3) Bourke et al. (2006); (4) Lee et
a smaller sample. Very recently, several extremely low al.(2009);(5)Dunhametal.(2010b);(6)Kauffmann
etal.(2005);(7)Kauffmannetal.(2011).
luminosity,Class0sourceshavebeendiscoveredincores
classified as starless based on Spitzer observations (e.g.,
ples of this point can be found in recent, detailed stud-
Chen et al. 2010; Enoch et al. 2010; Pineda et al. 2011;
ies of individual VeLLOs that use continuum radiative
Schnee et al. 2012; Chen et al. 2012), emphasizing that
transfer models to separate internal and external heat-
at least some of this difference may be due to a bias
ing and determine the intrinsic L . The observed L
against the lowest luminosity Class 0 sources in Spitzer- int bol
and model-derived L for six such sources are listed in
selected samples. This point is further emphasized by int
Table 3. For at least 5 and possibly all 6, the observed
the factthatthe excessoflow-luminosityClassI sources
L are above 0.1 L while the model-derived L are
occurs below our approximate completeness limit of 0.2 bol ⊙ int
below 0.1 L , qualifying them as VeLLOs. As a con-
L (see below), where any such comparisons are limited ⊙
⊙ sequence, the fraction of protostars with L ≤ 0.1 L
in utility. The true similarity of the Class 0 and Class bol ⊙
reportedinTable 2(0.07)doesnotimply that7%ofthe
I L distributions must be revisited once current and
bol sampleareVeLLOs;manymoreVeLLOswithL >0.1
futuresurveyswithHerschel andALMAdetectandchar- bol
L are likely present in the sample.
acterize the full population of extremely low luminosity ⊙
Wehavedecidednottoattempttocorrectourluminos-
protostars.
itydistributionforsourceinclinationorexternalheating,
We emphasize that the results presented here are the
since any such corrections would be model-dependent
observedbolometricluminositiesofprotostars,whichare
(and in the case of external heating would require de-
not the same as the intrinsic protostellar luminosities.
tailedmodelingofalllow-luminosityprotostars,aproject
Departures from spherical symmetry break the correla-
farbeyondthescopeofthispaper). Whatwepresentare
tionbetweenobservedandintrinsicbolometricluminosi-
simply the observed bolometric luminosities (after cor-
ties, and external heating from the interstellar radiation
recting for extinction). Theoreticalstudies that attempt
field breaks the correlation between bolometric and in-
to explain the observed protostellar luminosity distribu-
ternalluminosity. Regardingthe latter, externalheating
tion must take these considerations into account.
canadduptoseveraltenthsofasolarluminositydepend-
Sincetherelationshipbetweenthefluxesinthevarious
ingonthelocalstrengthoftheinterstellarradiationfield
Spitzer bandsandL dependsnotonlyondistancebut
andthecoremassavailabletobeheatedexternally(e.g., bol
also on the detailed spectral shape of each source, local
Evans et al. 2001) and can dominate the observed L
bol strengthoftheexternal(interstellar)radiationfield,and
for the lowest luminosity objects. A few specific exam-
total core mass available to be heated externally, there
8
is no one completeness limit for each cloud or for the
full c2d+GB dataset. In a detailed search for and study
of low luminosity protostars in the c2d survey, Dunham 0.20 c2d+GB (this study)
et al. (2008) found that the c2d data are sensitive to
protostars with L ≥ 4×10−3 (d/140pc)2 L . With c2d (Evans et al. 2009)
int ⊙
cloud distances ranging from 125−950 pc, the resulting
n 0.15
luminosity sensitivites range from 0.003− 0.18 L⊙, or Bi
0.003−0.04 L⊙ if IC5146 is omitted. However,this sen- er
sitivity is for L rather than L ; as discussed above, P
int bol n
the two are not the same for low luminosity protostars, o 0.10
with Lbol equal to or greater than Lint depending on acti
the details of the external heating. In another study, r
F
Enochetal.(2009)estimatedcompletenesslimitsofL
bol 0.05
∼ 0.01−0.05 L for protostars in the c2d clouds, al-
⊙
though they emphasized that there was significant un-
certaintyinderivingsuchlimits. Takingintoaccountall
0.00
oftheaboveinformation,weconservativelyestimatethat
oursampleisonlycompleteforL >0.2L (thesensi- --22 --11 00 11 22
bol ⊙
tivity limit forIC5146,the mostdistantcloud,using the log Lbol (LO •)
Dunham et al. [2008] relation and assuming no external
heating),andmarkthis limit witha solidverticalline in Fig.3.—HistogramshowingLboldistributionsinlogspacewith
1/3 dex bins. The shaded histogram shows the distribution de-
all figures presenting histograms of L . The existence
bol rivedinthisstudyforthe230protostarsinthecombinedc2dand
ofprotostarsbelow these limits willbe discussedin§5.3. GB samples (see Figure 1 for error bars). The dashed histogram
showsthedistributionforthe112protostars inthec2dsampleas
4. DISCUSSION derived by Evans et al. (2009). The solid vertical line shows the
approximatecompletenesslimitof0.2L⊙ forthec2d+GBsample.
In this section we discuss our results in comparison
to other observational and theoretical studies of proto-
stellar luminosities. In §4.1 we compare to two recent Wavelength (microns)
100 50 20 10
determinations of the observed protostellar luminosity
) 1.5
distribution, and in §4.2 we discuss several recent theo- -1z
reticalinvestigationsofprotostellarluminosities. Finally, H
in §4.3 we discuss the effects of missing far-infrared and -1s
submillimeter photometry on our derived luminosities. -2m
c 1.0
4.1. Comparison to Other Observations g
r
e
4.1.1. Comparison to c2d Results 1
2
-0
Evans et al. (2009) identified 112 protostars in the 1
(
c2d survey and calculated their observed bolometric lu- y 0.5
minosities. Our methods for identifying protostars, as- sit
n
sembling complete SEDs, and calculating L are very e
bol D
similar to theirs. All of their protostars are included x
u
in our study, but we have expanded to the rest of the Fl
star-formingcloudsobservedbythe GB surveyandthus 0.0
increased the number of protostars from 112 to 230. 00 11 22 33
We have also made three changes to the ancillary pho- Frequency (1013 Hz)
tometry included when assembling complete SEDs: (1)
we have included 12 and 22 µm photometry from the Fig. 4.—SpectralenergydistributionofNGC1333-IRAS2A,plot-
WISE all-sky survey, (2) we have included additional tedasSν versusνinlinearspace. Thelightshadedareashowsthe
resultofintegratingunderthecurvewhentheWISE 12and22µm
SHARC-II350µmphotometryfromatargetedsurveyof
photometry is included, whereas the dark shaded area shows the
nearby, low-mass star forming regions (Wu et al. 2007; extra amount amount added to the integral when no photometry
M.M. Dunham etal.2012,in preparation)that wasnot is available between 8 and 70 µm, as was the case for Evans et
yet available when Evans et al. (2009) completed their al.(2009).
study; and (3) we have not included any IRAS photom-
etry. The last change is motivated by the superiority of 1.3 L in this work and 1.5 L in the c2d-only sample
⊙ ⊙
WISE 12 and 22 µm and Spitzer 70 µm data to IRAS (Evans et al. 2009). A K-S test on the two distributions
12, 25, and 60 µm in essentially all cases, and the ex- returns a value of 0.33, indicating they are not signifi-
treme confusion from both nearby sources and ambient cantly different.
cloud emission in the IRAS 100 µm data. Despitetheirgeneralsimilarities,thetwodistributions
Figure 3 compares the L distributions from this do have different means: 4.3 L in this study versus
bol ⊙
workandfromEvansetal.(2009). Thenewdistribution 5.3 L in the c2d-only sample (Evans et al. 2009). The
⊙
obtained in this study has a similar shape and extent to mean is strongly influenced by the highest luminosity
the c2d-only distribution, except now with better statis- sources,severalofwhichwereoverestimatedbyEvanset
tics. The medians are also quite similar, with values of al.(2009). Tounderstandthe causeofthis overestimate,
9
wenotethatthereare14sourcesintheEvansetal.sam- ferent. Kryukova et al. do not compile complete SEDs
plesaturatedat24µmwithSpitzer andthuslackingany to use in calculating bolometric luminosities. Instead,
photometry between 8 and 70 µm. By including WISE for all sources they identify as protostars,they calculate
12 and 22 µm photometry, which was not available to L ,themid-infraredluminosityfromtheir2MASSand
MIR
Evansetal.,ourupdatedsamplefills inthis gap. Figure Spitzer 1.25−24 µm data, and α, the infrared spectral
4 plots the SED for NGC1333-IRAS2A (source 144 in indexcalculatedfrom3.6to24µm. Forsourceswithα≥
Table 4), one of the 14 sources saturated in the Spitzer 0.3 common to both their sample of protostars and the
24 µm observations. The SED is plotted as S versus ν Evansetal.(2009)sample,theythenderivethefollowing
ν
inlinearspaceratherthanthemoretypicalνS versusλ empirical relationship:
ν
inlogspacesincetheformeristhespaceinwhichthein-
tegralin Equation2 iscalculated. The lightshadedarea LMIR =(−0.466±0.014×log(α)+0.337±0.053)2 ,
shows the result of the integral when the WISE 12 and L
bol
22 µm photometry is included, whereas the dark shaded (4)
areashowsthe extraamountaddedto the integralwhen where L is from Evans et al. (2009). They use this
bol
no photometry is available between 8 and 70 µm. relation to calculate L for all protostars with α ≥
bol
As clearly demonstrated by Figure 4, omitting pho- 0.3, and the value of this relation at α = 0.3 to cal-
tometry between 8 and 70 µm can lead to significant culate L for all protostars with α < 0.3. At least
bol
overestimates of L . In the specific case of NGC1333- some of the discrepancy between our results and those
bol
IRAS2A, Evans et al. (2009) measured L = 76 L of Kryukovaet al. may arise because we have made sev-
bol ⊙
whereaswemeasureL =22L withtheWISE 12and eral changes to the SEDs used to calculate L , as de-
bol ⊙ bol
22 µm photometry included, a factor of 3.4 lower. Our scribed above in §4.1.1. To examine this possibility, we
measurement is consistent with previous measurements re-derivedthe aboveempiricalcorrelationusing our new
of L for this source whereas the Evans et al. value values of L for the sources common to both our sam-
bol bol
is not (e.g., Jørgensen et al. 2002). For the 14 sources ple and the Kryukova et al. sample, and obtained the
saturated at 24 µm, Evans et al. (2009) measured L following modification using a linear least-squares fit:
bol
ranging from 2.6 to 76 L , with a mean and median of
⊙
30 and 27 L⊙, respectively. For those same 14 sources LMIR =(−0.298±0.046×log(α)+0.270±0.013)2 .
and with WISE 12 and 22 µm photometry included, we L
bol
measure L ranging from 1.4 to 63 L , with a mean (5)
bol ⊙
and median of 20 and 13 L , respectively. Most of the We illustrate the effects of this modification in Figure 6,
⊙
decrease in the overall sample mean from 5.3 L to 4.3 which re-creates Figure 5 from Kryukova et al. (2012).
⊙
L is a result of correcting this overestimate for several Using our modified relationship between L , L ,
⊙ bol MIR
relatively high luminosity sources. and α, the changes to the Kryukova et al. (2012) L
bol
values range from an increase by a factor of 1.9 to a
4.1.2. Comparison to Kryukova Results
decrease by a factor of two, depending on α for each
Recently, Kryukova et al. (2012) presented observed source. The mean changeof allsources commonto both
protostellar luminosity distributions assembled from samples is an increase by a factor of 1.5. Such a change
Spitzer observations of 11 molecular clouds: the 7 c2d can explain much of the difference in means in the two
clouds (Chamaeleon II, Lupus I, Lupus III, Lupus IV, samples (4.2 L in this study compared to 2.3 L in
⊙ ⊙
Ophiuchus, Perseus, and Serpens), Taurus, and 3 mas- Kyuokova et al.), but cannot fully explain the excess of
sivestar-formingclouds(Orion,CepOB3,andMonR2). low-luminosity sources. Instead, the remainder of the
In total they identified 727 protostars in these clouds. discrepancy between our results lies in source selection.
Figure 5 compares our results. Kryukovaetal.identify43protostarsinthec2dclouds
The left panel of Figure 5 compares the L distribu- not identified by us or by Evans et al. (2009). Figure 7
bol
tions from this work and from Kryukova et al. (2012). shows the L distribution for these 43 sources. Most
bol
We use the contamination-subtracted L distributions haveL ≤1.0L ,andwhiletheeffectsofthese“extra”
bol bol ⊙
fromKryukovaetal.forthiscomparison. Thetwodistri- low-luminositysourcesaresignificantlymitigatedbysta-
butions are generally quite similar, but since Kryukova tistical contamination corrections included by Kryukova
et al. include three massive star-forming clouds in their et al., their net effect is to cause an excess of low-
sample, environmental effects may mask our ability to luminosity sources compared to our results. The main
properly compare the two results. Thus, the right panel difference between our method of selecting protostars
of Figure 5 compares the L distributions from this and that of Kryukova et al. is our requirement of at
bol
workandfromKryukovaetal.,wherenowbothsamples least one detection at λ ≥ 350 µm to ensure associa-
arerestrictedtothecloudscommontobothsamples(the tion with dense cores. Since the 43 sources shown in
c2d clouds). Figure 7 are not in our sample, they are not associated
Inspection of the right panel of Figure 5 clearly shows with (sub)millimeter detections and thus not associated
that, for the same clouds,Kryukovaet al.(2012)findan withknowndensecores. Byourdefinitionofaprotostar
observed distribution of protostellar luminosities that is (see §1), these are not protostars.
generally shifted to lower luminosities compared to our However, it is possible that at least some of these
results, with a much lower mean (2.3 L versus 4.2 L sources are in fact protostars associated with relatively
⊙ ⊙
in our sample) and much higher fraction of protostars low-mass cores not detected by the (sub)millimeter sur-
with L ≤ 0.1 L (22% versus 7% in our sample). A veys we used to compile complete SEDs. This is sup-
bol ⊙
K-Stestonthe twodistributions returnsavalue of0.01, ported by the fact that many such surveys have rela-
verifying that the two distributions are statistically dif- tively high completeness limits (for example, the 50%
10
c2d+GB (this study) c2d (this study)
0.20 0.20
Kryukova et al. (2012) c2d (Kryukova)
n n
Bi 0.15 Bi 0.15
er er
P P
n n
o o
cti 0.10 cti 0.10
a a
Fr Fr
0.05 0.05
0.00 0.00
--22 --11 00 11 22 --22 --11 00 11 22
log L (L ) log L (L )
bol O • bol O •
Fig.5.—HistogramshowingLboldistributionsinlogspacewith1/3dexbins. Inbothpanels,thesolidverticallinesshowtheapproximate
completenesslimitsof0.2L⊙ forthec2d+GBsample. Left: Theshadedhistogramshowsthedistributionderivedinthisstudyforthe230
protostarsinthecombinedc2dandGBsamples(seeFigure1forerrorbars). Thedashedhistogram showsthecontamination-subtracted
distributionforthe727protostarsidentifiedbyKryukovaetal.(2012). Right: Theshadedhistogramshowsthedistributionfromthiswork
whenonlyincludingthesourcesinthec2dclouds. Thedashedhistogramshowsthecontamination-subtracted distributionfromKryukova
etal.(2012) whenonlyincludingsourcesfromthesamec2dclouds.
1.0
100
0.8
)ol
b
L/LMIR 0.6 /LolMIR 10
qrt( Lb
s 0.4
0.2
1
0.0
-0.6 -0.4 -0.2 -0.0 0.2 0.4 0.6 -2.0 -1.5 -1.0 -0.5 0.0 0.5
log(alpha) log(alpha)
Fig.6.— Re-creation of Figure5 from Kryukova et al. (2012), showing p(LMIR/Lbol) (left) and Lbol/LMIR (right) vs. log(α) for the
protostarscommontobothoursampleandtheKryukovaetal.sample. LbolisfromthisworkwhereasLMIR (themid-infraredluminosity)
and α (the infraredspectral s lope) are given by Kryukova et al. (2012). The solid lines show the best-fit relation from Kryukova et al.,
whereasthedotted lineshowsthemodifiedrelationderivedusingournewvaluesofLbol (seetextfordetails).
completeness limits for the Bolocam 1.1 mm surveys of to estimate the contaminationfromgalaxiesand remove
Perseus, Serpens, and Ophiuchus are 0.8, 0.6, and 0.5 the effects of this contamination from their luminosity
M , respectively; Enoch et al. [2008]). On the other distribution. Such an estimate is a lower limit only be-
⊙
hand, some sources may be contaminants masquerading causeitdoesnottakeintoaccountthe factthatgalaxies
in the sample. Kryukova et al. (2012) made a careful in their cloud source catalogs are observed through the
attempt to correct for such contamination in a statisti- extraextinctionoftheclouditself,reddeningallgalaxies
cal sense, but the resulting corrections are highly uncer- andthusincreasingthenumberofgalaxieswiththeiras-
tain and may have been underestimated. For example, sumed colors of protostars. Furthermore, Heiderman et
they applied their protostar selection criteria to the cat- al.(2010)recently showedthat many sourcesselected as
alog produced by the Spitzer SWIRE Legacy survey of ClassIYSOsbytheirinfraredcolorslackedthe presence
the ELAIS N1 extragalactic field (Lonsdale et al. 2003) ofwarm,densegasandarethusnotprotostars(andmay