A Comprehensive Comparative Test of Seven Widely-Used Spectral Synthesis Models Against Multi-Band Photometry of Young Massive Star Clusters

1.8 MB·

by  A. Wofford

Mon.Not.R.Astron.Soc.000,1–31(2002) Printed18January2016 (MNLATEXstylefilev2.2)
A Comprehensive Comparative Test of Seven Widely-Used
Spectral Synthesis Models Against Multi-Band
Photometry of Young Massive Star Clusters
6
1
0
2 A. Wofford1⋆, S. Charlot1, G. Bruzual2, J.J Eldridge3, D. Calzetti4, A. Adamo5,

n
M. Cignoni6, S. E. de Mink7, D.A. Gouliermis8,9, K. Grasha4, E. K. Grebel10,
a
J J. Lee6, G. O¨stlin5, L.J. Smith6, L. Ubeda6, E. Zackrisson11
5
1 1Sorbonne Universit´es, UPMC-CNRS, UMR7095, Institut d’Astrophysique de Paris, F-75014 Paris, France
2Instituto de Radioastronoma y Astrofsica, UNAM, Campus Morelia, M´exico
] 3Dept. of Physics, Universityof Auckland, Auckland, NewZealand
A 4Dept. of Astronomy, University of Massachusetts – Amherst, Amherst, MA, USA
5Dept. of Astronomy, The Oskar KleinCentre, Stockholm University,AlbaNova University Centre, SE-106 91 Stockholm, Sweden
G
6Space Telescope Science Institute, Baltimore, MD, USA
h. 7Astronomical Institute Anton Pannekoek, Amsterdam University,Amsterdam, The Netherlands
p 8Universityof Heidelberg, Centre for Astronomy, Institute for Theoretical Astrophysics, Albert-Ueberle-Str.2, 69120 Heidelberg, Germany
- 9Max Planck Institute for Astronomy, K¨onigstuhl17, 69117 Heidelberg, Germany
o 10Astronomisches Rechen-Institut, Zentrumfu¨r Astronomie derUniversita¨t Heidelberg, M¨onchhofstr, Heidelberg, Germany
r 11Dept. of Physics and Astronomy, Uppsala University,Box 515, SE-751 20 Uppsala, Sweden
t
s
a
[
18January2016

1
v
0 ABSTRACT
5 We test the predictions ofspectralsynthesis models basedonsevendifferent massive-
8 star prescriptions against Legacy ExtraGalactic UV Survey (LEGUS) observations
3 of eight young massive clusters in two local galaxies, NGC 1566 and NGC 5253,
0
chosen because predictions of all seven models are available at the published galactic
.
1 metallicities. The high angular resolution, extensive cluster inventory and full near-
0 ultraviolet to near-infrared photometric coverage make the LEGUS dataset excellent
6 for this study. We account for both stellar and nebular emission in the models and
1 try two different prescriptions for attenuation by dust. From Bayesian fits of model
v: librariestotheobservations,wefindremarkablylowdispersioninthemedianE(B−V)
i (∼ 0.03mag), stellar masses (∼ 104M⊙) and ages (∼ 1Myr) derived for individual
X
clusters using different models, although maximum discrepancies in these quantities
r canreach0.09magandfactorsof2.8and2.5,respectively.Thisisforrangesinmedian
a properties of 0.05–0.54 mag, 1.8–10×104M⊙ and 1.6–40Myr spanned by the clusters
in our sample. In terms of best fit, the observations are slightly better reproduced
by models with interacting binaries and least well reproduced by models with single
rotating stars. Our study provides a first quantitative estimate of the accuracies and
uncertaintiesofthemostrecentspectralsynthesismodelsofyoungstellarpopulations,
demonstrates the good progress of models in fitting high-quality observations, and
highlights the needs for a larger cluster sample and more extensive tests of the model
parameter space.
Key words: (ISM:) Hii regions – galaxies: star clusters: general – galaxies: star
formation – ultraviolet: galaxies – stars: early-type.
1 INTRODUCTION
Young massive clusters (YMCs) are dense aggregates
⋆ E-mail:woff[email protected]
of young stars that are considered to be fundamen-
(cid:13)c 2002RAS
2 A. Wofford
tal building blocks of galaxies (Portegies Zwart et al. puted with the code PARSEC (Bressan et al. 2012) are able
2010). Determining accurate stellar masses and ages for to reproduce these loops at metallicities of Z = 0.001 and
large samples of individual YMCs in a wide range of Z = 0.004. This is accomplished by enhancing the over-
galaxy environments is essential for studies of star clus- shooting at the base of the convective envelope during the
ter populations and their evolution (Adamo et al. 2010; firstdredge-up,andinvokinglargemixinglengthsoftwoand
Bastian et al.2012;Fall & Chandar2012;Baumgardt et al. fourtimesthepressurescaleheight(Tang et al.2014).New
2013; Chandar et al. 2014), the star formation rates PARSEC massive-star tracks at other metallicities including
and spatially-resolved star formation histories of galaxies solar are presented in Chen et al. (2015).
(Glatt et al.2010;Wofford et al.2011;Chandaret al.2015; Second, nitrogen enhancements are observed on the
Calzetti et al. 2015a), and the effects on the evolution of surfaces of main sequence stars of typically 15M⊙
galaxiesoftheradiative,chemical,andmechanicalfeedback (Hunteret al. 2009). In such stars, this product of nuclear
of YMCs (e.g., Calzetti et al. 2015b, hereafter C15b). burning is not expected at the surface, since these stars do
The task of observing large samples of star clusters notdevelopstrongstellarwindscapableofexposingthecen-
in galaxies with a wide range of properties was recently tral nitrogen. The Geneva tracks for single rotating stars
completed by the Hubble Space Telescope (HST) Treasury (Georgy et al. 2013; Ekstr¨om et al. 2012) favor surface ni-
program, LEGUS: Legacy Extragalactic Ultraviolet Sur- trogen enhancements. This is because rotation effectively
vey (GO-13364; Calzetti et al. 2015a). LEGUS consists of mixes innerand outer stellar layers.
high spatial resolution (∼ 0.07”) images of portions of 50 Finally,itisnowwellestablishedthatmassivestarsare
nearby (6 13 Mpc) galaxies taken with the UVIS channel in binary systems with close to 70 per cent of them inter-
oftheWideFieldCameraThree(WFC3)inbroadbandfil- acting over the course of their evolution (e.g. Massey et al.
ters F275W (2704 ˚A), F336W (3355 ˚A), F438W (4325 ˚A), 2009; Sana et al. 2012, 2013; Chini et al. 2012). Processes
F555W(5308˚A),andF814W(8024˚A).Thesurveyincludes thatoccurduringtheevolutionofbinariesincludeenvelope
galaxiesofdifferentmorphologicaltypesandspansfactorsof removalfromthebinary,accretionofmassbythesecondary,
∼103inbothstarformationrate(SFR)andspecificstarfor- or even complete mergers (e.g. Podsiadlowski et al. 1992;
mationrate(sSFR),∼104instellarmass(∼107−1011M⊙), Langer 2012; deMink et al. 2013, 2014). Several indepen-
and∼102 in oxygenabundance(12+logO/H=7.2−9.2). dentgroupshavedevelopedevolutionarytracksthataccount
Someofthetargetsinthesurveyhavehighqualityarchival forinteractingbinariesandareusefulforspectralpopulation
images in some filters covering similar bandpasses required synthesismodels (e.g., Vanbeverenet al. 1998; Zhang et al.
by LEGUS, from the Wide Field Channel of HST’s Ad- 2004, 2005; and references therein). In this work, we use
vanced Camera for Surveys (ACS), or in fewer cases, from models computed with the Binary Population and Spectral
ACS’s High Resolution Channel (HRC). For such targets, Synthesis code, bpass, which is last described in Eldridge
LEGUS observedin filters thatcomplete thefivebandcov- etal.(inpreparation).Hereafter,wewillrefertosinglestar
erage. and binary tracks that are implemented in bpass as the
At thedistances of LEGUS galaxies (3−13Mpc), star Auckland tracks. Compared to other types of evolutionary
clusters cannot be resolved into individual stars, and they tracks, those including rotation and interacting binaries re-
usuallyappearaslightover-densities,slightlymoreextended quire the inclusion of a large number of uncertain physical
thanthestellarPSF.Insuchcases,astandardtechniquefor parameters.Howeverinthiswork,weusestandardrotating
deriving cluster masses and ages is the comparison of ob- and binary models and do not vary any of these uncertain
served and computed predictions of the integrated light of parameters to achieve a betterfit.
clusters in various photometric bands. Such predictions are Anothercontributortotheuncertaintiesofspectralsyn-
obtainedbyconvolvingspectralsynthesismodelswithfilter thesis models of massive-star populations are uncertainties
system throughputs. With regards to populations of mas- in models ofmassive-star atmospheres, which givetheindi-
sive stars, at fixed star formation history and stellar initial vidual spectra of the stars as a function of time. We delay
mass function (IMF), a major contributor to uncertainties thediscussion on this topic to furtherdown in thepaper.
inspectralsynthesismodelsofyoungpopulationsareuncer- In this work, we use near-ultraviolet (NUV) to near-
taintiesin massive-star evolutionary tracks(Leitherer et al. infraredphotometryoftwoavailableLEGUSgalaxies,spec-
2014).Foragiveninitialmetallicity,suchtracksprovidethe tralsynthesismodelsbasedonsevendifferentflavorsofmas-
mass, temperature, and luminosity of a star as a function sivestarevolution (older andstate-of-the-art),and twodif-
of age. Significant uncertainties still remain in massive-star ferent prescriptions for attenuation by dust, in order to ad-
evolution because of our poor knowledge of some complex dress five questions: 1) how well do the different spectral
physical processes, which still require an empirical calibra- synthesismodelsfittheobservations,2)isthereapreferred
tion,such astheefficiency of convectiveheat transport and flavor of massive star evolution, 3) is there a preferred pre-
interior mixing, and theeffects of close-binary interactions. scription for attenuation by dust, 4) how well do the ob-
In recent years, independent groups working on mas- servations and models constrain the YMC properties, and
sive star evolutionary tracks have attempted to reproduce 5) how different are the YMC properties obtained with the
threekeyobservationalconstraints.First,blueloopstarsare different models. The outline of our work is as follows. In
observed in the color-magnitude diagrams of nearby metal- the first four sections we describe the sample (Section 2),
poordwarfirregular star-forming galaxies. Themost recent observations (Section 3), models (Section 4), and method
Padova tracks for single non-rotating massive stars com- for comparing models to observations (Section 5). Section
(cid:13)c 2002RAS,MNRAS000,1–31
ATestof SevenWidely-Used SpectralSynthesis ModelsAgainstMulti-Band Photometryof YMCs. 3
6 presents our results, which are discussed in Section 7. Fi- relies on visual inspection in order to remove observational
nally, we providea summary and conclude in section 8. artifacts, contaminants, and false positives. Only a handful
ofgalaxieshadbeenvisuallyinspectedbytheLEGUSteam
atthetimeofthiswork.Ourfinalsampleiscomposedofsix
clustersingalaxyNGC1566andtwoclustersingalaxyNGC
2 SAMPLE
5253. Next we provide a brief description of these galaxies,
whose main properties, including morphology, redshift, dis-
Weselect oursample ofYMCsusingfourmain criteria and
cluster masses and ages determined by Adamo et al. (in tance, distance modulus, oxygen abundance, SFR, stellar
mass, and color excess associated with attenuation within
preparation, hereafter A16, NGC 1566) and C15b (NGC
theMilky Way are summarized in Table 1.
5253). The first criterion is that all clusters are detected
in the five LEGUS bands. The second criterion is that the
ages of clusters are 6 50Myr, which ensures that there
2.1 NGC 1566
are massive stars in the YMCs. For reference, 50 Myr
is the approximate main sequence life time of a single NGC1566isanearlyface-ongrand-designspiralgalaxythat
non-rotating 10 M⊙ star (Meynet et al. 1994). The third isthebrightest memberoftheDoradogroup(Agu¨ero et al.
criterion is that the clusters have masses of >5×104 M⊙, 2004). It has an intermediate-strength bar type (SAB)
which mitigates the effect of the stochastic sampling of the and hosts a low-luminosity AGN. Its Seyfert classification
stellar initial mass function (IMF, Cervin˜o & Luridiana varies between 1 and 2, depending on the activity phase
2004). A16 and C15b derive cluster properties using (Combes et al. 2014). Its globally-averaged gas-phase oxy-
yggdrasilspectralsynthesismodels(Zackrisson et al.2011, genabundanceis12+log(O/H)=8.63 or9.64, dependingon
http://ttt.astro.su.se/projects/yggdrasil/yggdrasil.html). thecalibrationthatisgiveninTable1.Weselectsixclusters
The version of yggdrasil used in the latter papers in this galaxy, excluding the nuclear region because of the
take model spectra of simple stellar populations (SSPs) presence of theAGN.
computed with starburst99 (Leithereret al. 1999;
V´azquez& Leitherer 2005), as input to photoioniza-
2.2 NGC 5253
tion models computed with cloudy (last described in
Ferland et al. 2013). An SSP corresponds to a system NGC5253isadwarfstarburstgalaxyofmorphologicaltype
where the stars are born instantaneously and are thus Im. It has a fairly flat (Westmoquetteet al. 2013) galac-
coeval. C15b and A16 obtain their cluster properties by tocentric profile of the gas-phase oxygen abundance with
using two different sets of older massive-star evolutionary a mean value of 12+Log(O/H)=8.25 (Monreal-Ibero et al.
tracks, which are described in detail in section 4, Padova 2012). This valueof theoxygen abundanceis similar to the
(Po) and Geneva (Go) tracks. In our sample, we include onereportedbyBresolin(2011),12+Log(O/H)=8.20±0.03.
all clusters with Po- or Go-based properties satisfying our Calzetti et al. (2015b) recently studied the brightest clus-
massandagecriteria.Thelastselectioncriterionisthatthe ters in this galaxy. Using extraordinarily well-sampled UV-
averageHii-regiongas-phaseoxygenabundanceof thehost to-near-IR spectral energy distributions and models which
galaxy must correspond closely to a metallicity for which aredescribedinsection4below,theyobtainunprecedented
massive-star tracks from the Padova, Geneva, and Auck- constraints on dust attenuations, ages, and masses of 11
landcollaborations exist.Theabovecollaborations recently clusters. In particular they find two clusters with ages and
released tracks which account for different non-standard masses satisfying our selection criteria and for which their
physicsthatarementionedintheintroduction.Sinceatthe fits are excellent, i.e., χ2 ∼ 1. Using their notation, the se-
time of writing Geneva tracks for rotating stars are only lected clusters are#5, which islocated within theso-called
available at Z = 0.002 (Georgy et al. 2013) and Z = 0.014 “radio nebula”; and #9, which is located outside of the
(Ekstr¨om et al. 2012), we focus on these two metallicity “radio nebula” but still within the starburst region of the
regimes. galaxy.
We encountered three main difficulties in assembling a
sample satisfying our requirements. Firstly, stellar popula-
tions with masses >5×104M⊙ and ages 650Myrare sig- 3 OBSERVATIONS
nificantlylessnumerousthanlessmassiveclustersofsimilar
ages. Cluster formation is a stochastic process undergoing NGC 1566 was observed in the fiveLEGUS broad band fil-
size-of-sample effects (Larsen 2002). More massive clusters ters, F275W, F336W, F438W, F555W, and F814W. NGC
aremorelikelytoformingalaxieswithhigherSFRs(Whit- 5253 was observed in WFC3/UVIS filters F275W and
more2000,Larsen2002)orwithinlongertimescales(Hunter F336W and ACS/HRC filters F435W (4311 ˚A), F550M
etal2003).Secondly,fortheSFRrangecoveredbytheLE- (5578 ˚A), and F814W (8115 ˚A), where observations in the
GUSsample, wedonotexpectalarge population of YMCs latter three filters are archival (PID 10609, PI Vacca). The
within the mass and age limits imposed in this work. In- pixel scales of WFC3/UVIS and ACS/HRC are 0.039 and
deed, YMCs satisfying our mass and age criteria are more 0.025 arcsec/pixel, respectively. In order to preserve the
frequentlyfoundinstarburstgalaxies, whichareaminority highestangularresolution,thealignedUVISimagesofNGC
intheLEGUSsample(Calzetti et al.2015a).Finally,atthe 5253wereallre-sampledtothepixelscaleofHRC.ForNGC
timeofwriting,theLEGUSmethodforfindingtrueclusters 1566, photometry is performed with a circular aperture of
(cid:13)c 2002RAS,MNRAS000,1–31
4 A. Wofford
Table 1.Galaxyproperties.
NGC Morph.a zb Dist.c µd 12+log(O/H)e SFR(UV)f M∗g E(B−V)MWh
# (Mpc) mag (PT)/(KK) (M⊙/yr) (M⊙) mag
1566 SABbc 0.005017 13.2 30.60 8.63/9.64 5.67 2.7E10 0.005
5253 Im 0.001358 3.15 27.78 8.25 0.10 2.2E8 0.049
aMorphologicaltypegivenbytheNASAExtragalacticDatabase, NED.
bRedshiftfromNED.
cDistance, as listed in table 1 of Calzetti etal. (2015a) and used in this work. We became aware of a
revised distance of NGC 1566 (∼18 Mpc, corresponding distance modulus of 31.28 mag) too late for
includingitinthepresentwork.
dDistancemodulus.
eForNGC1566,globallyaveragedabundance(Moustakas etal.2010).Thetwovalues,(PT)and(KK),
are the oxygen abundances on two calibration scales: the PT value, in the left-hand-side column, is
from the empirical calibration of Pilyugin&Thuan (2005), the KK value, in the right-hand-side col-
umn, is from the theoretical calibration of Kobulnicky&Kewley (2004). For NGC 5253, value from
Monreal-Iberoetal.(2012).
f Average star formation rate calculated from the GALEX far-UV, corrected for dust attenuation, as
describedinLeeetal.(2009).WenotethatSFRindicatorscalibratedfortheyouthofthestarburstin
NGC5253yieldSFR∼0.4M⊙ (Calzetti etal.2015b).
gStellar masses obtained from the extinction-corrected B-band luminosity, and color information, us-
ing the method described in Bothwelletal. (2009) and based on the mass-to-light ratio models of
Bell&deJong(2001).
hColor excess associated with extinction due to dust in MilkyWay. Uses Schlafly&Finkbeiner (2011)
extinction mapsandFitzpatrick(1999)reddeninglawwithRv=3.1.
0.156” in radius, with the background measured within an tainedinthispaperwiththemodelsusedtoselecttheclus-
annulusof0.273”ininnerradiusand0.039”inwidth;while ters in our sample, we adopt the same stellar and nebu-
for NGC 5253, the aperture is 0.125” in radius and the an- lar parameters as in the latter models, which are given in
nulus has an inner radius of 0.5” and width of 0.075”. For Zackrisson et al. (2011). In particular, we use simple stel-
NGC 5253, C15b provide luminosities, which we first con- lar populations (SSPs) with initial masses of 106M⊙ and
vert to the AB system, and then to the Vega system using an IMF such that the number of stars in the mass range
conversion factors 1.49, 1.18, -0.10, 0.03, and0.43 forfilters [m, m+dm] is given by N(m)dm ∝ m−α, where α = 1.3
F275W,F336W,F435W,F550M, andF814W,respectively. and α = 2.35 in the mass ranges 0.1–0.5M⊙ and 0.5–
The conversion factors were computed from the spectrum 100M⊙, respectively. We compute models for ages in the
of Vega that is described in Bohlin (2007). For NGC 1566, range 6 6 log(t/yr) 6 9 in steps of 0.1. In our approach,
theobservationalerrorsinthedifferentbandswereobtained the stars and ionized gas have the same metallicity. Since
by summing in quadrature the photometric error produced some of the stellar evolutionary tracks that we try are only
by IRAF’s task PHOT and the standard deviation derived available at two metallicities (Z = 0.002 and Z = 0.014),
from the aperture correction. For NGC 5253, the observa- in our Bayesian fitting approach, metallicity is not a free
tional errors are driven by crowding and uncertainties in parameter. Depending on the massive-star evolution flavor,
theaperturecorrection. Thelatterareatthelevelof±15% we use tracks corresponding to metallicities of Z = 0.002
for the UV–optical filters used here. When combined with or Z = 0.004 (NGC 5253) and Z = 0.014 or Z = 0.020
other smaller contributions, the total uncertainty amounts (NGC 1566). These metallicities roughly correspond to the
to±0.175 mag. Forfurtherdetails on thedata reduction of averagegas-phasemetallicitiesofthegalaxiesinoursample,
NGC 1566 and NGC 5253, see A16 and C15b, respectively. as gauged by their oxygen abundances (see Table 1).1 The
ThecoordinatesandphotometryoftheYMCsaregivenfor elemental abundances in the ionized nebula are scaled so-
each galaxy in Tables 2 and 3, respectively. larabundances.Weusethereferencesolarabundancesetof
Figure 1 shows LEGUS NUV images of the two galax- Asplundetal.(2009),whichcorrespondstoZ⊙ =0.014and
ies, where we have marked the locations of clusters in our 12+log(O/H) = 8.69. In the next subsections, we expand
sample.ForNGC1566,thefinalLEGUSIDsofclusterswere on the different model components, explain our procedure
not available at the time of writing and thus we assign our
own IDs, while for NGC 5253, we use the IDs of C15b, for
easier comparison with theirwork.
1 The highest O/H value of NGC 1566 given in Table 1 yields
a metallicity much larger than Z = 0.02 when adopting the
Asplundetal.(2009)solarabundancescale.WeexcludethisO/H
4 MODELS valueandadoptthelowestvaluecorrespondingtoZ=0.014,be-
cause some of the evolutionary tracks used in this work are not
The spectral synthesis models computed in this paper ac- available at Z > 0.014. However, if available, we use Z = 0.020
count for the contributions of stars, the ionized gas, and instead,becauseitisinbetweenthehigh/lowO/Hvaluesquoted
dust. For a more meaningful comparison of the results ob- forthisgalaxy.
(cid:13)c 2002RAS,MNRAS000,1–31
ATestof SevenWidely-Used SpectralSynthesis ModelsAgainstMulti-Band Photometryof YMCs. 5
Table 2.Location&photometryofclustersinNGC1566.
ID RAa DECa F275Wb F336Wb F438Wb F555Wb F814Wb
mag mag mag mag mag
1 64.98079821 -54.93067859 17.465±0.069 17.816±0.090 19.298±0.062 19.356±0.045 19.248±0.061
2 64.98077227 -54.93850353 17.541±0.070 17.875±0.092 19.307±0.064 19.325±0.047 19.088±0.060
3 65.01218101 -54.94103612 16.946±0.067 17.360±0.088 18.774±0.059 18.848±0.041 18.608±0.054
4 65.02378180 -54.94377250 18.410±0.069 18.682±0.090 20.054±0.061 20.069±0.043 19.800±0.058
5 65.00041201 -54.94429422 20.966±0.084 20.575±0.092 21.448±0.067 21.051±0.042 20.376±0.057
6 65.02121621 -54.95053748 17.697±0.069 18.047±0.089 19.506±0.060 19.554±0.042 19.406±0.055
aRightascension(RA)anddeclination(DEC)indecimalnotation(J2000).ThevalueswereobtainedfromtheF555W
frame(alignedandregistered).
bApparent Vega magnitudes and photometric errors. The listed photometry is from a 4-pixel radius aperture and
is corrected to infinite aperture and for foreground Milky Way extinction using E(B-V)=0.008. We use an average
aperturecorrection.
Table 3.Location&photometryofclustersinNGC5253.
ID RAa DECa F275Wb F336Wb F435Wb F550Mb F814Wb
mag mag mag mag mag
5 204.98328 -31.64015 17.048±0.175 17.072±0.175 18.013±0.175 17.938±0.175 16.905±0.175
9 204.9813 -31.64149 16.877±0.175 17.112±0.175 17.851±0.175 17.517±0.175 16.783±0.175
aRightascension(RA)anddeclination(DEC)indecimalnotation(J2000).ThevalueswereobtainedfromtheF336W
image.
bApparent Vega magnitude and photometric error based on luminosities reported in Calzetti et al. (2015b). The
listed photometry is from a 5-pixel radius aperture and is corrected to infinite aperture and for foreground Milky
WayextinctionusingE(B-V)=0.049.
for computing medium- and broad-band magnitudes, and ofthetracks(column5),thereferenceforeachsetoftracks
discusstheimpact of theionized gas on thepredictedmag- (column 6), and the references for the population synthe-
nitudesand colors. sis codes where thedifferent sets of tracksare implemented
(column 7). In this work, we do not account for pre-main
sequence stars. The implications of this approach are dis-
4.1 Stellar component
cussed in section 7.
As previously mentioned, we use cluster ages and
Inthiswork,wefocusonhowupdatesinmodelsofmassive-
massesobtainedbyA16(NGC1566)andC15b(NGC5253)
starevolutionaffectthederivedpropertiesofYMCs.Forthis
with Po and/or Go tracks for selecting the YMCs in our
purpose,wetestdifferentgenerationsandflavorsofmassive-
sample. For a more homogeneous comparison with the rest
star evolution (seven in total), where the different flavors
of models in the present paper, we re-determined Po and
account for different astrophysics. The main differences be-
Go based cluster properties. This is because A16 and C15b
tweenthesemodelslieinthestellarevolutionarytracksand
use different approaches for fitting the observations to the
stellar atmospheres employed.
Bayesian approach used in the present work. In Sections 6
and 7 we present our results and compare them to values
4.1.1 Massive-star evolutionary tracks reported in A16 and C15b.
We compute models using the following massive-star evo-
lutionary tracks: 1) older Padova for single non-rotating
4.1.2 Binary Tracks
stars (Bressan et al. 1993, Z = 0.020; Fagotto et al. 1994,
Z = 0.004); 2) older Geneva for single non-rotating stars TheAucklandbinarytracksusedinthisworkaredescribed
with high-mass loss (Meynet et al. 1994, Z = 0.020 and indetailinEldridgeetal.(inpreparation).Insummary,the
Z = 0.004); 3) new Padova for single non-rotating stars; bpass models are synthetic stellar populations that include
4) newGenevafor singlenon-rotatingstars; 5) newGeneva theevolutionarypathwaysfrominteractingbinarystarsthat
for single rotating stars; 6) new Auckland for single non- rotate.ThestellarmodelsarebasedonaversionoftheCam-
rotatingstars;and7)newAucklandforinteractingbinaries. bridge stars code, originally created by Eggleton (1971)
Hereafter, we will use the IDs provided in the first column andtheversionemployedhereisdescribedinEldridge et al.
of Table 4 to refer to these tracks and to spectral synthesis (2008). The combination of the evolution models into a
models which are based on these tracks. Table 4 also gives synthetic population are described in Eldridge & Stanway
thename of thecity associated with thetracks(column 2), (2009) and Eldridge (2012). However the version used here
adescriptionofthetypeofstellarevolution(column3),the has been significantly improved since these works and are
name of the spectral population synthesis code where each v2.0 to be described in detail in Eldridge et al. (in prepa-
flavor of tracks is implemented (column 4), the metallicity ration). The improvements include increasing the number
(cid:13)c 2002RAS,MNRAS000,1–31
6 A. Wofford
5
1
9
2
3
4 5
N
N
6
E
E
Figure1.WFC3UVISF275WimagesofgalaxiesNGC1566(left)andNGC5253(right).Wemarkthelocationsofclusterswithcircles
of radii 1.6” (∼100 pc, NGC 1566) and 1.25” (∼19 pc, NGC 5253). Photometry was extracted from circles of radii 10 times smaller.
Northisupandeastistotheleft.
Table 4.Massive-starevolutionarytracksandspectralsynthesiscodes.
IDa cityof typeoftrackc spectrald metallicitye Ref.f Ref.g
tracksb synthesis tracks pop.syn.
(1) (2) (3) (4) (5) (6) (7)
Po Padova singlenon-rotating starburst99 0.004/0.020 1 1
Go Geneva singlenon-rotating starburst99 0.002/0.020 2 1
Pn Padova singlenon-rotating galaxev 0.004/0.014 3 2
Gn Geneva singlenon-rotating starburst99 0.002/0.014 4 1
Gr Geneva singlerotatingk starburst99 0.002/0.014 4 1
An Auckland singlenon-rotating bpass 0.004/0.014 4 3
Ab Auckland interactingbinaries bpass 0.004/0.014 4 3
aIDofsetoftracksandmodelsbasedoncorrespondingsetoftracks.Thefirstletteristhe
letter of the city of the tracks. We use ”o”/”n” to designate older/newer versions of the
tracks.
bCitywheretracks werecomputed.
cTypeofevolutionofmassivestars.
dSpectralsynthesiscodewheretracksareimplemented.
eMetallicityoftracks.WeusethelowvalueforNGC5253[12+log(O/H)=8.25]andthehigh
valueforNGC1566[12+log(O/H)=8.63-9.64,dependingoncalibration].Usingthesolar
photosphericabundancesofAsplundetal.(2009),thecorrespondencebetweenmetallicity
andoxygenabundanceis:Z=0.002[12+log(O/H)=7.84],Z=0.004[12+log(O/H)=8.15],
Z=0.014[12+log(O/H)=8.69],andZ=0.020[12+log(O/H)=8.84].
f Reference for the massive-star tracks. 1. Bressanetal. 1993 (Z = 0.020), Fagotto etal.
1994(Z =0.004). 2.Meynetetal.1994(high-massloss).3.Tangetal.2014(Z=0.004);
Chenetal(Z=0.014,2014binprep.;highmetallicity).4.Ekstr¨ometal.2012(Z =0.014),
Georgyetal.2013(Z=0.002).5.Eldridgeetal.2008,inprep.(Z =0.004,Z=0.014).
gReferences and websites of population synthesis codes: 1. (Leithereretal. 2014);
http://www.stsci.edu/science/starburst99/docs/default.htm. 2. (Bruzual&Charlot 2003;
Charlot & Bruzual in prep.; www.iap.fr/ charlot/bc2003). 3. Eldridge et al. (2009, 2012,
inprep.);http://bpass.auckland.ac.nz/index.html
hRotationvelocityis40percentofthebreak-upvelocityonthezero-agemainsequence.
(cid:13)c 2002RAS,MNRAS000,1–31
ATestof SevenWidely-Used SpectralSynthesis ModelsAgainstMulti-Band Photometryof YMCs. 7
of stellar evolution models that can be calculated and the to be main sequence core hydrogen burning stars, rather
numberof stellar models at each metallicity. thanevolved stars (Conti et al. 1995; deKoter et al. 1997).
The models cover a range of masses and include Suchstarshavebeenfoundinclusterswithages 1−3Myr.
mass-loss rates by stellar winds from Vinket al. (2001), In models, these stars have luminosities of log L>6 under
deJager et al. (1988), and Nugis & Lamers (2000). The Galactic metallicity. In galaxev, for the 100 M⊙ tracks for
stellarwindmass-lossratesarescaledfromZ=0.020.Thisis instance,allWRphases(WNL,WNE,andWC)occuratlog
becauseinnearbymassivestarsthemass-lossratesaremore L>6.Inbpass,forahydrogenmassfractionofX >0.4,H-
likely to have a similar composition to other nearby mas- burning WR stars appear mostly as O stars but contribute
sive stars, as deduced by Nieva& Przybilla (2012), which to Heii wind emission lines. For X < 0.4, H-burning stars
is closer to a metal mass fraction of Z =0.020 rather than areincludedbuttherearenotmanyofthem.Theevolution
Z =0.014. Howeverthemasslossofthepopulationisdom- ofthesestarsmightbeverydifferentatlowmetallicities(cf.
inated by binary interactions and so the stellar winds only Hainich et al. 2015).
havea secondary effect. With regards to RSGs, an important effect of binary
The range of initial parameter distributions for the bi- evolutionistoreducethenumberofRSGstoaboutathirdof
nary starsare similar tothose inferred from binary popula- thenumberpredictedbysinglestarevolutionmodels.Inad-
tionsbySanaet al.(2012).Otherenhancementsincludeus- dition, the atmosphere models for RSGs are less important
ingthefullgridfrom thePotsdamWolf-Rayet(WR)model thanhowcoolthestellarmodelsbecome,whichislinkedto
atmosphere grid. Classical Wolf-Rayet stars have lost their the details of the assumed mixing length. Currently, Auck-
hydrogen envelope and can be the evolved descendants of: landbinarymodelsevolvetomuchcoolertemperaturesthan
a)singlestarswithinitialmassaboveathreshold(∼25M⊙ single-starmodels,asthestarsapproachcore-collapse.Work
at Z = 0.020) which increases as metallicity decreases; b) is in progress trying to understand the temperature of re-
secondary starsof interactingbinarysystems whichare ini- solved nearby RSGs as well as the temperatures of RSG
tially less massivethanthislimit andhaveaccreted enough progenitors.
mass to reach it; and c) primary stars of interacting binary
systems initially less massive than this limit, whose hydro-
gen envelope has been transferred to the secondary. The 4.1.4 Ionizing Fluxes of Stellar Populations
modelatmospheregridincludeslowermetallicitiesanddoes
Figure2showstheevolutionoftheHi,Hei,andHeiiioniz-
not rely on an extrapolation of theSolar metallicity grid to
ingrates(numberofphotonspersecond)fordifferentcombi-
theselowermetallicities,asinthepast.Thishasgreatlyim-
nations of stellar evolution model and metallicity. We show
proved the accuracy of the predicted spectra at young ages
whenWRstarsareasignificantcontributiontotheobserved results for SSPs of 106M⊙ in initial stellar mass. The Hi
ionizing rates are higher at low compared to high metallic-
light.
ity because at low metallicity, O stars are hotter and their
mass loss rates lower than at high metallicity. The lower
4.1.3 Stellar Atmospheres mass loss rates make the winds less dense so there is less
windblanketing.Thegreaterwindtransparencymeansthat
In addition to stellar evolutionary tracks, spectral popula-
more ionizing photons escape. Figure 2 also shows that the
tion synthesis requires spectra of individual stars, which in Hi ionizing rate of rotating models can be factors of a few
our case come from various empirical and/or theoretical li-
larger than all othersingle-star models from 3to ∼10 Myr,
braries.Theoreticallibrariesusestellaratmospheresthatare
whichwaspointedoutbyLeitherer et al.(2014)whencom-
characterized by parameters: luminosity, effective tempera-
paring Gr and Gn models. Finally, Fig. 2 shows that,while
ture, mass loss rate, surface gravity and chemical composi-
rotation and binaries both produce more ionizing photons,
tion of the atmosphere. Differences in stellar atmospheres
forGrmodels,theseareconcentratedatayoungage,while
can affect predictions of stellar-population fluxes in broad-
for Ab models, the ionizing flux is sustained to ages older
andmedium-bandfiltersviai)differencesinstellaropacities,
than10Myr.Thisisduetothreemainprocesses: rejuvena-
whichatfixedstellarparameters(effectivetemperature,sur-
tion,mergersandenveloperemoval.Thefirsttwoprocesses
facegravity,andmetallicity)slightlyaffecttheglobalshape
increasethemassofthesecondaryandprimarystarrespec-
of thespectrum; and ii) differences in fluxesof stellar spec-
tively,causingmoremassivestarstooccuratlateragesthan
tral lines, which in our case is not an issue because, a) LE-
expected in a single-star models. This is the dominant pro-
GUS bands do not contain strong stellar lines, and b) we
cessleadingtomorehydrogenionizingphotonsatlatetimes
use low-resolution spectra to convolvewith filters. For each
in thebinary-star models of Fig. 2. The third process leads
population synthesis code used in this work, Table 5 pro-
tothecreationofheliumstars,andlow-luminosityWRstars
vides references for the empirical and theoretical libraries
at later ages than normally possible in single-star models.
and codes used for the massive star spectra in this work.
We include this information for B main-sequence stars, O
main-sequence stars, WR stars, and cooler red supergiants
4.2 Nebular component
(RSG).
ThePoWR atmospheres usedin thiswork includeWR Tocomputethecontributionoftheionizedgastothephoto-
starsthatshowBalmerabsorptionlinesastheonesseenby metricfiltersusedinthispaper,weuseversion13.03ofpho-
Drissen et al. (1995) in NGC 3603, and which are believed toionization codecloudyandadoptthenebularparameters
(cid:13)c 2002RAS,MNRAS000,1–31
8 A. Wofford
Table 5.StellarAtmospheres.
spectral BMS OMS WR RSG
synthesis
galaxev Tlustya Tlusty+WM-Basicb PoWRc Milesd+UVBLUEe
starburst99 ATLAS9f WM-Basic CMFGENg BaSeL v3.1h
bpass BaSeL v3.1 WM-Basic PoWR BaSeL v3.1
aTlusty(Hubeny1988;Hubeny&Lanz1993;Hubenyetal.1994).
bWM-Basic (Pauldrachetal.2001;Smithetal.2002).
cPoWR (Gr¨afeneretal. 2002; Hamann&Gr¨afener 2003; Hamann&Gr¨afener
2004).
dMiles (RSG cooler than 104 K; optical; S´anchez-Bl´azquez etal. 2006;
Falco´n-Barrosoetal.2011)
eUVBLUE (RSG cooler than 104 K; UV; Rodr´ıguez-Merinoetal. 2005;
http://www.inaoep.mx/∼modelos/uvblue/uvblue.html).
f ATLAS9(Kurucz1992)
gCMFGEN(Hillier&Miller1998;Smithetal.2002).
hBaSeL v3.1(Lejeune etal.1997;Westeraetal.1999).
of Zackrisson et al. (2011), which are the nebular parame- and ionized gas. Weconsider E(B−V) values in the range
ters used by C15b and A16. The nebular parameters are: from 0 to 3, in steps of 0.01 mag.
hydrogendensity,n =100cm−3;innercloudradius,R =
H in
100R⊙(L/L⊙)1/2, where L is the bolometric luminosity of
the model stellar population; gas filling factor, ffill = 0.01, 4.4 Synthetic Magnitudes
meaning that the nebula’s Str¨omgren radius, R , is given
S
by R3 = 3Q /(4πn2 f α ), where Q is the number of We obtain magnitudes in the Vega system, m[z,t(z)],
S H H fill B H
by convolving the model spectra at redshift z,
ionizing photons per second and α the case-B hydrogen
B L [λ(1 + z)−1,t(z)], expressed in units of luminosity
recombination coefficient (e.g., Charlot & Longhetti 2001); λ
per unit wavelength, with the system filter through-
sphericalionizednebula,meaningthat∆r∼ R ,where∆r
S
puts, R(λ). For each LEGUS/WFC3/UVIS filter or
isthethethicknessofthenebula;anddust-freenebulawith
close ACS/HRC filter, the throughput curve was down-
nodepletion of elements in thegas onto dust grains. Inour
loaded from the Space Telescope Science Institute website
case,Q issetbythespectralshapeandtotalluminosityof
H
http://www.stsci.edu/∼WFC3/UVIS/SystemThroughput/
theinputSSPspectrum.Foradescriptionofhowtheabove
or http://www.stsci.edu/hst/acs/analysis/throughputs,
nebular parameters relate to the volume averaged ioniza-
respectively. Using the Vega spectrum C(λ) of Bohlin
tion parameter, which is a quantitynot used in the present
(2007), we write (e.g. Bruzual & Charlot 2003)
work but often used to describe models of spherical ionized
nebulae, we refer thereader toequation 3 of Panuzzo et al. ∞
(2003). −R∞dλλLλ([1λ+(z1+)4zπ)−d2L1,(tz()z)]R(λ)
m[z,t(z)]=−2.5log . (1)
∞
R dλλCλ(λ)R(λ)
−∞
4.3 Attenuation by dust
Dust attenuation is applied to the model spectrum prior to
In order to perform a meaningful comparison with the re-
convolution with filter system throughput curves. At fixed
sults of A16 and C15b, we do not include dust in the ion-
metallicity,dustattenuation,andmassive-starevolutionfla-
ized gas (see previous section), but we do account for the
vor,the total numberof models is 9331.
effects of dust on the emergent spectrum, by means of ei-
theranextinctionoranattenuationcurve.Extinctionrefers
to the effect of a uniform dust screen in front of the stars,
4.5 Impact of Ionized Gas on Magnitudes and
parametrized as F(λ) = F(λ) 10[−0.4∗E(B−V)∗k(λ)],
out model Colors
wherek(λ)istheextinctioncurve(e.g.Calzetti et al.2000);
while attenuation includes the effects of scattering on the Figure 3 shows the spectral features captured by the UVIS
absorptionprobabilityofphotonsinmixedstars-gasgeome- and HRC filters in the galaxies NGC 1566 (left panel,
tries(e.g.Wild et al.2011).Foreachgalaxy,wetrythestar- Z = 0.014) and NGC 5253 (right panel, Z = 0.004), as
burstattenuationlawofCalzettietal.(2000,weuseatotal illustrated using 3Myr-old Pn models at the redshifts of
to selective extinction R = 4.05); and an extinction law the galaxies. The vertical axes of both panels use arbitrary
V
based on the galaxy’s metallicity. Specifically, we use the units but identical scaling factors. Differences between the
Milky Way (MW) extinction law of Mathis (1990) for clus- two spectra arise from the dependance of stellar evolution
ters in NGC 1566, and the Small Magellanic Cloud (SMC) on metallicity. In particular, the ionizing rate at 3 Myr is
extinction law of Gordon et al. (2003) for clusters in NGC higher at low compared to high metallicity due to hotter
5253. In all cases, we assume equal attenuation of the stars O stars and less dense winds. Note the relatively stronger
(cid:13)c 2002RAS,MNRAS000,1–31
ATestof SevenWidely-Used SpectralSynthesis ModelsAgainstMulti-Band Photometryof YMCs. 9
Figure 2. Evolution of the number of photons emitted per second in the Hi, Hei, and Heii ionizing ranges (from top to bottom,
respectively)bySSPsofMcl=106M⊙.Weshowpredictionsfrommodelsthatarebasedondifferentstellarevolutionprescriptionsand
metallicities (lines of different styles, as specified in the legends). The low and high metallicity cases are shown on the left and right
panels,respectively.
collisionally-excited oxygen lines at low metallicity com- avoids contamination of broad-band photometry by strong
paredtohighmetallicity.ThisisbecauseatlowZ,theinter- Hβ and [Oiii]lines at the redshift of this galaxy.
stellar gas temperature is higher because of decreased cool-
As previously shown by, e.g., Charlot (1996),
ingfrommetals(Charlot & Longhetti2001).Ascanbeseen Zackrisson et al. (2001), Bergvall & O¨stlin (2002),
inFig.3,theuseofF550MinsteadofF555WforNGC5253 Anders& Fritze-v.Alvensleben (2003), O¨stlin et al.
(cid:13)c 2002RAS,MNRAS000,1–31
10 A. Wofford
Figure 3.Left.Synthetic spectrum correspondingtoanSSP ofage3MyrandmetallicityZ =0.014attheredshiftofNGC 1566(see
Section 4 for more details on the models). The spectrum uses new Padova evolutionary tracks. Prominent lines in the spectrum are
markedwithverticalgraylinesandlabeledatthetop.WeoverlayUVISsystemthroughputs.Right.SimilarbutmetallicityisZ=0.004
andredshiftisthatofNGC5253. WeoverlayUVISandHRCsystemthroughputs.
(2003) and Reines et al. (2010), the ionized gas associated 5 METHOD
with a young stellar population can provide a significant
We use Bayesian inference to constrain model param-
contribution to the total observed flux in a given filter
eters from the observations. For each cluster property,
through the contributions of both emission lines and the
i.e., reddening [E(B − V)], mass (M ) and age (t) we
recombination continuum. Figs. 4 and 5 quantify the cl
record two values, the best-fitting or minimum χ2 value,
contribution of the nebular emission in the filters used to
and the median of the posterior marginalized probabil-
observeNGC5253 and NGC1566, respectively.They show
ity distribution function. We use flat priors in E(B −V),
the magnitude difference as a function of age between a
log(M ) and log(t). Our errors around the median cor-
pure stellar population and stars+ionized gas. The left cl
respond to the 16th and 84th percentiles of the proba-
panel of fig. 11 in Reines et al. (2010) shows Go, Z =0.004
bility density function. The posterior marginalized proba-
predictionsforWFC3UVISfiltersF547MandF814W.The
bility distribution functions are computed as follows. Let
latter predictions can be compared to our Go, Z = 0.004
x = E(B − V), x = M and x = t . The clusters
predictions for close ACS HRC filters F550M and F814W, 1 2 cl 3
are observed in five photometric bands. For a given clus-
whichareplottedinthetop-rightpanelofFig.4.At3Myr,
ter, let y = (y , y , y ,y , y ) be the
our predictions are higher by about 0.2 mag relative to obs obs,1 obs,2 obs,3 obs,4 obs,5
fluxes obtained from the observed reddening-uncorrected
Reines et al.(2010).Thispresumablyarisesfromdifferences
apparent magnitudes, and σ = (σ , σ , σ , σ , σ ) the
in themodels, detectors, and filters. 1 2 3 4 5
corresponding flux errors. In addition, let y =
mod
In Figs. 6 and 7 we show the impact of the ionized (y , y , y , y , y ) represent a set of
mod,1 mod,2 mod,3 mod,4 mod,5
gas on colors for NGC 5253 and NGC 1566, respectively. fluxes obtained from a synthetic library of reddened mag-
Weshowmodelpredictionsforunattenuatedandattenuated nitudesthatmimictheredshiftanddistanceoftheobserva-
spectra, adopting E(B−V) = 0.5 mag in the latter case. tions. The syntheticfluxescorrespond tothe mass in living
We compare results based on starburst and alternative at- stars plus remnants at age t. We infer the marginal poste-
tenuations,asindicated inthecaptions. Weoverlay theob- rior probability distribution, p(x |y ;σ) for the physical
k obs
servations(symbols with errorbars).The rightpanelof fig. parameter x given the observations and errors, using ex-
k
11inReines et al.(2010)canbecomparedwithourGopre- pression:
dictions, which are plotted in the top-right panel of Fig. 6.
ThereisgeneralagreementbetweentheGopredictionspre- nk′ −χ2
sented in this work and in Reines et al. (2010). There are p(xk|yobs;σ)∝ Xexp( x12,x2,x3) (2)
clear differences from models based on different flavors of k′=1
massive star evolution. For instance, theAn models extend
significantly furthertotheupperright ofthecolor-color di- where nk′ is the number of possible values for the physical
agrams compared to other models in Fig. 6. Differences in parameter xk′ (k 6= k′), for a fixed metallicity, prescription
the color-color diagrams due to metallicity can be seen by for attenuation by dust, and set of tracks; and χ2x1,x2,x3 is
comparing the same models in Fig. 6 and 7. The diagrams obtained from:
show the importance of accounting for nebular emission at
5 (y −A · y )2
aaggeess fyoorutnhgeerbitnhaarny1m0oMdeylsr.for most models and even older χ2x1,x2,x3 =X obs,i x1,σx2i2,x3 mod,i (3)
i=1
(cid:13)c 2002RAS,MNRAS000,1–31