The young star cluster system of the Antennae galaxies

0.11 MB·English

by  Peter Anders

AstrophysicsandSpaceScience
DOI10.1007/s•••••-•••-••••-•
The young star cluster system of the Antennae galaxies
Peter Anders1 • Uta Fritze2 • Richard de Grijs3
8
0
0
2

n
a (cid:13)c Springer-Verlag••••
J

3 Abstract Anders and Fritze-v. Alvensleben2003;Bruzual and Charlot
2 Thestudyofyoungstarcluster(YSC)systems,pref- 2003). Therefore,observedspectrophotometricproper-

erentially in starburst and merging galaxies, has seen ties of SCs are relatively easy and straightforward to
]
h great interest in the recent past, as it provides im- interpret.
p portant input to models of star formation. However, SC formation is a major mode of all star forma-
-
o evensomebasicproperties(liketheluminosityfunction tion, and possibly even the dominant mode in strong
r [LF]) of YSC systems are still under debate. Here we starbursts triggered in gas-rich galaxy mergers (e.g.,
t
s study the photometric propertiesofthe YSC systemin Meurer 1995; de Grijs et al. 2003a). In addition, as
a
[ the nearest major merger system, the Antennae galax- SCs inherit and conserve the chemical abundances at
ies. We find evidence for the existence of a statistically the place and time of their birth up to old ages, they
1 significant turnover in the LF.
are excellent tracers of their parent galaxy’s properties
v
9 intermsofstarformationandchemicalenrichmenthis-
Keywords star clusters:general – galaxies: NGC
7 tory.
4038/39– methods: data analysis
6
One of the most basic and commonly used diagnos-
3
tics to explore the properties of entire SC systems is
.
1
1 Introduction their LF. While for old globular cluster systems the
0
Gaussian shape of their LFs (and indeed mass func-
8
0 Starclusters(SCs)formnearlyinstantaneouslythrough tions) is well established (see e.g. Ashman and Zepf
: the collapse of giant molecular gas clouds. Hence, 1998; Harris 1991), the situation for young SCs is still
v
i all stars within a SC are approximately coeval, share under discussion. While mainly LFs consistent with
X the same chemical composition, and therefore rep- a power-law, resembling the MF of nearby molecu-
r resent a simple stellar population. A small num- lar cloud cores, are quoted for YSC systems ranging
a
ber of parameters, in particular their initial chemi- from Galactic open cluster to YSCs in major mergers
cal composition and initial stellar mass function, are (e.g. van den Bergh and Lafontaine1984;Hunter et al.
enough to describe their colour and luminosity evolu- 2003;Schweizer and Seitzer1998;Whitmore et al.1999,
tion on the basis of a given set of stellar evolution-
seede Grijs et al.2003bforarecentcompilation),some
ary tracks or isochrones (e.g. Leitherer et al. 1999;
studies find deviations from a power-law or direct evi-
dence for Gaussian distributions (de Grijs and Anders
Peter Anders 2006;Fritze-v. Alvensleben1999;Goudfrooij et al.2004).
UtaFritze To contribute to our understanding of this astro-
RicharddeGrijs physicalquestion, which ties directly to the fundamen-
1Sterrenkundig Instituut, Universiteit Utrecht, P.O. Box 80000, tal physical conditions of star formation in general, we
NL-3508TAUtrecht,TheNetherlands performed an analysis of the young (∼ 0−100 Myr)
2Centre for Astrophysics Research, University of Hertfordshire, SC system formed during the, still ongoing, merging
CollegeLane,HatfieldAL109AB,UK process in the nearest major merger of two giant gas-
3Department of Physics & Astronomy, The University of rich spiral galaxies, the Antennae galaxies, i.e. NGC
Sheffield,HicksBuilding,HounsfieldRoad,SheffieldS37RH,UK
4038/39.
2
2 Photometry
1
We reanalysed the most homogeneous broad-band
dataset including the U band of the Antennae galaxies
0.8
available (providing photometry in UBVI), obtained
using HST/WFPC2 as part of programme GO-5962
n
o
(PI B. Whitmore). cti 0.6
a
We select our clusters based on: s fr
s
e
• a flux threshold compared to its surrounding eten
• cross-correlation between filters (i.e. a cluster must mpl 0.4
o
c
be detected in several bands)
U
• photometric errors < 0.2 mag in each band B
0.2 V
• size measurements using BAOlab (Larsen 1999) I
XID
XID + photo
Selecting only clusters in the (PSF-corrected) size 50 %
range R1/2 ≃ 5 to 25 pc (to avoid contamination by 020 22 24 26 28 30
single stars and cluster complexes, respectively) yields magnitude
asampleof365clusters,satisfyingallselectioncriteria.
Fig. 1 Completeness fractions for artificial clusters with
Utilisingtheclustersizesweapplysize-dependentaper- FWHM = ≃ 10 pc for a variety of selection criteria (see
ture corrections,as described in Anders et al. (2006). text for details)
different cluster sizes and regions within the galaxies
3 Completeness
contributecompletenesschangesof≃0.6and≃1mag,
In this study we are primarily interested in the LF of respectively. Thisverycomplexdependenceofthecom-
the young star clusters. A consistent and reliable de- pleteness fractions on a number of input parameters
termination of the completeness fraction as function of clearly shows the necessity to model the complete-
cluster magnitude and all other relevant selection cri- ness functions as realistically as possible, taking
teria/effects is therefore of prime importance. into account all cluster selection criteria. Forthe
We investigated the completeness for a number of same region as in Fig. 1, results for different cluster
restrictions, taking successively more cluster selection sizes and selection criteria are shown in Fig. 2.
criteriaintoaccount. Someoftheseresultsforartificial For additional details of the completeness determi-
clusterswithFWHM=≃10pcinacluster-richregion nation, see Anders et al. (2007).
in NGC 4038 are shown in Fig. 1.
First, we determined the completeness in each band
independently (labelled ”U”,”B”, ”V” and ”I” in Fig. 4 The model to analyse the luminosity
1). Taking the cross-correlation into account (labelled functions
”XID”) reduced the completeness to slightly below the
most limiting single-band completeness. By apply- In collaboration with statisticians from the University
ing the photometric uncertainties restriction (labelled of G¨ottingen we developed a suite of statistical tools.
”XID + photo”) the 50% completeness magnitude is ThisallowstofittheLFsofouryoungstarclusterswith
decreased by about ≃0.4 mag. either a Gaussian distribution or a power-law, access
Using stars instead of extended objects as test the accuracy of the fits and the fitted parameters, and
sourcesartificiallyincreasesthe50%completenessmag- establishthestatisticalsignificanceofdifferencesinthe
nitude by ≃1 mag. goodness-of-fitbetweenthetwofittedtestdistributions.
Computing completeness functions for 2 regions The fitting of a observed LF takes the photomet-
(characterisedbydifferentbackgroundandclusterden- ric error and appropriate completeness fractionof each
sity levels) and 2 sizes (characteristic for our observed individual cluster into account, and independently fits
cluster sample) and imposing the size restriction onto either a Gaussian distribution or a power-law to the
the artificial clusters (labelled ”XID photo size”) al- LF,yieldingasetofparameterscharacterisingthebest-
lowedustoattributerealisticcompletenessfractionsto fitting model anda likelihoodparameter for either test
each individual cluster, strengthening the subsequent distribution.
analysis of the LFs. Imposing the size restrictions We find a strong superiority (measured by the ratio
decrease the completeness by up to ≃ 1 mag, while ofthelikelihoodsforGaussianandpower-lawfits)ofthe
YSCsintheAntennae galaxies 3
1
1pix XID
1pix XID phot
1pix XID phot size
2pix XID
0.8 2pix XID phot
25p0i x% X ID phot size a.u.]
on F [
cti 0.6 L
a
s fr
s
e
n
e
et
pl 0.4
m
o
c
−14 −12 −10 −8
0.2
0
20 22 24 26 28 30
magnitude
Fig. 2 Completeness fractions for artificial clusters with
FWHM = ≃ 10 pc (= 1pixel) and FWHM = ≃ 20 pc (=
2pixel)foravarietyofselectioncriteria(seetextfordetails)
Gaussianfitcomparedtothepower-lawfit(seeSect. 5).
−14 −12 −10 −8
Tovalidatethisresult,wehavetotestitsstatisticalsig-
nificance, i.e. to determine the probability to achieve
such strong superiority of the Gaussian distribution if Fig.3 V-bandluminosityfunctionsforourAntennaeclus-
the trueunderlyingLFisapower-law,distortedbythe tersandfit: Observed(greensolidline),bestfit(blacksolid)
and underlying distribution of best fit (black dashed). Fit-
error distribution and the completeness function. In a
tedtest distributionsare: Gaussian (left panel)and power-
Monte-Carlo approach we draw artificial observations
law (right panel)
from the best-fitting power-law,use our statistics tools
tofitthemagainwithGaussianandpower-lawdistribu-
tion, anddetermine the likelihoodratioforthese artifi- the Gaussian fit when compared to the power-law fit).
cial observations. The likelihood ratio of the observed The Monte-Carlo analysis with 1000 test realisations
distribution compared to the distribution of likelihood drawn from the best-fitting power-law distribution re-
ratios from our artificial tests then yields the probabil- sultsinamaximumlikelihoodratioof11.4,hencenone
ity thatthe determinedsuperiorityofthe Gaussianfits of the test distributions can reproduce a superiority of
isconsistentwithanunderlyingpower-lawdistribution. the Gaussian distribution as strong as observed. Util-
The uncertainties of the best-fitting model parame- ising the properties of Bernoulli-distributed variables,
ters are estimated by bootstrapping. this corresponds to a probability < 0.5% that the un-
For additional details of the statistical model, see derlying distribution is still consistent with a power-
Anders et al. (2007). law.
This result is valid also for the other bands (except
the I band, for which the observations are significantly
5 Results shallower),andforseveralageandsizesubsamples. For
more details, see Anders et al. (2007).
As example, the LF in the V band for our YSC sam-
ple in the Antennae galaxies is shown in Fig. 3. From Acknowledgements WewouldliketothankNicolai
visualinspection,thebestfitwithanunderlyingGaus- Bissantz and Leif Boysen for their constant help and
sian distribution appears to represent the data better development of the statistics tools.
than the best fit with an underlying power-law distri-
bution.
Toquantifythis,weperformourlikelihoodratiotest
and the Monte-Carlo analysis. For the observed distri-
butions we find a likelihood ratio value of 23.1 (where
largervalues are equivalent to a stronger superiority of
4
References
Anders, P., Bissantz, N., Boysen, L., de Grijs, R., Fritze-
v. Alvensleben, U.: Mon. Not. R. Astron. Soc. 377, 91
(2007)
Anders, P., Fritze-v. Alvensleben, U.: Astron. Astrophys.
401, 1063 (2003)
Anders, P., Gieles, M., de Grijs, R.: Astron. Astrophys.
451, 375 (2006)
Ashman,K.M.,Zepf,S.E.: GlobularClusterSystems.Cam-
bridge astrophysics series ; 30. Cambridge University
Press, Cambridge, U.K. ; New York (1998)
Bruzual, G., Charlot, S.: Mon. Not. R. Astron. Soc. 344,
1000 (2003)
de Grijs, R., Anders, P.: Mon. Not. R. Astron. Soc. 366,
295 (2006)
deGrijs,R.,Anders,P.,Bastian,N.,Lynds,R.,Lamers,H.J.G.L.M.,
O’Neil, E.J.: Mon. Not. R. Astron. Soc. 343, 1285
(2003b)
deGrijs,R.,Lee,J.T.,ClemenciaMoraHerrera,M.,Fritze-
v. Alvensleben, U., Anders, P.: New Astronomy 8, 155
(2003a)
Fritze-v. Alvensleben, U.: Astron. Astrophys. 342, L25
(1999)
Goudfrooij,P.,Gilmore,D.,Whitmore,B.C.,Schweizer,F.:
Astrophys.J. Lett.613, L121 (2004)
Harris, W.E.: Annu. Rev. Astron. Astrophys. 29, 543
(1991)
Hunter, D.A., Elmegreen, B.G., Dupuy, T.J., Morton-
son, M.: Astron. J. 126, 1836 (2003)
Larsen, S.S.: Astron. Astrophys. Suppl. Ser. 139, 393
(1999)
Leitherer,C.,Schaerer,D.,Goldader,J.D.,Delgado,R.M.G.,
Robert,C.,Kune,D.F.,deMello,D.F.,Devost,D.,Heck-
man, T.M.: Astrophys.J. Suppl.Ser. 123, 3 (1999)
Meurer, G.R.: Nature 375, 742 (1995)
Schweizer, F., Seitzer, P.: Astron. J. 116, 2206 (1998)
van den Bergh, S., Lafontaine, A.: Astron. J. 89, 1822
(1984)
Whitmore, B.C., Zhang, Q., Leitherer, C., Fall, S.M.,
Schweizer,F.,Miller,B.W.: Astron.J. 118,1551(1999)
ThismanuscriptwaspreparedwiththeAASLATEXmacrosv5.2.