Cluster and nebular properties of the central star-forming region of NGC 1140
S. L. Moll, S. Mengel, R. de Grijs, L. J. Smith, P. A. Crowther
aa r X i v : . [ a s t r o - ph ] S e p Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 27 October 2018 (MN L A TEX style file v2.2)
Cluster and nebular properties of the central star-formingregion of NGC 1140 ⋆ S. L. Moll † , S. Mengel , R. de Grijs , , L. J. Smith , and P. A. Crowther Department of Physics and Astronomy, University of Sheffield, Sheffield S3 7RH European Southern Observatory, D-85748 Garching, Germany National Astronomical Observatories, Chinese Academy of Sciences, Beijing 100012, China Space Telescope Science Institute and European Space Agency, Baltimore, MD 21218, U.S.A. Department of Physics & Astronomy, University College London, London WC1E 6BT
ABSTRACT
We present new high spatial resolution
HST /ACS imaging of NGC 1140 and highspectral resolution VLT/UVES spectroscopy of its central star-forming region. Thecentral region contains several clusters, the two brightest of which are clusters 1 and6 from Hunter, O’Connell & Gallagher, located within star-forming knots A and B,respectively. Nebular analysis indicates that the knots have an LMC-like metallicity of12 + log O / H = 8 . ± .
09. According to continuum subtracted H α ACS imaging,cluster 1 dominates the nebular emission of the brighter knot A. Conversely, negligiblenebular emission in knot B originates from cluster 6. Evolutionary synthesis modellingimplies an age of 5 ± . ± . × M ⊙ is obtained. For this age and photometric mass, the modelling predictsthe presence of ∼ N (WR) /N (O) ∼ .
1, assuming that all the WR stars are located within cluster 1. Thevelocity dispersions of the clusters were measured from constituent red supergiants as σ ∼ ± − for cluster 1 and σ ∼ ± − for cluster 6. Combining σ withhalf-light radii of 8 ± . ± . ± × M ⊙ and (9 . ± . × M ⊙ for clusters 1and 6, respectively. The most likely reason for the difference between the dynamicaland photometric masses of cluster 1 is that the velocity dispersion of knot A is notdue solely to cluster 1, as assumed, but has an additional component associated withcluster 2. Key words: galaxies: individual: NGC 1140 – galaxies: starburst – galaxies: starclusters – stars: Wolf-Rayet
Violent bursts of star formation, which are characteristic ofstarburst galaxies, resemble the star-forming phase of younggalaxies in the early Universe. Nearby starbursts provide lo-cal templates to which distant star-forming galaxies maybe directly compared. Only a handful of starburst galax-ies are located within 10 Mpc, yet they produce around aquarter of the entire high-mass star population (Heckman ⋆ Based on observations collected at the European Southern Ob-servatory, Chile, under programme ESO 71.B-0058(A), and onobservations obtained with the NASA/ESA
Hubble Space Tele-scope , which is operated by the Association of Universities forResearch in Astronomy, Inc., under NASA contract NAS 5-26555. † E-mail:s.moll@sheffield.ac.uk ∼ c (cid:13) S. L. Moll et al. for an overview). Understanding the role of YMCs as can-didate proto-GCs is vital for our understanding of galaxyformation and evolution, as well as large-scale star forma-tion. Clusters also have the advantage of being simple tomodel – they can be approximated as a coeval, simple stel-lar population with a single metallicity. This makes themideal as probes of burst properties such as age, duration,chemical evolution and star-formation rate as well as con-straining the parameters of the stellar initial mass function(IMF).NGC 1140 is a low-metallicity WR galaxy at a dis-tance of ∼
20 Mpc and is a prime example of a nearbyanalogue of the star-forming galaxies identified at highredshifts. HI and optical observations indicate that thegalaxy has undergone a merger within the past 1 Gyrand it is thought that this event is responsible for theplethora of YMCs that the galaxy hosts. Star clustershave been imaged with both the Wide Field/PlanetaryCamera (WF/PC) and the Wide Field Planetary Cam-era 2 (WFPC2) on the Hubble Space Telescope (HST) byHunter, O’Connell & Gallagher (1994a) and de Grijs et al.(2004). They identified eight young luminous clusters in thecentre of the galaxy, and their studies indicated that theclusters have masses of up to a few × M ⊙ and ages thatlie in the range of a few to a few tens of Myr.This paper considers the properties of the two bright-est clusters within NGC 1140, clusters 1 and 6, and the twostar-forming knots in which these clusters are contained. Itis structured as follows. Details of our VLT and HST obser-vations and data reduction are given in Section 2, along withphotometry of the clusters. In Sections 3 and 4, we discussthe stellar and nebular properties of the two knots apparentin the VLT spectra, and in Section 5 we present the resultsof evolutionary synthesis modelling of cluster 1. The mas-sive star population of cluster 1 and of both knots A and B,and the star-formation rate of NGC 1140, are considered inSection 6. The dynamical masses of the clusters are deter-mined in Section 7 and we discuss our findings in Section 8.Finally, we summarise our results in Section 9.
We obtained high-resolution spectroscopy of the central re-gion of NGC 1140 with the VLT/Ultraviolet and VisualEchelle Spectrograph (UVES), in addition to high spatialresolution
Hubble Space Telescope (HST) / Advanced Cam-era for Surveys (ACS) imaging of NGC 1140. Fig. 1 showsthe region of the galaxy observed with UVES and labelsclusters 1 −
7, as designated by Hunter et al. (1994a). Itshows the location of the 1 ×
11 arcsec UVES slit, which wasaligned north-south over the two knots in the central regionof NGC 1140, hereafter called knots A and B. Knot A, whichcontains the clusters 1 and 2, is the brightest region in theoptical and near infrared. Knot B lies ∼ Based on the heliocentric velocity of the galaxy, cor-rected for the Virgocentric flow, and assuming H =70 km s − Mpc − ; adopted from the HyperLeda database athttp://leda.univ-lyon1.fr/
Table 1.
F625W ACS aperture stmag photometry of the bright-est central clusters of NGC 1140 and the regions designated knotsA and B. The F300W and F814W cluster magnitudes are takenfrom WFPC2 imaging of de Grijs et al. (2004).Cluster m F300W m F625W m F814W (mag) (mag) (mag)Cluster 1 16 . ± .
05 17 . ± . . ± . . ± .
05 18 . ± . . ± . . ± .
09 18 . ± . . ± . . ± .
51 19 . ± . . ± . . ± . . ± . . ± . . ± . . ± . . ± . NGC 1140 was observed on 21 August 2003 with ACS/WFCaboard
HST as part of GO programme 9892 (PI Jansen).Two exposures were taken – one of 950s duration with thenarrow-band F658N filter and one of 100s duration usingthe broad-band F625W filter. The data were reduced usingthe on-the-fly reduction pipeline.At the distance of NGC 1140, the F658N filter doesnot include the [N II] 6583 line, and, therefore, is essentiallyan H α plus continuum filter. We have corrected for the un-derlying continuum by subtracting the scaled F625W imagesuch that a background galaxy is removed. This continuumsubtracted F658N is included in Fig. 1. The high resolu-tion of the ACS data also shows that clusters 3 and 5 fromHunter et al. (1994a) are not single clusters.Aperture photometry was carried out on the drizzledF625W ACS image in Starlink’s gaia . Since the crowdednature of the field makes it difficult to position aperturesto include all of the light from the desired cluster withoutcontamination from its neighbours, several circular aper-tures with a range of radii and positions were consideredfor each region. Each measurement was sky subtracted us-ing a mean sky value obtained from several small aperturesof sky, and aperture corrected using the values contained inSirianni et al. (2005). As these aperture corrections are validfor point sources rather than for spatially extended clusters,our photometry may be an underestimate. However, sincethe light profile of the clusters is difficult to ascertain (seeSection 7.1), no correction for this effect is made.Table 1 presents our F625W photometry. The associ-ated uncertainty represents the range of results produced.Table 1 also contains F300W and F814W photometry.The F300W and F814W photometry of knots A and Bwere determined using archival WFPC2 images (GO 8645,PI Windhorst) using the same method as described forthe F625W images. The F300W and F814W cluster pho-tometry was taken from de Grijs et al. (2004). The pre-COSTAR WF/PC photometry of de Grijs et al. (2004) andHunter et al. (1994a) was not included due to its lower spa-tial resolution. The central region of NGC 1140 was observed with UVESon the VLT Kueyen Telescope (UT2) in Chile on 19-20September 2003. UVES is a cross-dispersed echelle spec- c (cid:13) , 000–000 luster and nebular properties of NGC 1140 Sky Knot A Sky Knot B SkyF625W F658N1 23 4 5 6 7 F625W N E N E E N Figure 1.
HST /ACS images of NGC 1140, at different cut levels (1 arcsec ∼
100 pc), marked as being either the F625W image or thecontinuum subtracted F658N image. The left-hand image shows the whole galaxy. The white box indicates the region that is magnifiedin the images in the right-hand panel. The lighter diagonal band across the image is the inter-CCD gap between the two detectors ofACS. Superimposed on each right-hand image is the 1 ×
11 arcsec VLT/UVES slit. The regions extracted as knots A and B and as skyare indicated by the solid and dashed lines, respectively. Clusters 1 − trograph, with two arms – a blue arm comprising a singleEEV CCD and a red arm comprising a mosaic of an EEVand a MIT-LL CCD. Using the dichroic beam splitter, redand blue data were taken simultaneously: cross-disperser l mm − was centred on 840 nm and cross-disperser l mm − was centred on 437 nm. Thus, datafrom the regions ∼ − ∼ − ∼ − × − and the blue data have ascale of 0.512 arcsec pixel − .The galaxy was observed with UVES for a total timeof 14700s on the first night and 16200s on the second night.The seeing averaged ∼ . ∼ . ∼ − .The data were reduced using the software package iraf .The data were bias corrected with a median bias frame, anddivided by a normalised flat field. The two knots were ex-tracted, each with an aperture of ∼ . Comparison red supergiant (RSG) spectra are required todetermine a cluster’s velocity dispersion based upon its RSGfeatures, which can then be used to calculate its dynamicalmass. Therefore, six Small Magellanic Cloud (SMC) red su-pergiants, taken from the Massey & Olsen (2003) catalogue,were observed with the setup described above. The catalogueID number of these stars, and their spectral types were 20133(M0 I), 64448 (K2-7 I), 71566 (K7 I), 50840 (M1-2 I), 66694(K5 I) and 71507 (K3-5 I).
Red supergiant and OB stellar features are visible in theUVES spectra of both knots. Wolf-Rayet (WR) features arealso present in the spectrum of knot A.
Knots A and B both show stellar absorption in the hydrogenBalmer series, arising from the presence of early-type stars,underlying strong nebular emission. The equivalent widths c (cid:13) , 000–000 S. L. Moll et al.
Figure 2.
Blue spectrum of knot A with the principal nebularlines and the Wolf-Rayet blue bump marked, corrected to thestellar rest-frame. Broad stellar absorption is apparent underlyingthe narrow nebular Balmer emission. of the underlying stellar absorption are W λ (H β ) ∼ λ (H γ ) ∼ λ (H δ ) ∼ λ λ λ ∼ . ∼ . ± Figure 3.
The normalised Ca II triplet lines of knot A (top) andknot B (bottom) after correction for a recession velocity of v r =1475 km s − . The λ λ
862 lines are offset by 0.5 and1.0, respectively. Nebular Paschen emission lines are indicated. of RSG features in the spectrum (see Section 3.2) makes itunlikely that the knot is younger than 3 Myr, for the caseof an instantaneous burst.
The Ca II triplet absorption lines ( λλ , , elf (emission-line fitting) rou-tine in Starlink’s dipso package. The line centres, widthsand intensity of the fits were allowed to vary freely, and themeasured line centres used to determine the recession ve-locity of the galaxy. A good fit could not be obtained tothe λ λ v r ∼ ± − . Our result is c (cid:13) , 000–000 luster and nebular properties of NGC 1140 in good agreement with other recent values of recessional ve-locity, such as v r = 1498 ±
33 km s − from optical measure-ments (de Vaucouleurs et al. 1991), v r = 1501 ± − from 21-cm HI observations (Haynes et al. 1998) and v r =1480 ± − from [Fe II] measurements (de Grijs et al.2004).After velocity correction, the equivalent widths of theCa II triplet lines were measured with elf , fixing the centralwavelengths of the lines. This yields values of W λ (8498) =0 . ± . λ (8542) = 1 . ± . λ (8662) =0 . ± . λ (8498) = 0 . ± . λ (8542) = 2 . ± . λ (8662) = 1 . ± . In this section we derive information on the gas dynamics ofthe knots from the profiles of the nebular lines. Extinctions,electron densities, temperatures and elemental abundancesare determined from the relative fluxes of the nebular lines.
Fig. 4 shows that the nebular H β profiles of knot A andknot B are very different. Knot A is dominated by one com-ponent at approximately v ≈ −
15 km s − , but also has aweaker underlying broad component. Knot B more clearlycomprises two components: a weaker component blueshiftedat v ≈ −
36 km s − with FWHM ∼
58 km s − and a brighterredshifted component at v ≈
28 km s − with FWHM ∼
47 km s − . These nebular profiles are representative of allthe strong nebular lines of the knots. Examination of the 2Dimage indicates that there is diffuse emission throughout thecentral region of the galaxy that probably corresponds to thebluer, weaker component of knot B and to the weak, broadcomponent of knot A. The stronger components of the knotsarise from more discrete emission from the clusters withinthese knots, as are visible in Fig. 1. Alternatively, the broadunderlying component of the knot A profiles may be due tohot cluster winds impacting on the surrounding interstellarclouds, as has been seen in other young star-forming regions(see Sidoli et al. 2006 for a review and Westmoquette et al.2007 for a detailed discussion). Turbulent broadening of theorder of 30 km s − was implied from the FWHMs of the maincomponents of the strong nebular lines of knots A and B. Nebular line fluxes were measured by fitting the nebularemission with Gaussian profiles using the elf routine, al-lowing line centres, widths and intensity to vary freely. Thestrong nebular emission of both knots A and B was fittedwith two Gaussians (Section 4.1).The internal extinctions of the knots, E ( B − V ) int ,were determined by comparing the observed flux ratios ofF(H γ ) / F(H β ) and F(H δ ) / F(H β ) to the intrinsic intensityratios predicted by assuming Case B recombination theoryfor electron densities of 10 cm − and a temperature of 10 K(Hummer & Storey 1987). As the diffuse emission contained
Figure 4.
The H β profiles for knots A and B, after correction to v r = 1475 km s − . Table 2.
Comparison between predicted intrinsic flux ratios,I λ / I(H β ) and observed flux ratios, F λ / F(H β ). For the fluxesquoted here, both components of the knot A profile were summed,while only the strong redshifted component of knot B was consid-ered. The observed ratios were first dereddened with a standardGalactic extinction law. Internal extinctions, E ( B − V ) int , werethen derived from predicted intensity ratios assuming a Howarth(1983) LMC extinction law.Line F λ / F(H β ) I λ / I(H β ) E ( B − V ) int (Observed) (Predicted) (mag)Knot A H γ . ± .
021 0.468 0 . ± . δ . ± .
011 0.259 0 . ± . γ . ± .
021 0.468 0 . ± . δ . ± .
011 0.259 0 . ± . within knot B is expected to have a very low extinction com-pared to the discrete emission from the clusters, the nebu-lar extinction was calculated based only on the stronger,redder component of knot B. Since the division betweencomponents is less well defined for knot A, both compo-nents were summed. The observed flux ratios were first cor-rected for foreground extinction of E ( B − V ) fore = 0 . γ and H δ to H β were considered since H α was not fluxcalibrated. Mean values of E ( B − V ) int = 0 . ± .
06 magfor knot A and E ( B − V ) int = 0 . ± .
08 mag for knot Bwere subsequently adopted.Table 3 presents observed and extinction corrected linefluxes, normalised to H β = 100. Here, both components weresummed for both knots, as the total emission within theknot was of interest. Lines with total fluxes of < .
2% of H β were not included. We concentrate on deriving the electrondensity, temperature and nebular abundances for knot A be-cause of the added complications introduced by the nebularstructure present in Knot B. c (cid:13) , 000–000 S. L. Moll et al.
Table 3.
Observed (F λ ) and intrinsic (I λ ) nebular line fluxes rel-ative to H β = 100 for knot A and knot B. The fluxes quotedhere are the summation of both components of the nebular lines,for both knots A and B. The intrinsic values are corrected for aforeground extinction of E ( B − V ) fore = 0 .
038 mag with a stan-dard Galactic extinction law and then for internal extinctions of E ( B − V ) int = 0 . ± .
06 mag for knot A and E ( B − V ) int =0 . ± .
08 mag for knot B using the Howarth (1983) LMC ex-tinction law. The flux of H β is in units of × − erg s − cm − .Knot A Knot BLine F λ I λ F λ I λ λ ± ± ± ± λ ± ±
10 110 ± ± λ ± ± ± ± λ ± ± ± ± λ ± ± ± ± λ ± ± ± ± λ δ ± ± ± ± λ γ ± ± ± ± λ ± ± ± ± λ ± ± ± ± λ ± ± ± ± λ ± ± ± ± λ ± ± λ β
100 100 100 100 λ ± ± ± ± β ± ± ± ± The electron density, N e , and electron temperature, T e , ofknot A were determined in the five-level atom calculator temden within iraf using the diagnostic line ratios of [O II]I( λ / I( λ λ / I( λ N e = 60 ±
50 cm − (consistent with the low densitylimit) and T e = 9700 ±
300 K.
Knot A abundances were calculated from standard [O II]and [O III] diagnostics, plus the values of N e and T e deter-mined in Section 4.3. These yield values of N (O + ) /N (H + ) =(8 . ± . × − and N (O ) /N (H + ) = (1 . ± . × − . These imply an abundance for knot A of12 + log O / H = 8 . ± .
09. We adopt this value forNGC 1140. It lies between the abundance of the Small Mag-ellanic Cloud (SMC) of 12 + log O / H = 8 .
13 and that ofthe LMC of 12 + log O / H = 8 .
37 (Russell & Dopita 1990).It agrees well with other direct abundance measurementsof NGC 1140 of 8 . ± .
06 (Izotov & Thuan 2004) and8 . ± .
07 (Nagao, Maiolino & Marconi 2006; recalculatedfrom the measurements of Izotov & Thuan 2004).As the determined abundance is most similar to theLMC abundance, we adopt an LMC-like metallicity with Z = 0 .
008 for spectral modelling purposes.
Table 4.
Equivalent widths and knot ages implied from theStarburst99 model. The equivalent width of Ca II is defined asW λ ( λ λ ( λ W λ (˚A) Age (Myr) W λ (˚A) Age (Myr)H α ±
29 5.5 ± ±
14 6.0 ± β ± ± ± ± ± ± ± > Having established that knot A has an LMC-like metallic-ity and an age of around 5 Myr, we computed v5.1 Star-burst99 evolutionary synthesis models (Leitherer et al. 1999;V´azquez & Leitherer 2005) for an instantaneous burst ofstar formation, with a total stellar mass of 10 M ⊙ and ametallicity of Z = 0 .
008 for ages between 1 and 10 Myr in0.5 Myr intervals. We adopted a Kroupa (2002) IMF (withslope α = 2 . . M/ M ⊙
100 and α = 1 . . M/ M ⊙ < .
5) and Padova stellar evolutionary tracks(Fagotto et al. 1994), which include careful consideration ofthe RSG phase.
The Starburst99 model predicts the equivalent widths forH α , H β and the Ca II triplet. The equivalent widths mea-sured for each knot and the ages implied from the Star-burst99 model are presented in Table 4.Fig. 1 and Table 1 help to put these equivalent widthsin context by showing the regions in which the continuumlight and the nebular emission originate. In knot A, both thestellar continuum and nebular emission are dominated bycluster 1, indicating that the H α and H β equivalent widthsare meaningful and can be applied to cluster 1. On the otherhand, the stellar continuum of knot B is dominated by clus-ter 6, while the nebular emission is due solely to cluster 5 anda second source of emission ∼ . − ∼ ∼
10 Myr, unless any gas associated with theseclusters has been removed by multiple supernovae at an evenearlier stage. The equivalent widths of the Ca II triplet forboth knots A and B are meaningful and can be applied tocluster 1 and 6, respectively, since these clusters dominatethe continuum in the region. However, only a lower limit tothe age of cluster 6 can be determined from the equivalentwidth of the Ca II triplet, because it does not vary stronglywith age after a few Myr. Furthermore, the contributionof RSGs to cluster evolution is not well understood (e.g.Massey & Olsen 2003), and as such RSGs do not provide areliable cluster age indicator.After considering both the age indication of 4 ± c (cid:13) , 000–000 luster and nebular properties of NGC 1140 Figure 5.
Fit of the reddened 5 Myr old model spectrum (Star-burst99) to ACS and WFPC2 photometry for cluster 1. Themodel spectrum was first reddened by the foreground extinctionof E ( B − V ) fore = 0 .
038 mag (Schlegel et al. 1998) using a stan-dard Galactic reddening law and then further reddened with a(Howarth 1983) LMC extinction law to best fit the plotted pho-tometry. age of 5 ± V − I colours,Hunter et al. (1994a) estimated the ages of the clusters inthe region of knot A as ∼ ∼
15 Myr, since this wasthe age determined by O’Connell, Gallagher III & Hunter(1994) for the YMCs in NGC 1569 and NGC 1705with the same colours. The near infrared CIRPASS dataof de Grijs et al. (2004) showed strong [Fe II] emissionthroughout the galaxy, while strong Br(12-4) and Br(14-4) was predominantly confined to knot A. Since the diffuse[Fe II] emission is likely associated with supernova remnants,while the Brackett nebular emission lines are associated withHII regions, de Grijs et al. (2004) argued that the ratio ofthese lines is a good age indicator. This would suggest thatknot B is several Myr older than knot A.
The model spectrum produced by the Starburst99 model fora 5 Myr old instantaneous burst was reddened to reproducethe ACS and WFPC2 photometry of cluster 1 (Table 1).The best fit, presented in Fig. 5, yields a value of inter-nal cluster reddening of E ( B − V ) int ∼ . ± .
04 magfor cluster 1. The uncertainty quoted here solely considersthe photometric uncertainties. The F300W and F625W pho-tometry was more heavily weighted than the F814W pho-tometry, since these are more sensitive to reddening. Ourvalue agrees well with the result of de Grijs et al. (2004) of E ( B − V ) int = 0 . − .
26 mag for cluster 2, determinedfrom spectral energy distribution fits to their WF/PC andWFPC2 photometry. It also agess with the nebular extinc-tion determined from the Balmer lines, given in Table 2.
The F625W magnitude for cluster 1 was converted intoan apparent V -band magnitude using the transformationin Sirianni et al. (2005), assuming the Starburst99 model V − R colour for a 5 Myr old burst and adopting the clus-ter extinction determined in Section 5.2. A distance of 20Mpc to NGC 1140 then implies a V -band luminosity of L V = (7 . ± . × L ⊙ for cluster 1. Comparing thisluminosity to the Starburst99 prediction yields a photomet-ric mass of (1 . ± . × M ⊙ . This is based on a KroupaIMF. In this section we determine the O star content of the knotsfrom the H β luminosity of the UVES spectroscopy. We es-timate the WR content of knot A from the blue bump seenin its spectrum and measure the star formation rate (SFR)of NGC 1140 from the continuum subtracted F658N ACSimaging. An estimate of the number of O stars within an individ-ual knot can be obtained from the H β luminosity, L (H β ).Assuming Case B recombination theory (Hummer & Storey1987) and that a ‘normal’ stellar population is the onlysource of ionising photons within the knot, the number ofequivalent O7V stars, N (O7V), contained within the knotis given by: N (O7V) = Q Obs0 Q O7V0 = L (H β ) × . × Q O7V0 (see e.g. Vacca 1994). Here, Q Obs0 is the observed totalLyman continuum luminosity of the knot and Q O7V0 isthe Lyman continuum flux of an individual O7V star.distance of 20 Mpc and using the flux values given inTable 3, we obtain L(H β ) = (5 . ± . × erg s − for knot A and L(H β ) = (3 . ± . × erg s − forknot B. Taking Q O7V0 = 10 . photons s − , as suggested byHadfield & Crowther (2006) for LMC metallicity O stars,implies that knot A contains 1300 ±
500 equivalent O7Vstars and that knot B contains 800 ±
500 equivalent O7Vstars. The Starburst99 modelling predicts an H β luminos-ity for cluster 1 of (4 . ± . × erg s − , showing thatcluster 1 dominates knot A in terms of its H β luminosity.This implies the presence of 1200 ±
300 equivalent O7V starswithin cluster 1.This number of equivalent O7V stars can be convertedinto the total number of O stars, N (O), using the time-dependent parameter, η ( t ): N (O) = N (O7V) η ( t ) . For an age of 5 Myr and a metallicity of Z = 0 . η = 0 .
2, implying that clus-ter 1 contains 5900 ± η = 0 . ± c (cid:13) , 000–000 S. L. Moll et al.
Figure 6.
The λ λ O stars. For knot B, which is dominated by the young clus-ter 5, the total number of O stars is comparable to the num-ber of equivalent O7V, since η ∼ < The number of WR stars within knot A can be obtainedby considering the luminosity of the blue WR bump. Thesimilarity in the strengths of N III 4640 and He II 4686in knot A suggests the presence of late-type WN (WNL)stars. The luminosity of λ . × erg s − , aftercorrecting for an extinction of E ( B − V ) int = 0 .
24 mag anda distance of 20 Mpc. Comparing this to the average lu-minosity of LMC WN7-9 stars of (7 . ± . × erg s − (Crowther & Hadfield 2006), indicates that knot A contains950 WNL stars, assuming no contribution to the blue bumpfrom WC stars. However, Guseva, Isotov & Thuan (2000)identify early-type WC (WCE) stars in their long-slit spec-trum of the entire central region of NGC 1140, by the pres-ence of C IV 5808 emission. Therefore, we empirically matchthe blue bump using various multiples of an LMC WN7-9star, plus an LMC WC4 star, finding a best fit of approx-imately 550 ±
100 WNL stars and 200 ±
50 WCE stars.This is shown in Fig. 6. We assume that all of the WR starswithin knot A are contained in cluster 1, since it is the mostmassive young cluster within this region.
We measure an observed H α flux of F(H α ) = (1 . ± . × − erg s − cm − for the whole galaxy, based upon aper-ture photometry of the continuum subtracted F658N image(see Section 2.1). Assuming a distance of 20 Mpc and a meanextinction of E ( B − V ) int = 0 .
19 mag, we find an H α lu-minosity of NGC 1140 of (9 ± × erg s − . This meanextinction was determined as an average of the nebular ex-tinctions of knots A and B (Section 4.2) weighted by the relative contributions to the H β flux of the central region of2:1 (taken from Table 3). Using the equation:SFR(M ⊙ yr − ) = 7 . × − × L(H α )(erg s − )= 1 . × − × Q Obs0 (photons s − )(Kennicutt 1998), we estimate the SFR of thegalaxy of 0 . ± . ⊙ yr − , in good agreementwith the SFR of 0 . ⊙ yr − , obtained fromground-based continuum subtracted H α imaging byHunter, van Woerden & Gallagher (1994b), after adjust-ment to our adopted distance. Our H α luminosity gives Q Obs0 = (6 . ± . × photons s − or 9000 ± The virial motions of the individual stars within a cluster,as measured by the line-of-sight velocity dispersion, give adynamical estimate of the cluster mass. This also requires aknowledge of the half-light radius of the cluster, which is thetwo-dimensional radius within which half of the projectedluminosity of the cluster is contained.
The half-light radii of clusters 1 and 6 were determinedfrom the drizzled ACS/
HST
F625W images using the pro-gram ishape (Larsen 1999). The routine convolves the point-spread function (PSF) of the camera with the desired lightprofile. ishape computes a χ minimisation between thecluster image and the model profile over a set fitting radius,iterating the FWHM, the ratio of the minor to major axesand the orientation of the fit. In order to obtain the bestfit, the user specifies the ‘clean radius’, which is the largestradius that contains no contamination by neighbouring clus-ters.Cluster 6 is more isolated than cluster 1, and is uncon-taminated by its neighbours up to a clean radius of fourpixels (1 pixel = 0.05 arcsec). Cluster 1, however, is onlyisolated up to two pixels. The half-light radii were com-puted by ishape for a range of profiles, using a PSF thatwas computed in Tiny Tim (Krist & Hook 1997), for fit-ting radii between two and twenty pixels with a clean ra-dius of two pixels for cluster 1 and four pixels for cluster 6.The profiles considered were a Gaussian profile, EFF pro-files (Elson, Fall & Freeman 1987) of index x , which takethe form: f ( z ) = 1(1 + z ) γ , γ = 0 . x and King (1962) profiles of index c of the form: f ( z ) = ( (cid:16) √ z − √ c (cid:17) z < c z > c . There was a sudden jump in the values of half-light radiiproduced for fitting radii of eight pixels and for nine pixels.Therefore, the means of the half-light radii computed by ishape for each profile for fitting radii between two and eightpixels were considered. These, along with the corresponding c (cid:13) , 000–000 luster and nebular properties of NGC 1140 Table 5.
Mean half-light radii computed by ishape for a rangeof model profiles for fitting radii between two and eight pixels,and the corresponding standard deviations (s.d.). Half-light radiiwere adopted on the basis of the best standard deviation.Cluster 1 Cluster 6Model Index r hl (pc) s.d. r hl (pc) s.d.Gaussian — 6.1 0.8 5.6 0.4EFF 15 9.5 0.5 8.4 0.5EFF 25 6.9 0.8 6.12 0.12King 5 6.6 0.8 5.79 0.12King 15 8.2 0.5 7.0 0.6King 30 10.4 0.4 8.8 1.2King 100 17.3 0.6 14.6 2.6Adopted 8 ± . ± . standard deviation, are contained in Table 5. As a good pro-file fit will not vary much with fitting radius, the profile withthe lowest standard deviation was adopted. For cluster 1 thiswas the King 30 profile. However, the standard deviation ofthis is quite large and similar to that produced for other pro-files. Discarding the outlying values produced by the King100 profile, the mean and standard deviation of the 42 re-sults produced by the other 6 profiles gives r hl = 8 ± hl =6 . ± . The line-of-sight velocity dispersion, σ , of a cluster can bedetermined by comparing the broadening in the lines of thecluster spectrum with respect to those in a red supergianttemplate spectrum. There are two main methods that quan-tify the comparison, and these are discussed below.The spectrum is compared to red supergiant templatesbecause the lines of these stars are broadened by onlya few km s − , by macro-turbulence in their atmospheres(Gray & Toner 1986). It is not appropriate to use ear-lier type supergiants and main sequence stars, as effectssuch as rotational broadening, macro-turbulence and micro-turbulence broaden the lines of these stars by amountscomparable to the anticipated cluster velocity dispersions.Therefore, only spectral regions redwards of ∼ χ between the normalised cluster spectrum and normalised,broadened template spectra. The broadening is achieved byconvolving the normalised template spectrum with a Gaus-sian of σ equal to the desired velocity broadening. Thebroadened template is multiplied by an optimum factor, sothat it produces the lowest possible reduced- χ between thebroadened template spectrum and the cluster spectrum. The Table 6.
Values of line-of sight velocity dispersion over the re-gion 8485 − χ minimisation ( χ ) technique was used. The Paschen emis-sion was masked out for the reduced χ minimisation technique.The results obtained for the cross-correlation technique are anupper limit on the cluster velocity dispersion.RSG Template Velocity dispersion (km s − )Cat No. Spectral Cluster1 cluster 6Type χ Xcor χ template is broadened by a range of suitable values and areduced- χ is similarly computed for each value of broaden-ing. The lowest of these χ values determines what amountof broadening produces the best match with the cluster fora given template, and so indicates the velocity dispersion ofthe cluster. This is repeated for a range of template spec-tral types. This method relies on a good match between therelative line strengths of the cluster and template and canthus be sensitive to the spectral type of the template star.The second method utilises the cross-correlation tech-nique of Tonry & Davis (1979). It requires that the spectrabeing considered are normalised, and continuum subtracted,to give a flat continuum at zero. The cluster is then cross-correlated with a red supergiant template over suitable spec-tral regions, and the FWHM of the resulting cross correla-tion function (CCF) is measured. The template spectrum isbroadened by a range of velocities. Each broadened templateis cross-correlated with the original, unbroadened templateand the FWHM of each CCF is measured. In this man-ner, the near-linear relationship between broadening and theFWHM of the CCF can be empirically calibrated to an ab-solute scale. This calibration is applied to the FWHM of theoriginal CCF, produced by cross-correlating the cluster spec-trum with the template spectrum, to determine the velocitydispersion of the cluster. This is repeated for each templatestar. The conversion factor differs with template spectrum.While this method is less sensitive to spectral type matching,it suffers from complications associated with the subjectiv-ity of fitting CCFs. These include factors such as selectingthe background level and fitting non-Gaussian CCFs, andare especially important when the CCF is weak.Since clusters 1 and 6 are the brightest cluster membersin knots A and B, we assume that the velocity dispersions ofthese clusters dominate the broadening of the RSG featuresapparent in the knots.Unfortunately, the youth of cluster 1 means that theRSG features in the spectrum are very weak, and many ofthe lines visible in the template spectra are absent in thecluster spectrum. The low signal-to-noise ratio of knot Bcauses a similar problem. The strongest RSG features, the c (cid:13) , 000–000 S. L. Moll et al.
Ca II triplet absorption lines, are clearly visible in the clusterspectra. However, these lines are saturated in the templatespectra. As the core of a saturated profile is narrower thanwould have been produced in a Gaussian profile, and it is thecore that produces the CCF signal, cross-correlation of theseregions tends to overestimate the cluster velocity dispersion(see e.g. Walcher et al. 2005). However, cross-correlation ofall other regions that both contain RSG features and lacktelluric features produce very noisy, non-Gaussian CCFs,which cannot be robustly fitted. Therefore, only the region8485 − χ minimisation could also only be com-puted for the region 8485 − χ minimisation was computed.The results of the cross-correlation and the reduced- χ min-imisation are listed in Table 6.The results show consistency over all the spectral types,with the cross-correlation of cluster 6 with the templatesproducing systematically higher velocity dispersions thanthe reduced- χ technique, as expected. There should notbe any systematic uncertainties in the velocity dispersioncalculated by reduced- χ minimisation, despite the use ofthe Ca II triplet lines. Mengel et al. (2002) found no dis-parity between the velocity dispersion results computed by χ minimisation for the strongest component of the Ca IItriplet and other individual absorption features for clus-ters in NGC 4038/4039. The mean and standard devia-tion of the velocity dispersions calculated from the reduced- χ minimisation from all six template stars give values of σ = 24 ± − for cluster 1 and σ = 26 ± − forcluster 6. The virial equation relates the virial mass of a cluster, M dyn ,to the line-of-sight velocity dispersion, σ , and the half-lightradius, r hl , of the cluster by the equation: M dyn ≈ ησ r hl G .
This equation assumes that a cluster is gravitationallybound, spherically symmetrical and virialised, that the ve-locity dispersion of the cluster is isotropic and that all of thestars contained in the cluster are single stars of equal mass.Spitzer (1987) showed that η = 9 .
75 for globular clusterswith a wide range of light profiles. However, mass segrega-tion and the presence of binary stars within the cluster cancause a large variation in the value of η (Boily et al. 2005;Fleck et al. 2006; Kouwenhoven & de Grijs, in prep.). Forthe young age and high mass of cluster 1, however, theseeffects are minimal. Table 7 summarises the cluster proper-ties. Table 7.
Summary of cluster properties of clusters 1 and 6.Property cluster 1 cluster 6 E ( B − V )(cluster) (mag) 0.24 ± ± hl (pc) 8 ± ± σ (km s − ) 24 ± ± M dyn (10 M ⊙ ) 10 ± ± L V (10 L ⊙ ) 7.3 ± L V /M dyn (( L V /M ) ⊙ ) 7.0 ± In this section we discuss the disparity found between thedynamical and photometric mass estimates. We also con-sider the reliability of stellar evolution models by comparingthem with observations.
The dynamical mass determined for cluster 1 is several timesgreater than the photometric mass found. To assess this dis-parity, Fig. 7 compares the light-to-dynamical mass ratioof the cluster to a model light-to-mass ratio at the knownage of the cluster. The model assumes Maraston (2005)SSPs, a Kroupa IMF and solar metallicity. Data for sev-eral other clusters whose velocity dispersions were measuredfrom UVES observations are also included. It is clear fromthis figure that many other young ( <
20 Myr) clusters alsohave light-to-dynamical mass ratios well below model pre-dictions, while the older ( >
20 Myr) clusters have ratios thatagree well with the canonical value.As discussed by Bastian et al. (2006) andGoodwin & Bastian (2006), the most likely explana-tion for the discrepancy between model and observationsseen in the sample of young clusters is a lack of virialequilibrium caused by violent relaxation after the formationof the cluster and from the expulsion of gas. This causesthe measured dynamical mass to overestimate the truemass. Since the older clusters, which are expected to bein virial equilibrium after surviving gas expulsion, lie onor near the model line, the photometric mass is likely tobe a good representation of the true cluster mass. Thissupports the assumption of a standard IMF. In this case,the disagreement between the dynamical mass and thephotometric mass of a cluster can be used to assess to whatdegree the cluster is out of virial equilibrium. This canbe parameterised by the effective star formation efficiency(eSFE) of the cluster, ǫ ; at the onset of gas expulsion, a clus-ter with an eSFE = ǫ has a velocity dispersion that is p /ǫ too large to be in virial equilibrium (Goodwin & Bastian2006). The model light-to-mass ratio assuming a range ofeffective star formation efficiencies are included in Fig. 7,taking the onset of gas expulsion as 2 Myr. These tracksdiffer from those presented by Goodwin & Bastian (2006)as they are not smoothed.An alternative explanation for clusters lying below thecanonical line in Fig. 7 is that these clusters have non-standard IMFs. In this situation, the dynamical mass repre-sents the true cluster mass, and the incorrect assumption of a c (cid:13) , 000–000 luster and nebular properties of NGC 1140 Figure 7.
Diagram of light to dynamical mass ratio againstage for cluster 1 in NGC 1140 (filled triangle; this work),and for clusters 1, 2 and 16 in NGC 4038/4039 (filled circles;Mengel et al. 2002), clusters 502 and 805 in NGC 5236 (opentriangles; Larsen & Richtler 2004), clusters W3 (Maraston et al.2004) and W30 (Bastian et al. 2006) in NGC 7252 (filled squares),cluster G114 in NGC 1316 (cross; Bastian et al. 2006) and clus-ters standard IMF causes the photometric mass to differ from thedynamical mass. Since the discrepant clusters lie below themodel line, they would have an over-abundance of low-massstars with respect to the Kroupa IMF. Such bottom-heavyclusters are more likely to remain bound than clusters witha standard Kroupa IMF and should be apparent in the sam-ple of old clusters. As they are not, this explanation seemsunlikely for the majority of these young clusters. However,a non-standard IMF can not be ruled out in the case ofNGC 1140–1.There are various potential explanations for the differ-ence between light-to-dynamical mass ratio and model light-to-mass ratio in the case of NGC 1140–1. These are exploredbelow:(i) Assuming that the difference bewteen dynamical andphotmetric mass is due solely to the cluster being out ofvirial equilibrium, it can be seen from Fig. 7 that cluster 1has an eSFE of around 10 − ∼
20 Myr and not remain bound. As such,the cluster would disperse over a relatively short timeframeand, therefore, not evolve into a second generation globularcluster.(ii) An underestimate of the cluster extinction would havecaused an underestimation of the photometric mass. How- ever, in order to increase the luminosity of the cluster bythe factor of ∼ E ( B − V ) int ≈ . hl ≈ . ishape for cluster 1, would still pro-duce a virial mass that was a factor of seven larger than thephotometric mass.(iv) The crowded nature of the region may have causedthe cluster velocity dispersion to be overestimated. Thebroadening of the RSG features seen in the knot spectramay not have been due solely to the virial motions of thebrightest cluster within the knot, as is assumed, but mayhave had contributions from the motions of other clusters.This situation can be modelled using the template spectra,and a reduced χ minimisation can be computed for thismodel knot. A template spectrum is broadened by 8 km s − to represent the broadening of cluster 1, as implied by itsphotometric mass and half-light radius. Added to this is thecluster 2 contribution, in the flux ratio of 1/1.7, as impliedby the F814W photometry in Table 1. Taking the velocitydispersion of cluster 2 to be equal to that of cluster 1, thevelocity dispersion found for knot A in Table 6 can be re-produced by introducing a relative velocity shift betweenthe two clusters of ∼
20 km s − . It is not possible to assesshow likely such a shift is. Cluster 2 appears to be older thancluster 1 because of the lack of nebular emission (Fig. 1).This then implies that cluster 2 could have a larger pho-tometric mass than cluster 1, despite its fainter luminosity,and thus have a larger velocity dispersion than adopted here.This would reduce the velocity shift required to produce acombined velocity dispersion equivalent to that observed forknot A. However, neither the photometric mass nor the shiftare known and the importance of the contribution of cluster2 to the velocity dispersion of knot A cannot be assessed.It is not possible to quantify the degree to which these fourfactors contribute to the difference between the light-to-dynamical mass ratio measured and the model in Fig. 7.However, the contribution from the neighbours of cluster 1is likely to be the most dominant factor for the large dynam-ical mass measured. As it is not possible to quantify this, itis also not possible to assess how long the cluster is likely tosurvive. The nebular properties of the two knots of NGC 1140, andthe massive star content of clusters 1 and 6 are presented inTable 8. There have been several other studies of young mas-sive stellar populations of nearby starbursts, including theultraviolet (UV) survey of Chandar, Leitherer & Tremonti(2004) and optical studies of Vacca & Conti (1992) andSchaerer, Contini & Kunth (1999a). However, the resultsof these surveys are not readily comparable to our results, c (cid:13) , 000–000 S. L. Moll et al.
Table 8.
Nebular knot properties and massive star content of NGC 1140 (this work), NGC 3125 (Hadfield & Crowther 2006) and Tol 89(Sidoli et al. 2006).Galaxy NGC 1140 NGC 3125 Tol 89 (NGC 5398)Knot A B A B A B E ( B − V ) (mag) 0.16 ± ± N e (cm − ) 60 ±
50 — 140 140 90 ±
40 150 ± T e (K) 9700 ±
300 — 10500 9800 10000 ±
300 9800 ± / H 8.29 ± . +0 . − . . +0 . − . N (O) ⋆ ± ±
500 4000 3200 685 2780Age (Myr) 5 ± > ∼ ∼ < . M phot (10 M ⊙ ) 1.1 ± . − . ∼ . N (O) † ± − ∼ N (WN) 550 ±
100 — 105 ∼
55 40 80 — N (WC) 200 ±
50 — 20 — 20 0 — N (WR) /N (O) 0.1 — 0.2 0.1 0.1 0.2 — ⋆ From the luminosity of the nebular H β emission of the knot. † From Starburst99, for the age and photometric mass of the cluster. due to the different techniques adopted, especially in deter-mining the WR populations.Chandar et al. (2004) derived the O and WR content ofthe central starburst regions for a large sample of galaxiesbased upon UV
HST spectroscopy and Starburst99 mod-els. Fits to far-UV spectral morphologies and continuumslopes provided representative ages and extinctions fromwhich O star populations were obtained, although individ-ual clusters were not distinguished. The number of WR starswas inferred from a calibration of He II 1640 line lumi-nosities. Individual clusters were also not considered in theground-based optical studies of Vacca & Conti (1992) andSchaerer et al. (1999a). Standard nebular techniques wereapplied to derive extinctions, representative ages and Ostar numbers. WR populations resulted from calibrations ofGalactic and LMC WR stars (Vacca 1994; Schaerer & Vacca1998), which differ from our calibration with solely LMCtemplates.Studies of individual massive clusters using commontechniques, including
HST imaging, have been carried outby Hadfield & Crowther (2006) and Sidoli et al. (2006) fortwo galaxies of LMC-like metallicity, the blue compact dwarfgalaxy NGC 3125 and the giant HII region (GHR) Tol 89located in the barred spiral galaxy NGC 5398. A comparisonof individual knots, as derived from nebular properties, andcluster masses and O star numbers, from Starburst99 mod-elling, is made in Table 8. WR populations follow from opti-cal calibrations of Crowther & Hadfield (2006) in all cases.NGC 1140–1 is a significantly more massive counter-part to NGC 3125–A1 and Tol 89–A1, with a similar, highratio of WR to O type stars, N (WR) /N (O) ∼ . − . N (WC) /N (WN) .
4. These empirical stellar populationscan be compared to predictions from evolutionary synthe-sis models (e.g. Starburst99), for which reduced WR popu-lations of N (WR) /N (O) ∼ . − . N (WC) /N (WN) ∼ − N (WR) /N (O) < . N (WC) /N (WN) ∼ Z = 0 . N (WR) /N (O) ∼ . − .
07. These models, however, pre-dict significantly more realistic ratios of WC to WN stars of N (WC) /N (WN) ∼ . − − We present new high spectral resolution VLT/UVES spec-troscopy and
HST /ACS imaging of the central region ofNGC 1140. It is apparent from the ACS imaging that thecentral region contains several clusters, although this is onlyresolved into two star-forming knots by the UVES spec-troscopy: knot A, which contains clusters 1 and 2, andknot B, which contains clusters 5, 6 and 7.Nebular analysis of knot A yields an LMC-like metal-licity of 12 + log O / H = 8 . ± .
09. Starburst99 mod-elling indicates an age of 5 ± . ± . × M ⊙ for cluster 1. Virial massesof (10 ± × M ⊙ and (9 . ± . × M ⊙ were deter-mined for clusters 1 and 6, respectively, using the half-lightradii determined from the F625W ACS image and the ve-locity dispersions determined by computing a reduced- χ minimisation between the cluster spectrum and RSG tem-plate spectra over a spectral region containing the Ca IItriplet. We interpret the difference between the dynamicaland photometric mass of cluster 1 as due to the crowdednature of knot A: the velocity dispersion measured may notrelate only to cluster 1, as assumed, but likely contains acomponent that arises from cluster 2, with a different sys-temic velocity to cluster 1.We find 6600 and 800 O stars within knots A andB, respectively, from the H β luminosities of the knots.Our Starburst99 model predicts 5900 O stars within clus-ter 1. This implies that >
90% of the O stars withinknot A are contained in cluster 1. Empricial fitting of the c (cid:13) , 000–000 luster and nebular properties of NGC 1140 blue bump of cluster 1 indicates that this cluster containsaround 550 WN stars and 200 WC stars, giving ratios of N (WR) /N (O) = 0 .
1, if all of the WR stars lie within clus-ter 1, and N (WC) /N (WN) = 0 .
4. The observed ratio of WRstars to O stars is significantly larger than predicted by cur-rent evolutionary models. The observed ratio of WC to WNstars is reproduced more successfully using Padova instanta-neous burst models than Geneva models, even allowing forrotational mixing.
ACKNOWLEDGMENTS
SM acknowledges financial support from PPARC/STFC.The Image Reduction and Analysis Facility iraf is dis-tributed by the National Optical Astronomy Observatories,which is operated by the Association of Universities for Re-search in Astronomy, Inc., under cooperative agreement withthe U.S. National Science Foundation. We would like tothank Søren Larsen for his help with the ishape analysisand Fabrizio Sidoli, who provided help with the UVES datareduction. We would also like to thank Nate Bastian andSimon Goodwin for supplying their effective star formationefficency models and for discussions on the subject.
REFERENCES
Bastian N., Saglia R. P., Goudfrooij P., Kissler-Patig M.,Schweizer F., Zoccali M., 2006, A&A, 448, 881Boily C. M., Lan¸con A., Deiters S., Heggie D. C., 2005,ApJ, 620, L27Chandar R., Leitherer C., Tremonti C. A., 2004, ApJ, 604,153Crowther P. A., 2007, ARA&A, 45, 177Crowther P. A., Hadfield L. J., 2006, A&A, 449, 711de Grijs R., Parmentier G., 2007, ChJA&A, 7, 155de Grijs R., Smith L. J., Bunker A., Sharp R. G., Gal-lagher III J. S., Anders P., Lan¸con A., O’Connell R. W.,Parry I. R., 2004, MNRAS, 352, 263de Vaucouleurs G., de Vaucouleurs A., Corwin H. G. J.,Buta R. J., Paturel G., Fouqu´e P., 1991, Third ReferenceCatalogue of Bright Galaxies, Version 3.9D´ıaz A. I., Terlevich E., Terlevich R., 1989, MNRAS, 239,325Elson R. A. W., Fall S. M., Freeman K. C., 1987, ApJ, 323,54Fagotto F., Bressan A., Bertelli G., Chiosi C., 1994, A&AS,105, 29Fleck J.-J., Boily C. M., Lan¸con A., Deiters S., 2006, MN-RAS, 369, 1392Goodwin S. P., Bastian N., 2006, MNRAS, 373, 752Gray D. F., Toner C. G., 1986, ApJ, 310, 277Guseva N. G., Isotov Y. I., Thuan T. X., 2000, ApJ, 531,776Hadfield L. J., Crowther P. A., 2006, MNRAS, 268, 1822Haynes M. P., Hogg D. E., Maddalena R. J., Roberts M. S.,van Zee L., 1998, AJ, 115, 62Heckman T. M., 1998, in Woodward C., Shull J., ThronsonH., eds, Starbursts: Lessons for the Origin and Evolutionof Galaxies and the Intergalactic Medium ASP ConferenceSeries vol. 158, San Francisco, p. 127 Ho L. C., Filippenko A. V., 1996a, ApJ, 466, L83Ho L. C., Filippenko A. V., 1996b, ApJ, 472, 600Howarth I. D., 1983, MNRAS, 203, 301Hummer D. G., Storey P. J., 1987, MNRAS, 224, 801Hunter D. A., O’Connell R. W., Gallagher III J. S., 1994a,AJ, 108, 84Hunter D. A., van Woerden H., Gallagher III J. S., 1994b,ApJS, 91, 79Izotov Y. I., Thuan T. X., 2004, ApJ, 602, 200Kennicutt R. C., 1998, ARA&A, 36, 189King I. R., 1962, AJ, 67, 471Krist J., Hook R., 1997, The Tiny Tim User’s Guide.STScI, BaltimoreKroupa P., 2002, Science, 295, 82Larsen S. S., 1999, A&AS, 139, 393Larsen S. S., Richtler T., 2004, A&A, 427, 495Leitherer C., Schaerer C., Goldader J. D., Gonz´alez Del-gado R. M., Robert C., Kune D. F., de Mello D. F., DevostD., Heckman T. M., 1999, ApJS, 123, 3Maraston C., 2005, MNRAS, 362, 799Maraston C., Bastian N., Saglia R. P., Kissler-Patig M.,Schweizer F., Goudfrooij P., 2004, A&A, 416, 176Massey P., Olsen K. A. G., 2003, AJ, 126, 2867Mengel S., Lehnert M. D., Thatte N., Genzel R., 2002,A&A, 383, 137Meynet G., Maeder A., Schaller G., Schaerer D., Charbon-nel C., 1994, A&AS, 103, 97Nagao T., Maiolino R., Marconi A., 2006, A&A, 459, 85O’Connell R. W., Gallagher III J. S., Hunter D. A., 1994,ApJ, 433, 65¨Ostlin G., Cumming R. J., Bergvall N., 2006, A&A, 261,471Russell S. C., Dopita M. A., 1990, ApJS, 74, 93Schaerer D., Contini T., Kunth D., 1999a, A&A, 341, 399Schaerer D., Contini T., Pindao M., 1999b, A&AS, 136, 35Schaerer D., Vacca W. D., 1998, ApJ, 497, 618Schlegel D. J., Finkbeiner D. P., Davis M., 1998, ApJ, 500,425Sidoli F. X., Smith L. J., Crowther P. A., 2006, MNRAS,370, 799Sirianni M. et al. 2005, PASP, 117, 836Spitzer L., 1987, Dynamical Evolution of Globular Clus-ters. Princeton University Press, PrincetonTonry J., Davis M., 1979, AJ, 84, 1511Vacca W., 1994, ApJ, 421, 140Vacca W. D., Conti P. S., 1992, ApJ, 401, 543V´azquez G. A., Leitherer C., 2005, ApJ, 621, 695V´azquez G. A., Leitherer C., Schaerer D., Meynet G.,Maeder A., 2007, ApJ, 663, 995Walcher C. J., van der Marel R. P., McLaughlin D., RixH. W., B¨oker T., H¨aring N., Ho L. C., Sarzi M., ShieldsJ. C., 2005, ApJ, 618, 237Westmoquette M. S., Smith L. J., Gallagher III J. S., ExterK. M., 2007, MNRAS, accepted c (cid:13)000