Table Of Content

Mon.Not.R.Astron.Soc.000,000–000 (0000) Printed5February2008 (MNLATEXstylefilev2.2) The Cluster Wind from Local Massive Star Clusters Ian R. Stevens, Joanna M. Hartwell School of Physicsand Astronomy, University of Birmingham, Edgbaston, Birmingham B15 2TT (E-mail: [email protected], [email protected]) 3 0 5February2008 0 2 n ABSTRACT a J Results of a study of the theoretically predicted and observedX-rayproperties of 3 localmassivestar clusters arepresented,with a focus on understanding the massand energy flow from these clusters into the ISM via a cluster wind. A simple theoretical 1 model,basedontheworkofChevalier&Clegg(1985),isusedtopredictthetheoretical v cluster properties, and these are compared to those obtained from recent Chandra 8 observations. The model includes the effect of lower energy transfer efficiency and 3 mass-loading. In spite of limited statistics, some general trends are indicated; the 0 observed temperature of the diffuse X-ray emission is lower than that predicted from 1 the stellar mass and energy input rates,but the predicted scaling of X-ray luminosity 0 with cluster parameters is seen. The implications of these results are discussed. 3 0 Key words: hydrodynamics: shock waves; open clusters and associations; stars: / h winds, outflows p - o r t s 1 INTRODUCTION dominate (cf. Leitherer & Heckman 1995). In many star- a burst galaxies, where a large numberof SSCs are seen, this v: Super Star Clusters are dense clusters of young massive cluster wind will be an important contributor to energising i stars,firstidentifiedinNGC1275byHoltzmanetal.(1992) the interstellar medium and possibly driving a hot super- X usingtheHubbleSpaceTelescope (HST),andsubsequently wind or outflow, such as those seen in M82 or NGC253 r in a wide range of star-forming galaxies, such as merg- (Lehnert,Heckman &Weaver1999; Strickland et al.2002). a ingsystems(NGC4038/4039;Whitmore&Schweizer1995), The efficiency with which stellar kinetic energy (from both dwarf galaxies (Henize 2-10; Johnson et al. 2000), classical stellar winds and supernovae) is converted into thermal en- starbursts (M82; Gallagher & Smith 1999) amongst many ergy within thecluster, which in turn can drivean outflow, other systems (see Whitmore 2000 for a review). These is an important and very difficult parameter to determine. extragalactic star clusters can contain many thousands of Intheliterature,severaldifferentvaluesorprescriptionsfor very young, energetic stars, and have stellar densities far thethermalizationefficiencyhavebeenused,oftenencapsu- greater than those seen in normal OB associations. The lated in a parameter η,which represents thefraction of the values quoted by Whitmore (2000) indicate ages for these kinetic energy of stars and supernova in the cluster that is star clusters of typically 1−10Myr, radii typically in the thermalized. The fraction of the kinetic energy that is not range of ∼ 1−6pc, total masses of stars in the cluster in thermalized is assumed to be lost to the system, primarily the range 103−106M⊙), with the central stellar densities via radiative losses. As we shall see radiative losses at X- reachingupto∼105M⊙ pc3.Itisclearthatinmanygalax- ray energies are likely to very small, however, at UV and iesasubstantialfractionoftheongoingstar-formation (and IR energies radiative losses are likely to be much larger. In hence mass/energy injection into the ISM) is occurring in fact, inhomogeneities in the complex flow within the clus- SSCs(Origliaetal.2001).TheGalactic(orlocalanalogues) tercould easily lead totheformation of denserregionsthat of these extragalactic SSCs are objects such as NGC3603, could contributetoradiating energy away from thesystem. R136in30DoradusandtheArchesclusterneartheGalactic In the literature, Strickland & Stevens (2000) argued Centre (see Figer, McLean & Morris 1999b). forveryefficientthermalization(η∼1;seealsoChevalier& The component stars of an SSC are believed to be Clegg 1985), whereas Bradamante, Matteucci & D’Ercole roughlycoeval,andwhentheclusterisveryyoungthemas- (1998) argued for a much lower thermalization efficiency sivestars intheclusterwill havestrongstellar windswhich of a few per cent. Other, more complex, prescriptions of will combine to produce an outflow from the cluster, which η havealso been proposed. For instance, Recchi, Matteucci is referred to as the cluster wind. As the cluster ages mass & D’Ercole (2001) use a model with η being very low in and energy input from supernova explosions will begin to the early evolution of a cluster, but as material is evacu- 2 I.R. Stevens, J.M. Hartwell ated by the cluster wind to create a low density region the describe outflows from starburst galaxies. The contrast be- thermalization efficiency rises to close to unity. In relation tweenthesemodelsisthatCantóetal.(2000)alsousedhy- to LMC superbubbles, Oey (1996) noted that the bubble drodynamic calculations with mass and energy input from dynamics implied a relatively low thermalization efficiency discrete stars within the cluster, whereas the Chevalier & (or put another way that there was less power driving the Clegg(1985)modelassumedmassandenergyinjectionuni- bubbles than implied by the stellar populations within the formlythroughoutthestarburstregion.Thedensestarclus- bubbles, see also Chu & Mac Low for a discussion of the ters considered here do not have a uniform distribution of X-ray properties of superbubbles). It would clearly be very stars and tend to be centrally concentrated, and mass and useful to have more direct observational constraints on this energy injection will tend to follow the stellar distribution. process. However,Cantóetal.(2000),whodidincludeanon-uniform Understandingthemassandenergylossprocessesfrom stellar distribution showed that theirnumerical simulations distant SSCs via the hot cluster wind can best be done at generallyreproducetheresultsoftheChevalier&Cleggtype X-ray energies. However, because of their distance and the model, and the formulation of Chevalier & Clegg (1985) is highstellardensity,evenChandrahasinsufficientresolution adopted here. One important difference between the Cantó to see what is going on in and around these systems, and et al. (2000) and Chevalier & Clegg (1985) models is that resolvethediffuseemission frompointsourceemission from the Cantó et al. (2000) have a much greater range of tem- stars in thecluster. However, local lower-mass analogues of peratures (and densities inside the cluster core). these SSCs in our Galaxy or Local Group can be used to Ifmassandenergyfrom starsisbeinginjectedviastel- study the mass and energy loss processes from stellar clus- lar winds into the volume of a stellar cluster of core radius ters.BystudyingthediffuseX-raypropertiesofstarclusters R then Chevalier &Clegg (1985) showed that thesolution c andthestellarenergyinjectionratefromstellarpopulations for theoutflow from this region can be written as intheclusteritshouldbepossibletoconstraintheefficiency −(3γ+1) (γ+1) with which the clusters can convert kinetic energy to ther- 3γ+1/M2 (5γ+1) γ−1+2/M2 2(5γ+1) r = (1) mal energy. Although not perfect analogues, the values de- 1+3γ 1+γ R c termined could havefar-reaching consequences, not just for (cid:18) (cid:19) (cid:18) (cid:19) the understandingof theprocesses within the star clusters, for r<Rc and butalso fortheenergyemission from larger structuressuch γ−1+2/M2 ((γ+1)/2(γ−1)) r 2 as starburst galaxies, where many SSCs are present. M(2/(γ−1)) = (2) 1+γ R It is also worth making the point that while the mass (cid:18) (cid:19) (cid:16) c(cid:17) andenergyinjectionratefrommassivestarsviatheirstellar for r > R , where M is the Mach number of the flow, γ is c winds can be estimated to a reasonable degree of accuracy, the adiabatic index (with γ = 5/3 assumed hereafter), R c thesameislesstrueforsupernova.Forinstance,itisworth the radius of the star cluster and r is the radius from the noting that the type IIn supernova SN1988Z is believed to centerofstarcluster.Fromtheseequationsandthemassand have radiated ∼ 1052 erg in its early evolution (Aretxaga energy injection rates, the velocity v(r), temperature T(r) et al. 1999), see also Chevalier & Fransson 2001), which is and mass density ρ(r) can be determined. The technique rather more than the usually assumed values for the total todosoinvolvessolvingeqn.1oreqn.2asappropriate,for energy injection from a SN. Consequently, looking at very theMachnumberateachradius,usingtheNewton-Raphson young clusters, where SN injection does not dominate, has method. From the Mach number, and using the integrated considerably more promise. forms of the mass and energy energy equations, the other In this paper the theory of cluster winds is devel- variablescanbedetermined(seeChevalier&Clegg1985for oped, particularly as it relates to X-ray emission from more details). Solar abundances and fully ionized material clusters (§ 2), and it is then applied to results from Chan- with a mean mass per particle of µ¯ = 10−24 gm are also dra observations of nearby SSCs or smaller stellar clusters assumed. (§3).Theresultsarediscussedin§4andsummarisedin§5. It isworth notingthat theconditions in thecentralre- gionsoftheclusterareagoodindicatoroftherecenthistory ofthecluster.Theflowtimewithintheclusterisshort,typically a few 103 yr, whereas the characteristic timescale of the larger scale bubble being blown by the cluster is much 2 THE X-RAY EMISSION FROM A CLUSTER longer(i.e.ofordertheageofthecluster).Thismeansthat WIND thecurrent X-raypropertiesshould givea betterhandleon When massive stars with strong winds are in close proxim- the transfer efficiency of stellar wind energy to the cluster ity, such as in a stellar cluster, the winds collide both with wind. each other and with the surrounding ISM, filling the sur- roundingvolumewithhotshockedgas.Eventuallyallofthe 2.1 The Stellar Mass and Energy Injection Rate available volume is filled and the hot gas escapes beyond the boundaries of the cluster. The development of such a For the young star clusters under consideration here the clusterwindhasbeennumericallymodelledbyCantó,Raga energy injection is dominated by stellar winds, especially & Rodríguez (2000) in a simulation carried out assuming a those of the massive OB and Wolf-Rayet (WR) stars, and spherical cluster consisting of ∼30 stars (see also Ozernoy, we assume that no supernovae have as yet exploded and Genzel & Usov 1997; Raga et al. 2001). contributedtotheclusterwind.Atlaterepochssupernovae The wind from a stellar cluster can also be described begin to dominate the energy injection (Leitherer & Heck- usingtheChevalier&Clegg(1985)modeloriginallyusedto man 1995). The Cluster Wind from Local Massive Star Clusters 3 Figure1.Theradialstructurefortheclusterwind,ascalculatedfromtheChevalier&Clegg(1985)model(seetextfordetails).Shown aretheradialprofilesforthewindvelocity(topleftpanel),gastemperature(topright),density(bottomleft)andthecumulativeX-ray luminosity(bottomright).Thestandardmodel(solidline)isforastarclusterwithastellarmassinjectionrateofM˙∗=10−4M⊙ yr−1, V¯∗ = 2000kms−1, η = 1 and no additional mass-loss (M˙cold = 0). The second model is identical, except with η = 0.1 (dashed line), andthethirdmodel(dotted line)hasη=1andM˙cold=2×10−4M⊙ yr−1. In principal, from the current observed massive star Ameanweighted terminalvelocityforthestarsintheclus- population in the clusters, the mass and energy injection ters is definedsuch that rate from the stellar winds of these stars can be estimated, 1/2 aelnsdewthheisrei,sththeereparroecesodmureeswigeniafidcoapntt.uHnocweretvaeinr,tiaessidnistchuessdeed- V¯∗ = PNi ∗M˙M∗˙iVi2! (5) termination of these values, related to uncertainties in understanding thestructureand evolution of massive stars. For some of the stellar clusters discussed in the next sectiontheindividualmass-lossratesandterminalvelocities The bulk of the mass and energy injection is assumed can be determined, and for those systems where there is to happen within a cluster core radius R . If the cluster c containsN∗ stars,eachstarwithamass-lossrateofM˙i and insufficient observational evidence stellar evolution models, such as the Starburst99 code (Leitherer et al. 1999) can be wind terminal velocity Vi with i = 1...N∗ then the total stellar wind mass injection rate, M˙∗, is used. Some qualifications should be made here. The method used for estimating the mass-loss rates and terminal veloc- N∗ M˙∗ = M˙i (3) ities of the cluster stars rests on our current best under- standingofthemass-lossparametersofmassivestars.There i X remainsignificantdeficienciesinourknowledgeofthestruc- and the total kinetic energy injection rate is tureand evolution of massive stars and theirwinds.For instance, the Starburst99 code only assumes a population of single stars and binary evolution could make a difference N∗ 1 E˙∗ = 2M˙iVi2 . (4) tkonotwhne tmhaastst-lhoesschraartaecstearnisdtictesromfitnhaelwveinlodcsitoyf.mInadsseievde,sittariss Xi 4 I.R. Stevens, J.M. Hartwell evolvewithtime,andthissnapshotcouldbemissingimpor- youngerclusterswherewewouldnotexpecttoseeemission tant early phases of evolution when the winds were much from SNRsand X-ray binaries. In thedata presented later, stronger.Inspiteofthesequalificationsweshallusecurrent thepointsourceshavebeenremovedfromtheX-raydatato best estimates and proceed accordingly. leave only the diffuse emission. It is however possible that In a cluster although the optically visible stars (and unresolvedpointsources,particularlyfromlowerluminosity theirwinds) will likelydominatetheenergyinjection, these pre-mainsequence(PMS)stars,willcontaminatethediffuse stars will not necessarily be the only sources of mass. Pro- emission. Wewill discuss thispoint more later. tostars (or other optically obscured objects) will also con- Fora given model, thebroad-bandX-rayluminosity of tribute mass as material is ablated off them by the cluster thecluster will be given by wind or through radiation. Examples of such objects may Rc be the ProPlyDs which have been seen in NGC3603 and L = 4πr2n n Λ(T)dr (8) X e i elsewhere (for instance, Brandner et al. 2000; Mu¨cke et al. Z0 2002). This cold mass being injected into the cluster wind whereΛ(T)istheemissivityofgasatatemperatureT,and can be accounted for by a term M˙ , so that the total cold n and n are the electron and ion number density respec- e i mass-loss rate into thecluster is tively. Given the solution for ρ(r), v(r) etc, eqn. 8 can be M˙tot =M˙∗+M˙cold (6) evaluated to derive the total X-ray luminosity (over a wide waveband) of the cluster. A more detailed analysis could This process can be considered as a mass-loading process calculatetheluminosityinspecificwavebands.However,the (see Hartquist et al. 1997 and references therein). majority of the emission will fall in the Chandra waveband As discussed in the introduction, to account for theef- and we will not complicate matters further. fects ofradiative losses in theconversion of stellar wind en- InFig.1theradialvariationofthevelocityandtemper- ergy into the cluster wind, we can introduce a parameter ature for a stellar cluster are shown, where the total stellar that we term the “thermalization efficiency” parameter η, mass-loss rate is 10−4M⊙ yr−1 and V¯∗ =2000km s−1 and such that the energy available to drive the cluster wind is the cluster core radius is 1pc. These values are similar to Eth=ηE∗, so that those of the clusters discussed in this paper. In this model N∗ thecentraltemperatureis 5.8×107K, thecentral ion num- E˙th=η 21M˙iVi2 = η2M˙∗V¯∗2 (7) breegriodnenwsiitthyinisa0.r6a7dcimus−R3, ainsdLthe=X5-.r1ay×l1u0m32ineorgsitsy−1o.fItmhe- i c X X portantly for our purposes, the cluster itself dominates the andallowη tovarybetween0and1.Thisparameterallows X-rayemission and thevolumeoutside R contributesonly for radiative losses. The energy radiated at X-ray energies c 15 per cent of the total X-ray luminosity. The reason for isunlikelytobeamajorlossmechanism,butUVradiation, thisisthesharpfalloffindensity(andtemperature)outside particularly from interfaces between hot and cold gas, may thecluster,whichmorethancompensatesfortheincreasing beabiggercontributor,ascould IRradiation. Asdiscussed volumeoftheemission region.Forasimpleanalysissuchas byRecchietal.(2001)wemightexpectthisparametertobe presented herethis level of accuracy is sufficient. relatively small in the early stages of cluster evolution, but Thecentraltemperatureoftheclusterwind(T )inthis toincreaseatlatertimes.Forsimplicity,inthesesimulations 0 model is given by η is assumed constant. The result of allowing η to be less trehdaunceuntihteytaenrdmiandadlictilounstaelr“wcoinldd”vmelaoscsit-iynjbeecltoiwonV¯w∗.illTbheistios TK0 =1.45×107 1000kV¯m∗ s−1 2 . (9) turn increases the density and often leads to an enhanced (cid:16) (cid:17) (cid:18) (cid:19) X-ray luminosity. Note that the slightly different numerical constant as com- The assumption of a constant thermalization efficiency pared to Cantó et al. (2000) is due to slightly different as- is also highly simplistic. Not only is thevaluelikely tovary sumptionsas to themean mass per particle. as a function of time, as the cluster wind evolves, it is also From the radial solution the emission weighted tem- likely to vary as a function of location within the cluster, perature for the cluster region (Tcl) can also be calculated, withadifferentthermalizationefficiencyinthedenserinner which will be somewhat lower than the central temper- regions to that in the more diffuse outer regions. Given the ature. For instance, for the standard model with V¯∗ = lackofdetailedinformationastohowηmightvary,weshall 2000km s−1, the emission weighted temperature is Tcl = treatitasaconstant.So,subjecttothisrecognitionthatthis 5.1×107 K. is a highly simplified treatment of a very complex problem Results of models illustrating the effect of η and mass- we shall proceed. loading on the expected cluster temperature are shown in Fig. 2, and show that, not surprisingly, both mass-loading andη<1leadtolowerclustertemperatures,butalso,often, 2.2 The Cluster X-ray Luminosity higher X-rayluminosities. Observationally,at X-raywavelengths,ayoungstellar clus- terwillhavebothpointsourceemissionassociatedwithindi- 2.3 X-ray Luminosity Scaling Relations vidualstars,suchasfromcollidingwindsinmassivebinaries, intrinsicemissionfromsingleearly-typestars,emissionasso- Giventhatthegastemperaturewithintheclusteriscompar- ciatedwithSNRs,X-raybinariesorpre-mainsequencestars, ativelyconstant (seeFig. 1)and hot(for V¯∗ ≥1000km s−1 anddiffuseemission(duetotheclusterwind).Herewecon- the temperatures are > 107K, so that the bremsstrahlung centrate only on the diffuse X-ray emission, and only from regionofthecoolingcurveisappropriate,i.e.Λ(T)∝T1/2) The Cluster Wind from Local Massive Star Clusters 5 Figure 2. The variation of the expected X-ray temperature of Figure 3. The theoretical cluster X-ray luminosity versus the theclustercoreversusthewindvelocityofthestellarcomponent. cluster wind scaling parameter Xcl = M˙2/(RcV¯∗). Results are Shownaremodel resultsforthe clustercentral temperature and shown for several different sets of models. The cluster radius is clusterluminosityforvaryingη andthecontributionofcoldgas. always1pcandV¯∗=2000kms−1.Theopentriangles(andsolid Thestandardclustermodelhas M˙∗=10−4M⊙ yr−1 andRc = line) are for the standard models, which have no mass injection 1pc.Themodelsrepresentedbyopentrianglesandmarkedwitha (M˙cold = 0) and η = 1, and stellar mass-loss rates of M˙∗ = solidlineareforη=1andV¯∗=500,1000,2000and3000kms−1 10−6,10−5,10−4and10−3M⊙ yr−1.Thefilledtriangles(dashed The models represented by filled triangles and marked with a line)havethesameparametersexceptη=0.1.Theopensquares dashedlinearethesamemodelsexceptwithη=0.1.Themodels (dashedline)haveη=1andM˙cold=M˙∗ (andhaveverysimilar represented by open squares and marked with a dotted line are results to the previous models). The filled squares (dot-dashed thesamemodelsexceptwithM˙cold=M˙∗ line)haveη=1andM˙cold=10M˙∗.Thecrosses(dot-dot-dashed lines)haveη=1andM˙cold=2×10−5M⊙ yr−1. the following simple scaling relationships for the X-ray luminosity of clusters can be derived (ignoring for the time being theeffects of η<1 and mass-loading). The gas density within thecluster, n∝M˙∗/(Rc2V¯∗), so that if we define a cluster wind scaling parameter X such cl that increasingdensity.Ofcourse,forη≪1theemissionwillnot M˙2 Xcl = RcV∗¯∗ (10) bvaeriaattiXon-raoyf tehneercgluiesstearnXd-LraXy lwuimllindoecsirteyasvee.rsRuessuthltissfcolursttheer then from eqn. (8) L ∝ X . We shall calculate X using scaling parameter are shown in Fig. 3, where models for a X cl cl the natural units, that is M˙ in M⊙ yr−1, Rc in pc and V¯∗ range of cluster wind scaling parameters Xcl are shown. in km s−1, so that for a model with M˙∗ =10−4M⊙ yr−1, The effect of mass-loading is similar to the effect of V¯∗ =2000km s−1 and Rc =1pc, then Xcl =5×10−12. lower valuesof η – bothlead toadrop intheaverageavail- At smaller velocities thedeviations from the T1/2 scal- able energy per particle. This is reflected in that models ing for the cooling curve will alter things. For tempera- with η = 0.1 have almost identical results to a model with tures in the range of 106.2 ≤ T ≤ 106.5 K, then roughly M˙cold = M˙∗. Mass-loading leads to a decrease in tempera- Λ(T)∝T−0.6andthecorrespondingexpressionforthevari- tureandanincreaseindensity.Modelresultsarealsoshown ation of theX-ray luminosity is in Fig. 3. Of note is that the fact that we get a very differ- entrelationshipbetweenL andX whenthemass-loading LX ∝ RcMV˙¯∗∗23.2 . (11) injecDtioisnenistacnognlsintagntthreatchoemrXbtihnaendaaccctloionnstaonftlomwuerltivpalleueosfMo˙f∗η. Thescalingeffectofvariationinηontheclusterproper- and mass-loading in an individual cluster may be difficult, tiescanalsobederived.Areductioninη effectivelyreduces but in a statistical sample the relative influence of both the amount of thermal energy available to the cluster as a effects may be apparent in star cluster X-ray luminosity– whole.Forthesameamountofmass-lossthiswillreducethe temperature plots. This is illustrated in Fig. 4, where for temperature, with T ∝ η, but also reduce the velocity of these simple models, different slopes in the cluster X-ray cl thecluster wind and increase thedensity. luminosity-temperaturerelationship are found when theef- For the bremsstrahlung dominant case (Λ(T) ∝ T1/2), fectsofη and mass-loading areisolated andvariedin asys- the density n ∝ M˙/(Rc2V¯∗) and V¯∗2 ∝ η, so that here tematic manner. Whether in reality the two effects can be L ∝η−0.5. Inthecase of Λ(T)∝T−0.6,then L ∝η−1.6. separated so cleanly is unclear. These models have also as- X X Somewhatcounter-intuitivelythen,areductioninηleadsto sumed a single cluster core radius. More realistic models, ariseinX-rayluminosity,duetothedominanteffectofthe with a range of parameters will show more scatter. 6 I.R. Stevens, J.M. Hartwell Table 1.Theobservedandpredictedpropertiesoftheclusterwindfromthe5starclustersunderconsideration.ThevaluesforM˙ and V¯∗ areobtainedusingeqns.3and5andaredescribedmoreinthetext.Thepredictedtemperature,kT0,iscalculatedusingeqn.9.The expression for the cluster wind scaling parameter, Xcl, is given by eqn. 10. The sources for observed values for LX and kTX for each clusteraredescribedinthetexttoo. PredictedValues Observedvalues Cluster M˙∗ V¯∗ Rc Xcl PredictedkT0 ObservedkTX LX (M⊙ yr−1) (kms−1) (pc) (keV) (keV) (ergs−1) NGC3603 2.3×10−4 2844 2.8 6.6×10−12 10.1 3.7 2.2×1034 R136 2.6×10−4 2110 2.0 1.6×10−11 5.6 2.1 5.5×1034 NGC346 1.7×10−5 2282 2.0 6.4×10−14 6.5 1.0 1.5×1034 Rosette 2.5×10−6 2173 4.6 6.3×10−16 5.9 0.6 8.0×1032 Arches 7.3×10−4 2810 0.2 9.5×10−10 9.9 0.8,6.4 5.0×1035 X-ray emission as seen with the Chandra satellite. In some casesresultsfromtheliteratureareused,whileinothersan analysis of thedata has been performed. 3.1 NGC3603 NGC3603isthemostmassivevisibleHIIregioninourgalaxy, andiscommonlyregardedasaGalacticanalogueforextragalactic SSCs. It contains a very dense concentration of stars of ages of ∼ 2−3Myr, including 3 WR stars in its core. In many ways it issimilartoR136in30Dorbutwithoutthesurroundingcluster halo(Moffatetal.1994).Weassumeadistanceof7kpc. The characteristics of the massive stars in NGC3603 have beenstudiedbyCrowther&Dessart(1998), andusingthisdata and Kudritzki & Puls (2000), the mass-loss rates and terminal velocitydatafor44starswithin2.8pcofthecentreofthecluster have been tabulated. These stars will dominate the stellar mass andenergylossintothecluster.Basedontheseresults,thetotal stellar mass-loss rate for NGC3603 is estimated as M˙∗ = 2.3× Figure 4. The theoretical stellar cluster X-ray luminosity- 21804−44kMm⊙s−yr1−.1Thanedcltuhsetemrceoarnewraedigihutseidsttaerkmenintaolbveel2o.c8iptyca(MsV¯o∗ffa=t temperature relationship. Shown are model results forthe mean etal.1994). cluster temperature and cluster luminosity for varying η and Results from a 50ksec ACIS-I Chandra observation of the contribution of cold gas. The standard cluster model has M˙∗ = 10−4M⊙ yr−1, V¯∗ = 2000kms−1 and Rc = 1pc. The NanGdCth3e6s0e3dhaatvaeshaolrweasdtyrobngeeenvipdreensceentoefdaindiffMuosffeattheertmaall.c(o2m00p2o)-, models represented by open triangles and marked with a solid nent, probably associated withacluster wind.To determine the line are for varying η, with models with η =1.0,0.8,0.6,0.4,0.2 spectral characteristics of the diffuse emission, point sources are and 0.1, with decreasing temperature corresponding to decreas- excluded and the diffuse spectrum extracted. The spectrum has ingη.Themodelsrepresentedbyfilledtrianglesandmarkedwith been fitted with an absorbed mekal model (wabs*mekal), with a dashed line are for variations with the amount of cold mate- Mri˙aclolidnje=ct1ed0−M6˙,c1o0ld−5a,n1d0−h4avaenηd=101−.30.MT⊙heyrm−o1d,ewl irtehsudltescraeraesifnogr NcoHrre=cted0.6lu7m×in1o0s2it2ycmis−L2X, =kT2X.2×=130.374keerVg,sa−n1d(tsheeeTaabbsolerp1t)i.on temperaturecorrespondingtoincreasingM˙cold. 3.2 R136 in 30 Doradus 3 THE SAMPLE OF NEARBY STAR R136is the central object of30 Doradus inthe LMC,andis re- CLUSTERS gardedastheclosestexampleofanSSC,andweassumeadistance of 50 kpc (cf. Feigelson 2001). The spectral types and mass-loss SSCs and other compact star clusters are relatively small ratesof39starsinR136wereobtainedfromastudybyCrowther objects with a typical size of a few pc, so even with the &Dessart(1998),andthecorrespondingwindvelocitiesfromKu- arcsecond resolution ofChandraonly thosethat arenearby dritzki& Puls(2000). Thecluster mass-lossrate isestimated to can be studied in any detail. In this study Chandra data be M˙∗ = 2.6×10−4M⊙ yr−1 and the mean weighted terminal forR136, NGC3603, NGC346, theRosetteNebulaand the velocityasV¯∗=2110kms−1.Theclustercoreradiusisassumed tobe2pc(Crowther &Dessart1998). Arches cluster are used. The definition of an SSC is rather The X-ray data of R136 was obtained from the Chandra vague and certainly the Rosette Nebula does not qualify. archive and consists of a 28ksec ACIS-I observation of 30 Do- However, although it is a much smaller cluster the physics radus.Thepointsourceswithinaregionofradius16′′ centredon should be thesame and so it is included. In this section we the cluster were removed as for NGC3603, and a spectrum was describe both the relevant properties of the stellar compo- extractedfromtheremainingemission,andfittedagainusingan nents of the cluster and the diffuse X-ray properties of the absorbed mekal model. The best fit spectrum to this emission The Cluster Wind from Local Massive Star Clusters 7 was an absorbed mekal model, with NH = 0.42×1022cm−2, assumethat thecluster coreradius is0.2pcandthat the cluster kTX =2.1keV,andabsorptioncorrectedLX =5.5×1034ergs−1 isatadistanceof8kpc(Figeretal.1999b). (seeTable1). Due to the lack of comprehensive data on the stars in the Arches cluster, we have adopted the simpler approach of calcu- lating M˙∗ and V¯∗ using the Starburst99 models1. Based on the resultsofFigeretal.(1999a),fortheArchesclusterweassumea 3.3 NGC346 top-heavy IMF(withα=1.6),solarabundances, highmass-loss NGC346 is the largest star formation region of the SMC, and stellar evolution tracks and a total stellar mass of 1.1×104M⊙ containsthemajorityoftheOstarsintheSMC(Massey,Parker (for stars in the mass range 1−100M⊙), and the appropriate & Garmany 1989). We assume a a distance of 59kpc (Mathew- values are taken for values in the range t ∼ 2−4.5Myr. For son, Ford & Visvanathan 1986). The spectral types and mag- times>3.5MyrmassandenergyinjectionfromSNbegintoplay nitudes of the stars in NGC346 were obtained from Massey et a role. For the Arches cluster, in this time span the derived al. (1989). Using a reddening correction of E(B−V) = 0.14, values of the mean weighted terminal velocity are in the range mass-loss rates were calculated from the expression in Howarth V¯∗=1800−2810kms−1,andthetotalstellarmass-lossisinthe & Prinja (1989), accounting for the effects of lower metallic- rangeM˙∗=(2.0−9.5)×10−4M⊙ yr−1.Thecorrespondinglarge ity on stellar mass-loss rates, with M˙(Z) ∝ (Z/Z⊙)0.5 (cf Ku- rangeintheclusterwindparameterisXcl=(0.85−25.0)×10−10. dritzki & Puls 2000 and using a value for the SMC metallicity From the Starburst99 simulations it is interesting to note ofZ⊙/5).Theclustermass-lossrateforNGC346isestimatedto that the value of V¯∗ ≥ 1500kms−1 for all times of relevance be M˙∗ = 1.7×10−5M⊙ yr−1 and the mean weighted terminal to young clusters. In the Starburst99 models only single stars velocityasV¯∗=2282kms−1.Theclustercoreradiusistakento are included and the energy injection rate from SN tails off at be2pc(Masseyetal.1989). t ∼ 35Myr (in this model). In models including binaries, mass Chandra observed NGC346 for a total of 100ksec in and energy injection from SN will continue to longer times due May 2001. The cluster lies close to a chip gaps of the ACIS-I totheprocessesofmasstransferandbinaryevolution. instrument, but diffuse emission from the cluster is still de- Forconvenience, wewilladoptthevaluesatatimeof4Myr tected. A more detailed analysis of the Chandra data of the for the Arches cluster, with V¯∗ = 2810kms−1, and M˙∗ = NGC346 region can be found in Nazé et al. (2002a; 2002b). 7.3×10−4M⊙ yr−1. These values are of course rather uncer- The extracted spectrum was best fit with an absorbed mekal tain, and but for instance we note that direct radio detections model, with NH = 0.42 × 1022cm−2, kTX = 1.0keV, and of a number of the stars in the Arches cluster by Lang, Goss & LX =1.5×1034ergs−1 (absorptioncorrected,seeTable1). Rodríguez (2001) might suggest a broadly similar value for the totalmass-lossratefromthecluster.However,intheirsimulation of the X-rayproperties of the Arches cluster, Raga et al. (2001) assumed a somewhat larger integrated mass-loss rate (60 stars 3.4 The Rosette Nebula, NGC2237 each with an individual mass-loss rate of 10−4M⊙ yr−1) but a lowermeanterminalvelocity(thewindofeachstarwasassumed The Rosette Nebula (NGC2237) contains an open cluster, tohave aterminalvelocity of1500kms−1). Thecalculations of NGC2244, and is one of the more massive diffuse nebulae. It Raga et al. (2001) did not include additional mass-loading and issituatedatadistanceof1.5kpc(Gregorio-Hetem etal.1998). inthefollowingsection weshallseethat toreproducethe X-ray The cluster wind parameters of the Rosette Nebula were calcu- properties of the Arches (and other) clusters we find that we do latedusingthepropertiesofthe26O,BandAstarslyingwithin need to include mass-loading (or a lower η), which has a sim- 4.6pcoftheclustercentre,whichcorrespondstotheroughextent ilar effect as an enhanced mass-loss rate and reduced terminal oftheX-rayemission. velocity. SpectraltypeswereobtainedfromSIMBAD,andthecorre- Results from a 51ksec ACIS-I Chandra observation of the spondingwindvelocitiesfromKudritzki&Puls(2000).Mass-loss Arches cluster have beenpresented byYusef-Zadehet al.(2002) rates wereevaluated usingthe relationshipof Howarth &Prinja and show three main emission regions probably associated (1989), assuming E(B−V) = 0.49 (Massey, Johnson, DeGioia- with the Arches cluster (designated A1–A3 by Yusef-Zadeh Eastwood1995).TheresultingvaluesfortheRosetteNebulaare et al. 2002). Two of these components are compact (A1 and M˙∗ =2.5×10−6M⊙ yr−1 and V¯∗ =2173kms−1. The cluster A2) and one coincides with the core of the cluster (A1). The coreradiusisassumedtobe4.6pc(seeabove). third component (A3) is more extended and part of it may Chandra observations of this region found a large number be due to the cluster wind and part may be associated with of point sources. On subtraction of these point sources, a weak fluorescent emission from a molecular cloud (C. Lang, private diffuse component was discovered (see Montmerle et al. 2002; communication). Yusef-Zadeh et al. (2002) fitted the spectrum Townsley et al. 2002, in preparation). Fitting this with an ofthiscomponentwithanadditionallinefeatureat6.4keV.The absorbedRaymond-Smithmodel,thebestfitparametersforthe total X-ray luminosity (0.2 − 10keV) from the 3 components diffuse emission were NH = 7×1021cm−2, kTX = 0.6keV and is 5×1035ergs−1, with component A1 being the most lumi- the absorption corrected luminositywas LX =8×1032ergs−1 nous. The spectra are fitted with a 2 component model, with inthe0.5−2keV waveband. kT1 ∼0.8keV and kT2 ∼6.4keV for the A1 component, and we shalluseboththesetemperaturevaluesintheanalysis,butonly the overall luminosity. The two compact X-raysources could be genuine point sources, associated with colliding stellar winds, or at least one of them could also be the unresolved core of a 3.5 The Arches Cluster clusterwind. The Arches cluster is located at a projected distance of ∼ 50pc from the Galactic centre (Cotera et al. 1996; Portegies-Zwart 2001), and contains ∼120 stars whichhave masses greater than 20M⊙ (Serabyn,Shupe&Figer1998).Itisoneofthedenseststar clusters known in the local group. Blum et al. (2001) have pre- senteda2µmstudyofthestellarpopulationoftheArchescluster, andinferredanagefortheclusterofbetween 2and4.5Myr.We 1 Availableathttp://www.stsci.edu/science/starburst99 8 I.R. Stevens, J.M. Hartwell Figure 5. Left: The fitted X-ray temperature of the diffuse X-ray emission of the sample of star clusters versus the predicted X-ray temperature calculated from the model described in § 2. The solid line depicts equality, that is extremely efficient conversion of the kinetic energy in the winds of the component stars to thermal energy. Two values of observed temperature for the Arches cluster are plotted (see text for details). Right: The observed X-rayluminosityLX versus the theoretical cluster windparameter Xcl (eqn.10) for thesampleclusters.Alsoplottedaretworesultsfromtwomodels showninFig.3,namelymodelswith η=1andM˙cold=M˙∗ (dotted line)andM˙cold=10M˙∗ (dot-dashed line). 4 RESULTS: COMPARISON OF THEORY AND Similarly,very low values of η could yield the same result (with OBSERVATION η ∼ 0.01), though this would be in conflict with the observed temperatures. From the theory developed in § 2 and the observational results It could be argued that this correlation is partly a conse- thatwehaveforthesmallnumberofclusters,presentedin§3we quence of the simple fact that we would expect more massive can now make some comparisons. In Table 1 we summarise the clusters to be more X-ray luminous, and undoubtedly the value theoretical andobservational dataforthesampleclusters. of Xcl is most dependent on the total stellar mass-loss rate of Fig. 5 (left panel) shows a plot of the observed X-ray tem- the cluster (M˙∗) and this will in turn scale with stellar mass. peraturesfromtheChandraobservationsversusthetheoretically Larger clusters may also have larger populations of unresolved predictedwindtemperaturesforeachcluster(specifically, values point sources making them appear more X-ray luminous. With of T0 are plotted, calculated using eqn. 9). It is clear that in all the verylimitedsamplewehave itisdifficult togofurther than casesbaronetheobservedtemperatureissignificantlylowerthan merely noting that there is a correlation between LX and Xcl. thatthevaluepredictedfromthestellarpopulationinthecluster. TheRosettenebulaappearstolieatasomewhathigherluminos- Theexception isthecaseofthehotter component oftheArches ity than its value of Xcl would imply. This could imply a larger cluster, where the discrepancy is much less. In the case of the comparative fraction of mass-loading as compared to the other Arches though, itshouldberememberedthat the best-fitmodel clusters(seeFig.3).Thelowobservedtemperature,ascompared to the X-ray spectrum had two components and the 2nd com- totheexpected value,wouldtallywiththis. ponent is also plotted and is much more in line with the other In Fig. 3 we postulated that it may be possible to dis- clusters. In fact, the cooler component (kT1 ∼ 0.8keV) is more criminate between the effects of a low thermalization efficiency luminous than the hotter component and itmay have been bet- and mass-loading by means of a stellar cluster X-rayluminosity tertousethatvalue.Also,R136israthermoredistantthanthe temperature diagram. In Fig. 6 we show a plot of the observed rest and the possibility of unresolved point sources contaminat- values for the cluster sample. The model predicted that the ing the spectrum is more. If we were to ignore both the Arches luminosityandclustertemperatureshouldbeinverselycorrelated cluster and R136 (of course, leaving only 3 data points) then a (thoughwithdifferentslopesdependingonwhethermass-loading morecoherentpicturemaybeemergingofconsistentlylowerob- or low thermalization efficiency was dominant). From Fig. 6 we servedtemperatures.Ashasbeennotedbefore,suchareduction canseeclearlythatthedatasuggestthatX-raytemperatureand inobservedtemperatureisanaturalconsequenceofη<1orthat luminosityareprobablynotwellcorrelated. significantmass-loadingisoccurring. In § 2 the cluster wind model was also used to make pre- dictions concerning the scaling of the cluster X-ray luminosity versustheclusterwindscalingparameter Xcl (seeeqn.10).The observedcorrelationbetweenthemeasuredLX andXclisshown 5 SUMMARY AND CONCLUSIONS in the right panel of Fig. 5. The most notable feature is that compared to the standard model in Fig. 3 the clusters are all Inthispaperwehavepresentedacomparisonbetweenthetheory overluminousbyoveranorderofmagnitude.Thesecondfeature ofoutflows fromyoung star clustersinacluster windandX-ray isthattheredoesseemtobeaclearcorrelationbetweenLX and observations of diffuse X-ray emission from these clusters. The Xcl, not necessarily linear, but apparently monotonic. For com- highspatialresolutionofChandraisnecessarytobegintodisen- parison,resultsfrommodelcalculations,alreadyshowninFig.3, tanglethediffuseemissionfrompointsourceemission.However, illustratethat the effects of mass-loading can roughly reproduce even with Chandra problems remain as to being sure that the the X-ray luminosities of the cluster, with M˙cold = 1−10M˙∗. diffuseemissionisgenuinelydiffuseandassociatedwiththeclus- The Cluster Wind from Local Massive Star Clusters 9 Conf.Ser.234,X-rayAstronomy2000,Astron.Soc.Pac.,San Francisco,p.131 FigerD.F.,KimS.S.,MorrisM.,SerabynE.,RichR.M.,McLean I.S.,1999,ApJ,525,750 FigerD.F.,McLeanI.S.,MorrisM.,1999,ApJ,514,202 GallagherJ.,SmithL.,1999, MNRAS,304,540 Gregorio-Hetem J., Montmerle T., Casanova S., Feigelson E., 1998,A&A,331,193 HartquistT.,DysonJ.,WilliamsR.,1997, ApJ,482,182 HoltzmanJ.,FaberS.,ShayaE.,LauerT.,GrotheJ.,HunterD., Baum W., Ewald S., Hester J., Light R., Lynds C., O’Neil E.J.,Westphal J.,1992,AJ,103,691 HowarthI.D.,PrinjaR.K.,1989,ApJS,69,527 JohnsonK.,LeithererC.,VaccaW.,ContiP.,2000,AJ,120,1273 KudritzkiR.-P.,PulsJ.,2000,ARA&A,38,613 LangC.C.,GossW.M.,RodríguezL.F.,2001, ApJL,551,L143 LehnertM.,HeckmanT.,Weaver K.,1999,ApJ,523,575 LeithererC.,HeckmanT.,1995,ApJS,96,9 LeithererC.,SchaererD.,GoldaderJ.,DelgadoR.,GonzalezR., CarmelleK.,DenisF., deMelloD.,Devost D.,Heckman T., Figure 6. A stellar cluster X-ray luminosity-temperature dia- 1999,ApJS,123,3 gram for the observed clusters, see text for more details. Note Massey P., Johnson K., DeGioia-Eastwood K., 1995, ApJ, 454, that the two values plotted for the Arches cluster correspond to 151 thetwofittedcomponents (seetextfordetails). MasseyP.,ParkerJ.,GarmanyC.,1989,AJ,98,1305 MathewsonD.,FordV.,VisvanathanN.,1986,AJ,301,664 MoffatA.,DrissenL.,SharaM.,1994,ApJ,436,183 Moffat A.F.J., Corcoran M.F., Stevens I.R., Marchenko S.V., terwind.Themajorproblemremainstheextremelysmallsample Skalkowski G., Mu¨cke A., Koribalski B.S., Ptak A., size.Moreobservationswillalleviatethisproblemtosomeextent, Mushotzky R., Pittard, J.M., Pollock A.M.T.,Brandner W., buttheobjectsdiscussedhererepresentthebestexamplesofsuch 2002,ApJ,573,191 aclusterwind. Montmerle T., Grosso N., Feigelson E.D., Townsley L., 2002, Thisanalysishasthrownupsomeinterestingresults,thedif- “NewVisionsoftheX-rayUniverseintheXMM-Newtonand fuseX-rayluminosityof the star clusters inthe sampleis corre- latedwiththeclusterwindscalingparameter Xcl=M˙∗2/(RcV¯∗) Mu¨cCkheanAd.,raKEorraib”a(leskdi.FB..,JeMnsoeffnatetAa.l,.)C,EorScAoraSnP-M48.8, (Sitnevpernesss)I., aspredicted,butthattheobservedX-raytemperatureisnotwell 2002,ApJ,571,366 correlatedwiththepredictedX-raytemperature(thoughabetter NazéY.,HartwellJ.M.,StevensI.R.,CorcoranM.F.,ChuY.-H., correlation maybe masked by the nature of the data). It isalso Koenigsberger G., Moffat A.F.J., Niemela V.S., 2002a, ApJ, clear that from these data it is very unclear as to what is going 580,225 on in the clusters; is there a low thermalization efficiency or is Nazé Y., Hartwell J.M., Stevens I.R., Manfroid J., Marchenko mass-loadingthedominantmechanism? S.,CorcoranM.F.,MoffatA.F.J.,SkalkowskiG.,2002b,ApJ, In this paper we have put together the results from several (submitted) Chandra observations of nearby massive young star clusters to OeyM.S.,1996,ApJ,467,666 investigate the clusterwind.It isclear that manymoreobserva- OrigliaL.,LeithererC.,AloisiA.,GreggioL.,TosiM.,2001,AJ, tionswillberequiredtotakethestudyfurther,tomakeaclearer 122,815 connectionbetweenthetheoryandobservationsofclusterwinds. OzernoyL.M.,Genzel R.,UsovV.V.,1997, MNRAS,288,237 Suchobservations willbenecessarytoreallyunderstandthe im- Portegies-Zwart S., 2001, In: Dynamics of Star Clusters and the pact that star clusters will have on the interstellar medium of MilkyWay,ASPConferenceSeries,Vol.228.Eds:S.Deiters, galaxiesofallshapes andsizes. B.Fuchs,R.Spurzem,A.Just,andR.Wielen.SanFrancisco, p.131 Raga A.C., Vela´zquez P.F., Canto´ J., Masciadri, E. Rodríguez L.F.,2001,ApJ,559,L33 REFERENCES RecchiS.,Matteucci F.,D’ErcoleA.,2001,MNRAS,322,800 Aretxaga I., Benetti S., Terlevich R., Fabian A., Cappellaro E., SerabynE.,ShupeD.,FigerD.,1998,Nature,394,448 TurattoM.,dellaValleM.,1999,MNRAS,309,343 Strickland D., Heckman T., Weaver K., Hoopes C., Dahlem M., BlumR.D.,SchaererD.,PasqualiA.,Heydari-MalayeriM.,Conti 2002,ApJ,568,689 P.S.,Schmutz W.,2001,AJ,122,1875 StricklandD.,Stevens I.,2000,MNRAS,314,511 BradamanteF.,Matteucci F.,D’ErcoleA.,1998,A&A,337,338 WhitmoreB.,2000,astro-ph/0012546 Brandner W., Grebel E.K., Chu Y-H., Dottori H., Brandl B., WhitmoreB.,SchweizerF.,1995,AJ,109,960 RichlingS.,YorkeH.W.,PointsS.D.,ZinneckerH.,2000,AJ, Yusef-ZadehF.,LawC.,WardleM.,WangQ.,FruscioneA.,Lang 119,292 C.,CoteraA.,2002,ApJ,570,665 Canto´J.,RagaA.,RodríguezL.,2000,ApJ,536,896 ChevalierR.,CleggA.,1985,Nature,317,44 ChevalierR.,FranssonC.,2001,ApJ,558,L27 ChuY-H.,MacLowM-M.,1990,ApJ,365,510 Cotera A.S., Erickson E.F., Colgan S.W.J., Simpson J.P., Allen D.A.,BurtonM.G.,1996,ApJ,461,750 CrowtherP.,DessartL.,1998,MNRAS,296,622 FeigelsonE.D.,2001,inGiacconiR.,SerioR.,StellaL.,eds,ASP