DraftversionFebruary2,2008 PreprinttypesetusingLATEXstyleemulateapjv.11/12/01 AN EXPLANATION FOR METALLICITY EFFECTS ON X-RAY BINARY PROPERTIES Thomas J. Maccarone1, Arunav Kundu2 and Stephen E. Zepf2 Draft version February 2, 2008 ABSTRACT We show that irradiationinduced stellar winds can explain two important metallicity effects in X-ray binaries - the higher numbers and the softer spectra of the X-raybinaries in metal richglobularclusters compared to the metal poor ones. As has been previously noted by Iben, Tutukov and Fedorova, the winds should be stronger at lower metallicity due to less efficient line cooling. This will speed up the evolution of the LMXBs in metal poor clusters, hence reducing their numbers. These winds can also 4 0 provideextramaterialneartheaccretingobjectwhichmaycreateanintrinsicabsorbertohardentheX- 0 rayspectraofthemetalpoorclustersystemsrelativetothemetalrichones,assuggestedbyobservations. 2 We outline some additional observational predictions of the model. n Subject headings: stars:winds,outflows – stars:neutron – stars:binaries:close – galaxies:star clusters – a globular clusters:general– X-rays:binaries J 6 1. introduction enhancement effects first associated with the metallicity 1 are, in fact, related directly to the metallicity and not to Globular clusters (GCs) are important laboratories for some other correlated property such as age, half-light ra- 1 studying stellar populations, both main sequence and ex- v otic. X-ray binaries are about 100 times overabundant in dius, ordistance from the center ofthe galaxy. It hasalso 3 been noted that the metal poor Large Magellanic Cloud GCs compared to in the field compared to the field (see 3 has a lower ratio of LMXBs to HMXBs than the more e.g. thecompilationofX-raybinarypropertiesinLiu,van 3 metalrichMilky Way (Cowley 1994;Iben, Tutukov & Fe- Paradijs& vanden Heuvel2001)due to dynamicaleffects 1 dorova1997- ITF97),which may be a combinationof the such as tidal capture and/or three and four body interac- 0 metallicityeffectsandthedifferenceintheirstarformation 4 tions (e.g. Clark 1975;Fabian, Pringle & Rees 1975; Hills histories. 0 1976). Furthermore, since detailed population studies of / X-ray binaries require more sources than the ∼ 100 in An additional metallicity effect can be found in the dif- h ferences of the soft X-ray spectra of these sources. It has the Local Group, the extragalactic case must be studied. p been found that the spectra of the blue (i.e. metal poor) Given that the ages and metallicities of individual extra- - GCs are harder than those of the red (i.e. metal rich) o galacticfieldstarsarenearlyimpossibletodetermine,but GCs inthe Milky WayandM31(Irwin& Bregman1998- r that these propertiescanbe inferredfromintegratedlight t IB) and in NGC 4472 (Maccarone, Kundu & Zepf 2003 - s of GCs, the GC X-ray binaries in other galaxies represent a MKZ03). On the other hand, the quiescent LMXBs seem an ideal place to study how age and metallicity affect X- : toshownoevidenceofametallicityeffectonthespectrum v ray binary populations. (e.g. Heinke et al. 2003), and the effect becomes much i Clearly, the dynamical properties of a GC are most im- X weaker or disappears at higher X-ray energies (Trinchieri portantforpredictingwhetheritwillhaveanX-raybinary r (e.g. Pooley et al. 2003; Heinke et al. 2003). In addition, et al. 1999; Di Stefano et al. 2002; Sidoli et al. 2001; a MKZ03). there also exists a strong residual correlationbetween the In this paper, we will show that the current theoretical number densities ofX-raybinariesandthe metallicities of explanations for the overabundance effect are unlikely to theGCs. ThiswasfirstsuggestedtobethecaseintheLo- match the typical factor of 3 difference between the prob- calGroup(Grindlay1993;Bellazzinietal. 1995),andwas abilities of metal poor and metal rich GCs’ having X-ray shown more conclusively and dramatically in NGC 4472 LMXBs. We willputforthanewscenarioinvokingtheef- (Kundu, Maccarone & Zepf 2002 - KMZ02). The pos- fects of irradiation induced winds (IIWs) can explain the sibility that this represents an age effect rather than a population difference. We will also show that IIWs can metallicity effect has been tested and found not to be the explain the previously unexplained spectral difference ef- case using data on NGC 3115 and NGC 4365 (Kundu et fects as a result of the same physical process, with the al. 2003 - K03). In NGC 3115, where the age spread is same parameter values. small, but the metallicity spread is large, the metallicity effect was found to be at least as strong as that in other 2. past theoretical work early type galaxies. Conversely, in NGC 4365, where the 2.1. IMF variations? age spread is large, the fraction of globular clusters with LMXBs did not vary much with age and was similar to Several attempts have been made to explain why metal thatofolderGCswithsimilarmetallicitiesinothergalax- rich GCs have more LMXBs, but as yet, none seems sat- ies. It thus seems most likely that the LMXB population isfactory. Grindlay (1993) suggested that the correlation might be due to a flatter initial mass function in higher 1 Astronomical Institute “Anton Pannekoek,” University of Amsterdam, Kruislaan 403, 1098 SJ, Amsterdam, The Netherlands; email: [email protected] 2 DepartmentofPhysicsandAstronomy,MichiganState University,EastLansingMI,48824; email: akundu, [email protected] 1 2 Maccarone,Kundu & Zepf metallicity GCs. The general consensus seems to be that of 10. An additional small increase in the tidal capture the IMF is fairly universal (e.g. Kroupa 2002), and star probabilitymightcomefromthe higherturnoffmassesfor formationtheorysuggeststhatanymetallicitydependence thestarsinmetalrichGCs(seee.g. Chaboyeretal. 1996), is likely to be such that metal rich stars have typically butthisfactorshouldbenomorethanabout5-10%. Stel- lower masses since the Jeans mass will be lower for the lar radius effects thus seem unlikely to make more than a more efficiently cooling metal rich gas (Larson 1998). Al- ∼60% difference (i.e. the product of the two 30% effects) thoughthepresentdaymassfunctionsofmetalrichGalac- inX-raybinarypopulationsasafunctionofmetallicity. In tic GCs are flatter than those of metal poor Galactic GCs fact, since tidal captures are not likely to produce binary (McClure et al. 1986),this seems to be largely due to the systems that are similar to the observed parameters, it factthatthemetalrichsystemsarelocatedmorecentrally seems more likely that the bulk ofthe LMXBs are formed withintheMilkyWayandhenceexperiencemoreextreme in three and four body exchanges (see e.g. Rasio et al. dynamical stripping of their low mass stars, as seen ob- 2000),andsotheoverabundancefactorinredclustersdue servationally (Piotto & Zoccali 1999) and expected theo- to larger stellar radii is likely to be closer to the factor of retically (Vesperini & Heggie 1997;Baumgardt& Makino 1.3 than 1.6. 2003). 3. the irradiation induced wind model 2.2. Effects of stellar radius ThemassdonorsinX-raybinariescanabsorbandrepro- Later, Bellazzini et al. (1995) suggested that the larger cessluminositiescomparabletotheirintrinsicluminosities. stellar radii of the stars in metal rich globular clusters The mass donor may then be “puffed up” to larger radii might contribute to their increased number densities of (e.g. Tutukov & Yungelson 1980; Podsiadlowski 1991; X-ray binaries. Larger stars should have a higher cross Harpaz & Rappaport1991)and that the extra kinetic en- section for tidal captures and should overflowtheir Roche ergy added to the envelope may drive off an evaporative lobes at larger separations. The physics of tidal capture wind with a velocity of order the escape velocity from the are a hotly contested issue and some workers have sug- stellarsurface(Arons1973;Rudermanetal. 1989;Tavani gested that tidal captures have great difficulty in form- &London1993;Pfahl,Rappaport&Podsiadlowski2003). ing systems similar to the X-ray binaries we observe (see A fraction (∼ 10%) of the gas lost by the mass donor will e.g. Rasio & Shapiro 1991; Rasio, Pfahl & Rappaport be accreted by the compact star in much the same way 2000 and references within), while others have suggested that more typical stellar winds are accreted in HMXBs. that the bulk of recycled pulsars may have been formed The lifetimes of LMXBs will also be accelerated by the in tidal capture systems (e.g. Di Stefano & Rappaport extra mass loss in the IIWs (Tavani 1991a). 1992). The alternative to formation of X-ray binaries by In addition to driving the evolution of the system, the tidal capturesis formationthroughexchangeinteractions. irradiationinduced wind will also leave a large amount of Thesearemostcommonlythree-bodyinteractionswherea gas in the environment of the X-ray binary. Most high neutron star encounters a binary system composedof two mass X-ray binaries show clear orbitalmodulations of the non-compactstarsandreplacesoneofthosetwostarsand X-ray flux, especially at soft X-ray energies, and this is formsabinarysystemwiththeother(seee.g. Clark1975; taken to be evidence of internal absorption by the mate- Hills 1976), or four-body interactions, where the neutron rial in the stellar wind. An IIW from a low mass star star is already in a binary system when it interacts with willlikelyhaveavelocitymuchcloserto the orbitalveloc- another binary system (see e.g. Mikkola 1984; see also ities in the binary system, so the orbital modulation will Fregeau et al. 2003 for a discussion of recent numerical not necessarily be as strong as in the HMXBs. Still, the work on 3-and 4-body interactions). increased absorption will have an effect on the spectrum. The easier Roche lobe overflow can be calculated using Let us summarize a few key past results on irradiation the standard assumption of a binary separation distribu- induced winds. It is generally found that these winds tionwhichislogarithmic,i.e. P(a)∝1/a. TakingKepler’s shouldbemoreimportantinlowmassX-raybinariesthan law andthe period-massrelation[equation(4.9)ofFrank, inhighmassX-raybinaries,sinceinlowmassX-raybina- King & Raine (1992 - FKR92)]we find that for a neutron ries the amount of radiation absorbed by the mass donor star accretor and a typical (i.e. a 0.6 M⊙) main sequence fromthe accretionflow may far exceedthe nuclear energy GC stardonor,the Rochelobe overfloworbitalseparation generation rate and may hence have a significant effect willbeabout3stellarradii,andvarieslittlewiththedonor on the structure of the donor star (see e.g. Tutukov & star’s mass. The separationat which Roche lobe overflow Yungelson 1980; Podsiadlowski 1991; ITF97); IIWs may will occur increases linearly with the stellar radius for a affect HMXBs as well (see e.g. Day & Stevens 1993), but fixed stellar mass. Given the logarithmic distribution of have been much less well studied and are far less likely to separations and a metallicity dependence of radius such dominate the overall mass transfer. Strong coronal winds that R∗ ∝Z1/8 (estimated from the stellar model interpo- are likely to result from IIWs and may be self-sustaining lationformulaeofToutetal. 1996),thecontributionofthe even if the mass donor does not fill its Roche lobe (see term related to the increase in the number of Roche lobe e.g. Basko & Sunyaev 1973; Arons 1973); in fact, ITF97 overflowingsystemswillbe1+1log(Zr)/log(rmax),which have found that self-sustaining winds can produce large 8 Zp rmin is about1.3 for a metallicity ratio of10. Numericalcalcu- luminosities even if the mass donor fills only ∼80% of its lations suggest that the tidal capture rate goes as R∗0.93, Roche lobe, in agreement with past results that a mass (e.g. Lee & Ostriker1986),so the number of neutron star donor need not fill its Roche lobe if it is sufficiently irra- binaries formed by tidal capture will go as Z0.12 which diated (Tavani, Ruderman & Shaham 1989). Some initial givesanotherfactorofabout1.3foraratioofmetallicities Rochelobeoverflowislikelytoberequiredinordertostart Metallicity/LMXB No. density correlations 3 accretion, but if the irradiation induced winds cause the loss rate for the case of a metal poor donor. Combining mass loss to be substantially faster than what would be equations(3)& (4),wefind that,forthe sameluminosity, caused by orbital and stellar evolution, then the star may M˙ Z Z ceaseto fill its Rochelobe evenasmasslossandaccretion r =( r)−0.35+ǫ−ǫ( r)−0.35. (5) continue. Metal rich stars can dissipate much of their ab- M˙p Zp Zp The fraction of the mass loss in the metal rich systems sorbedenergythroughlinecooling,whilemetalpoorstars dissipate this energy primarily through IIWs. Interpolat- coming from the IIW is (ZZpr)−0.35. ing from the stellar cooling rate tables of Sutherland & The number of X-ray binaries should scale as the for- Dopita (1993), the mass loss rate due to the IIWs should mation rate times the lifetime. The formation rate effects scaleasZ−0.35;thesystemlifetime shouldscaleasZ0.35 if have been studied by Bellazzini et al. (1995), and given themasslossisdominatedbythesewindsandthelifetime their lines of argument, we found in Section 2.2 that the is determined by the timescale for the mass donor to lose stellar radius effects should produce a difference by a fac- all its mass (ITF97). torofabout1.3(if exchangeinteractionsdominate)to1.6 As a caveat, we note that the treatment of ITF97, is (if tidal captures dominate). based in part on the analytical irradiation treatment in The effects of IIWs are predominantly on the system Iben, Tutukov & Yungelson (1995); while the results do lifetimes. For systems at a given luminosity, equation 5 agree well with the numerical work of Tavani & London showsthe differenceinmasslossrate,theinverseofwhich (1993) within the parameter space of their models, but gives the ratio of source lifetimes. Thus we find that ITF97 extrapolate outside this range. A more sophisti- Nr 1 = . (6) cbauttedisnculemaerlryicaoluttrseidaetmtheentsmcoapyeboefitnhoisrdpearpfeorr.futurework, Np (ZZpr)−0.35+ǫ−ǫ(ZZpr)−0.35 To compute the actual ratio of the number of red to blue Forawindemittedattheescapevelocityfromthemass GC X-raysources,it is necessaryto multiply the factorof donor, equation (4.35) of FKR92 shows that the fraction 1.3 to 1.6 from the stellar radius effects by the value from of the wind captured by the compact object will go as: equation 6, which should be about 2.1 for the typical pa- M˙ = 1(MCO)2(Rd)2 (1) rametersǫ=0.05and ZZpr=10. Thisgivesafactorbetween −M˙w 4 Md a about 2.6 and 3.4, although this factor is likely a slight overestimate, because even the X-ray sources in the most where M˙ is the mass accretionrate of the compactobject metalpoorGCsarelikelytohaveatleastsomeRochelobe M˙ is the wind mass loss rate, M is compact object’s w CO overflow contribution to their X-ray luminosities. mass, M and R are the mass and radius of the donor d d It has been assumed that the stellar wind velocity is star, and a is the orbital separation. Given typical values equal to the escape velocity from the surface of the star; of a = 4×1011 cm, Rd = 7×1010 cm, Md = 0.6M⊙ and this need not be the case. The extra wind energy for the MCO =1.4M⊙, about 5% of the wind is accreted. metal poor stars may be dissipated as a higher velocity To simplify the calculation, we make one additional as- wind rather than as a more dense wind. The fraction of sumption,thattheredGCsaccretefromacombinationof the mass lost that is accreted scales as v−4 (FKR92), Roche lobe overflow and IIWs, while the bright LMXBs wind whichalternativelyscalesastheinverseofthewindpower in blue GCs accrete from a self-sustaining wind. We will squared, for a constant mass loss rate. The luminosity of later revisit this assumption and show that it is not nec- the LMXBs in metal poor GCs would then be surpressed essaryinordertoreproduceroughlythe observations. We by a factor of about (Zr)0.70, while the lifetimes of the now define the notation for the following 5 equations - m˙ Zp two classes of systems would be about the same. Be- indicatesthemassaccretionrateontothecompactobject, cause the luminosity function has a slope of about −0.55 while M˙ is the mass loss rate. The subscript w indicates (KMZ02),theratioofthe numberofmetalrichandmetal windmassloss,RLindicatesRochelobeoverfloweffects,p poorclusterX-raysystemswouldthenbe (Zr)0.39,which, indicates systems with metalpoor donors,andr indicates Zp for the canonical factor of 10 difference between the two systems with metal rich donors, N indicates the number of systems and Z indicates metallicity. Then modesgivesafactorofabout2.5differenceintheexpected number of observed systems. We do note that the there m˙ =ǫM˙w+M˙RL, (2) might be systematic variations in the slope of the lumi- nosity functions as a function of metallicity, but absent where ǫ is the fractionof the mass lost in the wind that is accreted by the compact object and is given by equation measurementsoratheoreticalmodel, weassumethey will be the same. (1). Then, starting from the assumption outlined above We wish now to estimate the contribution to the col- thattheRochelobeoverflowcomponentisneglibleforthe metal poor systems, we have: umn density with which these systems will typically be observed due to the IIW. We find an average density of m˙p =ǫM˙w,p, (3) massinaspherearoundthe massdonorwithradiusequal while for metal rich stars, to the diameter of the orbit. For a path length of the orbital radius, the column density N is then: Z H m˙ =ǫ( r)−0.35M˙ +M˙ , (4) M˙ r w,p RL Z N = , (7) p H 8v r µ w orb where M˙ is the mass loss rate due to the Roche lobe where µ is the mean molecular weight of the gas in the RL overflowing component of the accretion flow, and is as- wind, v is the wind velocity, and r is the orbital sep- w orb sumedtobemuchsmallerthantheirradiationdrivenmass aration. This value, ≈ 6×1021cm−2, should give a good 4 Maccarone,Kundu & Zepf approximationoverorbitalphaseandinclinationanglefor manysystemswith differentvaluesofN andwith differ- H the column density observed. Edge-on sources might be entunderlyingspectra. Still,theroughinformationgiven, expected to havehigher columndensities, but the inclina- that the metal poor GCs have an intrinsic absorption of tion angles for the GC LMXBs are not well constrained. about 1021 cm−2 and that the column density is about 3 timesaslargeforthemetalpoorGCsasitisforthemetal 4. observational evidence for the scenario rich GCs seems to be a reasonable inference to draw from 4.1. NGC 4472 - number of sources thedata. Thetheoreticalmodelpredictsahighervaluefor the blue clusters, but the linear averaging tends to over- NGC 4472isthefirstgalaxywherethemetalrichmode estimate the effects on the spectrum, so the fact that our was shown to have a higher fraction of GCs with LMXBs crude calculation over-predicts the amount of absorption than the metal poor mode (KMZ02). It seems unlikely is to be expected. Furthermore, the fact that the gas is that the GCs in NGC 4472 span a wide range of ages; likely to be partly ionized and will make the fitted value they are all likely to be within a factor of 1.5 to 2 in age ofthe N less thanthe actualvalue,and alsosome ofthe H (Beasley et al. 2000;Cohen,Blakeslee & Cˆot´e2003). The gasmass will condense into a geometricallythin accretion metallicity effects on the number of globular cluster X- disk,andhencewillhaveaneffectofabsorbingX-raysonly ray sources thus are not likely due to an age difference ifthe inclinationangleis verylow. Thatthe values areon between the metal rich and metal poor GCs. The age the same order of magnitude and that the blue clusters measurements are rather sensitive to the stellar popula- haveabout3times asmuchabsorptioninthe fitsisabout tions models for the Balmer lines and there could still be as good an agreement as can be expected given the crude a rather substantial age difference between the two sam- modellingandtheconsiderabletheoreticaluncertaintiesin ples. More strict constraints are the age measurements the models of IIWs. of Puzia et al. (2002) which confirm that the correlation betweenmetallicityandLMXBspecificfrequencyinNGC 3115is due to metallicity andnot age(K03),butthis sys- 4.3. Local Group Sources temhasfewerX-raysources,sothe ratioofthe numberof Both the Milky Way and the Magellenic Clouds have LMXBs in metal rich and poor clusters cannot be as well been rather well studied in terms of their X-ray stellar determined. Finally there is the case of NGC 4365,where populations,andthey,showametallicitydifferenceonthe the ages do span a rather wide range, but do not seem sameorderasthatbetweenthe metalrichandmetalpoor to be strongly correlated with the LMXB number density modesforGCs-aboutafactorof10. Asonlyasmallfrac- (K03). tion of the X-ray sources in any of these galaxies is in a The two modes in color for NGC 4472peak at V −I of GC,thestarformationratehasasubstantialeffectonthe 0.98and1.23(KMZ02),correspondingto valuesof[Fe/H] relativenumberdensitiesofX-raybinaries,somerelycom- of -1.26 and -0.08, respectively, according to the scaling paringnumber countsperunit stellarmassis notlikelyto law of Kundu & Whitmore (1998). Defining the metal prove frutiful. However, one can be fairly confident that rich/metal poor mode boundary to be V − I=1.10, we the high mass X-ray binary population is not heavily af- find that 23 of the 450 metal rich GCs and only 7 of the fectedbyIIWsbecausetheluminositiesofhighmassstars 370 metal poor GCs contain X-ray sources. The metal are much larger than the intercepted and absorbed lumi- richGCsarethus2.7±1.2timesaslikelytocontainX-ray nosities. Therefore,the ratio of LMXBs to HMXBs might sources as the metal poor GCs. givearoughestimateofhowimportanttheIIWsare. The suggestionofITF97 thatthe difference ofthis ratiomight 4.2. NGC 4472 - source spectra be indicating that irradiation induced winds are playing Thespectraalsoshowadifferenceasafunctionofmetal- animportant role is therefore additionalevidence in favor licity. While the individual spectra cannot be easily mea- of this scenario. sured because of the low count rates, we have found that the summed spectra in NGC 4472 are harder in the blue 5. potential observational tests GCsthanintheredones(MKZ03). Ifweholdtheneutral hydrogen column in both cases to the Galactic value of Thismodelmakesseveraltestablepredictions. Thefirst 1.6×1020 cm−2 and fit a power law model to the data, we is that there should be a monotonic dependence between find a spectral index of 1.02±.27 for the blue GCs while thenumberofLMXBsperunitstellarmassandthemetal- we find a spectralindex of1.46±.10for the redGCs (90% licity; givenenoughstatistics we shouldsee a difference in errorcontours). Allowingthecolumntofloatfreelyforthe the LMXB specific frequency as a function of the metal- red GCs, we find N to be 4.8×1020 cm−2 and the power licityitself, andnotjustasafunctionofwhether acluster H law index to be 1.57. Then, we fix the power law index is in the metal rich or metal poor mode. Much new data for the blue GCs and find that the data is best fit with a has recently entered the Chandra archives, so it is now column density of 1.1×1021 cm−2. possible to test this prediction. There is already a corre- Wenotethatthesolutiontothespectraldifferenceprob- lation overa range of metallicities in the ROSAT spectral lem is not unique and is prone to numerous systematic indices of GC X-ray sources which does not show a “criti- uncertainties. The photoelectric absorption models we cal metallicity” (IB), so our model seems to pass this test have used assume a solar composition for the absorbing so far. medium, andthatthemedium iscold(i.e. completelyun- This scenario also predicts that neutron star LMXBs ionized). The underlying spectrum for accreting neutron willbe affectedfarmorethanothertypes of“dynamically stars and black holes in the 0.5-8 keV range is unlikely interesting”sources. Bluestragglersandcataclysmicvari- to be a single power law. Finally, we have averaged over ables will not generate high enough X-ray luminosities to Metallicity/LMXB No. density correlations 5 excite substantial IIWs. Black hole systems, because of absorption. Additionally,wenotethatthe fittingofSidoli the highermass compactobjects,will accretethe IIW gas etal. (2001)wasdoneusingthestandardassumptionthat much more efficiently. the absorber would be cold material of solar composition. IIWs have been suggested to explain why systems such Since BeppoSax is sensitive to absorption edges in the ∼ as 4U 1820-30show different period evolution than would fewkeVrange,thismaycauseasystematicerrorinthefit- be expected from conservative mass transfer driven by tedabsorptionvalue,asnotedabove. TheresultsofSidoli gravitational radition (Tavani 1991b). An alternative is et al. (2001) certainly place upper limits on the amount that the system is being effected by the interactions with of intrinsic absorption in the Milky Way’s LMXBs, but the GC potential(vander Klisetal. 1993). Gravitational theseupperlimitsaremostlytoohightoplacestrongcon- waveobservationsfromfuture missionssuchas LISA may straints on our model. A more sensitive test would come help break this degeneracy. fromXMM-NewtonspectraofM31globularclusterX-ray Additionally, our scenario predicts that the metallicity sources,where there will be little non-intrinsic absorption effects should be essentially the same for field sources as since the globular clusters will not be viewed through the they are for globular cluster sources. Given two galaxies disk of the Galaxy. In fact, many of the M31 globular with similar star formation histories, the more metal rich cluster X-ray sources show evidence for intrinsic absorp- galaxy should have more field X-ray binaries. The nat- tion,andtheonesthatdoarepredominantlyinmetalpoor ural way to test this hypothesis would be to look at the clusters (Irwin & Bregman 1999). field X-ray binary populations of elliptical galaxies,as (1) Differences inthe luminosity functions between redand they have very little recent star formation and (2) metal- blue globular cluster X-ray sources should also, in prin- licitytendstoscalewithgalaxymass. Apotentialproblem ciple, provide a way to discriminate between models for with this approachis thata fraction(and indeed, perhaps formation and evolution of their X-ray binaries. Unfortu- a large fraction)ofthe field X-raybinaries mayhave been nately,itisdifficultatthistimetomakeapredictionfrom created through stellar interactions in GCs and released ourscenario. Itisnotclearontheoreticalgroundswhether intothefieldthroughdynamicalejectionsorthroughtidal the extra wind energy in metal poor systems manifests it- destructionoftheGCs(seeMKZ03andreferenceswithin; self as a higher mass loss rate, yielding probably slightly seealsoGrindlay1988). Giventhatboththetidaldestruc- higher luminosities, albeit for much shorter amounts of tionrateandthemetallicityarelikelytobecorrelatedwith time, or as higher wind velocities, in which case the effi- themass,applyingthistestisnotstraightforward. Onthe ciency of wind capture is lower, so the luminosity will be other hand, the field sources of elliptical galaxies should lowerata givenmassloss rate, oras some combinationof showametallicityeffectontheirenergyspectraregardless the two. Furthermore, there is not yet a sufficiently large of concerns over formation processes. sample of X-ray binaries in globular clusters for making a A better test might then be to extend the suggestion good comparison of the luminosity functions. This does of ITF97 that the difference in the ratio of LMXBs to remain a good test to bear in mind for future work. HMXBs in the Milky Way and the LMC is due to the metallicity difference. Since most HMXBs are accretion 6. conclusions powered pulsars, a reasonably good separation between the bright ends of the luminosity distributions of HMXBs Wehaveoutlinedascenariowherebythetwometallicity andLMXBs shouldbe possiblygivengoodChandraspec- effects seeninLMXBs inglobularclusters,higher number traofnearbyspiralgalaxies. Wenotethatthisisageneric density in metal rich clusters, and harder low energy X- prediction of any model in which the metallicity effects ray spectral in metal poor clusters, can be explained via are strictly due to metallicity, but will provide a way to the same mechanism – irradiation induced stellar winds. distinguish between true metallicity effects and effects of We have presented additional feasible observational tests metallicitybeingcorrelatedwithmoredifficulttomeasure of this picture. While we have shown that the physics of parameters, such as the dynamics of the system. the irradiation induced winds required to reproduce the Past work on detailed spectral fitting on Milky Way observationsis consistentwith the mostrecenttheoretical GCs has been suggested to indicate that there is little work,we also note that this is a rather complicated prob- evidence for intrinsic absorption is these systems (Sidoli lemwhichisdeservingofconsiderableadditionalattention et al. 2001). We note that this is not in conflict with by experts in binary stellar structure and evolution. We our model. Sidoli et al (2001) did not obtain a satisfc- hope this paper will help to stimulate such work in the tory spectral fit to the data for M 15, the most metal future. poor of the Milky Way’s globular clusters with an X-ray source,probablybecauseBeppoSaxwasnotcapableofre- 7. acknowledgments solving the two brightX-raysourcesinthe cluster (White & Angelini 2001). The other two most metal poor clus- We acknowledge funding from NASA LTSA grants ters,NGC 1851andNGC 6712doshowexcessabsorption NAG5-12975 and NAG5-11319. We thank Michael Sip- in the X-rays compared with the optical, and the other ior, Simon Portegies Zwart and Jasinta Dewi for useful globularclustersinthe Milky Wayallhaveopticalextinc- comments onthis manuscript. We thank the referee,Josh tions significantly higher than the excess predicted by our Grindlay,forsuggestingthatweimprovetheclarityofthe model, so the fits would not be very sensitive to intrinsic paper and the depth of the background discussion. 6 Maccarone,Kundu & Zepf REFERENCES Arons,J.,1973,ApJ,184,539 Irwin,J.A.&Bregman,J.N.,1999,ApJL,510,21(IB) Bailyn,C.D.&Grindlay,J.E.,1990, ApJ,353,159 Kroupa,P.,2002,Science, 295,82 Basko,M.M.&Sunyaev, R.A.,1973,Ap&SS,23,117 Kundu, A., Maccarone, T.J. & Zepf, S.E., 2002, ApJL, 574, 5 Baumgardt,H.&Makino,J.,2003,MNRAS,340,227 (KMZ02) Beasley, M.A.,Sharples,R.M.,Bridges,T.J.,Hanes,D.A.,Zepf, Kundu,A.,Maccarone,T.J.,Zepf,S.E.&Puzia,T.H.,2003,ApJL, S.E.,Ashman,K.M.&Geisler,D.,2000,MNRAS,318,1249 589,81 Bellazzini,M.,Pasquali,A.,Federici,L.,Ferraro,F.R.,Pecci,F.Fusi, Kundu,A.&Whitmore,B.C.,1998,AJ,116,2841 1995,ApJ,439,687 Larson,R.B.,1998, MNRAS,301,569 Chaboyer, B., Demarque, P., Kernan, P. J., Krauss, L. M. & Maccarone, T.J., Kundu, A. & Zepf, S.E., 2003, ApJ, 586, 814 Sarajedini,A.,1996, MNRAS,283,683 (MKZ03) Clark,G.W.,1975, ApJL,199,143 Mikkola,S.,1984,MNRAS,207,115 Cohen, J.G.,Blakeslee,J.P.&Coˆt´e,P.,ApJ,592,866 Pfahl,E.,Rappaport, S.&Podsiadlowski,P.,2003, ApJ,597,1036 Cowley,A.P.,1994,inASPConf.56,InteractingBinarysystems,ed. Piotto,G.&Zoccali,M.,1999,A&A,345,485 A.W.Shafter(SanFrancisco:ASP),160 Podsiadlowski,P.,1991,Nature,350,136 Davies,M.B.&Hansen,B.,1998,MNRAS,301,15 Pooley,D.etal.,2003,ApJL,591,131L Day,C.S.R.&Stevens, I.R.,1993, ApJ,403,322 Puzia,T.H.,Zepf,S.E.,Kisslet-Patig,M.,Hilker,M.,Minniti,D.& DiStefano,R.,Kong,A.K.H.Garcia,M.R.,Barmby,P.,Greiner, Goodfroij,P.,2002,A&A,391,453 J.,Murray,S.S.&Primini,F.A.,2002,ApJ,570,618 Rasio,F.A.,Pfahl,E.D.&Rappaport,S.,2000,ApJL,532,47 DiStefano, R.&Rappaport, S.,1992,ApJ,396,587 Ruderman,M.,Shaham,J.,Tavani,M.&Eichler,D.1989,ApJ,343, Fabian,A.C.,Pringle,J.E.&Rees,M.J.,1975, MNRAS,172,15P 292 Frank, J., King, A. & Raine, D., 1992, Accretion Power in Sidoli,2001,L.,Parmar,A.N.,Oosterbroek,T.,Stella,L.,Verbunt, Astrophysics, Cambridge University Press: Cambridge, UK F.,Masetti,N.&DalFiume,D.A&A,368,451 (FKR92) Tavani,M.&London,R.,1993,ApJ,410,281 Fregeau, J.M.,Gu¨rkan, M.A.,Joshi,K.J.,Rasio,F.A.,2003,ApJ, Tavani,M.,1991a, ApJL,366,27 593,772 Tavani,M.,1991b, Nature,351,39 Grindlay,J.E.,1988, IAUSymp.126,pp.347 Tavani,M.,Ruderman,M.&Shaham,J.,1989,ApJL,342,31 Grindlay,J.E.,1993, ASPConferenceSeries48,156 Tout,C.,Pols,O.R.,Eggleton,P.P.,Han,Z.,1996,MNRAS,281,257 Harpaz,A.&Rappaport, S.,1991,ApJ,383,739 Trinchieri, G.; Israel, G. L.; Chiappetti, L.; Belloni, T.; Stella, L.; Heinke, C.O., Grindlay, J.E., Lugger, P.M., Cohn, H.N., Edmonds, Primini,F.;Fabbiano,P.;Pietsch,W.,1999,A&A,348,43 P.D.,Lloyd,D.A.&Cool,A.L.,ApJ,598,501 Tutukov, A.V.&Yungelson,L.R.,1980,SovietAstronomy,24,729 Hills,J.G.,1976,MNRAS,175,1P vanderKlis,M.Hasinger,G.;Verbunt,F.;vanParadijs,J.;Belloni, Iben,I. J., Tutukov, A. V., & Fedorova, A. V., 1997, ApJ, 486, 955 T.;Lewin,W.H.G.1993, A&ALetters,279,21 (ITF97) Vesperini,E.&Heggie,D.C.,1997,MNRAS,289,898 Iben, I.J.,Tutukov, A.V.&Yungelson, L.R.,1995,ApJS,100,233 White,N.E.&Angelini,L.,2001,ApJL,561,101