DraftversionFebruary1,2017 PreprinttypesetusingLATEXstyleAASTeX6v.1.0 ACHANDRAX-RAYCENSUSOFTHEINTERACTINGBINARIESINOLDOPENCLUSTERS–COLLINDER261 SmritiVats1andMaureenvandenBerg2,1 1AntonPannekoekInstituteforAstronomy,UniversityofAmsterdam,SciencePark904,1098XHAmsterdam,TheNetherlands;[email protected] and 2Harvard-SmithsonianCenterforAstrophysics,60GardenStreet,Cambridge,MA02138,USA;[email protected] 7 1 0 ABSTRACT 2 WepresentthefirstX-raystudyofCollinder261(Cr261),whichatanageof7Gyrisoneoftheoldestopen n clusters knownin the Galaxy. Our observationwith the ChandraX-ray Observatory is aimed at uncovering a J thecloseinteractingbinariesinCr261,andreachesalimitingX-rayluminosityofLX 4 1029ergs−1(0.3– ≈ × 1 7keV)forstarsinthecluster. We detect107sourceswithintheclusterhalf-massradiusrh, andweestimate 3 thatamongthesourceswithL &1030ergs 1, 26areassociatedwiththecluster. Weidentifyamixofactive X − ∼ binaries and candidate active binaries, candidate cataclysmic variables, and stars that have “straggled” from ] E the main locus of Cr261 in the colour-magnitudediagram. Based on a deep optical source catalogue of the H field,weestimatethatCr261hasanapproximatemassof6500 M ,roughlythesameastheoldopencluster ⊙ . NGC6791.TheX-rayemissivityofCr261issimilartothatofotheroldopenclusters,supportingthetrendthat h theyare more luminousin X-raysper unitmass thanold populationsof higher(globularclusters) and lower p - (thelocalneighbourhood)stellardensity.Thisimpliesthatthedynamicaldestructionofbinariesinthedensest o environmentsisnotsolelyresponsiblefortheobserveddifferencesinX-rayemissivity. r st Keywords:openclustersandassociations: individual(Collinder261);X-rays: binaries;binaries: close;stars: a activity;cataclysmicvariables [ 1 1. INTRODUCTION isaboutasoldastheSun(4 0.5Gyr;Dinescuetal.1995), v ± 1 OpenclusterswithagesinexcessofafewGyrarerelatively revealed a large number of X-ray sources among the clus- 8 ter members(Bellonietal. 1993). Manyof these turned out rareintheGalaxy(e.g.Kharchenkoetal.2013). Someaspect 8 to be close, tidally interacting binaries where the stellar ro- oftheirproperties(perhapstheirlargeinitialmassortheirlo- 8 tation is locked to the orbital period, and therefore kept at 0 cationoutoftheGalacticplane,wheretheyavoidinteractions . with large molecular clouds or the disruptive pull of exter- a level that can sustain magnetically active coronae. Sub- 1 sequent XMM-Newton (Gondoin 2005; Giardinoetal. 2008; 0 nal gravitational forces) helped them survive until old age. 7 Studies of old open clusters, with their well-developed sub- Gosnelletal. 2012), andChandra(vandenBergetal.2004, 1 2013;Giardinoetal.2008)observationsofoldopenclusters giantand giantbranches, have been a cornerstoneof stellar- v: evolution theory for many decades, thanks, in part, to their havedetected manysuch active binaries(ABs). ABs can be i binaries of two detached stars, or they can have a contact X accuratelymeasuredagesanddistances. orsemi-detachedconfigurationsuchasinWUMaandAlgol From the X-ray point of view, old open clusters are in- r a teresting for a number of reasons. First, X-ray observations binaries, respectively. In terms of number of sources, ABs are the most prominentX-ray source class in old openclus- efficiently detect different classes of close, interacting bina- ters,butotherclassesofinteractingbinaryarerepresentedas ries, enabling the study of processes such as tidal coupling well. In cataclysmic variables (CVs), the X-rays are the re- and the link between X-rays and rotation. The X-ray lumi- sult of accretion from a late-type main-sequencedonoronto nosity of late-type stars stronglydependson stellar rotation. a white dwarf. In fact, the first ROSAT observationof M67 Assingle starsage, theyspin downdueto magneticbraking wasaimedatstudyingtheX-raysfromaCVthatwasdiscov- (Pallavicini1989).Asaresult,theirX-rayemissiondecreases ered in the optical (Gillilandetal. 1991). The origin of the accordingly. An old star like our Sun ( 4.5 Gyr) has an X- ray luminosity of about 1026 to 1027 er∼g s 1 (0.1–2.4 keV; X-ray emission from more exotic open-cluster binaries, like − blue stragglers, is less well understood, but in X-rays they Peresetal.2000). Evenwiththedeepestexposuresofasen- aremoresimilartotheABsthantothemass-transfersources sitive X-ray telescope like the Chandra X-ray Observatory, (vandenBerg2013). thisisnearlyimpossibletodetectexceptfortheneareststars. A secondmotivefor studyingoldopenclustersin X-rays, Nevertheless, an early ROSAT observation of the old open isthattheir stellar densitieslie in betweenthose of thesolar clusterM67,whichliesat 840pc(Pasquinietal.2008)and ∼ 2 neighbourhood ( 0.1M pc 3) and dense globular clusters canonicalratio A /E(B V) = 3.1, and a neutral hydrogen − V ( 104M pc 3).∼This a⊙llows an investigation of the effect column density N = 1−.9 1021 cm 2 (Predehl&Schmitt − H − ≥ ⊙ × of stellar dynamics on the clusters’ close-binary population, 1995). The Galactic coordinates of Cr261 are l = 301.7 , ◦ in a poorly studied density regime. With the growing sam- b = 5.5 ; due to its low Galactic latitude and location to- ◦ − ple ofoldopenclustersstudiedin X-rays, itis nowpossible wards the bulge, the number of fore- and background stars to do simple statistics regarding the number of sources de- projected onto the cluster is high. Cluster membership is tected in each sourceclass. It was foundthat the numberof poorlyconstrainedforthemajorityofstarsinthefield.Cr261 CVsinM67andNGC6791scalewiththepresent-daycluster is includedin the star cluster catalogue of Kharchenkoetal. mass, pointingat a primordialorigin. For ABs, that propor- (2013), which lists structural parameters such as the overall tionality is not so obvious, raising the issue of whether dy- sizeoftheclusterandtheradiusofitscentralregion. Inthis namical interactions that break up or create binaries, play a work,we presentanestimate forthehalf-massradiusr and h role (vandenBergetal. 2013). The expectedlow encounter the approximate mass of Cr261, which, to our knowledge, ratesinopenclustersdonotseemtofavourthelatterexplana- havenotbeenreportedintheliteraturebefore.Theseparame- tion. Nevertheless,therearecluesthatdynamicalencounters tersfacilitateauniformcomparisonwiththeX-rayproperties shapethepropertiesofatleastsomebinaries.N-bodymodels ofotheroldGalacticclusters. ofM67(Hurleyetal.2005)suggestthatprimordialbinaries WepresenttheX-rayandopticalobservations,andthedata anddynamicalencountersarenecessaryto explaintheblue- reduction in Sect. 2. In Sect. 3 we describe the analysis, stragglerpopulationofM67. Someindividualsystems,such which includes the creation of the X-ray and optical source asthe likely tripleS1082in M67 (vandenBergetal. 2001; catalogues, their cross-correlationto identifycandidateopti- Sandquistetal. 2003) is also difficult to explain without in- calcounterpartstotheChandrasources,andthederivationof vokingencounters.Therefore,theoriginoftheX-raysources the structuralpropertiesof Cr261. Sect. 4 is focusedon the ofoldopenclustersmaynotbesolelyprimordial. X-raysourceclassification. InSect.5we discussourresults TheX-rayemissivity,ortheintegratedX-rayluminosityper in the context of the X-ray emission from other old stellar unitofmass,ofglobularclustersislowerthanthatofM67af- populations,andwesummariseourfindingsinSect.6. terremovingthecontributionfromluminouslow-massX-ray 2. OBSERVATIONSANDDATAREDUCTION binaries(LMXBs; e.g.Verbunt2001). Geetal.(2015)com- pared the X-ray emissivities of more diverse environments 2.1. X-rayObservations including dwarf elliptical galaxies and the local neighbour- Cr261 was observed with the Advanced CCD Imaging hood,andfoundthatoldopenclustersalsohavehigherX-ray Spectrometer(ACIS;Garmireetal.2003)onboardChandra emissivities than other old stellar populations. Various ex- starting 2009 November 9 14:50 UTC for a total exposure planationshavebeensuggested,relatingtoeithertheoverall timeof53.8ks(ObsID11308).Theobservationwasmadein mass-losshistoryoftheclusters,differencesindynamicalen- VeryFaint,Timedexposuremode,withasingleframeexpo- counterrates,ortheprocessesunderlyingtheX-rayemission. suretimeof3.2s. Kharchenkoetal.(2013)estimatethatthe Morestudyisneededtodeterminewhichofthesefactorsare radius1 of Cr261 is 14.1. This is considerably larger than ′ responsible. ∼ a single ACIS chip (8.4 8.4); therefore, we placed the cen- ′ ′ InordertoimprovethecensusofX-raysourcesinoldopen tre ofthe cluster (α ×= 12h38m06s.0, δ = 68 2201 ; 2000 2000 ◦ ′ ′′ clusters, we are undertakinga survey with Chandraof open Kharchenkoetal. 20132) close to the I3 aimpo−int, so that a clusterswithagesbetween3.5and10Gyr. Theobservations largercontiguouspartoftheclustercouldbeimaged(seeFig- are designed to reach a limiting luminosity of L 1030 X ure1).TheCCDsusedwereI0,I1,I2andI3fromtheACIS- ≈ erg s 1 (0.3–7 keV), or better, at the distance of the clus- − Iarray,andS2andS3fromtheACIS-Sarray. ters. As part of this survey, we have carried out the first X- Westartedthedatareductionwiththelevel-1eventfilepro- ray study of Collinder 261 (Cr261), and we present the re- ducedby the data processingpipelineof the ChandraX-ray sults of our efforts in this paper. With an estimated age of Center and used CIAO 4.5 with CALDB 4.5.5.1 calibration 6–7 Gyr (Bragaglia&Tosi 2006), Cr261 is one of the old- filesforfurtherprocessing.Tocreatethelevel-2eventfilewe est open clusters in the Galaxy, being superseded in age by used the chandra_reproscript. A background light curve NGC6791(8–9Gyr)andBerkeley17(8.5–10Gyr)only.The in the energy range 0.3–7 keV was created with the CIAO clustermetallicityisclosetosolar(Drazdauskasetal.2016), dmextract routine using source-free areas on the ACIS-I andreportedvaluesforthedistanceandreddeningliebetween chips, and was analysed with the lc_sigma_clip routine. 2.2–2.7 kpc and E(B V) 0.25 0.34, respectively (see Nobackgroundflareswithmorethan3σexcursionsfromthe − ≈ − e.g.Gozzolietal.1996,Carraroetal.1999,Bragaglia&Tosi averagebackgroundcountratewereobserved,hencethetotal 2006),withahighervalueofthereddeningconsideredmore plausible(Frieletal.2003). Inthispaper,weadoptadistance 1 Here we refer to the Kharchenkoetal. (2013) parameter r2, which is defined as the distance from the cluster centre where the projected stellar of2.5kpcandE(B V)=0.34,unlessstatedotherwise. The densitydropstotheaveragestellardensityofthefield. − latter correspondsto a V-band extinction A = 1.05 for the 2TheclustercentreisredeterminedinSect.3.3. V AChandrastudyoftheoldopenclusterCollinder261 3 Basicdatareductionstepsofbiassubtractionandflat-fielding wereperformedusingthebias,anddome-andsky-flatimages takenwithinonedayofthescienceexposures.WiththeMSC- CMATCHroutine,thegeometricdistortionoftheimageswas removed, and the eight individualchips of a given exposure werecombinedinto a single image. We createda master V- band image by stacking the two individual, slightly offset, 240-s V-band exposures. As a result, in the stacked V im- agethespacebetweentheindividualchipsoftheWFImosaic (23 widealongthelengthofthechips,and14 widealong ′′ ′′ theirwidth)islargely,butnotcompletely,filledin(seeFigure 1). ThestellarprofilesinoneoftheB-bandimagesofCr261 areverydistorted,whichpreventedusfrommodellingagood PSF.Sincethisdegradesthequalityofthederivedphotome- try, we opted to discard this low-qualityimage and use only asingle240-sBexposureforouranalysis. Therefore,ourB- bandcatalogueoftheCr261fieldhasnocoverageinthechip gaps. Figure1. Stacked WFI V-band image of Cr261. The four 3. ANALYSIS black squares show the ACIS-I field of view with chip IDs 3.1. X-raySourceDetectionandSourceCharacterisation marked. The solid white circle marks the core radius of WelimitedtheX-rayanalysistothedatafromchipsI0,I1, the cluster (r = 157 16 ), centred on the cluster centre c ′′ ± ′′ I2andI3. TheS2andS3chipsliefarfromtheI3aimpoint, (markedbytheredcross)asdeterminedbyus(seeSect.3.3). givingrise to largepositionalerrorson anysourcesdetected The dashed white circle marks the half-mass radius of the on them. Such large errors make it hard to identify optical cluster (r = 384 38 ). Blue rectangles show the off- h ′′ ± ′′ counterparts,andthustoclassifythesources. setfieldusedfordeterminingthebackgroundstellardensity. Sourcedetectionwasdonein a soft(0.3–2keV),hard(2– Small white rectangles are regionsof zero optical exposure. 7keV),andbroad(0.3–7keV)energyband.TheCIAOsource Northisup,easttotheleft. detectionroutinewavdetectwasrunforeightwaveletscales exposurewasusedforfurtheranalysis. ranging from 1.0 to 11.3 pixels, each increasing by a fac- tor of √2. Larger scales are better suited for more off-axis 2.2. OpticalObservations sources, where the PSF is wider or more distorted. Expo- suremapswerecomputedforanenergyvalueof1.5keV.The WeretrievedopticalimagesofCr261intheBandV bands wavdetectdetectionthreshold(sigthresh)wassetat10 7. fromtheESOpublicarchive.Thesedataweretakenaspartof − The corresponding expected number of spurious detections theESOImagingSurvey(EIS;programID164.O-0561).The per wavelet scale is 0.42 for all four ACIS chips combined, observationsof Cr261 were made using the Wide Field Im- ager(WFI),mountedonthe2.2mMPG/ESOtelescopeatLa or 3.35 in total for all wavelet scales. We ran wavdetect forthethreedifferentenergybandsandthencross-correlated Silla,Chile. TheWFIhasafieldofviewof34 33 covered ′ ′ × theresultingsourceliststoobtainamasterX-raysourcelist. byadetectorarrayofeight2k 4kCCDswithapixelscaleof 0.238pixel 1. TheCr261data×weretakenfrom2001June27 We detected113distinct X-raysources. To checkif we had ′′ − missed any real sources, we ran wavdetectagain for a de- 23:55UTCto2001June2800:38UTC,withatotalexposure tectionthresholdof 10 6, which increasesthe expectedtotal time of 510 s in the B and V filter each. In each filter, two − number of spurious detections to 33.5. We found a total of exposures of 240 s were taken, supplemented with a single 151 distinct X-ray sources with more than two counts (0.3– shortexposureof30stogetphotometryforthebrightstars. 7 keV) in this case. The positions of seven of the extra 38 Weonlyusedthelongexposuresforouranalysis. Theseeing sourcesarefoundtomatchthoseofshort-periodbinariesdis- duringtheobservationswas 1.15. ′′ ∼ coveredby Mazuretal. (1995) (see Sect. 3.4). Close, inter- ForreducingtheopticalimagesweusedtheImageReduc- tion and Analysis Facility (IRAF3) v2.16, supplemented by acting binaries are plausible real X-ray sources and indeed theexpectednumberofchancealignmentsbetweentheChan- theMSCREDpackageforhandlingandreducingmosaicdata. dradetectionsandthebinariesintheMazurcatalogueisvery 3 IRAF is distributed by the National Optical Astronomy Observatory, low(Sect.3.5). Itisthereforelikelythatatleasttheseseven which is operated by the Association of Universities for Research in As- additionalsourcesarereal;butgiventhe 34spuriousdetec- tronomy(AURA)underacooperativeagreementwiththeNationalScience ∼ Foundation. tionsthatareexpected,wedonotbelievethattherearemany 4 E E 50 50 0.6 1 2 3 4 5 k eV 0.6 1 2 3 4 5 k eV 2.5 2.5 a. b. V) 2.0 2.0 e k 3 0. E-75 V)/( 1.5 1.5 e k 3 0. E-25 3 ( 1.0 1.0 0.5 0.5 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.6 1 2 3 4 5 k eV 0.6 1 2 3 4 5 k eV 2.5 2.5 c. d. V) 2.0 2.0 e k 3 0. E-75 V)/( 1.5 1.5 e k 3 0. E-25 3 ( 1.0 1.0 0.5 0.5 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 log(E /0.3 keV)/log(7 keV/0.3 keV) log(E /0.3 keV)/log(7 keV/0.3 keV) 50 50 Figure2. Quantile diagrams with model grids representing a MeKaL plasma (left) and a power-law spectrum (right). The top panels show sources without, and the bottom panels show sources with candidate optical counterparts (see Table 2 for their classification). The plasma temperaturekT or photonindexΓ, and the columndensity N can be estimated from the location H ofasourcewithrespecttothegrid: bluecurvesrepresentlinesofconstant N normalisedinunitsof1022 cm 2 (N , where H − H,22 N 0.19cm 2istheclustervalue),whereasorangecurvesarelinesofconstantkT (labeledinunitsofkeV;left),andyellow H,22 − ≈ curvesarelinesofconstantΓ(right). ThemedianenergyE canbereadofffromthetopx-axis. Hereweshowsourceswith 50 twentynetcounts(0.3–7keV)ormore;errorbarsareshownonlyforthesourceswiththehighestandlowestnumberofcounts inagivenpanel. FilledcolouredsymbolsmarkX-raysourcesforwhichwehavefoundcandidateopticalcounterparts. Among them,greencirclesrepresentABsandcandidateABs(Sect.4.1),olivetrianglesareforcandidateCVsorAGNs(Sect.4.2),yellow four-pointstarsforcandidateBSSs,palereddownwardtrianglesforcandidateSSGs,maroondownwardtrianglesforcandidate YSS (Sect. 4.3), pale blue squaresforlikely non-membersof the cluster (Sect. 4.4), and deep blue diamondsforsourceswith uncertain classification (Sect. 4.5). Furthermore, sources CX18 and CX27 that have close-binary counterparts (Mazuretal. 1995)aremarkedwithalargerblackopencircle. morerealsourcesamongtheextradetections.Weflaggedthe tine provides us with the source positions, but is not op- sources that are only found for sigthresh=10 6, but kept timised to measure source counts. We determined the net − theminthemastersourcelist. sourcecountsusingACISExtract(Broosetal.2010,version For computing the positional uncertainties, required for 2013mar6).Alleventsbetween0.3and7keVwereextracted cross-correlationwithothersourcecatalogues,weusedEq.5 from regions enclosing 90% of the PSF at 1.5 keV. ACIS ∼ fromHongetal.(2005),whichgivesthe95%confidencera- Extractalso performsvariability characterisation based on a diuson the wavdetectposition, P . The wavdetectrou- Kolmogorov-Smirnov(K-S)testontheeventarrivaltimesfor err AChandrastudyoftheoldopenclusterCollinder261 5 sources with five counts or more that spend more than 90% In order to characterise the spectral properties of our X- ofthetotalexposuretimeontheACIS-Ichips4. Forexample, ray sourceswe use quantile analysis, which is optimisedfor asourcenearachipedgecouldeffectivelyhaveashorterex- sources with few counts (Hongetal. 2004). In this method, posuretimeifthetelescopedithermotionoccasionallymoves the median energy, E , and 25% and 75% quartile energies 50 it off the detector. There were 76 sources with no evidence (E and E , respectively)of the sourceevents’ energydis- 25 75 forvariability(0.05< P );foursourceswhichshowedpos- tributionareusedtodeterminespectralhardnessandspectral KS sible variability(0.005 < P < 0.05; CX9, CX13, CX64, shape. ConventionalX-ray hardness ratios use fixed energy KS CX93) and four which were likely variable (P < 0.005; values for defining hard and soft energy bands, and give re- KS CX63, CX91, CX120, CX137), where P is the proba- sultsoflittlemeaningifalleventslieineitherthesoftorthe KS bility for a source to have a constant count rate. The X-ray hard energy band. Details of the source properties are pre- lightcurvesofthevariablessuggestflare-likebehaviour,with sentedinTable1,whilequantilediagramsareshowninFig- alargefractionofthetotaleventsarrivinginarelativelyshort ure2forsourceswith 20netcounts(0.3–7keV). ≥ time interval. The brightest of these sources is CX63, for 3.2. OpticalSourceCatalogue whichthirteenofseventeeneventsarriveinthelast3.5hofthe observation (and nine of seventeen events in a single hour). The absolute astrometry of the optical images was tied Fortheotherthreesources,80%ormoreoftheeventsarrive to the International Celestial Reference System (ICRS). We within2.5–3h. X-rayflaresarecommonlyobservedinactive did this by computing an astrometric solution based on the late-typestarsorbinaries. Thisisconsistentwith ourclassi- positions of unsaturated stars in the field that are also in- ficationofCX120(aWUMabinaryandlikelynon-member cluded in the USNO CCD Astrograph Catalog 4 (UCAC4; ofthecluster)andCX91andCX137(likelyforegroundlate- Zachariasetal. 2013). For the V image we used 1912 un- typedwarfs).TheclassificationofCX63islesssecure,butit saturated stars and obtained rms residuals of 0.141 in right ′′ couldbealate-typestarorbinaryaswell. Thesesourcesare ascensionand0.166indeclinationinthesolution. Forthe B ′′ furtherdiscussedinSects.4.4and4.5. image,weused1773unsaturatedstarsandobtainedresiduals Onlyfivesourcesinourcataloguehavemorethan100net of0.157inrightascensionand0.173indeclination. ′′ ′′ counts(0.3–7keV),withthebrightestsourcehaving475net Forperformingphotometry,weusedtheDAOPHOTpackage counts. For the majority of our sources the spectrum of the in IRAF. Aftercreatinga sourcecatalogueforthe B andthe X-ray emission is therefore poorly constrained. We calcu- V imageseparately,wecross-matchedeachofthemwiththe latedunabsorbedfluxvaluesinthe0.3–7keVband,F ,for Gozzolietal.(1996)catalogueinordertoconvertourinstru- X,u each source from its net count rate and local rmf and arf mentalmagnitudestotheGozzolietal.calibratedmagnitudes response files using Sherpa. We assumed a 2 keV MeKaL intheJohnsonsystem. TheGozzolistudycoversaregionof model (xsmekal) attenuated by a neutral hydrogen column radius3.5aroundtheiradoptedclustercentre.Wefound2018 ′ density N = 1.9 1021 cm 2 (the value for Cr261) using matchesforthesourcesintheBcataloguewithinacalibrated- H − × thexstbabsmodel.TheMeKaLmodeldescribestheemission magnituderange13.7< B<24.0,and2276matchesfortheV froma hot, diffuse gasor optically thin plasma, as is appro- cataloguewithinarangeof13.0<V <22.5.Wemanuallyin- priate for ABs. Since the nature of our sources is unknown spectedallthematchedsourcesandfoundthatnoneappeared a priori, and the number of counts is too low to do any de- to be blended or saturated. Over these magnitude ranges, a tailedspectralfitting,weexploredtheeffectofusingdifferent constant offset provides a good transformation from instru- spectralmodelsonthederivedvaluesof F . We compared mental to calibrated magnitudes. The resulting WFI source X,u theunabsorbedfluxvaluesobtainedusingthe2keVMeKaL listsfortheentirefieldhavea calibratedmagnituderangeof model with those obtained using a 1 keV MeKaL model, a 12.9–23.5inVand13.7–24.6inB.Finally,wecross-matched 10keVthermalbremsstrahlungmodel(xsbrems)andapower- the BandV sourceliststomakeamasteropticalsourcelist. law model (xspowerlaw) with a photon index, Γ, set to 1.4; SomesourcesinthemastercatalogueweredetectedintheV thexstbabsmodelwasusedinallcases. Thefluxvaluesob- bandbutwerenotpresentinthesingleBimageduetoitschip tainedusingthesemodelswereabout6%smaller,40%larger, gapsandashorterexposure. and80%larger,respectively,thanthefluxvalueobtainedus- The colour-magnitudediagram (CMD) of Figure 3 shows ingthe2keVMeKaLmodel. TheX-raysensitivitylimit, as ourBandV photometryofstarsinsider (seenextsection). h definedby theunabsorbedflux ofthe faintestdetection, was foundto be 6 10 16 ergcm 2 s 1 forthe 2 keV MeKaL 3.3. EstimatefortheHalf-massRadiusandMassofCr261 − − − ∼ × modeland assumed cluster N , which correspondsto an X- OneofouraimsistocomparethenumberofX-raysources H ray luminosity of L 4 1029 erg s 1 (0.3–7 keV) at the in Cr261 with those detected in other old open clusters X − ≈ × adoptedclusterdistance(2.5kpc). and globular clusters. Making a uniform comparison be- tweenclustersrequiresanestimatefortheirmassesandstruc- 4basedontheevaluationoftheFRACEXPOkeywordgeneratedbymkarf tural parameters. An estimate for the King-profile (King inCIAO 1962)coreradiusr ofCr261wasderivedbyFroebrichetal. c 6 7 100 30 129 40 95 73 123 18 74 132 130 9 86 136 67 12 31 58 120 15 151 24 10 38 91 8 54 77 53 4 44 137 20 38 70 Figure3. Colour-magnitudediagramofCr261basedontheWFIphotometry. Thedifferentsymbolsandcolourshavethesame meaningasinFigure2. Furthermore,sourceswithclose-binarycounterpartsarecircledwithalargerblackopencircle;forthese stars,theBVphotometryisobtainedfrom(Mazuretal.1995,seeSect.4.1).Solidlinesrepresentisochrones(Bressanetal.2012; Chenetal.2014;Tangetal.2014;Chenetal.2015)fortheupperandlowerlimitsoftheclusterreddening(redforE(B V) = low − 0.25,andblackfor E(B V) = 0.34). Dottedlinesarethesameisochronesbutshiftedupwardby 0.75magtoindicatethe up − − limitforunresolvedphotometricbinaries.Thedashedlinerepresentsazero-agemain-sequenceisochrone.Errorbarsaresmaller than the symbols for some sources. The combineduncertainty in the location of the isochrone due to reddeningand distance uncertaintiesisshownintherectangleatthetopright.Forclarity,theABsanduncertainclassificationshavenotbeenlabeled. (2010)andKharchenkoetal.(2013), whofoundtwosignifi- motionsin the field of Cr261 (used by Kharchenkoet al. to cantlydifferentvalues,viz.52 and192 48 , respectively. weedoutpossible non-members)haverelativelylargeerrors ′′ ′′ ′′ ± Thesevaluesarebasedonthe2MASSnear-infraredcatalogue ( 9masyr 1,onaverage)anddonotdisplayacleardistinc- − ∼ (Skrutskieetal.2006),and(incaseofKharchenkoetal.only) tionbetweenclusterstarsandfieldstars.Wedecidedtoderive the optical PPMXL catalogue (Ro¨seretal. 2010). Both star ourownestimateofr andr ,withoutmakinguseofthePP- c h listsarerelativelyshallow,reaching.1magbelowtheCr261 MXLpropermotions. main-sequenceturnoff.Atthesametime,thePPMXLproper In order to estimate r for Cr261, we assumed that the h AChandrastudyoftheoldopenclusterCollinder261 7 Bellazzinietal. (2008). We calculated the integrated mag- nitude of stars inside r (i.e. I (V)) by summing the V-band h h fluxes of the stars inside r that satisfy the same magnitude h andcolourrestrictionsasoutlinedabove. Again,photometry of the offset fields provides a correction for the flux density of foregroundand backgroundstars. We converted I (V) to h theabsoluteintegratedV magnitudeofstarsinsider ,which h resulted in M (V) 3.6. Next, we compared this value h ≈ − with the theoretical curves for the evolution in time of the absoluteV magnitudeofsolar-metallicitystarclustersofvar- ious,constant,masses(Bellazzinietal.2008;Bragagliaetal. 2012). The age of Cr261 (7 Gyr) combined with our esti- mateforM (V),yieldsanapproximatevalueforhalftheclus- Figure4. Projected stellar density profile of Cr261 after cor- h ter mass of 4000–5500 M . The uncertaintystems fromthe rectionforthecontributionfromforegroundandbackground ⊙ rangespannedbythe theoreticalcurvescomputedfordiffer- stars (points connected with a dashed line). The solid blue ent initial-mass functions. As a final step, we have reduced line shows the best-fitting King profile (King 1962), which theinferredtotalmass(about8000–11000 M )withanem- has a central stellar density f = 0.0057 0.0007 stars per arcsecond2andacoreradiusr0 =157 1±6 . piricalscalingfactor. Thiswasmotivatedbyo⊙urfindingthat c ′′ ′′ ± the above method overestimates the masses of the old open clustersM67(byafactor1.1–1.7)andNGC188(byafactor stars are symmetrically distributed about the cluster centre of1.3–1.9),forwhichaccuratevirialmasseshavebeendeter- according to a King profile. In Figure 4 we plot the pro- mined(Gelleretal. 2008, 2015). Afterscaling, ourestimate jectedstellardensity f(r)versusradialoffsetfromthecluster forthetotalmassofCr261isabout5800–7200M . ⊙ centre r, computed in 50 -wide annular regions around the Obviously, our mass estimate should be consideredas ap- ′′ centre. Stars were selected from the region between two 7- proximate, only: we assumed that the total cluster mass is Gyrisochronesofsolarmetallicity(Z = 0.019;Bressanetal. containedwithinKharchenkoetal.’sclusterradiusr ,wehave 2 2012). One is modified for a distance of 2.5 kpc and red- no comprehensivelist of members, and the evolutionary se- denedbyE(B V)=0.34;thesecondisochroneisthesame, quencesfor M(V)fromBellazzinietal.(2008)maynotbea − but shifted upwardin the CMD by –0.75mag. This is done perfectmatchtoCr261(inmetallicityormassfunction).For to include the contribution from unresolved photometric bi- ourpurposes,though,thisestimateisgoodenough. naries (see Figure 3). To correct for the contribution from starsthatareunrelatedtothecluster,weestimatedthedensity 3.4. OpticalandX-rayCross-matching of stars within the same magnitude and colour limits, from Thepossible errorin the alignmentof Chandra’sabsolute a catalogue we created from offset fields in our WFI image astrometry to the ICRS is small5, but still allows for a sys- (the blue rectangles in Figure 1) that lie outside the cluster tematic offset between the X-ray positions and our ICRS- radius r (Kharchenkoetal. 2013). We fitted the function 2 calibrated optical positions. This systematic offset, or bore- f(r) = f (1+(r/r )2) 1 to this background-correctedradial 0 c − sight, canbe comparablein size to the randomerrorsonthe profile; here f is the central projected stellar density. This 0 X-raypositions(P ),andcancomplicatethesearchforop- function is the limit of the King profile for the assumption err ticalcounterpartsif notcorrectedfor. To calculatethe bore- that the tidal radius r is much larger than r . In the case t c sight, we used the 45 short-period (P < 3 d) close binaries of Cr261, the values of r 0.8 and r 22.2 as derived c ≈ ′ t ≈ ′ that were discovered by Mazuretal. (1995) in an optical- by Kharchenkoetal. (2013), support this assumption. Fit- variability study of the Cr261 field. The reason for using tingthe abovefunctionto the radial-densityprofile, givesus theshort-periodvariablesforcalculatingtheboresightisthat a centre for the cluster that is about 1.5 different from the ′ close binaries are plausible X-ray emitters and hence, there one given by Kharchenko et al., viz. α = 12h38m07s.1, 2000 isalowerchanceofspuriousdetectionsthatcouldaffectthe δ = 68 2333 ,withaformaluncertaintyofabout16 . 2000 − ◦ ′ ′′ ′′ boresightmeasurement(seealsoSect.3.5). Withthefinding We haveused this new cluster centrefor allpurposesin this charts in Mazur et al., we were able to identify all 45 vari- study.Thebest-fittingKingprofilehasr =157 16 ,con- c ′′± ′′ ablesamongtheWFIsources.TheirWFIpositionswerethen sistent with the value from Kharchenkoet al. Adoptingthis cross-matchedwiththeX-raycatalogue,whereweadopteda valueof r andassumingthatthe total massof the cluster is c 95% match radius that combinesthe error in the optical po- containedwithinr ,weusedEq.A3inFreireetal.(2005)to 2 computer . Wefindthatr =384 38 . h h ′′± ′′ 5 For ACIS-I, the 95% confidence radius on the alignment We estimated the mass of Cr261 from the integrated V is 0.9–1 within a distance of 3 from the aimpoint; see magnitude I(V) of the cluster, following the method of http:∼//c′x′c.ha′r′vard.edu/cal/ASPECT/celmon ′ 8 sitions6 and the random error on the X-ray positions (P ) which of the energy ranges considered in the Kim study is err in quadrature. To account for errors in the alignment, we closest to our broad band (0.3–7keV). To convertcounts to also add the 95% confidence radius on Chandra’s absolute fluxes we adopted a power-law spectrum with Γ = 1.4 and astrometry. Fifteen candidate counterparts were thus found, N =2.3 1021cm 2,i.e.equaltothetotalintegratedGalactic H − × whichwerethenusedtocalculatetheboresightfromtheav- columndensityalongthelineofsight(Marshalletal.2006). erageX-ray–opticalpositionaloffsets. AfterupdatingtheX- We calculated N forr < r , where mostX-raysourcesthat B c raypositionsforthisinitialboresight,thecross-matchingwas aretrulyassociatedwithCr261areexpectedtolie. Therea- repeated until the net boresightconverged. This method for son is that closer to the centre, the density of cluster stars is calculatingandcorrectingfortheboresightisoutlinedinde- simply higher; in addition, mass segregation makes the ra- tail in Sect. 3.3.1 of vandenBergetal. (2013). We found a dialdistributionofbinaries(andthuspotentialX-raysources) small boresight that is consistent with zero, viz. 0.06 0.07 moreconcentrated. For a 5-countdetectionlimit, we expect ′′ ′′ ± inrightascensionand0.09 0.08indeclination. N 9.5 3.1versus22sourcesactuallydetected. Fora10- ′′ ′′ B ± ≈ ± AftercorrectingtheX-raysourcepositionsforthe(almost countlimit, itisexpectedthat 5.7 2.4ofthetensources ∼ ± negligible)boresight,wematchedourX-raysourcelistwith detectedareextra-galactic.Intheregionr <r r ,45.7 6.8 c h ≤ ± the entire optical source list, again using 95% match radii. ofthe58sourcesdetectedabove5counts,or27.3 5.2ofthe ± For89uniqueX-raysources,wefound124opticalmatches; 38sourcesdetectedabove10countsareexpectedtobeextra- ofthelatter,104arepresentinboththeV andBimageswhile galactic. These numbers indicate that we do detect a popu- for20weonlyhaveaV orBdetection.Wealsoinspectedthe lationof, mainly faint, X-raysourcesthatis truly associated areaaroundeachX-raysourceintheWFIimagesbyeye,and withthecluster. discoveredthatfivemoreX-raysourceshavecandidateopti- GiventhelowGalacticlatitudeofCr261,afewforeground calcounterpartsthataresaturatedandthereforemissingfrom X-ray sources are also expected to contaminate our sample. our optical catalogue. Finally, we added to the list of can- TheexactnumberishardtoestimatesincethereisnoGalactic didatecounterpartssixopticalsourcesthatliejustoutsidethe X-raysourcedensitydistributionforthislatitudethatreaches 95%matchradius,butinsidethe3-σradius.Intotal,98ofthe downtoourdetectionlimit. WehaveusedthelogN logS − 151uniqueX-raysourceswerethusmatchedtooneormore curvesfromFigure9inEbisawaetal.(2005)forthesoftband opticalsources. Foracompletelistofcandidatecounterparts (0.5–2keV),andreadoffthelogN forafluxlimitthatcorre- andtheiropticalpropertieswerefertoTable2. spondstoa5-countdetectionemittinga2keVMeKaLspec- trum and N = 1.9 1021 cm 2. We expect 2.0 Galac- H − 3.5. FalsePositivesTest,BackgroundGalaxies,andGalactic × ∼ tic sources in the region inside r and 8.2 sources in the c Sources ∼ r < r < r annulus; this must be an upper limit since the c h To estimate the number of spurious matches between our EbisawafieldliesrightintheplanewhileCr261isafewde- X-ray and optical sources, we calculated the surface density grees off. Other factors, such as the difference between our of optical sources. Within r , the average density is 0.029 andEbisawa’ssoftband,anduncertaintiesintheX-rayspec- c sourcesarcsec 2,whilebetweenr andr itdropsslightlyto tralmodelandN ,alsoaffecttheaccuracyofthisnumber. − c h H 0.024 sources arcsec 2. Multiplying the optical source den- − sities with the totalarea coveredby the 95%errorcircles of 4. RESULTS theX-raysourcesinthetworegions,weexpect2.4spurious We usedthreecriteriatoclassifyourX-raysources. First, matchesamongthe23matchesthatwefindinthiscentralre- weconsideredthehardnessoftheX-rayspectrumasinferred gion,and11.6spuriousmatchesamongthe47matchesinthe from the energy quantiles. Coronally active stars and bina- outerregion.Similarly,weusethenumberofMazurvariables rieshavethermalX-rayspectrawithplasmatemperaturesthat per arcsec2 to estimate that the number of spurious matches generallydonotexceed3 4keV(e.g.Gu¨del2004). Thein- between X-ray sources and variables is 0.021 out of seven − tegrated Galactic column density in the direction of Cr261 matchesin the inner region,and0.022outof sevenmatches is 2.3 1021 cm 2; Galactic X-raysourceswithoutanyin- − fortheouterregion(oneX-ray–detectedMazurvariableslies ∼ × trinsicabsorptionshouldthereforehaveanN notlargerthan H outside r ). Therefore, all Mazur binaries that match with a h this. AsaresultofthesetemperatureandN constraints,the H Chandrasourcearelikelyrealcounterparts. expectedE valuesforcoronalsourcesarenotmuchhigher 50 In order to estimate the number of background galaxies than 1.5 keV7. On the other hand, accreting binaries with N amongourX-raydetections,we usedtherelationforthe ∼ B compactobjects,andAGNsoftenhaveintrinsicallyharderX- cumulative number density of high–galactic-latitude X-ray ray spectra, and sometimes are observedthroughadditional, sourcesaboveagivenfluxlimitS (Eq.5inKimetal.2007). localisedobscuringmaterial;inbothcases, theexpected E 50 WeadoptedthelogN logS relationforthe0.3–8keVband, B ishigherthan 1.5keV. − ∼ 6 Theerrors onthe optical positions that areadopted here arethe 1-σ errorsintheastrometriccalibrationgiveninSect.3.2,scaledtoa95%confi- 7 IntheMeKalgridofFigures2aand2c,thelocationkT 4keVand denceradiusassuminga2-Dgaussianerrordistribution NH ≈2×1021cm−2correspondstoE50≈1.5keV(seetopaxe≈s). AChandrastudyoftheoldopenclusterCollinder261 9 Secondly, we looked at the ratio of the unabsorbedX-ray toopticalflux,orthelimitsthereonforsourceswithoutcan- didateopticalcounterparts. We calculatedthisratiowith the equationlog(F /F ) = logF +V /2.5+5.44,wherethe X V u X,u 0 lasttermisthelogarithmoftheV-bandfluxforsourceswith V =0. Weadopteda2keVMeKaLmodeltocalculateX-ray fluxes,assumedN = 1.9 1021 cm 2 tocorrectforabsorp- H − × tion, and used V = V A = V 1.05. We caution that 0 V − − formostsources,N isunknown;iftheadoptedN islower H H (higher) than the actual N , the flux ratio is overestimated H (underestimated). Like E , thefluxratioismostlyusefulto 50 distinguish between coronal and accretion-poweredsources. Theformertypicallyhavelog(F /F ) . 1,withthemost X V u − active late-type dwarfsreaching valuesof about−0.5, while Figure5. The median energies E50 and X-ray–to–opticalflux the latter have log(FX/FV)u & −1 (Stockeetal. 1991). In- ratioslog(FX/FV)u showa trendoflowerflux ratiosforsoft deed, foroursources, theaveragefluxratio islower forsoft sourcesthanfor hardersources. For sourceswithoutoptical (E50 .1.5keV)thanforhard(E50 &1.5keV)sources(Figure counterparts,thelowerlimitonlog(FX/FV)u(shownasopen 5).AnopticalsourceinsidetheX-rayerrorcircleisnotneces- circles)was calculatedforthe detectionlimitV = 23.5. For sarilythetruecounterpart,butcanbeaspuriousmatch.Find- sourceswithmultiplecounterparts,therangeonlog(F /F ) X V u ing a relatively hard X-ray source with a low log(FX/FV)u is indicated with a vertical line. Error bars on the flux ratio value,cansignalsucharandomalignment. onlyincludestatisticalerrors,notanysystematicerrorsresult- For sources with candidate optical counterparts we also ing from uncertainties in the adopted X-ray spectral model. tookintoaccountthepositionofthesematchesintheCMD.In CX150 is an outlier and may be spuriously matched to the mostcases,thisworksreasonablywelltoseparateABsfrom starinitserrorcircle. Onlysourceswithσ <1areplotted. E50 AGNs (which can lie far off the cluster sequence) and CVs (which typically are blue). The position in the CMD doesa butwitherrorsonE thataretoolarge(&1keV)tomeaning- poorjobinseparatingclusterstarsfromfore-orbackground 50 fullyconstraintheirX-rayspectra;thesesourceswereclassi- stars. As can be seen in Figure 3, and also in the CMDs in fiedas“AB?”. CX25isanuncertainABbecauseitsposition Gozzolietal.(1996),theclusterstarsdonotclearlystandout. inthequantilediagramsuggestsanN thatisenhancedwith The lack of membershipinformationfor stars in the field of H respecttotheGalacticcolumn,whichisnotexpectedfortypi- Cr261limitstheclassificationofourChandrasources,aswe calABs. Finally,CX41andCX57haveE =2.2 0.5 keV discussbelow.Inthefollowing,X-rayfluxesandluminosities 50 ± and 2.5 0.5 keV, respectively; this is on the high side for refertothe0.3–7keVband. ± ABs,butgiventhelargeerrors,wealsoputthesetwosources 4.1. ActiveBinariesandCandidateActiveBinaries in the “AB?” category. With B V = 1.34 at V = 20.15, − CX41 is relatively blue, but not as offset from the main se- ForidentifyingpossibleABsinCr261,weselectedsources quenceasthesourcesdiscussedinSect.4.2;however,itisnot with candidate optical counterpartsthat lie along the cluster inconceivablethatthissourceisanAGNoraCV.Weexpect main sequence or sub-giant branch; if a source has multi- thatasignificantnumberof“AB”and“AB?”sourcesarefore- plematchesthatallsatisfythiscondition,itisalsoclassified andbackgroundactivestarsorbinaries. as a (candidate) AB. We allowed for the possible contribu- Ten ABs are matched to Mazur variables. CX27/V42, tiontothelightbyabinarycompanion,andforuncertainties CX89/V11, CX97/V38, and CX138/V21 are (semi- in the reddening, as indicated by the pairs of black and red )detached eclipsing binaries. The first three have periods of isochronesinFigure3. Theuncertaintyintheclusterdistance 1.3d orshorter,while thelightcurveofV21showseclipse- ( 350pc,basedontherangeofdistancesreportedintheliter- ∼ like eventswith an unconstrainedperiod. Themaximumor- ature),andinourabsolutephotometriccalibration(Sect.3.2) bitalperiodthatcanbetidallycircularisedin 7Gyr(i.e.the arealsosourcesofsystematicerror(seetheerrorbarinthetop ∼ age Cr261) is 15 d (Mathieuetal. 2004). Since the time rightofFigure3).Therefore,weclassifiedcandidatematches ∼ scale for tidal synchronisation is shorter than that for cir- that are only a little bit off the main sequence or sub-giant cularisation (Hut 1981; Zahn 1989), it is perfectly plausi- branchasABs,too. blethatatleastV42,V11,andV38containrapidlyrotating, Atotalof33Chandrasourcessatisfythephotometriccrite- and therefore X-ray–active, stars. CX93/V30, CX126/V13, riaoutlinedaboveandhaveE valueswithin1σof1.5 keV 50 CX133/V108, CX144/V24, CX147/V25, and CX149/V33 or lower; all have log(F /F ) . 1.4. We classified them X V u − as “AB” in Table2. Fouradditionalsources(CX62, CX80, 8CX133ismatchedwithtwostarsonthemainsequence;V10isthemore CX115, CX125) have similar optical and X-ray properties, likelycounterpartofthetwo. 10 arecontactbinariesoftheWUMatype. ForWUMa’s,adis- andX-rayproperties. However,since veryfewCVs inopen tanceconstraintcanbederivedfromtheknowncalibrationof clusters have been found, it is worthwhile to highlight any theabsolutemagnitudesintermsoforbitalperiodandB V candidates. Follow-up optical spectroscopy can confirm or − orV I colour(seee.g.Rucinski&Duerbeck1997). Mazur disprovewhetherasourceisaCVnornot. − et al. thus found that the distances to V30, V13, V25, V33, 4.3. CandidateBlueStragglers,YellowStragglers,and andlikelyV24,arecompatiblewiththatofCr261;theseare Sub-subgiants themostreliableclusterABsinoursample.Mazuretal.were inconclusiveregardingthedistancetoV10. Some of the brightest X-ray sources in old open clusters ThetimespanoftheWFIobservationsis 0.75h,withthe are membersthat lie off the main locus of the cluster in the ∼ B data taken first. For CX149/V33, with a period of 6.96 h CMD. These systems challengeour understandingof binary andalarge-amplitude( 0.8mag)lightcurve,thisspans 0.1 evolution,and,insomecases,wedonotunderstandwhythey ∼ ∼ in orbital phase. If our observations happened to be timed emit X-rays (vandenBergetal. 1999). Therefore, they de- around eclipse ingress (something we cannotcheck because servespecialattention. theephemerisisnotknownwithsufficientprecision),thiscan Blue stragglerstars (BSSs) are bluerandbrighterthanthe explain why we find a much bluer colour (B V = 0.72) main-sequenceturnoff(MSTO)ofacoevalpopulation.Their − thanMazuretal., whoreport B V = 1.09,i.e.rightonthe formationscenariosmustexplainhowthesestarsmanagedto − mainsequence.Wefoundsimilarcolourdifferencesforafew continuecore hydrogenburningfor a longer time than clus- other variables. Mazur et al. adopted a method that makes ter stars of similar mass. Mass transfer in a binary, direct their colours much less sensitive to non-simultaneous mea- collisions, and the merger of the close inner binary in a hi- surements. Therefore, in Figure 3, we plotted the variables erarchical triple driven by the Kozai-Lidov mechanism, are withtheirMazurphotometry(seeTable3),ifavailable. the three proposed formation channels (Davies 2015). For mostBSSs,itisnotclearwhich(ifany)ofthesechannelsap- 4.2. CandidateCataclysmicVariablesorAGN plies. The detection of X-rays in a bona-fide cluster BSS is Our mass estimate for Cr261 (5800–7200 M ) is similar asignofongoingbinaryinteractionandthusprovidesaclue ⊙ to the mass of NGC6791 (5000–7000 M ). If CVs in open tothecurrentsystemconfiguration. Thereisnostrictbright- ⊙ clusters are primordial, Cr261 would host a similar number nesslimitwithrespecttotheMSTOthatwecanusetoselect of CVs as NGC6791, i.e. 3 to 4 (vandenBerg 2013). CVs candidateBSSsinCr261.InM67,whichhasoneofthebest- typicallylietotheblueofthemainsequenceduetothelight studiedBSSspopulations,thebrightestBSS(F81; Leonard from the accretion disk, and possibly from the white dwarf 1996)lies 2.7magabovetheMSTOintheV band.Wetake ∼ (although in the B band, the blue excess is not always that the equivalent location in the CMD of Cr261, i.e. V 14, ≈ obvious, see e.g. Bassaetal. 2008). For eleven sources, the as a (somewhat arbitrary) limit, and consider brighter stars candidate optical counterpart(s) are blue with respect to the to be non-members. We thusfind eightmatcheswith candi- main sequence, and ten of them are possibly CVs: CX4, date BSSs: CX18/V45, CX67, CX73, CX74, CX86/V12, CX8, CX20, CX22, CX24, CX38, CX44, CX54, CX70, CX130, CX132, andCX136/V22. Exceptfor CX67, these andCX1519. Theotherbluesource,CX120,isnotamem- sources are soft (E . 1.4 keV) and all have log(F /F ) 50 X V u berofthecluster(seeSect.4.4). Weconsiderasourcetobe between –3.9 and –2.8. This is consistent with the proper- blueif it lies to theleftof the isochronethatis reddenedfor ties of ABs, and their X-rays are therefore likely the result thelowestpossibleclusterreddening. Inaddition,werequire of magnetic activity. Indeed, three sources are matched to thebluewardoffsetfromthisisochronetobeatleast0.13mag, (semi-)detachedeclipsing binaries with periods between 1.1 i.e.theerrorsontheabsolutephotometriccalibrationinVand and 2.1 d. V12 and V22 show Algol-typelightcurves. The Baddedinquadrature(seeSect.3.2). ideaofapossiblelinkbetweenAlgolsandBSSswasalready ThetensourceslistedabovehaveE valuesbetween 1.5 putforthbyMcCrea(1964). InanAlgolbinary,theoriginally 50 ∼ and2.8keV,log(F /F ) between–2.5and+1.0,andL be- lessmassivestarisnowobservedtobethemoremassiveone X V u X tween5 1029and3 1031ergs 1. Thisisconsistentwitha asaresultofthemassitreceivedfromitsRoche-lobefilling − × × CV classification, althoughlog(F /F ) 2.5forCX151 companion—herewemaybeseeingaBSS“inthemaking”. X V u ≈ − isonthelowsideforaCV;thismaysuggestadifferentsource It would therefore be particularly interesting to determine if class or the presenceofa randominterloperin the X-ray er- V12andV22areclustermembers. ror circle. We label these sources as candidateCVs (“CV?” ThecandidatecounterpartsofCX9andCX123liebetween in Table 2). Confusionwith other classes in this partof the theBSSsandredgiants.StarsinthisregionoftheCMDhave CMDismainlywithAGNs,whichoutnumbertheCVsinthe beendubbedyellowstragglersandmaybeBSSdescendants. field observed (Sect. 3.5) and can have similar blue colours All yellow stragglersin M67 are solid cluster membersand X-ray sources (Bellonietal. 1998). Their X-ray properties 9TheopticalmatchtoCX22,andoneofthematchestoCX24,areonly detectedinB,butthedetectionlimitinVimpliestheymustbeblue(B V. point at the presence of magnetic activity, and the same ap- 0.7). − pearstobethecaseforCX9andCX123.