MNRAS000,1–12(2017) Preprint23January2017 CompiledusingMNRASLATEXstylefilev3.0 Modeling Luminous-Blue-Variable Isolation 1⋆ 1 2 Mojgan Aghakhanlootakanloo, Jeremiah W. Murphy † and Nathan Smith ‡ 1Physics,FloridaStateUniversity,77ChieftanWay,Tallahassee,FL32306,USA 2StewardObservatory,933N.CherryAve,Tucson,AZ85719,USA 7 23January2017 1 0 2 ABSTRACT Observationsshowthatluminousbluevariables(LBVs)arefarmoredispersedthanmassive n O-type stars, and Smith&Tombleson (2015) suggested that these large separations are in- a J consistentwithasingle-starevolutionmodelofLBVs.Instead,theysuggestedthatthelarge distancesare mostconsistent with binaryevolutionscenarios. To test these suggestions,we 9 1 modeled young stellar clusters and their passive dissolution, and we find that, indeed, the standardsingle-starevolutionmodelisinconsistentwiththeobservedLBVenvironments.As ] single stars, the lifetimes of LBVs indicatedby their luminositieswouldbe far shorterthan R the ages required by their observed isolation. The resolution for this isolation is either that S LBVs are much less massive than their luminosity would suggest, or they are the result of . binary evolution.For the binaryscenarios, our crude models suggestthat LBVs are rejuve- h natedstars.Theyareeithertheresultofmergers,ortheyaremassgainersandreceivedakick p - whentheprimarystarexploded.Inthemergerscenario,LBVshavemoretimetodispersebe- o causetheyarethemergeroftwolessermassstars,inwhichtheprimaryhasanaveragemass r of about18 solar masses. In the mass-gainer and kick scenario, we find that LBV isolation t s isconsistentwitha wide rangeof kickvelocities,anywherefrom0 to∼200km/s.Ineither a scenario,binarityseemstoplayamajorroleintheisolationofLBVs. [ 1 Keywords: binaries:general–stars:evolution–stars:massive–stars:variables:general v 6 2 6 1 INTRODUCTION has been that most stars above 25-30 solar masses pass through 5 0 an LBV phase and experience eruptions as a means to transi- Initial mass is one of the primary characteristics that determines . tion from a hydrogen rich star to a H-poor Wolf-Rayet (WR) 1 astar’sevolutionandfate(Woosley&Heger2015);thereforeun- star (Humphreys&Davidson 1994). In this scenario, LBVs 0 derstanding mass-loss isimportant indeveloping acomplete the- experience high mass loss due to an unknown instability, which 7 ory of stellar evolution. Yet, understanding the physics and rela- may be driven by a high L/M ratio, near the Eddington limit 1 tiveimportanceofsteadyanderuptivemass-lossinthemostmas- (Humphreys&Davidson 1994; Vink 2012). On the other hand, v: sive stars remains a major challenge in stellar evolution theory. the LBV phenomenon might only be associated with rare cir- i There has been substantial progress in understanding mass-loss X cumstances if it is due to binary interaction (Smith&Tombleson via steady line-driven winds of hot stars (Kudritzki&Puls2000; 2015). r Pulsetal. 2008), and this effect is included in stellar evolution a Smith&Tombleson (2015) present new data that suggests models (Vinketal. 2001; Woosleyetal. 2002; Meynet&Maeder the latter. If LBVs mark a brief transitional phase at the end of 2005; Martins&Palacios 2013). However, the mass-loss rates of the main-sequence and before core-He burning WR stars, then redsupergiantsandtheroleoferuptivemasslossremainunclear, they should be found near other massive O-type stars. However, and the influence on stellar evolution remains uncertain (Smith Smith&Tombleson (2015) found that LBVs are quite isolated 2014;Smith&Owocki2006).TheLuminousbluevariable(LBV) fromO-typestars,andevenfartherawayfromOstarsthantheWR phenomenonisonesuchpoorlyconstrainedclassoferuptivestars. stars are. Given their isolation, Smith&Tombleson (2015) con- LBVs are luminous, unstable massive stars that suf- cludethattheLBVphenomenon isinconsistentwithasingle-star fer irregular variability and major mass-loss eruptions scenario and ismost consistent with binary scenarios. Inthis pa- (Humphreys&Davidson 1994). The mechanism of these per, we develop simple models for the dispersal of massive stars eruptions and the demographics of which stars experience these onthesky,inordertobetterconstrainwhethertheO-starandLBV is poorly constrained (Smith 2011, 2014). The traditional view distributionsrequirebinaryscenarios. Part of the reason that LBVs are poorly understood is be- cause there are very few examples. There are only 10 unob- ⋆ [email protected] † [email protected] scured in our galaxy and 19 known in the nearest galaxies, the ‡ [email protected] LMCandSMC(Smith&Tombleson2015).Eventhissmallsam- (cid:13)c 2017TheAuthors 2 Aghakhanlootakanlooet al. ple includes “candidate” LBVs (see below). Classifying vari- O stars, LBVsare isolatedin the Milky Wayand the Magellanic ous stars as LBVs or candidates can be somewhat controversial Clouds. Moreover, they find that LBVs appear to have a much (Humphreys&Davidson 1994; Weis 2003; Vink 2012); here we largerseparationthanevenWRstars,whicharethoughttobethe summarize their basic characteristics. LBVs are luminous, blue descendantsofLBVs.Theyconcludethattheirspatialdistribution massive stars with irregular or eruptive photometric variability. isinconsistentwithsingle-starevolutionandmostconsistentwith Stars that resemble LBVs in their physical properties and spec- binary models. Theysuggested a binary scenario in which LBVs tra,butlackthetell-talevariability,areusuallycalled“LBVcandi- arethekickedmassgainersormergerproducts. dates”.Thereasontheyaresometimesgroupedtogetherisbecause Humphreysetal.(2016)putforthadifferentinterpretationof itissuspectedthattheLBVinstabilitymaybeintermittent,sothat LBVlocations,suggestingthattheydonotruleoutthesingle-star candidatesaretemporarilydormantLBVsSmith(2011,2014). scenario.TheynotethatthesampleinSmith&Tombleson(2015) AlthoughthesignatureeruptivevariabilityofLBVswasiden- isamixtureoflessluminousLBVs,moreluminousclassicalLBVs, tified long ago, the physical theory of LBV eruptions is not yet andunconfirmedLBVs,andtheyproposethatseparatingthemal- clear.Forthemostpart,LBVsseemtoexperiencetwoclassesof leviates the conflict with single star models. From their point of eruptions; S Doradus (or S Dor) eruptions (1-2 mag) and Giant view, the single-star hypothesis still works because (1) the three eruptions(>2mag). most luminous stars of the sample that are classical LBVs (with SDoradusvariablestaketheirnamesakefromtheprototypi- initial masses greater than 50M ) have a distribution similar to ⊙ calLBVSDoradus(Hubble&Sandage1953).DuringSDorout- lateO-typestars,and(2)thelessluminousLBVs(withinitialmass bursts, LBVs make transitions in the HR Diagram (HRD) from ∼25−40M )arenotassociatedwithanyOstars,buthaveadistri- ⊙ theirnormal,hotquiescentstatetolowertemperatures(goingfrom butionsimilartoredsupergiants(RSGs),whichHumphreysetal. blue to red). In its quiescent state, an LBV has the spectrum of (2016)suggestisconsistentwiththembeingsingle-starsonapost- a B-type supergiant or a late Of-type/WN star (Walborn 1977; RSGphase.Humphreysetal.(2016)stronglysuggestedtodistin- Bohannan&Walborn 1989). In this state, LBVs are fainter (at guishbetweenthesetwotypesofLBVsinstatisticaltests.More- visual wavelengths) and blue with temperatures in the range of over,Humphreysetal.(2016)criticizedthatfiveoftheLMCstars 12,000to30,000K(Humphreys&Davidson1994).Intheirmax- (R81,R126,R84,Sk-69271andR99)areneitherLBVsnorcandi- imumvisiblestate,theirspectrumresemblesanF-typesupergiant dates. witharelativelyconstanttemperatureof∼ 8000K.SDorevents However, Smith (2016) showed that even using the LBVs were originally proposed to occur at constant bolometric lumi- sample subdivided as Humphreysetal. (2016) prefer does not nosity (Humphreys&Davidson 1994). So a change in tempera- changetheresultthatLBVsaretooisolatedforsinglestarevolution tureimpliesachangeinthephotosphericradius,L = 4πσR2T4. (overlookingthelackofstatisticalsignificance).Themostmassive Humphreys&Davidson (1994) suggested that the eruption is so LBVsappear tobeassociatedontheskywithlateO-typedwarfs optically thick that a pseudo photosphere forms in the wind or (point1above),which,however,haveinitialmasseslessthanhalf eruption.However,quantitativeestimatesofmass-lossratesshow ofthepresumedinitialmassesoftheclassicalLBVs.Similarly,the that theyare toolow toform alargeenough pseudo photosphere lower-luminosity LBVs have a similar distribution to RSGs, but (deKoteretal.1996;Grohetal.2009).Similarstudiesalsoimply theseRSGsaredominatedbystarsof10-15M (Smith2016). ⊙ that the bolometric luminosityisnot strictlyconstant (Grohetal. Humphreysetal. (2016) also stated that the observed LBV 2009). Instead, it has been suggested that the observed radius velocities seems to be too small to be consistent with the kicked changeofthephotospherecanbeapulsationorenvelopeinflation mass-gainerscenario,butSmith(2016)pointedoutthatwithouta drivenbytheFeopacitybump(Gra¨feneretal.2012). quantitativemodelforthevelocitydistributions,itwouldbediffi- Theotherdistinguishingtypeofvariabilityisintheformofgi- culttoruleanythinginorout.Inthispaper,wewillshowbothhigh anteruptionslikethe19thcenturyeruptionofηCar(Smith2011). andlowluminositiesLBVandLBVcandidateshavelargersepara- ThebasicdifferencefromSDoreventsisthatgianteruptionsshow tionsthanonewouldexpect,andin§4.4andFigure10weshow astrongincreaseinthebolometricluminosityandaremajorerup- that a wide range of kick velocities are consistent with the large tive mass loss events, whereas S Dor eruptions occur at roughly separations. constantluminosityandarenotmajormass-lossevents.Themass The main goal of this paper is to quantitatively constrain lossrateatSDormaximumisoforder10−4M yr−1orless(Wolf whethertherelativeisolationofLBVsisinconsistentwithasingle- ⊙ 1989; Grohetal. 2009). On the other hand, giant eruption mass star evolution model. We begin by reproducing and verifying lossrateistheorder of 10−1 to1M yr−1 (Owockietal.2004; Smith&Tombleson (2015) results (§ 2). In § 3, we introduce a ⊙ Smith&Owocki 2006; Smith 2014). It is unlikely that a normal simplemodel for young stellarclustersand their passive dissolu- line-driven stellar wind is responsible for the giant eruptions be- tion.Totestthismodel,wealsocomparetheseparationsbetween causethematerialishighlydenseandopticallythick(Owockietal. Ostarsforthemodelandobservations;wefindthatthemodelre- 2004; Smith&Owocki 2006). Instead, giant eruptions must be producessomegeneralpropertiesofthespatialdistributionofmas- continuum-drivensuper-Eddingtonwindsorhydrodynamicexplo- sivestars,butitislackinginotherways.Sinceweconstructedthe sions(Smith&Owocki2006).Bothoftheselackanexplanationof simplestmodelpossible,thisimpliesthatwemayimprovethedis- theunderlyingtrigger;thesuper-Eddingtonwindreliesuponanun- persion model and learn even more about the evolution of mas- explainedincreaseinthestar’sbolometricluminosity,whereasthe sive stars. Then in § 4, we present the primary consideration of explosivenatureofgianteruptionswouldrequiresignificantenergy thispaper;wecomparesingle-starevolutionandbinaryevolution deposition. There is much additional discussion about the nature inthecontextofclusterdissolution,andwefindthatthesingle-star of LBV giant eruptions in the literature (Humphreys&Davidson evolutionscenarioisinconsistentforinitialmassesappropriatefor 1994; Owockietal. 2004; Smith 2011, 2014; Smith&Owocki LBVluminosities.Wediscusstwobinaryevolutionchannelsthat 2006). are consistent with the relative isolation of LBVs. We then sum- Smith&Tombleson(2015)highlightaresultthatchangesthe marize,andwediscussfutureobservationstofurtherconstrainthe emphasis on the most likely models. They find that compared to binarymodels(§5). MNRAS000,1–12(2017) ModelingLBV Isolation 3 2 OBSERVATIONS Table 1. List of LBVs and LBV candidates adapted from Smith&Tombleson (2015). For the stars in the SMC, we rescale Inthefollowingsections,weexplorewhichtheoreticalmodelsare theirangularseparationby1.2asiftheyarelocatedatthedistanceofthe mostconsistentwiththedata,butbeforethat,weclearlydefine,an- LMC.ParenthesesinthenamerepresentLBVcandidatesandparentheses alyze,andcharacterizethedatainthissection.First,wereproduce in mass column specify the LBVs with relatively poorly constrained and verify the results of Smith&Tombleson (2015). Second, we luminosityandmass. furthercharacterizethedata,notingthatthedistributionsofnearest neighborsarelognormal.Sincelog-normaldistributionshavevery LBV(name) Galaxy(name) S(deg) Meff(M ) fewparameters,thisrestrictsthecomplexityandparametersofour ⊙ R143 LMC 0.00519 60 modelsin§3. R127 LMC 0.00475 90 Smith&Tombleson(2015)foundthatLBVsaremuchmore SDor LMC 0.0138 55 isolatedthanO-typeorWRstars,suggestingthatLBVsarenotan R81 LMC 0.1236 (40) intermediarystagebetweenthesetwoevolutionarystages.Inpar- R110 LMC 0.2805 30 ticular,theyfoundthatonaverage,thedistancefromLBVstothe R71 LMC 0.4448 29 nearestOstarisquitelarge(0.05deg).Forcomparison,theaverage MWC112 LMC 0.0892 (60) distancefromearlyOstarstothenearestOstaris0.002degrees, R85 LMC 0.0252 28 andfrommidandlateOstarsare0.008and0.010degreesrespec- (R84) LMC 0.1575 30 tively. If single early- and mid-type O stars are indeed the main (R99) LMC 0.0412 30 (R126) LMC 0.0358 (40) sequence progenitorsofLBVs,thenonewouldexpect thespatial (S61) LMC 0.1432 90 separations betweenLBVsandother O starstobenot toodiffer- (S119) LMC 0.3467 50 entthantheseparationbetweenearly-andmid-typeOstars.How- (Sk-69142a) LMC 0.0522 60 ever, the LBVseparations arean order of magnitude farther than (Sk-69279) LMC 0.0685 52 theearly- andmid-type separations. In fact,the LBVseparations (Sk-69271) LMC 0.040 50 are5timeslargerthaneventhelate-typeOstars,whichlivelonger HD5980 SMC 0.0191 150 andcaninprinciplemigratefarther. R40 SMC 0.1112 32 Smith&Tombleson (2015) quantified the difference in the (R4) SMC 0.0160 (30) distributions of separations by using the Kolmogorov-Smirnov (KS)test.ComparingthedistributionsofLBVstoearly-,mid-and late-type give P-values of 5.5e-9, 1.4e-4 and 4.4e-6 respectively. These values imply that O-stars and LBV distributions are quite different. If true, then these results have profound consequences andinthethird,theLBVdistributiononlycontainsthecandidates. forourunderstandingofLBVsandtheirplaceinmassivestarevo- When we include both confirmed and candidate LBVs, the LBV lution. In fact, Smith&Tombleson (2015) suggest that the most andO-stardistributionsareclearlynotdrawnfromthesamepar- natural explanation is that LBVsare the result of extreme binary ent distribution. However, omitting LBV candidates reduces the encounters.Laterwewilltestthisassertion,butfornowwerepro- distinctionsbetweenthedistributions.Onemightarguethatsince duceandverifytheirresults. LBVsrepresentalaterevolutionarystage,thenthespatialsepara- To verify the results of Smith&Tombleson (2015), we first tionsshouldbelarger,andtherefore,thedistributionsofearly-type definethedata.Thedataconsistsoftwomainparts;LBVsandO OstarsandLBVsshouldnotrepresentthesamedistribution.How- stars.TheirsampleincludesWRstars,sgB[e]stars,andRSGstoo, ever,wewillshowinsection4thatthelifetimesofmassivestarsare butwedonotdiscussthemherebecauseatthemoment,wewantto fartooshorttoexplaintheselargediscrepancies.Inourinitialas- keepourmodelsin§3&4simpleandwewillfocusjustonLBVs sessment,weagreewithSmith&Tombleson(2015),thelargesep- andOstars.TheirLBVsamplesinclude16starsintheLMC,and arations present a challenge to the single-star evolution scenario. three stars in the SMC. They did not consider Milky Way LBVs Inthenextsections,wewillpresenttheoreticalmodelstoquantify because it’s difficult to develop a homogeneous catalog. In their thisinconsistency. study,theyincludeLBVcandidateswithamassiveCSMshellthat Beforeweconstrainthemodels,notethattheseparationdis- likelyindicatesapreviousLBV-likegianteruption.LBVsandtheir tributions are log-normal (see Figure 1). In fact, this simple ob- importantparametersaresummarizedinTable1. servation greatly restricts the complexity of the models that we Smith&Tombleson (2015) gathered thepositions of O-type may explore in the next sections. If a variable, such as the sep- stars within 10◦ projected radius of 30 Dor from SIMBAD aration between stars shows a normal distribution, then there are database. They also used the revised Galactic O-star Catalog only two free parameters that describe the distribution, the mean (Ma´ızApella´nizetal. 2013) to check their O-star samples (not andthevariance.Inaddition,iftheseparationdependsuponother shownintheirpaper),butastheyclaimedthisdidnotchangetheir variablessuchasavelocitydistribution,thenthankstothecentral- overallresults.WecollectthesameO-starsamplesfromSIMBAD limittheorem,theseparationdistributionwillonlydependuponthe database. meanofthevarianceofthesecondaryvariablessuchastheveloc- After gathering the data, we find the distance from one star itydistribution.Thismeansthatwecannotproposeoverlycomplex tothenearest Ostar.Thebottompanel ofFigure3showsthere- modelsforthevelocitydistribution.Wewouldonlybeabletoinfer sultingcumulativedistributions;notethatthedistributionsforeach themeanand thevariance anyway. Fortunately, wemaymeasure typeofstarisquitedistinct.Infact,KStestsshowthatthesedis- theseparationfordifferenttypesofOstarsandotherevolutionary tributions very distinct. Our KS-test P values are listed in Table stages.Thismeansthatwemayinferthetemporalevolutioninad- 2.WeconsiderthreeKStests.Inone,wecomparetheseparation ditionthetothemeanandvariance.Whatevermodelswepropose, forbothconfirmedandcandidateLBVswithO-stardistribution.In theycannotbetooelaborate;wewillonlybeabletoinferthemean the second, the LBV distribution only includes confirmed LBVs, andvarianceofonequantityasafunctionoftime. MNRAS000,1–12(2017) 4 Aghakhanlootakanlooet al. Table2.Broadly,wereproducetheresultsofSmith&Tombleson(2015) whofoundthatthedistributionofseparationsbetweenLBVsandthenear- estOstarisquitedifferentthanthedistributionsfortheseparationsbetween OstarsandthenearestOstar.ThesecondrowshowstheresultsofourKS 5 tests between the LBV separations and the early-, mid-,and late-type O stars.OurresultsaresimilartothoseofSmith&Tombleson(2015)(first p ∼ 0.5 O2-O5(Early) row).LikeSmith&Tombleson(2015)weobtainthepositions ofOstars 4 andtheirspectraltypesfromSimbad.Smith&Tombleson(2015)updated thespectral types withtheGalactic O-starCatalog (Ma´ızApella´nizetal. 2013);however,wedidnot.Thisslightdifferenceinspectraltypingiswhat 3 causesthemodestdifferenceinPvalues.Ineithercase,theLBVseparations areinconsistentwithanyO-typeseparations.IfweexcludetheLBVcan- 2 didates(thirdrow),theconclusionsremainthesame,butthesignificanceis greatlyreduced. 1 DataSet EarlyO MidO LateO LBV+LBVc(Smith&Tombleson2015) 5.5e-9 1.4e-4 6.4e-06 LBV+LBVc(thiswork) 8.2e-08 2.8e-05 8.4e-05 LBV(thiswork) 9.2e-04 2.3e-02 5.7e-02 10 p ∼ 0.5 O6+O7(Mid) LBVc(thiswork) 6.4e-06 5.1e-05 2e-04 8 y c6 3 AGENERICMODELFORTHESPATIAL n DISTRIBUTIONOFTHESTARSINAPASSIVE ue4 DISPERSALCLUSTER q e r InordertomodeltherelativeisolationofLBVs,weneedtomodel F2 the dissolution of clusters and associations of massive stars. For severalreasons,wemodelthedissolutionofyoungstellarclusters withaminimumsetofparameters.Forone,theO-stardistributions p ∼ 0.4 O8+O9(Late) arelognormal.Therefore,thereareonlyafewparametersthatde- 20 scribethedatawhichonemayfit.Theonlydatathatwecanreliably fitarethemean,variance,andtimeevolutionoftheseparations.So 15 whatevermodelswedevelop,theyshouldnotbeoverlycomplex. Also, as far as we know, there are no simple self-consistent and testedmodelsforthedissolutionofclusters.Therefore,wepropose 10 asimplemodelofclusterdissolutionandadaptittoconsidertwo scenarios:clusterdissolutioninthecontextofsingle-starevolution, 5 and clusterdissolution withclosebinary interactions. Inthissec- tion,wepresentaclusterdispersalmodelconsideringonlysingle- 0 star evolution. Whileour dissolution models represent thespatial 10−4 10−3 10−2 10−1 100 distributionsreasonably wellincertainrespects, wenotethat our modelfailstomatchthedatainotherways.Thisimpliesthatour S [deg] modelismissingsomething.Inotherwords,wemaybeabletoin- fermorephysicsaboutthedissolutionofclustersfromthesimple spatialdistributionofOstars.Inthenextsection,wecontrastthe Figure1.Normality test. Thedistributions ofseparations forearly, mid, single-starmodelwithamodelthatconsidersbinarity. andlateOstarsareconsistentwithalognormaldistribution.Ineachplot, Ourmaingoalistointroduceamodelforyoungstellarclus- weshowtheprobability,p,thattheparentdistributionisalognormaldis- terswhichpredictsthespatial distributionof massivestars, espe- tribution. ciallyO stars.Westartbyconsidering thesimplest model. Inthe following,wemodeltheaveragedistancetothenearestOstarby nothingmorethanthepassivedispersalofacluster. willdriftawaywithaspeedroughlyequaltothevelocitydispersion Beforewediveintothedetailsofthemodel,itisworthchar- whentheclusterwasbound.Hence,vd ∼σv.Allthatislefttodois acterizingthescalesof atypical cluster.Webeginrightafter star estimateMT andR.AtypicalclusterhasR∼4pcandabout40O formation ends and consider a system of gas and stars that is in stars;ifonly∼1%ofthegasingiantmolecularcloudsformstars virialequilibrium.Inthiscase,wehave2T +U = 0,whereTis (Krumholz&Tan2007),thetotalmassofthemolecularcloud,MT thetotalthermalpluskineticenergy,andUisthegravitationalpo- isoforder2×105 M .Giventheseapproximations,weestimate tRenitsiabloeunnedrg,ya.nIdnitthiaellsyta,rtshehasvyesteamvewloicthitytodtaislpmerassisonMtThaatnsdcarlaedsiuass thattNheexdtr,ifwteveplroecsietyntis⊙aomfoorredesprevcdifi∼cd1i3s.s5o(lu2t×io1M0n5TmMo⊙de4Rplcto)12co.nvert 1 thegravitationalpotentialoftheentiresystemσv ∼(GMT/R)2. thisdispersalvelocityintoadistributionofseparationsasafunction Thenthesystemlosesgasmassbysomeformofstellarfeedback oftime.Ratherthanusingthisestimateforthedispersalvelocity, (UV radiation, stellar winds, etc.) and likely makes the stars un- wewillusethedataandourmodeltoinferthedispersalvelocities. bound. If the system looses all of the gas quickly, then the stars WeproposeaMonte-Carlomodelforthedissolutionoftheclusters. MNRAS000,1–12(2017) ModelingLBV Isolation 5 First,werandomlysampleN clustersuniformlyintimebetween cl 0Myrsand6Myrs.Foreachcluster,wedrawaclustermassfroma 80 60 1.0Myrs distributionofclustermasses.Then,weestimatethetotalnumber 40 hSi ∼ vdt of the stars (N∗), and for each cluster, we draw a distribution of 200 [NO(t)]1/2 stellarmasses(M )fromtheSalpeterdistribution. −20 First, we ran∗domly select a total number of O stars , N , −40 ∗ −60 e1foov9fre9er7e,a)acw,chehddMcNaclrcuclellusmtse∝teorrs.MitsTinct−hhlt2eeer,edSwsictshehtedrericbeihnuMttteihorcenlfnuifsunrmocthtmbieoenmrwoa(hEsfiscOlmhofsewtgtaherreesedfcnrolaur&wsetaeEtcrhf.hreHecmsoluiwozsve-- y[pc]−8246800000−803.−260M−y4r0s−20 0 20 40 60 80 24680000 6.3Myrs 0 0 ter,soourfirstorderofbusinessistoexpresstheSchechterfunc- −20 −20 tionintermsofthenumber ofOstars.Themassoftheclusteris −40 −40 Mandclm=axAimuMMm∗∗12mMas∗s−O1.3s5ta,rwthhaetrewMec∗1onasniddeMr. ∗In2taerremthseofmtihniism,uthme −−8600−80−60−40−20 0 20 40 60 80−−8600−80−60−40−20 0 20 40 60 80 totalnumbRerofOstarsbecomes x[pc] Figure 2. We Monte-Carlo model the separations between O stars and N = Mcl M∗2M −2.35dM . (1) LBVs by considering a random sample of dissolving clusters at random ∗ MM∗∗12M∗−1.35dM∗ ZM∗1 ∗ ∗ awgieths.iHtseorwenweagseh.oNwottheethOatsttahresaovfetrhargeeesreapnadroamtiolyngbeentwereaetendthceluOstesrtsa,resaicnh- R creaseswithagefortworeasons.First,theseparationsincreaseastheclus- Therefore,thetotalnumberofstarsintheclusterisproportionalto terdisperseswithadriftvelocityv overtimet.Second,Ostarsdisappear d themassofthecluster(N ∝Mcl),andwecaneasilytranslatethe astheyevolve. ∗ distributioninmasstoadistributioninthenumberofstarsforeach bcleutwsteere,ndd0NNca∗lnd∝1N,t∗h−e2n.tIhfeRt∗otiasldnruamwbnefrroomfstthaersuinniftohremclduisstterribiustion 3.1 COMPARINGTHEPASSIVESINGLE-STAR DISSOLUTIONMODELWITHTHEDATA 1 N = , (2) ∗ R (N max−1−N min−1)+N min−1 Next, we compare the passive dissolution of single stars to the ∗ ∗ ∗ ∗ LMCandSMCnearest-neighbordistributions.Figure3showsthe cumulative distribution for the separations for our simple disso- whereN andN arethemaximumandminimumnumber ∗max ∗min lutionmodel (toppanel) andfor theobservations (bottompanel). ofthestarsinthecluster. Forillustrationpurposes,wesetv to14.5km/s,makingthemod- d Foreachstar,wedrawthemassfromtheSalpeterInitialMass eleddistributionhaveaboutthesamemeanasthedata.Sofar,our Function; passivedissolutionmodelisingoodagreementwithobservations. Both the model and observations show a log-normal distribution M =( 1 )0.74 inseparations,andtheaverageseparationincreaseswithspectral- ∗ [Rm(Mmax−1.35−Mmin−1.35)]+Mmin−1.35 type,whichisexpectedsincelaterOstarsstarslivelongerandhave (3) moretimetodisperse. ,whereRmisrandomnumberbetween0and1. TheMean and Std.deviation of theseparations aretheonly Havingestablishedtheinitialconditions,wenowdescribethe informationwecangleanfromaGaussiandistribution,sonext,we evolution. The average separation between O stars depends upon investigatetheirbehaviors.Theprimaryparameterinourmodelis howmuchtheclusterhasdispersedandhowmanyOstarsareleft. v ,soinFigure4weplotthemean(bottompanel)andstd.devia- d SoweneedtomodelthedispersionoftheOstarsandtheirdisap- tion(toppanel)asafunctionofv .Thedashedlinesrepresentthe d pearance.Forthelifetimes,weusethePadovastellarevolutionary modeledmeanandstd.deviation,andthesolidbandsindicatethe models (Bertellietal. 2009). Figure 2 shows the spatial distribu- observedvalues.Theverticalaxes(µ andσ )inFigure4,arethe S S tionofanexamplemodel.ThespatialdistributionisGaussianwith meanandstandarddeviationinthelog;µ(logS)andσ(logS)to σ = vdt. Given the assumption that stars are coasting then the poweroftenrespectively.Thesolidbandsprovidesomeestimate individualvelocitiesarer/t.Withtheseassumptions,thenthedis- ofuncertaintyinourinferreddriftvelocity,webootstraptheobser- tributionofvelocitiesisGaussiantoo;p(v)= √2πtσ2e−r2/2σ2 . vations,givingavarianceforboththemeanandstd.deviation. Withourmodeldefined,ourfirsttaskistoconstrainwhether We draw three main conclusions from Figure 4. For one, theaveragedistancesbetweenLBVsandOstarsisconsistentwith the drift velocities that we infer by comparing our simple model the passive dissolution of a cluster with single-star evolution. To withthedataareroughlywhatwewouldexpect;seeourorder-of- compare ourmodels tothedata,wecalculatetheangular separa- magnitudeestimatein§3.Second,weinferlargerdriftvelocities tions, assuming that the clusters are at the distance of the LMC. forthelatertypeOstars.Thismayhintattheimportanceofbinary Furthermore,tobeconsistentwithSmith&Tombleson(2015),we evolutionandkicksamongtheO-typestarsthemselves.Third,our subdividethemodeledOstarsintoearly,mid,andlatetypesbased simplemodelisnotabletoreproducethevarianceinthedistribu- upontheirmasses.Toconvertfrommasstospectraltype,weused tions. This implies that something ismissing from our model. In Martinsetal.(2005)data.Early-typeOstarshavemassesgreater otherwords,thereismorethatwecanlearnabouttheevolutionof than34.17 M , late-typeOstarshavemasses6 24.15 M , and massivestarsinclustersfromtheirspatialdistributions. ⊙ ⊙ mid-typeOstarshavemassesinbetween. Inthenextsubsection, Wecan use theresultsin FigureMeanVariance to alsoinfer wetestwhetherourpassivesingle-stardissolutionmodelisconsis- that LBVisolation putsinteresting constraints on their evolution. tentwiththedata. The average separation for late-type O stars is 0.01 degrees. For MNRAS000,1–12(2017) 6 Aghakhanlootakanlooet al. 1.0 10 O2-O5(Early) Model O6+O7(Mid) 9 0.8 O8+O9(Late) 8 0.6 7 S σ 6 0.4 5 4 0.2 al t to 3 of 5 10 15 20 25 n tio OBS 0.020 c a O2-O5(Early) OBS Fr0.8 O6+O7(Mid) Model O8+O9(Late) 0.015 0.6 VLate∼17-20km/s 0.4 µS 0.010 VMid∼23-25km/s 0.2 0.005 LLBBVV 0.0 0.000 VEarly∼10-13km/s 10−3 10−2 10−1 100 5 10 15 20 25 ProjectedseparationtonearestO-typestar[deg] Vdrift [km/s] Figure3.Cumulativedistributionsfortheprojectedseparationtothenear- Figure4.Themean(bottompanel)andstd.deviation(toppanel)distanceto estOstar.Thetoppanelrepresentsthemodeleddistributionandthebottom thenearestneighborvs.driftvelocity.Inbothpanels,dashedlinesrepresent panelrepresentsthedata.Broadly,themodelreproducestheobservations; the passive dissolution model and solid lines represent the observational bothshowalog-normaldistribution,andtheaverageseparationincreases data(Smith&Tombleson2015).Wehighlightthreemainconclusions.1) withspectral-typebecausethelaterspectraltypeslastlonger. Thedriftvelocitiesthatweinferbycomparingoursimplemodelwiththe dataareroughlywhatweestimatedin§3.2)Weinferlargerdriftvelocities for the later type O stars, implying that binary evolution and kicks may LBVs,theaverageseparationisroughlyfivetimesbigger.Dimen- be important. 3) The passive dissolution model is not able to reproduce sionally,theaverageseparationshouldbeproportional tothedis- thevarianceinthedistributions, whichimpliesmissingphysicsfromour model. In another words, there is room to improve our model and learn persionvelocityandtheage,S ∼ vdtage.IfanLBVcomesfrom moreabouttheinterplaybetweenO-starevolutionandclusterdissolution. the most massive stars, then one would not expect them to have ageslargerthanthelate-typeOstars.Therefore,asaconservative estimate, let us assume that an LBV is an evolved very massive thatitisindeedinconsistentwithobservations.Inaddition,wecon- star which has about the same age as a low mass main sequence sidermodelsthatinvolvebinaryinteractionsinadispersingclus- Ostar.Underthisassumption,sincetheseparationsforLBVsare ter.Ouraimistodevelopmodelstoseewhetherbinaryscenarios 5timesbigger thanlate-typeOstars,thisimpliesthatthedisper- areconsistentwiththeLBVobservedseparations.Atthemoment, salvelocityis5timesbiggerthanthelate-typeOstarwhichisof there is very littleinformation other than the separations, so it is order 100km/s.Tobemorequantitative,inthenext sections,we notworthdevelopinganoverlycomplexmodelforbinaryinterac- extendthepassivemodeltoinfertheactualdispersal velocityfor tion.Wewouldnotbeabletoconstraintheextraparametersofthe LBVs.Alternatively,weconsiderbinaryscenarioswhichmaygive model.Therefore,wedevelop thesimplestbinarymodelstocon- aexplanationfortherelativelylargeisolationforLBVs. strainthedata. Inparticular, weconsider twosimplemodels that involvebinaryevolutioninadispersingcluster.Inthefirstmodel, weconsiderthataLBVistheproductofamergerandisarejuve- natedstar;inthesecondmodel,weconsiderthataLBVisamass 4 CLUSTERDISSOLUTIONWITHCLOSEBINARY gainerandreceivesakickwhenitsprimarycompanionexplodes. INTERACTIONS In§4.1,wefirstputtogetherananalyticmodelfortheaverage Intheprevioussection,wesuggestedthatthesingle-stardispersal separationbetweentwostarsversustime.Thenin§4.2weusethis model is inconsistent withthe isolation of LBVs.In this section, modeltoshowtheinconsistencyinthestandardsingle-starmodel, weputthepassivesingle-stardispersalmodeltothetest,andshow andweshowthatLBVsareeitheroverluminousgiventheirmass, MNRAS000,1–12(2017) ModelingLBV Isolation 7 stars,whichmayhaveadifferentdensitydistributionthanthefirst. Obviously,later’L’willrepresentLBVs.Inthiscase,thecombined densityis σO =vOt 3 nOL(r)= 2NπσO2 e−r2/2σO2 + 2NπσL2e−r2/2σL2 . (7) c O L p / Withthistwo-component expression for the density, we can ars σL =vLt evaluate the local separation, Sˆ via eq. (5) and then we can cal- culatetheaverageseparationfromeq.(4).Calculatingtheaverage t s # separationisnumericallystraightforward.However,withasmall butuseful assumption, wecanderive ananalyticestimatefor the averageseparation.Tomakeiteasiertocalculatetheintegralana- lytically,wemaketwoassumptions.First,weassumethattheaver- r[pc] ageseparationisroughlygivenbythescaleofoneoverthesquare rootoftheaveragedensity.Therefore, 1 Figure5.Twosimplespatial-distribution modelsforthederivationofour hS i≈ . (8) analyticscalings. OL hnOLi1/2 Second,becauseLBVsareextraordinarilyrarecomparedtotheO stars we assume that N ≫ N . By considering these two as- ortheyaretheproductofamergerandarearejuvenatedstar.Al- O L sumptionstheaveragedensityis ternatively, in§4.3, weuse theanalytic model to develop a kick model, and in § 4.4, we use this model to infer a potential kick N hn i≈ O . (9) velocityforLBVs.Insummary, weillustratethattheisolationof OL 2π(σ2 +σ2) O L LBVsisconsistentwithbinaryscenarioandisinconsistentwiththe Onceweplugthisintotheequationforhn i,eq.(8)leadstothe single-starmodel. OL averagedistancefromLBVstothenearestOstar: Toconstrainthemodels, wefirstderiveanalyticscalingsfor theaverageseparations,andthenweexplorewhetherthesescalings 2π(σ2 +σ2) 1/2 areconsistentwithsimplebinarymodels.First,weconsidersimple hSOLi≈ NO(t) L (10) models for the spatial distribution of two groups of stars, type O (cid:18) O (cid:19) andtypeL.EachhasaGaussianspatialdistributionwithitsown SoonwewillusetheseparationbetweenOstarstohelpcon- velocity dispersion vd ,which we label as vO and vL. Later, we strainthemodels for LBVs,so wenow derive an analyticmodel will consider two scenarios; one in which these average velocity forhSOi.TheaveragedensityforOstarsis dispersionsarethesame,andoneinwhichtheyaredifferent.For N (t) avisualrepresentationofthesesimplemodels,seeFigure5.Given hnOi= 4πOσ2 , (11) thesedistributions,wecalculatetheaverageseparationbetweena O andsotheaverageseparationbetweenOstarsisroughly starandtheneareststarinthesamegroup.Thenwecalculatethe averageseparationbetweenastaringroupOtoastaringroupL. 1 4π 1/2 Theaverageseparationbetweenstarsinthesamepopulationis hSOi≈ hn i1/2 = N (t) σO. (12) O (cid:18) O (cid:19) hSi=2π ∞Sp(r)rdr, (4) Tomakeuseoftheseexpressionsfortheaverageseparation, Z0 we need to compare the separations between two different tracer whereSistheseparation,andp(r)istheprobabilitydensityfunc- populations. Because the masses of mid-type O stars correspond tionp(r) = 2π1σ2e−r2/2σ2 whereσ = vt.Tocalculatethemean roughlytoinferredminimummassofLBVs,weusethemid-type valueoftheseparation,weneedtofindtheseparation(S).Oneway Ostaraverageseparationasareference; toestimatethedistancetothenearestneighboristousethespatial hS i 2 1 σ2 N (t ) densityofstars.Ingeneral,anestimateforthedistancetothenear- OL = 1+ L O O , (13) estneighbor is,Sˆ ≈ 1/n1/d,wherenisthenumber densityand (cid:18) hSOi (cid:19) 2(cid:18) σO2 (cid:19) NO(tL) d are the number of dimensions that we consider (ZeljkoIvezic wherewearecarefultoconsiderhowthenumberofOstarschanges 2014).Whenviewingclustersprojectedontothesky,d=2. withtimeandweevaluatethisfunctionattheageoftheLBVpopu- lationandthereferenceOstarpopulation.Thisequationrepresents Sˆ≈1/n1/2. (5) thegeneralexpressionrelatingage,theaverageseparations,andthe If we consider a simple density distribution, n(r) = drift(orkick)velocityofeachtracerpopulation. N exp−r2/2σ2, then the average separation in two di- Intheexpressionsfortheaverageseparations,theseparations 2πσ2 mensionalspaceis growduetotwoeffects:adriftvelocityandthedeathofOstars. Thedrift part is simplyproportional tot. Next, weexplicitly de- 2π 1/2 hSˆi=2 σ, (6) rive the number of O stars as a function of time, NO(t).Given (cid:18)N (cid:19) a mass function dN/dM, the total number of O stars is NO = whereN isthetotalnumberofstarsinthecluster. MM12 ddMNdM = αA+1(M2−α+1−M1−α+1),whereM1isthemin- Nowweconsidertwopopulationsofstars.Onewerepresent imummassforan−Ostar(∼16M ),M2 isthemaximummassO R ⊙ with’O’,whichrepresentsthelargestnumber oftracerstars.As star,whichisafunctionoftheageofthecluster,αistheslope(we thelabelsuggests,wewilllaterconsiderOstarsasalargenumber useSalpeter,2.35),andAisanormalizationconstant.Ifweassume oftracer stars.Theother, ’L’,representsamoreraresetoftracer apower-lawrelationshipbetweenmassofanOstaranditslifetime MNRAS000,1–12(2017) 8 Aghakhanlootakanlooet al. as an O star, then we can relate the age of an M2 O star to the ItisclearthatmostoftheLBVshavelargerseparationscompared ageofanM1Ostar:M2 =M1(tt1)−1/β.FromthePadovastellar towhatthepassivemodelpredicted. evolutionarymodels(Bertellietal.2009),wefindthatthevalueof Moreover, in the passive model, LBVs do not have enough β ∼1.5.Combiningtheseexpressions,wegetanequationforthe timetogettotheobservedaverageseparation.Figure7showsthe numberofOstarsasafunctionoftheageofthecluster,t, samepassivemodel,theobservedLBVseparation,butthistimewe simplifythepossibleagesofLBVsbyshowingtheagesfortheav- NO(t)= 1−Aα(1−(tt1)αβ−1)M11−α. (14) esirnaggleemstaasrs,tohfeonutrheLyBdVosnaomthpalev.eCelneoaurlgyh,itfimLBeVtosreevaochlvtehaeslaarngoersmepa-l WithanexplicitfunctionforthenumberofOstars,wemay arations.Instead,let’sconsiderhowoldanLBVwouldhavetobe nowderivetheequationrelatingseparation, age,andvelocity,in- inordertopassivelydispersetotheobservedseparations.Figure7 cluding explicitly all of the dependence on time. Substitutingthe showsthattheagewouldneedtobeabout7.3Myrs.Yetthisage expressionforNO(t),eq.(14)intothegeneralanalyticexpression, correspondstothemainsequenceturnofftimeforan18M staror eq.13,wefinallyarriveatthegeneralanalyticformula,explicitly thedeathof a27M star.Bothof thesevalues arebelow⊙theav- relatingseparation,age,andvelocity: eragemassof theLB⊙Vs,50M (§2).It isclear that considering ⊙ LBVs in the context of a standard single-star evolution is incon- α−1 hS i 2 1 v x 2 (1−x β ) sistentwiththeisolation.Ifwestickwiththeassumptionthatthe OL = 1+ L O , (15) (cid:18) hSOi (cid:19) 2 (cid:18)vOxO(cid:19) !(1−xαβ−1) driftvelocitiesoftheLBVsandtheOstarsarethesame,thenthere aretwopossiblesolutions.EitherLBVsareoverluminousorthey wherexO = ttO1 andx = ttL1.t1 (8Myrs)istheageofthemin- aretheresultofamergerandarerejuvenatedstars.SeeFigure8to imum mass and t (2.5 Myrs) is a reference age. We estimate visualizethemergermodel. O thesevaluesfromPadovastellarevolutionarymodels(Bertellietal. 2009). In the next subsections, we use our general analytic result, 4.3 KICKMODEL eq. (15), to explore what average separations one would expect AnotherbinarymodelthatisconsistentwiththeisolationofLBVs when we consider the passive dissolution in three scenarios, a isthekickmodel.Tovisualizethekickmodel,considerthebinary single-star evolution scenario, a binary scenario that involves a scenarioinFigure9.Inthismodel,theprimarystar,themoremas- merger,orabinaryscenariothatinvolvesakick. sivestar,evolvesfirstandtransfersmasstothesecondarystar.Ifthe moremassivestarismassiveenoughtoexplodeasacore-collapse supernova,thenthecompanionmayreceiveakick.Thiskickmay 4.1 PASSIVEMODEL beimpartedbyeitheranasymmetricexplosion,theBlaauwmech- Usingouranalyticestimatesfortheaverageseparation,weassume anism(Blaauw1961),oracombinationofboth.Inthismanuscript, thatthedispersalvelocityforLBVsandOstarsarethesameand wedonotmodelthebinaryevolutionandkickvelocities.Ratherwe estimatethe average separation for the passive single-star model. justassumethattherearetwopopulations,onemorenumerousand Comparingthismodeltotheobservations,wefindthatthepassive doesn’t receivekicks(theOstars),andonethat islessnumerous single-starmodelisinconsistentwiththeobservations.IfLBVsdo andwhosevelocitydistributionisdominatedbykicks. passivelydispersewiththesamevelocityastherestoftheOstars, Onceagain,wemayuseourgeneral analyticexpression, re- then we propose that LBVs are the product of a merger and are latingtheseparations,age,andvelocities,eq.(15),butthistimewe rejuvenatedstars. expressvLintermsoftheage,x=tL/t1,andthemeasuredvalues Inthiscase,weneedtoconsidertheaverageseparationwhen oftheseparations, the dispersal velocity for LBVsand O starsare the same. Inthis scenario,ourgeneralanalyticexpression,eq.(15),reducesto vL = xO 2 hSOLi 2 (1−xαβ−1) −1 1/2 . (17) v x hS i α−1 SL = 1 1+ x 2 (1−xOαβ−1) 1/2 . (16) O (cid:18) O (cid:19) (1−xOβ ) SO "2 (cid:18)xO(cid:19) ! (1−xαβ−1) # Smith&Tombleson(2015)showedtheaveragedistancefrom LBVsto the nearest O star is ∼ 6.5 larger than the average dis- Equation(16)representsthepassivemodel. tancefromOstartothenearestOstar.IftheageofLBVsaresim- ilartotheaverage mid-typeOstar, thenintheassumption of the kickmodel,thisimmediatelyimpliesthatvLisroughlyninetimes 4.2 INCONSISTENCYINTHEPASSIVEMODEL; largerthanvO.Inthenextsubsection,weestimatetheLBVs’drift IMPLIESMERGERANDREJUVENATION velocitygiventhismodel. Next, we use the passively dissolving solution, eq. (16), to show thattheisolationofLBVsisinconsistentwiththesingle-starsce- 4.4 ESTIMATIONANDINTERPRETATIONOFTHE nario. If LBVsaremassive starsabove 27 M and evolve asiso- ⊙ KICKVELOCITY latedstars,thenFigures7&8demonstratethatthemaximumages oftheseLBVsiswhollyinconsistentwiththelargeseparationsob- Figure10showstheinferredkickvelocityasafunctionoftheLBV servedformassivestars. age.Ifthemassgainerwhicheventually becomes theLBVgains Thepassive model predictsmuch lowerseparations thanthe littlemass, the there is littlediscrepancy between the zero main- observational data. See Figure 6 for an illustration. The orange sequence mass and the final mass. In thiscase, there is littledif- curve line represents the passive model, S in eq. (16). The ference between itsapparent age and itstruemain-sequence age. LBV solidbrownlineillustratestheLBVs’averageseparationobtained Thenitstrueageisrelativelyshortandtheonlywaytogetalarge fromthedatacomparedtothereferenceaverageseparation, SLBV . separationwithalargekickvelocity.Inthisscenario,wefindthat S0 MNRAS000,1–12(2017) ModelingLBV Isolation 9 Mdeath [M⊙] 050505 0 5 0 5 766554 4 3 3 2 MTO [M⊙] 00 5050 5 0 5 0 1100 5544 3 3 2 2 −−11 1100 ]] gg ee dd −−22 [[ 1100 SS gg oo SLBV ll LBVc −−33 1100 LBV passive model TO death −−44 1100 00 11 22 33 44 55 66 77 88 tt [[MMyyrrss]] Figure6.LBVisolation isinconsistent withthesingle-star modelandpassive dissolution ofthecluster. Thesolidbrownline showstheLBVs’average separationobtainedfromthedata.Theorangecurveshowsouranalyticdescription,eq.(16),fortheaverageseparationinthecontextofpassivedissolution. Thismodelrequiresareference;weusedthemid-typeOobservationsasthereference.Thelinesegmentsshowthepurportedmassesandallowableagesfor theLBVs(solidsegments)andLBVcandidates(dashedsegments).IfLBVmassestimatesarecorrect,thenevenwhenoneconsidersthemaximumagefor LBVs,theseparationsaremuchlargerthanwhatthesingle-starpassivemodelpredicts. thekickcanbeashighas200km/s.Ontheotherhand,ifthemass wepresentthescaleoftheproblem;inasubsequentpaper,wewill gainishigh,thenthetruemain-sequenceagewouldbemucholder modelthedistributionofobservedvelocitiesonewouldexpect. than the current mass imply. With a much older age the velocity require to get a large separation is much lower. It might even be zero,inwhichcase,theLBVistheproductofamerger.Thehor- izontalblacksolidlinerepresentstheaverageobservedseparation 5 SUMMARY formid-typeOstars.Thesolidbluelinelinerepresentsourmodel Smith&Tombleson(2015)foundthatLBVsaresurprisinglyiso- toinferthekickvelocity,eq.(17). latedfromotherOstars.Theysuggestedthattherelativeisolationis inconsistentwithasingle-starscenarioinwhichthemostmassive Thoughwepredictthatthekickvelocitiesmaybeashighas starsundergo an LBVphase on their way toevolving into aWR ∼200km/s,wenotethatthekickmaybequitelow,evennearzero. star. Instead, they suggested that abinary scenario is likely more Humphreysetal.(2016)arguedthatnoneoftheLBVsintheLarge consistentwiththerelativeisolationofLBVs.Inthispaper,wetest MagellanicCloudhavehighvelocities.Theysuggestthatmostof theseassertionsbydevelopingcrudemodelsforsingle-starandbi- theLBVvelocities(listedinTable3of Humphreysetal.(2016)) naryscenariosinthecontextofclusterdissolution.Evenwiththese areconsistentwiththesystemicvelocitiesofthelargeMagellanic crudemodels,wefindthattheLBVs’isolationisinconsistentwith Cloud,concludingthattheobservedvelocitiesareinconsistentwith thestandardpassivesingle-starevolutionmodel.EitherLBVsare thekick.InFigure10,weshowthatthekickvelocitymaybeany- overluminousandtheyaremuchlessmassivethanstandardmod- wherefrom0to∼200km/sdependingontheorbitalparametersat els would infer, or they are the result of binary interactions. We thetimeoftheSN,andhowmuchmasswastransferred.Tofurther findthattheLBVisolationismostconsistentwithtwobinarysce- constrainthemassgainerandkickmodel,onewillneedtoproperly narios;eitherLBVsaremassgainersandreceiveakickanywhere model binary evolution including explosions and kicks. For now, from0to∼200km/s,or theyaretheproduct ofmergersandare MNRAS000,1–12(2017) 10 Aghakhanlootakanlooet al. Mdeath [M⊙] 050505 0 5 0 5 766554 4 3 3 2 MTO [M⊙] 5050 5 0 5 0 00..1144 5544 3 3 2 2 00..1122 00..1100 MLBV(TO) ∼ 18 M⊙ MLBV(death) ∼ 27 M⊙ ]] 00..0088 MLBV∼ 50 M⊙ tLBV ∼ 7.3 Myrs gg ee dd [[ 00..0066 OBS SS 00..0044 Passive Model 00..0022 00..0000 00 11 22 33 44 55 66 77 88 t [Myrs] Figure7.TherelativeisolationofLBVsisconsistentwithabinarymergerinwhichtheLBVisarejuvenatedstar.Thesolidbrownlineshowstheobserved LBVaverageseparation. IfLBVspassively dissolvewiththerestofthecluster(orange line), thenweinfer anaverage ageforLBVsof7.3Myrs.This correspondstothemain-sequence-turn-offtimefora18M starandthedeathtimefora27M star.However,theaveragemassforLBVsestimatedfrom theirluminositiesisroughly50M .Starsthismassivedo⊙notlivelongenoughtopassivelydis⊙persetolargedistances.Ontheotherhand,ifLBVsarethe productsofamerger,andtheprima⊙ryhasamassbetweenabout18M and27M ,thentherejuvenatedstarcouldhaveahighluminosity,highmass,and oldageallowingittodispersetolargerdistances. ⊙ ⊙ rejuvenatedstars.Ofcourse,LBVsmayactuallyrepresentacom- km/s.Forthemid-andlate-type,werequiredriftvelocitiesonthe bination of thesetwoscenarios. Itisquitepossible that someare order of 24 and 18 km/s, respectively. The higher drift velocities massgainersandsomearetheproductofmergers. for later-typeO starssuggests that binarity and kicks mayplay a prominentroleinclusterdissolution.Infact,somefractionoflater- Inordertoconstrainthesemodels,wefirstreproducethere- type O stars may be mass gainers or the product of mergers. In sultsofSmith&Tombleson(2015).Similarly,weobtaintheposi- a future paper, we will investigate whether one can constrain the tionof allO starswithin10◦ projected radiusof 30Dor, andwe fractionofkicksandstrongbinaryinteraction. construct distributionsof distancestothenearest O star.Ourdis- tributionsareverysimilartotheirs.Wefindthat thedistributions UsingtheresultsoftheMonte-Carlosimulationsasaguide, for theLBVsand Ostarsareveryunlikely tobedrawnfromthe we develop an analytical model for the average separation as a sameparentdistributions.Inparticular,theaveragedistancetothe functionofdriftvelocityandtime.Theseanalyticscalingsstrongly nearestOstaris∼6.5timeslargerforLBVsthanOstars.Tobet- suggest that LBV isolation is inconsistent with single-star stellar terinformourmodels,wefurthercharacterizethedistributionsand evolution.Moreover,thesescalingssuggestthatLBVisolationsug- find that all of the nearest neighbor distributions are log normal. gests that either LBV stars have lower initial masses (and hence, Thefactthatthedistributionsaresimpleandlognormaldemands longer lifetimes) than one would infer from luminosities, or the thatourmodelsforclusterdissolutionarealsosimple. isolation is most consistent with some sort of binary interaction; WeproposesimpleMonte-Carloandanalyticmodelsforthe eitheramerger orkick.IfLBVshavethesamedispersion veloc- dispersalofopenclustersofOstars.Inthismodel,wesamplefrom itythatweinferfrommid-typeOstars,thenthetimetogettothe distributionsofclustersizes,theSalpeterIMFforstarsandrandom relatively large isolation is 7.3 Myrs. However, the average mass ages.TomatchtheobservedseparationsforearlyOstars,wefind ofLBVsis50M whichhasamaximumtimeof∼4.7Myrs.This ⊙ thattheearlytypeclustersneedadriftvelocityontheorderof11 isclearlyinconsistent. Ontheother hand, binaryinteractions can MNRAS000,1–12(2017)