DRAFTVERSIONFEBRUARY2,2008 PreprinttypesetusingLATEXstyleemulateapjv.12/14/05 GLAST TESTINGOFA PULSAR MODEL MATCHING H.E.S.S. OBSERVATIONS OFLS5039 AGNIESZKASIERPOWSKA-BARTOSIK1&DIEGOF.TORRES2,1 DraftversionFebruary2,2008 ABSTRACT LS5039isoneofahandfulofX-raybinariesthathavebeenrecentlydetectedathigh-energyγ-rays,inthis case, bytheHigh-EnergyStereoscopyArray(H.E.S.S.). Thenatureofthissystemisunknown: bothablack holeandapulsarhavebeeninvokedaspossiblecompactobjectcompanions. Hereweworkwithamodelof the high energy phenomenologyof the system in which it is assumed that the companion object is a pulsar 8 rotating around an O6.5V star in the ∼ 3.9 days orbit. The model assumes two differentsets of power-law 0 spectralparametersoftheinteractingprimaryleptonscorrespondingtothetwoorbitalphaseintervalsdefined 0 byH.E.S.S.ashavingdifferentgamma-rayspectraandvery-high-energy(VHE)cutoffs.WeshowtheH.E.S.S. 2 phenomenologyiscompletelyexplainedbythismodel.Wepresentpredictionsforphotonswithlowerenergies n (forE >1GeV),subjecttotestintheforthcomingmonthswiththeGLASTsatellite. WefindthatGLASTis a abletojudgeonthismodelwithinoneyear. J Subjectheadings:X-raybinaries(individualLS5039),γ-rays:observations,γ-rays:theory 9 ] 1. INTRODUCTION the orbit and indicate that the new modelcan nicely explain h the H.E.S.S. observations. We then present predictions for p TheGamma-rayLargeAreaSpaceTelescope(GLAST)is the flux and spectrum as functionof phase for photonswith - thenextgenerationγ-rayobservatorydueforlaunchinMay o lowerenergies,whereH.E.S.S.isnotsensitivebutGLASTis. 2008. Its primary instrument is the Large Area Telescope r LS 5039 is near the galactic center and has two other γ-ray t (LAT), which will measure γ-ray flux and spectra from 100 s MeVto ∼100GeV. LATis thesuccessorofthe EGRETex- sourcesnearby,aswellasasignificantgalacticdiffuseback- a ground:itisnotaneasytargetforGLAST(seeDubois2006). [ perimentthat flew onboardthe Compton satellite. LAT will have better angular resolution, greater effective area, wider Inordertostudyorbitalvariability,theneedtointegratelong 1 fieldofview,andbroaderenergycoveragethanEGRET.De- timeintervalsandmakeharderenergycutstobringthesignal v veloping predictions for the GLAST energy domain upon abovebackgroundwasreported. 7 which to test models of LS 5039 is then timely and worth 8 2. THEPULSARMODELWITHORBITALVARIABILITYINTHE exploring. 4 INTERACTINGPRIMARYSPECTRUM Inthepreviouswork(Sierpowska-Bartosik&Torres2007, 1 The main features of the model are described by wherewereferthereaderforreferences),undertheassump- . 1 tionthatLS5039iscomposedbyapulsarrotatingaroundan Sierpowska-Bartosik&Torres(2007). ForLS5039’sparam- 0 O6.5Vstar in the ∼ 3.9 daysorbit, we presentedthe results etersthestarwinddominatesovertheputativepulsar’s,anda 8 ofatheoreticalmodelingofthehighenergyphenomenology shockwrapsaroundthe latter. We assume thenthat the vol- 0 observedbyH.E.S.S.3 Themaindifferencebetweenthepre- umeofthesystemisshock-separatedasaresultofthecolli- : sionbetweenthepulsarandthemassivestarwinds. Between v vious model and the current one is that the former assumed thepulsarandtheshock,insidethepulsarwindzone(PWZ), i aconstantspectrumdescribingtheinteractingparticlesalong X relativisticleptonsareassumedtobefrozeninthemagnetized the orbit, whereas now it is let to vary. The previousmodel r (including detailed account of the system geometry, Klein- pulsarwindwhichpropagatesradiallyfromthepulsar.Inthis a PWZ,wecomputeKlein-NishinaICcascadesagainstthermal NishinainverseCompton(IC),γγabsorption,andcascading) wasabletodescribereasonablywelltherichdetailsfoundin radiationfromthemassivestar. Secondaryγ-raysmoveinto themassivestarwindregionandwhereassomeescapethebi- thesystem,bothfluxandspectrum-wise.However,almostthe completespectralenergydistributionatsuperiorconjunction, nary system, others get absorbed due to γγ process. These cascadesarefollowedbymeansofaMonteCarloprocedure. as well as other features of the observedorbital variation of We have computed geometric dependences upon the opaci- the H.E.S.S. energyspectra, remained to be consistently ex- tiesto theseprocesses. Forclosebinaries, theradiationfield plained. Withmodelsthatdonotentirelymatchtheobserved ofhotmassivestars(typeO,BeorWR,havingtypicalsurface dataatthehighestenergies,thestudyoflowerenergypredic- tionslackstestingpower(strictlyspeaking,thesemodelsare temperaturesintherangeTs ∼104−105Kandlineardimen- already ruled out by H.E.S.S. observations). Here, we ana- sionRs ∼10R⊙)dominatesalongthewholeorbitoverother possiblefields(e.g.,themagneticfieldorthethermalfieldof lyze and motivate the description of the VHE data allowing the neutron star). This thermal radiation field is anisotropic for a variable spectrum of interacting primary leptons along fore+e− injectedclose to the pulsar(theradiationsourceis 1InstitutdeCie`nciesdel’Espai(IEEC-CSIC),CampusUAB,TorreC5, misplacedwithrespecttotheelectroninjectionplace). 2aplanta,08193Barcelona,Spain.Email:[email protected] We assume that the injected power in relativistic leptons, 2Institucio´CatalanadeRecercaiEstudisAvanc¸ats(ICREA).Email:dtor- which initiate the cascading processes, is a fraction of the [email protected]γ-rayflux–consistentwiththeorbital spin-down luminosity, LSD. We particularize on the sim- plest case in which leptons are described, after being repro- timescale asdetermined byCasares etal. (2005)–andfastvariability dis- playedontopofthisperiodicbehavior,bothinfluxandspectrum(Aharonian cessed by losses which dominance can be a function of or- etal.2006). bitalphase,byapower-lawinenergythatmaybeconstant(as 2 A. SIERPOWSKA-BARTOSIK & D.F. TORRES inSierpowska-Bartosik&Torres2007)orvaryalongtheor- teractingpopulationremainsthe same. Giventhatthe losses bit. Here,wefocusonthecaseinwhichtwodifferentpower- dominancealongLS5039’sorbitcanbeafunctionofphase, laws are assumed for the interacting lepton population, cor- and in addition that also the opacity to pair production is respondingto the two orbitalintervalsproposedbyH.E.S.S. phase dependent, it seems natural to assume that in approx- aroundinferior(0.45<φ≤0.90)andsuperior(φ≤0.45and imating the distribution of the interacting lepton population φ > 0.90)conjunction(INFCandSUPC).Table1showsthe withapower-law,ithasadifferentindexalongthetwobroad- modelparameterschosen. In bothcases we assume a nomi- phaseintervalswhichqualitativeobservedfeaturesdifferthe nalvalueforLSD = 1037 ergs−1, andEmax=50TeV. Note most. In this context, the distinction between electron spec- thatLSDandthefractionofitthatisconvertedintorelativistic traatINFCandSUPCshouldbeunderstoodasanaverageof leptons(β) are obviouslyrelated. The position of the shock a smoother phase-dependence of the electron primary spec- (which defines the size of the PWZ where we compute the trum,whatcouldinturnbetestedwithfuturequalityofdata cascades) is defined by the parameter η = LSD/(M˙ Vwc), ifmorecomplexmodelingisavailable. where the wind velocity, Vw, dependson the radial distance fromthemassivestar,andM˙ isthestarmass-lossrate. Then, 3. MATCHINGOFH.E.S.S.RESULTSANDPREDICTIONSFOR GLAST M˙ Vw andLSD arealsoconnected. Differentsetsofparame- Fig. 1 shows the results of our model, both flux and ters(e.g.,asmallerLSD withahigherβ)givesimilarresults. spectrum-wise, compared with corresponding H.E.S.S. data This is the consequenceof a mild dependencewith η of the (obtained from Aharonian et al. 2006). The spectra for LS distancefromthepulsartotheshock.Inthemodelspresented here, only a small fraction (∼1%) of the pulsar’s LSD ends 5039ispresentedintwobroadorbitalphaseintervalsaround INFCandSUPC.Eachoftheexperimentaldatapointsrepre- upinrelativisticleptons.Thisisconsistentwithionscarrying sentsatypicaltimespanof28min.Thecorrespondingphases muchofthewindluminosity. for apastron, periastron, INFC, and SUPC are marked. Two Oncethepulsarinjectsrelativisticleptons(whatcoulditself periods are shown. To ease the comparison with our earlier be subject to orbital variability), or alternatively, a close-to- resultusingaconstantspectrumofinteractingparticlesalong the-pulsarshockacceleratesaprimarypopulationassumedto theorbit(Sierpowska-Bartosik& Torres2007)we includeit beisotropicinthepulsarrestframe(e.g.,Kirketal.1999),the (withafinergridding)alsointheseplots. interactingleptonpopulationarisesfromtheequilibriumbe- H.E.S.S.hasalsoprovidedtheevolutionofthenormaliza- tween injectionand the losses operatingin the system along tionandslopeofapower-lawfittothe0.2–5TeVdatain0.1 the orbital evolution. In this sense, SUPC and INFC of LS phase-binningalong the orbit (Aharonian et al. 2006). The 5039 may indeed represent qualitatively different physical use of a power-law fit was limited by low statistics in such scenarios,asnotedalreadybyAharonianetal. (2006). Let’s shorter sub-orbital intervals, i.e., higher-orderfunctional fit- consider first the maximum energy to which leptons are ac- tings such as a power-law with exponential cutoff were re- celeratedinthecasethelatterproceedsinashock. Thehigh ported to provide a no better fit and were not justified. To temperatureofthestarandthehardspectrumoftheVHEradi- directlycomparewiththeseresults,wehaveappliedthesame ationmeasuredimply(unlessaveryhardinjectionproceeds) approachto treat the modelpredictions, i.e., we fit a power- that IC musthappenin the Klein-Nishinaregime. For dom- law in the same energyrangeand phasebinning. Thiscom- inant IC cooling, shock acceleration and cooling timescales equalityimpliesthatEmax ∝(B/w)3.3,whereBisthemag- parisonisshowninFig. 2. Wefindarathergoodagreement betweenmodelpredictionsanddata. netic field and w is the target photon energy density. Since w ∝d−2,withdthebinaryseparation,andBisadecreasing Fig. 1 also shows the spectral energy distribution predic- functionofd,e.g.,B ∝d−1,Emaxincreasesbyafactor∼10 tionsextendedforenergiesabove1GeV.Thesteeperthepri- maryspectrumis,thehigherthefluxproducedatlowerener- from periastron (in the SUPC broad-orbital range) to apas- gies, whatisparticularlynotablefortheSUPC broadphase- tron (in the INFC broad-orbitalrange), leading to a spectral interval(seeboldandthinlinesofFig.1). Theminimumflux hardeningaroundINFC.Atthehighestenergies,synchrotron losses(whichtimescaleis∝ E−1)areexpectedtodominate neededforasourcetobedetectedbyGLASTaftera1-month and1-yearofoperationinall-skysurvey,foraE−2 source.4 andsteepenthespectrum. ThechangeoverenergyfromICto Atthesefluxlevels,a20%-uncertaintyinthedeterminationof synchrotrondomination of the radiative losses depend on B andd;ifBisnotexactly∝1/d,itmayalsointroduceaspec- theflux,aresultingsignificanceabout8σ,andaspectralindex determinedtoabout6%wouldbeachieved.Evenwhenthese tralhardeningatapastron.Finally,adiabaticlosseshappening sensitivities maybe slightly worse for a low-latitude source, withthetypicalscalesofthesystemassociatedwiththeflow if this model is correct, GLAST should be able to quantify patterns–i.e.,thestand-offdistanceofthepulsarwindshock the orbitalvariabilityafter a few monthsof integration. The (whichscaleswith, andislessthanthedistancebetweenthe stars), divided by the post-shock speed (∼ c/3)– generates factthatthesystemperiodicityis∼3.9daysallowsforafast build-up of integration time around each of the portions of a timescale that is independentof energy and, for LS 5039, theorbit.Amonth-integrationaroundSUPCwillbeobtained closetoorevensmallerthantheKlein-Nishinacoolingtimes. after ∼2 months of all-sky survey, putting this model to the In general, if leptonsenter the productionregion with a rate test soon after GLAST launch.5 Our model also predicts a describedbyapower-lawwithindexα0,inthecaseofdom- clearanti-correlationbetweentheexpectedoutputatTeVand inantradiativelosseseitherbysynchrotronorThompsonIC, GeV energies, which can be seen comparingthe lightcurves theslopeoftheinteractingpopulationissteppenedtoα0+1. shown in Figs. 1 and 3. The lower the energy range, the InthecaseofdominancebyKlein-NishinaIC,thedecreasein thecrosssectionissuchthatthedistributionindexisslightly 4www-glast.slac.stanford.edu/software/IS/glast latperformance.htm modified or even hardened at most by < 0.5 in index, with 5 LS5039ispositionally coincident with3EGJ1824-1514(Hartmanet hardeningdecreasing with the ratio U⋆/UB (Moderskiet al. al. 1999,Paredesetal. 2000),whoseaverageγ-rayfluxabove100MeVis 2005).Inthecaseofadiabaticdominance,theslopeofthein- ∼ 3.5×10−7 photonscm−2 s−1. Ourmodelpredictions areconsistent withthisobservationtoo. GLAST TESTINGOF A PULSARMODELOF LS5039 3 more anti-correlated the signal is, what is a consequence of fluenceoftheorbitalinclinationangleisminoralongallen- the phase-dependence of the cascading and absorption pro- ergyintervalswithinourtheoreticalmodel.Predictionsofthe cesses. A hardness ratio defined within the GLAST energy model at lower γ-ray energy domains are ready for testing domain does not vary as much as it does when constructed with GLAST. We find that GLAST is able to judge on this combiningthefluxesatlowandhighγ-rayenergies(afactor model within one year. Features both at the lightcurve and of 3versusanorderofmagnitude),butitmaybeusefulasa spectrallevel can be recognized. Additionaltests will come firstcheckbeforeintegrationtimeisachievedfordetermining at higher energy yet beyond the current reach, e.g., behav- thespectrum.ThisisshowninFig. 4. ior of the spectra at shorter phase intervals to be compared with data from the next generation of ground-based instru- 4. CONCLUDINGREMARKS ments: H.E.S.S.II,andtheplannedCerenkovTelescopeAr- AdetailedmodelingofLS5039,undertheassumptionthat ray(CTA). it contains a pulsar, is able to fully describe the challenging H.E.S.S.-observedphenomenologyfoundinthesystem. Ob- servedlightcurve,spectra(Fig. 1),andshort-timescalespec- tral variability (Fig. 2) are matched with a variable spec- We acknowledgethe use of IEEC-CSIC computercluster, trum of primary particles along the orbit, where two phase J. M. Paredes and O. Reimer for comments, and support by intervals are considered. It is interesting to see that the in- grantsMEC-AYA2006-00530andCSIC-PIE200750I029. REFERENCES AharonianF.,etal.2006,A&A460,743 KirkJ.G.,BallL.&SkjaeraasenO.,Astropart.Phys1999,10,31 DuboisR.2006,ProceedingsofScience(MQW06)068,intheProceedings Paredes,J.M.,Mart´ı,J.,Ribo´,M.,&Massi,M.2000,Science,288,2340 oftheVIMicroquasarWorkshop,Sept.18-22,2006,Como,Italy. Sierpowska-BartosikA.&TorresD.F.2007,ApJLetters671,145 Casares,J.,Ribo´,M.,Ribas,I.,etal.2005,MNRAS,364,899 Hartman,R.C.,etal.,1999,ApJS,123,79 ModerskiR.,SikoraM.,CoppiP.S.,&AharonianF.2005,MNRAS,845 4 A. SIERPOWSKA-BARTOSIK & D.F. TORRES TABLE1 MODELPARAMETERS Parameter Adoptedvalue Constantleptonspectrumalongtheorbit FractionofLSDinleptons:β 10−2 Slopeofthepower-law:Γe −2.0 Variableleptonspectrumalongtheorbit FractionofLSDinleptonsatINFCinterval:β 8.0×10−3 Slopeofthepower-lawatINFCinterval:Γe −1.9 FractionofLSDinleptonsatSUPCinterval:β 2.4×10−2 Slopeofthepower-lawatSUPCinterval:Γe −2.4 GLAST TESTINGOF A PULSARMODELOF LS5039 5 5 SUPC apastron INFC periastron SUPC apastron INFC 5 SUPC apastron INFC periastron SUPC apastron INFC -1 s ] 4 -1 s ] 4 -2 m -2 m c c ph 3 ph 3 -12 10 -12 10 V [ 2 V [ 2 e e T T 1 1 > 1 > 1 F F 0 0 i = 30o i = 60o 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 1,6 1,8 2,0 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 1,6 1,8 2,0 orbital phase orbital phase FIG.1.—Top:Lightcurvesforphotonswithenergyabove1TeVfordifferentbinaryinclinations(dot-dashed:i=300,solidlines:i=600),comparedwith H.E.S.S.data. Blacklinesstandforresultsobtainedwithavariableinteractingspectrumalongtheorbit;green(light)linescorrespondtotheconstantspectrum case.Bottom:SpectraaroundINFC(inred)andSUPC(inblue).H.E.S.S.dataisshowninthesamecolors.Resultsforbothcasesofinteractingelectronspectra aregiven.Shadedregionsareenergyrangesforwhichwestudythelightcurvesbelow.Thetwohorizontallinesbetween1and100GeVrepresentsensitivitiesof GLASTintheall-skysurveymode. 6 A. SIERPOWSKA-BARTOSIK & D.F. TORRES FIG. 2.—Shadedareasintheleft(right)panelshowthechangeinthenormalization(photonindex)ofapower-lawphotonspectrafittedtothetheoretical predictionforeachofthe0.1binsofphase.Thepresentedresultsareforthemodelofthevariableinteractingspectrum.Thetwodifferentcolorsoftheshading standforthetwoinclinationanglesconsidered. Thesizeoftheshadinggivesaccountoftheerrorinthefittingparameters. DatapointsrepresenttheH.E.S.S. resultsfortheequalprocedure:apower-lawfittotheobservationalspectraobtainedinthesamephasebinning. GLAST TESTINGOF A PULSARMODELOF LS5039 7 3-11-2-1F ( 100-10 GeV) [ 10 ph cm s ] 01234567 SUPC apastron INFC periastron SUPC apastron INFC -9-2-1F (10-100 GeV) [ 10 ph cm s ] 0123 SUPC apastron INFC periastron SUPC apastron INFC - 8-2-1F (1-10 GeV) [ 10 ph cm s ] 01234 SUPC apastron INFC periastron SUPC apastron INFC 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 1,6 1,8 2,0 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 1,6 1,8 2,0 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 1,6 1,8 2,0 orbital phase orbital phase orbital phase FIG. 3.—Predictedtheoreticallightcurvesfortheenergyintervals100GeV–1TeV,10–100GeV,and1–10GeV.Bothinclinationanglesandinteracting leptonspectraconsideredareshown(dot-dashed: i=300,solidlines:i=600,blacklines:variablespectrumofprimaryleptons,green(light)lines:constant spectrumalongtheorbit). 8 A. SIERPOWSKA-BARTOSIK & D.F. TORRES V)) 0,25 SUPC apastron INFC periastron SUPC apastron INFC V)) 0,03 SUPC apastron INFC periastron SUPC apastron INFC Ge 0,20 Ge 0 0 1 1 F(1- 0,15 F(1- 0,02 V)/ V)/ e e HR (F(10-100 G 000,,,001050 23 HR (F(10-10 G 00,,0001 o o i = 30 i = 30 o o -0,05 i = 60 i = 60 -0,01 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 1,6 1,8 2,0 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 1,6 1,8 2,0 orbital phase orbital phase FIG. 4.—Hardnessratiosasafunctionoforbitalphase,fortwoinclinationangles(dot-dashed: i=300,solidlines:i=600).Colorstylefollowsthatused inpreviousfigures.TheenergyregimesconsideredareF(1−10GeV),F(10−100GeV),andF(102−103GeV).