Astronomy&Astrophysicsmanuscriptno.Paper_rev2 c ESO2013 (cid:13) January15,2013 The average GeV-band Emission from Gamma-Ray Bursts J.Lange1 andM.Pohl1,2 1 DESY,15738Zeuthen,Germany 2 InstituteofPhysicsandAstronomy,UniversityofPotsdam,14476Potsdam,Germany Received/Accepted ABSTRACT 3 Aims.Weanalyzetheemissioninthe0.3–30GeVenergyrangeofGamma-RayBurstsdetectedwiththeFermiGamma-raySpace 1 Telescope.WeconcentrateonburststhatwerepreviouslyonlydetectedwiththeGamma-RayBurstMonitorinthekeVenergyrange. 0 TheseburstswillthenbecomparedtotheburststhatwereindividuallydetectedwiththeLargeAreaTelescopeathigherenergies. 2 Methods.ToestimatetheemissionoffaintGRBsweusenon-standardanalysismethodsandsumovermanyGRBstofindanaverage n signalwhichissignificantlyabovebackgroundlevel.Weuseasubsampleof99GRBslistedintheBurstCatalogfromthefirsttwo a yearsofobservation. J Results.Althoughmostlynotindividuallydetectable,theburstsnotdetectedbytheLargeAreaTelescopeonaverageemitasignificant fluxintheenergyrangefrom0.3GeVto30GeV,buttheircumulativeenergyfluenceisonly8%ofthatofallGRBs.Likewise,the 4 GeV-to-MeVfluxratioislessandtheGeV-bandspectraaresofter.WeconfirmthattheGeV-bandemissionlastsmuchlongerthan 1 theemissionfoundinthekeVenergyrange.TheaverageallskyenergyfluxfromGRBsintheGeVbandis6.4 10 4ergcm 2yr 1or − − − only 4%oftheenergyfluxofcosmicraysabovetheankleat1018.6eV. · ] ∼ E Keywords.Gamma-rayburst:general-Methods:statistical-Surveys H . h p 1. Introduction sionofGRBs. Therefore,weplacedthefocusonburststhatdo - not show a significant signal in the GeV range when analyzed o Since its launch in June 2008 the Fermi Gamma-ray Space individually. A subsample of 99 GRBs has been defined that r t Telescope has broadened our understanding of Gamma-Ray were in principle detectable with Fermi-LAT and occurred in s Bursts(GRBs).Thetwomaininstrumentsonboardofthesatel- regionsof low backgroundemission coming from other galac- a [ lite are the Gamma-Ray Burst Monitor (GBM, Meegan et al. ticandcosmicsources.Foreachburstthespectrumofexpected (2009)) and the Large Area Telescope (LAT, Atwood et al. backgroundphotonsandaneffectiveareaisestimated.Together 1 (2009)). Together they are capable of observing GRBs over 7 with the spectrum of observed photons we thus determine the v decades of energy. Especially the LAT that covers the high- fluencefromthesebursts. 4 energyregionfrom30MeVto300GeVhasthepotentialtopro- 1 videnewinsightintotheunderlyingphysicsofGRBs.However 9 onlyasmallfractionoftheGRBsdetectedwiththeGBMwere 2. Framework 2 1. individuallydetectedwiththeLAT(Gehrelsetal.2009). MostoftheGRBsdetectedbytheGBMdonottriggertheLAT. 0 WhereasthekeV-MeVemissionmaywellbequasi-thermal ThereforeitisexpectedthatmostoftheGRBsdonotshowasig- 3 emission,i.e.ofphotosphericorigin(Eichler&Levinson2000; nificantsignalabovebackground.Theburststhatdidnottrigger 1 Ryde & Pe’er 2009; Pe’Er et al. 2012), the GeV-band emis- the LATwill be referredto as GBM-detectedbursts.We deter- : sion is indicative of particle acceleration to very high energies minetheiremissionbycountingLAT-detectedphotonswithina v and can potentially probe whether or not GRBs are powerful certain time interval and a certain solid-angle element, hence- i X enoughtoprovideasignificantpartofultrahigh-energycosmic forth referred to as the Area of Integration (AOI). The choice r rays(Kotera&Olinto2011;Eichler&Pohl2011).Recently,the of the time window of observation and the AOI is important, a GeV-band photon output of GRBs was estimated on the basis becauseGeV emission canbe delayed(Abdoet al. 2010a)and of the relatively few GRBs, that are individually detected with theGRBpositionistypicallynotwelldeterminedbytheGBM. Fermi-LAT, and was found to be small in comparisonwith the ForlongobservationtimesandalargeAOIthesignalmightnot sourcepowerneededto sustain cosmic raysabovethe ankle in bedistinguishablefromthebackgroundwhileforshortobserva- thespectrum,at1018.6eV(Eichleretal.2010).Onesourceofun- tionsand a small AOI the emission could simply be missed. A certainty in this statement is the unknownlevel of high-energy fewgeneralconsiderationsarethereforeinorder. photonoutputofthemanyGRBsnotindividuallydetectedwith LAT(Waxman2010). 2.1.EnergyIntervals Here we re-analyze data of the Fermi-LAT detector with a viewtoinfertheGeV-bandhigh-energyemissionfromGRBs.In In general the LAT is designed to detect photons in an energy earlierstudiesthefocushadbeenplacedonstudyingeachburst rangeof30MeVto300GeV(Rando2009).Theanalysiscould individually (e.g. Abdo et al. 2011; Ackermann et al. 2012c; in principle be performed for the entire energy range, but the Zheng et al. 2012a,b). However, by analyzing many GRBs to- sensitivityofthedetectorvarieswithenergy.Theeffectivearea gether one is able to obtain more precise results for the emis- of the LAT has been derived through Monte Carlo simulations 1 J.Lange&M.Pohl:GeV-bandEmissionfromGRBs andverifiedwithflightdata(Ackermannetal.2012a),anditis ThereforetheradiusoftheAOIwillbethestatisticalerroradded isverylowforenergiesaround100MeVincomparisontoGeV inquadraturewith2 (PSF,68%containmentfor300MeV)and ◦ energies. Thusat low energiesthe numberof detected photons 3 (systematicerror).InthecaseofGRBpositioningwithother ◦ islow,buttheyrepresentalargeportionofthefluxbecausethe detectorsthan GBM the systematic uncertaintyis set to 0 . As ◦ actual photonflux is normally a decreasing functionof the en- afigureofmeritroughly 2/3ofallobservedGeVphotonsof ∼ ergy(Ackermannetal.2012b).Additionally,thereconstruction the GBM-detectedburstsare expectedto lie within theangular ofthephotonarrivaldirectionislesspreciseforlowerenergies, radius whichfurtherreducesthesignal-to-backgroundratio.Asaresult thedetectionsignificanceforenergiesbelow.300MeVshould α = σ2 +σ2 +σ2 . (1) beverylow andtheflux determinationuncertain.We therefore AOI q stat sys PSF analyzetheenergyintervalfrom300MeVto30GeVonly. As we sum the number of gamma-ray events over many GRBs, the global photon-detectionefficiency is approximately 2.2.TimeIntervals 2/3. RecentstudiesoftheGeV-bandemissionfromGRBssuggesta significantproductionofhighlyenergeticphotonslongafterthe 3. SampleDefinition prompt emission at keV–MeV energies (e.g. Abdo et al. 2009; For the analysis we use a sample of GRBs detected by the Rubtsovetal.2012),mandatingthattheGRBsbemonitoredfor GBM from the beginning of normal science operation of the long time intervals after the promptemission phase. The dura- FermisatelliteonAugust04,2008untilJuly9,2010aslistedin tionofGRBsischaracterizedbyT ,themidtimeinwhich90% 90 Paciesasetal. (2012).In thistimeinterval472burststriggered of the fluence is observedin the BATSE energyrange(50 keV theGBM.SomeoftheseGRBshavetobeexcludedbecausethey to 300keV). Thedurationofthe GBM emission phaseandthe didnotlieintheFieldofView(FoV)oftheLATorwerenotob- LAT emission phase is likely to be correlated and therefore it servedduringnormalscienceoperationofFermi.Itisalsouseful isreasonabletochoosetimeintervalsasmultiplesof T rather 90 to excludecertainGRBs thatareexpectedtohavea low signal thanabsolutetimeperiods. tobackgroundratio.Afterapplyingallcuts99burstsremain.14 ofthemhavebeenpreviouslydetectedbytheLAT. 2.3.AreaofIntegration From the GBM and other observatories the position of every 3.1.LATFieldofView burstisknownwithaprecisionofupto1 whiledatafromother ◦ The GBM can detect GRBs in almost every direction. On the observatoriesora combinedanalysismaypermita localization other hand the LAT has a FoV of 2.4 sr at 1 GeV after all in the arcsecond range. A comparisonof such localizationsin- ∼ analysiscutsforbackgroundrejectionhavebeenmade(Atwood dicates a systematic error in the positioning of the GRBs with et al. 2009). This means that only some of the GRBs detected GBM. The best fit for the systematic error is the combination by the GBM are actually in the FoV of the LAT and therefore of2Gaussianswithdispersion2.6 with72%weightand10.4 ◦ ◦ detectablebyit. Thesensitivityin termsoftheeffectiveareais with28%weight,tobeaddedinquadraturetothestatisticalerror adecreasingfunctionoftheangleθ betweenthephotonarrival (Paciesas et al. 2012).The expecteddistributionof GBM burst direction and the LAT boresight. Photons arriving at θ & 75 localizationerrorsareshowninFig.1.Forallotherlocalization ◦ likelyescapedetection.Itisthereforeusefultoconcentrateonly sourcesweneglectsystematicerrors. on GRBs where the angle between the GRB position and the LATboresightduringtheobservationwas 70 . ◦ ≤ 0.25 3.2.DataQuality 0.20 Some GRBs have to be rejected because of the quality of the D° dataeveniftheyareintheFoVoftheLAT.Asrecommendedby (cid:144)1 0.15 theLATteam,FermiisrequiredtobeoutsidetheSouthAtlantic @ y Anomaly (SAA) and in normal science data-taking mode dur- nsit 0.10 ingtheobservationtime.Additionally,anEarth-relativezenith- e D anglecutof100 isimposedtopreventthespill-inofemission ◦ from the Earth’slimb. All bursts are requiredto fulfill this cri- 0.05 terion at least in a time interval of 10 T starting from the 90 × beginningoftheT time. 90 0.00 0 5 10 15 20 25 30 3.3.BackgroundEmission Angle@°D Toavoidalargenumberofbackgroundphotonsitisreasonable Fig.1.TheexpectedGBMlocalizationaccuracyfor1 (dashed), ◦ to excludecertain regionsof sky that are knownto havea par- 5 (fine dashed) and 10 (continous) statistical error radius. A ◦ ◦ ticularly high background intensity in the high-energy region. normaldistributionisassumedforthestatisticalerror. Thedominantbackgroundsourcesarediffuseemissionfromthe galacticplaneaswellasbrightgamma-raysourceslikepulsars. The imperfect angular reconstruction of the LAT, charac- It hasbeen foundthat the photonintensity atGeV energiesin- terized by the point-spread function (PSF), may lead to a loss creasesbyroughlytwoordersofmagnitudewhenobservingthe of events, even if the actual GRB was located within the AOI. galacticplane(Ackermannetal.2012b).Regionswithagalactic 2 J.Lange&M.Pohl:GeV-bandEmissionfromGRBs latitudebof b < 5 +α willbeexcluded,whereα isthe 4.2.AnalysisoftheEmissionPhase ◦ AOI AOI | | radiusoftheAOIspecifiedinequation1.Itisalsousefultoex- FortheanalysisoftheemissionphasetheP7SOURCE_V6class cluderegionsaroundverybrightgamma-raysources.Themost is also used. The number of expected background photons is luminousonesaretheBlazar3C454.3,theVelaPulsar,theCrab calculated by computing an exposure map for the AOI. In this Pulsar,theGemingaPulsar,andthepulsarsPSRJ1709-4429and case the exposure map describes how each source at the sky PSRJ2021+4026.Theangularseparationtothesesourcesisre- contributes to the observed photon counts inside the AOI. It is quiredtobeatleast1 +α .ExceptfortheBlazar3C454.3all ◦ AOI definedas: thesesourcesliewithinthegalacticplaneandhavealreadybeen excluded. ǫ(E,pˆ)= dtdpˆ R(E,pˆ,pˆ ). (2) Z obs obs AOI 3.4.LocalizationError E is the energy, pˆ and pˆ the true and observeddirection obs Finally it is also useful to exclude GRBs with a high localiza- and R is the response derivedfrom the IRFs. This can be used tion error because a large AOI with correspondinglyhigh total to estimate the background photons inside the AOI from the backgroundwouldhavetobeconsidered.Byrequiringthatthe backgroundmodel.Energydispersionhasbeenneglectedinthis statisticalerrorradiusbesmallerthan5 thenumberofsamples step.TogetanestimateonthefluenceoftheGRBtheexposure ◦ isdecreasedbyafactorof 2/3.Thereisacorrelationbetween hastobeestimated.ThepossibleGRBpositionsaredistributed thestatisticalerrorradiusa∼ndthefluence.Therefore,byapply- throughoutthe AOI. The best estimate for the exposure of the ingthiscutontheerroranglemostlyfaintburstsarerejected. GRBistheintegraloftheexposureconvolvedwiththedistribu- tionfunctionφoftheexpectedlocalizationerrors(seeforexam- pleFig.1). 4. Method φ(δpˆ) ǫ(E)= dpˆ ǫ(E,pˆ) (3) Z 2πsin(δpˆ) 4.1.ModelingtheBackground Sky TheFermiScienceToolsareusedfortheanalysisofthebursts. Inthiswaytheerroroftheburstlocalizationandthefactthat AlthoughforthedurationofasingleGRBthebackgroundemis- some GRBs actuallylie outsidethe AOI is automaticallytaken sionisalmostnegligible,itbecomescrucialwhensummingover intoaccount. manyGRBs.Themainsourcesofbackgroundradiationarethe galacticandextragalacticdiffuseemission,misclassifiedevents, 5. Results andindividualsourceslikepulsars. Using binned likelihood analysis we construct a model of Itisusefultofirstlookattherawphotoncountsincomparison instrumentalbackgroundandthegamma-rayskybyfittingdata tothenumberofexpectedphotonsforGRBsnotdetectedbythe obtained in a time intervalof 3 106 s until 1 hourbefore the LAT.TheresultsareshowninTable1. × triggertimeofeachburst.Thepointinghistoryofthespacecraft duringtheemissionphaseoftheGRBthenpermitsaprediction T [T ] T [T ] GRBs n λ of the expected backgroundfor each burst. The eventfiles and start 90 stop 90 obs 5 2.5 75 4 2.35 thespacecraftfilesareavailableattheFermimissionwebsite1. − − 2.5 0 79 1 2.90 TheLAThasdifferentinstrumentresponsefunctions(IRFs).The − IRFs differespecially in the efficiencyin terms of the effective 0 2.5 85 13 3.53 area,thePSFandtheenergydispersion.Asrecommendedbythe 2.5 5 85 7 3.68 LAT team the P7SOURCE_V6 class is used together with the 5 7.5 85 8 3.83 corresponding galactic diffuse model (gal_2yearp7v6_v0) and 7.5 10 85 11 3.93 isotropicspectraltemplate(iso_p7v6source).Itoffersveryhigh 10 12.5 79 5 3.26 data quality with low background from misclassified photons 12.5 15 76 5 2.99 andafairlylargeeffectivearea. 15 17.5 76 6 3.02 17.5 20 73 3 2.92 The Region of Interest(ROI) for modelingthe background 20 22.5 71 8 2.36 emission is a circle around the burst location with a radius of 22.5 25 67 4 2.17 22 ; the energyrangeis 300 MeV to 30 GeV. Forthe analysis ◦ 25 27.5 64 3 1.92 only time intervals are used during which the LAT was not in 27.5 30 64 5 1.88 theSAA, innormalscience-operationmode,andin thenormal rangeof rockinganglesof notmorethan 52 . Additionallythe 30 32.5 63 1 1.78 ◦ ROI is requiredto notoverlapwith the Earth’slimb.The point 32.5 35 61 1 1.57 andextendedsourcesarelistedintheFermiLATSecondSource 0 10 85 39 14.97 Catalogwhichisbasedontwoyearsofobservations(Nolanetal. Table 1. The number of observed photons (n ) and the ex- obs 2012). As recommendedby the LAT team ten energy bins per pectedbackground(λ)fortheGBM-detectedburstsforthedif- decade and angular pixels of size 0.2 degrees are used. A first ferenttimeintervals.ThetimesaregiveninmultiplesoftheT 90 fit is performedwith the DRMNGB optimizerand the result is timeofeachburstandrelativetothestartofthe T timeinter- 90 then used for a second fit with the NEWMINUIT optimizer to val. Fortimes outside the 0 – 10 T time intervalsome GRBs 90 getmorepreciseresults. havetobeexcludedonaccountofcutsthatwereoriginallyonly appliedinthistimeinterval. 1 http://heasarc.gsfc.nasa.gov/ 3 J.Lange&M.Pohl:GeV-bandEmissionfromGRBs Aswecanseethereisasignificantemissioninthetimein- 300MeVand30GeVweassumetheemissionfollowsapower- terval from 0 to 10 times the T time. In this time λ = 14.97 lawwithindexof 2.3.Asweshallseebelow,thisindexisap- 90 − photonsare expectedand n = 39 observed.The chance proba- propriateforLAT-detectedGRBs,whereastheaverageemission bilityforabackgroundfluctuationis1.7 10 7,basedonPoisson ofGBM-detectedburstsisbetterdescribedwithapower-lawin- − × statistics dexof 3,inwhichcasethephotonfluencewouldbeunderesti- − matedbyabout10%.The68%-confidencestatisticaluncertainty P(n n ,λ)= ∞ λne λ . (4) of the fluence has been calculated using an incompletegamma ≥ obs X n! − function,i.e.theBayesianinversionofaPoissondistributionfor n=nobs uniformprior,andisconsiderablylargerthanthesystematicun- ThisisasignificantsignalwhensummedoverallGRBsbut certainty in estimating the fluence. The analysishas been done clearlynotsignificantforasingleburstforwhichthenumberof forthe GBM-andLAT-detectedburstsseparately.Theresult is observed counts is typically smaller than 1 and the number of showninTable3andgraphicallydisplayedinFigure2. observedphotonsmostly0or1.Forthe11observedshortbursts (T <2s)therewas1observedphotonand0.06expectedback- 90 groundphotonswhich is too low a signal to claim a detection. Time[T ] Fluence[γcm 2] 90 − Fortheremaining74longbursts(T > 2s)therewere38ob- 90 servedphotonsand14.91expectedfrombackground.Notethat GBM LAT before the GRB trigger we observe 5 events and expect 5.25. ( 5)–( 2.5) 0.93+1.12 10 5 0.64+3.47 10 5 Laterthan25T aftertriggerweobserve10eventsandexpect − − −1.30× − − × − 90 ( 2.5)–0 10.29+7.35 10 6 5.34+4.55 10 5 7.15, indicating that there is little, if any, emission at that late − − 5.40× − 5.98× − − − stage,andnoevidenceforanyactivitybeforetheGBMtrigger. 0–2.5 4.86−+11..9842×10−5 66.03−+44..1172×10−4 2.5–5 1.67+1.31 10 5 9.63+1.52 10 4 1.44× − 1.57× − − − 5.1.DetectionSignificance 5–7.5 2.07+1.38 10 5 5.75+1.15 10 4 −1.50× − −1.19× − One can search for significant signals of single GRBs in these 7.5–10 3.47+11..7621×10−5 15.34+65..4982×10−5 − − data.Forthispurposethenumberofobservedandpredictedpho- 10–12.5 0.92+1.17 10 5 33.64+8.69 10 5 tonsarecomparedforthe4timeintervalsbetween0to10T as −1.31× − −9.19× − 90 12.5–15 1.08+1.19 10 5 18.00+6.44 10 5 wellasthetotaltimeinterval.Asacross-checktheLAT-detected 1.34× − 7.00× − − − bteusrtsetds wfoirll5aldsoiffbeerentetstteimdewiinthtetrhvealssaamnedmtheetrheofdo.re99wbeuhrsatvsearaet 1175.5––172.05 01..0650+−+011...943440×1100−55 1113..7751+−+565...337731×1100−55 most 495 trials, because the time intervals are not all indepen- −1.17× − −6.05× − 20–22.5 3.18+1.57 10 5 14.42+5.98 10 5 dent.Thereforetheprobabilitythatthebackgroundemission,λ, 1.71× − 6.62× − − − reachesorexceedstheobservedemission, n ,bynormalfluc- 22.5–25 1.09+1.18 10 5 16.49+6.30 10 5 obs 1.36× − 6.90× − tuationsshouldbesmallerthan10−4fora5%post-trialschance 25–27.5 0.67+−1.06 10 5 1.96+−3.23 10 5 probability.All14burstspreviouslydetectedbytheLATshow −1.32× − −2.37× − 27.5–30 1.95+1.38 10 5 1.95+3.20 10 5 asignificantsignalinatleastonetimeinterval.Thisisnotsur- 1.55× − 2.36× − − − prisingastheseburstsactuallytriggeredtheLATandalsohave 30–32.5 5.00+8.66 10 6 4.48+3.72 10 5 − 6.37× − 4.90× − high-precisionlocalizationseitherbytheLATorotherobserva- − − 32.5–35 3.77+8.98 10 6 2.06+3.35 10 5 tories.Howevertherewere2otherburstswithsignificantemis- − 6.60× − 2.46× − − − sion:GRB100207BandGRB081009A.Thelatteronehasalso Table 3. We show the photon fluences for different time inter- been reportedas a candidate for a LAT-detected burst with the valsaroundthestartoftheT90 timeinterval.TheGBMfluence samemethodbutwithawiderenergyrange(E >100MeV)and istheweightedaveragefluenceofthe85burstsnotdetectedby longerobservationstimes(t = 1500s)(Rubtsovetal.2012). the LAT. The LAT fluence is the weighted average fluence of obs HoweverithasnotbeenaddedtotheLATBurstcatalogsofar. the14burstsdetectedbytheLAT,whichareindividuallylisted TheresultsofthisanalysisareshowninTable2. in Table 4. The LAT fluence is dominated by the three excep- tionally brightbursts GRB 080916C,GRB 090902Band GRB 090926A. GRB Time[T ] n λ P(n n ,λ) 90 obs ≥ obs 081009A 5–7.5 3 0.0366 7.93 10 6 − × 100207B 0–2.5 2 0.0123 7.54 10 5 − × BothGBM-andLAT-detectedburstsshowGeV-bandemis- Table 2. We show results for the two GRBs not listed in the sionlongafterthebulkofemissionintheBATSEenergyrange LATBurstcatalogthatshowsignificantemissionintheanalysis. (50keVto300keV),whicharisesduringtheT time.Whereas As before, n is the number of observed photons and λ the 90 obs thehigh-energyphotonfluenceintheT timeintervalfromthe expectednumberofphotonsfrombackground. 90 GBM-detected bursts is 2.38+1.25 10 5 γ cm 2, for the time 1.38 × − − interval from 0 to 25 times T− we find an average photon flu- 90 encefromtheGBM-detectedGRBsof19.98+4.33 10 5γcm 2. 4.77× − − Given that there could still be emission at la−ter times which is justtoolowtobedetected,theobservedfluenceduringthe T 5.2.PhotonFluence 90 timeisatmost11.91+6.87%ofthetotalGeV-bandemission.One 7.37 TheaveragephotonfluenceofGRBs canbecalculatedbysep- has to keep in mind−that this is the average fluence weighted arately summing the exposure and the observed photon and with the effective area of each burst. Since the Fermi satellite expected background counts in every energy bin. To estimate occasionally reorients for a very bright burst, our method may the cumulative exposure over the entire energy range between favorbrightburstsespeciallyforlatertimes. 4 J.Lange&M.Pohl:GeV-bandEmissionfromGRBs 8 5.3.AverageFluxfromGRBs We nowdeterminetheaverageemission fromGRBs in theen- 6 ergy range from 300 MeV to 30 GeV. GRBs not detected by -2Dcm 4 wthheerLeAasTLcAoTn-tdriebtuetcetewdiGthRBasflaucecnocuentoffo1r7a+−fl34.u7e×nc1e0o−f31γ20c+m77..58−2×, Γ 10−3γcm−2.Hencethetotalobservedfluenceintherange−from -5 300MeVto30GeVis137+8.3 10 3γcm 2.Since50%ofthe @10 2 observedfluencecomesfro−m8.7o×nly2−GRBs−(GRB090902Band Fn GRB 090926A) the statistical error on this estimate should be 0 roughly0.5/√2 35%. ≈ Toestimate the averageallskyflux ofGRB-producedGeV- bandgammaraysonehastoconsiderallcutsappliedintheanal- -2 ysis.First,someburstsweredroppedonaccountofdataquality 0 10 20 30 as discussed in section 3.2. This affected 27% of all available Time@T D 90 bursts giving a weight factor of 1.37. Then, the 5 cut on the ◦ Fig.2. Average fluences for different time intervals in units of statistical error radius of the burst location is compensated by T . Shown in black (grey) are the average fluences for the anadditionalfactor.Theburstsexcludedinthisstepaccountfor 90 GBM-detected(LAT-detected)bursts.ThefluencesfortheLAT 3% of the overall fluence observed in the GBM energy range. detectedburstswerescaledbyafactorof10 2tofitonthesame This gives a factor of 1.03, assuming that the fluences in the − plot. GBM energy range and in the LAT energy range are propor- tional. (We find less GeV emission than that, and thereforethe truecorrectionfactorissomewherebetween1and1.03).Finally, thelimitationtoburststhatoccurredintheLATFoV(θ 70 ) The photonfluence hasalso beenseparatelycalculatedand ≤ ◦ and were not too close to the galactic plane and the Blazar listedinTable4forallLAT-detectedburstsusingthedatafrom 3C454.3iscompensatedwithafactorof3.46.We combinethe 0to10timestheT timeinterval.Theresultsthusderivedare 90 three correction factors to derive the total efficiency factor as comparable to those previously found with standard methods. 4.88= 3.46 1.03 1.37.Thetotalhigh-energyfluenceoverthe TheBurstCatalog(Paciesasetal.2012)canbeusedtocompare entireskysh·ouldth·ereforebe(6.69 2.37) 10 1γcm 2.The thefluenceinthehigh-energyregionfrom300MeVto30GeV ± × − − effective observation time from August 04, 2008 until July 9, to that in the BATSE range from 50 keV to 300 keV obtained 2010was5.083 107 s,andhencethetotalaveragedfluxfrom with the GBM. To be noted from Table 4 is that the bursts not × allburstsovertheentireskyintheenergyrangefrom300MeV detected by the LAT are fainter in the high-energyregion than to30GeVis(13.2 4.7) 10 9γcm 2s 1. theLAT-detectedGRBswhennormalizedtothesamekeV-band ± × − − − We have also determined the energy fluence from GRBs. fluence. GRBs not detected by the LAT contribute with a fluence of 1.67 10 5 ergcm 2,whereasLAT-detectedGRBsaccountfor − − aflue×nceof1.93 10 4 ergcm 2.Theaveragephotonenergies GRB F [γcm 2] F [γcm 2] − − BATSE − [300MeV,30GeV] − × are 1GeVforLAT-detectedGRBsand 0.6GeVforGBM- 080825C 114.2±0.2 1.81+11..3670×10−3 dete∼ctedbursts,estimatedastheratioofth∼eenergyfluence and − 080916C 6.93 0.2 22.27+2.91 10 3 thephotonfluence.Iftheemissionspectrawerepowerlaws,then ± −2.96× − thespectralindices, s,wouldbes 2.3ands 3. 081024B 0.16±0.03 12.93+67..4580×10−4 The average allsky energLAyT ≃flux is GaBbMou≃t 6.4 − 090217 4.8±0.1 2.12+01..8071×10−3 10−4 ergcm−2yr−1, slightly larger than the estimate of· 090328 7.8 0.2 3.59+−1.29 10 3 Eichleretal.(2010),butstillcompatiblewithitconsideringthe ± −1.57× − uncertainties. 090510 1.11 0.05 11.99+1.68 10 3 ± 1.76× − − 090626 15.0 0.3 2.19+1.36 10 3 ± 1.05× − 6. SummaryandDiscussion − 090902B 29.5 0.3 35.87+3.69 10 3 ± −3.81× − In this study of high-energyemission from GRBs in the range 090926A 40.1 0.4 26.64+3.24 10 3 ± 3.36× − from300 MeV to 30 GeV the focushasbeen placedon GRBs − 091003 17.1±0.3 9.99+56..0857×10−4 notindividuallydetectedbytheLATandreferredtoasGBMde- 091031 3.1 0.2 2.55+−1.18 10 3 tected.Wedefinedasampleof85GRBslistedintheGBMcata- 100225A 1.9±0.1 0.95+−11..1006×10−3 log,74ofwhichcanberegardedaslongbursts(T90 >2s)while ± −0.96× − the remaining 11 GRBs are short (T90 < 2 s) bursts. We find 100325A 2.6 0.1 4.39+5.53 10 4 significant emission above background for the complete sam- 100414A 10.1± 0.4 6.83+−34..0432×10−3 ple. For the long bursts the emission is clearly visible while ± −3.65× − for the short bursts alone the results are statistically inconclu- AllGBMBursts 3.38 0.01 19.98+4.92 10 5 ± 4.77× − sive.Moreover,theGeV-bandemissionlastsconsiderablylonger − Table4.ComparisonofthefluencesfromseveralLAT-detected thantheT90 timeofthekeV–MeVenergyrange,infactatleast bursts and the averagefluence forthe 85 GBM-detected bursts to 10 times T90. A similar conclusion was previously reported with the fluences in the BATSE range taken from the GRB for individual GRBs like GRB 080916C (Abdo et al. 2009), Catalog (Paciesas et al. 2012)and derivedfroma fit to a Band GRB081024B(Abdoetal.2010b)orGRB940217(Hurleyetal. function. 1994).Altogetheronly(12 7)%ofthetotalphotonfluencein ± therangefrom300MeVto30GeVisemittedduringT .Since 90 5 J.Lange&M.Pohl:GeV-bandEmissionfromGRBs thisextendedemissionwasobservedwhensummingovermany GRBs,wecannotdiscriminatebetweencontinousemissionand asequenceofflares.GiventhatthenumberofGRBsinthesam- pleislargerthanthenumberofobservedphotons,itisalsoun- clear whether this delayed emission is a general feature or the productofasubsetofGRBs. Theratioofthefluenceinthehigh-energyregiontotheflu- enceintheBATSEenergyregionislower(6 10 5)forthebursts − notdetectedbytheLATthanforLAT-detecte·dbursts(4.7 10 4). − · The estimated spectra in the GeV band are softer for GBM- detected bursts (with equivalent photon index s 3) than for ≃ LAT-detected bursts (equivalent photon index s 2.3). There ≃ is no indication that GRBs produce particle spectra with typi- calindices s 2whichareoftenassumedinsourcemodelsof ≃ ultra-high-energycosmicrays.Likewise,thereisnoevidencefor apopulationofGRBsthatefficientlyaccelerateparticlestohigh energiesbutindividuallyemittooweaklyintheGeVbandfora detectionwithFermi-LAT. Altogether the bursts not detected by the LAT contribute roughly14%ofallGRB-producedphotons,and8%oftheemit- ted energy, in the energy range from 300 MeV to 30 GeV. Finally, we find that the average allsky gamma-ray flux com- ing from GRBs in this energy range is (13.16 4.65) 10 9 γ cm 2 s 1.Theaverageallskyenergyfluxfrom±GRBs i×n − − − the GeV band is only 4% of the energy flux of cosmic rays abovetheankleat1018∼.6eV. Acknowledgements. We acknowledge support by the Helmholtz Alliance for AstroparticlePhysicsHAPfundedbytheInitiativeandNetworkingFundofthe HelmholtzAssociation. References Abdo,A.A.,Ackermann,M.,Ajello,M.,etal.2010a,ApJ,712,558 Abdo,A.A.,Ackermann,M.,Ajello,M.,etal.2010b,ApJ,712,558 Abdo,A.A.,Ackermann,M.,Ajello,M.,etal.2011,ApJ,734,L27 Abdo,A.A.,Ackermann,M.,Arimoto,M.,etal.2009,Science,323,1688 Ackermann,M.,Ajello,M.,Albert,A.,etal.2012a,ApJS,203,4 Ackermann,M.,Ajello,M.,Atwood,W.B.,etal.2012b,ApJ,750,3 Ackermann,M.,Ajello,M.,Baldini,L.,etal.2012c,ApJ,754,121 Atwood,W.B.,Abdo,A.A.,Ackermann,M.,etal.2009,ApJ,697,1071 Eichler,D.,Guetta,D.,&Pohl,M.2010,ApJ,722,543 Eichler,D.&Levinson,A.2000,ApJ,529,146 Eichler,D.&Pohl,M.2011,ApJ,738,L21 Gehrels,N.,Ramirez-Ruiz,E.,&Fox,D.B.2009,ARA&A,47,567 Hurley,K.,Dingus,B.L.,Mukherjee,R.,etal.1994,Nature,372,652 Kotera,K.&Olinto,A.V.2011,ARA&A,49,119 Meegan,C.,Lichti,G.,Bhat,P.N.,etal.2009,ApJ,702,791 Nolan,P.L.,Abdo,A.A.,Ackermann,M.,etal.2012,ApJS,199,31 Paciesas,W.S.,Meegan,C.A.,vonKienlin,A.,etal.2012,ApJS,199,18 Pe’Er,A.,Zhang,B.-B.,Ryde,F.,etal.2012,MNRAS,420,468 Rando,R.2009,ArXive-prints0907.0626 Rubtsov,G.I.,Pshirkov,M.S.,&Tinyakov,P.G.2012,MNRAS,421,L14 Ryde,F.&Pe’er,A.2009,ApJ,702,1211 Waxman,E.2010,ArXive-prints1010.5007 Zheng,W.,Akerlof,C.W.,Pandey,S.B.,etal.2012a,ApJ,745,72 Zheng,W.,Akerlof,C.W.,Pandey,S.B.,etal.2012b,ApJ,756,64 6