Table Of Content

Astronomy & Astrophysics manuscript no. 15980 © ESO 2011 January 18, 2011 Letter to the Editor Twelve-hour spikes from the Crab Pevatron M. Balbo1,2, R. Walter1,2, C. Ferrigno1,2, P. Bordas1,3 1 INTEGRAL Science Data Centre, Université de Genève, Chemin d’Ecogia 16, CH-1290 Versoix, Switzerland 2 Observatoire de Genève, Université de Genève, Chemin des Maillettes 51, CH-1290 Sauverny, Switzerland 3 Institut fu¨r Astronomie und Astrophysik, Universita¨t Tu¨bingen, Sand 1, 72076 Tu¨bingen, Germany Received October 22, 2010; refereed December 9, 2010; accepted December 15, 2010 Abstract 1 Aims.TheCrabnebuladisplayedalargeγ-rayflareonSeptember18,2010.Tomorecloselyunderstandtheoriginofthis 1 phenomenon,weanalyzetheINTEGRAL(20-500keV)andFERMI(0.1-300GeV)datacollectedalmostsimultaneously 0 during the flare. 2 Methods. We divide the available data into three different sets, corresponding to the pre-flare period, the flare, and the subsequent quiescence. For each period, we perform timing and spectral analyses to differentiate between the n contributions of the pulsar and from the surrounding nebula to the γ-ray luminosity. a Results.NosignificantvariationsinthepulseprofileandspectralcharacteristicsaredetectedinthehardX-raydomain. J In contrast, we identify three separate enhancements in the γ-ray flux lasting for about 12 hours and separated by an 7 interval of about two days from each other. The spectral analysis shows that the flux enhancement, confined below 1 ∼1GeV,canbemodelledbyapower-lawwithahighenergyexponentialcut-off,whereeitherthecut-offenergyorthe modelnormalizationincreasedbyafactorof∼5relativetothepre-flareemission.Wealsoconfirmthattheγ-rayflare ] is not pulsed. E Conclusions. The timing and spectral analysis indicate that the γ-ray flare is due to synchrotron emission from a very H compact Pevatron located in the region of interaction between the pulsar wind and the surrounding nebula. These are . thehighestelectronenergiesevermeasuredinacosmicaccelerator.Thespectralpropertiesoftheflareareinterpreted h in the framework of a relativistically moving emitter and/or a harder emitting electron population. p - Key words. Acceleration of particles – Astroparticle physics – Magnetic fields – (Stars:) pulsars: individual: Crab – o Gamma rays: stars – X-rays: stars r t s a [ 1. Introduction statistically significant increase in the Crab flux was ob- 2 Withanintegratedluminosityofabout5×1038ergs−1 and served in the IBIS/ISGRI light curves between 20 and 400 keV, as well as in the Swift/BAT 15 − 50keV flux at a v a distance of ∼2kpc, the Crab supernova remnant is very level of 5% at the 1σ confidence level (Ferrigno et al. 2010; 7 bright from the radio domain to TeV energies (see e.g., Markwardt et al. 2010). 9 Hester 2008, for a review). It is powered by a pulsar spin- 3 ning on its axis in about 33ms that injects energetic elec- Swift/XRT observed the Crab region for 1ks on 2010 3 September 22 at 16:42 UT and did not reveal any signifi- trons into the surrounding nebula. Nearly all the nebular . cant variation in the source flux, spectrum, and pulse pro- 2 emission up to 0.4GeV is believed to be produced by syn- file (Evangelista et al. 2010). The ultraviolet and soft X- 1 chrotron cooling of these electrons in an average magnetic 0 fieldof∼300µG.Athigherenergies,inverseCompton(IC) ray images excluded the presence of any bright unknown 1 cooling dominates. field object that could have contributed to the γ-ray flux : (Heinke 2010). The near-infrared Crab flux in the J and H v The integrated high-energy flux of the nebula and the bands was also constant (Kanbach et al. 2010). Dedicated i pulsarhasbeenremarkablystableoverthepastfewdecades X pointing performed by RXTE did not show any significant andtheobjectisindeedusedasacalibrationsourceinsev- change in the overall spectral properties in the 3-20keV r eralexperiments(butsee Wilson-Hodgeetal.2010, forthe a band (Shaposhnikov et al. 2010). firstreportofasecularX-raytrend).On2010September22 the AGILE collaboration (Tavani et al. 2010) reported the A 5ks TOO Chandra observation was performed on first γ-ray flare from a source positionally consistent with 2010 September 28 to monitor the morphology of the in- the Crab, during which the flux above 100MeV was nearly ner nebula. There are no particular variations with respect double its normal value. The γ−ray flare from the direc- to the over 35 previous observations, with the possible ex- tion of the Crab was confirmed by the Fermi collaboration ception of an anomalous extension of a bright knot closer (Buehler et al. 2010). (down to 3”) to the pulsar, also noticed in one archival ob- During a period partially covering the γ-ray flare, the servation (Tennant et al. 2010). A subsequent HST optical Crab region was observed by the INTEGRAL satellite and observation of the Crab confirmed an increase in the emis- the Swift/BAT telescope during its routine sky survey. No sion about 3 arcsec south-east of the pulsar with respect to archival observations (Caraveo et al. 2010). However, it Send offprint requests to: [email protected] remainsunclearwhetherthisfeatureisrelatedtotheγ-ray 1 M. Balbo et al.: Twelve-hour spikes from the Crab Pevatron event. The Crab γ-ray flux returned to its usual level on 2010 September 23, less than a week after the onset of the flare (Hays et al. 2010). 2. Results 2.1. INTEGRAL The INTEGRAL satellite is equipped with several high- energy instruments covering the energy range from a few keVtoafewMeV.Here,weconcentrateontheresultsfrom the spectrometer SPI (Vedrenne et al. 2003) and the hard X-ray imager IBIS/ISGRI (Ubertini et al. 2003; Lebrun etal.2003).Wedividedtheavailableobservationsintotwo periods: the pre-flare from 2010 September 12 10:32 UT to Figure1. Upper panel: pulse profiles measured by 2010September1805:56UT(exposures:ISGRI280ks,SPI IBIS/ISGRI in the 20-40keV energy band: crosses and di- 336ks), and the flare from 2010 September 18 11:24 UT to amonds represent the pulse profile before and during the 2010 September 23 10:20 UT (exposures: ISGRI 89ks, SPI Crab flare, respectively. Lower panel: significance in stan- 109ks). The gap between the two data sets is due to the dard deviations of the difference between the pulse profile uncertainty in the exact time of the flare onset. The data during and before the flare. areanalysedusingtheoff-linescienceanalysissoftwarever- sion 9 provided by the ISDC (Courvoisier et al. 2003). The data analysis is performed using the SPIROS package and band, and 10% between 80 and 150keV. Above 150keV, theGEDSATtemplatebackgroundmodel,basedonthecount the sensitivity is limited to 30%. rateoftheeventssaturatingtheGermaniumdetector.The spectra obtained for both data sets are equal within the errors. For the pre-flare set, a best-fit to the data is ob- 2.2. FERMI tainedusingabrokenpower-lawmodelwithafixedenergy FERMI/LATisapair-conversiontelescopethatoperatesin break at 100 keV (Jourdain & Roques 2009). The spectral the30MeV-300GeVenergyrangewithunprecedentedsen- indices are Γ = 2.142 ± 0.019 and Γ = 2.194 ± 0.021 1 2 sitivityandresolution(Atwood&etal.2009).Inouranaly- (χ2 = 42/47). For the data-set corresponding to the flare, ses,weusedalltheeventsinthe0.1–300GeVenergyrange a broken power-law model is also adequate, whose best-fit belonging to the “diffuse” class, but rejected photons with indices are Γ =2.103±0.038 and Γ =2.149±0.041 (χ2 1 2 zenithangleslargerthan105◦ toavoidcontaminationfrom = 52/47). The total flux between 50 and 1000 keV is, re- the Earth’s bright γ-ray albedo. We limited our sample to spectively, 2.79+0.09 ×10−8 erg cm−2 s−1 before the flare −0.12 thedatatakenbetween2010September1toOctober6.The and 2.81+−00..1246×10−8 erg cm−2 s−1 during the flare, which analysiswasperformedusingtheScienceTools,providedby are equal within the uncertainties. the Fermi collaboration (Version v9r15p2). To obtain the Whilethe absolutecount ratein IBIS/ISGRIissubject sourceflux,weusedthegtliketool,whichexploitsamax- to a few percent variation due to instrumental systematic imum likelihood method (Mattox et al. 1996). The instru- effects,thebackground-subtractedpulseprofilenormalized ment response function is P6 V3 DIFFUSE and the Galactic to its average count rate can be considered a trustful esti- emissionisreproducedusingthemodelgll iem v02.fit2. matorofthepulseshape.Wecanthuscomparetherelative All sources listed in the Fermi-LAT first year catalogue contributionofthepulsarandnebularemissionsbeforeand (TheFermi-LATCollaboration2010)within7◦ oftheCrab during the γ-ray flare. We exploited the method of Segreto positionandwithaTSdetectiongreaterthan10aretaken & Ferrigno (2007) and used the ephemeris of Lyne et al. into account in the likelihood analysis. Each point in the (1993)1toextractbackground-subtractedpulseprofiles,ac- source light-curve (Fig. 2) is obtained by means of a single cumulatedin100phasebinsintheenergyranges20-40keV likelihoodanalysis.Owingtothelownumberofeventscol- and40-80keV,andin25binsinthe80-150and150-500keV lectedineverytimebin,weusedaPowerLaw23 torepresent energy ranges. the total emission of the Crab. To optimize the signal-to- The sum of the squared differences between the pulse noiseratio(S/N),weaccumulatedthelight-curveswithdif- profiles’normalizedratesdividedbythecorrespondingvari- ferent binnings (6h, 12h, 1d, 2d), finding that the 12-hour ances is distributed as a χ2 with a number of considered binning is the most adequate to follow the Crab count-rate phasebinsminusonedegreeoffreedom.Inallenergybands, evolution. In Fig. 2, the green dashed line indicates the av- wefoundthatthenormalizedpulseprofilesareequalwithin eragefluxfromtheCrabof(286±2)×10−8ph cm−2 s−1,as astatisticaluncertaintyofonestandarddeviation(seee.g., reported by Buehler et al. (2010). Three distinct peaks are Fig.1,whereχ2 =1.18for99d.o.f.).Wecanestimateour clearly visible, which exceed by a factor of nearly three the red sensitivity measuring a deviation from the template pulse pre-flarefluxlevel,correspondingtoadetectionsignificance profile, based on the assumption that the flux increase is > 15σ. The first steep increase occurs between 6:00 and constant throughout the phase. In the 20-40keV energy 18:00UTConSeptember18.In12hours,theflarereachesa range, we would be able to appreciate an increase in the un-pulsed emission as small as 4% at the 99% confidence 2 http://fermi.gsfc.nasa.gov/ssc/data/analysis/software/ level; our sensitivity diminishing to 6% in the 40-80keV /aux/gll iem v02.fit 3 http://fermi.gsfc.nasa.gov/ssc/data/analysis/scitools/ 1 http://www.jb.man.ac.uk/∼pulsar/crab.html /xml model defs.html#powerlaw2 2 M. Balbo et al.: Twelve-hour spikes from the Crab Pevatron Crab light curve (100 MeV - 300 GeV) Pulse profile: BEFORE the FLARE EMREMRnnMMeettaaSSrrnnii ee s s IINNFFOO 0000....4242 919144626266585899 Pulse profile: DURING the FLARE EMREMRnnMMeettaaSSrrnnii ee s s IINNFFOO 0000....4242 858577939355373777 -1-2 ] cmph s10!10-6 -1-1-2 bin cmph s000..34.455!10-6 -1-1-2 bin cmph s000..34.455!10-6 x [ 8 0.3 0.3 u Fl 0.25 0.25 6 0.2 0.2 0.15 0.15 4 0.1 0.1 0.05 0.05 2 0 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Phase Phase 0 55440 55445 55450 55455 55460 55465 55470 55475 Figure3. Exposure corrected Crab pulse profiles. Left Time [MJD] panel: before the flares. Right panel: during the flares. The horizontal green and red lines indicate the off-pulse aver- Figure2. Light curve of the Crab in the range 0.1–300 ages. GeV with 12h time bins. The green dashed line indicates the flux of the Crab estimated over all the Fermi operation period. The green solid line is the average flux between remained constant within the errors. This result confirms September 19 and 21, as reported by Buehler et al. (2010). the preliminary claim of Hays et al. (2010). The red lines represent the average flux before and during To study the high energy spectrum of the Crab, we ex- the flares (see Table 1). The time is expressed in Modified ploitthemaximumlikelihoodmethodforthedatasetlisted Julian Date. in Table 1. Following Abdo et al. (2010), we model the spectral emission from the Crab using two power-law com- ponents for the nebula, which represent respectively the maximumfluxof(8.6±1.1)×10−6ph cm−2 s−1.Thesecond synchrotron and IC emissions, and a power-law with an flux increase occurs about two days after, between 12:00 exponential cutoff to describe the pulsar contribution. As and 24:00 on September 20. The peak flux is (9.4±1.0)× the flare is not pulsed, we fix the parameter related to 10−6 ph cm−2 s−1. The last flux increase was detected be- the pulsar emission to these of Abdo et al. (2010), and tween 18:00 on September 21 and 18:00 on September 22, assume that the IC contribution of the nebula has not reaching a value of (7.5±1.1)×10−6 ph cm−2 s−1. In all changed during the flare. The only free parameters are cases, the flare decay time was about 1 day. On the basis the synchrotron contribution and the normalization of the ofthelightcurvereportedinFig.2,wedividethedatainto Galactic emission. The synchrotron emission is found to three different sets as reported in Table 1. increase from a flux of (5.6±1.3)×10−7ph cm−2 s−1 to (32.4±2.7)×10−7ph cm−2 s−1 in the 0.1-300 GeV band. Table1.TimeintervalsusedtoperformtheFermianalysis. 3. Discussion Todiscriminateamongthevariousmodelscapableofrepro- Pre-flare Flare Post-flare ducing the quasi-exponential turnover of the synchrotron Start 2010-09-01 2010-09-18 2010-09-26 emission of the nebula that peaks below the LAT en- [UTC] 00:00:00 00:00:00 00:00:00 ergy window, we studied the Fermi data complemented by archival CGRO/COMPTEL data (0.75-30 MeV). A single Stop 2010-09-16 2010-09-23 2010-10-05 power-lawcannotreproducethepre-flaredata(χ2/d.o.f.∼ [UTC] 23:59:59 23:59:59 23:59:59 44/15). A power-law with a high energy exponential cutoff can instead reproduce the data (χ2/d.o.f. ∼ 4/14). To model the nebular synchrotron spectrum, we used the following function In Fig. 3, we compare the pulse profile measured before and during the flares. We can estimate the contribution of dN (cid:18) E (cid:19)−Γ (cid:18) E (cid:19) =N exp − . (1) the nebula, by assuming that the pulsar emission is negli- dE 0 1GeV E cutoff gible in the phase ranges φ<0.25 and φ>0.8. Before the flares, the average rate of the nebular emission is (2.44± The best-fit solution yields Γ = −2.20 ± 0.08 and 0.29)×10−8 ph cm−2 s−1 bin−1 ,whereasduringtheflare N0 =(4.3±1.9)·10−10ph cm−2 s−1 MeV−1.Thedifference phase between the quiescent and flaring spectra can be under- it is (8.12±0.55)×10−8 ph cm−2 s−1 bin−1 . After sub- phase stoodbyconsideringtwodifferentextremecasesofeithera tracting the contribution from the nebula, we compute the constant power-law normalisation or a constant cutoff en- average rate of the pulsed emission: before the flare it is ergy. In the former case, an increase in the energy cutoff of (9.32 ± 0.66) ×10−8 ph cm−2 s−1 bin−1 , whilst during phase afactorofnearly5(from77±15MeVto367±45MeV)is the flares it is (8.52±0.90)×10−8 ph cm−2 s−1 bin−1 . needed (as illustrated in Fig. 4). This increase is averaged phase Duringtheflares,weconcludethatthefluxfromthenebula overthewholeflaringperiod,thusrepresentsalowerlimit, increasedbyafactor3.33±0.46.Incontrast,thepulsarflux since in each single flare the cutoff energy might have been 3 CRAMB .SEBDalbo et al.: Twelve-hour spikes from the Crab Pevatron crosses: BEFORE (black) - DURING (red) - AFTER (green) γ ∼ 3−10×109 in the emitting regions. Taking δ ∼ 2.3, 10-9 the comoving cooling timescale for those particles, taking anextremevalueB ∼2mG,is∼0.3d.Thecorresponding observer timescale would then be similar to the decay time ofthepeakspresentintheFermi lightcurveduringtheflar- -2m ] ingperiod,(cid:46)1d.Incontrast,theflarescouldberelatedto c -1erg s aitnyecnohualdncbeedoebletcatirnoendpboyprualaistiinogn,thanedcotnhteinspuuecmtranlovrmaraialibzial-- F [(cid:105) 10-10 tionbyafactorof∼5orbyaddingahardveryhighenergy (cid:105) electron population (leading to a photon index Γ < 1). In thiscase,aDopplerboostingmaynotberequired,andthe observeddurationoftheflarescouldcorrespondtothesyn- chrotron timescale of PeV electrons embedded in magnetic fields (cid:46)1mG. The duration of the three short flares limits the size 10-1 100 E [GeV]101 102 of the emitting region(s) to (cid:46) 1015 cm. The peak luminosity of these flares is higher/brighter than 1035 erg/s, Figure4.Crabspectralenergydistributioninthe100MeV i.e. (cid:62) 0.5‰ of the Crab spin-down luminosity, assuming - 300 GeV energy range. The points with error bars are an isotropic distribution. The distance between the emit- the Fermi detections before the flare (dark green), during ting region and the pulsar can thus be constrained to be the flare (red), and after the flare (light green). The black (cid:54) 6×1016 cm, i.e. not larger than 15% of the size of the dashedlinerepresentsthecontributionfromthepulsar.The bright synchrotron torus observed by Chandra and HST, black dot-dot-dashed line represents the IC emission from and probably consistent with the half-width of this torus. the nebula. The blue and magenta dot-dashed and solid The emitting region could therefore be linked to the inter- lines are the synchrotron nebula and the total emission be- actionzonebetweenthejetandthetorus,whichisfoundto fore and during the flare, respectively. Arrows indicate the have brightened in the HST image obtained on 2 October 95% confidence flux limits. (Caraveo et al. 2010). The three flares separated by two days could possibly be related to various emitting knots in this region. Alternatively, gamma-rays could be produced higher.Inthelattercase,thespectralvariabilitycanbeex- within the jet itself. However, if the emitting region were plainedbyraisingthecontinuumnormalizationbyafactor moving at relativistic speeds, the emission would be radi- of ∼5. atedwithinanangle∼1/δ.Forreasonablevaluesofthejet We note that the non-detection of any significant hard inclination angle with respect to the line of sight (see e.g. X-ray variability during the flare does not allow us to dif- Ng & Romani 2004), this scenario would make the flares ferentiate between the two possibilities as several electron difficult to detect. populationsareprobablypresentinthenebula.Asreported Toconclude,theflarerelativeshortdurations(<1day), by Aharonian & Atoyan (1998), the COMPTEL data are their soft spectrum, and the analysis of the pulse profile in characterized by a flattening of the spectrum that can be the 0.1-300GeV indicate that one or more compact por- ascribed to the synchrotron emission of a separate electron tions ((cid:46)1015 cm or < 0.1”) of the synchrotron nebula are population confined in compact regions such as wisps or responsiblefortheflares.Intheseregion(s),freshlyacceler- knots. The luminosity of this component is less than 1% of atedPeVelectronsarerapidlycooling,causingtheobserved that of the whole nebula. variability. Thecutoffinthesynchrotronspectrumoccursatachar- acteristic frequency ν ∼ 4.2×106Bγ2. Provided that Acknowledgements. PBhasbeensupportedbytheDLRgrant50OG peak synchrotron radiation is the dominant mechanism through 1001. which particles channel their energy, the maximum electron Lorentz factor obtained by equating t to t is sync accel References γ ∝(Bη)−1/2, where η ≥1 is the gyrofactor that char- max acterizes the acceleration rate γ˙ ≡γ/t and t = Abdo,A.A.,etal.2010,ApJ,708,1254 accel accel accel Aharonian,F.A.2000,NewA,5,377 ηE/q Bc. This makes ν independent of B, leading to e peak Aharonian,F.A.&Atoyan,A.M.1998,inNeutronStarsandPulsars: an electron synchrotron energy cutoff ≈ 160η−1 MeV (see ThirtyYearsaftertheDiscovery,ed.N.Shibazaki,439–+ e.g. Aharonian 2000). A higher value may imply that the Atwood,W.B.,etal.2009,ApJ,697,1071 conditions in the accelerator differ from those in the emis- Buehler,R.,etal.2010,TheAstronomer’sTelegram,2861 Caraveo,P.,etal.2010,TheAstronomer’sTelegram,2903 sionregion,e.g.thereisalowermagneticfieldintheformer, Courvoisier,T.,etal.2003,A&A,411,L53 orthatthesynchrotrongamma-raysareproducedinarela- Evangelista,Y.,etal.2010,TheAstronomer’sTelegram,2866 tivisticallymovingregion,whichproducesashiftintheen- Ferrigno,C.,etal.,2010,TheAstronomer’sTelegram,2856 ergycutofftohigherenergiesbythecorrespondingDoppler Hays,E.,etal.,2010,TheAstronomer’sTelegram,2879 Heinke,C.O.2010,TheAstronomer’sTelegram,2868 factor δ. In this scenario, a value δ ∼367/160≈2.3 would Hester,J.J.2008,ARA&A,46,127 be required to explain the energy cutoff obtained during Jourdain,E.&Roques,J.P.2009,ApJ,704,17 the flaring episode. Kanbach,G.,etal.2010,TheAstronomer’sTelegram,2867 On the other hand, magnetic fields at the level of be- Lebrun,F.,etal.2003,A&A,411,L141 Lyne,A.G.,etal.1993,MNRAS,265,1003 tween ∼ 300 µG and ∼ 2 mG are found in the syn- Markwardt,C.B.,etal.2010,TheAstronomer’sTelegram,2858 chrotron nebula and wisps, respectively (see e.g. Hester Mattox,J.R.,etal.1996,ApJ,461,396 2008). Synchrotron radiation at ∼ 1 GeV implies that Ng,C.&Romani,R.W.2004,ApJ,601,479 4 M. Balbo et al.: Twelve-hour spikes from the Crab Pevatron Segreto,A.&Ferrigno,C.2007,inESA-SP,622,633–+ Shaposhnikov,N.,etal.2010,TheAstronomer’sTelegram,2872 Tavani,M.,etal.2010,TheAstronomer’sTelegram,2855 Tennant,A.,etal.2010,TheAstronomer’sTelegram,2882 TheFermi-LATCollaboration,2010,ArXive-prints1002.2280 Ubertini,P.,etal.2003,A&A,411,L131 Vedrenne,G.,etal.2003,A&A,411,L63 Wilson-Hodge,C.A.,etal.2010,ArXive-prints1010.2679 5