Exploring the blazar zone in High Energy flares of FSRQs
L. Pacciani, F. Tavecchio, I. Donnarumma, A. Stamerra, L. Carrasco, E. Recillas, A. Porras, M. Uemura
aa r X i v : . [ a s t r o - ph . H E ] J un Draft version November 8, 2018
Preprint typeset using L A TEX style emulateapj v. 04/17/13
EXPLORING THE BLAZAR ZONE IN HIGH ENERGY FLARES OF FSRQS
L. Pacciani , F. Tavecchio , I. Donnarumma , A. Stamerra , L. Carrasco , E. Recillas , A. Porras ,M. Uemura INAF-Istituto di Astrofisica e Planetologia Spaziale, Via Fosso del Cavaliere 100, I-00133 Rome, Italy INAF-Osservatorio Astronomico di Brera, via E. Bianchi 46, I-23807, Merate, Italy INFN sez. Tor Vergata, Via della Ricerca Scientifica 1, I-00133 Rome, Italy INAF-Osservatorio Astrofisico di Torino, via P. Giuria 1, I-10125 Torino, Italy Instituto Nacional de Astrofisica, Optica y Electronica, Mexico, Luis E, Erro 1, Sta. Maria Tonantzintla, Puebla, CP 72840, Mexico and Hiroshima Astrophysical Science Center, Hiroshima University 1-3-1 Kagamiyama, Higashi-Hiroshima, 739-8526, Japan
Draft version November 8, 2018
ABSTRACTThe gamma-ray emission offers a powerful diagnostic tool to probe jets and their surroundings in flatspectrum radio quasars (FSRQ). In particular, sources emitting at high energies ( >
10 GeV) give usthe strongest constraints. This motivates us to start a systematic study of flares with bright emissionabove 10 GeV, examining archival data of
Fermi -LAT gamma-ray telescope. At the same time, webegan to trigger Target of Opportunity observations to the
Swift observatory at the occurrence ofhigh-energy flares, obtaining a wide coverage of the spectral energy distributions for several FSRQsduring flares. Among the others we investigate the SED of a peculiar flare of 3C 454.3, showing aremarkable hard gamma-ray spectrum, quite different from the brightest flares of this source, and abright flare of CTA 102. We modeled the SED in the framework of the one–zone leptonic model, usingalso archival optical spectroscopic data to derive the luminosity of the broad lines and thus estimatethe disk luminosity, from which the structural parameters of the FSRQ nucleus can be inferred.The model allowed us to evaluate the magnetic field intensity in the blazar zone, and to locatethe emitting region of gamma rays in the particular case in which gamma-ray spectra show neitherabsorption from the BLR, nor the Klein-Nishina curvature expected in leptonic models assuming theBLR as source of seed photons for the External Compton. For FSRQs bright above 10 GeV, we whereable to identify short periods lasting less than 1 day characterized by high rate of high energy gammarays, and hard gamma-ray spectra.We discussed the observed spectra and variability timescales in terms of injection and cooling ofenergetic particles, arguing that these flares could be triggered by magnetic reconnections events orturbulence in the flow.
Keywords: galaxies: active - galaxies: quasars: general - galaxies: Jets - radiation mechanism: nonthermal INTRODUCTION
In the framework of the unification scheme(Urry and Padovani 1995) of Active Galactic Nu-clei (AGN), blazars are the radio–loud AGNs withjets oriented close to the line of sight of the observer.Their emission encompasses the whole electromagneticspectrum, from radio band to gamma-ray energies. Asmall fraction ( ∼
50 objects) have been detected at TeVenergies with Cherenkov detectors (see, e.g., Holder2012).The Spectral Energy Distribution (SED) of blazarsshows two humps, whose origin is believed to be theboosted non-thermal emission from the relativistic jet,overwhelming the thermal components. In general,the synchrotron emission from energetic electrons in atangled magnetic field accounts for the low energy bumpof the SED. The emission mechanisms responsible forthe high-energy bump, peaking in gamma-rays are stillmatter of debate. Both hadronic and leptonic modelscan explain the observations (e.g., Boettcher et al.2013). The high-energy emission is explained in leptonicmodels as due to the Inverse–Compton (IC) scatteringof relativistic electrons of the jet with a seed photon field
EMAIL: [email protected] for both Bl Lac objects and for Flat Spectrum RadioQuasars (FSRQs). Photon field for the inverse-Comptonscattering can originate from the synchrotron emissionitself (Synchrotron Self Compton, SSC, Maraschi et al.1992; Marscher and Bloom 1992). The SED of BL Lacobjects is usually explained with SC and SSC emissions.The high energy bump of FSRQs is usually modeledin the External Compton (EC) scenario, with photonfields for the IC originating from a source external tothe jet. There are several sources of external photonfields that can play a role: the direct thermal radiationfrom the disk, the reprocessed disk emission fromthe broad line region (BLR) or from the moleculartorus (Blazejowski et al. 2000; Arbeiter et al. 2002;Sikora et al. 2002), the thermal radiation from an hotcorona (Sikora et al. 1994; Dermer & Schlickeiser 1993;Ghisellini and Tavecchio 2009).It is generally assumed (e.g., Ghisellini and Tavecchio2009, Sikora et al. 2009) that the intensity of each ex-ternal photon field depends on the distance of the emit-ting region from the Super Massive Black Hole (SMBH)and on the accretion disk luminosity. Particularly im-portant, these external radiation fields can also absorbthe gamma-ray photons, through the pair creation re-action γγ → e ± . In particular, the intense emission (e.g., Poutanen and Stern 2010) from the BLR can par-tially absorb gamma rays at least for FSRQs with themost luminous accretion disk (L disk ∼ - 10 ergs − ). Above 20 GeV/(1+z) a spherical shell BLR isvirtually opaque to gamma rays emitted in the center,mainly due to luminous H Ly α and continuum emissionof the BLR. Liu & Bai (2006) computed the BLR opticaldepth ( τ γγ ) as a function of the location of the gamma-ray dissipation region, assuming a BLR luminosity of L BLR =2.3 × erg/s, and a spherical shell geometryfor the BLR with internal radius R BLR , and external ra-dius R extBLR . They evaluated τ γγ ∼ . ∼
13) at 35 GeV,and ∼ ∼
16) at 50 GeV for gamma-ray photons emit-ted at the mid point R MBLR between internal and externalradius (at the internal radius) of a BLR with luminos-ity L BLR = 2 . × erg/s. Liu, Bai & Ma (2006) alsoevaluated the optical opacity for the case of 3C 279 witha fainter BLR ( L BLR = 2 . × erg/s). They ob-tained τ γγ ∼ R MBLR (at R BLR ) of theBLR shell.In the standard view, high energy emission from FSRQhas been located inside the BLR, whose intense emis-sion provides the ideal environment for a powerful ICemission. Isler et al. (2013) and L´eon-Tavares (2013)found marginal evidence that the gamma-emitting re-gion is located within the BLR during at least twoflares of 3C 454.3. However, there is growing ev-idence that, at least in some occasions or in somesources, the emission can occur much farther (up tofew pc) from the central SMBH. The SED modelingof FSRQs flares bounds the dissipation region at theedge, or outside the BLR in a few cases: for PKS1222+216 (dissipation region > > FERMI–LAT (Atwood et al. 2009). SAMPLE SELECTION
We searched for flares in the
FERMI – LAT dataarchive from all the FSRQs in the second
FERMI – LAT catalog (Nolan et al. 2012) with statistically relevantsignal above 10 GeV. Hereafter, we refer to these flaresas High Energy (HE) flares.We note that at high energy the point spread function(PSF) of the
FERMI–LAT is particularly narrow ( ∼ ◦ at 10 GeV, 95% containment, Ackermann et al.2013; see also the updated PSF information reported inthe FERMI–LAT web page ), and the background neg-ligible. A rough estimation for the order of magnitudeof the background rate in a circular region of radius of0.4 ◦ is ∼ × − cts/d ( ∼ × − ph cm − s − E >
10 GeV) at high latitude, with the satellite scanningthe whole sky. Disregarding differences of exposures todifferent regions of the sky due the scanning strategy,we note that three gamma-rays detected within a fewweek from a circular region of 0.4 ◦ around the knownposition of a FSRQ give a TS (Mattox et al. 1996) ∼ FERMI–LAT standard recipes. Ameaningful HE counts map showing the low backgroundlevel, and the small PSF is shown in figure 1. It hasbeen obtained collecting gamma-rays within a circularregion of radius 20 ◦ around 3C 454.3 and integratingduring an HE flare lasting ∼ ◦ , and 6background gamma-rays within an annular region ofinternal radius 0.4 ◦ and external radius 20 ◦ .A simple and not CPU–time consuming methodto search for HE flares is a two-level algorithm. Thefirst-level algorithm makes a rough analysis of the FERMI–LAT
HE photon list, disregarding differences ofthe exposures, and background. The small PSF, and lowbackground level allowed us to prepare the first–level ofthe flare search algorithm without the use of likelihood igh Energy flares of FSRQs FSRQ name z HE activity period ∆ t HE z >
10 GeV) energyGamma (%) ∗ photonsray (GeV)PKS 0250-225 1.49 2009-02-12 11:00 – 2009-03-07 20:00 9.4 3 0.36/8.4 26 18.4, 16.2PKS 0454-234 1.00 2012-11-24 09:00 – 2012-12-13 12:00 9.6 5 0.055/1.8 71 25.2, 19.3PKS 1502+106 1.84 2009-04-10 05:00 – 2009-05-14 15:00 12.1 9 4.6 × − /8.6 × −
108 19.9, 15.7B2 1520+31 1.49 2009-04-10 14:00 – 2009-04-27 02:00 6.6 4 0.18/6.4 44 27.1, 16.04C +38.41 1.81 2011-07-02 10:00 – 2011-07-13 09:00 3.9 3 0.031/1.7 34 13.9, 10.5B2 1846+32A 0.80 2010-10-16 08:00 – 2010-10-29 14:00 7.4 5 2.6 × − /1.2 × −
39 25.4, 23.5PMN J2345-1555 0.62 2013-04-15 23:00 – 2013-04-29 21:00 8.6 5 6.6 × − /0.29 68 96.8, 37.4CTA 102 1.49 2012-09-18 12:00 – 2012-10-03 21:00 6.2 9 1.7 × − /7.0 × −
136 21.8, 20.1PKS 0805-07 1.84 2009-05-14 13:00 – 2009-05-23 14:00 3.2 9 2.3 × − /1.6 × −
112 23.2, 20.63C 454.3 0.86 2013-09-23 10:00 – 2013-09-25 07:00 1.0 6 6.6 × − /2.2 × −
101 35.8, 28.4
Table 1
List of FSRQs of our sample, together with the HE activity period, the duration of the activity period in the host galaxy frame, and thenumber of HE gamma rays (E >
10 GeV) coming from a circle centered on the source position and of radius 0.4 ◦ . ∗ We give both thechance probability for a bunch of photons within the integration time, and the chance probability for the same bunch of photons withinthe integration time and occurring during an activity period at lower gamma-ray energy (for sake of simplicity, we assumed that all thesources were in an activity period for 1/3 of the whole
FERMI–LAT operations, except for 3C 454.3 where we assumed the source in anactive state for 1/2 of the whole
FERMI–LAT operations). TS significance is estimated through
FERMI–LAT standard analysis recipes.Last column reports the energy of the most energetic photons during the HE activity period.
Figure 1.
Smoothed HE Counts map obtained collecting gamma-rays within a circular region of radius 20 ◦ around 3C 454.3 andintegrating for 1.8 d during an HE flare. The two circles haveradius 0.4 ◦ and 20 ◦ . In the map there are 12 HE gamma-rays, 6of which lie within the 0.4 ◦ radius circular region. analysis, which is indeed necessary when the PSF is large(i.e., at lower energy for the FERMI–LAT , and wheneach Gamma–ray could be associated with more thanone source, or with diffuse–background (Mattox et al.1996). The first–level algorithm will search for HEflares from each known source based on the differentialtime between consecutive HE gamma-rays detectedin a circular region of radius 0.4 ◦ around the sourceposition, without taking care of the exposure, and ofbackground. We define the flaring activity period asthe period of time in which a bunch of HE gamma-raysis detected by FERMI–LAT with differential timesbetween consecutive gamma rays which is less than apredefined quantity ¯∆ t . ¯∆ t is defined to be of themean differential time of HE gamma rays from thesource evaluated from the whole FERMI–LAT archive(i.e., we assert the first-level trigger for the period of time for which the instantaneous photon rate from thesource is at least three times the mean rate from thesource). We re-define the flaring activity period byadding to both the edges of it half of the typical meandifferential time between consecutive events duringflare. We point out that this is only a rough pre-trigger,aimed at accept as much HE flares as possible, with anhigh discrimination factor and with the minimum CPUusage.The first-level trigger generates 0.14 false first-leveltriggers/year for a source with a mean HE countingrate of 0.01 d − , and accepting a detection with 3gamma-rays.The second-level algorithm is based on the refined anal-ysis of the FERMI–LAT data and obviously accounts foreffective area, source exposure, satellite scanning strat-egy, and background: it performs the standard analysisof the data within the already defined flaring activityperiod. The second-level runs only at the occurrence ofthe first-level trigger, considerably reducing the CPUtime needed to search for HE flares. The second–leveltrigger is asserted once the source is detected with a TS ≥
25 in the complementary
FERMI–LAT energy band,0.1–10 GeV.This two-levels algorithm has been applied to botharchival–, and incoming–data, downloaded twice everyday. It provides a non-complete sample. In fact itselects for HE flares starting from their detectability,which still depends on the instrument pointing strategyand Earth occultation.We found HE flares from about 60 FSRQs, and for morethan a dozen of sources we found archival multiwave-length observations during HE flares, or we awardedToO observations with
Swift (Gehrels et al. 2004) duringflares. We investigate here HE flares of FSRQs withsufficient multiwavelength coverage and for which wewere able to derive the disk luminosity of the sourcefrom BLR spectroscopy.The source list, the flaring activity period, the numberof HE gamma rays (E >
10 GeV) collected duringthe activity period, are reported in Table 1, togetherwith the estimated chance probability for a bunch ofgamma-rays to be detected within the reported inte-gration time and simultaneous to an activity period atlower gamma-ray energies. In table 1 we report also thethe likelihood signal significance TS estimated through
FERMI–LAT standard analysis recipes for E >
10 GeV. MULTIWAVELENGTH OBSERVATIONS
In Table 2 we report the timeline of the observatorycampaigns for our sample of flares.
Swift–UVOT obser-vations with all optical–UV filters were performed simul-taneous to the
Swift–XRT observations we report, un-less otherwise specified. All observations at GuillelmoHaro were performed with near IR K s , H, J filters, andall observations with SMARTS were performed with theoptical–NIR filters B, V, R, J, K, unless otherwise spec-ified. Dates in Table 2 and everywhere in the paper arereported in UTC.We report also some peculiar cases: for PKS 0805-07 theNIR observations where performed 10 days before theactivity period in gamma-rays. The results of a multifre-quency monitoring of 4C +38.41 covering almost 4 yearsof activity from Radio to Gamma-rays, and including theactivity period that we study in this paper, is reportedin Raiteri et al. (2012). DATA ANALYSIS
FERMI–LAT analysis
We already stated that we searched for HE flares onFSRQs of the second
FERMI–LAT catalog starting fromthe HE photon list. The alert task has been describedin the sample–selection section. The filtered photon listused in the first–level trigger has been prepared withthe latest standard
Fermi
Science tools (v9r27p1 untilNovember 2013, and v9r32p5 since this date) availableat the time of trigger. We filtered events of event class >
10 GeV, and we disre-garded photons from the Earth’s limb with a cut at 100 ◦ in the zenith angle.We performed the second–level trigger, and theoffline analysis on flaring sources with standard Fermi–LAT
Science tools v9r32p5, using the Pass 7(P7REP SOURCE V15) response functions. We disre-garded photons from the Earth’s limb, adopting a cut at100 ◦ in the zenith angle, and we allowed only events of event class
2. Light curves and spectra were obtainedperforming the unbinned likelihood analysis inside a re-gion of radius 20 ◦ around target–sources. Galactic diffuseand Extragalactic isotropic backgrounds were modeledusing gll iem v05 and iso source v05, respectively. Nontarget sources were taken from the second Fermi –LATcatalog (Nolan et al. 2012).We extracted light curves with energies >
300 MeV inorder to process data with a smaller PSF and reducebackground Gamma-rays from neighbour sources. Thisis a convenient choice for the massive analysis of tem-poral bins. and it causes negligible reduction of signalsignificance.
X-ray data analysis
Chandra (Weisskopf et al. 2002) observed PKS1502+106 in X-rays with the back-illuminated S3 CCD of the Advanced CCD Imaging Spectrometer (ACIS,Garmire et al. 2003). We reduced data and performedthe analysis making use of the Chandra InteractiveAnalysis of Observation (CIAO, Fruscione et al. 2006)version 4.3 software, with calibration version (CALDB)4.1.3. We extracted the source spectrum using a circularregion with a 5 arcsec radius centered on the sourceoptical position. The Background was taken from anearby circular region with a 15 arcsec radius. Weproduced Response Matrix (RMF) and created theAncillary Response File (ARF).We reduced
Swift–XRT (Burrows et al. 2005) datawith xrtpipeline version 0.12.3 software, and analysed itwith standard tools, using the most recent calibrationfiles available. We selected events with grade 0–2 forwindow timing data, and with grade 0-12 for photoncounting mode. We created the Ancillary Response filesusing xrtmkarf .We used an absorbed power–law to fit model to thesources X-ray spectra, with absorption fixed at thegalactic values (Kalberla et al. 2005). Data for PKS1502+106 show a statistically significant excess ( ∼ UV–optical–NIR data analysis
The
Swift–UVOT (Roming et al. 2005) data analysiswas performed with the FTOOLS tasks uvotimsumand uvotsource. magnitudes were evaluated throughaperture photometry within a circular region or radius5 arcsec around the source positions. Backgrounds wereestimated from nearby source–free regions, of radius 9arcsec.We performed simultaneous V and J -band imaging-polarimetry of PKS 1502+106 using the TRISPECinstrument attached to the Kanata V and J -band image. The data werereduced according to the standard procedure of CCDimages. The differential photometry was performedwith a comparison star located at R.A.= 15 : 04 : 36 . . V = 15 .
335 and J = 14 . igh Energy flares of FSRQs Figure 2.
Light curves of PKS 0250-225, PKS 0454-234, PKS 1502+106, B2 1520+31 in gamma rays. left: long integration, right: zoomaround the reported flare. Binsize is 4 d. source X-ray observing period optical–NIR observation/observatoryPKS 0250-225 2009-02-20 15:53 – 2009-02-22 11:14PKS 0454-234 2012-12-04 12:25 – 2012-12-04 19:30 2012-12-05 03:51/SMARTSPKS 1502+106 2009-04-09 12:47 – 2009-04-09 15:02 ∗ ∗∗ Table 2
Timeline of the multiwavelength campaigns on the sources. All X-ray observations where performed with Swift(except for ∗ performedwith Chandra). SMARTS observations where performed with B, V, R, J, K filters, unless otherwise specified; observations at GuillermoHaro where performed with K S , H, J filters. ∗∗ Swift observed with UV filters only.
Figure 3.
Light curves of 4C +38.41, B2 1846+32A, PMN J2345-1555 in gamma rays. left: long integration, right: zoom around thereported flare. Binsize is 4 d, except for the right panel for 4C +38.41 with binsize of 2 d. igh Energy flares of FSRQs Figure 4.
Light curves of CTA 102, PKS 0805-07, 3C 454.3 in gamma rays. left: long integration, right: zoom around the reported flare.Binsize is 4 d, except for the right panel for 3C 454.3 with binsize of 1 d, and for the core of the emission for CTA 102, with binsize of 0.17d. For 3C 454.3 we show the optical light curve with R filter taken within the Yale-SMARTS monitored blazars program. NIR observations of PKS 1502+106, B2 1846+32A,CTA 102, PKS 0805+07 were carried out withINAOE’s 2.1m
Guillermo Haro telescope equippedwith CANICA, a near IR camera. Standard NIR dif-ferential photometry was obtained for 5 arcmin squaredimages, centered on the objects of interest. The adoptedreference local photometric standards were those objectslisted in the 2Mass point source catalog (Skrutskie et al.2006). http://astro.inaoep.mx/observatorios/cananea/ NIR–optical–UV photometry is de-reddened us-ing the interstellar extinction curve proposed inFitzpatrick et al. (1999).For sources at redshift 0.9 or more, the absorptionlines from neutral Hydrogen of the intergalactic medium(IGM) enters the band of the
Swift–UVOT
UVW2 filter(see, e.g., Rau et al. 2012). We used the method pro-posed in Prochaska et al. (2009), to extrapolate at z < <
10% for the photome-try with UVW2, UVM2 and UVW2 filters at redshift of1.1, 1.4 and 1.7 respectively. These reported values arepessimistic because they are evaluated using the propermean free path at 912 ˚ A ( λ mfp ) for sources at redshift 2,instead that redshift <
2. Corrections are uncertain dueto the poor knowledge of the neutral Hydrogen columndensity of the IGM. Our sample contains sources withredshift up to 1.84. We do not try to correct optical–uvphotometry for our sample. We will show that for all thesources but one the corrections will not affect our results. RESULTS
Gamma-ray light curves
We anticipated some of the results in Table 1: the HEactivity for our sample lasts from a couple of days toabout one month. The gamma-ray light curves for oursources are shown in Figure 2, 3, 4, where we also high-light the HE activity period. We note that for the entiresample, the HE activity period corresponds to high ac-tivity also at lower energy. For CTA 102 the core of theactivity period is reported with temporal bins of 0.17 din order to show the fast variability of the source duringthe period of interest.In the last years 3C 454.3 exhibited flaring activitywith peak flux exceeding 10 − ph cm − s − for E >
100 MeV (Vercellone et al. 2011; Pacciani et al. 2010;Bonnoli et al. 2011). The comparison of the gamma rayand optical light curves for 3C 454.3 reveals one of thepeculiarity of the HE flare that we are investigating: itexhibited a gamma ray flux of 350 × − ph cm − s − (E >
100 MeV) which is an order of magnitude fainterthan during the brightest gamma-ray flares, but in opti-cal the flux is comparable to the peak emission observedduring the brightest gamma-ray flares.
Spectral Energy Distribution modeling
The gamma-ray spectra during flares of the sources weinvestigated are reported together with the other multi-wavelength simultaneous data in Figure 6 and Figure 7.For all sources we integrated the gamma-ray data for thewhole period of HE emission, except for B2 1846+32Awith gamma-ray data integrated for 4 days around theNIR and optical-UV observations, and for CTA 102 andPKS 0454-234 showing variability within the HE activityperiod. We integrated gamma-ray data between 2012-09-22 02:00 and 2012-09-24 14:00 for the former, andbetween 2012-12-02 00:00:00 and 2012-12-06 00:00:00 forthe latter.For PMN J2345-1555 we observed a peculiar flare sim-ilar to that reported in Ghisellini et al. (2013b) for thesame source, with synchrotron emission extended up to X-rays.We obtained a multiwavelength SED for an interestingflare of 3C 454.3, which reached a flux of 350 × − ph cm − s − (E >
100 MEV), with a flat gamma-rayspectrum up to 40 GeV and a photon index 1.82 ± FERMI–LAT data (see., e.g., Abdo et al. 2011;Finke & Dermer 2010), showing soft gamma-ray spectrawith a break or a cutoff in the GeV range.We found a similar recent flare in the archival data of3C 454.3. Integrating
FERMI–LAT data between 2013-04-05 17:00 and 2013-04-12:00, we obtained a flux of ∼ × − ph cm − s − (E >
100 MeV), and a gamma-ray photon index of 1.93 ± FERMI–LAT archive other activity periods of the sourceshowing the same spectral characteristics.By definition our sample is biased towards flares showingrelevant emission above 10 GeV. We showed in Table 1that the collected gamma rays with energy >
10 GeVcan not be explained by chance coincidences, thence theobserved emission at HE is intrinsic to flaring state ofthe sources. All the spectra show no, or at least negli-gible evidence of absorption, with the possible exceptionof 4C +38.41 (which will be discussed later in this pa-per), and B2 1520+031. If the gamma-ray emission wereproduced inside the BLR cavity, we expect absorptionfrom γγ absorption with the BLR photons: at the thresh-old energy (E Lyαthr =
25 GeV1+ z ) for γγ absorption with the HLy α target photons, the optical depth is of the order of τ ∼ σ T n Lyα R BLR (Tavecchio et al. 2013) where σ T isthe Thomson cross section, and the density of target pho-tons is n Lyα = L Lyα / πR BLR chν
Lyα , and assuming theblazar dissipation zone in the center of a spherical shellshaped BLR. The H Ly α luminosity can be estimatedstarting from the broad lines spectroscopy, using thetemplate reported in Francis et al. (1991) and the cor-rections proposed by Celotti et al. (1997). The accretiondisk luminosities (L disk ) have been evaluated assumingthe BLR luminosity to be L disk (Baldwin and Netzer1978). The internal radius of the BLR can be in-ferred from the relation connecting it to the disk lu-minosity as indicated by reverberation mapping studies(Bentz et al. 2009). Following Ghisellini and Tavecchio(2009) this can be written as: R BLR = 10 L . disk, cm,and R outBLR ∼ × R BLR . Liu & Bai (2006) performed arefined evaluation of the optical depth as a function ofthe blazar dissipation zone (i.e., outside the BLR cav-ity). We evaluate the optical opacity starting from theirfindings and interpolating for the disk luminosity of oursample.The interpolated optical depth at E
Lyαthr , 35 and 50 GeVfor all the sources is reported in Table 3, where we reportalso the broad line used to evaluate the accretion diskluminosity, and the reference to the original data. Theestimated γγ absorption for all the sources in our sam-ple will produce relevant features in the collected spec-tra if the emitting region is inside the cavity of the BLR( < R BLR ). Moreover, for the 5 objects with the highestdisk luminosities, the BLR opacity will produce relevantfeatures for an emitting region at R MBLR or beyond. Theopacity argument could not be used to exclude emittingregions located toward the outer edge of the BLR, as far igh Energy flares of FSRQs source τ γγ L disk luminosityname at R BLR at R MBLR estimatorat E Lyαthr at 35 GeV at 50 GeV at E
Lyαthr at 35 GeV at 50 GeV erg/sPKS 0250-225 3.7 6.3 7.8 0.6 1.1 1.6 5.3 Mg II (Shaw et al. 2012)PKS 0454-234 3.1 5.3 6.5 0.5 0.9 1.3 3.7 Mg II (Stickel et al. 1993)PKS 1502+106 6.2 10.6 13.1 1.1 1.9 2.7 15. Mg II , C IV (Shaw et al. 2012;Sbarrato et al. 2012)B2 1520+031 4.6 7.8 9.6 0.8 1.4 2.0 8. Mg II (Shaw et al. 2012)4C +38.41 11.4 19.4 23.9 2.0 3.5 4.9 50. Mg II , C IV (Stickel et al. 1993;Sbarrato et al. 2012)B2 1846+32A 3.0 5.1 6.2 0.5 0.9 1.3 3.4 Mg II (Shaw et al. 2012)PMN J2345-1555 2.0 3.4 4.1 0.3 0.6 0.9 1.5 Mg II , C IV (Stickel et al. 1993)CTA 102 10.3 17.6 21.7 1.8 3.1 4.5 41. ×
10 L
BLR (Pian et al. 2005)PKS 0805-07 7.9 13.5 16.6 1.4 2.4 3.4 24. Mg II , C IV (White et al. 1988)3C 454.3 9.2 15.8 19.4 1.6 2.8 4.0 33. ×
10 L
BLR (Sbarrato et al. 2012)
Table 3
Disk luminosity and optical depth for γγ absorption evaluated at E Lyαthr , 35 and 50 GeV for photons emitted at the internal radius( R BLR ) of a spherical shell of BLR, and for photons emitted in the middle between the internal and external radius of the shell( R MBLR = R BLR + R outBLR ) for our sample of FSRQs. The opacity is evaluated interpolating the results of Liu & Bai (2006) for the diskluminosities of our sample.The last column gives the broad lines used to evaluate the disk luminosity. as the opacity becomes negligible, and its effect on thegamma-ray spectra undetectable.In the framework of leptonic models, flares dissipating in-side the BLR cavity will emit gamma-ray photons subjectto the Klein-Nishina suppression above ∼ bulk is 15, and we assume a power-law electrondistribution with slope p=3 (corresponding to a photonspectral index α =1 in Thomson regime). The line ofsight forms an angle bulk with the jet bulk motion. Re-sults are reported in figure 5, normalized to the SED inthe Thomson approximation. In the jet frame, the up-stream boosted BLR energy density prevails in the SEDuntil the jet approaches R BLR . From about this dissi-pation region, the downstream BLR energy density con-tribution is no more negligible. For a dissipation regionat the edge or beyond the BLR, all the BLR photonscome from behind. The net effect is that starting from R BLR , the center of mass energy of the BLR photon -electron scattering starts to decrease for each directionof the incoming electrons, thence progressively mitigat-ing the Klein-Nishina suppression. For a jet dissipatingat R outBLR , the Klein-Nishina suppression is 3.7, 4.3, 5.2,6 for a gamma-ray energy of 25, 35, 50, 80 GeV respec-tively.Combining the two arguments (the model independent γγ absorption, and the Klein-Nishina suppression for theIC emission with seed photons from the BLR in leptonicmodels), the gamma-ray spectra we showed could onlybe explained assuming a dissipation region at the outeredge of the BLR or beyond. Our SED modeling, andtime-resolved spectra will give further clues on the loca-tion of the gamma-ray dissipation region. Figure 5.
Gamma-ray SED for the IC scattering with seed pho-tons from the BLR as a function of the dissipation region. FromBottom Up, the solid curves refer to a dissipation region located atthe center of the BLR cavity, at R BLR , R extBLR . Dot dashed curverefers to a dissipation region at R MBLR . From Bottom Up, the fourdashed curves refer to a dissipation region located at 3 . × R BLR ,5 × R BLR , 6 × R BLR , 8 × R BLR (corresponding to 0 . × R extBLR ,1 . × R extBLR , 1 . × R extBLR , 2 × R extBLR in our model). We modeled the SED for each epoch in the frameworkof leptonic models, and with the parametrization of thephoton field originating from the BLR and dusty torusprovided in Ghisellini and Tavecchio (2009) and alreadyused in Tavecchio et al. (2013). The emission region (the“blob”), assumed to be spherical with radius R and mov-ing with bulk Lorentz factor Γ bulk , carries a magneticfield with intensity B and a population of relativisticelectrons. The electron energy distribution is assumedto follow a smoothed broken power law function between γ min and γ max , with slopes n and n below and above abreak at γ break and normalization K . The blob velocityis assumed to form an angle θ v with respect to the lineof sight, so that relativistic amplification effects can befully specified by the relativistic Doppler factor δ .Given the disk luminosity of the FSRQ, the externalradiation fields interacting through External Compton toproduce gamma rays can be fully parametrized in termsof distance of the radiating ejecta from the SMBH.0As detailed in Tavecchio et al. (2013), with the suggestedparametrization, and with the information of the disk lu-minosity, we can directly obtain an estimate of the loca-tion of the radiating ejecta. We briefly recall the chainof arguments followed in Tavecchio et al. (2013).Let us assume the torus radiation field dominates theemission (e.g., the dissipation region is at least at theouter edge of the BLR), with a black body spectrumpeaked at 3 × eV. The position of the IC peakprovides the value of the Lorentz factor of the electronsemitting at the peak, γ p , which, in turn can be usedto derive the value of the magnetic field from thesynchrotron peak frequency, ν s ≃ . × Bγ δ/ (1 + z )Hz. Since, during HE outburst the IC maximum liesat energies much larger than those usually observedin FSRQ, the corresponding γ p will be larger, di-rectly implying a low magnetic field. A further stepcan be done exploiting the observed IC/synchrotronluminosity ratio, the so-called Compton dominance,directly proportional to the ratio of the external radi-ation and magnetic energy densities in the jet frame, L C /L s = U ′ ext /U ′ B = U ext Γ bulk /U ′ B . Since the magneticfield is known, the Compton dominance allows us toinfer the radiation field energy density and therefore,thanks to the link with the distance from the SMBH,the position of the emission region.If, contrary to our previous assumption, the BLR radia-tion field dominates the emission over the torus IR field,then the magnetic field must be enhanced by a factor ν BLR ν IR ∼
100 (really more than this value on accountof the KN suppression). To reproduce the observedCompton-dominance, we have to rise the U ′ ext (that nowcorresponds to U ′ BLR ) by a factor ν BLR ν IR × τ IR τ BLR ∼ U IR ) makinguse of the seed photon field from the BLR. In fact wehave to rise U ′ BLR by a factor 600 with respect to U ′ IR .Plausible values for the maximum of the ratio U ′ BLR U ′ IR are ∼
100 (Ghisellini and Tavecchio 2009, that adopted τ IR = 0 . ∼
600 (Sikora et al. 2009, that adopted τ IR = 0 . τ IR (e.g., τ IR ≤ U ′ BLR U ′ IR could beobtained putting the dissipation region at ∼ R BLR orbelow, but this range of parameter R diss is excluded bythe lack of γγ absorption and KN suppression in theobserved gamma-ray spectrum.More generally speaking, in leptonic SED modeling thecharacterization of the blazar zone as located outside theBLR (and irrespective of the blob radius, electron den-sity, and Γ bulk ), is a quite stable result, once the regioninside the BLR cavity is excluded. In fact, the optical-UV data show the synchrotron high-energy tail in theSEDs of all the flares, while the gamma-ray data can besubdivided into two subsets, showing (a) the EC HighEnergy tail, or (b) a flat EC spectrum with no fall atHigh Energy (with the exception of PKS 0250-225, B21520+031, 4C +38.41 for which the observations do notdetail the synchrotron emission). In case (a) , these data bind the B/Γ bulk ν ext , and the B / U ext Γ bulk ratios in themodeling, thence the estimate of the U ext and B/Γ bulk parameter does not depend on the estimate of Γ bulk (oncewe have identified the prevailing seed photon field), anddoes not depend on electron energy density and radiusof the emitting blob.Once we have established that the dissipation regionis at a distance R diss > R BLR (outside the BLR cav-ity), there are other two options to consider: i) the BLRprevails on the IR radiation field, then U ′ ext ∼ U ′ BLR ,which rapidly varies with R diss ( U ′ BLR drops of a fac-tor ∼
100 at R extBLR with respect to R BLR , and a furtherfactor 10 at ∼ . × R extBLR ). This circumstance makesthe localization of R diss a quite stable parameter. j) IRprevails on the BLR radiation field , then U ′ ext ∼ U ′ IR , U IR is constant until R diss < R torus , and ν ext = ν IR .Thence the estimate of the U ext parameter does not de-pend on the estimate of Γ bulk . Furthermore, assumingSSC emission dominates the X-ray range, we can con-sider the ratio of SSC to Synchrotron emission to furtherconstrain the model components: the product R blob × n eo (see, e.g., Dermer, Sturner, Schlickeiser 1997 for the for-mula and for the definition of n eo ) must be held constantin the modeling, due to the observational constraints.Now, using the formula for the power spectral density inDermer, Sturner, Schlickeiser (1997), we obtain that theproduct Γ bulk × R blob must be held constant (using theobservational constraints already obtained for R blob × n eo , B/ Γ bulk , and on U IR ). Reasonable values for Γ bulk arein the range 10-50 which are about a factor 2 aroundthe value we chose for our modeling (see below). Thismeans that we could obtain reasonable models varying R blob in the opposite direction with respect to Γ bulk (andat most by a factor 2 / ∼ − of Γ bulk ), andmaintaining the constraints from the measured quanti-ties. But the decreasing of the parameter R diss meets al-most a barrier to low values (at R diss ∼ R extblr ), where U blr rapidly varies and prevails on U IR . Thence R diss > R extblr .Remembering that we chose R blob ∼ . × R diss , thence R diss could be pushed further out with respect to ourmodeling, at the expense of a corresponding lowering ofΓ bulk , or accepting a lower value for the ratio R blob R diss . Sum-ming up, in case (j) R diss is constrained at lower values( R diss > R extblr ), and could be pushed further out thevalue we chose in our modeling by a factor 6 at most.We discussed case (b) in Pacciani et al. (2012). In thiscase, we obtain from the optical-UV and gamma-ray dataan upper limit for the B/Γ bulk ν ext ratio, and a lower limitfor R diss .For two of the flares with poor optical coverage of thesynchrotron emission (PKS 0250-225 and 4C +38.41),the optical data give at least a lower limit for the syn-chrotron peak height, and we were able to derive a lowerlimit for R diss from the SED modeling).The previous considerations apply only for one-zone lep-tonic models. We explicitly do not tried to model theSEDs with other modeling, such as the spine-sheath orhadronic models.We reproduced the observed flare SED adjusting theparameters to obtain the best agreement with data. Thederived model parameters are reported in Table 4. Notethat the radius of the BLR and the torus are not free pa- igh Energy flares of FSRQs source name L disk B γ min γ break γ max n n K δ Γ bulk L pkin L ekin L jet L disk R BLR R torus r blob d blob − erg/s G cm − erg/s erg/s cm cm cm cm[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16]PKS 0250-225 5.3 2.5 55 8 20 2.7 3.2 10 26 25 330 29 60 2.30 5.76 8 9PKS 0454-234 3.7 3.5 350 4 20 2.7 3.2 20 26 15 110 9.5 30 1.92 4.81 8 7.6PKS 1502+106 15 11 100 0.9 17 2 3.35 0.24 26 15 24 5.4 1 3.87 9.68 6 6B2 1520+031 8 6.3 100 0.9 15 2 3.5 0.3 25 14 13.3 3.1 2 2.83 7.07 4 44C +38.41 50 26 13 0.15 6.3 2.7 2.7 16 25 17 79 1.7 1.6 7.07 17.68 4. 4.B2 1846+32A 3.4 59 235 0.27 90 2.2 3.4 47 22 20 5.1 1.9 1.5 1.84 4.61 0.5 0.7PMN J2345-1555 1.5 21 18 7 300 2 4.3 0.03 26 21 4.6 0.42 3 1.22 3.06 2. 2CTA 102 41 14.3 1.7 3.3 30 2 3.55 5.6 20 14 920 8.5 20 2.03 5.09 1 1PKS 0805-07 24 31.5 5 8 50 2.3 3.5 2.4 20 15 460 6.0 20 4.90 12.25 2.5 2.53C454.3 33 28 2.5 4 14 2.25 4. 4.3 24 15 294 4.6 8.8 5.7 14.4 1.5 2.3 Table 4
Parameters of SED modeling of our sample of flares. [1] disk luminosity (10 erg/s), [2] magnetic field (mG) , [3] minimum randomLorentz factor of electrons, [4] break random Lorentz factor of electrons (10 ), [5] maximum random Lorentz factor of electrons (10 ), [6]low energy slope of the electron population, [7] high energy slope of the electron population, [8] electron density (10 cm − ), [9] Dopplerfactor, [10] bulk Lorentz factor, [11] kinetic power of protons (10 erg/s, assuming one cold proton per electron), [12] kinetic power ofelectrons (10 erg/s), [13] Broad Line region Radius (10 cm), [14] Molecular Torus Radius (10 cm), [15] Blob radius (10 cm), [16]Blob distance from the SMBH (10 cm). rameters but are fixed by L disk (which is determined bythe BLR luminosity). Moreover, we reduced the num-ber of free parameters assuming that the radius of theemitting region is roughly 1/10 of the distance from theSMBH. Source radii and Doppler factors can also be con-strained by the observed duration of the gamma-ray highstate, ∆ t high > t cross = R (1 + z ) /δc , estimated from thelight curves. Summing up, the adopted model has a totalof 9 free parameters.Our procedure naturally implies some degree of un-certainty. In this respect, the most critical point isthe position of the cut-off in the gamma-ray spectrum.Higher energies implies higher electron Lorentz factorsand, following the reasoning above, lower magnetic fieldswhich, implying lower external photon energy densities,leads to infer larger distances. In all cases we try toassume the case providing the most conservative esti-mate of the source distance d blob < R torus , at the costto slightly underestimate the flux at the highest ener-gies. This is also justified in view of the fact that, al-though close in time, the data collected in the SED arenot strictly simultaneous and (see below), the gamma-rayspectra show hints of variability at the highest energieson relatively short ( ∼ day) timescales. However, for twosources, PKS 0250-225 and PKS 0454-234, the shape ofthe SED does not allow for the solution correspondingto R BLR < d blob < R torus and forces the assumptionof large distances. The reason, similarly to PKS B1424-418 (Tavecchio et al. 2013), is the large separation be-tween the IC and synchrotron peak, which result in arather low magnetic field. We found dissipation regionslocated between 0.3 and 3 pc from the SMBH, with thepossible exception of B2 1846+32A. We obtained similarresults for GB6 J1239+0443 (Pacciani et al. 2012) andPKS B1424-418 (Tavecchio et al. 2013).The IGM Lyman Complex absorption affects the Swift–UVOT photometry with UVW2, UVM2 and UVW2 fil-ters starting at redshift 1.1, 1.4 and 1.7 respectively. Wenote that all the flares we reported for sources at z > ∼ d blob is at ∼ d blob at ∼
10 pc.
Spectral Variability
For the most powerful HE flares (CTA 102, 3C 454.3,PKS 0805-07, PKS 1502+106) we tried to identify theportion of the flaring period (period A) showing thehighest flux of energetic gamma-rays.For the same sources we used the sub-sequent portionsof the flaring period (period B, C, D) in order toinvestigate spectral differences, and possibly the spectralevolutions. The integration times for period A and B ofeach source are reported in Table 5. For PKS 0805-07,and PKS 1502+106 we enlarged the integration timeto 0.38 d in order to increase the statistics. In thesame Table we report the number of HE gamma raysfor period A, and the chance probability for a bunchof gamma-rays to occur within the integration time ofperiod A, assuming that the mean gamma-ray rate isthe rate of the whole activity period. We note howeverthat we evaluated the chance probability as a statisticalfluctuation of a flat gamma-ray distribution within thewhole activity period of the source. This is not the casefor CTA 102. For this source, we evaluated the meangamma-ray rate not from the whole activity period, butfrom the 2.2 days in which the overall source flux wasabove 2 × − ph cm − s − (E >
300 MeV).The gamma-ray spectra for period A and B, C, D areshown in Figure 8. In the following fit procedures, we2
Figure 6.
SEDs for the reported flares of PKS 0250-225, PKS 0454-234, PKS 1502+106, B2 1520+31, 4C +38.41, B2 1846+32A. Dottedlines represent Synchrotron emission, short dashed lines represent the disk emission and the SSC, long dashed lines represent the EC ontorus photon field. igh Energy flares of FSRQs Figure 7.
SEDs for the reported flares of PMN J2345-1555, CTA 102, PKS 0805-07, 3C 454.3. Dotted lines represent Synchrotronemission, short dashed lines represent the disk emission and the SSC, long dashed lines represent the EC on torus photon field.
Fermi–LAT
PSF at lower energies. This choice have been applied inHayashida et al. (2012).The spectral fit of period A with a powerlaw model forthe sources give an hard photon index (Γ ph ), reported inTable 5. We do not show fit with log-parabolic, broken–powerlaw, powerlaw with exponential-cutoff models,because of the low statistics accumulated for period Afor our sources: there is no gain in the log-likelihoodwith respect to the powerlaw model, and fit do notconstrain model parameters other than source flux andlow–energy photon index ( α for the log-parabola model).For the bright flares of CTA 102 and 3C 454.3, we triedto establish if there was spectral evolution from periodA to period B. As first step, we statistically tested thehypothesis that the spectra of period A and B havethe same shape. To compare data of the two periods,we cannot obtain useful information by direct spectralcomparison using the likelihood evaluated errors in eachband. If we restrict to emission below and above 10GeV (LE and HE respectively), we can use the LE datato normalize the spectrum of period B to the emissionof period A, and compare the emission in HE band forthe two periods. In HE band, we can disregard othersources, and galactic and extragalactic diffuse emission.Thence we can use Poisson statistics to establish theprobability P( shape A = shape B ) that a source witha given flux ratio (R f = F ( E> GeV ) F (0 . ÷ GeV ) ) could give riseto the observed flux and counts during periods A andB. In this evaluation, we assume that F(0.1 ÷
10 GeV)have Gaussian distribution. We report in Table 5the maximum value of P( shape A = shape B ) obtainedvarying R f . In this evaluation, the exposure of thesource in period A and B are taken into account.As second step we made spectral fitting of period B withpowerlaw, broken powerlaw, log-parabola, powerlawwith exponential cutoff models. Results are shownin Table 6. For both CTA 102 and 3C 454.3 thepowerlaw model is disfavoured. The fit with all theother models succesfully constrain parameters. Thefit with broken–powerlaw, powerlaw with exponentialcutoff, and logparabola models give hard low–energyphoton index ( α for the log-parabola model) which areconsistent with the powerlaw photon index evaluatedfor period A and reported in Table 5.We could define the period A for 4C +38.41 on thebasis of two HE photons only, it is reported in Table 5.In this case the statistics prevent to build a statisticallysignificant spectrum, and we enlarged the period Asymmetrically on lower and higher temporal ends, tohave a duration of 0.3 d. While the cumulative spectrumreported in Figure 6 seems to show absorption in the1–10 GeV energy band with a drop of a factor ∼ < γ , t obs = t ′ cool (1 + z ) /δ , can bewritten as: t obs = 3 m e c (1 + z )4 σ T U ′ IR γ δ , (1)where the energy density of the IR torus emission is: U ′ IR ≃ τ IR L disk πR IR c Γ bulk = 2 . × − Γ erg cm − , (2)in which τ IR ∼ . R IR interms of L disk . Inserting this into the previous equationwe finally obtain: t obs = 1 . × z Γ δ γ s . (3)Considering an observed gamma-ray energy E obs and anIR photon field peaked at ν IR , the corresponding Lorentzfactor of the emitting electrons can be derived as: γ ≃ × (cid:18) E obs (GeV) ν IR (Hz) × z Γ δ (cid:19) / . (4)We therefore obtain for E obs =30 GeV and ν IR = 3 × Hz: t obs = 4 × (1 + z ) / Γ / δ / s . (5)The expected timescales, smaller than a day, are in agree-ment with, or even smaller than, the values reported inTable 5. Caveat
An obvious problem of the cooling scenario (and, moregenerally, of the observed variability, see below) is that,assuming a single, homogeneous emission region, theobserved variability timescales would be dictated by themuch longer light-cross time t cross = R (1 + z ) /cδ , whichin the most extreme cases is of the order of 1–30 days.While (by construction) this timescale is in agreementwith the duration of the observed active phases (seeFigure 2, 3, 4), it is too long to explain the fasterspikes in the gamma-ray light curve and the observedspectral variability. A possibility to solve this problemis to abandon the one-zone framework and to assume,besides the large region, accounting for the long durationflare, other, much smaller sub-regions responsible for igh Energy flares of FSRQs Figure 8.
Gamma-ray spectra for period A (left panels) and periods B, C, D (right panels) for the most powerful HE emitters, and for 4C+38.41. Spectra for periods B, C, D have diamonds symbols, triangles, squares respectively (with black, red, and green colours respectivelyfor the online version). source period A ∆ t ph ∆ t (d) Prob(d) photons Prob. for period shape A =shape B (%) ∗∗∗∗ [0.2–10 GeV] B C D (%)PKS 1502+106 2009-05-06 05:20–2009-05-06 13:11 0.326(0.38) ∗ ± < ∗∗∗∗ ± ∗∗∗ ± ± ∗∗ < ∗ ± ± ∗∗ ∗ ± < Table 5
Integrations for period A and B as defined in the test for the most-powerful sources in the sample, and for 4C +38.41. We report alsoother integration periods (C,D) consecutive each other. Γ ph is the photon index estimated fitting data of period A with a powerlaw modelfrom 0.2 to 10 GeV. The last column is the estimate of probability that period A and B have the same spectral shape. ∗ Chanceprobability is evaluated both for a bunch of Gamma-rays within the integration time of period A, assuming as mean rate the rate of thewhole activity period, and for the same bunch of photons within the integration time and occurring during the whole activity period atHE. ∗ for PKS 1502+106, PKS 0805-07 and 4C +38.41 we enlarged the period A symmetrically on lower and higher temporal ends, tohave a duration of 0.38, 0.38 and 0.30 d, respectively. Photon index is reported in the energy range 0.2 – 10 GeV. ∗∗ photon index in thesame energy range for the whole HE activity period is also reported between brackets for 3C 454.3 and PKS 0805-07. ∗∗∗ For CTA 102there is a gap in the data, thence period C starts 5.08 d after the end of period B. ∗∗∗∗
To evaluate chance probability for CTA 102 we donot consider the whole activity period, but the period from 2012-09-21 21:36 to 2012-09-24 02:38, when the source was at the highest fluxin Gamma-ray.PowerlawSource F ( >
100 MeV) Γ ph TS(10 − ph cm s − )3C 454.3 222 ±
24 2.07 ± ±
37 2.05 ± >
100 MeV) Γ
LEph Γ HEph E break TS -2∆ L (10 − ph cm s − ) (GeV)3C 454.3 205 ±
24 1.80 ± ± ± ±
37 1.88 ± ± ± >
100 MeV) α β E break TS -2∆ L (10 − ph cm s − ) (GeV)3C 454.3 203 ±
24 1.55 ± ± ± ±
37 1.66 ± ± ± >
100 MeV) Γ
LEph Γ HEph cutoff TS -2∆ L (10 − ph cm s − ) (GeV)3C 454.3 203 ±
24 1.66 ± ± ±
37 1.88 ± ± Table 6
Fitting parameters for Gamma-ray spectra of CTA 102 and 3C 454.3 for period B. Fitting are performed with powerlaw, brokenpowerlaw, logparabola, powerlaw with exponential cutoff for energies above 200 MeV. ∆ L is the difference of the log likelihood of the fitwith respect to a single powerlaw fit. igh Energy flares of FSRQs slow-envelope emission in Giannios 2013). Therandomly oriented production of envelope-plasmoidstroughtout the whole blob mimics the randomly orientedelectron velocity distribution from the whole blob inleptonic models. Occasionally, much more powerfulevents produce small (compared to the whole blob)plasmoids whose emitted luminosity is comparable oreven larger than the envelope one, and oriented towardthe observer ( monster-plasmoids ). These events couldexplain the fast spikes in the light curve, and fastgamma-ray spectral variability that we showed for 4sources. Electron energy distribution is expected to besimilar for envelope- and monster-plasmoids.We just depicted a multi-zone scenario which, instead,we modeled in the framework of single-zone homoge-neous models, assuming large blob size ( ∼ / R diss ).We will show that the model we provided accountsfor the envelope-emission (assuming electrons acceler-ated throughout the whole blob), even if we showedfast-spikes during HE activity periods: i) the envelope-and monster-plasmoid will dissipate at the same R diss ,thence the γγ opacity and the Klein-Nishina argumentsapply simultaneously for the two-components of themodel, and the dissipation zone must be searched forat the outer edge of the BLR or beyond; ii) in ourSEDs we integrated gamma-ray data for long lastingperiods including the envelope emission, and the fastspikes which are expected to have a similar intensityto the envelope emission, and similar electron energydistribution; iii) the X-ray, NIR, and optical-UV datawhere never collected during HE fast spikes. So, atmost, we overestimated the gamma-ray spectra by afactor 2 for the envelope emission (because envelope- andmonster-plasmoid emission are expected to give similarintensities); iv) a correction in SED modeling to accountfor the fast-spikes will eventually cause the Comptondominance to be lowered for the envelope-emission, andthence the B/ Γ bulk to be raised of the same amount,and finally the product Γ bulk × R blob to be lowered by afactor ∼ R blob and/or Γ bulk (and thence of R diss ).A similar idea could be sketched for the turbulence sce-nario (Marscher 2013; Narayan & Piran 2012). Of coursethis scheme needs to be quantified and specified, a taskclearly beyond the aims of the present study. DISCUSSION AND CONCLUSION
Our systematic study enlarges the sample of FSRQsshowing evidence of gamma-ray flares occurring outsidethe BLR. The lack of the expected absorption signaturesdue to the interaction with the BLR radiation fieldconstraints the dissipation region outside the cavity ofthe BLR for all the sources, and outside the core ofthe BLR spherical shell for the 5 most powerful BroadLine emitters in our sample. This consideration doesnot depends on the emitting mechanism of FSRQs. Itis based on existing measurement of the Broad lineluminosities, and slightly depends on the BLR geometryand emission model (Liu & Bai 2006).In the framework of leptonic models for blazar emission,we showed that IC on BLR photon field is subject toKlein-Nishina suppression. The lack of this suppression,and of the γγ absorption in our spectra could only beexplained assuming a dissipation region at the outeredges of the BLR or further out.Furthermore, the leptonic SED modeling locates, insome cases, the radiating zone at parsec scale, assumingthe existence of the molecular torus for all the sources(Tristram et al. 2007; Cleary et al. 2007; Tristram et al.2014). If this is not the case, the solution must be putjust outside the broad line regions, because, in this hy-pothesis, the comoving intensity of the external radiationfield drops rapidly with the distance of the moving sourcefrom the SMBH.Once the region within the BLR cavity is excluded, thelocation of the blazar zone is a stable result of leptonicSED modeling for solutions found within the core of theBLR (between inner and outer shell radii of the BLR)because of the rapidly varying external radiation energydensity U blr from the inner to the outer radius of theBLR. The characterization of the blazar-zone as locatedoutside the BLR is also a stable result of leptonic model-ing, once the solution is found outside the external radiusof the BLR. But in this case, the refined localization de-pends on the chosen details in the modeling ( R blob /R diss ratio, the electron density, and Γ bulk ), because the exter-nal radiation field U IR is constant up to ∼ R IR .Here we assumed the emitting (envelope) region to en-compass the entire jet cross section, and a jet apertureangle θ j = r blob /d blob ≃ . ≈ / Γ bulk . Recent VLBIobservations (Clausen-Brown et al. 2013, Jorstad et al.2005, Pushkarev et al. 2009) indicate slightly lower val-ues, θ j ≈ . / Γ bulk . However, even these lower apertureangles would imply in several cases jet cross section ra-dius larger than ∼ cm (see Table 4), i.e. minimumvariability timescales of several days.As already noted, for several sources such longminimum variability timescales are inconsistent withthe much shorter variability timescale inferred from thegamma-ray light-curves. This problem cannot be easilysolved assuming that the region is located at smallerdistance from the SMBH, since the variability timescaleof < . × / (1 + z ) cm, which implies distancesin several cases smaller than the BLR radius. The sameproblem, i.e. the inconsistency between the observedvariability timescales and those expected from theestimate of the source dimension, has been encounteredin several other blazars, both BL Lacs and FSRQs (e.g.the BL Lac object Mrk 421, Gaidos et al. 1996, or PKS82155-304, Aharonian & Atohian 2007). For FSRQ themost extreme case is that of PKS 1222+216, whichshowed a doubling of the TeV flux in about 10 minutes(Aleksic et al. 2011).We point out that the location of the blazar zone inthis work has been achieved within the framework ofleptonic models, and we did not try other modeling suchas the spine-sheath or hadronic ones.We obtained periods of HE activity lasting from 1 d toabout a month. The HE activity periods coincide withactivity at lower energy. Within this period of activity wehave identified shorter periods with timescales of the or-der of 0.2–0.6 d characterized by brighter emission of HEgamma-rays. Focusing on these short periods, a ratherobvious consequence of our searching criterion is the rel-evant flux we derive above 10 GeV. We indeed evaluatedthat chance probability of our searching method is verylow for at least 2 flares (CTA 102 and PKS 0805-07) over4 sources for which we can try this study. For the othertwo flares that we examined in details (3C 454.3, andPKS 1502+106) the gamma-ray spectrum from 200 MeVto 10 GeV (i.e., below the energy threshold of our search-ing method) is rather hard. Thence the hard gamma-rayspectra obtained below 10 GeV are not biased results.They imply an energy spectral index <
1, suggesting thatthe emission derives from a fresh (i.e. not cooled) popu-lation of electrons injected/accelerated in the source.The spectral fit to gamma-ray spectra of periods Bfor the flares of CTA 102 and 3C 454.3 (Table 6), whencompared to periods A, reveals a break, or a curvature,or a cutoff. The limited statistics does not allow us todiscriminate among the different models. We note thatthe fit with a broken powerlaw gives ∆Γ ph ∼ ± ± ph = 0 .
5. The relatively long cooling time impliedby the data supports the view that the emission occursthrough the EC scattering in an environment with a lowenergy density of the target photons, such as that ofthe dusty torus. The spectra for period A of the othersources show similar trend (with the exception of PKS0805-07). The gamma-ray spectra of periods C, D givea further hint of the ongoing cooling. We note, however,that period A for CTA 102 corresponds to a fast flare inthe whole
FERMI–LAT energy band (not only at HE),as shown in figure 4. This consideration makes it hardto correlate periods A and B for this source in termsof slow–cooling, with period A encompassing the wholedevelopment of the gamma-ray flare.It’s worth mentioning the gamma-ray spectrum reportedin Tanaka et al. (2011) for PKS 1222+216, obtainedintegrating FERMI–LAT gamma-ray data for 8 daysaround the fast TeV flare studied in Aleksic et al.(2011). With this integration time, in the slow coolingdominated scenario, the cooled electron populationdominates the gamma-ray spectrum. Tanaka et al.(2011) reports ∆Γ ph = 0 . ± .
11 and they invoke theslow–cooling scenario for the TeV flare.Some of the gamma-ray spectra integrated on longperiods for 4C +38.41 and B2 1520+031 show peculiarstructures, reminiscent of absorption features. We performed the time-resolved study of the gamma-rayspectra of 4C +38.41 which showed the brighter HEflare. The analysis revealed, instead, that there was atleast one period in which the spectrum is flat (periodA), followed by periods in which the spectrum is soft(period B). As above this behaviour could be interpretedas due to the injection and subsequent cooling of high-energy electrons in the emitting region. The integratedspectra show both mechanisms in action, resultingin absorption–like features. It is not always possibleto time-resolve acceleration– from cooling–dominatedperiods. In fact, we reported in Table 1 that the HEgamma-rays are emitted on long timescales, possiblycausing the superpositions of accelerations and coolingphases.We did not attempt with this study to establish ifHE is a rare emission phase or not for FSRQs. Fromthe sub-sample of 10 FSRQs reported here, we triggered ∼
30 HE flares in total, ranging from 1 to 8 per sourcewithin 5.3 years of
FERMI–LAT operations. We willaddress this study on the whole
FERMI–LAT sample ina forthcoming paper.Summing up, our findings picture a scenario in which,during HE flares, the emission from FSRQ occurs in re-gions distant from the central BH. Further, we find ev-idence for the likely presence of substructures, respon-sible for the observed relatively rapid variability (fluxand spectrum). This framework is similar to that in-ferred from the observations of other FSRQ, especiallyduring the emission of TeV photons. Among the possi-ble theoretical scenarios advanced to explain such a phe-nomenology, we briefly discuss the reconnection modelof Giannios (2013) and that invoking turbulence in theflow (Marscher 2013; Narayan & Piran 2012). In the rel-ativistic magnetic reconnection scenario envisaged in Gi-annios (2013), long term high states are thought as the“envelope” of events of dissipation of magnetic field inthe jet through reconnection. During these events thereis the possibility that “monster plasmoids” form, whichaccounts for the fast flares. The discussion in Giannios(2013) was tailored to the case of the TeV flare of PKS1222+220, but the scenario should be applicable to thecases presented in our paper with small changes.Turbulence in the flow has been invoked as a possiblemechanism able to produce rapid flickering of thelight-curve both for FSRQ (Marscher 2013) and BL Lacobjects (Narayan & Piran 2012). In this scheme, smallcells of fluids characterized by fast turbulent speed canproduce the rapid flares, while the long-term emissionis produced by the larger active region encompassingthe whole jet. In the Marscher (2014) model, theactivity is triggered by the passage of the flow in are-collimation shock, thought to form at parsec scalewhen the jet internal pressure drops below that of aconfining medium. In this idea, the modulation ofthe injected power into the jet by the central engineaccounts for the long-term evolution of the emission,while the emission from single turbulent cells producethe rapid flares. Simulation along the lines of thosereported in Marscher (2014) could test if the scenariocan reproduce the phenomenology shown by LAT. igh Energy flares of FSRQs
E >
100 GeV) range (Ghisellini and Tavecchio2009), where the Imaging Cherenkov Telescopes (IACTs)systems operate. The detection by IACTs of the FS-RQs of our sample would allow to verify this hypoth-esis. The objects of this study have high redshift andthe absorption by the optical and IR Extragalactic Back-ground Light (EBL) has a non-negligible effect that mustbe taken into account. The unabsorbed fraction at 100GeV spans from a factor 0.75 for PKS 2345-1555 at z=0.6up to 0.1 at z=2. EBL absorption is more severe at 300GeV, where most of the present generation of Cherenkovtelescopes (MAGIC, HESS and VERITAS) have theirhighest sensitivity. At 300 GeV the unabsorbed fractionranges from a factor 0.1 at z=0.6 up to 10 − at z=2.Therefore the possibility of detection in the VHE rangerelies on one side on instruments with the lowest energythreshold such as MAGIC (Aleksic et al. 2012), and onthe other side on the high sensitivity of the Cherenkovtelescopes of future generation, such as the CherenkovTelescope Array (CTA, Actis et al. 2011).Assuming a sensitivity of ∼ − erg/cm /s at 100GeV for few hours of observations by present IACTs(Aleksic et al. 2012) and an unabsorbed fraction of ∼ . − .
01, the extrapolation of the spectra shown in fig-ure 8 shows that some of the sources with the hardestand highest flux, e.g. CTA 102, 3C 454.3, PMN J2355-1555, could be detected during a flaring episode, or couldyield a sound upper limit constraining the cut-off. Onthe other hand the fast drop of the flux at higher en-ergies would reduce the possibility to measure preciselythe spectrum at few hundreds of GeV, which is the neces-sary condition to disentangle the absorption of EBL fromthe internal absorption due to the IR radiation field. Toperform this measurement the existing IACTs should re-strict to the nearest sources, while most distant ones willbe the targets of the forthcoming CTA, which promisesa sensitivity up to an order of magnitude better than thepresent generation of Cherenkov telescopes.
ACKNOWLEDGEMENTS
L.P. and F.T. acknowledge partial financial contribu-tion from grant PRIN-INAF-2011. SMARTS observa-tions of LAT monitored blazars are supported by YaleUniversity and Fermi GI grant NNX 12AP15G, and theSMARTS blazar monitoring program is carried out byC.M.U., E.M., Michelle Buxton, Imran Hasan, J. I.,Charles Bailyn and Paolo Coppi. C.B., M.B. and theSMARTS 1.3m observing queue receive support fromNSF grant AST-0707627. We thank the Swift teamfor the ToO observations. We acknowledge all agen-cies and Institutes supporting the Fermi-LAT operationsand the Scientific analysis tools. This research has madeuse of data obtained from the Chandra Data Archiveand the Chandra Source Catalog, and software providedby the Chandra X-ray Center (CXC) in the applicationpackages CIAO, ChIPS, and Sherpa. This publicationmakes use of data products from the Two Micron AllSky Survey, which is a joint project of the University ofMassachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by theNational Aeronautics and Space Administration and theNational Science Foundation.REFERENCES
Abdo , A. A. et al., 2010, Nature, 463, 919;Abdo, A. A., et al., 2011, ApJL, 733, 26;Ackermann, M. et al., 2013, ApJ, 765, 54;Actis, M. et al., 2011, ExA, 32, 193;Agudo I. et al., 2011, ApJL, 726, 13;Agudo I. et al., 2011, ApJL, 735, 10;Aharonian, F. & Atoyan, A. M., 1981, AP&SS, 79, 321;Aharonian, F. et al., 2007, ApJL, 664, 71;Aleksic, J. et al., 2011, ApJL, 730, 8;Aleksic, J. et al., 2012, Astrop.Phys, 35, 435;Arbeiter, C., Pohl, M., Schlickeiser, R., 2002, A&A, 386, 415;Atwood, W. B. et al., 2009, ApJ, 697, 1071;Baldwin, J. A. and Netzer, H., 1978, ApJ, 226, 1;Begelman, M. C., Fabian, A. C.; Rees, M. J., 2008, MNRAS, 384,L19;Bentz, M. C., et al., 2009, ApJ, 705, 199;Blazejowski, M., Sikora, M., Moderski, R., Madejski, G. M., 2000,ApJ, 545, 107;Boettcher, M. et al., 2013, ApJ, 768, 54;Bonning, E. et al., 2012, ApJ,756,13;Bonnoli, G. et al., 2011, MNRAS, 410, 368;Bromberg, O. and Levinson A., 2009, ApJ 699, 1274;Burrows, D. N., et al., 2005, Space Sci. Rev., 120, 165;Buxton, M. M., et al.,2012, AJ, 143, 130;Cleary, K. et al., 2007, ApJ, 660, 117;Celotti, A., Padovani, P. and Ghisellini G., 1997, MNRAS, 286,415;Dermer, C. D.& Schlickeiser, R.,1993, ApJ, 416, 458;Dermer, C. D., Sturner, S. J., Schlickeiser, R.,1997, ApJS, 109,103;Dotson, A. et al., 2012, ApJL, 785, 15;Finke, J. D. & Dermer, C. D., 2010, ApJL, 714, 303;Fitzpatrick, E., L., 1999, PASP, 111, 63;Francis, P. et al., 1991, ApJ, 373, 465;Fruscione, A., et al., 2006, SPIE, 6270;Gaidos, J. A., et al., 1996, Nature, 383, 319;Garmire, G.P., et al., 2003, Proc. of the SPIE, 4851, 28;Gehrels N. et al., 2004, ApJ, 611, 1005;Ghisellini, G. and Tavecchio, F., 2009, MNRAS, 397, 985;Ghisellini, G. et al., 2010, MNRAS, 405, 387;Ghisellini, G. et al., 2013, MNRAS, 432, L66;Giannios, D., 2013, MNRAS, 431, 355;Hayashida, M. et al., 2012, ApJ, 754, 114;Holder, J., 2012, Astropart. Phys., 39–40, 61;Ikejiri, Y, et al., 2011, PASJ, 63, 639;Isler, J., et al, 2013, ApJ, 779, 100;Jorstad S. G. et al., 2010, ApJ, 715, 362;Jorstad S. G. et al., 2013, ApJ, 773, 147;Kalberla, P. M. W., et al., 2005, A&A, 440, 775;Landolt, A. U., 1992, AJ, 194, 340L;L´eon-Tavares, j., et al., 2013, ApJL, 763, 36;Liu, H. T. and Bai J. M., 2006, ApJL, 653, 1089;Liu, H. T., Bai J. M., Ma, L., 2008, ApJL, 688, 148;Maraschi, L., Ghisellini, G., and Celotti, A., 1992, ApJL, 397, 5;Marscher, A. P., and Bloom, S. D., 1992, Proceedings of TheCompton Observatory Science Workshop, p. 346;Marscher, A. P., 2013, 2012 Fermi Symposium proceedings -eConf C121028, arXiv:1304.2064;Marscher, A. P., 2014, ApJ, in press;Mattox, J. R. et al., 1996, ApJ, 461, 396;Morrison, R. & McCammon, D., 1983, ApJ, 270, 119;Nalewajko K. et al., 2012, ApJ, 760, 69;Narayan, R., Piran, T., 2012, MNRAS, 420, 604;Nolan, P. L. et al., 2012, ApJS, 199, 31;Pacciani, L. et al., 2010, ApJL, 716, 170;Pacciani L. et al., 2012, MNRAS, 425, 2015;Persson, S. E., et al. 1998, AJ, 116, 2475;Pian, E., Falomo, R., and Treves, A., 2005, MNRAS, 361, 919;Poutanen, J and Stern, B, 2010, ApJL, 717, 118;0