The blazar's divide and the properties of Fermi blazars
aa r X i v : . [ a s t r o - ph . C O ] D ec **FULL TITLE**ASP Conference Series, Vol. **VOLUME**, **YEAR OF PUBLICATION****NAMES OF EDITORS** The blazar’s divide and the properties of Fermi blazars
Gabriele Ghisellini
INAF – Osservatorio Astronomico di Brera
Abstract.
The LAT instrument, onboard the
Fermi satellite, in its first threemonths of operation detected more than 100 blazars at more than the 10 σ level.This is already a great improvement with respect to its predecessor, the instru-ment EGRET onboard the Compton Gamma Ray Observatory . Observationally,the new detections follow and confirm the so–called blazar sequence, relating thebolometric observed non–thermal luminosity to the overall shape of the spectralenergy distribution. We have studied the general physical properties of all thesebright
Fermi blazars, and found that their jets are matter dominated, carrying alarge total power that correlates with the luminosity of their accretion disks. Wesuggest that the division of blazars into the two subclasses of broad line emittingobjects (Flat Spectrum Radio Quasars) and line–less BL Lacs is a consequenceof a rather drastic change of the accretion mode, becoming radiatively inefficientbelow a critical value of the accretion rate, corresponding to a disk luminosity of ∼
1. The Fermi blazar sequence
The Large Area Telescope (LAT) on board the
Fermi Gamma Ray Space Tele-scope (Fermi) revealed in the first three months of operation 57 flat spectrumradio quasars (FSRQs), 42 BL Lac objects, and 5 blazars with uncertain classi-fication (Abdo et al. 2009a, hereafter A09; Foschini et al., these proceedings).Ghisellini et al. (2009a) showed that the spectral index α γ correlates withthe γ –ray luminosity L γ and that BL Lacs and FSRQs occupy different regionsof the α γ − L γ plane. There is a rather well defined boundary between BL Lacsand FSRQs as shown in Fig. 1. Empty circles and squares correspond to BLLac objects and FSRQs, respectively, while filled symbols indicate sources alsodetected in the TeV band. This correlation holds despite the large amplitudevariability of blazars, especially at high energies. Examples of how variabilitycan change the position of single sources in the α γ – L γ plane are shown in Fig.1 by the segments connecting the locations of specific sources at different times.Note that several sources “move” orthogonally to the correlation defined by theensemble of sources, i.e. they become harder when brighter (with the exceptionof 3C 454.3). The high and the low γ –ray states of single sources can be dra-matically different, and this implies that the distribution in luminosity withineach blazar class is largely affected by the variability of the sources.The exceptional case of BL Lac itself is shown in the right panel of Fig. 1.Its γ –ray luminosity varied by two orders of magnitude. Moreover the slope ofthe high energy emission varied from α γ ∼ . Figure 1.
Left panel: energy spectral index vs γ –ray luminosity for allblazars in the list of A09. Empty squares and cirles are BL Lacs and FSRQs,respectively. Filled symbols correspond to sources already detected in the TeVband. For a few blazars we show the observed range of γ –ray luminosity andspectral index, using past EGRET or AGILE observations. This is indicatedby a segment. The grey stripes at about L γ = 10 erg s − mark the dividebetween BL Lac objects and FSRQs. Right panel: the SEDs of BL Lac itselfillustrates the dramatic variability of blazars, especially at high energies.
Fermi –observed value of α γ ∼ . γ –ray state.Fig. 1 shows that BL Lacs and FSRQs separate at L γ ∼ erg s − , asindicated by the grey stripes. Furthermore there is a less clear-cut separation inspectral indices, occurring at α γ ≃ . L γ and flatter α γ . FSRQs, instead, peak at lower frequencies, andthe peak of their high energy emission (dominating their power output) is below100 MeV. In the LAT energy range they are steep, but powerful. Therefore theleft panel of Fig. 1 represents the γ –ray selected version of the blazar sequence. The other intriguing feature of Fig. 1 is the existence of a γ –ray luminositydividing BL Lacs from FSRQs. We have proposed that this is a consequenceof the change of the accretion regime, becoming radiatively inefficient below acritical disk luminosity, in units of Eddington. This reflects also in a critical(dividing) luminosity of the observed beamed emission, rather well tracked by L γ . To understand why in a simple way, assume that most of the bright blazarsdetected by the 3–months LAT survey have approximately the same black holemass. Assume also that the largest L γ correspond to jets with the largest powercarried in bulk motion of particles and fields. Finally, assume that the jet powerand the accretion rate are related. These three assumptions, that will be betterjustified later, imply that the most luminous blazars have the most powerful jetsand are accreting near Eddington. These are the FSRQs with L γ ∼ erg s − .The dividing L γ is a factor 100 less, so it should correspond to disks emittingat the 1% level of the Eddington level. Below this value we find BL Lacs, thathave no (or very weak) broad emission lines. If the disk becomes radiativelyinefficient at L d < − L Edd the broad line region receives a much decreasedionizing luminosity, and the lines become much weaker. The radiation energydensity of the lines becomes unimportant for the formation of the high energycontinuum (there are much less seed photons for the Inverse Compton process),implying: i) a reduced “Compton dominance” (i.e. the ratio of the Compton tosynchrotron luminosities); ii) less severe cooling for the emitting electrons, thatcan then achieve larger energies and then iii) a shift of both the synchrotron andthe Inverse Compton peak frequencies to larger values.According to this interpretation, it is the accretion mode that determinesthe “look” of the radiation produced by the jet, not a property of the jet itself.
2. General properties of the
Fermi blazars
We (Ghisellini et al. 2009b, hereafter G09) have studied and modelled almostall the
Fermi blazars detected in its first 3–months of operation. We excludedobjects without known redshift and a few with very few data available (insuffi-cient to construct a meaningful SED). In total, we studied 85 blazars (includingone Narrow Line Seyfert 1, see Abdo et al. 2009b, 2009c).Many of them were observed by the X–ray (XRT) and UV–optical (UVOT)telescopes onboard
Swift , and this was of great help in characterizing their SED.What was an exception in the EGRET era (and a result of huge efforts bymany people involved in multi-wavelength campaigns) is now routine. Fig. 2shows, for illustration, the SED of the FSRQ 2141+175. As can be seen, thesynchrotron spectrum peaks at very low frequencies, and the flux produced bythe accretion disk is well visible. In this case the data are good enough to fitthe optical–UV flux with a standard Shakura –Sunjaev (1973) accretion disk. Inturn, this allows to estimate the mass of the black hole and the accretion rate.For these FSRQs (with good optical–UV coverage) we can then study in areliable way the connection between the jet power and the accretion luminosity,also in units of the Eddington one. A first result is shown in the right panelof Fig. 2: the observed γ –ray luminosity L γ is related with the accretion diskluminosity L d . Note that for BL Lacs we have only an upper limit on L d (shownby the triangles).The grey stripe shows a linear relation above L d = 10 erg s − (withscatter, blazars can vary their non–thermal luminosity even by one or two orderof magnitude). This is appropriate for all Fermi
FSRQs. Below this criticalluminosity value there are only BL Lacs, and the grey stripe becomes L γ ∝ L / . This corresponds to assume that the jet power (and then the observedluminosity, for aligned sources) scales always as the accretion rate ˙ M , while thedisk luminosity, which is linear with ˙ M at high rates, scales as L d ∝ ˙ M below L d = 10 erg s − , so that L γ ∝ ˙ M ∝ L / (see Ghisellini & Tavecchio 2008).Note that, for a 10 M ⊙ black hole, this “dividing” luminosity corresponds to Figure 2.
Left panel: the SED of the FSRQ 2141+175 and the fitting model.We label the different components. Note how the synchrotron spectrum,peaking at low frequencies, makes the accretion disk flux “naked”. In thiscases the data are good enough for estimating both the black hole mass andthe accretion rate.
Right panel: the γ –ray luminosity as a function of theaccretion disk luminosity for Fermi blazar of the A09 sample. Red filledcircles are FSRQs, triangles are for BL Lacs with only an upper limit for theirdisk luminosity. The grey band corresponds to what expected if the FSRQswith L d > erg s − have standard accretion disks with L d > − L Edd and L γ ∝ L jet ∝ ˙ M ∝ L d , while BL Lac have “ADAF” like accretion with L d ∝ ˙ M . In this case their L γ ∝ L jet ∝ ˙ M ∝ L / . L d ∼ − L Edd . In the near future, when blazars with black holes of smallermasses will be observed, this clear–cut division will become fuzzier.
Several attempts have been done in the past to find the jet power and theaccretion disk luminosity in blazars and radio–loud objects in general (start-ing from Rawlings & Saunders 1991; Celotti et al. 1997; Cavaliere & D’Elia2002; Maraschi & Tavecchio 2003; Padovani & Landt 2003; Sambruna et al.2006; Allen et al. 2006; Celotti & Ghisellini 2008; Ghisellini & Tavecchio 2008;Kataoka et al. 2008). These works found large jet powers, often larger thanthe luminosity produced by the disk. However, there were two caveats: the firstconcerns the low energy end of the emitting particle distribution, where most ofthe electrons are. To the end of estimating the jet power, this is a crucial quan-tity if one assumes that there is one proton per electron (and this assumption isthe second caveat). But in powerful sources, for which the radiative cooling issevere, even low energy electrons cool in a light crossing time, leaving much lessuncertainty about the presence of low energy electrons, distributed in energy ∝ γ − .Sikora & Madehski (2000) and Celotti & Ghisellini (2008) argued thatelectron–positron pairs cannot be dynamically important, corresponding to alimit of a few pairs per proton. This issue (discussed at length in Celotti &Ghisellini 2008 and G09) can be understood looking at the left panel of Fig. 3, Figure 3.
Left panel:
The distribution of jet powers in the form of bulkmotion of cold protons ( P p ), emitting electrons ( P e ), magnetic field ( P B ) andradiation ( P r ). The bottom panel shows the distribution of disk luminosities L d . Grey shaded areas correspond to BL Lacs. Right panel: the total jetpower P jet vs the accretion disk luminosity L d . To estimate P jet , we haveassumed one proton per emitting electron. showing the histograms of the different forms of power carried by the jet. Theshaded areas correspond to BL Lacs. The crucial power, that is almost model–independent, is the power P r spent by the jet to produce its radiation. It is sim-ply the observed, beamed, bolometric luminosity multiplied by Γ /δ ∼ /δ .For FSRQs, the distribution of P r extends to larger values than the distributionof P e , the power carried by the jet in the form of emitting electrons. So theradiation we see cannot originate by electrons (or pairs) only. Can it come fromthe Poynting flux (by e.g. reconnection)? The distribution of P B is at slightlysmaller values than the distribution of P r , indicating that the Poynting flux can-not be at the origin of the radiation we see. As described in Celotti & Ghisellini(2008), this is a direct consequence of the large values of the Compton domi-nance (i.e. the ratio of the Compton to the synchrotron luminosity is small),since this limits the value of the magnetic field.To justify the power that the jet carries in radiation we are forced to considerprotons. If there is one proton per electron (i.e. no pairs), then P p for FSRQs isa factor ∼ P r , meaning an efficiency of 1–10% for the jet toconvert its bulk kinetic motion into radiation. This is reasonable: most of thejet power in FSRQs goes to form and energize the large radio structures, andnot into radiation.We then conclude that jets should be matter dominated, at least at thescale (hundreds of Schwarzschild radii from the black hole) where most of theirluminosity is produced. The bottom left panel of Fig. 3 shows the distributionof the disk luminosities. In this case the shaded area corresponds to upperlimit for BL Lac objects, and not to actual values. This L d distribution lies atintermediate values between P r and P p . Figure 4.
Left panel:
The average SED for BL Lacs (blue long dashed), andFSRQs (red solid) detected in the 3–months
Fermi survey, both in νF ν (top)and νL ν (bottom). Right panel: sketch illustrating P jet and L d as a functionof ˙ M / ˙ M Edd . It is assumed that the jet power always scales linearly with˙ M , while accretion rates below a critical value produce radiatively inefficientaccretion disks. In this case the object looks like a BL Lac (if aligned) or aFR I (if misaligned). The gray stripes indicate the critical ˙ M ˙ M Edd ∼ . L d /L Edd ∼ − . The right panel of Fig. 3 shows the total jet power P jet ≡ P p + P e + P B as a func-tion of the thermal disk luminosity. Arrows corresponds to BL Lacs for whichonly an upper limit on L d could be derived. The different symbols correspondsto blazars of different γ –ray luminosities, and one can see that L γ correlatesboth with P jet and L d .As discussed in G09, there is a significant correlation between P jet and L d for FSRQs, which remains highly significant even when excluding the commonredshift dependence. The slope of this correlation is consistent with being linear,and P jet is larger than L d for almost all sources, and must be much larger forBL Lacs.
3. Discussion
The first results of
Fermi confirm the idea that blazars form a sequence. Fig. 4shows the average model SED constructed for BL Lacs and FSRQs by averagingthe parameters obtained by fitting the sources one by one. It shows both the νF ν and νL ν representations. In the LAT energy range the average BL Lac has a flat( α γ <
1) spectrum, while FSRQs are steeper than unity. This is associated withthe larger Compton dominance in FSRQs, in turn associated with the presence ofexternal seed photons for the scattering process. Also shown (short dashed line)is the averaged disk spectrum of FSRQs, together with the spectrum producedby the X–ray corona and the re–emission of part of the disk optical–UV radiationby an absorbing torus.
The relation between P jet and L d strongly suggests that P jet ≈ ˙ M c (1)while the accretion disk luminosity L d ∼ . M c ˙ M ≥ ˙ M c L d ∼ . ˙ M ˙ M c ! c ˙ M ≤ ˙ M c (2)where the L d ∝ ˙ M dependence is appropriate for advection dominated accre-tion flows (ADAF, e.g. Narayan, Garcia & McClintock 1997). Radiatively ineffi-cient disk may also correspond to adiabatic inflow–outflows (ADIOS, Blandford& Begelman 1999) or a convection dominated flows (CDAF, Narayan, Igumen-shchev & Abramowicz 2000). At the other extreme of accretion rates (i.e. nearlyEddington) the density close to the hole may correspond to scattering opticaldepths larger than unity, trapping a fraction of photons and making them tobe swallowed by the black hole before escaping. Fig. 4 sketches the expectedbehavior of both P jet and L d as a function of ˙ M / ˙ M Edd , where ˙ M Edd ≡ L Edd /c .According to this scenario all radio loud objects of all powers have a jet powerproportional to ˙ M , irrespective of the accretion regimes. These instead affect theemitted disk luminosity L d at both ends of the ˙ M range. Below L d ∼ − L Edd ,corresponding to ˙ M ∼ . M Edd , the disk becomes radiatively inefficient, its ion-izing radiation is greatly reduced, as are the broad lines. These objects are BLLacs if pointing in our direction, and FR I radio–galaxies if they point some-where else. Above the critical ˙ M , jet powers and disk luminosities scale linearly,producing a FSRQ or a powerful FR II. The fact that the jet power correlates with L d , but tends to be larger thanthat, leads us to ask: What is the source of the power of the jet? Is it onlythe gravitational energy of the accreting matter or do we necessarily need alsothe rotational energy of a spinning black hole? In G09 we have discussed twopossible alternatives.The first possibility stems out from the idea by Jolley et al. (2009) andJolley & Kuncic (2008), who propose that, in jetted sources, a sizeable fractionof the accretion power goes to power the jet. As a result, the remaining powerfor the disk luminosity is less than usually estimated by setting L d = η ˙ M in c ,with η ∼ . L d is larger than what we have estimated. Also the total accretion poweris larger, and it is sufficient to explain the derived large jet powers.The second alternative is the more standard Blandford & Znajek (1978)scenario, in which jets are powered by the rotational energy of the spinning blackhole. In this scenario the correlation between jet power and disk luminosityis provided by the requirement of having a sufficiently strong magnetic field,anchored to the disk, to tap the spin energy of the hole. If the magnetic energydensity scales with the disk density, in turn linked to the accretion rate, then P jet should scale as L d . Acknowledgments.
I gratefully thank my collaborators A. Celotti, L. Fos-chini, G. Ghirlanda, L. Maraschi and F. Tavecchio. I thank the grant PRIN–INAF 2007 for partial funding.