High redshift Fermi blazars
G. Ghisellini, G. Tagliaferri, L. Foschini, G. Ghirlanda, F. Tavecchio, R. Della Ceca, F. Haardt, M. Volonteri, N. Gehrels
aa r X i v : . [ a s t r o - ph . C O ] S e p Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 12 September 2018 (MN L A TEX style file v2.2)
High redshift Fermi blazars
G. Ghisellini ⋆ , G. Tagliaferri , L. Foschini , G. Ghirlanda , F. Tavecchio ,R. Della Ceca , F. Haardt , , M. Volonteri , N. Gehrels INAF – Osservatorio Astronomico di Brera, Via Bianchi 46, I–23807 Merate, Italy INAF – Osservatorio Astronomico di Brera, Via Brera 28, I–20100 Milano, Italy Universit´a dell’Insubria, Dipartimento di Fisica e Matematica, Via Valleggio 11, I–22100 Como, Italy; INFN, Sezione di Milano–Bicocca, I–20126 Milano, Italy; Astronomy Department, University of Michigan, Ann Arbor, MI 48109 NASA Goddard Space Flight Center, Greenbelt, USA.
12 September 2018
ABSTRACT
With the release of the first year
Fermi catalogue, the number of blazars detected above 100MeV lying at high redshift has been largely increased. There are 28 blazars at z > inthe “clean” sample. All of them are Flat Spectrum Radio Quasars (FSRQs). We study andmodel their overall spectral energy distribution in order to find the physical parameters ofthe jet emitting region, and for all of them we estimate their black hole masses and accretionrates. We then compare the jet with the accretion disk properties, setting these sources in thebroader context of all the other bright γ –ray or hard X–ray blazars. We confirm that the jetpower correlates with the accretion luminosity. We find that the high energy emission peakshifts to smaller frequencies as the observed luminosity increases, according to the blazarsequence, making the hard X–ray band the most suitable for searching the most luminous anddistant blazars. Key words:
BL Lacertae objects: general — quasars: general — radiation mechanisms: non–thermal — gamma-rays: theory — X-rays: general
The Large Area Telescope (LAT) onboard the
Fermi satellite de-tected, after 11 months of all sky survey, more than 1,400 sources,presented in Abdo et al. (2010a), with roughly half of them beingBL Lac objects or Flat Spectrum Radio Quasars (FSRQs) (Abdoet al. 2010b, hereafter A10) and a few radio–galaxies, starbursts,galaxies, and Narrow Line Seyfert 1 galaxies. The correspondingcatalog of AGNs at high Galactic latitude ( | b | > ◦ ) is calledFirst LAT AGN Catalog (1LAC). BL Lacs and FSRQs are approx-imately present in equal number.With respect to the previous sample (LAT Bright AGN Sam-ple, hereafter LBAS), constructed after 3 months of survey (Abdoet al. 2009; hereafter A09), the number of detected blazars is about7 times larger, as a result of the lower limiting sensitivity, ob-tained with the longer exposure and the smaller required signifi-cance (from 10 σ of the first 3 months to the current 4 σ level).Correspondingly, also the number of high redshift blazars detectedin γ –rays increased: in the LBAS there were 5 blazars at z > (and none at z > ), while in the 1LAC catalogue there are 28sources at z > (and 2 at z > ).The increased number of high redshift γ –ray blazars allows ⋆ Email: [email protected] us to characterize them in a meaningful way, through their SpectralEnergy Distributions (SEDs) and their modelling. Indeed, the cov-erage at other frequencies (besides the
Fermi /LAT band) includesobservations by the
Swift satellite for all sources both in the optical–UV band (through the Optical–UV Telescope UVOT) and the softX–ray band (0.3–10 keV, through the X–Ray Telescope XRT).It is also interesting to compare the properties of the high red-shift blazars detected in γ –rays with the high– z blazars detectedin hard X–rays by the Burst Alert Telescope (BAT) instrument on-board the Swift satellite. All blazars at z > are FSRQs, so, upto now, high redshift “blazars” coincide with high redshift FSRQs,since no BL Lac objects with a measured redshifts z > has beendetected so far. There are 10 FSRQs at z > and 5 at z > inthe 3–year BAT all sky survey presented by Ajello et al. (2009),that have been studied in Ghisellini et al. (2010a, hereafter G10).The BAT and the LAT samples of high redshift blazars are ratherwell defined, since the sky coverage is quasi–uniform (excludingthe Galactic plane) and we can consider these samples as flux lim-ited. The main aims of the present paper are then to characterizethe properties of blazars detected at high energies at redshift greaterthan 2 and to see if we can understand the differences (if any) be-tween the blazars detected in the two bands ( γ –rays and hard X–rays). In G10, in fact, we suggested that the best way to select the c (cid:13) G. Ghisellini et al. most powerful blazars at large redshifts is through a survey in thehard X–ray band, rather than in the γ –ray one, but this was basedon small numbers. None of the 10 BAT blazars at z > is presentin the LBAS catalogue, and only 4 of them are in the 1LAC sample,i.e. have been detected after 11 months of survey by the Fermi /LATinstrument. In G10 we explained this through a change of the av-erage SED when the bolometric luminosity changes: by increasingit, the high energy hump of the SED peaks at smaller frequencies,in agreement with the blazar sequence as put forward by Fossati etal. (1998) and Donato et al. (2001), and interpreted in Ghisellini etal. (1998).We anticipate that our earlier suggestion remains valid, withimportant implications on the planned future hard X–ray surveymissions, such as
EXIST .In this paper we use a cosmology with h = Ω Λ = 0 . and Ω M = 0 . , and use the notation Q = 10 X Q x in cgs units (exceptfor the black hole masses, measured in solar mass units). We consider all blazars detected during the first year all–sky surveyof
Fermi and classified as “clean” in the catalogue of A10. Theseare all the blazars with | b | > ◦ , detected at more than the 4 σ level whose identification is secure and unique. In total the 1LACclean sample contains 599 sources (A10), of which 248 are FSRQs,all with a measured redshift, and 275 BL Lacs (116 with the red-shift measured). Among these, we selected the 27 blazars at z > as listed and classified by A10, plus an additional source, SWIFTJ1656.3–3302 ( z = 2 . ), that Ghirlanda et al. (2010) recently clas-sified as FSRQ among the unidentified 1LAC sources.Five of these were already present in the LBAS list, i.e. theblazars detected at more than the 10 σ level during the first 3 monthsof Fermi survey (A09). Four additional sources are present in the3–years survey of
Swift /BAT (A09) and they too have been studiedin G10.Table 1 lists all sources: the top 19 blazars are studied in thispaper, while the bottom 9 are the sources already present either inthe BAT or LBAS samples. In this paper we present the spectralenergy distributions (SED) and the modelling for the “new” ones,i.e. blazars not present in our previous study (G10).
For all blazars studied in this paper there are
Swift observations.Even when they were performed during the 11 months of the 1LACsurvey, they correspond to a “snapshot” of the optical–X–ray stateof the source, while the γ –ray data are an average over the 11months. Given the very rapid blazar variability, the SEDs con-structed in this way should be considered, in all cases, not simulta-neous (but the Swift
UVOT and XRT data are indeed simultaneous).The data were screened, cleaned and analysed with the soft-ware package HEASOFT v. 6.8, with the calibration database up-dated to 30 December 2009. The XRT data were processed with thestandard procedures (
XRTPIPELINE v.0.12.4 ). All sourceswere observed in photon counting (PC) mode and grade 0–12 (sin-gle to quadruple pixel) were selected. The channels with energiesbelow 0.2 keV and above 10 keV were excluded from the fit andthe spectra were rebinned in energy so to have at least 20–30 countsper bin in order to apply the χ test. When there are no sufficientcounts, then we applied the likelihood statistic in the form reported Name Alias z log L γ Ref0106+01 4C+01.02 2.107 48.7 00157–4614 PMN 2.287 47.9 00242+23 B2 2.243 48.0 00322+222 TXS 2.066 48.0 00420+022 PKS 2.277 47.9 00451–28 PKS 2.56 48.7 00458–02 PKS 2.291 48.1 00601–70 PKS 2.409 48.3 00625–5438 PMN 2.051 48.2 00907+230 TXS 2.661 48.3 00908+416 TXS 2.563 47.7 01149–084 PKS 2.367 47.7 01343+451 TXS 2.534 48.4 01344–1723 PMN 2.49 48.5 01537+2754 [WB92] 2.19 47.6 01656.3–3302 Swift 2.4 48.1 01959–4246 PMN 2.174 48.0 02118+188 TXS 2.18 48.1 02135–5006 PMN 2.181 48.1 00227–369 PKS 2.115 48.1 G090347–211 PKS 2.944 49.1 G090528+134 PKS 2.07 48.8 G090537-286 PKS 3.104 48.4 G100743+259 TXS 2.979 48.6 G100805+6144 CGRaBS 3.033 48.4 G100836+710 4C+71.07 2.218 48.5 G100917+449 TXS 2.19 48.4 G091329-049 PKS 2.15 48.5 G09
Table 1.
List of blazars at z > . The upper part of the table reports theblazars studied in this paper (denoted by “0” in the last column); while thebottom part reports blazars studied previously: in Ghisellini et al. (2010;G10) (blazars with z > detected by Swift /BAT); and in Ghisellini,Tavecchio & Ghirlanda (2009; G09) (blazars with z > in LBAS with L γ > erg s − ). by Cash (1979). Each spectrum was analysed through XSPEC v.12.5.1n with an absorbed power law model with a fixed Galacticcolumn density as measured by Kalberla et al. (2005). The com-puted errors represent the 90% confidence interval on the spectralparameters. Tab. 2 reports the log of the observations and the bestfit results of the X–ray data with a simple power law model. TheX–ray spectra displayed in the SED have been properly rebinned toensure the best visualization.UVOT (Roming et al. 2005) source counts were extractedfrom a circular region ′′ − sized centred on the source position,while the background was extracted from a larger circular nearbysource–free region. Data were integrated with the uvotimsum task and then analysed by using the uvotsource task. The ob-served magnitudes have been dereddened according to the formulaeby Cardelli et al. (1989) and converted into fluxes by using standardformulae and zero points from Poole et al. (2008). Tab. 3 lists theobserved magnitudes in the 6 filters of UVOT. To model the SEDs of the blazars in this sample we used the samemodel used in G10. It is a one–zone, leptonic model, fully dis-cussed in Ghisellini & Tavecchio (2009). In that paper we em- c (cid:13) , 000–000 igh redshift Fermi blazars Name Obs Date Exp N H Γ F obs0 . − χ /Cash d.o.f.DD/MM/YYYY [ks] [ cm − ] [ − erg cm − s − ]0106+01 2007–2008 a . ± . . ± . . ± . . ± . —/1.6 40242+23 11/03/2010 1.2 9.46 . ± . . ± . —/3.7 60322+222 25/03/2007 2.8 8.90 . ± . ± —/48.99 550420+022 27/03/2010 4.2 10.7 . ± . . ± . —/6.7 90451–28 27/10/2009 6.6 2.00 . ± . ± b . ± .
08 14 . ± . c . ± . . ± . . ± . . ± . —/37.8 330907+230 30/12/2009 7.8 4.83 . ± . . ± . —/5.7 130908+416 2010 d . ± . . ± . —/59.5 121149–084 10/11/2009 1.0 4.75 . ± . . ± . —/0.85 21343+451 2009 e . ± . . ± . —/36.26 411344–1723 f . . ± . —/— —1539+2744 17/03/2010 7.1 2.81 . ± . . ± . —/24.6 191656–3302 2006 g . ± . ± . ± . . ± . —/32.5 302118+188 2009 h . ± . . ± . . ± . . ± . —/12.4 11 Table 2.
Summary of XRT observations. The observation date column indicates the date of a single snapshot or the yearsduring which multiple snapshots were performed. The corresponding note reports the complete set of observations integrated.The column “Exp” indicates the effective exposure in ks, while N H is the Galactic absorption column in units of [ cm − ]from Kalberla et al. (2005). Γ is the photon index of the power law model [ F ( E ) ∝ E − Γ ], F obs0 . − is the observed(absorbed) flux. The two last columns indicate the results of the statistical analysis: the last column contains the degrees offreedom, while the last but one column displays the reduced χ or the value of the likelihood (Cash 1979), in the case therewere no sufficient counts to apply the χ test. a sum of observations of: 02/07/2007, 10/01/2008, 16/02/2008, 16/08/2009. b sum of observations of: 22/03/2007, 10/04/2007, 08/08/2007, 10/08/2007, 13/01/2008, 20/04/2008, 22/04/2008,26/10/2008, 22/04/2009. c sum of observations of: 26/12/2008, 08/01/2009. d sum of observations of: 21/02/2010, 25/02/2010. e sum of observations of: 06/03/2009, 01/10/2009, 04/10/2009. f Flux derived by using WebPIMMS with a rate of (5 ± × − c/s and fixed parameters. g sum of observations of: 09/06/2006, 13/06/2006. h sum of observations of: 08/01/2009, 13/01/2009.Source A V v b u uvw uvm uvw . ± .
12 18 . ± .
07 17 . ± .
06 18 . ± . . ± .
16 20 . ± . > . > .
26 20 . ± . ... ... ...0242+23 0.713 ... ... ... ... > . ...0322+222 0.722 > . > . > . > . > . > . > .
08 19 . ± .
34 19 . ± .
26 19 . ± .
31 20 . ± .
33 20 . ± . > . ... ...0458–02 0.251 . ± .
22 19 . ± .
14 19 . ± . ... ... ...0601–70 0.249 . ± .
22 19 . ± .
15 20 . ± .
25 20 . ± . ... ...0625–5438 0.472 . ± .
21 19 . ± .
18 18 . ± .
11 18 . ± .
11 19 . ± .
19 21 . ± . > . ... ...0908+416 0.056 > . > . > . > . > . > . > .
16 18 . ± .
24 19 . ± . ... ...1343+451 0.078 > . > . > . > . > . > . > . ...1537+2744 0.094 . ± . . ± .
11 19 . ± .
11 19 . ± .
14 20 . ± .
18 19 . ± . > . > . ... ...1959–4246 0.259 . ± .
07 19 . ± .
06 19 . ± . ... ... ...2118+188 0.393 > .
08 20 . ± . . ± .
18 20 . ± . > . > . > . > . > . ... ... ... Table 3.
Summary of
Swift /UVOT observed magnitudes. Lower limits are at σ level.c (cid:13) , 000–000 G. Ghisellini et al.
Figure 1.
SEDs of 0106+01 (=4C+01.02) together with the fitting mod-els, with parameters listed in Tab. 4. De–absorbed UVOT, XRT and BATdata are indicated by darker symbols (red in the electronic version), whilearchival data (from NED) are in light grey. Diamonds (and lower arrows,cyan in the electronic version) indicate UVOT data not de–absorbed by in-tervening Lyman– α clouds. The short–dashed line is the emission from theIR torus, the accretion disk and its X–ray corona. The solid thin (green) lineis the non–thermal emission (sum of synchrotron, SSC and EC). The longdashed and the dot–dashed grey lines are the synchrotron self–Compton(SSC) and the External Compton (EC) components, respectively. The thicksolid (blue) line is the sum of all components. The grey stripe in the γ –rayband corresponds to the Fermi /LAT sensitivity of the first 3 months (10 σ ,upper boundary) and for 11 months (4 σ , lower boundary). phasize the relative importance of the different sources of the seedphotons for the inverse Compton scattering process, and how theychange as a function of the distance of the emitting region from theblack hole. Here we briefly summarize the main characteristics ofthe model.The source is assumed spherical (radius R ) and located at adistance R diss from the central black hole. The emitting electronsare injected at a rate Q ( γ ) [cm − s − ] for a finite time equal to thelight crossing time R/c . The shape of Q ( γ ) we adopt is assumedto be a smoothly broken power law with a break at γ b : Q ( γ ) = Q ( γ/γ b ) − s γ/γ b ) − s + s (1)The emitting region is moving with a velocity βc correspond-ing to a bulk Lorentz factor Γ . We observe the source at the viewingangle θ v and the Doppler factor is δ = 1 / [Γ(1 − β cos θ v )] . Themagnetic field B is tangled and uniform throughout the emittingregion. We take into account several sources of radiation externallyto the jet: i) the broad line photons, assumed to re–emit 10% ofthe accretion luminosity from a shell–like distribution of clouds lo-cated at a distance R BLR = 10 L / , cm; ii) the IR emissionfrom a dusty torus, located at a distance R IR = 2 . × L / , cm; iii) the direct emission from the accretion disk, including itsX–ray corona. Furthermore we take into account the starlight con-tribution from the inner region of the host galaxy and the cosmicbackground radiation, but these photon sources are unimportant inour case. All these contributions are evaluated in the blob comov-ing frame, where we calculate the corresponding inverse Compton Figure 2.
SED of PMN 0157–4614, B2 0242+23 and TXS 0322+222. Sym-bols and lines as in Fig. 2. radiation from all these components, and then transform into theobserver frame.We calculate the energy distribution N ( γ ) [cm − ] of the emit-ting particles at the particular time R/c , when the injection processends. Our numerical code solves the continuity equation which in-cludes injection, radiative cooling and e ± pair production and re-processing. Ours is not a time dependent code: we give a “snap-shot” of the predicted SED at the time R/c , when the particle distri-bution N ( γ ) and consequently the produced flux are at their maxi-mum.For all sources in our sample, the radiative cooling time of theparticles is short, shorter than R/c even for low energetic particles.In Tab. 4 (last column) we have listed the values of γ c , that is theminimum value of the random Lorentz factor of electrons cooling c (cid:13) , 000–000 igh redshift Fermi blazars Figure 3.
SED of PKS 0420+022, PKS 0451–28, PKS 0458–02 and PKS0601–70. Symbols and lines as in Fig. 2.
Figure 4.
SED of PMN 0625–5438, TXS 0907+230, TXS 0908+416 andPKS 1149–084. Symbols and lines as in Fig. 2.c (cid:13) , 000–000
G. Ghisellini et al.
Figure 5.
SED of TXS 1343+451, PMN 1344–1723, [WB92] 1537+2754and SWIFT J1656–3302. Symbols and lines as in Fig. 2.
Figure 6.
SED of PMN 1959–4246, TXS 2118+188 and PMN 2135–5006Symbols and lines as in Fig. 2. in one light crossing time. Since it is always smaller than γ b , al-most all the energy injected in the form of relativistic electrons isradiated away. Most of the cooling is due to the inverse Comptonscattering with broad line photons, with a minor contribution fromthe synchrotron and the self–Compton process. Therefore we al-ways are in the fast cooling regime (i.e. γ c < γ b ). In this regimethe produced luminosity does not depend on the amount of the radi-ation energy density, but only on the energy content of the injectedrelativistic electrons.Another implication is that, at lower energies, the N ( γ ) distri-bution is proportional to γ − , while, above γ b , N ( γ ) ∝ γ − ( s +1) .The electrons emitting most of the observed radiation have energies c (cid:13) , 000–000 igh redshift Fermi blazars γ peak which is close to γ b (but these two energies are not exactlyequal, due to the curved injected spectrum).The accretion disk component is calculated assuming a stan-dard optically thick geometrically thin Shakura & Sunjaev (1973)disk. The emission is locally a black body. The temperature profileof the disk is given e.g. in Frank, King & Raine (2002). Since theoptical–UV emission is the sum of the accretion disk and the jetnon–thermal components, for a few sources there is some degener-acy when deriving the black hole mass and the accretion rate.We model at the same time the thermal disk (and IR torus) ra-diation and the non–thermal jet–emission. The link between thesetwo components is given by the amount of radiation energy den-sity (as seen in the comoving frame of the emitting blob) comingdirectly from the accretion disk or reprocessed by the BLR and theIR torus. This radiation energy density depends mainly on R diss ,but not on the adopted accretion rate or black hole mass (they arein any case chosen to reproduce the observed thermal disk lumi-nosity).By estimating the physical parameters of the source we cancalculate the power that the jet carries in the form of radiation ( P r ),magnetic field ( P B ), relativistic electrons ( P e ) and cold protons( P p ) assuming one proton per electron. These powers are calcu-lated according to: P i = πR Γ cU ′ i (2)where U ′ is the energy density of the i th component in the comov-ing frame. α absorption Being at z > the optical–UV flux of the blazars in our samplecould be affected by absorption of neutral hydrogen in interveningLyman– α absorption systems. To correct for this, we use the attenu-ation calculated in G10 specifically for the UVOT filters, illustratedin Fig. 3 of that paper.Full details of our calculation will be described in Haardt etal. (in preparation), together with a more refined treatment of themean attenuation and its variance around the mean. The currentprocedure is very crude, especially when the attenuation is large(i.e. optical depths larger than unity) because in such cases most ofthe attenuation is due to very few clouds, implying a large variance.However, we note that the variance of the attenuation is largelyreduced when the actual filter width is taken into account (Madau1995). Our absorption model results in a mean number of thicksystems which is < for z ∼ < , so we do not expect excessive off–set of the attenuation along individual line of sight with respect tothe mean value.When presenting the SED of our sources, we will show boththe fluxes and upper limits de–reddened for the extinction dueto our Galaxy and the fluxes (and upper limits) obtained by de–absorbing them with the τ eff shown in Fig. 3 in G10. Table 4 lists all parameters used to model the SEDs of our blazars,Table 5 lists the different forms of power carried by the jet andFig. 1–6 show the SEDs of the 19 blazars studied in this paperand the corresponding fitting model. In all figures we have markedwith a grey shaded area the
Fermi /LAT sensitivity, bounded on thebottom by considering one year of operation and a 5 σ detectionlevel, and on the top by considering 3 months and a 10 σ detection Figure 7.
Distribution of the black hole masses derived for the z > Fermi /LAT (top) and
Swift /BAT samples (bottom). The hatched area in thetop panel corresponds to the 4 blazars in common. level (this assumes a common energy spectral index of α γ ∼ , thesensitivity limit for other spectral indices is slightly different, seeFig. 9 in A10). All these sources were not detected in the first 3months, in fact the (11 months) γ –ray data points are very close tothe lower boundary of the grey area. There are exceptions: 0106+01(=4C+01.02) is brighter than the 3–months, 10 σ sensitivity limitseven if it has not been included in LBAS. This is due to a ratherstrong variability of the source, fainter in the first 3 months andbrighter soon after. The same occurred for 1344–1723 and 0451–28. The opposite happened for 0227–369, 0347–211 and 0528+134(i.e. they were brighter during the first 3 months), but their flux,averaged over 11 months, was in any case large enough to let theirinclusion in the 1LAC sample.Some of the sources have a sufficiently good IR–optical–UVcoverage to allow to see a peak of the SED in this band (seefor instance 0420+022; 0451–28; 0458–02; 0625–5428; 0907+230;0908+416; 1149–084; 1656–3302). The other sources have a SEDconsistent with a peak in this band, but the lack of data also al-lows for a peak at lower frequencies. We interpret the peak in theoptical band as due to the accretion disk, and assume its presencealso in those blazars where it is allowed, but not strictly required.By assuming a standard Shakura–Sunyaev (1973) disk we are thenable to estimate both the black hole mass and the accretion rate.This important point has been discussed in G10, in Ghisellini et al.(2009) (for S5 0014+813) and in Ghisellini & Tavecchio (2009).The radio data cannot be fitted by a simple one–zone modelspecialized to fit the bulk of the emission, since the latter must beemitted in a compact region, whose radio flux is self–absorbed upto hundreds of GHz. The radio emission should come from largerregions of the jet. On the other hand, when possible, we try to havesome “continuity” between the non–thermal model continuum andthe radio fluxes (i.e. the model, in its low frequency part, should notlie at too low or too high fluxes with respect to the radio data). Inthe following we briefly comment on the obtained parameters. Dissipation region —
The distance R diss at which most of the dis-sipation takes place is one of the key parameters for the shape c (cid:13) , 000–000 G. Ghisellini et al.
Name z R diss
M R
BLR P ′ i L d B Γ γ b γ max s s γ c [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]0106+01 2.107 900 (600) 5e9 866 0.08 75 (0.1) 1.13 14 300 5e3 0 3.1 2.00157–4614 2.287 195 (1.3e3) 5e8 274 0.015 7.5 (0.1) 1.54 15 200 2e3 –1 3.0 5.70242+23 2.243 420 (700) 2e9 812 0.022 66 (0.22) 2.13 15 220 2e3 0.5 3.1 2.60322+222 2.066 450 (500) 3e9 671 0.06 45 (0.1) 2.06 12 150 3e3 0.5 3.1 3.70420+022 2.277 210 (1.4e3) 5e8 725 0.02 52.5 (0.7) 3.79 15 300 2e3 –1 3.2 4.90451–28 2.56 540 (450) 4e9 1.1e3 0.24 120 (0.2) 2.66 10 180 2e3 0. 2.6 4.10458–02 2.291 472 (450) 3.5e9 606 0.07 37 (0.07) 2.14 10 200 5e3 0.8 3.0 4.90601–70 2.409 525 (500) 3.5e9 606 0.05 37 (0.07) 1.83 12.9 190 5e3 –1 3.1 2.80625–5438 2.051 270 (900) 1e9 648 0.03 42 (0.28) 2.64 15 240 5e3 0 4.0 3.90907+230 2.661 360 (1.5e3) 8e8 290 0.05 8.4 (0.07) 0.36 13 300 1.7e4 0.75 2.8 32.70908+416 2.563 180 (600) 1e9 346 0.025 12 (0.08) 1.06 14 150 3e3 0 3.1 7.01149–084 2.367 720 (600) 4e9 849 0.015 72 (0.12) 1.39 14 300 3e3 –1 3.0 1.81343+451 2.534 420 (700) 2e9 387 0.045 15 (0.05) 1.09 14 150 5e3 –1 2.8 6.51344–1723 2.409 330 (1.1e3) 1e9 274 0.027 7.5 (0.05) 0.89 13 1.4e3 8e3 –1 2.5 26.21537+2754 2.19 120 (400) 1e9 367 0.015 13.5 (0.09) 4.42 11.5 60 4e3 0.5 2.1 12.21656–3302 2.4 525 (700) 2.5e9 1.1e3 0.07 124 (0.33) 1.09 15 70 1e4 0.75 2.85 2.11959–4246 2.174 825 (500) 5.5e9 812 0.024 66 (0.08) 1.51 12.9 170 5e3 0 2.7 1.92118+188 2.18 270 (600) 1.5e9 424 0.022 18 (0.08) 1.85 14 250 1e4 0.5 2.8 4.62135–5006 2.181 189 (900) 7e8 324 0.023 10.5 (0.1) 2.02 14 180 2e3 –1 3.2 6.60227–369 2.115 420 (700) 2e9 547 0.08 30 (0.1) 1.5 14 200 5e3 0 3.1 3.00347–211 2.944 750 (500) 5e9 866 0.12 75 (0.1) 1.5 12.9 500 3e3 –1 3.0 1.80528+134 2.04 420 (1400) 1e9 866 0.13 75 (0.5) 2.6 13 150 3e3 –1 2.8 3.30537–286 3.104 420 (700) 2e9 735 0.13 54 (0.18) 1.92 15 50 2e3 –1 3 2.60743+259 2.979 1.65e3 (1.1e3) 5e9 866 0.24 75 (0.1) 0.1 15 200 5e3 0.75 2.6 1020805+614 3.033 270 (600) 1.5e9 581 0.15 34 (0.15) 2.54 14 60 3e3 –0.5 3 4.30836+710 2.172 540 (600) 3e9 1.5e3 0.22 225 (0.5) 3.28 14 90 2e3 –1 3.6 2.10917–449 2.1899 900 (500) 6e9 1341 0.1 180 (0.2) 1.95 12.9 50 4e3 –1 2.6 1.61329–049 2.15 450 (1e3) 1.5e9 822 0.07 67.5 (0.3) 1.4 15 300 5e3 1 3.3 2.5 Table 4.
List of parameters used to construct the theoretical SED. Not all of them are “input parameters” for the model, because R BLR is uniquely determinedfrom L d , and the cooling energy γ c is a derived parameter. Col. [1]: name; Col. [2]: redshift; Col. [3]: dissipation radius in units of cm and (in parenthesis)in units of Schwarzschild radii; Col. [4]: black hole mass in solar masses; Col. [5]: size of the BLR in units of cm; Col. [6]: power injected in the blobcalculated in the comoving frame, in units of erg s − ; Col. [7]: accretion disk luminosity in units of erg s − and (in parenthesis) in units of L Edd ;Col. [8]: magnetic field in Gauss; Col. [9]: bulk Lorentz factor at R diss ; Col. [10] and [11]: break and maximum random Lorentz factors of the injectedelectrons; Col. [12] and [13]: slopes of the injected electron distribution [ Q ( γ ) ] below and above γ b ; Col. [14] values of the minimum random Lorentz factorof those electrons cooling in one light crossing time. The total X–ray corona luminosity is assumed to be in the range 10–30 per cent of L d . Its spectral shapeis assumed to be always ∝ ν − exp( − hν/
150 keV) . The viewing angle θ v is ◦ for all sources. of the overall SED, since it controls the amount of energy den-sities as seen in the comoving frame (see Ghisellini & Tavecchio2009). For almost all sources we have R diss < R BLR , while for0106+01, 0907+230, 1343+451, 1344–1723, 1959–4246, R diss isslightly larger than R BLR , and for 0743+259 R diss ∼ R BLR .In all sources the dominant cooling is trough inverse Comp-ton off the seed photons of the BLR. This is true also for thefew blazars in which R diss > R BLR since, even if U ′ BLR seenin the comoving frame is smaller, also U B is smaller, imply-ing that the ratio U ′ BLR /U B is similar to the values in othersources [ U ′ BLR /U B ranges between ∼
30 (1344–1723) and ∼ γ c to be larger ( γ c = 102 , see Tab. 4).With larger still R diss ≫ R BLR , the main seed photons for theCompton scattering process would became the photons producedby the the IR torus (if it exists), but this case does not occur for oursources.
Compton dominance —
This is the ratio between the luminosityemitted at high frequencies and the synchrotron luminosity. Theaverage magnetic field is found to be of the order of 1 Gauss, with a corresponding magnetic energy density that is around two ordersof magnitude lower than the radiation energy density. Correspond-ingly, all sources are Compton dominated.
Black hole masses —
Fig. 7 shows the distribution of black holemasses for the 28 blazars at z > and compares them with the dis-tribution of masses for the high redshift BAT blazars. Although theblack hole masses of the BAT sample extend to larger values, thereare still too few sources to estimate if the two distributions are dif-ferent. It is interesting to note that all but 3 sources (0420+022,0907+230 and 2135–5006) have black hole masses greater than M ⊙ . In Ghisellini, Tavecchio & Ghirlanda (2009) we consid-ered the Fermi blazars of γ –ray luminosity L γ > erg s − ,finding, with the same method and model applied here, that for allthese sources the black hole mass was greater than a billion solarmasses. Therefore all blazars with L γ > erg s − have blackholes heavier than M ⊙ , while the vast majority, but not all,blazars at z > have such large black hole masses. We searchedin the literature other estimates of the black hole masses for theobjects in this sample, finding M = 2 . × M ⊙ for 0836+710(estimated by Liu, Jiang & Gu 2006), and other few limits for the c (cid:13) , 000–000 igh redshift Fermi blazars Name log P r log P B log P e log P p Table 5.
Logarithm of the jet power in the form of radiation ( P r ), Poyntingflux ( P B ), bulk motion of electrons ( P e ) and protons ( P p , assuming oneproton per emitting electron). Powers are in erg s − . The bottom part of thetable reports the data derived in G10 and G09. black hole masses for 0836+710 and 0528+134, that were howeverbased assuming an isotropic γ –ray emission. Disk luminosities —
We are considering very powerful blazars, sowe do expect large disk luminosities, not only on the basis of anexpected positive trend between the observed non–thermal (albeitbeamed) and the accretion luminosities, but also on the basis ofthe observed luminosities of the broad lines, that should linearlydepend on the accretion power. What is interesting is that all theFSRQs analyzed up to now (i.e. belonging to the LBAS sample orto the subset of high redshift 1LAC and BAT samples) have a ratio L d /L Edd between 10 − and 1. This can be seen in the mid panel ofFig. 8, that shows L d as a function of the derived black hole mass.The two dashed lines correspond to the Eddington and 1% Edding-ton luminosities. This confirms the idea of the “blazars’ divide” as aresult of the changing of the accretion mode (Ghisellini, Maraschi& Tavecchio 2009): from the standard Shakura–Sunyaev (appro-priate for all FSRQs) to the ADAF–like regime (appropriate for BLLacs). The z > blazars analysed here have L d /L Edd ratios rang-ing from 0.05 and 0.7. The exact values of the disk luminositiesderived here are the frequency integrated bolometric luminositiesof the assumed Shakura–Sunyaev accretion disk model that bestinterpolates the data. On the other hand, any other accretion diskmodel has to fit the data as well, implying that our values of L d are robust, and nearly model–independent, within the limit of theuncertainties of the observed data. Figure 8.
The observed γ –ray luminosity in the 0.1–10 GeV band, theaccretion luminosity L d and the total jet power P jet as a function of the de-rived black hole mass. All points correspond to FSRQs. Different symbolscorrespond to LBAS FSRQs with L γ > erg s − (red stars, analysedin Ghisellini, Tavecchio & Ghirlanda 2009); 1LAC FSRQs with z > (empty blue circles); BAT FSRQs with z > (black diamonds, G10) andthe LBAS FSRQs with L γ < erg s − (grey squares, Ghisellini et al.2010b). The γ –ray luminosity (top panel) is the observed beamed one, andit has not to be confused with P r , the power spent by the jet to produce theradiation we see. The mid panel shows that all FSRQs have disk luminosi-ties between 0.01 and 1 Eddington luminosity. The bottom panel shows that P jet can be larger (but not by a big factor) than the Eddington luminositycorresponding to the dashed line. Jet powers —
The values listed in Tab. 5 are very similar to the val-ues derived for other powerful
Fermi
FSRQs. They are not, how-ever, the absolutely greatest powers found. This can be seen in thebottom panel of Fig. 8, showing P jet as a function of the black holemass, and in the mid and bottom panels of Fig. 9, where we plot thepower of the jet spent in the form of radiation ( P r ) and the total jetpower P jet as a function of L d . We can compare the z > z > sources follow the distribution of the most luminous γ –ray blazars. Remarkably though, the z > BAT FSRQs appear c (cid:13) , 000–000 G. Ghisellini et al.
Figure 9.
Top: the observed γ –ray luminosity L γ as a function of theaccretion luminosity L d for the LBAS FSRQs (grey squares), the z > Swift /BAT (diamonds) FSRQs. The dashed lineindicates L γ = L d . Mid: the power P r as a function of L d . P r can beconsidered as a robust and almost model–independent lower limit to the jetpower. Bottom: the total jet power P jet = P B + P e + P p as a functionof L d . Almost all sources have P jet > L d . One cold proton per emittingelectron is assumed. to lie at the extreme of the distributions, being the more powerfulin L d , and among the most powerful in P r and P jet . For calculat-ing the power carried by the jet in the form of protons, we re–iteratethat we have assumed one cold proton per emitting electron: if thereexist a population of cold electrons, and no electron–positron pairs,than we underestimate P p and then P jet , while if there are no coldleptons but there are pairs then we overestimate P p . Finally, pro-tons are assumed cold for simplicity (and “economy”), but theycould be hot or even relativistic (if, e.g. shocks accelerate not onlyelectrons but also protons), and in such cases the power is underes-timated. For a detailed discussion about the presence of electron–positron pairs in blazars’ jets we refer to the discussions in Sikora& Madejski (2000), Celotti & Ghisellini (2008) and G10, whereone can finds arguments limiting the amount of pairs in the jet. Afew electrons–positrons per proton are possible, but not more. Jet powers vs accretion luminosities —
Fig. 9 shows that the corre-lations found in G10 between P r and/or P jet and L d are confirmed.We remind the reader that P r and L d are independent quantitieseven if the main radiation mechanism is the inverse Compton pro-cess using broad line photons as seeds, that in turn are proportionalto the accretion disk luminosity. This is because the radiative cool-ing of the emitting electrons is complete, implying that the pro-duced jet luminosity becomes independent on the amount of radi-ation energy density. In other words: in the fast cooling regime thejet always emits all the energy of its relativistic electrons, no matterthe amount of the luminosity of the accretion disk.A least square fit returns a chance probability P = 4 × − that log P r and log P jet are correlated with log L d (and the cor-relation are consistent with being linear). They remain significantalso when considering the common redshift dependence, althoughthe chance probability increases to P = 4 × − (for the P r – L d correlation) and to P = 10 − ( P jet – L d ).As expected, the 1LAC blazars at high redshifts are amongthe most powerful, even if there are blazars at lower redshifts withcomparable powers. This can be seen comparing the empty circles,corresponding to the 1LAC blazars of our sample, with the LBASFSRQs of L γ > erg s − (stars) and the BAT FSRQs at z > (black diamonds). There are a few sources with P r > L d , andseveral with P r ∼ L d . The jet in these blazars, only to produce theradiation we see, requires a power comparable to (or even largerthan) the disk luminosity. The P r power should be considered avery robust estimate of the minimum jet power: it is robust becauseit is almost model–independent ( P r ∼ L γ / Γ , see G10), and it isa lower limit because it corresponds to the entire jet power beingconverted into radiation at the γ –ray emitting zone. Indeed, if thereis one proton per emitting electron, the total jet power, dominatedby the bulk motion of cold protons, becomes a factor ∼ largerthan L d (bottom panel of Fig. 9), with the FSRQs of our sampledistributed in a large portion of the P jet – L d plane.We believe that the relation between both P r and P jet with L d is a key ingredient to understand the birth of jets: accretion must play a key role. Comparison with other models —
Several groups (Larionov et al.2008; Marscher et al. 2008, 2010; Sikora, Moderski & Madejski2008) proposed that the emitting region, especially during flares, isproduced at distances from the central black hole of the order of10–20 pc (much larger than what we assume) at the expected loca-tion of a reconfinement shock (e.g. Sokolov, Marscher & McHardy2004). On the basis of an observed peculiar behaviour of the polar-ization angle in the optical, Marscher et al. (2008) thus suggested c (cid:13) , 000–000 igh redshift Fermi blazars Figure 10.
Top: The ratio of the (rest frame) luminosities in the 0.1–100 GeV and 15–55 keV bands as a function of the rest frame peak fre-quency of the high energy bump. For sources not detected in one of the twobands, we have used the corresponding flux resulting from the modelling oftheir overall SED. Different symbols are from the sources detected only by
Fermi /LAT, by both
Swift /BAT and
Fermi /LAT and only by
Swift /BAT, aslabelled. For comparison we also show all FSRQs (grey dots) and BL Lacs(squares) in the LBAS sample. The continuous line shows the estimate us-ing the smoothly broken power law function (see text) with α x = 0 . and α γ = 1 . . Bottom panel: the same, but plotting observed peak frequen-cies. The continuous lines are for h z i = 2 . and h z i = 0 . . See how thesimple function in Eq. 3 interpolates well the data of FSRQs. The X and γ –ray SED of BL Lacs, instead, is not due to a single radiation process,since the hard X–rays are often due to the tail of the synchrotron emission.As a consequence, they show the opposite behavior of FSRQs: a smaller L X /L γ ratio when increasing ν peak , indicating an increasingly strongercontribution of the synchrotron flux to the hard X–ray band. that blobs ejected from the central region are forced by the mag-netic field to follow a helical path, accounting for the observed ro-tation of the polarization angle in the optical. Flares (at all wave-lengths) correspond to the passage of these blobs through a stand-ing conical shock, triggered by the compression of the plasma inthe shock. This has important consequences for the variability ofthe emission: since the emission region is located at large distances Figure 11.
The rest frame peak frequency ν peak of the high energy emis-sion as a function of the ν peak L ν peak luminosity. Symbols as in Fig. 10.A clear trend can be seen when considering BL Lacs and FSRQs together,while FSRQs only are characterized by a very large dispersion. High red-shifts FSRQs occupy the region of the largest luminosities and smallest ν peak . We indicate the two “outliers”: BL Lac and Ap Lib. from the central engine, its size is large, and the expected variabil-ity timescale cannot be very short. Assuming R diss = 15 pc, ajet aperture angle of θ jet = 3 ◦ and δ = 20 , we find a minimumvariability time scale of t var = θ jet R diss (1 + z ) / ( cδ ) of the orderof 1.5 (1+z) months, that for sources at z > implies a mini-mum variability timescales of 5–6 months. The main high energyemission mechanism is still the inverse Compton process, usingas seeds the IR radiation of a surrounding torus (Sikora, Moderski& Madejski 2008) with a possible important contribution from jetsynchrotron radiation (Marscher et al. 2008, 2010). The main dif-ficulty of these models concerns the expected variability, predictedto occur on a very long time scales, if the size of the emitting re-gion is proportional (through e.g. the opening angle of the jet), tothe distance of the source to the black hole. Instead the observed γ –ray flux in all strong γ –ray sources (the ones for which a reliablevariability behaviour can be established) varies on much shortertime scales, and factor 2 flux changes can occur even on 3–6 hours(see Tavecchio et al. 2010 for 3C 454.3 and PKS 1510–089; Bon-noli et al. 2010; Foschini et al. 2010 and Ackermann et al. 2010for 3C 454.3; Abdo et al. 2009b for PKS 1454–354; Abdo et al.2010c for PKS 1502+105). This indicates that the source is com-pact. This in turn suggests (although it does not prove) that its lo-cation cannot be too far from the black hole. It then also suggeststhat it is within the broad line region. In turn, this suggests thatthe broad lines are the main seeds for the inverse Compton scatter-ing process. Occasionally, though, dissipation could occur furtherout, where the main seeds are the infrared photons produced by aputative torus surrounding the accretion disks. Since the seed pho-tons have smaller frequencies, in these cases the produced high en-ergy spectrum suffers less from possible effects of the decreasing(with seed frequencies) scattering Klein–Nishina cross section andless from possible photon–photon interactions leading to electron–positron pair production. The decreased importance of both effectsmay account for high energy spectra extending, unbroken, up tohundreds of GeV. These cases should be characterized by a longervariability timescale.In the model of Marscher et al. (2008) a very short variability c (cid:13) , 000–000 G. Ghisellini et al. timescale still indicates a very compact emitting region, but never-theless located at a large distance from the central black hole.
Both the 1LAC sample and the BAT 3–years survey have a ratheruniform sky coverage, and both approximate a flux limited sample.The 1LAC sample has a limiting flux sensitivity that depends on the γ –ray spectral index of the sources, but since we are dealing withFSRQs only (whose α γ is contained in a relatively narrow range),we can consider this sample as flux limited. It is then interestingto compare the high redshift blazars (all of them are FSRQs) con-tained in the two samples. We remind that for blazars with z > ,we have 10 FSRQs in the BAT sample, 28 in the 1LAC, and 4 inboth. The X–ray to γ –ray SED of these sources is very similar:even if we have only 4 blazars in common, for all sources the SEDhas a high energy peak in the ∼ MeV–100 MeV band, with α x < and α γ > .Thus the spectrum can be approximated by a broken powerlaw. For illustration, consider the smoothly broken power law ofthe form L ( ν ) ∝ ( ν/ν b ) − α x ν/ν b ) α γ − α x (3)If the energy indices α x < and α γ > , the peak is at ν peak = ν b [(1 − α x ) / ( α γ − / ( α γ − α x ) . With this function we can easilycalculate the ratio of the BAT [15–55 keV] to LAT [0.1–100 GeV]luminosities as a function of ν peak , and see if it compares well tothe data. We alert the reader that by “data” we mean real observeddata when the source has been detected in the X–ray and γ –rayband, while, when the detection is missing, we mean the “data”coming from our fitting model described in § L X /L γ as a function of ν peak calculated using Eq.3 setting α x = 0 . and α γ = 1 . is plotted in Fig. 10 as a grey (or-ange in the electronic version) line, together with the points corre-sponding to high redshift blazars, studied in this paper and in G10.Furthermore, Fig. 10 reports also the data of all the LBAS blazarsstudied in Ghisellini et al. (2010b). These are FSRQs (filled greycircles) and BL Lacs (empty grey squares). In the top panel ν peak is in the rest frame of the source, while in the bottom panel ν peak is the observed one.Fig. 10 shows a remarkable agreement between the FSRQ dataand the simple broken power law of Eq. 3, both for high redshiftsand for less distant FSRQs. BL Lac objects, instead, are not wellrepresented by Eq. 3. In fact, in many BL Lacs the hard X–rays cor-respond to the (steep) tail of the synchrotron component. Thereforethe X and the γ –rays are produced by two different mechanisms,and Eq. 3 does not represent their overall high energy SED. In BLLacs the importance of X–rays increases increasing ν peak , becauseof the increasing importance of synchrotron flux in the hard X–rays.Coming back to high redshift FSRQs, only 4 sources have de-tection in both bands, while 24 FSRQs are detected only by theLAT and 6 only by the BAT instrument. Our model explains thelarge fraction of sources that are detected only in one of the twoinstruments as due to the different ν peak of the sources: if ν peak is large (above 10 MeV), F γ /F X is large and the source is rela-tively weak in the BAT band, while if ν peak <
10 MeV the sourcebecomes relatively weak in the LAT band and strong in the BAT.It is interesting to see if the derived ν peak correlates withthe bolometric luminosity. Fig. 11 shows that a trend indeed ex-ists: more powerful sources have smaller ν peak (we have used ν peak L ν peak as a proxy for the bolometric luminosity). But com-paring with all the LBAS FSRQs (grey dots) we see that the z > luminous FSRQs belong to a broader (i.e. more scattered) distribu-tion, and that the high– z BAT blazars are really the most extreme.We can conclude that i) there is a a trend between the highenergy peak and the peak luminosity, ii) that this correlation has alarge scatter, even if iii) the z > FSRQs show the same trendwith less scatter (but this may be due to the still small number) andfinally iv) the z > blazars in the 3–years BAT survey all lie in thehighest luminosity, smallest ν peak part of the plane.When more BAT detections of high redshift LAT blazars willbecome available (and, conversely, when LAT will detect moreBAT high– z blazars) this trend can be tested directly (i.e. with-out modelling the SED). The importance of this is two–fold: firstwe can estimate in a reasonable way the peak energy of the highenergy emission having the hard X–ray and the γ –ray luminosities,and second (and more important) we could conclude that the mostpowerful blazars can be more easily picked up trough hard X–raysurveys, as the one foreseen with the EXIST mission (Grindlay etal. 2010).
The total number of z > blazars with high energy informationis 34 (28 with a LAT detection, 6 with BAT, and 4 with both). It isstill a limited number, the tip of the iceberg of a much larger (andfainter) population, but it is derived from two well defined samples(LAT and BAT), that we can consider as flux limited and comingfrom two all sky surveys (excluding the galactic plane). The mainresult of studying them is that all the earlier findings concerningthe physical parameters of the jet emitting zone, the jet power, andthe correlation between the jet power and the disk luminosity areconfirmed.They are in agreement with the blazar sequence, i.e. their non–thermal SED are “redder” than less luminous blazar, with a largedominance of their high energy emission over the synchrotron one.This implies that the disk emission is left unhidden by the syn-chrotron flux, and this allows an estimate of the black hole massand the accretion rate. The uncertainties associated with these es-timates are relatively small within the assumption that the thermalcomponent is produced by a standard Shakura–Sunyaev disk withan associated non–spinning hole. In G10 we argued that in any casethe masses are not largely affected by this assumption, and in par-ticular that in the Kerr case the derived masses are not smaller, de-spite the greater accretion efficiency. The possibility of an intrin-sic collimation of the disk radiation appears more serious. If thedisk is not geometrically thin, but e.g. a flared disk, then we ex-pect a disk emission pattern concentrated along the normal to thedisk, i.e. along the jet axis. We argued previously (Ghisellini et al.2009) that this can be the case of S5 0014+813 at z = 3 . ,an extraordinary luminous blazars (detected by BAT) with an esti-mated “outrageous” black hole mass of × M ⊙ . And indeedwe found it to be an “outlier” in the P r – L d plane. Reverting theargument, it implies that the other FSQRs, obeying a well defined P r – L d trend, should have disk luminosities with a quasi–isotropicpattern, i.e. standard, not flaring, disks. The other severe uncertaintyon the mass estimation for objects at large redshifts is the amount ofattenuation of their optical–UV flux, due to intervening Lyman– α clouds. We have corrected for this assuming an average distributionof clouds, and when the attenuation is due to a few thick cloudsthe expected variance is large. While we plan to refine such esti- c (cid:13)000
The total number of z > blazars with high energy informationis 34 (28 with a LAT detection, 6 with BAT, and 4 with both). It isstill a limited number, the tip of the iceberg of a much larger (andfainter) population, but it is derived from two well defined samples(LAT and BAT), that we can consider as flux limited and comingfrom two all sky surveys (excluding the galactic plane). The mainresult of studying them is that all the earlier findings concerningthe physical parameters of the jet emitting zone, the jet power, andthe correlation between the jet power and the disk luminosity areconfirmed.They are in agreement with the blazar sequence, i.e. their non–thermal SED are “redder” than less luminous blazar, with a largedominance of their high energy emission over the synchrotron one.This implies that the disk emission is left unhidden by the syn-chrotron flux, and this allows an estimate of the black hole massand the accretion rate. The uncertainties associated with these es-timates are relatively small within the assumption that the thermalcomponent is produced by a standard Shakura–Sunyaev disk withan associated non–spinning hole. In G10 we argued that in any casethe masses are not largely affected by this assumption, and in par-ticular that in the Kerr case the derived masses are not smaller, de-spite the greater accretion efficiency. The possibility of an intrin-sic collimation of the disk radiation appears more serious. If thedisk is not geometrically thin, but e.g. a flared disk, then we ex-pect a disk emission pattern concentrated along the normal to thedisk, i.e. along the jet axis. We argued previously (Ghisellini et al.2009) that this can be the case of S5 0014+813 at z = 3 . ,an extraordinary luminous blazars (detected by BAT) with an esti-mated “outrageous” black hole mass of × M ⊙ . And indeedwe found it to be an “outlier” in the P r – L d plane. Reverting theargument, it implies that the other FSQRs, obeying a well defined P r – L d trend, should have disk luminosities with a quasi–isotropicpattern, i.e. standard, not flaring, disks. The other severe uncertaintyon the mass estimation for objects at large redshifts is the amount ofattenuation of their optical–UV flux, due to intervening Lyman– α clouds. We have corrected for this assuming an average distributionof clouds, and when the attenuation is due to a few thick cloudsthe expected variance is large. While we plan to refine such esti- c (cid:13)000 , 000–000 igh redshift Fermi blazars Figure 12.
The ratio between the luminosities in the BAT and LAT energyranges (i.e. 15–55 keV and 0.1–10 GeV) as a function of the disk luminos-ity, in Eddington units. Big dots are the 6 BAT blazars in the z > samplenot detected (yet) by Fermi /LAT, squares are BAT blazars of the same sam-ple already present in the 1LAC sample, triangles are the sources discussedin this paper, and the small grey dots are all the FSRQs present in the LBASsample discussed in G10. The dashed line is the best least square fit (chanceprobability P = 2 × − ). mates (Haardt et al. in preparation), it is unlikely that this impliesa systematic error on the mass estimates, leading to larger values:statistically, the mass distribution should not be seriously affectedby this uncertainty.Although affected by the same issues outlined above, also thecorrelation between the jet power and the disk luminosity shouldresist when a more refined treatment of the Lyman– α absorptionwill be available, by the same arguments. Therefore we can con-clude that the accretion rate is really a fundamental player in pow-ering the relativistic jet, and we refer the reader to Ghisellini et al.(2010b) for a detailed discussion about this finding. If true, theseideas lead us to suggest that powerful, high redshifts “true” (i.e.really lineless) BL Lacs do not exist.Another, at first sight surprising, result of our study is that thecorrelation between the peak frequency of the high energy emis-sion and the γ –ray luminosity at this peak frequency exists alsofor the z > , highly luminous FSRQs. It is surprising becausein very powerful sources the radiative cooling is complete (i.e. allelectrons with γ larger than an few cool in one light crossing time).Therefore the electrons responsible for the peak have energies thatare not fixed by radiative cooling, but by the injection function [i.e.by the Q ( γ ) function given in Eq. 1, and more precisely by thevalue of γ b ]. On the other hand, when considering all blazars inthe LBAS sample, we see that the scatter, for large luminosities, ismuch larger, so the apparent correlation for the z > blazars canbe due to small statistics. Clearly, it is a point to investigate further.It is interesting to ask what will be the best strategy to find themost luminous FSRQs at the largest redshifts. The interest lies inthe link between jet power and disk luminosity: finding the mostpowerful jets implies to find the most accreting systems, hencethe heaviest black holes. Since for each source pointing at Earth (i.e. a blazar) there must exist ∼ similar sources pointing inother directions, the finding of even a few blazars at large red-shifts with a large black hole mass can put very interesting con-straints on the black hole mass function of jetted sources. This is-sue has been discussed by us previously (G10), and we suggestedthat the existence of the blazar sequence, plus an important K–correction effect, makes the hard X–ray range the best band whereto search for the high– z heaviest black holes. Here we re–iteratethis suggestion offering a supplementary information. For each an-alyzed FSRQs (belonging to the LBAS sample, or in this paper),we have calculated the ratio between the expected BAT luminos-ity [15–55 keV, rest frame] and the observed [0.1–100 GeV, restframe] LAT luminosity. This ratio is not an observed quantity, sincevery few FSRQs (in comparison with LAT) have been detected byBAT, but it is a result of the model. Then, with the same model,we have calculated the disk luminosity in units of Eddington. Fig.12 shows the L BAT /L LAT ratio as a function of L d /L Edd . Wehave marked with different symbols the FSRQs analyzed in thispaper and the ones analysed previously (G10). We have excludedBL Lacs for which we have only an upper limit on the disk lu-minosity. Fig. 12 suggests a clear trend (albeit with some scatter):FSRQs accreting close to Eddington emit relatively more in thehard X–rays than above 100 MeV. Formally, a least square fit re-turns ( L BAT /L BAT ) ∝ ( L d /L Edd ) . with a chance probability P = 2 . × − (for 95 objects). This implies that hard X–ray sur-veys benefit of a positive bias when looking for blazars with blackholes accreting close to Eddington. In turn, at high– z , this impliesthe finding of the heaviest black holes, since it is very likely that atthose early epochs (e.g. z > ) all black holes are accreting closeto the Eddington rate. We summarize here our main conclusions: • The blazars detected by
Fermi at z > are all FSRQs, withtypical “red” SEDs. • These FSRQs are very luminous and powerful, but they are notat the very extreme of the distribution of luminosity and jet power. • These sources have heavy black holes ( M ∼ M ⊙ ) and ac-cretion luminosities greater than ∼
10% Eddington. When includ-ing all FSRQs in the LBAS sample, irrespective of redshift, theaccretion disk luminosities is greater than 1% Eddington. • The trend of redder SED when more luminous (i.e. one ofthe defining characteristics of the blazar sequence) is confirmed,and it is even present within the relatively small range of observedluminosity of the z > blazars. • The correlation between the jet power and the disk luminos-ity is confirmed and points to a crucial role played by accretion inpowering the jet. • FSRQs with accretion disks closer to the Eddington luminos-ity have jets emitting a “redder” SED, and therefore can be moreefficiently picked up by hard X–ray surveys (such as the one fore-seen by
EXIST ), rather than by surveys in the hard γ –ray band. ACKNOWLEDGMENTS
This work was partly financially supported by an ASI I/088/06/0)grants. This research made use of the NASA/IPAC ExtragalacticDatabase (NED) which is operated by the Jet Propulsion Labora-tory, Caltech, under contract with NASA, and of the
Swift public c (cid:13) , 000–000 G. Ghisellini et al. data made available by the HEASARC archive system. We alsothank the
Swift team for quickly approving and performing the re-quested ToO observations.
REFERENCES
Abdo A.A., Ackermann M., Ajello M. et al., 2009a , ApJ, 700, 597 (A09)Abdo A.A., Ackermann M., Ajello M. et al., 2009b, ApJ, 697, 934Abdo A.A., Ackermann M., Ajello M. et al., 2010a, ApJS, 188, 405Abdo A.A., Ackermann M., Ajello M. et al., 2010b, ApJ, 715, 429 (A10)Abdo A.A., Ackermann M., Ajello M. et al., 2010c, ApJ, 710, 810Ackermann M., Ajello M., Baldini L. et al., 2010, subm to ApJ (astro–ph/1007.0483)Ajello M., Costamante L., Sambruna R.M., et al., 2009, ApJ, 699, 603Bonnoli G., Ghisellini G., Foschini L., Tavecchio F., Ghirlanda G., 2010,MNRAS, in press (astro–ph/1003.3476)Cardelli J.A., Clayton G.C. & Mathis J.S., 1989, ApJ, 345, 245Cash W., 1979, ApJ, 228, 939Celotti A. & Ghisellini G., 2008, MNRAS, 385, 283Donato D., Ghisellini G., Tagliaferri G. & Fossati G., 2001, A&A, 375, 739Foschini L., Tagliaferri G., Ghisellini G., Ghirlanda G., Tavecchio F. & Bon-noli G., 2010, MNRAS, in press (astro–ph/1004.4518)Fossati G., Maraschi L., Celotti A., Comastri A. & Ghisellini G., 1998,MNRAS, 299, 433Frank J., King A. & Raine D.J., 2002, Accretion power in astrophysics,Cambridge (UK) (Cambridge University Press)Ghirlanda G., Ghisellini G., Tavecchio F., Foschini L. 2010, MNRAS, inpress (astro–ph/1003.5163)Ghisellini G., Celotti A., Fossati G., Maraschi L. & Comastri A., 1998,MNRAS, 301, 451Ghisellini G. & Tavecchio F., 2009, 397, 985 MNRASGhisellini G., Tavecchio F. & Ghirlanda G., 2009, MNRAS 399, 2041Ghisellini G., Foschini L., Volonteri M., Ghirlanda G., Haardt F., BurlonD., Tavecchio F., 2009, MNRAS, 399, L24Ghisellini G., Maraschi L. & Tavecchio F., 2009, MNRAS, 396, L105Ghisellini G., Della Ceca R., Volonteri M. et al., 2010a, MNRAS, 405, 387(G10)Ghisellini G., Tavecchio F., Foschini L., Ghirlanda G., Maraschi L., CelottiA., 2010b, MNRAS, 402, 497Grindlay J., Bloom J., Coppi P. et al., 2010, in X–ray Astronomy 2009:Present status, multiwavelength approach and future perspectives”, toappear in AIP Conf. Proc. (editors: A. Comastri, M. Cappi, L. Angelini)(astro–ph/1002.4823)Kalberla P.M.W., Burton W.B., Hartmann D., Arnal E.M., Bajaja E., MorrasR. & P¨oppel W.G.L., 2005, A&A, 440, 775Larionov V.M. et al., 2008, A&A 492, 389Liu Y., Jiang D.R. & Gu M.F., 2006, ApJ, 637, 669Madau P., 1995, ApJ, 441, 18Marscher A.P., et al. 2008, Nat., 452, 966Marscher A.P., et al. 2010, ApJ, 461, L126Poole T.S., Breeveld A.A., Page M.J. et al., 2008, MNRAS, 383, 627Roming P.W.A., Kennedy T.E., Mason K.O. et al., 2005, Space Sci. Rev.,120, 95Sikora M., Moderski R. & Madejski G.M., 2008, ApJ, 675, 71Sikora M. & Madjesi G., 2000, ApJ, 534, 109Shakura N.I. & Sunyaev R.A., 1973 A&A, 24, 337Sokolov A., Marscher A.P. & McHardy I.M., 2004, ApJ, 613, 725Tavecchio F. & Ghisellini G., 2008, MNRAS, 386, 945Tavecchio F., Ghisellini G., Bonnoli G. & Ghirlanda G., 2010, MNRAS,405, L94 c (cid:13)000